| Message ID | 1454671768-27062-1-git-send-email-mika.kahola@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 2/5/2016 4:59 PM, Mika Kahola wrote: > Skip DDI PLL selection if display type is DSI/MIPI. > > Signed-off-by: Mika Kahola <mika.kahola@intel.com> > --- > drivers/gpu/drm/i915/intel_display.c | 9 +++++++-- > 1 file changed, 7 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c > index d7de2a5..5da98b2 100644 > --- a/drivers/gpu/drm/i915/intel_display.c > +++ b/drivers/gpu/drm/i915/intel_display.c > @@ -9902,8 +9902,13 @@ static void broadwell_modeset_commit_cdclk(struct drm_atomic_state *old_state) > static int haswell_crtc_compute_clock(struct intel_crtc *crtc, > struct intel_crtc_state *crtc_state) > { > - if (!intel_ddi_pll_select(crtc, crtc_state)) > - return -EINVAL; > + struct intel_encoder *intel_encoder = > + intel_ddi_get_crtc_new_encoder(crtc_state); > + > + if (intel_encoder->type != INTEL_OUTPUT_DSI) { > + if (!intel_ddi_pll_select(crtc, crtc_state)) > + return -EINVAL; > + } > can this be moved inside bxt_ddi_pll_select ? we can avoid this check for other platforms that also execute this function. regards, Sivakumar > crtc->lowfreq_avail = false; >
On Tue, 09 Feb 2016, "Thulasimani, Sivakumar" <sivakumar.thulasimani@intel.com> wrote: > On 2/5/2016 4:59 PM, Mika Kahola wrote: >> Skip DDI PLL selection if display type is DSI/MIPI. >> >> Signed-off-by: Mika Kahola <mika.kahola@intel.com> >> --- >> drivers/gpu/drm/i915/intel_display.c | 9 +++++++-- >> 1 file changed, 7 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c >> index d7de2a5..5da98b2 100644 >> --- a/drivers/gpu/drm/i915/intel_display.c >> +++ b/drivers/gpu/drm/i915/intel_display.c >> @@ -9902,8 +9902,13 @@ static void broadwell_modeset_commit_cdclk(struct drm_atomic_state *old_state) >> static int haswell_crtc_compute_clock(struct intel_crtc *crtc, >> struct intel_crtc_state *crtc_state) >> { >> - if (!intel_ddi_pll_select(crtc, crtc_state)) >> - return -EINVAL; >> + struct intel_encoder *intel_encoder = >> + intel_ddi_get_crtc_new_encoder(crtc_state); >> + >> + if (intel_encoder->type != INTEL_OUTPUT_DSI) { >> + if (!intel_ddi_pll_select(crtc, crtc_state)) >> + return -EINVAL; >> + } >> > can this be moved inside bxt_ddi_pll_select ? we can avoid this check for > other platforms that also execute this function. I asked Mika to do it this way, but if you feel strongly about it I guess I could be persuaded otherwise too. My main point is, if we pass on DSI encoders to DDI functions in some cases but mostly not, it will muddy the waters and eventually people end up checking for "is dsi" all around DDI just because they can't be bothered to check if the functions are really called for DDI only or not. It's more of a maintainability concern than anything else. BR, Jani. > > regards, > Sivakumar >> crtc->lowfreq_avail = false; >> >
On 2/9/2016 12:02 PM, Jani Nikula wrote: > On Tue, 09 Feb 2016, "Thulasimani, Sivakumar" <sivakumar.thulasimani@intel.com> wrote: >> On 2/5/2016 4:59 PM, Mika Kahola wrote: >>> Skip DDI PLL selection if display type is DSI/MIPI. >>> >>> Signed-off-by: Mika Kahola <mika.kahola@intel.com> >>> --- >>> drivers/gpu/drm/i915/intel_display.c | 9 +++++++-- >>> 1 file changed, 7 insertions(+), 2 deletions(-) >>> >>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c >>> index d7de2a5..5da98b2 100644 >>> --- a/drivers/gpu/drm/i915/intel_display.c >>> +++ b/drivers/gpu/drm/i915/intel_display.c >>> @@ -9902,8 +9902,13 @@ static void broadwell_modeset_commit_cdclk(struct drm_atomic_state *old_state) >>> static int haswell_crtc_compute_clock(struct intel_crtc *crtc, >>> struct intel_crtc_state *crtc_state) >>> { >>> - if (!intel_ddi_pll_select(crtc, crtc_state)) >>> - return -EINVAL; >>> + struct intel_encoder *intel_encoder = >>> + intel_ddi_get_crtc_new_encoder(crtc_state); >>> + >>> + if (intel_encoder->type != INTEL_OUTPUT_DSI) { >>> + if (!intel_ddi_pll_select(crtc, crtc_state)) >>> + return -EINVAL; >>> + } >>> >> can this be moved inside bxt_ddi_pll_select ? we can avoid this check for >> other platforms that also execute this function. > I asked Mika to do it this way, but if you feel strongly about it I > guess I could be persuaded otherwise too. > > My main point is, if we pass on DSI encoders to DDI functions in some > cases but mostly not, it will muddy the waters and eventually people end > up checking for "is dsi" all around DDI just because they can't be > bothered to check if the functions are really called for DDI only or > not. It's more of a maintainability concern than anything else. > > BR, > Jani. > i am fine with this either way. i was thinking of avoid such checks in other platforms where it is not needed but your concern of too many is_dsi checks is valid as well. with that i am fine with this change as is. Reviewed-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> > >> regards, >> Sivakumar >>> crtc->lowfreq_avail = false; >>>
On Tue, Feb 09, 2016 at 01:16:18PM +0530, Thulasimani, Sivakumar wrote: > > > On 2/9/2016 12:02 PM, Jani Nikula wrote: > > On Tue, 09 Feb 2016, "Thulasimani, Sivakumar" <sivakumar.thulasimani@intel.com> wrote: > >> On 2/5/2016 4:59 PM, Mika Kahola wrote: > >>> Skip DDI PLL selection if display type is DSI/MIPI. > >>> > >>> Signed-off-by: Mika Kahola <mika.kahola@intel.com> > >>> --- > >>> drivers/gpu/drm/i915/intel_display.c | 9 +++++++-- > >>> 1 file changed, 7 insertions(+), 2 deletions(-) > >>> > >>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c > >>> index d7de2a5..5da98b2 100644 > >>> --- a/drivers/gpu/drm/i915/intel_display.c > >>> +++ b/drivers/gpu/drm/i915/intel_display.c > >>> @@ -9902,8 +9902,13 @@ static void broadwell_modeset_commit_cdclk(struct drm_atomic_state *old_state) > >>> static int haswell_crtc_compute_clock(struct intel_crtc *crtc, > >>> struct intel_crtc_state *crtc_state) > >>> { > >>> - if (!intel_ddi_pll_select(crtc, crtc_state)) > >>> - return -EINVAL; > >>> + struct intel_encoder *intel_encoder = > >>> + intel_ddi_get_crtc_new_encoder(crtc_state); > >>> + > >>> + if (intel_encoder->type != INTEL_OUTPUT_DSI) { > >>> + if (!intel_ddi_pll_select(crtc, crtc_state)) > >>> + return -EINVAL; > >>> + } > >>> > >> can this be moved inside bxt_ddi_pll_select ? we can avoid this check for > >> other platforms that also execute this function. > > I asked Mika to do it this way, but if you feel strongly about it I > > guess I could be persuaded otherwise too. > > > > My main point is, if we pass on DSI encoders to DDI functions in some > > cases but mostly not, it will muddy the waters and eventually people end > > up checking for "is dsi" all around DDI just because they can't be > > bothered to check if the functions are really called for DDI only or > > not. It's more of a maintainability concern than anything else. > > > > BR, > > Jani. > > > i am fine with this either way. i was thinking of avoid such checks > in other platforms where it is not needed but your concern of > too many is_dsi checks is valid as well. > with that i am fine with this change as is. > Reviewed-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> Another idea would be to use the clock_set thing to skip it, but I think historically that has only been used to skip the PLL calculations, not the PLL selection. So might be it would just confuse things more.
