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/sockets.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
  55 static bool is_daemon;
  56 static bool get_version; /* IOC getversion IOCTL supported */
  57
  58 static void GCC_FMT_ATTR(2, 3) do_log(int loglevel, const char *format, ...)
  59 {
  60     va_list ap;
  61
  62     va_start(ap, format);
  63     if (is_daemon) {
  64         vsyslog(LOG_CRIT, format, ap);
  65     } else {
  66         vfprintf(stderr, format, ap);
  67     }
  68     va_end(ap);
  69 }
  70
  71 static void do_perror(const char *string)
  72 {
  73     if (is_daemon) {
  74         syslog(LOG_CRIT, "%s:%s", string, strerror(errno));
  75     } else {
  76         fprintf(stderr, "%s:%s\n", string, strerror(errno));
  77     }
  78 }
  79
  80 static int do_cap_set(cap_value_t *cap_value, int size, int reset)
  81 {
  82     cap_t caps;
  83     if (reset) {
  84         /*
  85          * Start with an empty set and set permitted and effective
  86          */
  87         caps = cap_init();
  88         if (caps == NULL) {
  89             do_perror("cap_init");
  90             return -1;
  91         }
  92         if (cap_set_flag(caps, CAP_PERMITTED, size, cap_value, CAP_SET) < 0) {
  93             do_perror("cap_set_flag");
  94             goto error;
  95         }
  96     } else {
  97         caps = cap_get_proc();
  98         if (!caps) {
  99             do_perror("cap_get_proc");
 100             return -1;
 101         }
 102     }
 103     if (cap_set_flag(caps, CAP_EFFECTIVE, size, cap_value, CAP_SET) < 0) {
 104         do_perror("cap_set_flag");
 105         goto error;
 106     }
 107     if (cap_set_proc(caps) < 0) {
 108         do_perror("cap_set_proc");
 109         goto error;
 110     }
 111     cap_free(caps);
 112     return 0;
 113
 114 error:
 115     cap_free(caps);
 116     return -1;
 117 }
 118
 119 static int init_capabilities(void)
 120 {
 121     /* helper needs following capabilities only */
 122     cap_value_t cap_list[] = {
 123         CAP_CHOWN,
 124         CAP_DAC_OVERRIDE,
 125         CAP_FOWNER,
 126         CAP_FSETID,
 127         CAP_SETGID,
 128         CAP_MKNOD,
 129         CAP_SETUID,
 130     };
 131     return do_cap_set(cap_list, ARRAY_SIZE(cap_list), 1);
 132 }
 133
 134 static int socket_read(int sockfd, void *buff, ssize_t size)
 135 {
 136     ssize_t retval, total = 0;
 137
 138     while (size) {
 139         retval = read(sockfd, buff, size);
 140         if (retval == 0) {
 141             return -EIO;
 142         }
 143         if (retval < 0) {
 144             if (errno == EINTR) {
 145                 continue;
 146             }
 147             return -errno;
 148         }
 149         size -= retval;
 150         buff += retval;
 151         total += retval;
 152     }
 153     return total;
 154 }
 155
 156 static int socket_write(int sockfd, void *buff, ssize_t size)
 157 {
 158     ssize_t retval, total = 0;
 159
 160     while (size) {
 161         retval = write(sockfd, buff, size);
 162         if (retval < 0) {
 163             if (errno == EINTR) {
 164                 continue;
 165             }
 166             return -errno;
 167         }
 168         size -= retval;
 169         buff += retval;
 170         total += retval;
 171     }
 172     return total;
 173 }
 174
 175 static int read_request(int sockfd, struct iovec *iovec, ProxyHeader *header)
 176 {
 177     int retval;
 178
 179     /*
 180      * read the request header.
 181      */
 182     iovec->iov_len = 0;
 183     retval = socket_read(sockfd, iovec->iov_base, PROXY_HDR_SZ);
 184     if (retval < 0) {
 185         return retval;
 186     }
 187     iovec->iov_len = PROXY_HDR_SZ;
 188     retval = proxy_unmarshal(iovec, 0, "dd", &header->type, &header->size);
 189     if (retval < 0) {
 190         return retval;
 191     }
 192     /*
 193      * We can't process message.size > PROXY_MAX_IO_SZ.
