drm/i915: Park the breadcrumbs signaler across a GPU reset
authorChris Wilson <chris@chris-wilson.co.uk>
Sun, 12 Feb 2017 17:20:01 +0000 (17:20 +0000)
committerChris Wilson <chris@chris-wilson.co.uk>
Mon, 13 Feb 2017 11:18:28 +0000 (11:18 +0000)
The signal threads may be running concurrently with the GPU reset. The
completion from the GPU run asynchronous with the reset and two threads
may see different snapshots of the state, and the signaler may mark a
request as complete as we try to reset it. We don't tolerate 2 different
views of the same state and complain if we try to mark a request as
failed if it is already complete. Disable the signal threads during
reset to prevent this conflict (even though the conflict implies that
the state we resetting to is invalid, we have already made our
decision!).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99733
References: https://bugs.freedesktop.org/show_bug.cgi?id=99671
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212172002.23072-4-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/intel_breadcrumbs.c

index 3c2c2296c1dcf3c867f74a95a91dbd7fac7c32e5..730804ce96106c21a8fe138f8268694403363adb 100644 (file)
@@ -35,6 +35,7 @@
 #include "intel_frontbuffer.h"
 #include "intel_mocs.h"
 #include <linux/dma-fence-array.h>
+#include <linux/kthread.h>
 #include <linux/reservation.h>
 #include <linux/shmem_fs.h>
 #include <linux/slab.h>
@@ -2643,6 +2644,17 @@ int i915_gem_reset_prepare(struct drm_i915_private *dev_priv)
        for_each_engine(engine, dev_priv, id) {
                struct drm_i915_gem_request *request;
 
+               /* Prevent the signaler thread from updating the request
+                * state (by calling dma_fence_signal) as we are processing
+                * the reset. The write from the GPU of the seqno is
+                * asynchronous and the signaler thread may see a different
+                * value to us and declare the request complete, even though
+                * the reset routine have picked that request as the active
+                * (incomplete) request. This conflict is not handled
+                * gracefully!
+                */
+               kthread_park(engine->breadcrumbs.signaler);
+
                /* Prevent request submission to the hardware until we have
                 * completed the reset in i915_gem_reset_finish(). If a request
                 * is completed by one engine, it may then queue a request
@@ -2796,8 +2808,10 @@ void i915_gem_reset_finish(struct drm_i915_private *dev_priv)
 
        lockdep_assert_held(&dev_priv->drm.struct_mutex);
 
-       for_each_engine(engine, dev_priv, id)
+       for_each_engine(engine, dev_priv, id) {
                tasklet_enable(&engine->irq_tasklet);
+               kthread_unpark(engine->breadcrumbs.signaler);
+       }
 }
 
 static void nop_submit_request(struct drm_i915_gem_request *request)
index 9fd002bcebb692e4028a596598fd313360d52ea2..f5e05343110af3e3b7d1384d20b3587321169ec1 100644 (file)
@@ -490,6 +490,9 @@ static int intel_breadcrumbs_signaler(void *arg)
                                break;
 
                        schedule();
+
+                       if (kthread_should_park())
+                               kthread_parkme();
                }
        } while (1);
        __set_current_state(TASK_RUNNING);