You need to sign in or sign up before continuing.
libvirtd.c 86.2 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
    char *sock_dir;
    char *dir_prefix = NULL;
    int ret = -1;
    char *sock_dir_prefix = NULL;

D
Daniel P. Berrange 已提交
703
    if (unix_sock_dir) {
704
        sock_dir = unix_sock_dir;
D
Daniel P. Berrange 已提交
705 706 707 708 709 710 711
        /* Change the group ownership of /var/run/libvirt to unix_sock_gid */
        if (server->privileged) {
            if (chown(unix_sock_dir, -1, unix_sock_gid) < 0)
                VIR_ERROR(_("Failed to change group ownership of %s"),
                          unix_sock_dir);
        }
    } else {
712
        sock_dir = sockname;
713
        if (server->privileged) {
714 715 716 717 718 719 720 721 722
            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 {
723
            uid_t uid = geteuid();
724 725 726 727 728
            dir_prefix = virGetUserDirectory(NULL, uid);
            if (dir_prefix == NULL) {
                /* Do not diagnose here; virGetUserDirectory does that.  */
                goto snprintf_error;
            }
729

730 731 732 733
            if (snprintf(sock_dir, maxlen, "%s/.libvirt", dir_prefix) >= maxlen)
                goto snprintf_error;
        }
    }
734

735 736 737 738 739
    sock_dir_prefix = strdup (sock_dir);
    if (!sock_dir_prefix) {
        virReportOOMError(NULL);
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
740

741
    if (server->privileged) {
742 743 744 745
        if (snprintf (sockname, maxlen, "%s/libvirt-sock",
                      sock_dir_prefix) >= maxlen
            || (snprintf (roSockname, maxlen, "%s/libvirt-sock-ro",
                          sock_dir_prefix) >= maxlen))
746
            goto snprintf_error;
747
        unlink(sockname);
748
        unlink(roSockname);
D
Daniel P. Berrange 已提交
749
    } else {
750 751
        if (snprintf(sockname, maxlen, "@%s/libvirt-sock",
                     sock_dir_prefix) >= maxlen)
752
            goto snprintf_error;
753
    }
754

755 756 757 758 759 760 761
    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);
    }
762

763
    if (server->logDir == NULL)
764
        goto cleanup;
765 766

    ret = 0;
767 768

 snprintf_error:
769 770 771 772 773 774 775 776
    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
777 778
}

779 780 781 782 783 784
static void virshErrorHandler(void *opaque ATTRIBUTE_UNUSED, virErrorPtr err ATTRIBUTE_UNUSED)
{
    /* Don't do anything, since logging infrastructure already
     * took care of reporting the error */
}

785
static struct qemud_server *qemudInitialize(int sigread) {
D
Daniel P. Berrange 已提交
786 787
    struct qemud_server *server;

788
    if (VIR_ALLOC(server) < 0) {
789
        VIR_ERROR0(_("Failed to allocate struct qemud_server"));
D
Daniel P. Berrange 已提交
790
        return NULL;
791
    }
D
Daniel P. Berrange 已提交
792

793 794 795 796 797 798 799
    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);
800 801 802
        VIR_FREE(server);
    }

803
    server->privileged = geteuid() == 0 ? 1 : 0;
804
    server->sigread = sigread;
D
Daniel P. Berrange 已提交
805

806
    if (virEventInit() < 0) {
807
        VIR_ERROR0(_("Failed to initialize event system"));
808 809 810 811
        VIR_FREE(server);
        return NULL;
    }

812 813 814 815 816 817 818
    /*
     * 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.
     */
819 820 821
#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
822
     * If they try to open a connection for a module that
823 824 825 826
     * is not loaded they'll get a suitable error at that point
     */
    virDriverLoadModule("network");
    virDriverLoadModule("storage");
827
    virDriverLoadModule("nodedev");
828
    virDriverLoadModule("secret");
829 830 831
    virDriverLoadModule("qemu");
    virDriverLoadModule("lxc");
    virDriverLoadModule("uml");
D
Daniel Veillard 已提交
832
    virDriverLoadModule("one");
833
#else
834 835 836
#ifdef WITH_NETWORK
    networkRegister();
#endif
837 838 839
#ifdef WITH_NETCF
    interfaceRegister();
#endif
840 841
#ifdef WITH_STORAGE_DIR
    storageRegister();
842
#endif
843 844
#if defined(WITH_NODE_DEVICES) && \
    (defined(HAVE_HAL) || defined(HAVE_DEVKIT))
845 846
    nodedevRegister();
#endif
847
    secretRegister();
848 849 850 851 852 853 854 855 856
#ifdef WITH_QEMU
    qemuRegister();
#endif
#ifdef WITH_LXC
    lxcRegister();
#endif
#ifdef WITH_UML
    umlRegister();
#endif
D
Daniel Veillard 已提交
857 858 859
#ifdef WITH_ONE
    oneRegister();
#endif
860 861
#endif

862 863 864 865 866 867
    virEventRegisterImpl(virEventAddHandleImpl,
                         virEventUpdateHandleImpl,
                         virEventRemoveHandleImpl,
                         virEventAddTimeoutImpl,
                         virEventUpdateTimeoutImpl,
                         virEventRemoveTimeoutImpl);
868

869
    virStateInitialize(server->privileged);
870

871 872 873
    return server;
}

874
static int qemudNetworkInit(struct qemud_server *server) {
875 876
    char sockname[PATH_MAX];
    char roSockname[PATH_MAX];
877
#if HAVE_SASL
878 879 880 881 882 883 884 885 886
    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)
887
        goto cleanup;
888 889 890 891 892 893 894 895 896 897

    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) {
898 899
            VIR_ERROR(_("Failed to initialize SASL authentication %s"),
                      sasl_errstring(err, NULL, NULL));
900 901
            goto cleanup;
        }
902 903 904
    }
#endif

905
#if HAVE_POLKIT0
906 907 908
    if (auth_unix_rw == REMOTE_AUTH_POLKIT ||
        auth_unix_ro == REMOTE_AUTH_POLKIT) {
        DBusError derr;
909 910 911 912

        dbus_connection_set_change_sigpipe(FALSE);
        dbus_threads_init_default();

913 914 915
        dbus_error_init(&derr);
        server->sysbus = dbus_bus_get(DBUS_BUS_SYSTEM, &derr);
        if (!(server->sysbus)) {
916 917
            VIR_ERROR(_("Failed to connect to system bus for PolicyKit auth: %s"),
                      derr.message);
918 919 920
            dbus_error_free(&derr);
            goto cleanup;
        }
921
        dbus_connection_set_exit_on_disconnect(server->sysbus, FALSE);
922 923 924
    }
#endif

925
    if (ipsock) {
926
        if (listen_tcp && remoteListenTCP (server, listen_addr, tcp_port, QEMUD_SOCK_TYPE_TCP, auth_tcp) < 0)
927 928 929 930 931 932
            goto cleanup;

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

933
            if (remoteListenTCP (server, listen_addr, tls_port, QEMUD_SOCK_TYPE_TLS, auth_tls) < 0)
934 935
                goto cleanup;
        }
D
Daniel P. Berrange 已提交
936 937
    }