 194      * Treat it as fatal error
 195      */
 196     if (header->size > PROXY_MAX_IO_SZ) {
 197         return -ENOBUFS;
 198     }
 199     retval = socket_read(sockfd, iovec->iov_base + PROXY_HDR_SZ, header->size);
 200     if (retval < 0) {
 201         return retval;
 202     }
 203     iovec->iov_len += header->size;
 204     return 0;
 205 }
 206
 207 static int send_fd(int sockfd, int fd)
 208 {
 209     struct msghdr msg;
 210     struct iovec iov;
 211     int retval, data;
 212     struct cmsghdr *cmsg;
 213     union MsgControl msg_control;
 214
 215     iov.iov_base = &data;
 216     iov.iov_len = sizeof(data);
 217
 218     memset(&msg, 0, sizeof(msg));
 219     msg.msg_iov = &iov;
 220     msg.msg_iovlen = 1;
 221     /* No ancillary data on error */
 222     if (fd < 0) {
 223         /* fd is really negative errno if the request failed  */
 224         data = fd;
 225     } else {
 226         data = V9FS_FD_VALID;
 227         msg.msg_control = &msg_control;
 228         msg.msg_controllen = sizeof(msg_control);
 229
 230         cmsg = &msg_control.cmsg;
 231         cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 232         cmsg->cmsg_level = SOL_SOCKET;
 233         cmsg->cmsg_type = SCM_RIGHTS;
 234         memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
 235     }
 236
 237     do {
 238         retval = sendmsg(sockfd, &msg, 0);
 239     } while (retval < 0 && errno == EINTR);
 240     if (fd >= 0) {
 241         close(fd);
 242     }
 243     if (retval < 0) {
 244         return retval;
 245     }
 246     return 0;
 247 }
 248
 249 static int send_status(int sockfd, struct iovec *iovec, int status)
 250 {
 251     ProxyHeader header;
 252     int retval, msg_size;
 253
 254     if (status < 0) {
 255         header.type = T_ERROR;
 256     } else {
 257         header.type = T_SUCCESS;
 258     }
 259     header.size = sizeof(status);
 260     /*
 261      * marshal the return status. We don't check error.
 262      * because we are sure we have enough space for the status
 263      */
 264     msg_size = proxy_marshal(iovec, 0, "ddd", header.type,
 265                              header.size, status);
 266     if (msg_size < 0) {
 267         return msg_size;
 268     }
 269     retval = socket_write(sockfd, iovec->iov_base, msg_size);
 270     if (retval < 0) {
 271         return retval;
 272     }
 273     return 0;
 274 }
 275
 276 /*
 277  * from man 7 capabilities, section
 278  * Effect of User ID Changes on Capabilities:
 279  * If the effective user ID is changed from nonzero to 0, then the permitted
 280  * set is copied to the effective set.  If the effective user ID is changed
 281  * from 0 to nonzero, then all capabilities are are cleared from the effective
 282  * set.
 283  *
 284  * The setfsuid/setfsgid man pages warn that changing the effective user ID may
 285  * expose the program to unwanted signals, but this is not true anymore: for an
 286  * unprivileged (without CAP_KILL) program to send a signal, the real or
 287  * effective user ID of the sending process must equal the real or saved user
 288  * ID of the target process.  Even when dropping privileges, it is enough to
 289  * keep the saved UID to a "privileged" value and virtfs-proxy-helper won't
 290  * be exposed to signals.  So just use setresuid/setresgid.
 291  */
 292 static int setugid(int uid, int gid, int *suid, int *sgid)
 293 {
 294     int retval;
 295
 296     /*
 297      * We still need DAC_OVERRIDE because we don't change
 298      * supplementary group ids, and hence may be subjected DAC rules
 299      */
 300     cap_value_t cap_list[] = {
 301         CAP_DAC_OVERRIDE,
 302     };
 303
 304     *suid = geteuid();
 305     *sgid = getegid();
 306
 307     if (setresgid(-1, gid, *sgid) == -1) {
 308         retval = -errno;
 309         goto err_out;
 310     }
 311
 312     if (setresuid(-1, uid, *suid) == -1) {
 313         retval = -errno;
 314         goto err_sgid;
 315     }
 316
 317     if (uid != 0 || gid != 0) {
 318         if (do_cap_set(cap_list, ARRAY_SIZE(cap_list), 0) < 0) {
 319             retval = -errno;
 320             goto err_suid;
 321         }
 322     }
 323     return 0;
 324
 325 err_suid:
 326     if (setresuid(-1, *suid, *suid) == -1) {
 327         abort();
 328     }
 329 err_sgid:
 330     if (setresgid(-1, *sgid, *sgid) == -1) {
 331         abort();
 332     }
 333 err_out:
 334     return retval;
 335 }
 336
 337 /*
 338  * This is used to reset the ugid back with the saved values
 339  * There is nothing much we can do checking error values here.
