Re: [PATCH] taskstats: fix data-race

From: Balbir Singh <bsingharora@gmail.com>
To: Christian Brauner <christian.brauner@ubuntu.com>,
	syzbot+c5d03165a1bd1dead0c1@syzkaller.appspotmail.com
Cc: elver@google.com, linux-kernel@vger.kernel.org,
	syzkaller-bugs@googlegroups.com
Subject: Re: [PATCH] taskstats: fix data-race
Date: Wed, 06 Nov 2019 11:27:03 +1100	[thread overview]
Message-ID: <50abba6cb6643a8beaf29b233bea28d7b6eed5e5.camel@gmail.com> (raw)
In-Reply-To: <20191005112806.13960-1-christian.brauner@ubuntu.com>

On Sat, 2019-10-05 at 13:28 +0200, Christian Brauner wrote:
> When assiging and testing taskstats in taskstats
> taskstats_exit() there's a race around writing and reading sig->stats.
> 
> cpu0:
> task calls exit()
> do_exit()
> 	-> taskstats_exit()
> 		-> taskstats_tgid_alloc()
> The task takes sighand lock and assigns new stats to sig->stats.
> 
> cpu1:
> task catches signal
> do_exit()
> 	-> taskstats_tgid_alloc()
> 		-> taskstats_exit()
> The tasks reads sig->stats __without__ holding sighand lock seeing
> garbage.
> 
> Fix this by taking sighand lock when reading sig->stats.
> 

Hi

Sorry I am just catching up with my inbox, why is sig->stats garbage? I'll
need to go back and look, unsigned long (pointer) reads are atomic and the
expectation is that they would be NULL initially and then non NULL for thread
groups when first allocated. May be I am just being dense in the morning -
what am I missing?

Balbir Singh.