| 1 | // SPDX-License-Identifier: GPL-2.0-only |
| 2 | /* |
| 3 | * inode.c - part of tracefs, a pseudo file system for activating tracing |
| 4 | * |
| 5 | * Based on debugfs by: Greg Kroah-Hartman <greg@kroah.com> |
| 6 | * |
| 7 | * Copyright (C) 2014 Red Hat Inc, author: Steven Rostedt <srostedt@redhat.com> |
| 8 | * |
| 9 | * tracefs is the file system that is used by the tracing infrastructure. |
| 10 | */ |
| 11 | |
| 12 | #include <linux/module.h> |
| 13 | #include <linux/fs.h> |
| 14 | #include <linux/fs_context.h> |
| 15 | #include <linux/fs_parser.h> |
| 16 | #include <linux/kobject.h> |
| 17 | #include <linux/namei.h> |
| 18 | #include <linux/tracefs.h> |
| 19 | #include <linux/fsnotify.h> |
| 20 | #include <linux/security.h> |
| 21 | #include <linux/seq_file.h> |
| 22 | #include <linux/magic.h> |
| 23 | #include <linux/slab.h> |
| 24 | #include "internal.h" |
| 25 | |
| 26 | #define TRACEFS_DEFAULT_MODE 0700 |
| 27 | static struct kmem_cache *tracefs_inode_cachep __ro_after_init; |
| 28 | |
| 29 | static struct vfsmount *tracefs_mount; |
| 30 | static int tracefs_mount_count; |
| 31 | static bool tracefs_registered; |
| 32 | |
| 33 | /* |
| 34 | * Keep track of all tracefs_inodes in order to update their |
| 35 | * flags if necessary on a remount. |
| 36 | */ |
| 37 | static DEFINE_SPINLOCK(tracefs_inode_lock); |
| 38 | static LIST_HEAD(tracefs_inodes); |
| 39 | |
| 40 | static struct inode *tracefs_alloc_inode(struct super_block *sb) |
| 41 | { |
| 42 | struct tracefs_inode *ti; |
| 43 | unsigned long flags; |
| 44 | |
| 45 | ti = alloc_inode_sb(sb, tracefs_inode_cachep, GFP_KERNEL); |
| 46 | if (!ti) |
| 47 | return NULL; |
| 48 | |
| 49 | spin_lock_irqsave(&tracefs_inode_lock, flags); |
| 50 | list_add_rcu(new: &ti->list, head: &tracefs_inodes); |
| 51 | spin_unlock_irqrestore(lock: &tracefs_inode_lock, flags); |
| 52 | |
| 53 | return &ti->vfs_inode; |
| 54 | } |
| 55 | |
| 56 | static void tracefs_free_inode(struct inode *inode) |
| 57 | { |
| 58 | struct tracefs_inode *ti = get_tracefs(inode); |
| 59 | |
| 60 | kmem_cache_free(s: tracefs_inode_cachep, objp: ti); |
| 61 | } |
| 62 | |
| 63 | static void tracefs_destroy_inode(struct inode *inode) |
| 64 | { |
| 65 | struct tracefs_inode *ti = get_tracefs(inode); |
| 66 | unsigned long flags; |
| 67 | |
| 68 | spin_lock_irqsave(&tracefs_inode_lock, flags); |
| 69 | list_del_rcu(entry: &ti->list); |
| 70 | spin_unlock_irqrestore(lock: &tracefs_inode_lock, flags); |
| 71 | } |
| 72 | |
| 73 | static ssize_t default_read_file(struct file *file, char __user *buf, |
| 74 | size_t count, loff_t *ppos) |
| 75 | { |
| 76 | return 0; |
| 77 | } |
| 78 | |
| 79 | static ssize_t default_write_file(struct file *file, const char __user *buf, |
| 80 | size_t count, loff_t *ppos) |
| 81 | { |
| 82 | return count; |
| 83 | } |
| 84 | |
| 85 | static const struct file_operations tracefs_file_operations = { |
| 86 | .read = default_read_file, |
| 87 | .write = default_write_file, |
| 88 | .open = simple_open, |
| 89 | .llseek = noop_llseek, |
| 90 | }; |
| 91 | |
| 92 | static struct tracefs_dir_ops { |
| 93 | int (*mkdir)(const char *name); |
| 94 | int (*rmdir)(const char *name); |
| 95 | } tracefs_ops __ro_after_init; |
| 96 | |
| 97 | static char *get_dname(struct dentry *dentry) |
| 98 | { |
| 99 | const char *dname; |
| 100 | char *name; |
| 101 | int len = dentry->d_name.len; |
| 102 | |
| 103 | dname = dentry->d_name.name; |
| 104 | name = kmalloc(len + 1, GFP_KERNEL); |
| 105 | if (!name) |
| 106 | return NULL; |
| 107 | memcpy(name, dname, len); |
| 108 | name[len] = 0; |
| 109 | return name; |
| 110 | } |
| 111 | |
| 112 | static struct dentry *tracefs_syscall_mkdir(struct mnt_idmap *idmap, |
| 113 | struct inode *inode, struct dentry *dentry, |
| 114 | umode_t mode) |
| 115 | { |
| 116 | struct tracefs_inode *ti; |
| 117 | char *name; |
| 118 | int ret; |
| 119 | |
| 120 | name = get_dname(dentry); |
| 121 | if (!name) |
| 122 | return ERR_PTR(error: -ENOMEM); |
| 123 | |
| 124 | /* |
| 125 | * This is a new directory that does not take the default of |
| 126 | * the rootfs. It becomes the default permissions for all the |
| 127 | * files and directories underneath it. |
| 128 | */ |
| 129 | ti = get_tracefs(inode); |
| 130 | ti->flags |= TRACEFS_INSTANCE_INODE; |
| 131 | ti->private = inode; |
| 132 | |
| 133 | /* |
| 134 | * The mkdir call can call the generic functions that create |
| 135 | * the files within the tracefs system. It is up to the individual |
| 136 | * mkdir routine to handle races. |
| 137 | */ |
| 138 | inode_unlock(inode); |
| 139 | ret = tracefs_ops.mkdir(name); |
| 140 | inode_lock(inode); |
| 141 | |
| 142 | kfree(objp: name); |
| 143 | |
| 144 | return ERR_PTR(error: ret); |
| 145 | } |
| 146 | |
| 147 | static int tracefs_syscall_rmdir(struct inode *inode, struct dentry *dentry) |
| 148 | { |
| 149 | char *name; |
| 150 | int ret; |
| 151 | |
| 152 | name = get_dname(dentry); |
| 153 | if (!name) |
| 154 | return -ENOMEM; |
| 155 | |
| 156 | /* |
| 157 | * The rmdir call can call the generic functions that create |
| 158 | * the files within the tracefs system. It is up to the individual |
| 159 | * rmdir routine to handle races. |
| 160 | * This time we need to unlock not only the parent (inode) but |
| 161 | * also the directory that is being deleted. |
| 162 | */ |
| 163 | inode_unlock(inode); |
| 164 | inode_unlock(inode: d_inode(dentry)); |
| 165 | |
| 166 | ret = tracefs_ops.rmdir(name); |
| 167 | |
| 168 | inode_lock_nested(inode, subclass: I_MUTEX_PARENT); |
| 169 | inode_lock(inode: d_inode(dentry)); |
| 170 | |
| 171 | kfree(objp: name); |
| 172 | |
| 173 | return ret; |
| 174 | } |
| 175 | |
| 176 | static void set_tracefs_inode_owner(struct inode *inode) |
| 177 | { |
| 178 | struct tracefs_inode *ti = get_tracefs(inode); |
| 179 | struct inode *root_inode = ti->private; |
| 180 | kuid_t uid; |
| 181 | kgid_t gid; |
| 182 | |
| 183 | uid = root_inode->i_uid; |
| 184 | gid = root_inode->i_gid; |
| 185 | |
| 186 | /* |
| 187 | * If the root is not the mount point, then check the root's |
| 188 | * permissions. If it was never set, then default to the |
| 189 | * mount point. |
| 190 | */ |
| 191 | if (root_inode != d_inode(dentry: root_inode->i_sb->s_root)) { |
| 192 | struct tracefs_inode *rti; |
| 193 | |
| 194 | rti = get_tracefs(inode: root_inode); |
| 195 | root_inode = d_inode(dentry: root_inode->i_sb->s_root); |
| 196 | |
| 197 | if (!(rti->flags & TRACEFS_UID_PERM_SET)) |
| 198 | uid = root_inode->i_uid; |
| 199 | |
| 200 | if (!(rti->flags & TRACEFS_GID_PERM_SET)) |
| 201 | gid = root_inode->i_gid; |
| 202 | } |
| 203 | |
| 204 | /* |
| 205 | * If this inode has never been referenced, then update |
| 206 | * the permissions to the superblock. |
| 207 | */ |
| 208 | if (!(ti->flags & TRACEFS_UID_PERM_SET)) |
| 209 | inode->i_uid = uid; |
| 210 | |
| 211 | if (!(ti->flags & TRACEFS_GID_PERM_SET)) |
| 212 | inode->i_gid = gid; |
| 213 | } |
| 214 | |
| 215 | static int tracefs_permission(struct mnt_idmap *idmap, |
| 216 | struct inode *inode, int mask) |
| 217 | { |
| 218 | set_tracefs_inode_owner(inode); |
| 219 | return generic_permission(idmap, inode, mask); |
| 220 | } |
| 221 | |
| 222 | static int tracefs_getattr(struct mnt_idmap *idmap, |
| 223 | const struct path *path, struct kstat *stat, |
