Chapter 1: A typical SSH session

A typical SSH session

A SSH session goes through the following steps:

Before connecting to the server, you can set up if you wish one or other server public key authentication, i.e. DSA or RSA. You can choose cryptographic algorithms you trust and compression algorithms if any. You must of course set up the hostname.
The connection is established. A secure handshake is made, and resulting from it, a public key from the server is gained. You MUST verify that the public key is legitimate, using for instance the MD5 fingerprint or the known hosts file.
The client must authenticate: the classical ways are password, or public keys (from dsa and rsa key-pairs generated by openssh). If a SSH agent is running, it is possible to use it.
Now that the user has been authenticated, you must open one or several channels. Channels are different subways for information into a single ssh connection. Each channel has a standard stream (stdout) and an error stream (stderr). You can theoretically open an infinity of channels.
With the channel you opened, you can do several things:
- Execute a single command.
- Open a shell. You may want to request a pseudo-terminal before.
- Invoke the sftp subsystem to transfer files.
- Invoke the scp subsystem to transfer files.
- Invoke your own subsystem. This is outside the scope of this document, but can be done.
When everything is finished, just close the channels, and then the connection.

The sftp and scp subsystems use channels, but libssh hides them to the programmer. If you want to use those subsystems, instead of a channel, you’ll usually open a “sftp session” or a “scp session”.

Creating the session and setting options

The most important object in a SSH connection is the SSH session. In order to allocate a new SSH session, you use ssh_new(). Don’t forget to always verify that the allocation successed.

#include <libssh/libssh.h>
#include <stdlib.h>

int main()
{
  ssh_session my_ssh_session = ssh_new();
  if (my_ssh_session == NULL)
    exit(-1);
  ...
  ssh_free(my_ssh_session);
}

libssh follows the allocate-it-deallocate-it pattern. Each object that you allocate using xxxxx_new() must be deallocated using xxxxx_free(). In this case, ssh_new() does the allocation and ssh_free() does the contrary.

The ssh_options_set() function sets the options of the session. The most important options are:

SSH_OPTIONS_HOST: the name of the host you want to connect to
SSH_OPTIONS_PORT: the used port (default is port 22)
SSH_OPTIONS_USER: the system user under which you want to connect
SSH_OPTIONS_LOG_VERBOSITY: the quantity of messages that are printed

The complete list of options can be found in the documentation of ssh_options_set(). The only mandatory option is SSH_OPTIONS_HOST. If you don’t use SSH_OPTIONS_USER, the local username of your account will be used.

Here is a small example of how to use it:

#include <libssh/libssh.h>
#include <stdlib.h>

int main()
{
  ssh_session my_ssh_session;
  int verbosity = SSH_LOG_PROTOCOL;
  int port = 22;

  my_ssh_session = ssh_new();
  if (my_ssh_session == NULL)
    exit(-1);

  ssh_options_set(my_ssh_session, SSH_OPTIONS_HOST, "localhost");
  ssh_options_set(my_ssh_session, SSH_OPTIONS_LOG_VERBOSITY, &verbosity);
  ssh_options_set(my_ssh_session, SSH_OPTIONS_PORT, &port);

  ...

  ssh_free(my_ssh_session);
}

Please notice that all parameters are passed to ssh_options_set() as pointers, even if you need to set an integer value.

Connecting to the server

Once all settings have been made, you can connect using ssh_connect(). That function will return SSH_OK if the connection worked, SSH_ERROR otherwise.

You can get the English error string with ssh_get_error() in order to show the user what went wrong. Then, use ssh_disconnect() when you want to stop the session.

Here’s an example:

#include <libssh/libssh.h>
#include <stdlib.h>
#include <stdio.h>

int main()
{
  ssh_session my_ssh_session;
  int rc;

  my_ssh_session = ssh_new();
  if (my_ssh_session == NULL)
    exit(-1);

  ssh_options_set(my_ssh_session, SSH_OPTIONS_HOST, "localhost");

  rc = ssh_connect(my_ssh_session);
  if (rc != SSH_OK)
  {
    fprintf(stderr, "Error connecting to localhost: %s\n",
            ssh_get_error(my_ssh_session));
    exit(-1);
  }

  ...

  ssh_disconnect(my_ssh_session);
  ssh_free(my_ssh_session);
}

Authenticating the server

Once you’re connected, the following step is mandatory: you must check that the server you just connected to is known and safe to use (remember, SSH is about security and authentication).

