static void
sigchld_handler(int signr OVS_UNUSED)
{
struct process *p;
COVERAGE_INC(process_sigchld);
LIST_FOR_EACH (p, struct process, node, &all_processes) {
if (!p->exited) {
int retval, status;
do {
retval = waitpid(p->pid, &status, WNOHANG);
} while (retval == -1 && errno == EINTR);
if (retval == p->pid) {
p->exited = true;
p->status = status;
} else if (retval < 0) {
/* XXX We want to log something but we're in a signal
* handler. */
p->exited = true;
p->status = -1;
}
}
}
ignore(write(fds[1], "", 1));
}
static int
nl_sock_send__(struct nl_sock *sock, const struct ofpbuf *msg,
uint32_t nlmsg_seq, bool wait)
{
struct nlmsghdr *nlmsg = nl_msg_nlmsghdr(msg);
int error;
nlmsg->nlmsg_len = ofpbuf_size(msg);
nlmsg->nlmsg_seq = nlmsg_seq;
nlmsg->nlmsg_pid = sock->pid;
do {
int retval;
#ifdef _WIN32
bool result;
DWORD last_error = 0;
result = WriteFile(sock->handle, ofpbuf_data(msg), ofpbuf_size(msg),
&retval, NULL);
last_error = GetLastError();
if (last_error != ERROR_SUCCESS && !result) {
retval = -1;
errno = EAGAIN;
}
#else
retval = send(sock->fd, ofpbuf_data(msg), ofpbuf_size(msg), wait ? 0 : MSG_DONTWAIT);
#endif
error = retval < 0 ? errno : 0;
} while (error == EINTR);
log_nlmsg(__func__, error, ofpbuf_data(msg), ofpbuf_size(msg), sock->protocol);
if (!error) {
COVERAGE_INC(netlink_sent);
}
return error;
}
static void
resize(struct hmap *hmap, size_t new_mask, const char *where)
{
struct hmap tmp;
size_t i;
ovs_assert(is_pow2(new_mask + 1));
hmap_init(&tmp);
if (new_mask) {
tmp.buckets = xmalloc(sizeof *tmp.buckets * (new_mask + 1));
tmp.mask = new_mask;
for (i = 0; i <= tmp.mask; i++) {
tmp.buckets[i] = NULL;
}
}
for (i = 0; i <= hmap->mask; i++) {
struct hmap_node *node, *next;
int count = 0;
for (node = hmap->buckets[i]; node; node = next) {
next = node->next;
hmap_insert_fast(&tmp, node, node->hash);
count++;
}
if (count > 5) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10);
COVERAGE_INC(hmap_pathological);
VLOG_DBG_RL(&rl, "%s: %d nodes in bucket (%"PRIuSIZE" nodes, %"PRIuSIZE" buckets)",
where, count, hmap->n, hmap->mask + 1);
}
}
hmap_swap(hmap, &tmp);
hmap_destroy(&tmp);
}
int
process_run(char **argv,
const int keep_fds[], size_t n_keep_fds,
const int null_fds[], size_t n_null_fds,
int *status)
{
struct process *p;
int retval;
COVERAGE_INC(process_run);
retval = process_start(argv, keep_fds, n_keep_fds, null_fds, n_null_fds,
&p);
if (retval) {
*status = 0;
return retval;
}
while (!process_exited(p)) {
process_wait(p);
poll_block();
}
*status = process_status(p);
process_destroy(p);
return 0;
}
/* Attempts to make 'ml' learn from the fact that a frame from 'src_mac' was
* just observed arriving from 'src_port' on the given 'vlan'.
*
* Returns nonzero if we actually learned something from this, zero if it just
* confirms what we already knew. The nonzero return value is the tag of flows
* that now need revalidation.
*
* The 'vlan' parameter is used to maintain separate per-VLAN learning tables.
* Specify 0 if this behavior is undesirable.
*
* 'lock_type' specifies whether the entry should be locked or existing locks
* are check. */
tag_type
mac_learning_learn(struct mac_learning *ml,
const uint8_t src_mac[ETH_ADDR_LEN], uint16_t vlan,
uint16_t src_port, enum grat_arp_lock_type lock_type)
{
struct mac_entry *e;
struct list *bucket;
if (!is_learning_vlan(ml, vlan)) {
return 0;
}
if (eth_addr_is_multicast(src_mac)) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 30);
VLOG_DBG_RL(&rl, "multicast packet source "ETH_ADDR_FMT,
ETH_ADDR_ARGS(src_mac));
return 0;
}
bucket = mac_table_bucket(ml, src_mac, vlan);
e = search_bucket(bucket, src_mac, vlan);
if (!e) {
if (!list_is_empty(&ml->free)) {
e = mac_entry_from_lru_node(ml->free.next);
} else {
e = mac_entry_from_lru_node(ml->lrus.next);
list_remove(&e->hash_node);
}
memcpy(e->mac, src_mac, ETH_ADDR_LEN);
list_push_front(bucket, &e->hash_node);
e->port = -1;
e->vlan = vlan;
e->tag = make_unknown_mac_tag(ml, src_mac, vlan);
e->grat_arp_lock = TIME_MIN;
}
if (lock_type != GRAT_ARP_LOCK_CHECK || time_now() >= e->grat_arp_lock) {
/* Make the entry most-recently-used. */
list_remove(&e->lru_node);
list_push_back(&ml->lrus, &e->lru_node);
e->expires = time_now() + MAC_ENTRY_IDLE_TIME;
if (lock_type == GRAT_ARP_LOCK_SET) {
e->grat_arp_lock = time_now() + MAC_GRAT_ARP_LOCK_TIME;
}
/* Did we learn something? */
if (e->port != src_port) {
tag_type old_tag = e->tag;
e->port = src_port;
e->tag = tag_create_random();
COVERAGE_INC(mac_learning_learned);
return old_tag;
}
}
return 0;
}
/* Shrinks 'hindex', if necessary, to optimize the performance of iteration. */
void
hindex_shrink(struct hindex *hindex)
{
size_t new_mask = hindex_calc_mask(hindex->n_unique);
if (new_mask < hindex->mask) {
COVERAGE_INC(hindex_shrink);
hindex_resize(hindex, new_mask);
}
}
/* Expands 'hmap', if necessary, to optimize the performance of searches when
* it has up to 'n' elements. (But iteration will be slow in a hash map whose
* allocated capacity is much higher than its current number of nodes.)
*
* ('where' is used in debug logging. Commonly one would use hmap_reserve() to
* automatically provide the caller's source file and line number for
* 'where'.) */
void
hmap_reserve_at(struct hmap *hmap, size_t n, const char *where)
{
size_t new_mask = calc_mask(n);
if (new_mask > hmap->mask) {
COVERAGE_INC(hmap_reserve);
resize(hmap, new_mask, where);
}
}
/* Shrinks 'hmap', if necessary, to optimize the performance of iteration.
*
* ('where' is used in debug logging. Commonly one would use hmap_shrink() to
* automatically provide the caller's source file and line number for
* 'where'.) */
void
hmap_shrink_at(struct hmap *hmap, const char *where)
{
size_t new_mask = calc_mask(hmap->n);
if (new_mask < hmap->mask) {
COVERAGE_INC(hmap_shrink);
resize(hmap, new_mask, where);
}
}
/* Expands 'hindex', if necessary, to optimize the performance of searches when
* it has up to 'n' unique hashes. (But iteration will be slow in a hash index
* whose allocated capacity is much higher than its current number of
* nodes.) */
void
hindex_reserve(struct hindex *hindex, size_t n)
{
size_t new_mask = hindex_calc_mask(n);
if (new_mask > hindex->mask) {
COVERAGE_INC(hindex_reserve);
hindex_resize(hindex, new_mask);
}
}
/* Expands 'hindex', if necessary, to optimize the performance of searches. */
void
hindex_expand(struct hindex *hindex)
{
size_t new_mask = hindex_calc_mask(hindex->n_unique);
if (new_mask > hindex->mask) {
COVERAGE_INC(hindex_expand);
hindex_resize(hindex, new_mask);
}
}
void *
xmalloc(size_t size)
{
void *p = malloc(size ? size : 1);
COVERAGE_INC(util_xalloc);
if (p == NULL) {
out_of_memory();
}
return p;
}
void *
xcalloc(size_t count, size_t size)
{
void *p = count && size ? calloc(count, size) : malloc(1);
COVERAGE_INC(util_xalloc);
if (p == NULL) {
out_of_memory();
}
return p;
}
void *
xrealloc(void *p, size_t size)
{
p = realloc(p, size ? size : 1);
COVERAGE_INC(util_xalloc);
if (p == NULL) {
out_of_memory();
}
return p;
}
static uint32_t *
miniflow_alloc_values(struct miniflow *flow, int n)
{
if (n <= MINI_N_INLINE) {
return flow->inline_values;
} else {
COVERAGE_INC(miniflow_malloc);
return xmalloc(n * sizeof *flow->values);
}
}
void
mac_learning_run(struct mac_learning *ml, struct tag_set *set)
{
struct mac_entry *e;
while (get_lru(ml, &e) && time_now() >= e->expires) {
COVERAGE_INC(mac_learning_expired);
if (set) {
tag_set_add(set, e->tag);
}
free_mac_entry(ml, e);
}
}
/* Starts a subprocess with the arguments in the null-terminated argv[] array.
