freebsd-skq/usr.bin/csup/mux.c
lulf c6aa3ac446 - Move csup away from contrib/ and into usr.bin/. Software is no longer
contributed, and main development is happening in the FreeBSD repo.

Suggested by:	joel
2010-03-02 07:26:07 +00:00

1203 lines
26 KiB
C

/*-
* Copyright (c) 2003-2006, Maxime Henrion <mux@FreeBSD.org>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/socket.h>
#include <sys/uio.h>
#include <netinet/in.h>
#include <assert.h>
#include <errno.h>
#include <pthread.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "misc.h"
#include "mux.h"
/*
* Packet types.
*/
#define MUX_STARTUPREQ 0
#define MUX_STARTUPREP 1
#define MUX_CONNECT 2
#define MUX_ACCEPT 3
#define MUX_RESET 4
#define MUX_DATA 5
#define MUX_WINDOW 6
#define MUX_CLOSE 7
/*
* Header sizes.
*/
#define MUX_STARTUPHDRSZ 3
#define MUX_CONNECTHDRSZ 8
#define MUX_ACCEPTHDRSZ 8
#define MUX_RESETHDRSZ 2
#define MUX_DATAHDRSZ 4
#define MUX_WINDOWHDRSZ 6
#define MUX_CLOSEHDRSZ 2
#define MUX_PROTOVER 0 /* Protocol version. */
struct mux_header {
uint8_t type;
union {
struct {
uint16_t version;
} __packed mh_startup;
struct {
uint8_t id;
uint16_t mss;
uint32_t window;
} __packed mh_connect;
struct {
uint8_t id;
uint16_t mss;
uint32_t window;
} __packed mh_accept;
struct {
uint8_t id;
} __packed mh_reset;
struct {
uint8_t id;
uint16_t len;
} __packed mh_data;
struct {
uint8_t id;
uint32_t window;
} __packed mh_window;
struct {
uint8_t id;
} __packed mh_close;
} mh_u;
} __packed;
#define mh_startup mh_u.mh_startup
#define mh_connect mh_u.mh_connect
#define mh_accept mh_u.mh_accept
#define mh_reset mh_u.mh_reset
#define mh_data mh_u.mh_data
#define mh_window mh_u.mh_window
#define mh_close mh_u.mh_close
#define MUX_MAXCHAN 2
/* Channel states. */
#define CS_UNUSED 0
#define CS_LISTENING 1
#define CS_CONNECTING 2
#define CS_ESTABLISHED 3
#define CS_RDCLOSED 4
#define CS_WRCLOSED 5
#define CS_CLOSED 6
/* Channel flags. */
#define CF_CONNECT 0x01
#define CF_ACCEPT 0x02
#define CF_RESET 0x04
#define CF_WINDOW 0x08
#define CF_DATA 0x10
#define CF_CLOSE 0x20
#define CHAN_SBSIZE (16 * 1024) /* Send buffer size. */
#define CHAN_RBSIZE (16 * 1024) /* Receive buffer size. */
#define CHAN_MAXSEGSIZE 1024 /* Maximum segment size. */
/* Circular buffer. */
struct buf {
uint8_t *data;
size_t size;
size_t in;
size_t out;
};
struct chan {
int flags;
int state;
pthread_mutex_t lock;
struct mux *mux;
/* Receiver state variables. */
struct buf *recvbuf;
pthread_cond_t rdready;
uint32_t recvseq;
uint16_t recvmss;
/* Sender state variables. */
struct buf *sendbuf;
pthread_cond_t wrready;
uint32_t sendseq;
uint32_t sendwin;
uint16_t sendmss;
};
struct mux {
int closed;
int status;
int socket;
pthread_mutex_t lock;
pthread_cond_t done;
struct chan *channels[MUX_MAXCHAN];
int nchans;
/* Sender thread data. */
pthread_t sender;
pthread_cond_t sender_newwork;
pthread_cond_t sender_started;
int sender_waiting;
int sender_ready;
int sender_lastid;
/* Receiver thread data. */
pthread_t receiver;
};
static int sock_writev(int, struct iovec *, int);
static int sock_write(int, void *, size_t);
static ssize_t sock_read(int, void *, size_t);
static int sock_readwait(int, void *, size_t);
static int mux_init(struct mux *);
static void mux_lock(struct mux *);
static void mux_unlock(struct mux *);
static struct chan *chan_new(struct mux *);
static struct chan *chan_get(struct mux *, int);
static struct chan *chan_connect(struct mux *, int);
