2 * Routines for SMTP packet disassembly
4 * $Id: packet-smtp.c,v 1.34 2003/06/11 18:22:12 guy Exp $
6 * Copyright (c) 2000 by Richard Sharpe <rsharpe@ns.aus.com>
8 * Ethereal - Network traffic analyzer
9 * By Gerald Combs <gerald@ethereal.com>
10 * Copyright 1999 Gerald Combs
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version 2
15 * of the License, or (at your option) any later version.
17 * This program is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 * GNU General Public License for more details.
22 * You should have received a copy of the GNU General Public License
23 * along with this program; if not, write to the Free Software
24 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
37 #include <epan/packet.h>
38 #include <epan/conversation.h>
39 #include <epan/resolv.h>
41 #include <epan/strutil.h>
43 #define TCP_PORT_SMTP 25
45 void proto_reg_handoff_smtp(void);
47 static int proto_smtp = -1;
49 static int hf_smtp_req = -1;
50 static int hf_smtp_rsp = -1;
51 static int hf_smtp_req_command = -1;
52 static int hf_smtp_req_parameter = -1;
53 static int hf_smtp_rsp_code = -1;
54 static int hf_smtp_rsp_parameter = -1;
56 static int ett_smtp = -1;
57 static int ett_smtp_cmdresp = -1;
59 static int global_smtp_tcp_port = TCP_PORT_SMTP;
61 /* desegmentation of SMTP command and response lines */
62 static gboolean smtp_desegment = TRUE;
65 * A CMD is an SMTP command, MESSAGE is the message portion, and EOM is the
66 * last part of a message
69 #define SMTP_PDU_CMD 0
70 #define SMTP_PDU_MESSAGE 1
71 #define SMTP_PDU_EOM 2
73 struct smtp_proto_data {
77 static int smtp_packet_init_count = 100;
80 * State information stored with a conversation.
82 struct smtp_request_val {
83 gboolean reading_data; /* Reading message data, not commands */
84 guint16 crlf_seen; /* Have we seen a CRLF on the end of a packet */
87 static GMemChunk *smtp_request_vals = NULL;
88 static GMemChunk *smtp_packet_infos = NULL;
91 smtp_init_protocol(void)
93 if (smtp_request_vals)
94 g_mem_chunk_destroy(smtp_request_vals);
95 if (smtp_packet_infos)
96 g_mem_chunk_destroy(smtp_packet_infos);
98 smtp_request_vals = g_mem_chunk_new("smtp_request_vals",
99 sizeof(struct smtp_request_val),
100 smtp_packet_init_count * sizeof(struct smtp_request_val), G_ALLOC_AND_FREE);
101 smtp_packet_infos = g_mem_chunk_new("smtp_packet_infos",
102 sizeof(struct smtp_proto_data),
103 smtp_packet_init_count * sizeof(struct smtp_proto_data), G_ALLOC_AND_FREE);
108 dissect_smtp(tvbuff_t *tvb, packet_info *pinfo, proto_tree *tree)
110 struct smtp_proto_data *frame_data;
111 proto_tree *smtp_tree;
112 proto_tree *cmdresp_tree;
116 conversation_t *conversation;
117 struct smtp_request_val *request_val;
121 gint length_remaining;
122 gboolean eom_seen = FALSE;
124 gboolean is_continuation_line;
127 /* As there is no guarantee that we will only see frames in the
128 * the SMTP conversation once, and that we will see them in
129 * order - in Ethereal, the user could randomly click on frames
130 * in the conversation in any order in which they choose - we
131 * have to store information with each frame indicating whether
132 * it contains commands or data or an EOM indication.
134 * XXX - what about frames that contain *both*? TCP is a
135 * byte-stream protocol, and there are no guarantees that
136 * TCP segment boundaries will correspond to SMTP commands
137 * or EOM indications.
139 * We only need that for the client->server stream; responses
140 * are easy to manage.
142 * If we have per frame data, use that, else, we must be on the first
143 * pass, so we figure it out on the first pass.
