Mercurial > dropbear
view common-channel.c @ 1938:77bc00dcc19f default tip main master
Bump version to 2022.82
author | Matt Johnston <matt@ucc.asn.au> |
---|---|
date | Fri, 01 Apr 2022 14:43:27 +0800 |
parents | 1d86a58fb52d |
children |
line wrap: on
line source
/* * Dropbear SSH * * Copyright (c) 2002-2004 Matt Johnston * All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to deal * in the Software without restriction, including without limitation the rights * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell * copies of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ /* Handle the multiplexed channels, such as sessions, x11, agent connections */ #include "includes.h" #include "session.h" #include "packet.h" #include "ssh.h" #include "buffer.h" #include "circbuffer.h" #include "dbutil.h" #include "channel.h" #include "listener.h" #include "runopts.h" #include "netio.h" static void send_msg_channel_open_failure(unsigned int remotechan, int reason, const char *text, const char *lang); static void send_msg_channel_open_confirmation(const struct Channel* channel, unsigned int recvwindow, unsigned int recvmaxpacket); static int writechannel(struct Channel* channel, int fd, circbuffer *cbuf, const unsigned char *moredata, unsigned int *morelen); static void send_msg_channel_window_adjust(const struct Channel *channel, unsigned int incr); static void send_msg_channel_data(struct Channel *channel, int isextended); static void send_msg_channel_eof(struct Channel *channel); static void send_msg_channel_close(struct Channel *channel); static void remove_channel(struct Channel *channel); static unsigned int write_pending(const struct Channel * channel); static void check_close(struct Channel *channel); static void close_chan_fd(struct Channel *channel, int fd, int how); #define FD_UNINIT (-2) #define FD_CLOSED (-1) #define ERRFD_IS_READ(channel) ((channel)->extrabuf == NULL) #define ERRFD_IS_WRITE(channel) (!ERRFD_IS_READ(channel)) /* allow space for: * 1 byte byte SSH_MSG_CHANNEL_DATA * 4 bytes uint32 recipient channel * 4 bytes string data */ #define RECV_MAX_CHANNEL_DATA_LEN (RECV_MAX_PAYLOAD_LEN-(1+4+4)) /* Initialise all the channels */ void chaninitialise(const struct ChanType *chantypes[]) { /* may as well create space for a single channel */ ses.channels = (struct Channel**)m_malloc(sizeof(struct Channel*)); ses.chansize = 1; ses.channels[0] = NULL; ses.chancount = 0; ses.chantypes = chantypes; #if DROPBEAR_LISTENERS listeners_initialise(); #endif } /* Clean up channels, freeing allocated memory */ void chancleanup() { unsigned int i; TRACE(("enter chancleanup")) for (i = 0; i < ses.chansize; i++) { if (ses.channels[i] != NULL) { TRACE(("channel %d closing", i)) remove_channel(ses.channels[i]); } } m_free(ses.channels); TRACE(("leave chancleanup")) } /* Create a new channel entry, send a reply confirm or failure */ /* If remotechan, transwindow and transmaxpacket are not know (for a new * outgoing connection, with them to be filled on confirmation), they should * all be set to 0 */ static struct Channel* newchannel(unsigned int remotechan, const struct ChanType *type, unsigned int transwindow, unsigned int transmaxpacket) { struct Channel * newchan; unsigned int i, j; TRACE(("enter newchannel")) /* first see if we can use existing channels */ for (i = 0; i < ses.chansize; i++) { if (ses.channels[i] == NULL) { break; } } /* otherwise extend the list */ if (i == ses.chansize) { if (ses.chansize >= MAX_CHANNELS) { TRACE(("leave newchannel: max chans reached")) return NULL; } /* extend the channels */ ses.channels = (struct Channel**)m_realloc(ses.channels, (ses.chansize+CHAN_EXTEND_SIZE)*sizeof(struct Channel*)); ses.chansize += CHAN_EXTEND_SIZE; /* set the new channels to null */ for (j = i; j < ses.chansize; j++) { ses.channels[j] = NULL; } } newchan = (struct Channel*)m_malloc(sizeof(struct Channel)); newchan->type = type; newchan->index = i; newchan->sent_close = newchan->recv_close = 0; newchan->sent_eof = newchan->recv_eof = 0; newchan->remotechan = remotechan; newchan->transwindow = transwindow; newchan->transmaxpacket = transmaxpacket; newchan->typedata = NULL; newchan->writefd = FD_UNINIT; newchan->readfd = FD_UNINIT; newchan->errfd = FD_CLOSED; /* this isn't always set to start with */ newchan->await_open = 0; newchan->writebuf = cbuf_new(opts.recv_window); newchan->recvwindow = opts.recv_window; newchan->extrabuf = NULL; /* The user code can set it up */ newchan->recvdonelen = 0; newchan->recvmaxpacket = RECV_MAX_CHANNEL_DATA_LEN; newchan->prio = DROPBEAR_PRIO_NORMAL; ses.channels[i] = newchan; ses.chancount++; TRACE(("leave newchannel")) return newchan; } /* Returns the channel structure corresponding to the channel in the current * data packet (ses.payload must be positioned appropriately). * A valid channel is always returns, it will fail fatally with an unknown * channel */ static struct Channel* getchannel_msg(const char* kind) { unsigned int chan; chan = buf_getint(ses.payload); if (chan >= ses.chansize || ses.channels[chan] == NULL) { if (kind) { dropbear_exit("%s for unknown channel %d", kind, chan); } else { dropbear_exit("Unknown channel %d", chan); } } return ses.channels[chan]; } struct Channel* getchannel() { return getchannel_msg(NULL); } /* Iterate through the channels, performing IO if available */ void channelio(const fd_set *readfds, const fd_set *writefds) { /* Listeners such as TCP, X11, agent-auth */ struct Channel *channel; unsigned int i; /* foreach channel */ for (i = 0; i < ses.chansize; i++) { /* Close checking only needs to occur for channels that had IO events */ int do_check_close = 0; channel = ses.channels[i]; if (channel == NULL) { /* only process in-use channels */ continue; } /* read data and send it over the wire */ if (channel->readfd >= 0 && FD_ISSET(channel->readfd, readfds)) { TRACE(("send normal readfd")) send_msg_channel_data(channel, 0); do_check_close = 1; } /* read stderr data and send it over the wire */ if (ERRFD_IS_READ(channel) && channel->errfd >= 0 && FD_ISSET(channel->errfd, readfds)) { TRACE(("send normal errfd")) send_msg_channel_data(channel, 1); do_check_close = 1; } /* write to program/pipe stdin */ if (channel->writefd >= 0 && FD_ISSET(channel->writefd, writefds)) { writechannel(channel, channel->writefd, channel->writebuf, NULL, NULL); do_check_close = 1; } /* stderr for client mode */ if (ERRFD_IS_WRITE(channel) && channel->errfd >= 0 && FD_ISSET(channel->errfd, writefds)) { writechannel(channel, channel->errfd, channel->extrabuf, NULL, NULL); do_check_close = 1; } if (ses.channel_signal_pending) { /* SIGCHLD can change channel state for server sessions */ do_check_close = 1; } /* handle any channel closing etc */ if (do_check_close) { check_close(channel); } } #if DROPBEAR_LISTENERS handle_listeners(readfds); #endif } /* Returns true if there is data remaining to be written to stdin or * stderr of a channel's endpoint. */ static unsigned int write_pending(const struct Channel * channel) { if (channel->writefd >= 0 && cbuf_getused(channel->writebuf) > 0) { return 1; } else if (channel->errfd >= 0 && channel->extrabuf && cbuf_getused(channel->extrabuf) > 0) { return 1; } return 0; } /* EOF/close handling */ static void check_close(struct Channel *channel) { int close_allowed = 0; TRACE2(("check_close: writefd %d, readfd %d, errfd %d, sent_close %d, recv_close %d", channel->writefd, channel->readfd, channel->errfd, channel->sent_close, channel->recv_close)) TRACE2(("writebuf size %d extrabuf size %d", channel->writebuf ? cbuf_getused(channel->writebuf) : 0, channel->extrabuf ? cbuf_getused(channel->extrabuf) : 0)) /* if a type-specific check_close is defined we will only exit once that has been triggered. this is only used for a server "session" channel, to ensure that the shell has exited (and the exit status retrieved) before we close things up. */ if (!channel->type->check_close || channel->sent_close || channel->type->check_close(channel)) { close_allowed = 1; } /* In flushing mode we close FDs as soon as pipes are empty. This is used to drain out FDs when the process exits, in the case where the FD doesn't have EOF - "sleep 10&echo hello" case */ if (channel->flushing) { if (channel->readfd >= 0 && !fd_read_pending(channel->readfd)) { close_chan_fd(channel, channel->readfd, SHUT_RD); } if (ERRFD_IS_READ(channel) && channel->errfd >= 0 && !fd_read_pending(channel->errfd)) { close_chan_fd(channel, channel->errfd, SHUT_RD); } } if (channel->recv_close && !write_pending(channel) && close_allowed) { if (!channel->sent_close) { TRACE(("Sending MSG_CHANNEL_CLOSE in response to same.")) send_msg_channel_close(channel); } remove_channel(channel); return; } if ((channel->recv_eof && !write_pending(channel)) /* have a server "session" and child has exited */ || (channel->type->check_close && close_allowed)) { close_chan_fd(channel, channel->writefd, SHUT_WR); } /* If we're not going to send any more data, send EOF */ if (!channel->sent_eof && channel->readfd == FD_CLOSED && (ERRFD_IS_WRITE(channel) || channel->errfd == FD_CLOSED)) { send_msg_channel_eof(channel); } /* And if we can't receive any more data from them either, close up */ if (channel->readfd == FD_CLOSED && channel->writefd == FD_CLOSED && (ERRFD_IS_WRITE(channel) || channel->errfd == FD_CLOSED) && !channel->sent_close && close_allowed && !write_pending(channel)) { TRACE(("sending close, readfd is closed")) send_msg_channel_close(channel); } } /* Check whether a deferred (EINPROGRESS) connect() was successful, and * if so, set up the channel properly. Otherwise, the channel is cleaned up, so * it is important that the channel reference isn't used after a call to this * function */ void channel_connect_done(int result, int sock, void* user_data, const char* errstring) { struct Channel *channel = user_data; TRACE(("enter channel_connect_done")) if (result == DROPBEAR_SUCCESS) { channel->readfd = channel->writefd = sock; channel->bidir_fd = 1; channel->conn_pending = NULL; send_msg_channel_open_confirmation(channel, channel->recvwindow, channel->recvmaxpacket); TRACE(("leave channel_connect_done: success")) } else { send_msg_channel_open_failure(channel->remotechan, SSH_OPEN_CONNECT_FAILED, errstring, ""); remove_channel(channel); TRACE(("leave check_in_progress: fail. internal errstring: %s", errstring)) } } /* Send the close message and set the channel as closed */ static void send_msg_channel_close(struct Channel *channel) { TRACE(("enter send_msg_channel_close %p", (void*)channel)) if (channel->type->closehandler) { channel->type->closehandler(channel); } CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_CLOSE); buf_putint(ses.writepayload, channel->remotechan); encrypt_packet(); channel->sent_eof = 1; channel->sent_close = 1; close_chan_fd(channel, channel->readfd, SHUT_RD); close_chan_fd(channel, channel->errfd, SHUT_RDWR); close_chan_fd(channel, channel->writefd, SHUT_WR); TRACE(("leave send_msg_channel_close")) } /* call this when trans/eof channels are closed */ static void send_msg_channel_eof(struct Channel *channel) { TRACE(("enter send_msg_channel_eof")) CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_EOF); buf_putint(ses.writepayload, channel->remotechan); encrypt_packet(); channel->sent_eof = 1; TRACE(("leave send_msg_channel_eof")) } #ifndef HAVE_WRITEV static int writechannel_fallback(struct Channel* channel, int fd, circbuffer *cbuf, const unsigned char *UNUSED(moredata), unsigned int *morelen) { unsigned char *circ_p1, *circ_p2; unsigned int circ_len1, circ_len2; ssize_t written; if (morelen) { /* fallback doesn't consume moredata */ *morelen = 0; } /* Write the first portion of the circular buffer */ cbuf_readptrs(cbuf, &circ_p1, &circ_len1, &circ_p2, &circ_len2); written = write(fd, circ_p1, circ_len1); if (written < 0) { if (errno != EINTR && errno != EAGAIN) { TRACE(("channel IO write error fd %d %s", fd, strerror(errno))) close_chan_fd(channel, fd, SHUT_WR); return DROPBEAR_FAILURE; } } else { cbuf_incrread(cbuf, written); channel->recvdonelen += written; } return DROPBEAR_SUCCESS; } #endif /* !HAVE_WRITEV */ #ifdef HAVE_WRITEV static int writechannel_writev(struct Channel* channel, int fd, circbuffer *cbuf, const unsigned char *moredata, unsigned int *morelen) { struct iovec iov[3]; unsigned char *circ_p1, *circ_p2; unsigned int circ_len1, circ_len2; int io_count = 0; ssize_t written; cbuf_readptrs(cbuf, &circ_p1, &circ_len1, &circ_p2, &circ_len2); if (circ_len1 > 0) { TRACE(("circ1 %d", circ_len1)) iov[io_count].iov_base = circ_p1; iov[io_count].iov_len = circ_len1; io_count++; } if (circ_len2 > 0) { TRACE(("circ2 %d", circ_len2)) iov[io_count].iov_base = circ_p2; iov[io_count].iov_len = circ_len2; io_count++; } if (morelen) { assert(moredata); TRACE(("more %d", *morelen)) iov[io_count].iov_base = (void*)moredata; iov[io_count].iov_len = *morelen; io_count++; } if (io_count == 0) { /* writechannel may sometimes be called twice in a main loop iteration. From common_recv_msg_channel_data() then channelio(). The second call may not have any data to write, so we just return. */ TRACE(("leave writechannel, no data")) return DROPBEAR_SUCCESS; } if (morelen) { /* Default return value, none consumed */ *morelen = 0; } written = writev(fd, iov, io_count); if (written < 0) { if (errno != EINTR && errno != EAGAIN) { TRACE(("channel IO write error fd %d %s", fd, strerror(errno))) close_chan_fd(channel, fd, SHUT_WR); return DROPBEAR_FAILURE; } } else { int cbuf_written = MIN(circ_len1+circ_len2, (unsigned int)written); cbuf_incrread(cbuf, cbuf_written); if (morelen) { *morelen = written - cbuf_written; } channel->recvdonelen += written; } return DROPBEAR_SUCCESS; } #endif /* HAVE_WRITEV */ /* Called to write data out to the local side of the channel. Writes the circular buffer contents and also the "moredata" buffer if not null. Will ignore EAGAIN. Returns DROPBEAR_FAILURE if writing to fd had an error and the channel is being closed, DROPBEAR_SUCCESS otherwise */ static int writechannel(struct Channel* channel, int fd, circbuffer *cbuf, const unsigned char *moredata, unsigned int *morelen) { int ret = DROPBEAR_SUCCESS; TRACE(("enter writechannel fd %d", fd)) #ifdef HAVE_WRITEV ret = writechannel_writev(channel, fd, cbuf, moredata, morelen); #else ret = writechannel_fallback(channel, fd, cbuf, moredata, morelen); #endif /* Window adjust handling */ if (channel->recvdonelen >= RECV_WINDOWEXTEND) { send_msg_channel_window_adjust(channel, channel->recvdonelen); channel->recvwindow += channel->recvdonelen; channel->recvdonelen = 0; } dropbear_assert(channel->recvwindow <= opts.recv_window); dropbear_assert(channel->recvwindow <= cbuf_getavail(channel->writebuf)); dropbear_assert(channel->extrabuf == NULL || channel->recvwindow <= cbuf_getavail(channel->extrabuf)); TRACE(("leave writechannel")) return ret; } /* Set the file descriptors for the main select in session.c * This avoid channels which don't have any window available, are closed, etc*/ void setchannelfds(fd_set *readfds, fd_set *writefds, int allow_reads) { unsigned int i; struct Channel * channel; for (i = 0; i < ses.chansize; i++) { channel = ses.channels[i]; if (channel == NULL) { continue; } /* Stuff to put over the wire. Avoid queueing data to send if we're in the middle of a key re-exchange (!dataallowed), but still read from the FD if there's the possibility of "~."" to kill an interactive session (the read_mangler) */ if (channel->transwindow > 0 && ((ses.dataallowed && allow_reads) || channel->read_mangler)) { if (channel->readfd >= 0) { FD_SET(channel->readfd, readfds); } if (ERRFD_IS_READ(channel) && channel->errfd >= 0) { FD_SET(channel->errfd, readfds); } } /* Stuff from the wire */ if (channel->writefd >= 0 && cbuf_getused(channel->writebuf) > 0) { FD_SET(channel->writefd, writefds); } if (ERRFD_IS_WRITE(channel) && channel->errfd >= 0 && cbuf_getused(channel->extrabuf) > 0) { FD_SET(channel->errfd, writefds); } } /* foreach channel */ #if DROPBEAR_LISTENERS set_listener_fds(readfds); #endif } /* handle the channel EOF event, by closing the channel filedescriptor. The * channel isn't closed yet, it is left until the incoming (from the program * etc) FD is also EOF */ void recv_msg_channel_eof() { struct Channel * channel; TRACE(("enter recv_msg_channel_eof")) channel = getchannel_msg("EOF"); channel->recv_eof = 1; check_close(channel); TRACE(("leave recv_msg_channel_eof")) } /* Handle channel closure(), respond in kind and close the channels */ void recv_msg_channel_close() { struct Channel * channel; TRACE(("enter recv_msg_channel_close")) channel = getchannel_msg("Close"); channel->recv_eof = 1; channel->recv_close = 1; check_close(channel); TRACE(("leave recv_msg_channel_close")) } /* Remove a channel entry, this is only executed after both sides have sent * channel close */ static void remove_channel(struct Channel * channel) { TRACE(("enter remove_channel")) TRACE(("channel index is %d", channel->index)) cbuf_free(channel->writebuf); channel->writebuf = NULL; if (channel->extrabuf) { cbuf_free(channel->extrabuf); channel->extrabuf = NULL; } if (IS_DROPBEAR_SERVER || (channel->writefd != STDOUT_FILENO)) { /* close the FDs in case they haven't been done * yet (they might have been shutdown etc) */ TRACE(("CLOSE writefd %d", channel->writefd)) m_close(channel->writefd); TRACE(("CLOSE readfd %d", channel->readfd)) m_close(channel->readfd); TRACE(("CLOSE errfd %d", channel->errfd)) m_close(channel->errfd); } if (channel->type->cleanup) { channel->type->cleanup(channel); } if (channel->conn_pending) { cancel_connect(channel->conn_pending); } ses.channels[channel->index] = NULL; m_free(channel); ses.chancount--; update_channel_prio(); TRACE(("leave remove_channel")) } /* Handle channel specific requests, passing off to corresponding handlers * such as chansession or x11fwd */ void recv_msg_channel_request() { struct Channel *channel; channel = getchannel(); TRACE(("enter recv_msg_channel_request %p", (void*)channel)) if (channel->type->reqhandler) { channel->type->reqhandler(channel); } else { int wantreply; buf_eatstring(ses.payload); wantreply = buf_getbool(ses.payload); if (wantreply) { send_msg_channel_failure(channel); } } TRACE(("leave recv_msg_channel_request")) } /* Reads data from the server's program/shell/etc, and puts it in a * channel_data packet to send. * chan is the remote channel, isextended is 0 if it is normal data, 1 * if it is extended data. if it is extended, then the type is in * exttype */ static void send_msg_channel_data(struct Channel *channel, int isextended) { int len; size_t maxlen, size_pos; int fd; CHECKCLEARTOWRITE(); TRACE(("enter send_msg_channel_data")) dropbear_assert(!channel->sent_close); if (isextended) { fd = channel->errfd; } else { fd = channel->readfd; } TRACE(("enter send_msg_channel_data isextended %d fd %d", isextended, fd)) dropbear_assert(fd >= 0); maxlen = MIN(channel->transwindow, channel->transmaxpacket); /* -(1+4+4) is SSH_MSG_CHANNEL_DATA, channel number, string length, and * exttype if is extended */ maxlen = MIN(maxlen, ses.writepayload->size - 1 - 4 - 4 - (isextended ? 4 : 0)); TRACE(("maxlen %zd", maxlen)) if (maxlen == 0) { TRACE(("leave send_msg_channel_data: no window")) return; } buf_putbyte(ses.writepayload, isextended ? SSH_MSG_CHANNEL_EXTENDED_DATA : SSH_MSG_CHANNEL_DATA); buf_putint(ses.writepayload, channel->remotechan); if (isextended) { buf_putint(ses.writepayload, SSH_EXTENDED_DATA_STDERR); } /* a dummy size first ...