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