ACK: [SRU P][PATCH 0/1] Fix netdevsim crashing in nl_netdev test
Mehmet Basaran
mehmet.basaran at canonical.com
Fri Oct 10 10:31:49 UTC 2025
Acked-by: Mehmet Basaran <mehmet.basaran at canonical.com>
On 10/10/25 12:43, Stefan Bader wrote:
> BugLink: https://bugs.launchpad.net/bugs/2121866
>
> [SRU Justification]
>
> == Impact ==
> Since SRU cycle 2025.08.11 we saw the ubuntu_kselftest_net suite failing
> more or less harshly. The reason was that running nl_netdev.py was
> causing a crash:
>
> [ 5769.558910] RIP: 0010:mutex_lock+0x1c/0x50
> [ 5769.568304] napi_disable+0x1d/0x50
> [ 5769.568661] nsim_queue_stop+0x27/0x80 [netdevsim]
> [ 5769.569140] netdev_rx_queue_restart+0x14a/0x260
> [ 5769.569603] nsim_qreset_write+0x10d/0x150 [netdevsim]
> [ 5769.570113] full_proxy_write+0x5e/0xa0
> [ 5769.570499] vfs_write+0x10a/0x460
>
> == Fix ==
> It was found that this got introduced by
> "netdevsim: allow normal queue reset while down"
> Apparently it should be allowed but stopping and starting the queue(s)
> is not allowed while a device is not running.
> This was handled with the following upstream change:
> net: devmem: don't call queue stop / start when the interface is down
> However the submission was explicitly stating that is was not inteneded
> as a bug fix but a clarification. Therefor it never was picked into
> upstream stable.
>
> == Testcase ==
> Running ubuntu_kselftest_net (specifically nl_netdev.py) should no
> longer fail and produce an oops.
>
> == Regression Potential ==
> Potentially the behaviour of netdev_queue_restart() (resetting network
> devices) is impacted.
>
> -Stefan
>
>
> Jakub Kicinski (1):
> net: devmem: don't call queue stop / start when the interface is down
>
> include/net/netdev_queues.h | 4 ++++
> net/core/netdev_rx_queue.c | 18 +++++++++++-------
> 2 files changed, 15 insertions(+), 7 deletions(-)
>
More information about the kernel-team
mailing list