libvirtd.c 95.0 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>
50

51
#include "libvirt_internal.h"
52
#include "virterror_internal.h"
53
#include "files.h"
54

55 56
#define VIR_FROM_THIS VIR_FROM_QEMU

D
Daniel P. Berrange 已提交
57
#include "libvirtd.h"
58 59
#include "dispatch.h"

60
#include "util.h"
61
#include "uuid.h"
62
#include "remote_driver.h"
63
#include "conf.h"
D
Daniel P. Berrange 已提交
64
#include "event.h"
65
#include "memory.h"
66
#include "stream.h"
67
#include "hooks.h"
68
#include "virtaudit.h"
69
#ifdef HAVE_AVAHI
70
# include "mdns.h"
71
#endif
D
Daniel P. Berrange 已提交
72

73
#ifdef WITH_DRIVER_MODULES
74
# include "driver.h"
75
#else
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
# 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
103 104 105
# ifdef WITH_NWFILTER
#  include "nwfilter/nwfilter_driver.h"
# endif
106
#endif
107 108


J
John Levon 已提交
109
#ifdef __sun
110 111
# include <ucred.h>
# include <priv.h>
J
John Levon 已提交
112

113 114 115
# ifndef PRIV_VIRT_MANAGE
#  define PRIV_VIRT_MANAGE ((const char *)"virt_manage")
# endif
J
John Levon 已提交
116

117 118 119
# ifndef PRIV_XVM_CONTROL
#  define PRIV_XVM_CONTROL ((const char *)"xvm_control")
# endif
J
John Levon 已提交
120

121 122
# define PU_RESETGROUPS          0x0001  /* Remove supplemental groups */
# define PU_CLEARLIMITSET        0x0008  /* L=0 */
J
John Levon 已提交
123 124 125

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

126
# define SYSTEM_UID 60
J
John Levon 已提交
127 128 129 130 131 132 133 134 135 136 137 138 139

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 */

140 141
static int godaemon = 0;        /* -d: Be a daemon */
static int verbose = 0;         /* -v: Verbose mode */
142
static int timeout = -1;        /* -t: Shutdown timeout */
143
static int sigwrite = -1;       /* Signal handler pipe */
144
static int ipsock = 0;          /* -l  Listen for TCP/IP */
145

146
/* Defaults for configuration file elements */
147 148
static int listen_tls = 1;
static int listen_tcp = 0;
149
static char *listen_addr  = (char *) LIBVIRTD_LISTEN_ADDR;
150 151
static char *tls_port = (char *) LIBVIRTD_TLS_PORT;
static char *tcp_port = (char *) LIBVIRTD_TCP_PORT;
152

153 154
static char *unix_sock_dir = NULL;

155 156 157 158
#if HAVE_POLKIT
static int auth_unix_rw = REMOTE_AUTH_POLKIT;
static int auth_unix_ro = REMOTE_AUTH_POLKIT;
#else
159 160
static int auth_unix_rw = REMOTE_AUTH_NONE;
static int auth_unix_ro = REMOTE_AUTH_NONE;
161
#endif /* HAVE_POLKIT */
162 163 164 165 166 167 168
#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;

169
static int mdns_adv = 1;
170
static char *mdns_name = NULL;
171

172
static int tls_no_verify_certificate = 0;
173
static char **tls_allowed_dn_list = NULL;
174

175 176 177 178
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 *) "";
179 180 181 182

static gnutls_certificate_credentials_t x509_cred;
static gnutls_dh_params_t dh_params;

183 184 185 186
static int min_workers = 5;
static int max_workers = 20;
static int max_clients = 20;

187 188 189 190 191
/* 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;

192 193 194
static int audit_level = 1;
static int audit_logging = 0;

195
#define DH_BITS 1024
196

197 198
static sig_atomic_t sig_errors = 0;
static int sig_lasterrno = 0;
199
static const char *argv0;
200

201 202 203 204 205 206 207 208 209
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,
210
    VIR_DAEMON_ERR_HOOKS,
211
    VIR_DAEMON_ERR_AUDIT,
212 213 214 215 216 217 218 219 220 221 222 223 224

    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",
225
              "Unable to load configuration file",
226 227
              "Unable to look for hook scripts",
              "Unable to initialize audit system")
228

229 230
static void sig_handler(int sig, siginfo_t * siginfo,
                        void* context ATTRIBUTE_UNUSED) {
231
    int origerrno;
232
    int r;
233

234 235
    /* set the sig num in the struct */
    siginfo->si_signo = sig;
236 237

    origerrno = errno;
238
    r = safewrite(sigwrite, siginfo, sizeof(*siginfo));
239 240 241 242
    if (r == -1) {
        sig_errors++;
        sig_lasterrno = errno;
    }
243
    errno = origerrno;
D
Daniel P. Berrange 已提交
244
}
245

246 247
static void qemudDispatchClientEvent(int watch, int fd, int events, void *opaque);
static void qemudDispatchServerEvent(int watch, int fd, int events, void *opaque);
248
static int qemudStartWorker(struct qemud_server *server, struct qemud_worker *worker);
249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264

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;
    }
}

265
struct qemud_client_message *
266 267 268 269 270 271 272 273 274 275 276
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
277

278 279 280 281 282
static int
remoteCheckCertFile(const char *type, const char *file)
{
    struct stat sb;
    if (stat(file, &sb) < 0) {
283 284 285
        char ebuf[1024];
        VIR_ERROR(_("Cannot access %s '%s': %s"),
                  type, file, virStrerror(errno, ebuf, sizeof ebuf));
286 287 288 289 290
        return -1;
    }
    return 0;
}

291 292 293 294 295 296 297 298 299 300
static int
remoteInitializeGnuTLS (void)
{
    int err;

    /* Initialise GnuTLS. */
    gnutls_global_init ();

    err = gnutls_certificate_allocate_credentials (&x509_cred);
    if (err) {
301
        VIR_ERROR(_("gnutls_certificate_allocate_credentials: %s"),
302 303 304 305 306
                  gnutls_strerror (err));
        return -1;
    }

    if (ca_file && ca_file[0] != '\0') {
307 308 309
        if (remoteCheckCertFile("CA certificate", ca_file) < 0)
            return -1;

310 311 312 313
        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) {
314
            VIR_ERROR(_("gnutls_certificate_set_x509_trust_file: %s"),
315 316 317 318 319 320
                      gnutls_strerror (err));
            return -1;
        }
    }

    if (crl_file && crl_file[0] != '\0') {
J
Jim Meyering 已提交
321
        if (remoteCheckCertFile("CA revocation list", crl_file) < 0)
322 323
            return -1;

324
        DEBUG("loading CRL from %s", crl_file);
325 326 327
        err = gnutls_certificate_set_x509_crl_file (x509_cred, crl_file,
                                                    GNUTLS_X509_FMT_PEM);
        if (err < 0) {
328
            VIR_ERROR(_("gnutls_certificate_set_x509_crl_file: %s"),
329 330 331 332 333 334
                      gnutls_strerror (err));
            return -1;
        }
    }

    if (cert_file && cert_file[0] != '\0' && key_file && key_file[0] != '\0') {
335 336 337 338
        if (remoteCheckCertFile("server certificate", cert_file) < 0)
            return -1;
        if (remoteCheckCertFile("server key", key_file) < 0)
            return -1;
339
        DEBUG("loading cert and key from %s and %s", cert_file, key_file);
340 341 342 343 344
        err =
            gnutls_certificate_set_x509_key_file (x509_cred,
                                                  cert_file, key_file,
                                                  GNUTLS_X509_FMT_PEM);
        if (err < 0) {
345
            VIR_ERROR(_("gnutls_certificate_set_x509_key_file: %s"),
346 347 348 349 350 351 352 353 354 355 356 357
                      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) {
358
        VIR_ERROR(_("gnutls_dh_params_init: %s"), gnutls_strerror (err));
359 360 361 362
        return -1;
    }
    err = gnutls_dh_params_generate2 (dh_params, DH_BITS);
    if (err < 0) {
363
        VIR_ERROR(_("gnutls_dh_params_generate2: %s"), gnutls_strerror (err));
364 365 366 367 368 369 370 371
        return -1;
    }

    gnutls_certificate_set_dh_params (x509_cred, dh_params);

    return 0;
}

372
static void
373 374
qemudDispatchSignalEvent(int watch ATTRIBUTE_UNUSED,
                         int fd ATTRIBUTE_UNUSED,
375 376
                         int events ATTRIBUTE_UNUSED,
                         void *opaque) {
D
Daniel P. Berrange 已提交
377
    struct qemud_server *server = (struct qemud_server *)opaque;
378
    siginfo_t siginfo;
379

380
    virMutexLock(&server->lock);
381

382
    if (saferead(server->sigread, &siginfo, sizeof(siginfo)) != sizeof(siginfo)) {
383 384 385
        char ebuf[1024];
        VIR_ERROR(_("Failed to read from signal pipe: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
386
        virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
387
        return;
388 389
    }

390
    switch (siginfo.si_signo) {
391
    case SIGHUP:
392
        VIR_INFO0(_("Reloading configuration on SIGHUP"));
393 394
        virHookCall(VIR_HOOK_DRIVER_DAEMON, "-",
                    VIR_HOOK_DAEMON_OP_RELOAD, SIGHUP, "SIGHUP", NULL);
395
        if (virStateReload() < 0)
396
            VIR_WARN0("Error while reloading drivers");
397

398 399 400
        break;

    case SIGINT:
401
    case SIGQUIT:
402
    case SIGTERM:
403
        VIR_WARN("Shutting down on signal %d", siginfo.si_signo);
404
        server->quitEventThread = 1;
405 406 407
        break;

    default:
408
        VIR_INFO(_("Received unexpected signal %d"), siginfo.si_signo);
409 410 411
        break;
    }

412
    virMutexUnlock(&server->lock);
413 414
}

415

416 417 418 419 420
static int daemonForkIntoBackground(void) {
    int statuspipe[2];
    if (pipe(statuspipe) < 0)
        return -1;

D
Daniel P. Berrange 已提交
421 422 423 424 425 426
    int pid = fork();
    switch (pid) {
    case 0:
        {
            int stdinfd = -1;
            int stdoutfd = -1;
427
            int nextpid;
D
Daniel P. Berrange 已提交
428

429
            VIR_FORCE_CLOSE(statuspipe[0]);
430

431
            if ((stdinfd = open("/dev/null", O_RDONLY)) < 0)
D
Daniel P. Berrange 已提交
432
                goto cleanup;
433
            if ((stdoutfd = open("/dev/null", O_WRONLY)) < 0)
D
Daniel P. Berrange 已提交
434 435 436 437 438 439 440
                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;
441
            if (VIR_CLOSE(stdinfd) < 0)
D
Daniel P. Berrange 已提交
442
                goto cleanup;
443
            if (VIR_CLOSE(stdoutfd) < 0)
D
Daniel P. Berrange 已提交
444 445 446 447 448 449 450 451
                goto cleanup;

            if (setsid() < 0)
                goto cleanup;

            nextpid = fork();
            switch (nextpid) {
            case 0:
452
                return statuspipe[1];
D
Daniel P. Berrange 已提交
453 454 455
            case -1:
                return -1;
            default:
456
                _exit(0);
D
Daniel P. Berrange 已提交
457 458 459
            }

        cleanup:
460 461
            VIR_FORCE_CLOSE(stdoutfd);
            VIR_FORCE_CLOSE(stdinfd);
D
Daniel P. Berrange 已提交
462 463 464 465 466 467 468 469 470
            return -1;

