Re: [PATCH] drm/bridge: adv7511: Reset registers on hotplug

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Jul 4, 2018 at 10:15 AM, Daniel Vetter <daniel@xxxxxxxx> wrote:
> On Wed, Jul 4, 2018 at 4:06 PM, Rob Clark <robdclark@xxxxxxxxx> wrote:
>> On Wed, Jul 4, 2018 at 9:23 AM, Archit Taneja <architt@xxxxxxxxxxxxxx> wrote:
>>>
>>>
>>> On Wednesday 04 July 2018 12:28 AM, Rob Clark wrote:
>>>>
>>>> On Tue, Jul 3, 2018 at 12:56 PM, Sean Paul <seanpaul@xxxxxxxxxxxx> wrote:
>>>>>
>>>>> The bridge loses its hw state when the cable is unplugged. If we detect
>>>>> this case in the hpd handler, reset its state.
>>>>>
>>>>> Reported-by: Rob Clark <robdclark@xxxxxxxxx>
>>>>> Signed-off-by: Sean Paul <seanpaul@xxxxxxxxxxxx>
>>>>
>>>>
>>>> Tested-by: Rob Clark <robdclark@xxxxxxxxx>
>>>>
>>>>> ---
>>>>> This is the follow-up to "drm/msm: Disable mdp5 crtc when there are no
>>>>> active planes". I incorrectly assumed the modeset was required from the
>>>>> msm side, but Rob pointed out that he thought it might be a bridge
>>>>> issue.
>>>>
>>>>
>>>> To elaborate, this is an atomic userspace (weston), when it sees the
>>>> display disconnected, it removes the planes from the CRTC but does not
>>>> disable the CRTC.  So drm/msm sets up the LM to scanout a solid color,
>>>> and leaves the encoder and dsi cranking along happily.  When
>>>> replugging the display, the atomic helpers see the CRTC is still
>>>> active and (correctly) doesn't do a full modeset.  So the bridge is
>>>> not re-configured/re-enabled.
>>>
>>>
>>> The adv7511 connector's detect() op should have re-configured the
>>> registers. I'm assuming it was never called after the cable is
>>> plugged in again in the wetson usecase?
>>>
>>> With this patch, in the case of fb emulation/X, I'm wondering if
>>> we will end up trying to re-enable ADV7511 twice. First, in the new code
>>> in adv7511_hpd_work(), and later in adv7511_detect() (i.e, the
>>> connector's detect op).
>>>
>>
>> jfwiw, fbcon and things using legacy SETCRTC end up triggering a full
>> modeset, which somehow papers over the issue (because we go thru the
>> bridge disable/enable sequence).  So now there probably is a double
>> power cycle sequence, although it doesn't seem to be causing any harm
>> (fbcon and x11 still seem happy)
>
> Hm that's strange. Legacy setcrtc on atomic drivers should do the
> exact same opportunistic modeset avoidance as an atomic flip. The only
> difference is around the automagic resetting of the link_status, but
> msm doesn't use that. I suspect the real reason here is that fbcon and
> X aren't broken, and properly shut down the disconnected output before
> trying to re-enable it. This shouldn't have anything to do with atomic
> vs. legacy ioctl paths.

yes, probably it is because they shut down the CRTC on the disconnect,
but userspace using ATOMIC ioctl has to do that explicitly (and weston
does not).  So the only connection to ATOMIC vs legacy ioctl paths is
that atomic lets userspace not shut down the CRTC.

Fwiw, I did have a patch to make adv75xx use the link_status (it is
really the bridge, not msm, which should be doing this, since it is
100% about the bridge).  But since that relies on userspace support
(which weston does not have yet), and Sean's approach actually seems
to work out ok, I kinda like his approach better.

BR,
-R

> Also, in-kernel fbcon is 100% atomic nowadays, it doesn't use any of
> the legacy entry points anymore.
> -Daniel
>
>>
>> BR,
>> -R
>>
>>> I'm guessing the 'hpd' variable in the check in adv7511_detect() below
>>> should ideally be false, and avoid us trying to configure the registers
>>> again, but I'm not entirely sure.
>>>
>>> if (status == connector_status_connected && hpd && adv7511->powered) {
>>>         regcache_mark_dirty(adv7511->regmap);
>>>         ...
>>>
>>> In any case:
>>>
>>> Reviewed-by: Archit Taneja <architt@xxxxxxxxxxxxxx>
>>>
>>>>
>>>> Arguably this perhaps isn't what weston *wanted* to do, but in the end
>>>> the bug is in the bridge.
>>>>
>>>> We could possibly set the connector link_status to BAD as an
>>>> alternative.  But at least the build of weston I'm using atm doesn't
>>>> handle the link_status propery, this approach requires each atomic
>>>> userspace to handle this case.  Compared to Sean's approach which is
>>>> transparent to userspace, which seems attractive.
>>>>
>>>> BR,
>>>> -R
>>>>
>>>>>
>>>>>   drivers/gpu/drm/bridge/adv7511/adv7511_drv.c | 12 ++++++++++++
>>>>>   1 file changed, 12 insertions(+)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>>>> b/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>>>> index 73021b388e12..dd3ff2f2cdce 100644
>>>>> --- a/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>>>> +++ b/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>>>> @@ -429,6 +429,18 @@ static void adv7511_hpd_work(struct work_struct
>>>>> *work)
>>>>>          else
>>>>>                  status = connector_status_disconnected;
>>>>>
>>>>> +       /*
>>>>> +        * The bridge resets its registers on unplug. So when we get a
>>>>> plug
>>>>> +        * event and we're already supposed to be powered, cycle the
>>>>> bridge to
>>>>> +        * restore its state.
>>>>> +        */
>>>>> +       if (status == connector_status_connected &&
>>>>> +           adv7511->connector.status == connector_status_disconnected &&
>>>>> +           adv7511->powered) {
>>>>> +               regcache_mark_dirty(adv7511->regmap);
>>>>> +               adv7511_power_on(adv7511);
>>>>> +       }
>>>>> +
>>>>>          if (adv7511->connector.status != status) {
>>>>>                  adv7511->connector.status = status;
>>>>>                  if (status == connector_status_disconnected)
>>>>> --
>>>>> Sean Paul, Software Engineer, Google / Chromium OS
>>>>>
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe linux-arm-msm"
>>>> in
>>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>>
>>>
>> _______________________________________________
>> dri-devel mailing list
>> dri-devel@xxxxxxxxxxxxxxxxxxxxx
>> https://lists.freedesktop.org/mailman/listinfo/dri-devel
>
>
>
> --
> Daniel Vetter
> Software Engineer, Intel Corporation
> +41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel




[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux