wireshark/extcap/ssh-base.c

362 lines
11 KiB
C
Raw Permalink Normal View History

/* ssh-base.c
* ssh-base has base utility functions to connect to hosts via ssh
*
* Copyright 2016, Dario Lombardo
*
* Wireshark - Network traffic analyzer
* By Gerald Combs <gerald@wireshark.org>
* Copyright 1998 Gerald Combs
*
* SPDX-License-Identifier: GPL-2.0-or-later
*/
#include "config.h"
2021-12-04 22:15:31 +00:00
#define WS_LOG_DOMAIN LOG_DOMAIN_EXTCAP
#include "ssh-base.h"
#include <extcap/extcap-base.h>
#include <string.h>
#include <libssh/callbacks.h>
#include <ws_attributes.h>
Refactor our logging and extend the wslog API Experience has shown that: 1. The current logging methods are not very reliable or practical. A logging bitmask makes little sense as the user-facing interface (who would want debug but not crtical messages for example?); it's computer-friendly and user-unfriendly. More importantly the console log level preference is initialized too late in the startup process to be used for the logging subsystem and that fact raises a number of annoying and hard-to-fix usability issues. 2. Coding around G_MESSAGES_DEBUG to comply with our log level mask and not clobber the user's settings or not create unexpected log misses is unworkable and generally follows the principle of most surprise. The fact that G_MESSAGES_DEBUG="all" can leak to other programs using GLib is also annoying. 3. The non-structured GLib logging API is very opinionated and lacks configurability beyond replacing the log handler. 4. Windows GUI has some special code to attach to a console, but it would be nice to abstract away the rest under a single interface. 5. Using this logger seems to be noticeably faster. Deprecate the console log level preference and extend our API to implement a log handler in wsutil/wslog.h to provide easy-to-use, flexible and dependable logging during all execution phases. Log levels have a hierarchy, from most verbose to least verbose (debug to error). When a given level is set everything above that is also enabled. The log level can be set with an environment variable or a command line option (parsed as soon as possible but still later than the environment). The default log level is "message". Dissector logging is not included because it is not clear what log domain they should use. An explosion to thousands of domains is not desirable and putting everything in a single domain is probably too coarse and noisy. For now I think it makes sense to let them do their own thing using g_log_default_handler() and continue using the G_MESSAGES_DEBUG mechanism with specific domains for each individual dissector. In the future a mechanism may be added to selectively enable these domains at runtime while trying to avoid the problems introduced by G_MESSAGES_DEBUG.
2021-06-08 02:46:52 +01:00
#include <wsutil/wslog.h>
/*
* The unreleased 0.11.0 version of libssh has the ability to
* add algorithms to the default supported list by prepending
* "+" to the configuration list. For older versions, we have
* to specify all the algorithms we want, but as long as at
* least one succeeds the command won't fail. (That means that
* it's possible that we won't actually add support for SHA-1,
* say if it's running on a system in FIPS mode. We could parse
* the returned list to check.)
