This patch reduces the BUG_ON when pin_count reaches max to a warning.
The reason we want to just warn is because the unpin_work_fn work
function can get starved out during soft lockup. When the unpin worker
is starved, the pin_count rises, and the gem BUG_ON races the soft
lockup crash reporter.
The original intent of this BUG was to catch overflow cases when all of
the small bitfields in drm_i915_gem_object were combined into one
bitfield (commit 778c35444). Since pin leaks aren't an issue in steady
state, it should be fine to downgrade the error.
Signed-off-by: Sean Paul <seanp...@chromium.org>
---
drivers/gpu/drm/i915/i915_gem.c | 5 ++++-
1 files changed, 4 insertions(+), 1 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 6d2180c..65ff55d 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3016,7 +3016,10 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj,
{
int ret;
- BUG_ON(obj->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT);
+ if (obj->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT) {
+ WARN(obj->pin_count, "pin count has reached its max\n");
+ return obj->pin_count;
+ }
if (obj->gtt_space != NULL) {
if ((alignment && obj->gtt_offset & (alignment - 1)) ||
--
1.7.7.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