938
#ifdef HAVE_AVAHI
939
    if (server->privileged && mdns_adv) {
940
        struct libvirtd_mdns_group *group;
941
        struct qemud_socket *sock;
942 943 944 945 946
        int port = 0;

        server->mdns = libvirtd_mdns_new();

        if (!mdns_name) {
947
            char groupname[64], *localhost, *tmp;
948
            /* Extract the host part of the potentially FQDN */
949 950 951 952 953
            localhost = virGetHostname();
            if (localhost == NULL) {
                virReportOOMError(NULL);
                goto cleanup;
            }
954 955 956 957 958
            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);
959
            VIR_FREE(localhost);
960 961 962 963
        } else {
            group = libvirtd_mdns_add_group(server->mdns, mdns_name);
        }

964
        /*
965 966 967 968 969 970
         * 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) {
971
            if (sock->port != -1 && sock->type == QEMUD_SOCK_TYPE_TLS) {
972 973 974 975 976
                port = sock->port;
                break;
            }
            sock = sock->next;
        }
977

978 979 980 981 982 983 984 985 986
        /*
         * 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

987
    return 0;
D
Daniel P. Berrange 已提交
988 989

 cleanup:
990
    return -1;
D
Daniel P. Berrange 已提交
991 992
}

993

994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
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:
1021
  VIR_ERROR(_("remoteInitializeTLSSession: %s"),
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
            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;
1032
    char **wildcards;
1033 1034 1035 1036
    int err;

    err = gnutls_x509_crt_get_dn (cert, name, &namesize);
    if (err != 0) {
1037
        VIR_ERROR(_("remoteCheckDN: gnutls_x509_cert_get_dn: %s"),
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
                  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. */
1054
    DEBUG(_("remoteCheckDN: failed: client DN is %s"), name);
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068

    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){
1069
        VIR_ERROR(_("remoteCheckCertificate: verify failed: %s"),
1070 1071 1072 1073 1074 1075
                  gnutls_strerror (ret));
        return -1;
    }

    if (status != 0) {
        if (status & GNUTLS_CERT_INVALID)
1076 1077
            VIR_ERROR0(_("remoteCheckCertificate: "
                         "the client certificate is not trusted."));
1078 1079

        if (status & GNUTLS_CERT_SIGNER_NOT_FOUND)
1080 1081
            VIR_ERROR0(_("remoteCheckCertificate: the client "
                         "certificate has unknown issuer."));
1082 1083

        if (status & GNUTLS_CERT_REVOKED)
1084 1085
            VIR_ERROR0(_("remoteCheckCertificate: "
                         "the client certificate has been revoked."));
1086

1087
#ifndef GNUTLS_1_0_COMPAT
1088
        if (status & GNUTLS_CERT_INSECURE_ALGORITHM)
1089 1090
            VIR_ERROR0(_("remoteCheckCertificate: the client certificate"
                         " uses an insecure algorithm."));
1091
#endif
1092 1093 1094 1095 1096

        return -1;
    }

    if (gnutls_certificate_type_get (session) != GNUTLS_CRT_X509) {
1097
        VIR_ERROR0(_("remoteCheckCertificate: certificate is not X.509"));
1098 1099 1100 1101
        return -1;
    }

    if (!(certs = gnutls_certificate_get_peers(session, &nCerts))) {
1102
        VIR_ERROR0(_("remoteCheckCertificate: no peers"));
1103 1104 1105 1106 1107 1108 1109 1110 1111
        return -1;
    }

    now = time (NULL);

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

        if (gnutls_x509_crt_init (&cert) < 0) {
1112
            VIR_ERROR0(_("remoteCheckCertificate: gnutls_x509_crt_init failed"));
1113 1114 1115 1116 1117 1118 1119
            return -1;
        }

        if (gnutls_x509_crt_import(cert, &certs[i], GNUTLS_X509_FMT_DER) < 0) {
            gnutls_x509_crt_deinit (cert);
            return -1;
        }
1120

1121
        if (gnutls_x509_crt_get_expiration_time (cert) < now) {
1122 1123
            VIR_ERROR0(_("remoteCheckCertificate: "
                         "the client certificate has expired"));
1124 1125 1126
            gnutls_x509_crt_deinit (cert);
            return -1;
        }
1127

1128
        if (gnutls_x509_crt_get_activation_time (cert) > now) {
1129 1130
            VIR_ERROR0(_("remoteCheckCertificate: the client "
                         "certificate is not yet activated"));
1131 1132 1133 1134 1135 1136 1137
            gnutls_x509_crt_deinit (cert);
            return -1;
        }

        if (i == 0) {
            if (!remoteCheckDN (cert)) {
                /* This is the most common error: make it informative. */
1138
                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."));
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151
                gnutls_x509_crt_deinit (cert);
                return -1;
            }
        }
    }

    return 0;
}

/* Check the client's access. */
static int
remoteCheckAccess (struct qemud_client *client)
{
1152 1153
    struct qemud_client_message *confirm;

1154
    /* Verify client certificate. */
1155
    if (remoteCheckCertificate (client->tlssession) == -1) {
1156 1157
        VIR_ERROR0(_("remoteCheckCertificate: "
                     "failed to verify client's certificate"));
1158
        if (!tls_no_verify_certificate) return -1;
1159 1160
        else VIR_INFO0(_("remoteCheckCertificate: tls_no_verify_certificate "
                          "is set so the bad certificate is ignored"));
1161 1162
    }

1163 1164 1165 1166 1167 1168 1169 1170 1171
    if (client->tx) {
        VIR_INFO("%s",
                 _("client had unexpected data pending tx after access check"));
        return -1;
    }

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

1172 1173 1174 1175
    /* 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).
     */
1176 1177 1178 1179 1180 1181
    confirm->async = 1;
    confirm->bufferLength = 1;
    confirm->bufferOffset = 0;
    confirm->buffer[0] = '\1';

    client->tx = confirm;
1182 1183
    return 0;
}
D
Daniel P. Berrange 已提交
1184

1185 1186 1187 1188 1189 1190 1191
#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) {
1192
        char ebuf[1024];
1193
        VIR_ERROR(_("Failed to verify client credentials: %s"),
1194
                  virStrerror(errno, ebuf, sizeof ebuf));
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207
        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

1208

D
Daniel P. Berrange 已提交
1209 1210 1211
static int qemudDispatchServer(struct qemud_server *server, struct qemud_socket *sock) {
    int fd;
    struct sockaddr_storage addr;
1212
    socklen_t addrlen = (socklen_t) (sizeof addr);
D
Daniel P. Berrange 已提交
1213
    struct qemud_client *client;
1214
    int no_slow_start = 1;
D
Daniel P. Berrange 已提交
1215 1216

    if ((fd = accept(sock->fd, (struct sockaddr *)&addr, &addrlen)) < 0) {
1217
        char ebuf[1024];
D
Daniel P. Berrange 已提交
1218 1219
        if (errno == EAGAIN)
            return 0;
1220 1221
        VIR_ERROR(_("Failed to accept connection: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
D
Daniel P. Berrange 已提交
1222 1223 1224
        return -1;
    }

1225
    if (server->nclients >= max_clients) {
1226
        VIR_ERROR(_("Too many active clients (%d), dropping connection"), max_clients);
1227 1228 1229 1230
        close(fd);
        return -1;
    }

1231
    if (VIR_REALLOC_N(server->clients, server->nclients+1) < 0) {
1232
        VIR_ERROR0(_("Out of memory allocating clients"));
1233 1234 1235 1236
        close(fd);
        return -1;
    }

J
John Levon 已提交
1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
#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 */

1260 1261 1262 1263
    /* Disable Nagle.  Unix sockets will ignore this. */
    setsockopt (fd, IPPROTO_TCP, TCP_NODELAY, (void *)&no_slow_start,
                sizeof no_slow_start);