*/
#if LIBSSH_VERSION_INT >= SSH_VERSION_INT(0,11,0)
#define HOSTKEYS_SHA1 "+ssh-rsa"
#define KEY_EXCHANGE_SHA1 "+diffie-hellman-group14-sha1,diffie-hellman-group1-sha1,diffie-hellman-group-exchange-sha1"
#define HMAC_SHA1 "+hmac-sha1-etm@openssh.com,hmac-sha1"
#else
#define HOSTKEYS_SHA1 \
"ssh-ed25519," \
"ecdsa-sha2-nistp521," \
"ecdsa-sha2-nistp384," \
"ecdsa-sha2-nistp256," \
"sk-ssh-ed25519@openssh.com," \
"sk-ecdsa-sha2-nistp256@openssh.com," \
"rsa-sha2-512," \
"rsa-sha2-256," \
"ssh-rsa"
#define KEY_EXCHANGE_SHA1 \
"curve25519-sha256,curve25519-sha256@libssh.org," \
"ecdh-sha2-nistp256,ecdh-sha2-nistp384,ecdh-sha2-nistp521," \
"diffie-hellman-group18-sha512,diffie-hellman-group16-sha512," \
"diffie-hellman-group-exchange-sha256," \
"diffie-hellman-group14-sha256," \
"diffie-hellman-group-exchange-sha1," \
"diffie-hellman-group14-sha1,diffie-hellman-group1-sha1"
#define HMAC_SHA1 \
"hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com," \
"hmac-sha2-256,hmac-sha2-512," \
"hmac-sha1-etm@openssh.com,hmac-sha1"
#endif
static void extcap_log(int priority, const char *function, const char *buffer, void *userdata _U_)
{
enum ws_log_level level = LOG_LEVEL_DEBUG;
switch (priority) {
case SSH_LOG_TRACE:
level = LOG_LEVEL_NOISY;
break;
case SSH_LOG_DEBUG:
level = LOG_LEVEL_DEBUG;
break;
case SSH_LOG_INFO:
level = LOG_LEVEL_INFO;
break;
case SSH_LOG_WARN:
default:
/* Prior to 0.11.0 libssh prints far too much at SSH_LOG_WARN,
* including merely informational messages.
* Lower them to LOG_LEVEL_INFO, which won't get shown in the GUI
* and aren't shown by default. (Anything INFO and below goes to
* stdout due to ws_log_console_writer_set_use_stdout in extcap-base.c)
* After the following commit libssh only uses LOG_LEVEL_WARN for
* serious issues:
* https://gitlab.com/libssh/libssh-mirror/-/commit/657d9143d121dfff74f5a63f734d0096c7f37194
*/
#if LIBSSH_VERSION_INT < SSH_VERSION_INT(0,11,0)
level = LOG_LEVEL_INFO;
#else
level = LOG_LEVEL_WARNING;
#endif
break;
}
/* We set the libssh log level to specifically ask for this, so don't
* both checking the log level a second time.
*/
ws_log_write_always_full("libssh", level, NULL, 0, function, "%s", buffer);
}
void add_libssh_info(extcap_parameters * extcap_conf)
{
extcap_base_set_compiled_with(extcap_conf, "libssh version %s", SSH_STRINGIFY(LIBSSH_VERSION));
extcap_base_set_running_with(extcap_conf, "libssh version %s", ssh_version(0));
}
ssh_session create_ssh_connection(const ssh_params_t* ssh_params, char** err_info)
{
ssh_session sshs;
char* username = NULL;
unsigned port;
/* Open session and set options */
sshs = ssh_new();
if (sshs == NULL) {
*err_info = g_strdup("Can't create ssh session");
return NULL;
}
if (!ssh_params->host) {
*err_info = g_strdup("Hostname needed");
goto failure;
}
if (ssh_options_set(sshs, SSH_OPTIONS_HOST, ssh_params->host)) {
*err_info = ws_strdup_printf("Can't set the host: %s", ssh_params->host);
goto failure;
}
/* Load the configurations already present in the system configuration file. */
/* They will be overwritten by the user-provided configurations. */
if (ssh_options_parse_config(sshs, NULL) != 0) {
*err_info = g_strdup("Unable to load the configuration file");
goto failure;
}
ssh_options_set(sshs, SSH_OPTIONS_LOG_VERBOSITY, &ssh_params->debug);
ssh_set_log_callback(extcap_log);
if (ssh_params->ssh_sha1) {
if (ssh_options_set(sshs, SSH_OPTIONS_HOSTKEYS, HOSTKEYS_SHA1)) {
*err_info = ws_strdup_printf("Can't set host keys to allow SHA-1.");
goto failure;
}
if (ssh_options_set(sshs, SSH_OPTIONS_PUBLICKEY_ACCEPTED_TYPES, HOSTKEYS_SHA1)) {
*err_info = ws_strdup_printf("Can't set public key algorithms to allow SSH-RSA (SHA-1).");
goto failure;
}
if (ssh_options_set(sshs, SSH_OPTIONS_KEY_EXCHANGE, KEY_EXCHANGE_SHA1)) {
*err_info = ws_strdup_printf("Can't set key exchange methods to allow SHA-1.");
goto failure;
}
if (ssh_options_set(sshs, SSH_OPTIONS_HMAC_C_S, HMAC_SHA1)) {
*err_info = ws_strdup_printf("Can't set MAC client to server algorithms to allow SHA-1.");
goto failure;
}
if (ssh_options_set(sshs, SSH_OPTIONS_HMAC_S_C, HMAC_SHA1)) {
*err_info = ws_strdup_printf("Can't set MAC server to client algorithms to allow SHA-1.");
goto failure;
}
}
if (ssh_params->port != 0) {
port = ssh_params->port;
if (ssh_options_set(sshs, SSH_OPTIONS_PORT, &port)) {
*err_info = ws_strdup_printf("Can't set the port: %u", port);
goto failure;
}
}
if (ssh_params->proxycommand) {
#if defined(_WIN32) && (LIBSSH_VERSION_INT < SSH_VERSION_INT(0,11,0))
// ProxyCommand isn't supported on Windows:
// https://gitlab.com/libssh/libssh-mirror/-/issues/75
// ssh_options_set "succeeds" in setting it, though.