static void chan_lock(struct chan *);
static void chan_unlock(struct chan *);
static int chan_insert(struct mux *, struct chan *);
static void chan_free(struct chan *);
static struct buf *buf_new(size_t);
static size_t buf_count(struct buf *);
static size_t buf_avail(struct buf *);
static void buf_get(struct buf *, void *, size_t);
static void buf_put(struct buf *, const void *, size_t);
static void buf_free(struct buf *);
static void sender_wakeup(struct mux *);
static void *sender_loop(void *);
static int sender_waitforwork(struct mux *, int *);
static int sender_scan(struct mux *, int *);
static void sender_cleanup(void *);
static void *receiver_loop(void *);
static int
sock_writev(int s, struct iovec *iov, int iovcnt)
{
ssize_t nbytes;
again:
nbytes = writev(s, iov, iovcnt);
if (nbytes != -1) {
while (nbytes > 0 && (size_t)nbytes >= iov->iov_len) {
nbytes -= iov->iov_len;
iov++;
iovcnt--;
}
if (nbytes == 0)
return (0);
iov->iov_len -= nbytes;
iov->iov_base = (char *)iov->iov_base + nbytes;
} else if (errno != EINTR) {
return (-1);
}
goto again;
}
static int
sock_write(int s, void *buf, size_t size)
{
struct iovec iov;
int ret;
iov.iov_base = buf;
iov.iov_len = size;
ret = sock_writev(s, &iov, 1);
return (ret);
}
static ssize_t
sock_read(int s, void *buf, size_t size)
{
ssize_t nbytes;
again:
nbytes = read(s, buf, size);
if (nbytes == -1 && errno == EINTR)
goto again;
return (nbytes);
}
static int
sock_readwait(int s, void *buf, size_t size)
{
char *cp;
ssize_t nbytes;
size_t left;
cp = buf;
left = size;
while (left > 0) {
nbytes = sock_read(s, cp, left);
if (nbytes == 0) {
errno = ECONNRESET;
return (-1);
}
if (nbytes < 0)
return (-1);
left -= nbytes;
cp += nbytes;
}
return (0);
}
static void
mux_lock(struct mux *m)
{
int error;
error = pthread_mutex_lock(&m->lock);
assert(!error);
}
static void
mux_unlock(struct mux *m)
{
int error;
error = pthread_mutex_unlock(&m->lock);
assert(!error);
}
/* Create a TCP multiplexer on the given socket. */
struct mux *
mux_open(int sock, struct chan **chan)
{
struct mux *m;
struct chan *chan0;
int error;
m = xmalloc(sizeof(struct mux));
memset(m->channels, 0, sizeof(m->channels));
m->nchans = 0;
m->closed = 0;
m->status = -1;
m->socket = sock;
m->sender_waiting = 0;
m->sender_lastid = 0;
m->sender_ready = 0;
pthread_mutex_init(&m->lock, NULL);
pthread_cond_init(&m->done, NULL);
pthread_cond_init(&m->sender_newwork, NULL);
pthread_cond_init(&m->sender_started, NULL);
error = mux_init(m);
if (error)
goto bad;
chan0 = chan_connect(m, 0);
if (chan0 == NULL)
goto bad;
*chan = chan0;
return (m);
bad:
mux_shutdown(m, NULL, STATUS_FAILURE);
(void)mux_close(m);
return (NULL);
}
int
mux_close(struct mux *m)
{
struct chan *chan;
int i, status;
assert(m->closed);
for (i = 0; i < m->nchans; i++) {
chan = m->channels[i];
if (chan != NULL)
chan_free(chan);
}
pthread_cond_destroy(&m->sender_started);
pthread_cond_destroy(&m->sender_newwork);
pthread_cond_destroy(&m->done);
pthread_mutex_destroy(&m->lock);
status = m->status;
free(m);
return (status);
}
/* Close a channel. */
int
chan_close(struct chan *chan)
{
chan_lock(chan);
if (chan->state == CS_ESTABLISHED) {
chan->state = CS_WRCLOSED;
chan->flags |= CF_CLOSE;
} else if (chan->state == CS_RDCLOSED) {
chan->state = CS_CLOSED;
chan->flags |= CF_CLOSE;
} else if (chan->state == CS_WRCLOSED || chan->state == CS_CLOSED) {
chan_unlock(chan);
return (0);
} else {
chan_unlock(chan);
return (-1);
}
