fs/libfs.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
simple_getattr
simple_statfs
always_delete_dentry
simple_lookup
dcache_dir_open
dcache_dir_close
scan_positives
dcache_dir_lseek
dt_type
dcache_readdir
generic_read_dir
pseudo_fs_fill_super
pseudo_fs_get_tree
pseudo_fs_free
init_pseudo
simple_open
simple_link
simple_empty
simple_unlink
simple_rmdir
simple_rename
simple_setattr
simple_readpage
simple_write_begin
simple_write_end
simple_fill_super
simple_pin_fs
simple_release_fs
simple_read_from_buffer
simple_write_to_buffer
memory_read_from_buffer
simple_transaction_set
simple_transaction_get
simple_transaction_read
simple_transaction_release
simple_attr_open
simple_attr_release
simple_attr_read
simple_attr_write
generic_fh_to_dentry
generic_fh_to_parent
__generic_file_fsync
generic_file_fsync
generic_check_addressable
noop_fsync
noop_set_page_dirty
noop_invalidatepage
noop_direct_IO
kfree_link
anon_set_page_dirty
alloc_anon_inode
simple_nosetlease
simple_get_link
empty_dir_lookup
empty_dir_getattr
empty_dir_setattr
empty_dir_listxattr
empty_dir_llseek
empty_dir_readdir
make_empty_dir_inode
is_empty_dir_inode
   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *      fs/libfs.c
   4  *      Library for filesystems writers.
   5  */
   6 
   7 #include <linux/blkdev.h>
   8 #include <linux/export.h>
   9 #include <linux/pagemap.h>
  10 #include <linux/slab.h>
  11 #include <linux/cred.h>
  12 #include <linux/mount.h>
  13 #include <linux/vfs.h>
  14 #include <linux/quotaops.h>
  15 #include <linux/mutex.h>
  16 #include <linux/namei.h>
  17 #include <linux/exportfs.h>
  18 #include <linux/writeback.h>
  19 #include <linux/buffer_head.h> /* sync_mapping_buffers */
  20 #include <linux/fs_context.h>
  21 #include <linux/pseudo_fs.h>
  22 
  23 #include <linux/uaccess.h>
  24 
  25 #include "internal.h"
  26 
  27 int simple_getattr(const struct path *path, struct kstat *stat,
  28                    u32 request_mask, unsigned int query_flags)
  29 {
  30         struct inode *inode = d_inode(path->dentry);
  31         generic_fillattr(inode, stat);
  32         stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
  33         return 0;
  34 }
  35 EXPORT_SYMBOL(simple_getattr);
  36 
  37 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
  38 {
  39         buf->f_type = dentry->d_sb->s_magic;
  40         buf->f_bsize = PAGE_SIZE;
  41         buf->f_namelen = NAME_MAX;
  42         return 0;
  43 }
  44 EXPORT_SYMBOL(simple_statfs);
  45 
  46 /*
  47  * Retaining negative dentries for an in-memory filesystem just wastes
  48  * memory and lookup time: arrange for them to be deleted immediately.
  49  */
  50 int always_delete_dentry(const struct dentry *dentry)
  51 {
  52         return 1;
  53 }
  54 EXPORT_SYMBOL(always_delete_dentry);
  55 
  56 const struct dentry_operations simple_dentry_operations = {
  57         .d_delete = always_delete_dentry,
  58 };
  59 EXPORT_SYMBOL(simple_dentry_operations);
  60 
  61 /*
  62  * Lookup the data. This is trivial - if the dentry didn't already
  63  * exist, we know it is negative.  Set d_op to delete negative dentries.
  64  */
  65 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
  66 {
  67         if (dentry->d_name.len > NAME_MAX)
  68                 return ERR_PTR(-ENAMETOOLONG);
  69         if (!dentry->d_sb->s_d_op)
  70                 d_set_d_op(dentry, &simple_dentry_operations);
  71         d_add(dentry, NULL);
  72         return NULL;
  73 }
  74 EXPORT_SYMBOL(simple_lookup);
  75 
  76 int dcache_dir_open(struct inode *inode, struct file *file)
  77 {
  78         file->private_data = d_alloc_cursor(file->f_path.dentry);
  79 
  80         return file->private_data ? 0 : -ENOMEM;
  81 }
  82 EXPORT_SYMBOL(dcache_dir_open);
  83 
  84 int dcache_dir_close(struct inode *inode, struct file *file)
  85 {
  86         dput(file->private_data);
  87         return 0;
  88 }
  89 EXPORT_SYMBOL(dcache_dir_close);
  90 
  91 /* parent is locked at least shared */
  92 /*
  93  * Returns an element of siblings' list.
  94  * We are looking for <count>th positive after <p>; if
  95  * found, dentry is grabbed and returned to caller.
  96  * If no such element exists, NULL is returned.
  97  */
  98 static struct dentry *scan_positives(struct dentry *cursor,
  99                                         struct list_head *p,
 100                                         loff_t count,
 101                                         struct dentry *last)
 102 {
 103         struct dentry *dentry = cursor->d_parent, *found = NULL;
 104 
 105         spin_lock(&dentry->d_lock);
 106         while ((p = p->next) != &dentry->d_subdirs) {
 107                 struct dentry *d = list_entry(p, struct dentry, d_child);
 108                 // we must at least skip cursors, to avoid livelocks
 109                 if (d->d_flags & DCACHE_DENTRY_CURSOR)
 110                         continue;
 111                 if (simple_positive(d) && !--count) {
 112                         spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
 113                         if (simple_positive(d))
 114                                 found = dget_dlock(d);
 115                         spin_unlock(&d->d_lock);
 116                         if (likely(found))
 117                                 break;
 118                         count = 1;
 119                 }
 120                 if (need_resched()) {
 121                         list_move(&cursor->d_child, p);
 122                         p = &cursor->d_child;
 123                         spin_unlock(&dentry->d_lock);
 124                         cond_resched();
 125                         spin_lock(&dentry->d_lock);
 126                 }
 127         }
 128         spin_unlock(&dentry->d_lock);
 129         dput(last);
 130         return found;
 131 }
 132 
 133 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
 134 {
 135         struct dentry *dentry = file->f_path.dentry;
 136         switch (whence) {
 137                 case 1:
 138                         offset += file->f_pos;
 139                         /* fall through */
 140                 case 0:
 141                         if (offset >= 0)
 142                                 break;
 143                         /* fall through */
 144                 default:
 145                         return -EINVAL;
 146         }
 147         if (offset != file->f_pos) {
 148                 struct dentry *cursor = file->private_data;
 149                 struct dentry *to = NULL;
 150 
 151                 inode_lock_shared(dentry->d_inode);
 152 
 153                 if (offset > 2)
 154                         to = scan_positives(cursor, &dentry->d_subdirs,
 155                                             offset - 2, NULL);
 156                 spin_lock(&dentry->d_lock);
 157                 if (to)
 158                         list_move(&cursor->d_child, &to->d_child);
 159                 else
 160                         list_del_init(&cursor->d_child);
 161                 spin_unlock(&dentry->d_lock);
 162                 dput(to);
 163 
 164                 file->f_pos = offset;
 165 
 166                 inode_unlock_shared(dentry->d_inode);
 167         }
 168         return offset;
 169 }
 170 EXPORT_SYMBOL(dcache_dir_lseek);
 171 
 172 /* Relationship between i_mode and the DT_xxx types */
 173 static inline unsigned char dt_type(struct inode *inode)
 174 {
 175         return (inode->i_mode >> 12) & 15;
 176 }
 177 
 178 /*
 179  * Directory is locked and all positive dentries in it are safe, since
 180  * for ramfs-type trees they can't go away without unlink() or rmdir(),
 181  * both impossible due to the lock on directory.
 182  */
 183 
 184 int dcache_readdir(struct file *file, struct dir_context *ctx)
 185 {
 186         struct dentry *dentry = file->f_path.dentry;
 187         struct dentry *cursor = file->private_data;
 188         struct list_head *anchor = &dentry->d_subdirs;
 189         struct dentry *next = NULL;
 190         struct list_head *p;
 191 
 192         if (!dir_emit_dots(file, ctx))
 193                 return 0;
 194 
 195         if (ctx->pos == 2)
 196                 p = anchor;
 197         else if (!list_empty(&cursor->d_child))
 198                 p = &cursor->d_child;
 199         else
 200                 return 0;
 201 
 202         while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
 203                 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
 204                               d_inode(next)->i_ino, dt_type(d_inode(next))))
 205                         break;
 206                 ctx->pos++;
 207                 p = &next->d_child;
 208         }
 209         spin_lock(&dentry->d_lock);
 210         if (next)
 211                 list_move_tail(&cursor->d_child, &next->d_child);
 212         else
 213                 list_del_init(&cursor->d_child);
 214         spin_unlock(&dentry->d_lock);
 215         dput(next);
 216 
 217         return 0;
 218 }
 219 EXPORT_SYMBOL(dcache_readdir);
 220 
 221 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
 222 {
 223         return -EISDIR;
 224 }
 225 EXPORT_SYMBOL(generic_read_dir);
 226 
 227 const struct file_operations simple_dir_operations = {
 228         .open           = dcache_dir_open,
 229         .release        = dcache_dir_close,
 230         .llseek         = dcache_dir_lseek,
 231         .read           = generic_read_dir,
 232         .iterate_shared = dcache_readdir,
 233         .fsync          = noop_fsync,
 234 };
 235 EXPORT_SYMBOL(simple_dir_operations);
 236 
 237 const struct inode_operations simple_dir_inode_operations = {
 238         .lookup         = simple_lookup,
 239 };
 240 EXPORT_SYMBOL(simple_dir_inode_operations);
 241 
 242 static const struct super_operations simple_super_operations = {
 243         .statfs         = simple_statfs,
 244 };
 245 
 246 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
 247 {
 248         struct pseudo_fs_context *ctx = fc->fs_private;
 249         struct inode *root;
 250 
 251         s->s_maxbytes = MAX_LFS_FILESIZE;
 252         s->s_blocksize = PAGE_SIZE;
 253         s->s_blocksize_bits = PAGE_SHIFT;
 254         s->s_magic = ctx->magic;
 255         s->s_op = ctx->ops ?: &simple_super_operations;
 256         s->s_xattr = ctx->xattr;
 257         s->s_time_gran = 1;
 258         root = new_inode(s);
 259         if (!root)
 260                 return -ENOMEM;
 261 
 262         /*
 263          * since this is the first inode, make it number 1. New inodes created
 264          * after this must take care not to collide with it (by passing
 265          * max_reserved of 1 to iunique).
 266          */
 267         root->i_ino = 1;
 268         root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
 269         root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
 270         s->s_root = d_make_root(root);
 271         if (!s->s_root)
 272                 return -ENOMEM;
 273         s->s_d_op = ctx->dops;
 274         return 0;
 275 }
 276 
 277 static int pseudo_fs_get_tree(struct fs_context *fc)
 278 {
 279         return get_tree_nodev(fc, pseudo_fs_fill_super);
 280 }
 281 
 282 static void pseudo_fs_free(struct fs_context *fc)
 283 {
 284         kfree(fc->fs_private);
 285 }
 286 
 287 static const struct fs_context_operations pseudo_fs_context_ops = {
 288         .free           = pseudo_fs_free,
 289         .get_tree       = pseudo_fs_get_tree,
 290 };
 291 
 292 /*
 293  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
 294  * will never be mountable)
 295  */
 296 struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
 297                                         unsigned long magic)
 298 {
 299         struct pseudo_fs_context *ctx;
 300 
 301         ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
 302         if (likely(ctx)) {
 303                 ctx->magic = magic;
 304                 fc->fs_private = ctx;
 305                 fc->ops = &pseudo_fs_context_ops;
 306                 fc->sb_flags |= SB_NOUSER;
 307                 fc->global = true;
 308         }
 309         return ctx;
 310 }
 311 EXPORT_SYMBOL(init_pseudo);
 312 
 313 int simple_open(struct inode *inode, struct file *file)
 314 {
 315         if (inode->i_private)
 316                 file->private_data = inode->i_private;
 317         return 0;
 318 }
 319 EXPORT_SYMBOL(simple_open);
 320 
 321 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
 322 {
 323         struct inode *inode = d_inode(old_dentry);
 324 
 325         inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
 326         inc_nlink(inode);
 327         ihold(inode);
 328         dget(dentry);
 329         d_instantiate(dentry, inode);
 330         return 0;
 331 }
 332 EXPORT_SYMBOL(simple_link);
 333 
 334 int simple_empty(struct dentry *dentry)
 335 {
 336         struct dentry *child;
 337         int ret = 0;
 338 
 339         spin_lock(&dentry->d_lock);
 340         list_for_each_entry(child, &dentry->d_subdirs, d_child) {
 341                 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
 342                 if (simple_positive(child)) {
 343                         spin_unlock(&child->d_lock);
 344                         goto out;
 345                 }
 346                 spin_unlock(&child->d_lock);
 347         }
 348         ret = 1;
 349 out:
 350         spin_unlock(&dentry->d_lock);
 351         return ret;
 352 }
 353 EXPORT_SYMBOL(simple_empty);
 354 
 355 int simple_unlink(struct inode *dir, struct dentry *dentry)
 356 {
 357         struct inode *inode = d_inode(dentry);
 358 
 359         inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
 360         drop_nlink(inode);
 361         dput(dentry);
 362         return 0;
 363 }
 364 EXPORT_SYMBOL(simple_unlink);
 365 
 366 int simple_rmdir(struct inode *dir, struct dentry *dentry)
 367 {
 368         if (!simple_empty(dentry))
 369                 return -ENOTEMPTY;
 370 
 371         drop_nlink(d_inode(dentry));
 372         simple_unlink(dir, dentry);
 373         drop_nlink(dir);
 374         return 0;
 375 }
 376 EXPORT_SYMBOL(simple_rmdir);
 377 
 378 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
 379                   struct inode *new_dir, struct dentry *new_dentry,
 380                   unsigned int flags)
 381 {
 382         struct inode *inode = d_inode(old_dentry);
 383         int they_are_dirs = d_is_dir(old_dentry);
 384 
 385         if (flags & ~RENAME_NOREPLACE)
 386                 return -EINVAL;
 387 
 388         if (!simple_empty(new_dentry))
 389                 return -ENOTEMPTY;
 390 
 391         if (d_really_is_positive(new_dentry)) {
 392                 simple_unlink(new_dir, new_dentry);
 393                 if (they_are_dirs) {
 394                         drop_nlink(d_inode(new_dentry));
 395                         drop_nlink(old_dir);
 396                 }
 397         } else if (they_are_dirs) {
 398                 drop_nlink(old_dir);
 399                 inc_nlink(new_dir);
 400         }
 401 
 402         old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
 403                 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
 404 
 405         return 0;
 406 }
 407 EXPORT_SYMBOL(simple_rename);
 408 
 409 /**
 410  * simple_setattr - setattr for simple filesystem
 411  * @dentry: dentry
 412  * @iattr: iattr structure
 413  *
 414  * Returns 0 on success, -error on failure.
 415  *
 416  * simple_setattr is a simple ->setattr implementation without a proper
 417  * implementation of size changes.
 418  *
 419  * It can either be used for in-memory filesystems or special files
 420  * on simple regular filesystems.  Anything that needs to change on-disk
 421  * or wire state on size changes needs its own setattr method.
 422  */
 423 int simple_setattr(struct dentry *dentry, struct iattr *iattr)
 424 {
 425         struct inode *inode = d_inode(dentry);
 426         int error;
 427 
 428         error = setattr_prepare(dentry, iattr);
 429         if (error)
 430                 return error;
 431 
 432         if (iattr->ia_valid & ATTR_SIZE)
 433                 truncate_setsize(inode, iattr->ia_size);
 434         setattr_copy(inode, iattr);
 435         mark_inode_dirty(inode);
 436         return 0;
 437 }
 438 EXPORT_SYMBOL(simple_setattr);
 439 
 440 int simple_readpage(struct file *file, struct page *page)
 441 {
 442         clear_highpage(page);
 443         flush_dcache_page(page);
 444         SetPageUptodate(page);
 445         unlock_page(page);
 446         return 0;
 447 }
 448 EXPORT_SYMBOL(simple_readpage);
 449 
 450 int simple_write_begin(struct file *file, struct address_space *mapping,
 451                         loff_t pos, unsigned len, unsigned flags,
 452                         struct page **pagep, void **fsdata)
 453 {
 454         struct page *page;
 455         pgoff_t index;
 456 
 457         index = pos >> PAGE_SHIFT;
 458 
 459         page = grab_cache_page_write_begin(mapping, index, flags);
 460         if (!page)
 461                 return -ENOMEM;
 462 
 463         *pagep = page;
 464 
 465         if (!PageUptodate(page) && (len != PAGE_SIZE)) {
 466                 unsigned from = pos & (PAGE_SIZE - 1);
 467 
 468                 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
 469         }
 470         return 0;
 471 }
 472 EXPORT_SYMBOL(simple_write_begin);
 473 
 474 /**
 475  * simple_write_end - .write_end helper for non-block-device FSes
 476  * @file: See .write_end of address_space_operations
 477  * @mapping:            "
 478  * @pos:                "
 479  * @len:                "
 480  * @copied:             "
 481  * @page:               "
 482  * @fsdata:             "
 483  *
 484  * simple_write_end does the minimum needed for updating a page after writing is
 485  * done. It has the same API signature as the .write_end of
 486  * address_space_operations vector. So it can just be set onto .write_end for
 487  * FSes that don't need any other processing. i_mutex is assumed to be held.
 488  * Block based filesystems should use generic_write_end().
 489  * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
 490  * is not called, so a filesystem that actually does store data in .write_inode
 491  * should extend on what's done here with a call to mark_inode_dirty() in the
 492  * case that i_size has changed.
 493  *
 494  * Use *ONLY* with simple_readpage()
 495  */
 496 int simple_write_end(struct file *file, struct address_space *mapping,
 497                         loff_t pos, unsigned len, unsigned copied,
 498                         struct page *page, void *fsdata)
 499 {
 500         struct inode *inode = page->mapping->host;
 501         loff_t last_pos = pos + copied;
 502 
 503         /* zero the stale part of the page if we did a short copy */
 504         if (!PageUptodate(page)) {
 505                 if (copied < len) {
 506                         unsigned from = pos & (PAGE_SIZE - 1);
 507 
 508                         zero_user(page, from + copied, len - copied);
 509                 }
 510                 SetPageUptodate(page);
 511         }
 512         /*
 513          * No need to use i_size_read() here, the i_size
 514          * cannot change under us because we hold the i_mutex.
 515          */
 516         if (last_pos > inode->i_size)
 517                 i_size_write(inode, last_pos);
 518 
 519         set_page_dirty(page);
 520         unlock_page(page);
 521         put_page(page);
 522 
 523         return copied;
 524 }
 525 EXPORT_SYMBOL(simple_write_end);
 526 
 527 /*
 528  * the inodes created here are not hashed. If you use iunique to generate
 529  * unique inode values later for this filesystem, then you must take care
 530  * to pass it an appropriate max_reserved value to avoid collisions.
 531  */
 532 int simple_fill_super(struct super_block *s, unsigned long magic,
 533                       const struct tree_descr *files)
 534 {
 535         struct inode *inode;
 536         struct dentry *root;
 537         struct dentry *dentry;
 538         int i;
 539 
 540         s->s_blocksize = PAGE_SIZE;
 541         s->s_blocksize_bits = PAGE_SHIFT;
 542         s->s_magic = magic;
 543         s->s_op = &simple_super_operations;
 544         s->s_time_gran = 1;
 545 
 546         inode = new_inode(s);
 547         if (!inode)
 548                 return -ENOMEM;
 549         /*
 550          * because the root inode is 1, the files array must not contain an
 551          * entry at index 1
 552          */
 553         inode->i_ino = 1;
 554         inode->i_mode = S_IFDIR | 0755;
 555         inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
 556         inode->i_op = &simple_dir_inode_operations;
 557         inode->i_fop = &simple_dir_operations;
 558         set_nlink(inode, 2);
 559         root = d_make_root(inode);
 560         if (!root)
 561                 return -ENOMEM;
 562         for (i = 0; !files->name || files->name[0]; i++, files++) {
 563                 if (!files->name)
 564                         continue;
 565 
 566                 /* warn if it tries to conflict with the root inode */
 567                 if (unlikely(i == 1))
 568                         printk(KERN_WARNING "%s: %s passed in a files array"
 569                                 "with an index of 1!\n", __func__,
 570                                 s->s_type->name);
 571 
 572                 dentry = d_alloc_name(root, files->name);
 573                 if (!dentry)
 574                         goto out;
 575                 inode = new_inode(s);
 576                 if (!inode) {
 577                         dput(dentry);
 578                         goto out;
 579                 }
 580                 inode->i_mode = S_IFREG | files->mode;
 581                 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
 582                 inode->i_fop = files->ops;
 583                 inode->i_ino = i;
 584                 d_add(dentry, inode);
 585         }
 586         s->s_root = root;
 587         return 0;
 588 out:
 589         d_genocide(root);
 590         shrink_dcache_parent(root);
 591         dput(root);
 592         return -ENOMEM;
 593 }
 594 EXPORT_SYMBOL(simple_fill_super);
 595 
 596 static DEFINE_SPINLOCK(pin_fs_lock);
 597 
 598 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
 599 {
 600         struct vfsmount *mnt = NULL;
 601         spin_lock(&pin_fs_lock);
 602         if (unlikely(!*mount)) {
 603                 spin_unlock(&pin_fs_lock);
 604                 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
 605                 if (IS_ERR(mnt))
 606                         return PTR_ERR(mnt);
 607                 spin_lock(&pin_fs_lock);
 608                 if (!*mount)
 609                         *mount = mnt;
 610         }
 611         mntget(*mount);
 612         ++*count;
 613         spin_unlock(&pin_fs_lock);
 614         mntput(mnt);
 615         return 0;
 616 }
 617 EXPORT_SYMBOL(simple_pin_fs);
 618 
 619 void simple_release_fs(struct vfsmount **mount, int *count)
 620 {
 621         struct vfsmount *mnt;
 622         spin_lock(&pin_fs_lock);
 623         mnt = *mount;
 624         if (!--*count)
 625                 *mount = NULL;
 626         spin_unlock(&pin_fs_lock);
 627         mntput(mnt);
 628 }
 629 EXPORT_SYMBOL(simple_release_fs);
 630 
 631 /**
 632  * simple_read_from_buffer - copy data from the buffer to user space
 633  * @to: the user space buffer to read to
 634  * @count: the maximum number of bytes to read
 635  * @ppos: the current position in the buffer
 636  * @from: the buffer to read from
 637  * @available: the size of the buffer
 638  *
 639  * The simple_read_from_buffer() function reads up to @count bytes from the
 640  * buffer @from at offset @ppos into the user space address starting at @to.
 641  *
 642  * On success, the number of bytes read is returned and the offset @ppos is
 643  * advanced by this number, or negative value is returned on error.
 644  **/
 645 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
 646                                 const void *from, size_t available)
 647 {
 648         loff_t pos = *ppos;
 649         size_t ret;
 650 
 651         if (pos < 0)
 652                 return -EINVAL;
 653         if (pos >= available || !count)
 654                 return 0;
 655         if (count > available - pos)
 656                 count = available - pos;
 657         ret = copy_to_user(to, from + pos, count);
 658         if (ret == count)
 659                 return -EFAULT;
 660         count -= ret;
 661         *ppos = pos + count;
 662         return count;
 663 }
 664 EXPORT_SYMBOL(simple_read_from_buffer);
 665 
 666 /**
 667  * simple_write_to_buffer - copy data from user space to the buffer
 668  * @to: the buffer to write to
 669  * @available: the size of the buffer
 670  * @ppos: the current position in the buffer
 671  * @from: the user space buffer to read from
 672  * @count: the maximum number of bytes to read
 673  *
 674  * The simple_write_to_buffer() function reads up to @count bytes from the user
 675  * space address starting at @from into the buffer @to at offset @ppos.
 676  *
 677  * On success, the number of bytes written is returned and the offset @ppos is
 678  * advanced by this number, or negative value is returned on error.
 679  **/
 680 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
 681                 const void __user *from, size_t count)
 682 {
 683         loff_t pos = *ppos;
 684         size_t res;
 685 
 686         if (pos < 0)
 687                 return -EINVAL;
 688         if (pos >= available || !count)
 689                 return 0;
 690         if (count > available - pos)
 691                 count = available - pos;
 692         res = copy_from_user(to + pos, from, count);
 693         if (res == count)
 694                 return -EFAULT;
 695         count -= res;
 696         *ppos = pos + count;
 697         return count;
 698 }
 699 EXPORT_SYMBOL(simple_write_to_buffer);
 700 
 701 /**
 702  * memory_read_from_buffer - copy data from the buffer
 703  * @to: the kernel space buffer to read to
 704  * @count: the maximum number of bytes to read
 705  * @ppos: the current position in the buffer
 706  * @from: the buffer to read from
 707  * @available: the size of the buffer
 708  *
 709  * The memory_read_from_buffer() function reads up to @count bytes from the
 710  * buffer @from at offset @ppos into the kernel space address starting at @to.
 711  *
 712  * On success, the number of bytes read is returned and the offset @ppos is
 713  * advanced by this number, or negative value is returned on error.
 714  **/
 715 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
 716                                 const void *from, size_t available)
 717 {
 718         loff_t pos = *ppos;
 719 
 720         if (pos < 0)
 721                 return -EINVAL;
 722         if (pos >= available)
 723                 return 0;
 724         if (count > available - pos)
 725                 count = available - pos;
 726         memcpy(to, from + pos, count);
 727         *ppos = pos + count;
 728 
 729         return count;
 730 }
 731 EXPORT_SYMBOL(memory_read_from_buffer);
 732 
 733 /*
 734  * Transaction based IO.
 735  * The file expects a single write which triggers the transaction, and then
 736  * possibly a read which collects the result - which is stored in a
 737  * file-local buffer.
 738  */
 739 
 740 void simple_transaction_set(struct file *file, size_t n)
 741 {
 742         struct simple_transaction_argresp *ar = file->private_data;
 743 
 744         BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
 745 
 746         /*
 747          * The barrier ensures that ar->size will really remain zero until
 748          * ar->data is ready for reading.
 749          */
 750         smp_mb();
 751         ar->size = n;
 752 }
 753 EXPORT_SYMBOL(simple_transaction_set);
 754 
 755 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
 756 {
 757         struct simple_transaction_argresp *ar;
 758         static DEFINE_SPINLOCK(simple_transaction_lock);
 759 
 760         if (size > SIMPLE_TRANSACTION_LIMIT - 1)
 761                 return ERR_PTR(-EFBIG);
 762 
 763         ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
 764         if (!ar)
 765                 return ERR_PTR(-ENOMEM);
 766 
 767         spin_lock(&simple_transaction_lock);
 768 
 769         /* only one write allowed per open */
 770         if (file->private_data) {
 771                 spin_unlock(&simple_transaction_lock);
 772                 free_page((unsigned long)ar);
 773                 return ERR_PTR(-EBUSY);
 774         }
 775 
 776         file->private_data = ar;
 777 
 778         spin_unlock(&simple_transaction_lock);
 779 
 780         if (copy_from_user(ar->data, buf, size))
 781                 return ERR_PTR(-EFAULT);
 782 
 783         return ar->data;
 784 }
 785 EXPORT_SYMBOL(simple_transaction_get);
 786 
 787 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 788 {
 789         struct simple_transaction_argresp *ar = file->private_data;
 790 
 791         if (!ar)
 792                 return 0;
 793         return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
 794 }
 795 EXPORT_SYMBOL(simple_transaction_read);
 796 
 797 int simple_transaction_release(struct inode *inode, struct file *file)
 798 {
 799         free_page((unsigned long)file->private_data);
 800         return 0;
 801 }
 802 EXPORT_SYMBOL(simple_transaction_release);
 803 
 804 /* Simple attribute files */
 805 
 806 struct simple_attr {
 807         int (*get)(void *, u64 *);
 808         int (*set)(void *, u64);
 809         char get_buf[24];       /* enough to store a u64 and "\n\0" */
 810         char set_buf[24];
 811         void *data;
 812         const char *fmt;        /* format for read operation */
 813         struct mutex mutex;     /* protects access to these buffers */
 814 };
 815 
 816 /* simple_attr_open is called by an actual attribute open file operation
 817  * to set the attribute specific access operations. */
 818 int simple_attr_open(struct inode *inode, struct file *file,
 819                      int (*get)(void *, u64 *), int (*set)(void *, u64),
 820                      const char *fmt)
 821 {
 822         struct simple_attr *attr;
 823 
 824         attr = kzalloc(sizeof(*attr), GFP_KERNEL);
 825         if (!attr)
 826                 return -ENOMEM;
 827 
 828         attr->get = get;
 829         attr->set = set;
 830         attr->data = inode->i_private;
 831         attr->fmt = fmt;
 832         mutex_init(&attr->mutex);
 833 
 834         file->private_data = attr;
 835 
 836         return nonseekable_open(inode, file);
 837 }
 838 EXPORT_SYMBOL_GPL(simple_attr_open);
 839 
 840 int simple_attr_release(struct inode *inode, struct file *file)
 841 {
 842         kfree(file->private_data);
 843         return 0;
 844 }
 845 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only?  This?  Really? */
 846 
 847 /* read from the buffer that is filled with the get function */
 848 ssize_t simple_attr_read(struct file *file, char __user *buf,
 849                          size_t len, loff_t *ppos)
 850 {
 851         struct simple_attr *attr;
 852         size_t size;
 853         ssize_t ret;
 854 
 855         attr = file->private_data;
 856 
 857         if (!attr->get)
 858                 return -EACCES;
 859 
 860         ret = mutex_lock_interruptible(&attr->mutex);
 861         if (ret)
 862                 return ret;
 863 
 864         if (*ppos && attr->get_buf[0]) {
 865                 /* continued read */
 866                 size = strlen(attr->get_buf);
 867         } else {
 868                 /* first read */
 869                 u64 val;
 870                 ret = attr->get(attr->data, &val);
 871                 if (ret)
 872                         goto out;
 873 
 874                 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
 875                                  attr->fmt, (unsigned long long)val);
 876         }
 877 
 878         ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
 879 out:
 880         mutex_unlock(&attr->mutex);
 881         return ret;
 882 }
 883 EXPORT_SYMBOL_GPL(simple_attr_read);
 884 
 885 /* interpret the buffer as a number to call the set function with */
 886 ssize_t simple_attr_write(struct file *file, const char __user *buf,
 887                           size_t len, loff_t *ppos)
 888 {
 889         struct simple_attr *attr;
 890         u64 val;
 891         size_t size;
 892         ssize_t ret;
 893 
 894         attr = file->private_data;
 895         if (!attr->set)
 896                 return -EACCES;
 897 
 898         ret = mutex_lock_interruptible(&attr->mutex);
 899         if (ret)
 900                 return ret;
 901 
 902         ret = -EFAULT;
 903         size = min(sizeof(attr->set_buf) - 1, len);
 904         if (copy_from_user(attr->set_buf, buf, size))
 905                 goto out;
 906 
 907         attr->set_buf[size] = '\0';
 908         val = simple_strtoll(attr->set_buf, NULL, 0);
 909         ret = attr->set(attr->data, val);
 910         if (ret == 0)
 911                 ret = len; /* on success, claim we got the whole input */
 912 out:
 913         mutex_unlock(&attr->mutex);
 914         return ret;
 915 }
 916 EXPORT_SYMBOL_GPL(simple_attr_write);
 917 
 918 /**
 919  * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
 920  * @sb:         filesystem to do the file handle conversion on
 921  * @fid:        file handle to convert
 922  * @fh_len:     length of the file handle in bytes
 923  * @fh_type:    type of file handle
 924  * @get_inode:  filesystem callback to retrieve inode
 925  *
 926  * This function decodes @fid as long as it has one of the well-known
 927  * Linux filehandle types and calls @get_inode on it to retrieve the
 928  * inode for the object specified in the file handle.
 929  */
 930 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
 931                 int fh_len, int fh_type, struct inode *(*get_inode)
 932                         (struct super_block *sb, u64 ino, u32 gen))
 933 {
 934         struct inode *inode = NULL;
 935 
 936         if (fh_len < 2)
 937                 return NULL;
 938 
 939         switch (fh_type) {
 940         case FILEID_INO32_GEN:
 941         case FILEID_INO32_GEN_PARENT:
 942                 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
 943                 break;
 944         }
 945 
 946         return d_obtain_alias(inode);
 947 }
 948 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
 949 
 950 /**
 951  * generic_fh_to_parent - generic helper for the fh_to_parent export operation
 952  * @sb:         filesystem to do the file handle conversion on
 953  * @fid:        file handle to convert
 954  * @fh_len:     length of the file handle in bytes
 955  * @fh_type:    type of file handle
 956  * @get_inode:  filesystem callback to retrieve inode
 957  *
 958  * This function decodes @fid as long as it has one of the well-known
 959  * Linux filehandle types and calls @get_inode on it to retrieve the
 960  * inode for the _parent_ object specified in the file handle if it
 961  * is specified in the file handle, or NULL otherwise.
 962  */
 963 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
 964                 int fh_len, int fh_type, struct inode *(*get_inode)
 965                         (struct super_block *sb, u64 ino, u32 gen))
 966 {
 967         struct inode *inode = NULL;
 968 
 969         if (fh_len <= 2)
 970                 return NULL;
 971 
 972         switch (fh_type) {
 973         case FILEID_INO32_GEN_PARENT:
 974                 inode = get_inode(sb, fid->i32.parent_ino,
 975                                   (fh_len > 3 ? fid->i32.parent_gen : 0));
 976                 break;
 977         }
 978 
 979         return d_obtain_alias(inode);
 980 }
 981 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
 982 
 983 /**
 984  * __generic_file_fsync - generic fsync implementation for simple filesystems
 985  *
 986  * @file:       file to synchronize
 987  * @start:      start offset in bytes
 988  * @end:        end offset in bytes (inclusive)
 989  * @datasync:   only synchronize essential metadata if true
 990  *
 991  * This is a generic implementation of the fsync method for simple
 992  * filesystems which track all non-inode metadata in the buffers list
 993  * hanging off the address_space structure.
 994  */
 995 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
 996                                  int datasync)
 997 {
 998         struct inode *inode = file->f_mapping->host;
 999         int err;
1000         int ret;
1001 
1002         err = file_write_and_wait_range(file, start, end);
1003         if (err)
1004                 return err;
1005 
1006         inode_lock(inode);
1007         ret = sync_mapping_buffers(inode->i_mapping);
1008         if (!(inode->i_state & I_DIRTY_ALL))
1009                 goto out;
1010         if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1011                 goto out;
1012 
1013         err = sync_inode_metadata(inode, 1);
1014         if (ret == 0)
1015                 ret = err;
1016 
1017 out:
1018         inode_unlock(inode);
1019         /* check and advance again to catch errors after syncing out buffers */
1020         err = file_check_and_advance_wb_err(file);
1021         if (ret == 0)
1022                 ret = err;
1023         return ret;
1024 }
1025 EXPORT_SYMBOL(__generic_file_fsync);
1026 
1027 /**
1028  * generic_file_fsync - generic fsync implementation for simple filesystems
1029  *                      with flush
1030  * @file:       file to synchronize
1031  * @start:      start offset in bytes
1032  * @end:        end offset in bytes (inclusive)
1033  * @datasync:   only synchronize essential metadata if true
1034  *
1035  */
1036 
1037 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1038                        int datasync)
1039 {
1040         struct inode *inode = file->f_mapping->host;
1041         int err;
1042 
1043         err = __generic_file_fsync(file, start, end, datasync);
1044         if (err)
1045                 return err;
1046         return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
1047 }
1048 EXPORT_SYMBOL(generic_file_fsync);
1049 
1050 /**
1051  * generic_check_addressable - Check addressability of file system
1052  * @blocksize_bits:     log of file system block size
1053  * @num_blocks:         number of blocks in file system
1054  *
1055  * Determine whether a file system with @num_blocks blocks (and a
1056  * block size of 2**@blocksize_bits) is addressable by the sector_t
1057  * and page cache of the system.  Return 0 if so and -EFBIG otherwise.
1058  */
1059 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1060 {
1061         u64 last_fs_block = num_blocks - 1;
1062         u64 last_fs_page =
1063                 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1064 
1065         if (unlikely(num_blocks == 0))
1066                 return 0;
1067 
1068         if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1069                 return -EINVAL;
1070 
1071         if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1072             (last_fs_page > (pgoff_t)(~0ULL))) {
1073                 return -EFBIG;
1074         }
1075         return 0;
1076 }
1077 EXPORT_SYMBOL(generic_check_addressable);
1078 
1079 /*
1080  * No-op implementation of ->fsync for in-memory filesystems.
1081  */
1082 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1083 {
1084         return 0;
1085 }
1086 EXPORT_SYMBOL(noop_fsync);
1087 
1088 int noop_set_page_dirty(struct page *page)
1089 {
1090         /*
1091          * Unlike __set_page_dirty_no_writeback that handles dirty page
1092          * tracking in the page object, dax does all dirty tracking in
1093          * the inode address_space in response to mkwrite faults. In the
1094          * dax case we only need to worry about potentially dirty CPU
1095          * caches, not dirty page cache pages to write back.
1096          *
1097          * This callback is defined to prevent fallback to
1098          * __set_page_dirty_buffers() in set_page_dirty().
1099          */
1100         return 0;
1101 }
1102 EXPORT_SYMBOL_GPL(noop_set_page_dirty);
1103 
1104 void noop_invalidatepage(struct page *page, unsigned int offset,
1105                 unsigned int length)
1106 {
1107         /*
1108          * There is no page cache to invalidate in the dax case, however
1109          * we need this callback defined to prevent falling back to
1110          * block_invalidatepage() in do_invalidatepage().
1111          */
1112 }
1113 EXPORT_SYMBOL_GPL(noop_invalidatepage);
1114 
1115 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1116 {
1117         /*
1118          * iomap based filesystems support direct I/O without need for
1119          * this callback. However, it still needs to be set in
1120          * inode->a_ops so that open/fcntl know that direct I/O is
1121          * generally supported.
1122          */
1123         return -EINVAL;
1124 }
1125 EXPORT_SYMBOL_GPL(noop_direct_IO);
1126 
1127 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1128 void kfree_link(void *p)
1129 {
1130         kfree(p);
1131 }
1132 EXPORT_SYMBOL(kfree_link);
1133 
1134 /*
1135  * nop .set_page_dirty method so that people can use .page_mkwrite on
1136  * anon inodes.
1137  */
1138 static int anon_set_page_dirty(struct page *page)
1139 {
1140         return 0;
1141 };
1142 
1143 /*
1144  * A single inode exists for all anon_inode files. Contrary to pipes,
1145  * anon_inode inodes have no associated per-instance data, so we need
1146  * only allocate one of them.
1147  */
1148 struct inode *alloc_anon_inode(struct super_block *s)
1149 {
1150         static const struct address_space_operations anon_aops = {
1151                 .set_page_dirty = anon_set_page_dirty,
1152         };
1153         struct inode *inode = new_inode_pseudo(s);
1154 
1155         if (!inode)
1156                 return ERR_PTR(-ENOMEM);
1157 
1158         inode->i_ino = get_next_ino();
1159         inode->i_mapping->a_ops = &anon_aops;
1160 
1161         /*
1162          * Mark the inode dirty from the very beginning,
1163          * that way it will never be moved to the dirty
1164          * list because mark_inode_dirty() will think
1165          * that it already _is_ on the dirty list.
1166          */
1167         inode->i_state = I_DIRTY;
1168         inode->i_mode = S_IRUSR | S_IWUSR;
1169         inode->i_uid = current_fsuid();
1170         inode->i_gid = current_fsgid();
1171         inode->i_flags |= S_PRIVATE;
1172         inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1173         return inode;
1174 }
1175 EXPORT_SYMBOL(alloc_anon_inode);
1176 
1177 /**
1178  * simple_nosetlease - generic helper for prohibiting leases
1179  * @filp: file pointer
1180  * @arg: type of lease to obtain
1181  * @flp: new lease supplied for insertion
1182  * @priv: private data for lm_setup operation
1183  *
1184  * Generic helper for filesystems that do not wish to allow leases to be set.
1185  * All arguments are ignored and it just returns -EINVAL.
1186  */
1187 int
1188 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1189                   void **priv)
1190 {
1191         return -EINVAL;
1192 }
1193 EXPORT_SYMBOL(simple_nosetlease);
1194 
1195 /**
1196  * simple_get_link - generic helper to get the target of "fast" symlinks
1197  * @dentry: not used here
1198  * @inode: the symlink inode
1199  * @done: not used here
1200  *
1201  * Generic helper for filesystems to use for symlink inodes where a pointer to
1202  * the symlink target is stored in ->i_link.  NOTE: this isn't normally called,
1203  * since as an optimization the path lookup code uses any non-NULL ->i_link
1204  * directly, without calling ->get_link().  But ->get_link() still must be set,
1205  * to mark the inode_operations as being for a symlink.
1206  *
1207  * Return: the symlink target
1208  */
1209 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1210                             struct delayed_call *done)
1211 {
1212         return inode->i_link;
1213 }
1214 EXPORT_SYMBOL(simple_get_link);
1215 
1216 const struct inode_operations simple_symlink_inode_operations = {
1217         .get_link = simple_get_link,
1218 };
1219 EXPORT_SYMBOL(simple_symlink_inode_operations);
1220 
1221 /*
1222  * Operations for a permanently empty directory.
1223  */
1224 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1225 {
1226         return ERR_PTR(-ENOENT);
1227 }
1228 
1229 static int empty_dir_getattr(const struct path *path, struct kstat *stat,
1230                              u32 request_mask, unsigned int query_flags)
1231 {
1232         struct inode *inode = d_inode(path->dentry);
1233         generic_fillattr(inode, stat);
1234         return 0;
1235 }
1236 
1237 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr)
1238 {
1239         return -EPERM;
1240 }
1241 
1242 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1243 {
1244         return -EOPNOTSUPP;
1245 }
1246 
1247 static const struct inode_operations empty_dir_inode_operations = {
1248         .lookup         = empty_dir_lookup,
1249         .permission     = generic_permission,
1250         .setattr        = empty_dir_setattr,
1251         .getattr        = empty_dir_getattr,
1252         .listxattr      = empty_dir_listxattr,
1253 };
1254 
1255 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1256 {
1257         /* An empty directory has two entries . and .. at offsets 0 and 1 */
1258         return generic_file_llseek_size(file, offset, whence, 2, 2);
1259 }
1260 
1261 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1262 {
1263         dir_emit_dots(file, ctx);
1264         return 0;
1265 }
1266 
1267 static const struct file_operations empty_dir_operations = {
1268         .llseek         = empty_dir_llseek,
1269         .read           = generic_read_dir,
1270         .iterate_shared = empty_dir_readdir,
1271         .fsync          = noop_fsync,
1272 };
1273 
1274 
1275 void make_empty_dir_inode(struct inode *inode)
1276 {
1277         set_nlink(inode, 2);
1278         inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1279         inode->i_uid = GLOBAL_ROOT_UID;
1280         inode->i_gid = GLOBAL_ROOT_GID;
1281         inode->i_rdev = 0;
1282         inode->i_size = 0;
1283         inode->i_blkbits = PAGE_SHIFT;
1284         inode->i_blocks = 0;
1285 
1286         inode->i_op = &empty_dir_inode_operations;
1287         inode->i_opflags &= ~IOP_XATTR;
1288         inode->i_fop = &empty_dir_operations;
1289 }
1290 
1291 bool is_empty_dir_inode(struct inode *inode)
1292 {
1293         return (inode->i_fop == &empty_dir_operations) &&
1294                 (inode->i_op == &empty_dir_inode_operations);
1295 }
/* [<][>][^][v][top][bottom][index][help] */
root/fs/libfs.c

DEFINITIONS
