libvirtd.c 95.8 KB
Newer Older
D
Daniel P. Berrange 已提交
1
/*
D
Daniel P. Berrange 已提交
2
 * libvirtd.c: daemon start of day, guest process & i/o management
D
Daniel P. Berrange 已提交
3
 *
4
 * Copyright (C) 2006-2010 Red Hat, Inc.
D
Daniel P. Berrange 已提交
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library 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
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307  USA
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

24
#include <config.h>
25

D
Daniel P. Berrange 已提交
26 27 28 29 30 31 32 33 34 35
#include <sys/types.h>
#include <sys/wait.h>
#include <sys/stat.h>
#include <unistd.h>
#include <fcntl.h>
#include <limits.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <sys/poll.h>
#include <netinet/in.h>
36
#include <netinet/tcp.h>
D
Daniel P. Berrange 已提交
37 38 39 40
#include <netdb.h>
#include <stdlib.h>
#include <pwd.h>
#include <stdio.h>
41 42
#include <stdarg.h>
#include <syslog.h>
D
Daniel P. Berrange 已提交
43 44 45
#include <string.h>
#include <errno.h>
#include <getopt.h>
46
#include <fnmatch.h>
47
#include <grp.h>
48
#include <signal.h>
J
Jim Meyering 已提交
49
#include <netdb.h>
E
Eric Blake 已提交
50
#include <locale.h>
51

52
#include "libvirt_internal.h"
53
#include "virterror_internal.h"
54
#include "files.h"
55

56 57
#define VIR_FROM_THIS VIR_FROM_QEMU

D
Daniel P. Berrange 已提交
58
#include "libvirtd.h"
59 60
#include "dispatch.h"

61
#include "util.h"
62
#include "uuid.h"
63
#include "remote_driver.h"
64
#include "conf.h"
D
Daniel P. Berrange 已提交
65
#include "event.h"
66
#include "memory.h"
67
#include "stream.h"
68
#include "hooks.h"
69
#include "virtaudit.h"
70
#ifdef HAVE_AVAHI
71
# include "mdns.h"
72
#endif
D
Daniel P. Berrange 已提交
73

74
#ifdef WITH_DRIVER_MODULES
75
# include "driver.h"
76
#else
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
# ifdef WITH_QEMU
#  include "qemu/qemu_driver.h"
# endif
# ifdef WITH_LXC
#  include "lxc/lxc_driver.h"
# endif
# ifdef WITH_UML
#  include "uml/uml_driver.h"
# endif
# ifdef WITH_ONE
#  include "opennebula/one_driver.h"
# endif
# ifdef WITH_NETWORK
#  include "network/bridge_driver.h"
# endif
# ifdef WITH_NETCF
#  include "interface/netcf_driver.h"
# endif
# ifdef WITH_STORAGE_DIR
#  include "storage/storage_driver.h"
# endif
# ifdef WITH_NODE_DEVICES
#  include "node_device/node_device_driver.h"
# endif
# ifdef WITH_SECRETS
#  include "secret/secret_driver.h"
# endif
104 105 106
# ifdef WITH_NWFILTER
#  include "nwfilter/nwfilter_driver.h"
# endif
107
#endif
108 109


J
John Levon 已提交
110
#ifdef __sun
111 112
# include <ucred.h>
# include <priv.h>
J
John Levon 已提交
113

114 115 116
# ifndef PRIV_VIRT_MANAGE
#  define PRIV_VIRT_MANAGE ((const char *)"virt_manage")
# endif
J
John Levon 已提交
117

118 119 120
# ifndef PRIV_XVM_CONTROL
#  define PRIV_XVM_CONTROL ((const char *)"xvm_control")
# endif
J
John Levon 已提交
121

122 123
# define PU_RESETGROUPS          0x0001  /* Remove supplemental groups */
# define PU_CLEARLIMITSET        0x0008  /* L=0 */
J
John Levon 已提交
124 125 126

extern int __init_daemon_priv(int, uid_t, gid_t, ...);

127
# define SYSTEM_UID 60
J
John Levon 已提交
128 129 130 131 132 133 134 135 136 137 138 139 140

static gid_t unix_sock_gid = 60; /* Not used */
static int unix_sock_rw_mask = 0666;
static int unix_sock_ro_mask = 0666;

#else

static gid_t unix_sock_gid = 0; /* Only root by default */
static int unix_sock_rw_mask = 0700; /* Allow user only */
static int unix_sock_ro_mask = 0777; /* Allow world */

#endif /* __sun */

141 142
#include "configmake.h"

143 144
static int godaemon = 0;        /* -d: Be a daemon */
static int verbose = 0;         /* -v: Verbose mode */
145
static int timeout = -1;        /* -t: Shutdown timeout */
146
static int sigwrite = -1;       /* Signal handler pipe */
147
static int ipsock = 0;          /* -l  Listen for TCP/IP */
148

149
/* Defaults for configuration file elements */
150 151
static int listen_tls = 1;
static int listen_tcp = 0;
152
static char *listen_addr  = (char *) LIBVIRTD_LISTEN_ADDR;
153 154
static char *tls_port = (char *) LIBVIRTD_TLS_PORT;
static char *tcp_port = (char *) LIBVIRTD_TCP_PORT;
155

156 157
static char *unix_sock_dir = NULL;

158 159 160 161
#if HAVE_POLKIT
static int auth_unix_rw = REMOTE_AUTH_POLKIT;
static int auth_unix_ro = REMOTE_AUTH_POLKIT;
#else
162 163
static int auth_unix_rw = REMOTE_AUTH_NONE;
static int auth_unix_ro = REMOTE_AUTH_NONE;
164
#endif /* HAVE_POLKIT */
165 166 167 168 169 170 171
#if HAVE_SASL
static int auth_tcp = REMOTE_AUTH_SASL;
#else
static int auth_tcp = REMOTE_AUTH_NONE;
#endif
static int auth_tls = REMOTE_AUTH_NONE;

172
static int mdns_adv = 1;
173
static char *mdns_name = NULL;
174

175
static int tls_no_verify_certificate = 0;
176
static char **tls_allowed_dn_list = NULL;
177

178 179 180 181
static char *key_file = (char *) LIBVIRT_SERVERKEY;
static char *cert_file = (char *) LIBVIRT_SERVERCERT;
static char *ca_file = (char *) LIBVIRT_CACERT;
static char *crl_file = (char *) "";
182 183 184 185

static gnutls_certificate_credentials_t x509_cred;
static gnutls_dh_params_t dh_params;

186 187 188 189
static int min_workers = 5;
static int max_workers = 20;
static int max_clients = 20;

190 191 192 193 194
/* Total number of 'in-process' RPC calls allowed across all clients */
static int max_requests = 20;
/* Total number of 'in-process' RPC calls allowed by a single client*/
static int max_client_requests = 5;

195 196 197
static int audit_level = 1;
static int audit_logging = 0;

198
#define DH_BITS 1024
199

200 201
static sig_atomic_t sig_errors = 0;
static int sig_lasterrno = 0;
202
static const char *argv0;
203

204 205 206 207 208 209 210 211 212
enum {
    VIR_DAEMON_ERR_NONE = 0,
    VIR_DAEMON_ERR_PIDFILE,
    VIR_DAEMON_ERR_RUNDIR,
    VIR_DAEMON_ERR_INIT,
    VIR_DAEMON_ERR_SIGNAL,
    VIR_DAEMON_ERR_PRIVS,
    VIR_DAEMON_ERR_NETWORK,
    VIR_DAEMON_ERR_CONFIG,
213
    VIR_DAEMON_ERR_HOOKS,
214
    VIR_DAEMON_ERR_AUDIT,
215 216 217 218 219 220 221 222 223 224 225 226 227

    VIR_DAEMON_ERR_LAST
};

VIR_ENUM_DECL(virDaemonErr)
VIR_ENUM_IMPL(virDaemonErr, VIR_DAEMON_ERR_LAST,
              "Initialization successful",
              "Unable to obtain pidfile",
              "Unable to create rundir",
              "Unable to initialize libvirt",
              "Unable to setup signal handlers",
              "Unable to drop privileges",
              "Unable to initialize network sockets",
228
              "Unable to load configuration file",
229 230
              "Unable to look for hook scripts",
              "Unable to initialize audit system")
231

232 233
static void sig_handler(int sig, siginfo_t * siginfo,
                        void* context ATTRIBUTE_UNUSED) {
234
    int origerrno;
235
    int r;
236

237 238
    /* set the sig num in the struct */
    siginfo->si_signo = sig;
239 240

    origerrno = errno;
241
    r = safewrite(sigwrite, siginfo, sizeof(*siginfo));
242 243 244 245
    if (r == -1) {
        sig_errors++;
        sig_lasterrno = errno;
    }
246
    errno = origerrno;
D
Daniel P. Berrange 已提交
247
}
248

249 250
static void qemudDispatchClientEvent(int watch, int fd, int events, void *opaque);
static void qemudDispatchServerEvent(int watch, int fd, int events, void *opaque);
251
static int qemudStartWorker(struct qemud_server *server, struct qemud_worker *worker);
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267

void
qemudClientMessageQueuePush(struct qemud_client_message **queue,
                            struct qemud_client_message *msg)
{
    struct qemud_client_message *tmp = *queue;

    if (tmp) {
        while (tmp->next)
            tmp = tmp->next;
        tmp->next = msg;
    } else {
        *queue = msg;
    }
}

268
struct qemud_client_message *
269 270 271 272 273 274 275 276 277 278 279
qemudClientMessageQueueServe(struct qemud_client_message **queue)
{
    struct qemud_client_message *tmp = *queue;

    if (tmp) {
        *queue = tmp->next;
        tmp->next = NULL;
    }

    return tmp;
}
D
Daniel P. Berrange 已提交
280

281 282 283 284 285
static int
remoteCheckCertFile(const char *type, const char *file)
{
    struct stat sb;
    if (stat(file, &sb) < 0) {
286 287 288
        char ebuf[1024];
        VIR_ERROR(_("Cannot access %s '%s': %s"),
                  type, file, virStrerror(errno, ebuf, sizeof ebuf));
289 290 291 292 293
        return -1;
    }
    return 0;
}

294 295 296 297 298 299 300 301 302 303
static int
remoteInitializeGnuTLS (void)
{
    int err;

    /* Initialise GnuTLS. */
    gnutls_global_init ();

    err = gnutls_certificate_allocate_credentials (&x509_cred);
    if (err) {
304
        VIR_ERROR(_("gnutls_certificate_allocate_credentials: %s"),
305 306 307 308 309
                  gnutls_strerror (err));
        return -1;
    }

    if (ca_file && ca_file[0] != '\0') {
310 311 312
        if (remoteCheckCertFile("CA certificate", ca_file) < 0)
            return -1;

313 314 315 316
        qemudDebug ("loading CA cert from %s", ca_file);
        err = gnutls_certificate_set_x509_trust_file (x509_cred, ca_file,
                                                      GNUTLS_X509_FMT_PEM);
        if (err < 0) {
317
            VIR_ERROR(_("gnutls_certificate_set_x509_trust_file: %s"),
318 319 320 321 322 323
                      gnutls_strerror (err));
            return -1;
        }
    }

    if (crl_file && crl_file[0] != '\0') {
J
Jim Meyering 已提交
324
        if (remoteCheckCertFile("CA revocation list", crl_file) < 0)
325 326
            return -1;

327
        DEBUG("loading CRL from %s", crl_file);
328 329 330
        err = gnutls_certificate_set_x509_crl_file (x509_cred, crl_file,
                                                    GNUTLS_X509_FMT_PEM);
        if (err < 0) {
331
            VIR_ERROR(_("gnutls_certificate_set_x509_crl_file: %s"),
332 333 334 335 336 337
                      gnutls_strerror (err));
            return -1;
        }
    }

    if (cert_file && cert_file[0] != '\0' && key_file && key_file[0] != '\0') {
338 339 340 341
        if (remoteCheckCertFile("server certificate", cert_file) < 0)
            return -1;
        if (remoteCheckCertFile("server key", key_file) < 0)
            return -1;
342
        DEBUG("loading cert and key from %s and %s", cert_file, key_file);
343 344 345 346 347
        err =
            gnutls_certificate_set_x509_key_file (x509_cred,
                                                  cert_file, key_file,
                                                  GNUTLS_X509_FMT_PEM);
        if (err < 0) {
348
            VIR_ERROR(_("gnutls_certificate_set_x509_key_file: %s"),
349 350 351 352 353 354 355 356 357 358 359 360
                      gnutls_strerror (err));
            return -1;
        }
    }

    /* Generate Diffie Hellman parameters - for use with DHE
     * kx algorithms. These should be discarded and regenerated
     * once a day, once a week or once a month. Depending on the
     * security requirements.
     */
    err = gnutls_dh_params_init (&dh_params);
    if (err < 0) {
361
        VIR_ERROR(_("gnutls_dh_params_init: %s"), gnutls_strerror (err));
362 363 364 365
        return -1;
    }
    err = gnutls_dh_params_generate2 (dh_params, DH_BITS);
    if (err < 0) {
366
        VIR_ERROR(_("gnutls_dh_params_generate2: %s"), gnutls_strerror (err));
367 368 369 370 371 372 373 374
        return -1;
    }

    gnutls_certificate_set_dh_params (x509_cred, dh_params);

    return 0;
}

375
static void
376 377
qemudDispatchSignalEvent(int watch ATTRIBUTE_UNUSED,
                         int fd ATTRIBUTE_UNUSED,
378 379
                         int events ATTRIBUTE_UNUSED,
                         void *opaque) {
D
Daniel P. Berrange 已提交
380
    struct qemud_server *server = (struct qemud_server *)opaque;
381
    siginfo_t siginfo;
382

383
    virMutexLock(&server->lock);
384

385
    if (saferead(server->sigread, &siginfo, sizeof(siginfo)) != sizeof(siginfo)) {
386 387 388
        char ebuf[1024];
        VIR_ERROR(_("Failed to read from signal pipe: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
389
        virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
390
        return;
391 392
    }

393
    switch (siginfo.si_signo) {
394
    case SIGHUP:
395
        VIR_INFO0(_("Reloading configuration on SIGHUP"));
396 397
        virHookCall(VIR_HOOK_DRIVER_DAEMON, "-",
                    VIR_HOOK_DAEMON_OP_RELOAD, SIGHUP, "SIGHUP", NULL);
398
        if (virStateReload() < 0)
399
            VIR_WARN0("Error while reloading drivers");
400

401 402 403
        break;

    case SIGINT:
404
    case SIGQUIT:
405
    case SIGTERM:
406
        VIR_WARN("Shutting down on signal %d", siginfo.si_signo);
407
        server->quitEventThread = 1;
408 409 410
        break;

    default:
411
        VIR_INFO(_("Received unexpected signal %d"), siginfo.si_signo);
412 413 414
        break;
    }

415
    virMutexUnlock(&server->lock);
416 417
}

418

419 420 421 422 423
static int daemonForkIntoBackground(void) {
    int statuspipe[2];
    if (pipe(statuspipe) < 0)
        return -1;

D
Daniel P. Berrange 已提交
424 425 426 427 428 429
    int pid = fork();
    switch (pid) {
    case 0:
        {
            int stdinfd = -1;
            int stdoutfd = -1;
430
            int nextpid;
D
Daniel P. Berrange 已提交
431

