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 <glib.h>
  15 #include <sys/types.h>
  16 #include <sys/ioctl.h>
  17 #include <sys/wait.h>
  18 #include <unistd.h>
  19 #include <errno.h>
  20 #include <fcntl.h>
  21 #include <dirent.h>
  22 #include <stdio.h>
  23 #include <string.h>
  24 #include <sys/stat.h>
  25 #include <inttypes.h>
  26 #include "qga/guest-agent-core.h"
  27 #include "qga-qmp-commands.h"
  28 #include "qapi/qmp/qerror.h"
  29 #include "qemu/queue.h"
  30 #include "qemu/host-utils.h"
  31
  32 #ifndef CONFIG_HAS_ENVIRON
  33 #ifdef __APPLE__
  34 #include <crt_externs.h>
  35 #define environ (*_NSGetEnviron())
  36 #else
  37 extern char **environ;
  38 #endif
  39 #endif
  40
  41 #if defined(__linux__)
  42 #include <mntent.h>
  43 #include <linux/fs.h>
  44 #include <ifaddrs.h>
  45 #include <arpa/inet.h>
  46 #include <sys/socket.h>
  47 #include <net/if.h>
  48
  49 #ifdef FIFREEZE
  50 #define CONFIG_FSFREEZE
  51 #endif
  52 #ifdef FITRIM
  53 #define CONFIG_FSTRIM
  54 #endif
  55 #endif
  56
  57 static void ga_wait_child(pid_t pid, int *status, Error **errp)
  58 {
  59     pid_t rpid;
  60
  61     *status = 0;
  62
  63     do {
  64         rpid = waitpid(pid, status, 0);
  65     } while (rpid == -1 && errno == EINTR);
  66
  67     if (rpid == -1) {
  68         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
  69                          pid);
  70         return;
  71     }
  72
  73     g_assert(rpid == pid);
  74 }
  75
  76 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
  77 {
  78     const char *shutdown_flag;
  79     Error *local_err = NULL;
  80     pid_t pid;
  81     int status;
  82
  83     slog("guest-shutdown called, mode: %s", mode);
  84     if (!has_mode || strcmp(mode, "powerdown") == 0) {
  85         shutdown_flag = "-P";
  86     } else if (strcmp(mode, "halt") == 0) {
  87         shutdown_flag = "-H";
  88     } else if (strcmp(mode, "reboot") == 0) {
  89         shutdown_flag = "-r";
  90     } else {
  91         error_setg(errp,
  92                    "mode is invalid (valid values are: halt|powerdown|reboot");
  93         return;
  94     }
  95
  96     pid = fork();
  97     if (pid == 0) {
  98         /* child, start the shutdown */
  99         setsid();
 100         reopen_fd_to_null(0);
 101         reopen_fd_to_null(1);
 102         reopen_fd_to_null(2);
 103
 104         execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
 105                "hypervisor initiated shutdown", (char*)NULL, environ);
 106         _exit(EXIT_FAILURE);
 107     } else if (pid < 0) {
 108         error_setg_errno(errp, errno, "failed to create child process");
 109         return;
 110     }
 111
 112     ga_wait_child(pid, &status, &local_err);
 113     if (local_err) {
 114         error_propagate(errp, local_err);
 115         return;
 116     }
 117
 118     if (!WIFEXITED(status)) {
 119         error_setg(errp, "child process has terminated abnormally");
 120         return;
 121     }
 122
 123     if (WEXITSTATUS(status)) {
 124         error_setg(errp, "child process has failed to shutdown");
 125         return;
 126     }
 127
 128     /* succeeded */
 129 }
 130
 131 int64_t qmp_guest_get_time(Error **errp)
 132 {
 133    int ret;
 134    qemu_timeval tq;
 135    int64_t time_ns;
 136
 137    ret = qemu_gettimeofday(&tq);
 138    if (ret < 0) {
 139        error_setg_errno(errp, errno, "Failed to get time");
 140        return -1;
 141    }
 142
 143    time_ns = tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
 144    return time_ns;
 145 }
 146
 147 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
 148 {
 149     int ret;
 150     int status;
 151     pid_t pid;
 152     Error *local_err = NULL;
 153     struct timeval tv;
 154
 155     /* If user has passed a time, validate and set it. */
 156     if (has_time) {
 157         /* year-2038 will overflow in case time_t is 32bit */
 158         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
 159             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
 160             return;
 161         }
 162
 163         tv.tv_sec = time_ns / 1000000000;
 164         tv.tv_usec = (time_ns % 1000000000) / 1000;
 165
 166         ret = settimeofday(&tv, NULL);
 167         if (ret < 0) {
 168             error_setg_errno(errp, errno, "Failed to set time to guest");
 169             return;
 170         }
 171     }
 172
 173     /* Now, if user has passed a time to set and the system time is set, we
 174      * just need to synchronize the hardware clock. However, if no time was
 175      * passed, user is requesting the opposite: set the system time from the
 176      * hardware clock (RTC). */
 177     pid = fork();
 178     if (pid == 0) {
 179         setsid();
 180         reopen_fd_to_null(0);
 181         reopen_fd_to_null(1);
 182         reopen_fd_to_null(2);
 183
 184         /* Use '/sbin/hwclock -w' to set RTC from the system time,
 185          * or '/sbin/hwclock -s' to set the system time from RTC. */
 186         execle("/sbin/hwclock", "hwclock", has_time ? "-w" : "-s",
 187                NULL, environ);
 188         _exit(EXIT_FAILURE);
 189     } else if (pid < 0) {
 190         error_setg_errno(errp, errno, "failed to create child process");
 191         return;
 192     }
 193
 194     ga_wait_child(pid, &status, &local_err);
 195     if (local_err) {
 196         error_propagate(errp, local_err);
 197         return;
 198     }
 199
 200     if (!WIFEXITED(status)) {
 201         error_setg(errp, "child process has terminated abnormally");
 202         return;
 203     }
 204
 205     if (WEXITSTATUS(status)) {
 206         error_setg(errp, "hwclock failed to set hardware clock to system time");
 207         return;
 208     }
 209 }
 210
 211 typedef struct GuestFileHandle {
 212     uint64_t id;
 213     FILE *fh;
 214     QTAILQ_ENTRY(GuestFileHandle) next;
 215 } GuestFileHandle;
 216
 217 static struct {
 218     QTAILQ_HEAD(, GuestFileHandle) filehandles;
 219 } guest_file_state;
 220
 221 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
 222 {
 223     GuestFileHandle *gfh;
 224     int64_t handle;
 225
 226     handle = ga_get_fd_handle(ga_state, errp);
 227     if (handle < 0) {
 228         return -1;
 229     }
 230
 231     gfh = g_malloc0(sizeof(GuestFileHandle));
 232     gfh->id = handle;
 233     gfh->fh = fh;
 234     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
 235
 236     return handle;
 237 }
 238
 239 static GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
 240 {
 241     GuestFileHandle *gfh;
 242
 243     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
 244     {
 245         if (gfh->id == id) {
 246             return gfh;
 247         }
 248     }
 249
 250     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
 251     return NULL;
 252 }
 253
 254 typedef const char * const ccpc;
 255
 256 #ifndef O_BINARY
 257 #define O_BINARY 0
 258 #endif
 259
 260 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
 261 static const struct {
 262     ccpc *forms;
 263     int oflag_base;
 264 } guest_file_open_modes[] = {
 265     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
 266     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
 267     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
 268     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
 269     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
 270     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
 271     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
 272     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
 273     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
 274     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
 275     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
 276     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
 277 };
 278
 279 static int
 280 find_open_flag(const char *mode_str, Error **errp)
 281 {
 282     unsigned mode;
 283
 284     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
 285         ccpc *form;
 286
 287         form = guest_file_open_modes[mode].forms;
 288         while (*form != NULL && strcmp(*form, mode_str) != 0) {
 289             ++form;
 290         }
 291         if (*form != NULL) {
 292             break;
 293         }
 294     }
 295
 296     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
 297         error_setg(errp, "invalid file open mode '%s'", mode_str);
 298         return -1;
 299     }
 300     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
 301 }
 302
 303 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
 304                                S_IRGRP | S_IWGRP | \
 305                                S_IROTH | S_IWOTH)
 306
 307 static FILE *
 308 safe_open_or_create(const char *path, const char *mode, Error **errp)
 309 {
 310     Error *local_err = NULL;
 311     int oflag;
 312
 313     oflag = find_open_flag(mode, &local_err);
 314     if (local_err == NULL) {
 315         int fd;
 316
 317         /* If the caller wants / allows creation of a new file, we implement it
 318          * with a two step process: open() + (open() / fchmod()).
 319          *
 320          * First we insist on creating the file exclusively as a new file. If
 321          * that succeeds, we're free to set any file-mode bits on it. (The
 322          * motivation is that we want to set those file-mode bits independently
 323          * of the current umask.)
