#include <netinet/ether.h>
#include <stdbool.h>
#include <unistd.h>
-#include <linux/netlink.h>
-#include <linux/veth.h>
-#include <linux/if.h>
-#include <linux/ip.h>
-#include <linux/if_tunnel.h>
-#include <linux/if_bridge.h>
#include "util.h"
#include "refcnt.h"
#define GET_CONTAINER(m, i) ((i) < (m)->n_containers ? (struct rtattr*)((uint8_t*)(m)->hdr + (m)->container_offsets[i]) : NULL)
#define PUSH_CONTAINER(m, new) (m)->container_offsets[(m)->n_containers ++] = (uint8_t*)(new) - (uint8_t*)(m)->hdr;
+#define RTA_TYPE(rta) ((rta)->rta_type & NLA_TYPE_MASK)
+
static int message_new_empty(sd_rtnl *rtnl, sd_rtnl_message **ret) {
sd_rtnl_message *m;
}
int sd_rtnl_message_new_route(sd_rtnl *rtnl, sd_rtnl_message **ret,
- uint16_t nlmsg_type, unsigned char rtm_family) {
+ uint16_t nlmsg_type, int rtm_family,
+ unsigned char rtm_protocol) {
struct rtmsg *rtm;
int r;
return r;
if (nlmsg_type == RTM_NEWROUTE)
- (*ret)->hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
+ (*ret)->hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_APPEND;
rtm = NLMSG_DATA((*ret)->hdr);
rtm->rtm_scope = RT_SCOPE_UNIVERSE;
rtm->rtm_type = RTN_UNICAST;
rtm->rtm_table = RT_TABLE_MAIN;
- rtm->rtm_protocol = RTPROT_BOOT;
+ rtm->rtm_protocol = rtm_protocol;
return 0;
}
return 0;
}
+int sd_rtnl_message_link_set_family(sd_rtnl_message *m, unsigned family) {
+ struct ifinfomsg *ifi;
+
+ assert_return(m, -EINVAL);
+ assert_return(m->hdr, -EINVAL);
+ assert_return(rtnl_message_type_is_link(m->hdr->nlmsg_type), -EINVAL);
+
+ ifi = NLMSG_DATA(m->hdr);
+
+ ifi->ifi_family = family;
+
+ return 0;
+}
+
int sd_rtnl_message_new_link(sd_rtnl *rtnl, sd_rtnl_message **ret,
uint16_t nlmsg_type, int index) {
struct ifinfomsg *ifi;
return 0;
}
+int sd_rtnl_message_request_dump(sd_rtnl_message *m, int dump) {
+ assert_return(m, -EINVAL);
+ assert_return(m->hdr, -EINVAL);
+ assert_return(m->hdr->nlmsg_type == RTM_GETLINK ||
+ m->hdr->nlmsg_type == RTM_GETADDR ||
+ m->hdr->nlmsg_type == RTM_GETROUTE,
+ -EINVAL);
+
+ if (dump)
+ m->hdr->nlmsg_flags |= NLM_F_DUMP;
+ else
+ m->hdr->nlmsg_flags &= ~NLM_F_DUMP;
+
+ return 0;
+}
+
int sd_rtnl_message_addr_set_prefixlen(sd_rtnl_message *m, unsigned char prefixlen) {
struct ifaddrmsg *ifa;
return 0;
}
-int sd_rtnl_message_addr_get_family(sd_rtnl_message *m, unsigned char *family) {
+int sd_rtnl_message_addr_get_family(sd_rtnl_message *m, int *family) {
struct ifaddrmsg *ifa;
assert_return(m, -EINVAL);
return 0;
}
+int sd_rtnl_message_addr_get_prefixlen(sd_rtnl_message *m, unsigned char *prefixlen) {
+ struct ifaddrmsg *ifa;
+
+ assert_return(m, -EINVAL);
+ assert_return(m->hdr, -EINVAL);
+ assert_return(rtnl_message_type_is_addr(m->hdr->nlmsg_type), -EINVAL);
+ assert_return(prefixlen, -EINVAL);
+
+ ifa = NLMSG_DATA(m->hdr);
+
+ *prefixlen = ifa->ifa_prefixlen;
+
+ return 0;
+}
+
int sd_rtnl_message_addr_get_scope(sd_rtnl_message *m, unsigned char *scope) {
struct ifaddrmsg *ifa;
int sd_rtnl_message_new_addr(sd_rtnl *rtnl, sd_rtnl_message **ret,
uint16_t nlmsg_type, int index,
- unsigned char family) {
+ int family) {