146 /* Find out what conversation this packet is part of ... but only
147 * if we have no information on this packet, so find the per-frame
151 /* SMTP messages have a simple format ... */
153 request = pinfo -> destport == pinfo -> match_port;
156 * Get the first line from the buffer.
158 * Note that "tvb_find_line_end()" will, if it doesn't return
159 * -1, return a value that is not longer than what's in the buffer,
160 * and "tvb_find_line_end()" will always return a value that is not
161 * longer than what's in the buffer, so the "tvb_get_ptr()" call
162 * won't throw an exception.
164 linelen = tvb_find_line_end(tvb, offset, -1, &next_offset,
165 smtp_desegment && pinfo->can_desegment);
168 * We didn't find a line ending, and we're doing desegmentation;
169 * tell the TCP dissector where the data for this message starts
170 * in the data it handed us, and tell it we need one more byte
171 * (we may need more, but we'll try again if what we get next
172 * isn't enough), and return.
174 pinfo->desegment_offset = offset;
175 pinfo->desegment_len = 1;
178 line = tvb_get_ptr(tvb, offset, linelen);
180 frame_data = p_get_proto_data(pinfo->fd, proto_smtp);
184 conversation = find_conversation(&pinfo->src, &pinfo->dst, pinfo->ptype,
185 pinfo->srcport, pinfo->destport, 0);
186 if (conversation == NULL) { /* No conversation, create one */
187 conversation = conversation_new(&pinfo->src, &pinfo->dst, pinfo->ptype,
188 pinfo->srcport, pinfo->destport, 0);
193 * Is there a request structure attached to this conversation?
195 request_val = conversation_get_proto_data(conversation, proto_smtp);
200 * No - create one and attach it.
202 request_val = g_mem_chunk_alloc(smtp_request_vals);
203 request_val->reading_data = FALSE;
204 request_val->crlf_seen = 0;
206 conversation_add_proto_data(conversation, proto_smtp, request_val);
211 * Check whether or not this packet is an end of message packet
212 * We should look for CRLF.CRLF and they may be split.
213 * We have to keep in mind that we may see what we want on
214 * two passes through here ...
217 if (request_val->reading_data) {
220 * The order of these is important ... We want to avoid
221 * cases where there is a CRLF at the end of a packet and a
222 * .CRLF at the begining of the same packet.
225 if ((request_val->crlf_seen && tvb_strneql(tvb, offset, ".\r\n", 3) == 0) ||
226 tvb_strneql(tvb, offset, "\r\n.\r\n", 5) == 0) {
232 length_remaining = tvb_length_remaining(tvb, offset);
233 if (length_remaining == tvb_reported_length_remaining(tvb, offset) &&
234 tvb_strneql(tvb, offset + length_remaining - 2, "\r\n", 2) == 0) {
236 request_val->crlf_seen = 1;
241 request_val->crlf_seen = 0;
247 * OK, Check if we have seen a DATA request. We do it here for
248 * simplicity, but we have to be careful below.
253 frame_data = g_mem_chunk_alloc(smtp_packet_infos);
255 if (request_val->reading_data) {
257 * This is message data.
259 if (eom_seen) { /* Seen the EOM */
262 * Everything that comes after it is commands.
264 * XXX - what if the EOM isn't at the beginning of
265 * the TCP segment? It can occur anywhere....
267 frame_data->pdu_type = SMTP_PDU_EOM;
268 request_val->reading_data = FALSE;
271 * Message data with no EOM.
273 frame_data->pdu_type = SMTP_PDU_MESSAGE;
277 * This is commands - unless the capture started in the
278 * middle of a session, and we're in the middle of data.
279 * To quote RFC 821, "Command codes are four alphabetic
280 * characters"; if we don't see four alphabetic characters
281 * and, if there's anything else in the line, a space, we
282 * assume it's not a command.
283 * (We treat only A-Z and a-z as alphabetic.)