* argv[0] is used as the name of the process. Searches the PATH environment
* variable to find the program to execute.
*
* All file descriptors are closed before executing the subprocess, except for
* fds 0, 1, and 2 and the 'n_keep_fds' fds listed in 'keep_fds'. Also, any of
* the 'n_null_fds' fds listed in 'null_fds' are replaced by /dev/null.
*
* Returns 0 if successful, otherwise a positive errno value indicating the
* error. If successful, '*pp' is assigned a new struct process that may be
* used to query the process's status. On failure, '*pp' is set to NULL. */
int
process_start(char **argv,
const int keep_fds[], size_t n_keep_fds,
const int null_fds[], size_t n_null_fds,
struct process **pp)
{
sigset_t oldsigs;
pid_t pid;
int error;
*pp = NULL;
COVERAGE_INC(process_start);
error = process_prestart(argv);
if (error) {
return error;
}
block_sigchld(&oldsigs);
pid = fork();
if (pid < 0) {
unblock_sigchld(&oldsigs);
VLOG_WARN("fork failed: %s", strerror(errno));
return errno;
} else if (pid) {
/* Running in parent process. */
*pp = process_register(argv[0], pid);
unblock_sigchld(&oldsigs);
return 0;
} else {
/* Running in child process. */
int fd_max = get_max_fds();
int fd;
fatal_signal_fork();
unblock_sigchld(&oldsigs);
for (fd = 0; fd < fd_max; fd++) {
if (is_member(fd, null_fds, n_null_fds)) {
/* We can't use get_null_fd() here because we might have
* already closed its fd. */
int nullfd = open("/dev/null", O_RDWR);
dup2(nullfd, fd);
close(nullfd);
} else if (fd >= 3 && !is_member(fd, keep_fds, n_keep_fds)) {
close(fd);
}
}
execvp(argv[0], argv);
fprintf(stderr, "execvp(\"%s\") failed: %s\n",
argv[0], strerror(errno));
_exit(1);
}
}
/* Initializes 'flow' members from 'packet', 'skb_priority', 'tnl', and
* 'ofp_in_port'.
*
* Initializes 'packet' header pointers as follows:
*
* - packet->l2 to the start of the Ethernet header.
*
* - packet->l2_5 to the start of the MPLS shim header.
*
* - packet->l3 to just past the Ethernet header, or just past the
* vlan_header if one is present, to the first byte of the payload of the
* Ethernet frame.
*
* - packet->l4 to just past the IPv4 header, if one is present and has a
* correct length, and otherwise NULL.
*
* - packet->l7 to just past the TCP or UDP or ICMP header, if one is
* present and has a correct length, and otherwise NULL.
*/
void
flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t skb_mark,
const struct flow_tnl *tnl, uint16_t ofp_in_port,
struct flow *flow)
{
struct ofpbuf b = *packet;
struct eth_header *eth;
COVERAGE_INC(flow_extract);
memset(flow, 0, sizeof *flow);
if (tnl) {
ovs_assert(tnl != &flow->tunnel);
flow->tunnel = *tnl;
}
flow->in_port = ofp_in_port;
flow->skb_priority = skb_priority;
flow->skb_mark = skb_mark;
packet->l2 = b.data;
packet->l2_5 = NULL;
packet->l3 = NULL;
packet->l4 = NULL;
packet->l7 = NULL;
if (b.size < sizeof *eth) {
return;
}
/* Link layer. */
eth = b.data;
memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
/* dl_type, vlan_tci. */
ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
parse_vlan(&b, flow);
}
flow->dl_type = parse_ethertype(&b);
/* Parse mpls, copy l3 ttl. */
if (eth_type_mpls(flow->dl_type)) {
packet->l2_5 = b.data;
parse_mpls(&b, flow);
}
packet->l3 = b.data;
flow_extract_l3_onwards(packet, flow, flow->dl_type);
}
static int
nl_sock_send__(struct nl_sock *sock, const struct ofpbuf *msg,
uint32_t nlmsg_seq, bool wait)
{
struct nlmsghdr *nlmsg = nl_msg_nlmsghdr(msg);
int error;
nlmsg->nlmsg_len = msg->size;
nlmsg->nlmsg_seq = nlmsg_seq;
nlmsg->nlmsg_pid = sock->pid;
do {
int retval;
retval = send(sock->fd, msg->data, msg->size, wait ? 0 : MSG_DONTWAIT);
