1 // SPDX-License-Identifier: GPL-2.0-only 1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 2 /* 3 * fs/libfs.c 3 * fs/libfs.c 4 * Library for filesystems writers. 4 * Library for filesystems writers. 5 */ 5 */ 6 6 7 #include <linux/blkdev.h> 7 #include <linux/blkdev.h> 8 #include <linux/export.h> 8 #include <linux/export.h> 9 #include <linux/pagemap.h> 9 #include <linux/pagemap.h> 10 #include <linux/slab.h> 10 #include <linux/slab.h> 11 #include <linux/cred.h> 11 #include <linux/cred.h> 12 #include <linux/mount.h> 12 #include <linux/mount.h> 13 #include <linux/vfs.h> 13 #include <linux/vfs.h> 14 #include <linux/quotaops.h> 14 #include <linux/quotaops.h> 15 #include <linux/mutex.h> 15 #include <linux/mutex.h> 16 #include <linux/namei.h> 16 #include <linux/namei.h> 17 #include <linux/exportfs.h> 17 #include <linux/exportfs.h> 18 #include <linux/iversion.h> << 19 #include <linux/writeback.h> 18 #include <linux/writeback.h> 20 #include <linux/buffer_head.h> /* sync_mapping 19 #include <linux/buffer_head.h> /* sync_mapping_buffers */ 21 #include <linux/fs_context.h> 20 #include <linux/fs_context.h> 22 #include <linux/pseudo_fs.h> 21 #include <linux/pseudo_fs.h> 23 #include <linux/fsnotify.h> 22 #include <linux/fsnotify.h> 24 #include <linux/unicode.h> 23 #include <linux/unicode.h> 25 #include <linux/fscrypt.h> 24 #include <linux/fscrypt.h> 26 #include <linux/pidfs.h> << 27 25 28 #include <linux/uaccess.h> 26 #include <linux/uaccess.h> 29 27 30 #include "internal.h" 28 #include "internal.h" 31 29 32 int simple_getattr(struct mnt_idmap *idmap, co !! 30 int simple_getattr(const struct path *path, struct kstat *stat, 33 struct kstat *stat, u32 req !! 31 u32 request_mask, unsigned int query_flags) 34 unsigned int query_flags) << 35 { 32 { 36 struct inode *inode = d_inode(path->de 33 struct inode *inode = d_inode(path->dentry); 37 generic_fillattr(&nop_mnt_idmap, reque !! 34 generic_fillattr(inode, stat); 38 stat->blocks = inode->i_mapping->nrpag 35 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9); 39 return 0; 36 return 0; 40 } 37 } 41 EXPORT_SYMBOL(simple_getattr); 38 EXPORT_SYMBOL(simple_getattr); 42 39 43 int simple_statfs(struct dentry *dentry, struc 40 int simple_statfs(struct dentry *dentry, struct kstatfs *buf) 44 { 41 { 45 u64 id = huge_encode_dev(dentry->d_sb- << 46 << 47 buf->f_fsid = u64_to_fsid(id); << 48 buf->f_type = dentry->d_sb->s_magic; 42 buf->f_type = dentry->d_sb->s_magic; 49 buf->f_bsize = PAGE_SIZE; 43 buf->f_bsize = PAGE_SIZE; 50 buf->f_namelen = NAME_MAX; 44 buf->f_namelen = NAME_MAX; 51 return 0; 45 return 0; 52 } 46 } 53 EXPORT_SYMBOL(simple_statfs); 47 EXPORT_SYMBOL(simple_statfs); 54 48 55 /* 49 /* 56 * Retaining negative dentries for an in-memor 50 * Retaining negative dentries for an in-memory filesystem just wastes 57 * memory and lookup time: arrange for them to 51 * memory and lookup time: arrange for them to be deleted immediately. 58 */ 52 */ 59 int always_delete_dentry(const struct dentry * 53 int always_delete_dentry(const struct dentry *dentry) 60 { 54 { 61 return 1; 55 return 1; 62 } 56 } 63 EXPORT_SYMBOL(always_delete_dentry); 57 EXPORT_SYMBOL(always_delete_dentry); 64 58 65 const struct dentry_operations simple_dentry_o 59 const struct dentry_operations simple_dentry_operations = { 66 .d_delete = always_delete_dentry, 60 .d_delete = always_delete_dentry, 67 }; 61 }; 68 EXPORT_SYMBOL(simple_dentry_operations); 62 EXPORT_SYMBOL(simple_dentry_operations); 69 63 70 /* 64 /* 71 * Lookup the data. This is trivial - if the d 65 * Lookup the data. This is trivial - if the dentry didn't already 72 * exist, we know it is negative. Set d_op to 66 * exist, we know it is negative. Set d_op to delete negative dentries. 73 */ 67 */ 74 struct dentry *simple_lookup(struct inode *dir 68 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) 75 { 69 { 76 if (dentry->d_name.len > NAME_MAX) 70 if (dentry->d_name.len > NAME_MAX) 77 return ERR_PTR(-ENAMETOOLONG); 71 return ERR_PTR(-ENAMETOOLONG); 78 if (!dentry->d_sb->s_d_op) 72 if (!dentry->d_sb->s_d_op) 79 d_set_d_op(dentry, &simple_den 73 d_set_d_op(dentry, &simple_dentry_operations); 80 d_add(dentry, NULL); 74 d_add(dentry, NULL); 81 return NULL; 75 return NULL; 82 } 76 } 83 EXPORT_SYMBOL(simple_lookup); 77 EXPORT_SYMBOL(simple_lookup); 84 78 85 int dcache_dir_open(struct inode *inode, struc 79 int dcache_dir_open(struct inode *inode, struct file *file) 86 { 80 { 87 file->private_data = d_alloc_cursor(fi 81 file->private_data = d_alloc_cursor(file->f_path.dentry); 88 82 89 return file->private_data ? 0 : -ENOME 83 return file->private_data ? 0 : -ENOMEM; 90 } 84 } 91 EXPORT_SYMBOL(dcache_dir_open); 85 EXPORT_SYMBOL(dcache_dir_open); 92 86 93 int dcache_dir_close(struct inode *inode, stru 87 int dcache_dir_close(struct inode *inode, struct file *file) 94 { 88 { 95 dput(file->private_data); 89 dput(file->private_data); 96 return 0; 90 return 0; 97 } 91 } 98 EXPORT_SYMBOL(dcache_dir_close); 92 EXPORT_SYMBOL(dcache_dir_close); 99 93 100 /* parent is locked at least shared */ 94 /* parent is locked at least shared */ 101 /* 95 /* 102 * Returns an element of siblings' list. 96 * Returns an element of siblings' list. 103 * We are looking for <count>th positive after 97 * We are looking for <count>th positive after <p>; if 104 * found, dentry is grabbed and returned to ca 98 * found, dentry is grabbed and returned to caller. 105 * If no such element exists, NULL is returned 99 * If no such element exists, NULL is returned. 106 */ 100 */ 107 static struct dentry *scan_positives(struct de 101 static struct dentry *scan_positives(struct dentry *cursor, 108 struct !! 102 struct list_head *p, 109 loff_t 103 loff_t count, 110 struct 104 struct dentry *last) 111 { 105 { 112 struct dentry *dentry = cursor->d_pare 106 struct dentry *dentry = cursor->d_parent, *found = NULL; 113 107 114 spin_lock(&dentry->d_lock); 108 spin_lock(&dentry->d_lock); 115 while (*p) { !! 109 while ((p = p->next) != &dentry->d_subdirs) { 116 struct dentry *d = hlist_entry !! 110 struct dentry *d = list_entry(p, struct dentry, d_child); 117 p = &d->d_sib.next; << 118 // we must at least skip curso 111 // we must at least skip cursors, to avoid livelocks 119 if (d->d_flags & DCACHE_DENTRY 112 if (d->d_flags & DCACHE_DENTRY_CURSOR) 120 continue; 113 continue; 121 if (simple_positive(d) && !--c 114 if (simple_positive(d) && !--count) { 122 spin_lock_nested(&d->d 115 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); 123 if (simple_positive(d) 116 if (simple_positive(d)) 124 found = dget_d 117 found = dget_dlock(d); 125 spin_unlock(&d->d_lock 118 spin_unlock(&d->d_lock); 126 if (likely(found)) 119 if (likely(found)) 127 break; 120 break; 128 count = 1; 121 count = 1; 129 } 122 } 130 if (need_resched()) { 123 if (need_resched()) { 131 if (!hlist_unhashed(&c !! 124 list_move(&cursor->d_child, p); 132 __hlist_del(&c !! 125 p = &cursor->d_child; 133 hlist_add_behind(&curs << 134 p = &cursor->d_sib.nex << 135 spin_unlock(&dentry->d 126 spin_unlock(&dentry->d_lock); 136 cond_resched(); 127 cond_resched(); 137 spin_lock(&dentry->d_l 128 spin_lock(&dentry->d_lock); 138 } 129 } 139 } 130 } 140 spin_unlock(&dentry->d_lock); 131 spin_unlock(&dentry->d_lock); 141 dput(last); 132 dput(last); 142 return found; 133 return found; 143 } 134 } 144 135 145 loff_t dcache_dir_lseek(struct file *file, lof 136 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) 146 { 137 { 147 struct dentry *dentry = file->f_path.d 138 struct dentry *dentry = file->f_path.dentry; 148 switch (whence) { 139 switch (whence) { 149 case 1: 140 case 1: 150 offset += file->f_pos; 141 offset += file->f_pos; 151 fallthrough; 142 fallthrough; 152 case 0: 143 case 0: 153 if (offset >= 0) 144 if (offset >= 0) 154 break; 145 break; 155 fallthrough; 146 fallthrough; 156 default: 147 default: 157 return -EINVAL; 148 return -EINVAL; 158 } 149 } 159 if (offset != file->f_pos) { 150 if (offset != file->f_pos) { 160 struct dentry *cursor = file-> 151 struct dentry *cursor = file->private_data; 161 struct dentry *to = NULL; 152 struct dentry *to = NULL; 162 153 163 inode_lock_shared(dentry->d_in 154 inode_lock_shared(dentry->d_inode); 164 155 165 if (offset > 2) 156 if (offset > 2) 166 to = scan_positives(cu !! 157 to = scan_positives(cursor, &dentry->d_subdirs, 167 of 158 offset - 2, NULL); 168 spin_lock(&dentry->d_lock); 159 spin_lock(&dentry->d_lock); 169 hlist_del_init(&cursor->d_sib) << 170 if (to) 160 if (to) 171 hlist_add_behind(&curs !! 161 list_move(&cursor->d_child, &to->d_child); >> 162 else >> 163 list_del_init(&cursor->d_child); 172 spin_unlock(&dentry->d_lock); 164 spin_unlock(&dentry->d_lock); 173 dput(to); 165 dput(to); 174 166 175 file->f_pos = offset; 167 file->f_pos = offset; 176 168 177 inode_unlock_shared(dentry->d_ 169 inode_unlock_shared(dentry->d_inode); 178 } 170 } 179 return offset; 171 return offset; 180 } 172 } 181 EXPORT_SYMBOL(dcache_dir_lseek); 173 EXPORT_SYMBOL(dcache_dir_lseek); 182 174 >> 175 /* Relationship between i_mode and the DT_xxx types */ >> 176 static inline unsigned char dt_type(struct inode *inode) >> 177 { >> 178 return (inode->i_mode >> 12) & 15; >> 179 } >> 180 183 /* 181 /* 184 * Directory is locked and all positive dentri 182 * Directory is locked and all positive dentries in it are safe, since 185 * for ramfs-type trees they can't go away wit 183 * for ramfs-type trees they can't go away without unlink() or rmdir(), 186 * both impossible due to the lock on director 184 * both impossible due to the lock on directory. 187 */ 185 */ 188 186 189 int dcache_readdir(struct file *file, struct d 187 int dcache_readdir(struct file *file, struct dir_context *ctx) 190 { 188 { 191 struct dentry *dentry = file->f_path.d 189 struct dentry *dentry = file->f_path.dentry; 192 struct dentry *cursor = file->private_ 190 struct dentry *cursor = file->private_data; >> 191 struct list_head *anchor = &dentry->d_subdirs; 193 struct dentry *next = NULL; 192 struct dentry *next = NULL; 194 struct hlist_node **p; !! 193 struct list_head *p; 195 194 196 if (!dir_emit_dots(file, ctx)) 195 if (!dir_emit_dots(file, ctx)) 197 return 0; 196 return 0; 198 197 199 if (ctx->pos == 2) 198 if (ctx->pos == 2) 200 p = &dentry->d_children.first; !! 199 p = anchor; >> 200 else if (!list_empty(&cursor->d_child)) >> 201 p = &cursor->d_child; 201 else 202 else 202 p = &cursor->d_sib.next; !! 203 return 0; 203 204 204 while ((next = scan_positives(cursor, 205 while ((next = scan_positives(cursor, p, 1, next)) != NULL) { 205 if (!dir_emit(ctx, next->d_nam 206 if (!dir_emit(ctx, next->d_name.name, next->d_name.len, 206 d_inode(next)->i !! 207 d_inode(next)->i_ino, dt_type(d_inode(next)))) 207 fs_umode_to_dtyp << 208 break; 208 break; 209 ctx->pos++; 209 ctx->pos++; 210 p = &next->d_sib.next; !! 210 p = &next->d_child; 211 } 211 } 212 spin_lock(&dentry->d_lock); 212 spin_lock(&dentry->d_lock); 213 hlist_del_init(&cursor->d_sib); << 214 if (next) 213 if (next) 215 hlist_add_before(&cursor->d_si !! 