432
            VIR_FORCE_CLOSE(statuspipe[0]);
433

434
            if ((stdinfd = open("/dev/null", O_RDONLY)) < 0)
D
Daniel P. Berrange 已提交
435
                goto cleanup;
436
            if ((stdoutfd = open("/dev/null", O_WRONLY)) < 0)
D
Daniel P. Berrange 已提交
437 438 439 440 441 442 443
                goto cleanup;
            if (dup2(stdinfd, STDIN_FILENO) != STDIN_FILENO)
                goto cleanup;
            if (dup2(stdoutfd, STDOUT_FILENO) != STDOUT_FILENO)
                goto cleanup;
            if (dup2(stdoutfd, STDERR_FILENO) != STDERR_FILENO)
                goto cleanup;
444
            if (VIR_CLOSE(stdinfd) < 0)
D
Daniel P. Berrange 已提交
445
                goto cleanup;
446
            if (VIR_CLOSE(stdoutfd) < 0)
D
Daniel P. Berrange 已提交
447 448 449 450 451 452 453 454
                goto cleanup;

            if (setsid() < 0)
                goto cleanup;

            nextpid = fork();
            switch (nextpid) {
            case 0:
455
                return statuspipe[1];
D
Daniel P. Berrange 已提交
456 457 458
            case -1:
                return -1;
            default:
459
                _exit(0);
D
Daniel P. Berrange 已提交
460 461 462
            }

        cleanup:
463 464
            VIR_FORCE_CLOSE(stdoutfd);
            VIR_FORCE_CLOSE(stdinfd);
D
Daniel P. Berrange 已提交
465 466 467 468 469 470 471 472 473
            return -1;

        }

    case -1:
        return -1;

    default:
        {
474 475 476 477
            int got, exitstatus = 0;
            int ret;
            char status;

478
            VIR_FORCE_CLOSE(statuspipe[1]);
479 480 481

            /* We wait to make sure the first child forked successfully */
            if ((got = waitpid(pid, &exitstatus, 0)) < 0 ||
D
Daniel P. Berrange 已提交
482
                got != pid ||
483
                exitstatus != 0) {
D
Daniel P. Berrange 已提交
484 485
                return -1;
            }
486 487 488 489 490 491 492 493

            /* Now block until the second child initializes successfully */
        again:
            ret = read(statuspipe[0], &status, 1);
            if (ret == -1 && errno == EINTR)
                goto again;

            if (ret == 1 && status != 0) {
494
                fprintf(stderr,
495 496
                        _("%s: error: %s. Check /var/log/messages or run without "
                          "--daemon for more info.\n"), argv0,
497
                        virDaemonErrTypeToString(status));
498 499
            }
            _exit(ret == 1 && status == 0 ? 0 : 1);
D
Daniel P. Berrange 已提交
500 501 502 503
        }
    }
}

504 505 506
static int qemudWritePidFile(const char *pidFile) {
    int fd;
    FILE *fh;
507
    char ebuf[1024];
508 509 510 511 512

    if (pidFile[0] == '\0')
        return 0;

    if ((fd = open(pidFile, O_WRONLY|O_CREAT|O_EXCL, 0644)) < 0) {
513
        VIR_ERROR(_("Failed to open pid file '%s' : %s"),
514
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
515 516 517
        return -1;
    }

518
    if (!(fh = VIR_FDOPEN(fd, "w"))) {
519
        VIR_ERROR(_("Failed to fdopen pid file '%s' : %s"),
520
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
521
        VIR_FORCE_CLOSE(fd);
522 523 524 525
        return -1;
    }

    if (fprintf(fh, "%lu\n", (unsigned long)getpid()) < 0) {
526 527
        VIR_ERROR(_("%s: Failed to write to pid file '%s' : %s"),
                  argv0, pidFile, virStrerror(errno, ebuf, sizeof ebuf));
528
        VIR_FORCE_FCLOSE(fh);
529 530 531
        return -1;
    }

532
    if (VIR_FCLOSE(fh) == EOF) {
533 534
        VIR_ERROR(_("%s: Failed to close pid file '%s' : %s"),
                  argv0, pidFile, virStrerror(errno, ebuf, sizeof ebuf));
535 536 537 538 539 540
        return -1;
    }

    return 0;
}

D
Daniel P. Berrange 已提交
541
static int qemudListenUnix(struct qemud_server *server,
542
                           char *path, int readonly, int auth) {
543
    struct qemud_socket *sock;
D
Daniel P. Berrange 已提交
544
    mode_t oldmask;
545
    char ebuf[1024];
D
Daniel P. Berrange 已提交
546

547
    if (VIR_ALLOC(sock) < 0) {
548
        VIR_ERROR0(_("Failed to allocate memory for struct qemud_socket"));
D
Daniel P. Berrange 已提交
549
        return -1;
550
    }
D
Daniel P. Berrange 已提交
551 552

    sock->readonly = readonly;
553
    sock->type = QEMUD_SOCK_TYPE_UNIX;
554
    sock->auth = auth;
555
    sock->path = path;
556 557 558 559 560 561
    sock->addr.len = sizeof(sock->addr.data.un);
    if (!(sock->addrstr = strdup(path))) {
        VIR_ERROR(_("Failed to copy socket address: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
562

563
    if ((sock->fd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
564
        VIR_ERROR(_("Failed to create socket: %s"),
565
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
566
        goto cleanup;
567
    }
D
Daniel P. Berrange 已提交
568

569 570
    if (virSetCloseExec(sock->fd) < 0 ||
        virSetNonBlock(sock->fd) < 0)
D
Daniel P. Berrange 已提交
571
        goto cleanup;
D
Daniel P. Berrange 已提交
572

573 574
    sock->addr.data.un.sun_family = AF_UNIX;
    if (virStrcpyStatic(sock->addr.data.un.sun_path, path) == NULL) {
C
Chris Lalancette 已提交
575 576 577
        VIR_ERROR(_("Path %s too long for unix socket"), path);
        goto cleanup;
    }
578 579
    if (sock->addr.data.un.sun_path[0] == '@')
        sock->addr.data.un.sun_path[0] = '\0';
D
Daniel P. Berrange 已提交
580

581
    oldmask = umask(readonly ? ~unix_sock_ro_mask : ~unix_sock_rw_mask);
582
    if (bind(sock->fd, &sock->addr.data.sa, sock->addr.len) < 0) {
583
        VIR_ERROR(_("Failed to bind socket to '%s': %s"),
584
                  path, virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
585
        goto cleanup;
586
    }
D
Daniel P. Berrange 已提交
587
    umask(oldmask);
588 589 590 591 592 593 594 595

    /* chown() doesn't work for abstract sockets but we use them only
     * if libvirtd runs unprivileged
     */
    if (server->privileged && chown(path, -1, unix_sock_gid)) {
        VIR_ERROR(_("Failed to change group ID of '%s' to %d: %s"),
                  path, unix_sock_gid,
                  virStrerror(errno, ebuf, sizeof ebuf));
596 597
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
598

599
    if (listen(sock->fd, 30) < 0) {
600
        VIR_ERROR(_("Failed to listen for connections on '%s': %s"),
601
                  path, virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
602 603 604 605 606 607 608
        goto cleanup;
    }

    sock->next = server->sockets;
    server->sockets = sock;
    server->nsockets++;

D
Daniel P. Berrange 已提交
609
    return 0;
D
Daniel P. Berrange 已提交
610 611

 cleanup:
612
    VIR_FORCE_CLOSE(sock->fd);
613
    VIR_FREE(sock);
D
Daniel P. Berrange 已提交
614
    return -1;
D
Daniel P. Berrange 已提交
615 616
}

617 618
// See: http://people.redhat.com/drepper/userapi-ipv6.html
static int
619
remoteMakeSockets (int *fds, int max_fds, int *nfds_r, const char *node, const char *service)
620 621 622 623 624 625 626
{
    struct addrinfo *ai;
    struct addrinfo hints;
    memset (&hints, 0, sizeof hints);
    hints.ai_flags = AI_PASSIVE | AI_ADDRCONFIG;
    hints.ai_socktype = SOCK_STREAM;

627
    int e = getaddrinfo (node, service, &hints, &ai);
628
    if (e != 0) {
629
        VIR_ERROR(_("getaddrinfo: %s"), gai_strerror (e));
630 631 632 633 634
        return -1;
    }

    struct addrinfo *runp = ai;
    while (runp && *nfds_r < max_fds) {
635
        char ebuf[1024];
636 637 638
        fds[*nfds_r] = socket (runp->ai_family, runp->ai_socktype,
                               runp->ai_protocol);
        if (fds[*nfds_r] == -1) {
639
            VIR_ERROR(_("socket: %s"), virStrerror (errno, ebuf, sizeof ebuf));
640 641 642 643 644 645
            return -1;
        }

        int opt = 1;
        setsockopt (fds[*nfds_r], SOL_SOCKET, SO_REUSEADDR, &opt, sizeof opt);

646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
#ifdef IPV6_V6ONLY
        if (runp->ai_family == PF_INET6) {
            int on = 1;
            /*
             * Normally on Linux an INET6 socket will bind to the INET4
             * address too. If getaddrinfo returns results with INET4
             * first though, this will result in INET6 binding failing.
             * We can trivially cope with multiple server sockets, so
             * we force it to only listen on IPv6
             */
            setsockopt(fds[*nfds_r], IPPROTO_IPV6,IPV6_V6ONLY,
                       (void*)&on, sizeof on);
        }
#endif

661 662
        if (bind (fds[*nfds_r], runp->ai_addr, runp->ai_addrlen) == -1) {
            if (errno != EADDRINUSE) {
663
                VIR_ERROR(_("bind: %s"), virStrerror (errno, ebuf, sizeof ebuf));
664 665
                return -1;
            }
666
            VIR_FORCE_CLOSE(fds[*nfds_r]);
667
        } else {
668 669 670 671
            ++*nfds_r;
        }
        runp = runp->ai_next;
    }
672

673 674 675 676 677 678 679 680 681
    freeaddrinfo (ai);
    return 0;
}

/* Listen on the named/numbered TCP port.  On a machine with IPv4 and
 * IPv6 interfaces this may generate several sockets.
 */
static int
remoteListenTCP (struct qemud_server *server,
682
                 const char *addr,
683
                 const char *port,
684
                 int type,
685
                 int auth)
686 687 688
{
    int fds[2];
    int nfds = 0;
689
    int i;
690 691
    struct qemud_socket *sock;

692
    if (remoteMakeSockets (fds, 2, &nfds, addr, port) == -1)
693 694 695
        return -1;

    for (i = 0; i < nfds; ++i) {
696
        char ebuf[1024];
697

698
        if (VIR_ALLOC(sock) < 0) {
699 700
            VIR_ERROR(_("remoteListenTCP: calloc: %s"),
                      virStrerror (errno, ebuf, sizeof ebuf));
701
            goto cleanup;
702 703
        }

704
        sock->addr.len = sizeof(sock->addr.data.stor);
705 706 707 708 709 710
        sock->readonly = 0;
        sock->next = server->sockets;
        server->sockets = sock;
        server->nsockets++;

        sock->fd = fds[i];
711
        sock->type = type;
712
        sock->auth = auth;
D
Daniel P. Berrange 已提交
713

714
        if (getsockname(sock->fd, &sock->addr.data.sa, &sock->addr.len) < 0)
715
            goto cleanup;
716

717 718
        if (!(sock->addrstr = virSocketFormatAddrFull(&sock->addr, true, ";")))
            goto cleanup;
719

720 721
        if (virSetCloseExec(sock->fd) < 0 ||
            virSetNonBlock(sock->fd) < 0)
722
            goto cleanup;
723 724

        if (listen (sock->fd, 30) < 0) {
725 726
            VIR_ERROR(_("remoteListenTCP: listen: %s"),
                      virStrerror (errno, ebuf, sizeof ebuf));
727
            goto cleanup;
728
        }
729 730 731
    }

    return 0;
732 733 734

cleanup:
    for (i = 0; i < nfds; ++i)
735
        VIR_FORCE_CLOSE(fds[i]);
736
    return -1;
737 738 739
}

static int qemudInitPaths(struct qemud_server *server,
740 741
                          char **sockname,
                          char **roSockname)
J
John Levon 已提交
742
{
743
    char *base_dir_prefix = NULL;
744
    char *sock_dir_prefix = NULL;
745 746 747 748 749
    int ret = -1;

    /* The base_dir_prefix is the base under which all libvirtd
     * files live */
    if (server->privileged) {
750
        if (!(base_dir_prefix = strdup (LOCALSTATEDIR)))
751 752 753 754 755 756
            goto no_memory;
    } else {
        uid_t uid = geteuid();
        if (!(base_dir_prefix = virGetUserDirectory(uid)))
            goto cleanup;
    }
757

758 759
    /* The unix_sock_dir is the location under which all
     * unix domain sockets live */
D
Daniel P. Berrange 已提交
760
    if (unix_sock_dir) {
761 762 763
        if (!(sock_dir_prefix = strdup(unix_sock_dir)))
            goto no_memory;

D
Daniel P. Berrange 已提交
764 765 766 767 768 769 770
        /* Change the group ownership of /var/run/libvirt to unix_sock_gid */
        if (server->privileged) {
            if (chown(unix_sock_dir, -1, unix_sock_gid) < 0)
                VIR_ERROR(_("Failed to change group ownership of %s"),
                          unix_sock_dir);
        }
    } else {
771
        if (server->privileged) {
772 773 774
            if (virAsprintf(&sock_dir_prefix, "%s/run/libvirt",
                            base_dir_prefix) < 0)
                goto no_memory;
775
        } else {
776 777 778
            if (virAsprintf(&sock_dir_prefix, "%s/.libvirt",
                            base_dir_prefix) < 0)
                goto no_memory;
779 780
        }
    }
781

782
    if (server->privileged) {
783 784 785 786 787 788 789 790
        if (virAsprintf(sockname, "%s/libvirt-sock",
                        sock_dir_prefix) < 0)
            goto no_memory;
        if (virAsprintf(roSockname, "%s/libvirt-sock-ro",
                        sock_dir_prefix) < 0)
            goto no_memory;
        unlink(*sockname);
        unlink(*roSockname);
D
Daniel P. Berrange 已提交
791
    } else {
792 793 794 795 796 797
        if (virAsprintf(sockname, "@%s/libvirt-sock",
                        sock_dir_prefix) < 0)
            goto no_memory;
        /* There is no RO socket in unprivileged mode,
         * since the user always has full RW access
         * to their private instance */
798
    }
799

800
    if (server->privileged) {
801 802 803
        if (virAsprintf(&server->logDir, "%s/log/libvirt",
                        base_dir_prefix) < 0)
            goto no_memory;
804
    } else {
805 806 807
        if (virAsprintf(&server->logDir, "%s/.libvirt/log",
                        base_dir_prefix) < 0)
            goto no_memory;
808
    }
809

810
    ret = 0;
811

812 813 814
no_memory:
    if (ret != 0)
        virReportOOMError();
815 816

 cleanup:
817
    VIR_FREE(base_dir_prefix);
818
    VIR_FREE(sock_dir_prefix);
819
    return ret;
D
Daniel P. Berrange 已提交
820 821
}

822 823 824 825 826 827
static void virshErrorHandler(void *opaque ATTRIBUTE_UNUSED, virErrorPtr err ATTRIBUTE_UNUSED)
{
    /* Don't do anything, since logging infrastructure already
     * took care of reporting the error */
}

828
static struct qemud_server *qemudInitialize(void) {
D
Daniel P. Berrange 已提交
829 830
    struct qemud_server *server;

