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 /* Store disk device info specified by @sysfs into @fs */
1033 static void build_guest_fsinfo_for_real_device(char const *syspath,
1034                                                GuestFilesystemInfo *fs,
1035                                                Error **errp)
1036 {
1037     GuestDiskAddress *disk;
1038     GuestPCIAddress *pciaddr;
1039     GuestDiskAddressList *list = NULL;
1040     bool has_hwinf;
1041 #ifdef CONFIG_LIBUDEV
1042     struct udev *udev = NULL;
1043     struct udev_device *udevice = NULL;
1044 #endif
1045
1046     pciaddr = g_new0(GuestPCIAddress, 1);
1047     pciaddr->domain = -1;                       /* -1 means field is invalid */
1048     pciaddr->bus = -1;
1049     pciaddr->slot = -1;
1050     pciaddr->function = -1;
1051
1052     disk = g_new0(GuestDiskAddress, 1);
1053     disk->pci_controller = pciaddr;
1054     disk->bus_type = GUEST_DISK_BUS_TYPE_UNKNOWN;
1055
1056     list = g_new0(GuestDiskAddressList, 1);
1057     list->value = disk;
1058
1059 #ifdef CONFIG_LIBUDEV
1060     udev = udev_new();
1061     udevice = udev_device_new_from_syspath(udev, syspath);
1062     if (udev == NULL || udevice == NULL) {
1063         g_debug("failed to query udev");
1064     } else {
1065         const char *devnode, *serial;
1066         devnode = udev_device_get_devnode(udevice);
1067         if (devnode != NULL) {
1068             disk->dev = g_strdup(devnode);
1069             disk->has_dev = true;
1070         }
1071         serial = udev_device_get_property_value(udevice, "ID_SERIAL");
1072         if (serial != NULL && *serial != 0) {
1073             disk->serial = g_strdup(serial);
1074             disk->has_serial = true;
1075         }
1076     }
1077
1078     udev_unref(udev);
1079     udev_device_unref(udevice);
1080 #endif
1081
1082     if (strstr(syspath, "/devices/pci")) {
1083         has_hwinf = build_guest_fsinfo_for_pci_dev(syspath, disk, errp);
1084     } else if (strstr(syspath, "/virtio")) {
1085         has_hwinf = build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1086     } else {
1087         g_debug("Unsupported device type for '%s'", syspath);
1088         has_hwinf = false;
1089     }
1090
1091     if (has_hwinf || disk->has_dev || disk->has_serial) {
1092         list->next = fs->disk;
1093         fs->disk = list;
1094     } else {
1095         qapi_free_GuestDiskAddressList(list);
1096     }
1097 }
1098
1099 static void build_guest_fsinfo_for_device(char const *devpath,
1100                                           GuestFilesystemInfo *fs,
1101                                           Error **errp);
1102
1103 /* Store a list of slave devices of virtual volume specified by @syspath into
1104  * @fs */
1105 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1106                                                   GuestFilesystemInfo *fs,
1107                                                   Error **errp)
1108 {
1109     Error *err = NULL;
1110     DIR *dir;
1111     char *dirpath;
1112     struct dirent *entry;
1113
1114     dirpath = g_strdup_printf("%s/slaves", syspath);
1115     dir = opendir(dirpath);
1116     if (!dir) {
1117         if (errno != ENOENT) {
1118             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1119         }
1120         g_free(dirpath);
1121         return;
1122     }
1123
1124     for (;;) {
1125         errno = 0;
1126         entry = readdir(dir);
1127         if (entry == NULL) {
1128             if (errno) {
1129                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1130             }
1131             break;
1132         }
1133
1134         if (entry->d_type == DT_LNK) {
1135             char *path;
1136
1137             g_debug(" slave device '%s'", entry->d_name);
1138             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1139             build_guest_fsinfo_for_device(path, fs, &err);
1140             g_free(path);
1141
1142             if (err) {
1143                 error_propagate(errp, err);
1144                 break;
1145             }
1146         }
1147     }
1148
1149     g_free(dirpath);
1150     closedir(dir);
1151 }
1152
1153 static bool is_disk_virtual(const char *devpath, Error **errp)
1154 {
1155     g_autofree char *syspath = realpath(devpath, NULL);
1156
1157     if (!syspath) {
1158         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1159         return false;
1160     }
1161     return strstr(syspath, "/devices/virtual/block/") != NULL;
1162 }
1163
1164 /* Dispatch to functions for virtual/real device */
1165 static void build_guest_fsinfo_for_device(char const *devpath,
1166                                           GuestFilesystemInfo *fs,
1167                                           Error **errp)
1168 {
1169     ERRP_GUARD();
1170     g_autofree char *syspath = NULL;
1171     bool is_virtual = false;
1172
1173     syspath = realpath(devpath, NULL);
1174     if (!syspath) {
1175         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1176         return;
1177     }
1178
1179     if (!fs->name) {
1180         fs->name = g_path_get_basename(syspath);
1181     }
1182
1183     g_debug("  parse sysfs path '%s'", syspath);
1184     is_virtual = is_disk_virtual(syspath, errp);
1185     if (*errp != NULL) {
1186         return;
1187     }
1188     if (is_virtual) {
1189         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1190     } else {
1191         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1192     }
1193 }
1194
1195 #ifdef CONFIG_LIBUDEV
1196
1197 /*
1198  * Wrapper around build_guest_fsinfo_for_device() for getting just
1199  * the disk address.
1200  */
1201 static GuestDiskAddress *get_disk_address(const char *syspath, Error **errp)
1202 {
1203     g_autoptr(GuestFilesystemInfo) fs = NULL;
1204
1205     fs = g_new0(GuestFilesystemInfo, 1);
1206     build_guest_fsinfo_for_device(syspath, fs, errp);
1207     if (fs->disk != NULL) {
1208         return g_steal_pointer(&fs->disk->value);
1209     }
1210     return NULL;
1211 }
1212
1213 static char *get_alias_for_syspath(const char *syspath)
1214 {
1215     struct udev *udev = NULL;
1216     struct udev_device *udevice = NULL;
1217     char *ret = NULL;
1218
1219     udev = udev_new();
1220     if (udev == NULL) {
1221         g_debug("failed to query udev");
1222         goto out;
1223     }
1224     udevice = udev_device_new_from_syspath(udev, syspath);
1225     if (udevice == NULL) {
1226         g_debug("failed to query udev for path: %s", syspath);
1227         goto out;
1228     } else {
1229         const char *alias = udev_device_get_property_value(
1230             udevice, "DM_NAME");
1231         /*
1232          * NULL means there was an error and empty string means there is no
1233          * alias. In case of no alias we return NULL instead of empty string.