There are two ways of doing this:

The first way (recommended) is to use the ssh_is_server_known() function. This function will look into the known host file (~/.ssh/known_hosts on UNIX), look for the server hostname’s pattern, and determine whether this host is present or not in the list.
The second way is to use ssh_get_pubkey_hash() to get a binary version of the public key hash value. You can then use your own database to check if this public key is known and secure.

You can also use the ssh_get_pubkey_hash() to show the public key hash value to the user, in case he knows what the public key hash value is (some paranoid people write their public key hash values on paper before going abroad, just in case …).

If the remote host is being used to for the first time, you can ask the user whether he/she trusts it. Once he/she concluded that the host is valid and worth being added in the known hosts file, you use ssh_write_knownhost() to register it in the known hosts file, or any other way if you use your own database.

The following example is part of the examples suite available in the examples/ directory:

#include <errno.h>
#include <string.h>

int verify_knownhost(ssh_session session)
{
  int state, hlen;
  unsigned char *hash = NULL;
  char *hexa;
  char buf[10];

  state = ssh_is_server_known(session);

  hlen = ssh_get_pubkey_hash(session, &hash);
  if (hlen < 0)
    return -1;

  switch (state)
  {
    case SSH_SERVER_KNOWN_OK:
      break; /* ok */

    case SSH_SERVER_KNOWN_CHANGED:
      fprintf(stderr, "Host key for server changed: it is now:\n");
      ssh_print_hexa("Public key hash", hash, hlen);
      fprintf(stderr, "For security reasons, connection will be stopped\n");
      free(hash);
      return -1;

    case SSH_SERVER_FOUND_OTHER:
      fprintf(stderr, "The host key for this server was not found but an other"
        "type of key exists.\n");
      fprintf(stderr, "An attacker might change the default server key to"
        "confuse your client into thinking the key does not exist\n");
      free(hash);
      return -1;

    case SSH_SERVER_FILE_NOT_FOUND:
      fprintf(stderr, "Could not find known host file.\n");
      fprintf(stderr, "If you accept the host key here, the file will be"
       "automatically created.\n");
      /* fallback to SSH_SERVER_NOT_KNOWN behavior */

    case SSH_SERVER_NOT_KNOWN:
      hexa = ssh_get_hexa(hash, hlen);
      fprintf(stderr,"The server is unknown. Do you trust the host key?\n");
      fprintf(stderr, "Public key hash: %s\n", hexa);
      free(hexa);
      if (fgets(buf, sizeof(buf), stdin) == NULL)
      {
        free(hash);
        return -1;
      }
      if (strncasecmp(buf, "yes", 3) != 0)
      {
        free(hash);
        return -1;
      }
      if (ssh_write_knownhost(session) < 0)
      {
        fprintf(stderr, "Error %s\n", strerror(errno));
        free(hash);
        return -1;
      }
      break;

    case SSH_SERVER_ERROR:
      fprintf(stderr, "Error %s", ssh_get_error(session));
      free(hash);
      return -1;
  }

  free(hash);
  return 0;
}

Authenticating the user

The authentication process is the way a service provider can identify a user and verify his/her identity. The authorization process is about enabling the authenticated user the access to ressources. In SSH, the two concepts are linked. After authentication, the server can grant the user access to several ressources such as port forwarding, shell, sftp subsystem, and so on.

libssh supports several methods of authentication:

“none” method. This method allows to get the available authentications methods. It also gives the server a chance to authenticate the user with just his/her login. Some very old hardware uses this feature to fallback the user on a “telnet over SSH” style of login.
password method. A password is sent to the server, which accepts it or not.
keyboard-interactive method. The server sends several challenges to the user, who must answer correctly. This makes possible the authentication via a codebook for instance (“give code at 23:R on page 3”).
public key method. The host knows the public key of the user, and the user must prove he knows the associated private key. This can be done manually, or delegated to the SSH agent as we’ll see later.

All these methods can be combined. You can for instance force the user to authenticate with at least two of the authentication methods. In that case, one speaks of “Partial authentication”. A partial authentication is a response from authentication functions stating that your credential was accepted, but yet another one is required to get in.

