Re: [PATCH] KVM: MMU: Fix rmap_remove() race

[Andrea Arcangeli <andrea@qumranet.com>]

On Thu, Mar 27, 2008 at 03:56:56PM +0200, Avi Kivity wrote:
> That's not good.  We need to support the older userspace, for a while yet.
>
> Why is there a problem? IIRC it's just anonymous memory.

Problem is that for it to be unmapped __do_fault must call
page_add_new_anon_rmap on it. Even anon would be 1, that would mean
cow and that's clearly not what you want as it would lose the
visibility on the future guest writes. anon will be 0 if qemu is
reading for example, leaving pinned anonymous memory that mmu
notifiers won't be able to swap anymore. This is about providing the
guarantee to the admin, that if he enables the mmu notifiers in the
kernel, giving access to /dev/kvm won't give full mlock privileges
too. Not sure if anybody will care, but it's a bit like removing the
limit on tmpfs that only half ram can be pinned.

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace