Low RX performance for 40G Solarflare NICs

[Heitor Alves de Siqueira]

[Impact]
* Some 40G Solarflare NICs have low RX performance in some cases, due
  low RX recycle ring size
* RX recycle ring size is either 4096 for IOMMU, 16 for NOIOMMU
* The low fixed sizes can cause a high number of calls to alloc_pages,
  tanking performance for higher link speeds

[Test Plan]
* Users report that iperf3 is sufficient to showcase the bad RX performance
* For some setups, RX performance was around 15Gbps while TX stayed
  consistently above 30Gbps

[Fix]
* This patch sets the RX recycle ring size according to an adapter's
  maximum link speed
* Fix was introduced by commit:
  000fe940e51f "sfc: The size of the RX recycle ring should be more flexible"
* --!-- Commit is from net-next --!--

[Regression Potential]
* Regressions would show primarily as performance issues, as we're
  effectively changing ring sizes for all RX traffic
* It's possible to see increased calls to alloc_pages if ring sizes
  aren't being set correctly
* We should look out for excessive memory usage in the sfc driver due to
  the increased ring sizes

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This bug is awaiting verification that the linux/5.13.0-38.43 kernel in -proposed solves the problem. Please test the kernel and update this bug with the results. If the problem is solved, change the tag 'verification-needed-impish' to 'verification-done-impish'. If the problem still exists, change the tag 'verification-needed-impish' to 'verification-failed-impish'.

If verification is not done by 5 working days from today, this fix will be dropped from the source code, and this bug will be closed.

See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you!

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This bug is awaiting verification that the linux/5.4.0-106.120 kernel in -proposed solves the problem. Please test the kernel and update this bug with the results. If the problem is solved, change the tag 'verification-needed-focal' to 'verification-done-focal'. If the problem still exists, change the tag 'verification-needed-focal' to 'verification-failed-focal'.

If verification is not done by 5 working days from today, this fix will be dropped from the source code, and this bug will be closed.

See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you!

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This bug is awaiting verification that the linux-hwe-5.4/5.4.0-107.121~18.04.1 kernel in -proposed solves the problem. Please test the kernel and update this bug with the results. If the problem is solved, change the tag 'verification-needed-bionic' to 'verification-done-bionic'. If the problem still exists, change the tag 'verification-needed-bionic' to 'verification-failed-bionic'.

If verification is not done by 5 working days from today, this fix will be dropped from the source code, and this bug will be closed.

See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you!

[Heitor Alves de Siqueira (halves)]

Validated for Ubuntu Focal with the current kernel from focal-proposed.
Tested on a 10G NIC that's affected by this patch (device ID 0x0903). iperf3 shows that we're able to almost saturate the NIC successfully in both standard and reverse sender mode:

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.3 GBytes 8.83 Gbits/sec 266 sender
[ 5] 0.00-10.00 sec 10.3 GBytes 8.82 Gbits/sec receiver

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.8 GBytes 9.30 Gbits/sec 0 sender
[ 5] 0.00-10.00 sec 10.8 GBytes 9.30 Gbits/sec receiver

ubuntu@duduo:~$ uname -rv
5.4.0-109-generic #123-Ubuntu SMP Fri Apr 8 09:10:54 UTC 2022

[Heitor Alves de Siqueira (halves)]

Validated for Ubuntu Bionic with the current HWE kernel from bionic-proposed.
Tested on a 10G NIC that's affected by this patch (device ID 0x0903). iperf3 shows that we're able to almost saturate the NIC successfully in both standard and reverse sender mode:

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.2 GBytes 8.80 Gbits/sec 376 sender
[ 5] 0.00-10.00 sec 10.2 GBytes 8.80 Gbits/sec receiver

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.9 GBytes 9.39 Gbits/sec 0 sender
[ 5] 0.00-10.00 sec 10.9 GBytes 9.38 Gbits/sec receiver

ubuntu@duduo:~$ uname -rv
5.4.0-108-generic #122~18.04.1-Ubuntu SMP Wed Apr 6 16:57:12 UTC 2022

Other basic smoke testing confirms no major regressions.

