On Fri, Oct 11, 2019 at 04:51:13PM -0400, Lyude Paul wrote:
a little late but: i915 does have this hack (or rather-possible_crtcs with MST in i915 has been broken for a while and got fixed, but had to get reverted because of this issue), it's where this originally came from.
Hm since this is widespread I think we should check for this when we register connectors (either in drm_dev_register, or hotplugged ones). I think just validating that all encoder->possible_crtc match and WARN_ON if not would be really good.
2nd option would be to do that in the GETENCODERS ioctl. That would at least keep the encoders useful for driver-internal stuff. We could then un-revert the i915 patch again.
Either way I think we should have this hack + comment with links to the offending userspace in common code, not duplicated over all drivers. -Daniel
On Wed, 2019-10-09 at 17:01 +0200, Daniel Vetter wrote:
On Fri, Sep 27, 2019 at 11:27:41AM -0400, Sean Paul wrote:
On Thu, Sep 26, 2019 at 06:51:07PM -0400, Lyude Paul wrote:
This commit is seperate from the previous one to make it easier to revert in the future. Basically, there's multiple userspace applications that interpret possible_crtcs very wrong:
https://gitlab.freedesktop.org/xorg/xserver/merge_requests/277 https://gitlab.gnome.org/GNOME/mutter/issues/759
While work is ongoing to fix these issues in userspace, we need to report ->possible_crtcs incorrectly for now in order to avoid introducing a regression in in userspace. Once these issues get fixed, this commit should be reverted.
Signed-off-by: Lyude Paul lyude@redhat.com Cc: Ville Syrjälä ville.syrjala@linux.intel.com
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 11 +++++++++++ 1 file changed, 11 insertions(+)
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index b404f1ae6df7..fe8ac801d7a5 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -4807,6 +4807,17 @@ static int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm, if (!acrtc->mst_encoder) goto fail;
- /*
* FIXME: This is a hack to workaround the following issues:** https://gitlab.gnome.org/GNOME/mutter/issues/759* https://gitlab.freedesktop.org/xorg/xserver/merge_requests/277** One these issues are closed, this should be removedEven when these issues are closed, we'll still be introducing a regression if we revert this change. Time for actually_possible_crtcs? :)
You also might want to briefly explain the u/s bug in case the links go sour.
*/- acrtc->mst_encoder->base.possible_crtcs =
amdgpu_dm_get_encoder_crtc_mask(dm->adev);Why don't we put this hack in amdgpu_dm_dp_create_fake_mst_encoder()?
If we don't have the same hack for i915 mst I think we shouldn't merge this ... broken userspace is broken. -Daniel
-- Cheers, Lyude Paul