The q_vector names are based on the interface name with a driver prefix,
the type of q_vector setup, and the queue number. We previously set the
size of this variable to IFNAMSIZ + 9, which is incorrect, because we
actually include a minimum of 14 characters extra beyond the interface
name size.

New versions of GCC since 7 include a new warning that detects this
possible truncation and complains. We can fix this by increasing the
size in case our interface name is too large to avoid truncation. We
don't need to go beyond 14 because the compiler is smart enough to
realize our values can never exceed size of 1. We do go up to 15 here
because possible future changes may increase the number of queues beyond
one digit.

While we are here, also change some variables to be unsigned (since they
are never negative) and stop using an extra unnecessary %s format
specifier.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

According to the header file cpumask.h, we shouldn't be directly copying
a cpumask_t, since its a bitmap and might not be copied correctly. Lets
use the provided cpumask_copy() function instead.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

If we're going to bother initializing a variable to reference it we might
as well use it.
Signed-off-by: NAlan Brady <alan.brady@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The current name of vf_offload_flags indicates that the bitmap is
limited to offload related features. Make this more generic by renaming
it to vf_cap_flags, which allows for other capabilities besides
offloading to be added.
Signed-off-by: NStefan Assmann <sassmann@kpanic.de>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The number of flags found in pf->flags has grown quite large, and there
are a lot of different types of flags. Most of the flags are simply
hardware features which are enabled on some firmware or some MAC types.
Other flags are dynamic run-time flags which enable or disable certain
features of the driver.

Separate these two types of flags into pf->hw_features and pf->flags.
The hw_features list will contain a set of features which are enabled at
init time. This will not contain toggles or otherwise dynamically
changing features. These flags should not need atomic protections, as
they will be set once during init and then be essentially read only.

Everything else will remain in the flags variable. These flags may be
modified at any time during run time. A future patch may wish to convert
these flags into set_bit/clear_bit/test_bit or similar approach to
ensure atomic correctness.

The I40E_FLAG_MFP_ENABLED flag may be a good fit for hw_features but
currently is used by ethtool in the private flags settings, and thus has
been left as part of flags.

Additionally, I40E_FLAG_DCB_CAPABLE may be a good fit for the
hw_features but this patch has not tried to untangle it yet.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Currently i40evf_close() can return before state transitions to
__I40EVF_DOWN because of the latency involved in processing and
receiving response from PF driver and scheduling of VF watchdog_task.
Due to this inconsistency an immediate call to i40evf_open() fails
because state is still DOWN_PENDING.

When a VF interface is in up state and we try to add it as slave,
The bonding driver calls dev_close() and dev_open() in short duration
resulting in dev_open returning error. The ifenslave command needs
to be run again for dev_open to succeed.

This fix ensures that watchdog timer is scheduled immediately after
admin queue operations are scheduled in i40evf_down(). In addition a
wait condition is added at the end of i40evf_close so that function
wont return when state is still DOWN_PENDING. The timeout value is
chosen after some profiling and includes some buffer.
Signed-off-by: NSudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Now that the kernel supports double VLAN tags, we should at least play
nice. Adjust the max packet size to account for two VLAN tags, not just
one.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e hardware descriptor fields are in little-endian format. Driver
must use le32_to_cpu while evaluating these fields otherwise on
big-endian arch we end up evaluating incorrect values, cause errors
like:
i40evf 0000:03:0a.0: Expected response 24 from PF, received 402653184
i40evf 0000:03:0a.1: Expected response 7 from PF, received 117440512
Signed-off-by: NTushar Dave <tushar.n.dave@oracle.com>
Reviewed-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The variable num_active_queues represents the number of active queues we
have for the device. We assign this pretty early in i40evf_init_subtask.

Several code locations are written with loops over the tx_rings and
rx_rings structures, which don't get allocated until
i40evf_alloc_queues, and which get freed by i40evf_free_queues.

These call sites were written under the assumption that tx_rings and
rx_rings would always be allocated at least when num_active_queues is
non-zero.

