Jeff Kirsher says:
====================
1GbE Intel Wired LAN Driver Updates 2017-04-20
This series contains updates to e1000, e1000e, igb/vf and ixgb.
Tobias Klauser cleans up e1000, ixgb and igbvf from having a local
function or structure for netdev stats.
Bernd Faust fixes an issue for 82579 devices, where the clock frequency
was being incorrectly set for these devices. These devices only support
96MHz, so make sure they are set to use only that.
Yury Kylulin extends the work Jake and Alex did for ixgbe in MAC filter
handling into the igb driver.
Kim Tatt Chuah enables igb to wake up by packet and to read the necessary
Wake Up Status (WUS) and Wake Up Packet Memory (WUPM) registers.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Jeff Kirsher says:
====================
40GbE Intel Wired LAN Driver Updates 2017-04-19
This series contains updates to i40e and i40evf only, most notable being
the addition of trace points for BPF programs.
Tobias Klauser updates i40evf to use net_device stats struct instead
of a local private copy.
Preethi updates the VF driver to not enable receive checksum offload by
default for tunneled packets.
Alex fixes an issue he introduced when he converted the code over to
using the length field to determine if a descriptor was done or not.
Mitch adds the ability to dump additional information on the VFs, which
is not available through 'ip link show' using debugfs.
Scott adds trace points to the drivers so that BPF programs can be
attached for feature testing and verification.
Jingjing adds admin queue functions for Pipeline Personalization Profile
commands.
Jake does most of the heavy lifting in this series, starting with the
a reduction in the scope of the RTNL lock being held while resetting VFs
to allow multiple PFs to reset in a timely manner. Factored out the
direct queue modification so that we are able to re-use the code.
Reduced the wait time for admin queue commands to complete, since we were
waiting a minimum of a millisecond, when in practice the admin queue
command is processed often much faster. Cleaned up code (flag) we never
use. Make the code to resetting all the VFs optimized for parallel
computing instead of the current way is a serialized fashion, to help
reduce the time it takes.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of using a private copy of struct net_device_stats in
struct igbvf_adapter, use stats from struct net_device. Also remove the
now unnecessary .ndo_get_stats function.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Currently, in igb_resume(), igb driver ignores the Wake Up Status (WUS)
and Wake Up Packet Memory (WUPM) registers. This patch enables the igb
driver to read the WUPM if the controller was woken by a wake up packet
that is not more than 128 bytes long (maximum WUPM size), then pass it
up the kernel network stack.
Signed-off-by: Kim Tatt Chuah <kim.tatt.chuah@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add functionality for the VF to request up to 3 additional MAC filters.
This is done using existing E1000_VF_SET_MAC_ADDR message, but with
additional message info - E1000_VF_MAC_FILTER_CLR to clear all unicast
MAC filters previously set for this VF and E1000_VF_MAC_FILTER_ADD to
add MAC filter.
Additional filters can be added only in case if administrator did not
set VF MAC explicitly and allowed to change default MAC to the VF.
Due to the limited number of RAR entries reserve at least 3 MAC filters
for the PF.
If SRIOV is supported by the NIC after this change RAR entries starting
from 1 to (RAR MAX ENTRIES - NUM SRIOV VFS) will be used for PF and VF
MAC filters.
Signed-off-by: Yury Kylulin <yury.kylulin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Using the work which was done for ixgbe driver by Jacob Keller
commit 5d7daa35b9 ("ixgbe: improve mac filter handling") and Alexander
Duyck commit 0f079d2283 ("ixgbe: Use __dev_uc_sync and __dev_uc_unsync
for unicast addresses") and out-of-tree igb driver add functionality to
manage (add and delete) MAC filters.
Signed-off-by: Yury Kylulin <yury.kylulin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
After an upgrade to Linux kernel v4.x the hardware timestamps of the
82579 Gigabit Ethernet Controller are different than expected.
The values that are being read are almost four times as big as before
the kernel upgrade.
The difference is that after the upgrade the driver sets the clock
frequency to 25MHz, where before the upgrade it was set to 96MHz. Intel
confirmed that the correct frequency for this network adapter is 96MHz.
Signed-off-by: Bernd Faust <berndfaust@gmail.com>
Acked-by: Sasha Neftin <sasha.neftin@intel.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
ixgb_get_stats() just returns dev->stats so we can leave it
out altogether and let dev_get_stats() do the job.
Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
e1000_get_stats() just returns dev->stats so we can leave it
out altogether and let dev_get_stats() do the job.
Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This state bit was added as a way for DCB to avoid having to wait for
the queues to disable when handling LLDP events. The logic for this was
burried deep within stop Tx and stop Rx queue code. First, let's rename
it so that it does not appear to only affect Tx when infact it modifies
both Tx and Rx flow. Second we can move it up into the i40e_stop_rings()
function, and we can simply re-use the i40e_stop_rings_no_wait() so that
we don't have to bury the implementation as deep into the call stack.
An alternative might be to remove the state bit and instead attempt to
shut down everything directly in DCP flow. This, however, is not ideal
because it creates yet another separate shutdown routine that we'd have
to maintain. In the current implementation any changes will be made to
both flows.
Change-ID: I68e1ccb901af320862bca395e9c9746f08e8b17c
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When there are a lot of active VFs, it can take multiple seconds to
finish resetting all of them during certain flows., which can cause some
VFs to fail to wait long enough for the reset to occur. The user might
see messages like "Never saw reset" or "Reset never finished" and the VF
driver will stop functioning properly.
The naive solution would be to simply increase the wait timer. We can
get much more clever. Notice that i40e_reset_vf is run in a serialized
fashion, and includes lots of delays.
There are two prominent delays which take most of the time. First, when
we begin resetting VFs, we have multiple 10ms delays which accrue
because we reset each VF in a serial fashion. These delays accumulate to
almost 4 seconds when handling the maximum number of VFs (128).
Secondly, there is a massive 50ms delay for each time we disable queues
on a VSI. This delay is necessary to allow HW to finish disabling queues
before we restore functionality. However, just like with the first case,
we are paying the cost for each VF, rather than disabling all VFs and
waiting once.
Both of these can be fixed, but required some previous refactoring to
handle the special case. First, we will need the
i40e_vsi_wait_queues_disabled function which was previously DCB
specific. Second, we will need to implement our own
i40e_vsi_stop_rings_no_wait function which will handle the stopping of
rings without the delays.
Finally, implement an i40e_reset_all_vfs function, which will first
start the reset of all VFs, and pay the wait cost all at once, rather
than serially waiting for each VF before we start processing then next
one. After the VF has been reset, we'll disable all the VF queues, and
then wait for them to disable. Again, we'll organize the flow such that
we pay the wait cost only once.
Finally, after we've disabled queues we'll go ahead and begin restoring
VF functionality. The result is reducing the wait time by a large factor
and ensuring that VFs do not timeout when waiting in the VF driver.
Change-ID: Ia6e8cf8d98131b78aec89db78afb8d905c9b12be
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
A future patch is going to want to re-use some of the code in
i40e_reset_vf, so lets break up the beginning and ending parts into
their own helper functions. The first function will be used to
initialize the reset on a VF, while the second function will be used to
finalize the reset and restore functionality.
Change-ID: I48df808b8bf09de3c2ed8c521f57b3f0ab9e5907
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This flag was originally intended to be used to let some
driver code know when we were running from netpoll.
Ultimately this was not necessary and we never used it.
Let's remove it
Change-ID: I43b72483d91c1638071d2a7f389ab171ec5b796a
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When sending an adminq command, we wait for the command to complete in
a loop. This loop waits for an entire millisecond, when in practice the
adminq command is processed often much faster.
Change the loop to use i40e_usec_delay instead, and wait for 50 usecs
each time instead. This appears to be about the minimum time required,
based on some manual observation and testing.
The primary benefit of this change is reducing latency of various
operations in the PF driver, especially when related to having a large
number of VFs enabled.
For example, on Linux, when instantiating 128 VFs, the time to finish
the operation dropped from about 9 seconds down to under 6 seconds.
Additionally, the time it takes to finish a PF reset with 128 VFs
dropped from 5.1 seconds down to 0.7 seconds.
As the examples above show, a significant portion of the delay is wasted
waiting for admiqn operations which have already finished.
This patch shouldn't cause impact to functionality, as we still check
and keep waiting until the command does get processed. The only expected
change is an increase in CPU utilization as we now check for completion
far more times. However, in practice the commands appear to generally be
complete within the first delay window anyways.
Change-ID: If8af8388e100da0a14eaf9e1af3afadf73a958cf
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The check for I40E_CONFIG_BUSY state bit in the i40e_set_link_ksettings
function is fishy. First we can notice a few things about the check here.