The example below shows an authentication with password:

#include <libssh/libssh.h>
#include <stdlib.h>
#include <stdio.h>

int main()
{
  ssh_session my_ssh_session;
  int rc;
  char *password;

  // Open session and set options
  my_ssh_session = ssh_new();
  if (my_ssh_session == NULL)
    exit(-1);
  ssh_options_set(my_ssh_session, SSH_OPTIONS_HOST, "localhost");

  // Connect to server
  rc = ssh_connect(my_ssh_session);
  if (rc != SSH_OK)
  {
    fprintf(stderr, "Error connecting to localhost: %s\n",
            ssh_get_error(my_ssh_session));
    ssh_free(my_ssh_session);
    exit(-1);
  }

  // Verify the server's identity
  // For the source code of verify_knowhost(), check previous example
  if (verify_knownhost(my_ssh_session) < 0)
  {
    ssh_disconnect(my_ssh_session);
    ssh_free(my_ssh_session);
    exit(-1);
  }

  // Authenticate ourselves
  password = getpass("Password: ");
  rc = ssh_userauth_password(my_ssh_session, NULL, password);
  if (rc != SSH_AUTH_SUCCESS)
  {
    fprintf(stderr, "Error authenticating with password: %s\n",
            ssh_get_error(my_ssh_session));
    ssh_disconnect(my_ssh_session);
    ssh_free(my_ssh_session);
    exit(-1);
  }

  ...

  ssh_disconnect(my_ssh_session);
  ssh_free(my_ssh_session);
}

Doing something

At this point, the authenticity of both server and client is established. Time has come to take advantage of the many possibilities offered by the SSH protocol: execute a remote command, open remote shells, transfer files, forward ports, etc.

The example below shows how to execute a remote command:

int show_remote_processes(ssh_session session)
{
  ssh_channel channel;
  int rc;
  char buffer[256];
  int nbytes;

  channel = ssh_channel_new(session);
  if (channel == NULL)
    return SSH_ERROR;

  rc = ssh_channel_open_session(channel);
  if (rc != SSH_OK)
  {
    ssh_channel_free(channel);
    return rc;
  }

  rc = ssh_channel_request_exec(channel, "ps aux");
  if (rc != SSH_OK)
  {
    ssh_channel_close(channel);
    ssh_channel_free(channel);
    return rc;
  }

  nbytes = ssh_channel_read(channel, buffer, sizeof(buffer), 0);
  while (nbytes > 0)
  {
    if (write(1, buffer, nbytes) != (unsigned int) nbytes)
    {
      ssh_channel_close(channel);
      ssh_channel_free(channel);
      return SSH_ERROR;
    }
    nbytes = ssh_channel_read(channel, buffer, sizeof(buffer), 0);
  }

  if (nbytes < 0)
  {
    ssh_channel_close(channel);
    ssh_channel_free(channel);
    return SSH_ERROR;
  }

  ssh_channel_send_eof(channel);
  ssh_channel_close(channel);
  ssh_channel_free(channel);

  return SSH_OK;
}

Handling the errors

All the libssh functions which return an error value also set an English error message describing the problem.

Error values are typically SSH_ERROR for integer values, or NULL for pointers.

The function ssh_get_error() returns a pointer to the static error message.

ssh_error_code() returns the error code number : SSH_NO_ERROR, SSH_REQUEST_DENIED, SSH_INVALID_REQUEST, SSH_CONNECTION_LOST, SSH_FATAL, or SSH_INVALID_DATA. SSH_REQUEST_DENIED means the ssh server refused your request, but the situation is recoverable. The others mean something happened to the connection (some encryption problems, server problems, …). SSH_INVALID_REQUEST means the library got some garbage from server, but might be recoverable. SSH_FATAL means the connection has an important problem and isn’t probably recoverable.

Most of time, the error returned are SSH_FATAL, but some functions (generaly the ssh_request_xxx ones) may fail because of server denying request. In these cases, SSH_REQUEST_DENIED is returned.

For thread safety, errors are bound to ssh_session objects. As long as your ssh_session object is not NULL, you can retrieve the last error message and error code from the ssh_session using ssh_get_error() and ssh_get_error_code() respectively.

The SFTP subsystem has its own error codes, in addition to libssh ones.