solanum-vs-hackint-and-char.../ircd/wsproc.c

591 lines
12 KiB
C
Raw Normal View History

/*
2016-04-02 08:10:01 +00:00
* sslproc.c: An interface to wsockd
* Copyright (C) 2007 Aaron Sethman <androsyn@ratbox.org>
* Copyright (C) 2007 ircd-ratbox development team
*
* 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301
* USA
*/
#include <rb_lib.h>
#include "stdinc.h"
#include "s_conf.h"
#include "logger.h"
#include "listener.h"
#include "wsproc.h"
#include "s_serv.h"
#include "ircd.h"
#include "hash.h"
#include "client.h"
#include "send.h"
#include "packet.h"
static void ws_read_ctl(rb_fde_t * F, void *data);
static int wsockd_count;
#define MAXPASSFD 4
#define READSIZE 1024
typedef struct _ws_ctl_buf
{
rb_dlink_node node;
char *buf;
size_t buflen;
rb_fde_t *F[MAXPASSFD];
int nfds;
} ws_ctl_buf_t;
struct ws_ctl
{
rb_dlink_node node;
int cli_count;
rb_fde_t *F;
rb_fde_t *P;
pid_t pid;
rb_dlink_list readq;
rb_dlink_list writeq;
uint8_t shutdown;
uint8_t dead;
};
static rb_dlink_list wsock_daemons;
static inline uint32_t
buf_to_uint32(char *buf)
{
uint32_t x;
memcpy(&x, buf, sizeof(x));
return x;
}
static inline void
uint32_to_buf(char *buf, uint32_t x)
{
memcpy(buf, &x, sizeof(x));
return;
}
static ws_ctl_t *
allocate_ws_daemon(rb_fde_t * F, rb_fde_t * P, int pid)
{
ws_ctl_t *ctl;
if(F == NULL || pid < 0)
return NULL;
ctl = rb_malloc(sizeof(ws_ctl_t));
ctl->F = F;
ctl->P = P;
ctl->pid = pid;
wsockd_count++;
rb_dlinkAdd(ctl, &ctl->node, &wsock_daemons);
return ctl;
}
static void
free_ws_daemon(ws_ctl_t * ctl)
{
rb_dlink_node *ptr;
ws_ctl_buf_t *ctl_buf;
int x;
if(ctl->cli_count)
return;
RB_DLINK_FOREACH(ptr, ctl->readq.head)
{
ctl_buf = ptr->data;
for(x = 0; x < ctl_buf->nfds; x++)
rb_close(ctl_buf->F[x]);
rb_free(ctl_buf->buf);
rb_free(ctl_buf);
}
RB_DLINK_FOREACH(ptr, ctl->writeq.head)
{
ctl_buf = ptr->data;
for(x = 0; x < ctl_buf->nfds; x++)
rb_close(ctl_buf->F[x]);
rb_free(ctl_buf->buf);
rb_free(ctl_buf);
}
rb_close(ctl->F);
rb_close(ctl->P);
rb_dlinkDelete(&ctl->node, &wsock_daemons);
rb_free(ctl);
}
static char *wsockd_path;
static int wsockd_spin_count = 0;
static time_t last_spin;
static int wsockd_wait = 0;
void
restart_wsockd(void)
{
rb_dlink_node *ptr, *next;
ws_ctl_t *ctl;
RB_DLINK_FOREACH_SAFE(ptr, next, wsock_daemons.head)
{
ctl = ptr->data;
if(ctl->dead)
continue;
if(ctl->shutdown)
continue;
ctl->shutdown = 1;
wsockd_count--;
if(!ctl->cli_count)
{
rb_kill(ctl->pid, SIGKILL);
free_ws_daemon(ctl);
}
}
start_wsockd(ServerInfo.wsockd_count);
}
#if 0
static void
ws_killall(void)
{
rb_dlink_node *ptr, *next;
ws_ctl_t *ctl;
RB_DLINK_FOREACH_SAFE(ptr, next, wsock_daemons.head)
{
ctl = ptr->data;
if(ctl->dead)
continue;
ctl->dead = 1;
if(!ctl->shutdown)
wsockd_count--;
rb_kill(ctl->pid, SIGKILL);
if(!ctl->cli_count)
free_ws_daemon(ctl);
}
}
#endif
static void
ws_dead(ws_ctl_t * ctl)
{
if(ctl->dead)
return;
ctl->dead = 1;
rb_kill(ctl->pid, SIGKILL); /* make sure the process is really gone */
if(!ctl->shutdown)
{
wsockd_count--;
ilog(L_MAIN, "wsockd helper died - attempting to restart");
2020-11-08 19:30:41 +00:00
sendto_realops_snomask(SNO_GENERAL, L_NETWIDE, "wsockd helper died - attempting to restart");
start_wsockd(1);
}
}
static void
ws_do_pipe(rb_fde_t * F, void *data)
{
int retlen;
ws_ctl_t *ctl = data;
retlen = rb_write(F, "0", 1);
if(retlen == 0 || (retlen < 0 && !rb_ignore_errno(errno)))
{
ws_dead(ctl);
return;
}
rb_setselect(F, RB_SELECT_READ, ws_do_pipe, data);
}
static void
2016-04-02 08:10:01 +00:00
restart_wsockd_event(void *unused)
{
wsockd_spin_count = 0;
last_spin = 0;
wsockd_wait = 0;
if(ServerInfo.wsockd_count > get_wsockd_count())
{
int start = ServerInfo.wsockd_count - get_wsockd_count();
ilog(L_MAIN, "Attempting to restart wsockd processes");
2020-11-08 19:30:41 +00:00
sendto_realops_snomask(SNO_GENERAL, L_NETWIDE, "Attempting to restart wsockd processes");
start_wsockd(start);
}
}
int
start_wsockd(int count)
{
rb_fde_t *F1, *F2;
rb_fde_t *P1, *P2;
#ifdef _WIN32
const char *suffix = ".exe";
#else
const char *suffix = "";
#endif
char fullpath[PATH_MAX + 1];
char fdarg[6];
const char *parv[2];
char buf[128];
char s_pid[10];
pid_t pid;
int started = 0, i;
if(wsockd_wait)
return 0;
if(wsockd_spin_count > 20 && (rb_current_time() - last_spin < 5))
{
2016-04-02 08:10:01 +00:00
ilog(L_MAIN, "wsockd helper is spinning - will attempt to restart in 1 minute");
2020-11-08 19:30:41 +00:00
sendto_realops_snomask(SNO_GENERAL, L_NETWIDE,
2016-04-02 08:10:01 +00:00
"wsockd helper is spinning - will attempt to restart in 1 minute");
rb_event_add("restart_wsockd_event", restart_wsockd_event, NULL, 60);
wsockd_wait = 1;
return 0;
}
wsockd_spin_count++;
last_spin = rb_current_time();
if(wsockd_path == NULL)
{
snprintf(fullpath, sizeof(fullpath), "%s%cwsockd%s", ircd_paths[IRCD_PATH_LIBEXEC], RB_PATH_SEPARATOR, suffix);
if(access(fullpath, X_OK) == -1)
{
snprintf(fullpath, sizeof(fullpath), "%s%cbin%cwsockd%s",
ConfigFileEntry.dpath, RB_PATH_SEPARATOR, RB_PATH_SEPARATOR, suffix);
if(access(fullpath, X_OK) == -1)
{
ilog(L_MAIN,
2016-04-02 08:10:01 +00:00
"Unable to execute wsockd%s in %s or %s/bin",
suffix, ircd_paths[IRCD_PATH_LIBEXEC], ConfigFileEntry.dpath);
return 0;
}
}
wsockd_path = rb_strdup(fullpath);
}
2016-04-02 08:10:01 +00:00
rb_strlcpy(buf, "-ircd wsockd daemon", sizeof(buf));
parv[0] = buf;
parv[1] = NULL;
for(i = 0; i < count; i++)
{
ws_ctl_t *ctl;
if(rb_socketpair(AF_UNIX, SOCK_DGRAM, 0, &F1, &F2, "wsockd handle passing socket") == -1)
{
ilog(L_MAIN, "Unable to create wsockd - rb_socketpair failed: %s", strerror(errno));
return started;
}
rb_set_buffers(F1, READBUF_SIZE);
rb_set_buffers(F2, READBUF_SIZE);
snprintf(fdarg, sizeof(fdarg), "%d", rb_get_fd(F2));
rb_setenv("CTL_FD", fdarg, 1);
if(rb_pipe(&P1, &P2, "wsockd pipe") == -1)
{
ilog(L_MAIN, "Unable to create wsockd - rb_pipe failed: %s", strerror(errno));
return started;
}
snprintf(fdarg, sizeof(fdarg), "%d", rb_get_fd(P1));
rb_setenv("CTL_PIPE", fdarg, 1);
snprintf(s_pid, sizeof(s_pid), "%d", (int)getpid());
rb_setenv("CTL_PPID", s_pid, 1);
#ifdef _WIN32
SetHandleInformation((HANDLE) rb_get_fd(F2), HANDLE_FLAG_INHERIT, 1);
SetHandleInformation((HANDLE) rb_get_fd(P1), HANDLE_FLAG_INHERIT, 1);
#endif
pid = rb_spawn_process(wsockd_path, (const char **) parv);
if(pid == -1)
{
2016-04-02 08:10:01 +00:00
ilog(L_MAIN, "Unable to create wsockd: %s\n", strerror(errno));
rb_close(F1);
rb_close(F2);
rb_close(P1);
rb_close(P2);
return started;
}
started++;
rb_close(F2);
rb_close(P1);
ctl = allocate_ws_daemon(F1, P2, pid);
ws_read_ctl(ctl->F, ctl);
ws_do_pipe(P2, ctl);
}
return started;
}
static void
ws_process_dead_fd(ws_ctl_t * ctl, ws_ctl_buf_t * ctl_buf)
{
struct Client *client_p;
char reason[256];
uint32_t fd;
if(ctl_buf->buflen < 6)
return; /* bogus message..drop it.. XXX should warn here */
fd = buf_to_uint32(&ctl_buf->buf[1]);
rb_strlcpy(reason, &ctl_buf->buf[5], sizeof(reason));
client_p = find_cli_connid_hash(fd);
if(client_p == NULL)
return;
if(IsAnyServer(client_p) || IsRegistered(client_p))
{
/* read any last moment ERROR, QUIT or the like -- jilles */
if (!strcmp(reason, "Remote host closed the connection"))
read_packet(client_p->localClient->F, client_p);
if (IsAnyDead(client_p))
return;
}
exit_client(client_p, client_p, &me, reason);
}
static void
ws_process_cmd_recv(ws_ctl_t * ctl)
{
rb_dlink_node *ptr, *next;
ws_ctl_buf_t *ctl_buf;
if(ctl->dead)
return;
RB_DLINK_FOREACH_SAFE(ptr, next, ctl->readq.head)
{
ctl_buf = ptr->data;
switch (*ctl_buf->buf)
{
case 'D':
ws_process_dead_fd(ctl, ctl_buf);
break;
default:
2016-04-02 08:10:01 +00:00
ilog(L_MAIN, "Received invalid command from wsockd: %s", ctl_buf->buf);
2020-11-08 19:30:41 +00:00
sendto_realops_snomask(SNO_GENERAL, L_NETWIDE, "Received invalid command from wsockd");
break;
}
rb_dlinkDelete(ptr, &ctl->readq);
rb_free(ctl_buf->buf);
rb_free(ctl_buf);
}
}
static void
ws_read_ctl(rb_fde_t * F, void *data)
{
ws_ctl_buf_t *ctl_buf;
ws_ctl_t *ctl = data;
int retlen;
if(ctl->dead)
return;
do
{
ctl_buf = rb_malloc(sizeof(ws_ctl_buf_t));
ctl_buf->buf = rb_malloc(READSIZE);
retlen = rb_recv_fd_buf(ctl->F, ctl_buf->buf, READSIZE, ctl_buf->F, 4);
ctl_buf->buflen = retlen;
if(retlen <= 0)
{
rb_free(ctl_buf->buf);
rb_free(ctl_buf);
}
else
rb_dlinkAddTail(ctl_buf, &ctl_buf->node, &ctl->readq);
}
while(retlen > 0);
if(retlen == 0 || (retlen < 0 && !rb_ignore_errno(errno)))
{
ws_dead(ctl);
return;
}
ws_process_cmd_recv(ctl);
rb_setselect(ctl->F, RB_SELECT_READ, ws_read_ctl, ctl);
}
static ws_ctl_t *
which_wsockd(void)
{
ws_ctl_t *ctl, *lowest = NULL;
rb_dlink_node *ptr;
RB_DLINK_FOREACH(ptr, wsock_daemons.head)
{
ctl = ptr->data;
if(ctl->dead)
continue;
if(ctl->shutdown)
continue;
if(lowest == NULL)
{
lowest = ctl;
continue;
}
if(ctl->cli_count < lowest->cli_count)
lowest = ctl;
}
return (lowest);
}
static void
ws_write_ctl(rb_fde_t * F, void *data)
{
ws_ctl_t *ctl = data;
ws_ctl_buf_t *ctl_buf;
rb_dlink_node *ptr, *next;
int retlen, x;
if(ctl->dead)
return;
RB_DLINK_FOREACH_SAFE(ptr, next, ctl->writeq.head)
{
ctl_buf = ptr->data;
/* in theory unix sock_dgram shouldn't ever short write this.. */
retlen = rb_send_fd_buf(ctl->F, ctl_buf->F, ctl_buf->nfds, ctl_buf->buf, ctl_buf->buflen, ctl->pid);
if(retlen > 0)
{
rb_dlinkDelete(ptr, &ctl->writeq);
for(x = 0; x < ctl_buf->nfds; x++)
rb_close(ctl_buf->F[x]);
rb_free(ctl_buf->buf);
rb_free(ctl_buf);
}
if(retlen == 0 || (retlen < 0 && !rb_ignore_errno(errno)))
{
ws_dead(ctl);
return;
}
else
{
rb_setselect(ctl->F, RB_SELECT_WRITE, ws_write_ctl, ctl);
}
}
}
static void
ws_cmd_write_queue(ws_ctl_t * ctl, rb_fde_t ** F, int count, const void *buf, size_t buflen)
{
ws_ctl_buf_t *ctl_buf;
int x;
/* don't bother */
if(ctl->dead)
return;
ctl_buf = rb_malloc(sizeof(ws_ctl_buf_t));
ctl_buf->buf = rb_malloc(buflen);
memcpy(ctl_buf->buf, buf, buflen);
ctl_buf->buflen = buflen;
for(x = 0; x < count && x < MAXPASSFD; x++)
{
ctl_buf->F[x] = F[x];
}
ctl_buf->nfds = count;
rb_dlinkAddTail(ctl_buf, &ctl_buf->node, &ctl->writeq);
ws_write_ctl(ctl->F, ctl);
}
ws_ctl_t *
start_wsockd_accept(rb_fde_t * sslF, rb_fde_t * plainF, uint32_t id)
{
rb_fde_t *F[2];
ws_ctl_t *ctl;
char buf[5];
F[0] = sslF;
F[1] = plainF;
buf[0] = 'A';
uint32_to_buf(&buf[1], id);
ctl = which_wsockd();
if(!ctl)
return NULL;
ctl->cli_count++;
ws_cmd_write_queue(ctl, F, 2, buf, sizeof(buf));
return ctl;
}
void
wsockd_decrement_clicount(ws_ctl_t * ctl)
{
if(ctl == NULL)
return;
ctl->cli_count--;
if(ctl->shutdown && !ctl->cli_count)
{
ctl->dead = 1;
rb_kill(ctl->pid, SIGKILL);
}
if(ctl->dead && !ctl->cli_count)
{
free_ws_daemon(ctl);
}
}
static void
cleanup_dead_ws(void *unused)
{
rb_dlink_node *ptr, *next;
ws_ctl_t *ctl;
RB_DLINK_FOREACH_SAFE(ptr, next, wsock_daemons.head)
{
ctl = ptr->data;
if(ctl->dead && !ctl->cli_count)
{
free_ws_daemon(ctl);
}
}
}
int
get_wsockd_count(void)
{
return wsockd_count;
}
void
wsockd_foreach_info(void (*func)(void *data, pid_t pid, int cli_count, enum wsockd_status status), void *data)
{
rb_dlink_node *ptr, *next;
ws_ctl_t *ctl;
RB_DLINK_FOREACH_SAFE(ptr, next, wsock_daemons.head)
{
ctl = ptr->data;
func(data, ctl->pid, ctl->cli_count,
ctl->dead ? WSOCKD_DEAD :
(ctl->shutdown ? WSOCKD_SHUTDOWN : WSOCKD_ACTIVE));
}
}
void
init_wsockd(void)
{
rb_event_addish("cleanup_dead_ws", cleanup_dead_ws, NULL, 60);
}