831
    if (VIR_ALLOC(server) < 0) {
832
        VIR_ERROR0(_("Failed to allocate struct qemud_server"));
D
Daniel P. Berrange 已提交
833
        return NULL;
834
    }
D
Daniel P. Berrange 已提交
835

836 837 838
    server->privileged = geteuid() == 0 ? 1 : 0;
    server->sigread = server->sigwrite = -1;

839
    if (virMutexInit(&server->lock) < 0) {
840
        VIR_ERROR0(_("cannot initialize mutex"));
841
        VIR_FREE(server);
842
        return NULL;
843 844
    }
    if (virCondInit(&server->job) < 0) {
845
        VIR_ERROR0(_("cannot initialize condition variable"));
846
        virMutexDestroy(&server->lock);
847
        VIR_FREE(server);
848
        return NULL;
849 850
    }

851
    if (virEventInit() < 0) {
852
        VIR_ERROR0(_("Failed to initialize event system"));
853 854 855
        virMutexDestroy(&server->lock);
        if (virCondDestroy(&server->job) < 0)
        {}
856 857 858 859
        VIR_FREE(server);
        return NULL;
    }

860 861 862 863 864 865 866
    /*
     * Note that the order is important: the first ones have a higher
     * priority when calling virStateInitialize. We must register
     * the network, storage and nodedev drivers before any domain
     * drivers, since their resources must be auto-started before
     * any domains can be auto-started.
     */
867 868 869
#ifdef WITH_DRIVER_MODULES
    /* We don't care if any of these fail, because the whole point
     * is to allow users to only install modules they want to use.
D
Dan Kenigsberg 已提交
870
     * If they try to open a connection for a module that
871 872 873 874
     * is not loaded they'll get a suitable error at that point
     */
    virDriverLoadModule("network");
    virDriverLoadModule("storage");
875
    virDriverLoadModule("nodedev");
876
    virDriverLoadModule("secret");
877 878 879
    virDriverLoadModule("qemu");
    virDriverLoadModule("lxc");
    virDriverLoadModule("uml");
D
Daniel Veillard 已提交
880
    virDriverLoadModule("one");
881
    virDriverLoadModule("nwfilter");
882
#else
883
# ifdef WITH_NETWORK
884
    networkRegister();
885 886
# endif
# ifdef WITH_NETCF
887
    interfaceRegister();
888 889
# endif
# ifdef WITH_STORAGE_DIR
890
    storageRegister();
891 892
# endif
# if defined(WITH_NODE_DEVICES)
893
    nodedevRegister();
894 895
# endif
# ifdef WITH_SECRETS
896
    secretRegister();
897
# endif
898 899 900
# ifdef WITH_NWFILTER
    nwfilterRegister();
# endif
901
# ifdef WITH_QEMU
902
    qemuRegister();
903 904
# endif
# ifdef WITH_LXC
905
    lxcRegister();
906 907
# endif
# ifdef WITH_UML
908
    umlRegister();
909 910
# endif
# ifdef WITH_ONE
D
Daniel Veillard 已提交
911
    oneRegister();
912
# endif
913 914
#endif

915 916 917 918 919 920
    virEventRegisterImpl(virEventAddHandleImpl,
                         virEventUpdateHandleImpl,
                         virEventRemoveHandleImpl,
                         virEventAddTimeoutImpl,
                         virEventUpdateTimeoutImpl,
                         virEventRemoveTimeoutImpl);
921

922 923 924
    return server;
}

925
static int qemudNetworkInit(struct qemud_server *server) {
926 927
    char *sockname = NULL;
    char *roSockname = NULL;
928
#if HAVE_SASL
929 930 931
    int err;
#endif /* HAVE_SASL */

932
    if (qemudInitPaths(server, &sockname, &roSockname) < 0)
933 934 935
        goto cleanup;

    if (qemudListenUnix(server, sockname, 0, auth_unix_rw) < 0)
936
        goto cleanup;
937
    sockname = NULL;
938

939
    if (roSockname != NULL && qemudListenUnix(server, roSockname, 1, auth_unix_ro) < 0)
940
        goto cleanup;
941
    roSockname = NULL;
942 943 944 945 946 947 948

#if HAVE_SASL
    if (auth_unix_rw == REMOTE_AUTH_SASL ||
        auth_unix_ro == REMOTE_AUTH_SASL ||
        auth_tcp == REMOTE_AUTH_SASL ||
        auth_tls == REMOTE_AUTH_SASL) {
        if ((err = sasl_server_init(NULL, "libvirt")) != SASL_OK) {
949 950
            VIR_ERROR(_("Failed to initialize SASL authentication %s"),
                      sasl_errstring(err, NULL, NULL));
951 952
            goto cleanup;
        }
953 954 955
    }
#endif

956
#if HAVE_POLKIT0
957 958 959
    if (auth_unix_rw == REMOTE_AUTH_POLKIT ||
        auth_unix_ro == REMOTE_AUTH_POLKIT) {
        DBusError derr;
960 961 962 963

        dbus_connection_set_change_sigpipe(FALSE);
        dbus_threads_init_default();

964 965 966
        dbus_error_init(&derr);
        server->sysbus = dbus_bus_get(DBUS_BUS_SYSTEM, &derr);
        if (!(server->sysbus)) {
967 968
            VIR_ERROR(_("Failed to connect to system bus for PolicyKit auth: %s"),
                      derr.message);
969 970 971
            dbus_error_free(&derr);
            goto cleanup;
        }
972
        dbus_connection_set_exit_on_disconnect(server->sysbus, FALSE);
973 974 975
    }
#endif

976
    if (ipsock) {
977
        if (listen_tcp && remoteListenTCP (server, listen_addr, tcp_port, QEMUD_SOCK_TYPE_TCP, auth_tcp) < 0)
978 979 980 981 982 983
            goto cleanup;

        if (listen_tls) {
            if (remoteInitializeGnuTLS () < 0)
                goto cleanup;

984
            if (remoteListenTCP (server, listen_addr, tls_port, QEMUD_SOCK_TYPE_TLS, auth_tls) < 0)
985 986
                goto cleanup;
        }
D
Daniel P. Berrange 已提交
987 988
    }

989
#ifdef HAVE_AVAHI
990
    if (server->privileged && mdns_adv) {
991
        struct libvirtd_mdns_group *group;
992
        struct qemud_socket *sock;
993
        int port = 0;
994
        int ret;
995 996 997 998

        server->mdns = libvirtd_mdns_new();

        if (!mdns_name) {
999 1000
            char *groupname, *localhost, *tmp;

1001 1002
            localhost = virGetHostname(NULL);
            if (localhost == NULL)
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
                /* we couldn't resolve the hostname; assume that we are
                 * running in disconnected operation, and report a less
                 * useful Avahi string
                 */
                ret = virAsprintf(&groupname, "Virtualization Host");
            else {
                /* Extract the host part of the potentially FQDN */
                if ((tmp = strchr(localhost, '.')))
                    *tmp = '\0';
                ret = virAsprintf(&groupname, "Virtualization Host %s",
                                  localhost);
            }
            VIR_FREE(localhost);
            if (ret < 0) {
                virReportOOMError();
1018
                goto cleanup;
1019
            }
1020
            group = libvirtd_mdns_add_group(server->mdns, groupname);
1021
            VIR_FREE(groupname);
1022 1023 1024 1025
        } else {
            group = libvirtd_mdns_add_group(server->mdns, mdns_name);
        }

1026
        /*
1027 1028 1029 1030 1031 1032
         * See if there's a TLS enabled port we can advertise. Cowardly
         * don't bother to advertise TCP since we don't want people using
         * them for real world apps
         */
        sock = server->sockets;
        while (sock) {
1033 1034 1035
            if (virSocketGetPort(&sock->addr) != -1 &&
                sock->type == QEMUD_SOCK_TYPE_TLS) {
                port = virSocketGetPort(&sock->addr);
1036 1037 1038 1039
                break;
            }
            sock = sock->next;
        }
1040

1041 1042 1043 1044 1045 1046 1047 1048 1049
        /*
         * Add the primary entry - we choose SSH because its most likely to always
         * be available
         */
        libvirtd_mdns_add_entry(group, "_libvirt._tcp", port);
        libvirtd_mdns_start(server->mdns);
    }
#endif

1050
    return 0;
D
Daniel P. Berrange 已提交
1051 1052

 cleanup:
1053 1054
    VIR_FREE(sockname);
    VIR_FREE(roSockname);
1055
    return -1;
D
Daniel P. Berrange 已提交
1056 1057
}

1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
static int qemudNetworkEnable(struct qemud_server *server) {
    struct qemud_socket *sock;

    sock = server->sockets;
    while (sock) {
        if ((sock->watch = virEventAddHandleImpl(sock->fd,
                                                 VIR_EVENT_HANDLE_READABLE |
                                                 VIR_EVENT_HANDLE_ERROR |
                                                 VIR_EVENT_HANDLE_HANGUP,
                                                 qemudDispatchServerEvent,
                                                 server, NULL)) < 0) {
            VIR_ERROR0(_("Failed to add server event callback"));
            return -1;
        }

        sock = sock->next;
    }
    return 0;
}
1077

1078

1079 1080 1081
static gnutls_session_t
remoteInitializeTLSSession (void)
{
1082 1083
    gnutls_session_t session;
    int err;
1084

1085 1086
    err = gnutls_init (&session, GNUTLS_SERVER);
    if (err != 0) goto failed;
1087

1088 1089 1090 1091 1092
    /* avoid calling all the priority functions, since the defaults
     * are adequate.
     */
    err = gnutls_set_default_priority (session);
    if (err != 0) goto failed;
1093

1094 1095
    err = gnutls_credentials_set (session, GNUTLS_CRD_CERTIFICATE, x509_cred);
    if (err != 0) goto failed;
1096

1097 1098 1099
    /* request client certificate if any.
     */
    gnutls_certificate_server_set_request (session, GNUTLS_CERT_REQUEST);
1100

1101
    gnutls_dh_set_prime_bits (session, DH_BITS);
1102

1103
    return session;
1104

1105 1106 1107 1108
failed:
    VIR_ERROR(_("remoteInitializeTLSSession: %s"),
              gnutls_strerror (err));
    return NULL;
1109 1110 1111 1112
}

/* Check DN is on tls_allowed_dn_list. */
static int
1113
remoteCheckDN (const char *dname)
1114
{
1115
    char **wildcards;
1116 1117 1118 1119 1120 1121 1122

    /* If the list is not set, allow any DN. */
    wildcards = tls_allowed_dn_list;
    if (!wildcards)
        return 1;

    while (*wildcards) {
1123
        if (fnmatch (*wildcards, dname, 0) == 0)
1124 1125 1126 1127 1128
            return 1;
        wildcards++;
    }

    /* Print the client's DN. */
1129
    DEBUG(_("remoteCheckDN: failed: client DN is %s"), dname);
1130 1131 1132 1133 1134

    return 0; // Not found.
}

static int
1135
remoteCheckCertificate(struct qemud_client *client)
1136 1137 1138 1139 1140 1141
{
    int ret;
    unsigned int status;
    const gnutls_datum_t *certs;
    unsigned int nCerts, i;
    time_t now;
1142 1143 1144 1145
    char name[256];
    size_t namesize = sizeof name;

    memset(name, 0, namesize);
1146

1147 1148
    if ((ret = gnutls_certificate_verify_peers2 (client->tlssession, &status)) < 0){
        VIR_ERROR(_("Failed to verify certificate peers: %s"),
1149
                  gnutls_strerror (ret));
1150
        goto authdeny;
1151 1152 1153 1154
    }

    if (status != 0) {
        if (status & GNUTLS_CERT_INVALID)
1155
            VIR_ERROR0(_("The client certificate is not trusted."));
1156 1157

        if (status & GNUTLS_CERT_SIGNER_NOT_FOUND)
1158
            VIR_ERROR0(_("The client certificate has unknown issuer."));
1159 1160

        if (status & GNUTLS_CERT_REVOKED)
1161
            VIR_ERROR0(_("The client certificate has been revoked."));
1162

1163
#ifndef GNUTLS_1_0_COMPAT
1164
        if (status & GNUTLS_CERT_INSECURE_ALGORITHM)
1165
            VIR_ERROR0(_("The client certificate uses an insecure algorithm."));
1166
#endif
1167

1168
        goto authdeny;
1169 1170
    }

1171 1172 1173
    if (gnutls_certificate_type_get(client->tlssession) != GNUTLS_CRT_X509) {
        VIR_ERROR0(_("Only x509 certificates are supported"));
        goto authdeny;
1174 1175
    }

1176 1177 1178
    if (!(certs = gnutls_certificate_get_peers(client->tlssession, &nCerts))) {
        VIR_ERROR0(_("The certificate has no peers"));
        goto authdeny;
1179 1180 1181 1182 1183 1184 1185 1186
    }

    now = time (NULL);

    for (i = 0; i < nCerts; i++) {
        gnutls_x509_crt_t cert;

        if (gnutls_x509_crt_init (&cert) < 0) {
1187 1188
            VIR_ERROR0(_("Unable to initialize certificate"));
            goto authfail;
1189 1190 1191
        }

        if (gnutls_x509_crt_import(cert, &certs[i], GNUTLS_X509_FMT_DER) < 0) {
1192
            VIR_ERROR0(_("Unable to load certificate"));
1193
            gnutls_x509_crt_deinit (cert);
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
            goto authfail;
        }

        if (i == 0) {
            ret = gnutls_x509_crt_get_dn (cert, name, &namesize);
            if (ret != 0) {
                VIR_ERROR(_("Failed to get certificate distinguished name: %s"),
                          gnutls_strerror(ret));
                gnutls_x509_crt_deinit (cert);
                goto authfail;
            }

            if (!remoteCheckDN (name)) {
                /* This is the most common error: make it informative. */
                VIR_ERROR0(_("Client's Distinguished Name is not on the list "
                             "of allowed clients (tls_allowed_dn_list).  Use "
                             "'certtool -i --infile clientcert.pem' to view the"
                             "Distinguished Name field in the client certificate,"
                             "or run this daemon with --verbose option."));
                gnutls_x509_crt_deinit (cert);
                goto authdeny;
            }
1216
        }
1217

1218
        if (gnutls_x509_crt_get_expiration_time (cert) < now) {
1219
            VIR_ERROR0(_("The client certificate has expired"));
1220
            gnutls_x509_crt_deinit (cert);
1221
            goto authdeny;
1222
        }
1223

1224
        if (gnutls_x509_crt_get_activation_time (cert) > now) {
1225
            VIR_ERROR0(_("The client certificate is not yet active"));
1226
            gnutls_x509_crt_deinit (cert);
1227
            goto authdeny;
1228 1229 1230
        }
    }

M
Matthias Bolte 已提交
1231
    PROBE(CLIENT_TLS_ALLOW, "fd=%d, name=%s", client->fd, (char *)name);
1232
    return 0;
1233 1234

authdeny:
M
Matthias Bolte 已提交
1235
    PROBE(CLIENT_TLS_DENY, "fd=%d, name=%s", client->fd, (char *)name);
1236 1237 1238
    return -1;

authfail:
1239
    PROBE(CLIENT_TLS_FAIL, "fd=%d", client->fd);
1240
    return -1;
1241 1242 1243 1244 1245 1246
}

/* Check the client's access. */
static int
remoteCheckAccess (struct qemud_client *client)
{
1247 1248
    struct qemud_client_message *confirm;

