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 /*
 865  * Store disk device info for devices on the PCI bus.
 866  * Returns true if information has been stored, or false for failure.
 867  */
 868 static bool build_guest_fsinfo_for_pci_dev(char const *syspath,
 869                                            GuestDiskAddress *disk,
 870                                            Error **errp)
 871 {
 872     unsigned int pci[4], host, hosts[8], tgt[3];
 873     int i, nhosts = 0, pcilen;
 874     GuestPCIAddress *pciaddr = disk->pci_controller;
 875     bool has_ata = false, has_host = false, has_tgt = false;
 876     char *p, *q, *driver = NULL;
 877     bool ret = false;
 878
 879     p = strstr(syspath, "/devices/pci");
 880     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
 881                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
 882         g_debug("only pci device is supported: sysfs path '%s'", syspath);
 883         return false;
 884     }
 885
 886     p += 12 + pcilen;
 887     while (true) {
 888         driver = get_pci_driver(syspath, p - syspath, errp);
 889         if (driver && (g_str_equal(driver, "ata_piix") ||
 890                        g_str_equal(driver, "sym53c8xx") ||
 891                        g_str_equal(driver, "virtio-pci") ||
 892                        g_str_equal(driver, "ahci"))) {
 893             break;
 894         }
 895
 896         g_free(driver);
 897         if (sscanf(p, "/%x:%x:%x.%x%n",
 898                           pci, pci + 1, pci + 2, pci + 3, &pcilen) == 4) {
 899             p += pcilen;
 900             continue;
 901         }
 902
 903         g_debug("unsupported driver or sysfs path '%s'", syspath);
 904         return false;
 905     }
 906
 907     p = strstr(syspath, "/target");
 908     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
 909                     tgt, tgt + 1, tgt + 2) == 3) {
 910         has_tgt = true;
 911     }
 912
 913     p = strstr(syspath, "/ata");
 914     if (p) {
 915         q = p + 4;
 916         has_ata = true;
 917     } else {
 918         p = strstr(syspath, "/host");
 919         q = p + 5;
 920     }
 921     if (p && sscanf(q, "%u", &host) == 1) {
 922         has_host = true;
 923         nhosts = build_hosts(syspath, p, has_ata, hosts,
 924                              ARRAY_SIZE(hosts), errp);
 925         if (nhosts < 0) {
 926             goto cleanup;
 927         }
 928     }
 929
 930     pciaddr->domain = pci[0];
 931     pciaddr->bus = pci[1];
 932     pciaddr->slot = pci[2];
 933     pciaddr->function = pci[3];
 934
 935     if (strcmp(driver, "ata_piix") == 0) {
 936         /* a host per ide bus, target*:0:<unit>:0 */
 937         if (!has_host || !has_tgt) {
 938             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 939             goto cleanup;
 940         }
 941         for (i = 0; i < nhosts; i++) {
 942             if (host == hosts[i]) {
 943                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
 944                 disk->bus = i;
 945                 disk->unit = tgt[1];
 946                 break;
 947             }
 948         }
 949         if (i >= nhosts) {
 950             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 951             goto cleanup;
 952         }
 953     } else if (strcmp(driver, "sym53c8xx") == 0) {
 954         /* scsi(LSI Logic): target*:0:<unit>:0 */
 955         if (!has_tgt) {
 956             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 957             goto cleanup;
 958         }
 959         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 960         disk->unit = tgt[1];
 961     } else if (strcmp(driver, "virtio-pci") == 0) {
 962         if (has_tgt) {
 963             /* virtio-scsi: target*:0:0:<unit> */
 964             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 965             disk->unit = tgt[2];
 966         } else {
 967             /* virtio-blk: 1 disk per 1 device */
 968             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
 969         }
 970     } else if (strcmp(driver, "ahci") == 0) {
 971         /* ahci: 1 host per 1 unit */
 972         if (!has_host || !has_tgt) {
 973             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 974             goto cleanup;
 975         }
 976         for (i = 0; i < nhosts; i++) {
 977             if (host == hosts[i]) {
 978                 disk->unit = i;
 979                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
 980                 break;
 981             }
 982         }
 983         if (i >= nhosts) {
 984             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 985             goto cleanup;
 986         }
 987     } else {
 988         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
 989         goto cleanup;
 990     }
 991
 992     ret = true;
 993
 994 cleanup:
 995     g_free(driver);
 996     return ret;
 997 }
 998
 999 /*
1000  * Store disk device info for non-PCI virtio devices (for example s390x
1001  * channel I/O devices). Returns true if information has been stored, or
1002  * false for failure.
1003  */
1004 static bool build_guest_fsinfo_for_nonpci_virtio(char const *syspath,
1005                                                  GuestDiskAddress *disk,
1006                                                  Error **errp)
1007 {
1008     unsigned int tgt[3];
1009     char *p;
1010
1011     if (!strstr(syspath, "/virtio") || !strstr(syspath, "/block")) {
1012         g_debug("Unsupported virtio device '%s'", syspath);
1013         return false;
1014     }
1015
1016     p = strstr(syspath, "/target");
1017     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
1018                     &tgt[0], &tgt[1], &tgt[2]) == 3) {
1019         /* virtio-scsi: target*:0:<target>:<unit> */
1020         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
1021         disk->bus = tgt[0];
1022         disk->target = tgt[1];
1023         disk->unit = tgt[2];
1024     } else {
1025         /* virtio-blk: 1 disk per 1 device */
1026         disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
1027     }
1028
1029     return true;
1030 }
1031
1032 /*
1033  * Store disk device info for CCW devices (s390x channel I/O devices).
1034  * Returns true if information has been stored, or false for failure.
1035  */
1036 static bool build_guest_fsinfo_for_ccw_dev(char const *syspath,
1037                                            GuestDiskAddress *disk,
1038                                            Error **errp)
1039 {
1040     unsigned int cssid, ssid, subchno, devno;
1041     char *p;
1042
1043     p = strstr(syspath, "/devices/css");
1044     if (!p || sscanf(p + 12, "%*x/%x.%x.%x/%*x.%*x.%x/",
1045                      &cssid, &ssid, &subchno, &devno) < 4) {
1046         g_debug("could not parse ccw device sysfs path: %s", syspath);
1047         return false;
1048     }
1049
1050     disk->has_ccw_address = true;
1051     disk->ccw_address = g_new0(GuestCCWAddress, 1);
1052     disk->ccw_address->cssid = cssid;
1053     disk->ccw_address->ssid = ssid;
1054     disk->ccw_address->subchno = subchno;
1055     disk->ccw_address->devno = devno;
1056
1057     if (strstr(p, "/virtio")) {
1058         build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1059     }
1060
1061     return true;
1062 }
1063
1064 /* Store disk device info specified by @sysfs into @fs */
1065 static void build_guest_fsinfo_for_real_device(char const *syspath,
1066                                                GuestFilesystemInfo *fs,
1067                                                Error **errp)
1068 {
1069     GuestDiskAddress *disk;
1070     GuestPCIAddress *pciaddr;
1071     bool has_hwinf;
1072 #ifdef CONFIG_LIBUDEV
1073     struct udev *udev = NULL;
1074     struct udev_device *udevice = NULL;
1075 #endif
1076
1077     pciaddr = g_new0(GuestPCIAddress, 1);
1078     pciaddr->domain = -1;                       /* -1 means field is invalid */
1079     pciaddr->bus = -1;
1080     pciaddr->slot = -1;
1081     pciaddr->function = -1;
1082
1083     disk = g_new0(GuestDiskAddress, 1);
1084     disk->pci_controller = pciaddr;
1085     disk->bus_type = GUEST_DISK_BUS_TYPE_UNKNOWN;
1086
1087 #ifdef CONFIG_LIBUDEV
1088     udev = udev_new();
1089     udevice = udev_device_new_from_syspath(udev, syspath);
1090     if (udev == NULL || udevice == NULL) {
1091         g_debug("failed to query udev");
1092     } else {
1093         const char *devnode, *serial;
1094         devnode = udev_device_get_devnode(udevice);
1095         if (devnode != NULL) {
1096             disk->dev = g_strdup(devnode);
1097             disk->has_dev = true;
1098         }
1099         serial = udev_device_get_property_value(udevice, "ID_SERIAL");
1100         if (serial != NULL && *serial != 0) {
1101             disk->serial = g_strdup(serial);
1102             disk->has_serial = true;
1103         }
1104     }
1105
1106     udev_unref(udev);
1107     udev_device_unref(udevice);
1108 #endif
1109
1110     if (strstr(syspath, "/devices/pci")) {
1111         has_hwinf = build_guest_fsinfo_for_pci_dev(syspath, disk, errp);
1112     } else if (strstr(syspath, "/devices/css")) {
1113         has_hwinf = build_guest_fsinfo_for_ccw_dev(syspath, disk, errp);
1114     } else if (strstr(syspath, "/virtio")) {
1115         has_hwinf = build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1116     } else {
1117         g_debug("Unsupported device type for '%s'", syspath);
1118         has_hwinf = false;
1119     }
1120
1121     if (has_hwinf || disk->has_dev || disk->has_serial) {
1122         QAPI_LIST_PREPEND(fs->disk, disk);
1123     } else {
1124         qapi_free_GuestDiskAddress(disk);
1125     }
1126 }
1127
1128 static void build_guest_fsinfo_for_device(char const *devpath,
1129                                           GuestFilesystemInfo *fs,
1130                                           Error **errp);
1131
1132 /* Store a list of slave devices of virtual volume specified by @syspath into
1133  * @fs */
1134 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1135                                                   GuestFilesystemInfo *fs,
1136                                                   Error **errp)
1137 {
1138     Error *err = NULL;
1139     DIR *dir;
1140     char *dirpath;
1141     struct dirent *entry;
1142
1143     dirpath = g_strdup_printf("%s/slaves", syspath);
1144     dir = opendir(dirpath);
1145     if (!dir) {
1146         if (errno != ENOENT) {
1147             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1148         }
1149         g_free(dirpath);
1150         return;
1151     }
1152
1153     for (;;) {
1154         errno = 0;
1155         entry = readdir(dir);
1156         if (entry == NULL) {
1157             if (errno) {
1158                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1159             }
1160             break;
1161         }
1162
1163         if (entry->d_type == DT_LNK) {
1164             char *path;
1165
1166             g_debug(" slave device '%s'", entry->d_name);
1167             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1168             build_guest_fsinfo_for_device(path, fs, &err);
1169             g_free(path);
1170
1171             if (err) {
1172                 error_propagate(errp, err);
1173                 break;
1174             }
1175         }
1176     }
1177
1178     g_free(dirpath);
1179     closedir(dir);
1180 }
1181
1182 static bool is_disk_virtual(const char *devpath, Error **errp)
1183 {
1184     g_autofree char *syspath = realpath(devpath, NULL);
1185
1186     if (!syspath) {
1187         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1188         return false;
1189     }
1190     return strstr(syspath, "/devices/virtual/block/") != NULL;
1191 }
1192
1193 /* Dispatch to functions for virtual/real device */
1194 static void build_guest_fsinfo_for_device(char const *devpath,
1195                                           GuestFilesystemInfo *fs,
1196                                           Error **errp)
1197 {
1198     ERRP_GUARD();
1199     g_autofree char *syspath = NULL;
1200     bool is_virtual = false;
1201
1202     syspath = realpath(devpath, NULL);
1203     if (!syspath) {
1204         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1205         return;
1206     }
1207
1208     if (!fs->name) {
1209         fs->name = g_path_get_basename(syspath);
1210     }
1211
1212     g_debug("  parse sysfs path '%s'", syspath);
1213     is_virtual = is_disk_virtual(syspath, errp);
1214     if (*errp != NULL) {
1215         return;
1216     }
1217     if (is_virtual) {
1218         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1219     } else {
1220         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1221     }
1222 }
1223
1224 #ifdef CONFIG_LIBUDEV
1225
1226 /*
1227  * Wrapper around build_guest_fsinfo_for_device() for getting just
1228  * the disk address.
