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          * if a filesystem is mounted more than once (aka bind mount) a
1247          * consecutive attempt to freeze an already frozen filesystem will
1248          * return EBUSY.
1249          *
1250          * any other error means a failure to freeze a filesystem we
1251          * expect to be freezable, so return an error in those cases
1252          * and return system to thawed state.
1253          */
1254         ret = ioctl(fd, FIFREEZE);
1255         if (ret == -1) {
1256             if (errno != EOPNOTSUPP && errno != EBUSY) {
1257                 error_setg_errno(errp, errno, "failed to freeze %s",
1258                                  mount->dirname);
1259                 close(fd);
1260                 goto error;
1261             }
1262         } else {
1263             i++;
1264         }
1265         close(fd);
1266     }
1267
1268     free_fs_mount_list(&mounts);
1269     return i;
1270
1271 error:
1272     free_fs_mount_list(&mounts);
1273     qmp_guest_fsfreeze_thaw(NULL);
1274     return 0;
1275 }
1276
1277 /*
1278  * Walk list of frozen file systems in the guest, and thaw them.
1279  */
1280 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1281 {
1282     int ret;
1283     FsMountList mounts;
1284     FsMount *mount;
1285     int fd, i = 0, logged;
1286     Error *local_err = NULL;
1287
1288     QTAILQ_INIT(&mounts);
1289     build_fs_mount_list(&mounts, &local_err);
1290     if (local_err) {
1291         error_propagate(errp, local_err);
1292         return 0;
1293     }
1294
1295     QTAILQ_FOREACH(mount, &mounts, next) {
1296         logged = false;
1297         fd = qemu_open(mount->dirname, O_RDONLY);
1298         if (fd == -1) {
1299             continue;
1300         }
1301         /* we have no way of knowing whether a filesystem was actually unfrozen
1302          * as a result of a successful call to FITHAW, only that if an error
1303          * was returned the filesystem was *not* unfrozen by that particular
1304          * call.
1305          *
1306          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1307          * to unfreeze, continuing issuing FITHAW until an error is returned,
1308          * in which case either the filesystem is in an unfreezable state, or,
1309          * more likely, it was thawed previously (and remains so afterward).
1310          *
1311          * also, since the most recent successful call is the one that did
1312          * the actual unfreeze, we can use this to provide an accurate count
1313          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1314          * may * be useful for determining whether a filesystem was unfrozen
1315          * during the freeze/thaw phase by a process other than qemu-ga.
1316          */
1317         do {
1318             ret = ioctl(fd, FITHAW);
1319             if (ret == 0 && !logged) {
1320                 i++;
1321                 logged = true;
1322             }
1323         } while (ret == 0);
1324         close(fd);
1325     }
1326
1327     ga_unset_frozen(ga_state);
1328     free_fs_mount_list(&mounts);
1329
1330     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1331
1332     return i;
1333 }
1334
1335 static void guest_fsfreeze_cleanup(void)
1336 {
1337     Error *err = NULL;
1338
1339     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1340         qmp_guest_fsfreeze_thaw(&err);
1341         if (err) {
1342             slog("failed to clean up frozen filesystems: %s",
1343                  error_get_pretty(err));
1344             error_free(err);
1345         }
1346     }
1347 }
1348 #endif /* CONFIG_FSFREEZE */
1349
1350 #if defined(CONFIG_FSTRIM)
1351 /*
1352  * Walk list of mounted file systems in the guest, and trim them.
1353  */
1354 GuestFilesystemTrimResponse *
1355 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1356 {
1357     GuestFilesystemTrimResponse *response;
1358     GuestFilesystemTrimResultList *list;
1359     GuestFilesystemTrimResult *result;
1360     int ret = 0;
1361     FsMountList mounts;
1362     struct FsMount *mount;
1363     int fd;
1364     Error *local_err = NULL;
1365     struct fstrim_range r;
1366
1367     slog("guest-fstrim called");
1368
1369     QTAILQ_INIT(&mounts);
1370     build_fs_mount_list(&mounts, &local_err);
1371     if (local_err) {
1372         error_propagate(errp, local_err);
1373         return NULL;
1374     }
1375
1376     response = g_malloc0(sizeof(*response));
1377
1378     QTAILQ_FOREACH(mount, &mounts, next) {
1379         result = g_malloc0(sizeof(*result));
1380         result->path = g_strdup(mount->dirname);
1381
1382         list = g_malloc0(sizeof(*list));
1383         list->value = result;
1384         list->next = response->paths;
1385         response->paths = list;
1386
1387         fd = qemu_open(mount->dirname, O_RDONLY);
1388         if (fd == -1) {
1389             result->error = g_strdup_printf("failed to open: %s",
1390                                             strerror(errno));
1391             result->has_error = true;
1392             continue;
1393         }
1394
1395         /* We try to cull filesystems we know won't work in advance, but other
1396          * filesystems may not implement fstrim for less obvious reasons.
