mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2004, 2005 Topspin Communications. All rights reserved.
|
|
|
|
* Copyright (c) 2005, 2006, 2007, 2008 Mellanox Technologies.
|
|
|
|
* All rights reserved.
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/pci.h>
|
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/io.h>
|
2011-12-26 06:35:34 +07:00
|
|
|
#include <linux/slab.h>
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
#include <linux/mlx4/cmd.h>
|
|
|
|
#include <linux/mlx4/qp.h>
|
2012-07-19 05:33:49 +07:00
|
|
|
#include <linux/if_ether.h>
|
2012-09-06 05:50:49 +07:00
|
|
|
#include <linux/etherdevice.h>
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
#include "mlx4.h"
|
|
|
|
#include "fw.h"
|
|
|
|
|
|
|
|
#define MLX4_MAC_VALID (1ull << 63)
|
|
|
|
|
|
|
|
struct mac_res {
|
|
|
|
struct list_head list;
|
|
|
|
u64 mac;
|
|
|
|
u8 port;
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_common {
|
|
|
|
struct list_head list;
|
2012-07-05 11:03:41 +07:00
|
|
|
struct rb_node node;
|
2012-07-05 11:03:42 +07:00
|
|
|
u64 res_id;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int owner;
|
|
|
|
int state;
|
|
|
|
int from_state;
|
|
|
|
int to_state;
|
|
|
|
int removing;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum {
|
|
|
|
RES_ANY_BUSY = 1
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_gid {
|
|
|
|
struct list_head list;
|
|
|
|
u8 gid[16];
|
|
|
|
enum mlx4_protocol prot;
|
2012-02-15 13:23:16 +07:00
|
|
|
enum mlx4_steer_type steer;
|
2013-04-11 08:56:41 +07:00
|
|
|
u64 reg_id;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
enum res_qp_states {
|
|
|
|
RES_QP_BUSY = RES_ANY_BUSY,
|
|
|
|
|
|
|
|
/* QP number was allocated */
|
|
|
|
RES_QP_RESERVED,
|
|
|
|
|
|
|
|
/* ICM memory for QP context was mapped */
|
|
|
|
RES_QP_MAPPED,
|
|
|
|
|
|
|
|
/* QP is in hw ownership */
|
|
|
|
RES_QP_HW
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_qp {
|
|
|
|
struct res_common com;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct res_cq *rcq;
|
|
|
|
struct res_cq *scq;
|
|
|
|
struct res_srq *srq;
|
|
|
|
struct list_head mcg_list;
|
|
|
|
spinlock_t mcg_spl;
|
|
|
|
int local_qpn;
|
2013-03-21 12:55:55 +07:00
|
|
|
atomic_t ref_count;
|
2013-06-27 23:05:21 +07:00
|
|
|
u32 qpc_flags;
|
|
|
|
u8 sched_queue;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
enum res_mtt_states {
|
|
|
|
RES_MTT_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_MTT_ALLOCATED,
|
|
|
|
};
|
|
|
|
|
|
|
|
static inline const char *mtt_states_str(enum res_mtt_states state)
|
|
|
|
{
|
|
|
|
switch (state) {
|
|
|
|
case RES_MTT_BUSY: return "RES_MTT_BUSY";
|
|
|
|
case RES_MTT_ALLOCATED: return "RES_MTT_ALLOCATED";
|
|
|
|
default: return "Unknown";
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
struct res_mtt {
|
|
|
|
struct res_common com;
|
|
|
|
int order;
|
|
|
|
atomic_t ref_count;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum res_mpt_states {
|
|
|
|
RES_MPT_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_MPT_RESERVED,
|
|
|
|
RES_MPT_MAPPED,
|
|
|
|
RES_MPT_HW,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_mpt {
|
|
|
|
struct res_common com;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
int key;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum res_eq_states {
|
|
|
|
RES_EQ_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_EQ_RESERVED,
|
|
|
|
RES_EQ_HW,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_eq {
|
|
|
|
struct res_common com;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum res_cq_states {
|
|
|
|
RES_CQ_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_CQ_ALLOCATED,
|
|
|
|
RES_CQ_HW,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_cq {
|
|
|
|
struct res_common com;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
atomic_t ref_count;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum res_srq_states {
|
|
|
|
RES_SRQ_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_SRQ_ALLOCATED,
|
|
|
|
RES_SRQ_HW,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_srq {
|
|
|
|
struct res_common com;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct res_cq *cq;
|
|
|
|
atomic_t ref_count;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum res_counter_states {
|
|
|
|
RES_COUNTER_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_COUNTER_ALLOCATED,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_counter {
|
|
|
|
struct res_common com;
|
|
|
|
int port;
|
|
|
|
};
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
enum res_xrcdn_states {
|
|
|
|
RES_XRCD_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_XRCD_ALLOCATED,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_xrcdn {
|
|
|
|
struct res_common com;
|
|
|
|
int port;
|
|
|
|
};
|
|
|
|
|
2012-07-05 11:03:47 +07:00
|
|
|
enum res_fs_rule_states {
|
|
|
|
RES_FS_RULE_BUSY = RES_ANY_BUSY,
|
|
|
|
RES_FS_RULE_ALLOCATED,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct res_fs_rule {
|
|
|
|
struct res_common com;
|
2013-03-21 12:55:55 +07:00
|
|
|
int qpn;
|
2012-07-05 11:03:47 +07:00
|
|
|
};
|
|
|
|
|
2012-07-05 11:03:41 +07:00
|
|
|
static void *res_tracker_lookup(struct rb_root *root, u64 res_id)
|
|
|
|
{
|
|
|
|
struct rb_node *node = root->rb_node;
|
|
|
|
|
|
|
|
while (node) {
|
|
|
|
struct res_common *res = container_of(node, struct res_common,
|
|
|
|
node);
|
|
|
|
|
|
|
|
if (res_id < res->res_id)
|
|
|
|
node = node->rb_left;
|
|
|
|
else if (res_id > res->res_id)
|
|
|
|
node = node->rb_right;
|
|
|
|
else
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int res_tracker_insert(struct rb_root *root, struct res_common *res)
|
|
|
|
{
|
|
|
|
struct rb_node **new = &(root->rb_node), *parent = NULL;
|
|
|
|
|
|
|
|
/* Figure out where to put new node */
|
|
|
|
while (*new) {
|
|
|
|
struct res_common *this = container_of(*new, struct res_common,
|
|
|
|
node);
|
|
|
|
|
|
|
|
parent = *new;
|
|
|
|
if (res->res_id < this->res_id)
|
|
|
|
new = &((*new)->rb_left);
|
|
|
|
else if (res->res_id > this->res_id)
|
|
|
|
new = &((*new)->rb_right);
|
|
|
|
else
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Add new node and rebalance tree. */
|
|
|
|
rb_link_node(&res->node, parent, new);
|
|
|
|
rb_insert_color(&res->node, root);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
enum qp_transition {
|
|
|
|
QP_TRANS_INIT2RTR,
|
|
|
|
QP_TRANS_RTR2RTS,
|
|
|
|
QP_TRANS_RTS2RTS,
|
|
|
|
QP_TRANS_SQERR2RTS,
|
|
|
|
QP_TRANS_SQD2SQD,
|
|
|
|
QP_TRANS_SQD2RTS
|
|
|
|
};
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
/* For Debug uses */
|
|
|
|
static const char *ResourceType(enum mlx4_resource rt)
|
|
|
|
{
|
|
|
|
switch (rt) {
|
|
|
|
case RES_QP: return "RES_QP";
|
|
|
|
case RES_CQ: return "RES_CQ";
|
|
|
|
case RES_SRQ: return "RES_SRQ";
|
|
|
|
case RES_MPT: return "RES_MPT";
|
|
|
|
case RES_MTT: return "RES_MTT";
|
|
|
|
case RES_MAC: return "RES_MAC";
|
|
|
|
case RES_EQ: return "RES_EQ";
|
|
|
|
case RES_COUNTER: return "RES_COUNTER";
|
2012-07-05 11:03:47 +07:00
|
|
|
case RES_FS_RULE: return "RES_FS_RULE";
|
2012-05-15 17:35:03 +07:00
|
|
|
case RES_XRCD: return "RES_XRCD";
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
default: return "Unknown resource type !!!";
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_init_resource_tracker(struct mlx4_dev *dev)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
int i;
|
|
|
|
int t;
|
|
|
|
|
|
|
|
priv->mfunc.master.res_tracker.slave_list =
|
|
|
|
kzalloc(dev->num_slaves * sizeof(struct slave_list),
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!priv->mfunc.master.res_tracker.slave_list)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
for (i = 0 ; i < dev->num_slaves; i++) {
|
|
|
|
for (t = 0; t < MLX4_NUM_OF_RESOURCE_TYPE; ++t)
|
|
|
|
INIT_LIST_HEAD(&priv->mfunc.master.res_tracker.
|
|
|
|
slave_list[i].res_list[t]);
|
|
|
|
mutex_init(&priv->mfunc.master.res_tracker.slave_list[i].mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
mlx4_dbg(dev, "Started init_resource_tracker: %ld slaves\n",
|
|
|
|
dev->num_slaves);
|
|
|
|
for (i = 0 ; i < MLX4_NUM_OF_RESOURCE_TYPE; i++)
|
2012-07-05 11:03:41 +07:00
|
|
|
priv->mfunc.master.res_tracker.res_tree[i] = RB_ROOT;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
spin_lock_init(&priv->mfunc.master.res_tracker.lock);
|
|
|
|
return 0 ;
|
|
|
|
}
|
|
|
|
|
2012-05-15 17:35:02 +07:00
|
|
|
void mlx4_free_resource_tracker(struct mlx4_dev *dev,
|
|
|
|
enum mlx4_res_tracker_free_type type)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (priv->mfunc.master.res_tracker.slave_list) {
|
2012-05-15 17:35:02 +07:00
|
|
|
if (type != RES_TR_FREE_STRUCTS_ONLY)
|
|
|
|
for (i = 0 ; i < dev->num_slaves; i++)
|
|
|
|
if (type == RES_TR_FREE_ALL ||
|
|
|
|
dev->caps.function != i)
|
|
|
|
mlx4_delete_all_resources_for_slave(dev, i);
|
|
|
|
|
|
|
|
if (type != RES_TR_FREE_SLAVES_ONLY) {
|
|
|
|
kfree(priv->mfunc.master.res_tracker.slave_list);
|
|
|
|
priv->mfunc.master.res_tracker.slave_list = NULL;
|
|
|
|
}
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
static void update_pkey_index(struct mlx4_dev *dev, int slave,
|
|
|
|
struct mlx4_cmd_mailbox *inbox)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
2012-08-03 15:40:43 +07:00
|
|
|
u8 sched = *(u8 *)(inbox->buf + 64);
|
|
|
|
u8 orig_index = *(u8 *)(inbox->buf + 35);
|
|
|
|
u8 new_index;
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
int port;
|
|
|
|
|
|
|
|
port = (sched >> 6 & 1) + 1;
|
|
|
|
|
|
|
|
new_index = priv->virt2phys_pkey[slave][port - 1][orig_index];
|
|
|
|
*(u8 *)(inbox->buf + 35) = new_index;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void update_gid(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *inbox,
|
|
|
|
u8 slave)
|
|
|
|
{
|
|
|
|
struct mlx4_qp_context *qp_ctx = inbox->buf + 8;
|
|
|
|
enum mlx4_qp_optpar optpar = be32_to_cpu(*(__be32 *) inbox->buf);
|
|
|
|
u32 ts = (be32_to_cpu(qp_ctx->flags) >> 16) & 0xff;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
if (MLX4_QP_ST_UD == ts)
|
|
|
|
qp_ctx->pri_path.mgid_index = 0x80 | slave;
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
if (MLX4_QP_ST_RC == ts || MLX4_QP_ST_UC == ts) {
|
|
|
|
if (optpar & MLX4_QP_OPTPAR_PRIMARY_ADDR_PATH)
|
|
|
|
qp_ctx->pri_path.mgid_index = slave & 0x7F;
|
|
|
|
if (optpar & MLX4_QP_OPTPAR_ALT_ADDR_PATH)
|
|
|
|
qp_ctx->alt_path.mgid_index = slave & 0x7F;
|
|
|
|
}
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
|
2013-04-25 12:22:28 +07:00
|
|
|
static int update_vport_qp_param(struct mlx4_dev *dev,
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
2013-06-27 23:05:21 +07:00
|
|
|
u8 slave, u32 qpn)
|
2013-04-25 12:22:28 +07:00
|
|
|
{
|
|
|
|
struct mlx4_qp_context *qpc = inbox->buf + 8;
|
|
|
|
struct mlx4_vport_oper_state *vp_oper;
|
|
|
|
struct mlx4_priv *priv;
|
|
|
|
u32 qp_type;
|
|
|
|
int port;
|
|
|
|
|
|
|
|
port = (qpc->pri_path.sched_queue & 0x40) ? 2 : 1;
|
|
|
|
priv = mlx4_priv(dev);
|
|
|
|
vp_oper = &priv->mfunc.master.vf_oper[slave].vport[port];
|
|
|
|
|
|
|
|
if (MLX4_VGT != vp_oper->state.default_vlan) {
|
|
|
|
qp_type = (be32_to_cpu(qpc->flags) >> 16) & 0xff;
|
2013-06-27 23:05:21 +07:00
|
|
|
if (MLX4_QP_ST_RC == qp_type ||
|
|
|
|
(MLX4_QP_ST_UD == qp_type &&
|
|
|
|
!mlx4_is_qp_reserved(dev, qpn)))
|
2013-04-25 12:22:28 +07:00
|
|
|
return -EINVAL;
|
|
|
|
|
2013-06-27 23:05:21 +07:00
|
|
|
/* the reserved QPs (special, proxy, tunnel)
|
|
|
|
* do not operate over vlans
|
|
|
|
*/
|
|
|
|
if (mlx4_is_qp_reserved(dev, qpn))
|
|
|
|
return 0;
|
|
|
|
|
2013-05-09 05:22:35 +07:00
|
|
|
/* force strip vlan by clear vsd */
|
|
|
|
qpc->param3 &= ~cpu_to_be32(MLX4_STRIP_VLAN);
|
2013-06-27 23:05:22 +07:00
|
|
|
|
|
|
|
if (vp_oper->state.link_state == IFLA_VF_LINK_STATE_DISABLE &&
|
|
|
|
dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_UPDATE_QP) {
|
|
|
|
qpc->pri_path.vlan_control =
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_PRIO_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_UNTAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_PRIO_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_UNTAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_TAGGED;
|
|
|
|
} else if (0 != vp_oper->state.default_vlan) {
|
2013-05-09 05:22:35 +07:00
|
|
|
qpc->pri_path.vlan_control =
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_PRIO_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_UNTAGGED;
|
|
|
|
} else { /* priority tagged */
|
|
|
|
qpc->pri_path.vlan_control =
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_TAGGED;
|
|
|
|
}
|
|
|
|
|
|
|
|
qpc->pri_path.fvl_rx |= MLX4_FVL_RX_FORCE_ETH_VLAN;
|
2013-04-25 12:22:28 +07:00
|
|
|
qpc->pri_path.vlan_index = vp_oper->vlan_idx;
|
2013-05-09 05:22:35 +07:00
|
|
|
qpc->pri_path.fl |= MLX4_FL_CV | MLX4_FL_ETH_HIDE_CQE_VLAN;
|
|
|
|
qpc->pri_path.feup |= MLX4_FEUP_FORCE_ETH_UP | MLX4_FVL_FORCE_ETH_VLAN;
|
2013-04-25 12:22:28 +07:00
|
|
|
qpc->pri_path.sched_queue &= 0xC7;
|
|
|
|
qpc->pri_path.sched_queue |= (vp_oper->state.default_qos) << 3;
|
|
|
|
}
|
2013-04-25 12:22:29 +07:00
|
|
|
if (vp_oper->state.spoofchk) {
|
2013-05-09 05:22:35 +07:00
|
|
|
qpc->pri_path.feup |= MLX4_FSM_FORCE_ETH_SRC_MAC;
|
2013-04-25 12:22:29 +07:00
|
|
|
qpc->pri_path.grh_mylmc = (0x80 & qpc->pri_path.grh_mylmc) + vp_oper->mac_idx;
|
|
|
|
}
|
2013-04-25 12:22:28 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
static int mpt_mask(struct mlx4_dev *dev)
|
|
|
|
{
|
|
|
|
return dev->caps.num_mpts - 1;
|
|
|
|
}
|
|
|
|
|
2013-03-21 12:55:54 +07:00
|
|
|
static void *find_res(struct mlx4_dev *dev, u64 res_id,
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
enum mlx4_resource type)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
2012-07-05 11:03:41 +07:00
|
|
|
return res_tracker_lookup(&priv->mfunc.master.res_tracker.res_tree[type],
|
|
|
|
res_id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:42 +07:00
|
|
|
static int get_res(struct mlx4_dev *dev, int slave, u64 res_id,