214 list_move_tail(&cursor->d_child, &next->d_child); >> 215 else >> 216 list_del_init(&cursor->d_child); 216 spin_unlock(&dentry->d_lock); 217 spin_unlock(&dentry->d_lock); 217 dput(next); 218 dput(next); 218 219 219 return 0; 220 return 0; 220 } 221 } 221 EXPORT_SYMBOL(dcache_readdir); 222 EXPORT_SYMBOL(dcache_readdir); 222 223 223 ssize_t generic_read_dir(struct file *filp, ch 224 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) 224 { 225 { 225 return -EISDIR; 226 return -EISDIR; 226 } 227 } 227 EXPORT_SYMBOL(generic_read_dir); 228 EXPORT_SYMBOL(generic_read_dir); 228 229 229 const struct file_operations simple_dir_operat 230 const struct file_operations simple_dir_operations = { 230 .open = dcache_dir_open, 231 .open = dcache_dir_open, 231 .release = dcache_dir_close, 232 .release = dcache_dir_close, 232 .llseek = dcache_dir_lseek, 233 .llseek = dcache_dir_lseek, 233 .read = generic_read_dir, 234 .read = generic_read_dir, 234 .iterate_shared = dcache_readdir, 235 .iterate_shared = dcache_readdir, 235 .fsync = noop_fsync, 236 .fsync = noop_fsync, 236 }; 237 }; 237 EXPORT_SYMBOL(simple_dir_operations); 238 EXPORT_SYMBOL(simple_dir_operations); 238 239 239 const struct inode_operations simple_dir_inode 240 const struct inode_operations simple_dir_inode_operations = { 240 .lookup = simple_lookup, 241 .lookup = simple_lookup, 241 }; 242 }; 242 EXPORT_SYMBOL(simple_dir_inode_operations); 243 EXPORT_SYMBOL(simple_dir_inode_operations); 243 244 244 /* 0 is '.', 1 is '..', so always start with o << 245 enum { << 246 DIR_OFFSET_MIN = 2, << 247 }; << 248 << 249 static void offset_set(struct dentry *dentry, << 250 { << 251 dentry->d_fsdata = (void *)offset; << 252 } << 253 << 254 static long dentry2offset(struct dentry *dentr << 255 { << 256 return (long)dentry->d_fsdata; << 257 } << 258 << 259 static struct lock_class_key simple_offset_loc << 260 << 261 /** << 262 * simple_offset_init - initialize an offset_c << 263 * @octx: directory offset map to be initializ << 264 * << 265 */ << 266 void simple_offset_init(struct offset_ctx *oct << 267 { << 268 mt_init_flags(&octx->mt, MT_FLAGS_ALLO << 269 lockdep_set_class(&octx->mt.ma_lock, & << 270 octx->next_offset = DIR_OFFSET_MIN; << 271 } << 272 << 273 /** << 274 * simple_offset_add - Add an entry to a direc << 275 * @octx: directory offset ctx to be updated << 276 * @dentry: new dentry being added << 277 * << 278 * Returns zero on success. @octx and the dent << 279 * Otherwise, a negative errno value is return << 280 */ << 281 int simple_offset_add(struct offset_ctx *octx, << 282 { << 283 unsigned long offset; << 284 int ret; << 285 << 286 if (dentry2offset(dentry) != 0) << 287 return -EBUSY; << 288 << 289 ret = mtree_alloc_cyclic(&octx->mt, &o << 290 LONG_MAX, &oc << 291 if (ret < 0) << 292 return ret; << 293 << 294 offset_set(dentry, offset); << 295 return 0; << 296 } << 297 << 298 static int simple_offset_replace(struct offset << 299 long offset) << 300 { << 301 int ret; << 302 << 303 ret = mtree_store(&octx->mt, offset, d << 304 if (ret) << 305 return ret; << 306 offset_set(dentry, offset); << 307 return 0; << 308 } << 309 << 310 /** << 311 * simple_offset_remove - Remove an entry to a << 312 * @octx: directory offset ctx to be updated << 313 * @dentry: dentry being removed << 314 * << 315 */ << 316 void simple_offset_remove(struct offset_ctx *o << 317 { << 318 long offset; << 319 << 320 offset = dentry2offset(dentry); << 321 if (offset == 0) << 322 return; << 323 << 324 mtree_erase(&octx->mt, offset); << 325 offset_set(dentry, 0); << 326 } << 327 << 328 /** << 329 * simple_offset_empty - Check if a dentry can << 330 * @dentry: dentry to be tested << 331 * << 332 * Returns 0 if @dentry is a non-empty directo << 333 */ << 334 int simple_offset_empty(struct dentry *dentry) << 335 { << 336 struct inode *inode = d_inode(dentry); << 337 struct offset_ctx *octx; << 338 struct dentry *child; << 339 unsigned long index; << 340 int ret = 1; << 341 << 342 if (!inode || !S_ISDIR(inode->i_mode)) << 343 return ret; << 344 << 345 index = DIR_OFFSET_MIN; << 346 octx = inode->i_op->get_offset_ctx(ino << 347 mt_for_each(&octx->mt, child, index, L << 348 spin_lock(&child->d_lock); << 349 if (simple_positive(child)) { << 350 spin_unlock(&child->d_ << 351 ret = 0; << 352 break; << 353 } << 354 spin_unlock(&child->d_lock); << 355 } << 356 << 357 return ret; << 358 } << 359 << 360 /** << 361 * simple_offset_rename - handle directory off << 362 * @old_dir: parent directory of source entry << 363 * @old_dentry: dentry of source entry << 364 * @new_dir: parent_directory of destination e << 365 * @new_dentry: dentry of destination << 366 * << 367 * Caller provides appropriate serialization. << 368 * << 369 * User space expects the directory offset val << 370 * (new) directory entry to be unchanged after << 371 * << 372 * Returns zero on success, a negative errno v << 373 */ << 374 int simple_offset_rename(struct inode *old_dir << 375 struct inode *new_dir << 376 { << 377 struct offset_ctx *old_ctx = old_dir-> << 378 struct offset_ctx *new_ctx = new_dir-> << 379 long new_offset = dentry2offset(new_de << 380 << 381 simple_offset_remove(old_ctx, old_dent << 382 << 383 if (new_offset) { << 384 offset_set(new_dentry, 0); << 385 return simple_offset_replace(n << 386 } << 387 return simple_offset_add(new_ctx, old_ << 388 } << 389 << 390 /** << 391 * simple_offset_rename_exchange - exchange re << 392 * @old_dir: parent of dentry being moved << 393 * @old_dentry: dentry being moved << 394 * @new_dir: destination parent << 395 * @new_dentry: destination dentry << 396 * << 397 * This API preserves the directory offset val << 398 * appropriate serialization. << 399 * << 400 * Returns zero on success. Otherwise a negati << 401 * rename is rolled back. << 402 */ << 403 int simple_offset_rename_exchange(struct inode << 404 struct dentr << 405 struct inode << 406 struct dentr << 407 { << 408 struct offset_ctx *old_ctx = old_dir-> << 409 struct offset_ctx *new_ctx = new_dir-> << 410 long old_index = dentry2offset(old_den << 411 long new_index = dentry2offset(new_den << 412 int ret; << 413 << 414 simple_offset_remove(old_ctx, old_dent << 415 simple_offset_remove(new_ctx, new_dent << 416 << 417 ret = simple_offset_replace(new_ctx, o << 418 if (ret) << 419 goto out_restore; << 420 << 421 ret = simple_offset_replace(old_ctx, n << 422 if (ret) { << 423 simple_offset_remove(new_ctx, << 424 goto out_restore; << 425 } << 426 << 427 ret = simple_rename_exchange(old_dir, << 428 if (ret) { << 429 simple_offset_remove(new_ctx, << 430 simple_offset_remove(old_ctx, << 431 goto out_restore; << 432 } << 433 return 0; << 434 << 435 out_restore: << 436 (void)simple_offset_replace(old_ctx, o << 437 (void)simple_offset_replace(new_ctx, n << 438 return ret; << 439 } << 440 << 441 /** << 442 * simple_offset_destroy - Release offset map << 443 * @octx: directory offset ctx that is about t << 444 * << 445 * During fs teardown (eg. umount), a director << 446 * contain entries. xa_destroy() cleans out an << 447 */ << 448 void simple_offset_destroy(struct offset_ctx * << 449 { << 450 mtree_destroy(&octx->mt); << 451 } << 452 << 453 static int offset_dir_open(struct inode *inode << 454 { << 455 struct offset_ctx *ctx = inode->i_op-> << 456 << 457 file->private_data = (void *)ctx->next << 458 return 0; << 459 } << 460 << 461 /** << 462 * offset_dir_llseek - Advance the read positi << 463 * @file: an open directory whose position is << 464 * @offset: a byte offset << 465 * @whence: enumerator describing the starting << 466 * << 467 * SEEK_END, SEEK_DATA, and SEEK_HOLE are not << 468 * << 469 * Returns the updated read position if succes << 470 * negative errno is returned and the read pos << 471 */ << 472 static loff_t offset_dir_llseek(struct file *f << 473 { << 474 struct inode *inode = file->f_inode; << 475 struct offset_ctx *ctx = inode->i_op-> << 476 << 477 switch (whence) { << 478 case SEEK_CUR: << 479 offset += file->f_pos; << 480 fallthrough; << 481 case SEEK_SET: << 482 if (offset >= 0) << 483 break; << 484 fallthrough; << 485 default: << 486 return -EINVAL; << 487 } << 488 << 489 /* In this case, ->private_data is pro << 490 if (!offset) << 491 file->private_data = (void *)c << 492 return vfs_setpos(file, offset, LONG_M << 493 } << 494 << 495 static struct dentry *offset_find_next(struct << 496 { << 497 MA_STATE(mas, &octx->mt, offset, offse << 498 struct dentry *child, *found = NULL; << 499 << 500 rcu_read_lock(); << 501 child = mas_find(&mas, LONG_MAX); << 502 if (!child) << 503 goto out; << 504 spin_lock(&child->d_lock); << 505 if (simple_positive(child)) << 506 found = dget_dlock(child); << 507 spin_unlock(&child->d_lock); << 508 out: << 509 rcu_read_unlock(); << 510 return found; << 511 } << 512 << 513 static bool offset_dir_emit(struct dir_context << 514 { << 515 struct inode *inode = d_inode(dentry); << 516 long offset = dentry2offset(dentry); << 517 << 518 return ctx->actor(ctx, dentry->d_name. << 519 inode->i_ino, fs_umo << 520 } << 521 << 522 static void offset_iterate_dir(struct inode *i << 523 { << 524 struct offset_ctx *octx = inode->i_op- << 525 struct dentry *dentry; << 526 << 527 while (true) { << 528 dentry = offset_find_next(octx << 529 if (!dentry) << 530 return; << 531 << 532 if (dentry2offset(dentry) >= l << 533 dput(dentry); << 534 return; << 535 } << 536 << 537 if (!offset_dir_emit(ctx, dent << 538 dput(dentry); << 539 return; << 540 } << 541 << 542 ctx->pos = dentry2offset(dentr << 543 dput(dentry); << 544 } << 545 } << 546 << 547 /** << 548 * offset_readdir - Emit entries starting at o << 549 * @file: an open directory to iterate over << 550 * @ctx: directory iteration context << 551 * << 552 * Caller must hold @file's i_rwsem to prevent << 553 * entries during this call. << 554 * << 555 * On entry, @ctx->pos contains an offset that << 556 * to be read from the directory. << 557 * << 558 * The operation continues until there are no << 559 * until the ctx->actor indicates there is no << 560 * output buffer. << 561 * << 562 * On return, @ctx->pos contains an offset tha << 563 * in this directory when offset_readdir() is << 564 * << 565 * Return values: << 566 * %0 - Complete << 567 */ << 568 static int offset_readdir(struct file *file, s << 569 { << 570 struct dentry *dir = file->f_path.dent << 571 long last_index = (long)file->private_ << 572 << 573 lockdep_assert_held(&d_inode(dir)->i_r << 574 << 575 if (!dir_emit_dots(file, ctx)) << 576 return 0; << 577 << 578 offset_iterate_dir(d_inode(dir), ctx, << 579 return 0; << 580 } << 581 << 582 const struct file_operations simple_offset_dir << 583 .open = offset_dir_open, << 584 .llseek = offset_dir_llseek, << 585 .iterate_shared = offset_readdir, << 586 .read = generic_read_dir, << 587 .fsync = noop_fsync, << 588 }; << 589 << 590 static struct dentry *find_next_child(struct d 245 static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev) 591 { 246 { 592 struct dentry *child = NULL, *d; !! 247 struct dentry *child = NULL; >> 248 struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs; 593 249 594 spin_lock(&parent->d_lock); 250 spin_lock(&parent->d_lock); 595 d = prev ? d_next_sibling(prev) : d_fi !! 251 while ((p = p->next) != &parent->d_subdirs) { 596 hlist_for_each_entry_from(d, d_sib) { !! 252 struct dentry *d = container_of(p, struct dentry, d_child); 597 if (simple_positive(d)) { 253 if (simple_positive(d)) { 598 spin_lock_nested(&d->d 254 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); 599 if (simple_positive(d) 255 if (simple_positive(d)) 600 child = dget_d 256 child = dget_dlock(d); 601 spin_unlock(&d->d_lock 257 spin_unlock(&d->d_lock); 602 if (likely(child)) 258 if (likely(child)) 603 break; 259 break; 604 } 260 } 605 } 261 } 606 spin_unlock(&parent->d_lock); 262 spin_unlock(&parent->d_lock); 607 dput(prev); 263 dput(prev); 608 return child; 264 return child; 609 } 265 } 610 266 611 void simple_recursive_removal(struct dentry *d 267 void simple_recursive_removal(struct dentry *dentry, 612 void (*callback) 268 void (*callback)(struct dentry *)) 613 { 269 { 614 struct dentry *this = dget(dentry); 270 struct dentry *this = dget(dentry); 615 while (true) { 271 while (true) { 616 struct dentry *victim = NULL, 272 struct dentry *victim = NULL, *child; 617 struct inode *inode = this->d_ 273 struct inode *inode = this->d_inode; 618 274 619 inode_lock(inode); 275 inode_lock(inode); 620 if (d_is_dir(this)) 276 if (d_is_dir(this)) 621 inode->i_flags |= S_DE 277 inode->i_flags |= S_DEAD; 622 while ((child = find_next_chil 278 while ((child = find_next_child(this, victim)) == NULL) { 623 // kill and ascend 279 // kill and ascend 624 // update metadata whi 280 // update metadata while it's still locked 625 inode_set_ctime_curren !! 281 inode->i_ctime = current_time(inode); 626 clear_nlink(inode); 282 clear_nlink(inode); 627 inode_unlock(inode); 283 inode_unlock(inode); 628 victim = this; 284 victim = this; 629 this = this->d_parent; 285 this = this->d_parent; 630 inode = this->d_inode; 286 inode = this->d_inode; 631 inode_lock(inode); 287 inode_lock(inode); 632 if (simple_positive(vi 288 if (simple_positive(victim)) { 633 d_invalidate(v 289 d_invalidate(victim); // avoid lost mounts 634 if (d_is_dir(v 290 if (d_is_dir(victim)) 635 fsnoti 291 fsnotify_rmdir(inode, victim); 636 else 292 else 637 fsnoti 293 fsnotify_unlink(inode, victim); 638 if (callback) 294 if (callback) 639 callba 295 callback(victim); 640 dput(victim); 296 dput(victim); // unpin it 641 } 297 } 642 if (victim == dentry) 298 if (victim == dentry) { 643 inode_set_mtim !! 299 inode->i_ctime = inode->i_mtime = 644 !! 300 current_time(inode); 645 if (d_is_dir(d 301 if (d_is_dir(dentry)) 646 drop_n 302 drop_nlink(inode); 647 inode_unlock(i 303 inode_unlock(inode); 648 dput(dentry); 304 dput(dentry); 649 return; 305 return; 650 } 306 } 651 } 307 } 652 inode_unlock(inode); 308 inode_unlock(inode); 653 this = child; 309 this = child; 654 } 310 } 655 } 311 } 656 EXPORT_SYMBOL(simple_recursive_removal); 312 EXPORT_SYMBOL(simple_recursive_removal); 657 313 658 static const struct super_operations simple_su 314 static const struct super_operations simple_super_operations = { 659 .statfs = simple_statfs, 315 .statfs = simple_statfs, 660 }; 316 }; 661 317 662 static int pseudo_fs_fill_super(struct super_b 318 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc) 663 { 319 { 664 struct pseudo_fs_context *ctx = fc->fs 320 struct pseudo_fs_context *ctx = fc->fs_private; 665 struct inode *root; 321 struct inode *root; 666 322 667 s->s_maxbytes = MAX_LFS_FILESIZE; 323 s->s_maxbytes = MAX_LFS_FILESIZE; 668 s->s_blocksize = PAGE_SIZE; 324 s->s_blocksize = PAGE_SIZE; 669 s->s_blocksize_bits = PAGE_SHIFT; 325 s->s_blocksize_bits = PAGE_SHIFT; 670 s->s_magic = ctx->magic; 326 s->s_magic = ctx->magic; 671 s->s_op = ctx->ops ?: &simple_super_op 327 s->s_op = ctx->ops ?: &simple_super_operations; 672 s->s_xattr = ctx->xattr; 328 s->s_xattr = ctx->xattr; 673 s->s_time_gran = 1; 329 s->s_time_gran = 1; 674 root = new_inode(s); 330 root = new_inode(s); 675 if (!root) 331 if (!root) 676 return -ENOMEM; 332 return -ENOMEM; 677 333 678 /* 334 /* 679 * since this is the first inode, make 335 * since this is the first inode, make it number 1. New inodes created 680 * after this must take care not to co 336 * after this must take care not to collide with it (by passing 681 * max_reserved of 1 to iunique). 