1229  */
1230 static GuestDiskAddress *get_disk_address(const char *syspath, Error **errp)
1231 {
1232     g_autoptr(GuestFilesystemInfo) fs = NULL;
1233
1234     fs = g_new0(GuestFilesystemInfo, 1);
1235     build_guest_fsinfo_for_device(syspath, fs, errp);
1236     if (fs->disk != NULL) {
1237         return g_steal_pointer(&fs->disk->value);
1238     }
1239     return NULL;
1240 }
1241
1242 static char *get_alias_for_syspath(const char *syspath)
1243 {
1244     struct udev *udev = NULL;
1245     struct udev_device *udevice = NULL;
1246     char *ret = NULL;
1247
1248     udev = udev_new();
1249     if (udev == NULL) {
1250         g_debug("failed to query udev");
1251         goto out;
1252     }
1253     udevice = udev_device_new_from_syspath(udev, syspath);
1254     if (udevice == NULL) {
1255         g_debug("failed to query udev for path: %s", syspath);
1256         goto out;
1257     } else {
1258         const char *alias = udev_device_get_property_value(
1259             udevice, "DM_NAME");
1260         /*
1261          * NULL means there was an error and empty string means there is no
1262          * alias. In case of no alias we return NULL instead of empty string.
1263          */
1264         if (alias == NULL) {
1265             g_debug("failed to query udev for device alias for: %s",
1266                 syspath);
1267         } else if (*alias != 0) {
1268             ret = g_strdup(alias);
1269         }
1270     }
1271
1272 out:
1273     udev_unref(udev);
1274     udev_device_unref(udevice);
1275     return ret;
1276 }
1277
1278 static char *get_device_for_syspath(const char *syspath)
1279 {
1280     struct udev *udev = NULL;
1281     struct udev_device *udevice = NULL;
1282     char *ret = NULL;
1283
1284     udev = udev_new();
1285     if (udev == NULL) {
1286         g_debug("failed to query udev");
1287         goto out;
1288     }
1289     udevice = udev_device_new_from_syspath(udev, syspath);
1290     if (udevice == NULL) {
1291         g_debug("failed to query udev for path: %s", syspath);
1292         goto out;
1293     } else {
1294         ret = g_strdup(udev_device_get_devnode(udevice));
1295     }
1296
1297 out:
1298     udev_unref(udev);
1299     udev_device_unref(udevice);
1300     return ret;
1301 }
1302
1303 static void get_disk_deps(const char *disk_dir, GuestDiskInfo *disk)
1304 {
1305     g_autofree char *deps_dir = NULL;
1306     const gchar *dep;
1307     GDir *dp_deps = NULL;
1308
1309     /* List dependent disks */
1310     deps_dir = g_strdup_printf("%s/slaves", disk_dir);
1311     g_debug("  listing entries in: %s", deps_dir);
1312     dp_deps = g_dir_open(deps_dir, 0, NULL);
1313     if (dp_deps == NULL) {
1314         g_debug("failed to list entries in %s", deps_dir);
1315         return;
1316     }
1317     disk->has_dependencies = true;
1318     while ((dep = g_dir_read_name(dp_deps)) != NULL) {
1319         g_autofree char *dep_dir = NULL;
1320         char *dev_name;
1321
1322         /* Add dependent disks */
1323         dep_dir = g_strdup_printf("%s/%s", deps_dir, dep);
1324         dev_name = get_device_for_syspath(dep_dir);
1325         if (dev_name != NULL) {
1326             g_debug("  adding dependent device: %s", dev_name);
1327             QAPI_LIST_PREPEND(disk->dependencies, dev_name);
1328         }
1329     }
1330     g_dir_close(dp_deps);
1331 }
1332
1333 /*
1334  * Detect partitions subdirectory, name is "<disk_name><number>" or
1335  * "<disk_name>p<number>"
1336  *
1337  * @disk_name -- last component of /sys path (e.g. sda)
1338  * @disk_dir -- sys path of the disk (e.g. /sys/block/sda)
1339  * @disk_dev -- device node of the disk (e.g. /dev/sda)
1340  */
1341 static GuestDiskInfoList *get_disk_partitions(
1342     GuestDiskInfoList *list,
1343     const char *disk_name, const char *disk_dir,
1344     const char *disk_dev)
1345 {
1346     GuestDiskInfoList *ret = list;
1347     struct dirent *de_disk;
1348     DIR *dp_disk = NULL;
1349     size_t len = strlen(disk_name);
1350
1351     dp_disk = opendir(disk_dir);
1352     while ((de_disk = readdir(dp_disk)) != NULL) {
1353         g_autofree char *partition_dir = NULL;
1354         char *dev_name;
1355         GuestDiskInfo *partition;
1356
1357         if (!(de_disk->d_type & DT_DIR)) {
1358             continue;
1359         }
1360
1361         if (!(strncmp(disk_name, de_disk->d_name, len) == 0 &&
1362             ((*(de_disk->d_name + len) == 'p' &&
1363             isdigit(*(de_disk->d_name + len + 1))) ||
1364                 isdigit(*(de_disk->d_name + len))))) {
1365             continue;
1366         }
1367
1368         partition_dir = g_strdup_printf("%s/%s",
1369             disk_dir, de_disk->d_name);
1370         dev_name = get_device_for_syspath(partition_dir);
1371         if (dev_name == NULL) {
1372             g_debug("Failed to get device name for syspath: %s",
1373                 disk_dir);
1374             continue;
1375         }
1376         partition = g_new0(GuestDiskInfo, 1);
1377         partition->name = dev_name;
1378         partition->partition = true;
1379         /* Add parent disk as dependent for easier tracking of hierarchy */
1380         QAPI_LIST_PREPEND(partition->dependencies, g_strdup(disk_dev));
1381
1382         QAPI_LIST_PREPEND(ret, partition);
1383     }
1384     closedir(dp_disk);
1385
1386     return ret;
1387 }
1388
1389 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1390 {
1391     GuestDiskInfoList *ret = NULL;
1392     GuestDiskInfo *disk;
1393     DIR *dp = NULL;
1394     struct dirent *de = NULL;
1395
1396     g_debug("listing /sys/block directory");
1397     dp = opendir("/sys/block");
1398     if (dp == NULL) {
1399         error_setg_errno(errp, errno, "Can't open directory \"/sys/block\"");
1400         return NULL;
1401     }
1402     while ((de = readdir(dp)) != NULL) {
1403         g_autofree char *disk_dir = NULL, *line = NULL,
1404             *size_path = NULL;
1405         char *dev_name;
1406         Error *local_err = NULL;
1407         if (de->d_type != DT_LNK) {
1408             g_debug("  skipping entry: %s", de->d_name);
1409             continue;
1410         }
1411
1412         /* Check size and skip zero-sized disks */
1413         g_debug("  checking disk size");
1414         size_path = g_strdup_printf("/sys/block/%s/size", de->d_name);
1415         if (!g_file_get_contents(size_path, &line, NULL, NULL)) {
1416             g_debug("  failed to read disk size");
1417             continue;
1418         }
1419         if (g_strcmp0(line, "0\n") == 0) {
1420             g_debug("  skipping zero-sized disk");
1421             continue;
1422         }
1423
1424         g_debug("  adding %s", de->d_name);
1425         disk_dir = g_strdup_printf("/sys/block/%s", de->d_name);
1426         dev_name = get_device_for_syspath(disk_dir);
1427         if (dev_name == NULL) {
1428             g_debug("Failed to get device name for syspath: %s",
1429                 disk_dir);
1430             continue;
1431         }
1432         disk = g_new0(GuestDiskInfo, 1);
1433         disk->name = dev_name;
1434         disk->partition = false;
1435         disk->alias = get_alias_for_syspath(disk_dir);
1436         disk->has_alias = (disk->alias != NULL);
1437         QAPI_LIST_PREPEND(ret, disk);
1438
1439         /* Get address for non-virtual devices */
1440         bool is_virtual = is_disk_virtual(disk_dir, &local_err);
1441         if (local_err != NULL) {
1442             g_debug("  failed to check disk path, ignoring error: %s",
1443                 error_get_pretty(local_err));
1444             error_free(local_err);
1445             local_err = NULL;
1446             /* Don't try to get the address */
1447             is_virtual = true;
1448         }
1449         if (!is_virtual) {
1450             disk->address = get_disk_address(disk_dir, &local_err);
1451             if (local_err != NULL) {
