source3/modules/vfs_glusterfs.c

   1 /*
   2    Unix SMB/CIFS implementation.
   3
   4    Wrap GlusterFS GFAPI calls in vfs functions.
   5
   6    Copyright (c) 2013 Anand Avati <avati@redhat.com>
   7
   8    This program is free software; you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation; either version 3 of the License, or
  11    (at your option) any later version.
  12
  13    This program is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License
  19    along with this program.  If not, see <http://www.gnu.org/licenses/>.
  20 */
  21
  22 /**
  23  * @file   vfs_glusterfs.c
  24  * @author Anand Avati <avati@redhat.com>
  25  * @date   May 2013
  26  * @brief  Samba VFS module for glusterfs
  27  *
  28  * @todo
  29  *   - sendfile/recvfile support
  30  *
  31  * A Samba VFS module for GlusterFS, based on Gluster's libgfapi.
  32  * This is a "bottom" vfs module (not something to be stacked on top of
  33  * another module), and translates (most) calls to the closest actions
  34  * available in libgfapi.
  35  *
  36  */
  37
  38 #include "includes.h"
  39 #include "smbd/smbd.h"
  40 #include <stdio.h>
  41 #include "api/glfs.h"
  42 #include "lib/util/dlinklist.h"
  43 #include "lib/util/tevent_unix.h"
  44 #include "lib/tevent/tevent_internal.h"
  45 #include "smbd/globals.h"
  46 #include "lib/sys_rw.h"
  47
  48 #define DEFAULT_VOLFILE_SERVER "localhost"
  49
  50 static int read_fd = -1;
  51 static int write_fd = -1;
  52 static struct tevent_fd *aio_read_event = NULL;
  53
  54 /**
  55  * Helper to convert struct stat to struct stat_ex.
  56  */
  57 static void smb_stat_ex_from_stat(struct stat_ex *dst, const struct stat *src)
  58 {
  59         ZERO_STRUCTP(dst);
  60
  61         dst->st_ex_dev = src->st_dev;
  62         dst->st_ex_ino = src->st_ino;
  63         dst->st_ex_mode = src->st_mode;
  64         dst->st_ex_nlink = src->st_nlink;
  65         dst->st_ex_uid = src->st_uid;
  66         dst->st_ex_gid = src->st_gid;
  67         dst->st_ex_rdev = src->st_rdev;
  68         dst->st_ex_size = src->st_size;
  69         dst->st_ex_atime.tv_sec = src->st_atime;
  70         dst->st_ex_mtime.tv_sec = src->st_mtime;
  71         dst->st_ex_ctime.tv_sec = src->st_ctime;
  72         dst->st_ex_btime.tv_sec = src->st_mtime;
  73         dst->st_ex_blksize = src->st_blksize;
  74         dst->st_ex_blocks = src->st_blocks;
  75 #ifdef STAT_HAVE_NSEC
  76         dst->st_ex_atime.tv_nsec = src->st_atime_nsec;
  77         dst->st_ex_mtime.tv_nsec = src->st_mtime_nsec;
  78         dst->st_ex_ctime.tv_nsec = src->st_ctime_nsec;
  79         dst->st_ex_btime.tv_nsec = src->st_mtime_nsec;
  80 #endif
  81 }
  82
  83 /* pre-opened glfs_t */
  84
  85 static struct glfs_preopened {
  86         char *volume;
  87         char *connectpath;
  88         glfs_t *fs;
  89         int ref;
  90         struct glfs_preopened *next, *prev;
  91 } *glfs_preopened;
  92
  93
  94 static int glfs_set_preopened(const char *volume, const char *connectpath, glfs_t *fs)
  95 {
  96         struct glfs_preopened *entry = NULL;
  97
  98         entry = talloc_zero(NULL, struct glfs_preopened);
  99         if (!entry) {
 100                 errno = ENOMEM;
 101                 return -1;
 102         }
 103
 104         entry->volume = talloc_strdup(entry, volume);
 105         if (!entry->volume) {
 106                 talloc_free(entry);
 107                 errno = ENOMEM;
 108                 return -1;
 109         }
 110
 111         entry->connectpath = talloc_strdup(entry, connectpath);
 112         if (entry->connectpath == NULL) {
 113                 talloc_free(entry);
 114                 errno = ENOMEM;
 115                 return -1;
 116         }
 117
 118         entry->fs = fs;
 119         entry->ref = 1;
 120
 121         DLIST_ADD(glfs_preopened, entry);
 122
 123         return 0;
 124 }
 125
 126 static glfs_t *glfs_find_preopened(const char *volume, const char *connectpath)
 127 {
 128         struct glfs_preopened *entry = NULL;
 129
 130         for (entry = glfs_preopened; entry; entry = entry->next) {
 131                 if (strcmp(entry->volume, volume) == 0 &&
 132                     strcmp(entry->connectpath, connectpath) == 0)
 133                 {
 134                         entry->ref++;
 135                         return entry->fs;
 136                 }
 137         }
 138
 139         return NULL;
 140 }
 141
 142 static void glfs_clear_preopened(glfs_t *fs)
 143 {
 144         struct glfs_preopened *entry = NULL;
 145
 146         for (entry = glfs_preopened; entry; entry = entry->next) {
 147                 if (entry->fs == fs) {
 148                         if (--entry->ref)
 149                                 return;
 150
 151                         DLIST_REMOVE(glfs_preopened, entry);
 152
 153                         glfs_fini(entry->fs);
 154                         talloc_free(entry);
 155                 }
 156         }
 157 }
 158
 159 /* Disk Operations */
 160
 161 static int vfs_gluster_connect(struct vfs_handle_struct *handle,
 162                                const char *service,
 163                                const char *user)
 164 {
 165         const char *volfile_server;
 166         const char *volume;
 167         char *logfile;
 168         int loglevel;
 169         glfs_t *fs = NULL;
 170         TALLOC_CTX *tmp_ctx;
 171         int ret = 0;
 172
 173         tmp_ctx = talloc_new(NULL);
 174         if (tmp_ctx == NULL) {
 175                 ret = -1;
 176                 goto done;
 177         }
 178         logfile = lp_parm_talloc_string(tmp_ctx, SNUM(handle->conn), "glusterfs",
 179                                        "logfile", NULL);
 180
 181         loglevel = lp_parm_int(SNUM(handle->conn), "glusterfs", "loglevel", -1);
 182
 183         volfile_server = lp_parm_const_string(SNUM(handle->conn), "glusterfs",
 184                                                "volfile_server", NULL);
 185         if (volfile_server == NULL) {
 186                 volfile_server = DEFAULT_VOLFILE_SERVER;
 187         }
 188
 189         volume = lp_parm_const_string(SNUM(handle->conn), "glusterfs", "volume",
 190                                       NULL);
 191         if (volume == NULL) {
 192                 volume = service;
 193         }
 194
 195         fs = glfs_find_preopened(volume, handle->conn->connectpath);
 196         if (fs) {
 197                 goto done;
 198         }
 199
 200         fs = glfs_new(volume);
 201         if (fs == NULL) {
 202                 ret = -1;
 203                 goto done;
 204         }
 205
 206         ret = glfs_set_volfile_server(fs, "tcp", volfile_server, 0);
 207         if (ret < 0) {
 208                 DEBUG(0, ("Failed to set volfile_server %s\n", volfile_server));
 209                 goto done;
 210         }
 211
 212         ret = glfs_set_xlator_option(fs, "*-md-cache", "cache-posix-acl",
 213                                      "true");
 214         if (ret < 0) {
 215                 DEBUG(0, ("%s: Failed to set xlator options\n", volume));
 216                 goto done;
 217         }
 218
 219
 220         ret = glfs_set_xlator_option(fs, "*-snapview-client",
 221                                      "snapdir-entry-path",
 222                                      handle->conn->connectpath);
 223         if (ret < 0) {
 224                 DEBUG(0, ("%s: Failed to set xlator option:"
 225                           " snapdir-entry-path\n", volume));
 226                 glfs_fini(fs);
 227                 return -1;
 228         }
 229
 230         ret = glfs_set_logging(fs, logfile, loglevel);
 231         if (ret < 0) {
 232                 DEBUG(0, ("%s: Failed to set logfile %s loglevel %d\n",
 233                           volume, logfile, loglevel));
 234                 goto done;
 235         }
 236
 237         ret = glfs_init(fs);
 238         if (ret < 0) {
 239                 DEBUG(0, ("%s: Failed to initialize volume (%s)\n",
 240                           volume, strerror(errno)));
 241                 goto done;
 242         }
 243
 244         ret = glfs_set_preopened(volume, handle->conn->connectpath, fs);
 245         if (ret < 0) {
 246                 DEBUG(0, ("%s: Failed to register volume (%s)\n",
 247                           volume, strerror(errno)));
 248                 goto done;
 249         }
 250 done:
 251         talloc_free(tmp_ctx);
 252         if (ret < 0) {
 253                 if (fs)
 254                         glfs_fini(fs);
 255                 return -1;
 256         } else {
