Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Dmitry Monakhov <dmtrmonakhov@yandex-team.ru>
To:     Tejun Heo <tj@kernel.org>
Cc:     linux-kernel@vger.kernel.org, axboe@kernel.dk
In-Reply-To: <YKZ5i5P8fL8bgq7G@slm.duckdns.org>
References: <20210513082827.1818-1-dmtrmonakhov@yandex-team.ru> <YKZ5i5P8fL8bgq7G@slm.duckdns.org>
Subject: Re: [PATCH] blk-throttle: fix race between submitter and throttler thread
MIME-Version: 1.0
Date:   Thu, 20 May 2021 20:00:51 +0300
Message-Id: <1394051621530051@iva4-92c901fae84c.qloud-c.yandex.net>
Content-Transfer-Encoding: 7bit
Content-Type: text/plain
Precedence: bulk


> Hello, Dmitry.
> 
> This generally looks good to me. A couple nits below.
> 
>> @@ -277,6 +277,8 @@ static struct bio *__bio_chain_endio(struct bio *bio)
>> {
>> struct bio *parent = bio->bi_private;
>>
>> + BUG_ON(!bio_flagged(parent, BIO_CHAIN));
> 
> Let's do WARN_ON_ONCE().
If we hit this point when this mean that  ->bio_end_io  will be called for parent bio.
Which likely result in use-after-free for that bio and silent data corruption for bio's pages
So IMHO BUG_ON is more appropriate here. What do you think?
> 
>> @@ -2270,6 +2285,8 @@ bool blk_throtl_bio(struct bio *bio)
>>
>> td->nr_queued[rw]++;
>> throtl_add_bio_tg(bio, qn, tg);
> 
> Can you add some comment here explaining how now that the bio is added for
> throttling, there are two accessors of it and the bio must not be modified
> without holding the lock?
Sound reasonable, will be back with updated comments.
> 
> Thank you.
> 
> --
> tejun