libvirtd.c 85.9 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, 2007, 2008, 2009 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

54 55
#define VIR_FROM_THIS VIR_FROM_QEMU

D
Daniel P. Berrange 已提交
56
#include "libvirtd.h"
57 58
#include "dispatch.h"

59
#include "util.h"
60
#include "remote_driver.h"
61
#include "conf.h"
D
Daniel P. Berrange 已提交
62
#include "event.h"
63
#include "memory.h"
64
#include "stream.h"
65 66 67
#ifdef HAVE_AVAHI
#include "mdns.h"
#endif
D
Daniel P. Berrange 已提交
68

69 70 71
#ifdef WITH_DRIVER_MODULES
#include "driver.h"
#else
72
#ifdef WITH_QEMU
73
#include "qemu/qemu_driver.h"
74 75
#endif
#ifdef WITH_LXC
76
#include "lxc/lxc_driver.h"
77
#endif
78
#ifdef WITH_UML
D
Daniel P. Berrange 已提交
79
#include "uml/uml_driver.h"
80
#endif
D
Daniel Veillard 已提交
81 82 83
#ifdef WITH_ONE
#include "opennebula/one_driver.h"
#endif
84
#ifdef WITH_NETWORK
85
#include "network/bridge_driver.h"
86
#endif
87
#ifdef WITH_NETCF
88
#include "interface/netcf_driver.h"
89
#endif
90
#ifdef WITH_STORAGE_DIR
91
#include "storage/storage_driver.h"
92
#endif
93
#ifdef WITH_NODE_DEVICES
94
#include "node_device/node_device_driver.h"
95
#endif
96
#include "secret/secret_driver.h"
97
#endif
98 99


J
John Levon 已提交
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
#ifdef __sun
#include <ucred.h>
#include <priv.h>

#ifndef PRIV_VIRT_MANAGE
#define PRIV_VIRT_MANAGE ((const char *)"virt_manage")
#endif

#ifndef PRIV_XVM_CONTROL
#define PRIV_XVM_CONTROL ((const char *)"xvm_control")
#endif

#define PU_RESETGROUPS          0x0001  /* Remove supplemental groups */
#define PU_CLEARLIMITSET        0x0008  /* L=0 */

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

#define SYSTEM_UID 60

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 */

131 132
static int godaemon = 0;        /* -d: Be a daemon */
static int verbose = 0;         /* -v: Verbose mode */
133
static int timeout = -1;        /* -t: Shutdown timeout */
134
static int sigwrite = -1;       /* Signal handler pipe */
135
static int ipsock = 0;          /* -l  Listen for TCP/IP */
136

137
/* Defaults for configuration file elements */
138 139
static int listen_tls = 1;
static int listen_tcp = 0;
140
static char *listen_addr  = (char *) LIBVIRTD_LISTEN_ADDR;
141 142
static char *tls_port = (char *) LIBVIRTD_TLS_PORT;
static char *tcp_port = (char *) LIBVIRTD_TCP_PORT;
143

144 145
static char *unix_sock_dir = NULL;

146 147 148 149
#if HAVE_POLKIT
static int auth_unix_rw = REMOTE_AUTH_POLKIT;
static int auth_unix_ro = REMOTE_AUTH_POLKIT;
#else
150 151
static int auth_unix_rw = REMOTE_AUTH_NONE;
static int auth_unix_ro = REMOTE_AUTH_NONE;
152
#endif /* HAVE_POLKIT */
153 154 155 156 157 158 159
#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;

160
static int mdns_adv = 1;
161
static char *mdns_name = NULL;
162

163
static int tls_no_verify_certificate = 0;
164
static char **tls_allowed_dn_list = NULL;
165

166 167 168 169
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 *) "";
170 171 172 173

static gnutls_certificate_credentials_t x509_cred;
static gnutls_dh_params_t dh_params;

174 175 176 177
static int min_workers = 5;
static int max_workers = 20;
static int max_clients = 20;

178 179 180 181 182
/* 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;

183
#define DH_BITS 1024
184

185 186 187
static sig_atomic_t sig_errors = 0;
static int sig_lasterrno = 0;

188 189
static void sig_handler(int sig, siginfo_t * siginfo,
                        void* context ATTRIBUTE_UNUSED) {
190
    int origerrno;
191
    int r;
192

193 194
    /* set the sig num in the struct */
    siginfo->si_signo = sig;
195 196

    origerrno = errno;
197
    r = safewrite(sigwrite, siginfo, sizeof(*siginfo));
198 199 200 201
    if (r == -1) {
        sig_errors++;
        sig_lasterrno = errno;
    }
202
    errno = origerrno;
D
Daniel P. Berrange 已提交
203
}
204

205 206
static void qemudDispatchClientEvent(int watch, int fd, int events, void *opaque);
static void qemudDispatchServerEvent(int watch, int fd, int events, void *opaque);
207
static int qemudStartWorker(struct qemud_server *server, struct qemud_worker *worker);
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223

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;
    }
}

224
struct qemud_client_message *
225 226 227 228 229 230 231 232 233 234 235
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
236

237 238 239 240 241
static int
remoteCheckCertFile(const char *type, const char *file)
{
    struct stat sb;
    if (stat(file, &sb) < 0) {
242 243 244
        char ebuf[1024];
        VIR_ERROR(_("Cannot access %s '%s': %s"),
                  type, file, virStrerror(errno, ebuf, sizeof ebuf));
245 246 247 248 249
        return -1;
    }
    return 0;
}

250 251 252 253 254 255 256 257 258 259
static int
remoteInitializeGnuTLS (void)
{
    int err;

    /* Initialise GnuTLS. */
    gnutls_global_init ();

    err = gnutls_certificate_allocate_credentials (&x509_cred);
    if (err) {
260
        VIR_ERROR(_("gnutls_certificate_allocate_credentials: %s"),
261 262 263 264 265
                  gnutls_strerror (err));
        return -1;
    }

    if (ca_file && ca_file[0] != '\0') {
266 267 268
        if (remoteCheckCertFile("CA certificate", ca_file) < 0)
            return -1;

269 270 271 272
        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) {
273
            VIR_ERROR(_("gnutls_certificate_set_x509_trust_file: %s"),
274 275 276 277 278 279
                      gnutls_strerror (err));
            return -1;
        }
    }

    if (crl_file && crl_file[0] != '\0') {
J
Jim Meyering 已提交
280
        if (remoteCheckCertFile("CA revocation list", crl_file) < 0)
281 282
            return -1;

283
        DEBUG("loading CRL from %s", crl_file);
284 285 286
        err = gnutls_certificate_set_x509_crl_file (x509_cred, crl_file,
                                                    GNUTLS_X509_FMT_PEM);
        if (err < 0) {
287
            VIR_ERROR(_("gnutls_certificate_set_x509_crl_file: %s"),
288 289 290 291 292 293
                      gnutls_strerror (err));
            return -1;
        }
    }

    if (cert_file && cert_file[0] != '\0' && key_file && key_file[0] != '\0') {
294 295 296 297
        if (remoteCheckCertFile("server certificate", cert_file) < 0)
            return -1;
        if (remoteCheckCertFile("server key", key_file) < 0)
            return -1;
298
        DEBUG("loading cert and key from %s and %s", cert_file, key_file);
299 300 301 302 303
        err =
            gnutls_certificate_set_x509_key_file (x509_cred,
                                                  cert_file, key_file,
                                                  GNUTLS_X509_FMT_PEM);
        if (err < 0) {
304
            VIR_ERROR(_("gnutls_certificate_set_x509_key_file: %s"),
305 306 307 308 309 310 311 312 313 314 315 316
                      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) {
317
        VIR_ERROR(_("gnutls_dh_params_init: %s"), gnutls_strerror (err));
318 319 320 321
        return -1;
    }
    err = gnutls_dh_params_generate2 (dh_params, DH_BITS);
    if (err < 0) {
322
        VIR_ERROR(_("gnutls_dh_params_generate2: %s"), gnutls_strerror (err));
323 324 325 326 327 328 329 330
        return -1;
    }

    gnutls_certificate_set_dh_params (x509_cred, dh_params);

    return 0;
}

331
static void
332 333
qemudDispatchSignalEvent(int watch ATTRIBUTE_UNUSED,
                         int fd ATTRIBUTE_UNUSED,
334 335
                         int events ATTRIBUTE_UNUSED,
                         void *opaque) {
D
Daniel P. Berrange 已提交
336
    struct qemud_server *server = (struct qemud_server *)opaque;
337
    siginfo_t siginfo;
338 339
    int ret;

340
    virMutexLock(&server->lock);
341

342
    if (saferead(server->sigread, &siginfo, sizeof(siginfo)) != sizeof(siginfo)) {
343 344 345
        char ebuf[1024];
        VIR_ERROR(_("Failed to read from signal pipe: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
346
        virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
347
        return;
348 349
    }

350 351
    ret = 0;

352
    switch (siginfo.si_signo) {
353
    case SIGHUP:
354
        VIR_INFO0(_("Reloading configuration on SIGHUP"));
355
        if (virStateReload() < 0)
356
            VIR_WARN0(_("Error while reloading drivers"));
357 358 359
        break;

    case SIGINT:
360
    case SIGQUIT:
361
    case SIGTERM:
362
        VIR_WARN(_("Shutting down on signal %d"), siginfo.si_signo);
363 364 365 366
        server->shutdown = 1;
        break;

    default:
367
        VIR_INFO(_("Received unexpected signal %d"), siginfo.si_signo);
368 369 370
        break;
    }

D
Daniel P. Berrange 已提交
371 372
    if (ret != 0)
        server->shutdown = 1;
373

374
    virMutexUnlock(&server->lock);
375 376
}

377

D
Daniel P. Berrange 已提交
378 379 380 381 382 383 384
static int qemudGoDaemon(void) {
    int pid = fork();
    switch (pid) {
    case 0:
        {
            int stdinfd = -1;
            int stdoutfd = -1;
385
            int nextpid;
D
Daniel P. Berrange 已提交
386

387
            if ((stdinfd = open("/dev/null", O_RDONLY)) < 0)
D
Daniel P. Berrange 已提交
388
                goto cleanup;
389
            if ((stdoutfd = open("/dev/null", O_WRONLY)) < 0)
D
Daniel P. Berrange 已提交
390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
                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;
            if (close(stdinfd) < 0)
                goto cleanup;
            stdinfd = -1;
            if (close(stdoutfd) < 0)
                goto cleanup;
            stdoutfd = -1;

            if (setsid() < 0)
                goto cleanup;

            nextpid = fork();
            switch (nextpid) {
            case 0:
                return 0;
            case -1:
                return -1;
            default:
414
                _exit(0);
D
Daniel P. Berrange 已提交
415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
            }

        cleanup:
            if (stdoutfd != -1)
                close(stdoutfd);
            if (stdinfd != -1)
                close(stdinfd);
            return -1;

        }

    case -1:
        return -1;

    default:
        {
            int got, status = 0;
            /* We wait to make sure the next child forked
               successfully */
            if ((got = waitpid(pid, &status, 0)) < 0 ||
                got != pid ||
                status != 0) {
                return -1;
            }
439
            _exit(0);
D
Daniel P. Berrange 已提交
440 441 442 443
        }
    }
}

444 445 446
static int qemudWritePidFile(const char *pidFile) {
    int fd;
    FILE *fh;
447
    char ebuf[1024];
448 449 450 451 452

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

    if ((fd = open(pidFile, O_WRONLY|O_CREAT|O_EXCL, 0644)) < 0) {
453
        VIR_ERROR(_("Failed to open pid file '%s' : %s"),
454
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
455 456 457 458
        return -1;
    }

    if (!(fh = fdopen(fd, "w"))) {
459
        VIR_ERROR(_("Failed to fdopen pid file '%s' : %s"),
460
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
461 462 463 464 465
        close(fd);
        return -1;
    }

    if (fprintf(fh, "%lu\n", (unsigned long)getpid()) < 0) {
466
        VIR_ERROR(_("Failed to write to pid file '%s' : %s"),
467
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
468
        fclose(fh);
469 470 471 472
        return -1;
    }

    if (fclose(fh) == EOF) {
473
        VIR_ERROR(_("Failed to close pid file '%s' : %s"),
474
                  pidFile, virStrerror(errno, ebuf, sizeof ebuf));
475 476 477 478 479 480
        return -1;
    }

    return 0;
}

D
Daniel P. Berrange 已提交
481
static int qemudListenUnix(struct qemud_server *server,
482
                           const char *path, int readonly, int auth) {
483
    struct qemud_socket *sock;
D
Daniel P. Berrange 已提交
484 485
    struct sockaddr_un addr;
    mode_t oldmask;
486
    gid_t oldgrp;
487
    char ebuf[1024];
D
Daniel P. Berrange 已提交
488

489
    if (VIR_ALLOC(sock) < 0) {
490
        VIR_ERROR("%s", _("Failed to allocate memory for struct qemud_socket"));
D
Daniel P. Berrange 已提交
491
        return -1;
492
    }
D
Daniel P. Berrange 已提交
493 494

    sock->readonly = readonly;
495
    sock->port = -1;
496
    sock->type = QEMUD_SOCK_TYPE_UNIX;
497
    sock->auth = auth;
D
Daniel P. Berrange 已提交
498

499
    if ((sock->fd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
500
        VIR_ERROR(_("Failed to create socket: %s"),
501
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
502
        goto cleanup;
503
    }
D
Daniel P. Berrange 已提交
504

505 506
    if (virSetCloseExec(sock->fd) < 0 ||
        virSetNonBlock(sock->fd) < 0)
D
Daniel P. Berrange 已提交
507
        goto cleanup;
D
Daniel P. Berrange 已提交
508 509 510

    memset(&addr, 0, sizeof(addr));
    addr.sun_family = AF_UNIX;
C
Chris Lalancette 已提交
511 512 513 514
    if (virStrcpyStatic(addr.sun_path, path) == NULL) {
        VIR_ERROR(_("Path %s too long for unix socket"), path);
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
515 516 517
    if (addr.sun_path[0] == '@')
        addr.sun_path[0] = '\0';

518
    oldgrp = getgid();
519
    oldmask = umask(readonly ? ~unix_sock_ro_mask : ~unix_sock_rw_mask);
520
    if (server->privileged)
521 522
        setgid(unix_sock_gid);

523
    if (bind(sock->fd, (struct sockaddr *)&addr, sizeof(addr)) < 0) {
524
        VIR_ERROR(_("Failed to bind socket to '%s': %s"),
525
                  path, virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
526
        goto cleanup;
527
    }
D
Daniel P. Berrange 已提交
528
    umask(oldmask);
529
    if (server->privileged)
530
        setgid(oldgrp);
D
Daniel P. Berrange 已提交
531

532
    if (listen(sock->fd, 30) < 0) {
533
        VIR_ERROR(_("Failed to listen for connections on '%s': %s"),
534
                  path, virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
535 536 537
        goto cleanup;
    }

538 539 540 541 542
    if ((sock->watch = virEventAddHandleImpl(sock->fd,
                                             VIR_EVENT_HANDLE_READABLE |
                                             VIR_EVENT_HANDLE_ERROR |
                                             VIR_EVENT_HANDLE_HANGUP,
                                             qemudDispatchServerEvent,
543
                                             server, NULL)) < 0) {
544
        VIR_ERROR0(_("Failed to add server event callback"));
D
Daniel P. Berrange 已提交
545
        goto cleanup;
546
    }
D
Daniel P. Berrange 已提交
547

D
Daniel P. Berrange 已提交
548 549 550 551
    sock->next = server->sockets;
    server->sockets = sock;
    server->nsockets++;

D
Daniel P. Berrange 已提交
552
    return 0;
D
Daniel P. Berrange 已提交
553 554 555 556 557 558

 cleanup:
    if (sock->fd)
        close(sock->fd);
    free(sock);
    return -1;
D
Daniel P. Berrange 已提交
559 560
}

561 562
// See: http://people.redhat.com/drepper/userapi-ipv6.html
static int
563
remoteMakeSockets (int *fds, int max_fds, int *nfds_r, const char *node, const char *service)
564 565 566 567 568 569 570
{
    struct addrinfo *ai;
    struct addrinfo hints;
    memset (&hints, 0, sizeof hints);
    hints.ai_flags = AI_PASSIVE | AI_ADDRCONFIG;
    hints.ai_socktype = SOCK_STREAM;

571
    int e = getaddrinfo (node, service, &hints, &ai);
572
    if (e != 0) {
573
        VIR_ERROR(_("getaddrinfo: %s\n"), gai_strerror (e));
574 575 576 577 578
        return -1;
    }

    struct addrinfo *runp = ai;
    while (runp && *nfds_r < max_fds) {
579
        char ebuf[1024];
580 581 582
        fds[*nfds_r] = socket (runp->ai_family, runp->ai_socktype,
                               runp->ai_protocol);
        if (fds[*nfds_r] == -1) {
583
            VIR_ERROR(_("socket: %s"), virStrerror (errno, ebuf, sizeof ebuf));
584 585 586 587 588 589 590 591
            return -1;
        }

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

        if (bind (fds[*nfds_r], runp->ai_addr, runp->ai_addrlen) == -1) {
            if (errno != EADDRINUSE) {
592
                VIR_ERROR(_("bind: %s"), virStrerror (errno, ebuf, sizeof ebuf));
593 594 595
                return -1;
            }
            close (fds[*nfds_r]);
596
        } else {
597 598 599 600
            ++*nfds_r;
        }
        runp = runp->ai_next;
    }
601

602 603 604 605 606 607 608 609 610
    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,
611
                 const char *addr,
612
                 const char *port,
613
                 int type,
614
                 int auth)
615 616 617
{
    int fds[2];
    int nfds = 0;
618
    int i;
619 620
    struct qemud_socket *sock;

621
    if (remoteMakeSockets (fds, 2, &nfds, addr, port) == -1)
622 623 624
        return -1;

    for (i = 0; i < nfds; ++i) {
625 626 627 628 629 630 631 632
        union {
            struct sockaddr_storage sa_stor;
            struct sockaddr sa;
            struct sockaddr_in sa_in;
#ifdef AF_INET6
            struct sockaddr_in6 sa_in6;
#endif
        } s;
633
        char ebuf[1024];
634
        socklen_t salen = sizeof(s);
635

636
        if (VIR_ALLOC(sock) < 0) {
637 638
            VIR_ERROR(_("remoteListenTCP: calloc: %s"),
                      virStrerror (errno, ebuf, sizeof ebuf));
639
            goto cleanup;
640 641 642 643 644 645 646 647
        }

        sock->readonly = 0;
        sock->next = server->sockets;
        server->sockets = sock;
        server->nsockets++;

        sock->fd = fds[i];
648
        sock->type = type;
649
        sock->auth = auth;
D
Daniel P. Berrange 已提交
650

651
        if (getsockname(sock->fd, &s.sa, &salen) < 0)
652
            goto cleanup;
653

654 655
        if (s.sa.sa_family == AF_INET) {
            sock->port = htons(s.sa_in.sin_port);
656
#ifdef AF_INET6
657 658
        } else if (s.sa.sa_family == AF_INET6)
            sock->port = htons(s.sa_in6.sin6_port);
659
#endif
660 661 662
        else
            sock->port = -1;

663 664
        if (virSetCloseExec(sock->fd) < 0 ||
            virSetNonBlock(sock->fd) < 0)
665
            goto cleanup;
666 667

        if (listen (sock->fd, 30) < 0) {
668 669
            VIR_ERROR(_("remoteListenTCP: listen: %s"),
                      virStrerror (errno, ebuf, sizeof ebuf));
670
            goto cleanup;
671
        }
D
Daniel P. Berrange 已提交
672

673 674 675 676 677
        if ((sock->watch = virEventAddHandleImpl(sock->fd,
                                                 VIR_EVENT_HANDLE_READABLE |
                                                 VIR_EVENT_HANDLE_ERROR |
                                                 VIR_EVENT_HANDLE_HANGUP,
                                                 qemudDispatchServerEvent,
678
                                                 server, NULL)) < 0) {
679
            VIR_ERROR0(_("Failed to add server event callback"));
680
            goto cleanup;
D
Daniel P. Berrange 已提交
681 682
        }

683 684 685
    }

    return 0;
686 687 688

cleanup:
    for (i = 0; i < nfds; ++i)
C
Cole Robinson 已提交
689
        close(fds[i]);
690
    return -1;
691 692 693 694 695
}

static int qemudInitPaths(struct qemud_server *server,
                          char *sockname,
                          char *roSockname,
J
John Levon 已提交
696 697
                          int maxlen)
{
698 699 700 701 702 703 704 705 706
    char *sock_dir;
    char *dir_prefix = NULL;
    int ret = -1;
    char *sock_dir_prefix = NULL;

    if (unix_sock_dir)
        sock_dir = unix_sock_dir;
    else {
        sock_dir = sockname;
707
        if (server->privileged) {
708 709 710 711 712 713 714 715 716
            dir_prefix = strdup (LOCAL_STATE_DIR);
            if (dir_prefix == NULL) {
                virReportOOMError(NULL);
                goto cleanup;
            }
            if (snprintf (sock_dir, maxlen, "%s/run/libvirt",
                          dir_prefix) >= maxlen)
                goto snprintf_error;
        } else {
717
            uid_t uid = geteuid();
718 719 720 721 722
            dir_prefix = virGetUserDirectory(NULL, uid);
            if (dir_prefix == NULL) {
                /* Do not diagnose here; virGetUserDirectory does that.  */
                goto snprintf_error;
            }
723