1264 1265
    if (virSetCloseExec(fd) < 0 ||
        virSetNonBlock(fd) < 0) {
D
Daniel P. Berrange 已提交
1266 1267 1268 1269
        close(fd);
        return -1;
    }

1270
    if (VIR_ALLOC(client) < 0)
1271
        goto cleanup;
1272 1273 1274
    if (virMutexInit(&client->lock) < 0) {
        VIR_ERROR("%s", _("cannot initialize mutex"));
        VIR_FREE(client);
1275
        goto cleanup;
1276
    }
1277

1278
    client->magic = QEMUD_CLIENT_MAGIC;
D
Daniel P. Berrange 已提交
1279 1280
    client->fd = fd;
    client->readonly = sock->readonly;
1281
    client->type = sock->type;
1282
    client->auth = sock->auth;
1283 1284 1285
    memcpy (&client->addr, &addr, sizeof addr);
    client->addrlen = addrlen;

1286 1287 1288 1289 1290 1291
    /* Prepare one for packet receive */
    if (VIR_ALLOC(client->rx) < 0)
        goto cleanup;
    client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;


1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
#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;

1304
        /* Client is running as root, so disable auth */
1305
        if (uid == 0) {
1306
            VIR_INFO(_("Turn off polkit auth for privileged client %d"), pid);
1307 1308 1309 1310 1311
            client->auth = REMOTE_AUTH_NONE;
        }
    }
#endif

1312
    if (client->type != QEMUD_SOCK_TYPE_TLS) {
1313
        /* Plain socket, so prepare to read first message */
1314
        if (qemudRegisterClientEvent (server, client) < 0)
D
Daniel P. Berrange 已提交
1315
            goto cleanup;
1316 1317 1318
    } else {
        int ret;

1319 1320
        client->tlssession = remoteInitializeTLSSession ();
        if (client->tlssession == NULL)
D
Daniel P. Berrange 已提交
1321
            goto cleanup;
1322

1323
        gnutls_transport_set_ptr (client->tlssession,
1324 1325 1326
                                  (gnutls_transport_ptr_t) (long) fd);

        /* Begin the TLS handshake. */
1327
        ret = gnutls_handshake (client->tlssession);
1328
        if (ret == 0) {
1329 1330
            client->handshake = 0;

1331 1332
            /* Unlikely, but ...  Next step is to check the certificate. */
            if (remoteCheckAccess (client) == -1)
D
Daniel P. Berrange 已提交
1333 1334
                goto cleanup;

1335
            /* Handshake & cert check OK,  so prepare to read first message */
1336
            if (qemudRegisterClientEvent(server, client) < 0)
D
Daniel P. Berrange 已提交
1337
                goto cleanup;
1338
        } else if (ret == GNUTLS_E_INTERRUPTED || ret == GNUTLS_E_AGAIN) {
1339 1340
            /* Most likely, need to do more handshake data */
            client->handshake = 1;
D
Daniel P. Berrange 已提交
1341

1342
            if (qemudRegisterClientEvent (server, client) < 0)
D
Daniel P. Berrange 已提交
1343
                goto cleanup;
1344
        } else {
1345
            VIR_ERROR(_("TLS handshake failed: %s"),
1346
                      gnutls_strerror (ret));
D
Daniel P. Berrange 已提交
1347
            goto cleanup;
1348 1349
        }
    }
D
Daniel P. Berrange 已提交
1350

1351
    server->clients[server->nclients++] = client;
D
Daniel P. Berrange 已提交
1352

1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
    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
1367
    return 0;
1368

D
Daniel P. Berrange 已提交
1369
 cleanup:
1370 1371
    if (client &&
        client->tlssession) gnutls_deinit (client->tlssession);
1372
    close (fd);
1373 1374
    VIR_FREE(client->rx);
    VIR_FREE(client);
1375
    return -1;
D
Daniel P. Berrange 已提交
1376 1377 1378
}


1379 1380 1381 1382 1383 1384 1385
/*
 * 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
 */
1386
void qemudDispatchClientFailure(struct qemud_client *client) {
1387 1388 1389 1390
    if (client->watch != -1) {
        virEventRemoveHandleImpl(client->watch);
        client->watch = -1;
    }
D
Daniel P. Berrange 已提交
1391

1392 1393
    /* Deregister event delivery callback */
    if(client->conn) {
1394
        DEBUG0("Deregistering to relay remote events");
1395 1396 1397
        virConnectDomainEventDeregister(client->conn, remoteRelayDomainEvent);
    }

1398
#if HAVE_SASL
1399 1400 1401 1402
    if (client->saslconn) {
        sasl_dispose(&client->saslconn);
        client->saslconn = NULL;
    }
1403
    free(client->saslUsername);
1404
    client->saslUsername = NULL;
1405
#endif
1406 1407 1408 1409 1410 1411 1412 1413
    if (client->tlssession) {
        gnutls_deinit (client->tlssession);
        client->tlssession = NULL;
    }
    if (client->fd != -1) {
        close(client->fd);
        client->fd = -1;
    }
1414 1415 1416 1417 1418 1419 1420 1421
}


/* Caller must hold server lock */
static struct qemud_client *qemudPendingJob(struct qemud_server *server)
{
    int i;
    for (i = 0 ; i < server->nclients ; i++) {
1422
        virMutexLock(&server->clients[i]->lock);
1423
        if (server->clients[i]->dx) {
1424 1425 1426
            /* Delibrately don't unlock client - caller wants the lock */
            return server->clients[i];
        }
1427
        virMutexUnlock(&server->clients[i]->lock);
1428 1429
    }
    return NULL;
D
Daniel P. Berrange 已提交
1430 1431
}

1432 1433
static void *qemudWorker(void *data)
{
1434 1435
    struct qemud_worker *worker = data;
    struct qemud_server *server = worker->server;
1436 1437

    while (1) {
1438
        struct qemud_client *client = NULL;
1439
        struct qemud_client_message *msg;
1440

1441
        virMutexLock(&server->lock);
1442 1443
        while (((client = qemudPendingJob(server)) == NULL) &&
               !worker->quitRequest) {
1444 1445 1446 1447 1448
            if (virCondWait(&server->job, &server->lock) < 0) {
                virMutexUnlock(&server->lock);
                return NULL;
            }
        }
1449 1450 1451 1452 1453 1454 1455
        if (worker->quitRequest) {
            if (client)
                virMutexUnlock(&client->lock);
            virMutexUnlock(&server->lock);
            return NULL;
        }
        worker->processingCall = 1;
1456
        virMutexUnlock(&server->lock);
1457 1458 1459 1460

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

1461
        /* Remove our message from dispatch queue while we use it */
1462
        msg = qemudClientMessageQueueServe(&client->dx);
1463 1464 1465

        /* This function drops the lock during dispatch,
         * and re-acquires it before returning */
1466
        if (remoteDispatchClientRequest (server, client, msg) < 0) {
1467
            VIR_FREE(msg);
1468 1469 1470 1471 1472
            qemudDispatchClientFailure(client);
            client->refs--;
            virMutexUnlock(&client->lock);
            continue;
        }
1473 1474

        client->refs--;
1475
        virMutexUnlock(&client->lock);
1476 1477 1478 1479

        virMutexLock(&server->lock);
        worker->processingCall = 0;
        virMutexUnlock(&server->lock);
1480 1481
    }
}
D
Daniel P. Berrange 已提交
1482

1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509
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
1510

1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
/*
 * 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) {
1524 1525
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1526 1527 1528
        qemudDispatchClientFailure(client);
        return -1;
    }
1529 1530 1531

