Re: [PATCH] NCR53C9x ESP: C99 designated initializers

From: Doug Ledford (
Date: Mon Nov 11 2002 - 15:31:37 EST

On Mon, Nov 11, 2002 at 09:24:32AM -0800, Linus Torvalds wrote:
> On 11 Nov 2002, Alan Cox wrote:
> >
> > The stupid thing is we take the lock then call the eh function then drop
> > it. You can drop the lock, wait and retake it. I need to fix a couple of
> > other drivers to do a proper wait and in much the same way.
> Hmm.. I wonder if the thing should disable the queue (plug it)

It does this already

> and release
> the lock before calling reset.

We call into the driver with the lock held for the sake of consistency.
When you get right down to it, you can write a driver who has no more
concept of SMP locking than spin_lock_irq(host->host_lock); in the isr
routine. They need to know to drop the lock before calling scsi_sleep()
in the eh routines, or they can drop the lock and sleep on some wait queue
as well if they wish. But, in general, in depth locking knowledge is not
required. Since I've ported the old aic7xxx driver to the new scheme, it
is for certain an improvement over the old method in terms of ease of
programming a driver to the API.

> I assume we don't want any new requests at
> this point anyway, and having the low-level drivers know about stopping
> the queue etc sounds like a bad idea..

Right, both already handled.

> Of course, I suspect that it is potentially a bad idea to have the reset
> functionality at the SCSI level _at_all_.

It saves a lot of code duplication. There's basically only two existing
reset strategies around. A) I'm a stupid card with a SCSI bus and all
SCSI busses and plain SCSI devices hang pretty much the same way, so the
mid layer can write a reasonable strategy routine and just have the low
level driver fill in the implementation hooks (abort hook, bus reset hook,
full host reset hook, etc) and B) I'm an intelligent firmware and you just
let me do my thing, in which case the mid layer can't really do anything
and the low level drivers only need to leave the hooks undefined because
if the firmware looses the command we're bust already.

> As usual, the higher layers
> don't really know what is going on, and the lower levels on smarter cards
> are likely to be doing the right thing on their own, no?

Lower levels on smart cards assume that if the card ever looses a command
then you just want a watchdog to reset the machine anyway.

> (Yes, we should improve the infrastructure for having per-command timeouts
> etc, but the reset/abort callbacks have always been strange)
> Linus

  Doug Ledford <>     919-754-3700 x44233
         Red Hat, Inc. 
         1801 Varsity Dr.
         Raleigh, NC 27606
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

This archive was generated by hypermail 2b29 : Fri Nov 15 2002 - 22:00:23 EST