fsdev/virtfs-proxy-helper.c

   1 /*
   2  * Helper for QEMU Proxy FS Driver
   3  * Copyright IBM, Corp. 2011
   4  *
   5  * Authors:
   6  * M. Mohan Kumar <mohan@in.ibm.com>
   7  *
   8  * This work is licensed under the terms of the GNU GPL, version 2. See
   9  * the COPYING file in the top-level directory.
  10  */
  11
  12 #include <sys/resource.h>
  13 #include <getopt.h>
  14 #include <syslog.h>
  15 #include <sys/capability.h>
  16 #include <sys/fsuid.h>
  17 #include <sys/vfs.h>
  18 #include <sys/ioctl.h>
  19 #include <linux/fs.h>
  20 #ifdef CONFIG_LINUX_MAGIC_H
  21 #include <linux/magic.h>
  22 #endif
  23 #include "qemu-common.h"
  24 #include "qemu_socket.h"
  25 #include "qemu-xattr.h"
  26 #include "virtio-9p-marshal.h"
  27 #include "hw/9pfs/virtio-9p-proxy.h"
  28 #include "fsdev/virtio-9p-marshal.h"
  29
  30 #define PROGNAME "virtfs-proxy-helper"
  31
  32 #ifndef XFS_SUPER_MAGIC
  33 #define XFS_SUPER_MAGIC  0x58465342
  34 #endif
  35 #ifndef EXT2_SUPER_MAGIC
  36 #define EXT2_SUPER_MAGIC 0xEF53
  37 #endif
  38 #ifndef REISERFS_SUPER_MAGIC
  39 #define REISERFS_SUPER_MAGIC 0x52654973
  40 #endif
  41 #ifndef BTRFS_SUPER_MAGIC
  42 #define BTRFS_SUPER_MAGIC 0x9123683E
  43 #endif
  44
  45 static struct option helper_opts[] = {
  46     {"fd", required_argument, NULL, 'f'},
  47     {"path", required_argument, NULL, 'p'},
  48     {"nodaemon", no_argument, NULL, 'n'},
  49     {"socket", required_argument, NULL, 's'},
  50     {"uid", required_argument, NULL, 'u'},
  51     {"gid", required_argument, NULL, 'g'},
  52 };
  53
  54 static bool is_daemon;
  55 static bool get_version; /* IOC getversion IOCTL supported */
  56
  57 static void GCC_FMT_ATTR(2, 3) do_log(int loglevel, const char *format, ...)
  58 {
  59     va_list ap;
  60
  61     va_start(ap, format);
  62     if (is_daemon) {
  63         vsyslog(LOG_CRIT, format, ap);
  64     } else {
  65         vfprintf(stderr, format, ap);
  66     }
  67     va_end(ap);
  68 }
  69
  70 static void do_perror(const char *string)
  71 {
  72     if (is_daemon) {
  73         syslog(LOG_CRIT, "%s:%s", string, strerror(errno));
  74     } else {
  75         fprintf(stderr, "%s:%s\n", string, strerror(errno));
  76     }
  77 }
  78
  79 static int do_cap_set(cap_value_t *cap_value, int size, int reset)
  80 {
  81     cap_t caps;
  82     if (reset) {
  83         /*
  84          * Start with an empty set and set permitted and effective
  85          */
  86         caps = cap_init();
  87         if (caps == NULL) {
  88             do_perror("cap_init");
  89             return -1;
  90         }
  91         if (cap_set_flag(caps, CAP_PERMITTED, size, cap_value, CAP_SET) < 0) {
  92             do_perror("cap_set_flag");
  93             goto error;
  94         }
  95     } else {
  96         caps = cap_get_proc();
  97         if (!caps) {
  98             do_perror("cap_get_proc");
  99             return -1;
 100         }
 101     }
 102     if (cap_set_flag(caps, CAP_EFFECTIVE, size, cap_value, CAP_SET) < 0) {
 103         do_perror("cap_set_flag");
 104         goto error;
 105     }
 106     if (cap_set_proc(caps) < 0) {
 107         do_perror("cap_set_proc");
 108         goto error;
 109     }
 110     cap_free(caps);
 111     return 0;
 112
 113 error:
 114     cap_free(caps);
 115     return -1;
 116 }
 117
 118 static int init_capabilities(void)
 119 {
 120     /* helper needs following capbabilities only */
 121     cap_value_t cap_list[] = {
 122         CAP_CHOWN,
 123         CAP_DAC_OVERRIDE,
 124         CAP_FOWNER,
 125         CAP_FSETID,
 126         CAP_SETGID,
 127         CAP_MKNOD,
 128         CAP_SETUID,
 129     };
 130     return do_cap_set(cap_list, ARRAY_SIZE(cap_list), 1);
 131 }
 132
 133 static int socket_read(int sockfd, void *buff, ssize_t size)
 134 {
 135     ssize_t retval, total = 0;
 136
 137     while (size) {
 138         retval = read(sockfd, buff, size);
 139         if (retval == 0) {
 140             return -EIO;
 141         }
 142         if (retval < 0) {
 143             if (errno == EINTR) {
 144                 continue;
 145             }
 146             return -errno;
 147         }
 148         size -= retval;
 149         buff += retval;
 150         total += retval;
 151     }
 152     return total;
 153 }
 154
 155 static int socket_write(int sockfd, void *buff, ssize_t size)
 156 {
 157     ssize_t retval, total = 0;
 158
 159     while (size) {
 160         retval = write(sockfd, buff, size);
 161         if (retval < 0) {
 162             if (errno == EINTR) {
 163                 continue;
 164             }
 165             return -errno;
 166         }
 167         size -= retval;
 168         buff += retval;
 169         total += retval;
 170     }
 171     return total;
 172 }
 173
 174 static int read_request(int sockfd, struct iovec *iovec, ProxyHeader *header)
 175 {
 176     int retval;
 177
 178     /*
 179      * read the request header.
