2007-05-09 08:00:38 +07:00
|
|
|
/*
|
|
|
|
|
* Copyright (c) 2004, 2005 Topspin Communications. All rights reserved.
|
2008-07-26 00:32:52 +07:00
|
|
|
* Copyright (c) 2005, 2006, 2007, 2008 Mellanox Technologies. All rights reserved.
|
2007-05-09 08:00:38 +07:00
|
|
|
* Copyright (c) 2005, 2006, 2007 Cisco Systems, Inc. All rights reserved.
|
|
|
|
|
*
|
|
|
|
|
* This software is available to you under a choice of one of two
|
|
|
|
|
* licenses. You may choose to be licensed under the terms of the GNU
|
|
|
|
|
* General Public License (GPL) Version 2, available from the file
|
|
|
|
|
* COPYING in the main directory of this source tree, or the
|
|
|
|
|
* OpenIB.org BSD license below:
|
|
|
|
|
*
|
|
|
|
|
* Redistribution and use in source and binary forms, with or
|
|
|
|
|
* without modification, are permitted provided that the following
|
|
|
|
|
* conditions are met:
|
|
|
|
|
*
|
|
|
|
|
* - Redistributions of source code must retain the above
|
|
|
|
|
* copyright notice, this list of conditions and the following
|
|
|
|
|
* disclaimer.
|
|
|
|
|
*
|
|
|
|
|
* - Redistributions in binary form must reproduce the above
|
|
|
|
|
* copyright notice, this list of conditions and the following
|
|
|
|
|
* disclaimer in the documentation and/or other materials
|
|
|
|
|
* provided with the distribution.
|
|
|
|
|
*
|
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
|
|
|
|
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
|
|
|
|
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
|
|
|
|
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
|
|
|
|
|
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
|
|
|
|
|
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
|
|
|
|
|
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
|
|
|
* SOFTWARE.
|
|
|
|
|
*/
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
#include <linux/etherdevice.h>
|
2007-05-09 08:00:38 +07:00
|
|
|
#include <linux/mlx4/cmd.h>
|
2011-07-04 02:21:01 +07:00
|
|
|
#include <linux/module.h>
|
2009-09-25 01:03:03 +07:00
|
|
|
#include <linux/cache.h>
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
#include "fw.h"
|
|
|
|
|
#include "icm.h"
|
|
|
|
|
|
2007-06-08 13:24:36 +07:00
|
|
|
enum {
|
2007-06-18 22:15:02 +07:00
|
|
|
MLX4_COMMAND_INTERFACE_MIN_REV = 2,
|
|
|
|
|
MLX4_COMMAND_INTERFACE_MAX_REV = 3,
|
|
|
|
|
MLX4_COMMAND_INTERFACE_NEW_PORT_CMDS = 3,
|
2007-06-08 13:24:36 +07:00
|
|
|
};
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
extern void __buggy_use_of_MLX4_GET(void);
|
|
|
|
|
extern void __buggy_use_of_MLX4_PUT(void);
|
|
|
|
|
|
2015-04-02 20:31:18 +07:00
|
|
|
static bool enable_qos = true;
|
2008-07-23 04:19:37 +07:00
|
|
|
module_param(enable_qos, bool, 0444);
|
2015-04-02 20:31:18 +07:00
|
|
|
MODULE_PARM_DESC(enable_qos, "Enable Enhanced QoS support (default: on)");
|
2008-07-23 04:19:37 +07:00
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
#define MLX4_GET(dest, source, offset) \
|
|
|
|
|
do { \
|
|
|
|
|
void *__p = (char *) (source) + (offset); \
|
2015-04-30 03:52:51 +07:00
|
|
|
u64 val; \
|
2007-05-09 08:00:38 +07:00
|
|
|
switch (sizeof (dest)) { \
|
|
|
|
|
case 1: (dest) = *(u8 *) __p; break; \
|
|
|
|
|
case 2: (dest) = be16_to_cpup(__p); break; \
|
|
|
|
|
case 4: (dest) = be32_to_cpup(__p); break; \
|
2015-04-30 03:52:51 +07:00
|
|
|
case 8: val = get_unaligned((u64 *)__p); \
|
|
|
|
|
(dest) = be64_to_cpu(val); break; \
|
2007-05-09 08:00:38 +07:00
|
|
|
default: __buggy_use_of_MLX4_GET(); \
|
|
|
|
|
} \
|
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
|
|
#define MLX4_PUT(dest, source, offset) \
|
|
|
|
|
do { \
|
|
|
|
|
void *__d = ((char *) (dest) + (offset)); \
|
|
|
|
|
switch (sizeof(source)) { \
|
|
|
|
|
case 1: *(u8 *) __d = (source); break; \
|
|
|
|
|
case 2: *(__be16 *) __d = cpu_to_be16(source); break; \
|
|
|
|
|
case 4: *(__be32 *) __d = cpu_to_be32(source); break; \
|
|
|
|
|
case 8: *(__be64 *) __d = cpu_to_be64(source); break; \
|
|
|
|
|
default: __buggy_use_of_MLX4_PUT(); \
|
|
|
|
|
} \
|
|
|
|
|
} while (0)
|
|
|
|
|
|
2011-06-15 21:41:42 +07:00
|
|
|
static void dump_dev_cap_flags(struct mlx4_dev *dev, u64 flags)
|
2007-05-09 08:00:38 +07:00
|
|
|
{
|
|
|
|
|
static const char *fname[] = {
|
|
|
|
|
[ 0] = "RC transport",
|
|
|
|
|
[ 1] = "UC transport",
|
|
|
|
|
[ 2] = "UD transport",
|
2007-10-10 09:59:13 +07:00
|
|
|
[ 3] = "XRC transport",
|
2007-05-09 08:00:38 +07:00
|
|
|
[ 6] = "SRQ support",
|
|
|
|
|
[ 7] = "IPoIB checksum offload",
|
|
|
|
|
[ 8] = "P_Key violation counter",
|
|
|
|
|
[ 9] = "Q_Key violation counter",
|
2013-04-07 10:44:06 +07:00
|
|
|
[12] = "Dual Port Different Protocol (DPDP) support",
|
2009-11-13 02:19:44 +07:00
|
|
|
[15] = "Big LSO headers",
|
2007-05-09 08:00:38 +07:00
|
|
|
[16] = "MW support",
|
|
|
|
|
[17] = "APM support",
|
|
|
|
|
[18] = "Atomic ops support",
|
|
|
|
|
[19] = "Raw multicast support",
|
|
|
|
|
[20] = "Address vector port checking support",
|
|
|
|
|
[21] = "UD multicast support",
|
2011-07-08 02:19:29 +07:00
|
|
|
[30] = "IBoE support",
|
|
|
|
|
[32] = "Unicast loopback support",
|
2011-10-18 08:50:42 +07:00
|
|
|
[34] = "FCS header control",
|
2015-01-27 20:58:08 +07:00
|
|
|
[37] = "Wake On LAN (port1) support",
|
|
|
|
|
[38] = "Wake On LAN (port2) support",
|
2011-07-08 02:19:29 +07:00
|
|
|
[40] = "UDP RSS support",
|
|
|
|
|
[41] = "Unicast VEP steering support",
|
2011-06-15 21:47:14 +07:00
|
|
|
[42] = "Multicast VEP steering support",
|
|
|
|
|
[48] = "Counters support",
|
2015-04-02 20:31:06 +07:00
|
|
|
[52] = "RSS IP fragments support",
|
2013-04-07 10:44:07 +07:00
|
|
|
[53] = "Port ETS Scheduler support",
|
2013-04-07 10:44:06 +07:00
|
|
|
[55] = "Port link type sensing support",
|
mlx4: Use port management change event instead of smp_snoop
The port management change event can replace smp_snoop. If the
capability bit for this event is set in dev-caps, the event is used
(by the driver setting the PORT_MNG_CHG_EVENT bit in the async event
mask in the MAP_EQ fw command). In this case, when the driver passes
incoming SMP PORT_INFO SET mads to the FW, the FW generates port
management change events to signal any changes to the driver.
If the FW generates these events, smp_snoop shouldn't be invoked in
ib_process_mad(), or duplicate events will occur (once from the
FW-generated event, and once from smp_snoop).
In the case where the FW does not generate port management change
events smp_snoop needs to be invoked to create these events. The flow
in smp_snoop has been modified to make use of the same procedures as
in the fw-generated-event event case to generate the port management
events (LID change, Client-rereg, Pkey change, and/or GID change).
Port management change event handling required changing the
mlx4_ib_event and mlx4_dispatch_event prototypes; the "param" argument
(last argument) had to be changed to unsigned long in order to
accomodate passing the EQE pointer.
We also needed to move the definition of struct mlx4_eqe from
net/mlx4.h to file device.h -- to make it available to the IB driver,
to handle port management change events.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 15:21:40 +07:00
|
|
|
[59] = "Port management change event support",
|
2012-10-21 21:59:24 +07:00
|
|
|
[61] = "64 byte EQE support",
|
|
|
|
|
[62] = "64 byte CQE support",
|
2007-05-09 08:00:38 +07:00
|
|
|
};
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "DEV_CAP flags:\n");
|
2007-05-19 22:51:57 +07:00
|
|
|
for (i = 0; i < ARRAY_SIZE(fname); ++i)
|
2011-06-15 21:41:42 +07:00
|
|
|
if (fname[i] && (flags & (1LL << i)))
|
2007-05-09 08:00:38 +07:00
|
|
|
mlx4_dbg(dev, " %s\n", fname[i]);
|
|
|
|
|
}
|
|
|
|
|
|
2012-04-29 21:04:25 +07:00
|
|
|
static void dump_dev_cap_flags2(struct mlx4_dev *dev, u64 flags)
|
|
|
|
|
{
|
|
|
|
|
static const char * const fname[] = {
|
|
|
|
|
[0] = "RSS support",
|
|
|
|
|
[1] = "RSS Toeplitz Hash Function support",
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
[2] = "RSS XOR Hash Function support",
|
2014-03-12 22:16:30 +07:00
|
|
|
[3] = "Device managed flow steering support",
|
2013-04-23 13:06:47 +07:00
|
|
|
[4] = "Automatic MAC reassignment support",
|
2013-05-09 05:22:34 +07:00
|
|
|
[5] = "Time stamping support",
|
|
|
|
|
[6] = "VST (control vlan insertion/stripping) support",
|
2013-06-27 23:05:21 +07:00
|
|
|
[7] = "FSM (MAC anti-spoofing) support",
|
2013-12-23 21:09:43 +07:00
|
|
|
[8] = "Dynamic QP updates support",
|
2014-03-12 22:16:30 +07:00
|
|
|
[9] = "Device managed flow steering IPoIB support",
|
2014-06-01 15:53:50 +07:00
|
|
|
[10] = "TCP/IP offloads/flow-steering for VXLAN support",
|
2014-09-18 15:50:59 +07:00
|
|
|
[11] = "MAD DEMUX (Secure-Host) support",
|
|
|
|
|
[12] = "Large cache line (>64B) CQE stride support",
|
2014-10-27 16:37:37 +07:00
|
|
|
[13] = "Large cache line (>64B) EQE stride support",
|
2014-10-27 16:37:38 +07:00
|
|
|
[14] = "Ethernet protocol control support",
|
2014-11-02 21:26:17 +07:00
|
|
|
[15] = "Ethernet Backplane autoneg support",
|
2014-11-13 19:45:32 +07:00
|
|
|
[16] = "CONFIG DEV support",
|
2014-11-13 19:45:33 +07:00
|
|
|
[17] = "Asymmetric EQs support",
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
[18] = "More than 80 VFs support",
|
2015-01-27 20:57:59 +07:00
|
|
|
[19] = "Performance optimized for limited rule configuration flow steering support",
|
2015-02-03 21:48:32 +07:00
|
|
|
[20] = "Recoverable error events support",
|
2015-03-06 01:16:12 +07:00
|
|
|
[21] = "Port Remap support",
|
2015-03-18 19:57:34 +07:00
|
|
|
[22] = "QCN support",
|
2015-03-30 21:45:25 +07:00
|
|
|
[23] = "QP rate limiting support",
|
2015-04-02 20:31:13 +07:00
|
|
|
[24] = "Ethernet Flow control statistics support",
|
|
|
|
|
[25] = "Granular QoS per VF support",
|
2015-04-02 20:31:17 +07:00
|
|
|
[26] = "Port ETS Scheduler support",
|
2015-04-02 20:31:20 +07:00
|
|
|
[27] = "Port beacon support",
|
2015-04-02 20:31:22 +07:00
|
|
|
[28] = "RX-ALL support",
|
2015-07-27 18:46:31 +07:00
|
|
|
[29] = "802.1ad offload support",
|
2015-10-15 18:44:38 +07:00
|
|
|
[31] = "Modifying loopback source checks using UPDATE_QP support",
|
|
|
|
|
[32] = "Loopback source checks support",
|
2012-04-29 21:04:25 +07:00
|
|
|
};
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < ARRAY_SIZE(fname); ++i)
|
|
|
|
|
if (fname[i] && (flags & (1LL << i)))
|
|
|
|
|
mlx4_dbg(dev, " %s\n", fname[i]);
|
|
|
|
|
}
|
|
|
|
|
|
2008-07-15 13:48:53 +07:00
|
|
|
int mlx4_MOD_STAT_CFG(struct mlx4_dev *dev, struct mlx4_mod_stat_cfg *cfg)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *inbox;
|
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
|
|
#define MOD_STAT_CFG_IN_SIZE 0x100
|
|
|
|
|
|
|
|
|
|
#define MOD_STAT_CFG_PG_SZ_M_OFFSET 0x002
|
|
|
|
|
#define MOD_STAT_CFG_PG_SZ_OFFSET 0x003
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
inbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
MLX4_PUT(inbox, cfg->log_pg_sz, MOD_STAT_CFG_PG_SZ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, cfg->log_pg_sz_m, MOD_STAT_CFG_PG_SZ_M_OFFSET);
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, 0, 0, MLX4_CMD_MOD_STAT_CFG,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2008-07-15 13:48:53 +07:00
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 19:45:31 +07:00
|
|
|
int mlx4_QUERY_FUNC(struct mlx4_dev *dev, struct mlx4_func *func, int slave)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
u8 in_modifier;
|
|
|
|
|
u8 field;
|
|
|
|
|
u16 field16;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
#define QUERY_FUNC_BUS_OFFSET 0x00
|
|
|
|
|
#define QUERY_FUNC_DEVICE_OFFSET 0x01
|
|
|
|
|
#define QUERY_FUNC_FUNCTION_OFFSET 0x01
|
|
|
|
|
#define QUERY_FUNC_PHYSICAL_FUNCTION_OFFSET 0x03
|
|
|
|
|
#define QUERY_FUNC_RSVD_EQS_OFFSET 0x04
|
|
|
|
|
#define QUERY_FUNC_MAX_EQ_OFFSET 0x06
|
|
|
|
|
#define QUERY_FUNC_RSVD_UARS_OFFSET 0x0b
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
in_modifier = slave;
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, in_modifier, 0,
|
|
|
|
|
MLX4_CMD_QUERY_FUNC,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_BUS_OFFSET);
|
|
|
|
|
func->bus = field & 0xf;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_DEVICE_OFFSET);
|
|
|
|
|
func->device = field & 0xf1;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_FUNCTION_OFFSET);
|
|
|
|
|
func->function = field & 0x7;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_PHYSICAL_FUNCTION_OFFSET);
|
|
|
|
|
func->physical_function = field & 0xf;
|
|
|
|
|
MLX4_GET(field16, outbox, QUERY_FUNC_RSVD_EQS_OFFSET);
|
|
|
|
|
func->rsvd_eqs = field16 & 0xffff;
|
|
|
|
|
MLX4_GET(field16, outbox, QUERY_FUNC_MAX_EQ_OFFSET);
|
|
|
|
|
func->max_eq = field16 & 0xffff;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_RSVD_UARS_OFFSET);
|
|
|
|
|
func->rsvd_uars = field & 0x0f;
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Bus: %d, Device: %d, Function: %d, Physical function: %d, Max EQs: %d, Reserved EQs: %d, Reserved UARs: %d\n",
|
|
|
|
|
func->bus, func->device, func->function, func->physical_function,
|
|
|
|
|
func->max_eq, func->rsvd_eqs, func->rsvd_uars);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int mlx4_QUERY_FUNC_CAP_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:23 +07:00
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2014-05-29 20:31:03 +07:00
|
|
|
u8 field, port;
|
|
|
|
|
u32 size, proxy_qp, qkey;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int err = 0;
|
2014-11-13 19:45:32 +07:00
|
|
|
struct mlx4_func func;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
|
|
|
|
#define QUERY_FUNC_CAP_FLAGS_OFFSET 0x0
|
|
|
|
|
#define QUERY_FUNC_CAP_NUM_PORTS_OFFSET 0x1
|
|
|
|
|
#define QUERY_FUNC_CAP_PF_BHVR_OFFSET 0x4
|
2012-06-19 15:21:43 +07:00
|
|
|
#define QUERY_FUNC_CAP_FMR_OFFSET 0x8
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
#define QUERY_FUNC_CAP_QP_QUOTA_OFFSET_DEP 0x10
|
|
|
|
|
#define QUERY_FUNC_CAP_CQ_QUOTA_OFFSET_DEP 0x14
|
|
|
|
|
#define QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET_DEP 0x18
|
|
|
|
|
#define QUERY_FUNC_CAP_MPT_QUOTA_OFFSET_DEP 0x20
|
|
|
|
|
#define QUERY_FUNC_CAP_MTT_QUOTA_OFFSET_DEP 0x24
|
|
|
|
|
#define QUERY_FUNC_CAP_MCG_QUOTA_OFFSET_DEP 0x28
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
#define QUERY_FUNC_CAP_MAX_EQ_OFFSET 0x2c
|
2012-09-23 23:18:24 +07:00
|
|
|
#define QUERY_FUNC_CAP_RESERVED_EQ_OFFSET 0x30
|
2015-01-27 20:58:00 +07:00
|
|
|
#define QUERY_FUNC_CAP_QP_RESD_LKEY_OFFSET 0x48
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
#define QUERY_FUNC_CAP_QP_QUOTA_OFFSET 0x50
|
|
|
|
|
#define QUERY_FUNC_CAP_CQ_QUOTA_OFFSET 0x54
|
|
|
|
|
#define QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET 0x58
|
|
|
|
|
#define QUERY_FUNC_CAP_MPT_QUOTA_OFFSET 0x60
|
|
|
|
|
#define QUERY_FUNC_CAP_MTT_QUOTA_OFFSET 0x64
|
|
|
|
|
#define QUERY_FUNC_CAP_MCG_QUOTA_OFFSET 0x68
|
|
|
|
|
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:54 +07:00
|
|
|
#define QUERY_FUNC_CAP_EXTRA_FLAGS_OFFSET 0x6c
|
|
|
|
|
|
2012-06-19 15:21:43 +07:00
|
|
|
#define QUERY_FUNC_CAP_FMR_FLAG 0x80
|
|
|
|
|
#define QUERY_FUNC_CAP_FLAG_RDMA 0x40
|
|
|
|
|
#define QUERY_FUNC_CAP_FLAG_ETH 0x80
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAG_QUOTAS 0x10
|
2015-01-27 20:58:00 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAG_RESD_LKEY 0x08
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:54 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAG_VALID_MAILBOX 0x04
|
|
|
|
|
|
|
|
|
|
#define QUERY_FUNC_CAP_EXTRA_FLAGS_BF_QP_ALLOC_FLAG (1UL << 31)
|
net/mlx4: Add A0 hybrid steering
A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.
In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.
When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.
Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).
Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.
When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.