Lets fix this by moving the assignment into the function where we
allocate queues. We'll use a temporary variable for storage so that we
don't assign the value in the adapter structure until after the rings
have been set up.

Finally, when we free the queues, we'll clear the value to ensure that
we do not loop over the rings memory that no longer exists.

This resolves a possible NULL pointer dereference in
i40evf_get_ethtool_stats which could occur if the VF fails to recover
from a reset, and then a user requests statistics.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add device ID define and mac_type assignment needed for
Adaptive Virtual Function (VF Base Mode Support).

Also, update version to v3.0.0 in order to indicate
clearly that this is the first driver supporting the AVF
device ID.
Signed-off-by: NPreethi Banala <preethi.banala@intel.com>
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch implements the complete version of the virtchnl.h file
with final renames, and fixes the related code in i40e and i40evf.

It also expands comments, and adds details on the usage of
certain fields.

In addition, due to the changes a couple of casts are needed
to prevent errors found by sparse after renaming some fields.
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This morphs all the i40e and i40evf references to/in virtchnl.h
to be generic, using only automated methods. Updates all the
callers to use the new names.  A followup patch provides separate
clean ups for messy line conversions from these "automatic"
changes, to make them more reviewable.

Was executed with the following sed script:
sed -i -f transform_script drivers/net/ethernet/intel/i40e/i40e_client.c
sed -i -f transform_script drivers/net/ethernet/intel/i40e/i40e_prototype.h
sed -i -f transform_script drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
sed -i -f transform_script drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.h
sed -i -f transform_script drivers/net/ethernet/intel/i40evf/i40e_common.c
sed -i -f transform_script drivers/net/ethernet/intel/i40evf/i40e_prototype.h
sed -i -f transform_script drivers/net/ethernet/intel/i40evf/i40evf.h
sed -i -f transform_script drivers/net/ethernet/intel/i40evf/i40evf_client.c
sed -i -f transform_script drivers/net/ethernet/intel/i40evf/i40evf_main.c
sed -i -f transform_script drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
sed -i -f transform_script include/linux/avf/virtchnl.h

transform_script:
----8<----
s/I40E_VIRTCHNL_SUPPORTED_QTYPES/SAVE_ME_SUPPORTED_QTYPES/g
s/I40E_VIRTCHNL_VF_CAP/SAVE_ME_VF_CAP/g

s/I40E_VIRTCHNL_/VIRTCHNL_/g
s/i40e_virtchnl_/virtchnl_/g
s/i40e_vfr_/virtchnl_vfr_/g
s/I40E_VFR_/VIRTCHNL_VFR_/g

s/VIRTCHNL_OP_ADD_ETHER_ADDRESS/VIRTCHNL_OP_ADD_ETH_ADDR/g
s/VIRTCHNL_OP_DEL_ETHER_ADDRESS/VIRTCHNL_OP_DEL_ETH_ADDR/g
s/VIRTCHNL_OP_FCOE/VIRTCHNL_OP_RSVD/g

s/SAVE_ME_SUPPORTED_QTYPES/I40E_VIRTCHNL_SUPPORTED_QTYPES/g
s/SAVE_ME_VF_CAP/I40E_VIRTCHNL_VF_CAP/g
----8<----
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

On architectures with larger pages, we get a warning about an unused variable:

drivers/net/ethernet/intel/i40evf/i40evf_main.c: In function 'i40evf_configure_rx':
drivers/net/ethernet/intel/i40evf/i40evf_main.c:690:21: error: unused variable 'netdev' [-Werror=unused-variable]

This moves the declaration into the #ifdef to avoid the warning.

Fixes: dab86afd ("i40e/i40evf: Change the way we limit the maximum frame size for Rx")
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Acked-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This matches the ordering of how we free stuff during reset and remove.
It also makes logical sense because we set the interrupts based on the
number of queues. Currently this doesn't really matter in practice.
However a future patch moves the assignment of num_active_queues into
i40evf_alloc_queues, which is required by
i40evf_set_interrupt_capability.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

We already set pairs to the value of adapter->num_active_queues. This
value is limited by vsi_res->num_queue_pairs and num_online_cpus(). This
means that pairs by definition is already smaller than
num_online_cpus()*2, so we don't even need to bother with this check.