First a similar check was introduced by commit
'c7d05ca89f8e ("i40e: driver ethtool core")'
Later a commit introducing the link settings was added by commit
'bf9c71417f72 ("i40e: Implement set_settings for ethtool")'
However, this second check was against vsi->state instead of pf->state,
and also failed to set the bit, it only checks. That indicates the locking
was not quite correct. The only other place that the state bit
in vsi->state gets used is to protect the filter list.
Since this code does not care about the mac filter list, and seems
clear the original code should have set the pf->state bit. Fix these
issues by using pf->state correctly, and by actually setting the bit
so that we properly lock as expected.
Since these checks occur while holding the rtnl_lock(), lets also add a
timeout so that we don't potentially softlock the system.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
A future patch will need to be able to handle controlling queues without
waiting until all VSIs are handled. Factor out the direct queue
modification so that we can easily re-use this code. The result is also
a bit easier to read since we don't embed multiple single-letter loop
counters.
Change-ID: Id923cbfa43127b1c24d8ed4f809b1012c736d9ac
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
We made some effort to reduce the RTNL lock scope when resetting and
rebuilding the PF. Unfortunately we still held the RTNL lock during the
VF reset operation, which meant that multiple PFs could not reset in
parallel due to the global lock. For now, further reduce the scope by
not holding the RTNL lock while resetting VFs. This allows multiple PFs
to reset in a timely manner.
Change-ID: I2fbf823a0063f24dff67676cad09f0bbf83ee4ce
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds tracepoints to the i40e and i40evf drivers to which
BPF programs can be attached for feature testing and verification.
It's expected that an attached BPF program will identify and count or
log some interesting subset of traffic. The bcc-tools package is
helpful there for containing all the BPF arcana in a handy Python
wrapper. Though you can make these tracepoints log trace messages, the
messages themselves probably won't be very useful (other to verify the
tracepoint is being called while you're debugging your BPF program).
The idea here is that tracepoints have such low performance cost when
disabled that we can leave these in the upstream drivers. This may
eventually enable the instrumentation of unmodified customer systems
should the need arise to verify a NIC feature is working as expected.
In general this enables one set of feature verification tools to be
used on these drivers whether they're built with the kernel or
separately.
Users are advised against using these tracepoints for anything other
than a diagnostic tool. They have a performance impact when enabled,
and their exact placement and form may change as we see how well they
work in practice for the purposes above.
Change-ID: Id6014a7322c0e6d08068114dd20bd156f2f6435e
Signed-off-by: Scott Peterson <scott.d.peterson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Dump some internal state about VFs through debugfs. This provides
information not available with 'ip link show'. To use, write "dump vf
<id>" to the command file, or just "dump vf" to dump information on all
of the VFs.
Change-ID: Ibe32b7f4ae55d4358c0b903217475f708ada1ecd
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch fixes an issue I introduced when I converted the code over to
using the length field to determine if a descriptor was done or not. It
turns out that we are also processing programming descriptors in the Rx
path and need to have these processed even though the length field will be
0 on these packets. What will happen with a programming descriptor is that
we will receive a descriptor that has the SPH bit set, and the header
length and packet length fields cleared.
To account for this we should be checking for the bit for split header
being set even though we aren't actually using header split. This bit is
set in the length field to indicate if a programming descriptor response is
contained in the descriptor. Since we don't support header split we don't
need to perform the extra checks of using a fixed value for the entire
length field.
In addition I am moving the function for checking if a filter is a
programming status filter into the i40e_txrx.c file since there is no
longer support for FCoE it doesn't make sense to keep this file in i40e.h.
Change-ID: I12c359c3dc70adb9d6b92b27324bb2c7f04c1a06
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Rx checksum offload for tunneled packets was never being negotiated or
requested by VF. This capability was assumed by default and enabled in
current hardware for VF. Going forward, this feature needs to be disabled
or advanced ptypes should be negotiated with PF in the future.
Change-ID: I9e54cfa8a90e03ab6956db4412f1e337ccd2c2e0
Signed-off-by: Preethi Banala <preethi.banala@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Instead of using a private copy of struct net_device_stats in
struct i40evf_adapter, use stats from struct net_device. Also remove the
now unnecessary .ndo_get_stats function.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
I just found that when we had changed the Rx path to check for length
instead of the DD bit we introduced an issue in ixgbe_dump since we were no
longer clearing the status bits.
To correct this I am updating ixgbe_dump to look for the length bits in the
descriptor since that is what we are using in the Rx path.
