Merge branch 'suspend-irqs-during-application-busy-periods'

Joe Damato says:

====================
Suspend IRQs during application busy periods

This series introduces a new mechanism, IRQ suspension, which allows
network applications using epoll to mask IRQs during periods of high
traffic while also reducing tail latency (compared to existing
mechanisms, see below) during periods of low traffic. In doing so, this
balances CPU consumption with network processing efficiency.

Martin Karsten (CC'd) and I have been collaborating on this series for
several months and have appreciated the feedback from the community on
our RFC [1]. We've updated the cover letter and kernel documentation in
an attempt to more clearly explain how this mechanism works, how
applications can use it, and how it compares to existing mechanisms in
the kernel.

I briefly mentioned this idea at netdev conf 2024 (for those who were
there) and Martin described this idea in an earlier paper presented at
Sigmetrics 2024 [2].

~ The short explanation (TL;DR)

We propose adding a new napi config parameter: irq_suspend_timeout to
help balance CPU usage and network processing efficiency when using IRQ
deferral and napi busy poll.

If this parameter is set to a non-zero value *and* a user application
has enabled preferred busy poll on a busy poll context (via the
EPIOCSPARAMS ioctl introduced in commit 18e2bf0edf4d ("eventpoll: Add
epoll ioctl for epoll_params")), then application calls to epoll_wait
for that context will cause device IRQs and softirq processing to be
suspended as long as epoll_wait successfully retrieves data from the
NAPI. Each time data is retrieved, the irq_suspend_timeout is deferred.

If/when network traffic subsides and epoll_wait returns no data, IRQ
suspension is immediately reverted back to the existing
napi_defer_hard_irqs and gro_flush_timeout mechanism which was
introduced in commit 6f8b12d661d0 ("net: napi: add hard irqs deferral
feature")).

The irq_suspend_timeout serves as a safety mechanism. If userland takes
a long time processing data, irq_suspend_timeout will fire and restart
normal NAPI processing.

For a more in depth explanation, please continue reading.

~ Comparison with existing mechanisms

Interrupt mitigation can be accomplished in napi software, by setting
napi_defer_hard_irqs and gro_flush_timeout, or via interrupt coalescing
in the NIC. This can be quite efficient, but in both cases, a fixed
timeout (or packet count) needs to be configured. However, a fixed
timeout cannot effectively support both low- and high-load situations:

At low load, an application typically processes a few requests and then
waits to receive more input data. In this scenario, a large timeout will
cause unnecessary latency.

At high load, an application typically processes many requests before
being ready to receive more input data. In this case, a small timeout
will likely fire prematurely and trigger irq/softirq processing, which
interferes with the application's execution. This causes overhead, most
likely due to cache contention.

While NICs attempt to provide adaptive interrupt coalescing schemes,
these cannot properly take into account application-level processing.

An alternative packet delivery mechanism is busy-polling, which results
in perfect alignment of application processing and network polling. It
delivers optimal performance (throughput and latency), but results in
100% cpu utilization and is thus inefficient for below-capacity
workloads.

We propose to add a new packet delivery mode that properly alternates
between busy polling and interrupt-based delivery depending on busy and
idle periods of the application. During a busy period, the system
operates in busy-polling mode, which avoids interference. During an idle
period, the system falls back to interrupt deferral, but with a small
timeout to avoid excessive latencies. This delivery mode can also be
viewed as an extension of basic interrupt deferral, but alternating
between a small and a very large timeout.

This delivery mode is efficient, because it avoids softirq execution
interfering with application processing during busy periods. It can be
used with blocking epoll_wait to conserve cpu cycles during idle
periods. The effect of alternating between busy and idle periods is that
performance (throughput and latency) is very close to full busy polling,
while cpu utilization is lower and very close to interrupt mitigation.

~ Usage details

IRQ suspension is introduced via a per-NAPI configuration parameter that
controls the maximum time that IRQs can be suspended.

Here's how it is intended to work:
  - The user application (or system administrator) uses the netdev-genl
    netlink interface to set the pre-existing napi_defer_hard_irqs and
    gro_flush_timeout NAPI config parameters to enable IRQ deferral.

  - The user application (or system administrator) sets the proposed
    irq_suspend_timeout parameter via the netdev-genl netlink interface
    to a larger value than gro_flush_timeout to enable IRQ suspension.

  - The user application issues the existing epoll ioctl to set the
    prefer_busy_poll flag on the epoll context.

  - The user application then calls epoll_wait to busy poll for network
    events, as it normally would.

  - If epoll_wait returns events to userland, IRQs are suspended for the
    duration of irq_suspend_timeout.

  - If epoll_wait finds no events and the thread is about to go to
    sleep, IRQ handling using napi_defer_hard_irqs and gro_flush_timeout
    is resumed.

As long as epoll_wait is retrieving events, IRQs (and softirq
processing) for the NAPI being polled remain disabled. When network
traffic reduces, eventually a busy poll loop in the kernel will retrieve
no data. When this occurs, regular IRQ deferral using gro_flush_timeout
for the polled NAPI is re-enabled.

Unless IRQ suspension is continued by subsequent calls to epoll_wait, it
automatically times out after the irq_suspend_timeout timer expires.
Regular deferral is also immediately re-enabled when the epoll context
is destroyed.

~ Usage scenario

The target scenario for IRQ suspension as packet delivery mode is a
system that runs a dominant application with substantial network I/O.
The target application can be configured to receive input data up to a
certain batch size (via epoll_wait maxevents parameter) and this batch
size determines the worst-case latency that application requests might
experience. Because packet delivery is suspended during the target
application's processing, the batch size also determines the worst-case
latency of concurrent applications using the same RX queue(s).

gro_flush_timeout should be set as small as possible, but large enough to
make sure that a single request is likely not being interfered with.

irq_suspend_timeout is largely a safety mechanism against misbehaving
applications. It should be set large enough to cover the processing of an
entire application batch, i.e., the factor between gro_flush_timeout and
irq_suspend_timeout should roughly correspond to the maximum batch size
that the target application would process in one go.

~ Important call out in the implementation

  - Enabling per epoll-context preferred busy poll will now effectively
    lead to a nonblocking iteration through napi_busy_loop, even when
    busy_poll_usecs is 0. See patch 4.

~ Benchmark configs & descriptions

The changes were benchmarked with memcached [3] using the benchmarking
tool mutilate [4].

To facilitate benchmarking, a small patch [5] was applied to memcached
1.6.29 to allow setting per-epoll context preferred busy poll and other
settings via environment variables. Another small patch [6] was applied
to libevent to enable full busy-polling.

Multiple scenarios were benchmarked as described below and the scripts
used for producing these results can be found on github [7] (note: all
scenarios use NAPI-based traffic splitting via SO_INCOMING_ID by passing
-N to memcached):

  - base:
    - no other options enabled
  - deferX:
    - set defer_hard_irqs to 100
    - set gro_flush_timeout to X,000
  - napibusy:
    - set defer_hard_irqs to 100
    - set gro_flush_timeout to 200,000
    - enable busy poll via the existing ioctl (busy_poll_usecs = 64,
      busy_poll_budget = 64, prefer_busy_poll = true)
  - fullbusy:
    - set defer_hard_irqs to 100
    - set gro_flush_timeout to 5,000,000
    - enable busy poll via the existing ioctl (busy_poll_usecs = 1000,
      busy_poll_budget = 64, prefer_busy_poll = true)
    - change memcached's nonblocking epoll_wait invocation (via
      libevent) to using a 1 ms timeout
  - suspend0:
    - set defer_hard_irqs to 0
    - set gro_flush_timeout to 0
    - set irq_suspend_timeout to 20,000,000
    - enable busy poll via the existing ioctl (busy_poll_usecs = 0,
      busy_poll_budget = 64, prefer_busy_poll = true)
  - suspendX:
    - set defer_hard_irqs to 100
    - set gro_flush_timeout to X,000
    - set irq_suspend_timeout to 20,000,000
    - enable busy poll via the existing ioctl (busy_poll_usecs = 0,
      busy_poll_budget = 64, prefer_busy_poll = true)

~ Benchmark results

Tested on:

Single socket AMD EPYC 7662 64-Core Processor
Hyperthreading disabled
4 NUMA Zones (NPS=4)
16 CPUs per NUMA zone (64 cores total)
2 x Dual port 100gbps Mellanox Technologies ConnectX-5 Ex EN NIC

The test machine is configured such that a single interface has 8 RX
queues. The queues' IRQs and memcached are pinned to CPUs that are
NUMA-local to the interface which is under test. The NIC's interrupt
coalescing configuration is left at boot-time defaults.