In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:57 +07:00
|
|
|
#define QUERY_FUNC_CAP_EXTRA_FLAGS_A0_QP_ALLOC_FLAG (1UL << 30)
|
2012-06-19 15:21:43 +07:00
|
|
|
|
|
|
|
|
/* when opcode modifier = 1 */
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
#define QUERY_FUNC_CAP_PHYS_PORT_OFFSET 0x3
|
2014-05-29 20:31:03 +07:00
|
|
|
#define QUERY_FUNC_CAP_PRIV_VF_QKEY_OFFSET 0x4
|
2013-12-20 02:20:10 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAGS0_OFFSET 0x8
|
|
|
|
|
#define QUERY_FUNC_CAP_FLAGS1_OFFSET 0xc
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
#define QUERY_FUNC_CAP_QP0_TUNNEL 0x10
|
|
|
|
|
#define QUERY_FUNC_CAP_QP0_PROXY 0x14
|
|
|
|
|
#define QUERY_FUNC_CAP_QP1_TUNNEL 0x18
|
|
|
|
|
#define QUERY_FUNC_CAP_QP1_PROXY 0x1c
|
2013-12-20 02:20:12 +07:00
|
|
|
#define QUERY_FUNC_CAP_PHYS_PORT_ID 0x28
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
|
2013-12-20 02:20:10 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAGS1_FORCE_MAC 0x40
|
|
|
|
|
#define QUERY_FUNC_CAP_FLAGS1_FORCE_VLAN 0x80
|
2013-12-20 02:20:11 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAGS1_NIC_INFO 0x10
|
2014-05-29 20:31:03 +07:00
|
|
|
#define QUERY_FUNC_CAP_VF_ENABLE_QP0 0x08
|
2012-06-19 15:21:43 +07:00
|
|
|
|
2013-12-20 02:20:10 +07:00
|
|
|
#define QUERY_FUNC_CAP_FLAGS0_FORCE_PHY_WQE_GID 0x80
|
2014-11-13 19:45:32 +07:00
|
|
|
#define QUERY_FUNC_CAP_SUPPORTS_NON_POWER_OF_2_NUM_EQS (1 << 31)
|
2015-07-27 18:46:31 +07:00
|
|
|
#define QUERY_FUNC_CAP_PHV_BIT 0x40
|
2012-06-19 15:21:43 +07:00
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
if (vhcr->op_modifier == 1) {
|
2014-03-19 23:11:52 +07:00
|
|
|
struct mlx4_active_ports actv_ports =
|
|
|
|
|
mlx4_get_active_ports(dev, slave);
|
|
|
|
|
int converted_port = mlx4_slave_convert_port(
|
|
|
|
|
dev, slave, vhcr->in_modifier);
|
|
|
|
|
|
|
|
|
|
if (converted_port < 0)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
vhcr->in_modifier = converted_port;
|
|
|
|
|
/* phys-port = logical-port */
|
|
|
|
|
field = vhcr->in_modifier -
|
|
|
|
|
find_first_bit(actv_ports.ports, dev->caps.num_ports);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_PHYS_PORT_OFFSET);
|
|
|
|
|
|
2014-05-29 20:31:03 +07:00
|
|
|
port = vhcr->in_modifier;
|
|
|
|
|
proxy_qp = dev->phys_caps.base_proxy_sqpn + 8 * slave + port - 1;
|
|
|
|
|
|
|
|
|
|
/* Set nic_info bit to mark new fields support */
|
|
|
|
|
field = QUERY_FUNC_CAP_FLAGS1_NIC_INFO;
|
|
|
|
|
|
|
|
|
|
if (mlx4_vf_smi_enabled(dev, slave, port) &&
|
|
|
|
|
!mlx4_get_parav_qkey(dev, proxy_qp, &qkey)) {
|
|
|
|
|
field |= QUERY_FUNC_CAP_VF_ENABLE_QP0;
|
|
|
|
|
MLX4_PUT(outbox->buf, qkey,
|
|
|
|
|
QUERY_FUNC_CAP_PRIV_VF_QKEY_OFFSET);
|
|
|
|
|
}
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_FLAGS1_OFFSET);
|
|
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
/* size is now the QP number */
|
2014-05-29 20:31:03 +07:00
|
|
|
size = dev->phys_caps.base_tunnel_sqpn + 8 * slave + port - 1;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP0_TUNNEL);
|
|
|
|
|
|
|
|
|
|
size += 2;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP1_TUNNEL);
|
|
|
|
|
|
2014-05-29 20:31:03 +07:00
|
|
|
MLX4_PUT(outbox->buf, proxy_qp, QUERY_FUNC_CAP_QP0_PROXY);
|
|
|
|
|
proxy_qp += 2;
|
|
|
|
|
MLX4_PUT(outbox->buf, proxy_qp, QUERY_FUNC_CAP_QP1_PROXY);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
|
2013-12-20 02:20:12 +07:00
|
|
|
MLX4_PUT(outbox->buf, dev->caps.phys_port_id[vhcr->in_modifier],
|
|
|
|
|
QUERY_FUNC_CAP_PHYS_PORT_ID);
|
|
|
|
|
|
2015-07-27 18:46:31 +07:00
|
|
|
if (dev->caps.phv_bit[port]) {
|
|
|
|
|
field = QUERY_FUNC_CAP_PHV_BIT;
|
|
|
|
|
MLX4_PUT(outbox->buf, field,
|
|
|
|
|
QUERY_FUNC_CAP_FLAGS0_OFFSET);
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
} else if (vhcr->op_modifier == 0) {
|
2014-03-19 23:11:52 +07:00
|
|
|
struct mlx4_active_ports actv_ports =
|
|
|
|
|
mlx4_get_active_ports(dev, slave);
|
2015-01-27 20:58:00 +07:00
|
|
|
/* enable rdma and ethernet interfaces, new quota locations,
|
|
|
|
|
* and reserved lkey
|
|
|
|
|
*/
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
field = (QUERY_FUNC_CAP_FLAG_ETH | QUERY_FUNC_CAP_FLAG_RDMA |
|
2015-01-27 20:58:00 +07:00
|
|
|
QUERY_FUNC_CAP_FLAG_QUOTAS | QUERY_FUNC_CAP_FLAG_VALID_MAILBOX |
|
|
|
|
|
QUERY_FUNC_CAP_FLAG_RESD_LKEY);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_FLAGS_OFFSET);
|
|
|
|
|
|
2014-03-19 23:11:52 +07:00
|
|
|
field = min(
|
|
|
|
|
bitmap_weight(actv_ports.ports, dev->caps.num_ports),
|
|
|
|
|
dev->caps.num_ports);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_NUM_PORTS_OFFSET);
|
|
|
|
|
|
2012-10-21 21:59:24 +07:00
|
|
|
size = dev->caps.function_caps; /* set PF behaviours */
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_PF_BHVR_OFFSET);
|
|
|
|
|
|
2012-06-19 15:21:43 +07:00
|
|
|
field = 0; /* protected FMR support not available as yet */
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_FMR_OFFSET);
|
|
|
|
|
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:23 +07:00
|
|
|
size = priv->mfunc.master.res_tracker.res_alloc[RES_QP].quota[slave];
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP_QUOTA_OFFSET);
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
size = dev->caps.num_qps;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP_QUOTA_OFFSET_DEP);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:23 +07:00
|
|
|
size = priv->mfunc.master.res_tracker.res_alloc[RES_SRQ].quota[slave];
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET);
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
size = dev->caps.num_srqs;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET_DEP);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:23 +07:00
|
|
|
size = priv->mfunc.master.res_tracker.res_alloc[RES_CQ].quota[slave];
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_CQ_QUOTA_OFFSET);
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
size = dev->caps.num_cqs;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_CQ_QUOTA_OFFSET_DEP);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2014-11-13 19:45:32 +07:00
|
|
|
if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SYS_EQS) ||
|
|
|
|
|
mlx4_QUERY_FUNC(dev, &func, slave)) {
|
|
|
|
|
size = vhcr->in_modifier &
|
|
|
|
|
QUERY_FUNC_CAP_SUPPORTS_NON_POWER_OF_2_NUM_EQS ?
|
|
|
|
|
dev->caps.num_eqs :
|
|
|
|
|
rounddown_pow_of_two(dev->caps.num_eqs);
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MAX_EQ_OFFSET);
|
|
|
|
|
size = dev->caps.reserved_eqs;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_RESERVED_EQ_OFFSET);
|
|
|
|
|
} else {
|
|
|
|
|
size = vhcr->in_modifier &
|
|
|
|
|
QUERY_FUNC_CAP_SUPPORTS_NON_POWER_OF_2_NUM_EQS ?
|
|
|
|
|
func.max_eq :
|
|
|
|
|
rounddown_pow_of_two(func.max_eq);
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MAX_EQ_OFFSET);
|
|
|
|
|
size = func.rsvd_eqs;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_RESERVED_EQ_OFFSET);
|
|
|
|
|
}
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:23 +07:00
|
|
|
size = priv->mfunc.master.res_tracker.res_alloc[RES_MPT].quota[slave];
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MPT_QUOTA_OFFSET);
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
size = dev->caps.num_mpts;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MPT_QUOTA_OFFSET_DEP);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Structures and init/teardown for VF resource quotas
This is step #1 for implementing SRIOV resource quotas for VFs.
Quotas are implemented per resource type for VFs and the PF, to prevent
any entity from simply grabbing all the resources for itself and leaving
the other entities unable to obtain such resources.
Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC,
VLAN, and Counters.
The quota system works as follows:
Each entity (VF or PF) is given a max number of a given resource (its quota),
and a guaranteed minimum number for each resource (starvation prevention).
For QPs, CQs, SRQs, MPTs and MTTs:
50% of the available quantity for the resource is divided equally among
the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module
parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool.
The other 50% is the "free pool", allocated on a first-come-first-serve basis.
For each VF/PF, resources are first allocated from its "guaranteed-minimum"
pool. When that pool is exhausted, the driver attempts to allocate from
the resource "free-pool".
The quota (i.e., max) for the VFs and the PF is:
The free-pool amount (50% of the real max) + the guaranteed minimum
For MACs:
Guarantee 2 MACs per VF/PF per port. As a result, since we have only
128 MACs per port, reduce the allowable number of VFs from 64 to 63.
Any remaining MACs are put into a free pool.
For VLANs:
For the PF, the per-port quota is 128 and guarantee is 64
(to allow the PF to register at least a VLAN per VF in VST mode).
For the VFs, the per-port quota is 64 and the guarantee is 0.
We assume that VGT VFs are trusted not to abuse the VLAN resource.
For Counters:
For all functions (PF and VFs), the quota is 128 and the guarantee is 0.
In this patch, we define the needed structures, which are added to the
resource-tracker struct. In addition, we do initialization
for the resource quota, and adjust the query_device response to use quotas
rather than resource maxima.
As part of the implementation, we introduce a new field in
mlx4_dev: quotas. This field holds the resource quotas used
to report maxima to the upper layers (ib_core, via query_device).
The HCA maxima of these values are passed to the VFs (via
QUERY_HCA) so that they may continue to use these in handling
QPs, CQs, SRQs and MPTs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:23 +07:00
|
|
|
size = priv->mfunc.master.res_tracker.res_alloc[RES_MTT].quota[slave];
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MTT_QUOTA_OFFSET);
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
size = dev->caps.num_mtts;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MTT_QUOTA_OFFSET_DEP);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
|
|
|
|
size = dev->caps.num_mgms + dev->caps.num_amgms;
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET);
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET_DEP);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
net/mlx4: Add A0 hybrid steering
A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.
In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.
When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.
Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).
Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.
When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.
In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:57 +07:00
|
|
|
size = QUERY_FUNC_CAP_EXTRA_FLAGS_BF_QP_ALLOC_FLAG |
|
|
|
|
|
QUERY_FUNC_CAP_EXTRA_FLAGS_A0_QP_ALLOC_FLAG;
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:54 +07:00
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_EXTRA_FLAGS_OFFSET);
|
2015-01-27 20:58:00 +07:00
|
|
|
|
|
|
|
|
size = dev->caps.reserved_lkey + ((slave << 8) & 0xFF00);
|
|
|
|
|
MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP_RESD_LKEY_OFFSET);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
} else
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-13 19:45:28 +07:00
|
|
|
int mlx4_QUERY_FUNC_CAP(struct mlx4_dev *dev, u8 gen_or_port,
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
struct mlx4_func_cap *func_cap)
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
u8 field, op_modifier;
|
2014-05-29 20:31:03 +07:00
|
|
|
u32 size, qkey;
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
int err = 0, quotas = 0;
|
2014-11-13 19:45:32 +07:00
|
|
|
u32 in_modifier;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
op_modifier = !!gen_or_port; /* 0 = general, 1 = logical port */
|
2014-11-13 19:45:32 +07:00
|
|
|
in_modifier = op_modifier ? gen_or_port :
|
|
|
|
|
QUERY_FUNC_CAP_SUPPORTS_NON_POWER_OF_2_NUM_EQS;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
|
2014-11-13 19:45:32 +07:00
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, in_modifier, op_modifier,
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_CMD_QUERY_FUNC_CAP,
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
if (!op_modifier) {
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_CAP_FLAGS_OFFSET);
|
|
|
|
|
if (!(field & (QUERY_FUNC_CAP_FLAG_ETH | QUERY_FUNC_CAP_FLAG_RDMA))) {
|
|
|
|
|
mlx4_err(dev, "The host supports neither eth nor rdma interfaces\n");
|
|
|
|
|
err = -EPROTONOSUPPORT;
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
func_cap->flags = field;
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
quotas = !!(func_cap->flags & QUERY_FUNC_CAP_FLAG_QUOTAS);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_CAP_NUM_PORTS_OFFSET);
|
|
|
|
|
func_cap->num_ports = field;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_PF_BHVR_OFFSET);
|
|
|
|
|
func_cap->pf_context_behaviour = size;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
if (quotas) {
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP_QUOTA_OFFSET);
|
|
|
|
|
func_cap->qp_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET);
|
|
|
|
|
func_cap->srq_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_CQ_QUOTA_OFFSET);
|
|
|
|
|
func_cap->cq_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MPT_QUOTA_OFFSET);
|
|
|
|
|
func_cap->mpt_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MTT_QUOTA_OFFSET);
|
|
|
|
|
func_cap->mtt_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET);
|
|
|
|
|
func_cap->mcg_quota = size & 0xFFFFFF;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
} else {
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP_QUOTA_OFFSET_DEP);
|
|
|
|
|
func_cap->qp_quota = size & 0xFFFFFF;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET_DEP);
|
|
|
|
|
func_cap->srq_quota = size & 0xFFFFFF;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
net/mlx4_core: Fix quota handling in the QUERY_FUNC_CAP wrapper
In current kernels, the mlx4 driver running on a VM does not
differentiate between max resource numbers for the HCA and
max quotas -- it simply takes the quota values passed to it
as max-resource values.
However, the driver actually requires the VFs to be aware of
the actual number of resources that the HCA was initialized with,
for QPs, CQs, SRQs and MPTs.
For QPs, CQs and SRQs, the reason is that in completion handling
the driver must know which of the 24 bits are the actual resource
number, and which are "padding" bits.
For MPTs, also, the driver assumes knowledge of the number of MPTs
in the system.
The previous commit fixes the quota logic on the VM for the quota values
passed to it by QUERY_FUNC_CAPS.
For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers
from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed
in QUERY_FUNC_CAPS are used to report max resource number values
in the response to ib_query_device.
However, the Hypervisor driver must consider that VMs
may be running previous kernels, and compatibility must be preserved.
To resolve the incompatibility with previous kernels running on VMs,
we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the
deprecated fields, we pass the max-resource values from INIT_HCA
The quota fields are moved to a new location, and the current kernel
driver takes the proper values from that location. There is
also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox;
if this flag is set, the (VM) driver takes the quota values from the
new location.
VMs running previous kernels will work properly, except that the max resource
numbers reported in ib_query_device for these resources will be
too high. The Hypervisor driver will, however, enforce the quotas
for these VMs.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-11-03 15:03:24 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_CQ_QUOTA_OFFSET_DEP);
|
|
|
|
|
func_cap->cq_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MPT_QUOTA_OFFSET_DEP);
|
|
|
|
|
func_cap->mpt_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MTT_QUOTA_OFFSET_DEP);
|
|
|
|
|
func_cap->mtt_quota = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET_DEP);
|
|
|
|
|
func_cap->mcg_quota = size & 0xFFFFFF;
|
|
|
|
|
}
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_MAX_EQ_OFFSET);
|
|
|
|
|
func_cap->max_eq = size & 0xFFFFFF;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_RESERVED_EQ_OFFSET);
|
|
|
|
|
func_cap->reserved_eq = size & 0xFFFFFF;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2015-01-27 20:58:00 +07:00
|
|
|
if (func_cap->flags & QUERY_FUNC_CAP_FLAG_RESD_LKEY) {
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP_RESD_LKEY_OFFSET);
|
|
|
|
|
func_cap->reserved_lkey = size;
|
|
|
|
|
} else {
|
|
|
|
|
func_cap->reserved_lkey = 0;
|
|
|
|
|
}
|
|
|
|
|
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:54 +07:00
|
|
|
func_cap->extra_flags = 0;
|
|
|
|
|
|
|
|
|
|
/* Mailbox data from 0x6c and onward should only be treated if
|
|
|
|
|
* QUERY_FUNC_CAP_FLAG_VALID_MAILBOX is set in func_cap->flags
|
|
|
|
|
*/
|
|
|
|
|
if (func_cap->flags & QUERY_FUNC_CAP_FLAG_VALID_MAILBOX) {
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_EXTRA_FLAGS_OFFSET);
|
|
|
|
|
if (size & QUERY_FUNC_CAP_EXTRA_FLAGS_BF_QP_ALLOC_FLAG)
|
|
|
|
|
func_cap->extra_flags |= MLX4_QUERY_FUNC_FLAGS_BF_RES_QP;
|
net/mlx4: Add A0 hybrid steering
A0 hybrid steering is a form of high performance flow steering.
By using this mode, mlx4 cards use a fast limited table based steering,
in order to enable fast steering of unicast packets to a QP.
In order to implement A0 hybrid steering we allocate resources
from different zones:
(1) General range
(2) Special MAC-assigned QPs [RSS, Raw-Ethernet] each has its own region.
When we create a rss QP or a raw ethernet (A0 steerable and BF ready) QP,
we try hard to allocate the QP from range (2). Otherwise, we try hard not
to allocate from this range. However, when the system is pushed to its
limits and one needs every resource, the allocator uses every region it can.
Meaning, when we run out of raw-eth qps, the allocator allocates from the
general range (and the special-A0 area is no longer active). If we run out
of RSS qps, the mechanism tries to allocate from the raw-eth QP zone. If that
is also exhausted, the allocator will allocate from the general range
(and the A0 region is no longer active).
Note that if a raw-eth qp is allocated from the general range, it attempts
to allocate the range such that bits 6 and 7 (blueflame bits) in the
QP number are not set.
When the feature is used in SRIOV, the VF has to notify the PF what
kind of QP attributes it needs. In order to do that, along with the
"Eth QP blueflame" bit, we reserve a new "A0 steerable QP". According
to the combination of these bits, the PF tries to allocate a suitable QP.
In order to maintain backward compatibility (with older PFs), the PF
notifies which QP attributes it supports via QUERY_FUNC_CAP command.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:57 +07:00
|
|
|
if (size & QUERY_FUNC_CAP_EXTRA_FLAGS_A0_QP_ALLOC_FLAG)
|
|
|
|
|
func_cap->extra_flags |= MLX4_QUERY_FUNC_FLAGS_A0_RES_QP;
|
net/mlx4: Change QP allocation scheme
When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.
The current Ethernet driver code reserves a Tx QP range with 256b alignment.
This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.
This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.
The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:
1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
and request "BF enabled QPs" if BF is supported for the function
2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0] - number of QPs
b. param1[31-24] - flags controlling QPs reservation
Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.
Bits 24-30 of the flags are currently reserved.
When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.
In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:57:54 +07:00
|
|
|
}
|
|
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
goto out;
|
|
|
|
|
}
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
/* logical port query */
|
|
|
|
|
if (gen_or_port > dev->caps.num_ports) {
|
|
|
|
|
err = -EINVAL;
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2013-12-20 02:20:11 +07:00
|
|
|
MLX4_GET(func_cap->flags1, outbox, QUERY_FUNC_CAP_FLAGS1_OFFSET);
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
if (dev->caps.port_type[gen_or_port] == MLX4_PORT_TYPE_ETH) {
|
2014-05-29 20:31:00 +07:00
|
|
|
if (func_cap->flags1 & QUERY_FUNC_CAP_FLAGS1_FORCE_VLAN) {
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
mlx4_err(dev, "VLAN is enforced on this port\n");
|
|
|
|
|
err = -EPROTONOSUPPORT;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
goto out;
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
}
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2013-12-20 02:20:11 +07:00
|
|
|
if (func_cap->flags1 & QUERY_FUNC_CAP_FLAGS1_FORCE_MAC) {
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
mlx4_err(dev, "Force mac is enabled on this port\n");
|
|
|
|
|
err = -EPROTONOSUPPORT;
|
|
|
|
|
goto out;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
}
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
} else if (dev->caps.port_type[gen_or_port] == MLX4_PORT_TYPE_IB) {
|
2013-12-20 02:20:10 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_CAP_FLAGS0_OFFSET);
|
|
|
|
|
if (field & QUERY_FUNC_CAP_FLAGS0_FORCE_PHY_WQE_GID) {
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_err(dev, "phy_wqe_gid is enforced on this ib port\n");
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
err = -EPROTONOSUPPORT;
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
}
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_CAP_PHYS_PORT_OFFSET);
|
|
|
|
|
func_cap->physical_port = field;
|
|
|
|
|
if (func_cap->physical_port != gen_or_port) {
|
|
|
|
|
err = -ENOSYS;
|
|
|
|
|
goto out;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
}
|
|
|
|
|
|
2014-05-29 20:31:03 +07:00
|
|
|
if (func_cap->flags1 & QUERY_FUNC_CAP_VF_ENABLE_QP0) {
|
|
|
|
|
MLX4_GET(qkey, outbox, QUERY_FUNC_CAP_PRIV_VF_QKEY_OFFSET);
|
|
|
|
|
func_cap->qp0_qkey = qkey;
|
|
|
|
|
} else {
|
|
|
|
|
func_cap->qp0_qkey = 0;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
Previously, the structure of a guest's proxy QPs followed the
structure of the PPF special qps (qp0 port 1, qp0 port 2, qp1 port 1,
qp1 port 2, ...). The guest then did offset calculations on the
sqp_base qp number that the PPF passed to it in QUERY_FUNC_CAP().
This is now changed so that the guest does no offset calculations
regarding proxy or tunnel QPs to use. This change frees the PPF from
needing to adhere to a specific order in allocating proxy and tunnel
QPs.
Now QUERY_FUNC_CAP provides each port individually with its proxy
qp0, proxy qp1, tunnel qp0, and tunnel qp1 QP numbers, and these are
used directly where required (with no offset calculations).
To accomplish this change, several fields were added to the phys_caps
structure for use by the PPF and by non-SR-IOV mode:
base_sqpn -- in non-sriov mode, this was formerly sqp_start.
base_proxy_sqpn -- the first physical proxy qp number -- used by PPF
base_tunnel_sqpn -- the first physical tunnel qp number -- used by PPF.
