[libvirt] [PATCH 1/3] conf: ABI: Hugepage backing definition is not guest ABI

[John Ferlan]

On 02/16/2015 02:50 PM, Peter Krempa wrote:
> The backing of the vm's memory isn't influencing the guest ABI thus
> shouldn't be checked.
> ---
>  src/conf/domain_conf.c | 24 ------------------------
>  1 file changed, 24 deletions(-)
> 

Just to put it "officially there" - ACK...

some thoughts below FWIW


John

> diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
> index fe73a5f..9a7060d 100644
> --- a/src/conf/domain_conf.c
> +++ b/src/conf/domain_conf.c
> @@ -15969,30 +15969,6 @@ virDomainDefCheckABIStability(virDomainDefPtr src,
>                         dst->mem.cur_balloon, src->mem.cur_balloon);
>          goto error;
>      }
> -    if (src->mem.nhugepages != dst->mem.nhugepages) {
> -        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
> -                       _("Target domain huge pages count %zu does not match source %zu"),
> -                       dst->mem.nhugepages, src->mem.nhugepages);
> -        goto error;
> -    }

Looks like this hunk was originally added by commit id 'f2fc1eee' when
it was one huge chunk... I'm not overly familiar with how hugepages
work, but agree that it doesn't seem to cause ABI issues.

> -    for (i = 0; i < src->mem.nhugepages; i++) {
> -        virDomainHugePagePtr src_huge = &src->mem.hugepages[i];
> -        virDomainHugePagePtr dst_huge = &dst->mem.hugepages[i];
> -
> -        if (src_huge->size != dst_huge->size) {
> -            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
> -                           _("Target domain huge page size %llu "
> -                             "does not match source %llu"),
> -                           dst_huge->size, src_huge->size);
> -            goto error;
> -        }
> -
> -        if (!virBitmapEqual(src_huge->nodemask, dst_huge->nodemask)) {
> -            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> -                           _("Target huge page nodemask does not match source"));
> -            goto error;
> -        }
> -    }

Looks like this hunk was added by commit id '136ad497' when the one
chunk was split with the extra mask capability.  Again, I agree it
doesn't appear to cause ABI issues.

> 
>      if (src->vcpus != dst->vcpus) {
>          virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
>