 257                 DEBUG(0, ("%s: Initialized volume from server %s\n",
 258                          volume, volfile_server));
 259                 handle->data = fs;
 260                 return 0;
 261         }
 262 }
 263
 264 static void vfs_gluster_disconnect(struct vfs_handle_struct *handle)
 265 {
 266         glfs_t *fs = NULL;
 267
 268         fs = handle->data;
 269
 270         glfs_clear_preopened(fs);
 271 }
 272
 273 static uint64_t vfs_gluster_disk_free(struct vfs_handle_struct *handle,
 274                                       const char *path, bool small_query,
 275                                       uint64_t *bsize_p, uint64_t *dfree_p,
 276                                       uint64_t *dsize_p)
 277 {
 278         struct statvfs statvfs = { 0, };
 279         int ret;
 280
 281         ret = glfs_statvfs(handle->data, path, &statvfs);
 282         if (ret < 0) {
 283                 return -1;
 284         }
 285
 286         if (bsize_p != NULL) {
 287                 *bsize_p = (uint64_t)statvfs.f_bsize; /* Block size */
 288         }
 289         if (dfree_p != NULL) {
 290                 *dfree_p = (uint64_t)statvfs.f_bavail; /* Available Block units */
 291         }
 292         if (dsize_p != NULL) {
 293                 *dsize_p = (uint64_t)statvfs.f_blocks; /* Total Block units */
 294         }
 295
 296         return (uint64_t)statvfs.f_bavail;
 297 }
 298
 299 static int vfs_gluster_get_quota(struct vfs_handle_struct *handle,
 300                                  enum SMB_QUOTA_TYPE qtype, unid_t id,
 301                                  SMB_DISK_QUOTA *qt)
 302 {
 303         errno = ENOSYS;
 304         return -1;
 305 }
 306
 307 static int
 308 vfs_gluster_set_quota(struct vfs_handle_struct *handle,
 309                       enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
 310 {
 311         errno = ENOSYS;
 312         return -1;
 313 }
 314
 315 static int vfs_gluster_statvfs(struct vfs_handle_struct *handle,
 316                                const char *path,
 317                                struct vfs_statvfs_struct *vfs_statvfs)
 318 {
 319         struct statvfs statvfs = { 0, };
 320         int ret;
 321
 322         ret = glfs_statvfs(handle->data, path, &statvfs);
 323         if (ret < 0) {
 324                 DEBUG(0, ("glfs_statvfs(%s) failed: %s\n",
 325                           path, strerror(errno)));
 326                 return -1;
 327         }
 328
 329         ZERO_STRUCTP(vfs_statvfs);
 330
 331         vfs_statvfs->OptimalTransferSize = statvfs.f_frsize;
 332         vfs_statvfs->BlockSize = statvfs.f_bsize;
 333         vfs_statvfs->TotalBlocks = statvfs.f_blocks;
 334         vfs_statvfs->BlocksAvail = statvfs.f_bfree;
 335         vfs_statvfs->UserBlocksAvail = statvfs.f_bavail;
 336         vfs_statvfs->TotalFileNodes = statvfs.f_files;
 337         vfs_statvfs->FreeFileNodes = statvfs.f_ffree;
 338         vfs_statvfs->FsIdentifier = statvfs.f_fsid;
 339         vfs_statvfs->FsCapabilities =
 340             FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
 341
 342         return ret;
 343 }
 344
 345 static uint32_t vfs_gluster_fs_capabilities(struct vfs_handle_struct *handle,
 346                                             enum timestamp_set_resolution *p_ts_res)
 347 {
 348         uint32_t caps = FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
 349
 350 #ifdef STAT_HAVE_NSEC
 351         *p_ts_res = TIMESTAMP_SET_NT_OR_BETTER;
 352 #endif
 353
 354         return caps;
 355 }
 356
 357 static DIR *vfs_gluster_opendir(struct vfs_handle_struct *handle,
 358                                 const char *path, const char *mask,
 359                                 uint32 attributes)
 360 {
 361         glfs_fd_t *fd;
 362
 363         fd = glfs_opendir(handle->data, path);
 364         if (fd == NULL) {
 365                 DEBUG(0, ("glfs_opendir(%s) failed: %s\n",
 366                           path, strerror(errno)));
 367         }
 368
 369         return (DIR *) fd;
 370 }
 371
 372 static DIR *vfs_gluster_fdopendir(struct vfs_handle_struct *handle,
 373                                   files_struct *fsp, const char *mask,
 374                                   uint32 attributes)
 375 {
 376         return (DIR *) *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
 377 }
 378
 379 static int vfs_gluster_closedir(struct vfs_handle_struct *handle, DIR *dirp)
 380 {
 381         return glfs_closedir((void *)dirp);
 382 }
 383
 384 static struct dirent *vfs_gluster_readdir(struct vfs_handle_struct *handle,
 385                                           DIR *dirp, SMB_STRUCT_STAT *sbuf)
 386 {
 387         static char direntbuf[512];
 388         int ret;
 389         struct stat stat;
 390         struct dirent *dirent = 0;
 391
 392         if (sbuf != NULL) {
 393                 ret = glfs_readdirplus_r((void *)dirp, &stat, (void *)direntbuf,
 394                                          &dirent);
 395         } else {
 396                 ret = glfs_readdir_r((void *)dirp, (void *)direntbuf, &dirent);
 397         }
 398
 399         if ((ret < 0) || (dirent == NULL)) {
 400                 return NULL;
 401         }
 402
 403         if (sbuf != NULL) {
 404                 smb_stat_ex_from_stat(sbuf, &stat);
 405         }
 406
 407         return dirent;
 408 }
 409
 410 static long vfs_gluster_telldir(struct vfs_handle_struct *handle, DIR *dirp)
 411 {
 412         return glfs_telldir((void *)dirp);
 413 }
 414
 415 static void vfs_gluster_seekdir(struct vfs_handle_struct *handle, DIR *dirp,
 416                                 long offset)
 417 {
 418         glfs_seekdir((void *)dirp, offset);
 419 }
 420
 421 static void vfs_gluster_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
 422 {
 423         glfs_seekdir((void *)dirp, 0);
 424 }
 425
 426 static void vfs_gluster_init_search_op(struct vfs_handle_struct *handle,
 427                                        DIR *dirp)
 428 {
 429         return;
 430 }
 431
 432 static int vfs_gluster_mkdir(struct vfs_handle_struct *handle, const char *path,
 433                              mode_t mode)
 434 {
 435         return glfs_mkdir(handle->data, path, mode);
 436 }
 437
 438 static int vfs_gluster_rmdir(struct vfs_handle_struct *handle, const char *path)
 439 {
 440         return glfs_rmdir(handle->data, path);
 441 }
 442
 443 static int vfs_gluster_open(struct vfs_handle_struct *handle,
 444                             struct smb_filename *smb_fname, files_struct *fsp,
 445                             int flags, mode_t mode)
 446 {
 447         glfs_fd_t *glfd;
 448         glfs_fd_t **p_tmp;
 449
 450         if (flags & O_DIRECTORY) {
 451                 glfd = glfs_opendir(handle->data, smb_fname->base_name);
 452         } else if (flags & O_CREAT) {
 453                 glfd = glfs_creat(handle->data, smb_fname->base_name, flags,
 454                                   mode);
 455         } else {
 456                 glfd = glfs_open(handle->data, smb_fname->base_name, flags);
 457         }
 458
 459         if (glfd == NULL) {
 460                 return -1;
 461         }
 462         p_tmp = (glfs_fd_t **)VFS_ADD_FSP_EXTENSION(handle, fsp,
 463                                                           glfs_fd_t *, NULL);
 464         *p_tmp = glfd;
 465         /* An arbitrary value for error reporting, so you know its us. */
 466         return 13371337;
 467 }
 468
 469 static int vfs_gluster_close(struct vfs_handle_struct *handle,
 470                              files_struct *fsp)
 471 {
 472         glfs_fd_t *glfd;
 473         glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
 474         VFS_REMOVE_FSP_EXTENSION(handle, fsp);
 475         return glfs_close(glfd);
 476 }
 477
 478 static ssize_t vfs_gluster_read(struct vfs_handle_struct *handle,
 479                                 files_struct *fsp, void *data, size_t n)
 480 {
 481         return glfs_read(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
 482 }
 483
 484 static ssize_t vfs_gluster_pread(struct vfs_handle_struct *handle,
 485                                  files_struct *fsp, void *data, size_t n,
 486                                  off_t offset)
 487 {
 488         return glfs_pread(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
 489 }
 490
 491 struct glusterfs_aio_state {
 492         ssize_t ret;
 493         int err;
 494 };
 495
 496 /*
 497  * This function is the callback that will be called on glusterfs
 498  * threads once the async IO submitted is complete. To notify
 499  * Samba of the completion we use a pipe pased queue.