| 224 | u32 request_mask, unsigned int flags) |
| 225 | { |
| 226 | struct inode *inode = d_backing_inode(upper: path->dentry); |
| 227 | |
| 228 | set_tracefs_inode_owner(inode); |
| 229 | generic_fillattr(idmap, request_mask, inode, stat); |
| 230 | return 0; |
| 231 | } |
| 232 | |
| 233 | static int tracefs_setattr(struct mnt_idmap *idmap, struct dentry *dentry, |
| 234 | struct iattr *attr) |
| 235 | { |
| 236 | unsigned int ia_valid = attr->ia_valid; |
| 237 | struct inode *inode = d_inode(dentry); |
| 238 | struct tracefs_inode *ti = get_tracefs(inode); |
| 239 | |
| 240 | if (ia_valid & ATTR_UID) |
| 241 | ti->flags |= TRACEFS_UID_PERM_SET; |
| 242 | |
| 243 | if (ia_valid & ATTR_GID) |
| 244 | ti->flags |= TRACEFS_GID_PERM_SET; |
| 245 | |
| 246 | return simple_setattr(idmap, dentry, attr); |
| 247 | } |
| 248 | |
| 249 | static const struct inode_operations tracefs_instance_dir_inode_operations = { |
| 250 | .lookup = simple_lookup, |
| 251 | .mkdir = tracefs_syscall_mkdir, |
| 252 | .rmdir = tracefs_syscall_rmdir, |
| 253 | .permission = tracefs_permission, |
| 254 | .getattr = tracefs_getattr, |
| 255 | .setattr = tracefs_setattr, |
| 256 | }; |
| 257 | |
| 258 | static const struct inode_operations tracefs_dir_inode_operations = { |
| 259 | .lookup = simple_lookup, |
| 260 | .permission = tracefs_permission, |
| 261 | .getattr = tracefs_getattr, |
| 262 | .setattr = tracefs_setattr, |
| 263 | }; |
| 264 | |
| 265 | static const struct inode_operations tracefs_file_inode_operations = { |
| 266 | .permission = tracefs_permission, |
| 267 | .getattr = tracefs_getattr, |
| 268 | .setattr = tracefs_setattr, |
| 269 | }; |
| 270 | |
| 271 | struct inode *tracefs_get_inode(struct super_block *sb) |
| 272 | { |
| 273 | struct inode *inode = new_inode(sb); |
| 274 | if (inode) { |
| 275 | inode->i_ino = get_next_ino(); |
| 276 | simple_inode_init_ts(inode); |
| 277 | } |
| 278 | return inode; |
| 279 | } |
| 280 | |
| 281 | struct tracefs_fs_info { |
| 282 | kuid_t uid; |
| 283 | kgid_t gid; |
| 284 | umode_t mode; |
| 285 | /* Opt_* bitfield. */ |
| 286 | unsigned int opts; |
| 287 | }; |
| 288 | |
| 289 | enum { |
| 290 | Opt_uid, |
| 291 | Opt_gid, |
| 292 | Opt_mode, |
| 293 | }; |
| 294 | |
| 295 | static const struct fs_parameter_spec tracefs_param_specs[] = { |
| 296 | fsparam_gid ("gid" , Opt_gid), |
| 297 | fsparam_u32oct ("mode" , Opt_mode), |
| 298 | fsparam_uid ("uid" , Opt_uid), |
| 299 | {} |
| 300 | }; |
| 301 | |
| 302 | static int tracefs_parse_param(struct fs_context *fc, struct fs_parameter *param) |
| 303 | { |
| 304 | struct tracefs_fs_info *opts = fc->s_fs_info; |
| 305 | struct fs_parse_result result; |
| 306 | int opt; |
| 307 | |
| 308 | opt = fs_parse(fc, desc: tracefs_param_specs, param, result: &result); |
| 309 | if (opt < 0) |
| 310 | return opt; |
| 311 | |
| 312 | switch (opt) { |
| 313 | case Opt_uid: |
| 314 | opts->uid = result.uid; |
| 315 | break; |
| 316 | case Opt_gid: |
| 317 | opts->gid = result.gid; |
| 318 | break; |
| 319 | case Opt_mode: |
| 320 | opts->mode = result.uint_32 & S_IALLUGO; |
| 321 | break; |
| 322 | /* |
| 323 | * We might like to report bad mount options here; |
| 324 | * but traditionally tracefs has ignored all mount options |
| 325 | */ |
| 326 | } |
| 327 | |
| 328 | opts->opts |= BIT(opt); |
| 329 | |
| 330 | return 0; |
| 331 | } |
| 332 | |
| 333 | static int tracefs_apply_options(struct super_block *sb, bool remount) |
| 334 | { |
| 335 | struct tracefs_fs_info *fsi = sb->s_fs_info; |
| 336 | struct inode *inode = d_inode(dentry: sb->s_root); |
| 337 | struct tracefs_inode *ti; |