chan_unlock(chan);
sender_wakeup(chan->mux);
return (0);
}
void
chan_wait(struct chan *chan)
{
chan_lock(chan);
while (chan->state != CS_CLOSED)
pthread_cond_wait(&chan->rdready, &chan->lock);
chan_unlock(chan);
}
/* Returns the ID of an available channel in the listening state. */
int
chan_listen(struct mux *m)
{
struct chan *chan;
int i;
mux_lock(m);
for (i = 0; i < m->nchans; i++) {
chan = m->channels[i];
chan_lock(chan);
if (chan->state == CS_UNUSED) {
mux_unlock(m);
chan->state = CS_LISTENING;
chan_unlock(chan);
return (i);
}
chan_unlock(chan);
}
mux_unlock(m);
chan = chan_new(m);
chan->state = CS_LISTENING;
i = chan_insert(m, chan);
if (i == -1)
chan_free(chan);
return (i);
}
struct chan *
chan_accept(struct mux *m, int id)
{
struct chan *chan;
chan = chan_get(m, id);
while (chan->state == CS_LISTENING)
pthread_cond_wait(&chan->rdready, &chan->lock);
if (chan->state != CS_ESTABLISHED) {
errno = ECONNRESET;
chan_unlock(chan);
return (NULL);
}
chan_unlock(chan);
return (chan);
}
/* Read bytes from a channel. */
ssize_t
chan_read(struct chan *chan, void *buf, size_t size)
{
char *cp;
size_t count, n;
cp = buf;
chan_lock(chan);
for (;;) {
if (chan->state == CS_RDCLOSED || chan->state == CS_CLOSED) {
chan_unlock(chan);
return (0);
}
if (chan->state != CS_ESTABLISHED &&
chan->state != CS_WRCLOSED) {
chan_unlock(chan);
errno = EBADF;
return (-1);
}
count = buf_count(chan->recvbuf);
if (count > 0)
break;
pthread_cond_wait(&chan->rdready, &chan->lock);
}
n = min(count, size);
buf_get(chan->recvbuf, cp, n);
chan->recvseq += n;
chan->flags |= CF_WINDOW;
chan_unlock(chan);
/* We need to wake up the sender so that it sends a window update. */
sender_wakeup(chan->mux);
return (n);
}
/* Write bytes to a channel. */
ssize_t
chan_write(struct chan *chan, const void *buf, size_t size)
{
const char *cp;
size_t avail, n, pos;
pos = 0;
cp = buf;
chan_lock(chan);
while (pos < size) {
for (;;) {
if (chan->state != CS_ESTABLISHED &&
chan->state != CS_RDCLOSED) {
chan_unlock(chan);
errno = EPIPE;
return (-1);
}
avail = buf_avail(chan->sendbuf);
if (avail > 0)
break;
pthread_cond_wait(&chan->wrready, &chan->lock);
}
n = min(avail, size - pos);
buf_put(chan->sendbuf, cp + pos, n);
pos += n;
}
chan_unlock(chan);
sender_wakeup(chan->mux);
return (size);
}
/*
* Internal channel API.
*/
static struct chan *
chan_connect(struct mux *m, int id)
{
struct chan *chan;
chan = chan_get(m, id);
if (chan->state != CS_UNUSED) {
chan_unlock(chan);
return (NULL);
}
chan->state = CS_CONNECTING;
chan->flags |= CF_CONNECT;
chan_unlock(chan);
sender_wakeup(m);
chan_lock(chan);
while (chan->state == CS_CONNECTING)
pthread_cond_wait(&chan->wrready, &chan->lock);
if (chan->state != CS_ESTABLISHED) {
chan_unlock(chan);
return (NULL);
}
chan_unlock(chan);
return (chan);
}
/*
* Get a channel from its ID, creating it if necessary.
* The channel is returned locked.
*/
static struct chan *
chan_get(struct mux *m, int id)
{
struct chan *chan;
assert(id < MUX_MAXCHAN);
mux_lock(m);
chan = m->channels[id];
if (chan == NULL) {
chan = chan_new(m);
m->channels[id] = chan;
m->nchans++;
}
chan_lock(chan);
mux_unlock(m);
return (chan);
}
/* Lock a channel. */
static void
chan_lock(struct chan *chan)
{
int error;
error = pthread_mutex_lock(&chan->lock);
assert(!error);
}
/* Unlock a channel. */
static void
chan_unlock(struct chan *chan)
{
int error;
error = pthread_mutex_unlock(&chan->lock);
assert(!error);
}
/*
* Create a new channel.
