[PATCH v7 09/15] drm/bridge: analogix_dp: Add support to get panel from the DP AUX bus

[Damon Ding]

Hi Doug,

On 2025/2/25 9:41, Doug Anderson wrote:
> Hi,
> 
> On Mon, Feb 24, 2025 at 12:14 AM Damon Ding <damon.ding at rock-chips.com> wrote:
>>
>> The main modification is moving the DP AUX initialization from function
>> analogix_dp_bind() to analogix_dp_probe(). In order to get the EDID of
>> eDP panel during probing, it is also needed to advance PM operations to
>> ensure that eDP controller and phy are prepared for AUX transmission.
>>
>> Additionally, add support for &drm_dp_aux.wait_hpd_asserted() to help
>> confirm the HPD state before doing AUX transfers.
> 
> Maybe move the addition of the analogix_dpaux_wait_hpd_asserted() to a
> separate patch?
> 

It would be a good idea.

> 
>> Signed-off-by: Damon Ding <damon.ding at rock-chips.com>
>>
>> ---
>>
>> Changes in v4:
>> - Use done_probing() to call drm_of_find_panel_or_bridge() and
>>    component_add() when getting panel from the DP AUX bus
>>
>> Changes in v5:
>> - Advance PM operations to make eDP AUX work well
>>
>> Changes in v6:
>> - Use devm_pm_runtime_enable() instead of devm_add_action_or_reset()
>> - Add a new function analogix_dp_remove() to ensure symmetry for PM
>>    operations
>>
>> Changes in v7:
>> - Fix the misspelling of word 'operations' in commit message
>> - Remove the check related to CONFIG_PM
>> - Remove the unnecessary call to pm_runtime_dont_use_autosuspend() if
>>    devm_pm_runtime_enable() fails
>> - Remove unnecessary function analogix_dp_remove()
>> - Add new function analogix_dpaux_wait_hpd_asserted()
>> ---
>>   .../drm/bridge/analogix/analogix_dp_core.c    | 37 ++++++++++++++-----
>>   1 file changed, 28 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c b/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c
>> index a57e06d303a1..ff81c37dbe1d 100644
>> --- a/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c
>> +++ b/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c
>> @@ -1548,6 +1548,22 @@ static ssize_t analogix_dpaux_transfer(struct drm_dp_aux *aux,
>>          return ret;
>>   }
>>
>> +static int analogix_dpaux_wait_hpd_asserted(struct drm_dp_aux *aux, unsigned long wait_us)
>> +{
>> +       struct analogix_dp_device *dp = to_dp(aux);
>> +       int val;
>> +       int ret;
>> +
>> +       pm_runtime_get_sync(dp->dev);
>> +
>> +       ret = readx_poll_timeout(analogix_dp_detect_hpd, dp, val, !val, wait_us / 100, wait_us);
> 
> More than happy if someone else wants to overrule me, but it seems
> weird that you're looping over a function that already has a loop.
> Shouldn't you be calling analogix_dp_get_plug_in_status() directly?
> ...and if "dp->force_hpd" you probably shouldn't be polling at all. If
> HPD is not hooked up I think we decided in sn65dsi86 that we should
> just sleep for the maximum time (choosing a conservative value if told
> to wait forever).
> 

Yes, I think there is no need to use analogix_dp_detect_hpd(), and the 
code as you recommended is better:

static int analogix_dpaux_wait_hpd_asserted(struct drm_dp_aux *aux, 
unsigned long wait_us)
{
	struct analogix_dp_device *dp = to_dp(aux);
	int val;
	int ret;

	if (dp->force_hpd)
		return 0;

	pm_runtime_get_sync(dp->dev);

	ret = readx_poll_timeout(analogix_dp_get_plug_in_status, dp, val, !val,
				 wait_us / 100, wait_us);

	pm_runtime_mark_last_busy(dp->dev);
	pm_runtime_put_autosuspend(dp->dev);

	return ret;
}

> 
> Aside from that and the idea of moving
> analogix_dpaux_wait_hpd_asserted() to a separate patch this looks good
> to me.
> 

Best regards
Damon