On 03/03/2017 10:00 AM, Cédric Bosdonnat wrote:
virNetlinkCommand() processes only one response message, while some netlink commands like routes dumping need to process several ones. Add virNetlinkDumpCommand() as a virNetlinkCommand() sister. --- src/libvirt_private.syms | 1 + src/util/virnetlink.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++++ src/util/virnetlink.h | 9 ++++++++ 3 files changed, 65 insertions(+)
diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index bce0487ab..71143851c 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -2117,6 +2117,7 @@ virNetDevVPortProfileOpTypeToString; # util/virnetlink.h virNetlinkCommand; virNetlinkDelLink; +virNetlinkDumpCommand; virNetlinkDumpLink; virNetlinkEventAddClient; virNetlinkEventRemoveClient; diff --git a/src/util/virnetlink.c b/src/util/virnetlink.c index 5fb49251c..4747ba5a4 100644 --- a/src/util/virnetlink.c +++ b/src/util/virnetlink.c @@ -335,6 +335,49 @@ int virNetlinkCommand(struct nl_msg *nl_msg, return ret; }
+int +virNetlinkDumpCommand(struct nl_msg *nl_msg, + virNetlinkDumpCallback callback, + uint32_t src_pid, uint32_t dst_pid, + unsigned int protocol, unsigned int groups, + void *opaque) +{ + int ret = -1; + bool end = false; + int len = 0; + struct nlmsghdr *resp = NULL; + struct nlmsghdr *msg = NULL; + + struct sockaddr_nl nladdr = { + .nl_family = AF_NETLINK, + .nl_pid = dst_pid, + .nl_groups = 0, + }; + virNetlinkHandle *nlhandle = NULL; + + if (!(nlhandle = virNetlinkDoCommand(nl_msg, src_pid, nladdr, + protocol, groups))) + goto cleanup; + + while (!end) { + len = nl_recv(nlhandle, &nladdr, (unsigned char **)&resp, NULL); + + for (msg = resp; NLMSG_OK(msg, len); msg = NLMSG_NEXT(msg, len)) { + if (msg->nlmsg_type == NLMSG_DONE) + end = true; + + if (callback(msg, opaque) < 0) + goto cleanup;
This is relying on the callback function to throw an error in order to get us out of the loop (assuming we never get an NLMSG_DONE). I don't see the callback that's written in patch 4 doing anything along the lines of virNetlinkGetErrorCode() though. Maybe you should call that function from here prior to the callback - that would eliminate the need to separately do that level of error checking in all the future consumers of this new function. Otherwise I think it looks okay (I can't claim to have great knowledge of multipart netlink messages though, so...)