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