fs/afs/inode.c

   1 /*
   2  * Copyright (c) 2002 Red Hat, Inc. All rights reserved.
   3  *
   4  * This software may be freely redistributed under the terms of the
   5  * GNU General Public License.
   6  *
   7  * You should have received a copy of the GNU General Public License
   8  * along with this program; if not, write to the Free Software
   9  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  10  *
  11  * Authors: David Woodhouse <dwmw2@infradead.org>
  12  *          David Howells <dhowells@redhat.com>
  13  *
  14  */
  15
  16 #include <linux/kernel.h>
  17 #include <linux/module.h>
  18 #include <linux/init.h>
  19 #include <linux/fs.h>
  20 #include <linux/pagemap.h>
  21 #include <linux/sched.h>
  22 #include <linux/mount.h>
  23 #include <linux/namei.h>
  24 #include "internal.h"
  25
  26 static const struct inode_operations afs_symlink_inode_operations = {
  27         .get_link       = page_get_link,
  28         .listxattr      = afs_listxattr,
  29 };
  30
  31 /*
  32  * map the AFS file status to the inode member variables
  33  */
  34 static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key)
  35 {
  36         struct inode *inode = AFS_VNODE_TO_I(vnode);
  37         bool changed;
  38
  39         _debug("FS: ft=%d lk=%d sz=%llu ver=%Lu mod=%hu",
  40                vnode->status.type,
  41                vnode->status.nlink,
  42                (unsigned long long) vnode->status.size,
  43                vnode->status.data_version,
  44                vnode->status.mode);
  45
  46         read_seqlock_excl(&vnode->cb_lock);
  47
  48         switch (vnode->status.type) {
  49         case AFS_FTYPE_FILE:
  50                 inode->i_mode   = S_IFREG | vnode->status.mode;
  51                 inode->i_op     = &afs_file_inode_operations;
  52                 inode->i_fop    = &afs_file_operations;
  53                 break;
  54         case AFS_FTYPE_DIR:
  55                 inode->i_mode   = S_IFDIR | vnode->status.mode;
  56                 inode->i_op     = &afs_dir_inode_operations;
  57                 inode->i_fop    = &afs_dir_file_operations;
  58                 break;
  59         case AFS_FTYPE_SYMLINK:
  60                 /* Symlinks with a mode of 0644 are actually mountpoints. */
  61                 if ((vnode->status.mode & 0777) == 0644) {
  62                         inode->i_flags |= S_AUTOMOUNT;
  63
  64                         set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
  65
  66                         inode->i_mode   = S_IFDIR | 0555;
  67                         inode->i_op     = &afs_mntpt_inode_operations;
  68                         inode->i_fop    = &afs_mntpt_file_operations;
  69                 } else {
  70                         inode->i_mode   = S_IFLNK | vnode->status.mode;
  71                         inode->i_op     = &afs_symlink_inode_operations;
  72                 }
  73                 inode_nohighmem(inode);
  74                 break;
  75         default:
  76                 printk("kAFS: AFS vnode with undefined type\n");
  77                 read_sequnlock_excl(&vnode->cb_lock);
  78                 return -EBADMSG;
  79         }
  80
  81         changed = (vnode->status.size != inode->i_size);
  82
  83         set_nlink(inode, vnode->status.nlink);
  84         inode->i_uid            = vnode->status.owner;
  85         inode->i_gid            = vnode->status.group;
  86         inode->i_size           = vnode->status.size;
  87         inode->i_ctime.tv_sec   = vnode->status.mtime_client;
  88         inode->i_ctime.tv_nsec  = 0;
  89         inode->i_atime          = inode->i_mtime = inode->i_ctime;
  90         inode->i_blocks         = 0;
  91         inode->i_generation     = vnode->fid.unique;
  92         inode->i_version        = vnode->status.data_version;
  93         inode->i_mapping->a_ops = &afs_fs_aops;
  94
  95         read_sequnlock_excl(&vnode->cb_lock);
  96
  97 #ifdef CONFIG_AFS_FSCACHE
  98         if (changed)
  99                 fscache_attr_changed(vnode->cache);
 100 #endif
 101         return 0;
 102 }
 103
 104 /*
 105  * Fetch file status from the volume.
 106  */
 107 int afs_fetch_status(struct afs_vnode *vnode, struct key *key)
 108 {
 109         struct afs_fs_cursor fc;
 110         int ret;
 111
 112         _enter("%s,{%x:%u.%u,S=%lx}",
 113                vnode->volume->name,
 114                vnode->fid.vid, vnode->fid.vnode, vnode->fid.unique,
 115                vnode->flags);
 116
 117         ret = -ERESTARTSYS;
 118         if (afs_begin_vnode_operation(&fc, vnode, key)) {
 119                 while (afs_select_fileserver(&fc)) {
 120                         fc.cb_break = vnode->cb_break + vnode->cb_s_break;
 121                         afs_fs_fetch_file_status(&fc, NULL);
 122                 }
 123
 124                 afs_check_for_remote_deletion(&fc, fc.vnode);
 125                 afs_vnode_commit_status(&fc, vnode, fc.cb_break);
 126                 ret = afs_end_vnode_operation(&fc);
 127         }
 128
 129         _leave(" = %d", ret);
 130         return ret;
 131 }
 132
 133 /*
 134  * iget5() comparator
 135  */
 136 int afs_iget5_test(struct inode *inode, void *opaque)
 137 {
 138         struct afs_iget_data *data = opaque;
 139
 140         return inode->i_ino == data->fid.vnode &&
 141                 inode->i_generation == data->fid.unique;
 142 }
 143
 144 /*
 145  * iget5() comparator for inode created by autocell operations
 146  *
 147  * These pseudo inodes don't match anything.
 148  */
 149 static int afs_iget5_autocell_test(struct inode *inode, void *opaque)
 150 {
 151         return 0;
 152 }
 153
 154 /*
 155  * iget5() inode initialiser
 156  */
 157 static int afs_iget5_set(struct inode *inode, void *opaque)
 158 {
 159         struct afs_iget_data *data = opaque;
 160         struct afs_vnode *vnode = AFS_FS_I(inode);
 161
 162         inode->i_ino = data->fid.vnode;
 163         inode->i_generation = data->fid.unique;
 164         vnode->fid = data->fid;
 165         vnode->volume = data->volume;
 166
 167         return 0;
 168 }
 169
 170 /*
 171  * inode retrieval for autocell
 172  */
 173 struct inode *afs_iget_autocell(struct inode *dir, const char *dev_name,
 174                                 int namesz, struct key *key)
 175 {
 176         struct afs_iget_data data;
 177         struct afs_super_info *as;
 178         struct afs_vnode *vnode;
 179         struct super_block *sb;
 180         struct inode *inode;
 181         static atomic_t afs_autocell_ino;
 182
 183         _enter("{%x:%u},%*.*s,",
 184                AFS_FS_I(dir)->fid.vid, AFS_FS_I(dir)->fid.vnode,
 185                namesz, namesz, dev_name ?: "");
 186
 187         sb = dir->i_sb;
 188         as = sb->s_fs_info;
 189         data.volume = as->volume;
 190         data.fid.vid = as->volume->vid;
 191         data.fid.unique = 0;
 192         data.fid.vnode = 0;
 193
 194         inode = iget5_locked(sb, atomic_inc_return(&afs_autocell_ino),
 195                              afs_iget5_autocell_test, afs_iget5_set,
 196                              &data);
 197         if (!inode) {
 198                 _leave(" = -ENOMEM");
 199                 return ERR_PTR(-ENOMEM);
 200         }
 201
 202         _debug("GOT INODE %p { ino=%lu, vl=%x, vn=%x, u=%x }",
 203                inode, inode->i_ino, data.fid.vid, data.fid.vnode,
 204                data.fid.unique);
 205
 206         vnode = AFS_FS_I(inode);
 207
 208         /* there shouldn't be an existing inode */
 209         BUG_ON(!(inode->i_state & I_NEW));
 210
 211         inode->i_size           = 0;
 212         inode->i_mode           = S_IFDIR | S_IRUGO | S_IXUGO;
 213         inode->i_op             = &afs_autocell_inode_operations;
 214         set_nlink(inode, 2);
 215         inode->i_uid            = GLOBAL_ROOT_UID;
 216         inode->i_gid            = GLOBAL_ROOT_GID;
 217         inode->i_ctime.tv_sec   = get_seconds();
 218         inode->i_ctime.tv_nsec  = 0;
 219         inode->i_atime          = inode->i_mtime = inode->i_ctime;
 220         inode->i_blocks         = 0;
 221         inode->i_version        = 0;
 222         inode->i_generation     = 0;
 223
 224         set_bit(AFS_VNODE_PSEUDODIR, &vnode->flags);