struct ifaddrmsg *ifa;
int r;
assert_return(rtnl_message_type_is_addr(nlmsg_type), -EINVAL);
- assert_return(index > 0, -EINVAL);
- assert_return(family == AF_INET || family == AF_INET6, -EINVAL);
+ assert_return((nlmsg_type == RTM_GETADDR && index == 0) ||
+ index > 0, -EINVAL);
+ assert_return((nlmsg_type == RTM_GETADDR && family == AF_UNSPEC) ||
+ family == AF_INET || family == AF_INET6, -EINVAL);
assert_return(ret, -EINVAL);
r = message_new(rtnl, ret, nlmsg_type);
if (r < 0)
return r;
- if (nlmsg_type == RTM_GETADDR && family == AF_INET)
+ if (nlmsg_type == RTM_GETADDR)
(*ret)->hdr->nlmsg_flags |= NLM_F_DUMP;
ifa = NLMSG_DATA((*ret)->hdr);
}
int sd_rtnl_message_new_addr_update(sd_rtnl *rtnl, sd_rtnl_message **ret,
- int index, unsigned char family) {
+ int index, int family) {
int r;
r = sd_rtnl_message_new_addr(rtnl, ret, RTM_NEWADDR, index, family);
return 0;
}
+int sd_rtnl_message_link_get_type(sd_rtnl_message *m, unsigned *type) {
+ struct ifinfomsg *ifi;
+
+ assert_return(m, -EINVAL);
+ assert_return(m->hdr, -EINVAL);
+ assert_return(rtnl_message_type_is_link(m->hdr->nlmsg_type), -EINVAL);
+ assert_return(type, -EINVAL);
+
+ ifi = NLMSG_DATA(m->hdr);
+
+ *type = ifi->ifi_type;
+
+ return 0;
+}
+
/* If successful the updated message will be correctly aligned, if
unsuccessful the old message is untouched. */
static int add_rtattr(sd_rtnl_message *m, unsigned short type, const void *data, size_t data_length) {
- uint32_t rta_length, message_length;
+ uint32_t rta_length;
+ size_t message_length, padding_length;
struct nlmsghdr *new_hdr;
struct rtattr *rta;
char *padding;
unsigned i;
+ int offset;
assert(m);
assert(m->hdr);
assert(!m->sealed);
assert(NLMSG_ALIGN(m->hdr->nlmsg_len) == m->hdr->nlmsg_len);
- assert(!data || data_length > 0);
- assert(data || m->n_containers < RTNL_CONTAINER_DEPTH);
+ assert(!data || data_length);
+
+ /* get offset of the new attribute */
+ offset = m->hdr->nlmsg_len;
/* get the size of the new rta attribute (with padding at the end) */
rta_length = RTA_LENGTH(data_length);
/* get the new message size (with padding at the end) */
- message_length = m->hdr->nlmsg_len + RTA_ALIGN(rta_length);
+ message_length = offset + RTA_ALIGN(rta_length);
/* realloc to fit the new attribute */
new_hdr = realloc(m->hdr, message_length);
m->hdr = new_hdr;
/* get pointer to the attribute we are about to add */
- rta = (struct rtattr *) ((uint8_t *) m->hdr + m->hdr->nlmsg_len);
+ rta = (struct rtattr *) ((uint8_t *) m->hdr + offset);
/* if we are inside containers, extend them */
for (i = 0; i < m->n_containers; i++)
- GET_CONTAINER(m, i)->rta_len += message_length - m->hdr->nlmsg_len;
+ GET_CONTAINER(m, i)->rta_len += message_length - offset;
/* fill in the attribute */
rta->rta_type = type;
rta->rta_len = rta_length;
- if (!data) {
- //TODO: simply return this value rather than check for !data
- /* this is the start of a new container */
- m->container_offsets[m->n_containers ++] = m->hdr->nlmsg_len;
- } else {
+ if (data)
/* we don't deal with the case where the user lies about the type
