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 "qemu/osdep.h"
  13 #include <sys/resource.h>
  14 #include <getopt.h>
  15 #include <syslog.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 <cap-ng.h>
  24 #include "qemu-common.h"
  25 #include "qemu/sockets.h"
  26 #include "qemu/xattr.h"
  27 #include "9p-iov-marshal.h"
  28 #include "hw/9pfs/9p-proxy.h"
  29 #include "fsdev/9p-iov-marshal.h"
  30
  31 #define PROGNAME "virtfs-proxy-helper"
  32
  33 #ifndef XFS_SUPER_MAGIC
  34 #define XFS_SUPER_MAGIC  0x58465342
  35 #endif
  36 #ifndef EXT2_SUPER_MAGIC
  37 #define EXT2_SUPER_MAGIC 0xEF53
  38 #endif
  39 #ifndef REISERFS_SUPER_MAGIC
  40 #define REISERFS_SUPER_MAGIC 0x52654973
  41 #endif
  42 #ifndef BTRFS_SUPER_MAGIC
  43 #define BTRFS_SUPER_MAGIC 0x9123683E
  44 #endif
  45
  46 static const struct option helper_opts[] = {
  47     {"fd", required_argument, NULL, 'f'},
  48     {"path", required_argument, NULL, 'p'},
  49     {"nodaemon", no_argument, NULL, 'n'},
  50     {"socket", required_argument, NULL, 's'},
  51     {"uid", required_argument, NULL, 'u'},
  52     {"gid", required_argument, NULL, 'g'},
  53     {},
  54 };
  55
  56 static bool is_daemon;
  57 static bool get_version; /* IOC getversion IOCTL supported */
  58 static char *prog_name;
  59
  60 static void GCC_FMT_ATTR(2, 3) do_log(int loglevel, const char *format, ...)
  61 {
  62     va_list ap;
  63
  64     va_start(ap, format);
  65     if (is_daemon) {
  66         vsyslog(LOG_CRIT, format, ap);
  67     } else {
  68         vfprintf(stderr, format, ap);
  69     }
  70     va_end(ap);
  71 }
  72
  73 static void do_perror(const char *string)
  74 {
  75     if (is_daemon) {
  76         syslog(LOG_CRIT, "%s:%s", string, strerror(errno));
  77     } else {
  78         fprintf(stderr, "%s:%s\n", string, strerror(errno));
  79     }
  80 }
  81
  82 static int init_capabilities(void)
  83 {
  84     /* helper needs following capabilities only */
  85     int cap_list[] = {
  86         CAP_CHOWN,
  87         CAP_DAC_OVERRIDE,
  88         CAP_FOWNER,
  89         CAP_FSETID,
  90         CAP_SETGID,
  91         CAP_MKNOD,
  92         CAP_SETUID,
  93     };
  94     int i;
  95
  96     capng_clear(CAPNG_SELECT_BOTH);
  97     for (i = 0; i < ARRAY_SIZE(cap_list); i++) {
  98         if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED,
  99                          cap_list[i]) < 0) {
 100             do_perror("capng_update");
 101             return -1;
 102         }
 103     }
 104     if (capng_apply(CAPNG_SELECT_BOTH) < 0) {
 105         do_perror("capng_apply");
 106         return -1;
 107     }
 108
 109     /* Prepare effective set for setugid.  */
 110     for (i = 0; i < ARRAY_SIZE(cap_list); i++) {
 111         if (cap_list[i] == CAP_DAC_OVERRIDE) {
 112             continue;
 113         }
 114
 115         if (capng_update(CAPNG_DROP, CAPNG_EFFECTIVE,
 116                          cap_list[i]) < 0) {
 117             do_perror("capng_update");
 118             return -1;
 119         }
 120     }
 121     return 0;
 122 }
 123
 124 static int socket_read(int sockfd, void *buff, ssize_t size)
 125 {
 126     ssize_t retval, total = 0;
 127
 128     while (size) {
 129         retval = read(sockfd, buff, size);
 130         if (retval == 0) {
 131             return -EIO;
 132         }
 133         if (retval < 0) {
 134             if (errno == EINTR) {
 135                 continue;
 136             }
 137             return -errno;
 138         }
 139         size -= retval;
 140         buff += retval;
 141         total += retval;
 142     }
 143     return total;
 144 }
 145
 146 static int socket_write(int sockfd, void *buff, ssize_t size)
 147 {
 148     ssize_t retval, total = 0;
 149
 150     while (size) {
 151         retval = write(sockfd, buff, size);
 152         if (retval < 0) {
 153             if (errno == EINTR) {
 154                 continue;
 155             }
 156             return -errno;
 157         }
 158         size -= retval;
 159         buff += retval;
 160         total += retval;
 161     }
 162     return total;
 163 }
 164
 165 static int read_request(int sockfd, struct iovec *iovec, ProxyHeader *header)
 166 {
 167     int retval;
 168
 169     /*
 170      * read the request header.
