| Message ID | 1454894009-15466-4-git-send-email-mario.kleiner.de@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Feb 08, 2016 at 02:13:26AM +0100, Mario Kleiner wrote: > Changes to drm_update_vblank_count() in Linux 4.4 broke the > behaviour of the pre/post modeset functions as the new update > code doesn't deal with hw vblank counter resets inbetween calls > to drm_vblank_pre_modeset an drm_vblank_post_modeset, as it > should. > > This causes mistreatment of such hw counter resets as counter > wraparound, and thereby large forward jumps of the software > vblank counter which in turn cause vblank event dispatching > and vblank waits to fail/hang --> userspace clients hang. > > This symptom was reported on radeon-kms to cause a infinite > hang of KDE Plasma 5 shell's login procedure, preventing users > from logging in. > > Fix this by detecting when drm_update_vblank_count() is called > inside a pre->post modeset interval. If so, clamp valid vblank > increments to the safe values 0 and 1, pretty much restoring > the update behavior of the old update code of Linux 4.3 and > earlier. Also reset the last recorded hw vblank count at call > to drm_vblank_post_modeset() to be safe against hw that after > modesetting, dpms on etc. only fires its first vblank irq after > drm_vblank_post_modeset() was already called. > > Reported-by: Vlastimil Babka <vbabka@suse.cz> > Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> > Cc: <stable@vger.kernel.org> # 4.4+ > Cc: michel@daenzer.net > Cc: vbabka@suse.cz > Cc: ville.syrjala@linux.intel.com > Cc: daniel.vetter@ffwll.ch > Cc: dri-devel@lists.freedesktop.org > Cc: alexander.deucher@amd.com > Cc: christian.koenig@amd.com We need to untangle the new vblank stuff that assumes solide (atomic) drivers and all the old stuff much more. But that can be done later on. Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> > --- > drivers/gpu/drm/drm_irq.c | 16 ++++++++++++++++ > 1 file changed, 16 insertions(+) > > diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c > index aa2c74b..5c27ad3 100644 > --- a/drivers/gpu/drm/drm_irq.c > +++ b/drivers/gpu/drm/drm_irq.c > @@ -222,6 +222,21 @@ static void drm_update_vblank_count(struct drm_device *dev, unsigned int pipe, > } > > /* > + * Within a drm_vblank_pre_modeset - drm_vblank_post_modeset > + * interval? If so then vblank irqs keep running and it will likely > + * happen that the hardware vblank counter is not trustworthy as it > + * might reset at some point in that interval and vblank timestamps > + * are not trustworthy either in that interval. Iow. this can result > + * in a bogus diff >> 1 which must be avoided as it would cause > + * random large forward jumps of the software vblank counter. > + */ > + if (diff > 1 && (vblank->inmodeset & 0x2)) { > + DRM_DEBUG_VBL("clamping vblank bump to 1 on crtc %u: diffr=%u" > + " due to pre-modeset.\n", pipe, diff); > + diff = 1; > + } > + > + /* > * Restrict the bump of the software vblank counter to a safe maximum > * value of +1 whenever there is the possibility that concurrent readers > * of vblank timestamps could be active at the moment, as the current > @@ -1573,6 +1588,7 @@ void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe) > if (vblank->inmodeset) { > spin_lock_irqsave(&dev->vbl_lock, irqflags); > dev->vblank_disable_allowed = true; > + drm_reset_vblank_timestamp(dev, pipe); > spin_unlock_irqrestore(&dev->vbl_lock, irqflags); > > if (vblank->inmodeset & 0x2) > -- > 1.9.1 >
On 02/08/2016 02:13 AM, Mario Kleiner wrote: > Changes to drm_update_vblank_count() in Linux 4.4 broke the > behaviour of the pre/post modeset functions as the new update > code doesn't deal with hw vblank counter resets inbetween calls > to drm_vblank_pre_modeset an drm_vblank_post_modeset, as it > should. > > This causes mistreatment of such hw counter resets as counter > wraparound, and thereby large forward jumps of the software > vblank counter which in turn cause vblank event dispatching > and vblank waits to fail/hang --> userspace clients hang. > > This symptom was reported on radeon-kms to cause a infinite > hang of KDE Plasma 5 shell's login procedure, preventing users > from logging in. > > Fix this by detecting when drm_update_vblank_count() is called > inside a pre->post modeset interval. If so, clamp valid vblank > increments to the safe values 0 and 1, pretty much restoring > the update behavior of the old update code of Linux 4.3 and > earlier. Also reset the last recorded hw vblank count at call > to drm_vblank_post_modeset() to be safe against hw that after > modesetting, dpms on etc. only fires its first vblank irq after > drm_vblank_post_modeset() was already called. > > Reported-by: Vlastimil Babka <vbabka@suse.cz> FWIW, I've applied the whole patchset to 4.4 and the kde5 login problem didn't occur. I can test the next version too. Thanks, Vlastimil
diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index aa2c74b..5c27ad3 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -222,6 +222,21 @@ static void drm_update_vblank_count(struct drm_device *dev, unsigned int pipe, } /* + * Within a drm_vblank_pre_modeset - drm_vblank_post_modeset + * interval? If so then vblank irqs keep running and it will likely + * happen that the hardware vblank counter is not trustworthy as it + * might reset at some point in that interval and vblank timestamps + * are not trustworthy either in that interval. Iow. this can result + * in a bogus diff >> 1 which must be avoided as it would cause + * random large forward jumps of the software vblank counter. + */ + if (diff > 1 && (vblank->inmodeset & 0x2)) { + DRM_DEBUG_VBL("clamping vblank bump to 1 on crtc %u: diffr=%u" + " due to pre-modeset.\n", pipe, diff); + diff = 1; + } + + /* * Restrict the bump of the software vblank counter to a safe maximum * value of +1 whenever there is the possibility that concurrent readers * of vblank timestamps could be active at the moment, as the current @@ -1573,6 +1588,7 @@ void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe) if (vblank->inmodeset) { spin_lock_irqsave(&dev->vbl_lock, irqflags); dev->vblank_disable_allowed = true; + drm_reset_vblank_timestamp(dev, pipe); spin_unlock_irqrestore(&dev->vbl_lock, irqflags); if (vblank->inmodeset & 0x2)
Changes to drm_update_vblank_count() in Linux 4.4 broke the behaviour of the pre/post modeset functions as the new update code doesn't deal with hw vblank counter resets inbetween calls to drm_vblank_pre_modeset an drm_vblank_post_modeset, as it should. This causes mistreatment of such hw counter resets as counter wraparound, and thereby large forward jumps of the software vblank counter which in turn cause vblank event dispatching and vblank waits to fail/hang --> userspace clients hang. This symptom was reported on radeon-kms to cause a infinite hang of KDE Plasma 5 shell's login procedure, preventing users from logging in. Fix this by detecting when drm_update_vblank_count() is called inside a pre->post modeset interval. If so, clamp valid vblank increments to the safe values 0 and 1, pretty much restoring the update behavior of the old update code of Linux 4.3 and earlier. Also reset the last recorded hw vblank count at call to drm_vblank_post_modeset() to be safe against hw that after modesetting, dpms on etc. only fires its first vblank irq after drm_vblank_post_modeset() was already called. Reported-by: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: <stable@vger.kernel.org> # 4.4+ Cc: michel@daenzer.net Cc: vbabka@suse.cz Cc: ville.syrjala@linux.intel.com Cc: daniel.vetter@ffwll.ch Cc: dri-devel@lists.freedesktop.org Cc: alexander.deucher@amd.com Cc: christian.koenig@amd.com --- drivers/gpu/drm/drm_irq.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)