* and gives us too little data (so don't do that)
- */
+ */
padding = mempcpy(RTA_DATA(rta), data, data_length);
- /* make sure also the padding at the end of the message is initialized */
- memzero(padding,
- (uint8_t *) m->hdr + message_length - (uint8_t *) padding);
+ else {
+ /* if no data was passed, make sure we still initialize the padding
+ note that we can have data_length > 0 (used by some containers) */
+ padding = RTA_DATA(rta);
}
+ /* make sure also the padding at the end of the message is initialized */
+ padding_length = (uint8_t*)m->hdr + message_length - (uint8_t*)padding;
+ memzero(padding, padding_length);
+
/* update message size */
m->hdr->nlmsg_len = message_length;
- return 0;
+ return offset;
}
static int message_attribute_has_type(sd_rtnl_message *m, uint16_t attribute_type, uint16_t data_type) {
assert_return(m, -EINVAL);
assert_return(!m->sealed, -EPERM);
+ assert_return(m->n_containers < RTNL_CONTAINER_DEPTH, -ERANGE);
r = message_attribute_has_type(m, type, NLA_NESTED);
if (r < 0)
if (r < 0)
return r;
- r = add_rtattr(m, type, NULL, size);
+ r = add_rtattr(m, type | NLA_F_NESTED, NULL, size);
if (r < 0)
return r;
+ m->container_offsets[m->n_containers ++] = r;
+
return 0;
}
if (r < 0)
return r;
+ m->container_offsets[m->n_containers ++] = r;
+
return 0;
}
return RTA_PAYLOAD(rta);
}
-int sd_rtnl_message_read_string(sd_rtnl_message *m, unsigned short type, char **data) {
+int sd_rtnl_message_read_string(sd_rtnl_message *m, unsigned short type, const char **data) {
int r;
void *attr_data;
else if (strnlen(attr_data, r) >= (size_t) r)
return -EIO;
- *data = (char *) attr_data;
+ *data = (const char *) attr_data;
return 0;
}
return r;
} else if (nl_type->type == NLA_UNION) {
const NLTypeSystemUnion *type_system_union;
- char *key;
+ const char *key;
r = type_system_get_type_system_union(m->container_type_system[m->n_containers],
&type_system_union,
return err->error;
}
-static int message_receive_need(sd_rtnl *rtnl, size_t *need) {
- assert(rtnl);
- assert(need);
-
- /* ioctl(rtnl->fd, FIONREAD, &need)
- Does not appear to work on netlink sockets. libnl uses
- MSG_PEEK instead. I don't know if that is worth the
- extra roundtrip.
-
- For now we simply use the maximum message size the kernel
- may use (NLMSG_GOODSIZE), and then realloc to the actual
- size after reading the message (hence avoiding huge memory
- usage in case many small messages are kept around) */
- *need = page_size();
- if (*need > 8192UL)
- *need = 8192UL;
-
- return 0;
-}
-
int rtnl_message_parse(sd_rtnl_message *m,
size_t **rta_offset_tb,
unsigned short *rta_tb_size,
*rta_tb_size = max + 1;
for (; RTA_OK(rta, rt_len); rta = RTA_NEXT(rta, rt_len)) {
- type = rta->rta_type;
+ type = RTA_TYPE(rta);
/* if the kernel is newer than the headers we used
when building, we ignore out-of-range attributes
return k;
}
-/* On success, the number of bytes received is returned and *ret points to the received message
- * which has a valid header and the correct size.
- * If nothing useful was received 0 is returned.
- * On failure, a negative error code is returned.
