if (!(expint = expand_string(istring)))
{
- if (expand_string_forcedfail) return TRUE;
+ if (f.expand_string_forcedfail) return TRUE;
addr->transport_return = PANIC;
addr->message = string_sprintf("failed to expand \"interface\" "
"option for %s: %s", msg, expand_string_message);
struct tcp_info tinfo;
socklen_t len = sizeof(tinfo);
-if (getsockopt(sock, IPPROTO_TCP, TCP_INFO, &tinfo, &len) == 0)
+switch (tcp_out_fastopen)
{
- switch (tcp_out_fastopen)
- {
- /* This is a somewhat dubious detection method; totally undocumented so likely
- to fail in future kernels. There seems to be no documented way. What we really
- want to know is if the server sent smtp-banner data before our ACK of his SYN,ACK
- hit him. What this (possibly?) detects is whether we sent a TFO cookie with our
- SYN, as distinct from a TFO request. This gets a false-positive when the server
- key is rotated; we send the old one (which this test sees) but the server returns
- the new one and does not send its SMTP banner before we ACK his SYN,ACK.
- To force that rotation case:
- '# echo -n "00000000-00000000-00000000-0000000" >/proc/sys/net/ipv4/tcp_fastopen_key'
- The kernel seems to be counting unack'd packets. */
-
- case 1:
- if (tinfo.tcpi_unacked > 1)
- {
- DEBUG(D_transport|D_v)
- debug_printf("TCP_FASTOPEN tcpi_unacked %d\n", tinfo.tcpi_unacked);
- tcp_out_fastopen = 2;
- }
- break;
-
-#ifdef notdef /* This seems to always fire, meaning that we cannot tell
- whether the server accepted data we sent. For now assume
- that it did. */
+ /* This is a somewhat dubious detection method; totally undocumented so likely
+ to fail in future kernels. There seems to be no documented way. What we really
+ want to know is if the server sent smtp-banner data before our ACK of his SYN,ACK
+ hit him. What this (possibly?) detects is whether we sent a TFO cookie with our
+ SYN, as distinct from a TFO request. This gets a false-positive when the server
+ key is rotated; we send the old one (which this test sees) but the server returns
+ the new one and does not send its SMTP banner before we ACK his SYN,ACK.
+ To force that rotation case:
+ '# echo -n "00000000-00000000-00000000-0000000" >/proc/sys/net/ipv4/tcp_fastopen_key'
+ The kernel seems to be counting unack'd packets. */
+
+ case TFO_ATTEMPTED_NODATA:
+ if ( getsockopt(sock, IPPROTO_TCP, TCP_INFO, &tinfo, &len) == 0
+ && tinfo.tcpi_state == TCP_SYN_SENT
+ && tinfo.tcpi_unacked > 1
+ )
+ {
+ DEBUG(D_transport|D_v)
+ debug_printf("TCP_FASTOPEN tcpi_unacked %d\n", tinfo.tcpi_unacked);
+ tcp_out_fastopen = TFO_USED_NODATA;
+ }
+ break;
- /* If there was data-on-SYN but we had to retrasnmit it, declare no TFO */
+ /* When called after waiting for received data we should be able
+ to tell if data we sent was accepted. */
- case 2:
- if (!(tinfo.tcpi_options & TCPI_OPT_SYN_DATA))
+ case TFO_ATTEMPTED_DATA:
+ if ( getsockopt(sock, IPPROTO_TCP, TCP_INFO, &tinfo, &len) == 0
+ && tinfo.tcpi_state == TCP_ESTABLISHED
+ )
+ if (tinfo.tcpi_options & TCPI_OPT_SYN_DATA)
+ {
+ DEBUG(D_transport|D_v) debug_printf("TFO: data was acked\n");
+ tcp_out_fastopen = TFO_USED_DATA;
+ }
+ else
{
DEBUG(D_transport|D_v) debug_printf("TFO: had to retransmit\n");
- tcp_out_fastopen = 0;
+ tcp_out_fastopen = TFO_NOT_USED;
}
- break;
-#endif
- }
-
+ break;
}
# endif
}
/* Arguments as for smtp_connect(), plus
- early_data if non-NULL, data to be sent - preferably in the TCP SYN segment
+ early_data if non-NULL, idenmpotent data to be sent -
+ preferably in the TCP SYN segment
Returns: connected socket number, or -1 with errno set
*/
/* Connect to the remote host, and add keepalive to the socket before returning
it, if requested. If the build supports TFO, request it - and if the caller
-requested some early-data then include that in the TFO request. */
+requested some early-data then include that in the TFO request. If there is
+early-data but no TFO support, send it after connecting. */
else
{
#ifdef TCP_FASTOPEN
- if (verify_check_given_host(&ob->hosts_try_fastopen, host) == OK)
+ if (verify_check_given_host(CUSS &ob->hosts_try_fastopen, host) == OK)
fastopen_blob = early_data ? early_data : &tcp_fastopen_nodata;
#endif
if (ip_connect(sock, host_af, host->address, port, timeout, fastopen_blob) < 0)
save_errno = errno;
else if (early_data && !fastopen_blob && early_data->data && early_data->len)