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
enum mlx4_resource type,
|
|
|
|
void *res)
|
|
|
|
{
|
|
|
|
struct res_common *r;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
r = find_res(dev, res_id, type);
|
|
|
|
if (!r) {
|
|
|
|
err = -ENONET;
|
|
|
|
goto exit;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (r->state == RES_ANY_BUSY) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto exit;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (r->owner != slave) {
|
|
|
|
err = -EPERM;
|
|
|
|
goto exit;
|
|
|
|
}
|
|
|
|
|
|
|
|
r->from_state = r->state;
|
|
|
|
r->state = RES_ANY_BUSY;
|
|
|
|
|
|
|
|
if (res)
|
|
|
|
*((struct res_common **)res) = r;
|
|
|
|
|
|
|
|
exit:
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_get_slave_from_resource_id(struct mlx4_dev *dev,
|
|
|
|
enum mlx4_resource type,
|
2012-07-05 11:03:42 +07:00
|
|
|
u64 res_id, int *slave)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
|
|
|
|
struct res_common *r;
|
|
|
|
int err = -ENOENT;
|
|
|
|
int id = res_id;
|
|
|
|
|
|
|
|
if (type == RES_QP)
|
|
|
|
id &= 0x7fffff;
|
2011-12-19 11:03:05 +07:00
|
|
|
spin_lock(mlx4_tlock(dev));
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
r = find_res(dev, id, type);
|
|
|
|
if (r) {
|
|
|
|
*slave = r->owner;
|
|
|
|
err = 0;
|
|
|
|
}
|
2011-12-19 11:03:05 +07:00
|
|
|
spin_unlock(mlx4_tlock(dev));
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:42 +07:00
|
|
|
static void put_res(struct mlx4_dev *dev, int slave, u64 res_id,
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
enum mlx4_resource type)
|
|
|
|
{
|
|
|
|
struct res_common *r;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
r = find_res(dev, res_id, type);
|
|
|
|
if (r)
|
|
|
|
r->state = r->from_state;
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_qp_tr(int id)
|
|
|
|
{
|
|
|
|
struct res_qp *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_QP_RESERVED;
|
2012-02-15 13:22:57 +07:00
|
|
|
ret->local_qpn = id;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
INIT_LIST_HEAD(&ret->mcg_list);
|
|
|
|
spin_lock_init(&ret->mcg_spl);
|
2013-03-21 12:55:55 +07:00
|
|
|
atomic_set(&ret->ref_count, 0);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_mtt_tr(int id, int order)
|
|
|
|
{
|
|
|
|
struct res_mtt *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->order = order;
|
|
|
|
ret->com.state = RES_MTT_ALLOCATED;
|
|
|
|
atomic_set(&ret->ref_count, 0);
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_mpt_tr(int id, int key)
|
|
|
|
{
|
|
|
|
struct res_mpt *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_MPT_RESERVED;
|
|
|
|
ret->key = key;
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_eq_tr(int id)
|
|
|
|
{
|
|
|
|
struct res_eq *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_EQ_RESERVED;
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_cq_tr(int id)
|
|
|
|
{
|
|
|
|
struct res_cq *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_CQ_ALLOCATED;
|
|
|
|
atomic_set(&ret->ref_count, 0);
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_srq_tr(int id)
|
|
|
|
{
|
|
|
|
struct res_srq *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_SRQ_ALLOCATED;
|
|
|
|
atomic_set(&ret->ref_count, 0);
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_common *alloc_counter_tr(int id)
|
|
|
|
{
|
|
|
|
struct res_counter *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_COUNTER_ALLOCATED;
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
static struct res_common *alloc_xrcdn_tr(int id)
|
|
|
|
{
|
|
|
|
struct res_xrcdn *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_XRCD_ALLOCATED;
|
|
|
|
|
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
2013-03-21 12:55:55 +07:00
|
|
|
static struct res_common *alloc_fs_rule_tr(u64 id, int qpn)
|
2012-07-05 11:03:47 +07:00
|
|
|
{
|
|
|
|
struct res_fs_rule *ret;
|
|
|
|
|
|
|
|
ret = kzalloc(sizeof *ret, GFP_KERNEL);
|
|
|
|
if (!ret)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
ret->com.res_id = id;
|
|
|
|
ret->com.state = RES_FS_RULE_ALLOCATED;
|
2013-03-21 12:55:55 +07:00
|
|
|
ret->qpn = qpn;
|
2012-07-05 11:03:47 +07:00
|
|
|
return &ret->com;
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:42 +07:00
|
|
|
static struct res_common *alloc_tr(u64 id, enum mlx4_resource type, int slave,
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int extra)
|
|
|
|
{
|
|
|
|
struct res_common *ret;
|
|
|
|
|
|
|
|
switch (type) {
|
|
|
|
case RES_QP:
|
|
|
|
ret = alloc_qp_tr(id);
|
|
|
|
break;
|
|
|
|
case RES_MPT:
|
|
|
|
ret = alloc_mpt_tr(id, extra);
|
|
|
|
break;
|
|
|
|
case RES_MTT:
|
|
|
|
ret = alloc_mtt_tr(id, extra);
|
|
|
|
break;
|
|
|
|
case RES_EQ:
|
|
|
|
ret = alloc_eq_tr(id);
|
|
|
|
break;
|
|
|
|
case RES_CQ:
|
|
|
|
ret = alloc_cq_tr(id);
|
|
|
|
break;
|
|
|
|
case RES_SRQ:
|
|
|
|
ret = alloc_srq_tr(id);
|
|
|
|
break;
|
|
|
|
case RES_MAC:
|
|
|
|
printk(KERN_ERR "implementation missing\n");
|
|
|
|
return NULL;
|
|
|
|
case RES_COUNTER:
|
|
|
|
ret = alloc_counter_tr(id);
|
|
|
|
break;
|
2012-05-15 17:35:03 +07:00
|
|
|
case RES_XRCD:
|
|
|
|
ret = alloc_xrcdn_tr(id);
|
|
|
|
break;
|
2012-07-05 11:03:47 +07:00
|
|
|
case RES_FS_RULE:
|
2013-03-21 12:55:55 +07:00
|
|
|
ret = alloc_fs_rule_tr(id, extra);
|
2012-07-05 11:03:47 +07:00
|
|
|
break;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
default:
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
if (ret)
|
|
|
|
ret->owner = slave;
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:42 +07:00
|
|
|
static int add_res_range(struct mlx4_dev *dev, int slave, u64 base, int count,
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
enum mlx4_resource type, int extra)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
int err;
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct res_common **res_arr;
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
2012-07-05 11:03:41 +07:00
|
|
|
struct rb_root *root = &tracker->res_tree[type];
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
res_arr = kzalloc(count * sizeof *res_arr, GFP_KERNEL);
|
|
|
|
if (!res_arr)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
for (i = 0; i < count; ++i) {
|
|
|
|
res_arr[i] = alloc_tr(base + i, type, slave, extra);
|
|
|
|
if (!res_arr[i]) {
|
|
|
|
for (--i; i >= 0; --i)
|
|
|
|
kfree(res_arr[i]);
|
|
|
|
|
|
|
|
kfree(res_arr);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
for (i = 0; i < count; ++i) {
|
|
|
|
if (find_res(dev, base + i, type)) {
|
|
|
|
err = -EEXIST;
|
|
|
|
goto undo;
|
|
|
|
}
|
2012-07-05 11:03:41 +07:00
|
|
|
err = res_tracker_insert(root, res_arr[i]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto undo;
|
|
|
|
list_add_tail(&res_arr[i]->list,
|
|
|
|
&tracker->slave_list[slave].res_list[type]);
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(res_arr);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
undo:
|
|
|
|
for (--i; i >= base; --i)
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&res_arr[i]->node, root);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
for (i = 0; i < count; ++i)
|
|
|
|
kfree(res_arr[i]);
|
|
|
|
|
|
|
|
kfree(res_arr);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_qp_ok(struct res_qp *res)
|
|
|
|
{
|
2013-03-21 12:55:55 +07:00
|
|
|
if (res->com.state == RES_QP_BUSY || atomic_read(&res->ref_count) ||
|
|
|
|
!list_empty(&res->mcg_list)) {
|
|
|
|
pr_err("resource tracker: fail to remove qp, state %d, ref_count %d\n",
|
|
|
|
res->com.state, atomic_read(&res->ref_count));
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
return -EBUSY;
|
2013-03-21 12:55:55 +07:00
|
|
|
} else if (res->com.state != RES_QP_RESERVED) {
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
return -EPERM;
|
2013-03-21 12:55:55 +07:00
|
|
|
}
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_mtt_ok(struct res_mtt *res, int order)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_MTT_BUSY ||
|
|
|
|
atomic_read(&res->ref_count)) {
|
|
|
|
printk(KERN_DEBUG "%s-%d: state %s, ref_count %d\n",
|
|
|
|
__func__, __LINE__,
|
|
|
|
mtt_states_str(res->com.state),
|
|
|
|
atomic_read(&res->ref_count));
|
|
|
|
return -EBUSY;
|
|
|
|
} else if (res->com.state != RES_MTT_ALLOCATED)
|
|
|
|
return -EPERM;
|
|
|
|
else if (res->order != order)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_mpt_ok(struct res_mpt *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_MPT_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_MPT_RESERVED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_eq_ok(struct res_eq *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_MPT_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_MPT_RESERVED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_counter_ok(struct res_counter *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_COUNTER_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_COUNTER_ALLOCATED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
static int remove_xrcdn_ok(struct res_xrcdn *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_XRCD_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_XRCD_ALLOCATED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:47 +07:00
|
|
|
static int remove_fs_rule_ok(struct res_fs_rule *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_FS_RULE_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_FS_RULE_ALLOCATED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
static int remove_cq_ok(struct res_cq *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_CQ_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_CQ_ALLOCATED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_srq_ok(struct res_srq *res)
|
|
|
|
{
|
|
|
|
if (res->com.state == RES_SRQ_BUSY)
|
|
|
|
return -EBUSY;
|
|
|
|
else if (res->com.state != RES_SRQ_ALLOCATED)
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int remove_ok(struct res_common *res, enum mlx4_resource type, int extra)
|
|
|
|
{
|
|
|
|
switch (type) {
|
|
|
|
case RES_QP:
|
|
|
|
return remove_qp_ok((struct res_qp *)res);
|
|
|
|
case RES_CQ:
|
|
|
|
return remove_cq_ok((struct res_cq *)res);
|
|
|
|
case RES_SRQ:
|
|
|
|
return remove_srq_ok((struct res_srq *)res);
|
|
|
|
case RES_MPT:
|
|
|
|
return remove_mpt_ok((struct res_mpt *)res);
|
|
|
|
case RES_MTT:
|
|
|
|
return remove_mtt_ok((struct res_mtt *)res, extra);
|
|
|
|
case RES_MAC:
|
|
|
|
return -ENOSYS;
|
|
|
|
case RES_EQ:
|
|
|
|
return remove_eq_ok((struct res_eq *)res);
|
|
|
|
case RES_COUNTER:
|
|
|
|
return remove_counter_ok((struct res_counter *)res);
|
2012-05-15 17:35:03 +07:00
|
|
|
case RES_XRCD:
|
|
|
|
return remove_xrcdn_ok((struct res_xrcdn *)res);
|
2012-07-05 11:03:47 +07:00
|
|
|
case RES_FS_RULE:
|
|
|
|
return remove_fs_rule_ok((struct res_fs_rule *)res);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:42 +07:00
|
|
|
static int rem_res_range(struct mlx4_dev *dev, int slave, u64 base, int count,
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
enum mlx4_resource type, int extra)
|
|
|
|
{
|
2012-07-05 11:03:42 +07:00
|
|
|
u64 i;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int err;
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_common *r;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
for (i = base; i < base + count; ++i) {
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[type], i);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (!r) {
|
|
|
|
err = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (r->owner != slave) {
|
|
|
|
err = -EPERM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
err = remove_ok(r, type, extra);
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = base; i < base + count; ++i) {
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[type], i);
|
|
|
|
rb_erase(&r->node, &tracker->res_tree[type]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&r->list);
|
|
|
|
kfree(r);
|
|
|
|
}
|
|
|
|
err = 0;
|
|
|
|
|
|
|
|
out:
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_res_start_move_to(struct mlx4_dev *dev, int slave, int qpn,
|
|
|
|
enum res_qp_states state, struct res_qp **qp,
|
|
|
|
int alloc)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_qp *r;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[RES_QP], qpn);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (!r)
|
|
|
|
err = -ENOENT;
|
|
|
|
else if (r->com.owner != slave)
|
|
|
|
err = -EPERM;
|
|
|
|
else {
|
|
|
|
switch (state) {
|
|
|
|
case RES_QP_BUSY:
|
2012-07-05 11:03:42 +07:00
|
|
|
mlx4_dbg(dev, "%s: failed RES_QP, 0x%llx\n",
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
__func__, r->com.res_id);
|
|
|
|
err = -EBUSY;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_QP_RESERVED:
|
|
|
|
if (r->com.state == RES_QP_MAPPED && !alloc)
|
|
|
|
break;
|
|
|
|
|
2012-07-05 11:03:42 +07:00
|
|
|
mlx4_dbg(dev, "failed RES_QP, 0x%llx\n", r->com.res_id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_QP_MAPPED:
|
|
|
|
if ((r->com.state == RES_QP_RESERVED && alloc) ||
|
|
|
|
r->com.state == RES_QP_HW)
|
|
|
|
break;
|
|
|
|
else {
|
2012-07-05 11:03:42 +07:00
|
|
|
mlx4_dbg(dev, "failed RES_QP, 0x%llx\n",
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
r->com.res_id);
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_QP_HW:
|
|
|
|
if (r->com.state != RES_QP_MAPPED)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
r->com.from_state = r->com.state;
|
|
|
|
r->com.to_state = state;
|
|
|
|
r->com.state = RES_QP_BUSY;
|
|
|
|
if (qp)
|
2012-06-04 19:44:16 +07:00
|
|
|
*qp = r;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mr_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
|
|
|
|
enum res_mpt_states state, struct res_mpt **mpt)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_mpt *r;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[RES_MPT], index);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (!r)
|
|
|
|
err = -ENOENT;
|
|
|
|
else if (r->com.owner != slave)
|
|
|
|
err = -EPERM;
|
|
|
|
else {
|
|
|
|
switch (state) {
