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

Re: [Xen-devel] [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate

Hi Stefano,

On 2/6/19 7:35 PM, Stefano Stabellini wrote:
On Wed, 6 Feb 2019, Julien Grall wrote:
However, I think we should add some sanity check in arch_set_info_guest for
our peace of mind. For guest entry/exit, rather than zero the upper 32-bits I
would also add sanity check in enter_hypervisor_head and leave_hypervisor_tail
but only in debug build. Any opinions?

Definitely we should have sanity checks.

I will add this one and ...

In the official public headers, I can't find anything telling you the size of
each arguments and the number of arguments. Instead you have to look at Xen
code to know the exact number of arguments and the size. Did I miss anything?

No, it is like you wrote. I think I should have pushed the discussion
further and added more information to the Xen public headers back in
those days.

... and this one in my list of things to improve.

In conclusion, if you and other maintainers prefer unsigned long I'll
drop my reservation.

The summary of my e-mail is:
     - we need to add sanity check (or zero) upper-bits for 32-bit guest
     - unsigned long should be fine for 64-bit only features

I take that you mean that unsigned long should be fine, and would also
allow us to introduce 64-bit only features in the future, right?

that's correct.

You are
*not* saying that unsigned long should only be used with 64-bit

unsigned long == uint32_t on 32-bit. So it would be silly of me to suggest that :).

     - we need to document the behavior of each hypercall and provide
guidelines for new one.

None of this is specific to Argo and I would be happy to defer this as a
follow-up series.

Assuming my understanding is right, I agree with you.

You understood right.


Julien Grall

Xen-devel mailing list



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