 171      */
 172     iovec->iov_len = 0;
 173     retval = socket_read(sockfd, iovec->iov_base, PROXY_HDR_SZ);
 174     if (retval < 0) {
 175         return retval;
 176     }
 177     iovec->iov_len = PROXY_HDR_SZ;
 178     retval = proxy_unmarshal(iovec, 0, "dd", &header->type, &header->size);
 179     if (retval < 0) {
 180         return retval;
 181     }
 182     /*
 183      * We can't process message.size > PROXY_MAX_IO_SZ.
 184      * Treat it as fatal error
 185      */
 186     if (header->size > PROXY_MAX_IO_SZ) {
 187         return -ENOBUFS;
 188     }
 189     retval = socket_read(sockfd, iovec->iov_base + PROXY_HDR_SZ, header->size);
 190     if (retval < 0) {
 191         return retval;
 192     }
 193     iovec->iov_len += header->size;
 194     return 0;
 195 }
 196
 197 static int send_fd(int sockfd, int fd)
 198 {
 199     struct msghdr msg;
 200     struct iovec iov;
 201     int retval, data;
 202     struct cmsghdr *cmsg;
 203     union MsgControl msg_control;
 204
 205     iov.iov_base = &data;
 206     iov.iov_len = sizeof(data);
 207
 208     memset(&msg, 0, sizeof(msg));
 209     msg.msg_iov = &iov;
 210     msg.msg_iovlen = 1;
 211     /* No ancillary data on error */
 212     if (fd < 0) {
 213         /* fd is really negative errno if the request failed  */
 214         data = fd;
 215     } else {
 216         data = V9FS_FD_VALID;
 217         msg.msg_control = &msg_control;
 218         msg.msg_controllen = sizeof(msg_control);
 219
 220         cmsg = &msg_control.cmsg;
 221         cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
 222         cmsg->cmsg_level = SOL_SOCKET;
 223         cmsg->cmsg_type = SCM_RIGHTS;
 224         memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
 225     }
 226
 227     do {
 228         retval = sendmsg(sockfd, &msg, 0);
 229     } while (retval < 0 && errno == EINTR);
 230     if (fd >= 0) {
 231         close(fd);
 232     }
 233     if (retval < 0) {
 234         return retval;
 235     }
 236     return 0;
 237 }
 238
 239 static int send_status(int sockfd, struct iovec *iovec, int status)
 240 {
 241     ProxyHeader header;
 242     int retval, msg_size;
 243
 244     if (status < 0) {
 245         header.type = T_ERROR;
 246     } else {
 247         header.type = T_SUCCESS;
 248     }
 249     header.size = sizeof(status);
 250     /*
 251      * marshal the return status. We don't check error.
 252      * because we are sure we have enough space for the status
 253      */
 254     msg_size = proxy_marshal(iovec, 0, "ddd", header.type,
 255                              header.size, status);
 256     if (msg_size < 0) {
 257         return msg_size;
 258     }
 259     retval = socket_write(sockfd, iovec->iov_base, msg_size);
 260     if (retval < 0) {
 261         return retval;
 262     }
 263     return 0;
 264 }
 265
 266 /*
 267  * from man 7 capabilities, section
 268  * Effect of User ID Changes on Capabilities:
 269  * If the effective user ID is changed from nonzero to 0, then the permitted
 270  * set is copied to the effective set.  If the effective user ID is changed
 271  * from 0 to nonzero, then all capabilities are are cleared from the effective
 272  * set.
 273  *
 274  * The setfsuid/setfsgid man pages warn that changing the effective user ID may
 275  * expose the program to unwanted signals, but this is not true anymore: for an
 276  * unprivileged (without CAP_KILL) program to send a signal, the real or
 277  * effective user ID of the sending process must equal the real or saved user
 278  * ID of the target process.  Even when dropping privileges, it is enough to
 279  * keep the saved UID to a "privileged" value and virtfs-proxy-helper won't
 280  * be exposed to signals.  So just use setresuid/setresgid.
 281  */
 282 static int setugid(int uid, int gid, int *suid, int *sgid)
 283 {
 284     int retval;
 285
 286     *suid = geteuid();
 287     *sgid = getegid();
 288
 289     if (setresgid(-1, gid, *sgid) == -1) {
 290         return -errno;
 291     }
 292
 293     if (setresuid(-1, uid, *suid) == -1) {
 294         retval = -errno;
 295         goto err_sgid;
 296     }
 297
 298     if (uid == 0 && gid == 0) {
 299         /* Linux has already copied the permitted set to the effective set.  */
 300         return 0;
 301     }
 302
 303     /*
 304      * All capabilities have been cleared from the effective set.  However
 305      * we still need DAC_OVERRIDE because we don't change supplementary
 306      * group ids, and hence may be subject to DAC rules.  init_capabilities
 307      * left the set of capabilities that we want in libcap-ng's state.
 308      */
 309     if (capng_apply(CAPNG_SELECT_CAPS) < 0) {
 310         retval = -errno;
 311         do_perror("capng_apply");
 312         goto err_suid;
 313     }
 314     return 0;
 315
 316 err_suid:
 317     if (setresuid(-1, *suid, *suid) == -1) {
 318         abort();
 319     }
 320 err_sgid:
 321     if (setresgid(-1, *sgid, *sgid) == -1) {
 322         abort();
 323     }
 324     return retval;
 325 }
 326
 327 /*
 328  * This is used to reset the ugid back with the saved values
 329  * There is nothing much we can do checking error values here.
 330  */
 331 static void resetugid(int suid, int sgid)
 332 {
 333     if (setresgid(-1, sgid, sgid) == -1) {
 334         abort();
 335     }
 336     if (setresuid(-1, suid, suid) == -1) {
 337         abort();
 338     }
 339 }
 340
 341 /*
 342  * send response in two parts
 343  * 1) ProxyHeader
 344  * 2) Response or error status
 345  * This function should be called with marshaled response
 346  * send_response constructs header part and error part only.
 347  * send response sends {ProxyHeader,Response} if the request was success
 348  * otherwise sends {ProxyHeader,error status}
 349  */
 350 static int send_response(int sock, struct iovec *iovec, int size)
 351 {
 352     int retval;
 353     ProxyHeader header;
 354
 355     /*
 356      * If response size exceeds available iovec->iov_len,
 357      * we return ENOBUFS
 358      */
 359     if (size > PROXY_MAX_IO_SZ) {
 360         size = -ENOBUFS;
 361     }
 362
 363     if (size < 0) {
 364         /*
 365          * In case of error we would not have got the error encoded
 366          * already so encode the error here.
 367          */
 368         header.type = T_ERROR;
 369         header.size = sizeof(size);
 370         proxy_marshal(iovec, PROXY_HDR_SZ, "d", size);
 371     } else {
 372         header.type = T_SUCCESS;
 373         header.size = size;
 374     }
 375     proxy_marshal(iovec, 0, "dd", header.type, header.size);
 376     retval = socket_write(sock, iovec->iov_base, header.size + PROXY_HDR_SZ);
 377     if (retval < 0) {
 378         return retval;
 379     }
 380     return 0;
 381 }
 382
 383 /*
 384  * gets generation number
 385  * returns -errno on failure and sizeof(generation number) on success
 386  */
 387 static int do_getversion(struct iovec *iovec, struct iovec *out_iovec)
 388 {
 389     uint64_t version;
 390     int retval = -ENOTTY;
 391 #ifdef FS_IOC_GETVERSION
 392     int fd;
 393     V9fsString path;
 394 #endif
 395
 396
 397     /* no need to issue ioctl */
 398     if (!get_version) {
 399         version = 0;
 400         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
 401         return retval;
 402     }
 403 #ifdef FS_IOC_GETVERSION
 404     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
 405     if (retval < 0) {
 406         return retval;
 407     }
 408
 409     fd = open(path.data, O_RDONLY);
 410     if (fd < 0) {
 411         retval = -errno;
 412         goto err_out;
 413     }
 414     if (ioctl(fd, FS_IOC_GETVERSION, &version) < 0) {
 415         retval = -errno;
 416     } else {
 417         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
 418     }
 419     close(fd);
 420 err_out:
 421     v9fs_string_free(&path);
 422 #endif
 423     return retval;
 424 }
 425
 426 static int do_getxattr(int type, struct iovec *iovec, struct iovec *out_iovec)
 427 {
 428     int size = 0, offset, retval;
 429     V9fsString path, name, xattr;
 430
 431     v9fs_string_init(&xattr);
 432     v9fs_string_init(&path);
 433     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "ds", &size, &path);
 434     if (retval < 0) {
 435         return retval;
 436     }
 437     offset = PROXY_HDR_SZ + retval;
 438
 439     if (size) {
 440         xattr.data = g_malloc(size);
 441         xattr.size = size;
 442     }
 443     switch (type) {
 444     case T_LGETXATTR:
 445         v9fs_string_init(&name);
 446         retval = proxy_unmarshal(iovec, offset, "s", &name);
 447         if (retval > 0) {
 448             retval = lgetxattr(path.data, name.data, xattr.data, size);
 449             if (retval < 0) {
 450                 retval = -errno;