724 725 726 727
            if (snprintf(sock_dir, maxlen, "%s/.libvirt", dir_prefix) >= maxlen)
                goto snprintf_error;
        }
    }
728

729 730 731 732 733
    sock_dir_prefix = strdup (sock_dir);
    if (!sock_dir_prefix) {
        virReportOOMError(NULL);
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
734

735
    if (server->privileged) {
736 737 738 739
        if (snprintf (sockname, maxlen, "%s/libvirt-sock",
                      sock_dir_prefix) >= maxlen
            || (snprintf (roSockname, maxlen, "%s/libvirt-sock-ro",
                          sock_dir_prefix) >= maxlen))
740
            goto snprintf_error;
741
        unlink(sockname);
742
        unlink(roSockname);
D
Daniel P. Berrange 已提交
743
    } else {
744 745
        if (snprintf(sockname, maxlen, "@%s/libvirt-sock",
                     sock_dir_prefix) >= maxlen)
746
            goto snprintf_error;
747
    }
748

749 750 751 752 753 754 755
    if (server->privileged) {
        if (!(server->logDir = strdup (LOCAL_STATE_DIR "/log/libvirt")))
            virReportOOMError(NULL);
    } else {
        if (virAsprintf(&server->logDir, "%s/.libvirt/log", dir_prefix) < 0)
            virReportOOMError(NULL);
    }
756

757
    if (server->logDir == NULL)
758
        goto cleanup;
759 760

    ret = 0;
761 762

 snprintf_error:
763 764 765 766 767 768 769 770
    if (ret)
        VIR_ERROR("%s",
                  _("Resulting path too long for buffer in qemudInitPaths()"));

 cleanup:
    free (dir_prefix);
    free (sock_dir_prefix);
    return ret;
D
Daniel P. Berrange 已提交
771 772
}

773
static struct qemud_server *qemudInitialize(int sigread) {
D
Daniel P. Berrange 已提交
774 775
    struct qemud_server *server;

776
    if (VIR_ALLOC(server) < 0) {
777
        VIR_ERROR0(_("Failed to allocate struct qemud_server"));
D
Daniel P. Berrange 已提交
778
        return NULL;
779
    }
D
Daniel P. Berrange 已提交
780

781 782 783 784 785 786 787
    if (virMutexInit(&server->lock) < 0) {
        VIR_ERROR("%s", _("cannot initialize mutex"));
        VIR_FREE(server);
    }
    if (virCondInit(&server->job) < 0) {
        VIR_ERROR("%s", _("cannot initialize condition variable"));
        virMutexDestroy(&server->lock);
788 789 790
        VIR_FREE(server);
    }

791
    server->privileged = geteuid() == 0 ? 1 : 0;
792
    server->sigread = sigread;
D
Daniel P. Berrange 已提交
793

794
    if (virEventInit() < 0) {
795
        VIR_ERROR0(_("Failed to initialize event system"));
796 797 798 799
        VIR_FREE(server);
        return NULL;
    }

800 801
    virInitialize();

802 803 804 805 806 807 808
    /*
     * 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.
     */
809 810 811
#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
812
     * If they try to open a connection for a module that
813 814 815 816
     * is not loaded they'll get a suitable error at that point
     */
    virDriverLoadModule("network");
    virDriverLoadModule("storage");
817
    virDriverLoadModule("nodedev");
818
    virDriverLoadModule("secret");
819 820 821
    virDriverLoadModule("qemu");
    virDriverLoadModule("lxc");
    virDriverLoadModule("uml");
D
Daniel Veillard 已提交
822
    virDriverLoadModule("one");
823
#else
824 825 826
#ifdef WITH_NETWORK
    networkRegister();
#endif
827 828 829
#ifdef WITH_NETCF
    interfaceRegister();
#endif
830 831
#ifdef WITH_STORAGE_DIR
    storageRegister();
832
#endif
833 834
#if defined(WITH_NODE_DEVICES) && \
    (defined(HAVE_HAL) || defined(HAVE_DEVKIT))
835 836
    nodedevRegister();
#endif
837
    secretRegister();
838 839 840 841 842 843 844 845 846
#ifdef WITH_QEMU
    qemuRegister();
#endif
#ifdef WITH_LXC
    lxcRegister();
#endif
#ifdef WITH_UML
    umlRegister();
#endif
D
Daniel Veillard 已提交
847 848 849
#ifdef WITH_ONE
    oneRegister();
#endif
850 851
#endif

852 853 854 855 856 857
    virEventRegisterImpl(virEventAddHandleImpl,
                         virEventUpdateHandleImpl,
                         virEventRemoveHandleImpl,
                         virEventAddTimeoutImpl,
                         virEventUpdateTimeoutImpl,
                         virEventRemoveTimeoutImpl);
858

859
    virStateInitialize(server->privileged);
860

861 862 863
    return server;
}

864
static int qemudNetworkInit(struct qemud_server *server) {
865 866
    char sockname[PATH_MAX];
    char roSockname[PATH_MAX];
867
#if HAVE_SASL
868 869 870 871 872 873 874 875 876
    int err;
#endif /* HAVE_SASL */

    roSockname[0] = '\0';

    if (qemudInitPaths(server, sockname, roSockname, PATH_MAX) < 0)
        goto cleanup;

    if (qemudListenUnix(server, sockname, 0, auth_unix_rw) < 0)
877
        goto cleanup;
878 879 880 881 882 883 884 885 886 887

    if (roSockname[0] != '\0' && qemudListenUnix(server, roSockname, 1, auth_unix_ro) < 0)
        goto cleanup;

#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) {
888 889
            VIR_ERROR(_("Failed to initialize SASL authentication %s"),
                      sasl_errstring(err, NULL, NULL));
890 891
            goto cleanup;
        }
892 893 894
    }
#endif

895
#if HAVE_POLKIT0
896 897 898
    if (auth_unix_rw == REMOTE_AUTH_POLKIT ||
        auth_unix_ro == REMOTE_AUTH_POLKIT) {
        DBusError derr;
899 900 901 902

        dbus_connection_set_change_sigpipe(FALSE);
        dbus_threads_init_default();

903 904 905
        dbus_error_init(&derr);
        server->sysbus = dbus_bus_get(DBUS_BUS_SYSTEM, &derr);
        if (!(server->sysbus)) {
906 907
            VIR_ERROR(_("Failed to connect to system bus for PolicyKit auth: %s"),
                      derr.message);
908 909 910
            dbus_error_free(&derr);
            goto cleanup;
        }
911
        dbus_connection_set_exit_on_disconnect(server->sysbus, FALSE);
912 913 914
    }
#endif

915
    if (ipsock) {
916
        if (listen_tcp && remoteListenTCP (server, listen_addr, tcp_port, QEMUD_SOCK_TYPE_TCP, auth_tcp) < 0)
917 918 919 920 921 922
            goto cleanup;

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

923
            if (remoteListenTCP (server, listen_addr, tls_port, QEMUD_SOCK_TYPE_TLS, auth_tls) < 0)
924 925
                goto cleanup;
        }
D
Daniel P. Berrange 已提交
926 927
    }

928
#ifdef HAVE_AVAHI
929
    if (server->privileged && mdns_adv) {
930
        struct libvirtd_mdns_group *group;
931
        struct qemud_socket *sock;
932 933 934 935 936
        int port = 0;

        server->mdns = libvirtd_mdns_new();

        if (!mdns_name) {
937
            char groupname[64], *localhost, *tmp;
938
            /* Extract the host part of the potentially FQDN */
939 940 941 942 943
            localhost = virGetHostname();
            if (localhost == NULL) {
                virReportOOMError(NULL);
                goto cleanup;
            }
944 945 946 947 948
            if ((tmp = strchr(localhost, '.')))
                *tmp = '\0';
            snprintf(groupname, sizeof(groupname)-1, "Virtualization Host %s", localhost);
            groupname[sizeof(groupname)-1] = '\0';
            group = libvirtd_mdns_add_group(server->mdns, groupname);
949
            VIR_FREE(localhost);
950 951 952 953
        } else {
            group = libvirtd_mdns_add_group(server->mdns, mdns_name);
        }

954
        /*
955 956 957 958 959 960
         * 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) {
961
            if (sock->port != -1 && sock->type == QEMUD_SOCK_TYPE_TLS) {
962 963 964 965 966
                port = sock->port;
                break;
            }
            sock = sock->next;
        }
967

968 969 970 971 972 973 974 975 976
        /*
         * 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

977
    return 0;
D
Daniel P. Berrange 已提交
978 979

 cleanup:
980
    return -1;
D
Daniel P. Berrange 已提交
981 982
}

983

984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
static gnutls_session_t
remoteInitializeTLSSession (void)
{
  gnutls_session_t session;
  int err;

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

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

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

  /* request client certificate if any.
   */
  gnutls_certificate_server_set_request (session, GNUTLS_CERT_REQUEST);

  gnutls_dh_set_prime_bits (session, DH_BITS);

  return session;

 failed:
1011
  VIR_ERROR(_("remoteInitializeTLSSession: %s"),
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
            gnutls_strerror (err));
  return NULL;
}

/* Check DN is on tls_allowed_dn_list. */
static int
remoteCheckDN (gnutls_x509_crt_t cert)
{
    char name[256];
    size_t namesize = sizeof name;
1022
    char **wildcards;
1023 1024 1025 1026
    int err;

    err = gnutls_x509_crt_get_dn (cert, name, &namesize);
    if (err != 0) {
1027
        VIR_ERROR(_("remoteCheckDN: gnutls_x509_cert_get_dn: %s"),
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
                  gnutls_strerror (err));
        return 0;
    }

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

    while (*wildcards) {
        if (fnmatch (*wildcards, name, 0) == 0)
            return 1;
        wildcards++;
    }

