signal_handler.c revision 9c5366ba55b1553b2d66f48e3d14fbd274a2944d 
1/*
2 * Copyright (C) 2010 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17#include <stdio.h>
18#include <errno.h>
19#include <signal.h>
20#include <unistd.h>
21#include <fcntl.h>
22#include <sys/types.h>
23#include <sys/socket.h>
24#include <sys/wait.h>
25#include <cutils/sockets.h>
26#include <sys/reboot.h>
27
28#include "init.h"
29
30static int signal_fd = -1;
31static int signal_recv_fd = -1;
32
33static void sigchld_handler(int s)
34{
35    write(signal_fd, &s, 1);
36}
37
38void signal_init(void)
39{
40    int s[2];
41
42    struct sigaction act;
43
44    act.sa_handler = sigchld_handler;
45    act.sa_flags = SA_NOCLDSTOP;
46    act.sa_mask = 0;
47    act.sa_restorer = NULL;
48    sigaction(SIGCHLD, &act, 0);
49
50    /* create a signalling mechanism for the sigchld handler */
51    if (socketpair(AF_UNIX, SOCK_STREAM, 0, s) == 0) {
52        signal_fd = s[0];
53        signal_recv_fd = s[1];
54        fcntl(s[0], F_SETFD, FD_CLOEXEC);
55        fcntl(s[0], F_SETFL, O_NONBLOCK);
56        fcntl(s[1], F_SETFD, FD_CLOEXEC);
57        fcntl(s[1], F_SETFL, O_NONBLOCK);
58    }
59}
60
61#define CRITICAL_CRASH_THRESHOLD    4       /* if we crash >4 times ... */
62#define CRITICAL_CRASH_WINDOW       (4*60)  /* ... in 4 minutes, goto recovery*/
63
64static int wait_for_one_process(int block)
65{
66    pid_t pid;
67    int status;
68    struct service *svc;
69    struct socketinfo *si;
70    time_t now;
71    struct listnode *node;
72    struct command *cmd;
73
74    while ( (pid = waitpid(-1, &status, block ? 0 : WNOHANG)) == -1 && errno == EINTR );
75    if (pid <= 0) return -1;
76    INFO("waitpid returned pid %d, status = %08x\n", pid, status);
77
78    svc = service_find_by_pid(pid);
79    if (!svc) {
80        ERROR("untracked pid %d exited\n", pid);
81        return 0;
82    }
83
84    NOTICE("process '%s', pid %d exited\n", svc->name, pid);
85
86    if (!(svc->flags & SVC_ONESHOT)) {
87        kill(-pid, SIGKILL);
88        NOTICE("process '%s' killing any children in process group\n", svc->name);
89    }
90
91    /* remove any sockets we may have created */
92    for (si = svc->sockets; si; si = si->next) {
93        char tmp[128];
94        snprintf(tmp, sizeof(tmp), ANDROID_SOCKET_DIR"/%s", si->name);
95        unlink(tmp);
96    }
97
98    svc->pid = 0;
99    svc->flags &= (~SVC_RUNNING);
100
101        /* oneshot processes go into the disabled state on exit */
102    if (svc->flags & SVC_ONESHOT) {
103        svc->flags |= SVC_DISABLED;
104    }
105
106        /* disabled processes do not get restarted automatically */
107    if (svc->flags & SVC_DISABLED) {
108        notify_service_state(svc->name, "stopped");
109        return 0;
110    }
111
112    now = gettime();
113    if (svc->flags & SVC_CRITICAL) {
114        if (svc->time_crashed + CRITICAL_CRASH_WINDOW >= now) {
115            if (++svc->nr_crashed > CRITICAL_CRASH_THRESHOLD) {
116                ERROR("critical process '%s' exited %d times in %d minutes; "
117                      "rebooting into recovery mode\n", svc->name,
118                      CRITICAL_CRASH_THRESHOLD, CRITICAL_CRASH_WINDOW / 60);
119                sync();
120                __reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2,
121                         LINUX_REBOOT_CMD_RESTART2, "recovery");
122                return 0;
123            }
124        } else {
125            svc->time_crashed = now;
126            svc->nr_crashed = 1;
127        }
128    }
129
130    svc->flags |= SVC_RESTARTING;
131
132    /* Execute all onrestart commands for this service. */
133    list_for_each(node, &svc->onrestart.commands) {
134        cmd = node_to_item(node, struct command, clist);
135        cmd->func(cmd->nargs, cmd->args);
136    }
137    notify_service_state(svc->name, "restarting");
138    return 0;
139}
140
141void handle_signal(void)
142{
143    char tmp[32];
144
145    /* we got a SIGCHLD - reap and restart as needed */
146    read(signal_recv_fd, tmp, sizeof(tmp));
147    while (!wait_for_one_process(0))
148        ;
149}
150
151int get_signal_fd()
152{
153    return signal_recv_fd;
154}
155