- */
-int socket_read_message(sd_rtnl *rtnl) {
- _cleanup_rtnl_message_unref_ sd_rtnl_message *first = NULL;
- sd_rtnl_message *previous = NULL;
- _cleanup_free_ void *buffer = NULL;
- uint8_t cred_buffer[CMSG_SPACE(sizeof(struct ucred))];
- struct iovec iov = {};
+static int socket_recv_message(int fd, struct iovec *iov, uint32_t *_group, bool peek) {
+ uint8_t cred_buffer[CMSG_SPACE(sizeof(struct ucred)) +
+ CMSG_SPACE(sizeof(struct nl_pktinfo))];
struct msghdr msg = {
- .msg_iov = &iov,
+ .msg_iov = iov,
.msg_iovlen = 1,
.msg_control = cred_buffer,
.msg_controllen = sizeof(cred_buffer),
};
struct cmsghdr *cmsg;
+ uint32_t group = 0;
bool auth = false;
- struct nlmsghdr *new_msg;
- size_t need, len;
- int r, ret = 0;
-
- assert(rtnl);
-
- r = message_receive_need(rtnl, &need);
- if (r < 0)
- return r;
-
- buffer = malloc0(need);
- if (!buffer)
- return -ENOMEM;
+ int r;
- iov.iov_base = buffer;
- iov.iov_len = need;
+ assert(fd >= 0);
+ assert(iov);
- r = recvmsg(rtnl->fd, &msg, 0);
- if (r < 0)
+ r = recvmsg(fd, &msg, MSG_TRUNC | (peek ? MSG_PEEK : 0));
+ if (r < 0) {
/* no data */
+ if (errno == ENOBUFS)
+ log_debug("rtnl: kernel receive buffer overrun");
+
return (errno == EAGAIN) ? 0 : -errno;
- else if (r == 0)
+ } else if (r == 0)
/* connection was closed by the kernel */
return -ECONNRESET;
- else
- len = (size_t)r;
for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
if (cmsg->cmsg_level == SOL_SOCKET &&
struct ucred *ucred = (void *)CMSG_DATA(cmsg);
/* from the kernel */
- if (ucred->uid == 0 && ucred->pid == 0) {
+ if (ucred->uid == 0 && ucred->pid == 0)
auth = true;
- break;
- }
+ } else if (cmsg->cmsg_level == SOL_NETLINK &&
+ cmsg->cmsg_type == NETLINK_PKTINFO &&
+ cmsg->cmsg_len == CMSG_LEN(sizeof(struct nl_pktinfo))) {
+ struct nl_pktinfo *pktinfo = (void *)CMSG_DATA(cmsg);
+
+ /* multi-cast group */
+ group = pktinfo->group;
}
}
/* not from the kernel, ignore */
return 0;
- for (new_msg = buffer; NLMSG_OK(new_msg, len); new_msg = NLMSG_NEXT(new_msg, len)) {
+ if (group)
+ *_group = group;
+
+ return r;
+}
+
+/* On success, the number of bytes received is returned and *ret points to the received message
+ * which has a valid header and the correct size.
+ * If nothing useful was received 0 is returned.
+ * On failure, a negative error code is returned.