 500  */
 501 static void aio_glusterfs_done(glfs_fd_t *fd, ssize_t ret, void *data)
 502 {
 503         struct tevent_req *req = NULL;
 504         struct glusterfs_aio_state *state = NULL;
 505         int sts = 0;
 506
 507         req = talloc_get_type_abort(data, struct tevent_req);
 508         state = tevent_req_data(req, struct glusterfs_aio_state);
 509
 510         if (ret < 0) {
 511                 state->ret = -1;
 512                 state->err = errno;
 513         } else {
 514                 state->ret = ret;
 515                 state->err = 0;
 516         }
 517
 518         /*
 519          * Write the pointer to each req that needs to be completed
 520          * by calling tevent_req_done(). tevent_req_done() cannot
 521          * be called here, as it is not designed to be executed
 522          * in the multithread environment, tevent_req_done() must be
 523          * executed from the smbd main thread.
 524          */
 525
 526         sts = sys_write (write_fd, &req, sizeof(struct tevent_req *));
 527         if (sts < 0) {
 528                 DEBUG(0,("\nWrite to pipe failed (%s)", strerror(errno)));
 529         }
 530
 531         return;
 532 }
 533
 534 /*
 535  * Read each req off the pipe and process it.
 536  */
 537 static void aio_tevent_fd_done(struct tevent_context *event_ctx,
 538                                 struct tevent_fd *fde,
 539                                 uint16 flags, void *data)
 540 {
 541         struct tevent_req *req = NULL;
 542         int sts = 0;
 543
 544         sts = sys_read(read_fd, &req, sizeof(struct tevent_req *));
 545         if (sts < 0) {
 546                 DEBUG(0,("\nRead from pipe failed (%s)", strerror(errno)));
 547         }
 548
 549         if (req) {
 550                 tevent_req_done(req);
 551         }
 552         return;
 553 }
 554
 555 static bool init_gluster_aio(struct vfs_handle_struct *handle)
 556 {
 557         int fds[2];
 558         int ret = -1;
 559
 560         if (read_fd != -1) {
 561                 /*
 562                  * Already initialized.
 563                  */
 564                 return true;
 565         }
 566
 567         ret = pipe(fds);
 568         if (ret == -1) {
 569                 goto fail;
 570         }
 571
 572         read_fd = fds[0];
 573         write_fd = fds[1];
 574
 575         aio_read_event = tevent_add_fd(handle->conn->sconn->ev_ctx,
 576                                         NULL,
 577                                         read_fd,
 578                                         TEVENT_FD_READ,
 579                                         aio_tevent_fd_done,
 580                                         NULL);
 581         if (aio_read_event == NULL) {
 582                 goto fail;
 583         }
 584
 585         return true;
 586 fail:
 587         TALLOC_FREE(aio_read_event);
 588         if (read_fd != -1) {
 589                 close(read_fd);
 590                 close(write_fd);
 591                 read_fd = -1;
 592                 write_fd = -1;
 593         }
 594         return false;
 595 }
 596
 597 static struct tevent_req *vfs_gluster_pread_send(struct vfs_handle_struct
 598                                                  *handle, TALLOC_CTX *mem_ctx,
 599                                                  struct tevent_context *ev,
 600                                                  files_struct *fsp, void *data,
 601                                                  size_t n, off_t offset)
 602 {
 603         struct tevent_req *req = NULL;
 604         struct glusterfs_aio_state *state = NULL;
 605         int ret = 0;
 606
 607         req = tevent_req_create(mem_ctx, &state, struct glusterfs_aio_state);
 608         if (req == NULL) {
 609                 return NULL;
 610         }
 611
 612         if (!init_gluster_aio(handle)) {
 613                 tevent_req_error(req, EIO);
 614                 return tevent_req_post(req, ev);
 615         }
 616         ret = glfs_pread_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 617                                 fsp), data, n, offset, 0, aio_glusterfs_done,
 618                                 req);
 619         if (ret < 0) {
 620                 tevent_req_error(req, -ret);
 621                 return tevent_req_post(req, ev);
 622         }
 623
 624         return req;
 625 }
 626
 627 static ssize_t vfs_gluster_write(struct vfs_handle_struct *handle,
 628                                  files_struct *fsp, const void *data, size_t n)
 629 {
 630         return glfs_write(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
 631 }
 632
 633 static ssize_t vfs_gluster_pwrite(struct vfs_handle_struct *handle,
 634                                   files_struct *fsp, const void *data,
 635                                   size_t n, off_t offset)
 636 {
 637         return glfs_pwrite(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
 638 }
 639
 640 static struct tevent_req *vfs_gluster_pwrite_send(struct vfs_handle_struct
 641                                                   *handle, TALLOC_CTX *mem_ctx,
 642                                                   struct tevent_context *ev,
 643                                                   files_struct *fsp,
 644                                                   const void *data, size_t n,
 645                                                   off_t offset)
 646 {
 647         struct tevent_req *req = NULL;
 648         struct glusterfs_aio_state *state = NULL;
 649         int ret = 0;
 650
 651         req = tevent_req_create(mem_ctx, &state, struct glusterfs_aio_state);
 652         if (req == NULL) {
 653                 return NULL;
 654         }
 655         if (!init_gluster_aio(handle)) {
 656                 tevent_req_error(req, EIO);
 657                 return tevent_req_post(req, ev);
 658         }
 659         ret = glfs_pwrite_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 660                                 fsp), data, n, offset, 0, aio_glusterfs_done,
 661                                 req);
 662         if (ret < 0) {
 663                 tevent_req_error(req, -ret);
 664                 return tevent_req_post(req, ev);
 665         }
 666         return req;
 667 }
 668
 669 static ssize_t vfs_gluster_recv(struct tevent_req *req, int *err)
 670 {
 671         struct glusterfs_aio_state *state = NULL;
 672
 673         state = tevent_req_data(req, struct glusterfs_aio_state);
 674         if (state == NULL) {
 675                 return -1;
 676         }
 677
 678         if (tevent_req_is_unix_error(req, err)) {
 679                 return -1;
 680         }
 681         if (state->ret == -1) {
 682                 *err = state->err;
 683         }
 684         return state->ret;
 685 }
 686
 687 static off_t vfs_gluster_lseek(struct vfs_handle_struct *handle,
 688                                files_struct *fsp, off_t offset, int whence)
 689 {
 690         return glfs_lseek(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset, whence);
 691 }
 692
 693 static ssize_t vfs_gluster_sendfile(struct vfs_handle_struct *handle, int tofd,
 694                                     files_struct *fromfsp,
 695                                     const DATA_BLOB *hdr,
 696                                     off_t offset, size_t n)
 697 {
 698         errno = ENOTSUP;
 699         return -1;
 700 }
 701
 702 static ssize_t vfs_gluster_recvfile(struct vfs_handle_struct *handle,
 703                                     int fromfd, files_struct *tofsp,
 704                                     off_t offset, size_t n)
 705 {
 706         errno = ENOTSUP;
 707         return -1;
 708 }
 709
 710 static int vfs_gluster_rename(struct vfs_handle_struct *handle,
 711                               const struct smb_filename *smb_fname_src,
 712                               const struct smb_filename *smb_fname_dst)
 713 {
 714         return glfs_rename(handle->data, smb_fname_src->base_name,
 715                            smb_fname_dst->base_name);
 716 }
 717
 718 static int vfs_gluster_fsync(struct vfs_handle_struct *handle,
 719                              files_struct *fsp)
 720 {
 721         return glfs_fsync(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp));
 722 }
 723
 724 static struct tevent_req *vfs_gluster_fsync_send(struct vfs_handle_struct
 725                                                  *handle, TALLOC_CTX *mem_ctx,
 726                                                  struct tevent_context *ev,
 727                                                  files_struct *fsp)
 728 {
 729         struct tevent_req *req = NULL;
 730         struct glusterfs_aio_state *state = NULL;
 731         int ret = 0;
 732
 733         req = tevent_req_create(mem_ctx, &state, struct glusterfs_aio_state);
 734         if (req == NULL) {
 735                 return NULL;
 736         }
 737         if (!init_gluster_aio(handle)) {
 738                 tevent_req_error(req, EIO);
 739                 return tevent_req_post(req, ev);
 740         }
 741         ret = glfs_fsync_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 742                                 fsp), aio_glusterfs_done, req);
 743         if (ret < 0) {
 744                 tevent_req_error(req, -ret);
 745                 return tevent_req_post(req, ev);
 746         }
 747         return req;
 748 }
 749
 750 static int vfs_gluster_fsync_recv(struct tevent_req *req, int *err)
 751 {
 752         /*
 753          * Use implicit conversion ssize_t->int
 754          */
 755         return vfs_gluster_recv(req, err);
 756 }
 757
 758 static int vfs_gluster_stat(struct vfs_handle_struct *handle,
 759                             struct smb_filename *smb_fname)
 760 {
 761         struct stat st;
 762         int ret;
 763
 764         ret = glfs_stat(handle->data, smb_fname->base_name, &st);
 765         if (ret == 0) {
 766                 smb_stat_ex_from_stat(&smb_fname->st, &st);
 767         }
 768         if (ret < 0 && errno != ENOENT) {
 769                 DEBUG(0, ("glfs_stat(%s) failed: %s\n",
 770                           smb_fname->base_name, strerror(errno)));
 771         }
 772         return ret;
 773 }
 774
 775 static int vfs_gluster_fstat(struct vfs_handle_struct *handle,
 776                              files_struct *fsp, SMB_STRUCT_STAT *sbuf)
 777 {
 778         struct stat st;
 779         int ret;
 780
 781         ret = glfs_fstat(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), &st);
 782         if (ret == 0) {
 783                 smb_stat_ex_from_stat(sbuf, &st);
 784         }
 785         if (ret < 0) {
 786                 DEBUG(0, ("glfs_fstat(%d) failed: %s\n",
 787                           fsp->fh->fd, strerror(errno)));
 788         }
 789         return ret;
 790 }
 791
 792 static int vfs_gluster_lstat(struct vfs_handle_struct *handle,
 793                              struct smb_filename *smb_fname)
 794 {
 795         struct stat st;
 796         int ret;
 797
 798         ret = glfs_lstat(handle->data, smb_fname->base_name, &st);
 799         if (ret == 0) {
 800                 smb_stat_ex_from_stat(&smb_fname->st, &st);
 801         }
 802         if (ret < 0 && errno != ENOENT) {
 803                 DEBUG(0, ("glfs_lstat(%s) failed: %s\n",
 804                           smb_fname->base_name, strerror(errno)));
 805         }
 806         return ret;
 807 }
 808
 809 static uint64_t vfs_gluster_get_alloc_size(struct vfs_handle_struct *handle,
 810                                            files_struct *fsp,
 811                                            const SMB_STRUCT_STAT *sbuf)
 812 {
 813         return sbuf->st_ex_blocks * 512;
 814 }
 815
 816 static int vfs_gluster_unlink(struct vfs_handle_struct *handle,
 817                               const struct smb_filename *smb_fname)
 818 {
 819         return glfs_unlink(handle->data, smb_fname->base_name);
 820 }
 821
 822 static int vfs_gluster_chmod(struct vfs_handle_struct *handle,
 823                              const char *path, mode_t mode)
 824 {
 825         return glfs_chmod(handle->data, path, mode);
 826 }
 827
 828 static int vfs_gluster_fchmod(struct vfs_handle_struct *handle,
 829                               files_struct *fsp, mode_t mode)
 830 {
 831         return glfs_fchmod(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), mode);
 832 }
 833
 834 static int vfs_gluster_chown(struct vfs_handle_struct *handle,
 835                              const char *path, uid_t uid, gid_t gid)
 836 {
 837         return glfs_chown(handle->data, path, uid, gid);
 838 }
 839
 840 static int vfs_gluster_fchown(struct vfs_handle_struct *handle,
 841                               files_struct *fsp, uid_t uid, gid_t gid)
 842 {
 843         return glfs_fchown(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), uid, gid);
 844 }
 845
 846 static int vfs_gluster_lchown(struct vfs_handle_struct *handle,
 847                               const char *path, uid_t uid, gid_t gid)
 848 {
 849         return glfs_lchown(handle->data, path, uid, gid);
 850 }
 851
 852 static int vfs_gluster_chdir(struct vfs_handle_struct *handle, const char *path)
 853 {
 854         return glfs_chdir(handle->data, path);
 855 }
 856
 857 static char *vfs_gluster_getwd(struct vfs_handle_struct *handle)
 858 {
 859         char *cwd;
 860         char *ret;
 861
 862         cwd = SMB_CALLOC_ARRAY(char, PATH_MAX);
 863         if (cwd == NULL) {
 864                 return NULL;
 865         }
 866
 867         ret = glfs_getcwd(handle->data, cwd, PATH_MAX - 1);
 868         if (ret == 0) {
 869                 free(cwd);
 870         }
 871         return ret;
 872 }
 873
 874 static int vfs_gluster_ntimes(struct vfs_handle_struct *handle,
 875                               const struct smb_filename *smb_fname,
 876                               struct smb_file_time *ft)
 877 {
 878         struct timespec times[2];
 879
 880         if (null_timespec(ft->atime)) {
 881                 times[0].tv_sec = smb_fname->st.st_ex_atime.tv_sec;
 882                 times[0].tv_nsec = smb_fname->st.st_ex_atime.tv_nsec;
 883         } else {
 884                 times[0].tv_sec = ft->atime.tv_sec;
 885                 times[0].tv_nsec = ft->atime.tv_nsec;
 886         }
 887
 888         if (null_timespec(ft->mtime)) {
 889                 times[1].tv_sec = smb_fname->st.st_ex_mtime.tv_sec;
 890                 times[1].tv_nsec = smb_fname->st.st_ex_mtime.tv_nsec;
 891         } else {
 892                 times[1].tv_sec = ft->mtime.tv_sec;
 893                 times[1].tv_nsec = ft->mtime.tv_nsec;
 894         }
 895
 896         if ((timespec_compare(&times[0],
 897                               &smb_fname->st.st_ex_atime) == 0) &&
 898             (timespec_compare(&times[1],
 899                               &smb_fname->st.st_ex_mtime) == 0)) {
 900                 return 0;
 901         }
 902
 903         return glfs_utimens(handle->data, smb_fname->base_name, times);
 904 }
 905
 906 static int vfs_gluster_ftruncate(struct vfs_handle_struct *handle,
 907                                  files_struct *fsp, off_t offset)
 908 {
 909         return glfs_ftruncate(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset);
 910 }
 911
 912 static int vfs_gluster_fallocate(struct vfs_handle_struct *handle,
 913                                  struct files_struct *fsp,
 914                                  enum vfs_fallocate_mode mode,
 915                                  off_t offset, off_t len)
 916 {
 917         errno = ENOTSUP;
 918         return -1;
 919 }
 920
 921 static char *vfs_gluster_realpath(struct vfs_handle_struct *handle,
 922                                   const char *path)
 923 {
 924         return glfs_realpath(handle->data, path, 0);
 925 }
 926
 927 static bool vfs_gluster_lock(struct vfs_handle_struct *handle,
 928                              files_struct *fsp, int op, off_t offset,
 929                              off_t count, int type)
 930 {
 931         struct flock flock = { 0, };
 932         int ret;
 933
 934         flock.l_type = type;
 935         flock.l_whence = SEEK_SET;
 936         flock.l_start = offset;
 937         flock.l_len = count;
 938         flock.l_pid = 0;
 939
 940         ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), op, &flock);
 941
 942         if (op == F_GETLK) {
 943                 /* lock query, true if someone else has locked */
 944                 if ((ret != -1) &&
 945                     (flock.l_type != F_UNLCK) &&
 946                     (flock.l_pid != 0) && (flock.l_pid != getpid()))
 947                         return true;