 324          *
 325          * If the exclusive creation fails because the file already exists
 326          * (EEXIST is not possible for any other reason), we just attempt to
 327          * open the file, but in this case we won't be allowed to change the
 328          * file-mode bits on the preexistent file.
 329          *
 330          * The pathname should never disappear between the two open()s in
 331          * practice. If it happens, then someone very likely tried to race us.
 332          * In this case just go ahead and report the ENOENT from the second
 333          * open() to the caller.
 334          *
 335          * If the caller wants to open a preexistent file, then the first
 336          * open() is decisive and its third argument is ignored, and the second
 337          * open() and the fchmod() are never called.
 338          */
 339         fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
 340         if (fd == -1 && errno == EEXIST) {
 341             oflag &= ~(unsigned)O_CREAT;
 342             fd = open(path, oflag);
 343         }
 344
 345         if (fd == -1) {
 346             error_setg_errno(&local_err, errno, "failed to open file '%s' "
 347                              "(mode: '%s')", path, mode);
 348         } else {
 349             qemu_set_cloexec(fd);
 350
 351             if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
 352                 error_setg_errno(&local_err, errno, "failed to set permission "
 353                                  "0%03o on new file '%s' (mode: '%s')",
 354                                  (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
 355             } else {
 356                 FILE *f;
 357
 358                 f = fdopen(fd, mode);
 359                 if (f == NULL) {
 360                     error_setg_errno(&local_err, errno, "failed to associate "
 361                                      "stdio stream with file descriptor %d, "
 362                                      "file '%s' (mode: '%s')", fd, path, mode);
 363                 } else {
 364                     return f;
 365                 }
 366             }
 367
 368             close(fd);
 369             if (oflag & O_CREAT) {
 370                 unlink(path);
 371             }
 372         }
 373     }
 374
 375     error_propagate(errp, local_err);
 376     return NULL;
 377 }
 378
 379 static int guest_file_toggle_flags(int fd, int flags, bool set, Error **err)
 380 {
 381     int ret, old_flags;
 382
 383     old_flags = fcntl(fd, F_GETFL);
 384     if (old_flags == -1) {
 385         error_set_errno(err, errno, QERR_QGA_COMMAND_FAILED,
 386                         "failed to fetch filehandle flags");
 387         return -1;
 388     }
 389
 390     ret = fcntl(fd, F_SETFL, set ? (old_flags | flags) : (old_flags & ~flags));
 391     if (ret == -1) {
 392         error_set_errno(err, errno, QERR_QGA_COMMAND_FAILED,
 393                         "failed to set filehandle flags");
 394         return -1;
 395     }
 396
 397     return ret;
 398 }
 399
 400 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
 401                             Error **errp)
 402 {
 403     FILE *fh;
 404     Error *local_err = NULL;
 405     int64_t handle;
 406
 407     if (!has_mode) {
 408         mode = "r";
 409     }
 410     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
 411     fh = safe_open_or_create(path, mode, &local_err);
 412     if (local_err != NULL) {
 413         error_propagate(errp, local_err);
 414         return -1;
 415     }
 416
 417     /* set fd non-blocking to avoid common use cases (like reading from a
 418      * named pipe) from hanging the agent
 419      */
 420     if (guest_file_toggle_flags(fileno(fh), O_NONBLOCK, true, errp) < 0) {
 421         fclose(fh);
 422         return -1;
 423     }
 424
 425     handle = guest_file_handle_add(fh, errp);
 426     if (handle < 0) {
 427         fclose(fh);
 428         return -1;
 429     }
 430
 431     slog("guest-file-open, handle: %" PRId64, handle);
 432     return handle;
 433 }
 434
 435 void qmp_guest_file_close(int64_t handle, Error **errp)
 436 {
 437     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 438     int ret;
 439
 440     slog("guest-file-close called, handle: %" PRId64, handle);
 441     if (!gfh) {
 442         return;
 443     }
 444
 445     ret = fclose(gfh->fh);
 446     if (ret == EOF) {
 447         error_setg_errno(errp, errno, "failed to close handle");
 448         return;
 449     }
 450
 451     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
 452     g_free(gfh);
 453 }
 454
 455 struct GuestFileRead *qmp_guest_file_read(int64_t handle, bool has_count,
 456                                           int64_t count, Error **errp)
 457 {
 458     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 459     GuestFileRead *read_data = NULL;
 460     guchar *buf;
 461     FILE *fh;
 462     size_t read_count;
 463
 464     if (!gfh) {
 465         return NULL;
 466     }
 467
 468     if (!has_count) {
 469         count = QGA_READ_COUNT_DEFAULT;
 470     } else if (count < 0) {
 471         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 472                    count);
 473         return NULL;
 474     }
 475
 476     fh = gfh->fh;
 477     buf = g_malloc0(count+1);
 478     read_count = fread(buf, 1, count, fh);
 479     if (ferror(fh)) {
 480         error_setg_errno(errp, errno, "failed to read file");
 481         slog("guest-file-read failed, handle: %" PRId64, handle);
 482     } else {
 483         buf[read_count] = 0;
 484         read_data = g_malloc0(sizeof(GuestFileRead));
 485         read_data->count = read_count;
 486         read_data->eof = feof(fh);
 487         if (read_count) {
 488             read_data->buf_b64 = g_base64_encode(buf, read_count);
 489         }
 490     }
 491     g_free(buf);
 492     clearerr(fh);
 493
 494     return read_data;
 495 }
 496
 497 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
 498                                      bool has_count, int64_t count,
 499                                      Error **errp)
 500 {
 501     GuestFileWrite *write_data = NULL;
 502     guchar *buf;
 503     gsize buf_len;
 504     int write_count;
 505     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 506     FILE *fh;
 507
 508     if (!gfh) {
 509         return NULL;
 510     }
 511
 512     fh = gfh->fh;
 513     buf = g_base64_decode(buf_b64, &buf_len);
 514
 515     if (!has_count) {
 516         count = buf_len;
 517     } else if (count < 0 || count > buf_len) {
 518         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 519                    count);
 520         g_free(buf);
 521         return NULL;
 522     }
 523
 524     write_count = fwrite(buf, 1, count, fh);
 525     if (ferror(fh)) {
 526         error_setg_errno(errp, errno, "failed to write to file");
 527         slog("guest-file-write failed, handle: %" PRId64, handle);
 528     } else {
 529         write_data = g_malloc0(sizeof(GuestFileWrite));
 530         write_data->count = write_count;
 531         write_data->eof = feof(fh);
 532     }
 533     g_free(buf);
 534     clearerr(fh);
 535
 536     return write_data;
 537 }
 538
 539 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
 540                                           int64_t whence, Error **errp)
 541 {
 542     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 543     GuestFileSeek *seek_data = NULL;
 544     FILE *fh;
 545     int ret;
 546
 547     if (!gfh) {
 548         return NULL;
 549     }
 550
 551     fh = gfh->fh;
 552     ret = fseek(fh, offset, whence);
 553     if (ret == -1) {
 554         error_setg_errno(errp, errno, "failed to seek file");
 555     } else {
 556         seek_data = g_new0(GuestFileSeek, 1);
 557         seek_data->position = ftell(fh);
 558         seek_data->eof = feof(fh);
 559     }
 560     clearerr(fh);
 561
 562     return seek_data;
 563 }
 564
 565 void qmp_guest_file_flush(int64_t handle, Error **errp)
 566 {
 567     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 568     FILE *fh;
 569     int ret;
 570
 571     if (!gfh) {
 572         return;
 573     }
 574
 575     fh = gfh->fh;
 576     ret = fflush(fh);
 577     if (ret == EOF) {
 578         error_setg_errno(errp, errno, "failed to flush file");
 579     }
 580 }
 581
 582 static void guest_file_init(void)
 583 {
 584     QTAILQ_INIT(&guest_file_state.filehandles);
 585 }
 586
 587 /* linux-specific implementations. avoid this if at all possible. */
 588 #if defined(__linux__)
 589
 590 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
 591 typedef struct FsMount {
 592     char *dirname;
 593     char *devtype;
 594     unsigned int devmajor, devminor;
 595     QTAILQ_ENTRY(FsMount) next;
 596 } FsMount;
 597
 598 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
 599
 600 static void free_fs_mount_list(FsMountList *mounts)
 601 {
 602      FsMount *mount, *temp;
 603
 604      if (!mounts) {
 605          return;
 606      }
 607
 608      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
 609          QTAILQ_REMOVE(mounts, mount, next);
 610          g_free(mount->dirname);
 611          g_free(mount->devtype);
 612          g_free(mount);
 613      }
 614 }
 615
 616 static int dev_major_minor(const char *devpath,
 617                            unsigned int *devmajor, unsigned int *devminor)
 618 {
 619     struct stat st;
 620
 621     *devmajor = 0;
 622     *devminor = 0;
 623
 624     if (stat(devpath, &st) < 0) {
 625         slog("failed to stat device file '%s': %s", devpath, strerror(errno));
 626         return -1;
 627     }
 628     if (S_ISDIR(st.st_mode)) {
 629         /* It is bind mount */
 630         return -2;
 631     }
 632     if (S_ISBLK(st.st_mode)) {
 633         *devmajor = major(st.st_rdev);
 634         *devminor = minor(st.st_rdev);
 635         return 0;
 636     }
 637     return -1;
 638 }
 639
 640 /*
 641  * Walk the mount table and build a list of local file systems
 642  */
 643 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
 644 {
 645     struct mntent *ment;
 646     FsMount *mount;
 647     char const *mtab = "/proc/self/mounts";
 648     FILE *fp;
 649     unsigned int devmajor, devminor;
 650
 651     fp = setmntent(mtab, "r");
 652     if (!fp) {
 653         error_setg(errp, "failed to open mtab file: '%s'", mtab);
 654         return;
 655     }
 656
 657     while ((ment = getmntent(fp))) {
 658         /*
 659          * An entry which device name doesn't start with a '/' is
 660          * either a dummy file system or a network file system.