+ */
+int socket_read_message(sd_rtnl *rtnl) {
+ _cleanup_rtnl_message_unref_ sd_rtnl_message *first = NULL;
+ struct iovec iov = {};
+ uint32_t group = 0;
+ bool multi_part = false, done = false;
+ struct nlmsghdr *new_msg;
+ size_t len;
+ int r;
+ unsigned i = 0;
+
+ assert(rtnl);
+ assert(rtnl->rbuffer);
+ assert(rtnl->rbuffer_allocated >= sizeof(struct nlmsghdr));
+
+ /* read nothing, just get the pending message size */
+ r = socket_recv_message(rtnl->fd, &iov, &group, true);
+ if (r <= 0)
+ return r;
+ else
+ len = (size_t)r;
+
+ /* make room for the pending message */
+ if (!greedy_realloc((void **)&rtnl->rbuffer,
+ &rtnl->rbuffer_allocated,
+ len, sizeof(uint8_t)))
+ return -ENOMEM;
+
+ iov.iov_base = rtnl->rbuffer;
+ iov.iov_len = rtnl->rbuffer_allocated;
+
+ /* read the pending message */
+ r = socket_recv_message(rtnl->fd, &iov, &group, false);
+ if (r <= 0)
+ return r;
+ else
+ len = (size_t)r;
+
+ if (len > rtnl->rbuffer_allocated)
+ /* message did not fit in read buffer */
+ return -EIO;
+
+ if (NLMSG_OK(rtnl->rbuffer, len) && rtnl->rbuffer->nlmsg_flags & NLM_F_MULTI) {
+ multi_part = true;
+
+ for (i = 0; i < rtnl->rqueue_partial_size; i++) {
+ if (rtnl_message_get_serial(rtnl->rqueue_partial[i]) ==
+ rtnl->rbuffer->nlmsg_seq) {
+ first = rtnl->rqueue_partial[i];
+ break;
+ }
+ }
+ }
+
+ for (new_msg = rtnl->rbuffer; NLMSG_OK(new_msg, len); new_msg = NLMSG_NEXT(new_msg, len)) {
_cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
const NLType *nl_type;
- if (new_msg->nlmsg_pid && new_msg->nlmsg_pid != rtnl->sockaddr.nl.nl_pid)
+ if (!group && new_msg->nlmsg_pid != rtnl->sockaddr.nl.nl_pid)
/* not broadcast and not for us */
continue;
- /* silently drop noop messages */
if (new_msg->nlmsg_type == NLMSG_NOOP)
+ /* silently drop noop messages */
continue;
- /* finished reading multi-part message */
- if (new_msg->nlmsg_type == NLMSG_DONE)
+ if (new_msg->nlmsg_type == NLMSG_DONE) {
+ /* finished reading multi-part message */
+ done = true;
break;
+ }
/* check that we support this message type */
r = type_system_get_type(NULL, &nl_type, new_msg->nlmsg_type);
if (r < 0)
return r;
- if (!first)
- first = m;
- else {
- assert(previous);
-
- previous->next = m;
- }
- previous = m;
+ /* push the message onto the multi-part message stack */
+ if (first)
+ m->next = first;
+ first = m;
m = NULL;
+ }
- ret += new_msg->nlmsg_len;
+ if (len)
+ log_debug("sd-rtnl: discarding %zu bytes of incoming message", len);
- /* not a multi-part message, so stop reading*/
- if (!(new_msg->nlmsg_flags & NLM_F_MULTI))
- break;
- }
+ if (!first)
+ return 0;
- r = rtnl_rqueue_make_room(rtnl);
- if (r < 0)
- return r;
+ if (!multi_part || done) {
+ /* we got a complete message, push it on the read queue */
+ r = rtnl_rqueue_make_room(rtnl);
+ if (r < 0)
+ return r;
+
+ rtnl->rqueue[rtnl->rqueue_size ++] = first;
+ first = NULL;
+
+ if (multi_part && (i < rtnl->rqueue_partial_size)) {
+ /* remove the message form the partial read queue */
+ memmove(rtnl->rqueue_partial + i,rtnl->rqueue_partial + i + 1,
+ sizeof(sd_rtnl_message*) * (rtnl->rqueue_partial_size - i - 1));
+ rtnl->rqueue_partial_size --;
+ }
- rtnl->rqueue[rtnl->rqueue_size ++] = first;
- first = NULL;
+ return 1;
+ } else {
+ /* we only got a partial multi-part message, push it on the
+ partial read queue */
+ if (i < rtnl->rqueue_partial_size) {
+ rtnl->rqueue_partial[i] = first;
+ } else {
+ r = rtnl_rqueue_partial_make_room(rtnl);
+ if (r < 0)
+ return r;
+
+ rtnl->rqueue_partial[rtnl->rqueue_partial_size ++] = first;
+ }
+ first = NULL;
- return ret;
+ return 0;
+ }
}
int sd_rtnl_message_rewind(sd_rtnl_message *m) {