| Message ID | CAAuj=_fEjHCsGYzdA20LvP_292oaTHEC4PE4uNFdWHE4UkBMJw@mail.gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915/display: Allow tighter hblank span | expand |
On Sun, Feb 18, 2024 at 07:12:24PM +0100, Bas S wrote: > I ran into an issue with the i915 driver not being able to drive a display > with this specific modeline: Could you please file a bug on this? https://drm.pages.freedesktop.org/intel-docs/how-to-file-i915-bugs.html The limits below are per platform/display-architecture and probably forgotten to get updates for newer platforms. But having more information about your platform would be very helpful. Thanks, Rodrigo. > > [drm]] Modeline "1920x720": 60 120980 1920 1932 1936 1948 720 723 733 1035 > 0x48 0x9 > [drm:drm_mode_prune_invalid [drm]] Not using 1920x720 mode: H_ILLEGAL > > After some investigation I found that intel_mode_valid (and in newer > kernels, intel_cpu_transcoder_mode_valid) returns MODE_H_ILLEGAL due to > (htotal - hdisplay) being lower than 32. > The modeline in question indeed does not satisfy this constraint, as > HTOTAL(1948) - HDISPLAY(1920) equals 28. > Changing the driver code to allow for a hblank span of 28 pixels or lower > resulted in the driver successfully rendering to the display. > As such I propose this patch to allow for a tighter hblank span. > > Nb: I am uncertain if the hblank span of 32 pixels has been chosen > deliberately and what the side-effects could be of lowering this value. > Any insights into this or alternative solutions would be very much > appreciated! I also considered introducing a kernel module parameter to > optionally loosen these mode constraints. > > The referenced modeline is present in a line of ultrawide signage displays > and has been known to work on other graphics drivers/OSes. > > Signed-off-by: Sebastiaan Schalbroeck <[1]schalbroeck@gmail.com> > --- > drivers/gpu/drm/i915/display/intel_display.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c > b/drivers/gpu/drm/i915/display/intel_display.c > index b10aad1..f6aba1d 100644 > --- a/drivers/gpu/drm/i915/display/intel_display.c > +++ b/drivers/gpu/drm/i915/display/intel_display.c > @@ -7745,13 +7745,13 @@ enum drm_mode_status > intel_cpu_transcoder_mode_valid(struct drm_i915_private *de > */ > if (DISPLAY_VER(dev_priv) >= 5) { > if (mode->hdisplay < 64 || > - mode->htotal - mode->hdisplay < 32) > + mode->htotal - mode->hdisplay < 28) > return MODE_H_ILLEGAL; > > if (mode->vtotal - mode->vdisplay < 5) > return MODE_V_ILLEGAL; > } else { > - if (mode->htotal - mode->hdisplay < 32) > + if (mode->htotal - mode->hdisplay < 28) > return MODE_H_ILLEGAL; > > if (mode->vtotal - mode->vdisplay < 3) > > -- > 2.39.2 > > References > > Visible links > 1. mailto:schalbroeck@gmail.com
diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c index b10aad1..f6aba1d 100644 --- a/drivers/gpu/drm/i915/display/intel_display.c +++ b/drivers/gpu/drm/i915/display/intel_display.c @@ -7745,13 +7745,13 @@ enum drm_mode_status intel_cpu_transcoder_mode_valid(struct drm_i915_private *de */ if (DISPLAY_VER(dev_priv) >= 5) { if (mode->hdisplay < 64 || - mode->htotal - mode->hdisplay < 32) + mode->htotal - mode->hdisplay < 28) return MODE_H_ILLEGAL; if (mode->vtotal - mode->vdisplay < 5) return MODE_V_ILLEGAL; } else { - if (mode->htotal - mode->hdisplay < 32) + if (mode->htotal - mode->hdisplay < 28) return MODE_H_ILLEGAL;
I ran into an issue with the i915 driver not being able to drive a display with this specific modeline: [drm]] Modeline "1920x720": 60 120980 1920 1932 1936 1948 720 723 733 1035 0x48 0x9 [drm:drm_mode_prune_invalid [drm]] Not using 1920x720 mode: H_ILLEGAL After some investigation I found that intel_mode_valid (and in newer kernels, intel_cpu_transcoder_mode_valid) returns MODE_H_ILLEGAL due to (htotal - hdisplay) being lower than 32. The modeline in question indeed does not satisfy this constraint, as HTOTAL(1948) - HDISPLAY(1920) equals 28. Changing the driver code to allow for a hblank span of 28 pixels or lower resulted in the driver successfully rendering to the display. As such I propose this patch to allow for a tighter hblank span. Nb: I am uncertain if the hblank span of 32 pixels has been chosen deliberately and what the side-effects could be of lowering this value. Any insights into this or alternative solutions would be very much appreciated! I also considered introducing a kernel module parameter to optionally loosen these mode constraints. The referenced modeline is present in a line of ultrawide signage displays and has been known to work on other graphics drivers/OSes. Signed-off-by: Sebastiaan Schalbroeck <schalbroeck@gmail.com> --- drivers/gpu/drm/i915/display/intel_display.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) if (mode->vtotal - mode->vdisplay < 3) -- 2.39.2