3:52 p.m.
Scott Feldman <scofeldm(a)cisco.com> wrote on 05/20/2010 03:22:12 PM:
Next steps for V3, etc:
1) merge with Stefan's latest patch, as much as possible
Yes. Try to make it a patch on top of my V5 patch...
2) assign VM device mac addr to eth and set macvtap mac addr to
random
Hm, why is this necessary? I have tested the macvtap device and always set
it to the address that libvirt assigned to it and that happened to be the
same as the one inside the VM? Is this something specific to your
underlying driver?
3) figure out where host_uuid lives
I posted a message a while ago. It's not anywhere right now and should
probably be provided via libvirt.conf overriding dmidecode, but dmidecode
being used if its output is valid and resorting to a temporary (between
libvirt restart) host uuid if everything fails...
diff --git a/src/util/macvtap.c b/src/util/macvtap.c
index 1f8dd29..7803bc8 100644
--- a/src/util/macvtap.c
+++ b/src/util/macvtap.c
@@ -85,14 +85,14 @@ static void nlClose(int fd)
* buffer will be returned.
*/
static
-int nlComm(struct nlmsghdr *nlmsg,
+int nlComm(struct nlmsghdr *nlmsg, int nlgroups,
char **respbuf, int *respbuflen)
{
int rc = 0;
struct sockaddr_nl nladdr = {
.nl_family = AF_NETLINK,
.nl_pid = 0,
- .nl_groups = 0,
+ .nl_groups = nlgroups,
I think if only you use this function, nlgroups can be 0 here. I would use
different code to use multicast for talking to lldpad.
};
int rcvChunkSize = 1024; // expecting less than that
int rcvoffset = 0;
@@ -192,6 +192,27 @@ nlAppend(struct nlmsghdr *nlm, int totlen,
const void *data, int datalen)
return pos;
}
+#define NLMSG_TAIL(nmsg) \
+ ((struct rtattr *) (((char *) (nmsg)) +
NLMSG_ALIGN((nmsg)->nlmsg_len)))
+
+static struct rtattr *nlNest(struct nlmsghdr *nlm, int totlen, int
type)
+{
+ struct rtattr *nest = NLMSG_TAIL(nlm);
+
+ if (nlm->nlmsg_len + NLMSG_ALIGN(sizeof(*nest)) > totlen)
+ return NULL;
+ nest->rta_type = type;
+ nest->rta_len = RTA_LENGTH(0);
+ nlm->nlmsg_len += sizeof(*nest);
+ nlAlign(nlm);
+ return nest;
+}
+
+static int nlNestEnd(struct nlmsghdr *nlm, struct rtattr *nest)
+{
+ nest->rta_len = (char *)NLMSG_TAIL(nlm) - (char *)nest;
+ return nlm->nlmsg_len;
+}
At some point in the future someone may want to change all this netlink
stuff to use libnl -- except for maybe in the case of the multicast
message that I am sending in one of my patches and using select with a
timeout to wait for a response from (future) lldpad where I am not sure
whether an equivalent function exists in libnl...
+#if 0
+static int
+get_host_uuid(char *host_uuid, int len)
this is not active code ... skipping it
+
+static int
+portProfileIdMcast(const char *linkdev,
+ const char request,
+ const char *profileid,
+ const unsigned short *instance_uuid,
+ const unsigned short *host_uuid)
+{
+ struct rtattr *port_self;
+ char nlmsgbuf[512];
+ struct nlmsghdr *nlm = (struct nlmsghdr *)nlmsgbuf, *resp;
+ struct ifinfomsg infomsg = { .ifi_family = AF_UNSPEC };
+ char rtattbuf[256];
+ struct rtattr *rta;
+ char *recvbuf = NULL;
+ int recvbuflen;
+ struct nlmsgerr *err;
+ int rc = 0;
+
+ memset(&nlmsgbuf, 0, sizeof(nlmsgbuf));
+ nlInit(nlm, NLM_F_REQUEST, RTM_SETLINK);
+
+ if (!nlAppend(nlm, sizeof(nlmsgbuf), &infomsg, sizeof(infomsg)))
+ goto buffer_too_small;
+
+ rta = rtattrCreate(rtattbuf, sizeof(rtattbuf), IFLA_IFNAME,
+ linkdev, strlen(linkdev) + 1);
+ if (!rta)
+ goto buffer_too_small;
+
+ if (!nlAppend(nlm, sizeof(nlmsgbuf), rtattbuf, rta->rta_len))
+ goto buffer_too_small;
+
+ port_self = nlNest(nlm, sizeof(nlmsgbuf), IFLA_PORT_SELF);
+ if (!port_self)
+ return -1;
+
+ switch (request) {
+ case PORT_REQUEST_ASSOCIATE:
+ if (profileid && strlen(profileid)) {
+ rta = rtattrCreate(rtattbuf, sizeof(rtattbuf),
+ IFLA_PORT_PROFILE,
+ profileid, strlen(profileid) + 1);
+ if (!rta)
+ goto buffer_too_small;
+
+ if (!nlAppend(nlm, sizeof(nlmsgbuf), rtattbuf,
rta->rta_len))
+ goto buffer_too_small;
+ }
+ if (instance_uuid) {
+ rta = rtattrCreate(rtattbuf, sizeof(rtattbuf),
+ IFLA_PORT_INSTANCE_UUID,
+ instance_uuid, PORT_UUID_MAX);
+ if (!rta)
+ goto buffer_too_small;
+
+ if (!nlAppend(nlm, sizeof(nlmsgbuf), rtattbuf,
rta->rta_len))
+ goto buffer_too_small;
+ }
+ if (host_uuid) {
+ rta = rtattrCreate(rtattbuf, sizeof(rtattbuf),
+ IFLA_PORT_HOST_UUID,
+ host_uuid, PORT_UUID_MAX);
+ if (!rta)
+ goto buffer_too_small;
+
+ if (!nlAppend(nlm, sizeof(nlmsgbuf), rtattbuf,
rta->rta_len))
+ goto buffer_too_small;
+ }
+ break;
+ case PORT_REQUEST_DISASSOCIATE:
+ break;
+ }
+
+ rta = rtattrCreate(rtattbuf, sizeof(rtattbuf), IFLA_PORT_REQUEST,
+ &request, 1);
+ if (!rta)
+ goto buffer_too_small;
+
+ if (!nlAppend(nlm, sizeof(nlmsgbuf), rtattbuf, rta->rta_len))
+ goto buffer_too_small;
+
+ nlNestEnd(nlm, port_self);
+
+ if (nlComm(nlm, RTNLGRP_LINK, &recvbuf, &recvbuflen) < 0)
+ return -1;
+
I know that it doesn't work like this in the netlink multicast case where
libvirt will receive an error from the kernel and hopefully some time
later a response from lldpad. Still it would be nice if we could use
common code here for both technologies, but the unicast vs. multicast
receiver code makes it quite different. Have a look at my patches.
Stefan