[libvirt] [PATCH 1/2] conf: Check if number of vCPUs fits in the storage variable

[Eric Blake]

On 01/22/2013 08:07 AM, John Ferlan wrote:
> On 01/22/2013 09:31 AM, Peter Krempa wrote:
>> The count of vCPUs for a domain is extracted as a usingned long variable
>> but is stored in a unsigned short. If the actual number was too large,
>> a faulty number was stored.

>> +        if (count == 0 || (unsigned short) count != count) {
> 
> maxvcpus is a 'unsigned short' and count is an 'unsigned long', thus if
> def->maxvcpus != count after this point, then we have the overflow,
> right?  Or would the compiler "adjust" that comparison behind our back
> on an if check?

For unsigned types, the C standard guarantees that overflow wraps
around, and that casting a larger type down to a smaller type in order
to compare the same number is required to tell you if overflow happened,
at all optimization levels.  This code is valid.

For signed types, the C standard says overflow leads to unspecified
behavior, so all bets are off.  Thankfully, this isn't dealing with
signed types.

> ACK - I think what you've done is right, although perhaps someone with a
> bit more knowledge of what the compiler does could pipe in (I'm curious
> too).

I think the ACK stands.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 621 bytes
Desc: OpenPGP digital signature
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20130123/fe443c1c/attachment-0001.sig>