s3: Add support for AF_NETLINK addr notifications

Via an AF_NETLINK socket, the Linux kernel can inform us when IP addresses are
added or dropped.

This will first be used in winbind, it was triggered by clustering with ctdb.
When winbind is connected to a domain controller and ctdb decides to move away
the IP address that winbind used locally for the connection to the DC, the next
request will run into a timeout. winbind sends out its request, but the
response will never arrive: The IP is gone.

It will also be interesting for more reliable online/offline detection, but
this is something for future winbind refactoring.

1 files changed, 271 insertions, 0 deletions

diff --git a/source3/lib/addrchange.c b/source3/lib/addrchange.c
new file mode 100644
index 0000000000..cf415d8c7b
--- /dev/null
+++ b/source3/lib/addrchange.c
@@ -0,0 +1,271 @@
+/*
+ * Samba Unix/Linux SMB client library
+ * Copyright (C) Volker Lendecke 2011
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "includes.h"
+#include "lib/addrchange.h"
+
+#if HAVE_LINUX_RTNETLINK_H
+
+#include "linux/netlink.h"
+#include "linux/rtnetlink.h"
+#include "lib/async_req/async_sock.h"
+
+struct addrchange_context {
+	int sock;
+	uint8_t *buf;
+};
+
+static int addrchange_context_destructor(struct addrchange_context *c);
+
+NTSTATUS addrchange_context_create(TALLOC_CTX *mem_ctx,
+				   struct addrchange_context **pctx)
+{
+	struct addrchange_context *ctx;
+	struct sockaddr_nl addr;
+	NTSTATUS status;
+	int res;
+
+	ctx = talloc(mem_ctx, struct addrchange_context);
+	if (ctx == NULL) {
+		return NT_STATUS_NO_MEMORY;
+	}
+
+	ctx->sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
+	if (ctx->sock == -1) {
+		status = map_nt_error_from_unix(errno);
+		goto fail;
+	}
+	talloc_set_destructor(ctx, addrchange_context_destructor);
+
+	/*
+	 * We're interested in address changes
+	 */
+	ZERO_STRUCT(addr);
+	addr.nl_family = AF_NETLINK;
+	addr.nl_groups = RTMGRP_IPV6_IFADDR | RTMGRP_IPV4_IFADDR;
+
+	res = bind(ctx->sock, (struct sockaddr *)(void *)&addr, sizeof(addr));
+	if (res == -1) {
+		status = map_nt_error_from_unix(errno);
+		goto fail;
+	}
+
+	*pctx = ctx;
+	return NT_STATUS_OK;
+fail:
+	TALLOC_FREE(ctx);
+	return status;
+}
+
+static int addrchange_context_destructor(struct addrchange_context *c)
+{
+	if (c->sock != -1) {
+		close(c->sock);
+		c->sock = 0;
+	}
+	return 0;
+}
+
+struct addrchange_state {
+	uint8_t buf[8192];
+	struct sockaddr_storage fromaddr;
+	socklen_t fromaddr_len;
+
+	enum addrchange_type type;
+	struct sockaddr_storage addr;
+};
+
+static void addrchange_done(struct tevent_req *subreq);
+
+struct tevent_req *addrchange_send(TALLOC_CTX *mem_ctx,
+				   struct tevent_context *ev,
+				   struct addrchange_context *ctx)
+{
+	struct tevent_req *req, *subreq;
+	struct addrchange_state *state;
+
+	req = tevent_req_create(mem_ctx, &state, struct addrchange_state);
+	if (req == NULL) {
+		return NULL;
+	}
+
+	state->fromaddr_len = sizeof(state->fromaddr);
+
+	subreq = recvfrom_send(state, ev, ctx->sock,
+			       state->buf, sizeof(state->buf), 0,
+			       &state->fromaddr, &state->fromaddr_len);
+	if (tevent_req_nomem(subreq, req)) {
+		return tevent_req_post(req, ev);
+	}
+	tevent_req_set_callback(subreq, addrchange_done, req);
+	return req;
+}
+
+static void addrchange_done(struct tevent_req *subreq)
+{
+	struct tevent_req *req = tevent_req_callback_data(
+		subreq, struct tevent_req);
+	struct addrchange_state *state = tevent_req_data(
+		req, struct addrchange_state);
+	struct nlmsghdr *h;
+	struct ifaddrmsg *ifa;
+	struct rtattr *rta;
