newio/src/s_bsd.c

/*
 *  ircd-hybrid: an advanced Internet Relay Chat Daemon(ircd).
 *  s_bsd.c: Network functions.
 *
 *  Copyright (C) 2002 by the past and present ircd coders, and others.
 *  Copyright (C) 1999 Thomas Helvey
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307
 *  USA
 *
 *  $Id$
 */

#include "stdinc.h"
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/tcp.h>
#include "list.h"
#include "fdlist.h"
#include "s_bsd.h"
#include "client.h"
#include "dbuf.h"
#include "event.h"
#include "irc_string.h"
#include "ircd.h"
#include "listener.h"
#include "numeric.h"
#include "packet.h"
#include "irc_res.h"
#include "restart.h"
#include "s_auth.h"
#include "conf.h"
#include "log.h"
#include "s_serv.h"
#include "send.h"
#include "memory.h"
#include "s_user.h"
#include "msgq.h"


dlink_list connection_list;

static void client_sock_callback(struct Event *);
static void client_timer_callback(struct Event *);

/** Temporary buffer for reading data from a peer. */
static char readbuf[SERVER_TCP_WINDOW];

/*
 * report_error text constants
 */
const char* const ACCEPT_ERROR_MSG    = "error accepting connection for %s: %s";
const char* const BIND_ERROR_MSG      = "bind error for %s: %s";
const char* const CONNECT_ERROR_MSG   = "connect to host %s failed: %s";
const char* const CONNLIMIT_ERROR_MSG = "connect limit exceeded for %s: %s";
const char* const LISTEN_ERROR_MSG    = "listen error for %s: %s";
const char* const NONB_ERROR_MSG      = "error setting non-blocking for %s: %s";
const char* const PEERNAME_ERROR_MSG  = "getpeername failed for %s: %s";
const char* const POLL_ERROR_MSG      = "poll error for %s: %s";
const char* const REGISTER_ERROR_MSG  = "registering %s: %s";
const char* const REUSEADDR_ERROR_MSG = "error setting SO_REUSEADDR for %s: %s";
const char* const SELECT_ERROR_MSG    = "select error for %s: %s";
const char* const SETBUFS_ERROR_MSG   = "error setting buffer size for %s: %s";
const char* const SOCKET_ERROR_MSG    = "error creating socket for %s: %s";
const char* const TOS_ERROR_MSG       = "error setting TOS for %s: %s";


static int
is_blocked(int error)
{
  return EWOULDBLOCK == error
#ifdef ENOMEM
    || ENOMEM == error
#endif
#ifdef ENOBUFS
    || ENOBUFS == error
#endif
    || EAGAIN == error;
}

/* get_sockerr - get the error value from the socket or the current errno
 *
 * Get the *real* error from the socket (well try to anyway..).
 * This may only work when SO_DEBUG is enabled but its worth the
 * gamble anyway.
 */
int
get_sockerr(int fd)
{
  int errtmp = errno;
#ifdef SO_ERROR
  int err = 0;
  socklen_t len = sizeof(err);

  if (-1 < fd && !getsockopt(fd, SOL_SOCKET, SO_ERROR, &err, &len))
  {
    if (err)
      errtmp = err;
  }
  errno = errtmp;
#endif
  return errtmp;
}

/** Mark a socket's address as reusable.
 * @param[in] fd %Socket file descriptor to manipulate.
 * @return Non-zero on success, or zero on failure.
 */
int
os_set_reuseaddr(int fd)
{
  unsigned int opt = 1;

  return !setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(opt));
}

/** Set a socket's send and receive buffer sizes.
 * @param[in] fd %Socket file descriptor to manipulate.
 * @param[in] ssize New send buffer size.
 * @param[in] rsize New receive buffer size.
 * @return Non-zero on success, or zero on failure.
 */
int
os_set_sockbufs(int fd, unsigned int ssize, unsigned int rsize)
{
  unsigned int sopt = ssize;
  unsigned int ropt = rsize;

  return !setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &ropt, sizeof(ropt)) &&
         !setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &sopt, sizeof(sopt));
}

/** Set a socket's "type of service" value.
 * @param[in] fd %Socket file descriptor to manipulate.
 * @param[in] tos New type of service value to use.
 * @return Non-zero on success, or zero on failure.
 */
int
os_set_tos(int fd, int tos)
{
#if defined(IP_TOS) && defined(IPPROTO_IP)
  unsigned int opt = tos;

  return !setsockopt(fd, IPPROTO_IP, IP_TOS, &opt, sizeof(opt));
#else
  return 1;
#endif
}

/** Disable IP options on a socket.
 * @param[in] fd %Socket file descriptor to manipulate.
 * @return Non-zero on success, or zero on failure.
 */
int
os_disable_options(int fd)
{
#if defined(IP_OPTIONS) && defined(IPPROTO_IP)
  return !setsockopt(fd, IPPROTO_IP, IP_OPTIONS, NULL, 0);
#else
  return 1;
#endif
}

/** Start listening on a socket.
 * @param[in] fd Disconnected file descriptor.
 * @param[in] backlog Maximum number of un-accept()ed connections to keep.
 * @return Non-zero on success; zero on error.
 */
int
os_set_listen(int fd, int backlog)
{
  return !listen(fd, backlog);
}