    /* Print the client's DN. */
1044
    DEBUG(_("remoteCheckDN: failed: client DN is %s"), name);
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058

    return 0; // Not found.
}

static int
remoteCheckCertificate (gnutls_session_t session)
{
    int ret;
    unsigned int status;
    const gnutls_datum_t *certs;
    unsigned int nCerts, i;
    time_t now;

    if ((ret = gnutls_certificate_verify_peers2 (session, &status)) < 0){
1059
        VIR_ERROR(_("remoteCheckCertificate: verify failed: %s"),
1060 1061 1062 1063 1064 1065
                  gnutls_strerror (ret));
        return -1;
    }

    if (status != 0) {
        if (status & GNUTLS_CERT_INVALID)
1066 1067
            VIR_ERROR0(_("remoteCheckCertificate: "
                         "the client certificate is not trusted."));
1068 1069

        if (status & GNUTLS_CERT_SIGNER_NOT_FOUND)
1070 1071
            VIR_ERROR0(_("remoteCheckCertificate: the client "
                         "certificate has unknown issuer."));
1072 1073

        if (status & GNUTLS_CERT_REVOKED)
1074 1075
            VIR_ERROR0(_("remoteCheckCertificate: "
                         "the client certificate has been revoked."));
1076

1077
#ifndef GNUTLS_1_0_COMPAT
1078
        if (status & GNUTLS_CERT_INSECURE_ALGORITHM)
1079 1080
            VIR_ERROR0(_("remoteCheckCertificate: the client certificate"
                         " uses an insecure algorithm."));
1081
#endif
1082 1083 1084 1085 1086

        return -1;
    }

    if (gnutls_certificate_type_get (session) != GNUTLS_CRT_X509) {
1087
        VIR_ERROR0(_("remoteCheckCertificate: certificate is not X.509"));
1088 1089 1090 1091
        return -1;
    }

    if (!(certs = gnutls_certificate_get_peers(session, &nCerts))) {
1092
        VIR_ERROR0(_("remoteCheckCertificate: no peers"));
1093 1094 1095 1096 1097 1098 1099 1100 1101
        return -1;
    }

    now = time (NULL);

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

        if (gnutls_x509_crt_init (&cert) < 0) {
1102
            VIR_ERROR0(_("remoteCheckCertificate: gnutls_x509_crt_init failed"));
1103 1104 1105 1106 1107 1108 1109
            return -1;
        }

        if (gnutls_x509_crt_import(cert, &certs[i], GNUTLS_X509_FMT_DER) < 0) {
            gnutls_x509_crt_deinit (cert);
            return -1;
        }
1110

1111
        if (gnutls_x509_crt_get_expiration_time (cert) < now) {
1112 1113
            VIR_ERROR0(_("remoteCheckCertificate: "
                         "the client certificate has expired"));
1114 1115 1116
            gnutls_x509_crt_deinit (cert);
            return -1;
        }
1117

1118
        if (gnutls_x509_crt_get_activation_time (cert) > now) {
1119 1120
            VIR_ERROR0(_("remoteCheckCertificate: the client "
                         "certificate is not yet activated"));
1121 1122 1123 1124 1125 1126 1127
            gnutls_x509_crt_deinit (cert);
            return -1;
        }

        if (i == 0) {
            if (!remoteCheckDN (cert)) {
                /* This is the most common error: make it informative. */
1128
                VIR_ERROR0(_("remoteCheckCertificate: client's Distinguished Name is not on the list of allowed clients (tls_allowed_dn_list).  Use 'openssl x509 -in clientcert.pem -text' to view the Distinguished Name field in the client certificate, or run this daemon with --verbose option."));
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
                gnutls_x509_crt_deinit (cert);
                return -1;
            }
        }
    }

    return 0;
}

/* Check the client's access. */
static int
remoteCheckAccess (struct qemud_client *client)
{
1142 1143
    struct qemud_client_message *confirm;

1144
    /* Verify client certificate. */
1145
    if (remoteCheckCertificate (client->tlssession) == -1) {
1146 1147
        VIR_ERROR0(_("remoteCheckCertificate: "
                     "failed to verify client's certificate"));
1148
        if (!tls_no_verify_certificate) return -1;
1149 1150
        else VIR_INFO0(_("remoteCheckCertificate: tls_no_verify_certificate "
                          "is set so the bad certificate is ignored"));
1151 1152
    }

1153 1154 1155 1156 1157 1158 1159 1160 1161
    if (client->tx) {
        VIR_INFO("%s",
                 _("client had unexpected data pending tx after access check"));
        return -1;
    }

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

1162 1163 1164 1165
    /* 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).
     */
1166 1167 1168 1169 1170 1171
    confirm->async = 1;
    confirm->bufferLength = 1;
    confirm->bufferOffset = 0;
    confirm->buffer[0] = '\1';

    client->tx = confirm;
1172 1173
    return 0;
}
D
Daniel P. Berrange 已提交
1174

1175 1176 1177 1178 1179 1180 1181
#if HAVE_POLKIT
int qemudGetSocketIdentity(int fd, uid_t *uid, pid_t *pid) {
#ifdef SO_PEERCRED
    struct ucred cr;
    unsigned int cr_len = sizeof (cr);

    if (getsockopt (fd, SOL_SOCKET, SO_PEERCRED, &cr, &cr_len) < 0) {
1182
        char ebuf[1024];
1183
        VIR_ERROR(_("Failed to verify client credentials: %s"),
1184
                  virStrerror(errno, ebuf, sizeof ebuf));
1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
        return -1;
    }

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

1198

D
Daniel P. Berrange 已提交
1199 1200 1201
static int qemudDispatchServer(struct qemud_server *server, struct qemud_socket *sock) {
    int fd;
    struct sockaddr_storage addr;
1202
    socklen_t addrlen = (socklen_t) (sizeof addr);
D
Daniel P. Berrange 已提交
1203
    struct qemud_client *client;
1204
    int no_slow_start = 1;
D
Daniel P. Berrange 已提交
1205 1206

    if ((fd = accept(sock->fd, (struct sockaddr *)&addr, &addrlen)) < 0) {
1207
        char ebuf[1024];
D
Daniel P. Berrange 已提交
1208 1209
        if (errno == EAGAIN)
            return 0;
1210 1211
        VIR_ERROR(_("Failed to accept connection: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
1212 1213 1214
        return -1;
    }

1215
    if (server->nclients >= max_clients) {
1216
        VIR_ERROR(_("Too many active clients (%d), dropping connection"), max_clients);
1217 1218 1219 1220
        close(fd);
        return -1;
    }

1221
    if (VIR_REALLOC_N(server->clients, server->nclients+1) < 0) {
1222
        VIR_ERROR0(_("Out of memory allocating clients"));
1223 1224 1225 1226
        close(fd);
        return -1;
    }

J
John Levon 已提交
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
#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);
            close (fd);
            return -1;
        }

        if (!priv_ismember (privs, PRIV_VIRT_MANAGE)) {
            ucred_free (ucred);
            close (fd);
            return -1;
        }

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

1250 1251 1252 1253
    /* Disable Nagle.  Unix sockets will ignore this. */
    setsockopt (fd, IPPROTO_TCP, TCP_NODELAY, (void *)&no_slow_start,
                sizeof no_slow_start);

1254 1255
    if (virSetCloseExec(fd) < 0 ||
        virSetNonBlock(fd) < 0) {
D
Daniel P. Berrange 已提交
1256 1257 1258 1259
        close(fd);
        return -1;
    }

1260
    if (VIR_ALLOC(client) < 0)
1261
        goto cleanup;
1262 1263 1264
    if (virMutexInit(&client->lock) < 0) {
        VIR_ERROR("%s", _("cannot initialize mutex"));
        VIR_FREE(client);
1265
        goto cleanup;
1266
    }
1267

1268
    client->magic = QEMUD_CLIENT_MAGIC;
D
Daniel P. Berrange 已提交
1269 1270
    client->fd = fd;
    client->readonly = sock->readonly;
1271
    client->type = sock->type;
1272
    client->auth = sock->auth;
1273 1274 1275
    memcpy (&client->addr, &addr, sizeof addr);
    client->addrlen = addrlen;

1276 1277 1278 1279 1280 1281
    /* Prepare one for packet receive */
    if (VIR_ALLOC(client->rx) < 0)
        goto cleanup;
    client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;


1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293
#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)
            goto cleanup;

1294
        /* Client is running as root, so disable auth */
1295
        if (uid == 0) {
1296
            VIR_INFO(_("Turn off polkit auth for privileged client %d"), pid);
1297 1298 1299 1300 1301
            client->auth = REMOTE_AUTH_NONE;
        }
    }
#endif

1302
    if (client->type != QEMUD_SOCK_TYPE_TLS) {
1303
        /* Plain socket, so prepare to read first message */
1304
        if (qemudRegisterClientEvent (server, client) < 0)
D
Daniel P. Berrange 已提交
1305
            goto cleanup;
1306 1307 1308
    } else {
        int ret;

1309 1310
        client->tlssession = remoteInitializeTLSSession ();
        if (client->tlssession == NULL)
D
Daniel P. Berrange 已提交
1311
            goto cleanup;
1312

1313
        gnutls_transport_set_ptr (client->tlssession,
1314 1315 1316
                                  (gnutls_transport_ptr_t) (long) fd);

        /* Begin the TLS handshake. */
1317
        ret = gnutls_handshake (client->tlssession);
1318
        if (ret == 0) {
1319 1320
            client->handshake = 0;

1321 1322
            /* Unlikely, but ...  Next step is to check the certificate. */
            if (remoteCheckAccess (client) == -1)
D
Daniel P. Berrange 已提交
1323 1324
                goto cleanup;

1325
            /* Handshake & cert check OK,  so prepare to read first message */
1326
            if (qemudRegisterClientEvent(server, client) < 0)
D
Daniel P. Berrange 已提交
1327
                goto cleanup;
1328
        } else if (ret == GNUTLS_E_INTERRUPTED || ret == GNUTLS_E_AGAIN) {
1329 1330
            /* Most likely, need to do more handshake data */
            client->handshake = 1;
D
Daniel P. Berrange 已提交
1331

1332
            if (qemudRegisterClientEvent (server, client) < 0)
D
Daniel P. Berrange 已提交
1333
                goto cleanup;
1334
        } else {
1335
            VIR_ERROR(_("TLS handshake failed: %s"),
1336
                      gnutls_strerror (ret));
D
Daniel P. Berrange 已提交
1337
            goto cleanup;
1338 1339
        }
    }
D
Daniel P. Berrange 已提交
1340

1341
    server->clients[server->nclients++] = client;
D
Daniel P. Berrange 已提交
1342

1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
    if (server->nclients > server->nactiveworkers &&
        server->nactiveworkers < server->nworkers) {
        int i;
        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
1357
    return 0;
1358

D
Daniel P. Berrange 已提交
1359
 cleanup:
1360 1361
    if (client &&
        client->tlssession) gnutls_deinit (client->tlssession);
1362
    close (fd);
1363 1364
    VIR_FREE(client->rx);
    VIR_FREE(client);
1365
    return -1;
D
Daniel P. Berrange 已提交
1366 1367 1368
}


1369 1370 1371 1372 1373 1374 1375
/*
 * 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
 */
1376
void qemudDispatchClientFailure(struct qemud_client *client) {
1377 1378 1379 1380
    if (client->watch != -1) {
        virEventRemoveHandleImpl(client->watch);
        client->watch = -1;
    }
D
Daniel P. Berrange 已提交
1381

1382 1383
    /* Deregister event delivery callback */
    if(client->conn) {
1384
        DEBUG0("Deregistering to relay remote events");
1385 1386 1387
        virConnectDomainEventDeregister(client->conn, remoteRelayDomainEvent);
    }

1388
#if HAVE_SASL
1389 1390 1391 1392
    if (client->saslconn) {
        sasl_dispose(&client->saslconn);
        client->saslconn = NULL;
    }
1393
    free(client->saslUsername);
1394
    client->saslUsername = NULL;
1395
#endif
1396 1397 1398 1399 1400 1401 1402 1403
    if (client->tlssession) {
        gnutls_deinit (client->tlssession);
        client->tlssession = NULL;
    }
    if (client->fd != -1) {
        close(client->fd);
        client->fd = -1;
    }
1404 1405 1406 1407 1408 1409 1410 1411
}


/* Caller must hold server lock */
static struct qemud_client *qemudPendingJob(struct qemud_server *server)
{
    int i;
    for (i = 0 ; i < server->nclients ; i++) {
1412
        virMutexLock(&server->clients[i]->lock);
1413
        if (server->clients[i]->dx) {
1414 1415 1416
            /* Delibrately don't unlock client - caller wants the lock */
            return server->clients[i];
        }
1417
        virMutexUnlock(&server->clients[i]->lock);
1418 1419
    }
    return NULL;
D
Daniel P. Berrange 已提交
1420 1421
}

1422 1423
static void *qemudWorker(void *data)
{
1424 1425
    struct qemud_worker *worker = data;
    struct qemud_server *server = worker->server;
1426 1427

    while (1) {
1428
        struct qemud_client *client = NULL;
1429
        struct qemud_client_message *msg;
1430

1431
        virMutexLock(&server->lock);
1432 1433
        while (((client = qemudPendingJob(server)) == NULL) &&
               !worker->quitRequest) {
1434 1435 1436 1437 1438
            if (virCondWait(&server->job, &server->lock) < 0) {
                virMutexUnlock(&server->lock);
                return NULL;
            }
        }
1439 1440 1441 1442 1443 1444 1445
        if (worker->quitRequest) {
            if (client)
                virMutexUnlock(&client->lock);
            virMutexUnlock(&server->lock);
            return NULL;
        }
        worker->processingCall = 1;
1446
        virMutexUnlock(&server->lock);
1447 1448 1449 1450

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

1451
        /* Remove our message from dispatch queue while we use it */
1452
        msg = qemudClientMessageQueueServe(&client->dx);
1453 1454 1455

        /* This function drops the lock during dispatch,
         * and re-acquires it before returning */
1456
        if (remoteDispatchClientRequest (server, client, msg) < 0) {
1457
            VIR_FREE(msg);
1458 1459 1460 1461 1462
            qemudDispatchClientFailure(client);
            client->refs--;
            virMutexUnlock(&client->lock);
            continue;
        }
1463 1464

        client->refs--;
1465
        virMutexUnlock(&client->lock);
1466 1467 1468 1469

        virMutexLock(&server->lock);
        worker->processingCall = 0;
        virMutexUnlock(&server->lock);
1470 1471
    }
}
D
Daniel P. Berrange 已提交
1472

1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
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
1500

1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513
/*
 * 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) {
1514 1515
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1516 1517 1518
        qemudDispatchClientFailure(client);
        return -1;
    }
1519 1520 1521

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

1522
    if (!client->tlssession) {
1523
        char ebuf[1024];
1524 1525 1526 1527 1528 1529
        ret = read (client->fd, data, len);
        if (ret == -1 && (errno == EAGAIN ||
                          errno == EINTR))
            return 0;
        if (ret <= 0) {
            if (ret != 0)
1530 1531
                VIR_ERROR(_("read: %s"),
                          virStrerror (errno, ebuf, sizeof ebuf));
1532
            qemudDispatchClientFailure(client);
1533 1534 1535
            return -1;
        }
    } else {
1536
        ret = gnutls_record_recv (client->tlssession, data, len);
1537 1538 1539 1540 1541 1542 1543 1544 1545

        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);
1546 1547
            return -1;
        }
D
Daniel P. Berrange 已提交
1548
    }
1549

1550 1551 1552
    return ret;
}

1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570
/*
 * 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
1571 1572
}

1573
#if HAVE_SASL
1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
/*
 * 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;
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593

    /* 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) {
1594
        int ret;
1595
        char encoded[8192];
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
        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);
1608
            return -1;
1609
        }
1610 1611 1612 1613 1614 1615

        client->saslDecodedOffset = 0;
    }

    /* Some buffered decoded data to return now */
    got = client->saslDecodedLength - client->saslDecodedOffset;
1616
    want = client->rx->bufferLength - client->rx->bufferOffset;
1617 1618 1619 1620

    if (want > got)
        want = got;

1621
    memcpy(client->rx->buffer + client->rx->bufferOffset,
1622 1623
           client->saslDecoded + client->saslDecodedOffset, want);
    client->saslDecodedOffset += want;
1624
    client->rx->bufferOffset += want;
1625 1626 1627 1628 1629 1630

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

1631
    return want;
1632 1633 1634
}
#endif

1635 1636 1637 1638 1639
/*
 * 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) {
1640 1641
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_READ)
1642
        return qemudClientReadSASL(client);
1643 1644
    else
#endif
1645
        return qemudClientReadPlain(client);
1646 1647 1648
}


1649 1650 1651 1652 1653
/*
 * Read data until we get a complete message to process
 */
static void qemudDispatchClientRead(struct qemud_server *server,
                                    struct qemud_client *client) {
1654
    /*qemudDebug ("qemudDispatchClientRead: mode = %d", client->mode);*/
D
Daniel P. Berrange 已提交
1655

1656 1657 1658
readmore:
    if (qemudClientRead(client) < 0)
        return; /* Error */
1659

1660 1661
    if (client->rx->bufferOffset < client->rx->bufferLength)
        return; /* Still not read enough */
1662

1663 1664 1665 1666
    /* Either done with length word header */
    if (client->rx->bufferLength == REMOTE_MESSAGE_HEADER_XDR_LEN) {
        unsigned int len;
        XDR x;
1667

1668
        xdrmem_create(&x, client->rx->buffer, client->rx->bufferLength, XDR_DECODE);
1669

1670 1671
        if (!xdr_u_int(&x, &len)) {
            xdr_destroy (&x);
1672
            DEBUG0("Failed to decode packet length");
1673
            qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1674 1675
            return;
        }
1676
        xdr_destroy (&x);
1677

1678 1679 1680
        if (len < REMOTE_MESSAGE_HEADER_XDR_LEN) {
            DEBUG("Packet length %u too small", len);
            qemudDispatchClientFailure(client);
1681
            return;
D
Daniel P. Berrange 已提交
1682
        }
1683

1684 1685 1686 1687 1688 1689
        /* 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);
1690 1691 1692
            return;
        }

1693 1694
        /* Prepare to read rest of message */
        client->rx->bufferLength += len;
1695

1696
        qemudUpdateClientEvent(client);
D
Daniel P. Berrange 已提交
1697

1698 1699 1700 1701 1702
        /* Try and read payload immediately instead of going back
           into poll() because chances are the data is already
           waiting for us */
        goto readmore;
    } else {
1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715
        /* 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) {
1716 1717 1718
            int ret;
            ret = (filter->query)(client, msg, filter->opaque);
            if (ret == 1) {
1719 1720
                msg = NULL;
                break;
1721 1722 1723 1724
            } else if (ret == -1) {
                VIR_FREE(msg);
                qemudDispatchClientFailure(client);
                return;
1725 1726 1727 1728
            }
            filter = filter->next;
        }

1729
        /* Move completed message to the end of the dispatch queue */
1730 1731
        if (msg)
            qemudClientMessageQueuePush(&client->dx, msg);
1732 1733 1734 1735 1736 1737
        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
1738
        } else {
1739 1740 1741
            if (client->rx)
                client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;

1742 1743 1744 1745
            qemudUpdateClientEvent(client);

            /* Tell one of the workers to get on with it... */
            virCondSignal(&server->job);
D
Daniel P. Berrange 已提交
1746
        }
D
Daniel P. Berrange 已提交
1747 1748 1749 1750
    }
}


1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
/*
 * 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) {
1764 1765
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1766 1767 1768 1769
        qemudDispatchClientFailure(client);
        return -1;
    }

1770
    if (!client->tlssession) {
1771
        char ebuf[1024];
1772 1773 1774
        if ((ret = write(client->fd, data, len)) == -1) {
            if (errno == EAGAIN || errno == EINTR)
                return 0;
1775
            VIR_ERROR(_("write: %s"), virStrerror (errno, ebuf, sizeof ebuf));
1776
            qemudDispatchClientFailure(client);
1777 1778 1779
            return -1;
        }
    } else {
1780
        ret = gnutls_record_send (client->tlssession, data, len);
1781 1782 1783 1784 1785 1786 1787
        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);
1788 1789
            return -1;
        }
D
Daniel P. Berrange 已提交
1790
    }
1791 1792
    return ret;
}
1793

1794

1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810
/*
 * 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
1811 1812 1813
}


1814
#if HAVE_SASL
1815 1816 1817 1818 1819 1820 1821 1822 1823
/*
 * Send client->tx using SASL encoding
 *
 * Returns:
 *   -1 on error
 *    0 on EAGAIN
 *    n number of bytes
 */
static int qemudClientWriteSASL(struct qemud_client *client) {
1824 1825 1826 1827
    int ret;

    /* Not got any pending encoded data, so we need to encode raw stuff */
    if (client->saslEncoded == NULL) {
1828 1829 1830
        ret = sasl_encode(client->saslconn,
                          client->tx->buffer + client->tx->bufferOffset,
                          client->tx->bufferLength - client->tx->bufferOffset,
1831 1832 1833
                          &client->saslEncoded,
                          &client->saslEncodedLength);

1834 1835 1836 1837 1838 1839 1840
        if (ret != SASL_OK) {
            VIR_ERROR(_("failed to encode SASL data %s"),
                      sasl_errstring(ret, NULL, NULL));
            qemudDispatchClientFailure(client);
            return -1;
        }