1397          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1398          * will be reported (e.g. CD-ROMs).
1399          * Any other error means an unexpected error.
1400          */
1401         r.start = 0;
1402         r.len = -1;
1403         r.minlen = has_minimum ? minimum : 0;
1404         ret = ioctl(fd, FITRIM, &r);
1405         if (ret == -1) {
1406             result->has_error = true;
1407             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1408                 result->error = g_strdup("trim not supported");
1409             } else {
1410                 result->error = g_strdup_printf("failed to trim: %s",
1411                                                 strerror(errno));
1412             }
1413             close(fd);
1414             continue;
1415         }
1416
1417         result->has_minimum = true;
1418         result->minimum = r.minlen;
1419         result->has_trimmed = true;
1420         result->trimmed = r.len;
1421         close(fd);
1422     }
1423
1424     free_fs_mount_list(&mounts);
1425     return response;
1426 }
1427 #endif /* CONFIG_FSTRIM */
1428
1429
1430 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1431 #define SUSPEND_SUPPORTED 0
1432 #define SUSPEND_NOT_SUPPORTED 1
1433
1434 static void bios_supports_mode(const char *pmutils_bin, const char *pmutils_arg,
1435                                const char *sysfile_str, Error **errp)
1436 {
1437     Error *local_err = NULL;
1438     char *pmutils_path;
1439     pid_t pid;
1440     int status;
1441
1442     pmutils_path = g_find_program_in_path(pmutils_bin);
1443
1444     pid = fork();
1445     if (!pid) {
1446         char buf[32]; /* hopefully big enough */
1447         ssize_t ret;
1448         int fd;
1449
1450         setsid();
1451         reopen_fd_to_null(0);
1452         reopen_fd_to_null(1);
1453         reopen_fd_to_null(2);
1454
1455         if (pmutils_path) {
1456             execle(pmutils_path, pmutils_bin, pmutils_arg, NULL, environ);
1457         }
1458
1459         /*
1460          * If we get here either pm-utils is not installed or execle() has
1461          * failed. Let's try the manual method if the caller wants it.
1462          */
1463
1464         if (!sysfile_str) {
1465             _exit(SUSPEND_NOT_SUPPORTED);
1466         }
1467
1468         fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1469         if (fd < 0) {
1470             _exit(SUSPEND_NOT_SUPPORTED);
1471         }
1472
1473         ret = read(fd, buf, sizeof(buf)-1);
1474         if (ret <= 0) {
1475             _exit(SUSPEND_NOT_SUPPORTED);
1476         }
1477         buf[ret] = '\0';
1478
1479         if (strstr(buf, sysfile_str)) {
1480             _exit(SUSPEND_SUPPORTED);
1481         }
1482
1483         _exit(SUSPEND_NOT_SUPPORTED);
1484     } else if (pid < 0) {
1485         error_setg_errno(errp, errno, "failed to create child process");
1486         goto out;
1487     }
1488
1489     ga_wait_child(pid, &status, &local_err);
1490     if (local_err) {
1491         error_propagate(errp, local_err);
1492         goto out;
1493     }
1494
1495     if (!WIFEXITED(status)) {
1496         error_setg(errp, "child process has terminated abnormally");
1497         goto out;
1498     }
1499
1500     switch (WEXITSTATUS(status)) {
1501     case SUSPEND_SUPPORTED:
1502         goto out;
1503     case SUSPEND_NOT_SUPPORTED:
1504         error_setg(errp,
1505                    "the requested suspend mode is not supported by the guest");
1506         goto out;
1507     default:
1508         error_setg(errp,
1509                    "the helper program '%s' returned an unexpected exit status"
1510                    " code (%d)", pmutils_path, WEXITSTATUS(status));
1511         goto out;
1512     }
1513
1514 out:
1515     g_free(pmutils_path);
1516 }
1517
1518 static void guest_suspend(const char *pmutils_bin, const char *sysfile_str,
1519                           Error **errp)
1520 {
1521     Error *local_err = NULL;
1522     char *pmutils_path;
1523     pid_t pid;
1524     int status;
1525
1526     pmutils_path = g_find_program_in_path(pmutils_bin);
1527
1528     pid = fork();
1529     if (pid == 0) {
1530         /* child */
1531         int fd;
1532
1533         setsid();
1534         reopen_fd_to_null(0);
1535         reopen_fd_to_null(1);
1536         reopen_fd_to_null(2);
1537
1538         if (pmutils_path) {
1539             execle(pmutils_path, pmutils_bin, NULL, environ);
1540         }
1541
1542         /*
1543          * If we get here either pm-utils is not installed or execle() has
1544          * failed. Let's try the manual method if the caller wants it.
1545          */
1546
1547         if (!sysfile_str) {
1548             _exit(EXIT_FAILURE);
1549         }
1550
1551         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1552         if (fd < 0) {
1553             _exit(EXIT_FAILURE);
1554         }
1555
1556         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1557             _exit(EXIT_FAILURE);
1558         }
1559
1560         _exit(EXIT_SUCCESS);
1561     } else if (pid < 0) {
1562         error_setg_errno(errp, errno, "failed to create child process");
1563         goto out;
1564     }
1565
1566     ga_wait_child(pid, &status, &local_err);
1567     if (local_err) {
1568         error_propagate(errp, local_err);
1569         goto out;
1570     }
1571
1572     if (!WIFEXITED(status)) {
1573         error_setg(errp, "child process has terminated abnormally");
1574         goto out;
1575     }
1576
1577     if (WEXITSTATUS(status)) {
1578         error_setg(errp, "child process has failed to suspend");
1579         goto out;
1580     }
1581
1582 out:
1583     g_free(pmutils_path);
1584 }
1585
1586 void qmp_guest_suspend_disk(Error **errp)
1587 {
1588     Error *local_err = NULL;
1589
1590     bios_supports_mode("pm-is-supported", "--hibernate", "disk", &local_err);
1591     if (local_err) {
1592         error_propagate(errp, local_err);
1593         return;
1594     }
1595
1596     guest_suspend("pm-hibernate", "disk", errp);
1597 }
1598
1599 void qmp_guest_suspend_ram(Error **errp)
1600 {
1601     Error *local_err = NULL;
1602
1603     bios_supports_mode("pm-is-supported", "--suspend", "mem", &local_err);
1604     if (local_err) {
1605         error_propagate(errp, local_err);
1606         return;
1607     }
1608
1609     guest_suspend("pm-suspend", "mem", errp);
1610 }
1611
1612 void qmp_guest_suspend_hybrid(Error **errp)
1613 {
1614     Error *local_err = NULL;
1615
1616     bios_supports_mode("pm-is-supported", "--suspend-hybrid", NULL,
1617                        &local_err);
1618     if (local_err) {
1619         error_propagate(errp, local_err);
1620         return;
1621     }
1622
1623     guest_suspend("pm-suspend-hybrid", NULL, errp);
1624 }
1625
1626 static GuestNetworkInterfaceList *
1627 guest_find_interface(GuestNetworkInterfaceList *head,
1628                      const char *name)
1629 {
1630     for (; head; head = head->next) {
1631         if (strcmp(head->value->name, name) == 0) {
1632             break;
1633         }
1634     }
1635
1636     return head;
1637 }
1638
1639 /*
1640  * Build information about guest interfaces
1641  */
1642 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1643 {
1644     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1645     struct ifaddrs *ifap, *ifa;
1646
1647     if (getifaddrs(&ifap) < 0) {
1648         error_setg_errno(errp, errno, "getifaddrs failed");
1649         goto error;
1650     }
1651
1652     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1653         GuestNetworkInterfaceList *info;
1654         GuestIpAddressList **address_list = NULL, *address_item = NULL;
1655         char addr4[INET_ADDRSTRLEN];
1656         char addr6[INET6_ADDRSTRLEN];
1657         int sock;
1658         struct ifreq ifr;
1659         unsigned char *mac_addr;
1660         void *p;
1661
1662         g_debug("Processing %s interface", ifa->ifa_name);
1663
1664         info = guest_find_interface(head, ifa->ifa_name);
1665
1666         if (!info) {
1667             info = g_malloc0(sizeof(*info));
1668             info->value = g_malloc0(sizeof(*info->value));
1669             info->value->name = g_strdup(ifa->ifa_name);
1670
1671             if (!cur_item) {
1672                 head = cur_item = info;
1673             } else {
1674                 cur_item->next = info;
1675                 cur_item = info;
1676             }
1677         }
1678
1679         if (!info->value->has_hardware_address &&
1680             ifa->ifa_flags & SIOCGIFHWADDR) {
1681             /* we haven't obtained HW address yet */
1682             sock = socket(PF_INET, SOCK_STREAM, 0);
1683             if (sock == -1) {
1684                 error_setg_errno(errp, errno, "failed to create socket");
1685                 goto error;
1686             }
1687
1688             memset(&ifr, 0, sizeof(ifr));
1689             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1690             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1691                 error_setg_errno(errp, errno,
1692                                  "failed to get MAC address of %s",
1693                                  ifa->ifa_name);
1694                 close(sock);
1695                 goto error;
1696             }
1697
1698             close(sock);
1699             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
1700
1701             info->value->hardware_address =
1702                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
1703                                 (int) mac_addr[0], (int) mac_addr[1],
1704                                 (int) mac_addr[2], (int) mac_addr[3],
1705                                 (int) mac_addr[4], (int) mac_addr[5]);
1706
1707             info->value->has_hardware_address = true;
1708         }
1709
1710         if (ifa->ifa_addr &&
1711             ifa->ifa_addr->sa_family == AF_INET) {
1712             /* interface with IPv4 address */
1713             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1714             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
1715                 error_setg_errno(errp, errno, "inet_ntop failed");
1716                 goto error;
1717             }
1718
1719             address_item = g_malloc0(sizeof(*address_item));
1720             address_item->value = g_malloc0(sizeof(*address_item->value));
1721             address_item->value->ip_address = g_strdup(addr4);
1722             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
1723
1724             if (ifa->ifa_netmask) {
1725                 /* Count the number of set bits in netmask.
1726                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1727                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
1728                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
1729             }
1730         } else if (ifa->ifa_addr &&
1731                    ifa->ifa_addr->sa_family == AF_INET6) {
1732             /* interface with IPv6 address */
1733             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
1734             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
1735                 error_setg_errno(errp, errno, "inet_ntop failed");
1736                 goto error;
1737             }
1738
1739             address_item = g_malloc0(sizeof(*address_item));
1740             address_item->value = g_malloc0(sizeof(*address_item->value));
1741             address_item->value->ip_address = g_strdup(addr6);
1742             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
1743
1744             if (ifa->ifa_netmask) {
1745                 /* Count the number of set bits in netmask.
1746                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1747                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
1748                 address_item->value->prefix =
1749                     ctpop32(((uint32_t *) p)[0]) +
1750                     ctpop32(((uint32_t *) p)[1]) +
1751                     ctpop32(((uint32_t *) p)[2]) +
1752                     ctpop32(((uint32_t *) p)[3]);
1753             }
1754         }
1755
1756         if (!address_item) {
1757             continue;
1758         }
1759
1760         address_list = &info->value->ip_addresses;
1761
1762         while (*address_list && (*address_list)->next) {
1763             address_list = &(*address_list)->next;
1764         }
1765
1766         if (!*address_list) {
1767             *address_list = address_item;
1768         } else {
1769             (*address_list)->next = address_item;
1770         }
1771
1772         info->value->has_ip_addresses = true;
1773
1774
1775     }
1776
1777     freeifaddrs(ifap);
1778     return head;
1779
1780 error:
1781     freeifaddrs(ifap);
1782     qapi_free_GuestNetworkInterfaceList(head);
1783     return NULL;
1784 }
1785
1786 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
1787
1788 static long sysconf_exact(int name, const char *name_str, Error **errp)
1789 {
1790     long ret;
1791
1792     errno = 0;
1793     ret = sysconf(name);
1794     if (ret == -1) {
1795         if (errno == 0) {
1796             error_setg(errp, "sysconf(%s): value indefinite", name_str);
1797         } else {
1798             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
1799         }
1800     }
1801     return ret;
1802 }
1803
1804 /* Transfer online/offline status between @vcpu and the guest system.
1805  *
1806  * On input either @errp or *@errp must be NULL.
1807  *
1808  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
1809  * - R: vcpu->logical_id
1810  * - W: vcpu->online
1811  * - W: vcpu->can_offline
1812  *
1813  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
1814  * - R: vcpu->logical_id
1815  * - R: vcpu->online
1816  *
1817  * Written members remain unmodified on error.