/** Attempt to read from a non-blocking socket.
 * @param[in] fd File descriptor to read from.
 * @param[out] buf Output buffer to read into.
 * @param[in] length Number of bytes to read.
 * @param[out] count_out Receives number of bytes actually read.
 * @return An IOResult value indicating status.
 */
IOResult
os_recv_nonb(int fd, char *buf, unsigned int length, unsigned int *count_out)
{
  int res = 0;

  assert(buf);
  assert(count_out);

  if ((res = recv(fd, buf, length, 0)) > 0)
  {
    *count_out = (unsigned int)res;
    return IO_SUCCESS;
  }

  if (res == 0)
  {
    *count_out = 0;
    errno = 0;  /* Or ECONNRESET? */
    return IO_FAILURE;
  }

  *count_out = 0;
  return is_blocked(errno) ? IO_BLOCKED : IO_FAILURE;
}

/** Attempt to read from a non-blocking UDP socket.
 * @param[in] fd File descriptor to read from.
 * @param[out] buf Output buffer to read into.
 * @param[in] length Number of bytes to read.
 * @param[out] length_out Receives number of bytes actually read.
 * @param[out] addr_out Peer address that sent the message.
 * @return An IOResult value indicating status.
 */
IOResult
os_recvfrom_nonb(int fd, char *buf, unsigned int length,
                 unsigned int *length_out,
                 struct irc_ssaddr *addr)
{
  unsigned int len = sizeof(struct irc_ssaddr);
  int res = 0;

  assert(buf);
  assert(length_out);
  assert(addr_out);

  res = recvfrom(fd, buf, length, 0, (struct sockaddr *)addr, &len);

  if (res > -1)
  {
    *length_out = res;
    return IO_SUCCESS;
  }

  *length_out = 0;
  return is_blocked(errno) ? IO_BLOCKED : IO_FAILURE;
}

/** Attempt to write on a connected socket.
 * @param[in] fd File descriptor to write to.
 * @param[in] buf Output buffer to send from.
 * @param[in] length Number of bytes to write.
 * @param[out] count_out Receives number of bytes actually written.
 * @return An IOResult value indicating status.
 */
IOResult
os_send_nonb(int fd, const char *buf, unsigned int length,
             unsigned int *count_out)
{
  int res = 0;

  assert(buf);
  assert(count_out);

  if ((res = send(fd, buf, length, 0)) > -1)
  {
    *count_out = (unsigned) res;
    return IO_SUCCESS;
  }

  *count_out = 0;
  return is_blocked(errno) ? IO_BLOCKED : IO_FAILURE;
}

/** Attempt a vectored write on a connected socket.
 * @param[in] fd File descriptor to write to.
 * @param[in] buf Message queue to send from.
 * @param[out] count_in Number of bytes mapped from \a buf.
 * @param[out] count_out Receives number of bytes actually written.
 * @return An IOResult value indicating status.
 */
IOResult
os_sendv_nonb(int fd, struct MsgQ *buf, unsigned int *count_in,
              unsigned int *count_out)
{
  int res = 0, count = 0;
  struct iovec iov[IOV_MAX];

  assert(buf);
  assert(count_in);
  assert(count_out);

  *count_in = 0;
  count = msgq_mapiov(buf, iov, IOV_MAX, count_in);

  if ((res = writev(fd, iov, count)) > -1)
  {
    *count_out = (unsigned int)res;
    return IO_SUCCESS;
  }

  *count_out = 0;
  return is_blocked(errno) ? IO_BLOCKED : IO_FAILURE;
}

/*
 * report_error - report an error from an errno. 
 * Record error to log and also send a copy to all *LOCAL* opers online.
 *
 *        text        is a *format* string for outputing error. It must
 *                contain only two '%s', the first will be replaced
 *                by the sockhost from the client_p, and the latter will
 *                be taken from sys_errlist[errno].
 *
 *        client_p        if not NULL, is the *LOCAL* client associated with
 *                the error.
 *
 * Cannot use perror() within daemon. stderr is closed in
 * ircd and cannot be used. And, worse yet, it might have
 * been reassigned to a normal connection...
 * 
 * Actually stderr is still there IFF ircd was run with -s --Rodder
 */

void
report_error(int level, const char *text, const char *who, int error)
{
  int errtmp = errno; /* debug may change 'errno' */
  const char *errmsg = (error) ? strerror(error) : "";

  if (!errmsg)
    errmsg = "Unknown error";

  if (EmptyString(who))
    who = "unknown";

  sendto_realops_flags(UMODE_DEBUG, level, SEND_NOTICE,
                       text, who, errmsg);
  ilog(LOG_TYPE_IRCD, text, who, errmsg);
  errno = errtmp;
}

/*
 * setup_socket()
 *
 * Set the socket non-blocking, and other wonderful bits.
 */
static void
setup_socket(int fd)
{
  int opt = 1;

  setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &opt, sizeof(opt));