The current code in the PPF still adheres to the previous layout of
sqps, proxy-sqps and tunnel-sqps. However, the PPF can change this
layout without affecting VF or (paravirtualized) PF code.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-03 15:40:57 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP0_TUNNEL);
|
|
|
|
|
func_cap->qp0_tunnel_qpn = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP0_PROXY);
|
|
|
|
|
func_cap->qp0_proxy_qpn = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP1_TUNNEL);
|
|
|
|
|
func_cap->qp1_tunnel_qpn = size & 0xFFFFFF;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP1_PROXY);
|
|
|
|
|
func_cap->qp1_proxy_qpn = size & 0xFFFFFF;
|
|
|
|
|
|
2013-12-20 02:20:12 +07:00
|
|
|
if (func_cap->flags1 & QUERY_FUNC_CAP_FLAGS1_NIC_INFO)
|
|
|
|
|
MLX4_GET(func_cap->phys_port_id, outbox,
|
|
|
|
|
QUERY_FUNC_CAP_PHYS_PORT_ID);
|
|
|
|
|
|
2015-07-27 18:46:31 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_FUNC_CAP_FLAGS0_OFFSET);
|
|
|
|
|
func_cap->flags |= (field & QUERY_FUNC_CAP_PHV_BIT);
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
/* All other resources are allocated by the master, but we still report
|
|
|
|
|
* 'num' and 'reserved' capabilities as follows:
|
|
|
|
|
* - num remains the maximum resource index
|
|
|
|
|
* - 'num - reserved' is the total available objects of a resource, but
|
|
|
|
|
* resource indices may be less than 'reserved'
|
|
|
|
|
* TODO: set per-resource quotas */
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
u8 field;
|
2011-07-08 02:19:29 +07:00
|
|
|
u32 field32, flags, ext_flags;
|
2007-05-09 08:00:38 +07:00
|
|
|
u16 size;
|
|
|
|
|
u16 stat_rate;
|
|
|
|
|
int err;
|
2007-06-18 22:15:02 +07:00
|
|
|
int i;
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
#define QUERY_DEV_CAP_OUT_SIZE 0x100
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_SRQ_SZ_OFFSET 0x10
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_QP_SZ_OFFSET 0x11
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_QP_OFFSET 0x12
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_QP_OFFSET 0x13
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_SRQ_OFFSET 0x14
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_SRQ_OFFSET 0x15
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_EEC_OFFSET 0x16
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_EEC_OFFSET 0x17
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_CQ_SZ_OFFSET 0x19
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_CQ_OFFSET 0x1a
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_CQ_OFFSET 0x1b
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_MPT_OFFSET 0x1d
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_EQ_OFFSET 0x1e
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_EQ_OFFSET 0x1f
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_MTT_OFFSET 0x20
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_MRW_SZ_OFFSET 0x21
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_MRW_OFFSET 0x22
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_MTT_SEG_OFFSET 0x23
|
2014-11-13 19:45:32 +07:00
|
|
|
#define QUERY_DEV_CAP_NUM_SYS_EQ_OFFSET 0x26
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_AV_OFFSET 0x27
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_REQ_QP_OFFSET 0x29
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_RES_QP_OFFSET 0x2b
|
2008-04-17 11:09:27 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_GSO_OFFSET 0x2d
|
2012-04-29 21:04:25 +07:00
|
|
|
#define QUERY_DEV_CAP_RSS_OFFSET 0x2e
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_RDMA_OFFSET 0x2f
|
|
|
|
|
#define QUERY_DEV_CAP_RSZ_SRQ_OFFSET 0x33
|
2015-04-02 20:31:20 +07:00
|
|
|
#define QUERY_DEV_CAP_PORT_BEACON_OFFSET 0x34
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_ACK_DELAY_OFFSET 0x35
|
|
|
|
|
#define QUERY_DEV_CAP_MTU_WIDTH_OFFSET 0x36
|
|
|
|
|
#define QUERY_DEV_CAP_VL_PORT_OFFSET 0x37
|
2007-06-26 19:55:28 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_MSG_SZ_OFFSET 0x38
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_GID_OFFSET 0x3b
|
|
|
|
|
#define QUERY_DEV_CAP_RATE_SUPPORT_OFFSET 0x3c
|
2013-04-23 13:06:47 +07:00
|
|
|
#define QUERY_DEV_CAP_CQ_TS_SUPPORT_OFFSET 0x3e
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_PKEY_OFFSET 0x3f
|
2011-07-08 02:19:29 +07:00
|
|
|
#define QUERY_DEV_CAP_EXT_FLAGS_OFFSET 0x40
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_FLAGS_OFFSET 0x44
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_UAR_OFFSET 0x48
|
|
|
|
|
#define QUERY_DEV_CAP_UAR_SZ_OFFSET 0x49
|
|
|
|
|
#define QUERY_DEV_CAP_PAGE_SZ_OFFSET 0x4b
|
|
|
|
|
#define QUERY_DEV_CAP_BF_OFFSET 0x4c
|
|
|
|
|
#define QUERY_DEV_CAP_LOG_BF_REG_SZ_OFFSET 0x4d
|
|
|
|
|
#define QUERY_DEV_CAP_LOG_MAX_BF_REGS_PER_PAGE_OFFSET 0x4e
|
|
|
|
|
#define QUERY_DEV_CAP_LOG_MAX_BF_PAGES_OFFSET 0x4f
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_SG_SQ_OFFSET 0x51
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_DESC_SZ_SQ_OFFSET 0x52
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_SG_RQ_OFFSET 0x55
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_DESC_SZ_RQ_OFFSET 0x56
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_QP_MCG_OFFSET 0x61
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_MCG_OFFSET 0x62
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_MCG_OFFSET 0x63
|
|
|
|
|
#define QUERY_DEV_CAP_RSVD_PD_OFFSET 0x64
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_PD_OFFSET 0x65
|
2011-06-02 23:01:33 +07:00
|
|
|
#define QUERY_DEV_CAP_RSVD_XRC_OFFSET 0x66
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_XRC_OFFSET 0x67
|
2011-06-15 21:47:14 +07:00
|
|
|
#define QUERY_DEV_CAP_MAX_COUNTERS_OFFSET 0x68
|
2015-03-30 21:45:25 +07:00
|
|
|
#define QUERY_DEV_CAP_PORT_FLOWSTATS_COUNTERS_OFFSET 0x70
|
2013-04-25 12:22:28 +07:00
|
|
|
#define QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET 0x70
|
2013-11-07 20:25:14 +07:00
|
|
|
#define QUERY_DEV_CAP_FLOW_STEERING_IPOIB_OFFSET 0x74
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
#define QUERY_DEV_CAP_FLOW_STEERING_RANGE_EN_OFFSET 0x76
|
|
|
|
|
#define QUERY_DEV_CAP_FLOW_STEERING_MAX_QP_OFFSET 0x77
|
2014-09-18 15:50:59 +07:00
|
|
|
#define QUERY_DEV_CAP_CQ_EQ_CACHE_LINE_STRIDE 0x7a
|
2015-03-06 01:16:12 +07:00
|
|
|
#define QUERY_DEV_CAP_ECN_QCN_VER_OFFSET 0x7b
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_RDMARC_ENTRY_SZ_OFFSET 0x80
|
|
|
|
|
#define QUERY_DEV_CAP_QPC_ENTRY_SZ_OFFSET 0x82
|
|
|
|
|
#define QUERY_DEV_CAP_AUX_ENTRY_SZ_OFFSET 0x84
|
|
|
|
|
#define QUERY_DEV_CAP_ALTC_ENTRY_SZ_OFFSET 0x86
|
|
|
|
|
#define QUERY_DEV_CAP_EQC_ENTRY_SZ_OFFSET 0x88
|
|
|
|
|
#define QUERY_DEV_CAP_CQC_ENTRY_SZ_OFFSET 0x8a
|
|
|
|
|
#define QUERY_DEV_CAP_SRQ_ENTRY_SZ_OFFSET 0x8c
|
|
|
|
|
#define QUERY_DEV_CAP_C_MPT_ENTRY_SZ_OFFSET 0x8e
|
|
|
|
|
#define QUERY_DEV_CAP_MTT_ENTRY_SZ_OFFSET 0x90
|
|
|
|
|
#define QUERY_DEV_CAP_D_MPT_ENTRY_SZ_OFFSET 0x92
|
2008-07-23 22:12:26 +07:00
|
|
|
#define QUERY_DEV_CAP_BMME_FLAGS_OFFSET 0x94
|
2014-11-02 21:26:17 +07:00
|
|
|
#define QUERY_DEV_CAP_CONFIG_DEV_OFFSET 0x94
|
2015-07-27 18:46:31 +07:00
|
|
|
#define QUERY_DEV_CAP_PHV_EN_OFFSET 0x96
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_DEV_CAP_RSVD_LKEY_OFFSET 0x98
|
|
|
|
|
#define QUERY_DEV_CAP_MAX_ICM_SZ_OFFSET 0xa0
|
2014-10-27 16:37:38 +07:00
|
|
|
#define QUERY_DEV_CAP_ETH_BACKPL_OFFSET 0x9c
|
2013-01-31 06:07:10 +07:00
|
|
|
#define QUERY_DEV_CAP_FW_REASSIGN_MAC 0x9d
|
2013-12-23 21:09:43 +07:00
|
|
|
#define QUERY_DEV_CAP_VXLAN 0x9e
|
2014-06-01 15:53:50 +07:00
|
|
|
#define QUERY_DEV_CAP_MAD_DEMUX_OFFSET 0xb0
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
#define QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_BASE_OFFSET 0xa8
|
|
|
|
|
#define QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_RANGE_OFFSET 0xac
|
2015-03-18 19:57:34 +07:00
|
|
|
#define QUERY_DEV_CAP_QP_RATE_LIMIT_NUM_OFFSET 0xcc
|
|
|
|
|
#define QUERY_DEV_CAP_QP_RATE_LIMIT_MAX_OFFSET 0xd0
|
|
|
|
|
#define QUERY_DEV_CAP_QP_RATE_LIMIT_MIN_OFFSET 0xd2
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2012-04-29 21:04:25 +07:00
|
|
|
dev_cap->flags2 = 0;
|
2007-05-09 08:00:38 +07:00
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_QUERY_DEV_CAP,
|
2012-05-30 16:14:55 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_QP_OFFSET);
|
|
|
|
|
dev_cap->reserved_qps = 1 << (field & 0xf);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_QP_OFFSET);
|
|
|
|
|
dev_cap->max_qps = 1 << (field & 0x1f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_SRQ_OFFSET);
|
|
|
|
|
dev_cap->reserved_srqs = 1 << (field >> 4);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_SRQ_OFFSET);
|
|
|
|
|
dev_cap->max_srqs = 1 << (field & 0x1f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_CQ_SZ_OFFSET);
|
|
|
|
|
dev_cap->max_cq_sz = 1 << field;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_CQ_OFFSET);
|
|
|
|
|
dev_cap->reserved_cqs = 1 << (field & 0xf);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_CQ_OFFSET);
|
|
|
|
|
dev_cap->max_cqs = 1 << (field & 0x1f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_MPT_OFFSET);
|
|
|
|
|
dev_cap->max_mpts = 1 << (field & 0x3f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_EQ_OFFSET);
|
2014-11-13 19:45:27 +07:00
|
|
|
dev_cap->reserved_eqs = 1 << (field & 0xf);
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_EQ_OFFSET);
|
2007-12-10 10:25:23 +07:00
|
|
|
dev_cap->max_eqs = 1 << (field & 0xf);
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_MTT_OFFSET);
|
|
|
|
|
dev_cap->reserved_mtts = 1 << (field >> 4);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_MRW_SZ_OFFSET);
|
|
|
|
|
dev_cap->max_mrw_sz = 1 << field;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_MRW_OFFSET);
|
|
|
|
|
dev_cap->reserved_mrws = 1 << (field & 0xf);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_MTT_SEG_OFFSET);
|
|
|
|
|
dev_cap->max_mtt_seg = 1 << (field & 0x3f);
|
2014-11-13 19:45:32 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_NUM_SYS_EQ_OFFSET);
|
|
|
|
|
dev_cap->num_sys_eqs = size & 0xfff;
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_REQ_QP_OFFSET);
|
|
|
|
|
dev_cap->max_requester_per_qp = 1 << (field & 0x3f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_RES_QP_OFFSET);
|
|
|
|
|
dev_cap->max_responder_per_qp = 1 << (field & 0x3f);
|
2008-04-17 11:09:27 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_GSO_OFFSET);
|
|
|
|
|
field &= 0x1f;
|
|
|
|
|
if (!field)
|
|
|
|
|
dev_cap->max_gso_sz = 0;
|
|
|
|
|
else
|
|
|
|
|
dev_cap->max_gso_sz = 1 << field;
|
|
|
|
|
|
2012-04-29 21:04:25 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSS_OFFSET);
|
|
|
|
|
if (field & 0x20)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_RSS_XOR;
|
|
|
|
|
if (field & 0x10)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_RSS_TOP;
|
|
|
|
|
field &= 0xf;
|
|
|
|
|
if (field) {
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_RSS;
|
|
|
|
|
dev_cap->max_rss_tbl_sz = 1 << field;
|
|
|
|
|
} else
|
|
|
|
|
dev_cap->max_rss_tbl_sz = 0;
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_RDMA_OFFSET);
|
|
|
|
|
dev_cap->max_rdma_global = 1 << (field & 0x3f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_ACK_DELAY_OFFSET);
|
|
|
|
|
dev_cap->local_ca_ack_delay = field & 0x1f;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_VL_PORT_OFFSET);
|
|
|
|
|
dev_cap->num_ports = field & 0xf;
|
2007-06-26 19:55:28 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_MSG_SZ_OFFSET);
|
2015-04-21 19:46:34 +07:00
|
|
|
dev_cap->max_msg_sz = 1 << (field & 0x1f);
|
2015-03-30 21:45:25 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_PORT_FLOWSTATS_COUNTERS_OFFSET);
|
|
|
|
|
if (field & 0x10)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_FLOWSTATS_EN;
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_FLOW_STEERING_RANGE_EN_OFFSET);
|
|
|
|
|
if (field & 0x80)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_FS_EN;
|
|
|
|
|
dev_cap->fs_log_max_ucast_qp_range_size = field & 0x1f;
|
2015-04-02 20:31:20 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_PORT_BEACON_OFFSET);
|
|
|
|
|
if (field & 0x80)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_PORT_BEACON;
|
2013-11-07 20:25:14 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_FLOW_STEERING_IPOIB_OFFSET);
|
|
|
|
|
if (field & 0x80)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_DMFS_IPOIB;
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_FLOW_STEERING_MAX_QP_OFFSET);
|
|
|
|
|
dev_cap->fs_max_num_qp_per_entry = field;
|
2015-03-06 01:16:12 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_ECN_QCN_VER_OFFSET);
|
|
|
|
|
if (field & 0x1)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_QCN;
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(stat_rate, outbox, QUERY_DEV_CAP_RATE_SUPPORT_OFFSET);
|
|
|
|
|
dev_cap->stat_rate_support = stat_rate;
|
2013-04-23 13:06:47 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_CQ_TS_SUPPORT_OFFSET);
|
|
|
|
|
if (field & 0x80)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_TS;
|
2011-07-08 02:19:29 +07:00
|
|
|
MLX4_GET(ext_flags, outbox, QUERY_DEV_CAP_EXT_FLAGS_OFFSET);
|
2011-06-15 21:41:42 +07:00
|
|
|
MLX4_GET(flags, outbox, QUERY_DEV_CAP_FLAGS_OFFSET);
|
2011-07-08 02:19:29 +07:00
|
|
|
dev_cap->flags = flags | (u64)ext_flags << 32;
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_UAR_OFFSET);
|
|
|
|
|
dev_cap->reserved_uars = field >> 4;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_UAR_SZ_OFFSET);
|
|
|
|
|
dev_cap->uar_size = 1 << ((field & 0x3f) + 20);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_PAGE_SZ_OFFSET);
|
|
|
|
|
dev_cap->min_page_sz = 1 << field;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_BF_OFFSET);
|
|
|
|
|
if (field & 0x80) {
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_LOG_BF_REG_SZ_OFFSET);
|
|
|
|
|
dev_cap->bf_reg_size = 1 << (field & 0x1f);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_LOG_MAX_BF_REGS_PER_PAGE_OFFSET);
|
2011-01-11 08:42:05 +07:00
|
|
|
if ((1 << (field & 0x3f)) > (PAGE_SIZE / dev_cap->bf_reg_size))
|
2010-11-10 19:52:37 +07:00
|
|
|
field = 3;
|
2007-05-09 08:00:38 +07:00
|
|
|
dev_cap->bf_regs_per_page = 1 << (field & 0x3f);
|
|
|
|
|
} else {
|
|
|
|
|
dev_cap->bf_reg_size = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_SG_SQ_OFFSET);
|
|
|
|
|
dev_cap->max_sq_sg = field;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_MAX_DESC_SZ_SQ_OFFSET);
|
|
|
|
|
dev_cap->max_sq_desc_sz = size;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_QP_MCG_OFFSET);
|
|
|
|
|
dev_cap->max_qp_per_mcg = 1 << field;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_MCG_OFFSET);
|
|
|
|
|
dev_cap->reserved_mgms = field & 0xf;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_MCG_OFFSET);
|
|
|
|
|
dev_cap->max_mcgs = 1 << field;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_PD_OFFSET);
|
|
|
|
|
dev_cap->reserved_pds = field >> 4;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_PD_OFFSET);
|
|
|
|
|
dev_cap->max_pds = 1 << (field & 0x3f);
|
2011-06-02 23:01:33 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_XRC_OFFSET);
|
|
|
|
|
dev_cap->reserved_xrcds = field >> 4;
|
2012-08-23 21:09:04 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_XRC_OFFSET);
|
2011-06-02 23:01:33 +07:00
|
|
|
dev_cap->max_xrcds = 1 << (field & 0x1f);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_RDMARC_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->rdmarc_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_QPC_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->qpc_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_AUX_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->aux_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_ALTC_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->altc_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_EQC_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->eqc_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_CQC_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->cqc_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_SRQ_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->srq_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_C_MPT_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->cmpt_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_MTT_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->mtt_entry_sz = size;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_D_MPT_ENTRY_SZ_OFFSET);
|
|
|
|
|
dev_cap->dmpt_entry_sz = size;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_SRQ_SZ_OFFSET);
|
|
|
|
|
dev_cap->max_srq_sz = 1 << field;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_QP_SZ_OFFSET);
|
|
|
|
|
dev_cap->max_qp_sz = 1 << field;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_RSZ_SRQ_OFFSET);
|
|
|
|
|
dev_cap->resize_srq = field & 1;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_SG_RQ_OFFSET);
|
|
|
|
|
dev_cap->max_rq_sg = field;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_MAX_DESC_SZ_RQ_OFFSET);
|
|
|
|
|
dev_cap->max_rq_desc_sz = size;
|
2014-09-18 15:50:59 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_CQ_EQ_CACHE_LINE_STRIDE);
|
2015-04-02 20:31:13 +07:00
|
|
|
if (field & (1 << 4))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_QOS_VPP;
|
2014-10-27 16:37:37 +07:00
|
|
|
if (field & (1 << 5))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_ETH_PROT_CTRL;
|
2014-09-18 15:50:59 +07:00
|
|
|
if (field & (1 << 6))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_CQE_STRIDE;
|
|
|
|
|
if (field & (1 << 7))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_EQE_STRIDE;
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(dev_cap->bmme_flags, outbox,
|
|
|
|
|
QUERY_DEV_CAP_BMME_FLAGS_OFFSET);
|
2015-02-03 21:48:32 +07:00
|
|
|
if (dev_cap->bmme_flags & MLX4_FLAG_PORT_REMAP)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_PORT_REMAP;
|
2014-11-02 21:26:17 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_CONFIG_DEV_OFFSET);
|
|
|
|
|
if (field & 0x20)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_CONFIG_DEV;
|
2015-04-02 20:31:22 +07:00
|
|
|
if (field & (1 << 2))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_IGNORE_FCS;
|
2015-07-27 18:46:31 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_PHV_EN_OFFSET);
|
|
|
|
|
if (field & 0x80)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_PHV_EN;
|
|
|
|
|
if (field & 0x40)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN;
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(dev_cap->reserved_lkey, outbox,
|
|
|
|
|
QUERY_DEV_CAP_RSVD_LKEY_OFFSET);
|
2014-10-27 16:37:38 +07:00
|
|
|
MLX4_GET(field32, outbox, QUERY_DEV_CAP_ETH_BACKPL_OFFSET);
|
|
|
|
|
if (field32 & (1 << 0))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_ETH_BACKPL_AN_REP;
|
2015-01-27 20:57:59 +07:00
|
|
|
if (field32 & (1 << 7))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_RECOVERABLE_ERROR_EVENT;
|
2013-01-31 06:07:10 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_FW_REASSIGN_MAC);
|
|
|
|
|
if (field & 1<<6)
|
2013-10-15 21:55:22 +07:00
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN;
|
2013-12-23 21:09:43 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_VXLAN);
|
|
|
|
|
if (field & 1<<3)
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS;
|
2015-04-02 20:31:17 +07:00
|
|
|
if (field & (1 << 5))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_ETS_CFG;
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(dev_cap->max_icm_sz, outbox,
|
|
|
|
|
QUERY_DEV_CAP_MAX_ICM_SZ_OFFSET);
|
2011-06-15 21:47:14 +07:00
|
|
|
if (dev_cap->flags & MLX4_DEV_CAP_FLAG_COUNTERS)
|
|
|
|
|
MLX4_GET(dev_cap->max_counters, outbox,
|
|
|
|
|
QUERY_DEV_CAP_MAX_COUNTERS_OFFSET);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2014-06-01 15:53:50 +07:00
|
|
|
MLX4_GET(field32, outbox,
|
|
|
|
|
QUERY_DEV_CAP_MAD_DEMUX_OFFSET);
|
|
|
|
|
if (field32 & (1 << 0))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_MAD_DEMUX;
|
|
|
|
|
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
MLX4_GET(dev_cap->dmfs_high_rate_qpn_base, outbox,
|
|
|
|
|
QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_BASE_OFFSET);
|
|
|
|
|
dev_cap->dmfs_high_rate_qpn_base &= MGM_QPN_MASK;
|
|
|
|
|
MLX4_GET(dev_cap->dmfs_high_rate_qpn_range, outbox,
|
|
|
|
|
QUERY_DEV_CAP_DMFS_HIGH_RATE_QPN_RANGE_OFFSET);
|
|
|
|
|
dev_cap->dmfs_high_rate_qpn_range &= MGM_QPN_MASK;
|
|
|
|
|
|
2015-03-18 19:57:34 +07:00
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_QP_RATE_LIMIT_NUM_OFFSET);
|
|
|
|
|
dev_cap->rl_caps.num_rates = size;
|
|
|
|
|
if (dev_cap->rl_caps.num_rates) {
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_QP_RATE_LIMIT_MAX_OFFSET);
|
|
|
|
|
dev_cap->rl_caps.max_val = size & 0xfff;
|
|
|
|
|
dev_cap->rl_caps.max_unit = size >> 14;
|
|
|
|
|
MLX4_GET(size, outbox, QUERY_DEV_CAP_QP_RATE_LIMIT_MIN_OFFSET);
|
|
|
|
|
dev_cap->rl_caps.min_val = size & 0xfff;
|
|
|
|
|
dev_cap->rl_caps.min_unit = size >> 14;
|
|
|
|
|
}
|
|
|
|
|
|
2013-04-25 12:22:28 +07:00
|
|
|
MLX4_GET(field32, outbox, QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET);
|
2013-06-27 23:05:21 +07:00
|
|
|
if (field32 & (1 << 16))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_UPDATE_QP;
|
2015-10-15 18:44:38 +07:00
|
|
|
if (field32 & (1 << 18))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB;
|
|
|
|
|
if (field32 & (1 << 19))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_LB_SRC_CHK;
|
2013-04-25 12:22:28 +07:00
|
|
|
if (field32 & (1 << 26))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_VLAN_CONTROL;
|
2013-04-25 12:22:29 +07:00
|
|
|
if (field32 & (1 << 20))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_FSM;
|
2014-11-13 19:45:33 +07:00
|
|
|
if (field32 & (1 << 21))
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_80_VFS;
|
2013-04-25 12:22:28 +07:00
|
|
|
|
2014-12-11 15:57:59 +07:00
|
|
|
for (i = 1; i <= dev_cap->num_ports; i++) {
|
|
|
|
|
err = mlx4_QUERY_PORT(dev, i, dev_cap->port_cap + i);
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
2007-06-18 22:15:02 +07:00
|
|
|
}
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
/*
|
|
|
|
|
* Each UAR has 4 EQ doorbells; so if a UAR is reserved, then
|
|
|
|
|
* we can't use any EQs whose doorbell falls on that page,
|
|
|
|
|
* even if the EQ itself isn't reserved.