*/ size_pos = ses.writepayload->pos; buf_putint(ses.writepayload, 0); /* read the data */ len = read(fd, buf_getwriteptr(ses.writepayload, maxlen), maxlen); if (len <= 0) { if (len == 0 || errno != EINTR) { /* This will also get hit in the case of EAGAIN. The only time we expect to receive EAGAIN is when we're flushing a FD, in which case it can be treated the same as EOF */ close_chan_fd(channel, fd, SHUT_RD); } buf_setpos(ses.writepayload, 0); buf_setlen(ses.writepayload, 0); TRACE(("leave send_msg_channel_data: len %d read err %d or EOF for fd %d", len, errno, fd)) return; } if (channel->read_mangler) { channel->read_mangler(channel, buf_getwriteptr(ses.writepayload, len), &len); if (len == 0) { buf_setpos(ses.writepayload, 0); buf_setlen(ses.writepayload, 0); return; } } TRACE(("send_msg_channel_data: len %d fd %d", len, fd)) buf_incrwritepos(ses.writepayload, len); /* ... real size here */ buf_setpos(ses.writepayload, size_pos); buf_putint(ses.writepayload, len); channel->transwindow -= len; encrypt_packet(); TRACE(("leave send_msg_channel_data")) } /* We receive channel data */ void recv_msg_channel_data() { struct Channel *channel; channel = getchannel(); common_recv_msg_channel_data(channel, channel->writefd, channel->writebuf); } /* Shared for data and stderr data - when we receive data, put it in a buffer * for writing to the local file descriptor */ void common_recv_msg_channel_data(struct Channel *channel, int fd, circbuffer * cbuf) { unsigned int datalen; unsigned int maxdata; unsigned int buflen; unsigned int len; unsigned int consumed; int res; TRACE(("enter recv_msg_channel_data")) if (channel->recv_eof) { dropbear_exit("Received data after eof"); } if (fd < 0 || !cbuf) { /* If we have encountered failed write, the far side might still * be sending data without having yet received our close notification. * We just drop the data. */ return; } datalen = buf_getint(ses.payload); TRACE(("length %d", datalen)) maxdata = cbuf_getavail(cbuf); /* Whilst the spec says we "MAY ignore data past the end" this could * lead to corrupted file transfers etc (chunks missed etc). It's better to * just die horribly */ if (datalen > maxdata) { dropbear_exit("Oversized packet"); } dropbear_assert(channel->recvwindow >= datalen); channel->recvwindow -= datalen; dropbear_assert(channel->recvwindow <= opts.recv_window); /* Attempt to write the data immediately without having to put it in the circular buffer */ consumed = datalen; res = writechannel(channel, fd, cbuf, buf_getptr(ses.payload, datalen), &consumed); datalen -= consumed; buf_incrpos(ses.payload, consumed); /* We may have to run throught twice, if the buffer wraps around. Can't * just "leave it for next time" like with writechannel, since this * is payload data. * If the writechannel() failed then remaining data is discarded */ if (res == DROPBEAR_SUCCESS) { len = datalen; while (len > 0) { buflen = cbuf_writelen(cbuf); buflen = MIN(buflen, len); memcpy(cbuf_writeptr(cbuf, buflen), buf_getptr(ses.payload, buflen), buflen); cbuf_incrwrite(cbuf, buflen); buf_incrpos(ses.payload, buflen); len -= buflen; } } TRACE(("leave recv_msg_channel_data")) } /* Increment the outgoing data window for a channel - the remote end limits * the amount of data which may be transmitted, this window is decremented * as data is sent, and incremented upon receiving window-adjust messages */ void recv_msg_channel_window_adjust() { struct Channel * channel; unsigned int incr; channel = getchannel(); incr = buf_getint(ses.payload); TRACE(("received window increment %d", incr)) incr = MIN(incr, TRANS_MAX_WIN_INCR); channel->transwindow += incr; channel->transwindow = MIN(channel->transwindow, TRANS_MAX_WINDOW); } /* Increment the incoming data window for a channel, and let the remote * end know */ static void send_msg_channel_window_adjust(const struct Channel* channel, unsigned int incr) { TRACE(("sending window adjust %d", incr)) CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_WINDOW_ADJUST); buf_putint(ses.writepayload, channel->remotechan); buf_putint(ses.writepayload, incr); encrypt_packet(); } /* Handle a new channel request, performing any channel-type-specific setup */ void recv_msg_channel_open() { char *type; unsigned int typelen; unsigned int remotechan, transwindow, transmaxpacket; struct Channel *channel; const struct ChanType **cp; const struct ChanType *chantype; unsigned int errtype = SSH_OPEN_UNKNOWN_CHANNEL_TYPE; int ret; TRACE(("enter