 948                 /* not me */
 949                 return false;
 950         }
 951
 952         if (ret == -1) {
 953                 return false;
 954         }
 955
 956         return true;
 957 }
 958
 959 static int vfs_gluster_kernel_flock(struct vfs_handle_struct *handle,
 960                                     files_struct *fsp, uint32 share_mode,
 961                                     uint32_t access_mask)
 962 {
 963         errno = ENOSYS;
 964         return -1;
 965 }
 966
 967 static int vfs_gluster_linux_setlease(struct vfs_handle_struct *handle,
 968                                       files_struct *fsp, int leasetype)
 969 {
 970         errno = ENOSYS;
 971         return -1;
 972 }
 973
 974 static bool vfs_gluster_getlock(struct vfs_handle_struct *handle,
 975                                 files_struct *fsp, off_t *poffset,
 976                                 off_t *pcount, int *ptype, pid_t *ppid)
 977 {
 978         struct flock flock = { 0, };
 979         int ret;
 980
 981         flock.l_type = *ptype;
 982         flock.l_whence = SEEK_SET;
 983         flock.l_start = *poffset;
 984         flock.l_len = *pcount;
 985         flock.l_pid = 0;
 986
 987         ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), F_GETLK, &flock);
 988
 989         if (ret == -1) {
 990                 return false;
 991         }
 992
 993         *ptype = flock.l_type;
 994         *poffset = flock.l_start;
 995         *pcount = flock.l_len;
 996         *ppid = flock.l_pid;
 997
 998         return true;
 999 }
1000
1001 static int vfs_gluster_symlink(struct vfs_handle_struct *handle,
1002                                const char *oldpath, const char *newpath)
1003 {
1004         return glfs_symlink(handle->data, oldpath, newpath);
1005 }
1006
1007 static int vfs_gluster_readlink(struct vfs_handle_struct *handle,
1008                                 const char *path, char *buf, size_t bufsiz)
1009 {
1010         return glfs_readlink(handle->data, path, buf, bufsiz);
1011 }
1012
1013 static int vfs_gluster_link(struct vfs_handle_struct *handle,
1014                             const char *oldpath, const char *newpath)
1015 {
1016         return glfs_link(handle->data, oldpath, newpath);
1017 }
1018
1019 static int vfs_gluster_mknod(struct vfs_handle_struct *handle, const char *path,
1020                              mode_t mode, SMB_DEV_T dev)
1021 {
1022         return glfs_mknod(handle->data, path, mode, dev);
1023 }
1024
1025 static NTSTATUS vfs_gluster_notify_watch(struct vfs_handle_struct *handle,
1026                                          struct sys_notify_context *ctx,
1027                                          const char *path, uint32_t *filter,
1028                                          uint32_t *subdir_filter,
1029                                          void (*callback) (struct sys_notify_context *ctx,
1030                                                            void *private_data,
1031                                                            struct notify_event *ev),
1032                                          void *private_data, void *handle_p)
1033 {
1034         return NT_STATUS_NOT_IMPLEMENTED;
1035 }
1036
1037 static int vfs_gluster_chflags(struct vfs_handle_struct *handle,
1038                                const char *path, unsigned int flags)
1039 {
1040         errno = ENOSYS;
1041         return -1;
1042 }
1043
1044 static int vfs_gluster_get_real_filename(struct vfs_handle_struct *handle,
1045                                          const char *path, const char *name,
1046                                          TALLOC_CTX *mem_ctx, char **found_name)
1047 {
1048         int ret;
1049         char key_buf[NAME_MAX + 64];
1050         char val_buf[NAME_MAX + 1];
1051
1052         if (strlen(name) >= NAME_MAX) {
1053                 errno = ENAMETOOLONG;
1054                 return -1;
1055         }
1056
1057         snprintf(key_buf, NAME_MAX + 64,
1058                  "glusterfs.get_real_filename:%s", name);
1059
1060         ret = glfs_getxattr(handle->data, path, key_buf, val_buf, NAME_MAX + 1);
1061         if (ret == -1) {
1062                 if (errno == ENODATA) {
1063                         errno = EOPNOTSUPP;
1064                 }
1065                 return -1;
1066         }
1067
1068         *found_name = talloc_strdup(mem_ctx, val_buf);
1069         if (found_name[0] == NULL) {
1070                 errno = ENOMEM;
1071                 return -1;
1072         }
1073         return 0;
1074 }
1075
1076 static const char *vfs_gluster_connectpath(struct vfs_handle_struct *handle,
1077                                            const char *filename)
1078 {
1079         return handle->conn->connectpath;
1080 }
1081
1082 /* EA Operations */
1083
1084 static ssize_t vfs_gluster_getxattr(struct vfs_handle_struct *handle,
1085                                     const char *path, const char *name,
1086                                     void *value, size_t size)
1087 {
1088         return glfs_getxattr(handle->data, path, name, value, size);
1089 }
1090
1091 static ssize_t vfs_gluster_fgetxattr(struct vfs_handle_struct *handle,
1092                                      files_struct *fsp, const char *name,
1093                                      void *value, size_t size)
1094 {
1095         return glfs_fgetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size);
1096 }
1097
1098 static ssize_t vfs_gluster_listxattr(struct vfs_handle_struct *handle,
1099                                      const char *path, char *list, size_t size)
1100 {
1101         return glfs_listxattr(handle->data, path, list, size);
1102 }
1103
1104 static ssize_t vfs_gluster_flistxattr(struct vfs_handle_struct *handle,
1105                                       files_struct *fsp, char *list,
1106                                       size_t size)
1107 {
1108         return glfs_flistxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), list, size);
1109 }
1110
1111 static int vfs_gluster_removexattr(struct vfs_handle_struct *handle,
1112                                    const char *path, const char *name)
1113 {
1114         return glfs_removexattr(handle->data, path, name);
1115 }
1116
1117 static int vfs_gluster_fremovexattr(struct vfs_handle_struct *handle,
1118                                     files_struct *fsp, const char *name)
1119 {
1120         return glfs_fremovexattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name);
1121 }
1122
1123 static int vfs_gluster_setxattr(struct vfs_handle_struct *handle,
1124                                 const char *path, const char *name,
1125                                 const void *value, size_t size, int flags)
1126 {
1127         return glfs_setxattr(handle->data, path, name, value, size, flags);
1128 }
1129
1130 static int vfs_gluster_fsetxattr(struct vfs_handle_struct *handle,
1131                                  files_struct *fsp, const char *name,
1132                                  const void *value, size_t size, int flags)
1133 {
1134         return glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size,
1135                               flags);
1136 }
1137
1138 /* AIO Operations */
1139
1140 static bool vfs_gluster_aio_force(struct vfs_handle_struct *handle,
1141                                   files_struct *fsp)
1142 {
1143         return false;
1144 }
1145
1146 /* Offline Operations */
1147
1148 static bool vfs_gluster_is_offline(struct vfs_handle_struct *handle,
1149                                    const struct smb_filename *fname,
1150                                    SMB_STRUCT_STAT *sbuf)
1151 {
1152         return false;
1153 }
1154
1155 static int vfs_gluster_set_offline(struct vfs_handle_struct *handle,
1156                                    const struct smb_filename *fname)
1157 {
1158         errno = ENOTSUP;
1159         return -1;
1160 }
1161
1162 /*
1163   Gluster ACL Format:
1164
1165   Size = 4 (header) + N * 8 (entry)
1166
1167   Offset  Size    Field (Little Endian)
1168   -------------------------------------
1169   0-3     4-byte  Version
1170
1171   4-5     2-byte  Entry-1 tag
1172   6-7     2-byte  Entry-1 perm
1173   8-11    4-byte  Entry-1 id
1174
1175   12-13   2-byte  Entry-2 tag
1176   14-15   2-byte  Entry-2 perm
1177   16-19   4-byte  Entry-2 id
1178
1179   ...
