fs/ceph/super.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2
   3 #include <linux/ceph/ceph_debug.h>
   4
   5 #include <linux/backing-dev.h>
   6 #include <linux/ctype.h>
   7 #include <linux/fs.h>
   8 #include <linux/inet.h>
   9 #include <linux/in6.h>
  10 #include <linux/module.h>
  11 #include <linux/mount.h>
  12 #include <linux/fs_context.h>
  13 #include <linux/fs_parser.h>
  14 #include <linux/sched.h>
  15 #include <linux/seq_file.h>
  16 #include <linux/slab.h>
  17 #include <linux/statfs.h>
  18 #include <linux/string.h>
  19
  20 #include "super.h"
  21 #include "mds_client.h"
  22 #include "cache.h"
  23
  24 #include <linux/ceph/ceph_features.h>
  25 #include <linux/ceph/decode.h>
  26 #include <linux/ceph/mon_client.h>
  27 #include <linux/ceph/auth.h>
  28 #include <linux/ceph/debugfs.h>
  29
  30 static DEFINE_SPINLOCK(ceph_fsc_lock);
  31 static LIST_HEAD(ceph_fsc_list);
  32
  33 /*
  34  * Ceph superblock operations
  35  *
  36  * Handle the basics of mounting, unmounting.
  37  */
  38
  39 /*
  40  * super ops
  41  */
  42 static void ceph_put_super(struct super_block *s)
  43 {
  44         struct ceph_fs_client *fsc = ceph_sb_to_client(s);
  45
  46         dout("put_super\n");
  47         ceph_mdsc_close_sessions(fsc->mdsc);
  48 }
  49
  50 static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
  51 {
  52         struct ceph_fs_client *fsc = ceph_inode_to_client(d_inode(dentry));
  53         struct ceph_mon_client *monc = &fsc->client->monc;
  54         struct ceph_statfs st;
  55         int i, err;
  56         u64 data_pool;
  57
  58         if (fsc->mdsc->mdsmap->m_num_data_pg_pools == 1) {
  59                 data_pool = fsc->mdsc->mdsmap->m_data_pg_pools[0];
  60         } else {
  61                 data_pool = CEPH_NOPOOL;
  62         }
  63
  64         dout("statfs\n");
  65         err = ceph_monc_do_statfs(monc, data_pool, &st);
  66         if (err < 0)
  67                 return err;
  68
  69         /* fill in kstatfs */
  70         buf->f_type = CEPH_SUPER_MAGIC;  /* ?? */
  71
  72         /*
  73          * express utilization in terms of large blocks to avoid
  74          * overflow on 32-bit machines.
  75          *
  76          * NOTE: for the time being, we make bsize == frsize to humor
  77          * not-yet-ancient versions of glibc that are broken.
  78          * Someday, we will probably want to report a real block
  79          * size...  whatever that may mean for a network file system!
  80          */
  81         buf->f_bsize = 1 << CEPH_BLOCK_SHIFT;
  82         buf->f_frsize = 1 << CEPH_BLOCK_SHIFT;
  83
  84         /*
  85          * By default use root quota for stats; fallback to overall filesystem
  86          * usage if using 'noquotadf' mount option or if the root dir doesn't
  87          * have max_bytes quota set.
  88          */
  89         if (ceph_test_mount_opt(fsc, NOQUOTADF) ||
  90             !ceph_quota_update_statfs(fsc, buf)) {
  91                 buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10);
  92                 buf->f_bfree = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
  93                 buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
  94         }
  95
  96         buf->f_files = le64_to_cpu(st.num_objects);
  97         buf->f_ffree = -1;
  98         buf->f_namelen = NAME_MAX;
  99
 100         /* Must convert the fsid, for consistent values across arches */
 101         buf->f_fsid.val[0] = 0;
 102         mutex_lock(&monc->mutex);
 103         for (i = 0 ; i < sizeof(monc->monmap->fsid) / sizeof(__le32) ; ++i)
 104                 buf->f_fsid.val[0] ^= le32_to_cpu(((__le32 *)&monc->monmap->fsid)[i]);
 105         mutex_unlock(&monc->mutex);
 106
 107         /* fold the fs_cluster_id into the upper bits */
 108         buf->f_fsid.val[1] = monc->fs_cluster_id;
 109
 110         return 0;
 111 }
 112
 113 static int ceph_sync_fs(struct super_block *sb, int wait)
 114 {
 115         struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
 116
 117         if (!wait) {
 118                 dout("sync_fs (non-blocking)\n");
 119                 ceph_flush_dirty_caps(fsc->mdsc);
 120                 dout("sync_fs (non-blocking) done\n");
 121                 return 0;
 122         }
 123
 124         dout("sync_fs (blocking)\n");
 125         ceph_osdc_sync(&fsc->client->osdc);
 126         ceph_mdsc_sync(fsc->mdsc);
 127         dout("sync_fs (blocking) done\n");
 128         return 0;
 129 }
 130
 131 /*
 132  * mount options
 133  */
 134 enum {
 135         Opt_wsize,
 136         Opt_rsize,
 137         Opt_rasize,
 138         Opt_caps_wanted_delay_min,
 139         Opt_caps_wanted_delay_max,
 140         Opt_caps_max,
 141         Opt_readdir_max_entries,
 142         Opt_readdir_max_bytes,
 143         Opt_congestion_kb,
 144         /* int args above */
 145         Opt_snapdirname,
 146         Opt_mds_namespace,
 147         Opt_recover_session,
 148         Opt_source,
 149         /* string args above */
 150         Opt_dirstat,
 151         Opt_rbytes,
 152         Opt_asyncreaddir,
 153         Opt_dcache,
 154         Opt_ino32,
 155         Opt_fscache,
 156         Opt_poolperm,
 157         Opt_require_active_mds,
 158         Opt_acl,
 159         Opt_quotadf,
 160         Opt_copyfrom,
 161         Opt_wsync,
 162 };
 163
 164 enum ceph_recover_session_mode {
 165         ceph_recover_session_no,
 166         ceph_recover_session_clean
 167 };
 168
 169 static const struct constant_table ceph_param_recover[] = {
 170         { "no",         ceph_recover_session_no },
 171         { "clean",      ceph_recover_session_clean },
 172         {}
 173 };
 174
 175 static const struct fs_parameter_spec ceph_mount_parameters[] = {
 176         fsparam_flag_no ("acl",                         Opt_acl),
 177         fsparam_flag_no ("asyncreaddir",                Opt_asyncreaddir),
 178         fsparam_s32     ("caps_max",                    Opt_caps_max),
 179         fsparam_u32     ("caps_wanted_delay_max",       Opt_caps_wanted_delay_max),
 180         fsparam_u32     ("caps_wanted_delay_min",       Opt_caps_wanted_delay_min),
 181         fsparam_u32     ("write_congestion_kb",         Opt_congestion_kb),
 182         fsparam_flag_no ("copyfrom",                    Opt_copyfrom),
 183         fsparam_flag_no ("dcache",                      Opt_dcache),
 184         fsparam_flag_no ("dirstat",                     Opt_dirstat),
 185         fsparam_flag_no ("fsc",                         Opt_fscache), // fsc|nofsc
 186         fsparam_string  ("fsc",                         Opt_fscache), // fsc=...
