| Message ID | 162854cb-c7bd-d9ce-9fa0-9a6cd89c621b@xs4all.nl (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Dec 15, 2017 at 04:21:50PM +0100, Hans Verkuil wrote: > When I connected my cubieboard running 4.15-rc1 to my 4k display I got no picture. Some > digging found that there is no check against the upper pixelclock limit of the HDMI > output, so X selects a 4kp60 format at 594 MHz, which obviously won't work. > > The patch below adds a check for the upper bound of what this hardware can do, and > it checks if the requested tmds clock can be obtained. > > It also allows for the +/- 0.5% pixel clock variation that the HDMI spec permits. > > That code is based on commit 22d0be2a557e53a22feb484e8fce255fe09e6ad5 from > Jose Abreu for drm/arc. > > Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> > Thanks-to: Jose Abreu <Jose.Abreu@synopsys.com> > --- > Changes since v2: > - Allow for the 0.5% variation. > --- > drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c | 19 +++++++++++++++++++ > 1 file changed, 19 insertions(+) > > diff --git a/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c b/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c > index dda904ec0534..a7e9eb93c378 100644 > --- a/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c > +++ b/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c > @@ -208,8 +208,27 @@ static int sun4i_hdmi_get_modes(struct drm_connector *connector) > return ret; > } > > +static int sun4i_hdmi_mode_valid(struct drm_connector *connector, > + struct drm_display_mode *mode) > +{ This only catches bad modes in the probe code, userspace can still request bad modes directly through atomic or SETCRTC ioctls. Please read the kerneldoc for drm_connector_helper_funcs->mode_valid and put your shiny mode_valid callback into the right vtable. Cheers, Daniel > + struct sun4i_hdmi *hdmi = drm_connector_to_sun4i_hdmi(connector); > + unsigned long rate = mode->clock * 1000; > + long diff = rate / 200; /* +-0.5% allowed by HDMI spec */ > + long rounded_rate; > + > + /* 165 MHz is the typical max pixelclock frequency for HDMI <= 1.2 */ > + if (rate > 165000000) > + return MODE_CLOCK_HIGH; > + rounded_rate = clk_round_rate(hdmi->tmds_clk, rate); > + if (max(rounded_rate, rate) - min(rounded_rate, rate) < diff && > + rounded_rate > 0) > + return MODE_OK; > + return MODE_NOCLOCK; > +} > + > static const struct drm_connector_helper_funcs sun4i_hdmi_connector_helper_funcs = { > .get_modes = sun4i_hdmi_get_modes, > + .mode_valid = sun4i_hdmi_mode_valid, > }; > > static enum drm_connector_status > -- > 2.15.0 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/dri-devel
Hi, On Fri, Dec 15, 2017 at 04:21:50PM +0100, Hans Verkuil wrote: > When I connected my cubieboard running 4.15-rc1 to my 4k display I got no picture. Some > digging found that there is no check against the upper pixelclock limit of the HDMI > output, so X selects a 4kp60 format at 594 MHz, which obviously won't work. > > The patch below adds a check for the upper bound of what this hardware can do, and > it checks if the requested tmds clock can be obtained. > > It also allows for the +/- 0.5% pixel clock variation that the HDMI spec permits. > > That code is based on commit 22d0be2a557e53a22feb484e8fce255fe09e6ad5 from > Jose Abreu for drm/arc. > > Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> > Thanks-to: Jose Abreu <Jose.Abreu@synopsys.com> I've fixed the checkpatch warnings, and the compilation warning about the type mismatch (between long and unsigned long), and applied. Thanks! Maxime
diff --git a/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c b/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c index dda904ec0534..a7e9eb93c378 100644 --- a/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c +++ b/drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c @@ -208,8 +208,27 @@ static int sun4i_hdmi_get_modes(struct drm_connector *connector) return ret; } +static int sun4i_hdmi_mode_valid(struct drm_connector *connector, + struct drm_display_mode *mode) +{ + struct sun4i_hdmi *hdmi = drm_connector_to_sun4i_hdmi(connector); + unsigned long rate = mode->clock * 1000; + long diff = rate / 200; /* +-0.5% allowed by HDMI spec */ + long rounded_rate; + + /* 165 MHz is the typical max pixelclock frequency for HDMI <= 1.2 */ + if (rate > 165000000) + return MODE_CLOCK_HIGH; + rounded_rate = clk_round_rate(hdmi->tmds_clk, rate); + if (max(rounded_rate, rate) - min(rounded_rate, rate) < diff && + rounded_rate > 0) + return MODE_OK; + return MODE_NOCLOCK; +} + static const struct drm_connector_helper_funcs sun4i_hdmi_connector_helper_funcs = { .get_modes = sun4i_hdmi_get_modes, + .mode_valid = sun4i_hdmi_mode_valid, }; static enum drm_connector_status
When I connected my cubieboard running 4.15-rc1 to my 4k display I got no picture. Some digging found that there is no check against the upper pixelclock limit of the HDMI output, so X selects a 4kp60 format at 594 MHz, which obviously won't work. The patch below adds a check for the upper bound of what this hardware can do, and it checks if the requested tmds clock can be obtained. It also allows for the +/- 0.5% pixel clock variation that the HDMI spec permits. That code is based on commit 22d0be2a557e53a22feb484e8fce255fe09e6ad5 from Jose Abreu for drm/arc. Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Thanks-to: Jose Abreu <Jose.Abreu@synopsys.com> --- Changes since v2: - Allow for the 0.5% variation. --- drivers/gpu/drm/sun4i/sun4i_hdmi_enc.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)