        }

    case -1:
        return -1;

    default:
        {
471 472 473 474
            int got, exitstatus = 0;
            int ret;
            char status;

475
            VIR_FORCE_CLOSE(statuspipe[1]);
476 477 478

            /* We wait to make sure the first child forked successfully */
            if ((got = waitpid(pid, &exitstatus, 0)) < 0 ||
D
Daniel P. Berrange 已提交
479
                got != pid ||
480
                exitstatus != 0) {
D
Daniel P. Berrange 已提交
481 482
                return -1;
            }
483 484 485 486 487 488 489 490

            /* 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) {
491
                fprintf(stderr,
492 493
                        _("%s: error: %s. Check /var/log/messages or run without "
                          "--daemon for more info.\n"), argv0,
494
                        virDaemonErrTypeToString(status));
495 496
            }
            _exit(ret == 1 && status == 0 ? 0 : 1);
D
Daniel P. Berrange 已提交
497 498 499 500
        }
    }
}

501 502 503
static int qemudWritePidFile(const char *pidFile) {
    int fd;
    FILE *fh;
504
    char ebuf[1024];
505 506 507 508 509

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

    if ((fd = open(pidFile, O_WRONLY|O_CREAT|O_EXCL, 0644)) < 0) {
510
        VIR_ERROR(_("Failed to open pid file '%s' : %s"),
511
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
512 513 514
        return -1;
    }

515
    if (!(fh = VIR_FDOPEN(fd, "w"))) {
516
        VIR_ERROR(_("Failed to fdopen pid file '%s' : %s"),
517
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
518
        VIR_FORCE_CLOSE(fd);
519 520 521 522
        return -1;
    }

    if (fprintf(fh, "%lu\n", (unsigned long)getpid()) < 0) {
523 524
        VIR_ERROR(_("%s: Failed to write to pid file '%s' : %s"),
                  argv0, pidFile, virStrerror(errno, ebuf, sizeof ebuf));
525
        VIR_FORCE_FCLOSE(fh);
526 527 528
        return -1;
    }

529
    if (VIR_FCLOSE(fh) == EOF) {
530 531
        VIR_ERROR(_("%s: Failed to close pid file '%s' : %s"),
                  argv0, pidFile, virStrerror(errno, ebuf, sizeof ebuf));
532 533 534 535 536 537
        return -1;
    }

    return 0;
}

D
Daniel P. Berrange 已提交
538
static int qemudListenUnix(struct qemud_server *server,
539
                           char *path, int readonly, int auth) {
540
    struct qemud_socket *sock;
D
Daniel P. Berrange 已提交
541
    mode_t oldmask;
542
    gid_t oldgrp;
543
    char ebuf[1024];
D
Daniel P. Berrange 已提交
544

545
    if (VIR_ALLOC(sock) < 0) {
546
        VIR_ERROR0(_("Failed to allocate memory for struct qemud_socket"));
D
Daniel P. Berrange 已提交
547
        return -1;
548
    }
D
Daniel P. Berrange 已提交
549 550

    sock->readonly = readonly;
551
    sock->type = QEMUD_SOCK_TYPE_UNIX;
552
    sock->auth = auth;
553
    sock->path = path;
554 555 556 557 558 559
    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
560

561
    if ((sock->fd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
562
        VIR_ERROR(_("Failed to create socket: %s"),
563
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
564
        goto cleanup;
565
    }
D
Daniel P. Berrange 已提交
566

567 568
    if (virSetCloseExec(sock->fd) < 0 ||
        virSetNonBlock(sock->fd) < 0)
D
Daniel P. Berrange 已提交
569
        goto cleanup;
D
Daniel P. Berrange 已提交
570

571 572
    sock->addr.data.un.sun_family = AF_UNIX;
    if (virStrcpyStatic(sock->addr.data.un.sun_path, path) == NULL) {
C
Chris Lalancette 已提交
573 574 575
        VIR_ERROR(_("Path %s too long for unix socket"), path);
        goto cleanup;
    }
576 577
    if (sock->addr.data.un.sun_path[0] == '@')
        sock->addr.data.un.sun_path[0] = '\0';
D
Daniel P. Berrange 已提交
578

579
    oldgrp = getgid();
580
    oldmask = umask(readonly ? ~unix_sock_ro_mask : ~unix_sock_rw_mask);
581 582 583 584
    if (server->privileged && setgid(unix_sock_gid)) {
        VIR_ERROR(_("Failed to set group ID to %d"), unix_sock_gid);
        goto cleanup;
    }
585

586
    if (bind(sock->fd, &sock->addr.data.sa, sock->addr.len) < 0) {
587
        VIR_ERROR(_("Failed to bind socket to '%s': %s"),
588
                  path, virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
589
        goto cleanup;
590
    }
D
Daniel P. Berrange 已提交
591
    umask(oldmask);
592 593 594 595
    if (server->privileged && setgid(oldgrp)) {
        VIR_ERROR(_("Failed to restore group ID to %d"), oldgrp);
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
596

597
    if (listen(sock->fd, 30) < 0) {
598
        VIR_ERROR(_("Failed to listen for connections on '%s': %s"),
599
                  path, virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
600 601 602 603 604 605 606
        goto cleanup;
    }

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

D
Daniel P. Berrange 已提交
607
    return 0;
D
Daniel P. Berrange 已提交
608 609

 cleanup:
610
    VIR_FORCE_CLOSE(sock->fd);
611
    VIR_FREE(sock);
D
Daniel P. Berrange 已提交
612
    return -1;
D
Daniel P. Berrange 已提交
613 614
}

615 616
// See: http://people.redhat.com/drepper/userapi-ipv6.html
static int
617
remoteMakeSockets (int *fds, int max_fds, int *nfds_r, const char *node, const char *service)
618 619 620 621 622 623 624
{
    struct addrinfo *ai;
    struct addrinfo hints;
    memset (&hints, 0, sizeof hints);
    hints.ai_flags = AI_PASSIVE | AI_ADDRCONFIG;
    hints.ai_socktype = SOCK_STREAM;

625
    int e = getaddrinfo (node, service, &hints, &ai);
626
    if (e != 0) {
627
        VIR_ERROR(_("getaddrinfo: %s"), gai_strerror (e));
628 629 630 631 632
        return -1;
    }

    struct addrinfo *runp = ai;
    while (runp && *nfds_r < max_fds) {
633
        char ebuf[1024];
634 635 636
        fds[*nfds_r] = socket (runp->ai_family, runp->ai_socktype,
                               runp->ai_protocol);
        if (fds[*nfds_r] == -1) {
637
            VIR_ERROR(_("socket: %s"), virStrerror (errno, ebuf, sizeof ebuf));
638 639 640 641 642 643
            return -1;
        }

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

644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
#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

659 660
        if (bind (fds[*nfds_r], runp->ai_addr, runp->ai_addrlen) == -1) {
            if (errno != EADDRINUSE) {
661
                VIR_ERROR(_("bind: %s"), virStrerror (errno, ebuf, sizeof ebuf));
662 663
                return -1;
            }
664
            VIR_FORCE_CLOSE(fds[*nfds_r]);
665
        } else {
666 667 668 669
            ++*nfds_r;
        }
        runp = runp->ai_next;
    }
670

671 672 673 674 675 676 677 678 679
    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,
680
                 const char *addr,
681
                 const char *port,
682
                 int type,
683
                 int auth)
684 685 686
{
    int fds[2];
    int nfds = 0;
687
    int i;
688 689
    struct qemud_socket *sock;

690
    if (remoteMakeSockets (fds, 2, &nfds, addr, port) == -1)
691 692 693
        return -1;

    for (i = 0; i < nfds; ++i) {
694
        char ebuf[1024];
695

696
        if (VIR_ALLOC(sock) < 0) {
697 698
            VIR_ERROR(_("remoteListenTCP: calloc: %s"),
                      virStrerror (errno, ebuf, sizeof ebuf));
699
            goto cleanup;
700 701
        }

702
        sock->addr.len = sizeof(sock->addr.data.stor);
703 704 705 706 707 708
        sock->readonly = 0;
        sock->next = server->sockets;
        server->sockets = sock;
        server->nsockets++;

        sock->fd = fds[i];
709
        sock->type = type;
710
        sock->auth = auth;
D
Daniel P. Berrange 已提交
711

712
        if (getsockname(sock->fd, &sock->addr.data.sa, &sock->addr.len) < 0)
713
            goto cleanup;
714

715 716
        if (!(sock->addrstr = virSocketFormatAddrFull(&sock->addr, true, ";")))
            goto cleanup;
717

718 719
        if (virSetCloseExec(sock->fd) < 0 ||
            virSetNonBlock(sock->fd) < 0)
720
            goto cleanup;
721 722

        if (listen (sock->fd, 30) < 0) {
723 724
            VIR_ERROR(_("remoteListenTCP: listen: %s"),
                      virStrerror (errno, ebuf, sizeof ebuf));
725
            goto cleanup;
726
        }
727 728 729
    }

    return 0;
730 731 732

cleanup:
    for (i = 0; i < nfds; ++i)
733
        VIR_FORCE_CLOSE(fds[i]);
734
    return -1;
735 736 737
}

static int qemudInitPaths(struct qemud_server *server,
738 739
                          char **sockname,
                          char **roSockname)
J
John Levon 已提交
740
{
741
    char *base_dir_prefix = NULL;
742
    char *sock_dir_prefix = NULL;
743 744 745 746 747 748 749 750 751 752 753 754
    int ret = -1;

    /* The base_dir_prefix is the base under which all libvirtd
     * files live */
    if (server->privileged) {
        if (!(base_dir_prefix = strdup (LOCAL_STATE_DIR)))
            goto no_memory;
    } else {
        uid_t uid = geteuid();
        if (!(base_dir_prefix = virGetUserDirectory(uid)))
            goto cleanup;
    }
755

756 757
    /* The unix_sock_dir is the location under which all
     * unix domain sockets live */
D
Daniel P. Berrange 已提交
758
    if (unix_sock_dir) {
759 760 761
        if (!(sock_dir_prefix = strdup(unix_sock_dir)))
            goto no_memory;

D
Daniel P. Berrange 已提交
762 763 764 765 766 767 768
        /* 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 {
769
        if (server->privileged) {
770 771 772
            if (virAsprintf(&sock_dir_prefix, "%s/run/libvirt",
                            base_dir_prefix) < 0)
                goto no_memory;
773
        } else {
774 775 776
            if (virAsprintf(&sock_dir_prefix, "%s/.libvirt",
                            base_dir_prefix) < 0)
                goto no_memory;
777 778
        }
    }
779

780
    if (server->privileged) {
781 782 783 784 785 786 787 788
        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
789
    } else {
790 791 792 793 794 795
        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 */
796
    }
797

798
    if (server->privileged) {
799 800 801
        if (virAsprintf(&server->logDir, "%s/log/libvirt",
                        base_dir_prefix) < 0)
            goto no_memory;
802
    } else {
803 804 805
        if (virAsprintf(&server->logDir, "%s/.libvirt/log",
                        base_dir_prefix) < 0)
            goto no_memory;
806
    }
807