#ifdef IPTOS_LOWDELAY
  opt = IPTOS_LOWDELAY;
  setsockopt(fd, IPPROTO_IP, IP_TOS, &opt, sizeof(opt));
#endif

  fcntl(fd, F_SETFL, fcntl(fd, F_GETFL, 0) | O_NONBLOCK);
}

/*
 * close_connection
 *        Close the physical connection. This function must make
 *        MyConnect(client_p) == FALSE, and set client_p->from == NULL.
 */
void
close_connection(struct Client *client_p)
{
  dlink_node *ptr = NULL;

  assert(client_p);

  if (!IsDead(client_p))
  {
    /* attempt to flush any pending dbufs. Evil, but .. -- adrian */
    /* there is still a chance that we might send data to this socket
     * even if it is marked as blocked (COMM_SELECT_READ handler is called
     * before COMM_SELECT_WRITE). Let's try, nothing to lose.. -adx
     */
    ClearSendqBlocked(client_p);
    send_queued_write(client_p);
  }

  if (IsClient(client_p))
  {
    ++ServerStats.is_cl;
    ServerStats.is_cbs += client_p->localClient->send.bytes;
    ServerStats.is_cbr += client_p->localClient->recv.bytes;
    ServerStats.is_cti += CurrentTime - client_p->localClient->firsttime;
  }
  else if (IsServer(client_p))
  {
    ++ServerStats.is_sv;
    ServerStats.is_sbs += client_p->localClient->send.bytes;
    ServerStats.is_sbr += client_p->localClient->recv.bytes;
    ServerStats.is_sti += CurrentTime - client_p->localClient->firsttime;

    DLINK_FOREACH(ptr, server_items.head)
    {
      struct MaskItem *conf = ptr->data;

      if (irccmp(conf->name, client_p->name))
        continue;

      /*
       * Reset next-connect cycle of all connect{} blocks that match
       * this servername.
       */
      conf->until = CurrentTime + conf->class->con_freq;
    }
  }
  else
    ++ServerStats.is_ni;

  if (s_fd(&client_p->localClient->socket)> -1)
  {
    flush_connections(client_p);
    dlinkDelete(&client_p->localClient->node, &connection_list);
    close(s_fd(&client_p->localClient->socket));
    socket_del(&client_p->localClient->socket);  /* Queue a socket delete */
    s_fd(&client_p->localClient->socket) = -1;
  }

  AddFlag(client_p, FLAGS_DEADSOCKET);

  MsgQClear(&client_p->localClient->sendQ);
  client_drop_sendq(client_p->localClient);
  DBufClear(&client_p->localClient->recvQ);

  MyFree(client_p->localClient->passwd);
  detach_conf(client_p, CONF_CLIENT|CONF_OPER|CONF_SERVER);
  client_p->from = NULL; /* ...this should catch them! >:) --msa */

  if (client_p->localClient->listener)
  {
    assert(0 < client_p->localClient->listener->ref_count);
    if (0 == --client_p->localClient->listener->ref_count &&
        !client_p->localClient->listener->active)
      free_listener(client_p->localClient->listener);
  }
}

#ifdef HAVE_LIBCRYPTO
/*
 * ssl_handshake - let OpenSSL initialize the protocol. Register for
 * read/write events if necessary.
 */
static void
ssl_handshake(int fd, struct Client *client_p)
{
  X509 *cert = NULL;
  int ret = SSL_accept(client_p->localClient->fd.ssl);
  int err = SSL_get_error(client_p->localClient->fd.ssl, ret);

  ilog(LOG_TYPE_IRCD, "SSL Error %d %s", err, ERR_error_string(err, NULL));

  if ((cert = SSL_get_peer_certificate(client_p->localClient->fd.ssl)))
  {
    int res = SSL_get_verify_result(client_p->localClient->fd.ssl);
    char buf[EVP_MAX_MD_SIZE * 2 + 1] = { '\0' };
    unsigned char md[EVP_MAX_MD_SIZE] = { '\0' };

    if (res == X509_V_OK || res == X509_V_ERR_SELF_SIGNED_CERT_IN_CHAIN ||
        res == X509_V_ERR_UNABLE_TO_VERIFY_LEAF_SIGNATURE ||
        res == X509_V_ERR_DEPTH_ZERO_SELF_SIGNED_CERT)
    {
      unsigned int i = 0, n = 0;

      if (X509_digest(cert, EVP_sha256(), md, &n))
      {
        for (; i < n; ++i)
          snprintf(buf + 2 * i, 3, "%02X", md[i]);
        client_p->certfp = xstrdup(buf);
      }
    }
    else
      ilog(LOG_TYPE_IRCD, "Client %s!%s@%s gave bad SSL client certificate: %d",
           client_p->name, client_p->username, client_p->host, res);
    X509_free(cert);
  }

  if (ret <= 0)
  {
    switch (SSL_get_error(client_p->localClient->fd.ssl, ret))
    {
      case SSL_ERROR_WANT_WRITE:
        comm_setselect(&client_p->localClient->fd, COMM_SELECT_WRITE,
                       (PF *) ssl_handshake, client_p, 0);
        return;

      case SSL_ERROR_WANT_READ:
        comm_setselect(&client_p->localClient->fd, COMM_SELECT_READ,
                       (PF *) ssl_handshake, client_p, 0);
        return;

      default:
        exit_client(client_p, client_p, "Error during SSL handshake");
        return;
    }
  }

  start_auth(client_p);
}
#endif

/*
 * add_connection - creates a client which has just connected to us on 
 * the given fd. The sockhost field is initialized with the ip# of the host.
 * An unique id is calculated now, in case it is needed for auth.
 * The client is sent to the auth module for verification, and not put in
 * any client list yet.
 */
void
add_connection(struct Listener *listener, struct irc_ssaddr *irn, int fd)
{
  struct Client *new_client = make_client(NULL);

  fd_open(&new_client->localClient->fd, fd, 1,
          (listener->flags & LISTENER_SSL) ?
          "Incoming SSL connection" : "Incoming connection");

