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/util/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, uint64_t *bsize_p,
 275                                       uint64_t *dfree_p, uint64_t *dsize_p)
 276 {
 277         struct statvfs statvfs = { 0, };
 278         int ret;
 279
 280         ret = glfs_statvfs(handle->data, path, &statvfs);
 281         if (ret < 0) {
 282                 return -1;
 283         }
 284
 285         if (bsize_p != NULL) {
 286                 *bsize_p = (uint64_t)statvfs.f_bsize; /* Block size */
 287         }
 288         if (dfree_p != NULL) {
 289                 *dfree_p = (uint64_t)statvfs.f_bavail; /* Available Block units */
 290         }
 291         if (dsize_p != NULL) {
 292                 *dsize_p = (uint64_t)statvfs.f_blocks; /* Total Block units */
 293         }
 294
 295         return (uint64_t)statvfs.f_bavail;
 296 }
 297
 298 static int vfs_gluster_get_quota(struct vfs_handle_struct *handle,
 299                                  enum SMB_QUOTA_TYPE qtype, unid_t id,
 300                                  SMB_DISK_QUOTA *qt)
 301 {
 302         errno = ENOSYS;
 303         return -1;
 304 }
 305
 306 static int
 307 vfs_gluster_set_quota(struct vfs_handle_struct *handle,
 308                       enum SMB_QUOTA_TYPE qtype, unid_t id, SMB_DISK_QUOTA *qt)
 309 {
 310         errno = ENOSYS;
 311         return -1;
 312 }
 313
 314 static int vfs_gluster_statvfs(struct vfs_handle_struct *handle,
 315                                const char *path,
 316                                struct vfs_statvfs_struct *vfs_statvfs)
 317 {
 318         struct statvfs statvfs = { 0, };
 319         int ret;
 320
 321         ret = glfs_statvfs(handle->data, path, &statvfs);
 322         if (ret < 0) {
 323                 DEBUG(0, ("glfs_statvfs(%s) failed: %s\n",
 324                           path, strerror(errno)));
 325                 return -1;
 326         }
 327
 328         ZERO_STRUCTP(vfs_statvfs);
 329
 330         vfs_statvfs->OptimalTransferSize = statvfs.f_frsize;
 331         vfs_statvfs->BlockSize = statvfs.f_bsize;
 332         vfs_statvfs->TotalBlocks = statvfs.f_blocks;
 333         vfs_statvfs->BlocksAvail = statvfs.f_bfree;
 334         vfs_statvfs->UserBlocksAvail = statvfs.f_bavail;
 335         vfs_statvfs->TotalFileNodes = statvfs.f_files;
 336         vfs_statvfs->FreeFileNodes = statvfs.f_ffree;
 337         vfs_statvfs->FsIdentifier = statvfs.f_fsid;
 338         vfs_statvfs->FsCapabilities =
 339             FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
 340
 341         return ret;
 342 }
 343
 344 static uint32_t vfs_gluster_fs_capabilities(struct vfs_handle_struct *handle,
 345                                             enum timestamp_set_resolution *p_ts_res)
 346 {
 347         uint32_t caps = FILE_CASE_SENSITIVE_SEARCH | FILE_CASE_PRESERVED_NAMES;
 348
 349 #ifdef STAT_HAVE_NSEC
 350         *p_ts_res = TIMESTAMP_SET_NT_OR_BETTER;
 351 #endif
 352
 353         return caps;
 354 }
 355
 356 static DIR *vfs_gluster_opendir(struct vfs_handle_struct *handle,
 357                                 const char *path, const char *mask,
 358                                 uint32_t attributes)
 359 {
 360         glfs_fd_t *fd;
 361
 362         fd = glfs_opendir(handle->data, path);
 363         if (fd == NULL) {
 364                 DEBUG(0, ("glfs_opendir(%s) failed: %s\n",
 365                           path, strerror(errno)));
 366         }
 367
 368         return (DIR *) fd;
 369 }
 370
 371 static DIR *vfs_gluster_fdopendir(struct vfs_handle_struct *handle,
 372                                   files_struct *fsp, const char *mask,
 373                                   uint32_t attributes)
 374 {
 375         return (DIR *) *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
 376 }
 377
 378 static int vfs_gluster_closedir(struct vfs_handle_struct *handle, DIR *dirp)
 379 {
 380         return glfs_closedir((void *)dirp);
 381 }
 382
 383 static struct dirent *vfs_gluster_readdir(struct vfs_handle_struct *handle,
 384                                           DIR *dirp, SMB_STRUCT_STAT *sbuf)
 385 {
 386         static char direntbuf[512];
 387         int ret;
 388         struct stat stat;
 389         struct dirent *dirent = 0;
 390
 391         if (sbuf != NULL) {
 392                 ret = glfs_readdirplus_r((void *)dirp, &stat, (void *)direntbuf,
 393                                          &dirent);
 394         } else {
 395                 ret = glfs_readdir_r((void *)dirp, (void *)direntbuf, &dirent);
 396         }
 397
 398         if ((ret < 0) || (dirent == NULL)) {
 399                 return NULL;
 400         }
 401
 402         if (sbuf != NULL) {
 403                 smb_stat_ex_from_stat(sbuf, &stat);
 404         }
 405
 406         return dirent;
 407 }
 408
 409 static long vfs_gluster_telldir(struct vfs_handle_struct *handle, DIR *dirp)
 410 {
 411         return glfs_telldir((void *)dirp);
 412 }
 413
 414 static void vfs_gluster_seekdir(struct vfs_handle_struct *handle, DIR *dirp,
 415                                 long offset)
 416 {
 417         glfs_seekdir((void *)dirp, offset);
 418 }
 419
 420 static void vfs_gluster_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
 421 {
 422         glfs_seekdir((void *)dirp, 0);
 423 }
 424
 425 static void vfs_gluster_init_search_op(struct vfs_handle_struct *handle,
 426                                        DIR *dirp)
 427 {
 428         return;
 429 }
 430
 431 static int vfs_gluster_mkdir(struct vfs_handle_struct *handle, const char *path,
 432                              mode_t mode)
 433 {
 434         return glfs_mkdir(handle->data, path, mode);
 435 }
 436
 437 static int vfs_gluster_rmdir(struct vfs_handle_struct *handle, const char *path)
 438 {
 439         return glfs_rmdir(handle->data, path);
 440 }
 441
 442 static int vfs_gluster_open(struct vfs_handle_struct *handle,
 443                             struct smb_filename *smb_fname, files_struct *fsp,
 444                             int flags, mode_t mode)
 445 {
 446         glfs_fd_t *glfd;
 447         glfs_fd_t **p_tmp;
 448
 449         if (flags & O_DIRECTORY) {
 450                 glfd = glfs_opendir(handle->data, smb_fname->base_name);
 451         } else if (flags & O_CREAT) {
 452                 glfd = glfs_creat(handle->data, smb_fname->base_name, flags,
 453                                   mode);
 454         } else {
 455                 glfd = glfs_open(handle->data, smb_fname->base_name, flags);
 456         }
 457
 458         if (glfd == NULL) {
 459                 return -1;
 460         }
 461         p_tmp = (glfs_fd_t **)VFS_ADD_FSP_EXTENSION(handle, fsp,
 462                                                           glfs_fd_t *, NULL);
 463         *p_tmp = glfd;
 464         /* An arbitrary value for error reporting, so you know its us. */
 465         return 13371337;
 466 }
 467
 468 static int vfs_gluster_close(struct vfs_handle_struct *handle,
 469                              files_struct *fsp)
 470 {
 471         glfs_fd_t *glfd;
 472         glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
 473         VFS_REMOVE_FSP_EXTENSION(handle, fsp);
 474         return glfs_close(glfd);
 475 }
 476
 477 static ssize_t vfs_gluster_read(struct vfs_handle_struct *handle,
 478                                 files_struct *fsp, void *data, size_t n)
 479 {
 480         return glfs_read(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
 481 }
 482
 483 static ssize_t vfs_gluster_pread(struct vfs_handle_struct *handle,
 484                                  files_struct *fsp, void *data, size_t n,
 485                                  off_t offset)
 486 {
 487         return glfs_pread(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
 488 }
 489
 490 struct glusterfs_aio_state;
 491
 492 struct glusterfs_aio_wrapper {
 493         struct glusterfs_aio_state *state;
 494 };
 495
 496 struct glusterfs_aio_state {
 497         ssize_t ret;
 498         int err;
 499         struct tevent_req *req;
 500         bool cancelled;
 501 };
 502
 503 static int aio_wrapper_destructor(struct glusterfs_aio_wrapper *wrap)
 504 {
 505         wrap->state->cancelled = true;
 506
 507         return 0;
 508 }
 509
 510 /*
 511  * This function is the callback that will be called on glusterfs
 512  * threads once the async IO submitted is complete. To notify
 513  * Samba of the completion we use a pipe based queue.
