Doug Ledford wrote:
> >
> > is the driver re-entrant such that one thread could be servicing the NQ
> >
> > device while other threads would be putting queued commands on the bus
> >
> > for other queueing devices?
>
> Heh, this isn't NT here....there are no threads in that driver and we wouldn't
> be so insane as to use a thread per device or thread per command.
---
Not like NT? Can't tell if youi meant it as a good or a bad
thing -- but I wasn't thinking of a per-device or command thread, but
on a 256P processor, I could see multiple kernel threads all potentially
writing to SCSI devices simultaneously. There may be some point at which
those requests (not the entire transaction) must be serialized to go out
on the bus, but hopefully that serialized portion would be next to
nill. Of course to match the 256P system I'd assume an I/O infrastruction
that can support multi-gigabyte sustained transfer. Seems like
re-entrant SCSI code would be a big bonus, or a bottleneck if not there,
no? Or is the bottleneck going to be elsewhere?
-linda
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