  /*
   * copy address to 'sockhost' as a string, copy it to host too
   * so we have something valid to put into error messages...
   */
  memcpy(&new_client->localClient->ip, irn, sizeof(struct irc_ssaddr));

  getnameinfo((struct sockaddr *)&new_client->localClient->ip,
              new_client->localClient->ip.ss_len, new_client->sockhost, 
              sizeof(new_client->sockhost), NULL, 0, NI_NUMERICHOST);
  new_client->localClient->aftype = new_client->localClient->ip.ss.ss_family;

#ifdef HAVE_LIBGEOIP
  /* XXX IPV6 SUPPORT XXX */
  if (irn->ss.ss_family == AF_INET && geoip_ctx)
  {
    const struct sockaddr_in *v4 = (const struct sockaddr_in *)&new_client->localClient->ip;
    new_client->localClient->country_id = GeoIP_id_by_ipnum(geoip_ctx, (unsigned long)ntohl(v4->sin_addr.s_addr));
  }
#endif

  if (new_client->sockhost[0] == ':' && new_client->sockhost[1] == ':')
  {
    strlcpy(new_client->host, "0", sizeof(new_client->host));
    strlcpy(new_client->host+1, new_client->sockhost, sizeof(new_client->host)-1);
    memmove(new_client->sockhost+1, new_client->sockhost, sizeof(new_client->sockhost)-1);
    new_client->sockhost[0] = '0';
  }
  else
    strlcpy(new_client->host, new_client->sockhost, sizeof(new_client->host));

  new_client->localClient->listener = listener;
  ++listener->ref_count;

#ifdef HAVE_LIBCRYPTO
  if (listener->flags & LISTENER_SSL)
  {
    if ((new_client->localClient->fd.ssl = SSL_new(ServerInfo.server_ctx)) == NULL)
    {
      ilog(LOG_TYPE_IRCD, "SSL_new() ERROR! -- %s",
           ERR_error_string(ERR_get_error(), NULL));

      SetDead(new_client);
      exit_client(new_client, new_client, "SSL_new failed");
      return;
    }

    AddFlag(new_client, FLAGS_SSL);
    SSL_set_fd(new_client->localClient->fd.ssl, fd);
    ssl_handshake(0, new_client);
  }
  else
#endif
    start_auth(new_client);
#ifdef NEWIO

  s_fd(&new_client->localClient->socket) = fd;
  if (!socket_add(&new_client->localClient->socket, client_sock_callback,
                  new_client->localClient, SS_CONNECTED, 0, fd))
  {
    ++ServerStats.is_ref;
    write(fd, register_message, strlen(register_message));
    close(fd);
    s_fd(&new_client->localClient->socket) = -1;
    return;
  }

  new_client->localClient->freeflag |= FREEFLAG_SOCKET;
  new_client->localClient->listener = listener;
  ++listener->ref_count;

  /* if we've made it this far we can put the client on the auth query pile */
  start_auth(new_client);
#endif
}

/** Determines whether to tell the events engine we're interested in
 * writable events.
 * \param client_p Client for which to decide this.
 */
void
update_write(struct Client *client_p)
{
  /*
   * If there are messages that need to be sent along, or if the client
   * is in the middle of a /list, then we need to tell the engine that
   * we're interested in writable events--otherwise, we need to drop
   * that interest.
   */
  socket_events(&client_p->localClient->socket,
                ((MsgQLength(&client_p->localClient->sendQ) || client_p->localClient->list_task) ?
                 SOCK_ACTION_ADD : SOCK_ACTION_DEL) | SOCK_EVENT_WRITABLE);
}

/** Read a 'packet' of data from a connection and process it.  Read in
 * 8k chunks to give a better performance rating (for server
 * connections).  Do some tricky stuff for client connections to make
 * sure they don't do any flooding >:-) -avalon
 * @param client_p Client from which to read data.
 * @param socket_ready If non-zero, more data can be read from the client's socket.
 * @return Positive number on success, zero on connection-fatal failure, negative
 *   if user is killed.
 */
static int
read_packet(struct Client *client_p, int socket_ready)
{
  unsigned int dolen = 0, length = 0;

  if (socket_ready && !(IsClient(client_p) &&
                        DBufLength&(client_p->localClient->recvQ) > feature_int(FEAT_CLIENT_FLOOD)))
  {
    switch (os_recv_nonb(cli_fd(client_p), readbuf, sizeof(readbuf), &length))
    {
      case IO_SUCCESS:
        if (length)
        {
          client_p->localClient->lasttime = CurrentTime;

          if (client_p->localClient->lasttime > client_p->localClient->since)
            client_p->localClient->since = CurrentTime;

          ClearPingSent(client_p);
          DelFlag(client_p, FLAGS_NOEWLINE);
        }
        break;

      case IO_BLOCKED:
        break;
      case IO_FAILURE:
        client_p->localClient->error = errno;
        /* AddFlag(client_p, FLAG_DEADSOCKET); */
        return 0;
    }
  }