|
|
|
|
|
*/
|
2014-11-13 19:45:32 +07:00
|
|
|
if (dev_cap->num_sys_eqs == 0)
|
|
|
|
|
dev_cap->reserved_eqs = max(dev_cap->reserved_uars * 4,
|
|
|
|
|
dev_cap->reserved_eqs);
|
|
|
|
|
else
|
|
|
|
|
dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_SYS_EQS;
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2014-12-14 21:18:05 +07:00
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void mlx4_dev_cap_dump(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
|
|
|
|
|
{
|
|
|
|
|
if (dev_cap->bf_reg_size > 0)
|
|
|
|
|
mlx4_dbg(dev, "BlueFlame available (reg size %d, regs/page %d)\n",
|
|
|
|
|
dev_cap->bf_reg_size, dev_cap->bf_regs_per_page);
|
|
|
|
|
else
|
|
|
|
|
mlx4_dbg(dev, "BlueFlame not available\n");
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Base MM extensions: flags %08x, rsvd L_Key %08x\n",
|
|
|
|
|
dev_cap->bmme_flags, dev_cap->reserved_lkey);
|
2007-05-09 08:00:38 +07:00
|
|
|
mlx4_dbg(dev, "Max ICM size %lld MB\n",
|
|
|
|
|
(unsigned long long) dev_cap->max_icm_sz >> 20);
|
|
|
|
|
mlx4_dbg(dev, "Max QPs: %d, reserved QPs: %d, entry size: %d\n",
|
|
|
|
|
dev_cap->max_qps, dev_cap->reserved_qps, dev_cap->qpc_entry_sz);
|
|
|
|
|
mlx4_dbg(dev, "Max SRQs: %d, reserved SRQs: %d, entry size: %d\n",
|
|
|
|
|
dev_cap->max_srqs, dev_cap->reserved_srqs, dev_cap->srq_entry_sz);
|
|
|
|
|
mlx4_dbg(dev, "Max CQs: %d, reserved CQs: %d, entry size: %d\n",
|
|
|
|
|
dev_cap->max_cqs, dev_cap->reserved_cqs, dev_cap->cqc_entry_sz);
|
2014-11-13 19:45:32 +07:00
|
|
|
mlx4_dbg(dev, "Num sys EQs: %d, max EQs: %d, reserved EQs: %d, entry size: %d\n",
|
|
|
|
|
dev_cap->num_sys_eqs, dev_cap->max_eqs, dev_cap->reserved_eqs,
|
|
|
|
|
dev_cap->eqc_entry_sz);
|
2007-05-09 08:00:38 +07:00
|
|
|
mlx4_dbg(dev, "reserved MPTs: %d, reserved MTTs: %d\n",
|
|
|
|
|
dev_cap->reserved_mrws, dev_cap->reserved_mtts);
|
|
|
|
|
mlx4_dbg(dev, "Max PDs: %d, reserved PDs: %d, reserved UARs: %d\n",
|
|
|
|
|
dev_cap->max_pds, dev_cap->reserved_pds, dev_cap->reserved_uars);
|
|
|
|
|
mlx4_dbg(dev, "Max QP/MCG: %d, reserved MGMs: %d\n",
|
|
|
|
|
dev_cap->max_pds, dev_cap->reserved_mgms);
|
|
|
|
|
mlx4_dbg(dev, "Max CQEs: %d, max WQEs: %d, max SRQ WQEs: %d\n",
|
|
|
|
|
dev_cap->max_cq_sz, dev_cap->max_qp_sz, dev_cap->max_srq_sz);
|
|
|
|
|
mlx4_dbg(dev, "Local CA ACK delay: %d, max MTU: %d, port width cap: %d\n",
|
2014-12-11 15:57:59 +07:00
|
|
|
dev_cap->local_ca_ack_delay, 128 << dev_cap->port_cap[1].ib_mtu,
|
|
|
|
|
dev_cap->port_cap[1].max_port_width);
|
2007-05-09 08:00:38 +07:00
|
|
|
mlx4_dbg(dev, "Max SQ desc size: %d, max SQ S/G: %d\n",
|
|
|
|
|
dev_cap->max_sq_desc_sz, dev_cap->max_sq_sg);
|
|
|
|
|
mlx4_dbg(dev, "Max RQ desc size: %d, max RQ S/G: %d\n",
|
|
|
|
|
dev_cap->max_rq_desc_sz, dev_cap->max_rq_sg);
|
2008-04-17 11:09:27 +07:00
|
|
|
mlx4_dbg(dev, "Max GSO size: %d\n", dev_cap->max_gso_sz);
|
2011-06-15 21:47:14 +07:00
|
|
|
mlx4_dbg(dev, "Max counters: %d\n", dev_cap->max_counters);
|
2012-04-29 21:04:25 +07:00
|
|
|
mlx4_dbg(dev, "Max RSS Table size: %d\n", dev_cap->max_rss_tbl_sz);
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
mlx4_dbg(dev, "DMFS high rate steer QPn base: %d\n",
|
|
|
|
|
dev_cap->dmfs_high_rate_qpn_base);
|
|
|
|
|
mlx4_dbg(dev, "DMFS high rate steer QPn range: %d\n",
|
|
|
|
|
dev_cap->dmfs_high_rate_qpn_range);
|
2015-03-18 19:57:34 +07:00
|
|
|
|
|
|
|
|
if (dev_cap->flags2 & MLX4_DEV_CAP_FLAG2_QP_RATE_LIMIT) {
|
|
|
|
|
struct mlx4_rate_limit_caps *rl_caps = &dev_cap->rl_caps;
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "QP Rate-Limit: #rates %d, unit/val max %d/%d, min %d/%d\n",
|
|
|
|
|
rl_caps->num_rates, rl_caps->max_unit, rl_caps->max_val,
|
|
|
|
|
rl_caps->min_unit, rl_caps->min_val);
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
dump_dev_cap_flags(dev, dev_cap->flags);
|
2012-04-29 21:04:25 +07:00
|
|
|
dump_dev_cap_flags2(dev, dev_cap->flags2);
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
|
|
|
|
|
2014-12-11 15:57:59 +07:00
|
|
|
int mlx4_QUERY_PORT(struct mlx4_dev *dev, int port, struct mlx4_port_cap *port_cap)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
u8 field;
|
|
|
|
|
u32 field32;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
if (dev->flags & MLX4_FLAG_OLD_PORT_CMDS) {
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_QUERY_DEV_CAP,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_VL_PORT_OFFSET);
|
|
|
|
|
port_cap->max_vl = field >> 4;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MTU_WIDTH_OFFSET);
|
|
|
|
|
port_cap->ib_mtu = field >> 4;
|
|
|
|
|
port_cap->max_port_width = field & 0xf;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_GID_OFFSET);
|
|
|
|
|
port_cap->max_gids = 1 << (field & 0xf);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_DEV_CAP_MAX_PKEY_OFFSET);
|
|
|
|
|
port_cap->max_pkeys = 1 << (field & 0xf);
|
|
|
|
|
} else {
|
|
|
|
|
#define QUERY_PORT_SUPPORTED_TYPE_OFFSET 0x00
|
|
|
|
|
#define QUERY_PORT_MTU_OFFSET 0x01
|
|
|
|
|
#define QUERY_PORT_ETH_MTU_OFFSET 0x02
|
|
|
|
|
#define QUERY_PORT_WIDTH_OFFSET 0x06
|
|
|
|
|
#define QUERY_PORT_MAX_GID_PKEY_OFFSET 0x07
|
|
|
|
|
#define QUERY_PORT_MAX_MACVLAN_OFFSET 0x0a
|
|
|
|
|
#define QUERY_PORT_MAX_VL_OFFSET 0x0b
|
|
|
|
|
#define QUERY_PORT_MAC_OFFSET 0x10
|
|
|
|
|
#define QUERY_PORT_TRANS_VENDOR_OFFSET 0x18
|
|
|
|
|
#define QUERY_PORT_WAVELENGTH_OFFSET 0x1c
|
|
|
|
|
#define QUERY_PORT_TRANS_CODE_OFFSET 0x20
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, port, 0, MLX4_CMD_QUERY_PORT,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_SUPPORTED_TYPE_OFFSET);
|
2015-12-06 23:07:38 +07:00
|
|
|
port_cap->link_state = (field & 0x80) >> 7;
|
2014-12-11 15:57:59 +07:00
|
|
|
port_cap->supported_port_types = field & 3;
|
|
|
|
|
port_cap->suggested_type = (field >> 3) & 1;
|
|
|
|
|
port_cap->default_sense = (field >> 4) & 1;
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
port_cap->dmfs_optimized_state = (field >> 5) & 1;
|
2014-12-11 15:57:59 +07:00
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_MTU_OFFSET);
|
|
|
|
|
port_cap->ib_mtu = field & 0xf;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_WIDTH_OFFSET);
|
|
|
|
|
port_cap->max_port_width = field & 0xf;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_MAX_GID_PKEY_OFFSET);
|
|
|
|
|
port_cap->max_gids = 1 << (field >> 4);
|
|
|
|
|
port_cap->max_pkeys = 1 << (field & 0xf);
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_MAX_VL_OFFSET);
|
|
|
|
|
port_cap->max_vl = field & 0xf;
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_MAX_MACVLAN_OFFSET);
|
|
|
|
|
port_cap->log_max_macs = field & 0xf;
|
|
|
|
|
port_cap->log_max_vlans = field >> 4;
|
|
|
|
|
MLX4_GET(port_cap->eth_mtu, outbox, QUERY_PORT_ETH_MTU_OFFSET);
|
|
|
|
|
MLX4_GET(port_cap->def_mac, outbox, QUERY_PORT_MAC_OFFSET);
|
|
|
|
|
MLX4_GET(field32, outbox, QUERY_PORT_TRANS_VENDOR_OFFSET);
|
|
|
|
|
port_cap->trans_type = field32 >> 24;
|
|
|
|
|
port_cap->vendor_oui = field32 & 0xffffff;
|
|
|
|
|
MLX4_GET(port_cap->wavelength, outbox, QUERY_PORT_WAVELENGTH_OFFSET);
|
|
|
|
|
MLX4_GET(port_cap->trans_code, outbox, QUERY_PORT_TRANS_CODE_OFFSET);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2015-03-30 21:45:25 +07:00
|
|
|
#define DEV_CAP_EXT_2_FLAG_PFC_COUNTERS (1 << 28)
|
2014-12-11 15:57:52 +07:00
|
|
|
#define DEV_CAP_EXT_2_FLAG_VLAN_CONTROL (1 << 26)
|
|
|
|
|
#define DEV_CAP_EXT_2_FLAG_80_VFS (1 << 21)
|
|
|
|
|
#define DEV_CAP_EXT_2_FLAG_FSM (1 << 20)
|
|
|
|
|
|
2012-05-30 16:14:53 +07:00
|
|
|
int mlx4_QUERY_DEV_CAP_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
2012-08-03 15:40:50 +07:00
|
|
|
u64 flags;
|
2012-05-30 16:14:53 +07:00
|
|
|
int err = 0;
|
|
|
|
|
u8 field;
|
2015-03-18 19:57:34 +07:00
|
|
|
u16 field16;
|
2014-12-11 15:57:52 +07:00
|
|
|
u32 bmme_flags, field32;
|
2014-03-19 23:11:52 +07:00
|
|
|
int real_port;
|
|
|
|
|
int slave_port;
|
|
|
|
|
int first_port;
|
|
|
|
|
struct mlx4_active_ports actv_ports;
|
2012-05-30 16:14:53 +07:00
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, outbox->dma, 0, 0, MLX4_CMD_QUERY_DEV_CAP,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
|
2013-02-06 23:19:10 +07:00
|
|
|
/* add port mng change event capability and disable mw type 1
|
|
|
|
|
* unconditionally to slaves
|
|
|
|
|
*/
|
2012-08-03 15:40:50 +07:00
|
|
|
MLX4_GET(flags, outbox->buf, QUERY_DEV_CAP_EXT_FLAGS_OFFSET);
|
|
|
|
|
flags |= MLX4_DEV_CAP_FLAG_PORT_MNG_CHG_EV;
|
2013-02-06 23:19:10 +07:00
|
|
|
flags &= ~MLX4_DEV_CAP_FLAG_MEM_WINDOW;
|
2014-03-19 23:11:52 +07:00
|
|
|
actv_ports = mlx4_get_active_ports(dev, slave);
|
|
|
|
|
first_port = find_first_bit(actv_ports.ports, dev->caps.num_ports);
|
|
|
|
|
for (slave_port = 0, real_port = first_port;
|
|
|
|
|
real_port < first_port +
|
|
|
|
|
bitmap_weight(actv_ports.ports, dev->caps.num_ports);
|
|
|
|
|
++real_port, ++slave_port) {
|
|
|
|
|
if (flags & (MLX4_DEV_CAP_FLAG_WOL_PORT1 << real_port))
|
|
|
|
|
flags |= MLX4_DEV_CAP_FLAG_WOL_PORT1 << slave_port;
|
|
|
|
|
else
|
|
|
|
|
flags &= ~(MLX4_DEV_CAP_FLAG_WOL_PORT1 << slave_port);
|
|
|
|
|
}
|
|
|
|
|
for (; slave_port < dev->caps.num_ports; ++slave_port)
|
|
|
|
|
flags &= ~(MLX4_DEV_CAP_FLAG_WOL_PORT1 << slave_port);
|
2015-04-02 20:31:06 +07:00
|
|
|
|
|
|
|
|
/* Not exposing RSS IP fragments to guests */
|
|
|
|
|
flags &= ~MLX4_DEV_CAP_FLAG_RSS_IP_FRAG;
|
2012-08-03 15:40:50 +07:00
|
|
|
MLX4_PUT(outbox->buf, flags, QUERY_DEV_CAP_EXT_FLAGS_OFFSET);
|
|
|
|
|
|
2014-03-19 23:11:52 +07:00
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_VL_PORT_OFFSET);
|
|
|
|
|
field &= ~0x0F;
|
|
|
|
|
field |= bitmap_weight(actv_ports.ports, dev->caps.num_ports) & 0x0F;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_VL_PORT_OFFSET);
|
|
|
|
|
|
2013-04-25 12:22:23 +07:00
|
|
|
/* For guests, disable timestamp */
|
|
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_CQ_TS_SUPPORT_OFFSET);
|
|
|
|
|
field &= 0x7f;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_CQ_TS_SUPPORT_OFFSET);
|
|
|
|
|
|
2015-04-02 20:31:17 +07:00
|
|
|
/* For guests, disable vxlan tunneling and QoS support */
|
2014-03-06 23:28:16 +07:00
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_VXLAN);
|
2015-04-02 20:31:17 +07:00
|
|
|
field &= 0xd7;
|
2013-12-23 21:09:43 +07:00
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_VXLAN);
|
|
|
|
|
|
2015-04-02 20:31:20 +07:00
|
|
|
/* For guests, disable port BEACON */
|
|
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_PORT_BEACON_OFFSET);
|
|
|
|
|
field &= 0x7f;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_PORT_BEACON_OFFSET);
|
|
|
|
|
|
2012-05-30 16:14:53 +07:00
|
|
|
/* For guests, report Blueflame disabled */
|
|
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_BF_OFFSET);
|
|
|
|
|
field &= 0x7f;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_BF_OFFSET);
|
|
|
|
|
|
2015-02-03 21:48:32 +07:00
|
|
|
/* For guests, disable mw type 2 and port remap*/
|
2014-03-06 23:28:16 +07:00
|
|
|
MLX4_GET(bmme_flags, outbox->buf, QUERY_DEV_CAP_BMME_FLAGS_OFFSET);
|
2013-02-06 23:19:10 +07:00
|
|
|
bmme_flags &= ~MLX4_BMME_FLAG_TYPE_2_WIN;
|
2015-02-03 21:48:32 +07:00
|
|
|
bmme_flags &= ~MLX4_FLAG_PORT_REMAP;
|
2013-02-06 23:19:10 +07:00
|
|
|
MLX4_PUT(outbox->buf, bmme_flags, QUERY_DEV_CAP_BMME_FLAGS_OFFSET);
|
|
|
|
|
|
2013-03-07 10:46:53 +07:00
|
|
|
/* turn off device-managed steering capability if not enabled */
|
|
|
|
|
if (dev->caps.steering_mode != MLX4_STEERING_MODE_DEVICE_MANAGED) {
|
|
|
|
|
MLX4_GET(field, outbox->buf,
|
|
|
|
|
QUERY_DEV_CAP_FLOW_STEERING_RANGE_EN_OFFSET);
|
|
|
|
|
field &= 0x7f;
|
|
|
|
|
MLX4_PUT(outbox->buf, field,
|
|
|
|
|
QUERY_DEV_CAP_FLOW_STEERING_RANGE_EN_OFFSET);
|
|
|
|
|
}
|
2013-11-07 20:25:14 +07:00
|
|
|
|
|
|
|
|
/* turn off ipoib managed steering for guests */
|
2014-03-06 23:28:16 +07:00
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_FLOW_STEERING_IPOIB_OFFSET);
|
2013-11-07 20:25:14 +07:00
|
|
|
field &= ~0x80;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_FLOW_STEERING_IPOIB_OFFSET);
|
|
|
|
|
|
2014-12-11 15:57:52 +07:00
|
|
|
/* turn off host side virt features (VST, FSM, etc) for guests */
|
|
|
|
|
MLX4_GET(field32, outbox->buf, QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET);
|
|
|
|
|
field32 &= ~(DEV_CAP_EXT_2_FLAG_VLAN_CONTROL | DEV_CAP_EXT_2_FLAG_80_VFS |
|
2015-03-30 21:45:25 +07:00
|
|
|
DEV_CAP_EXT_2_FLAG_FSM | DEV_CAP_EXT_2_FLAG_PFC_COUNTERS);
|
2014-12-11 15:57:52 +07:00
|
|
|
MLX4_PUT(outbox->buf, field32, QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET);
|
|
|
|
|
|
2015-03-06 01:16:12 +07:00
|
|
|
/* turn off QCN for guests */
|
|
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_ECN_QCN_VER_OFFSET);
|
|
|
|
|
field &= 0xfe;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_ECN_QCN_VER_OFFSET);
|
|
|
|
|
|
2015-03-18 19:57:34 +07:00
|
|
|
/* turn off QP max-rate limiting for guests */
|
|
|
|
|
field16 = 0;
|
|
|
|
|
MLX4_PUT(outbox->buf, field16, QUERY_DEV_CAP_QP_RATE_LIMIT_NUM_OFFSET);
|
|
|
|
|
|
2015-04-02 20:31:13 +07:00
|
|
|
/* turn off QoS per VF support for guests */
|
|
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_CQ_EQ_CACHE_LINE_STRIDE);
|
|
|
|
|
field &= 0xef;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_CQ_EQ_CACHE_LINE_STRIDE);
|
|
|
|
|
|
2015-04-02 20:31:22 +07:00
|
|
|
/* turn off ignore FCS feature for guests */
|
|
|
|
|
MLX4_GET(field, outbox->buf, QUERY_DEV_CAP_CONFIG_DEV_OFFSET);
|
|
|
|
|
field &= 0xfb;
|
|
|
|
|
MLX4_PUT(outbox->buf, field, QUERY_DEV_CAP_CONFIG_DEV_OFFSET);
|
|
|
|
|
|
2012-05-30 16:14:53 +07:00
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int mlx4_QUERY_PORT_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
2013-04-25 12:22:26 +07:00
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
u64 def_mac;
|
|
|
|
|
u8 port_type;
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 15:21:44 +07:00
|
|
|
u16 short_field;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int err;
|
2013-06-13 17:19:11 +07:00
|
|
|
int admin_link_state;
|
2014-03-19 23:11:52 +07:00
|
|
|
int port = mlx4_slave_convert_port(dev, slave,
|
|
|
|
|
vhcr->in_modifier & 0xFF);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2012-06-19 15:21:43 +07:00
|
|
|
#define MLX4_VF_PORT_NO_LINK_SENSE_MASK 0xE0
|
2013-06-13 17:19:11 +07:00
|
|
|
#define MLX4_PORT_LINK_UP_MASK 0x80
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 15:21:44 +07:00
|
|
|
#define QUERY_PORT_CUR_MAX_PKEY_OFFSET 0x0c
|
|
|
|
|
#define QUERY_PORT_CUR_MAX_GID_OFFSET 0x0e
|
2011-12-29 14:42:39 +07:00
|
|
|
|
2014-03-19 23:11:52 +07:00
|
|
|
if (port < 0)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
2014-09-30 16:03:49 +07:00
|
|
|
/* Protect against untrusted guests: enforce that this is the
|
|
|
|
|
* QUERY_PORT general query.
|
|
|
|
|
*/
|
|
|
|
|
if (vhcr->op_modifier || vhcr->in_modifier & ~0xFF)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
vhcr->in_modifier = port;
|
2014-03-19 23:11:52 +07:00
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
err = mlx4_cmd_box(dev, 0, outbox->dma, vhcr->in_modifier, 0,
|
|
|
|
|
MLX4_CMD_QUERY_PORT, MLX4_CMD_TIME_CLASS_B,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
if (!err && dev->caps.function != slave) {
|
2013-08-01 23:55:00 +07:00
|
|
|
def_mac = priv->mfunc.master.vf_oper[slave].vport[vhcr->in_modifier].state.mac;
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, def_mac, QUERY_PORT_MAC_OFFSET);
|
|
|
|
|
|
|
|
|
|
/* get port type - currently only eth is enabled */
|
|
|
|
|
MLX4_GET(port_type, outbox->buf,
|
|
|
|
|
QUERY_PORT_SUPPORTED_TYPE_OFFSET);
|
|
|
|
|
|
2012-06-19 15:21:43 +07:00
|
|
|
/* No link sensing allowed */
|
|
|
|
|
port_type &= MLX4_VF_PORT_NO_LINK_SENSE_MASK;
|
|
|
|
|
/* set port type to currently operating port type */
|
|
|
|
|
port_type |= (dev->caps.port_type[vhcr->in_modifier] & 0x3);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2013-06-13 17:19:11 +07:00
|
|
|
admin_link_state = priv->mfunc.master.vf_oper[slave].vport[vhcr->in_modifier].state.link_state;
|
|
|
|
|
if (IFLA_VF_LINK_STATE_ENABLE == admin_link_state)
|
|
|
|
|
port_type |= MLX4_PORT_LINK_UP_MASK;
|
|
|
|
|
else if (IFLA_VF_LINK_STATE_DISABLE == admin_link_state)
|
|
|
|
|
port_type &= ~MLX4_PORT_LINK_UP_MASK;
|
2015-12-06 23:07:38 +07:00
|
|
|
else if (IFLA_VF_LINK_STATE_AUTO == admin_link_state && mlx4_is_bonded(dev)) {
|
|
|
|
|
int other_port = (port == 1) ? 2 : 1;
|
|
|
|
|
struct mlx4_port_cap port_cap;
|
|
|
|
|
|
|
|
|
|
err = mlx4_QUERY_PORT(dev, other_port, &port_cap);
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
port_type |= (port_cap.link_state << 7);
|
|
|
|
|
}
|
2013-06-13 17:19:11 +07:00
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_PUT(outbox->buf, port_type,
|
|
|
|
|
QUERY_PORT_SUPPORTED_TYPE_OFFSET);
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 15:21:44 +07:00
|
|
|
|
2014-03-12 17:00:39 +07:00
|
|
|
if (dev->caps.port_type[vhcr->in_modifier] == MLX4_PORT_TYPE_ETH)
|
2014-03-19 23:11:52 +07:00
|
|
|
short_field = mlx4_get_slave_num_gids(dev, slave, port);
|
2014-03-12 17:00:39 +07:00
|
|
|
else
|
|
|
|
|
short_field = 1; /* slave max gids */
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 15:21:44 +07:00
|
|
|
MLX4_PUT(outbox->buf, short_field,
|
|
|
|
|
QUERY_PORT_CUR_MAX_GID_OFFSET);
|
|
|
|
|
|
|
|
|
|
short_field = dev->caps.pkey_table_len[vhcr->in_modifier];
|
|
|
|
|
MLX4_PUT(outbox->buf, short_field,
|
|
|
|
|
QUERY_PORT_CUR_MAX_PKEY_OFFSET);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
}
|
2015-12-06 23:07:38 +07:00
|
|
|
out:
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
To allow easy paravirtualization of P_Key and GID table sizes, keep
paravirtualized sizes in mlx4_dev->caps, but save the actual physical
sizes from FW in struct: mlx4_dev->phys_cap.