 340  */
 341 static void resetugid(int suid, int sgid)
 342 {
 343     if (setresgid(-1, sgid, sgid) == -1) {
 344         abort();
 345     }
 346     if (setresuid(-1, suid, suid) == -1) {
 347         abort();
 348     }
 349 }
 350
 351 /*
 352  * send response in two parts
 353  * 1) ProxyHeader
 354  * 2) Response or error status
 355  * This function should be called with marshaled response
 356  * send_response constructs header part and error part only.
 357  * send response sends {ProxyHeader,Response} if the request was success
 358  * otherwise sends {ProxyHeader,error status}
 359  */
 360 static int send_response(int sock, struct iovec *iovec, int size)
 361 {
 362     int retval;
 363     ProxyHeader header;
 364
 365     /*
 366      * If response size exceeds available iovec->iov_len,
 367      * we return ENOBUFS
 368      */
 369     if (size > PROXY_MAX_IO_SZ) {
 370         size = -ENOBUFS;
 371     }
 372
 373     if (size < 0) {
 374         /*
 375          * In case of error we would not have got the error encoded
 376          * already so encode the error here.
 377          */
 378         header.type = T_ERROR;
 379         header.size = sizeof(size);
 380         proxy_marshal(iovec, PROXY_HDR_SZ, "d", size);
 381     } else {
 382         header.type = T_SUCCESS;
 383         header.size = size;
 384     }
 385     proxy_marshal(iovec, 0, "dd", header.type, header.size);
 386     retval = socket_write(sock, iovec->iov_base, header.size + PROXY_HDR_SZ);
 387     if (retval < 0) {
 388         return retval;
 389     }
 390     return 0;
 391 }
 392
 393 /*
 394  * gets generation number
 395  * returns -errno on failure and sizeof(generation number) on success
 396  */
 397 static int do_getversion(struct iovec *iovec, struct iovec *out_iovec)
 398 {
 399     uint64_t version;
 400     int retval = -ENOTTY;
 401 #ifdef FS_IOC_GETVERSION
 402     int fd;
 403     V9fsString path;
 404 #endif
 405
 406
 407     /* no need to issue ioctl */
 408     if (!get_version) {
 409         version = 0;
 410         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
 411         return retval;
 412     }
 413 #ifdef FS_IOC_GETVERSION
 414     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
 415     if (retval < 0) {
 416         return retval;
 417     }
 418
 419     fd = open(path.data, O_RDONLY);
 420     if (fd < 0) {
 421         retval = -errno;
 422         goto err_out;
 423     }
 424     if (ioctl(fd, FS_IOC_GETVERSION, &version) < 0) {
 425         retval = -errno;
 426     } else {
 427         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
 428     }
 429     close(fd);
 430 err_out:
 431     v9fs_string_free(&path);
 432 #endif
 433     return retval;
 434 }
 435
 436 static int do_getxattr(int type, struct iovec *iovec, struct iovec *out_iovec)
 437 {
 438     int size = 0, offset, retval;
 439     V9fsString path, name, xattr;
 440
 441     v9fs_string_init(&xattr);
 442     v9fs_string_init(&path);
 443     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "ds", &size, &path);
 444     if (retval < 0) {
 445         return retval;
 446     }
 447     offset = PROXY_HDR_SZ + retval;
 448
 449     if (size) {
 450         xattr.data = g_malloc(size);
 451         xattr.size = size;
 452     }
 453     switch (type) {
 454     case T_LGETXATTR:
 455         v9fs_string_init(&name);
 456         retval = proxy_unmarshal(iovec, offset, "s", &name);
 457         if (retval > 0) {
 458             retval = lgetxattr(path.data, name.data, xattr.data, size);
 459             if (retval < 0) {
 460                 retval = -errno;
 461             } else {
 462                 xattr.size = retval;
 463             }
 464         }
 465         v9fs_string_free(&name);
 466         break;
 467     case T_LLISTXATTR:
 468         retval = llistxattr(path.data, xattr.data, size);
 469         if (retval < 0) {
 470             retval = -errno;
 471         } else {
 472             xattr.size = retval;
 473         }
 474         break;
 475     }
 476     if (retval < 0) {
 477         goto err_out;
 478     }
 479
