qga/commands-posix.c

   1 /*
   2  * QEMU Guest Agent POSIX-specific command implementations
   3  *
   4  * Copyright IBM Corp. 2011
   5  *
   6  * Authors:
   7  *  Michael Roth      <mdroth@linux.vnet.ibm.com>
   8  *  Michal Privoznik  <mprivozn@redhat.com>
   9  *
  10  * This work is licensed under the terms of the GNU GPL, version 2 or later.
  11  * See the COPYING file in the top-level directory.
  12  */
  13
  14 #include "qemu/osdep.h"
  15 #include <sys/ioctl.h>
  16 #include <sys/utsname.h>
  17 #include <sys/wait.h>
  18 #include <dirent.h>
  19 #include "qemu-common.h"
  20 #include "guest-agent-core.h"
  21 #include "qga-qapi-commands.h"
  22 #include "qapi/error.h"
  23 #include "qapi/qmp/qerror.h"
  24 #include "qemu/queue.h"
  25 #include "qemu/host-utils.h"
  26 #include "qemu/sockets.h"
  27 #include "qemu/base64.h"
  28 #include "qemu/cutils.h"
  29
  30 #ifdef HAVE_UTMPX
  31 #include <utmpx.h>
  32 #endif
  33
  34 #ifndef CONFIG_HAS_ENVIRON
  35 #ifdef __APPLE__
  36 #include <crt_externs.h>
  37 #define environ (*_NSGetEnviron())
  38 #else
  39 extern char **environ;
  40 #endif
  41 #endif
  42
  43 #if defined(__linux__)
  44 #include <mntent.h>
  45 #include <linux/fs.h>
  46 #include <ifaddrs.h>
  47 #include <arpa/inet.h>
  48 #include <sys/socket.h>
  49 #include <net/if.h>
  50 #include <sys/statvfs.h>
  51
  52 #ifdef CONFIG_LIBUDEV
  53 #include <libudev.h>
  54 #endif
  55
  56 #ifdef FIFREEZE
  57 #define CONFIG_FSFREEZE
  58 #endif
  59 #ifdef FITRIM
  60 #define CONFIG_FSTRIM
  61 #endif
  62 #endif
  63
  64 static void ga_wait_child(pid_t pid, int *status, Error **errp)
  65 {
  66     pid_t rpid;
  67
  68     *status = 0;
  69
  70     do {
  71         rpid = waitpid(pid, status, 0);
  72     } while (rpid == -1 && errno == EINTR);
  73
  74     if (rpid == -1) {
  75         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
  76                          pid);
  77         return;
  78     }
  79
  80     g_assert(rpid == pid);
  81 }
  82
  83 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
  84 {
  85     const char *shutdown_flag;
  86     Error *local_err = NULL;
  87     pid_t pid;
  88     int status;
  89
  90     slog("guest-shutdown called, mode: %s", mode);
  91     if (!has_mode || strcmp(mode, "powerdown") == 0) {
  92         shutdown_flag = "-P";
  93     } else if (strcmp(mode, "halt") == 0) {
  94         shutdown_flag = "-H";
  95     } else if (strcmp(mode, "reboot") == 0) {
  96         shutdown_flag = "-r";
  97     } else {
  98         error_setg(errp,
  99                    "mode is invalid (valid values are: halt|powerdown|reboot");
 100         return;
 101     }
 102
 103     pid = fork();
 104     if (pid == 0) {
 105         /* child, start the shutdown */
 106         setsid();
 107         reopen_fd_to_null(0);
 108         reopen_fd_to_null(1);
 109         reopen_fd_to_null(2);
 110
 111         execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
 112                "hypervisor initiated shutdown", (char*)NULL, environ);
 113         _exit(EXIT_FAILURE);
 114     } else if (pid < 0) {
 115         error_setg_errno(errp, errno, "failed to create child process");
 116         return;
 117     }
 118
 119     ga_wait_child(pid, &status, &local_err);
 120     if (local_err) {
 121         error_propagate(errp, local_err);
 122         return;
 123     }
 124
 125     if (!WIFEXITED(status)) {
 126         error_setg(errp, "child process has terminated abnormally");
 127         return;
 128     }
 129
 130     if (WEXITSTATUS(status)) {
 131         error_setg(errp, "child process has failed to shutdown");
 132         return;
 133     }
 134
 135     /* succeeded */
 136 }
 137
 138 int64_t qmp_guest_get_time(Error **errp)
 139 {
 140    int ret;
 141    qemu_timeval tq;
 142
 143    ret = qemu_gettimeofday(&tq);
 144    if (ret < 0) {
 145        error_setg_errno(errp, errno, "Failed to get time");
 146        return -1;
 147    }
 148
 149    return tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
 150 }
 151
 152 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
 153 {
 154     int ret;
 155     int status;
 156     pid_t pid;
 157     Error *local_err = NULL;
 158     struct timeval tv;
 159     static const char hwclock_path[] = "/sbin/hwclock";
 160     static int hwclock_available = -1;
 161
 162     if (hwclock_available < 0) {
 163         hwclock_available = (access(hwclock_path, X_OK) == 0);
 164     }
 165
 166     if (!hwclock_available) {
 167         error_setg(errp, QERR_UNSUPPORTED);
 168         return;
 169     }
 170
 171     /* If user has passed a time, validate and set it. */
 172     if (has_time) {
 173         GDate date = { 0, };
 174
 175         /* year-2038 will overflow in case time_t is 32bit */
 176         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
 177             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
 178             return;
 179         }
 180
 181         tv.tv_sec = time_ns / 1000000000;
 182         tv.tv_usec = (time_ns % 1000000000) / 1000;
 183         g_date_set_time_t(&date, tv.tv_sec);
 184         if (date.year < 1970 || date.year >= 2070) {
 185             error_setg_errno(errp, errno, "Invalid time");
 186             return;
 187         }
 188
 189         ret = settimeofday(&tv, NULL);
 190         if (ret < 0) {
 191             error_setg_errno(errp, errno, "Failed to set time to guest");
 192             return;
 193         }
 194     }
 195
 196     /* Now, if user has passed a time to set and the system time is set, we
 197      * just need to synchronize the hardware clock. However, if no time was
 198      * passed, user is requesting the opposite: set the system time from the
 199      * hardware clock (RTC). */
 200     pid = fork();
 201     if (pid == 0) {
 202         setsid();
 203         reopen_fd_to_null(0);
 204         reopen_fd_to_null(1);
 205         reopen_fd_to_null(2);
 206
 207         /* Use '/sbin/hwclock -w' to set RTC from the system time,
 208          * or '/sbin/hwclock -s' to set the system time from RTC. */
 209         execle(hwclock_path, "hwclock", has_time ? "-w" : "-s",
 210                NULL, environ);
 211         _exit(EXIT_FAILURE);
 212     } else if (pid < 0) {
 213         error_setg_errno(errp, errno, "failed to create child process");
 214         return;
 215     }
 216
 217     ga_wait_child(pid, &status, &local_err);
 218     if (local_err) {
 219         error_propagate(errp, local_err);
 220         return;
 221     }
 222
 223     if (!WIFEXITED(status)) {
 224         error_setg(errp, "child process has terminated abnormally");
 225         return;
 226     }
 227
 228     if (WEXITSTATUS(status)) {
 229         error_setg(errp, "hwclock failed to set hardware clock to system time");
 230         return;
 231     }
 232 }
 233
 234 typedef enum {
 235     RW_STATE_NEW,
 236     RW_STATE_READING,
 237     RW_STATE_WRITING,
 238 } RwState;
 239
 240 typedef struct GuestFileHandle {
 241     uint64_t id;
 242     FILE *fh;
 243     RwState state;
 244     QTAILQ_ENTRY(GuestFileHandle) next;
 245 } GuestFileHandle;
 246
 247 static struct {
 248     QTAILQ_HEAD(, GuestFileHandle) filehandles;
 249 } guest_file_state = {
 250     .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
 251 };
 252
 253 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
 254 {
 255     GuestFileHandle *gfh;
 256     int64_t handle;
 257
 258     handle = ga_get_fd_handle(ga_state, errp);
 259     if (handle < 0) {
 260         return -1;
 261     }
 262
 263     gfh = g_new0(GuestFileHandle, 1);
 264     gfh->id = handle;
 265     gfh->fh = fh;
 266     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
 267
 268     return handle;
 269 }
 270
 271 static GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
 272 {
 273     GuestFileHandle *gfh;
 274
 275     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
 276     {
 277         if (gfh->id == id) {
 278             return gfh;
 279         }
 280     }
 281
 282     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
 283     return NULL;
 284 }
 285
 286 typedef const char * const ccpc;
 287
 288 #ifndef O_BINARY
 289 #define O_BINARY 0
 290 #endif
 291
 292 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
 293 static const struct {
 294     ccpc *forms;
 295     int oflag_base;
 296 } guest_file_open_modes[] = {
 297     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
 298     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
 299     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
 300     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
 301     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
 302     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
 303     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
 304     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
 305     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
 306     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
 307     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
 308     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
 309 };
 310
 311 static int
 312 find_open_flag(const char *mode_str, Error **errp)
 313 {
 314     unsigned mode;
 315
 316     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
 317         ccpc *form;
 318
 319         form = guest_file_open_modes[mode].forms;
 320         while (*form != NULL && strcmp(*form, mode_str) != 0) {
 321             ++form;
 322         }
 323         if (*form != NULL) {
 324             break;
 325         }
 326     }
 327
 328     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
 329         error_setg(errp, "invalid file open mode '%s'", mode_str);
 330         return -1;
 331     }
 332     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
 333 }
 334
 335 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
 336                                S_IRGRP | S_IWGRP | \
 337                                S_IROTH | S_IWOTH)
 338
 339 static FILE *
 340 safe_open_or_create(const char *path, const char *mode, Error **errp)
 341 {
 342     Error *local_err = NULL;
 343     int oflag;
 344
 345     oflag = find_open_flag(mode, &local_err);
 346     if (local_err == NULL) {
 347         int fd;
 348
 349         /* If the caller wants / allows creation of a new file, we implement it
 350          * with a two step process: open() + (open() / fchmod()).
 351          *
 352          * First we insist on creating the file exclusively as a new file. If
 353          * that succeeds, we're free to set any file-mode bits on it. (The
 354          * motivation is that we want to set those file-mode bits independently
 355          * of the current umask.)
 356          *
 357          * If the exclusive creation fails because the file already exists
 358          * (EEXIST is not possible for any other reason), we just attempt to
 359          * open the file, but in this case we won't be allowed to change the
 360          * file-mode bits on the preexistent file.
 361          *
 362          * The pathname should never disappear between the two open()s in
 363          * practice. If it happens, then someone very likely tried to race us.
 364          * In this case just go ahead and report the ENOENT from the second
 365          * open() to the caller.
 366          *
 367          * If the caller wants to open a preexistent file, then the first
 368          * open() is decisive and its third argument is ignored, and the second
 369          * open() and the fchmod() are never called.