337 * max_reserved of 1 to iunique). 682 */ 338 */ 683 root->i_ino = 1; 339 root->i_ino = 1; 684 root->i_mode = S_IFDIR | S_IRUSR | S_I 340 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; 685 simple_inode_init_ts(root); !! 341 root->i_atime = root->i_mtime = root->i_ctime = current_time(root); 686 s->s_root = d_make_root(root); 342 s->s_root = d_make_root(root); 687 if (!s->s_root) 343 if (!s->s_root) 688 return -ENOMEM; 344 return -ENOMEM; 689 s->s_d_op = ctx->dops; 345 s->s_d_op = ctx->dops; 690 return 0; 346 return 0; 691 } 347 } 692 348 693 static int pseudo_fs_get_tree(struct fs_contex 349 static int pseudo_fs_get_tree(struct fs_context *fc) 694 { 350 { 695 return get_tree_nodev(fc, pseudo_fs_fi 351 return get_tree_nodev(fc, pseudo_fs_fill_super); 696 } 352 } 697 353 698 static void pseudo_fs_free(struct fs_context * 354 static void pseudo_fs_free(struct fs_context *fc) 699 { 355 { 700 kfree(fc->fs_private); 356 kfree(fc->fs_private); 701 } 357 } 702 358 703 static const struct fs_context_operations pseu 359 static const struct fs_context_operations pseudo_fs_context_ops = { 704 .free = pseudo_fs_free, 360 .free = pseudo_fs_free, 705 .get_tree = pseudo_fs_get_tree, 361 .get_tree = pseudo_fs_get_tree, 706 }; 362 }; 707 363 708 /* 364 /* 709 * Common helper for pseudo-filesystems (sockf 365 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that 710 * will never be mountable) 366 * will never be mountable) 711 */ 367 */ 712 struct pseudo_fs_context *init_pseudo(struct f 368 struct pseudo_fs_context *init_pseudo(struct fs_context *fc, 713 unsign 369 unsigned long magic) 714 { 370 { 715 struct pseudo_fs_context *ctx; 371 struct pseudo_fs_context *ctx; 716 372 717 ctx = kzalloc(sizeof(struct pseudo_fs_ 373 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL); 718 if (likely(ctx)) { 374 if (likely(ctx)) { 719 ctx->magic = magic; 375 ctx->magic = magic; 720 fc->fs_private = ctx; 376 fc->fs_private = ctx; 721 fc->ops = &pseudo_fs_context_o 377 fc->ops = &pseudo_fs_context_ops; 722 fc->sb_flags |= SB_NOUSER; 378 fc->sb_flags |= SB_NOUSER; 723 fc->global = true; 379 fc->global = true; 724 } 380 } 725 return ctx; 381 return ctx; 726 } 382 } 727 EXPORT_SYMBOL(init_pseudo); 383 EXPORT_SYMBOL(init_pseudo); 728 384 729 int simple_open(struct inode *inode, struct fi 385 int simple_open(struct inode *inode, struct file *file) 730 { 386 { 731 if (inode->i_private) 387 if (inode->i_private) 732 file->private_data = inode->i_ 388 file->private_data = inode->i_private; 733 return 0; 389 return 0; 734 } 390 } 735 EXPORT_SYMBOL(simple_open); 391 EXPORT_SYMBOL(simple_open); 736 392 737 int simple_link(struct dentry *old_dentry, str 393 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) 738 { 394 { 739 struct inode *inode = d_inode(old_dent 395 struct inode *inode = d_inode(old_dentry); 740 396 741 inode_set_mtime_to_ts(dir, !! 397 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); 742 inode_set_ctime_ << 743 inc_nlink(inode); 398 inc_nlink(inode); 744 ihold(inode); 399 ihold(inode); 745 dget(dentry); 400 dget(dentry); 746 d_instantiate(dentry, inode); 401 d_instantiate(dentry, inode); 747 return 0; 402 return 0; 748 } 403 } 749 EXPORT_SYMBOL(simple_link); 404 EXPORT_SYMBOL(simple_link); 750 405 751 int simple_empty(struct dentry *dentry) 406 int simple_empty(struct dentry *dentry) 752 { 407 { 753 struct dentry *child; 408 struct dentry *child; 754 int ret = 0; 409 int ret = 0; 755 410 756 spin_lock(&dentry->d_lock); 411 spin_lock(&dentry->d_lock); 757 hlist_for_each_entry(child, &dentry->d !! 412 list_for_each_entry(child, &dentry->d_subdirs, d_child) { 758 spin_lock_nested(&child->d_loc 413 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); 759 if (simple_positive(child)) { 414 if (simple_positive(child)) { 760 spin_unlock(&child->d_ 415 spin_unlock(&child->d_lock); 761 goto out; 416 goto out; 762 } 417 } 763 spin_unlock(&child->d_lock); 418 spin_unlock(&child->d_lock); 764 } 419 } 765 ret = 1; 420 ret = 1; 766 out: 421 out: 767 spin_unlock(&dentry->d_lock); 422 spin_unlock(&dentry->d_lock); 768 return ret; 423 return ret; 769 } 424 } 770 EXPORT_SYMBOL(simple_empty); 425 EXPORT_SYMBOL(simple_empty); 771 426 772 int simple_unlink(struct inode *dir, struct de 427 int simple_unlink(struct inode *dir, struct dentry *dentry) 773 { 428 { 774 struct inode *inode = d_inode(dentry); 429 struct inode *inode = d_inode(dentry); 775 430 776 inode_set_mtime_to_ts(dir, !! 431 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); 777 inode_set_ctime_ << 778 drop_nlink(inode); 432 drop_nlink(inode); 779 dput(dentry); 433 dput(dentry); 780 return 0; 434 return 0; 781 } 435 } 782 EXPORT_SYMBOL(simple_unlink); 436 EXPORT_SYMBOL(simple_unlink); 783 437 784 int simple_rmdir(struct inode *dir, struct den 438 int simple_rmdir(struct inode *dir, struct dentry *dentry) 785 { 439 { 786 if (!simple_empty(dentry)) 440 if (!simple_empty(dentry)) 787 return -ENOTEMPTY; 441 return -ENOTEMPTY; 788 442 789 drop_nlink(d_inode(dentry)); 443 drop_nlink(d_inode(dentry)); 790 simple_unlink(dir, dentry); 444 simple_unlink(dir, dentry); 791 drop_nlink(dir); 445 drop_nlink(dir); 792 return 0; 446 return 0; 793 } 447 } 794 EXPORT_SYMBOL(simple_rmdir); 448 EXPORT_SYMBOL(simple_rmdir); 795 449 796 /** !! 450 int simple_rename(struct inode *old_dir, struct dentry *old_dentry, 797 * simple_rename_timestamp - update the variou !! 451 struct inode *new_dir, struct dentry *new_dentry, 798 * @old_dir: old parent directory !! 452 unsigned int flags) 799 * @old_dentry: dentry that is being renamed << 800 * @new_dir: new parent directory << 801 * @new_dentry: target for rename << 802 * << 803 * POSIX mandates that the old and new parent << 804 * mtime updated, and that inodes of @old_dent << 805 * their ctime updated. << 806 */ << 807 void simple_rename_timestamp(struct inode *old << 808 struct inode *new << 809 { << 810 struct inode *newino = d_inode(new_den << 811 << 812 inode_set_mtime_to_ts(old_dir, inode_s << 813 if (new_dir != old_dir) << 814 inode_set_mtime_to_ts(new_dir, << 815 inode_se << 816 inode_set_ctime_current(d_inode(old_de << 817 if (newino) << 818 inode_set_ctime_current(newino << 819 } << 820 EXPORT_SYMBOL_GPL(simple_rename_timestamp); << 821 << 822 int simple_rename_exchange(struct inode *old_d << 823 struct inode *new_d << 824 { << 825 bool old_is_dir = d_is_dir(old_dentry) << 826 bool new_is_dir = d_is_dir(new_dentry) << 827 << 828 if (old_dir != new_dir && old_is_dir ! << 829 if (old_is_dir) { << 830 drop_nlink(old_dir); << 831 inc_nlink(new_dir); << 832 } else { << 833 drop_nlink(new_dir); << 834 inc_nlink(old_dir); << 835 } << 836 } << 837 simple_rename_timestamp(old_dir, old_d << 838 return 0; << 839 } << 840 EXPORT_SYMBOL_GPL(simple_rename_exchange); << 841 << 842 int simple_rename(struct mnt_idmap *idmap, str << 843 struct dentry *old_dentry, s << 844 struct dentry *new_dentry, u << 845 { 453 { >> 454 struct inode *inode = d_inode(old_dentry); 846 int they_are_dirs = d_is_dir(old_dentr 455 int they_are_dirs = d_is_dir(old_dentry); 847 456 848 if (flags & ~(RENAME_NOREPLACE | RENAM !! 457 if (flags & ~RENAME_NOREPLACE) 849 return -EINVAL; 458 return -EINVAL; 850 459 851 if (flags & RENAME_EXCHANGE) << 852 return simple_rename_exchange( << 853 << 854 if (!simple_empty(new_dentry)) 460 if (!simple_empty(new_dentry)) 855 return -ENOTEMPTY; 461 return -ENOTEMPTY; 856 462 857 if (d_really_is_positive(new_dentry)) 463 if (d_really_is_positive(new_dentry)) { 858 simple_unlink(new_dir, new_den 464 simple_unlink(new_dir, new_dentry); 859 if (they_are_dirs) { 465 if (they_are_dirs) { 860 drop_nlink(d_inode(new 466 drop_nlink(d_inode(new_dentry)); 861 drop_nlink(old_dir); 467 drop_nlink(old_dir); 862 } 468 } 863 } else if (they_are_dirs) { 469 } else if (they_are_dirs) { 864 drop_nlink(old_dir); 470 drop_nlink(old_dir); 865 inc_nlink(new_dir); 471 inc_nlink(new_dir); 866 } 472 } 867 473 868 simple_rename_timestamp(old_dir, old_d !! 474 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime = >> 475 new_dir->i_mtime = inode->i_ctime = current_time(old_dir); >> 476 869 return 0; 477 return 0; 870 } 478 } 871 EXPORT_SYMBOL(simple_rename); 479 EXPORT_SYMBOL(simple_rename); 872 480 873 /** 481 /** 874 * simple_setattr - setattr for simple filesys 482 * simple_setattr - setattr for simple filesystem 875 * @idmap: idmap of the target mount << 876 * @dentry: dentry 483 * @dentry: dentry 877 * @iattr: iattr structure 484 * @iattr: iattr structure 878 * 485 * 879 * Returns 0 on success, -error on failure. 486 * Returns 0 on success, -error on failure. 880 * 487 * 881 * simple_setattr is a simple ->setattr implem 488 * simple_setattr is a simple ->setattr implementation without a proper 882 * implementation of size changes. 489 * implementation of size changes. 883 * 490 * 884 * It can either be used for in-memory filesys 491 * It can either be used for in-memory filesystems or special files 885 * on simple regular filesystems. Anything th 492 * on simple regular filesystems. Anything that needs to change on-disk 886 * or wire state on size changes needs its own 493 * or wire state on size changes needs its own setattr method. 887 */ 494 */ 888 int simple_setattr(struct mnt_idmap *idmap, st !! 495 int simple_setattr(struct dentry *dentry, struct iattr *iattr) 889 struct iattr *iattr) << 890 { 496 { 891 struct inode *inode = d_inode(dentry); 497 struct inode *inode = d_inode(dentry); 892 int error; 498 int error; 893 499 894 error = setattr_prepare(idmap, dentry, !! 500 error = setattr_prepare(dentry, iattr); 895 if (error) 501 if (error) 896 return error; 502 return error; 897 503 898 if (iattr->ia_valid & ATTR_SIZE) 504 if (iattr->ia_valid & ATTR_SIZE) 899 truncate_setsize(inode, iattr- 505 truncate_setsize(inode, iattr->ia_size); 900 setattr_copy(idmap, inode, iattr); !! 506 setattr_copy(inode, iattr); 901 mark_inode_dirty(inode); 507 mark_inode_dirty(inode); 902 return 0; 508 return 0; 903 } 509 } 904 EXPORT_SYMBOL(simple_setattr); 510 EXPORT_SYMBOL(simple_setattr); 905 511 906 static int simple_read_folio(struct file *file !! 512 int simple_readpage(struct file *file, struct page *page) 907 { 513 { 908 folio_zero_range(folio, 0, folio_size( !! 514 clear_highpage(page); 909 flush_dcache_folio(folio); !! 515 flush_dcache_page(page); 910 folio_mark_uptodate(folio); !! 516 SetPageUptodate(page); 911 folio_unlock(folio); !! 517 unlock_page(page); 912 return 0; 518 return 0; 913 } 519 } >> 520 EXPORT_SYMBOL(simple_readpage); 914 521 915 int simple_write_begin(struct file *file, stru 522 int simple_write_begin(struct file *file, struct address_space *mapping, 916 loff_t pos, unsigned l !! 523 loff_t pos, unsigned len, unsigned flags, 917 struct folio **foliop, !! 524 struct page **pagep, void **fsdata) 918 { 525 { 919 struct folio *folio; !! 526 struct page *page; >> 527 pgoff_t index; 920 528 921 folio = __filemap_get_folio(mapping, p !! 529 index = pos >> PAGE_SHIFT; 922 mapping_gfp_mask(mappi << 923 if (IS_ERR(folio)) << 924 return PTR_ERR(folio); << 925 530 926 *foliop = folio; !! 531 page = grab_cache_page_write_begin(mapping, index, flags); >> 532 if (!page) >> 533 return -ENOMEM; >> 534 >> 535 *pagep = page; 927 536 928 if (!folio_test_uptodate(folio) && (le !! 537 if (!PageUptodate(page) && (len != PAGE_SIZE)) { 929 size_t from = offset_in_folio( !! 538 unsigned from = pos & (PAGE_SIZE - 1); 930 539 931 folio_zero_segments(folio, 0, !! 540 zero_user_segments(page, 0, from, from + len, PAGE_SIZE); 932 from + len, fo << 933 } 541 } 934 return 0; 542 return 0; 935 } 543 } 936 EXPORT_SYMBOL(simple_write_begin); 544 EXPORT_SYMBOL(simple_write_begin); 937 545 938 /** 546 /** 939 * simple_write_end - .write_end helper for no 547 * simple_write_end - .write_end helper for non-block-device FSes 940 * @file: See .write_end of address_space_oper 548 * @file: See .write_end of address_space_operations 941 * @mapping: " 549 * @mapping: " 942 * @pos: " 550 * @pos: " 943 * @len: " 551 * @len: " 944 * @copied: " 552 * @copied: " 945 * @folio: " !! 553 * @page: " 946 * @fsdata: " 554 * @fsdata: " 947 * 555 * 948 * simple_write_end does the minimum needed fo !! 556 * simple_write_end does the minimum needed for updating a page after writing is 949 * writing is done. It has the same API signat !! 557 * done. It has the same API signature as the .write_end of 950 * address_space_operations vector. So it can 558 * address_space_operations vector. So it can just be set onto .write_end for 951 * FSes that don't need any other processing. 559 * FSes that don't need any other processing. i_mutex is assumed to be held. 952 * Block based filesystems should use generic_ 560 * Block based filesystems should use generic_write_end(). 953 * NOTE: Even though i_size might get updated 561 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty 954 * is not called, so a filesystem that actuall 562 * is not called, so a filesystem that actually does store data in .write_inode 955 * should extend on what's done here with a ca 563 * should extend on what's done here with a call to mark_inode_dirty() in the 956 * case that i_size has changed. 564 * case that i_size has changed. 957 * 565 * 958 * Use *ONLY* with simple_read_folio() !! 566 * Use *ONLY* with simple_readpage() 959 */ 567 */ 960 static int simple_write_end(struct file *file, !! 568 int simple_write_end(struct file *file, struct address_space *mapping, 961 loff_t pos, unsigned l 569 loff_t pos, unsigned len, unsigned copied, 962 struct folio *folio, v !! 570 struct page *page, void *fsdata) 963 { 571 { 964 struct inode *inode = folio->mapping-> !! 572 struct inode *inode = page->mapping->host; 965 loff_t last_pos = pos + copied; 573 loff_t last_pos = pos + copied; 966 574 967 /* zero the stale part of the folio if !! 575 /* zero the stale part of the page if we did a short copy */ 968 if (!folio_test_uptodate(folio)) { !! 576 if (!PageUptodate(page)) { 969 if (copied < len) { 577 if (copied < len) { 970 size_t from = offset_i !! 578 unsigned from = pos & (PAGE_SIZE - 1); 971 579 972 folio_zero_range(folio !! 580 zero_user(page, from + copied, len - copied); 973 } 581 } 974 folio_mark_uptodate(folio); !! 582 SetPageUptodate(page); 975 } 583 } 976 /* 584 /* 977 * No need to use i_size_read() here, 585 * No need to use i_size_read() here, the i_size 978 * cannot change under us because we h 586 * cannot change under us because we hold the i_mutex. 979 */ 587 */ 980 if (last_pos > inode->i_size) 588 if (last_pos > inode->i_size) 981 i_size_write(inode, last_pos); 589 i_size_write(inode, last_pos); 982 590 983 folio_mark_dirty(folio); !! 591 set_page_dirty(page); 984 folio_unlock(folio); !! 592 unlock_page(page); 985 folio_put(folio); !! 593 put_page(page); 986 594 987 return copied; 595 return copied; 988 } 596 } 989 !! 597 EXPORT_SYMBOL(simple_write_end); 990 /* << 991 * Provides ramfs-style behavior: data in the << 992 */ << 993 const struct address_space_operations ram_aops << 994 .read_folio = simple_read_folio, << 995 .write_begin = simple_write_begin, << 996 .write_end = simple_write_end, << 997 .dirty_folio = noop_dirty_folio, << 998 }; << 999 EXPORT_SYMBOL(ram_aops); << 1000 598 1001 /* 599 /* 1002 * the inodes created here are not hashed. If 600 * the inodes created here are not hashed. If you use iunique to generate 1003 * unique inode values later for this filesys 601 * unique inode values later for this filesystem, then you must take care 1004 * to pass it an appropriate max_reserved val 602 * to pass it an appropriate max_reserved value to avoid collisions. 1005 */ 603 */ 1006 int simple_fill_super(struct super_block *s, 604 int simple_fill_super(struct super_block *s, unsigned long magic, 1007 const struct tree_descr 605 const struct tree_descr *files) 1008 { 606 { 1009 struct inode *inode; 607 struct inode *inode; >> 608 struct dentry *root; 1010 struct dentry *dentry; 609 struct dentry *dentry; 1011 int i; 610 int i; 1012 611 1013 s->s_blocksize = PAGE_SIZE; 612 s->s_blocksize = PAGE_SIZE; 1014 s->s_blocksize_bits = PAGE_SHIFT; 613 s->s_blocksize_bits = PAGE_SHIFT; 1015 s->s_magic = magic; 614 s->s_magic = magic; 1016 s->s_op = &simple_super_operations; 615 s->s_op = &simple_super_operations; 1017 s->s_time_gran = 1; 616 s->s_time_gran = 1; 1018 617 1019 inode = new_inode(s); 618 inode = new_inode(s); 1020 if (!inode) 619 if (!inode) 1021 return -ENOMEM; 620 return -ENOMEM; 1022 /* 621 /* 1023 * because the root inode is 1, the f 622 * because the root inode is 1, the files array must not contain an 1024 * entry at index 1 623 * entry at index 1 1025 */ 624 */ 1026 inode->i_ino = 1; 625 inode->i_ino = 1; 1027 inode->i_mode = S_IFDIR | 0755; 626 inode->i_mode = S_IFDIR | 0755; 1028 simple_inode_init_ts(inode); !! 627 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); 1029 inode->i_op = &simple_dir_inode_opera 628 inode->i_op = &simple_dir_inode_operations; 1030 inode->i_fop = &simple_dir_operations 629 inode->i_fop = &simple_dir_operations; 1031 set_nlink(inode, 2); 630 set_nlink(inode, 2); 1032 s->s_root = d_make_root(inode); !! 631 root = d_make_root(inode); 1033 if (!s->s_root) !! 632 if (!root) 1034 return -ENOMEM; 633 return -ENOMEM; 1035 for (i = 0; !files->name || files->na 634 for (i = 0; !files->name || files->name[0]; i++, files++) { 1036 if (!files->name) 635 if (!files->name) 1037 continue; 636 continue; 1038 637 1039 /* warn if it tries to confli 638 /* warn if it tries to conflict with the root inode */ 1040 if (unlikely(i == 1)) 639 if (unlikely(i == 1)) 1041 printk(KERN_WARNING " 640 printk(KERN_WARNING "%s: %s passed in a files array" 1042 "with an inde 641 "with an index of 1!\n", __func__, 1043 s->s_type->na 642 s->s_type->name); 1044 643 1045 dentry = d_alloc_name(s->s_ro !! 644 dentry = d_alloc_name(root, files->name); 1046 if (!dentry) 645 if (!dentry) 1047 return -ENOMEM; !! 646 goto out; 1048 inode = new_inode(s); 647 inode = new_inode(s); 1049 if (!inode) { 648 if (!inode) { 1050 dput(dentry); 649 dput(dentry); 1051 return -ENOMEM; !! 650 goto out; 1052 } 651 } 1053 inode->i_mode = S_IFREG | fil 652 inode->i_mode = S_IFREG | files->mode; 1054 simple_inode_init_ts(inode); !! 653 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); 1055 inode->i_fop = files->ops; 654 inode->i_fop = files->ops; 1056 inode->i_ino = i; 655 inode->i_ino = i; 1057 d_add(dentry, inode); 656 d_add(dentry, inode); 1058 } 657 } >> 658 s->s_root = root; 1059 return 0; 659 return 0; >> 660 out: >> 661 d_genocide(root); >> 662 shrink_dcache_parent(root); >> 663 dput(root); >> 664 return -ENOMEM; 1060 } 665 } 1061 EXPORT_SYMBOL(simple_fill_super); 666 EXPORT_SYMBOL(simple_fill_super); 1062 667 1063 static DEFINE_SPINLOCK(pin_fs_lock); 668 static DEFINE_SPINLOCK(pin_fs_lock); 1064 669 1065 int simple_pin_fs(struct file_system_type *ty 670 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) 1066 { 671 { 1067 struct vfsmount *mnt = NULL; 672 struct vfsmount *mnt = NULL; 1068 spin_lock(&pin_fs_lock); 673 spin_lock(&pin_fs_lock); 1069 if (unlikely(!*mount)) { 674 if (unlikely(!*mount)) { 1070 spin_unlock(&pin_fs_lock); 675 spin_unlock(&pin_fs_lock); 1071 mnt = vfs_kern_mount(type, SB 676 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL); 1072 if (IS_ERR(mnt)) 677 if (IS_ERR(mnt)) 1073 return PTR_ERR(mnt); 678 return PTR_ERR(mnt); 1074 spin_lock(&pin_fs_lock); 679 spin_lock(&pin_fs_lock); 1075 if (!*mount) 680 if (!*mount) 1076 *mount = mnt; 681 *mount = mnt; 1077 } 682 } 1078 mntget(*mount); 683 mntget(*mount); 1079 ++*count; 684 ++*count; 1080 spin_unlock(&pin_fs_lock); 685 spin_unlock(&pin_fs_lock); 1081 mntput(mnt); 686 mntput(mnt); 1082 return 0; 687 return 0; 1083 } 688 } 1084 EXPORT_SYMBOL(simple_pin_fs); 689 EXPORT_SYMBOL(simple_pin_fs); 1085 690 1086 void simple_release_fs(struct vfsmount **moun 691 void simple_release_fs(struct vfsmount **mount, int *count) 1087 { 692 { 1088 struct vfsmount *mnt; 693 struct vfsmount *mnt; 1089 spin_lock(&pin_fs_lock); 694 spin_lock(&pin_fs_lock); 1090 mnt = *mount; 695 mnt = *mount; 1091 if (!--*count) 696 if (!--*count) 1092 *mount = NULL; 697 *mount = NULL; 1093 spin_unlock(&pin_fs_lock); 698 spin_unlock(&pin_fs_lock); 1094 mntput(mnt); 699 mntput(mnt); 1095 } 700 } 1096 EXPORT_SYMBOL(simple_release_fs); 701 EXPORT_SYMBOL(simple_release_fs); 1097 702 1098 /** 703 /** 1099 * simple_read_from_buffer - copy data from t 704 * simple_read_from_buffer - copy data from the buffer to user space 1100 * @to: the user space buffer to read to 705 * @to: the user space buffer to read to 1101 * @count: the maximum number of bytes to rea 706 * @count: the maximum number of bytes to read 1102 * @ppos: the current position in the buffer 707 * @ppos: the current position in the buffer 1103 * @from: the buffer to read from 708 * @from: the buffer to read from 1104 * @available: the size of the buffer 709 * @available: the size of the buffer 1105 * 710 * 1106 * The simple_read_from_buffer() function rea 711 * The simple_read_from_buffer() function reads up to @count bytes from the 1107 * buffer @from at offset @ppos into the user 712 * buffer @from at offset @ppos into the user space address starting at @to. 1108 * 713 * 1109 * On success, the number of bytes read is re 714 * On success, the number of bytes read is returned and the offset @ppos is 1110 * advanced by this number, or negative value 715 * advanced by this number, or negative value is returned on error. 1111 **/ 716 **/ 1112 ssize_t simple_read_from_buffer(void __user * 717 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, 1113 const void *f 718 const void *from, size_t available) 1114 { 719 { 1115 loff_t pos = *ppos; 720 loff_t pos = *ppos; 1116 size_t ret; 721 size_t ret; 1117 722 1118 if (pos < 0) 723 if (pos < 0) 1119 return -EINVAL; 724 return -EINVAL; 1120 if (pos >= available || !count) 725 if (pos >= available || !count) 1121 return 0; 726 return 0; 1122 if (count > available - pos) 727 if (count > available - pos) 1123 count = available - pos; 728 count = available - pos; 1124 ret = copy_to_user(to, from + pos, co 729 ret = copy_to_user(to, from + pos, count); 1125 if (ret == count) 730 if (ret == count) 1126 return -EFAULT; 731 return -EFAULT; 1127 count -= ret; 732 count -= ret; 1128 *ppos = pos + count; 733 *ppos = pos + count; 1129 return count; 734 return count; 1130 } 735 } 1131 EXPORT_SYMBOL(simple_read_from_buffer); 736 EXPORT_SYMBOL(simple_read_from_buffer); 1132 737 1133 /** 738 /** 1134 * simple_write_to_buffer - copy data from us 739 * simple_write_to_buffer - copy data from user space to the buffer 1135 * @to: the buffer to write to 740 * @to: the buffer to write to 1136 * @available: the size of the buffer 741 * @available: the size of the buffer 1137 * @ppos: the current position in the buffer 742 * @ppos: the current position in the buffer 1138 * @from: the user space buffer to read from 743 * @from: the user space buffer to read from 1139 * @count: the maximum number of bytes to rea 744 * @count: the maximum number of bytes to read 1140 * 745 * 1141 * The simple_write_to_buffer() function read 746 * The simple_write_to_buffer() function reads up to @count bytes from the user 1142 * space address starting at @from into the b 747 * space address starting at @from into the buffer @to at offset @ppos. 1143 * 748 * 1144 * On success, the number of bytes written is 749 * On success, the number of bytes written is returned and the offset @ppos is 1145 * advanced by this number, or negative value 750 * advanced by this number, or negative value is returned on error. 1146 **/ 751 **/ 1147 ssize_t simple_write_to_buffer(void *to, size 752 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 1148 const void __user *from, size 753 const void __user *from, size_t count) 1149 { 754 { 1150 loff_t pos = *ppos; 755 loff_t pos = *ppos; 1151 size_t res; 756 size_t res; 1152 757 1153 if (pos < 0) 758 if (pos < 0) 1154 return -EINVAL; 759 return -EINVAL; 1155 if (pos >= available || !count) 760 if (pos >= available || !count) 1156 return 0; 761 return 0; 1157 if (count > available - pos) 762 if (count > available - pos) 1158 count = available - pos; 763 count = available - pos; 1159 res = copy_from_user(to + pos, from, 764 res = copy_from_user(to + pos, from, count); 1160 if (res == count) 765 if (res == count) 1161 return -EFAULT; 766 return -EFAULT; 1162 count -= res; 767 count -= res; 1163 *ppos = pos + count; 768 *ppos = pos + count; 1164 return count; 769 return count; 1165 } 770 } 1166 EXPORT_SYMBOL(simple_write_to_buffer); 771 EXPORT_SYMBOL(simple_write_to_buffer); 1167 772 1168 /** 773 /** 1169 * memory_read_from_buffer - copy data from t 774 * memory_read_from_buffer - copy data from the buffer 1170 * @to: the kernel space buffer to read to 775 * @to: the kernel space buffer to read to 1171 * @count: the maximum number of bytes to rea 776 * @count: the maximum number of bytes to read 1172 * @ppos: the current position in the buffer 777 * @ppos: the current position in the buffer 1173 * @from: the buffer to read from 778 * @from: the buffer to read from 1174 * @available: the size of the buffer 779 * @available: the size of the buffer 1175 * 780 * 1176 * The memory_read_from_buffer() function rea 781 * The memory_read_from_buffer() function reads up to @count bytes from the 1177 * buffer @from at offset @ppos into the kern 782 * buffer @from at offset @ppos into the kernel space address starting at @to. 1178 * 783 * 1179 * On success, the number of bytes read is re 784 * On success, the number of bytes read is returned and the offset @ppos is 1180 * advanced by this number, or negative value 785 * advanced by this number, or negative value is returned on error. 1181 **/ 786 **/ 1182 ssize_t memory_read_from_buffer(void *to, siz 787 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, 1183 const void *f 788 const void *from, size_t available) 1184 { 789 { 1185 loff_t pos = *ppos; 790 loff_t pos = *ppos; 1186 791 1187 if (pos < 0) 792 if (pos < 0) 1188 return -EINVAL; 793 return -EINVAL; 1189 if (pos >= available) 794 if (pos >= available) 1190 return 0; 795 return 0; 1191 if (count > available - pos) 796 if (count > available - pos) 1192 count = available - pos; 797 count = available - pos; 1193 memcpy(to, from + pos, count); 798 memcpy(to, from + pos, count); 1194 *ppos = pos + count; 799 *ppos = pos + count; 1195 800 1196 return count; 801 return count; 1197 } 802 } 1198 EXPORT_SYMBOL(memory_read_from_buffer); 803 EXPORT_SYMBOL(memory_read_from_buffer); 1199 804 1200 /* 805 /* 1201 * Transaction based IO. 806 * Transaction based IO. 1202 * The file expects a single write which trig 807 * The file expects a single write which triggers the transaction, and then 1203 * possibly a read which collects the result 808 * possibly a read which collects the result - which is stored in a 1204 * file-local buffer. 809 * file-local buffer. 1205 */ 810 */ 1206 811 1207 void simple_transaction_set(struct file *file 812 void simple_transaction_set(struct file *file, size_t n) 1208 { 813 { 1209 struct simple_transaction_argresp *ar 814 struct simple_transaction_argresp *ar = file->private_data; 1210 815 1211 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); 816 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); 1212 817 1213 /* 818 /* 1214 * The barrier ensures that ar->size 819 * The barrier ensures that ar->size will really remain zero until 1215 * ar->data is ready for reading. 820 * ar->data is ready for reading. 1216 */ 821 */ 1217 smp_mb(); 822 smp_mb(); 1218 ar->size = n; 823 ar->size = n; 1219 } 824 } 1220 EXPORT_SYMBOL(simple_transaction_set); 825 EXPORT_SYMBOL(simple_transaction_set); 1221 826 1222 char *simple_transaction_get(struct file *fil 827 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) 1223 { 828 { 1224 struct simple_transaction_argresp *ar 829 struct simple_transaction_argresp *ar; 1225 static DEFINE_SPINLOCK(simple_transac 830 static DEFINE_SPINLOCK(simple_transaction_lock); 1226 831 1227 if (size > SIMPLE_TRANSACTION_LIMIT - 832 if (size > SIMPLE_TRANSACTION_LIMIT - 1) 1228 return ERR_PTR(-EFBIG); 833 return ERR_PTR(-EFBIG); 1229 834 1230 ar = (struct simple_transaction_argre 835 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); 1231 if (!ar) 836 if (!ar) 1232 return ERR_PTR(-ENOMEM); 837 return ERR_PTR(-ENOMEM); 1233 838 1234 spin_lock(&simple_transaction_lock); 839 spin_lock(&simple_transaction_lock); 1235 840 1236 /* only one write allowed per open */ 841 /* only one write allowed per open */ 1237 if (file->private_data) { 842 if (file->private_data) { 1238 spin_unlock(&simple_transacti 843 spin_unlock(&simple_transaction_lock); 1239 free_page((unsigned long)ar); 844 free_page((unsigned long)ar); 1240 return ERR_PTR(-EBUSY); 845 return ERR_PTR(-EBUSY); 1241 } 846 } 1242 847 1243 file->private_data = ar; 848 file->private_data = ar; 1244 849 1245 spin_unlock(&simple_transaction_lock) 850 spin_unlock(&simple_transaction_lock); 1246 851 1247 if (copy_from_user(ar->data, buf, siz 852 if (copy_from_user(ar->data, buf, size)) 1248 return ERR_PTR(-EFAULT); 853 return ERR_PTR(-EFAULT); 1249 854 1250 return ar->data; 855 return ar->data; 1251 } 856 } 1252 EXPORT_SYMBOL(simple_transaction_get); 857 EXPORT_SYMBOL(simple_transaction_get); 1253 858 1254 ssize_t simple_transaction_read(struct file * 859 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) 1255 { 860 { 1256 struct simple_transaction_argresp *ar 861 struct simple_transaction_argresp *ar = file->private_data; 1257 862 1258 if (!ar) 863 if (!ar) 1259 return 0; 864 return 0; 1260 return simple_read_from_buffer(buf, s 865 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); 1261 } 866 } 1262 EXPORT_SYMBOL(simple_transaction_read); 867 EXPORT_SYMBOL(simple_transaction_read); 1263 868 1264 int simple_transaction_release(struct inode * 869 int simple_transaction_release(struct inode *inode, struct file *file) 1265 { 870 { 1266 free_page((unsigned long)file->privat 871 free_page((unsigned long)file->private_data); 1267 return 0; 872 return 0; 1268 } 873 } 1269 EXPORT_SYMBOL(simple_transaction_release); 874 EXPORT_SYMBOL(simple_transaction_release); 1270 875 1271 /* Simple attribute files */ 876 /* Simple attribute files */ 1272 877 1273 struct simple_attr { 878 struct simple_attr { 1274 int (*get)(void *, u64 *); 879 int (*get)(void *, u64 *); 1275 int (*set)(void *, u64); 880 int (*set)(void *, u64); 1276 char get_buf[24]; /* enough to 881 char get_buf[24]; /* enough to store a u64 and "\n\0" */ 1277 char set_buf[24]; 882 char set_buf[24]; 1278 void *data; 883 void *data; 1279 const char *fmt; /* format for 884 const char *fmt; /* format for read operation */ 1280 struct mutex mutex; /* protects a 885 struct mutex mutex; /* protects access to these buffers */ 1281 }; 886 }; 1282 887 1283 /* simple_attr_open is called by an actual at 888 /* simple_attr_open is called by an actual attribute open file operation 1284 * to set the attribute specific access opera 889 * to set the attribute specific access operations. */ 1285 int simple_attr_open(struct inode *inode, str 890 int simple_attr_open(struct inode *inode, struct file *file, 1286 int (*get)(void *, u64 * 891 int (*get)(void *, u64 *), int (*set)(void *, u64), 1287 const char *fmt) 892 const char *fmt) 1288 { 893 { 1289 struct simple_attr *attr; 894 struct simple_attr *attr; 1290 895 1291 attr = kzalloc(sizeof(*attr), GFP_KER 896 attr = kzalloc(sizeof(*attr), GFP_KERNEL); 1292 if (!attr) 897 if (!attr) 1293 return -ENOMEM; 898 return -ENOMEM; 1294 899 1295 attr->get = get; 900 attr->get = get; 1296 attr->set = set; 901 attr->set = set; 1297 attr->data = inode->i_private; 902 attr->data = inode->i_private; 1298 attr->fmt = fmt; 903 attr->fmt = fmt; 1299 mutex_init(&attr->mutex); 904 mutex_init(&attr->mutex); 1300 905 1301 file->private_data = attr; 906 file->private_data = attr; 1302 907 1303 return nonseekable_open(inode, file); 908 return nonseekable_open(inode, file); 1304 } 909 } 1305 EXPORT_SYMBOL_GPL(simple_attr_open); 910 EXPORT_SYMBOL_GPL(simple_attr_open); 1306 911 1307 int simple_attr_release(struct inode *inode, 912 int simple_attr_release(struct inode *inode, struct file *file) 1308 { 913 { 1309 kfree(file->private_data); 914 kfree(file->private_data); 1310 return 0; 915 return 0; 1311 } 916 } 1312 EXPORT_SYMBOL_GPL(simple_attr_release); /* GP 917 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ 1313 918 1314 /* read from the buffer that is filled with t 919 /* read from the buffer that is filled with the get function */ 1315 ssize_t simple_attr_read(struct file *file, c 920 ssize_t simple_attr_read(struct file *file, char __user *buf, 1316 size_t len, loff_t * 921 size_t len, loff_t *ppos) 1317 { 922 { 1318 struct simple_attr *attr; 923 struct simple_attr *attr; 1319 size_t size; 924 size_t size; 1320 ssize_t ret; 925 ssize_t ret; 1321 926 1322 attr = file->private_data; 927 attr = file->private_data; 1323 928 1324 if (!attr->get) 929 if (!attr->get) 1325 return -EACCES; 930 return -EACCES; 1326 931 1327 ret = mutex_lock_interruptible(&attr- 932 ret = mutex_lock_interruptible(&attr->mutex); 1328 if (ret) 933 if (ret) 1329 return ret; 934 return ret; 1330 935 1331 if (*ppos && attr->get_buf[0]) { 936 if (*ppos && attr->get_buf[0]) { 1332 /* continued read */ 937 /* continued read */ 1333 size = strlen(attr->get_buf); 938 size = strlen(attr->get_buf); 1334 } else { 939 } else { 1335 /* first read */ 940 /* first read */ 1336 u64 val; 941 u64 val; 1337 ret = attr->get(attr->data, & 942 ret = attr->get(attr->data, &val); 1338 if (ret) 943 if (ret) 1339 goto out; 944 goto out; 1340 945 1341 size = scnprintf(attr->get_bu 946 size = scnprintf(attr->get_buf, sizeof(attr->get_buf), 1342 attr->fmt, ( 947 attr->fmt, (unsigned long long)val); 1343 } 948 } 1344 949 1345 ret = simple_read_from_buffer(buf, le 950 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); 1346 out: 951 out: 1347 mutex_unlock(&attr->mutex); 952 mutex_unlock(&attr->mutex); 1348 return ret; 953 return ret; 1349 } 954 } 1350 EXPORT_SYMBOL_GPL(simple_attr_read); 955 EXPORT_SYMBOL_GPL(simple_attr_read); 1351 956 1352 /* interpret the buffer as a number to call t 957 /* interpret the buffer as a number to call the set function with */ 1353 static ssize_t simple_attr_write_xsigned(stru 958 static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf, 1354 size_t len, loff_t 959 size_t len, loff_t *ppos, bool is_signed) 1355 { 960 { 1356 struct simple_attr *attr; 961 struct simple_attr *attr; 1357 unsigned long long val; 962 unsigned long long val; 1358 size_t size; 963 size_t size; 1359 ssize_t ret; 964 ssize_t ret; 1360 965 1361 attr = file->private_data; 966 attr = file->private_data; 1362 if (!attr->set) 967 if (!attr->set) 1363 return -EACCES; 968 return -EACCES; 1364 969 1365 ret = mutex_lock_interruptible(&attr- 970 ret = mutex_lock_interruptible(&attr->mutex); 1366 if (ret) 971 if (ret) 1367 return ret; 972 return ret; 1368 973 1369 ret = -EFAULT; 974 ret = -EFAULT; 1370 size = min(sizeof(attr->set_buf) - 1, 975 size = min(sizeof(attr->set_buf) - 1, len); 1371 if (copy_from_user(attr->set_buf, buf 976 if (copy_from_user(attr->set_buf, buf, size)) 1372 goto out; 977 goto out; 1373 978 1374 attr->set_buf[size] = '\0'; 979 attr->set_buf[size] = '\0'; 1375 if (is_signed) 980 if (is_signed) 1376 ret = kstrtoll(attr->set_buf, 981 ret = kstrtoll(attr->set_buf, 0, &val); 1377 else 982 else 1378 ret = kstrtoull(attr->set_buf 983 ret = kstrtoull(attr->set_buf, 0, &val); 1379 if (ret) 984 if (ret) 1380 goto out; 985 goto out; 1381 ret = attr->set(attr->data, val); 986 ret = attr->set(attr->data, val); 1382 if (ret == 0) 987 if (ret == 0) 1383 ret = len; /* on success, cla 988 ret = len; /* on success, claim we got the whole input */ 1384 out: 989 out: 1385 mutex_unlock(&attr->mutex); 990 mutex_unlock(&attr->mutex); 1386 return ret; 991 return ret; 1387 } 992 } 1388 993 1389 ssize_t simple_attr_write(struct file *file, 994 ssize_t simple_attr_write(struct file *file, const char __user *buf, 1390 size_t len, loff_t 995 size_t len, loff_t *ppos) 1391 { 996 { 1392 return simple_attr_write_xsigned(file 997 return simple_attr_write_xsigned(file, buf, len, ppos, false); 1393 } 998 } 1394 EXPORT_SYMBOL_GPL(simple_attr_write); 999 EXPORT_SYMBOL_GPL(simple_attr_write); 1395 1000 1396 ssize_t simple_attr_write_signed(struct file 1001 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, 1397 size_t len, loff_t 1002 size_t len, loff_t *ppos) 1398 { 1003 { 1399 return simple_attr_write_xsigned(file 1004 return simple_attr_write_xsigned(file, buf, len, ppos, true); 1400 } 1005 } 1401 EXPORT_SYMBOL_GPL(simple_attr_write_signed); 1006 EXPORT_SYMBOL_GPL(simple_attr_write_signed); 1402 1007 1403 /** 1008 /** 1404 * generic_encode_ino32_fh - generic export_o << 1405 * @inode: the object to encode << 1406 * @fh: where to store the file handle f << 1407 * @max_len: maximum length to store there (i << 1408 * @parent: parent directory inode, if wante << 1409 * << 1410 * This generic encode_fh function assumes th << 1411 * is suitable for locating an inode, and tha << 1412 * can be used to check that it is still vali << 1413 * filehandle fragment where export_decode_fh << 1414 */ << 1415 int generic_encode_ino32_fh(struct inode *ino << 1416 struct inode *par << 1417 { << 1418 struct fid *fid = (void *)fh; << 1419 int len = *max_len; << 1420 int type = FILEID_INO32_GEN; << 1421 << 1422 if (parent && (len < 4)) { << 1423 *max_len = 4; << 1424 return FILEID_INVALID; << 1425 } else if (len < 2) { << 1426 *max_len = 2; << 1427 return FILEID_INVALID; << 1428 } << 1429 << 1430 len = 2; << 1431 fid->i32.ino = inode->i_ino; << 1432 fid->i32.gen = inode->i_generation; << 1433 if (parent) { << 1434 fid->i32.parent_ino = parent- << 1435 fid->i32.parent_gen = parent- << 1436 len = 4; << 1437 type = FILEID_INO32_GEN_PAREN << 1438 } << 1439 *max_len = len; << 1440 return type; << 1441 } << 1442 EXPORT_SYMBOL_GPL(generic_encode_ino32_fh); << 1443 << 1444 /** << 1445 * generic_fh_to_dentry - generic helper for 1009 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation 1446 * @sb: filesystem to do the file han 1010 * @sb: filesystem to do the file handle conversion on 1447 * @fid: file handle to convert 1011 * @fid: file handle to convert 1448 * @fh_len: length of the file handle in 1012 * @fh_len: length of the file handle in bytes 1449 * @fh_type: type of file handle 1013 * @fh_type: type of file handle 1450 * @get_inode: filesystem callback to retrie 1014 * @get_inode: filesystem callback to retrieve inode 1451 * 1015 * 1452 * This function decodes @fid as long as it h 1016 * This function decodes @fid as long as it has one of the well-known 1453 * Linux filehandle types and calls @get_inod 1017 * Linux filehandle types and calls @get_inode on it to retrieve the 1454 * inode for the object specified in the file 1018 * inode for the object specified in the file handle. 1455 */ 1019 */ 1456 struct dentry *generic_fh_to_dentry(struct su 1020 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, 1457 int fh_len, int fh_type, stru 1021 int fh_len, int fh_type, struct inode *(*get_inode) 1458 (struct super_block * 1022 (struct super_block *sb, u64 ino, u32 gen)) 1459 { 1023 { 1460 struct inode *inode = NULL; 1024 struct inode *inode = NULL; 1461 1025 1462 if (fh_len < 2) 1026 if (fh_len < 2) 1463 return NULL; 1027 return NULL; 1464 1028 1465 switch (fh_type) { 1029 switch (fh_type) { 1466 case FILEID_INO32_GEN: 1030 case FILEID_INO32_GEN: 1467 case FILEID_INO32_GEN_PARENT: 1031 case FILEID_INO32_GEN_PARENT: 1468 inode = get_inode(sb, fid->i3 1032 inode = get_inode(sb, fid->i32.ino, fid->i32.gen); 1469 break; 1033 break; 1470 } 1034 } 1471 1035 1472 return d_obtain_alias(inode); 1036 return d_obtain_alias(inode); 1473 } 1037 } 1474 EXPORT_SYMBOL_GPL(generic_fh_to_dentry); 1038 EXPORT_SYMBOL_GPL(generic_fh_to_dentry); 1475 1039 1476 /** 1040 /** 1477 * generic_fh_to_parent - generic helper for 1041 * generic_fh_to_parent - generic helper for the fh_to_parent export operation 1478 * @sb: filesystem to do the file han 1042 * @sb: filesystem to do the file handle conversion on 1479 * @fid: file handle to convert 1043 * @fid: file handle to convert 1480 * @fh_len: length of the file handle in 1044 * @fh_len: length of the file handle in bytes 1481 * @fh_type: type of file handle 1045 * @fh_type: type of file handle 1482 * @get_inode: filesystem callback to retrie 1046 * @get_inode: filesystem callback to retrieve inode 1483 * 1047 * 1484 * This function decodes @fid as long as it h 1048 * This function decodes @fid as long as it has one of the well-known 1485 * Linux filehandle types and calls @get_inod 1049 * Linux filehandle types and calls @get_inode on it to retrieve the 1486 * inode for the _parent_ object specified in 1050 * inode for the _parent_ object specified in the file handle if it 1487 * is specified in the file handle, or NULL o 1051 * is specified in the file handle, or NULL otherwise. 1488 */ 1052 */ 1489 struct dentry *generic_fh_to_parent(struct su 1053 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, 1490 int fh_len, int fh_type, stru 1054 int fh_len, int fh_type, struct inode *(*get_inode) 1491 (struct super_block * 1055 (struct super_block *sb, u64 ino, u32 gen)) 1492 { 1056 { 1493 struct inode *inode = NULL; 1057 struct inode *inode = NULL; 1494 1058 1495 if (fh_len <= 2) 1059 if (fh_len <= 2) 1496 return NULL; 1060 return NULL; 1497 1061 1498 switch (fh_type) { 1062 switch (fh_type) { 1499 case FILEID_INO32_GEN_PARENT: 1063 case FILEID_INO32_GEN_PARENT: 1500 inode = get_inode(sb, fid->i3 1064 inode = get_inode(sb, fid->i32.parent_ino, 1501 (fh_len > 3 1065 (fh_len > 3 ? fid->i32.parent_gen : 0)); 1502 break; 1066 break; 1503 } 1067 } 1504 1068 1505 return d_obtain_alias(inode); 1069 return d_obtain_alias(inode); 1506 } 1070 } 1507 EXPORT_SYMBOL_GPL(generic_fh_to_parent); 1071 EXPORT_SYMBOL_GPL(generic_fh_to_parent); 1508 1072 1509 /** 1073 /** 1510 * __generic_file_fsync - generic fsync imple 1074 * __generic_file_fsync - generic fsync implementation for simple filesystems 1511 * 1075 * 1512 * @file: file to synchronize 1076 * @file: file to synchronize 1513 * @start: start offset in bytes 1077 * @start: start offset in bytes 1514 * @end: end offset in bytes (inclusiv 1078 * @end: end offset in bytes (inclusive) 1515 * @datasync: only synchronize essential me 1079 * @datasync: only synchronize essential metadata if true 1516 * 1080 * 1517 * This is a generic implementation of the fs 1081 * This is a generic implementation of the fsync method for simple 1518 * filesystems which track all non-inode meta 1082 * filesystems which track all non-inode metadata in the buffers list 1519 * hanging off the address_space structure. 1083 * hanging off the address_space structure. 1520 */ 1084 */ 1521 int __generic_file_fsync(struct file *file, l 1085 int __generic_file_fsync(struct file *file, loff_t start, loff_t end, 1522 int datasync 1086 int datasync) 1523 { 1087 { 1524 struct inode *inode = file->f_mapping 1088 struct inode *inode = file->f_mapping->host; 1525 int err; 1089 int err; 1526 int ret; 1090 int ret; 1527 1091 1528 err = file_write_and_wait_range(file, 1092 err = file_write_and_wait_range(file, start, end); 1529 if (err) 1093 if (err) 1530 return err; 1094 return err; 1531 1095 1532 inode_lock(inode); 1096 inode_lock(inode); 1533 ret = sync_mapping_buffers(inode->i_m 1097 ret = sync_mapping_buffers(inode->i_mapping); 1534 if (!(inode->i_state & I_DIRTY_ALL)) 1098 if (!(inode->i_state & I_DIRTY_ALL)) 1535 goto out; 1099 goto out; 1536 if (datasync && !(inode->i_state & I_ 1100 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) 1537 goto out; 1101 goto out; 1538 1102 1539 err = sync_inode_metadata(inode, 1); 1103 err = sync_inode_metadata(inode, 1); 1540 if (ret == 0) 1104 if (ret == 0) 1541 ret = err; 1105 ret = err; 1542 1106 1543 out: 1107 out: 1544 inode_unlock(inode); 1108 inode_unlock(inode); 1545 /* check and advance again to catch e 1109 /* check and advance again to catch errors after syncing out buffers */ 1546 err = file_check_and_advance_wb_err(f 1110 err = file_check_and_advance_wb_err(file); 1547 if (ret == 0) 1111 if (ret == 0) 1548 ret = err; 1112 ret = err; 1549 return ret; 1113 return ret; 1550 } 1114 } 1551 EXPORT_SYMBOL(__generic_file_fsync); 1115 EXPORT_SYMBOL(__generic_file_fsync); 1552 1116 1553 /** 1117 /** 1554 * generic_file_fsync - generic fsync impleme 1118 * generic_file_fsync - generic fsync implementation for simple filesystems 1555 * with flush 1119 * with flush 1556 * @file: file to synchronize 1120 * @file: file to synchronize 1557 * @start: start offset in bytes 1121 * @start: start offset in bytes 1558 * @end: end offset in bytes (inclusiv 1122 * @end: end offset in bytes (inclusive) 1559 * @datasync: only synchronize essential me 1123 * @datasync: only synchronize essential metadata if true 1560 * 1124 * 1561 */ 1125 */ 1562 1126 1563 int generic_file_fsync(struct file *file, lof 1127 int generic_file_fsync(struct file *file, loff_t start, loff_t end, 1564 int datasync) 1128 int datasync) 1565 { 1129 { 1566 struct inode *inode = file->f_mapping 1130 struct inode *inode = file->f_mapping->host; 1567 int err; 1131 int err; 1568 1132 1569 err = __generic_file_fsync(file, star 1133 err = __generic_file_fsync(file, start, end, datasync); 1570 if (err) 1134 if (err) 1571 return err; 1135 return err; 1572 return blkdev_issue_flush(inode->i_sb !! 1136 return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL); 1573 } 1137 } 1574 EXPORT_SYMBOL(generic_file_fsync); 1138 EXPORT_SYMBOL(generic_file_fsync); 1575 1139 1576 /** 1140 /** 1577 * generic_check_addressable - Check addressa 1141 * generic_check_addressable - Check addressability of file system 1578 * @blocksize_bits: log of file system bl 1142 * @blocksize_bits: log of file system block size 1579 * @num_blocks: number of blocks in f 1143 * @num_blocks: number of blocks in file system 1580 * 1144 * 1581 * Determine whether a file system with @num_ 1145 * Determine whether a file system with @num_blocks blocks (and a 1582 * block size of 2**@blocksize_bits) is addre 1146 * block size of 2**@blocksize_bits) is addressable by the sector_t 1583 * and page cache of the system. Return 0 if 1147 * and page cache of the system. Return 0 if so and -EFBIG otherwise. 1584 */ 1148 */ 1585 int generic_check_addressable(unsigned blocks 1149 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) 1586 { 1150 { 1587 u64 last_fs_block = num_blocks - 1; 1151 u64 last_fs_block = num_blocks - 1; 1588 u64 last_fs_page = 1152 u64 last_fs_page = 1589 last_fs_block >> (PAGE_SHIFT 1153 last_fs_block >> (PAGE_SHIFT - blocksize_bits); 1590 1154 1591 if (unlikely(num_blocks == 0)) 1155 if (unlikely(num_blocks == 0)) 1592 return 0; 1156 return 0; 1593 1157 1594 if ((blocksize_bits < 9) || (blocksiz 1158 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) 1595 return -EINVAL; 1159 return -EINVAL; 1596 1160 1597 if ((last_fs_block > (sector_t)(~0ULL 1161 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || 1598 (last_fs_page > (pgoff_t)(~0ULL)) 1162 (last_fs_page > (pgoff_t)(~0ULL))) { 1599 return -EFBIG; 1163 return -EFBIG; 1600 } 1164 } 1601 return 0; 1165 return 0; 1602 } 1166 } 1603 EXPORT_SYMBOL(generic_check_addressable); 1167 EXPORT_SYMBOL(generic_check_addressable); 1604 1168 1605 /* 1169 /* 1606 * No-op implementation of ->fsync for in-mem 1170 * No-op implementation of ->fsync for in-memory filesystems. 1607 */ 1171 */ 1608 int noop_fsync(struct file *file, loff_t star 1172 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) 1609 { 1173 { 1610 return 0; 1174 return 0; 1611 } 1175 } 1612 EXPORT_SYMBOL(noop_fsync); 1176 EXPORT_SYMBOL(noop_fsync); 1613 1177 >> 1178 int noop_set_page_dirty(struct page *page) >> 1179 { >> 1180 /* >> 1181 * Unlike __set_page_dirty_no_writeback that handles dirty page >> 1182 * tracking in the page object, dax does all dirty tracking in >> 1183 * the inode address_space in response to mkwrite faults. In the >> 1184 * dax case we only need to worry about potentially dirty CPU >> 1185 * caches, not dirty page cache pages to write back. >> 1186 * >> 1187 * This callback is defined to prevent fallback to >> 1188 * __set_page_dirty_buffers() in set_page_dirty(). >> 1189 */ >> 1190 return 0; >> 1191 } >> 1192 EXPORT_SYMBOL_GPL(noop_set_page_dirty); >> 1193 >> 1194 void noop_invalidatepage(struct page *page, unsigned int offset, >> 1195 unsigned int length) >> 1196 { >> 1197 /* >> 1198 * There is no page cache to invalidate in the dax case, however >> 1199 * we need this callback defined to prevent falling back to >> 1200 * block_invalidatepage() in do_invalidatepage(). >> 1201 */ >> 1202 } >> 1203 EXPORT_SYMBOL_GPL(noop_invalidatepage); >> 1204 1614 ssize_t noop_direct_IO(struct kiocb *iocb, st 1205 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter) 1615 { 1206 { 1616 /* 1207 /* 1617 * iomap based filesystems support di 1208 * iomap based filesystems support direct I/O without need for 1618 * this callback. However, it still n 1209 * this callback. However, it still needs to be set in 1619 * inode->a_ops so that open/fcntl kn 1210 * inode->a_ops so that open/fcntl know that direct I/O is 1620 * generally supported. 1211 * generally supported. 1621 */ 1212 */ 1622 return -EINVAL; 1213 return -EINVAL; 1623 } 1214 } 1624 EXPORT_SYMBOL_GPL(noop_direct_IO); 1215 EXPORT_SYMBOL_GPL(noop_direct_IO); 1625 1216 1626 /* Because kfree isn't assignment-compatible 1217 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ 1627 void kfree_link(void *p) 1218 void kfree_link(void *p) 1628 { 1219 { 1629 kfree(p); 1220 kfree(p); 1630 } 1221 } 1631 EXPORT_SYMBOL(kfree_link); 1222 EXPORT_SYMBOL(kfree_link); 1632 1223 >> 1224 /* >> 1225 * nop .set_page_dirty method so that people can use .page_mkwrite on >> 1226 * anon inodes. >> 1227 */ >> 1228 static int anon_set_page_dirty(struct page *page) >> 1229 { >> 1230 return 0; >> 1231 }; >> 1232 >> 1233 /* >> 1234 * A single inode exists for all anon_inode files. Contrary to pipes, >> 1235 * anon_inode inodes have no associated per-instance data, so we need >> 1236 * only allocate one of them. >> 1237 */ 1633 struct inode *alloc_anon_inode(struct super_b 1238 struct inode *alloc_anon_inode(struct super_block *s) 1634 { 1239 { 1635 static const struct address_space_ope 1240 static const struct address_space_operations anon_aops = { 1636 .dirty_folio = noop_dirty_ !! 1241 .set_page_dirty = anon_set_page_dirty, 1637 }; 1242 }; 1638 struct inode *inode = new_inode_pseud 1243 struct inode *inode = new_inode_pseudo(s); 1639 1244 1640 if (!inode) 1245 if (!inode) 1641 return ERR_PTR(-ENOMEM); 1246 return ERR_PTR(-ENOMEM); 1642 1247 1643 inode->i_ino = get_next_ino(); 1248 inode->i_ino = get_next_ino(); 1644 inode->i_mapping->a_ops = &anon_aops; 1249 inode->i_mapping->a_ops = &anon_aops; 1645 1250 1646 /* 1251 /* 1647 * Mark the inode dirty from the very 1252 * Mark the inode dirty from the very beginning, 1648 * that way it will never be moved to 1253 * that way it will never be moved to the dirty 1649 * list because mark_inode_dirty() wi 1254 * list because mark_inode_dirty() will think 1650 * that it already _is_ on the dirty 1255 * that it already _is_ on the dirty list. 1651 */ 1256 */ 1652 inode->i_state = I_DIRTY; 1257 inode->i_state = I_DIRTY; 1653 inode->i_mode = S_IRUSR | S_IWUSR; 1258 inode->i_mode = S_IRUSR | S_IWUSR; 1654 inode->i_uid = current_fsuid(); 1259 inode->i_uid = current_fsuid(); 1655 inode->i_gid = current_fsgid(); 1260 inode->i_gid = current_fsgid(); 1656 inode->i_flags |= S_PRIVATE; 1261 inode->i_flags |= S_PRIVATE; 1657 simple_inode_init_ts(inode); !! 1262 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); 1658 return inode; 1263 return inode; 1659 } 1264 } 1660 EXPORT_SYMBOL(alloc_anon_inode); 1265 EXPORT_SYMBOL(alloc_anon_inode); 1661 1266 1662 /** 1267 /** 1663 * simple_nosetlease - generic helper for pro 1268 * simple_nosetlease - generic helper for prohibiting leases 1664 * @filp: file pointer 1269 * @filp: file pointer 1665 * @arg: type of lease to obtain 1270 * @arg: type of lease to obtain 1666 * @flp: new lease supplied for insertion 1271 * @flp: new lease supplied for insertion 1667 * @priv: private data for lm_setup operation 1272 * @priv: private data for lm_setup operation 1668 * 1273 * 1669 * Generic helper for filesystems that do not 1274 * Generic helper for filesystems that do not wish to allow leases to be set. 1670 * All arguments are ignored and it just retu 1275 * All arguments are ignored and it just returns -EINVAL. 1671 */ 1276 */ 1672 int 1277 int 1673 simple_nosetlease(struct file *filp, int arg, !! 1278 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp, 1674 void **priv) 1279 void **priv) 1675 { 1280 { 1676 return -EINVAL; 1281 return -EINVAL; 1677 } 1282 } 1678 EXPORT_SYMBOL(simple_nosetlease); 1283 EXPORT_SYMBOL(simple_nosetlease); 1679 1284 1680 /** 1285 /** 1681 * simple_get_link - generic helper to get th 1286 * simple_get_link - generic helper to get the target of "fast" symlinks 1682 * @dentry: not used here 1287 * @dentry: not used here 1683 * @inode: the symlink inode 1288 * @inode: the symlink inode 1684 * @done: not used here 1289 * @done: not used here 1685 * 1290 * 1686 * Generic helper for filesystems to use for 1291 * Generic helper for filesystems to use for symlink inodes where a pointer to 1687 * the symlink target is stored in ->i_link. 1292 * the symlink target is stored in ->i_link. NOTE: this isn't normally called, 1688 * since as an optimization the path lookup c 1293 * since as an optimization the path lookup code uses any non-NULL ->i_link 1689 * directly, without calling ->get_link(). B 1294 * directly, without calling ->get_link(). But ->get_link() still must be set, 1690 * to mark the inode_operations as being for 1295 * to mark the inode_operations as being for a symlink. 1691 * 1296 * 1692 * Return: the symlink target 1297 * Return: the symlink target 1693 */ 1298 */ 1694 const char *simple_get_link(struct dentry *de 1299 const char *simple_get_link(struct dentry *dentry, struct inode *inode, 1695 struct delayed_ca 1300 struct delayed_call *done) 1696 { 1301 { 1697 return inode->i_link; 1302 return inode->i_link; 1698 } 1303 } 1699 EXPORT_SYMBOL(simple_get_link); 1304 EXPORT_SYMBOL(simple_get_link); 1700 1305 1701 const struct inode_operations simple_symlink_ 1306 const struct inode_operations simple_symlink_inode_operations = { 1702 .get_link = simple_get_link, 1307 .get_link = simple_get_link, 1703 }; 1308 }; 1704 EXPORT_SYMBOL(simple_symlink_inode_operations 1309 EXPORT_SYMBOL(simple_symlink_inode_operations); 1705 1310 1706 /* 1311 /* 1707 * Operations for a permanently empty directo 1312 * Operations for a permanently empty directory. 1708 */ 1313 */ 1709 static struct dentry *empty_dir_lookup(struct 1314 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) 1710 { 1315 { 1711 return ERR_PTR(-ENOENT); 1316 return ERR_PTR(-ENOENT); 1712 } 1317 } 1713 1318 1714 static int empty_dir_getattr(struct mnt_idmap !! 1319 static int empty_dir_getattr(const struct path *path, struct kstat *stat, 1715 const struct pat << 1716 u32 request_mask 1320 u32 request_mask, unsigned int query_flags) 1717 { 1321 { 1718 struct inode *inode = d_inode(path->d 1322 struct inode *inode = d_inode(path->dentry); 1719 generic_fillattr(&nop_mnt_idmap, requ !! 1323 generic_fillattr(inode, stat); 1720 return 0; 1324 return 0; 1721 } 1325 } 1722 1326 1723 static int empty_dir_setattr(struct mnt_idmap !! 1327 static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr) 1724 struct dentry *d << 1725 { 1328 { 1726 return -EPERM; 1329 return -EPERM; 1727 } 1330 } 1728 1331 1729 static ssize_t empty_dir_listxattr(struct den 1332 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) 1730 { 1333 { 1731 return -EOPNOTSUPP; 1334 return -EOPNOTSUPP; 1732 } 1335 } 1733 1336 1734 static const struct inode_operations empty_di 1337 static const struct inode_operations empty_dir_inode_operations = { 1735 .lookup = empty_dir_lookup, 1338 .lookup = empty_dir_lookup, 1736 .permission = generic_permission, 1339 .permission = generic_permission, 1737 .setattr = empty_dir_setattr, 1340 .setattr = empty_dir_setattr, 1738 .getattr = empty_dir_getattr, 1341 .getattr = empty_dir_getattr, 1739 .listxattr = empty_dir_listxattr 1342 .listxattr = empty_dir_listxattr, 1740 }; 1343 }; 1741 1344 1742 static loff_t empty_dir_llseek(struct file *f 1345 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) 1743 { 1346 { 1744 /* An empty directory has two entries 1347 /* An empty directory has two entries . and .. at offsets 0 and 1 */ 1745 return generic_file_llseek_size(file, 1348 return generic_file_llseek_size(file, offset, whence, 2, 2); 1746 } 1349 } 1747 1350 1748 static int empty_dir_readdir(struct file *fil 1351 static int empty_dir_readdir(struct file *file, struct dir_context *ctx) 1749 { 1352 { 1750 dir_emit_dots(file, ctx); 1353 dir_emit_dots(file, ctx); 1751 return 0; 1354 return 0; 1752 } 1355 } 1753 1356 1754 static const struct file_operations empty_dir 1357 static const struct file_operations empty_dir_operations = { 1755 .llseek = empty_dir_llseek, 1358 .llseek = empty_dir_llseek, 1756 .read = generic_read_dir, 1359 .read = generic_read_dir, 1757 .iterate_shared = empty_dir_readdir, 1360 .iterate_shared = empty_dir_readdir, 1758 .fsync = noop_fsync, 1361 .fsync = noop_fsync, 1759 }; 1362 }; 1760 1363 1761 1364 1762 void make_empty_dir_inode(struct inode *inode 1365 void make_empty_dir_inode(struct inode *inode) 1763 { 1366 { 1764 set_nlink(inode, 2); 1367 set_nlink(inode, 2); 1765 inode->i_mode = S_IFDIR | S_IRUGO | S 1368 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; 1766 inode->i_uid = GLOBAL_ROOT_UID; 1369 inode->i_uid = GLOBAL_ROOT_UID; 1767 inode->i_gid = GLOBAL_ROOT_GID; 1370 inode->i_gid = GLOBAL_ROOT_GID; 1768 inode->i_rdev = 0; 1371 inode->i_rdev = 0; 1769 inode->i_size = 0; 1372 inode->i_size = 0; 1770 inode->i_blkbits = PAGE_SHIFT; 1373 inode->i_blkbits = PAGE_SHIFT; 1771 inode->i_blocks = 0; 1374 inode->i_blocks = 0; 1772 1375 1773 inode->i_op = &empty_dir_inode_operat 1376 inode->i_op = &empty_dir_inode_operations; 1774 inode->i_opflags &= ~IOP_XATTR; 1377 inode->i_opflags &= ~IOP_XATTR; 1775 inode->i_fop = &empty_dir_operations; 1378 inode->i_fop = &empty_dir_operations; 1776 } 1379 } 1777 1380 1778 bool is_empty_dir_inode(struct inode *inode) 1381 bool is_empty_dir_inode(struct inode *inode) 1779 { 1382 { 1780 return (inode->i_fop == &empty_dir_op 1383 return (inode->i_fop == &empty_dir_operations) && 1781 (inode->i_op == &empty_dir_in 1384 (inode->i_op == &empty_dir_inode_operations); 1782 } 1385 } 1783 1386 1784 #if IS_ENABLED(CONFIG_UNICODE) !! 1387 #ifdef CONFIG_UNICODE >> 1388 /* >> 1389 * Determine if the name of a dentry should be casefolded. >> 1390 * >> 1391 * Return: if names will need casefolding >> 1392 */ >> 1393 static bool needs_casefold(const struct inode *dir) >> 1394 { >> 1395 return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding; >> 1396 } >> 1397 1785 /** 1398 /** 1786 * generic_ci_d_compare - generic d_compare i 1399 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems 1787 * @dentry: dentry whose name we are chec 1400 * @dentry: dentry whose name we are checking against 1788 * @len: len of name of dentry 1401 * @len: len of name of dentry 1789 * @str: str pointer to name of dentry 1402 * @str: str pointer to name of dentry 1790 * @name: Name to compare against 1403 * @name: Name to compare against 1791 * 1404 * 1792 * Return: 0 if names match, 1 if mismatch, o 1405 * Return: 0 if names match, 1 if mismatch, or -ERRNO 1793 */ 1406 */ 1794 static int generic_ci_d_compare(const struct !! 1407 int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, 1795 const char *s !! 1408 const char *str, const struct qstr *name) 1796 { 1409 { 1797 const struct dentry *parent; !! 1410 const struct dentry *parent = READ_ONCE(dentry->d_parent); 1798 const struct inode *dir; !! 1411 const struct inode *dir = READ_ONCE(parent->d_inode); >> 1412 const struct super_block *sb = dentry->d_sb; >> 1413 const struct unicode_map *um = sb->s_encoding; >> 1414 struct qstr qstr = QSTR_INIT(str, len); 1799 char strbuf[DNAME_INLINE_LEN]; 1415 char strbuf[DNAME_INLINE_LEN]; 1800 struct qstr qstr; !! 1416 int ret; 1801 << 1802 /* << 1803 * Attempt a case-sensitive match fir << 1804 * should cover most lookups, includi << 1805 * applications that expect a case-se << 1806 * << 1807 * This comparison is safe under RCU << 1808 * guarantees the consistency between << 1809 * __d_lookup_rcu_op_compare() for de << 1810 */ << 1811 if (len == name->len && !memcmp(str, << 1812 return 0; << 1813 << 1814 parent = READ_ONCE(dentry->d_parent); << 1815 dir = READ_ONCE(parent->d_inode); << 1816 if (!dir || !IS_CASEFOLDED(dir)) << 1817 return 1; << 1818 1417 >> 1418 if (!dir || !needs_casefold(dir)) >> 1419 goto fallback; 1819 /* 1420 /* 1820 * If the dentry name is stored in-li 1421 * If the dentry name is stored in-line, then it may be concurrently 1821 * modified by a rename. If this hap 1422 * modified by a rename. If this happens, the VFS will eventually retry 1822 * the lookup, so it doesn't matter w 1423 * the lookup, so it doesn't matter what ->d_compare() returns. 1823 * However, it's unsafe to call utf8_ 1424 * However, it's unsafe to call utf8_strncasecmp() with an unstable 1824 * string. Therefore, we have to cop 1425 * string. Therefore, we have to copy the name into a temporary buffer. 1825 */ 1426 */ 1826 if (len <= DNAME_INLINE_LEN - 1) { 1427 if (len <= DNAME_INLINE_LEN - 1) { 1827 memcpy(strbuf, str, len); 1428 memcpy(strbuf, str, len); 1828 strbuf[len] = 0; 1429 strbuf[len] = 0; 1829 str = strbuf; !! 1430 qstr.name = strbuf; 1830 /* prevent compiler from opti 1431 /* prevent compiler from optimizing out the temporary buffer */ 1831 barrier(); 1432 barrier(); 1832 } 1433 } 1833 qstr.len = len; !! 1434 ret = utf8_strncasecmp(um, name, &qstr); 1834 qstr.name = str; !! 1435 if (ret >= 0) >> 1436 return ret; 1835 1437 1836 return utf8_strncasecmp(dentry->d_sb- !! 1438 if (sb_has_strict_encoding(sb)) >> 1439 return -EINVAL; >> 1440 fallback: >> 1441 if (len != name->len) >> 1442 return 1; >> 1443 return !!memcmp(str, name->name, len); 1837 } 1444 } >> 1445 EXPORT_SYMBOL(generic_ci_d_compare); 1838 1446 1839 /** 1447 /** 1840 * generic_ci_d_hash - generic d_hash impleme 1448 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems 1841 * @dentry: dentry of the parent director 1449 * @dentry: dentry of the parent directory 1842 * @str: qstr of name whose hash we sh 1450 * @str: qstr of name whose hash we should fill in 1843 * 1451 * 1844 * Return: 0 if hash was successful or unchan 1452 * Return: 0 if hash was successful or unchanged, and -EINVAL on error 1845 */ 1453 */ 1846 static int generic_ci_d_hash(const struct den !! 1454 int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) 1847 { 1455 { 1848 const struct inode *dir = READ_ONCE(d 1456 const struct inode *dir = READ_ONCE(dentry->d_inode); 1849 struct super_block *sb = dentry->d_sb 1457 struct super_block *sb = dentry->d_sb; 1850 const struct unicode_map *um = sb->s_ 1458 const struct unicode_map *um = sb->s_encoding; 1851 int ret; !! 1459 int ret = 0; 1852 1460 1853 if (!dir || !IS_CASEFOLDED(dir)) !! 1461 if (!dir || !needs_casefold(dir)) 1854 return 0; 1462 return 0; 1855 1463 1856 ret = utf8_casefold_hash(um, dentry, 1464 ret = utf8_casefold_hash(um, dentry, str); 1857 if (ret < 0 && sb_has_strict_encoding 1465 if (ret < 0 && sb_has_strict_encoding(sb)) 1858 return -EINVAL; 1466 return -EINVAL; 1859 return 0; 1467 return 0; 1860 } 1468 } 1861 !! 1469 EXPORT_SYMBOL(generic_ci_d_hash); 1862 static const struct dentry_operations generic << 1863 .d_hash = generic_ci_d_hash, << 1864 .d_compare = generic_ci_d_compare, << 1865 #ifdef CONFIG_FS_ENCRYPTION << 1866 .d_revalidate = fscrypt_d_revalidate, << 1867 #endif << 1868 }; << 1869 << 1870 /** << 1871 * generic_ci_match() - Match a name (case-in << 1872 * This is a filesystem helper for comparison << 1873 * generic_ci_d_compare should be used in VFS << 1874 * << 1875 * @parent: Inode of the parent of the dirent << 1876 * @name: name under lookup. << 1877 * @folded_name: Optional pre-folded name und << 1878 * @de_name: Dirent name. << 1879 * @de_name_len: dirent name length. << 1880 * << 1881 * Test whether a case-insensitive directory << 1882 * being searched. If @folded_name is provid << 1883 * recalculating the casefold of @name. << 1884 * << 1885 * Return: > 0 if the directory entry matches << 1886 * < 0 on error. << 1887 */ << 1888 int generic_ci_match(const struct inode *pare << 1889 const struct qstr *name, << 1890 const struct qstr *folde << 1891 const u8 *de_name, u32 d << 1892 { << 1893 const struct super_block *sb = parent << 1894 const struct unicode_map *um = sb->s_ << 1895 struct fscrypt_str decrypted_name = F << 1896 struct qstr dirent = QSTR_INIT(de_nam << 1897 int res = 0; << 1898 << 1899 if (IS_ENCRYPTED(parent)) { << 1900 const struct fscrypt_str encr << 1901 FSTR_INIT((u8 *) de_n << 1902 << 1903 if (WARN_ON_ONCE(!fscrypt_has << 1904 return -EINVAL; << 1905 << 1906 decrypted_name.name = kmalloc << 1907 if (!decrypted_name.name) << 1908 return -ENOMEM; << 1909 res = fscrypt_fname_disk_to_u << 1910 << 1911 if (res < 0) { << 1912 kfree(decrypted_name. << 1913 return res; << 1914 } << 1915 dirent.name = decrypted_name. << 1916 dirent.len = decrypted_name.l << 1917 } << 1918 << 1919 /* << 1920 * Attempt a case-sensitive match fir << 1921 * should cover most lookups, includi << 1922 * applications that expect a case-se << 1923 */ << 1924 << 1925 if (dirent.len == name->len && << 1926 !memcmp(name->name, dirent.name, << 1927 goto out; << 1928 << 1929 if (folded_name->name) << 1930 res = utf8_strncasecmp_folded << 1931 else << 1932 res = utf8_strncasecmp(um, na << 1933 << 1934 out: << 1935 kfree(decrypted_name.name); << 1936 if (res < 0 && sb_has_strict_encoding << 1937 pr_err_ratelimited("Directory << 1938 return 0; << 1939 } << 1940 return !res; << 1941 } << 1942 EXPORT_SYMBOL(generic_ci_match); << 1943 #endif << 1944 << 1945 #ifdef CONFIG_FS_ENCRYPTION << 1946 static const struct dentry_operations generic << 1947 .d_revalidate = fscrypt_d_revalidate, << 1948 }; << 1949 #endif << 1950 << 1951 /** << 1952 * generic_set_sb_d_ops - helper for choosing << 1953 * filesystem-wide dentry operations for the << 1954 * @sb: superblock to be configured << 1955 * << 1956 * Filesystems supporting casefolding and/or << 1957 * helper at mount-time to configure sb->s_d_ << 1958 * operations required for the enabled featur << 1959 * called after these have been configured, b << 1960 * is created. << 1961 */ << 1962 void generic_set_sb_d_ops(struct super_block << 1963 { << 1964 #if IS_ENABLED(CONFIG_UNICODE) << 1965 if (sb->s_encoding) { << 1966 sb->s_d_op = &generic_ci_dent << 1967 return; << 1968 } << 1969 #endif 1470 #endif 1970 #ifdef CONFIG_FS_ENCRYPTION << 1971 if (sb->s_cop) { << 1972 sb->s_d_op = &generic_encrypt << 1973 return; << 1974 } << 1975 #endif << 1976 } << 1977 EXPORT_SYMBOL(generic_set_sb_d_ops); << 1978 << 1979 /** << 1980 * inode_maybe_inc_iversion - increments i_ve << 1981 * @inode: inode with the i_version that shou << 1982 * @force: increment the counter even if it's << 1983 * << 1984 * Every time the inode is modified, the i_ve << 1985 * changed by any observer. << 1986 * << 1987 * If "force" is set or the QUERIED flag is s << 1988 * the value, and clear the queried flag. << 1989 * << 1990 * In the common case where neither is set, t << 1991 * updating i_version. << 1992 * << 1993 * If this function returns false, and no oth << 1994 * can avoid logging the metadata. << 1995 */ << 1996 bool inode_maybe_inc_iversion(struct inode *i << 1997 { << 1998 u64 cur, new; << 1999 << 2000 /* << 2001 * The i_version field is not strictl << 2002 * information, but the legacy inode_ << 2003 * to serialize increments. << 2004 * << 2005 * We add a full memory barrier to en << 2006 * with other state is preserved (eit << 2007 * or explicitly from smp_mb if we do << 2008 * the former). << 2009 * << 2010 * These barriers pair with inode_que << 2011 */ << 2012 cur = inode_peek_iversion_raw(inode); << 2013 if (!force && !(cur & I_VERSION_QUERI << 2014 smp_mb(); << 2015 cur = inode_peek_iversion_raw << 2016 } << 2017 << 2018 do { << 2019 /* If flag is clear then we n << 2020 if (!force && !(cur & I_VERSI << 2021 return false; << 2022 << 2023 /* Since lowest bit is flag, << 2024 new = (cur & ~I_VERSION_QUERI << 2025 } while (!atomic64_try_cmpxchg(&inode << 2026 return true; << 2027 } << 2028 EXPORT_SYMBOL(inode_maybe_inc_iversion); << 2029 << 2030 /** << 2031 * inode_query_iversion - read i_version for << 2032 * @inode: inode from which i_version should << 2033 * << 2034 * Read the inode i_version counter. This sho << 2035 * to store the returned i_version for later << 2036 * that a later query of the i_version will r << 2037 * anything has changed. << 2038 * << 2039 * In this implementation, we fetch the curre << 2040 * then try to swap it into place with a cmpx << 2041 * that fails, we try again with the newly fe << 2042 */ << 2043 u64 inode_query_iversion(struct inode *inode) << 2044 { << 2045 u64 cur, new; << 2046 bool fenced = false; << 2047 << 2048 /* << 2049 * Memory barriers (implicit in cmpxc << 2050 * inode_maybe_inc_iversion(), see th << 2051 */ << 2052 cur = inode_peek_iversion_raw(inode); << 2053 do { << 2054 /* If flag is already set, th << 2055 if (cur & I_VERSION_QUERIED) << 2056 if (!fenced) << 2057 smp_mb(); << 2058 break; << 2059 } << 2060 << 2061 fenced = true; << 2062 new = cur | I_VERSION_QUERIED << 2063 } while (!atomic64_try_cmpxchg(&inode << 2064 return cur >> I_VERSION_QUERIED_SHIFT << 2065 } << 2066 EXPORT_SYMBOL(inode_query_iversion); << 2067 << 2068 ssize_t direct_write_fallback(struct kiocb *i << 2069 ssize_t direct_written, ssize << 2070 { << 2071 struct address_space *mapping = iocb- << 2072 loff_t pos = iocb->ki_pos - buffered_ << 2073 loff_t end = iocb->ki_pos - 1; << 2074 int err; << 2075 << 2076 /* << 2077 * If the buffered write fallback ret << 2078 * the number of bytes which were wri << 2079 * code if that was zero. << 2080 * << 2081 * Note that this differs from normal << 2082 * return -EFOO even if some bytes we << 2083 */ << 2084 if (unlikely(buffered_written < 0)) { << 2085 if (direct_written) << 2086 return direct_written << 2087 return buffered_written; << 2088 } << 2089 << 2090 /* << 2091 * We need to ensure that the page ca << 2092 * invalidated to preserve the expect << 2093 */ << 2094 err = filemap_write_and_wait_range(ma << 2095 if (err < 0) { << 2096 /* << 2097 * We don't know how much we << 2098 * bytes which were direct-wr << 2099 */ << 2100 iocb->ki_pos -= buffered_writ << 2101 if (direct_written) << 2102 return direct_written << 2103 return err; << 2104 } << 2105 invalidate_mapping_pages(mapping, pos << 2106 return direct_written + buffered_writ << 2107 } << 2108 EXPORT_SYMBOL_GPL(direct_write_fallback); << 2109 << 2110 /** << 2111 * simple_inode_init_ts - initialize the time << 2112 * @inode: inode to be initialized << 2113 * << 2114 * When a new inode is created, most filesyst << 2115 * current time. Add a helper to do this. << 2116 */ << 2117 struct timespec64 simple_inode_init_ts(struct << 2118 { << 2119 struct timespec64 ts = inode_set_ctim << 2120 << 2121 inode_set_atime_to_ts(inode, ts); << 2122 inode_set_mtime_to_ts(inode, ts); << 2123 return ts; << 2124 } << 2125 EXPORT_SYMBOL(simple_inode_init_ts); << 2126 << 2127 static inline struct dentry *get_stashed_dent << 2128 { << 2129 struct dentry *dentry; << 2130 << 2131 guard(rcu)(); << 2132 dentry = rcu_dereference(*stashed); << 2133 if (!dentry) << 2134 return NULL; << 2135 if (!lockref_get_not_dead(&dentry->d_ << 2136 return NULL; << 2137 return dentry; << 2138 } << 2139 << 2140 static struct dentry *prepare_anon_dentry(str << 2141 str << 2142 voi << 2143 { << 2144 struct dentry *dentry; << 2145 struct inode *inode; << 2146 const struct stashed_operations *sops << 2147 int ret; << 2148 << 2149 inode = new_inode_pseudo(sb); << 2150 if (!inode) { << 2151 sops->put_data(data); << 2152 return ERR_PTR(-ENOMEM); << 2153 } << 2154 << 2155 inode->i_flags |= S_IMMUTABLE; << 2156 inode->i_mode = S_IFREG; << 2157 simple_inode_init_ts(inode); << 2158 << 2159 ret = sops->init_inode(inode, data); << 2160 if (ret < 0) { << 2161 iput(inode); << 2162 return ERR_PTR(ret); << 2163 } << 2164 << 2165 /* Notice when this is changed. */ << 2166 WARN_ON_ONCE(!S_ISREG(inode->i_mode)) << 2167 WARN_ON_ONCE(!IS_IMMUTABLE(inode)); << 2168 << 2169 dentry = d_alloc_anon(sb); << 2170 if (!dentry) { << 2171 iput(inode); << 2172 return ERR_PTR(-ENOMEM); << 2173 } << 2174 << 2175 /* Store address of location where de << 2176 dentry->d_fsdata = stashed; << 2177 << 2178 /* @data is now owned by the fs */ << 2179 d_instantiate(dentry, inode); << 2180 return dentry; << 2181 } << 2182 << 2183 static struct dentry *stash_dentry(struct den << 2184 struct den << 2185 { << 2186 guard(rcu)(); << 2187 for (;;) { << 2188 struct dentry *old; << 2189 << 2190 /* Assume any old dentry was << 2191 old = cmpxchg(stashed, NULL, << 2192 if (likely(!old)) << 2193 return dentry; << 2194 << 2195 /* Check if somebody else ins << 2196 if (lockref_get_not_dead(&old << 2197 return old; << 2198 << 2199 /* There's an old dead dentry << 2200 if (likely(try_cmpxchg(stashe << 2201 return dentry; << 2202 } << 2203 } << 2204 << 2205 /** << 2206 * path_from_stashed - create path from stash << 2207 * @stashed: where to retrieve or stash de << 2208 * @mnt: mnt of the filesystems to use << 2209 * @data: data to store in inode->i_pri << 2210 * @path: path to create << 2211 * << 2212 * The function tries to retrieve a stashed d << 2213 * is still valid then it will be reused. If << 2214 * will allocate a new dentry and inode. It w << 2215 * can reuse an existing dentry in case one h << 2216 * update @stashed with the newly added dentr << 2217 * << 2218 * Special-purpose helper for nsfs and pidfs. << 2219 * << 2220 * Return: On success zero and on failure a n << 2221 */ << 2222 int path_from_stashed(struct dentry **stashed << 2223 struct path *path) << 2224 { << 2225 struct dentry *dentry; << 2226 const struct stashed_operations *sops << 2227 << 2228 /* See if dentry can be reused. */ << 2229 path->dentry = get_stashed_dentry(sta << 2230 if (path->dentry) { << 2231 sops->put_data(data); << 2232 goto out_path; << 2233 } << 2234 << 2235 /* Allocate a new dentry. */ << 2236 dentry = prepare_anon_dentry(stashed, << 2237 if (IS_ERR(dentry)) << 2238 return PTR_ERR(dentry); << 2239 << 2240 /* Added a new dentry. @data is now o << 2241 path->dentry = stash_dentry(stashed, << 2242 if (path->dentry != dentry) << 2243 dput(dentry); << 2244 << 2245 out_path: << 2246 WARN_ON_ONCE(path->dentry->d_fsdata ! << 2247 WARN_ON_ONCE(d_inode(path->dentry)->i << 2248 path->mnt = mntget(mnt); << 2249 return 0; << 2250 } << 2251 << 2252 void stashed_dentry_prune(struct dentry *dent << 2253 { << 2254 struct dentry **stashed = dentry->d_f << 2255 struct inode *inode = d_inode(dentry) << 2256 << 2257 if (WARN_ON_ONCE(!stashed)) << 2258 return; << 2259 << 2260 if (!inode) << 2261 return; << 2262 << 2263 /* << 2264 * Only replace our own @dentry as so << 2265 * already cleared out @dentry and st << 2266 * dentry in there. << 2267 */ << 2268 cmpxchg(stashed, dentry, NULL); << 2269 } << 2270 1471
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.