| 338 | bool update_uid, update_gid; |
| 339 | umode_t tmp_mode; |
| 340 | |
| 341 | /* |
| 342 | * On remount, only reset mode/uid/gid if they were provided as mount |
| 343 | * options. |
| 344 | */ |
| 345 | |
| 346 | if (!remount || fsi->opts & BIT(Opt_mode)) { |
| 347 | tmp_mode = READ_ONCE(inode->i_mode) & ~S_IALLUGO; |
| 348 | tmp_mode |= fsi->mode; |
| 349 | WRITE_ONCE(inode->i_mode, tmp_mode); |
| 350 | } |
| 351 | |
| 352 | if (!remount || fsi->opts & BIT(Opt_uid)) |
| 353 | inode->i_uid = fsi->uid; |
| 354 | |
| 355 | if (!remount || fsi->opts & BIT(Opt_gid)) |
| 356 | inode->i_gid = fsi->gid; |
| 357 | |
| 358 | if (remount && (fsi->opts & BIT(Opt_uid) || fsi->opts & BIT(Opt_gid))) { |
| 359 | |
| 360 | update_uid = fsi->opts & BIT(Opt_uid); |
| 361 | update_gid = fsi->opts & BIT(Opt_gid); |
| 362 | |
| 363 | rcu_read_lock(); |
| 364 | list_for_each_entry_rcu(ti, &tracefs_inodes, list) { |
| 365 | if (update_uid) { |
| 366 | ti->flags &= ~TRACEFS_UID_PERM_SET; |
| 367 | ti->vfs_inode.i_uid = fsi->uid; |
| 368 | } |
| 369 | |
| 370 | if (update_gid) { |
| 371 | ti->flags &= ~TRACEFS_GID_PERM_SET; |
| 372 | ti->vfs_inode.i_gid = fsi->gid; |
| 373 | } |
| 374 | |
| 375 | /* |
| 376 | * Note, the above ti->vfs_inode updates are |
| 377 | * used in eventfs_remount() so they must come |
| 378 | * before calling it. |
| 379 | */ |
| 380 | if (ti->flags & TRACEFS_EVENT_INODE) |
| 381 | eventfs_remount(ti, update_uid, update_gid); |
| 382 | } |
| 383 | rcu_read_unlock(); |
| 384 | } |
| 385 | |
| 386 | return 0; |
| 387 | } |
| 388 | |
| 389 | static int tracefs_reconfigure(struct fs_context *fc) |
| 390 | { |
| 391 | struct super_block *sb = fc->root->d_sb; |
| 392 | struct tracefs_fs_info *sb_opts = sb->s_fs_info; |
| 393 | struct tracefs_fs_info *new_opts = fc->s_fs_info; |
| 394 | |
| 395 | if (!new_opts) |
| 396 | return 0; |
| 397 | |
| 398 | sync_filesystem(sb); |
| 399 | /* structure copy of new mount options to sb */ |
| 400 | *sb_opts = *new_opts; |
| 401 | |
| 402 | return tracefs_apply_options(sb, remount: true); |
| 403 | } |
| 404 | |
| 405 | static int tracefs_show_options(struct seq_file *m, struct dentry *root) |
| 406 | { |
| 407 | struct tracefs_fs_info *fsi = root->d_sb->s_fs_info; |
| 408 | |
| 409 | if (!uid_eq(left: fsi->uid, GLOBAL_ROOT_UID)) |
| 410 | seq_printf(m, ",uid=%u" , |
| 411 | from_kuid_munged(&init_user_ns, fsi->uid)); |
| 412 | if (!gid_eq(fsi->gid, GLOBAL_ROOT_GID)) |
| 413 | seq_printf(m, ",gid=%u" , |
| 414 | from_kgid_munged(&init_user_ns, fsi->gid)); |
| 415 | if (fsi->mode != TRACEFS_DEFAULT_MODE) |
| 416 | seq_printf(m, ",mode=%o" , fsi->mode); |
| 417 | |
| 418 | return 0; |
| 419 | } |
| 420 | |
| 421 | static int tracefs_drop_inode(struct inode *inode) |
| 422 | { |
| 423 | struct tracefs_inode *ti = get_tracefs(inode); |
| 424 | |
| 425 | /* |
| 426 | * This inode is being freed and cannot be used for |
| 427 | * eventfs. Clear the flag so that it doesn't call into |
| 428 | * eventfs during the remount flag updates. The eventfs_inode |
| 429 | * gets freed after an RCU cycle, so the content will still |
| 430 | * be safe if the iteration is going on now. |
| 431 | */ |
| 432 | ti->flags &= ~TRACEFS_EVENT_INODE; |
| 433 | |
| 434 | return 1; |
| 435 | } |
| 436 | |
| 437 | static const struct super_operations tracefs_super_operations = { |
| 438 | .alloc_inode = tracefs_alloc_inode, |
| 439 | .free_inode = tracefs_free_inode, |
| 440 | .destroy_inode = tracefs_destroy_inode, |
| 441 | .drop_inode = tracefs_drop_inode, |
| 442 | .statfs = simple_statfs, |
| 443 | .show_options = tracefs_show_options, |
| 444 | }; |
| 445 | |
| 446 | /* |