 370          */
 371         fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
 372         if (fd == -1 && errno == EEXIST) {
 373             oflag &= ~(unsigned)O_CREAT;
 374             fd = open(path, oflag);
 375         }
 376
 377         if (fd == -1) {
 378             error_setg_errno(&local_err, errno, "failed to open file '%s' "
 379                              "(mode: '%s')", path, mode);
 380         } else {
 381             qemu_set_cloexec(fd);
 382
 383             if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
 384                 error_setg_errno(&local_err, errno, "failed to set permission "
 385                                  "0%03o on new file '%s' (mode: '%s')",
 386                                  (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
 387             } else {
 388                 FILE *f;
 389
 390                 f = fdopen(fd, mode);
 391                 if (f == NULL) {
 392                     error_setg_errno(&local_err, errno, "failed to associate "
 393                                      "stdio stream with file descriptor %d, "
 394                                      "file '%s' (mode: '%s')", fd, path, mode);
 395                 } else {
 396                     return f;
 397                 }
 398             }
 399
 400             close(fd);
 401             if (oflag & O_CREAT) {
 402                 unlink(path);
 403             }
 404         }
 405     }
 406
 407     error_propagate(errp, local_err);
 408     return NULL;
 409 }
 410
 411 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
 412                             Error **errp)
 413 {
 414     FILE *fh;
 415     Error *local_err = NULL;
 416     int64_t handle;
 417
 418     if (!has_mode) {
 419         mode = "r";
 420     }
 421     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
 422     fh = safe_open_or_create(path, mode, &local_err);
 423     if (local_err != NULL) {
 424         error_propagate(errp, local_err);
 425         return -1;
 426     }
 427
 428     /* set fd non-blocking to avoid common use cases (like reading from a
 429      * named pipe) from hanging the agent
 430      */
 431     qemu_set_nonblock(fileno(fh));
 432
 433     handle = guest_file_handle_add(fh, errp);
 434     if (handle < 0) {
 435         fclose(fh);
 436         return -1;
 437     }
 438
 439     slog("guest-file-open, handle: %" PRId64, handle);
 440     return handle;
 441 }
 442
 443 void qmp_guest_file_close(int64_t handle, Error **errp)
 444 {
 445     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 446     int ret;
 447
 448     slog("guest-file-close called, handle: %" PRId64, handle);
 449     if (!gfh) {
 450         return;
 451     }
 452
 453     ret = fclose(gfh->fh);
 454     if (ret == EOF) {
 455         error_setg_errno(errp, errno, "failed to close handle");
 456         return;
 457     }
 458
 459     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
 460     g_free(gfh);
 461 }
 462
 463 struct GuestFileRead *qmp_guest_file_read(int64_t handle, bool has_count,
 464                                           int64_t count, Error **errp)
 465 {
 466     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 467     GuestFileRead *read_data = NULL;
 468     guchar *buf;
 469     FILE *fh;
 470     size_t read_count;
 471
 472     if (!gfh) {
 473         return NULL;
 474     }
 475
 476     if (!has_count) {
 477         count = QGA_READ_COUNT_DEFAULT;
 478     } else if (count < 0 || count >= UINT32_MAX) {
 479         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 480                    count);
 481         return NULL;
 482     }
 483
 484     fh = gfh->fh;
 485
 486     /* explicitly flush when switching from writing to reading */
 487     if (gfh->state == RW_STATE_WRITING) {
 488         int ret = fflush(fh);
 489         if (ret == EOF) {
 490             error_setg_errno(errp, errno, "failed to flush file");
 491             return NULL;
 492         }
 493         gfh->state = RW_STATE_NEW;
 494     }
 495
 496     buf = g_malloc0(count+1);
 497     read_count = fread(buf, 1, count, fh);
 498     if (ferror(fh)) {
 499         error_setg_errno(errp, errno, "failed to read file");
 500         slog("guest-file-read failed, handle: %" PRId64, handle);
 501     } else {
 502         buf[read_count] = 0;
 503         read_data = g_new0(GuestFileRead, 1);
 504         read_data->count = read_count;
 505         read_data->eof = feof(fh);
 506         if (read_count) {
 507             read_data->buf_b64 = g_base64_encode(buf, read_count);
 508         }
 509         gfh->state = RW_STATE_READING;
 510     }
 511     g_free(buf);
 512     clearerr(fh);
 513
 514     return read_data;
 515 }
 516
 517 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
 518                                      bool has_count, int64_t count,
 519                                      Error **errp)
 520 {
 521     GuestFileWrite *write_data = NULL;
 522     guchar *buf;
 523     gsize buf_len;
 524     int write_count;
 525     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 526     FILE *fh;
 527
 528     if (!gfh) {
 529         return NULL;
 530     }
 531
 532     fh = gfh->fh;
 533
 534     if (gfh->state == RW_STATE_READING) {
 535         int ret = fseek(fh, 0, SEEK_CUR);
 536         if (ret == -1) {
 537             error_setg_errno(errp, errno, "failed to seek file");
 538             return NULL;
 539         }
 540         gfh->state = RW_STATE_NEW;
 541     }
 542
 543     buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
 544     if (!buf) {
 545         return NULL;
 546     }
 547
 548     if (!has_count) {
 549         count = buf_len;
 550     } else if (count < 0 || count > buf_len) {
 551         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 552                    count);
 553         g_free(buf);
 554         return NULL;
 555     }
 556
 557     write_count = fwrite(buf, 1, count, fh);
 558     if (ferror(fh)) {
 559         error_setg_errno(errp, errno, "failed to write to file");
 560         slog("guest-file-write failed, handle: %" PRId64, handle);
 561     } else {
 562         write_data = g_new0(GuestFileWrite, 1);
 563         write_data->count = write_count;
 564         write_data->eof = feof(fh);
 565         gfh->state = RW_STATE_WRITING;
 566     }
 567     g_free(buf);
 568     clearerr(fh);
 569
 570     return write_data;
 571 }
 572
 573 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
 574                                           GuestFileWhence *whence_code,
 575                                           Error **errp)
 576 {
 577     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 578     GuestFileSeek *seek_data = NULL;
 579     FILE *fh;
 580     int ret;
 581     int whence;
 582     Error *err = NULL;
 583
 584     if (!gfh) {
 585         return NULL;
 586     }
 587
 588     /* We stupidly exposed 'whence':'int' in our qapi */
 589     whence = ga_parse_whence(whence_code, &err);
 590     if (err) {
 591         error_propagate(errp, err);
 592         return NULL;
 593     }
 594
 595     fh = gfh->fh;
 596     ret = fseek(fh, offset, whence);
 597     if (ret == -1) {
 598         error_setg_errno(errp, errno, "failed to seek file");
 599         if (errno == ESPIPE) {
 600             /* file is non-seekable, stdio shouldn't be buffering anyways */
 601             gfh->state = RW_STATE_NEW;
 602         }
 603     } else {
 604         seek_data = g_new0(GuestFileSeek, 1);
 605         seek_data->position = ftell(fh);
 606         seek_data->eof = feof(fh);
 607         gfh->state = RW_STATE_NEW;
 608     }
 609     clearerr(fh);
 610
 611     return seek_data;
 612 }
 613
 614 void qmp_guest_file_flush(int64_t handle, Error **errp)
 615 {
 616     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 617     FILE *fh;
 618     int ret;
 619
 620     if (!gfh) {
 621         return;
 622     }
 623
 624     fh = gfh->fh;
 625     ret = fflush(fh);
 626     if (ret == EOF) {
 627         error_setg_errno(errp, errno, "failed to flush file");
 628     } else {
 629         gfh->state = RW_STATE_NEW;
 630     }
 631 }
 632
 633 /* linux-specific implementations. avoid this if at all possible. */
 634 #if defined(__linux__)
 635
 636 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
 637 typedef struct FsMount {
 638     char *dirname;
 639     char *devtype;
 640     unsigned int devmajor, devminor;
 641     QTAILQ_ENTRY(FsMount) next;
 642 } FsMount;
 643
 644 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
 645
 646 static void free_fs_mount_list(FsMountList *mounts)
 647 {
 648      FsMount *mount, *temp;
 649
 650      if (!mounts) {
 651          return;
 652      }
 653
 654      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
 655          QTAILQ_REMOVE(mounts, mount, next);
 656          g_free(mount->dirname);
 657          g_free(mount->devtype);
 658          g_free(mount);
 659      }
 660 }
 661
 662 static int dev_major_minor(const char *devpath,
 663                            unsigned int *devmajor, unsigned int *devminor)
 664 {
 665     struct stat st;
 666
 667     *devmajor = 0;
 668     *devminor = 0;
 669
 670     if (stat(devpath, &st) < 0) {
 671         slog("failed to stat device file '%s': %s", devpath, strerror(errno));
 672         return -1;
 673     }
 674     if (S_ISDIR(st.st_mode)) {
 675         /* It is bind mount */
 676         return -2;
 677     }
 678     if (S_ISBLK(st.st_mode)) {
 679         *devmajor = major(st.st_rdev);
 680         *devminor = minor(st.st_rdev);
 681         return 0;
 682     }
 683     return -1;
 684 }
 685
 686 /*
 687  * Walk the mount table and build a list of local file systems
 688  */
 689 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
 690 {
 691     struct mntent *ment;
 692     FsMount *mount;
 693     char const *mtab = "/proc/self/mounts";
 694     FILE *fp;
 695     unsigned int devmajor, devminor;
 696
 697     fp = setmntent(mtab, "r");
 698     if (!fp) {
 699         error_setg(errp, "failed to open mtab file: '%s'", mtab);
 700         return;
 701     }
 702
 703     while ((ment = getmntent(fp))) {
 704         /*
 705          * An entry which device name doesn't start with a '/' is
 706          * either a dummy file system or a network file system.
 707          * Add special handling for smbfs and cifs as is done by
 708          * coreutils as well.
 709          */
 710         if ((ment->mnt_fsname[0] != '/') ||
 711             (strcmp(ment->mnt_type, "smbfs") == 0) ||
 712             (strcmp(ment->mnt_type, "cifs") == 0)) {
 713             continue;
 714         }
 715         if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
 716             /* Skip bind mounts */
 717             continue;
 718         }
 719
 720         mount = g_new0(FsMount, 1);
 721         mount->dirname = g_strdup(ment->mnt_dir);
 722         mount->devtype = g_strdup(ment->mnt_type);
 723         mount->devmajor = devmajor;
 724         mount->devminor = devminor;
 725
 726         QTAILQ_INSERT_TAIL(mounts, mount, next);
 727     }
 728
 729     endmntent(fp);
 730 }
 731
 732 static void decode_mntname(char *name, int len)
 733 {
 734     int i, j = 0;
 735     for (i = 0; i <= len; i++) {
 736         if (name[i] != '\\') {
 737             name[j++] = name[i];
 738         } else if (name[i + 1] == '\\') {
 739             name[j++] = '\\';
 740             i++;
 741         } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
 742                    name[i + 2] >= '0' && name[i + 2] <= '7' &&
 743                    name[i + 3] >= '0' && name[i + 3] <= '7') {
 744             name[j++] = (name[i + 1] - '0') * 64 +
 745                         (name[i + 2] - '0') * 8 +
 746                         (name[i + 3] - '0');
 747             i += 3;
 748         } else {
 749             name[j++] = name[i];
 750         }
 751     }
 752 }
 753
 754 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
 755 {
 756     FsMount *mount;
 757     char const *mountinfo = "/proc/self/mountinfo";
 758     FILE *fp;
 759     char *line = NULL, *dash;
 760     size_t n;
 761     char check;
 762     unsigned int devmajor, devminor;
 763     int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
 764
 765     fp = fopen(mountinfo, "r");
 766     if (!fp) {
 767         build_fs_mount_list_from_mtab(mounts, errp);
 768         return;
 769     }
 770
 771     while (getline(&line, &n, fp) != -1) {
 772         ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
 773                      &devmajor, &devminor, &dir_s, &dir_e, &check);
 774         if (ret < 3) {
 775             continue;
 776         }
 777         dash = strstr(line + dir_e, " - ");
 778         if (!dash) {
 779             continue;
 780         }
 781         ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
 782                      &type_s, &type_e, &dev_s, &dev_e, &check);
 783         if (ret < 1) {
 784             continue;
 785         }
 786         line[dir_e] = 0;
 787         dash[type_e] = 0;
 788         dash[dev_e] = 0;
 789         decode_mntname(line + dir_s, dir_e - dir_s);
 790         decode_mntname(dash + dev_s, dev_e - dev_s);
 791         if (devmajor == 0) {
 792             /* btrfs reports major number = 0 */
 793             if (strcmp("btrfs", dash + type_s) != 0 ||
 794                 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
 795                 continue;
 796             }
 797         }
 798
 799         mount = g_new0(FsMount, 1);
 800         mount->dirname = g_strdup(line + dir_s);
 801         mount->devtype = g_strdup(dash + type_s);
 802         mount->devmajor = devmajor;
 803         mount->devminor = devminor;
 804
 805         QTAILQ_INSERT_TAIL(mounts, mount, next);
 806     }
 807     free(line);
 808
 809     fclose(fp);
 810 }
 811 #endif
 812
 813 #if defined(CONFIG_FSFREEZE)
 814
 815 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
 816 {
 817     char *path;
 818     char *dpath;
 819     char *driver = NULL;
 820     char buf[PATH_MAX];
 821     ssize_t len;
 822
 823     path = g_strndup(syspath, pathlen);
 824     dpath = g_strdup_printf("%s/driver", path);
 825     len = readlink(dpath, buf, sizeof(buf) - 1);
 826     if (len != -1) {
 827         buf[len] = 0;
 828         driver = g_path_get_basename(buf);
 829     }
 830     g_free(dpath);
 831     g_free(path);
 832     return driver;
 833 }
 834
 835 static int compare_uint(const void *_a, const void *_b)
 836 {
 837     unsigned int a = *(unsigned int *)_a;
 838     unsigned int b = *(unsigned int *)_b;
 839
 840     return a < b ? -1 : a > b ? 1 : 0;
 841 }
 842
 843 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
 844 static int build_hosts(char const *syspath, char const *host, bool ata,
 845                        unsigned int *hosts, int hosts_max, Error **errp)
 846 {
 847     char *path;
 848     DIR *dir;
 849     struct dirent *entry;
 850     int i = 0;
 851
 852     path = g_strndup(syspath, host - syspath);
 853     dir = opendir(path);
 854     if (!dir) {
 855         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
 856         g_free(path);
 857         return -1;
 858     }
 859
 860     while (i < hosts_max) {
 861         entry = readdir(dir);
 862         if (!entry) {
 863             break;
 864         }
 865         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
 866             ++i;
 867         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
 868             ++i;
 869         }
 870     }
 871
 872     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
 873
 874     g_free(path);
 875     closedir(dir);
 876     return i;
 877 }
 878
 879 /* Store disk device info specified by @sysfs into @fs */
 880 static void build_guest_fsinfo_for_real_device(char const *syspath,
 881                                                GuestFilesystemInfo *fs,
 882                                                Error **errp)
 883 {
 884     unsigned int pci[4], host, hosts[8], tgt[3];
 885     int i, nhosts = 0, pcilen;
 886     GuestDiskAddress *disk;
 887     GuestPCIAddress *pciaddr;
 888     GuestDiskAddressList *list = NULL;
 889     bool has_ata = false, has_host = false, has_tgt = false;
 890     char *p, *q, *driver = NULL;
 891 #ifdef CONFIG_LIBUDEV
 892     struct udev *udev = NULL;
 893     struct udev_device *udevice = NULL;
 894 #endif
 895
 896     p = strstr(syspath, "/devices/pci");
 897     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
 898                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
 899         g_debug("only pci device is supported: sysfs path '%s'", syspath);
 900         return;
 901     }
 902
 903     p += 12 + pcilen;
 904     while (true) {
 905         driver = get_pci_driver(syspath, p - syspath, errp);
 906         if (driver && (g_str_equal(driver, "ata_piix") ||
 907                        g_str_equal(driver, "sym53c8xx") ||
 908                        g_str_equal(driver, "virtio-pci") ||
 909                        g_str_equal(driver, "ahci"))) {
 910             break;
 911         }
 912
 913         g_free(driver);
 914         if (sscanf(p, "/%x:%x:%x.%x%n",
 915                           pci, pci + 1, pci + 2, pci + 3, &pcilen) == 4) {
 916             p += pcilen;
 917             continue;
 918         }
 919
 920         g_debug("unsupported driver or sysfs path '%s'", syspath);
 921         return;
 922     }
 923
 924     p = strstr(syspath, "/target");
 925     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
 926                     tgt, tgt + 1, tgt + 2) == 3) {
 927         has_tgt = true;
 928     }
 929
 930     p = strstr(syspath, "/ata");
 931     if (p) {
 932         q = p + 4;
 933         has_ata = true;
 934     } else {
 935         p = strstr(syspath, "/host");
 936         q = p + 5;
 937     }
 938     if (p && sscanf(q, "%u", &host) == 1) {
 939         has_host = true;
 940         nhosts = build_hosts(syspath, p, has_ata, hosts,
 941                              ARRAY_SIZE(hosts), errp);
 942         if (nhosts < 0) {
 943             goto cleanup;
 944         }
 945     }
 946
 947     pciaddr = g_malloc0(sizeof(*pciaddr));
 948     pciaddr->domain = pci[0];
 949     pciaddr->bus = pci[1];
 950     pciaddr->slot = pci[2];
 951     pciaddr->function = pci[3];
 952
 953     disk = g_malloc0(sizeof(*disk));
 954     disk->pci_controller = pciaddr;
 955
 956     list = g_malloc0(sizeof(*list));
 957     list->value = disk;
 958
 959 #ifdef CONFIG_LIBUDEV
 960     udev = udev_new();
 961     udevice = udev_device_new_from_syspath(udev, syspath);
 962     if (udev == NULL || udevice == NULL) {
 963         g_debug("failed to query udev");
 964     } else {
 965         const char *devnode, *serial;
 966         devnode = udev_device_get_devnode(udevice);
 967         if (devnode != NULL) {
 968             disk->dev = g_strdup(devnode);
 969             disk->has_dev = true;
 970         }
 971         serial = udev_device_get_property_value(udevice, "ID_SERIAL");
 972         if (serial != NULL && *serial != 0) {
 973             disk->serial = g_strdup(serial);
 974             disk->has_serial = true;
 975         }
 976     }
 977 #endif
 978
 979     if (strcmp(driver, "ata_piix") == 0) {
 980         /* a host per ide bus, target*:0:<unit>:0 */
 981         if (!has_host || !has_tgt) {
 982             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 983             goto cleanup;
 984         }
 985         for (i = 0; i < nhosts; i++) {
 986             if (host == hosts[i]) {
 987                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
 988                 disk->bus = i;
 989                 disk->unit = tgt[1];
 990                 break;
 991             }
 992         }
 993         if (i >= nhosts) {
 994             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 995             goto cleanup;
 996         }
 997     } else if (strcmp(driver, "sym53c8xx") == 0) {
 998         /* scsi(LSI Logic): target*:0:<unit>:0 */
 999         if (!has_tgt) {
1000             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
1001             goto cleanup;
1002         }
1003         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
1004         disk->unit = tgt[1];
1005     } else if (strcmp(driver, "virtio-pci") == 0) {
1006         if (has_tgt) {
1007             /* virtio-scsi: target*:0:0:<unit> */
1008             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
1009             disk->unit = tgt[2];
1010         } else {
1011             /* virtio-blk: 1 disk per 1 device */
1012             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
1013         }
1014     } else if (strcmp(driver, "ahci") == 0) {
1015         /* ahci: 1 host per 1 unit */
1016         if (!has_host || !has_tgt) {
1017             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
1018             goto cleanup;
1019         }
1020         for (i = 0; i < nhosts; i++) {
1021             if (host == hosts[i]) {
1022                 disk->unit = i;
1023                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
1024                 break;
1025             }
1026         }
1027         if (i >= nhosts) {
1028             g_debug("no host for '%s' (driver '%s')", syspath, driver);
1029             goto cleanup;
1030         }
1031     } else {
1032         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
1033         goto cleanup;
1034     }
1035
1036     list->next = fs->disk;
1037     fs->disk = list;
1038     goto out;
1039
1040 cleanup:
1041     if (list) {
1042         qapi_free_GuestDiskAddressList(list);
1043     }
1044 out:
1045     g_free(driver);
1046 #ifdef CONFIG_LIBUDEV
1047     udev_unref(udev);
1048     udev_device_unref(udevice);
1049 #endif
1050     return;
1051 }
1052
1053 static void build_guest_fsinfo_for_device(char const *devpath,
1054                                           GuestFilesystemInfo *fs,
1055                                           Error **errp);
1056
1057 /* Store a list of slave devices of virtual volume specified by @syspath into
1058  * @fs */
1059 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1060                                                   GuestFilesystemInfo *fs,
1061                                                   Error **errp)
1062 {
1063     DIR *dir;
1064     char *dirpath;
1065     struct dirent *entry;
1066
1067     dirpath = g_strdup_printf("%s/slaves", syspath);
1068     dir = opendir(dirpath);
1069     if (!dir) {
1070         if (errno != ENOENT) {
1071             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1072         }
1073         g_free(dirpath);
1074         return;
1075     }
1076
1077     for (;;) {
1078         errno = 0;
1079         entry = readdir(dir);
1080         if (entry == NULL) {
1081             if (errno) {
1082                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1083             }
1084             break;
1085         }
1086
1087         if (entry->d_type == DT_LNK) {
1088             char *path;
1089
1090             g_debug(" slave device '%s'", entry->d_name);
1091             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1092             build_guest_fsinfo_for_device(path, fs, errp);
1093             g_free(path);
1094
1095             if (*errp) {
1096                 break;
1097             }
1098         }
1099     }
1100
1101     g_free(dirpath);
1102     closedir(dir);
1103 }
1104
1105 /* Dispatch to functions for virtual/real device */
1106 static void build_guest_fsinfo_for_device(char const *devpath,
1107                                           GuestFilesystemInfo *fs,
1108                                           Error **errp)
1109 {
1110     char *syspath = realpath(devpath, NULL);
1111
1112     if (!syspath) {
1113         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1114         return;
1115     }
1116
1117     if (!fs->name) {
1118         fs->name = g_path_get_basename(syspath);
1119     }
1120
1121     g_debug("  parse sysfs path '%s'", syspath);
1122
1123     if (strstr(syspath, "/devices/virtual/block/")) {
1124         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1125     } else {
1126         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1127     }
1128
1129     free(syspath);
1130 }
1131
1132 /* Return a list of the disk device(s)' info which @mount lies on */
1133 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1134                                                Error **errp)
1135 {
1136     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1137     struct statvfs buf;
1138     unsigned long used, nonroot_total, fr_size;
1139     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1140                                     mount->devmajor, mount->devminor);
1141
1142     fs->mountpoint = g_strdup(mount->dirname);
1143     fs->type = g_strdup(mount->devtype);
1144     build_guest_fsinfo_for_device(devpath, fs, errp);
1145
1146     if (statvfs(fs->mountpoint, &buf) == 0) {
1147         fr_size = buf.f_frsize;
1148         used = buf.f_blocks - buf.f_bfree;
1149         nonroot_total = used + buf.f_bavail;
1150         fs->used_bytes = used * fr_size;
1151         fs->total_bytes = nonroot_total * fr_size;
1152
1153         fs->has_total_bytes = true;
1154         fs->has_used_bytes = true;
1155     }
1156
1157     g_free(devpath);
1158
1159     return fs;
1160 }
1161
1162 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1163 {
1164     FsMountList mounts;
1165     struct FsMount *mount;
1166     GuestFilesystemInfoList *new, *ret = NULL;
1167     Error *local_err = NULL;
1168
1169     QTAILQ_INIT(&mounts);
1170     build_fs_mount_list(&mounts, &local_err);
1171     if (local_err) {
1172         error_propagate(errp, local_err);
1173         return NULL;
1174     }
1175
1176     QTAILQ_FOREACH(mount, &mounts, next) {
1177         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1178
1179         new = g_malloc0(sizeof(*ret));
1180         new->value = build_guest_fsinfo(mount, &local_err);
1181         new->next = ret;
1182         ret = new;
1183         if (local_err) {
1184             error_propagate(errp, local_err);
1185             qapi_free_GuestFilesystemInfoList(ret);
1186             ret = NULL;
1187             break;
1188         }
1189     }
1190
1191     free_fs_mount_list(&mounts);
1192     return ret;
1193 }
1194
1195
1196 typedef enum {
1197     FSFREEZE_HOOK_THAW = 0,
1198     FSFREEZE_HOOK_FREEZE,
1199 } FsfreezeHookArg;
1200
1201 static const char *fsfreeze_hook_arg_string[] = {
1202     "thaw",
1203     "freeze",
1204 };
1205
1206 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1207 {
1208     int status;
1209     pid_t pid;
1210     const char *hook;
1211     const char *arg_str = fsfreeze_hook_arg_string[arg];
1212     Error *local_err = NULL;
1213
1214     hook = ga_fsfreeze_hook(ga_state);
1215     if (!hook) {
1216         return;
1217     }
1218     if (access(hook, X_OK) != 0) {
1219         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1220         return;
1221     }
1222
1223     slog("executing fsfreeze hook with arg '%s'", arg_str);
1224     pid = fork();
1225     if (pid == 0) {
1226         setsid();
1227         reopen_fd_to_null(0);
1228         reopen_fd_to_null(1);
1229         reopen_fd_to_null(2);
1230
1231         execle(hook, hook, arg_str, NULL, environ);
1232         _exit(EXIT_FAILURE);
1233     } else if (pid < 0) {
1234         error_setg_errno(errp, errno, "failed to create child process");
1235         return;
1236     }
1237
1238     ga_wait_child(pid, &status, &local_err);
1239     if (local_err) {
1240         error_propagate(errp, local_err);
1241         return;
1242     }
1243
1244     if (!WIFEXITED(status)) {
1245         error_setg(errp, "fsfreeze hook has terminated abnormally");
1246         return;
1247     }
1248
1249     status = WEXITSTATUS(status);
1250     if (status) {
1251         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1252         return;
1253     }
1254 }
1255
1256 /*
1257  * Return status of freeze/thaw
1258  */
1259 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1260 {
1261     if (ga_is_frozen(ga_state)) {
1262         return GUEST_FSFREEZE_STATUS_FROZEN;
1263     }
1264
1265     return GUEST_FSFREEZE_STATUS_THAWED;
1266 }
1267
1268 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1269 {
1270     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1271 }
1272
1273 /*
1274  * Walk list of mounted file systems in the guest, and freeze the ones which
1275  * are real local file systems.
1276  */
1277 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1278                                        strList *mountpoints,
1279                                        Error **errp)
1280 {
1281     int ret = 0, i = 0;
1282     strList *list;
1283     FsMountList mounts;
1284     struct FsMount *mount;
1285     Error *local_err = NULL;
1286     int fd;
1287
1288     slog("guest-fsfreeze called");
1289
1290     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1291     if (local_err) {
1292         error_propagate(errp, local_err);
1293         return -1;
1294     }
1295
1296     QTAILQ_INIT(&mounts);
1297     build_fs_mount_list(&mounts, &local_err);
1298     if (local_err) {
1299         error_propagate(errp, local_err);
1300         return -1;
1301     }
1302
1303     /* cannot risk guest agent blocking itself on a write in this state */
1304     ga_set_frozen(ga_state);
1305
1306     QTAILQ_FOREACH_REVERSE(mount, &mounts, next) {
1307         /* To issue fsfreeze in the reverse order of mounts, check if the
1308          * mount is listed in the list here */
1309         if (has_mountpoints) {
1310             for (list = mountpoints; list; list = list->next) {
1311                 if (strcmp(list->value, mount->dirname) == 0) {
1312                     break;
1313                 }
1314             }
1315             if (!list) {
1316                 continue;
1317             }
1318         }
1319
1320         fd = qemu_open(mount->dirname, O_RDONLY);
1321         if (fd == -1) {
1322             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1323             goto error;
1324         }
1325
1326         /* we try to cull filesystems we know won't work in advance, but other
1327          * filesystems may not implement fsfreeze for less obvious reasons.
1328          * these will report EOPNOTSUPP. we simply ignore these when tallying
1329          * the number of frozen filesystems.
1330          * if a filesystem is mounted more than once (aka bind mount) a
1331          * consecutive attempt to freeze an already frozen filesystem will
1332          * return EBUSY.
1333          *
1334          * any other error means a failure to freeze a filesystem we
1335          * expect to be freezable, so return an error in those cases
1336          * and return system to thawed state.
1337          */
1338         ret = ioctl(fd, FIFREEZE);
1339         if (ret == -1) {
1340             if (errno != EOPNOTSUPP && errno != EBUSY) {
1341                 error_setg_errno(errp, errno, "failed to freeze %s",
1342                                  mount->dirname);
1343                 close(fd);
1344                 goto error;
1345             }
1346         } else {
1347             i++;
1348         }
1349         close(fd);
1350     }
1351
1352     free_fs_mount_list(&mounts);
1353     /* We may not issue any FIFREEZE here.
1354      * Just unset ga_state here and ready for the next call.
1355      */
1356     if (i == 0) {
1357         ga_unset_frozen(ga_state);
1358     }
1359     return i;
1360
1361 error:
1362     free_fs_mount_list(&mounts);
1363     qmp_guest_fsfreeze_thaw(NULL);
1364     return 0;
1365 }
1366
1367 /*
1368  * Walk list of frozen file systems in the guest, and thaw them.
1369  */
1370 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1371 {
1372     int ret;
1373     FsMountList mounts;
1374     FsMount *mount;
1375     int fd, i = 0, logged;
1376     Error *local_err = NULL;
1377
1378     QTAILQ_INIT(&mounts);
1379     build_fs_mount_list(&mounts, &local_err);
1380     if (local_err) {
1381         error_propagate(errp, local_err);
1382         return 0;
1383     }
1384
1385     QTAILQ_FOREACH(mount, &mounts, next) {
1386         logged = false;
1387         fd = qemu_open(mount->dirname, O_RDONLY);
1388         if (fd == -1) {
1389             continue;
1390         }
1391         /* we have no way of knowing whether a filesystem was actually unfrozen
1392          * as a result of a successful call to FITHAW, only that if an error
1393          * was returned the filesystem was *not* unfrozen by that particular
1394          * call.
1395          *
1396          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1397          * to unfreeze, continuing issuing FITHAW until an error is returned,
1398          * in which case either the filesystem is in an unfreezable state, or,
1399          * more likely, it was thawed previously (and remains so afterward).
1400          *
1401          * also, since the most recent successful call is the one that did
1402          * the actual unfreeze, we can use this to provide an accurate count
1403          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1404          * may * be useful for determining whether a filesystem was unfrozen
1405          * during the freeze/thaw phase by a process other than qemu-ga.
1406          */
1407         do {
1408             ret = ioctl(fd, FITHAW);
1409             if (ret == 0 && !logged) {
1410                 i++;
1411                 logged = true;
1412             }
1413         } while (ret == 0);
1414         close(fd);
1415     }
1416
1417     ga_unset_frozen(ga_state);
1418     free_fs_mount_list(&mounts);
1419
1420     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1421
1422     return i;
1423 }
1424
1425 static void guest_fsfreeze_cleanup(void)
1426 {
1427     Error *err = NULL;
1428
1429     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1430         qmp_guest_fsfreeze_thaw(&err);
1431         if (err) {
1432             slog("failed to clean up frozen filesystems: %s",
1433                  error_get_pretty(err));
1434             error_free(err);
1435         }
1436     }
1437 }
1438 #endif /* CONFIG_FSFREEZE */
1439
1440 #if defined(CONFIG_FSTRIM)
1441 /*
1442  * Walk list of mounted file systems in the guest, and trim them.
1443  */
1444 GuestFilesystemTrimResponse *
1445 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1446 {
1447     GuestFilesystemTrimResponse *response;
1448     GuestFilesystemTrimResultList *list;
1449     GuestFilesystemTrimResult *result;
1450     int ret = 0;
1451     FsMountList mounts;
1452     struct FsMount *mount;
1453     int fd;
1454     Error *local_err = NULL;
1455     struct fstrim_range r;
1456
1457     slog("guest-fstrim called");
1458
1459     QTAILQ_INIT(&mounts);
1460     build_fs_mount_list(&mounts, &local_err);
1461     if (local_err) {
1462         error_propagate(errp, local_err);
1463         return NULL;
1464     }
1465
1466     response = g_malloc0(sizeof(*response));
1467
1468     QTAILQ_FOREACH(mount, &mounts, next) {
1469         result = g_malloc0(sizeof(*result));
1470         result->path = g_strdup(mount->dirname);
1471
1472         list = g_malloc0(sizeof(*list));
1473         list->value = result;
1474         list->next = response->paths;
1475         response->paths = list;
1476
1477         fd = qemu_open(mount->dirname, O_RDONLY);
1478         if (fd == -1) {
1479             result->error = g_strdup_printf("failed to open: %s",
1480                                             strerror(errno));
1481             result->has_error = true;
1482             continue;
1483         }
1484
1485         /* We try to cull filesystems we know won't work in advance, but other
1486          * filesystems may not implement fstrim for less obvious reasons.
1487          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1488          * will be reported (e.g. CD-ROMs).
1489          * Any other error means an unexpected error.
1490          */
1491         r.start = 0;
1492         r.len = -1;
1493         r.minlen = has_minimum ? minimum : 0;
1494         ret = ioctl(fd, FITRIM, &r);
1495         if (ret == -1) {
1496             result->has_error = true;
1497             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1498                 result->error = g_strdup("trim not supported");
1499             } else {
1500                 result->error = g_strdup_printf("failed to trim: %s",
1501                                                 strerror(errno));
1502             }
1503             close(fd);
1504             continue;
1505         }
1506
1507         result->has_minimum = true;
1508         result->minimum = r.minlen;
1509         result->has_trimmed = true;
1510         result->trimmed = r.len;
1511         close(fd);
1512     }
1513
1514     free_fs_mount_list(&mounts);
1515     return response;
1516 }
1517 #endif /* CONFIG_FSTRIM */
1518
1519
1520 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1521 #define SUSPEND_SUPPORTED 0
1522 #define SUSPEND_NOT_SUPPORTED 1
1523
1524 typedef enum {
1525     SUSPEND_MODE_DISK = 0,
1526     SUSPEND_MODE_RAM = 1,
1527     SUSPEND_MODE_HYBRID = 2,
1528 } SuspendMode;
1529
1530 /*
1531  * Executes a command in a child process using g_spawn_sync,
1532  * returning an int >= 0 representing the exit status of the
1533  * process.
1534  *
1535  * If the program wasn't found in path, returns -1.
1536  *
1537  * If a problem happened when creating the child process,
1538  * returns -1 and errp is set.
1539  */
1540 static int run_process_child(const char *command[], Error **errp)
1541 {
1542     int exit_status, spawn_flag;
1543     GError *g_err = NULL;
1544     bool success;
1545
1546     spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1547                  G_SPAWN_STDERR_TO_DEV_NULL;
1548
1549     success =  g_spawn_sync(NULL, (char **)command, environ, spawn_flag,
1550                             NULL, NULL, NULL, NULL,
1551                             &exit_status, &g_err);
1552
1553     if (success) {
1554         return WEXITSTATUS(exit_status);
1555     }
1556
1557     if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1558         error_setg(errp, "failed to create child process, error '%s'",
1559                    g_err->message);
1560     }
1561
1562     g_error_free(g_err);
1563     return -1;
1564 }
1565
1566 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1567 {
1568     Error *local_err = NULL;
1569     const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1570                                      "systemd-hybrid-sleep"};
1571     const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1572     int status;
1573
1574     status = run_process_child(cmd, &local_err);
1575
1576     /*
1577      * systemctl status uses LSB return codes so we can expect
1578      * status > 0 and be ok. To assert if the guest has support
1579      * for the selected suspend mode, status should be < 4. 4 is
1580      * the code for unknown service status, the return value when
1581      * the service does not exist. A common value is status = 3
1582      * (program is not running).
1583      */
1584     if (status > 0 && status < 4) {
1585         return true;
1586     }
1587
1588     error_propagate(errp, local_err);
1589     return false;
1590 }
1591
1592 static void systemd_suspend(SuspendMode mode, Error **errp)
1593 {
1594     Error *local_err = NULL;
1595     const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1596     const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1597     int status;
1598
1599     status = run_process_child(cmd, &local_err);
1600
1601     if (status == 0) {
1602         return;
1603     }
1604
1605     if ((status == -1) && !local_err) {
1606         error_setg(errp, "the helper program 'systemctl %s' was not found",
1607                    systemctl_args[mode]);
1608         return;
1609     }
1610
1611     if (local_err) {
1612         error_propagate(errp, local_err);
1613     } else {
1614         error_setg(errp, "the helper program 'systemctl %s' returned an "
1615                    "unexpected exit status code (%d)",
1616                    systemctl_args[mode], status);
1617     }
1618 }
1619
1620 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1621 {
1622     Error *local_err = NULL;
1623     const char *pmutils_args[3] = {"--hibernate", "--suspend",
1624                                    "--suspend-hybrid"};
1625     const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1626     int status;
1627
1628     status = run_process_child(cmd, &local_err);
1629
1630     if (status == SUSPEND_SUPPORTED) {
1631         return true;
1632     }
1633
1634     if ((status == -1) && !local_err) {
1635         return false;
1636     }
1637
1638     if (local_err) {
1639         error_propagate(errp, local_err);
1640     } else {
1641         error_setg(errp,
1642                    "the helper program '%s' returned an unexpected exit"
1643                    " status code (%d)", "pm-is-supported", status);
1644     }
1645
1646     return false;
1647 }
1648
1649 static void pmutils_suspend(SuspendMode mode, Error **errp)
1650 {
1651     Error *local_err = NULL;
1652     const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1653                                        "pm-suspend-hybrid"};
1654     const char *cmd[2] = {pmutils_binaries[mode], NULL};
1655     int status;
1656
1657     status = run_process_child(cmd, &local_err);
1658
1659     if (status == 0) {
1660         return;
1661     }
1662
1663     if ((status == -1) && !local_err) {
1664         error_setg(errp, "the helper program '%s' was not found",
1665                    pmutils_binaries[mode]);
1666         return;
1667     }
1668
1669     if (local_err) {
1670         error_propagate(errp, local_err);
1671     } else {
1672         error_setg(errp,
1673                    "the helper program '%s' returned an unexpected exit"
1674                    " status code (%d)", pmutils_binaries[mode], status);
1675     }
1676 }
1677
1678 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
1679 {
1680     const char *sysfile_strs[3] = {"disk", "mem", NULL};
1681     const char *sysfile_str = sysfile_strs[mode];
1682     char buf[32]; /* hopefully big enough */
1683     int fd;
1684     ssize_t ret;
1685
1686     if (!sysfile_str) {
1687         error_setg(errp, "unknown guest suspend mode");
1688         return false;
1689     }
1690
1691     fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1692     if (fd < 0) {
1693         return false;
1694     }
1695
1696     ret = read(fd, buf, sizeof(buf) - 1);
1697     close(fd);
1698     if (ret <= 0) {
1699         return false;
1700     }
1701     buf[ret] = '\0';
1702
1703     if (strstr(buf, sysfile_str)) {
1704         return true;
1705     }
1706     return false;
1707 }
1708
1709 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
1710 {
1711     Error *local_err = NULL;
1712     const char *sysfile_strs[3] = {"disk", "mem", NULL};
1713     const char *sysfile_str = sysfile_strs[mode];
1714     pid_t pid;
1715     int status;
1716
1717     if (!sysfile_str) {
1718         error_setg(errp, "unknown guest suspend mode");
1719         return;
1720     }
1721
1722     pid = fork();
1723     if (!pid) {
1724         /* child */
1725         int fd;
1726
1727         setsid();
1728         reopen_fd_to_null(0);
1729         reopen_fd_to_null(1);
1730         reopen_fd_to_null(2);
1731
1732         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1733         if (fd < 0) {
1734             _exit(EXIT_FAILURE);
1735         }
1736
1737         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1738             _exit(EXIT_FAILURE);
1739         }
1740
1741         _exit(EXIT_SUCCESS);
1742     } else if (pid < 0) {
1743         error_setg_errno(errp, errno, "failed to create child process");
1744         return;
1745     }
1746
1747     ga_wait_child(pid, &status, &local_err);
1748     if (local_err) {
1749         error_propagate(errp, local_err);
1750         return;
1751     }
1752
1753     if (WEXITSTATUS(status)) {
1754         error_setg(errp, "child process has failed to suspend");
1755     }
1756
1757 }
1758
1759 static void guest_suspend(SuspendMode mode, Error **errp)
1760 {
1761     Error *local_err = NULL;
1762     bool mode_supported = false;
1763
1764     if (systemd_supports_mode(mode, &local_err)) {
1765         mode_supported = true;
1766         systemd_suspend(mode, &local_err);
1767     }
1768
1769     if (!local_err) {
1770         return;
1771     }
1772
1773     error_free(local_err);
1774
1775     if (pmutils_supports_mode(mode, &local_err)) {
1776         mode_supported = true;
1777         pmutils_suspend(mode, &local_err);
1778     }
1779
1780     if (!local_err) {
1781         return;
1782     }
1783
1784     error_free(local_err);
1785
1786     if (linux_sys_state_supports_mode(mode, &local_err)) {
1787         mode_supported = true;
1788         linux_sys_state_suspend(mode, &local_err);
1789     }
1790
1791     if (!mode_supported) {
1792         error_setg(errp,
1793                    "the requested suspend mode is not supported by the guest");
1794     } else {
1795         error_propagate(errp, local_err);
1796     }
1797 }
1798
1799 void qmp_guest_suspend_disk(Error **errp)
1800 {
1801     guest_suspend(SUSPEND_MODE_DISK, errp);
1802 }
1803
1804 void qmp_guest_suspend_ram(Error **errp)
1805 {
1806     guest_suspend(SUSPEND_MODE_RAM, errp);
1807 }
1808
1809 void qmp_guest_suspend_hybrid(Error **errp)
1810 {
1811     guest_suspend(SUSPEND_MODE_HYBRID, errp);
1812 }
1813
1814 static GuestNetworkInterfaceList *
1815 guest_find_interface(GuestNetworkInterfaceList *head,
1816                      const char *name)
1817 {
1818     for (; head; head = head->next) {
1819         if (strcmp(head->value->name, name) == 0) {
1820             break;
1821         }
1822     }
1823
1824     return head;
1825 }
1826
1827 static int guest_get_network_stats(const char *name,
1828                        GuestNetworkInterfaceStat *stats)
1829 {
1830     int name_len;
1831     char const *devinfo = "/proc/net/dev";