Results:

Results are shown below. The mechanism added by this series is
represented by the 'suspend' cases. Data presented shows a summary over
nearly 10 runs of each test case [8] using the scripts on github [7].
For latency, the median is shown. For throughput and CPU utilization,
the average is shown.

The results also include cycles-per-query (cpq) and
instruction-per-query (ipq) metrics, following the methodology proposed
in [2], to augment the CPU utilization numbers, which could be skewed
due to frequency scaling. We find that this does not appear to be the
case as CPU utilization and low-level metrics show similar trends.

These results were captured using the scripts on github [7] to
illustrate how this approach compares with other pre-existing
mechanisms. This data is not to be interpreted as scientific data
captured in a fully isolated lab setting, but instead as best effort,
illustrative information comparing and contrasting tradeoffs.

The absolute QPS results shift between submissions, but the
relative differences are equivalent. As patches are rebased,
several factors likely influence overall performance.

Compare:
- Throughput (MAX) and latencies of base vs suspend.
- CPU usage of napibusy and fullbusy during lower load (200K, 400K for
  example) vs suspend.
- Latency of the defer variants vs suspend as timeout and load
  increases.
- suspend0, which sets defer_hard_irqs and gro_flush_timeout to 0, has
  nearly the same performance as the base case (this is FAQ item #1).

The overall takeaway is that the suspend variants provide a superior
combination of high throughput, low latency, and low cpu utilization
compared to all other variants. Each of the suspend variants works very
well, but some fine-tuning between latency and cpu utilization is still
possible by tuning the small timeout (gro_flush_timeout).

Note: we've reorganized the results to make comparison among testcases
with the same load easier.

  testcase  load     qps  avglat  95%lat  99%lat     cpu     cpq     ipq
      base  200K  199946     112     239     416      26   12973   11343
   defer10  200K  199971      54     124     142      29   19412   17460
   defer20  200K  199986      60     130     153      26   15644   14095
   defer50  200K  200025      79     144     182      23   12122   11632
  defer200  200K  199999     164     254     309      19    8923    9635
  fullbusy  200K  199998      46     118     133     100   43658   23133
  napibusy  200K  199983     100     237     277      56   24840   24716
  suspend0  200K  200020     105     249     432      30   14264   11796
 suspend10  200K  199950      53     123     141      32   19518   16903
 suspend20  200K  200037      58     126     151      30   16426   14736
 suspend50  200K  199961      73     136     177      26   13310   12633
suspend200  200K  199998     149     251     306      21    9566   10203

  testcase  load     qps  avglat  95%lat  99%lat     cpu     cpq     ipq
      base  400K  400014     139     269     707      41    9476    9343
   defer10  400K  400016      59     133     166      53   13991   12989
   defer20  400K  399952      67     140     172      47   12063   11644
   defer50  400K  400007      87     162     198      39    9384    9880
  defer200  400K  399979     181     274     330      31    7089    8430
  fullbusy  400K  399987      50     123     156     100   21827   16037
  napibusy  400K  400014      76     222     272      83   18185   16529
  suspend0  400K  400015     127     350     776      47   10699    9603
 suspend10  400K  400023      57     129     164      54   13758   13178
 suspend20  400K  400043      62     135     169      49   12071   11826
 suspend50  400K  400071      76     149     186      42   10011   10301
suspend200  400K  399961     154     269     327      34    7827    8774

  testcase  load     qps  avglat  95%lat  99%lat     cpu     cpq     ipq
      base  600K  599951     149     266     574      61    9265    8876
   defer10  600K  600006      71     147     203      76   11866   10936
   defer20  600K  600123      76     152     203      66   10430   10342
   defer50  600K  600162      95     172     217      54    8526    9142
  defer200  600K  599942     200     301     357      46    6977    8212
  fullbusy  600K  599990      55     127     177     100   14551   13983
  napibusy  600K  600035      63     160     250      96   13937   14140
  suspend0  600K  599903     127     320     732      68   10166    8963
 suspend10  600K  599908      63     137     192      69   10902   11100
 suspend20  600K  599961      66     141     194      65    9976   10370
 suspend50  600K  599973      80     159     204      57    8678    9381
suspend200  600K  600010     157     277     346      48    7133    8381

  testcase  load     qps  avglat  95%lat  99%lat     cpu     cpq     ipq
      base  800K  800039     181     300     536      87    9585    8304
   defer10  800K  800038     181     530     939      96   10564    8970
   defer20  800K  800029     112     225     329      90   10056    8935
   defer50  800K  799999     120     208     296      82    9234    8562
  defer200  800K  800066     227     338     401      63    7117    8129
  fullbusy  800K  800040      61     134     190     100   10913   12608
  napibusy  800K  799944      64     141     214      99   10828   12588
  suspend0  800K  799911     126     248     509      85    9346    8498
 suspend10  800K  800006      69     143     200      83    9410    9845
 suspend20  800K  800120      74     150     207      78    8786    9454
 suspend50  800K  799989      87     168     224      71    7946    8833
suspend200  800K  799987     160     292     357      62    6923    8229

  testcase  load     qps  avglat  95%lat  99%lat     cpu     cpq     ipq
      base 1000K  906879    4079    5751    6216      98    9496    7904
   defer10 1000K  860849    3643    6274    6730      99   10040    8676
   defer20 1000K  896063    3298    5840    6349      98    9620    8237
   defer50 1000K  919782    2962    5513    5807      97    9284    7951
  defer200 1000K  970941    3059    5348    5984      95    8593    7959
  fullbusy 1000K  999950      70     150     207     100    8732   10777
  napibusy 1000K  999996      78     154     223     100    8722   10656
  suspend0 1000K  949706    2666    5770    6660      99    9071    8046
 suspend10 1000K 1000024      80     160     220      92    8137    9035
 suspend20 1000K 1000059      83     165     226      89    7850    8804
 suspend50 1000K  999955      95     180     240      84    7411    8459
suspend200 1000K  999914     163     299     366      77    6833    8078

  testcase  load     qps  avglat  95%lat  99%lat     cpu     cpq     ipq
      base   MAX 1037654    4184    5453    5810     100    8411    7938
   defer10   MAX  905607    4840    6151    6380     100    9639    8431
   defer20   MAX  986463    4455    5594    5796     100    8848    8110
   defer50   MAX 1077030    4000    5073    5299     100    8104    7920
  defer200   MAX 1040728    4152    5385    5765     100    8379    7849
  fullbusy   MAX 1247536    3518    3935    3984     100    6998    7930
  napibusy   MAX 1136310    3799    7756    9964     100    7670    7877
  suspend0   MAX 1057509    4132    5724    6185     100    8253    7918
 suspend10   MAX 1215147    3580    3957    4041     100    7185    7944
 suspend20   MAX 1216469    3576    3953    3988     100    7175    7950
 suspend50   MAX 1215871    3577    3961    4075     100    7181    7949
suspend200   MAX 1216882    3556    3951    3988     100    7175    7955

~ FAQ

  - Why is a new parameter needed? Does irq_suspend_timeout override
    gro_flush_timeout?

    Using the suspend mechanism causes the system to alternate between
    polling mode and irq-driven packet delivery. During busy periods,
    irq_suspend_timeout overrides gro_flush_timeout and keeps the system
    busy polling, but when epoll finds no events, the setting of
    gro_flush_timeout and napi_defer_hard_irqs determine the next step.

    There are essentially three possible loops for network processing and
    packet delivery:

    1) hardirq -> softirq -> napi poll; basic interrupt delivery
    2) timer -> softirq -> napi poll; deferred irq processing
    3) epoll -> busy-poll -> napi poll; busy looping

    Loop 2 can take control from Loop 1, if gro_flush_timeout and
    napi_defer_hard_irqs are set.

    If gro_flush_timeout and napi_defer_hard_irqs are set, Loops 2 and
    3 "wrestle" with each other for control. During busy periods,
    irq_suspend_timeout is used as timer in Loop 2, which essentially
    tilts this in favour of Loop 3.