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

1532
    if (!client->tlssession) {
1533
        char ebuf[1024];
1534 1535 1536 1537 1538 1539
        ret = read (client->fd, data, len);
        if (ret == -1 && (errno == EAGAIN ||
                          errno == EINTR))
            return 0;
        if (ret <= 0) {
            if (ret != 0)
1540 1541
                VIR_ERROR(_("read: %s"),
                          virStrerror (errno, ebuf, sizeof ebuf));
1542
            qemudDispatchClientFailure(client);
1543 1544 1545
            return -1;
        }
    } else {
1546
        ret = gnutls_record_recv (client->tlssession, data, len);
1547 1548 1549 1550 1551 1552 1553 1554 1555

        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);
1556 1557
            return -1;
        }
D
Daniel P. Berrange 已提交
1558
    }
1559

1560 1561 1562
    return ret;
}

1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
/*
 * 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
1581 1582
}

1583
#if HAVE_SASL
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593
/*
 * 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;
1594 1595 1596 1597 1598 1599 1600 1601 1602 1603

    /* 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) {
1604
        int ret;
1605
        char encoded[8192];
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
        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);
1618
            return -1;
1619
        }
1620 1621 1622 1623 1624 1625

        client->saslDecodedOffset = 0;
    }

    /* Some buffered decoded data to return now */
    got = client->saslDecodedLength - client->saslDecodedOffset;
1626
    want = client->rx->bufferLength - client->rx->bufferOffset;
1627 1628 1629 1630

    if (want > got)
        want = got;

1631
    memcpy(client->rx->buffer + client->rx->bufferOffset,
1632 1633
           client->saslDecoded + client->saslDecodedOffset, want);
    client->saslDecodedOffset += want;
1634
    client->rx->bufferOffset += want;
1635 1636 1637 1638 1639 1640

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

1641
    return want;
1642 1643 1644
}
#endif

1645 1646 1647 1648 1649
/*
 * 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) {
1650 1651
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_READ)
1652
        return qemudClientReadSASL(client);
1653 1654
    else
#endif
1655
        return qemudClientReadPlain(client);
1656 1657 1658
}


1659 1660 1661 1662 1663
/*
 * Read data until we get a complete message to process
 */
static void qemudDispatchClientRead(struct qemud_server *server,
                                    struct qemud_client *client) {
1664
    /*qemudDebug ("qemudDispatchClientRead: mode = %d", client->mode);*/
D
Daniel P. Berrange 已提交
1665

1666 1667 1668
readmore:
    if (qemudClientRead(client) < 0)
        return; /* Error */
1669

1670 1671
    if (client->rx->bufferOffset < client->rx->bufferLength)
        return; /* Still not read enough */
1672

1673 1674 1675 1676
    /* Either done with length word header */
    if (client->rx->bufferLength == REMOTE_MESSAGE_HEADER_XDR_LEN) {
        unsigned int len;
        XDR x;
1677

1678
        xdrmem_create(&x, client->rx->buffer, client->rx->bufferLength, XDR_DECODE);
1679

1680 1681
        if (!xdr_u_int(&x, &len)) {
            xdr_destroy (&x);
1682
            DEBUG0("Failed to decode packet length");
1683
            qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1684 1685
            return;
        }
1686
        xdr_destroy (&x);
1687

1688 1689 1690
        if (len < REMOTE_MESSAGE_HEADER_XDR_LEN) {
            DEBUG("Packet length %u too small", len);
            qemudDispatchClientFailure(client);
1691
            return;
D
Daniel P. Berrange 已提交
1692
        }
1693

1694 1695 1696 1697 1698 1699
        /* 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);
1700 1701 1702
            return;
        }

1703 1704
        /* Prepare to read rest of message */
        client->rx->bufferLength += len;
1705

1706
        qemudUpdateClientEvent(client);
D
Daniel P. Berrange 已提交
1707

1708 1709 1710 1711 1712
        /* Try and read payload immediately instead of going back
           into poll() because chances are the data is already
           waiting for us */
        goto readmore;
    } else {
1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
        /* 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) {
1726 1727 1728
            int ret;
            ret = (filter->query)(client, msg, filter->opaque);
            if (ret == 1) {
1729 1730
                msg = NULL;
                break;
1731 1732 1733 1734
            } else if (ret == -1) {
                VIR_FREE(msg);
                qemudDispatchClientFailure(client);
                return;
1735 1736 1737 1738
            }
            filter = filter->next;
        }

1739
        /* Move completed message to the end of the dispatch queue */
1740 1741
        if (msg)
            qemudClientMessageQueuePush(&client->dx, msg);
1742 1743 1744 1745 1746 1747
        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
1748
        } else {
1749 1750 1751
            if (client->rx)
                client->rx->bufferLength = REMOTE_MESSAGE_HEADER_XDR_LEN;

1752 1753 1754 1755
            qemudUpdateClientEvent(client);

            /* Tell one of the workers to get on with it... */
            virCondSignal(&server->job);
D
Daniel P. Berrange 已提交
1756
        }
D
Daniel P. Berrange 已提交
1757 1758 1759 1760
    }
}


1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773
/*
 * 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) {
1774 1775
        VIR_ERROR(_("unexpected negative length request %lld"),
                  (long long int) len);
1776 1777 1778 1779
        qemudDispatchClientFailure(client);
        return -1;
    }

1780
    if (!client->tlssession) {
1781
        char ebuf[1024];
1782 1783 1784
        if ((ret = write(client->fd, data, len)) == -1) {
            if (errno == EAGAIN || errno == EINTR)
                return 0;
1785
            VIR_ERROR(_("write: %s"), virStrerror (errno, ebuf, sizeof ebuf));
1786
            qemudDispatchClientFailure(client);
1787 1788 1789
            return -1;
        }
    } else {
1790
        ret = gnutls_record_send (client->tlssession, data, len);
1791 1792 1793 1794 1795 1796 1797
        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);
1798 1799
            return -1;
        }
D
Daniel P. Berrange 已提交
1800
    }
1801 1802
    return ret;
}
1803

1804

1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820
/*
 * 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
1821 1822 1823
}


1824
#if HAVE_SASL
1825 1826 1827 1828 1829 1830 1831 1832 1833
/*
 * Send client->tx using SASL encoding
 *
 * Returns:
 *   -1 on error
 *    0 on EAGAIN
 *    n number of bytes
 */
static int qemudClientWriteSASL(struct qemud_client *client) {
1834 1835 1836 1837
    int ret;

    /* Not got any pending encoded data, so we need to encode raw stuff */
    if (client->saslEncoded == NULL) {
1838 1839 1840
        ret = sasl_encode(client->saslconn,
                          client->tx->buffer + client->tx->bufferOffset,
                          client->tx->bufferLength - client->tx->bufferOffset,
1841 1842 1843
                          &client->saslEncoded,
                          &client->saslEncodedLength);

1844 1845 1846 1847 1848 1849 1850
        if (ret != SASL_OK) {
            VIR_ERROR(_("failed to encode SASL data %s"),
                      sasl_errstring(ret, NULL, NULL));
            qemudDispatchClientFailure(client);
            return -1;
        }

1851 1852 1853 1854
        client->saslEncodedOffset = 0;
    }

    /* Send some of the encoded stuff out on the wire */
1855
    ret = qemudClientWriteBuf(client,
1856 1857 1858
                              client->saslEncoded + client->saslEncodedOffset,
                              client->saslEncodedLength - client->saslEncodedOffset);

