qga/commands-posix.c

   1 /*
   2  * QEMU Guest Agent POSIX-specific command implementations
   3  *
   4  * Copyright IBM Corp. 2011
   5  *
   6  * Authors:
   7  *  Michael Roth      <mdroth@linux.vnet.ibm.com>
   8  *  Michal Privoznik  <mprivozn@redhat.com>
   9  *
  10  * This work is licensed under the terms of the GNU GPL, version 2 or later.
  11  * See the COPYING file in the top-level directory.
  12  */
  13
  14 #include "qemu/osdep.h"
  15 #include <sys/ioctl.h>
  16 #include <sys/utsname.h>
  17 #include <sys/wait.h>
  18 #include <dirent.h>
  19 #include "qemu-common.h"
  20 #include "guest-agent-core.h"
  21 #include "qga-qapi-commands.h"
  22 #include "qapi/error.h"
  23 #include "qapi/qmp/qerror.h"
  24 #include "qemu/queue.h"
  25 #include "qemu/host-utils.h"
  26 #include "qemu/sockets.h"
  27 #include "qemu/base64.h"
  28 #include "qemu/cutils.h"
  29 #include "commands-common.h"
  30
  31 #ifdef HAVE_UTMPX
  32 #include <utmpx.h>
  33 #endif
  34
  35 #ifndef CONFIG_HAS_ENVIRON
  36 #ifdef __APPLE__
  37 #include <crt_externs.h>
  38 #define environ (*_NSGetEnviron())
  39 #else
  40 extern char **environ;
  41 #endif
  42 #endif
  43
  44 #if defined(__linux__)
  45 #include <mntent.h>
  46 #include <linux/fs.h>
  47 #include <ifaddrs.h>
  48 #include <arpa/inet.h>
  49 #include <sys/socket.h>
  50 #include <net/if.h>
  51 #include <sys/statvfs.h>
  52
  53 #ifdef CONFIG_LIBUDEV
  54 #include <libudev.h>
  55 #endif
  56
  57 #ifdef FIFREEZE
  58 #define CONFIG_FSFREEZE
  59 #endif
  60 #ifdef FITRIM
  61 #define CONFIG_FSTRIM
  62 #endif
  63 #endif
  64
  65 static void ga_wait_child(pid_t pid, int *status, Error **errp)
  66 {
  67     pid_t rpid;
  68
  69     *status = 0;
  70
  71     do {
  72         rpid = waitpid(pid, status, 0);
  73     } while (rpid == -1 && errno == EINTR);
  74
  75     if (rpid == -1) {
  76         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
  77                          pid);
  78         return;
  79     }
  80
  81     g_assert(rpid == pid);
  82 }
  83
  84 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
  85 {
  86     const char *shutdown_flag;
  87     Error *local_err = NULL;
  88     pid_t pid;
  89     int status;
  90
  91     slog("guest-shutdown called, mode: %s", mode);
  92     if (!has_mode || strcmp(mode, "powerdown") == 0) {
  93         shutdown_flag = "-P";
  94     } else if (strcmp(mode, "halt") == 0) {
  95         shutdown_flag = "-H";
  96     } else if (strcmp(mode, "reboot") == 0) {
  97         shutdown_flag = "-r";
  98     } else {
  99         error_setg(errp,
 100                    "mode is invalid (valid values are: halt|powerdown|reboot");
 101         return;
 102     }
 103
 104     pid = fork();
 105     if (pid == 0) {
 106         /* child, start the shutdown */
 107         setsid();
 108         reopen_fd_to_null(0);
 109         reopen_fd_to_null(1);
 110         reopen_fd_to_null(2);
 111
 112         execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
 113                "hypervisor initiated shutdown", (char*)NULL, environ);
 114         _exit(EXIT_FAILURE);
 115     } else if (pid < 0) {
 116         error_setg_errno(errp, errno, "failed to create child process");
 117         return;
 118     }
 119
 120     ga_wait_child(pid, &status, &local_err);
 121     if (local_err) {
 122         error_propagate(errp, local_err);
 123         return;
 124     }
 125
 126     if (!WIFEXITED(status)) {
 127         error_setg(errp, "child process has terminated abnormally");
 128         return;
 129     }
 130
 131     if (WEXITSTATUS(status)) {
 132         error_setg(errp, "child process has failed to shutdown");
 133         return;
 134     }
 135
 136     /* succeeded */
 137 }
 138
 139 int64_t qmp_guest_get_time(Error **errp)
 140 {
 141    int ret;
 142    qemu_timeval tq;
 143
 144    ret = qemu_gettimeofday(&tq);
 145    if (ret < 0) {
 146        error_setg_errno(errp, errno, "Failed to get time");
 147        return -1;
 148    }
 149
 150    return tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
 151 }
 152
 153 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
 154 {
 155     int ret;
 156     int status;
 157     pid_t pid;
 158     Error *local_err = NULL;
 159     struct timeval tv;
 160     static const char hwclock_path[] = "/sbin/hwclock";
 161     static int hwclock_available = -1;
 162
 163     if (hwclock_available < 0) {
 164         hwclock_available = (access(hwclock_path, X_OK) == 0);
 165     }
 166
 167     if (!hwclock_available) {
 168         error_setg(errp, QERR_UNSUPPORTED);
 169         return;
 170     }
 171
 172     /* If user has passed a time, validate and set it. */
 173     if (has_time) {
 174         GDate date = { 0, };
 175
 176         /* year-2038 will overflow in case time_t is 32bit */
 177         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
 178             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
 179             return;
 180         }
 181
 182         tv.tv_sec = time_ns / 1000000000;
 183         tv.tv_usec = (time_ns % 1000000000) / 1000;
 184         g_date_set_time_t(&date, tv.tv_sec);
 185         if (date.year < 1970 || date.year >= 2070) {
 186             error_setg_errno(errp, errno, "Invalid time");
 187             return;
 188         }
 189
 190         ret = settimeofday(&tv, NULL);
 191         if (ret < 0) {
 192             error_setg_errno(errp, errno, "Failed to set time to guest");
 193             return;
 194         }
 195     }
 196
 197     /* Now, if user has passed a time to set and the system time is set, we
 198      * just need to synchronize the hardware clock. However, if no time was
 199      * passed, user is requesting the opposite: set the system time from the
 200      * hardware clock (RTC). */
 201     pid = fork();
 202     if (pid == 0) {
 203         setsid();
 204         reopen_fd_to_null(0);
 205         reopen_fd_to_null(1);
 206         reopen_fd_to_null(2);
 207
 208         /* Use '/sbin/hwclock -w' to set RTC from the system time,
 209          * or '/sbin/hwclock -s' to set the system time from RTC. */
 210         execle(hwclock_path, "hwclock", has_time ? "-w" : "-s",
 211                NULL, environ);
 212         _exit(EXIT_FAILURE);
 213     } else if (pid < 0) {
 214         error_setg_errno(errp, errno, "failed to create child process");
 215         return;
 216     }
 217
 218     ga_wait_child(pid, &status, &local_err);
 219     if (local_err) {
 220         error_propagate(errp, local_err);
 221         return;
 222     }
 223
 224     if (!WIFEXITED(status)) {
 225         error_setg(errp, "child process has terminated abnormally");
 226         return;
 227     }
 228
 229     if (WEXITSTATUS(status)) {
 230         error_setg(errp, "hwclock failed to set hardware clock to system time");
 231         return;
 232     }
 233 }
 234
 235 typedef enum {
 236     RW_STATE_NEW,
 237     RW_STATE_READING,
 238     RW_STATE_WRITING,
 239 } RwState;
 240
 241 struct GuestFileHandle {
 242     uint64_t id;
 243     FILE *fh;
 244     RwState state;
 245     QTAILQ_ENTRY(GuestFileHandle) next;
 246 };
 247
 248 static struct {
 249     QTAILQ_HEAD(, GuestFileHandle) filehandles;
 250 } guest_file_state = {
 251     .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
 252 };
 253
 254 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
 255 {
 256     GuestFileHandle *gfh;
 257     int64_t handle;
 258
 259     handle = ga_get_fd_handle(ga_state, errp);
 260     if (handle < 0) {
 261         return -1;
 262     }
 263
 264     gfh = g_new0(GuestFileHandle, 1);
 265     gfh->id = handle;
 266     gfh->fh = fh;
 267     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
 268
 269     return handle;
 270 }
 271
 272 GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
 273 {
 274     GuestFileHandle *gfh;
 275
 276     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
 277     {
 278         if (gfh->id == id) {
 279             return gfh;
 280         }
 281     }
 282
 283     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
 284     return NULL;
 285 }
 286
 287 typedef const char * const ccpc;
 288
 289 #ifndef O_BINARY
 290 #define O_BINARY 0
 291 #endif
 292
 293 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
 294 static const struct {
 295     ccpc *forms;
 296     int oflag_base;
 297 } guest_file_open_modes[] = {
 298     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
 299     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
 300     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
 301     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
 302     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
 303     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
 304     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
 305     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
 306     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
 307     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
 308     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
 309     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
 310 };
 311
 312 static int
 313 find_open_flag(const char *mode_str, Error **errp)
 314 {
 315     unsigned mode;
 316
 317     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
 318         ccpc *form;
 319
 320         form = guest_file_open_modes[mode].forms;
 321         while (*form != NULL && strcmp(*form, mode_str) != 0) {
 322             ++form;
 323         }
 324         if (*form != NULL) {
 325             break;
 326         }
 327     }
 328
 329     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
 330         error_setg(errp, "invalid file open mode '%s'", mode_str);
 331         return -1;
 332     }
 333     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
 334 }
 335
 336 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
 337                                S_IRGRP | S_IWGRP | \
 338                                S_IROTH | S_IWOTH)
 339
 340 static FILE *
 341 safe_open_or_create(const char *path, const char *mode, Error **errp)
 342 {
 343     Error *local_err = NULL;
 344     int oflag;
 345
 346     oflag = find_open_flag(mode, &local_err);
 347     if (local_err == NULL) {
 348         int fd;
 349
 350         /* If the caller wants / allows creation of a new file, we implement it
 351          * with a two step process: open() + (open() / fchmod()).
 352          *
 353          * First we insist on creating the file exclusively as a new file. If
 354          * that succeeds, we're free to set any file-mode bits on it. (The
 355          * motivation is that we want to set those file-mode bits independently
 356          * of the current umask.)
 357          *
 358          * If the exclusive creation fails because the file already exists
 359          * (EEXIST is not possible for any other reason), we just attempt to
 360          * open the file, but in this case we won't be allowed to change the
 361          * file-mode bits on the preexistent file.
 362          *
 363          * The pathname should never disappear between the two open()s in
 364          * practice. If it happens, then someone very likely tried to race us.
 365          * In this case just go ahead and report the ENOENT from the second
 366          * open() to the caller.
 367          *
 368          * If the caller wants to open a preexistent file, then the first
 369          * open() is decisive and its third argument is ignored, and the second
 370          * open() and the fchmod() are never called.
