signal_handler.c revision 3899e9fc01cf608f19f716749c54cc5c4d17e766 
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#include "util.h"
30
31static int signal_fd = -1;
32static int signal_recv_fd = -1;
33
34static void sigchld_handler(int s)
35{
36    write(signal_fd, &s, 1);
37}
38
39void signal_init(void)
40{
41    int s[2];
42
43    struct sigaction act;
44
45    act.sa_handler = sigchld_handler;
46    act.sa_flags = SA_NOCLDSTOP;
47    act.sa_mask = 0;
48    act.sa_restorer = NULL;
49    sigaction(SIGCHLD, &act, 0);
50
51    /* create a signalling mechanism for the sigchld handler */
52    if (socketpair(AF_UNIX, SOCK_STREAM, 0, s) == 0) {
53        signal_fd = s[0];
54        signal_recv_fd = s[1];
55        fcntl(s[0], F_SETFD, FD_CLOEXEC);
56        fcntl(s[0], F_SETFL, O_NONBLOCK);
57        fcntl(s[1], F_SETFD, FD_CLOEXEC);
58        fcntl(s[1], F_SETFL, O_NONBLOCK);
59    }
60}
61
62#define CRITICAL_CRASH_THRESHOLD    4       /* if we crash >4 times ... */
63#define CRITICAL_CRASH_WINDOW       (4*60)  /* ... in 4 minutes, goto recovery*/
64
65static int wait_for_one_process(int block)
66{
67    pid_t pid;
68    int status;
69    struct service *svc;
70    struct socketinfo *si;
71    time_t now;
72    struct listnode *node;
73    struct command *cmd;
74
75    while ( (pid = waitpid(-1, &status, block ? 0 : WNOHANG)) == -1 && errno == EINTR );
76    if (pid <= 0) return -1;
77    INFO("waitpid returned pid %d, status = %08x\n", pid, status);
78
79    svc = service_find_by_pid(pid);
80    if (!svc) {
81        ERROR("untracked pid %d exited\n", pid);
82        return 0;
83    }
84
85    NOTICE("process '%s', pid %d exited\n", svc->name, pid);
86
87    if (!(svc->flags & SVC_ONESHOT)) {
88        kill(-pid, SIGKILL);
89        NOTICE("process '%s' killing any children in process group\n", svc->name);
90    }
91
92    /* remove any sockets we may have created */
93    for (si = svc->sockets; si; si = si->next) {
94        char tmp[128];
95        snprintf(tmp, sizeof(tmp), ANDROID_SOCKET_DIR"/%s", si->name);
96        unlink(tmp);
97    }
98
99    svc->pid = 0;
100    svc->flags &= (~SVC_RUNNING);
101
102        /* oneshot processes go into the disabled state on exit */
103    if (svc->flags & SVC_ONESHOT) {
104        svc->flags |= SVC_DISABLED;
105    }
106
107        /* disabled processes do not get restarted automatically */
108    if (svc->flags & SVC_DISABLED) {
109        notify_service_state(svc->name, "stopped");
110        return 0;
111    }
112
113    now = gettime();
114    if (svc->flags & SVC_CRITICAL) {
115        if (svc->time_crashed + CRITICAL_CRASH_WINDOW >= now) {
116            if (++svc->nr_crashed > CRITICAL_CRASH_THRESHOLD) {
117                ERROR("critical process '%s' exited %d times in %d minutes; "
118                      "rebooting into recovery mode\n", svc->name,
119                      CRITICAL_CRASH_THRESHOLD, CRITICAL_CRASH_WINDOW / 60);
120                sync();
121                __reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2,
122                         LINUX_REBOOT_CMD_RESTART2, "recovery");
123                return 0;
124            }
125        } else {
126            svc->time_crashed = now;
127            svc->nr_crashed = 1;
128        }
129    }
130
131    svc->flags |= SVC_RESTARTING;
132
133    /* Execute all onrestart commands for this service. */
134    list_for_each(node, &svc->onrestart.commands) {
135        cmd = node_to_item(node, struct command, clist);
136        cmd->func(cmd->nargs, cmd->args);
137    }
138    notify_service_state(svc->name, "restarting");
139    return 0;
140}
141
142void handle_signal(void)
143{
144    char tmp[32];
145
146    /* we got a SIGCHLD - reap and restart as needed */
147    read(signal_recv_fd, tmp, sizeof(tmp));
148    while (!wait_for_one_process(0))
149        ;
150}
151
152int get_signal_fd()
153{
154    return signal_recv_fd;
155}
156