1859 1860
    if (ret <= 0)
        return ret; /* -1 error, 0 == egain */
1861 1862 1863 1864 1865 1866 1867 1868

    /* 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;
1869 1870 1871

        /* Mark as complete, so caller detects completion */
        client->tx->bufferOffset = client->tx->bufferLength;
1872 1873
    }

1874
    return ret;
1875 1876 1877
}
#endif

1878 1879 1880 1881 1882 1883 1884 1885 1886
/*
 * 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) {
1887 1888
#if HAVE_SASL
    if (client->saslSSF & QEMUD_SASL_SSF_WRITE)
1889
        return qemudClientWriteSASL(client);
1890 1891
    else
#endif
1892
        return qemudClientWritePlain(client);
1893 1894 1895
}


1896 1897 1898 1899
void
qemudClientMessageRelease(struct qemud_client *client,
                          struct qemud_client_message *msg)
{
1900 1901 1902
    if (msg->streamTX) {
        remoteStreamMessageFinished(client, msg);
    } else if (!msg->async)
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
        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);
}


1921 1922 1923 1924 1925
/*
 * Process all queued client->tx messages until
 * we would block on I/O
 */
static void
1926
qemudDispatchClientWrite(struct qemud_client *client) {
1927 1928
    while (client->tx) {
        ssize_t ret;
1929

1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943
        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);

1944
            qemudClientMessageRelease(client, reply);
1945

1946 1947
            if (client->closing)
                qemudDispatchClientFailure(client);
1948
         }
1949
    }
1950
}
1951

1952
static void
1953
qemudDispatchClientHandshake(struct qemud_client *client) {
1954 1955 1956 1957
    int ret;
    /* Continue the handshake. */
    ret = gnutls_handshake (client->tlssession);
    if (ret == 0) {
1958 1959
        client->handshake = 0;

1960 1961 1962
        /* Finished.  Next step is to check the certificate. */
        if (remoteCheckAccess (client) == -1)
            qemudDispatchClientFailure(client);
1963 1964
        else
            qemudUpdateClientEvent(client);
1965 1966 1967 1968 1969
    } 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 */
1970
        qemudUpdateClientEvent (client);
1971 1972 1973 1974 1975
    } else {
        /* Fatal error in handshake */
        VIR_ERROR(_("TLS handshake failed: %s"),
                  gnutls_strerror (ret));
        qemudDispatchClientFailure(client);
D
Daniel P. Berrange 已提交
1976 1977 1978
    }
}

1979
static void
1980
qemudDispatchClientEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
1981
    struct qemud_server *server = (struct qemud_server *)opaque;
1982 1983
    struct qemud_client *client = NULL;
    int i;
1984

1985
    virMutexLock(&server->lock);
1986

1987
    for (i = 0 ; i < server->nclients ; i++) {
1988
        virMutexLock(&server->clients[i]->lock);
1989 1990
        if (server->clients[i]->watch == watch) {
            client = server->clients[i];
D
Daniel P. Berrange 已提交
1991
            break;
1992
        }
1993
        virMutexUnlock(&server->clients[i]->lock);
D
Daniel P. Berrange 已提交
1994
    }
1995

1996 1997
    virMutexUnlock(&server->lock);

1998
    if (!client) {
D
Daniel P. Berrange 已提交
1999
        return;
2000 2001
    }

2002 2003
    if (client->fd != fd) {
        virMutexUnlock(&client->lock);
2004
        return;
2005 2006 2007 2008 2009
    }

    if (events & (VIR_EVENT_HANDLE_WRITABLE |
                  VIR_EVENT_HANDLE_READABLE)) {
        if (client->handshake) {
2010
            qemudDispatchClientHandshake(client);
2011 2012
        } else {
            if (events & VIR_EVENT_HANDLE_WRITABLE)
2013
                qemudDispatchClientWrite(client);
2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
            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);
2024

2025
    virMutexUnlock(&client->lock);
D
Daniel P. Berrange 已提交
2026 2027
}

2028 2029 2030 2031 2032 2033

/*
 * @client: a locked client object
 */
static int
qemudCalculateHandleMode(struct qemud_client *client) {
2034 2035 2036
    int mode = 0;

    if (client->handshake) {
2037
        if (gnutls_record_get_direction (client->tlssession) == 0)
2038
            mode |= VIR_EVENT_HANDLE_READABLE;
2039
        else
2040 2041 2042 2043 2044 2045
            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;
2046

2047 2048 2049 2050
        /* 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;
2051 2052
    }

2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
    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
2071 2072 2073 2074

    return 0;
}

2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
/*
 * @client: a locked client object
 */
void qemudUpdateClientEvent(struct qemud_client *client) {
    int mode;

    mode = qemudCalculateHandleMode(client);

    virEventUpdateHandleImpl(client->watch, mode);
}


2087
static void
2088
qemudDispatchServerEvent(int watch, int fd, int events, void *opaque) {
D
Daniel P. Berrange 已提交
2089
    struct qemud_server *server = (struct qemud_server *)opaque;
2090 2091
    struct qemud_socket *sock;

2092
    virMutexLock(&server->lock);
2093 2094

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2095

D
Daniel P. Berrange 已提交
2096
    while (sock) {
2097
        if (sock->watch == watch)
D
Daniel P. Berrange 已提交
2098 2099 2100
            break;

        sock = sock->next;
D
Daniel P. Berrange 已提交
2101
    }
D
Daniel P. Berrange 已提交
2102

2103
    if (sock && sock->fd == fd && events)
D
Daniel P. Berrange 已提交
2104
        qemudDispatchServer(server, sock);
2105

2106
    virMutexUnlock(&server->lock);
D
Daniel P. Berrange 已提交
2107 2108 2109
}


2110
static int qemudOneLoop(void) {
2111
    sig_atomic_t errors;
D
Daniel P. Berrange 已提交
2112

D
Daniel P. Berrange 已提交
2113
    if (virEventRunOnce() < 0)
D
Daniel P. Berrange 已提交
2114 2115
        return -1;

2116 2117 2118
    /* Check for any signal handling errors and log them. */
    errors = sig_errors;
    if (errors) {
2119
        char ebuf[1024];
2120
        sig_errors -= errors;
2121
        VIR_ERROR(_("Signal handler reported %d errors: last error: %s"),
2122
                  errors, virStrerror (sig_lasterrno, ebuf, sizeof ebuf));
2123 2124 2125
        return -1;
    }

D
Daniel P. Berrange 已提交
2126 2127 2128
    return 0;
}

2129
static void qemudInactiveTimer(int timerid, void *data) {
2130
    struct qemud_server *server = (struct qemud_server *)data;
2131 2132 2133 2134 2135 2136 2137

    if (virStateActive() ||
        server->clients) {
        DEBUG0("Timer expired but still active, not shutting down");
        virEventUpdateTimeoutImpl(timerid, -1);
    } else {
        DEBUG0("Timer expired and inactive, shutting down");
2138 2139 2140 2141
        server->shutdown = 1;
    }
}

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158
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);
    }

2159 2160 2161
    while (client->streams)
        remoteRemoveClientStream(client, client->streams);

2162 2163 2164 2165 2166 2167
    if (client->conn)
        virConnectClose(client->conn);
    virMutexDestroy(&client->lock);
    VIR_FREE(client);
}

2168
static int qemudRunLoop(struct qemud_server *server) {
2169
    int timerid = -1;
2170
    int ret = -1, i;
2171
    int timerActive = 0;
2172