Fixes: c3630cc40b ("ixgbe: Use length to determine if descriptor is done")
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch increases the headroom allocated when using build_skb on a
system with 4K pages. Specifically the breakdown of headroom versus cache
size is as follows:
L1 Cache Size Headroom
64 192
64, NET_IP_ALIGN == 2 194
128 128
128, NET_IP_ALIGN == 2 130
256 512
256, NET_IP_ALIGN == 2 258
I stopped at supporting only a cache line size of 256 as that was the
largest cache size I could find supported in the kernel.
With this we are guaranteeing at least 128 bytes of headroom to spare in
the frame. This should be enough for us to insert a couple of IPv6 headers
if needed which is likely enough room for anything XDP should need.
I'm leaving the padding for systems with pages larger than 4K unmodified
for now. XDP currently isn't really setup to work on those types of
systems so we can cross that bridge when we get there.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
We did not have a check in place for MMNGC.MNG_VETO when setting up link
on X550EM_X KR devices which resulted in link loss for the BMC when
loading the driver.
This patch adds a check for ixgbe_check_reset_blocked() in setup_link()
since in that case there is no PHY reset function.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The ethtool api {get|set}_settings is deprecated.
We move this driver to new api {get|set}_link_ksettings.
As I don't have the hardware, I'd be very pleased if
someone may test this patch.
Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Remove the Marvell 1145 PHY define as we have never had a device that
supports it and have no plan to in the future. The existence of this
define has caused confusing on whether or not this PHY was supported
by ixgbe.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Avoid setting adapter->num_vfs early in the init code path when
using the max_vfs module parameter by passing it to ixgbe_enable_sriov()
as a function parameter.
This fixes an issue where if we failed to allocate vfinfo in
__ixgbe_enable_sriov() the driver will crash with NULL pointer in
ixgbe_disable_sriov() when attempting to free the vfinfo struct based
on adapter->num_vfs. Also it cleans up the assignment of adapter->num_vfs
since now it will only be set in __ixgbe_enable_sriov() and cleared in
ixgbe_disable_sriov().
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Since we exit at the end of the block, we can save a level of
indentation by performing an early return, and make the next several
sections of code more legible, with fewer 80 character line breaks.
Also moved allocating vfinfo at the beginning and the notification
for enabling SRIOV at the end of the function when we know that it
will succeed.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Move the code allocating memory for list of MAC addresses that
the VFs can use for MACVLAN into its own function.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add default setting for mac->ops.setup_link on x550em_a MAC types.
This fixes a link issue on KR parts.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
We forgot to indicate some of the supported speed on the X553
backplane. This patch attempts to correct for that.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch add support for X552 XFI backplane interface. The XFI
backplane requires a custom tuned link. HW/FW owns the link config
for XF backplane and SW must not interfere with it.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The initial patches supporting X553 sgmii forgot some details. This patch
should cover those missing spots.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The KX4 PHY is configured by the NVM. Currently, the driver is overwriting
the config; remove the code associated with KX4 configuration.
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
As pr_cont output can be interleaved by other processes,
using pr_cont should be avoided where possible.
Miscellanea:
- Use a temporary pointer to hold the next descriptions and
consolidate the pr_cont uses
- Use the temporary buffer to hold the 8 u32 register values and
emit those in a single go
- Coalesce formats and logging neatening around those changes
- Fix a defective output for the rx ring entry description when
also emitting rx_buffer_info data
This reduces overall object size a tiny bit too.
$ size drivers/net/ethernet/intel/ixgbe/*.o*
text data bss dec hex filename
62167 728 12 62907 f5bb drivers/net/ethernet/intel/ixgbe/ixgbe_main.o.new
62273 728 12 63013 f625 drivers/net/ethernet/intel/ixgbe/ixgbe_main.o.old
Signed-off-by: Joe Perches <joe@perches.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When DCB is enabled, add checks to ensure creation of number of VF's is
valid based on the traffic classes configured by the device.
Signed-off-by: Usha Ketineni <usha.k.ketineni@intel.com>
Tested-by: Ronald Bynoe <ronald.j.bynoe@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch is meant to improve the performance of the Rx path.
Specifically by using build_skb we have several distinct advantages.
In the case of small frames we were previously using a copy-break approach.
This means that we were allocating a page fragment to use for skb->head,
and were having to copy the packet into that region. Both of those calls
are now avoided since we just build the skb around the data.