1180
1181  */
1182
1183 /* header version */
1184 #define GLUSTER_ACL_VERSION 2
1185
1186 /* perm bits */
1187 #define GLUSTER_ACL_READ    0x04
1188 #define GLUSTER_ACL_WRITE   0x02
1189 #define GLUSTER_ACL_EXECUTE 0x01
1190
1191 /* tag values */
1192 #define GLUSTER_ACL_UNDEFINED_TAG  0x00
1193 #define GLUSTER_ACL_USER_OBJ       0x01
1194 #define GLUSTER_ACL_USER           0x02
1195 #define GLUSTER_ACL_GROUP_OBJ      0x04
1196 #define GLUSTER_ACL_GROUP          0x08
1197 #define GLUSTER_ACL_MASK           0x10
1198 #define GLUSTER_ACL_OTHER          0x20
1199
1200 #define GLUSTER_ACL_UNDEFINED_ID  (-1)
1201
1202 #define GLUSTER_ACL_HEADER_SIZE    4
1203 #define GLUSTER_ACL_ENTRY_SIZE     8
1204
1205 #define GLUSTER_ACL_SIZE(n)       (GLUSTER_ACL_HEADER_SIZE + (n * GLUSTER_ACL_ENTRY_SIZE))
1206
1207 static SMB_ACL_T mode_to_smb_acls(const struct stat *mode, TALLOC_CTX *mem_ctx)
1208 {
1209         struct smb_acl_t *result;
1210         int count;
1211
1212         count = 3;
1213         result = sys_acl_init(mem_ctx);
1214         if (!result) {
1215                 errno = ENOMEM;
1216                 return NULL;
1217         }
1218
1219         result->acl = talloc_array(result, struct smb_acl_entry, count);
1220         if (!result->acl) {
1221                 errno = ENOMEM;
1222                 talloc_free(result);
1223                 return NULL;
1224         }
1225
1226         result->count = count;
1227
1228         result->acl[0].a_type = SMB_ACL_USER_OBJ;
1229         result->acl[0].a_perm = (mode->st_mode & S_IRWXU) >> 6;;
1230
1231         result->acl[1].a_type = SMB_ACL_GROUP_OBJ;
1232         result->acl[1].a_perm = (mode->st_mode & S_IRWXG) >> 3;;
1233
1234         result->acl[2].a_type = SMB_ACL_OTHER;
1235         result->acl[2].a_perm = mode->st_mode & S_IRWXO;;
1236
1237         return result;
1238 }
1239
1240 static SMB_ACL_T gluster_to_smb_acl(const char *buf, size_t xattr_size,
1241                                     TALLOC_CTX *mem_ctx)
1242 {
1243         int count;
1244         size_t size;
1245         struct smb_acl_entry *smb_ace;
1246         struct smb_acl_t *result;
1247         int i;
1248         int offset;
1249         uint16_t tag;
1250         uint16_t perm;
1251         uint32_t id;
1252
1253         size = xattr_size;
1254
1255         if (size < GLUSTER_ACL_HEADER_SIZE) {
1256                 /* ACL should be at least as big as the header (4 bytes) */
1257                 errno = EINVAL;
1258                 return NULL;
1259         }
1260
1261         size -= GLUSTER_ACL_HEADER_SIZE; /* size of header = 4 bytes */
1262
1263         if (size % GLUSTER_ACL_ENTRY_SIZE) {
1264                 /* Size of entries must strictly be a multiple of
1265                    size of an ACE (8 bytes)
1266                 */
1267                 errno = EINVAL;
1268                 return NULL;
1269         }
1270
1271         count = size / GLUSTER_ACL_ENTRY_SIZE;
1272
1273         /* Version is the first 4 bytes of the ACL */
1274         if (IVAL(buf, 0) != GLUSTER_ACL_VERSION) {
1275                 DEBUG(0, ("Unknown gluster ACL version: %d\n",
1276                           IVAL(buf, 0)));
1277                 return NULL;
1278         }
1279         offset = GLUSTER_ACL_HEADER_SIZE;
1280
1281         result = sys_acl_init(mem_ctx);
1282         if (!result) {
1283                 errno = ENOMEM;
1284                 return NULL;
1285         }
1286
1287         result->acl = talloc_array(result, struct smb_acl_entry, count);
1288         if (!result->acl) {
1289                 errno = ENOMEM;
1290                 talloc_free(result);
1291                 return NULL;
1292         }
1293
1294         result->count = count;
1295
1296         smb_ace = result->acl;
1297
1298         for (i = 0; i < count; i++) {
1299                 /* TAG is the first 2 bytes of an entry */
1300                 tag = SVAL(buf, offset);
1301                 offset += 2;
1302
1303                 /* PERM is the next 2 bytes of an entry */
1304                 perm = SVAL(buf, offset);
1305                 offset += 2;
1306
1307                 /* ID is the last 4 bytes of an entry */
1308                 id = IVAL(buf, offset);
1309                 offset += 4;
1310
1311                 switch(tag) {
1312                 case GLUSTER_ACL_USER:
1313                         smb_ace->a_type = SMB_ACL_USER;
1314                         break;
1315                 case GLUSTER_ACL_USER_OBJ:
1316                         smb_ace->a_type = SMB_ACL_USER_OBJ;
1317                         break;
1318                 case GLUSTER_ACL_GROUP:
1319                         smb_ace->a_type = SMB_ACL_GROUP;
1320                         break;
1321                 case GLUSTER_ACL_GROUP_OBJ:
1322                         smb_ace->a_type = SMB_ACL_GROUP_OBJ;
1323                         break;
1324                 case GLUSTER_ACL_OTHER:
1325                         smb_ace->a_type = SMB_ACL_OTHER;
1326                         break;
1327                 case GLUSTER_ACL_MASK:
1328                         smb_ace->a_type = SMB_ACL_MASK;
1329                         break;
1330                 default:
1331                         DEBUG(0, ("unknown tag type %d\n", (unsigned int) tag));
1332                         return NULL;
1333                 }
1334
1335
1336                 switch(smb_ace->a_type) {
1337                 case SMB_ACL_USER:
1338                         smb_ace->info.user.uid = id;
1339                         break;
1340                 case SMB_ACL_GROUP:
1341                         smb_ace->info.group.gid = id;
1342                         break;
1343                 default:
1344                         break;
1345                 }
1346
1347                 smb_ace->a_perm = 0;
1348                 smb_ace->a_perm |=
1349                         ((perm & GLUSTER_ACL_READ) ? SMB_ACL_READ : 0);
1350                 smb_ace->a_perm |=
1351                         ((perm & GLUSTER_ACL_WRITE) ? SMB_ACL_WRITE : 0);
1352                 smb_ace->a_perm |=
1353                         ((perm & GLUSTER_ACL_EXECUTE) ? SMB_ACL_EXECUTE : 0);
1354
1355                 smb_ace++;
1356         }
1357
1358         return result;
1359 }
1360
1361
1362 static int gluster_ace_cmp(const void *left, const void *right)
1363 {
1364         int ret = 0;
1365         uint16_t tag_left, tag_right;
1366         uint32_t id_left, id_right;
1367
1368         /*
1369           Sorting precedence:
1370
1371            - Smaller TAG values must be earlier.
1372
1373            - Within same TAG, smaller identifiers must be earlier, E.g:
1374              UID 0 entry must be earlier than UID 200
1375              GID 17 entry must be earlier than GID 19
1376         */
1377
1378         /* TAG is the first element in the entry */
1379         tag_left = SVAL(left, 0);
1380         tag_right = SVAL(right, 0);
1381
1382         ret = (tag_left - tag_right);
1383         if (!ret) {
1384                 /* ID is the third element in the entry, after two short
1385                    integers (tag and perm), i.e at offset 4.
1386                 */
1387                 id_left = IVAL(left, 4);
1388                 id_right = IVAL(right, 4);
1389                 ret = id_left - id_right;
1390         }
1391
1392         return ret;
1393 }
1394
1395
1396 static ssize_t smb_to_gluster_acl(SMB_ACL_T theacl, char *buf, size_t len)
1397 {
1398         ssize_t size;
1399         struct smb_acl_entry *smb_ace;
1400         int i;
1401         int count;
1402         uint16_t tag;
1403         uint16_t perm;
1404         uint32_t id;
1405         int offset;
1406
1407         count = theacl->count;
1408
1409         size = GLUSTER_ACL_HEADER_SIZE + (count * GLUSTER_ACL_ENTRY_SIZE);
1410         if (!buf) {
1411                 return size;
1412         }
1413
1414         if (len < size) {
1415                 errno = ERANGE;
1416                 return -1;
1417         }
1418
1419         smb_ace = theacl->acl;
1420
1421         /* Version is the first 4 bytes of the ACL */
1422         SIVAL(buf, 0, GLUSTER_ACL_VERSION);
1423         offset = GLUSTER_ACL_HEADER_SIZE;
1424
1425         for (i = 0; i < count; i++) {
1426                 /* Calculate tag */
1427                 switch(smb_ace->a_type) {
1428                 case SMB_ACL_USER:
1429                         tag = GLUSTER_ACL_USER;
1430                         break;
1431                 case SMB_ACL_USER_OBJ:
1432                         tag = GLUSTER_ACL_USER_OBJ;
1433                         break;
1434                 case SMB_ACL_GROUP:
1435                         tag = GLUSTER_ACL_GROUP;
1436                         break;
1437                 case SMB_ACL_GROUP_OBJ:
1438                         tag = GLUSTER_ACL_GROUP_OBJ;
1439                         break;
1440                 case SMB_ACL_OTHER:
1441                         tag = GLUSTER_ACL_OTHER;
1442                         break;
1443                 case SMB_ACL_MASK:
1444                         tag = GLUSTER_ACL_MASK;
1445                         break;
1446                 default:
1447                         DEBUG(0, ("Unknown tag value %d\n",
1448                                   smb_ace->a_type));
1449                         errno = EINVAL;
1450                         return -1;
1451                 }
1452
1453
1454                 /* Calculate id */
1455                 switch(smb_ace->a_type) {
1456                 case SMB_ACL_USER:
1457                         id = smb_ace->info.user.uid;
1458                         break;
1459                 case SMB_ACL_GROUP:
1460                         id = smb_ace->info.group.gid;
1461                         break;
1462                 default:
1463                         id = GLUSTER_ACL_UNDEFINED_ID;
1464                         break;
1465                 }
1466
1467                 /* Calculate perm */
1468                 perm = 0;
1469
1470                 perm |=
1471                         ((smb_ace->a_perm & SMB_ACL_READ) ? GLUSTER_ACL_READ : 0);
1472                 perm |=
1473                         ((smb_ace->a_perm & SMB_ACL_WRITE) ? GLUSTER_ACL_WRITE : 0);
1474                 perm |=
1475                         ((smb_ace->a_perm & SMB_ACL_EXECUTE) ? GLUSTER_ACL_EXECUTE : 0);
1476
1477
1478                 /* TAG is the first 2 bytes of an entry */
1479                 SSVAL(buf, offset, tag);
1480                 offset += 2;
1481
1482                 /* PERM is the next 2 bytes of an entry */
1483                 SSVAL(buf, offset, perm);
1484                 offset += 2;
1485
1486                 /* ID is the last 4 bytes of an entry */
1487                 SIVAL(buf, offset, id);
1488                 offset += 4;
1489
1490                 smb_ace++;
1491         }
1492
1493         /* Skip the header, sort @count number of 8-byte entries */
1494         qsort(buf+GLUSTER_ACL_HEADER_SIZE, count, GLUSTER_ACL_ENTRY_SIZE,
1495               gluster_ace_cmp);
1496
1497         return size;
1498 }
1499
1500
1501 static SMB_ACL_T vfs_gluster_sys_acl_get_file(struct vfs_handle_struct *handle,
1502                                               const char *path_p,
1503                                               SMB_ACL_TYPE_T type,
1504                                               TALLOC_CTX *mem_ctx)
1505 {
1506         struct smb_acl_t *result;
1507         struct stat st;
1508         char *buf;
1509         const char *key;
1510         ssize_t ret, size = GLUSTER_ACL_SIZE(20);
1511
1512         switch (type) {
1513         case SMB_ACL_TYPE_ACCESS:
1514                 key = "system.posix_acl_access";
1515                 break;
1516         case SMB_ACL_TYPE_DEFAULT:
1517                 key = "system.posix_acl_default";
1518                 break;
1519         default:
1520                 errno = EINVAL;
1521                 return NULL;
1522         }
1523
1524         buf = alloca(size);
1525         if (!buf) {
1526                 return NULL;
1527         }
1528
1529         ret = glfs_getxattr(handle->data, path_p, key, buf, size);
1530         if (ret == -1 && errno == ERANGE) {
1531                 ret = glfs_getxattr(handle->data, path_p, key, 0, 0);
1532                 if (ret > 0) {
1533                         buf = alloca(ret);
1534                         if (!buf) {
1535                                 return NULL;
1536                         }
1537                         ret = glfs_getxattr(handle->data, path_p, key, buf, ret);
1538                 }
1539         }
1540
1541         /* retrieving the ACL from the xattr has finally failed, do a
1542          * mode-to-acl mapping */
1543
1544         if (ret == -1 && errno == ENODATA) {
1545                 ret = glfs_stat(handle->data, path_p, &st);
1546                 if (ret == 0) {
1547                         result = mode_to_smb_acls(&st, mem_ctx);
1548                         return result;
1549                 }
1550         }
1551
1552         if (ret <= 0) {
1553                 return NULL;
1554         }
1555
1556         result = gluster_to_smb_acl(buf, ret, mem_ctx);
1557
1558         return result;
1559 }
1560
1561 static SMB_ACL_T vfs_gluster_sys_acl_get_fd(struct vfs_handle_struct *handle,
1562                                             struct files_struct *fsp,
1563                                             TALLOC_CTX *mem_ctx)
1564 {
1565         struct smb_acl_t *result;
1566         struct stat st;
1567         ssize_t ret, size = GLUSTER_ACL_SIZE(20);
1568         char *buf;
1569         glfs_fd_t *glfd;
1570
1571         glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
1572
1573         buf = alloca(size);
1574         if (!buf) {
1575                 return NULL;
1576         }
1577
1578         ret = glfs_fgetxattr(glfd, "system.posix_acl_access", buf, size);
1579         if (ret == -1 && errno == ERANGE) {
1580                 ret = glfs_fgetxattr(glfd, "system.posix_acl_access", 0, 0);
1581                 if (ret > 0) {
1582                         buf = alloca(ret);
1583                         if (!buf) {
1584                                 return NULL;
1585                         }
1586                         ret = glfs_fgetxattr(glfd, "system.posix_acl_access",
1587                                              buf, ret);
1588                 }
1589         }
1590
1591         /* retrieving the ACL from the xattr has finally failed, do a
1592          * mode-to-acl mapping */
1593
1594         if (ret == -1 && errno == ENODATA) {
1595                 ret = glfs_fstat(glfd, &st);
1596                 if (ret == 0) {
1597                         result = mode_to_smb_acls(&st, mem_ctx);
1598                         return result;
1599                 }
1600         }
1601
1602         if (ret <= 0) {
1603                 return NULL;
1604         }
1605
1606         result = gluster_to_smb_acl(buf, ret, mem_ctx);
1607
1608         return result;
1609 }
1610
1611 static int vfs_gluster_sys_acl_set_file(struct vfs_handle_struct *handle,
1612                                         const char *name,
1613                                         SMB_ACL_TYPE_T acltype,
1614                                         SMB_ACL_T theacl)
1615 {
1616         int ret;
1617         const char *key;
1618         char *buf;
1619         ssize_t size;
1620
1621         switch (acltype) {
1622         case SMB_ACL_TYPE_ACCESS:
1623                 key = "system.posix_acl_access";
1624                 break;
1625         case SMB_ACL_TYPE_DEFAULT:
1626                 key = "system.posix_acl_default";
1627                 break;
1628         default:
1629                 errno = EINVAL;
1630                 return -1;
1631         }
1632
1633         size = smb_to_gluster_acl(theacl, 0, 0);
1634         buf = alloca(size);
1635
1636         size = smb_to_gluster_acl(theacl, buf, size);
1637         if (size == -1) {
1638                 return -1;
1639         }
1640
1641         ret = glfs_setxattr(handle->data, name, key, buf, size, 0);
1642
1643         return ret;
1644 }
1645
1646 static int vfs_gluster_sys_acl_set_fd(struct vfs_handle_struct *handle,
1647                                       struct files_struct *fsp,
1648                                       SMB_ACL_T theacl)
1649 {
1650         int ret;
1651         char *buf;
1652         ssize_t size;
1653
1654         size = smb_to_gluster_acl(theacl, 0, 0);
1655         buf = alloca(size);
1656
1657         size = smb_to_gluster_acl(theacl, buf, size);
1658         if (size == -1) {
1659                 return -1;
1660         }
1661
1662         ret = glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp),
1663                              "system.posix_acl_access", buf, size, 0);
1664         return ret;
1665 }
1666
1667 static int vfs_gluster_sys_acl_delete_def_file(struct vfs_handle_struct *handle,
1668                                                const char *path)
1669 {
1670         return glfs_removexattr(handle->data, path, "system.posix_acl_default");
1671 }
1672
1673 static struct vfs_fn_pointers glusterfs_fns = {
1674
1675         /* Disk Operations */
1676
1677         .connect_fn = vfs_gluster_connect,
1678         .disconnect_fn = vfs_gluster_disconnect,
1679         .disk_free_fn = vfs_gluster_disk_free,
1680         .get_quota_fn = vfs_gluster_get_quota,
1681         .set_quota_fn = vfs_gluster_set_quota,
1682         .statvfs_fn = vfs_gluster_statvfs,
1683         .fs_capabilities_fn = vfs_gluster_fs_capabilities,
1684
1685         .get_dfs_referrals_fn = NULL,
1686
1687         /* Directory Operations */
1688
1689         .opendir_fn = vfs_gluster_opendir,
1690         .fdopendir_fn = vfs_gluster_fdopendir,
1691         .readdir_fn = vfs_gluster_readdir,
1692         .seekdir_fn = vfs_gluster_seekdir,
1693         .telldir_fn = vfs_gluster_telldir,
1694         .rewind_dir_fn = vfs_gluster_rewinddir,
1695         .mkdir_fn = vfs_gluster_mkdir,
1696         .rmdir_fn = vfs_gluster_rmdir,
1697         .closedir_fn = vfs_gluster_closedir,
1698         .init_search_op_fn = vfs_gluster_init_search_op,
1699
1700         /* File Operations */
1701
1702         .open_fn = vfs_gluster_open,
1703         .create_file_fn = NULL,
1704         .close_fn = vfs_gluster_close,
1705         .read_fn = vfs_gluster_read,
1706         .pread_fn = vfs_gluster_pread,
1707         .pread_send_fn = vfs_gluster_pread_send,
1708         .pread_recv_fn = vfs_gluster_recv,
1709         .write_fn = vfs_gluster_write,
1710         .pwrite_fn = vfs_gluster_pwrite,
1711         .pwrite_send_fn = vfs_gluster_pwrite_send,
1712         .pwrite_recv_fn = vfs_gluster_recv,
1713         .lseek_fn = vfs_gluster_lseek,
1714         .sendfile_fn = vfs_gluster_sendfile,
1715         .recvfile_fn = vfs_gluster_recvfile,
1716         .rename_fn = vfs_gluster_rename,
1717         .fsync_fn = vfs_gluster_fsync,
1718         .fsync_send_fn = vfs_gluster_fsync_send,
1719         .fsync_recv_fn = vfs_gluster_fsync_recv,
1720
1721         .stat_fn = vfs_gluster_stat,
1722         .fstat_fn = vfs_gluster_fstat,
1723         .lstat_fn = vfs_gluster_lstat,
1724         .get_alloc_size_fn = vfs_gluster_get_alloc_size,
1725         .unlink_fn = vfs_gluster_unlink,
1726
1727         .chmod_fn = vfs_gluster_chmod,
1728         .fchmod_fn = vfs_gluster_fchmod,
1729         .chown_fn = vfs_gluster_chown,
1730         .fchown_fn = vfs_gluster_fchown,
1731         .lchown_fn = vfs_gluster_lchown,
1732         .chdir_fn = vfs_gluster_chdir,
1733         .getwd_fn = vfs_gluster_getwd,
1734         .ntimes_fn = vfs_gluster_ntimes,
1735         .ftruncate_fn = vfs_gluster_ftruncate,
1736         .fallocate_fn = vfs_gluster_fallocate,
1737         .lock_fn = vfs_gluster_lock,
1738         .kernel_flock_fn = vfs_gluster_kernel_flock,
1739         .linux_setlease_fn = vfs_gluster_linux_setlease,
1740         .getlock_fn = vfs_gluster_getlock,
1741         .symlink_fn = vfs_gluster_symlink,
1742         .readlink_fn = vfs_gluster_readlink,
1743         .link_fn = vfs_gluster_link,
1744         .mknod_fn = vfs_gluster_mknod,
1745         .realpath_fn = vfs_gluster_realpath,
1746         .notify_watch_fn = vfs_gluster_notify_watch,
1747         .chflags_fn = vfs_gluster_chflags,
1748         .file_id_create_fn = NULL,
1749         .copy_chunk_send_fn = NULL,
1750         .copy_chunk_recv_fn = NULL,
1751         .streaminfo_fn = NULL,
1752         .get_real_filename_fn = vfs_gluster_get_real_filename,
1753         .connectpath_fn = vfs_gluster_connectpath,
1754
1755         .brl_lock_windows_fn = NULL,
1756         .brl_unlock_windows_fn = NULL,
1757         .brl_cancel_windows_fn = NULL,
1758         .strict_lock_fn = NULL,
1759         .strict_unlock_fn = NULL,
1760         .translate_name_fn = NULL,
1761         .fsctl_fn = NULL,
1762
1763         /* NT ACL Operations */
1764         .fget_nt_acl_fn = NULL,
1765         .get_nt_acl_fn = NULL,
1766         .fset_nt_acl_fn = NULL,
1767         .audit_file_fn = NULL,
1768
1769         /* Posix ACL Operations */
1770         .chmod_acl_fn = NULL,   /* passthrough to default */
1771         .fchmod_acl_fn = NULL,  /* passthrough to default */
1772         .sys_acl_get_file_fn = vfs_gluster_sys_acl_get_file,
1773         .sys_acl_get_fd_fn = vfs_gluster_sys_acl_get_fd,
1774         .sys_acl_blob_get_file_fn = posix_sys_acl_blob_get_file,
1775         .sys_acl_blob_get_fd_fn = posix_sys_acl_blob_get_fd,
1776         .sys_acl_set_file_fn = vfs_gluster_sys_acl_set_file,
1777         .sys_acl_set_fd_fn = vfs_gluster_sys_acl_set_fd,
1778         .sys_acl_delete_def_file_fn = vfs_gluster_sys_acl_delete_def_file,
1779
1780         /* EA Operations */
1781         .getxattr_fn = vfs_gluster_getxattr,
1782         .fgetxattr_fn = vfs_gluster_fgetxattr,
1783         .listxattr_fn = vfs_gluster_listxattr,
1784         .flistxattr_fn = vfs_gluster_flistxattr,
1785         .removexattr_fn = vfs_gluster_removexattr,
1786         .fremovexattr_fn = vfs_gluster_fremovexattr,
1787         .setxattr_fn = vfs_gluster_setxattr,
1788         .fsetxattr_fn = vfs_gluster_fsetxattr,
1789
1790         /* AIO Operations */
1791         .aio_force_fn = vfs_gluster_aio_force,
1792
1793         /* Offline Operations */
1794         .is_offline_fn = vfs_gluster_is_offline,
1795         .set_offline_fn = vfs_gluster_set_offline,
1796
1797         /* Durable handle Operations */
1798         .durable_cookie_fn = NULL,
1799         .durable_disconnect_fn = NULL,
1800         .durable_reconnect_fn = NULL,
1801 };
1802
1803 NTSTATUS vfs_glusterfs_init(void);
1804 NTSTATUS vfs_glusterfs_init(void)
1805 {
1806         return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
1807                                 "glusterfs", &glusterfs_fns);
1808 }