*/
static struct chan *
chan_new(struct mux *m)
{
struct chan *chan;
chan = xmalloc(sizeof(struct chan));
chan->state = CS_UNUSED;
chan->flags = 0;
chan->mux = m;
chan->sendbuf = buf_new(CHAN_SBSIZE);
chan->sendseq = 0;
chan->sendwin = 0;
chan->sendmss = 0;
chan->recvbuf = buf_new(CHAN_RBSIZE);
chan->recvseq = 0;
chan->recvmss = CHAN_MAXSEGSIZE;
pthread_mutex_init(&chan->lock, NULL);
pthread_cond_init(&chan->rdready, NULL);
pthread_cond_init(&chan->wrready, NULL);
return (chan);
}
/* Free any resources associated with a channel. */
static void
chan_free(struct chan *chan)
{
pthread_cond_destroy(&chan->rdready);
pthread_cond_destroy(&chan->wrready);
pthread_mutex_destroy(&chan->lock);
buf_free(chan->recvbuf);
buf_free(chan->sendbuf);
free(chan);
}
/* Insert the new channel in the channel list. */
static int
chan_insert(struct mux *m, struct chan *chan)
{
int i;
mux_lock(m);
for (i = 0; i < MUX_MAXCHAN; i++) {
if (m->channels[i] == NULL) {
m->channels[i] = chan;
m->nchans++;
mux_unlock(m);
return (i);
}
}
errno = ENOBUFS;
return (-1);
}
/*
* Initialize the multiplexer protocol.
*
* This means negotiating protocol version and starting
* the receiver and sender threads.
*/
static int
mux_init(struct mux *m)
{
struct mux_header mh;
int error;
mh.type = MUX_STARTUPREQ;
mh.mh_startup.version = htons(MUX_PROTOVER);
error = sock_write(m->socket, &mh, MUX_STARTUPHDRSZ);
if (error)
return (-1);
error = sock_readwait(m->socket, &mh, MUX_STARTUPHDRSZ);
if (error)
return (-1);
if (mh.type != MUX_STARTUPREP ||
ntohs(mh.mh_startup.version) != MUX_PROTOVER)
return (-1);
mux_lock(m);
error = pthread_create(&m->sender, NULL, sender_loop, m);
if (error) {
mux_unlock(m);
return (-1);
}
/*
* Make sure the sender thread has run and is waiting for new work
* before going on. Otherwise, it might lose the race and a
* request, which will cause a deadlock.
*/
while (!m->sender_ready)
pthread_cond_wait(&m->sender_started, &m->lock);
mux_unlock(m);
error = pthread_create(&m->receiver, NULL, receiver_loop, m);
if (error)
return (-1);
return (0);
}
/*
* Close all the channels, terminate the sender and receiver thread.
* This is an important function because it is used everytime we need
* to wake up all the worker threads to abort the program.
*
* This function accepts an error message that will be printed if the
* multiplexer wasn't already closed. This is useful because it ensures
* that only the first error message will be printed, and that it will
* be printed before doing the actual shutdown work. If this is a
* normal shutdown, NULL can be passed instead.
*
* The "status" parameter of the first mux_shutdown() call is retained
* and then returned by mux_close(), so that the main thread can know
* what type of error happened in the end, if any.