 661          * Add special handling for smbfs and cifs as is done by
 662          * coreutils as well.
 663          */
 664         if ((ment->mnt_fsname[0] != '/') ||
 665             (strcmp(ment->mnt_type, "smbfs") == 0) ||
 666             (strcmp(ment->mnt_type, "cifs") == 0)) {
 667             continue;
 668         }
 669         if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
 670             /* Skip bind mounts */
 671             continue;
 672         }
 673
 674         mount = g_malloc0(sizeof(FsMount));
 675         mount->dirname = g_strdup(ment->mnt_dir);
 676         mount->devtype = g_strdup(ment->mnt_type);
 677         mount->devmajor = devmajor;
 678         mount->devminor = devminor;
 679
 680         QTAILQ_INSERT_TAIL(mounts, mount, next);
 681     }
 682
 683     endmntent(fp);
 684 }
 685
 686 static void decode_mntname(char *name, int len)
 687 {
 688     int i, j = 0;
 689     for (i = 0; i <= len; i++) {
 690         if (name[i] != '\\') {
 691             name[j++] = name[i];
 692         } else if (name[i + 1] == '\\') {
 693             name[j++] = '\\';
 694             i++;
 695         } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
 696                    name[i + 2] >= '0' && name[i + 2] <= '7' &&
 697                    name[i + 3] >= '0' && name[i + 3] <= '7') {
 698             name[j++] = (name[i + 1] - '0') * 64 +
 699                         (name[i + 2] - '0') * 8 +
 700                         (name[i + 3] - '0');
 701             i += 3;
 702         } else {
 703             name[j++] = name[i];
 704         }
 705     }
 706 }
 707
 708 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
 709 {
 710     FsMount *mount;
 711     char const *mountinfo = "/proc/self/mountinfo";
 712     FILE *fp;
 713     char *line = NULL, *dash;
 714     size_t n;
 715     char check;
 716     unsigned int devmajor, devminor;
 717     int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
 718
 719     fp = fopen(mountinfo, "r");
 720     if (!fp) {
 721         build_fs_mount_list_from_mtab(mounts, errp);
 722         return;
 723     }
 724
 725     while (getline(&line, &n, fp) != -1) {
 726         ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
 727                      &devmajor, &devminor, &dir_s, &dir_e, &check);
 728         if (ret < 3) {
 729             continue;
 730         }
 731         dash = strstr(line + dir_e, " - ");
 732         if (!dash) {
 733             continue;
 734         }
 735         ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
 736                      &type_s, &type_e, &dev_s, &dev_e, &check);
 737         if (ret < 1) {
 738             continue;
 739         }
 740         line[dir_e] = 0;
 741         dash[type_e] = 0;
 742         dash[dev_e] = 0;
 743         decode_mntname(line + dir_s, dir_e - dir_s);
 744         decode_mntname(dash + dev_s, dev_e - dev_s);
 745         if (devmajor == 0) {
 746             /* btrfs reports major number = 0 */
 747             if (strcmp("btrfs", dash + type_s) != 0 ||
 748                 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
 749                 continue;
 750             }
 751         }
 752
 753         mount = g_malloc0(sizeof(FsMount));
 754         mount->dirname = g_strdup(line + dir_s);
 755         mount->devtype = g_strdup(dash + type_s);
 756         mount->devmajor = devmajor;
 757         mount->devminor = devminor;
 758
 759         QTAILQ_INSERT_TAIL(mounts, mount, next);
 760     }
 761     free(line);
 762
 763     fclose(fp);
 764 }
 765 #endif
 766
 767 #if defined(CONFIG_FSFREEZE)
 768
 769 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
 770 {
 771     char *path;
 772     char *dpath;
 773     char *driver = NULL;
 774     char buf[PATH_MAX];
 775     ssize_t len;
 776
 777     path = g_strndup(syspath, pathlen);
 778     dpath = g_strdup_printf("%s/driver", path);
 779     len = readlink(dpath, buf, sizeof(buf) - 1);
 780     if (len != -1) {
 781         buf[len] = 0;
 782         driver = g_strdup(basename(buf));
 783     }
 784     g_free(dpath);
 785     g_free(path);
 786     return driver;
 787 }
 788
 789 static int compare_uint(const void *_a, const void *_b)
 790 {
 791     unsigned int a = *(unsigned int *)_a;
 792     unsigned int b = *(unsigned int *)_b;
 793
 794     return a < b ? -1 : a > b ? 1 : 0;
 795 }
 796
 797 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
 798 static int build_hosts(char const *syspath, char const *host, bool ata,
 799                        unsigned int *hosts, int hosts_max, Error **errp)
 800 {
 801     char *path;
 802     DIR *dir;
 803     struct dirent *entry;
 804     int i = 0;
 805
 806     path = g_strndup(syspath, host - syspath);
 807     dir = opendir(path);
 808     if (!dir) {
 809         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
 810         g_free(path);
 811         return -1;
 812     }
 813
 814     while (i < hosts_max) {
 815         entry = readdir(dir);
 816         if (!entry) {
 817             break;
 818         }
 819         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
 820             ++i;
 821         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
 822             ++i;
 823         }
 824     }
 825
 826     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
 827
 828     g_free(path);
 829     closedir(dir);
 830     return i;
 831 }
 832
 833 /* Store disk device info specified by @sysfs into @fs */
 834 static void build_guest_fsinfo_for_real_device(char const *syspath,
 835                                                GuestFilesystemInfo *fs,
 836                                                Error **errp)
 837 {
 838     unsigned int pci[4], host, hosts[8], tgt[3];
 839     int i, nhosts = 0, pcilen;
 840     GuestDiskAddress *disk;
 841     GuestPCIAddress *pciaddr;
 842     GuestDiskAddressList *list = NULL;
 843     bool has_ata = false, has_host = false, has_tgt = false;
 844     char *p, *q, *driver = NULL;
 845
 846     p = strstr(syspath, "/devices/pci");
 847     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
 848                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
 849         g_debug("only pci device is supported: sysfs path \"%s\"", syspath);
 850         return;
 851     }
 852
 853     driver = get_pci_driver(syspath, (p + 12 + pcilen) - syspath, errp);
 854     if (!driver) {
 855         goto cleanup;
 856     }
 857
 858     p = strstr(syspath, "/target");
 859     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
 860                     tgt, tgt + 1, tgt + 2) == 3) {
 861         has_tgt = true;
 862     }
 863
 864     p = strstr(syspath, "/ata");
 865     if (p) {
 866         q = p + 4;
 867         has_ata = true;
 868     } else {
 869         p = strstr(syspath, "/host");
 870         q = p + 5;
 871     }
 872     if (p && sscanf(q, "%u", &host) == 1) {
 873         has_host = true;
 874         nhosts = build_hosts(syspath, p, has_ata, hosts,
 875                              sizeof(hosts) / sizeof(hosts[0]), errp);
 876         if (nhosts < 0) {
 877             goto cleanup;
 878         }
 879     }
 880
 881     pciaddr = g_malloc0(sizeof(*pciaddr));
 882     pciaddr->domain = pci[0];
 883     pciaddr->bus = pci[1];
 884     pciaddr->slot = pci[2];
 885     pciaddr->function = pci[3];
 886
 887     disk = g_malloc0(sizeof(*disk));
 888     disk->pci_controller = pciaddr;
 889
 890     list = g_malloc0(sizeof(*list));
 891     list->value = disk;
 892
 893     if (strcmp(driver, "ata_piix") == 0) {
 894         /* a host per ide bus, target*:0:<unit>:0 */
 895         if (!has_host || !has_tgt) {
 896             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 897             goto cleanup;
 898         }
 899         for (i = 0; i < nhosts; i++) {
 900             if (host == hosts[i]) {
 901                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
 902                 disk->bus = i;
 903                 disk->unit = tgt[1];
 904                 break;
 905             }
 906         }
 907         if (i >= nhosts) {
 908             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 909             goto cleanup;
 910         }
 911     } else if (strcmp(driver, "sym53c8xx") == 0) {
 912         /* scsi(LSI Logic): target*:0:<unit>:0 */
 913         if (!has_tgt) {
 914             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 915             goto cleanup;
 916         }
 917         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 918         disk->unit = tgt[1];
 919     } else if (strcmp(driver, "virtio-pci") == 0) {
 920         if (has_tgt) {
 921             /* virtio-scsi: target*:0:0:<unit> */
 922             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 923             disk->unit = tgt[2];
 924         } else {
 925             /* virtio-blk: 1 disk per 1 device */
 926             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
 927         }
 928     } else if (strcmp(driver, "ahci") == 0) {