In addition, in SR-IOV mode, do the following:
1. Reduce reported P_Key table size by 1.
This is done to reserve the highest P_Key index for internal use,
for declaring an invalid P_Key in P_Key paravirtualization.
We require a P_Key index which always contain an invalid P_Key
value for this purpose (i.e., one which cannot be modified by
the subnet manager). The way to do this is to reduce the
P_Key table size reported to the subnet manager by 1, so that
it will not attempt to access the P_Key at index #127.
2. Paravirtualize the GID table size to 1. Thus, each guest sees
only a single GID (at its paravirtualized index 0).
In addition, since we are paravirtualizing the GID table size to 1, we
add paravirtualization of the master GID event here (i.e., we do not
do ib_dispatch_event() for the GUID change event on the master, since
its (only) GUID never changes).
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-06-19 15:21:44 +07:00
|
|
|
int mlx4_get_slave_pkey_gid_tbl_len(struct mlx4_dev *dev, u8 port,
|
|
|
|
|
int *gid_tbl_len, int *pkey_tbl_len)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
u16 field;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, port, 0,
|
|
|
|
|
MLX4_CMD_QUERY_PORT, MLX4_CMD_TIME_CLASS_B,
|
|
|
|
|
MLX4_CMD_WRAPPED);
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_CUR_MAX_GID_OFFSET);
|
|
|
|
|
*gid_tbl_len = field;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(field, outbox, QUERY_PORT_CUR_MAX_PKEY_OFFSET);
|
|
|
|
|
*pkey_tbl_len = field;
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(mlx4_get_slave_pkey_gid_tbl_len);
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
int mlx4_map_cmd(struct mlx4_dev *dev, u16 op, struct mlx4_icm *icm, u64 virt)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
struct mlx4_icm_iter iter;
|
|
|
|
|
__be64 *pages;
|
|
|
|
|
int lg;
|
|
|
|
|
int nent = 0;
|
|
|
|
|
int i;
|
|
|
|
|
int err = 0;
|
|
|
|
|
int ts = 0, tc = 0;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
pages = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
for (mlx4_icm_first(icm, &iter);
|
|
|
|
|
!mlx4_icm_last(&iter);
|
|
|
|
|
mlx4_icm_next(&iter)) {
|
|
|
|
|
/*
|
|
|
|
|
* We have to pass pages that are aligned to their
|
|
|
|
|
* size, so find the least significant 1 in the
|
|
|
|
|
* address or size and use that as our log2 size.
|
|
|
|
|
*/
|
|
|
|
|
lg = ffs(mlx4_icm_addr(&iter) | mlx4_icm_size(&iter)) - 1;
|
|
|
|
|
if (lg < MLX4_ICM_PAGE_SHIFT) {
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_warn(dev, "Got FW area not aligned to %d (%llx/%lx)\n",
|
|
|
|
|
MLX4_ICM_PAGE_SIZE,
|
|
|
|
|
(unsigned long long) mlx4_icm_addr(&iter),
|
|
|
|
|
mlx4_icm_size(&iter));
|
2007-05-09 08:00:38 +07:00
|
|
|
err = -EINVAL;
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < mlx4_icm_size(&iter) >> lg; ++i) {
|
|
|
|
|
if (virt != -1) {
|
|
|
|
|
pages[nent * 2] = cpu_to_be64(virt);
|
|
|
|
|
virt += 1 << lg;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
pages[nent * 2 + 1] =
|
|
|
|
|
cpu_to_be64((mlx4_icm_addr(&iter) + (i << lg)) |
|
|
|
|
|
(lg - MLX4_ICM_PAGE_SHIFT));
|
|
|
|
|
ts += 1 << (lg - 10);
|
|
|
|
|
++tc;
|
|
|
|
|
|
|
|
|
|
if (++nent == MLX4_MAILBOX_SIZE / 16) {
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, nent, 0, op,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_B,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
nent = 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (nent)
|
2011-12-13 11:10:51 +07:00
|
|
|
err = mlx4_cmd(dev, mailbox->dma, nent, 0, op,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
|
case MLX4_CMD_MAP_FA:
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_dbg(dev, "Mapped %d chunks/%d KB for FW\n", tc, ts);
|
2007-05-09 08:00:38 +07:00
|
|
|
break;
|
|
|
|
|
case MLX4_CMD_MAP_ICM_AUX:
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_dbg(dev, "Mapped %d chunks/%d KB for ICM aux\n", tc, ts);
|
2007-05-09 08:00:38 +07:00
|
|
|
break;
|
|
|
|
|
case MLX4_CMD_MAP_ICM:
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_dbg(dev, "Mapped %d chunks/%d KB at %llx for ICM\n",
|
|
|
|
|
tc, ts, (unsigned long long) virt - (ts << 10));
|
2007-05-09 08:00:38 +07:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_MAP_FA(struct mlx4_dev *dev, struct mlx4_icm *icm)
|
|
|
|
|
{
|
|
|
|
|
return mlx4_map_cmd(dev, MLX4_CMD_MAP_FA, icm, -1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_UNMAP_FA(struct mlx4_dev *dev)
|
|
|
|
|
{
|
2011-12-13 11:10:51 +07:00
|
|
|
return mlx4_cmd(dev, 0, 0, 0, MLX4_CMD_UNMAP_FA,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
int mlx4_RUN_FW(struct mlx4_dev *dev)
|
|
|
|
|
{
|
2011-12-13 11:10:51 +07:00
|
|
|
return mlx4_cmd(dev, 0, 0, 0, MLX4_CMD_RUN_FW,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_QUERY_FW(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_fw *fw = &mlx4_priv(dev)->fw;
|
|
|
|
|
struct mlx4_cmd *cmd = &mlx4_priv(dev)->cmd;
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
int err = 0;
|
|
|
|
|
u64 fw_ver;
|
2007-06-08 13:24:36 +07:00
|
|
|
u16 cmd_if_rev;
|
2007-05-09 08:00:38 +07:00
|
|
|
u8 lg;
|
|
|
|
|
|
|
|
|
|
#define QUERY_FW_OUT_SIZE 0x100
|
|
|
|
|
#define QUERY_FW_VER_OFFSET 0x00
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
#define QUERY_FW_PPF_ID 0x09
|
2007-06-08 13:24:36 +07:00
|
|
|
#define QUERY_FW_CMD_IF_REV_OFFSET 0x0a
|
2007-05-09 08:00:38 +07:00
|
|
|
#define QUERY_FW_MAX_CMD_OFFSET 0x0f
|
|
|
|
|
#define QUERY_FW_ERR_START_OFFSET 0x30
|
|
|
|
|
#define QUERY_FW_ERR_SIZE_OFFSET 0x38
|
|
|
|
|
#define QUERY_FW_ERR_BAR_OFFSET 0x3c
|
|
|
|
|
|
|
|
|
|
#define QUERY_FW_SIZE_OFFSET 0x00
|
|
|
|
|
#define QUERY_FW_CLR_INT_BASE_OFFSET 0x20
|
|
|
|
|
#define QUERY_FW_CLR_INT_BAR_OFFSET 0x28
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
#define QUERY_FW_COMM_BASE_OFFSET 0x40
|
|
|
|
|
#define QUERY_FW_COMM_BAR_OFFSET 0x48
|
|
|
|
|
|
2013-04-23 13:06:48 +07:00
|
|
|
#define QUERY_FW_CLOCK_OFFSET 0x50
|
|
|
|
|
#define QUERY_FW_CLOCK_BAR 0x58
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_QUERY_FW,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(fw_ver, outbox, QUERY_FW_VER_OFFSET);
|
|
|
|
|
/*
|
2007-06-04 09:47:10 +07:00
|
|
|
* FW subminor version is at more significant bits than minor
|
2007-05-09 08:00:38 +07:00
|
|
|
* version, so swap here.
|
|
|
|
|
*/
|
|
|
|
|
dev->caps.fw_ver = (fw_ver & 0xffff00000000ull) |
|
|
|
|
|
((fw_ver & 0xffff0000ull) >> 16) |
|
|
|
|
|
((fw_ver & 0x0000ffffull) << 16);
|
|
|
|
|
|
2012-06-19 15:21:33 +07:00
|
|
|
MLX4_GET(lg, outbox, QUERY_FW_PPF_ID);
|
|
|
|
|
dev->caps.function = lg;
|
|
|
|
|
|
2012-05-30 16:14:53 +07:00
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
goto out;
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
|
2007-06-08 13:24:36 +07:00
|
|
|
MLX4_GET(cmd_if_rev, outbox, QUERY_FW_CMD_IF_REV_OFFSET);
|
2007-06-18 22:15:02 +07:00
|
|
|
if (cmd_if_rev < MLX4_COMMAND_INTERFACE_MIN_REV ||
|
|
|
|
|
cmd_if_rev > MLX4_COMMAND_INTERFACE_MAX_REV) {
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_err(dev, "Installed FW has unsupported command interface revision %d\n",
|
2007-06-08 13:24:36 +07:00
|
|
|
cmd_if_rev);
|
|
|
|
|
mlx4_err(dev, "(Installed FW version is %d.%d.%03d)\n",
|
|
|
|
|
(int) (dev->caps.fw_ver >> 32),
|
|
|
|
|
(int) (dev->caps.fw_ver >> 16) & 0xffff,
|
|
|
|
|
(int) dev->caps.fw_ver & 0xffff);
|
2014-05-08 02:52:57 +07:00
|
|
|
mlx4_err(dev, "This driver version supports only revisions %d to %d\n",
|
2007-06-18 22:15:02 +07:00
|
|
|
MLX4_COMMAND_INTERFACE_MIN_REV, MLX4_COMMAND_INTERFACE_MAX_REV);
|
2007-06-08 13:24:36 +07:00
|
|
|
err = -ENODEV;
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
if (cmd_if_rev < MLX4_COMMAND_INTERFACE_NEW_PORT_CMDS)
|
|
|
|
|
dev->flags |= MLX4_FLAG_OLD_PORT_CMDS;
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_GET(lg, outbox, QUERY_FW_MAX_CMD_OFFSET);
|
|
|
|
|
cmd->max_cmds = 1 << lg;
|
|
|
|
|
|
2007-06-08 13:24:36 +07:00
|
|
|
mlx4_dbg(dev, "FW version %d.%d.%03d (cmd intf rev %d), max commands %d\n",
|
2007-05-09 08:00:38 +07:00
|
|
|
(int) (dev->caps.fw_ver >> 32),
|
|
|
|
|
(int) (dev->caps.fw_ver >> 16) & 0xffff,
|
|
|
|
|
(int) dev->caps.fw_ver & 0xffff,
|
2007-06-08 13:24:36 +07:00
|
|
|
cmd_if_rev, cmd->max_cmds);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
MLX4_GET(fw->catas_offset, outbox, QUERY_FW_ERR_START_OFFSET);
|
|
|
|
|
MLX4_GET(fw->catas_size, outbox, QUERY_FW_ERR_SIZE_OFFSET);
|
|
|
|
|
MLX4_GET(fw->catas_bar, outbox, QUERY_FW_ERR_BAR_OFFSET);
|
|
|
|
|
fw->catas_bar = (fw->catas_bar >> 6) * 2;
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Catastrophic error buffer at 0x%llx, size 0x%x, BAR %d\n",
|
|
|
|
|
(unsigned long long) fw->catas_offset, fw->catas_size, fw->catas_bar);
|
|
|
|
|
|
|
|
|
|
MLX4_GET(fw->fw_pages, outbox, QUERY_FW_SIZE_OFFSET);
|
|
|
|
|
MLX4_GET(fw->clr_int_base, outbox, QUERY_FW_CLR_INT_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(fw->clr_int_bar, outbox, QUERY_FW_CLR_INT_BAR_OFFSET);
|
|
|
|
|
fw->clr_int_bar = (fw->clr_int_bar >> 6) * 2;
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
MLX4_GET(fw->comm_base, outbox, QUERY_FW_COMM_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(fw->comm_bar, outbox, QUERY_FW_COMM_BAR_OFFSET);
|
|
|
|
|
fw->comm_bar = (fw->comm_bar >> 6) * 2;
|
|
|
|
|
mlx4_dbg(dev, "Communication vector bar:%d offset:0x%llx\n",
|
|
|
|
|
fw->comm_bar, fw->comm_base);
|
2007-05-09 08:00:38 +07:00
|
|
|
mlx4_dbg(dev, "FW size %d KB\n", fw->fw_pages >> 2);
|
|
|
|
|
|
2013-04-23 13:06:48 +07:00
|
|
|
MLX4_GET(fw->clock_offset, outbox, QUERY_FW_CLOCK_OFFSET);
|
|
|
|
|
MLX4_GET(fw->clock_bar, outbox, QUERY_FW_CLOCK_BAR);
|
|
|
|
|
fw->clock_bar = (fw->clock_bar >> 6) * 2;
|
|
|
|
|
mlx4_dbg(dev, "Internal clock bar:%d offset:0x%llx\n",
|
|
|
|
|
fw->clock_bar, fw->clock_offset);
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
/*
|
|
|
|
|
* Round up number of system pages needed in case
|
|
|
|
|
* MLX4_ICM_PAGE_SIZE < PAGE_SIZE.
|
|
|
|
|
*/
|
|
|
|
|
fw->fw_pages =
|
|
|
|
|
ALIGN(fw->fw_pages, PAGE_SIZE / MLX4_ICM_PAGE_SIZE) >>
|
|
|
|
|
(PAGE_SHIFT - MLX4_ICM_PAGE_SHIFT);
|
|
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Clear int @ %llx, BAR %d\n",
|
|
|
|
|
(unsigned long long) fw->clr_int_base, fw->clr_int_bar);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2012-05-30 16:14:53 +07:00
|
|
|
int mlx4_QUERY_FW_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
|
|
|
|
u8 *outbuf;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
outbuf = outbox->buf;
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, outbox->dma, 0, 0, MLX4_CMD_QUERY_FW,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
|
2012-06-19 15:21:33 +07:00
|
|
|
/* for slaves, set pci PPF ID to invalid and zero out everything
|
|
|
|
|
* else except FW version */
|
2012-05-30 16:14:53 +07:00
|
|
|
outbuf[0] = outbuf[1] = 0;
|
|
|
|
|
memset(&outbuf[8], 0, QUERY_FW_OUT_SIZE - 8);
|
2012-06-19 15:21:33 +07:00
|
|
|
outbuf[QUERY_FW_PPF_ID] = MLX4_INVALID_SLAVE_ID;
|
|
|
|
|
|
2012-05-30 16:14:53 +07:00
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
static void get_board_id(void *vsd, char *board_id)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
#define VSD_OFFSET_SIG1 0x00
|
|
|
|
|
#define VSD_OFFSET_SIG2 0xde
|
|
|
|
|
#define VSD_OFFSET_MLX_BOARD_ID 0xd0
|
|
|
|
|
#define VSD_OFFSET_TS_BOARD_ID 0x20
|
|
|
|
|
|
|
|
|
|
#define VSD_SIGNATURE_TOPSPIN 0x5ad
|
|
|
|
|
|
|
|
|
|
memset(board_id, 0, MLX4_BOARD_ID_LEN);
|
|
|
|
|
|
|
|
|
|
if (be16_to_cpup(vsd + VSD_OFFSET_SIG1) == VSD_SIGNATURE_TOPSPIN &&
|
|
|
|
|
be16_to_cpup(vsd + VSD_OFFSET_SIG2) == VSD_SIGNATURE_TOPSPIN) {
|
|
|
|
|
strlcpy(board_id, vsd + VSD_OFFSET_TS_BOARD_ID, MLX4_BOARD_ID_LEN);
|
|
|
|
|
} else {
|
|
|
|
|
/*
|
|
|
|
|
* The board ID is a string but the firmware byte
|
|
|
|
|
* swaps each 4-byte word before passing it back to
|
|
|
|
|
* us. Therefore we need to swab it before printing.