 514  */
 515 static void aio_glusterfs_done(glfs_fd_t *fd, ssize_t ret, void *data)
 516 {
 517         struct glusterfs_aio_state *state = NULL;
 518         int sts = 0;
 519
 520         state = (struct glusterfs_aio_state *)data;
 521
 522         if (ret < 0) {
 523                 state->ret = -1;
 524                 state->err = errno;
 525         } else {
 526                 state->ret = ret;
 527                 state->err = 0;
 528         }
 529
 530         /*
 531          * Write the state pointer to glusterfs_aio_state to the
 532          * pipe, so we can call tevent_req_done() from the main thread,
 533          * because tevent_req_done() is not designed to be executed in
 534          * the multithread environment, so tevent_req_done() must be
 535          * executed from the smbd main thread.
 536          *
 537          * write(2) on pipes with sizes under _POSIX_PIPE_BUF
 538          * in size is atomic, without this, the use op pipes in this
 539          * code would not work.
 540          *
 541          * sys_write is a thin enough wrapper around write(2)
 542          * that we can trust it here.
 543          */
 544
 545         sts = sys_write(write_fd, &state, sizeof(struct glusterfs_aio_state *));
 546         if (sts < 0) {
 547                 DEBUG(0,("\nWrite to pipe failed (%s)", strerror(errno)));
 548         }
 549
 550         return;
 551 }
 552
 553 /*
 554  * Read each req off the pipe and process it.
 555  */
 556 static void aio_tevent_fd_done(struct tevent_context *event_ctx,
 557                                 struct tevent_fd *fde,
 558                                 uint16_t flags, void *data)
 559 {
 560         struct tevent_req *req = NULL;
 561         struct glusterfs_aio_state *state = NULL;
 562         int sts = 0;
 563
 564         /*
 565          * read(2) on pipes is atomic if the needed data is available
 566          * in the pipe, per SUS and POSIX.  Because we always write
 567          * to the pipe in sizeof(struct tevent_req *) chunks, we can
 568          * always read in those chunks, atomically.
 569          *
 570          * sys_read is a thin enough wrapper around read(2) that we
 571          * can trust it here.
 572          */
 573
 574         sts = sys_read(read_fd, &state, sizeof(struct glusterfs_aio_state *));
 575
 576         if (sts < 0) {
 577                 DEBUG(0,("\nRead from pipe failed (%s)", strerror(errno)));
 578         }
 579
 580         /* if we've cancelled the op, there is no req, so just clean up. */
 581         if (state->cancelled == true) {
 582                 TALLOC_FREE(state);
 583                 return;
 584         }
 585
 586         req = state->req;
 587
 588         if (req) {
 589                 tevent_req_done(req);
 590         }
 591         return;
 592 }
 593
 594 static bool init_gluster_aio(struct vfs_handle_struct *handle)
 595 {
 596         int fds[2];
 597         int ret = -1;
 598
 599         if (read_fd != -1) {
 600                 /*
 601                  * Already initialized.
 602                  */
 603                 return true;
 604         }
 605
 606         ret = pipe(fds);
 607         if (ret == -1) {
 608                 goto fail;
 609         }
 610
 611         read_fd = fds[0];
 612         write_fd = fds[1];
 613
 614         aio_read_event = tevent_add_fd(handle->conn->sconn->ev_ctx,
 615                                         NULL,
 616                                         read_fd,
 617                                         TEVENT_FD_READ,
 618                                         aio_tevent_fd_done,
 619                                         NULL);
 620         if (aio_read_event == NULL) {
 621                 goto fail;
 622         }
 623
 624         return true;
 625 fail:
 626         TALLOC_FREE(aio_read_event);
 627         if (read_fd != -1) {
 628                 close(read_fd);
 629                 close(write_fd);
 630                 read_fd = -1;
 631                 write_fd = -1;
 632         }
 633         return false;
 634 }
 635
 636 static struct glusterfs_aio_state *aio_state_create(TALLOC_CTX *mem_ctx)
 637 {
 638         struct tevent_req *req = NULL;
 639         struct glusterfs_aio_state *state = NULL;
 640         struct glusterfs_aio_wrapper *wrapper = NULL;
 641
 642         req = tevent_req_create(mem_ctx, &wrapper, struct glusterfs_aio_wrapper);
 643
 644         if (req == NULL) {
 645                 return NULL;
 646         }
 647
 648         state = talloc(NULL, struct glusterfs_aio_state);
 649
 650         if (state == NULL) {
 651                 TALLOC_FREE(req);
 652                 return NULL;
 653         }
 654
 655         talloc_set_destructor(wrapper, aio_wrapper_destructor);
 656         state->cancelled = false;
 657         state->ret = 0;
 658         state->err = 0;
 659         state->req = req;
 660
 661         wrapper->state = state;
 662
 663         return state;
 664 }
 665
 666 static struct tevent_req *vfs_gluster_pread_send(struct vfs_handle_struct
 667                                                   *handle, TALLOC_CTX *mem_ctx,
 668                                                   struct tevent_context *ev,
 669                                                   files_struct *fsp,
 670                                                   void *data, size_t n,
 671                                                   off_t offset)
 672 {
 673         struct glusterfs_aio_state *state = NULL;
 674         struct tevent_req *req = NULL;
 675         int ret = 0;
 676
 677         state = aio_state_create(mem_ctx);
 678
 679         if (state == NULL) {
 680                 return NULL;
 681         }
 682
 683         req = state->req;
 684
 685         if (!init_gluster_aio(handle)) {
 686                 tevent_req_error(req, EIO);
 687                 return tevent_req_post(req, ev);
 688         }
 689
 690         ret = glfs_pread_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 691                                 fsp), data, n, offset, 0, aio_glusterfs_done,
 692                                 state);
 693         if (ret < 0) {
 694                 tevent_req_error(req, -ret);
 695                 return tevent_req_post(req, ev);
 696         }
 697
 698         return req;
 699 }
 700
 701 static struct tevent_req *vfs_gluster_pwrite_send(struct vfs_handle_struct
 702                                                   *handle, TALLOC_CTX *mem_ctx,
 703                                                   struct tevent_context *ev,
 704                                                   files_struct *fsp,
 705                                                   const void *data, size_t n,
 706                                                   off_t offset)