 180      */
 181     iovec->iov_len = 0;
 182     retval = socket_read(sockfd, iovec->iov_base, PROXY_HDR_SZ);
 183     if (retval < 0) {
 184         return retval;
 185     }
 186     iovec->iov_len = PROXY_HDR_SZ;
 187     retval = proxy_unmarshal(iovec, 0, "dd", &header->type, &header->size);
 188     if (retval < 0) {
 189         return retval;
 190     }
 191     /*
 192      * We can't process message.size > PROXY_MAX_IO_SZ.
 193      * Treat it as fatal error
 194      */
 195     if (header->size > PROXY_MAX_IO_SZ) {
 196         return -ENOBUFS;
 197     }
 198     retval = socket_read(sockfd, iovec->iov_base + PROXY_HDR_SZ, header->size);
 199     if (retval < 0) {
 200         return retval;
 201     }
 202     iovec->iov_len += header->size;
 203     return 0;
 204 }
 205
 206 static int send_fd(int sockfd, int fd)
 207 {
 208     struct msghdr msg;
 209     struct iovec iov;
 210     int retval, data;
 211     struct cmsghdr *cmsg;
 212     union MsgControl msg_control;
 213
 214     iov.iov_base = &data;
 215     iov.iov_len = sizeof(data);
 216
 217     memset(&msg, 0, sizeof(msg));
 218     msg.msg_iov = &iov;
 219     msg.msg_iovlen = 1;
 220     /* No ancillary data on error */
 221     if (fd < 0) {
 222         /* fd is really negative errno if the request failed  */
 223         data = fd;
 224     } else {
 225         data = V9FS_FD_VALID;
 226         msg.msg_control = &msg_control;
 227         msg.msg_controllen = sizeof(msg_control);
 228
 229         cmsg = &msg_control.cmsg;
 230         cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 231         cmsg->cmsg_level = SOL_SOCKET;
 232         cmsg->cmsg_type = SCM_RIGHTS;
 233         memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
 234     }
 235
 236     do {
 237         retval = sendmsg(sockfd, &msg, 0);
 238     } while (retval < 0 && errno == EINTR);
 239     if (fd >= 0) {
 240         close(fd);
 241     }
 242     if (retval < 0) {
 243         return retval;
 244     }
 245     return 0;
 246 }
 247
 248 static int send_status(int sockfd, struct iovec *iovec, int status)
 249 {
 250     ProxyHeader header;
 251     int retval, msg_size;;
 252
 253     if (status < 0) {
 254         header.type = T_ERROR;
 255     } else {
 256         header.type = T_SUCCESS;
 257     }
 258     header.size = sizeof(status);
 259     /*
 260      * marshal the return status. We don't check error.
 261      * because we are sure we have enough space for the status
 262      */
 263     msg_size = proxy_marshal(iovec, 0, "ddd", header.type,
 264                              header.size, status);
 265     retval = socket_write(sockfd, iovec->iov_base, msg_size);
 266     if (retval < 0) {
 267         return retval;
 268     }
 269     return 0;
 270 }
 271
 272 /*
 273  * from man 7 capabilities, section
 274  * Effect of User ID Changes on Capabilities:
 275  * If the effective user ID is changed from nonzero to 0, then the permitted
 276  * set is copied to the effective set.  If the effective user ID is changed
 277  * from 0 to nonzero, then all capabilities are are cleared from the effective
 278  * set.
 279  *
 280  * The setfsuid/setfsgid man pages warn that changing the effective user ID may
 281  * expose the program to unwanted signals, but this is not true anymore: for an
 282  * unprivileged (without CAP_KILL) program to send a signal, the real or
 283  * effective user ID of the sending process must equal the real or saved user
 284  * ID of the target process.  Even when dropping privileges, it is enough to
 285  * keep the saved UID to a "privileged" value and virtfs-proxy-helper won't
 286  * be exposed to signals.  So just use setresuid/setresgid.
 287  */
 288 static int setugid(int uid, int gid, int *suid, int *sgid)
 289 {
 290     int retval;
 291
 292     /*
 293      * We still need DAC_OVERRIDE because we don't change
 294      * supplementary group ids, and hence may be subjected DAC rules
 295      */
 296     cap_value_t cap_list[] = {
 297         CAP_DAC_OVERRIDE,
 298     };
 299
 300     *suid = geteuid();
 301     *sgid = getegid();
 302
 303     if (setresgid(-1, gid, *sgid) == -1) {
 304         retval = -errno;
 305         goto err_out;
 306     }
 307
 308     if (setresuid(-1, uid, *suid) == -1) {
 309         retval = -errno;
 310         goto err_sgid;
 311     }
 312
 313     if (uid != 0 || gid != 0) {
 314         if (do_cap_set(cap_list, ARRAY_SIZE(cap_list), 0) < 0) {
 315             retval = -errno;
 316             goto err_suid;
 317         }
 318     }
 319     return 0;
 320
 321 err_suid:
 322     if (setresuid(-1, *suid, *suid) == -1) {
 323         abort();
 324     }
 325 err_sgid:
 326     if (setresgid(-1, *sgid, *sgid) == -1) {
 327         abort();
 328     }
 329 err_out:
 330     return retval;
 331 }
 332
 333 /*
 334  * This is used to reset the ugid back with the saved values
 335  * There is nothing much we can do checking error values here.
 336  */
 337 static void resetugid(int suid, int sgid)
 338 {
 339     if (setresgid(-1, sgid, sgid) == -1) {
 340         abort();
 341     }
 342     if (setresuid(-1, suid, suid) == -1) {
 343         abort();
 344     }
 345 }
 346
 347 /*
 348  * send response in two parts
 349  * 1) ProxyHeader
 350  * 2) Response or error status
 351  * This function should be called with marshaled response
 352  * send_response constructs header part and error part only.
 353  * send response sends {ProxyHeader,Response} if the request was success
 354  * otherwise sends {ProxyHeader,error status}
 355  */
 356 static int send_response(int sock, struct iovec *iovec, int size)
 357 {
 358     int retval;
 359     ProxyHeader header;
 360
 361     /*
 362      * If response size exceeds available iovec->iov_len,
 363      * we return ENOBUFS
 364      */
 365     if (size > PROXY_MAX_IO_SZ) {
 366         size = -ENOBUFS;
 367     }
 368
 369     if (size < 0) {
 370         /*
 371          * In case of error we would not have got the error encoded
 372          * already so encode the error here.
 373          */
 374         header.type = T_ERROR;
 375         header.size = sizeof(size);
 376         proxy_marshal(iovec, PROXY_HDR_SZ, "d", size);
 377     } else {
 378         header.type = T_SUCCESS;
 379         header.size = size;
 380     }
 381     proxy_marshal(iovec, 0, "dd", header.type, header.size);
 382     retval = socket_write(sock, iovec->iov_base, header.size + PROXY_HDR_SZ);
 383     if (retval < 0) {
 384         return retval;;
 385     }
 386     return 0;
 387 }
 388
 389 /*
 390  * gets generation number
 391  * returns -errno on failure and sizeof(generation number) on success
 392  */
 393 static int do_getversion(struct iovec *iovec, struct iovec *out_iovec)
 394 {
 395     uint64_t version;
 396     int retval = -ENOTTY;
 397 #ifdef FS_IOC_GETVERSION
 398     int fd;
 399     V9fsString path;
 400 #endif
 401
 402
 403     /* no need to issue ioctl */
 404     if (!get_version) {
 405         version = 0;
 406         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
 407         return retval;
 408     }
 409 #ifdef FS_IOC_GETVERSION
 410     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
 411     if (retval < 0) {
 412         return retval;
 413     }
 414
 415     fd = open(path.data, O_RDONLY);
 416     if (fd < 0) {
 417         retval = -errno;
 418         goto err_out;
 419     }
 420     if (ioctl(fd, FS_IOC_GETVERSION, &version) < 0) {
 421         retval = -errno;
 422     } else {
 423         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
 424     }
 425     close(fd);
 426 err_out:
 427     v9fs_string_free(&path);
 428 #endif
 429     return retval;
 430 }
 431
 432 static int do_getxattr(int type, struct iovec *iovec, struct iovec *out_iovec)
 433 {
 434     int size = 0, offset, retval;
 435     V9fsString path, name, xattr;
 436
 437     v9fs_string_init(&xattr);
 438     v9fs_string_init(&path);
 439     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "ds", &size, &path);
 440     if (retval < 0) {
 441         return retval;
 442     }
 443     offset = PROXY_HDR_SZ + retval;
 444
 445     if (size) {
 446         xattr.data = g_malloc(size);
 447         xattr.size = size;
 448     }
 449     switch (type) {
 450     case T_LGETXATTR:
 451         v9fs_string_init(&name);
 452         retval = proxy_unmarshal(iovec, offset, "s", &name);
 453         if (retval > 0) {
 454             retval = lgetxattr(path.data, name.data, xattr.data, size);
 455             if (retval < 0) {
 456                 retval = -errno;
 457             } else {
 458                 xattr.size = retval;
 459             }
 460         }
 461         v9fs_string_free(&name);
 462         break;
 463     case T_LLISTXATTR:
 464         retval = llistxattr(path.data, xattr.data, size);
 465         if (retval < 0) {
 466             retval = -errno;
 467         } else {
 468             xattr.size = retval;
 469         }
 470         break;
 471     }
 472     if (retval < 0) {
 473         goto err_out;
 474     }
 475
 476     if (!size) {
 477         proxy_marshal(out_iovec, PROXY_HDR_SZ, "d", retval);
 478         retval = sizeof(retval);
 479     } else {
 480         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &xattr);
 481     }
 482 err_out:
 483     v9fs_string_free(&xattr);
 484     v9fs_string_free(&path);
 485     return retval;
 486 }
 487
 488 static void stat_to_prstat(ProxyStat *pr_stat, struct stat *stat)
 489 {
 490     memset(pr_stat, 0, sizeof(*pr_stat));
 491     pr_stat->st_dev = stat->st_dev;
 492     pr_stat->st_ino = stat->st_ino;
 493     pr_stat->st_nlink = stat->st_nlink;
 494     pr_stat->st_mode = stat->st_mode;
 495     pr_stat->st_uid = stat->st_uid;
 496     pr_stat->st_gid = stat->st_gid;
 497     pr_stat->st_rdev = stat->st_rdev;
 498     pr_stat->st_size = stat->st_size;
 499     pr_stat->st_blksize = stat->st_blksize;
 500     pr_stat->st_blocks = stat->st_blocks;
 501     pr_stat->st_atim_sec = stat->st_atim.tv_sec;
 502     pr_stat->st_atim_nsec = stat->st_atim.tv_nsec;
 503     pr_stat->st_mtim_sec = stat->st_mtim.tv_sec;
 504     pr_stat->st_mtim_nsec = stat->st_mtim.tv_nsec;
 505     pr_stat->st_ctim_sec = stat->st_ctim.tv_sec;
 506     pr_stat->st_ctim_nsec = stat->st_ctim.tv_nsec;
 507 }
 508
 509 static void statfs_to_prstatfs(ProxyStatFS *pr_stfs, struct statfs *stfs)
 510 {
 511     memset(pr_stfs, 0, sizeof(*pr_stfs));
 512     pr_stfs->f_type = stfs->f_type;
 513     pr_stfs->f_bsize = stfs->f_bsize;
 514     pr_stfs->f_blocks = stfs->f_blocks;
 515     pr_stfs->f_bfree = stfs->f_bfree;
 516     pr_stfs->f_bavail = stfs->f_bavail;
 517     pr_stfs->f_files = stfs->f_files;
 518     pr_stfs->f_ffree = stfs->f_ffree;
 519     pr_stfs->f_fsid[0] = stfs->f_fsid.__val[0];
 520     pr_stfs->f_fsid[1] = stfs->f_fsid.__val[1];
 521     pr_stfs->f_namelen = stfs->f_namelen;
 522     pr_stfs->f_frsize = stfs->f_frsize;
 523 }
 524
 525 /*
 526  * Gets stat/statfs information and packs in out_iovec structure
 527  * on success returns number of bytes packed in out_iovec struture
 528  * otherwise returns -errno
 529  */
 530 static int do_stat(int type, struct iovec *iovec, struct iovec *out_iovec)
 531 {
 532     int retval;
 533     V9fsString path;
 534     ProxyStat pr_stat;
 535     ProxyStatFS pr_stfs;
 536     struct stat st_buf;
 537     struct statfs stfs_buf;
 538
 539     v9fs_string_init(&path);
 540     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
 541     if (retval < 0) {
 542         return retval;
 543     }
 544
 545     switch (type) {
 546     case T_LSTAT:
 547         retval = lstat(path.data, &st_buf);
 548         if (retval < 0) {
 549             retval = -errno;
 550         } else {
 551             stat_to_prstat(&pr_stat, &st_buf);
 552             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
 553                                    "qqqdddqqqqqqqqqq", pr_stat.st_dev,
 554                                    pr_stat.st_ino, pr_stat.st_nlink,
 555                                    pr_stat.st_mode, pr_stat.st_uid,
 556                                    pr_stat.st_gid, pr_stat.st_rdev,
 557                                    pr_stat.st_size, pr_stat.st_blksize,
 558                                    pr_stat.st_blocks,
 559                                    pr_stat.st_atim_sec, pr_stat.st_atim_nsec,
 560                                    pr_stat.st_mtim_sec, pr_stat.st_mtim_nsec,