808
    ret = 0;
809

810 811 812
no_memory:
    if (ret != 0)
        virReportOOMError();
813 814

 cleanup:
815
    VIR_FREE(base_dir_prefix);
816
    VIR_FREE(sock_dir_prefix);
817
    return ret;
D
Daniel P. Berrange 已提交
818 819
}

820 821 822 823 824 825
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 */
}

826
static struct qemud_server *qemudInitialize(void) {
D
Daniel P. Berrange 已提交
827 828
    struct qemud_server *server;

829
    if (VIR_ALLOC(server) < 0) {
830
        VIR_ERROR0(_("Failed to allocate struct qemud_server"));
D
Daniel P. Berrange 已提交
831
        return NULL;
832
    }
D
Daniel P. Berrange 已提交
833

834 835 836
    server->privileged = geteuid() == 0 ? 1 : 0;
    server->sigread = server->sigwrite = -1;

837
    if (virMutexInit(&server->lock) < 0) {
838
        VIR_ERROR0(_("cannot initialize mutex"));
839
        VIR_FREE(server);
840
        return NULL;
841 842
    }
    if (virCondInit(&server->job) < 0) {
843
        VIR_ERROR0(_("cannot initialize condition variable"));
844
        virMutexDestroy(&server->lock);
845
        VIR_FREE(server);
846
        return NULL;
847 848
    }

849
    if (virEventInit() < 0) {
850
        VIR_ERROR0(_("Failed to initialize event system"));
851 852 853
        virMutexDestroy(&server->lock);
        if (virCondDestroy(&server->job) < 0)
        {}
854 855 856 857
        VIR_FREE(server);
        return NULL;
    }

858 859 860 861 862 863 864
    /*
     * 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.
     */
865 866 867
#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
868
     * If they try to open a connection for a module that
869 870 871 872
     * is not loaded they'll get a suitable error at that point
     */
    virDriverLoadModule("network");
    virDriverLoadModule("storage");
873
    virDriverLoadModule("nodedev");
874
    virDriverLoadModule("secret");
875 876 877
    virDriverLoadModule("qemu");
    virDriverLoadModule("lxc");
    virDriverLoadModule("uml");
D
Daniel Veillard 已提交
878
    virDriverLoadModule("one");
879
    virDriverLoadModule("nwfilter");
880
#else
881
# ifdef WITH_NETWORK
882
    networkRegister();
883 884
# endif
# ifdef WITH_NETCF
885
    interfaceRegister();
886 887
# endif
# ifdef WITH_STORAGE_DIR
888
    storageRegister();
889 890
# endif
# if defined(WITH_NODE_DEVICES)
891
    nodedevRegister();
892 893
# endif
# ifdef WITH_SECRETS
894
    secretRegister();
895
# endif
896 897 898
# ifdef WITH_NWFILTER
    nwfilterRegister();
# endif
899
# ifdef WITH_QEMU
900
    qemuRegister();
901 902
# endif
# ifdef WITH_LXC
903
    lxcRegister();
904 905
# endif
# ifdef WITH_UML
906
    umlRegister();
907 908
# endif
# ifdef WITH_ONE
D
Daniel Veillard 已提交
909
    oneRegister();
910
# endif
911 912
#endif

913 914 915 916 917 918
    virEventRegisterImpl(virEventAddHandleImpl,
                         virEventUpdateHandleImpl,
                         virEventRemoveHandleImpl,
                         virEventAddTimeoutImpl,
                         virEventUpdateTimeoutImpl,
                         virEventRemoveTimeoutImpl);
919

920 921 922
    return server;
}

923
static int qemudNetworkInit(struct qemud_server *server) {
924 925
    char *sockname = NULL;
    char *roSockname = NULL;
926
#if HAVE_SASL
927 928 929
    int err;
#endif /* HAVE_SASL */

930
    if (qemudInitPaths(server, &sockname, &roSockname) < 0)
931 932 933
        goto cleanup;

    if (qemudListenUnix(server, sockname, 0, auth_unix_rw) < 0)
934
        goto cleanup;
935
    sockname = NULL;
936

937
    if (roSockname != NULL && qemudListenUnix(server, roSockname, 1, auth_unix_ro) < 0)
938
        goto cleanup;
939
    roSockname = NULL;
940 941 942 943 944 945 946

#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) {
947 948
            VIR_ERROR(_("Failed to initialize SASL authentication %s"),
                      sasl_errstring(err, NULL, NULL));
949 950
            goto cleanup;
        }
951 952 953
    }
#endif

954
#if HAVE_POLKIT0
955 956 957
    if (auth_unix_rw == REMOTE_AUTH_POLKIT ||
        auth_unix_ro == REMOTE_AUTH_POLKIT) {
        DBusError derr;
958 959 960 961

        dbus_connection_set_change_sigpipe(FALSE);
        dbus_threads_init_default();

962 963 964
        dbus_error_init(&derr);
        server->sysbus = dbus_bus_get(DBUS_BUS_SYSTEM, &derr);
        if (!(server->sysbus)) {
965 966
            VIR_ERROR(_("Failed to connect to system bus for PolicyKit auth: %s"),
                      derr.message);
967 968 969
            dbus_error_free(&derr);
            goto cleanup;
        }
970
        dbus_connection_set_exit_on_disconnect(server->sysbus, FALSE);
971 972 973
    }
#endif

974
    if (ipsock) {
975
        if (listen_tcp && remoteListenTCP (server, listen_addr, tcp_port, QEMUD_SOCK_TYPE_TCP, auth_tcp) < 0)
976 977 978 979 980 981
            goto cleanup;

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

982
            if (remoteListenTCP (server, listen_addr, tls_port, QEMUD_SOCK_TYPE_TLS, auth_tls) < 0)
983 984
                goto cleanup;
        }
D
Daniel P. Berrange 已提交
985 986
    }

987
#ifdef HAVE_AVAHI
988
    if (server->privileged && mdns_adv) {
989
        struct libvirtd_mdns_group *group;
990
        struct qemud_socket *sock;
991
        int port = 0;
992
        int ret;
993 994 995 996

        server->mdns = libvirtd_mdns_new();

        if (!mdns_name) {
997 998
            char *groupname, *localhost, *tmp;

999 1000
            localhost = virGetHostname(NULL);
            if (localhost == NULL)
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
                /* 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();
1016
                goto cleanup;
1017
            }
1018
            group = libvirtd_mdns_add_group(server->mdns, groupname);
1019
            VIR_FREE(groupname);
1020 1021 1022 1023
        } else {
            group = libvirtd_mdns_add_group(server->mdns, mdns_name);
        }

1024
        /*
1025 1026 1027 1028 1029 1030
         * 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) {
1031 1032 1033
            if (virSocketGetPort(&sock->addr) != -1 &&
                sock->type == QEMUD_SOCK_TYPE_TLS) {
                port = virSocketGetPort(&sock->addr);
1034 1035 1036 1037
                break;
            }
            sock = sock->next;
        }
1038

1039 1040 1041 1042 1043 1044 1045 1046 1047
        /*
         * 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

1048
    return 0;
D
Daniel P. Berrange 已提交
1049 1050

 cleanup:
1051 1052
    VIR_FREE(sockname);
    VIR_FREE(roSockname);
1053
    return -1;
D
Daniel P. Berrange 已提交
1054 1055
}

1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
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;
}
1075

1076

1077 1078 1079
static gnutls_session_t
remoteInitializeTLSSession (void)
{
1080 1081
    gnutls_session_t session;
    int err;
1082

1083 1084
    err = gnutls_init (&session, GNUTLS_SERVER);
    if (err != 0) goto failed;
1085

1086 1087 1088 1089 1090
    /* avoid calling all the priority functions, since the defaults
     * are adequate.
     */
    err = gnutls_set_default_priority (session);
    if (err != 0) goto failed;
1091

1092 1093
    err = gnutls_credentials_set (session, GNUTLS_CRD_CERTIFICATE, x509_cred);
    if (err != 0) goto failed;
1094

1095 1096 1097
    /* request client certificate if any.
     */
    gnutls_certificate_server_set_request (session, GNUTLS_CERT_REQUEST);
1098

1099
    gnutls_dh_set_prime_bits (session, DH_BITS);
1100

1101
    return session;
1102

1103 1104 1105 1106
failed:
    VIR_ERROR(_("remoteInitializeTLSSession: %s"),
              gnutls_strerror (err));
    return NULL;
1107 1108 1109 1110
}

/* Check DN is on tls_allowed_dn_list. */
static int
1111
remoteCheckDN (const char *dname)
1112
{
1113
    char **wildcards;
1114 1115 1116 1117 1118 1119 1120

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

    while (*wildcards) {
1121
        if (fnmatch (*wildcards, dname, 0) == 0)
1122 1123 1124 1125 1126
            return 1;
        wildcards++;
    }

    /* Print the client's DN. */
1127
    DEBUG(_("remoteCheckDN: failed: client DN is %s"), dname);
1128 1129 1130 1131 1132

    return 0; // Not found.
}

static int
1133
remoteCheckCertificate(struct qemud_client *client)
1134 1135 1136 1137 1138 1139
{
    int ret;
    unsigned int status;
    const gnutls_datum_t *certs;
    unsigned int nCerts, i;
    time_t now;
1140 1141 1142 1143
    char name[256];
    size_t namesize = sizeof name;

    memset(name, 0, namesize);
1144

1145 1146
    if ((ret = gnutls_certificate_verify_peers2 (client->tlssession, &status)) < 0){
        VIR_ERROR(_("Failed to verify certificate peers: %s"),
1147
                  gnutls_strerror (ret));
1148
        goto authdeny;
1149 1150 1151 1152
    }

    if (status != 0) {
        if (status & GNUTLS_CERT_INVALID)
1153
            VIR_ERROR0(_("The client certificate is not trusted."));
1154 1155

        if (status & GNUTLS_CERT_SIGNER_NOT_FOUND)
1156
            VIR_ERROR0(_("The client certificate has unknown issuer."));
1157 1158

        if (status & GNUTLS_CERT_REVOKED)
1159
            VIR_ERROR0(_("The client certificate has been revoked."));
1160

1161
#ifndef GNUTLS_1_0_COMPAT
1162
        if (status & GNUTLS_CERT_INSECURE_ALGORITHM)
1163
            VIR_ERROR0(_("The client certificate uses an insecure algorithm."));
1164
#endif
1165

1166
        goto authdeny;
1167 1168
    }

1169 1170 1171
    if (gnutls_certificate_type_get(client->tlssession) != GNUTLS_CRT_X509) {
        VIR_ERROR0(_("Only x509 certificates are supported"));
        goto authdeny;
1172 1173
    }

1174 1175 1176
    if (!(certs = gnutls_certificate_get_peers(client->tlssession, &nCerts))) {
        VIR_ERROR0(_("The certificate has no peers"));
        goto authdeny;
1177 1178 1179 1180 1181 1182 1183 1184
    }

    now = time (NULL);