error = retval < 0 ? errno : 0;
} while (error == EINTR);
log_nlmsg(__func__, error, msg->data, msg->size, sock->protocol);
if (!error) {
COVERAGE_INC(netlink_sent);
}
return error;
}
/* Reallocates 'hindex''s array of buckets to use bitwise mask 'new_mask'. */
static void
hindex_resize(struct hindex *hindex, size_t new_mask)
{
struct hindex tmp;
size_t i;
ovs_assert(is_pow2(new_mask + 1));
ovs_assert(new_mask != SIZE_MAX);
hindex_init(&tmp);
if (new_mask) {
tmp.buckets = xmalloc(sizeof *tmp.buckets * (new_mask + 1));
tmp.mask = new_mask;
for (i = 0; i <= tmp.mask; i++) {
tmp.buckets[i] = NULL;
}
}
for (i = 0; i <= hindex->mask; i++) {
struct hindex_node *node, *next;
int count;
count = 0;
for (node = hindex->buckets[i]; node; node = next) {
struct hindex_node **head = &tmp.buckets[node->hash & tmp.mask];
next = node->d;
node->d = *head;
*head = node;
count++;
}
if (count > 5) {
COVERAGE_INC(hindex_pathological);
}
}
tmp.n_unique = hindex->n_unique;
hindex_swap(hindex, &tmp);
hindex_destroy(&tmp);
}
/* Starts the process whose arguments are given in the null-terminated array
* 'argv' and waits for it to exit. On success returns 0 and stores the
* process exit value (suitable for passing to process_status_msg()) in
* '*status'. On failure, returns a positive errno value and stores 0 in
* '*status'.
*
* If 'stdout_log' is nonnull, then the subprocess's output to stdout (up to a
* limit of PROCESS_MAX_CAPTURE bytes) is captured in a memory buffer, which
* when this function returns 0 is stored as a null-terminated string in
* '*stdout_log'. The caller is responsible for freeing '*stdout_log' (by
* passing it to free()). When this function returns an error, '*stdout_log'
* is set to NULL.
*
* If 'stderr_log' is nonnull, then it is treated like 'stdout_log' except
* that it captures the subprocess's output to stderr. */
int
process_run_capture(char **argv, char **stdout_log, char **stderr_log,
int *status)
{
struct stream s_stdout, s_stderr;
sigset_t oldsigs;
pid_t pid;
int error;
COVERAGE_INC(process_run_capture);
if (stdout_log) {
*stdout_log = NULL;
}
if (stderr_log) {
*stderr_log = NULL;
}
*status = 0;
error = process_prestart(argv);
if (error) {
return error;
}
error = stream_open(&s_stdout);
if (error) {
return error;
}
error = stream_open(&s_stderr);
if (error) {
stream_close(&s_stdout);
return error;
}
block_sigchld(&oldsigs);
pid = fork();
if (pid < 0) {
int error = errno;
unblock_sigchld(&oldsigs);
VLOG_WARN("fork failed: %s", strerror(error));
stream_close(&s_stdout);
stream_close(&s_stderr);
*status = 0;
return error;
} else if (pid) {
/* Running in parent process. */
struct process *p;
p = process_register(argv[0], pid);
unblock_sigchld(&oldsigs);
close(s_stdout.fds[1]);
close(s_stderr.fds[1]);
while (!process_exited(p)) {
stream_read(&s_stdout);
stream_read(&s_stderr);
stream_wait(&s_stdout);
stream_wait(&s_stderr);
process_wait(p);
poll_block();
}
stream_read(&s_stdout);
stream_read(&s_stderr);
if (stdout_log) {
*stdout_log = ds_steal_cstr(&s_stdout.log);
}
if (stderr_log) {
*stderr_log = ds_steal_cstr(&s_stderr.log);
}
stream_close(&s_stdout);
stream_close(&s_stderr);
*status = process_status(p);
process_destroy(p);
return 0;
} else {
/* Running in child process. */
int max_fds;
int i;
fatal_signal_fork();
unblock_sigchld(&oldsigs);
dup2(get_null_fd(), 0);
dup2(s_stdout.fds[1], 1);
dup2(s_stderr.fds[1], 2);
max_fds = get_max_fds();
for (i = 3; i < max_fds; i++) {
close(i);
}
execvp(argv[0], argv);
fprintf(stderr, "execvp(\"%s\") failed: %s\n",
argv[0], strerror(errno));
exit(EXIT_FAILURE);
}
}
/* Initializes 'flow' members from 'packet', 'tun_id', and 'ofp_in_port'.