1818  */
1819 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
1820                           Error **errp)
1821 {
1822     char *dirpath;
1823     int dirfd;
1824
1825     dirpath = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
1826                               vcpu->logical_id);
1827     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
1828     if (dirfd == -1) {
1829         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
1830     } else {
1831         static const char fn[] = "online";
1832         int fd;
1833         int res;
1834
1835         fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
1836         if (fd == -1) {
1837             if (errno != ENOENT) {
1838                 error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
1839             } else if (sys2vcpu) {
1840                 vcpu->online = true;
1841                 vcpu->can_offline = false;
1842             } else if (!vcpu->online) {
1843                 error_setg(errp, "logical processor #%" PRId64 " can't be "
1844                            "offlined", vcpu->logical_id);
1845             } /* otherwise pretend successful re-onlining */
1846         } else {
1847             unsigned char status;
1848
1849             res = pread(fd, &status, 1, 0);
1850             if (res == -1) {
1851                 error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
1852             } else if (res == 0) {
1853                 error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
1854                            fn);
1855             } else if (sys2vcpu) {
1856                 vcpu->online = (status != '0');
1857                 vcpu->can_offline = true;
1858             } else if (vcpu->online != (status != '0')) {
1859                 status = '0' + vcpu->online;
1860                 if (pwrite(fd, &status, 1, 0) == -1) {
1861                     error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
1862                                      fn);
1863                 }
1864             } /* otherwise pretend successful re-(on|off)-lining */
1865
1866             res = close(fd);
1867             g_assert(res == 0);
1868         }
1869
1870         res = close(dirfd);
1871         g_assert(res == 0);
1872     }
1873
1874     g_free(dirpath);
1875 }
1876
1877 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
1878 {
1879     int64_t current;
1880     GuestLogicalProcessorList *head, **link;
1881     long sc_max;
1882     Error *local_err = NULL;
1883
1884     current = 0;
1885     head = NULL;
1886     link = &head;
1887     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
1888
1889     while (local_err == NULL && current < sc_max) {
1890         GuestLogicalProcessor *vcpu;
1891         GuestLogicalProcessorList *entry;
1892
1893         vcpu = g_malloc0(sizeof *vcpu);
1894         vcpu->logical_id = current++;
1895         vcpu->has_can_offline = true; /* lolspeak ftw */
1896         transfer_vcpu(vcpu, true, &local_err);
1897
1898         entry = g_malloc0(sizeof *entry);
1899         entry->value = vcpu;
1900
1901         *link = entry;
1902         link = &entry->next;
1903     }
1904
1905     if (local_err == NULL) {
1906         /* there's no guest with zero VCPUs */
1907         g_assert(head != NULL);
1908         return head;
1909     }
1910
1911     qapi_free_GuestLogicalProcessorList(head);
1912     error_propagate(errp, local_err);
1913     return NULL;
1914 }
1915
1916 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
1917 {
1918     int64_t processed;
1919     Error *local_err = NULL;
1920
1921     processed = 0;
1922     while (vcpus != NULL) {
1923         transfer_vcpu(vcpus->value, false, &local_err);
1924         if (local_err != NULL) {
1925             break;
1926         }
1927         ++processed;
1928         vcpus = vcpus->next;
1929     }
1930
1931     if (local_err != NULL) {
1932         if (processed == 0) {
1933             error_propagate(errp, local_err);
1934         } else {
1935             error_free(local_err);
1936         }
1937     }
1938
1939     return processed;
1940 }
1941
1942 void qmp_guest_set_user_password(const char *username,
1943                                  const char *password,
1944                                  bool crypted,
1945                                  Error **errp)
1946 {
1947     Error *local_err = NULL;
1948     char *passwd_path = NULL;
1949     pid_t pid;
1950     int status;
1951     int datafd[2] = { -1, -1 };
1952     char *rawpasswddata = NULL;
1953     size_t rawpasswdlen;
1954     char *chpasswddata = NULL;
1955     size_t chpasswdlen;
1956
1957     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
1958     if (!rawpasswddata) {
1959         return;
1960     }
1961     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
1962     rawpasswddata[rawpasswdlen] = '\0';
1963
1964     if (strchr(rawpasswddata, '\n')) {
1965         error_setg(errp, "forbidden characters in raw password");
1966         goto out;
1967     }
1968
1969     if (strchr(username, '\n') ||
1970         strchr(username, ':')) {
1971         error_setg(errp, "forbidden characters in username");
1972         goto out;
1973     }
1974
1975     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
1976     chpasswdlen = strlen(chpasswddata);
1977
1978     passwd_path = g_find_program_in_path("chpasswd");
1979
1980     if (!passwd_path) {
1981         error_setg(errp, "cannot find 'passwd' program in PATH");