 929         /* ahci: 1 host per 1 unit */
 930         if (!has_host || !has_tgt) {
 931             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 932             goto cleanup;
 933         }
 934         for (i = 0; i < nhosts; i++) {
 935             if (host == hosts[i]) {
 936                 disk->unit = i;
 937                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
 938                 break;
 939             }
 940         }
 941         if (i >= nhosts) {
 942             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 943             goto cleanup;
 944         }
 945     } else {
 946         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
 947         goto cleanup;
 948     }
 949
 950     list->next = fs->disk;
 951     fs->disk = list;
 952     g_free(driver);
 953     return;
 954
 955 cleanup:
 956     if (list) {
 957         qapi_free_GuestDiskAddressList(list);
 958     }
 959     g_free(driver);
 960 }
 961
 962 static void build_guest_fsinfo_for_device(char const *devpath,
 963                                           GuestFilesystemInfo *fs,
 964                                           Error **errp);
 965
 966 /* Store a list of slave devices of virtual volume specified by @syspath into
 967  * @fs */
 968 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
 969                                                   GuestFilesystemInfo *fs,
 970                                                   Error **errp)
 971 {
 972     DIR *dir;
 973     char *dirpath;
 974     struct dirent *entry;
 975
 976     dirpath = g_strdup_printf("%s/slaves", syspath);
 977     dir = opendir(dirpath);
 978     if (!dir) {
 979         error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
 980         g_free(dirpath);
 981         return;
 982     }
 983
 984     for (;;) {
 985         errno = 0;
 986         entry = readdir(dir);
 987         if (entry == NULL) {
 988             if (errno) {
 989                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
 990             }
 991             break;
 992         }
 993
 994         if (entry->d_type == DT_LNK) {
 995             char *path;
 996
 997             g_debug(" slave device '%s'", entry->d_name);
 998             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
 999             build_guest_fsinfo_for_device(path, fs, errp);
1000             g_free(path);
1001
1002             if (*errp) {
1003                 break;
1004             }
1005         }
1006     }
1007
1008     g_free(dirpath);
1009     closedir(dir);
1010 }
1011
1012 /* Dispatch to functions for virtual/real device */
1013 static void build_guest_fsinfo_for_device(char const *devpath,
1014                                           GuestFilesystemInfo *fs,
1015                                           Error **errp)
1016 {
1017     char *syspath = realpath(devpath, NULL);
1018
1019     if (!syspath) {
1020         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1021         return;
1022     }
1023
1024     if (!fs->name) {
1025         fs->name = g_strdup(basename(syspath));
1026     }
1027
1028     g_debug("  parse sysfs path '%s'", syspath);
1029
1030     if (strstr(syspath, "/devices/virtual/block/")) {
1031         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1032     } else {
1033         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1034     }
1035
1036     free(syspath);
1037 }
1038
1039 /* Return a list of the disk device(s)' info which @mount lies on */
1040 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1041                                                Error **errp)
1042 {
1043     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1044     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1045                                     mount->devmajor, mount->devminor);
1046
1047     fs->mountpoint = g_strdup(mount->dirname);
1048     fs->type = g_strdup(mount->devtype);
1049     build_guest_fsinfo_for_device(devpath, fs, errp);
1050
1051     g_free(devpath);
1052     return fs;
1053 }
1054
1055 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1056 {
1057     FsMountList mounts;
1058     struct FsMount *mount;
1059     GuestFilesystemInfoList *new, *ret = NULL;
1060     Error *local_err = NULL;
1061
1062     QTAILQ_INIT(&mounts);
1063     build_fs_mount_list(&mounts, &local_err);
1064     if (local_err) {
1065         error_propagate(errp, local_err);
1066         return NULL;
1067     }
1068
1069     QTAILQ_FOREACH(mount, &mounts, next) {
1070         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1071
1072         new = g_malloc0(sizeof(*ret));
1073         new->value = build_guest_fsinfo(mount, &local_err);
1074         new->next = ret;
1075         ret = new;
1076         if (local_err) {
1077             error_propagate(errp, local_err);
1078             qapi_free_GuestFilesystemInfoList(ret);
1079             ret = NULL;
1080             break;
1081         }
1082     }
1083
1084     free_fs_mount_list(&mounts);
1085     return ret;
1086 }
1087
1088
1089 typedef enum {
1090     FSFREEZE_HOOK_THAW = 0,
1091     FSFREEZE_HOOK_FREEZE,
1092 } FsfreezeHookArg;
1093
1094 static const char *fsfreeze_hook_arg_string[] = {
1095     "thaw",
1096     "freeze",
1097 };
1098
1099 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1100 {
1101     int status;
1102     pid_t pid;
1103     const char *hook;
1104     const char *arg_str = fsfreeze_hook_arg_string[arg];
1105     Error *local_err = NULL;
1106
1107     hook = ga_fsfreeze_hook(ga_state);
1108     if (!hook) {
1109         return;
1110     }
1111     if (access(hook, X_OK) != 0) {
1112         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1113         return;
1114     }
1115
1116     slog("executing fsfreeze hook with arg '%s'", arg_str);
1117     pid = fork();
1118     if (pid == 0) {
1119         setsid();
1120         reopen_fd_to_null(0);
1121         reopen_fd_to_null(1);
1122         reopen_fd_to_null(2);
1123
1124         execle(hook, hook, arg_str, NULL, environ);
1125         _exit(EXIT_FAILURE);
1126     } else if (pid < 0) {
1127         error_setg_errno(errp, errno, "failed to create child process");
1128         return;
1129     }
1130
1131     ga_wait_child(pid, &status, &local_err);
1132     if (local_err) {
1133         error_propagate(errp, local_err);
1134         return;
1135     }
1136
1137     if (!WIFEXITED(status)) {
1138         error_setg(errp, "fsfreeze hook has terminated abnormally");
1139         return;
1140     }
1141
1142     status = WEXITSTATUS(status);
1143     if (status) {
1144         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1145         return;
1146     }
1147 }
1148
1149 /*
1150  * Return status of freeze/thaw
1151  */
1152 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1153 {
1154     if (ga_is_frozen(ga_state)) {
1155         return GUEST_FSFREEZE_STATUS_FROZEN;
1156     }
1157
1158     return GUEST_FSFREEZE_STATUS_THAWED;
1159 }
1160
1161 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1162 {
1163     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1164 }
1165
1166 /*
1167  * Walk list of mounted file systems in the guest, and freeze the ones which
1168  * are real local file systems.
1169  */
1170 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1171                                        strList *mountpoints,
1172                                        Error **errp)
1173 {
1174     int ret = 0, i = 0;
1175     strList *list;
1176     FsMountList mounts;
1177     struct FsMount *mount;
1178     Error *local_err = NULL;
1179     int fd;
1180
1181     slog("guest-fsfreeze called");
1182
1183     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1184     if (local_err) {
1185         error_propagate(errp, local_err);
1186         return -1;
1187     }
1188
1189     QTAILQ_INIT(&mounts);
1190     build_fs_mount_list(&mounts, &local_err);
1191     if (local_err) {
1192         error_propagate(errp, local_err);
1193         return -1;
1194     }
1195
1196     /* cannot risk guest agent blocking itself on a write in this state */
1197     ga_set_frozen(ga_state);
1198
1199     QTAILQ_FOREACH_REVERSE(mount, &mounts, FsMountList, next) {
1200         /* To issue fsfreeze in the reverse order of mounts, check if the
1201          * mount is listed in the list here */
1202         if (has_mountpoints) {
1203             for (list = mountpoints; list; list = list->next) {
1204                 if (strcmp(list->value, mount->dirname) == 0) {
1205                     break;
1206                 }
1207             }
1208             if (!list) {
1209                 continue;
1210             }
1211         }
1212
1213         fd = qemu_open(mount->dirname, O_RDONLY);
1214         if (fd == -1) {
1215             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1216             goto error;
1217         }
1218
1219         /* we try to cull filesytems we know won't work in advance, but other
1220          * filesytems may not implement fsfreeze for less obvious reasons.
