qga/commands-posix.c

   1 /*
   2  * QEMU Guest Agent POSIX-specific command implementations
   3  *
   4  * Copyright IBM Corp. 2011
   5  *
   6  * Authors:
   7  *  Michael Roth      <mdroth@linux.vnet.ibm.com>
   8  *  Michal Privoznik  <mprivozn@redhat.com>
   9  *
  10  * This work is licensed under the terms of the GNU GPL, version 2 or later.
  11  * See the COPYING file in the top-level directory.
  12  */
  13
  14 #include "qemu/osdep.h"
  15 #include <sys/ioctl.h>
  16 #include <sys/utsname.h>
  17 #include <sys/wait.h>
  18 #include <dirent.h>
  19 #include "qga/guest-agent-core.h"
  20 #include "qga-qapi-commands.h"
  21 #include "qapi/error.h"
  22 #include "qapi/qmp/qerror.h"
  23 #include "qemu/queue.h"
  24 #include "qemu/host-utils.h"
  25 #include "qemu/sockets.h"
  26 #include "qemu/base64.h"
  27 #include "qemu/cutils.h"
  28
  29 #ifdef HAVE_UTMPX
  30 #include <utmpx.h>
  31 #endif
  32
  33 #ifndef CONFIG_HAS_ENVIRON
  34 #ifdef __APPLE__
  35 #include <crt_externs.h>
  36 #define environ (*_NSGetEnviron())
  37 #else
  38 extern char **environ;
  39 #endif
  40 #endif
  41
  42 #if defined(__linux__)
  43 #include <mntent.h>
  44 #include <linux/fs.h>
  45 #include <ifaddrs.h>
  46 #include <arpa/inet.h>
  47 #include <sys/socket.h>
  48 #include <net/if.h>
  49
  50 #ifdef FIFREEZE
  51 #define CONFIG_FSFREEZE
  52 #endif
  53 #ifdef FITRIM
  54 #define CONFIG_FSTRIM
  55 #endif
  56 #endif
  57
  58 static void ga_wait_child(pid_t pid, int *status, Error **errp)
  59 {
  60     pid_t rpid;
  61
  62     *status = 0;
  63
  64     do {
  65         rpid = waitpid(pid, status, 0);
  66     } while (rpid == -1 && errno == EINTR);
  67
  68     if (rpid == -1) {
  69         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
  70                          pid);
  71         return;
  72     }
  73
  74     g_assert(rpid == pid);
  75 }
  76
  77 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
  78 {
  79     const char *shutdown_flag;
  80     Error *local_err = NULL;
  81     pid_t pid;
  82     int status;
  83
  84     slog("guest-shutdown called, mode: %s", mode);
  85     if (!has_mode || strcmp(mode, "powerdown") == 0) {
  86         shutdown_flag = "-P";
  87     } else if (strcmp(mode, "halt") == 0) {
  88         shutdown_flag = "-H";
  89     } else if (strcmp(mode, "reboot") == 0) {
  90         shutdown_flag = "-r";
  91     } else {
  92         error_setg(errp,
  93                    "mode is invalid (valid values are: halt|powerdown|reboot");
  94         return;
  95     }
  96
  97     pid = fork();
  98     if (pid == 0) {
  99         /* child, start the shutdown */
 100         setsid();
 101         reopen_fd_to_null(0);
 102         reopen_fd_to_null(1);
 103         reopen_fd_to_null(2);
 104
 105         execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
 106                "hypervisor initiated shutdown", (char*)NULL, environ);
 107         _exit(EXIT_FAILURE);
 108     } else if (pid < 0) {
 109         error_setg_errno(errp, errno, "failed to create child process");
 110         return;
 111     }
 112
 113     ga_wait_child(pid, &status, &local_err);
 114     if (local_err) {
 115         error_propagate(errp, local_err);
 116         return;
 117     }
 118
 119     if (!WIFEXITED(status)) {
 120         error_setg(errp, "child process has terminated abnormally");
 121         return;
 122     }
 123
 124     if (WEXITSTATUS(status)) {
 125         error_setg(errp, "child process has failed to shutdown");
 126         return;
 127     }
 128
 129     /* succeeded */
 130 }
 131
 132 int64_t qmp_guest_get_time(Error **errp)
 133 {
 134    int ret;
 135    qemu_timeval tq;
 136
 137    ret = qemu_gettimeofday(&tq);
 138    if (ret < 0) {
 139        error_setg_errno(errp, errno, "Failed to get time");
 140        return -1;
 141    }
 142
 143    return tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
 144 }
 145
 146 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
 147 {
 148     int ret;
 149     int status;
 150     pid_t pid;
 151     Error *local_err = NULL;
 152     struct timeval tv;
 153
 154     /* If user has passed a time, validate and set it. */
 155     if (has_time) {
 156         GDate date = { 0, };
 157
 158         /* year-2038 will overflow in case time_t is 32bit */
 159         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
 160             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
 161             return;
 162         }
 163
 164         tv.tv_sec = time_ns / 1000000000;
 165         tv.tv_usec = (time_ns % 1000000000) / 1000;
 166         g_date_set_time_t(&date, tv.tv_sec);
 167         if (date.year < 1970 || date.year >= 2070) {
 168             error_setg_errno(errp, errno, "Invalid time");
 169             return;
 170         }
 171
 172         ret = settimeofday(&tv, NULL);
 173         if (ret < 0) {
 174             error_setg_errno(errp, errno, "Failed to set time to guest");
 175             return;
 176         }
 177     }
 178
 179     /* Now, if user has passed a time to set and the system time is set, we
 180      * just need to synchronize the hardware clock. However, if no time was
 181      * passed, user is requesting the opposite: set the system time from the
 182      * hardware clock (RTC). */
 183     pid = fork();
 184     if (pid == 0) {
 185         setsid();
 186         reopen_fd_to_null(0);
 187         reopen_fd_to_null(1);
 188         reopen_fd_to_null(2);
 189
 190         /* Use '/sbin/hwclock -w' to set RTC from the system time,
 191          * or '/sbin/hwclock -s' to set the system time from RTC. */
 192         execle("/sbin/hwclock", "hwclock", has_time ? "-w" : "-s",
 193                NULL, environ);
 194         _exit(EXIT_FAILURE);
 195     } else if (pid < 0) {
 196         error_setg_errno(errp, errno, "failed to create child process");
 197         return;
 198     }
 199
 200     ga_wait_child(pid, &status, &local_err);
 201     if (local_err) {
 202         error_propagate(errp, local_err);
 203         return;
 204     }
 205
 206     if (!WIFEXITED(status)) {
 207         error_setg(errp, "child process has terminated abnormally");
 208         return;
 209     }
 210
 211     if (WEXITSTATUS(status)) {
 212         error_setg(errp, "hwclock failed to set hardware clock to system time");
 213         return;
 214     }
 215 }
 216
 217 typedef enum {
 218     RW_STATE_NEW,
 219     RW_STATE_READING,
 220     RW_STATE_WRITING,
 221 } RwState;
 222
 223 typedef struct GuestFileHandle {
 224     uint64_t id;
 225     FILE *fh;
 226     RwState state;
 227     QTAILQ_ENTRY(GuestFileHandle) next;
 228 } GuestFileHandle;
 229
 230 static struct {
 231     QTAILQ_HEAD(, GuestFileHandle) filehandles;
 232 } guest_file_state = {
 233     .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
 234 };
 235
 236 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
 237 {
 238     GuestFileHandle *gfh;
 239     int64_t handle;
 240
 241     handle = ga_get_fd_handle(ga_state, errp);
 242     if (handle < 0) {
 243         return -1;
 244     }
 245
 246     gfh = g_new0(GuestFileHandle, 1);
 247     gfh->id = handle;
 248     gfh->fh = fh;
 249     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
 250
 251     return handle;
 252 }
 253
 254 static GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
 255 {
 256     GuestFileHandle *gfh;
 257
 258     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
 259     {
 260         if (gfh->id == id) {
 261             return gfh;
 262         }
 263     }
 264
 265     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
 266     return NULL;
 267 }
 268
 269 typedef const char * const ccpc;
 270
 271 #ifndef O_BINARY
 272 #define O_BINARY 0
 273 #endif
 274
 275 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
 276 static const struct {
 277     ccpc *forms;
 278     int oflag_base;
 279 } guest_file_open_modes[] = {
 280     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
 281     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
 282     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
 283     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
 284     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
 285     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
 286     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
 287     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
 288     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
 289     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
 290     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
 291     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
 292 };
 293
 294 static int
 295 find_open_flag(const char *mode_str, Error **errp)
 296 {
 297     unsigned mode;
 298
 299     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
 300         ccpc *form;
 301
 302         form = guest_file_open_modes[mode].forms;
 303         while (*form != NULL && strcmp(*form, mode_str) != 0) {
 304             ++form;
 305         }
 306         if (*form != NULL) {
 307             break;
 308         }
 309     }
 310
 311     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
 312         error_setg(errp, "invalid file open mode '%s'", mode_str);
 313         return -1;
 314     }
 315     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
 316 }
 317
 318 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
 319                                S_IRGRP | S_IWGRP | \
 320                                S_IROTH | S_IWOTH)
 321
 322 static FILE *
 323 safe_open_or_create(const char *path, const char *mode, Error **errp)
 324 {
 325     Error *local_err = NULL;
 326     int oflag;
 327
 328     oflag = find_open_flag(mode, &local_err);
 329     if (local_err == NULL) {
 330         int fd;
 331
 332         /* If the caller wants / allows creation of a new file, we implement it
 333          * with a two step process: open() + (open() / fchmod()).
 334          *
 335          * First we insist on creating the file exclusively as a new file. If
 336          * that succeeds, we're free to set any file-mode bits on it. (The
 337          * motivation is that we want to set those file-mode bits independently
 338          * of the current umask.)
 339          *
 340          * If the exclusive creation fails because the file already exists
 341          * (EEXIST is not possible for any other reason), we just attempt to
 342          * open the file, but in this case we won't be allowed to change the
 343          * file-mode bits on the preexistent file.
 344          *
 345          * The pathname should never disappear between the two open()s in
 346          * practice. If it happens, then someone very likely tried to race us.
 347          * In this case just go ahead and report the ENOENT from the second
 348          * open() to the caller.