|
|
|
|
case RES_MPT_BUSY:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT_RESERVED:
|
|
|
|
if (r->com.state != RES_MPT_MAPPED)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT_MAPPED:
|
|
|
|
if (r->com.state != RES_MPT_RESERVED &&
|
|
|
|
r->com.state != RES_MPT_HW)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT_HW:
|
|
|
|
if (r->com.state != RES_MPT_MAPPED)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
r->com.from_state = r->com.state;
|
|
|
|
r->com.to_state = state;
|
|
|
|
r->com.state = RES_MPT_BUSY;
|
|
|
|
if (mpt)
|
2012-06-04 19:44:16 +07:00
|
|
|
*mpt = r;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int eq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
|
|
|
|
enum res_eq_states state, struct res_eq **eq)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_eq *r;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[RES_EQ], index);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (!r)
|
|
|
|
err = -ENOENT;
|
|
|
|
else if (r->com.owner != slave)
|
|
|
|
err = -EPERM;
|
|
|
|
else {
|
|
|
|
switch (state) {
|
|
|
|
case RES_EQ_BUSY:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_EQ_RESERVED:
|
|
|
|
if (r->com.state != RES_EQ_HW)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_EQ_HW:
|
|
|
|
if (r->com.state != RES_EQ_RESERVED)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
r->com.from_state = r->com.state;
|
|
|
|
r->com.to_state = state;
|
|
|
|
r->com.state = RES_EQ_BUSY;
|
|
|
|
if (eq)
|
|
|
|
*eq = r;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cq_res_start_move_to(struct mlx4_dev *dev, int slave, int cqn,
|
|
|
|
enum res_cq_states state, struct res_cq **cq)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_cq *r;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[RES_CQ], cqn);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (!r)
|
|
|
|
err = -ENOENT;
|
|
|
|
else if (r->com.owner != slave)
|
|
|
|
err = -EPERM;
|
|
|
|
else {
|
|
|
|
switch (state) {
|
|
|
|
case RES_CQ_BUSY:
|
|
|
|
err = -EBUSY;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_CQ_ALLOCATED:
|
|
|
|
if (r->com.state != RES_CQ_HW)
|
|
|
|
err = -EINVAL;
|
|
|
|
else if (atomic_read(&r->ref_count))
|
|
|
|
err = -EBUSY;
|
|
|
|
else
|
|
|
|
err = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_CQ_HW:
|
|
|
|
if (r->com.state != RES_CQ_ALLOCATED)
|
|
|
|
err = -EINVAL;
|
|
|
|
else
|
|
|
|
err = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
r->com.from_state = r->com.state;
|
|
|
|
r->com.to_state = state;
|
|
|
|
r->com.state = RES_CQ_BUSY;
|
|
|
|
if (cq)
|
|
|
|
*cq = r;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int srq_res_start_move_to(struct mlx4_dev *dev, int slave, int index,
|
|
|
|
enum res_cq_states state, struct res_srq **srq)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_srq *r;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[RES_SRQ], index);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (!r)
|
|
|
|
err = -ENOENT;
|
|
|
|
else if (r->com.owner != slave)
|
|
|
|
err = -EPERM;
|
|
|
|
else {
|
|
|
|
switch (state) {
|
|
|
|
case RES_SRQ_BUSY:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_SRQ_ALLOCATED:
|
|
|
|
if (r->com.state != RES_SRQ_HW)
|
|
|
|
err = -EINVAL;
|
|
|
|
else if (atomic_read(&r->ref_count))
|
|
|
|
err = -EBUSY;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_SRQ_HW:
|
|
|
|
if (r->com.state != RES_SRQ_ALLOCATED)
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
r->com.from_state = r->com.state;
|
|
|
|
r->com.to_state = state;
|
|
|
|
r->com.state = RES_SRQ_BUSY;
|
|
|
|
if (srq)
|
|
|
|
*srq = r;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void res_abort_move(struct mlx4_dev *dev, int slave,
|
|
|
|
enum mlx4_resource type, int id)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_common *r;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[type], id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (r && (r->owner == slave))
|
|
|
|
r->state = r->from_state;
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void res_end_move(struct mlx4_dev *dev, int slave,
|
|
|
|
enum mlx4_resource type, int id)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_common *r;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
r = res_tracker_lookup(&tracker->res_tree[type], id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (r && (r->owner == slave))
|
|
|
|
r->state = r->to_state;
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static int valid_reserved(struct mlx4_dev *dev, int slave, int qpn)
|
|
|
|
{
|
2012-08-03 15:40:41 +07:00
|
|
|
return mlx4_is_qp_reserved(dev, qpn) &&
|
|
|
|
(mlx4_is_master(dev) || mlx4_is_guest_proxy(dev, slave, qpn));
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
static int fw_reserved(struct mlx4_dev *dev, int qpn)
|
|
|
|
{
|
|
|
|
return qpn < dev->caps.reserved_qps_cnt[MLX4_QP_REGION_FW];
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int count;
|
|
|
|
int align;
|
|
|
|
int base;
|
|
|
|
int qpn;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE:
|
|
|
|
count = get_param_l(&in_param);
|
|
|
|
align = get_param_h(&in_param);
|
|
|
|
err = __mlx4_qp_reserve_range(dev, count, align, &base);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, base, count, RES_QP, 0);
|
|
|
|
if (err) {
|
|
|
|
__mlx4_qp_release_range(dev, base, count);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
set_param_l(out_param, base);
|
|
|
|
break;
|
|
|
|
case RES_OP_MAP_ICM:
|
|
|
|
qpn = get_param_l(&in_param) & 0x7fffff;
|
|
|
|
if (valid_reserved(dev, slave, qpn)) {
|
|
|
|
err = add_res_range(dev, slave, qpn, 1, RES_QP, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = qp_res_start_move_to(dev, slave, qpn, RES_QP_MAPPED,
|
|
|
|
NULL, 1);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
if (!fw_reserved(dev, qpn)) {
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
err = __mlx4_qp_alloc_icm(dev, qpn);
|
|
|
|
if (err) {
|
|
|
|
res_abort_move(dev, slave, RES_QP, qpn);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
res_end_move(dev, slave, RES_QP, qpn);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mtt_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int err = -EINVAL;
|
|
|
|
int base;
|
|
|
|
int order;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE_AND_MAP)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
order = get_param_l(&in_param);
|
|
|
|
base = __mlx4_alloc_mtt_range(dev, order);
|
|
|
|
if (base == -1)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, base, 1, RES_MTT, order);
|
|
|
|
if (err)
|
|
|
|
__mlx4_free_mtt_range(dev, base, order);
|
|
|
|
else
|
|
|
|
set_param_l(out_param, base);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mpt_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int err = -EINVAL;
|
|
|
|
int index;
|
|
|
|
int id;
|
|
|
|
struct res_mpt *mpt;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE:
|
2013-02-06 23:19:08 +07:00
|
|
|
index = __mlx4_mpt_reserve(dev);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (index == -1)
|
|
|
|
break;
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, id, 1, RES_MPT, index);
|
|
|
|
if (err) {
|
2013-02-06 23:19:08 +07:00
|
|
|
__mlx4_mpt_release(dev, index);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
set_param_l(out_param, index);
|
|
|
|
break;
|
|
|
|
case RES_OP_MAP_ICM:
|
|
|
|
index = get_param_l(&in_param);
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
err = mr_res_start_move_to(dev, slave, id,
|
|
|
|
RES_MPT_MAPPED, &mpt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2013-02-06 23:19:08 +07:00
|
|
|
err = __mlx4_mpt_alloc_icm(dev, mpt->key);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err) {
|
|
|
|
res_abort_move(dev, slave, RES_MPT, id);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
res_end_move(dev, slave, RES_MPT, id);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cq_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int cqn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE_AND_MAP:
|
|
|
|
err = __mlx4_cq_alloc_icm(dev, &cqn);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, cqn, 1, RES_CQ, 0);
|
|
|
|
if (err) {
|
|
|
|
__mlx4_cq_free_icm(dev, cqn);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
set_param_l(out_param, cqn);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int srq_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int srqn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE_AND_MAP:
|
|
|
|
err = __mlx4_srq_alloc_icm(dev, &srqn);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, srqn, 1, RES_SRQ, 0);
|
|
|
|
if (err) {
|
|
|
|
__mlx4_srq_free_icm(dev, srqn);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
set_param_l(out_param, srqn);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mac_add_to_slave(struct mlx4_dev *dev, int slave, u64 mac, int port)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct mac_res *res;
|
|
|
|
|
|
|
|
res = kzalloc(sizeof *res, GFP_KERNEL);
|
|
|
|
if (!res)
|
|
|
|
return -ENOMEM;
|
|
|
|
res->mac = mac;
|
|
|
|
res->port = (u8) port;
|
|
|
|
list_add_tail(&res->list,
|
|
|
|
&tracker->slave_list[slave].res_list[RES_MAC]);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void mac_del_from_slave(struct mlx4_dev *dev, int slave, u64 mac,
|
|
|
|
int port)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *mac_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_MAC];
|
|
|
|
struct mac_res *res, *tmp;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(res, tmp, mac_list, list) {
|
|
|
|
if (res->mac == mac && res->port == (u8) port) {
|
|
|
|
list_del(&res->list);
|
|
|
|
kfree(res);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rem_slave_macs(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *mac_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_MAC];
|
|
|
|
struct mac_res *res, *tmp;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(res, tmp, mac_list, list) {
|
|
|
|
list_del(&res->list);
|
|
|
|
__mlx4_unregister_mac(dev, res->port, res->mac);
|
|
|
|
kfree(res);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mac_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int err = -EINVAL;
|
|
|
|
int port;
|
|
|
|
u64 mac;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE_AND_MAP)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
port = get_param_l(out_param);
|
|
|
|
mac = in_param;
|
|
|
|
|
|
|
|
err = __mlx4_register_mac(dev, port, mac);
|
|
|
|
if (err >= 0) {
|
|
|
|
set_param_l(out_param, err);
|
|
|
|
err = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
err = mac_add_to_slave(dev, slave, mac, port);
|
|
|
|
if (err)
|
|
|
|
__mlx4_unregister_mac(dev, port, mac);
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:21 +07:00
|
|
|
static int vlan_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
static int counter_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
u32 index;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
err = __mlx4_counter_alloc(dev, &index);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, index, 1, RES_COUNTER, 0);
|
|
|
|
if (err)
|
|
|
|
__mlx4_counter_free(dev, index);
|
|
|
|
else
|
|
|
|
set_param_l(out_param, index);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int xrcdn_alloc_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
u32 xrcdn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
err = __mlx4_xrcd_alloc(dev, &xrcdn);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, xrcdn, 1, RES_XRCD, 0);
|
|
|
|
if (err)
|
|
|
|
__mlx4_xrcd_free(dev, xrcdn);
|
|
|
|
else
|
|
|
|
set_param_l(out_param, xrcdn);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mlx4_ALLOC_RES_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int alop = vhcr->op_modifier;
|
|
|
|
|
|
|
|
switch (vhcr->in_modifier) {
|
|
|
|
case RES_QP:
|
|
|
|
err = qp_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MTT:
|
|
|
|
err = mtt_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT:
|
|
|
|
err = mpt_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_CQ:
|
|
|
|
err = cq_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_SRQ:
|
|
|
|
err = srq_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MAC:
|
|
|
|
err = mac_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
2011-12-13 11:16:21 +07:00
|
|
|
case RES_VLAN:
|
|
|
|
err = vlan_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
case RES_COUNTER:
|
|
|
|
err = counter_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_XRCD:
|
|
|
|
err = xrcdn_alloc_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int count;
|
|
|
|
int base;
|
|
|
|
int qpn;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE:
|
|
|
|
base = get_param_l(&in_param) & 0x7fffff;
|
|
|
|
count = get_param_h(&in_param);
|
|
|
|
err = rem_res_range(dev, slave, base, count, RES_QP, 0);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
__mlx4_qp_release_range(dev, base, count);
|
|
|
|
break;
|
|
|
|
case RES_OP_MAP_ICM:
|
|
|
|
qpn = get_param_l(&in_param) & 0x7fffff;
|
|
|
|
err = qp_res_start_move_to(dev, slave, qpn, RES_QP_RESERVED,
|
|
|
|
NULL, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
if (!fw_reserved(dev, qpn))
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
__mlx4_qp_free_icm(dev, qpn);
|
|
|
|
|
|
|
|
res_end_move(dev, slave, RES_QP, qpn);
|
|
|
|
|
|
|
|
if (valid_reserved(dev, slave, qpn))
|
|
|
|
err = rem_res_range(dev, slave, qpn, 1, RES_QP, 0);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mtt_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int err = -EINVAL;
|
|
|
|
int base;
|
|
|
|
int order;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE_AND_MAP)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
base = get_param_l(&in_param);
|
|
|
|
order = get_param_h(&in_param);
|
|
|
|
err = rem_res_range(dev, slave, base, 1, RES_MTT, order);
|
|
|
|
if (!err)
|
|
|
|
__mlx4_free_mtt_range(dev, base, order);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mpt_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param)
|
