Re: [patch] nfs: use GFP_NOFS preloads for radix-tree insertion

[Peter Zijlstra <peterz@infradead.org>]

On Wed, 2007-11-14 at 05:20 +0100, Nick Piggin wrote:
> On Tue, Nov 13, 2007 at 11:55:45AM +0100, Peter Zijlstra wrote:
> > 
> > On Thu, 2007-11-08 at 07:56 +0100, Nick Piggin wrote:
> > > Here is the NFS version. I guess Trond should ack it before you pick it
> > > up.
> > > 
> > > --
> > > 
> > > NFS should use GFP_NOFS mode radix tree preloads rather than GFP_ATOMIC
> > > allocations at radix-tree insertion-time. This is important to reduce the
> > > atomic memory requirement.
> > 
> > In another mail you said:
> > 
> > > Anyway we can also simplify the code because the insertion can't fail with a
> > > preload.
> > 
> > Can we please avoid adding strict dependencies on that as the preload
> > API is unsupportable in -rt.
> 
> You can surely support it. You just have to do per-thread preloads if you
> want preemption left on.

Well, true, but that would mean adding stuff to task_struct, not the end
of the world I guess.

But as it is leaving the error handling on each individual
radix_tree_insert() allows us to just use GFP_KERNEL for everything.

The other, nicer option, is to do preload on the radix_tree_context
object instead.


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>