    If gro_flush_timeout and napi_defer_hard_irqs are not set, Loop 3
    cannot take control from Loop 1.

    Therefore, setting gro_flush_timeout and napi_defer_hard_irqs is the
    recommended usage, because otherwise setting irq_suspend_timeout
    might not have any discernible effect.

    This is shown in the results above: compare suspend0 with the base
    case. Note that the lack of napi_defer_hard_irqs and
    gro_flush_timeout produce similar results for both, which encourages
    the use of napi_defer_hard_irqs and gro_flush_timeout in addition to
    irq_suspend_timeout.

  - Can the new timeout value be threaded through the new epoll ioctl ?

    It is possible, but presents challenges for userspace. User
    applications must ensure that the file descriptors added to epoll
    contexts have the same NAPI ID to support busy polling.

    An epoll context is not permanently tied to any particular NAPI ID.
    So, a user application could decide to clear the file descriptors
    from the context and add a new set of file descriptors with a
    different NAPI ID to the context. Busy polling would work as
    expected, but the meaning of the suspend timeout becomes ambiguous
    because IRQs are not inherently associated with epoll contexts, but
    rather with the NAPI. The user program would need to reissue the
    ioctl to set the irq_suspend_timeout, but the napi_defer_hard_irqs
    and gro_flush_timeout settings would come from the NAPI's
    napi_config (which are set either by sysfs or by netlink). Such an
    interface seems awkard to use from a user perspective.

    Further, IRQs are related to NAPIs, which is why they are stored in
    the napi_config space. Putting the irq_suspend_timeout in
    the epoll context while other IRQ deferral mechanisms remain in the
    NAPI's napi_config space seems like an odd design choice.

    We've opted to keep all of the IRQ deferral parameters together and
    place the irq_suspend_timeout in napi_config. This has nice benefits
    for userspace: if a user app were to remove all file descriptors
    from an epoll context and add new file descriptors with a new NAPI ID,
    the correct suspend timeout for that NAPI ID would be used automatically
    without the user application needing to do anything (like re-issuing an
    ioctl, for example). All IRQ deferral related parameters are in one
    place and can all be set the same way: with netlink.

  - Can irq suspend be built by combining NIC coalescing and
    gro_flush_timeout ?

    No. The problem is that the long timeout must engage if and only if
    prefer-busy is active.

    When using NIC coalescing for the short timeout (without
    napi_defer_hard_irqs/gro_flush_timeout), an interrupt after an idle
    period will trigger softirq, which will run napi polling. At this
    point, prefer-busy is not active, so NIC interrupts would be
    re-enabled. Then it is not possible for the longer timeout to
    interject to switch control back to polling. In other words, only by
    using the software timer for the short timeout, it is possible to
    extend the timeout without having to reprogram the NIC timer or
    reach down directly and disable interrupts.

    Using gro_flush_timeout for the long timeout also has problems, for
    the same underlying reason. In the current napi implementation,
    gro_flush_timeout is not tied to prefer-busy. We'd either have to
    change that and in the process modify the existing deferral
    mechanism, or introduce a state variable to determine whether
    gro_flush_timeout is used as long timeout for irq suspend or whether
    it is used for its default purpose. In an earlier version, we did
    try something similar to the latter and made it work, but it ends up
    being a lot more convoluted than our current proposal.

  - Isn't it already possible to combine busy looping with irq deferral?

    Yes, in fact enabling irq deferral via napi_defer_hard_irqs and
    gro_flush_timeout is a precondition for prefer_busy_poll to have an
    effect. If the application also uses a tight busy loop with
    essentially nonblocking epoll_wait (accomplished with a very short
    timeout parameter), this is the fullbusy case shown in the results.
    An application using blocking epoll_wait is shown as the napibusy
    case in the results. It's a hybrid approach that provides limited
    latency benefits compared to the base case and plain irq deferral,
    but not as good as fullbusy or suspend.

~ Special thanks

Several people were involved in earlier stages of the development of this
mechanism whom we'd like to thank:

  - Peter Cai (CC'd), for the initial kernel patch and his contributions
    to the paper.

  - Mohammadamin Shafie (CC'd), for testing various versions of the kernel
    patch and providing helpful feedback.

Thanks,
Martin and Joe

[1]: https://lore.kernel.org/netdev/20240812125717.413108-1-jdamato@fastly.com/
[2]: https://doi.org/10.1145/3626780
[3]: https://github.com/memcached/memcached/blob/master/doc/napi_ids.txt
[4]: https://github.com/leverich/mutilate
[5]: https://raw.githubusercontent.com/martinkarsten/irqsuspend/main/patches/memcached.patch
[6]: https://raw.githubusercontent.com/martinkarsten/irqsuspend/main/patches/libevent.patch
[7]: https://github.com/martinkarsten/irqsuspend
[8]: https://github.com/martinkarsten/irqsuspend/tree/main/results

v8: https://lore.kernel.org/20241108045337.292905-1-jdamato@fastly.com
v7: https://lore.kernel.org/20241108023912.98416-1-jdamato@fastly.com
v6: https://lore.kernel.org/20241104215542.215919-1-jdamato@fastly.com
v5: https://lore.kernel.org/20241103052421.518856-1-jdamato@fastly.com
v4: https://lore.kernel.org/20241102005214.32443-1-jdamato@fastly.com
v3: https://lore.kernel.org/20241101004846.32532-1-jdamato@fastly.com
v2: https://lore.kernel.org/20241021015311.95468-1-jdamato@fastly.com
====================

Link: https://patch.msgid.link/20241109050245.191288-1-jdamato@fastly.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

15 files changed, 816 insertions, 6 deletions

diff --git a/Documentation/netlink/specs/netdev.yaml b/Documentation/netlink/specs/netdev.yaml
index f9cb97d6106c..cbb544bd6c84 100644
--- a/Documentation/netlink/specs/netdev.yaml
+++ b/Documentation/netlink/specs/netdev.yaml
@@ -263,6 +263,11 @@ attribute-sets:
              the end of a NAPI cycle. This may add receive latency in exchange
              for reducing the number of frames processed by the network stack.
         type: uint
+      -
+        name: irq-suspend-timeout
+        doc: The timeout, in nanoseconds, of how long to suspend irq
+             processing, if event polling finds events
+        type: uint
   -
     name: queue
     attributes:
@@ -653,6 +658,7 @@ operations:
             - pid
             - defer-hard-irqs
             - gro-flush-timeout
+            - irq-suspend-timeout
       dump:
         request:
           attributes:
@@ -704,6 +710,7 @@ operations:
             - id
             - defer-hard-irqs
             - gro-flush-timeout
+            - irq-suspend-timeout
 
 kernel-family:
   headers: [ "linux/list.h"]
diff --git a/Documentation/networking/napi.rst b/Documentation/networking/napi.rst
index dfa5d549be9c..02720dd71a76 100644
--- a/Documentation/networking/napi.rst
+++ b/Documentation/networking/napi.rst
@@ -192,6 +192,33 @@ is reused to control the delay of the timer, while
 ``napi_defer_hard_irqs`` controls the number of consecutive empty polls
 before NAPI gives up and goes back to using hardware IRQs.
 
+The above parameters can also be set on a per-NAPI basis using netlink via
+netdev-genl. When used with netlink and configured on a per-NAPI basis, the
+parameters mentioned above use hyphens instead of underscores:
+``gro-flush-timeout`` and ``napi-defer-hard-irqs``.
+
+Per-NAPI configuration can be done programmatically in a user application
+or by using a script included in the kernel source tree:
+``tools/net/ynl/cli.py``.
+
+For example, using the script:
+
+.. code-block:: bash
+
+  $ kernel-source/tools/net/ynl/cli.py \
+            --spec Documentation/netlink/specs/netdev.yaml \
+            --do napi-set \
+            --json='{"id": 345,
+                     "defer-hard-irqs": 111,
+                     "gro-flush-timeout": 11111}'
+
+Similarly, the parameter ``irq-suspend-timeout`` can be set using netlink
+via netdev-genl. There is no global sysfs parameter for this value.
+
+``irq-suspend-timeout`` is used to determine how long an application can
+completely suspend IRQs. It is used in combination with SO_PREFER_BUSY_POLL,
+which can be set on a per-epoll context basis with ``EPIOCSPARAMS`` ioctl.
+
 .. _poll:
 
 Busy polling
@@ -207,6 +234,46 @@ selected sockets or using the global ``net.core.busy_poll`` and
 ``net.core.busy_read`` sysctls. An io_uring API for NAPI busy polling
 also exists.
 