1221          * these will report EOPNOTSUPP. we simply ignore these when tallying
1222          * the number of frozen filesystems.
1223          *
1224          * any other error means a failure to freeze a filesystem we
1225          * expect to be freezable, so return an error in those cases
1226          * and return system to thawed state.
1227          */
1228         ret = ioctl(fd, FIFREEZE);
1229         if (ret == -1) {
1230             if (errno != EOPNOTSUPP) {
1231                 error_setg_errno(errp, errno, "failed to freeze %s",
1232                                  mount->dirname);
1233                 close(fd);
1234                 goto error;
1235             }
1236         } else {
1237             i++;
1238         }
1239         close(fd);
1240     }
1241
1242     free_fs_mount_list(&mounts);
1243     return i;
1244
1245 error:
1246     free_fs_mount_list(&mounts);
1247     qmp_guest_fsfreeze_thaw(NULL);
1248     return 0;
1249 }
1250
1251 /*
1252  * Walk list of frozen file systems in the guest, and thaw them.
1253  */
1254 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1255 {
1256     int ret;
1257     FsMountList mounts;
1258     FsMount *mount;
1259     int fd, i = 0, logged;
1260     Error *local_err = NULL;
1261
1262     QTAILQ_INIT(&mounts);
1263     build_fs_mount_list(&mounts, &local_err);
1264     if (local_err) {
1265         error_propagate(errp, local_err);
1266         return 0;
1267     }
1268
1269     QTAILQ_FOREACH(mount, &mounts, next) {
1270         logged = false;
1271         fd = qemu_open(mount->dirname, O_RDONLY);
1272         if (fd == -1) {
1273             continue;
1274         }
1275         /* we have no way of knowing whether a filesystem was actually unfrozen
1276          * as a result of a successful call to FITHAW, only that if an error
1277          * was returned the filesystem was *not* unfrozen by that particular
1278          * call.
1279          *
1280          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1281          * to unfreeze, continuing issuing FITHAW until an error is returned,
1282          * in which case either the filesystem is in an unfreezable state, or,
1283          * more likely, it was thawed previously (and remains so afterward).
1284          *
1285          * also, since the most recent successful call is the one that did
1286          * the actual unfreeze, we can use this to provide an accurate count
1287          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1288          * may * be useful for determining whether a filesystem was unfrozen
1289          * during the freeze/thaw phase by a process other than qemu-ga.
1290          */
1291         do {
1292             ret = ioctl(fd, FITHAW);
1293             if (ret == 0 && !logged) {
1294                 i++;
1295                 logged = true;
1296             }
1297         } while (ret == 0);
1298         close(fd);
1299     }
1300
1301     ga_unset_frozen(ga_state);
1302     free_fs_mount_list(&mounts);
1303
1304     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1305
1306     return i;
1307 }
1308
1309 static void guest_fsfreeze_cleanup(void)
1310 {
1311     Error *err = NULL;
1312
1313     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1314         qmp_guest_fsfreeze_thaw(&err);
1315         if (err) {
1316             slog("failed to clean up frozen filesystems: %s",
1317                  error_get_pretty(err));
1318             error_free(err);
1319         }
1320     }
1321 }
1322 #endif /* CONFIG_FSFREEZE */
1323
1324 #if defined(CONFIG_FSTRIM)
1325 /*
1326  * Walk list of mounted file systems in the guest, and trim them.
1327  */
1328 void qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1329 {
1330     int ret = 0;
1331     FsMountList mounts;
1332     struct FsMount *mount;
1333     int fd;
1334     Error *local_err = NULL;
1335     struct fstrim_range r = {
1336         .start = 0,
1337         .len = -1,
1338         .minlen = has_minimum ? minimum : 0,
1339     };
1340
1341     slog("guest-fstrim called");
1342
1343     QTAILQ_INIT(&mounts);
1344     build_fs_mount_list(&mounts, &local_err);
1345     if (local_err) {
1346         error_propagate(errp, local_err);
1347         return;
1348     }
1349
1350     QTAILQ_FOREACH(mount, &mounts, next) {
1351         fd = qemu_open(mount->dirname, O_RDONLY);
1352         if (fd == -1) {
1353             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1354             goto error;
1355         }
1356
1357         /* We try to cull filesytems we know won't work in advance, but other
1358          * filesytems may not implement fstrim for less obvious reasons.  These
1359          * will report EOPNOTSUPP; we simply ignore these errors.  Any other
1360          * error means an unexpected error, so return it in those cases.  In
1361          * some other cases ENOTTY will be reported (e.g. CD-ROMs).
1362          */
1363         ret = ioctl(fd, FITRIM, &r);
1364         if (ret == -1) {
1365             if (errno != ENOTTY && errno != EOPNOTSUPP) {
1366                 error_setg_errno(errp, errno, "failed to trim %s",
1367                                  mount->dirname);
1368                 close(fd);
1369                 goto error;
1370             }
1371         }
1372         close(fd);
1373     }
1374
1375 error:
1376     free_fs_mount_list(&mounts);
1377 }
1378 #endif /* CONFIG_FSTRIM */
1379
1380
1381 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1382 #define SUSPEND_SUPPORTED 0
1383 #define SUSPEND_NOT_SUPPORTED 1
1384
1385 static void bios_supports_mode(const char *pmutils_bin, const char *pmutils_arg,
1386                                const char *sysfile_str, Error **errp)
1387 {
1388     Error *local_err = NULL;
1389     char *pmutils_path;
1390     pid_t pid;
1391     int status;
1392
1393     pmutils_path = g_find_program_in_path(pmutils_bin);
1394
1395     pid = fork();
1396     if (!pid) {
1397         char buf[32]; /* hopefully big enough */
1398         ssize_t ret;
1399         int fd;
1400
1401         setsid();
1402         reopen_fd_to_null(0);
1403         reopen_fd_to_null(1);
1404         reopen_fd_to_null(2);
1405
1406         if (pmutils_path) {
1407             execle(pmutils_path, pmutils_bin, pmutils_arg, NULL, environ);
1408         }
1409
1410         /*
1411          * If we get here either pm-utils is not installed or execle() has
1412          * failed. Let's try the manual method if the caller wants it.
1413          */
1414
1415         if (!sysfile_str) {
1416             _exit(SUSPEND_NOT_SUPPORTED);
1417         }
1418
1419         fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1420         if (fd < 0) {
1421             _exit(SUSPEND_NOT_SUPPORTED);
1422         }
1423
1424         ret = read(fd, buf, sizeof(buf)-1);
1425         if (ret <= 0) {
1426             _exit(SUSPEND_NOT_SUPPORTED);
1427         }
1428         buf[ret] = '\0';
1429
1430         if (strstr(buf, sysfile_str)) {
1431             _exit(SUSPEND_SUPPORTED);
1432         }
1433
1434         _exit(SUSPEND_NOT_SUPPORTED);
1435     } else if (pid < 0) {
1436         error_setg_errno(errp, errno, "failed to create child process");
1437         goto out;
1438     }
1439
1440     ga_wait_child(pid, &status, &local_err);
1441     if (local_err) {
1442         error_propagate(errp, local_err);
1443         goto out;
1444     }
1445
1446     if (!WIFEXITED(status)) {
1447         error_setg(errp, "child process has terminated abnormally");
1448         goto out;
1449     }
1450
1451     switch (WEXITSTATUS(status)) {
1452     case SUSPEND_SUPPORTED:
1453         goto out;
1454     case SUSPEND_NOT_SUPPORTED:
1455         error_setg(errp,
1456                    "the requested suspend mode is not supported by the guest");
1457         goto out;
1458     default:
1459         error_setg(errp,
1460                    "the helper program '%s' returned an unexpected exit status"
1461                    " code (%d)", pmutils_path, WEXITSTATUS(status));
1462         goto out;
1463     }
1464
1465 out:
1466     g_free(pmutils_path);
1467 }
1468
1469 static void guest_suspend(const char *pmutils_bin, const char *sysfile_str,
1470                           Error **errp)
1471 {
1472     Error *local_err = NULL;
1473     char *pmutils_path;
1474     pid_t pid;
1475     int status;
1476
1477     pmutils_path = g_find_program_in_path(pmutils_bin);
1478
1479     pid = fork();
1480     if (pid == 0) {
1481         /* child */
1482         int fd;
1483
1484         setsid();
1485         reopen_fd_to_null(0);
1486         reopen_fd_to_null(1);
1487         reopen_fd_to_null(2);
1488
1489         if (pmutils_path) {
1490             execle(pmutils_path, pmutils_bin, NULL, environ);
1491         }
1492
1493         /*
1494          * If we get here either pm-utils is not installed or execle() has
1495          * failed. Let's try the manual method if the caller wants it.