1234          */
1235         if (alias == NULL) {
1236             g_debug("failed to query udev for device alias for: %s",
1237                 syspath);
1238         } else if (*alias != 0) {
1239             ret = g_strdup(alias);
1240         }
1241     }
1242
1243 out:
1244     udev_unref(udev);
1245     udev_device_unref(udevice);
1246     return ret;
1247 }
1248
1249 static char *get_device_for_syspath(const char *syspath)
1250 {
1251     struct udev *udev = NULL;
1252     struct udev_device *udevice = NULL;
1253     char *ret = NULL;
1254
1255     udev = udev_new();
1256     if (udev == NULL) {
1257         g_debug("failed to query udev");
1258         goto out;
1259     }
1260     udevice = udev_device_new_from_syspath(udev, syspath);
1261     if (udevice == NULL) {
1262         g_debug("failed to query udev for path: %s", syspath);
1263         goto out;
1264     } else {
1265         ret = g_strdup(udev_device_get_devnode(udevice));
1266     }
1267
1268 out:
1269     udev_unref(udev);
1270     udev_device_unref(udevice);
1271     return ret;
1272 }
1273
1274 static void get_disk_deps(const char *disk_dir, GuestDiskInfo *disk)
1275 {
1276     g_autofree char *deps_dir = NULL;
1277     const gchar *dep;
1278     GDir *dp_deps = NULL;
1279
1280     /* List dependent disks */
1281     deps_dir = g_strdup_printf("%s/slaves", disk_dir);
1282     g_debug("  listing entries in: %s", deps_dir);
1283     dp_deps = g_dir_open(deps_dir, 0, NULL);
1284     if (dp_deps == NULL) {
1285         g_debug("failed to list entries in %s", deps_dir);
1286         return;
1287     }
1288     while ((dep = g_dir_read_name(dp_deps)) != NULL) {
1289         g_autofree char *dep_dir = NULL;
1290         strList *dep_item = NULL;
1291         char *dev_name;
1292
1293         /* Add dependent disks */
1294         dep_dir = g_strdup_printf("%s/%s", deps_dir, dep);
1295         dev_name = get_device_for_syspath(dep_dir);
1296         if (dev_name != NULL) {
1297             g_debug("  adding dependent device: %s", dev_name);
1298             dep_item = g_new0(strList, 1);
1299             dep_item->value = dev_name;
1300             dep_item->next = disk->dependents;
1301             disk->dependents = dep_item;
1302         }
1303     }
1304     g_dir_close(dp_deps);
1305 }
1306
1307 /*
1308  * Detect partitions subdirectory, name is "<disk_name><number>" or
1309  * "<disk_name>p<number>"
1310  *
1311  * @disk_name -- last component of /sys path (e.g. sda)
1312  * @disk_dir -- sys path of the disk (e.g. /sys/block/sda)
1313  * @disk_dev -- device node of the disk (e.g. /dev/sda)
1314  */
1315 static GuestDiskInfoList *get_disk_partitions(
1316     GuestDiskInfoList *list,
1317     const char *disk_name, const char *disk_dir,
1318     const char *disk_dev)
1319 {
1320     GuestDiskInfoList *item, *ret = list;
1321     struct dirent *de_disk;
1322     DIR *dp_disk = NULL;
1323     size_t len = strlen(disk_name);
1324
1325     dp_disk = opendir(disk_dir);
1326     while ((de_disk = readdir(dp_disk)) != NULL) {
1327         g_autofree char *partition_dir = NULL;
1328         char *dev_name;
1329         GuestDiskInfo *partition;
1330
1331         if (!(de_disk->d_type & DT_DIR)) {
1332             continue;
1333         }
1334
1335         if (!(strncmp(disk_name, de_disk->d_name, len) == 0 &&
1336             ((*(de_disk->d_name + len) == 'p' &&
1337             isdigit(*(de_disk->d_name + len + 1))) ||
1338                 isdigit(*(de_disk->d_name + len))))) {
1339             continue;
1340         }
1341
1342         partition_dir = g_strdup_printf("%s/%s",
1343             disk_dir, de_disk->d_name);
1344         dev_name = get_device_for_syspath(partition_dir);
1345         if (dev_name == NULL) {
1346             g_debug("Failed to get device name for syspath: %s",
1347                 disk_dir);
1348             continue;
1349         }
1350         partition = g_new0(GuestDiskInfo, 1);
1351         partition->name = dev_name;
1352         partition->partition = true;
1353         /* Add parent disk as dependent for easier tracking of hierarchy */
1354         partition->dependents = g_new0(strList, 1);
1355         partition->dependents->value = g_strdup(disk_dev);
1356
1357         item = g_new0(GuestDiskInfoList, 1);
1358         item->value = partition;
1359         item->next = ret;
1360         ret = item;
1361
1362     }
1363     closedir(dp_disk);
1364
1365     return ret;
1366 }
1367
1368 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1369 {
1370     GuestDiskInfoList *item, *ret = NULL;
1371     GuestDiskInfo *disk;
1372     DIR *dp = NULL;
1373     struct dirent *de = NULL;
1374
1375     g_debug("listing /sys/block directory");
1376     dp = opendir("/sys/block");
1377     if (dp == NULL) {
1378         error_setg_errno(errp, errno, "Can't open directory \"/sys/block\"");
1379         return NULL;
1380     }
1381     while ((de = readdir(dp)) != NULL) {
1382         g_autofree char *disk_dir = NULL, *line = NULL,
1383             *size_path = NULL;
1384         char *dev_name;
1385         Error *local_err = NULL;
1386         if (de->d_type != DT_LNK) {
1387             g_debug("  skipping entry: %s", de->d_name);
1388             continue;
1389         }
1390
1391         /* Check size and skip zero-sized disks */
1392         g_debug("  checking disk size");
1393         size_path = g_strdup_printf("/sys/block/%s/size", de->d_name);
1394         if (!g_file_get_contents(size_path, &line, NULL, NULL)) {
1395             g_debug("  failed to read disk size");
1396             continue;
1397         }
1398         if (g_strcmp0(line, "0\n") == 0) {
1399             g_debug("  skipping zero-sized disk");
1400             continue;
1401         }
1402
1403         g_debug("  adding %s", de->d_name);
1404         disk_dir = g_strdup_printf("/sys/block/%s", de->d_name);
1405         dev_name = get_device_for_syspath(disk_dir);
1406         if (dev_name == NULL) {
1407             g_debug("Failed to get device name for syspath: %s",
1408                 disk_dir);
1409             continue;
1410         }
1411         disk = g_new0(GuestDiskInfo, 1);
1412         disk->name = dev_name;
1413         disk->partition = false;
1414         disk->alias = get_alias_for_syspath(disk_dir);
1415         disk->has_alias = (disk->alias != NULL);
1416         item = g_new0(GuestDiskInfoList, 1);
1417         item->value = disk;
1418         item->next = ret;
1419         ret = item;
1420
1421         /* Get address for non-virtual devices */
1422         bool is_virtual = is_disk_virtual(disk_dir, &local_err);
1423         if (local_err != NULL) {
1424             g_debug("  failed to check disk path, ignoring error: %s",
1425                 error_get_pretty(local_err));
1426             error_free(local_err);
1427             local_err = NULL;
1428             /* Don't try to get the address */
1429             is_virtual = true;
1430         }
1431         if (!is_virtual) {
1432             disk->address = get_disk_address(disk_dir, &local_err);
1433             if (local_err != NULL) {
1434                 g_debug("  failed to get device info, ignoring error: %s",
1435                     error_get_pretty(local_err));
1436                 error_free(local_err);
1437                 local_err = NULL;
1438             } else if (disk->address != NULL) {
1439                 disk->has_address = true;
1440             }
1441         }
1442
1443         get_disk_deps(disk_dir, disk);
1444         ret = get_disk_partitions(ret, de->d_name, disk_dir, dev_name);
1445     }
1446
1447     closedir(dp);
1448
1449     return ret;
1450 }
1451
1452 #else
1453
1454 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1455 {
1456     error_setg(errp, QERR_UNSUPPORTED);
1457     return NULL;
1458 }
1459
1460 #endif
1461
1462 /* Return a list of the disk device(s)' info which @mount lies on */
1463 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1464                                                Error **errp)
1465 {
1466     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1467     struct statvfs buf;
1468     unsigned long used, nonroot_total, fr_size;
1469     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1470                                     mount->devmajor, mount->devminor);
1471
1472     fs->mountpoint = g_strdup(mount->dirname);
1473     fs->type = g_strdup(mount->devtype);
1474     build_guest_fsinfo_for_device(devpath, fs, errp);
1475
1476     if (statvfs(fs->mountpoint, &buf) == 0) {
1477         fr_size = buf.f_frsize;
1478         used = buf.f_blocks - buf.f_bfree;
1479         nonroot_total = used + buf.f_bavail;
1480         fs->used_bytes = used * fr_size;
1481         fs->total_bytes = nonroot_total * fr_size;
1482
1483         fs->has_total_bytes = true;
1484         fs->has_used_bytes = true;
1485     }
1486
1487     g_free(devpath);
1488
1489     return fs;
1490 }
1491
1492 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1493 {
1494     FsMountList mounts;
1495     struct FsMount *mount;
1496     GuestFilesystemInfoList *new, *ret = NULL;
1497     Error *local_err = NULL;
1498
1499     QTAILQ_INIT(&mounts);
1500     build_fs_mount_list(&mounts, &local_err);
1501     if (local_err) {
1502         error_propagate(errp, local_err);
1503         return NULL;
1504     }
1505
1506     QTAILQ_FOREACH(mount, &mounts, next) {
1507         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1508
1509         new = g_malloc0(sizeof(*ret));
1510         new->value = build_guest_fsinfo(mount, &local_err);
1511         new->next = ret;
1512         ret = new;
1513         if (local_err) {
1514             error_propagate(errp, local_err);
1515             qapi_free_GuestFilesystemInfoList(ret);
1516             ret = NULL;
1517             break;
1518         }
1519     }
1520
1521     free_fs_mount_list(&mounts);
1522     return ret;
1523 }
1524
1525
1526 typedef enum {
1527     FSFREEZE_HOOK_THAW = 0,
1528     FSFREEZE_HOOK_FREEZE,
1529 } FsfreezeHookArg;
1530
1531 static const char *fsfreeze_hook_arg_string[] = {
1532     "thaw",
1533     "freeze",
1534 };
1535
1536 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1537 {
1538     int status;
1539     pid_t pid;
1540     const char *hook;
1541     const char *arg_str = fsfreeze_hook_arg_string[arg];
1542     Error *local_err = NULL;
1543
1544     hook = ga_fsfreeze_hook(ga_state);
1545     if (!hook) {
1546         return;
1547     }
1548     if (access(hook, X_OK) != 0) {
1549         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1550         return;
1551     }
1552
1553     slog("executing fsfreeze hook with arg '%s'", arg_str);
1554     pid = fork();
1555     if (pid == 0) {
1556         setsid();
1557         reopen_fd_to_null(0);
1558         reopen_fd_to_null(1);
1559         reopen_fd_to_null(2);
1560
1561         execle(hook, hook, arg_str, NULL, environ);
1562         _exit(EXIT_FAILURE);
1563     } else if (pid < 0) {
1564         error_setg_errno(errp, errno, "failed to create child process");
1565         return;
1566     }
1567
1568     ga_wait_child(pid, &status, &local_err);
1569     if (local_err) {
1570         error_propagate(errp, local_err);
1571         return;
1572     }
1573
1574     if (!WIFEXITED(status)) {
1575         error_setg(errp, "fsfreeze hook has terminated abnormally");
1576         return;
1577     }
1578
1579     status = WEXITSTATUS(status);
1580     if (status) {
1581         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1582         return;
1583     }
1584 }
1585
1586 /*
1587  * Return status of freeze/thaw
1588  */
1589 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1590 {
1591     if (ga_is_frozen(ga_state)) {
1592         return GUEST_FSFREEZE_STATUS_FROZEN;
1593     }
1594
1595     return GUEST_FSFREEZE_STATUS_THAWED;
1596 }
1597
1598 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1599 {
1600     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1601 }
1602
1603 /*
1604  * Walk list of mounted file systems in the guest, and freeze the ones which
1605  * are real local file systems.