 225         set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
 226         inode->i_flags |= S_AUTOMOUNT | S_NOATIME;
 227         unlock_new_inode(inode);
 228         _leave(" = %p", inode);
 229         return inode;
 230 }
 231
 232 /*
 233  * inode retrieval
 234  */
 235 struct inode *afs_iget(struct super_block *sb, struct key *key,
 236                        struct afs_fid *fid, struct afs_file_status *status,
 237                        struct afs_callback *cb, struct afs_cb_interest *cbi)
 238 {
 239         struct afs_iget_data data = { .fid = *fid };
 240         struct afs_super_info *as;
 241         struct afs_vnode *vnode;
 242         struct inode *inode;
 243         int ret;
 244
 245         _enter(",{%x:%u.%u},,", fid->vid, fid->vnode, fid->unique);
 246
 247         as = sb->s_fs_info;
 248         data.volume = as->volume;
 249
 250         inode = iget5_locked(sb, fid->vnode, afs_iget5_test, afs_iget5_set,
 251                              &data);
 252         if (!inode) {
 253                 _leave(" = -ENOMEM");
 254                 return ERR_PTR(-ENOMEM);
 255         }
 256
 257         _debug("GOT INODE %p { vl=%x vn=%x, u=%x }",
 258                inode, fid->vid, fid->vnode, fid->unique);
 259
 260         vnode = AFS_FS_I(inode);
 261
 262         /* deal with an existing inode */
 263         if (!(inode->i_state & I_NEW)) {
 264                 _leave(" = %p", inode);
 265                 return inode;
 266         }
 267
 268         if (!status) {
 269                 /* it's a remotely extant inode */
 270                 ret = afs_fetch_status(vnode, key);
 271                 if (ret < 0)
 272                         goto bad_inode;
 273         } else {
 274                 /* it's an inode we just created */
 275                 memcpy(&vnode->status, status, sizeof(vnode->status));
 276
 277                 if (!cb) {
 278                         /* it's a symlink we just created (the fileserver
 279                          * didn't give us a callback) */
 280                         vnode->cb_version = 0;
 281                         vnode->cb_type = 0;
 282                         vnode->cb_expires_at = 0;
 283                 } else {
 284                         vnode->cb_version = cb->version;
 285                         vnode->cb_type = cb->type;
 286                         vnode->cb_expires_at = cb->expiry;
 287                         vnode->cb_interest = afs_get_cb_interest(cbi);
 288                         set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
 289                 }
 290
 291                 vnode->cb_expires_at += ktime_get_real_seconds();
 292         }
 293
 294         /* set up caching before mapping the status, as map-status reads the
 295          * first page of symlinks to see if they're really mountpoints */
 296         inode->i_size = vnode->status.size;
 297 #ifdef CONFIG_AFS_FSCACHE
 298         vnode->cache = fscache_acquire_cookie(vnode->volume->cache,
 299                                               &afs_vnode_cache_index_def,
 300                                               vnode, true);
 301 #endif
 302
 303         ret = afs_inode_map_status(vnode, key);
 304         if (ret < 0)
 305                 goto bad_inode;
 306
 307         /* success */
 308         clear_bit(AFS_VNODE_UNSET, &vnode->flags);
 309         inode->i_flags |= S_NOATIME;
 310         unlock_new_inode(inode);
 311         _leave(" = %p [CB { v=%u t=%u }]", inode, vnode->cb_version, vnode->cb_type);
 312         return inode;
 313
 314         /* failure */
 315 bad_inode:
 316 #ifdef CONFIG_AFS_FSCACHE
 317         fscache_relinquish_cookie(vnode->cache, 0);
 318         vnode->cache = NULL;
 319 #endif
 320         iget_failed(inode);
 321         _leave(" = %d [bad]", ret);
 322         return ERR_PTR(ret);
 323 }
 324
 325 /*
 326  * mark the data attached to an inode as obsolete due to a write on the server
 327  * - might also want to ditch all the outstanding writes and dirty pages
 328  */
 329 void afs_zap_data(struct afs_vnode *vnode)
 330 {
 331         _enter("{%x:%u}", vnode->fid.vid, vnode->fid.vnode);
 332