|
|
|
{
|
|
|
|
int err = -EINVAL;
|
|
|
|
int index;
|
|
|
|
int id;
|
|
|
|
struct res_mpt *mpt;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE:
|
|
|
|
index = get_param_l(&in_param);
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
err = get_res(dev, slave, id, RES_MPT, &mpt);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
index = mpt->key;
|
|
|
|
put_res(dev, slave, id, RES_MPT);
|
|
|
|
|
|
|
|
err = rem_res_range(dev, slave, id, 1, RES_MPT, 0);
|
|
|
|
if (err)
|
|
|
|
break;
|
2013-02-06 23:19:08 +07:00
|
|
|
__mlx4_mpt_release(dev, index);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
break;
|
|
|
|
case RES_OP_MAP_ICM:
|
|
|
|
index = get_param_l(&in_param);
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
err = mr_res_start_move_to(dev, slave, id,
|
|
|
|
RES_MPT_RESERVED, &mpt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2013-02-06 23:19:08 +07:00
|
|
|
__mlx4_mpt_free_icm(dev, mpt->key);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
res_end_move(dev, slave, RES_MPT, id);
|
|
|
|
return err;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cq_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int cqn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE_AND_MAP:
|
|
|
|
cqn = get_param_l(&in_param);
|
|
|
|
err = rem_res_range(dev, slave, cqn, 1, RES_CQ, 0);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
|
|
|
|
__mlx4_cq_free_icm(dev, cqn);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int srq_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int srqn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE_AND_MAP:
|
|
|
|
srqn = get_param_l(&in_param);
|
|
|
|
err = rem_res_range(dev, slave, srqn, 1, RES_SRQ, 0);
|
|
|
|
if (err)
|
|
|
|
break;
|
|
|
|
|
|
|
|
__mlx4_srq_free_icm(dev, srqn);
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mac_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int port;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case RES_OP_RESERVE_AND_MAP:
|
|
|
|
port = get_param_l(out_param);
|
|
|
|
mac_del_from_slave(dev, slave, in_param, port);
|
|
|
|
__mlx4_unregister_mac(dev, port, in_param);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
err = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:21 +07:00
|
|
|
static int vlan_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
static int counter_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int index;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
index = get_param_l(&in_param);
|
|
|
|
err = rem_res_range(dev, slave, index, 1, RES_COUNTER, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
__mlx4_counter_free(dev, index);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int xrcdn_free_res(struct mlx4_dev *dev, int slave, int op, int cmd,
|
|
|
|
u64 in_param, u64 *out_param)
|
|
|
|
{
|
|
|
|
int xrcdn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (op != RES_OP_RESERVE)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
xrcdn = get_param_l(&in_param);
|
|
|
|
err = rem_res_range(dev, slave, xrcdn, 1, RES_XRCD, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
__mlx4_xrcd_free(dev, xrcdn);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mlx4_FREE_RES_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)
|
|
|
|
{
|
|
|
|
int err = -EINVAL;
|
|
|
|
int alop = vhcr->op_modifier;
|
|
|
|
|
|
|
|
switch (vhcr->in_modifier) {
|
|
|
|
case RES_QP:
|
|
|
|
err = qp_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MTT:
|
|
|
|
err = mtt_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT:
|
|
|
|
err = mpt_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_CQ:
|
|
|
|
err = cq_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_SRQ:
|
|
|
|
err = srq_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MAC:
|
|
|
|
err = mac_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
2011-12-13 11:16:21 +07:00
|
|
|
case RES_VLAN:
|
|
|
|
err = vlan_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
case RES_COUNTER:
|
|
|
|
err = counter_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_XRCD:
|
|
|
|
err = xrcdn_free_res(dev, slave, vhcr->op_modifier, alop,
|
|
|
|
vhcr->in_param, &vhcr->out_param);
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* ugly but other choices are uglier */
|
|
|
|
static int mr_phys_mpt(struct mlx4_mpt_entry *mpt)
|
|
|
|
{
|
|
|
|
return (be32_to_cpu(mpt->flags) >> 9) & 1;
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
static int mr_get_mtt_addr(struct mlx4_mpt_entry *mpt)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
2011-12-13 11:16:56 +07:00
|
|
|
return (int)be64_to_cpu(mpt->mtt_addr) & 0xfffffff8;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static int mr_get_mtt_size(struct mlx4_mpt_entry *mpt)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(mpt->mtt_sz);
|
|
|
|
}
|
|
|
|
|
2013-02-06 23:19:10 +07:00
|
|
|
static u32 mr_get_pd(struct mlx4_mpt_entry *mpt)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(mpt->pd_flags) & 0x00ffffff;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mr_is_fmr(struct mlx4_mpt_entry *mpt)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(mpt->pd_flags) & MLX4_MPT_PD_FLAG_FAST_REG;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mr_is_bind_enabled(struct mlx4_mpt_entry *mpt)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(mpt->flags) & MLX4_MPT_FLAG_BIND_ENABLE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mr_is_region(struct mlx4_mpt_entry *mpt)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(mpt->flags) & MLX4_MPT_FLAG_REGION;
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
static int qp_get_mtt_addr(struct mlx4_qp_context *qpc)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
return be32_to_cpu(qpc->mtt_base_addr_l) & 0xfffffff8;
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
static int srq_get_mtt_addr(struct mlx4_srq_context *srqc)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
return be32_to_cpu(srqc->mtt_base_addr_l) & 0xfffffff8;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_get_mtt_size(struct mlx4_qp_context *qpc)
|
|
|
|
{
|
|
|
|
int page_shift = (qpc->log_page_size & 0x3f) + 12;
|
|
|
|
int log_sq_size = (qpc->sq_size_stride >> 3) & 0xf;
|
|
|
|
int log_sq_sride = qpc->sq_size_stride & 7;
|
|
|
|
int log_rq_size = (qpc->rq_size_stride >> 3) & 0xf;
|
|
|
|
int log_rq_stride = qpc->rq_size_stride & 7;
|
|
|
|
int srq = (be32_to_cpu(qpc->srqn) >> 24) & 1;
|
|
|
|
int rss = (be32_to_cpu(qpc->flags) >> 13) & 1;
|
|
|
|
int xrc = (be32_to_cpu(qpc->local_qpn) >> 23) & 1;
|
|
|
|
int sq_size;
|
|
|
|
int rq_size;
|
|
|
|
int total_pages;
|
|
|
|
int total_mem;
|
|
|
|
int page_offset = (be32_to_cpu(qpc->params2) >> 6) & 0x3f;
|
|
|
|
|
|
|
|
sq_size = 1 << (log_sq_size + log_sq_sride + 4);
|
|
|
|
rq_size = (srq|rss|xrc) ? 0 : (1 << (log_rq_size + log_rq_stride + 4));
|
|
|
|
total_mem = sq_size + rq_size;
|
|
|
|
total_pages =
|
|
|
|
roundup_pow_of_two((total_mem + (page_offset << 6)) >>
|
|
|
|
page_shift);
|
|
|
|
|
|
|
|
return total_pages;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int check_mtt_range(struct mlx4_dev *dev, int slave, int start,
|
|
|
|
int size, struct res_mtt *mtt)
|
|
|
|
{
|
2011-12-13 11:16:56 +07:00
|
|
|
int res_start = mtt->com.res_id;
|
|
|
|
int res_size = (1 << mtt->order);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
if (start < res_start || start + size > res_start + res_size)
|
|
|
|
return -EPERM;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_SW2HW_MPT_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int index = vhcr->in_modifier;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct res_mpt *mpt;
|
2011-12-13 11:16:56 +07:00
|
|
|
int mtt_base = mr_get_mtt_addr(inbox->buf) / dev->caps.mtt_entry_sz;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int phys;
|
|
|
|
int id;
|
2013-02-06 23:19:10 +07:00
|
|
|
u32 pd;
|
|
|
|
int pd_slave;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
err = mr_res_start_move_to(dev, slave, id, RES_MPT_HW, &mpt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2013-02-06 23:19:10 +07:00
|
|
|
/* Disable memory windows for VFs. */
|
|
|
|
if (!mr_is_region(inbox->buf)) {
|
|
|
|
err = -EPERM;
|
|
|
|
goto ex_abort;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Make sure that the PD bits related to the slave id are zeros. */
|
|
|
|
pd = mr_get_pd(inbox->buf);
|
|
|
|
pd_slave = (pd >> 17) & 0x7f;
|
|
|
|
if (pd_slave != 0 && pd_slave != slave) {
|
|
|
|
err = -EPERM;
|
|
|
|
goto ex_abort;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mr_is_fmr(inbox->buf)) {
|
|
|
|
/* FMR and Bind Enable are forbidden in slave devices. */
|
|
|
|
if (mr_is_bind_enabled(inbox->buf)) {
|
|
|
|
err = -EPERM;
|
|
|
|
goto ex_abort;
|
|
|
|
}
|
|
|
|
/* FMR and Memory Windows are also forbidden. */
|
|
|
|
if (!mr_is_region(inbox->buf)) {
|
|
|
|
err = -EPERM;
|
|
|
|
goto ex_abort;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
phys = mr_phys_mpt(inbox->buf);
|
|
|
|
if (!phys) {
|
2011-12-13 11:16:56 +07:00
|
|
|
err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
|
|
|
|
err = check_mtt_range(dev, slave, mtt_base,
|
|
|
|
mr_get_mtt_size(inbox->buf), mtt);
|
|
|
|
if (err)
|
|
|
|
goto ex_put;
|
|
|
|
|
|
|
|
mpt->mtt = mtt;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_put;
|
|
|
|
|
|
|
|
if (!phys) {
|
|
|
|
atomic_inc(&mtt->ref_count);
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
}
|
|
|
|
|
|
|
|
res_end_move(dev, slave, RES_MPT, id);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_put:
|
|
|
|
if (!phys)
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_MPT, id);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_HW2SW_MPT_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int index = vhcr->in_modifier;
|
|
|
|
struct res_mpt *mpt;
|
|
|
|
int id;
|
|
|
|
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
err = mr_res_start_move_to(dev, slave, id, RES_MPT_MAPPED, &mpt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
|
|
|
|
if (mpt->mtt)
|
|
|
|
atomic_dec(&mpt->mtt->ref_count);
|
|
|
|
|
|
|
|
res_end_move(dev, slave, RES_MPT, id);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_MPT, id);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_QUERY_MPT_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int index = vhcr->in_modifier;
|
|
|
|
struct res_mpt *mpt;
|
|
|
|
int id;
|
|
|
|
|
|
|
|
id = index & mpt_mask(dev);
|
|
|
|
err = get_res(dev, slave, id, RES_MPT, &mpt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (mpt->com.from_state != RES_MPT_HW) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
|
|
|
|
out:
|
|
|
|
put_res(dev, slave, id, RES_MPT);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_get_rcqn(struct mlx4_qp_context *qpc)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(qpc->cqn_recv) & 0xffffff;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_get_scqn(struct mlx4_qp_context *qpc)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(qpc->cqn_send) & 0xffffff;
|
|
|
|
}
|
|
|
|
|
|
|
|
static u32 qp_get_srqn(struct mlx4_qp_context *qpc)
|
|
|
|
{
|
|
|
|
return be32_to_cpu(qpc->srqn) & 0x1ffffff;
|
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
static void adjust_proxy_tun_qkey(struct mlx4_dev *dev, struct mlx4_vhcr *vhcr,
|
|
|
|
struct mlx4_qp_context *context)
|
|
|
|
{
|
|
|
|
u32 qpn = vhcr->in_modifier & 0xffffff;
|
|
|
|
u32 qkey = 0;
|
|
|
|
|
|
|
|
if (mlx4_get_parav_qkey(dev, qpn, &qkey))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* adjust qkey in qp context */
|
|
|
|
context->qkey = cpu_to_be32(qkey);
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mlx4_RST2INIT_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int qpn = vhcr->in_modifier & 0x7fffff;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct res_qp *qp;
|
|
|
|
struct mlx4_qp_context *qpc = inbox->buf + 8;
|
2011-12-13 11:16:56 +07:00
|
|
|
int mtt_base = qp_get_mtt_addr(qpc) / dev->caps.mtt_entry_sz;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mtt_size = qp_get_mtt_size(qpc);
|
|
|
|
struct res_cq *rcq;
|
|
|
|
struct res_cq *scq;
|
|
|
|
int rcqn = qp_get_rcqn(qpc);
|
|
|
|
int scqn = qp_get_scqn(qpc);
|
|
|
|
u32 srqn = qp_get_srqn(qpc) & 0xffffff;
|
|
|
|
int use_srq = (qp_get_srqn(qpc) >> 24) & 1;
|
|
|
|
struct res_srq *srq;
|
|
|
|
int local_qpn = be32_to_cpu(qpc->local_qpn) & 0xffffff;
|
|
|
|
|
|
|
|
err = qp_res_start_move_to(dev, slave, qpn, RES_QP_HW, &qp, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
qp->local_qpn = local_qpn;
|
2013-06-27 23:05:21 +07:00
|
|
|
qp->sched_queue = 0;
|
|
|
|
qp->qpc_flags = be32_to_cpu(qpc->flags);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
|
|
|
|
err = check_mtt_range(dev, slave, mtt_base, mtt_size, mtt);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_mtt;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, rcqn, RES_CQ, &rcq);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_mtt;
|
|
|
|
|
|
|
|
if (scqn != rcqn) {
|
|
|
|
err = get_res(dev, slave, scqn, RES_CQ, &scq);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_rcq;
|
|
|
|
} else
|
|
|
|
scq = rcq;
|
|
|
|
|
|
|
|
if (use_srq) {
|
|
|
|
err = get_res(dev, slave, srqn, RES_SRQ, &srq);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_scq;
|
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, qpc);
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_srq;
|
|
|
|
atomic_inc(&mtt->ref_count);
|
|
|
|
qp->mtt = mtt;
|
|
|
|
atomic_inc(&rcq->ref_count);
|
|
|
|
qp->rcq = rcq;
|
|
|
|
atomic_inc(&scq->ref_count);
|
|
|
|
qp->scq = scq;
|
|
|
|
|
|
|
|
if (scqn != rcqn)
|
|
|
|
put_res(dev, slave, scqn, RES_CQ);
|
|
|
|
|
|
|
|
if (use_srq) {
|
|
|
|
atomic_inc(&srq->ref_count);
|
|
|
|
put_res(dev, slave, srqn, RES_SRQ);
|
|
|
|
qp->srq = srq;
|
|
|
|
}
|
|
|
|
put_res(dev, slave, rcqn, RES_CQ);
|
2011-12-13 11:16:56 +07:00
|
|
|
put_res(dev, slave, mtt_base, RES_MTT);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
