| Message ID | 20161020143952.2538-1-tiwai@suse.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Oct 20, 2016 at 04:39:52PM +0200, Takashi Iwai wrote: > Since 4.7 kernel, we've seen the error messages like > > kernel: [TTM] Buffer eviction failed > kernel: qxl 0000:00:02.0: object_init failed for (4026540032, 0x00000001) > kernel: [drm:qxl_alloc_bo_reserved [qxl]] *ERROR* failed to allocate VRAM BO > > on QXL when switching and accessing on VT. The culprit was the > generic deferred_io code (qxl driver switched to it since 4.7). > There is a race between the dirty clip update and the call of > callback. > > In drm_fb_helper_dirty(), the dirty clip is updated in the spinlock, > while it kicks off the update worker outside the spinlock. Meanwhile > the update worker clears the dirty clip in the spinlock, too. Thus, > when drm_fb_helper_dirty() is called concurrently, schedule_work() is > called after the clip is cleared in the first worker call. > > This patch addresses it by validating the clip before calling the > dirty fb callback. > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98322 > Bugzilla: https://bugzilla.suse.com/show_bug.cgi?id=1003298 > Fixes: eaa434defaca ('drm/fb-helper: Add fb_deferred_io support') > Cc: <stable@vger.kernel.org> > Signed-off-by: Takashi Iwai <tiwai@suse.de> > --- > drivers/gpu/drm/drm_fb_helper.c | 13 +++++++++---- > 1 file changed, 9 insertions(+), 4 deletions(-) > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c > index 03414bde1f15..d790d205129e 100644 > --- a/drivers/gpu/drm/drm_fb_helper.c > +++ b/drivers/gpu/drm/drm_fb_helper.c > @@ -636,15 +636,20 @@ static void drm_fb_helper_dirty_work(struct work_struct *work) > dirty_work); > struct drm_clip_rect *clip = &helper->dirty_clip; > struct drm_clip_rect clip_copy; > + bool dirty; > unsigned long flags; > > spin_lock_irqsave(&helper->dirty_lock, flags); > - clip_copy = *clip; > - clip->x1 = clip->y1 = ~0; > - clip->x2 = clip->y2 = 0; > + dirty = (clip->x1 < clip->x2 && clip->y1 < clip->y2); > + if (dirty) { > + clip_copy = *clip; > + clip->x1 = clip->y1 = ~0; > + clip->x2 = clip->y2 = 0; > + } > spin_unlock_irqrestore(&helper->dirty_lock, flags); > > - helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, &clip_copy, 1); > + if (dirty) Could do it the other way too, ie. just make the copy, and then check the copy (can be done after dropping the lock even). Would avoid having to add the 'dirty' variable. > + helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, &clip_copy, 1); > } > > /** > -- > 2.10.1
On Thu, 20 Oct 2016 16:56:04 +0200, Ville Syrjälä wrote: > > On Thu, Oct 20, 2016 at 04:39:52PM +0200, Takashi Iwai wrote: > > Since 4.7 kernel, we've seen the error messages like > > > > kernel: [TTM] Buffer eviction failed > > kernel: qxl 0000:00:02.0: object_init failed for (4026540032, 0x00000001) > > kernel: [drm:qxl_alloc_bo_reserved [qxl]] *ERROR* failed to allocate VRAM BO > > > > on QXL when switching and accessing on VT. The culprit was the > > generic deferred_io code (qxl driver switched to it since 4.7). > > There is a race between the dirty clip update and the call of > > callback. > > > > In drm_fb_helper_dirty(), the dirty clip is updated in the spinlock, > > while it kicks off the update worker outside the spinlock. Meanwhile > > the update worker clears the dirty clip in the spinlock, too. Thus, > > when drm_fb_helper_dirty() is called concurrently, schedule_work() is > > called after the clip is cleared in the first worker call. > > > > This patch addresses it by validating the clip before calling the > > dirty fb callback. > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98322 > > Bugzilla: https://bugzilla.suse.com/show_bug.cgi?id=1003298 > > Fixes: eaa434defaca ('drm/fb-helper: Add fb_deferred_io support') > > Cc: <stable@vger.kernel.org> > > Signed-off-by: Takashi Iwai <tiwai@suse.de> > > --- > > drivers/gpu/drm/drm_fb_helper.c | 13 +++++++++---- > > 1 file changed, 9 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c > > index 03414bde1f15..d790d205129e 100644 > > --- a/drivers/gpu/drm/drm_fb_helper.c > > +++ b/drivers/gpu/drm/drm_fb_helper.c > > @@ -636,15 +636,20 @@ static void drm_fb_helper_dirty_work(struct work_struct *work) > > dirty_work); > > struct drm_clip_rect *clip = &helper->dirty_clip; > > struct drm_clip_rect clip_copy; > > + bool dirty; > > unsigned long flags; > > > > spin_lock_irqsave(&helper->dirty_lock, flags); > > - clip_copy = *clip; > > - clip->x1 = clip->y1 = ~0; > > - clip->x2 = clip->y2 = 0; > > + dirty = (clip->x1 < clip->x2 && clip->y1 < clip->y2); > > + if (dirty) { > > + clip_copy = *clip; > > + clip->x1 = clip->y1 = ~0; > > + clip->x2 = clip->y2 = 0; > > + } > > spin_unlock_irqrestore(&helper->dirty_lock, flags); > > > > - helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, &clip_copy, 1); > > + if (dirty) > > Could do it the other way too, ie. just make the copy, and then check the > copy (can be done after dropping the lock even). Would avoid having to > add the 'dirty' variable. Sounds good. Let me try... Takashi