 349          *
 350          * If the caller wants to open a preexistent file, then the first
 351          * open() is decisive and its third argument is ignored, and the second
 352          * open() and the fchmod() are never called.
 353          */
 354         fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
 355         if (fd == -1 && errno == EEXIST) {
 356             oflag &= ~(unsigned)O_CREAT;
 357             fd = open(path, oflag);
 358         }
 359
 360         if (fd == -1) {
 361             error_setg_errno(&local_err, errno, "failed to open file '%s' "
 362                              "(mode: '%s')", path, mode);
 363         } else {
 364             qemu_set_cloexec(fd);
 365
 366             if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
 367                 error_setg_errno(&local_err, errno, "failed to set permission "
 368                                  "0%03o on new file '%s' (mode: '%s')",
 369                                  (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
 370             } else {
 371                 FILE *f;
 372
 373                 f = fdopen(fd, mode);
 374                 if (f == NULL) {
 375                     error_setg_errno(&local_err, errno, "failed to associate "
 376                                      "stdio stream with file descriptor %d, "
 377                                      "file '%s' (mode: '%s')", fd, path, mode);
 378                 } else {
 379                     return f;
 380                 }
 381             }
 382
 383             close(fd);
 384             if (oflag & O_CREAT) {
 385                 unlink(path);
 386             }
 387         }
 388     }
 389
 390     error_propagate(errp, local_err);
 391     return NULL;
 392 }
 393
 394 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
 395                             Error **errp)
 396 {
 397     FILE *fh;
 398     Error *local_err = NULL;
 399     int64_t handle;
 400
 401     if (!has_mode) {
 402         mode = "r";
 403     }
 404     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
 405     fh = safe_open_or_create(path, mode, &local_err);
 406     if (local_err != NULL) {
 407         error_propagate(errp, local_err);
 408         return -1;
 409     }
 410
 411     /* set fd non-blocking to avoid common use cases (like reading from a
 412      * named pipe) from hanging the agent
 413      */
 414     qemu_set_nonblock(fileno(fh));
 415
 416     handle = guest_file_handle_add(fh, errp);
 417     if (handle < 0) {
 418         fclose(fh);
 419         return -1;
 420     }
 421
 422     slog("guest-file-open, handle: %" PRId64, handle);
 423     return handle;
 424 }
 425
 426 void qmp_guest_file_close(int64_t handle, Error **errp)
 427 {
 428     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 429     int ret;
 430
 431     slog("guest-file-close called, handle: %" PRId64, handle);
 432     if (!gfh) {
 433         return;
 434     }
 435
 436     ret = fclose(gfh->fh);
 437     if (ret == EOF) {
 438         error_setg_errno(errp, errno, "failed to close handle");
 439         return;
 440     }
 441
 442     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
 443     g_free(gfh);
 444 }
 445
 446 struct GuestFileRead *qmp_guest_file_read(int64_t handle, bool has_count,
 447                                           int64_t count, Error **errp)
 448 {
 449     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 450     GuestFileRead *read_data = NULL;
 451     guchar *buf;
 452     FILE *fh;
 453     size_t read_count;
 454
 455     if (!gfh) {
 456         return NULL;
 457     }
 458
 459     if (!has_count) {
 460         count = QGA_READ_COUNT_DEFAULT;
 461     } else if (count < 0) {
 462         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 463                    count);
 464         return NULL;
 465     }
 466
 467     fh = gfh->fh;
 468
 469     /* explicitly flush when switching from writing to reading */
 470     if (gfh->state == RW_STATE_WRITING) {
 471         int ret = fflush(fh);
 472         if (ret == EOF) {
 473             error_setg_errno(errp, errno, "failed to flush file");
 474             return NULL;
 475         }
 476         gfh->state = RW_STATE_NEW;
 477     }
 478
 479     buf = g_malloc0(count+1);
 480     read_count = fread(buf, 1, count, fh);
 481     if (ferror(fh)) {
 482         error_setg_errno(errp, errno, "failed to read file");
 483         slog("guest-file-read failed, handle: %" PRId64, handle);
 484     } else {
 485         buf[read_count] = 0;
 486         read_data = g_new0(GuestFileRead, 1);
 487         read_data->count = read_count;
 488         read_data->eof = feof(fh);
 489         if (read_count) {
 490             read_data->buf_b64 = g_base64_encode(buf, read_count);
 491         }
 492         gfh->state = RW_STATE_READING;
 493     }
 494     g_free(buf);
 495     clearerr(fh);
 496
 497     return read_data;
 498 }
 499
 500 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
 501                                      bool has_count, int64_t count,
 502                                      Error **errp)
 503 {
 504     GuestFileWrite *write_data = NULL;
 505     guchar *buf;
 506     gsize buf_len;
 507     int write_count;
 508     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 509     FILE *fh;
 510
 511     if (!gfh) {
 512         return NULL;
 513     }
 514
 515     fh = gfh->fh;
 516
 517     if (gfh->state == RW_STATE_READING) {
 518         int ret = fseek(fh, 0, SEEK_CUR);
 519         if (ret == -1) {
 520             error_setg_errno(errp, errno, "failed to seek file");
 521             return NULL;
 522         }
 523         gfh->state = RW_STATE_NEW;
 524     }
 525
 526     buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
 527     if (!buf) {
 528         return NULL;
 529     }
 530
 531     if (!has_count) {
 532         count = buf_len;
 533     } else if (count < 0 || count > buf_len) {
 534         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 535                    count);
 536         g_free(buf);
 537         return NULL;
 538     }
 539
 540     write_count = fwrite(buf, 1, count, fh);
 541     if (ferror(fh)) {
 542         error_setg_errno(errp, errno, "failed to write to file");
 543         slog("guest-file-write failed, handle: %" PRId64, handle);
 544     } else {
 545         write_data = g_new0(GuestFileWrite, 1);
 546         write_data->count = write_count;
 547         write_data->eof = feof(fh);
 548         gfh->state = RW_STATE_WRITING;
 549     }
 550     g_free(buf);
 551     clearerr(fh);
 552
 553     return write_data;
 554 }
 555
 556 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
 557                                           GuestFileWhence *whence_code,
 558                                           Error **errp)
 559 {
 560     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 561     GuestFileSeek *seek_data = NULL;
 562     FILE *fh;
 563     int ret;
 564     int whence;
 565     Error *err = NULL;
 566
 567     if (!gfh) {
 568         return NULL;
 569     }
 570
 571     /* We stupidly exposed 'whence':'int' in our qapi */
 572     whence = ga_parse_whence(whence_code, &err);
 573     if (err) {
 574         error_propagate(errp, err);
 575         return NULL;
 576     }
 577
 578     fh = gfh->fh;
 579     ret = fseek(fh, offset, whence);
 580     if (ret == -1) {
 581         error_setg_errno(errp, errno, "failed to seek file");
 582         if (errno == ESPIPE) {
 583             /* file is non-seekable, stdio shouldn't be buffering anyways */
 584             gfh->state = RW_STATE_NEW;
 585         }
 586     } else {
 587         seek_data = g_new0(GuestFileSeek, 1);
 588         seek_data->position = ftell(fh);
 589         seek_data->eof = feof(fh);
 590         gfh->state = RW_STATE_NEW;
 591     }
 592     clearerr(fh);
 593
 594     return seek_data;
 595 }
 596
 597 void qmp_guest_file_flush(int64_t handle, Error **errp)
 598 {
 599     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 600     FILE *fh;
 601     int ret;
 602
 603     if (!gfh) {
 604         return;
 605     }
 606
 607     fh = gfh->fh;
 608     ret = fflush(fh);
 609     if (ret == EOF) {
 610         error_setg_errno(errp, errno, "failed to flush file");
 611     } else {
 612         gfh->state = RW_STATE_NEW;
 613     }
 614 }
 615
 616 /* linux-specific implementations. avoid this if at all possible. */
 617 #if defined(__linux__)
 618
 619 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
 620 typedef struct FsMount {
 621     char *dirname;
 622     char *devtype;
 623     unsigned int devmajor, devminor;
 624     QTAILQ_ENTRY(FsMount) next;
 625 } FsMount;
 626
 627 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
 628
 629 static void free_fs_mount_list(FsMountList *mounts)
 630 {
 631      FsMount *mount, *temp;
 632
 633      if (!mounts) {
 634          return;
 635      }
 636
 637      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
 638          QTAILQ_REMOVE(mounts, mount, next);
 639          g_free(mount->dirname);
 640          g_free(mount->devtype);
 641          g_free(mount);
 642      }
 643 }
 644
 645 static int dev_major_minor(const char *devpath,
 646                            unsigned int *devmajor, unsigned int *devminor)
 647 {
 648     struct stat st;
 649
 650     *devmajor = 0;
 651     *devminor = 0;
 652
 653     if (stat(devpath, &st) < 0) {
 654         slog("failed to stat device file '%s': %s", devpath, strerror(errno));
 655         return -1;
 656     }
 657     if (S_ISDIR(st.st_mode)) {
 658         /* It is bind mount */
 659         return -2;
 660     }
 661     if (S_ISBLK(st.st_mode)) {
 662         *devmajor = major(st.st_rdev);
 663         *devminor = minor(st.st_rdev);
 664         return 0;
 665     }
 666     return -1;
 667 }
 668
 669 /*
 670  * Walk the mount table and build a list of local file systems
 671  */
 672 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
 673 {
 674     struct mntent *ment;
 675     FsMount *mount;
 676     char const *mtab = "/proc/self/mounts";
 677     FILE *fp;
 678     unsigned int devmajor, devminor;
 679
 680     fp = setmntent(mtab, "r");
 681     if (!fp) {
 682         error_setg(errp, "failed to open mtab file: '%s'", mtab);
 683         return;
 684     }
 685
 686     while ((ment = getmntent(fp))) {
 687         /*
 688          * An entry which device name doesn't start with a '/' is
 689          * either a dummy file system or a network file system.
 690          * Add special handling for smbfs and cifs as is done by
 691          * coreutils as well.
