Skip to content

R
raspberrypi-kernel

openeuler / raspberrypi-kernel

通知 14
Star 1
Fork 0
R
raspberrypi-kernel
切换分支/标签
  1. 28 8月, 2017 8 次提交
    • J
      i40e/i40evf: avoid dynamic ITR updates when polling or low packet rate · 742c9875
      由 Jacob Keller 提交于 
      The dynamic ITR algorithm depends on a calculation of usecs which
assumes that the interrupts have been firing constantly at the interrupt
throttle rate. This is not guaranteed because we could have a low packet
rate, or have been polling in software.

We'll estimate whether this is the case by using jiffies to determine if
we've been too long. If the time difference of jiffies is larger we are
guaranteed to have an incorrect calculation. If the time difference of
jiffies is smaller we might have been polling some but the difference
shouldn't affect the calculation too much.

This ensures that we don't get stuck in BULK latency during certain rare
situations where we receive bursts of packets that force us into NAPI
polling.
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>
      742c9875
    • J
      i40e/i40evf: remove ULTRA latency mode · 0a2c7722
      由 Jacob Keller 提交于 
      Since commit c56625d5 ("i40e/i40evf: change dynamic interrupt
thresholds") a new higher latency ITR setting called I40E_ULTRA_LATENCY
was added with a cryptic comment about how it was meant for adjusting Rx
more aggressively when streaming small packets.

This mode was attempting to calculate packets per second and then kick
in when we have a huge number of small packets.

Unfortunately, the ULTRA setting was kicking in for workloads it wasn't
intended for including single-thread UDP_STREAM workloads.

This wasn't caught for a variety of reasons. First, the ip_defrag
routines were improved somewhat which makes the UDP_STREAM test still
reasonable at 10GbE, even when dropped down to 8k interrupts a second.
Additionally, some other obvious workloads appear to work fine, such
as TCP_STREAM.

The number 40k doesn't make sense for a number of reasons. First, we
absolutely can do more than 40k packets per second. Second, we calculate
the value inline in an integer, which sometimes can overflow resulting
in using incorrect values.

If we fix this overflow it makes it even more likely that we'll enter
ULTRA mode which is the opposite of what we want.

The ULTRA mode was added originally as a way to reduce CPU utilization
during a small packet workload where we weren't keeping up anyways. It
should never have been kicking in during these other workloads.

Given the issues outlined above, let's remove the ULTRA latency mode. If
necessary, a better solution to the CPU utilization issue for small
packet workloads will be added in a future patch.
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>
      0a2c7722
    • J
      i40e: invert logic for checking incorrect cpu vs irq affinity · 6d977729
      由 Jacob Keller 提交于 
      In commit 96db776a ("i40e/vf: fix interrupt affinity bug")
we added some code to force exit of polling in case we did
not have the correct CPU. This is important since it was possible for
the IRQ affinity to be changed while the CPU is pegged at 100%. This can
result in the polling routine being stuck on the wrong CPU until
traffic finally stops.

Unfortunately, the implementation, "if the CPU is correct, exit as
normal, otherwise, fall-through to the end-polling exit" is incredibly
confusing to reason about. In this case, the normal flow looks like the
exception, while the exception actually occurs far away from the if
statement and comment.

We recently discovered and fixed a bug in this code because we were
incorrectly initializing the affinity mask.

Re-write the code so that the exceptional case is handled at the check,
rather than having the logic be spread through the regular exit flow.
This does end up with minor code duplication, but the resulting code is
much easier to reason about.

The new logic is identical, but inverted. If we are running on a CPU not
in our affinity mask, we'll exit polling. However, the code flow is much
easier to understand.

Note that we don't actually have to check for MSI-X, because in the MSI
case we'll only have one q_vector, but its default affinity mask should
be correct as it includes all CPUs when it's initialized. Further, we
could at some point add code to setup the notifier for the non-MSI-X
case and enable this workaround for that case too, if desired, though
there isn't much gain since its unlikely to be the common case.
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>
      6d977729
    • J
      i40e: initialize our affinity_mask based on cpu_possible_mask · 759dc4a7
      由 Jacob Keller 提交于 
      On older kernels a call to irq_set_affinity_hint does not guarantee that
the IRQ affinity will be set. If nothing else on the system sets the IRQ
affinity this can result in a bug in the i40e_napi_poll() routine where
we notice that our interrupt fired on the "wrong" CPU according to our
internal affinity_mask variable.

This results in a bug where we continuously tell NAPI to stop polling to
move the interrupt to a new CPU, but the CPU never changes because our
affinity mask does not match the actual mask setup for the IRQ.

The root problem is a mismatched affinity mask value. So lets initialize
the value to cpu_possible_mask instead. This ensures that prior to the
first time we get an IRQ affinity notification we'll have the mask set
to include every possible CPU.

We use cpu_possible_mask instead of cpu_online_mask since the former is
almost certainly never going to change, while the later might change
after we've made a copy.
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>
      759dc4a7
    • M
      i40e/i40evf: support for VF VLAN tag stripping control · 8774370d
      由 Mariusz Stachura 提交于 
      This patch gives VF capability to control VLAN tag stripping via
ethtool. As rx-vlan-offload was fixed before, now the VF is able to
change it using "ethtool --offload <IF> rxvlan on/off" settings.
Signed-off-by: NMariusz Stachura <mariusz.stachura@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
      8774370d
    • J
      i40evf: fix possible snprintf truncation of q_vector->name · 696ac80a
      由 Jacob Keller 提交于 
      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>
      696ac80a
    • J
      i40e: prevent snprintf format specifier truncation · b5d5504a
      由 Jacob Keller 提交于 
      Increase the size of the prefix buffer so that it can hold enough
characters for every possible input. Although 20 is enough for all
expected inputs, it is possible for the values to be larger than
expected, resulting in a possibly truncated string. Additionally, lets
use sizeof(prefix) in order to ensure we use the correct size if we need
to change the array length in the future.

New versions of GCC starting at 7 now include warnings to prevent
truncation unless you handle the return code. At most 27 bytes can be
written here, so lets just increase the buffer size even if for all
expected hw->bus.* values we only needed 20.
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>
      b5d5504a
    • M
      i40e: Store the requested FEC information · ed601f66
      由 Mariusz Stachura 提交于 
      Store information about FEC modes, that were requested. It will be used
in printing link status information function and this way there is no
need to call admin queue there.
Signed-off-by: NMariusz Stachura <mariusz.stachura@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
      ed601f66
  3. 26 8月, 2017 7 次提交
  5. 26 7月, 2017 4 次提交
  7. 21 6月, 2017 2 次提交
  9. 06 6月, 2017 1 次提交
  11. 02 6月, 2017 6 次提交
  13. 31 5月, 2017 3 次提交
  15. 30 4月, 2017 7 次提交
  17. 20 4月, 2017 2 次提交
    • J
      i40e: remove I40E_FLAG_IN_NETPOLL entirely · 1de81c2d
      由 Jacob Keller 提交于 
      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: 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>
      1de81c2d
    • J
      i40e: reduce wait time for adminq command completion · 9e3f23f4
      由 Jacob Keller 提交于 
      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: 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>
      9e3f23f4