[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250512132641.GC31781@lst.de>
Date: Mon, 12 May 2025 15:26:41 +0200
From: Christoph Hellwig <hch@....de>
To: Yu Kuai <yukuai1@...weicloud.com>
Cc: Christoph Hellwig <hch@....de>, xni@...hat.com, colyli@...nel.org,
agk@...hat.com, snitzer@...nel.org, mpatocka@...hat.com,
song@...nel.org, linux-kernel@...r.kernel.org,
dm-devel@...ts.linux.dev, linux-raid@...r.kernel.org,
yi.zhang@...wei.com, yangerkun@...wei.com, johnny.chenyi@...wei.com,
"yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [PATCH RFC md-6.16 v3 15/19] md/md-llbitmap: implement APIs to
dirty bits and clear bits
On Mon, May 12, 2025 at 04:23:55PM +0800, Yu Kuai wrote:
>>> + INIT_WORK_ONSTACK(&unplug_work.work, llbitmap_unplug_fn);
>>> + queue_work(md_llbitmap_unplug_wq, &unplug_work.work);
>>> + wait_for_completion(&done);
>>> + destroy_work_on_stack(&unplug_work.work);
>>
>> Why is this deferring the work to a workqueue, but then synchronously
>> waits on it?
>
> This is the same as old bitmap, by the fact that issue new IO and wait
> for such IO to be done from submit_bio() context will deadlock.
>
> 1) bitmap bio must be done before this bio can be issued;
> 2) bitmap bio will be added to current->bio_list, and wait for this bio
> to be issued;
>
> Do you have a better sulution to this problem?
A bew block layer API that bypasses bio_list maybe? I.e. export
__submit_bio with a better name and a kerneldoc detailing the narrow
use case.
Powered by blists - more mailing lists