On 9/23/24 1:03 AM, peter.wang@xxxxxxxxxxxx wrote:
diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c index b5c7bc50a27e..b42079c3d634 100644 --- a/drivers/ufs/core/ufshcd.c +++ b/drivers/ufs/core/ufshcd.c @@ -5404,7 +5404,10 @@ ufshcd_transfer_rsp_status(struct ufs_hba *hba, struct ufshcd_lrb *lrbp, } break; case OCS_ABORTED: - result |= DID_ABORT << 16; + if (hba->quirks & UFSHCD_QUIRK_OCS_ABORTED) + result |= DID_REQUEUE << 16; + else + result |= DID_ABORT << 16; dev_warn(hba->dev, "OCS aborted from controller = %x for tag %d\n", ocs, lrbp->task_tag);
I think the approach of this patch is racy: the cmd->result assignment by ufshcd_transfer_rsp_status() races with the cmd->result assignment by ufshcd_abort_one(). How about addressing this race as follows? * In ufshcd_compl_one_cqe(), if the OCS_ABORTED status is encountered, set a completion. * In the code that aborts SCSI commands, for MediaTek controllers only, wait for that completion (based on a quirk). * Instead of introducing an if-statement in ufshcd_transfer_rsp_status(), rely on the cmd->result value assigned by ufshcd_abort_one(). Thanks, Bart.