*/
void
mux_shutdown(struct mux *m, const char *errmsg, int status)
{
pthread_t self, sender, receiver;
struct chan *chan;
const char *name;
void *val;
int i, ret;
mux_lock(m);
if (m->closed) {
mux_unlock(m);
return;
}
m->closed = 1;
m->status = status;
self = pthread_self();
sender = m->sender;
receiver = m->receiver;
if (errmsg != NULL) {
if (pthread_equal(self, receiver))
name = "Receiver";
else if (pthread_equal(self, sender))
name = "Sender";
else
name = NULL;
if (name == NULL)
lprintf(-1, "%s\n", errmsg);
else
lprintf(-1, "%s: %s\n", name, errmsg);
}
for (i = 0; i < MUX_MAXCHAN; i++) {
if (m->channels[i] != NULL) {
chan = m->channels[i];
chan_lock(chan);
if (chan->state != CS_UNUSED) {
chan->state = CS_CLOSED;
chan->flags = 0;
pthread_cond_broadcast(&chan->rdready);
pthread_cond_broadcast(&chan->wrready);
}
chan_unlock(chan);
}
}
mux_unlock(m);
if (!pthread_equal(self, receiver)) {
ret = pthread_cancel(receiver);
assert(!ret);
pthread_join(receiver, &val);
assert(val == PTHREAD_CANCELED);
}
if (!pthread_equal(self, sender)) {
ret = pthread_cancel(sender);
assert(!ret);
pthread_join(sender, &val);
assert(val == PTHREAD_CANCELED);
}
}
static void
sender_wakeup(struct mux *m)
{
int waiting;
mux_lock(m);
waiting = m->sender_waiting;
mux_unlock(m);
/*
* We don't care about the race here: if the sender was
* waiting and is not anymore, we'll just send a useless
* signal; if he wasn't waiting then he won't go to sleep
* before having sent what we want him to.
*/
if (waiting)
pthread_cond_signal(&m->sender_newwork);
}
static void *
sender_loop(void *arg)
{
struct iovec iov[3];
struct mux_header mh;
struct mux *m;
struct chan *chan;
struct buf *buf;
uint32_t winsize;
uint16_t hdrsize, size, len;
int error, id, iovcnt, what = 0;
m = (struct mux *)arg;
what = 0;
again:
id = sender_waitforwork(m, &what);
chan = chan_get(m, id);
hdrsize = size = 0;
switch (what) {
case CF_CONNECT:
mh.type = MUX_CONNECT;
mh.mh_connect.id = id;
mh.mh_connect.mss = htons(chan->recvmss);
mh.mh_connect.window = htonl(chan->recvseq +
chan->recvbuf->size);
hdrsize = MUX_CONNECTHDRSZ;
break;
case CF_ACCEPT:
mh.type = MUX_ACCEPT;
mh.mh_accept.id = id;
mh.mh_accept.mss = htons(chan->recvmss);
mh.mh_accept.window = htonl(chan->recvseq +
chan->recvbuf->size);
hdrsize = MUX_ACCEPTHDRSZ;
break;
case CF_RESET:
mh.type = MUX_RESET;
mh.mh_reset.id = id;
hdrsize = MUX_RESETHDRSZ;
break;
case CF_WINDOW:
mh.type = MUX_WINDOW;
mh.mh_window.id = id;
mh.mh_window.window = htonl(chan->recvseq +
chan->recvbuf->size);
hdrsize = MUX_WINDOWHDRSZ;
break;
case CF_DATA:
mh.type = MUX_DATA;
mh.mh_data.id = id;
size = min(buf_count(chan->sendbuf), chan->sendmss);
winsize = chan->sendwin - chan->sendseq;
if (winsize < size)
size = winsize;
mh.mh_data.len = htons(size);
hdrsize = MUX_DATAHDRSZ;
break;
case CF_CLOSE:
mh.type = MUX_CLOSE;
mh.mh_close.id = id;
hdrsize = MUX_CLOSEHDRSZ;
break;
}
if (size > 0) {
assert(mh.type == MUX_DATA);
/*
* Older FreeBSD versions (and maybe other OSes) have the
* iov_base field defined as char *. Cast to char * to
* silence a warning in this case.
*/
iov[0].iov_base = (char *)&mh;
iov[0].iov_len = hdrsize;
iovcnt = 1;
/* We access the buffer directly to avoid some copying. */
buf = chan->sendbuf;
len = min(size, buf->size + 1 - buf->out);
iov[iovcnt].iov_base = buf->data + buf->out;
iov[iovcnt].iov_len = len;
iovcnt++;
if (size > len) {
/* Wrapping around. */
iov[iovcnt].iov_base = buf->data;
iov[iovcnt].iov_len = size - len;
iovcnt++;
}
/*
* Since we're the only thread sending bytes from the
* buffer and modifying buf->out, it's safe to unlock
* here during I/O. It avoids keeping the channel lock
* too long, since write() might block.
