/* Drains possibly pending incoming data on the file descriptor attached to the
* connection and update the connection's flags accordingly. This is used to
* know whether we need to disable lingering on close. Returns non-zero if it
* is safe to close without disabling lingering, otherwise zero. The SOCK_RD_SH
* flag may also be updated if the incoming shutdown was reported by the drain()
* function.
*/
int conn_sock_drain(struct connection *conn)
{
if (!conn_ctrl_ready(conn))
return 1;
if (conn->flags & (CO_FL_ERROR | CO_FL_SOCK_RD_SH))
return 1;
if (fdtab[conn->t.sock.fd].ev & (FD_POLL_ERR|FD_POLL_HUP)) {
fdtab[conn->t.sock.fd].linger_risk = 0;
}
else {
if (!fd_recv_ready(conn->t.sock.fd))
return 0;
/* disable draining if we were called and have no drain function */
if (!conn->ctrl->drain) {
__conn_data_stop_recv(conn);
return 0;
}
if (conn->ctrl->drain(conn->t.sock.fd) <= 0)
return 0;
}
conn->flags |= CO_FL_SOCK_RD_SH;
return 1;
}
/*
* This function propagates a null read received on a socket-based connection.
* It updates the stream interface. If the stream interface has SI_FL_NOHALF,
* the close is also forwarded to the write side as an abort. This function is
* still socket-specific as it handles a setsockopt() call to set the SO_LINGER
* state on the socket.
*/
void stream_sock_read0(struct stream_interface *si)
{
si->ib->flags &= ~CF_SHUTR_NOW;
if (si->ib->flags & CF_SHUTR)
return;
si->ib->flags |= CF_SHUTR;
si->ib->rex = TICK_ETERNITY;
si->flags &= ~SI_FL_WAIT_ROOM;
if (si->state != SI_ST_EST && si->state != SI_ST_CON)
return;
if (si->ob->flags & CF_SHUTW)
goto do_close;
if (si->flags & SI_FL_NOHALF) {
/* we want to immediately forward this close to the write side */
if (si->flags & SI_FL_NOLINGER) {
si->flags &= ~SI_FL_NOLINGER;
setsockopt(si->conn->t.sock.fd, SOL_SOCKET, SO_LINGER,
(struct linger *) &nolinger, sizeof(struct linger));
}
/* force flag on ssl to keep session in cache */
if (si->conn->xprt->shutw)
si->conn->xprt->shutw(si->conn, 0);
goto do_close;
}
/* otherwise that's just a normal read shutdown */
__conn_data_stop_recv(si->conn);
return;
do_close:
/* OK we completely close the socket here just as if we went through si_shut[rw]() */
conn_full_close(si->conn);
si->ib->flags &= ~CF_SHUTR_NOW;
si->ib->flags |= CF_SHUTR;
si->ib->rex = TICK_ETERNITY;
si->ob->flags &= ~CF_SHUTW_NOW;
si->ob->flags |= CF_SHUTW;
si->ob->wex = TICK_ETERNITY;
si->flags &= ~(SI_FL_WAIT_DATA | SI_FL_WAIT_ROOM);
si->state = SI_ST_DIS;
si->exp = TICK_ETERNITY;
if (si->release)
si->release(si);
return;
}
/*
* This function propagates a null read received on a socket-based connection.
* It updates the stream interface. If the stream interface has SI_FL_NOHALF,
* the close is also forwarded to the write side as an abort. This function is
* still socket-specific as it handles a setsockopt() call to set the SO_LINGER
* state on the socket.