1606  */
1607 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1608                                        strList *mountpoints,
1609                                        Error **errp)
1610 {
1611     int ret = 0, i = 0;
1612     strList *list;
1613     FsMountList mounts;
1614     struct FsMount *mount;
1615     Error *local_err = NULL;
1616     int fd;
1617
1618     slog("guest-fsfreeze called");
1619
1620     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1621     if (local_err) {
1622         error_propagate(errp, local_err);
1623         return -1;
1624     }
1625
1626     QTAILQ_INIT(&mounts);
1627     build_fs_mount_list(&mounts, &local_err);
1628     if (local_err) {
1629         error_propagate(errp, local_err);
1630         return -1;
1631     }
1632
1633     /* cannot risk guest agent blocking itself on a write in this state */
1634     ga_set_frozen(ga_state);
1635
1636     QTAILQ_FOREACH_REVERSE(mount, &mounts, next) {
1637         /* To issue fsfreeze in the reverse order of mounts, check if the
1638          * mount is listed in the list here */
1639         if (has_mountpoints) {
1640             for (list = mountpoints; list; list = list->next) {
1641                 if (strcmp(list->value, mount->dirname) == 0) {
1642                     break;
1643                 }
1644             }
1645             if (!list) {
1646                 continue;
1647             }
1648         }
1649
1650         fd = qemu_open_old(mount->dirname, O_RDONLY);
1651         if (fd == -1) {
1652             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1653             goto error;
1654         }
1655
1656         /* we try to cull filesystems we know won't work in advance, but other
1657          * filesystems may not implement fsfreeze for less obvious reasons.
1658          * these will report EOPNOTSUPP. we simply ignore these when tallying
1659          * the number of frozen filesystems.
1660          * if a filesystem is mounted more than once (aka bind mount) a
1661          * consecutive attempt to freeze an already frozen filesystem will
1662          * return EBUSY.
1663          *
1664          * any other error means a failure to freeze a filesystem we
1665          * expect to be freezable, so return an error in those cases
1666          * and return system to thawed state.
1667          */
1668         ret = ioctl(fd, FIFREEZE);
1669         if (ret == -1) {
1670             if (errno != EOPNOTSUPP && errno != EBUSY) {
1671                 error_setg_errno(errp, errno, "failed to freeze %s",
1672                                  mount->dirname);
1673                 close(fd);
1674                 goto error;
1675             }
1676         } else {
1677             i++;
1678         }
1679         close(fd);
1680     }
1681
1682     free_fs_mount_list(&mounts);
1683     /* We may not issue any FIFREEZE here.
1684      * Just unset ga_state here and ready for the next call.
1685      */
1686     if (i == 0) {
1687         ga_unset_frozen(ga_state);
1688     }
1689     return i;
1690
1691 error:
1692     free_fs_mount_list(&mounts);
1693     qmp_guest_fsfreeze_thaw(NULL);
1694     return 0;
1695 }
1696
1697 /*
1698  * Walk list of frozen file systems in the guest, and thaw them.
1699  */
1700 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1701 {
1702     int ret;
1703     FsMountList mounts;
1704     FsMount *mount;
1705     int fd, i = 0, logged;
1706     Error *local_err = NULL;
1707
1708     QTAILQ_INIT(&mounts);
1709     build_fs_mount_list(&mounts, &local_err);
1710     if (local_err) {
1711         error_propagate(errp, local_err);
1712         return 0;
1713     }
1714
1715     QTAILQ_FOREACH(mount, &mounts, next) {
1716         logged = false;
1717         fd = qemu_open_old(mount->dirname, O_RDONLY);
1718         if (fd == -1) {
1719             continue;
1720         }
1721         /* we have no way of knowing whether a filesystem was actually unfrozen
1722          * as a result of a successful call to FITHAW, only that if an error
1723          * was returned the filesystem was *not* unfrozen by that particular
1724          * call.
1725          *
1726          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1727          * to unfreeze, continuing issuing FITHAW until an error is returned,
1728          * in which case either the filesystem is in an unfreezable state, or,
1729          * more likely, it was thawed previously (and remains so afterward).
1730          *
1731          * also, since the most recent successful call is the one that did
1732          * the actual unfreeze, we can use this to provide an accurate count
1733          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1734          * may * be useful for determining whether a filesystem was unfrozen
1735          * during the freeze/thaw phase by a process other than qemu-ga.
1736          */
1737         do {
1738             ret = ioctl(fd, FITHAW);
1739             if (ret == 0 && !logged) {
1740                 i++;
1741                 logged = true;
1742             }
1743         } while (ret == 0);
1744         close(fd);
1745     }
1746
1747     ga_unset_frozen(ga_state);
1748     free_fs_mount_list(&mounts);
1749
1750     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1751
1752     return i;
1753 }
1754
1755 static void guest_fsfreeze_cleanup(void)
1756 {
1757     Error *err = NULL;
1758
1759     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1760         qmp_guest_fsfreeze_thaw(&err);
1761         if (err) {
1762             slog("failed to clean up frozen filesystems: %s",
1763                  error_get_pretty(err));
1764             error_free(err);
1765         }
1766     }
1767 }
1768 #endif /* CONFIG_FSFREEZE */
1769
1770 #if defined(CONFIG_FSTRIM)
1771 /*
1772  * Walk list of mounted file systems in the guest, and trim them.
1773  */
1774 GuestFilesystemTrimResponse *
1775 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1776 {
1777     GuestFilesystemTrimResponse *response;
1778     GuestFilesystemTrimResultList *list;
1779     GuestFilesystemTrimResult *result;
1780     int ret = 0;
1781     FsMountList mounts;
1782     struct FsMount *mount;
1783     int fd;
1784     Error *local_err = NULL;
1785     struct fstrim_range r;
1786
1787     slog("guest-fstrim called");
1788
1789     QTAILQ_INIT(&mounts);
1790     build_fs_mount_list(&mounts, &local_err);
1791     if (local_err) {
1792         error_propagate(errp, local_err);
1793         return NULL;
1794     }
1795
1796     response = g_malloc0(sizeof(*response));
1797
1798     QTAILQ_FOREACH(mount, &mounts, next) {
1799         result = g_malloc0(sizeof(*result));
1800         result->path = g_strdup(mount->dirname);
1801
1802         list = g_malloc0(sizeof(*list));
1803         list->value = result;
1804         list->next = response->paths;
1805         response->paths = list;
1806
1807         fd = qemu_open_old(mount->dirname, O_RDONLY);
1808         if (fd == -1) {
1809             result->error = g_strdup_printf("failed to open: %s",
1810                                             strerror(errno));
1811             result->has_error = true;
1812             continue;
1813         }
1814
1815         /* We try to cull filesystems we know won't work in advance, but other
1816          * filesystems may not implement fstrim for less obvious reasons.
1817          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1818          * will be reported (e.g. CD-ROMs).
1819          * Any other error means an unexpected error.
1820          */
1821         r.start = 0;
1822         r.len = -1;
1823         r.minlen = has_minimum ? minimum : 0;
1824         ret = ioctl(fd, FITRIM, &r);
1825         if (ret == -1) {
1826             result->has_error = true;
1827             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1828                 result->error = g_strdup("trim not supported");
1829             } else {
1830                 result->error = g_strdup_printf("failed to trim: %s",
1831                                                 strerror(errno));
1832             }
1833             close(fd);
1834             continue;
1835         }
1836
1837         result->has_minimum = true;
1838         result->minimum = r.minlen;
1839         result->has_trimmed = true;
1840         result->trimmed = r.len;
1841         close(fd);
1842     }
1843
1844     free_fs_mount_list(&mounts);
1845     return response;
1846 }
1847 #endif /* CONFIG_FSTRIM */
1848
1849
1850 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1851 #define SUSPEND_SUPPORTED 0
1852 #define SUSPEND_NOT_SUPPORTED 1
1853
1854 typedef enum {
1855     SUSPEND_MODE_DISK = 0,
1856     SUSPEND_MODE_RAM = 1,
1857     SUSPEND_MODE_HYBRID = 2,
1858 } SuspendMode;
1859
1860 /*
1861  * Executes a command in a child process using g_spawn_sync,
1862  * returning an int >= 0 representing the exit status of the
1863  * process.
1864  *
1865  * If the program wasn't found in path, returns -1.
1866  *
1867  * If a problem happened when creating the child process,
1868  * returns -1 and errp is set.
1869  */
1870 static int run_process_child(const char *command[], Error **errp)
1871 {
1872     int exit_status, spawn_flag;
1873     GError *g_err = NULL;
1874     bool success;
1875
1876     spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1877                  G_SPAWN_STDERR_TO_DEV_NULL;
1878
1879     success =  g_spawn_sync(NULL, (char **)command, environ, spawn_flag,
1880                             NULL, NULL, NULL, NULL,
1881                             &exit_status, &g_err);
1882
1883     if (success) {
1884         return WEXITSTATUS(exit_status);
1885     }
1886
1887     if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1888         error_setg(errp, "failed to create child process, error '%s'",
1889                    g_err->message);
1890     }
1891
1892     g_error_free(g_err);
1893     return -1;
1894 }
1895
1896 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1897 {
1898     const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1899                                      "systemd-hybrid-sleep"};
1900     const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1901     int status;
1902
1903     status = run_process_child(cmd, errp);
1904
1905     /*
1906      * systemctl status uses LSB return codes so we can expect
1907      * status > 0 and be ok. To assert if the guest has support
1908      * for the selected suspend mode, status should be < 4. 4 is
1909      * the code for unknown service status, the return value when
1910      * the service does not exist. A common value is status = 3
1911      * (program is not running).
1912      */
1913     if (status > 0 && status < 4) {
1914         return true;
1915     }
1916
1917     return false;
1918 }
1919
1920 static void systemd_suspend(SuspendMode mode, Error **errp)
1921 {
1922     Error *local_err = NULL;
1923     const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1924     const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1925     int status;
1926
1927     status = run_process_child(cmd, &local_err);
1928
1929     if (status == 0) {
1930         return;
1931     }
1932
1933     if ((status == -1) && !local_err) {
1934         error_setg(errp, "the helper program 'systemctl %s' was not found",
1935                    systemctl_args[mode]);
1936         return;
1937     }
1938
1939     if (local_err) {
1940         error_propagate(errp, local_err);
1941     } else {
1942         error_setg(errp, "the helper program 'systemctl %s' returned an "
1943                    "unexpected exit status code (%d)",
1944                    systemctl_args[mode], status);
1945     }
1946 }
1947
1948 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1949 {
1950     Error *local_err = NULL;
1951     const char *pmutils_args[3] = {"--hibernate", "--suspend",
1952                                    "--suspend-hybrid"};
1953     const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1954     int status;
1955
1956     status = run_process_child(cmd, &local_err);
1957
1958     if (status == SUSPEND_SUPPORTED) {
1959         return true;
1960     }
1961
1962     if ((status == -1) && !local_err) {
1963         return false;
1964     }
1965
1966     if (local_err) {
1967         error_propagate(errp, local_err);
1968     } else {
1969         error_setg(errp,
1970                    "the helper program '%s' returned an unexpected exit"
1971                    " status code (%d)", "pm-is-supported", status);
1972     }
1973
1974     return false;
1975 }
1976
1977 static void pmutils_suspend(SuspendMode mode, Error **errp)
1978 {
1979     Error *local_err = NULL;
1980     const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1981                                        "pm-suspend-hybrid"};
1982     const char *cmd[2] = {pmutils_binaries[mode], NULL};
1983     int status;
1984
1985     status = run_process_child(cmd, &local_err);
1986
1987     if (status == 0) {
1988         return;
1989     }
1990
1991     if ((status == -1) && !local_err) {
1992         error_setg(errp, "the helper program '%s' was not found",
1993                    pmutils_binaries[mode]);
1994         return;
1995     }
1996
1997     if (local_err) {
1998         error_propagate(errp, local_err);
1999     } else {
2000         error_setg(errp,
2001                    "the helper program '%s' returned an unexpected exit"
2002                    " status code (%d)", pmutils_binaries[mode], status);
2003     }
2004 }
2005
2006 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
2007 {
2008     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2009     const char *sysfile_str = sysfile_strs[mode];
2010     char buf[32]; /* hopefully big enough */
2011     int fd;
2012     ssize_t ret;
2013
2014     if (!sysfile_str) {
2015         error_setg(errp, "unknown guest suspend mode");
2016         return false;
2017     }
2018
2019     fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
2020     if (fd < 0) {
2021         return false;
2022     }
2023
2024     ret = read(fd, buf, sizeof(buf) - 1);
2025     close(fd);
2026     if (ret <= 0) {
2027         return false;
2028     }
2029     buf[ret] = '\0';
2030
2031     if (strstr(buf, sysfile_str)) {
2032         return true;
2033     }
2034     return false;
2035 }
2036
2037 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
2038 {
2039     Error *local_err = NULL;
2040     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2041     const char *sysfile_str = sysfile_strs[mode];
2042     pid_t pid;
2043     int status;
2044
2045     if (!sysfile_str) {
2046         error_setg(errp, "unknown guest suspend mode");
2047         return;
2048     }
2049
2050     pid = fork();
2051     if (!pid) {
2052         /* child */
2053         int fd;
2054
2055         setsid();
2056         reopen_fd_to_null(0);
2057         reopen_fd_to_null(1);
2058         reopen_fd_to_null(2);
2059
2060         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
2061         if (fd < 0) {
2062             _exit(EXIT_FAILURE);
2063         }
2064
2065         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
2066             _exit(EXIT_FAILURE);
2067         }
2068
2069         _exit(EXIT_SUCCESS);
2070     } else if (pid < 0) {
2071         error_setg_errno(errp, errno, "failed to create child process");
2072         return;
2073     }
2074
2075     ga_wait_child(pid, &status, &local_err);
2076     if (local_err) {
2077         error_propagate(errp, local_err);
2078         return;
2079     }
2080
2081     if (WEXITSTATUS(status)) {
2082         error_setg(errp, "child process has failed to suspend");
2083     }
2084
2085 }
2086
2087 static void guest_suspend(SuspendMode mode, Error **errp)
2088 {
2089     Error *local_err = NULL;
2090     bool mode_supported = false;
2091
2092     if (systemd_supports_mode(mode, &local_err)) {
2093         mode_supported = true;
2094         systemd_suspend(mode, &local_err);
2095     }
2096
2097     if (!local_err) {
2098         return;
2099     }
2100
2101     error_free(local_err);
2102     local_err = NULL;
2103
2104     if (pmutils_supports_mode(mode, &local_err)) {
2105         mode_supported = true;
2106         pmutils_suspend(mode, &local_err);
2107     }
2108
2109     if (!local_err) {
2110         return;
2111     }
2112
2113     error_free(local_err);
2114     local_err = NULL;
2115
2116     if (linux_sys_state_supports_mode(mode, &local_err)) {
2117         mode_supported = true;
2118         linux_sys_state_suspend(mode, &local_err);
2119     }
2120
2121     if (!mode_supported) {
2122         error_free(local_err);
2123         error_setg(errp,
2124                    "the requested suspend mode is not supported by the guest");
2125     } else {
2126         error_propagate(errp, local_err);
2127     }
2128 }
2129
2130 void qmp_guest_suspend_disk(Error **errp)
2131 {
2132     guest_suspend(SUSPEND_MODE_DISK, errp);
2133 }
2134
2135 void qmp_guest_suspend_ram(Error **errp)
2136 {
2137     guest_suspend(SUSPEND_MODE_RAM, errp);
2138 }
2139
2140 void qmp_guest_suspend_hybrid(Error **errp)
2141 {
2142     guest_suspend(SUSPEND_MODE_HYBRID, errp);
2143 }
2144
2145 static GuestNetworkInterfaceList *
2146 guest_find_interface(GuestNetworkInterfaceList *head,
2147                      const char *name)
2148 {
2149     for (; head; head = head->next) {
2150         if (strcmp(head->value->name, name) == 0) {
2151             break;
2152         }
2153     }
2154
2155     return head;
2156 }
2157
2158 static int guest_get_network_stats(const char *name,
2159                        GuestNetworkInterfaceStat *stats)
2160 {
2161     int name_len;
2162     char const *devinfo = "/proc/net/dev";
2163     FILE *fp;
2164     char *line = NULL, *colon;
2165     size_t n = 0;
2166     fp = fopen(devinfo, "r");
2167     if (!fp) {
2168         return -1;
2169     }
2170     name_len = strlen(name);
2171     while (getline(&line, &n, fp) != -1) {
2172         long long dummy;
2173         long long rx_bytes;
2174         long long rx_packets;
2175         long long rx_errs;
2176         long long rx_dropped;
2177         long long tx_bytes;
2178         long long tx_packets;
2179         long long tx_errs;
2180         long long tx_dropped;
2181         char *trim_line;
2182         trim_line = g_strchug(line);
2183         if (trim_line[0] == '\0') {
2184             continue;
2185         }
2186         colon = strchr(trim_line, ':');
2187         if (!colon) {
2188             continue;
2189         }
2190         if (colon - name_len  == trim_line &&
2191            strncmp(trim_line, name, name_len) == 0) {
2192             if (sscanf(colon + 1,
2193                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
2194                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
2195                   &dummy, &dummy, &dummy, &dummy,
2196                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
2197                   &dummy, &dummy, &dummy, &dummy) != 16) {
2198                 continue;
2199             }
2200             stats->rx_bytes = rx_bytes;
2201             stats->rx_packets = rx_packets;
2202             stats->rx_errs = rx_errs;
2203             stats->rx_dropped = rx_dropped;
2204             stats->tx_bytes = tx_bytes;
2205             stats->tx_packets = tx_packets;
2206             stats->tx_errs = tx_errs;
2207             stats->tx_dropped = tx_dropped;
2208             fclose(fp);
2209             g_free(line);
2210             return 0;
2211         }
2212     }
2213     fclose(fp);
2214     g_free(line);
2215     g_debug("/proc/net/dev: Interface '%s' not found", name);
2216     return -1;
2217 }
2218
2219 /*
2220  * Build information about guest interfaces
2221  */
2222 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2223 {
2224     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
2225     struct ifaddrs *ifap, *ifa;
2226
2227     if (getifaddrs(&ifap) < 0) {
2228         error_setg_errno(errp, errno, "getifaddrs failed");
2229         goto error;
2230     }
2231
2232     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
2233         GuestNetworkInterfaceList *info;
2234         GuestIpAddressList **address_list = NULL, *address_item = NULL;
2235         GuestNetworkInterfaceStat  *interface_stat = NULL;
2236         char addr4[INET_ADDRSTRLEN];
2237         char addr6[INET6_ADDRSTRLEN];
2238         int sock;
2239         struct ifreq ifr;
2240         unsigned char *mac_addr;
2241         void *p;
2242
2243         g_debug("Processing %s interface", ifa->ifa_name);
2244
2245         info = guest_find_interface(head, ifa->ifa_name);
2246
2247         if (!info) {
2248             info = g_malloc0(sizeof(*info));
2249             info->value = g_malloc0(sizeof(*info->value));
2250             info->value->name = g_strdup(ifa->ifa_name);
2251
2252             if (!cur_item) {
2253                 head = cur_item = info;
2254             } else {
2255                 cur_item->next = info;
2256                 cur_item = info;
2257             }
2258         }
2259
2260         if (!info->value->has_hardware_address &&
2261             ifa->ifa_flags & SIOCGIFHWADDR) {
2262             /* we haven't obtained HW address yet */
2263             sock = socket(PF_INET, SOCK_STREAM, 0);
2264             if (sock == -1) {
2265                 error_setg_errno(errp, errno, "failed to create socket");
2266                 goto error;
2267             }
2268
2269             memset(&ifr, 0, sizeof(ifr));
2270             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
2271             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
2272                 error_setg_errno(errp, errno,
2273                                  "failed to get MAC address of %s",
2274                                  ifa->ifa_name);
2275                 close(sock);
2276                 goto error;
2277             }
2278
2279             close(sock);
2280             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
2281
2282             info->value->hardware_address =
2283                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
2284                                 (int) mac_addr[0], (int) mac_addr[1],
2285                                 (int) mac_addr[2], (int) mac_addr[3],
2286                                 (int) mac_addr[4], (int) mac_addr[5]);
2287
2288             info->value->has_hardware_address = true;
2289         }
2290
2291         if (ifa->ifa_addr &&
2292             ifa->ifa_addr->sa_family == AF_INET) {
2293             /* interface with IPv4 address */
2294             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
2295             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
2296                 error_setg_errno(errp, errno, "inet_ntop failed");
2297                 goto error;
2298             }
2299
2300             address_item = g_malloc0(sizeof(*address_item));
2301             address_item->value = g_malloc0(sizeof(*address_item->value));
2302             address_item->value->ip_address = g_strdup(addr4);
2303             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
2304
2305             if (ifa->ifa_netmask) {
2306                 /* Count the number of set bits in netmask.
2307                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2308                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
2309                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
2310             }
2311         } else if (ifa->ifa_addr &&
2312                    ifa->ifa_addr->sa_family == AF_INET6) {
2313             /* interface with IPv6 address */
2314             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
2315             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
2316                 error_setg_errno(errp, errno, "inet_ntop failed");
2317                 goto error;
2318             }
2319
2320             address_item = g_malloc0(sizeof(*address_item));
2321             address_item->value = g_malloc0(sizeof(*address_item->value));
2322             address_item->value->ip_address = g_strdup(addr6);
2323             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
2324
2325             if (ifa->ifa_netmask) {
2326                 /* Count the number of set bits in netmask.
2327                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2328                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
2329                 address_item->value->prefix =
2330                     ctpop32(((uint32_t *) p)[0]) +
2331                     ctpop32(((uint32_t *) p)[1]) +
2332                     ctpop32(((uint32_t *) p)[2]) +
2333                     ctpop32(((uint32_t *) p)[3]);
2334             }
2335         }
2336
2337         if (!address_item) {
2338             continue;
2339         }
2340
2341         address_list = &info->value->ip_addresses;
2342
2343         while (*address_list && (*address_list)->next) {
2344             address_list = &(*address_list)->next;
2345         }
2346
2347         if (!*address_list) {
2348             *address_list = address_item;
2349         } else {
2350             (*address_list)->next = address_item;
2351         }
2352
2353         info->value->has_ip_addresses = true;
2354
2355         if (!info->value->has_statistics) {
2356             interface_stat = g_malloc0(sizeof(*interface_stat));
2357             if (guest_get_network_stats(info->value->name,
2358                 interface_stat) == -1) {
2359                 info->value->has_statistics = false;
2360                 g_free(interface_stat);
2361             } else {
2362                 info->value->statistics = interface_stat;
2363                 info->value->has_statistics = true;
2364             }
2365         }
2366     }
2367
2368     freeifaddrs(ifap);
2369     return head;
2370
2371 error:
2372     freeifaddrs(ifap);
2373     qapi_free_GuestNetworkInterfaceList(head);
2374     return NULL;
2375 }
2376
2377 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
2378
2379 static long sysconf_exact(int name, const char *name_str, Error **errp)
2380 {
2381     long ret;
2382
2383     errno = 0;
2384     ret = sysconf(name);
2385     if (ret == -1) {
2386         if (errno == 0) {
2387             error_setg(errp, "sysconf(%s): value indefinite", name_str);
2388         } else {
2389             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
2390         }
2391     }
2392     return ret;
2393 }
2394
2395 /* Transfer online/offline status between @vcpu and the guest system.
2396  *
2397  * On input either @errp or *@errp must be NULL.
2398  *
2399  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
2400  * - R: vcpu->logical_id
2401  * - W: vcpu->online
2402  * - W: vcpu->can_offline
2403  *
2404  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
2405  * - R: vcpu->logical_id
2406  * - R: vcpu->online
2407  *
2408  * Written members remain unmodified on error.
2409  */
2410 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
2411                           char *dirpath, Error **errp)
2412 {
2413     int fd;
2414     int res;
2415     int dirfd;
2416     static const char fn[] = "online";
2417
2418     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2419     if (dirfd == -1) {
2420         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2421         return;
2422     }
2423
2424     fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2425     if (fd == -1) {
2426         if (errno != ENOENT) {
2427             error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2428         } else if (sys2vcpu) {
2429             vcpu->online = true;
2430             vcpu->can_offline = false;
2431         } else if (!vcpu->online) {
2432             error_setg(errp, "logical processor #%" PRId64 " can't be "
2433                        "offlined", vcpu->logical_id);
2434         } /* otherwise pretend successful re-onlining */
2435     } else {
2436         unsigned char status;
2437
2438         res = pread(fd, &status, 1, 0);
2439         if (res == -1) {
2440             error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2441         } else if (res == 0) {
2442             error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2443                        fn);
2444         } else if (sys2vcpu) {
2445             vcpu->online = (status != '0');
2446             vcpu->can_offline = true;
2447         } else if (vcpu->online != (status != '0')) {
2448             status = '0' + vcpu->online;
2449             if (pwrite(fd, &status, 1, 0) == -1) {
2450                 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2451                                  fn);
2452             }
2453         } /* otherwise pretend successful re-(on|off)-lining */
2454
2455         res = close(fd);
2456         g_assert(res == 0);
2457     }
2458
2459     res = close(dirfd);
2460     g_assert(res == 0);
2461 }
2462
2463 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2464 {
2465     int64_t current;
2466     GuestLogicalProcessorList *head, **link;
2467     long sc_max;
2468     Error *local_err = NULL;
2469
2470     current = 0;
2471     head = NULL;
2472     link = &head;
2473     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
2474
2475     while (local_err == NULL && current < sc_max) {
2476         GuestLogicalProcessor *vcpu;
2477         GuestLogicalProcessorList *entry;
2478         int64_t id = current++;
2479         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2480                                      id);
2481
2482         if (g_file_test(path, G_FILE_TEST_EXISTS)) {
2483             vcpu = g_malloc0(sizeof *vcpu);
2484             vcpu->logical_id = id;
2485             vcpu->has_can_offline = true; /* lolspeak ftw */
2486             transfer_vcpu(vcpu, true, path, &local_err);
2487             entry = g_malloc0(sizeof *entry);
2488             entry->value = vcpu;
2489             *link = entry;
2490             link = &entry->next;
2491         }
2492         g_free(path);
2493     }
2494
2495     if (local_err == NULL) {
2496         /* there's no guest with zero VCPUs */
2497         g_assert(head != NULL);
2498         return head;
2499     }
2500
2501     qapi_free_GuestLogicalProcessorList(head);
2502     error_propagate(errp, local_err);
2503     return NULL;
2504 }
2505
2506 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2507 {
2508     int64_t processed;
2509     Error *local_err = NULL;
2510
2511     processed = 0;
2512     while (vcpus != NULL) {
2513         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2514                                      vcpus->value->logical_id);
2515
2516         transfer_vcpu(vcpus->value, false, path, &local_err);
2517         g_free(path);
2518         if (local_err != NULL) {
2519             break;
2520         }
2521         ++processed;
2522         vcpus = vcpus->next;
2523     }
2524
2525     if (local_err != NULL) {
2526         if (processed == 0) {
2527             error_propagate(errp, local_err);
2528         } else {
2529             error_free(local_err);
2530         }
2531     }
2532
2533     return processed;
2534 }
2535
2536 void qmp_guest_set_user_password(const char *username,
2537                                  const char *password,
2538                                  bool crypted,
2539                                  Error **errp)
2540 {
2541     Error *local_err = NULL;
2542     char *passwd_path = NULL;
2543     pid_t pid;
2544     int status;
2545     int datafd[2] = { -1, -1 };
2546     char *rawpasswddata = NULL;
2547     size_t rawpasswdlen;
2548     char *chpasswddata = NULL;
2549     size_t chpasswdlen;
2550
2551     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2552     if (!rawpasswddata) {
2553         return;
2554     }
2555     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2556     rawpasswddata[rawpasswdlen] = '\0';
2557
2558     if (strchr(rawpasswddata, '\n')) {
2559         error_setg(errp, "forbidden characters in raw password");
2560         goto out;
2561     }
2562
2563     if (strchr(username, '\n') ||
2564         strchr(username, ':')) {
2565         error_setg(errp, "forbidden characters in username");
2566         goto out;
2567     }
2568
2569     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2570     chpasswdlen = strlen(chpasswddata);
2571
2572     passwd_path = g_find_program_in_path("chpasswd");
2573
2574     if (!passwd_path) {
2575         error_setg(errp, "cannot find 'passwd' program in PATH");
2576         goto out;
2577     }
2578
2579     if (pipe(datafd) < 0) {
2580         error_setg(errp, "cannot create pipe FDs");
2581         goto out;
2582     }
2583
2584     pid = fork();
2585     if (pid == 0) {
2586         close(datafd[1]);
2587         /* child */
2588         setsid();
2589         dup2(datafd[0], 0);
2590         reopen_fd_to_null(1);
2591         reopen_fd_to_null(2);
2592
2593         if (crypted) {
2594             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2595         } else {
2596             execle(passwd_path, "chpasswd", NULL, environ);
2597         }
2598         _exit(EXIT_FAILURE);
2599     } else if (pid < 0) {
2600         error_setg_errno(errp, errno, "failed to create child process");
2601         goto out;
2602     }
2603     close(datafd[0]);
2604     datafd[0] = -1;
2605
2606     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2607         error_setg_errno(errp, errno, "cannot write new account password");
2608         goto out;
2609     }
2610     close(datafd[1]);
2611     datafd[1] = -1;
2612
2613     ga_wait_child(pid, &status, &local_err);
2614     if (local_err) {
2615         error_propagate(errp, local_err);
2616         goto out;
2617     }
2618
2619     if (!WIFEXITED(status)) {
2620         error_setg(errp, "child process has terminated abnormally");
2621         goto out;
2622     }
2623
2624     if (WEXITSTATUS(status)) {
2625         error_setg(errp, "child process has failed to set user password");
2626         goto out;
2627     }
2628
2629 out:
2630     g_free(chpasswddata);
2631     g_free(rawpasswddata);
2632     g_free(passwd_path);
2633     if (datafd[0] != -1) {
2634         close(datafd[0]);
2635     }
2636     if (datafd[1] != -1) {
2637         close(datafd[1]);
2638     }
2639 }
2640
2641 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2642                                int size, Error **errp)
2643 {
2644     int fd;
2645     int res;
2646
2647     errno = 0;
2648     fd = openat(dirfd, pathname, O_RDONLY);
2649     if (fd == -1) {
2650         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2651         return;
2652     }
2653
2654     res = pread(fd, buf, size, 0);
2655     if (res == -1) {
2656         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2657     } else if (res == 0) {
2658         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2659     }
2660     close(fd);
2661 }
2662
2663 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2664                                 const char *buf, int size, Error **errp)
2665 {
2666     int fd;
2667
2668     errno = 0;
2669     fd = openat(dirfd, pathname, O_WRONLY);
2670     if (fd == -1) {
2671         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2672         return;
2673     }
2674
2675     if (pwrite(fd, buf, size, 0) == -1) {
2676         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2677     }
2678
2679     close(fd);
2680 }
2681
2682 /* Transfer online/offline status between @mem_blk and the guest system.
2683  *
2684  * On input either @errp or *@errp must be NULL.
2685  *
2686  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2687  * - R: mem_blk->phys_index
2688  * - W: mem_blk->online
2689  * - W: mem_blk->can_offline
2690  *
2691  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2692  * - R: mem_blk->phys_index
2693  * - R: mem_blk->online
2694  *-  R: mem_blk->can_offline
2695  * Written members remain unmodified on error.
2696  */
2697 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2698                                   GuestMemoryBlockResponse *result,
2699                                   Error **errp)
2700 {
2701     char *dirpath;
2702     int dirfd;
2703     char *status;
2704     Error *local_err = NULL;
2705
2706     if (!sys2memblk) {
2707         DIR *dp;
2708
2709         if (!result) {
2710             error_setg(errp, "Internal error, 'result' should not be NULL");
2711             return;
2712         }
2713         errno = 0;
2714         dp = opendir("/sys/devices/system/memory/");
2715          /* if there is no 'memory' directory in sysfs,
2716          * we think this VM does not support online/offline memory block,
2717          * any other solution?
2718          */
2719         if (!dp) {
2720             if (errno == ENOENT) {
2721                 result->response =
2722                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2723             }
2724             goto out1;
2725         }
2726         closedir(dp);
2727     }
2728
2729     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2730                               mem_blk->phys_index);
2731     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2732     if (dirfd == -1) {
2733         if (sys2memblk) {
2734             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2735         } else {
2736             if (errno == ENOENT) {
2737                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2738             } else {
2739                 result->response =
2740                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2741             }
2742         }
2743         g_free(dirpath);
2744         goto out1;
2745     }
2746     g_free(dirpath);
2747
2748     status = g_malloc0(10);
2749     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2750     if (local_err) {
2751         /* treat with sysfs file that not exist in old kernel */
2752         if (errno == ENOENT) {
2753             error_free(local_err);
2754             if (sys2memblk) {
2755                 mem_blk->online = true;
2756                 mem_blk->can_offline = false;
2757             } else if (!mem_blk->online) {
2758                 result->response =
2759                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2760             }
2761         } else {
2762             if (sys2memblk) {
2763                 error_propagate(errp, local_err);
2764             } else {
2765                 error_free(local_err);
2766                 result->response =
2767                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2768             }
2769         }
2770         goto out2;
2771     }
2772
2773     if (sys2memblk) {
2774         char removable = '0';
2775
2776         mem_blk->online = (strncmp(status, "online", 6) == 0);
2777
2778         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2779         if (local_err) {
2780             /* if no 'removable' file, it doesn't support offline mem blk */
2781             if (errno == ENOENT) {
2782                 error_free(local_err);
2783                 mem_blk->can_offline = false;
2784             } else {
2785                 error_propagate(errp, local_err);
2786             }
2787         } else {
2788             mem_blk->can_offline = (removable != '0');
2789         }
2790     } else {
2791         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2792             const char *new_state = mem_blk->online ? "online" : "offline";
2793
2794             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2795                                 &local_err);
2796             if (local_err) {
2797                 error_free(local_err);
2798                 result->response =
2799                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2800                 goto out2;
2801             }
2802
2803             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2804             result->has_error_code = false;
2805         } /* otherwise pretend successful re-(on|off)-lining */
2806     }
2807     g_free(status);
2808     close(dirfd);
2809     return;
2810
2811 out2:
2812     g_free(status);
2813     close(dirfd);
2814 out1:
2815     if (!sys2memblk) {
2816         result->has_error_code = true;
2817         result->error_code = errno;
2818     }
2819 }
2820
2821 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2822 {
2823     GuestMemoryBlockList *head, **link;
2824     Error *local_err = NULL;
2825     struct dirent *de;
2826     DIR *dp;
2827
2828     head = NULL;
2829     link = &head;
2830
2831     dp = opendir("/sys/devices/system/memory/");
2832     if (!dp) {
2833         /* it's ok if this happens to be a system that doesn't expose
2834          * memory blocks via sysfs, but otherwise we should report
2835          * an error
2836          */
2837         if (errno != ENOENT) {
2838             error_setg_errno(errp, errno, "Can't open directory"
2839                              "\"/sys/devices/system/memory/\"");
2840         }
2841         return NULL;
2842     }
2843
2844     /* Note: the phys_index of memory block may be discontinuous,
2845      * this is because a memblk is the unit of the Sparse Memory design, which
2846      * allows discontinuous memory ranges (ex. NUMA), so here we should
2847      * traverse the memory block directory.
2848      */
2849     while ((de = readdir(dp)) != NULL) {
2850         GuestMemoryBlock *mem_blk;
2851         GuestMemoryBlockList *entry;
2852
2853         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2854             !(de->d_type & DT_DIR)) {
2855             continue;
2856         }
2857
2858         mem_blk = g_malloc0(sizeof *mem_blk);
2859         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2860         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2861         mem_blk->has_can_offline = true; /* lolspeak ftw */
2862         transfer_memory_block(mem_blk, true, NULL, &local_err);
2863         if (local_err) {
2864             break;
2865         }
2866
2867         entry = g_malloc0(sizeof *entry);
2868         entry->value = mem_blk;
2869
2870         *link = entry;
2871         link = &entry->next;
2872     }
2873
2874     closedir(dp);
2875     if (local_err == NULL) {
2876         /* there's no guest with zero memory blocks */
2877         if (head == NULL) {
2878             error_setg(errp, "guest reported zero memory blocks!");
2879         }
2880         return head;
2881     }
2882
2883     qapi_free_GuestMemoryBlockList(head);
2884     error_propagate(errp, local_err);
2885     return NULL;
2886 }
2887
2888 GuestMemoryBlockResponseList *
2889 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2890 {
2891     GuestMemoryBlockResponseList *head, **link;
2892     Error *local_err = NULL;
2893
2894     head = NULL;
2895     link = &head;
2896
2897     while (mem_blks != NULL) {
2898         GuestMemoryBlockResponse *result;
2899         GuestMemoryBlockResponseList *entry;
2900         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2901
2902         result = g_malloc0(sizeof(*result));
2903         result->phys_index = current_mem_blk->phys_index;
2904         transfer_memory_block(current_mem_blk, false, result, &local_err);
2905         if (local_err) { /* should never happen */
2906             goto err;
2907         }
2908         entry = g_malloc0(sizeof *entry);
2909         entry->value = result;
2910
2911         *link = entry;
2912         link = &entry->next;
2913         mem_blks = mem_blks->next;
2914     }
2915
2916     return head;
2917 err:
2918     qapi_free_GuestMemoryBlockResponseList(head);
2919     error_propagate(errp, local_err);
2920     return NULL;
2921 }
2922
2923 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2924 {
2925     Error *local_err = NULL;
2926     char *dirpath;
2927     int dirfd;
2928     char *buf;
2929     GuestMemoryBlockInfo *info;
2930
2931     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2932     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2933     if (dirfd == -1) {
2934         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2935         g_free(dirpath);
2936         return NULL;
2937     }
2938     g_free(dirpath);
2939
2940     buf = g_malloc0(20);
2941     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2942     close(dirfd);
2943     if (local_err) {
2944         g_free(buf);
2945         error_propagate(errp, local_err);
2946         return NULL;
2947     }
2948
2949     info = g_new0(GuestMemoryBlockInfo, 1);
2950     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2951
2952     g_free(buf);
2953
2954     return info;
2955 }
2956
2957 #else /* defined(__linux__) */
2958
2959 void qmp_guest_suspend_disk(Error **errp)
2960 {
2961     error_setg(errp, QERR_UNSUPPORTED);
2962 }
2963
2964 void qmp_guest_suspend_ram(Error **errp)
2965 {
2966     error_setg(errp, QERR_UNSUPPORTED);
2967 }
2968
2969 void qmp_guest_suspend_hybrid(Error **errp)
2970 {
2971     error_setg(errp, QERR_UNSUPPORTED);
2972 }
2973
2974 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2975 {
2976     error_setg(errp, QERR_UNSUPPORTED);
2977     return NULL;
2978 }
2979
2980 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2981 {
2982     error_setg(errp, QERR_UNSUPPORTED);
2983     return NULL;
2984 }
2985
2986 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2987 {
2988     error_setg(errp, QERR_UNSUPPORTED);
2989     return -1;
2990 }
2991
2992 void qmp_guest_set_user_password(const char *username,
2993                                  const char *password,
2994                                  bool crypted,
2995                                  Error **errp)
2996 {
2997     error_setg(errp, QERR_UNSUPPORTED);
2998 }
2999
3000 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
3001 {
3002     error_setg(errp, QERR_UNSUPPORTED);
3003     return NULL;
3004 }
3005
3006 GuestMemoryBlockResponseList *
3007 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
3008 {
3009     error_setg(errp, QERR_UNSUPPORTED);
3010     return NULL;
3011 }
3012
3013 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
3014 {
3015     error_setg(errp, QERR_UNSUPPORTED);
3016     return NULL;
3017 }
3018
3019 #endif
3020
3021 #if !defined(CONFIG_FSFREEZE)
3022
3023 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
3024 {
3025     error_setg(errp, QERR_UNSUPPORTED);
3026     return NULL;
3027 }
3028
3029 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
3030 {
3031     error_setg(errp, QERR_UNSUPPORTED);
3032
3033     return 0;
3034 }
3035
3036 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
3037 {
3038     error_setg(errp, QERR_UNSUPPORTED);
3039
3040     return 0;
3041 }
3042
3043 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
3044                                        strList *mountpoints,
3045                                        Error **errp)
3046 {
3047     error_setg(errp, QERR_UNSUPPORTED);
3048
3049     return 0;
3050 }
3051
3052 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
3053 {
3054     error_setg(errp, QERR_UNSUPPORTED);
3055
3056     return 0;
3057 }
3058
3059 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
3060 {
3061     error_setg(errp, QERR_UNSUPPORTED);
3062     return NULL;
3063 }
3064
3065 #endif /* CONFIG_FSFREEZE */
3066
3067 #if !defined(CONFIG_FSTRIM)
3068 GuestFilesystemTrimResponse *
3069 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
3070 {
3071     error_setg(errp, QERR_UNSUPPORTED);
3072     return NULL;
3073 }
3074 #endif
3075
3076 /* add unsupported commands to the blacklist */
3077 GList *ga_command_blacklist_init(GList *blacklist)
3078 {
3079 #if !defined(__linux__)
3080     {
3081         const char *list[] = {
3082             "guest-suspend-disk", "guest-suspend-ram",
3083             "guest-suspend-hybrid", "guest-network-get-interfaces",
3084             "guest-get-vcpus", "guest-set-vcpus",
3085             "guest-get-memory-blocks", "guest-set-memory-blocks",
3086             "guest-get-memory-block-size", "guest-get-memory-block-info",
3087             NULL};
3088         char **p = (char **)list;
3089
3090         while (*p) {
3091             blacklist = g_list_append(blacklist, g_strdup(*p++));
3092         }
3093     }
3094 #endif
3095
3096 #if !defined(CONFIG_FSFREEZE)
3097     {
3098         const char *list[] = {
3099             "guest-get-fsinfo", "guest-fsfreeze-status",
3100             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
3101             "guest-fsfreeze-thaw", "guest-get-fsinfo",
3102             "guest-get-disks", NULL};
3103         char **p = (char **)list;
3104
3105         while (*p) {
3106             blacklist = g_list_append(blacklist, g_strdup(*p++));
3107         }
3108     }
3109 #endif
3110
3111 #if !defined(CONFIG_FSTRIM)
3112     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
3113 #endif
3114
3115     blacklist = g_list_append(blacklist, g_strdup("guest-get-devices"));
3116
3117     return blacklist;
3118 }
3119
3120 /* register init/cleanup routines for stateful command groups */
3121 void ga_command_state_init(GAState *s, GACommandState *cs)
3122 {
3123 #if defined(CONFIG_FSFREEZE)
3124     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
3125 #endif
3126 }
3127
3128 #ifdef HAVE_UTMPX
3129
3130 #define QGA_MICRO_SECOND_TO_SECOND 1000000
3131
3132 static double ga_get_login_time(struct utmpx *user_info)
3133 {
3134     double seconds = (double)user_info->ut_tv.tv_sec;
3135     double useconds = (double)user_info->ut_tv.tv_usec;
3136     useconds /= QGA_MICRO_SECOND_TO_SECOND;
3137     return seconds + useconds;
3138 }
3139
3140 GuestUserList *qmp_guest_get_users(Error **errp)
3141 {
3142     GHashTable *cache = NULL;
3143     GuestUserList *head = NULL, *cur_item = NULL;
3144     struct utmpx *user_info = NULL;
3145     gpointer value = NULL;
3146     GuestUser *user = NULL;
3147     GuestUserList *item = NULL;
3148     double login_time = 0;
3149
3150     cache = g_hash_table_new(g_str_hash, g_str_equal);
3151     setutxent();
3152
3153     for (;;) {
3154         user_info = getutxent();
3155         if (user_info == NULL) {
3156             break;
3157         } else if (user_info->ut_type != USER_PROCESS) {
3158             continue;
3159         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
3160             value = g_hash_table_lookup(cache, user_info->ut_user);
3161             user = (GuestUser *)value;
3162             login_time = ga_get_login_time(user_info);
3163             /* We're ensuring the earliest login time to be sent */
3164             if (login_time < user->login_time) {
3165                 user->login_time = login_time;
3166             }
3167             continue;
3168         }
3169
3170         item = g_new0(GuestUserList, 1);
3171         item->value = g_new0(GuestUser, 1);
3172         item->value->user = g_strdup(user_info->ut_user);
3173         item->value->login_time = ga_get_login_time(user_info);
3174
3175         g_hash_table_insert(cache, item->value->user, item->value);
3176
3177         if (!cur_item) {
3178             head = cur_item = item;
3179         } else {
3180             cur_item->next = item;
3181             cur_item = item;
3182         }
3183     }
3184     endutxent();
3185     g_hash_table_destroy(cache);
3186     return head;
3187 }
3188
3189 #else
3190
3191 GuestUserList *qmp_guest_get_users(Error **errp)
3192 {
3193     error_setg(errp, QERR_UNSUPPORTED);
3194     return NULL;
3195 }
3196
3197 #endif
3198
3199 /* Replace escaped special characters with theire real values. The replacement
3200  * is done in place -- returned value is in the original string.
3201  */
3202 static void ga_osrelease_replace_special(gchar *value)
3203 {
3204     gchar *p, *p2, quote;
3205
3206     /* Trim the string at first space or semicolon if it is not enclosed in
3207      * single or double quotes. */
3208     if ((value[0] != '"') || (value[0] == '\'')) {
3209         p = strchr(value, ' ');
3210         if (p != NULL) {
3211             *p = 0;
3212         }
3213         p = strchr(value, ';');
3214         if (p != NULL) {
3215             *p = 0;
3216         }
3217         return;
3218     }
3219
3220     quote = value[0];
3221     p2 = value;
3222     p = value + 1;
3223     while (*p != 0) {
3224         if (*p == '\\') {
3225             p++;
3226             switch (*p) {
3227             case '$':
3228             case '\'':
3229             case '"':
3230             case '\\':
3231             case '`':
3232                 break;
3233             default:
3234                 /* Keep literal backslash followed by whatever is there */
3235                 p--;
3236                 break;
3237             }
3238         } else if (*p == quote) {
3239             *p2 = 0;
3240             break;
3241         }
3242         *(p2++) = *(p++);
3243     }
3244 }
3245
3246 static GKeyFile *ga_parse_osrelease(const char *fname)
3247 {
3248     gchar *content = NULL;
3249     gchar *content2 = NULL;
3250     GError *err = NULL;
3251     GKeyFile *keys = g_key_file_new();
3252     const char *group = "[os-release]\n";
3253
3254     if (!g_file_get_contents(fname, &content, NULL, &err)) {
3255         slog("failed to read '%s', error: %s", fname, err->message);
3256         goto fail;
3257     }
3258
3259     if (!g_utf8_validate(content, -1, NULL)) {
3260         slog("file is not utf-8 encoded: %s", fname);
3261         goto fail;
3262     }
3263     content2 = g_strdup_printf("%s%s", group, content);
3264
3265     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
3266                                    &err)) {
3267         slog("failed to parse file '%s', error: %s", fname, err->message);
3268         goto fail;
3269     }
3270
3271     g_free(content);
3272     g_free(content2);
3273     return keys;
3274
3275 fail:
3276     g_error_free(err);
3277     g_free(content);
3278     g_free(content2);
3279     g_key_file_free(keys);
3280     return NULL;
3281 }
3282
3283 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
3284 {
3285     GuestOSInfo *info = NULL;
3286     struct utsname kinfo;
3287     GKeyFile *osrelease = NULL;
3288     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
3289
3290     info = g_new0(GuestOSInfo, 1);
3291
3292     if (uname(&kinfo) != 0) {
3293         error_setg_errno(errp, errno, "uname failed");
3294     } else {
3295         info->has_kernel_version = true;
3296         info->kernel_version = g_strdup(kinfo.version);
3297         info->has_kernel_release = true;
3298         info->kernel_release = g_strdup(kinfo.release);
3299         info->has_machine = true;
3300         info->machine = g_strdup(kinfo.machine);
3301     }
3302
3303     if (qga_os_release != NULL) {
3304         osrelease = ga_parse_osrelease(qga_os_release);
3305     } else {
3306         osrelease = ga_parse_osrelease("/etc/os-release");
3307         if (osrelease == NULL) {
3308             osrelease = ga_parse_osrelease("/usr/lib/os-release");
3309         }
3310     }
3311
3312     if (osrelease != NULL) {
3313         char *value;
3314
3315 #define GET_FIELD(field, osfield) do { \
3316     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
3317     if (value != NULL) { \
3318         ga_osrelease_replace_special(value); \
3319         info->has_ ## field = true; \
3320         info->field = value; \
3321     } \
3322 } while (0)
3323         GET_FIELD(id, "ID");
3324         GET_FIELD(name, "NAME");
3325         GET_FIELD(pretty_name, "PRETTY_NAME");
3326         GET_FIELD(version, "VERSION");
3327         GET_FIELD(version_id, "VERSION_ID");
3328         GET_FIELD(variant, "VARIANT");
3329         GET_FIELD(variant_id, "VARIANT_ID");
3330 #undef GET_FIELD
3331
3332         g_key_file_free(osrelease);
3333     }
3334
3335     return info;
3336 }
3337
3338 GuestDeviceInfoList *qmp_guest_get_devices(Error **errp)
3339 {
3340     error_setg(errp, QERR_UNSUPPORTED);
3341
3342     return NULL;
3343 }