Lets just remove it and update the comment.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Instead of assuming our flags fit within an unsigned long, use
DECLARE_BITMAP which will ensure that we always allocate enough space.
Additionally, use __I40E_STATE_SIZE__ markers as the last element of the
enumeration so that the size of the BITMAP is compile-time assigned
rather than programmer-time assigned. This ensures that potential future
flag additions do not actually overrun the array. This is especially
important as 32bit systems would only have 32bit longs instead of 64bit
longs as we generally have assumed in the prior code.

This change also removes a dereference of the state fields throughout
the code, so it does have a bit of code churn. The conversions were
automated using sed replacements with an alternation

  s/&(vsi->back|vsi|pf)->state/\1->state/
  s/&adapter->vsi.state/adapter->vsi.state/

For debugfs, we modify the printing so that we can display chunks of the
state value on new lines. This ensures that we can print the entire set
of state values. Additionally, we now print them as 08lx to ensure that
they display nicely.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Avoid using the same named flags for both vsi->state and pf->state. This
makes code review easier, as it is more likely that future authors will
use the correct state field when checking bits. Previous commits already
found issues with at least one check, and possibly others may be
incorrect.

This reduces confusion as it is more clear what each flag represents,
and which flags are valid for which state field.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The delay was added because of a desire to ensure that the VF driver can
finish up removing. However, pci_disable_sriov already has its own
ssleep() call that will sleep for an entire second, so there is no
reason to add extra delay on top of this by using msleep here. In
practice, an msleep() won't have a huge impact on timing but there is no
real value in keeping it, so lets just simplify the code and remove it.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff 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: NScott Peterson <scott.d.peterson@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff 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: NTobias Klauser <tklauser@distanz.ch>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff 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: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff 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: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch adds a capability negotiation between VF and PF using ENCAP/
ENCAP_CSUM offload flags in order for the VF to support outer checksum
and TSO offloads for encapsulated packets. These capabilities were assumed
by default and enabled in current hardware. Going forward, these features
needs to be negotiated with PF before advertising to the stack.
Additionally, strip out the mac.type checks for X722 since outer checksums
are enabled based on the ENCAP_CSUM offload negotiation flag and maintain
consistency between drivers in how the features are configured.

Change-ID: Ie380a6f57eca557a2bb575b66b12fae36d308920
Signed-off-by: NPreethi Banala <preethi.banala@intel.com>
Signed-off-by: NAlan Brady <alan.brady@intel.com>
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch changes the way we handle the maximum frame size for the Rx
path.  Previously we were rounding up to 2K for a 1500 MTU and then brining
the max frame size down to MTU plus a fixed amount.  With this patch
applied what we now do is limit the maximum frame to 1.5K minus the value
for NET_IP_ALIGN for standard MTU, and for any MTU greater than 1500 we
allow up to the maximum frame size.  This makes the behavior more
consistent with the other drivers such as igb which had similar logic.  In
addition it reduces the test matrix for MTU since we only have two max
frame sizes that are handled for Rx now.

Change-ID: I23a9d3c857e7df04b0ef28c64df63e659c013f3f
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NBimmy Pujari <bimmy.pujari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

In preparation for upcoming RDMA-capable hardware, add a client
interface to the VF driver. This is a slightly-simplified version
of the PF client interface, with the names changed to protect the
innocent.

Due to the nature of the VF<->PF interactions, the client interface
sometimes needs to call back into itself to pass messages. Because
of this, we can't use the coarse-grained locking like the PF's
client interface uses. Instead, we handle all client interactions
in a separate thread so the watchdog can still run and process
virtual channel messages.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NAnjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: NAvinash Dayanand <avinash.dayanand@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add a comment to reduce confusion.

