Re: [PATCH] mmu notifiers #v5

[Andrea Arcangeli <andrea@qumranet.com>]

On Sat, Feb 02, 2008 at 09:14:57PM -0600, Jack Steiner wrote:
> Also, most (but not all) applications that use the GRU do not usually do
> anything that requires frequent flushing (fortunately). The GRU is intended
> for HPC-like applications. These don't usually do frequent map/unmap
> operations or anything else that requires a lot of flushes.
> 
> I expect that KVM is a lot different.

I don't think so. invalidate_page/pages/range_start,end is a slow and
unfrequent path for KVM (or alternatively the ranges are very small in
which case _range_start/end won't payoff compared to _pages). Whenever
invalidate_page[s] become a fast path, we're generally I/O
bound. get_user_pages is always the fast path instead. I thought it
was much more important that get_user_pages scale as well as it does
now and that the KVM page fault isn't serialized with a mutex, than
whatever invalidate side optimization. get_user_pages may run
frequently from all vcpus even if there are no invalidates and no
memory pressure and I don't mean only during startup.

> I have most of the GRU code working with the latest mmuops patch. I still
> have a list of loose ends that I'll get to next week. The most important is
> the exact handling of the range invalidates. The code that I currently have
> works (mostly) but has a few endcases that will cause problems. Once I
> finish, I'll be glad to send you snippets of the code (or all of it) if you
> would like to take a look.

Sure.