res_end_move(dev, slave, RES_QP, qpn);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_put_srq:
|
|
|
|
if (use_srq)
|
|
|
|
put_res(dev, slave, srqn, RES_SRQ);
|
|
|
|
ex_put_scq:
|
|
|
|
if (scqn != rcqn)
|
|
|
|
put_res(dev, slave, scqn, RES_CQ);
|
|
|
|
ex_put_rcq:
|
|
|
|
put_res(dev, slave, rcqn, RES_CQ);
|
|
|
|
ex_put_mtt:
|
2011-12-13 11:16:56 +07:00
|
|
|
put_res(dev, slave, mtt_base, RES_MTT);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_QP, qpn);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
static int eq_get_mtt_addr(struct mlx4_eq_context *eqc)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
return be32_to_cpu(eqc->mtt_base_addr_l) & 0xfffffff8;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int eq_get_mtt_size(struct mlx4_eq_context *eqc)
|
|
|
|
{
|
|
|
|
int log_eq_size = eqc->log_eq_size & 0x1f;
|
|
|
|
int page_shift = (eqc->log_page_size & 0x3f) + 12;
|
|
|
|
|
|
|
|
if (log_eq_size + 5 < page_shift)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 1 << (log_eq_size + 5 - page_shift);
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
static int cq_get_mtt_addr(struct mlx4_cq_context *cqc)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
return be32_to_cpu(cqc->mtt_base_addr_l) & 0xfffffff8;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cq_get_mtt_size(struct mlx4_cq_context *cqc)
|
|
|
|
{
|
|
|
|
int log_cq_size = (be32_to_cpu(cqc->logsize_usrpage) >> 24) & 0x1f;
|
|
|
|
int page_shift = (cqc->log_page_size & 0x3f) + 12;
|
|
|
|
|
|
|
|
if (log_cq_size + 5 < page_shift)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 1 << (log_cq_size + 5 - page_shift);
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_SW2HW_EQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int eqn = vhcr->in_modifier;
|
|
|
|
int res_id = (slave << 8) | eqn;
|
|
|
|
struct mlx4_eq_context *eqc = inbox->buf;
|
2011-12-13 11:16:56 +07:00
|
|
|
int mtt_base = eq_get_mtt_addr(eqc) / dev->caps.mtt_entry_sz;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mtt_size = eq_get_mtt_size(eqc);
|
|
|
|
struct res_eq *eq;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
|
|
|
|
err = add_res_range(dev, slave, res_id, 1, RES_EQ, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
err = eq_res_start_move_to(dev, slave, res_id, RES_EQ_HW, &eq);
|
|
|
|
if (err)
|
|
|
|
goto out_add;
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto out_move;
|
|
|
|
|
|
|
|
err = check_mtt_range(dev, slave, mtt_base, mtt_size, mtt);
|
|
|
|
if (err)
|
|
|
|
goto out_put;
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto out_put;
|
|
|
|
|
|
|
|
atomic_inc(&mtt->ref_count);
|
|
|
|
eq->mtt = mtt;
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
res_end_move(dev, slave, RES_EQ, res_id);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_put:
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
out_move:
|
|
|
|
res_abort_move(dev, slave, RES_EQ, res_id);
|
|
|
|
out_add:
|
|
|
|
rem_res_range(dev, slave, res_id, 1, RES_EQ, 0);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int get_containing_mtt(struct mlx4_dev *dev, int slave, int start,
|
|
|
|
int len, struct res_mtt **res)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
int err = -EINVAL;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry(mtt, &tracker->slave_list[slave].res_list[RES_MTT],
|
|
|
|
com.list) {
|
|
|
|
if (!check_mtt_range(dev, slave, start, len, mtt)) {
|
|
|
|
*res = mtt;
|
|
|
|
mtt->com.from_state = mtt->com.state;
|
|
|
|
mtt->com.state = RES_MTT_BUSY;
|
|
|
|
err = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
static int verify_qp_parameters(struct mlx4_dev *dev,
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
enum qp_transition transition, u8 slave)
|
|
|
|
{
|
|
|
|
u32 qp_type;
|
|
|
|
struct mlx4_qp_context *qp_ctx;
|
|
|
|
enum mlx4_qp_optpar optpar;
|
|
|
|
|
|
|
|
qp_ctx = inbox->buf + 8;
|
|
|
|
qp_type = (be32_to_cpu(qp_ctx->flags) >> 16) & 0xff;
|
|
|
|
optpar = be32_to_cpu(*(__be32 *) inbox->buf);
|
|
|
|
|
|
|
|
switch (qp_type) {
|
|
|
|
case MLX4_QP_ST_RC:
|
|
|
|
case MLX4_QP_ST_UC:
|
|
|
|
switch (transition) {
|
|
|
|
case QP_TRANS_INIT2RTR:
|
|
|
|
case QP_TRANS_RTR2RTS:
|
|
|
|
case QP_TRANS_RTS2RTS:
|
|
|
|
case QP_TRANS_SQD2SQD:
|
|
|
|
case QP_TRANS_SQD2RTS:
|
|
|
|
if (slave != mlx4_master_func_num(dev))
|
|
|
|
/* slaves have only gid index 0 */
|
|
|
|
if (optpar & MLX4_QP_OPTPAR_PRIMARY_ADDR_PATH)
|
|
|
|
if (qp_ctx->pri_path.mgid_index)
|
|
|
|
return -EINVAL;
|
|
|
|
if (optpar & MLX4_QP_OPTPAR_ALT_ADDR_PATH)
|
|
|
|
if (qp_ctx->alt_path.mgid_index)
|
|
|
|
return -EINVAL;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mlx4_WRITE_MTT_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_mtt mtt;
|
|
|
|
__be64 *page_list = inbox->buf;
|
|
|
|
u64 *pg_list = (u64 *)page_list;
|
|
|
|
int i;
|
|
|
|
struct res_mtt *rmtt = NULL;
|
|
|
|
int start = be64_to_cpu(page_list[0]);
|
|
|
|
int npages = vhcr->in_modifier;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = get_containing_mtt(dev, slave, start, npages, &rmtt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
/* Call the SW implementation of write_mtt:
|
|
|
|
* - Prepare a dummy mtt struct
|
|
|
|
* - Translate inbox contents to simple addresses in host endianess */
|
2011-12-13 11:16:56 +07:00
|
|
|
mtt.offset = 0; /* TBD this is broken but I don't handle it since
|
|
|
|
we don't really use it */
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
mtt.order = 0;
|
|
|
|
mtt.page_shift = 0;
|
|
|
|
for (i = 0; i < npages; ++i)
|
|
|
|
pg_list[i + 2] = (be64_to_cpu(page_list[i + 2]) & ~1ULL);
|
|
|
|
|
|
|
|
err = __mlx4_write_mtt(dev, &mtt, be64_to_cpu(page_list[0]), npages,
|
|
|
|
((u64 *)page_list + 2));
|
|
|
|
|
|
|
|
if (rmtt)
|
|
|
|
put_res(dev, slave, rmtt->com.res_id, RES_MTT);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_HW2SW_EQ_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)
|
|
|
|
{
|
|
|
|
int eqn = vhcr->in_modifier;
|
|
|
|
int res_id = eqn | (slave << 8);
|
|
|
|
struct res_eq *eq;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = eq_res_start_move_to(dev, slave, res_id, RES_EQ_RESERVED, &eq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, eq->mtt->com.res_id, RES_MTT, NULL);
|
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_put;
|
|
|
|
|
|
|
|
atomic_dec(&eq->mtt->ref_count);
|
|
|
|
put_res(dev, slave, eq->mtt->com.res_id, RES_MTT);
|
|
|
|
res_end_move(dev, slave, RES_EQ, res_id);
|
|
|
|
rem_res_range(dev, slave, res_id, 1, RES_EQ, 0);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_put:
|
|
|
|
put_res(dev, slave, eq->mtt->com.res_id, RES_MTT);
|
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_EQ, res_id);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_GEN_EQE(struct mlx4_dev *dev, int slave, struct mlx4_eqe *eqe)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_slave_event_eq_info *event_eq;
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
u32 in_modifier = 0;
|
|
|
|
int err;
|
|
|
|
int res_id;
|
|
|
|
struct res_eq *req;
|
|
|
|
|
|
|
|
if (!priv->mfunc.master.slave_state)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2012-01-19 16:45:46 +07:00
|
|
|
event_eq = &priv->mfunc.master.slave_state[slave].event_eq[eqe->type];
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
/* Create the event only if the slave is registered */
|
2012-01-19 16:45:46 +07:00
|
|
|
if (event_eq->eqn < 0)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
mutex_lock(&priv->mfunc.master.gen_eqe_mutex[slave]);
|
|
|
|
res_id = (slave << 8) | event_eq->eqn;
|
|
|
|
err = get_res(dev, slave, res_id, RES_EQ, &req);
|
|
|
|
if (err)
|
|
|
|
goto unlock;
|
|
|
|
|
|
|
|
if (req->com.from_state != RES_EQ_HW) {
|
|
|
|
err = -EINVAL;
|
|
|
|
goto put;
|
|
|
|
}
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
if (IS_ERR(mailbox)) {
|
|
|
|
err = PTR_ERR(mailbox);
|
|
|
|
goto put;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (eqe->type == MLX4_EVENT_TYPE_CMD) {
|
|
|
|
++event_eq->token;
|
|
|
|
eqe->event.cmd.token = cpu_to_be16(event_eq->token);
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(mailbox->buf, (u8 *) eqe, 28);
|
|
|
|
|
|
|
|
in_modifier = (slave & 0xff) | ((event_eq->eqn & 0xff) << 16);
|
|
|
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma, in_modifier, 0,
|
|
|
|
MLX4_CMD_GEN_EQE, MLX4_CMD_TIME_CLASS_B,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
put_res(dev, slave, res_id, RES_EQ);
|
|
|
|
mutex_unlock(&priv->mfunc.master.gen_eqe_mutex[slave]);
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
return err;
|
|
|
|
|
|
|
|
put:
|
|
|
|
put_res(dev, slave, res_id, RES_EQ);
|
|
|
|
|
|
|
|
unlock:
|
|
|
|
mutex_unlock(&priv->mfunc.master.gen_eqe_mutex[slave]);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_QUERY_EQ_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)
|
|
|
|
{
|
|
|
|
int eqn = vhcr->in_modifier;
|
|
|
|
int res_id = eqn | (slave << 8);
|
|
|
|
struct res_eq *eq;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, res_id, RES_EQ, &eq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (eq->com.from_state != RES_EQ_HW) {
|
|
|
|
err = -EINVAL;
|
|
|
|
goto ex_put;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
|
|
|
|
ex_put:
|
|
|
|
put_res(dev, slave, res_id, RES_EQ);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_SW2HW_CQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int cqn = vhcr->in_modifier;
|
|
|
|
struct mlx4_cq_context *cqc = inbox->buf;
|
2011-12-13 11:16:56 +07:00
|
|
|
int mtt_base = cq_get_mtt_addr(cqc) / dev->caps.mtt_entry_sz;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
struct res_cq *cq;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
|
|
|
|
err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_HW, &cq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2011-12-13 11:16:56 +07:00
|
|
|
err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto out_move;
|
|
|
|
err = check_mtt_range(dev, slave, mtt_base, cq_get_mtt_size(cqc), mtt);
|
|
|
|
if (err)
|
|
|
|
goto out_put;
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto out_put;
|
|
|
|
atomic_inc(&mtt->ref_count);
|
|
|
|
cq->mtt = mtt;
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
res_end_move(dev, slave, RES_CQ, cqn);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_put:
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
out_move:
|
|
|
|
res_abort_move(dev, slave, RES_CQ, cqn);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_HW2SW_CQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int cqn = vhcr->in_modifier;
|
|
|
|
struct res_cq *cq;
|
|
|
|
|
|
|
|
err = cq_res_start_move_to(dev, slave, cqn, RES_CQ_ALLOCATED, &cq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto out_move;
|
|
|
|
atomic_dec(&cq->mtt->ref_count);
|
|
|
|
res_end_move(dev, slave, RES_CQ, cqn);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_move:
|
|
|
|
res_abort_move(dev, slave, RES_CQ, cqn);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_QUERY_CQ_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)
|
|
|
|
{
|
|
|
|
int cqn = vhcr->in_modifier;
|
|
|
|
struct res_cq *cq;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, cqn, RES_CQ, &cq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (cq->com.from_state != RES_CQ_HW)
|
|
|
|
goto ex_put;
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
ex_put:
|
|
|
|
put_res(dev, slave, cqn, RES_CQ);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int handle_resize(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 res_cq *cq)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct res_mtt *orig_mtt;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct mlx4_cq_context *cqc = inbox->buf;
|
2011-12-13 11:16:56 +07:00
|
|
|
int mtt_base = cq_get_mtt_addr(cqc) / dev->caps.mtt_entry_sz;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
err = get_res(dev, slave, cq->mtt->com.res_id, RES_MTT, &orig_mtt);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (orig_mtt != cq->mtt) {
|
|
|
|
err = -EINVAL;
|
|
|
|
goto ex_put;
|
|
|
|
}
|
|
|
|
|
2011-12-13 11:16:56 +07:00
|
|
|
err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto ex_put;
|
|
|
|
|
|
|
|
err = check_mtt_range(dev, slave, mtt_base, cq_get_mtt_size(cqc), mtt);
|
|
|
|
if (err)
|
|
|
|
goto ex_put1;
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_put1;
|
|
|
|
atomic_dec(&orig_mtt->ref_count);
|
|
|
|
put_res(dev, slave, orig_mtt->com.res_id, RES_MTT);
|
|
|
|
atomic_inc(&mtt->ref_count);
|
|
|
|
cq->mtt = mtt;
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_put1:
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
ex_put:
|
|
|
|
put_res(dev, slave, orig_mtt->com.res_id, RES_MTT);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_MODIFY_CQ_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)
|
|
|
|
{
|
|
|
|
int cqn = vhcr->in_modifier;
|
|
|
|
struct res_cq *cq;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, cqn, RES_CQ, &cq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (cq->com.from_state != RES_CQ_HW)
|
|
|
|
goto ex_put;
|
|
|
|
|
|
|
|
if (vhcr->op_modifier == 0) {
|
|
|
|
err = handle_resize(dev, slave, vhcr, inbox, outbox, cmd, cq);
|
2012-03-07 12:56:35 +07:00
|
|
|
goto ex_put;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
ex_put:
|
|
|
|
put_res(dev, slave, cqn, RES_CQ);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int srq_get_mtt_size(struct mlx4_srq_context *srqc)
|
|
|
|
{
|
|
|
|
int log_srq_size = (be32_to_cpu(srqc->state_logsize_srqn) >> 24) & 0xf;
|
|
|
|
int log_rq_stride = srqc->logstride & 7;
|
|
|
|
int page_shift = (srqc->log_page_size & 0x3f) + 12;
|
|
|
|
|
|
|
|
if (log_srq_size + log_rq_stride + 4 < page_shift)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 1 << (log_srq_size + log_rq_stride + 4 - page_shift);
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_SW2HW_SRQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int srqn = vhcr->in_modifier;
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct res_srq *srq;
|
|
|
|
struct mlx4_srq_context *srqc = inbox->buf;
|
2011-12-13 11:16:56 +07:00
|
|
|
int mtt_base = srq_get_mtt_addr(srqc) / dev->caps.mtt_entry_sz;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
if (srqn != (be32_to_cpu(srqc->state_logsize_srqn) & 0xffffff))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_HW, &srq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2011-12-13 11:16:56 +07:00
|
|
|
err = get_res(dev, slave, mtt_base, RES_MTT, &mtt);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
err = check_mtt_range(dev, slave, mtt_base, srq_get_mtt_size(srqc),
|
|
|
|
mtt);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_mtt;
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_put_mtt;
|
|
|
|
|
|
|
|
atomic_inc(&mtt->ref_count);
|
|
|
|
srq->mtt = mtt;
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
res_end_move(dev, slave, RES_SRQ, srqn);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_put_mtt:
|
|
|
|
put_res(dev, slave, mtt->com.res_id, RES_MTT);
|
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_SRQ, srqn);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_HW2SW_SRQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int srqn = vhcr->in_modifier;
|
|
|
|
struct res_srq *srq;
|
|
|
|
|
|
|
|
err = srq_res_start_move_to(dev, slave, srqn, RES_SRQ_ALLOCATED, &srq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
atomic_dec(&srq->mtt->ref_count);
|
|
|
|
if (srq->cq)
|
|
|
|
atomic_dec(&srq->cq->ref_count);
|
|
|
|
res_end_move(dev, slave, RES_SRQ, srqn);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_SRQ, srqn);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_QUERY_SRQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int srqn = vhcr->in_modifier;
|
|
|
|
struct res_srq *srq;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, srqn, RES_SRQ, &srq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
if (srq->com.from_state != RES_SRQ_HW) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
out:
|
|
|
|
put_res(dev, slave, srqn, RES_SRQ);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_ARM_SRQ_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int srqn = vhcr->in_modifier;
|
|
|
|
struct res_srq *srq;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, srqn, RES_SRQ, &srq);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (srq->com.from_state != RES_SRQ_HW) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
out:
|
|
|
|
put_res(dev, slave, srqn, RES_SRQ);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_GEN_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int qpn = vhcr->in_modifier & 0x7fffff;
|
|
|
|
struct res_qp *qp;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, qpn, RES_QP, &qp);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
if (qp->com.from_state != RES_QP_HW) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
out:
|
|
|
|
put_res(dev, slave, qpn, RES_QP);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
int mlx4_INIT2INIT_QP_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_qp_context *context = inbox->buf + 8;
|
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, context);
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
|
|
|
return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mlx4_INIT2RTR_QP_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 15:40:43 +07:00
|
|
|
int err;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
struct mlx4_qp_context *qpc = inbox->buf + 8;
|
2013-06-27 23:05:21 +07:00
|
|
|
int qpn = vhcr->in_modifier & 0x7fffff;
|
|
|
|
struct res_qp *qp;
|
|
|
|
u8 orig_sched_queue;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
err = verify_qp_parameters(dev, inbox, QP_TRANS_INIT2RTR, slave);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
|
|
|
update_gid(dev, inbox, (u8)slave);
|
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, qpc);
|
2013-06-27 23:05:21 +07:00
|
|
|
orig_sched_queue = qpc->pri_path.sched_queue;
|
|
|
|
err = update_vport_qp_param(dev, inbox, slave, qpn);
|
2013-04-25 12:22:28 +07:00
|
|
|
if (err)
|
|
|
|
return err;
|
2012-08-03 15:40:43 +07:00
|
|
|
|
2013-06-27 23:05:21 +07:00
|
|
|
err = get_res(dev, slave, qpn, RES_QP, &qp);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
if (qp->com.from_state != RES_QP_HW) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
out:
|
|
|
|
/* if no error, save sched queue value passed in by VF. This is
|
|
|
|
* essentially the QOS value provided by the VF. This will be useful
|
|
|
|
* if we allow dynamic changes from VST back to VGT
|
|
|
|
*/
|
|
|
|
if (!err)
|
|
|
|
qp->sched_queue = orig_sched_queue;
|
|
|
|
|
|
|
|
put_res(dev, slave, qpn, RES_QP);
|
|
|
|
return err;
|
2012-08-03 15:40:43 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_RTR2RTS_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct mlx4_qp_context *context = inbox->buf + 8;
|
|
|
|
|
|
|
|
err = verify_qp_parameters(dev, inbox, QP_TRANS_RTR2RTS, slave);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
|
|
|
update_gid(dev, inbox, (u8)slave);
|
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, context);
|
|
|
|
return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_RTS2RTS_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct mlx4_qp_context *context = inbox->buf + 8;
|
|
|
|
|
|
|
|
err = verify_qp_parameters(dev, inbox, QP_TRANS_RTS2RTS, slave);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
|
|
|
update_gid(dev, inbox, (u8)slave);
|
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, context);
|
|
|
|
return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int mlx4_SQERR2RTS_QP_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_qp_context *context = inbox->buf + 8;
|
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, context);
|
|
|
|
return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_SQD2SQD_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct mlx4_qp_context *context = inbox->buf + 8;
|
|
|
|
|
|
|
|
err = verify_qp_parameters(dev, inbox, QP_TRANS_SQD2SQD, slave);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, context);
|
|
|
|
update_gid(dev, inbox, (u8)slave);
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
|
|
|
return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_SQD2RTS_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct mlx4_qp_context *context = inbox->buf + 8;
|
|
|
|
|
|
|
|
err = verify_qp_parameters(dev, inbox, QP_TRANS_SQD2RTS, slave);
|
|
|
|
if (err)
|
|
|
|
return err;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
2012-08-03 15:40:43 +07:00
|
|
|
adjust_proxy_tun_qkey(dev, vhcr, context);
|
|
|
|
update_gid(dev, inbox, (u8)slave);
|
|
|
|
update_pkey_index(dev, slave, inbox);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
return mlx4_GEN_QP_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_2RST_QP_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int qpn = vhcr->in_modifier & 0x7fffff;
|
|
|
|
struct res_qp *qp;
|
|
|
|
|
|
|
|
err = qp_res_start_move_to(dev, slave, qpn, RES_QP_MAPPED, &qp, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
if (err)
|
|
|
|
goto ex_abort;
|
|
|
|
|
|
|
|
atomic_dec(&qp->mtt->ref_count);
|
|
|
|
atomic_dec(&qp->rcq->ref_count);
|
|
|
|
atomic_dec(&qp->scq->ref_count);
|
|
|
|
if (qp->srq)
|
|
|
|
atomic_dec(&qp->srq->ref_count);
|
|
|
|
res_end_move(dev, slave, RES_QP, qpn);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ex_abort:
|
|
|
|
res_abort_move(dev, slave, RES_QP, qpn);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res_gid *find_gid(struct mlx4_dev *dev, int slave,
|
|
|
|
struct res_qp *rqp, u8 *gid)
|
|
|
|
{
|
|
|
|
struct res_gid *res;
|
|
|
|
|
|
|
|
list_for_each_entry(res, &rqp->mcg_list, list) {
|
|
|
|
if (!memcmp(res->gid, gid, 16))
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int add_mcg_res(struct mlx4_dev *dev, int slave, struct res_qp *rqp,
|
2012-02-15 13:23:16 +07:00
|
|
|
u8 *gid, enum mlx4_protocol prot,
|
2013-04-11 08:56:41 +07:00
|
|
|
enum mlx4_steer_type steer, u64 reg_id)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
struct res_gid *res;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
res = kzalloc(sizeof *res, GFP_KERNEL);
|
|
|
|
if (!res)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
spin_lock_irq(&rqp->mcg_spl);
|
|
|
|
if (find_gid(dev, slave, rqp, gid)) {
|
|
|
|
kfree(res);
|
|
|
|
err = -EEXIST;
|
|
|
|
} else {
|
|
|
|
memcpy(res->gid, gid, 16);
|
|
|
|
res->prot = prot;
|
2012-02-15 13:23:16 +07:00
|
|
|
res->steer = steer;
|
2013-04-11 08:56:41 +07:00
|
|
|
res->reg_id = reg_id;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_add_tail(&res->list, &rqp->mcg_list);
|
|
|
|
err = 0;
|
|
|
|
}
|
|
|
|
spin_unlock_irq(&rqp->mcg_spl);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int rem_mcg_res(struct mlx4_dev *dev, int slave, struct res_qp *rqp,
|
2012-02-15 13:23:16 +07:00
|
|
|
u8 *gid, enum mlx4_protocol prot,
|
2013-04-11 08:56:41 +07:00
|
|
|
enum mlx4_steer_type steer, u64 *reg_id)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
{
|
|
|
|
struct res_gid *res;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
spin_lock_irq(&rqp->mcg_spl);
|
|
|
|
res = find_gid(dev, slave, rqp, gid);
|
2012-02-15 13:23:16 +07:00
|
|
|
if (!res || res->prot != prot || res->steer != steer)
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
err = -EINVAL;
|
|
|
|
else {
|
2013-04-11 08:56:41 +07:00
|
|
|
*reg_id = res->reg_id;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&res->list);
|
|
|
|
kfree(res);
|
|
|
|
err = 0;
|
|
|
|
}
|
|
|
|
spin_unlock_irq(&rqp->mcg_spl);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2013-04-11 08:56:41 +07:00
|
|
|
static int qp_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
|
|
|
|
int block_loopback, enum mlx4_protocol prot,
|
|
|
|
enum mlx4_steer_type type, u64 *reg_id)
|
|
|
|
{
|
|
|
|
switch (dev->caps.steering_mode) {
|
|
|
|
case MLX4_STEERING_MODE_DEVICE_MANAGED:
|
|
|
|
return mlx4_trans_to_dmfs_attach(dev, qp, gid, gid[5],
|
|
|
|
block_loopback, prot,
|
|
|
|
reg_id);
|
|
|
|
case MLX4_STEERING_MODE_B0:
|
|
|
|
return mlx4_qp_attach_common(dev, qp, gid,
|
|
|
|
block_loopback, prot, type);
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int qp_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
|
|
|
|
enum mlx4_protocol prot, enum mlx4_steer_type type,
|
|
|
|
u64 reg_id)
|
|
|
|
{
|
|
|
|
switch (dev->caps.steering_mode) {
|
|
|
|
case MLX4_STEERING_MODE_DEVICE_MANAGED:
|
|
|
|
return mlx4_flow_detach(dev, reg_id);
|
|
|
|
case MLX4_STEERING_MODE_B0:
|
|
|
|
return mlx4_qp_detach_common(dev, qp, gid, prot, type);
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int mlx4_QP_ATTACH_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_qp qp; /* dummy for calling attach/detach */
|
|
|
|
u8 *gid = inbox->buf;
|
|
|
|
enum mlx4_protocol prot = (vhcr->in_modifier >> 28) & 0x7;
|
2012-05-15 17:34:57 +07:00
|
|
|
int err;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int qpn;
|
|
|
|
struct res_qp *rqp;
|
2013-04-11 08:56:41 +07:00
|
|
|
u64 reg_id = 0;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
int attach = vhcr->op_modifier;
|
|
|
|
int block_loopback = vhcr->in_modifier >> 31;
|
|
|
|
u8 steer_type_mask = 2;
|
2012-02-15 13:22:49 +07:00
|
|
|
enum mlx4_steer_type type = (gid[7] & steer_type_mask) >> 1;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
|
|
|
qpn = vhcr->in_modifier & 0xffffff;
|
|
|
|
err = get_res(dev, slave, qpn, RES_QP, &rqp);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
qp.qpn = qpn;
|
|
|
|
if (attach) {
|
2013-04-11 08:56:41 +07:00
|
|
|
err = qp_attach(dev, &qp, gid, block_loopback, prot,
|
|
|
|
type, ®_id);
|
|
|
|
if (err) {
|
|
|
|
pr_err("Fail to attach rule to qp 0x%x\n", qpn);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
goto ex_put;
|
2013-04-11 08:56:41 +07:00
|
|
|
}
|
|
|
|
err = add_mcg_res(dev, slave, rqp, gid, prot, type, reg_id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
2013-04-11 08:56:41 +07:00
|
|
|
goto ex_detach;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
} else {
|
2013-04-11 08:56:41 +07:00
|
|
|
err = rem_mcg_res(dev, slave, rqp, gid, prot, type, ®_id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
if (err)
|
|
|
|
goto ex_put;
|
|
|
|
|
2013-04-11 08:56:41 +07:00
|
|
|
err = qp_detach(dev, &qp, gid, prot, type, reg_id);
|
|
|
|
if (err)
|
|
|
|
pr_err("Fail to detach rule from qp 0x%x reg_id = 0x%llx\n",
|
|
|
|
qpn, reg_id);
|
|
|
|
}
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
put_res(dev, slave, qpn, RES_QP);
|
2013-04-11 08:56:41 +07:00
|
|
|
return err;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
|
2013-04-11 08:56:41 +07:00
|
|
|
ex_detach:
|
|
|
|
qp_detach(dev, &qp, gid, prot, type, reg_id);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
ex_put:
|
|
|
|
put_res(dev, slave, qpn, RES_QP);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2012-09-06 05:50:49 +07:00
|
|
|
/*
|
|
|
|
* MAC validation for Flow Steering rules.
|
|
|
|
* VF can attach rules only with a mac address which is assigned to it.
|
|
|
|
*/
|
|
|
|
static int validate_eth_header_mac(int slave, struct _rule_hw *eth_header,
|
|
|
|
struct list_head *rlist)
|
|
|
|
{
|
|
|
|
struct mac_res *res, *tmp;
|
|
|
|
__be64 be_mac;
|
|
|
|
|
|
|
|
/* make sure it isn't multicast or broadcast mac*/
|
|
|
|
if (!is_multicast_ether_addr(eth_header->eth.dst_mac) &&
|
|
|
|
!is_broadcast_ether_addr(eth_header->eth.dst_mac)) {
|
|
|
|
list_for_each_entry_safe(res, tmp, rlist, list) {
|
|
|
|
be_mac = cpu_to_be64(res->mac << 16);
|
|
|
|
if (!memcmp(&be_mac, eth_header->eth.dst_mac, ETH_ALEN))
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
pr_err("MAC %pM doesn't belong to VF %d, Steering rule rejected\n",
|
|
|
|
eth_header->eth.dst_mac, slave);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* In case of missing eth header, append eth header with a MAC address
|
|
|
|
* assigned to the VF.