* Initializes 'packet' header pointers as follows:
*
* - packet->l2 to the start of the Ethernet header.
*
* - packet->l3 to just past the Ethernet header, or just past the
* vlan_header if one is present, to the first byte of the payload of the
* Ethernet frame.
*
* - packet->l4 to just past the IPv4 header, if one is present and has a
* correct length, and otherwise NULL.
*
* - packet->l7 to just past the TCP or UDP or ICMP header, if one is
* present and has a correct length, and otherwise NULL.
*/
void
flow_extract(struct ofpbuf *packet, uint32_t priority, ovs_be64 tun_id,
uint16_t ofp_in_port, struct flow *flow)
{
struct ofpbuf b = *packet;
struct eth_header *eth;
COVERAGE_INC(flow_extract);
memset(flow, 0, sizeof *flow);
flow->tun_id = tun_id;
flow->in_port = ofp_in_port;
flow->priority = priority;
packet->l2 = b.data;
packet->l3 = NULL;
packet->l4 = NULL;
packet->l7 = NULL;
if (b.size < sizeof *eth) {
return;
}
/* Link layer. */
eth = b.data;
memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
/* dl_type, vlan_tci. */
ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
parse_vlan(&b, flow);
}
flow->dl_type = parse_ethertype(&b);
/* Network layer. */
packet->l3 = b.data;
if (flow->dl_type == htons(ETH_TYPE_IP)) {
const struct ip_header *nh = pull_ip(&b);
if (nh) {
packet->l4 = b.data;
flow->nw_src = get_unaligned_be32(&nh->ip_src);
flow->nw_dst = get_unaligned_be32(&nh->ip_dst);
flow->nw_proto = nh->ip_proto;
flow->nw_tos = nh->ip_tos;
if (IP_IS_FRAGMENT(nh->ip_frag_off)) {
flow->nw_frag = FLOW_NW_FRAG_ANY;
if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) {
flow->nw_frag |= FLOW_NW_FRAG_LATER;
}
}
flow->nw_ttl = nh->ip_ttl;
if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) {
if (flow->nw_proto == IPPROTO_TCP) {
parse_tcp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_UDP) {
parse_udp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_ICMP) {
const struct icmp_header *icmp = pull_icmp(&b);
if (icmp) {
flow->tp_src = htons(icmp->icmp_type);
flow->tp_dst = htons(icmp->icmp_code);
packet->l7 = b.data;
}
}
}
}
} else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
if (parse_ipv6(&b, flow)) {
return;
}
packet->l4 = b.data;
if (flow->nw_proto == IPPROTO_TCP) {
parse_tcp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_UDP) {
parse_udp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_ICMPV6) {
if (parse_icmpv6(&b, flow)) {
packet->l7 = b.data;
}
}
} else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
const struct arp_eth_header *arp = pull_arp(&b);
if (arp && arp->ar_hrd == htons(1)
&& arp->ar_pro == htons(ETH_TYPE_IP)
&& arp->ar_hln == ETH_ADDR_LEN
&& arp->ar_pln == 4) {
/* We only match on the lower 8 bits of the opcode. */
if (ntohs(arp->ar_op) <= 0xff) {
flow->nw_proto = ntohs(arp->ar_op);
}
if ((flow->nw_proto == ARP_OP_REQUEST)
|| (flow->nw_proto == ARP_OP_REPLY)) {
flow->nw_src = arp->ar_spa;
flow->nw_dst = arp->ar_tpa;
memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
}
}
}
}
/* Initializes 'flow' members from 'packet', 'skb_priority', 'tnl', and
* 'in_port'.
*
* Initializes 'packet' header pointers as follows:
*
* - packet->l2 to the start of the Ethernet header.
*
* - packet->l2_5 to the start of the MPLS shim header.
*
* - packet->l3 to just past the Ethernet header, or just past the
* vlan_header if one is present, to the first byte of the payload of the
* Ethernet frame.
*
* - packet->l4 to just past the IPv4 header, if one is present and has a
* correct length, and otherwise NULL.
*
* - packet->l7 to just past the TCP/UDP/SCTP/ICMP header, if one is
* present and has a correct length, and otherwise NULL.