 480     if (!size) {
 481         proxy_marshal(out_iovec, PROXY_HDR_SZ, "d", retval);
 482         retval = sizeof(retval);
 483     } else {
 484         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &xattr);
 485     }
 486 err_out:
 487     v9fs_string_free(&xattr);
 488     v9fs_string_free(&path);
 489     return retval;
 490 }
 491
 492 static void stat_to_prstat(ProxyStat *pr_stat, struct stat *stat)
 493 {
 494     memset(pr_stat, 0, sizeof(*pr_stat));
 495     pr_stat->st_dev = stat->st_dev;
 496     pr_stat->st_ino = stat->st_ino;
 497     pr_stat->st_nlink = stat->st_nlink;
 498     pr_stat->st_mode = stat->st_mode;
 499     pr_stat->st_uid = stat->st_uid;
 500     pr_stat->st_gid = stat->st_gid;
 501     pr_stat->st_rdev = stat->st_rdev;
 502     pr_stat->st_size = stat->st_size;
 503     pr_stat->st_blksize = stat->st_blksize;
 504     pr_stat->st_blocks = stat->st_blocks;
 505     pr_stat->st_atim_sec = stat->st_atim.tv_sec;
 506     pr_stat->st_atim_nsec = stat->st_atim.tv_nsec;
 507     pr_stat->st_mtim_sec = stat->st_mtim.tv_sec;
 508     pr_stat->st_mtim_nsec = stat->st_mtim.tv_nsec;
 509     pr_stat->st_ctim_sec = stat->st_ctim.tv_sec;
 510     pr_stat->st_ctim_nsec = stat->st_ctim.tv_nsec;
 511 }
 512
 513 static void statfs_to_prstatfs(ProxyStatFS *pr_stfs, struct statfs *stfs)
 514 {
 515     memset(pr_stfs, 0, sizeof(*pr_stfs));
 516     pr_stfs->f_type = stfs->f_type;
 517     pr_stfs->f_bsize = stfs->f_bsize;
 518     pr_stfs->f_blocks = stfs->f_blocks;
 519     pr_stfs->f_bfree = stfs->f_bfree;
 520     pr_stfs->f_bavail = stfs->f_bavail;
 521     pr_stfs->f_files = stfs->f_files;
 522     pr_stfs->f_ffree = stfs->f_ffree;
 523     pr_stfs->f_fsid[0] = stfs->f_fsid.__val[0];
 524     pr_stfs->f_fsid[1] = stfs->f_fsid.__val[1];
 525     pr_stfs->f_namelen = stfs->f_namelen;
 526     pr_stfs->f_frsize = stfs->f_frsize;
 527 }
 528
 529 /*
 530  * Gets stat/statfs information and packs in out_iovec structure
 531  * on success returns number of bytes packed in out_iovec struture
 532  * otherwise returns -errno
 533  */
 534 static int do_stat(int type, struct iovec *iovec, struct iovec *out_iovec)
 535 {
 536     int retval;
 537     V9fsString path;
 538     ProxyStat pr_stat;
 539     ProxyStatFS pr_stfs;
 540     struct stat st_buf;
 541     struct statfs stfs_buf;
 542
 543     v9fs_string_init(&path);
 544     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
 545     if (retval < 0) {
 546         return retval;
 547     }
 548
 549     switch (type) {
 550     case T_LSTAT:
 551         retval = lstat(path.data, &st_buf);
 552         if (retval < 0) {
 553             retval = -errno;
 554         } else {
 555             stat_to_prstat(&pr_stat, &st_buf);
 556             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
 557                                    "qqqdddqqqqqqqqqq", pr_stat.st_dev,
 558                                    pr_stat.st_ino, pr_stat.st_nlink,
 559                                    pr_stat.st_mode, pr_stat.st_uid,
 560                                    pr_stat.st_gid, pr_stat.st_rdev,
 561                                    pr_stat.st_size, pr_stat.st_blksize,
 562                                    pr_stat.st_blocks,
 563                                    pr_stat.st_atim_sec, pr_stat.st_atim_nsec,
 564                                    pr_stat.st_mtim_sec, pr_stat.st_mtim_nsec,
 565                                    pr_stat.st_ctim_sec, pr_stat.st_ctim_nsec);
 566         }
 567         break;
 568     case T_STATFS:
 569         retval = statfs(path.data, &stfs_buf);
 570         if (retval < 0) {
 571             retval = -errno;
 572         } else {
 573             statfs_to_prstatfs(&pr_stfs, &stfs_buf);
 574             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
 575                                    "qqqqqqqqqqq", pr_stfs.f_type,
 576                                    pr_stfs.f_bsize, pr_stfs.f_blocks,
 577                                    pr_stfs.f_bfree, pr_stfs.f_bavail,
 578                                    pr_stfs.f_files, pr_stfs.f_ffree,