1841 1842 1843 1844
        client->saslEncodedOffset = 0;
    }

    /* Send some of the encoded stuff out on the wire */
1845
    ret = qemudClientWriteBuf(client,
1846 1847 1848
                              client->saslEncoded + client->saslEncodedOffset,
                              client->saslEncodedLength - client->saslEncodedOffset);

1849 1850
    if (ret <= 0)
        return ret; /* -1 error, 0 == egain */
1851 1852 1853 1854 1855 1856 1857 1858

    /* 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;
1859 1860 1861

        /* Mark as complete, so caller detects completion */
        client->tx->bufferOffset = client->tx->bufferLength;
1862 1863
    }

1864
    return ret;
1865 1866 1867
}
#endif

1868 1869 1870 1871 1872 1873 1874 1875 1876
/*
 * 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) {
1877 1878
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_WRITE)
1879
        return qemudClientWriteSASL(client);
1880 1881
    else
#endif
1882
        return qemudClientWritePlain(client);
1883 1884 1885
}


1886 1887 1888 1889
void
qemudClientMessageRelease(struct qemud_client *client,
                          struct qemud_client_message *msg)
{
1890 1891 1892
    if (msg->streamTX) {
        remoteStreamMessageFinished(client, msg);
    } else if (!msg->async)
1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
        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);
}


1911 1912 1913 1914 1915
/*
 * Process all queued client->tx messages until
 * we would block on I/O
 */
static void
1916
qemudDispatchClientWrite(struct qemud_client *client) {
1917 1918
    while (client->tx) {
        ssize_t ret;
1919

1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933
        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);

1934
            qemudClientMessageRelease(client, reply);
1935

1936 1937
            if (client->closing)
                qemudDispatchClientFailure(client);
1938
         }
1939
    }
1940
}
1941

1942
static void
1943
qemudDispatchClientHandshake(struct qemud_client *client) {
1944 1945 1946 1947
    int ret;
    /* Continue the handshake. */
    ret = gnutls_handshake (client->tlssession);
    if (ret == 0) {
1948 1949
        client->handshake = 0;

1950 1951 1952
        /* Finished.  Next step is to check the certificate. */
        if (remoteCheckAccess (client) == -1)
            qemudDispatchClientFailure(client);
1953 1954
        else
            qemudUpdateClientEvent(client);
1955 1956 1957 1958 1959
    } 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 */
1960
        qemudUpdateClientEvent (client);
1961 1962 1963 1964 1965
    } else {
        /* Fatal error in handshake */
        VIR_ERROR(_("TLS handshake failed: %s"),
                  gnutls_strerror (ret));
        qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1966 1967 1968
    }
}

1969
static void
1970
qemudDispatchClientEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
1971
    struct qemud_server *server = (struct qemud_server *)opaque;
1972 1973
    struct qemud_client *client = NULL;
    int i;
1974

1975
    virMutexLock(&server->lock);
1976

1977
    for (i = 0 ; i < server->nclients ; i++) {
1978
        virMutexLock(&server->clients[i]->lock);
1979 1980
        if (server->clients[i]->watch == watch) {
            client = server->clients[i];
D
Daniel P. Berrange 已提交
1981
            break;
1982
        }
1983
        virMutexUnlock(&server->clients[i]->lock);
D
Daniel P. Berrange 已提交
1984
    }
1985

1986 1987
    virMutexUnlock(&server->lock);

1988
    if (!client) {
D
Daniel P. Berrange 已提交
1989
        return;
1990 1991
    }

1992 1993
    if (client->fd != fd) {
        virMutexUnlock(&client->lock);
1994
        return;
1995 1996 1997 1998 1999
    }

    if (events & (VIR_EVENT_HANDLE_WRITABLE |
                  VIR_EVENT_HANDLE_READABLE)) {
        if (client->handshake) {
2000
            qemudDispatchClientHandshake(client);
2001 2002
        } else {
            if (events & VIR_EVENT_HANDLE_WRITABLE)
2003
                qemudDispatchClientWrite(client);
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
            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);
2014

2015
    virMutexUnlock(&client->lock);
D
Daniel P. Berrange 已提交
2016 2017
}

2018 2019 2020 2021 2022 2023

/*
 * @client: a locked client object
 */
static int
qemudCalculateHandleMode(struct qemud_client *client) {
2024 2025 2026
    int mode = 0;

    if (client->handshake) {
2027
        if (gnutls_record_get_direction (client->tlssession) == 0)
2028
            mode |= VIR_EVENT_HANDLE_READABLE;
2029
        else
2030 2031 2032 2033 2034 2035
            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;
2036

2037 2038 2039 2040
        /* 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;
2041 2042
    }

2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060
    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
2061 2062 2063 2064

    return 0;
}

2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076
/*
 * @client: a locked client object
 */
void qemudUpdateClientEvent(struct qemud_client *client) {
    int mode;

    mode = qemudCalculateHandleMode(client);

    virEventUpdateHandleImpl(client->watch, mode);
}


2077
static void
2078
qemudDispatchServerEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
2079
    struct qemud_server *server = (struct qemud_server *)opaque;
2080 2081
    struct qemud_socket *sock;

2082
    virMutexLock(&server->lock);
2083 2084

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2085

D
Daniel P. Berrange 已提交
2086
    while (sock) {
2087
        if (sock->watch == watch)
D
Daniel P. Berrange 已提交
2088 2089 2090
            break;

        sock = sock->next;
D
Daniel P. Berrange 已提交
2091
    }
D
Daniel P. Berrange 已提交
2092

2093
    if (sock && sock->fd == fd && events)
D
Daniel P. Berrange 已提交
2094
        qemudDispatchServer(server, sock);
2095

2096
    virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
2097 2098 2099
}


2100
static int qemudOneLoop(void) {
2101
    sig_atomic_t errors;
D
Daniel P. Berrange 已提交
2102

D
Daniel P. Berrange 已提交
2103
    if (virEventRunOnce() < 0)
D
Daniel P. Berrange 已提交
2104 2105
        return -1;

2106 2107 2108
    /* Check for any signal handling errors and log them. */
    errors = sig_errors;
    if (errors) {
2109
        char ebuf[1024];
2110
        sig_errors -= errors;
2111
        VIR_ERROR(_("Signal handler reported %d errors: last error: %s"),
2112
                  errors, virStrerror (sig_lasterrno, ebuf, sizeof ebuf));
2113 2114 2115
        return -1;
    }

D
Daniel P. Berrange 已提交
2116 2117 2118
    return 0;
}

2119
static void qemudInactiveTimer(int timerid, void *data) {
2120
    struct qemud_server *server = (struct qemud_server *)data;
2121 2122 2123 2124 2125 2126 2127

    if (virStateActive() ||
        server->clients) {
        DEBUG0("Timer expired but still active, not shutting down");
        virEventUpdateTimeoutImpl(timerid, -1);
    } else {
        DEBUG0("Timer expired and inactive, shutting down");
2128 2129 2130 2131
        server->shutdown = 1;
    }
}

2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148
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);
    }

2149 2150 2151
    while (client->streams)
        remoteRemoveClientStream(client, client->streams);

2152 2153 2154 2155 2156 2157
    if (client->conn)
        virConnectClose(client->conn);
    virMutexDestroy(&client->lock);
    VIR_FREE(client);
}

2158
static int qemudRunLoop(struct qemud_server *server) {
2159
    int timerid = -1;
2160
    int ret = -1, i;
2161
    int timerActive = 0;
2162

2163
    virMutexLock(&server->lock);
2164

2165 2166 2167 2168 2169 2170 2171 2172
    if (timeout > 0 &&
        (timerid = virEventAddTimeoutImpl(-1,
                                          qemudInactiveTimer,
                                          server, NULL)) < 0) {
        VIR_ERROR0(_("Failed to register shutdown timeout"));
        return -1;
    }

2173 2174 2175 2176
    if (min_workers > max_workers)
        max_workers = min_workers;

    server->nworkers = max_workers;
2177
    if (VIR_ALLOC_N(server->workers, server->nworkers) < 0) {
2178
        VIR_ERROR0(_("Failed to allocate workers"));
2179 2180 2181
        return -1;
    }

2182 2183 2184 2185
    for (i = 0 ; i < min_workers ; i++) {
        if (qemudStartWorker(server, &server->workers[i]) < 0)
            goto cleanup;
        server->nactiveworkers++;
2186
    }
2187 2188 2189 2190 2191 2192

    for (;;) {
        /* A shutdown timeout is specified, so check
         * if any drivers have active state, if not
         * shutdown after timeout seconds
         */
2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207
        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;
                }
            }
2208 2209
        }

2210
        virMutexUnlock(&server->lock);
2211 2212 2213
        if (qemudOneLoop() < 0) {
            virMutexLock(&server->lock);
            DEBUG0("Loop iteration error, exiting\n");
2214
            break;
2215
        }
2216
        virMutexLock(&server->lock);
2217 2218 2219 2220

    reprocess:
        for (i = 0 ; i < server->nclients ; i++) {
            int inactive;
2221
            virMutexLock(&server->clients[i]->lock);
2222 2223
            inactive = server->clients[i]->fd == -1
                && server->clients[i]->refs == 0;
2224
            virMutexUnlock(&server->clients[i]->lock);
2225
            if (inactive) {
2226
                qemudFreeClient(server->clients[i]);
2227
                server->nclients--;
2228
                if (i < server->nclients)
2229 2230
                    memmove(server->clients + i,
                            server->clients + i + 1,
2231 2232 2233 2234 2235
                            sizeof (*server->clients) * (server->nclients - i));

                if (VIR_REALLOC_N(server->clients,
                                  server->nclients) < 0) {
                    ; /* ignore */
2236
                }
2237
                goto reprocess;
2238 2239
            }
        }
D
Daniel P. Berrange 已提交
2240

2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260
        /* 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--;
            }
        }

2261 2262 2263 2264
        if (server->shutdown) {
            ret = 0;
            break;
        }
2265 2266
    }

2267
cleanup:
2268
    for (i = 0 ; i < server->nworkers ; i++) {
2269 2270 2271 2272 2273 2274
        if (!server->workers[i].hasThread)
            continue;

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

2275
        virMutexUnlock(&server->lock);
2276
        pthread_join(server->workers[i].thread, NULL);
2277
        virMutexLock(&server->lock);
2278
        server->workers[i].hasThread = 0;
2279
    }
2280
    VIR_FREE(server->workers);
2281

2282
    virMutexUnlock(&server->lock);
2283
    return ret;
D
Daniel P. Berrange 已提交
2284 2285 2286
}

static void qemudCleanup(struct qemud_server *server) {
2287 2288
    struct qemud_socket *sock;