In the case of large frames the gains are much more significant.
Specifically we were having to allocate skb->head, and copy the headers as
before. However in addition we were having to parse the header using
eth_get_headlen which could be quite expensive. All of this is avoided by
building the frame around the data. I have seen gains as high as 30% when
using VXLAN for instance due to just header pulling overhead.
Finally with all this in place it also sets us up to start looking at
enabling XDP. Specifically we now have a path in which the data is in the
page and the frame is built around it. So if we parse it with XDP before
we call build_skb we can take care of any necessary processing there.
Change-ID: Id4bdd618e94473d41f892417e5d8019639e421e3
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds padding to the start of frames to make room for headroom
for us to eventually start using build_skb. Right now we guarantee at
least NET_SKB_PAD + NET_IP_ALIGN, however we allocate more space if more is
available. For example on x86 the headroom should be 192 bytes.
On systems that have too large of a cache line size to support storing 1.5K
padding and shared info we default to using 3K buffers and reserve
everything that isn't used for skb_shared_info or the data buffer for
headroom.
Change-ID: I33c641c9a1ea10cf7cc484c2d20985368d2d709a
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
There are situations where adding padding to the front and back of an Rx
buffer will require that we add additional padding. Specifically if
NET_IP_ALIGN is non-zero, or the MTU size is larger than 7.5K we would need
to use 2K buffers which leaves us with no room for the padding.
To preemptively address these cases I am adding support for 3K buffers to
the Rx path so that we can provide the additional padding needed in the
event of NET_IP_ALIGN being non-zero or a cache line being greater than 64.
Change-ID: I938bc1ba611285428df39a613cd66f98e60b55c7
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Since an early commit a few flags have no longer
been used. Remove these definitions to reduce code clutter.
Change-ID: I3589be4622574e747013cd4dc403e18b039f4965
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The I40E_FLAG_NEED_LINK_UPDATE was never used. Remove the flag
definitions.
Change-ID: If59d0c6b4af85ca27281f3183c54b055adb439a4
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
We can simply check both Tx and Rx queues in a single loop, rather than
repeating the loop twice.
Change-ID: Ic06f26b0e3c2620e0e33c1a2999edda488e647ad
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Look up the MAC address from the eth_get_platform_mac_address() function
first before checking what the firmware provides. We already handle the
case of re-writing the MAC-VLAN filter, so there is no need to add extra
code for this. However, update the comment where we do this to indicate
that it does impact the Open Firmware MAC address case.
Change-ID: I73e59fbe0b0e7e6f3ee9f5170d0bd3a4d5faf4db
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch greatly reduces the unneeded complexity in the
i40e_detect_recover_hung_queue code path. The previous implementation
set a 'hung bit' which would then get cleared while polling. If the
detection routine was called a second time with the bit already set, we
would issue a software interrupt. This patch makes it such that if
interrupts are disabled and we have pending TX descriptors, we trigger a
software interrupt since in, the worst case, queues are already clean
and we have an extra interrupt.
Additionally this patch removes the workaround for lost interrupts as
calling napi_reschedule in this context can cause software interrupts to
fire on the wrong CPU.
Change-ID: Iae108582a3ceb6229ed1d22e4ed6e69cf97aad8d
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Previously rtnl lock was held during whole reset procedure that
was stopping other PFs running their reset procedures. In the result
reset was not handled properly and host reset was the only way
to recover.
Change-ID: I23c0771c0303caaa7bd64badbf0c667e25142954
Signed-off-by: Maciej Sosin <maciej.sosin@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This is a minor cleanup so that we are always updating pf->flags when we
make a change to the private flags instead of updating a mix of either
pf->flags and/or pf->hw_disabled_flags.
In addition I went through and cleaned out all the spots where we were
using the X722 define in regards to this flag.
Lastly since we changed the logic I went through and flushed out any
redundancy and cleaned up the handling of the flags in the Tx path.
Change-ID: I79ff95a7272bb2533251ff11ef91e89ccb80b610
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Re-word the error message displayed when adding a filter with an
invalid flow type. Additionally, report a distinct error message when
the IPv4 protocol is at fault.
Change-ID: Iba3d85b87f8d383c97c8bdd180df34a6adf3ee67
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The client interface is only intended for use on devices that support
iWarp. Only register with the client if this is the case.
This fixes a panic when loading i40iw on X710 devices.
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Reported-by: Stefan Assmann <sassmann@kpanic.de>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
