#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;
-static int message_new_empty(sd_rtnl *rtnl, sd_rtnl_message **ret, size_t initial_size) {
+static int message_new_empty(sd_rtnl *rtnl, sd_rtnl_message **ret) {
sd_rtnl_message *m;
assert_return(ret, -EINVAL);
- assert_return(initial_size >= sizeof(struct nlmsghdr), -EINVAL);
/* Note that 'rtnl' is curretly unused, if we start using it internally
we must take care to avoid problems due to mutual references between
if (!m)
return -ENOMEM;
- m->hdr = malloc0(initial_size);
- if (!m->hdr) {
- free(m);
- return -ENOMEM;
- }
-
m->n_ref = REFCNT_INIT;
- m->hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
m->sealed = false;
*ret = m;
}
int message_new(sd_rtnl *rtnl, sd_rtnl_message **ret, uint16_t type) {
+ _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
const NLType *nl_type;
size_t size;
int r;
assert(nl_type->type == NLA_NESTED);
- size = NLMSG_SPACE(nl_type->size);
-
- r = message_new_empty(rtnl, ret, size);
+ r = message_new_empty(rtnl, &m);
if (r < 0)
return r;
- (*ret)->container_type_system[0] = nl_type->type_system;
- (*ret)->hdr->nlmsg_len = size;
- (*ret)->hdr->nlmsg_type = type;
+ size = NLMSG_SPACE(nl_type->size);
+
+ assert(size >= sizeof(struct nlmsghdr));
+ m->hdr = malloc0(size);
+ if (!m->hdr)
+ return -ENOMEM;
+
+ m->hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
+
+ m->container_type_system[0] = nl_type->type_system;
+ m->hdr->nlmsg_len = size;
+ m->hdr->nlmsg_type = type;
+
+ *ret = m;
+ m = NULL;
return 0;
}
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);
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) {
+ 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(family, -EINVAL);
+
+ ifa = NLMSG_DATA(m->hdr);
+
+ *family = ifa->ifa_family;
+
+ 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;
+
+ assert_return(m, -EINVAL);
+ assert_return(m->hdr, -EINVAL);
+ assert_return(rtnl_message_type_is_addr(m->hdr->nlmsg_type), -EINVAL);
+ assert_return(scope, -EINVAL);
+
+ ifa = NLMSG_DATA(m->hdr);
+
+ *scope = ifa->ifa_scope;
+
+ return 0;
+}
+
+int sd_rtnl_message_addr_get_flags(sd_rtnl_message *m, unsigned char *flags) {
+ 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(flags, -EINVAL);
+
+ ifa = NLMSG_DATA(m->hdr);
+
+ *flags = ifa->ifa_flags;
+
+ return 0;
+}
+
+int sd_rtnl_message_addr_get_ifindex(sd_rtnl_message *m, int *ifindex) {
+ 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(ifindex, -EINVAL);
+
+ ifa = NLMSG_DATA(m->hdr);
+
+ *ifindex = ifa->ifa_index;
+
+ return 0;
+}
+
int sd_rtnl_message_new_addr(sd_rtnl *rtnl, sd_rtnl_message **ret,
uint16_t nlmsg_type, int index,
unsigned char family) {
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);
return 0;
}
+int sd_rtnl_message_new_addr_update(sd_rtnl *rtnl, sd_rtnl_message **ret,
+ int index, unsigned char family) {
+ int r;
+
+ r = sd_rtnl_message_new_addr(rtnl, ret, RTM_NEWADDR, index, family);
+ if (r < 0)
+ return r;
+
+ (*ret)->hdr->nlmsg_flags |= NLM_F_REPLACE;
+
+ return 0;
+}
+
sd_rtnl_message *sd_rtnl_message_ref(sd_rtnl_message *m) {
if (m)
assert_se(REFCNT_INC(m->n_ref) >= 2);
for (i = 0; i <= m->n_containers; i++)
free(m->rta_offset_tb[i]);
+ sd_rtnl_message_unref(m->next);
+
free(m);
}
/* 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);
+ data_length = 0;
}
+ /* 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) {
return 0;
}
+int sd_rtnl_message_append_cache_info(sd_rtnl_message *m, unsigned short type, const struct ifa_cacheinfo *info) {
+ int r;
+
+ assert_return(m, -EINVAL);
+ assert_return(!m->sealed, -EPERM);
+ assert_return(info, -EINVAL);
+
+ r = message_attribute_has_type(m, type, NLA_CACHE_INFO);
+ if (r < 0)
+ return r;
+
+ r = add_rtattr(m, type, info, sizeof(struct ifa_cacheinfo));
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
int sd_rtnl_message_open_container(sd_rtnl_message *m, unsigned short type) {
size_t size;
int r;
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;
+ m->container_offsets[m->n_containers ++] = r;
+
return 0;
}
if (r < 0)
return r;
+ m->container_offsets[m->n_containers ++] = r;
+
return 0;
}
return 0;
}
+int sd_rtnl_message_read_cache_info(sd_rtnl_message *m, unsigned short type, struct ifa_cacheinfo *info) {
+ int r;
+ void *attr_data;
+
+ r = message_attribute_has_type(m, type, NLA_CACHE_INFO);
+ if (r < 0)
+ return r;
+
+ r = rtnl_message_read_internal(m, type, &attr_data);
+ if (r < 0)
+ return r;
+ else if ((size_t)r < sizeof(struct ifa_cacheinfo))
+ return -EIO;
+
+ memcpy(info, attr_data, sizeof(struct ifa_cacheinfo));
+
+ return 0;
+}
+
int sd_rtnl_message_read_in_addr(sd_rtnl_message *m, unsigned short type, struct in_addr *data) {
int r;
void *attr_data;
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,
unsigned short type;
size_t *tb;
- tb = (size_t *) new0(size_t *, max);
+ tb = new0(size_t, max + 1);
if(!tb)
return -ENOMEM;
- *rta_tb_size = max;
+ *rta_tb_size = max + 1;
for (; RTA_OK(rta, rt_len); rta = RTA_NEXT(rta, rt_len)) {
type = rta->rta_type;
- if (type > max) {
- log_debug("rtnl: message parse - ignore out of range attribute type");
+ /* if the kernel is newer than the headers we used
+ when building, we ignore out-of-range attributes
+ */
+ if (type > max)
continue;
- }
if (tb[type])
log_debug("rtnl: message parse - overwriting repeated attribute");
* If nothing useful was received 0 is returned.
* On failure, a negative error code is returned.
*/
-int socket_read_message(sd_rtnl *nl, sd_rtnl_message **ret) {
- _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
- const NLType *nl_type;
- struct nlmsghdr *new_hdr;
- union {
- struct sockaddr sa;
- struct sockaddr_nl nl;
- } addr;
- socklen_t addr_len;
- size_t need, len;
+int socket_read_message(sd_rtnl *rtnl) {
+ _cleanup_rtnl_message_unref_ sd_rtnl_message *first = NULL;
+ uint8_t cred_buffer[CMSG_SPACE(sizeof(struct ucred))];
+ struct iovec iov = {};
+ struct msghdr msg = {
+ .msg_iov = &iov,
+ .msg_iovlen = 1,
+ .msg_control = cred_buffer,
+ .msg_controllen = sizeof(cred_buffer),
+ };
+ struct cmsghdr *cmsg;
+ bool auth = false, multi_part = false, done = false;
+ struct nlmsghdr *new_msg;
+ size_t len;
int r;
+ unsigned i = 0;
- assert(nl);
- assert(ret);
+ assert(rtnl);
+ assert(rtnl->rbuffer);
+ assert(rtnl->rbuffer_allocated >= sizeof(struct nlmsghdr));
- r = message_receive_need(nl, &need);
+ /* read nothing, just get the pending message size */
+ r = recvmsg(rtnl->fd, &msg, MSG_PEEK | MSG_TRUNC);
if (r < 0)
- return r;
+ /* no data */
+ return (errno == EAGAIN) ? 0 : -errno;
+ else if (r == 0)
+ /* connection was closed by the kernel */
+ return -ECONNRESET;
+ else
+ len = (size_t)r;
- r = message_new_empty(nl, &m, need);
- if (r < 0)
- return r;
+ /* make room for the pending message */
+ if (!greedy_realloc((void **)&rtnl->rbuffer,
+ &rtnl->rbuffer_allocated,
+ len, sizeof(uint8_t)))
+ return -ENOMEM;
- addr_len = sizeof(addr);
+ iov.iov_base = rtnl->rbuffer;
+ iov.iov_len = rtnl->rbuffer_allocated;
- r = recvfrom(nl->fd, m->hdr, need,
- 0, &addr.sa, &addr_len);
+ r = recvmsg(rtnl->fd, &msg, MSG_TRUNC);
if (r < 0)
- return (errno == EAGAIN) ? 0 : -errno; /* no data */
+ /* no data */
+ return (errno == EAGAIN) ? 0 : -errno;
else if (r == 0)
- return -ECONNRESET; /* connection was closed by the kernel */
- else if (addr_len != sizeof(addr.nl) ||
- addr.nl.nl_family != AF_NETLINK)
- return -EIO; /* not a netlink message */
- else if (addr.nl.nl_pid != 0)
- return 0; /* not from the kernel */
- else if ((size_t) r < sizeof(struct nlmsghdr) ||
- (size_t) r < m->hdr->nlmsg_len)
- return -EIO; /* too small (we do accept too big though) */
- else if (m->hdr->nlmsg_pid && m->hdr->nlmsg_pid != nl->sockaddr.nl.nl_pid)
- return 0; /* not broadcast and not for us */
+ /* connection was closed by the kernel */
+ return -ECONNRESET;
else
- len = (size_t) r;
+ len = (size_t)r;
- /* silently drop noop messages */
- if (m->hdr->nlmsg_type == NLMSG_NOOP)
- return 0;
+ if (len > rtnl->rbuffer_allocated)
+ /* message did not fit in read buffer */
+ return -EIO;
- /* check that we support this message type */
- r = type_system_get_type(NULL, &nl_type, m->hdr->nlmsg_type);
- if (r < 0) {
- if (r == -ENOTSUP)
- log_debug("sd-rtnl: ignored message with unknown type: %u",
- m->hdr->nlmsg_type);
+ for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
+ if (cmsg->cmsg_level == SOL_SOCKET &&
+ cmsg->cmsg_type == SCM_CREDENTIALS &&
+ cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) {
+ struct ucred *ucred = (void *)CMSG_DATA(cmsg);
+
+ /* from the kernel */
+ if (ucred->uid == 0 && ucred->pid == 0) {
+ auth = true;
+ break;
+ }
+ }
+ }
+ if (!auth)
+ /* not from the kernel, ignore */
return 0;
+
+ 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;
+ }
+ }
}
- /* check that the size matches the message type */
- if (len < NLMSG_LENGTH(nl_type->size))
- return -EIO;
+ 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;
- /* we probably allocated way too much memory, give it back */
- new_hdr = realloc(m->hdr, len);
- if (!new_hdr)
- return -ENOMEM;
- m->hdr = new_hdr;
+ if (new_msg->nlmsg_pid && new_msg->nlmsg_pid != rtnl->sockaddr.nl.nl_pid)
+ /* not broadcast and not for us */
+ continue;
- /* seal and parse the top-level message */
- r = sd_rtnl_message_rewind(m);
- if (r < 0)
- return r;
+ if (new_msg->nlmsg_type == NLMSG_NOOP)
+ /* silently drop noop messages */
+ continue;
- *ret = m;
- m = NULL;
+ 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) {
+ if (r == -ENOTSUP)
+ log_debug("sd-rtnl: ignored message with unknown type: %u",
+ new_msg->nlmsg_type);
+
+ continue;
+ }
+
+ /* check that the size matches the message type */
+ if (new_msg->nlmsg_len < NLMSG_LENGTH(nl_type->size))
+ continue;
- return len;
+ r = message_new_empty(rtnl, &m);
+ if (r < 0)
+ return r;
+
+ m->hdr = memdup(new_msg, new_msg->nlmsg_len);
+ if (!m->hdr)
+ return -ENOMEM;
+
+ /* seal and parse the top-level message */
+ r = sd_rtnl_message_rewind(m);
+ if (r < 0)
+ return r;
+
+ /* push the message onto the multi-part message stack */
+ if (first)
+ m->next = first;
+ first = m;
+ m = NULL;
+ }
+
+ if (len)
+ log_debug("sd-rtnl: discarding %zu bytes of incoming message", len);
+
+ if (!first)
+ return 0;
+
+ 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 --;
+ }
+
+ 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 0;
+ }
}
int sd_rtnl_message_rewind(sd_rtnl_message *m) {
&m->rta_offset_tb[m->n_containers],
&m->rta_tb_size[m->n_containers],
type_system->max,
- (char*)NLMSG_DATA(m->hdr) + NLMSG_ALIGN(type->size),
+ (struct rtattr*)((uint8_t*)NLMSG_DATA(m->hdr) +
+ NLMSG_ALIGN(type->size)),
NLMSG_PAYLOAD(m->hdr, type->size));
if (r < 0)
return r;
m->sealed = true;
}
+
+sd_rtnl_message *sd_rtnl_message_next(sd_rtnl_message *m) {
+ assert_return(m, NULL);
+
+ return m->next;
+}