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>

In commit a75e8005 ("i40e: queue-specific settings for interrupt
moderation") the i40e driver gained support for setting interrupt
moderation values per queue. This patch adds support for this feature
to the i40evf driver as well. In addition, a few changes are made to
the i40e implementation to add function header documentation comments,
as well.

This behaves in a similar fashion to the implementation in i40e. Thus,
requesting the moderation value when no queue is provided will report
queue 0 value, while setting the value without a queue will set all
queues at once.

Change-ID: I1f310a57c8e6c84a8524c178d44d1b7a6d3a848e
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>

Several defines and code comments were indented with spaces instead
of tabs, correct the issue to make indentation consistent.

Change-ID: I0dc6bbb990ec4a9e856acc9ec526d876181f092c
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>

The PF driver tells us the link speed, so do something with that
information. Add link speed to log messages, and report speed through
ethtool.

Change-Id: I279dc9540cc5203376406050a3e8d67e128d5882
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>

This patch enables a feature to enable/disable all multicast
for a trusted VF.

Change-Id: I926eba7f8850c8d40f8ad7e08bbe4056bbd3985f
Signed-off-by: NAnjali Singhai Jain <anjali.singhai@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This is part 2 of the Rx refactor series, just including
changes to i40evf.

This refactor aligns the receive routine with the one in
ixgbe which was highly optimized.  This reduces the code
we have to maintain and allows for (hopefully) more readable
and maintainable RX hot path.

In order to do this:
- consolidate the receive path into a single function that doesn't
  use packet split but *does* use pages for Rx buffers.
- remove the old _1buf routine
- consolidate several routines into helper functions
- remove VF ethtool control over packet split
- remove priv_flags interface since it is unused
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>

As part of preparation for the rx-refactor, remove the
packet split receive routine and ancillary code.

Some of the split related context set up code stays in
i40e_virtchnl_pf.c in case an older VF driver tries to load
and still wants to use packet split.
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>

Add necessary Linux Ethernet driver support for promiscuous mode
operation. Add a flag so the VF knows it is in promiscuous mode
and two state flags to discreetly track multicast and unicast
promiscuous states.

Change-Id: Ib2f2dc7a7582304fec90fc917ebb7ded21ba1de4
Signed-off-by: NAnjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: NGreg Rose <gregory.v.rose@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>

If the PF driver reports proper support, allow the PF driver to
configure RSS on the behalf of the VF driver. This will allow for RSS
support on future hardware without changes to the VF driver.

Unfortunately, the old RSS code still needs to stay as the driver needs
to be compatible with PF drivers that don't support this interface. But
this change still simplifies the data structures a bunch and makes this
code simpler to read and maintain.

Change-ID: I0375aad40788ecdc0cb24d5cfeccf07804e69771
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>

Under some circumstances the driver remove function may be called before
the driver is fully initialized. So we can't assume that we know where
our towel is at, or that all of the data structures are initialized.

To ensure that we don't panic, check that the vsi_res pointer is valid
before dereferencing it. Then drink beer and eat peanuts.

Change-ID: If697b4db57348e39f9538793e16aa755e3e1af03
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>

Support packet split receive on VFs. This is off by default but can be
enabled using ethtool private flags. Because we need to trigger a reset
from outside of i40evf_main.c, create a new function to do so, and
export it.

Also update copyright year in file headers.

Change-ID: I721aa5d70113d3d6d94102e5f31526f6fc57cbbb
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>

In some modes, bonding would not enslave VF interfaces. This is due to
bonding calling change_mtu and the immediately calling open. Because of
the asynchronous nature of the admin queue mechanism, the VF returns
-EBUSY to the open call, because it knows the previous operation hasn't
finished yet. This causes bonding to fail with a less-than-useful error
message.

To fix this, remove the check for pending operations at the beginning of
open. But this introduces a new bug where the driver will panic on a
quick close/open cycle. To fix that, we add a new driver state,
__I40EVF_DOWN_PENDING, that the driver enters when down is called. The
driver finally transitions to a fully DOWN state when it receives
confirmation from the PF driver that all the queues are disabled. This
allows open to complete even if there is a pending mtu change, and
bonding is finally happy.

Change-ID: I06f4c7e435d5bacbfceaa7c3f209e0ff04be21cc
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>

Future devices will allow for more queue pairs, so allocate a netdev
that can handle them. While we're at it, get rid of the separate
MAX_TX/MAX_RX defines. Since we always get matched queue pairs, having
these makes no sense.

Change-ID: I0e3556cd9a962506e509eb7c0afa36b329e8cb51
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>

Instead of awkwardly keeping a fixed array of pointers in the adapter
struct and then allocating ring structs individually, just keep a single
pointer and allocate a single blob for the arrays. This simplifies code,
shrinks the adapter structure, and future-proofs the driver by not
limiting the number of rings we can handle.