*/
void
flow_extract(struct ofpbuf *packet, uint32_t skb_priority, uint32_t pkt_mark,
const struct flow_tnl *tnl, const union flow_in_port *in_port,
struct flow *flow)
{
struct ofpbuf b = *packet;
struct eth_header *eth;
COVERAGE_INC(flow_extract);
memset(flow, 0, sizeof *flow);
if (tnl) {
ovs_assert(tnl != &flow->tunnel);
flow->tunnel = *tnl;
}
if (in_port) {
flow->in_port = *in_port;
}
flow->skb_priority = skb_priority;
flow->pkt_mark = pkt_mark;
packet->l2 = b.data;
packet->l2_5 = NULL;
packet->l3 = NULL;
packet->l4 = NULL;
packet->l7 = NULL;
if (b.size < sizeof *eth) {
return;
}
/* Link layer. */
eth = b.data;
memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
/* dl_type, vlan_tci. */
ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
parse_vlan(&b, flow);
}
flow->dl_type = parse_ethertype(&b);
/* Parse mpls, copy l3 ttl. */
if (eth_type_mpls(flow->dl_type)) {
packet->l2_5 = b.data;
parse_mpls(&b, flow);
}
/* Network layer. */
packet->l3 = b.data;
if (flow->dl_type == htons(ETH_TYPE_IP)) {
const struct ip_header *nh = pull_ip(&b);
if (nh) {
packet->l4 = b.data;
flow->nw_src = get_16aligned_be32(&nh->ip_src);
flow->nw_dst = get_16aligned_be32(&nh->ip_dst);
flow->nw_proto = nh->ip_proto;
flow->nw_tos = nh->ip_tos;
if (IP_IS_FRAGMENT(nh->ip_frag_off)) {
flow->nw_frag = FLOW_NW_FRAG_ANY;
if (nh->ip_frag_off & htons(IP_FRAG_OFF_MASK)) {
flow->nw_frag |= FLOW_NW_FRAG_LATER;
}
}
flow->nw_ttl = nh->ip_ttl;
if (!(nh->ip_frag_off & htons(IP_FRAG_OFF_MASK))) {
if (flow->nw_proto == IPPROTO_TCP) {
parse_tcp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_UDP) {
parse_udp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_SCTP) {
parse_sctp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_ICMP) {
const struct icmp_header *icmp = pull_icmp(&b);
if (icmp) {
flow->tp_src = htons(icmp->icmp_type);
flow->tp_dst = htons(icmp->icmp_code);
packet->l7 = b.data;
}
}
}
}
} else if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
if (parse_ipv6(&b, flow)) {
return;
}
packet->l4 = b.data;
if (flow->nw_proto == IPPROTO_TCP) {
parse_tcp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_UDP) {
parse_udp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_SCTP) {
parse_sctp(packet, &b, flow);
} else if (flow->nw_proto == IPPROTO_ICMPV6) {
if (parse_icmpv6(&b, flow)) {
packet->l7 = b.data;
}
}
} else if (flow->dl_type == htons(ETH_TYPE_ARP) ||
flow->dl_type == htons(ETH_TYPE_RARP)) {
const struct arp_eth_header *arp = pull_arp(&b);
if (arp && arp->ar_hrd == htons(1)
&& arp->ar_pro == htons(ETH_TYPE_IP)
&& arp->ar_hln == ETH_ADDR_LEN
&& arp->ar_pln == 4) {
/* We only match on the lower 8 bits of the opcode. */
if (ntohs(arp->ar_op) <= 0xff) {
flow->nw_proto = ntohs(arp->ar_op);
}
flow->nw_src = get_16aligned_be32(&arp->ar_spa);
flow->nw_dst = get_16aligned_be32(&arp->ar_tpa);
memcpy(flow->arp_sha, arp->ar_sha, ETH_ADDR_LEN);
memcpy(flow->arp_tha, arp->ar_tha, ETH_ADDR_LEN);
}
}
}
static int
nl_sock_recv__(struct nl_sock *sock, struct ofpbuf *buf, bool wait)
{
/* We can't accurately predict the size of the data to be received. The
* caller is supposed to have allocated enough space in 'buf' to handle the
* "typical" case. To handle exceptions, we make available enough space in
* 'tail' to allow Netlink messages to be up to 64 kB long (a reasonable
* figure since that's the maximum length of a Netlink attribute). */
struct nlmsghdr *nlmsghdr;
#ifdef _WIN32
#define MAX_STACK_LENGTH 81920
uint8_t tail[MAX_STACK_LENGTH];
#else
uint8_t tail[65536];
#endif
struct iovec iov[2];
struct msghdr msg;
ssize_t retval;
int error;
ovs_assert(buf->allocated >= sizeof *nlmsghdr);
ofpbuf_clear(buf);
iov[0].iov_base = ofpbuf_base(buf);
iov[0].iov_len = buf->allocated;
iov[1].iov_base = tail;
iov[1].iov_len = sizeof tail;
memset(&msg, 0, sizeof msg);
msg.msg_iov = iov;
msg.msg_iovlen = 2;
/* Receive a Netlink message from the kernel.
