On 10/31/2013 06:33 PM, Daniel P. Berrange wrote:
On Thu, Oct 31, 2013 at 04:40:11PM +0100, 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(-)
Have you tested this on a system where the IPv6 kernel module is not present, and on a system with IPv6 kmod is present, but no interfaces have any IPv6 addresses (not even link-local ones). I'm concerned this is going to trigger failures on IPv4 only hosts.
When the ipv6 module is not present, socket(AF_INET6) returns EAFNOSUPPORT. With sysctl net.ipv6.conf.all.disable_ipv6=1 even bind succeeds.
diff --git a/src/util/virportallocator.c b/src/util/virportallocator.c index 1922ea6..24b4bff 100644 --- a/src/util/virportallocator.c +++ b/src/util/virportallocator.c
+ if (family == AF_INET6) { + addr = (struct sockaddr*)&addr6; + addrlen = sizeof(addr6); + ipv6 = true; + } else if (family == AF_INET) { + addr = (struct sockaddr*)&addr4; + addrlen = sizeof(addr4);
family can be AF_INET or AF_INET6 here...
+ } else { + virReportError(VIR_ERR_INTERNAL_ERROR, _("Unknown family %d"), family); + return -1; + }
*used = false;
fd = socket(PF_INET, SOCK_STREAM, 0);
...but you're hardcoding PF_INET here. I'm not sure it is valid to use an PF_INET socket with an AF_INET6 address in bind().
Ouch, rebase error. That can't work. s/PF_INET/family/
if (fd < 0) { + if (errno == EAFNOSUPPORT) + return 0; virReportSystemError(errno, "%s", _("Unable to open test socket")); goto cleanup; }
Jan