*/
void stream_sock_read0(struct stream_interface *si)
{
struct connection *conn = __objt_conn(si->end);
si->ib->flags &= ~CF_SHUTR_NOW;
if (si->ib->flags & CF_SHUTR)
return;
si->ib->flags |= CF_SHUTR;
si->ib->rex = TICK_ETERNITY;
si->flags &= ~SI_FL_WAIT_ROOM;
if (si->state != SI_ST_EST && si->state != SI_ST_CON)
return;
if (si->ob->flags & CF_SHUTW)
goto do_close;
if (si->flags & SI_FL_NOHALF) {
/* we want to immediately forward this close to the write side */
/* force flag on ssl to keep session in cache */
if (conn->xprt->shutw)
conn->xprt->shutw(conn, 0);
goto do_close;
}
/* otherwise that's just a normal read shutdown */
if (conn_ctrl_ready(conn))
fdtab[conn->t.sock.fd].linger_risk = 0;
__conn_data_stop_recv(conn);
return;
do_close:
/* OK we completely close the socket here just as if we went through si_shut[rw]() */
conn_full_close(conn);
si->ib->flags &= ~CF_SHUTR_NOW;
si->ib->flags |= CF_SHUTR;
si->ib->rex = TICK_ETERNITY;
si->ob->flags &= ~CF_SHUTW_NOW;
si->ob->flags |= CF_SHUTW;
si->ob->wex = TICK_ETERNITY;
si->flags &= ~(SI_FL_WAIT_DATA | SI_FL_WAIT_ROOM);
si->state = SI_ST_DIS;
si->exp = TICK_ETERNITY;
return;
}
/*
* This function propagates a null read received on a socket-based connection.
* It updates the stream interface. If the stream interface has SI_FL_NOHALF,
* the close is also forwarded to the write side as an abort.
*/
void stream_sock_read0(struct stream_interface *si)
{
struct connection *conn = __objt_conn(si->end);
struct channel *ic = si_ic(si);
struct channel *oc = si_oc(si);
HA_DBG("stream_sock_read0\n");
ic->flags &= ~CF_SHUTR_NOW;
if (ic->flags & CF_SHUTR)
return;
ic->flags |= CF_SHUTR;
ic->rex = TICK_ETERNITY;
si->flags &= ~SI_FL_WAIT_ROOM;
if (si->state != SI_ST_EST && si->state != SI_ST_CON)
return;
if (oc->flags & CF_SHUTW)
goto do_close;
if (si->flags & SI_FL_NOHALF) {
/* we want to immediately forward this close to the write side */
/* force flag on ssl to keep stream in cache */
conn_data_shutw_hard(conn);
goto do_close;
}
/* otherwise that's just a normal read shutdown */
__conn_data_stop_recv(conn);
return;
do_close:
/* OK we completely close the socket here just as if we went through si_shut[rw]() */
conn_full_close(conn);
ic->flags &= ~CF_SHUTR_NOW;
ic->flags |= CF_SHUTR;
ic->rex = TICK_ETERNITY;
oc->flags &= ~CF_SHUTW_NOW;
oc->flags |= CF_SHUTW;
oc->wex = TICK_ETERNITY;
si->flags &= ~(SI_FL_WAIT_DATA | SI_FL_WAIT_ROOM);
si->state = SI_ST_DIS;
si->exp = TICK_ETERNITY;
return;
}
/* Tiny I/O callback called on recv/send I/O events on idle connections.
* It simply sets the CO_FL_SOCK_RD_SH flag so that si_idle_conn_wake_cb()
* is notified and can kill the connection.
*/
static void si_idle_conn_null_cb(struct connection *conn)
{
if (conn->flags & (CO_FL_ERROR | CO_FL_SOCK_RD_SH))
return;
if (fdtab[conn->t.sock.fd].ev & (FD_POLL_ERR|FD_POLL_HUP)) {
fdtab[conn->t.sock.fd].linger_risk = 0;
conn->flags |= CO_FL_SOCK_RD_SH;
}
else {
conn_drain(conn);
}
/* disable draining if we were called and have no drain function */
if (!conn->ctrl->drain)
__conn_data_stop_recv(conn);
}
/* Callback to be used by connection I/O handlers upon completion. It propagates
* connection flags to the stream interface, updates the stream (which may or
* may not take this opportunity to try to forward data), then update the
* connection's polling based on the channels and stream interface's final
* states. The function always returns 0.
*/
static int si_conn_wake_cb(struct connection *conn)
{
struct stream_interface *si = conn->owner;
struct channel *ic = si_ic(si);
struct channel *oc = si_oc(si);
/* First step, report to the stream-int what was detected at the
* connection layer : errors and connection establishment.
