blob: b635ee5adbccede861d2fa88d7a75da496baea2a [file] [log] [blame]
Thomas Gleixner457c8992019-05-19 13:08:55 +01001// SPDX-License-Identifier: GPL-2.0-only
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * fs/libfs.c
4 * Library for filesystems writers.
5 */
6
Fabian Frederickac13a822014-06-04 16:06:27 -07007#include <linux/blkdev.h>
Paul Gortmaker630d9c42011-11-16 23:57:37 -05008#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/pagemap.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090010#include <linux/slab.h>
Ingo Molnar5b825c32017-02-02 17:54:15 +010011#include <linux/cred.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/mount.h>
13#include <linux/vfs.h>
npiggin@suse.de7bb46a62010-05-27 01:05:33 +100014#include <linux/quotaops.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080015#include <linux/mutex.h>
Al Viro87dc8002013-09-16 10:30:04 -040016#include <linux/namei.h>
Christoph Hellwig25961102007-10-21 16:42:05 -070017#include <linux/exportfs.h>
Andrew Morton5ca14832022-09-09 13:57:41 -070018#include <linux/iversion.h>
Al Virod5aacad2009-06-07 14:56:44 -040019#include <linux/writeback.h>
Al Viroff01bb42011-09-16 02:31:11 -040020#include <linux/buffer_head.h> /* sync_mapping_buffers */
David Howells31d6d5c2019-03-25 16:38:23 +000021#include <linux/fs_context.h>
22#include <linux/pseudo_fs.h>
Al Viroa3d1e7e2019-11-18 09:43:10 -050023#include <linux/fsnotify.h>
Daniel Rosenbergc8438432020-07-08 02:12:35 -070024#include <linux/unicode.h>
25#include <linux/fscrypt.h>
Christian Braunerb28ddcc2024-02-19 16:30:57 +010026#include <linux/pidfs.h>
Ingo Molnar7cf34c72006-03-23 03:00:36 -080027
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080028#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029
Al Viroa4464db2011-07-07 15:03:58 -040030#include "internal.h"
31
Christian Braunerb74d24f2023-01-13 12:49:12 +010032int simple_getattr(struct mnt_idmap *idmap, const struct path *path,
Christian Brauner549c7292021-01-21 14:19:43 +010033 struct kstat *stat, u32 request_mask,
34 unsigned int query_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070035{
David Howellsa528d352017-01-31 16:46:22 +000036 struct inode *inode = d_inode(path->dentry);
Jeff Layton0d72b9282023-08-07 15:38:33 -040037 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +030038 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 return 0;
40}
Al Viro12f38872013-09-15 21:20:49 -040041EXPORT_SYMBOL(simple_getattr);
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
David Howells726c3342006-06-23 02:02:58 -070043int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070044{
Amir Goldsteinae62bcb2023-10-23 17:30:49 +030045 u64 id = huge_encode_dev(dentry->d_sb->s_dev);
46
47 buf->f_fsid = u64_to_fsid(id);
David Howells726c3342006-06-23 02:02:58 -070048 buf->f_type = dentry->d_sb->s_magic;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +030049 buf->f_bsize = PAGE_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -070050 buf->f_namelen = NAME_MAX;
51 return 0;
52}
Al Viro12f38872013-09-15 21:20:49 -040053EXPORT_SYMBOL(simple_statfs);
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
55/*
56 * Retaining negative dentries for an in-memory filesystem just wastes
57 * memory and lookup time: arrange for them to be deleted immediately.
58 */
Al Virob26d4cd2013-10-25 18:47:37 -040059int always_delete_dentry(const struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -070060{
61 return 1;
62}
Al Virob26d4cd2013-10-25 18:47:37 -040063EXPORT_SYMBOL(always_delete_dentry);
64
65const struct dentry_operations simple_dentry_operations = {
66 .d_delete = always_delete_dentry,
67};
68EXPORT_SYMBOL(simple_dentry_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
70/*
71 * Lookup the data. This is trivial - if the dentry didn't already
72 * exist, we know it is negative. Set d_op to delete negative dentries.
