/*
* this function is called on a read event from a listen socket, corresponding
* to an accept. It tries to accept as many connections as possible.
* It returns a positive value upon success, 0 if the connection needs to be
* closed and ignored, or a negative value upon critical failure.
*/
int peer_accept(struct session *s)
{
/* we have a dedicated I/O handler for the stats */
stream_int_register_handler(&s->si[1], &peer_applet);
copy_target(&s->target, &s->si[1].conn.target); // for logging only
s->si[1].conn.data_ctx = s;
s->si[1].applet.st0 = PEER_SESSION_ACCEPT;
tv_zero(&s->logs.tv_request);
s->logs.t_queue = 0;
s->logs.t_connect = 0;
s->logs.t_data = 0;
s->logs.t_close = 0;
s->logs.bytes_in = s->logs.bytes_out = 0;
s->logs.prx_queue_size = 0;/* we get the number of pending conns before us */
s->logs.srv_queue_size = 0; /* we will get this number soon */
s->req->flags |= CF_READ_DONTWAIT; /* we plan to read small requests */
if (s->listener->timeout) {
s->req->rto = *s->listener->timeout;
s->rep->wto = *s->listener->timeout;
}
return 1;
}
/* tv_udpate_date: sets <date> to system time, and sets <now> to something as
* close as possible to real time, following a monotonic function. The main
* principle consists in detecting backwards and forwards time jumps and adjust
* an offset to correct them. This function should be called once after each
* poll, and never farther apart than MAX_DELAY_MS*2. The poll's timeout should
* be passed in <max_wait>, and the return value in <interrupted> (a non-zero
* value means that we have not expired the timeout). Calling it with (-1,*)
* sets both <date> and <now> to current date, and calling it with (0,1) simply
* updates the values.
*/
REGPRM2 void tv_update_date(int max_wait, int interrupted)
{
static struct timeval tv_offset; /* warning: signed offset! */
struct timeval adjusted, deadline;
gettimeofday(&date, NULL);
if (unlikely(max_wait < 0)) {
tv_zero(&tv_offset);
adjusted = date;
after_poll = date;
samp_time = idle_time = 0;
idle_pct = 100;
goto to_ms;
}
__tv_add(&adjusted, &date, &tv_offset);
if (unlikely(__tv_islt(&adjusted, &now))) {
goto fixup; /* jump in the past */
}
/* OK we did not jump backwards, let's see if we have jumped too far
* forwards. The poll value was in <max_wait>, we accept that plus
* MAX_DELAY_MS to cover additional time.
*/
_tv_ms_add(&deadline, &now, max_wait + MAX_DELAY_MS);
if (likely(__tv_islt(&adjusted, &deadline)))
goto to_ms; /* OK time is within expected range */
fixup:
/* Large jump. If the poll was interrupted, we consider that the date
* has not changed (immediate wake-up), otherwise we add the poll
* time-out to the previous date. The new offset is recomputed.
*/
_tv_ms_add(&adjusted, &now, interrupted ? 0 : max_wait);
tv_offset.tv_sec = adjusted.tv_sec - date.tv_sec;
tv_offset.tv_usec = adjusted.tv_usec - date.tv_usec;
if (tv_offset.tv_usec < 0) {
tv_offset.tv_usec += 1000000;
tv_offset.tv_sec--;
}
to_ms:
now = adjusted;
curr_sec_ms = now.tv_usec / 1000; /* ms of current second */
curr_sec_ms_scaled = curr_sec_ms * 4294971; /* ms * 2^32 / 1000 */
now_ms = now.tv_sec * 1000 + curr_sec_ms;
return;
}
/* Finish a session accept() for a proxy (TCP or HTTP). It returns a negative
* value in case of a critical failure which must cause the listener to be
* disabled, a positive value in case of success, or zero if it is a success
* but the session must be closed ASAP (eg: monitoring).
*/
int frontend_accept(struct session *s)
{
int cfd = s->si[0].fd;
tv_zero(&s->logs.tv_request);
s->logs.t_queue = -1;
s->logs.t_connect = -1;
s->logs.t_data = -1;
s->logs.t_close = 0;
s->logs.bytes_in = s->logs.bytes_out = 0;
s->logs.prx_queue_size = 0; /* we get the number of pending conns before us */
s->logs.srv_queue_size = 0; /* we will get this number soon */
/* FIXME: the logs are horribly complicated now, because they are
* defined in <p>, <p>, and later <be> and <be>.