*/
if (conn->flags & CO_FL_ERROR)
si->flags |= SI_FL_ERR;
if (unlikely(!(conn->flags & (CO_FL_WAIT_L4_CONN | CO_FL_WAIT_L6_CONN | CO_FL_CONNECTED)))) {
si->exp = TICK_ETERNITY;
oc->flags |= CF_WRITE_NULL;
}
/* Second step : update the stream-int and channels, try to forward any
* pending data, then possibly wake the stream up based on the new
* stream-int status.
*/
stream_int_notify(si);
/* Third step : update the connection's polling status based on what
* was done above (eg: maybe some buffers got emptied).
*/
if (channel_is_empty(oc))
__conn_data_stop_send(conn);
if (si->flags & SI_FL_WAIT_ROOM) {
__conn_data_stop_recv(conn);
}
else if ((ic->flags & (CF_SHUTR|CF_READ_PARTIAL|CF_DONT_READ)) == CF_READ_PARTIAL &&
channel_may_recv(ic)) {
__conn_data_want_recv(conn);
}
return 0;
}
/* This function is used for inter-stream-interface calls. It is called by the
* consumer to inform the producer side that it may be interested in checking
* for free space in the buffer. Note that it intentionally does not update
* timeouts, so that we can still check them later at wake-up. This function is
* dedicated to connection-based stream interfaces.
*/
static void stream_int_chk_rcv_conn(struct stream_interface *si)
{
struct channel *ib = si->ib;
if (unlikely(si->state > SI_ST_EST || (ib->flags & CF_SHUTR)))
return;
conn_refresh_polling_flags(si->conn);
if ((ib->flags & CF_DONT_READ) || channel_full(ib)) {
/* stop reading */
if (!(ib->flags & CF_DONT_READ)) /* full */
si->flags |= SI_FL_WAIT_ROOM;
__conn_data_stop_recv(si->conn);
}
else {
/* (re)start reading */
si->flags &= ~SI_FL_WAIT_ROOM;
__conn_data_want_recv(si->conn);
}
conn_cond_update_data_polling(si->conn);
}
/* This function is used for inter-stream-interface calls. It is called by the
* consumer to inform the producer side that it may be interested in checking
* for free space in the buffer. Note that it intentionally does not update
* timeouts, so that we can still check them later at wake-up. This function is
* dedicated to connection-based stream interfaces.
*/
static void stream_int_chk_rcv_conn(struct stream_interface *si)
{
struct channel *ic = si_ic(si);
struct connection *conn = __objt_conn(si->end);
if (unlikely(si->state > SI_ST_EST || (ic->flags & CF_SHUTR)))
return;
conn_refresh_polling_flags(conn);
if ((ic->flags & CF_DONT_READ) || !channel_may_recv(ic)) {
/* stop reading */
if (!(ic->flags & CF_DONT_READ)) /* full */
si->flags |= SI_FL_WAIT_ROOM;
__conn_data_stop_recv(conn);
}
else {
/* (re)start reading */
si->flags &= ~SI_FL_WAIT_ROOM;
__conn_data_want_recv(conn);
}
conn_cond_update_data_polling(conn);
}
/* Updates the polling status of a connection outside of the connection handler
* based on the channel's flags and the stream interface's flags. It needs to be
* called once after the channels' flags have settled down and the stream has
* been updated. It is not designed to be called from within the connection
* handler itself.
*/
void stream_int_update_conn(struct stream_interface *si)
{
struct channel *ic = si_ic(si);
struct channel *oc = si_oc(si);
struct connection *conn = __objt_conn(si->end);
if (!(ic->flags & CF_SHUTR)) {
/* Read not closed */
if ((ic->flags & CF_DONT_READ) || !channel_may_recv(ic))
__conn_data_stop_recv(conn);
else
__conn_data_want_recv(conn);
}
if (!(oc->flags & CF_SHUTW)) {
/* Write not closed */
if (channel_is_empty(oc))
__conn_data_stop_send(conn);
else
__conn_data_want_send(conn);
}
conn_cond_update_data_polling(conn);
}
/* Callback to be used by connection I/O handlers upon completion. It differs from
* the update function in that it is designed to be called by lower layers after I/O
* events have been completed. It will also try to wake the associated task up if
* an important event requires special handling. It relies on the connection handler
* to commit any polling updates. The function always returns 0.