 371          */
 372         fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
 373         if (fd == -1 && errno == EEXIST) {
 374             oflag &= ~(unsigned)O_CREAT;
 375             fd = open(path, oflag);
 376         }
 377
 378         if (fd == -1) {
 379             error_setg_errno(&local_err, errno, "failed to open file '%s' "
 380                              "(mode: '%s')", path, mode);
 381         } else {
 382             qemu_set_cloexec(fd);
 383
 384             if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
 385                 error_setg_errno(&local_err, errno, "failed to set permission "
 386                                  "0%03o on new file '%s' (mode: '%s')",
 387                                  (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
 388             } else {
 389                 FILE *f;
 390
 391                 f = fdopen(fd, mode);
 392                 if (f == NULL) {
 393                     error_setg_errno(&local_err, errno, "failed to associate "
 394                                      "stdio stream with file descriptor %d, "
 395                                      "file '%s' (mode: '%s')", fd, path, mode);
 396                 } else {
 397                     return f;
 398                 }
 399             }
 400
 401             close(fd);
 402             if (oflag & O_CREAT) {
 403                 unlink(path);
 404             }
 405         }
 406     }
 407
 408     error_propagate(errp, local_err);
 409     return NULL;
 410 }
 411
 412 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
 413                             Error **errp)
 414 {
 415     FILE *fh;
 416     Error *local_err = NULL;
 417     int64_t handle;
 418
 419     if (!has_mode) {
 420         mode = "r";
 421     }
 422     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
 423     fh = safe_open_or_create(path, mode, &local_err);
 424     if (local_err != NULL) {
 425         error_propagate(errp, local_err);
 426         return -1;
 427     }
 428
 429     /* set fd non-blocking to avoid common use cases (like reading from a
 430      * named pipe) from hanging the agent
 431      */
 432     qemu_set_nonblock(fileno(fh));
 433
 434     handle = guest_file_handle_add(fh, errp);
 435     if (handle < 0) {
 436         fclose(fh);
 437         return -1;
 438     }
 439
 440     slog("guest-file-open, handle: %" PRId64, handle);
 441     return handle;
 442 }
 443
 444 void qmp_guest_file_close(int64_t handle, Error **errp)
 445 {
 446     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 447     int ret;
 448
 449     slog("guest-file-close called, handle: %" PRId64, handle);
 450     if (!gfh) {
 451         return;
 452     }
 453
 454     ret = fclose(gfh->fh);
 455     if (ret == EOF) {
 456         error_setg_errno(errp, errno, "failed to close handle");
 457         return;
 458     }
 459
 460     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
 461     g_free(gfh);
 462 }
 463
 464 GuestFileRead *guest_file_read_unsafe(GuestFileHandle *gfh,
 465                                       int64_t count, Error **errp)
 466 {
 467     GuestFileRead *read_data = NULL;
 468     guchar *buf;
 469     FILE *fh = gfh->fh;
 470     size_t read_count;
 471
 472     /* explicitly flush when switching from writing to reading */
 473     if (gfh->state == RW_STATE_WRITING) {
 474         int ret = fflush(fh);
 475         if (ret == EOF) {
 476             error_setg_errno(errp, errno, "failed to flush file");
 477             return NULL;
 478         }
 479         gfh->state = RW_STATE_NEW;
 480     }
 481
 482     buf = g_malloc0(count+1);
 483     read_count = fread(buf, 1, count, fh);
 484     if (ferror(fh)) {
 485         error_setg_errno(errp, errno, "failed to read file");
 486     } else {
 487         buf[read_count] = 0;
 488         read_data = g_new0(GuestFileRead, 1);
 489         read_data->count = read_count;
 490         read_data->eof = feof(fh);
 491         if (read_count) {
 492             read_data->buf_b64 = g_base64_encode(buf, read_count);
 493         }
 494         gfh->state = RW_STATE_READING;
 495     }
 496     g_free(buf);
 497     clearerr(fh);
 498
 499     return read_data;
 500 }
 501
 502 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
 503                                      bool has_count, int64_t count,
 504                                      Error **errp)
 505 {
 506     GuestFileWrite *write_data = NULL;
 507     guchar *buf;
 508     gsize buf_len;
 509     int write_count;
 510     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 511     FILE *fh;
 512
 513     if (!gfh) {
 514         return NULL;
 515     }
 516
 517     fh = gfh->fh;
 518
 519     if (gfh->state == RW_STATE_READING) {
 520         int ret = fseek(fh, 0, SEEK_CUR);
 521         if (ret == -1) {
 522             error_setg_errno(errp, errno, "failed to seek file");
 523             return NULL;
 524         }
 525         gfh->state = RW_STATE_NEW;
 526     }
 527
 528     buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
 529     if (!buf) {
 530         return NULL;
 531     }
 532
 533     if (!has_count) {
 534         count = buf_len;
 535     } else if (count < 0 || count > buf_len) {
 536         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
 537                    count);
 538         g_free(buf);
 539         return NULL;
 540     }
 541
 542     write_count = fwrite(buf, 1, count, fh);
 543     if (ferror(fh)) {
 544         error_setg_errno(errp, errno, "failed to write to file");
 545         slog("guest-file-write failed, handle: %" PRId64, handle);
 546     } else {
 547         write_data = g_new0(GuestFileWrite, 1);
 548         write_data->count = write_count;
 549         write_data->eof = feof(fh);
 550         gfh->state = RW_STATE_WRITING;
 551     }
 552     g_free(buf);
 553     clearerr(fh);
 554
 555     return write_data;
 556 }
 557
 558 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
 559                                           GuestFileWhence *whence_code,
 560                                           Error **errp)
 561 {
 562     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 563     GuestFileSeek *seek_data = NULL;
 564     FILE *fh;
 565     int ret;
 566     int whence;
 567     Error *err = NULL;
 568
 569     if (!gfh) {
 570         return NULL;
 571     }
 572
 573     /* We stupidly exposed 'whence':'int' in our qapi */
 574     whence = ga_parse_whence(whence_code, &err);
 575     if (err) {
 576         error_propagate(errp, err);
 577         return NULL;
 578     }
 579
 580     fh = gfh->fh;
 581     ret = fseek(fh, offset, whence);
 582     if (ret == -1) {
 583         error_setg_errno(errp, errno, "failed to seek file");
 584         if (errno == ESPIPE) {
 585             /* file is non-seekable, stdio shouldn't be buffering anyways */
 586             gfh->state = RW_STATE_NEW;
 587         }
 588     } else {
 589         seek_data = g_new0(GuestFileSeek, 1);
 590         seek_data->position = ftell(fh);
 591         seek_data->eof = feof(fh);
 592         gfh->state = RW_STATE_NEW;
 593     }
 594     clearerr(fh);
 595
 596     return seek_data;
 597 }
 598
 599 void qmp_guest_file_flush(int64_t handle, Error **errp)
 600 {
 601     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
 602     FILE *fh;
 603     int ret;
 604
 605     if (!gfh) {
 606         return;
 607     }
 608
 609     fh = gfh->fh;
 610     ret = fflush(fh);
 611     if (ret == EOF) {
 612         error_setg_errno(errp, errno, "failed to flush file");
 613     } else {
 614         gfh->state = RW_STATE_NEW;
 615     }
 616 }
 617
 618 /* linux-specific implementations. avoid this if at all possible. */
 619 #if defined(__linux__)
 620
 621 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
 622 typedef struct FsMount {
 623     char *dirname;
 624     char *devtype;
 625     unsigned int devmajor, devminor;
 626     QTAILQ_ENTRY(FsMount) next;
 627 } FsMount;
 628
 629 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
 630
 631 static void free_fs_mount_list(FsMountList *mounts)
 632 {
 633      FsMount *mount, *temp;
 634
 635      if (!mounts) {
 636          return;
 637      }
 638
 639      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
 640          QTAILQ_REMOVE(mounts, mount, next);
 641          g_free(mount->dirname);
 642          g_free(mount->devtype);
 643          g_free(mount);
 644      }
 645 }
 646
 647 static int dev_major_minor(const char *devpath,
 648                            unsigned int *devmajor, unsigned int *devminor)
 649 {
 650     struct stat st;
 651
 652     *devmajor = 0;
 653     *devminor = 0;
 654
 655     if (stat(devpath, &st) < 0) {
 656         slog("failed to stat device file '%s': %s", devpath, strerror(errno));
 657         return -1;
 658     }
 659     if (S_ISDIR(st.st_mode)) {
 660         /* It is bind mount */
 661         return -2;
 662     }
 663     if (S_ISBLK(st.st_mode)) {
 664         *devmajor = major(st.st_rdev);
 665         *devminor = minor(st.st_rdev);
 666         return 0;
 667     }
 668     return -1;
 669 }
 670
 671 /*
 672  * Walk the mount table and build a list of local file systems
 673  */
 674 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
 675 {
 676     struct mntent *ment;
 677     FsMount *mount;
 678     char const *mtab = "/proc/self/mounts";
 679     FILE *fp;
 680     unsigned int devmajor, devminor;
 681
 682     fp = setmntent(mtab, "r");
 683     if (!fp) {
 684         error_setg(errp, "failed to open mtab file: '%s'", mtab);
 685         return;
 686     }
 687
 688     while ((ment = getmntent(fp))) {
 689         /*
 690          * An entry which device name doesn't start with a '/' is
 691          * either a dummy file system or a network file system.
 692          * Add special handling for smbfs and cifs as is done by
 693          * coreutils as well.
 694          */
 695         if ((ment->mnt_fsname[0] != '/') ||
 696             (strcmp(ment->mnt_type, "smbfs") == 0) ||
 697             (strcmp(ment->mnt_type, "cifs") == 0)) {
 698             continue;
 699         }
 700         if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
 701             /* Skip bind mounts */
 702             continue;
 703         }
 704
 705         mount = g_new0(FsMount, 1);
 706         mount->dirname = g_strdup(ment->mnt_dir);
 707         mount->devtype = g_strdup(ment->mnt_type);
 708         mount->devmajor = devmajor;
 709         mount->devminor = devminor;
 710
 711         QTAILQ_INSERT_TAIL(mounts, mount, next);
 712     }
 713
 714     endmntent(fp);
 715 }
 716
 717 static void decode_mntname(char *name, int len)
 718 {
 719     int i, j = 0;
 720     for (i = 0; i <= len; i++) {
 721         if (name[i] != '\\') {
 722             name[j++] = name[i];
 723         } else if (name[i + 1] == '\\') {
 724             name[j++] = '\\';
 725             i++;
 726         } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
 727                    name[i + 2] >= '0' && name[i + 2] <= '7' &&
 728                    name[i + 3] >= '0' && name[i + 3] <= '7') {
 729             name[j++] = (name[i + 1] - '0') * 64 +
 730                         (name[i + 2] - '0') * 8 +
 731                         (name[i + 3] - '0');
 732             i += 3;
 733         } else {
 734             name[j++] = name[i];
 735         }
 736     }
 737 }
 738
 739 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
 740 {
 741     FsMount *mount;
 742     char const *mountinfo = "/proc/self/mountinfo";
 743     FILE *fp;
 744     char *line = NULL, *dash;
 745     size_t n;
 746     char check;
 747     unsigned int devmajor, devminor;
 748     int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
 749
 750     fp = fopen(mountinfo, "r");
 751     if (!fp) {
 752         build_fs_mount_list_from_mtab(mounts, errp);
 753         return;
 754     }
 755
 756     while (getline(&line, &n, fp) != -1) {
 757         ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
 758                      &devmajor, &devminor, &dir_s, &dir_e, &check);
 759         if (ret < 3) {
 760             continue;
 761         }
 762         dash = strstr(line + dir_e, " - ");
 763         if (!dash) {
 764             continue;
 765         }
 766         ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
 767                      &type_s, &type_e, &dev_s, &dev_e, &check);
 768         if (ret < 1) {
 769             continue;
 770         }
 771         line[dir_e] = 0;
 772         dash[type_e] = 0;
 773         dash[dev_e] = 0;
 774         decode_mntname(line + dir_s, dir_e - dir_s);
 775         decode_mntname(dash + dev_s, dev_e - dev_s);
 776         if (devmajor == 0) {
 777             /* btrfs reports major number = 0 */
 778             if (strcmp("btrfs", dash + type_s) != 0 ||
 779                 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
 780                 continue;
 781             }
 782         }
 783
 784         mount = g_new0(FsMount, 1);
 785         mount->dirname = g_strdup(line + dir_s);
 786         mount->devtype = g_strdup(dash + type_s);
 787         mount->devmajor = devmajor;
 788         mount->devminor = devminor;
 789
 790         QTAILQ_INSERT_TAIL(mounts, mount, next);
 791     }
 792     free(line);
 793
 794     fclose(fp);
 795 }
 796 #endif
 797
 798 #if defined(CONFIG_FSFREEZE)
 799
 800 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
 801 {
 802     char *path;
 803     char *dpath;
 804     char *driver = NULL;
 805     char buf[PATH_MAX];
 806     ssize_t len;
 807
 808     path = g_strndup(syspath, pathlen);
 809     dpath = g_strdup_printf("%s/driver", path);
 810     len = readlink(dpath, buf, sizeof(buf) - 1);
 811     if (len != -1) {
 812         buf[len] = 0;
 813         driver = g_path_get_basename(buf);
 814     }
 815     g_free(dpath);
 816     g_free(path);
 817     return driver;
 818 }
 819
 820 static int compare_uint(const void *_a, const void *_b)
 821 {
 822     unsigned int a = *(unsigned int *)_a;
 823     unsigned int b = *(unsigned int *)_b;
 824
 825     return a < b ? -1 : a > b ? 1 : 0;
 826 }
 827
 828 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
 829 static int build_hosts(char const *syspath, char const *host, bool ata,
 830                        unsigned int *hosts, int hosts_max, Error **errp)
 831 {
 832     char *path;
 833     DIR *dir;
 834     struct dirent *entry;
 835     int i = 0;
 836
 837     path = g_strndup(syspath, host - syspath);
 838     dir = opendir(path);
 839     if (!dir) {
 840         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
 841         g_free(path);
 842         return -1;
 843     }
 844
 845     while (i < hosts_max) {
 846         entry = readdir(dir);
 847         if (!entry) {
 848             break;
 849         }
 850         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
 851             ++i;
 852         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
 853             ++i;
 854         }
 855     }
 856
 857     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
 858
 859     g_free(path);
 860     closedir(dir);
 861     return i;
 862 }
 863
 864 /*
 865  * Store disk device info for devices on the PCI bus.