 451             } else {
 452                 xattr.size = retval;
 453             }
 454         }
 455         v9fs_string_free(&name);
 456         break;
 457     case T_LLISTXATTR:
 458         retval = llistxattr(path.data, xattr.data, size);
 459         if (retval < 0) {
 460             retval = -errno;
 461         } else {
 462             xattr.size = retval;
 463         }
 464         break;
 465     }
 466     if (retval < 0) {
 467         goto err_out;
 468     }
 469
 470     if (!size) {
 471         proxy_marshal(out_iovec, PROXY_HDR_SZ, "d", retval);
 472         retval = sizeof(retval);
 473     } else {
 474         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &xattr);
 475     }
 476 err_out:
 477     v9fs_string_free(&xattr);
 478     v9fs_string_free(&path);
 479     return retval;
 480 }
 481
 482 static void stat_to_prstat(ProxyStat *pr_stat, struct stat *stat)
 483 {
 484     memset(pr_stat, 0, sizeof(*pr_stat));
 485     pr_stat->st_dev = stat->st_dev;
 486     pr_stat->st_ino = stat->st_ino;
 487     pr_stat->st_nlink = stat->st_nlink;
 488     pr_stat->st_mode = stat->st_mode;
 489     pr_stat->st_uid = stat->st_uid;
 490     pr_stat->st_gid = stat->st_gid;
 491     pr_stat->st_rdev = stat->st_rdev;
 492     pr_stat->st_size = stat->st_size;
 493     pr_stat->st_blksize = stat->st_blksize;
 494     pr_stat->st_blocks = stat->st_blocks;
 495     pr_stat->st_atim_sec = stat->st_atim.tv_sec;
 496     pr_stat->st_atim_nsec = stat->st_atim.tv_nsec;
 497     pr_stat->st_mtim_sec = stat->st_mtim.tv_sec;
 498     pr_stat->st_mtim_nsec = stat->st_mtim.tv_nsec;
 499     pr_stat->st_ctim_sec = stat->st_ctim.tv_sec;
 500     pr_stat->st_ctim_nsec = stat->st_ctim.tv_nsec;
 501 }
 502
 503 static void statfs_to_prstatfs(ProxyStatFS *pr_stfs, struct statfs *stfs)
 504 {
 505     memset(pr_stfs, 0, sizeof(*pr_stfs));
 506     pr_stfs->f_type = stfs->f_type;
 507     pr_stfs->f_bsize = stfs->f_bsize;
 508     pr_stfs->f_blocks = stfs->f_blocks;
 509     pr_stfs->f_bfree = stfs->f_bfree;
 510     pr_stfs->f_bavail = stfs->f_bavail;
 511     pr_stfs->f_files = stfs->f_files;
 512     pr_stfs->f_ffree = stfs->f_ffree;
 513     pr_stfs->f_fsid[0] = stfs->f_fsid.__val[0];
 514     pr_stfs->f_fsid[1] = stfs->f_fsid.__val[1];
 515     pr_stfs->f_namelen = stfs->f_namelen;
 516     pr_stfs->f_frsize = stfs->f_frsize;
 517 }
 518
 519 /*
 520  * Gets stat/statfs information and packs in out_iovec structure
 521  * on success returns number of bytes packed in out_iovec structure
 522  * otherwise returns -errno
 523  */
 524 static int do_stat(int type, struct iovec *iovec, struct iovec *out_iovec)
 525 {
 526     int retval;
 527     V9fsString path;
 528     ProxyStat pr_stat;
 529     ProxyStatFS pr_stfs;
 530     struct stat st_buf;
 531     struct statfs stfs_buf;
 532
 533     v9fs_string_init(&path);
 534     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
 535     if (retval < 0) {
 536         return retval;
 537     }
 538
 539     switch (type) {
 540     case T_LSTAT:
 541         retval = lstat(path.data, &st_buf);
 542         if (retval < 0) {
 543             retval = -errno;
 544         } else {
 545             stat_to_prstat(&pr_stat, &st_buf);
 546             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
 547                                    "qqqdddqqqqqqqqqq", pr_stat.st_dev,
 548                                    pr_stat.st_ino, pr_stat.st_nlink,
 549                                    pr_stat.st_mode, pr_stat.st_uid,
 550                                    pr_stat.st_gid, pr_stat.st_rdev,
 551                                    pr_stat.st_size, pr_stat.st_blksize,
 552                                    pr_stat.st_blocks,
 553                                    pr_stat.st_atim_sec, pr_stat.st_atim_nsec,
 554                                    pr_stat.st_mtim_sec, pr_stat.st_mtim_nsec,
 555                                    pr_stat.st_ctim_sec, pr_stat.st_ctim_nsec);
 556         }
 557         break;
 558     case T_STATFS:
 559         retval = statfs(path.data, &stfs_buf);
 560         if (retval < 0) {
 561             retval = -errno;
 562         } else {
 563             statfs_to_prstatfs(&pr_stfs, &stfs_buf);
 564             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
 565                                    "qqqqqqqqqqq", pr_stfs.f_type,
 566                                    pr_stfs.f_bsize, pr_stfs.f_blocks,
 567                                    pr_stfs.f_bfree, pr_stfs.f_bavail,
 568                                    pr_stfs.f_files, pr_stfs.f_ffree,
 569                                    pr_stfs.f_fsid[0], pr_stfs.f_fsid[1],
 570                                    pr_stfs.f_namelen, pr_stfs.f_frsize);
 571         }
 572         break;
 573     }
 574     v9fs_string_free(&path);
 575     return retval;
 576 }
 577
 578 static int do_readlink(struct iovec *iovec, struct iovec *out_iovec)
 579 {
 580     char *buffer;
 581     int size, retval;
 582     V9fsString target, path;
 583
 584     v9fs_string_init(&path);
 585     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &size);
 586     if (retval < 0) {
 587         v9fs_string_free(&path);
 588         return retval;
 589     }
 590     buffer = g_malloc(size);
 591     v9fs_string_init(&target);
 592     retval = readlink(path.data, buffer, size - 1);
 593     if (retval > 0) {
 594         buffer[retval] = '\0';
 595         v9fs_string_sprintf(&target, "%s", buffer);
 596         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &target);
 597     } else {
 598         retval = -errno;
 599     }
 600     g_free(buffer);
 601     v9fs_string_free(&target);
 602     v9fs_string_free(&path);
 603     return retval;
 604 }
 605
 606 /*
 607  * create other filesystem objects and send 0 on success
 608  * return -errno on error
 609  */
 610 static int do_create_others(int type, struct iovec *iovec)
 611 {
 612     dev_t rdev;
 613     int retval = 0;
 614     int offset = PROXY_HDR_SZ;
 615     V9fsString oldpath, path;
 616     int mode, uid, gid, cur_uid, cur_gid;
 617
 618     v9fs_string_init(&path);
 619     v9fs_string_init(&oldpath);
 620
 621     retval = proxy_unmarshal(iovec, offset, "dd", &uid, &gid);
 622     if (retval < 0) {
 623         return retval;
 624     }
 625     offset += retval;
 626     retval = setugid(uid, gid, &cur_uid, &cur_gid);
 627     if (retval < 0) {
 628         goto unmarshal_err_out;
 629     }
 630     switch (type) {
 631     case T_MKNOD:
 632         retval = proxy_unmarshal(iovec, offset, "sdq", &path, &mode, &rdev);
 633         if (retval < 0) {
 634             goto err_out;
 635         }
 636         retval = mknod(path.data, mode, rdev);
 637         break;
 638     case T_MKDIR:
 639         retval = proxy_unmarshal(iovec, offset, "sd", &path, &mode);
 640         if (retval < 0) {
 641             goto err_out;
 642         }
 643         retval = mkdir(path.data, mode);
 644         break;
 645     case T_SYMLINK:
 646         retval = proxy_unmarshal(iovec, offset, "ss", &oldpath, &path);
 647         if (retval < 0) {
 648             goto err_out;
 649         }
 650         retval = symlink(oldpath.data, path.data);
 651         break;
 652     }
 653     if (retval < 0) {
 654         retval = -errno;
 655     }
 656
 657 err_out:
 658     resetugid(cur_uid, cur_gid);
 659 unmarshal_err_out:
 660     v9fs_string_free(&path);
 661     v9fs_string_free(&oldpath);
 662     return retval;
 663 }
 664
 665 /*
 666  * create a file and send fd on success
 667  * return -errno on error
 668  */
 669 static int do_create(struct iovec *iovec)
 670 {
 671     int ret;
 672     V9fsString path;
 673     int flags, mode, uid, gid, cur_uid, cur_gid;
 674
 675     v9fs_string_init(&path);
 676     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sdddd",
 677                           &path, &flags, &mode, &uid, &gid);
 678     if (ret < 0) {
 679         goto unmarshal_err_out;
 680     }
 681     ret = setugid(uid, gid, &cur_uid, &cur_gid);
 682     if (ret < 0) {
 683         goto unmarshal_err_out;
 684     }
 685     ret = open(path.data, flags, mode);
 686     if (ret < 0) {
 687         ret = -errno;
 688     }
 689
 690     resetugid(cur_uid, cur_gid);
 691 unmarshal_err_out:
 692     v9fs_string_free(&path);
 693     return ret;
 694 }
 695
 696 /*
 697  * open a file and send fd on success
 698  * return -errno on error
 699  */
 700 static int do_open(struct iovec *iovec)
 701 {
 702     int flags, ret;
 703     V9fsString path;
 704
 705     v9fs_string_init(&path);
 706     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &flags);
 707     if (ret < 0) {
 708         goto err_out;
 709     }
 710     ret = open(path.data, flags);
 711     if (ret < 0) {
 712         ret = -errno;
 713     }
 714 err_out:
 715     v9fs_string_free(&path);
 716     return ret;
 717 }
 718
 719 /* create unix domain socket and return the descriptor */
 720 static int proxy_socket(const char *path, uid_t uid, gid_t gid)
 721 {
 722     int sock, client;
 723     struct sockaddr_un proxy, qemu;
 724     socklen_t size;
 725
 726     /* requested socket already exists, refuse to start */
 727     if (!access(path, F_OK)) {
 728         do_log(LOG_CRIT, "socket already exists\n");
 729         return -1;
 730     }
 731
 732     if (strlen(path) >= sizeof(proxy.sun_path)) {
 733         do_log(LOG_CRIT, "UNIX domain socket path exceeds %zu characters\n",
 734                sizeof(proxy.sun_path));
 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         goto error;
 753     }
 754     if (chown(proxy.sun_path, uid, gid) < 0) {
 755         do_perror("chown");
 756         goto error;
 757     }
 758     if (listen(sock, 1) < 0) {
 759         do_perror("listen");
 760         goto error;
 761     }
 762
 763     size = sizeof(qemu);
 764     client = accept(sock, (struct sockaddr *)&qemu, &size);
 765     if (client < 0) {
 766         do_perror("accept");
 767         goto error;
 768     }
 769     close(sock);
 770     return client;
 771
 772 error:
 773     close(sock);
 774     return -1;
 775 }
 776
 777 static void usage(void)
 778 {
 779     fprintf(stderr, "usage: %s\n"
 780             " -p|--path <path> 9p path to export\n"
 781             " {-f|--fd <socket-descriptor>} socket file descriptor to be used\n"
 782             " {-s|--socket <socketname> socket file used for communication\n"
 783             " \t-u|--uid <uid> -g|--gid <gid>} - uid:gid combination to give "
 784             " access to this socket\n"
 785             " \tNote: -s & -f can not be used together\n"
 786             " [-n|--nodaemon] Run as a normal program\n",
 787             prog_name);
 788 }
 789
 790 static int process_reply(int sock, int type,
 791                          struct iovec *out_iovec, int retval)
 792 {
 793     switch (type) {
 794     case T_OPEN:
 795     case T_CREATE:
 796         if (send_fd(sock, retval) < 0) {
 797             return -1;
 798         }
 799         break;
 800     case T_MKNOD:
 801     case T_MKDIR:
 802     case T_SYMLINK:
 803     case T_LINK:
 804     case T_CHMOD:
 805     case T_CHOWN:
 806     case T_TRUNCATE:
 807     case T_UTIME:
 808     case T_RENAME:
 809     case T_REMOVE:
 810     case T_LSETXATTR:
 811     case T_LREMOVEXATTR:
 812         if (send_status(sock, out_iovec, retval) < 0) {
 813             return -1;
 814         }
 815         break;
 816     case T_LSTAT:
 817     case T_STATFS:
 818     case T_READLINK:
 819     case T_LGETXATTR:
 820     case T_LLISTXATTR:
 821     case T_GETVERSION:
 822         if (send_response(sock, out_iovec, retval) < 0) {
 823             return -1;
 824         }
 825         break;
 826     default:
 827         return -1;
 828         break;
 829     }
 830     return 0;
 831 }
 832
 833 static int process_requests(int sock)
 834 {
 835     int flags;
 836     int size = 0;
 837     int retval = 0;
 838     uint64_t offset;
 839     ProxyHeader header;
 840     int mode, uid, gid;
 841     V9fsString name, value;
 842     struct timespec spec[2];
 843     V9fsString oldpath, path;
 844     struct iovec in_iovec, out_iovec;
 845
 846     in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
 847     in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
 848     out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
 849     out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
 850
 851     while (1) {
 852         /*
 853          * initialize the header type, so that we send
 854          * response to proper request type.
 855          */
 856         header.type = 0;
 857         retval = read_request(sock, &in_iovec, &header);
 858         if (retval < 0) {
 859             goto err_out;
 860         }
 861
 862         switch (header.type) {
 863         case T_OPEN:
 864             retval = do_open(&in_iovec);
 865             break;
 866         case T_CREATE:
 867             retval = do_create(&in_iovec);
 868             break;
 869         case T_MKNOD:
 870         case T_MKDIR:
 871         case T_SYMLINK:
 872             retval = do_create_others(header.type, &in_iovec);
 873             break;
 874         case T_LINK:
 875             v9fs_string_init(&path);
 876             v9fs_string_init(&oldpath);
 877             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 878                                      "ss", &oldpath, &path);
 879             if (retval > 0) {
 880                 retval = link(oldpath.data, path.data);
 881                 if (retval < 0) {
 882                     retval = -errno;
 883                 }
 884             }
 885             v9fs_string_free(&oldpath);
 886             v9fs_string_free(&path);
 887             break;
 888         case T_LSTAT:
 889         case T_STATFS:
 890             retval = do_stat(header.type, &in_iovec, &out_iovec);
 891             break;
 892         case T_READLINK:
 893             retval = do_readlink(&in_iovec, &out_iovec);
 894             break;
 895         case T_CHMOD:
 896             v9fs_string_init(&path);
 897             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 898                                      "sd", &path, &mode);
 899             if (retval > 0) {
 900                 retval = chmod(path.data, mode);
 901                 if (retval < 0) {
 902                     retval = -errno;
 903                 }
 904             }
 905             v9fs_string_free(&path);
 906             break;
 907         case T_CHOWN:
 908             v9fs_string_init(&path);
 909             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sdd", &path,
 910                                      &uid, &gid);
 911             if (retval > 0) {
 912                 retval = lchown(path.data, uid, gid);
 913                 if (retval < 0) {
 914                     retval = -errno;
 915                 }
 916             }
 917             v9fs_string_free(&path);
 918             break;
 919         case T_TRUNCATE:
 920             v9fs_string_init(&path);
 921             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sq",
 922                                      &path, &offset);
 923             if (retval > 0) {
 924                 retval = truncate(path.data, offset);
 925                 if (retval < 0) {
 926                     retval = -errno;
 927                 }
 928             }
 929             v9fs_string_free(&path);
 930             break;
 931         case T_UTIME:
 932             v9fs_string_init(&path);
 933             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sqqqq", &path,
 934                                      &spec[0].tv_sec, &spec[0].tv_nsec,
 935                                      &spec[1].tv_sec, &spec[1].tv_nsec);
 936             if (retval > 0) {
 937                 retval = utimensat(AT_FDCWD, path.data, spec,
 938                                    AT_SYMLINK_NOFOLLOW);
 939                 if (retval < 0) {
 940                     retval = -errno;
 941                 }
 942             }
 943             v9fs_string_free(&path);
 944             break;
 945         case T_RENAME:
 946             v9fs_string_init(&path);
 947             v9fs_string_init(&oldpath);
 948             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
 949                                      "ss", &oldpath, &path);
 950             if (retval > 0) {
 951                 retval = rename(oldpath.data, path.data);
 952                 if (retval < 0) {
 953                     retval = -errno;
 954                 }
 955             }
 956             v9fs_string_free(&oldpath);
 957             v9fs_string_free(&path);
 958             break;
 959         case T_REMOVE:
 960             v9fs_string_init(&path);
 961             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "s", &path);
 962             if (retval > 0) {
 963                 retval = remove(path.data);
 964                 if (retval < 0) {
 965                     retval = -errno;
 966                 }
 967             }
 968             v9fs_string_free(&path);
 969             break;
 970         case T_LGETXATTR:
 971         case T_LLISTXATTR:
 972             retval = do_getxattr(header.type, &in_iovec, &out_iovec);
 973             break;
 974         case T_LSETXATTR:
 975             v9fs_string_init(&path);
 976             v9fs_string_init(&name);
 977             v9fs_string_init(&value);
 978             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sssdd", &path,
 979                                      &name, &value, &size, &flags);
 980             if (retval > 0) {
 981                 retval = lsetxattr(path.data,
 982                                    name.data, value.data, size, flags);
 983                 if (retval < 0) {
 984                     retval = -errno;
 985                 }
 986             }
 987             v9fs_string_free(&path);
 988             v9fs_string_free(&name);
 989             v9fs_string_free(&value);
 990             break;
 991         case T_LREMOVEXATTR:
 992             v9fs_string_init(&path);
 993             v9fs_string_init(&name);
 994             retval = proxy_unmarshal(&in_iovec,
 995                                      PROXY_HDR_SZ, "ss", &path, &name);
 996             if (retval > 0) {
 997                 retval = lremovexattr(path.data, name.data);
 998                 if (retval < 0) {
 999                     retval = -errno;
1000                 }
1001             }
1002             v9fs_string_free(&path);
1003             v9fs_string_free(&name);
1004             break;
1005         case T_GETVERSION:
1006             retval = do_getversion(&in_iovec, &out_iovec);
1007             break;
1008         default:
1009             goto err_out;
1010             break;
1011         }
1012
1013         if (process_reply(sock, header.type, &out_iovec, retval) < 0) {
1014             goto err_out;
1015         }
1016     }
1017 err_out:
1018     g_free(in_iovec.iov_base);
1019     g_free(out_iovec.iov_base);
1020     return -1;
1021 }
1022
1023 int main(int argc, char **argv)
1024 {
1025     int sock;
1026     uid_t own_u;
1027     gid_t own_g;
1028     char *rpath = NULL;
1029     char *sock_name = NULL;
1030     struct stat stbuf;
1031     int c, option_index;
1032 #ifdef FS_IOC_GETVERSION
1033     int retval;
1034     struct statfs st_fs;
1035 #endif
1036
1037     prog_name = g_path_get_basename(argv[0]);
1038
1039     is_daemon = true;
1040     sock = -1;
1041     own_u = own_g = -1;
1042     while (1) {
1043         option_index = 0;
1044         c = getopt_long(argc, argv, "p:nh?f:s:u:g:", helper_opts,
1045                         &option_index);
1046         if (c == -1) {
1047             break;
1048         }
1049         switch (c) {
1050         case 'p':
1051             rpath = g_strdup(optarg);
1052             break;
1053         case 'n':
1054             is_daemon = false;
1055             break;
1056         case 'f':
1057             sock = atoi(optarg);
1058             break;
1059         case 's':
1060             sock_name = g_strdup(optarg);
1061             break;
1062         case 'u':
1063             own_u = atoi(optarg);
1064             break;
1065         case 'g':
1066             own_g = atoi(optarg);
1067             break;
1068         case '?':
1069         case 'h':
1070         default:
1071             usage();
1072             exit(EXIT_FAILURE);
1073         }
1074     }
1075
1076     /* Parameter validation */
1077     if ((sock_name == NULL && sock == -1) || rpath == NULL) {
1078         fprintf(stderr, "socket, socket descriptor or path not specified\n");
1079         usage();
1080         return -1;
1081     }
1082
1083     if (sock_name && sock != -1) {
1084         fprintf(stderr, "both named socket and socket descriptor specified\n");
1085         usage();
1086         exit(EXIT_FAILURE);
1087     }
1088
1089     if (sock_name && (own_u == -1 || own_g == -1)) {
1090         fprintf(stderr, "owner uid:gid not specified, ");
1091         fprintf(stderr,
1092                 "owner uid:gid specifies who can access the socket file\n");
1093         usage();
1094         exit(EXIT_FAILURE);
1095     }
1096
1097     if (lstat(rpath, &stbuf) < 0) {
1098         fprintf(stderr, "invalid path \"%s\" specified, %s\n",
1099                 rpath, strerror(errno));
1100         exit(EXIT_FAILURE);
1101     }
1102
1103     if (!S_ISDIR(stbuf.st_mode)) {
1104         fprintf(stderr, "specified path \"%s\" is not directory\n", rpath);
1105         exit(EXIT_FAILURE);
1106     }
1107
1108     if (is_daemon) {
1109         if (daemon(0, 0) < 0) {
1110             fprintf(stderr, "daemon call failed\n");
1111             exit(EXIT_FAILURE);
1112         }
1113         openlog(PROGNAME, LOG_PID, LOG_DAEMON);
1114     }
1115
1116     do_log(LOG_INFO, "Started\n");
1117     if (sock_name) {
1118         sock = proxy_socket(sock_name, own_u, own_g);
1119         if (sock < 0) {
1120             goto error;
1121         }
1122     }
1123
1124     if (chroot(rpath) < 0) {
1125         do_perror("chroot");
1126         goto error;
1127     }
1128     if (chdir("/") < 0) {
1129         do_perror("chdir");
1130         goto error;
1131     }
1132
1133     get_version = false;
1134 #ifdef FS_IOC_GETVERSION
1135     /* check whether underlying FS support IOC_GETVERSION */
1136     retval = statfs("/", &st_fs);
1137     if (!retval) {
1138         switch (st_fs.f_type) {
1139         case EXT2_SUPER_MAGIC:
1140         case BTRFS_SUPER_MAGIC:
1141         case REISERFS_SUPER_MAGIC:
1142         case XFS_SUPER_MAGIC:
1143             get_version = true;
1144             break;
1145         }
1146     }
1147 #endif
1148
1149     umask(0);
1150     if (init_capabilities() < 0) {
1151         goto error;
1152     }
1153
1154     process_requests(sock);
1155 error:
1156     g_free(rpath);
1157     g_free(sock_name);
1158     do_log(LOG_INFO, "Done\n");
1159     closelog();
1160     return 0;
1161 }