1496          */
1497
1498         if (!sysfile_str) {
1499             _exit(EXIT_FAILURE);
1500         }
1501
1502         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1503         if (fd < 0) {
1504             _exit(EXIT_FAILURE);
1505         }
1506
1507         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1508             _exit(EXIT_FAILURE);
1509         }
1510
1511         _exit(EXIT_SUCCESS);
1512     } else if (pid < 0) {
1513         error_setg_errno(errp, errno, "failed to create child process");
1514         goto out;
1515     }
1516
1517     ga_wait_child(pid, &status, &local_err);
1518     if (local_err) {
1519         error_propagate(errp, local_err);
1520         goto out;
1521     }
1522
1523     if (!WIFEXITED(status)) {
1524         error_setg(errp, "child process has terminated abnormally");
1525         goto out;
1526     }
1527
1528     if (WEXITSTATUS(status)) {
1529         error_setg(errp, "child process has failed to suspend");
1530         goto out;
1531     }
1532
1533 out:
1534     g_free(pmutils_path);
1535 }
1536
1537 void qmp_guest_suspend_disk(Error **errp)
1538 {
1539     Error *local_err = NULL;
1540
1541     bios_supports_mode("pm-is-supported", "--hibernate", "disk", &local_err);
1542     if (local_err) {
1543         error_propagate(errp, local_err);
1544         return;
1545     }
1546
1547     guest_suspend("pm-hibernate", "disk", errp);
1548 }
1549
1550 void qmp_guest_suspend_ram(Error **errp)
1551 {
1552     Error *local_err = NULL;
1553
1554     bios_supports_mode("pm-is-supported", "--suspend", "mem", &local_err);
1555     if (local_err) {
1556         error_propagate(errp, local_err);
1557         return;
1558     }
1559
1560     guest_suspend("pm-suspend", "mem", errp);
1561 }
1562
1563 void qmp_guest_suspend_hybrid(Error **errp)
1564 {
1565     Error *local_err = NULL;
1566
1567     bios_supports_mode("pm-is-supported", "--suspend-hybrid", NULL,
1568                        &local_err);
1569     if (local_err) {
1570         error_propagate(errp, local_err);
1571         return;
1572     }
1573
1574     guest_suspend("pm-suspend-hybrid", NULL, errp);
1575 }
1576
1577 static GuestNetworkInterfaceList *
1578 guest_find_interface(GuestNetworkInterfaceList *head,
1579                      const char *name)
1580 {
1581     for (; head; head = head->next) {
1582         if (strcmp(head->value->name, name) == 0) {
1583             break;
1584         }
1585     }
1586
1587     return head;
1588 }
1589
1590 /*
1591  * Build information about guest interfaces
1592  */
1593 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1594 {
1595     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1596     struct ifaddrs *ifap, *ifa;
1597
1598     if (getifaddrs(&ifap) < 0) {
1599         error_setg_errno(errp, errno, "getifaddrs failed");
1600         goto error;
1601     }
1602
1603     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1604         GuestNetworkInterfaceList *info;
1605         GuestIpAddressList **address_list = NULL, *address_item = NULL;
1606         char addr4[INET_ADDRSTRLEN];
1607         char addr6[INET6_ADDRSTRLEN];
1608         int sock;
1609         struct ifreq ifr;
1610         unsigned char *mac_addr;
1611         void *p;
1612
1613         g_debug("Processing %s interface", ifa->ifa_name);
1614
1615         info = guest_find_interface(head, ifa->ifa_name);
1616
1617         if (!info) {
1618             info = g_malloc0(sizeof(*info));
1619             info->value = g_malloc0(sizeof(*info->value));
1620             info->value->name = g_strdup(ifa->ifa_name);
1621
1622             if (!cur_item) {
1623                 head = cur_item = info;
1624             } else {
1625                 cur_item->next = info;
1626                 cur_item = info;
1627             }
1628         }
1629
1630         if (!info->value->has_hardware_address &&
1631             ifa->ifa_flags & SIOCGIFHWADDR) {
1632             /* we haven't obtained HW address yet */
1633             sock = socket(PF_INET, SOCK_STREAM, 0);
1634             if (sock == -1) {
1635                 error_setg_errno(errp, errno, "failed to create socket");
1636                 goto error;
1637             }
1638
1639             memset(&ifr, 0, sizeof(ifr));
1640             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1641             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1642                 error_setg_errno(errp, errno,
1643                                  "failed to get MAC address of %s",
1644                                  ifa->ifa_name);
1645                 close(sock);
1646                 goto error;
1647             }
1648
1649             close(sock);
1650             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
1651
1652             info->value->hardware_address =
1653                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
1654                                 (int) mac_addr[0], (int) mac_addr[1],
1655                                 (int) mac_addr[2], (int) mac_addr[3],
1656                                 (int) mac_addr[4], (int) mac_addr[5]);
1657
1658             info->value->has_hardware_address = true;
1659         }
1660
1661         if (ifa->ifa_addr &&
1662             ifa->ifa_addr->sa_family == AF_INET) {
1663             /* interface with IPv4 address */
1664             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1665             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
1666                 error_setg_errno(errp, errno, "inet_ntop failed");
1667                 goto error;
1668             }
1669
1670             address_item = g_malloc0(sizeof(*address_item));
1671             address_item->value = g_malloc0(sizeof(*address_item->value));
1672             address_item->value->ip_address = g_strdup(addr4);
1673             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
1674
1675             if (ifa->ifa_netmask) {
1676                 /* Count the number of set bits in netmask.
1677                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1678                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
1679                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
1680             }
1681         } else if (ifa->ifa_addr &&
1682                    ifa->ifa_addr->sa_family == AF_INET6) {
1683             /* interface with IPv6 address */
1684             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
1685             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
1686                 error_setg_errno(errp, errno, "inet_ntop failed");
1687                 goto error;
1688             }
1689
1690             address_item = g_malloc0(sizeof(*address_item));
1691             address_item->value = g_malloc0(sizeof(*address_item->value));
1692             address_item->value->ip_address = g_strdup(addr6);
1693             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
1694
1695             if (ifa->ifa_netmask) {
1696                 /* Count the number of set bits in netmask.
1697                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1698                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
1699                 address_item->value->prefix =
1700                     ctpop32(((uint32_t *) p)[0]) +
1701                     ctpop32(((uint32_t *) p)[1]) +
1702                     ctpop32(((uint32_t *) p)[2]) +
1703                     ctpop32(((uint32_t *) p)[3]);
1704             }
1705         }
1706
1707         if (!address_item) {
1708             continue;
1709         }
1710
1711         address_list = &info->value->ip_addresses;
1712
1713         while (*address_list && (*address_list)->next) {
1714             address_list = &(*address_list)->next;
1715         }
1716
1717         if (!*address_list) {
1718             *address_list = address_item;
1719         } else {
1720             (*address_list)->next = address_item;
1721         }
1722
1723         info->value->has_ip_addresses = true;
1724
1725
1726     }
1727
1728     freeifaddrs(ifap);
1729     return head;
1730
1731 error:
1732     freeifaddrs(ifap);
1733     qapi_free_GuestNetworkInterfaceList(head);
1734     return NULL;
1735 }
1736
1737 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
1738
1739 static long sysconf_exact(int name, const char *name_str, Error **errp)
1740 {
1741     long ret;
1742
1743     errno = 0;
1744     ret = sysconf(name);
1745     if (ret == -1) {
1746         if (errno == 0) {
1747             error_setg(errp, "sysconf(%s): value indefinite", name_str);
1748         } else {
1749             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
1750         }
1751     }
1752     return ret;
1753 }
1754
1755 /* Transfer online/offline status between @vcpu and the guest system.
1756  *
1757  * On input either @errp or *@errp must be NULL.
1758  *
1759  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
1760  * - R: vcpu->logical_id
1761  * - W: vcpu->online
1762  * - W: vcpu->can_offline
1763  *
1764  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
1765  * - R: vcpu->logical_id
1766  * - R: vcpu->online
1767  *
1768  * Written members remain unmodified on error.