 866  * Returns true if information has been stored, or false for failure.
 867  */
 868 static bool build_guest_fsinfo_for_pci_dev(char const *syspath,
 869                                            GuestDiskAddress *disk,
 870                                            Error **errp)
 871 {
 872     unsigned int pci[4], host, hosts[8], tgt[3];
 873     int i, nhosts = 0, pcilen;
 874     GuestPCIAddress *pciaddr = disk->pci_controller;
 875     bool has_ata = false, has_host = false, has_tgt = false;
 876     char *p, *q, *driver = NULL;
 877     bool ret = false;
 878
 879     p = strstr(syspath, "/devices/pci");
 880     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
 881                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
 882         g_debug("only pci device is supported: sysfs path '%s'", syspath);
 883         return false;
 884     }
 885
 886     p += 12 + pcilen;
 887     while (true) {
 888         driver = get_pci_driver(syspath, p - syspath, errp);
 889         if (driver && (g_str_equal(driver, "ata_piix") ||
 890                        g_str_equal(driver, "sym53c8xx") ||
 891                        g_str_equal(driver, "virtio-pci") ||
 892                        g_str_equal(driver, "ahci"))) {
 893             break;
 894         }
 895
 896         g_free(driver);
 897         if (sscanf(p, "/%x:%x:%x.%x%n",
 898                           pci, pci + 1, pci + 2, pci + 3, &pcilen) == 4) {
 899             p += pcilen;
 900             continue;
 901         }
 902
 903         g_debug("unsupported driver or sysfs path '%s'", syspath);
 904         return false;
 905     }
 906
 907     p = strstr(syspath, "/target");
 908     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
 909                     tgt, tgt + 1, tgt + 2) == 3) {
 910         has_tgt = true;
 911     }
 912
 913     p = strstr(syspath, "/ata");
 914     if (p) {
 915         q = p + 4;
 916         has_ata = true;
 917     } else {
 918         p = strstr(syspath, "/host");
 919         q = p + 5;
 920     }
 921     if (p && sscanf(q, "%u", &host) == 1) {
 922         has_host = true;
 923         nhosts = build_hosts(syspath, p, has_ata, hosts,
 924                              ARRAY_SIZE(hosts), errp);
 925         if (nhosts < 0) {
 926             goto cleanup;
 927         }
 928     }
 929
 930     pciaddr->domain = pci[0];
 931     pciaddr->bus = pci[1];
 932     pciaddr->slot = pci[2];
 933     pciaddr->function = pci[3];
 934
 935     if (strcmp(driver, "ata_piix") == 0) {
 936         /* a host per ide bus, target*:0:<unit>:0 */
 937         if (!has_host || !has_tgt) {
 938             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 939             goto cleanup;
 940         }
 941         for (i = 0; i < nhosts; i++) {
 942             if (host == hosts[i]) {
 943                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
 944                 disk->bus = i;
 945                 disk->unit = tgt[1];
 946                 break;
 947             }
 948         }
 949         if (i >= nhosts) {
 950             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 951             goto cleanup;
 952         }
 953     } else if (strcmp(driver, "sym53c8xx") == 0) {
 954         /* scsi(LSI Logic): target*:0:<unit>:0 */
 955         if (!has_tgt) {
 956             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 957             goto cleanup;
 958         }
 959         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 960         disk->unit = tgt[1];
 961     } else if (strcmp(driver, "virtio-pci") == 0) {
 962         if (has_tgt) {
 963             /* virtio-scsi: target*:0:0:<unit> */
 964             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
 965             disk->unit = tgt[2];
 966         } else {
 967             /* virtio-blk: 1 disk per 1 device */
 968             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
 969         }
 970     } else if (strcmp(driver, "ahci") == 0) {
 971         /* ahci: 1 host per 1 unit */
 972         if (!has_host || !has_tgt) {
 973             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
 974             goto cleanup;
 975         }
 976         for (i = 0; i < nhosts; i++) {
 977             if (host == hosts[i]) {
 978                 disk->unit = i;
 979                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
 980                 break;
 981             }
 982         }
 983         if (i >= nhosts) {
 984             g_debug("no host for '%s' (driver '%s')", syspath, driver);
 985             goto cleanup;
 986         }
 987     } else {
 988         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
 989         goto cleanup;
 990     }
 991
 992     ret = true;
 993
 994 cleanup:
 995     g_free(driver);
 996     return ret;
 997 }
 998
 999 /*
1000  * Store disk device info for non-PCI virtio devices (for example s390x
1001  * channel I/O devices). Returns true if information has been stored, or
1002  * false for failure.
1003  */
1004 static bool build_guest_fsinfo_for_nonpci_virtio(char const *syspath,
1005                                                  GuestDiskAddress *disk,
1006                                                  Error **errp)
1007 {
1008     unsigned int tgt[3];
1009     char *p;
1010
1011     if (!strstr(syspath, "/virtio") || !strstr(syspath, "/block")) {
1012         g_debug("Unsupported virtio device '%s'", syspath);
1013         return false;
1014     }
1015
1016     p = strstr(syspath, "/target");
1017     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
1018                     &tgt[0], &tgt[1], &tgt[2]) == 3) {
1019         /* virtio-scsi: target*:0:<target>:<unit> */
1020         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
1021         disk->bus = tgt[0];
1022         disk->target = tgt[1];
1023         disk->unit = tgt[2];
1024     } else {
1025         /* virtio-blk: 1 disk per 1 device */
1026         disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
1027     }
1028
1029     return true;
1030 }
1031
1032 /* Store disk device info specified by @sysfs into @fs */
1033 static void build_guest_fsinfo_for_real_device(char const *syspath,
1034                                                GuestFilesystemInfo *fs,
1035                                                Error **errp)
1036 {
1037     GuestDiskAddress *disk;
1038     GuestPCIAddress *pciaddr;
1039     GuestDiskAddressList *list = NULL;
1040     bool has_hwinf;
1041 #ifdef CONFIG_LIBUDEV
1042     struct udev *udev = NULL;
1043     struct udev_device *udevice = NULL;
1044 #endif
1045
1046     pciaddr = g_new0(GuestPCIAddress, 1);
1047     pciaddr->domain = -1;                       /* -1 means field is invalid */
1048     pciaddr->bus = -1;
1049     pciaddr->slot = -1;
1050     pciaddr->function = -1;
1051
1052     disk = g_new0(GuestDiskAddress, 1);
1053     disk->pci_controller = pciaddr;
1054     disk->bus_type = GUEST_DISK_BUS_TYPE_UNKNOWN;
1055
1056     list = g_new0(GuestDiskAddressList, 1);
1057     list->value = disk;
1058
1059 #ifdef CONFIG_LIBUDEV
1060     udev = udev_new();
1061     udevice = udev_device_new_from_syspath(udev, syspath);
1062     if (udev == NULL || udevice == NULL) {
1063         g_debug("failed to query udev");
1064     } else {
1065         const char *devnode, *serial;
1066         devnode = udev_device_get_devnode(udevice);
1067         if (devnode != NULL) {
1068             disk->dev = g_strdup(devnode);
1069             disk->has_dev = true;
1070         }
1071         serial = udev_device_get_property_value(udevice, "ID_SERIAL");
1072         if (serial != NULL && *serial != 0) {
1073             disk->serial = g_strdup(serial);
1074             disk->has_serial = true;
1075         }
1076     }
1077
1078     udev_unref(udev);
1079     udev_device_unref(udevice);
1080 #endif
1081
1082     if (strstr(syspath, "/devices/pci")) {
1083         has_hwinf = build_guest_fsinfo_for_pci_dev(syspath, disk, errp);
1084     } else if (strstr(syspath, "/virtio")) {
1085         has_hwinf = build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1086     } else {
1087         g_debug("Unsupported device type for '%s'", syspath);
1088         has_hwinf = false;
1089     }
1090
1091     if (has_hwinf || disk->has_dev || disk->has_serial) {
1092         list->next = fs->disk;
1093         fs->disk = list;
1094     } else {
1095         qapi_free_GuestDiskAddressList(list);
1096     }
1097 }
1098
1099 static void build_guest_fsinfo_for_device(char const *devpath,
1100                                           GuestFilesystemInfo *fs,
1101                                           Error **errp);
1102
1103 /* Store a list of slave devices of virtual volume specified by @syspath into
1104  * @fs */
1105 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1106                                                   GuestFilesystemInfo *fs,
1107                                                   Error **errp)
1108 {
1109     Error *err = NULL;
1110     DIR *dir;
1111     char *dirpath;
1112     struct dirent *entry;
1113
1114     dirpath = g_strdup_printf("%s/slaves", syspath);
1115     dir = opendir(dirpath);
1116     if (!dir) {
1117         if (errno != ENOENT) {
1118             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1119         }
1120         g_free(dirpath);
1121         return;
1122     }
1123
1124     for (;;) {
1125         errno = 0;
1126         entry = readdir(dir);
1127         if (entry == NULL) {
1128             if (errno) {
1129                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1130             }
1131             break;
1132         }
1133
1134         if (entry->d_type == DT_LNK) {
1135             char *path;
1136
1137             g_debug(" slave device '%s'", entry->d_name);
1138             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1139             build_guest_fsinfo_for_device(path, fs, &err);
1140             g_free(path);
1141
1142             if (err) {
1143                 error_propagate(errp, err);
1144                 break;
1145             }
1146         }
1147     }
1148
1149     g_free(dirpath);
1150     closedir(dir);
1151 }
1152
1153 static bool is_disk_virtual(const char *devpath, Error **errp)
1154 {
1155     g_autofree char *syspath = realpath(devpath, NULL);
1156
1157     if (!syspath) {
1158         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1159         return false;
1160     }
1161     return strstr(syspath, "/devices/virtual/block/") != NULL;
1162 }
1163
1164 /* Dispatch to functions for virtual/real device */
1165 static void build_guest_fsinfo_for_device(char const *devpath,
1166                                           GuestFilesystemInfo *fs,
1167                                           Error **errp)
1168 {
1169     ERRP_GUARD();
1170     g_autofree char *syspath = NULL;
1171     bool is_virtual = false;
1172
1173     syspath = realpath(devpath, NULL);
1174     if (!syspath) {
1175         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1176         return;
1177     }
1178
1179     if (!fs->name) {
1180         fs->name = g_path_get_basename(syspath);
1181     }
1182
1183     g_debug("  parse sysfs path '%s'", syspath);
1184     is_virtual = is_disk_virtual(syspath, errp);
1185     if (*errp != NULL) {
1186         return;
1187     }
1188     if (is_virtual) {
1189         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1190     } else {
1191         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1192     }
1193 }
1194
1195 #ifdef CONFIG_LIBUDEV
1196
1197 /*
1198  * Wrapper around build_guest_fsinfo_for_device() for getting just
1199  * the disk address.
1200  */
1201 static GuestDiskAddress *get_disk_address(const char *syspath, Error **errp)
1202 {
1203     g_autoptr(GuestFilesystemInfo) fs = NULL;
1204
1205     fs = g_new0(GuestFilesystemInfo, 1);
1206     build_guest_fsinfo_for_device(syspath, fs, errp);
1207     if (fs->disk != NULL) {
1208         return g_steal_pointer(&fs->disk->value);
1209     }
1210     return NULL;
1211 }
1212
1213 static char *get_alias_for_syspath(const char *syspath)
1214 {
1215     struct udev *udev = NULL;
1216     struct udev_device *udevice = NULL;
1217     char *ret = NULL;
1218
1219     udev = udev_new();
1220     if (udev == NULL) {
1221         g_debug("failed to query udev");
1222         goto out;
1223     }
1224     udevice = udev_device_new_from_syspath(udev, syspath);
1225     if (udevice == NULL) {
1226         g_debug("failed to query udev for path: %s", syspath);
1227         goto out;
1228     } else {
1229         const char *alias = udev_device_get_property_value(
1230             udevice, "DM_NAME");
1231         /*
1232          * NULL means there was an error and empty string means there is no
1233          * alias. In case of no alias we return NULL instead of empty string.
1234          */
1235         if (alias == NULL) {
1236             g_debug("failed to query udev for device alias for: %s",
1237                 syspath);
1238         } else if (*alias != 0) {
1239             ret = g_strdup(alias);
1240         }
1241     }
1242
1243 out:
1244     udev_unref(udev);
1245     udev_device_unref(udevice);
1246     return ret;
1247 }
1248
1249 static char *get_device_for_syspath(const char *syspath)
1250 {
1251     struct udev *udev = NULL;
1252     struct udev_device *udevice = NULL;
1253     char *ret = NULL;
1254
1255     udev = udev_new();
1256     if (udev == NULL) {
1257         g_debug("failed to query udev");
1258         goto out;
1259     }
1260     udevice = udev_device_new_from_syspath(udev, syspath);
1261     if (udevice == NULL) {
1262         g_debug("failed to query udev for path: %s", syspath);
1263         goto out;
1264     } else {
1265         ret = g_strdup(udev_device_get_devnode(udevice));
1266     }
1267
1268 out:
1269     udev_unref(udev);
1270     udev_device_unref(udevice);
1271     return ret;
1272 }
1273
1274 static void get_disk_deps(const char *disk_dir, GuestDiskInfo *disk)
1275 {
1276     g_autofree char *deps_dir = NULL;
1277     const gchar *dep;
1278     GDir *dp_deps = NULL;
1279
1280     /* List dependent disks */
1281     deps_dir = g_strdup_printf("%s/slaves", disk_dir);
1282     g_debug("  listing entries in: %s", deps_dir);
1283     dp_deps = g_dir_open(deps_dir, 0, NULL);
1284     if (dp_deps == NULL) {
1285         g_debug("failed to list entries in %s", deps_dir);
1286         return;
1287     }
1288     while ((dep = g_dir_read_name(dp_deps)) != NULL) {
1289         g_autofree char *dep_dir = NULL;
1290         strList *dep_item = NULL;
1291         char *dev_name;
1292
1293         /* Add dependent disks */
1294         dep_dir = g_strdup_printf("%s/%s", deps_dir, dep);
1295         dev_name = get_device_for_syspath(dep_dir);
1296         if (dev_name != NULL) {
1297             g_debug("  adding dependent device: %s", dev_name);
1298             dep_item = g_new0(strList, 1);
1299             dep_item->value = dev_name;
1300             dep_item->next = disk->dependents;
1301             disk->dependents = dep_item;
1302         }
1303     }
1304     g_dir_close(dp_deps);
1305 }
1306
1307 /*
1308  * Detect partitions subdirectory, name is "<disk_name><number>" or
1309  * "<disk_name>p<number>"
1310  *
1311  * @disk_name -- last component of /sys path (e.g. sda)
1312  * @disk_dir -- sys path of the disk (e.g. /sys/block/sda)
1313  * @disk_dev -- device node of the disk (e.g. /dev/sda)
1314  */
1315 static GuestDiskInfoList *get_disk_partitions(
1316     GuestDiskInfoList *list,
1317     const char *disk_name, const char *disk_dir,
1318     const char *disk_dev)
1319 {
1320     GuestDiskInfoList *item, *ret = list;
1321     struct dirent *de_disk;
1322     DIR *dp_disk = NULL;
1323     size_t len = strlen(disk_name);
1324
1325     dp_disk = opendir(disk_dir);
1326     while ((de_disk = readdir(dp_disk)) != NULL) {
1327         g_autofree char *partition_dir = NULL;
1328         char *dev_name;
1329         GuestDiskInfo *partition;
1330
1331         if (!(de_disk->d_type & DT_DIR)) {
1332             continue;
1333         }
1334
1335         if (!(strncmp(disk_name, de_disk->d_name, len) == 0 &&
1336             ((*(de_disk->d_name + len) == 'p' &&
1337             isdigit(*(de_disk->d_name + len + 1))) ||
1338                 isdigit(*(de_disk->d_name + len))))) {
1339             continue;
1340         }
1341
1342         partition_dir = g_strdup_printf("%s/%s",
1343             disk_dir, de_disk->d_name);
1344         dev_name = get_device_for_syspath(partition_dir);
1345         if (dev_name == NULL) {
1346             g_debug("Failed to get device name for syspath: %s",
1347                 disk_dir);
1348             continue;
1349         }
1350         partition = g_new0(GuestDiskInfo, 1);
1351         partition->name = dev_name;
1352         partition->partition = true;
1353         /* Add parent disk as dependent for easier tracking of hierarchy */
1354         partition->dependents = g_new0(strList, 1);
1355         partition->dependents->value = g_strdup(disk_dev);
1356
1357         item = g_new0(GuestDiskInfoList, 1);
1358         item->value = partition;
1359         item->next = ret;
1360         ret = item;
1361
1362     }
1363     closedir(dp_disk);
1364
1365     return ret;
1366 }
1367
1368 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1369 {
1370     GuestDiskInfoList *item, *ret = NULL;
1371     GuestDiskInfo *disk;
1372     DIR *dp = NULL;
1373     struct dirent *de = NULL;
1374
1375     g_debug("listing /sys/block directory");
1376     dp = opendir("/sys/block");
1377     if (dp == NULL) {
1378         error_setg_errno(errp, errno, "Can't open directory \"/sys/block\"");
1379         return NULL;
1380     }
1381     while ((de = readdir(dp)) != NULL) {
1382         g_autofree char *disk_dir = NULL, *line = NULL,
1383             *size_path = NULL;
1384         char *dev_name;
1385         Error *local_err = NULL;
1386         if (de->d_type != DT_LNK) {
1387             g_debug("  skipping entry: %s", de->d_name);
1388             continue;
1389         }
1390
1391         /* Check size and skip zero-sized disks */
1392         g_debug("  checking disk size");
1393         size_path = g_strdup_printf("/sys/block/%s/size", de->d_name);
1394         if (!g_file_get_contents(size_path, &line, NULL, NULL)) {
1395             g_debug("  failed to read disk size");
1396             continue;
1397         }
1398         if (g_strcmp0(line, "0\n") == 0) {
1399             g_debug("  skipping zero-sized disk");
1400             continue;
1401         }
1402
1403         g_debug("  adding %s", de->d_name);
1404         disk_dir = g_strdup_printf("/sys/block/%s", de->d_name);
1405         dev_name = get_device_for_syspath(disk_dir);
1406         if (dev_name == NULL) {
1407             g_debug("Failed to get device name for syspath: %s",
1408                 disk_dir);
1409             continue;
1410         }
1411         disk = g_new0(GuestDiskInfo, 1);
1412         disk->name = dev_name;
1413         disk->partition = false;
1414         disk->alias = get_alias_for_syspath(disk_dir);
1415         disk->has_alias = (disk->alias != NULL);
1416         item = g_new0(GuestDiskInfoList, 1);
1417         item->value = disk;
1418         item->next = ret;
1419         ret = item;
1420
1421         /* Get address for non-virtual devices */
1422         bool is_virtual = is_disk_virtual(disk_dir, &local_err);
1423         if (local_err != NULL) {
1424             g_debug("  failed to check disk path, ignoring error: %s",
1425                 error_get_pretty(local_err));
1426             error_free(local_err);
1427             local_err = NULL;
1428             /* Don't try to get the address */
1429             is_virtual = true;
1430         }
1431         if (!is_virtual) {
1432             disk->address = get_disk_address(disk_dir, &local_err);
1433             if (local_err != NULL) {
1434                 g_debug("  failed to get device info, ignoring error: %s",
1435                     error_get_pretty(local_err));
1436                 error_free(local_err);
1437                 local_err = NULL;
1438             } else if (disk->address != NULL) {
1439                 disk->has_address = true;
1440             }
1441         }
1442
1443         get_disk_deps(disk_dir, disk);
1444         ret = get_disk_partitions(ret, de->d_name, disk_dir, dev_name);
1445     }
1446     return ret;
1447 }
1448
1449 #else
1450
1451 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1452 {
1453     error_setg(errp, QERR_UNSUPPORTED);
1454     return NULL;
1455 }
1456
1457 #endif
1458
1459 /* Return a list of the disk device(s)' info which @mount lies on */
1460 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1461                                                Error **errp)
1462 {
1463     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1464     struct statvfs buf;
1465     unsigned long used, nonroot_total, fr_size;
1466     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1467                                     mount->devmajor, mount->devminor);
1468
1469     fs->mountpoint = g_strdup(mount->dirname);
1470     fs->type = g_strdup(mount->devtype);
1471     build_guest_fsinfo_for_device(devpath, fs, errp);
1472
1473     if (statvfs(fs->mountpoint, &buf) == 0) {
1474         fr_size = buf.f_frsize;
1475         used = buf.f_blocks - buf.f_bfree;
1476         nonroot_total = used + buf.f_bavail;
1477         fs->used_bytes = used * fr_size;
1478         fs->total_bytes = nonroot_total * fr_size;
1479
1480         fs->has_total_bytes = true;
1481         fs->has_used_bytes = true;
1482     }
1483
1484     g_free(devpath);
1485
1486     return fs;
1487 }
1488
1489 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1490 {
1491     FsMountList mounts;
1492     struct FsMount *mount;
1493     GuestFilesystemInfoList *new, *ret = NULL;
1494     Error *local_err = NULL;
1495
1496     QTAILQ_INIT(&mounts);
1497     build_fs_mount_list(&mounts, &local_err);
1498     if (local_err) {
1499         error_propagate(errp, local_err);
1500         return NULL;
1501     }
1502
1503     QTAILQ_FOREACH(mount, &mounts, next) {
1504         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1505
1506         new = g_malloc0(sizeof(*ret));
1507         new->value = build_guest_fsinfo(mount, &local_err);
1508         new->next = ret;
1509         ret = new;
1510         if (local_err) {
1511             error_propagate(errp, local_err);
1512             qapi_free_GuestFilesystemInfoList(ret);
1513             ret = NULL;
1514             break;
1515         }
1516     }
1517
1518     free_fs_mount_list(&mounts);
1519     return ret;
1520 }
1521
1522
1523 typedef enum {
1524     FSFREEZE_HOOK_THAW = 0,
1525     FSFREEZE_HOOK_FREEZE,
1526 } FsfreezeHookArg;
1527
1528 static const char *fsfreeze_hook_arg_string[] = {
1529     "thaw",
1530     "freeze",
1531 };
1532
1533 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1534 {
1535     int status;
1536     pid_t pid;
1537     const char *hook;
1538     const char *arg_str = fsfreeze_hook_arg_string[arg];
1539     Error *local_err = NULL;
1540
1541     hook = ga_fsfreeze_hook(ga_state);
1542     if (!hook) {
1543         return;
1544     }
1545     if (access(hook, X_OK) != 0) {
1546         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1547         return;
1548     }
1549
1550     slog("executing fsfreeze hook with arg '%s'", arg_str);
1551     pid = fork();
1552     if (pid == 0) {
1553         setsid();
1554         reopen_fd_to_null(0);
1555         reopen_fd_to_null(1);
1556         reopen_fd_to_null(2);
1557
1558         execle(hook, hook, arg_str, NULL, environ);
1559         _exit(EXIT_FAILURE);
1560     } else if (pid < 0) {
1561         error_setg_errno(errp, errno, "failed to create child process");
1562         return;
1563     }
1564
1565     ga_wait_child(pid, &status, &local_err);
1566     if (local_err) {
1567         error_propagate(errp, local_err);
1568         return;
1569     }
1570
1571     if (!WIFEXITED(status)) {
1572         error_setg(errp, "fsfreeze hook has terminated abnormally");
1573         return;
1574     }
1575
1576     status = WEXITSTATUS(status);
1577     if (status) {
1578         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1579         return;
1580     }
1581 }
1582
1583 /*
1584  * Return status of freeze/thaw
1585  */
1586 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1587 {
1588     if (ga_is_frozen(ga_state)) {
1589         return GUEST_FSFREEZE_STATUS_FROZEN;
1590     }
1591
1592     return GUEST_FSFREEZE_STATUS_THAWED;
1593 }
1594
1595 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1596 {
1597     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1598 }
1599
1600 /*
1601  * Walk list of mounted file systems in the guest, and freeze the ones which
1602  * are real local file systems.
1603  */
1604 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1605                                        strList *mountpoints,
1606                                        Error **errp)
1607 {
1608     int ret = 0, i = 0;
1609     strList *list;
1610     FsMountList mounts;
1611     struct FsMount *mount;
1612     Error *local_err = NULL;
1613     int fd;
1614
1615     slog("guest-fsfreeze called");
1616
1617     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1618     if (local_err) {
1619         error_propagate(errp, local_err);
1620         return -1;
1621     }
1622
1623     QTAILQ_INIT(&mounts);
1624     build_fs_mount_list(&mounts, &local_err);
1625     if (local_err) {
1626         error_propagate(errp, local_err);
1627         return -1;
1628     }
1629
1630     /* cannot risk guest agent blocking itself on a write in this state */
1631     ga_set_frozen(ga_state);
1632
1633     QTAILQ_FOREACH_REVERSE(mount, &mounts, next) {
1634         /* To issue fsfreeze in the reverse order of mounts, check if the
1635          * mount is listed in the list here */
1636         if (has_mountpoints) {
1637             for (list = mountpoints; list; list = list->next) {
1638                 if (strcmp(list->value, mount->dirname) == 0) {
1639                     break;
1640                 }
1641             }
1642             if (!list) {
1643                 continue;
1644             }
1645         }
1646
1647         fd = qemu_open_old(mount->dirname, O_RDONLY);
1648         if (fd == -1) {
1649             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1650             goto error;
1651         }
1652
1653         /* we try to cull filesystems we know won't work in advance, but other
1654          * filesystems may not implement fsfreeze for less obvious reasons.
1655          * these will report EOPNOTSUPP. we simply ignore these when tallying
1656          * the number of frozen filesystems.
1657          * if a filesystem is mounted more than once (aka bind mount) a
1658          * consecutive attempt to freeze an already frozen filesystem will
1659          * return EBUSY.
1660          *
1661          * any other error means a failure to freeze a filesystem we
1662          * expect to be freezable, so return an error in those cases
1663          * and return system to thawed state.
1664          */
1665         ret = ioctl(fd, FIFREEZE);
1666         if (ret == -1) {
1667             if (errno != EOPNOTSUPP && errno != EBUSY) {
1668                 error_setg_errno(errp, errno, "failed to freeze %s",
1669                                  mount->dirname);
1670                 close(fd);
1671                 goto error;
1672             }
1673         } else {
1674             i++;
1675         }
1676         close(fd);
1677     }
1678
1679     free_fs_mount_list(&mounts);
1680     /* We may not issue any FIFREEZE here.
1681      * Just unset ga_state here and ready for the next call.
1682      */
1683     if (i == 0) {
1684         ga_unset_frozen(ga_state);
1685     }
1686     return i;
1687
1688 error:
1689     free_fs_mount_list(&mounts);
1690     qmp_guest_fsfreeze_thaw(NULL);
1691     return 0;
1692 }
1693
1694 /*
1695  * Walk list of frozen file systems in the guest, and thaw them.
1696  */
1697 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1698 {
1699     int ret;
1700     FsMountList mounts;
1701     FsMount *mount;
1702     int fd, i = 0, logged;
1703     Error *local_err = NULL;
1704
1705     QTAILQ_INIT(&mounts);
1706     build_fs_mount_list(&mounts, &local_err);
1707     if (local_err) {
1708         error_propagate(errp, local_err);
1709         return 0;
1710     }
1711
1712     QTAILQ_FOREACH(mount, &mounts, next) {
1713         logged = false;
1714         fd = qemu_open_old(mount->dirname, O_RDONLY);
1715         if (fd == -1) {
1716             continue;
1717         }
1718         /* we have no way of knowing whether a filesystem was actually unfrozen
1719          * as a result of a successful call to FITHAW, only that if an error
1720          * was returned the filesystem was *not* unfrozen by that particular
1721          * call.
1722          *
1723          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1724          * to unfreeze, continuing issuing FITHAW until an error is returned,
1725          * in which case either the filesystem is in an unfreezable state, or,
1726          * more likely, it was thawed previously (and remains so afterward).
1727          *
1728          * also, since the most recent successful call is the one that did
1729          * the actual unfreeze, we can use this to provide an accurate count
1730          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1731          * may * be useful for determining whether a filesystem was unfrozen
1732          * during the freeze/thaw phase by a process other than qemu-ga.
1733          */
1734         do {
1735             ret = ioctl(fd, FITHAW);
1736             if (ret == 0 && !logged) {
1737                 i++;
1738                 logged = true;
1739             }
1740         } while (ret == 0);
1741         close(fd);
1742     }
1743
1744     ga_unset_frozen(ga_state);
1745     free_fs_mount_list(&mounts);
1746
1747     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1748
1749     return i;
1750 }
1751
1752 static void guest_fsfreeze_cleanup(void)
1753 {
1754     Error *err = NULL;
1755
1756     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1757         qmp_guest_fsfreeze_thaw(&err);
1758         if (err) {
1759             slog("failed to clean up frozen filesystems: %s",
1760                  error_get_pretty(err));
1761             error_free(err);
1762         }
1763     }
1764 }
1765 #endif /* CONFIG_FSFREEZE */
1766
1767 #if defined(CONFIG_FSTRIM)
1768 /*
1769  * Walk list of mounted file systems in the guest, and trim them.
1770  */
1771 GuestFilesystemTrimResponse *
1772 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1773 {
1774     GuestFilesystemTrimResponse *response;
1775     GuestFilesystemTrimResultList *list;
1776     GuestFilesystemTrimResult *result;
1777     int ret = 0;
1778     FsMountList mounts;
1779     struct FsMount *mount;
1780     int fd;
1781     Error *local_err = NULL;
1782     struct fstrim_range r;
1783
1784     slog("guest-fstrim called");
1785
1786     QTAILQ_INIT(&mounts);
1787     build_fs_mount_list(&mounts, &local_err);
1788     if (local_err) {
1789         error_propagate(errp, local_err);
1790         return NULL;
1791     }
1792
1793     response = g_malloc0(sizeof(*response));
1794
1795     QTAILQ_FOREACH(mount, &mounts, next) {
1796         result = g_malloc0(sizeof(*result));
1797         result->path = g_strdup(mount->dirname);
1798
1799         list = g_malloc0(sizeof(*list));
1800         list->value = result;
1801         list->next = response->paths;
1802         response->paths = list;
1803
1804         fd = qemu_open_old(mount->dirname, O_RDONLY);
1805         if (fd == -1) {
1806             result->error = g_strdup_printf("failed to open: %s",
1807                                             strerror(errno));
1808             result->has_error = true;
1809             continue;
1810         }
1811
1812         /* We try to cull filesystems we know won't work in advance, but other
1813          * filesystems may not implement fstrim for less obvious reasons.
1814          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1815          * will be reported (e.g. CD-ROMs).
1816          * Any other error means an unexpected error.
1817          */
1818         r.start = 0;
1819         r.len = -1;
1820         r.minlen = has_minimum ? minimum : 0;
1821         ret = ioctl(fd, FITRIM, &r);
1822         if (ret == -1) {
1823             result->has_error = true;
1824             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1825                 result->error = g_strdup("trim not supported");
1826             } else {
1827                 result->error = g_strdup_printf("failed to trim: %s",
1828                                                 strerror(errno));
1829             }
1830             close(fd);
1831             continue;
1832         }
1833
1834         result->has_minimum = true;
1835         result->minimum = r.minlen;
1836         result->has_trimmed = true;
1837         result->trimmed = r.len;
1838         close(fd);
1839     }
1840
1841     free_fs_mount_list(&mounts);
1842     return response;
1843 }
1844 #endif /* CONFIG_FSTRIM */
1845
1846
1847 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1848 #define SUSPEND_SUPPORTED 0
1849 #define SUSPEND_NOT_SUPPORTED 1
1850
1851 typedef enum {
1852     SUSPEND_MODE_DISK = 0,
1853     SUSPEND_MODE_RAM = 1,
1854     SUSPEND_MODE_HYBRID = 2,
1855 } SuspendMode;
1856
1857 /*
1858  * Executes a command in a child process using g_spawn_sync,
1859  * returning an int >= 0 representing the exit status of the
1860  * process.
1861  *
1862  * If the program wasn't found in path, returns -1.
1863  *
1864  * If a problem happened when creating the child process,
1865  * returns -1 and errp is set.
1866  */
1867 static int run_process_child(const char *command[], Error **errp)
1868 {
1869     int exit_status, spawn_flag;
1870     GError *g_err = NULL;
1871     bool success;
1872
1873     spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1874                  G_SPAWN_STDERR_TO_DEV_NULL;
1875
1876     success =  g_spawn_sync(NULL, (char **)command, environ, spawn_flag,
1877                             NULL, NULL, NULL, NULL,
1878                             &exit_status, &g_err);
1879
1880     if (success) {
1881         return WEXITSTATUS(exit_status);
1882     }
1883
1884     if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1885         error_setg(errp, "failed to create child process, error '%s'",
1886                    g_err->message);
1887     }
1888
1889     g_error_free(g_err);
1890     return -1;
1891 }
1892
1893 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1894 {
1895     const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1896                                      "systemd-hybrid-sleep"};
1897     const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1898     int status;
1899
1900     status = run_process_child(cmd, errp);
1901
1902     /*
1903      * systemctl status uses LSB return codes so we can expect
1904      * status > 0 and be ok. To assert if the guest has support
1905      * for the selected suspend mode, status should be < 4. 4 is
1906      * the code for unknown service status, the return value when
1907      * the service does not exist. A common value is status = 3
1908      * (program is not running).
1909      */
1910     if (status > 0 && status < 4) {
1911         return true;
1912     }
1913
1914     return false;
1915 }
1916
1917 static void systemd_suspend(SuspendMode mode, Error **errp)
1918 {
1919     Error *local_err = NULL;
1920     const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1921     const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1922     int status;
1923
1924     status = run_process_child(cmd, &local_err);
1925
1926     if (status == 0) {
1927         return;
1928     }
1929
1930     if ((status == -1) && !local_err) {
1931         error_setg(errp, "the helper program 'systemctl %s' was not found",
1932                    systemctl_args[mode]);
1933         return;
1934     }
1935
1936     if (local_err) {
1937         error_propagate(errp, local_err);
1938     } else {
1939         error_setg(errp, "the helper program 'systemctl %s' returned an "
1940                    "unexpected exit status code (%d)",
1941                    systemctl_args[mode], status);
1942     }
1943 }
1944
1945 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1946 {
1947     Error *local_err = NULL;
1948     const char *pmutils_args[3] = {"--hibernate", "--suspend",
1949                                    "--suspend-hybrid"};
1950     const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1951     int status;
1952
1953     status = run_process_child(cmd, &local_err);
1954
1955     if (status == SUSPEND_SUPPORTED) {
1956         return true;
1957     }
1958
1959     if ((status == -1) && !local_err) {
1960         return false;
1961     }
1962
1963     if (local_err) {
1964         error_propagate(errp, local_err);
1965     } else {
1966         error_setg(errp,
1967                    "the helper program '%s' returned an unexpected exit"
1968                    " status code (%d)", "pm-is-supported", status);
1969     }
1970
1971     return false;
1972 }
1973
1974 static void pmutils_suspend(SuspendMode mode, Error **errp)
1975 {
1976     Error *local_err = NULL;
1977     const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1978                                        "pm-suspend-hybrid"};
1979     const char *cmd[2] = {pmutils_binaries[mode], NULL};
1980     int status;
1981
1982     status = run_process_child(cmd, &local_err);
1983
1984     if (status == 0) {
1985         return;
1986     }
1987
1988     if ((status == -1) && !local_err) {
1989         error_setg(errp, "the helper program '%s' was not found",
1990                    pmutils_binaries[mode]);
1991         return;
1992     }
1993
1994     if (local_err) {
1995         error_propagate(errp, local_err);
1996     } else {
1997         error_setg(errp,
1998                    "the helper program '%s' returned an unexpected exit"
1999                    " status code (%d)", pmutils_binaries[mode], status);
2000     }
2001 }
2002
2003 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
2004 {
2005     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2006     const char *sysfile_str = sysfile_strs[mode];
2007     char buf[32]; /* hopefully big enough */
2008     int fd;
2009     ssize_t ret;
2010
2011     if (!sysfile_str) {
2012         error_setg(errp, "unknown guest suspend mode");
2013         return false;
2014     }
2015
2016     fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
2017     if (fd < 0) {
2018         return false;
2019     }
2020
2021     ret = read(fd, buf, sizeof(buf) - 1);
2022     close(fd);
2023     if (ret <= 0) {
2024         return false;
2025     }
2026     buf[ret] = '\0';
2027
2028     if (strstr(buf, sysfile_str)) {
2029         return true;
2030     }
2031     return false;
2032 }
2033
2034 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
2035 {
2036     Error *local_err = NULL;
2037     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2038     const char *sysfile_str = sysfile_strs[mode];
2039     pid_t pid;
2040     int status;
2041
2042     if (!sysfile_str) {
2043         error_setg(errp, "unknown guest suspend mode");
2044         return;
2045     }
2046
2047     pid = fork();
2048     if (!pid) {
2049         /* child */
2050         int fd;
2051
2052         setsid();
2053         reopen_fd_to_null(0);
2054         reopen_fd_to_null(1);
2055         reopen_fd_to_null(2);
2056
2057         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
2058         if (fd < 0) {
2059             _exit(EXIT_FAILURE);
2060         }
2061
2062         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
2063             _exit(EXIT_FAILURE);
2064         }
2065
2066         _exit(EXIT_SUCCESS);
2067     } else if (pid < 0) {
2068         error_setg_errno(errp, errno, "failed to create child process");
2069         return;
2070     }
2071
2072     ga_wait_child(pid, &status, &local_err);
2073     if (local_err) {
2074         error_propagate(errp, local_err);
2075         return;
2076     }
2077
2078     if (WEXITSTATUS(status)) {
2079         error_setg(errp, "child process has failed to suspend");
2080     }
2081
2082 }
2083
2084 static void guest_suspend(SuspendMode mode, Error **errp)
2085 {
2086     Error *local_err = NULL;
2087     bool mode_supported = false;
2088
2089     if (systemd_supports_mode(mode, &local_err)) {
2090         mode_supported = true;
2091         systemd_suspend(mode, &local_err);
2092     }
2093
2094     if (!local_err) {
2095         return;
2096     }
2097
2098     error_free(local_err);
2099     local_err = NULL;
2100
2101     if (pmutils_supports_mode(mode, &local_err)) {
2102         mode_supported = true;
2103         pmutils_suspend(mode, &local_err);
2104     }
2105
2106     if (!local_err) {
2107         return;
2108     }
2109
2110     error_free(local_err);
2111     local_err = NULL;
2112
2113     if (linux_sys_state_supports_mode(mode, &local_err)) {
2114         mode_supported = true;
2115         linux_sys_state_suspend(mode, &local_err);
2116     }
2117
2118     if (!mode_supported) {
2119         error_free(local_err);
2120         error_setg(errp,
2121                    "the requested suspend mode is not supported by the guest");
2122     } else {
2123         error_propagate(errp, local_err);
2124     }
2125 }
2126
2127 void qmp_guest_suspend_disk(Error **errp)
2128 {
2129     guest_suspend(SUSPEND_MODE_DISK, errp);
2130 }
2131
2132 void qmp_guest_suspend_ram(Error **errp)
2133 {
2134     guest_suspend(SUSPEND_MODE_RAM, errp);
2135 }
2136
2137 void qmp_guest_suspend_hybrid(Error **errp)
2138 {
2139     guest_suspend(SUSPEND_MODE_HYBRID, errp);
2140 }
2141
2142 static GuestNetworkInterfaceList *
2143 guest_find_interface(GuestNetworkInterfaceList *head,
2144                      const char *name)
2145 {
2146     for (; head; head = head->next) {
2147         if (strcmp(head->value->name, name) == 0) {
2148             break;
2149         }
2150     }
2151
2152     return head;
2153 }
2154
2155 static int guest_get_network_stats(const char *name,
2156                        GuestNetworkInterfaceStat *stats)
2157 {
2158     int name_len;
2159     char const *devinfo = "/proc/net/dev";
2160     FILE *fp;
2161     char *line = NULL, *colon;
2162     size_t n = 0;
2163     fp = fopen(devinfo, "r");
2164     if (!fp) {
2165         return -1;
2166     }
2167     name_len = strlen(name);
2168     while (getline(&line, &n, fp) != -1) {
2169         long long dummy;
2170         long long rx_bytes;
2171         long long rx_packets;
2172         long long rx_errs;
2173         long long rx_dropped;
2174         long long tx_bytes;
2175         long long tx_packets;
2176         long long tx_errs;
2177         long long tx_dropped;
2178         char *trim_line;
2179         trim_line = g_strchug(line);
2180         if (trim_line[0] == '\0') {
2181             continue;
2182         }
2183         colon = strchr(trim_line, ':');
2184         if (!colon) {
2185             continue;
2186         }
2187         if (colon - name_len  == trim_line &&
2188            strncmp(trim_line, name, name_len) == 0) {
2189             if (sscanf(colon + 1,
2190                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
2191                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
2192                   &dummy, &dummy, &dummy, &dummy,
2193                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
2194                   &dummy, &dummy, &dummy, &dummy) != 16) {
2195                 continue;
2196             }
2197             stats->rx_bytes = rx_bytes;
2198             stats->rx_packets = rx_packets;
2199             stats->rx_errs = rx_errs;
2200             stats->rx_dropped = rx_dropped;
2201             stats->tx_bytes = tx_bytes;
2202             stats->tx_packets = tx_packets;
2203             stats->tx_errs = tx_errs;
2204             stats->tx_dropped = tx_dropped;
2205             fclose(fp);
2206             g_free(line);
2207             return 0;
2208         }
2209     }
2210     fclose(fp);
2211     g_free(line);
2212     g_debug("/proc/net/dev: Interface '%s' not found", name);
2213     return -1;
2214 }
2215
2216 /*
2217  * Build information about guest interfaces
2218  */
2219 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2220 {
2221     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
2222     struct ifaddrs *ifap, *ifa;
2223
2224     if (getifaddrs(&ifap) < 0) {
2225         error_setg_errno(errp, errno, "getifaddrs failed");
2226         goto error;
2227     }
2228
2229     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
2230         GuestNetworkInterfaceList *info;
2231         GuestIpAddressList **address_list = NULL, *address_item = NULL;
2232         GuestNetworkInterfaceStat  *interface_stat = NULL;
2233         char addr4[INET_ADDRSTRLEN];
2234         char addr6[INET6_ADDRSTRLEN];
2235         int sock;
2236         struct ifreq ifr;
2237         unsigned char *mac_addr;
2238         void *p;
2239
2240         g_debug("Processing %s interface", ifa->ifa_name);
2241
2242         info = guest_find_interface(head, ifa->ifa_name);
2243
2244         if (!info) {
2245             info = g_malloc0(sizeof(*info));
2246             info->value = g_malloc0(sizeof(*info->value));
2247             info->value->name = g_strdup(ifa->ifa_name);
2248
2249             if (!cur_item) {
2250                 head = cur_item = info;
2251             } else {
2252                 cur_item->next = info;
2253                 cur_item = info;
2254             }
2255         }
2256
2257         if (!info->value->has_hardware_address &&
2258             ifa->ifa_flags & SIOCGIFHWADDR) {
2259             /* we haven't obtained HW address yet */
2260             sock = socket(PF_INET, SOCK_STREAM, 0);
2261             if (sock == -1) {
2262                 error_setg_errno(errp, errno, "failed to create socket");
2263                 goto error;
2264             }
2265
2266             memset(&ifr, 0, sizeof(ifr));
2267             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
2268             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
2269                 error_setg_errno(errp, errno,
2270                                  "failed to get MAC address of %s",
2271                                  ifa->ifa_name);
2272                 close(sock);
2273                 goto error;
2274             }
2275
2276             close(sock);
2277             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
2278
2279             info->value->hardware_address =
2280                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
2281                                 (int) mac_addr[0], (int) mac_addr[1],
2282                                 (int) mac_addr[2], (int) mac_addr[3],
2283                                 (int) mac_addr[4], (int) mac_addr[5]);
2284
2285             info->value->has_hardware_address = true;
2286         }
2287
2288         if (ifa->ifa_addr &&
2289             ifa->ifa_addr->sa_family == AF_INET) {
2290             /* interface with IPv4 address */
2291             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
2292             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
2293                 error_setg_errno(errp, errno, "inet_ntop failed");
2294                 goto error;
2295             }
2296
2297             address_item = g_malloc0(sizeof(*address_item));
2298             address_item->value = g_malloc0(sizeof(*address_item->value));
2299             address_item->value->ip_address = g_strdup(addr4);
2300             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
2301
2302             if (ifa->ifa_netmask) {
2303                 /* Count the number of set bits in netmask.
2304                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2305                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
2306                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
2307             }
2308         } else if (ifa->ifa_addr &&
2309                    ifa->ifa_addr->sa_family == AF_INET6) {
2310             /* interface with IPv6 address */
2311             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
2312             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
2313                 error_setg_errno(errp, errno, "inet_ntop failed");
2314                 goto error;
2315             }
2316
2317             address_item = g_malloc0(sizeof(*address_item));
2318             address_item->value = g_malloc0(sizeof(*address_item->value));
2319             address_item->value->ip_address = g_strdup(addr6);
2320             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
2321
2322             if (ifa->ifa_netmask) {
2323                 /* Count the number of set bits in netmask.
2324                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2325                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
2326                 address_item->value->prefix =
2327                     ctpop32(((uint32_t *) p)[0]) +
2328                     ctpop32(((uint32_t *) p)[1]) +
2329                     ctpop32(((uint32_t *) p)[2]) +
2330                     ctpop32(((uint32_t *) p)[3]);
2331             }
2332         }
2333
2334         if (!address_item) {
2335             continue;
2336         }
2337
2338         address_list = &info->value->ip_addresses;
2339
2340         while (*address_list && (*address_list)->next) {
2341             address_list = &(*address_list)->next;
2342         }
2343
2344         if (!*address_list) {
2345             *address_list = address_item;
2346         } else {
2347             (*address_list)->next = address_item;
2348         }
2349
2350         info->value->has_ip_addresses = true;
2351
2352         if (!info->value->has_statistics) {
2353             interface_stat = g_malloc0(sizeof(*interface_stat));
2354             if (guest_get_network_stats(info->value->name,
2355                 interface_stat) == -1) {
2356                 info->value->has_statistics = false;
2357                 g_free(interface_stat);
2358             } else {
2359                 info->value->statistics = interface_stat;
2360                 info->value->has_statistics = true;
2361             }
2362         }
2363     }
2364
2365     freeifaddrs(ifap);
2366     return head;
2367
2368 error:
2369     freeifaddrs(ifap);
2370     qapi_free_GuestNetworkInterfaceList(head);
2371     return NULL;
2372 }
2373
2374 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
2375
2376 static long sysconf_exact(int name, const char *name_str, Error **errp)
2377 {
2378     long ret;
2379
2380     errno = 0;
2381     ret = sysconf(name);
2382     if (ret == -1) {
2383         if (errno == 0) {
2384             error_setg(errp, "sysconf(%s): value indefinite", name_str);
2385         } else {
2386             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
2387         }
2388     }
2389     return ret;
2390 }
2391
2392 /* Transfer online/offline status between @vcpu and the guest system.
2393  *
2394  * On input either @errp or *@errp must be NULL.
2395  *
2396  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
2397  * - R: vcpu->logical_id
2398  * - W: vcpu->online
2399  * - W: vcpu->can_offline
2400  *
2401  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
2402  * - R: vcpu->logical_id
2403  * - R: vcpu->online
2404  *
2405  * Written members remain unmodified on error.
2406  */
2407 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
2408                           char *dirpath, Error **errp)
2409 {
2410     int fd;
2411     int res;
2412     int dirfd;
2413     static const char fn[] = "online";
2414
2415     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2416     if (dirfd == -1) {
2417         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2418         return;
2419     }
2420
2421     fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2422     if (fd == -1) {
2423         if (errno != ENOENT) {
2424             error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2425         } else if (sys2vcpu) {
2426             vcpu->online = true;
2427             vcpu->can_offline = false;
2428         } else if (!vcpu->online) {
2429             error_setg(errp, "logical processor #%" PRId64 " can't be "
2430                        "offlined", vcpu->logical_id);
2431         } /* otherwise pretend successful re-onlining */
2432     } else {
2433         unsigned char status;
2434
2435         res = pread(fd, &status, 1, 0);
2436         if (res == -1) {
2437             error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2438         } else if (res == 0) {
2439             error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2440                        fn);
2441         } else if (sys2vcpu) {
2442             vcpu->online = (status != '0');
2443             vcpu->can_offline = true;
2444         } else if (vcpu->online != (status != '0')) {
2445             status = '0' + vcpu->online;
2446             if (pwrite(fd, &status, 1, 0) == -1) {
2447                 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2448                                  fn);
2449             }
2450         } /* otherwise pretend successful re-(on|off)-lining */
2451
2452         res = close(fd);
2453         g_assert(res == 0);
2454     }
2455
2456     res = close(dirfd);
2457     g_assert(res == 0);
2458 }
2459
2460 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2461 {
2462     int64_t current;
2463     GuestLogicalProcessorList *head, **link;
2464     long sc_max;
2465     Error *local_err = NULL;
2466
2467     current = 0;
2468     head = NULL;
2469     link = &head;
2470     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
2471
2472     while (local_err == NULL && current < sc_max) {
2473         GuestLogicalProcessor *vcpu;
2474         GuestLogicalProcessorList *entry;
2475         int64_t id = current++;
2476         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2477                                      id);
2478
2479         if (g_file_test(path, G_FILE_TEST_EXISTS)) {
2480             vcpu = g_malloc0(sizeof *vcpu);
2481             vcpu->logical_id = id;
2482             vcpu->has_can_offline = true; /* lolspeak ftw */
2483             transfer_vcpu(vcpu, true, path, &local_err);
2484             entry = g_malloc0(sizeof *entry);
2485             entry->value = vcpu;
2486             *link = entry;
2487             link = &entry->next;
2488         }
2489         g_free(path);
2490     }
2491
2492     if (local_err == NULL) {
2493         /* there's no guest with zero VCPUs */
2494         g_assert(head != NULL);
2495         return head;
2496     }
2497
2498     qapi_free_GuestLogicalProcessorList(head);
2499     error_propagate(errp, local_err);
2500     return NULL;
2501 }
2502
2503 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2504 {
2505     int64_t processed;
2506     Error *local_err = NULL;
2507
2508     processed = 0;
2509     while (vcpus != NULL) {
2510         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2511                                      vcpus->value->logical_id);
2512
2513         transfer_vcpu(vcpus->value, false, path, &local_err);
2514         g_free(path);
2515         if (local_err != NULL) {
2516             break;
2517         }
2518         ++processed;
2519         vcpus = vcpus->next;
2520     }
2521
2522     if (local_err != NULL) {
2523         if (processed == 0) {
2524             error_propagate(errp, local_err);
2525         } else {
2526             error_free(local_err);
2527         }
2528     }
2529
2530     return processed;
2531 }
2532
2533 void qmp_guest_set_user_password(const char *username,
2534                                  const char *password,
2535                                  bool crypted,
2536                                  Error **errp)
2537 {
2538     Error *local_err = NULL;
2539     char *passwd_path = NULL;
2540     pid_t pid;
2541     int status;
2542     int datafd[2] = { -1, -1 };
2543     char *rawpasswddata = NULL;
2544     size_t rawpasswdlen;
2545     char *chpasswddata = NULL;
2546     size_t chpasswdlen;
2547
2548     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2549     if (!rawpasswddata) {
2550         return;
2551     }
2552     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2553     rawpasswddata[rawpasswdlen] = '\0';
2554
2555     if (strchr(rawpasswddata, '\n')) {
2556         error_setg(errp, "forbidden characters in raw password");
2557         goto out;
2558     }
2559
2560     if (strchr(username, '\n') ||
2561         strchr(username, ':')) {
2562         error_setg(errp, "forbidden characters in username");
2563         goto out;
2564     }
2565
2566     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2567     chpasswdlen = strlen(chpasswddata);
2568
2569     passwd_path = g_find_program_in_path("chpasswd");
2570
2571     if (!passwd_path) {
2572         error_setg(errp, "cannot find 'passwd' program in PATH");
2573         goto out;
2574     }
2575
2576     if (pipe(datafd) < 0) {
2577         error_setg(errp, "cannot create pipe FDs");
2578         goto out;
2579     }
2580
2581     pid = fork();
2582     if (pid == 0) {
2583         close(datafd[1]);
2584         /* child */
2585         setsid();
2586         dup2(datafd[0], 0);
2587         reopen_fd_to_null(1);
2588         reopen_fd_to_null(2);
2589
2590         if (crypted) {
2591             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2592         } else {
2593             execle(passwd_path, "chpasswd", NULL, environ);
2594         }
2595         _exit(EXIT_FAILURE);
2596     } else if (pid < 0) {
2597         error_setg_errno(errp, errno, "failed to create child process");
2598         goto out;
2599     }
2600     close(datafd[0]);
2601     datafd[0] = -1;
2602
2603     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2604         error_setg_errno(errp, errno, "cannot write new account password");
2605         goto out;
2606     }
2607     close(datafd[1]);
2608     datafd[1] = -1;
2609
2610     ga_wait_child(pid, &status, &local_err);
2611     if (local_err) {
2612         error_propagate(errp, local_err);
2613         goto out;
2614     }
2615
2616     if (!WIFEXITED(status)) {
2617         error_setg(errp, "child process has terminated abnormally");
2618         goto out;
2619     }
2620
2621     if (WEXITSTATUS(status)) {
2622         error_setg(errp, "child process has failed to set user password");
2623         goto out;
2624     }
2625
2626 out:
2627     g_free(chpasswddata);
2628     g_free(rawpasswddata);
2629     g_free(passwd_path);
2630     if (datafd[0] != -1) {
2631         close(datafd[0]);
2632     }
2633     if (datafd[1] != -1) {
2634         close(datafd[1]);
2635     }
2636 }
2637
2638 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2639                                int size, Error **errp)
2640 {
2641     int fd;
2642     int res;
2643
2644     errno = 0;
2645     fd = openat(dirfd, pathname, O_RDONLY);
2646     if (fd == -1) {
2647         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2648         return;
2649     }
2650
2651     res = pread(fd, buf, size, 0);
2652     if (res == -1) {
2653         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2654     } else if (res == 0) {
2655         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2656     }
2657     close(fd);
2658 }
2659
2660 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2661                                 const char *buf, int size, Error **errp)
2662 {
2663     int fd;
2664
2665     errno = 0;
2666     fd = openat(dirfd, pathname, O_WRONLY);
2667     if (fd == -1) {
2668         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2669         return;
2670     }
2671
2672     if (pwrite(fd, buf, size, 0) == -1) {
2673         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2674     }
2675
2676     close(fd);
2677 }
2678
2679 /* Transfer online/offline status between @mem_blk and the guest system.
2680  *
2681  * On input either @errp or *@errp must be NULL.
2682  *
2683  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2684  * - R: mem_blk->phys_index
2685  * - W: mem_blk->online
2686  * - W: mem_blk->can_offline
2687  *
2688  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2689  * - R: mem_blk->phys_index
2690  * - R: mem_blk->online
2691  *-  R: mem_blk->can_offline
2692  * Written members remain unmodified on error.
2693  */
2694 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2695                                   GuestMemoryBlockResponse *result,
2696                                   Error **errp)
2697 {
2698     char *dirpath;
2699     int dirfd;
2700     char *status;
2701     Error *local_err = NULL;
2702
2703     if (!sys2memblk) {
2704         DIR *dp;
2705
2706         if (!result) {
2707             error_setg(errp, "Internal error, 'result' should not be NULL");
2708             return;
2709         }
2710         errno = 0;
2711         dp = opendir("/sys/devices/system/memory/");
2712          /* if there is no 'memory' directory in sysfs,
2713          * we think this VM does not support online/offline memory block,
2714          * any other solution?
2715          */
2716         if (!dp) {
2717             if (errno == ENOENT) {
2718                 result->response =
2719                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2720             }
2721             goto out1;
2722         }
2723         closedir(dp);
2724     }
2725
2726     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2727                               mem_blk->phys_index);
2728     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2729     if (dirfd == -1) {
2730         if (sys2memblk) {
2731             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2732         } else {
2733             if (errno == ENOENT) {
2734                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2735             } else {
2736                 result->response =
2737                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2738             }
2739         }
2740         g_free(dirpath);
2741         goto out1;
2742     }
2743     g_free(dirpath);
2744
2745     status = g_malloc0(10);
2746     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2747     if (local_err) {
2748         /* treat with sysfs file that not exist in old kernel */
2749         if (errno == ENOENT) {
2750             error_free(local_err);
2751             if (sys2memblk) {
2752                 mem_blk->online = true;
2753                 mem_blk->can_offline = false;
2754             } else if (!mem_blk->online) {
2755                 result->response =
2756                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2757             }
2758         } else {
2759             if (sys2memblk) {
2760                 error_propagate(errp, local_err);
2761             } else {
2762                 error_free(local_err);
2763                 result->response =
2764                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2765             }
2766         }
2767         goto out2;
2768     }
2769
2770     if (sys2memblk) {
2771         char removable = '0';
2772
2773         mem_blk->online = (strncmp(status, "online", 6) == 0);
2774
2775         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2776         if (local_err) {
2777             /* if no 'removable' file, it doesn't support offline mem blk */
2778             if (errno == ENOENT) {
2779                 error_free(local_err);
2780                 mem_blk->can_offline = false;
2781             } else {
2782                 error_propagate(errp, local_err);
2783             }
2784         } else {
2785             mem_blk->can_offline = (removable != '0');
2786         }
2787     } else {
2788         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2789             const char *new_state = mem_blk->online ? "online" : "offline";
2790
2791             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2792                                 &local_err);
2793             if (local_err) {
2794                 error_free(local_err);
2795                 result->response =
2796                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2797                 goto out2;
2798             }
2799
2800             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2801             result->has_error_code = false;
2802         } /* otherwise pretend successful re-(on|off)-lining */
2803     }
2804     g_free(status);
2805     close(dirfd);
2806     return;
2807
2808 out2:
2809     g_free(status);
2810     close(dirfd);
2811 out1:
2812     if (!sys2memblk) {
2813         result->has_error_code = true;
2814         result->error_code = errno;
2815     }
2816 }
2817
2818 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2819 {
2820     GuestMemoryBlockList *head, **link;
2821     Error *local_err = NULL;
2822     struct dirent *de;
2823     DIR *dp;
2824
2825     head = NULL;
2826     link = &head;
2827
2828     dp = opendir("/sys/devices/system/memory/");
2829     if (!dp) {
2830         /* it's ok if this happens to be a system that doesn't expose
2831          * memory blocks via sysfs, but otherwise we should report
2832          * an error
2833          */
2834         if (errno != ENOENT) {
2835             error_setg_errno(errp, errno, "Can't open directory"
2836                              "\"/sys/devices/system/memory/\"");
2837         }
2838         return NULL;
2839     }
2840
2841     /* Note: the phys_index of memory block may be discontinuous,
2842      * this is because a memblk is the unit of the Sparse Memory design, which
2843      * allows discontinuous memory ranges (ex. NUMA), so here we should
2844      * traverse the memory block directory.
