Message ID | 1404415623-7198-1-git-send-email-imirkin@alum.mit.edu (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Fri, Jul 4, 2014 at 5:27 AM, Ilia Mirkin <imirkin@alum.mit.edu> wrote: > Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> > --- > > Based on a recent discussion in #radeon, and also my own observation that the > 'full' scaling causes no end of confusion among users. > > See https://bugs.freedesktop.org/show_bug.cgi?id=80868 for some more details, > although it is more radeon-specific. > > Side-note: the scaling mode update handler disallows setting the mode to none > for LVDS... should it be the same for eDP as well? Or perhaps relax the > restriction and let people do whatever they want? We do need some improvement here, for sure. What we do currently is confusing for a lot of cases, like where people select a (non-preferred) 24Hz mode or something and still get 60 because we always use the preferred mode on the backend (except for the 'None' scaling mode). I somewhat think we should probably default to something like "if mode came from edid (or the user?), no scaling" and "if it's from the random list of modes we add to fixed panels to keep old games happy, scale to closest edid mode / preferred mode / something".. I'm not entirely sure what the best choices are to be honest. Ben. > > drm/nouveau_connector.c | 7 +++++-- > 1 file changed, 5 insertions(+), 2 deletions(-) > > diff --git a/drm/nouveau_connector.c b/drm/nouveau_connector.c > index 1fa222e..295e222 100644 > --- a/drm/nouveau_connector.c > +++ b/drm/nouveau_connector.c > @@ -1192,6 +1192,8 @@ nouveau_connector_create(struct drm_device *dev, int index) > disp->color_vibrance_property, > 150); > > + nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; > + > switch (nv_connector->type) { > case DCB_CONNECTOR_VGA: > if (nv_device(drm->device)->card_type >= NV_50) { > @@ -1203,10 +1205,11 @@ nouveau_connector_create(struct drm_device *dev, int index) > case DCB_CONNECTOR_TV_0: > case DCB_CONNECTOR_TV_1: > case DCB_CONNECTOR_TV_3: > - nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; > break; > default: > - nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; > + if (type == DRM_MODE_CONNECTOR_LVDS || > + type == DRM_MODE_CONNECTOR_eDP) > + nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; > > drm_object_attach_property(&connector->base, > dev->mode_config.scaling_mode_property, > -- > 1.8.5.5 > > _______________________________________________ > Nouveau mailing list > Nouveau@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/nouveau
On Thu, Jul 3, 2014 at 8:27 PM, Ben Skeggs <skeggsb@gmail.com> wrote: > On Fri, Jul 4, 2014 at 5:27 AM, Ilia Mirkin <imirkin@alum.mit.edu> wrote: >> Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> >> --- >> >> Based on a recent discussion in #radeon, and also my own observation that the >> 'full' scaling causes no end of confusion among users. >> >> See https://bugs.freedesktop.org/show_bug.cgi?id=80868 for some more details, >> although it is more radeon-specific. >> >> Side-note: the scaling mode update handler disallows setting the mode to none >> for LVDS... should it be the same for eDP as well? Or perhaps relax the >> restriction and let people do whatever they want? > We do need some improvement here, for sure. What we do currently is > confusing for a lot of cases, like where people select a > (non-preferred) 24Hz mode or something and still get 60 because we > always use the preferred mode on the backend (except for the 'None' > scaling mode). Right. That was the confusion I was largely referring to. > > I somewhat think we should probably default to something like "if mode > came from edid (or the user?), no scaling" and "if it's from the > random list of modes we add to fixed panels to keep old games happy, > scale to closest edid mode / preferred mode / something".. I'm not > entirely sure what the best choices are to be honest. What are "fixed panels" (in the sense of... how does a driver determine this)? I figured it was just LVDS/eDP == fixed. Everything else == not fixed. At least that's what radeon does, according to... someone (Alex, IIRC). Unfortunately scaling mode is a connector-level property. So it may be confusing if the user were to set it to something, then connect a different screen, and have it all of a sudden reset itself to whatever we feel the default should be. I guess we could keep track of a "has the user futzed with this" bit. I guess what you're saying is more of a per-mode property (of which no such thing currently exists)... -ilia > > Ben. > >> >> drm/nouveau_connector.c | 7 +++++-- >> 1 file changed, 5 insertions(+), 2 deletions(-) >> >> diff --git a/drm/nouveau_connector.c b/drm/nouveau_connector.c >> index 1fa222e..295e222 100644 >> --- a/drm/nouveau_connector.c >> +++ b/drm/nouveau_connector.c >> @@ -1192,6 +1192,8 @@ nouveau_connector_create(struct drm_device *dev, int index) >> disp->color_vibrance_property, >> 150); >> >> + nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; >> + >> switch (nv_connector->type) { >> case DCB_CONNECTOR_VGA: >> if (nv_device(drm->device)->card_type >= NV_50) { >> @@ -1203,10 +1205,11 @@ nouveau_connector_create(struct drm_device *dev, int index) >> case DCB_CONNECTOR_TV_0: >> case DCB_CONNECTOR_TV_1: >> case DCB_CONNECTOR_TV_3: >> - nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; >> break; >> default: >> - nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; >> + if (type == DRM_MODE_CONNECTOR_LVDS || >> + type == DRM_MODE_CONNECTOR_eDP) >> + nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; >> >> drm_object_attach_property(&connector->base, >> dev->mode_config.scaling_mode_property, >> -- >> 1.8.5.5 >> >> _______________________________________________ >> Nouveau mailing list >> Nouveau@lists.freedesktop.org >> http://lists.freedesktop.org/mailman/listinfo/nouveau
diff --git a/drm/nouveau_connector.c b/drm/nouveau_connector.c index 1fa222e..295e222 100644 --- a/drm/nouveau_connector.c +++ b/drm/nouveau_connector.c @@ -1192,6 +1192,8 @@ nouveau_connector_create(struct drm_device *dev, int index) disp->color_vibrance_property, 150); + nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; + switch (nv_connector->type) { case DCB_CONNECTOR_VGA: if (nv_device(drm->device)->card_type >= NV_50) { @@ -1203,10 +1205,11 @@ nouveau_connector_create(struct drm_device *dev, int index) case DCB_CONNECTOR_TV_0: case DCB_CONNECTOR_TV_1: case DCB_CONNECTOR_TV_3: - nv_connector->scaling_mode = DRM_MODE_SCALE_NONE; break; default: - nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; + if (type == DRM_MODE_CONNECTOR_LVDS || + type == DRM_MODE_CONNECTOR_eDP) + nv_connector->scaling_mode = DRM_MODE_SCALE_FULLSCREEN; drm_object_attach_property(&connector->base, dev->mode_config.scaling_mode_property,
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu> --- Based on a recent discussion in #radeon, and also my own observation that the 'full' scaling causes no end of confusion among users. See https://bugs.freedesktop.org/show_bug.cgi?id=80868 for some more details, although it is more radeon-specific. Side-note: the scaling mode update handler disallows setting the mode to none for LVDS... should it be the same for eDP as well? Or perhaps relax the restriction and let people do whatever they want? drm/nouveau_connector.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)