drm/i915: Close race between processing unpin task and queueing the flip

Before queuing the flip but crucially after attaching the unpin-work to
the crtc, we continue to setup the unpin-work. However, should the
hardware fire early, we see the connected unpin-work and queue the task.
The task then promptly runs and unpins the fb before we finish taking
the required references or even pinning it... Havoc.

To close the race, we use the flip-pending atomic to indicate when the
flip is finally setup and enqueued. So during the flip-done processing,
we can check more accurately whether the flip was expected.

v2: Add the appropriate mb() to ensure that the writes to the page-flip
worker are complete prior to marking it active and emitting the MI_FLIP.
On the read side, the mb should be enforced by the spinlocks.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
[danvet: Review the barriers a bit, we need a write barrier both
before and after updating ->pending. Similarly we need a read barrier
in the interrupt handler both before and after reading ->pending. With
well-ordered irqs only one barrier in each place should be required,
but since this patch explicitly sets out to combat spurious interrupts
with is staged activation of the unpin work we need to go full-bore on
the barriers, too. Discussed with Chris Wilson on irc and changes
acked by him.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 8afc0dd7de649a4daec0ca58e58d1c183a52bb3f..e6a11ca85eafa913f97d315493abd995fd0a764c 100644 (file)
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -317,7 +317,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)
                        seq_printf(m, "No flip due on pipe %c (plane %c)\n",
                                   pipe, plane);
                } else {
-                       if (!work->pending) {
+                       if (atomic_read(&work->pending) < INTEL_FLIP_COMPLETE) {
                                seq_printf(m, "Flip queued on pipe %c (plane %c)\n",
                                           pipe, plane);
                        } else {
@@ -328,7 +328,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)
                                seq_printf(m, "Stall check enabled, ");
                        else
                                seq_printf(m, "Stall check waiting for page flip ioctl, ");
-                       seq_printf(m, "%d prepares\n", work->pending);
+                       seq_printf(m, "%d prepares\n", atomic_read(&work->pending));
 
                        if (work->old_fb_obj) {
                                struct drm_i915_gem_object *obj = work->old_fb_obj;

diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 6cd3dc94b47236f2033a637cfc2fd5762453a49a..a4dc97f8b9f0f7d08fd0c2525e039d9893d76735 100644 (file)
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -1466,7 +1466,9 @@ static void i915_pageflip_stall_check(struct drm_device *dev, int pipe)
        spin_lock_irqsave(&dev->event_lock, flags);
        work = intel_crtc->unpin_work;
 
-       if (work == NULL || work->pending || !work->enable_stall_check) {
+       if (work == NULL ||
+           atomic_read(&work->pending) >= INTEL_FLIP_COMPLETE ||
+           !work->enable_stall_check) {
                /* Either the pending flip IRQ arrived, or we're too early. Don't check */
                spin_unlock_irqrestore(&dev->event_lock, flags);
                return;

diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 78d12c471472f17c50f257e15a107180002ec851..f8ee3d16109ce250c7d740a90471523465ad1f9b 100644 (file)
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -6929,11 +6929,18 @@ static void do_intel_finish_page_flip(struct drm_device *dev,
 
        spin_lock_irqsave(&dev->event_lock, flags);
        work = intel_crtc->unpin_work;
-       if (work == NULL || !work->pending) {
+
+       /* Ensure we don't miss a work->pending update ... */
+       smp_rmb();
+
+       if (work == NULL || atomic_read(&work->pending) < INTEL_FLIP_COMPLETE) {
                spin_unlock_irqrestore(&dev->event_lock, flags);
                return;
        }
 
+       /* and that the unpin work is consistent wrt ->pending. */
+       smp_rmb();
+
        intel_crtc->unpin_work = NULL;
 
        if (work->event)
@@ -6977,16 +6984,25 @@ void intel_prepare_page_flip(struct drm_device *dev, int plane)
                to_intel_crtc(dev_priv->plane_to_crtc_mapping[plane]);
        unsigned long flags;
 