285 #define ISALPHA(c) (((c) >= 'A' && (c) <= 'Z') || \
286 ((c) >= 'a' && (c) <= 'z'))
287 if (linelen >= 4 && ISALPHA(line[0]) && ISALPHA(line[1]) &&
288 ISALPHA(line[2]) && ISALPHA(line[3]) &&
289 (linelen == 4 || line[4] == ' ')) {
290 if (strncasecmp(line, "DATA", 4) == 0) {
294 * This is a command, but everything that comes after it,
295 * until an EOM, is data.
297 frame_data->pdu_type = SMTP_PDU_CMD;
298 request_val->reading_data = TRUE;
305 frame_data->pdu_type = SMTP_PDU_CMD;
311 * Assume it's message data.
314 frame_data->pdu_type = SMTP_PDU_MESSAGE;
320 p_add_proto_data(pinfo->fd, proto_smtp, frame_data);
326 * From here, we simply add items to the tree and info to the info
330 if (check_col(pinfo->cinfo, COL_PROTOCOL))
331 col_set_str(pinfo->cinfo, COL_PROTOCOL, "SMTP");
333 if (check_col(pinfo->cinfo, COL_INFO)) { /* Add the appropriate type here */
336 * If it is a request, we have to look things up, otherwise, just
337 * display the right things
342 /* We must have frame_data here ... */
344 switch (frame_data->pdu_type) {
345 case SMTP_PDU_MESSAGE:
347 col_set_str(pinfo->cinfo, COL_INFO, "Message Body");
352 col_add_fstr(pinfo->cinfo, COL_INFO, "EOM: %s",
353 format_text(line, linelen));
358 col_add_fstr(pinfo->cinfo, COL_INFO, "Command: %s",
359 format_text(line, linelen));
367 col_add_fstr(pinfo->cinfo, COL_INFO, "Response: %s",
368 format_text(line, linelen));
373 if (tree) { /* Build the tree info ... */
375 ti = proto_tree_add_item(tree, proto_smtp, tvb, offset, -1, FALSE);
376 smtp_tree = proto_item_add_subtree(ti, ett_smtp);
380 * Check out whether or not we can see a command in there ...
381 * What we are looking for is not data_seen and the word DATA
384 * We will see DATA and request_val->data_seen when we process the
385 * tree view after we have seen a DATA packet when processing
386 * the packet list pane.
388 * On the first pass, we will not have any info on the packets
389 * On second and subsequent passes, we will.
392 switch (frame_data->pdu_type) {
394 case SMTP_PDU_MESSAGE:
398 * Put its lines into the protocol tree, a line at a time.
400 while (tvb_offset_exists(tvb, offset)) {
403 * Find the end of the line.
405 tvb_find_line_end(tvb, offset, -1, &next_offset, FALSE);
410 proto_tree_add_text(smtp_tree, tvb, offset, next_offset - offset,
412 tvb_format_text(tvb, offset, next_offset - offset));
415 * Step to the next line.
417 offset = next_offset;
426 * End-of-message-body indicator.
428 * XXX - what about stuff after the first line?
429 * Unlikely, as the client should wait for a response to the
430 * DATA command this terminates before sending another
431 * request, but we should probably handle it.
433 proto_tree_add_text(smtp_tree, tvb, offset, linelen,
434 "EOM: %s", format_text(line, linelen));
443 * XXX - what about stuff after the first line?
444 * Unlikely, as the client should wait for a response to the
445 * previous command before sending another request, but we
446 * should probably handle it.
452 proto_tree_add_boolean_hidden(smtp_tree, hf_smtp_req, tvb,
455 * Put the command line into the protocol tree.
457 ti = proto_tree_add_text(smtp_tree, tvb, offset, next_offset - offset,
459 tvb_format_text(tvb, offset, next_offset - offset));
460 cmdresp_tree = proto_item_add_subtree(ti, ett_smtp_cmdresp);
462 proto_tree_add_item(cmdresp_tree, hf_smtp_req_command, tvb,
463 offset, cmdlen, FALSE);
465 proto_tree_add_item(cmdresp_tree, hf_smtp_req_parameter, tvb,
466 offset + 5, linelen - 5, FALSE);
475 * Process the response, a line at a time, until we hit a line
476 * that doesn't have a continuation indication on it.
