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