From 6bd9e877ce5322baf84c991c4509b49cf39a9b57 Mon Sep 17 00:00:00 2001 From: Vitaly Prosyak Date: Tue, 20 Oct 2015 15:02:03 -0400 Subject: drm/amdgpu: Move MMIO flip out of spinlocked region Prior actual MMIO flip we need to acquire DAL mutex to guard our target state which get modified on reset mode. Assign page flip status before actual flip to handle the possible race condition with interrupt. Signed-off-by: Vitaly Prosyak Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_display.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index acd066d0a805..531627b0a900 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -118,12 +118,12 @@ static void amdgpu_flip_work_func(struct work_struct *__work) spin_lock_irqsave(&crtc->dev->event_lock, flags); }; - /* do the flip (mmio) */ - adev->mode_info.funcs->page_flip(adev, work->crtc_id, work->base); /* set the flip status */ amdgpuCrtc->pflip_status = AMDGPU_FLIP_SUBMITTED; - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); + + /* Do the flip (mmio) */ + adev->mode_info.funcs->page_flip(adev, work->crtc_id, work->base); } /* -- cgit v1.2.3 From 6800e2ea2e74635e5386fc4bb43b0b69edd85a19 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 11 Feb 2016 14:51:47 +0100 Subject: drm/amdgpu: stop calling amdgpu_gpu_reset from the flip code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We don't return -EDEADLK any more. Signed-off-by: Christian König Acked-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_display.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 531627b0a900..1d6ef1a95e74 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -38,20 +38,12 @@ static void amdgpu_flip_wait_fence(struct amdgpu_device *adev, struct fence **f) { - struct amdgpu_fence *fence; long r; if (*f == NULL) return; - fence = to_amdgpu_fence(*f); - if (fence) { - r = fence_wait(&fence->base, false); - if (r == -EDEADLK) - r = amdgpu_gpu_reset(adev); - } else - r = fence_wait(*f, false); - + r = fence_wait(*f, false); if (r) DRM_ERROR("failed to wait on page flip fence (%ld)!\n", r); -- cgit v1.2.3 From c3874b758ad52708ffdbab3e7c58e70167ce5024 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 11 Feb 2016 15:48:30 +0100 Subject: drm/amdgpu: stop blocking for page filp fences MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just register an callback and reschedule the work item if necessary. Signed-off-by: Christian König Acked-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 43 ++++++++++++++++++----------- 1 file changed, 27 insertions(+), 16 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_display.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 1d6ef1a95e74..edf91593ac87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -35,24 +35,32 @@ #include #include -static void amdgpu_flip_wait_fence(struct amdgpu_device *adev, - struct fence **f) +static void amdgpu_flip_callback(struct fence *f, struct fence_cb *cb) { - long r; + struct amdgpu_flip_work *work = + container_of(cb, struct amdgpu_flip_work, cb); + struct amdgpu_device *adev = work->adev; + struct amdgpu_crtc *amdgpu_crtc = adev->mode_info.crtcs[work->crtc_id]; - if (*f == NULL) - return; + fence_put(f); + queue_work(amdgpu_crtc->pflip_queue, &work->flip_work); +} - r = fence_wait(*f, false); - if (r) - DRM_ERROR("failed to wait on page flip fence (%ld)!\n", r); +static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work, + struct fence **f) +{ + struct fence *fence= *f; + + if (fence == NULL) + return false; - /* We continue with the page flip even if we failed to wait on - * the fence, otherwise the DRM core and userspace will be - * confused about which BO the CRTC is scanning out - */ - fence_put(*f); *f = NULL; + + if (!fence_add_callback(fence, &work->cb, amdgpu_flip_callback)) + return true; + + fence_put(*f); + return false; } static void amdgpu_flip_work_func(struct work_struct *__work) @@ -68,9 +76,12 @@ static void amdgpu_flip_work_func(struct work_struct *__work) int