[PATCH v4 12/12] drm/ast: Implement polling for VGA and SIL164 connectors

[Jani Nikula]

On Tue, 19 Mar 2024, Thomas Zimmermann <tzimmermann at suse.de> wrote:
> Hi
>
> Am 19.03.24 um 10:37 schrieb Maxime Ripard:
>> Hi,
>>
>> On Tue, Mar 19, 2024 at 09:00:32AM +0100, Thomas Zimmermann wrote:
>>> Implement polling for VGA and SIL164 connectors. Set the flag
>>> DRM_CONNECTOR_POLL_DISCONNECT for each to detect the removal of the
>>> monitor cable. Implement struct drm_connector_helper_funcs.detect_ctx
>>> for each type of connector by testing for EDID data. The code for
>>> both types of connectors is identical for now. Maybe this can later
>>> become a common helper function for various drivers.
>>>
>>> Signed-off-by: Thomas Zimmermann <tzimmermann at suse.de>
>>> ---
>>>   drivers/gpu/drm/ast/ast_mode.c | 36 ++++++++++++++++++++++++++++++++--
>>>   1 file changed, 34 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/ast/ast_mode.c b/drivers/gpu/drm/ast/ast_mode.c
>>> index 71cc681d6188f..f740b8706a38b 100644
>>> --- a/drivers/gpu/drm/ast/ast_mode.c
>>> +++ b/drivers/gpu/drm/ast/ast_mode.c
>>> @@ -1344,8 +1344,24 @@ static int ast_crtc_init(struct drm_device *dev)
>>>    * VGA Connector
>>>    */
>>>   
>>> +static int ast_vga_connector_helper_detect_ctx(struct drm_connector *connector,
>>> +					       struct drm_modeset_acquire_ctx *ctx,
>>> +					       bool force)
>>> +{
>>> +	enum drm_connector_status status = connector_status_disconnected;
>>> +	const struct drm_edid *edid;
>>> +
>>> +	edid = drm_edid_read(connector);
>>> +	if (edid)
>>> +		status = connector_status_connected;
>>> +	drm_edid_free(edid);
>>> +
>>> +	return status;
>>> +}
>>> +
>> Yeah, I think it would be worth turning it into a helper. We have a
>> number of drivers using some variation of that already
>> (display-connector, loongson).
>>
>> It's probably better to use drm_probe_ddc here too instead of parsing
>> and updating the EDID property each time we call detect.
>
> I see. I'll update the patch accordingly. drm_probe_ddc() doesn't update 
> the property, which is probably better anyway.

The struct drm_edid based drm_edid_read() and friends do *not* parse the
EDID (apart from what's necessary to read the EDID) nor update the EDID
property. You need to call drm_edid_connector_update() separately for
that. This is by design.

The struct edid based drm_get_edid() does parse and update the property,
and I think adding that was a mistake that a lot of drivers later
depended on, and couldn't be removed.

As a design consideration, IMO it's also a fine approach to read and
cache the EDID and update the property at ->detect, and then use
drm_edid_connector_add_modes() in ->get_modes to avoid re-reading the
EDID at that time. It's not uncommon to need stuff from the EDID between
those hooks.


BR,
Jani.


-- 
Jani Nikula, Intel