1769  */
1770 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
1771                           Error **errp)
1772 {
1773     char *dirpath;
1774     int dirfd;
1775
1776     dirpath = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
1777                               vcpu->logical_id);
1778     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
1779     if (dirfd == -1) {
1780         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
1781     } else {
1782         static const char fn[] = "online";
1783         int fd;
1784         int res;
1785
1786         fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
1787         if (fd == -1) {
1788             if (errno != ENOENT) {
1789                 error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
1790             } else if (sys2vcpu) {
1791                 vcpu->online = true;
1792                 vcpu->can_offline = false;
1793             } else if (!vcpu->online) {
1794                 error_setg(errp, "logical processor #%" PRId64 " can't be "
1795                            "offlined", vcpu->logical_id);
1796             } /* otherwise pretend successful re-onlining */
1797         } else {
1798             unsigned char status;
1799
1800             res = pread(fd, &status, 1, 0);
1801             if (res == -1) {
1802                 error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
1803             } else if (res == 0) {
1804                 error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
1805                            fn);
1806             } else if (sys2vcpu) {
1807                 vcpu->online = (status != '0');
1808                 vcpu->can_offline = true;
1809             } else if (vcpu->online != (status != '0')) {
1810                 status = '0' + vcpu->online;
1811                 if (pwrite(fd, &status, 1, 0) == -1) {
1812                     error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
1813                                      fn);
1814                 }
1815             } /* otherwise pretend successful re-(on|off)-lining */
1816
1817             res = close(fd);
1818             g_assert(res == 0);
1819         }
1820
1821         res = close(dirfd);
1822         g_assert(res == 0);
1823     }
1824
1825     g_free(dirpath);
1826 }
1827
1828 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
1829 {
1830     int64_t current;
1831     GuestLogicalProcessorList *head, **link;
1832     long sc_max;
1833     Error *local_err = NULL;
1834
1835     current = 0;
1836     head = NULL;
1837     link = &head;
1838     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
1839
1840     while (local_err == NULL && current < sc_max) {
1841         GuestLogicalProcessor *vcpu;
1842         GuestLogicalProcessorList *entry;
1843
1844         vcpu = g_malloc0(sizeof *vcpu);
1845         vcpu->logical_id = current++;
1846         vcpu->has_can_offline = true; /* lolspeak ftw */
1847         transfer_vcpu(vcpu, true, &local_err);
1848
1849         entry = g_malloc0(sizeof *entry);
1850         entry->value = vcpu;
1851
1852         *link = entry;
1853         link = &entry->next;
1854     }
1855
1856     if (local_err == NULL) {
1857         /* there's no guest with zero VCPUs */
1858         g_assert(head != NULL);
1859         return head;
1860     }
1861
1862     qapi_free_GuestLogicalProcessorList(head);
1863     error_propagate(errp, local_err);
1864     return NULL;
1865 }
1866
1867 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
1868 {
1869     int64_t processed;
1870     Error *local_err = NULL;
1871
1872     processed = 0;
1873     while (vcpus != NULL) {
1874         transfer_vcpu(vcpus->value, false, &local_err);
1875         if (local_err != NULL) {
1876             break;
1877         }
1878         ++processed;
1879         vcpus = vcpus->next;
1880     }
1881
1882     if (local_err != NULL) {
1883         if (processed == 0) {
1884             error_propagate(errp, local_err);
1885         } else {
1886             error_free(local_err);
1887         }
1888     }
1889
1890     return processed;
1891 }
1892
1893 void qmp_guest_set_user_password(const char *username,
1894                                  const char *password,
1895                                  bool crypted,
1896                                  Error **errp)
1897 {
1898     Error *local_err = NULL;
1899     char *passwd_path = NULL;
1900     pid_t pid;
1901     int status;
1902     int datafd[2] = { -1, -1 };
1903     char *rawpasswddata = NULL;
1904     size_t rawpasswdlen;
1905     char *chpasswddata = NULL;
1906     size_t chpasswdlen;
1907
1908     rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen);
1909     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
1910     rawpasswddata[rawpasswdlen] = '\0';
1911
1912     if (strchr(rawpasswddata, '\n')) {
1913         error_setg(errp, "forbidden characters in raw password");
1914         goto out;
1915     }
1916
1917     if (strchr(username, '\n') ||
1918         strchr(username, ':')) {
1919         error_setg(errp, "forbidden characters in username");
1920         goto out;
1921     }
1922
1923     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
1924     chpasswdlen = strlen(chpasswddata);
1925
1926     passwd_path = g_find_program_in_path("chpasswd");
1927
1928     if (!passwd_path) {
1929         error_setg(errp, "cannot find 'passwd' program in PATH");
1930         goto out;
1931     }
1932
1933     if (pipe(datafd) < 0) {
1934         error_setg(errp, "cannot create pipe FDs");
1935         goto out;
1936     }
1937
1938     pid = fork();
1939     if (pid == 0) {
1940         close(datafd[1]);
1941         /* child */
1942         setsid();
1943         dup2(datafd[0], 0);
1944         reopen_fd_to_null(1);
1945         reopen_fd_to_null(2);
1946
1947         if (crypted) {
1948             execle(passwd_path, "chpasswd", "-e", NULL, environ);
1949         } else {
1950             execle(passwd_path, "chpasswd", NULL, environ);
1951         }
1952         _exit(EXIT_FAILURE);
1953     } else if (pid < 0) {
1954         error_setg_errno(errp, errno, "failed to create child process");
1955         goto out;
1956     }
1957     close(datafd[0]);
1958     datafd[0] = -1;
1959
1960     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
1961         error_setg_errno(errp, errno, "cannot write new account password");
1962         goto out;
1963     }
1964     close(datafd[1]);
1965     datafd[1] = -1;
1966
1967     ga_wait_child(pid, &status, &local_err);
1968     if (local_err) {
1969         error_propagate(errp, local_err);
1970         goto out;
1971     }
1972
1973     if (!WIFEXITED(status)) {
1974         error_setg(errp, "child process has terminated abnormally");
1975         goto out;
1976     }
1977
1978     if (WEXITSTATUS(status)) {
1979         error_setg(errp, "child process has failed to set user password");
1980         goto out;
1981     }
1982
1983 out:
1984     g_free(chpasswddata);
1985     g_free(rawpasswddata);
1986     g_free(passwd_path);
1987     if (datafd[0] != -1) {
1988         close(datafd[0]);
1989     }
1990     if (datafd[1] != -1) {
1991         close(datafd[1]);
1992     }
1993 }
1994
1995 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
1996                                int size, Error **errp)
1997 {
1998     int fd;
1999     int res;
2000
2001     errno = 0;
2002     fd = openat(dirfd, pathname, O_RDONLY);
2003     if (fd == -1) {
2004         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2005         return;
2006     }
2007
2008     res = pread(fd, buf, size, 0);
2009     if (res == -1) {
2010         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2011     } else if (res == 0) {
2012         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2013     }
2014     close(fd);
2015 }
2016
2017 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2018                                 const char *buf, int size, Error **errp)
2019 {
2020     int fd;
2021
2022     errno = 0;
2023     fd = openat(dirfd, pathname, O_WRONLY);
2024     if (fd == -1) {
2025         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2026         return;
2027     }
2028
2029     if (pwrite(fd, buf, size, 0) == -1) {
2030         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2031     }
2032
2033     close(fd);
2034 }
2035
2036 /* Transfer online/offline status between @mem_blk and the guest system.
2037  *
2038  * On input either @errp or *@errp must be NULL.
2039  *
2040  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2041  * - R: mem_blk->phys_index
2042  * - W: mem_blk->online
2043  * - W: mem_blk->can_offline
2044  *
2045  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2046  * - R: mem_blk->phys_index
2047  * - R: mem_blk->online
2048  *-  R: mem_blk->can_offline
2049  * Written members remain unmodified on error.
2050  */
2051 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2052                                   GuestMemoryBlockResponse *result,
2053                                   Error **errp)
2054 {
2055     char *dirpath;
2056     int dirfd;
2057     char *status;
2058     Error *local_err = NULL;
2059
2060     if (!sys2memblk) {
2061         DIR *dp;
2062
2063         if (!result) {
2064             error_setg(errp, "Internal error, 'result' should not be NULL");
2065             return;
2066         }
2067         errno = 0;
2068         dp = opendir("/sys/devices/system/memory/");
2069          /* if there is no 'memory' directory in sysfs,
2070          * we think this VM does not support online/offline memory block,
2071          * any other solution?
