On Wed, 2018-08-08 at 01:44 +-0800, Ming Lei wrote: +AD4- --- a/drivers/scsi/scsi+AF8-lib.c +AD4- +-+-+- b/drivers/scsi/scsi+AF8-lib.c +AD4- +AEAAQA- -278,16 +-278,16 +AEAAQA- int +AF8AXw-scsi+AF8-execute(struct scsi+AF8-device +ACo-sdev, const unsigned char +ACo-cmd, +AD4- struct request +ACo-req+ADs- +AD4- struct scsi+AF8-request +ACo-rq+ADs- +AD4- int ret +AD0- DRIVER+AF8-ERROR +ADwAPA- 24+ADs- +AD4- +- struct request+AF8-queue +ACo-q +AD0- sdev-+AD4-host-+AD4-admin+AF8-q+ADs- +AD4- +AD4- - req +AD0- blk+AF8-get+AF8-request(sdev-+AD4-request+AF8-queue, +AD4- +- req +AD0- blk+AF8-get+AF8-request(q, +AD4- data+AF8-direction +AD0APQ- DMA+AF8-TO+AF8-DEVICE ? +AD4- REQ+AF8-OP+AF8-SCSI+AF8-OUT : REQ+AF8-OP+AF8-SCSI+AF8-IN, BLK+AF8-MQ+AF8-REQ+AF8-PREEMPT)+ADs- The above looks weird to me. Why are all RQF+AF8-PREEMPT requests sent to the admin queue instead of only RQF+AF8-PM requests? +AD4- +AEAAQA- -299,6 +-299,8 +AEAAQA- int +AF8AXw-scsi+AF8-execute(struct scsi+AF8-device +ACo-sdev, const unsigned char +ACo-cmd, +AD4- req-+AD4-cmd+AF8-flags +AHwAPQ- flags+ADs- +AD4- req-+AD4-rq+AF8-flags +AHwAPQ- rq+AF8-flags +AHw- RQF+AF8-QUIET+ADs- +AD4- +AD4- +- atomic+AF8-inc(+ACY-sdev-+AD4-nr+AF8-admin+AF8-pending)+ADs- Why has a new counter been introduced to keep track of admin requests instead of using q+AF8-usage+AF8-counter? Thanks, Bart.