1832     FILE *fp;
1833     char *line = NULL, *colon;
1834     size_t n = 0;
1835     fp = fopen(devinfo, "r");
1836     if (!fp) {
1837         return -1;
1838     }
1839     name_len = strlen(name);
1840     while (getline(&line, &n, fp) != -1) {
1841         long long dummy;
1842         long long rx_bytes;
1843         long long rx_packets;
1844         long long rx_errs;
1845         long long rx_dropped;
1846         long long tx_bytes;
1847         long long tx_packets;
1848         long long tx_errs;
1849         long long tx_dropped;
1850         char *trim_line;
1851         trim_line = g_strchug(line);
1852         if (trim_line[0] == '\0') {
1853             continue;
1854         }
1855         colon = strchr(trim_line, ':');
1856         if (!colon) {
1857             continue;
1858         }
1859         if (colon - name_len  == trim_line &&
1860            strncmp(trim_line, name, name_len) == 0) {
1861             if (sscanf(colon + 1,
1862                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
1863                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
1864                   &dummy, &dummy, &dummy, &dummy,
1865                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
1866                   &dummy, &dummy, &dummy, &dummy) != 16) {
1867                 continue;
1868             }
1869             stats->rx_bytes = rx_bytes;
1870             stats->rx_packets = rx_packets;
1871             stats->rx_errs = rx_errs;
1872             stats->rx_dropped = rx_dropped;
1873             stats->tx_bytes = tx_bytes;
1874             stats->tx_packets = tx_packets;
1875             stats->tx_errs = tx_errs;
1876             stats->tx_dropped = tx_dropped;
1877             fclose(fp);
1878             g_free(line);
1879             return 0;
1880         }
1881     }
1882     fclose(fp);
1883     g_free(line);
1884     g_debug("/proc/net/dev: Interface '%s' not found", name);
1885     return -1;
1886 }
1887
1888 /*
1889  * Build information about guest interfaces
1890  */
1891 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1892 {
1893     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1894     struct ifaddrs *ifap, *ifa;
1895
1896     if (getifaddrs(&ifap) < 0) {
1897         error_setg_errno(errp, errno, "getifaddrs failed");
1898         goto error;
1899     }
1900
1901     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1902         GuestNetworkInterfaceList *info;
1903         GuestIpAddressList **address_list = NULL, *address_item = NULL;
1904         GuestNetworkInterfaceStat  *interface_stat = NULL;
1905         char addr4[INET_ADDRSTRLEN];
1906         char addr6[INET6_ADDRSTRLEN];
1907         int sock;
1908         struct ifreq ifr;
1909         unsigned char *mac_addr;
1910         void *p;
1911
1912         g_debug("Processing %s interface", ifa->ifa_name);
1913
1914         info = guest_find_interface(head, ifa->ifa_name);
1915
1916         if (!info) {
1917             info = g_malloc0(sizeof(*info));
1918             info->value = g_malloc0(sizeof(*info->value));
1919             info->value->name = g_strdup(ifa->ifa_name);
1920
1921             if (!cur_item) {
1922                 head = cur_item = info;
1923             } else {
1924                 cur_item->next = info;
1925                 cur_item = info;
1926             }
1927         }
1928
1929         if (!info->value->has_hardware_address &&
1930             ifa->ifa_flags & SIOCGIFHWADDR) {
1931             /* we haven't obtained HW address yet */
1932             sock = socket(PF_INET, SOCK_STREAM, 0);
1933             if (sock == -1) {
1934                 error_setg_errno(errp, errno, "failed to create socket");
1935                 goto error;
1936             }
1937
1938             memset(&ifr, 0, sizeof(ifr));
1939             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1940             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1941                 error_setg_errno(errp, errno,
1942                                  "failed to get MAC address of %s",
1943                                  ifa->ifa_name);
1944                 close(sock);
1945                 goto error;
1946             }
1947
1948             close(sock);
1949             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
1950
1951             info->value->hardware_address =
1952                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
1953                                 (int) mac_addr[0], (int) mac_addr[1],
1954                                 (int) mac_addr[2], (int) mac_addr[3],
1955                                 (int) mac_addr[4], (int) mac_addr[5]);
1956
1957             info->value->has_hardware_address = true;
1958         }
1959
1960         if (ifa->ifa_addr &&
1961             ifa->ifa_addr->sa_family == AF_INET) {
1962             /* interface with IPv4 address */
1963             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1964             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
1965                 error_setg_errno(errp, errno, "inet_ntop failed");
1966                 goto error;
1967             }
1968
1969             address_item = g_malloc0(sizeof(*address_item));
1970             address_item->value = g_malloc0(sizeof(*address_item->value));
1971             address_item->value->ip_address = g_strdup(addr4);
1972             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
1973
1974             if (ifa->ifa_netmask) {
1975                 /* Count the number of set bits in netmask.
1976                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1977                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
1978                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
1979             }
1980         } else if (ifa->ifa_addr &&
1981                    ifa->ifa_addr->sa_family == AF_INET6) {
1982             /* interface with IPv6 address */
1983             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
1984             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
1985                 error_setg_errno(errp, errno, "inet_ntop failed");
1986                 goto error;
1987             }
1988
1989             address_item = g_malloc0(sizeof(*address_item));
1990             address_item->value = g_malloc0(sizeof(*address_item->value));
1991             address_item->value->ip_address = g_strdup(addr6);
1992             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
1993
1994             if (ifa->ifa_netmask) {
1995                 /* Count the number of set bits in netmask.
1996                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1997                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
1998                 address_item->value->prefix =
1999                     ctpop32(((uint32_t *) p)[0]) +
2000                     ctpop32(((uint32_t *) p)[1]) +
2001                     ctpop32(((uint32_t *) p)[2]) +
2002                     ctpop32(((uint32_t *) p)[3]);
2003             }
2004         }
2005
2006         if (!address_item) {
2007             continue;
2008         }
2009
2010         address_list = &info->value->ip_addresses;
2011
2012         while (*address_list && (*address_list)->next) {
2013             address_list = &(*address_list)->next;
2014         }
2015
2016         if (!*address_list) {
2017             *address_list = address_item;
2018         } else {
2019             (*address_list)->next = address_item;
2020         }
2021
2022         info->value->has_ip_addresses = true;
2023
2024         if (!info->value->has_statistics) {
2025             interface_stat = g_malloc0(sizeof(*interface_stat));
2026             if (guest_get_network_stats(info->value->name,
2027                 interface_stat) == -1) {
2028                 info->value->has_statistics = false;
2029                 g_free(interface_stat);
2030             } else {
2031                 info->value->statistics = interface_stat;
2032                 info->value->has_statistics = true;
2033             }
2034         }
2035     }
2036
2037     freeifaddrs(ifap);
2038     return head;
2039
2040 error:
2041     freeifaddrs(ifap);
2042     qapi_free_GuestNetworkInterfaceList(head);
2043     return NULL;
2044 }
2045
2046 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
2047
2048 static long sysconf_exact(int name, const char *name_str, Error **errp)
2049 {
2050     long ret;
2051
2052     errno = 0;
2053     ret = sysconf(name);
2054     if (ret == -1) {
2055         if (errno == 0) {
2056             error_setg(errp, "sysconf(%s): value indefinite", name_str);
2057         } else {
2058             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
2059         }
2060     }
2061     return ret;
2062 }
2063
2064 /* Transfer online/offline status between @vcpu and the guest system.
2065  *
2066  * On input either @errp or *@errp must be NULL.
2067  *
2068  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
2069  * - R: vcpu->logical_id
2070  * - W: vcpu->online
2071  * - W: vcpu->can_offline
2072  *
2073  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
2074  * - R: vcpu->logical_id
2075  * - R: vcpu->online
2076  *
2077  * Written members remain unmodified on error.
2078  */
2079 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
2080                           char *dirpath, Error **errp)
2081 {
2082     int fd;
2083     int res;
2084     int dirfd;
2085     static const char fn[] = "online";
2086
2087     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2088     if (dirfd == -1) {
2089         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2090         return;
2091     }
2092
2093     fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2094     if (fd == -1) {
2095         if (errno != ENOENT) {
2096             error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2097         } else if (sys2vcpu) {
2098             vcpu->online = true;
2099             vcpu->can_offline = false;
2100         } else if (!vcpu->online) {
2101             error_setg(errp, "logical processor #%" PRId64 " can't be "
2102                        "offlined", vcpu->logical_id);
2103         } /* otherwise pretend successful re-onlining */
2104     } else {
2105         unsigned char status;
2106
2107         res = pread(fd, &status, 1, 0);
2108         if (res == -1) {
2109             error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2110         } else if (res == 0) {
2111             error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2112                        fn);
2113         } else if (sys2vcpu) {
2114             vcpu->online = (status != '0');
2115             vcpu->can_offline = true;
2116         } else if (vcpu->online != (status != '0')) {
2117             status = '0' + vcpu->online;
2118             if (pwrite(fd, &status, 1, 0) == -1) {
2119                 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2120                                  fn);
2121             }
2122         } /* otherwise pretend successful re-(on|off)-lining */
2123
2124         res = close(fd);
2125         g_assert(res == 0);
2126     }
2127
2128     res = close(dirfd);
2129     g_assert(res == 0);
2130 }
2131
2132 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2133 {
2134     int64_t current;
2135     GuestLogicalProcessorList *head, **link;
2136     long sc_max;
2137     Error *local_err = NULL;
2138
2139     current = 0;
2140     head = NULL;
2141     link = &head;
2142     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
2143
2144     while (local_err == NULL && current < sc_max) {
2145         GuestLogicalProcessor *vcpu;
2146         GuestLogicalProcessorList *entry;
2147         int64_t id = current++;