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

        if (gnutls_x509_crt_init (&cert) < 0) {
1185 1186
            VIR_ERROR0(_("Unable to initialize certificate"));
            goto authfail;
1187 1188 1189
        }

        if (gnutls_x509_crt_import(cert, &certs[i], GNUTLS_X509_FMT_DER) < 0) {
1190
            VIR_ERROR0(_("Unable to load certificate"));
1191
            gnutls_x509_crt_deinit (cert);
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
            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;
            }
1214
        }
1215

1216
        if (gnutls_x509_crt_get_expiration_time (cert) < now) {
1217
            VIR_ERROR0(_("The client certificate has expired"));
1218
            gnutls_x509_crt_deinit (cert);
1219
            goto authdeny;
1220
        }
1221

1222
        if (gnutls_x509_crt_get_activation_time (cert) > now) {
1223
            VIR_ERROR0(_("The client certificate is not yet active"));
1224
            gnutls_x509_crt_deinit (cert);
1225
            goto authdeny;
1226 1227 1228
        }
    }

M
Matthias Bolte 已提交
1229
    PROBE(CLIENT_TLS_ALLOW, "fd=%d, name=%s", client->fd, (char *)name);
1230
    return 0;
1231 1232

authdeny:
M
Matthias Bolte 已提交
1233
    PROBE(CLIENT_TLS_DENY, "fd=%d, name=%s", client->fd, (char *)name);
1234 1235 1236
    return -1;

authfail:
1237
    PROBE(CLIENT_TLS_FAIL, "fd=%d", client->fd);
1238
    return -1;
1239 1240 1241 1242 1243 1244
}

/* Check the client's access. */
static int
remoteCheckAccess (struct qemud_client *client)
{
1245 1246
    struct qemud_client_message *confirm;

1247
    /* Verify client certificate. */
1248
    if (remoteCheckCertificate (client) == -1) {
1249 1250
        VIR_ERROR0(_("remoteCheckCertificate: "
                     "failed to verify client's certificate"));
1251
        if (!tls_no_verify_certificate) return -1;
1252 1253
        else VIR_INFO0(_("remoteCheckCertificate: tls_no_verify_certificate "
                          "is set so the bad certificate is ignored"));
1254 1255
    }

1256 1257 1258 1259 1260 1261 1262 1263 1264
    if (client->tx) {
        VIR_INFO("%s",
                 _("client had unexpected data pending tx after access check"));
        return -1;
    }

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

1265 1266 1267 1268
    /* 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).
     */
1269 1270 1271 1272 1273 1274
    confirm->async = 1;
    confirm->bufferLength = 1;
    confirm->bufferOffset = 0;
    confirm->buffer[0] = '\1';

    client->tx = confirm;
1275 1276
    return 0;
}
D
Daniel P. Berrange 已提交
1277

1278 1279
#if HAVE_POLKIT
int qemudGetSocketIdentity(int fd, uid_t *uid, pid_t *pid) {
1280
# ifdef SO_PEERCRED
1281 1282 1283 1284
    struct ucred cr;
    unsigned int cr_len = sizeof (cr);

    if (getsockopt (fd, SOL_SOCKET, SO_PEERCRED, &cr, &cr_len) < 0) {
1285
        char ebuf[1024];
1286
        VIR_ERROR(_("Failed to verify client credentials: %s"),
1287
                  virStrerror(errno, ebuf, sizeof ebuf));
1288 1289 1290 1291 1292
        return -1;
    }

    *pid = cr.pid;
    *uid = cr.uid;
1293
# else
1294
    /* XXX Many more OS support UNIX socket credentials we could port to. See dbus ....*/
1295 1296
#  error "UNIX socket credentials not supported/implemented on this platform yet..."
# endif
1297 1298 1299 1300
    return 0;
}
#endif

1301

D
Daniel P. Berrange 已提交
1302 1303
static int qemudDispatchServer(struct qemud_server *server, struct qemud_socket *sock) {
    int fd;
1304 1305
    virSocketAddr addr;
    char *addrstr = NULL;
1306
    struct qemud_client *client = NULL;
1307
    int no_slow_start = 1;
1308
    int i;
D
Daniel P. Berrange 已提交
1309

1310 1311
    addr.len = sizeof(addr.data.stor);
    if ((fd = accept(sock->fd, &addr.data.sa, &addr.len)) < 0) {
1312
        char ebuf[1024];
D
Daniel P. Berrange 已提交
1313 1314
        if (errno == EAGAIN)
            return 0;
1315 1316
        VIR_ERROR(_("Failed to accept connection: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
1317 1318
        return -1;
    }
1319 1320 1321 1322
    if (!(addrstr = virSocketFormatAddrFull(&addr, true, ";"))) {
        VIR_ERROR0(_("Failed to format addresss: out of memory"));
        goto error;
    }
D
Daniel P. Berrange 已提交
1323

1324 1325
    PROBE(CLIENT_CONNECT, "fd=%d, readonly=%d localAddr=%s remoteAddr=%s",
          fd, sock->readonly, sock->addrstr, addrstr);
1326

1327
    if (server->nclients >= max_clients) {
1328 1329
        VIR_ERROR(_("Too many active clients (%d), dropping connection from %s"),
                  max_clients, addrstr);
1330
        goto error;
1331 1332
    }

1333
    if (VIR_REALLOC_N(server->clients, server->nclients+1) < 0) {
1334
        VIR_ERROR0(_("Out of memory allocating clients"));
1335
        goto error;
1336 1337
    }

J
John Levon 已提交
1338 1339 1340 1341 1342 1343 1344 1345 1346
#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);
1347
            goto error;
J
John Levon 已提交
1348 1349 1350 1351
        }

        if (!priv_ismember (privs, PRIV_VIRT_MANAGE)) {
            ucred_free (ucred);
1352
            goto error;
J
John Levon 已提交
1353 1354 1355 1356 1357 1358
        }

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

1359 1360 1361 1362
    /* Disable Nagle.  Unix sockets will ignore this. */
    setsockopt (fd, IPPROTO_TCP, TCP_NODELAY, (void *)&no_slow_start,
                sizeof no_slow_start);

1363 1364
    if (virSetCloseExec(fd) < 0 ||
        virSetNonBlock(fd) < 0) {
1365
        goto error;
D
Daniel P. Berrange 已提交
1366 1367
    }

1368
    if (VIR_ALLOC(client) < 0)
1369
        goto error;
1370
    if (virMutexInit(&client->lock) < 0) {
1371
        VIR_ERROR0(_("cannot initialize mutex"));
1372
        goto error;
1373
    }
1374

1375
    client->magic = QEMUD_CLIENT_MAGIC;
D
Daniel P. Berrange 已提交
1376 1377
    client->fd = fd;
    client->readonly = sock->readonly;
1378
    client->type = sock->type;
1379
    client->auth = sock->auth;
1380 1381 1382
    client->addr = addr;
    client->addrstr = addrstr;
    addrstr = NULL;
1383

1384 1385 1386 1387
    for (i = 0 ; i < VIR_DOMAIN_EVENT_ID_LAST ; i++) {
        client->domainEventCallbackID[i] = -1;
    }

1388 1389
    /* Prepare one for packet receive */
    if (VIR_ALLOC(client->rx) < 0)
1390
        goto error;
1391 1392 1393
    client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;


1394 1395 1396 1397 1398 1399 1400 1401 1402 1403
#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)
1404
            goto error;
1405

1406
        /* Client is running as root, so disable auth */
1407
        if (uid == 0) {
1408 1409
            VIR_INFO(_("Turn off polkit auth for privileged client pid %d from %s"),
                     pid, addrstr);
1410 1411 1412 1413 1414
            client->auth = REMOTE_AUTH_NONE;
        }
    }
#endif

1415
    if (client->type != QEMUD_SOCK_TYPE_TLS) {
1416
        /* Plain socket, so prepare to read first message */
1417
        if (qemudRegisterClientEvent (server, client) < 0)
1418
            goto error;
1419 1420 1421
    } else {
        int ret;

1422 1423
        client->tlssession = remoteInitializeTLSSession ();
        if (client->tlssession == NULL)
1424
            goto error;
1425

1426
        gnutls_transport_set_ptr (client->tlssession,
1427 1428 1429
                                  (gnutls_transport_ptr_t) (long) fd);

        /* Begin the TLS handshake. */
1430
        ret = gnutls_handshake (client->tlssession);
1431
        if (ret == 0) {
1432 1433
            client->handshake = 0;

1434 1435
            /* Unlikely, but ...  Next step is to check the certificate. */
            if (remoteCheckAccess (client) == -1)
1436
                goto error;
D
Daniel P. Berrange 已提交
1437

1438
            /* Handshake & cert check OK,  so prepare to read first message */
1439
            if (qemudRegisterClientEvent(server, client) < 0)
1440
                goto error;
1441
        } else if (ret == GNUTLS_E_INTERRUPTED || ret == GNUTLS_E_AGAIN) {
1442 1443
            /* Most likely, need to do more handshake data */
            client->handshake = 1;
D
Daniel P. Berrange 已提交
1444

1445
            if (qemudRegisterClientEvent (server, client) < 0)
1446
                goto error;
1447
        } else {
1448
            PROBE(CLIENT_TLS_FAIL, "fd=%d", client->fd);
1449 1450
            VIR_ERROR(_("TLS handshake failed for client %s: %s"),
                      addrstr, gnutls_strerror (ret));
1451
            goto error;
1452 1453
        }
    }
D
Daniel P. Berrange 已提交
1454

1455
    server->clients[server->nclients++] = client;
D
Daniel P. Berrange 已提交
1456

1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469
    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
1470
    return 0;
1471

1472 1473 1474
error:
    if (client) {
        if (client->tlssession) gnutls_deinit (client->tlssession);
1475 1476
        if (client) {
            VIR_FREE(client->addrstr);
1477
            VIR_FREE(client->rx);
1478
        }
1479 1480
        VIR_FREE(client);
    }
1481
    VIR_FREE(addrstr);
1482
    VIR_FORCE_CLOSE(fd);
1483
    PROBE(CLIENT_DISCONNECT, "fd=%d", fd);
1484
    return -1;
D
Daniel P. Berrange 已提交
1485 1486 1487
}


1488 1489 1490 1491 1492 1493 1494
/*
 * 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
 */
1495
void qemudDispatchClientFailure(struct qemud_client *client) {
1496 1497 1498 1499
    if (client->watch != -1) {
        virEventRemoveHandleImpl(client->watch);
        client->watch = -1;
    }
D
Daniel P. Berrange 已提交
1500

1501
    /* Deregister event delivery callback */
1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512
    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;
        }
1513 1514
    }

1515
#if HAVE_SASL
1516 1517 1518 1519
    if (client->saslconn) {
        sasl_dispose(&client->saslconn);
        client->saslconn = NULL;
    }
1520
    VIR_FREE(client->saslUsername);
1521
#endif
1522 1523 1524 1525 1526
    if (client->tlssession) {
        gnutls_deinit (client->tlssession);
        client->tlssession = NULL;
    }
    if (client->fd != -1) {
1527
        PROBE(CLIENT_DISCONNECT, "fd=%d", client->fd);
1528
        VIR_FORCE_CLOSE(client->fd);
1529
    }
1530
    VIR_FREE(client->addrstr);
1531 1532 1533 1534 1535 1536 1537 1538
}