2173
    virMutexLock(&server->lock);
2174

2175 2176 2177 2178 2179 2180 2181 2182
    if (timeout > 0 &&
        (timerid = virEventAddTimeoutImpl(-1,
                                          qemudInactiveTimer,
                                          server, NULL)) < 0) {
        VIR_ERROR0(_("Failed to register shutdown timeout"));
        return -1;
    }

2183 2184 2185 2186
    if (min_workers > max_workers)
        max_workers = min_workers;

    server->nworkers = max_workers;
2187
    if (VIR_ALLOC_N(server->workers, server->nworkers) < 0) {
2188
        VIR_ERROR0(_("Failed to allocate workers"));
2189 2190 2191
        return -1;
    }

2192 2193 2194 2195
    for (i = 0 ; i < min_workers ; i++) {
        if (qemudStartWorker(server, &server->workers[i]) < 0)
            goto cleanup;
        server->nactiveworkers++;
2196
    }
2197 2198 2199 2200 2201 2202

    for (;;) {
        /* A shutdown timeout is specified, so check
         * if any drivers have active state, if not
         * shutdown after timeout seconds
         */
2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217
        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;
                }
            }
2218 2219
        }

2220
        virMutexUnlock(&server->lock);
2221 2222 2223
        if (qemudOneLoop() < 0) {
            virMutexLock(&server->lock);
            DEBUG0("Loop iteration error, exiting\n");
2224
            break;
2225
        }
2226
        virMutexLock(&server->lock);
2227 2228 2229 2230

    reprocess:
        for (i = 0 ; i < server->nclients ; i++) {
            int inactive;
2231
            virMutexLock(&server->clients[i]->lock);
2232 2233
            inactive = server->clients[i]->fd == -1
                && server->clients[i]->refs == 0;
2234
            virMutexUnlock(&server->clients[i]->lock);
2235
            if (inactive) {
2236
                qemudFreeClient(server->clients[i]);
2237
                server->nclients--;
2238
                if (i < server->nclients)
2239 2240
                    memmove(server->clients + i,
                            server->clients + i + 1,
2241 2242 2243 2244 2245
                            sizeof (*server->clients) * (server->nclients - i));

                if (VIR_REALLOC_N(server->clients,
                                  server->nclients) < 0) {
                    ; /* ignore */
2246
                }
2247
                goto reprocess;
2248 2249
            }
        }
D
Daniel P. Berrange 已提交
2250

2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270
        /* 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--;
            }
        }

2271 2272 2273 2274
        if (server->shutdown) {
            ret = 0;
            break;
        }
2275 2276
    }

2277
cleanup:
2278
    for (i = 0 ; i < server->nworkers ; i++) {
2279 2280 2281 2282 2283 2284
        if (!server->workers[i].hasThread)
            continue;

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

2285
        virMutexUnlock(&server->lock);
2286
        pthread_join(server->workers[i].thread, NULL);
2287
        virMutexLock(&server->lock);
2288
        server->workers[i].hasThread = 0;
2289
    }
2290
    VIR_FREE(server->workers);
2291

2292
    virMutexUnlock(&server->lock);
2293
    return ret;
D
Daniel P. Berrange 已提交
2294 2295 2296
}

static void qemudCleanup(struct qemud_server *server) {
2297 2298
    struct qemud_socket *sock;

2299
    close(server->sigread);
2300 2301

    sock = server->sockets;
D
Daniel P. Berrange 已提交
2302
    while (sock) {
2303
        struct qemud_socket *next = sock->next;
2304 2305
        if (sock->watch)
            virEventRemoveHandleImpl(sock->watch);
D
Daniel P. Berrange 已提交
2306
        close(sock->fd);
2307 2308
        free(sock);
        sock = next;
D
Daniel P. Berrange 已提交
2309
    }
2310
    free(server->logDir);
2311

2312
#ifdef HAVE_SASL
2313 2314 2315
    if (server->saslUsernameWhitelist) {
        char **list = server->saslUsernameWhitelist;
        while (*list) {
2316
            free(*list);
2317 2318
            list++;
        }
J
Jim Meyering 已提交
2319
        free(server->saslUsernameWhitelist);
2320
    }
2321
#endif
2322

2323 2324 2325 2326 2327
#if HAVE_POLKIT0
        if (server->sysbus)
            dbus_connection_unref(server->sysbus);
#endif

2328
    virStateCleanup();
2329

2330 2331 2332 2333 2334 2335
    if (virCondDestroy(&server->job) < 0) {
        ;
    }
    virMutexDestroy(&server->lock);

    VIR_FREE(server);
D
Daniel P. Berrange 已提交
2336 2337
}

2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353
/* 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:
2354
        if (VIR_ALLOC_N(list, 2) < 0) {
2355
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2356 2357 2358 2359 2360
            return -1;
        }
        list[0] = strdup (p->str);
        list[1] = NULL;
        if (list[0] == NULL) {
2361
            VIR_ERROR(_("failed to allocate memory for %s config list value"),
2362
                      key);
2363
            VIR_FREE(list);
2364 2365 2366 2367 2368 2369 2370 2371 2372
            return -1;
        }
        break;

    case VIR_CONF_LIST: {
        int i, len = 0;
        virConfValuePtr pp;
        for (pp = p->list; pp; pp = pp->next)
            len++;
2373
        if (VIR_ALLOC_N(list, 1+len) < 0) {
2374
            VIR_ERROR(_("failed to allocate memory for %s config list"), key);
2375 2376 2377 2378
            return -1;
        }
        for (i = 0, pp = p->list; pp; ++i, pp = pp->next) {
            if (pp->type != VIR_CONF_STRING) {
2379 2380
                VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
                            " must be a string or list of strings\n"),
2381
                          filename, key);
2382
                VIR_FREE(list);
2383 2384 2385 2386 2387 2388
                return -1;
            }
            list[i] = strdup (pp->str);
            if (list[i] == NULL) {
                int j;
                for (j = 0 ; j < i ; j++)
2389 2390
                    VIR_FREE(list[j]);
                VIR_FREE(list);
2391 2392
                VIR_ERROR(_("failed to allocate memory for %s config list value"),
                          key);
2393 2394 2395 2396 2397 2398 2399 2400 2401
                return -1;
            }

        }
        list[i] = NULL;
        break;
    }

    default:
2402 2403
        VIR_ERROR(_("remoteReadConfigFile: %s: %s:"
                    " must be a string or list of strings\n"),
2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417
                  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) {
2418
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: invalid type:"
2419
                    " got %s; expected %s\n"), filename, key,
2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
                  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) {                                   \
2439 2440 2441
                char ebuf[1024];                                        \
                VIR_ERROR(_("remoteReadConfigFile: %s\n"),		\
                          virStrerror(errno, ebuf, sizeof ebuf));       \
2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457
                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)

2458 2459 2460 2461 2462 2463 2464 2465

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

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

2466
    if (checkType (p, filename, key, VIR_CONF_STRING) < 0)
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476
        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;
2477 2478 2479 2480
#endif
#if HAVE_POLKIT
    } else if (STREQ(p->str, "polkit")) {
        *auth = REMOTE_AUTH_POLKIT;
2481 2482
#endif
    } else {
2483
        VIR_ERROR(_("remoteReadConfigFile: %s: %s: unsupported auth %s\n"),
2484
                  filename, key, p->str);
2485 2486 2487 2488 2489 2490
        return -1;
    }

    return 0;
}

2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510
#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