|
|
|
|
|
*/
|
2015-04-30 03:52:51 +07:00
|
|
|
u32 *bid_u32 = (u32 *)board_id;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < 4; ++i) {
|
|
|
|
|
u32 *addr;
|
|
|
|
|
u32 val;
|
|
|
|
|
|
|
|
|
|
addr = (u32 *) (vsd + VSD_OFFSET_MLX_BOARD_ID + i * 4);
|
|
|
|
|
val = get_unaligned(addr);
|
|
|
|
|
val = swab32(val);
|
|
|
|
|
put_unaligned(val, &bid_u32[i]);
|
|
|
|
|
}
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_QUERY_ADAPTER(struct mlx4_dev *dev, struct mlx4_adapter *adapter)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
#define QUERY_ADAPTER_OUT_SIZE 0x100
|
|
|
|
|
#define QUERY_ADAPTER_INTA_PIN_OFFSET 0x10
|
|
|
|
|
#define QUERY_ADAPTER_VSD_OFFSET 0x20
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_QUERY_ADAPTER,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(adapter->inta_pin, outbox, QUERY_ADAPTER_INTA_PIN_OFFSET);
|
|
|
|
|
|
|
|
|
|
get_board_id(outbox + QUERY_ADAPTER_VSD_OFFSET / 4,
|
|
|
|
|
adapter->board_id);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_INIT_HCA(struct mlx4_dev *dev, struct mlx4_init_hca_param *param)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
__be32 *inbox;
|
|
|
|
|
int err;
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
static const u8 a0_dmfs_hw_steering[] = {
|
|
|
|
|
[MLX4_STEERING_DMFS_A0_DEFAULT] = 0,
|
|
|
|
|
[MLX4_STEERING_DMFS_A0_DYNAMIC] = 1,
|
|
|
|
|
[MLX4_STEERING_DMFS_A0_STATIC] = 2,
|
|
|
|
|
[MLX4_STEERING_DMFS_A0_DISABLE] = 3
|
|
|
|
|
};
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
#define INIT_HCA_IN_SIZE 0x200
|
|
|
|
|
#define INIT_HCA_VERSION_OFFSET 0x000
|
|
|
|
|
#define INIT_HCA_VERSION 2
|
2013-12-23 21:09:43 +07:00
|
|
|
#define INIT_HCA_VXLAN_OFFSET 0x0c
|
2009-09-25 01:03:03 +07:00
|
|
|
#define INIT_HCA_CACHELINE_SZ_OFFSET 0x0e
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_FLAGS_OFFSET 0x014
|
2015-01-27 20:57:59 +07:00
|
|
|
#define INIT_HCA_RECOVERABLE_ERROR_EVENT_OFFSET 0x018
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_QPC_OFFSET 0x020
|
|
|
|
|
#define INIT_HCA_QPC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x10)
|
|
|
|
|
#define INIT_HCA_LOG_QP_OFFSET (INIT_HCA_QPC_OFFSET + 0x17)
|
|
|
|
|
#define INIT_HCA_SRQC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x28)
|
|
|
|
|
#define INIT_HCA_LOG_SRQ_OFFSET (INIT_HCA_QPC_OFFSET + 0x2f)
|
|
|
|
|
#define INIT_HCA_CQC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x30)
|
|
|
|
|
#define INIT_HCA_LOG_CQ_OFFSET (INIT_HCA_QPC_OFFSET + 0x37)
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
#define INIT_HCA_EQE_CQE_OFFSETS (INIT_HCA_QPC_OFFSET + 0x38)
|
2014-09-18 15:50:59 +07:00
|
|
|
#define INIT_HCA_EQE_CQE_STRIDE_OFFSET (INIT_HCA_QPC_OFFSET + 0x3b)
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_ALTC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x40)
|
|
|
|
|
#define INIT_HCA_AUXC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x50)
|
|
|
|
|
#define INIT_HCA_EQC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x60)
|
|
|
|
|
#define INIT_HCA_LOG_EQ_OFFSET (INIT_HCA_QPC_OFFSET + 0x67)
|
2014-11-13 19:45:32 +07:00
|
|
|
#define INIT_HCA_NUM_SYS_EQS_OFFSET (INIT_HCA_QPC_OFFSET + 0x6a)
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_RDMARC_BASE_OFFSET (INIT_HCA_QPC_OFFSET + 0x70)
|
|
|
|
|
#define INIT_HCA_LOG_RD_OFFSET (INIT_HCA_QPC_OFFSET + 0x77)
|
|
|
|
|
#define INIT_HCA_MCAST_OFFSET 0x0c0
|
|
|
|
|
#define INIT_HCA_MC_BASE_OFFSET (INIT_HCA_MCAST_OFFSET + 0x00)
|
|
|
|
|
#define INIT_HCA_LOG_MC_ENTRY_SZ_OFFSET (INIT_HCA_MCAST_OFFSET + 0x12)
|
|
|
|
|
#define INIT_HCA_LOG_MC_HASH_SZ_OFFSET (INIT_HCA_MCAST_OFFSET + 0x16)
|
2011-03-23 05:38:31 +07:00
|
|
|
#define INIT_HCA_UC_STEERING_OFFSET (INIT_HCA_MCAST_OFFSET + 0x18)
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_LOG_MC_TABLE_SZ_OFFSET (INIT_HCA_MCAST_OFFSET + 0x1b)
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
#define INIT_HCA_DEVICE_MANAGED_FLOW_STEERING_EN 0x6
|
|
|
|
|
#define INIT_HCA_FS_PARAM_OFFSET 0x1d0
|
|
|
|
|
#define INIT_HCA_FS_BASE_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x00)
|
|
|
|
|
#define INIT_HCA_FS_LOG_ENTRY_SZ_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x12)
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
#define INIT_HCA_FS_A0_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x18)
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
#define INIT_HCA_FS_LOG_TABLE_SZ_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x1b)
|
|
|
|
|
#define INIT_HCA_FS_ETH_BITS_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x21)
|
|
|
|
|
#define INIT_HCA_FS_ETH_NUM_ADDRS_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x22)
|
|
|
|
|
#define INIT_HCA_FS_IB_BITS_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x25)
|
|
|
|
|
#define INIT_HCA_FS_IB_NUM_ADDRS_OFFSET (INIT_HCA_FS_PARAM_OFFSET + 0x26)
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_TPT_OFFSET 0x0f0
|
|
|
|
|
#define INIT_HCA_DMPT_BASE_OFFSET (INIT_HCA_TPT_OFFSET + 0x00)
|
2013-02-06 23:19:11 +07:00
|
|
|
#define INIT_HCA_TPT_MW_OFFSET (INIT_HCA_TPT_OFFSET + 0x08)
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_HCA_LOG_MPT_SZ_OFFSET (INIT_HCA_TPT_OFFSET + 0x0b)
|
|
|
|
|
#define INIT_HCA_MTT_BASE_OFFSET (INIT_HCA_TPT_OFFSET + 0x10)
|
|
|
|
|
#define INIT_HCA_CMPT_BASE_OFFSET (INIT_HCA_TPT_OFFSET + 0x18)
|
|
|
|
|
#define INIT_HCA_UAR_OFFSET 0x120
|
|
|
|
|
#define INIT_HCA_LOG_UAR_SZ_OFFSET (INIT_HCA_UAR_OFFSET + 0x0a)
|
|
|
|
|
#define INIT_HCA_UAR_PAGE_SZ_OFFSET (INIT_HCA_UAR_OFFSET + 0x0b)
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
inbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
*((u8 *) mailbox->buf + INIT_HCA_VERSION_OFFSET) = INIT_HCA_VERSION;
|
|
|
|
|
|
2009-09-25 01:03:03 +07:00
|
|
|
*((u8 *) mailbox->buf + INIT_HCA_CACHELINE_SZ_OFFSET) =
|
|
|
|
|
(ilog2(cache_line_size()) - 4) << 5;
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
#if defined(__LITTLE_ENDIAN)
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) &= ~cpu_to_be32(1 << 1);
|
|
|
|
|
#elif defined(__BIG_ENDIAN)
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |= cpu_to_be32(1 << 1);
|
|
|
|
|
#else
|
|
|
|
|
#error Host endianness not defined
|
|
|
|
|
#endif
|
|
|
|
|
/* Check port for UD address vector: */
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |= cpu_to_be32(1);
|
|
|
|
|
|
2008-04-17 11:01:10 +07:00
|
|
|
/* Enable IPoIB checksumming if we can: */
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_IPOIB_CSUM)
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |= cpu_to_be32(1 << 3);
|
|
|
|
|
|
2008-07-23 04:19:37 +07:00
|
|
|
/* Enable QoS support if module parameter set */
|
2015-04-02 20:31:18 +07:00
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_ETS_CFG && enable_qos)
|
2008-07-23 04:19:37 +07:00
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |= cpu_to_be32(1 << 2);
|
|
|
|
|
|
2011-06-15 21:47:14 +07:00
|
|
|
/* enable counters */
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_COUNTERS)
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |= cpu_to_be32(1 << 4);
|
|
|
|
|
|
2015-04-02 20:31:06 +07:00
|
|
|
/* Enable RSS spread to fragmented IP packets when supported */
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_RSS_IP_FRAG)
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |= cpu_to_be32(1 << 13);
|
|
|
|
|
|
2012-10-21 21:59:24 +07:00
|
|
|
/* CX3 is capable of extending CQEs/EQEs from 32 to 64 bytes */
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_64B_EQE) {
|
|
|
|
|
*(inbox + INIT_HCA_EQE_CQE_OFFSETS / 4) |= cpu_to_be32(1 << 29);
|
|
|
|
|
dev->caps.eqe_size = 64;
|
|
|
|
|
dev->caps.eqe_factor = 1;
|
|
|
|
|
} else {
|
|
|
|
|
dev->caps.eqe_size = 32;
|
|
|
|
|
dev->caps.eqe_factor = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (dev->caps.flags & MLX4_DEV_CAP_FLAG_64B_CQE) {
|
|
|
|
|
*(inbox + INIT_HCA_EQE_CQE_OFFSETS / 4) |= cpu_to_be32(1 << 30);
|
|
|
|
|
dev->caps.cqe_size = 64;
|
2014-09-18 15:50:59 +07:00
|
|
|
dev->caps.userspace_caps |= MLX4_USER_DEV_CAP_LARGE_CQE;
|
2012-10-21 21:59:24 +07:00
|
|
|
} else {
|
|
|
|
|
dev->caps.cqe_size = 32;
|
|
|
|
|
}
|
|
|
|
|
|
2014-09-18 15:50:59 +07:00
|
|
|
/* CX3 is capable of extending CQEs\EQEs to strides larger than 64B */
|
|
|
|
|
if ((dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_EQE_STRIDE) &&
|
|
|
|
|
(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_CQE_STRIDE)) {
|
|
|
|
|
dev->caps.eqe_size = cache_line_size();
|
|
|
|
|
dev->caps.cqe_size = cache_line_size();
|
|
|
|
|
dev->caps.eqe_factor = 0;
|
|
|
|
|
MLX4_PUT(inbox, (u8)((ilog2(dev->caps.eqe_size) - 5) << 4 |
|
|
|
|
|
(ilog2(dev->caps.eqe_size) - 5)),
|
|
|
|
|
INIT_HCA_EQE_CQE_STRIDE_OFFSET);
|
|
|
|
|
|
|
|
|
|
/* User still need to know to support CQE > 32B */
|
|
|
|
|
dev->caps.userspace_caps |= MLX4_USER_DEV_CAP_LARGE_CQE;
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-27 20:57:59 +07:00
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_RECOVERABLE_ERROR_EVENT)
|
|
|
|
|
*(inbox + INIT_HCA_RECOVERABLE_ERROR_EVENT_OFFSET / 4) |= cpu_to_be32(1 << 31);
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
/* QPC/EEC/CQC/EQC/RDMARC attributes */
|
|
|
|
|
|
|
|
|
|
MLX4_PUT(inbox, param->qpc_base, INIT_HCA_QPC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_num_qps, INIT_HCA_LOG_QP_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->srqc_base, INIT_HCA_SRQC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_num_srqs, INIT_HCA_LOG_SRQ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->cqc_base, INIT_HCA_CQC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_num_cqs, INIT_HCA_LOG_CQ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->altc_base, INIT_HCA_ALTC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->auxc_base, INIT_HCA_AUXC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->eqc_base, INIT_HCA_EQC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_num_eqs, INIT_HCA_LOG_EQ_OFFSET);
|
2014-11-13 19:45:32 +07:00
|
|
|
MLX4_PUT(inbox, param->num_sys_eqs, INIT_HCA_NUM_SYS_EQS_OFFSET);
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_PUT(inbox, param->rdmarc_base, INIT_HCA_RDMARC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_rd_per_qp, INIT_HCA_LOG_RD_OFFSET);
|
|
|
|
|
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
/* steering attributes */
|
|
|
|
|
if (dev->caps.steering_mode ==
|
|
|
|
|
MLX4_STEERING_MODE_DEVICE_MANAGED) {
|
|
|
|
|
*(inbox + INIT_HCA_FLAGS_OFFSET / 4) |=
|
|
|
|
|
cpu_to_be32(1 <<
|
|
|
|
|
INIT_HCA_DEVICE_MANAGED_FLOW_STEERING_EN);
|
|
|
|
|
|
|
|
|
|
MLX4_PUT(inbox, param->mc_base, INIT_HCA_FS_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_mc_entry_sz,
|
|
|
|
|
INIT_HCA_FS_LOG_ENTRY_SZ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_mc_table_sz,
|
|
|
|
|
INIT_HCA_FS_LOG_TABLE_SZ_OFFSET);
|
|
|
|
|
/* Enable Ethernet flow steering
|
|
|
|
|
* with udp unicast and tcp unicast
|
|
|
|
|
*/
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
if (dev->caps.dmfs_high_steer_mode !=
|
|
|
|
|
MLX4_STEERING_DMFS_A0_STATIC)
|
|
|
|
|
MLX4_PUT(inbox,
|
|
|
|
|
(u8)(MLX4_FS_UDP_UC_EN | MLX4_FS_TCP_UC_EN),
|
|
|
|
|
INIT_HCA_FS_ETH_BITS_OFFSET);
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
MLX4_PUT(inbox, (u16) MLX4_FS_NUM_OF_L2_ADDR,
|
|
|
|
|
INIT_HCA_FS_ETH_NUM_ADDRS_OFFSET);
|
|
|
|
|
/* Enable IPoIB flow steering
|
|
|
|
|
* with udp unicast and tcp unicast
|
|
|
|
|
*/
|
2013-01-31 06:07:09 +07:00
|
|
|
MLX4_PUT(inbox, (u8) (MLX4_FS_UDP_UC_EN | MLX4_FS_TCP_UC_EN),
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
INIT_HCA_FS_IB_BITS_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, (u16) MLX4_FS_NUM_OF_L2_ADDR,
|
|
|
|
|
INIT_HCA_FS_IB_NUM_ADDRS_OFFSET);
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
|
|
|
|
|
if (dev->caps.dmfs_high_steer_mode !=
|
|
|
|
|
MLX4_STEERING_DMFS_A0_NOT_SUPPORTED)
|
|
|
|
|
MLX4_PUT(inbox,
|
|
|
|
|
((u8)(a0_dmfs_hw_steering[dev->caps.dmfs_high_steer_mode]
|
|
|
|
|
<< 6)),
|
|
|
|
|
INIT_HCA_FS_A0_OFFSET);
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
} else {
|
|
|
|
|
MLX4_PUT(inbox, param->mc_base, INIT_HCA_MC_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_mc_entry_sz,
|
|
|
|
|
INIT_HCA_LOG_MC_ENTRY_SZ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_mc_hash_sz,
|
|
|
|
|
INIT_HCA_LOG_MC_HASH_SZ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->log_mc_table_sz,
|
|
|
|
|
INIT_HCA_LOG_MC_TABLE_SZ_OFFSET);
|
|
|
|
|
if (dev->caps.steering_mode == MLX4_STEERING_MODE_B0)
|
|
|
|
|
MLX4_PUT(inbox, (u8) (1 << 3),
|
|
|
|
|
INIT_HCA_UC_STEERING_OFFSET);
|
|
|
|
|
}
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
/* TPT attributes */
|
|
|
|
|
|
|
|
|
|
MLX4_PUT(inbox, param->dmpt_base, INIT_HCA_DMPT_BASE_OFFSET);
|
2013-02-06 23:19:11 +07:00
|
|
|
MLX4_PUT(inbox, param->mw_enabled, INIT_HCA_TPT_MW_OFFSET);
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_PUT(inbox, param->log_mpt_sz, INIT_HCA_LOG_MPT_SZ_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->mtt_base, INIT_HCA_MTT_BASE_OFFSET);
|
|
|
|
|
MLX4_PUT(inbox, param->cmpt_base, INIT_HCA_CMPT_BASE_OFFSET);
|
|
|
|
|
|
|
|
|
|
/* UAR attributes */
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
MLX4_PUT(inbox, param->uar_page_sz, INIT_HCA_UAR_PAGE_SZ_OFFSET);
|
2007-05-09 08:00:38 +07:00
|
|
|
MLX4_PUT(inbox, param->log_uar_sz, INIT_HCA_LOG_UAR_SZ_OFFSET);
|
|
|
|
|
|
2013-12-23 21:09:43 +07:00
|
|
|
/* set parser VXLAN attributes */
|
|
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS) {
|
|
|
|
|
u8 parser_params = 0;
|
|
|
|
|
MLX4_PUT(inbox, parser_params, INIT_HCA_VXLAN_OFFSET);
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-27 20:58:02 +07:00
|
|
|
err = mlx4_cmd(dev, mailbox->dma, 0, 0, MLX4_CMD_INIT_HCA,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
|
|
|
|
if (err)
|
|
|
|
|
mlx4_err(dev, "INIT_HCA returns %d\n", err);
|
|
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
int mlx4_QUERY_HCA(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_init_hca_param *param)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
__be32 *outbox;
|
2012-12-07 00:11:59 +07:00
|
|
|
u32 dword_field;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
int err;
|
2012-10-21 21:59:24 +07:00
|
|
|
u8 byte_field;
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
static const u8 a0_dmfs_query_hw_steering[] = {
|
|
|
|
|
[0] = MLX4_STEERING_DMFS_A0_DEFAULT,
|
|
|
|
|
[1] = MLX4_STEERING_DMFS_A0_DYNAMIC,
|
|
|
|
|
[2] = MLX4_STEERING_DMFS_A0_STATIC,
|
|
|
|
|
[3] = MLX4_STEERING_DMFS_A0_DISABLE
|
|
|
|
|
};
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
|
|
|
|
|
#define QUERY_HCA_GLOBAL_CAPS_OFFSET 0x04
|
2013-04-23 13:06:48 +07:00
|
|
|
#define QUERY_HCA_CORE_CLOCK_OFFSET 0x0c
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0,
|
|
|
|
|
MLX4_CMD_QUERY_HCA,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B,
|
|
|
|
|
!mlx4_is_slave(dev));
|
|
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(param->global_caps, outbox, QUERY_HCA_GLOBAL_CAPS_OFFSET);
|
2013-04-23 13:06:48 +07:00
|
|
|
MLX4_GET(param->hca_core_clock, outbox, QUERY_HCA_CORE_CLOCK_OFFSET);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
|
|
|
|
|
/* QPC/EEC/CQC/EQC/RDMARC attributes */
|
|
|
|
|
|
|
|
|
|
MLX4_GET(param->qpc_base, outbox, INIT_HCA_QPC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_num_qps, outbox, INIT_HCA_LOG_QP_OFFSET);
|
|
|
|
|
MLX4_GET(param->srqc_base, outbox, INIT_HCA_SRQC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_num_srqs, outbox, INIT_HCA_LOG_SRQ_OFFSET);
|
|
|
|
|
MLX4_GET(param->cqc_base, outbox, INIT_HCA_CQC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_num_cqs, outbox, INIT_HCA_LOG_CQ_OFFSET);
|
|
|
|
|
MLX4_GET(param->altc_base, outbox, INIT_HCA_ALTC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->auxc_base, outbox, INIT_HCA_AUXC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->eqc_base, outbox, INIT_HCA_EQC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_num_eqs, outbox, INIT_HCA_LOG_EQ_OFFSET);
|
2014-11-13 19:45:32 +07:00
|
|
|
MLX4_GET(param->num_sys_eqs, outbox, INIT_HCA_NUM_SYS_EQS_OFFSET);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
MLX4_GET(param->rdmarc_base, outbox, INIT_HCA_RDMARC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_rd_per_qp, outbox, INIT_HCA_LOG_RD_OFFSET);
|
|
|
|
|
|
2012-12-07 00:11:59 +07:00
|
|
|
MLX4_GET(dword_field, outbox, INIT_HCA_FLAGS_OFFSET);
|
|
|
|
|
if (dword_field & (1 << INIT_HCA_DEVICE_MANAGED_FLOW_STEERING_EN)) {
|
|
|
|
|
param->steering_mode = MLX4_STEERING_MODE_DEVICE_MANAGED;
|
|
|
|
|
} else {
|
|
|
|
|
MLX4_GET(byte_field, outbox, INIT_HCA_UC_STEERING_OFFSET);
|
|
|
|
|
if (byte_field & 0x8)
|
|
|
|
|
param->steering_mode = MLX4_STEERING_MODE_B0;
|
|
|
|
|
else
|
|
|
|
|
param->steering_mode = MLX4_STEERING_MODE_A0;
|
|
|
|
|
}
|
2015-04-02 20:31:06 +07:00
|
|
|
|
|
|
|
|
if (dword_field & (1 << 13))
|
|
|
|
|
param->rss_ip_frags = 1;
|
|
|
|
|
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
/* steering attributes */
|
2012-12-07 00:11:59 +07:00
|
|
|
if (param->steering_mode == MLX4_STEERING_MODE_DEVICE_MANAGED) {
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
MLX4_GET(param->mc_base, outbox, INIT_HCA_FS_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_mc_entry_sz, outbox,
|
|
|
|
|
INIT_HCA_FS_LOG_ENTRY_SZ_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_mc_table_sz, outbox,
|
|
|
|
|
INIT_HCA_FS_LOG_TABLE_SZ_OFFSET);
|
net/mlx4: Add support for A0 steering
Add the required firmware commands for A0 steering and a way to enable
that. The firmware support focuses on INIT_HCA, QUERY_HCA, QUERY_PORT,
QUERY_DEV_CAP and QUERY_FUNC_CAP commands. Those commands are used
to configure and query the device.
The different A0 DMFS (steering) modes are:
Static - optimized performance, but flow steering rules are
limited. This mode should be choosed explicitly by the user
in order to be used.
Dynamic - this mode should be explicitly choosed by the user.
In this mode, the FW works in optimized steering mode as long as
it can and afterwards automatically drops to classic (full) DMFS.
Disable - this mode should be explicitly choosed by the user.
The user instructs the system not to use optimized steering, even if
the FW supports Dynamic A0 DMFS (and thus will be able to use optimized
steering in Default A0 DMFS mode).
Default - this mode is implicitly choosed. In this mode, if the FW
supports Dynamic A0 DMFS, it'll work in this mode. Otherwise, it'll
work at Disable A0 DMFS mode.
Under SRIOV configuration, when the A0 steering mode is enabled,
older guest VF drivers who aren't using the RX QP allocation flag
(MLX4_RESERVE_A0_QP) will get a QP from the general range and
fail when attempting to register a steering rule. To avoid that,
the PF context behaviour is changed once on A0 static mode, to
require support for the allocation flag in VF drivers too.
In order to enable A0 steering, we use log_num_mgm_entry_size param.
If the value of the parameter is not positive, we treat the absolute
value of log_num_mgm_entry_size as a bit field. Setting bit 2 of this
bit field enables static A0 steering.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-12-11 15:58:00 +07:00
|
|
|
MLX4_GET(byte_field, outbox,
|
|
|
|
|
INIT_HCA_FS_A0_OFFSET);
|
|
|
|
|
param->dmfs_high_steer_mode =
|
|
|
|
|
a0_dmfs_query_hw_steering[(byte_field >> 6) & 3];
|
{NET, IB}/mlx4: Add device managed flow steering firmware API
The driver is modified to support three operation modes.
If supported by firmware use the device managed flow steering
API, that which we call device managed steering mode. Else, if
the firmware supports the B0 steering mode use it, and finally,
if none of the above, use the A0 steering mode.
When the steering mode is device managed, the code is modified
such that L2 based rules set by the mlx4_en driver for Ethernet
unicast and multicast, and the IB stack multicast attach calls
done through the mlx4_ib driver are all routed to use the device
managed API.
When attaching rule using device managed flow steering API,
the firmware returns a 64 bit registration id, which is to be
provided during detach.
Currently the firmware is always programmed during HCA initialization
to use standard L2 hashing. Future work should be done to allow
configuring the flow-steering hash function with common, non
proprietary means.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-07-05 11:03:46 +07:00
|
|
|
} else {
|
|
|
|
|
MLX4_GET(param->mc_base, outbox, INIT_HCA_MC_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_mc_entry_sz, outbox,
|
|
|
|
|
INIT_HCA_LOG_MC_ENTRY_SZ_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_mc_hash_sz, outbox,
|
|
|
|
|
INIT_HCA_LOG_MC_HASH_SZ_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_mc_table_sz, outbox,
|
|
|
|
|
INIT_HCA_LOG_MC_TABLE_SZ_OFFSET);
|
|
|
|
|
}
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
|
2012-10-21 21:59:24 +07:00
|
|
|
/* CX3 is capable of extending CQEs/EQEs from 32 to 64 bytes */
|
|
|
|
|
MLX4_GET(byte_field, outbox, INIT_HCA_EQE_CQE_OFFSETS);
|
|
|
|
|
if (byte_field & 0x20) /* 64-bytes eqe enabled */
|
|
|
|
|
param->dev_cap_enabled |= MLX4_DEV_CAP_64B_EQE_ENABLED;
|
|
|
|
|
if (byte_field & 0x40) /* 64-bytes cqe enabled */
|
|
|
|
|
param->dev_cap_enabled |= MLX4_DEV_CAP_64B_CQE_ENABLED;
|
|
|
|
|
|
2014-09-18 15:50:59 +07:00
|
|
|
/* CX3 is capable of extending CQEs\EQEs to strides larger than 64B */
|
|
|
|
|
MLX4_GET(byte_field, outbox, INIT_HCA_EQE_CQE_STRIDE_OFFSET);
|
|
|
|
|
if (byte_field) {
|
2014-12-16 18:28:54 +07:00
|
|
|
param->dev_cap_enabled |= MLX4_DEV_CAP_EQE_STRIDE_ENABLED;
|
|
|
|
|
param->dev_cap_enabled |= MLX4_DEV_CAP_CQE_STRIDE_ENABLED;
|
2014-09-18 15:50:59 +07:00
|
|
|
param->cqe_size = 1 << ((byte_field &
|
|
|
|
|
MLX4_CQE_SIZE_MASK_STRIDE) + 5);
|
|
|
|
|
param->eqe_size = 1 << (((byte_field &
|
|
|
|
|
MLX4_EQE_SIZE_MASK_STRIDE) >> 4) + 5);
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
/* TPT attributes */
|
|
|
|
|
|
|
|
|
|
MLX4_GET(param->dmpt_base, outbox, INIT_HCA_DMPT_BASE_OFFSET);
|
2013-02-06 23:19:11 +07:00
|
|
|
MLX4_GET(param->mw_enabled, outbox, INIT_HCA_TPT_MW_OFFSET);
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
MLX4_GET(param->log_mpt_sz, outbox, INIT_HCA_LOG_MPT_SZ_OFFSET);
|
|
|
|
|
MLX4_GET(param->mtt_base, outbox, INIT_HCA_MTT_BASE_OFFSET);
|
|
|
|
|
MLX4_GET(param->cmpt_base, outbox, INIT_HCA_CMPT_BASE_OFFSET);
|
|
|
|
|
|
|
|
|
|
/* UAR attributes */
|
|
|
|
|
|
|
|
|
|
MLX4_GET(param->uar_page_sz, outbox, INIT_HCA_UAR_PAGE_SZ_OFFSET);
|
|
|
|
|
MLX4_GET(param->log_uar_sz, outbox, INIT_HCA_LOG_UAR_SZ_OFFSET);
|
|
|
|
|
|
2015-07-27 18:46:31 +07:00
|
|
|
/* phv_check enable */
|
|
|
|
|
MLX4_GET(byte_field, outbox, INIT_HCA_CACHELINE_SZ_OFFSET);
|
|
|
|
|
if (byte_field & 0x2)
|
|
|
|
|
param->phv_check_en = 1;
|
mlx4_core: Modify driver initialization flow to accommodate SRIOV for Ethernet
1. Added module parameters sr_iov and probe_vf for controlling enablement of
SRIOV mode.