1982         goto out;
1983     }
1984
1985     if (pipe(datafd) < 0) {
1986         error_setg(errp, "cannot create pipe FDs");
1987         goto out;
1988     }
1989
1990     pid = fork();
1991     if (pid == 0) {
1992         close(datafd[1]);
1993         /* child */
1994         setsid();
1995         dup2(datafd[0], 0);
1996         reopen_fd_to_null(1);
1997         reopen_fd_to_null(2);
1998
1999         if (crypted) {
2000             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2001         } else {
2002             execle(passwd_path, "chpasswd", NULL, environ);
2003         }
2004         _exit(EXIT_FAILURE);
2005     } else if (pid < 0) {
2006         error_setg_errno(errp, errno, "failed to create child process");
2007         goto out;
2008     }
2009     close(datafd[0]);
2010     datafd[0] = -1;
2011
2012     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2013         error_setg_errno(errp, errno, "cannot write new account password");
2014         goto out;
2015     }
2016     close(datafd[1]);
2017     datafd[1] = -1;
2018
2019     ga_wait_child(pid, &status, &local_err);
2020     if (local_err) {
2021         error_propagate(errp, local_err);
2022         goto out;
2023     }
2024
2025     if (!WIFEXITED(status)) {
2026         error_setg(errp, "child process has terminated abnormally");
2027         goto out;
2028     }
2029
2030     if (WEXITSTATUS(status)) {
2031         error_setg(errp, "child process has failed to set user password");
2032         goto out;
2033     }
2034
2035 out:
2036     g_free(chpasswddata);
2037     g_free(rawpasswddata);
2038     g_free(passwd_path);
2039     if (datafd[0] != -1) {
2040         close(datafd[0]);
2041     }
2042     if (datafd[1] != -1) {
2043         close(datafd[1]);
2044     }
2045 }
2046
2047 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2048                                int size, Error **errp)
2049 {
2050     int fd;
2051     int res;
2052
2053     errno = 0;
2054     fd = openat(dirfd, pathname, O_RDONLY);
2055     if (fd == -1) {
2056         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2057         return;
2058     }
2059
2060     res = pread(fd, buf, size, 0);
2061     if (res == -1) {
2062         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2063     } else if (res == 0) {
2064         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2065     }
2066     close(fd);
2067 }
2068
2069 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2070                                 const char *buf, int size, Error **errp)
2071 {
2072     int fd;
2073
2074     errno = 0;
2075     fd = openat(dirfd, pathname, O_WRONLY);
2076     if (fd == -1) {
2077         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2078         return;
2079     }
2080
2081     if (pwrite(fd, buf, size, 0) == -1) {
2082         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2083     }
2084
2085     close(fd);
2086 }
2087
2088 /* Transfer online/offline status between @mem_blk and the guest system.
2089  *
2090  * On input either @errp or *@errp must be NULL.
2091  *
2092  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2093  * - R: mem_blk->phys_index
2094  * - W: mem_blk->online
2095  * - W: mem_blk->can_offline
2096  *
2097  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2098  * - R: mem_blk->phys_index
2099  * - R: mem_blk->online
2100  *-  R: mem_blk->can_offline
2101  * Written members remain unmodified on error.
2102  */
2103 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2104                                   GuestMemoryBlockResponse *result,
2105                                   Error **errp)
2106 {
2107     char *dirpath;
2108     int dirfd;
2109     char *status;
2110     Error *local_err = NULL;
2111
2112     if (!sys2memblk) {
2113         DIR *dp;
2114
2115         if (!result) {
2116             error_setg(errp, "Internal error, 'result' should not be NULL");
2117             return;
2118         }
2119         errno = 0;
2120         dp = opendir("/sys/devices/system/memory/");
2121          /* if there is no 'memory' directory in sysfs,
2122          * we think this VM does not support online/offline memory block,
2123          * any other solution?
2124          */
2125         if (!dp && errno == ENOENT) {
2126             result->response =
2127                 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2128             goto out1;
2129         }
2130         closedir(dp);
2131     }
2132
2133     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2134                               mem_blk->phys_index);
2135     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2136     if (dirfd == -1) {
2137         if (sys2memblk) {
2138             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2139         } else {
2140             if (errno == ENOENT) {
2141                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2142             } else {
2143                 result->response =
2144                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2145             }
2146         }
2147         g_free(dirpath);
2148         goto out1;
2149     }
2150     g_free(dirpath);
2151
2152     status = g_malloc0(10);
2153     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2154     if (local_err) {
2155         /* treat with sysfs file that not exist in old kernel */
2156         if (errno == ENOENT) {
2157             error_free(local_err);