1249
    /* Verify client certificate. */
1250
    if (remoteCheckCertificate (client) == -1) {
1251 1252
        VIR_ERROR0(_("remoteCheckCertificate: "
                     "failed to verify client's certificate"));
1253
        if (!tls_no_verify_certificate) return -1;
1254 1255
        else VIR_INFO0(_("remoteCheckCertificate: tls_no_verify_certificate "
                          "is set so the bad certificate is ignored"));
1256 1257
    }

1258 1259 1260 1261 1262 1263 1264 1265 1266
    if (client->tx) {
        VIR_INFO("%s",
                 _("client had unexpected data pending tx after access check"));
        return -1;
    }

    if (VIR_ALLOC(confirm) < 0)
        return -1;

1267 1268 1269 1270
    /* Checks have succeeded.  Write a '\1' byte back to the client to
     * indicate this (otherwise the socket is abruptly closed).
     * (NB. The '\1' byte is sent in an encrypted record).
     */
1271 1272 1273 1274 1275 1276
    confirm->async = 1;
    confirm->bufferLength = 1;
    confirm->bufferOffset = 0;
    confirm->buffer[0] = '\1';

    client->tx = confirm;
1277 1278
    return 0;
}
D
Daniel P. Berrange 已提交
1279

1280 1281
#if HAVE_POLKIT
int qemudGetSocketIdentity(int fd, uid_t *uid, pid_t *pid) {
1282
# ifdef SO_PEERCRED
1283 1284 1285 1286
    struct ucred cr;
    unsigned int cr_len = sizeof (cr);

    if (getsockopt (fd, SOL_SOCKET, SO_PEERCRED, &cr, &cr_len) < 0) {
1287
        char ebuf[1024];
1288
        VIR_ERROR(_("Failed to verify client credentials: %s"),
1289
                  virStrerror(errno, ebuf, sizeof ebuf));
1290 1291 1292 1293 1294
        return -1;
    }

    *pid = cr.pid;
    *uid = cr.uid;
1295
# else
1296
    /* XXX Many more OS support UNIX socket credentials we could port to. See dbus ....*/
1297 1298
#  error "UNIX socket credentials not supported/implemented on this platform yet..."
# endif
1299 1300 1301 1302
    return 0;
}
#endif

1303

D
Daniel P. Berrange 已提交
1304 1305
static int qemudDispatchServer(struct qemud_server *server, struct qemud_socket *sock) {
    int fd;
1306 1307
    virSocketAddr addr;
    char *addrstr = NULL;
1308
    struct qemud_client *client = NULL;
1309
    int no_slow_start = 1;
1310
    int i;
D
Daniel P. Berrange 已提交
1311

1312 1313
    addr.len = sizeof(addr.data.stor);
    if ((fd = accept(sock->fd, &addr.data.sa, &addr.len)) < 0) {
1314
        char ebuf[1024];
D
Daniel P. Berrange 已提交
1315 1316
        if (errno == EAGAIN)
            return 0;
1317 1318
        VIR_ERROR(_("Failed to accept connection: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
1319 1320
        return -1;
    }
1321 1322 1323 1324
    if (!(addrstr = virSocketFormatAddrFull(&addr, true, ";"))) {
        VIR_ERROR0(_("Failed to format addresss: out of memory"));
        goto error;
    }
D
Daniel P. Berrange 已提交
1325

1326 1327
    PROBE(CLIENT_CONNECT, "fd=%d, readonly=%d localAddr=%s remoteAddr=%s",
          fd, sock->readonly, sock->addrstr, addrstr);
1328

1329
    if (server->nclients >= max_clients) {
1330 1331
        VIR_ERROR(_("Too many active clients (%d), dropping connection from %s"),
                  max_clients, addrstr);
1332
        goto error;
1333 1334
    }

1335 1336
    if (VIR_RESIZE_N(server->clients, server->nclients_max,
                     server->nclients, 1) < 0) {
1337
        VIR_ERROR0(_("Out of memory allocating clients"));
1338
        goto error;
1339 1340
    }

J
John Levon 已提交
1341 1342 1343 1344 1345 1346 1347 1348 1349
#ifdef __sun
    {
        ucred_t *ucred = NULL;
        const priv_set_t *privs;

        if (getpeerucred (fd, &ucred) == -1 ||
            (privs = ucred_getprivset (ucred, PRIV_EFFECTIVE)) == NULL) {
            if (ucred != NULL)
                ucred_free (ucred);
1350
            goto error;
J
John Levon 已提交
1351 1352 1353 1354
        }

        if (!priv_ismember (privs, PRIV_VIRT_MANAGE)) {
            ucred_free (ucred);
1355
            goto error;
J
John Levon 已提交
1356 1357 1358 1359 1360 1361
        }

        ucred_free (ucred);
    }
#endif /* __sun */

1362 1363 1364 1365
    /* Disable Nagle.  Unix sockets will ignore this. */
    setsockopt (fd, IPPROTO_TCP, TCP_NODELAY, (void *)&no_slow_start,
                sizeof no_slow_start);

1366 1367
    if (virSetCloseExec(fd) < 0 ||
        virSetNonBlock(fd) < 0) {
1368
        goto error;
D
Daniel P. Berrange 已提交
1369 1370
    }

1371
    if (VIR_ALLOC(client) < 0)
1372
        goto error;
1373
    if (virMutexInit(&client->lock) < 0) {
1374
        VIR_ERROR0(_("cannot initialize mutex"));
1375
        goto error;
1376
    }
1377

1378
    client->magic = QEMUD_CLIENT_MAGIC;
D
Daniel P. Berrange 已提交
1379 1380
    client->fd = fd;
    client->readonly = sock->readonly;
1381
    client->type = sock->type;
1382
    client->auth = sock->auth;
1383 1384 1385
    client->addr = addr;
    client->addrstr = addrstr;
    addrstr = NULL;
1386

1387 1388 1389 1390
    for (i = 0 ; i < VIR_DOMAIN_EVENT_ID_LAST ; i++) {
        client->domainEventCallbackID[i] = -1;
    }

1391 1392
    /* Prepare one for packet receive */
    if (VIR_ALLOC(client->rx) < 0)
1393
        goto error;
1394 1395 1396
    client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;


1397 1398 1399 1400 1401 1402 1403 1404 1405 1406
#if HAVE_POLKIT
    /* Only do policy checks for non-root - allow root user
       through with no checks, as a fail-safe - root can easily
       change policykit policy anyway, so its pointless trying
       to restrict root */
    if (client->auth == REMOTE_AUTH_POLKIT) {
        uid_t uid;
        pid_t pid;

        if (qemudGetSocketIdentity(client->fd, &uid, &pid) < 0)
1407
            goto error;
1408

1409
        /* Client is running as root, so disable auth */
1410
        if (uid == 0) {
1411
            VIR_INFO(_("Turn off polkit auth for privileged client pid %d from %s"),
E
Eric Blake 已提交
1412
                     pid, client->addrstr);
1413 1414 1415 1416 1417
            client->auth = REMOTE_AUTH_NONE;
        }
    }
#endif

1418
    if (client->type != QEMUD_SOCK_TYPE_TLS) {
1419
        /* Plain socket, so prepare to read first message */
1420
        if (qemudRegisterClientEvent (server, client) < 0)
1421
            goto error;
1422 1423 1424
    } else {
        int ret;

1425 1426
        client->tlssession = remoteInitializeTLSSession ();
        if (client->tlssession == NULL)
1427
            goto error;
1428

1429
        gnutls_transport_set_ptr (client->tlssession,
1430 1431 1432
                                  (gnutls_transport_ptr_t) (long) fd);

        /* Begin the TLS handshake. */
1433
        ret = gnutls_handshake (client->tlssession);
1434
        if (ret == 0) {
1435 1436
            client->handshake = 0;

1437 1438
            /* Unlikely, but ...  Next step is to check the certificate. */
            if (remoteCheckAccess (client) == -1)
1439
                goto error;
D
Daniel P. Berrange 已提交
1440

1441
            /* Handshake & cert check OK,  so prepare to read first message */
1442
            if (qemudRegisterClientEvent(server, client) < 0)
1443
                goto error;
1444
        } else if (ret == GNUTLS_E_INTERRUPTED || ret == GNUTLS_E_AGAIN) {
1445 1446
            /* Most likely, need to do more handshake data */
            client->handshake = 1;
D
Daniel P. Berrange 已提交
1447

1448
            if (qemudRegisterClientEvent (server, client) < 0)
1449
                goto error;
1450
        } else {
1451
            PROBE(CLIENT_TLS_FAIL, "fd=%d", client->fd);
1452
            VIR_ERROR(_("TLS handshake failed for client %s: %s"),
E
Eric Blake 已提交
1453
                      client->addrstr, gnutls_strerror (ret));
1454
            goto error;
1455 1456
        }
    }
D
Daniel P. Berrange 已提交
1457

1458
    server->clients[server->nclients++] = client;
D
Daniel P. Berrange 已提交
1459

1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472
    if (server->nclients > server->nactiveworkers &&
        server->nactiveworkers < server->nworkers) {
        for (i = 0 ; i < server->nworkers ; i++) {
            if (!server->workers[i].hasThread) {
                if (qemudStartWorker(server, &server->workers[i]) < 0)
                    return -1;
                server->nactiveworkers++;
                break;
            }
        }
    }


D
Daniel P. Berrange 已提交
1473
    return 0;
1474

1475 1476 1477
error:
    if (client) {
        if (client->tlssession) gnutls_deinit (client->tlssession);
1478 1479
        if (client) {
            VIR_FREE(client->addrstr);
1480
            VIR_FREE(client->rx);
1481
        }
1482 1483
        VIR_FREE(client);
    }
1484
    VIR_FREE(addrstr);
1485
    VIR_FORCE_CLOSE(fd);
1486
    PROBE(CLIENT_DISCONNECT, "fd=%d", fd);
1487
    return -1;
D
Daniel P. Berrange 已提交
1488 1489 1490
}


1491 1492 1493 1494 1495 1496 1497
/*
 * You must hold lock for at least the client
 * We don't free stuff here, merely disconnect the client's
 * network socket & resources.
 * We keep the libvirt connection open until any async
 * jobs have finished, then clean it up elsehwere
 */
1498
void qemudDispatchClientFailure(struct qemud_client *client) {
1499 1500 1501 1502
    if (client->watch != -1) {
        virEventRemoveHandleImpl(client->watch);
        client->watch = -1;
    }
D
Daniel P. Berrange 已提交
1503

1504
    /* Deregister event delivery callback */
1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515
    if (client->conn) {
        int i;

        for (i = 0 ; i < VIR_DOMAIN_EVENT_ID_LAST ; i++) {
            if (client->domainEventCallbackID[i] != -1) {
                DEBUG("Deregistering to relay remote events %d", i);
                virConnectDomainEventDeregisterAny(client->conn,
                                                   client->domainEventCallbackID[i]);
            }
            client->domainEventCallbackID[i] = -1;
        }
1516 1517
    }

1518
#if HAVE_SASL
1519 1520 1521 1522
    if (client->saslconn) {
        sasl_dispose(&client->saslconn);
        client->saslconn = NULL;
    }
1523
    VIR_FREE(client->saslUsername);
1524
#endif
1525 1526 1527 1528 1529
    if (client->tlssession) {
        gnutls_deinit (client->tlssession);
        client->tlssession = NULL;
    }
    if (client->fd != -1) {
1530
        PROBE(CLIENT_DISCONNECT, "fd=%d", client->fd);
1531
        VIR_FORCE_CLOSE(client->fd);
1532
    }
1533
    VIR_FREE(client->addrstr);
1534 1535 1536 1537 1538 1539 1540 1541
}


/* Caller must hold server lock */
static struct qemud_client *qemudPendingJob(struct qemud_server *server)
{
    int i;
    for (i = 0 ; i < server->nclients ; i++) {
1542
        virMutexLock(&server->clients[i]->lock);
1543
        if (server->clients[i]->dx) {
1544 1545 1546
            /* Delibrately don't unlock client - caller wants the lock */
            return server->clients[i];
        }
1547
        virMutexUnlock(&server->clients[i]->lock);
1548 1549
    }
    return NULL;
D
Daniel P. Berrange 已提交
1550 1551
}

1552 1553
static void *qemudWorker(void *data)
{
1554 1555
    struct qemud_worker *worker = data;
    struct qemud_server *server = worker->server;
1556 1557

    while (1) {
1558
        struct qemud_client *client = NULL;
1559
        struct qemud_client_message *msg;
1560

1561
        virMutexLock(&server->lock);
1562 1563 1564
        while ((client = qemudPendingJob(server)) == NULL) {
            if (worker->quitRequest ||
                virCondWait(&server->job, &server->lock) < 0) {
1565 1566 1567 1568
                virMutexUnlock(&server->lock);
                return NULL;
            }
        }
1569
        if (worker->quitRequest) {
1570
            virMutexUnlock(&client->lock);
1571 1572 1573 1574
            virMutexUnlock(&server->lock);
            return NULL;
        }
        worker->processingCall = 1;
1575
        virMutexUnlock(&server->lock);
1576 1577 1578 1579

        /* We own a locked client now... */
        client->refs++;

1580
        /* Remove our message from dispatch queue while we use it */
1581
        msg = qemudClientMessageQueueServe(&client->dx);
1582 1583 1584

        /* This function drops the lock during dispatch,
         * and re-acquires it before returning */
1585
        if (remoteDispatchClientRequest (server, client, msg) < 0) {
1586
            VIR_FREE(msg);
1587 1588 1589 1590 1591
            qemudDispatchClientFailure(client);
            client->refs--;
            virMutexUnlock(&client->lock);
            continue;
        }
1592 1593

        client->refs--;
1594
        virMutexUnlock(&client->lock);
1595 1596 1597 1598

        virMutexLock(&server->lock);
        worker->processingCall = 0;
        virMutexUnlock(&server->lock);
1599 1600
    }
}
D
Daniel P. Berrange 已提交
1601

E
Eric Blake 已提交
1602 1603 1604 1605
static int
qemudStartWorker(struct qemud_server *server,
                 struct qemud_worker *worker)
{
1606
    pthread_attr_t attr;
E
Eric Blake 已提交
1607 1608 1609 1610
    int ret = -1;

    if (pthread_attr_init(&attr) != 0)
        return -1;
1611 1612 1613 1614
    /* We want to join workers, so don't detach them */
    /*pthread_attr_setdetachstate(&attr, 1);*/

    if (worker->hasThread)
E
Eric Blake 已提交
1615
        goto cleanup;
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627

    worker->server = server;
    worker->hasThread = 1;
    worker->quitRequest = 0;
    worker->processingCall = 0;

    if (pthread_create(&worker->thread,
                       &attr,
                       qemudWorker,
                       worker) != 0) {
        worker->hasThread = 0;
        worker->server = NULL;
E
Eric Blake 已提交
1628
        goto cleanup;
1629 1630
    }

E
Eric Blake 已提交
1631 1632 1633 1634
    ret = 0;
cleanup:
    pthread_attr_destroy(&attr);
    return ret;
1635 1636
}

D
Daniel P. Berrange 已提交
1637

1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
/*
 * Read data into buffer using wire decoding (plain or TLS)
 *
 * Returns:
 *   -1 on error or EOF
 *    0 on EAGAIN
 *    n number of bytes
 */
static ssize_t qemudClientReadBuf(struct qemud_client *client,
                                  char *data, ssize_t len) {
    ssize_t ret;

    if (len < 0) {
1651 1652
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1653 1654 1655
        qemudDispatchClientFailure(client);
        return -1;
    }
1656 1657 1658

