fs/libfs.c

   1 /*
   2  *      fs/libfs.c
   3  *      Library for filesystems writers.
   4  */
   5
   6 #include <linux/module.h>
   7 #include <linux/pagemap.h>
   8 #include <linux/mount.h>
   9 #include <linux/vfs.h>
  10 #include <linux/mutex.h>
  11
  12 #include <asm/uaccess.h>
  13
  14 int simple_getattr(struct vfsmount *mnt, struct dentry *dentry,
  15                    struct kstat *stat)
  16 {
  17         struct inode *inode = dentry->d_inode;
  18         generic_fillattr(inode, stat);
  19         stat->blocks = inode->i_mapping->nrpages << (PAGE_CACHE_SHIFT - 9);
  20         return 0;
  21 }
  22
  23 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
  24 {
  25         buf->f_type = dentry->d_sb->s_magic;
  26         buf->f_bsize = PAGE_CACHE_SIZE;
  27         buf->f_namelen = NAME_MAX;
  28         return 0;
  29 }
  30
  31 /*
  32  * Retaining negative dentries for an in-memory filesystem just wastes
  33  * memory and lookup time: arrange for them to be deleted immediately.
  34  */
  35 static int simple_delete_dentry(struct dentry *dentry)
  36 {
  37         return 1;
  38 }
  39
  40 /*
  41  * Lookup the data. This is trivial - if the dentry didn't already
  42  * exist, we know it is negative.  Set d_op to delete negative dentries.
  43  */
  44 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
  45 {
  46         static struct dentry_operations simple_dentry_operations = {
  47                 .d_delete = simple_delete_dentry,
  48         };
  49
  50         if (dentry->d_name.len > NAME_MAX)
  51                 return ERR_PTR(-ENAMETOOLONG);
  52         dentry->d_op = &simple_dentry_operations;
  53         d_add(dentry, NULL);
  54         return NULL;
  55 }
  56
  57 int simple_sync_file(struct file * file, struct dentry *dentry, int datasync)
  58 {
  59         return 0;
  60 }
  61
  62 int dcache_dir_open(struct inode *inode, struct file *file)
  63 {
  64         static struct qstr cursor_name = {.len = 1, .name = "."};
  65
  66         file->private_data = d_alloc(file->f_dentry, &cursor_name);
  67
  68         return file->private_data ? 0 : -ENOMEM;
  69 }
  70
  71 int dcache_dir_close(struct inode *inode, struct file *file)
  72 {
  73         dput(file->private_data);
  74         return 0;
  75 }
  76
  77 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int origin)
  78 {
  79         mutex_lock(&file->f_dentry->d_inode->i_mutex);
  80         switch (origin) {
  81                 case 1:
  82                         offset += file->f_pos;
  83                 case 0:
  84                         if (offset >= 0)
  85                                 break;
  86                 default:
  87                         mutex_unlock(&file->f_dentry->d_inode->i_mutex);
  88                         return -EINVAL;
  89         }
  90         if (offset != file->f_pos) {
  91                 file->f_pos = offset;
  92                 if (file->f_pos >= 2) {
  93                         struct list_head *p;
  94                         struct dentry *cursor = file->private_data;
  95                         loff_t n = file->f_pos - 2;
  96
  97                         spin_lock(&dcache_lock);
  98                         list_del(&cursor->d_u.d_child);
  99                         p = file->f_dentry->d_subdirs.next;
 100                         while (n && p != &file->f_dentry->d_subdirs) {
 101                                 struct dentry *next;
 102                                 next = list_entry(p, struct dentry, d_u.d_child);
 103                                 if (!d_unhashed(next) && next->d_inode)
 104                                         n--;
 105                                 p = p->next;
 106                         }
 107                         list_add_tail(&cursor->d_u.d_child, p);
 108                         spin_unlock(&dcache_lock);
 109                 }
 110         }
 111         mutex_unlock(&file->f_dentry->d_inode->i_mutex);
 112         return offset;
 113 }
 114
 115 /* Relationship between i_mode and the DT_xxx types */
 116 static inline unsigned char dt_type(struct inode *inode)
 117 {
 118         return (inode->i_mode >> 12) & 15;
 119 }
 120
 121 /*
 122  * Directory is locked and all positive dentries in it are safe, since
 123  * for ramfs-type trees they can't go away without unlink() or rmdir(),
 124  * both impossible due to the lock on directory.