*/
static int si_conn_wake_cb(struct connection *conn)
{
struct stream_interface *si = conn->owner;
DPRINTF(stderr, "%s: si=%p, si->state=%d ib->flags=%08x ob->flags=%08x\n",
__FUNCTION__,
si, si->state, si->ib->flags, si->ob->flags);
if (conn->flags & CO_FL_ERROR)
si->flags |= SI_FL_ERR;
/* check for recent connection establishment */
if (unlikely(!(conn->flags & (CO_FL_WAIT_L4_CONN | CO_FL_WAIT_L6_CONN | CO_FL_CONNECTED)))) {
si->exp = TICK_ETERNITY;
si->ob->flags |= CF_WRITE_NULL;
}
/* process consumer side */
if (channel_is_empty(si->ob)) {
if (((si->ob->flags & (CF_SHUTW|CF_SHUTW_NOW)) == CF_SHUTW_NOW) &&
(si->state == SI_ST_EST))
stream_int_shutw_conn(si);
__conn_data_stop_send(conn);
si->ob->wex = TICK_ETERNITY;
}
if ((si->ob->flags & (CF_SHUTW|CF_SHUTW_NOW)) == 0 && !channel_full(si->ob))
si->flags |= SI_FL_WAIT_DATA;
if (si->ob->flags & CF_WRITE_ACTIVITY) {
/* update timeouts if we have written something */
if ((si->ob->flags & (CF_SHUTW|CF_WRITE_PARTIAL)) == CF_WRITE_PARTIAL &&
!channel_is_empty(si->ob))
if (tick_isset(si->ob->wex))
si->ob->wex = tick_add_ifset(now_ms, si->ob->wto);
if (!(si->flags & SI_FL_INDEP_STR))
if (tick_isset(si->ib->rex))
si->ib->rex = tick_add_ifset(now_ms, si->ib->rto);
if (likely((si->ob->flags & (CF_SHUTW|CF_WRITE_PARTIAL|CF_DONT_READ)) == CF_WRITE_PARTIAL &&
!channel_full(si->ob) &&
(si->ob->prod->flags & SI_FL_WAIT_ROOM)))
si_chk_rcv(si->ob->prod);
}
/* process producer side.
* We might have some data the consumer is waiting for.
* We can do fast-forwarding, but we avoid doing this for partial
* buffers, because it is very likely that it will be done again
* immediately afterwards once the following data is parsed (eg:
* HTTP chunking).
*/
if (((si->ib->flags & CF_READ_PARTIAL) && !channel_is_empty(si->ib)) &&
(si->ib->pipe /* always try to send spliced data */ ||
(si->ib->buf->i == 0 && (si->ib->cons->flags & SI_FL_WAIT_DATA)))) {
int last_len = si->ib->pipe ? si->ib->pipe->data : 0;
si_chk_snd(si->ib->cons);
/* check if the consumer has freed some space either in the
* buffer or in the pipe.
