On 10/31/2013 06:40 PM, Eric Blake wrote:
On 10/31/2013 09:40 AM, Ján Tomko wrote:
Also try to bind on IPv6 to check if the port is occupied.
https://bugzilla.redhat.com/show_bug.cgi?id=1025407 --- src/util/virportallocator.c | 45 ++++++++++++++++++++++++++++++++++++++------- 1 file changed, 38 insertions(+), 7 deletions(-)
Since it is listed as a bugzilla fix, I could probably live with this in 1.1.4; but it would be nice to get a second opinion (after all, this is not a regression, so it could be argued that we are finally implementing a missing feature).
fd = socket(PF_INET, SOCK_STREAM, 0); if (fd < 0) { + if (errno == EAFNOSUPPORT) + return 0;
As to the code, I didn't spot any errors. I do wonder if checking for EAFNOSUPPORT on every iteration for finding a free port could be made more efficient by remembering up front if ipv6 is even enabled, and skipping the attempt in that case (that is, a running system is unlikely to change which subset of [ipv4, ipv6] that it currently supports except at boot). But there's probably more in the code base that we could optimize if we know that a particular family is unsupported for the given process.
I've tried to check up front if ipv6 is enabled in v1: https://www.redhat.com/archives/libvir-list/2013-October/msg00010.html Jan