1452                 g_debug("  failed to get device info, ignoring error: %s",
1453                     error_get_pretty(local_err));
1454                 error_free(local_err);
1455                 local_err = NULL;
1456             } else if (disk->address != NULL) {
1457                 disk->has_address = true;
1458             }
1459         }
1460
1461         get_disk_deps(disk_dir, disk);
1462         ret = get_disk_partitions(ret, de->d_name, disk_dir, dev_name);
1463     }
1464
1465     closedir(dp);
1466
1467     return ret;
1468 }
1469
1470 #else
1471
1472 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1473 {
1474     error_setg(errp, QERR_UNSUPPORTED);
1475     return NULL;
1476 }
1477
1478 #endif
1479
1480 /* Return a list of the disk device(s)' info which @mount lies on */
1481 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1482                                                Error **errp)
1483 {
1484     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1485     struct statvfs buf;
1486     unsigned long used, nonroot_total, fr_size;
1487     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1488                                     mount->devmajor, mount->devminor);
1489
1490     fs->mountpoint = g_strdup(mount->dirname);
1491     fs->type = g_strdup(mount->devtype);
1492     build_guest_fsinfo_for_device(devpath, fs, errp);
1493
1494     if (statvfs(fs->mountpoint, &buf) == 0) {
1495         fr_size = buf.f_frsize;
1496         used = buf.f_blocks - buf.f_bfree;
1497         nonroot_total = used + buf.f_bavail;
1498         fs->used_bytes = used * fr_size;
1499         fs->total_bytes = nonroot_total * fr_size;
1500
1501         fs->has_total_bytes = true;
1502         fs->has_used_bytes = true;
1503     }
1504
1505     g_free(devpath);
1506
1507     return fs;
1508 }
1509
1510 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1511 {
1512     FsMountList mounts;
1513     struct FsMount *mount;
1514     GuestFilesystemInfoList *ret = NULL;
1515     Error *local_err = NULL;
1516
1517     QTAILQ_INIT(&mounts);
1518     build_fs_mount_list(&mounts, &local_err);
1519     if (local_err) {
1520         error_propagate(errp, local_err);
1521         return NULL;
1522     }
1523
1524     QTAILQ_FOREACH(mount, &mounts, next) {
1525         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1526
1527         QAPI_LIST_PREPEND(ret, build_guest_fsinfo(mount, &local_err));
1528         if (local_err) {
1529             error_propagate(errp, local_err);
1530             qapi_free_GuestFilesystemInfoList(ret);
1531             ret = NULL;
1532             break;
1533         }
1534     }
1535
1536     free_fs_mount_list(&mounts);
1537     return ret;
1538 }
1539
1540
1541 typedef enum {
1542     FSFREEZE_HOOK_THAW = 0,
1543     FSFREEZE_HOOK_FREEZE,
1544 } FsfreezeHookArg;
1545
1546 static const char *fsfreeze_hook_arg_string[] = {
1547     "thaw",
1548     "freeze",
1549 };
1550
1551 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1552 {
1553     int status;
1554     pid_t pid;
1555     const char *hook;
1556     const char *arg_str = fsfreeze_hook_arg_string[arg];
1557     Error *local_err = NULL;
1558
1559     hook = ga_fsfreeze_hook(ga_state);
1560     if (!hook) {
1561         return;
1562     }
1563     if (access(hook, X_OK) != 0) {
1564         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1565         return;
1566     }
1567
1568     slog("executing fsfreeze hook with arg '%s'", arg_str);
1569     pid = fork();
1570     if (pid == 0) {
1571         setsid();
1572         reopen_fd_to_null(0);
1573         reopen_fd_to_null(1);
1574         reopen_fd_to_null(2);
1575
1576         execle(hook, hook, arg_str, NULL, environ);
1577         _exit(EXIT_FAILURE);
1578     } else if (pid < 0) {
1579         error_setg_errno(errp, errno, "failed to create child process");
1580         return;
1581     }
1582
1583     ga_wait_child(pid, &status, &local_err);
1584     if (local_err) {
1585         error_propagate(errp, local_err);
1586         return;
1587     }
1588
1589     if (!WIFEXITED(status)) {
1590         error_setg(errp, "fsfreeze hook has terminated abnormally");
1591         return;
1592     }
1593
1594     status = WEXITSTATUS(status);
1595     if (status) {
1596         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1597         return;
1598     }
1599 }
1600
1601 /*
1602  * Return status of freeze/thaw
1603  */
1604 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1605 {
1606     if (ga_is_frozen(ga_state)) {
1607         return GUEST_FSFREEZE_STATUS_FROZEN;
1608     }
1609
1610     return GUEST_FSFREEZE_STATUS_THAWED;
1611 }
1612
1613 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1614 {
1615     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1616 }
1617
1618 /*
1619  * Walk list of mounted file systems in the guest, and freeze the ones which
1620  * are real local file systems.
1621  */
1622 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1623                                        strList *mountpoints,
1624                                        Error **errp)
1625 {
1626     int ret = 0, i = 0;
1627     strList *list;
1628     FsMountList mounts;
1629     struct FsMount *mount;
1630     Error *local_err = NULL;
1631     int fd;
1632
1633     slog("guest-fsfreeze called");
1634
1635     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1636     if (local_err) {
1637         error_propagate(errp, local_err);
1638         return -1;
1639     }
1640
1641     QTAILQ_INIT(&mounts);
1642     build_fs_mount_list(&mounts, &local_err);
1643     if (local_err) {
1644         error_propagate(errp, local_err);
1645         return -1;
1646     }
1647
1648     /* cannot risk guest agent blocking itself on a write in this state */
1649     ga_set_frozen(ga_state);
1650
1651     QTAILQ_FOREACH_REVERSE(mount, &mounts, next) {
1652         /* To issue fsfreeze in the reverse order of mounts, check if the
1653          * mount is listed in the list here */
1654         if (has_mountpoints) {
1655             for (list = mountpoints; list; list = list->next) {
1656                 if (strcmp(list->value, mount->dirname) == 0) {
1657                     break;
1658                 }
1659             }
1660             if (!list) {
1661                 continue;
1662             }
1663         }
1664
1665         fd = qemu_open_old(mount->dirname, O_RDONLY);
1666         if (fd == -1) {
1667             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1668             goto error;
1669         }
1670
1671         /* we try to cull filesystems we know won't work in advance, but other
1672          * filesystems may not implement fsfreeze for less obvious reasons.
1673          * these will report EOPNOTSUPP. we simply ignore these when tallying
1674          * the number of frozen filesystems.
1675          * if a filesystem is mounted more than once (aka bind mount) a
1676          * consecutive attempt to freeze an already frozen filesystem will
1677          * return EBUSY.
1678          *
1679          * any other error means a failure to freeze a filesystem we
1680          * expect to be freezable, so return an error in those cases
1681          * and return system to thawed state.
1682          */
1683         ret = ioctl(fd, FIFREEZE);
1684         if (ret == -1) {
1685             if (errno != EOPNOTSUPP && errno != EBUSY) {
1686                 error_setg_errno(errp, errno, "failed to freeze %s",
1687                                  mount->dirname);
1688                 close(fd);
1689                 goto error;
1690             }
1691         } else {
1692             i++;
1693         }
1694         close(fd);
1695     }
1696
1697     free_fs_mount_list(&mounts);
1698     /* We may not issue any FIFREEZE here.
1699      * Just unset ga_state here and ready for the next call.
1700      */
1701     if (i == 0) {
1702         ga_unset_frozen(ga_state);
1703     }
1704     return i;
1705
1706 error:
1707     free_fs_mount_list(&mounts);
1708     qmp_guest_fsfreeze_thaw(NULL);
1709     return 0;
1710 }
1711
1712 /*
1713  * Walk list of frozen file systems in the guest, and thaw them.
1714  */
1715 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1716 {
1717     int ret;
1718     FsMountList mounts;
1719     FsMount *mount;
1720     int fd, i = 0, logged;
1721     Error *local_err = NULL;
1722
1723     QTAILQ_INIT(&mounts);
1724     build_fs_mount_list(&mounts, &local_err);
1725     if (local_err) {
1726         error_propagate(errp, local_err);
1727         return 0;
1728     }
1729
1730     QTAILQ_FOREACH(mount, &mounts, next) {
1731         logged = false;
1732         fd = qemu_open_old(mount->dirname, O_RDONLY);
1733         if (fd == -1) {
1734             continue;
1735         }
1736         /* we have no way of knowing whether a filesystem was actually unfrozen
1737          * as a result of a successful call to FITHAW, only that if an error
1738          * was returned the filesystem was *not* unfrozen by that particular
1739          * call.
1740          *
1741          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1742          * to unfreeze, continuing issuing FITHAW until an error is returned,
1743          * in which case either the filesystem is in an unfreezable state, or,
1744          * more likely, it was thawed previously (and remains so afterward).
1745          *
1746          * also, since the most recent successful call is the one that did
1747          * the actual unfreeze, we can use this to provide an accurate count
1748          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1749          * may * be useful for determining whether a filesystem was unfrozen
1750          * during the freeze/thaw phase by a process other than qemu-ga.
1751          */
1752         do {
1753             ret = ioctl(fd, FITHAW);
1754             if (ret == 0 && !logged) {
1755                 i++;
1756                 logged = true;
1757             }
1758         } while (ret == 0);
1759         close(fd);
1760     }
1761
1762     ga_unset_frozen(ga_state);
1763     free_fs_mount_list(&mounts);
1764
1765     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1766
1767     return i;
1768 }
1769
1770 static void guest_fsfreeze_cleanup(void)
1771 {
1772     Error *err = NULL;
1773
1774     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1775         qmp_guest_fsfreeze_thaw(&err);
1776         if (err) {
1777             slog("failed to clean up frozen filesystems: %s",
1778                  error_get_pretty(err));
1779             error_free(err);
1780         }
1781     }
1782 }
1783 #endif /* CONFIG_FSFREEZE */
1784
1785 #if defined(CONFIG_FSTRIM)
1786 /*
1787  * Walk list of mounted file systems in the guest, and trim them.
1788  */
1789 GuestFilesystemTrimResponse *
1790 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1791 {
1792     GuestFilesystemTrimResponse *response;
1793     GuestFilesystemTrimResult *result;
1794     int ret = 0;
1795     FsMountList mounts;
1796     struct FsMount *mount;
1797     int fd;
1798     Error *local_err = NULL;
1799     struct fstrim_range r;
1800
1801     slog("guest-fstrim called");
1802
1803     QTAILQ_INIT(&mounts);
1804     build_fs_mount_list(&mounts, &local_err);
1805     if (local_err) {
1806         error_propagate(errp, local_err);
1807         return NULL;
1808     }
1809
1810     response = g_malloc0(sizeof(*response));
1811
1812     QTAILQ_FOREACH(mount, &mounts, next) {
1813         result = g_malloc0(sizeof(*result));
1814         result->path = g_strdup(mount->dirname);
1815
1816         QAPI_LIST_PREPEND(response->paths, result);
1817
1818         fd = qemu_open_old(mount->dirname, O_RDONLY);
1819         if (fd == -1) {
1820             result->error = g_strdup_printf("failed to open: %s",
1821                                             strerror(errno));
1822             result->has_error = true;
1823             continue;
1824         }
1825
1826         /* We try to cull filesystems we know won't work in advance, but other
1827          * filesystems may not implement fstrim for less obvious reasons.
1828          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1829          * will be reported (e.g. CD-ROMs).
1830          * Any other error means an unexpected error.
1831          */
1832         r.start = 0;
1833         r.len = -1;
1834         r.minlen = has_minimum ? minimum : 0;
1835         ret = ioctl(fd, FITRIM, &r);
1836         if (ret == -1) {
1837             result->has_error = true;
1838             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1839                 result->error = g_strdup("trim not supported");
1840             } else {
1841                 result->error = g_strdup_printf("failed to trim: %s",
1842                                                 strerror(errno));
1843             }
1844             close(fd);
1845             continue;
1846         }
1847
1848         result->has_minimum = true;
1849         result->minimum = r.minlen;
1850         result->has_trimmed = true;
1851         result->trimmed = r.len;
1852         close(fd);
1853     }
1854
1855     free_fs_mount_list(&mounts);
1856     return response;
1857 }
1858 #endif /* CONFIG_FSTRIM */
1859
1860
1861 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1862 #define SUSPEND_SUPPORTED 0
1863 #define SUSPEND_NOT_SUPPORTED 1
1864
1865 typedef enum {
1866     SUSPEND_MODE_DISK = 0,
1867     SUSPEND_MODE_RAM = 1,
1868     SUSPEND_MODE_HYBRID = 2,
1869 } SuspendMode;
1870
1871 /*
1872  * Executes a command in a child process using g_spawn_sync,
1873  * returning an int >= 0 representing the exit status of the
1874  * process.
1875  *
1876  * If the program wasn't found in path, returns -1.
1877  *
1878  * If a problem happened when creating the child process,
1879  * returns -1 and errp is set.
1880  */
1881 static int run_process_child(const char *command[], Error **errp)
1882 {
1883     int exit_status, spawn_flag;
1884     GError *g_err = NULL;
1885     bool success;
1886
1887     spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1888                  G_SPAWN_STDERR_TO_DEV_NULL;
1889
1890     success =  g_spawn_sync(NULL, (char **)command, environ, spawn_flag,
1891                             NULL, NULL, NULL, NULL,
1892                             &exit_status, &g_err);
1893
1894     if (success) {
1895         return WEXITSTATUS(exit_status);
1896     }
1897
1898     if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1899         error_setg(errp, "failed to create child process, error '%s'",
1900                    g_err->message);
1901     }
1902
1903     g_error_free(g_err);
1904     return -1;
1905 }
1906
1907 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1908 {
1909     const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1910                                      "systemd-hybrid-sleep"};
1911     const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1912     int status;
1913
1914     status = run_process_child(cmd, errp);
1915
1916     /*
1917      * systemctl status uses LSB return codes so we can expect
1918      * status > 0 and be ok. To assert if the guest has support
1919      * for the selected suspend mode, status should be < 4. 4 is
1920      * the code for unknown service status, the return value when
1921      * the service does not exist. A common value is status = 3
1922      * (program is not running).
1923      */
1924     if (status > 0 && status < 4) {
1925         return true;
1926     }
1927
1928     return false;
1929 }
1930
1931 static void systemd_suspend(SuspendMode mode, Error **errp)
1932 {
1933     Error *local_err = NULL;
1934     const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1935     const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1936     int status;
1937
1938     status = run_process_child(cmd, &local_err);
1939
1940     if (status == 0) {
1941         return;
1942     }
1943
1944     if ((status == -1) && !local_err) {
1945         error_setg(errp, "the helper program 'systemctl %s' was not found",
1946                    systemctl_args[mode]);
1947         return;
1948     }
1949
1950     if (local_err) {
1951         error_propagate(errp, local_err);
1952     } else {
1953         error_setg(errp, "the helper program 'systemctl %s' returned an "
1954                    "unexpected exit status code (%d)",
1955                    systemctl_args[mode], status);
1956     }
1957 }
1958
1959 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1960 {
1961     Error *local_err = NULL;
1962     const char *pmutils_args[3] = {"--hibernate", "--suspend",
1963                                    "--suspend-hybrid"};
1964     const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1965     int status;
1966
1967     status = run_process_child(cmd, &local_err);
1968
1969     if (status == SUSPEND_SUPPORTED) {
1970         return true;
1971     }
1972
1973     if ((status == -1) && !local_err) {
1974         return false;
1975     }
1976
1977     if (local_err) {
1978         error_propagate(errp, local_err);
1979     } else {
1980         error_setg(errp,
1981                    "the helper program '%s' returned an unexpected exit"
1982                    " status code (%d)", "pm-is-supported", status);
1983     }
1984
1985     return false;
1986 }
1987
1988 static void pmutils_suspend(SuspendMode mode, Error **errp)
1989 {
1990     Error *local_err = NULL;
1991     const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1992                                        "pm-suspend-hybrid"};
1993     const char *cmd[2] = {pmutils_binaries[mode], NULL};
1994     int status;
1995
1996     status = run_process_child(cmd, &local_err);
1997
1998     if (status == 0) {
1999         return;
2000     }
2001
2002     if ((status == -1) && !local_err) {
2003         error_setg(errp, "the helper program '%s' was not found",
2004                    pmutils_binaries[mode]);
2005         return;
2006     }
2007
2008     if (local_err) {
2009         error_propagate(errp, local_err);
2010     } else {
2011         error_setg(errp,
2012                    "the helper program '%s' returned an unexpected exit"
2013                    " status code (%d)", pmutils_binaries[mode], status);
2014     }
2015 }
2016
2017 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
2018 {
2019     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2020     const char *sysfile_str = sysfile_strs[mode];
2021     char buf[32]; /* hopefully big enough */
2022     int fd;
2023     ssize_t ret;
2024
2025     if (!sysfile_str) {
2026         error_setg(errp, "unknown guest suspend mode");
2027         return false;
2028     }
2029
2030     fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
2031     if (fd < 0) {
2032         return false;
2033     }
2034
2035     ret = read(fd, buf, sizeof(buf) - 1);
2036     close(fd);
2037     if (ret <= 0) {
2038         return false;
2039     }
2040     buf[ret] = '\0';
2041
2042     if (strstr(buf, sysfile_str)) {
2043         return true;
2044     }
2045     return false;
2046 }
2047
2048 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
2049 {
2050     Error *local_err = NULL;
2051     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2052     const char *sysfile_str = sysfile_strs[mode];
2053     pid_t pid;
2054     int status;
2055
2056     if (!sysfile_str) {
2057         error_setg(errp, "unknown guest suspend mode");
2058         return;
2059     }
2060
2061     pid = fork();
2062     if (!pid) {
2063         /* child */
2064         int fd;
2065
2066         setsid();
2067         reopen_fd_to_null(0);
2068         reopen_fd_to_null(1);
2069         reopen_fd_to_null(2);
2070
2071         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
2072         if (fd < 0) {
2073             _exit(EXIT_FAILURE);
2074         }
2075
2076         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
2077             _exit(EXIT_FAILURE);
2078         }
2079
2080         _exit(EXIT_SUCCESS);
2081     } else if (pid < 0) {
2082         error_setg_errno(errp, errno, "failed to create child process");
2083         return;
2084     }
2085
2086     ga_wait_child(pid, &status, &local_err);
2087     if (local_err) {
2088         error_propagate(errp, local_err);
2089         return;
2090     }
2091
2092     if (WEXITSTATUS(status)) {
2093         error_setg(errp, "child process has failed to suspend");
2094     }
2095
2096 }
2097
2098 static void guest_suspend(SuspendMode mode, Error **errp)
2099 {
2100     Error *local_err = NULL;
2101     bool mode_supported = false;
2102
2103     if (systemd_supports_mode(mode, &local_err)) {
2104         mode_supported = true;
2105         systemd_suspend(mode, &local_err);
2106     }
2107
2108     if (!local_err) {
2109         return;
2110     }
2111
2112     error_free(local_err);
2113     local_err = NULL;
2114
2115     if (pmutils_supports_mode(mode, &local_err)) {
2116         mode_supported = true;
2117         pmutils_suspend(mode, &local_err);
2118     }
2119
2120     if (!local_err) {
2121         return;
2122     }
2123
2124     error_free(local_err);
2125     local_err = NULL;
2126
2127     if (linux_sys_state_supports_mode(mode, &local_err)) {
2128         mode_supported = true;
2129         linux_sys_state_suspend(mode, &local_err);
2130     }
2131
2132     if (!mode_supported) {
2133         error_free(local_err);
2134         error_setg(errp,
2135                    "the requested suspend mode is not supported by the guest");
2136     } else {
2137         error_propagate(errp, local_err);
2138     }
2139 }
2140
2141 void qmp_guest_suspend_disk(Error **errp)
2142 {
2143     guest_suspend(SUSPEND_MODE_DISK, errp);
2144 }
2145
2146 void qmp_guest_suspend_ram(Error **errp)
2147 {
2148     guest_suspend(SUSPEND_MODE_RAM, errp);
2149 }
2150
2151 void qmp_guest_suspend_hybrid(Error **errp)
2152 {
2153     guest_suspend(SUSPEND_MODE_HYBRID, errp);
2154 }
2155
2156 static GuestNetworkInterface *
2157 guest_find_interface(GuestNetworkInterfaceList *head,
2158                      const char *name)
2159 {
2160     for (; head; head = head->next) {
2161         if (strcmp(head->value->name, name) == 0) {
2162             return head->value;
2163         }
2164     }
2165
2166     return NULL;
2167 }
2168
2169 static int guest_get_network_stats(const char *name,
2170                        GuestNetworkInterfaceStat *stats)
2171 {
2172     int name_len;
2173     char const *devinfo = "/proc/net/dev";
2174     FILE *fp;
2175     char *line = NULL, *colon;
2176     size_t n = 0;
2177     fp = fopen(devinfo, "r");
2178     if (!fp) {
2179         return -1;
2180     }
2181     name_len = strlen(name);
2182     while (getline(&line, &n, fp) != -1) {
2183         long long dummy;
2184         long long rx_bytes;
2185         long long rx_packets;
2186         long long rx_errs;
2187         long long rx_dropped;
2188         long long tx_bytes;
2189         long long tx_packets;
2190         long long tx_errs;
2191         long long tx_dropped;
2192         char *trim_line;
2193         trim_line = g_strchug(line);
2194         if (trim_line[0] == '\0') {
2195             continue;
2196         }
2197         colon = strchr(trim_line, ':');
2198         if (!colon) {
2199             continue;
2200         }
2201         if (colon - name_len  == trim_line &&
2202            strncmp(trim_line, name, name_len) == 0) {
2203             if (sscanf(colon + 1,
2204                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
2205                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
2206                   &dummy, &dummy, &dummy, &dummy,
2207                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
2208                   &dummy, &dummy, &dummy, &dummy) != 16) {
2209                 continue;
2210             }
2211             stats->rx_bytes = rx_bytes;
2212             stats->rx_packets = rx_packets;
2213             stats->rx_errs = rx_errs;
2214             stats->rx_dropped = rx_dropped;
2215             stats->tx_bytes = tx_bytes;
2216             stats->tx_packets = tx_packets;
2217             stats->tx_errs = tx_errs;
2218             stats->tx_dropped = tx_dropped;
2219             fclose(fp);
2220             g_free(line);
2221             return 0;
2222         }
2223     }
2224     fclose(fp);
2225     g_free(line);
2226     g_debug("/proc/net/dev: Interface '%s' not found", name);
2227     return -1;
2228 }
2229
2230 /*
2231  * Build information about guest interfaces
2232  */
2233 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2234 {
2235     GuestNetworkInterfaceList *head = NULL, **tail = &head;
2236     struct ifaddrs *ifap, *ifa;
2237
2238     if (getifaddrs(&ifap) < 0) {
2239         error_setg_errno(errp, errno, "getifaddrs failed");
2240         goto error;
2241     }
2242
2243     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
2244         GuestNetworkInterface *info;
2245         GuestIpAddressList **address_tail;
2246         GuestIpAddress *address_item = NULL;
2247         GuestNetworkInterfaceStat *interface_stat = NULL;
2248         char addr4[INET_ADDRSTRLEN];
2249         char addr6[INET6_ADDRSTRLEN];
2250         int sock;
2251         struct ifreq ifr;
2252         unsigned char *mac_addr;
2253         void *p;
2254
2255         g_debug("Processing %s interface", ifa->ifa_name);
2256
2257         info = guest_find_interface(head, ifa->ifa_name);
2258
2259         if (!info) {
2260             info = g_malloc0(sizeof(*info));
2261             info->name = g_strdup(ifa->ifa_name);
2262
2263             QAPI_LIST_APPEND(tail, info);
2264         }
2265
2266         if (!info->has_hardware_address && ifa->ifa_flags & SIOCGIFHWADDR) {
2267             /* we haven't obtained HW address yet */
2268             sock = socket(PF_INET, SOCK_STREAM, 0);
2269             if (sock == -1) {
2270                 error_setg_errno(errp, errno, "failed to create socket");
2271                 goto error;
2272             }
2273
2274             memset(&ifr, 0, sizeof(ifr));
2275             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->name);
2276             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
2277                 error_setg_errno(errp, errno,
2278                                  "failed to get MAC address of %s",
2279                                  ifa->ifa_name);
2280                 close(sock);
2281                 goto error;
2282             }
2283
2284             close(sock);
2285             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
2286
2287             info->hardware_address =
2288                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
2289                                 (int) mac_addr[0], (int) mac_addr[1],
2290                                 (int) mac_addr[2], (int) mac_addr[3],
2291                                 (int) mac_addr[4], (int) mac_addr[5]);
2292
2293             info->has_hardware_address = true;
2294         }
2295
2296         if (ifa->ifa_addr &&
2297             ifa->ifa_addr->sa_family == AF_INET) {
2298             /* interface with IPv4 address */
2299             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
2300             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
2301                 error_setg_errno(errp, errno, "inet_ntop failed");
2302                 goto error;
2303             }
2304
2305             address_item = g_malloc0(sizeof(*address_item));
2306             address_item->ip_address = g_strdup(addr4);
2307             address_item->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
2308
2309             if (ifa->ifa_netmask) {
2310                 /* Count the number of set bits in netmask.
2311                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2312                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
2313                 address_item->prefix = ctpop32(((uint32_t *) p)[0]);
2314             }
2315         } else if (ifa->ifa_addr &&
2316                    ifa->ifa_addr->sa_family == AF_INET6) {
2317             /* interface with IPv6 address */
2318             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
2319             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
2320                 error_setg_errno(errp, errno, "inet_ntop failed");
2321                 goto error;
2322             }
2323
2324             address_item = g_malloc0(sizeof(*address_item));
2325             address_item->ip_address = g_strdup(addr6);
2326             address_item->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
2327
2328             if (ifa->ifa_netmask) {
2329                 /* Count the number of set bits in netmask.
2330                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2331                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
2332                 address_item->prefix =
2333                     ctpop32(((uint32_t *) p)[0]) +
2334                     ctpop32(((uint32_t *) p)[1]) +
2335                     ctpop32(((uint32_t *) p)[2]) +
2336                     ctpop32(((uint32_t *) p)[3]);
2337             }
2338         }
2339
2340         if (!address_item) {
2341             continue;
2342         }
2343
2344         address_tail = &info->ip_addresses;
2345         while (*address_tail) {
2346             address_tail = &(*address_tail)->next;
2347         }
2348         QAPI_LIST_APPEND(address_tail, address_item);
2349
2350         info->has_ip_addresses = true;
2351
2352         if (!info->has_statistics) {
2353             interface_stat = g_malloc0(sizeof(*interface_stat));
2354             if (guest_get_network_stats(info->name, interface_stat) == -1) {
2355                 info->has_statistics = false;
2356                 g_free(interface_stat);
2357             } else {
2358                 info->statistics = interface_stat;
2359                 info->has_statistics = true;
2360             }
2361         }
2362     }
2363
2364     freeifaddrs(ifap);
2365     return head;
2366
2367 error:
2368     freeifaddrs(ifap);
2369     qapi_free_GuestNetworkInterfaceList(head);
2370     return NULL;
2371 }
2372
2373 /* Transfer online/offline status between @vcpu and the guest system.
2374  *
2375  * On input either @errp or *@errp must be NULL.
2376  *
2377  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
2378  * - R: vcpu->logical_id
2379  * - W: vcpu->online
2380  * - W: vcpu->can_offline
2381  *
2382  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
2383  * - R: vcpu->logical_id
2384  * - R: vcpu->online
2385  *
2386  * Written members remain unmodified on error.
2387  */
2388 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
2389                           char *dirpath, Error **errp)
2390 {
2391     int fd;
2392     int res;
2393     int dirfd;
2394     static const char fn[] = "online";
2395
2396     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2397     if (dirfd == -1) {
2398         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2399         return;
2400     }
2401
2402     fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2403     if (fd == -1) {
2404         if (errno != ENOENT) {
2405             error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2406         } else if (sys2vcpu) {
2407             vcpu->online = true;
2408             vcpu->can_offline = false;
2409         } else if (!vcpu->online) {
2410             error_setg(errp, "logical processor #%" PRId64 " can't be "
2411                        "offlined", vcpu->logical_id);
2412         } /* otherwise pretend successful re-onlining */
2413     } else {
2414         unsigned char status;
2415
2416         res = pread(fd, &status, 1, 0);
2417         if (res == -1) {
2418             error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2419         } else if (res == 0) {
2420             error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2421                        fn);
2422         } else if (sys2vcpu) {
2423             vcpu->online = (status != '0');
2424             vcpu->can_offline = true;
2425         } else if (vcpu->online != (status != '0')) {
2426             status = '0' + vcpu->online;
2427             if (pwrite(fd, &status, 1, 0) == -1) {
2428                 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2429                                  fn);
2430             }
2431         } /* otherwise pretend successful re-(on|off)-lining */
2432
2433         res = close(fd);
2434         g_assert(res == 0);
2435     }
2436
2437     res = close(dirfd);
2438     g_assert(res == 0);
2439 }
2440
2441 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2442 {
2443     GuestLogicalProcessorList *head, **tail;
2444     const char *cpu_dir = "/sys/devices/system/cpu";
2445     const gchar *line;
2446     g_autoptr(GDir) cpu_gdir = NULL;
2447     Error *local_err = NULL;
2448
2449     head = NULL;
2450     tail = &head;
2451     cpu_gdir = g_dir_open(cpu_dir, 0, NULL);
2452
2453     if (cpu_gdir == NULL) {
2454         error_setg_errno(errp, errno, "failed to list entries: %s", cpu_dir);
2455         return NULL;
2456     }
2457
2458     while (local_err == NULL && (line = g_dir_read_name(cpu_gdir)) != NULL) {
2459         GuestLogicalProcessor *vcpu;
2460         int64_t id;
2461         if (sscanf(line, "cpu%" PRId64, &id)) {
2462             g_autofree char *path = g_strdup_printf("/sys/devices/system/cpu/"
2463                                                     "cpu%" PRId64 "/", id);
2464             vcpu = g_malloc0(sizeof *vcpu);
2465             vcpu->logical_id = id;
2466             vcpu->has_can_offline = true; /* lolspeak ftw */
2467             transfer_vcpu(vcpu, true, path, &local_err);
2468             QAPI_LIST_APPEND(tail, vcpu);
2469         }
2470     }
2471
2472     if (local_err == NULL) {
2473         /* there's no guest with zero VCPUs */
2474         g_assert(head != NULL);
2475         return head;
2476     }
2477
2478     qapi_free_GuestLogicalProcessorList(head);
2479     error_propagate(errp, local_err);
2480     return NULL;
2481 }
2482
2483 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2484 {
2485     int64_t processed;
2486     Error *local_err = NULL;
2487
2488     processed = 0;
2489     while (vcpus != NULL) {
2490         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2491                                      vcpus->value->logical_id);
2492
2493         transfer_vcpu(vcpus->value, false, path, &local_err);
2494         g_free(path);
2495         if (local_err != NULL) {
2496             break;
2497         }
2498         ++processed;
2499         vcpus = vcpus->next;
2500     }
2501
2502     if (local_err != NULL) {
2503         if (processed == 0) {
2504             error_propagate(errp, local_err);
2505         } else {
2506             error_free(local_err);
2507         }
2508     }
2509
2510     return processed;
2511 }
2512
2513 void qmp_guest_set_user_password(const char *username,
2514                                  const char *password,
2515                                  bool crypted,
2516                                  Error **errp)
2517 {
2518     Error *local_err = NULL;
2519     char *passwd_path = NULL;
2520     pid_t pid;
2521     int status;
2522     int datafd[2] = { -1, -1 };
2523     char *rawpasswddata = NULL;
2524     size_t rawpasswdlen;
2525     char *chpasswddata = NULL;
2526     size_t chpasswdlen;
2527
2528     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2529     if (!rawpasswddata) {
2530         return;
2531     }
2532     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2533     rawpasswddata[rawpasswdlen] = '\0';
2534
2535     if (strchr(rawpasswddata, '\n')) {
2536         error_setg(errp, "forbidden characters in raw password");
2537         goto out;
2538     }
2539
2540     if (strchr(username, '\n') ||
2541         strchr(username, ':')) {
2542         error_setg(errp, "forbidden characters in username");
2543         goto out;
2544     }
2545
2546     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2547     chpasswdlen = strlen(chpasswddata);
2548
2549     passwd_path = g_find_program_in_path("chpasswd");
2550
2551     if (!passwd_path) {
2552         error_setg(errp, "cannot find 'passwd' program in PATH");
2553         goto out;
2554     }
2555
2556     if (pipe(datafd) < 0) {
2557         error_setg(errp, "cannot create pipe FDs");
2558         goto out;
2559     }
2560
2561     pid = fork();
2562     if (pid == 0) {
2563         close(datafd[1]);
2564         /* child */
2565         setsid();
2566         dup2(datafd[0], 0);
2567         reopen_fd_to_null(1);
2568         reopen_fd_to_null(2);
2569
2570         if (crypted) {
2571             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2572         } else {
2573             execle(passwd_path, "chpasswd", NULL, environ);
2574         }
2575         _exit(EXIT_FAILURE);
2576     } else if (pid < 0) {
2577         error_setg_errno(errp, errno, "failed to create child process");
2578         goto out;
2579     }
2580     close(datafd[0]);
2581     datafd[0] = -1;
2582
2583     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2584         error_setg_errno(errp, errno, "cannot write new account password");
2585         goto out;
2586     }
2587     close(datafd[1]);
2588     datafd[1] = -1;
2589
2590     ga_wait_child(pid, &status, &local_err);
2591     if (local_err) {
2592         error_propagate(errp, local_err);
2593         goto out;
2594     }
2595
2596     if (!WIFEXITED(status)) {
2597         error_setg(errp, "child process has terminated abnormally");
2598         goto out;
2599     }
2600
2601     if (WEXITSTATUS(status)) {
2602         error_setg(errp, "child process has failed to set user password");
2603         goto out;
2604     }
2605
2606 out:
2607     g_free(chpasswddata);
2608     g_free(rawpasswddata);
2609     g_free(passwd_path);
2610     if (datafd[0] != -1) {
2611         close(datafd[0]);
2612     }
2613     if (datafd[1] != -1) {
2614         close(datafd[1]);
2615     }
2616 }
2617
2618 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2619                                int size, Error **errp)
2620 {
2621     int fd;
2622     int res;
2623
2624     errno = 0;
2625     fd = openat(dirfd, pathname, O_RDONLY);
2626     if (fd == -1) {
2627         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2628         return;
2629     }
2630
2631     res = pread(fd, buf, size, 0);
2632     if (res == -1) {
2633         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2634     } else if (res == 0) {
2635         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2636     }
2637     close(fd);
2638 }
2639
2640 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2641                                 const char *buf, int size, Error **errp)
2642 {
2643     int fd;
2644
2645     errno = 0;
2646     fd = openat(dirfd, pathname, O_WRONLY);
2647     if (fd == -1) {
2648         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2649         return;
2650     }
2651
2652     if (pwrite(fd, buf, size, 0) == -1) {
2653         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2654     }
2655
2656     close(fd);
2657 }
2658
2659 /* Transfer online/offline status between @mem_blk and the guest system.