*/
chan_unlock(chan);
error = sock_writev(m->socket, iov, iovcnt);
if (error)
goto bad;
chan_lock(chan);
chan->sendseq += size;
buf->out += size;
if (buf->out > buf->size)
buf->out -= buf->size + 1;
pthread_cond_signal(&chan->wrready);
chan_unlock(chan);
} else {
chan_unlock(chan);
error = sock_write(m->socket, &mh, hdrsize);
if (error)
goto bad;
}
goto again;
bad:
if (error == EPIPE)
mux_shutdown(m, strerror(errno), STATUS_TRANSIENTFAILURE);
else
mux_shutdown(m, strerror(errno), STATUS_FAILURE);
return (NULL);
}
static void
sender_cleanup(void *arg)
{
struct mux *m;
m = (struct mux *)arg;
mux_unlock(m);
}
static int
sender_waitforwork(struct mux *m, int *what)
{
int id;
mux_lock(m);
pthread_cleanup_push(sender_cleanup, m);
if (!m->sender_ready) {
pthread_cond_signal(&m->sender_started);
m->sender_ready = 1;
}
while ((id = sender_scan(m, what)) == -1) {
m->sender_waiting = 1;
pthread_cond_wait(&m->sender_newwork, &m->lock);
}
m->sender_waiting = 0;
pthread_cleanup_pop(1);
return (id);
}
/*
* Scan for work to do for the sender. Has to be called with
* the multiplexer lock held.
*/
static int
sender_scan(struct mux *m, int *what)
{
struct chan *chan;
int id;
if (m->nchans <= 0)
return (-1);
id = m->sender_lastid;
do {
id++;
if (id >= m->nchans)
id = 0;
chan = m->channels[id];
chan_lock(chan);
if (chan->state != CS_UNUSED) {
if (chan->sendseq != chan->sendwin &&
buf_count(chan->sendbuf) > 0)
chan->flags |= CF_DATA;
if (chan->flags) {
/* By order of importance. */
if (chan->flags & CF_CONNECT)
*what = CF_CONNECT;
else if (chan->flags & CF_ACCEPT)
*what = CF_ACCEPT;
else if (chan->flags & CF_RESET)
*what = CF_RESET;
else if (chan->flags & CF_WINDOW)
*what = CF_WINDOW;
else if (chan->flags & CF_DATA)
*what = CF_DATA;
else if (chan->flags & CF_CLOSE)
*what = CF_CLOSE;
chan->flags &= ~*what;
chan_unlock(chan);
m->sender_lastid = id;
return (id);
}
}
chan_unlock(chan);
} while (id != m->sender_lastid);
return (-1);
}
/* Read the rest of a packet header depending on its type. */
#define SOCK_READREST(s, mh, hsize) \
sock_readwait(s, (char *)&mh + sizeof(mh.type), (hsize) - sizeof(mh.type))
void *
receiver_loop(void *arg)
{
struct mux_header mh;
struct mux *m;
struct chan *chan;
struct buf *buf;
uint16_t size, len;
int error;
m = (struct mux *)arg;
while ((error = sock_readwait(m->socket, &mh.type,
sizeof(mh.type))) == 0) {
switch (mh.type) {
case MUX_CONNECT:
error = SOCK_READREST(m->socket, mh, MUX_CONNECTHDRSZ);
if (error)
goto bad;
chan = chan_get(m, mh.mh_connect.id);
if (chan->state == CS_LISTENING) {
chan->state = CS_ESTABLISHED;
chan->sendmss = ntohs(mh.mh_connect.mss);
chan->sendwin = ntohl(mh.mh_connect.window);
chan->flags |= CF_ACCEPT;
pthread_cond_signal(&chan->rdready);
} else
chan->flags |= CF_RESET;
chan_unlock(chan);
sender_wakeup(m);
break;
case MUX_ACCEPT:
error = SOCK_READREST(m->socket, mh, MUX_ACCEPTHDRSZ);
if (error)
goto bad;
chan = chan_get(m, mh.mh_accept.id);
if (chan->state == CS_CONNECTING) {
chan->sendmss = ntohs(mh.mh_accept.mss);
chan->sendwin = ntohl(mh.mh_accept.window);
chan->state = CS_ESTABLISHED;
pthread_cond_signal(&chan->wrready);
chan_unlock(chan);
} else {
chan->flags |= CF_RESET;
chan_unlock(chan);
sender_wakeup(m);
}
break;
case MUX_RESET:
error = SOCK_READREST(m->socket, mh, MUX_RESETHDRSZ);
if (error)
goto bad;
goto badproto;
case MUX_WINDOW:
error = SOCK_READREST(m->socket, mh, MUX_WINDOWHDRSZ);
if (error)
goto bad;
chan = chan_get(m, mh.mh_window.id);
if (chan->state == CS_ESTABLISHED ||
chan->state == CS_RDCLOSED) {
chan->sendwin = ntohl(mh.mh_window.window);
chan_unlock(chan);
sender_wakeup(m);
} else {
chan_unlock(chan);
}
break;
case MUX_DATA:
error = SOCK_READREST(m->socket, mh, MUX_DATAHDRSZ);
if (error)
goto bad;
chan = chan_get(m, mh.mh_data.id);
len = ntohs(mh.mh_data.len);
buf = chan->recvbuf;
if ((chan->state != CS_ESTABLISHED &&
chan->state != CS_WRCLOSED) ||
(len > buf_avail(buf) ||
len > chan->recvmss)) {
chan_unlock(chan);
goto badproto;
return (NULL);
}
/*
* Similarly to the sender code, it's safe to
* unlock the channel here.
