linux-kernel - Re: [PATCH v6 1/2] blktrace: Fix potentail deadlock between delete & sysfs ops

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: <20170919204135.GA7411@infradead.org>
Date:   Tue, 19 Sep 2017 13:41:35 -0700
From:   Christoph Hellwig <hch@...radead.org>
To:     Waiman Long <longman@...hat.com>
Cc:     Christoph Hellwig <hch@...radead.org>,
        Jens Axboe <axboe@...nel.dk>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ingo Molnar <mingo@...nel.org>, linux-kernel@...r.kernel.org,
        linux-fsdevel@...r.kernel.org, linux-block@...r.kernel.org,
        linux-nilfs@...r.kernel.org, cluster-devel@...hat.com,
        Bart Van Assche <Bart.VanAssche@....com>
Subject: Re: [PATCH v6 1/2] blktrace: Fix potentail deadlock between delete &
 sysfs ops

On Tue, Sep 19, 2017 at 11:58:34AM -0400, Waiman Long wrote:
> I was trying not to add a new mutex to a structure just for blktrace as
> it is an optional feature that is enabled only if the
> CONFIG_BLK_DEV_IO_TRACE config option is defined and it will only need
> to be taken occasionally.

So?  Make the lock optional, too.

> As filesystem freeze looks orthogonal to blktrace and the mutex also
> looks to be used sparingly, I think it is a good match to overload it to
> control blktrace as well.

If the functionally is orthogonal that is a reason not to share a lock,
not to the contrary.

> I could modify the patch to use a mutex in the request_queue structure.
> The current sysfs_lock mutex has about 74 references. So I am not
> totally sure if it is safe to reuse. So the only option is to add
> something like

> 
> #ifdef CONFIG_BLK_DEV_IO_TRACE
> struct mutex blktrace_mutex;
> #endif
> 
> to the request_queue structure. That structure is large enough that
> adding a mutex won't increase the size much percentage-wise.

Call it blk_trace mutex and move it right next to the blk_trace
structure:

ifdef CONFIG_BLK_DEV_IO_TRACE
        struct blk_trace        *blk_trace;
	struct mutex		blk_trace_mutex;
#endif

which makes it completely obvious to any read what you are protecting
with it.