*/
if (!channel_full(si->ib) &&
(!last_len || !si->ib->pipe || si->ib->pipe->data < last_len))
si->flags &= ~SI_FL_WAIT_ROOM;
}
if (si->flags & SI_FL_WAIT_ROOM) {
__conn_data_stop_recv(conn);
si->ib->rex = TICK_ETERNITY;
}
else if ((si->ib->flags & (CF_SHUTR|CF_READ_PARTIAL|CF_DONT_READ)) == CF_READ_PARTIAL &&
!channel_full(si->ib)) {
/* we must re-enable reading if si_chk_snd() has freed some space */
__conn_data_want_recv(conn);
if (!(si->ib->flags & CF_READ_NOEXP) && tick_isset(si->ib->rex))
si->ib->rex = tick_add_ifset(now_ms, si->ib->rto);
}
/* wake the task up only when needed */
if (/* changes on the production side */
(si->ib->flags & (CF_READ_NULL|CF_READ_ERROR)) ||
si->state != SI_ST_EST ||
(si->flags & SI_FL_ERR) ||
((si->ib->flags & CF_READ_PARTIAL) &&
(!si->ib->to_forward || si->ib->cons->state != SI_ST_EST)) ||
/* changes on the consumption side */
(si->ob->flags & (CF_WRITE_NULL|CF_WRITE_ERROR)) ||
((si->ob->flags & CF_WRITE_ACTIVITY) &&
((si->ob->flags & CF_SHUTW) ||
((si->ob->flags & CF_WAKE_WRITE) &&
(si->ob->prod->state != SI_ST_EST ||
(channel_is_empty(si->ob) && !si->ob->to_forward)))))) {
task_wakeup(si->owner, TASK_WOKEN_IO);
}
if (si->ib->flags & CF_READ_ACTIVITY)
si->ib->flags &= ~CF_READ_DONTWAIT;
return 0;
}
/*
* This is the callback which is called by the connection layer to receive data
* into the buffer from the connection. It iterates over the transport layer's
* rcv_buf function.
*/
static void si_conn_recv_cb(struct connection *conn)
{
struct stream_interface *si = conn->owner;
struct channel *chn = si->ib;
int ret, max, cur_read;
int read_poll = MAX_READ_POLL_LOOPS;
/* stop immediately on errors. Note that we DON'T want to stop on
* POLL_ERR, as the poller might report a write error while there
* are still data available in the recv buffer. This typically
* happens when we send too large a request to a backend server
* which rejects it before reading it all.
*/
if (conn->flags & CO_FL_ERROR)
return;
/* stop here if we reached the end of data */
if (conn_data_read0_pending(conn))
goto out_shutdown_r;
/* maybe we were called immediately after an asynchronous shutr */
if (chn->flags & CF_SHUTR)
return;
cur_read = 0;
if ((chn->flags & (CF_STREAMER | CF_STREAMER_FAST)) && !chn->buf->o &&
global.tune.idle_timer &&
(unsigned short)(now_ms - chn->last_read) >= global.tune.idle_timer) {
/* The buffer was empty and nothing was transferred for more
* than one second. This was caused by a pause and not by
* congestion. Reset any streaming mode to reduce latency.
*/
chn->xfer_small = 0;
chn->xfer_large = 0;
chn->flags &= ~(CF_STREAMER | CF_STREAMER_FAST);
}
/* First, let's see if we may splice data across the channel without
* using a buffer.
*/
if (conn->xprt->rcv_pipe &&
(chn->pipe || chn->to_forward >= MIN_SPLICE_FORWARD) &&
chn->flags & CF_KERN_SPLICING) {
if (buffer_not_empty(chn->buf)) {
/* We're embarrassed, there are already data pending in
* the buffer and we don't want to have them at two
* locations at a time. Let's indicate we need some
* place and ask the consumer to hurry.
*/
goto abort_splice;
}
if (unlikely(chn->pipe == NULL)) {
if (pipes_used >= global.maxpipes || !(chn->pipe = get_pipe())) {
chn->flags &= ~CF_KERN_SPLICING;
goto abort_splice;
}
}
ret = conn->xprt->rcv_pipe(conn, chn->pipe, chn->to_forward);
if (ret < 0) {
/* splice not supported on this end, let's disable it */
chn->flags &= ~CF_KERN_SPLICING;
goto abort_splice;
}
if (ret > 0) {
if (chn->to_forward != CHN_INFINITE_FORWARD)
chn->to_forward -= ret;
chn->total += ret;
cur_read += ret;
chn->flags |= CF_READ_PARTIAL;
}
if (conn_data_read0_pending(conn))
goto out_shutdown_r;
if (conn->flags & CO_FL_ERROR)
return;
if (conn->flags & CO_FL_WAIT_ROOM) {
/* the pipe is full or we have read enough data that it
* could soon be full. Let's stop before needing to poll.