+ {
+ HDEBUG(D_transport|D_acl|D_v)
+ debug_printf("sending %ld nonTFO early-data\n", (long)early_data->len);
+
+#ifdef TCP_QUICKACK
+ (void) setsockopt(sock, IPPROTO_TCP, TCP_QUICKACK, US &off, sizeof(off));
+#endif
if (send(sock, early_data->data, early_data->len, 0) < 0)
save_errno = errno;
+ }
}
/* Either bind() or connect() failed */
return -1;
}
-/* Both bind() and connect() succeeded */
+/* Both bind() and connect() succeeded, and any early-data */
else
{
union sockaddr_46 interface_sock;
EXIM_SOCKLEN_T size = sizeof(interface_sock);
+
HDEBUG(D_transport|D_acl|D_v) debug_printf_indent("connected\n");
if (getsockname(sock, (struct sockaddr *)(&interface_sock), &size) == 0)
sending_ip_address = host_ntoa(-1, &interface_sock, NULL, &sending_port);
close(sock);
return -1;
}
+
if (ob->keepalive) ip_keepalive(sock, host->address, TRUE);
#ifdef TCP_FASTOPEN
- if (fastopen_blob) tfo_out_check(sock);
+ tfo_out_check(sock);
#endif
return sock;
}
host->address will always be an IPv4 address.
Arguments:
- host host item containing name and address and port
- host_af AF_INET or AF_INET6
- interface outgoing interface address or NULL
- timeout timeout value or 0
- tb transport
+ sc details for making connection: host, af, interface, transport
+ early_data if non-NULL, data to be sent - preferably in the TCP SYN segment
Returns: connected socket number, or -1 with errno set
*/
int
-smtp_connect(host_item *host, int host_af, uschar *interface,
- int timeout, transport_instance * tb)
+smtp_connect(smtp_connect_args * sc, const blob * early_data)
{
-int port = host->port;
-#ifdef SUPPORT_SOCKS
-smtp_transport_options_block * ob =
- (smtp_transport_options_block *)tb->options_block;
-#endif
+int port = sc->host->port;
+smtp_transport_options_block * ob = sc->ob;
-callout_address = string_sprintf("[%s]:%d", host->address, port);
+callout_address = string_sprintf("[%s]:%d", sc->host->address, port);
HDEBUG(D_transport|D_acl|D_v)
{
uschar * s = US" ";
- if (interface) s = string_sprintf(" from %s ", interface);
+ if (sc->interface) s = string_sprintf(" from %s ", sc->interface);
#ifdef SUPPORT_SOCKS
if (ob->socks_proxy) s = string_sprintf("%svia proxy ", s);
#endif
- debug_printf_indent("Connecting to %s %s%s... ", host->name, callout_address, s);
+ debug_printf_indent("Connecting to %s %s%s... ", sc->host->name, callout_address, s);
}
/* Create and connect the socket */
#ifdef SUPPORT_SOCKS
if (ob->socks_proxy)
- return socks_sock_connect(host, host_af, port, interface, tb, timeout);
+ {
+ int sock = socks_sock_connect(sc->host, sc->host_af, port, sc->interface,
+ sc->tblock, ob->connect_timeout);
+
+ if (sock >= 0)
+ {
+ if (early_data && early_data->data && early_data->len)
+ if (send(sock, early_data->data, early_data->len, 0) < 0)
+ {
+ int save_errno = errno;
+ HDEBUG(D_transport|D_acl|D_v)
+ {
+ debug_printf_indent("failed: %s", CUstrerror(save_errno));
+ if (save_errno == ETIMEDOUT)
+ debug_printf(" (timeout=%s)", readconf_printtime(ob->connect_timeout));
+ debug_printf("\n");
+ }
+ (void)close(sock);
+ sock = -1;
+ errno = save_errno;
+ }
+ }
+ return sock;
+ }
#endif
-return smtp_sock_connect(host, host_af, port, interface, tb, timeout, NULL);
+return smtp_sock_connect(sc->host, sc->host_af, port, sc->interface,
+ sc->tblock, ob->connect_timeout, early_data);
}
more ? " (more expected)" : "");
#ifdef SUPPORT_TLS
-if (tls_out.active == outblock->sock)
- rc = tls_write(FALSE, outblock->buffer, n, more);
+if (outblock->cctx->tls_ctx)
+ rc = tls_write(outblock->cctx->tls_ctx, outblock->buffer, n, more);
else
#endif
- rc = send(outblock->sock, outblock->buffer, n,
+
+ {
+ if (outblock->conn_args)
+ {
+ blob early_data = { .data = outblock->buffer, .len = n };
+
+ /* We ignore the more-flag if we're doing a connect with early-data, which
+ means we won't get BDAT+data. A pity, but wise due to the idempotency
+ requirement: TFO with data can, in rare cases, replay the data to the
+ receiver. */
+
+ if ( (outblock->cctx->sock = smtp_connect(outblock->conn_args, &early_data))
+ < 0)
+ return FALSE;
+ outblock->conn_args = NULL;
+ rc = n;
+ }
+ else
+
+ rc = send(outblock->cctx->sock, outblock->buffer, n,
#ifdef MSG_MORE
- more ? MSG_MORE : 0
+ more ? MSG_MORE : 0
#else
- 0
+ 0
#endif
- );
+ );
+ }
if (rc <= 0)
{
any error message.
