Thread (37 messages) 37 messages, 4 authors, 2021-10-19

Re: [PATCH 06/14] block: store elevator state in request

From: Jens Axboe <axboe@kernel.dk>
Date: 2021-10-19 22:26:11
Also in: kernelci, lkml

On 10/19/21 4:21 PM, Guillaume Tucker wrote:
Hi Jens,

On 17/10/2021 02:37, Jens Axboe wrote:
quoted
Add an rq private RQF_ELV flag, which tells the block layer that this
request was initialized on a queue that has an IO scheduler attached.
This allows for faster checking in the fast path, rather than having to
deference rq->q later on.

Elevator switching does full quiesce of the queue before detaching an
IO scheduler, so it's safe to cache this in the request itself.
A kernelci.org automated bisection found that this patch
introduced a regression in next-20211019 with a NULL pointer
dereference, which only seems to be affecting QEMU but across all
architectures.

More details about the regression can be found here:

  https://linux.kernelci.org/test/job/next/branch/master/kernel/next-20211019/plan/baseline/
  https://linux.kernelci.org/test/case/id/616ea20eb7104071c43358ea/

See also all the test jobs involved in the automated bisection:

  https://lava.collabora.co.uk/scheduler/device_type/qemu?dt_search=bisection-287

If you do send a fix, please include this trailer:

  Reported-by: "kernelci.org bot" [off-list ref]


Please let us know if this seems like a valid bisection result
and if you need any help to debug the issue or try a fix.
This got fixed yesterday, current tree is fine.

-- 
Jens Axboe
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help