  /*
   * For server connections, we process as many as we can without
   * worrying about the time of day or anything :)
   */
  if (length > 0 && IsServer(client_p))
    return server_dopacket(client_p, readbuf, length);
  else if (length > 0 && (IsHandshake(client_p) || IsConnecting(client_p)))
    return connect_dopacket(client_p, readbuf, length);
  else
  {
    /*
     * Before we even think of parsing what we just read, stick
     * it on the end of the receive queue and do it when its
     * turn comes around.
     */
    if (length > 0 && !dbuf_put(&client_p->localClient->recvQ, readbuf, length))
      return exit_client(client_p, client_p, "dbuf_put fail");

    if (DBufLength(&client_p->localClient_recvQ) > feature_int(FEAT_CLIENT_FLOOD))
      return exit_client(client_p, client_p, "Excess Flood");

    while (DBufLength(&client_p->localClient->recvQ) && !NoNewLine(client_p) &&
           (IsTrusted(client_p) || cli_since(client_p) - CurrentTime < 10))
    {
      dolen = dbuf_getmsg(&client_p->localClient->recvQ,
                          client_p->localClient->buffer, IRCD_BUFSIZE);

      /*
       * Devious looking...whats it do ? well..if a client
       * sends a *long* message without any CR or LF, then
       * dbuf_getmsg fails and we pull it out using this
       * loop which just gets the next 512 bytes and then
       * deletes the rest of the buffer contents.
       * -avalon
       */
      if (dolen == 0)
      {
        if (DBufLength(&client_p->localClient->recvQ) < 510)
          AddFlag(client_p, FLAGS_NONEWLINE);
        else
        {
          /*
           * More than 512 bytes in the line - drop the input and yell
           * at the client.
           */
          DBufClear(&client_p->localClient->recvQ);
          send_reply(client_p, ERR_INPUTTOOLONG);
        }
      }
      else if (client_dopacket(client_p, dolen) == CPTR_KILLED)
        return CPTR_KILLED;

      /*
       * If it has become registered as a Server
       * then skip the per-message parsing below.
       */
      if (IsHandshake(client_p) || IsServer(client_p))
      {
        while (-1)
        {
          dolen = dbuf_get(&client_p->localClient->recvQ, readbuf, sizeof(readbuf));

          if (dolen <= 0)
            return 1;
          if (dolen == 0)
          {
            if (DBufLength(&client_p->localClient->recvQ) < 510)
              AddFlag(client_p, FLAGS_NONEWLINE);
            else
              DBufClear(&client_p->localClient->recvQ);
          }
          else if ((IsServer(client_p) &&
                    server_dopacket(client_p, readbuf, dolen) == CPTR_KILLED) ||
                   (!IsServer(client_p) &&
                    connect_dopacket(client_p, readbuf, dolen) == CPTR_KILLED))
            return CPTR_KILLED;
        }
      }
    }

    /* If there's still data to process, wait 2 seconds first */
    if (DBufLength(&client_p->localClient->recvQ) && !HasFlag(client_p, FLAGS_NONEWLINE) &&
        !t_onqueue(&client_p->localClient->proc))
    {
      client_p->localClient->freeflag |= FREEFLAG_TIMER;
      timer_add(&client_p->localClient->proc, client_timer_callback, client_p->localClient,
                TT_RELATIVE, 2);
    }
  }
  return 1;
}

/** Attempt to send a sequence of bytes to the connection.
 * As a side effect, updates \a cptr's FLAG_BLOCKED setting
 * and sendB/sendK fields.
 * @param cptr Client that should receive data.
 * @param buf Message buffer to send to client.
 * @return Negative on connection-fatal error; otherwise
 *  number of bytes sent.
 */
unsigned int
deliver_it(struct Client *client_p, struct MsgQ *buf)
{
  unsigned int bytes_written = 0, bytes_count = 0;

  assert(client_p);

  switch (os_sendv_nonb(cli_fd(cptr), buf, &bytes_count, &bytes_written))
  {
    case IO_SUCCESS:
      DelFlag(client_p, FLAG_BLOCKED);

      client_p->localClient->send.bytes += bytes_written;
      me.localClient->send.bytes        += bytes_written;

      /* A partial write implies that future writes will block. */
      if (bytes_written < bytes_count)
        AddFlag(client_p, FLAG_BLOCKED);
      break;
    case IO_BLOCKED:
      AddFlag(client_p, FLAG_BLOCKED);
      break;
    case IO_FAILURE:
      client_p->localClient->error = errno;
      AddFlag(client_p, FLAG_DEADSOCKET);
      break;
  }

  return bytes_written;
}

/*
 * stolen from squid - its a neat (but overused! :) routine which we
 * can use to see whether we can ignore this errno or not. It is
 * generally useful for non-blocking network IO related errnos.
 *     -- adrian
 */
int
ignoreErrno(int ierrno)
{
  switch (ierrno)
  {
    case EINPROGRESS:
    case EWOULDBLOCK:
#if EAGAIN != EWOULDBLOCK
    case EAGAIN:
#endif
    case EALREADY:
    case EINTR:
#ifdef ERESTART
    case ERESTART:
#endif
        return 1;
    default:
        return 0;
  }
}

/*
 * comm_settimeout() - set the socket timeout
 *
 * Set the timeout for the fd
 */
void
comm_settimeout(fde_t *fd, time_t timeout, PF *callback, void *cbdata)
{
  assert(fd->flags.open);

  fd->timeout = CurrentTime + (timeout / 1000);
  fd->timeout_handler = callback;
  fd->timeout_data = cbdata;
}