| 447 | * It would be cleaner if eventfs had its own dentry ops. |
| 448 | * |
| 449 | * Note that d_revalidate is called potentially under RCU, |
| 450 | * so it can't take the eventfs mutex etc. It's fine - if |
| 451 | * we open a file just as it's marked dead, things will |
| 452 | * still work just fine, and just see the old stale case. |
| 453 | */ |
| 454 | static void tracefs_d_release(struct dentry *dentry) |
| 455 | { |
| 456 | if (dentry->d_fsdata) |
| 457 | eventfs_d_release(dentry); |
| 458 | } |
| 459 | |
| 460 | static int tracefs_d_revalidate(struct inode *inode, const struct qstr *name, |
| 461 | struct dentry *dentry, unsigned int flags) |
| 462 | { |
| 463 | struct eventfs_inode *ei = dentry->d_fsdata; |
| 464 | |
| 465 | return !(ei && ei->is_freed); |
| 466 | } |
| 467 | |
| 468 | static int tracefs_d_delete(const struct dentry *dentry) |
| 469 | { |
| 470 | /* |
| 471 | * We want to keep eventfs dentries around but not tracefs |
| 472 | * ones. eventfs dentries have content in d_fsdata. |
| 473 | * Use d_fsdata to determine if it's a eventfs dentry or not. |
| 474 | */ |
| 475 | return dentry->d_fsdata == NULL; |
| 476 | } |
| 477 | |
| 478 | static const struct dentry_operations tracefs_dentry_operations = { |
| 479 | .d_revalidate = tracefs_d_revalidate, |
| 480 | .d_release = tracefs_d_release, |
| 481 | .d_delete = tracefs_d_delete, |
| 482 | }; |
| 483 | |
| 484 | static int tracefs_fill_super(struct super_block *sb, struct fs_context *fc) |
| 485 | { |
| 486 | static const struct tree_descr trace_files[] = {{"" }}; |
| 487 | int err; |
| 488 | |
| 489 | err = simple_fill_super(sb, TRACEFS_MAGIC, trace_files); |
| 490 | if (err) |
| 491 | return err; |
| 492 | |
| 493 | sb->s_op = &tracefs_super_operations; |
| 494 | set_default_d_op(sb, &tracefs_dentry_operations); |
| 495 | |
| 496 | return 0; |
| 497 | } |
| 498 | |
| 499 | static int tracefs_get_tree(struct fs_context *fc) |
| 500 | { |
| 501 | int err = get_tree_single(fc, fill_super: tracefs_fill_super); |
| 502 | |
| 503 | if (err) |
| 504 | return err; |
| 505 | |
| 506 | return tracefs_reconfigure(fc); |
| 507 | } |
| 508 | |
| 509 | static void tracefs_free_fc(struct fs_context *fc) |
| 510 | { |
| 511 | kfree(objp: fc->s_fs_info); |
| 512 | } |
| 513 | |
| 514 | static const struct fs_context_operations tracefs_context_ops = { |
| 515 | .free = tracefs_free_fc, |
| 516 | .parse_param = tracefs_parse_param, |
| 517 | .get_tree = tracefs_get_tree, |
| 518 | .reconfigure = tracefs_reconfigure, |
| 519 | }; |
| 520 | |
| 521 | static int tracefs_init_fs_context(struct fs_context *fc) |
| 522 | { |
| 523 | struct tracefs_fs_info *fsi; |
| 524 | |
| 525 | fsi = kzalloc(sizeof(struct tracefs_fs_info), GFP_KERNEL); |
| 526 | if (!fsi) |
| 527 | return -ENOMEM; |
| 528 | |
| 529 | fsi->mode = TRACEFS_DEFAULT_MODE; |
| 530 | |
| 531 | fc->s_fs_info = fsi; |
| 532 | fc->ops = &tracefs_context_ops; |
| 533 | return 0; |
| 534 | } |
| 535 | |
| 536 | static struct file_system_type trace_fs_type = { |
| 537 | .owner = THIS_MODULE, |
| 538 | .name = "tracefs" , |
| 539 | .init_fs_context = tracefs_init_fs_context, |
| 540 | .parameters = tracefs_param_specs, |
| 541 | .kill_sb = kill_anon_super, |
| 542 | }; |
| 543 | MODULE_ALIAS_FS("tracefs" ); |
| 544 | |
| 545 | struct dentry *tracefs_start_creating(const char *name, struct dentry *parent) |
| 546 | { |
| 547 | struct dentry *dentry; |
| 548 | int error; |
| 549 | |
| 550 | pr_debug("tracefs: creating file '%s'\n" ,name); |
| 551 | |
| 552 | error = simple_pin_fs(&trace_fs_type, mount: &tracefs_mount, |
| 553 | count: &tracefs_mount_count); |
| 554 | if (error) |
| 555 | return ERR_PTR(error); |
