linux-kernel - Re: [PATCH] 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: <4841e079-c518-cdff-2fbd-661a9ef36722@redhat.com>
Date:   Wed, 16 Aug 2017 14:46:42 -0400
From:   Waiman Long <longman@...hat.com>
To:     Steven Rostedt <rostedt@...dmis.org>
Cc:     Jens Axboe <axboe@...nel.dk>,
        Jeff Layton <jlayton@...chiereds.net>,
        "J. Bruce Fields" <bfields@...ldses.org>,
        Ingo Molnar <mingo@...nel.org>, linux-kernel@...r.kernel.org,
        linux-block@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH] blktrace: Fix potentail deadlock between delete & sysfs
 ops

On 08/16/2017 02:17 PM, Steven Rostedt wrote:
> On Wed, 16 Aug 2017 14:14:36 -0400
> Waiman Long <longman@...hat.com> wrote:
>
>>> You also just turned the mutex into a spinlock. What happens if we just
>>> preempted the owner of bdev->bd_mutex and are an RT task with higher
>>> priority? This will turn into a live lock.
>>>  
>>>> +		schedule();
>>>> +	}
>>>>    
>> That is OK because I used schedule() instead of cpu_relax() for
>> inserting delay.
> Please explain to me how that is OK? schedule is a nop if the current
> task is the highest priority task running, and it preempted the owner
> of the lock. Nothing will actually schedule.
>
> -- Steve

I haven't been thinking about RT tasks. You are right that it can be a
problem in this case. I think I will have to revert back to use
mutex_lock() if a RT task is running. Though in this case, the lock
inversion problem will still be there. However, it is highly unlikely
that a RT task will need to read write the block trace sysfs files.

Thanks for the input.

Cheers,
Longman