On Thu, Jan 07, 2016 at 03:11:38PM +0100, Bart Van Assche wrote: > Sorry but I do not agree that checking CMD_T_ABORTED without locking is > wrong. The worst that can happen if CMD_T_ABORTED is checked without > locking is that if this test happens within a few nanoseconds after the > CMD_T_ABORTED flag has been set that target_complete_cmd() proceeds with > executing the command instead of aborting it. This behavior is compliant > with the SCSI specs and would also occur if sending the ABORT command from > the initiator to the target is delayed slightly. The Linux kernel contains > many optimizations of this kind, namely not using locking to check a flag > when it is safe to do leave out locking. Yes. The only case where we need to take a lock on the read side is if we need to test a specific combination of two flags, and even that is dubious if they are in the same cache line. -- 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
