On Thu, 2017-12-21 at 17:59 +0000, Bart Van Assche wrote: > On Thu, 2017-12-21 at 09:33 -0800, Raghava Aditya Renukunta wrote: > > +static void aac_wait_for_io_completion(struct aac_dev *aac) > > +{ > > + unsigned long flagv = 0; > > + int i = 0; > > + > > + for (i = 60; i; --i) { > > + struct scsi_device *dev; > > + struct scsi_cmnd *command; > > + int active = 0; > > + > > + __shost_for_each_device(dev, aac->scsi_host_ptr) { > > + spin_lock_irqsave(&dev->list_lock, flagv); > > + list_for_each_entry(command, &dev->cmd_list, list) { > > + if (command->SCp.phase == AAC_OWNER_FIRMWARE) { > > + active++; > > + break; > > + } > > + } > > + spin_unlock_irqrestore(&dev->list_lock, flagv); > > + if (active) > > + break; > > + > > + } > > + /* > > + * We can exit If all the commands are complete > > + */ > > + if (active == 0) > > + break; > > + ssleep(1); > > + } > > +} > > Have you considered to call scsi_target_block() and scsi_target_unblock() > instead > of implementing functionality like the above in a SCSI LLD?
(replying to my own e-mail) It seems like I misread your code - calling scsi_target_block() and scsi_target_unblock() would not be sufficient. But calling blk_mq_freeze_queue() and blk_mq_unfreeze_queue() should be sufficient. The following commit made these functions work not only for scsi-mq but also for legacy scsi queues: commit 055f6e18e08f ("block: Make q_usage_counter also track legacy requests"). Bart.