diff mbox

drm: Check for connector->state NULL in drm_atomic_add_affected_connectors

Message ID 1457135033-11791-1-git-send-email-dianders@chromium.org (mailing list archive)
State New, archived
Headers show

Commit Message

Doug Anderson March 4, 2016, 11:43 p.m. UTC
On a system I'm doing development on I found a crash.  The crawl looked
like:

  PC is at drm_atomic_add_affected_connectors+0x98/0xe8
  ...
  drm_atomic_add_affected_connectors from __drm_atomic_helper_set_config+0x218/0x344
  __drm_atomic_helper_set_config from restore_fbdev_mode+0x108/0x250
  restore_fbdev_mode from drm_fb_helper_restore_fbdev_mode_unlocked+0x3c/0x80
  drm_fb_helper_restore_fbdev_mode_unlocked from rockchip_drm_lastclose+0x1c/0x20
  rockchip_drm_lastclose from drm_lastclose+0x4c/0x104
  drm_lastclose from drm_release+0x424/0x47c
  drm_release from __fput+0xf8/0x1d4
  __fput from ____fput+0x18/0x1c
  ____fput from task_work_run+0xa8/0xbc
  task_work_run from do_exit+0x448/0x91c
  do_exit from do_group_exit+0x5c/0xcc
  do_group_exit from get_signal+0x4dc/0x57c
  get_signal from do_signal+0x9c/0x3b4
  do_signal from do_work_pending+0x60/0xb8
  do_work_pending from slow_work_pending+0xc/0x20

I found that I could fix the crash by checking connector->state against
NULL.  This isn't code I'm familiar with and I didn't dig too deep, so
I'd appreciate any opinions about whether this is a sane thing to do.
Note that I don't actually have graphics up on my system at the moment,
so perhaps this is all just a symptom of the strange state I'm in.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
---
 drivers/gpu/drm/drm_atomic.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Daniel Vetter March 6, 2016, 4:29 p.m. UTC | #1
On Fri, Mar 04, 2016 at 03:43:53PM -0800, Douglas Anderson wrote:
> On a system I'm doing development on I found a crash.  The crawl looked
> like:
> 
>   PC is at drm_atomic_add_affected_connectors+0x98/0xe8
>   ...
>   drm_atomic_add_affected_connectors from __drm_atomic_helper_set_config+0x218/0x344
>   __drm_atomic_helper_set_config from restore_fbdev_mode+0x108/0x250
>   restore_fbdev_mode from drm_fb_helper_restore_fbdev_mode_unlocked+0x3c/0x80
>   drm_fb_helper_restore_fbdev_mode_unlocked from rockchip_drm_lastclose+0x1c/0x20
>   rockchip_drm_lastclose from drm_lastclose+0x4c/0x104
>   drm_lastclose from drm_release+0x424/0x47c
>   drm_release from __fput+0xf8/0x1d4
>   __fput from ____fput+0x18/0x1c
>   ____fput from task_work_run+0xa8/0xbc
>   task_work_run from do_exit+0x448/0x91c
>   do_exit from do_group_exit+0x5c/0xcc
>   do_group_exit from get_signal+0x4dc/0x57c
>   get_signal from do_signal+0x9c/0x3b4
>   do_signal from do_work_pending+0x60/0xb8
>   do_work_pending from slow_work_pending+0xc/0x20
> 
> I found that I could fix the crash by checking connector->state against
> NULL.  This isn't code I'm familiar with and I didn't dig too deep, so
> I'd appreciate any opinions about whether this is a sane thing to do.
> Note that I don't actually have graphics up on my system at the moment,
> so perhaps this is all just a symptom of the strange state I'm in.
> 
> Signed-off-by: Douglas Anderson <dianders@chromium.org>

This is a driver bug - under atomic the assumption is that there is
_always_ a current software state. Most driver set up the initial
"everything off" state in the ->reset functions.
-Daniel

