From: Chris Wilson <[email protected]>
Avoid trying to invalidate the TLB in the middle of performing an
engine reset, as this may result in the reset timing out. Currently,
the TLB invalidate is only serialised by its own mutex, forgoing the
uncore lock, but we can take the uncore->lock as well to serialise
the mmio access, thereby serialising with the GDRST.
Tested on a NUC5i7RYB, BIOS RYBDWi35.86A.0380.2019.0517.1530 with
i915 selftest/hangcheck.
Cc: [email protected]
Fixes: 7938d61591d3 ("drm/i915: Flush TLBs before releasing backing store")
Reported-by: Mauro Carvalho Chehab <[email protected]>
Tested-by: Mauro Carvalho Chehab <[email protected]>
Reviewed-by: Mauro Carvalho Chehab <[email protected]>
Signed-off-by: Chris Wilson <[email protected]>
Cc: Tvrtko Ursulin <[email protected]>
Acked-by: Thomas Hellström <[email protected]>
Reviewed-by: Andi Shyti <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
---
See [PATCH v2 0/3] at: https://lore.kernel.org/all/[email protected]/
drivers/gpu/drm/i915/gt/intel_gt.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
index 30c60cd960e8..7e57a90b4095 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt.c
@@ -952,6 +952,8 @@ void intel_gt_invalidate_tlbs(struct intel_gt *gt)
mutex_lock(>->tlb_invalidate_lock);
intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL);
+ spin_lock_irq(&uncore->lock); /* seralise invalidate with GT reset */
+
awake = 0;
for_each_engine(engine, gt, id) {
struct reg_and_bit rb;
@@ -967,6 +969,8 @@ void intel_gt_invalidate_tlbs(struct intel_gt *gt)
awake |= engine->mask;
}
+ spin_unlock_irq(&uncore->lock);
+
for_each_engine_masked(engine, gt, awake, tmp) {
struct reg_and_bit rb;
--
2.36.1