drm/i915: Avoid div by zero when pixel clock is large
Subsystems (2)
PATCH 1/1 drm/i915: Avoid div by zero when pixel clock is large
From: Ville Syrjälä <[email protected]>
Make sure the line_time_us isn't zero in the gmch watermarks code as that would cause a div by zero. This can be triggered by specifying a very fast pixel clock for the mode.
At some point we should probably just switch over to using the same math we use on PCH platforms which avoids such intermediate rounded results.
Also we should verify the user provided mode much more rigorously. At the moment we accept pretty much anything.
--- drivers/gpu/drm/i915/intel_pm.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +4 -4
@@ -1131,7 +1131,7 @@ static bool g4x_compute_wm0(struct drm_device *dev, *plane_wm = display->max_wm;/* Use the large buffer method to calculate cursor watermark */ - line_time_us = ((htotal *1000 ) / clock); + line_time_us =max (htotal *1000 / clock,1 ); line_count = (cursor_latency_ns / line_time_us +1000 ) /1000 ; entries = line_count *64 * pixel_size; tlb_miss = cursor->fifo_size*cursor->cacheline_size - hdisplay *8 ;
@@ -1207,7 +1207,7 @@ static bool g4x_compute_srwm(struct drm_device *dev, hdisplay =to_intel_crtc (crtc)->config.pipe_src_w; pixel_size = crtc->fb->bits_per_pixel /8 ; - line_time_us = (htotal *1000 ) / clock; + line_time_us =max (htotal *1000 / clock,1 ); line_count = (latency_ns / line_time_us +1000 ) /1000 ; line_size = hdisplay * pixel_size;
@@ -1440,7 +1440,7 @@ static void i965_update_wm(struct drm_crtc *unused_crtc)unsigned long line_time_us;int entries; - line_time_us = ((htotal *1000 ) / clock); + line_time_us =max (htotal *1000 / clock,1 );/* Use ns/us then divide to preserve precision */ entries = (((sr_latency_ns / line_time_us) + 1000) / 1000) *
@@ -1566,7 +1566,7 @@ static void i9xx_update_wm(struct drm_crtc *unused_crtc)unsigned long line_time_us;int entries; - line_time_us = (htotal *1000 ) / clock; + line_time_us =max (htotal *1000 / clock,1 );/* Use ns/us then divide to preserve precision */ entries = (((sr_latency_ns / line_time_us) + 1000) / 1000) *
Signature · 6 lines
1.8.3.2 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 5 lines · On Fri, Feb 14, 2014 at 02:18:57PM +0200, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
Make sure the
line_time_usisn't zero in the gmch watermarks code as that would cause a div by zero. This can be triggered by specifying a very fast pixel clock for the mode.
Very fast but valid? If it was just very fast, we should have rejected the mode line. On the other hand, if it was valid, maybe we should adjust the bias (* 1000 / 1000) so that the computation remains reasonably accurate because we will need a larger FIFO to accommodate the higher clock rate.
Anyway, even with the larger bias, such a safe guard is sensisble. I can't argue against this patch, but I just wonder if you opened a can of worms.
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Quoted 8 lines · Chris Wilson
Quoted 5 lines · On Fri, Feb 14, 2014 at 02:18:57PM +0200, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
Make sure the
line_time_usisn't zero in the gmch watermarks code as that would cause a div by zero. This can be triggered by specifying a very fast pixel clock for the mode.Very fast but valid?
74.25 GHz. I'm not sure which way you would classify that :)
Quoted 2 lines
If it was just very fast, we should have rejected
the mode line.
Except we do more or less zero checks on the mode supplied to the setcrtc ioctl. Something we should really get fixed at some point.
Signature · 2 lines
Ville Syrjälä Intel OTC
Daniel Vetter
AppliedQuoted 11 lines · Ville Syrjälä
Quoted 8 lines · Chris Wilson
Quoted 5 lines · On Fri, Feb 14, 2014 at 02:18:57PM +0200, [email protected] wrote:
From: Ville Syrjälä <[email protected]>
Make sure the
line_time_usisn't zero in the gmch watermarks code as that would cause a div by zero. This can be triggered by specifying a very fast pixel clock for the mode.Very fast but valid?
74.25 GHz. I'm not sure which way you would classify that :)
I've added this to the commit message ...
Quoted 5 lines
Quoted 2 lines
If it was just very fast, we should have rejected
the mode line.Except we do more or less zero checks on the mode supplied to the setcrtc ioctl. Something we should really get fixed at some point.
Queued for -next, thanks for the patch.
-Daniel
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch