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