2148         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2149                                      id);
2150
2151         if (g_file_test(path, G_FILE_TEST_EXISTS)) {
2152             vcpu = g_malloc0(sizeof *vcpu);
2153             vcpu->logical_id = id;
2154             vcpu->has_can_offline = true; /* lolspeak ftw */
2155             transfer_vcpu(vcpu, true, path, &local_err);
2156             entry = g_malloc0(sizeof *entry);
2157             entry->value = vcpu;
2158             *link = entry;
2159             link = &entry->next;
2160         }
2161         g_free(path);
2162     }
2163
2164     if (local_err == NULL) {
2165         /* there's no guest with zero VCPUs */
2166         g_assert(head != NULL);
2167         return head;
2168     }
2169
2170     qapi_free_GuestLogicalProcessorList(head);
2171     error_propagate(errp, local_err);
2172     return NULL;
2173 }
2174
2175 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2176 {
2177     int64_t processed;
2178     Error *local_err = NULL;
2179
2180     processed = 0;
2181     while (vcpus != NULL) {
2182         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2183                                      vcpus->value->logical_id);
2184
2185         transfer_vcpu(vcpus->value, false, path, &local_err);
2186         g_free(path);
2187         if (local_err != NULL) {
2188             break;
2189         }
2190         ++processed;
2191         vcpus = vcpus->next;
2192     }
2193
2194     if (local_err != NULL) {
2195         if (processed == 0) {
2196             error_propagate(errp, local_err);
2197         } else {
2198             error_free(local_err);
2199         }
2200     }
2201
2202     return processed;
2203 }
2204
2205 void qmp_guest_set_user_password(const char *username,
2206                                  const char *password,
2207                                  bool crypted,
2208                                  Error **errp)
2209 {
2210     Error *local_err = NULL;
2211     char *passwd_path = NULL;
2212     pid_t pid;
2213     int status;
2214     int datafd[2] = { -1, -1 };
2215     char *rawpasswddata = NULL;
2216     size_t rawpasswdlen;
2217     char *chpasswddata = NULL;
2218     size_t chpasswdlen;
2219
2220     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2221     if (!rawpasswddata) {
2222         return;
2223     }
2224     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2225     rawpasswddata[rawpasswdlen] = '\0';
2226
2227     if (strchr(rawpasswddata, '\n')) {
2228         error_setg(errp, "forbidden characters in raw password");
2229         goto out;
2230     }
2231
2232     if (strchr(username, '\n') ||
2233         strchr(username, ':')) {
2234         error_setg(errp, "forbidden characters in username");
2235         goto out;
2236     }
2237
2238     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2239     chpasswdlen = strlen(chpasswddata);
2240
2241     passwd_path = g_find_program_in_path("chpasswd");
2242
2243     if (!passwd_path) {
2244         error_setg(errp, "cannot find 'passwd' program in PATH");
2245         goto out;
2246     }
2247
2248     if (pipe(datafd) < 0) {
2249         error_setg(errp, "cannot create pipe FDs");
2250         goto out;
2251     }
2252
2253     pid = fork();
2254     if (pid == 0) {
2255         close(datafd[1]);
2256         /* child */
2257         setsid();
2258         dup2(datafd[0], 0);
2259         reopen_fd_to_null(1);
2260         reopen_fd_to_null(2);
2261
2262         if (crypted) {
2263             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2264         } else {
2265             execle(passwd_path, "chpasswd", NULL, environ);
2266         }
2267         _exit(EXIT_FAILURE);
2268     } else if (pid < 0) {
2269         error_setg_errno(errp, errno, "failed to create child process");
2270         goto out;
2271     }
2272     close(datafd[0]);
2273     datafd[0] = -1;
2274
2275     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2276         error_setg_errno(errp, errno, "cannot write new account password");
2277         goto out;
2278     }
2279     close(datafd[1]);
2280     datafd[1] = -1;
2281
2282     ga_wait_child(pid, &status, &local_err);
2283     if (local_err) {
2284         error_propagate(errp, local_err);
2285         goto out;
2286     }
2287
2288     if (!WIFEXITED(status)) {
2289         error_setg(errp, "child process has terminated abnormally");
2290         goto out;
2291     }
2292
2293     if (WEXITSTATUS(status)) {
2294         error_setg(errp, "child process has failed to set user password");
2295         goto out;
2296     }
2297
2298 out:
2299     g_free(chpasswddata);
2300     g_free(rawpasswddata);
2301     g_free(passwd_path);
2302     if (datafd[0] != -1) {
2303         close(datafd[0]);
2304     }
2305     if (datafd[1] != -1) {
2306         close(datafd[1]);
2307     }
2308 }
2309
2310 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2311                                int size, Error **errp)
2312 {
2313     int fd;
2314     int res;
2315
2316     errno = 0;
2317     fd = openat(dirfd, pathname, O_RDONLY);
2318     if (fd == -1) {
2319         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2320         return;
2321     }
2322
2323     res = pread(fd, buf, size, 0);
2324     if (res == -1) {
2325         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2326     } else if (res == 0) {
2327         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2328     }
2329     close(fd);
2330 }
2331
2332 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2333                                 const char *buf, int size, Error **errp)
2334 {
2335     int fd;
2336
2337     errno = 0;
2338     fd = openat(dirfd, pathname, O_WRONLY);
2339     if (fd == -1) {
2340         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2341         return;
2342     }
2343
2344     if (pwrite(fd, buf, size, 0) == -1) {
2345         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2346     }
2347
2348     close(fd);
2349 }
2350
2351 /* Transfer online/offline status between @mem_blk and the guest system.
2352  *
2353  * On input either @errp or *@errp must be NULL.
2354  *
2355  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2356  * - R: mem_blk->phys_index
2357  * - W: mem_blk->online
2358  * - W: mem_blk->can_offline
2359  *
2360  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2361  * - R: mem_blk->phys_index
2362  * - R: mem_blk->online
2363  *-  R: mem_blk->can_offline
2364  * Written members remain unmodified on error.
2365  */
2366 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2367                                   GuestMemoryBlockResponse *result,
2368                                   Error **errp)
2369 {
2370     char *dirpath;
2371     int dirfd;
2372     char *status;
2373     Error *local_err = NULL;
2374
2375     if (!sys2memblk) {
2376         DIR *dp;
2377
2378         if (!result) {
2379             error_setg(errp, "Internal error, 'result' should not be NULL");
2380             return;
2381         }
2382         errno = 0;
2383         dp = opendir("/sys/devices/system/memory/");
2384          /* if there is no 'memory' directory in sysfs,
2385          * we think this VM does not support online/offline memory block,
2386          * any other solution?
2387          */
2388         if (!dp) {
2389             if (errno == ENOENT) {
2390                 result->response =
2391                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2392             }
2393             goto out1;
2394         }
2395         closedir(dp);
2396     }
2397
2398     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2399                               mem_blk->phys_index);
2400     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2401     if (dirfd == -1) {
2402         if (sys2memblk) {
2403             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2404         } else {
2405             if (errno == ENOENT) {
2406                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2407             } else {
2408                 result->response =
2409                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2410             }
2411         }
2412         g_free(dirpath);
2413         goto out1;
2414     }
2415     g_free(dirpath);
2416
2417     status = g_malloc0(10);
2418     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2419     if (local_err) {
2420         /* treat with sysfs file that not exist in old kernel */
2421         if (errno == ENOENT) {
2422             error_free(local_err);
2423             if (sys2memblk) {
2424                 mem_blk->online = true;
2425                 mem_blk->can_offline = false;
2426             } else if (!mem_blk->online) {
2427                 result->response =
2428                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2429             }
2430         } else {
2431             if (sys2memblk) {
2432                 error_propagate(errp, local_err);
2433             } else {
2434                 result->response =
2435                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2436             }
2437         }
2438         goto out2;
2439     }
2440
2441     if (sys2memblk) {
2442         char removable = '0';
2443
2444         mem_blk->online = (strncmp(status, "online", 6) == 0);
2445
2446         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2447         if (local_err) {
2448             /* if no 'removable' file, it doesn't support offline mem blk */
2449             if (errno == ENOENT) {
2450                 error_free(local_err);
2451                 mem_blk->can_offline = false;
2452             } else {
2453                 error_propagate(errp, local_err);
2454             }
2455         } else {
2456             mem_blk->can_offline = (removable != '0');
2457         }
2458     } else {
2459         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2460             const char *new_state = mem_blk->online ? "online" : "offline";
2461
2462             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2463                                 &local_err);
2464             if (local_err) {
2465                 error_free(local_err);
2466                 result->response =
2467                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2468                 goto out2;
2469             }
2470
2471             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2472             result->has_error_code = false;
2473         } /* otherwise pretend successful re-(on|off)-lining */
2474     }
2475     g_free(status);
2476     close(dirfd);
2477     return;
2478
2479 out2:
2480     g_free(status);
2481     close(dirfd);
2482 out1:
2483     if (!sys2memblk) {
2484         result->has_error_code = true;
2485         result->error_code = errno;
2486     }
2487 }
2488
2489 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2490 {
2491     GuestMemoryBlockList *head, **link;
2492     Error *local_err = NULL;
2493     struct dirent *de;
2494     DIR *dp;
2495
2496     head = NULL;
2497     link = &head;
2498
2499     dp = opendir("/sys/devices/system/memory/");
2500     if (!dp) {
2501         /* it's ok if this happens to be a system that doesn't expose
2502          * memory blocks via sysfs, but otherwise we should report
2503          * an error
2504          */
2505         if (errno != ENOENT) {
2506             error_setg_errno(errp, errno, "Can't open directory"
2507                              "\"/sys/devices/system/memory/\"");
2508         }
2509         return NULL;
2510     }
2511
2512     /* Note: the phys_index of memory block may be discontinuous,
2513      * this is because a memblk is the unit of the Sparse Memory design, which
2514      * allows discontinuous memory ranges (ex. NUMA), so here we should
2515      * traverse the memory block directory.