*/
chan_unlock(chan);
size = min(buf->size + 1 - buf->in, len);
error = sock_readwait(m->socket,
buf->data + buf->in, size);
if (error)
goto bad;
if (len > size) {
/* Wrapping around. */
error = sock_readwait(m->socket,
buf->data, len - size);
if (error)
goto bad;
}
chan_lock(chan);
buf->in += len;
if (buf->in > buf->size)
buf->in -= buf->size + 1;
pthread_cond_signal(&chan->rdready);
chan_unlock(chan);
break;
case MUX_CLOSE:
error = SOCK_READREST(m->socket, mh, MUX_CLOSEHDRSZ);
if (error)
goto bad;
chan = chan_get(m, mh.mh_close.id);
if (chan->state == CS_ESTABLISHED)
chan->state = CS_RDCLOSED;
else if (chan->state == CS_WRCLOSED)
chan->state = CS_CLOSED;
else
goto badproto;
pthread_cond_signal(&chan->rdready);
chan_unlock(chan);
break;
default:
goto badproto;
}
}
bad:
if (errno == ECONNRESET || errno == ECONNABORTED)
mux_shutdown(m, strerror(errno), STATUS_TRANSIENTFAILURE);
else
mux_shutdown(m, strerror(errno), STATUS_FAILURE);
return (NULL);
badproto:
mux_shutdown(m, "Protocol error", STATUS_FAILURE);
return (NULL);
}
/*
* Circular buffers API.
*/
static struct buf *
buf_new(size_t size)
{
struct buf *buf;
buf = xmalloc(sizeof(struct buf));
buf->data = xmalloc(size + 1);
buf->size = size;
buf->in = 0;
buf->out = 0;
return (buf);
}
static void
buf_free(struct buf *buf)
{
free(buf->data);
free(buf);
}
/* Number of bytes stored in the buffer. */
static size_t
buf_count(struct buf *buf)
{
size_t count;
if (buf->in >= buf->out)
count = buf->in - buf->out;
else
count = buf->size + 1 + buf->in - buf->out;
return (count);
}
/* Number of bytes available in the buffer. */
static size_t
buf_avail(struct buf *buf)
{
size_t avail;
if (buf->out > buf->in)
avail = buf->out - buf->in - 1;
else
avail = buf->size + buf->out - buf->in;
return (avail);
}
static void
buf_put(struct buf *buf, const void *data, size_t size)
{
const char *cp;
size_t len;
assert(size > 0);
assert(buf_avail(buf) >= size);
cp = data;
len = buf->size + 1 - buf->in;
if (len < size) {
/* Wrapping around. */
memcpy(buf->data + buf->in, cp, len);
memcpy(buf->data, cp + len, size - len);
} else {
/* Not wrapping around. */
memcpy(buf->data + buf->in, cp, size);
}
buf->in += size;
if (buf->in > buf->size)
buf->in -= buf->size + 1;
}
static void
buf_get(struct buf *buf, void *data, size_t size)
{
char *cp;
size_t len;
assert(size > 0);
assert(buf_count(buf) >= size);
cp = data;
len = buf->size + 1 - buf->out;
if (len < size) {
/* Wrapping around. */
memcpy(cp, buf->data + buf->out, len);
memcpy(cp + len, buf->data, size - len);
} else {
/* Not wrapping around. */
memcpy(cp, buf->data + buf->out, size);
}
buf->out += size;
if (buf->out > buf->size)
buf->out -= buf->size + 1;
}