    /*qemudDebug ("qemudClientRead: len = %d", len);*/

1659
    if (!client->tlssession) {
1660
        char ebuf[1024];
1661 1662 1663 1664 1665 1666
        ret = read (client->fd, data, len);
        if (ret == -1 && (errno == EAGAIN ||
                          errno == EINTR))
            return 0;
        if (ret <= 0) {
            if (ret != 0)
1667 1668
                VIR_ERROR(_("read: %s"),
                          virStrerror (errno, ebuf, sizeof ebuf));
1669
            qemudDispatchClientFailure(client);
1670 1671 1672
            return -1;
        }
    } else {
1673
        ret = gnutls_record_recv (client->tlssession, data, len);
1674 1675 1676 1677 1678 1679 1680 1681 1682

        if (ret < 0 && (ret == GNUTLS_E_AGAIN ||
                        ret == GNUTLS_E_INTERRUPTED))
            return 0;
        if (ret <= 0) {
            if (ret != 0)
                VIR_ERROR(_("gnutls_record_recv: %s"),
                          gnutls_strerror (ret));
            qemudDispatchClientFailure(client);
1683 1684
            return -1;
        }
D
Daniel P. Berrange 已提交
1685
    }
1686

1687 1688 1689
    return ret;
}

1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
/*
 * Read data into buffer without decoding
 *
 * Returns:
 *   -1 on error or EOF
 *    0 on EAGAIN
 *    n number of bytes
 */
static ssize_t qemudClientReadPlain(struct qemud_client *client) {
    ssize_t ret;
    ret = qemudClientReadBuf(client,
                             client->rx->buffer + client->rx->bufferOffset,
                             client->rx->bufferLength - client->rx->bufferOffset);
    if (ret <= 0)
        return ret; /* -1 error, 0 eagain */

    client->rx->bufferOffset += ret;
    return ret;
D
Daniel P. Berrange 已提交
1708 1709
}

1710
#if HAVE_SASL
1711 1712 1713 1714 1715 1716 1717 1718 1719 1720
/*
 * Read data into buffer decoding with SASL
 *
 * Returns:
 *   -1 on error or EOF
 *    0 on EAGAIN
 *    n number of bytes
 */
static ssize_t qemudClientReadSASL(struct qemud_client *client) {
    ssize_t got, want;
1721 1722 1723 1724 1725 1726 1727 1728 1729 1730

    /* We're doing a SSF data read, so now its times to ensure
     * future writes are under SSF too.
     *
     * cf remoteSASLCheckSSF in remote.c
     */
    client->saslSSF |= QEMUD_SASL_SSF_WRITE;

    /* Need to read some more data off the wire */
    if (client->saslDecoded == NULL) {
1731
        int ret;
1732
        char encoded[8192];
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
        ssize_t encodedLen = sizeof(encoded);
        encodedLen = qemudClientReadBuf(client, encoded, encodedLen);

        if (encodedLen <= 0)
            return encodedLen;

        ret = sasl_decode(client->saslconn, encoded, encodedLen,
                          &client->saslDecoded, &client->saslDecodedLength);
        if (ret != SASL_OK) {
            VIR_ERROR(_("failed to decode SASL data %s"),
                      sasl_errstring(ret, NULL, NULL));
            qemudDispatchClientFailure(client);
1745
            return -1;
1746
        }
1747 1748 1749 1750 1751 1752

        client->saslDecodedOffset = 0;
    }

    /* Some buffered decoded data to return now */
    got = client->saslDecodedLength - client->saslDecodedOffset;
1753
    want = client->rx->bufferLength - client->rx->bufferOffset;
1754 1755 1756 1757

    if (want > got)
        want = got;

1758
    memcpy(client->rx->buffer + client->rx->bufferOffset,
1759 1760
           client->saslDecoded + client->saslDecodedOffset, want);
    client->saslDecodedOffset += want;
1761
    client->rx->bufferOffset += want;
1762 1763 1764 1765 1766 1767

    if (client->saslDecodedOffset == client->saslDecodedLength) {
        client->saslDecoded = NULL;
        client->saslDecodedOffset = client->saslDecodedLength = 0;
    }

1768
    return want;
1769 1770 1771
}
#endif

1772 1773 1774 1775 1776
/*
 * Read as much data off wire as possible till we fill our
 * buffer, or would block on I/O
 */
static ssize_t qemudClientRead(struct qemud_client *client) {
1777 1778
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_READ)
1779
        return qemudClientReadSASL(client);
1780 1781
    else
#endif
1782
        return qemudClientReadPlain(client);
1783 1784 1785
}


1786 1787 1788 1789 1790
/*
 * Read data until we get a complete message to process
 */
static void qemudDispatchClientRead(struct qemud_server *server,
                                    struct qemud_client *client) {
1791
    /*qemudDebug ("qemudDispatchClientRead: mode = %d", client->mode);*/
D
Daniel P. Berrange 已提交
1792

1793 1794 1795
readmore:
    if (qemudClientRead(client) < 0)
        return; /* Error */
1796

1797 1798
    if (client->rx->bufferOffset < client->rx->bufferLength)
        return; /* Still not read enough */
1799

1800 1801 1802 1803
    /* Either done with length word header */
    if (client->rx->bufferLength == REMOTE_MESSAGE_HEADER_XDR_LEN) {
        unsigned int len;
        XDR x;
1804

1805
        xdrmem_create(&x, client->rx->buffer, client->rx->bufferLength, XDR_DECODE);
1806

1807 1808
        if (!xdr_u_int(&x, &len)) {
            xdr_destroy (&x);
1809
            DEBUG0("Failed to decode packet length");
1810
            qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1811 1812
            return;
        }
1813
        xdr_destroy (&x);
1814

1815 1816 1817
        if (len < REMOTE_MESSAGE_HEADER_XDR_LEN) {
            DEBUG("Packet length %u too small", len);
            qemudDispatchClientFailure(client);
1818
            return;
D
Daniel P. Berrange 已提交
1819
        }
1820

1821 1822 1823 1824 1825 1826
        /* Length includes the size of the length word itself */
        len -= REMOTE_MESSAGE_HEADER_XDR_LEN;

        if (len > REMOTE_MESSAGE_MAX) {
            DEBUG("Packet length %u too large", len);
            qemudDispatchClientFailure(client);
1827 1828 1829
            return;
        }

1830 1831
        /* Prepare to read rest of message */
        client->rx->bufferLength += len;
1832

1833
        qemudUpdateClientEvent(client);
D
Daniel P. Berrange 已提交
1834

1835 1836 1837 1838 1839
        /* Try and read payload immediately instead of going back
           into poll() because chances are the data is already
           waiting for us */
        goto readmore;
    } else {
1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
        /* Grab the completed message */
        struct qemud_client_message *msg = qemudClientMessageQueueServe(&client->rx);
        struct qemud_client_filter *filter;

        /* Decode the header so we can use it for routing decisions */
        if (remoteDecodeClientMessageHeader(msg) < 0) {
            VIR_FREE(msg);
            qemudDispatchClientFailure(client);
        }

        /* Check if any filters match this message */
        filter = client->filters;
        while (filter) {
1853 1854 1855
            int ret;
            ret = (filter->query)(client, msg, filter->opaque);
            if (ret == 1) {
1856 1857
                msg = NULL;
                break;
1858 1859 1860 1861
            } else if (ret == -1) {
                VIR_FREE(msg);
                qemudDispatchClientFailure(client);
                return;
1862 1863 1864 1865
            }
            filter = filter->next;
        }

1866
        /* Move completed message to the end of the dispatch queue */
1867 1868
        if (msg)
            qemudClientMessageQueuePush(&client->dx, msg);
1869 1870 1871 1872 1873 1874
        client->nrequests++;

        /* Possibly need to create another receive buffer */
        if ((client->nrequests < max_client_requests &&
             VIR_ALLOC(client->rx) < 0)) {
            qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1875
        } else {
1876 1877 1878
            if (client->rx)
                client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;

1879 1880 1881 1882
            qemudUpdateClientEvent(client);

            /* Tell one of the workers to get on with it... */
            virCondSignal(&server->job);
D
Daniel P. Berrange 已提交
1883
        }
D
Daniel P. Berrange 已提交
1884 1885 1886 1887
    }
}


1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
/*
 * Send a chunk of data using wire encoding (plain or TLS)
 *
 * Returns:
 *   -1 on error
 *    0 on EAGAIN
 *    n number of bytes
 */
static ssize_t qemudClientWriteBuf(struct qemud_client *client,
                                   const char *data, ssize_t len) {
    ssize_t ret;

    if (len < 0) {
1901 1902
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1903 1904 1905 1906
        qemudDispatchClientFailure(client);
        return -1;
    }

1907
    if (!client->tlssession) {
1908
        char ebuf[1024];
1909 1910 1911
        if ((ret = write(client->fd, data, len)) == -1) {
            if (errno == EAGAIN || errno == EINTR)
                return 0;
1912
            VIR_ERROR(_("write: %s"), virStrerror (errno, ebuf, sizeof ebuf));
1913
            qemudDispatchClientFailure(client);
1914 1915 1916
            return -1;
        }
    } else {
1917
        ret = gnutls_record_send (client->tlssession, data, len);
1918 1919 1920 1921 1922 1923 1924
        if (ret < 0) {
            if (ret == GNUTLS_E_INTERRUPTED ||
                ret == GNUTLS_E_AGAIN)
                return 0;

            VIR_ERROR(_("gnutls_record_send: %s"), gnutls_strerror (ret));
            qemudDispatchClientFailure(client);
1925 1926
            return -1;
        }
D
Daniel P. Berrange 已提交
1927
    }
1928 1929
    return ret;
}
1930

1931

1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
/*
 * Send client->tx using no encoding
 *
 * Returns:
 *   -1 on error or EOF
 *    0 on EAGAIN
 *    n number of bytes
 */
static int qemudClientWritePlain(struct qemud_client *client) {
    int ret = qemudClientWriteBuf(client,
                                  client->tx->buffer + client->tx->bufferOffset,
                                  client->tx->bufferLength - client->tx->bufferOffset);
    if (ret <= 0)
        return ret; /* -1 error, 0 = egain */
    client->tx->bufferOffset += ret;
    return ret;
D
Daniel P. Berrange 已提交
1948 1949 1950
}


1951
#if HAVE_SASL
1952 1953 1954 1955 1956 1957 1958 1959 1960
/*
 * Send client->tx using SASL encoding
 *
 * Returns:
 *   -1 on error
 *    0 on EAGAIN
 *    n number of bytes
 */
static int qemudClientWriteSASL(struct qemud_client *client) {
1961 1962 1963 1964
    int ret;

    /* Not got any pending encoded data, so we need to encode raw stuff */
    if (client->saslEncoded == NULL) {
1965 1966 1967
        ret = sasl_encode(client->saslconn,
                          client->tx->buffer + client->tx->bufferOffset,
                          client->tx->bufferLength - client->tx->bufferOffset,
1968 1969 1970
                          &client->saslEncoded,
                          &client->saslEncodedLength);

1971 1972 1973 1974 1975 1976 1977
        if (ret != SASL_OK) {
            VIR_ERROR(_("failed to encode SASL data %s"),
                      sasl_errstring(ret, NULL, NULL));
            qemudDispatchClientFailure(client);
            return -1;
        }

1978 1979 1980 1981
        client->saslEncodedOffset = 0;
    }

    /* Send some of the encoded stuff out on the wire */
1982
    ret = qemudClientWriteBuf(client,
1983 1984 1985
                              client->saslEncoded + client->saslEncodedOffset,
                              client->saslEncodedLength - client->saslEncodedOffset);

1986 1987
    if (ret <= 0)
        return ret; /* -1 error, 0 == egain */
1988 1989 1990 1991 1992 1993 1994 1995

    /* Note how much we sent */
    client->saslEncodedOffset += ret;

    /* Sent all encoded, so update raw buffer to indicate completion */
    if (client->saslEncodedOffset == client->saslEncodedLength) {
        client->saslEncoded = NULL;
        client->saslEncodedOffset = client->saslEncodedLength = 0;
1996 1997 1998

        /* Mark as complete, so caller detects completion */
        client->tx->bufferOffset = client->tx->bufferLength;
1999 2000
    }

2001
    return ret;
2002 2003 2004
}
#endif

2005 2006 2007 2008 2009 2010 2011 2012 2013
/*
 * Send as much data in the client->tx as possible
 *
 * Returns:
 *   -1 on error or EOF
 *    0 on EAGAIN
 *    n number of bytes
 */
static ssize_t qemudClientWrite(struct qemud_client *client) {
2014 2015
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_WRITE)
2016
        return qemudClientWriteSASL(client);
2017 2018
    else
#endif
2019
        return qemudClientWritePlain(client);
2020 2021 2022
}


2023 2024 2025 2026
void
qemudClientMessageRelease(struct qemud_client *client,
                          struct qemud_client_message *msg)
{
2027 2028 2029
    if (msg->streamTX) {
        remoteStreamMessageFinished(client, msg);
    } else if (!msg->async)
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047
        client->nrequests--;

    /* See if the recv queue is currently throttled */
    if (!client->rx &&
        client->nrequests < max_client_requests) {
        /* Reset message record for next RX attempt */
        memset(msg, 0, sizeof(*msg));
        client->rx = msg;
        /* Get ready to receive next message */
        client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;
    } else {
        VIR_FREE(msg);
    }

    qemudUpdateClientEvent(client);
}


2048 2049 2050 2051 2052
/*
 * Process all queued client->tx messages until
 * we would block on I/O
 */
static void
2053
qemudDispatchClientWrite(struct qemud_client *client) {
2054 2055
    while (client->tx) {
        ssize_t ret;
2056

2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
        ret = qemudClientWrite(client);
        if (ret < 0) {
            qemudDispatchClientFailure(client);
            return;
        }
        if (ret == 0)
            return; /* Would block on write EAGAIN */

        if (client->tx->bufferOffset == client->tx->bufferLength) {
            struct qemud_client_message *reply;

            /* Get finished reply from head of tx queue */
            reply = qemudClientMessageQueueServe(&client->tx);

2071
            qemudClientMessageRelease(client, reply);
2072

2073 2074
            if (client->closing)
                qemudDispatchClientFailure(client);
2075
         }
2076
    }
2077
}
2078

2079
static void
2080
qemudDispatchClientHandshake(struct qemud_client *client) {
2081 2082 2083 2084
    int ret;
    /* Continue the handshake. */
    ret = gnutls_handshake (client->tlssession);
    if (ret == 0) {
2085 2086
        client->handshake = 0;

2087 2088 2089
        /* Finished.  Next step is to check the certificate. */
        if (remoteCheckAccess (client) == -1)
            qemudDispatchClientFailure(client);
2090 2091
        else
            qemudUpdateClientEvent(client);
2092 2093 2094 2095 2096
    } else if (ret == GNUTLS_E_AGAIN ||
               ret == GNUTLS_E_INTERRUPTED) {
        /* Carry on waiting for more handshake. Update
           the events just in case handshake data flow
           direction has changed */
2097
        qemudUpdateClientEvent (client);
2098
    } else {
2099
        PROBE(CLIENT_TLS_FAIL, "fd=%d", client->fd);
2100 2101 2102 2103
        /* Fatal error in handshake */
        VIR_ERROR(_("TLS handshake failed: %s"),
                  gnutls_strerror (ret));
        qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
2104 2105 2106
    }
}