|
|
|
|
*/
|
|
|
|
static int add_eth_header(struct mlx4_dev *dev, int slave,
|
|
|
|
struct mlx4_cmd_mailbox *inbox,
|
|
|
|
struct list_head *rlist, int header_id)
|
|
|
|
{
|
|
|
|
struct mac_res *res, *tmp;
|
|
|
|
u8 port;
|
|
|
|
struct mlx4_net_trans_rule_hw_ctrl *ctrl;
|
|
|
|
struct mlx4_net_trans_rule_hw_eth *eth_header;
|
|
|
|
struct mlx4_net_trans_rule_hw_ipv4 *ip_header;
|
|
|
|
struct mlx4_net_trans_rule_hw_tcp_udp *l4_header;
|
|
|
|
__be64 be_mac = 0;
|
|
|
|
__be64 mac_msk = cpu_to_be64(MLX4_MAC_MASK << 16);
|
|
|
|
|
|
|
|
ctrl = (struct mlx4_net_trans_rule_hw_ctrl *)inbox->buf;
|
2013-01-31 06:07:02 +07:00
|
|
|
port = ctrl->port;
|
2012-09-06 05:50:49 +07:00
|
|
|
eth_header = (struct mlx4_net_trans_rule_hw_eth *)(ctrl + 1);
|
|
|
|
|
|
|
|
/* Clear a space in the inbox for eth header */
|
|
|
|
switch (header_id) {
|
|
|
|
case MLX4_NET_TRANS_RULE_ID_IPV4:
|
|
|
|
ip_header =
|
|
|
|
(struct mlx4_net_trans_rule_hw_ipv4 *)(eth_header + 1);
|
|
|
|
memmove(ip_header, eth_header,
|
|
|
|
sizeof(*ip_header) + sizeof(*l4_header));
|
|
|
|
break;
|
|
|
|
case MLX4_NET_TRANS_RULE_ID_TCP:
|
|
|
|
case MLX4_NET_TRANS_RULE_ID_UDP:
|
|
|
|
l4_header = (struct mlx4_net_trans_rule_hw_tcp_udp *)
|
|
|
|
(eth_header + 1);
|
|
|
|
memmove(l4_header, eth_header, sizeof(*l4_header));
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
list_for_each_entry_safe(res, tmp, rlist, list) {
|
|
|
|
if (port == res->port) {
|
|
|
|
be_mac = cpu_to_be64(res->mac << 16);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!be_mac) {
|
|
|
|
pr_err("Failed adding eth header to FS rule, Can't find matching MAC for port %d .\n",
|
|
|
|
port);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(eth_header, 0, sizeof(*eth_header));
|
|
|
|
eth_header->size = sizeof(*eth_header) >> 2;
|
|
|
|
eth_header->id = cpu_to_be16(__sw_id_hw[MLX4_NET_TRANS_RULE_ID_ETH]);
|
|
|
|
memcpy(eth_header->dst_mac, &be_mac, ETH_ALEN);
|
|
|
|
memcpy(eth_header->dst_mac_msk, &mac_msk, ETH_ALEN);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:45 +07:00
|
|
|
int mlx4_QP_FLOW_STEERING_ATTACH_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)
|
|
|
|
{
|
2012-09-06 05:50:49 +07:00
|
|
|
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *rlist = &tracker->slave_list[slave].res_list[RES_MAC];
|
2012-07-05 11:03:47 +07:00
|
|
|
int err;
|
2012-12-07 00:11:57 +07:00
|
|
|
int qpn;
|
2013-03-21 12:55:55 +07:00
|
|
|
struct res_qp *rqp;
|
2012-09-06 05:50:49 +07:00
|
|
|
struct mlx4_net_trans_rule_hw_ctrl *ctrl;
|
|
|
|
struct _rule_hw *rule_header;
|
|
|
|
int header_id;
|
2012-07-05 11:03:47 +07:00
|
|
|
|
{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
|
|
|
if (dev->caps.steering_mode !=
|
|
|
|
MLX4_STEERING_MODE_DEVICE_MANAGED)
|
|
|
|
return -EOPNOTSUPP;
|
2012-07-05 11:03:47 +07:00
|
|
|
|
2012-09-06 05:50:49 +07:00
|
|
|
ctrl = (struct mlx4_net_trans_rule_hw_ctrl *)inbox->buf;
|
2012-12-07 00:11:57 +07:00
|
|
|
qpn = be32_to_cpu(ctrl->qpn) & 0xffffff;
|
2013-03-21 12:55:55 +07:00
|
|
|
err = get_res(dev, slave, qpn, RES_QP, &rqp);
|
2012-12-07 00:11:57 +07:00
|
|
|
if (err) {
|
|
|
|
pr_err("Steering rule with qpn 0x%x rejected.\n", qpn);
|
|
|
|
return err;
|
|
|
|
}
|
2012-09-06 05:50:49 +07:00
|
|
|
rule_header = (struct _rule_hw *)(ctrl + 1);
|
|
|
|
header_id = map_hw_to_sw_id(be16_to_cpu(rule_header->id));
|
|
|
|
|
|
|
|
switch (header_id) {
|
|
|
|
case MLX4_NET_TRANS_RULE_ID_ETH:
|
2012-12-07 00:11:57 +07:00
|
|
|
if (validate_eth_header_mac(slave, rule_header, rlist)) {
|
|
|
|
err = -EINVAL;
|
|
|
|
goto err_put;
|
|
|
|
}
|
2012-09-06 05:50:49 +07:00
|
|
|
break;
|
2012-10-03 22:38:48 +07:00
|
|
|
case MLX4_NET_TRANS_RULE_ID_IB:
|
|
|
|
break;
|
2012-09-06 05:50:49 +07:00
|
|
|
case MLX4_NET_TRANS_RULE_ID_IPV4:
|
|
|
|
case MLX4_NET_TRANS_RULE_ID_TCP:
|
|
|
|
case MLX4_NET_TRANS_RULE_ID_UDP:
|
|
|
|
pr_warn("Can't attach FS rule without L2 headers, adding L2 header.\n");
|
2012-12-07 00:11:57 +07:00
|
|
|
if (add_eth_header(dev, slave, inbox, rlist, header_id)) {
|
|
|
|
err = -EINVAL;
|
|
|
|
goto err_put;
|
|
|
|
}
|
2012-09-06 05:50:49 +07:00
|
|
|
vhcr->in_modifier +=
|
|
|
|
sizeof(struct mlx4_net_trans_rule_hw_eth) >> 2;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
pr_err("Corrupted mailbox.\n");
|
2012-12-07 00:11:57 +07:00
|
|
|
err = -EINVAL;
|
|
|
|
goto err_put;
|
2012-09-06 05:50:49 +07:00
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:47 +07:00
|
|
|
err = mlx4_cmd_imm(dev, inbox->dma, &vhcr->out_param,
|
|
|
|
vhcr->in_modifier, 0,
|
|
|
|
MLX4_QP_FLOW_STEERING_ATTACH, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
if (err)
|
2012-12-07 00:11:57 +07:00
|
|
|
goto err_put;
|
2012-07-05 11:03:47 +07:00
|
|
|
|
2013-03-21 12:55:55 +07:00
|
|
|
err = add_res_range(dev, slave, vhcr->out_param, 1, RES_FS_RULE, qpn);
|
2012-07-05 11:03:47 +07:00
|
|
|
if (err) {
|
|
|
|
mlx4_err(dev, "Fail to add flow steering resources.\n ");
|
|
|
|
/* detach rule*/
|
|
|
|
mlx4_cmd(dev, vhcr->out_param, 0, 0,
|
2012-12-07 00:11:58 +07:00
|
|
|
MLX4_QP_FLOW_STEERING_DETACH, MLX4_CMD_TIME_CLASS_A,
|
2012-07-05 11:03:47 +07:00
|
|
|
MLX4_CMD_NATIVE);
|
2013-03-21 12:55:55 +07:00
|
|
|
goto err_put;
|
2012-07-05 11:03:47 +07:00
|
|
|
}
|
2013-03-21 12:55:55 +07:00
|
|
|
atomic_inc(&rqp->ref_count);
|
2012-12-07 00:11:57 +07:00
|
|
|
err_put:
|
|
|
|
put_res(dev, slave, qpn, RES_QP);
|
2012-07-05 11:03:47 +07:00
|
|
|
return err;
|
2012-07-05 11:03:45 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int mlx4_QP_FLOW_STEERING_DETACH_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)
|
|
|
|
{
|
2012-07-05 11:03:47 +07:00
|
|
|
int err;
|
2013-03-21 12:55:55 +07:00
|
|
|
struct res_qp *rqp;
|
|
|
|
struct res_fs_rule *rrule;
|
2012-07-05 11:03:47 +07:00
|
|
|
|
{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
|
|
|
if (dev->caps.steering_mode !=
|
|
|
|
MLX4_STEERING_MODE_DEVICE_MANAGED)
|
|
|
|
return -EOPNOTSUPP;
|
2012-07-05 11:03:47 +07:00
|
|
|
|
2013-03-21 12:55:55 +07:00
|
|
|
err = get_res(dev, slave, vhcr->in_param, RES_FS_RULE, &rrule);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
/* Release the rule form busy state before removal */
|
|
|
|
put_res(dev, slave, vhcr->in_param, RES_FS_RULE);
|
|
|
|
err = get_res(dev, slave, rrule->qpn, RES_QP, &rqp);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2012-07-05 11:03:47 +07:00
|
|
|
err = rem_res_range(dev, slave, vhcr->in_param, 1, RES_FS_RULE, 0);
|
|
|
|
if (err) {
|
|
|
|
mlx4_err(dev, "Fail to remove flow steering resources.\n ");
|
2013-03-21 12:55:55 +07:00
|
|
|
goto out;
|
2012-07-05 11:03:47 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
err = mlx4_cmd(dev, vhcr->in_param, 0, 0,
|
|
|
|
MLX4_QP_FLOW_STEERING_DETACH, MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
2013-03-21 12:55:55 +07:00
|
|
|
if (!err)
|
|
|
|
atomic_dec(&rqp->ref_count);
|
|
|
|
out:
|
|
|
|
put_res(dev, slave, rrule->qpn, RES_QP);
|
2012-07-05 11:03:47 +07:00
|
|
|
return err;
|
2012-07-05 11:03:45 +07:00
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
enum {
|
|
|
|
BUSY_MAX_RETRIES = 10
|
|
|
|
};
|
|
|
|
|
|
|
|
int mlx4_QUERY_IF_STAT_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)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
int index = vhcr->in_modifier & 0xffff;
|
|
|
|
|
|
|
|
err = get_res(dev, slave, index, RES_COUNTER, NULL);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
|
|
|
|
put_res(dev, slave, index, RES_COUNTER);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void detach_qp(struct mlx4_dev *dev, int slave, struct res_qp *rqp)
|
|
|
|
{
|
|
|
|
struct res_gid *rgid;
|
|
|
|
struct res_gid *tmp;
|
|
|
|
struct mlx4_qp qp; /* dummy for calling attach/detach */
|
|
|
|
|
|
|
|
list_for_each_entry_safe(rgid, tmp, &rqp->mcg_list, list) {
|
2013-04-11 08:56:41 +07:00
|
|
|
switch (dev->caps.steering_mode) {
|
|
|
|
case MLX4_STEERING_MODE_DEVICE_MANAGED:
|
|
|
|
mlx4_flow_detach(dev, rgid->reg_id);
|
|
|
|
break;
|
|
|
|
case MLX4_STEERING_MODE_B0:
|
|
|
|
qp.qpn = rqp->local_qpn;
|
|
|
|
(void) mlx4_qp_detach_common(dev, &qp, rgid->gid,
|
|
|
|
rgid->prot, rgid->steer);
|
|
|
|
break;
|
|
|
|
}
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&rgid->list);
|
|
|
|
kfree(rgid);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int _move_all_busy(struct mlx4_dev *dev, int slave,
|
|
|
|
enum mlx4_resource type, int print)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker =
|
|
|
|
&priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *rlist = &tracker->slave_list[slave].res_list[type];
|
|
|
|
struct res_common *r;
|
|
|
|
struct res_common *tmp;
|
|
|
|
int busy;
|
|
|
|
|
|
|
|
busy = 0;
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(r, tmp, rlist, list) {
|
|
|
|
if (r->owner == slave) {
|
|
|
|
if (!r->removing) {
|
|
|
|
if (r->state == RES_ANY_BUSY) {
|
|
|
|
if (print)
|
|
|
|
mlx4_dbg(dev,
|
2012-07-05 11:03:42 +07:00
|
|
|
"%s id 0x%llx is busy\n",
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
ResourceType(type),
|
|
|
|
r->res_id);
|
|
|
|
++busy;
|
|
|
|
} else {
|
|
|
|
r->from_state = r->state;
|
|
|
|
r->state = RES_ANY_BUSY;
|
|
|
|
r->removing = 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
|
|
|
|
return busy;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int move_all_busy(struct mlx4_dev *dev, int slave,
|
|
|
|
enum mlx4_resource type)
|
|
|
|
{
|
|
|
|
unsigned long begin;
|
|
|
|
int busy;
|
|
|
|
|
|
|
|
begin = jiffies;
|
|
|
|
do {
|
|
|
|
busy = _move_all_busy(dev, slave, type, 0);
|
|
|
|
if (time_after(jiffies, begin + 5 * HZ))
|
|
|
|
break;
|
|
|
|
if (busy)
|
|
|
|
cond_resched();
|
|
|
|
} while (busy);
|
|
|
|
|
|
|
|
if (busy)
|
|
|
|
busy = _move_all_busy(dev, slave, type, 1);
|
|
|
|
|
|
|
|
return busy;
|
|
|
|
}
|
|
|
|
static void rem_slave_qps(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *qp_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_QP];
|
|
|
|
struct res_qp *qp;
|
|
|
|
struct res_qp *tmp;
|
|
|
|
int state;
|
|
|
|
u64 in_param;
|
|
|
|
int qpn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_QP);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_qps: Could not move all qps to busy"
|
|
|
|
"for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(qp, tmp, qp_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (qp->com.owner == slave) {
|
|
|
|
qpn = qp->com.res_id;
|
|
|
|
detach_qp(dev, slave, qp);
|
|
|
|
state = qp->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_QP_RESERVED:
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&qp->com.node,
|
|
|
|
&tracker->res_tree[RES_QP]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&qp->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(qp);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
case RES_QP_MAPPED:
|
|
|
|
if (!valid_reserved(dev, slave, qpn))
|
|
|
|
__mlx4_qp_free_icm(dev, qpn);
|
|
|
|
state = RES_QP_RESERVED;
|
|
|
|
break;
|
|
|
|
case RES_QP_HW:
|
|
|
|
in_param = slave;
|
|
|
|
err = mlx4_cmd(dev, in_param,
|
|
|
|
qp->local_qpn, 2,
|
|
|
|
MLX4_CMD_2RST_QP,
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
if (err)
|
|
|
|
mlx4_dbg(dev, "rem_slave_qps: failed"
|
|
|
|
" to move slave %d qpn %d to"
|
|
|
|
" reset\n", slave,
|
|
|
|
qp->local_qpn);
|
|
|
|
atomic_dec(&qp->rcq->ref_count);
|
|
|
|
atomic_dec(&qp->scq->ref_count);
|
|
|
|
atomic_dec(&qp->mtt->ref_count);
|
|
|
|
if (qp->srq)
|
|
|
|
atomic_dec(&qp->srq->ref_count);
|
|
|
|
state = RES_QP_MAPPED;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rem_slave_srqs(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *srq_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_SRQ];
|
|
|
|
struct res_srq *srq;
|
|
|
|
struct res_srq *tmp;
|
|
|
|
int state;
|
|
|
|
u64 in_param;
|
|
|
|
LIST_HEAD(tlist);
|
|
|
|
int srqn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_SRQ);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_srqs: Could not move all srqs to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(srq, tmp, srq_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (srq->com.owner == slave) {
|
|
|
|
srqn = srq->com.res_id;
|
|
|
|
state = srq->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_SRQ_ALLOCATED:
|
|
|
|
__mlx4_srq_free_icm(dev, srqn);
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&srq->com.node,
|
|
|
|
&tracker->res_tree[RES_SRQ]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&srq->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(srq);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_SRQ_HW:
|
|
|
|
in_param = slave;
|
|
|
|
err = mlx4_cmd(dev, in_param, srqn, 1,
|
|
|
|
MLX4_CMD_HW2SW_SRQ,
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
if (err)
|
|
|
|
mlx4_dbg(dev, "rem_slave_srqs: failed"
|
|
|
|
" to move slave %d srq %d to"
|
|
|
|
" SW ownership\n",
|
|
|
|
slave, srqn);
|
|
|
|
|
|
|
|
atomic_dec(&srq->mtt->ref_count);
|
|
|
|
if (srq->cq)
|
|
|
|
atomic_dec(&srq->cq->ref_count);
|
|
|
|
state = RES_SRQ_ALLOCATED;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rem_slave_cqs(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *cq_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_CQ];
|
|
|
|
struct res_cq *cq;
|
|
|
|
struct res_cq *tmp;
|
|
|
|
int state;
|
|
|
|
u64 in_param;
|
|
|
|
LIST_HEAD(tlist);
|
|
|
|
int cqn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_CQ);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_cqs: Could not move all cqs to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(cq, tmp, cq_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (cq->com.owner == slave && !atomic_read(&cq->ref_count)) {
|
|
|
|
cqn = cq->com.res_id;
|
|
|
|
state = cq->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_CQ_ALLOCATED:
|
|
|
|
__mlx4_cq_free_icm(dev, cqn);
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&cq->com.node,
|
|
|
|
&tracker->res_tree[RES_CQ]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&cq->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(cq);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_CQ_HW:
|
|
|
|
in_param = slave;
|
|
|
|
err = mlx4_cmd(dev, in_param, cqn, 1,
|
|
|
|
MLX4_CMD_HW2SW_CQ,
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