recv_msg_channel_open")) /* get the packet contents */ type = buf_getstring(ses.payload, &typelen); remotechan = buf_getint(ses.payload); transwindow = buf_getint(ses.payload); transwindow = MIN(transwindow, TRANS_MAX_WINDOW); transmaxpacket = buf_getint(ses.payload); transmaxpacket = MIN(transmaxpacket, TRANS_MAX_PAYLOAD_LEN); /* figure what type of packet it is */ if (typelen > MAX_NAME_LEN) { goto failure; } /* Get the channel type. Client and server style invokation will set up a * different list for ses.chantypes at startup. We just iterate through * this list and find the matching name */ for (cp = &ses.chantypes[0], chantype = (*cp); chantype != NULL; cp++, chantype = (*cp)) { if (strcmp(type, chantype->name) == 0) { break; } } if (chantype == NULL) { TRACE(("No matching type for '%s'", type)) goto failure; } TRACE(("matched type '%s'", type)) /* create the channel */ channel = newchannel(remotechan, chantype, transwindow, transmaxpacket); if (channel == NULL) { TRACE(("newchannel returned NULL")) errtype = SSH_OPEN_RESOURCE_SHORTAGE; goto failure; } if (channel->type->inithandler) { ret = channel->type->inithandler(channel); if (ret == SSH_OPEN_IN_PROGRESS) { /* We'll send the confirmation later */ goto cleanup; } if (ret > 0) { errtype = ret; remove_channel(channel); TRACE(("inithandler returned failure %d", ret)) goto failure; } } update_channel_prio(); /* success */ send_msg_channel_open_confirmation(channel, channel->recvwindow, channel->recvmaxpacket); goto cleanup; failure: TRACE(("recv_msg_channel_open failure")) send_msg_channel_open_failure(remotechan, errtype, "", ""); cleanup: m_free(type); TRACE(("leave recv_msg_channel_open")) } /* Send a failure message */ void send_msg_channel_failure(const struct Channel *channel) { TRACE(("enter send_msg_channel_failure")) if (channel->sent_close) { TRACE(("Skipping sending msg_channel_failure for closed channel")) return; } CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_FAILURE); buf_putint(ses.writepayload, channel->remotechan); encrypt_packet(); TRACE(("leave send_msg_channel_failure")) } /* Send a success message */ void send_msg_channel_success(const struct Channel *channel) { TRACE(("enter send_msg_channel_success")) if (channel->sent_close) { TRACE(("Skipping sending msg_channel_success for closed channel")) return; } CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_SUCCESS); buf_putint(ses.writepayload, channel->remotechan); encrypt_packet(); TRACE(("leave send_msg_channel_success")) } /* Send a channel open failure message, with a corresponding reason * code (usually resource shortage or unknown chan type) */ static void send_msg_channel_open_failure(unsigned int remotechan, int reason, const char *text, const char *lang) { TRACE(("enter send_msg_channel_open_failure")) CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_OPEN_FAILURE); buf_putint(ses.writepayload, remotechan); buf_putint(ses.writepayload, reason); buf_putstring(ses.writepayload, text, strlen(text)); buf_putstring(ses.writepayload, lang, strlen(lang)); encrypt_packet(); TRACE(("leave send_msg_channel_open_failure")) } /* Confirm a channel open, and let the remote end know what number we've * allocated and the receive parameters */ static void send_msg_channel_open_confirmation(const struct Channel* channel, unsigned int recvwindow, unsigned int recvmaxpacket) { TRACE(("enter send_msg_channel_open_confirmation")) CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_OPEN_CONFIRMATION); buf_putint(ses.writepayload, channel->remotechan); buf_putint(ses.writepayload, channel->index); buf_putint(ses.writepayload, recvwindow); buf_putint(ses.writepayload, recvmaxpacket); encrypt_packet(); TRACE(("leave send_msg_channel_open_confirmation")) } /* close a fd, how is SHUT_RD or SHUT_WR */ static void close_chan_fd(struct Channel *channel, int fd, int how) { int closein = 0, closeout = 0; if (channel->bidir_fd) { TRACE(("SHUTDOWN(%d, %d)", fd, how)) shutdown(fd, how); if (how == 