 579                                    pr_stfs.f_fsid[0], pr_stfs.f_fsid[1],
 580                                    pr_stfs.f_namelen, pr_stfs.f_frsize);
 581         }
 582         break;
 583     }
 584     v9fs_string_free(&path);
 585     return retval;
 586 }
 587
 588 static int do_readlink(struct iovec *iovec, struct iovec *out_iovec)
 589 {
 590     char *buffer;
 591     int size, retval;
 592     V9fsString target, path;
 593
 594     v9fs_string_init(&path);
 595     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &size);
 596     if (retval < 0) {
 597         v9fs_string_free(&path);
 598         return retval;
 599     }
 600     buffer = g_malloc(size);
 601     v9fs_string_init(&target);
 602     retval = readlink(path.data, buffer, size - 1);
 603     if (retval > 0) {
 604         buffer[retval] = '\0';
 605         v9fs_string_sprintf(&target, "%s", buffer);
 606         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &target);
 607     } else {
 608         retval = -errno;
 609     }
 610     g_free(buffer);
 611     v9fs_string_free(&target);
 612     v9fs_string_free(&path);
 613     return retval;
 614 }
 615
 616 /*
 617  * create other filesystem objects and send 0 on success
 618  * return -errno on error
 619  */
 620 static int do_create_others(int type, struct iovec *iovec)
 621 {
 622     dev_t rdev;
 623     int retval = 0;
 624     int offset = PROXY_HDR_SZ;
 625     V9fsString oldpath, path;
 626     int mode, uid, gid, cur_uid, cur_gid;
 627
 628     v9fs_string_init(&path);
 629     v9fs_string_init(&oldpath);
 630
 631     retval = proxy_unmarshal(iovec, offset, "dd", &uid, &gid);
 632     if (retval < 0) {
 633         return retval;
 634     }
 635     offset += retval;
 636     retval = setugid(uid, gid, &cur_uid, &cur_gid);
 637     if (retval < 0) {
 638         goto unmarshal_err_out;
 639     }
 640     switch (type) {
 641     case T_MKNOD:
 642         retval = proxy_unmarshal(iovec, offset, "sdq", &path, &mode, &rdev);
 643         if (retval < 0) {
 644             goto err_out;
 645         }
 646         retval = mknod(path.data, mode, rdev);
 647         break;
 648     case T_MKDIR:
 649         retval = proxy_unmarshal(iovec, offset, "sd", &path, &mode);
 650         if (retval < 0) {
 651             goto err_out;
 652         }
 653         retval = mkdir(path.data, mode);
 654         break;
 655     case T_SYMLINK:
 656         retval = proxy_unmarshal(iovec, offset, "ss", &oldpath, &path);
 657         if (retval < 0) {
 658             goto err_out;
 659         }
 660         retval = symlink(oldpath.data, path.data);
 661         break;
 662     }
 663     if (retval < 0) {
 664         retval = -errno;
 665     }
 666
 667 err_out:
 668     resetugid(cur_uid, cur_gid);
 669 unmarshal_err_out:
 670     v9fs_string_free(&path);
 671     v9fs_string_free(&oldpath);
 672     return retval;
 673 }
 674
 675 /*
 676  * create a file and send fd on success
 677  * return -errno on error
 678  */
 679 static int do_create(struct iovec *iovec)
 680 {
 681     int ret;
 682     V9fsString path;
 683     int flags, mode, uid, gid, cur_uid, cur_gid;
 684
 685     v9fs_string_init(&path);
 686     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sdddd",
 687                           &path, &flags, &mode, &uid, &gid);
 688     if (ret < 0) {
 689         goto unmarshal_err_out;
 690     }
 691     ret = setugid(uid, gid, &cur_uid, &cur_gid);
 692     if (ret < 0) {
 693         goto unmarshal_err_out;
 694     }
 695     ret = open(path.data, flags, mode);
 696     if (ret < 0) {
 697         ret = -errno;
 698     }
 699
 700     resetugid(cur_uid, cur_gid);
 701 unmarshal_err_out:
 702     v9fs_string_free(&path);
 703     return ret;
 704 }
 705
 706 /*
 707  * open a file and send fd on success
 708  * return -errno on error
 709  */
 710 static int do_open(struct iovec *iovec)
 711 {
 712     int flags, ret;
 713     V9fsString path;
 714
 715     v9fs_string_init(&path);
 716     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &flags);
 717     if (ret < 0) {
 718         goto err_out;
 719     }
