Return-Path: <linux-bluetooth-owner@vger.kernel.org>
Date: Thu, 22 Jun 2017 07:32:43 +0300
From: Matias Karhumaa <matias.karhumaa@gmail.com>
To: Luiz Augusto von Dentz <luiz.dentz@gmail.com>,
        linux-bluetooth@vger.kernel.org
Subject: Re: [PATCH] obexd: Fix null pointer dereference.
Message-ID: <20170622043241.GA31744@makarhum-e6330>
References: <20170621061358.GA17661@makarhum-e6330>
 <20170621094148.GA27878@x1c>
 <CABBYNZLfN-u0xOqut3f35jaX5xRwusY5SW0oL1M_XzOF283oLQ@mail.gmail.com>
 <CABBYNZJOO+v0FHh2F65iMk6kDZedwG2bOXRn5Ycr4uu=qx9NHw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <CABBYNZJOO+v0FHh2F65iMk6kDZedwG2bOXRn5Ycr4uu=qx9NHw@mail.gmail.com>
Sender: linux-bluetooth-owner@vger.kernel.org
List-ID: <linux-bluetooth.vger.kernel.org>

On Wed, Jun 21, 2017 at 10:31:39PM +0300, Luiz Augusto von Dentz wrote:
> Hi Matias,
> 
> On Wed, Jun 21, 2017 at 8:36 PM, Luiz Augusto von Dentz
> <luiz.dentz@gmail.com> wrote:
> > Hi Johan,
> >
> > On Wed, Jun 21, 2017 at 12:41 PM, Johan Hedberg <johan.hedberg@gmail.com> wrote:
> >> Hi Matias,
> >>
> >> On Wed, Jun 21, 2017, Matias Karhumaa wrote:
> >>> By sending OPP Put request before CONNECT we were able to cause
> >>> SIGSEGV in obexd. Crash was caused by null pointer dereference.
> >>>
> >>> gdb output:
> >>>
> >>> Program received signal SIGSEGV, Segmentation fault.
> >>> manager_request_authorization (transfer=transfer@entry=0x0, new_folder=new_folder@entry=0x7fffffffda18, new_name=new_name@entry=0x7fffffffda20) at obexd/src/manager.c:677
> >>> 677           struct obex_session *os = transfer->session;
> >>> (gdb) bt
> >>> *#0  manager_request_authorization (transfer=transfer@entry=0x0, new_folder=new_folder@entry=0x7fffffffda18, new_name=new_name@entry=0x7fffffffda20) at obexd/src/manager.c:677
> >>> *#1  0x000000000041b7a5 in opp_chkput (os=0x67de60, user_data=0x0) at obexd/plugins/opp.c:80
> >>> *#2  0x0000000000426cc5 in check_put (obex=0x678a50, req=0x679250, user_data=0x67de60) at obexd/src/obex.c:831
> >>> *#3  cmd_put (obex=0x678a50, req=0x679250, user_data=0x67de60) at obexd/src/obex.c:887
> >>> *#4  0x00000000004145e7 in handle_request (req=0x679250, obex=0x678a50) at gobex/gobex.c:1199
> >>> *#5  incoming_data (io=<optimized out>, cond=<optimized out>, user_data=0x678a50) at gobex/gobex.c:1375
> >>> *#6  0x00007ffff749204a in g_main_dispatch (context=0x674810) at /build/glib2.0-prJhLS/glib2.0-2.48.2/./glib/gmain.c:3154
> >>> *#7  g_main_context_dispatch (context=context@entry=0x674810) at /build/glib2.0-prJhLS/glib2.0-2.48.2/./glib/gmain.c:3769
> >>> *#8  0x00007ffff74923f0 in g_main_context_iterate (context=0x674810, block=block@entry=1, dispatch=dispatch@entry=1, self=<optimized out>)
> >>>     at /build/glib2.0-prJhLS/glib2.0-2.48.2/./glib/gmain.c:3840
> >>> *#9  0x00007ffff7492712 in g_main_loop_run (loop=0x66fdf0) at /build/glib2.0-prJhLS/glib2.0-2.48.2/./glib/gmain.c:4034
> >>> *#10 0x000000000040dd0f in main (argc=1, argv=0x7fffffffde08) at obexd/src/main.c:322
> >>>
> >>> Crash was found using Synopsys Defensics Obex Server test suite.
> >>> ---
> >>>  obexd/src/obex.c | 2 +-
> >>>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>>
> >>> diff --git a/obexd/src/obex.c b/obexd/src/obex.c
> >>> index 788bffc..91fa838 100644
> >>> --- a/obexd/src/obex.c
> >>> +++ b/obexd/src/obex.c
> >>> @@ -825,7 +825,7 @@ static gboolean check_put(GObex *obex, GObexPacket *req, void *user_data)
> >>>       struct obex_session *os = user_data;
> >>>       int ret;
> >>>
> >>> -     if (os->service->chkput == NULL)
> >>> +     if (os->service->chkput == NULL || os->service_data == NULL)
> >>>               goto done;
> >>
> >> As far as I understand, os->service_data is the OBEX service-specific
> >> context, which I think can in principle validly be NULL. Also, isn't it
> >> so that OBEX Object Push permits PUT without a preceding CONNECT (at
> >> least that's what I remember from the times I was working with OBEX)?
> >
> > Yep, OPP can start a transfer without a CONNECT so the service_data
> > will need to be instantiated directly on PUT if that wasn't created
> > yet.
> >
> >> It seems to me that the bug is with opp.c passing the service_data to
> >> manager_request_authorization(), and service_data is expected to be the
> >> obex_transfer object. However, currently the code creates this object
> >> only upon CONNECT (in opp_connect).
> >>
> >> I think one possible way to solve this would be to trigger a call to
> >> os->service->connect if CONNECT hasn't been explicitly issued, however
> >> then the code needs to track this in some other way since service_data
> >> seems too unreliable.
> 
> How about the following:
> 
> https://gist.github.com/Vudentz/1736d6af9608b9332b93858d92a3feff

Your fix seems solid to me and fixes the crash with OPP. I checked ftp.c
also and saw potentially similar problem in ftp_chkput. Should we call
os->service->connect also in case of OBEX_FTP?

Best Regards,
Matias Karhumaa