 692          */
 693         if ((ment->mnt_fsname[0] != '/') ||
 694             (strcmp(ment->mnt_type, "smbfs") == 0) ||
 695             (strcmp(ment->mnt_type, "cifs") == 0)) {
 696             continue;
 697         }
 698         if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
 699             /* Skip bind mounts */
 700             continue;
 701         }
 702
 703         mount = g_new0(FsMount, 1);
 704         mount->dirname = g_strdup(ment->mnt_dir);
 705         mount->devtype = g_strdup(ment->mnt_type);
 706         mount->devmajor = devmajor;
 707         mount->devminor = devminor;
 708
 709         QTAILQ_INSERT_TAIL(mounts, mount, next);
 710     }
 711
 712     endmntent(fp);
 713 }
 714
 715 static void decode_mntname(char *name, int len)
 716 {
 717     int i, j = 0;
 718     for (i = 0; i <= len; i++) {
 719         if (name[i] != '\\') {
 720             name[j++] = name[i];
 721         } else if (name[i + 1] == '\\') {
 722             name[j++] = '\\';
 723             i++;
 724         } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
 725                    name[i + 2] >= '0' && name[i + 2] <= '7' &&
 726                    name[i + 3] >= '0' && name[i + 3] <= '7') {
 727             name[j++] = (name[i + 1] - '0') * 64 +
 728                         (name[i + 2] - '0') * 8 +
 729                         (name[i + 3] - '0');
 730             i += 3;
 731         } else {
 732             name[j++] = name[i];
 733         }
 734     }
 735 }
 736
 737 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
 738 {
 739     FsMount *mount;
 740     char const *mountinfo = "/proc/self/mountinfo";
 741     FILE *fp;
 742     char *line = NULL, *dash;
 743     size_t n;
 744     char check;
 745     unsigned int devmajor, devminor;
 746     int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
 747
 748     fp = fopen(mountinfo, "r");
 749     if (!fp) {
 750         build_fs_mount_list_from_mtab(mounts, errp);
 751         return;
 752     }
 753
 754     while (getline(&line, &n, fp) != -1) {
 755         ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
 756                      &devmajor, &devminor, &dir_s, &dir_e, &check);
 757         if (ret < 3) {
 758             continue;
 759         }
 760         dash = strstr(line + dir_e, " - ");
 761         if (!dash) {
 762             continue;
 763         }
 764         ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
 765                      &type_s, &type_e, &dev_s, &dev_e, &check);
 766         if (ret < 1) {
 767             continue;
 768         }
 769         line[dir_e] = 0;
 770         dash[type_e] = 0;
 771         dash[dev_e] = 0;
 772         decode_mntname(line + dir_s, dir_e - dir_s);
 773         decode_mntname(dash + dev_s, dev_e - dev_s);
 774         if (devmajor == 0) {
 775             /* btrfs reports major number = 0 */
 776             if (strcmp("btrfs", dash + type_s) != 0 ||
 777                 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
 778                 continue;
 779             }
 780         }
 781
 782         mount = g_new0(FsMount, 1);
 783         mount->dirname = g_strdup(line + dir_s);
 784         mount->devtype = g_strdup(dash + type_s);
 785         mount->devmajor = devmajor;
 786         mount->devminor = devminor;
 787
 788         QTAILQ_INSERT_TAIL(mounts, mount, next);
 789     }
 790     free(line);
 791
 792     fclose(fp);
 793 }
 794 #endif
 795
 796 #if defined(CONFIG_FSFREEZE)
 797
 798 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
 799 {
 800     char *path;
 801     char *dpath;
 802     char *driver = NULL;
 803     char buf[PATH_MAX];
 804     ssize_t len;
 805
 806     path = g_strndup(syspath, pathlen);
 807     dpath = g_strdup_printf("%s/driver", path);
 808     len = readlink(dpath, buf, sizeof(buf) - 1);
 809     if (len != -1) {
 810         buf[len] = 0;
 811         driver = g_path_get_basename(buf);
 812     }
 813     g_free(dpath);
 814     g_free(path);
 815     return driver;
 816 }
 817
 818 static int compare_uint(const void *_a, const void *_b)
 819 {
 820     unsigned int a = *(unsigned int *)_a;
 821     unsigned int b = *(unsigned int *)_b;
 822
 823     return a < b ? -1 : a > b ? 1 : 0;
 824 }
 825
 826 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
 827 static int build_hosts(char const *syspath, char const *host, bool ata,
 828                        unsigned int *hosts, int hosts_max, Error **errp)
 829 {
 830     char *path;
 831     DIR *dir;
 832     struct dirent *entry;
 833     int i = 0;
 834
 835     path = g_strndup(syspath, host - syspath);
 836     dir = opendir(path);
 837     if (!dir) {
 838         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
 839         g_free(path);
 840         return -1;
 841     }
 842
 843     while (i < hosts_max) {
 844         entry = readdir(dir);
 845         if (!entry) {
 846             break;
 847         }
 848         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
 849             ++i;
 850         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
 851             ++i;
 852         }
 853     }
 854
 855     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
 856
 857     g_free(path);
 858     closedir(dir);
 859     return i;
 860 }
 861
 862 /* Store disk device info specified by @sysfs into @fs */
 863 static void build_guest_fsinfo_for_real_device(char const *syspath,
 864                                                GuestFilesystemInfo *fs,
 865                                                Error **errp)
 866 {
 867     unsigned int pci[4], host, hosts[8], tgt[3];
 868     int i, nhosts = 0, pcilen;
 869     GuestDiskAddress *disk;
 870     GuestPCIAddress *pciaddr;
 871     GuestDiskAddressList *list = NULL;
 872     bool has_ata = false, has_host = false, has_tgt = false;
 873     char *p, *q, *driver = NULL;
 874
 875     p = strstr(syspath, "/devices/pci");
 876     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
 877                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
 878         g_debug("only pci device is supported: sysfs path \"%s\"", syspath);
 879         return;
 880     }
 881
 882     driver = get_pci_driver(syspath, (p + 12 + pcilen) - syspath, errp);
 883     if (!driver) {
 884         goto cleanup;
 885     }
 886
 887     p = strstr(syspath, "/target");
 888     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
 889                     tgt, tgt + 1, tgt + 2) == 3) {
 890         has_tgt = true;
 891     }
 892
 893     p = strstr(syspath, "/ata");
 894     if (p) {
 895         q = p + 4;
 896         has_ata = true;
 897     } else {
 898         p = strstr(syspath, "/host");
 899         q = p + 5;
 900     }
 901     if (p && sscanf(q, "%u", &host) == 1) {
 902         has_host = true;
 903         nhosts = build_hosts(syspath, p, has_ata, hosts,
 904                              ARRAY_SIZE(hosts), errp);
 905         if (nhosts < 0) {
 906             goto cleanup;
 907         }
 908     }
 909
 910     pciaddr = g_malloc0(sizeof(*pciaddr));
 911     pciaddr->domain = pci[0];
 912     pciaddr->bus = pci[1];
 913     pciaddr->slot = pci[2];
 914     pciaddr->function = pci[3];
 915
 916     disk = g_malloc0(sizeof(*disk));
 917     disk->pci_controller = pciaddr;
 918
 919     list = g_malloc0(sizeof(*list));
 920     list->value = disk;
 921
 922     if (strcmp(driver, "ata_piix") == 0) {
 923         /* a host per ide bus, target*:0:<unit>:0 */
 924         if (!has_host || !has_tgt) {
 925             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 926             goto cleanup;
 927         }
 928         for (i = 0; i < nhosts; i++) {
 929             if (host == hosts[i]) {
 930                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
 931                 disk->bus = i;
 932                 disk->unit = tgt[1];
 933                 break;
 934             }
 935         }
 936         if (i >= nhosts) {
 937             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 938             goto cleanup;
 939         }
 940     } else if (strcmp(driver, "sym53c8xx") == 0) {
 941         /* scsi(LSI Logic): target*:0:<unit>:0 */
 942         if (!has_tgt) {
 943             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 944             goto cleanup;
 945         }
 946         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 947         disk->unit = tgt[1];
 948     } else if (strcmp(driver, "virtio-pci") == 0) {
 949         if (has_tgt) {
 950             /* virtio-scsi: target*:0:0:<unit> */
 951             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 952             disk->unit = tgt[2];
 953         } else {
 954             /* virtio-blk: 1 disk per 1 device */
 955             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
 956         }
 957     } else if (strcmp(driver, "ahci") == 0) {
 958         /* ahci: 1 host per 1 unit */
 959         if (!has_host || !has_tgt) {
 960             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 961             goto cleanup;
 962         }
 963         for (i = 0; i < nhosts; i++) {
 964             if (host == hosts[i]) {
 965                 disk->unit = i;
 966                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
 967                 break;
 968             }
 969         }
 970         if (i >= nhosts) {
 971             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 972             goto cleanup;
 973         }
 974     } else {
 975         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
 976         goto cleanup;
 977     }
 978
 979     list->next = fs->disk;
 980     fs->disk = list;
 981     g_free(driver);
 982     return;
 983
 984 cleanup:
 985     if (list) {
 986         qapi_free_GuestDiskAddressList(list);
 987     }
 988     g_free(driver);
 989 }
 990
 991 static void build_guest_fsinfo_for_device(char const *devpath,
 992                                           GuestFilesystemInfo *fs,
 993                                           Error **errp);
 994
 995 /* Store a list of slave devices of virtual volume specified by @syspath into
 996  * @fs */
 997 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
 998                                                   GuestFilesystemInfo *fs,
 999                                                   Error **errp)
1000 {
1001     DIR *dir;
1002     char *dirpath;
1003     struct dirent *entry;
1004
1005     dirpath = g_strdup_printf("%s/slaves", syspath);
1006     dir = opendir(dirpath);
1007     if (!dir) {
1008         if (errno != ENOENT) {
1009             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1010         }
1011         g_free(dirpath);
1012         return;
1013     }
1014
1015     for (;;) {
1016         errno = 0;
1017         entry = readdir(dir);
1018         if (entry == NULL) {
1019             if (errno) {
1020                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1021             }
1022             break;
1023         }
1024
1025         if (entry->d_type == DT_LNK) {
1026             char *path;
1027
1028             g_debug(" slave device '%s'", entry->d_name);
1029             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1030             build_guest_fsinfo_for_device(path, fs, errp);
1031             g_free(path);
1032
1033             if (*errp) {
1034                 break;
1035             }
1036         }
1037     }
1038
1039     g_free(dirpath);
1040     closedir(dir);
1041 }
1042
1043 /* Dispatch to functions for virtual/real device */
1044 static void build_guest_fsinfo_for_device(char const *devpath,