2289
    close(server->sigread);
2290 2291

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2292
    while (sock) {
2293
        struct qemud_socket *next = sock->next;
2294 2295
        if (sock->watch)
            virEventRemoveHandleImpl(sock->watch);
D
Daniel P. Berrange 已提交
2296
        close(sock->fd);
2297 2298
        free(sock);
        sock = next;
D
Daniel P. Berrange 已提交
2299
    }
2300
    free(server->logDir);
2301

2302
#ifdef HAVE_SASL
2303 2304 2305
    if (server->saslUsernameWhitelist) {
        char **list = server->saslUsernameWhitelist;
        while (*list) {
2306
            free(*list);
2307 2308
            list++;
        }
J
Jim Meyering 已提交
2309
        free(server->saslUsernameWhitelist);
2310
    }
2311
#endif
2312

2313 2314 2315 2316 2317
#if HAVE_POLKIT0
        if (server->sysbus)
            dbus_connection_unref(server->sysbus);
#endif

2318
    virStateCleanup();
2319

2320 2321 2322 2323 2324 2325
    if (virCondDestroy(&server->job) < 0) {
        ;
    }
    virMutexDestroy(&server->lock);

    VIR_FREE(server);
D
Daniel P. Berrange 已提交
2326 2327
}

2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
/* 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:
2344
        if (VIR_ALLOC_N(list, 2) < 0) {
2345
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2346 2347 2348 2349 2350
            return -1;
        }
        list[0] = strdup (p->str);
        list[1] = NULL;
        if (list[0] == NULL) {
2351
            VIR_ERROR(_("failed to allocate memory for %s config list value"),
2352
                      key);
2353
            VIR_FREE(list);
2354 2355 2356 2357 2358 2359 2360 2361 2362
            return -1;
        }
        break;

    case VIR_CONF_LIST: {
        int i, len = 0;
        virConfValuePtr pp;
        for (pp = p->list; pp; pp = pp->next)
            len++;
2363
        if (VIR_ALLOC_N(list, 1+len) < 0) {
2364
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2365 2366 2367 2368
            return -1;
        }
        for (i = 0, pp = p->list; pp; ++i, pp = pp->next) {
            if (pp->type != VIR_CONF_STRING) {
2369 2370
                VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
                            " must be a string or list of strings\n"),
2371
                          filename, key);
2372
                VIR_FREE(list);
2373 2374 2375 2376 2377 2378
                return -1;
            }
            list[i] = strdup (pp->str);
            if (list[i] == NULL) {
                int j;
                for (j = 0 ; j < i ; j++)
2379 2380
                    VIR_FREE(list[j]);
                VIR_FREE(list);
2381 2382
                VIR_ERROR(_("failed to allocate memory for %s config list value"),
                          key);
2383 2384 2385 2386 2387 2388 2389 2390 2391
                return -1;
            }

        }
        list[i] = NULL;
        break;
    }

    default:
2392 2393
        VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
                    " must be a string or list of strings\n"),
2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407
                  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) {
2408
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: invalid type:"
2409
                    " got %s; expected %s\n"), filename, key,
2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
                  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) {                                   \
2429 2430 2431
                char ebuf[1024];                                        \
                VIR_ERROR(_("remoteReadConfigFile: %s\n"),		\
                          virStrerror(errno, ebuf, sizeof ebuf));       \
2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447
                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)

2448 2449 2450 2451 2452 2453 2454 2455

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

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

2456
    if (checkType (p, filename, key, VIR_CONF_STRING) < 0)
2457 2458 2459 2460 2461 2462 2463 2464 2465 2466
        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;
2467 2468 2469 2470
#endif
#if HAVE_POLKIT
    } else if (STREQ(p->str, "polkit")) {
        *auth = REMOTE_AUTH_POLKIT;
2471 2472
#endif
    } else {
2473
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: unsupported auth %s\n"),
2474
                  filename, key, p->str);
2475 2476 2477 2478 2479 2480
        return -1;
    }

    return 0;
}

2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
#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

2501 2502 2503 2504 2505 2506
/*
 * 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.
 */
2507
static int
2508 2509 2510
qemudSetLogging(virConfPtr conf, const char *filename)
{
    int log_level = 0;
2511 2512
    char *log_filters = NULL;
    char *log_outputs = NULL;
2513 2514
    int ret = -1;

2515 2516
    virLogReset();

2517
    /*
2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529
     * 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.
2530
     */
2531
    GET_CONF_INT (conf, filename, log_level);
2532 2533 2534
    if (log_level != 0)
        virLogSetDefaultPriority(log_level);

2535
    virLogSetFromEnv();
2536 2537 2538 2539

    if (virLogGetNbFilters() == 0) {
        GET_CONF_STR (conf, filename, log_filters);
        virLogParseFilters(log_filters);
2540
    }
2541

2542 2543 2544
    if (virLogGetNbOutputs() == 0) {
        GET_CONF_STR (conf, filename, log_outputs);
        virLogParseOutputs(log_outputs);
2545
    }
2546 2547

    /*
2548 2549
     * If no defined outputs, then direct to syslog when running
     * as daemon. Otherwise the default output is stderr.
2550
     */
2551
    if (virLogGetNbOutputs() == 0) {
2552
        char *tmp = NULL;
2553
        if (godaemon) {
2554 2555
            if (virAsprintf (&tmp, "%d:syslog:libvirtd",
                             virLogGetDefaultPriority()) < 0)
2556 2557
                goto free_and_fail;
        } else {
2558 2559
            if (virAsprintf (&tmp, "%d:stderr",
                             virLogGetDefaultPriority()) < 0)
2560
                goto free_and_fail;
2561
        }
2562 2563
        virLogParseOutputs(tmp);
        VIR_FREE(tmp);
2564
    }
2565 2566 2567 2568 2569 2570 2571

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

2572 2573
    ret = 0;

2574 2575 2576
free_and_fail:
    VIR_FREE(log_filters);
    VIR_FREE(log_outputs);
2577
    return(ret);
2578
}
2579

2580 2581 2582 2583
/* Read the config file if it exists.
 * Only used in the remote case, hence the name.
 */
static int
2584
remoteReadConfigFile (struct qemud_server *server, const char *filename)
2585 2586 2587
{
    virConfPtr conf;

2588 2589 2590 2591 2592
    /* 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;
2593
    char *buf = NULL;
2594

2595 2596
#if HAVE_POLKIT
    /* Change the default back to no auth for non-root */
2597
    if (!server->privileged && auth_unix_rw == REMOTE_AUTH_POLKIT)
2598
        auth_unix_rw = REMOTE_AUTH_NONE;
2599
    if (!server->privileged && auth_unix_ro == REMOTE_AUTH_POLKIT)
2600 2601 2602
        auth_unix_ro = REMOTE_AUTH_NONE;
#endif

2603
    conf = virConfReadFile (filename, 0);
2604
    if (!conf) return -1;
2605

2606 2607 2608
    /*
     * First get all the logging settings and activate them
     */
2609 2610
    if (qemudSetLogging(conf, filename) < 0)
        goto free_and_fail;
2611

2612 2613
    GET_CONF_INT (conf, filename, listen_tcp);
    GET_CONF_INT (conf, filename, listen_tls);
2614 2615
    GET_CONF_STR (conf, filename, tls_port);
    GET_CONF_STR (conf, filename, tcp_port);
2616
    GET_CONF_STR (conf, filename, listen_addr);
J
Jim Meyering 已提交
2617

2618
    if (remoteConfigGetAuth(conf, "auth_unix_rw", &auth_unix_rw, filename) < 0)
J
Jim Meyering 已提交
2619
        goto free_and_fail;
2620 2621 2622 2623 2624 2625 2626
#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
2627
    if (remoteConfigGetAuth(conf, "auth_unix_ro", &auth_unix_ro, filename) < 0)
J
Jim Meyering 已提交
2628
        goto free_and_fail;
2629
    if (remoteConfigGetAuth(conf, "auth_tcp", &auth_tcp, filename) < 0)
J
Jim Meyering 已提交
2630
        goto free_and_fail;
2631
    if (remoteConfigGetAuth(conf, "auth_tls", &auth_tls, filename) < 0)
J
Jim Meyering 已提交
2632
        goto free_and_fail;
2633

2634 2635
    GET_CONF_STR (conf, filename, unix_sock_group);
    if (unix_sock_group) {
2636
        if (!server->privileged) {
2637
            VIR_WARN0(_("Cannot set group when not running as root"));
2638
        } else {
2639
            int ret;
2640
            struct group grpdata, *grp;
2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661
            size_t maxbuf = sysconf(_SC_GETGR_R_SIZE_MAX);

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

            if (VIR_ALLOC_N(buf, maxbuf) < 0) {
                VIR_ERROR("%s", _("Failed to allocate memory for buffer"));
                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) {
                        VIR_ERROR("%s", _("Failed to reallocate enough memory for buffer"));
                        goto free_and_fail;
                    }
            }

            if (ret != 0 || !grp) {
2662
                VIR_ERROR(_("Failed to lookup group '%s'"), unix_sock_group);
2663
                goto free_and_fail;
2664 2665
            }
            unix_sock_gid = grp->gr_gid;
2666
            VIR_FREE (buf);
2667
        }
2668 2669
        free (unix_sock_group);
        unix_sock_group = NULL;
2670 2671
    }

2672 2673
    GET_CONF_STR (conf, filename, unix_sock_ro_perms);
    if (unix_sock_ro_perms) {
2674
        if (virStrToLong_i (unix_sock_ro_perms, NULL, 8, &unix_sock_ro_mask) != 0) {
2675
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_ro_perms);
2676
            goto free_and_fail;
2677
        }
2678 2679
        free (unix_sock_ro_perms);
        unix_sock_ro_perms = NULL;
2680 2681
    }

2682 2683
    GET_CONF_STR (conf, filename, unix_sock_rw_perms);
    if (unix_sock_rw_perms) {
2684
        if (virStrToLong_i (unix_sock_rw_perms, NULL, 8, &unix_sock_rw_mask) != 0) {
2685
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_rw_perms);
2686
            goto free_and_fail;
2687
        }
2688 2689
        free (unix_sock_rw_perms);
        unix_sock_rw_perms = NULL;
2690 2691
    }

2692 2693
    GET_CONF_STR (conf, filename, unix_sock_dir);

2694 2695
    GET_CONF_INT (conf, filename, mdns_adv);
    GET_CONF_STR (conf, filename, mdns_name);
2696

2697
    GET_CONF_INT (conf, filename, tls_no_verify_certificate);
2698

2699 2700 2701 2702
    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);
2703