On Thu, Oct 20, 2016 at 05:00:35PM +0200, Takashi Iwai wrote: > On Thu, 20 Oct 2016 16:56:04 +0200, > Ville Syrjälä wrote: > > > > On Thu, Oct 20, 2016 at 04:39:52PM +0200, Takashi Iwai wrote: > > > Since 4.7 kernel, we've seen the error messages like > > > > > > kernel: [TTM] Buffer eviction failed > > > kernel: qxl 0000:00:02.0: object_init failed for (4026540032, 0x00000001) > > > kernel: [drm:qxl_alloc_bo_reserved [qxl]] *ERROR* failed to allocate VRAM BO > > > > > > on QXL when switching and accessing on VT. The culprit was the > > > generic deferred_io code (qxl driver switched to it since 4.7). > > > There is a race between the dirty clip update and the call of > > > callback. > > > > > > In drm_fb_helper_dirty(), the dirty clip is updated in the spinlock, > > > while it kicks off the update worker outside the spinlock. Meanwhile > > > the update worker clears the dirty clip in the spinlock, too. Thus, > > > when drm_fb_helper_dirty() is called concurrently, schedule_work() is > > > called after the clip is cleared in the first worker call. > > > > > > This patch addresses it by validating the clip before calling the > > > dirty fb callback. > > > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98322 > > > Bugzilla: https://bugzilla.suse.com/show_bug.cgi?id=1003298 > > > Fixes: eaa434defaca ('drm/fb-helper: Add fb_deferred_io support') > > > Cc: <stable@vger.kernel.org> > > > Signed-off-by: Takashi Iwai <tiwai@suse.de> > > > --- > > > drivers/gpu/drm/drm_fb_helper.c | 13 +++++++++---- > > > 1 file changed, 9 insertions(+), 4 deletions(-) > > > > > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c > > > index 03414bde1f15..d790d205129e 100644 > > > --- a/drivers/gpu/drm/drm_fb_helper.c > > > +++ b/drivers/gpu/drm/drm_fb_helper.c > > > @@ -636,15 +636,20 @@ static void drm_fb_helper_dirty_work(struct work_struct *work) > > > dirty_work); > > > struct drm_clip_rect *clip = &helper->dirty_clip; > > > struct drm_clip_rect clip_copy; > > > + bool dirty; > > > unsigned long flags; > > > > > > spin_lock_irqsave(&helper->dirty_lock, flags); > > > - clip_copy = *clip; > > > - clip->x1 = clip->y1 = ~0; > > > - clip->x2 = clip->y2 = 0; > > > + dirty = (clip->x1 < clip->x2 && clip->y1 < clip->y2); > > > + if (dirty) { > > > + clip_copy = *clip; > > > + clip->x1 = clip->y1 = ~0; > > > + clip->x2 = clip->y2 = 0; > > > + } > > > spin_unlock_irqrestore(&helper->dirty_lock, flags); > > > > > > - helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, &clip_copy, 1); > > > + if (dirty) > > > > Could do it the other way too, ie. just make the copy, and then check the > > copy (can be done after dropping the lock even). Would avoid having to > > add the 'dirty' variable. > > Sounds good. Let me try... Another thing: How do we prevent userspace from doing the same, i.e. submitting an empty rectangle? Do we need to patch up drm_mode_dirtyfb_ioctl too? Not much point if we fix this bug in the fb helpers and leave the barn door wide open for userspace to oops drivers ;-) -Daniel > > > Takashi
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 03414bde1f15..d790d205129e 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -636,15 +636,20 @@ static void drm_fb_helper_dirty_work(struct work_struct *work) dirty_work); struct drm_clip_rect *clip = &helper->dirty_clip; struct drm_clip_rect clip_copy; + bool dirty; unsigned long flags; spin_lock_irqsave(&helper->dirty_lock, flags); - clip_copy = *clip; - clip->x1 = clip->y1 = ~0; - clip->x2 = clip->y2 = 0; + dirty = (clip->x1 < clip->x2 && clip->y1 < clip->y2); + if (dirty) { + clip_copy = *clip; + clip->x1 = clip->y1 = ~0; + clip->x2 = clip->y2 = 0; + } spin_unlock_irqrestore(&helper->dirty_lock, flags); - helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, &clip_copy, 1); + if (dirty) + helper->fb->funcs->dirty(helper->fb, NULL, 0, 0, &clip_copy, 1); } /**
Since 4.7 kernel, we've seen the error messages like kernel: [TTM] Buffer eviction failed kernel: qxl 0000:00:02.0: object_init failed for (4026540032, 0x00000001) kernel: [drm:qxl_alloc_bo_reserved [qxl]] *ERROR* failed to allocate VRAM BO on QXL when switching and accessing on VT. The culprit was the generic deferred_io code (qxl driver switched to it since 4.7). There is a race between the dirty clip update and the call of callback. In drm_fb_helper_dirty(), the dirty clip is updated in the spinlock, while it kicks off the update worker outside the spinlock. Meanwhile the update worker clears the dirty clip in the spinlock, too. Thus, when drm_fb_helper_dirty() is called concurrently, schedule_work() is called after the clip is cleared in the first worker call. This patch addresses it by validating the clip before calling the dirty fb callback. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98322 Bugzilla: https://bugzilla.suse.com/show_bug.cgi?id=1003298 Fixes: eaa434defaca ('drm/fb-helper: Add fb_deferred_io support') Cc: <stable@vger.kernel.org> Signed-off-by: Takashi Iwai <tiwai@suse.de> --- drivers/gpu/drm/drm_fb_helper.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)