1045                                           GuestFilesystemInfo *fs,
1046                                           Error **errp)
1047 {
1048     char *syspath = realpath(devpath, NULL);
1049
1050     if (!syspath) {
1051         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1052         return;
1053     }
1054
1055     if (!fs->name) {
1056         fs->name = g_path_get_basename(syspath);
1057     }
1058
1059     g_debug("  parse sysfs path '%s'", syspath);
1060
1061     if (strstr(syspath, "/devices/virtual/block/")) {
1062         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1063     } else {
1064         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1065     }
1066
1067     free(syspath);
1068 }
1069
1070 /* Return a list of the disk device(s)' info which @mount lies on */
1071 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1072                                                Error **errp)
1073 {
1074     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1075     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1076                                     mount->devmajor, mount->devminor);
1077
1078     fs->mountpoint = g_strdup(mount->dirname);
1079     fs->type = g_strdup(mount->devtype);
1080     build_guest_fsinfo_for_device(devpath, fs, errp);
1081
1082     g_free(devpath);
1083     return fs;
1084 }
1085
1086 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1087 {
1088     FsMountList mounts;
1089     struct FsMount *mount;
1090     GuestFilesystemInfoList *new, *ret = NULL;
1091     Error *local_err = NULL;
1092
1093     QTAILQ_INIT(&mounts);
1094     build_fs_mount_list(&mounts, &local_err);
1095     if (local_err) {
1096         error_propagate(errp, local_err);
1097         return NULL;
1098     }
1099
1100     QTAILQ_FOREACH(mount, &mounts, next) {
1101         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1102
1103         new = g_malloc0(sizeof(*ret));
1104         new->value = build_guest_fsinfo(mount, &local_err);
1105         new->next = ret;
1106         ret = new;
1107         if (local_err) {
1108             error_propagate(errp, local_err);
1109             qapi_free_GuestFilesystemInfoList(ret);
1110             ret = NULL;
1111             break;
1112         }
1113     }
1114
1115     free_fs_mount_list(&mounts);
1116     return ret;
1117 }
1118
1119
1120 typedef enum {
1121     FSFREEZE_HOOK_THAW = 0,
1122     FSFREEZE_HOOK_FREEZE,
1123 } FsfreezeHookArg;
1124
1125 static const char *fsfreeze_hook_arg_string[] = {
1126     "thaw",
1127     "freeze",
1128 };
1129
1130 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1131 {
1132     int status;
1133     pid_t pid;
1134     const char *hook;
1135     const char *arg_str = fsfreeze_hook_arg_string[arg];
1136     Error *local_err = NULL;
1137
1138     hook = ga_fsfreeze_hook(ga_state);
1139     if (!hook) {
1140         return;
1141     }
1142     if (access(hook, X_OK) != 0) {
1143         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1144         return;
1145     }
1146
1147     slog("executing fsfreeze hook with arg '%s'", arg_str);
1148     pid = fork();
1149     if (pid == 0) {
1150         setsid();
1151         reopen_fd_to_null(0);
1152         reopen_fd_to_null(1);
1153         reopen_fd_to_null(2);
1154
1155         execle(hook, hook, arg_str, NULL, environ);
1156         _exit(EXIT_FAILURE);
1157     } else if (pid < 0) {
1158         error_setg_errno(errp, errno, "failed to create child process");
1159         return;
1160     }
1161
1162     ga_wait_child(pid, &status, &local_err);
1163     if (local_err) {
1164         error_propagate(errp, local_err);
1165         return;
1166     }
1167
1168     if (!WIFEXITED(status)) {
1169         error_setg(errp, "fsfreeze hook has terminated abnormally");
1170         return;
1171     }
1172
1173     status = WEXITSTATUS(status);
1174     if (status) {
1175         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1176         return;
1177     }
1178 }
1179
1180 /*
1181  * Return status of freeze/thaw
1182  */
1183 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1184 {
1185     if (ga_is_frozen(ga_state)) {
1186         return GUEST_FSFREEZE_STATUS_FROZEN;
1187     }
1188
1189     return GUEST_FSFREEZE_STATUS_THAWED;
1190 }
1191
1192 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1193 {
1194     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1195 }
1196
1197 /*
1198  * Walk list of mounted file systems in the guest, and freeze the ones which
1199  * are real local file systems.
1200  */
1201 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1202                                        strList *mountpoints,
1203                                        Error **errp)
1204 {
1205     int ret = 0, i = 0;
1206     strList *list;
1207     FsMountList mounts;
1208     struct FsMount *mount;
1209     Error *local_err = NULL;
1210     int fd;
1211
1212     slog("guest-fsfreeze called");
1213
1214     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1215     if (local_err) {
1216         error_propagate(errp, local_err);
1217         return -1;
1218     }
1219
1220     QTAILQ_INIT(&mounts);
1221     build_fs_mount_list(&mounts, &local_err);
1222     if (local_err) {
1223         error_propagate(errp, local_err);
1224         return -1;
1225     }
1226
1227     /* cannot risk guest agent blocking itself on a write in this state */
1228     ga_set_frozen(ga_state);
1229
1230     QTAILQ_FOREACH_REVERSE(mount, &mounts, FsMountList, next) {
1231         /* To issue fsfreeze in the reverse order of mounts, check if the
1232          * mount is listed in the list here */
1233         if (has_mountpoints) {
1234             for (list = mountpoints; list; list = list->next) {
1235                 if (strcmp(list->value, mount->dirname) == 0) {
1236                     break;
1237                 }
1238             }
1239             if (!list) {
1240                 continue;
1241             }
1242         }
1243
1244         fd = qemu_open(mount->dirname, O_RDONLY);
1245         if (fd == -1) {
1246             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1247             goto error;
1248         }
1249
1250         /* we try to cull filesystems we know won't work in advance, but other
1251          * filesystems may not implement fsfreeze for less obvious reasons.
1252          * these will report EOPNOTSUPP. we simply ignore these when tallying
1253          * the number of frozen filesystems.
1254          * if a filesystem is mounted more than once (aka bind mount) a
1255          * consecutive attempt to freeze an already frozen filesystem will
1256          * return EBUSY.
1257          *
1258          * any other error means a failure to freeze a filesystem we
1259          * expect to be freezable, so return an error in those cases
1260          * and return system to thawed state.
1261          */
1262         ret = ioctl(fd, FIFREEZE);
1263         if (ret == -1) {
1264             if (errno != EOPNOTSUPP && errno != EBUSY) {
1265                 error_setg_errno(errp, errno, "failed to freeze %s",
1266                                  mount->dirname);
1267                 close(fd);
1268                 goto error;
1269             }
1270         } else {
1271             i++;
1272         }
1273         close(fd);
1274     }
1275
1276     free_fs_mount_list(&mounts);
1277     return i;
1278
1279 error:
1280     free_fs_mount_list(&mounts);
1281     qmp_guest_fsfreeze_thaw(NULL);
1282     return 0;
1283 }
1284
1285 /*
1286  * Walk list of frozen file systems in the guest, and thaw them.
1287  */
1288 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1289 {
1290     int ret;
1291     FsMountList mounts;
1292     FsMount *mount;
1293     int fd, i = 0, logged;
1294     Error *local_err = NULL;
1295
1296     QTAILQ_INIT(&mounts);
1297     build_fs_mount_list(&mounts, &local_err);
1298     if (local_err) {
1299         error_propagate(errp, local_err);
1300         return 0;
1301     }
1302
1303     QTAILQ_FOREACH(mount, &mounts, next) {
1304         logged = false;
1305         fd = qemu_open(mount->dirname, O_RDONLY);
1306         if (fd == -1) {
1307             continue;
1308         }
1309         /* we have no way of knowing whether a filesystem was actually unfrozen
1310          * as a result of a successful call to FITHAW, only that if an error
1311          * was returned the filesystem was *not* unfrozen by that particular
1312          * call.
1313          *
1314          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1315          * to unfreeze, continuing issuing FITHAW until an error is returned,
1316          * in which case either the filesystem is in an unfreezable state, or,
1317          * more likely, it was thawed previously (and remains so afterward).
1318          *
1319          * also, since the most recent successful call is the one that did
1320          * the actual unfreeze, we can use this to provide an accurate count
1321          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1322          * may * be useful for determining whether a filesystem was unfrozen
1323          * during the freeze/thaw phase by a process other than qemu-ga.
1324          */
1325         do {
1326             ret = ioctl(fd, FITHAW);
1327             if (ret == 0 && !logged) {
1328                 i++;
1329                 logged = true;
1330             }
1331         } while (ret == 0);
1332         close(fd);
1333     }
1334
1335     ga_unset_frozen(ga_state);
1336     free_fs_mount_list(&mounts);
1337
1338     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1339
1340     return i;
1341 }
1342
1343 static void guest_fsfreeze_cleanup(void)
1344 {
1345     Error *err = NULL;
1346
1347     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1348         qmp_guest_fsfreeze_thaw(&err);
1349         if (err) {
1350             slog("failed to clean up frozen filesystems: %s",
1351                  error_get_pretty(err));
1352             error_free(err);
1353         }
1354     }
1355 }
1356 #endif /* CONFIG_FSFREEZE */
1357
1358 #if defined(CONFIG_FSTRIM)
1359 /*
1360  * Walk list of mounted file systems in the guest, and trim them.
1361  */
1362 GuestFilesystemTrimResponse *
1363 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1364 {
1365     GuestFilesystemTrimResponse *response;
1366     GuestFilesystemTrimResultList *list;
1367     GuestFilesystemTrimResult *result;
1368     int ret = 0;
1369     FsMountList mounts;
1370     struct FsMount *mount;
1371     int fd;
1372     Error *local_err = NULL;
1373     struct fstrim_range r;
1374
1375     slog("guest-fstrim called");
1376
1377     QTAILQ_INIT(&mounts);
1378     build_fs_mount_list(&mounts, &local_err);
1379     if (local_err) {
1380         error_propagate(errp, local_err);
1381         return NULL;
1382     }
1383
1384     response = g_malloc0(sizeof(*response));
1385
1386     QTAILQ_FOREACH(mount, &mounts, next) {
1387         result = g_malloc0(sizeof(*result));
1388         result->path = g_strdup(mount->dirname);
1389
1390         list = g_malloc0(sizeof(*list));
1391         list->value = result;
1392         list->next = response->paths;
1393         response->paths = list;
1394
1395         fd = qemu_open(mount->dirname, O_RDONLY);
1396         if (fd == -1) {
1397             result->error = g_strdup_printf("failed to open: %s",
1398                                             strerror(errno));
1399             result->has_error = true;
1400             continue;
1401         }
1402
1403         /* We try to cull filesystems we know won't work in advance, but other
1404          * filesystems may not implement fstrim for less obvious reasons.
1405          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1406          * will be reported (e.g. CD-ROMs).
1407          * Any other error means an unexpected error.
1408          */
1409         r.start = 0;
1410         r.len = -1;
1411         r.minlen = has_minimum ? minimum : 0;
1412         ret = ioctl(fd, FITRIM, &r);
1413         if (ret == -1) {
1414             result->has_error = true;
1415             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1416                 result->error = g_strdup("trim not supported");
1417             } else {
1418                 result->error = g_strdup_printf("failed to trim: %s",
1419                                                 strerror(errno));
1420             }
1421             close(fd);
1422             continue;
1423         }
1424
1425         result->has_minimum = true;
1426         result->minimum = r.minlen;
1427         result->has_trimmed = true;
1428         result->trimmed = r.len;
1429         close(fd);
1430     }
1431
1432     free_fs_mount_list(&mounts);
1433     return response;
1434 }
1435 #endif /* CONFIG_FSTRIM */
1436
1437
1438 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1439 #define SUSPEND_SUPPORTED 0
1440 #define SUSPEND_NOT_SUPPORTED 1
1441
1442 static void bios_supports_mode(const char *pmutils_bin, const char *pmutils_arg,
1443                                const char *sysfile_str, Error **errp)
1444 {
1445     Error *local_err = NULL;
1446     char *pmutils_path;
1447     pid_t pid;
1448     int status;
1449
1450     pmutils_path = g_find_program_in_path(pmutils_bin);
1451
1452     pid = fork();
1453     if (!pid) {
1454         char buf[32]; /* hopefully big enough */
1455         ssize_t ret;
1456         int fd;
1457
1458         setsid();
1459         reopen_fd_to_null(0);
1460         reopen_fd_to_null(1);
1461         reopen_fd_to_null(2);
1462
1463         if (pmutils_path) {
1464             execle(pmutils_path, pmutils_bin, pmutils_arg, NULL, environ);
1465         }
1466
1467         /*
1468          * If we get here either pm-utils is not installed or execle() has
1469          * failed. Let's try the manual method if the caller wants it.
1470          */
1471
1472         if (!sysfile_str) {
1473             _exit(SUSPEND_NOT_SUPPORTED);
1474         }
1475
1476         fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1477         if (fd < 0) {
1478             _exit(SUSPEND_NOT_SUPPORTED);
1479         }
1480
1481         ret = read(fd, buf, sizeof(buf)-1);
1482         if (ret <= 0) {
1483             _exit(SUSPEND_NOT_SUPPORTED);
1484         }
1485         buf[ret] = '\0';
1486
1487         if (strstr(buf, sysfile_str)) {
1488             _exit(SUSPEND_SUPPORTED);
1489         }
1490
1491         _exit(SUSPEND_NOT_SUPPORTED);
1492     } else if (pid < 0) {
1493         error_setg_errno(errp, errno, "failed to create child process");
1494         goto out;
1495     }
1496
1497     ga_wait_child(pid, &status, &local_err);
1498     if (local_err) {
1499         error_propagate(errp, local_err);
1500         goto out;
1501     }
1502
1503     if (!WIFEXITED(status)) {
1504         error_setg(errp, "child process has terminated abnormally");
1505         goto out;
1506     }
1507
1508     switch (WEXITSTATUS(status)) {
1509     case SUSPEND_SUPPORTED:
1510         goto out;
1511     case SUSPEND_NOT_SUPPORTED:
1512         error_setg(errp,
1513                    "the requested suspend mode is not supported by the guest");
1514         goto out;
1515     default:
1516         error_setg(errp,
1517                    "the helper program '%s' returned an unexpected exit status"
1518                    " code (%d)", pmutils_path, WEXITSTATUS(status));
1519         goto out;
1520     }
1521
1522 out:
1523     g_free(pmutils_path);
1524 }
1525
1526 static void guest_suspend(const char *pmutils_bin, const char *sysfile_str,
1527                           Error **errp)
1528 {
1529     Error *local_err = NULL;
1530     char *pmutils_path;
1531     pid_t pid;
1532     int status;
1533
1534     pmutils_path = g_find_program_in_path(pmutils_bin);
1535
1536     pid = fork();
1537     if (pid == 0) {
1538         /* child */
1539         int fd;
1540
1541         setsid();
1542         reopen_fd_to_null(0);
1543         reopen_fd_to_null(1);
1544         reopen_fd_to_null(2);
1545
1546         if (pmutils_path) {
1547             execle(pmutils_path, pmutils_bin, NULL, environ);
1548         }
1549
1550         /*
1551          * If we get here either pm-utils is not installed or execle() has
1552          * failed. Let's try the manual method if the caller wants it.
1553          */
1554
1555         if (!sysfile_str) {
1556             _exit(EXIT_FAILURE);
1557         }
1558
1559         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1560         if (fd < 0) {
1561             _exit(EXIT_FAILURE);
1562         }
1563
1564         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1565             _exit(EXIT_FAILURE);
1566         }
1567
1568         _exit(EXIT_SUCCESS);
1569     } else if (pid < 0) {
1570         error_setg_errno(errp, errno, "failed to create child process");
1571         goto out;
1572     }
1573
1574     ga_wait_child(pid, &status, &local_err);
1575     if (local_err) {
1576         error_propagate(errp, local_err);
1577         goto out;
1578     }
1579
1580     if (!WIFEXITED(status)) {
1581         error_setg(errp, "child process has terminated abnormally");
1582         goto out;
1583     }
1584
1585     if (WEXITSTATUS(status)) {
1586         error_setg(errp, "child process has failed to suspend");
1587         goto out;
1588     }
1589
1590 out:
1591     g_free(pmutils_path);
1592 }
1593
1594 void qmp_guest_suspend_disk(Error **errp)
1595 {
1596     Error *local_err = NULL;
1597
1598     bios_supports_mode("pm-is-supported", "--hibernate", "disk", &local_err);
1599     if (local_err) {
1600         error_propagate(errp, local_err);
1601         return;
1602     }
1603
1604     guest_suspend("pm-hibernate", "disk", errp);
1605 }
1606
1607 void qmp_guest_suspend_ram(Error **errp)
1608 {
1609     Error *local_err = NULL;
1610
1611     bios_supports_mode("pm-is-supported", "--suspend", "mem", &local_err);
1612     if (local_err) {
1613         error_propagate(errp, local_err);
1614         return;
1615     }
1616
1617     guest_suspend("pm-suspend", "mem", errp);
1618 }
1619
1620 void qmp_guest_suspend_hybrid(Error **errp)
1621 {
1622     Error *local_err = NULL;
1623
1624     bios_supports_mode("pm-is-supported", "--suspend-hybrid", NULL,
1625                        &local_err);
1626     if (local_err) {
1627         error_propagate(errp, local_err);
1628         return;
1629     }
1630
1631     guest_suspend("pm-suspend-hybrid", NULL, errp);
1632 }
1633
1634 static GuestNetworkInterfaceList *
1635 guest_find_interface(GuestNetworkInterfaceList *head,
1636                      const char *name)
1637 {
1638     for (; head; head = head->next) {
1639         if (strcmp(head->value->name, name) == 0) {
1640             break;
1641         }
1642     }
1643
1644     return head;
1645 }
1646
1647 static int guest_get_network_stats(const char *name,
1648                        GuestNetworkInterfaceStat *stats)
1649 {
1650     int name_len;
1651     char const *devinfo = "/proc/net/dev";
1652     FILE *fp;
1653     char *line = NULL, *colon;
1654     size_t n = 0;
1655     fp = fopen(devinfo, "r");
1656     if (!fp) {
1657         return -1;
1658     }
1659     name_len = strlen(name);
1660     while (getline(&line, &n, fp) != -1) {
1661         long long dummy;
1662         long long rx_bytes;
1663         long long rx_packets;
1664         long long rx_errs;
1665         long long rx_dropped;
1666         long long tx_bytes;
1667         long long tx_packets;
1668         long long tx_errs;
1669         long long tx_dropped;
1670         char *trim_line;
1671         trim_line = g_strchug(line);
1672         if (trim_line[0] == '\0') {
1673             continue;
1674         }
1675         colon = strchr(trim_line, ':');
1676         if (!colon) {
1677             continue;
1678         }
1679         if (colon - name_len  == trim_line &&
1680            strncmp(trim_line, name, name_len) == 0) {
1681             if (sscanf(colon + 1,
1682                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
1683                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
1684                   &dummy, &dummy, &dummy, &dummy,
1685                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
1686                   &dummy, &dummy, &dummy, &dummy) != 16) {
1687                 continue;
1688             }
1689             stats->rx_bytes = rx_bytes;
1690             stats->rx_packets = rx_packets;
1691             stats->rx_errs = rx_errs;
1692             stats->rx_dropped = rx_dropped;
1693             stats->tx_bytes = tx_bytes;
1694             stats->tx_packets = tx_packets;
1695             stats->tx_errs = tx_errs;
1696             stats->tx_dropped = tx_dropped;
1697             fclose(fp);
1698             g_free(line);
1699             return 0;
1700         }
1701     }
1702     fclose(fp);
1703     g_free(line);
1704     g_debug("/proc/net/dev: Interface '%s' not found", name);
1705     return -1;
1706 }
1707
1708 /*
1709  * Build information about guest interfaces
1710  */
1711 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1712 {
1713     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1714     struct ifaddrs *ifap, *ifa;
1715
1716     if (getifaddrs(&ifap) < 0) {
1717         error_setg_errno(errp, errno, "getifaddrs failed");
1718         goto error;
1719     }
1720
1721     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1722         GuestNetworkInterfaceList *info;
1723         GuestIpAddressList **address_list = NULL, *address_item = NULL;
1724         GuestNetworkInterfaceStat  *interface_stat = NULL;
1725         char addr4[INET_ADDRSTRLEN];
1726         char addr6[INET6_ADDRSTRLEN];
1727         int sock;
1728         struct ifreq ifr;
1729         unsigned char *mac_addr;
1730         void *p;
1731
1732         g_debug("Processing %s interface", ifa->ifa_name);
1733
1734         info = guest_find_interface(head, ifa->ifa_name);
1735
1736         if (!info) {
1737             info = g_malloc0(sizeof(*info));
1738             info->value = g_malloc0(sizeof(*info->value));
1739             info->value->name = g_strdup(ifa->ifa_name);
1740
1741             if (!cur_item) {
1742                 head = cur_item = info;
1743             } else {
1744                 cur_item->next = info;
1745                 cur_item = info;
1746             }
1747         }
1748
1749         if (!info->value->has_hardware_address &&
1750             ifa->ifa_flags & SIOCGIFHWADDR) {
1751             /* we haven't obtained HW address yet */
1752             sock = socket(PF_INET, SOCK_STREAM, 0);
1753             if (sock == -1) {
1754                 error_setg_errno(errp, errno, "failed to create socket");
1755                 goto error;
1756             }
1757
1758             memset(&ifr, 0, sizeof(ifr));
1759             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1760             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1761                 error_setg_errno(errp, errno,
1762                                  "failed to get MAC address of %s",
1763                                  ifa->ifa_name);
1764                 close(sock);
1765                 goto error;
1766             }
1767
1768             close(sock);
1769             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
1770
1771             info->value->hardware_address =
1772                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
1773                                 (int) mac_addr[0], (int) mac_addr[1],
1774                                 (int) mac_addr[2], (int) mac_addr[3],
1775                                 (int) mac_addr[4], (int) mac_addr[5]);
1776
1777             info->value->has_hardware_address = true;
1778         }
1779
1780         if (ifa->ifa_addr &&
1781             ifa->ifa_addr->sa_family == AF_INET) {
1782             /* interface with IPv4 address */
1783             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1784             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
1785                 error_setg_errno(errp, errno, "inet_ntop failed");
1786                 goto error;
1787             }
1788
1789             address_item = g_malloc0(sizeof(*address_item));
1790             address_item->value = g_malloc0(sizeof(*address_item->value));
1791             address_item->value->ip_address = g_strdup(addr4);
1792             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
1793
1794             if (ifa->ifa_netmask) {
1795                 /* Count the number of set bits in netmask.
