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                                  const char *path,
 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 struct smb_filename *smb_fname,
 359                                 const char *mask,
 360                                 uint32_t attributes)
 361 {
 362         glfs_fd_t *fd;
 363
 364         fd = glfs_opendir(handle->data, smb_fname->base_name);
 365         if (fd == NULL) {
 366                 DEBUG(0, ("glfs_opendir(%s) failed: %s\n",
 367                           smb_fname->base_name, strerror(errno)));
 368         }
 369
 370         return (DIR *) fd;
 371 }
 372
 373 static DIR *vfs_gluster_fdopendir(struct vfs_handle_struct *handle,
 374                                   files_struct *fsp, const char *mask,
 375                                   uint32_t attributes)
 376 {
 377         return (DIR *) *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
 378 }
 379
 380 static int vfs_gluster_closedir(struct vfs_handle_struct *handle, DIR *dirp)
 381 {
 382         return glfs_closedir((void *)dirp);
 383 }
 384
 385 static struct dirent *vfs_gluster_readdir(struct vfs_handle_struct *handle,
 386                                           DIR *dirp, SMB_STRUCT_STAT *sbuf)
 387 {
 388         static char direntbuf[512];
 389         int ret;
 390         struct stat stat;
 391         struct dirent *dirent = 0;
 392
 393         if (sbuf != NULL) {
 394                 ret = glfs_readdirplus_r((void *)dirp, &stat, (void *)direntbuf,
 395                                          &dirent);
 396         } else {
 397                 ret = glfs_readdir_r((void *)dirp, (void *)direntbuf, &dirent);
 398         }
 399
 400         if ((ret < 0) || (dirent == NULL)) {
 401                 return NULL;
 402         }
 403
 404         if (sbuf != NULL) {
 405                 smb_stat_ex_from_stat(sbuf, &stat);
 406         }
 407
 408         return dirent;
 409 }
 410
 411 static long vfs_gluster_telldir(struct vfs_handle_struct *handle, DIR *dirp)
 412 {
 413         return glfs_telldir((void *)dirp);
 414 }
 415
 416 static void vfs_gluster_seekdir(struct vfs_handle_struct *handle, DIR *dirp,
 417                                 long offset)
 418 {
 419         glfs_seekdir((void *)dirp, offset);
 420 }
 421
 422 static void vfs_gluster_rewinddir(struct vfs_handle_struct *handle, DIR *dirp)
 423 {
 424         glfs_seekdir((void *)dirp, 0);
 425 }
 426
 427 static void vfs_gluster_init_search_op(struct vfs_handle_struct *handle,
 428                                        DIR *dirp)
 429 {
 430         return;
 431 }
 432
 433 static int vfs_gluster_mkdir(struct vfs_handle_struct *handle,
 434                              const struct smb_filename *smb_fname,
 435                              mode_t mode)
 436 {
 437         return glfs_mkdir(handle->data, smb_fname->base_name, mode);
 438 }
 439
 440 static int vfs_gluster_rmdir(struct vfs_handle_struct *handle,
 441                         const struct smb_filename *smb_fname)
 442 {
 443         return glfs_rmdir(handle->data, smb_fname->base_name);
 444 }
 445
 446 static int vfs_gluster_open(struct vfs_handle_struct *handle,
 447                             struct smb_filename *smb_fname, files_struct *fsp,
 448                             int flags, mode_t mode)
 449 {
 450         glfs_fd_t *glfd;
 451         glfs_fd_t **p_tmp;
 452
 453         if (flags & O_DIRECTORY) {
 454                 glfd = glfs_opendir(handle->data, smb_fname->base_name);
 455         } else if (flags & O_CREAT) {
 456                 glfd = glfs_creat(handle->data, smb_fname->base_name, flags,
 457                                   mode);
 458         } else {
 459                 glfd = glfs_open(handle->data, smb_fname->base_name, flags);
 460         }
 461
 462         if (glfd == NULL) {
 463                 return -1;
 464         }
 465         p_tmp = (glfs_fd_t **)VFS_ADD_FSP_EXTENSION(handle, fsp,
 466                                                           glfs_fd_t *, NULL);
 467         *p_tmp = glfd;
 468         /* An arbitrary value for error reporting, so you know its us. */
 469         return 13371337;
 470 }
 471
 472 static int vfs_gluster_close(struct vfs_handle_struct *handle,
 473                              files_struct *fsp)
 474 {
 475         glfs_fd_t *glfd;
 476         glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
 477         VFS_REMOVE_FSP_EXTENSION(handle, fsp);
 478         return glfs_close(glfd);
 479 }
 480
 481 static ssize_t vfs_gluster_read(struct vfs_handle_struct *handle,
 482                                 files_struct *fsp, void *data, size_t n)
 483 {
 484         return glfs_read(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
 485 }
 486
 487 static ssize_t vfs_gluster_pread(struct vfs_handle_struct *handle,
 488                                  files_struct *fsp, void *data, size_t n,
 489                                  off_t offset)
 490 {
 491         return glfs_pread(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
 492 }
 493
 494 struct glusterfs_aio_state;
 495
 496 struct glusterfs_aio_wrapper {
 497         struct glusterfs_aio_state *state;
 498 };
 499
 500 struct glusterfs_aio_state {
 501         ssize_t ret;
 502         struct tevent_req *req;
 503         bool cancelled;
 504         struct vfs_aio_state vfs_aio_state;
 505         struct timespec start;
 506 };
 507
 508 static int aio_wrapper_destructor(struct glusterfs_aio_wrapper *wrap)
 509 {
 510         wrap->state->cancelled = true;
 511
 512         return 0;
 513 }
 514
 515 /*
 516  * This function is the callback that will be called on glusterfs
 517  * threads once the async IO submitted is complete. To notify
 518  * Samba of the completion we use a pipe based queue.
 519  */
 520 static void aio_glusterfs_done(glfs_fd_t *fd, ssize_t ret, void *data)
 521 {
 522         struct glusterfs_aio_state *state = NULL;
 523         int sts = 0;
 524         struct timespec end;
 525
 526         state = (struct glusterfs_aio_state *)data;
 527
 528         clock_gettime_mono(&end);
 529
 530         if (ret < 0) {
 531                 state->ret = -1;
 532                 state->vfs_aio_state.error = errno;
 533         } else {
 534                 state->ret = ret;
 535         }
 536         state->vfs_aio_state.duration = nsec_time_diff(&end, &state->start);
 537
 538         /*
 539          * Write the state pointer to glusterfs_aio_state to the
 540          * pipe, so we can call tevent_req_done() from the main thread,
 541          * because tevent_req_done() is not designed to be executed in
 542          * the multithread environment, so tevent_req_done() must be
 543          * executed from the smbd main thread.
 544          *
 545          * write(2) on pipes with sizes under _POSIX_PIPE_BUF
 546          * in size is atomic, without this, the use op pipes in this
 547          * code would not work.
 548          *
 549          * sys_write is a thin enough wrapper around write(2)
 550          * that we can trust it here.
 551          */
 552
 553         sts = sys_write(write_fd, &state, sizeof(struct glusterfs_aio_state *));
 554         if (sts < 0) {
 555                 DEBUG(0,("\nWrite to pipe failed (%s)", strerror(errno)));
 556         }
 557
 558         return;
 559 }
 560
 561 /*
 562  * Read each req off the pipe and process it.
 563  */
 564 static void aio_tevent_fd_done(struct tevent_context *event_ctx,
 565                                 struct tevent_fd *fde,
 566                                 uint16_t flags, void *data)
 567 {
 568         struct tevent_req *req = NULL;
 569         struct glusterfs_aio_state *state = NULL;
 570         int sts = 0;
 571
 572         /*
 573          * read(2) on pipes is atomic if the needed data is available
 574          * in the pipe, per SUS and POSIX.  Because we always write
 575          * to the pipe in sizeof(struct tevent_req *) chunks, we can
 576          * always read in those chunks, atomically.
 577          *
 578          * sys_read is a thin enough wrapper around read(2) that we
 579          * can trust it here.
 580          */
 581
 582         sts = sys_read(read_fd, &state, sizeof(struct glusterfs_aio_state *));
 583
 584         if (sts < 0) {
 585                 DEBUG(0,("\nRead from pipe failed (%s)", strerror(errno)));
 586         }
 587
 588         /* if we've cancelled the op, there is no req, so just clean up. */
 589         if (state->cancelled == true) {
 590                 TALLOC_FREE(state);
 591                 return;
 592         }
 593
 594         req = state->req;
 595
 596         if (req) {
 597                 tevent_req_done(req);
 598         }
 599         return;
 600 }
 601
 602 static bool init_gluster_aio(struct vfs_handle_struct *handle)
 603 {
 604         int fds[2];
 605         int ret = -1;
 606
 607         if (read_fd != -1) {
 608                 /*
 609                  * Already initialized.
 610                  */
 611                 return true;
 612         }
 613
 614         ret = pipe(fds);
 615         if (ret == -1) {
 616                 goto fail;
 617         }
 618
 619         read_fd = fds[0];
 620         write_fd = fds[1];
 621
 622         aio_read_event = tevent_add_fd(handle->conn->sconn->ev_ctx,
 623                                         NULL,
 624                                         read_fd,
 625                                         TEVENT_FD_READ,
 626                                         aio_tevent_fd_done,
 627                                         NULL);
 628         if (aio_read_event == NULL) {
 629                 goto fail;
 630         }
 631
 632         return true;
 633 fail:
 634         TALLOC_FREE(aio_read_event);
 635         if (read_fd != -1) {
 636                 close(read_fd);
 637                 close(write_fd);
 638                 read_fd = -1;
 639                 write_fd = -1;
 640         }
 641         return false;
 642 }
 643
 644 static struct glusterfs_aio_state *aio_state_create(TALLOC_CTX *mem_ctx)
 645 {
 646         struct tevent_req *req = NULL;
 647         struct glusterfs_aio_state *state = NULL;
 648         struct glusterfs_aio_wrapper *wrapper = NULL;
 649
 650         req = tevent_req_create(mem_ctx, &wrapper, struct glusterfs_aio_wrapper);
 651
 652         if (req == NULL) {
 653                 return NULL;
 654         }
 655
 656         state = talloc_zero(NULL, struct glusterfs_aio_state);
 657
 658         if (state == NULL) {
 659                 TALLOC_FREE(req);
 660                 return NULL;
 661         }
 662
 663         talloc_set_destructor(wrapper, aio_wrapper_destructor);
 664         state->cancelled = false;
 665         state->req = req;
 666
 667         wrapper->state = state;
 668
 669         return state;
 670 }
 671
 672 static struct tevent_req *vfs_gluster_pread_send(struct vfs_handle_struct
 673                                                   *handle, TALLOC_CTX *mem_ctx,
 674                                                   struct tevent_context *ev,
 675                                                   files_struct *fsp,
 676                                                   void *data, size_t n,
 677                                                   off_t offset)
 678 {
 679         struct glusterfs_aio_state *state = NULL;
 680         struct tevent_req *req = NULL;
 681         int ret = 0;
 682
 683         state = aio_state_create(mem_ctx);
 684
 685         if (state == NULL) {
 686                 return NULL;
 687         }
 688
 689         req = state->req;
 690
 691         if (!init_gluster_aio(handle)) {
 692                 tevent_req_error(req, EIO);
 693                 return tevent_req_post(req, ev);
 694         }
 695
 696         clock_gettime_mono(&state->start);
 697         ret = glfs_pread_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 698                                 fsp), data, n, offset, 0, aio_glusterfs_done,
 699                                 state);
 700         if (ret < 0) {
 701                 tevent_req_error(req, -ret);
 702                 return tevent_req_post(req, ev);
 703         }
 704
 705         return req;
 706 }
 707
 708 static struct tevent_req *vfs_gluster_pwrite_send(struct vfs_handle_struct
 709                                                   *handle, TALLOC_CTX *mem_ctx,
 710                                                   struct tevent_context *ev,
 711                                                   files_struct *fsp,
 712                                                   const void *data, size_t n,
 713                                                   off_t offset)
 714 {
 715         struct glusterfs_aio_state *state = NULL;
 716         struct tevent_req *req = NULL;
 717         int ret = 0;
 718
 719         state = aio_state_create(mem_ctx);
 720
 721         if (state == NULL) {
 722                 return NULL;
 723         }
 724
 725         req = state->req;
 726
 727         if (!init_gluster_aio(handle)) {
 728                 tevent_req_error(req, EIO);
 729                 return tevent_req_post(req, ev);
 730         }
 731
 732         clock_gettime_mono(&state->start);
 733         ret = glfs_pwrite_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 734                                 fsp), data, n, offset, 0, aio_glusterfs_done,
 735                                 state);
 736         if (ret < 0) {
 737                 tevent_req_error(req, -ret);
 738                 return tevent_req_post(req, ev);
 739         }
 740
 741         return req;
 742 }
 743
 744 static ssize_t vfs_gluster_recv(struct tevent_req *req,
 745                                 struct vfs_aio_state *vfs_aio_state)
 746 {
 747         struct glusterfs_aio_state *state = NULL;
 748         struct glusterfs_aio_wrapper *wrapper = NULL;
 749         int ret = 0;
 750
 751         wrapper = tevent_req_data(req, struct glusterfs_aio_wrapper);
 752
 753         if (wrapper == NULL) {
 754                 return -1;
 755         }
 756
 757         state = wrapper->state;
 758
 759         if (state == NULL) {
 760                 return -1;
 761         }
 762
 763         if (tevent_req_is_unix_error(req, &vfs_aio_state->error)) {
 764                 return -1;
 765         }
 766
 767         *vfs_aio_state = state->vfs_aio_state;
 768         ret = state->ret;
 769
 770         /* Clean up the state, it is in a NULL context. */
 771
 772         TALLOC_FREE(state);
 773
 774         return ret;
 775 }
 776
 777 static ssize_t vfs_gluster_write(struct vfs_handle_struct *handle,
 778                                  files_struct *fsp, const void *data, size_t n)
 779 {
 780         return glfs_write(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, 0);
 781 }
 782
 783 static ssize_t vfs_gluster_pwrite(struct vfs_handle_struct *handle,
 784                                   files_struct *fsp, const void *data,
 785                                   size_t n, off_t offset)
 786 {
 787         return glfs_pwrite(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), data, n, offset, 0);
 788 }
 789
 790 static off_t vfs_gluster_lseek(struct vfs_handle_struct *handle,
 791                                files_struct *fsp, off_t offset, int whence)
 792 {
 793         return glfs_lseek(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset, whence);
 794 }
 795
 796 static ssize_t vfs_gluster_sendfile(struct vfs_handle_struct *handle, int tofd,
 797                                     files_struct *fromfsp,
 798                                     const DATA_BLOB *hdr,
 799                                     off_t offset, size_t n)
 800 {
 801         errno = ENOTSUP;
 802         return -1;
 803 }
 804
 805 static ssize_t vfs_gluster_recvfile(struct vfs_handle_struct *handle,
 806                                     int fromfd, files_struct *tofsp,
 807                                     off_t offset, size_t n)
 808 {
 809         errno = ENOTSUP;
 810         return -1;
 811 }
 812
 813 static int vfs_gluster_rename(struct vfs_handle_struct *handle,
 814                               const struct smb_filename *smb_fname_src,
 815                               const struct smb_filename *smb_fname_dst)
 816 {
 817         return glfs_rename(handle->data, smb_fname_src->base_name,
 818                            smb_fname_dst->base_name);
 819 }
 820
 821 static int vfs_gluster_fsync(struct vfs_handle_struct *handle,
 822                              files_struct *fsp)
 823 {
 824         return glfs_fsync(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp));
 825 }
 826
 827 static struct tevent_req *vfs_gluster_fsync_send(struct vfs_handle_struct
 828                                                  *handle, TALLOC_CTX *mem_ctx,
 829                                                  struct tevent_context *ev,
 830                                                  files_struct *fsp)
 831 {
 832         struct tevent_req *req = NULL;
 833         struct glusterfs_aio_state *state = NULL;
 834         int ret = 0;
 835
 836         state = aio_state_create(mem_ctx);
 837
 838         if (state == NULL) {
 839                 return NULL;
 840         }
 841
 842         req = state->req;
 843
 844         if (!init_gluster_aio(handle)) {
 845                 tevent_req_error(req, EIO);
 846                 return tevent_req_post(req, ev);
 847         }
 848
 849         clock_gettime_mono(&state->start);
 850         ret = glfs_fsync_async(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle,
 851                                 fsp), aio_glusterfs_done, req);
 852         if (ret < 0) {
 853                 tevent_req_error(req, -ret);
 854                 return tevent_req_post(req, ev);
 855         }
 856         return req;
 857 }
 858
 859 static int vfs_gluster_fsync_recv(struct tevent_req *req,
 860                                   struct vfs_aio_state *vfs_aio_state)
 861 {
 862         /*
 863          * Use implicit conversion ssize_t->int
 864          */
 865         return vfs_gluster_recv(req, vfs_aio_state);
 866 }
 867
 868 static int vfs_gluster_stat(struct vfs_handle_struct *handle,
 869                             struct smb_filename *smb_fname)
 870 {
 871         struct stat st;
 872         int ret;
 873
 874         ret = glfs_stat(handle->data, smb_fname->base_name, &st);
 875         if (ret == 0) {
 876                 smb_stat_ex_from_stat(&smb_fname->st, &st);
 877         }
 878         if (ret < 0 && errno != ENOENT) {
 879                 DEBUG(0, ("glfs_stat(%s) failed: %s\n",
 880                           smb_fname->base_name, strerror(errno)));
 881         }
 882         return ret;
 883 }
 884
 885 static int vfs_gluster_fstat(struct vfs_handle_struct *handle,
 886                              files_struct *fsp, SMB_STRUCT_STAT *sbuf)
 887 {
 888         struct stat st;
 889         int ret;
 890
 891         ret = glfs_fstat(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), &st);
 892         if (ret == 0) {
 893                 smb_stat_ex_from_stat(sbuf, &st);
 894         }
 895         if (ret < 0) {
 896                 DEBUG(0, ("glfs_fstat(%d) failed: %s\n",
 897                           fsp->fh->fd, strerror(errno)));
 898         }
 899         return ret;
 900 }
 901
 902 static int vfs_gluster_lstat(struct vfs_handle_struct *handle,
 903                              struct smb_filename *smb_fname)
 904 {
 905         struct stat st;
 906         int ret;
 907
 908         ret = glfs_lstat(handle->data, smb_fname->base_name, &st);
 909         if (ret == 0) {
 910                 smb_stat_ex_from_stat(&smb_fname->st, &st);
 911         }
 912         if (ret < 0 && errno != ENOENT) {
 913                 DEBUG(0, ("glfs_lstat(%s) failed: %s\n",
 914                           smb_fname->base_name, strerror(errno)));
 915         }
 916         return ret;
 917 }
 918
 919 static uint64_t vfs_gluster_get_alloc_size(struct vfs_handle_struct *handle,
 920                                            files_struct *fsp,
 921                                            const SMB_STRUCT_STAT *sbuf)
 922 {
 923         return sbuf->st_ex_blocks * 512;
 924 }
 925
 926 static int vfs_gluster_unlink(struct vfs_handle_struct *handle,
 927                               const struct smb_filename *smb_fname)
 928 {
 929         return glfs_unlink(handle->data, smb_fname->base_name);
 930 }
 931
 932 static int vfs_gluster_chmod(struct vfs_handle_struct *handle,
 933                                 const struct smb_filename *smb_fname,
 934                                 mode_t mode)
 935 {
 936         return glfs_chmod(handle->data, smb_fname->base_name, mode);
 937 }
 938
 939 static int vfs_gluster_fchmod(struct vfs_handle_struct *handle,
 940                               files_struct *fsp, mode_t mode)
 941 {
 942         return glfs_fchmod(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), mode);
 943 }
 944
 945 static int vfs_gluster_chown(struct vfs_handle_struct *handle,
 946                              const char *path, uid_t uid, gid_t gid)
 947 {
 948         return glfs_chown(handle->data, path, uid, gid);
 949 }
 950
 951 static int vfs_gluster_fchown(struct vfs_handle_struct *handle,
 952                               files_struct *fsp, uid_t uid, gid_t gid)
 953 {
 954         return glfs_fchown(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), uid, gid);
 955 }
 956
 957 static int vfs_gluster_lchown(struct vfs_handle_struct *handle,
 958                               const char *path, uid_t uid, gid_t gid)
 959 {
 960         return glfs_lchown(handle->data, path, uid, gid);
 961 }
 962
 963 static int vfs_gluster_chdir(struct vfs_handle_struct *handle, const char *path)
 964 {
 965         return glfs_chdir(handle->data, path);
 966 }
 967
 968 static char *vfs_gluster_getwd(struct vfs_handle_struct *handle)
 969 {
 970         char *cwd;
 971         char *ret;
 972
 973         cwd = SMB_CALLOC_ARRAY(char, PATH_MAX);
 974         if (cwd == NULL) {
 975                 return NULL;
 976         }
 977
 978         ret = glfs_getcwd(handle->data, cwd, PATH_MAX - 1);
 979         if (ret == 0) {
 980                 free(cwd);
 981         }
 982         return ret;
 983 }
 984
 985 static int vfs_gluster_ntimes(struct vfs_handle_struct *handle,
 986                               const struct smb_filename *smb_fname,
 987                               struct smb_file_time *ft)
 988 {
 989         struct timespec times[2];
 990
 991         if (null_timespec(ft->atime)) {
 992                 times[0].tv_sec = smb_fname->st.st_ex_atime.tv_sec;
 993                 times[0].tv_nsec = smb_fname->st.st_ex_atime.tv_nsec;
 994         } else {
 995                 times[0].tv_sec = ft->atime.tv_sec;
 996                 times[0].tv_nsec = ft->atime.tv_nsec;
 997         }
 998
 999         if (null_timespec(ft->mtime)) {
1000                 times[1].tv_sec = smb_fname->st.st_ex_mtime.tv_sec;
1001                 times[1].tv_nsec = smb_fname->st.st_ex_mtime.tv_nsec;
1002         } else {
1003                 times[1].tv_sec = ft->mtime.tv_sec;
1004                 times[1].tv_nsec = ft->mtime.tv_nsec;
1005         }
1006
1007         if ((timespec_compare(&times[0],
1008                               &smb_fname->st.st_ex_atime) == 0) &&
1009             (timespec_compare(&times[1],
1010                               &smb_fname->st.st_ex_mtime) == 0)) {
1011                 return 0;
1012         }
1013
1014         return glfs_utimens(handle->data, smb_fname->base_name, times);
1015 }
1016
1017 static int vfs_gluster_ftruncate(struct vfs_handle_struct *handle,
1018                                  files_struct *fsp, off_t offset)
1019 {
1020         return glfs_ftruncate(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), offset);
1021 }
1022
1023 static int vfs_gluster_fallocate(struct vfs_handle_struct *handle,
1024                                  struct files_struct *fsp,
1025                                  uint32_t mode,
1026                                  off_t offset, off_t len)
1027 {
1028         /* TODO: add support using glfs_fallocate() and glfs_zerofill() */
1029         errno = ENOTSUP;
1030         return -1;
1031 }
1032
1033 static char *vfs_gluster_realpath(struct vfs_handle_struct *handle,
1034                                   const char *path)
1035 {
1036         return glfs_realpath(handle->data, path, 0);
1037 }
1038
1039 static bool vfs_gluster_lock(struct vfs_handle_struct *handle,
1040                              files_struct *fsp, int op, off_t offset,
1041                              off_t count, int type)
1042 {
1043         struct flock flock = { 0, };
1044         int ret;
1045
1046         flock.l_type = type;
1047         flock.l_whence = SEEK_SET;
1048         flock.l_start = offset;
1049         flock.l_len = count;
1050         flock.l_pid = 0;
1051
1052         ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), op, &flock);
1053
1054         if (op == F_GETLK) {
1055                 /* lock query, true if someone else has locked */
1056                 if ((ret != -1) &&
1057                     (flock.l_type != F_UNLCK) &&
1058                     (flock.l_pid != 0) && (flock.l_pid != getpid()))
1059                         return true;
1060                 /* not me */
1061                 return false;
1062         }
1063
1064         if (ret == -1) {
1065                 return false;
1066         }
1067
1068         return true;
1069 }
1070
1071 static int vfs_gluster_kernel_flock(struct vfs_handle_struct *handle,
1072                                     files_struct *fsp, uint32_t share_mode,
1073                                     uint32_t access_mask)
1074 {
1075         errno = ENOSYS;
1076         return -1;
1077 }
1078
1079 static int vfs_gluster_linux_setlease(struct vfs_handle_struct *handle,
1080                                       files_struct *fsp, int leasetype)
1081 {
1082         errno = ENOSYS;
1083         return -1;
1084 }
1085
1086 static bool vfs_gluster_getlock(struct vfs_handle_struct *handle,
1087                                 files_struct *fsp, off_t *poffset,
1088                                 off_t *pcount, int *ptype, pid_t *ppid)
1089 {
1090         struct flock flock = { 0, };
1091         int ret;
1092
1093         flock.l_type = *ptype;
1094         flock.l_whence = SEEK_SET;
1095         flock.l_start = *poffset;
1096         flock.l_len = *pcount;
1097         flock.l_pid = 0;
1098
1099         ret = glfs_posix_lock(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), F_GETLK, &flock);
1100
1101         if (ret == -1) {
1102                 return false;
1103         }
1104
1105         *ptype = flock.l_type;
1106         *poffset = flock.l_start;
1107         *pcount = flock.l_len;
1108         *ppid = flock.l_pid;
1109
1110         return true;
1111 }
1112
1113 static int vfs_gluster_symlink(struct vfs_handle_struct *handle,
1114                                const char *oldpath, const char *newpath)
1115 {
1116         return glfs_symlink(handle->data, oldpath, newpath);
1117 }
1118
1119 static int vfs_gluster_readlink(struct vfs_handle_struct *handle,
1120                                 const char *path, char *buf, size_t bufsiz)
1121 {
1122         return glfs_readlink(handle->data, path, buf, bufsiz);
1123 }
1124
1125 static int vfs_gluster_link(struct vfs_handle_struct *handle,
1126                             const char *oldpath, const char *newpath)
1127 {
1128         return glfs_link(handle->data, oldpath, newpath);
1129 }
1130
1131 static int vfs_gluster_mknod(struct vfs_handle_struct *handle, const char *path,
1132                              mode_t mode, SMB_DEV_T dev)
1133 {
1134         return glfs_mknod(handle->data, path, mode, dev);
1135 }
1136
1137 static int vfs_gluster_chflags(struct vfs_handle_struct *handle,
1138                                const char *path, unsigned int flags)
1139 {
1140         errno = ENOSYS;
1141         return -1;
1142 }
1143
1144 static int vfs_gluster_get_real_filename(struct vfs_handle_struct *handle,
1145                                          const char *path, const char *name,
1146                                          TALLOC_CTX *mem_ctx, char **found_name)
1147 {
1148         int ret;
1149         char key_buf[NAME_MAX + 64];
1150         char val_buf[NAME_MAX + 1];
1151
1152         if (strlen(name) >= NAME_MAX) {
1153                 errno = ENAMETOOLONG;
1154                 return -1;
1155         }
1156
1157         snprintf(key_buf, NAME_MAX + 64,
1158                  "glusterfs.get_real_filename:%s", name);
1159
1160         ret = glfs_getxattr(handle->data, path, key_buf, val_buf, NAME_MAX + 1);
1161         if (ret == -1) {
1162                 if (errno == ENODATA) {
1163                         errno = EOPNOTSUPP;
1164                 }
1165                 return -1;
1166         }
1167
1168         *found_name = talloc_strdup(mem_ctx, val_buf);
1169         if (found_name[0] == NULL) {
1170                 errno = ENOMEM;
1171                 return -1;
1172         }
1173         return 0;
1174 }
1175
1176 static const char *vfs_gluster_connectpath(struct vfs_handle_struct *handle,
1177                                            const char *filename)
1178 {
1179         return handle->conn->connectpath;
1180 }
1181
1182 /* EA Operations */
1183
1184 static ssize_t vfs_gluster_getxattr(struct vfs_handle_struct *handle,
1185                                     const char *path, const char *name,
1186                                     void *value, size_t size)
1187 {
1188         return glfs_getxattr(handle->data, path, name, value, size);
1189 }
1190
1191 static ssize_t vfs_gluster_fgetxattr(struct vfs_handle_struct *handle,
1192                                      files_struct *fsp, const char *name,
1193                                      void *value, size_t size)
1194 {
1195         return glfs_fgetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size);
1196 }
1197
1198 static ssize_t vfs_gluster_listxattr(struct vfs_handle_struct *handle,
1199                                      const char *path, char *list, size_t size)
1200 {
1201         return glfs_listxattr(handle->data, path, list, size);
1202 }
1203
1204 static ssize_t vfs_gluster_flistxattr(struct vfs_handle_struct *handle,
1205                                       files_struct *fsp, char *list,
1206                                       size_t size)
1207 {
1208         return glfs_flistxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), list, size);
1209 }
1210
1211 static int vfs_gluster_removexattr(struct vfs_handle_struct *handle,
1212                                    const char *path, const char *name)
1213 {
1214         return glfs_removexattr(handle->data, path, name);
1215 }
1216
1217 static int vfs_gluster_fremovexattr(struct vfs_handle_struct *handle,
1218                                     files_struct *fsp, const char *name)
1219 {
1220         return glfs_fremovexattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name);
1221 }
1222
1223 static int vfs_gluster_setxattr(struct vfs_handle_struct *handle,
1224                                 const char *path, const char *name,
1225                                 const void *value, size_t size, int flags)
1226 {
1227         return glfs_setxattr(handle->data, path, name, value, size, flags);
1228 }
1229
1230 static int vfs_gluster_fsetxattr(struct vfs_handle_struct *handle,
1231                                  files_struct *fsp, const char *name,
1232                                  const void *value, size_t size, int flags)
1233 {
1234         return glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp), name, value, size,
1235                               flags);
1236 }
1237
1238 /* AIO Operations */
1239
1240 static bool vfs_gluster_aio_force(struct vfs_handle_struct *handle,
1241                                   files_struct *fsp)
1242 {
1243         return false;
1244 }
1245
1246 /* Offline Operations */
1247
1248 static bool vfs_gluster_is_offline(struct vfs_handle_struct *handle,
1249                                    const struct smb_filename *fname,
1250                                    SMB_STRUCT_STAT *sbuf)
1251 {
1252         return false;
1253 }
1254
1255 static int vfs_gluster_set_offline(struct vfs_handle_struct *handle,
1256                                    const struct smb_filename *fname)
1257 {
1258         errno = ENOTSUP;
1259         return -1;
1260 }
1261
1262 /*
1263   Gluster ACL Format:
1264
1265   Size = 4 (header) + N * 8 (entry)
1266
1267   Offset  Size    Field (Little Endian)
1268   -------------------------------------
1269   0-3     4-byte  Version
1270
1271   4-5     2-byte  Entry-1 tag
1272   6-7     2-byte  Entry-1 perm
1273   8-11    4-byte  Entry-1 id
1274
1275   12-13   2-byte  Entry-2 tag
1276   14-15   2-byte  Entry-2 perm
1277   16-19   4-byte  Entry-2 id
1278
1279   ...
1280
1281  */
1282
1283 /* header version */
1284 #define GLUSTER_ACL_VERSION 2
1285
1286 /* perm bits */
1287 #define GLUSTER_ACL_READ    0x04
1288 #define GLUSTER_ACL_WRITE   0x02
1289 #define GLUSTER_ACL_EXECUTE 0x01
1290
1291 /* tag values */
1292 #define GLUSTER_ACL_UNDEFINED_TAG  0x00
1293 #define GLUSTER_ACL_USER_OBJ       0x01
1294 #define GLUSTER_ACL_USER           0x02
1295 #define GLUSTER_ACL_GROUP_OBJ      0x04
1296 #define GLUSTER_ACL_GROUP          0x08
1297 #define GLUSTER_ACL_MASK           0x10
1298 #define GLUSTER_ACL_OTHER          0x20
1299
1300 #define GLUSTER_ACL_UNDEFINED_ID  (-1)
1301
1302 #define GLUSTER_ACL_HEADER_SIZE    4
1303 #define GLUSTER_ACL_ENTRY_SIZE     8
1304
1305 #define GLUSTER_ACL_SIZE(n)       (GLUSTER_ACL_HEADER_SIZE + (n * GLUSTER_ACL_ENTRY_SIZE))
1306
1307 static SMB_ACL_T mode_to_smb_acls(const struct stat *mode, TALLOC_CTX *mem_ctx)
1308 {
1309         struct smb_acl_t *result;
1310         int count;
1311
1312         count = 3;
1313         result = sys_acl_init(mem_ctx);
1314         if (!result) {
1315                 errno = ENOMEM;
1316                 return NULL;
1317         }
1318
1319         result->acl = talloc_array(result, struct smb_acl_entry, count);
1320         if (!result->acl) {
1321                 errno = ENOMEM;
1322                 talloc_free(result);
1323                 return NULL;
1324         }
1325
1326         result->count = count;
1327
1328         result->acl[0].a_type = SMB_ACL_USER_OBJ;
1329         result->acl[0].a_perm = (mode->st_mode & S_IRWXU) >> 6;;
1330
1331         result->acl[1].a_type = SMB_ACL_GROUP_OBJ;
1332         result->acl[1].a_perm = (mode->st_mode & S_IRWXG) >> 3;;
1333
1334         result->acl[2].a_type = SMB_ACL_OTHER;
1335         result->acl[2].a_perm = mode->st_mode & S_IRWXO;;
1336
1337         return result;
1338 }
1339
1340 static SMB_ACL_T gluster_to_smb_acl(const char *buf, size_t xattr_size,
1341                                     TALLOC_CTX *mem_ctx)
1342 {
1343         int count;
1344         size_t size;
1345         struct smb_acl_entry *smb_ace;
1346         struct smb_acl_t *result;
1347         int i;
1348         int offset;
1349         uint16_t tag;
1350         uint16_t perm;
1351         uint32_t id;
1352
1353         size = xattr_size;
1354
1355         if (size < GLUSTER_ACL_HEADER_SIZE) {
1356                 /* ACL should be at least as big as the header (4 bytes) */
1357                 errno = EINVAL;
1358                 return NULL;
1359         }
1360
1361         size -= GLUSTER_ACL_HEADER_SIZE; /* size of header = 4 bytes */
1362
1363         if (size % GLUSTER_ACL_ENTRY_SIZE) {
1364                 /* Size of entries must strictly be a multiple of
1365                    size of an ACE (8 bytes)
1366                 */
1367                 errno = EINVAL;
1368                 return NULL;
1369         }
1370
1371         count = size / GLUSTER_ACL_ENTRY_SIZE;
1372
1373         /* Version is the first 4 bytes of the ACL */
1374         if (IVAL(buf, 0) != GLUSTER_ACL_VERSION) {
1375                 DEBUG(0, ("Unknown gluster ACL version: %d\n",
1376                           IVAL(buf, 0)));
1377                 return NULL;
1378         }
1379         offset = GLUSTER_ACL_HEADER_SIZE;
1380
1381         result = sys_acl_init(mem_ctx);
1382         if (!result) {
1383                 errno = ENOMEM;
1384                 return NULL;
1385         }
1386
1387         result->acl = talloc_array(result, struct smb_acl_entry, count);
1388         if (!result->acl) {
1389                 errno = ENOMEM;
1390                 talloc_free(result);
1391                 return NULL;
1392         }
1393
1394         result->count = count;
1395
1396         smb_ace = result->acl;
1397
1398         for (i = 0; i < count; i++) {
1399                 /* TAG is the first 2 bytes of an entry */
1400                 tag = SVAL(buf, offset);
1401                 offset += 2;
1402
1403                 /* PERM is the next 2 bytes of an entry */
1404                 perm = SVAL(buf, offset);
1405                 offset += 2;
1406
1407                 /* ID is the last 4 bytes of an entry */
1408                 id = IVAL(buf, offset);
1409                 offset += 4;
1410
1411                 switch(tag) {
1412                 case GLUSTER_ACL_USER:
1413                         smb_ace->a_type = SMB_ACL_USER;
1414                         break;
1415                 case GLUSTER_ACL_USER_OBJ:
1416                         smb_ace->a_type = SMB_ACL_USER_OBJ;
1417                         break;
1418                 case GLUSTER_ACL_GROUP:
1419                         smb_ace->a_type = SMB_ACL_GROUP;
1420                         break;
1421                 case GLUSTER_ACL_GROUP_OBJ:
1422                         smb_ace->a_type = SMB_ACL_GROUP_OBJ;
1423                         break;
1424                 case GLUSTER_ACL_OTHER:
1425                         smb_ace->a_type = SMB_ACL_OTHER;
1426                         break;
1427                 case GLUSTER_ACL_MASK:
1428                         smb_ace->a_type = SMB_ACL_MASK;
1429                         break;
1430                 default:
1431                         DEBUG(0, ("unknown tag type %d\n", (unsigned int) tag));
1432                         return NULL;
1433                 }
1434
1435
1436                 switch(smb_ace->a_type) {
1437                 case SMB_ACL_USER:
1438                         smb_ace->info.user.uid = id;
1439                         break;
1440                 case SMB_ACL_GROUP:
1441                         smb_ace->info.group.gid = id;
1442                         break;
1443                 default:
1444                         break;
1445                 }
1446
1447                 smb_ace->a_perm = 0;
1448                 smb_ace->a_perm |=
1449                         ((perm & GLUSTER_ACL_READ) ? SMB_ACL_READ : 0);
1450                 smb_ace->a_perm |=
1451                         ((perm & GLUSTER_ACL_WRITE) ? SMB_ACL_WRITE : 0);
1452                 smb_ace->a_perm |=
1453                         ((perm & GLUSTER_ACL_EXECUTE) ? SMB_ACL_EXECUTE : 0);
1454
1455                 smb_ace++;
1456         }
1457
1458         return result;
1459 }
1460
1461
1462 static int gluster_ace_cmp(const void *left, const void *right)
1463 {
1464         int ret = 0;
1465         uint16_t tag_left, tag_right;
1466         uint32_t id_left, id_right;
1467
1468         /*
1469           Sorting precedence:
1470
1471            - Smaller TAG values must be earlier.
1472
1473            - Within same TAG, smaller identifiers must be earlier, E.g:
1474              UID 0 entry must be earlier than UID 200
1475              GID 17 entry must be earlier than GID 19
1476         */
1477
1478         /* TAG is the first element in the entry */
1479         tag_left = SVAL(left, 0);
1480         tag_right = SVAL(right, 0);
1481
1482         ret = (tag_left - tag_right);
1483         if (!ret) {
1484                 /* ID is the third element in the entry, after two short
1485                    integers (tag and perm), i.e at offset 4.
1486                 */
1487                 id_left = IVAL(left, 4);
1488                 id_right = IVAL(right, 4);
1489                 ret = id_left - id_right;
1490         }
1491
1492         return ret;
1493 }
1494
1495
1496 static ssize_t smb_to_gluster_acl(SMB_ACL_T theacl, char *buf, size_t len)
1497 {
1498         ssize_t size;
1499         struct smb_acl_entry *smb_ace;
1500         int i;
1501         int count;
1502         uint16_t tag;
1503         uint16_t perm;
1504         uint32_t id;
1505         int offset;
1506
1507         count = theacl->count;
1508
1509         size = GLUSTER_ACL_HEADER_SIZE + (count * GLUSTER_ACL_ENTRY_SIZE);
1510         if (!buf) {
1511                 return size;
1512         }
1513
1514         if (len < size) {
1515                 errno = ERANGE;
1516                 return -1;
1517         }
1518
1519         smb_ace = theacl->acl;
1520
1521         /* Version is the first 4 bytes of the ACL */
1522         SIVAL(buf, 0, GLUSTER_ACL_VERSION);
1523         offset = GLUSTER_ACL_HEADER_SIZE;
1524
1525         for (i = 0; i < count; i++) {
1526                 /* Calculate tag */
1527                 switch(smb_ace->a_type) {
1528                 case SMB_ACL_USER:
1529                         tag = GLUSTER_ACL_USER;
1530                         break;
1531                 case SMB_ACL_USER_OBJ:
1532                         tag = GLUSTER_ACL_USER_OBJ;
1533                         break;
1534                 case SMB_ACL_GROUP:
1535                         tag = GLUSTER_ACL_GROUP;
1536                         break;
1537                 case SMB_ACL_GROUP_OBJ:
1538                         tag = GLUSTER_ACL_GROUP_OBJ;
1539                         break;
1540                 case SMB_ACL_OTHER:
1541                         tag = GLUSTER_ACL_OTHER;
1542                         break;
1543                 case SMB_ACL_MASK:
1544                         tag = GLUSTER_ACL_MASK;
1545                         break;
1546                 default:
1547                         DEBUG(0, ("Unknown tag value %d\n",
1548                                   smb_ace->a_type));
1549                         errno = EINVAL;
1550                         return -1;
1551                 }
1552
1553
1554                 /* Calculate id */
1555                 switch(smb_ace->a_type) {
1556                 case SMB_ACL_USER:
1557                         id = smb_ace->info.user.uid;
1558                         break;
1559                 case SMB_ACL_GROUP:
1560                         id = smb_ace->info.group.gid;
1561                         break;
1562                 default:
1563                         id = GLUSTER_ACL_UNDEFINED_ID;
1564                         break;
1565                 }
1566
1567                 /* Calculate perm */
1568                 perm = 0;
1569
1570                 perm |=
1571                         ((smb_ace->a_perm & SMB_ACL_READ) ? GLUSTER_ACL_READ : 0);
1572                 perm |=
1573                         ((smb_ace->a_perm & SMB_ACL_WRITE) ? GLUSTER_ACL_WRITE : 0);
1574                 perm |=
1575                         ((smb_ace->a_perm & SMB_ACL_EXECUTE) ? GLUSTER_ACL_EXECUTE : 0);
1576
1577
1578                 /* TAG is the first 2 bytes of an entry */
1579                 SSVAL(buf, offset, tag);
1580                 offset += 2;
1581
1582                 /* PERM is the next 2 bytes of an entry */
1583                 SSVAL(buf, offset, perm);
1584                 offset += 2;
1585
1586                 /* ID is the last 4 bytes of an entry */
1587                 SIVAL(buf, offset, id);
1588                 offset += 4;
1589
1590                 smb_ace++;
1591         }
1592
1593         /* Skip the header, sort @count number of 8-byte entries */
1594         qsort(buf+GLUSTER_ACL_HEADER_SIZE, count, GLUSTER_ACL_ENTRY_SIZE,
1595               gluster_ace_cmp);
1596
1597         return size;
1598 }
1599
1600
1601 static SMB_ACL_T vfs_gluster_sys_acl_get_file(struct vfs_handle_struct *handle,
1602                                               const char *path_p,
1603                                               SMB_ACL_TYPE_T type,
1604                                               TALLOC_CTX *mem_ctx)
1605 {
1606         struct smb_acl_t *result;
1607         struct stat st;
1608         char *buf;
1609         const char *key;
1610         ssize_t ret, size = GLUSTER_ACL_SIZE(20);
1611
1612         switch (type) {
1613         case SMB_ACL_TYPE_ACCESS:
1614                 key = "system.posix_acl_access";
1615                 break;
1616         case SMB_ACL_TYPE_DEFAULT:
1617                 key = "system.posix_acl_default";
1618                 break;
1619         default:
1620                 errno = EINVAL;
1621                 return NULL;
1622         }
1623
1624         buf = alloca(size);
1625         if (!buf) {
1626                 return NULL;
1627         }
1628
1629         ret = glfs_getxattr(handle->data, path_p, key, buf, size);
1630         if (ret == -1 && errno == ERANGE) {
1631                 ret = glfs_getxattr(handle->data, path_p, key, 0, 0);
1632                 if (ret > 0) {
1633                         buf = alloca(ret);
1634                         if (!buf) {
1635                                 return NULL;
1636                         }
1637                         ret = glfs_getxattr(handle->data, path_p, key, buf, ret);
1638                 }
1639         }
1640
1641         /* retrieving the ACL from the xattr has finally failed, do a
1642          * mode-to-acl mapping */
1643
1644         if (ret == -1 && errno == ENODATA) {
1645                 ret = glfs_stat(handle->data, path_p, &st);
1646                 if (ret == 0) {
1647                         result = mode_to_smb_acls(&st, mem_ctx);
1648                         return result;
1649                 }
1650         }
1651
1652         if (ret <= 0) {
1653                 return NULL;
1654         }
1655
1656         result = gluster_to_smb_acl(buf, ret, mem_ctx);
1657
1658         return result;
1659 }
1660
1661 static SMB_ACL_T vfs_gluster_sys_acl_get_fd(struct vfs_handle_struct *handle,
1662                                             struct files_struct *fsp,
1663                                             TALLOC_CTX *mem_ctx)
1664 {
1665         struct smb_acl_t *result;
1666         struct stat st;
1667         ssize_t ret, size = GLUSTER_ACL_SIZE(20);
1668         char *buf;
1669         glfs_fd_t *glfd;
1670
1671         glfd = *(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp);
1672
1673         buf = alloca(size);
1674         if (!buf) {
1675                 return NULL;
1676         }
1677
1678         ret = glfs_fgetxattr(glfd, "system.posix_acl_access", buf, size);
1679         if (ret == -1 && errno == ERANGE) {
1680                 ret = glfs_fgetxattr(glfd, "system.posix_acl_access", 0, 0);
1681                 if (ret > 0) {
1682                         buf = alloca(ret);
1683                         if (!buf) {
1684                                 return NULL;
1685                         }
1686                         ret = glfs_fgetxattr(glfd, "system.posix_acl_access",
1687                                              buf, ret);
1688                 }
1689         }
1690
1691         /* retrieving the ACL from the xattr has finally failed, do a
1692          * mode-to-acl mapping */
1693
1694         if (ret == -1 && errno == ENODATA) {
1695                 ret = glfs_fstat(glfd, &st);
1696                 if (ret == 0) {
1697                         result = mode_to_smb_acls(&st, mem_ctx);
1698                         return result;
1699                 }
1700         }
1701
1702         if (ret <= 0) {
1703                 return NULL;
1704         }
1705
1706         result = gluster_to_smb_acl(buf, ret, mem_ctx);
1707
1708         return result;
1709 }
1710
1711 static int vfs_gluster_sys_acl_set_file(struct vfs_handle_struct *handle,
1712                                         const char *name,
1713                                         SMB_ACL_TYPE_T acltype,
1714                                         SMB_ACL_T theacl)
1715 {
1716         int ret;
1717         const char *key;
1718         char *buf;
1719         ssize_t size;
1720
1721         switch (acltype) {
1722         case SMB_ACL_TYPE_ACCESS:
1723                 key = "system.posix_acl_access";
1724                 break;
1725         case SMB_ACL_TYPE_DEFAULT:
1726                 key = "system.posix_acl_default";
1727                 break;
1728         default:
1729                 errno = EINVAL;
1730                 return -1;
1731         }
1732
1733         size = smb_to_gluster_acl(theacl, 0, 0);
1734         buf = alloca(size);
1735
1736         size = smb_to_gluster_acl(theacl, buf, size);
1737         if (size == -1) {
1738                 return -1;
1739         }
1740
1741         ret = glfs_setxattr(handle->data, name, key, buf, size, 0);
1742
1743         return ret;
1744 }
1745
1746 static int vfs_gluster_sys_acl_set_fd(struct vfs_handle_struct *handle,
1747                                       struct files_struct *fsp,
1748                                       SMB_ACL_T theacl)
1749 {
1750         int ret;
1751         char *buf;
1752         ssize_t size;
1753
1754         size = smb_to_gluster_acl(theacl, 0, 0);
1755         buf = alloca(size);
1756
1757         size = smb_to_gluster_acl(theacl, buf, size);
1758         if (size == -1) {
1759                 return -1;
1760         }
1761
1762         ret = glfs_fsetxattr(*(glfs_fd_t **)VFS_FETCH_FSP_EXTENSION(handle, fsp),
1763                              "system.posix_acl_access", buf, size, 0);
1764         return ret;
1765 }
1766
1767 static int vfs_gluster_sys_acl_delete_def_file(struct vfs_handle_struct *handle,
1768                                                const char *path)
1769 {
1770         return glfs_removexattr(handle->data, path, "system.posix_acl_default");
1771 }
1772
1773 static struct vfs_fn_pointers glusterfs_fns = {
1774
1775         /* Disk Operations */
1776
1777         .connect_fn = vfs_gluster_connect,
1778         .disconnect_fn = vfs_gluster_disconnect,
1779         .disk_free_fn = vfs_gluster_disk_free,
1780         .get_quota_fn = vfs_gluster_get_quota,
1781         .set_quota_fn = vfs_gluster_set_quota,
1782         .statvfs_fn = vfs_gluster_statvfs,
1783         .fs_capabilities_fn = vfs_gluster_fs_capabilities,
1784
1785         .get_dfs_referrals_fn = NULL,
1786
1787         /* Directory Operations */
1788
1789         .opendir_fn = vfs_gluster_opendir,
1790         .fdopendir_fn = vfs_gluster_fdopendir,
1791         .readdir_fn = vfs_gluster_readdir,
1792         .seekdir_fn = vfs_gluster_seekdir,
1793         .telldir_fn = vfs_gluster_telldir,
1794         .rewind_dir_fn = vfs_gluster_rewinddir,
1795         .mkdir_fn = vfs_gluster_mkdir,
1796         .rmdir_fn = vfs_gluster_rmdir,
1797         .closedir_fn = vfs_gluster_closedir,
1798         .init_search_op_fn = vfs_gluster_init_search_op,
1799
1800         /* File Operations */
1801
1802         .open_fn = vfs_gluster_open,
1803         .create_file_fn = NULL,
1804         .close_fn = vfs_gluster_close,
1805         .read_fn = vfs_gluster_read,
1806         .pread_fn = vfs_gluster_pread,
1807         .pread_send_fn = vfs_gluster_pread_send,
1808         .pread_recv_fn = vfs_gluster_recv,
1809         .write_fn = vfs_gluster_write,
1810         .pwrite_fn = vfs_gluster_pwrite,
1811         .pwrite_send_fn = vfs_gluster_pwrite_send,
1812         .pwrite_recv_fn = vfs_gluster_recv,
1813         .lseek_fn = vfs_gluster_lseek,
1814         .sendfile_fn = vfs_gluster_sendfile,
1815         .recvfile_fn = vfs_gluster_recvfile,
1816         .rename_fn = vfs_gluster_rename,
1817         .fsync_fn = vfs_gluster_fsync,
1818         .fsync_send_fn = vfs_gluster_fsync_send,
1819         .fsync_recv_fn = vfs_gluster_fsync_recv,
1820
1821         .stat_fn = vfs_gluster_stat,
1822         .fstat_fn = vfs_gluster_fstat,
1823         .lstat_fn = vfs_gluster_lstat,
1824         .get_alloc_size_fn = vfs_gluster_get_alloc_size,
1825         .unlink_fn = vfs_gluster_unlink,
1826
1827         .chmod_fn = vfs_gluster_chmod,
1828         .fchmod_fn = vfs_gluster_fchmod,
1829         .chown_fn = vfs_gluster_chown,
1830         .fchown_fn = vfs_gluster_fchown,
1831         .lchown_fn = vfs_gluster_lchown,
1832         .chdir_fn = vfs_gluster_chdir,
1833         .getwd_fn = vfs_gluster_getwd,
1834         .ntimes_fn = vfs_gluster_ntimes,
1835         .ftruncate_fn = vfs_gluster_ftruncate,
1836         .fallocate_fn = vfs_gluster_fallocate,
1837         .lock_fn = vfs_gluster_lock,
1838         .kernel_flock_fn = vfs_gluster_kernel_flock,
1839         .linux_setlease_fn = vfs_gluster_linux_setlease,
1840         .getlock_fn = vfs_gluster_getlock,
1841         .symlink_fn = vfs_gluster_symlink,
1842         .readlink_fn = vfs_gluster_readlink,
1843         .link_fn = vfs_gluster_link,
1844         .mknod_fn = vfs_gluster_mknod,
1845         .realpath_fn = vfs_gluster_realpath,
1846         .chflags_fn = vfs_gluster_chflags,
1847         .file_id_create_fn = NULL,
1848         .copy_chunk_send_fn = NULL,
1849         .copy_chunk_recv_fn = NULL,
1850         .streaminfo_fn = NULL,
1851         .get_real_filename_fn = vfs_gluster_get_real_filename,
1852         .connectpath_fn = vfs_gluster_connectpath,
1853
1854         .brl_lock_windows_fn = NULL,
1855         .brl_unlock_windows_fn = NULL,
1856         .brl_cancel_windows_fn = NULL,
1857         .strict_lock_fn = NULL,
1858         .strict_unlock_fn = NULL,
1859         .translate_name_fn = NULL,
1860         .fsctl_fn = NULL,
1861
1862         /* NT ACL Operations */
1863         .fget_nt_acl_fn = NULL,
1864         .get_nt_acl_fn = NULL,
1865         .fset_nt_acl_fn = NULL,
1866         .audit_file_fn = NULL,
1867
1868         /* Posix ACL Operations */
1869         .chmod_acl_fn = NULL,   /* passthrough to default */
1870         .fchmod_acl_fn = NULL,  /* passthrough to default */
1871         .sys_acl_get_file_fn = vfs_gluster_sys_acl_get_file,
1872         .sys_acl_get_fd_fn = vfs_gluster_sys_acl_get_fd,
1873         .sys_acl_blob_get_file_fn = posix_sys_acl_blob_get_file,
1874         .sys_acl_blob_get_fd_fn = posix_sys_acl_blob_get_fd,
1875         .sys_acl_set_file_fn = vfs_gluster_sys_acl_set_file,
1876         .sys_acl_set_fd_fn = vfs_gluster_sys_acl_set_fd,
1877         .sys_acl_delete_def_file_fn = vfs_gluster_sys_acl_delete_def_file,
1878
1879         /* EA Operations */
1880         .getxattr_fn = vfs_gluster_getxattr,
1881         .fgetxattr_fn = vfs_gluster_fgetxattr,
1882         .listxattr_fn = vfs_gluster_listxattr,
1883         .flistxattr_fn = vfs_gluster_flistxattr,
1884         .removexattr_fn = vfs_gluster_removexattr,
1885         .fremovexattr_fn = vfs_gluster_fremovexattr,
1886         .setxattr_fn = vfs_gluster_setxattr,
1887         .fsetxattr_fn = vfs_gluster_fsetxattr,
1888
1889         /* AIO Operations */
1890         .aio_force_fn = vfs_gluster_aio_force,
1891
1892         /* Offline Operations */
1893         .is_offline_fn = vfs_gluster_is_offline,
1894         .set_offline_fn = vfs_gluster_set_offline,
1895
1896         /* Durable handle Operations */
1897         .durable_cookie_fn = NULL,
1898         .durable_disconnect_fn = NULL,
1899         .durable_reconnect_fn = NULL,
1900 };
1901
1902 NTSTATUS vfs_glusterfs_init(void);
1903 NTSTATUS vfs_glusterfs_init(void)
1904 {
1905         return smb_register_vfs(SMB_VFS_INTERFACE_VERSION,
1906                                 "glusterfs", &glusterfs_fns);
1907 }