blktests failures with v6.12 kernel

[Nilay Shroff]

Hi Shinichiro,

On 11/26/24 07:49, Shinichiro Kawasaki wrote:
> Hi all,
> 
> I ran the latest blktests (git hash: 83781f257857) with the v6.12 kernel. Also,
> I checked CKI project runs for the kernel. I observed five failures below.
> 
> Comparing with the previous report using the v6.12-rc1 kernel [1], two failures
> were resolved: nvme/014 and srp group. On the other hand, four new failures were
> observed.
> 
> [1] https://lore.kernel.org/linux-block/xpe6bea7rakpyoyfvspvin2dsozjmjtjktpph7rep3h25tv7fb@ooz4cu5z6bq6/
> 
> 
> List of failures
> ================
> #1: nvme/031 (fc transport)
> #2: nvme/037 (fc transport)
> #3: nvme/041 (fc transport)
> #4: nvme/052 (loop transport)
> #5: throtl/001 (CKI project, s390 arch)
> 
> Failure description
> ===================
> 
<snip>
> #4: nvme/052 (loop transport)
> 
>   The test case fails due to the "BUG: sleeping function called from invalid
>   context" [5]. A fix candidate was posted which sets NVME_F_BLOCKING to loop
>   transport, but it is not the best solution [6]. It is desired to have a better
>   fix and the test case to confirm it.
> 
>   [5] https://lore.kernel.org/linux-nvme/tqcy3sveity7p56v7ywp7ssyviwcb3w4623cnxj3knoobfcanq@yxgt2mjkbkam/
>   [6] https://lore.kernel.org/linux-nvme/20241017172052.2603389-1-kbusch@meta.com/
> 
I have developed a blktest to recreate this bug and also probable fix as suggested by Christoph[1].
I will send the blktest and patch later today.

[1] https://lore.kernel.org/linux-nvme/20241022070252.GA11389@lst.de/

Thanks,
--Nilay