> ---
>  drivers/gpu/drm/drm_atomic.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> index 8fb469c4e4b8..3377d7ddc6d1 100644
> --- a/drivers/gpu/drm/drm_atomic.c
> +++ b/drivers/gpu/drm/drm_atomic.c
> @@ -1147,7 +1147,7 @@ drm_atomic_add_affected_connectors(struct drm_atomic_state *state,
>  	 * current configuration.
>  	 */
>  	drm_for_each_connector(connector, state->dev) {
> -		if (connector->state->crtc != crtc)
> +		if (!connector->state || connector->state->crtc != crtc)
>  			continue;
>  
>  		conn_state = drm_atomic_get_connector_state(state, connector);
> -- 
> 2.7.0.rc3.207.g0ac5344
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
Ville Syrjala March 6, 2016, 5:30 p.m. UTC | #2
On Sun, Mar 06, 2016 at 05:29:51PM +0100, Daniel Vetter wrote:
> On Fri, Mar 04, 2016 at 03:43:53PM -0800, Douglas Anderson wrote:
> > On a system I'm doing development on I found a crash.  The crawl looked
> > like:
> > 
> >   PC is at drm_atomic_add_affected_connectors+0x98/0xe8
> >   ...
> >   drm_atomic_add_affected_connectors from __drm_atomic_helper_set_config+0x218/0x344
> >   __drm_atomic_helper_set_config from restore_fbdev_mode+0x108/0x250
> >   restore_fbdev_mode from drm_fb_helper_restore_fbdev_mode_unlocked+0x3c/0x80
> >   drm_fb_helper_restore_fbdev_mode_unlocked from rockchip_drm_lastclose+0x1c/0x20
> >   rockchip_drm_lastclose from drm_lastclose+0x4c/0x104
> >   drm_lastclose from drm_release+0x424/0x47c
> >   drm_release from __fput+0xf8/0x1d4
> >   __fput from ____fput+0x18/0x1c
> >   ____fput from task_work_run+0xa8/0xbc
> >   task_work_run from do_exit+0x448/0x91c
> >   do_exit from do_group_exit+0x5c/0xcc
> >   do_group_exit from get_signal+0x4dc/0x57c
> >   get_signal from do_signal+0x9c/0x3b4
> >   do_signal from do_work_pending+0x60/0xb8
> >   do_work_pending from slow_work_pending+0xc/0x20
> > 
> > I found that I could fix the crash by checking connector->state against
> > NULL.  This isn't code I'm familiar with and I didn't dig too deep, so
> > I'd appreciate any opinions about whether this is a sane thing to do.
> > Note that I don't actually have graphics up on my system at the moment,
> > so perhaps this is all just a symptom of the strange state I'm in.
> > 
> > Signed-off-by: Douglas Anderson <dianders@chromium.org>
> 
> This is a driver bug - under atomic the assumption is that there is
> _always_ a current software state. Most driver set up the initial
> "everything off" state in the ->reset functions.

Except if it's due to kzalloc() failing in the atomic helper reset
implementation. There's no way to propagate the error upwards, so
the current implementation just silently sets state=NULL and goes on
like nothing happened.
Doug Anderson March 8, 2016, 12:05 a.m. UTC | #3
Daniel,

On Sun, Mar 6, 2016 at 8:29 AM, Daniel Vetter <daniel@ffwll.ch> wrote:
> On Fri, Mar 04, 2016 at 03:43:53PM -0800, Douglas Anderson wrote:
>> On a system I'm doing development on I found a crash.  The crawl looked
>> like:
>>
>>   PC is at drm_atomic_add_affected_connectors+0x98/0xe8
>>   ...
>>   drm_atomic_add_affected_connectors from __drm_atomic_helper_set_config+0x218/0x344
>>   __drm_atomic_helper_set_config from restore_fbdev_mode+0x108/0x250
>>   restore_fbdev_mode from drm_fb_helper_restore_fbdev_mode_unlocked+0x3c/0x80
>>   drm_fb_helper_restore_fbdev_mode_unlocked from rockchip_drm_lastclose+0x1c/0x20
>>   rockchip_drm_lastclose from drm_lastclose+0x4c/0x104
>>   drm_lastclose from drm_release+0x424/0x47c
>>   drm_release from __fput+0xf8/0x1d4
>>   __fput from ____fput+0x18/0x1c
>>   ____fput from task_work_run+0xa8/0xbc
>>   task_work_run from do_exit+0x448/0x91c
>>   do_exit from do_group_exit+0x5c/0xcc
>>   do_group_exit from get_signal+0x4dc/0x57c
>>   get_signal from do_signal+0x9c/0x3b4
>>   do_signal from do_work_pending+0x60/0xb8
>>   do_work_pending from slow_work_pending+0xc/0x20
>>
>> I found that I could fix the crash by checking connector->state against
>> NULL.  This isn't code I'm familiar with and I didn't dig too deep, so
>> I'd appreciate any opinions about whether this is a sane thing to do.
>> Note that I don't actually have graphics up on my system at the moment,
>> so perhaps this is all just a symptom of the strange state I'm in.
>>
>> Signed-off-by: Douglas Anderson <dianders@chromium.org>
>
> This is a driver bug - under atomic the assumption is that there is
> _always_ a current software state. Most driver set up the initial
> "everything off" state in the ->reset functions.