 707 {
 708         struct glusterfs_aio_state *state = NULL;
 709         struct tevent_req *req = NULL;
 710         int ret = 0;
 711
 712         state = aio_state_create(mem_ctx);
 713
 714         if (state == NULL) {
 715                 return NULL;
 716         }
 717
 718         req = state->req;
 719
 720         if (!init_gluster_aio(handle)) {
 721                 tevent_req_error(req, EIO);
 722                 return tevent_req_post(req, ev);
 723         }
 724
 725         ret = glfs_pwrite_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 726                                 fsp), data, n, offset, 0, aio_glusterfs_done,
 727                                 state);
 728         if (ret < 0) {
 729                 tevent_req_error(req, -ret);
 730                 return tevent_req_post(req, ev);
 731         }
 732
 733         return req;
 734 }
 735
 736 static ssize_t vfs_gluster_recv(struct tevent_req *req, int *err)
 737 {
 738         struct glusterfs_aio_state *state = NULL;
 739         struct glusterfs_aio_wrapper *wrapper = NULL;
 740         int ret = 0;
 741
 742         wrapper = tevent_req_data(req, struct glusterfs_aio_wrapper);
 743
 744         if (wrapper == NULL) {
 745                 return -1;
 746         }
 747
 748         state = wrapper->state;
 749
 750         if (state == NULL) {
 751                 return -1;
 752         }
 753
 754         if (tevent_req_is_unix_error(req, err)) {
 755                 return -1;
 756         }
 757         if (state->ret == -1) {
 758                 *err = state->err;
 759         }
 760
 761         ret = state->ret;
 762
 763         /* Clean up the state, it is in a NULL context. */
 764
 765         TALLOC_FREE(state);
 766
 767         return ret;
 768 }
 769
 770 static ssize_t vfs_gluster_write(struct vfs_handle_struct *handle,
 771                                  files_struct *fsp, const void *data, size_t n)
 772 {
 773         return glfs_write(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
 774 }
 775
 776 static ssize_t vfs_gluster_pwrite(struct vfs_handle_struct *handle,
 777                                   files_struct *fsp, const void *data,
 778                                   size_t n, off_t offset)
 779 {
 780         return glfs_pwrite(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
 781 }
 782
 783 static off_t vfs_gluster_lseek(struct vfs_handle_struct *handle,
 784                                files_struct *fsp, off_t offset, int whence)
 785 {
 786         return glfs_lseek(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset, whence);
 787 }
 788
 789 static ssize_t vfs_gluster_sendfile(struct vfs_handle_struct *handle, int tofd,
 790                                     files_struct *fromfsp,
 791                                     const DATA_BLOB *hdr,
 792                                     off_t offset, size_t n)
 793 {
 794         errno = ENOTSUP;
 795         return -1;
 796 }
 797
 798 static ssize_t vfs_gluster_recvfile(struct vfs_handle_struct *handle,
 799                                     int fromfd, files_struct *tofsp,
 800                                     off_t offset, size_t n)
 801 {
 802         errno = ENOTSUP;
 803         return -1;
 804 }
 805
 806 static int vfs_gluster_rename(struct vfs_handle_struct *handle,
 807                               const struct smb_filename *smb_fname_src,
 808                               const struct smb_filename *smb_fname_dst)
 809 {
 810         return glfs_rename(handle->data, smb_fname_src->base_name,
 811                            smb_fname_dst->base_name);
 812 }
 813
 814 static int vfs_gluster_fsync(struct vfs_handle_struct *handle,
 815                              files_struct *fsp)
 816 {
 817         return glfs_fsync(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp));
 818 }
 819
 820 static struct tevent_req *vfs_gluster_fsync_send(struct vfs_handle_struct
 821                                                  *handle, TALLOC_CTX *mem_ctx,
 822                                                  struct tevent_context *ev,
 823                                                  files_struct *fsp)
 824 {
 825         struct tevent_req *req = NULL;
 826         struct glusterfs_aio_state *state = NULL;
 827         int ret = 0;
 828
 829         state = aio_state_create(mem_ctx);
 830
 831         if (state == NULL) {
 832                 return NULL;
 833         }
 834
 835         req = state->req;
 836
 837         if (!init_gluster_aio(handle)) {
 838                 tevent_req_error(req, EIO);
 839                 return tevent_req_post(req, ev);
 840         }
 841         ret = glfs_fsync_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 842                                 fsp), aio_glusterfs_done, req);
 843         if (ret < 0) {
 844                 tevent_req_error(req, -ret);
 845                 return tevent_req_post(req, ev);
 846         }
 847         return req;
 848 }
 849
 850 static int vfs_gluster_fsync_recv(struct tevent_req *req, int *err)
 851 {
 852         /*
 853          * Use implicit conversion ssize_t->int
 854          */
 855         return vfs_gluster_recv(req, err);
 856 }
 857
 858 static int vfs_gluster_stat(struct vfs_handle_struct *handle,
 859                             struct smb_filename *smb_fname)
 860 {
 861         struct stat st;
 862         int ret;
 863
 864         ret = glfs_stat(handle->data, smb_fname->base_name, &st);
 865         if (ret == 0) {
 866                 smb_stat_ex_from_stat(&smb_fname->st, &st);
 867         }
 868         if (ret < 0 && errno != ENOENT) {
 869                 DEBUG(0, ("glfs_stat(%s) failed: %s\n",
 870                           smb_fname->base_name, strerror(errno)));
 871         }
 872         return ret;
 873 }
 874
 875 static int vfs_gluster_fstat(struct vfs_handle_struct *handle,
 876                              files_struct *fsp, SMB_STRUCT_STAT *sbuf)
 877 {
 878         struct stat st;
 879         int ret;
 880
 881         ret = glfs_fstat(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), &st);
 882         if (ret == 0) {
 883                 smb_stat_ex_from_stat(sbuf, &st);
 884         }
 885         if (ret < 0) {
 886                 DEBUG(0, ("glfs_fstat(%d) failed: %s\n",
 887                           fsp->fh->fd, strerror(errno)));
 888         }
 889         return ret;
 890 }
 891
 892 static int vfs_gluster_lstat(struct vfs_handle_struct *handle,
 893                              struct smb_filename *smb_fname)
 894 {
 895         struct stat st;
 896         int ret;
 897
 898         ret = glfs_lstat(handle->data, smb_fname->base_name, &st);
 899         if (ret == 0) {
 900                 smb_stat_ex_from_stat(&smb_fname->st, &st);