2660  *
2661  * On input either @errp or *@errp must be NULL.
2662  *
2663  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2664  * - R: mem_blk->phys_index
2665  * - W: mem_blk->online
2666  * - W: mem_blk->can_offline
2667  *
2668  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2669  * - R: mem_blk->phys_index
2670  * - R: mem_blk->online
2671  *-  R: mem_blk->can_offline
2672  * Written members remain unmodified on error.
2673  */
2674 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2675                                   GuestMemoryBlockResponse *result,
2676                                   Error **errp)
2677 {
2678     char *dirpath;
2679     int dirfd;
2680     char *status;
2681     Error *local_err = NULL;
2682
2683     if (!sys2memblk) {
2684         DIR *dp;
2685
2686         if (!result) {
2687             error_setg(errp, "Internal error, 'result' should not be NULL");
2688             return;
2689         }
2690         errno = 0;
2691         dp = opendir("/sys/devices/system/memory/");
2692          /* if there is no 'memory' directory in sysfs,
2693          * we think this VM does not support online/offline memory block,
2694          * any other solution?
2695          */
2696         if (!dp) {
2697             if (errno == ENOENT) {
2698                 result->response =
2699                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2700             }
2701             goto out1;
2702         }
2703         closedir(dp);
2704     }
2705
2706     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2707                               mem_blk->phys_index);
2708     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2709     if (dirfd == -1) {
2710         if (sys2memblk) {
2711             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2712         } else {
2713             if (errno == ENOENT) {
2714                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2715             } else {
2716                 result->response =
2717                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2718             }
2719         }
2720         g_free(dirpath);
2721         goto out1;
2722     }
2723     g_free(dirpath);
2724
2725     status = g_malloc0(10);
2726     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2727     if (local_err) {
2728         /* treat with sysfs file that not exist in old kernel */
2729         if (errno == ENOENT) {
2730             error_free(local_err);
2731             if (sys2memblk) {
2732                 mem_blk->online = true;
2733                 mem_blk->can_offline = false;
2734             } else if (!mem_blk->online) {
2735                 result->response =
2736                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2737             }
2738         } else {
2739             if (sys2memblk) {
2740                 error_propagate(errp, local_err);
2741             } else {
2742                 error_free(local_err);
2743                 result->response =
2744                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2745             }
2746         }
2747         goto out2;
2748     }
2749
2750     if (sys2memblk) {
2751         char removable = '0';
2752
2753         mem_blk->online = (strncmp(status, "online", 6) == 0);
2754
2755         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2756         if (local_err) {
2757             /* if no 'removable' file, it doesn't support offline mem blk */
2758             if (errno == ENOENT) {
2759                 error_free(local_err);
2760                 mem_blk->can_offline = false;
2761             } else {
2762                 error_propagate(errp, local_err);
2763             }
2764         } else {
2765             mem_blk->can_offline = (removable != '0');
2766         }
2767     } else {
2768         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2769             const char *new_state = mem_blk->online ? "online" : "offline";
2770
2771             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2772                                 &local_err);
2773             if (local_err) {
2774                 error_free(local_err);
2775                 result->response =
2776                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2777                 goto out2;
2778             }
2779
2780             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2781             result->has_error_code = false;
2782         } /* otherwise pretend successful re-(on|off)-lining */
2783     }
2784     g_free(status);
2785     close(dirfd);
2786     return;
2787
2788 out2:
2789     g_free(status);
2790     close(dirfd);
2791 out1:
2792     if (!sys2memblk) {
2793         result->has_error_code = true;
2794         result->error_code = errno;
2795     }
2796 }
2797
2798 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2799 {
2800     GuestMemoryBlockList *head, **tail;
2801     Error *local_err = NULL;
2802     struct dirent *de;
2803     DIR *dp;
2804
2805     head = NULL;
2806     tail = &head;
2807
2808     dp = opendir("/sys/devices/system/memory/");
2809     if (!dp) {
2810         /* it's ok if this happens to be a system that doesn't expose
2811          * memory blocks via sysfs, but otherwise we should report
2812          * an error
2813          */
2814         if (errno != ENOENT) {
2815             error_setg_errno(errp, errno, "Can't open directory"
2816                              "\"/sys/devices/system/memory/\"");
2817         }
2818         return NULL;
2819     }
2820
2821     /* Note: the phys_index of memory block may be discontinuous,
2822      * this is because a memblk is the unit of the Sparse Memory design, which
2823      * allows discontinuous memory ranges (ex. NUMA), so here we should
2824      * traverse the memory block directory.
2825      */
2826     while ((de = readdir(dp)) != NULL) {
2827         GuestMemoryBlock *mem_blk;
2828
2829         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2830             !(de->d_type & DT_DIR)) {
2831             continue;
2832         }
2833
2834         mem_blk = g_malloc0(sizeof *mem_blk);
2835         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2836         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2837         mem_blk->has_can_offline = true; /* lolspeak ftw */
2838         transfer_memory_block(mem_blk, true, NULL, &local_err);
2839         if (local_err) {
2840             break;
2841         }
2842
2843         QAPI_LIST_APPEND(tail, mem_blk);
2844     }
2845
2846     closedir(dp);
2847     if (local_err == NULL) {
2848         /* there's no guest with zero memory blocks */
2849         if (head == NULL) {
2850             error_setg(errp, "guest reported zero memory blocks!");
2851         }
2852         return head;
2853     }
2854
2855     qapi_free_GuestMemoryBlockList(head);
2856     error_propagate(errp, local_err);
2857     return NULL;
2858 }
2859
2860 GuestMemoryBlockResponseList *
2861 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2862 {
2863     GuestMemoryBlockResponseList *head, **tail;
2864     Error *local_err = NULL;
2865
2866     head = NULL;
2867     tail = &head;
2868
2869     while (mem_blks != NULL) {
2870         GuestMemoryBlockResponse *result;
2871         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2872
2873         result = g_malloc0(sizeof(*result));
2874         result->phys_index = current_mem_blk->phys_index;
2875         transfer_memory_block(current_mem_blk, false, result, &local_err);
2876         if (local_err) { /* should never happen */
2877             goto err;
2878         }
2879
2880         QAPI_LIST_APPEND(tail, result);
2881         mem_blks = mem_blks->next;
2882     }
2883
2884     return head;
2885 err:
2886     qapi_free_GuestMemoryBlockResponseList(head);
2887     error_propagate(errp, local_err);
2888     return NULL;
2889 }
2890
2891 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2892 {
2893     Error *local_err = NULL;
2894     char *dirpath;
2895     int dirfd;
2896     char *buf;
2897     GuestMemoryBlockInfo *info;
2898
2899     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2900     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2901     if (dirfd == -1) {
2902         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2903         g_free(dirpath);
2904         return NULL;
2905     }
2906     g_free(dirpath);
2907
2908     buf = g_malloc0(20);
2909     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2910     close(dirfd);
2911     if (local_err) {
2912         g_free(buf);
2913         error_propagate(errp, local_err);
2914         return NULL;
2915     }
2916
2917     info = g_new0(GuestMemoryBlockInfo, 1);
2918     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2919
2920     g_free(buf);
2921
2922     return info;
2923 }
2924
2925 #else /* defined(__linux__) */
2926
2927 void qmp_guest_suspend_disk(Error **errp)
2928 {
2929     error_setg(errp, QERR_UNSUPPORTED);
2930 }
2931
2932 void qmp_guest_suspend_ram(Error **errp)
2933 {
2934     error_setg(errp, QERR_UNSUPPORTED);
2935 }
2936
2937 void qmp_guest_suspend_hybrid(Error **errp)
2938 {
2939     error_setg(errp, QERR_UNSUPPORTED);
2940 }
2941
2942 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2943 {
2944     error_setg(errp, QERR_UNSUPPORTED);
2945     return NULL;
2946 }
2947
2948 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2949 {
2950     error_setg(errp, QERR_UNSUPPORTED);
2951     return NULL;
2952 }
2953
2954 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2955 {
2956     error_setg(errp, QERR_UNSUPPORTED);
2957     return -1;
2958 }
2959
2960 void qmp_guest_set_user_password(const char *username,
2961                                  const char *password,
2962                                  bool crypted,
2963                                  Error **errp)
2964 {
2965     error_setg(errp, QERR_UNSUPPORTED);
2966 }
2967
2968 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2969 {
2970     error_setg(errp, QERR_UNSUPPORTED);
2971     return NULL;
2972 }
2973
2974 GuestMemoryBlockResponseList *
2975 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2976 {
2977     error_setg(errp, QERR_UNSUPPORTED);
2978     return NULL;
2979 }
2980
2981 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2982 {
2983     error_setg(errp, QERR_UNSUPPORTED);
2984     return NULL;
2985 }
2986
2987 #endif
2988
2989 #if !defined(CONFIG_FSFREEZE)
2990
2991 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2992 {
2993     error_setg(errp, QERR_UNSUPPORTED);
2994     return NULL;
2995 }
2996
2997 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2998 {
2999     error_setg(errp, QERR_UNSUPPORTED);
3000
3001     return 0;
3002 }
3003
3004 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
3005 {
3006     error_setg(errp, QERR_UNSUPPORTED);
3007
3008     return 0;
3009 }
3010
3011 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
3012                                        strList *mountpoints,
3013                                        Error **errp)
3014 {
3015     error_setg(errp, QERR_UNSUPPORTED);
3016
3017     return 0;
3018 }
3019
3020 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
3021 {
3022     error_setg(errp, QERR_UNSUPPORTED);
3023
3024     return 0;
3025 }
3026
3027 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
3028 {
3029     error_setg(errp, QERR_UNSUPPORTED);
3030     return NULL;
3031 }
3032
3033 #endif /* CONFIG_FSFREEZE */
3034
3035 #if !defined(CONFIG_FSTRIM)
3036 GuestFilesystemTrimResponse *
3037 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
3038 {
3039     error_setg(errp, QERR_UNSUPPORTED);
3040     return NULL;
3041 }
3042 #endif
3043
3044 /* add unsupported commands to the blacklist */
3045 GList *ga_command_blacklist_init(GList *blacklist)
3046 {
3047 #if !defined(__linux__)
3048     {
3049         const char *list[] = {
3050             "guest-suspend-disk", "guest-suspend-ram",
3051             "guest-suspend-hybrid", "guest-network-get-interfaces",
3052             "guest-get-vcpus", "guest-set-vcpus",
3053             "guest-get-memory-blocks", "guest-set-memory-blocks",
3054             "guest-get-memory-block-size", "guest-get-memory-block-info",
3055             NULL};
3056         char **p = (char **)list;
3057
3058         while (*p) {
3059             blacklist = g_list_append(blacklist, g_strdup(*p++));
3060         }
3061     }
3062 #endif
3063
3064 #if !defined(CONFIG_FSFREEZE)
3065     {
3066         const char *list[] = {
3067             "guest-get-fsinfo", "guest-fsfreeze-status",
3068             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
3069             "guest-fsfreeze-thaw", "guest-get-fsinfo",
3070             "guest-get-disks", NULL};
3071         char **p = (char **)list;
3072
3073         while (*p) {
3074             blacklist = g_list_append(blacklist, g_strdup(*p++));
3075         }
3076     }
3077 #endif
3078
3079 #if !defined(CONFIG_FSTRIM)
3080     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
3081 #endif
3082
3083     blacklist = g_list_append(blacklist, g_strdup("guest-get-devices"));
3084
3085     return blacklist;
3086 }
3087
3088 /* register init/cleanup routines for stateful command groups */
3089 void ga_command_state_init(GAState *s, GACommandState *cs)
3090 {
3091 #if defined(CONFIG_FSFREEZE)
3092     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
3093 #endif
3094 }
3095
3096 #ifdef HAVE_UTMPX
3097
3098 #define QGA_MICRO_SECOND_TO_SECOND 1000000
3099
3100 static double ga_get_login_time(struct utmpx *user_info)
3101 {
3102     double seconds = (double)user_info->ut_tv.tv_sec;
3103     double useconds = (double)user_info->ut_tv.tv_usec;
3104     useconds /= QGA_MICRO_SECOND_TO_SECOND;
3105     return seconds + useconds;
3106 }
3107
3108 GuestUserList *qmp_guest_get_users(Error **errp)
3109 {
3110     GHashTable *cache = NULL;
3111     GuestUserList *head = NULL, **tail = &head;
3112     struct utmpx *user_info = NULL;
3113     gpointer value = NULL;
3114     GuestUser *user = NULL;
3115     double login_time = 0;
3116
3117     cache = g_hash_table_new(g_str_hash, g_str_equal);
3118     setutxent();
3119
3120     for (;;) {
3121         user_info = getutxent();
3122         if (user_info == NULL) {
3123             break;
3124         } else if (user_info->ut_type != USER_PROCESS) {
3125             continue;
3126         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
3127             value = g_hash_table_lookup(cache, user_info->ut_user);
3128             user = (GuestUser *)value;
3129             login_time = ga_get_login_time(user_info);
3130             /* We're ensuring the earliest login time to be sent */
3131             if (login_time < user->login_time) {
3132                 user->login_time = login_time;
3133             }
3134             continue;
3135         }
3136
3137         user = g_new0(GuestUser, 1);
3138         user->user = g_strdup(user_info->ut_user);
3139         user->login_time = ga_get_login_time(user_info);
3140
3141         g_hash_table_insert(cache, user->user, user);
3142
3143         QAPI_LIST_APPEND(tail, user);
3144     }
3145     endutxent();
3146     g_hash_table_destroy(cache);
3147     return head;
3148 }
3149
3150 #else
3151
3152 GuestUserList *qmp_guest_get_users(Error **errp)
3153 {
3154     error_setg(errp, QERR_UNSUPPORTED);
3155     return NULL;
3156 }
3157
3158 #endif
3159
3160 /* Replace escaped special characters with theire real values. The replacement
3161  * is done in place -- returned value is in the original string.
3162  */
3163 static void ga_osrelease_replace_special(gchar *value)
3164 {
3165     gchar *p, *p2, quote;
3166
3167     /* Trim the string at first space or semicolon if it is not enclosed in
3168      * single or double quotes. */
3169     if ((value[0] != '"') || (value[0] == '\'')) {
3170         p = strchr(value, ' ');
3171         if (p != NULL) {
3172             *p = 0;
3173         }
3174         p = strchr(value, ';');
3175         if (p != NULL) {
3176             *p = 0;
3177         }
3178         return;
3179     }
3180
3181     quote = value[0];
3182     p2 = value;
3183     p = value + 1;
3184     while (*p != 0) {
3185         if (*p == '\\') {
3186             p++;
3187             switch (*p) {
3188             case '$':
3189             case '\'':
3190             case '"':
3191             case '\\':
3192             case '`':
3193                 break;
3194             default:
3195                 /* Keep literal backslash followed by whatever is there */
3196                 p--;
3197                 break;
3198             }
3199         } else if (*p == quote) {
3200             *p2 = 0;
3201             break;
3202         }
3203         *(p2++) = *(p++);
3204     }
3205 }
3206
3207 static GKeyFile *ga_parse_osrelease(const char *fname)
3208 {
3209     gchar *content = NULL;
3210     gchar *content2 = NULL;
3211     GError *err = NULL;
3212     GKeyFile *keys = g_key_file_new();
3213     const char *group = "[os-release]\n";
3214
3215     if (!g_file_get_contents(fname, &content, NULL, &err)) {
3216         slog("failed to read '%s', error: %s", fname, err->message);
3217         goto fail;
3218     }
3219
3220     if (!g_utf8_validate(content, -1, NULL)) {
3221         slog("file is not utf-8 encoded: %s", fname);
3222         goto fail;
3223     }
3224     content2 = g_strdup_printf("%s%s", group, content);
3225
3226     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
3227                                    &err)) {
3228         slog("failed to parse file '%s', error: %s", fname, err->message);
3229         goto fail;
3230     }
3231
3232     g_free(content);
3233     g_free(content2);
3234     return keys;
3235
3236 fail:
3237     g_error_free(err);
3238     g_free(content);
3239     g_free(content2);
3240     g_key_file_free(keys);
3241     return NULL;
3242 }
3243
3244 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
3245 {
3246     GuestOSInfo *info = NULL;
3247     struct utsname kinfo;
3248     GKeyFile *osrelease = NULL;
3249     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
3250
3251     info = g_new0(GuestOSInfo, 1);
3252
3253     if (uname(&kinfo) != 0) {
3254         error_setg_errno(errp, errno, "uname failed");
3255     } else {
3256         info->has_kernel_version = true;
3257         info->kernel_version = g_strdup(kinfo.version);
3258         info->has_kernel_release = true;
3259         info->kernel_release = g_strdup(kinfo.release);
3260         info->has_machine = true;
3261         info->machine = g_strdup(kinfo.machine);
3262     }
3263
3264     if (qga_os_release != NULL) {
3265         osrelease = ga_parse_osrelease(qga_os_release);
3266     } else {
3267         osrelease = ga_parse_osrelease("/etc/os-release");
3268         if (osrelease == NULL) {
3269             osrelease = ga_parse_osrelease("/usr/lib/os-release");
3270         }
3271     }
3272
3273     if (osrelease != NULL) {
3274         char *value;
3275
3276 #define GET_FIELD(field, osfield) do { \
3277     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
3278     if (value != NULL) { \
3279         ga_osrelease_replace_special(value); \
3280         info->has_ ## field = true; \
3281         info->field = value; \
3282     } \
3283 } while (0)
3284         GET_FIELD(id, "ID");
3285         GET_FIELD(name, "NAME");
3286         GET_FIELD(pretty_name, "PRETTY_NAME");
3287         GET_FIELD(version, "VERSION");
3288         GET_FIELD(version_id, "VERSION_ID");
3289         GET_FIELD(variant, "VARIANT");
3290         GET_FIELD(variant_id, "VARIANT_ID");
3291 #undef GET_FIELD
3292
3293         g_key_file_free(osrelease);
3294     }
3295
3296     return info;
3297 }
3298
3299 GuestDeviceInfoList *qmp_guest_get_devices(Error **errp)
3300 {
3301     error_setg(errp, QERR_UNSUPPORTED);
3302
3303     return NULL;
3304 }