2. Increased default max num-qps, num-mpts and log_num_macs to accomodate
SRIOV mode
3. Added port_type_array as a module parameter to allow driver startup with
ports configured as desired.
In SRIOV mode, only ETH is supported, and this array is ignored; otherwise,
for the case where the FW supports both port types (ETH and IB), the
port_type_array parameter is used.
By default, the port_type_array is set to configure both ports as IB.
4. When running in sriov mode, the master needs to initialize the ICM eq table
to hold the eq's for itself and also for all the slaves.
5. mlx4_set_port_mask() now invoked from mlx4_init_hca, instead of in mlx4_dev_cap.
6. Introduced sriov VF (slave) device startup/teardown logic (mainly procedures
mlx4_init_slave, mlx4_slave_exit, mlx4_slave_cap, mlx4_slave_exit and flow
modifications in __mlx4_init_one, mlx4_init_hca, and mlx4_setup_hca).
VFs obtain their startup information from the PF (master) device via the
comm channel.
7. In SRIOV mode (both PF and VF), MSI_X must be enabled, or the driver
aborts loading the device.
8. Do not allow setting port type via sysfs when running in SRIOV mode.
9. mlx4_get_ownership: Currently, only one PF is supported by the driver.
If the HCA is burned with FW which enables more than one PF, only one
of the PFs is allowed to run. The first one up grabs a FW ownership
semaphone -- all other PFs will find that semaphore taken, and the
driver will not allow them to run.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Liran Liss <liranl@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:18:30 +07:00
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2015-01-27 20:58:09 +07:00
|
|
|
static int mlx4_hca_core_clock_update(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
__be32 *outbox;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox)) {
|
|
|
|
|
mlx4_warn(dev, "hca_core_clock mailbox allocation failed\n");
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
}
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0,
|
|
|
|
|
MLX4_CMD_QUERY_HCA,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B,
|
|
|
|
|
!mlx4_is_slave(dev));
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_warn(dev, "hca_core_clock update failed\n");
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
MLX4_GET(dev->caps.hca_core_clock, outbox, QUERY_HCA_CORE_CLOCK_OFFSET);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2012-08-03 15:40:53 +07:00
|
|
|
/* for IB-type ports only in SRIOV mode. Checks that both proxy QP0
|
|
|
|
|
* and real QP0 are active, so that the paravirtualized QP0 is ready
|
|
|
|
|
* to operate */
|
|
|
|
|
static int check_qp0_state(struct mlx4_dev *dev, int function, int port)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
/* irrelevant if not infiniband */
|
|
|
|
|
if (priv->mfunc.master.qp0_state[port].proxy_qp0_active &&
|
|
|
|
|
priv->mfunc.master.qp0_state[port].qp0_active)
|
|
|
|
|
return 1;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int mlx4_INIT_PORT_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2014-03-19 23:11:52 +07:00
|
|
|
int port = mlx4_slave_convert_port(dev, slave, vhcr->in_modifier);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int err;
|
|
|
|
|
|
2014-03-19 23:11:52 +07:00
|
|
|
if (port < 0)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
if (priv->mfunc.master.slave_state[slave].init_port_mask & (1 << port))
|
|
|
|
|
return 0;
|
|
|
|
|
|
2012-08-03 15:40:53 +07:00
|
|
|
if (dev->caps.port_mask[port] != MLX4_PORT_TYPE_IB) {
|
|
|
|
|
/* Enable port only if it was previously disabled */
|
|
|
|
|
if (!priv->mfunc.master.init_port_ref[port]) {
|
|
|
|
|
err = mlx4_cmd(dev, 0, port, 0, MLX4_CMD_INIT_PORT,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
priv->mfunc.master.slave_state[slave].init_port_mask |= (1 << port);
|
|
|
|
|
} else {
|
|
|
|
|
if (slave == mlx4_master_func_num(dev)) {
|
|
|
|
|
if (check_qp0_state(dev, slave, port) &&
|
|
|
|
|
!priv->mfunc.master.qp0_state[port].port_active) {
|
|
|
|
|
err = mlx4_cmd(dev, 0, port, 0, MLX4_CMD_INIT_PORT,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
priv->mfunc.master.qp0_state[port].port_active = 1;
|
|
|
|
|
priv->mfunc.master.slave_state[slave].init_port_mask |= (1 << port);
|
|
|
|
|
}
|
|
|
|
|
} else
|
|
|
|
|
priv->mfunc.master.slave_state[slave].init_port_mask |= (1 << port);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
}
|
|
|
|
|
++priv->mfunc.master.init_port_ref[port];
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
int mlx4_INIT_PORT(struct mlx4_dev *dev, int port)
|
2007-05-09 08:00:38 +07:00
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 *inbox;
|
|
|
|
|
int err;
|
|
|
|
|
u32 flags;
|
2007-06-18 22:15:02 +07:00
|
|
|
u16 field;
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
if (dev->flags & MLX4_FLAG_OLD_PORT_CMDS) {
|
2007-05-09 08:00:38 +07:00
|
|
|
#define INIT_PORT_IN_SIZE 256
|
|
|
|
|
#define INIT_PORT_FLAGS_OFFSET 0x00
|
|
|
|
|
#define INIT_PORT_FLAG_SIG (1 << 18)
|
|
|
|
|
#define INIT_PORT_FLAG_NG (1 << 17)
|
|
|
|
|
#define INIT_PORT_FLAG_G0 (1 << 16)
|
|
|
|
|
#define INIT_PORT_VL_SHIFT 4
|
|
|
|
|
#define INIT_PORT_PORT_WIDTH_SHIFT 8
|
|
|
|
|
#define INIT_PORT_MTU_OFFSET 0x04
|
|
|
|
|
#define INIT_PORT_MAX_GID_OFFSET 0x06
|
|
|
|
|
#define INIT_PORT_MAX_PKEY_OFFSET 0x0a
|
|
|
|
|
#define INIT_PORT_GUID0_OFFSET 0x10
|
|
|
|
|
#define INIT_PORT_NODE_GUID_OFFSET 0x18
|
|
|
|
|
#define INIT_PORT_SI_GUID_OFFSET 0x20
|
|
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
inbox = mailbox->buf;
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
flags = 0;
|
|
|
|
|
flags |= (dev->caps.vl_cap[port] & 0xf) << INIT_PORT_VL_SHIFT;
|
|
|
|
|
flags |= (dev->caps.port_width_cap[port] & 0xf) << INIT_PORT_PORT_WIDTH_SHIFT;
|
|
|
|
|
MLX4_PUT(inbox, flags, INIT_PORT_FLAGS_OFFSET);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2008-10-23 00:56:48 +07:00
|
|
|
field = 128 << dev->caps.ib_mtu_cap[port];
|
2007-06-18 22:15:02 +07:00
|
|
|
MLX4_PUT(inbox, field, INIT_PORT_MTU_OFFSET);
|
|
|
|
|
field = dev->caps.gid_table_len[port];
|
|
|
|
|
MLX4_PUT(inbox, field, INIT_PORT_MAX_GID_OFFSET);
|
|
|
|
|
field = dev->caps.pkey_table_len[port];
|
|
|
|
|
MLX4_PUT(inbox, field, INIT_PORT_MAX_PKEY_OFFSET);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
err = mlx4_cmd(dev, mailbox->dma, port, 0, MLX4_CMD_INIT_PORT,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2007-06-18 22:15:02 +07:00
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
} else
|
|
|
|
|
err = mlx4_cmd(dev, 0, port, 0, MLX4_CMD_INIT_PORT,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
|
2007-05-09 08:00:38 +07:00
|
|
|
|
2015-01-27 20:58:09 +07:00
|
|
|
if (!err)
|
|
|
|
|
mlx4_hca_core_clock_update(dev);
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_INIT_PORT);
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int mlx4_CLOSE_PORT_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
2014-03-19 23:11:52 +07:00
|
|
|
int port = mlx4_slave_convert_port(dev, slave, vhcr->in_modifier);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
int err;
|
|
|
|
|
|
2014-03-19 23:11:52 +07:00
|
|
|
if (port < 0)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
if (!(priv->mfunc.master.slave_state[slave].init_port_mask &
|
|
|
|
|
(1 << port)))
|
|
|
|
|
return 0;
|
|
|
|
|
|
2012-08-03 15:40:53 +07:00
|
|
|
if (dev->caps.port_mask[port] != MLX4_PORT_TYPE_IB) {
|
|
|
|
|
if (priv->mfunc.master.init_port_ref[port] == 1) {
|
|
|
|
|
err = mlx4_cmd(dev, 0, port, 0, MLX4_CMD_CLOSE_PORT,
|
2015-01-27 20:58:02 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2012-08-03 15:40:53 +07:00
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
priv->mfunc.master.slave_state[slave].init_port_mask &= ~(1 << port);
|
|
|
|
|
} else {
|
|
|
|
|
/* infiniband port */
|
|
|
|
|
if (slave == mlx4_master_func_num(dev)) {
|
|
|
|
|
if (!priv->mfunc.master.qp0_state[port].qp0_active &&
|
|
|
|
|
priv->mfunc.master.qp0_state[port].port_active) {
|
|
|
|
|
err = mlx4_cmd(dev, 0, port, 0, MLX4_CMD_CLOSE_PORT,
|
2015-01-27 20:58:02 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2012-08-03 15:40:53 +07:00
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
priv->mfunc.master.slave_state[slave].init_port_mask &= ~(1 << port);
|
|
|
|
|
priv->mfunc.master.qp0_state[port].port_active = 0;
|
|
|
|
|
}
|
|
|
|
|
} else
|
|
|
|
|
priv->mfunc.master.slave_state[slave].init_port_mask &= ~(1 << port);
|
mlx4_core: Added FW commands and their wrappers for supporting SRIOV
The following commands are added here:
1. QUERY_FUNC_CAP and its wrapper. This function is used by VFs when
they start up to receive configuration information from the PF, such
as resource quotas for this VF, which ports should be used (currently
two), what protocol is running on the port (currently Ethernet ONLY,
or port not active).
2. QUERY_PORT and its wrapper. Previously, this FW command was invoked directly
by the ETH driver (en_port.c) using mlx4_cmd_box. Virtualization is now
required here (the VF's MAC address must be substituted for the PFs
MAC address returned by the FW). We changed the invocation
in the ETH driver to use mlx4_QUERY_PORT, and added the wrapper.
3. QUERY_HCA. Used by the VF to determine how the HCA was initialized.
For now, we need only the multicast table member entry size
(log2_mc_table_entry_sz, in the ConnectX PRM). No wrapper is needed
here, because the data may be passed as is to the VF without modification).
In this command, we have added a GLOBAL_CAPS field for passing required
configuration information from FW to a VF (this field is to allow safely
adding new SRIOV capabilities which require support in VF drivers, too).
Bits will set here by FW in response to PF-driver configuration commands which
will activate as yet undefined new SRIOV features. The VF will test to see that
all required capabilities indicated by this field are supported (i.e., if a bit
is set and the VF driver does not recognize that bit, it must abort
its initialization). Currently, no bits are set.
4. Added a CLOSE_PORT wrapper. The PF context needs to keep track of how many VF contexts
have the port open. The PF context will not actually issue the FW close port command
until the last port user issues a CLOSE_PORT request.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Marcel Apfelbaum <marcela@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:12:40 +07:00
|
|
|
}
|
|
|
|
|
--priv->mfunc.master.init_port_ref[port];
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
int mlx4_CLOSE_PORT(struct mlx4_dev *dev, int port)
|
|
|
|
|
{
|
2015-01-27 20:58:02 +07:00
|
|
|
return mlx4_cmd(dev, 0, port, 0, MLX4_CMD_CLOSE_PORT,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_CLOSE_PORT);
|
|
|
|
|
|
|
|
|
|
int mlx4_CLOSE_HCA(struct mlx4_dev *dev, int panic)
|
|
|
|
|
{
|
2015-01-27 20:58:02 +07:00
|
|
|
return mlx4_cmd(dev, 0, 0, panic, MLX4_CMD_CLOSE_HCA,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
|
|
|
|
|
2014-03-27 19:02:03 +07:00
|
|
|
struct mlx4_config_dev {
|
|
|
|
|
__be32 update_flags;
|
2014-11-02 21:26:17 +07:00
|
|
|
__be32 rsvd1[3];
|
2014-03-27 19:02:03 +07:00
|
|
|
__be16 vxlan_udp_dport;
|
|
|
|
|
__be16 rsvd2;
|
2015-02-03 21:48:32 +07:00
|
|
|
__be32 rsvd3;
|
|
|
|
|
__be32 roce_flags;
|
|
|
|
|
__be32 rsvd4[25];
|
|
|
|
|
__be16 rsvd5;
|
|
|
|
|
u8 rsvd6;
|
2014-11-02 21:26:17 +07:00
|
|
|
u8 rx_checksum_val;
|
2014-03-27 19:02:03 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
#define MLX4_VXLAN_UDP_DPORT (1 << 0)
|
2015-02-03 21:48:32 +07:00
|
|
|
#define MLX4_DISABLE_RX_PORT BIT(18)
|
2014-03-27 19:02:03 +07:00
|
|
|
|
2014-11-02 21:26:17 +07:00
|
|
|
static int mlx4_CONFIG_DEV_set(struct mlx4_dev *dev, struct mlx4_config_dev *config_dev)
|
2014-03-27 19:02:03 +07:00
|
|
|
{
|
|
|
|
|
int err;
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
|
|
|
|
|
memcpy(mailbox->buf, config_dev, sizeof(*config_dev));
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, 0, 0, MLX4_CMD_CONFIG_DEV,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-11-02 21:26:17 +07:00
|
|
|
static int mlx4_CONFIG_DEV_get(struct mlx4_dev *dev, struct mlx4_config_dev *config_dev)
|
|
|
|
|
{
|
|
|
|
|
int err;
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 1, MLX4_CMD_CONFIG_DEV,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
if (!err)
|
|
|
|
|
memcpy(config_dev, mailbox->buf, sizeof(*config_dev));
|
|
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Conversion between the HW values and the actual functionality.
|
|
|
|
|
* The value represented by the array index,
|
|
|
|
|
* and the functionality determined by the flags.
|
|
|
|
|
*/
|
|
|
|
|
static const u8 config_dev_csum_flags[] = {
|
|
|
|
|
[0] = 0,
|
|
|
|
|
[1] = MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP,
|
|
|
|
|
[2] = MLX4_RX_CSUM_MODE_VAL_NON_TCP_UDP |
|
|
|
|
|
MLX4_RX_CSUM_MODE_L4,
|
|
|
|
|
[3] = MLX4_RX_CSUM_MODE_L4 |
|
|
|
|
|
MLX4_RX_CSUM_MODE_IP_OK_IP_NON_TCP_UDP |
|
|
|
|
|
MLX4_RX_CSUM_MODE_MULTI_VLAN
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
int mlx4_config_dev_retrieval(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_config_dev_params *params)
|
|
|
|
|
{
|
2015-02-03 22:57:16 +07:00
|
|
|
struct mlx4_config_dev config_dev = {0};
|
2014-11-02 21:26:17 +07:00
|
|
|
int err;
|
|
|
|
|
u8 csum_mask;
|
|
|
|
|
|
|
|
|
|
#define CONFIG_DEV_RX_CSUM_MODE_MASK 0x7
|
|
|
|
|
#define CONFIG_DEV_RX_CSUM_MODE_PORT1_BIT_OFFSET 0
|
|
|
|
|
#define CONFIG_DEV_RX_CSUM_MODE_PORT2_BIT_OFFSET 4
|
|
|
|
|
|
|
|
|
|
if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_CONFIG_DEV))
|
|
|
|
|
return -ENOTSUPP;
|
|
|
|
|
|
|
|
|
|
err = mlx4_CONFIG_DEV_get(dev, &config_dev);
|
|
|
|
|
if (err)
|
|
|
|
|
return err;
|
|
|
|
|
|
|
|
|
|
csum_mask = (config_dev.rx_checksum_val >> CONFIG_DEV_RX_CSUM_MODE_PORT1_BIT_OFFSET) &
|
|
|
|
|
CONFIG_DEV_RX_CSUM_MODE_MASK;
|
|
|
|
|
|
|
|
|
|
if (csum_mask >= sizeof(config_dev_csum_flags)/sizeof(config_dev_csum_flags[0]))
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
params->rx_csum_flags_port_1 = config_dev_csum_flags[csum_mask];
|
|
|
|
|
|
|
|
|
|
csum_mask = (config_dev.rx_checksum_val >> CONFIG_DEV_RX_CSUM_MODE_PORT2_BIT_OFFSET) &
|
|
|
|
|
CONFIG_DEV_RX_CSUM_MODE_MASK;
|
|
|
|
|
|
|
|
|
|
if (csum_mask >= sizeof(config_dev_csum_flags)/sizeof(config_dev_csum_flags[0]))
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
params->rx_csum_flags_port_2 = config_dev_csum_flags[csum_mask];
|
|
|
|
|
|
|
|
|
|
params->vxlan_udp_dport = be16_to_cpu(config_dev.vxlan_udp_dport);
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_config_dev_retrieval);
|
|
|
|
|
|
2014-03-27 19:02:03 +07:00
|
|
|
int mlx4_config_vxlan_port(struct mlx4_dev *dev, __be16 udp_port)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_config_dev config_dev;
|
|
|
|
|
|
|
|
|
|
memset(&config_dev, 0, sizeof(config_dev));
|
|
|
|
|
config_dev.update_flags = cpu_to_be32(MLX4_VXLAN_UDP_DPORT);
|
|
|
|
|
config_dev.vxlan_udp_dport = udp_port;
|
|
|
|
|
|
2014-11-02 21:26:17 +07:00
|
|
|
return mlx4_CONFIG_DEV_set(dev, &config_dev);
|
2014-03-27 19:02:03 +07:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_config_vxlan_port);
|
|
|
|
|
|
2015-02-03 21:48:32 +07:00
|
|
|
#define CONFIG_DISABLE_RX_PORT BIT(15)
|
|
|
|
|
int mlx4_disable_rx_port_check(struct mlx4_dev *dev, bool dis)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_config_dev config_dev;
|
|
|
|
|
|
|
|
|
|
memset(&config_dev, 0, sizeof(config_dev));
|
|
|
|
|
config_dev.update_flags = cpu_to_be32(MLX4_DISABLE_RX_PORT);
|
|
|
|
|
if (dis)
|
|
|
|
|
config_dev.roce_flags =
|
|
|
|
|
cpu_to_be32(CONFIG_DISABLE_RX_PORT);
|
|
|
|
|
|
|
|
|
|
return mlx4_CONFIG_DEV_set(dev, &config_dev);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_virt2phy_port_map(struct mlx4_dev *dev, u32 port1, u32 port2)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
struct {
|
|
|
|
|
__be32 v_port1;
|
|
|
|
|
__be32 v_port2;
|
|
|
|
|
} *v2p;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
|
|
v2p = mailbox->buf;
|
|
|
|
|
v2p->v_port1 = cpu_to_be32(port1);
|
|
|
|
|
v2p->v_port2 = cpu_to_be32(port2);
|
|
|
|
|
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, 0,
|
|
|
|
|
MLX4_SET_PORT_VIRT2PHY, MLX4_CMD_VIRT_PORT_MAP,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2014-03-27 19:02:03 +07:00
|
|
|
|
2007-05-09 08:00:38 +07:00
|
|
|
int mlx4_SET_ICM_SIZE(struct mlx4_dev *dev, u64 icm_size, u64 *aux_pages)
|
|
|
|
|
{
|
|
|
|
|
int ret = mlx4_cmd_imm(dev, icm_size, aux_pages, 0, 0,
|
|
|
|
|
MLX4_CMD_SET_ICM_SIZE,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
if (ret)
|
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Round up number of system pages needed in case
|
|
|
|
|
* MLX4_ICM_PAGE_SIZE < PAGE_SIZE.
