On Fri, 10 Dec 2021, Tvrtko Ursulin tvrtko.ursulin@linux.intel.com wrote:
On 09/12/2021 19:14, Daniele Ceraolo Spurio wrote:
On 12/9/2021 10:48 AM, Matthew Brost wrote:
s/ce/cn/ when grabbing guc_state.lock before calling clr_context_registered.
Fixes: 0f7976506de61 ("drm/i915/guc: Rework and simplify locking") Signed-off-by: Matthew Brost matthew.brost@intel.com Cc: stable@vger.kernel.org
I think Cc: stable is not needed here:
$ git tag --contains 0f7976506de61 drm-intel-fixes-2021-11-18 drm-intel-gt-next-2021-10-08 drm-intel-gt-next-2021-10-21 drm-intel-gt-next-2021-11-22 drm-intel-next-2021-10-15 drm-intel-next-fixes-2021-11-09 v5.16-rc1 v5.16-rc2 v5.16-rc3 v5.16-rc4
'dim fixes 0f7976506de61' concurs.
BR, Jani.
So still can hit 5.16 via fixes. Rodrigo, did I get this right and you will be able to pick it up next week or so?
Reviewed-by: Daniele Ceraolo Spurio daniele.ceraolospurio@intel.com
I'm assuming we didn't see any splat from the lockdep assert in clr_context_registered in our CI runs because we never hit this case as it requires 64k+ contexts. Maybe we can add a selftest to purposely exercise this path? Not a blocker for merging this fix.
Was the bug found by inspection or reported?
Given the buggy function is called steal_guc_id, so if the implication is there is no testing for guc id stealing, then it indeed please add some coverage ASAP.
Regards,
Tvrtko
Daniele
drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c index 1f9d4fde421f..9b7b4f4e0d91 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -1937,9 +1937,9 @@ static int steal_guc_id(struct intel_guc *guc, struct intel_context *ce) list_del_init(&cn->guc_id.link); ce->guc_id = cn->guc_id; - spin_lock(&ce->guc_state.lock); + spin_lock(&cn->guc_state.lock); clr_context_registered(cn); - spin_unlock(&ce->guc_state.lock); + spin_unlock(&cn->guc_state.lock); set_context_guc_id_invalid(cn);