2511 2512 2513 2514 2515 2516
/*
 * 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.
 */
2517
static int
2518 2519 2520
qemudSetLogging(virConfPtr conf, const char *filename)
{
    int log_level = 0;
2521 2522
    char *log_filters = NULL;
    char *log_outputs = NULL;
2523 2524
    int ret = -1;

2525 2526
    virLogReset();

2527
    /*
2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
     * 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.
2540
     */
2541
    GET_CONF_INT (conf, filename, log_level);
2542 2543 2544
    if (log_level != 0)
        virLogSetDefaultPriority(log_level);

2545
    virLogSetFromEnv();
2546 2547 2548 2549

    if (virLogGetNbFilters() == 0) {
        GET_CONF_STR (conf, filename, log_filters);
        virLogParseFilters(log_filters);
2550
    }
2551

2552 2553 2554
    if (virLogGetNbOutputs() == 0) {
        GET_CONF_STR (conf, filename, log_outputs);
        virLogParseOutputs(log_outputs);
2555
    }
2556 2557

    /*
2558 2559
     * If no defined outputs, then direct to syslog when running
     * as daemon. Otherwise the default output is stderr.
2560
     */
2561
    if (virLogGetNbOutputs() == 0) {
2562
        char *tmp = NULL;
2563
        if (godaemon) {
2564 2565
            if (virAsprintf (&tmp, "%d:syslog:libvirtd",
                             virLogGetDefaultPriority()) < 0)
2566 2567
                goto free_and_fail;
        } else {
2568 2569
            if (virAsprintf (&tmp, "%d:stderr",
                             virLogGetDefaultPriority()) < 0)
2570
                goto free_and_fail;
2571
        }
2572 2573
        virLogParseOutputs(tmp);
        VIR_FREE(tmp);
2574
    }
2575 2576 2577 2578 2579 2580 2581

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

2582 2583
    ret = 0;

2584 2585 2586
free_and_fail:
    VIR_FREE(log_filters);
    VIR_FREE(log_outputs);
2587
    return(ret);
2588
}
2589

2590 2591 2592 2593
/* Read the config file if it exists.
 * Only used in the remote case, hence the name.
 */
static int
2594
remoteReadConfigFile (struct qemud_server *server, const char *filename)
2595 2596 2597
{
    virConfPtr conf;

2598 2599 2600 2601 2602
    /* 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;
2603
    char *buf = NULL;
2604

2605 2606
#if HAVE_POLKIT
    /* Change the default back to no auth for non-root */
2607
    if (!server->privileged && auth_unix_rw == REMOTE_AUTH_POLKIT)
2608
        auth_unix_rw = REMOTE_AUTH_NONE;
2609
    if (!server->privileged && auth_unix_ro == REMOTE_AUTH_POLKIT)
2610 2611 2612
        auth_unix_ro = REMOTE_AUTH_NONE;
#endif

2613
    conf = virConfReadFile (filename, 0);
2614
    if (!conf) return -1;
2615

2616 2617 2618
    /*
     * First get all the logging settings and activate them
     */
2619 2620
    if (qemudSetLogging(conf, filename) < 0)
        goto free_and_fail;
2621

2622 2623
    GET_CONF_INT (conf, filename, listen_tcp);
    GET_CONF_INT (conf, filename, listen_tls);
2624 2625
    GET_CONF_STR (conf, filename, tls_port);
    GET_CONF_STR (conf, filename, tcp_port);
2626
    GET_CONF_STR (conf, filename, listen_addr);
J
Jim Meyering 已提交
2627

2628
    if (remoteConfigGetAuth(conf, "auth_unix_rw", &auth_unix_rw, filename) < 0)
J
Jim Meyering 已提交
2629
        goto free_and_fail;
2630 2631 2632 2633 2634 2635 2636
#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
2637
    if (remoteConfigGetAuth(conf, "auth_unix_ro", &auth_unix_ro, filename) < 0)
J
Jim Meyering 已提交
2638
        goto free_and_fail;
2639
    if (remoteConfigGetAuth(conf, "auth_tcp", &auth_tcp, filename) < 0)
J
Jim Meyering 已提交
2640
        goto free_and_fail;
2641
    if (remoteConfigGetAuth(conf, "auth_tls", &auth_tls, filename) < 0)
J
Jim Meyering 已提交
2642
        goto free_and_fail;
2643

2644 2645
    GET_CONF_STR (conf, filename, unix_sock_group);
    if (unix_sock_group) {
2646
        if (!server->privileged) {
2647
            VIR_WARN0(_("Cannot set group when not running as root"));
2648
        } else {
2649
            int ret;
2650
            struct group grpdata, *grp;
2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671
            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) {
2672
                VIR_ERROR(_("Failed to lookup group '%s'"), unix_sock_group);
2673
                goto free_and_fail;
2674 2675
            }
            unix_sock_gid = grp->gr_gid;
2676
            VIR_FREE (buf);
2677
        }
2678 2679
        free (unix_sock_group);
        unix_sock_group = NULL;
2680 2681
    }

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

2692 2693
    GET_CONF_STR (conf, filename, unix_sock_rw_perms);
    if (unix_sock_rw_perms) {
2694
        if (virStrToLong_i (unix_sock_rw_perms, NULL, 8, &unix_sock_rw_mask) != 0) {
2695
            VIR_ERROR(_("Failed to parse mode '%s'"), unix_sock_rw_perms);
2696
            goto free_and_fail;
2697
        }
2698 2699
        free (unix_sock_rw_perms);
        unix_sock_rw_perms = NULL;
2700 2701
    }

2702 2703
    GET_CONF_STR (conf, filename, unix_sock_dir);

2704 2705
    GET_CONF_INT (conf, filename, mdns_adv);
    GET_CONF_STR (conf, filename, mdns_name);
2706

2707
    GET_CONF_INT (conf, filename, tls_no_verify_certificate);
2708

2709 2710 2711 2712
    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);
2713

2714 2715 2716
    if (remoteConfigGetStringList (conf, "tls_allowed_dn_list",
                                   &tls_allowed_dn_list, filename) < 0)
        goto free_and_fail;
2717

2718
    if (remoteReadSaslAllowedUsernameList (conf, server, filename) < 0)
2719
        goto free_and_fail;
2720

2721 2722 2723 2724 2725

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

2726 2727 2728
    GET_CONF_INT (conf, filename, max_requests);
    GET_CONF_INT (conf, filename, max_client_requests);

2729 2730
    virConfFree (conf);
    return 0;
2731

2732 2733 2734 2735 2736 2737 2738
 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);
2739
    VIR_FREE (buf);
2740

2741 2742 2743 2744
    /* 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
2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755
       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;
2756 2757
}

2758 2759 2760 2761 2762 2763 2764
/* 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
2765 2766 2767 2768 2769 2770 2771 2772
#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)) {
2773
        VIR_ERROR0(_("additional privileges are required\n"));
J
John Levon 已提交
2774 2775 2776 2777 2778
        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)) {
2779
        VIR_ERROR0(_("failed to set reduced privileges\n"));
J
John Levon 已提交
2780 2781 2782 2783 2784 2785 2786 2787 2788
        return -1;
    }