Change-ID: I3d5819c0f3f5174680442ae54398a073d4a61f4f
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

When the i40evf_remove() calls netdev close, the device doesn't actually
close - it schedules the work for the watchdog to perform. Since we're
stopping the watchdog, this work doesn't get done. However, we're
resetting the part, so we can free resources after the reset request has
gone through. This plugs a memory leak.

Change-ID: Id5335dcaf76ce00d2a4c3d26e9faf711d7f051cf
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Currently i40e_bus_info has PCI device and function info only and log
messages print device number as bus number. Added field to provide bus
number info and modified log statements to print bus, device and
function information.

Change-ID: I811617cee2714cc0d6bade8d369f57040990756f
Signed-off-by: NSudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NBimmy Pujari <bimmy.pujari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This device ID was intended for use when running Linux VF drivers under
Hyper-V, but we have determined that it is not necessary. Since it is
unused, and will never be used, remove it.

Change-ID: I74998ab4237db043cd400547bb54a0a5e2a37ea5
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NBimmy Pujari <bimmy.pujari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Update the functions which free msix_entries and q_vectors so that they
are safe against NULL values. This allows calling code to not care
whether these have already been freed when disabling and freeing them.

Change-ID: I31bfd1c0da18023d971b618edc6fb049721f3298
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

It is possible for msix_entries to be freed by a previous suspend/remove
before a VF is closed.  This patch fixes the issue by checking for NULL
before dereferencing msix_entries and returning early in the case where
it is NULL within the i40evf_close code path.  Without this patch it is
possible to trigger a kernel panic through NULL dereference.

Change-ID: I92a2746e82533a889e25f91578eac9abd0388ae2
Signed-off-by: NAlan Brady <alan.brady@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The i40evf_reset_task function is a couple hundred lines and it has
a separable block that disables VF.  Move that block to a new
i40evf_disable_vf function to shorten i40evf_reset_task a bit.
Signed-off-by: NJoe Perches <joe@perches.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change makes it so that we are much more robust about defining what we
can and cannot offload.  Previously we were performing no checks.  This
should bring us up to parity with the i40e PF driver.

In addition the device only supports GSO as long as the MSS is 64 or
greater.  We were not checking this so an MSS less than that was resulting
in Tx hangs.

Change-ID: If533553ec92fc6ba694eab6ac81fdaf3004f3592
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Remove the second call to msleep outside the loop, and move the msleep
within the loop as the first step. This guarantees that a single loop
will wait the minimum time first, and then after the reset finishes we
no longer need an extra msleep.

Change-ID: Ib2086f0a142402b614f67846bc091754203a0b9a
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NBimmy Pujari <bimmy.pujari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NBimmy Pujari <bimmy.pujari@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

There exists a bug in which a 'perfect storm' can occur and cause
interrupts to fail to be correctly affinitized. This causes unexpected
behavior and has a substantial impact on performance when it happens.

The bug occurs if there is heavy traffic, any number of CPUs that have
an i40e interrupt are pegged at 100%, and the interrupt afffinity for
those CPUs is changed.  Instead of moving to the new CPU, the interrupt
continues to be polled while there is heavy traffic.

The bug is most readily realized as the driver is first brought up and
all interrupts start on CPU0. If there is heavy traffic and the
interrupt starts polling before the interrupt is affinitized, the
interrupt will be stuck on CPU0 until traffic stops. The bug, however,
can also be wrought out more simply by affinitizing all the interrupts
to a single CPU and then attempting to move any of those interrupts off
while there is heavy traffic.

This patch fixes the bug by registering for update notifications from
the kernel when the interrupt affinity changes. When that fires, we
cache the intended affinity mask. Then, while polling, if the cpu is
pegged at 100% and we failed to clean the rings, we check to make sure
we have the correct affinity and stop polling if we're firing on the
wrong CPU.  When the kernel successfully moves the interrupt, it will
start polling on the correct CPU. The performance impact is minimal
since the only time this section gets executed is when performance is
already compromised by the CPU.

Change-ID: I4410a880159b9dba1f8297aa72bef36dca34e830
Signed-off-by: NAlan Brady <alan.brady@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>