2107
static void
2108
qemudDispatchClientEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
2109
    struct qemud_server *server = (struct qemud_server *)opaque;
2110 2111
    struct qemud_client *client = NULL;
    int i;
2112

2113
    virMutexLock(&server->lock);
2114

2115
    for (i = 0 ; i < server->nclients ; i++) {
2116
        virMutexLock(&server->clients[i]->lock);
2117 2118
        if (server->clients[i]->watch == watch) {
            client = server->clients[i];
D
Daniel P. Berrange 已提交
2119
            break;
2120
        }
2121
        virMutexUnlock(&server->clients[i]->lock);
D
Daniel P. Berrange 已提交
2122
    }
2123

2124 2125
    virMutexUnlock(&server->lock);

2126
    if (!client) {
D
Daniel P. Berrange 已提交
2127
        return;
2128 2129
    }

2130 2131
    if (client->fd != fd) {
        virMutexUnlock(&client->lock);
2132
        return;
2133 2134 2135 2136 2137
    }

    if (events & (VIR_EVENT_HANDLE_WRITABLE |
                  VIR_EVENT_HANDLE_READABLE)) {
        if (client->handshake) {
2138
            qemudDispatchClientHandshake(client);
2139 2140
        } else {
            if (events & VIR_EVENT_HANDLE_WRITABLE)
2141
                qemudDispatchClientWrite(client);
2142 2143 2144 2145 2146 2147 2148 2149 2150 2151
            if (events & VIR_EVENT_HANDLE_READABLE)
                qemudDispatchClientRead(server, client);
        }
    }

    /* NB, will get HANGUP + READABLE at same time upon
     * disconnect */
    if (events & (VIR_EVENT_HANDLE_ERROR |
                  VIR_EVENT_HANDLE_HANGUP))
        qemudDispatchClientFailure(client);
2152

2153
    virMutexUnlock(&client->lock);
D
Daniel P. Berrange 已提交
2154 2155
}

2156 2157 2158 2159 2160 2161

/*
 * @client: a locked client object
 */
static int
qemudCalculateHandleMode(struct qemud_client *client) {
2162 2163 2164
    int mode = 0;

    if (client->handshake) {
2165
        if (gnutls_record_get_direction (client->tlssession) == 0)
2166
            mode |= VIR_EVENT_HANDLE_READABLE;
2167
        else
2168 2169 2170 2171 2172 2173
            mode |= VIR_EVENT_HANDLE_WRITABLE;
    } else {
        /* If there is a message on the rx queue then
         * we're wanting more input */
        if (client->rx)
            mode |= VIR_EVENT_HANDLE_READABLE;
2174

2175 2176 2177 2178
        /* If there are one or more messages to send back to client,
           then monitor for writability on socket */
        if (client->tx)
            mode |= VIR_EVENT_HANDLE_WRITABLE;
2179 2180
    }

2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198
    return mode;
}

/*
 * @server: a locked or unlocked server object
 * @client: a locked client object
 */
int qemudRegisterClientEvent(struct qemud_server *server,
                             struct qemud_client *client) {
    int mode;

    mode = qemudCalculateHandleMode(client);

    if ((client->watch = virEventAddHandleImpl(client->fd,
                                               mode,
                                               qemudDispatchClientEvent,
                                               server, NULL)) < 0)
        return -1;
D
Daniel P. Berrange 已提交
2199 2200 2201 2202

    return 0;
}

2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214
/*
 * @client: a locked client object
 */
void qemudUpdateClientEvent(struct qemud_client *client) {
    int mode;

    mode = qemudCalculateHandleMode(client);

    virEventUpdateHandleImpl(client->watch, mode);
}


2215
static void
2216
qemudDispatchServerEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
2217
    struct qemud_server *server = (struct qemud_server *)opaque;
2218 2219
    struct qemud_socket *sock;

2220
    virMutexLock(&server->lock);
2221 2222

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2223

D
Daniel P. Berrange 已提交
2224
    while (sock) {
2225
        if (sock->watch == watch)
D
Daniel P. Berrange 已提交
2226 2227 2228
            break;

        sock = sock->next;
D
Daniel P. Berrange 已提交
2229
    }
D
Daniel P. Berrange 已提交
2230

2231
    if (sock && sock->fd == fd && events)
D
Daniel P. Berrange 已提交
2232
        qemudDispatchServer(server, sock);
2233

2234
    virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
2235 2236 2237
}


2238
static int qemudOneLoop(void) {
2239
    sig_atomic_t errors;
D
Daniel P. Berrange 已提交
2240

D
Daniel P. Berrange 已提交
2241
    if (virEventRunOnce() < 0)
D
Daniel P. Berrange 已提交
2242 2243
        return -1;

2244 2245 2246
    /* Check for any signal handling errors and log them. */
    errors = sig_errors;
    if (errors) {
2247
        char ebuf[1024];
2248
        sig_errors -= errors;
2249
        VIR_ERROR(_("Signal handler reported %d errors: last error: %s"),
2250
                  errors, virStrerror (sig_lasterrno, ebuf, sizeof ebuf));
2251 2252 2253
        return -1;
    }

D
Daniel P. Berrange 已提交
2254 2255 2256
    return 0;
}

2257
static void qemudInactiveTimer(int timerid, void *data) {
2258
    struct qemud_server *server = (struct qemud_server *)data;
2259 2260 2261 2262 2263 2264 2265

    if (virStateActive() ||
        server->clients) {
        DEBUG0("Timer expired but still active, not shutting down");
        virEventUpdateTimeoutImpl(timerid, -1);
    } else {
        DEBUG0("Timer expired and inactive, shutting down");
2266
        server->quitEventThread = 1;
2267 2268 2269
    }
}

2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
static void qemudFreeClient(struct qemud_client *client) {
    while (client->rx) {
        struct qemud_client_message *msg
            = qemudClientMessageQueueServe(&client->rx);
        VIR_FREE(msg);
    }
    while (client->dx) {
        struct qemud_client_message *msg
            = qemudClientMessageQueueServe(&client->dx);
        VIR_FREE(msg);
    }
    while (client->tx) {
        struct qemud_client_message *msg
            = qemudClientMessageQueueServe(&client->tx);
        VIR_FREE(msg);
    }

2287 2288 2289
    while (client->streams)
        remoteRemoveClientStream(client, client->streams);

2290 2291 2292
    if (client->conn)
        virConnectClose(client->conn);
    virMutexDestroy(&client->lock);
E
Eric Blake 已提交
2293
    VIR_FREE(client->addrstr);
2294 2295 2296
    VIR_FREE(client);
}

2297 2298
static void *qemudRunLoop(void *opaque) {
    struct qemud_server *server = opaque;
2299
    int timerid = -1;
2300
    int i;
2301
    int timerActive = 0;
2302

2303
    virMutexLock(&server->lock);
2304

2305 2306 2307 2308 2309
    if (timeout > 0 &&
        (timerid = virEventAddTimeoutImpl(-1,
                                          qemudInactiveTimer,
                                          server, NULL)) < 0) {
        VIR_ERROR0(_("Failed to register shutdown timeout"));
2310
        return NULL;
2311 2312
    }

2313 2314 2315 2316
    if (min_workers > max_workers)
        max_workers = min_workers;

    server->nworkers = max_workers;
2317
    if (VIR_ALLOC_N(server->workers, server->nworkers) < 0) {
2318
        VIR_ERROR0(_("Failed to allocate workers"));
2319
        return NULL;
2320 2321
    }

2322 2323 2324 2325
    for (i = 0 ; i < min_workers ; i++) {
        if (qemudStartWorker(server, &server->workers[i]) < 0)
            goto cleanup;
        server->nactiveworkers++;
2326
    }
2327

2328
    for (;!server->quitEventThread;) {
2329 2330 2331 2332
        /* A shutdown timeout is specified, so check
         * if any drivers have active state, if not
         * shutdown after timeout seconds
         */
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
        if (timeout > 0) {
            if (timerActive) {
                if (server->clients) {
                    DEBUG("Deactivating shutdown timer %d", timerid);
                    virEventUpdateTimeoutImpl(timerid, -1);
                    timerActive = 0;
                }
            } else {
                if (!virStateActive() &&
                    !server->clients) {
                    DEBUG("Activating shutdown timer %d", timerid);
                    virEventUpdateTimeoutImpl(timerid, timeout * 1000);
                    timerActive = 1;
                }
            }
2348 2349
        }

2350
        virMutexUnlock(&server->lock);
2351 2352
        if (qemudOneLoop() < 0) {
            virMutexLock(&server->lock);
2353
            DEBUG0("Loop iteration error, exiting");
2354
            break;
2355
        }
2356
        virMutexLock(&server->lock);
2357 2358 2359 2360

    reprocess:
        for (i = 0 ; i < server->nclients ; i++) {
            int inactive;
2361
            virMutexLock(&server->clients[i]->lock);
2362 2363
            inactive = server->clients[i]->fd == -1
                && server->clients[i]->refs == 0;
2364
            virMutexUnlock(&server->clients[i]->lock);
2365
            if (inactive) {
2366
                qemudFreeClient(server->clients[i]);
2367
                server->nclients--;
2368
                if (i < server->nclients)
2369 2370
                    memmove(server->clients + i,
                            server->clients + i + 1,
2371 2372
                            sizeof (*server->clients) * (server->nclients - i));

2373 2374
                VIR_SHRINK_N(server->clients, server->nclients_max,
                             server->nclients_max - server->nclients);
2375
                goto reprocess;
2376 2377
            }
        }
D
Daniel P. Berrange 已提交
2378

2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397
        /* If number of active workers exceeds both the min_workers
         * threshold and the number of clients, then kill some
         * off */
        for (i = 0 ; (i < server->nworkers &&
                      server->nactiveworkers > server->nclients &&
                      server->nactiveworkers > min_workers) ; i++) {

            if (server->workers[i].hasThread &&
                !server->workers[i].processingCall) {
                server->workers[i].quitRequest = 1;

                virCondBroadcast(&server->job);
                virMutexUnlock(&server->lock);
                pthread_join(server->workers[i].thread, NULL);
                virMutexLock(&server->lock);
                server->workers[i].hasThread = 0;
                server->nactiveworkers--;
            }
        }
2398 2399
    }

2400
cleanup:
2401
    for (i = 0 ; i < server->nworkers ; i++) {
2402 2403 2404 2405 2406 2407
        if (!server->workers[i].hasThread)
            continue;

        server->workers[i].quitRequest = 1;
        virCondBroadcast(&server->job);

2408
        virMutexUnlock(&server->lock);
2409
        pthread_join(server->workers[i].thread, NULL);
2410
        virMutexLock(&server->lock);
2411
        server->workers[i].hasThread = 0;
2412
    }
2413
    VIR_FREE(server->workers);
2414 2415 2416 2417
    for (i = 0; i < server->nclients; i++)
        qemudFreeClient(server->clients[i]);
    server->nclients = 0;
    VIR_SHRINK_N(server->clients, server->nclients_max, server->nclients_max);
2418

2419
    virMutexUnlock(&server->lock);
2420
    return NULL;
D
Daniel P. Berrange 已提交
2421 2422
}

2423

E
Eric Blake 已提交
2424 2425 2426
static int
qemudStartEventLoop(struct qemud_server *server)
{
2427
    pthread_attr_t attr;
E
Eric Blake 已提交
2428 2429 2430 2431
    int ret = -1;

    if (pthread_attr_init(&attr) != 0)
        return -1;
2432 2433 2434 2435 2436 2437 2438
    /* We want to join the eventloop, so don't detach it */
    /*pthread_attr_setdetachstate(&attr, 1);*/

    if (pthread_create(&server->eventThread,
                       &attr,
                       qemudRunLoop,
                       server) != 0)
E
Eric Blake 已提交
2439
        goto cleanup;
2440 2441 2442

    server->hasEventThread = 1;

E
Eric Blake 已提交
2443 2444 2445 2446
    ret = 0;
cleanup:
    pthread_attr_destroy(&attr);
    return ret;
2447 2448 2449
}


D
Daniel P. Berrange 已提交
2450
static void qemudCleanup(struct qemud_server *server) {
2451 2452
    struct qemud_socket *sock;

2453 2454
    VIR_FORCE_CLOSE(server->sigread);
    VIR_FORCE_CLOSE(server->sigwrite);
2455 2456

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2457
    while (sock) {
2458
        struct qemud_socket *next = sock->next;
2459 2460
        if (sock->watch)
            virEventRemoveHandleImpl(sock->watch);
2461
        VIR_FORCE_CLOSE(sock->fd);
2462 2463 2464 2465 2466 2467 2468

        /* Unlink unix domain sockets which are not in
         * the abstract namespace */
        if (sock->path &&
            sock->path[0] != '@')
            unlink(sock->path);
        VIR_FREE(sock->path);
2469
        VIR_FREE(sock->addrstr);
2470

2471
        VIR_FREE(sock);
2472
        sock = next;
D
Daniel P. Berrange 已提交
2473
    }
2474
    VIR_FREE(server->logDir);
2475

2476
#ifdef HAVE_SASL
2477 2478 2479
    if (server->saslUsernameWhitelist) {
        char **list = server->saslUsernameWhitelist;
        while (*list) {
2480
            VIR_FREE(*list);
2481 2482
            list++;
        }
2483
        VIR_FREE(server->saslUsernameWhitelist);
2484
    }
2485
#endif
2486

2487 2488 2489 2490 2491
#if HAVE_POLKIT0
        if (server->sysbus)
            dbus_connection_unref(server->sysbus);
#endif

2492
    virStateCleanup();
2493

2494 2495 2496 2497 2498 2499
    if (virCondDestroy(&server->job) < 0) {
        ;
    }
    virMutexDestroy(&server->lock);

    VIR_FREE(server);
D
Daniel P. Berrange 已提交
2500 2501
}

2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
/* Allocate an array of malloc'd strings from the config file, filename
 * (used only in diagnostics), using handle "conf".  Upon error, return -1
 * and free any allocated memory.  Otherwise, save the array in *list_arg
 * and return 0.
 */
static int
remoteConfigGetStringList(virConfPtr conf, const char *key, char ***list_arg,
                          const char *filename)
{
    char **list;
    virConfValuePtr p = virConfGetValue (conf, key);
    if (!p)
        return 0;

    switch (p->type) {
    case VIR_CONF_STRING:
2518
        if (VIR_ALLOC_N(list, 2) < 0) {
2519
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2520 2521 2522 2523 2524
            return -1;
        }
        list[0] = strdup (p->str);
        list[1] = NULL;
        if (list[0] == NULL) {
2525
            VIR_ERROR(_("failed to allocate memory for %s config list value"),
2526
                      key);
2527
            VIR_FREE(list);
2528 2529 2530 2531 2532 2533 2534 2535 2536
            return -1;
        }
        break;

    case VIR_CONF_LIST: {
        int i, len = 0;
        virConfValuePtr pp;
        for (pp = p->list; pp; pp = pp->next)
            len++;
2537
        if (VIR_ALLOC_N(list, 1+len) < 0) {
2538
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2539 2540 2541 2542
            return -1;
        }
        for (i = 0, pp = p->list; pp; ++i, pp = pp->next) {
            if (pp->type != VIR_CONF_STRING) {
2543
                VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
2544
                            " must be a string or list of strings"),
2545
                          filename, key);
2546
                VIR_FREE(list);
2547 2548 2549 2550 2551 2552
                return -1;
            }
            list[i] = strdup (pp->str);
            if (list[i] == NULL) {
                int j;
                for (j = 0 ; j < i ; j++)
2553 2554
                    VIR_FREE(list[j]);
                VIR_FREE(list);
2555 2556
                VIR_ERROR(_("failed to allocate memory for %s config list value"),
                          key);
2557 2558 2559 2560 2561 2562 2563 2564 2565
                return -1;
            }