if (err)
|
|
|
|
mlx4_dbg(dev, "rem_slave_cqs: failed"
|
|
|
|
" to move slave %d cq %d to"
|
|
|
|
" SW ownership\n",
|
|
|
|
slave, cqn);
|
|
|
|
atomic_dec(&cq->mtt->ref_count);
|
|
|
|
state = RES_CQ_ALLOCATED;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rem_slave_mrs(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *mpt_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_MPT];
|
|
|
|
struct res_mpt *mpt;
|
|
|
|
struct res_mpt *tmp;
|
|
|
|
int state;
|
|
|
|
u64 in_param;
|
|
|
|
LIST_HEAD(tlist);
|
|
|
|
int mptn;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_MPT);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_mrs: Could not move all mpts to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(mpt, tmp, mpt_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (mpt->com.owner == slave) {
|
|
|
|
mptn = mpt->com.res_id;
|
|
|
|
state = mpt->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_MPT_RESERVED:
|
2013-02-06 23:19:08 +07:00
|
|
|
__mlx4_mpt_release(dev, mpt->key);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&mpt->com.node,
|
|
|
|
&tracker->res_tree[RES_MPT]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&mpt->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(mpt);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT_MAPPED:
|
2013-02-06 23:19:08 +07:00
|
|
|
__mlx4_mpt_free_icm(dev, mpt->key);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
state = RES_MPT_RESERVED;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_MPT_HW:
|
|
|
|
in_param = slave;
|
|
|
|
err = mlx4_cmd(dev, in_param, mptn, 0,
|
|
|
|
MLX4_CMD_HW2SW_MPT,
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
if (err)
|
|
|
|
mlx4_dbg(dev, "rem_slave_mrs: failed"
|
|
|
|
" to move slave %d mpt %d to"
|
|
|
|
" SW ownership\n",
|
|
|
|
slave, mptn);
|
|
|
|
if (mpt->mtt)
|
|
|
|
atomic_dec(&mpt->mtt->ref_count);
|
|
|
|
state = RES_MPT_MAPPED;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rem_slave_mtts(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker =
|
|
|
|
&priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *mtt_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_MTT];
|
|
|
|
struct res_mtt *mtt;
|
|
|
|
struct res_mtt *tmp;
|
|
|
|
int state;
|
|
|
|
LIST_HEAD(tlist);
|
|
|
|
int base;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_MTT);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_mtts: Could not move all mtts to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(mtt, tmp, mtt_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (mtt->com.owner == slave) {
|
|
|
|
base = mtt->com.res_id;
|
|
|
|
state = mtt->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_MTT_ALLOCATED:
|
|
|
|
__mlx4_free_mtt_range(dev, base,
|
|
|
|
mtt->order);
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&mtt->com.node,
|
|
|
|
&tracker->res_tree[RES_MTT]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&mtt->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(mtt);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
2012-07-05 11:03:47 +07:00
|
|
|
static void rem_slave_fs_rule(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker =
|
|
|
|
&priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *fs_rule_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_FS_RULE];
|
|
|
|
struct res_fs_rule *fs_rule;
|
|
|
|
struct res_fs_rule *tmp;
|
|
|
|
int state;
|
|
|
|
u64 base;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_FS_RULE);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_fs_rule: Could not move all mtts to busy for slave %d\n",
|
|
|
|
slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(fs_rule, tmp, fs_rule_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (fs_rule->com.owner == slave) {
|
|
|
|
base = fs_rule->com.res_id;
|
|
|
|
state = fs_rule->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_FS_RULE_ALLOCATED:
|
|
|
|
/* detach rule */
|
|
|
|
err = mlx4_cmd(dev, base, 0, 0,
|
|
|
|
MLX4_QP_FLOW_STEERING_DETACH,
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
rb_erase(&fs_rule->com.node,
|
|
|
|
&tracker->res_tree[RES_FS_RULE]);
|
|
|
|
list_del(&fs_rule->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(fs_rule);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
static void rem_slave_eqs(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *eq_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_EQ];
|
|
|
|
struct res_eq *eq;
|
|
|
|
struct res_eq *tmp;
|
|
|
|
int err;
|
|
|
|
int state;
|
|
|
|
LIST_HEAD(tlist);
|
|
|
|
int eqn;
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_EQ);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_eqs: Could not move all eqs to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(eq, tmp, eq_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (eq->com.owner == slave) {
|
|
|
|
eqn = eq->com.res_id;
|
|
|
|
state = eq->com.from_state;
|
|
|
|
while (state != 0) {
|
|
|
|
switch (state) {
|
|
|
|
case RES_EQ_RESERVED:
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&eq->com.node,
|
|
|
|
&tracker->res_tree[RES_EQ]);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
list_del(&eq->com.list);
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
kfree(eq);
|
|
|
|
state = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RES_EQ_HW:
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
if (IS_ERR(mailbox)) {
|
|
|
|
cond_resched();
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
err = mlx4_cmd_box(dev, slave, 0,
|
|
|
|
eqn & 0xff, 0,
|
|
|
|
MLX4_CMD_HW2SW_EQ,
|
|
|
|
MLX4_CMD_TIME_CLASS_A,
|
|
|
|
MLX4_CMD_NATIVE);
|
2012-05-15 17:35:04 +07:00
|
|
|
if (err)
|
|
|
|
mlx4_dbg(dev, "rem_slave_eqs: failed"
|
|
|
|
" to move slave %d eqs %d to"
|
|
|
|
" SW ownership\n", slave, eqn);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
2012-05-15 17:35:04 +07:00
|
|
|
atomic_dec(&eq->mtt->ref_count);
|
|
|
|
state = RES_EQ_RESERVED;
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
state = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
2012-05-15 17:35:03 +07:00
|
|
|
static void rem_slave_counters(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *counter_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_COUNTER];
|
|
|
|
struct res_counter *counter;
|
|
|
|
struct res_counter *tmp;
|
|
|
|
int err;
|
|
|
|
int index;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_COUNTER);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_counters: Could not move all counters to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(counter, tmp, counter_list, com.list) {
|
|
|
|
if (counter->com.owner == slave) {
|
|
|
|
index = counter->com.res_id;
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&counter->com.node,
|
|
|
|
&tracker->res_tree[RES_COUNTER]);
|
2012-05-15 17:35:03 +07:00
|
|
|
list_del(&counter->com.list);
|
|
|
|
kfree(counter);
|
|
|
|
__mlx4_counter_free(dev, index);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void rem_slave_xrcdns(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
struct mlx4_resource_tracker *tracker = &priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *xrcdn_list =
|
|
|
|
&tracker->slave_list[slave].res_list[RES_XRCD];
|
|
|
|
struct res_xrcdn *xrcd;
|
|
|
|
struct res_xrcdn *tmp;
|
|
|
|
int err;
|
|
|
|
int xrcdn;
|
|
|
|
|
|
|
|
err = move_all_busy(dev, slave, RES_XRCD);
|
|
|
|
if (err)
|
|
|
|
mlx4_warn(dev, "rem_slave_xrcdns: Could not move all xrcdns to "
|
|
|
|
"busy for slave %d\n", slave);
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(xrcd, tmp, xrcdn_list, com.list) {
|
|
|
|
if (xrcd->com.owner == slave) {
|
|
|
|
xrcdn = xrcd->com.res_id;
|
2012-07-05 11:03:41 +07:00
|
|
|
rb_erase(&xrcd->com.node, &tracker->res_tree[RES_XRCD]);
|
2012-05-15 17:35:03 +07:00
|
|
|
list_del(&xrcd->com.list);
|
|
|
|
kfree(xrcd);
|
|
|
|
__mlx4_xrcd_free(dev, xrcdn);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
void mlx4_delete_all_resources_for_slave(struct mlx4_dev *dev, int slave)
|
|
|
|
{
|
|
|
|
struct mlx4_priv *priv = mlx4_priv(dev);
|
|
|
|
|
|
|
|
mutex_lock(&priv->mfunc.master.res_tracker.slave_list[slave].mutex);
|
|
|
|
/*VLAN*/
|
|
|
|
rem_slave_macs(dev, slave);
|
2013-03-21 12:55:52 +07:00
|
|
|
rem_slave_fs_rule(dev, slave);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
rem_slave_qps(dev, slave);
|
|
|
|
rem_slave_srqs(dev, slave);
|
|
|
|
rem_slave_cqs(dev, slave);
|
|
|
|
rem_slave_mrs(dev, slave);
|
|
|
|
rem_slave_eqs(dev, slave);
|
|
|
|
rem_slave_mtts(dev, slave);
|
2012-05-15 17:35:03 +07:00
|
|
|
rem_slave_counters(dev, slave);
|
|
|
|
rem_slave_xrcdns(dev, slave);
|
mlx4_core: resource tracking for HCA resources used by guests
The resource tracker is used to track usage of HCA resources by the different
guests.
Virtual functions (VFs) are attached to guest operating systems but
resources are allocated from the same pool and are assigned to VFs. It is
essential that hostile/buggy guests not be able to affect the operation of
other VFs, possibly attached to other guest OSs since ConnectX firmware is not
tolerant to misuse of resources.
The resource tracker module associates each resource with a VF and maintains
state information for the allocated object. It also defines allowed state
transitions and enforces them.
Relationships between resources are also referred to. For example, CQs are
pointed to by QPs, so it is forbidden to destroy a CQ if a QP refers to it.
ICM memory is always accessible through the primary function and hence it is
allocated by the owner of the primary function.
When a guest dies, an FLR is generated for all the VFs it owns and all the
resources it used are freed.
The tracked resource types are: QPs, CQs, SRQs, MPTs, MTTs, MACs, RES_EQs,
and XRCDNs.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
2011-12-13 11:15:24 +07:00
|
|
|
mutex_unlock(&priv->mfunc.master.res_tracker.slave_list[slave].mutex);
|
|
|
|
}
|
2013-06-27 23:05:21 +07:00
|
|
|
|
|
|
|
void mlx4_vf_immed_vlan_work_handler(struct work_struct *_work)
|
|
|
|
{
|
|
|
|
struct mlx4_vf_immed_vlan_work *work =
|
|
|
|
container_of(_work, struct mlx4_vf_immed_vlan_work, work);
|
|
|
|
struct mlx4_cmd_mailbox *mailbox;
|
|
|
|
struct mlx4_update_qp_context *upd_context;
|
|
|
|
struct mlx4_dev *dev = &work->priv->dev;
|
|
|
|
struct mlx4_resource_tracker *tracker =
|
|
|
|
&work->priv->mfunc.master.res_tracker;
|
|
|
|
struct list_head *qp_list =
|
|
|
|
&tracker->slave_list[work->slave].res_list[RES_QP];
|
|
|
|
struct res_qp *qp;
|
|
|
|
struct res_qp *tmp;
|
|
|
|
u64 qp_mask = ((1ULL << MLX4_UPD_QP_PATH_MASK_ETH_TX_BLOCK_UNTAGGED) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_ETH_TX_BLOCK_1P) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_ETH_TX_BLOCK_TAGGED) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_ETH_RX_BLOCK_UNTAGGED) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_ETH_RX_BLOCK_1P) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_ETH_RX_BLOCK_TAGGED) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_VLAN_INDEX) |
|
|
|
|
(1ULL << MLX4_UPD_QP_PATH_MASK_SCHED_QUEUE));
|
|
|
|
|
|
|
|
int err;
|
|
|
|
int port, errors = 0;
|
|
|
|
u8 vlan_control;
|
|
|
|
|
|
|
|
if (mlx4_is_slave(dev)) {
|
|
|
|
mlx4_warn(dev, "Trying to update-qp in slave %d\n",
|
|
|
|
work->slave);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
mailbox = mlx4_alloc_cmd_mailbox(dev);
|
|
|
|
if (IS_ERR(mailbox))
|
|
|
|
goto out;
|
2013-06-27 23:05:22 +07:00
|
|
|
if (work->flags & MLX4_VF_IMMED_VLAN_FLAG_LINK_DISABLE) /* block all */
|
|
|
|
vlan_control = MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_PRIO_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_TX_BLOCK_UNTAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_PRIO_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_UNTAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_TAGGED;
|
|
|
|
else if (!work->vlan_id)
|
2013-06-27 23:05:21 +07:00
|
|
|
vlan_control = MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_TAGGED;
|
|
|
|
else
|
|
|
|
vlan_control = MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_PRIO_TAGGED |
|
|
|
|
MLX4_VLAN_CTRL_ETH_RX_BLOCK_UNTAGGED;
|
|
|
|
|
|
|
|
upd_context = mailbox->buf;
|
|
|
|
upd_context->primary_addr_path_mask = cpu_to_be64(qp_mask);
|
|
|
|
upd_context->qp_context.pri_path.vlan_control = vlan_control;
|
|
|
|
upd_context->qp_context.pri_path.vlan_index = work->vlan_ix;
|
|
|
|
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
list_for_each_entry_safe(qp, tmp, qp_list, com.list) {
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
if (qp->com.owner == work->slave) {
|
|
|
|
if (qp->com.from_state != RES_QP_HW ||
|
|
|
|
!qp->sched_queue || /* no INIT2RTR trans yet */
|
|
|
|
mlx4_is_qp_reserved(dev, qp->local_qpn) ||
|
|
|
|
qp->qpc_flags & (1 << MLX4_RSS_QPC_FLAG_OFFSET)) {
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
port = (qp->sched_queue >> 6 & 1) + 1;
|
|
|
|
if (port != work->port) {
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
upd_context->qp_context.pri_path.sched_queue =
|
|
|
|
qp->sched_queue & 0xC7;
|
|
|
|
upd_context->qp_context.pri_path.sched_queue |=
|
|
|
|
((work->qos & 0x7) << 3);
|
|
|
|
|
|
|
|
err = mlx4_cmd(dev, mailbox->dma,
|
|
|
|
qp->local_qpn & 0xffffff,
|
|
|
|
0, MLX4_CMD_UPDATE_QP,
|
|
|
|
MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
|
|
|
|
if (err) {
|
|
|
|
mlx4_info(dev, "UPDATE_QP failed for slave %d, "
|
|
|
|
"port %d, qpn %d (%d)\n",
|
|
|
|
work->slave, port, qp->local_qpn,
|
|
|
|
err);
|
|
|
|
errors++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_lock_irq(mlx4_tlock(dev));
|
|
|
|
}
|
|
|
|
spin_unlock_irq(mlx4_tlock(dev));
|
|
|
|
mlx4_free_cmd_mailbox(dev, mailbox);
|
|
|
|
|
|
|
|
if (errors)
|
|
|
|
mlx4_err(dev, "%d UPDATE_QP failures for slave %d, port %d\n",
|
|
|
|
errors, work->slave, work->port);
|
|
|
|
|
|
|
|
/* unregister previous vlan_id if needed and we had no errors
|
|
|
|
* while updating the QPs
|
|
|
|
*/
|
|
|
|
if (work->flags & MLX4_VF_IMMED_VLAN_FLAG_VLAN && !errors &&
|
|
|
|
NO_INDX != work->orig_vlan_ix)
|
|
|
|
__mlx4_unregister_vlan(&work->priv->dev, work->port,
|
|
|
|
work->orig_vlan_ix);
|
|
|
|
out:
|
|
|
|
kfree(work);
|
|
|
|
return;
|
|
|
|
}
|