Re: GFS2 and DLM

[Nathan Scott <nathans@sgi.com>]

On Tue, Jun 27, 2006 at 10:22:40AM +0200, Ingo Molnar wrote:
> yeah, you are right - sorry about that.

No worries.

> which is in essence an open-coded GFP_NOFAIL implementation. Here's what 
> __GFP_NOFAIL does:

Heh, trust me, I know very well what the code does here.

> and since XFS makes use of KM_SLEEP in 130+ callsites, that means it is 
> in essence using GFP_NOFAIL massively!

Their locations have been carefully audited and understood.  The original
issue here was IRIX being able to do a very good of preventing kernel
memory allocation failures, which I suspect caused the original XFS guys
to be fairly relaxed in their handling of memory allocation failures.
Its caused us no end of pain with the Linux port, I assure you.

But, I digress.  If Christoph says you have a problem, you probably do;
he really is not biased XFS's way - he's driven us to make many changes
leading up to our major merge points too, and he will happily point out
crappy code in XFS as well. ;-)

cheers.

-- 
Nathan