        }
        list[i] = NULL;
        break;
    }

    default:
2566
        VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
2567
                    " must be a string or list of strings"),
2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581
                  filename, key);
        return -1;
    }

    *list_arg = list;
    return 0;
}

/* A helper function used by each of the following macros.  */
static int
checkType (virConfValuePtr p, const char *filename,
           const char *key, virConfType required_type)
{
    if (p->type != required_type) {
2582
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: invalid type:"
2583
                    " got %s; expected %s"), filename, key,
2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602
                  virConfTypeName (p->type),
                  virConfTypeName (required_type));
        return -1;
    }
    return 0;
}

/* If there is no config data for the key, #var_name, then do nothing.
   If there is valid data of type VIR_CONF_STRING, and strdup succeeds,
   store the result in var_name.  Otherwise, (i.e. invalid type, or strdup
   failure), give a diagnostic and "goto" the cleanup-and-fail label.  */
#define GET_CONF_STR(conf, filename, var_name)                          \
    do {                                                                \
        virConfValuePtr p = virConfGetValue (conf, #var_name);          \
        if (p) {                                                        \
            if (checkType (p, filename, #var_name, VIR_CONF_STRING) < 0) \
                goto free_and_fail;                                     \
            (var_name) = strdup (p->str);                               \
            if ((var_name) == NULL) {                                   \
2603
                char ebuf[1024];                                        \
2604
                VIR_ERROR(_("remoteReadConfigFile: %s"),		\
2605
                          virStrerror(errno, ebuf, sizeof ebuf));       \
2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621
                goto free_and_fail;                                     \
            }                                                           \
        }                                                               \
    } while (0)

/* Like GET_CONF_STR, but for integral values.  */
#define GET_CONF_INT(conf, filename, var_name)                          \
    do {                                                                \
        virConfValuePtr p = virConfGetValue (conf, #var_name);          \
        if (p) {                                                        \
            if (checkType (p, filename, #var_name, VIR_CONF_LONG) < 0)  \
                goto free_and_fail;                                     \
            (var_name) = p->l;                                          \
        }                                                               \
    } while (0)

2622 2623 2624 2625 2626 2627 2628 2629

static int remoteConfigGetAuth(virConfPtr conf, const char *key, int *auth, const char *filename) {
    virConfValuePtr p;

    p = virConfGetValue (conf, key);
    if (!p)
        return 0;

2630
    if (checkType (p, filename, key, VIR_CONF_STRING) < 0)
2631 2632 2633 2634 2635 2636 2637 2638 2639 2640
        return -1;

    if (!p->str)
        return 0;

    if (STREQ(p->str, "none")) {
        *auth = REMOTE_AUTH_NONE;
#if HAVE_SASL
    } else if (STREQ(p->str, "sasl")) {
        *auth = REMOTE_AUTH_SASL;
2641 2642 2643 2644
#endif
#if HAVE_POLKIT
    } else if (STREQ(p->str, "polkit")) {
        *auth = REMOTE_AUTH_POLKIT;
2645 2646
#endif
    } else {
2647
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: unsupported auth %s"),
2648
                  filename, key, p->str);
2649 2650 2651 2652 2653 2654
        return -1;
    }

    return 0;
}

2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674
#ifdef HAVE_SASL
static inline int
remoteReadSaslAllowedUsernameList (virConfPtr conf,
                                   struct qemud_server *server,
                                   const char *filename)
{
    return
        remoteConfigGetStringList (conf, "sasl_allowed_username_list",
                                   &server->saslUsernameWhitelist, filename);
}
#else
static inline int
remoteReadSaslAllowedUsernameList (virConfPtr conf ATTRIBUTE_UNUSED,
                                   struct qemud_server *server ATTRIBUTE_UNUSED,
                                   const char *filename ATTRIBUTE_UNUSED)
{
    return 0;
}
#endif

2675 2676 2677 2678 2679 2680
/*
 * Set up the logging environment
 * By default if daemonized all errors go to syslog and the logging
 * is also saved onto the logfile libvird.log, but if verbose or error
 * debugging is asked for then output informations or debug.
 */
2681
static int
2682 2683 2684
qemudSetLogging(virConfPtr conf, const char *filename)
{
    int log_level = 0;
2685 2686
    char *log_filters = NULL;
    char *log_outputs = NULL;
2687 2688
    int ret = -1;

2689 2690
    virLogReset();

2691
    /*
2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703
     * Libvirtd's order of precedence is:
     * cmdline > environment > config
     *
     * In order to achieve this, we must process configuration in
     * different order for the log level versus the filters and
     * outputs. Because filters and outputs append, we have to look at
     * the environment first and then only check the config file if
     * there was no result from the environment. The default output is
     * then applied only if there was no setting from either of the
     * first two. Because we don't have a way to determine if the log
     * level has been set, we must process variables in the opposite
     * order, each one overriding the previous.
2704
     */
2705
    GET_CONF_INT (conf, filename, log_level);
2706 2707 2708
    if (log_level != 0)
        virLogSetDefaultPriority(log_level);

2709
    virLogSetFromEnv();
2710 2711 2712 2713

    if (virLogGetNbFilters() == 0) {
        GET_CONF_STR (conf, filename, log_filters);
        virLogParseFilters(log_filters);
2714
    }
2715

2716 2717 2718
    if (virLogGetNbOutputs() == 0) {
        GET_CONF_STR (conf, filename, log_outputs);
        virLogParseOutputs(log_outputs);
2719
    }
2720 2721

    /*
2722 2723
     * If no defined outputs, then direct to syslog when running
     * as daemon. Otherwise the default output is stderr.
2724
     */
2725
    if (virLogGetNbOutputs() == 0) {
2726
        char *tmp = NULL;
2727
        if (godaemon) {
2728 2729
            if (virAsprintf (&tmp, "%d:syslog:libvirtd",
                             virLogGetDefaultPriority()) < 0)
2730 2731
                goto free_and_fail;
        } else {
2732 2733
            if (virAsprintf (&tmp, "%d:stderr",
                             virLogGetDefaultPriority()) < 0)
2734
                goto free_and_fail;
2735
        }
2736 2737
        virLogParseOutputs(tmp);
        VIR_FREE(tmp);
2738
    }
2739 2740 2741 2742 2743 2744 2745

    /*
     * Command line override for --verbose
     */
    if ((verbose) && (virLogGetDefaultPriority() > VIR_LOG_INFO))
        virLogSetDefaultPriority(VIR_LOG_INFO);

2746 2747
    ret = 0;

2748 2749 2750
free_and_fail:
    VIR_FREE(log_filters);
    VIR_FREE(log_outputs);
2751
    return(ret);
2752
}
2753

2754 2755 2756 2757
/* Read the config file if it exists.
 * Only used in the remote case, hence the name.
 */
static int
2758
remoteReadConfigFile (struct qemud_server *server, const char *filename)
2759 2760 2761
{
    virConfPtr conf;

2762 2763 2764 2765 2766
    /* The following variable names must match the corresponding
       configuration strings.  */
    char *unix_sock_ro_perms = NULL;
    char *unix_sock_rw_perms = NULL;
    char *unix_sock_group = NULL;
2767
    char *buf = NULL;
2768
    char *host_uuid = NULL;
2769

2770 2771
#if HAVE_POLKIT
    /* Change the default back to no auth for non-root */
2772
    if (!server->privileged && auth_unix_rw == REMOTE_AUTH_POLKIT)
2773
        auth_unix_rw = REMOTE_AUTH_NONE;
2774
    if (!server->privileged && auth_unix_ro == REMOTE_AUTH_POLKIT)
2775 2776 2777
        auth_unix_ro = REMOTE_AUTH_NONE;
#endif

2778
    conf = virConfReadFile (filename, 0);
2779
    if (!conf) return -1;
2780

2781 2782 2783
    /*
     * First get all the logging settings and activate them
     */
2784 2785
    if (qemudSetLogging(conf, filename) < 0)
        goto free_and_fail;
2786

2787 2788
    GET_CONF_INT (conf, filename, listen_tcp);
    GET_CONF_INT (conf, filename, listen_tls);
2789 2790
    GET_CONF_STR (conf, filename, tls_port);
    GET_CONF_STR (conf, filename, tcp_port);
2791
    GET_CONF_STR (conf, filename, listen_addr);
J
Jim Meyering 已提交
2792

2793
    if (remoteConfigGetAuth(conf, "auth_unix_rw", &auth_unix_rw, filename) < 0)
J
Jim Meyering 已提交
2794
        goto free_and_fail;
2795 2796 2797 2798 2799 2800 2801
#if HAVE_POLKIT
    /* Change default perms to be wide-open if PolicyKit is enabled.
     * Admin can always override in config file
     */
    if (auth_unix_rw == REMOTE_AUTH_POLKIT)
        unix_sock_rw_mask = 0777;
#endif
2802
    if (remoteConfigGetAuth(conf, "auth_unix_ro", &auth_unix_ro, filename) < 0)
J
Jim Meyering 已提交
2803
        goto free_and_fail;
2804
    if (remoteConfigGetAuth(conf, "auth_tcp", &auth_tcp, filename) < 0)
J
Jim Meyering 已提交
2805
        goto free_and_fail;
2806
    if (remoteConfigGetAuth(conf, "auth_tls", &auth_tls, filename) < 0)
J
Jim Meyering 已提交
2807
        goto free_and_fail;
2808

2809 2810
    GET_CONF_STR (conf, filename, unix_sock_group);
    if (unix_sock_group) {
2811
        if (!server->privileged) {
2812
            VIR_WARN0("Cannot set group when not running as root");
2813
        } else {
2814
            int ret;
2815
            struct group grpdata, *grp;
2816 2817 2818 2819 2820 2821
            size_t maxbuf = sysconf(_SC_GETGR_R_SIZE_MAX);

            if (maxbuf == -1)
                maxbuf = 1024;

            if (VIR_ALLOC_N(buf, maxbuf) < 0) {
2822
                VIR_ERROR0(_("Failed to allocate memory for buffer"));
2823 2824 2825 2826 2827 2828 2829 2830
                goto free_and_fail;
            }

            while ((ret = getgrnam_r(unix_sock_group, &grpdata,
                                     buf, maxbuf,
                                     &grp)) == ERANGE) {
                    maxbuf *= 2;
                    if (maxbuf > 65536 || VIR_REALLOC_N(buf, maxbuf) < 0) {
2831
                        VIR_ERROR0(_("Failed to reallocate enough memory for buffer"));
2832 2833 2834 2835 2836
                        goto free_and_fail;
                    }
            }

            if (ret != 0 || !grp) {
2837
                VIR_ERROR(_("Failed to lookup group '%s'"), unix_sock_group);
2838
                goto free_and_fail;
2839 2840
            }
            unix_sock_gid = grp->gr_gid;
2841
            VIR_FREE(buf);
2842
        }
2843
        VIR_FREE(unix_sock_group);
2844 2845
    }

2846 2847
    GET_CONF_STR (conf, filename, unix_sock_ro_perms);
    if (unix_sock_ro_perms) {
2848
        if (virStrToLong_i (unix_sock_ro_perms, NULL, 8, &unix_sock_ro_mask) != 0) {
2849
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_ro_perms);
2850
            goto free_and_fail;
2851
        }
2852
        VIR_FREE(unix_sock_ro_perms);
2853 2854
    }

2855 2856
    GET_CONF_STR (conf, filename, unix_sock_rw_perms);
    if (unix_sock_rw_perms) {
2857
        if (virStrToLong_i (unix_sock_rw_perms, NULL, 8, &unix_sock_rw_mask) != 0) {
2858
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_rw_perms);
2859
            goto free_and_fail;
2860
        }
2861
        VIR_FREE(unix_sock_rw_perms);
2862 2863
    }

2864 2865
    GET_CONF_STR (conf, filename, unix_sock_dir);

2866 2867
    GET_CONF_INT (conf, filename, mdns_adv);
    GET_CONF_STR (conf, filename, mdns_name);
2868

2869
    GET_CONF_INT (conf, filename, tls_no_verify_certificate);
2870

2871 2872 2873 2874
    GET_CONF_STR (conf, filename, key_file);
    GET_CONF_STR (conf, filename, cert_file);
    GET_CONF_STR (conf, filename, ca_file);
    GET_CONF_STR (conf, filename, crl_file);
2875

2876 2877 2878
    if (remoteConfigGetStringList (conf, "tls_allowed_dn_list",
                                   &tls_allowed_dn_list, filename) < 0)
        goto free_and_fail;
2879

2880
    if (remoteReadSaslAllowedUsernameList (conf, server, filename) < 0)
2881
        goto free_and_fail;
2882

2883 2884 2885 2886 2887

    GET_CONF_INT (conf, filename, min_workers);
    GET_CONF_INT (conf, filename, max_workers);
    GET_CONF_INT (conf, filename, max_clients);

2888 2889 2890
    GET_CONF_INT (conf, filename, max_requests);
    GET_CONF_INT (conf, filename, max_client_requests);

2891 2892 2893
    GET_CONF_INT (conf, filename, audit_level);
    GET_CONF_INT (conf, filename, audit_logging);

2894
    GET_CONF_STR (conf, filename, host_uuid);
2895 2896
    if (virSetHostUUIDStr(host_uuid)) {
        VIR_ERROR(_("invalid host UUID: %s"), host_uuid);
2897
        goto free_and_fail;
2898
    }
2899 2900 2901

    VIR_FREE(host_uuid);

2902 2903
    virConfFree (conf);
    return 0;
2904

2905 2906
 free_and_fail:
    virConfFree (conf);
2907
    VIR_FREE(host_uuid);
2908 2909 2910 2911 2912
    VIR_FREE(mdns_name);
    VIR_FREE(unix_sock_ro_perms);
    VIR_FREE(unix_sock_rw_perms);
    VIR_FREE(unix_sock_group);
    VIR_FREE(buf);
2913

2914 2915 2916 2917
    /* Don't bother trying to free listen_addr, tcp_port, tls_port, key_file,
       cert_file, ca_file, or crl_file, since they are initialized to
       non-malloc'd strings.  Besides, these are static variables, and callers
       are unlikely to call this function more than once, so there wouldn't
2918 2919 2920 2921 2922
       even be a real leak.  */

    if (tls_allowed_dn_list) {
        int i;
        for (i = 0; tls_allowed_dn_list[i]; i++)
2923 2924
            VIR_FREE(tls_allowed_dn_list[i]);
        VIR_FREE(tls_allowed_dn_list);
2925 2926 2927
    }

    return -1;
2928 2929
}

2930 2931
/* Display version information. */
static void
2932
version (void)
2933 2934 2935 2936
{
    printf ("%s (%s) %s\n", argv0, PACKAGE_NAME, PACKAGE_VERSION);
}