/* Caller must hold server lock */
static struct qemud_client *qemudPendingJob(struct qemud_server *server)
{
    int i;
    for (i = 0 ; i < server->nclients ; i++) {
1539
        virMutexLock(&server->clients[i]->lock);
1540
        if (server->clients[i]->dx) {
1541 1542 1543
            /* Delibrately don't unlock client - caller wants the lock */
            return server->clients[i];
        }
1544
        virMutexUnlock(&server->clients[i]->lock);
1545 1546
    }
    return NULL;
D
Daniel P. Berrange 已提交
1547 1548
}

1549 1550
static void *qemudWorker(void *data)
{
1551 1552
    struct qemud_worker *worker = data;
    struct qemud_server *server = worker->server;
1553 1554

    while (1) {
1555
        struct qemud_client *client = NULL;
1556
        struct qemud_client_message *msg;
1557

1558
        virMutexLock(&server->lock);
1559 1560 1561
        while ((client = qemudPendingJob(server)) == NULL) {
            if (worker->quitRequest ||
                virCondWait(&server->job, &server->lock) < 0) {
1562 1563 1564 1565
                virMutexUnlock(&server->lock);
                return NULL;
            }
        }
1566
        if (worker->quitRequest) {
1567
            virMutexUnlock(&client->lock);
1568 1569 1570 1571
            virMutexUnlock(&server->lock);
            return NULL;
        }
        worker->processingCall = 1;
1572
        virMutexUnlock(&server->lock);
1573 1574 1575 1576

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

1577
        /* Remove our message from dispatch queue while we use it */
1578
        msg = qemudClientMessageQueueServe(&client->dx);
1579 1580 1581

        /* This function drops the lock during dispatch,
         * and re-acquires it before returning */
1582
        if (remoteDispatchClientRequest (server, client, msg) < 0) {
1583
            VIR_FREE(msg);
1584 1585 1586 1587 1588
            qemudDispatchClientFailure(client);
            client->refs--;
            virMutexUnlock(&client->lock);
            continue;
        }
1589 1590

        client->refs--;
1591
        virMutexUnlock(&client->lock);
1592 1593 1594 1595

        virMutexLock(&server->lock);
        worker->processingCall = 0;
        virMutexUnlock(&server->lock);
1596 1597
    }
}
D
Daniel P. Berrange 已提交
1598

1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625
static int qemudStartWorker(struct qemud_server *server,
                            struct qemud_worker *worker) {
    pthread_attr_t attr;
    pthread_attr_init(&attr);
    /* We want to join workers, so don't detach them */
    /*pthread_attr_setdetachstate(&attr, 1);*/

    if (worker->hasThread)
        return -1;

    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;
        return -1;
    }

    return 0;
}

D
Daniel P. Berrange 已提交
1626

1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
/*
 * 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) {
1640 1641
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1642 1643 1644
        qemudDispatchClientFailure(client);
        return -1;
    }
1645 1646 1647

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

1648
    if (!client->tlssession) {
1649
        char ebuf[1024];
1650 1651 1652 1653 1654 1655
        ret = read (client->fd, data, len);
        if (ret == -1 && (errno == EAGAIN ||
                          errno == EINTR))
            return 0;
        if (ret <= 0) {
            if (ret != 0)
1656 1657
                VIR_ERROR(_("read: %s"),
                          virStrerror (errno, ebuf, sizeof ebuf));
1658
            qemudDispatchClientFailure(client);
1659 1660 1661
            return -1;
        }
    } else {
1662
        ret = gnutls_record_recv (client->tlssession, data, len);
1663 1664 1665 1666 1667 1668 1669 1670 1671

        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);
1672 1673
            return -1;
        }
D
Daniel P. Berrange 已提交
1674
    }
1675

1676 1677 1678
    return ret;
}

1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696
/*
 * 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
1697 1698
}

1699
#if HAVE_SASL
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
/*
 * 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;
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719

    /* 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) {
1720
        int ret;
1721
        char encoded[8192];
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
        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);
1734
            return -1;
1735
        }
1736 1737 1738 1739 1740 1741

        client->saslDecodedOffset = 0;
    }

    /* Some buffered decoded data to return now */
    got = client->saslDecodedLength - client->saslDecodedOffset;
1742
    want = client->rx->bufferLength - client->rx->bufferOffset;
1743 1744 1745 1746

    if (want > got)
        want = got;

1747
    memcpy(client->rx->buffer + client->rx->bufferOffset,
1748 1749
           client->saslDecoded + client->saslDecodedOffset, want);
    client->saslDecodedOffset += want;
1750
    client->rx->bufferOffset += want;
1751 1752 1753 1754 1755 1756

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

1757
    return want;
1758 1759 1760
}
#endif

1761 1762 1763 1764 1765
/*
 * 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) {
1766 1767
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_READ)
1768
        return qemudClientReadSASL(client);
1769 1770
    else
#endif
1771
        return qemudClientReadPlain(client);
1772 1773 1774
}


1775 1776 1777 1778 1779
/*
 * Read data until we get a complete message to process
 */
static void qemudDispatchClientRead(struct qemud_server *server,
                                    struct qemud_client *client) {
1780
    /*qemudDebug ("qemudDispatchClientRead: mode = %d", client->mode);*/
D
Daniel P. Berrange 已提交
1781

1782 1783 1784
readmore:
    if (qemudClientRead(client) < 0)
        return; /* Error */
1785

1786 1787
    if (client->rx->bufferOffset < client->rx->bufferLength)
        return; /* Still not read enough */
1788

1789 1790 1791 1792
    /* Either done with length word header */
    if (client->rx->bufferLength == REMOTE_MESSAGE_HEADER_XDR_LEN) {
        unsigned int len;
        XDR x;
1793

1794
        xdrmem_create(&x, client->rx->buffer, client->rx->bufferLength, XDR_DECODE);
1795

1796 1797
        if (!xdr_u_int(&x, &len)) {
            xdr_destroy (&x);
1798
            DEBUG0("Failed to decode packet length");
1799
            qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1800 1801
            return;
        }
1802
        xdr_destroy (&x);
1803

1804 1805 1806
        if (len < REMOTE_MESSAGE_HEADER_XDR_LEN) {
            DEBUG("Packet length %u too small", len);
            qemudDispatchClientFailure(client);
1807
            return;
D
Daniel P. Berrange 已提交
1808
        }
1809

1810 1811 1812 1813 1814 1815
        /* 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);
1816 1817 1818
            return;
        }

1819 1820
        /* Prepare to read rest of message */
        client->rx->bufferLength += len;
1821

1822
        qemudUpdateClientEvent(client);
D
Daniel P. Berrange 已提交
1823

1824 1825 1826 1827 1828
        /* Try and read payload immediately instead of going back
           into poll() because chances are the data is already
           waiting for us */
        goto readmore;
    } else {
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
        /* 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) {
1842 1843 1844
            int ret;
            ret = (filter->query)(client, msg, filter->opaque);
            if (ret == 1) {
1845 1846
                msg = NULL;
                break;
1847 1848 1849 1850
            } else if (ret == -1) {
                VIR_FREE(msg);
                qemudDispatchClientFailure(client);
                return;
1851 1852 1853 1854
            }
            filter = filter->next;
        }

1855
        /* Move completed message to the end of the dispatch queue */
1856 1857
        if (msg)
            qemudClientMessageQueuePush(&client->dx, msg);
1858 1859 1860 1861 1862 1863
        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
1864
        } else {
1865 1866 1867
            if (client->rx)
                client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;

1868 1869 1870 1871
            qemudUpdateClientEvent(client);

            /* Tell one of the workers to get on with it... */
            virCondSignal(&server->job);
D
Daniel P. Berrange 已提交
1872
        }
D
Daniel P. Berrange 已提交
1873 1874 1875 1876
    }
}


1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
/*
 * 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) {
1890 1891
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1892 1893 1894 1895
        qemudDispatchClientFailure(client);
        return -1;
    }

1896
    if (!client->tlssession) {
1897
        char ebuf[1024];
1898 1899 1900
        if ((ret = write(client->fd, data, len)) == -1) {
            if (errno == EAGAIN || errno == EINTR)
                return 0;
1901
            VIR_ERROR(_("write: %s"), virStrerror (errno, ebuf, sizeof ebuf));
1902
            qemudDispatchClientFailure(client);
1903 1904 1905
            return -1;
        }
    } else {
1906
        ret = gnutls_record_send (client->tlssession, data, len);
1907 1908 1909 1910 1911 1912 1913
        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);
1914 1915
            return -1;
        }
D
Daniel P. Berrange 已提交
1916
    }
1917 1918
    return ret;
}
1919

1920

1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
/*
 * 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
1937 1938 1939
}


1940
#if HAVE_SASL
1941 1942 1943 1944 1945 1946 1947 1948 1949
/*
 * Send client->tx using SASL encoding
 *
 * Returns:
 *   -1 on error
 *    0 on EAGAIN
 *    n number of bytes
 */
static int qemudClientWriteSASL(struct qemud_client *client) {
1950 1951 1952 1953
    int ret;

    /* Not got any pending encoded data, so we need to encode raw stuff */
    if (client->saslEncoded == NULL) {
1954 1955 1956
        ret = sasl_encode(client->saslconn,
                          client->tx->buffer + client->tx->bufferOffset,
                          client->tx->bufferLength - client->tx->bufferOffset,
1957 1958 1959
                          &client->saslEncoded,
                          &client->saslEncodedLength);

1960 1961 1962 1963 1964 1965 1966
        if (ret != SASL_OK) {
            VIR_ERROR(_("failed to encode SASL data %s"),
                      sasl_errstring(ret, NULL, NULL));
            qemudDispatchClientFailure(client);
            return -1;
        }

1967 1968 1969 1970
        client->saslEncodedOffset = 0;
    }

    /* Send some of the encoded stuff out on the wire */
1971
    ret = qemudClientWriteBuf(client,
1972 1973 1974
                              client->saslEncoded + client->saslEncodedOffset,
                              client->saslEncodedLength - client->saslEncodedOffset);

1975 1976
    if (ret <= 0)
        return ret; /* -1 error, 0 == egain */
1977 1978 1979 1980 1981 1982 1983 1984

    /* 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;
1985 1986 1987

        /* Mark as complete, so caller detects completion */
        client->tx->bufferOffset = client->tx->bufferLength;
1988 1989
    }

1990
    return ret;
1991 1992 1993
}
#endif

1994 1995 1996 1997 1998 1999 2000 2001 2002
/*
 * 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) {
2003 2004
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_WRITE)
2005
        return qemudClientWriteSASL(client);
2006 2007
    else
#endif
2008
        return qemudClientWritePlain(client);
2009 2010 2011
}


2012 2013 2014 2015
void
qemudClientMessageRelease(struct qemud_client *client,
                          struct qemud_client_message *msg)
{
2016 2017 2018
    if (msg->streamTX) {
        remoteStreamMessageFinished(client, msg);
    } else if (!msg->async)
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
        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);
}


2037 2038 2039 2040 2041
/*
 * Process all queued client->tx messages until
 * we would block on I/O
 */
static void
2042
qemudDispatchClientWrite(struct qemud_client *client) {
2043 2044
    while (client->tx) {
        ssize_t ret;
2045

2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
        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);