|
|
|
|
|
*/
|
|
|
|
|
*aux_pages = ALIGN(*aux_pages, PAGE_SIZE / MLX4_ICM_PAGE_SIZE) >>
|
|
|
|
|
(PAGE_SHIFT - MLX4_ICM_PAGE_SHIFT);
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_NOP(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
/* Input modifier of 0x1f means "finish as soon as possible." */
|
2015-01-27 20:58:02 +07:00
|
|
|
return mlx4_cmd(dev, 0, 0x1f, 0, MLX4_CMD_NOP, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
2007-05-09 08:00:38 +07:00
|
|
|
}
|
2011-03-23 05:37:59 +07:00
|
|
|
|
2013-12-20 02:20:12 +07:00
|
|
|
int mlx4_get_phys_port_id(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
u8 port;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
u32 in_mod;
|
|
|
|
|
u32 guid_hi, guid_lo;
|
|
|
|
|
int err, ret = 0;
|
|
|
|
|
#define MOD_STAT_CFG_PORT_OFFSET 8
|
|
|
|
|
#define MOD_STAT_CFG_GUID_H 0X14
|
|
|
|
|
#define MOD_STAT_CFG_GUID_L 0X1c
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
for (port = 1; port <= dev->caps.num_ports; port++) {
|
|
|
|
|
in_mod = port << MOD_STAT_CFG_PORT_OFFSET;
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, in_mod, 0x2,
|
|
|
|
|
MLX4_CMD_MOD_STAT_CFG, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Fail to get port %d uplink guid\n",
|
|
|
|
|
port);
|
|
|
|
|
ret = err;
|
|
|
|
|
} else {
|
|
|
|
|
MLX4_GET(guid_hi, outbox, MOD_STAT_CFG_GUID_H);
|
|
|
|
|
MLX4_GET(guid_lo, outbox, MOD_STAT_CFG_GUID_L);
|
|
|
|
|
dev->caps.phys_port_id[port] = (u64)guid_lo |
|
|
|
|
|
(u64)guid_hi << 32;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
2011-03-23 05:37:59 +07:00
|
|
|
#define MLX4_WOL_SETUP_MODE (5 << 28)
|
|
|
|
|
int mlx4_wol_read(struct mlx4_dev *dev, u64 *config, int port)
|
|
|
|
|
{
|
|
|
|
|
u32 in_mod = MLX4_WOL_SETUP_MODE | port << 8;
|
|
|
|
|
|
|
|
|
|
return mlx4_cmd_imm(dev, 0, config, in_mod, 0x3,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_MOD_STAT_CFG, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
2011-03-23 05:37:59 +07:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_wol_read);
|
|
|
|
|
|
|
|
|
|
int mlx4_wol_write(struct mlx4_dev *dev, u64 config, int port)
|
|
|
|
|
{
|
|
|
|
|
u32 in_mod = MLX4_WOL_SETUP_MODE | port << 8;
|
|
|
|
|
|
|
|
|
|
return mlx4_cmd(dev, config, in_mod, 0x1, MLX4_CMD_MOD_STAT_CFG,
|
2011-12-13 11:10:51 +07:00
|
|
|
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
|
2011-03-23 05:37:59 +07:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_wol_write);
|
2013-07-28 22:54:21 +07:00
|
|
|
|
|
|
|
|
enum {
|
|
|
|
|
ADD_TO_MCG = 0x26,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
void mlx4_opreq_action(struct work_struct *work)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_priv *priv = container_of(work, struct mlx4_priv,
|
|
|
|
|
opreq_task);
|
|
|
|
|
struct mlx4_dev *dev = &priv->dev;
|
|
|
|
|
int num_tasks = atomic_read(&priv->opreq_count);
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
struct mlx4_mgm *mgm;
|
|
|
|
|
u32 *outbox;
|
|
|
|
|
u32 modifier;
|
|
|
|
|
u16 token;
|
|
|
|
|
u16 type;
|
|
|
|
|
int err;
|
|
|
|
|
u32 num_qps;
|
|
|
|
|
struct mlx4_qp qp;
|
|
|
|
|
int i;
|
|
|
|
|
u8 rem_mcg;
|
|
|
|
|
u8 prot;
|
|
|
|
|
|
|
|
|
|
#define GET_OP_REQ_MODIFIER_OFFSET 0x08
|
|
|
|
|
#define GET_OP_REQ_TOKEN_OFFSET 0x14
|
|
|
|
|
#define GET_OP_REQ_TYPE_OFFSET 0x1a
|
|
|
|
|
#define GET_OP_REQ_DATA_OFFSET 0x20
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox)) {
|
|
|
|
|
mlx4_err(dev, "Failed to allocate mailbox for GET_OP_REQ\n");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
outbox = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
while (num_tasks) {
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0,
|
|
|
|
|
MLX4_CMD_GET_OP_REQ, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
if (err) {
|
2013-09-30 21:19:09 +07:00
|
|
|
mlx4_err(dev, "Failed to retrieve required operation: %d\n",
|
2013-07-28 22:54:21 +07:00
|
|
|
err);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
MLX4_GET(modifier, outbox, GET_OP_REQ_MODIFIER_OFFSET);
|
|
|
|
|
MLX4_GET(token, outbox, GET_OP_REQ_TOKEN_OFFSET);
|
|
|
|
|
MLX4_GET(type, outbox, GET_OP_REQ_TYPE_OFFSET);
|
|
|
|
|
type &= 0xfff;
|
|
|
|
|
|
|
|
|
|
switch (type) {
|
|
|
|
|
case ADD_TO_MCG:
|
|
|
|
|
if (dev->caps.steering_mode ==
|
|
|
|
|
MLX4_STEERING_MODE_DEVICE_MANAGED) {
|
|
|
|
|
mlx4_warn(dev, "ADD MCG operation is not supported in DEVICE_MANAGED steering mode\n");
|
|
|
|
|
err = EPERM;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
mgm = (struct mlx4_mgm *)((u8 *)(outbox) +
|
|
|
|
|
GET_OP_REQ_DATA_OFFSET);
|
|
|
|
|
num_qps = be32_to_cpu(mgm->members_count) &
|
|
|
|
|
MGM_QPN_MASK;
|
|
|
|
|
rem_mcg = ((u8 *)(&mgm->members_count))[0] & 1;
|
|
|
|
|
prot = ((u8 *)(&mgm->members_count))[0] >> 6;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < num_qps; i++) {
|
|
|
|
|
qp.qpn = be32_to_cpu(mgm->qp[i]);
|
|
|
|
|
if (rem_mcg)
|
|
|
|
|
err = mlx4_multicast_detach(dev, &qp,
|
|
|
|
|
mgm->gid,
|
|
|
|
|
prot, 0);
|
|
|
|
|
else
|
|
|
|
|
err = mlx4_multicast_attach(dev, &qp,
|
|
|
|
|
mgm->gid,
|
|
|
|
|
mgm->gid[5]
|
|
|
|
|
, 0, prot,
|
|
|
|
|
NULL);
|
|
|
|
|
if (err)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
mlx4_warn(dev, "Bad type for required operation\n");
|
|
|
|
|
err = EINVAL;
|
|
|
|
|
break;
|
|
|
|
|
}
|
2014-03-02 15:25:03 +07:00
|
|
|
err = mlx4_cmd(dev, 0, ((u32) err |
|
|
|
|
|
(__force u32)cpu_to_be32(token) << 16),
|
2013-07-28 22:54:21 +07:00
|
|
|
1, MLX4_CMD_GET_OP_REQ, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_err(dev, "Failed to acknowledge required request: %d\n",
|
|
|
|
|
err);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
memset(outbox, 0, 0xffc);
|
|
|
|
|
num_tasks = atomic_dec_return(&priv->opreq_count);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
}
|
2014-06-01 15:53:50 +07:00
|
|
|
|
|
|
|
|
static int mlx4_check_smp_firewall_active(struct mlx4_dev *dev,
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox)
|
|
|
|
|
{
|
|
|
|
|
#define MLX4_CMD_MAD_DEMUX_SET_ATTR_OFFSET 0x10
|
|
|
|
|
#define MLX4_CMD_MAD_DEMUX_GETRESP_ATTR_OFFSET 0x20
|
|
|
|
|
#define MLX4_CMD_MAD_DEMUX_TRAP_ATTR_OFFSET 0x40
|
|
|
|
|
#define MLX4_CMD_MAD_DEMUX_TRAP_REPRESS_ATTR_OFFSET 0x70
|
|
|
|
|
|
|
|
|
|
u32 set_attr_mask, getresp_attr_mask;
|
|
|
|
|
u32 trap_attr_mask, traprepress_attr_mask;
|
|
|
|
|
|
|
|
|
|
MLX4_GET(set_attr_mask, mailbox->buf,
|
|
|
|
|
MLX4_CMD_MAD_DEMUX_SET_ATTR_OFFSET);
|
|
|
|
|
mlx4_dbg(dev, "SMP firewall set_attribute_mask = 0x%x\n",
|
|
|
|
|
set_attr_mask);
|
|
|
|
|
|
|
|
|
|
MLX4_GET(getresp_attr_mask, mailbox->buf,
|
|
|
|
|
MLX4_CMD_MAD_DEMUX_GETRESP_ATTR_OFFSET);
|
|
|
|
|
mlx4_dbg(dev, "SMP firewall getresp_attribute_mask = 0x%x\n",
|
|
|
|
|
getresp_attr_mask);
|
|
|
|
|
|
|
|
|
|
MLX4_GET(trap_attr_mask, mailbox->buf,
|
|
|
|
|
MLX4_CMD_MAD_DEMUX_TRAP_ATTR_OFFSET);
|
|
|
|
|
mlx4_dbg(dev, "SMP firewall trap_attribute_mask = 0x%x\n",
|
|
|
|
|
trap_attr_mask);
|
|
|
|
|
|
|
|
|
|
MLX4_GET(traprepress_attr_mask, mailbox->buf,
|
|
|
|
|
MLX4_CMD_MAD_DEMUX_TRAP_REPRESS_ATTR_OFFSET);
|
|
|
|
|
mlx4_dbg(dev, "SMP firewall traprepress_attribute_mask = 0x%x\n",
|
|
|
|
|
traprepress_attr_mask);
|
|
|
|
|
|
|
|
|
|
if (set_attr_mask && getresp_attr_mask && trap_attr_mask &&
|
|
|
|
|
traprepress_attr_mask)
|
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int mlx4_config_mad_demux(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
int secure_host_active;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
/* Check if mad_demux is supported */
|
|
|
|
|
if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_MAD_DEMUX))
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox)) {
|
|
|
|
|
mlx4_warn(dev, "Failed to allocate mailbox for cmd MAD_DEMUX");
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Query mad_demux to find out which MADs are handled by internal sma */
|
|
|
|
|
err = mlx4_cmd_box(dev, 0, mailbox->dma, 0x01 /* subn mgmt class */,
|
|
|
|
|
MLX4_CMD_MAD_DEMUX_QUERY_RESTR, MLX4_CMD_MAD_DEMUX,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_warn(dev, "MLX4_CMD_MAD_DEMUX: query restrictions failed (%d)\n",
|
|
|
|
|
err);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
secure_host_active = mlx4_check_smp_firewall_active(dev, mailbox);
|
|
|
|
|
|
|
|
|
|
/* Config mad_demux to handle all MADs returned by the query above */
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, 0x01 /* subn mgmt class */,
|
|
|
|
|
MLX4_CMD_MAD_DEMUX_CONFIG, MLX4_CMD_MAD_DEMUX,
|
|
|
|
|
MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
|
|
|
|
|
if (err) {
|
|
|
|
|
mlx4_warn(dev, "MLX4_CMD_MAD_DEMUX: configure failed (%d)\n", err);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (secure_host_active)
|
|
|
|
|
mlx4_warn(dev, "HCA operating in secure-host mode. SMP firewall activated.\n");
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
2014-10-27 16:37:37 +07:00
|
|
|
|
|
|
|
|
/* Access Reg commands */
|
|
|
|
|
enum mlx4_access_reg_masks {
|
|
|
|
|
MLX4_ACCESS_REG_STATUS_MASK = 0x7f,
|
|
|
|
|
MLX4_ACCESS_REG_METHOD_MASK = 0x7f,
|
|
|
|
|
MLX4_ACCESS_REG_LEN_MASK = 0x7ff
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
struct mlx4_access_reg {
|
|
|
|
|
__be16 constant1;
|
|
|
|
|
u8 status;
|
|
|
|
|
u8 resrvd1;
|
|
|
|
|
__be16 reg_id;
|
|
|
|
|
u8 method;
|
|
|
|
|
u8 constant2;
|
|
|
|
|
__be32 resrvd2[2];
|
|
|
|
|
__be16 len_const;
|
|
|
|
|
__be16 resrvd3;
|
|
|
|
|
#define MLX4_ACCESS_REG_HEADER_SIZE (20)
|
|
|
|
|
u8 reg_data[MLX4_MAILBOX_SIZE-MLX4_ACCESS_REG_HEADER_SIZE];
|
|
|
|
|
} __attribute__((__packed__));
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* mlx4_ACCESS_REG - Generic access reg command.
|
|
|
|
|
* @dev: mlx4_dev.
|
|
|
|
|
* @reg_id: register ID to access.
|
|
|
|
|
* @method: Access method Read/Write.
|
|
|
|
|
* @reg_len: register length to Read/Write in bytes.
|
|
|
|
|
* @reg_data: reg_data pointer to Read/Write From/To.
|
|
|
|
|
*
|
|
|
|
|
* Access ConnectX registers FW command.
|
|
|
|
|
* Returns 0 on success and copies outbox mlx4_access_reg data
|
|
|
|
|
* field into reg_data or a negative error code.
|
|
|
|
|
*/
|
|
|
|
|
static int mlx4_ACCESS_REG(struct mlx4_dev *dev, u16 reg_id,
|
|
|
|
|
enum mlx4_access_reg_method method,
|
|
|
|
|
u16 reg_len, void *reg_data)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox, *outbox;
|
|
|
|
|
struct mlx4_access_reg *inbuf, *outbuf;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
inbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(inbox))
|
|
|
|
|
return PTR_ERR(inbox);
|
|
|
|
|
|
|
|
|
|
outbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(outbox)) {
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, inbox);
|
|
|
|
|
return PTR_ERR(outbox);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
inbuf = inbox->buf;
|
|
|
|
|
outbuf = outbox->buf;
|
|
|
|
|
|
|
|
|
|
inbuf->constant1 = cpu_to_be16(0x1<<11 | 0x4);
|
|
|
|
|
inbuf->constant2 = 0x1;
|
|
|
|
|
inbuf->reg_id = cpu_to_be16(reg_id);
|
|
|
|
|
inbuf->method = method & MLX4_ACCESS_REG_METHOD_MASK;
|
|
|
|
|
|
|
|
|
|
reg_len = min(reg_len, (u16)(sizeof(inbuf->reg_data)));
|
|
|
|
|
inbuf->len_const =
|
|
|
|
|
cpu_to_be16(((reg_len/4 + 1) & MLX4_ACCESS_REG_LEN_MASK) |
|
|
|
|
|
((0x3) << 12));
|
|
|
|
|
|
|
|
|
|
memcpy(inbuf->reg_data, reg_data, reg_len);
|
|
|
|
|
err = mlx4_cmd_box(dev, inbox->dma, outbox->dma, 0, 0,
|
|
|
|
|
MLX4_CMD_ACCESS_REG, MLX4_CMD_TIME_CLASS_C,
|
2014-11-02 21:26:13 +07:00
|
|
|
MLX4_CMD_WRAPPED);
|
2014-10-27 16:37:37 +07:00
|
|
|
if (err)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
if (outbuf->status & MLX4_ACCESS_REG_STATUS_MASK) {
|
|
|
|
|
err = outbuf->status & MLX4_ACCESS_REG_STATUS_MASK;
|
|
|
|
|
mlx4_err(dev,
|
|
|
|
|
"MLX4_CMD_ACCESS_REG(%x) returned REG status (%x)\n",
|
|
|
|
|
reg_id, err);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
memcpy(reg_data, outbuf->reg_data, reg_len);
|
|
|
|
|
out:
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, inbox);
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, outbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* ConnectX registers IDs */
|
|
|
|
|
enum mlx4_reg_id {
|
|
|
|
|
MLX4_REG_ID_PTYS = 0x5004,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* mlx4_ACCESS_PTYS_REG - Access PTYs (Port Type and Speed)
|
|
|
|
|
* register
|
|
|
|
|
* @dev: mlx4_dev.
|
|
|
|
|
* @method: Access method Read/Write.
|
|
|
|
|
* @ptys_reg: PTYS register data pointer.
|
|
|
|
|
*
|
|
|
|
|
* Access ConnectX PTYS register, to Read/Write Port Type/Speed
|
|
|
|
|
* configuration
|
|
|
|
|
* Returns 0 on success or a negative error code.
|
|
|
|
|
*/
|
|
|
|
|
int mlx4_ACCESS_PTYS_REG(struct mlx4_dev *dev,
|
|
|
|
|
enum mlx4_access_reg_method method,
|
|
|
|
|
struct mlx4_ptys_reg *ptys_reg)
|
|
|
|
|
{
|
|
|
|
|
return mlx4_ACCESS_REG(dev, MLX4_REG_ID_PTYS,
|
|
|
|
|
method, sizeof(*ptys_reg), ptys_reg);
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_ACCESS_PTYS_REG);
|
2014-11-02 21:26:13 +07:00
|
|
|
|
|
|
|
|
int mlx4_ACCESS_REG_wrapper(struct mlx4_dev *dev, int slave,
|
|
|
|
|
struct mlx4_vhcr *vhcr,
|
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
|
struct mlx4_cmd_mailbox *outbox,
|
|
|
|
|
struct mlx4_cmd_info *cmd)
|
|
|
|
|
{
|
|
|
|
|
struct mlx4_access_reg *inbuf = inbox->buf;
|
|
|
|
|
u8 method = inbuf->method & MLX4_ACCESS_REG_METHOD_MASK;
|
|
|
|
|
u16 reg_id = be16_to_cpu(inbuf->reg_id);
|
|
|
|
|
|
|
|
|
|
if (slave != mlx4_master_func_num(dev) &&
|
|
|
|
|
method == MLX4_ACCESS_REG_WRITE)
|
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
|
|
if (reg_id == MLX4_REG_ID_PTYS) {
|
|
|
|
|
struct mlx4_ptys_reg *ptys_reg =
|
|
|
|
|
(struct mlx4_ptys_reg *)inbuf->reg_data;
|
|
|
|
|
|
|
|
|
|
ptys_reg->local_port =
|
|
|
|
|
mlx4_slave_convert_port(dev, slave,
|
|
|
|
|
ptys_reg->local_port);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return mlx4_cmd_box(dev, inbox->dma, outbox->dma, vhcr->in_modifier,
|
|
|
|
|
0, MLX4_CMD_ACCESS_REG, MLX4_CMD_TIME_CLASS_C,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
}
|
2015-07-27 18:46:31 +07:00
|
|
|
|
|
|
|
|
static int mlx4_SET_PORT_phv_bit(struct mlx4_dev *dev, u8 port, u8 phv_bit)
|
|
|
|
|
{
|
|
|
|
|
#define SET_PORT_GEN_PHV_VALID 0x10
|
|
|
|
|
#define SET_PORT_GEN_PHV_EN 0x80
|
|
|
|
|
|
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
struct mlx4_set_port_general_context *context;
|
|
|
|
|
u32 in_mod;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
|
return PTR_ERR(mailbox);
|
|
|
|
|
context = mailbox->buf;
|
|
|
|
|
|
|
|
|
|
context->v_ignore_fcs |= SET_PORT_GEN_PHV_VALID;
|
|
|
|
|
if (phv_bit)
|
|
|
|
|
context->phv_en |= SET_PORT_GEN_PHV_EN;
|
|
|
|
|
|
|
|
|
|
in_mod = MLX4_SET_PORT_GENERAL << 8 | port;
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, in_mod, MLX4_SET_PORT_ETH_OPCODE,
|
|
|
|
|
MLX4_CMD_SET_PORT, MLX4_CMD_TIME_CLASS_B,
|
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int get_phv_bit(struct mlx4_dev *dev, u8 port, int *phv)
|
|
|
|
|
{
|
|
|
|
|
int err;
|
|
|
|
|
struct mlx4_func_cap func_cap;
|
|
|
|
|
|
|
|
|
|
memset(&func_cap, 0, sizeof(func_cap));
|
2015-07-28 17:19:19 +07:00
|
|
|
err = mlx4_QUERY_FUNC_CAP(dev, port, &func_cap);
|
2015-07-27 18:46:31 +07:00
|
|
|
if (!err)
|
|
|
|
|
*phv = func_cap.flags & QUERY_FUNC_CAP_PHV_BIT;
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(get_phv_bit);
|
|
|
|
|
|
|
|
|
|
int set_phv_bit(struct mlx4_dev *dev, u8 port, int new_val)
|
|
|
|
|
{
|
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
|
|
if (mlx4_is_slave(dev))
|
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
|
|
if (dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_PHV_EN &&
|
|
|
|
|
!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_SKIP_OUTER_VLAN)) {
|
|
|
|
|
ret = mlx4_SET_PORT_phv_bit(dev, port, new_val);
|
|
|
|
|
if (!ret)
|
|
|
|
|
dev->caps.phv_bit[port] = new_val;
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return -EOPNOTSUPP;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(set_phv_bit);
|
net/mlx4_core: Replace VF zero mac with random mac in mlx4_core
By design, when no default MAC addresses are set in the Hypervisor for VFs,
the VFs are passed zero-macs. When such a MAC is received by the VF, it
generates a random MAC address and registers that MAC address
with the Hypervisor.
This random mac generation is currently done in the mlx4_en module.
There is a problem, though, if the mlx4_ib module is loaded by a VF before
the mlx4_en module. In this case, for RoCE, mlx4_ib will see the un-replaced
zero-mac and register that zero-mac as part of QP1 initialization.
Having a zero-mac in the port's MAC table creates problems for a
Baseboard Management Console. The BMC occasionally sends packets with a
zero-mac destination MAC. If there is a zero-mac present in the port's
MAC table, the FW will send such BMC packets to the host driver rather than
to the wire, and BMC will stop working.
To address this problem, we move the replacement of zero-mac addresses
with random-mac addresses to procedure mlx4_slave_cap(), which is part of the
driver startup for VFs, and is before activation of mlx4_ib and mlx4_en.
As a result, zero-mac addresses will never be registered in the port MAC table
by the driver.
In addition, when mlx4_en does initialize the net device, it needs to set
the NET_ADDR_RANDOM flag in the netdev structure if the address was
randomly generated. This is done so that udev on the VM does not create
a new device name after each VF probe (VM boot and such). To accomplish this,
we add a per-port flag in mlx4_dev which gets set whenever mlx4_core replaces
a zero-mac with a randomly-generated mac. This flag is examined when mlx4_en
initializes the net-device.
Fix was suggested by Matan Barak <matanb@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-10-14 21:43:48 +07:00
|
|
|
|
|
|
|
|
void mlx4_replace_zero_macs(struct mlx4_dev *dev)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
u8 mac_addr[ETH_ALEN];
|
|
|
|
|
|
|
|
|
|
dev->port_random_macs = 0;
|
|
|
|
|
for (i = 1; i <= dev->caps.num_ports; ++i)
|
|
|
|
|
if (!dev->caps.def_mac[i] &&
|
|
|
|
|
dev->caps.port_type[i] == MLX4_PORT_TYPE_ETH) {
|
|
|
|
|
eth_random_addr(mac_addr);
|
|
|
|
|
dev->port_random_macs |= 1 << i;
|
|
|
|
|
dev->caps.def_mac[i] = mlx4_mac_to_u64(mac_addr);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(mlx4_replace_zero_macs);
|