+epoll-based busy polling
+------------------------
+
+It is possible to trigger packet processing directly from calls to
+``epoll_wait``. In order to use this feature, a user application must ensure
+all file descriptors which are added to an epoll context have the same NAPI ID.
+
+If the application uses a dedicated acceptor thread, the application can obtain
+the NAPI ID of the incoming connection using SO_INCOMING_NAPI_ID and then
+distribute that file descriptor to a worker thread. The worker thread would add
+the file descriptor to its epoll context. This would ensure each worker thread
+has an epoll context with FDs that have the same NAPI ID.
+
+Alternatively, if the application uses SO_REUSEPORT, a bpf or ebpf program can
+be inserted to distribute incoming connections to threads such that each thread
+is only given incoming connections with the same NAPI ID. Care must be taken to
+carefully handle cases where a system may have multiple NICs.
+
+In order to enable busy polling, there are two choices:
+
+1. ``/proc/sys/net/core/busy_poll`` can be set with a time in useconds to busy
+   loop waiting for events. This is a system-wide setting and will cause all
+   epoll-based applications to busy poll when they call epoll_wait. This may
+   not be desirable as many applications may not have the need to busy poll.
+
+2. Applications using recent kernels can issue an ioctl on the epoll context
+   file descriptor to set (``EPIOCSPARAMS``) or get (``EPIOCGPARAMS``) ``struct
+   epoll_params``:, which user programs can define as follows:
+
+.. code-block:: c
+
+  struct epoll_params {
+      uint32_t busy_poll_usecs;
+      uint16_t busy_poll_budget;
+      uint8_t prefer_busy_poll;
+
+      /* pad the struct to a multiple of 64bits */
+      uint8_t __pad;
+  };
+
 IRQ mitigation
 ---------------
 
@@ -222,12 +289,111 @@ Such applications can pledge to the kernel that they will perform a busy
 polling operation periodically, and the driver should keep the device IRQs
 permanently masked. This mode is enabled by using the ``SO_PREFER_BUSY_POLL``
 socket option. To avoid system misbehavior the pledge is revoked
-if ``gro_flush_timeout`` passes without any busy poll call.
+if ``gro_flush_timeout`` passes without any busy poll call. For epoll-based
+busy polling applications, the ``prefer_busy_poll`` field of ``struct
+epoll_params`` can be set to 1 and the ``EPIOCSPARAMS`` ioctl can be issued to
+enable this mode. See the above section for more details.
 
 The NAPI budget for busy polling is lower than the default (which makes
 sense given the low latency intention of normal busy polling). This is
 not the case with IRQ mitigation, however, so the budget can be adjusted
-with the ``SO_BUSY_POLL_BUDGET`` socket option.
+with the ``SO_BUSY_POLL_BUDGET`` socket option. For epoll-based busy polling
+applications, the ``busy_poll_budget`` field can be adjusted to the desired value
+in ``struct epoll_params`` and set on a specific epoll context using the ``EPIOCSPARAMS``
+ioctl. See the above section for more details.
+
+It is important to note that choosing a large value for ``gro_flush_timeout``
+will defer IRQs to allow for better batch processing, but will induce latency
+when the system is not fully loaded. Choosing a small value for
+``gro_flush_timeout`` can cause interference of the user application which is
+attempting to busy poll by device IRQs and softirq processing. This value
+should be chosen carefully with these tradeoffs in mind. epoll-based busy
+polling applications may be able to mitigate how much user processing happens
+by choosing an appropriate value for ``maxevents``.
+
+Users may want to consider an alternate approach, IRQ suspension, to help deal
+with these tradeoffs.
+
+IRQ suspension
+--------------
+
+IRQ suspension is a mechanism wherein device IRQs are masked while epoll
+triggers NAPI packet processing.
+
+While application calls to epoll_wait successfully retrieve events, the kernel will
+defer the IRQ suspension timer. If the kernel does not retrieve any events
+while busy polling (for example, because network traffic levels subsided), IRQ
+suspension is disabled and the IRQ mitigation strategies described above are
+engaged.
+
+This allows users to balance CPU consumption with network processing
+efficiency.
+
+To use this mechanism:
+
+  1. The per-NAPI config parameter ``irq-suspend-timeout`` should be set to the
+     maximum time (in nanoseconds) the application can have its IRQs
+     suspended. This is done using netlink, as described above. This timeout
+     serves as a safety mechanism to restart IRQ driver interrupt processing if
+     the application has stalled. This value should be chosen so that it covers
+     the amount of time the user application needs to process data from its
+     call to epoll_wait, noting that applications can control how much data
+     they retrieve by setting ``max_events`` when calling epoll_wait.
+
+  2. The sysfs parameter or per-NAPI config parameters ``gro_flush_timeout``
+     and ``napi_defer_hard_irqs`` can be set to low values. They will be used
+     to defer IRQs after busy poll has found no data.
+
+  3. The ``prefer_busy_poll`` flag must be set to true. This can be done using
+     the ``EPIOCSPARAMS`` ioctl as described above.
+
+  4. The application uses epoll as described above to trigger NAPI packet
+     processing.
+
+As mentioned above, as long as subsequent calls to epoll_wait return events to
+userland, the ``irq-suspend-timeout`` is deferred and IRQs are disabled. This
+allows the application to process data without interference.
+
+Once a call to epoll_wait results in no events being found, IRQ suspension is
+automatically disabled and the ``gro_flush_timeout`` and
+``napi_defer_hard_irqs`` mitigation mechanisms take over.
+
+It is expected that ``irq-suspend-timeout`` will be set to a value much larger
+than ``gro_flush_timeout`` as ``irq-suspend-timeout`` should suspend IRQs for
+the duration of one userland processing cycle.
+
+While it is not stricly necessary to use ``napi_defer_hard_irqs`` and
+``gro_flush_timeout`` to use IRQ suspension, their use is strongly
+recommended.
+
+IRQ suspension causes the system to alternate between polling mode and
+irq-driven packet delivery. During busy periods, ``irq-suspend-timeout``
+overrides ``gro_flush_timeout`` and keeps the system busy polling, but when
+epoll finds no events, the setting of ``gro_flush_timeout`` and
+``napi_defer_hard_irqs`` determine the next step.
+
+There are essentially three possible loops for network processing and
+packet delivery:
+
+1) hardirq -> softirq -> napi poll; basic interrupt delivery
+2) timer -> softirq -> napi poll; deferred irq processing
+3) epoll -> busy-poll -> napi poll; busy looping
+
+Loop 2 can take control from Loop 1, if ``gro_flush_timeout`` and
+``napi_defer_hard_irqs`` are set.
+
+If ``gro_flush_timeout`` and ``napi_defer_hard_irqs`` are set, Loops 2
+and 3 "wrestle" with each other for control.
+
+During busy periods, ``irq-suspend-timeout`` is used as timer in Loop 2,
+which essentially tilts network processing in favour of Loop 3.
+
+If ``gro_flush_timeout`` and ``napi_defer_hard_irqs`` are not set, Loop 3
+cannot take control from Loop 1.
+
+Therefore, setting ``gro_flush_timeout`` and ``napi_defer_hard_irqs`` is
+the recommended usage, because otherwise setting ``irq-suspend-timeout``
+might not have any discernible effect.
 
