Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 3 Mar 2020 08:35:23 -0500
From:   Steven Rostedt <rostedt@goodmis.org>
To:     Bob Liu <bob.liu@oracle.com>
Cc:     Cengiz Can <cengiz@kernel.wtf>, Jens Axboe <axboe@kernel.dk>,
        Ingo Molnar <mingo@redhat.com>, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] blktrace: fix dereference after null check
Message-ID: <20200303083523.78233c24@gandalf.local.home>
In-Reply-To: <ec24c6d8-617f-1460-0420-bc2ac3f346c6@oracle.com>
References: <20200303073358.57799-1-cengiz@kernel.wtf>
        <ec24c6d8-617f-1460-0420-bc2ac3f346c6@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 3 Mar 2020 21:29:08 +0800
Bob Liu <bob.liu@oracle.com> wrote:

> On 3/3/20 3:33 PM, Cengiz Can wrote:
> > There was a recent change in blktrace.c that added a RCU protection to
> > `q->blk_trace` in order to fix a use-after-free issue during access.
> > 
> > However the change missed an edge case that can lead to dereferencing of
> > `bt` pointer even when it's NULL:
> > 
> > ```
> >         bt->act_mask = value; // bt can still be NULL here
> > ```
> > 
> > Added a reassignment into the NULL check block to fix the issue.
> > 
> > Fixes: c780e86dd48 ("blktrace: Protect q->blk_trace with RCU")
> > 
> > Signed-off-by: Cengiz Can <cengiz@kernel.wtf>
> > ---
> >  Huge thanks goes to Steven Rostedt for his assistance.
> > 
> >  kernel/trace/blktrace.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
> > index 4560878f0bac..29ea88f10b87 100644
> > --- a/kernel/trace/blktrace.c
> > +++ b/kernel/trace/blktrace.c
> > @@ -1896,8 +1896,10 @@ static ssize_t sysfs_blk_trace_attr_store(struct device *dev,
> >  	}
> > 
> >  	ret = 0;
> > -	if (bt == NULL)
> > +	if (bt == NULL) {
> >  		ret = blk_trace_setup_queue(q, bdev);
> > +		bt = q->blk_trace;  
> 
> The return value 'ret' should be judged, it's wrong to set 'bt' if blk_trace_setup_queue()
> return failure.

Why? If ret is an error, q is still valid, and bt would just be garbage. bt
is ignored below if ret is anything but zero. Why add an unnecessary if
condition here?

That said, the bt assignment still needs rcu annotation:

		bt = rcu_dereference_protected(q->blk_trace,
				lockdep_is_held(&q->blk_trace_mutex));

-- Steve


> 
> > +	}
> > 
> >  	if (ret == 0) {
> >  		if (attr == &dev_attr_act_mask)
> > --
> > 2.25.1
> >