/*
 * comm_setflush() - set a flush function
 *
 * A flush function is simply a function called if found during
 * comm_timeouts(). Its basically a second timeout, except in this case
 * I'm too lazy to implement multiple timeout functions! :-)
 * its kinda nice to have it separate, since this is designed for
 * flush functions, and when comm_close() is implemented correctly
 * with close functions, we _actually_ don't call comm_close() here ..
 * -- originally Adrian's notes
 * comm_close() is replaced with fd_close() in fdlist.c 
 */
void
comm_setflush(fde_t *fd, time_t timeout, PF *callback, void *cbdata)
{
  assert(fd->flags.open);

  fd->flush_timeout = CurrentTime + (timeout / 1000);
  fd->flush_handler = callback;
  fd->flush_data = cbdata;
}

/*
 * comm_checktimeouts() - check the socket timeouts
 *
 * All this routine does is call the given callback/cbdata, without closing
 * down the file descriptor. When close handlers have been implemented,
 * this will happen.
 */
void
comm_checktimeouts(void *notused)
{
  int i;
  fde_t *F;
  PF *hdl;
  void *data;

  for (i = 0; i < FD_HASH_SIZE; i++)
    for (F = fd_hash[i]; F != NULL; F = fd_next_in_loop)
    {
      assert(F->flags.open);
      fd_next_in_loop = F->hnext;

      /* check flush functions */
      if (F->flush_handler && F->flush_timeout > 0 &&
          F->flush_timeout < CurrentTime)
      {
        hdl = F->flush_handler;
        data = F->flush_data;
        comm_setflush(F, 0, NULL, NULL);
        hdl(F, data);
      }

      /* check timeouts */
      if (F->timeout_handler && F->timeout > 0 &&
          F->timeout < CurrentTime)
      {
        /* Call timeout handler */
        hdl = F->timeout_handler;
        data = F->timeout_data;
        comm_settimeout(F, 0, NULL, NULL);
        hdl(F, data);
      }
    }
}

/*
 * void comm_connect_tcp(int fd, const char *host, unsigned short port,
 *                       struct sockaddr *clocal, int socklen,
 *                       CNCB *callback, void *data, int aftype, int timeout)
 * Input: An fd to connect with, a host and port to connect to,
 *        a local sockaddr to connect from + length(or NULL to use the
 *        default), a callback, the data to pass into the callback, the
 *        address family.
 * Output: None.
 * Side-effects: A non-blocking connection to the host is started, and
 *               if necessary, set up for selection. The callback given
 *               may be called now, or it may be called later.
 */
void
comm_connect_tcp(fde_t *fd, const char *host, unsigned short port,
                 struct sockaddr *clocal, int socklen, CNCB *callback,
                 void *data, int aftype, int timeout)
{
  struct addrinfo hints, *res;
  char portname[PORTNAMELEN + 1];

  assert(callback);
  fd->connect.callback = callback;
  fd->connect.data = data;

  fd->connect.hostaddr.ss.ss_family = aftype;
  fd->connect.hostaddr.ss_port = htons(port);

  /* Note that we're using a passed sockaddr here. This is because
   * generally you'll be bind()ing to a sockaddr grabbed from
   * getsockname(), so this makes things easier.
   * XXX If NULL is passed as local, we should later on bind() to the
   * virtual host IP, for completeness.
   *   -- adrian
   */
  if ((clocal != NULL) && (bind(fd->fd, clocal, socklen) < 0))
  { 
    /* Failure, call the callback with COMM_ERR_BIND */
    comm_connect_callback(fd, COMM_ERR_BIND);
    /* ... and quit */
    return;
  }

  /* Next, if we have been given an IP, get the addr and skip the
   * DNS check (and head direct to comm_connect_tryconnect().
   */
  memset(&hints, 0, sizeof(hints));
  hints.ai_family = AF_UNSPEC;
  hints.ai_socktype = SOCK_STREAM;
  hints.ai_flags = AI_PASSIVE | AI_NUMERICHOST;

  snprintf(portname, sizeof(portname), "%d", port);

  if (getaddrinfo(host, portname, &hints, &res))
  {
    /* Send the DNS request, for the next level */
    if (aftype == AF_INET6)
      gethost_byname_type(comm_connect_dns_callback, fd, host, T_AAAA);
    else
      gethost_byname_type(comm_connect_dns_callback, fd, host, T_A);
  }
  else
  {
    /* We have a valid IP, so we just call tryconnect */
    /* Make sure we actually set the timeout here .. */
    assert(res != NULL);
    memcpy(&fd->connect.hostaddr, res->ai_addr, res->ai_addrlen);
    fd->connect.hostaddr.ss_len = res->ai_addrlen;
    fd->connect.hostaddr.ss.ss_family = res->ai_family;
    freeaddrinfo(res);
    comm_settimeout(fd, timeout*1000, comm_connect_timeout, NULL);
    comm_connect_tryconnect(fd, NULL);
  }
}

/*
 * comm_connect_callback() - call the callback, and continue with life
 */
static void
comm_connect_callback(fde_t *fd, int status)
{
  CNCB *hdl;

  /* This check is gross..but probably necessary */
  if (fd->connect.callback == NULL)
    return;

  /* Clear the connect flag + handler */
  hdl = fd->connect.callback;
  fd->connect.callback = NULL;

  /* Clear the timeout handler */
  comm_settimeout(fd, 0, NULL, NULL);

  /* Call the handler */
  hdl(fd, status, fd->connect.data);
}

/*
 * comm_connect_timeout() - this gets called when the socket connection
 * times out. This *only* can be called once connect() is initially
 * called ..
 */
static void
comm_connect_timeout(fde_t *fd, void *notused)
{
  /* error! */
  comm_connect_callback(fd, COMM_ERR_TIMEOUT);
}

/*
 * comm_connect_dns_callback() - called at the completion of the DNS request
 *
 * The DNS request has completed, so if we've got an error, return it,
 * otherwise we initiate the connect()
 */
static void
comm_connect_dns_callback(void *vptr, const struct irc_ssaddr *addr, const char *name)
{
  fde_t *F = vptr;

  if (name == NULL)
  {
    comm_connect_callback(F, COMM_ERR_DNS);
    return;
  }

  /* No error, set a 10 second timeout */
  comm_settimeout(F, 30*1000, comm_connect_timeout, NULL);

  /* Copy over the DNS reply info so we can use it in the connect() */
  /*
   * Note we don't fudge the refcount here, because we aren't keeping
   * the DNS record around, and the DNS cache is gone anyway.. 
   *     -- adrian
   */
  memcpy(&F->connect.hostaddr, addr, addr->ss_len);
  /* The cast is hacky, but safe - port offset is same on v4 and v6 */
  ((struct sockaddr_in *) &F->connect.hostaddr)->sin_port =
    F->connect.hostaddr.ss_port;
  F->connect.hostaddr.ss_len = addr->ss_len;

  /* Now, call the tryconnect() routine to try a connect() */
  comm_connect_tryconnect(F, NULL);
}

/* static void comm_connect_tryconnect(int fd, void *notused)
 * Input: The fd, the handler data(unused).
 * Output: None.
 * Side-effects: Try and connect with pending connect data for the FD. If
 *               we succeed or get a fatal error, call the callback.
 *               Otherwise, it is still blocking or something, so register
 *               to select for a write event on this FD.
 */
static void
comm_connect_tryconnect(fde_t *fd, void *notused)
{
  int retval;

  /* This check is needed or re-entrant s_bsd_* like sigio break it. */
  if (fd->connect.callback == NULL)
    return;

  /* Try the connect() */
  retval = connect(fd->fd, (struct sockaddr *) &fd->connect.hostaddr, 
    fd->connect.hostaddr.ss_len);

  /* Error? */
  if (retval < 0)
  {
    /*
     * If we get EISCONN, then we've already connect()ed the socket,
     * which is a good thing.
     *   -- adrian
     */
    if (errno == EISCONN)
      comm_connect_callback(fd, COMM_OK);
    else if (ignoreErrno(errno))
      /* Ignore error? Reschedule */
      comm_setselect(fd, COMM_SELECT_WRITE, comm_connect_tryconnect,
                     NULL, 0);
    else
      /* Error? Fail with COMM_ERR_CONNECT */
      comm_connect_callback(fd, COMM_ERR_CONNECT);
    return;
  }

  /* If we get here, we've suceeded, so call with COMM_OK */
  comm_connect_callback(fd, COMM_OK);
}

/*
 * comm_errorstr() - return an error string for the given error condition
 */
const char *
comm_errstr(int error)
{
  if (error < 0 || error >= COMM_ERR_MAX)
    return "Invalid error number!";
  return comm_err_str[error];
}

/*
 * comm_open() - open a socket
 *
 * This is a highly highly cut down version of squid's comm_open() which
 * for the most part emulates socket(), *EXCEPT* it fails if we're about
 * to run out of file descriptors.
 */
int
comm_open(fde_t *F, int family, int sock_type, int proto, const char *note)
{
  int fd;

  /* First, make sure we aren't going to run out of file descriptors */
  if (number_fd >= hard_fdlimit)
  {
    errno = ENFILE;
    return -1;
  }

  /*
   * Next, we try to open the socket. We *should* drop the reserved FD
   * limit if/when we get an error, but we can deal with that later.
   * XXX !!! -- adrian
   */
  fd = socket(family, sock_type, proto);
  if (fd < 0)
    return -1; /* errno will be passed through, yay.. */

  setup_socket(fd);

  /* update things in our fd tracking */
  fd_open(F, fd, 1, note);
  return 0;
}

/*
 * comm_accept() - accept an incoming connection
 *
 * This is a simple wrapper for accept() which enforces FD limits like
 * comm_open() does. Returned fd must be either closed or tagged with
 * fd_open (this function no longer does it).
 */
int
comm_accept(struct Listener *lptr, struct irc_ssaddr *pn)
{
  int newfd;
  socklen_t addrlen = sizeof(struct irc_ssaddr);

  if (number_fd >= hard_fdlimit)
  {
    errno = ENFILE;
    return -1;
  }

  /*
   * Next, do the accept(). if we get an error, we should drop the
   * reserved fd limit, but we can deal with that when comm_open()
   * also does it. XXX -- adrian
   */
  newfd = accept(lptr->fd.fd, (struct sockaddr *)pn, &addrlen);
  if (newfd < 0)
    return -1;

#ifdef IPV6
  remove_ipv6_mapping(pn);
#else
  pn->ss_len = addrlen;
#endif

  execute_callback(setup_socket_cb, newfd);

  /* .. and return */
  return newfd;
}

/* 
 * remove_ipv6_mapping() - Removes IPv4-In-IPv6 mapping from an address
 * OSes with IPv6 mapping listening on both
 * AF_INET and AF_INET6 map AF_INET connections inside AF_INET6 structures
 * 
 */
#ifdef IPV6
void
remove_ipv6_mapping(struct irc_ssaddr *addr)
{
  if (addr->ss.ss_family == AF_INET6)
  {
    if (IN6_IS_ADDR_V4MAPPED(&((struct sockaddr_in6 *)addr)->sin6_addr))
    {
      struct sockaddr_in6 v6; 
      struct sockaddr_in *v4 = (struct sockaddr_in *)addr;

      memcpy(&v6, addr, sizeof(v6));
      memset(v4, 0, sizeof(struct sockaddr_in));
      memcpy(&v4->sin_addr, &v6.sin6_addr.s6_addr[12], sizeof(v4->sin_addr));

      addr->ss.ss_family = AF_INET;
      addr->ss_len = sizeof(struct sockaddr_in);
    }
    else 
      addr->ss_len = sizeof(struct sockaddr_in6);
  }
  else
    addr->ss_len = sizeof(struct sockaddr_in);
} 
#endif

/** Process events on a client socket.
 * @param ev Socket event structure that has a struct Connection as
 *   its associated data.
 */
static void
client_sock_callback(struct Event *ev)
{
  struct Client *client_p = NULL;
  struct Connection *con = NULL;
  const char *fmt = "%s";
  const char *fallback = NULL;

  assert(ev_socket(ev));
  assert(s_data(ev_socket(ev)));

  con = (struct Connection *)s_data(ev_socket(ev));

  assert(con->localClient || ev_type(ev) == ET_DESTROY);

  client_p = con->client_p;

  assert(!client_p || con == client_p->localClient);

  switch (ev_type(ev))
  {
    case ET_DESTROY:
      con->freeflag &= ~FREEFLAG_SOCKET;

      if (!con_freeflag(con) && !client_p)
        free_connection(con);
      break;

    case ET_CONNECT:  /* Socket connection completed */
      if (!completed_connection(client_p) || IsDead(client_p))
        fallback = client_p->info;
      break;

    case ET_ERROR:  /* An error occurred */
      fallback = client_p->info;
      client_p->error = ev_data(ev);

      /*
       * If the OS told us we have a bad file descriptor, we should
       * record that for future reference.
       */
      if (client_p->error == EBADF)
        s_fd(&client_p->localClient->socket) = -1;

      if (s_state(&con->socket) == SS_CONNECTING)
      {
        completed_connection(client_p);

        /*
         * For some reason, the os_get_sockerr() in completed_connect()
         * can return 0 even when ev_data(ev) indicates a real error, so
         * re-assign the client error here.
         */
        client_p->error = ev_data(ev);
        break;
      }

      /*FALLTHROUGH*/
    case ET_EOF:  /* End of file on socket */
      SetFlag(client_p, FLAG_DEADSOCKET);

      if ((IsServer(client_p) || IsHandshake(client_p)) && client_p->error == 0)
      {
        exit_client_msg(client_p, client_p, "Server %s closed the connection (%s)",
                        client_p->name, client_p->servptr->->last_error_msg);
        return;
      }
      else
      {
        fmt = "Read error: %s";
        fallback = "EOF from client";
      }

      break;

    case ET_WRITE: /* Socket is writable */
      DelFlag(client_p, FLAGS_BLOCKED);

      if (cli_listing(client_p) && MsgQLength(&client_p->sendQ) < 2048)
        list_next_channels(client_p);
      send_queued(client_p);
      break;

    case ET_READ:  /* Socket is readable */
      if (!IsDead(client_p))
      {
        if (read_packet(client_p, 1) == 0)  /* Error while reading packet */
          fallback = "EOF from client";
      }

      break;

    default:
      assert(0 && "Unrecognized socket event in client_sock_callback()");
      break;
  }

  assert(!client_p || !client_p->localClient || con == client_p->localClient);

  if (fallback)
  {
    const char *msg = (client_p->error) ? strerror(client_p->error) : fallback;

    if (!msg)
      msg = "Unknown error";
    exit_client_msg(client_p, client_p, fmt, msg);
  }
}

/** Process a timer on client socket.
 * @param ev Timer event that has a struct Connection as its
 * associated data.
 */
static void
client_timer_callback(struct Event *ev)
{
  struct Client *client_p = NULL;
  struct Connection *con = NULL;

  assert(ev_timer(ev));
  assert(t_data(ev_timer(ev)));
  assert(ET_DESTROY == ev_type(ev) || ET_EXPIRE == ev_type(ev));

  con = (struct Connection *)t_data(ev_timer(ev));

  assert(con->client || ev_type(ev) == ET_DESTROY);

  client_p = con->client;

  assert(!client_p || con == client_p->localClient);

  if (ev_type(ev) == ET_DESTROY)
  {
    conf->freeflag &= ~FREEFLAG_TIMER;  /* Timer has expired... */

    if (!con->freeflag && !client_p)
      free_connection(con);  /* Client is being destroyed */
  }
  else
    read_packet(client_p, 0);  /* read_packet will re-add timer if needed */

  assert(!client_p || !client_p->localClient || con == client_p->localClient);
}
Revision:	2391
Committed:	Thu Jul 11 20:57:08 2013 UTC (10 years, 9 months ago) by michael
Content type:	text/x-csrc
File size:	38955 byte(s)
Log Message:	- ioengine changes as of 11JUN13