 125  */
 126
 127 int dcache_readdir(struct file * filp, void * dirent, filldir_t filldir)
 128 {
 129         struct dentry *dentry = filp->f_dentry;
 130         struct dentry *cursor = filp->private_data;
 131         struct list_head *p, *q = &cursor->d_u.d_child;
 132         ino_t ino;
 133         int i = filp->f_pos;
 134
 135         switch (i) {
 136                 case 0:
 137                         ino = dentry->d_inode->i_ino;
 138                         if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
 139                                 break;
 140                         filp->f_pos++;
 141                         i++;
 142                         /* fallthrough */
 143                 case 1:
 144                         ino = parent_ino(dentry);
 145                         if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
 146                                 break;
 147                         filp->f_pos++;
 148                         i++;
 149                         /* fallthrough */
 150                 default:
 151                         spin_lock(&dcache_lock);
 152                         if (filp->f_pos == 2)
 153                                 list_move(q, &dentry->d_subdirs);
 154
 155                         for (p=q->next; p != &dentry->d_subdirs; p=p->next) {
 156                                 struct dentry *next;
 157                                 next = list_entry(p, struct dentry, d_u.d_child);
 158                                 if (d_unhashed(next) || !next->d_inode)
 159                                         continue;
 160
 161                                 spin_unlock(&dcache_lock);
 162                                 if (filldir(dirent, next->d_name.name, next->d_name.len, filp->f_pos, next->d_inode->i_ino, dt_type(next->d_inode)) < 0)
 163                                         return 0;
 164                                 spin_lock(&dcache_lock);
 165                                 /* next is still alive */
 166                                 list_move(q, p);
 167                                 p = q;
 168                                 filp->f_pos++;
 169                         }
 170                         spin_unlock(&dcache_lock);
 171         }
 172         return 0;
 173 }
 174
 175 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
 176 {
 177         return -EISDIR;
 178 }
 179
 180 const struct file_operations simple_dir_operations = {
 181         .open           = dcache_dir_open,
 182         .release        = dcache_dir_close,
 183         .llseek         = dcache_dir_lseek,
 184         .read           = generic_read_dir,
 185         .readdir        = dcache_readdir,
 186         .fsync          = simple_sync_file,
 187 };
 188
 189 struct inode_operations simple_dir_inode_operations = {
 190         .lookup         = simple_lookup,
 191 };
 192
 193 /*
 194  * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
 195  * will never be mountable)
 196  */
 197 int get_sb_pseudo(struct file_system_type *fs_type, char *name,
 198         struct super_operations *ops, unsigned long magic,
 199         struct vfsmount *mnt)
 200 {
 201         struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
 202         static struct super_operations default_ops = {.statfs = simple_statfs};
 203         struct dentry *dentry;
 204         struct inode *root;
 205         struct qstr d_name = {.name = name, .len = strlen(name)};
 206
 207         if (IS_ERR(s))
 208                 return PTR_ERR(s);
 209
 210         s->s_flags = MS_NOUSER;
 211         s->s_maxbytes = ~0ULL;
 212         s->s_blocksize = 1024;
 213         s->s_blocksize_bits = 10;
 214         s->s_magic = magic;
 215         s->s_op = ops ? ops : &default_ops;
 216         s->s_time_gran = 1;
 217         root = new_inode(s);
 218         if (!root)
 219                 goto Enomem;
 220         root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
 221         root->i_uid = root->i_gid = 0;
 222         root->i_atime = root->i_mtime = root->i_ctime = CURRENT_TIME;
 223         dentry = d_alloc(NULL, &d_name);
 224         if (!dentry) {
 225                 iput(root);
 226                 goto Enomem;
 227         }
 228         dentry->d_sb = s;
 229         dentry->d_parent = dentry;
 230         d_instantiate(dentry, root);
 231         s->s_root = dentry;
 232         s->s_flags |= MS_ACTIVE;
 233         return simple_set_mnt(mnt, s);
 234
 235 Enomem:
 236         up_write(&s->s_umount);
 237         deactivate_super(s);
 238         return -ENOMEM;
 239 }
 240
 241 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
 242 {
 243         struct inode *inode = old_dentry->d_inode;
 244
 245         inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
 246         inc_nlink(inode);
 247         atomic_inc(&inode->i_count);
 248         dget(dentry);
 249         d_instantiate(dentry, inode);
 250         return 0;
 251 }
 252
 253 static inline int simple_positive(struct dentry *dentry)
 254 {
 255         return dentry->d_inode && !d_unhashed(dentry);
 256 }
 257
 258 int simple_empty(struct dentry *dentry)
 259 {
 260         struct dentry *child;
 261         int ret = 0;
 262
 263         spin_lock(&dcache_lock);
 264         list_for_each_entry(child, &dentry->d_subdirs, d_u.d_child)
 265                 if (simple_positive(child))
 266                         goto out;
 267         ret = 1;
 268 out:
 269         spin_unlock(&dcache_lock);
 270         return ret;
 271 }
 272
 273 int simple_unlink(struct inode *dir, struct dentry *dentry)
 274 {
 275         struct inode *inode = dentry->d_inode;
 276
 277         inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
 278         drop_nlink(inode);
 279         dput(dentry);
 280         return 0;
 281 }
 282
 283 int simple_rmdir(struct inode *dir, struct dentry *dentry)
 284 {
 285         if (!simple_empty(dentry))
 286                 return -ENOTEMPTY;
 287
 288         drop_nlink(dentry->d_inode);
 289         simple_unlink(dir, dentry);
 290         drop_nlink(dir);
 291         return 0;
 292 }
 293
 294 int simple_rename(struct inode *old_dir, struct dentry *old_dentry,
 295                 struct inode *new_dir, struct dentry *new_dentry)
 296 {
 297         struct inode *inode = old_dentry->d_inode;
 298         int they_are_dirs = S_ISDIR(old_dentry->d_inode->i_mode);
 299
 300         if (!simple_empty(new_dentry))
 301                 return -ENOTEMPTY;
 302
 303         if (new_dentry->d_inode) {
 304                 simple_unlink(new_dir, new_dentry);
 305                 if (they_are_dirs)
 306                         drop_nlink(old_dir);
 307         } else if (they_are_dirs) {
 308                 drop_nlink(old_dir);
 309                 inc_nlink(new_dir);
 310         }
 311
 312         old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
 313                 new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
 314
 315         return 0;
 316 }
 317
 318 int simple_readpage(struct file *file, struct page *page)
 319 {
 320         clear_highpage(page);
 321         flush_dcache_page(page);
 322         SetPageUptodate(page);
 323         unlock_page(page);
 324         return 0;
 325 }
 326
 327 int simple_prepare_write(struct file *file, struct page *page,
 328                         unsigned from, unsigned to)
 329 {
 330         if (!PageUptodate(page)) {
 331                 if (to - from != PAGE_CACHE_SIZE) {
 332                         void *kaddr = kmap_atomic(page, KM_USER0);
 333                         memset(kaddr, 0, from);
 334                         memset(kaddr + to, 0, PAGE_CACHE_SIZE - to);
 335                         flush_dcache_page(page);
 336                         kunmap_atomic(kaddr, KM_USER0);
 337                 }
 338                 SetPageUptodate(page);
 339         }
 340         return 0;
 341 }
 342
 343 int simple_commit_write(struct file *file, struct page *page,
 344                         unsigned offset, unsigned to)
 345 {
 346         struct inode *inode = page->mapping->host;
 347         loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
 348
 349         /*
 350          * No need to use i_size_read() here, the i_size
 351          * cannot change under us because we hold the i_mutex.
 352          */
 353         if (pos > inode->i_size)
 354                 i_size_write(inode, pos);
 355         set_page_dirty(page);
 356         return 0;
 357 }
 358
 359 int simple_fill_super(struct super_block *s, int magic, struct tree_descr *files)
 360 {
 361         static struct super_operations s_ops = {.statfs = simple_statfs};
 362         struct inode *inode;
 363         struct dentry *root;
 364         struct dentry *dentry;
 365         int i;
 366
 367         s->s_blocksize = PAGE_CACHE_SIZE;
 368         s->s_blocksize_bits = PAGE_CACHE_SHIFT;
 369         s->s_magic = magic;
 370         s->s_op = &s_ops;
 371         s->s_time_gran = 1;
 372
 373         inode = new_inode(s);
 374         if (!inode)
 375                 return -ENOMEM;
 376         inode->i_mode = S_IFDIR | 0755;
 377         inode->i_uid = inode->i_gid = 0;
 378         inode->i_blocks = 0;
 379         inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
 380         inode->i_op = &simple_dir_inode_operations;
 381         inode->i_fop = &simple_dir_operations;
 382         inode->i_nlink = 2;
 383         root = d_alloc_root(inode);
 384         if (!root) {
 385                 iput(inode);
 386                 return -ENOMEM;
 387         }
 388         for (i = 0; !files->name || files->name[0]; i++, files++) {
 389                 if (!files->name)
 390                         continue;
 391                 dentry = d_alloc_name(root, files->name);
 392                 if (!dentry)
 393                         goto out;
 394                 inode = new_inode(s);
 395                 if (!inode)
 396                         goto out;
 397                 inode->i_mode = S_IFREG | files->mode;
 398                 inode->i_uid = inode->i_gid = 0;
 399                 inode->i_blocks = 0;
 400                 inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
 401                 inode->i_fop = files->ops;
 402                 inode->i_ino = i;
 403                 d_add(dentry, inode);
 404         }
 405         s->s_root = root;
 406         return 0;
 407 out:
 408         d_genocide(root);
 409         dput(root);
 410         return -ENOMEM;
 411 }
 412
 413 static DEFINE_SPINLOCK(pin_fs_lock);
 414
 415 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
 416 {
 417         struct vfsmount *mnt = NULL;
 418         spin_lock(&pin_fs_lock);
 419         if (unlikely(!*mount)) {
 420                 spin_unlock(&pin_fs_lock);
 421                 mnt = vfs_kern_mount(type, 0, type->name, NULL);
 422                 if (IS_ERR(mnt))
 423                         return PTR_ERR(mnt);
 424                 spin_lock(&pin_fs_lock);
 425                 if (!*mount)
 426                         *mount = mnt;
 427         }
 428         mntget(*mount);
 429         ++*count;
 430         spin_unlock(&pin_fs_lock);
 431         mntput(mnt);
 432         return 0;
 433 }
 434
 435 void simple_release_fs(struct vfsmount **mount, int *count)
 436 {
 437         struct vfsmount *mnt;
 438         spin_lock(&pin_fs_lock);
 439         mnt = *mount;
 440         if (!--*count)
 441                 *mount = NULL;
 442         spin_unlock(&pin_fs_lock);
 443         mntput(mnt);
 444 }
 445
 446 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
 447                                 const void *from, size_t available)
 448 {
 449         loff_t pos = *ppos;
 450         if (pos < 0)
 451                 return -EINVAL;
 452         if (pos >= available)
 453                 return 0;
 454         if (count > available - pos)
 455                 count = available - pos;
 456         if (copy_to_user(to, from + pos, count))
 457                 return -EFAULT;
 458         *ppos = pos + count;
 459         return count;
 460 }
 461
 462 /*
 463  * Transaction based IO.
 464  * The file expects a single write which triggers the transaction, and then
 465  * possibly a read which collects the result - which is stored in a
 466  * file-local buffer.
 467  */
 468 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
 469 {
 470         struct simple_transaction_argresp *ar;
 471         static DEFINE_SPINLOCK(simple_transaction_lock);
 472
 473         if (size > SIMPLE_TRANSACTION_LIMIT - 1)
 474                 return ERR_PTR(-EFBIG);
 475
 476         ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
 477         if (!ar)
 478                 return ERR_PTR(-ENOMEM);
 479
 480         spin_lock(&simple_transaction_lock);
 481
 482         /* only one write allowed per open */
 483         if (file->private_data) {
 484                 spin_unlock(&simple_transaction_lock);
 485                 free_page((unsigned long)ar);
 486                 return ERR_PTR(-EBUSY);
 487         }
 488
 489         file->private_data = ar;
 490
 491         spin_unlock(&simple_transaction_lock);
 492
 493         if (copy_from_user(ar->data, buf, size))
 494                 return ERR_PTR(-EFAULT);
 495
 496         return ar->data;
 497 }
 498
 499 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
 500 {
 501         struct simple_transaction_argresp *ar = file->private_data;
 502
 503         if (!ar)
 504                 return 0;
 505         return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
 506 }
 507
 508 int simple_transaction_release(struct inode *inode, struct file *file)
 509 {
 510         free_page((unsigned long)file->private_data);
 511         return 0;
 512 }
 513
 514 /* Simple attribute files */
 515
 516 struct simple_attr {
 517         u64 (*get)(void *);
 518         void (*set)(void *, u64);
 519         char get_buf[24];       /* enough to store a u64 and "\n\0" */
 520         char set_buf[24];
 521         void *data;
 522         const char *fmt;        /* format for read operation */
 523         struct mutex mutex;     /* protects access to these buffers */
 524 };
 525
 526 /* simple_attr_open is called by an actual attribute open file operation
 527  * to set the attribute specific access operations. */
 528 int simple_attr_open(struct inode *inode, struct file *file,
 529                      u64 (*get)(void *), void (*set)(void *, u64),
 530                      const char *fmt)
 531 {
 532         struct simple_attr *attr;
 533
 534         attr = kmalloc(sizeof(*attr), GFP_KERNEL);
 535         if (!attr)
 536                 return -ENOMEM;
 537
 538         attr->get = get;
 539         attr->set = set;
 540         attr->data = inode->i_private;
 541         attr->fmt = fmt;
 542         mutex_init(&attr->mutex);
 543
 544         file->private_data = attr;
 545
 546         return nonseekable_open(inode, file);
 547 }
 548
 549 int simple_attr_close(struct inode *inode, struct file *file)
 550 {
 551         kfree(file->private_data);
 552         return 0;
 553 }
 554
 555 /* read from the buffer that is filled with the get function */
 556 ssize_t simple_attr_read(struct file *file, char __user *buf,
 557                          size_t len, loff_t *ppos)
 558 {
 559         struct simple_attr *attr;
 560         size_t size;
 561         ssize_t ret;
 562
 563         attr = file->private_data;
 564
 565         if (!attr->get)
 566                 return -EACCES;
 567
 568         mutex_lock(&attr->mutex);
 569         if (*ppos) /* continued read */
 570                 size = strlen(attr->get_buf);
 571         else      /* first read */
 572                 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
 573                                  attr->fmt,
 574                                  (unsigned long long)attr->get(attr->data));
 575
 576         ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
 577         mutex_unlock(&attr->mutex);
 578         return ret;
 579 }
 580
 581 /* interpret the buffer as a number to call the set function with */
 582 ssize_t simple_attr_write(struct file *file, const char __user *buf,
 583                           size_t len, loff_t *ppos)
 584 {
 585         struct simple_attr *attr;
 586         u64 val;
 587         size_t size;
 588         ssize_t ret;
 589
 590         attr = file->private_data;
 591
 592         if (!attr->set)
 593                 return -EACCES;
 594
 595         mutex_lock(&attr->mutex);
 596         ret = -EFAULT;
 597         size = min(sizeof(attr->set_buf) - 1, len);
 598         if (copy_from_user(attr->set_buf, buf, size))
 599                 goto out;
 600
 601         ret = len; /* claim we got the whole input */
 602         attr->set_buf[size] = '\0';
 603         val = simple_strtol(attr->set_buf, NULL, 0);
 604         attr->set(attr->data, val);
 605 out:
 606         mutex_unlock(&attr->mutex);
 607         return ret;
 608 }
 609
 610 EXPORT_SYMBOL(dcache_dir_close);
 611 EXPORT_SYMBOL(dcache_dir_lseek);
 612 EXPORT_SYMBOL(dcache_dir_open);
 613 EXPORT_SYMBOL(dcache_readdir);
 614 EXPORT_SYMBOL(generic_read_dir);
 615 EXPORT_SYMBOL(get_sb_pseudo);
 616 EXPORT_SYMBOL(simple_commit_write);
 617 EXPORT_SYMBOL(simple_dir_inode_operations);
 618 EXPORT_SYMBOL(simple_dir_operations);
 619 EXPORT_SYMBOL(simple_empty);
 620 EXPORT_SYMBOL(d_alloc_name);
 621 EXPORT_SYMBOL(simple_fill_super);
 622 EXPORT_SYMBOL(simple_getattr);
 623 EXPORT_SYMBOL(simple_link);
 624 EXPORT_SYMBOL(simple_lookup);
 625 EXPORT_SYMBOL(simple_pin_fs);
 626 EXPORT_SYMBOL(simple_prepare_write);
 627 EXPORT_SYMBOL(simple_readpage);
 628 EXPORT_SYMBOL(simple_release_fs);
 629 EXPORT_SYMBOL(simple_rename);
 630 EXPORT_SYMBOL(simple_rmdir);
 631 EXPORT_SYMBOL(simple_statfs);
 632 EXPORT_SYMBOL(simple_sync_file);
 633 EXPORT_SYMBOL(simple_unlink);
 634 EXPORT_SYMBOL(simple_read_from_buffer);
 635 EXPORT_SYMBOL(simple_transaction_get);
 636 EXPORT_SYMBOL(simple_transaction_read);
 637 EXPORT_SYMBOL(simple_transaction_release);
 638 EXPORT_SYMBOL_GPL(simple_attr_open);
 639 EXPORT_SYMBOL_GPL(simple_attr_close);
 640 EXPORT_SYMBOL_GPL(simple_attr_read);
 641 EXPORT_SYMBOL_GPL(simple_attr_write);