vpos, hpos, stat, min_udelay; struct drm_vblank_crtc *vblank = &crtc->dev->vblank[work->crtc_id]; - amdgpu_flip_wait_fence(adev, &work->excl); + if (amdgpu_flip_handle_fence(work, &work->excl)) + return; + for (i = 0; i < work->shared_count; ++i) - amdgpu_flip_wait_fence(adev, &work->shared[i]); + if (amdgpu_flip_handle_fence(work, &work->shared[i])) + return; /* We borrow the event spin lock for protecting flip_status */ spin_lock_irqsave(&crtc->dev->event_lock, flags); @@ -234,7 +245,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, /* update crtc fb */ crtc->primary->fb = fb; spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - queue_work(amdgpu_crtc->pflip_queue, &work->flip_work); + amdgpu_flip_work_func(&work->flip_work); return 0; vblank_cleanup: -- cgit v1.2.3 From 87d58c11ec17234c4636a588466260a95fec59cb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 11 Feb 2016 17:31:37 +0100 Subject: drm/amdgpu: remove page flip work queue v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just use the system queue now that we don't block any more. v2: handle DAL as well. v3: agd: split DAL changes out Signed-off-by: Christian König Reviewed-by: Mykola Lysenko Acked-by: Alex Deucher (v1) --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_display.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index edf91593ac87..2cb53c24dec0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -39,11 +39,9 @@ static void amdgpu_flip_callback(struct fence *f, struct fence_cb *cb) { struct amdgpu_flip_work *work = container_of(cb, struct amdgpu_flip_work, cb); - struct amdgpu_device *adev = work->adev; - struct amdgpu_crtc *amdgpu_crtc = adev->mode_info.crtcs[work->crtc_id]; fence_put(f); - queue_work(amdgpu_crtc->pflip_queue, &work->flip_work); + schedule_work(&work->flip_work); } static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work, -- cgit v1.2.3 From e1d09dc0ccc6c91e3916476f636edb76da1f65bb Mon Sep 17 00:00:00 2001 From: Mario Kleiner Date: Fri, 19 Feb 2016 02:06:39 +0100 Subject: drm/amdgpu: Don't hang in amdgpu_flip_work_func on disabled crtc. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This fixes a regression introduced in Linux 4.4. This is a port of the same fix for radeon-kms in the patch "drm/radeon: Don't hang in radeon_flip_work_func on disabled crtc. (v2)" Limit the amount of time amdgpu_flip_work_func can delay programming a page flip, by both limiting the maximum amount of time per wait cycle and the maximum number of wait cycles. Continue the flip if the limit is exceeded, even if that may result in a visual or timing glitch. This is to prevent a hang of page flips, as reported in fdo bug #93746: Disconnecting a DisplayPort display in parallel to a kms pageflip getting queued can cause the following hang of page flips and thereby an unusable desktop: 1. kms pageflip ioctl() queues pageflip -> queues execution of amdgpu_flip_work_func. 2. Hotunplug of display causes the driver to DPMS OFF the unplugged display. Display engine shuts down, scanout no longer moves, but stays at its resting position at start line of vblank. 3. amdgpu_flip_work_func executes while crtc is off, and due to the non-moving scanout position, the new flip delay code introduced into Linux 4.4 by commit 8e36f9d33c13 ("drm/amdgpu: Fixup hw vblank counter/ts..") enters an infinite wait loop. 4. After reconnecting the display, the pageflip continues to hang in 3. and the display doesn't update its view of the desktop. This patch fixes the Linux 4.4 regression from fdo bug #93746 Reported-by: Bernd Steinhauser Signed-off-by: Mario Kleiner Cc: # 4.4+ Cc: Michel Dänzer Cc: Alex Deucher Reviewed-by: Michel Dänzer Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_display.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index acd066d0a805..8297bc319369 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -72,8 +72,8 @@ static void amdgpu_flip_work_func(struct work_struct *__work) struct drm_crtc *crtc = &amdgpuCrtc->base; unsigned long flags; - unsigned i; - int vpos, hpos, stat, min_udelay; + unsigned i, repcnt = 4; + int vpos, hpos, stat, min_udelay = 0; struct drm_vblank_crtc *vblank = &crtc->dev->vblank[work->crtc_id]; amdgpu_flip_wait_fence(adev, &work->excl); @@ -96,7 +96,7 @@ static void amdgpu_flip_work_func(struct work_struct *__work) * In practice this won't execute very often unless on very fast * machines because the time window for this to happen is very small. */ - for (;;) { + while (amdgpuCrtc->enabled && repcnt--) { /* GET_DISTANCE_TO_VBLANKSTART returns distance to real vblank * start in hpos, and to the "fudged earlier" vblank start in * vpos. @@ -114,10 +114,22 @@ static void amdgpu_flip_work_func(struct work_struct *__work) /* Sleep at least until estimated real start of hw vblank */ spin_unlock_irqrestore(&crtc->dev->event_lock, flags); min_udelay = (-hpos + 1) * max(vblank->linedur_ns / 1000, 5); + if (min_udelay > vblank->framedur_ns / 2000) { + /* Don't wait ridiculously long - something is wrong */ + repcnt = 0; + break; + } usleep_range(min_udelay, 2 * min_udelay); spin_lock_irqsave(&crtc->dev->event_lock, flags); }; + if (!repcnt) + DRM_DEBUG_DRIVER("Delay problem on crtc %d: min_udelay %d, " + "framedur %d, linedur %d, stat %d, vpos %d, " + "hpos %d\n", work->crtc_id, min_udelay, + vblank->framedur_ns / 1000, + vblank->linedur_ns / 1000, stat, vpos, hpos); + /* do the flip (mmio) */ adev->mode_info.funcs->page_flip(adev, work->crtc_id, work->base); /* set the flip status */ -- cgit v1.2.3 From 90e94b160c7f647ddffda707f5e3c0c66c170df8 Mon Sep 17 00:00:00 2001 From: Mario Kleiner Date: Tue, 1 Mar 2016 21:31:16 +0100 Subject: drm/amdgpu: Fix error handling in amdgpu_flip_work_func. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The patch e1d09dc0ccc6: "drm/amdgpu: Don't hang in amdgpu_flip_work_func on disabled crtc." from Feb 19, 2016, leads to the following static checker warning, as reported by Dan Carpenter in https://lists.freedesktop.org/archives/dri-devel/2016-February/101987.html drivers/gpu/drm/amd/amdgpu/amdgpu_display.c:127 amdgpu_flip_work_func() warn: should this be 'repcnt == -1' drivers/gpu/drm/amd/amdgpu/amdgpu_display.c:136 amdgpu_flip_work_func() error: double unlock 'spin_lock:&crtc->dev->event_lock' drivers/gpu/drm/amd/amdgpu/amdgpu_display.c:136 amdgpu_flip_work_func() error: double unlock 'irqsave:flags' This patch fixes both reported problems: Change post-decrement of repcnt to pre-decrement, so it can't underflow anymore, but still performs up to three repetitions - three is the maximum one could expect in practice. Move the spin_unlock_irqrestore to where it actually belongs. Reviewed-by: Michel Dänzer Reported-by: Dan Carpenter Signed-off-by: Mario Kleiner Cc: # 4.4+ Cc: Michel Dänzer Cc: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_display.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 8297bc319369..1846d65b7285 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -96,7 +96,7 @@ static void amdgpu_flip_work_func(struct work_struct *__work) * In practice this won't execute very often unless on very fast * machines because the time window for this to happen is very small. */ - while (amdgpuCrtc->enabled && repcnt--) { + while (amdgpuCrtc->enabled && --repcnt) { /* GET_DISTANCE_TO_VBLANKSTART returns distance to real vblank * start in hpos, and to the "fudged earlier" vblank start in * vpos. @@ -112,13 +112,13 @@ static void amdgpu_flip_work_func(struct work_struct *__work) break; /* Sleep at least until estimated real start of hw vblank */ - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); min_udelay = (-hpos + 1) * max(vblank->linedur_ns / 1000, 5); if (min_udelay > vblank->framedur_ns / 2000) { /* Don't wait ridiculously long - something is wrong */ repcnt = 0; break; } + spin_unlock_irqrestore(&crtc->dev->event_lock, flags); usleep_range(min_udelay, 2 * min_udelay); spin_lock_irqsave(&crtc->dev->event_lock, flags); }; -- cgit v1.2.3