Change-ID: I31334ff911a6474954232cfe4bc98ccca3c769ff
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>

Change the queue_vector array from a statically-sized member of the
adapter structure to a dynamically-allocated and -sized array.

This reduces the size of the adapter structure, and allows us to support
any number of queue vectors in the future without changing the code.

Change-ID: I08dc622cb2f2ad01e832e51c1ad9b86524730693
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>

This patch adds new fields to i40e_vsi to store user configured
RSS config data and code to use it.

Change-ID: Ic5d3db8d9df52182b560248f8cdca9c5c7546879
Signed-off-by: NHelin Zhang <helin.zhang@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

There are two ways to get RSS, this patch implements two functions
with the same input parameters, and creates a more generic function
for getting RSS configuration.

Change-ID: I12d3b712c21455d47dd0a5aae58fc9b7c680db59
Signed-off-by: NHelin Zhang <helin.zhang@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

There are two ways to configure RSS, this patch adjusts those two
functions with the same input parameters, and creates a more
generic function for configuring RSS.

Change-ID: Iace73bdeba4831909979bef221011060ab327f71
Signed-off-by: NHelin Zhang <helin.zhang@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The adaptive ITR (interrupt throttle rate) algorithm was adjusting
the hardware's interrupt rate too frequently.  This caused a lot
of variation in the interrupt rate for fairly constant workloads.

Change the code to have a counter and adjust only once every N
number of interrupts.

Change-ID: I0460f1f86571037484eca5aca36ac4d889cb8389
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 some devices, in some systems, in some configurations, the VFs would
fail to initialize the first time you loaded the driver.

To correct this, increase the delay time for the init task slightly, and
wait longer before giving up.

If we enable VFs and load the VF driver in the same kernel as the PF
driver, we can totally overwhelm the PF driver with AQ requests because
all of the instances try to initialize at the same time.

To help alleviate this, stagger the initial scheduling of the init task
using the PCIe function as a multiplier. We mask off the function to
only three bits so no instance has to wait too long.

With these two changes, initializing 128 VFs on a single device goes
from four minutes to just a few seconds.

Change-ID: If3d8720c1c4e838ab36d8781d9ec295a62380936
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>

The code in i40e and i40evf is using an "IN_NETPOLL" flag that has never
added any value due to the fact that the Rx clean-up is handled in NAPI.
As such the flag was set, the queue was scheduled via NAPI, and then polled
from the netpoll controller and if any Rx packets were processed the were
processed in the wrong context.

In addition the flag itself just added an unneeded conditional to the
hot-path so it can safely be dropped and save us a few instructions.
Signed-off-by: NAlexander Duyck <aduyck@mirantis.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The driver was not correctly handling calls to its ndo_set_mac_address
method. It did not properly check to see if the override would be
allowed by the PF driver, and never removed the old address from its
filter list.

Add a new flag to the adapter struct which is set if the MAC address is
assigned by the PF. Check this flag and don't allow the MAC address to
be changed if it is set. Search for and properly remove the filter
for the old MAC address when the new one is set.

Change-ID: I817bf620c869c5a80e6a7eab65c9cbad1dc89799
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>

The VF really doesn't care about the QOS handle but it will in the
future. Since the VF only uses TC0, send it that handle. On the VF
side, save the handle and use it to populate the QOS params when we call
into the client interface.

Change-ID: I76f41b070baeaa09b19383e9168bc677837e0761
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>

Add missings spaces after declarations, remove another __func__ use,
remove uncessary braces, remove unneeded breaks, and useless returns,
and generally fix up some code.

Change-ID: Ie715d6b64976c50e1c21531685fe0a2bd38c4244
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NShannon Nelson <shannon.nelson@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

X722 fixes an issue from X710 where TX descriptor WB would not happen if
the interrupts were disabled. In order for the write backs to happen a
bit needs to be set in the dynamic interrupt control register called
WB_ON_ITR. With this feature, the SW driver need not arm SW interrupts to
work around the issue in X710.
Signed-off-by: NAnjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: NCatherine Sullivan <catherine.sullivan@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

X722 uses the admin queue to configure RSS. This patch adds the necessary
flow changes to configure RSS through AQ. It also adds the separate VMDQ2
lookup tables and hash key programming for X722.

X722 also exposes a different set of PCTYPES for RSS, this patch
accommodates those changes.
Signed-off-by: NAnjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: NCatherine Sullivan <catherine.sullivan@intel.com>
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add capabilities flags specific to X722.
Signed-off-by: NAnjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: NCatherine Sullivan <catherine.sullivan@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Use macros for abstracting (1 << foo) to BIT(foo)
and (1ULL << foo64) to BIT_ULL(foo64) in order to match
better with kernel requirements.

NOTE: the adminq_cmd.h file was not modified on purpose because
of the dependency upon firmware for that file.

