[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v12 05/11] x86/mm: add HYPERVISOR_memory_op to acquire guest resources



Hi,

On 19/10/17 16:47, Jan Beulich wrote:
On 19.10.17 at 17:37, <julien.grall@xxxxxxxxxx> wrote:
Hi,

On 19/10/17 16:11, Jan Beulich wrote:
On 19.10.17 at 16:49, <Paul.Durrant@xxxxxxxxxx> wrote:
I'd prefer to make the whole thing x86-only since that's the only platform
on which I can test it, and indeed the code used to be x86-only. Jan
objected
to this so all I'm trying to achieve is that it builds for ARM. Please can
you and
Jan reach agreement on where the code should live and how, if at all, it
should be #ifdef-ed?

I am quite surprised of "it is tools-only" so it is fine to not protect
it even if it is x86 only. That's probably going to bite us in the future.


So, this appears to have reached an impasse. I don't know how to proceed
without having to also fix priv mapping for x86, which is a yak far too
large
for me to shave at the moment.

Julien,

why is it that you are making refcounting on p2m insertion / removal
a requirement for this series? We all know it's not there right now
(and similarly also not for the IOMMU, which might affect ARM as well
unless you always use shared page tables), and it's - as Paul validly
says - unrelated to his series.

Well, we do at least have refcounting for foreign mapping on Arm. So if
a foreign domain remove the mapping, the page will stay allocated until
the last mapping is removed. For IOMMU, page tables are for the moment
always shared.

If you don't want to fix x86 now, then it is fine. But I would
appreciate if you don't spread that on Arm.

To give you a bit more context, I was ready to implement the Arm version
of set_foreign_p2m_entry (it is quite trivial) to append at the end of
this series. But given that refcounting is not done, I am more reluctant
to do that.

I don't understand: The refcounting is to be done by ARM-specific
code anyway, i.e. by the implementation of set_foreign_p2m_entry(),
not its caller. At least that's what I would have expected.

I thought I said it before, but it looks like not. Assuming the MFN is always baked by a domain, the prototype would likely need to be extended and take the foreign domain.

If it is not the case, we would need to find another way to do refcounting.

Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.