On Thu, 2017-02-09 at 03:28 -0800, Nicholas A. Bellinger wrote: > As-is, this patch breaks active I/O configfs shutdown of > /sys/kernel/config/target/$FABRIC/$WWN/$TPGT/lun/$LUN_ID/$DEV_SYMLINK. > > The reason why cmd->se_lun->lun_ref is dropped, and se_cmd->se_lun is > cleared before the response hand-off back to the fabric driver is so > that ../$WWN/$TPGT/lun/$LUN_ID/ shutdown can occur from configfs, > without having to wait for se_cmd->cmd_kref to reach zero. > > Depending on initiator fabric ACKs coming back to drop the last > se_cmd->cmd_kref in order to release se_lun->lun_ref once user has > explicitly requested ../$WWN/$TPGT/lun/$LUN_ID doesn't work in practice > for a scale-out distributed use-case. > > This must be able to happen asychronously from the fabric ACK. There can only be a delay between the last kref_put() call from the target driver and the final kref_put() if a task management function is being processed or a session is being shut down. The time needed to process any of these actions is finite so shutdown of a target driver while I/O is in progress still works. Bart.-- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
