[dm-devel] Re: [PATCH] dm-rwlock.patch (Re: 2.6.4-rc1-mm1: queue-congestion-dm-implementation patch)

[Andrew Morton]

Miquel van Smoorenburg <miquels at cistron.nl> wrote:
>
> On Mon, 08 Mar 2004 21:19:54, Joe Thornber wrote:
> > On Mon, Mar 08, 2004 at 08:12:18PM +0100, Miquel van Smoorenburg wrote:
> > > I also wasn't sure how bad it is for md->map to change during
> > > dm_request. With your patch, __split_bio() has a pointer to a
> > > refcounted copy of the table in ci.map, but multiple calls to
> > > __split_bio from the same dm_request could still end up with
> > > different maps, which could be bad (I don't know).
> > 
> > No, dm_request() only calls __dm_request() and hence __split_bio()
> > once.
> 
> Ofcourse you're right. I was thinking of __map_bio(), which is called
> several times, but it doesn't take a table as argument ofcourse.

I'll plunk Joe's update to your patch into the next -mm.  I'm not testing
DM personally, so I'd appreciate it if you could keep an eye on what's
happening in there and run regression tests.

Thanks.