2158             if (sys2memblk) {
2159                 mem_blk->online = true;
2160                 mem_blk->can_offline = false;
2161             } else if (!mem_blk->online) {
2162                 result->response =
2163                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2164             }
2165         } else {
2166             if (sys2memblk) {
2167                 error_propagate(errp, local_err);
2168             } else {
2169                 result->response =
2170                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2171             }
2172         }
2173         goto out2;
2174     }
2175
2176     if (sys2memblk) {
2177         char removable = '0';
2178
2179         mem_blk->online = (strncmp(status, "online", 6) == 0);
2180
2181         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2182         if (local_err) {
2183             /* if no 'removable' file, it doesn't support offline mem blk */
2184             if (errno == ENOENT) {
2185                 error_free(local_err);
2186                 mem_blk->can_offline = false;
2187             } else {
2188                 error_propagate(errp, local_err);
2189             }
2190         } else {
2191             mem_blk->can_offline = (removable != '0');
2192         }
2193     } else {
2194         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2195             char *new_state = mem_blk->online ? g_strdup("online") :
2196                                                 g_strdup("offline");
2197
2198             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2199                                 &local_err);
2200             g_free(new_state);
2201             if (local_err) {
2202                 error_free(local_err);
2203                 result->response =
2204                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2205                 goto out2;
2206             }
2207
2208             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2209             result->has_error_code = false;
2210         } /* otherwise pretend successful re-(on|off)-lining */
2211     }
2212     g_free(status);
2213     close(dirfd);
2214     return;
2215
2216 out2:
2217     g_free(status);
2218     close(dirfd);
2219 out1:
2220     if (!sys2memblk) {
2221         result->has_error_code = true;
2222         result->error_code = errno;
2223     }
2224 }
2225
2226 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2227 {
2228     GuestMemoryBlockList *head, **link;
2229     Error *local_err = NULL;
2230     struct dirent *de;
2231     DIR *dp;
2232
2233     head = NULL;
2234     link = &head;
2235
2236     dp = opendir("/sys/devices/system/memory/");
2237     if (!dp) {
2238         /* it's ok if this happens to be a system that doesn't expose
2239          * memory blocks via sysfs, but otherwise we should report
2240          * an error
2241          */
2242         if (errno != ENOENT) {
2243             error_setg_errno(errp, errno, "Can't open directory"
2244                              "\"/sys/devices/system/memory/\"");
2245         }
2246         return NULL;
2247     }
2248
2249     /* Note: the phys_index of memory block may be discontinuous,
2250      * this is because a memblk is the unit of the Sparse Memory design, which
2251      * allows discontinuous memory ranges (ex. NUMA), so here we should
2252      * traverse the memory block directory.
2253      */
2254     while ((de = readdir(dp)) != NULL) {
2255         GuestMemoryBlock *mem_blk;
2256         GuestMemoryBlockList *entry;
2257
2258         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2259             !(de->d_type & DT_DIR)) {
2260             continue;
2261         }
2262
2263         mem_blk = g_malloc0(sizeof *mem_blk);
2264         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2265         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2266         mem_blk->has_can_offline = true; /* lolspeak ftw */
2267         transfer_memory_block(mem_blk, true, NULL, &local_err);
2268
2269         entry = g_malloc0(sizeof *entry);
2270         entry->value = mem_blk;
2271
2272         *link = entry;
2273         link = &entry->next;
2274     }
2275
2276     closedir(dp);
2277     if (local_err == NULL) {
2278         /* there's no guest with zero memory blocks */
2279         if (head == NULL) {
2280             error_setg(errp, "guest reported zero memory blocks!");
2281         }
2282         return head;
2283     }
2284
2285     qapi_free_GuestMemoryBlockList(head);
2286     error_propagate(errp, local_err);
2287     return NULL;
2288 }
2289
2290 GuestMemoryBlockResponseList *
2291 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2292 {
2293     GuestMemoryBlockResponseList *head, **link;
2294     Error *local_err = NULL;
2295
2296     head = NULL;
2297     link = &head;
2298
2299     while (mem_blks != NULL) {
2300         GuestMemoryBlockResponse *result;
2301         GuestMemoryBlockResponseList *entry;
2302         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2303
2304         result = g_malloc0(sizeof(*result));
2305         result->phys_index = current_mem_blk->phys_index;
2306         transfer_memory_block(current_mem_blk, false, result, &local_err);
2307         if (local_err) { /* should never happen */
2308             goto err;
2309         }
2310         entry = g_malloc0(sizeof *entry);
2311         entry->value = result;
2312
2313         *link = entry;