*/
si->flags |= SI_FL_WAIT_ROOM;
__conn_data_stop_recv(conn);
}
/* splice not possible (anymore), let's go on on standard copy */
}
abort_splice:
if (chn->pipe && unlikely(!chn->pipe->data)) {
put_pipe(chn->pipe);
chn->pipe = NULL;
}
/* Important note : if we're called with POLL_IN|POLL_HUP, it means the read polling
* was enabled, which implies that the recv buffer was not full. So we have a guarantee
* that if such an event is not handled above in splice, it will be handled here by
* recv().
*/
while (!(conn->flags & (CO_FL_ERROR | CO_FL_SOCK_RD_SH | CO_FL_DATA_RD_SH | CO_FL_WAIT_ROOM | CO_FL_HANDSHAKE))) {
max = bi_avail(chn);
if (!max) {
si->flags |= SI_FL_WAIT_ROOM;
break;
}
ret = conn->xprt->rcv_buf(conn, chn->buf, max);
if (ret <= 0)
break;
cur_read += ret;
/* if we're allowed to directly forward data, we must update ->o */
if (chn->to_forward && !(chn->flags & (CF_SHUTW|CF_SHUTW_NOW))) {
unsigned long fwd = ret;
if (chn->to_forward != CHN_INFINITE_FORWARD) {
if (fwd > chn->to_forward)
fwd = chn->to_forward;
chn->to_forward -= fwd;
}
b_adv(chn->buf, fwd);
}
chn->flags |= CF_READ_PARTIAL;
chn->total += ret;
if (channel_full(chn)) {
si->flags |= SI_FL_WAIT_ROOM;
break;
}
if ((chn->flags & CF_READ_DONTWAIT) || --read_poll <= 0) {
si->flags |= SI_FL_WAIT_ROOM;
__conn_data_stop_recv(conn);
break;
}
/* if too many bytes were missing from last read, it means that
* it's pointless trying to read again because the system does
* not have them in buffers.
*/
if (ret < max) {
/* if a streamer has read few data, it may be because we
* have exhausted system buffers. It's not worth trying
* again.
*/
if (chn->flags & CF_STREAMER)
break;
/* if we read a large block smaller than what we requested,
* it's almost certain we'll never get anything more.
*/
if (ret >= global.tune.recv_enough)
break;
}
} /* while !flags */
if (conn->flags & CO_FL_ERROR)
return;
if (cur_read) {
if ((chn->flags & (CF_STREAMER | CF_STREAMER_FAST)) &&
(cur_read <= chn->buf->size / 2)) {
chn->xfer_large = 0;
chn->xfer_small++;
if (chn->xfer_small >= 3) {
/* we have read less than half of the buffer in
* one pass, and this happened at least 3 times.
* This is definitely not a streamer.
*/
chn->flags &= ~(CF_STREAMER | CF_STREAMER_FAST);
}
else if (chn->xfer_small >= 2) {
/* if the buffer has been at least half full twice,
* we receive faster than we send, so at least it
* is not a "fast streamer".
*/
chn->flags &= ~CF_STREAMER_FAST;
}
}
else if (!(chn->flags & CF_STREAMER_FAST) &&
(cur_read >= chn->buf->size - global.tune.maxrewrite)) {
/* we read a full buffer at once */
chn->xfer_small = 0;
chn->xfer_large++;
if (chn->xfer_large >= 3) {
/* we call this buffer a fast streamer if it manages
* to be filled in one call 3 consecutive times.
*/
chn->flags |= (CF_STREAMER | CF_STREAMER_FAST);
}
}
else {
chn->xfer_small = 0;
chn->xfer_large = 0;
}
chn->last_read = now_ms;
}
if (conn_data_read0_pending(conn))
/* connection closed */
goto out_shutdown_r;
return;
out_shutdown_r:
/* we received a shutdown */
chn->flags |= CF_READ_NULL;
if (chn->flags & CF_AUTO_CLOSE)
channel_shutw_now(chn);
stream_sock_read0(si);
conn_data_read0(conn);
return;
}