Re: [RFC, PATCH] mm: map few pages around fault address if they arein page cache

[Kirill A. Shutemov]

On Fri, Feb 07, 2014 at 09:32:00AM -0800, Andi Kleen wrote:
> > I haven't tested it much, but my kvm boots. There're few places where code
> > should be fixed. __do_fault() and filemap_fault() are too ugly and need to
> > be cleaned.
> > 
> > I don't have any performance data yet.
> > 
> > Any thoughts?
> 
> It seems very drastic to do it unconditionally. How about at least a simple
> stream detection heuristic and perhaps also madvise?

We already have readahead here it can be reused here. But see below.

> There are some extreme cases where workloads could use a lot more memory
> than before, if they access their memory sparsely in the right pattern.

Have you noticied that we don't actually allocate any memory: only reuse
what's already there. Sure, it will increase VmSize, but do we care?

-- 
 Kirill A. Shutemov
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/