2072          */
2073         if (!dp && errno == ENOENT) {
2074             result->response =
2075                 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2076             goto out1;
2077         }
2078         closedir(dp);
2079     }
2080
2081     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2082                               mem_blk->phys_index);
2083     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2084     if (dirfd == -1) {
2085         if (sys2memblk) {
2086             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2087         } else {
2088             if (errno == ENOENT) {
2089                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2090             } else {
2091                 result->response =
2092                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2093             }
2094         }
2095         g_free(dirpath);
2096         goto out1;
2097     }
2098     g_free(dirpath);
2099
2100     status = g_malloc0(10);
2101     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2102     if (local_err) {
2103         /* treat with sysfs file that not exist in old kernel */
2104         if (errno == ENOENT) {
2105             error_free(local_err);
2106             if (sys2memblk) {
2107                 mem_blk->online = true;
2108                 mem_blk->can_offline = false;
2109             } else if (!mem_blk->online) {
2110                 result->response =
2111                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2112             }
2113         } else {
2114             if (sys2memblk) {
2115                 error_propagate(errp, local_err);
2116             } else {
2117                 result->response =
2118                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2119             }
2120         }
2121         goto out2;
2122     }
2123
2124     if (sys2memblk) {
2125         char removable = '0';
2126
2127         mem_blk->online = (strncmp(status, "online", 6) == 0);
2128
2129         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2130         if (local_err) {
2131             /* if no 'removable' file, it does't support offline mem blk */
2132             if (errno == ENOENT) {
2133                 error_free(local_err);
2134                 mem_blk->can_offline = false;
2135             } else {
2136                 error_propagate(errp, local_err);
2137             }
2138         } else {
2139             mem_blk->can_offline = (removable != '0');
2140         }
2141     } else {
2142         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2143             char *new_state = mem_blk->online ? g_strdup("online") :
2144                                                 g_strdup("offline");
2145
2146             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2147                                 &local_err);
2148             g_free(new_state);
2149             if (local_err) {
2150                 error_free(local_err);
2151                 result->response =
2152                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2153                 goto out2;
2154             }
2155
2156             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2157             result->has_error_code = false;
2158         } /* otherwise pretend successful re-(on|off)-lining */
2159     }
2160     g_free(status);
2161     close(dirfd);
2162     return;
2163
2164 out2:
2165     g_free(status);
2166     close(dirfd);
2167 out1:
2168     if (!sys2memblk) {
2169         result->has_error_code = true;
2170         result->error_code = errno;
2171     }
2172 }
2173
2174 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2175 {
2176     GuestMemoryBlockList *head, **link;
2177     Error *local_err = NULL;
2178     struct dirent *de;
2179     DIR *dp;
2180
2181     head = NULL;
2182     link = &head;
2183
2184     dp = opendir("/sys/devices/system/memory/");
2185     if (!dp) {
2186         error_setg_errno(errp, errno, "Can't open directory"
2187                          "\"/sys/devices/system/memory/\"\n");
2188         return NULL;
2189     }
2190
2191     /* Note: the phys_index of memory block may be discontinuous,
2192      * this is because a memblk is the unit of the Sparse Memory design, which
2193      * allows discontinuous memory ranges (ex. NUMA), so here we should
2194      * traverse the memory block directory.
2195      */
2196     while ((de = readdir(dp)) != NULL) {
2197         GuestMemoryBlock *mem_blk;
2198         GuestMemoryBlockList *entry;
2199
2200         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2201             !(de->d_type & DT_DIR)) {
2202             continue;
2203         }
2204
2205         mem_blk = g_malloc0(sizeof *mem_blk);
2206         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2207         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2208         mem_blk->has_can_offline = true; /* lolspeak ftw */
2209         transfer_memory_block(mem_blk, true, NULL, &local_err);
2210
2211         entry = g_malloc0(sizeof *entry);
2212         entry->value = mem_blk;
2213
2214         *link = entry;
2215         link = &entry->next;
2216     }
2217
2218     closedir(dp);
2219     if (local_err == NULL) {
2220         /* there's no guest with zero memory blocks */
2221         if (head == NULL) {
2222             error_setg(errp, "guest reported zero memory blocks!");
2223         }
2224         return head;
2225     }
2226
2227     qapi_free_GuestMemoryBlockList(head);
2228     error_propagate(errp, local_err);
2229     return NULL;
2230 }
2231
2232 GuestMemoryBlockResponseList *
2233 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2234 {
2235     GuestMemoryBlockResponseList *head, **link;
2236     Error *local_err = NULL;
2237
2238     head = NULL;
2239     link = &head;
2240
2241     while (mem_blks != NULL) {
2242         GuestMemoryBlockResponse *result;
2243         GuestMemoryBlockResponseList *entry;
2244         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2245
2246         result = g_malloc0(sizeof(*result));
2247         result->phys_index = current_mem_blk->phys_index;
2248         transfer_memory_block(current_mem_blk, false, result, &local_err);
2249         if (local_err) { /* should never happen */
2250             goto err;
2251         }
2252         entry = g_malloc0(sizeof *entry);
2253         entry->value = result;
2254
2255         *link = entry;
2256         link = &entry->next;
2257         mem_blks = mem_blks->next;
2258     }
2259
2260     return head;
2261 err:
2262     qapi_free_GuestMemoryBlockResponseList(head);
2263     error_propagate(errp, local_err);
2264     return NULL;
2265 }
2266
2267 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2268 {
2269     Error *local_err = NULL;
2270     char *dirpath;
2271     int dirfd;
2272     char *buf;
2273     GuestMemoryBlockInfo *info;
2274
2275     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2276     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2277     if (dirfd == -1) {
2278         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2279         g_free(dirpath);
2280         return NULL;
2281     }
2282     g_free(dirpath);
2283
2284     buf = g_malloc0(20);
2285     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2286     if (local_err) {
2287         g_free(buf);
2288         error_propagate(errp, local_err);
2289         return NULL;
2290     }
2291
2292     info = g_new0(GuestMemoryBlockInfo, 1);
2293     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2294
2295     g_free(buf);
2296
2297     return info;
2298 }
2299
2300 #else /* defined(__linux__) */
2301
2302 void qmp_guest_suspend_disk(Error **errp)
2303 {
2304     error_set(errp, QERR_UNSUPPORTED);
2305 }
2306
2307 void qmp_guest_suspend_ram(Error **errp)
2308 {
2309     error_set(errp, QERR_UNSUPPORTED);
2310 }
2311
2312 void qmp_guest_suspend_hybrid(Error **errp)
2313 {
2314     error_set(errp, QERR_UNSUPPORTED);
2315 }
2316
2317 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2318 {
2319     error_set(errp, QERR_UNSUPPORTED);
2320     return NULL;
2321 }
2322
2323 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2324 {
2325     error_set(errp, QERR_UNSUPPORTED);
2326     return NULL;
2327 }
2328
2329 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2330 {
2331     error_set(errp, QERR_UNSUPPORTED);
2332     return -1;
2333 }
2334
2335 void qmp_guest_set_user_password(const char *username,
2336                                  const char *password,
2337                                  bool crypted,
2338                                  Error **errp)
2339 {
2340     error_set(errp, QERR_UNSUPPORTED);
2341 }
2342
2343 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2344 {
2345     error_set(errp, QERR_UNSUPPORTED);
2346     return NULL;
2347 }
2348
2349 GuestMemoryBlockResponseList *
2350 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2351 {
2352     error_set(errp, QERR_UNSUPPORTED);
2353     return NULL;
2354 }
2355
2356 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2357 {
2358     error_set(errp, QERR_UNSUPPORTED);
2359     return NULL;
2360 }
2361
2362 #endif
2363
2364 #if !defined(CONFIG_FSFREEZE)
2365
2366 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2367 {
2368     error_set(errp, QERR_UNSUPPORTED);
2369     return NULL;
2370 }
2371
2372 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2373 {
2374     error_set(errp, QERR_UNSUPPORTED);
2375
2376     return 0;
2377 }
2378
2379 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2380 {
2381     error_set(errp, QERR_UNSUPPORTED);
2382
2383     return 0;
2384 }
2385
2386 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2387                                        strList *mountpoints,
2388                                        Error **errp)
2389 {
2390     error_set(errp, QERR_UNSUPPORTED);
2391
2392     return 0;
2393 }
2394
2395 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2396 {
2397     error_set(errp, QERR_UNSUPPORTED);
2398
2399     return 0;
2400 }
2401 #endif /* CONFIG_FSFREEZE */
2402
2403 #if !defined(CONFIG_FSTRIM)
2404 void qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
2405 {
2406     error_set(errp, QERR_UNSUPPORTED);
2407 }
2408 #endif
2409
2410 /* add unsupported commands to the blacklist */
2411 GList *ga_command_blacklist_init(GList *blacklist)
2412 {
2413 #if !defined(__linux__)
2414     {
2415         const char *list[] = {
2416             "guest-suspend-disk", "guest-suspend-ram",
2417             "guest-suspend-hybrid", "guest-network-get-interfaces",
2418             "guest-get-vcpus", "guest-set-vcpus",
2419             "guest-get-memory-blocks", "guest-set-memory-blocks",
2420             "guest-get-memory-block-size", NULL};
2421         char **p = (char **)list;
2422
2423         while (*p) {
2424             blacklist = g_list_append(blacklist, *p++);
2425         }
2426     }
2427 #endif
2428
2429 #if !defined(CONFIG_FSFREEZE)
2430     {
2431         const char *list[] = {
2432             "guest-get-fsinfo", "guest-fsfreeze-status",
2433             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
2434             "guest-fsfreeze-thaw", "guest-get-fsinfo", NULL};
2435         char **p = (char **)list;
2436
2437         while (*p) {
2438             blacklist = g_list_append(blacklist, *p++);
2439         }
2440     }
2441 #endif
2442
2443 #if !defined(CONFIG_FSTRIM)
2444     blacklist = g_list_append(blacklist, (char *)"guest-fstrim");
2445 #endif
2446
2447     return blacklist;
2448 }
2449
2450 /* register init/cleanup routines for stateful command groups */
2451 void ga_command_state_init(GAState *s, GACommandState *cs)
2452 {
2453 #if defined(CONFIG_FSFREEZE)
2454     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
2455 #endif
2456     ga_command_state_add(cs, guest_file_init, NULL);
2457 }