Ah, I see what the problem is.  It looks like the main Rockchip
subsystem moved to atomic but the patch to support that in dw_hdmi
never landed.  If I pick <https://patchwork.kernel.org/patch/7868951/>
then my crash goes away.  :)

In case it's not obvious, please consider $SUBJECT patch abandoned.  Thanks!

-Doug
Doug Anderson March 8, 2016, 12:09 a.m. UTC | #4
Hi,

On Mon, Mar 7, 2016 at 4:05 PM, Doug Anderson <dianders@chromium.org> wrote:
> Daniel,
>
> On Sun, Mar 6, 2016 at 8:29 AM, Daniel Vetter <daniel@ffwll.ch> wrote:
>> On Fri, Mar 04, 2016 at 03:43:53PM -0800, Douglas Anderson wrote:
>>> On a system I'm doing development on I found a crash.  The crawl looked
>>> like:
>>>
>>>   PC is at drm_atomic_add_affected_connectors+0x98/0xe8
>>>   ...
>>>   drm_atomic_add_affected_connectors from __drm_atomic_helper_set_config+0x218/0x344
>>>   __drm_atomic_helper_set_config from restore_fbdev_mode+0x108/0x250
>>>   restore_fbdev_mode from drm_fb_helper_restore_fbdev_mode_unlocked+0x3c/0x80
>>>   drm_fb_helper_restore_fbdev_mode_unlocked from rockchip_drm_lastclose+0x1c/0x20
>>>   rockchip_drm_lastclose from drm_lastclose+0x4c/0x104
>>>   drm_lastclose from drm_release+0x424/0x47c
>>>   drm_release from __fput+0xf8/0x1d4
>>>   __fput from ____fput+0x18/0x1c
>>>   ____fput from task_work_run+0xa8/0xbc
>>>   task_work_run from do_exit+0x448/0x91c
>>>   do_exit from do_group_exit+0x5c/0xcc
>>>   do_group_exit from get_signal+0x4dc/0x57c
>>>   get_signal from do_signal+0x9c/0x3b4
>>>   do_signal from do_work_pending+0x60/0xb8
>>>   do_work_pending from slow_work_pending+0xc/0x20
>>>
>>> I found that I could fix the crash by checking connector->state against
>>> NULL.  This isn't code I'm familiar with and I didn't dig too deep, so
>>> I'd appreciate any opinions about whether this is a sane thing to do.
>>> Note that I don't actually have graphics up on my system at the moment,
>>> so perhaps this is all just a symptom of the strange state I'm in.
>>>
>>> Signed-off-by: Douglas Anderson <dianders@chromium.org>
>>
>> This is a driver bug - under atomic the assumption is that there is
>> _always_ a current software state. Most driver set up the initial
>> "everything off" state in the ->reset functions.
>
> Ah, I see what the problem is.  It looks like the main Rockchip
> subsystem moved to atomic but the patch to support that in dw_hdmi
> never landed.  If I pick <https://patchwork.kernel.org/patch/7868951/>
> then my crash goes away.  :)
>
> In case it's not obvious, please consider $SUBJECT patch abandoned.  Thanks!

Argh.  ...or the needed patch landed but I didn't pick it back in my
backport.  Even dumber.  :(

-Doug
diff mbox

Patch

diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
index 8fb469c4e4b8..3377d7ddc6d1 100644
--- a/drivers/gpu/drm/drm_atomic.c
+++ b/drivers/gpu/drm/drm_atomic.c
@@ -1147,7 +1147,7 @@  drm_atomic_add_affected_connectors(struct drm_atomic_state *state,
 	 * current configuration.
 	 */
 	drm_for_each_connector(connector, state->dev) {
-		if (connector->state->crtc != crtc)
+		if (!connector->state || connector->state->crtc != crtc)
 			continue;
 
 		conn_state = drm_atomic_get_connector_state(state, connector);