1796                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1797                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
1798                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
1799             }
1800         } else if (ifa->ifa_addr &&
1801                    ifa->ifa_addr->sa_family == AF_INET6) {
1802             /* interface with IPv6 address */
1803             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
1804             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
1805                 error_setg_errno(errp, errno, "inet_ntop failed");
1806                 goto error;
1807             }
1808
1809             address_item = g_malloc0(sizeof(*address_item));
1810             address_item->value = g_malloc0(sizeof(*address_item->value));
1811             address_item->value->ip_address = g_strdup(addr6);
1812             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
1813
1814             if (ifa->ifa_netmask) {
1815                 /* Count the number of set bits in netmask.
1816                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1817                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
1818                 address_item->value->prefix =
1819                     ctpop32(((uint32_t *) p)[0]) +
1820                     ctpop32(((uint32_t *) p)[1]) +
1821                     ctpop32(((uint32_t *) p)[2]) +
1822                     ctpop32(((uint32_t *) p)[3]);
1823             }
1824         }
1825
1826         if (!address_item) {
1827             continue;
1828         }
1829
1830         address_list = &info->value->ip_addresses;
1831
1832         while (*address_list && (*address_list)->next) {
1833             address_list = &(*address_list)->next;
1834         }
1835
1836         if (!*address_list) {
1837             *address_list = address_item;
1838         } else {
1839             (*address_list)->next = address_item;
1840         }
1841
1842         info->value->has_ip_addresses = true;
1843
1844         if (!info->value->has_statistics) {
1845             interface_stat = g_malloc0(sizeof(*interface_stat));
1846             if (guest_get_network_stats(info->value->name,
1847                 interface_stat) == -1) {
1848                 info->value->has_statistics = false;
1849                 g_free(interface_stat);
1850             } else {
1851                 info->value->statistics = interface_stat;
1852                 info->value->has_statistics = true;
1853             }
1854         }
1855     }
1856
1857     freeifaddrs(ifap);
1858     return head;
1859
1860 error:
1861     freeifaddrs(ifap);
1862     qapi_free_GuestNetworkInterfaceList(head);
1863     return NULL;
1864 }
1865
1866 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
1867
1868 static long sysconf_exact(int name, const char *name_str, Error **errp)
1869 {
1870     long ret;
1871
1872     errno = 0;
1873     ret = sysconf(name);
1874     if (ret == -1) {
1875         if (errno == 0) {
1876             error_setg(errp, "sysconf(%s): value indefinite", name_str);
1877         } else {
1878             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
1879         }
1880     }
1881     return ret;
1882 }
1883
1884 /* Transfer online/offline status between @vcpu and the guest system.
1885  *
1886  * On input either @errp or *@errp must be NULL.
1887  *
1888  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
1889  * - R: vcpu->logical_id
1890  * - W: vcpu->online
1891  * - W: vcpu->can_offline
1892  *
1893  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
1894  * - R: vcpu->logical_id
1895  * - R: vcpu->online
1896  *
1897  * Written members remain unmodified on error.
1898  */
1899 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
1900                           Error **errp)
1901 {
1902     char *dirpath;
1903     int dirfd;
1904
1905     dirpath = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
1906                               vcpu->logical_id);
1907     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
1908     if (dirfd == -1) {
1909         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
1910     } else {
1911         static const char fn[] = "online";
1912         int fd;
1913         int res;
1914
1915         fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
1916         if (fd == -1) {
1917             if (errno != ENOENT) {
1918                 error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
1919             } else if (sys2vcpu) {
1920                 vcpu->online = true;
1921                 vcpu->can_offline = false;
1922             } else if (!vcpu->online) {
1923                 error_setg(errp, "logical processor #%" PRId64 " can't be "
1924                            "offlined", vcpu->logical_id);
1925             } /* otherwise pretend successful re-onlining */
1926         } else {
1927             unsigned char status;
1928
1929             res = pread(fd, &status, 1, 0);
1930             if (res == -1) {
1931                 error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
1932             } else if (res == 0) {
1933                 error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
1934                            fn);
1935             } else if (sys2vcpu) {
1936                 vcpu->online = (status != '0');
1937                 vcpu->can_offline = true;
1938             } else if (vcpu->online != (status != '0')) {
1939                 status = '0' + vcpu->online;
1940                 if (pwrite(fd, &status, 1, 0) == -1) {
1941                     error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
1942                                      fn);
1943                 }
1944             } /* otherwise pretend successful re-(on|off)-lining */
1945
1946             res = close(fd);
1947             g_assert(res == 0);
1948         }
1949
1950         res = close(dirfd);
1951         g_assert(res == 0);
1952     }
1953
1954     g_free(dirpath);
1955 }
1956
1957 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
1958 {
1959     int64_t current;
1960     GuestLogicalProcessorList *head, **link;
1961     long sc_max;
1962     Error *local_err = NULL;
1963
1964     current = 0;
1965     head = NULL;
1966     link = &head;
1967     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
1968
1969     while (local_err == NULL && current < sc_max) {
1970         GuestLogicalProcessor *vcpu;
1971         GuestLogicalProcessorList *entry;
1972
1973         vcpu = g_malloc0(sizeof *vcpu);
1974         vcpu->logical_id = current++;
1975         vcpu->has_can_offline = true; /* lolspeak ftw */
1976         transfer_vcpu(vcpu, true, &local_err);
1977
1978         entry = g_malloc0(sizeof *entry);
1979         entry->value = vcpu;
1980
1981         *link = entry;
1982         link = &entry->next;
1983     }
1984
1985     if (local_err == NULL) {
1986         /* there's no guest with zero VCPUs */
1987         g_assert(head != NULL);
1988         return head;
1989     }
1990
1991     qapi_free_GuestLogicalProcessorList(head);
1992     error_propagate(errp, local_err);
1993     return NULL;
1994 }
1995
1996 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
1997 {
1998     int64_t processed;
1999     Error *local_err = NULL;
2000
2001     processed = 0;
2002     while (vcpus != NULL) {
2003         transfer_vcpu(vcpus->value, false, &local_err);
2004         if (local_err != NULL) {
2005             break;
2006         }
2007         ++processed;
2008         vcpus = vcpus->next;
2009     }
2010
2011     if (local_err != NULL) {
2012         if (processed == 0) {
2013             error_propagate(errp, local_err);
2014         } else {
2015             error_free(local_err);
2016         }
2017     }
2018
2019     return processed;
2020 }
2021
2022 void qmp_guest_set_user_password(const char *username,
2023                                  const char *password,
2024                                  bool crypted,
2025                                  Error **errp)
2026 {
2027     Error *local_err = NULL;
2028     char *passwd_path = NULL;
2029     pid_t pid;
2030     int status;
2031     int datafd[2] = { -1, -1 };
2032     char *rawpasswddata = NULL;
2033     size_t rawpasswdlen;
2034     char *chpasswddata = NULL;
2035     size_t chpasswdlen;
2036
2037     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2038     if (!rawpasswddata) {
2039         return;
2040     }
2041     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2042     rawpasswddata[rawpasswdlen] = '\0';
2043
2044     if (strchr(rawpasswddata, '\n')) {
2045         error_setg(errp, "forbidden characters in raw password");
2046         goto out;
2047     }
2048
2049     if (strchr(username, '\n') ||
2050         strchr(username, ':')) {
2051         error_setg(errp, "forbidden characters in username");
2052         goto out;
2053     }
2054
2055     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2056     chpasswdlen = strlen(chpasswddata);
2057
2058     passwd_path = g_find_program_in_path("chpasswd");
2059
2060     if (!passwd_path) {
2061         error_setg(errp, "cannot find 'passwd' program in PATH");
2062         goto out;
2063     }
2064
2065     if (pipe(datafd) < 0) {
2066         error_setg(errp, "cannot create pipe FDs");
2067         goto out;
2068     }
2069
2070     pid = fork();
2071     if (pid == 0) {
2072         close(datafd[1]);
2073         /* child */
2074         setsid();
2075         dup2(datafd[0], 0);
2076         reopen_fd_to_null(1);
2077         reopen_fd_to_null(2);
2078
2079         if (crypted) {
2080             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2081         } else {
2082             execle(passwd_path, "chpasswd", NULL, environ);
2083         }
2084         _exit(EXIT_FAILURE);
2085     } else if (pid < 0) {
2086         error_setg_errno(errp, errno, "failed to create child process");
2087         goto out;
2088     }
2089     close(datafd[0]);
2090     datafd[0] = -1;
2091
2092     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2093         error_setg_errno(errp, errno, "cannot write new account password");
2094         goto out;
2095     }
2096     close(datafd[1]);
2097     datafd[1] = -1;
2098
2099     ga_wait_child(pid, &status, &local_err);
2100     if (local_err) {
2101         error_propagate(errp, local_err);
2102         goto out;
2103     }
2104
2105     if (!WIFEXITED(status)) {
2106         error_setg(errp, "child process has terminated abnormally");
2107         goto out;
2108     }
2109
2110     if (WEXITSTATUS(status)) {
2111         error_setg(errp, "child process has failed to set user password");
2112         goto out;
2113     }
2114
2115 out:
2116     g_free(chpasswddata);
2117     g_free(rawpasswddata);
2118     g_free(passwd_path);
2119     if (datafd[0] != -1) {
2120         close(datafd[0]);
2121     }
2122     if (datafd[1] != -1) {
2123         close(datafd[1]);
2124     }
2125 }
2126
2127 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2128                                int size, Error **errp)
2129 {
2130     int fd;
2131     int res;
2132
2133     errno = 0;
2134     fd = openat(dirfd, pathname, O_RDONLY);
2135     if (fd == -1) {
2136         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2137         return;
2138     }
2139
2140     res = pread(fd, buf, size, 0);
2141     if (res == -1) {
2142         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2143     } else if (res == 0) {
2144         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2145     }
2146     close(fd);
2147 }
2148
2149 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2150                                 const char *buf, int size, Error **errp)
2151 {
2152     int fd;
2153
2154     errno = 0;
2155     fd = openat(dirfd, pathname, O_WRONLY);
2156     if (fd == -1) {
2157         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2158         return;
2159     }
2160
2161     if (pwrite(fd, buf, size, 0) == -1) {
2162         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2163     }
2164
2165     close(fd);
2166 }
2167
2168 /* Transfer online/offline status between @mem_blk and the guest system.