*/
s->do_log = sess_log;
/* default error reporting function, may be changed by analysers */
s->srv_error = default_srv_error;
/* Adjust some socket options */
if (s->listener->addr.ss_family == AF_INET || s->listener->addr.ss_family == AF_INET6) {
if (setsockopt(cfd, IPPROTO_TCP, TCP_NODELAY,
(char *) &one, sizeof(one)) == -1)
goto out_return;
if (s->fe->options & PR_O_TCP_CLI_KA)
setsockopt(cfd, SOL_SOCKET, SO_KEEPALIVE,
(char *) &one, sizeof(one));
if (s->fe->options & PR_O_TCP_NOLING)
setsockopt(cfd, SOL_SOCKET, SO_LINGER,
(struct linger *) &nolinger, sizeof(struct linger));
#if defined(TCP_MAXSEG)
if (s->listener->maxseg < 0) {
/* we just want to reduce the current MSS by that value */
int mss;
socklen_t mss_len = sizeof(mss);
if (getsockopt(cfd, IPPROTO_TCP, TCP_MAXSEG, &mss, &mss_len) == 0) {
mss += s->listener->maxseg; /* remember, it's < 0 */
setsockopt(cfd, IPPROTO_TCP, TCP_MAXSEG, &mss, sizeof(mss));
}
}
#endif
}
if (global.tune.client_sndbuf)
setsockopt(cfd, SOL_SOCKET, SO_SNDBUF, &global.tune.client_sndbuf, sizeof(global.tune.client_sndbuf));
if (global.tune.client_rcvbuf)
setsockopt(cfd, SOL_SOCKET, SO_RCVBUF, &global.tune.client_rcvbuf, sizeof(global.tune.client_rcvbuf));
if (s->fe->mode == PR_MODE_HTTP) {
/* the captures are only used in HTTP frontends */
if (unlikely(s->fe->nb_req_cap > 0 &&
(s->txn.req.cap = pool_alloc2(s->fe->req_cap_pool)) == NULL))
goto out_return; /* no memory */
if (unlikely(s->fe->nb_rsp_cap > 0 &&
(s->txn.rsp.cap = pool_alloc2(s->fe->rsp_cap_pool)) == NULL))
goto out_free_reqcap; /* no memory */
}
if (s->fe->acl_requires & ACL_USE_L7_ANY) {
/* we have to allocate header indexes only if we know
* that we may make use of them. This of course includes
* (mode == PR_MODE_HTTP).
*/
s->txn.hdr_idx.size = global.tune.max_http_hdr;
if (unlikely((s->txn.hdr_idx.v = pool_alloc2(pool2_hdr_idx)) == NULL))
goto out_free_rspcap; /* no memory */
/* and now initialize the HTTP transaction state */
http_init_txn(s);
}
if ((s->fe->mode == PR_MODE_TCP || s->fe->mode == PR_MODE_HTTP)
&& (!LIST_ISEMPTY(&s->fe->logsrvs))) {
if (likely(s->fe->to_log)) {
/* we have the client ip */
if (s->logs.logwait & LW_CLIP)
if (!(s->logs.logwait &= ~LW_CLIP))
s->do_log(s);
}
else {
char pn[INET6_ADDRSTRLEN], sn[INET6_ADDRSTRLEN];
if (!(s->flags & SN_FRT_ADDR_SET))
get_frt_addr(s);
switch (addr_to_str(&s->req->prod->addr.from, pn, sizeof(pn))) {
case AF_INET:
case AF_INET6:
addr_to_str(&s->req->prod->addr.to, sn, sizeof(sn));
send_log(s->fe, LOG_INFO, "Connect from %s:%d to %s:%d (%s/%s)\n",
pn, get_host_port(&s->req->prod->addr.from),
sn, get_host_port(&s->req->prod->addr.to),
s->fe->id, (s->fe->mode == PR_MODE_HTTP) ? "HTTP" : "TCP");
break;
case AF_UNIX:
/* UNIX socket, only the destination is known */
send_log(s->fe, LOG_INFO, "Connect to unix:%d (%s/%s)\n",
s->listener->luid,
s->fe->id, (s->fe->mode == PR_MODE_HTTP) ? "HTTP" : "TCP");
break;
}
}
}
if (unlikely((global.mode & MODE_DEBUG) && (!(global.mode & MODE_QUIET) || (global.mode & MODE_VERBOSE)))) {
char pn[INET6_ADDRSTRLEN];
int len = 0;
if (!(s->flags & SN_FRT_ADDR_SET))
get_frt_addr(s);
switch (addr_to_str(&s->req->prod->addr.from, pn, sizeof(pn))) {
case AF_INET:
case AF_INET6:
len = sprintf(trash, "%08x:%s.accept(%04x)=%04x from [%s:%d]\n",
s->uniq_id, s->fe->id, (unsigned short)s->listener->fd, (unsigned short)cfd,
pn, get_host_port(&s->req->prod->addr.from));
break;
case AF_UNIX:
/* UNIX socket, only the destination is known */
len = sprintf(trash, "%08x:%s.accept(%04x)=%04x from [unix:%d]\n",
s->uniq_id, s->fe->id, (unsigned short)s->listener->fd, (unsigned short)cfd,
s->listener->luid);
break;
}
write(1, trash, len);
}
if (s->fe->mode == PR_MODE_HTTP)
s->req->flags |= BF_READ_DONTWAIT; /* one read is usually enough */
/* note: this should not happen anymore since there's always at least the switching rules */
if (!s->req->analysers) {
buffer_auto_connect(s->req); /* don't wait to establish connection */