2704 2705 2706
    if (remoteConfigGetStringList (conf, "tls_allowed_dn_list",
                                   &tls_allowed_dn_list, filename) < 0)
        goto free_and_fail;
2707

2708
    if (remoteReadSaslAllowedUsernameList (conf, server, filename) < 0)
2709
        goto free_and_fail;
2710

2711 2712 2713 2714 2715

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

2716 2717 2718
    GET_CONF_INT (conf, filename, max_requests);
    GET_CONF_INT (conf, filename, max_client_requests);

2719 2720
    virConfFree (conf);
    return 0;
2721

2722 2723 2724 2725 2726 2727 2728
 free_and_fail:
    virConfFree (conf);
    free (mdns_name);
    mdns_name = NULL;
    free (unix_sock_ro_perms);
    free (unix_sock_rw_perms);
    free (unix_sock_group);
2729
    VIR_FREE (buf);
2730

2731 2732 2733 2734
    /* 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
2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
       even be a real leak.  */

    if (tls_allowed_dn_list) {
        int i;
        for (i = 0; tls_allowed_dn_list[i]; i++)
            free (tls_allowed_dn_list[i]);
        free (tls_allowed_dn_list);
        tls_allowed_dn_list = NULL;
    }

    return -1;
2746 2747
}

2748 2749 2750 2751 2752 2753 2754
/* Display version information. */
static void
version (const char *argv0)
{
    printf ("%s (%s) %s\n", argv0, PACKAGE_NAME, PACKAGE_VERSION);
}

J
John Levon 已提交
2755 2756 2757 2758 2759 2760 2761 2762
#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)) {
2763
        VIR_ERROR0(_("additional privileges are required\n"));
J
John Levon 已提交
2764 2765 2766 2767 2768
        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)) {
2769
        VIR_ERROR0(_("failed to set reduced privileges\n"));
J
John Levon 已提交
2770 2771 2772 2773 2774 2775 2776 2777 2778
        return -1;
    }

    return 0;
}
#else
#define qemudSetupPrivs() 0
#endif

2779 2780 2781 2782 2783
/* Print command-line usage. */
static void
usage (const char *argv0)
{
    fprintf (stderr,
2784 2785 2786 2787 2788 2789 2790
             "\n\
Usage:\n\
  %s [options]\n\
\n\
Options:\n\
  -v | --verbose         Verbose messages.\n\
  -d | --daemon          Run as a daemon & write PID file.\n\
2791 2792
  -l | --listen          Listen for TCP/IP connections.\n\
  -t | --timeout <secs>  Exit after timeout period.\n\
2793
  -f | --config <file>   Configuration file.\n\
2794
     | --version         Display version information.\n\
2795 2796
  -p | --pid-file <file> Change name of PID file.\n\
\n\
2797
libvirt management daemon:\n\
2798 2799 2800
\n\
  Default paths:\n\
\n\
2801
    Configuration file (unless overridden by -c):\n\
2802 2803
      " SYSCONF_DIR "/libvirt/libvirtd.conf\n\
\n\
2804
    Sockets (as root):\n\
2805 2806
      " LOCAL_STATE_DIR "/run/libvirt/libvirt-sock\n\
      " LOCAL_STATE_DIR "/run/libvirt/libvirt-sock-ro\n\
2807 2808 2809
\n\
    Sockets (as non-root):\n\
      $HOME/.libvirt/libvirt-sock (in UNIX abstract namespace)\n\
2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822
\n\
    TLS:\n\
      CA certificate:     " LIBVIRT_CACERT "\n\
      Server certificate: " LIBVIRT_SERVERCERT "\n\
      Server private key: " LIBVIRT_SERVERKEY "\n\
\n\
    PID file (unless overridden by --pid-file):\n\
      %s\n\
\n",
             argv0,
             REMOTE_PID_FILE[0] != '\0'
               ? REMOTE_PID_FILE
               : "(disabled in ./configure)");
2823 2824
}

2825 2826 2827 2828
enum {
    OPT_VERSION = 129
};

D
Daniel P. Berrange 已提交
2829 2830
#define MAX_LISTEN 5
int main(int argc, char **argv) {
2831
    struct qemud_server *server = NULL;
2832 2833
    struct sigaction sig_action;
    int sigpipe[2];
2834
    const char *pid_file = NULL;
2835
    const char *remote_config_file = NULL;
2836
    int ret = 1;
D
Daniel P. Berrange 已提交
2837 2838 2839 2840

    struct option opts[] = {
        { "verbose", no_argument, &verbose, 1},
        { "daemon", no_argument, &godaemon, 1},
2841
        { "listen", no_argument, &ipsock, 1},
2842
        { "config", required_argument, NULL, 'f'},
2843 2844
        { "timeout", required_argument, NULL, 't'},
        { "pid-file", required_argument, NULL, 'p'},
2845
        { "version", no_argument, NULL, OPT_VERSION },
2846
        { "help", no_argument, NULL, '?' },
D
Daniel P. Berrange 已提交
2847 2848 2849 2850 2851 2852 2853 2854
        {0, 0, 0, 0}
    };

    while (1) {
        int optidx = 0;
        int c;
        char *tmp;

2855
        c = getopt_long(argc, argv, "ldf:p:t:v", opts, &optidx);
D
Daniel P. Berrange 已提交
2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870

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

        switch (c) {
        case 0:
            /* Got one of the flags */
            break;
        case 'v':
            verbose = 1;
            break;
        case 'd':
            godaemon = 1;
            break;
2871 2872
        case 'l':
            ipsock = 1;
D
Daniel P. Berrange 已提交
2873 2874 2875
            break;

        case 't':
2876
            if (virStrToLong_i(optarg, &tmp, 10, &timeout) != 0
2877 2878 2879
                || timeout <= 0
                /* Ensure that we can multiply by 1000 without overflowing.  */
                || timeout > INT_MAX / 1000)
D
Daniel P. Berrange 已提交
2880 2881
                timeout = -1;
            break;
2882 2883

        case 'p':
2884 2885 2886 2887 2888
            pid_file = optarg;
            break;

        case 'f':
            remote_config_file = optarg;
2889 2890
            break;

2891 2892 2893 2894
        case OPT_VERSION:
            version (argv[0]);
            return 0;

D
Daniel P. Berrange 已提交
2895
        case '?':
2896
            usage (argv[0]);
D
Daniel P. Berrange 已提交
2897 2898 2899
            return 2;

        default:
2900 2901 2902
            fprintf (stderr, "libvirtd: internal error: unknown flag: %c\n",
                     c);
            exit (1);
D
Daniel P. Berrange 已提交
2903 2904 2905
        }
    }

2906 2907 2908 2909
    if (remote_config_file == NULL) {
        static const char *default_config_file
            = SYSCONF_DIR "/libvirt/libvirtd.conf";
        remote_config_file =
2910
            (access(default_config_file, R_OK) == 0
2911 2912 2913 2914
             ? default_config_file
             : "/dev/null");
    }

2915
    if (godaemon) {
2916
        char ebuf[1024];
2917
        if (qemudGoDaemon() < 0) {
2918 2919
            VIR_ERROR(_("Failed to fork as daemon: %s"),
                      virStrerror(errno, ebuf, sizeof ebuf));
2920 2921 2922 2923 2924 2925
            goto error1;
        }
    }

    /* If running as root and no PID file is set, use the default */
    if (pid_file == NULL &&
2926
        geteuid() == 0 &&
2927 2928 2929 2930 2931 2932 2933 2934
        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 &&
        qemudWritePidFile (pid_file) < 0)
        goto error1;

2935
    if (pipe(sigpipe) < 0 ||
2936 2937 2938 2939
        virSetNonBlock(sigpipe[0]) < 0 ||
        virSetNonBlock(sigpipe[1]) < 0 ||
        virSetCloseExec(sigpipe[0]) < 0 ||
        virSetCloseExec(sigpipe[1]) < 0) {
2940 2941 2942
        char ebuf[1024];
        VIR_ERROR(_("Failed to create pipe: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
2943
        goto error2;
2944
    }
2945 2946
    sigwrite = sigpipe[1];

2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959
    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);

J
John Levon 已提交
2960
    /* Ensure the rundir exists (on tmpfs on some systems) */
2961
    if (geteuid() == 0) {
J
John Levon 已提交
2962 2963 2964 2965
        const char *rundir = LOCAL_STATE_DIR "/run/libvirt";

        if (mkdir (rundir, 0755)) {
            if (errno != EEXIST) {
2966 2967 2968
                char ebuf[1024];
                VIR_ERROR(_("unable to create rundir %s: %s"), rundir,
                          virStrerror(errno, ebuf, sizeof(ebuf)));
J
John Levon 已提交
2969 2970 2971 2972 2973
                return -1;
            }
        }
    }

2974 2975 2976 2977 2978 2979
    /* 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
     */
J
John Levon 已提交
2980 2981 2982
    if (qemudSetupPrivs() < 0)
        goto error2;

2983 2984
    if (!(server = qemudInitialize(sigpipe[0]))) {
        ret = 2;
2985
        goto error2;
2986
    }
2987

2988 2989
    /* Read the config file (if it exists). */
    if (remoteReadConfigFile (server, remote_config_file) < 0)
2990
        goto error2;
D
Daniel P. Berrange 已提交
2991

2992
    /* Change the group ownership of /var/run/libvirt to unix_sock_gid */
2993
    if (unix_sock_dir && server->privileged) {
2994 2995 2996
        if (chown(unix_sock_dir, -1, unix_sock_gid) < 0)
            VIR_ERROR(_("Failed to change group ownership of %s"),
                      unix_sock_dir);
2997 2998
    }

2999
    if (virEventAddHandleImpl(sigpipe[0],
3000
                              VIR_EVENT_HANDLE_READABLE,
3001
                              qemudDispatchSignalEvent,
3002
                              server, NULL) < 0) {
3003
        VIR_ERROR0(_("Failed to register callback for signal pipe"));
D
Daniel P. Berrange 已提交
3004 3005 3006 3007
        ret = 3;
        goto error2;
    }

3008
    if (qemudNetworkInit(server) < 0) {
3009 3010 3011 3012
        ret = 2;
        goto error2;
    }

3013
    qemudRunLoop(server);
D
Daniel P. Berrange 已提交
3014

3015 3016
    ret = 0;

3017
error2:
3018 3019
    if (server)
        qemudCleanup(server);
3020 3021 3022 3023 3024
    if (pid_file)
        unlink (pid_file);
    close(sigwrite);

error1:
3025
    virLogShutdown();
3026
    return ret;
D
Daniel P. Berrange 已提交
3027
}