[Heitor Alves de Siqueira (halves)]

Validated for Ubuntu Impish with the current kernel from impish-proposed.
Tested on a 10G NIC that's affected by this patch (device ID 0x0903). iperf3 shows that we're able to almost saturate the NIC successfully in both standard and reverse sender mode:

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.1 GBytes 8.69 Gbits/sec 990 sender
[ 5] 0.00-9.99 sec 10.1 GBytes 8.69 Gbits/sec receiver

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 10.8 GBytes 9.30 Gbits/sec 155 sender
[ 5] 0.00-10.00 sec 10.8 GBytes 9.30 Gbits/sec receiver

ubuntu@duduo:~$ uname -rv
5.13.0-40-generic #45-Ubuntu SMP Tue Mar 29 14:48:14 UTC 2022

Other basic smoke testing confirms no major regressions.

[Heitor Alves de Siqueira (halves)]

In addition to my own validations above, impacted users that have 40G NICs have also confirmed the patch behaves as expected without major regressions.

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 5.13.0-40.45

---------------
linux (5.13.0-40.45) impish; urgency=medium

  * impish/linux: 5.13.0-40.45 -proposed tracker (LP: #1966701)

  * CVE-2022-1016
    - netfilter: nf_tables: initialize registers in nft_do_chain()

  * CVE-2022-1015
    - netfilter: nf_tables: validate registers coming from userspace.

  * audit: improve audit queue handling when "audit=1" on cmdline
    (LP: #1965723) // Impish update: upstream stable patchset 2022-03-22
    (LP: #1966021)
    - audit: improve audit queue handling when "audit=1" on cmdline

  * PS/2 Keyboard wakeup from s2idle not functioning on AMD Yellow Carp platform
    (LP: #1961739)
    - PM: s2idle: ACPI: Fix wakeup interrupts handling

  * Low RX performance for 40G Solarflare NICs (LP: #1964512)
    - SAUCE: sfc: The size of the RX recycle ring should be more flexible

  * [UBUNTU 20.04] Fix SIGP processing on KVM/s390 (LP: #1962578)
    - KVM: s390: Simplify SIGP Set Arch handling
    - KVM: s390: Add a routine for setting userspace CPU state

  * Move virtual graphics drivers from linux-modules-extra to linux-modules
    (LP: #1960633)
    - [Packaging] Move VM DRM drivers into modules

  * Impish update: upstream stable patchset 2022-03-09 (LP: #1964422)
    - bnx2x: Utilize firmware 7.13.21.0
    - bnx2x: Invalidate fastpath HSI version for VFs
    - rcu: Tighten rcu_advance_cbs_nowake() checks
    - select: Fix indefinitely sleeping task in poll_schedule_timeout()
    - drm/amdgpu: Use correct VIEWPORT_DIMENSION for DCN2
    - arm64/bpf: Remove 128MB limit for BPF JIT programs
    - Bluetooth: refactor malicious adv data check
    - net: sfp: ignore disabled SFP node
    - net: stmmac: skip only stmmac_ptp_register when resume from suspend
    - s390/hypfs: include z/VM guests with access control group set
    - bpf: Guard against accessing NULL pt_regs in bpf_get_task_stack()
    - scsi: zfcp: Fix failed recovery on gone remote port with non-NPIV FCP
      devices
    - udf: Restore i_lenAlloc when inode expansion fails
    - udf: Fix NULL ptr deref when converting from inline format
    - efi: runtime: avoid EFIv2 runtime services on Apple x86 machines
    - PM: wakeup: simplify the output logic of pm_show_wakelocks()
    - tracing/histogram: Fix a potential memory leak for kstrdup()
    - tracing: Don't inc err_log entry count if entry allocation fails
    - ceph: properly put ceph_string reference after async create attempt
    - ceph: set pool_ns in new inode layout for async creates
    - fsnotify: fix fsnotify hooks in pseudo filesystems
    - Revert "KVM: SVM: avoid infinite loop on NPF from bad address"
    - perf/x86/intel/uncore: Fix CAS_COUNT_WRITE issue for ICX
    - drm/etnaviv: relax submit size limits
    - KVM: x86: Update vCPU's runtime CPUID on write to MSR_IA32_XSS
    - netfilter: nft_payload: do not update layer 4 checksum when mangling
      fragments
    - serial: 8250: of: Fix mapped region size when using reg-offset property
    - serial: stm32: fix software flow control transfer
    - tty: n_gsm: fix SW flow control encoding/handling
    - tty: Add support for Brainboxes UC cards.
    - usb-storage: Add unusual-devs...

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 5.4.0-109.123

---------------
linux (5.4.0-109.123) focal; urgency=medium

  * focal/linux: 5.4.0-109.123 -proposed tracker (LP: #1968290)

  * USB devices not detected during boot on USB 3.0 hubs (LP: #1968210)
    - SAUCE: Revert "Revert "xhci: Set HCD flag to defer primary roothub
      registration""
    - SAUCE: Revert "Revert "usb: core: hcd: Add support for deferring roothub
      registration""

linux (5.4.0-108.122) focal; urgency=medium

  * focal/linux: 5.4.0-108.122 -proposed tracker (LP: #1966740)

  * Packaging resync (LP: #1786013)
    - [Packaging] resync dkms-build{,--nvidia-N} from LRMv5
    - debian/dkms-versions -- update from kernel-versions (main/2022.03.21)

  * Low RX performance for 40G Solarflare NICs (LP: #1964512)
    - SAUCE: sfc: The size of the RX recycle ring should be more flexible

  * [UBUNTU 20.04] KVM: Enable storage key checking for intercepted instruction
    (LP: #1962831)
    - selftests: kvm: add _vm_ioctl
    - selftests: kvm: Introduce the TEST_FAIL macro
    - KVM: selftests: Add GUEST_ASSERT variants to pass values to host
    - KVM: s390: gaccess: Refactor gpa and length calculation
    - KVM: s390: gaccess: Refactor access address range check
    - KVM: s390: gaccess: Cleanup access to guest pages
    - s390/uaccess: introduce bit field for OAC specifier
    - s390/uaccess: fix compile error
    - s390/uaccess: Add copy_from/to_user_key functions
    - KVM: s390: Honor storage keys when accessing guest memory
    - KVM: s390: handle_tprot: Honor storage keys
    - KVM: s390: selftests: Test TEST PROTECTION emulation
    - KVM: s390: Add optional storage key checking to MEMOP IOCTL
    - KVM: s390: Add vm IOCTL for key checked guest absolute memory access
    - KVM: s390: Rename existing vcpu memop functions
    - KVM: s390: Add capability for storage key extension of MEM_OP IOCTL
    - KVM: s390: Update api documentation for memop ioctl
    - KVM: s390: Clarify key argument for MEM_OP in api docs
    - KVM: s390: Add missing vm MEM_OP size check

  * 【sec-0911】 fail to reset sec module (LP: #1943301)
    - crypto: hisilicon/sec2 - Add workqueue for SEC driver.
    - crypto: hisilicon/sec2 - update SEC initialization and reset

  * Lots of hisi_qm zombie task slow down system after stress test
    (LP: #1932117)
    - crypto: hisilicon - Use one workqueue per qm instead of per qp

  * Lots of hisi_qm zombie task slow down system after stress test
    (LP: #1932117) // 【sec-0911】 fail to reset sec module (LP: #1943301)
    - crypto: hisilicon - Unify hardware error init/uninit into QM

  * [UBUNTU 20.04] Fix SIGP processing on KVM/s390 (LP: #1962578)
    - KVM: s390: Simplify SIGP Set Arch handling
    - KVM: s390: Add a routine for setting userspace CPU state

  * Move virtual graphics drivers from linux-modules-extra to linux-modules
    (LP: #1960633)
    - [Packaging] Move VM DRM drivers into modules

  * Focal update: v5.4.178 upstream stable release (LP: #1964634)
    - audit: improve audit queue handling when "audit=1" on cmdline
    - ASoC: ops: Reject out of bounds values in snd_soc_put_volsw()
    - ASoC: ops: Reject out of bounds values in snd_...