On 26.07.2009, at 14:10, Avi Kivity wrote:
On 07/26/2009 02:55 PM, Alexander Graf wrote:
I will need some acks from ppc people. Obviously for the non-kvm
bits, but also for the kvm bits as I am not qualified to review
ppc code.
Right, FWIW Ben would actually even prefer to take the whole thing
in his tree.
That's likely to cause conflicts if some kvm API changes and needs
modifications to this code. Perhaps the best option is for the non-
kvm changes to go through the ppc tree, and for me to duplicate (but
not push) them.
I agree, but that's for you two to decide I guess :-). Let's just wait
for Ben to wake up.
- use MMU Notifiers
What's the plan here?
Implement MMU Notifiers as soon as I fully understand them? :-)
I'm aware of the basic concept, but the callbacks still seem
somewhat magical to me.
Ask and I'll do my best to answer.
Cool, next week :-). The current version works fine without. And I
need a weekend for a change :-D.
- use u64* for dirty log
Is this a userspace interface issue? if so it will need to be
addressed before merging.
Yes, on big endian having a 64-bit kernel and 32-bit userspace
breaks when dirty log is ulong*. Nobody saw this until now, because
it's not a big deal on little endian.
I sent a patch doing the qemu side of things already, but haven't
went through the kvm bits yet. Basically we can't use set_bit and
test_bit for the dirty log, because they require us to have the
bitmap as ulong*.
Yuck. What do we do? Implement set_bit_u64() and friends?
That's what I'd think we should do, yes.
For qemu I sent a patch that always uses "uint64_t" because that works
fine with the current code (32-bit on 64-bit uses the right bytes for
LE, 64-bit on 64-bit doesn't change). PPC is not implement yet for
anyone either and for now we only have dirty logging with 64 bit
kernels and 32-bit or 64-bit userspace, so that's fine too.
As soon as someone implement dirty logging on a 32-bit big endian
target, it will break though, so before that day comes we need to
change set_bit to set_bit_64.
Alex
--
To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html