 333         /* nuke all the non-dirty pages that aren't locked, mapped or being
 334          * written back in a regular file and completely discard the pages in a
 335          * directory or symlink */
 336         if (S_ISREG(vnode->vfs_inode.i_mode))
 337                 invalidate_remote_inode(&vnode->vfs_inode);
 338         else
 339                 invalidate_inode_pages2(vnode->vfs_inode.i_mapping);
 340 }
 341
 342 /*
 343  * validate a vnode/inode
 344  * - there are several things we need to check
 345  *   - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
 346  *     symlink)
 347  *   - parent dir metadata changed (security changes)
 348  *   - dentry data changed (write, truncate)
 349  *   - dentry metadata changed (security changes)
 350  */
 351 int afs_validate(struct afs_vnode *vnode, struct key *key)
 352 {
 353         time64_t now = ktime_get_real_seconds();
 354         bool valid = false;
 355         int ret;
 356
 357         _enter("{v={%x:%u} fl=%lx},%x",
 358                vnode->fid.vid, vnode->fid.vnode, vnode->flags,
 359                key_serial(key));
 360
 361         /* Quickly check the callback state.  Ideally, we'd use read_seqbegin
 362          * here, but we have no way to pass the net namespace to the RCU
 363          * cleanup for the server record.
 364          */
 365         read_seqlock_excl(&vnode->cb_lock);
 366
 367         if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
 368                 if (vnode->cb_s_break != vnode->cb_interest->server->cb_s_break) {
 369                         vnode->cb_s_break = vnode->cb_interest->server->cb_s_break;
 370                 } else if (!test_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags) &&
 371                            !test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags) &&
 372                            vnode->cb_expires_at - 10 > now) {
 373                                 valid = true;
 374                 }
 375         } else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
 376                 valid = true;
 377         }
 378
 379         read_sequnlock_excl(&vnode->cb_lock);
 380
 381         if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
 382                 clear_nlink(&vnode->vfs_inode);
 383
 384         if (valid)
 385                 goto valid;
 386
 387         mutex_lock(&vnode->validate_lock);
 388
 389         /* if the promise has expired, we need to check the server again to get
 390          * a new promise - note that if the (parent) directory's metadata was
 391          * changed then the security may be different and we may no longer have
 392          * access */
 393         if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
 394                 _debug("not promised");
 395                 ret = afs_fetch_status(vnode, key);
 396                 if (ret < 0) {
 397                         if (ret == -ENOENT) {
 398                                 set_bit(AFS_VNODE_DELETED, &vnode->flags);
 399                                 ret = -ESTALE;
 400                         }
 401                         goto error_unlock;
 402                 }
 403                 _debug("new promise [fl=%lx]", vnode->flags);
 404         }
 405
 406         if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
 407                 _debug("file already deleted");
 408                 ret = -ESTALE;
 409                 goto error_unlock;
 410         }
 411
 412         /* if the vnode's data version number changed then its contents are
 413          * different */
 414         if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
 415                 afs_zap_data(vnode);
 416
 417         clear_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags);
 418         mutex_unlock(&vnode->validate_lock);
 419 valid:
 420         _leave(" = 0");
 421         return 0;
 422
 423 error_unlock:
 424         mutex_unlock(&vnode->validate_lock);
 425         _leave(" = %d", ret);
 426         return ret;
 427 }
 428
 429 /*
 430  * read the attributes of an inode
 431  */
 432 int afs_getattr(const struct path *path, struct kstat *stat,
 433                 u32 request_mask, unsigned int query_flags)
 434 {
 435         struct inode *inode = d_inode(path->dentry);
 436         struct afs_vnode *vnode = AFS_FS_I(inode);
 437         int seq = 0;
 438
 439         _enter("{ ino=%lu v=%u }", inode->i_ino, inode->i_generation);
 440
 441         do {
 442                 read_seqbegin_or_lock(&vnode->cb_lock, &seq);
 443                 generic_fillattr(inode, stat);
 444         } while (need_seqretry(&vnode->cb_lock, seq));
 445
 446         done_seqretry(&vnode->cb_lock, seq);
 447         return 0;
 448 }
 449
 450 /*
 451  * discard an AFS inode
 452  */
 453 int afs_drop_inode(struct inode *inode)
 454 {
 455         _enter("");
 456
 457         if (test_bit(AFS_VNODE_PSEUDODIR, &AFS_FS_I(inode)->flags))
 458                 return generic_delete_inode(inode);
 459         else
 460                 return generic_drop_inode(inode);
 461 }
 462
 463 /*
 464  * clear an AFS inode
 465  */
 466 void afs_evict_inode(struct inode *inode)
 467 {
 468         struct afs_vnode *vnode;
 469
 470         vnode = AFS_FS_I(inode);
 471
 472         _enter("{%x:%u.%d}",
 473                vnode->fid.vid,
 474                vnode->fid.vnode,
 475                vnode->fid.unique);
 476
 477         _debug("CLEAR INODE %p", inode);
 478
 479         ASSERTCMP(inode->i_ino, ==, vnode->fid.vnode);
 480
 481         truncate_inode_pages_final(&inode->i_data);
 482         clear_inode(inode);
 483
 484         if (vnode->cb_interest) {
 485                 afs_put_cb_interest(afs_i2net(inode), vnode->cb_interest);
 486                 vnode->cb_interest = NULL;
 487         }
 488
 489         while (!list_empty(&vnode->wb_keys)) {
 490                 struct afs_wb_key *wbk = list_entry(vnode->wb_keys.next,
 491                                                     struct afs_wb_key, vnode_link);
 492                 list_del(&wbk->vnode_link);
 493                 afs_put_wb_key(wbk);
 494         }
 495
 496 #ifdef CONFIG_AFS_FSCACHE
 497         fscache_relinquish_cookie(vnode->cache, 0);
 498         vnode->cache = NULL;
 499 #endif
 500
 501         afs_put_permits(vnode->permit_cache);
 502         _leave("");
 503 }
 504
 505 /*
 506  * set the attributes of an inode
 507  */
 508 int afs_setattr(struct dentry *dentry, struct iattr *attr)
 509 {
 510         struct afs_fs_cursor fc;
 511         struct afs_vnode *vnode = AFS_FS_I(d_inode(dentry));
 512         struct key *key;
 513         int ret;
 514
 515         _enter("{%x:%u},{n=%pd},%x",
 516                vnode->fid.vid, vnode->fid.vnode, dentry,
 517                attr->ia_valid);
 518
 519         if (!(attr->ia_valid & (ATTR_SIZE | ATTR_MODE | ATTR_UID | ATTR_GID |
 520                                 ATTR_MTIME))) {
 521                 _leave(" = 0 [unsupported]");
 522                 return 0;
 523         }
 524
 525         /* flush any dirty data outstanding on a regular file */
 526         if (S_ISREG(vnode->vfs_inode.i_mode))
 527                 filemap_write_and_wait(vnode->vfs_inode.i_mapping);
 528
 529         if (attr->ia_valid & ATTR_FILE) {
 530                 key = afs_file_key(attr->ia_file);
 531         } else {
 532                 key = afs_request_key(vnode->volume->cell);
 533                 if (IS_ERR(key)) {
 534                         ret = PTR_ERR(key);
 535                         goto error;
 536                 }
 537         }
 538
 539         ret = -ERESTARTSYS;
 540         if (afs_begin_vnode_operation(&fc, vnode, key)) {
 541                 while (afs_select_fileserver(&fc)) {
 542                         fc.cb_break = vnode->cb_break + vnode->cb_s_break;
 543                         afs_fs_setattr(&fc, attr);
 544                 }
 545
 546                 afs_check_for_remote_deletion(&fc, fc.vnode);
 547                 afs_vnode_commit_status(&fc, vnode, fc.cb_break);
 548                 ret = afs_end_vnode_operation(&fc);
 549         }
 550
 551         if (!(attr->ia_valid & ATTR_FILE))
 552                 key_put(key);
 553
 554 error:
 555         _leave(" = %d", ret);
 556         return ret;
 557 }