*
* This works around a kernel bug in which the kernel returns an error code
* as if it were the number of bytes read. It doesn't actually modify
* anything in the receive buffer in that case, so we can initialize the
* Netlink header with an impossible message length and then, upon success,
* check whether it changed. */
nlmsghdr = ofpbuf_base(buf);
do {
nlmsghdr->nlmsg_len = UINT32_MAX;
#ifdef _WIN32
boolean result = false;
DWORD last_error = 0;
result = ReadFile(sock->handle, tail, MAX_STACK_LENGTH, &retval, NULL);
last_error = GetLastError();
if (last_error != ERROR_SUCCESS && !result) {
retval = -1;
errno = EAGAIN;
} else {
ofpbuf_put(buf, tail, retval);
}
#else
retval = recvmsg(sock->fd, &msg, wait ? 0 : MSG_DONTWAIT);
#endif
error = (retval < 0 ? errno
: retval == 0 ? ECONNRESET /* not possible? */
: nlmsghdr->nlmsg_len != UINT32_MAX ? 0
: retval);
} while (error == EINTR);
if (error) {
if (error == ENOBUFS) {
/* Socket receive buffer overflow dropped one or more messages that
* the kernel tried to send to us. */
COVERAGE_INC(netlink_overflow);
}
return error;
}
if (msg.msg_flags & MSG_TRUNC) {
VLOG_ERR_RL(&rl, "truncated message (longer than %"PRIuSIZE" bytes)",
sizeof tail);
return E2BIG;
}
if (retval < sizeof *nlmsghdr
|| nlmsghdr->nlmsg_len < sizeof *nlmsghdr
|| nlmsghdr->nlmsg_len > retval) {
VLOG_ERR_RL(&rl, "received invalid nlmsg (%"PRIuSIZE" bytes < %"PRIuSIZE")",
retval, sizeof *nlmsghdr);
return EPROTO;
}
#ifndef _WIN32
ofpbuf_set_size(buf, MIN(retval, buf->allocated));
if (retval > buf->allocated) {
COVERAGE_INC(netlink_recv_jumbo);
ofpbuf_put(buf, tail, retval - buf->allocated);
}
#endif
log_nlmsg(__func__, 0, ofpbuf_data(buf), ofpbuf_size(buf), sock->protocol);
COVERAGE_INC(netlink_received);
return 0;
}
static int
nl_sock_recv__(struct nl_sock *sock, struct ofpbuf *buf, bool wait)
{
/* We can't accurately predict the size of the data to be received. The
* caller is supposed to have allocated enough space in 'buf' to handle the
* "typical" case. To handle exceptions, we make available enough space in
* 'tail' to allow Netlink messages to be up to 64 kB long (a reasonable
* figure since that's the maximum length of a Netlink attribute). */
struct nlmsghdr *nlmsghdr;
uint8_t tail[65536];
struct iovec iov[2];
struct msghdr msg;
ssize_t retval;
ovs_assert(buf->allocated >= sizeof *nlmsghdr);
ofpbuf_clear(buf);
iov[0].iov_base = buf->base;
iov[0].iov_len = buf->allocated;
iov[1].iov_base = tail;
iov[1].iov_len = sizeof tail;
memset(&msg, 0, sizeof msg);
msg.msg_iov = iov;
msg.msg_iovlen = 2;
do {
retval = recvmsg(sock->fd, &msg, wait ? 0 : MSG_DONTWAIT);
} while (retval < 0 && errno == EINTR);
if (retval < 0) {
int error = errno;
if (error == ENOBUFS) {
/* Socket receive buffer overflow dropped one or more messages that
* the kernel tried to send to us. */
COVERAGE_INC(netlink_overflow);
}
return error;
}
if (msg.msg_flags & MSG_TRUNC) {
VLOG_ERR_RL(&rl, "truncated message (longer than %zu bytes)",
sizeof tail);
return E2BIG;
}
nlmsghdr = buf->data;
if (retval < sizeof *nlmsghdr
|| nlmsghdr->nlmsg_len < sizeof *nlmsghdr
|| nlmsghdr->nlmsg_len > retval) {
VLOG_ERR_RL(&rl, "received invalid nlmsg (%zd bytes < %zu)",
retval, sizeof *nlmsghdr);
return EPROTO;
}
buf->size = MIN(retval, buf->allocated);
if (retval > buf->allocated) {
COVERAGE_INC(netlink_recv_jumbo);
ofpbuf_put(buf, tail, retval - buf->allocated);
}
log_nlmsg(__func__, 0, buf->data, buf->size, sock->protocol);
COVERAGE_INC(netlink_received);
return 0;
}
/* 'tun_id' is in network byte order, while 'in_port' is in host byte order.