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

2789 2790 2791 2792 2793
/* Print command-line usage. */
static void
usage (const char *argv0)
{
    fprintf (stderr,
2794 2795 2796 2797 2798 2799 2800
             "\n\
Usage:\n\
  %s [options]\n\
\n\
Options:\n\
  -v | --verbose         Verbose messages.\n\
  -d | --daemon          Run as a daemon & write PID file.\n\
2801 2802
  -l | --listen          Listen for TCP/IP connections.\n\
  -t | --timeout <secs>  Exit after timeout period.\n\
2803
  -f | --config <file>   Configuration file.\n\
2804
     | --version         Display version information.\n\
2805 2806
  -p | --pid-file <file> Change name of PID file.\n\
\n\
2807
libvirt management daemon:\n\
2808 2809 2810
\n\
  Default paths:\n\
\n\
2811
    Configuration file (unless overridden by -c):\n\
2812 2813
      " SYSCONF_DIR "/libvirt/libvirtd.conf\n\
\n\
2814
    Sockets (as root):\n\
2815 2816
      " LOCAL_STATE_DIR "/run/libvirt/libvirt-sock\n\
      " LOCAL_STATE_DIR "/run/libvirt/libvirt-sock-ro\n\
2817 2818 2819
\n\
    Sockets (as non-root):\n\
      $HOME/.libvirt/libvirt-sock (in UNIX abstract namespace)\n\
2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832
\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)");
2833 2834
}

2835 2836 2837 2838
enum {
    OPT_VERSION = 129
};

D
Daniel P. Berrange 已提交
2839 2840
#define MAX_LISTEN 5
int main(int argc, char **argv) {
2841
    struct qemud_server *server = NULL;
2842 2843
    struct sigaction sig_action;
    int sigpipe[2];
2844
    const char *pid_file = NULL;
2845
    const char *remote_config_file = NULL;
2846
    int ret = 1;
D
Daniel P. Berrange 已提交
2847 2848 2849 2850

    struct option opts[] = {
        { "verbose", no_argument, &verbose, 1},
        { "daemon", no_argument, &godaemon, 1},
2851
        { "listen", no_argument, &ipsock, 1},
2852
        { "config", required_argument, NULL, 'f'},
2853 2854
        { "timeout", required_argument, NULL, 't'},
        { "pid-file", required_argument, NULL, 'p'},
2855
        { "version", no_argument, NULL, OPT_VERSION },
2856
        { "help", no_argument, NULL, '?' },
D
Daniel P. Berrange 已提交
2857 2858 2859
        {0, 0, 0, 0}
    };

D
Daniel P. Berrange 已提交
2860 2861
    virInitialize();

D
Daniel P. Berrange 已提交
2862 2863 2864 2865 2866
    while (1) {
        int optidx = 0;
        int c;
        char *tmp;

2867
        c = getopt_long(argc, argv, "ldf:p:t:v", opts, &optidx);
D
Daniel P. Berrange 已提交
2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882

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

        switch (c) {
        case 0:
            /* Got one of the flags */
            break;
        case 'v':
            verbose = 1;
            break;
        case 'd':
            godaemon = 1;
            break;
2883 2884
        case 'l':
            ipsock = 1;
D
Daniel P. Berrange 已提交
2885 2886 2887
            break;

        case 't':
2888
            if (virStrToLong_i(optarg, &tmp, 10, &timeout) != 0
2889 2890 2891
                || timeout <= 0
                /* Ensure that we can multiply by 1000 without overflowing.  */
                || timeout > INT_MAX / 1000)
D
Daniel P. Berrange 已提交
2892 2893
                timeout = -1;
            break;
2894 2895

        case 'p':
2896 2897 2898 2899 2900
            pid_file = optarg;
            break;

        case 'f':
            remote_config_file = optarg;
2901 2902
            break;

2903 2904 2905 2906
        case OPT_VERSION:
            version (argv[0]);
            return 0;

D
Daniel P. Berrange 已提交
2907
        case '?':
2908
            usage (argv[0]);
D
Daniel P. Berrange 已提交
2909 2910 2911
            return 2;

        default:
2912 2913 2914
            fprintf (stderr, "libvirtd: internal error: unknown flag: %c\n",
                     c);
            exit (1);
D
Daniel P. Berrange 已提交
2915 2916 2917
        }
    }

2918 2919 2920 2921
    if (remote_config_file == NULL) {
        static const char *default_config_file
            = SYSCONF_DIR "/libvirt/libvirtd.conf";
        remote_config_file =
2922
            (access(default_config_file, R_OK) == 0
2923 2924 2925 2926
             ? default_config_file
             : "/dev/null");
    }

2927
    if (godaemon) {
2928
        char ebuf[1024];
2929
        if (qemudGoDaemon() < 0) {
2930 2931
            VIR_ERROR(_("Failed to fork as daemon: %s"),
                      virStrerror(errno, ebuf, sizeof ebuf));
2932 2933 2934 2935 2936 2937
            goto error1;
        }
    }

    /* If running as root and no PID file is set, use the default */
    if (pid_file == NULL &&
2938
        geteuid() == 0 &&
2939 2940 2941 2942 2943 2944 2945 2946
        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;

2947
    if (pipe(sigpipe) < 0 ||
2948 2949 2950 2951
        virSetNonBlock(sigpipe[0]) < 0 ||
        virSetNonBlock(sigpipe[1]) < 0 ||
        virSetCloseExec(sigpipe[0]) < 0 ||
        virSetCloseExec(sigpipe[1]) < 0) {
2952 2953 2954
        char ebuf[1024];
        VIR_ERROR(_("Failed to create pipe: %s"),
                  virStrerror(errno, ebuf, sizeof ebuf));
2955
        goto error2;
2956
    }
2957 2958
    sigwrite = sigpipe[1];

2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971
    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
2972
    /* Ensure the rundir exists (on tmpfs on some systems) */
2973
    if (geteuid() == 0) {
J
John Levon 已提交
2974 2975 2976 2977
        const char *rundir = LOCAL_STATE_DIR "/run/libvirt";

        if (mkdir (rundir, 0755)) {
            if (errno != EEXIST) {
2978 2979 2980
                char ebuf[1024];
                VIR_ERROR(_("unable to create rundir %s: %s"), rundir,
                          virStrerror(errno, ebuf, sizeof(ebuf)));
J
John Levon 已提交
2981 2982 2983 2984 2985
                return -1;
            }
        }
    }

2986 2987 2988 2989 2990 2991
    /* 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
2992 2993 2994
    if (qemudSetupPrivs() < 0)
        goto error2;

2995 2996
    if (!(server = qemudInitialize(sigpipe[0]))) {
        ret = 2;
2997
        goto error2;
2998
    }
2999

3000 3001
    /* Read the config file (if it exists). */
    if (remoteReadConfigFile (server, remote_config_file) < 0)
3002
        goto error2;
D
Daniel P. Berrange 已提交
3003

3004 3005 3006
    /* Disable error func, now logging is setup */
    virSetErrorFunc(NULL, virshErrorHandler);

3007
    if (virEventAddHandleImpl(sigpipe[0],
3008
                              VIR_EVENT_HANDLE_READABLE,
3009
                              qemudDispatchSignalEvent,
3010
                              server, NULL) < 0) {
3011
        VIR_ERROR0(_("Failed to register callback for signal pipe"));
D
Daniel P. Berrange 已提交
3012 3013 3014 3015
        ret = 3;
        goto error2;
    }

3016
    if (qemudNetworkInit(server) < 0) {
3017 3018 3019 3020
        ret = 2;
        goto error2;
    }

3021
    qemudRunLoop(server);
D
Daniel P. Berrange 已提交
3022

3023 3024
    ret = 0;

3025
error2:
3026 3027
    if (server)
        qemudCleanup(server);
3028 3029 3030 3031 3032
    if (pid_file)
        unlink (pid_file);
    close(sigwrite);

error1:
3033
    virLogShutdown();
3034
    return ret;
D
Daniel P. Berrange 已提交
3035
}