[bug report][bisected] blktests nvme/tcp failed on the latest linux-block/for-next

[Maurizio Lombardi]

čt 19. 10. 2023 v 7:29 odesílatel Yi Zhang <yi.zhang at redhat.com> napsal:
>
> Hi Hannes
> Bisect shows it was introduced from this commit:
>
> commit 675b453e024154dd547921c6e6d5b58747ba7e0e
> Author: Hannes Reinecke <hare at suse.de>
> Date:   Thu Aug 24 16:39:23 2023 +0200
>
>     nvmet-tcp: enable TLS handshake upcall

st 18. 10. 2023 v 16:08 odesílatel Yi Zhang <yi.zhang at redhat.com> napsal:
>
> [ 5161.405594] ODEBUG: assert_init not available (active state 0)
> [ 5161.521808] Workqueue: nvmet-wq nvmet_tcp_release_queue_work [nvmet_tcp]
> [ 5161.697797]  __timer_delete+0x70/0x170
> [ 5161.701550]  ? __pfx___timer_delete+0x10/0x10
> [ 5161.705907]  ? __mutex_lock+0x741/0x14b0
> [ 5161.709833]  ? try_to_grab_pending+0x70/0x80
> [ 5161.714105]  ? rcu_is_watching+0x11/0xb0
> [ 5161.718032]  try_to_grab_pending+0x46/0x80
> [ 5161.722132]  __cancel_work_timer+0xa0/0x460
> [ 5161.726318]  ? mutex_lock_io_nested+0x1273/0x1310
> [ 5161.731024]  ? __pfx___cancel_work_timer+0x10/0x10
> [ 5161.735816]  ? do_raw_write_trylock+0xb5/0x190
> [ 5161.740260]  ? __pfx_do_raw_write_trylock+0x10/0x10
> [ 5161.745141]  ? rcu_is_watching+0x11/0xb0
> [ 5161.749066]  ? trace_irq_enable.constprop.0+0x13d/0x180
> [ 5161.754292]  ? __pfx_sk_stream_write_space+0x10/0x10
> [ 5161.759258]  nvmet_tcp_release_queue_work+0x2db/0x1350 [nvmet_tcp]

Maybe the problem is that nvmet_tcp_release_queue_work() always calls
cancel_delayed_work_sync() against the tls_handshake_tmo_work timer,
but the latter is initialized only if
CONFIG_NVME_TARGET_TCP_TLS is defined.

Maurizio