We get a lockdep splat when the reset mutex is held, because it can be
taken from fence_wait. This conflicts with the mmu notifier we have,
because we recurse between reset mutex and mmap lock -> mmu notifier.
Remove this recursion by calling revoke_mmaps before taking the lock.
The reset code still needs fixing, as taking mmap locks during reset
is not allowed.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
[danvet: Add FIXME.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20210323155059.628690-64-maarten.lankhorst@linux.intel.com
err = __intel_gt_reset(gt, ALL_ENGINES);
for (i = 0; err && i < RESET_MAX_RETRIES; i++) {
msleep(10 * (i + 1));
err = __intel_gt_reset(gt, ALL_ENGINES);
for (i = 0; err && i < RESET_MAX_RETRIES; i++) {
msleep(10 * (i + 1));
might_sleep();
GEM_BUG_ON(!test_bit(I915_RESET_BACKOFF, >->reset.flags));
might_sleep();
GEM_BUG_ON(!test_bit(I915_RESET_BACKOFF, >->reset.flags));
+
+ /*
+ * FIXME: Revoking cpu mmap ptes cannot be done from a dma_fence
+ * critical section like gpu reset.
+ */
+ gt_revoke(gt);
+
mutex_lock(>->reset.mutex);
/* Clear any previous failed attempts at recovery. Time to try again. */
mutex_lock(>->reset.mutex);
/* Clear any previous failed attempts at recovery. Time to try again. */