[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Fri, 3 Mar 2023 10:21:49 +0800
From: Yu Kuai <yukuai1@...weicloud.com>
To: Keith Busch <kbusch@...nel.org>, Yu Kuai <yukuai1@...weicloud.com>
Cc: hch@....de, ming.lei@...hat.com, axboe@...nel.dk,
linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
yi.zhang@...wei.com, yangerkun@...wei.com,
"yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [PATCH] blk-mq: quiesce queue while reallocating hctxs
Hi,
在 2023/03/03 9:24, Keith Busch 写道:
> On Tue, Feb 21, 2023 at 05:24:36PM +0800, Yu Kuai wrote:
>> From: Yu Kuai <yukuai3@...wei.com>
>>
>> commit 8237c01f1696 ("blk-mq: use quiesced elevator switch when
>> reinitializing queues") add quiesce queue while switching elevator,
>> however, if old elevator is none, queue is still not quiesced. Hence
>> reallocating hctxs can concurrent with run queue. Fix it by also
>> quiesce queue in the beginning of __blk_mq_update_nr_hw_queues().
>
> Is this actually fixing anything? The quiesced elevator switch was to prevent
> use-after-free from an elevator being torn down, but if you are not switching
> elevators, then what resource does quiescing protect?
> .
>
What I can think of is hctx itself. run queue can fetch a hctx that will
be exited, I did't found any real problems yet, but I don't think it
is not good to run work for a exited or reused hctx.
Thanks,
Kuai
Powered by blists - more mailing lists