| 556 | |
| 557 | /* If the parent is not specified, we create it in the root. |
| 558 | * We need the root dentry to do this, which is in the super |
| 559 | * block. A pointer to that is in the struct vfsmount that we |
| 560 | * have around. |
| 561 | */ |
| 562 | if (!parent) |
| 563 | parent = tracefs_mount->mnt_root; |
| 564 | |
| 565 | dentry = simple_start_creating(parent, name); |
| 566 | if (IS_ERR(ptr: dentry)) |
| 567 | simple_release_fs(mount: &tracefs_mount, count: &tracefs_mount_count); |
| 568 | |
| 569 | return dentry; |
| 570 | } |
| 571 | |
| 572 | struct dentry *tracefs_failed_creating(struct dentry *dentry) |
| 573 | { |
| 574 | simple_done_creating(dentry); |
| 575 | simple_release_fs(mount: &tracefs_mount, count: &tracefs_mount_count); |
| 576 | return NULL; |
| 577 | } |
| 578 | |
| 579 | struct dentry *tracefs_end_creating(struct dentry *dentry) |
| 580 | { |
| 581 | simple_done_creating(dentry); |
| 582 | return dentry; // borrowed |
| 583 | } |
| 584 | |
| 585 | /* Find the inode that this will use for default */ |
| 586 | static struct inode *instance_inode(struct dentry *parent, struct inode *inode) |
| 587 | { |
| 588 | struct tracefs_inode *ti; |
| 589 | |
| 590 | /* If parent is NULL then use root inode */ |
| 591 | if (!parent) |
| 592 | return d_inode(dentry: inode->i_sb->s_root); |
| 593 | |
| 594 | /* Find the inode that is flagged as an instance or the root inode */ |
| 595 | while (!IS_ROOT(parent)) { |
| 596 | ti = get_tracefs(inode: d_inode(dentry: parent)); |
| 597 | if (ti->flags & TRACEFS_INSTANCE_INODE) |
| 598 | break; |
| 599 | parent = parent->d_parent; |
| 600 | } |
| 601 | |
| 602 | return d_inode(dentry: parent); |
| 603 | } |
| 604 | |
| 605 | /** |
| 606 | * tracefs_create_file - create a file in the tracefs filesystem |
| 607 | * @name: a pointer to a string containing the name of the file to create. |
| 608 | * @mode: the permission that the file should have. |
| 609 | * @parent: a pointer to the parent dentry for this file. This should be a |
| 610 | * directory dentry if set. If this parameter is NULL, then the |
| 611 | * file will be created in the root of the tracefs filesystem. |
| 612 | * @data: a pointer to something that the caller will want to get to later |
| 613 | * on. The inode.i_private pointer will point to this value on |
| 614 | * the open() call. |
| 615 | * @fops: a pointer to a struct file_operations that should be used for |
| 616 | * this file. |
| 617 | * |
| 618 | * This is the basic "create a file" function for tracefs. It allows for a |
| 619 | * wide range of flexibility in creating a file, or a directory (if you want |
| 620 | * to create a directory, the tracefs_create_dir() function is |
| 621 | * recommended to be used instead.) |
| 622 | * |
| 623 | * This function will return a pointer to a dentry if it succeeds. This |
| 624 | * pointer must be passed to the tracefs_remove() function when the file is |
| 625 | * to be removed (no automatic cleanup happens if your module is unloaded, |
| 626 | * you are responsible here.) If an error occurs, %NULL will be returned. |
| 627 | * |
| 628 | * If tracefs is not enabled in the kernel, the value -%ENODEV will be |
| 629 | * returned. |
| 630 | */ |
| 631 | struct dentry *tracefs_create_file(const char *name, umode_t mode, |
| 632 | struct dentry *parent, void *data, |
| 633 | const struct file_operations *fops) |
| 634 | { |
| 635 | struct tracefs_inode *ti; |
| 636 | struct dentry *dentry; |
| 637 | struct inode *inode; |
| 638 | |
| 639 | if (security_locked_down(what: LOCKDOWN_TRACEFS)) |
| 640 | return NULL; |
| 641 | |
| 642 | if (!(mode & S_IFMT)) |