Arguments:
- outblock contains buffer for pipelining, and socket
+ sx SMTP connection, contains buffer for pipelining, and socket
mode buffer, write-with-more-likely, write
format a format, starting with one of
of HELO, MAIL FROM, RCPT TO, DATA, ".", or QUIT.
*/
int
-smtp_write_command(smtp_outblock * outblock, int mode, const char *format, ...)
+smtp_write_command(void * sx, int mode, const char *format, ...)
{
+smtp_outblock * outblock = &((smtp_context *)sx)->outblock;
int count;
int rc = 0;
va_list ap;
uschar *p = buffer;
uschar *ptr = inblock->ptr;
uschar *ptrend = inblock->ptrend;
-int sock = inblock->sock;
+client_conn_ctx * cctx = inblock->cctx;
/* Loop for reading multiple packets or reading another packet after emptying
a previously-read one. */
/* Need to read a new input packet. */
- if((rc = ip_recv(sock, inblock->buffer, inblock->buffersize, timeout)) <= 0)
+ if((rc = ip_recv(cctx, inblock->buffer, inblock->buffersize, timeout)) <= 0)
{
DEBUG(D_deliver|D_transport|D_acl)
debug_printf_indent(errno ? " SMTP(%s)<<\n" : " SMTP(closed)<<\n",
the error code will be in errno.
Arguments:
- inblock the SMTP input block (contains holding buffer, socket, etc.)
+ sx the SMTP connection (contains input block with holding buffer,
+ socket, etc.)
buffer where to put the response
size the size of the buffer
okdigit the expected first digit of the response
Returns: TRUE if a valid, non-error response was received; else FALSE
*/
+/*XXX could move to smtp transport; no other users */
BOOL
-smtp_read_response(smtp_inblock *inblock, uschar *buffer, int size, int okdigit,
+smtp_read_response(void * sx0, uschar *buffer, int size, int okdigit,
int timeout)
{
+smtp_context * sx = sx0;
uschar *ptr = buffer;
-int count;
+int count = 0;
errno = 0; /* Ensure errno starts out zero */
+#ifdef EXPERIMENTAL_PIPE_CONNECT
+if (sx->pending_BANNER || sx->pending_EHLO)
+ if (smtp_reap_early_pipe(sx, &count) != OK)
+ {
+ DEBUG(D_transport) debug_printf("failed reaping pipelined cmd responsess\n");
+ return FALSE;
+ }
+#endif
+
/* This is a loop to read and concatenate the lines that make up a multi-line
response. */
for (;;)
{
- if ((count = read_response_line(inblock, ptr, size, timeout)) < 0)
+ if ((count = read_response_line(&sx->inblock, ptr, size, timeout)) < 0)
return FALSE;
HDEBUG(D_transport|D_acl|D_v)
- debug_printf_indent(" %s %s\n", (ptr == buffer)? "SMTP<<" : " ", ptr);
+ debug_printf_indent(" %s %s\n", ptr == buffer ? "SMTP<<" : " ", ptr);
/* Check the format of the response: it must start with three digits; if
these are followed by a space or end of line, the response is complete. If
size -= count + 1;
}
+#ifdef TCP_FASTOPEN
+ tfo_out_check(sx->cctx.sock);
+#endif
+
/* Return a value that depends on the SMTP return code. On some systems a
non-zero value of errno has been seen at this point, so ensure it is zero,
because the caller of this function looks at errno when FALSE is returned, to
git://git.exim.org / exim.git / blobdiff