// On 0.11.0 and later, ssh_connect will fail.
// Before 0.11.0, libssh silently ignores the option.
// https://gitlab.com/libssh/libssh-mirror/-/commit/bed4438695df2f4635617fc45f802d782fdd8479
*err_info = ws_strdup_printf("Can't set the ProxyCommand: %s (libssh does not support proxy commands on Windows.)", ssh_params->proxycommand);
goto failure;
#endif
if (ssh_options_set(sshs, SSH_OPTIONS_PROXYCOMMAND, ssh_params->proxycommand)) {
*err_info = ws_strdup_printf("Can't set the ProxyCommand: %s", ssh_params->proxycommand);
goto failure;
}
}
if (ssh_params->username) {
if (ssh_options_set(sshs, SSH_OPTIONS_USER, ssh_params->username)) {
*err_info = ws_strdup_printf("Can't set the username: %s", ssh_params->username);
goto failure;
}
}
ssh_options_get(sshs, SSH_OPTIONS_USER, &username);
ssh_options_get_port(sshs, &port);
Refactor our logging and extend the wslog API Experience has shown that: 1. The current logging methods are not very reliable or practical. A logging bitmask makes little sense as the user-facing interface (who would want debug but not crtical messages for example?); it's computer-friendly and user-unfriendly. More importantly the console log level preference is initialized too late in the startup process to be used for the logging subsystem and that fact raises a number of annoying and hard-to-fix usability issues. 2. Coding around G_MESSAGES_DEBUG to comply with our log level mask and not clobber the user's settings or not create unexpected log misses is unworkable and generally follows the principle of most surprise. The fact that G_MESSAGES_DEBUG="all" can leak to other programs using GLib is also annoying. 3. The non-structured GLib logging API is very opinionated and lacks configurability beyond replacing the log handler. 4. Windows GUI has some special code to attach to a console, but it would be nice to abstract away the rest under a single interface. 5. Using this logger seems to be noticeably faster. Deprecate the console log level preference and extend our API to implement a log handler in wsutil/wslog.h to provide easy-to-use, flexible and dependable logging during all execution phases. Log levels have a hierarchy, from most verbose to least verbose (debug to error). When a given level is set everything above that is also enabled. The log level can be set with an environment variable or a command line option (parsed as soon as possible but still later than the environment). The default log level is "message". Dissector logging is not included because it is not clear what log domain they should use. An explosion to thousands of domains is not desirable and putting everything in a single domain is probably too coarse and noisy. For now I think it makes sense to let them do their own thing using g_log_default_handler() and continue using the G_MESSAGES_DEBUG mechanism with specific domains for each individual dissector. In the future a mechanism may be added to selectively enable these domains at runtime while trying to avoid the problems introduced by G_MESSAGES_DEBUG.
