Re: [PATCH] block: fix request.queuelist usage in flush

[Friedrich Weber <f.weber@proxmox.com>]

On 05/06/2024 12:30, Chengming Zhou wrote:
> On 2024/6/5 16:45, Friedrich Weber wrote:
>> Hi,
>>
>> On 04/06/2024 08:47, Chengming Zhou wrote:
>>> Friedrich Weber reported a kernel crash problem and bisected to commit
>>> 81ada09cc25e ("blk-flush: reuse rq queuelist in flush state machine").
>>>
>>> The root cause is that we use "list_move_tail(&rq->queuelist, pending)"
>>> in the PREFLUSH/POSTFLUSH sequences. But rq->queuelist.next == xxx since
>>> it's popped out from plug->cached_rq in __blk_mq_alloc_requests_batch().
>>> We don't initialize its queuelist just for this first request, although
>>> the queuelist of all later popped requests will be initialized.
>>>
>>> Fix it by changing to use "list_add_tail(&rq->queuelist, pending)" so
>>> rq->queuelist doesn't need to be initialized. It should be ok since rq
>>> can't be on any list when PREFLUSH or POSTFLUSH, has no move actually.
>>>
>>> Please note the commit 81ada09cc25e ("blk-flush: reuse rq queuelist in
>>> flush state machine") also has another requirement that no drivers would
>>> touch rq->queuelist after blk_mq_end_request() since we will reuse it to
>>> add rq to the post-flush pending list in POSTFLUSH. If this is not true,
>>> we will have to revert that commit IMHO.
>>
>> Unfortunately, with this patch applied to kernel 6.9 I get a different
>> crash [2] on a Debian 12 (virtual) machine with root on LVM on boot (no
>> software RAID involved). See [1] for lsblk and findmnt output. addr2line
>> says:
> 
> Sorry, which commit is your kernel? Is mainline tag v6.9 or at some commit?

Yes, by "kernel 6.9" I meant mainline tag v6.9, so commit a38297e3fb01.

If I boot this mainline kernel v6.9 in a Debian (virtual) machine with
root on LVM, I do not get a crash. If I apply the patch "block: fix
request.queuelist usage in flush" on top of this mainline kernel v6.9,
and boot the Debian machine into that patched kernel, I get a crash on boot.

> And is it reproducible using the mainline kernel v6.10-rc2?

I'll test mainline kernel v6.10-rc2, and "block: fix request.queuelist
usage in flush" applied on top of v6.10-rc2, and get back to you.

>> # addr2line -f -e /usr/lib/debug/vmlinux-6.9.0-patch0604-nodebuglist+
>> blk_mq_request_bypass_insert+0x20
> 
> I think here should use blk_mq_insert_request+0x120, instead of the
> blk_mq_request_bypass_insert+0x20, which has "?" at the beginning.
> 

Right, sorry:

# addr2line -f -e /usr/lib/debug/vmlinux-6.9.0-patch0604-nodebuglist+
blk_mq_insert_request+0x120
blk_mq_insert_request
[...]/linux/block/blk-mq.c:2539

which refers to this line [1]:

		blk_mq_request_bypass_insert(rq, BLK_MQ_INSERT_AT_HEAD);

Thanks!

Friedrich

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/block/blk-mq.c?h=v6.9#n2539