Re: [libvirt] [PATCH] Fix initialization of current vcpus in libxl driver
On 05/24/2011 09:14 AM, Daniel Veillard wrote:
On Mon, May 23, 2011 at 04:43:19PM -0600, Eric Blake wrote:
> On 05/23/2011 04:26 PM, Jim Fehlig wrote:
>> b_info->hvm = hvm;
>> b_info->max_vcpus = def->maxvcpus;
>> - b_info->cur_vcpus = def->vcpus;
>> + b_info->cur_vcpus = (1 << def->vcpus) - 1;
>
> That's a compilation pitfall if def->vcpus is exactly 32.
>
> Shoot; xenFormatSxpr in xenxs/xen_sxpr.c has the same bug.
What would be the best way to fix ? cast to a larger int and
recast the result or provide some kind of helper function for
bitmaps ?
Maybe special-case:
if (def->vcpus == 32)
b_info->cur_vcpus = (uint32_t) -1;
else
b_info->cur_vcpus = (1 << def->vcpus) - 1;
Or build it up differently (although it gets harder to read):
b_info->cur_vcpus = ((1 << (def->vcpus - 1)) - 1) * 2 + 1;
As long as we rework it to avoid the undefined overflow of 1 << 32, when
we really just want 32 bits set.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org
Attachments:
- signature.asc (application/pgp-signature — 619 bytes)