Change-ID: I73ee2e48c880d671948aad19bd53ca6b2ac558fc
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NCatherine Sullivan <catherine.sullivan@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The most common type of reset that the VF will encounter is a PF reset
that cascades down into a VF reset for each VF. In this case, the VF
will always be assigned the same VSI and recovery is fairly simple.

However, in the case of 'bigger' resets, such as a Core or EMP reset,
when the device is reinitialized, it's probable that the VF will NOT get
the same VSI. When this happens, the VF will not be able to recover, as
it will continue to request resources for its original VSI.

Add an extra state to the admin queue state machine so that the driver
can re-request its configuration information at runtime. During reset
recovery, set this bit in the aq_required field, and fetch the (possibly
new) configuration information before attempting to bring the driver
back up. Since the driver doesn't know what kind of reset it has
encountered, this step is done even for a PF reset, but it doesn't hurt
anything - it just gets the same VSI back.

Change-ID: I915d59ffb40375215117362f4ac7a37811aba748
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Now that we've rolled the virtual channel API version to 1.1, add some
macros to test what version is being used by our partner in crime. For the
VF, add some macros to determine what our device capabilities are.

Change-ID: I79f6683d4c23bd76a8ad9fd492776fcc1208e1dc
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Down was requesting queue disables, but then exited immediately
without waiting for the queues to actually disable.  This could
allow any function called after i40evf_down to run immediately,
including i40evf_up, and causes a memory leak.

Removing the whole reinit_locked function is the best way
to go about this, and allows for the driver to handle the
state changes by requesting reset from the periodic timer.

Also, add a couple WARN_ONs in slow path to help us recognize
if we re-introduce this issue or missed any cases.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The aq_pending field in the adapter structure is actually redundant with
the current_op field. Remove the aq_pending field and expunge all traces
of it from the official record. This simplifies the code significantly,
especially in the virtual channel completion routine.

Change-ID: Ib2957c8c19882bd0cecc6fcd133912c24b46a1ff
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

When the VF interface is closed, we cannot immediately free our rings
and RX buffers, because the hardware hasn't yet stopped accessing this
memory. This shows up as a panic or memory corruption when the device is
brought down while under heavy stress.

To fix this, delay releasing resources until we receive acknowledgment
from the PF driver that the rings have indeed been stopped. Because of
this delay, we also need to check to make sure that all of our admin
queue requests have been handled before allowing the device to be
opened.

Change-ID: I44edd35529ce2fa2a9512437a3a8e6f14ed8ed63
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NJim Young <james.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch is the result of running checkpatch on the i40evf driver with
the --strict option. The vast majority of changes are adding/removing
blank lines, aligning function parameters, and correcting over-long
lines.

The only possible functional change is changing the flags member of the
adapter structure to be non-volatile. However, according to the kernel
documentation, this is not necessary and the volatile should be removed.

Change-ID: Ie8c6414800924f529bef831e8845292b970fe2ed
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

It's kind of silly to configure and attempt to use a bunch of queue
pairs when you're running on a single (virtual) CPU. Instead of
unconditionally configuring all of the queues that the PF gives us,
clamp the number of queue pairs to the number of CPUs.

Change-ID: I321714c9e15072ee76de8f95ab9a81f86ed347d1
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NPatrick Lu <patrick.lu@intel.com>
Tested-by: NJim Young <jamesx.m.young@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

i40evf_set_ringparam was broken in several ways. First, it only changed
the size of the first ring, and second, changing the ring size would
often result in a panic because it would change the count before
deallocating resources, causing the driver to either free nonexistent
buffers, or leak leftover buffers.

Fix this by storing the descriptor count in the adapter structure, and
updating the count for each ring each time we allocate them. This
ensures that we always free the right size ring, and always end up with
the requested count when the device is (re)opened.

Change-ID: I298396cd3d452ba8509d9f2d33a93f25868a9a55
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Hardware requires descriptors to be allocated in groups of 32.

Change-ID: I752ccc96769d1bd8d3018c004b8aeff464045bf2
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NSibai Li <sibai.li@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Adding the appropriate GNU General Public License header and
update copyright year to 2014.

Change-ID: I769dd2d37d70350afd0c8727ae2859c0fd340361
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: NCatherine Sullivan <catherine.sullivan@intel.com>
Tested-by: NSibai Li <sibai.li@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The Tx watchdog handler runs in interrupt context, so it would cause an
oops when sending an admin queue message to request a reset, because the
admin queue functions use spinlocks.

Instead, set a flag and let the reset task handle sending the request.

Change-ID: I65879470b72963d9c308edfb8f45ac4fbba2c14f
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NCatherine Sullivan <catherine.sullivan@intel.com>
Tested-by: NSibai Li <sibai.li@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Keep the descriptor ring size in the actual ring structs instead of in
the adapter struct. This enables us to use common tx and rx code with
the i40e PF driver.

Also update copyrights.

Change-ID: I2861e599b2b4c76441c062ea14400f4750f54d0e
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: NSibai Li <sibai.li@intel.com>
Signed-off-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>