On Tue, 2011-12-20 at 10:13 +0000, Bart Van Assche wrote: > On Mon, Dec 19, 2011 at 10:32 PM, David Dillow <[email protected]> wrote: > > I still think this is already solved in user space, but the new > > reconnect model you've implemented doesn't match up with the expected > > semantics. It'd be better to match the rest of the SCSI stack for this. > > Sorry, but I'm afraid there is no "rest of the SCSI stack > [semantics]". I've been following the semantics implemented in > open-iscsi. If the semantics implemented for FC differ of those > proposed in this patch set, then that means that the semantics > implemented for FC differ of those for open-iscsi.
FC, SAS, SPI, (current) SRP, and FCoE currently use the same semantic to check if the transport is still alive -- user space initiates activity, usually via multipathd. You are changing SRP to match the iSCSI stack, and because SRP does not have a NOP command, you have to fake one up using Test Unit Ready to the first LUN to come through slave_alloc(). You kill the entire connection if that LUN has issues and causes a timeout, though the rest of the LUNs may be fine and able to communicate. This is all handled properly by existing user space code; there's no need for it to be in kernel, especially if you have to fake it while going against the grain. -- Dave Dillow National Center for Computational Science Oak Ridge National Laboratory (865) 241-6602 office -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