 901         }
 902         if (ret < 0 && errno != ENOENT) {
 903                 DEBUG(0, ("glfs_lstat(%s) failed: %s\n",
 904                           smb_fname->base_name, strerror(errno)));
 905         }
 906         return ret;
 907 }
 908
 909 static uint64_t vfs_gluster_get_alloc_size(struct vfs_handle_struct *handle,
 910                                            files_struct *fsp,
 911                                            const SMB_STRUCT_STAT *sbuf)
 912 {
 913         return sbuf->st_ex_blocks * 512;
 914 }
 915
 916 static int vfs_gluster_unlink(struct vfs_handle_struct *handle,
 917                               const struct smb_filename *smb_fname)
 918 {
 919         return glfs_unlink(handle->data, smb_fname->base_name);
 920 }
 921
 922 static int vfs_gluster_chmod(struct vfs_handle_struct *handle,
 923                              const char *path, mode_t mode)
 924 {
 925         return glfs_chmod(handle->data, path, mode);
 926 }
 927
 928 static int vfs_gluster_fchmod(struct vfs_handle_struct *handle,
 929                               files_struct *fsp, mode_t mode)
 930 {
 931         return glfs_fchmod(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), mode);
 932 }
 933
 934 static int vfs_gluster_chown(struct vfs_handle_struct *handle,
 935                              const char *path, uid_t uid, gid_t gid)
 936 {
 937         return glfs_chown(handle->data, path, uid, gid);
 938 }
 939
 940 static int vfs_gluster_fchown(struct vfs_handle_struct *handle,
 941                               files_struct *fsp, uid_t uid, gid_t gid)
 942 {
 943         return glfs_fchown(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), uid, gid);
 944 }
 945
 946 static int vfs_gluster_lchown(struct vfs_handle_struct *handle,
 947                               const char *path, uid_t uid, gid_t gid)
 948 {
 949         return glfs_lchown(handle->data, path, uid, gid);
 950 }
 951
 952 static int vfs_gluster_chdir(struct vfs_handle_struct *handle, const char *path)
 953 {
 954         return glfs_chdir(handle->data, path);
 955 }
 956
 957 static char *vfs_gluster_getwd(struct vfs_handle_struct *handle)
 958 {
 959         char *cwd;
 960         char *ret;
 961
 962         cwd = SMB_CALLOC_ARRAY(char, PATH_MAX);
 963         if (cwd == NULL) {
 964                 return NULL;
 965         }
 966
 967         ret = glfs_getcwd(handle->data, cwd, PATH_MAX - 1);
 968         if (ret == 0) {
 969                 free(cwd);
 970         }
 971         return ret;
 972 }
 973
 974 static int vfs_gluster_ntimes(struct vfs_handle_struct *handle,
 975                               const struct smb_filename *smb_fname,
 976                               struct smb_file_time *ft)
 977 {
 978         struct timespec times[2];
 979
 980         if (null_timespec(ft->atime)) {
 981                 times[0].tv_sec = smb_fname->st.st_ex_atime.tv_sec;
 982                 times[0].tv_nsec = smb_fname->st.st_ex_atime.tv_nsec;
 983         } else {
 984                 times[0].tv_sec = ft->atime.tv_sec;
 985                 times[0].tv_nsec = ft->atime.tv_nsec;
 986         }
 987
 988         if (null_timespec(ft->mtime)) {
 989                 times[1].tv_sec = smb_fname->st.st_ex_mtime.tv_sec;
 990                 times[1].tv_nsec = smb_fname->st.st_ex_mtime.tv_nsec;
 991         } else {
 992                 times[1].tv_sec = ft->mtime.tv_sec;
 993                 times[1].tv_nsec = ft->mtime.tv_nsec;
 994         }
 995
 996         if ((timespec_compare(&times[0],
 997                               &smb_fname->st.st_ex_atime) == 0) &&
 998             (timespec_compare(&times[1],
 999                               &smb_fname->st.st_ex_mtime) == 0)) {
1000                 return 0;
1001         }
1002
1003         return glfs_utimens(handle->data, smb_fname->base_name, times);
1004 }
1005
1006 static int vfs_gluster_ftruncate(struct vfs_handle_struct *handle,
1007                                  files_struct *fsp, off_t offset)
1008 {
1009         return glfs_ftruncate(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset);
1010 }
1011
1012 static int vfs_gluster_fallocate(struct vfs_handle_struct *handle,
1013                                  struct files_struct *fsp,
1014                                  uint32_t mode,
1015                                  off_t offset, off_t len)
1016 {
1017         /* TODO: add support using glfs_fallocate() and glfs_zerofill() */
1018         errno = ENOTSUP;
1019         return -1;
1020 }
1021
1022 static char *vfs_gluster_realpath(struct vfs_handle_struct *handle,
1023                                   const char *path)
1024 {
1025         return glfs_realpath(handle->data, path, 0);
1026 }
1027
1028 static bool vfs_gluster_lock(struct vfs_handle_struct *handle,
1029                              files_struct *fsp, int op, off_t offset,
1030                              off_t count, int type)
1031 {
1032         struct flock flock = { 0, };
1033         int ret;
1034
1035         flock.l_type = type;
1036         flock.l_whence = SEEK_SET;
1037         flock.l_start = offset;
1038         flock.l_len = count;
1039         flock.l_pid = 0;
1040
1041         ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), op, &flock);
1042
1043         if (op == F_GETLK) {
1044                 /* lock query, true if someone else has locked */
1045                 if ((ret != -1) &&
1046                     (flock.l_type != F_UNLCK) &&
1047                     (flock.l_pid != 0) && (flock.l_pid != getpid()))
1048                         return true;
1049                 /* not me */
1050                 return false;
1051         }
1052
1053         if (ret == -1) {
1054                 return false;
1055         }
1056
1057         return true;
1058 }
1059
1060 static int vfs_gluster_kernel_flock(struct vfs_handle_struct *handle,
1061                                     files_struct *fsp, uint32_t share_mode,
1062                                     uint32_t access_mask)
1063 {
1064         errno = ENOSYS;
1065         return -1;
1066 }
1067
1068 static int vfs_gluster_linux_setlease(struct vfs_handle_struct *handle,
1069                                       files_struct *fsp, int leasetype)
1070 {
1071         errno = ENOSYS;
1072         return -1;
1073 }
1074
1075 static bool vfs_gluster_getlock(struct vfs_handle_struct *handle,
1076                                 files_struct *fsp, off_t *poffset,
1077                                 off_t *pcount, int *ptype, pid_t *ppid)
1078 {
1079         struct flock flock = { 0, };
1080         int ret;
1081
1082         flock.l_type = *ptype;
1083         flock.l_whence = SEEK_SET;
1084         flock.l_start = *poffset;
1085         flock.l_len = *pcount;
1086         flock.l_pid = 0;
1087
1088         ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), F_GETLK, &flock);
1089
1090         if (ret == -1) {
1091                 return false;
1092         }
1093
1094         *ptype = flock.l_type;
1095         *poffset = flock.l_start;
1096         *pcount = flock.l_len;
1097         *ppid = flock.l_pid;
1098
1099         return true;
1100 }
1101
1102 static int vfs_gluster_symlink(struct vfs_handle_struct *handle,
1103                                const char *oldpath, const char *newpath)
1104 {
1105         return glfs_symlink(handle->data, oldpath, newpath);
1106 }
1107
1108 static int vfs_gluster_readlink(struct vfs_handle_struct *handle,
1109                                 const char *path, char *buf, size_t bufsiz)
1110 {
1111         return glfs_readlink(handle->data, path, buf, bufsiz);
1112 }
1113
1114 static int vfs_gluster_link(struct vfs_handle_struct *handle,
1115                             const char *oldpath, const char *newpath)
1116 {
1117         return glfs_link(handle->data, oldpath, newpath);
1118 }
1119
1120 static int vfs_gluster_mknod(struct vfs_handle_struct *handle, const char *path,
1121                              mode_t mode, SMB_DEV_T dev)
1122 {
1123         return glfs_mknod(handle->data, path, mode, dev);
1124 }
1125
1126 static int vfs_gluster_chflags(struct vfs_handle_struct *handle,
1127                                const char *path, unsigned int flags)
1128 {
1129         errno = ENOSYS;
1130         return -1;
1131 }
1132
1133 static int vfs_gluster_get_real_filename(struct vfs_handle_struct *handle,
1134                                          const char *path, const char *name,
1135                                          TALLOC_CTX *mem_ctx, char **found_name)
1136 {
1137         int ret;
1138         char key_buf[NAME_MAX + 64];
1139         char val_buf[NAME_MAX + 1];
1140
1141         if (strlen(name) >= NAME_MAX) {
1142                 errno = ENAMETOOLONG;
1143                 return -1;
1144         }
1145
1146         snprintf(key_buf, NAME_MAX + 64,
1147                  "glusterfs.get_real_filename:%s", name);
1148
1149         ret = glfs_getxattr(handle->data, path, key_buf, val_buf, NAME_MAX + 1);
1150         if (ret == -1) {
1151                 if (errno == ENODATA) {
1152                         errno = EOPNOTSUPP;
1153                 }
1154                 return -1;
1155         }
1156
1157         *found_name = talloc_strdup(mem_ctx, val_buf);
1158         if (found_name[0] == NULL) {
1159                 errno = ENOMEM;
1160                 return -1;
1161         }
1162         return 0;
1163 }
1164
1165 static const char *vfs_gluster_connectpath(struct vfs_handle_struct *handle,
1166                                            const char *filename)
1167 {
1168         return handle->conn->connectpath;
1169 }
1170
1171 /* EA Operations */
1172
1173 static ssize_t vfs_gluster_getxattr(struct vfs_handle_struct *handle,
1174                                     const char *path, const char *name,
1175                                     void *value, size_t size)
1176 {
1177         return glfs_getxattr(handle->data, path, name, value, size);
1178 }
1179
1180 static ssize_t vfs_gluster_fgetxattr(struct vfs_handle_struct *handle,
1181                                      files_struct *fsp, const char *name,
1182                                      void *value, size_t size)
1183 {
1184         return glfs_fgetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size);
1185 }
1186
1187 static ssize_t vfs_gluster_listxattr(struct vfs_handle_struct *handle,
1188                                      const char *path, char *list, size_t size)
1189 {
1190         return glfs_listxattr(handle->data, path, list, size);
1191 }
1192
1193 static ssize_t vfs_gluster_flistxattr(struct vfs_handle_struct *handle,
1194                                       files_struct *fsp, char *list,
1195                                       size_t size)
1196 {
1197         return glfs_flistxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), list, size);
1198 }
1199
1200 static int vfs_gluster_removexattr(struct vfs_handle_struct *handle,
1201                                    const char *path, const char *name)
1202 {
1203         return glfs_removexattr(handle->data, path, name);
1204 }
1205
1206 static int vfs_gluster_fremovexattr(struct vfs_handle_struct *handle,
1207                                     files_struct *fsp, const char *name)
1208 {
1209         return glfs_fremovexattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name);
1210 }
1211
1212 static int vfs_gluster_setxattr(struct vfs_handle_struct *handle,
1213                                 const char *path, const char *name,
1214                                 const void *value, size_t size, int flags)
1215 {
1216         return glfs_setxattr(handle->data, path, name, value, size, flags);
1217 }
1218
1219 static int vfs_gluster_fsetxattr(struct vfs_handle_struct *handle,
1220                                  files_struct *fsp, const char *name,
1221                                  const void *value, size_t size, int flags)
1222 {
1223         return glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size,
1224                               flags);
1225 }
1226
1227 /* AIO Operations */
1228
1229 static bool vfs_gluster_aio_force(struct vfs_handle_struct *handle,
1230                                   files_struct *fsp)
1231 {
1232         return false;
1233 }
1234
1235 /* Offline Operations */
1236
1237 static bool vfs_gluster_is_offline(struct vfs_handle_struct *handle,
1238                                    const struct smb_filename *fname,
1239                                    SMB_STRUCT_STAT *sbuf)
1240 {
1241         return false;
1242 }
1243
1244 static int vfs_gluster_set_offline(struct vfs_handle_struct *handle,
1245                                    const struct smb_filename *fname)
1246 {
1247         errno = ENOTSUP;
1248         return -1;
1249 }
1250
1251 /*
1252   Gluster ACL Format:
1253
1254   Size = 4 (header) + N * 8 (entry)
1255
1256   Offset  Size    Field (Little Endian)
1257   -------------------------------------
1258   0-3     4-byte  Version
1259
1260   4-5     2-byte  Entry-1 tag
1261   6-7     2-byte  Entry-1 perm
1262   8-11    4-byte  Entry-1 id
1263
1264   12-13   2-byte  Entry-2 tag
1265   14-15   2-byte  Entry-2 perm
1266   16-19   4-byte  Entry-2 id
1267
1268   ...
1269
1270  */
1271
1272 /* header version */
1273 #define GLUSTER_ACL_VERSION 2
1274
1275 /* perm bits */
1276 #define GLUSTER_ACL_READ    0x04
1277 #define GLUSTER_ACL_WRITE   0x02
1278 #define GLUSTER_ACL_EXECUTE 0x01
1279
1280 /* tag values */
1281 #define GLUSTER_ACL_UNDEFINED_TAG  0x00
1282 #define GLUSTER_ACL_USER_OBJ       0x01
1283 #define GLUSTER_ACL_USER           0x02
1284 #define GLUSTER_ACL_GROUP_OBJ      0x04
1285 #define GLUSTER_ACL_GROUP          0x08
1286 #define GLUSTER_ACL_MASK           0x10
1287 #define GLUSTER_ACL_OTHER          0x20
1288
1289 #define GLUSTER_ACL_UNDEFINED_ID  (-1)
1290
1291 #define GLUSTER_ACL_HEADER_SIZE    4
1292 #define GLUSTER_ACL_ENTRY_SIZE     8
1293
1294 #define GLUSTER_ACL_SIZE(n)       (GLUSTER_ACL_HEADER_SIZE + (n * GLUSTER_ACL_ENTRY_SIZE))
1295
1296 static SMB_ACL_T mode_to_smb_acls(const struct stat *mode, TALLOC_CTX *mem_ctx)
1297 {
1298         struct smb_acl_t *result;
1299         int count;
1300
1301         count = 3;
1302         result = sys_acl_init(mem_ctx);
1303         if (!result) {
1304                 errno = ENOMEM;
1305                 return NULL;
1306         }
1307
1308         result->acl = talloc_array(result, struct smb_acl_entry, count);
1309         if (!result->acl) {
1310                 errno = ENOMEM;
1311                 talloc_free(result);
1312                 return NULL;
1313         }
1314
1315         result->count = count;
1316
1317         result->acl[0].a_type = SMB_ACL_USER_OBJ;
1318         result->acl[0].a_perm = (mode->st_mode & S_IRWXU) >> 6;;
1319
1320         result->acl[1].a_type = SMB_ACL_GROUP_OBJ;
1321         result->acl[1].a_perm = (mode->st_mode & S_IRWXG) >> 3;;
1322
1323         result->acl[2].a_type = SMB_ACL_OTHER;
1324         result->acl[2].a_perm = mode->st_mode & S_IRWXO;;
1325
1326         return result;
1327 }
1328
1329 static SMB_ACL_T gluster_to_smb_acl(const char *buf, size_t xattr_size,
1330                                     TALLOC_CTX *mem_ctx)
1331 {
1332         int count;
1333         size_t size;
1334         struct smb_acl_entry *smb_ace;
1335         struct smb_acl_t *result;
1336         int i;
1337         int offset;
1338         uint16_t tag;
1339         uint16_t perm;
1340         uint32_t id;
1341
1342         size = xattr_size;
1343
1344         if (size < GLUSTER_ACL_HEADER_SIZE) {
1345                 /* ACL should be at least as big as the header (4 bytes) */
1346                 errno = EINVAL;
1347                 return NULL;
1348         }
1349
1350         size -= GLUSTER_ACL_HEADER_SIZE; /* size of header = 4 bytes */
1351
1352         if (size % GLUSTER_ACL_ENTRY_SIZE) {
1353                 /* Size of entries must strictly be a multiple of
1354                    size of an ACE (8 bytes)
1355                 */
1356                 errno = EINVAL;
1357                 return NULL;
1358         }
1359
1360         count = size / GLUSTER_ACL_ENTRY_SIZE;
1361
1362         /* Version is the first 4 bytes of the ACL */
1363         if (IVAL(buf, 0) != GLUSTER_ACL_VERSION) {
1364                 DEBUG(0, ("Unknown gluster ACL version: %d\n",
1365                           IVAL(buf, 0)));
1366                 return NULL;
1367         }
1368         offset = GLUSTER_ACL_HEADER_SIZE;
1369
1370         result = sys_acl_init(mem_ctx);
1371         if (!result) {
1372                 errno = ENOMEM;
1373                 return NULL;
1374         }
1375
1376         result->acl = talloc_array(result, struct smb_acl_entry, count);
1377         if (!result->acl) {
1378                 errno = ENOMEM;
1379                 talloc_free(result);
1380                 return NULL;
1381         }
1382
1383         result->count = count;
1384
1385         smb_ace = result->acl;
1386
1387         for (i = 0; i < count; i++) {
1388                 /* TAG is the first 2 bytes of an entry */
1389                 tag = SVAL(buf, offset);
1390                 offset += 2;
1391
1392                 /* PERM is the next 2 bytes of an entry */
1393                 perm = SVAL(buf, offset);
1394                 offset += 2;
1395
1396                 /* ID is the last 4 bytes of an entry */
1397                 id = IVAL(buf, offset);
1398                 offset += 4;
1399
1400                 switch(tag) {
1401                 case GLUSTER_ACL_USER:
1402                         smb_ace->a_type = SMB_ACL_USER;
1403                         break;
1404                 case GLUSTER_ACL_USER_OBJ:
1405                         smb_ace->a_type = SMB_ACL_USER_OBJ;
1406                         break;
1407                 case GLUSTER_ACL_GROUP:
1408                         smb_ace->a_type = SMB_ACL_GROUP;
1409                         break;
1410                 case GLUSTER_ACL_GROUP_OBJ:
1411                         smb_ace->a_type = SMB_ACL_GROUP_OBJ;
1412                         break;
1413                 case GLUSTER_ACL_OTHER:
1414                         smb_ace->a_type = SMB_ACL_OTHER;
1415                         break;
1416                 case GLUSTER_ACL_MASK:
1417                         smb_ace->a_type = SMB_ACL_MASK;
1418                         break;
1419                 default:
1420                         DEBUG(0, ("unknown tag type %d\n", (unsigned int) tag));
1421                         return NULL;
1422                 }
1423
1424
1425                 switch(smb_ace->a_type) {
1426                 case SMB_ACL_USER:
1427                         smb_ace->info.user.uid = id;
1428                         break;
1429                 case SMB_ACL_GROUP:
1430                         smb_ace->info.group.gid = id;
1431                         break;
1432                 default:
1433                         break;
1434                 }
1435
1436                 smb_ace->a_perm = 0;
1437                 smb_ace->a_perm |=
1438                         ((perm & GLUSTER_ACL_READ) ? SMB_ACL_READ : 0);
1439                 smb_ace->a_perm |=
1440                         ((perm & GLUSTER_ACL_WRITE) ? SMB_ACL_WRITE : 0);
1441                 smb_ace->a_perm |=
1442                         ((perm & GLUSTER_ACL_EXECUTE) ? SMB_ACL_EXECUTE : 0);
1443
1444                 smb_ace++;
1445         }
1446
1447         return result;
1448 }
1449
1450
1451 static int gluster_ace_cmp(const void *left, const void *right)
1452 {
1453         int ret = 0;
1454         uint16_t tag_left, tag_right;
1455         uint32_t id_left, id_right;
1456
1457         /*
1458           Sorting precedence:
1459
1460            - Smaller TAG values must be earlier.
1461
1462            - Within same TAG, smaller identifiers must be earlier, E.g:
1463              UID 0 entry must be earlier than UID 200
1464              GID 17 entry must be earlier than GID 19
1465         */
1466
1467         /* TAG is the first element in the entry */
1468         tag_left = SVAL(left, 0);
1469         tag_right = SVAL(right, 0);
1470
1471         ret = (tag_left - tag_right);
1472         if (!ret) {
1473                 /* ID is the third element in the entry, after two short
1474                    integers (tag and perm), i.e at offset 4.
1475                 */
1476                 id_left = IVAL(left, 4);
1477                 id_right = IVAL(right, 4);
1478                 ret = id_left - id_right;
1479         }
1480
1481         return ret;
1482 }
1483
1484
1485 static ssize_t smb_to_gluster_acl(SMB_ACL_T theacl, char *buf, size_t len)
1486 {
1487         ssize_t size;
1488         struct smb_acl_entry *smb_ace;
1489         int i;
1490         int count;
1491         uint16_t tag;
1492         uint16_t perm;
1493         uint32_t id;
1494         int offset;
1495
1496         count = theacl->count;
1497
1498         size = GLUSTER_ACL_HEADER_SIZE + (count * GLUSTER_ACL_ENTRY_SIZE);
1499         if (!buf) {
1500                 return size;
1501         }
1502
1503         if (len < size) {
1504                 errno = ERANGE;
1505                 return -1;
1506         }
1507
1508         smb_ace = theacl->acl;
1509
1510         /* Version is the first 4 bytes of the ACL */
1511         SIVAL(buf, 0, GLUSTER_ACL_VERSION);
1512         offset = GLUSTER_ACL_HEADER_SIZE;
1513
1514         for (i = 0; i < count; i++) {
1515                 /* Calculate tag */
1516                 switch(smb_ace->a_type) {
1517                 case SMB_ACL_USER:
1518                         tag = GLUSTER_ACL_USER;
1519                         break;
1520                 case SMB_ACL_USER_OBJ:
1521                         tag = GLUSTER_ACL_USER_OBJ;
1522                         break;
1523                 case SMB_ACL_GROUP:
1524                         tag = GLUSTER_ACL_GROUP;
1525                         break;
1526                 case SMB_ACL_GROUP_OBJ:
1527                         tag = GLUSTER_ACL_GROUP_OBJ;
1528                         break;
1529                 case SMB_ACL_OTHER:
1530                         tag = GLUSTER_ACL_OTHER;
1531                         break;
1532                 case SMB_ACL_MASK:
1533                         tag = GLUSTER_ACL_MASK;
1534                         break;
1535                 default:
1536                         DEBUG(0, ("Unknown tag value %d\n",
1537                                   smb_ace->a_type));
1538                         errno = EINVAL;
1539                         return -1;
1540                 }
1541
1542
1543                 /* Calculate id */
1544                 switch(smb_ace->a_type) {
1545                 case SMB_ACL_USER:
1546                         id = smb_ace->info.user.uid;
1547                         break;
1548                 case SMB_ACL_GROUP:
1549                         id = smb_ace->info.group.gid;
1550                         break;
1551                 default:
1552                         id = GLUSTER_ACL_UNDEFINED_ID;
1553                         break;
1554                 }
1555
1556                 /* Calculate perm */
1557                 perm = 0;
1558
1559                 perm |=
1560                         ((smb_ace->a_perm & SMB_ACL_READ) ? GLUSTER_ACL_READ : 0);
1561                 perm |=
1562                         ((smb_ace->a_perm & SMB_ACL_WRITE) ? GLUSTER_ACL_WRITE : 0);
1563                 perm |=
1564                         ((smb_ace->a_perm & SMB_ACL_EXECUTE) ? GLUSTER_ACL_EXECUTE : 0);
1565
1566
1567                 /* TAG is the first 2 bytes of an entry */
1568                 SSVAL(buf, offset, tag);
1569                 offset += 2;
1570
1571                 /* PERM is the next 2 bytes of an entry */
1572                 SSVAL(buf, offset, perm);
1573                 offset += 2;
1574
1575                 /* ID is the last 4 bytes of an entry */
1576                 SIVAL(buf, offset, id);
1577                 offset += 4;
1578
1579                 smb_ace++;
1580         }
1581
1582         /* Skip the header, sort @count number of 8-byte entries */
1583         qsort(buf+GLUSTER_ACL_HEADER_SIZE, count, GLUSTER_ACL_ENTRY_SIZE,
1584               gluster_ace_cmp);
1585
1586         return size;
1587 }
1588
1589
1590 static SMB_ACL_T vfs_gluster_sys_acl_get_file(struct vfs_handle_struct *handle,
1591                                               const char *path_p,
1592                                               SMB_ACL_TYPE_T type,
1593                                               TALLOC_CTX *mem_ctx)
1594 {
1595         struct smb_acl_t *result;
1596         struct stat st;
1597         char *buf;
1598         const char *key;
1599         ssize_t ret, size = GLUSTER_ACL_SIZE(20);
1600
1601         switch (type) {
1602         case SMB_ACL_TYPE_ACCESS:
1603                 key = "system.posix_acl_access";
1604                 break;
1605         case SMB_ACL_TYPE_DEFAULT:
1606                 key = "system.posix_acl_default";
1607                 break;
1608         default:
1609                 errno = EINVAL;
1610                 return NULL;
1611         }
1612
1613         buf = alloca(size);
1614         if (!buf) {
1615                 return NULL;
1616         }
1617
1618         ret = glfs_getxattr(handle->data, path_p, key, buf, size);
1619         if (ret == -1 && errno == ERANGE) {
1620                 ret = glfs_getxattr(handle->data, path_p, key, 0, 0);
1621                 if (ret > 0) {
1622                         buf = alloca(ret);
1623                         if (!buf) {
1624                                 return NULL;
1625                         }
1626                         ret = glfs_getxattr(handle->data, path_p, key, buf, ret);
1627                 }
1628         }
1629
1630         /* retrieving the ACL from the xattr has finally failed, do a
1631          * mode-to-acl mapping */
1632
1633         if (ret == -1 && errno == ENODATA) {
1634                 ret = glfs_stat(handle->data, path_p, &st);
1635                 if (ret == 0) {
1636                         result = mode_to_smb_acls(&st, mem_ctx);
1637                         return result;
1638                 }
1639         }
1640
1641         if (ret <= 0) {
1642                 return NULL;
1643         }
1644
1645         result = gluster_to_smb_acl(buf, ret, mem_ctx);
1646
1647         return result;
1648 }
1649
1650 static SMB_ACL_T vfs_gluster_sys_acl_get_fd(struct vfs_handle_struct *handle,
1651                                             struct files_struct *fsp,
1652                                             TALLOC_CTX *mem_ctx)
1653 {
1654         struct smb_acl_t *result;
1655         struct stat st;
1656         ssize_t ret, size = GLUSTER_ACL_SIZE(20);
1657         char *buf;
1658         glfs_fd_t *glfd;
1659
1660         glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
1661
1662         buf = alloca(size);
1663         if (!buf) {
1664                 return NULL;
1665         }
1666
1667         ret = glfs_fgetxattr(glfd, "system.posix_acl_access", buf, size);
1668         if (ret == -1 && errno == ERANGE) {
1669                 ret = glfs_fgetxattr(glfd, "system.posix_acl_access", 0, 0);
1670                 if (ret > 0) {
1671                         buf = alloca(ret);
1672                         if (!buf) {
1673                                 return NULL;
1674                         }
1675                         ret = glfs_fgetxattr(glfd, "system.posix_acl_access",
1676                                              buf, ret);
1677                 }
1678         }
1679
1680         /* retrieving the ACL from the xattr has finally failed, do a
1681          * mode-to-acl mapping */
1682
1683         if (ret == -1 && errno == ENODATA) {
1684                 ret = glfs_fstat(glfd, &st);
1685                 if (ret == 0) {
1686                         result = mode_to_smb_acls(&st, mem_ctx);
1687                         return result;
1688                 }
1689         }
1690
1691         if (ret <= 0) {
1692                 return NULL;
1693         }
1694
1695         result = gluster_to_smb_acl(buf, ret, mem_ctx);
1696
1697         return result;
1698 }
1699
1700 static int vfs_gluster_sys_acl_set_file(struct vfs_handle_struct *handle,
1701                                         const char *name,
1702                                         SMB_ACL_TYPE_T acltype,
1703                                         SMB_ACL_T theacl)
1704 {
1705         int ret;
1706         const char *key;
1707         char *buf;
1708         ssize_t size;
1709
1710         switch (acltype) {
1711         case SMB_ACL_TYPE_ACCESS:
1712                 key = "system.posix_acl_access";
1713                 break;
1714         case SMB_ACL_TYPE_DEFAULT:
1715                 key = "system.posix_acl_default";
1716                 break;
1717         default:
1718                 errno = EINVAL;
1719                 return -1;
1720         }
1721
1722         size = smb_to_gluster_acl(theacl, 0, 0);
1723         buf = alloca(size);
1724
1725         size = smb_to_gluster_acl(theacl, buf, size);
1726         if (size == -1) {
1727                 return -1;
1728         }
1729
1730         ret = glfs_setxattr(handle->data, name, key, buf, size, 0);
1731
1732         return ret;
1733 }
1734
1735 static int vfs_gluster_sys_acl_set_fd(struct vfs_handle_struct *handle,
1736                                       struct files_struct *fsp,
1737                                       SMB_ACL_T theacl)
1738 {
1739         int ret;
1740         char *buf;
1741         ssize_t size;
1742
1743         size = smb_to_gluster_acl(theacl, 0, 0);
1744         buf = alloca(size);
1745
1746         size = smb_to_gluster_acl(theacl, buf, size);
1747         if (size == -1) {
1748                 return -1;
1749         }
1750
1751         ret = glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp),
1752                              "system.posix_acl_access", buf, size, 0);
1753         return ret;
1754 }
1755
1756 static int vfs_gluster_sys_acl_delete_def_file(struct vfs_handle_struct *handle,
1757                                                const char *path)
1758 {
1759         return glfs_removexattr(handle->data, path, "system.posix_acl_default");
1760 }
1761
1762 static struct vfs_fn_pointers glusterfs_fns = {
1763
1764         /* Disk Operations */
1765
1766         .connect_fn = vfs_gluster_connect,
1767         .disconnect_fn = vfs_gluster_disconnect,
1768         .disk_free_fn = vfs_gluster_disk_free,
1769         .get_quota_fn = vfs_gluster_get_quota,
1770         .set_quota_fn = vfs_gluster_set_quota,
1771         .statvfs_fn = vfs_gluster_statvfs,
1772         .fs_capabilities_fn = vfs_gluster_fs_capabilities,
1773
1774         .get_dfs_referrals_fn = NULL,
1775
1776         /* Directory Operations */
1777
1778         .opendir_fn = vfs_gluster_opendir,
1779         .fdopendir_fn = vfs_gluster_fdopendir,
1780         .readdir_fn = vfs_gluster_readdir,
1781         .seekdir_fn = vfs_gluster_seekdir,
1782         .telldir_fn = vfs_gluster_telldir,
1783         .rewind_dir_fn = vfs_gluster_rewinddir,
1784         .mkdir_fn = vfs_gluster_mkdir,
1785         .rmdir_fn = vfs_gluster_rmdir,
1786         .closedir_fn = vfs_gluster_closedir,
1787         .init_search_op_fn = vfs_gluster_init_search_op,
1788
1789         /* File Operations */
1790
1791         .open_fn = vfs_gluster_open,
1792         .create_file_fn = NULL,
1793         .close_fn = vfs_gluster_close,
1794         .read_fn = vfs_gluster_read,
1795         .pread_fn = vfs_gluster_pread,
1796         .pread_send_fn = vfs_gluster_pread_send,
1797         .pread_recv_fn = vfs_gluster_recv,
1798         .write_fn = vfs_gluster_write,
1799         .pwrite_fn = vfs_gluster_pwrite,
1800         .pwrite_send_fn = vfs_gluster_pwrite_send,
1801         .pwrite_recv_fn = vfs_gluster_recv,
1802         .lseek_fn = vfs_gluster_lseek,
1803         .sendfile_fn = vfs_gluster_sendfile,
1804         .recvfile_fn = vfs_gluster_recvfile,
1805         .rename_fn = vfs_gluster_rename,
1806         .fsync_fn = vfs_gluster_fsync,
1807         .fsync_send_fn = vfs_gluster_fsync_send,
1808         .fsync_recv_fn = vfs_gluster_fsync_recv,
1809
1810         .stat_fn = vfs_gluster_stat,
1811         .fstat_fn = vfs_gluster_fstat,
1812         .lstat_fn = vfs_gluster_lstat,
1813         .get_alloc_size_fn = vfs_gluster_get_alloc_size,
1814         .unlink_fn = vfs_gluster_unlink,
1815
1816         .chmod_fn = vfs_gluster_chmod,
1817         .fchmod_fn = vfs_gluster_fchmod,
1818         .chown_fn = vfs_gluster_chown,
1819         .fchown_fn = vfs_gluster_fchown,
1820         .lchown_fn = vfs_gluster_lchown,
1821         .chdir_fn = vfs_gluster_chdir,
1822         .getwd_fn = vfs_gluster_getwd,
1823         .ntimes_fn = vfs_gluster_ntimes,
1824         .ftruncate_fn = vfs_gluster_ftruncate,
1825         .fallocate_fn = vfs_gluster_fallocate,
1826         .lock_fn = vfs_gluster_lock,
1827         .kernel_flock_fn = vfs_gluster_kernel_flock,
1828         .linux_setlease_fn = vfs_gluster_linux_setlease,
1829         .getlock_fn = vfs_gluster_getlock,
1830         .symlink_fn = vfs_gluster_symlink,
1831         .readlink_fn = vfs_gluster_readlink,
1832         .link_fn = vfs_gluster_link,
1833         .mknod_fn = vfs_gluster_mknod,
1834         .realpath_fn = vfs_gluster_realpath,
1835         .chflags_fn = vfs_gluster_chflags,
1836         .file_id_create_fn = NULL,
1837         .copy_chunk_send_fn = NULL,
1838         .copy_chunk_recv_fn = NULL,
1839         .streaminfo_fn = NULL,
1840         .get_real_filename_fn = vfs_gluster_get_real_filename,
1841         .connectpath_fn = vfs_gluster_connectpath,
1842
1843         .brl_lock_windows_fn = NULL,
1844         .brl_unlock_windows_fn = NULL,
1845         .brl_cancel_windows_fn = NULL,
1846         .strict_lock_fn = NULL,
1847         .strict_unlock_fn = NULL,
1848         .translate_name_fn = NULL,
1849         .fsctl_fn = NULL,
1850
1851         /* NT ACL Operations */
1852         .fget_nt_acl_fn = NULL,
1853         .get_nt_acl_fn = NULL,
1854         .fset_nt_acl_fn = NULL,
1855         .audit_file_fn = NULL,
1856
1857         /* Posix ACL Operations */
1858         .chmod_acl_fn = NULL,   /* passthrough to default */
1859         .fchmod_acl_fn = NULL,  /* passthrough to default */
1860         .sys_acl_get_file_fn = vfs_gluster_sys_acl_get_file,
1861         .sys_acl_get_fd_fn = vfs_gluster_sys_acl_get_fd,
1862         .sys_acl_blob_get_file_fn = posix_sys_acl_blob_get_file,
1863         .sys_acl_blob_get_fd_fn = posix_sys_acl_blob_get_fd,
1864         .sys_acl_set_file_fn = vfs_gluster_sys_acl_set_file,
1865         .sys_acl_set_fd_fn = vfs_gluster_sys_acl_set_fd,
1866         .sys_acl_delete_def_file_fn = vfs_gluster_sys_acl_delete_def_file,
1867
1868         /* EA Operations */
1869         .getxattr_fn = vfs_gluster_getxattr,
1870         .fgetxattr_fn = vfs_gluster_fgetxattr,
1871         .listxattr_fn = vfs_gluster_listxattr,
1872         .flistxattr_fn = vfs_gluster_flistxattr,
1873         .removexattr_fn = vfs_gluster_removexattr,
1874         .fremovexattr_fn = vfs_gluster_fremovexattr,
1875         .setxattr_fn = vfs_gluster_setxattr,
1876         .fsetxattr_fn = vfs_gluster_fsetxattr,
1877
1878         /* AIO Operations */
1879         .aio_force_fn = vfs_gluster_aio_force,
1880
1881         /* Offline Operations */
1882         .is_offline_fn = vfs_gluster_is_offline,
1883         .set_offline_fn = vfs_gluster_set_offline,
1884
1885         /* Durable handle Operations */
1886         .durable_cookie_fn = NULL,
1887         .durable_disconnect_fn = NULL,
1888         .durable_reconnect_fn = NULL,
1889 };
1890
1891 NTSTATUS vfs_glusterfs_init(void);
1892 NTSTATUS vfs_glusterfs_init(void)
1893 {
1894         return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
1895                                 "glusterfs", &glusterfs_fns);
1896 }