 561                                    pr_stat.st_ctim_sec, pr_stat.st_ctim_nsec);
 562         }
 563         break;
 564     case T_STATFS:
 565         retval = statfs(path.data, &stfs_buf);
 566         if (retval < 0) {
 567             retval = -errno;
 568         } else {
 569             statfs_to_prstatfs(&pr_stfs, &stfs_buf);
 570             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
 571                                    "qqqqqqqqqqq", pr_stfs.f_type,
 572                                    pr_stfs.f_bsize, pr_stfs.f_blocks,
 573                                    pr_stfs.f_bfree, pr_stfs.f_bavail,
 574                                    pr_stfs.f_files, pr_stfs.f_ffree,
 575                                    pr_stfs.f_fsid[0], pr_stfs.f_fsid[1],
 576                                    pr_stfs.f_namelen, pr_stfs.f_frsize);
 577         }
 578         break;
 579     }
 580     v9fs_string_free(&path);
 581     return retval;
 582 }
 583
 584 static int do_readlink(struct iovec *iovec, struct iovec *out_iovec)
 585 {
 586     char *buffer;
 587     int size, retval;
 588     V9fsString target, path;
 589
 590     v9fs_string_init(&path);
 591     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &size);
 592     if (retval < 0) {
 593         v9fs_string_free(&path);
 594         return retval;
 595     }
 596     buffer = g_malloc(size);
 597     v9fs_string_init(&target);
 598     retval = readlink(path.data, buffer, size);
 599     if (retval > 0) {
 600         buffer[retval] = '\0';
 601         v9fs_string_sprintf(&target, "%s", buffer);
 602         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &target);
 603     } else {
 604         retval = -errno;
 605     }
 606     g_free(buffer);
 607     v9fs_string_free(&target);
 608     v9fs_string_free(&path);
 609     return retval;
 610 }
 611
 612 /*
 613  * create other filesystem objects and send 0 on success
 614  * return -errno on error
 615  */
 616 static int do_create_others(int type, struct iovec *iovec)
 617 {
 618     dev_t rdev;
 619     int retval = 0;
 620     int offset = PROXY_HDR_SZ;
 621     V9fsString oldpath, path;
 622     int mode, uid, gid, cur_uid, cur_gid;
 623
 624     v9fs_string_init(&path);
 625     v9fs_string_init(&oldpath);
 626
 627     retval = proxy_unmarshal(iovec, offset, "dd", &uid, &gid);
 628     if (retval < 0) {
 629         return retval;
 630     }
 631     offset += retval;
 632     retval = setugid(uid, gid, &cur_uid, &cur_gid);
 633     if (retval < 0) {
 634         goto unmarshal_err_out;
 635     }
 636     switch (type) {
 637     case T_MKNOD:
 638         retval = proxy_unmarshal(iovec, offset, "sdq", &path, &mode, &rdev);
 639         if (retval < 0) {
 640             goto err_out;
 641         }
 642         retval = mknod(path.data, mode, rdev);
 643         break;
 644     case T_MKDIR:
 645         retval = proxy_unmarshal(iovec, offset, "sd", &path, &mode);
 646         if (retval < 0) {
 647             goto err_out;
 648         }
 649         retval = mkdir(path.data, mode);
 650         break;
 651     case T_SYMLINK:
 652         retval = proxy_unmarshal(iovec, offset, "ss", &oldpath, &path);
 653         if (retval < 0) {
 654             goto err_out;
 655         }
 656         retval = symlink(oldpath.data, path.data);
 657         break;
 658     }
 659     if (retval < 0) {
 660         retval = -errno;
 661     }
 662
 663 err_out:
 664     resetugid(cur_uid, cur_gid);
 665 unmarshal_err_out:
 666     v9fs_string_free(&path);
 667     v9fs_string_free(&oldpath);
 668     return retval;
 669 }
 670
 671 /*
 672  * create a file and send fd on success
 673  * return -errno on error
 674  */
 675 static int do_create(struct iovec *iovec)
 676 {
 677     int ret;
 678     V9fsString path;
 679     int flags, mode, uid, gid, cur_uid, cur_gid;
 680
 681     v9fs_string_init(&path);
 682     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sdddd",
 683                           &path, &flags, &mode, &uid, &gid);
 684     if (ret < 0) {
 685         goto unmarshal_err_out;
 686     }
 687     ret = setugid(uid, gid, &cur_uid, &cur_gid);
 688     if (ret < 0) {
 689         goto unmarshal_err_out;
 690     }
 691     ret = open(path.data, flags, mode);
 692     if (ret < 0) {
 693         ret = -errno;
 694     }
 695
 696     resetugid(cur_uid, cur_gid);
 697 unmarshal_err_out:
 698     v9fs_string_free(&path);
 699     return ret;
 700 }
 701
 702 /*
 703  * open a file and send fd on success
 704  * return -errno on error
 705  */
 706 static int do_open(struct iovec *iovec)
 707 {
 708     int flags, ret;
 709     V9fsString path;
 710
 711     v9fs_string_init(&path);
 712     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &flags);
 713     if (ret < 0) {
 714         goto err_out;
 715     }
 716     ret = open(path.data, flags);
 717     if (ret < 0) {
 718         ret = -errno;
 719     }
 720 err_out:
 721     v9fs_string_free(&path);
 722     return ret;
 723 }
 724
 725 /* create unix domain socket and return the descriptor */
 726 static int proxy_socket(const char *path, uid_t uid, gid_t gid)
 727 {
 728     int sock, client;
 729     struct sockaddr_un proxy, qemu;
 730     socklen_t size;
 731
 732     /* requested socket already exists, refuse to start */
 733     if (!access(path, F_OK)) {
 734         do_log(LOG_CRIT, "socket already exists\n");
 735         return -1;
 736     }
 737
 738     sock = socket(AF_UNIX, SOCK_STREAM, 0);
 739     if (sock < 0) {
 740         do_perror("socket");
 741         return -1;
 742     }
 743
 744     /* mask other part of mode bits */
 745     umask(7);
 746
 747     proxy.sun_family = AF_UNIX;
 748     strcpy(proxy.sun_path, path);
 749     if (bind(sock, (struct sockaddr *)&proxy,
 750             sizeof(struct sockaddr_un)) < 0) {
 751         do_perror("bind");
 752         return -1;
 753     }
 754     if (chown(proxy.sun_path, uid, gid) < 0) {
 755         do_perror("chown");
 756         return -1;
 757     }
 758     if (listen(sock, 1) < 0) {
 759         do_perror("listen");
 760         return -1;
 761     }
 762
 763     client = accept(sock, (struct sockaddr *)&qemu, &size);
 764     if (client < 0) {
 765         do_perror("accept");
 766         return -1;
 767     }
 768     return client;
 769 }
 770
 771 static void usage(char *prog)
 772 {
 773     fprintf(stderr, "usage: %s\n"
 774             " -p|--path <path> 9p path to export\n"
 775             " {-f|--fd <socket-descriptor>} socket file descriptor to be used\n"
 776             " {-s|--socket <socketname> socket file used for communication\n"
 777             " \t-u|--uid <uid> -g|--gid <gid>} - uid:gid combination to give "
 778             " access to this socket\n"
 779             " \tNote: -s & -f can not be used together\n"
 780             " [-n|--nodaemon] Run as a normal program\n",
 781             basename(prog));
 782 }
 783
 784 static int process_reply(int sock, int type,
 785                          struct iovec *out_iovec, int retval)
 786 {
 787     switch (type) {
 788     case T_OPEN:
 789     case T_CREATE:
 790         if (send_fd(sock, retval) < 0) {
 791             return -1;
 792         }
 793         break;
 794     case T_MKNOD:
 795     case T_MKDIR:
 796     case T_SYMLINK:
 797     case T_LINK:
 798     case T_CHMOD:
 799     case T_CHOWN:
 800     case T_TRUNCATE:
 801     case T_UTIME:
 802     case T_RENAME:
 803     case T_REMOVE:
 804     case T_LSETXATTR:
 805     case T_LREMOVEXATTR:
 806         if (send_status(sock, out_iovec, retval) < 0) {
 807             return -1;
 808         }
 809         break;
 810     case T_LSTAT:
 811     case T_STATFS:
 812     case T_READLINK:
 813     case T_LGETXATTR:
 814     case T_LLISTXATTR:
 815     case T_GETVERSION:
 816         if (send_response(sock, out_iovec, retval) < 0) {
 817             return -1;
 818         }
 819         break;
 820     default:
 821         return -1;
 822         break;
 823     }
 824     return 0;
 825 }
 826
 827 static int process_requests(int sock)
 828 {
 829     int flags;
 830     int size = 0;
 831     int retval = 0;
 832     uint64_t offset;
 833     ProxyHeader header;
 834     int mode, uid, gid;
 835     V9fsString name, value;
 836     struct timespec spec[2];
 837     V9fsString oldpath, path;
 838     struct iovec in_iovec, out_iovec;
 839
 840     in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
 841     in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
 842     out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
 843     out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
 844
 845     while (1) {
 846         /*
 847          * initialize the header type, so that we send
 848          * response to proper request type.