2314         link = &entry->next;
2315         mem_blks = mem_blks->next;
2316     }
2317
2318     return head;
2319 err:
2320     qapi_free_GuestMemoryBlockResponseList(head);
2321     error_propagate(errp, local_err);
2322     return NULL;
2323 }
2324
2325 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2326 {
2327     Error *local_err = NULL;
2328     char *dirpath;
2329     int dirfd;
2330     char *buf;
2331     GuestMemoryBlockInfo *info;
2332
2333     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2334     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2335     if (dirfd == -1) {
2336         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2337         g_free(dirpath);
2338         return NULL;
2339     }
2340     g_free(dirpath);
2341
2342     buf = g_malloc0(20);
2343     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2344     close(dirfd);
2345     if (local_err) {
2346         g_free(buf);
2347         error_propagate(errp, local_err);
2348         return NULL;
2349     }
2350
2351     info = g_new0(GuestMemoryBlockInfo, 1);
2352     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2353
2354     g_free(buf);
2355
2356     return info;
2357 }
2358
2359 #else /* defined(__linux__) */
2360
2361 void qmp_guest_suspend_disk(Error **errp)
2362 {
2363     error_setg(errp, QERR_UNSUPPORTED);
2364 }
2365
2366 void qmp_guest_suspend_ram(Error **errp)
2367 {
2368     error_setg(errp, QERR_UNSUPPORTED);
2369 }
2370
2371 void qmp_guest_suspend_hybrid(Error **errp)
2372 {
2373     error_setg(errp, QERR_UNSUPPORTED);
2374 }
2375
2376 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2377 {
2378     error_setg(errp, QERR_UNSUPPORTED);
2379     return NULL;
2380 }
2381
2382 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2383 {
2384     error_setg(errp, QERR_UNSUPPORTED);
2385     return NULL;
2386 }
2387
2388 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2389 {
2390     error_setg(errp, QERR_UNSUPPORTED);
2391     return -1;
2392 }
2393
2394 void qmp_guest_set_user_password(const char *username,
2395                                  const char *password,
2396                                  bool crypted,
2397                                  Error **errp)
2398 {
2399     error_setg(errp, QERR_UNSUPPORTED);
2400 }
2401
2402 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2403 {
2404     error_setg(errp, QERR_UNSUPPORTED);
2405     return NULL;
2406 }
2407
2408 GuestMemoryBlockResponseList *
2409 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2410 {
2411     error_setg(errp, QERR_UNSUPPORTED);
2412     return NULL;
2413 }
2414
2415 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2416 {
2417     error_setg(errp, QERR_UNSUPPORTED);
2418     return NULL;
2419 }
2420
2421 #endif
2422
2423 #if !defined(CONFIG_FSFREEZE)
2424
2425 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2426 {
2427     error_setg(errp, QERR_UNSUPPORTED);
2428     return NULL;
2429 }
2430
2431 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2432 {
2433     error_setg(errp, QERR_UNSUPPORTED);
2434
2435     return 0;
2436 }
2437
2438 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2439 {
2440     error_setg(errp, QERR_UNSUPPORTED);
2441
2442     return 0;
2443 }
2444
2445 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2446                                        strList *mountpoints,
2447                                        Error **errp)
2448 {
2449     error_setg(errp, QERR_UNSUPPORTED);
2450
2451     return 0;
2452 }
2453
2454 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2455 {
2456     error_setg(errp, QERR_UNSUPPORTED);
2457
2458     return 0;
2459 }
2460 #endif /* CONFIG_FSFREEZE */
2461
2462 #if !defined(CONFIG_FSTRIM)
2463 GuestFilesystemTrimResponse *
2464 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
2465 {
2466     error_setg(errp, QERR_UNSUPPORTED);
2467     return NULL;
2468 }
2469 #endif
2470
2471 /* add unsupported commands to the blacklist */
2472 GList *ga_command_blacklist_init(GList *blacklist)
2473 {
2474 #if !defined(__linux__)
2475     {
2476         const char *list[] = {
2477             "guest-suspend-disk", "guest-suspend-ram",
2478             "guest-suspend-hybrid", "guest-network-get-interfaces",
2479             "guest-get-vcpus", "guest-set-vcpus",
2480             "guest-get-memory-blocks", "guest-set-memory-blocks",
2481             "guest-get-memory-block-size", NULL};
2482         char **p = (char **)list;
2483
2484         while (*p) {
2485             blacklist = g_list_append(blacklist, g_strdup(*p++));
2486         }
2487     }
2488 #endif
2489
2490 #if !defined(CONFIG_FSFREEZE)
2491     {
2492         const char *list[] = {
2493             "guest-get-fsinfo", "guest-fsfreeze-status",
2494             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
2495             "guest-fsfreeze-thaw", "guest-get-fsinfo", NULL};
2496         char **p = (char **)list;
2497
2498         while (*p) {
2499             blacklist = g_list_append(blacklist, g_strdup(*p++));
2500         }
2501     }
2502 #endif
2503
2504 #if !defined(CONFIG_FSTRIM)
2505     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
2506 #endif
2507
2508     return blacklist;
2509 }
2510
2511 /* register init/cleanup routines for stateful command groups */
2512 void ga_command_state_init(GAState *s, GACommandState *cs)
2513 {
2514 #if defined(CONFIG_FSFREEZE)
2515     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
2516 #endif
2517 }