2516      */
2517     while ((de = readdir(dp)) != NULL) {
2518         GuestMemoryBlock *mem_blk;
2519         GuestMemoryBlockList *entry;
2520
2521         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2522             !(de->d_type & DT_DIR)) {
2523             continue;
2524         }
2525
2526         mem_blk = g_malloc0(sizeof *mem_blk);
2527         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2528         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2529         mem_blk->has_can_offline = true; /* lolspeak ftw */
2530         transfer_memory_block(mem_blk, true, NULL, &local_err);
2531
2532         entry = g_malloc0(sizeof *entry);
2533         entry->value = mem_blk;
2534
2535         *link = entry;
2536         link = &entry->next;
2537     }
2538
2539     closedir(dp);
2540     if (local_err == NULL) {
2541         /* there's no guest with zero memory blocks */
2542         if (head == NULL) {
2543             error_setg(errp, "guest reported zero memory blocks!");
2544         }
2545         return head;
2546     }
2547
2548     qapi_free_GuestMemoryBlockList(head);
2549     error_propagate(errp, local_err);
2550     return NULL;
2551 }
2552
2553 GuestMemoryBlockResponseList *
2554 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2555 {
2556     GuestMemoryBlockResponseList *head, **link;
2557     Error *local_err = NULL;
2558
2559     head = NULL;
2560     link = &head;
2561
2562     while (mem_blks != NULL) {
2563         GuestMemoryBlockResponse *result;
2564         GuestMemoryBlockResponseList *entry;
2565         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2566
2567         result = g_malloc0(sizeof(*result));
2568         result->phys_index = current_mem_blk->phys_index;
2569         transfer_memory_block(current_mem_blk, false, result, &local_err);
2570         if (local_err) { /* should never happen */
2571             goto err;
2572         }
2573         entry = g_malloc0(sizeof *entry);
2574         entry->value = result;
2575
2576         *link = entry;
2577         link = &entry->next;
2578         mem_blks = mem_blks->next;
2579     }
2580
2581     return head;
2582 err:
2583     qapi_free_GuestMemoryBlockResponseList(head);
2584     error_propagate(errp, local_err);
2585     return NULL;
2586 }
2587
2588 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2589 {
2590     Error *local_err = NULL;
2591     char *dirpath;
2592     int dirfd;
2593     char *buf;
2594     GuestMemoryBlockInfo *info;
2595
2596     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2597     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2598     if (dirfd == -1) {
2599         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2600         g_free(dirpath);
2601         return NULL;
2602     }
2603     g_free(dirpath);
2604
2605     buf = g_malloc0(20);
2606     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2607     close(dirfd);
2608     if (local_err) {
2609         g_free(buf);
2610         error_propagate(errp, local_err);
2611         return NULL;
2612     }
2613
2614     info = g_new0(GuestMemoryBlockInfo, 1);
2615     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2616
2617     g_free(buf);
2618
2619     return info;
2620 }
2621
2622 #else /* defined(__linux__) */
2623
2624 void qmp_guest_suspend_disk(Error **errp)
2625 {
2626     error_setg(errp, QERR_UNSUPPORTED);
2627 }
2628
2629 void qmp_guest_suspend_ram(Error **errp)
2630 {
2631     error_setg(errp, QERR_UNSUPPORTED);
2632 }
2633
2634 void qmp_guest_suspend_hybrid(Error **errp)
2635 {
2636     error_setg(errp, QERR_UNSUPPORTED);
2637 }
2638
2639 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2640 {
2641     error_setg(errp, QERR_UNSUPPORTED);
2642     return NULL;
2643 }
2644
2645 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2646 {
2647     error_setg(errp, QERR_UNSUPPORTED);
2648     return NULL;
2649 }
2650
2651 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2652 {
2653     error_setg(errp, QERR_UNSUPPORTED);
2654     return -1;
2655 }
2656
2657 void qmp_guest_set_user_password(const char *username,
2658                                  const char *password,
2659                                  bool crypted,
2660                                  Error **errp)
2661 {
2662     error_setg(errp, QERR_UNSUPPORTED);
2663 }
2664
2665 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2666 {
2667     error_setg(errp, QERR_UNSUPPORTED);
2668     return NULL;
2669 }
2670
2671 GuestMemoryBlockResponseList *
2672 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2673 {
2674     error_setg(errp, QERR_UNSUPPORTED);
2675     return NULL;
2676 }
2677
2678 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2679 {
2680     error_setg(errp, QERR_UNSUPPORTED);
2681     return NULL;
2682 }
2683
2684 #endif
2685
2686 #if !defined(CONFIG_FSFREEZE)
2687
2688 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2689 {
2690     error_setg(errp, QERR_UNSUPPORTED);
2691     return NULL;
2692 }
2693
2694 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2695 {
2696     error_setg(errp, QERR_UNSUPPORTED);
2697
2698     return 0;
2699 }
2700
2701 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2702 {
2703     error_setg(errp, QERR_UNSUPPORTED);
2704
2705     return 0;
2706 }
2707
2708 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2709                                        strList *mountpoints,
2710                                        Error **errp)
2711 {
2712     error_setg(errp, QERR_UNSUPPORTED);
2713
2714     return 0;
2715 }
2716
2717 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2718 {
2719     error_setg(errp, QERR_UNSUPPORTED);
2720
2721     return 0;
2722 }
2723 #endif /* CONFIG_FSFREEZE */
2724
2725 #if !defined(CONFIG_FSTRIM)
2726 GuestFilesystemTrimResponse *
2727 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
2728 {
2729     error_setg(errp, QERR_UNSUPPORTED);
2730     return NULL;
2731 }
2732 #endif
2733
2734 /* add unsupported commands to the blacklist */
2735 GList *ga_command_blacklist_init(GList *blacklist)
2736 {
2737 #if !defined(__linux__)
2738     {
2739         const char *list[] = {
2740             "guest-suspend-disk", "guest-suspend-ram",
2741             "guest-suspend-hybrid", "guest-network-get-interfaces",
2742             "guest-get-vcpus", "guest-set-vcpus",
2743             "guest-get-memory-blocks", "guest-set-memory-blocks",
2744             "guest-get-memory-block-size", "guest-get-memory-block-info",
2745             NULL};
2746         char **p = (char **)list;
2747
2748         while (*p) {
2749             blacklist = g_list_append(blacklist, g_strdup(*p++));
2750         }
2751     }
2752 #endif
2753
2754 #if !defined(CONFIG_FSFREEZE)
2755     {
2756         const char *list[] = {
2757             "guest-get-fsinfo", "guest-fsfreeze-status",
2758             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
2759             "guest-fsfreeze-thaw", "guest-get-fsinfo", NULL};
2760         char **p = (char **)list;
2761
2762         while (*p) {
2763             blacklist = g_list_append(blacklist, g_strdup(*p++));
2764         }
2765     }
2766 #endif
2767
2768 #if !defined(CONFIG_FSTRIM)
2769     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
2770 #endif
2771
2772     return blacklist;
2773 }
2774
2775 /* register init/cleanup routines for stateful command groups */
2776 void ga_command_state_init(GAState *s, GACommandState *cs)
2777 {
2778 #if defined(CONFIG_FSFREEZE)
2779     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
2780 #endif
2781 }
2782
2783 #ifdef HAVE_UTMPX
2784
2785 #define QGA_MICRO_SECOND_TO_SECOND 1000000
2786
2787 static double ga_get_login_time(struct utmpx *user_info)
2788 {
2789     double seconds = (double)user_info->ut_tv.tv_sec;
2790     double useconds = (double)user_info->ut_tv.tv_usec;
2791     useconds /= QGA_MICRO_SECOND_TO_SECOND;
2792     return seconds + useconds;
2793 }
2794
2795 GuestUserList *qmp_guest_get_users(Error **err)
2796 {
2797     GHashTable *cache = NULL;
2798     GuestUserList *head = NULL, *cur_item = NULL;
2799     struct utmpx *user_info = NULL;
2800     gpointer value = NULL;
2801     GuestUser *user = NULL;
2802     GuestUserList *item = NULL;
2803     double login_time = 0;
2804
2805     cache = g_hash_table_new(g_str_hash, g_str_equal);
2806     setutxent();
2807
2808     for (;;) {
2809         user_info = getutxent();
2810         if (user_info == NULL) {
2811             break;
2812         } else if (user_info->ut_type != USER_PROCESS) {
2813             continue;
2814         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
2815             value = g_hash_table_lookup(cache, user_info->ut_user);
2816             user = (GuestUser *)value;
2817             login_time = ga_get_login_time(user_info);
2818             /* We're ensuring the earliest login time to be sent */
2819             if (login_time < user->login_time) {
2820                 user->login_time = login_time;
2821             }
2822             continue;
2823         }
2824
2825         item = g_new0(GuestUserList, 1);
2826         item->value = g_new0(GuestUser, 1);
2827         item->value->user = g_strdup(user_info->ut_user);
2828         item->value->login_time = ga_get_login_time(user_info);
2829
2830         g_hash_table_insert(cache, item->value->user, item->value);
2831
2832         if (!cur_item) {
2833             head = cur_item = item;
2834         } else {
2835             cur_item->next = item;
2836             cur_item = item;
2837         }
2838     }
2839     endutxent();
2840     g_hash_table_destroy(cache);
2841     return head;
2842 }
2843
2844 #else
2845
2846 GuestUserList *qmp_guest_get_users(Error **errp)
2847 {
2848     error_setg(errp, QERR_UNSUPPORTED);
2849     return NULL;
2850 }
2851
2852 #endif
2853
2854 /* Replace escaped special characters with theire real values. The replacement
2855  * is done in place -- returned value is in the original string.
2856  */
2857 static void ga_osrelease_replace_special(gchar *value)
2858 {
2859     gchar *p, *p2, quote;
2860
2861     /* Trim the string at first space or semicolon if it is not enclosed in
2862      * single or double quotes. */
2863     if ((value[0] != '"') || (value[0] == '\'')) {
2864         p = strchr(value, ' ');
2865         if (p != NULL) {
2866             *p = 0;
2867         }
2868         p = strchr(value, ';');
2869         if (p != NULL) {
2870             *p = 0;
2871         }
2872         return;
2873     }
2874
2875     quote = value[0];
2876     p2 = value;
2877     p = value + 1;
2878     while (*p != 0) {
2879         if (*p == '\\') {
2880             p++;
2881             switch (*p) {
2882             case '$':
2883             case '\'':
2884             case '"':
2885             case '\\':
2886             case '`':
2887                 break;
2888             default:
2889                 /* Keep literal backslash followed by whatever is there */
2890                 p--;
2891                 break;
2892             }
2893         } else if (*p == quote) {
2894             *p2 = 0;
2895             break;
2896         }
2897         *(p2++) = *(p++);
2898     }
2899 }
2900
2901 static GKeyFile *ga_parse_osrelease(const char *fname)
2902 {
2903     gchar *content = NULL;
2904     gchar *content2 = NULL;
2905     GError *err = NULL;
2906     GKeyFile *keys = g_key_file_new();
2907     const char *group = "[os-release]\n";
2908
2909     if (!g_file_get_contents(fname, &content, NULL, &err)) {
2910         slog("failed to read '%s', error: %s", fname, err->message);
2911         goto fail;
2912     }
2913
2914     if (!g_utf8_validate(content, -1, NULL)) {
2915         slog("file is not utf-8 encoded: %s", fname);
2916         goto fail;
2917     }
2918     content2 = g_strdup_printf("%s%s", group, content);
2919
2920     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
2921                                    &err)) {
2922         slog("failed to parse file '%s', error: %s", fname, err->message);
2923         goto fail;
2924     }
2925
2926     g_free(content);
2927     g_free(content2);
2928     return keys;
2929
2930 fail:
2931     g_error_free(err);
2932     g_free(content);
2933     g_free(content2);
2934     g_key_file_free(keys);
2935     return NULL;
2936 }
2937
2938 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
2939 {
2940     GuestOSInfo *info = NULL;
2941     struct utsname kinfo;
2942     GKeyFile *osrelease = NULL;
2943     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
2944
2945     info = g_new0(GuestOSInfo, 1);
2946
2947     if (uname(&kinfo) != 0) {
2948         error_setg_errno(errp, errno, "uname failed");
2949     } else {
2950         info->has_kernel_version = true;
2951         info->kernel_version = g_strdup(kinfo.version);
2952         info->has_kernel_release = true;
2953         info->kernel_release = g_strdup(kinfo.release);
2954         info->has_machine = true;
2955         info->machine = g_strdup(kinfo.machine);
2956     }
2957
2958     if (qga_os_release != NULL) {
2959         osrelease = ga_parse_osrelease(qga_os_release);
2960     } else {
2961         osrelease = ga_parse_osrelease("/etc/os-release");
2962         if (osrelease == NULL) {
2963             osrelease = ga_parse_osrelease("/usr/lib/os-release");
2964         }
2965     }
2966
2967     if (osrelease != NULL) {
2968         char *value;
2969
2970 #define GET_FIELD(field, osfield) do { \
2971     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
2972     if (value != NULL) { \
2973         ga_osrelease_replace_special(value); \
2974         info->has_ ## field = true; \
2975         info->field = value; \
2976     } \
2977 } while (0)
2978         GET_FIELD(id, "ID");
2979         GET_FIELD(name, "NAME");
2980         GET_FIELD(pretty_name, "PRETTY_NAME");
2981         GET_FIELD(version, "VERSION");
2982         GET_FIELD(version_id, "VERSION_ID");
2983         GET_FIELD(variant, "VARIANT");
2984         GET_FIELD(variant_id, "VARIANT_ID");
2985 #undef GET_FIELD
2986
2987         g_key_file_free(osrelease);
2988     }
2989
2990     return info;
2991 }