 849          */
 850         header.type = 0;
 851         retval = read_request(sock, &in_iovec, &header);
 852         if (retval < 0) {
 853             goto err_out;
 854         }
 855
 856         switch (header.type) {
 857         case T_OPEN:
 858             retval = do_open(&in_iovec);
 859             break;
 860         case T_CREATE:
 861             retval = do_create(&in_iovec);
 862             break;
 863         case T_MKNOD:
 864         case T_MKDIR:
 865         case T_SYMLINK:
 866             retval = do_create_others(header.type, &in_iovec);
 867             break;
 868         case T_LINK:
 869             v9fs_string_init(&path);
 870             v9fs_string_init(&oldpath);
 871             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 872                                      "ss", &oldpath, &path);
 873             if (retval > 0) {
 874                 retval = link(oldpath.data, path.data);
 875                 if (retval < 0) {
 876                     retval = -errno;
 877                 }
 878             }
 879             v9fs_string_free(&oldpath);
 880             v9fs_string_free(&path);
 881             break;
 882         case T_LSTAT:
 883         case T_STATFS:
 884             retval = do_stat(header.type, &in_iovec, &out_iovec);
 885             break;
 886         case T_READLINK:
 887             retval = do_readlink(&in_iovec, &out_iovec);
 888             break;
 889         case T_CHMOD:
 890             v9fs_string_init(&path);
 891             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 892                                      "sd", &path, &mode);
 893             if (retval > 0) {
 894                 retval = chmod(path.data, mode);
 895                 if (retval < 0) {
 896                     retval = -errno;
 897                 }
 898             }
 899             v9fs_string_free(&path);
 900             break;
 901         case T_CHOWN:
 902             v9fs_string_init(&path);
 903             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sdd", &path,
 904                                      &uid, &gid);
 905             if (retval > 0) {
 906                 retval = lchown(path.data, uid, gid);
 907                 if (retval < 0) {
 908                     retval = -errno;
 909                 }
 910             }
 911             v9fs_string_free(&path);
 912             break;
 913         case T_TRUNCATE:
 914             v9fs_string_init(&path);
 915             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sq",
 916                                      &path, &offset);
 917             if (retval > 0) {
 918                 retval = truncate(path.data, offset);
 919                 if (retval < 0) {
 920                     retval = -errno;
 921                 }
 922             }
 923             v9fs_string_free(&path);
 924             break;
 925         case T_UTIME:
 926             v9fs_string_init(&path);
 927             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sqqqq", &path,
 928                                      &spec[0].tv_sec, &spec[0].tv_nsec,
 929                                      &spec[1].tv_sec, &spec[1].tv_nsec);
 930             if (retval > 0) {
 931                 retval = qemu_utimens(path.data, spec);
 932                 if (retval < 0) {
 933                     retval = -errno;
 934                 }
 935             }
 936             v9fs_string_free(&path);
 937             break;
 938         case T_RENAME:
 939             v9fs_string_init(&path);
 940             v9fs_string_init(&oldpath);
 941             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 942                                      "ss", &oldpath, &path);
 943             if (retval > 0) {
 944                 retval = rename(oldpath.data, path.data);
 945                 if (retval < 0) {
 946                     retval = -errno;
 947                 }
 948             }
 949             v9fs_string_free(&oldpath);
 950             v9fs_string_free(&path);
 951             break;
 952         case T_REMOVE:
 953             v9fs_string_init(&path);
 954             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "s", &path);
 955             if (retval > 0) {
 956                 retval = remove(path.data);
 957                 if (retval < 0) {
 958                     retval = -errno;
 959                 }
 960             }
 961             v9fs_string_free(&path);
 962             break;
 963         case T_LGETXATTR:
 964         case T_LLISTXATTR:
 965             retval = do_getxattr(header.type, &in_iovec, &out_iovec);
 966             break;
 967         case T_LSETXATTR:
 968             v9fs_string_init(&path);
 969             v9fs_string_init(&name);
 970             v9fs_string_init(&value);
 971             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sssdd", &path,
 972                                      &name, &value, &size, &flags);
 973             if (retval > 0) {
 974                 retval = lsetxattr(path.data,
 975                                    name.data, value.data, size, flags);
 976                 if (retval < 0) {