2169  *
2170  * On input either @errp or *@errp must be NULL.
2171  *
2172  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2173  * - R: mem_blk->phys_index
2174  * - W: mem_blk->online
2175  * - W: mem_blk->can_offline
2176  *
2177  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2178  * - R: mem_blk->phys_index
2179  * - R: mem_blk->online
2180  *-  R: mem_blk->can_offline
2181  * Written members remain unmodified on error.
2182  */
2183 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2184                                   GuestMemoryBlockResponse *result,
2185                                   Error **errp)
2186 {
2187     char *dirpath;
2188     int dirfd;
2189     char *status;
2190     Error *local_err = NULL;
2191
2192     if (!sys2memblk) {
2193         DIR *dp;
2194
2195         if (!result) {
2196             error_setg(errp, "Internal error, 'result' should not be NULL");
2197             return;
2198         }
2199         errno = 0;
2200         dp = opendir("/sys/devices/system/memory/");
2201          /* if there is no 'memory' directory in sysfs,
2202          * we think this VM does not support online/offline memory block,
2203          * any other solution?
2204          */
2205         if (!dp) {
2206             if (errno == ENOENT) {
2207                 result->response =
2208                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2209             }
2210             goto out1;
2211         }
2212         closedir(dp);
2213     }
2214
2215     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2216                               mem_blk->phys_index);
2217     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2218     if (dirfd == -1) {
2219         if (sys2memblk) {
2220             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2221         } else {
2222             if (errno == ENOENT) {
2223                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2224             } else {
2225                 result->response =
2226                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2227             }
2228         }
2229         g_free(dirpath);
2230         goto out1;
2231     }
2232     g_free(dirpath);
2233
2234     status = g_malloc0(10);
2235     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2236     if (local_err) {
2237         /* treat with sysfs file that not exist in old kernel */
2238         if (errno == ENOENT) {
2239             error_free(local_err);
2240             if (sys2memblk) {
2241                 mem_blk->online = true;
2242                 mem_blk->can_offline = false;
2243             } else if (!mem_blk->online) {
2244                 result->response =
2245                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2246             }
2247         } else {
2248             if (sys2memblk) {
2249                 error_propagate(errp, local_err);
2250             } else {
2251                 result->response =
2252                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2253             }
2254         }
2255         goto out2;
2256     }
2257
2258     if (sys2memblk) {
2259         char removable = '0';
2260
2261         mem_blk->online = (strncmp(status, "online", 6) == 0);
2262
2263         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2264         if (local_err) {
2265             /* if no 'removable' file, it doesn't support offline mem blk */
2266             if (errno == ENOENT) {
2267                 error_free(local_err);
2268                 mem_blk->can_offline = false;
2269             } else {
2270                 error_propagate(errp, local_err);
2271             }
2272         } else {
2273             mem_blk->can_offline = (removable != '0');
2274         }
2275     } else {
2276         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2277             const char *new_state = mem_blk->online ? "online" : "offline";
2278
2279             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2280                                 &local_err);
2281             if (local_err) {
2282                 error_free(local_err);
2283                 result->response =
2284                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2285                 goto out2;
2286             }
2287
2288             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2289             result->has_error_code = false;
2290         } /* otherwise pretend successful re-(on|off)-lining */
2291     }
2292     g_free(status);
2293     close(dirfd);
2294     return;
2295
2296 out2:
2297     g_free(status);
2298     close(dirfd);
2299 out1:
2300     if (!sys2memblk) {
2301         result->has_error_code = true;
2302         result->error_code = errno;
2303     }
2304 }
2305
2306 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2307 {
2308     GuestMemoryBlockList *head, **link;
2309     Error *local_err = NULL;
2310     struct dirent *de;
2311     DIR *dp;
2312
2313     head = NULL;
2314     link = &head;
2315
2316     dp = opendir("/sys/devices/system/memory/");
2317     if (!dp) {
2318         /* it's ok if this happens to be a system that doesn't expose
2319          * memory blocks via sysfs, but otherwise we should report
2320          * an error
2321          */
2322         if (errno != ENOENT) {
2323             error_setg_errno(errp, errno, "Can't open directory"
2324                              "\"/sys/devices/system/memory/\"");
2325         }
2326         return NULL;
2327     }
2328
2329     /* Note: the phys_index of memory block may be discontinuous,
2330      * this is because a memblk is the unit of the Sparse Memory design, which
2331      * allows discontinuous memory ranges (ex. NUMA), so here we should
2332      * traverse the memory block directory.
2333      */
2334     while ((de = readdir(dp)) != NULL) {
2335         GuestMemoryBlock *mem_blk;
2336         GuestMemoryBlockList *entry;
2337
2338         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2339             !(de->d_type & DT_DIR)) {
2340             continue;
2341         }
2342
2343         mem_blk = g_malloc0(sizeof *mem_blk);
2344         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2345         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2346         mem_blk->has_can_offline = true; /* lolspeak ftw */
2347         transfer_memory_block(mem_blk, true, NULL, &local_err);
2348
2349         entry = g_malloc0(sizeof *entry);
2350         entry->value = mem_blk;
2351
2352         *link = entry;
2353         link = &entry->next;
2354     }
2355
2356     closedir(dp);
2357     if (local_err == NULL) {
2358         /* there's no guest with zero memory blocks */
2359         if (head == NULL) {
2360             error_setg(errp, "guest reported zero memory blocks!");
2361         }
2362         return head;
2363     }
2364
2365     qapi_free_GuestMemoryBlockList(head);
2366     error_propagate(errp, local_err);
2367     return NULL;
2368 }
2369
2370 GuestMemoryBlockResponseList *
2371 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2372 {
2373     GuestMemoryBlockResponseList *head, **link;
2374     Error *local_err = NULL;
2375
2376     head = NULL;
2377     link = &head;
2378
2379     while (mem_blks != NULL) {
2380         GuestMemoryBlockResponse *result;
2381         GuestMemoryBlockResponseList *entry;
2382         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2383
2384         result = g_malloc0(sizeof(*result));
2385         result->phys_index = current_mem_blk->phys_index;
2386         transfer_memory_block(current_mem_blk, false, result, &local_err);
2387         if (local_err) { /* should never happen */
2388             goto err;
2389         }
2390         entry = g_malloc0(sizeof *entry);
2391         entry->value = result;
2392
2393         *link = entry;
2394         link = &entry->next;
2395         mem_blks = mem_blks->next;
2396     }
2397
2398     return head;
2399 err:
2400     qapi_free_GuestMemoryBlockResponseList(head);
2401     error_propagate(errp, local_err);
2402     return NULL;
2403 }
2404
2405 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2406 {
2407     Error *local_err = NULL;
2408     char *dirpath;
2409     int dirfd;
2410     char *buf;
2411     GuestMemoryBlockInfo *info;
2412
2413     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2414     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2415     if (dirfd == -1) {
2416         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2417         g_free(dirpath);
2418         return NULL;
2419     }
2420     g_free(dirpath);
2421
2422     buf = g_malloc0(20);
2423     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2424     close(dirfd);
2425     if (local_err) {
2426         g_free(buf);
2427         error_propagate(errp, local_err);
2428         return NULL;
2429     }
2430
2431     info = g_new0(GuestMemoryBlockInfo, 1);
2432     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2433
2434     g_free(buf);
2435
2436     return info;
2437 }
2438
2439 #else /* defined(__linux__) */
2440
2441 void qmp_guest_suspend_disk(Error **errp)
2442 {
2443     error_setg(errp, QERR_UNSUPPORTED);
2444 }
2445
2446 void qmp_guest_suspend_ram(Error **errp)
2447 {
2448     error_setg(errp, QERR_UNSUPPORTED);
2449 }
2450
2451 void qmp_guest_suspend_hybrid(Error **errp)
2452 {
2453     error_setg(errp, QERR_UNSUPPORTED);
2454 }
2455
2456 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2457 {
2458     error_setg(errp, QERR_UNSUPPORTED);
2459     return NULL;
2460 }
2461
2462 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2463 {
2464     error_setg(errp, QERR_UNSUPPORTED);
2465     return NULL;
2466 }
2467
2468 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2469 {
2470     error_setg(errp, QERR_UNSUPPORTED);
2471     return -1;
2472 }
2473
2474 void qmp_guest_set_user_password(const char *username,
2475                                  const char *password,