0) { closeout = 1; } else { closein = 1; } } else { TRACE(("CLOSE some fd %d", fd)) m_close(fd); closein = closeout = 1; } if (closeout && (fd == channel->readfd)) { channel->readfd = FD_CLOSED; } if (closeout && ERRFD_IS_READ(channel) && (fd == channel->errfd)) { channel->errfd = FD_CLOSED; } if (closein && fd == channel->writefd) { channel->writefd = FD_CLOSED; } if (closein && ERRFD_IS_WRITE(channel) && (fd == channel->errfd)) { channel->errfd = FD_CLOSED; } /* if we called shutdown on it and all references are gone, then we * need to close() it to stop it lingering */ if (channel->bidir_fd && channel->readfd == FD_CLOSED && channel->writefd == FD_CLOSED && channel->errfd == FD_CLOSED) { TRACE(("CLOSE (finally) of %d", fd)) m_close(fd); } } #if (DROPBEAR_LISTENERS) || (DROPBEAR_CLIENT) /* Create a new channel, and start the open request. This is intended * for X11, agent, tcp forwarding, and should be filled with channel-specific * options, with the calling function calling encrypt_packet() after * completion. It is mandatory for the caller to encrypt_packet() if * a channel is returned. NULL is returned on failure. */ int send_msg_channel_open_init(int fd, const struct ChanType *type) { struct Channel* chan; TRACE(("enter send_msg_channel_open_init()")) chan = newchannel(0, type, 0, 0); if (!chan) { TRACE(("leave send_msg_channel_open_init() - FAILED in newchannel()")) return DROPBEAR_FAILURE; } /* Outbound opened channels don't make use of in-progress connections, * we can set it up straight away */ /* set fd non-blocking */ setnonblocking(fd); chan->writefd = chan->readfd = fd; ses.maxfd = MAX(ses.maxfd, fd); chan->bidir_fd = 1; chan->await_open = 1; /* now open the channel connection */ CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_OPEN); buf_putstring(ses.writepayload, type->name, strlen(type->name)); buf_putint(ses.writepayload, chan->index); buf_putint(ses.writepayload, opts.recv_window); buf_putint(ses.writepayload, RECV_MAX_CHANNEL_DATA_LEN); TRACE(("leave send_msg_channel_open_init()")) return DROPBEAR_SUCCESS; } /* Confirmation that our channel open request was * successful*/ void recv_msg_channel_open_confirmation() { struct Channel * channel; int ret; TRACE(("enter recv_msg_channel_open_confirmation")) channel = getchannel(); if (!channel->await_open) { dropbear_exit("Unexpected channel reply"); } channel->await_open = 0; channel->remotechan = buf_getint(ses.payload); channel->transwindow = buf_getint(ses.payload); channel->transmaxpacket = buf_getint(ses.payload); TRACE(("new chan remote %d local %d", channel->remotechan, channel->index)) /* Run the inithandler callback */ if (channel->type->inithandler) { ret = channel->type->inithandler(channel); if (ret > 0) { remove_channel(channel); TRACE(("inithandler returned failure %d", ret)) return; } } update_channel_prio(); TRACE(("leave recv_msg_channel_open_confirmation")) } /* Notification that our channel open request failed */ void recv_msg_channel_open_failure() { struct Channel * channel; channel = getchannel(); if (!channel->await_open) { dropbear_exit("Unexpected channel reply"); } channel->await_open = 0; remove_channel(channel); } #endif /* DROPBEAR_LISTENERS */ void send_msg_request_success() { CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_REQUEST_SUCCESS); encrypt_packet(); } void send_msg_request_failure() { CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_REQUEST_FAILURE); encrypt_packet(); } struct Channel* get_any_ready_channel() { size_t i; if (ses.chancount == 0) { return NULL; } for (i = 0; i < ses.chansize; i++) { struct Channel *chan = ses.channels[i]; if (chan && !(chan->sent_eof || chan->recv_eof) && !(chan->await_open)) { return chan; } } return NULL; } void start_send_channel_request(const struct Channel *channel, const char *type) { CHECKCLEARTOWRITE(); buf_putbyte(ses.writepayload, SSH_MSG_CHANNEL_REQUEST); buf_putint(ses.writepayload, channel->remotechan); buf_putstring(ses.writepayload, type, strlen(type)); }