478 proto_tree_add_boolean_hidden(smtp_tree, hf_smtp_rsp, tvb,
481 while (tvb_offset_exists(tvb, offset)) {
484 * Find the end of the line.
486 linelen = tvb_find_line_end(tvb, offset, -1, &next_offset, FALSE);
489 * Put it into the protocol tree.
491 ti = proto_tree_add_text(smtp_tree, tvb, offset,
492 next_offset - offset, "Response: %s",
493 tvb_format_text(tvb, offset,
494 next_offset - offset));
495 cmdresp_tree = proto_item_add_subtree(ti, ett_smtp_cmdresp);
498 * Is it a continuation line?
500 is_continuation_line =
501 (linelen >= 4 && tvb_get_guint8(tvb, offset + 3) == '-');
504 * Put the response code and parameters into the protocol tree.
506 line = tvb_get_ptr(tvb, offset, linelen);
507 if (linelen >= 3 && isdigit(line[0]) && isdigit(line[1])
508 && isdigit(line[2])) {
510 * We have a 3-digit response code.
512 code = (line[0] - '0')*100 + (line[1] - '0')*10 + (line[2] - '0');
513 proto_tree_add_uint(cmdresp_tree, hf_smtp_rsp_code, tvb, offset, 3,
517 proto_tree_add_item(cmdresp_tree, hf_smtp_rsp_parameter, tvb,
518 offset + 4, linelen - 4, FALSE);
523 * Step past this line.
525 offset = next_offset;
528 * If it's not a continuation line, quit.
530 if (!is_continuation_line)
539 /* Register all the bits needed by the filtering engine */
542 proto_register_smtp(void)
544 static hf_register_info hf[] = {
546 { "Request", "smtp.req", FT_BOOLEAN, BASE_NONE, NULL, 0x0, "", HFILL }},
549 { "Response", "smtp.rsp", FT_BOOLEAN, BASE_NONE, NULL, 0x0, "", HFILL }},
551 { &hf_smtp_req_command,
552 { "Command", "smtp.req.command", FT_STRING, BASE_NONE, NULL, 0x0,
555 { &hf_smtp_req_parameter,
556 { "Request parameter", "smtp.req.parameter", FT_STRING, BASE_NONE, NULL, 0x0,
560 { "Response code", "smtp.response.code", FT_UINT32, BASE_DEC, NULL, 0x0,
563 { &hf_smtp_rsp_parameter,
564 { "Response parameter", "smtp.rsp.parameter", FT_STRING, BASE_NONE, NULL, 0x0,
567 static gint *ett[] = {
571 module_t *smtp_module;
573 /* No Configuration options to register? */
575 proto_smtp = proto_register_protocol("Simple Mail Transfer Protocol",
578 proto_register_field_array(proto_smtp, hf, array_length(hf));
579 proto_register_subtree_array(ett, array_length(ett));
580 register_init_routine(&smtp_init_protocol);
582 smtp_module = prefs_register_protocol(proto_smtp, NULL);
583 prefs_register_bool_preference(smtp_module, "desegment_lines",
584 "Desegment all SMTP command and response lines spanning multiple TCP segments",
585 "Whether the SMTP dissector should desegment all command and response lines spanning multiple TCP segments",
589 /* The registration hand-off routine */
591 proto_reg_handoff_smtp(void)
593 static int smtp_prefs_initialized = FALSE;
594 static dissector_handle_t smtp_handle;
595 static int tcp_port = 0;
597 if (!smtp_prefs_initialized) {
599 smtp_handle = create_dissector_handle(dissect_smtp, proto_smtp);
601 smtp_prefs_initialized = TRUE;
606 dissector_delete("tcp.port", tcp_port, smtp_handle);
610 tcp_port = global_smtp_tcp_port;
612 dissector_add("tcp.port", global_smtp_tcp_port, smtp_handle);