2021-06-08 02:46:52 +01:00
ws_log(LOG_DOMAIN_CAPCHILD, LOG_LEVEL_INFO, "Opening ssh connection to %s@%s:%u", username,
ssh_params->host, port);
ssh_string_free_char(username);
/* Connect to server */
if (ssh_connect(sshs) != SSH_OK) {
*err_info = ws_strdup_printf("Connection error: %s", ssh_get_error(sshs));
goto failure;
}
/* If a public key path has been provided, try to authenticate using it */
if (ssh_params->sshkey_path) {
ssh_key pkey = ssh_key_new();
int ret;
Refactor our logging and extend the wslog API Experience has shown that: 1. The current logging methods are not very reliable or practical. A logging bitmask makes little sense as the user-facing interface (who would want debug but not crtical messages for example?); it's computer-friendly and user-unfriendly. More importantly the console log level preference is initialized too late in the startup process to be used for the logging subsystem and that fact raises a number of annoying and hard-to-fix usability issues. 2. Coding around G_MESSAGES_DEBUG to comply with our log level mask and not clobber the user's settings or not create unexpected log misses is unworkable and generally follows the principle of most surprise. The fact that G_MESSAGES_DEBUG="all" can leak to other programs using GLib is also annoying. 3. The non-structured GLib logging API is very opinionated and lacks configurability beyond replacing the log handler. 4. Windows GUI has some special code to attach to a console, but it would be nice to abstract away the rest under a single interface. 5. Using this logger seems to be noticeably faster. Deprecate the console log level preference and extend our API to implement a log handler in wsutil/wslog.h to provide easy-to-use, flexible and dependable logging during all execution phases. Log levels have a hierarchy, from most verbose to least verbose (debug to error). When a given level is set everything above that is also enabled. The log level can be set with an environment variable or a command line option (parsed as soon as possible but still later than the environment). The default log level is "message". Dissector logging is not included because it is not clear what log domain they should use. An explosion to thousands of domains is not desirable and putting everything in a single domain is probably too coarse and noisy. For now I think it makes sense to let them do their own thing using g_log_default_handler() and continue using the G_MESSAGES_DEBUG mechanism with specific domains for each individual dissector. In the future a mechanism may be added to selectively enable these domains at runtime while trying to avoid the problems introduced by G_MESSAGES_DEBUG.
2021-06-08 02:46:52 +01:00
ws_info("Connecting using public key in %s...", ssh_params->sshkey_path);
ret = ssh_pki_import_privkey_file(ssh_params->sshkey_path, ssh_params->sshkey_passphrase, NULL, NULL, &pkey);
switch (ret) {
case SSH_OK:
if (ssh_userauth_publickey(sshs, NULL, pkey) == SSH_AUTH_SUCCESS) {
Refactor our logging and extend the wslog API Experience has shown that: 1. The current logging methods are not very reliable or practical. A logging bitmask makes little sense as the user-facing interface (who would want debug but not crtical messages for example?); it's computer-friendly and user-unfriendly. More importantly the console log level preference is initialized too late in the startup process to be used for the logging subsystem and that fact raises a number of annoying and hard-to-fix usability issues. 2. Coding around G_MESSAGES_DEBUG to comply with our log level mask and not clobber the user's settings or not create unexpected log misses is unworkable and generally follows the principle of most surprise. The fact that G_MESSAGES_DEBUG="all" can leak to other programs using GLib is also annoying. 3. The non-structured GLib logging API is very opinionated and lacks configurability beyond replacing the log handler. 4. Windows GUI has some special code to attach to a console, but it would be nice to abstract away the rest under a single interface. 5. Using this logger seems to be noticeably faster. Deprecate the console log level preference and extend our API to implement a log handler in wsutil/wslog.h to provide easy-to-use, flexible and dependable logging during all execution phases. Log levels have a hierarchy, from most verbose to least verbose (debug to error). When a given level is set everything above that is also enabled. The log level can be set with an environment variable or a command line option (parsed as soon as possible but still later than the environment). The default log level is "message". Dissector logging is not included because it is not clear what log domain they should use. An explosion to thousands of domains is not desirable and putting everything in a single domain is probably too coarse and noisy. For now I think it makes sense to let them do their own thing using g_log_default_handler() and continue using the G_MESSAGES_DEBUG mechanism with specific domains for each individual dissector. In the future a mechanism may be added to selectively enable these domains at runtime while trying to avoid the problems introduced by G_MESSAGES_DEBUG.
