[dm-devel] Re: [RFC] patch to dm-io.c

Dave Olien dmo at osdl.org
Thu Jun 17 17:58:06 UTC 2004


I had thought that might be the reason for the code duplication.
However for me, testing snapshot where the system is running out of
KVA, it actually seems to work better to not have the extra pool.
(it has 16gig physical memory, so KVA is in tight supply).

It's a difficult call.  I could put the private pool back. In which
case, I'd solve the other problem a little differently.

What do you think?

By the way, I notice my post last night didn't go out on dm-devel
because I'm not a member? I thought I was on the dm-devel list.

On Thu, Jun 17, 2004 at 04:35:27PM +0100, Alasdair G Kergon wrote:
> On Wed, Jun 16, 2004 at 11:22:57PM -0700, Dave Olien wrote:
> > It seems to work.  Let me know what you think.
> 
> Have you tested this under memory pressure?
> 
> The reason Joe used separate pools of bios for each client was
> to guarantee that it would not deadlock and still make reasonable
> progress.
> 
> [Devices can be stacked, remember.]
> 
> Alasdair
> -- 
> agk at redhat.com



More information about the dm-devel mailing list