* These byte orders are the same as they are in struct odp_flow_key.
*
* Initializes packet header pointers as follows:
*
* - packet->l2 to the start of the Ethernet header.
*
* - packet->l3 to just past the Ethernet header, or just past the
* vlan_header if one is present, to the first byte of the payload of the
* Ethernet frame.
*
* - packet->l4 to just past the IPv4 header, if one is present and has a
* correct length, and otherwise NULL.
*
* - packet->l7 to just past the TCP or UDP or ICMP header, if one is
* present and has a correct length, and otherwise NULL.
*/
int
flow_extract(struct ofpbuf *packet, uint32_t tun_id, uint16_t in_port,
flow_t *flow)
{
struct ofpbuf b = *packet;
struct eth_header *eth;
int retval = 0;
COVERAGE_INC(flow_extract);
memset(flow, 0, sizeof *flow);
flow->tun_id = tun_id;
flow->in_port = in_port;
flow->dl_vlan = htons(OFP_VLAN_NONE);
packet->l2 = b.data;
packet->l3 = NULL;
packet->l4 = NULL;
packet->l7 = NULL;
if (b.size < sizeof *eth) {
return 0;
}
/* Link layer. */
eth = b.data;
memcpy(flow->dl_src, eth->eth_src, ETH_ADDR_LEN);
memcpy(flow->dl_dst, eth->eth_dst, ETH_ADDR_LEN);
/* dl_type, dl_vlan, dl_vlan_pcp. */
ofpbuf_pull(&b, ETH_ADDR_LEN * 2);
if (eth->eth_type == htons(ETH_TYPE_VLAN)) {
parse_vlan(&b, flow);
}
flow->dl_type = parse_ethertype(&b);
/* Network layer. */
packet->l3 = b.data;
if (flow->dl_type == htons(ETH_TYPE_IP)) {
const struct ip_header *nh = pull_ip(&b);
if (nh) {
flow->nw_src = get_unaligned_u32(&nh->ip_src);
flow->nw_dst = get_unaligned_u32(&nh->ip_dst);
flow->nw_tos = nh->ip_tos & IP_DSCP_MASK;
flow->nw_proto = nh->ip_proto;
packet->l4 = b.data;
if (!IP_IS_FRAGMENT(nh->ip_frag_off)) {
if (flow->nw_proto == IP_TYPE_TCP) {
const struct tcp_header *tcp = pull_tcp(&b);
if (tcp) {
flow->tp_src = tcp->tcp_src;
flow->tp_dst = tcp->tcp_dst;
packet->l7 = b.data;
}
} else if (flow->nw_proto == IP_TYPE_UDP) {
const struct udp_header *udp = pull_udp(&b);
if (udp) {
flow->tp_src = udp->udp_src;
flow->tp_dst = udp->udp_dst;
packet->l7 = b.data;
}
} else if (flow->nw_proto == IP_TYPE_ICMP) {
const struct icmp_header *icmp = pull_icmp(&b);
if (icmp) {
flow->icmp_type = htons(icmp->icmp_type);
flow->icmp_code = htons(icmp->icmp_code);
packet->l7 = b.data;
}
}
} else {
retval = 1;
}
}
} else if (flow->dl_type == htons(ETH_TYPE_ARP)) {
const struct arp_eth_header *arp = pull_arp(&b);
if (arp && arp->ar_hrd == htons(1)
&& arp->ar_pro == htons(ETH_TYPE_IP)
&& arp->ar_hln == ETH_ADDR_LEN
&& arp->ar_pln == 4) {
/* We only match on the lower 8 bits of the opcode. */
if (ntohs(arp->ar_op) <= 0xff) {
flow->nw_proto = ntohs(arp->ar_op);
}
if ((flow->nw_proto == ARP_OP_REQUEST)
|| (flow->nw_proto == ARP_OP_REPLY)) {
flow->nw_src = arp->ar_spa;
flow->nw_dst = arp->ar_tpa;
}
}
}
return retval;
}