J
John Levon 已提交
2937 2938 2939 2940 2941 2942 2943 2944
#ifdef __sun
static int
qemudSetupPrivs (void)
{
    chown ("/var/run/libvirt", SYSTEM_UID, SYSTEM_UID);

    if (__init_daemon_priv (PU_RESETGROUPS | PU_CLEARLIMITSET,
        SYSTEM_UID, SYSTEM_UID, PRIV_XVM_CONTROL, NULL)) {
2945
        VIR_ERROR0(_("additional privileges are required"));
J
John Levon 已提交
2946 2947 2948 2949 2950
        return -1;
    }

    if (priv_set (PRIV_OFF, PRIV_ALLSETS, PRIV_FILE_LINK_ANY, PRIV_PROC_INFO,
        PRIV_PROC_SESSION, PRIV_PROC_EXEC, PRIV_PROC_FORK, NULL)) {
2951
        VIR_ERROR0(_("failed to set reduced privileges"));
J
John Levon 已提交
2952 2953 2954 2955 2956 2957
        return -1;
    }

    return 0;
}
#else
2958
# define qemudSetupPrivs() 0
J
John Levon 已提交
2959 2960
#endif

2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015

/*
 * Doing anything non-trivial in signal handlers is pretty dangerous,
 * since there are very few async-signal safe POSIX funtions. To
 * deal with this we setup a very simple signal handler. It simply
 * writes the signal number to a pipe. The main event loop then sees
 * the signal on the pipe and can safely do the processing from
 * event loop context
 */
static int
daemonSetupSignals(struct qemud_server *server)
{
    struct sigaction sig_action;
    int sigpipe[2];

    if (pipe(sigpipe) < 0)
        return -1;

    if (virSetNonBlock(sigpipe[0]) < 0 ||
        virSetNonBlock(sigpipe[1]) < 0 ||
        virSetCloseExec(sigpipe[0]) < 0 ||
        virSetCloseExec(sigpipe[1]) < 0) {
        char ebuf[1024];
        VIR_ERROR(_("Failed to create pipe: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
        goto error;
    }

    sig_action.sa_sigaction = sig_handler;
    sig_action.sa_flags = SA_SIGINFO;
    sigemptyset(&sig_action.sa_mask);

    sigaction(SIGHUP, &sig_action, NULL);
    sigaction(SIGINT, &sig_action, NULL);
    sigaction(SIGQUIT, &sig_action, NULL);
    sigaction(SIGTERM, &sig_action, NULL);

    sig_action.sa_handler = SIG_IGN;
    sigaction(SIGPIPE, &sig_action, NULL);

    if (virEventAddHandleImpl(sigpipe[0],
                              VIR_EVENT_HANDLE_READABLE,
                              qemudDispatchSignalEvent,
                              server, NULL) < 0) {
        VIR_ERROR0(_("Failed to register callback for signal pipe"));
        goto error;
    }

    server->sigread = sigpipe[0];
    server->sigwrite = sigpipe[1];
    sigwrite = sigpipe[1];

    return 0;

error:
3016 3017
    VIR_FORCE_CLOSE(sigpipe[0]);
    VIR_FORCE_CLOSE(sigpipe[1]);
3018 3019 3020
    return -1;
}

3021 3022
/* Print command-line usage. */
static void
3023
usage (void)
3024 3025
{
    fprintf (stderr,
3026
             _("\n\
3027 3028 3029 3030 3031 3032
Usage:\n\
  %s [options]\n\
\n\
Options:\n\
  -v | --verbose         Verbose messages.\n\
  -d | --daemon          Run as a daemon & write PID file.\n\
3033 3034
  -l | --listen          Listen for TCP/IP connections.\n\
  -t | --timeout <secs>  Exit after timeout period.\n\
3035
  -f | --config <file>   Configuration file.\n\
3036
     | --version         Display version information.\n\
3037 3038
  -p | --pid-file <file> Change name of PID file.\n\
\n\
3039
libvirt management daemon:\n\
3040 3041 3042
\n\
  Default paths:\n\
\n\
W
Wolfgang Mauerer 已提交
3043
    Configuration file (unless overridden by -f):\n\
3044
      %s/libvirt/libvirtd.conf\n\
3045
\n\
3046
    Sockets (as root):\n\
3047 3048
      %s/run/libvirt/libvirt-sock\n\
      %s/run/libvirt/libvirt-sock-ro\n\
3049 3050 3051
\n\
    Sockets (as non-root):\n\
      $HOME/.libvirt/libvirt-sock (in UNIX abstract namespace)\n\
3052 3053
\n\
    TLS:\n\
3054 3055 3056
      CA certificate:     %s\n\
      Server certificate: %s\n\
      Server private key: %s\n\
3057 3058 3059
\n\
    PID file (unless overridden by --pid-file):\n\
      %s\n\
3060 3061
\n"),
               argv0,
3062 3063 3064
               SYSCONFDIR,
               LOCALSTATEDIR,
               LOCALSTATEDIR,
3065 3066 3067 3068 3069 3070
               LIBVIRT_CACERT,
               LIBVIRT_SERVERCERT,
               LIBVIRT_SERVERKEY,
               (REMOTE_PID_FILE[0] != '\0'
                ? REMOTE_PID_FILE
                : _("(disabled in ./configure)")));
3071 3072
}

3073 3074 3075 3076
enum {
    OPT_VERSION = 129
};

D
Daniel P. Berrange 已提交
3077 3078
#define MAX_LISTEN 5
int main(int argc, char **argv) {
3079
    struct qemud_server *server = NULL;
3080
    const char *pid_file = NULL;
3081
    const char *remote_config_file = NULL;
3082
    int statuswrite = -1;
3083
    int ret = 1;
3084
    argv0 = argv[0];
D
Daniel P. Berrange 已提交
3085 3086 3087 3088

    struct option opts[] = {
        { "verbose", no_argument, &verbose, 1},
        { "daemon", no_argument, &godaemon, 1},
3089
        { "listen", no_argument, &ipsock, 1},
3090
        { "config", required_argument, NULL, 'f'},
3091 3092
        { "timeout", required_argument, NULL, 't'},
        { "pid-file", required_argument, NULL, 'p'},
3093
        { "version", no_argument, NULL, OPT_VERSION },
3094
        { "help", no_argument, NULL, '?' },
D
Daniel P. Berrange 已提交
3095 3096 3097
        {0, 0, 0, 0}
    };

E
Eric Blake 已提交
3098 3099 3100 3101 3102 3103
    if (setlocale (LC_ALL, "") == NULL ||
        bindtextdomain (PACKAGE, LOCALEDIR) == NULL ||
        textdomain(PACKAGE) == NULL ||
        virInitialize() < 0) {
        fprintf(stderr, _("%s: initialization failed\n"), argv0);
        exit(EXIT_FAILURE);
3104
    }
D
Daniel P. Berrange 已提交
3105

D
Daniel P. Berrange 已提交
3106 3107 3108 3109 3110
    while (1) {
        int optidx = 0;
        int c;
        char *tmp;

3111
        c = getopt_long(argc, argv, "ldf:p:t:v", opts, &optidx);
D
Daniel P. Berrange 已提交
3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126

        if (c == -1) {
            break;
        }

        switch (c) {
        case 0:
            /* Got one of the flags */
            break;
        case 'v':
            verbose = 1;
            break;
        case 'd':
            godaemon = 1;
            break;
3127 3128
        case 'l':
            ipsock = 1;
D
Daniel P. Berrange 已提交
3129 3130 3131
            break;

        case 't':
3132
            if (virStrToLong_i(optarg, &tmp, 10, &timeout) != 0
3133 3134 3135
                || timeout <= 0
                /* Ensure that we can multiply by 1000 without overflowing.  */
                || timeout > INT_MAX / 1000)
D
Daniel P. Berrange 已提交
3136 3137
                timeout = -1;
            break;
3138 3139

        case 'p':
3140 3141 3142 3143 3144
            pid_file = optarg;
            break;

        case 'f':
            remote_config_file = optarg;
3145 3146
            break;

3147
        case OPT_VERSION:
3148
            version ();
3149 3150
            return 0;

D
Daniel P. Berrange 已提交
3151
        case '?':
3152
            usage ();
D
Daniel P. Berrange 已提交
3153 3154 3155
            return 2;

        default:
3156 3157
            fprintf (stderr, _("%s: internal error: unknown flag: %c\n"),
                     argv0, c);
3158
            exit (EXIT_FAILURE);
D
Daniel P. Berrange 已提交
3159 3160 3161
        }
    }

3162 3163
    if (remote_config_file == NULL) {
        static const char *default_config_file
3164
            = SYSCONFDIR "/libvirt/libvirtd.conf";
3165
        remote_config_file =
3166
            (access(default_config_file, R_OK) == 0
3167 3168 3169 3170
             ? default_config_file
             : "/dev/null");
    }

3171
    if (godaemon) {
3172
        char ebuf[1024];
3173 3174 3175 3176 3177 3178 3179

        if (chdir("/") < 0) {
            VIR_ERROR(_("cannot change to root directory: %s"),
                      virStrerror(errno, ebuf, sizeof(ebuf)));
            goto error;
        }

3180
        if ((statuswrite = daemonForkIntoBackground()) < 0) {
3181 3182
            VIR_ERROR(_("Failed to fork as daemon: %s"),
                      virStrerror(errno, ebuf, sizeof ebuf));
3183
            goto error;
3184 3185 3186 3187 3188
        }
    }

    /* If running as root and no PID file is set, use the default */
    if (pid_file == NULL &&
3189
        geteuid() == 0 &&
3190 3191 3192 3193 3194
        REMOTE_PID_FILE[0] != '\0')
        pid_file = REMOTE_PID_FILE;

    /* If we have a pidfile set, claim it now, exiting if already taken */
    if (pid_file != NULL &&
3195 3196 3197
        qemudWritePidFile (pid_file) < 0) {
        pid_file = NULL; /* Prevent unlinking of someone else's pid ! */
        ret = VIR_DAEMON_ERR_PIDFILE;
3198
        goto error;
3199
    }
3200

J
John Levon 已提交
3201
    /* Ensure the rundir exists (on tmpfs on some systems) */
3202
    if (geteuid() == 0) {
3203
        const char *rundir = LOCALSTATEDIR "/run/libvirt";
J
John Levon 已提交
3204 3205 3206

        if (mkdir (rundir, 0755)) {
            if (errno != EEXIST) {
3207 3208 3209
                char ebuf[1024];
                VIR_ERROR(_("unable to create rundir %s: %s"), rundir,
                          virStrerror(errno, ebuf, sizeof(ebuf)));
3210 3211
                ret = VIR_DAEMON_ERR_RUNDIR;
                goto error;
J
John Levon 已提交
3212 3213 3214 3215
            }
        }
    }

3216 3217 3218 3219 3220 3221
    /* Beyond this point, nothing should rely on using
     * getuid/geteuid() == 0, for privilege level checks.
     * It must all use the flag 'server->privileged'
     * which is also passed into all libvirt stateful
     * drivers
     */
3222 3223
    if (qemudSetupPrivs() < 0) {
        ret = VIR_DAEMON_ERR_PRIVS;
3224
        goto error;
3225
    }
J
John Levon 已提交
3226

3227
    if (!(server = qemudInitialize())) {
3228
        ret = VIR_DAEMON_ERR_INIT;
3229
        goto error;
3230
    }
3231

3232 3233
    if ((daemonSetupSignals(server)) < 0) {
        ret = VIR_DAEMON_ERR_SIGNAL;
3234
        goto error;
3235
    }
3236

3237
    /* Read the config file (if it exists). */
3238 3239
    if (remoteReadConfigFile (server, remote_config_file) < 0) {
        ret = VIR_DAEMON_ERR_CONFIG;
3240
        goto error;
3241
    }
D
Daniel P. Berrange 已提交
3242

3243 3244 3245 3246 3247 3248 3249 3250 3251 3252
    if (audit_level) {
        if (virAuditOpen() < 0) {
            if (audit_level > 1) {
                ret = VIR_DAEMON_ERR_AUDIT;
                goto error;
            }
        }
    }
    virAuditLog(audit_logging);

3253
    /* setup the hooks if any */
3254
    if (virHookInitialize() < 0) {
3255 3256 3257 3258
        ret = VIR_DAEMON_ERR_HOOKS;
        goto error;
    }

3259 3260 3261
    /* Disable error func, now logging is setup */
    virSetErrorFunc(NULL, virshErrorHandler);

3262 3263 3264 3265 3266 3267 3268 3269
    /*
     * Call the daemon startup hook
     * TODO: should we abort the daemon startup if the script returned
     *       an error ?
     */
    virHookCall(VIR_HOOK_DRIVER_DAEMON, "-", VIR_HOOK_DAEMON_OP_START,
                0, "start", NULL);

3270
    if (qemudNetworkInit(server) < 0) {
3271
        ret = VIR_DAEMON_ERR_NETWORK;
3272
        goto error;
3273 3274
    }

3275 3276 3277 3278 3279 3280 3281 3282 3283
    /* Tell parent of daemon that basic initialization is complete
     * In particular we're ready to accept net connections & have
     * written the pidfile
     */
    if (statuswrite != -1) {
        char status = 0;
        while (write(statuswrite, &status, 1) == -1 &&
               errno == EINTR)
            ;
3284
        VIR_FORCE_CLOSE(statuswrite);
3285 3286
    }

3287 3288 3289
    /* Start the event loop in a background thread, since
     * state initialization needs events to be being processed */
    if (qemudStartEventLoop(server) < 0) {
3290
        VIR_ERROR0(_("Event thread startup failed"));
3291 3292 3293
        goto error;
    }

3294 3295 3296 3297 3298
    /* Start the stateful HV drivers
     * This is delibrately done after telling the parent process
     * we're ready, since it can take a long time and this will
     * seriously delay OS bootup process */
    if (virStateInitialize(server->privileged) < 0) {
3299
        VIR_ERROR0(_("Driver state initialization failed"));
3300
        goto shutdown;
3301
    }
D
Daniel P. Berrange 已提交
3302

3303 3304 3305
    /* Start accepting new clients from network */
    virMutexLock(&server->lock);
    if (qemudNetworkEnable(server) < 0) {
3306
        VIR_ERROR0(_("Network event loop enablement failed"));
3307 3308 3309 3310
        goto shutdown;
    }
    virMutexUnlock(&server->lock);

3311 3312
    ret = 0;

3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328
shutdown:
    /* In a non-0 shutdown scenario we need to tell event loop
     * to quit immediately. Otherwise in normal case we just
     * sit in the thread join forever. Sure this means the
     * main thread doesn't do anything useful ever, but that's
     * not too much of drain on resources
     */
    if (ret != 0) {
        virMutexLock(&server->lock);
        if (server->hasEventThread)
            /* This SIGQUIT triggers the shutdown process */
            kill(getpid(), SIGQUIT);
        virMutexUnlock(&server->lock);
    }
    pthread_join(server->eventThread, NULL);

3329 3330 3331
    virHookCall(VIR_HOOK_DRIVER_DAEMON, "-", VIR_HOOK_DAEMON_OP_SHUTDOWN,
                0, "shutdown", NULL);

3332
error:
3333 3334 3335 3336 3337 3338 3339 3340
    if (statuswrite != -1) {
        if (ret != 0) {
            /* Tell parent of daemon what failed */
            char status = ret;
            while (write(statuswrite, &status, 1) == -1 &&
                   errno == EINTR)
                ;
        }
3341
        VIR_FORCE_CLOSE(statuswrite);
3342
    }
3343 3344
    if (server)
        qemudCleanup(server);
3345 3346
    if (pid_file)
        unlink (pid_file);
3347
    virLogShutdown();
3348
    return ret;
D
Daniel P. Berrange 已提交
3349
}