 187         fsparam_flag_no ("ino32",                       Opt_ino32),
 188         fsparam_string  ("mds_namespace",               Opt_mds_namespace),
 189         fsparam_flag_no ("poolperm",                    Opt_poolperm),
 190         fsparam_flag_no ("quotadf",                     Opt_quotadf),
 191         fsparam_u32     ("rasize",                      Opt_rasize),
 192         fsparam_flag_no ("rbytes",                      Opt_rbytes),
 193         fsparam_u32     ("readdir_max_bytes",           Opt_readdir_max_bytes),
 194         fsparam_u32     ("readdir_max_entries",         Opt_readdir_max_entries),
 195         fsparam_enum    ("recover_session",             Opt_recover_session, ceph_param_recover),
 196         fsparam_flag_no ("require_active_mds",          Opt_require_active_mds),
 197         fsparam_u32     ("rsize",                       Opt_rsize),
 198         fsparam_string  ("snapdirname",                 Opt_snapdirname),
 199         fsparam_string  ("source",                      Opt_source),
 200         fsparam_u32     ("wsize",                       Opt_wsize),
 201         fsparam_flag_no ("wsync",                       Opt_wsync),
 202         {}
 203 };
 204
 205 struct ceph_parse_opts_ctx {
 206         struct ceph_options             *copts;
 207         struct ceph_mount_options       *opts;
 208 };
 209
 210 /*
 211  * Remove adjacent slashes and then the trailing slash, unless it is
 212  * the only remaining character.
 213  *
 214  * E.g. "//dir1////dir2///" --> "/dir1/dir2", "///" --> "/".
 215  */
 216 static void canonicalize_path(char *path)
 217 {
 218         int i, j = 0;
 219
 220         for (i = 0; path[i] != '\0'; i++) {
 221                 if (path[i] != '/' || j < 1 || path[j - 1] != '/')
 222                         path[j++] = path[i];
 223         }
 224
 225         if (j > 1 && path[j - 1] == '/')
 226                 j--;
 227         path[j] = '\0';
 228 }
 229
 230 /*
 231  * Parse the source parameter.  Distinguish the server list from the path.
 232  *
 233  * The source will look like:
 234  *     <server_spec>[,<server_spec>...]:[<path>]
 235  * where
 236  *     <server_spec> is <ip>[:<port>]
 237  *     <path> is optional, but if present must begin with '/'
 238  */
 239 static int ceph_parse_source(struct fs_parameter *param, struct fs_context *fc)
 240 {
 241         struct ceph_parse_opts_ctx *pctx = fc->fs_private;
 242         struct ceph_mount_options *fsopt = pctx->opts;
 243         char *dev_name = param->string, *dev_name_end;
 244         int ret;
 245
 246         dout("%s '%s'\n", __func__, dev_name);
 247         if (!dev_name || !*dev_name)
 248                 return invalfc(fc, "Empty source");
 249
 250         dev_name_end = strchr(dev_name, '/');
 251         if (dev_name_end) {
 252                 /*
 253                  * The server_path will include the whole chars from userland
 254                  * including the leading '/'.
 255                  */
 256                 kfree(fsopt->server_path);
 257                 fsopt->server_path = kstrdup(dev_name_end, GFP_KERNEL);
 258                 if (!fsopt->server_path)
 259                         return -ENOMEM;
 260
 261                 canonicalize_path(fsopt->server_path);
 262         } else {
 263                 dev_name_end = dev_name + strlen(dev_name);
 264         }
 265
 266         dev_name_end--;         /* back up to ':' separator */
 267         if (dev_name_end < dev_name || *dev_name_end != ':')
 268                 return invalfc(fc, "No path or : separator in source");
 269
 270         dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name);
 271         if (fsopt->server_path)
 272                 dout("server path '%s'\n", fsopt->server_path);
 273
 274         ret = ceph_parse_mon_ips(param->string, dev_name_end - dev_name,
 275                                  pctx->copts, fc->log.log);
 276         if (ret)
 277                 return ret;
 278
 279         fc->source = param->string;
 280         param->string = NULL;
 281         return 0;
 282 }
 283
 284 static int ceph_parse_mount_param(struct fs_context *fc,
 285                                   struct fs_parameter *param)
 286 {
 287         struct ceph_parse_opts_ctx *pctx = fc->fs_private;
 288         struct ceph_mount_options *fsopt = pctx->opts;
 289         struct fs_parse_result result;
 290         unsigned int mode;
 291         int token, ret;
 292
 293         ret = ceph_parse_param(param, pctx->copts, fc->log.log);
 294         if (ret != -ENOPARAM)
 295                 return ret;
 296
 297         token = fs_parse(fc, ceph_mount_parameters, param, &result);
 298         dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
 299         if (token < 0)
 300                 return token;
 301
 302         switch (token) {
 303         case Opt_snapdirname:
 304                 kfree(fsopt->snapdir_name);
 305                 fsopt->snapdir_name = param->string;
 306                 param->string = NULL;
 307                 break;
 308         case Opt_mds_namespace:
 309                 kfree(fsopt->mds_namespace);
 310                 fsopt->mds_namespace = param->string;
 311                 param->string = NULL;
 312                 break;
 313         case Opt_recover_session:
 314                 mode = result.uint_32;
 315                 if (mode == ceph_recover_session_no)
 316                         fsopt->flags &= ~CEPH_MOUNT_OPT_CLEANRECOVER;
 317                 else if (mode == ceph_recover_session_clean)
 318                         fsopt->flags |= CEPH_MOUNT_OPT_CLEANRECOVER;
 319                 else
 320                         BUG();
 321                 break;
 322         case Opt_source:
 323                 if (fc->source)
 324                         return invalfc(fc, "Multiple sources specified");
 325                 return ceph_parse_source(param, fc);
 326         case Opt_wsize:
 327                 if (result.uint_32 < PAGE_SIZE ||
 328                     result.uint_32 > CEPH_MAX_WRITE_SIZE)
 329                         goto out_of_range;
 330                 fsopt->wsize = ALIGN(result.uint_32, PAGE_SIZE);
 331                 break;
 332         case Opt_rsize:
 333                 if (result.uint_32 < PAGE_SIZE ||
 334                     result.uint_32 > CEPH_MAX_READ_SIZE)
 335                         goto out_of_range;
 336                 fsopt->rsize = ALIGN(result.uint_32, PAGE_SIZE);
 337                 break;
 338         case Opt_rasize:
 339                 fsopt->rasize = ALIGN(result.uint_32, PAGE_SIZE);
 340                 break;
 341         case Opt_caps_wanted_delay_min:
 342                 if (result.uint_32 < 1)
 343                         goto out_of_range;
 344                 fsopt->caps_wanted_delay_min = result.uint_32;
 345                 break;
 346         case Opt_caps_wanted_delay_max:
 347                 if (result.uint_32 < 1)
 348                         goto out_of_range;
 349                 fsopt->caps_wanted_delay_max = result.uint_32;
 350                 break;
 351         case Opt_caps_max:
 352                 if (result.int_32 < 0)
 353                         goto out_of_range;
 354                 fsopt->caps_max = result.int_32;
 355                 break;
 356         case Opt_readdir_max_entries:
 357                 if (result.uint_32 < 1)
 358                         goto out_of_range;
 359                 fsopt->max_readdir = result.uint_32;
 360                 break;
 361         case Opt_readdir_max_bytes:
 362                 if (result.uint_32 < PAGE_SIZE && result.uint_32 != 0)
 363                         goto out_of_range;
 364                 fsopt->max_readdir_bytes = result.uint_32;
 365                 break;
 366         case Opt_congestion_kb:
 367                 if (result.uint_32 < 1024) /* at least 1M */
 368                         goto out_of_range;
 369                 fsopt->congestion_kb = result.uint_32;
 370                 break;
 371         case Opt_dirstat:
 372                 if (!result.negated)
 373                         fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT;
 374                 else
 375                         fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT;
 376                 break;
 377         case Opt_rbytes:
 378                 if (!result.negated)
 379                         fsopt->flags |= CEPH_MOUNT_OPT_RBYTES;
 380                 else
 381                         fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES;
 382                 break;
 383         case Opt_asyncreaddir:
 384                 if (!result.negated)
 385                         fsopt->flags &= ~CEPH_MOUNT_OPT_NOASYNCREADDIR;
 386                 else
 387                         fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
 388                 break;
 389         case Opt_dcache:
 390                 if (!result.negated)
 391                         fsopt->flags |= CEPH_MOUNT_OPT_DCACHE;
 392                 else
 393                         fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
 394                 break;
 395         case Opt_ino32:
 396                 if (!result.negated)
 397                         fsopt->flags |= CEPH_MOUNT_OPT_INO32;
 398                 else
 399                         fsopt->flags &= ~CEPH_MOUNT_OPT_INO32;
 400                 break;
 401
 402         case Opt_fscache:
 403 #ifdef CONFIG_CEPH_FSCACHE
 404                 kfree(fsopt->fscache_uniq);
 405                 fsopt->fscache_uniq = NULL;
 406                 if (result.negated) {
 407                         fsopt->flags &= ~CEPH_MOUNT_OPT_FSCACHE;
 408                 } else {
 409                         fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE;
 410                         fsopt->fscache_uniq = param->string;
 411                         param->string = NULL;
 412                 }
 413                 break;
 414 #else
 415                 return invalfc(fc, "fscache support is disabled");
 416 #endif
 417         case Opt_poolperm:
 418                 if (!result.negated)
 419                         fsopt->flags &= ~CEPH_MOUNT_OPT_NOPOOLPERM;
 420                 else
 421                         fsopt->flags |= CEPH_MOUNT_OPT_NOPOOLPERM;
 422                 break;
 423         case Opt_require_active_mds:
 424                 if (!result.negated)
 425                         fsopt->flags &= ~CEPH_MOUNT_OPT_MOUNTWAIT;
 426                 else
 427                         fsopt->flags |= CEPH_MOUNT_OPT_MOUNTWAIT;
 428                 break;
 429         case Opt_quotadf:
 430                 if (!result.negated)
 431                         fsopt->flags &= ~CEPH_MOUNT_OPT_NOQUOTADF;
 432                 else
 433                         fsopt->flags |= CEPH_MOUNT_OPT_NOQUOTADF;
 434                 break;
 435         case Opt_copyfrom:
 436                 if (!result.negated)
 437                         fsopt->flags &= ~CEPH_MOUNT_OPT_NOCOPYFROM;
 438                 else
 439                         fsopt->flags |= CEPH_MOUNT_OPT_NOCOPYFROM;
 440                 break;
 441         case Opt_acl:
 442                 if (!result.negated) {
 443 #ifdef CONFIG_CEPH_FS_POSIX_ACL
 444                         fc->sb_flags |= SB_POSIXACL;
 445 #else
 446                         return invalfc(fc, "POSIX ACL support is disabled");
 447 #endif
 448                 } else {
 449                         fc->sb_flags &= ~SB_POSIXACL;
 450                 }
 451                 break;
 452         case Opt_wsync:
 453                 if (!result.negated)
 454                         fsopt->flags &= ~CEPH_MOUNT_OPT_ASYNC_DIROPS;
 455                 else
 456                         fsopt->flags |= CEPH_MOUNT_OPT_ASYNC_DIROPS;
 457                 break;
 458         default:
 459                 BUG();
 460         }
 461         return 0;
 462
 463 out_of_range:
 464         return invalfc(fc, "%s out of range", param->key);
 465 }
 466
 467 static void destroy_mount_options(struct ceph_mount_options *args)
 468 {
 469         dout("destroy_mount_options %p\n", args);
 470         if (!args)
 471                 return;
 472
 473         kfree(args->snapdir_name);
 474         kfree(args->mds_namespace);
 475         kfree(args->server_path);
 476         kfree(args->fscache_uniq);
 477         kfree(args);
 478 }
 479
 480 static int strcmp_null(const char *s1, const char *s2)
 481 {
 482         if (!s1 && !s2)
 483                 return 0;
 484         if (s1 && !s2)
 485                 return -1;
 486         if (!s1 && s2)
 487                 return 1;
 488         return strcmp(s1, s2);
 489 }
 490
 491 static int compare_mount_options(struct ceph_mount_options *new_fsopt,
 492                                  struct ceph_options *new_opt,
 493                                  struct ceph_fs_client *fsc)
 494 {
 495         struct ceph_mount_options *fsopt1 = new_fsopt;
 496         struct ceph_mount_options *fsopt2 = fsc->mount_options;
 497         int ofs = offsetof(struct ceph_mount_options, snapdir_name);
 498         int ret;
 499
 500         ret = memcmp(fsopt1, fsopt2, ofs);
 501         if (ret)
 502                 return ret;
 503
 504         ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name);
 505         if (ret)
 506                 return ret;
 507
 508         ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace);
 509         if (ret)
 510                 return ret;
 511
 512         ret = strcmp_null(fsopt1->server_path, fsopt2->server_path);
 513         if (ret)
 514                 return ret;
 515
 516         ret = strcmp_null(fsopt1->fscache_uniq, fsopt2->fscache_uniq);
 517         if (ret)
 518                 return ret;
 519
 520         return ceph_compare_options(new_opt, fsc->client);
 521 }
 522
 523 /**
 524  * ceph_show_options - Show mount options in /proc/mounts
 525  * @m: seq_file to write to
 526  * @root: root of that (sub)tree
 527  */
 528 static int ceph_show_options(struct seq_file *m, struct dentry *root)
 529 {
 530         struct ceph_fs_client *fsc = ceph_sb_to_client(root->d_sb);
 531         struct ceph_mount_options *fsopt = fsc->mount_options;
 532         size_t pos;
 533         int ret;
 534
 535         /* a comma between MNT/MS and client options */
 536         seq_putc(m, ',');
 537         pos = m->count;
 538
 539         ret = ceph_print_client_options(m, fsc->client, false);
 540         if (ret)
 541                 return ret;
 542
 543         /* retract our comma if no client options */
 544         if (m->count == pos)
 545                 m->count--;
 546
 547         if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT)
 548                 seq_puts(m, ",dirstat");
 549         if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES))
 550                 seq_puts(m, ",rbytes");
 551         if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
 552                 seq_puts(m, ",noasyncreaddir");
 553         if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
 554                 seq_puts(m, ",nodcache");
 555         if (fsopt->flags & CEPH_MOUNT_OPT_INO32)
 556                 seq_puts(m, ",ino32");
 557         if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) {
 558                 seq_show_option(m, "fsc", fsopt->fscache_uniq);
 559         }
 560         if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
 561                 seq_puts(m, ",nopoolperm");
 562         if (fsopt->flags & CEPH_MOUNT_OPT_NOQUOTADF)
 563                 seq_puts(m, ",noquotadf");
 564
 565 #ifdef CONFIG_CEPH_FS_POSIX_ACL
 566         if (root->d_sb->s_flags & SB_POSIXACL)
 567                 seq_puts(m, ",acl");
 568         else
 569                 seq_puts(m, ",noacl");
 570 #endif
 571
 572         if ((fsopt->flags & CEPH_MOUNT_OPT_NOCOPYFROM) == 0)
 573                 seq_puts(m, ",copyfrom");
 574
 575         if (fsopt->mds_namespace)
 576                 seq_show_option(m, "mds_namespace", fsopt->mds_namespace);
 577
 578         if (fsopt->flags & CEPH_MOUNT_OPT_CLEANRECOVER)
 579                 seq_show_option(m, "recover_session", "clean");
 580
 581         if (!(fsopt->flags & CEPH_MOUNT_OPT_ASYNC_DIROPS))
 582                 seq_puts(m, ",wsync");
 583
 584         if (fsopt->wsize != CEPH_MAX_WRITE_SIZE)
 585                 seq_printf(m, ",wsize=%u", fsopt->wsize);
 586         if (fsopt->rsize != CEPH_MAX_READ_SIZE)
 587                 seq_printf(m, ",rsize=%u", fsopt->rsize);
 588         if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
 589                 seq_printf(m, ",rasize=%u", fsopt->rasize);
 590         if (fsopt->congestion_kb != default_congestion_kb())
 591                 seq_printf(m, ",write_congestion_kb=%u", fsopt->congestion_kb);
 592         if (fsopt->caps_max)
 593                 seq_printf(m, ",caps_max=%d", fsopt->caps_max);
 594         if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
 595                 seq_printf(m, ",caps_wanted_delay_min=%u",
 596                          fsopt->caps_wanted_delay_min);
 597         if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
 598                 seq_printf(m, ",caps_wanted_delay_max=%u",
 599                            fsopt->caps_wanted_delay_max);
 600         if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT)
 601                 seq_printf(m, ",readdir_max_entries=%u", fsopt->max_readdir);
 602         if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
 603                 seq_printf(m, ",readdir_max_bytes=%u", fsopt->max_readdir_bytes);
 604         if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
 605                 seq_show_option(m, "snapdirname", fsopt->snapdir_name);
 606
 607         return 0;
 608 }
 609
 610 /*
 611  * handle any mon messages the standard library doesn't understand.
 612  * return error if we don't either.
 613  */
 614 static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg)
 615 {
 616         struct ceph_fs_client *fsc = client->private;
 617         int type = le16_to_cpu(msg->hdr.type);
 618
 619         switch (type) {
 620         case CEPH_MSG_MDS_MAP:
 621                 ceph_mdsc_handle_mdsmap(fsc->mdsc, msg);
 622                 return 0;
 623         case CEPH_MSG_FS_MAP_USER:
 624                 ceph_mdsc_handle_fsmap(fsc->mdsc, msg);
 625                 return 0;
 626         default:
 627                 return -1;
 628         }
 629 }
 630
 631 /*
 632  * create a new fs client
 633  *
 634  * Success or not, this function consumes @fsopt and @opt.
 635  */
 636 static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
 637                                         struct ceph_options *opt)
 638 {
 639         struct ceph_fs_client *fsc;
 640         int err;
 641
 642         fsc = kzalloc(sizeof(*fsc), GFP_KERNEL);
 643         if (!fsc) {
 644                 err = -ENOMEM;
 645                 goto fail;
 646         }
 647
 648         fsc->client = ceph_create_client(opt, fsc);
 649         if (IS_ERR(fsc->client)) {
 650                 err = PTR_ERR(fsc->client);
 651                 goto fail;
 652         }
 653         opt = NULL; /* fsc->client now owns this */
 654
 655         fsc->client->extra_mon_dispatch = extra_mon_dispatch;
 656         ceph_set_opt(fsc->client, ABORT_ON_FULL);
 657
 658         if (!fsopt->mds_namespace) {
 659                 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
 660                                    0, true);
 661         } else {
 662                 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_FSMAP,
 663                                    0, false);
 664         }
 665
 666         fsc->mount_options = fsopt;
 667
 668         fsc->sb = NULL;
 669         fsc->mount_state = CEPH_MOUNT_MOUNTING;
 670         fsc->filp_gen = 1;
 671         fsc->have_copy_from2 = true;
 672
 673         atomic_long_set(&fsc->writeback_count, 0);
 674
 675         err = -ENOMEM;
 676         /*
 677          * The number of concurrent works can be high but they don't need
 678          * to be processed in parallel, limit concurrency.
 679          */
 680         fsc->inode_wq = alloc_workqueue("ceph-inode", WQ_UNBOUND, 0);
 681         if (!fsc->inode_wq)
 682                 goto fail_client;
 683         fsc->cap_wq = alloc_workqueue("ceph-cap", 0, 1);
 684         if (!fsc->cap_wq)
 685                 goto fail_inode_wq;
 686
 687         spin_lock(&ceph_fsc_lock);
 688         list_add_tail(&fsc->metric_wakeup, &ceph_fsc_list);
 689         spin_unlock(&ceph_fsc_lock);
 690
 691         return fsc;
 692
 693 fail_inode_wq:
 694         destroy_workqueue(fsc->inode_wq);
 695 fail_client:
 696         ceph_destroy_client(fsc->client);
 697 fail:
 698         kfree(fsc);
 699         if (opt)
 700                 ceph_destroy_options(opt);
 701         destroy_mount_options(fsopt);
 702         return ERR_PTR(err);
 703 }
 704
 705 static void flush_fs_workqueues(struct ceph_fs_client *fsc)
 706 {
 707         flush_workqueue(fsc->inode_wq);
 708         flush_workqueue(fsc->cap_wq);
 709 }
 710
 711 static void destroy_fs_client(struct ceph_fs_client *fsc)
 712 {
 713         dout("destroy_fs_client %p\n", fsc);
 714
 715         spin_lock(&ceph_fsc_lock);
 716         list_del(&fsc->metric_wakeup);
 717         spin_unlock(&ceph_fsc_lock);
 718
 719         ceph_mdsc_destroy(fsc);
 720         destroy_workqueue(fsc->inode_wq);
 721         destroy_workqueue(fsc->cap_wq);
 722
 723         destroy_mount_options(fsc->mount_options);
 724
 725         ceph_destroy_client(fsc->client);
 726
 727         kfree(fsc);
 728         dout("destroy_fs_client %p done\n", fsc);
 729 }
 730
 731 /*
 732  * caches
 733  */
 734 struct kmem_cache *ceph_inode_cachep;
 735 struct kmem_cache *ceph_cap_cachep;
 736 struct kmem_cache *ceph_cap_flush_cachep;
 737 struct kmem_cache *ceph_dentry_cachep;
 738 struct kmem_cache *ceph_file_cachep;
 739 struct kmem_cache *ceph_dir_file_cachep;
 740 struct kmem_cache *ceph_mds_request_cachep;
 741 mempool_t *ceph_wb_pagevec_pool;
 742
 743 static void ceph_inode_init_once(void *foo)
 744 {
 745         struct ceph_inode_info *ci = foo;
 746         inode_init_once(&ci->vfs_inode);
 747 }
 748
 749 static int __init init_caches(void)
 750 {
 751         int error = -ENOMEM;
 752
 753         ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
 754                                       sizeof(struct ceph_inode_info),
 755                                       __alignof__(struct ceph_inode_info),
 756                                       SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
 757                                       SLAB_ACCOUNT, ceph_inode_init_once);
 758         if (!ceph_inode_cachep)
 759                 return -ENOMEM;
 760
 761         ceph_cap_cachep = KMEM_CACHE(ceph_cap, SLAB_MEM_SPREAD);
 762         if (!ceph_cap_cachep)
 763                 goto bad_cap;
 764         ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush,
 765                                            SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
 766         if (!ceph_cap_flush_cachep)
 767                 goto bad_cap_flush;
 768
 769         ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
 770                                         SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
 771         if (!ceph_dentry_cachep)
 772                 goto bad_dentry;
 773
 774         ceph_file_cachep = KMEM_CACHE(ceph_file_info, SLAB_MEM_SPREAD);
 775         if (!ceph_file_cachep)
 776                 goto bad_file;
 777
 778         ceph_dir_file_cachep = KMEM_CACHE(ceph_dir_file_info, SLAB_MEM_SPREAD);
 779         if (!ceph_dir_file_cachep)
 780                 goto bad_dir_file;
 781
 782         ceph_mds_request_cachep = KMEM_CACHE(ceph_mds_request, SLAB_MEM_SPREAD);
 783         if (!ceph_mds_request_cachep)
 784                 goto bad_mds_req;
 785
 786         ceph_wb_pagevec_pool = mempool_create_kmalloc_pool(10, CEPH_MAX_WRITE_SIZE >> PAGE_SHIFT);
 787         if (!ceph_wb_pagevec_pool)
 788                 goto bad_pagevec_pool;
 789
 790         return 0;
 791
 792 bad_pagevec_pool:
 793         kmem_cache_destroy(ceph_mds_request_cachep);
 794 bad_mds_req:
 795         kmem_cache_destroy(ceph_dir_file_cachep);
 796 bad_dir_file:
 797         kmem_cache_destroy(ceph_file_cachep);
 798 bad_file:
 799         kmem_cache_destroy(ceph_dentry_cachep);
 800 bad_dentry:
 801         kmem_cache_destroy(ceph_cap_flush_cachep);
 802 bad_cap_flush:
 803         kmem_cache_destroy(ceph_cap_cachep);
 804 bad_cap:
 805         kmem_cache_destroy(ceph_inode_cachep);
 806         return error;
 807 }
 808
 809 static void destroy_caches(void)
 810 {
 811         /*
 812          * Make sure all delayed rcu free inodes are flushed before we
 813          * destroy cache.
 814          */
 815         rcu_barrier();
 816
 817         kmem_cache_destroy(ceph_inode_cachep);
 818         kmem_cache_destroy(ceph_cap_cachep);
 819         kmem_cache_destroy(ceph_cap_flush_cachep);
 820         kmem_cache_destroy(ceph_dentry_cachep);
 821         kmem_cache_destroy(ceph_file_cachep);
 822         kmem_cache_destroy(ceph_dir_file_cachep);
 823         kmem_cache_destroy(ceph_mds_request_cachep);
 824         mempool_destroy(ceph_wb_pagevec_pool);
 825 }
 826
 827 static void __ceph_umount_begin(struct ceph_fs_client *fsc)
 828 {
 829         ceph_osdc_abort_requests(&fsc->client->osdc, -EIO);
 830         ceph_mdsc_force_umount(fsc->mdsc);
 831         fsc->filp_gen++; // invalidate open files
 832 }
 833
 834 /*
 835  * ceph_umount_begin - initiate forced umount.  Tear down the
 836  * mount, skipping steps that may hang while waiting for server(s).
 837  */
 838 void ceph_umount_begin(struct super_block *sb)
 839 {
 840         struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
 841
 842         dout("ceph_umount_begin - starting forced umount\n");
 843         if (!fsc)
 844                 return;
 845         fsc->mount_state = CEPH_MOUNT_SHUTDOWN;
 846         __ceph_umount_begin(fsc);
 847 }
 848
 849 static const struct super_operations ceph_super_ops = {
 850         .alloc_inode    = ceph_alloc_inode,
 851         .free_inode     = ceph_free_inode,
 852         .write_inode    = ceph_write_inode,
 853         .drop_inode     = generic_delete_inode,
 854         .evict_inode    = ceph_evict_inode,
 855         .sync_fs        = ceph_sync_fs,
 856         .put_super      = ceph_put_super,
 857         .show_options   = ceph_show_options,
 858         .statfs         = ceph_statfs,
 859         .umount_begin   = ceph_umount_begin,
 860 };
 861
 862 /*
 863  * Bootstrap mount by opening the root directory.  Note the mount
 864  * @started time from caller, and time out if this takes too long.
 865  */
 866 static struct dentry *open_root_dentry(struct ceph_fs_client *fsc,
 867                                        const char *path,
 868                                        unsigned long started)
 869 {
 870         struct ceph_mds_client *mdsc = fsc->mdsc;
 871         struct ceph_mds_request *req = NULL;
 872         int err;
 873         struct dentry *root;
 874
 875         /* open dir */
 876         dout("open_root_inode opening '%s'\n", path);
 877         req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
 878         if (IS_ERR(req))
 879                 return ERR_CAST(req);
 880         req->r_path1 = kstrdup(path, GFP_NOFS);
 881         if (!req->r_path1) {
 882                 root = ERR_PTR(-ENOMEM);
 883                 goto out;
 884         }
 885
 886         req->r_ino1.ino = CEPH_INO_ROOT;
 887         req->r_ino1.snap = CEPH_NOSNAP;
 888         req->r_started = started;
 889         req->r_timeout = fsc->client->options->mount_timeout;
 890         req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
 891         req->r_num_caps = 2;
 892         err = ceph_mdsc_do_request(mdsc, NULL, req);
 893         if (err == 0) {
 894                 struct inode *inode = req->r_target_inode;
 895                 req->r_target_inode = NULL;
 896                 dout("open_root_inode success\n");
 897                 root = d_make_root(inode);
 898                 if (!root) {
 899                         root = ERR_PTR(-ENOMEM);
 900                         goto out;
 901                 }
 902                 dout("open_root_inode success, root dentry is %p\n", root);
 903         } else {
 904                 root = ERR_PTR(err);
 905         }
 906 out:
 907         ceph_mdsc_put_request(req);
 908         return root;
 909 }
 910
 911 /*
 912  * mount: join the ceph cluster, and open root directory.
 913  */
 914 static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc,
 915                                       struct fs_context *fc)
 916 {
 917         int err;
 918         unsigned long started = jiffies;  /* note the start time */
 919         struct dentry *root;
 920
 921         dout("mount start %p\n", fsc);
 922         mutex_lock(&fsc->client->mount_mutex);
 923
 924         if (!fsc->sb->s_root) {
 925                 const char *path = fsc->mount_options->server_path ?
 926                                      fsc->mount_options->server_path + 1 : "";
 927
 928                 err = __ceph_open_session(fsc->client, started);
 929                 if (err < 0)
 930                         goto out;
 931
 932                 /* setup fscache */
 933                 if (fsc->mount_options->flags & CEPH_MOUNT_OPT_FSCACHE) {
 934                         err = ceph_fscache_register_fs(fsc, fc);
 935                         if (err < 0)
 936                                 goto out;
 937                 }
 938
 939                 dout("mount opening path '%s'\n", path);
 940
 941                 ceph_fs_debugfs_init(fsc);
 942
 943                 root = open_root_dentry(fsc, path, started);
 944                 if (IS_ERR(root)) {
 945                         err = PTR_ERR(root);
 946                         goto out;
 947                 }
 948                 fsc->sb->s_root = dget(root);
 949         } else {
 950                 root = dget(fsc->sb->s_root);
 951         }
 952
 953         fsc->mount_state = CEPH_MOUNT_MOUNTED;
 954         dout("mount success\n");
 955         mutex_unlock(&fsc->client->mount_mutex);
 956         return root;
 957
 958 out:
 959         mutex_unlock(&fsc->client->mount_mutex);
 960         return ERR_PTR(err);
 961 }
 962
 963 static int ceph_set_super(struct super_block *s, struct fs_context *fc)
 964 {
 965         struct ceph_fs_client *fsc = s->s_fs_info;
 966         int ret;
 967
 968         dout("set_super %p\n", s);
 969
 970         s->s_maxbytes = MAX_LFS_FILESIZE;
 971
 972         s->s_xattr = ceph_xattr_handlers;
 973         fsc->sb = s;
 974         fsc->max_file_size = 1ULL << 40; /* temp value until we get mdsmap */
 975
 976         s->s_op = &ceph_super_ops;
 977         s->s_d_op = &ceph_dentry_ops;
 978         s->s_export_op = &ceph_export_ops;
 979
 980         s->s_time_gran = 1;
 981         s->s_time_min = 0;
 982         s->s_time_max = U32_MAX;
 983
 984         ret = set_anon_super_fc(s, fc);
 985         if (ret != 0)
 986                 fsc->sb = NULL;
 987         return ret;
 988 }
 989
 990 /*
 991  * share superblock if same fs AND options
 992  */
 993 static int ceph_compare_super(struct super_block *sb, struct fs_context *fc)
 994 {
 995         struct ceph_fs_client *new = fc->s_fs_info;
 996         struct ceph_mount_options *fsopt = new->mount_options;
 997         struct ceph_options *opt = new->client->options;
 998         struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
 999
1000         dout("ceph_compare_super %p\n", sb);
1001
1002         if (compare_mount_options(fsopt, opt, fsc)) {
1003                 dout("monitor(s)/mount options don't match\n");
1004                 return 0;
1005         }
1006         if ((opt->flags & CEPH_OPT_FSID) &&
1007             ceph_fsid_compare(&opt->fsid, &fsc->client->fsid)) {
1008                 dout("fsid doesn't match\n");
1009                 return 0;
1010         }
1011         if (fc->sb_flags != (sb->s_flags & ~SB_BORN)) {
1012                 dout("flags differ\n");
1013                 return 0;
1014         }
1015
1016         if (fsc->blocklisted && !ceph_test_mount_opt(fsc, CLEANRECOVER)) {
1017                 dout("client is blocklisted (and CLEANRECOVER is not set)\n");
1018                 return 0;
1019         }
1020
1021         if (fsc->mount_state == CEPH_MOUNT_SHUTDOWN) {
1022                 dout("client has been forcibly unmounted\n");
1023                 return 0;
1024         }
1025
1026         return 1;
1027 }
1028
1029 /*
1030  * construct our own bdi so we can control readahead, etc.
1031  */
1032 static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
1033
1034 static int ceph_setup_bdi(struct super_block *sb, struct ceph_fs_client *fsc)
1035 {
1036         int err;
1037
1038         err = super_setup_bdi_name(sb, "ceph-%ld",
1039                                    atomic_long_inc_return(&bdi_seq));
1040         if (err)
1041                 return err;
1042
1043         /* set ra_pages based on rasize mount option? */
1044         sb->s_bdi->ra_pages = fsc->mount_options->rasize >> PAGE_SHIFT;
1045
1046         /* set io_pages based on max osd read size */
1047         sb->s_bdi->io_pages = fsc->mount_options->rsize >> PAGE_SHIFT;
1048
1049         return 0;
1050 }
1051
1052 static int ceph_get_tree(struct fs_context *fc)
1053 {
1054         struct ceph_parse_opts_ctx *pctx = fc->fs_private;
1055         struct super_block *sb;
1056         struct ceph_fs_client *fsc;
1057         struct dentry *res;
1058         int (*compare_super)(struct super_block *, struct fs_context *) =
1059                 ceph_compare_super;
1060         int err;
1061
1062         dout("ceph_get_tree\n");
1063
1064         if (!fc->source)
1065                 return invalfc(fc, "No source");
1066
1067         /* create client (which we may/may not use) */
1068         fsc = create_fs_client(pctx->opts, pctx->copts);
1069         pctx->opts = NULL;
1070         pctx->copts = NULL;
1071         if (IS_ERR(fsc)) {
1072                 err = PTR_ERR(fsc);
1073                 goto out_final;
1074         }
1075
1076         err = ceph_mdsc_init(fsc);
1077         if (err < 0)
1078                 goto out;
1079
1080         if (ceph_test_opt(fsc->client, NOSHARE))
1081                 compare_super = NULL;
1082
1083         fc->s_fs_info = fsc;
1084         sb = sget_fc(fc, compare_super, ceph_set_super);
1085         fc->s_fs_info = NULL;
1086         if (IS_ERR(sb)) {
1087                 err = PTR_ERR(sb);
1088                 goto out;
1089         }
1090
1091         if (ceph_sb_to_client(sb) != fsc) {
1092                 destroy_fs_client(fsc);
1093                 fsc = ceph_sb_to_client(sb);
1094                 dout("get_sb got existing client %p\n", fsc);
1095         } else {
1096                 dout("get_sb using new client %p\n", fsc);
1097                 err = ceph_setup_bdi(sb, fsc);
1098                 if (err < 0)
1099                         goto out_splat;
1100         }
1101
1102         res = ceph_real_mount(fsc, fc);
1103         if (IS_ERR(res)) {
1104                 err = PTR_ERR(res);
1105                 goto out_splat;
1106         }
1107         dout("root %p inode %p ino %llx.%llx\n", res,
1108              d_inode(res), ceph_vinop(d_inode(res)));
1109         fc->root = fsc->sb->s_root;
1110         return 0;
1111
1112 out_splat:
1113         if (!ceph_mdsmap_is_cluster_available(fsc->mdsc->mdsmap)) {
1114                 pr_info("No mds server is up or the cluster is laggy\n");
1115                 err = -EHOSTUNREACH;
1116         }
1117
1118         ceph_mdsc_close_sessions(fsc->mdsc);
1119         deactivate_locked_super(sb);
1120         goto out_final;
1121
1122 out:
1123         destroy_fs_client(fsc);
1124 out_final:
1125         dout("ceph_get_tree fail %d\n", err);
1126         return err;
1127 }
1128
1129 static void ceph_free_fc(struct fs_context *fc)
1130 {
1131         struct ceph_parse_opts_ctx *pctx = fc->fs_private;
1132
1133         if (pctx) {
1134                 destroy_mount_options(pctx->opts);
1135                 ceph_destroy_options(pctx->copts);
1136                 kfree(pctx);
1137         }
1138 }
1139
1140 static int ceph_reconfigure_fc(struct fs_context *fc)
1141 {
1142         struct ceph_parse_opts_ctx *pctx = fc->fs_private;
1143         struct ceph_mount_options *fsopt = pctx->opts;
1144         struct ceph_fs_client *fsc = ceph_sb_to_client(fc->root->d_sb);
1145
1146         if (fsopt->flags & CEPH_MOUNT_OPT_ASYNC_DIROPS)
1147                 ceph_set_mount_opt(fsc, ASYNC_DIROPS);
1148         else
1149                 ceph_clear_mount_opt(fsc, ASYNC_DIROPS);
1150
1151         sync_filesystem(fc->root->d_sb);
1152         return 0;
1153 }
1154
1155 static const struct fs_context_operations ceph_context_ops = {
1156         .free           = ceph_free_fc,
1157         .parse_param    = ceph_parse_mount_param,
1158         .get_tree       = ceph_get_tree,
1159         .reconfigure    = ceph_reconfigure_fc,
1160 };
1161
1162 /*
1163  * Set up the filesystem mount context.
1164  */
1165 static int ceph_init_fs_context(struct fs_context *fc)
1166 {
1167         struct ceph_parse_opts_ctx *pctx;
1168         struct ceph_mount_options *fsopt;
1169
1170         pctx = kzalloc(sizeof(*pctx), GFP_KERNEL);
1171         if (!pctx)
1172                 return -ENOMEM;
1173
1174         pctx->copts = ceph_alloc_options();
1175         if (!pctx->copts)
1176                 goto nomem;
1177
1178         pctx->opts = kzalloc(sizeof(*pctx->opts), GFP_KERNEL);
1179         if (!pctx->opts)
1180                 goto nomem;
1181
1182         fsopt = pctx->opts;
1183         fsopt->flags = CEPH_MOUNT_OPT_DEFAULT;
1184
1185         fsopt->wsize = CEPH_MAX_WRITE_SIZE;
1186         fsopt->rsize = CEPH_MAX_READ_SIZE;
1187         fsopt->rasize = CEPH_RASIZE_DEFAULT;
1188         fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
1189         if (!fsopt->snapdir_name)
1190                 goto nomem;
1191
1192         fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
1193         fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
1194         fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT;
1195         fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
1196         fsopt->congestion_kb = default_congestion_kb();
1197
1198 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1199         fc->sb_flags |= SB_POSIXACL;
1200 #endif
1201
1202         fc->fs_private = pctx;
1203         fc->ops = &ceph_context_ops;
1204         return 0;
1205
1206 nomem:
1207         destroy_mount_options(pctx->opts);
1208         ceph_destroy_options(pctx->copts);
1209         kfree(pctx);
1210         return -ENOMEM;
1211 }
1212
1213 static void ceph_kill_sb(struct super_block *s)
1214 {
1215         struct ceph_fs_client *fsc = ceph_sb_to_client(s);
1216
1217         dout("kill_sb %p\n", s);
1218
1219         ceph_mdsc_pre_umount(fsc->mdsc);
1220         flush_fs_workqueues(fsc);
1221
1222         kill_anon_super(s);
1223
1224         fsc->client->extra_mon_dispatch = NULL;
1225         ceph_fs_debugfs_cleanup(fsc);
1226
1227         ceph_fscache_unregister_fs(fsc);
1228
1229         destroy_fs_client(fsc);
1230 }
1231
1232 static struct file_system_type ceph_fs_type = {
1233         .owner          = THIS_MODULE,
1234         .name           = "ceph",
1235         .init_fs_context = ceph_init_fs_context,
1236         .kill_sb        = ceph_kill_sb,
1237         .fs_flags       = FS_RENAME_DOES_D_MOVE,
1238 };
1239 MODULE_ALIAS_FS("ceph");
1240
1241 int ceph_force_reconnect(struct super_block *sb)
1242 {
1243         struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
1244         int err = 0;
1245
1246         fsc->mount_state = CEPH_MOUNT_RECOVER;
1247         __ceph_umount_begin(fsc);
1248
1249         /* Make sure all page caches get invalidated.
1250          * see remove_session_caps_cb() */
1251         flush_workqueue(fsc->inode_wq);
1252
1253         /* In case that we were blocklisted. This also reset
1254          * all mon/osd connections */
1255         ceph_reset_client_addr(fsc->client);
1256
1257         ceph_osdc_clear_abort_err(&fsc->client->osdc);
1258
1259         fsc->blocklisted = false;
1260         fsc->mount_state = CEPH_MOUNT_MOUNTED;
1261
1262         if (sb->s_root) {
1263                 err = __ceph_do_getattr(d_inode(sb->s_root), NULL,
1264                                         CEPH_STAT_CAP_INODE, true);
1265         }
1266         return err;
1267 }
1268
1269 static int __init init_ceph(void)
1270 {
1271         int ret = init_caches();
1272         if (ret)
1273                 goto out;
1274
1275         ceph_flock_init();
1276         ret = register_filesystem(&ceph_fs_type);
1277         if (ret)
1278                 goto out_caches;
1279
1280         pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
1281
1282         return 0;
1283
1284 out_caches:
1285         destroy_caches();
1286 out:
1287         return ret;
1288 }
1289
1290 static void __exit exit_ceph(void)
1291 {
1292         dout("exit_ceph\n");
1293         unregister_filesystem(&ceph_fs_type);
1294         destroy_caches();
1295 }
1296
1297 static int param_set_metrics(const char *val, const struct kernel_param *kp)
1298 {
1299         struct ceph_fs_client *fsc;
1300         int ret;
1301
1302         ret = param_set_bool(val, kp);
1303         if (ret) {
1304                 pr_err("Failed to parse sending metrics switch value '%s'\n",
1305                        val);
1306                 return ret;
1307         } else if (!disable_send_metrics) {
1308                 // wake up all the mds clients
1309                 spin_lock(&ceph_fsc_lock);
1310                 list_for_each_entry(fsc, &ceph_fsc_list, metric_wakeup) {
1311                         metric_schedule_delayed(&fsc->mdsc->metric);
1312                 }
1313                 spin_unlock(&ceph_fsc_lock);
1314         }
1315
1316         return 0;
1317 }
1318
1319 static const struct kernel_param_ops param_ops_metrics = {
1320         .set = param_set_metrics,
1321         .get = param_get_bool,
1322 };
1323
1324 bool disable_send_metrics = false;
1325 module_param_cb(disable_send_metrics, &param_ops_metrics, &disable_send_metrics, 0644);
1326 MODULE_PARM_DESC(disable_send_metrics, "Enable sending perf metrics to ceph cluster (default: on)");
1327
1328 module_init(init_ceph);
1329 module_exit(exit_ceph);
1330
1331 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
1332 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
1333 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
1334 MODULE_DESCRIPTION("Ceph filesystem for Linux");
1335 MODULE_LICENSE("GPL");