+	ssize_t received;
+	int len;
+	int err;
+	bool found;
+
+	received = recvfrom_recv(subreq, &err);
+	TALLOC_FREE(subreq);
+	if (received == -1) {
+		DEBUG(10, ("recvfrom returned %s\n", strerror(errno)));
+		tevent_req_nterror(req, map_nt_error_from_unix(err));
+		return;
+	}
+	if (received < sizeof(struct nlmsghdr)) {
+		DEBUG(10, ("received %d, expected at least %d\n",
+			   (int)received, (int)sizeof(struct nlmsghdr)));
+		tevent_req_nterror(req, NT_STATUS_UNEXPECTED_IO_ERROR);
+		return;
+	}
+
+	h = (struct nlmsghdr *)state->buf;
+	if (h->nlmsg_len < sizeof(struct nlmsghdr)) {
+		DEBUG(10, ("nlmsg_len=%d, expected at least %d\n",
+			   (int)h->nlmsg_len, (int)sizeof(struct nlmsghdr)));
+		tevent_req_nterror(req, NT_STATUS_UNEXPECTED_IO_ERROR);
+		return;
+	}
+	if (h->nlmsg_len > received) {
+		DEBUG(10, ("nlmsg_len=%d, expected at most %d\n",
+			   (int)h->nlmsg_len, (int)received));
+		tevent_req_nterror(req, NT_STATUS_UNEXPECTED_IO_ERROR);
+		return;
+	}
+	switch (h->nlmsg_type) {
+	case RTM_NEWADDR:
+		state->type = ADDRCHANGE_ADD;
+		break;
+	case RTM_DELADDR:
+		state->type = ADDRCHANGE_DEL;
+		break;
+	default:
+		DEBUG(10, ("Got unexpected type %d\n", h->nlmsg_type));
+		tevent_req_nterror(req, NT_STATUS_INVALID_ADDRESS);
+		return;
+	}
+
+	if (h->nlmsg_len < sizeof(struct nlmsghdr)+sizeof(struct ifaddrmsg)) {
+		DEBUG(10, ("nlmsg_len=%d, expected at least %d\n",
+			   (int)h->nlmsg_len,
+			   (int)(sizeof(struct nlmsghdr)
+				 +sizeof(struct ifaddrmsg))));
+		tevent_req_nterror(req, NT_STATUS_UNEXPECTED_IO_ERROR);
+		return;
+	}
+
+	ifa = (struct ifaddrmsg *)NLMSG_DATA(h);
+
+	state->addr.ss_family = ifa->ifa_family;
+
+	rta = IFA_RTA(ifa);
+	len = h->nlmsg_len - sizeof(struct nlmsghdr) + sizeof(struct ifaddrmsg);
+
+	found = false;
+
+	for (rta = IFA_RTA(ifa); RTA_OK(rta, len); rta = RTA_NEXT(rta, len)) {
+
+		if ((rta->rta_type != IFA_LOCAL)
+		    && (rta->rta_type != IFA_ADDRESS)) {
+			continue;
+		}
+
+		switch (ifa->ifa_family) {
+		case AF_INET: {
+			struct sockaddr_in *v4_addr;
+			v4_addr = (struct sockaddr_in *)(void *)&state->addr;
+
+			if (RTA_PAYLOAD(rta) != sizeof(uint32_t)) {
+				continue;
+			}
+			v4_addr->sin_addr.s_addr = *(uint32_t *)RTA_DATA(rta);
+			found = true;
+			break;
+		}
+		case AF_INET6: {
+			struct sockaddr_in6 *v6_addr;
+			v6_addr = (struct sockaddr_in6 *)(void *)&state->addr;
+
+			if (RTA_PAYLOAD(rta) !=
+			    sizeof(v6_addr->sin6_addr.s6_addr)) {
+				continue;
+			}
+			memcpy(v6_addr->sin6_addr.s6_addr, RTA_DATA(rta),
+			       sizeof(v6_addr->sin6_addr.s6_addr));
+			found = true;
+			break;
+		}
+		}
+	}
+
+	if (!found) {
+		tevent_req_nterror(req, NT_STATUS_INVALID_ADDRESS);
+		return;
+	}
+
+	tevent_req_done(req);
+}
+
+NTSTATUS addrchange_recv(struct tevent_req *req, enum addrchange_type *type,
+			 struct sockaddr_storage *addr)
+{
+	struct addrchange_state *state = tevent_req_data(
+		req, struct addrchange_state);
+	NTSTATUS status;
+
+	if (tevent_req_is_nterror(req, &status)) {
+		return status;
+	}
+
+	*type = state->type;
+	*addr = state->addr;
+	return NT_STATUS_OK;
+}
+
+#else
+
+NTSTATUS addrchange_context_create(TALLOC_CTX *mem_ctx,
+				   struct addrchange_context **pctx)
+{
+	return NT_STATUS_NOT_SUPPORTED;
+}
+
+struct tevent_req *addrchange_send(TALLOC_CTX *mem_ctx,
+				   struct tevent_context *ev,
+				   struct addrchange_context *ctx)
+{
+	return NULL;
+}
+
+NTSTATUS addrchange_recv(struct tevent_req *req, enum addrchange_type *type,
+			 struct sockaddr_storage *addr)
+{
+	return NT_STATUS_NOT_IMPLEMENTED;
+}
+
+#endif