2060
            qemudClientMessageRelease(client, reply);
2061

2062 2063
            if (client->closing)
                qemudDispatchClientFailure(client);
2064
         }
2065
    }
2066
}
2067

2068
static void
2069
qemudDispatchClientHandshake(struct qemud_client *client) {
2070 2071 2072 2073
    int ret;
    /* Continue the handshake. */
    ret = gnutls_handshake (client->tlssession);
    if (ret == 0) {
2074 2075
        client->handshake = 0;

2076 2077 2078
        /* Finished.  Next step is to check the certificate. */
        if (remoteCheckAccess (client) == -1)
            qemudDispatchClientFailure(client);
2079 2080
        else
            qemudUpdateClientEvent(client);
2081 2082 2083 2084 2085
    } 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 */
2086
        qemudUpdateClientEvent (client);
2087
    } else {
2088
        PROBE(CLIENT_TLS_FAIL, "fd=%d", client->fd);
2089 2090 2091 2092
        /* Fatal error in handshake */
        VIR_ERROR(_("TLS handshake failed: %s"),
                  gnutls_strerror (ret));
        qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
2093 2094 2095
    }
}

2096
static void
2097
qemudDispatchClientEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
2098
    struct qemud_server *server = (struct qemud_server *)opaque;
2099 2100
    struct qemud_client *client = NULL;
    int i;
2101

2102
    virMutexLock(&server->lock);
2103

2104
    for (i = 0 ; i < server->nclients ; i++) {
2105
        virMutexLock(&server->clients[i]->lock);
2106 2107
        if (server->clients[i]->watch == watch) {
            client = server->clients[i];
D
Daniel P. Berrange 已提交
2108
            break;
2109
        }
2110
        virMutexUnlock(&server->clients[i]->lock);
D
Daniel P. Berrange 已提交
2111
    }
2112

2113 2114
    virMutexUnlock(&server->lock);

2115
    if (!client) {
D
Daniel P. Berrange 已提交
2116
        return;
2117 2118
    }

2119 2120
    if (client->fd != fd) {
        virMutexUnlock(&client->lock);
2121
        return;
2122 2123 2124 2125 2126
    }

    if (events & (VIR_EVENT_HANDLE_WRITABLE |
                  VIR_EVENT_HANDLE_READABLE)) {
        if (client->handshake) {
2127
            qemudDispatchClientHandshake(client);
2128 2129
        } else {
            if (events & VIR_EVENT_HANDLE_WRITABLE)
2130
                qemudDispatchClientWrite(client);
2131 2132 2133 2134 2135 2136 2137 2138 2139 2140
            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);
2141

2142
    virMutexUnlock(&client->lock);
D
Daniel P. Berrange 已提交
2143 2144
}

2145 2146 2147 2148 2149 2150

/*
 * @client: a locked client object
 */
static int
qemudCalculateHandleMode(struct qemud_client *client) {
2151 2152 2153
    int mode = 0;

    if (client->handshake) {
2154
        if (gnutls_record_get_direction (client->tlssession) == 0)
2155
            mode |= VIR_EVENT_HANDLE_READABLE;
2156
        else
2157 2158 2159 2160 2161 2162
            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;
2163

2164 2165 2166 2167
        /* 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;
2168 2169
    }

2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
    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
2188 2189 2190 2191

    return 0;
}

2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203
/*
 * @client: a locked client object
 */
void qemudUpdateClientEvent(struct qemud_client *client) {
    int mode;

    mode = qemudCalculateHandleMode(client);

    virEventUpdateHandleImpl(client->watch, mode);
}


2204
static void
2205
qemudDispatchServerEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
2206
    struct qemud_server *server = (struct qemud_server *)opaque;
2207 2208
    struct qemud_socket *sock;

2209
    virMutexLock(&server->lock);
2210 2211

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2212

D
Daniel P. Berrange 已提交
2213
    while (sock) {
2214
        if (sock->watch == watch)
D
Daniel P. Berrange 已提交
2215 2216 2217
            break;

        sock = sock->next;
D
Daniel P. Berrange 已提交
2218
    }
D
Daniel P. Berrange 已提交
2219

2220
    if (sock && sock->fd == fd && events)
D
Daniel P. Berrange 已提交
2221
        qemudDispatchServer(server, sock);
2222

2223
    virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
2224 2225 2226
}


2227
static int qemudOneLoop(void) {
2228
    sig_atomic_t errors;
D
Daniel P. Berrange 已提交
2229

D
Daniel P. Berrange 已提交
2230
    if (virEventRunOnce() < 0)
D
Daniel P. Berrange 已提交
2231 2232
        return -1;

2233 2234 2235
    /* Check for any signal handling errors and log them. */
    errors = sig_errors;
    if (errors) {
2236
        char ebuf[1024];
2237
        sig_errors -= errors;
2238
        VIR_ERROR(_("Signal handler reported %d errors: last error: %s"),
2239
                  errors, virStrerror (sig_lasterrno, ebuf, sizeof ebuf));
2240 2241 2242
        return -1;
    }

D
Daniel P. Berrange 已提交
2243 2244 2245
    return 0;
}

2246
static void qemudInactiveTimer(int timerid, void *data) {
2247
    struct qemud_server *server = (struct qemud_server *)data;
2248 2249 2250 2251 2252 2253 2254

    if (virStateActive() ||
        server->clients) {
        DEBUG0("Timer expired but still active, not shutting down");
        virEventUpdateTimeoutImpl(timerid, -1);
    } else {
        DEBUG0("Timer expired and inactive, shutting down");
2255
        server->quitEventThread = 1;
2256 2257 2258
    }
}

2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275
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);
    }

2276 2277 2278
    while (client->streams)
        remoteRemoveClientStream(client, client->streams);

2279 2280 2281 2282 2283 2284
    if (client->conn)
        virConnectClose(client->conn);
    virMutexDestroy(&client->lock);
    VIR_FREE(client);
}

2285 2286
static void *qemudRunLoop(void *opaque) {
    struct qemud_server *server = opaque;
2287
    int timerid = -1;
2288
    int i;
2289
    int timerActive = 0;
2290

2291
    virMutexLock(&server->lock);
2292

2293 2294 2295 2296 2297
    if (timeout > 0 &&
        (timerid = virEventAddTimeoutImpl(-1,
                                          qemudInactiveTimer,
                                          server, NULL)) < 0) {
        VIR_ERROR0(_("Failed to register shutdown timeout"));
2298
        return NULL;
2299 2300
    }

2301 2302 2303 2304
    if (min_workers > max_workers)
        max_workers = min_workers;

    server->nworkers = max_workers;
2305
    if (VIR_ALLOC_N(server->workers, server->nworkers) < 0) {
2306
        VIR_ERROR0(_("Failed to allocate workers"));
2307
        return NULL;
2308 2309
    }

2310 2311 2312 2313
    for (i = 0 ; i < min_workers ; i++) {
        if (qemudStartWorker(server, &server->workers[i]) < 0)
            goto cleanup;
        server->nactiveworkers++;
2314
    }
2315

2316
    for (;!server->quitEventThread;) {
2317 2318 2319 2320
        /* A shutdown timeout is specified, so check
         * if any drivers have active state, if not
         * shutdown after timeout seconds
         */
2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
        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;
                }
            }
2336 2337
        }

2338
        virMutexUnlock(&server->lock);
2339 2340
        if (qemudOneLoop() < 0) {
            virMutexLock(&server->lock);
2341
            DEBUG0("Loop iteration error, exiting");
2342
            break;
2343
        }
2344
        virMutexLock(&server->lock);
2345 2346 2347 2348

    reprocess:
        for (i = 0 ; i < server->nclients ; i++) {
            int inactive;
2349
            virMutexLock(&server->clients[i]->lock);
2350 2351
            inactive = server->clients[i]->fd == -1
                && server->clients[i]->refs == 0;
2352
            virMutexUnlock(&server->clients[i]->lock);
2353
            if (inactive) {
2354
                qemudFreeClient(server->clients[i]);
2355
                server->nclients--;
2356
                if (i < server->nclients)
2357 2358
                    memmove(server->clients + i,
                            server->clients + i + 1,
2359 2360 2361 2362 2363
                            sizeof (*server->clients) * (server->nclients - i));

                if (VIR_REALLOC_N(server->clients,
                                  server->nclients) < 0) {
                    ; /* ignore */
2364
                }
2365
                goto reprocess;
2366 2367
            }
        }
D
Daniel P. Berrange 已提交
2368

2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
        /* 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--;
            }
        }
2388 2389
    }

2390
cleanup:
2391
    for (i = 0 ; i < server->nworkers ; i++) {
2392 2393 2394 2395 2396 2397
        if (!server->workers[i].hasThread)
            continue;

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

2398
        virMutexUnlock(&server->lock);
2399
        pthread_join(server->workers[i].thread, NULL);
2400
        virMutexLock(&server->lock);
2401
        server->workers[i].hasThread = 0;
2402
    }
2403
    VIR_FREE(server->workers);
2404

2405
    virMutexUnlock(&server->lock);
2406
    return NULL;
D
Daniel P. Berrange 已提交
2407 2408
}

2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427

static int qemudStartEventLoop(struct qemud_server *server) {
    pthread_attr_t attr;
    pthread_attr_init(&attr);
    /* 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)
        return -1;

    server->hasEventThread = 1;

    return 0;
}


D
Daniel P. Berrange 已提交
2428
static void qemudCleanup(struct qemud_server *server) {
2429 2430
    struct qemud_socket *sock;

2431 2432
    VIR_FORCE_CLOSE(server->sigread);
    VIR_FORCE_CLOSE(server->sigwrite);
2433 2434

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2435
    while (sock) {
2436
        struct qemud_socket *next = sock->next;
2437 2438
        if (sock->watch)
            virEventRemoveHandleImpl(sock->watch);
2439
        VIR_FORCE_CLOSE(sock->fd);
2440 2441 2442 2443 2444 2445 2446

        /* Unlink unix domain sockets which are not in
         * the abstract namespace */
        if (sock->path &&
            sock->path[0] != '@')
            unlink(sock->path);
        VIR_FREE(sock->path);
2447
        VIR_FREE(sock->addrstr);
2448

2449
        VIR_FREE(sock);
2450
        sock = next;
D
Daniel P. Berrange 已提交
2451
    }
2452
    VIR_FREE(server->logDir);
2453

2454
#ifdef HAVE_SASL
2455 2456 2457
    if (server->saslUsernameWhitelist) {
        char **list = server->saslUsernameWhitelist;
        while (*list) {
2458
            VIR_FREE(*list);
2459 2460
            list++;
        }
2461
        VIR_FREE(server->saslUsernameWhitelist);
2462
    }
2463
#endif
2464

2465 2466 2467 2468 2469
#if HAVE_POLKIT0
        if (server->sysbus)
            dbus_connection_unref(server->sysbus);
#endif

2470
    virStateCleanup();
2471

2472 2473 2474 2475 2476 2477
    if (virCondDestroy(&server->job) < 0) {
        ;
    }
    virMutexDestroy(&server->lock);