 720     ret = open(path.data, flags);
 721     if (ret < 0) {
 722         ret = -errno;
 723     }
 724 err_out:
 725     v9fs_string_free(&path);
 726     return ret;
 727 }
 728
 729 /* create unix domain socket and return the descriptor */
 730 static int proxy_socket(const char *path, uid_t uid, gid_t gid)
 731 {
 732     int sock, client;
 733     struct sockaddr_un proxy, qemu;
 734     socklen_t size;
 735
 736     /* requested socket already exists, refuse to start */
 737     if (!access(path, F_OK)) {
 738         do_log(LOG_CRIT, "socket already exists\n");
 739         return -1;
 740     }
 741
 742     g_assert(strlen(path) < sizeof(proxy.sun_path));
 743     sock = socket(AF_UNIX, SOCK_STREAM, 0);
 744     if (sock < 0) {
 745         do_perror("socket");
 746         return -1;
 747     }
 748
 749     /* mask other part of mode bits */
 750     umask(7);
 751
 752     proxy.sun_family = AF_UNIX;
 753     strcpy(proxy.sun_path, path);
 754     if (bind(sock, (struct sockaddr *)&proxy,
 755             sizeof(struct sockaddr_un)) < 0) {
 756         do_perror("bind");
 757         goto error;
 758     }
 759     if (chown(proxy.sun_path, uid, gid) < 0) {
 760         do_perror("chown");
 761         goto error;
 762     }
 763     if (listen(sock, 1) < 0) {
 764         do_perror("listen");
 765         goto error;
 766     }
 767
 768     size = sizeof(qemu);
 769     client = accept(sock, (struct sockaddr *)&qemu, &size);
 770     if (client < 0) {
 771         do_perror("accept");
 772         goto error;
 773     }
 774     close(sock);
 775     return client;
 776
 777 error:
 778     close(sock);
 779     return -1;
 780 }
 781
 782 static void usage(char *prog)
 783 {
 784     fprintf(stderr, "usage: %s\n"
 785             " -p|--path <path> 9p path to export\n"
 786             " {-f|--fd <socket-descriptor>} socket file descriptor to be used\n"
 787             " {-s|--socket <socketname> socket file used for communication\n"
 788             " \t-u|--uid <uid> -g|--gid <gid>} - uid:gid combination to give "
 789             " access to this socket\n"
 790             " \tNote: -s & -f can not be used together\n"
 791             " [-n|--nodaemon] Run as a normal program\n",
 792             basename(prog));
 793 }
 794
 795 static int process_reply(int sock, int type,
 796                          struct iovec *out_iovec, int retval)
 797 {
 798     switch (type) {
 799     case T_OPEN:
 800     case T_CREATE:
 801         if (send_fd(sock, retval) < 0) {
 802             return -1;
 803         }
 804         break;
 805     case T_MKNOD:
 806     case T_MKDIR:
 807     case T_SYMLINK:
 808     case T_LINK:
 809     case T_CHMOD:
 810     case T_CHOWN:
 811     case T_TRUNCATE:
 812     case T_UTIME:
 813     case T_RENAME:
 814     case T_REMOVE:
 815     case T_LSETXATTR:
 816     case T_LREMOVEXATTR:
 817         if (send_status(sock, out_iovec, retval) < 0) {
 818             return -1;
 819         }
 820         break;
 821     case T_LSTAT:
 822     case T_STATFS:
 823     case T_READLINK:
 824     case T_LGETXATTR:
 825     case T_LLISTXATTR:
 826     case T_GETVERSION:
 827         if (send_response(sock, out_iovec, retval) < 0) {
 828             return -1;
 829         }
 830         break;
 831     default:
 832         return -1;
 833         break;
 834     }
 835     return 0;
 836 }
 837
 838 static int process_requests(int sock)
 839 {
 840     int flags;
 841     int size = 0;
 842     int retval = 0;
 843     uint64_t offset;
 844     ProxyHeader header;
 845     int mode, uid, gid;
 846     V9fsString name, value;
 847     struct timespec spec[2];
 848     V9fsString oldpath, path;
 849     struct iovec in_iovec, out_iovec;
 850
 851     in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
 852     in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
 853     out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
 854     out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
 855
 856     while (1) {
 857         /*
 858          * initialize the header type, so that we send
 859          * response to proper request type.