 .. _threaded:
 
diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index 1ae4542f0bd8..83bcb559b89f 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -420,7 +420,9 @@ static bool busy_loop_ep_timeout(unsigned long start_time,
 
 static bool ep_busy_loop_on(struct eventpoll *ep)
 {
-	return !!READ_ONCE(ep->busy_poll_usecs) || net_busy_loop_on();
+	return !!READ_ONCE(ep->busy_poll_usecs) ||
+	       READ_ONCE(ep->prefer_busy_poll) ||
+	       net_busy_loop_on();
 }
 
 static bool ep_busy_loop_end(void *p, unsigned long start_time)
@@ -455,6 +457,8 @@ static bool ep_busy_loop(struct eventpoll *ep, int nonblock)
 		 * it back in when we have moved a socket with a valid NAPI
 		 * ID onto the ready list.
 		 */
+		if (prefer_busy_poll)
+			napi_resume_irqs(napi_id);
 		ep->napi_id = 0;
 		return false;
 	}
@@ -538,6 +542,22 @@ static long ep_eventpoll_bp_ioctl(struct file *file, unsigned int cmd,
 	}
 }
 
+static void ep_suspend_napi_irqs(struct eventpoll *ep)
+{
+	unsigned int napi_id = READ_ONCE(ep->napi_id);
+
+	if (napi_id >= MIN_NAPI_ID && READ_ONCE(ep->prefer_busy_poll))
+		napi_suspend_irqs(napi_id);
+}
+
+static void ep_resume_napi_irqs(struct eventpoll *ep)
+{
+	unsigned int napi_id = READ_ONCE(ep->napi_id);
+
+	if (napi_id >= MIN_NAPI_ID && READ_ONCE(ep->prefer_busy_poll))
+		napi_resume_irqs(napi_id);
+}
+
 #else
 
 static inline bool ep_busy_loop(struct eventpoll *ep, int nonblock)
@@ -555,6 +575,14 @@ static long ep_eventpoll_bp_ioctl(struct file *file, unsigned int cmd,
 	return -EOPNOTSUPP;
 }
 
+static void ep_suspend_napi_irqs(struct eventpoll *ep)
+{
+}
+
+static void ep_resume_napi_irqs(struct eventpoll *ep)
+{
+}
+
 #endif /* CONFIG_NET_RX_BUSY_POLL */
 
 /*
@@ -786,6 +814,7 @@ static bool ep_refcount_dec_and_test(struct eventpoll *ep)
 
 static void ep_free(struct eventpoll *ep)
 {
+	ep_resume_napi_irqs(ep);
 	mutex_destroy(&ep->mtx);
 	free_uid(ep->user);
 	wakeup_source_unregister(ep->ws);
@@ -2003,8 +2032,11 @@ static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events,
 			 * trying again in search of more luck.
 			 */
 			res = ep_send_events(ep, events, maxevents);
-			if (res)
+			if (res) {
+				if (res > 0)
+					ep_suspend_napi_irqs(ep);
 				return res;
+			}
 		}
 
 		if (timed_out)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index df4483598628..0aae346d919e 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -348,6 +348,7 @@ struct gro_list {
  */
 struct napi_config {
 	u64 gro_flush_timeout;
+	u64 irq_suspend_timeout;
 	u32 defer_hard_irqs;
 	unsigned int napi_id;
 };
@@ -384,6 +385,7 @@ struct napi_struct {
 	struct hrtimer		timer;
 	struct task_struct	*thread;
 	unsigned long		gro_flush_timeout;
+	unsigned long		irq_suspend_timeout;
 	u32			defer_hard_irqs;
 	/* control-path-only fields follow */
 	struct list_head	dev_list;
diff --git a/include/net/busy_poll.h b/include/net/busy_poll.h
index f03040baaefd..c858270141bc 100644
--- a/include/net/busy_poll.h
+++ b/include/net/busy_poll.h
@@ -52,6 +52,9 @@ void napi_busy_loop_rcu(unsigned int napi_id,
 			bool (*loop_end)(void *, unsigned long),
 			void *loop_end_arg, bool prefer_busy_poll, u16 budget);
 
+void napi_suspend_irqs(unsigned int napi_id);
+void napi_resume_irqs(unsigned int napi_id);
+
 #else /* CONFIG_NET_RX_BUSY_POLL */
 static inline unsigned long net_busy_loop_on(void)
 {
diff --git a/include/uapi/linux/netdev.h b/include/uapi/linux/netdev.h
index e3ebb49f60d2..e4be227d3ad6 100644
--- a/include/uapi/linux/netdev.h
+++ b/include/uapi/linux/netdev.h
@@ -124,6 +124,7 @@ enum {
 	NETDEV_A_NAPI_PID,
 	NETDEV_A_NAPI_DEFER_HARD_IRQS,
 	NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT,
+	NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT,
 
 	__NETDEV_A_NAPI_MAX,
 	NETDEV_A_NAPI_MAX = (__NETDEV_A_NAPI_MAX - 1)
diff --git a/net/core/dev.c b/net/core/dev.c
index 6a31152e4606..13d00fc10f55 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6507,6 +6507,43 @@ void napi_busy_loop(unsigned int napi_id,
 }
 EXPORT_SYMBOL(napi_busy_loop);
 
+void napi_suspend_irqs(unsigned int napi_id)
+{
+	struct napi_struct *napi;
+
+	rcu_read_lock();
+	napi = napi_by_id(napi_id);
+	if (napi) {
+		unsigned long timeout = napi_get_irq_suspend_timeout(napi);
+
+		if (timeout)
+			hrtimer_start(&napi->timer, ns_to_ktime(timeout),
+				      HRTIMER_MODE_REL_PINNED);
+	}
+	rcu_read_unlock();
+}
+
+void napi_resume_irqs(unsigned int napi_id)
+{
+	struct napi_struct *napi;
+
+	rcu_read_lock();
+	napi = napi_by_id(napi_id);
+	if (napi) {
+		/* If irq_suspend_timeout is set to 0 between the call to
+		 * napi_suspend_irqs and now, the original value still
+		 * determines the safety timeout as intended and napi_watchdog
+		 * will resume irq processing.
+		 */
+		if (napi_get_irq_suspend_timeout(napi)) {
+			local_bh_disable();
+			napi_schedule(napi);
+			local_bh_enable();
+		}
+	}
+	rcu_read_unlock();
+}
+
 #endif /* CONFIG_NET_RX_BUSY_POLL */
 
 static void __napi_hash_add_with_id(struct napi_struct *napi,
@@ -6666,6 +6703,7 @@ static void napi_restore_config(struct napi_struct *n)
 {
 	n->defer_hard_irqs = n->config->defer_hard_irqs;
 	n->gro_flush_timeout = n->config->gro_flush_timeout;
+	n->irq_suspend_timeout = n->config->irq_suspend_timeout;
 	/* a NAPI ID might be stored in the config, if so use it. if not, use
 	 * napi_hash_add to generate one for us. It will be saved to the config
 	 * in napi_disable.
@@ -6680,6 +6718,7 @@ static void napi_save_config(struct napi_struct *n)
 {
 	n->config->defer_hard_irqs = n->defer_hard_irqs;
 	n->config->gro_flush_timeout = n->gro_flush_timeout;
+	n->config->irq_suspend_timeout = n->irq_suspend_timeout;
 	n->config->napi_id = n->napi_id;
 	napi_hash_del(n);
 }
diff --git a/net/core/dev.h b/net/core/dev.h
index 7881bced70a9..d043dee25a68 100644
--- a/net/core/dev.h
+++ b/net/core/dev.h
@@ -236,6 +236,31 @@ static inline void netdev_set_gro_flush_timeout(struct net_device *netdev,
 		netdev->napi_config[i].gro_flush_timeout = timeout;
 }
 
+/**
+ * napi_get_irq_suspend_timeout - get the irq_suspend_timeout
+ * @n: napi struct to get the irq_suspend_timeout from
+ *
+ * Return: the per-NAPI value of the irq_suspend_timeout field.
+ */
+static inline unsigned long
+napi_get_irq_suspend_timeout(const struct napi_struct *n)
+{
+	return READ_ONCE(n->irq_suspend_timeout);
+}
+
+/**
+ * napi_set_irq_suspend_timeout - set the irq_suspend_timeout for a napi
+ * @n: napi struct to set the irq_suspend_timeout
+ * @timeout: timeout value to set
+ *
+ * napi_set_irq_suspend_timeout sets the per-NAPI irq_suspend_timeout
+ */
+static inline void napi_set_irq_suspend_timeout(struct napi_struct *n,
+						unsigned long timeout)
+{
+	WRITE_ONCE(n->irq_suspend_timeout, timeout);
+}
+
 int rps_cpumask_housekeeping(struct cpumask *mask);
 
 #if defined(CONFIG_DEBUG_NET) && defined(CONFIG_BPF_SYSCALL)
diff --git a/net/core/netdev-genl-gen.c b/net/core/netdev-genl-gen.c
index 21de7e10be16..a89cbd8d87c3 100644
--- a/net/core/netdev-genl-gen.c
+++ b/net/core/netdev-genl-gen.c
@@ -92,10 +92,11 @@ static const struct nla_policy netdev_bind_rx_nl_policy[NETDEV_A_DMABUF_FD + 1]
 };
 