2021-06-08 02:46:52 +01:00
ws_info("done");
ssh_key_free(pkey);
return sshs;
}
ws_info("failed (%s)", ssh_get_error(sshs));
break;
case SSH_EOF:
ws_warning("Error importing key from %s. File doesn't exist or permission denied.",
ssh_params->sshkey_path);
break;
case SSH_ERROR:
/* Unfortunately we can't call ssh_get_error() on the
* key to determine why import failed.
*/
ws_warning("Error importing key from %s. Make sure it is a valid"
" private key file and any necessary passphrase is configured.",
ssh_params->sshkey_path);
break;
default:
ws_warning("Unknown error from ssh_pki_import_privkey_file");
}
ssh_key_free(pkey);
}
/* Workaround: it may happen that libssh closes socket in meantime and any next ssh_ call fails so we should detect it in advance */
2023-01-13 20:32:34 +00:00
if (ssh_get_fd(sshs) != (socket_t)-1) {
/* If a password has been provided and all previous attempts failed, try to use it */
if (ssh_params->password) {
ws_info("Connecting using password...");
if (ssh_userauth_password(sshs, ssh_params->username, ssh_params->password) == SSH_AUTH_SUCCESS) {
ws_info("done");
return sshs;
}
ws_info("failed");
}
} else {
ws_info("ssh connection closed before password authentication");
}
/* Workaround: it may happen that libssh closes socket in meantime and any next ssh_ call fails so we should detect it in advance */
2023-01-13 20:32:34 +00:00
if (ssh_get_fd(sshs) != (socket_t)-1) {
/* Try to authenticate using standard public key */
ws_info("Connecting using standard public key...");
if (ssh_userauth_publickey_auto(sshs, NULL, NULL) == SSH_AUTH_SUCCESS) {
Refactor our logging and extend the wslog API Experience has shown that: 1. The current logging methods are not very reliable or practical. A logging bitmask makes little sense as the user-facing interface (who would want debug but not crtical messages for example?); it's computer-friendly and user-unfriendly. More importantly the console log level preference is initialized too late in the startup process to be used for the logging subsystem and that fact raises a number of annoying and hard-to-fix usability issues. 2. Coding around G_MESSAGES_DEBUG to comply with our log level mask and not clobber the user's settings or not create unexpected log misses is unworkable and generally follows the principle of most surprise. The fact that G_MESSAGES_DEBUG="all" can leak to other programs using GLib is also annoying. 3. The non-structured GLib logging API is very opinionated and lacks configurability beyond replacing the log handler. 4. Windows GUI has some special code to attach to a console, but it would be nice to abstract away the rest under a single interface. 5. Using this logger seems to be noticeably faster. Deprecate the console log level preference and extend our API to implement a log handler in wsutil/wslog.h to provide easy-to-use, flexible and dependable logging during all execution phases. Log levels have a hierarchy, from most verbose to least verbose (debug to error). When a given level is set everything above that is also enabled. The log level can be set with an environment variable or a command line option (parsed as soon as possible but still later than the environment). The default log level is "message". Dissector logging is not included because it is not clear what log domain they should use. An explosion to thousands of domains is not desirable and putting everything in a single domain is probably too coarse and noisy. For now I think it makes sense to let them do their own thing using g_log_default_handler() and continue using the G_MESSAGES_DEBUG mechanism with specific domains for each individual dissector. In the future a mechanism may be added to selectively enable these domains at runtime while trying to avoid the problems introduced by G_MESSAGES_DEBUG.