buffer_auto_close(s->req); /* let the producer forward close requests */
}
s->req->rto = s->fe->timeout.client;
s->rep->wto = s->fe->timeout.client;
fdtab[cfd].flags = FD_FL_TCP | FD_FL_TCP_NODELAY;
if (s->fe->options & PR_O_TCP_NOLING)
fdtab[cfd].flags |= FD_FL_TCP_NOLING;
if (unlikely((s->fe->mode == PR_MODE_HTTP && (s->flags & SN_MONITOR)) ||
(s->fe->mode == PR_MODE_HEALTH && ((s->fe->options2 & PR_O2_CHK_ANY) == PR_O2_HTTP_CHK)))) {
/* Either we got a request from a monitoring system on an HTTP instance,
* or we're in health check mode with the 'httpchk' option enabled. In
* both cases, we return a fake "HTTP/1.0 200 OK" response and we exit.
*/
struct chunk msg;
chunk_initstr(&msg, "HTTP/1.0 200 OK\r\n\r\n");
stream_int_retnclose(&s->si[0], &msg); /* forge a 200 response */
s->req->analysers = 0;
s->task->expire = s->rep->wex;
EV_FD_CLR(cfd, DIR_RD);
}
else if (unlikely(s->fe->mode == PR_MODE_HEALTH)) { /* health check mode, no client reading */
struct chunk msg;
chunk_initstr(&msg, "OK\n");
stream_int_retnclose(&s->si[0], &msg); /* forge an "OK" response */
s->req->analysers = 0;
s->task->expire = s->rep->wex;
EV_FD_CLR(cfd, DIR_RD);
}
/* everything's OK, let's go on */
return 1;
/* Error unrolling */
out_free_rspcap:
pool_free2(s->fe->rsp_cap_pool, s->txn.rsp.cap);
out_free_reqcap:
pool_free2(s->fe->req_cap_pool, s->txn.req.cap);
out_return:
return -1;
}
/* Finish a session accept() for a proxy (TCP or HTTP). It returns a negative
* value in case of a critical failure which must cause the listener to be
* disabled, a positive value in case of success, or zero if it is a success
* but the session must be closed ASAP (eg: monitoring).
*/
int frontend_accept(struct session *s)
{
int cfd = s->si[0].conn->t.sock.fd;
tv_zero(&s->logs.tv_request);
s->logs.t_queue = -1;
s->logs.t_connect = -1;
s->logs.t_data = -1;
s->logs.t_close = 0;
s->logs.bytes_in = s->logs.bytes_out = 0;
s->logs.prx_queue_size = 0; /* we get the number of pending conns before us */
s->logs.srv_queue_size = 0; /* we will get this number soon */
/* FIXME: the logs are horribly complicated now, because they are
* defined in <p>, <p>, and later <be> and <be>.
*/
s->do_log = sess_log;
/* default error reporting function, may be changed by analysers */
s->srv_error = default_srv_error;
/* Adjust some socket options */
if (s->listener->addr.ss_family == AF_INET || s->listener->addr.ss_family == AF_INET6) {
if (setsockopt(cfd, IPPROTO_TCP, TCP_NODELAY,
(char *) &one, sizeof(one)) == -1)
goto out_return;
if (s->fe->options & PR_O_TCP_CLI_KA)
setsockopt(cfd, SOL_SOCKET, SO_KEEPALIVE,
(char *) &one, sizeof(one));
if (s->fe->options & PR_O_TCP_NOLING)
setsockopt(cfd, SOL_SOCKET, SO_LINGER,
(struct linger *) &nolinger, sizeof(struct linger));
#if defined(TCP_MAXSEG)
if (s->listener->maxseg < 0) {
/* we just want to reduce the current MSS by that value */
int mss;
socklen_t mss_len = sizeof(mss);
if (getsockopt(cfd, IPPROTO_TCP, TCP_MAXSEG, &mss, &mss_len) == 0) {
mss += s->listener->maxseg; /* remember, it's < 0 */
setsockopt(cfd, IPPROTO_TCP, TCP_MAXSEG, &mss, sizeof(mss));
}
}
#endif
}
if (global.tune.client_sndbuf)
setsockopt(cfd, SOL_SOCKET, SO_SNDBUF, &global.tune.client_sndbuf, sizeof(global.tune.client_sndbuf));
if (global.tune.client_rcvbuf)
setsockopt(cfd, SOL_SOCKET, SO_RCVBUF, &global.tune.client_rcvbuf, sizeof(global.tune.client_rcvbuf));
if (s->fe->mode == PR_MODE_HTTP) {
/* the captures are only used in HTTP frontends */
if (unlikely(s->fe->nb_req_cap > 0 &&
(s->txn.req.cap = pool_alloc2(s->fe->req_cap_pool)) == NULL))
goto out_return; /* no memory */
if (unlikely(s->fe->nb_rsp_cap > 0 &&
(s->txn.rsp.cap = pool_alloc2(s->fe->rsp_cap_pool)) == NULL))
goto out_free_reqcap; /* no memory */
}
if (s->fe->http_needed) {
/* we have to allocate header indexes only if we know
* that we may make use of them. This of course includes
* (mode == PR_MODE_HTTP).
