On 04.08.2016 19:56, Daniel Vetter wrote:
> On Thu, Aug 04, 2016 at 12:39:41PM +0900, Michel Dänzer wrote:
>> From: Michel Dänzer <michel.daenzer at amd.com>
>>
>> @@ -543,15 +549,25 @@ struct drm_color_lut {
>> * 'as soon as possible', meaning that it not delay waiting for vblank.
>> * This may cause tearing on the screen.
>> *
>> - * The reserved field must be zero until we figure out something
>> - * clever to use it for.
>> + * The sequence field must be zero unless either of the
>> + * DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE/RELATIVE flags is specified. When
>> + * the ABSOLUTE flag is specified, the sequence field denotes the absolute
>> + * vblank sequence when the flip should take effect. When the RELATIVE
>> + * flag is specified, the sequence field denotes the relative (to the
>> + * current one when the ioctl is called) vblank sequence when the flip
>> + * should take effect. NOTE: DRM_IOCTL_WAIT_VBLANK must still be used to
>> + * make sure the vblank sequence before the target one has passed before
>> + * calling this ioctl. The purpose of the
>> + * DRM_MODE_PAGE_FLIP_TARGET_ABSOLUTE/RELATIVE flags is merely to clarify
>> + * the target for when code dealing with a page flip runs during a
>> + * vertical blank period.
>> */
>>
>> struct drm_mode_crtc_page_flip {
>> __u32 crtc_id;
>> __u32 fb_id;
>> __u32 flags;
>> - __u32 reserved;
>> + __u32 sequence;
>
> Might break abi somewhere. I think it'd be better to create a
> struct drm_mode_crtc_page_flip2 with the renamed field.
It doesn't break ABI. I guess you mean there might be userspace code
referencing the reserved field? Such code would have to be open-coding
drmModePageFlip instead of using it. And it turns out you're right, e.g.
SNA actually does that and would fail to compile... Will be fixed in v2.
> Reviewed-by: Daniel Vetter <daniel.vetter at ffwll.ch>
Thanks! Will it be okay to merge this via Alex's tree?
--
Earthling Michel Dänzer | http://www.amd.com
Libre software enthusiast | Mesa and X developer
