linux-kernel - Re: [PATCH] modules: fix livelock in add_unformed

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: <ba46f7c1-caee-4237-b6c5-7edec0eaaac3@redhat.com>
Date:   Wed, 22 May 2019 13:08:56 -0400
From:   Prarit Bhargava <prarit@...hat.com>
To:     Barret Rhoden <brho@...gle.com>, Jessica Yu <jeyu@...nel.org>
Cc:     linux-kernel@...r.kernel.org,
        Heiko Carstens <heiko.carstens@...ibm.com>,
        David Arcari <darcari@...hat.com>
Subject: Re: [PATCH] modules: fix livelock in add_unformed_module()



On 5/13/19 10:37 AM, Barret Rhoden wrote:
> Hi -
> 

Hey Barret, my apologies for not getting back to you earlier.  I got caught up
in something that took me away from this issue.

> On 5/13/19 7:23 AM, Prarit Bhargava wrote:
> [snip]
>> A module is loaded once for each cpu.
> 
> Does one CPU succeed in loading the module, and the others fail with EEXIST?
> 
>> My follow-up patch changes from wait_event_interruptible() to
>> wait_event_interruptible_timeout() so the CPUs are no longer sleeping and can
>> make progress on other tasks, which changes the return values from
>> wait_event_interruptible().
>>
>> https://marc.info/?l=linux-kernel&m=155724085927589&w=2
>>
>> I believe this also takes your concern into account?
> 
> That patch might work for me, but I think it papers over the bug where the check
> on old->state that you make before sleeping (was COMING || UNFORMED, now !LIVE)
> doesn't match the check to wake up in finished_loading().
> 
> The reason the issue might not show up in practice is that your patch basically
> polls, so the condition checks in finished_loading() are only a quicker exit.
> 
> If you squash my patch into yours, I think it will cover that case. Though if
> polling is the right answer here, it also raises the question of whether or not
> we even need finished_loading().
> 

The more I look at this I think you're right.  Let me do some additional testing
with your patch + my original patch.

P.


> Barret