| 643 | mode |= S_IFREG; |
| 644 | BUG_ON(!S_ISREG(mode)); |
| 645 | dentry = tracefs_start_creating(name, parent); |
| 646 | |
| 647 | if (IS_ERR(ptr: dentry)) |
| 648 | return NULL; |
| 649 | |
| 650 | inode = tracefs_get_inode(sb: dentry->d_sb); |
| 651 | if (unlikely(!inode)) |
| 652 | return tracefs_failed_creating(dentry); |
| 653 | |
| 654 | ti = get_tracefs(inode); |
| 655 | ti->private = instance_inode(parent, inode); |
| 656 | |
| 657 | inode->i_mode = mode; |
| 658 | inode->i_op = &tracefs_file_inode_operations; |
| 659 | inode->i_fop = fops ? fops : &tracefs_file_operations; |
| 660 | inode->i_private = data; |
| 661 | inode->i_uid = d_inode(dentry: dentry->d_parent)->i_uid; |
| 662 | inode->i_gid = d_inode(dentry: dentry->d_parent)->i_gid; |
| 663 | d_make_persistent(dentry, inode); |
| 664 | fsnotify_create(dir: d_inode(dentry: dentry->d_parent), dentry); |
| 665 | return tracefs_end_creating(dentry); |
| 666 | } |
| 667 | |
| 668 | static struct dentry *__create_dir(const char *name, struct dentry *parent, |
| 669 | const struct inode_operations *ops) |
| 670 | { |
| 671 | struct tracefs_inode *ti; |
| 672 | struct dentry *dentry = tracefs_start_creating(name, parent); |
| 673 | struct inode *inode; |
| 674 | |
| 675 | if (IS_ERR(ptr: dentry)) |
| 676 | return NULL; |
| 677 | |
| 678 | inode = tracefs_get_inode(sb: dentry->d_sb); |
| 679 | if (unlikely(!inode)) |
| 680 | return tracefs_failed_creating(dentry); |
| 681 | |
| 682 | /* Do not set bits for OTH */ |
| 683 | inode->i_mode = S_IFDIR | S_IRWXU | S_IRUSR| S_IRGRP | S_IXUSR | S_IXGRP; |
| 684 | inode->i_op = ops; |
| 685 | inode->i_fop = &simple_dir_operations; |
| 686 | inode->i_uid = d_inode(dentry: dentry->d_parent)->i_uid; |
| 687 | inode->i_gid = d_inode(dentry: dentry->d_parent)->i_gid; |
| 688 | |
| 689 | ti = get_tracefs(inode); |
| 690 | ti->private = instance_inode(parent, inode); |
| 691 | |
| 692 | /* directory inodes start off with i_nlink == 2 (for "." entry) */ |
| 693 | inc_nlink(inode); |
| 694 | d_make_persistent(dentry, inode); |
| 695 | inc_nlink(inode: d_inode(dentry: dentry->d_parent)); |
| 696 | fsnotify_mkdir(dir: d_inode(dentry: dentry->d_parent), dentry); |
| 697 | return tracefs_end_creating(dentry); |
| 698 | } |
| 699 | |
| 700 | /** |
| 701 | * tracefs_create_dir - create a directory in the tracefs filesystem |
| 702 | * @name: a pointer to a string containing the name of the directory to |
| 703 | * create. |
| 704 | * @parent: a pointer to the parent dentry for this file. This should be a |
| 705 | * directory dentry if set. If this parameter is NULL, then the |
| 706 | * directory will be created in the root of the tracefs filesystem. |
| 707 | * |
| 708 | * This function creates a directory in tracefs with the given name. |
| 709 | * |
| 710 | * This function will return a pointer to a dentry if it succeeds. This |
| 711 | * pointer must be passed to the tracefs_remove() function when the file is |
| 712 | * to be removed. If an error occurs, %NULL will be returned. |
| 713 | * |
| 714 | * If tracing is not enabled in the kernel, the value -%ENODEV will be |
| 715 | * returned. |
| 716 | */ |
| 717 | struct dentry *tracefs_create_dir(const char *name, struct dentry *parent) |
| 718 | { |
| 719 | if (security_locked_down(what: LOCKDOWN_TRACEFS)) |
| 720 | return NULL; |
| 721 | |
| 722 | return __create_dir(name, parent, ops: &tracefs_dir_inode_operations); |
| 723 | } |
| 724 | |
| 725 | /** |
| 726 | * tracefs_create_instance_dir - create the tracing instances directory |
| 727 | * @name: The name of the instances directory to create |
| 728 | * @parent: The parent directory that the instances directory will exist |
