On 6/23/21 9:16 PM, Can Guo wrote: > On 2021-06-24 05:33, Bart Van Assche wrote: >> On 6/23/21 12:35 AM, Can Guo wrote: >>> @@ -2737,7 +2737,7 @@ static int ufshcd_queuecommand(struct Scsi_Host >>> *host, struct scsi_cmnd *cmd) >>> * err handler blocked for too long. So, just fail the scsi cmd >>> * sent from PM ops, err handler can recover PM error anyways. >>> */ >>> - if (hba->wlu_pm_op_in_progress) { >>> + if (cmd->request->rq_flags & RQF_PM) { >>> hba->force_reset = true; >>> set_host_byte(cmd, DID_BAD_TARGET); >>> cmd->scsi_done(cmd); >> >> I'm still concerned that the above code may trigger data corruption. I >> prefer that the above code is removed instead of being modified. > > Removing the change will lead to deadlock when error handling prepare > calls pm_runtime_get_sync(). > > RQF_PM is only given to requests sent from power management operations, > during which the specific device/LU is suspending/resuming, meaning no > data transaction is ongoing. How can fast failing a PM request trigger > data corruption? Right, the above code only affects power management requests so there is no risk for data corruption. Thanks, Bart.