 977                     retval = -errno;
 978                 }
 979             }
 980             v9fs_string_free(&path);
 981             v9fs_string_free(&name);
 982             v9fs_string_free(&value);
 983             break;
 984         case T_LREMOVEXATTR:
 985             v9fs_string_init(&path);
 986             v9fs_string_init(&name);
 987             retval = proxy_unmarshal(&in_iovec,
 988                                      PROXY_HDR_SZ, "ss", &path, &name);
 989             if (retval > 0) {
 990                 retval = lremovexattr(path.data, name.data);
 991                 if (retval < 0) {
 992                     retval = -errno;
 993                 }
 994             }
 995             v9fs_string_free(&path);
 996             v9fs_string_free(&name);
 997             break;
 998         case T_GETVERSION:
 999             retval = do_getversion(&in_iovec, &out_iovec);
1000             break;
1001         default:
1002             goto err_out;
1003             break;
1004         }
1005
1006         if (process_reply(sock, header.type, &out_iovec, retval) < 0) {
1007             goto err_out;
1008         }
1009     }
1010 err_out:
1011     g_free(in_iovec.iov_base);
1012     g_free(out_iovec.iov_base);
1013     return -1;
1014 }
1015
1016 int main(int argc, char **argv)
1017 {
1018     int sock;
1019     uid_t own_u;
1020     gid_t own_g;
1021     char *rpath = NULL;
1022     char *sock_name = NULL;
1023     struct stat stbuf;
1024     int c, option_index;
1025 #ifdef FS_IOC_GETVERSION
1026     int retval;
1027     struct statfs st_fs;
1028 #endif
1029
1030     is_daemon = true;
1031     sock = -1;
1032     own_u = own_g = -1;
1033     while (1) {
1034         option_index = 0;
1035         c = getopt_long(argc, argv, "p:nh?f:s:u:g:", helper_opts,
1036                         &option_index);
1037         if (c == -1) {
1038             break;
1039         }
1040         switch (c) {
1041         case 'p':
1042             rpath = strdup(optarg);
1043             break;
1044         case 'n':
1045             is_daemon = false;
1046             break;
1047         case 'f':
1048             sock = atoi(optarg);
1049             break;
1050         case 's':
1051             sock_name = strdup(optarg);
1052             break;
1053         case 'u':
1054             own_u = atoi(optarg);
1055             break;
1056         case 'g':
1057             own_g = atoi(optarg);
1058             break;
1059         case '?':
1060         case 'h':
1061         default:
1062             usage(argv[0]);
1063             exit(EXIT_FAILURE);
1064         }
1065     }
1066
1067     /* Parameter validation */
1068     if ((sock_name == NULL && sock == -1) || rpath == NULL) {
1069         fprintf(stderr, "socket, socket descriptor or path not specified\n");
1070         usage(argv[0]);
1071         return -1;
1072     }
1073
1074     if (sock_name && sock != -1) {
1075         fprintf(stderr, "both named socket and socket descriptor specified\n");
1076         usage(argv[0]);
1077         exit(EXIT_FAILURE);
1078     }
1079
1080     if (sock_name && (own_u == -1 || own_g == -1)) {
1081         fprintf(stderr, "owner uid:gid not specified, ");
1082         fprintf(stderr,
1083                 "owner uid:gid specifies who can access the socket file\n");
1084         usage(argv[0]);
1085         exit(EXIT_FAILURE);
1086     }
1087
1088     if (lstat(rpath, &stbuf) < 0) {
1089         fprintf(stderr, "invalid path \"%s\" specified, %s\n",
1090                 rpath, strerror(errno));
1091         exit(EXIT_FAILURE);
1092     }
1093
1094     if (!S_ISDIR(stbuf.st_mode)) {
1095         fprintf(stderr, "specified path \"%s\" is not directory\n", rpath);
1096         exit(EXIT_FAILURE);
1097     }
1098
1099     if (is_daemon) {
1100         if (daemon(0, 0) < 0) {
1101             fprintf(stderr, "daemon call failed\n");
1102             exit(EXIT_FAILURE);
1103         }
1104         openlog(PROGNAME, LOG_PID, LOG_DAEMON);
1105     }
1106
1107     do_log(LOG_INFO, "Started\n");
1108     if (sock_name) {
1109         sock = proxy_socket(sock_name, own_u, own_g);
1110         if (sock < 0) {
1111             goto error;
1112         }
1113     }
1114
1115     get_version = false;
1116 #ifdef FS_IOC_GETVERSION
1117     /* check whether underlying FS support IOC_GETVERSION */
1118     retval = statfs(rpath, &st_fs);
1119     if (!retval) {
1120         switch (st_fs.f_type) {
1121         case EXT2_SUPER_MAGIC:
1122         case BTRFS_SUPER_MAGIC:
1123         case REISERFS_SUPER_MAGIC:
1124         case XFS_SUPER_MAGIC:
1125             get_version = true;
1126             break;
1127         }
1128     }
1129 #endif
1130
1131     if (chdir("/") < 0) {
1132         do_perror("chdir");
1133         goto error;
1134     }
1135     if (chroot(rpath) < 0) {
1136         do_perror("chroot");
1137         goto error;
1138     }
1139     umask(0);
1140
1141     if (init_capabilities() < 0) {
1142         goto error;
1143     }
1144
1145     process_requests(sock);
1146 error:
1147     do_log(LOG_INFO, "Done\n");
1148     closelog();
1149     return 0;
1150 }