| 729 | * @mkdir: The function to call when a mkdir is performed. |
| 730 | * @rmdir: The function to call when a rmdir is performed. |
| 731 | * |
| 732 | * Only one instances directory is allowed. |
| 733 | * |
| 734 | * The instances directory is special as it allows for mkdir and rmdir |
| 735 | * to be done by userspace. When a mkdir or rmdir is performed, the inode |
| 736 | * locks are released and the methods passed in (@mkdir and @rmdir) are |
| 737 | * called without locks and with the name of the directory being created |
| 738 | * within the instances directory. |
| 739 | * |
| 740 | * Returns the dentry of the instances directory. |
| 741 | */ |
| 742 | __init struct dentry *tracefs_create_instance_dir(const char *name, |
| 743 | struct dentry *parent, |
| 744 | int (*mkdir)(const char *name), |
| 745 | int (*rmdir)(const char *name)) |
| 746 | { |
| 747 | struct dentry *dentry; |
| 748 | |
| 749 | /* Only allow one instance of the instances directory. */ |
| 750 | if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir)) |
| 751 | return NULL; |
| 752 | |
| 753 | dentry = __create_dir(name, parent, ops: &tracefs_instance_dir_inode_operations); |
| 754 | if (!dentry) |
| 755 | return NULL; |
| 756 | |
| 757 | tracefs_ops.mkdir = mkdir; |
| 758 | tracefs_ops.rmdir = rmdir; |
| 759 | |
| 760 | return dentry; |
| 761 | } |
| 762 | |
| 763 | static void remove_one(struct dentry *victim) |
| 764 | { |
| 765 | simple_release_fs(mount: &tracefs_mount, count: &tracefs_mount_count); |
| 766 | } |
| 767 | |
| 768 | /** |
| 769 | * tracefs_remove - recursively removes a directory |
| 770 | * @dentry: a pointer to a the dentry of the directory to be removed. |
| 771 | * |
| 772 | * This function recursively removes a directory tree in tracefs that |
| 773 | * was previously created with a call to another tracefs function |
| 774 | * (like tracefs_create_file() or variants thereof.) |
| 775 | */ |
| 776 | void tracefs_remove(struct dentry *dentry) |
| 777 | { |
| 778 | if (IS_ERR_OR_NULL(ptr: dentry)) |
| 779 | return; |
| 780 | |
| 781 | simple_pin_fs(&trace_fs_type, mount: &tracefs_mount, count: &tracefs_mount_count); |
| 782 | simple_recursive_removal(dentry, callback: remove_one); |
| 783 | simple_release_fs(mount: &tracefs_mount, count: &tracefs_mount_count); |
| 784 | } |
| 785 | |
| 786 | /** |
| 787 | * tracefs_initialized - Tells whether tracefs has been registered |
| 788 | */ |
| 789 | bool tracefs_initialized(void) |
| 790 | { |
| 791 | return tracefs_registered; |
| 792 | } |
| 793 | |
| 794 | static void init_once(void *foo) |
| 795 | { |
| 796 | struct tracefs_inode *ti = (struct tracefs_inode *) foo; |
| 797 | |
| 798 | /* inode_init_once() calls memset() on the vfs_inode portion */ |
| 799 | inode_init_once(&ti->vfs_inode); |
| 800 | |
| 801 | /* Zero out the rest */ |
| 802 | memset_after(ti, 0, vfs_inode); |
| 803 | } |
| 804 | |
| 805 | static int __init tracefs_init(void) |
| 806 | { |
| 807 | int retval; |
| 808 | |
| 809 | tracefs_inode_cachep = kmem_cache_create("tracefs_inode_cache" , |
| 810 | sizeof(struct tracefs_inode), |
| 811 | 0, (SLAB_RECLAIM_ACCOUNT| |
| 812 | SLAB_ACCOUNT), |
| 813 | init_once); |
| 814 | if (!tracefs_inode_cachep) |
| 815 | return -ENOMEM; |
| 816 | |
| 817 | retval = sysfs_create_mount_point(parent_kobj: kernel_kobj, name: "tracing" ); |
| 818 | if (retval) |
| 819 | return -EINVAL; |
| 820 | |
| 821 | retval = register_filesystem(&trace_fs_type); |
| 822 | if (!retval) |
| 823 | tracefs_registered = true; |
| 824 | |
| 825 | return retval; |
| 826 | } |
| 827 | core_initcall(tracefs_init); |
| 828 | |