 /* NETDEV_CMD_NAPI_SET - do */
-static const struct nla_policy netdev_napi_set_nl_policy[NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT + 1] = {
+static const struct nla_policy netdev_napi_set_nl_policy[NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT + 1] = {
 	[NETDEV_A_NAPI_ID] = { .type = NLA_U32, },
 	[NETDEV_A_NAPI_DEFER_HARD_IRQS] = NLA_POLICY_FULL_RANGE(NLA_U32, &netdev_a_napi_defer_hard_irqs_range),
 	[NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT] = { .type = NLA_UINT, },
+	[NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT] = { .type = NLA_UINT, },
 };
 
 /* Ops table for netdev */
@@ -186,7 +187,7 @@ static const struct genl_split_ops netdev_nl_ops[] = {
 		.cmd		= NETDEV_CMD_NAPI_SET,
 		.doit		= netdev_nl_napi_set_doit,
 		.policy		= netdev_napi_set_nl_policy,
-		.maxattr	= NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT,
+		.maxattr	= NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT,
 		.flags		= GENL_ADMIN_PERM | GENL_CMD_CAP_DO,
 	},
 };
diff --git a/net/core/netdev-genl.c b/net/core/netdev-genl.c
index b49c3b4e5fbe..765ce7c9d73b 100644
--- a/net/core/netdev-genl.c
+++ b/net/core/netdev-genl.c
@@ -161,6 +161,7 @@ static int
 netdev_nl_napi_fill_one(struct sk_buff *rsp, struct napi_struct *napi,
 			const struct genl_info *info)
 {
+	unsigned long irq_suspend_timeout;
 	unsigned long gro_flush_timeout;
 	u32 napi_defer_hard_irqs;
 	void *hdr;
@@ -196,6 +197,11 @@ netdev_nl_napi_fill_one(struct sk_buff *rsp, struct napi_struct *napi,
 			napi_defer_hard_irqs))
 		goto nla_put_failure;
 
+	irq_suspend_timeout = napi_get_irq_suspend_timeout(napi);
+	if (nla_put_uint(rsp, NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT,
+			 irq_suspend_timeout))
+		goto nla_put_failure;
+
 	gro_flush_timeout = napi_get_gro_flush_timeout(napi);
 	if (nla_put_uint(rsp, NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT,
 			 gro_flush_timeout))
@@ -306,6 +312,7 @@ int netdev_nl_napi_get_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
 static int
 netdev_nl_napi_set_config(struct napi_struct *napi, struct genl_info *info)
 {
+	u64 irq_suspend_timeout = 0;
 	u64 gro_flush_timeout = 0;
 	u32 defer = 0;
 
@@ -314,6 +321,11 @@ netdev_nl_napi_set_config(struct napi_struct *napi, struct genl_info *info)
 		napi_set_defer_hard_irqs(napi, defer);
 	}
 
+	if (info->attrs[NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT]) {
+		irq_suspend_timeout = nla_get_uint(info->attrs[NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT]);
+		napi_set_irq_suspend_timeout(napi, irq_suspend_timeout);
+	}
+
 	if (info->attrs[NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT]) {
 		gro_flush_timeout = nla_get_uint(info->attrs[NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT]);
 		napi_set_gro_flush_timeout(napi, gro_flush_timeout);
diff --git a/tools/include/uapi/linux/netdev.h b/tools/include/uapi/linux/netdev.h
index e3ebb49f60d2..e4be227d3ad6 100644
--- a/tools/include/uapi/linux/netdev.h
+++ b/tools/include/uapi/linux/netdev.h
@@ -124,6 +124,7 @@ enum {
 	NETDEV_A_NAPI_PID,
 	NETDEV_A_NAPI_DEFER_HARD_IRQS,
 	NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT,
+	NETDEV_A_NAPI_IRQ_SUSPEND_TIMEOUT,
 
 	__NETDEV_A_NAPI_MAX,
 	NETDEV_A_NAPI_MAX = (__NETDEV_A_NAPI_MAX - 1)
diff --git a/tools/testing/selftests/net/.gitignore b/tools/testing/selftests/net/.gitignore
index a78debbd1fe7..48973e78d46b 100644
--- a/tools/testing/selftests/net/.gitignore
+++ b/tools/testing/selftests/net/.gitignore
@@ -2,6 +2,7 @@
 bind_bhash
 bind_timewait
 bind_wildcard
+busy_poller
 cmsg_sender
 diag_uid
 epoll_busy_poll
diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
index 9322b904ad00..2b2a5ec7fa6a 100644
--- a/tools/testing/selftests/net/Makefile
+++ b/tools/testing/selftests/net/Makefile
@@ -97,6 +97,11 @@ TEST_PROGS += fq_band_pktlimit.sh
 TEST_PROGS += vlan_hw_filter.sh
 TEST_PROGS += bpf_offload.py
 TEST_PROGS += ipv6_route_update_soft_lockup.sh
+TEST_PROGS += busy_poll_test.sh
+
+# YNL files, must be before "include ..lib.mk"
+YNL_GEN_FILES := busy_poller
+TEST_GEN_FILES += $(YNL_GEN_FILES)
 
 TEST_FILES := settings
 TEST_FILES += in_netns.sh lib.sh net_helper.sh setup_loopback.sh setup_veth.sh
@@ -107,6 +112,10 @@ TEST_INCLUDES := forwarding/lib.sh
 