On Tue, 09 Feb 2016, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote: > On Tue, Feb 09, 2016 at 01:16:18PM +0530, Thulasimani, Sivakumar wrote: >> >> >> On 2/9/2016 12:02 PM, Jani Nikula wrote: >> > On Tue, 09 Feb 2016, "Thulasimani, Sivakumar" <sivakumar.thulasimani@intel.com> wrote: >> >> On 2/5/2016 4:59 PM, Mika Kahola wrote: >> >>> Skip DDI PLL selection if display type is DSI/MIPI. >> >>> >> >>> Signed-off-by: Mika Kahola <mika.kahola@intel.com> >> >>> --- >> >>> drivers/gpu/drm/i915/intel_display.c | 9 +++++++-- >> >>> 1 file changed, 7 insertions(+), 2 deletions(-) >> >>> >> >>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c >> >>> index d7de2a5..5da98b2 100644 >> >>> --- a/drivers/gpu/drm/i915/intel_display.c >> >>> +++ b/drivers/gpu/drm/i915/intel_display.c >> >>> @@ -9902,8 +9902,13 @@ static void broadwell_modeset_commit_cdclk(struct drm_atomic_state *old_state) >> >>> static int haswell_crtc_compute_clock(struct intel_crtc *crtc, >> >>> struct intel_crtc_state *crtc_state) >> >>> { >> >>> - if (!intel_ddi_pll_select(crtc, crtc_state)) >> >>> - return -EINVAL; >> >>> + struct intel_encoder *intel_encoder = >> >>> + intel_ddi_get_crtc_new_encoder(crtc_state); >> >>> + >> >>> + if (intel_encoder->type != INTEL_OUTPUT_DSI) { >> >>> + if (!intel_ddi_pll_select(crtc, crtc_state)) >> >>> + return -EINVAL; >> >>> + } >> >>> >> >> can this be moved inside bxt_ddi_pll_select ? we can avoid this check for >> >> other platforms that also execute this function. >> > I asked Mika to do it this way, but if you feel strongly about it I >> > guess I could be persuaded otherwise too. >> > >> > My main point is, if we pass on DSI encoders to DDI functions in some >> > cases but mostly not, it will muddy the waters and eventually people end >> > up checking for "is dsi" all around DDI just because they can't be >> > bothered to check if the functions are really called for DDI only or >> > not. It's more of a maintainability concern than anything else. >> > >> > BR, >> > Jani. >> > >> i am fine with this either way. i was thinking of avoid such checks >> in other platforms where it is not needed but your concern of >> too many is_dsi checks is valid as well. >> with that i am fine with this change as is. >> Reviewed-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> > > Another idea would be to use the clock_set thing to skip it, but > I think historically that has only been used to skip the PLL > calculations, not the PLL selection. So might be it would just confuse > things more. I just pushed this one. Thanks for the patch and review. BR, Jani.
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index d7de2a5..5da98b2 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9902,8 +9902,13 @@ static void broadwell_modeset_commit_cdclk(struct drm_atomic_state *old_state) static int haswell_crtc_compute_clock(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state) { - if (!intel_ddi_pll_select(crtc, crtc_state)) - return -EINVAL; + struct intel_encoder *intel_encoder = + intel_ddi_get_crtc_new_encoder(crtc_state); + + if (intel_encoder->type != INTEL_OUTPUT_DSI) { + if (!intel_ddi_pll_select(crtc, crtc_state)) + return -EINVAL; + } crtc->lowfreq_avail = false;
Skip DDI PLL selection if display type is DSI/MIPI. Signed-off-by: Mika Kahola <mika.kahola@intel.com> --- drivers/gpu/drm/i915/intel_display.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)