linux-kernel - Re: [PATCH 3/3] blk-mq: fix wait condition for tagset wait completed check

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <04ca03dd-d240-458a-a049-8cf0ea7f9dcc@flourine.local>
Date: Fri, 31 Jan 2025 09:46:46 +0100
From: Daniel Wagner <dwagner@...e.de>
To: Sagi Grimberg <sagi@...mberg.me>
Cc: Daniel Wagner <wagi@...nel.org>, Keith Busch <kbusch@...nel.org>, 
	Jens Axboe <axboe@...nel.dk>, Christoph Hellwig <hch@....de>, Ming Lei <ming.lei@...hat.com>, 
	James Smart <james.smart@...adcom.com>, Hannes Reinecke <hare@...e.de>, linux-nvme@...ts.infradead.org, 
	linux-kernel@...r.kernel.org, linux-block@...r.kernel.org
Subject: Re: [PATCH 3/3] blk-mq: fix wait condition for tagset wait completed
 check

On Fri, Jan 31, 2025 at 10:13:47AM +0200, Sagi Grimberg wrote:
> On 28/01/2025 18:34, Daniel Wagner wrote:
> > blk_mq_tagset_count_completed_reqs returns the number of completed
> > requests. The only user of this function is
> > blk_mq_tagset_wait_completed_request which wants to know how many
> > request are not yet completed. Thus return the number of in flight
> > requests and terminate the wait loop when there is no inflight request.
> > 
> > Fixes: f9934a80f91d ("blk-mq: introduce blk_mq_tagset_wait_completed_request()")
> 
> Can you please describe what this patch is fixing? i.e. what is the observed
> bug?
> It is not clear (to me) from the patch.

I have to double check again my reasoning after reading Nilay's reply.

The problem I am running into with my wip tp4129 patchset is that
requests are pending an newly introduce requeue_list queue and never get
canceled because these requests stay in the COMPLETE state at the
moment. This blocked the shutdown path.

I don't think this problem exists right now in upstream but I was under
the impression that the check is incorrect here. I mean the
only request we need to cancel are the ones which are not idle, no?

Anyway, I'll have to go back and do some more homework.