 include ../lib.mk
 
+# YNL build
+YNL_GENS := netdev
+include ynl.mk
+
 $(OUTPUT)/epoll_busy_poll: LDLIBS += -lcap
 $(OUTPUT)/reuseport_bpf_numa: LDLIBS += -lnuma
 $(OUTPUT)/tcp_mmap: LDLIBS += -lpthread -lcrypto
diff --git a/tools/testing/selftests/net/busy_poll_test.sh b/tools/testing/selftests/net/busy_poll_test.sh
new file mode 100755
index 000000000000..7db292ec4884
--- /dev/null
+++ b/tools/testing/selftests/net/busy_poll_test.sh
@@ -0,0 +1,165 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+source net_helper.sh
+
+NSIM_SV_ID=$((256 + RANDOM % 256))
+NSIM_SV_SYS=/sys/bus/netdevsim/devices/netdevsim$NSIM_SV_ID
+NSIM_CL_ID=$((512 + RANDOM % 256))
+NSIM_CL_SYS=/sys/bus/netdevsim/devices/netdevsim$NSIM_CL_ID
+
+NSIM_DEV_SYS_NEW=/sys/bus/netdevsim/new_device
+NSIM_DEV_SYS_DEL=/sys/bus/netdevsim/del_device
+NSIM_DEV_SYS_LINK=/sys/bus/netdevsim/link_device
+NSIM_DEV_SYS_UNLINK=/sys/bus/netdevsim/unlink_device
+
+SERVER_IP=192.168.1.1
+CLIENT_IP=192.168.1.2
+SERVER_PORT=48675
+
+# busy poll config
+MAX_EVENTS=8
+BUSY_POLL_USECS=0
+BUSY_POLL_BUDGET=16
+PREFER_BUSY_POLL=1
+
+# IRQ deferral config
+NAPI_DEFER_HARD_IRQS=100
+GRO_FLUSH_TIMEOUT=50000
+SUSPEND_TIMEOUT=20000000
+
+setup_ns()
+{
+	set -e
+	ip netns add nssv
+	ip netns add nscl
+
+	NSIM_SV_NAME=$(find $NSIM_SV_SYS/net -maxdepth 1 -type d ! \
+		-path $NSIM_SV_SYS/net -exec basename {} \;)
+	NSIM_CL_NAME=$(find $NSIM_CL_SYS/net -maxdepth 1 -type d ! \
+		-path $NSIM_CL_SYS/net -exec basename {} \;)
+
+	# ensure the server has 1 queue
+	ethtool -L $NSIM_SV_NAME combined 1 2>/dev/null
+
+	ip link set $NSIM_SV_NAME netns nssv
+	ip link set $NSIM_CL_NAME netns nscl
+
+	ip netns exec nssv ip addr add "${SERVER_IP}/24" dev $NSIM_SV_NAME
+	ip netns exec nscl ip addr add "${CLIENT_IP}/24" dev $NSIM_CL_NAME
+
+	ip netns exec nssv ip link set dev $NSIM_SV_NAME up
+	ip netns exec nscl ip link set dev $NSIM_CL_NAME up
+
+	set +e
+}
+
+cleanup_ns()
+{
+	ip netns del nscl
+	ip netns del nssv
+}
+
+test_busypoll()
+{
+	suspend_value=${1:-0}
+	tmp_file=$(mktemp)
+	out_file=$(mktemp)
+
+	# fill a test file with random data
+	dd if=/dev/urandom of=${tmp_file} bs=1M count=1 2> /dev/null
+
+	timeout -k 1s 30s ip netns exec nssv ./busy_poller         \
+					     -p${SERVER_PORT}      \
+					     -b${SERVER_IP}        \
+					     -m${MAX_EVENTS}       \
+					     -u${BUSY_POLL_USECS}  \
+					     -P${PREFER_BUSY_POLL} \
+					     -g${BUSY_POLL_BUDGET} \
+					     -i${NSIM_SV_IFIDX}    \
+					     -s${suspend_value}    \
+					     -o${out_file}&
+
+	wait_local_port_listen nssv ${SERVER_PORT} tcp
+
+	ip netns exec nscl socat -u $tmp_file TCP:${SERVER_IP}:${SERVER_PORT}
+
+	wait
+
+	tmp_file_md5sum=$(md5sum $tmp_file | cut -f1 -d' ')
+	out_file_md5sum=$(md5sum $out_file | cut -f1 -d' ')
+
+	if [ "$tmp_file_md5sum" = "$out_file_md5sum" ]; then
+		res=0
+	else
+		echo "md5sum mismatch"
+		echo "input file md5sum: ${tmp_file_md5sum}";
+		echo "output file md5sum: ${out_file_md5sum}";
+		res=1
+	fi
+
+	rm $out_file $tmp_file
+
+	return $res
+}
+
+test_busypoll_with_suspend()
+{
+	test_busypoll ${SUSPEND_TIMEOUT}
+
+	return $?
+}
+
+###
+### Code start
+###
+
+modprobe netdevsim
+
+# linking
+
+echo $NSIM_SV_ID > $NSIM_DEV_SYS_NEW
+echo $NSIM_CL_ID > $NSIM_DEV_SYS_NEW
+udevadm settle
+
+setup_ns
+
+NSIM_SV_FD=$((256 + RANDOM % 256))
+exec {NSIM_SV_FD}</var/run/netns/nssv
+NSIM_SV_IFIDX=$(ip netns exec nssv cat /sys/class/net/$NSIM_SV_NAME/ifindex)
+
+NSIM_CL_FD=$((256 + RANDOM % 256))
+exec {NSIM_CL_FD}</var/run/netns/nscl
+NSIM_CL_IFIDX=$(ip netns exec nscl cat /sys/class/net/$NSIM_CL_NAME/ifindex)
+
+echo "$NSIM_SV_FD:$NSIM_SV_IFIDX $NSIM_CL_FD:$NSIM_CL_IFIDX" > \
+     $NSIM_DEV_SYS_LINK
+
+if [ $? -ne 0 ]; then
+	echo "linking netdevsim1 with netdevsim2 should succeed"
+	cleanup_ns
+	exit 1
+fi
+
+test_busypoll
+if [ $? -ne 0 ]; then
+	echo "test_busypoll failed"
+	cleanup_ns
+	exit 1
+fi
+
+test_busypoll_with_suspend
+if [ $? -ne 0 ]; then
+	echo "test_busypoll_with_suspend failed"
+	cleanup_ns
+	exit 1
+fi
+
+echo "$NSIM_SV_FD:$NSIM_SV_IFIDX" > $NSIM_DEV_SYS_UNLINK
+
+echo $NSIM_CL_ID > $NSIM_DEV_SYS_DEL
+
+cleanup_ns
+
+modprobe -r netdevsim
+
+exit 0
diff --git a/tools/testing/selftests/net/busy_poller.c b/tools/testing/selftests/net/busy_poller.c
new file mode 100644
index 000000000000..99b0e8c17fca
--- /dev/null
+++ b/tools/testing/selftests/net/busy_poller.c
@@ -0,0 +1,346 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <assert.h>
+#include <errno.h>
+#include <error.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <limits.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+#include <ynl.h>
+
+#include <arpa/inet.h>
+#include <netinet/in.h>
+
+#include <sys/epoll.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+
+#include <linux/genetlink.h>
+#include <linux/netlink.h>
+
+#include "netdev-user.h"
+
+/* The below ifdef blob is required because:
+ *
+ * - sys/epoll.h does not (yet) have the ioctl definitions included. So,
+ *   systems with older glibcs will not have them available. However,
+ *   sys/epoll.h does include the type definition for epoll_data, which is
+ *   needed by the user program (e.g. epoll_event.data.fd)
+ *
+ * - linux/eventpoll.h does not define the epoll_data type, it is simply an
+ *   opaque __u64. It does, however, include the ioctl definition.
+ *
+ * Including both headers is impossible (types would be redefined), so I've
+ * opted instead to take sys/epoll.h, and include the blob below.
+ *
+ * Someday, when glibc is globally up to date, the blob below can be removed.
+ */
+#if !defined(EPOLL_IOC_TYPE)
+struct epoll_params {
+	uint32_t busy_poll_usecs;
+	uint16_t busy_poll_budget;
+	uint8_t prefer_busy_poll;
+
+	/* pad the struct to a multiple of 64bits */
+	uint8_t __pad;
+};
+
+#define EPOLL_IOC_TYPE 0x8A
+#define EPIOCSPARAMS _IOW(EPOLL_IOC_TYPE, 0x01, struct epoll_params)
+#define EPIOCGPARAMS _IOR(EPOLL_IOC_TYPE, 0x02, struct epoll_params)
+#endif
+
+static uint32_t cfg_port = 8000;
+static struct in_addr cfg_bind_addr = { .s_addr = INADDR_ANY };
+static char *cfg_outfile;
+static int cfg_max_events = 8;
+static int cfg_ifindex;
+
+/* busy poll params */
+static uint32_t cfg_busy_poll_usecs;
+static uint32_t cfg_busy_poll_budget;
+static uint32_t cfg_prefer_busy_poll;
+
+/* IRQ params */
+static uint32_t cfg_defer_hard_irqs;
+static uint64_t cfg_gro_flush_timeout;
+static uint64_t cfg_irq_suspend_timeout;
+
+static void usage(const char *filepath)
+{
+	error(1, 0,
+	      "Usage: %s -p<port> -b<addr> -m<max_events> -u<busy_poll_usecs> -P<prefer_busy_poll> -g<busy_poll_budget> -o<outfile> -d<defer_hard_irqs> -r<gro_flush_timeout> -s<irq_suspend_timeout> -i<ifindex>",
+	      filepath);
+}
+
+static void parse_opts(int argc, char **argv)
+{
+	int ret;
+	int c;
+
+	if (argc <= 1)
+		usage(argv[0]);
+
+	while ((c = getopt(argc, argv, "p:m:b:u:P:g:o:d:r:s:i:")) != -1) {
+		switch (c) {
+		case 'u':