2021-06-08 02:46:52 +01:00
ws_info("done");
return sshs;
}
Refactor our logging and extend the wslog API Experience has shown that: 1. The current logging methods are not very reliable or practical. A logging bitmask makes little sense as the user-facing interface (who would want debug but not crtical messages for example?); it's computer-friendly and user-unfriendly. More importantly the console log level preference is initialized too late in the startup process to be used for the logging subsystem and that fact raises a number of annoying and hard-to-fix usability issues. 2. Coding around G_MESSAGES_DEBUG to comply with our log level mask and not clobber the user's settings or not create unexpected log misses is unworkable and generally follows the principle of most surprise. The fact that G_MESSAGES_DEBUG="all" can leak to other programs using GLib is also annoying. 3. The non-structured GLib logging API is very opinionated and lacks configurability beyond replacing the log handler. 4. Windows GUI has some special code to attach to a console, but it would be nice to abstract away the rest under a single interface. 5. Using this logger seems to be noticeably faster. Deprecate the console log level preference and extend our API to implement a log handler in wsutil/wslog.h to provide easy-to-use, flexible and dependable logging during all execution phases. Log levels have a hierarchy, from most verbose to least verbose (debug to error). When a given level is set everything above that is also enabled. The log level can be set with an environment variable or a command line option (parsed as soon as possible but still later than the environment). The default log level is "message". Dissector logging is not included because it is not clear what log domain they should use. An explosion to thousands of domains is not desirable and putting everything in a single domain is probably too coarse and noisy. For now I think it makes sense to let them do their own thing using g_log_default_handler() and continue using the G_MESSAGES_DEBUG mechanism with specific domains for each individual dissector. In the future a mechanism may be added to selectively enable these domains at runtime while trying to avoid the problems introduced by G_MESSAGES_DEBUG.
2021-06-08 02:46:52 +01:00
ws_info("failed");
} else {
ws_info("ssh connection closed before public key authentication");
}
*err_info = ws_strdup_printf("Can't find a valid authentication. Disconnecting.");
/* All authentication failed. Disconnect and return */
ssh_disconnect(sshs);
failure:
ssh_free(sshs);
return NULL;
}
int ssh_channel_printf(ssh_channel channel, const char* fmt, ...)
{
char* buf;
va_list arg;
int ret = EXIT_SUCCESS;
va_start(arg, fmt);
buf = ws_strdup_vprintf(fmt, arg);
ws_debug("%s", buf);
if (ssh_channel_write(channel, buf, (uint32_t)strlen(buf)) == SSH_ERROR)
ret = EXIT_FAILURE;
va_end(arg);
g_free(buf);
return ret;
}
void ssh_cleanup(ssh_session* sshs, ssh_channel* channel)
{
if (*channel) {
ssh_channel_send_eof(*channel);
ssh_channel_close(*channel);
ssh_channel_free(*channel);
*channel = NULL;
}
if (*sshs) {
ssh_disconnect(*sshs);
ssh_free(*sshs);
*sshs = NULL;
}
}
ssh_params_t* ssh_params_new(void)
{
return g_new0(ssh_params_t, 1);
}
void ssh_params_free(ssh_params_t* ssh_params)
{
if (!ssh_params)
return;
g_free(ssh_params->host);
g_free(ssh_params->username);
g_free(ssh_params->password);
g_free(ssh_params->sshkey_path);
g_free(ssh_params->sshkey_passphrase);
g_free(ssh_params->proxycommand);
g_free(ssh_params);
}
void ssh_params_set_log_level(ssh_params_t* ssh_params, enum ws_log_level level)
{
switch (level) {
case LOG_LEVEL_NOISY:
ssh_params->debug = SSH_LOG_TRACE;
break;
case LOG_LEVEL_DEBUG:
ssh_params->debug = SSH_LOG_DEBUG;
break;
case LOG_LEVEL_INFO:
ssh_params->debug = SSH_LOG_INFO;
break;
default:
ssh_params->debug = SSH_LOG_WARN;
}
}
/*
* Editor modelines - https://www.wireshark.org/tools/modelines.html
*
* Local variables:
* c-basic-offset: 8
* tab-width: 8
* indent-tabs-mode: t
* End:
*
* vi: set shiftwidth=8 tabstop=8 noexpandtab:
* :indentSize=8:tabSize=8:noTabs=false:
*/