Re: smp dead lock of io_request_lock/queue_lock patch

[Christoph Hellwig <hch@infradead.org>]

On Sat, Jan 17, 2004 at 02:21:43PM -0500, Doug Ledford wrote:
> I should pay more attention to what goes on in 2.6.  That was a crap
> decision.  There are drivers out there that use global data and need one
> lock across all controllers.  Putting a lock in the scsi host struct for
> per controller locks is a waste of space.  Let the driver decide if it
> needs a global driver lock, a per controller lock (in which case it
> should be part of the driver private host struct), or just the global
> io_request_lock.  It really is up to the driver and putting it into the
> scsi host struct doesn't make sense.

I tend to disagree.  Giving the driver control over the lock used by the
midlayer is just wrong.  If it has global state (which a good driver better
shouldn't) it's up to the driver to protect it.  That we still have a way
to override that lock is the big bug rather, and in fact only three drivers
are doing that, and they all override it with a lock that's per-HBA anyway.

> 
> -- 
>   Doug Ledford <dledford@redhat.com>     919-754-3700 x44233
>          Red Hat, Inc.
>          1801 Varsity Dr.
>          Raleigh, NC 27606
> 
> 
---end quoted text---

-- 
Christoph Hellwig <hch@lst.de>		-	Freelance Hacker
Contact me for driver hacking and kernel development consulting