[PATCH xserver 2/3] modesetting: ms_covering_crtc: Allow calling on non modesetting Screens

[Hans de Goede]

Hi,

On 27-08-16 20:03, Peter Wu wrote:
> On Wed, Aug 24, 2016 at 03:30:11PM +0200, Hans de Goede wrote:
>> 99% of the code in ms_covering_crtc is video-driver agnostic. Add a
>> screen_is_ms parameter when when FALSE skips the one ms specific check,
>> this will allow calling ms_covering_crtc on slave GPUs.
>>
>> Signed-off-by: Hans de Goede <hdegoede at redhat.com>
>> ---
>>  hw/xfree86/drivers/modesetting/vblank.c | 12 +++++++++---
>>  1 file changed, 9 insertions(+), 3 deletions(-)
>>
>> diff --git a/hw/xfree86/drivers/modesetting/vblank.c b/hw/xfree86/drivers/modesetting/vblank.c
>> index e738497..ec60ac4 100644
>> --- a/hw/xfree86/drivers/modesetting/vblank.c
>> +++ b/hw/xfree86/drivers/modesetting/vblank.c
>> @@ -97,7 +97,7 @@ ms_crtc_on(xf86CrtcPtr crtc)
>>   */
>>
>>  static xf86CrtcPtr
>> -ms_covering_crtc(ScreenPtr pScreen, BoxPtr box)
>> +ms_covering_crtc(ScreenPtr pScreen, BoxPtr box, Bool screen_is_ms)
>>  {
>>      ScrnInfoPtr scrn = xf86ScreenToScrn(pScreen);
>>      xf86CrtcConfigPtr xf86_config = XF86_CRTC_CONFIG_PTR(scrn);
>> @@ -105,14 +105,20 @@ ms_covering_crtc(ScreenPtr pScreen, BoxPtr box)
>>      int coverage, best_coverage;
>>      int c;
>>      BoxRec crtc_box, cover_box;
>> +    Bool crtc_on;
>>
>>      best_crtc = NULL;
>>      best_coverage = 0;
>>      for (c = 0; c < xf86_config->num_crtc; c++) {
>>          crtc = xf86_config->crtc[c];
>>
>> +        if (screen_is_ms)
>> +            crtc_on = ms_crtc_on(crtc);
>> +        else
>> +            crtc_on = crtc->enabled;
>
> This will skip the check whether a screen is on or off via DPMS, right?

Correct, but we do check that the primary fallback crtc is enabled
in patch 3/3 (which introduces the first caller with screen_is_ms = FALSE)
and it is reasonable (if not ideal) to assume that all screens are
on or off at the same time.

> If the DPMS property is somehow not valid for output slaves, shouldn't
> that be fixed instead?

The dpms state checked by ms_crtc_on is hold in a modesetting driver
private data struct, and the slaves we're calling ms_covering_crtc()
on in patch 3/3 may very well use a different driver and then we
would end up interpreting that driver's private data as if it is
modesetting driver data...

> (Please correct/educate me if I am wrong, this is all new for me ;))

No problem, thanks for taking a look at these patches, I hope the
above helps explain things (calling ms_covering_crtc on the crtc-s
of slaves is a bit of a hack, we must be careful to only access
core Xorg / xfree86 data and not driver specific data).

Regards,

Hans



>
> Kind regards,
> Peter
>
>> +
>>          /* If the CRTC is off, treat it as not covering */
>> -        if (!ms_crtc_on(crtc))
>> +        if (!crtc_on)
>>              continue;
>>
>>          ms_crtc_box(crtc, &crtc_box);
>> @@ -137,7 +143,7 @@ ms_dri2_crtc_covering_drawable(DrawablePtr pDraw)
>>      box.x2 = box.x1 + pDraw->width;
>>      box.y2 = box.y1 + pDraw->height;
>>
>> -    return ms_covering_crtc(pScreen, &box);
>> +    return ms_covering_crtc(pScreen, &box, TRUE);
>>  }
>>
>>  static Bool
>> --
>> 2.9.3
>>
>