[SRU][O][PATCH 0/3] Fix BQL crashes, most frequently observed on GCP instances

[Koichiro Den]

BugLink: https://bugs.launchpad.net/bugs/2089684

SRU Justification

[Impact]

Since upstream commit c8bd1f7f3e61 ("virtio_net: add support for Byte Queue
Limits"), BQL crashes have been observed. One crash pattern was addressed
by upstream commit b96ed2c97c79 ("virtio_net: move netdev_tx_reset_queue()
call before RX napi enable"), but other patterns remain unresolved. Most
notably, on GCP instances, frequent boot test failures with BQL crashes
have been found. To end users, this issue typically appears as extremely
slow instance boot time. Even when booting is successful, the instance
remains susceptible to kernel panics under certain conditions. So this
issue needs to be resolved. Only Oracular is affected.

[Fix]

The issue is resolved by the following patch series:
https://lore.kernel.org/all/20241206011047.923923-1-koichiro.den@canonical.com/
6.11.y branch is now EOL, so the patch series will not land in upstream
stable tree.

[Test Plan]

Reboot some GCP instances with the fix applied multiple times to verify
that the BQL crash no longer occurs.

[Where problems could occur]

The fix impacts only virtio-net, so any regression would appear as
unexpected behavior in virtio-net or potentially cause a kernel crash.

[Other Info]

[PATCH 1/3] resolves the issue observed on GCP.
[PATCH 2/3] + [PATCH 3/3] resolve similar issues which have not been
observed on our testing infrastructure but still worth applying to
prevent potential kernel panic due to BQL crash.


Koichiro Den (3):
  virtio_net: correct netdev_tx_reset_queue() invocation point
  virtio_ring: add a func argument 'recycle_done' to virtqueue_resize()
  virtio_net: ensure netdev_tx_reset_queue is called on tx ring resize

 drivers/net/virtio_net.c     | 24 +++++++++++++++++++++---
 drivers/virtio/virtio_ring.c |  6 +++++-
 include/linux/virtio.h       |  3 ++-
 3 files changed, 28 insertions(+), 5 deletions(-)

-- 
2.43.0