2845      */
2846     while ((de = readdir(dp)) != NULL) {
2847         GuestMemoryBlock *mem_blk;
2848         GuestMemoryBlockList *entry;
2849
2850         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2851             !(de->d_type & DT_DIR)) {
2852             continue;
2853         }
2854
2855         mem_blk = g_malloc0(sizeof *mem_blk);
2856         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2857         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2858         mem_blk->has_can_offline = true; /* lolspeak ftw */
2859         transfer_memory_block(mem_blk, true, NULL, &local_err);
2860         if (local_err) {
2861             break;
2862         }
2863
2864         entry = g_malloc0(sizeof *entry);
2865         entry->value = mem_blk;
2866
2867         *link = entry;
2868         link = &entry->next;
2869     }
2870
2871     closedir(dp);
2872     if (local_err == NULL) {
2873         /* there's no guest with zero memory blocks */
2874         if (head == NULL) {
2875             error_setg(errp, "guest reported zero memory blocks!");
2876         }
2877         return head;
2878     }
2879
2880     qapi_free_GuestMemoryBlockList(head);
2881     error_propagate(errp, local_err);
2882     return NULL;
2883 }
2884
2885 GuestMemoryBlockResponseList *
2886 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2887 {
2888     GuestMemoryBlockResponseList *head, **link;
2889     Error *local_err = NULL;
2890
2891     head = NULL;
2892     link = &head;
2893
2894     while (mem_blks != NULL) {
2895         GuestMemoryBlockResponse *result;
2896         GuestMemoryBlockResponseList *entry;
2897         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2898
2899         result = g_malloc0(sizeof(*result));
2900         result->phys_index = current_mem_blk->phys_index;
2901         transfer_memory_block(current_mem_blk, false, result, &local_err);
2902         if (local_err) { /* should never happen */
2903             goto err;
2904         }
2905         entry = g_malloc0(sizeof *entry);
2906         entry->value = result;
2907
2908         *link = entry;
2909         link = &entry->next;
2910         mem_blks = mem_blks->next;
2911     }
2912
2913     return head;
2914 err:
2915     qapi_free_GuestMemoryBlockResponseList(head);
2916     error_propagate(errp, local_err);
2917     return NULL;
2918 }
2919
2920 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2921 {
2922     Error *local_err = NULL;
2923     char *dirpath;
2924     int dirfd;
2925     char *buf;
2926     GuestMemoryBlockInfo *info;
2927
2928     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2929     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2930     if (dirfd == -1) {
2931         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2932         g_free(dirpath);
2933         return NULL;
2934     }
2935     g_free(dirpath);
2936
2937     buf = g_malloc0(20);
2938     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2939     close(dirfd);
2940     if (local_err) {
2941         g_free(buf);
2942         error_propagate(errp, local_err);
2943         return NULL;
2944     }
2945
2946     info = g_new0(GuestMemoryBlockInfo, 1);
2947     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2948
2949     g_free(buf);
2950
2951     return info;
2952 }
2953
2954 #else /* defined(__linux__) */
2955
2956 void qmp_guest_suspend_disk(Error **errp)
2957 {
2958     error_setg(errp, QERR_UNSUPPORTED);
2959 }
2960
2961 void qmp_guest_suspend_ram(Error **errp)
2962 {
2963     error_setg(errp, QERR_UNSUPPORTED);
2964 }
2965
2966 void qmp_guest_suspend_hybrid(Error **errp)
2967 {
2968     error_setg(errp, QERR_UNSUPPORTED);
2969 }
2970
2971 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2972 {
2973     error_setg(errp, QERR_UNSUPPORTED);
2974     return NULL;
2975 }
2976
2977 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2978 {
2979     error_setg(errp, QERR_UNSUPPORTED);
2980     return NULL;
2981 }
2982
2983 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2984 {
2985     error_setg(errp, QERR_UNSUPPORTED);
2986     return -1;
2987 }
2988
2989 void qmp_guest_set_user_password(const char *username,
2990                                  const char *password,
2991                                  bool crypted,
2992                                  Error **errp)
2993 {
2994     error_setg(errp, QERR_UNSUPPORTED);
2995 }
2996
2997 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2998 {
2999     error_setg(errp, QERR_UNSUPPORTED);
3000     return NULL;
3001 }
3002
3003 GuestMemoryBlockResponseList *
3004 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
3005 {
3006     error_setg(errp, QERR_UNSUPPORTED);
3007     return NULL;
3008 }
3009
3010 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
3011 {
3012     error_setg(errp, QERR_UNSUPPORTED);
3013     return NULL;
3014 }
3015
3016 #endif
3017
3018 #if !defined(CONFIG_FSFREEZE)
3019
3020 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
3021 {
3022     error_setg(errp, QERR_UNSUPPORTED);
3023     return NULL;
3024 }
3025
3026 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
3027 {
3028     error_setg(errp, QERR_UNSUPPORTED);
3029
3030     return 0;
3031 }
3032
3033 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
3034 {
3035     error_setg(errp, QERR_UNSUPPORTED);
3036
3037     return 0;
3038 }
3039
3040 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
3041                                        strList *mountpoints,
3042                                        Error **errp)
3043 {
3044     error_setg(errp, QERR_UNSUPPORTED);
3045
3046     return 0;
3047 }
3048
3049 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
3050 {
3051     error_setg(errp, QERR_UNSUPPORTED);
3052
3053     return 0;
3054 }
3055
3056 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
3057 {
3058     error_setg(errp, QERR_UNSUPPORTED);
3059     return NULL;
3060 }
3061
3062 #endif /* CONFIG_FSFREEZE */
3063
3064 #if !defined(CONFIG_FSTRIM)
3065 GuestFilesystemTrimResponse *
3066 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
3067 {
3068     error_setg(errp, QERR_UNSUPPORTED);
3069     return NULL;
3070 }
3071 #endif
3072
3073 /* add unsupported commands to the blacklist */
3074 GList *ga_command_blacklist_init(GList *blacklist)
3075 {
3076 #if !defined(__linux__)
3077     {
3078         const char *list[] = {
3079             "guest-suspend-disk", "guest-suspend-ram",
3080             "guest-suspend-hybrid", "guest-network-get-interfaces",
3081             "guest-get-vcpus", "guest-set-vcpus",
3082             "guest-get-memory-blocks", "guest-set-memory-blocks",
3083             "guest-get-memory-block-size", "guest-get-memory-block-info",
3084             NULL};
3085         char **p = (char **)list;
3086
3087         while (*p) {
3088             blacklist = g_list_append(blacklist, g_strdup(*p++));
3089         }
3090     }
3091 #endif
3092
3093 #if !defined(CONFIG_FSFREEZE)
3094     {
3095         const char *list[] = {
3096             "guest-get-fsinfo", "guest-fsfreeze-status",
3097             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
3098             "guest-fsfreeze-thaw", "guest-get-fsinfo",
3099             "guest-get-disks", NULL};
3100         char **p = (char **)list;
3101
3102         while (*p) {
3103             blacklist = g_list_append(blacklist, g_strdup(*p++));
3104         }
3105     }
3106 #endif
3107
3108 #if !defined(CONFIG_FSTRIM)
3109     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
3110 #endif
3111
3112     blacklist = g_list_append(blacklist, g_strdup("guest-get-devices"));
3113
3114     return blacklist;
3115 }
3116
3117 /* register init/cleanup routines for stateful command groups */
3118 void ga_command_state_init(GAState *s, GACommandState *cs)
3119 {
3120 #if defined(CONFIG_FSFREEZE)
3121     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
3122 #endif
3123 }
3124
3125 #ifdef HAVE_UTMPX
3126
3127 #define QGA_MICRO_SECOND_TO_SECOND 1000000
3128
3129 static double ga_get_login_time(struct utmpx *user_info)
3130 {
3131     double seconds = (double)user_info->ut_tv.tv_sec;
3132     double useconds = (double)user_info->ut_tv.tv_usec;
3133     useconds /= QGA_MICRO_SECOND_TO_SECOND;
3134     return seconds + useconds;
3135 }
3136
3137 GuestUserList *qmp_guest_get_users(Error **errp)
3138 {
3139     GHashTable *cache = NULL;
3140     GuestUserList *head = NULL, *cur_item = NULL;
3141     struct utmpx *user_info = NULL;
3142     gpointer value = NULL;
3143     GuestUser *user = NULL;
3144     GuestUserList *item = NULL;
3145     double login_time = 0;
3146
3147     cache = g_hash_table_new(g_str_hash, g_str_equal);
3148     setutxent();
3149
3150     for (;;) {
3151         user_info = getutxent();
3152         if (user_info == NULL) {
3153             break;
3154         } else if (user_info->ut_type != USER_PROCESS) {
3155             continue;
3156         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
3157             value = g_hash_table_lookup(cache, user_info->ut_user);
3158             user = (GuestUser *)value;
3159             login_time = ga_get_login_time(user_info);
3160             /* We're ensuring the earliest login time to be sent */
3161             if (login_time < user->login_time) {
3162                 user->login_time = login_time;
3163             }
3164             continue;
3165         }
3166
3167         item = g_new0(GuestUserList, 1);
3168         item->value = g_new0(GuestUser, 1);
3169         item->value->user = g_strdup(user_info->ut_user);
3170         item->value->login_time = ga_get_login_time(user_info);
3171
3172         g_hash_table_insert(cache, item->value->user, item->value);
3173
3174         if (!cur_item) {
3175             head = cur_item = item;
3176         } else {
3177             cur_item->next = item;
3178             cur_item = item;
3179         }
3180     }
3181     endutxent();
3182     g_hash_table_destroy(cache);
3183     return head;
3184 }
3185
3186 #else
3187
3188 GuestUserList *qmp_guest_get_users(Error **errp)
3189 {
3190     error_setg(errp, QERR_UNSUPPORTED);
3191     return NULL;
3192 }
3193
3194 #endif
3195
3196 /* Replace escaped special characters with theire real values. The replacement
3197  * is done in place -- returned value is in the original string.
3198  */
3199 static void ga_osrelease_replace_special(gchar *value)
3200 {
3201     gchar *p, *p2, quote;
3202
3203     /* Trim the string at first space or semicolon if it is not enclosed in
3204      * single or double quotes. */
3205     if ((value[0] != '"') || (value[0] == '\'')) {
3206         p = strchr(value, ' ');
3207         if (p != NULL) {
3208             *p = 0;
3209         }
3210         p = strchr(value, ';');
3211         if (p != NULL) {
3212             *p = 0;
3213         }
3214         return;
3215     }
3216
3217     quote = value[0];
3218     p2 = value;
3219     p = value + 1;
3220     while (*p != 0) {
3221         if (*p == '\\') {
3222             p++;
3223             switch (*p) {
3224             case '$':
3225             case '\'':
3226             case '"':
3227             case '\\':
3228             case '`':
3229                 break;
3230             default:
3231                 /* Keep literal backslash followed by whatever is there */
3232                 p--;
3233                 break;
3234             }
3235         } else if (*p == quote) {
3236             *p2 = 0;
3237             break;
3238         }
3239         *(p2++) = *(p++);
3240     }
3241 }
3242
3243 static GKeyFile *ga_parse_osrelease(const char *fname)
3244 {
3245     gchar *content = NULL;
3246     gchar *content2 = NULL;
3247     GError *err = NULL;
3248     GKeyFile *keys = g_key_file_new();
3249     const char *group = "[os-release]\n";
3250
3251     if (!g_file_get_contents(fname, &content, NULL, &err)) {
3252         slog("failed to read '%s', error: %s", fname, err->message);
3253         goto fail;
3254     }
3255
3256     if (!g_utf8_validate(content, -1, NULL)) {
3257         slog("file is not utf-8 encoded: %s", fname);
3258         goto fail;
3259     }
3260     content2 = g_strdup_printf("%s%s", group, content);
3261
3262     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
3263                                    &err)) {
3264         slog("failed to parse file '%s', error: %s", fname, err->message);
3265         goto fail;
3266     }
3267
3268     g_free(content);
3269     g_free(content2);
3270     return keys;
3271
3272 fail:
3273     g_error_free(err);
3274     g_free(content);
3275     g_free(content2);
3276     g_key_file_free(keys);
3277     return NULL;
3278 }
3279
3280 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
3281 {
3282     GuestOSInfo *info = NULL;
3283     struct utsname kinfo;
3284     GKeyFile *osrelease = NULL;
3285     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
3286
3287     info = g_new0(GuestOSInfo, 1);
3288
3289     if (uname(&kinfo) != 0) {
3290         error_setg_errno(errp, errno, "uname failed");
3291     } else {
3292         info->has_kernel_version = true;
3293         info->kernel_version = g_strdup(kinfo.version);
3294         info->has_kernel_release = true;
3295         info->kernel_release = g_strdup(kinfo.release);
3296         info->has_machine = true;
3297         info->machine = g_strdup(kinfo.machine);
3298     }
3299
3300     if (qga_os_release != NULL) {
3301         osrelease = ga_parse_osrelease(qga_os_release);
3302     } else {
3303         osrelease = ga_parse_osrelease("/etc/os-release");
3304         if (osrelease == NULL) {
3305             osrelease = ga_parse_osrelease("/usr/lib/os-release");
3306         }
3307     }
3308
3309     if (osrelease != NULL) {
3310         char *value;
3311
3312 #define GET_FIELD(field, osfield) do { \
3313     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
3314     if (value != NULL) { \
3315         ga_osrelease_replace_special(value); \
3316         info->has_ ## field = true; \
3317         info->field = value; \
3318     } \
3319 } while (0)
3320         GET_FIELD(id, "ID");
3321         GET_FIELD(name, "NAME");
3322         GET_FIELD(pretty_name, "PRETTY_NAME");
3323         GET_FIELD(version, "VERSION");
3324         GET_FIELD(version_id, "VERSION_ID");
3325         GET_FIELD(variant, "VARIANT");
3326         GET_FIELD(variant_id, "VARIANT_ID");
3327 #undef GET_FIELD
3328
3329         g_key_file_free(osrelease);
3330     }
3331
3332     return info;
3333 }
3334
3335 GuestDeviceInfoList *qmp_guest_get_devices(Error **errp)
3336 {
3337     error_setg(errp, QERR_UNSUPPORTED);
3338
3339     return NULL;
3340 }