On Monday 15 October 2007 21:07, Andi Kleen wrote:
quoted text
> On Tue, Oct 16, 2007 at 12:56:46AM +1000, Nick Piggin wrote:
> > Is this true even if you don't write through those old mappings?
>
> I think it happened for reads too.  It is a little counter intuitive
> because in theory the CPU doesn't need to write back non dirty lines,
> but in the one case which took so long to debug exactly this happened
> somehow.
>
> At it is undefined for reads and writes in the architecture so
> better be safe than sorry.

Yes, typo. I meant reads or writes.


quoted text
> And x86 CPUs are out of order and do speculative executation
> and that can lead to arbitary memory accesses even if the code
> never touches an particular address.
>
> Newer Intel CPUs have something called self-snoop which was supposed
> to handle this; but in some situations it doesn't seem to catch it
> either.

Fair enough, so we have to have this lazy tlb flush hook for
Xen/PAT/etc. I don't think it should be much problem to
implement.


quoted text
> > Is DRM or AGP then not also broken with lazy highmem flushing, or
> > how do they solve that?
>
> AGP doesn't allocate highmem pages.  Not sure about the DRM code.

Hmm, OK. It looks like DRM vmallocs memory (which gives highmem).
-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/