+       /* NB: An MMIO update of the plane base pointer will also
+        * generate a page-flip completion irq, i.e. every modeset
+        * is also accompanied by a spurious intel_prepare_page_flip().
+        */
        spin_lock_irqsave(&dev->event_lock, flags);
-       if (intel_crtc->unpin_work) {
-               if ((++intel_crtc->unpin_work->pending) > 1)
-                       DRM_ERROR("Prepared flip multiple times\n");
-       } else {
-               DRM_DEBUG_DRIVER("preparing flip with no unpin work?\n");
-       }
+       if (intel_crtc->unpin_work)
+               atomic_inc_not_zero(&intel_crtc->unpin_work->pending);
        spin_unlock_irqrestore(&dev->event_lock, flags);
 }
 
+inline static void intel_mark_page_flip_active(struct intel_crtc *intel_crtc)
+{
+       /* Ensure that the work item is consistent when activating it ... */
+       smp_wmb();
+       atomic_set(&intel_crtc->unpin_work->pending, INTEL_FLIP_PENDING);
+       /* and that it is marked active as soon as the irq could fire. */
+       smp_wmb();
+}
+
 static int intel_gen2_queue_flip(struct drm_device *dev,
                                 struct drm_crtc *crtc,
                                 struct drm_framebuffer *fb,
@@ -7020,6 +7036,8 @@ static int intel_gen2_queue_flip(struct drm_device *dev,
        intel_ring_emit(ring, fb->pitches[0]);
        intel_ring_emit(ring, obj->gtt_offset + intel_crtc->dspaddr_offset);
        intel_ring_emit(ring, 0); /* aux display base address, unused */
+
+       intel_mark_page_flip_active(intel_crtc);
        intel_ring_advance(ring);
        return 0;
 
@@ -7060,6 +7078,7 @@ static int intel_gen3_queue_flip(struct drm_device *dev,
        intel_ring_emit(ring, obj->gtt_offset + intel_crtc->dspaddr_offset);
        intel_ring_emit(ring, MI_NOOP);
 
+       intel_mark_page_flip_active(intel_crtc);
        intel_ring_advance(ring);
        return 0;
 
@@ -7106,6 +7125,8 @@ static int intel_gen4_queue_flip(struct drm_device *dev,
        pf = 0;
        pipesrc = I915_READ(PIPESRC(intel_crtc->pipe)) & 0x0fff0fff;
        intel_ring_emit(ring, pf | pipesrc);
+
+       intel_mark_page_flip_active(intel_crtc);
        intel_ring_advance(ring);
        return 0;
 
@@ -7148,6 +7169,8 @@ static int intel_gen6_queue_flip(struct drm_device *dev,
        pf = 0;
        pipesrc = I915_READ(PIPESRC(intel_crtc->pipe)) & 0x0fff0fff;
        intel_ring_emit(ring, pf | pipesrc);
+
+       intel_mark_page_flip_active(intel_crtc);
        intel_ring_advance(ring);
        return 0;
 
@@ -7202,6 +7225,8 @@ static int intel_gen7_queue_flip(struct drm_device *dev,
        intel_ring_emit(ring, (fb->pitches[0] | obj->tiling_mode));
        intel_ring_emit(ring, obj->gtt_offset + intel_crtc->dspaddr_offset);
        intel_ring_emit(ring, (MI_NOOP));
+
+       intel_mark_page_flip_active(intel_crtc);
        intel_ring_advance(ring);
        return 0;
 

diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 522061ca06856bc01aa8072ed15ac48209075da5..3915ca9abd414aec7d821134eb4c1240d863b392 100644 (file)
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -401,7 +401,10 @@ struct intel_unpin_work {
        struct drm_i915_gem_object *old_fb_obj;
        struct drm_i915_gem_object *pending_flip_obj;
        struct drm_pending_vblank_event *event;
-       int pending;
+       atomic_t pending;
+#define INTEL_FLIP_INACTIVE    0
+#define INTEL_FLIP_PENDING     1
+#define INTEL_FLIP_COMPLETE    2
        bool enable_stall_check;
 };