[Intel-gfx] [PATCH] drm/dp/mst: fix kernel oops when turning off secondary monitor

[chris at chris-wilson.co.uk]

On Mon, Dec 05, 2016 at 09:39:49PM +0000, Pandiyan, Dhinakaran wrote:
> On Mon, 2016-12-05 at 08:02 +0000, Chris Wilson wrote:
> > On Sun, Dec 04, 2016 at 07:31:18PM -0600, Pierre-Louis Bossart wrote:
> > > 100% reproducible issue found on SKL SkullCanyon NUC with two external
> > > DP daisy-chained monitors in DP/MST mode. When turning off or changing
> > > the input of the second monitor the machine stops with a kernel
> > > oops. This issue happened with 4.8.8 as well as drm/drm-intel-nightly.
> > > 
> > > This issue is traced to an inconsistent control flow in
> > > drm_dp_update_payload_part1(): the 'port' pointer is set to NULL at
> > > the same time as'req_payload.num_slots' is set to zero, but the pointer
> > > is dereferenced even when req_payload.num_slot is zero.
> > > 
> > > Fix by adding test condition to make sure both variables
> > > are used consistently. This removes the kernel oops.
> > > 
> > > There are still annoying cases where the primary display goes black
> > > when the secondary display is turned off but it can be recovered from
> > > by playing with the monitor inputs and power buttons.
> > > 
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98990
> > > Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart at linux.intel.com>
> > > ---
> > >  drivers/gpu/drm/drm_dp_mst_topology.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > index aa64448..5481fde 100644
> > > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > > @@ -1815,7 +1815,7 @@ int drm_dp_update_payload_part1(struct drm_dp_mst_topology_mgr *mgr)
> > >  				drm_dp_create_payload_step1(mgr, mgr->proposed_vcpis[i]->vcpi, &req_payload);
> > >  				mgr->payloads[i].num_slots = req_payload.num_slots;
> > >  				mgr->payloads[i].vcpi = req_payload.vcpi;
> > > -			} else if (mgr->payloads[i].num_slots) {
> > > +			} else if (mgr->payloads[i].num_slots && port != NULL) {
> > >  				mgr->payloads[i].num_slots = 0;
> > >  				drm_dp_destroy_payload_step1(mgr, port, port->vcpi.vcpi, &mgr->payloads[i]);
> > 
> > s/port->vcpi.vcpi/mgr->payloads[i].vcpi/ here looks to be the correct
> > fix.
> > -Chris
> > 
> 
> Hmm, not sure if that is the correct fix either. With port = NULL,
> doesn't look like we send drm_dp_payload_send_msg(..., pbn = 0).
> Although, we do update the payload table via DPCD 

I'll leave that for you to work out if something needs to be sent after
the downstream disappeared.
 
> Also, if port is set to  NULL, I wonder if we are messing up the
> reference counting. Because, this is done below.
> ...
> 	if (port)
> 		drm_dp_put_port(port);

This is ok, as this is clearing the local ref which is only obtained if
port != NULL. (If port == NULL, we have no local ref which we need to
put.)
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre