summaryrefslogtreecommitdiffstats
path: root/src/userdb/userdbd-manager.c
blob: 27db358c728902daf08528ed8777184eea27a265 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
/* SPDX-License-Identifier: LGPL-2.1-or-later */

#include <sys/wait.h>

#include "sd-daemon.h"

#include "env-util.h"
#include "fd-util.h"
#include "fs-util.h"
#include "mkdir.h"
#include "process-util.h"
#include "set.h"
#include "signal-util.h"
#include "socket-util.h"
#include "stdio-util.h"
#include "umask-util.h"
#include "userdbd-manager.h"

#define LISTEN_TIMEOUT_USEC (25 * USEC_PER_SEC)

static int start_workers(Manager *m, bool explicit_request);

static int on_sigchld(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
        Manager *m = ASSERT_PTR(userdata);

        assert(s);

        for (;;) {
                siginfo_t siginfo = {};
                bool removed = false;

                if (waitid(P_ALL, 0, &siginfo, WNOHANG|WEXITED) < 0) {
                        if (errno == ECHILD)
                                break;

                        log_warning_errno(errno, "Failed to invoke waitid(): %m");
                        break;
                }
                if (siginfo.si_pid == 0)
                        break;

                if (set_remove(m->workers_dynamic, PID_TO_PTR(siginfo.si_pid)))
                        removed = true;
                if (set_remove(m->workers_fixed, PID_TO_PTR(siginfo.si_pid)))
                        removed = true;

                if (!removed) {
                        log_warning("Weird, got SIGCHLD for unknown child " PID_FMT ", ignoring.", siginfo.si_pid);
                        continue;
                }

                if (siginfo.si_code == CLD_EXITED) {
                        if (siginfo.si_status == EXIT_SUCCESS)
                                log_debug("Worker " PID_FMT " exited successfully.", siginfo.si_pid);
                        else
                                log_warning("Worker " PID_FMT " died with a failure exit status %i, ignoring.", siginfo.si_pid, siginfo.si_status);
                } else if (siginfo.si_code == CLD_KILLED)
                        log_warning("Worker " PID_FMT " was killed by signal %s, ignoring.", siginfo.si_pid, signal_to_string(siginfo.si_status));
                else if (siginfo.si_code == CLD_DUMPED)
                        log_warning("Worker " PID_FMT " dumped core by signal %s, ignoring.", siginfo.si_pid, signal_to_string(siginfo.si_status));
                else
                        log_warning("Can't handle SIGCHLD of this type");
        }

        (void) start_workers(m, false); /* Fill up workers again if we fell below the low watermark */
        return 0;
}

static int on_sigusr2(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
        Manager *m = ASSERT_PTR(userdata);

        assert(s);

        (void) start_workers(m, /* explicit_request=*/ true); /* Workers told us there's more work, let's add one more worker as long as we are below the high watermark */
        return 0;
}

int manager_new(Manager **ret) {
        _cleanup_(manager_freep) Manager *m = NULL;
        int r;

        m = new(Manager, 1);
        if (!m)
                return -ENOMEM;

        *m = (Manager) {
                .listen_fd = -1,
                .worker_ratelimit = {
                        .interval = 2 * USEC_PER_SEC,
                        .burst = 2500,
                },
        };

        r = sd_event_new(&m->event);
        if (r < 0)
                return r;

        r = sd_event_add_signal(m->event, NULL, SIGINT, NULL, NULL);
        if (r < 0)
                return r;

        r = sd_event_add_signal(m->event, NULL, SIGTERM, NULL, NULL);
        if (r < 0)
                return r;

        (void) sd_event_set_watchdog(m->event, true);

        m->workers_fixed = set_new(NULL);
        m->workers_dynamic = set_new(NULL);

        if (!m->workers_fixed || !m->workers_dynamic)
                return -ENOMEM;

        r = sd_event_add_signal(m->event, &m->sigusr2_event_source, SIGUSR2, on_sigusr2, m);
        if (r < 0)
                return r;

        r = sd_event_add_signal(m->event, &m->sigchld_event_source, SIGCHLD, on_sigchld, m);
        if (r < 0)
                return r;

        *ret = TAKE_PTR(m);
        return 0;
}

Manager* manager_free(Manager *m) {
        if (!m)
                return NULL;

        set_free(m->workers_fixed);
        set_free(m->workers_dynamic);

        sd_event_source_disable_unref(m->sigusr2_event_source);
        sd_event_source_disable_unref(m->sigchld_event_source);

        safe_close(m->listen_fd);

        sd_event_unref(m->event);

        return mfree(m);
}

static size_t manager_current_workers(Manager *m) {
        assert(m);

        return set_size(m->workers_fixed) + set_size(m->workers_dynamic);
}

static int start_one_worker(Manager *m) {
        bool fixed;
        pid_t pid;
        int r;

        assert(m);

        fixed = set_size(m->workers_fixed) < USERDB_WORKERS_MIN;

        r = safe_fork("(sd-worker)", FORK_RESET_SIGNALS|FORK_DEATHSIG|FORK_LOG, &pid);
        if (r < 0)
                return log_error_errno(r, "Failed to fork new worker child: %m");
        if (r == 0) {
                /* Child */

                log_close();

                r = close_all_fds(&m->listen_fd, 1);
                if (r < 0) {
                        log_error_errno(r, "Failed to close fds in child: %m");
                        _exit(EXIT_FAILURE);
                }

                log_open();

                if (m->listen_fd == 3) {
                        r = fd_cloexec(3, false);
                        if (r < 0) {
                                log_error_errno(r, "Failed to turn off O_CLOEXEC for fd 3: %m");
                                _exit(EXIT_FAILURE);
                        }
                } else {
                        if (dup2(m->listen_fd, 3) < 0) { /* dup2() creates with O_CLOEXEC off */
                                log_error_errno(errno, "Failed to move listen fd to 3: %m");
                                _exit(EXIT_FAILURE);
                        }

                        safe_close(m->listen_fd);
                }

                r = setenvf("LISTEN_PID", /* overwrite= */ true, PID_FMT, pid);
                if (r < 0) {
                        log_error_errno(r, "Failed to set $LISTEN_PID: %m");
                        _exit(EXIT_FAILURE);
                }

                if (setenv("LISTEN_FDS", "1", 1) < 0) {
                        log_error_errno(errno, "Failed to set $LISTEN_FDS: %m");
                        _exit(EXIT_FAILURE);
                }


                if (setenv("USERDB_FIXED_WORKER", one_zero(fixed), 1) < 0) {
                        log_error_errno(errno, "Failed to set $USERDB_FIXED_WORKER: %m");
                        _exit(EXIT_FAILURE);
                }

                /* execl("/home/lennart/projects/systemd/build/systemd-userwork", "systemd-userwork", "xxxxxxxxxxxxxxxx", NULL); /\* With some extra space rename_process() can make use of *\/ */
                /* execl("/usr/bin/valgrind", "valgrind", "/home/lennart/projects/systemd/build/systemd-userwork", "systemd-userwork", "xxxxxxxxxxxxxxxx", NULL); /\* With some extra space rename_process() can make use of *\/ */

                execl(SYSTEMD_USERWORK_PATH, "systemd-userwork", "xxxxxxxxxxxxxxxx", NULL); /* With some extra space rename_process() can make use of */
                log_error_errno(errno, "Failed start worker process: %m");
                _exit(EXIT_FAILURE);
        }

        if (fixed)
                r = set_put(m->workers_fixed, PID_TO_PTR(pid));
        else
                r = set_put(m->workers_dynamic, PID_TO_PTR(pid));
        if (r < 0)
                return log_error_errno(r, "Failed to add child process to set: %m");

        return 0;
}

static int start_workers(Manager *m, bool explicit_request) {
        int r;

        assert(m);

        for (;;)  {
                size_t n;

                n = manager_current_workers(m);
                if (n >= USERDB_WORKERS_MIN && (!explicit_request || n >= USERDB_WORKERS_MAX))
                        break;

                if (!ratelimit_below(&m->worker_ratelimit)) {
                        /* If we keep starting workers too often, let's fail the whole daemon, something is wrong */
                        sd_event_exit(m->event, EXIT_FAILURE);

                        return log_error_errno(SYNTHETIC_ERRNO(EUCLEAN), "Worker threads requested too frequently, something is wrong.");
                }

                r = start_one_worker(m);
                if (r < 0)
                        return r;

                explicit_request = false;
        }

        return 0;
}

int manager_startup(Manager *m) {
        int n, r;

        assert(m);
        assert(m->listen_fd < 0);

        n = sd_listen_fds(false);
        if (n < 0)
                return log_error_errno(n, "Failed to determine number of passed file descriptors: %m");
        if (n > 1)
                return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Expected one listening fd, got %i.", n);
        if (n == 1)
                m->listen_fd = SD_LISTEN_FDS_START;
        else {
                union sockaddr_union sockaddr = {
                        .un.sun_family = AF_UNIX,
                        .un.sun_path = "/run/systemd/userdb/io.systemd.Multiplexer",
                };

                r = mkdir_p("/run/systemd/userdb", 0755);
                if (r < 0)
                        return log_error_errno(r, "Failed to create /run/systemd/userdb: %m");

                m->listen_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0);
                if (m->listen_fd < 0)
                        return log_error_errno(errno, "Failed to bind on socket: %m");

                (void) sockaddr_un_unlink(&sockaddr.un);

                RUN_WITH_UMASK(0000)
                        if (bind(m->listen_fd, &sockaddr.sa, SOCKADDR_UN_LEN(sockaddr.un)) < 0)
                                return log_error_errno(errno, "Failed to bind socket: %m");

                r = symlink_idempotent("io.systemd.Multiplexer",
                                       "/run/systemd/userdb/io.systemd.NameServiceSwitch", false);
                if (r < 0)
                        return log_error_errno(r, "Failed to bind io.systemd.Multiplexer: %m");

                r = symlink_idempotent("io.systemd.Multiplexer",
                                       "/run/systemd/userdb/io.systemd.DropIn", false);
                if (r < 0)
                        return log_error_errno(r, "Failed to bind io.systemd.Multiplexer: %m");

                if (listen(m->listen_fd, SOMAXCONN_DELUXE) < 0)
                        return log_error_errno(errno, "Failed to listen on socket: %m");
        }

        /* Let's make sure every accept() call on this socket times out after 25s. This allows workers to be
         * GC'ed on idle */
        if (setsockopt(m->listen_fd, SOL_SOCKET, SO_RCVTIMEO, TIMEVAL_STORE(LISTEN_TIMEOUT_USEC), sizeof(struct timeval)) < 0)
                return log_error_errno(errno, "Failed to se SO_RCVTIMEO: %m");

        return start_workers(m, /* explicit_request= */ false);
}