*/
s->txn.hdr_idx.size = global.tune.max_http_hdr;
if (unlikely((s->txn.hdr_idx.v = pool_alloc2(pool2_hdr_idx)) == NULL))
goto out_free_rspcap; /* no memory */
/* and now initialize the HTTP transaction state */
http_init_txn(s);
}
if ((s->fe->mode == PR_MODE_TCP || s->fe->mode == PR_MODE_HTTP)
&& (!LIST_ISEMPTY(&s->fe->logsrvs))) {
if (likely(!LIST_ISEMPTY(&s->fe->logformat))) {
/* we have the client ip */
if (s->logs.logwait & LW_CLIP)
if (!(s->logs.logwait &= ~(LW_CLIP|LW_INIT)))
s->do_log(s);
}
else {
char pn[INET6_ADDRSTRLEN], sn[INET6_ADDRSTRLEN];
conn_get_from_addr(s->req->prod->conn);
conn_get_to_addr(s->req->prod->conn);
switch (addr_to_str(&s->req->prod->conn->addr.from, pn, sizeof(pn))) {
case AF_INET:
case AF_INET6:
addr_to_str(&s->req->prod->conn->addr.to, sn, sizeof(sn));
send_log(s->fe, LOG_INFO, "Connect from %s:%d to %s:%d (%s/%s)\n",
pn, get_host_port(&s->req->prod->conn->addr.from),
sn, get_host_port(&s->req->prod->conn->addr.to),
s->fe->id, (s->fe->mode == PR_MODE_HTTP) ? "HTTP" : "TCP");
break;
case AF_UNIX:
/* UNIX socket, only the destination is known */
send_log(s->fe, LOG_INFO, "Connect to unix:%d (%s/%s)\n",
s->listener->luid,
s->fe->id, (s->fe->mode == PR_MODE_HTTP) ? "HTTP" : "TCP");
break;
}
}
}
if (unlikely((global.mode & MODE_DEBUG) && (!(global.mode & MODE_QUIET) || (global.mode & MODE_VERBOSE)))) {
char pn[INET6_ADDRSTRLEN];
conn_get_from_addr(s->req->prod->conn);
switch (addr_to_str(&s->req->prod->conn->addr.from, pn, sizeof(pn))) {
case AF_INET:
case AF_INET6:
chunk_printf(&trash, "%08x:%s.accept(%04x)=%04x from [%s:%d]\n",
s->uniq_id, s->fe->id, (unsigned short)s->listener->fd, (unsigned short)cfd,
pn, get_host_port(&s->req->prod->conn->addr.from));
break;
case AF_UNIX:
/* UNIX socket, only the destination is known */
chunk_printf(&trash, "%08x:%s.accept(%04x)=%04x from [unix:%d]\n",
s->uniq_id, s->fe->id, (unsigned short)s->listener->fd, (unsigned short)cfd,
s->listener->luid);
break;
}
if (write(1, trash.str, trash.len) < 0) /* shut gcc warning */;
}
if (s->fe->mode == PR_MODE_HTTP)
s->req->flags |= CF_READ_DONTWAIT; /* one read is usually enough */
/* note: this should not happen anymore since there's always at least the switching rules */
if (!s->req->analysers) {
channel_auto_connect(s->req); /* don't wait to establish connection */
channel_auto_close(s->req); /* let the producer forward close requests */
}
s->req->rto = s->fe->timeout.client;
s->rep->wto = s->fe->timeout.client;
/* everything's OK, let's go on */
return 1;
/* Error unrolling */
out_free_rspcap:
pool_free2(s->fe->rsp_cap_pool, s->txn.rsp.cap);
out_free_reqcap:
pool_free2(s->fe->req_cap_pool, s->txn.req.cap);
out_return:
return -1;
}