2476                                  bool crypted,
2477                                  Error **errp)
2478 {
2479     error_setg(errp, QERR_UNSUPPORTED);
2480 }
2481
2482 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2483 {
2484     error_setg(errp, QERR_UNSUPPORTED);
2485     return NULL;
2486 }
2487
2488 GuestMemoryBlockResponseList *
2489 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2490 {
2491     error_setg(errp, QERR_UNSUPPORTED);
2492     return NULL;
2493 }
2494
2495 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2496 {
2497     error_setg(errp, QERR_UNSUPPORTED);
2498     return NULL;
2499 }
2500
2501 #endif
2502
2503 #if !defined(CONFIG_FSFREEZE)
2504
2505 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2506 {
2507     error_setg(errp, QERR_UNSUPPORTED);
2508     return NULL;
2509 }
2510
2511 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2512 {
2513     error_setg(errp, QERR_UNSUPPORTED);
2514
2515     return 0;
2516 }
2517
2518 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2519 {
2520     error_setg(errp, QERR_UNSUPPORTED);
2521
2522     return 0;
2523 }
2524
2525 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2526                                        strList *mountpoints,
2527                                        Error **errp)
2528 {
2529     error_setg(errp, QERR_UNSUPPORTED);
2530
2531     return 0;
2532 }
2533
2534 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2535 {
2536     error_setg(errp, QERR_UNSUPPORTED);
2537
2538     return 0;
2539 }
2540 #endif /* CONFIG_FSFREEZE */
2541
2542 #if !defined(CONFIG_FSTRIM)
2543 GuestFilesystemTrimResponse *
2544 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
2545 {
2546     error_setg(errp, QERR_UNSUPPORTED);
2547     return NULL;
2548 }
2549 #endif
2550
2551 /* add unsupported commands to the blacklist */
2552 GList *ga_command_blacklist_init(GList *blacklist)
2553 {
2554 #if !defined(__linux__)
2555     {
2556         const char *list[] = {
2557             "guest-suspend-disk", "guest-suspend-ram",
2558             "guest-suspend-hybrid", "guest-network-get-interfaces",
2559             "guest-get-vcpus", "guest-set-vcpus",
2560             "guest-get-memory-blocks", "guest-set-memory-blocks",
2561             "guest-get-memory-block-size", NULL};
2562         char **p = (char **)list;
2563
2564         while (*p) {
2565             blacklist = g_list_append(blacklist, g_strdup(*p++));
2566         }
2567     }
2568 #endif
2569
2570 #if !defined(CONFIG_FSFREEZE)
2571     {
2572         const char *list[] = {
2573             "guest-get-fsinfo", "guest-fsfreeze-status",
2574             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
2575             "guest-fsfreeze-thaw", "guest-get-fsinfo", NULL};
2576         char **p = (char **)list;
2577
2578         while (*p) {
2579             blacklist = g_list_append(blacklist, g_strdup(*p++));
2580         }
2581     }
2582 #endif
2583
2584 #if !defined(CONFIG_FSTRIM)
2585     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
2586 #endif
2587
2588     return blacklist;
2589 }
2590
2591 /* register init/cleanup routines for stateful command groups */
2592 void ga_command_state_init(GAState *s, GACommandState *cs)
2593 {
2594 #if defined(CONFIG_FSFREEZE)
2595     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
2596 #endif
2597 }
2598
2599 #ifdef HAVE_UTMPX
2600
2601 #define QGA_MICRO_SECOND_TO_SECOND 1000000
2602
2603 static double ga_get_login_time(struct utmpx *user_info)
2604 {
2605     double seconds = (double)user_info->ut_tv.tv_sec;
2606     double useconds = (double)user_info->ut_tv.tv_usec;
2607     useconds /= QGA_MICRO_SECOND_TO_SECOND;
2608     return seconds + useconds;
2609 }
2610
2611 GuestUserList *qmp_guest_get_users(Error **err)
2612 {
2613     GHashTable *cache = NULL;
2614     GuestUserList *head = NULL, *cur_item = NULL;
2615     struct utmpx *user_info = NULL;
2616     gpointer value = NULL;
2617     GuestUser *user = NULL;
2618     GuestUserList *item = NULL;
2619     double login_time = 0;
2620
2621     cache = g_hash_table_new(g_str_hash, g_str_equal);
2622     setutxent();
2623
2624     for (;;) {
2625         user_info = getutxent();
2626         if (user_info == NULL) {
2627             break;
2628         } else if (user_info->ut_type != USER_PROCESS) {
2629             continue;
2630         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
2631             value = g_hash_table_lookup(cache, user_info->ut_user);
2632             user = (GuestUser *)value;
2633             login_time = ga_get_login_time(user_info);
2634             /* We're ensuring the earliest login time to be sent */
2635             if (login_time < user->login_time) {
2636                 user->login_time = login_time;
2637             }
2638             continue;
2639         }
2640
2641         item = g_new0(GuestUserList, 1);
2642         item->value = g_new0(GuestUser, 1);
2643         item->value->user = g_strdup(user_info->ut_user);
2644         item->value->login_time = ga_get_login_time(user_info);
2645
2646         g_hash_table_insert(cache, item->value->user, item->value);
2647
2648         if (!cur_item) {
2649             head = cur_item = item;
2650         } else {
2651             cur_item->next = item;
2652             cur_item = item;
2653         }
2654     }
2655     endutxent();
2656     g_hash_table_destroy(cache);
2657     return head;
2658 }
2659
2660 #else
2661
2662 GuestUserList *qmp_guest_get_users(Error **errp)
2663 {
2664     error_setg(errp, QERR_UNSUPPORTED);
2665     return NULL;
2666 }
2667
2668 #endif
2669
2670 /* Replace escaped special characters with theire real values. The replacement
2671  * is done in place -- returned value is in the original string.
2672  */
2673 static void ga_osrelease_replace_special(gchar *value)
2674 {
2675     gchar *p, *p2, quote;
2676
2677     /* Trim the string at first space or semicolon if it is not enclosed in
2678      * single or double quotes. */
2679     if ((value[0] != '"') || (value[0] == '\'')) {
2680         p = strchr(value, ' ');
2681         if (p != NULL) {
2682             *p = 0;
2683         }
2684         p = strchr(value, ';');
2685         if (p != NULL) {
2686             *p = 0;
2687         }
2688         return;
2689     }
2690
2691     quote = value[0];
2692     p2 = value;
2693     p = value + 1;
2694     while (*p != 0) {
2695         if (*p == '\\') {
2696             p++;
2697             switch (*p) {
2698             case '$':
2699             case '\'':
2700             case '"':
2701             case '\\':
2702             case '`':
2703                 break;
2704             default:
2705                 /* Keep literal backslash followed by whatever is there */
2706                 p--;
2707                 break;
2708             }
2709         } else if (*p == quote) {
2710             *p2 = 0;
2711             break;
2712         }
2713         *(p2++) = *(p++);
2714     }
2715 }
2716
2717 static GKeyFile *ga_parse_osrelease(const char *fname)
2718 {
2719     gchar *content = NULL;
2720     gchar *content2 = NULL;
2721     GError *err = NULL;
2722     GKeyFile *keys = g_key_file_new();
2723     const char *group = "[os-release]\n";
2724
2725     if (!g_file_get_contents(fname, &content, NULL, &err)) {
2726         slog("failed to read '%s', error: %s", fname, err->message);
2727         goto fail;
2728     }
2729
2730     if (!g_utf8_validate(content, -1, NULL)) {
2731         slog("file is not utf-8 encoded: %s", fname);
2732         goto fail;
2733     }
2734     content2 = g_strdup_printf("%s%s", group, content);
2735
2736     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
2737                                    &err)) {
2738         slog("failed to parse file '%s', error: %s", fname, err->message);
2739         goto fail;
2740     }
2741
2742     g_free(content);
2743     g_free(content2);
2744     return keys;
2745
2746 fail:
2747     g_error_free(err);
2748     g_free(content);
2749     g_free(content2);
2750     g_key_file_free(keys);
2751     return NULL;
2752 }
2753
2754 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
2755 {
2756     GuestOSInfo *info = NULL;
2757     struct utsname kinfo;
2758     GKeyFile *osrelease = NULL;
2759     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
2760
2761     info = g_new0(GuestOSInfo, 1);
2762
2763     if (uname(&kinfo) != 0) {
2764         error_setg_errno(errp, errno, "uname failed");
2765     } else {
2766         info->has_kernel_version = true;
2767         info->kernel_version = g_strdup(kinfo.version);
2768         info->has_kernel_release = true;
2769         info->kernel_release = g_strdup(kinfo.release);
2770         info->has_machine = true;
2771         info->machine = g_strdup(kinfo.machine);
2772     }
2773
2774     if (qga_os_release != NULL) {
2775         osrelease = ga_parse_osrelease(qga_os_release);
2776     } else {
2777         osrelease = ga_parse_osrelease("/etc/os-release");
2778         if (osrelease == NULL) {
2779             osrelease = ga_parse_osrelease("/usr/lib/os-release");
2780         }
2781     }
2782
2783     if (osrelease != NULL) {
2784         char *value;
2785
2786 #define GET_FIELD(field, osfield) do { \
2787     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
2788     if (value != NULL) { \
2789         ga_osrelease_replace_special(value); \
2790         info->has_ ## field = true; \
2791         info->field = value; \
2792     } \
2793 } while (0)
2794         GET_FIELD(id, "ID");
2795         GET_FIELD(name, "NAME");
2796         GET_FIELD(pretty_name, "PRETTY_NAME");
2797         GET_FIELD(version, "VERSION");
2798         GET_FIELD(version_id, "VERSION_ID");
2799         GET_FIELD(variant, "VARIANT");
2800         GET_FIELD(variant_id, "VARIANT_ID");
2801 #undef GET_FIELD
2802
2803         g_key_file_free(osrelease);
2804     }
2805
2806     return info;
2807 }