    VIR_FREE(server);
D
Daniel P. Berrange 已提交
2478 2479
}

2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495
/* 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:
2496
        if (VIR_ALLOC_N(list, 2) < 0) {
2497
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2498 2499 2500 2501 2502
            return -1;
        }
        list[0] = strdup (p->str);
        list[1] = NULL;
        if (list[0] == NULL) {
2503
            VIR_ERROR(_("failed to allocate memory for %s config list value"),
2504
                      key);
2505
            VIR_FREE(list);
2506 2507 2508 2509 2510 2511 2512 2513 2514
            return -1;
        }
        break;

    case VIR_CONF_LIST: {
        int i, len = 0;
        virConfValuePtr pp;
        for (pp = p->list; pp; pp = pp->next)
            len++;
2515
        if (VIR_ALLOC_N(list, 1+len) < 0) {
2516
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2517 2518 2519 2520
            return -1;
        }
        for (i = 0, pp = p->list; pp; ++i, pp = pp->next) {
            if (pp->type != VIR_CONF_STRING) {
2521
                VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
2522
                            " must be a string or list of strings"),
2523
                          filename, key);
2524
                VIR_FREE(list);
2525 2526 2527 2528 2529 2530
                return -1;
            }
            list[i] = strdup (pp->str);
            if (list[i] == NULL) {
                int j;
                for (j = 0 ; j < i ; j++)
2531 2532
                    VIR_FREE(list[j]);
                VIR_FREE(list);
2533 2534
                VIR_ERROR(_("failed to allocate memory for %s config list value"),
                          key);
2535 2536 2537 2538 2539 2540 2541 2542 2543
                return -1;
            }

        }
        list[i] = NULL;
        break;
    }

    default:
2544
        VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
2545
                    " must be a string or list of strings"),
2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559
                  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) {
2560
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: invalid type:"
2561
                    " got %s; expected %s"), filename, key,
2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580
                  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) {                                   \
2581
                char ebuf[1024];                                        \
2582
                VIR_ERROR(_("remoteReadConfigFile: %s"),		\
2583
                          virStrerror(errno, ebuf, sizeof ebuf));       \
2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599
                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)

2600 2601 2602 2603 2604 2605 2606 2607

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

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

2608
    if (checkType (p, filename, key, VIR_CONF_STRING) < 0)
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
        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;
2619 2620 2621 2622
#endif
#if HAVE_POLKIT
    } else if (STREQ(p->str, "polkit")) {
        *auth = REMOTE_AUTH_POLKIT;
2623 2624
#endif
    } else {
2625
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: unsupported auth %s"),
2626
                  filename, key, p->str);
2627 2628 2629 2630 2631 2632
        return -1;
    }

    return 0;
}

2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652
#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

2653 2654 2655 2656 2657 2658
/*
 * 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.
 */
2659
static int
2660 2661 2662
qemudSetLogging(virConfPtr conf, const char *filename)
{
    int log_level = 0;
2663 2664
    char *log_filters = NULL;
    char *log_outputs = NULL;
2665 2666
    int ret = -1;

2667 2668
    virLogReset();

2669
    /*
2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681
     * 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.
2682
     */
2683
    GET_CONF_INT (conf, filename, log_level);
2684 2685 2686
    if (log_level != 0)
        virLogSetDefaultPriority(log_level);

2687
    virLogSetFromEnv();
2688 2689 2690 2691

    if (virLogGetNbFilters() == 0) {
        GET_CONF_STR (conf, filename, log_filters);
        virLogParseFilters(log_filters);
2692
    }
2693

2694 2695 2696
    if (virLogGetNbOutputs() == 0) {
        GET_CONF_STR (conf, filename, log_outputs);
        virLogParseOutputs(log_outputs);
2697
    }
2698 2699

    /*
2700 2701
     * If no defined outputs, then direct to syslog when running
     * as daemon. Otherwise the default output is stderr.
2702
     */
2703
    if (virLogGetNbOutputs() == 0) {
2704
        char *tmp = NULL;
2705
        if (godaemon) {
2706 2707
            if (virAsprintf (&tmp, "%d:syslog:libvirtd",
                             virLogGetDefaultPriority()) < 0)
2708 2709
                goto free_and_fail;
        } else {
2710 2711
            if (virAsprintf (&tmp, "%d:stderr",
                             virLogGetDefaultPriority()) < 0)
2712
                goto free_and_fail;
2713
        }
2714 2715
        virLogParseOutputs(tmp);
        VIR_FREE(tmp);
2716
    }
2717 2718 2719 2720 2721 2722 2723

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

2724 2725
    ret = 0;

2726 2727 2728
free_and_fail:
    VIR_FREE(log_filters);
    VIR_FREE(log_outputs);
2729
    return(ret);
2730
}
2731

2732 2733 2734 2735
/* Read the config file if it exists.
 * Only used in the remote case, hence the name.
 */
static int
2736
remoteReadConfigFile (struct qemud_server *server, const char *filename)
2737 2738 2739
{
    virConfPtr conf;

2740 2741 2742 2743 2744
    /* 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;
2745
    char *buf = NULL;
2746
    char *host_uuid = NULL;
2747

2748 2749
#if HAVE_POLKIT
    /* Change the default back to no auth for non-root */
2750
    if (!server->privileged && auth_unix_rw == REMOTE_AUTH_POLKIT)
2751
        auth_unix_rw = REMOTE_AUTH_NONE;
2752
    if (!server->privileged && auth_unix_ro == REMOTE_AUTH_POLKIT)
2753 2754 2755
        auth_unix_ro = REMOTE_AUTH_NONE;
#endif

2756
    conf = virConfReadFile (filename, 0);
2757
    if (!conf) return -1;
2758

2759 2760 2761
    /*
     * First get all the logging settings and activate them
     */
2762 2763
    if (qemudSetLogging(conf, filename) < 0)
        goto free_and_fail;
2764

2765 2766
    GET_CONF_INT (conf, filename, listen_tcp);
    GET_CONF_INT (conf, filename, listen_tls);
2767 2768
    GET_CONF_STR (conf, filename, tls_port);
    GET_CONF_STR (conf, filename, tcp_port);
2769
    GET_CONF_STR (conf, filename, listen_addr);
J
Jim Meyering 已提交
2770

2771
    if (remoteConfigGetAuth(conf, "auth_unix_rw", &auth_unix_rw, filename) < 0)
J
Jim Meyering 已提交
2772
        goto free_and_fail;
2773 2774 2775 2776 2777 2778 2779
#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
2780
    if (remoteConfigGetAuth(conf, "auth_unix_ro", &auth_unix_ro, filename) < 0)
J
Jim Meyering 已提交
2781
        goto free_and_fail;
2782
    if (remoteConfigGetAuth(conf, "auth_tcp", &auth_tcp, filename) < 0)
J
Jim Meyering 已提交
2783
        goto free_and_fail;
2784
    if (remoteConfigGetAuth(conf, "auth_tls", &auth_tls, filename) < 0)
J
Jim Meyering 已提交
2785
        goto free_and_fail;
2786

2787 2788
    GET_CONF_STR (conf, filename, unix_sock_group);
    if (unix_sock_group) {
2789
        if (!server->privileged) {
2790
            VIR_WARN0("Cannot set group when not running as root");
2791
        } else {
2792
            int ret;
2793
            struct group grpdata, *grp;
2794 2795 2796 2797 2798 2799
            size_t maxbuf = sysconf(_SC_GETGR_R_SIZE_MAX);

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

            if (VIR_ALLOC_N(buf, maxbuf) < 0) {
2800
                VIR_ERROR0(_("Failed to allocate memory for buffer"));
2801 2802 2803 2804 2805 2806 2807 2808
                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) {
2809
                        VIR_ERROR0(_("Failed to reallocate enough memory for buffer"));
2810 2811 2812 2813 2814
                        goto free_and_fail;
                    }
            }

            if (ret != 0 || !grp) {
2815
                VIR_ERROR(_("Failed to lookup group '%s'"), unix_sock_group);
2816
                goto free_and_fail;
2817 2818
            }
            unix_sock_gid = grp->gr_gid;
2819
            VIR_FREE(buf);
2820
        }
2821
        VIR_FREE(unix_sock_group);
2822 2823
    }

2824 2825
    GET_CONF_STR (conf, filename, unix_sock_ro_perms);
    if (unix_sock_ro_perms) {
2826
        if (virStrToLong_i (unix_sock_ro_perms, NULL, 8, &unix_sock_ro_mask) != 0) {
2827
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_ro_perms);
2828
            goto free_and_fail;
2829
        }
2830
        VIR_FREE(unix_sock_ro_perms);
2831 2832
    }

2833 2834
    GET_CONF_STR (conf, filename, unix_sock_rw_perms);
    if (unix_sock_rw_perms) {
2835
        if (virStrToLong_i (unix_sock_rw_perms, NULL, 8, &unix_sock_rw_mask) != 0) {
2836
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_rw_perms);
2837
            goto free_and_fail;
2838
        }
2839
        VIR_FREE(unix_sock_rw_perms);
2840 2841
    }

2842 2843
    GET_CONF_STR (conf, filename, unix_sock_dir);

2844 2845
    GET_CONF_INT (conf, filename, mdns_adv);
    GET_CONF_STR (conf, filename, mdns_name);
2846

2847
    GET_CONF_INT (conf, filename, tls_no_verify_certificate);
2848

2849 2850 2851 2852
    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);
2853

2854 2855 2856
    if (remoteConfigGetStringList (conf, "tls_allowed_dn_list",
                                   &tls_allowed_dn_list, filename) < 0)
        goto free_and_fail;
2857

2858
    if (remoteReadSaslAllowedUsernameList (conf, server, filename) < 0)
2859
        goto free_and_fail;
2860

2861 2862 2863 2864 2865

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

2866 2867 2868
    GET_CONF_INT (conf, filename, max_requests);
    GET_CONF_INT (conf, filename, max_client_requests);

2869 2870 2871
    GET_CONF_INT (conf, filename, audit_level);
    GET_CONF_INT (conf, filename, audit_logging);

2872
    GET_CONF_STR (conf, filename, host_uuid);
2873 2874
    if (virSetHostUUIDStr(host_uuid)) {
        VIR_ERROR(_("invalid host UUID: %s"), host_uuid);
2875
        goto free_and_fail;
2876
    }
2877 2878 2879

    VIR_FREE(host_uuid);

2880 2881
    virConfFree (conf);
    return 0;
2882

2883 2884
 free_and_fail:
    virConfFree (conf);
2885
    VIR_FREE(host_uuid);
2886 2887 2888 2889 2890
    VIR_FREE(mdns_name);
    VIR_FREE(unix_sock_ro_perms);
    VIR_FREE(unix_sock_rw_perms);
    VIR_FREE(unix_sock_group);
    VIR_FREE(buf);
2891

2892 2893 2894 2895
    /* 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
2896 2897 2898 2899 2900
       even be a real leak.  */

    if (tls_allowed_dn_list) {
        int i;
        for (i = 0; tls_allowed_dn_list[i]; i++)
2901 2902
            VIR_FREE(tls_allowed_dn_list[i]);
        VIR_FREE(tls_allowed_dn_list);
2903 2904 2905
    }

    return -1;
2906 2907
}

2908 2909
/* Display version information. */
static void
2910
version (void)
2911 2912 2913 2914
{
    printf ("%s (%s) %s\n", argv0, PACKAGE_NAME, PACKAGE_VERSION);
}

J
John Levon 已提交
2915 2916 2917 2918 2919 2920 2921 2922
#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)) {
2923
        VIR_ERROR0(_("additional privileges are required"));
J
John Levon 已提交
2924 2925 2926 2927 2928
        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)) {
2929
        VIR_ERROR0(_("failed to set reduced privileges"));
J
John Levon 已提交
2930 2931 2932 2933 2934 2935
        return -1;
    }

    return 0;
}
#else
2936
# define qemudSetupPrivs() 0
J
John Levon 已提交
2937 2938
#endif

2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 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

/*
 * 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);
    sigaction(SIGCHLD, &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:
2995 2996
    VIR_FORCE_CLOSE(sigpipe[0]);
    VIR_FORCE_CLOSE(sigpipe[1]);
2997 2998 2999
    return -1;
}

3000 3001
/* Print command-line usage. */
static void
3002
usage (void)
3003 3004
{
    fprintf (stderr,
3005
             _("\n\
3006 3007 3008 3009 3010 3011
Usage:\n\
  %s [options]\n\
\n\
Options:\n\
  -v | --verbose         Verbose messages.\n\
  -d | --daemon          Run as a daemon & write PID file.\n\
3012 3013
  -l | --listen          Listen for TCP/IP connections.\n\
  -t | --timeout <secs>  Exit after timeout period.\n\
3014
  -f | --config <file>   Configuration file.\n\
3015
     | --version         Display version information.\n\
3016 3017
  -p | --pid-file <file> Change name of PID file.\n\
\n\
3018
libvirt management daemon:\n\
3019 3020 3021
\n\
  Default paths:\n\
\n\
W
Wolfgang Mauerer 已提交
3022
    Configuration file (unless overridden by -f):\n\
3023
      %s/libvirt/libvirtd.conf\n\
3024
\n\
3025
    Sockets (as root):\n\
3026 3027
      %s/run/libvirt/libvirt-sock\n\
      %s/run/libvirt/libvirt-sock-ro\n\
3028 3029 3030
\n\
    Sockets (as non-root):\n\
      $HOME/.libvirt/libvirt-sock (in UNIX abstract namespace)\n\
3031 3032
\n\
    TLS:\n\
3033 3034 3035
      CA certificate:     %s\n\
      Server certificate: %s\n\
      Server private key: %s\n\
3036 3037 3038
\n\
    PID file (unless overridden by --pid-file):\n\
      %s\n\
3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049
\n"),
               argv0,
               SYSCONF_DIR,
               LOCAL_STATE_DIR,
               LOCAL_STATE_DIR,
               LIBVIRT_CACERT,
               LIBVIRT_SERVERCERT,
               LIBVIRT_SERVERKEY,
               (REMOTE_PID_FILE[0] != '\0'
                ? REMOTE_PID_FILE
                : _("(disabled in ./configure)")));
3050 3051
}

3052 3053 3054 3055
enum {
    OPT_VERSION = 129
};

D
Daniel P. Berrange 已提交
3056 3057
#define MAX_LISTEN 5
int main(int argc, char **argv) {
3058
    struct qemud_server *server = NULL;
3059
    const char *pid_file = NULL;
3060
    const char *remote_config_file = NULL;
3061
    int statuswrite = -1;
3062
    int ret = 1;
3063
    argv0 = argv[0];
D
Daniel P. Berrange 已提交
3064 3065 3066 3067

    struct option opts[] = {
        { "verbose", no_argument, &verbose, 1},
        { "daemon", no_argument, &godaemon, 1},
3068
        { "listen", no_argument, &ipsock, 1},
3069
        { "config", required_argument, NULL, 'f'},
3070 3071
        { "timeout", required_argument, NULL, 't'},
        { "pid-file", required_argument, NULL, 'p'},
3072
        { "version", no_argument, NULL, OPT_VERSION },
3073
        { "help", no_argument, NULL, '?' },
D
Daniel P. Berrange 已提交
3074 3075 3076
        {0, 0, 0, 0}
    };

3077
    if (virInitialize() < 0) {
3078
        fprintf (stderr, _("%s: initialization failed\n"), argv0);
3079 3080
        exit (EXIT_FAILURE);
    }
D
Daniel P. Berrange 已提交
3081

D
Daniel P. Berrange 已提交
3082 3083 3084 3085 3086
    while (1) {
        int optidx = 0;
        int c;
        char *tmp;

3087
        c = getopt_long(argc, argv, "ldf:p:t:v", opts, &optidx);
D
Daniel P. Berrange 已提交
3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102

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

        switch (c) {
        case 0:
            /* Got one of the flags */
            break;
        case 'v':
            verbose = 1;
            break;
        case 'd':
            godaemon = 1;
            break;
3103 3104
        case 'l':
            ipsock = 1;
D
Daniel P. Berrange 已提交
3105 3106 3107
            break;

        case 't':
3108
            if (virStrToLong_i(optarg, &tmp, 10, &timeout) != 0
3109 3110 3111
                || timeout <= 0
                /* Ensure that we can multiply by 1000 without overflowing.  */
                || timeout > INT_MAX / 1000)
D
Daniel P. Berrange 已提交
3112 3113
                timeout = -1;
            break;
3114 3115

        case 'p':
3116 3117 3118 3119 3120
            pid_file = optarg;
            break;

        case 'f':
            remote_config_file = optarg;
3121 3122
            break;

3123
        case OPT_VERSION:
3124
            version ();
3125 3126
            return 0;

D
Daniel P. Berrange 已提交
3127
        case '?':
3128
            usage ();
D
Daniel P. Berrange 已提交
3129 3130 3131
            return 2;

        default:
3132 3133
            fprintf (stderr, _("%s: internal error: unknown flag: %c\n"),
                     argv0, c);
3134
            exit (EXIT_FAILURE);
D
Daniel P. Berrange 已提交
3135 3136 3137
        }
    }

3138 3139 3140 3141
    if (remote_config_file == NULL) {
        static const char *default_config_file
            = SYSCONF_DIR "/libvirt/libvirtd.conf";
        remote_config_file =
3142
            (access(default_config_file, R_OK) == 0
3143 3144 3145 3146
             ? default_config_file
             : "/dev/null");
    }

3147
    if (godaemon) {
3148
        char ebuf[1024];
3149
        if ((statuswrite = daemonForkIntoBackground()) < 0) {
3150 3151
            VIR_ERROR(_("Failed to fork as daemon: %s"),
                      virStrerror(errno, ebuf, sizeof ebuf));
3152
            goto error;
3153 3154 3155 3156 3157
        }
    }

    /* If running as root and no PID file is set, use the default */
    if (pid_file == NULL &&
3158
        geteuid() == 0 &&
3159 3160 3161 3162 3163
        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 &&
3164 3165 3166
        qemudWritePidFile (pid_file) < 0) {
        pid_file = NULL; /* Prevent unlinking of someone else's pid ! */
        ret = VIR_DAEMON_ERR_PIDFILE;
3167
        goto error;
3168
    }
3169

J
John Levon 已提交
3170
    /* Ensure the rundir exists (on tmpfs on some systems) */
3171
    if (geteuid() == 0) {
J
John Levon 已提交
3172 3173 3174 3175
        const char *rundir = LOCAL_STATE_DIR "/run/libvirt";

        if (mkdir (rundir, 0755)) {
            if (errno != EEXIST) {
3176 3177 3178
                char ebuf[1024];
                VIR_ERROR(_("unable to create rundir %s: %s"), rundir,
                          virStrerror(errno, ebuf, sizeof(ebuf)));
3179 3180
                ret = VIR_DAEMON_ERR_RUNDIR;
                goto error;
J
John Levon 已提交
3181 3182 3183 3184
            }
        }
    }

3185 3186 3187 3188 3189 3190
    /* 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
     */
3191 3192
    if (qemudSetupPrivs() < 0) {
        ret = VIR_DAEMON_ERR_PRIVS;
3193
        goto error;
3194
    }
J
John Levon 已提交
3195

3196
    if (!(server = qemudInitialize())) {
3197
        ret = VIR_DAEMON_ERR_INIT;
3198
        goto error;
3199
    }
3200

3201 3202
    if ((daemonSetupSignals(server)) < 0) {
        ret = VIR_DAEMON_ERR_SIGNAL;
3203
        goto error;
3204
    }
3205

3206
    /* Read the config file (if it exists). */
3207 3208
    if (remoteReadConfigFile (server, remote_config_file) < 0) {
        ret = VIR_DAEMON_ERR_CONFIG;
3209
        goto error;
3210
    }
D
Daniel P. Berrange 已提交
3211

3212 3213 3214 3215 3216 3217 3218 3219 3220 3221
    if (audit_level) {
        if (virAuditOpen() < 0) {
            if (audit_level > 1) {
                ret = VIR_DAEMON_ERR_AUDIT;
                goto error;
            }
        }
    }
    virAuditLog(audit_logging);

3222
    /* setup the hooks if any */
3223
    if (virHookInitialize() < 0) {
3224 3225 3226 3227
        ret = VIR_DAEMON_ERR_HOOKS;
        goto error;
    }

3228 3229 3230
    /* Disable error func, now logging is setup */
    virSetErrorFunc(NULL, virshErrorHandler);

3231 3232 3233 3234 3235 3236 3237 3238
    /*
     * 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);

3239
    if (qemudNetworkInit(server) < 0) {
3240
        ret = VIR_DAEMON_ERR_NETWORK;
3241
        goto error;
3242 3243
    }

3244 3245 3246 3247 3248 3249 3250 3251 3252
    /* 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)
            ;
3253
        VIR_FORCE_CLOSE(statuswrite);
3254 3255
    }

3256 3257 3258
    /* Start the event loop in a background thread, since
     * state initialization needs events to be being processed */
    if (qemudStartEventLoop(server) < 0) {
3259
        VIR_ERROR0(_("Event thread startup failed"));
3260 3261 3262
        goto error;
    }

3263 3264 3265 3266 3267
    /* 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) {
3268
        VIR_ERROR0(_("Driver state initialization failed"));
3269
        goto shutdown;
3270
    }
D
Daniel P. Berrange 已提交
3271

3272 3273 3274
    /* Start accepting new clients from network */
    virMutexLock(&server->lock);
    if (qemudNetworkEnable(server) < 0) {
3275
        VIR_ERROR0(_("Network event loop enablement failed"));
3276 3277 3278 3279
        goto shutdown;
    }
    virMutexUnlock(&server->lock);

3280 3281
    ret = 0;

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297
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);

3298 3299 3300
    virHookCall(VIR_HOOK_DRIVER_DAEMON, "-", VIR_HOOK_DAEMON_OP_SHUTDOWN,
                0, "shutdown", NULL);

3301
error:
3302 3303 3304 3305 3306 3307 3308 3309
    if (statuswrite != -1) {
        if (ret != 0) {
            /* Tell parent of daemon what failed */
            char status = ret;
            while (write(statuswrite, &status, 1) == -1 &&
                   errno == EINTR)
                ;
        }
3310
        VIR_FORCE_CLOSE(statuswrite);
3311
    }
3312 3313
    if (server)
        qemudCleanup(server);
3314 3315
    if (pid_file)
        unlink (pid_file);
3316
    virLogShutdown();
3317
    return ret;
D
Daniel P. Berrange 已提交
3318
}