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 <linux/iversion.h>
  25 #include "internal.h"
  26
  27 static const struct inode_operations afs_symlink_inode_operations = {
  28         .get_link       = page_get_link,
  29         .listxattr      = afs_listxattr,
  30 };
  31
  32 /*
  33  * map the AFS file status to the inode member variables
  34  */
  35 static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key)
  36 {
  37         struct inode *inode = AFS_VNODE_TO_I(vnode);
  38         bool changed;
  39
  40         _debug("FS: ft=%d lk=%d sz=%llu ver=%Lu mod=%hu",
  41                vnode->status.type,
  42                vnode->status.nlink,
  43                (unsigned long long) vnode->status.size,
  44                vnode->status.data_version,
  45                vnode->status.mode);
  46
  47         read_seqlock_excl(&vnode->cb_lock);
  48
  49         switch (vnode->status.type) {
  50         case AFS_FTYPE_FILE:
  51                 inode->i_mode   = S_IFREG | vnode->status.mode;
  52                 inode->i_op     = &afs_file_inode_operations;
  53                 inode->i_fop    = &afs_file_operations;
  54                 break;
  55         case AFS_FTYPE_DIR:
  56                 inode->i_mode   = S_IFDIR | vnode->status.mode;
  57                 inode->i_op     = &afs_dir_inode_operations;
  58                 inode->i_fop    = &afs_dir_file_operations;
  59                 break;
  60         case AFS_FTYPE_SYMLINK:
  61                 /* Symlinks with a mode of 0644 are actually mountpoints. */
  62                 if ((vnode->status.mode & 0777) == 0644) {
  63                         inode->i_flags |= S_AUTOMOUNT;
  64
  65                         set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
  66
  67                         inode->i_mode   = S_IFDIR | 0555;
  68                         inode->i_op     = &afs_mntpt_inode_operations;
  69                         inode->i_fop    = &afs_mntpt_file_operations;
  70                 } else {
  71                         inode->i_mode   = S_IFLNK | vnode->status.mode;
  72                         inode->i_op     = &afs_symlink_inode_operations;
  73                 }
  74                 inode_nohighmem(inode);
  75                 break;
  76         default:
  77                 printk("kAFS: AFS vnode with undefined type\n");
  78                 read_sequnlock_excl(&vnode->cb_lock);
  79                 return -EBADMSG;
  80         }
  81
  82         changed = (vnode->status.size != inode->i_size);
  83
  84         set_nlink(inode, vnode->status.nlink);
  85         inode->i_uid            = vnode->status.owner;
  86         inode->i_gid            = vnode->status.group;
  87         inode->i_size           = vnode->status.size;
  88         inode->i_ctime.tv_sec   = vnode->status.mtime_client;
  89         inode->i_ctime.tv_nsec  = 0;
  90         inode->i_atime          = inode->i_mtime = inode->i_ctime;
  91         inode->i_blocks         = 0;
  92         inode->i_generation     = vnode->fid.unique;
  93         inode_set_iversion_raw(inode, vnode->status.data_version);
  94         inode->i_mapping->a_ops = &afs_fs_aops;
  95
  96         read_sequnlock_excl(&vnode->cb_lock);
  97
  98 #ifdef CONFIG_AFS_FSCACHE
  99         if (changed)
 100                 fscache_attr_changed(vnode->cache);
 101 #endif
 102         return 0;
 103 }
 104
 105 /*
 106  * Fetch file status from the volume.
 107  */
 108 int afs_fetch_status(struct afs_vnode *vnode, struct key *key)
 109 {
 110         struct afs_fs_cursor fc;
 111         int ret;
 112
 113         _enter("%s,{%x:%u.%u,S=%lx}",
 114                vnode->volume->name,
 115                vnode->fid.vid, vnode->fid.vnode, vnode->fid.unique,
 116                vnode->flags);
 117
 118         ret = -ERESTARTSYS;
 119         if (afs_begin_vnode_operation(&fc, vnode, key)) {
 120                 while (afs_select_fileserver(&fc)) {
 121                         fc.cb_break = vnode->cb_break + vnode->cb_s_break;
 122                         afs_fs_fetch_file_status(&fc, NULL);
 123                 }
 124
 125                 afs_check_for_remote_deletion(&fc, fc.vnode);
 126                 afs_vnode_commit_status(&fc, vnode, fc.cb_break);
 127                 ret = afs_end_vnode_operation(&fc);
 128         }
 129
 130         _leave(" = %d", ret);
 131         return ret;
 132 }
 133
 134 /*
 135  * iget5() comparator
 136  */
 137 int afs_iget5_test(struct inode *inode, void *opaque)
 138 {
 139         struct afs_iget_data *data = opaque;
 140
 141         return inode->i_ino == data->fid.vnode &&
 142                 inode->i_generation == data->fid.unique;
 143 }
 144
 145 /*
 146  * iget5() comparator for inode created by autocell operations
 147  *
 148  * These pseudo inodes don't match anything.
 149  */
 150 static int afs_iget5_pseudo_dir_test(struct inode *inode, void *opaque)
 151 {
 152         return 0;
 153 }
 154
 155 /*
 156  * iget5() inode initialiser
 157  */
 158 static int afs_iget5_set(struct inode *inode, void *opaque)
 159 {
 160         struct afs_iget_data *data = opaque;
 161         struct afs_vnode *vnode = AFS_FS_I(inode);
 162
 163         inode->i_ino = data->fid.vnode;
 164         inode->i_generation = data->fid.unique;
 165         vnode->fid = data->fid;
 166         vnode->volume = data->volume;
 167
 168         return 0;
 169 }
 170
 171 /*
 172  * Create an inode for a dynamic root directory or an autocell dynamic
 173  * automount dir.
 174  */
 175 struct inode *afs_iget_pseudo_dir(struct super_block *sb, bool root)
 176 {
 177         struct afs_iget_data data;
 178         struct afs_super_info *as;
 179         struct afs_vnode *vnode;
 180         struct inode *inode;
 181         static atomic_t afs_autocell_ino;
 182
 183         _enter("");
 184
 185         as = sb->s_fs_info;
 186         if (as->volume) {
 187                 data.volume = as->volume;
 188                 data.fid.vid = as->volume->vid;
 189         }
 190         if (root) {
 191                 data.fid.vnode = 1;
 192                 data.fid.unique = 1;
 193         } else {
 194                 data.fid.vnode = atomic_inc_return(&afs_autocell_ino);
 195                 data.fid.unique = 0;
 196         }
 197
 198         inode = iget5_locked(sb, data.fid.vnode,
 199                              afs_iget5_pseudo_dir_test, afs_iget5_set,
 200                              &data);
 201         if (!inode) {
 202                 _leave(" = -ENOMEM");
 203                 return ERR_PTR(-ENOMEM);
 204         }
 205
 206         _debug("GOT INODE %p { ino=%lu, vl=%x, vn=%x, u=%x }",
 207                inode, inode->i_ino, data.fid.vid, data.fid.vnode,
 208                data.fid.unique);
 209
 210         vnode = AFS_FS_I(inode);
 211
 212         /* there shouldn't be an existing inode */
 213         BUG_ON(!(inode->i_state & I_NEW));
 214
 215         inode->i_size           = 0;
 216         inode->i_mode           = S_IFDIR | S_IRUGO | S_IXUGO;
 217         if (root) {
 218                 inode->i_op     = &afs_dynroot_inode_operations;
 219                 inode->i_fop    = &afs_dynroot_file_operations;
 220         } else {
 221                 inode->i_op     = &afs_autocell_inode_operations;
 222         }
 223         set_nlink(inode, 2);
 224         inode->i_uid            = GLOBAL_ROOT_UID;
 225         inode->i_gid            = GLOBAL_ROOT_GID;
 226         inode->i_ctime.tv_sec   = get_seconds();
 227         inode->i_ctime.tv_nsec  = 0;
 228         inode->i_atime          = inode->i_mtime = inode->i_ctime;
 229         inode->i_blocks         = 0;
 230         inode_set_iversion_raw(inode, 0);
 231         inode->i_generation     = 0;
 232
 233         set_bit(AFS_VNODE_PSEUDODIR, &vnode->flags);
 234         if (!root) {
 235                 set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
 236                 inode->i_flags |= S_AUTOMOUNT;
 237         }
 238
 239         inode->i_flags |= S_NOATIME;
 240         unlock_new_inode(inode);
 241         _leave(" = %p", inode);
 242         return inode;
 243 }
 244
 245 /*
 246  * inode retrieval
 247  */
 248 struct inode *afs_iget(struct super_block *sb, struct key *key,
 249                        struct afs_fid *fid, struct afs_file_status *status,
 250                        struct afs_callback *cb, struct afs_cb_interest *cbi)
 251 {
 252         struct afs_iget_data data = { .fid = *fid };
 253         struct afs_super_info *as;
 254         struct afs_vnode *vnode;
 255         struct inode *inode;
 256         int ret;
 257
 258         _enter(",{%x:%u.%u},,", fid->vid, fid->vnode, fid->unique);
 259
 260         as = sb->s_fs_info;
 261         data.volume = as->volume;
 262
 263         inode = iget5_locked(sb, fid->vnode, afs_iget5_test, afs_iget5_set,
 264                              &data);
 265         if (!inode) {
 266                 _leave(" = -ENOMEM");
 267                 return ERR_PTR(-ENOMEM);
 268         }
 269
 270         _debug("GOT INODE %p { vl=%x vn=%x, u=%x }",
 271                inode, fid->vid, fid->vnode, fid->unique);
 272
 273         vnode = AFS_FS_I(inode);
 274
 275         /* deal with an existing inode */
 276         if (!(inode->i_state & I_NEW)) {
 277                 _leave(" = %p", inode);
 278                 return inode;
 279         }
 280
 281         if (!status) {
 282                 /* it's a remotely extant inode */
 283                 ret = afs_fetch_status(vnode, key);
 284                 if (ret < 0)
 285                         goto bad_inode;
 286         } else {
 287                 /* it's an inode we just created */
 288                 memcpy(&vnode->status, status, sizeof(vnode->status));
 289
 290                 if (!cb) {
 291                         /* it's a symlink we just created (the fileserver
 292                          * didn't give us a callback) */
 293                         vnode->cb_version = 0;
 294                         vnode->cb_type = 0;
 295                         vnode->cb_expires_at = 0;
 296                 } else {
 297                         vnode->cb_version = cb->version;
 298                         vnode->cb_type = cb->type;
 299                         vnode->cb_expires_at = cb->expiry;
 300                         vnode->cb_interest = afs_get_cb_interest(cbi);
 301                         set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
 302                 }
 303
 304                 vnode->cb_expires_at += ktime_get_real_seconds();
 305         }
 306
 307         /* set up caching before mapping the status, as map-status reads the
 308          * first page of symlinks to see if they're really mountpoints */
 309         inode->i_size = vnode->status.size;
 310 #ifdef CONFIG_AFS_FSCACHE
 311         vnode->cache = fscache_acquire_cookie(vnode->volume->cache,
 312                                               &afs_vnode_cache_index_def,
 313                                               vnode, true);
 314 #endif
 315
 316         ret = afs_inode_map_status(vnode, key);
 317         if (ret < 0)
 318                 goto bad_inode;
 319
 320         /* success */
 321         clear_bit(AFS_VNODE_UNSET, &vnode->flags);
 322         inode->i_flags |= S_NOATIME;
 323         unlock_new_inode(inode);
 324         _leave(" = %p [CB { v=%u t=%u }]", inode, vnode->cb_version, vnode->cb_type);
 325         return inode;
 326
 327         /* failure */
 328 bad_inode:
 329 #ifdef CONFIG_AFS_FSCACHE
 330         fscache_relinquish_cookie(vnode->cache, 0);
 331         vnode->cache = NULL;
 332 #endif
 333         iget_failed(inode);
 334         _leave(" = %d [bad]", ret);
 335         return ERR_PTR(ret);
 336 }
 337
 338 /*
 339  * mark the data attached to an inode as obsolete due to a write on the server
 340  * - might also want to ditch all the outstanding writes and dirty pages
 341  */
 342 void afs_zap_data(struct afs_vnode *vnode)
 343 {
 344         _enter("{%x:%u}", vnode->fid.vid, vnode->fid.vnode);
 345
 346         /* nuke all the non-dirty pages that aren't locked, mapped or being
 347          * written back in a regular file and completely discard the pages in a
 348          * directory or symlink */
 349         if (S_ISREG(vnode->vfs_inode.i_mode))
 350                 invalidate_remote_inode(&vnode->vfs_inode);
 351         else
 352                 invalidate_inode_pages2(vnode->vfs_inode.i_mapping);
 353 }
 354
 355 /*
 356  * validate a vnode/inode
 357  * - there are several things we need to check
 358  *   - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
 359  *     symlink)
 360  *   - parent dir metadata changed (security changes)
 361  *   - dentry data changed (write, truncate)
 362  *   - dentry metadata changed (security changes)
 363  */
 364 int afs_validate(struct afs_vnode *vnode, struct key *key)
 365 {
 366         time64_t now = ktime_get_real_seconds();
 367         bool valid = false;
 368         int ret;
 369
 370         _enter("{v={%x:%u} fl=%lx},%x",
 371                vnode->fid.vid, vnode->fid.vnode, vnode->flags,
 372                key_serial(key));
 373
 374         /* Quickly check the callback state.  Ideally, we'd use read_seqbegin
 375          * here, but we have no way to pass the net namespace to the RCU
 376          * cleanup for the server record.
 377          */
 378         read_seqlock_excl(&vnode->cb_lock);
 379
 380         if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
 381                 if (vnode->cb_s_break != vnode->cb_interest->server->cb_s_break) {
 382                         vnode->cb_s_break = vnode->cb_interest->server->cb_s_break;
 383                 } else if (!test_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags) &&
 384                            !test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags) &&
 385                            vnode->cb_expires_at - 10 > now) {
 386                                 valid = true;
 387                 }
 388         } else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
 389                 valid = true;
 390         }
 391
 392         read_sequnlock_excl(&vnode->cb_lock);
 393
 394         if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
 395                 clear_nlink(&vnode->vfs_inode);
 396
 397         if (valid)
 398                 goto valid;
 399
 400         mutex_lock(&vnode->validate_lock);
 401
 402         /* if the promise has expired, we need to check the server again to get
 403          * a new promise - note that if the (parent) directory's metadata was
 404          * changed then the security may be different and we may no longer have
 405          * access */
 406         if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
 407                 _debug("not promised");
 408                 ret = afs_fetch_status(vnode, key);
 409                 if (ret < 0) {
 410                         if (ret == -ENOENT) {
 411                                 set_bit(AFS_VNODE_DELETED, &vnode->flags);
 412                                 ret = -ESTALE;
 413                         }
 414                         goto error_unlock;
 415                 }
 416                 _debug("new promise [fl=%lx]", vnode->flags);
 417         }
 418
 419         if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
 420                 _debug("file already deleted");
 421                 ret = -ESTALE;
 422                 goto error_unlock;
 423         }
 424
 425         /* if the vnode's data version number changed then its contents are
 426          * different */
 427         if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
 428                 afs_zap_data(vnode);
 429
 430         clear_bit(AFS_VNODE_DIR_MODIFIED, &vnode->flags);
 431         mutex_unlock(&vnode->validate_lock);
 432 valid:
 433         _leave(" = 0");
 434         return 0;
 435
 436 error_unlock:
 437         mutex_unlock(&vnode->validate_lock);
 438         _leave(" = %d", ret);
 439         return ret;
 440 }
 441
 442 /*
 443  * read the attributes of an inode
 444  */
 445 int afs_getattr(const struct path *path, struct kstat *stat,
 446                 u32 request_mask, unsigned int query_flags)
 447 {
 448         struct inode *inode = d_inode(path->dentry);
 449         struct afs_vnode *vnode = AFS_FS_I(inode);
 450         int seq = 0;
 451
 452         _enter("{ ino=%lu v=%u }", inode->i_ino, inode->i_generation);
 453
 454         do {
 455                 read_seqbegin_or_lock(&vnode->cb_lock, &seq);
 456                 generic_fillattr(inode, stat);
 457         } while (need_seqretry(&vnode->cb_lock, seq));
 458
 459         done_seqretry(&vnode->cb_lock, seq);
 460         return 0;
 461 }
 462
 463 /*
 464  * discard an AFS inode
 465  */
 466 int afs_drop_inode(struct inode *inode)
 467 {
 468         _enter("");
 469
 470         if (test_bit(AFS_VNODE_PSEUDODIR, &AFS_FS_I(inode)->flags))
 471                 return generic_delete_inode(inode);
 472         else
 473                 return generic_drop_inode(inode);
 474 }
 475
 476 /*
 477  * clear an AFS inode
 478  */
 479 void afs_evict_inode(struct inode *inode)
 480 {
 481         struct afs_vnode *vnode;
 482
 483         vnode = AFS_FS_I(inode);
 484
 485         _enter("{%x:%u.%d}",
 486                vnode->fid.vid,
 487                vnode->fid.vnode,
 488                vnode->fid.unique);
 489
 490         _debug("CLEAR INODE %p", inode);
 491
 492         ASSERTCMP(inode->i_ino, ==, vnode->fid.vnode);
 493
 494         truncate_inode_pages_final(&inode->i_data);
 495         clear_inode(inode);
 496
 497         if (vnode->cb_interest) {
 498                 afs_put_cb_interest(afs_i2net(inode), vnode->cb_interest);
 499                 vnode->cb_interest = NULL;
 500         }
 501
 502         while (!list_empty(&vnode->wb_keys)) {
 503                 struct afs_wb_key *wbk = list_entry(vnode->wb_keys.next,
 504                                                     struct afs_wb_key, vnode_link);
 505                 list_del(&wbk->vnode_link);
 506                 afs_put_wb_key(wbk);
 507         }
 508
 509 #ifdef CONFIG_AFS_FSCACHE
 510         fscache_relinquish_cookie(vnode->cache, 0);
 511         vnode->cache = NULL;
 512 #endif
 513
 514         afs_put_permits(vnode->permit_cache);
 515         _leave("");
 516 }
 517
 518 /*
 519  * set the attributes of an inode
 520  */
 521 int afs_setattr(struct dentry *dentry, struct iattr *attr)
 522 {
 523         struct afs_fs_cursor fc;
 524         struct afs_vnode *vnode = AFS_FS_I(d_inode(dentry));
 525         struct key *key;
 526         int ret;
 527
 528         _enter("{%x:%u},{n=%pd},%x",
 529                vnode->fid.vid, vnode->fid.vnode, dentry,
 530                attr->ia_valid);
 531
 532         if (!(attr->ia_valid & (ATTR_SIZE | ATTR_MODE | ATTR_UID | ATTR_GID |
 533                                 ATTR_MTIME))) {
 534                 _leave(" = 0 [unsupported]");
 535                 return 0;
 536         }
 537
 538         /* flush any dirty data outstanding on a regular file */
 539         if (S_ISREG(vnode->vfs_inode.i_mode))
 540                 filemap_write_and_wait(vnode->vfs_inode.i_mapping);
 541
 542         if (attr->ia_valid & ATTR_FILE) {
 543                 key = afs_file_key(attr->ia_file);
 544         } else {
 545                 key = afs_request_key(vnode->volume->cell);
 546                 if (IS_ERR(key)) {
 547                         ret = PTR_ERR(key);
 548                         goto error;
 549                 }
 550         }
 551
 552         ret = -ERESTARTSYS;
 553         if (afs_begin_vnode_operation(&fc, vnode, key)) {
 554                 while (afs_select_fileserver(&fc)) {
 555                         fc.cb_break = vnode->cb_break + vnode->cb_s_break;
 556                         afs_fs_setattr(&fc, attr);
 557                 }
 558
 559                 afs_check_for_remote_deletion(&fc, fc.vnode);
 560                 afs_vnode_commit_status(&fc, vnode, fc.cb_break);
 561                 ret = afs_end_vnode_operation(&fc);
 562         }
 563
 564         if (!(attr->ia_valid & ATTR_FILE))
 565                 key_put(key);
 566
 567 error:
 568         _leave(" = %d", ret);
 569         return ret;
 570 }