 860          */
 861         header.type = 0;
 862         retval = read_request(sock, &in_iovec, &header);
 863         if (retval < 0) {
 864             goto err_out;
 865         }
 866
 867         switch (header.type) {
 868         case T_OPEN:
 869             retval = do_open(&in_iovec);
 870             break;
 871         case T_CREATE:
 872             retval = do_create(&in_iovec);
 873             break;
 874         case T_MKNOD:
 875         case T_MKDIR:
 876         case T_SYMLINK:
 877             retval = do_create_others(header.type, &in_iovec);
 878             break;
 879         case T_LINK:
 880             v9fs_string_init(&path);
 881             v9fs_string_init(&oldpath);
 882             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 883                                      "ss", &oldpath, &path);
 884             if (retval > 0) {
 885                 retval = link(oldpath.data, path.data);
 886                 if (retval < 0) {
 887                     retval = -errno;
 888                 }
 889             }
 890             v9fs_string_free(&oldpath);
 891             v9fs_string_free(&path);
 892             break;
 893         case T_LSTAT:
 894         case T_STATFS:
 895             retval = do_stat(header.type, &in_iovec, &out_iovec);
 896             break;
 897         case T_READLINK:
 898             retval = do_readlink(&in_iovec, &out_iovec);
 899             break;
 900         case T_CHMOD:
 901             v9fs_string_init(&path);
 902             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 903                                      "sd", &path, &mode);
 904             if (retval > 0) {
 905                 retval = chmod(path.data, mode);
 906                 if (retval < 0) {
 907                     retval = -errno;
 908                 }
 909             }
 910             v9fs_string_free(&path);
 911             break;
 912         case T_CHOWN:
 913             v9fs_string_init(&path);
 914             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sdd", &path,
 915                                      &uid, &gid);
 916             if (retval > 0) {
 917                 retval = lchown(path.data, uid, gid);
 918                 if (retval < 0) {
 919                     retval = -errno;
 920                 }
 921             }
 922             v9fs_string_free(&path);
 923             break;
 924         case T_TRUNCATE:
 925             v9fs_string_init(&path);
 926             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sq",
 927                                      &path, &offset);
 928             if (retval > 0) {
 929                 retval = truncate(path.data, offset);
 930                 if (retval < 0) {
 931                     retval = -errno;
 932                 }
 933             }
 934             v9fs_string_free(&path);
 935             break;
 936         case T_UTIME:
 937             v9fs_string_init(&path);
 938             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sqqqq", &path,
 939                                      &spec[0].tv_sec, &spec[0].tv_nsec,
 940                                      &spec[1].tv_sec, &spec[1].tv_nsec);
 941             if (retval > 0) {
 942                 retval = qemu_utimens(path.data, spec);
 943                 if (retval < 0) {
 944                     retval = -errno;
 945                 }
 946             }
 947             v9fs_string_free(&path);
 948             break;
 949         case T_RENAME:
 950             v9fs_string_init(&path);
 951             v9fs_string_init(&oldpath);
 952             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 953                                      "ss", &oldpath, &path);
 954             if (retval > 0) {
 955                 retval = rename(oldpath.data, path.data);
 956                 if (retval < 0) {
 957                     retval = -errno;
 958                 }
 959             }
 960             v9fs_string_free(&oldpath);
 961             v9fs_string_free(&path);
 962             break;
 963         case T_REMOVE:
 964             v9fs_string_init(&path);
 965             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "s", &path);
 966             if (retval > 0) {
 967                 retval = remove(path.data);
 968                 if (retval < 0) {
 969                     retval = -errno;
 970                 }
 971             }
 972             v9fs_string_free(&path);
 973             break;
 974         case T_LGETXATTR:
 975         case T_LLISTXATTR:
 976             retval = do_getxattr(header.type, &in_iovec, &out_iovec);
 977             break;
 978         case T_LSETXATTR:
 979             v9fs_string_init(&path);
 980             v9fs_string_init(&name);
 981             v9fs_string_init(&value);
 982             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sssdd", &path,
 983                                      &name, &value, &size, &flags);
 984             if (retval > 0) {
 985                 retval = lsetxattr(path.data,
 986                                    name.data, value.data, size, flags);
 987                 if (retval < 0) {
 988                     retval = -errno;
 989                 }
 990             }
 991             v9fs_string_free(&path);
 992             v9fs_string_free(&name);
 993             v9fs_string_free(&value);
 994             break;
 995         case T_LREMOVEXATTR:
 996             v9fs_string_init(&path);
 997             v9fs_string_init(&name);
 998             retval = proxy_unmarshal(&in_iovec,
 999                                      PROXY_HDR_SZ, "ss", &path, &name);
1000             if (retval > 0) {
1001                 retval = lremovexattr(path.data, name.data);
1002                 if (retval < 0) {
1003                     retval = -errno;
1004                 }
1005             }
1006             v9fs_string_free(&path);
1007             v9fs_string_free(&name);
1008             break;
1009         case T_GETVERSION:
1010             retval = do_getversion(&in_iovec, &out_iovec);
1011             break;
1012         default:
1013             goto err_out;
1014             break;
1015         }
1016
1017         if (process_reply(sock, header.type, &out_iovec, retval) < 0) {
1018             goto err_out;
1019         }
1020     }
1021 err_out:
1022     g_free(in_iovec.iov_base);
1023     g_free(out_iovec.iov_base);
1024     return -1;
1025 }
1026
1027 int main(int argc, char **argv)
1028 {
1029     int sock;
1030     uid_t own_u;
1031     gid_t own_g;
1032     char *rpath = NULL;
1033     char *sock_name = NULL;
1034     struct stat stbuf;
1035     int c, option_index;
1036 #ifdef FS_IOC_GETVERSION
1037     int retval;
1038     struct statfs st_fs;
1039 #endif
1040
1041     is_daemon = true;
1042     sock = -1;
1043     own_u = own_g = -1;
1044     while (1) {
1045         option_index = 0;
1046         c = getopt_long(argc, argv, "p:nh?f:s:u:g:", helper_opts,
1047                         &option_index);
1048         if (c == -1) {
1049             break;
1050         }
1051         switch (c) {
1052         case 'p':
1053             rpath = g_strdup(optarg);
1054             break;
1055         case 'n':
1056             is_daemon = false;
1057             break;
1058         case 'f':
1059             sock = atoi(optarg);
1060             break;
1061         case 's':
1062             sock_name = g_strdup(optarg);
1063             break;
1064         case 'u':
1065             own_u = atoi(optarg);
1066             break;
1067         case 'g':
1068             own_g = atoi(optarg);
1069             break;
1070         case '?':
1071         case 'h':
1072         default:
1073             usage(argv[0]);
1074             exit(EXIT_FAILURE);
1075         }
1076     }
1077
1078     /* Parameter validation */
1079     if ((sock_name == NULL && sock == -1) || rpath == NULL) {
1080         fprintf(stderr, "socket, socket descriptor or path not specified\n");
1081         usage(argv[0]);
1082         return -1;
1083     }
1084
1085     if (sock_name && sock != -1) {
1086         fprintf(stderr, "both named socket and socket descriptor specified\n");
1087         usage(argv[0]);
1088         exit(EXIT_FAILURE);
1089     }
1090
1091     if (sock_name && (own_u == -1 || own_g == -1)) {
1092         fprintf(stderr, "owner uid:gid not specified, ");
1093         fprintf(stderr,
1094                 "owner uid:gid specifies who can access the socket file\n");
1095         usage(argv[0]);
1096         exit(EXIT_FAILURE);
1097     }
1098
1099     if (lstat(rpath, &stbuf) < 0) {
1100         fprintf(stderr, "invalid path \"%s\" specified, %s\n",
1101                 rpath, strerror(errno));
1102         exit(EXIT_FAILURE);
1103     }
1104
1105     if (!S_ISDIR(stbuf.st_mode)) {
1106         fprintf(stderr, "specified path \"%s\" is not directory\n", rpath);
1107         exit(EXIT_FAILURE);
1108     }
1109
1110     if (is_daemon) {
1111         if (daemon(0, 0) < 0) {
1112             fprintf(stderr, "daemon call failed\n");
1113             exit(EXIT_FAILURE);
1114         }
1115         openlog(PROGNAME, LOG_PID, LOG_DAEMON);
1116     }
1117
1118     do_log(LOG_INFO, "Started\n");
1119     if (sock_name) {
1120         sock = proxy_socket(sock_name, own_u, own_g);
1121         if (sock < 0) {
1122             goto error;
1123         }
1124     }
1125
1126     get_version = false;
1127 #ifdef FS_IOC_GETVERSION
1128     /* check whether underlying FS support IOC_GETVERSION */
1129     retval = statfs(rpath, &st_fs);
1130     if (!retval) {
1131         switch (st_fs.f_type) {
1132         case EXT2_SUPER_MAGIC:
1133         case BTRFS_SUPER_MAGIC:
1134         case REISERFS_SUPER_MAGIC:
1135         case XFS_SUPER_MAGIC:
1136             get_version = true;
1137             break;
1138         }
1139     }
1140 #endif
1141
1142     if (chdir("/") < 0) {
1143         do_perror("chdir");
1144         goto error;
1145     }
1146     if (chroot(rpath) < 0) {
1147         do_perror("chroot");
1148         goto error;
1149     }
1150     umask(0);
1151
1152     if (init_capabilities() < 0) {
1153         goto error;
1154     }
1155
1156     process_requests(sock);
1157 error:
1158     do_log(LOG_INFO, "Done\n");
1159     closelog();
1160     return 0;
1161 }