+			cfg_busy_poll_usecs = strtoul(optarg, NULL, 0);
+			if (cfg_busy_poll_usecs == ULONG_MAX ||
+			    cfg_busy_poll_usecs > UINT32_MAX)
+				error(1, ERANGE, "busy_poll_usecs too large");
+			break;
+		case 'P':
+			cfg_prefer_busy_poll = strtoul(optarg, NULL, 0);
+			if (cfg_prefer_busy_poll == ULONG_MAX ||
+			    cfg_prefer_busy_poll > 1)
+				error(1, ERANGE,
+				      "prefer busy poll should be 0 or 1");
+			break;
+		case 'g':
+			cfg_busy_poll_budget = strtoul(optarg, NULL, 0);
+			if (cfg_busy_poll_budget == ULONG_MAX ||
+			    cfg_busy_poll_budget > UINT16_MAX)
+				error(1, ERANGE,
+				      "busy poll budget must be [0, UINT16_MAX]");
+			break;
+		case 'p':
+			cfg_port = strtoul(optarg, NULL, 0);
+			if (cfg_port > UINT16_MAX)
+				error(1, ERANGE, "port must be <= 65535");
+			break;
+		case 'b':
+			ret = inet_aton(optarg, &cfg_bind_addr);
+			if (ret == 0)
+				error(1, errno,
+				      "bind address %s invalid", optarg);
+			break;
+		case 'o':
+			cfg_outfile = strdup(optarg);
+			if (!cfg_outfile)
+				error(1, 0, "outfile invalid");
+			break;
+		case 'm':
+			cfg_max_events = strtol(optarg, NULL, 0);
+
+			if (cfg_max_events == LONG_MIN ||
+			    cfg_max_events == LONG_MAX ||
+			    cfg_max_events <= 0)
+				error(1, ERANGE,
+				      "max events must be > 0 and < LONG_MAX");
+			break;
+		case 'd':
+			cfg_defer_hard_irqs = strtoul(optarg, NULL, 0);
+
+			if (cfg_defer_hard_irqs == ULONG_MAX ||
+			    cfg_defer_hard_irqs > INT32_MAX)
+				error(1, ERANGE,
+				      "defer_hard_irqs must be <= INT32_MAX");
+			break;
+		case 'r':
+			cfg_gro_flush_timeout = strtoull(optarg, NULL, 0);
+
+			if (cfg_gro_flush_timeout == ULLONG_MAX)
+				error(1, ERANGE,
+				      "gro_flush_timeout must be < ULLONG_MAX");
+			break;
+		case 's':
+			cfg_irq_suspend_timeout = strtoull(optarg, NULL, 0);
+
+			if (cfg_irq_suspend_timeout == ULLONG_MAX)
+				error(1, ERANGE,
+				      "irq_suspend_timeout must be < ULLONG_MAX");
+			break;
+		case 'i':
+			cfg_ifindex = strtoul(optarg, NULL, 0);
+			if (cfg_ifindex == ULONG_MAX)
+				error(1, ERANGE,
+				      "ifindex must be < ULONG_MAX");
+			break;
+		}
+	}
+
+	if (!cfg_ifindex)
+		usage(argv[0]);
+
+	if (optind != argc)
+		usage(argv[0]);
+}
+
+static void epoll_ctl_add(int epfd, int fd, uint32_t events)
+{
+	struct epoll_event ev;
+
+	ev.events = events;
+	ev.data.fd = fd;
+	if (epoll_ctl(epfd, EPOLL_CTL_ADD, fd, &ev) == -1)
+		error(1, errno, "epoll_ctl add fd: %d", fd);
+}
+
+static void setnonblock(int sockfd)
+{
+	int flags;
+
+	flags = fcntl(sockfd, F_GETFL, 0);
+
+	if (fcntl(sockfd, F_SETFL, flags | O_NONBLOCK) == -1)
+		error(1, errno, "unable to set socket to nonblocking mode");
+}
+
+static void write_chunk(int fd, char *buf, ssize_t buflen)
+{
+	ssize_t remaining = buflen;
+	char *buf_offset = buf;
+	ssize_t writelen = 0;
+	ssize_t write_result;
+
+	while (writelen < buflen) {
+		write_result = write(fd, buf_offset, remaining);
+		if (write_result == -1)
+			error(1, errno, "unable to write data to outfile");
+
+		writelen += write_result;
+		remaining -= write_result;
+		buf_offset += write_result;
+	}
+}
+
+static void setup_queue(void)
+{
+	struct netdev_napi_get_list *napi_list = NULL;
+	struct netdev_napi_get_req_dump *req = NULL;
+	struct netdev_napi_set_req *set_req = NULL;
+	struct ynl_sock *ys;
+	struct ynl_error yerr;
+	uint32_t napi_id;
+
+	ys = ynl_sock_create(&ynl_netdev_family, &yerr);
+	if (!ys)
+		error(1, 0, "YNL: %s", yerr.msg);
+
+	req = netdev_napi_get_req_dump_alloc();
+	netdev_napi_get_req_dump_set_ifindex(req, cfg_ifindex);
+	napi_list = netdev_napi_get_dump(ys, req);
+
+	/* assume there is 1 NAPI configured and take the first */
+	if (napi_list->obj._present.id)
+		napi_id = napi_list->obj.id;
+	else
+		error(1, 0, "napi ID not present?");
+
+	set_req = netdev_napi_set_req_alloc();
+	netdev_napi_set_req_set_id(set_req, napi_id);
+	netdev_napi_set_req_set_defer_hard_irqs(set_req, cfg_defer_hard_irqs);
+	netdev_napi_set_req_set_gro_flush_timeout(set_req,
+						  cfg_gro_flush_timeout);
+	netdev_napi_set_req_set_irq_suspend_timeout(set_req,
+						    cfg_irq_suspend_timeout);
+
+	if (netdev_napi_set(ys, set_req))
+		error(1, 0, "can't set NAPI params: %s\n", yerr.msg);
+
+	netdev_napi_get_list_free(napi_list);
+	netdev_napi_get_req_dump_free(req);
+	netdev_napi_set_req_free(set_req);
+	ynl_sock_destroy(ys);
+}
+
+static void run_poller(void)
+{
+	struct epoll_event events[cfg_max_events];
+	struct epoll_params epoll_params = {0};
+	struct sockaddr_in server_addr;
+	int i, epfd, nfds;
+	ssize_t readlen;
+	int outfile_fd;
+	char buf[1024];
+	int sockfd;
+	int conn;
+	int val;
+
+	outfile_fd = open(cfg_outfile, O_WRONLY | O_CREAT, 0644);
+	if (outfile_fd == -1)
+		error(1, errno, "unable to open outfile: %s", cfg_outfile);
+
+	sockfd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+	if (sockfd == -1)
+		error(1, errno, "unable to create listen socket");
+
+	server_addr.sin_family = AF_INET;
+	server_addr.sin_port = htons(cfg_port);
+	server_addr.sin_addr = cfg_bind_addr;
+
+	/* these values are range checked during parse_opts, so casting is safe
+	 * here
+	 */
+	epoll_params.busy_poll_usecs = cfg_busy_poll_usecs;
+	epoll_params.busy_poll_budget = (uint16_t)cfg_busy_poll_budget;
+	epoll_params.prefer_busy_poll = (uint8_t)cfg_prefer_busy_poll;
+	epoll_params.__pad = 0;
+
+	val = 1;
+	if (setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val)))
+		error(1, errno, "poller setsockopt reuseaddr");
+
+	setnonblock(sockfd);
+
+	if (bind(sockfd, (struct sockaddr *)&server_addr,
+		 sizeof(struct sockaddr_in)))
+		error(0, errno, "poller bind to port: %d\n", cfg_port);
+
+	if (listen(sockfd, 1))
+		error(1, errno, "poller listen");
+
+	epfd = epoll_create1(0);
+	if (ioctl(epfd, EPIOCSPARAMS, &epoll_params) == -1)
+		error(1, errno, "unable to set busy poll params");
+
+	epoll_ctl_add(epfd, sockfd, EPOLLIN | EPOLLOUT | EPOLLET);
+
+	for (;;) {
+		nfds = epoll_wait(epfd, events, cfg_max_events, -1);
+		for (i = 0; i < nfds; i++) {
+			if (events[i].data.fd == sockfd) {
+				conn = accept(sockfd, NULL, NULL);
+				if (conn == -1)
+					error(1, errno,
+					      "accepting incoming connection failed");
+
+				setnonblock(conn);
+				epoll_ctl_add(epfd, conn,
+					      EPOLLIN | EPOLLET | EPOLLRDHUP |
+					      EPOLLHUP);
+			} else if (events[i].events & EPOLLIN) {
+				for (;;) {
+					readlen = read(events[i].data.fd, buf,
+						       sizeof(buf));
+					if (readlen > 0)
+						write_chunk(outfile_fd, buf,
+							    readlen);
+					else
+						break;
+				}
+			} else {
+				/* spurious event ? */
+			}
+			if (events[i].events & (EPOLLRDHUP | EPOLLHUP)) {
+				epoll_ctl(epfd, EPOLL_CTL_DEL,
+					  events[i].data.fd, NULL);
+				close(events[i].data.fd);
+				close(outfile_fd);
+				return;
+			}
+		}
+	}
+}
+
+int main(int argc, char *argv[])
+{
+	parse_opts(argc, argv);
+	setup_queue();
+	run_poller();
+	return 0;
+}