73 */
Al Viro00cd8dd2012-06-10 17:13:09 -040074struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070075{
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 if (dentry->d_name.len > NAME_MAX)
77 return ERR_PTR(-ENAMETOOLONG);
Al Viro74931da2013-07-14 17:43:25 +040078 if (!dentry->d_sb->s_d_op)
79 d_set_d_op(dentry, &simple_dentry_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 d_add(dentry, NULL);
81 return NULL;
82}
Al Viro12f38872013-09-15 21:20:49 -040083EXPORT_SYMBOL(simple_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Linus Torvalds1da177e2005-04-16 15:20:36 -070085int dcache_dir_open(struct inode *inode, struct file *file)
86{
Al Viroba65dc52016-06-10 11:32:47 -040087 file->private_data = d_alloc_cursor(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
89 return file->private_data ? 0 : -ENOMEM;
90}
Al Viro12f38872013-09-15 21:20:49 -040091EXPORT_SYMBOL(dcache_dir_open);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93int dcache_dir_close(struct inode *inode, struct file *file)
94{
95 dput(file->private_data);
96 return 0;
97}
Al Viro12f38872013-09-15 21:20:49 -040098EXPORT_SYMBOL(dcache_dir_close);
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Al Viro4f42c1b2016-06-06 19:37:13 -0400100/* parent is locked at least shared */
Al Virod4f4de52019-09-15 12:12:39 -0400101/*
102 * Returns an element of siblings' list.
103 * We are looking for <count>th positive after <p>; if
Al Viro26b6c982019-09-20 16:32:42 -0400104 * found, dentry is grabbed and returned to caller.
105 * If no such element exists, NULL is returned.
Al Virod4f4de52019-09-15 12:12:39 -0400106 */
Al Viro26b6c982019-09-20 16:32:42 -0400107static struct dentry *scan_positives(struct dentry *cursor,
Al Viroda549bd2023-11-07 02:00:39 -0500108 struct hlist_node **p,
Al Virod4f4de52019-09-15 12:12:39 -0400109 loff_t count,
Al Viro26b6c982019-09-20 16:32:42 -0400110 struct dentry *last)
Al Viro4f42c1b2016-06-06 19:37:13 -0400111{
Al Virod4f4de52019-09-15 12:12:39 -0400112 struct dentry *dentry = cursor->d_parent, *found = NULL;
Al Viro4f42c1b2016-06-06 19:37:13 -0400113
Al Virod4f4de52019-09-15 12:12:39 -0400114 spin_lock(&dentry->d_lock);
Al Viroda549bd2023-11-07 02:00:39 -0500115 while (*p) {
116 struct dentry *d = hlist_entry(*p, struct dentry, d_sib);
117 p = &d->d_sib.next;
Al Virod4f4de52019-09-15 12:12:39 -0400118 // we must at least skip cursors, to avoid livelocks
119 if (d->d_flags & DCACHE_DENTRY_CURSOR)
120 continue;
121 if (simple_positive(d) && !--count) {
122 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
123 if (simple_positive(d))
124 found = dget_dlock(d);
125 spin_unlock(&d->d_lock);
126 if (likely(found))
127 break;
128 count = 1;
129 }
130 if (need_resched()) {
Al Viroda549bd2023-11-07 02:00:39 -0500131 if (!hlist_unhashed(&cursor->d_sib))
132 __hlist_del(&cursor->d_sib);
133 hlist_add_behind(&cursor->d_sib, &d->d_sib);
134 p = &cursor->d_sib.next;
Al Virod4f4de52019-09-15 12:12:39 -0400135 spin_unlock(&dentry->d_lock);
136 cond_resched();
137 spin_lock(&dentry->d_lock);
Al Viro4f42c1b2016-06-06 19:37:13 -0400138 }
139 }
Al Virod4f4de52019-09-15 12:12:39 -0400140 spin_unlock(&dentry->d_lock);
Al Viro26b6c982019-09-20 16:32:42 -0400141 dput(last);
142 return found;
Al Viro4f42c1b2016-06-06 19:37:13 -0400143}
144
Andrew Morton965c8e52012-12-17 15:59:39 -0800145loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146{
Nick Piggin2fd6b7f2011-01-07 17:49:34 +1100147 struct dentry *dentry = file->f_path.dentry;
Andrew Morton965c8e52012-12-17 15:59:39 -0800148 switch (whence) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 case 1:
150 offset += file->f_pos;
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -0500151 fallthrough;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 case 0:
153 if (offset >= 0)
154 break;
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -0500155 fallthrough;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 default:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 return -EINVAL;
158 }
159 if (offset != file->f_pos) {
Al Virod4f4de52019-09-15 12:12:39 -0400160 struct dentry *cursor = file->private_data;
161 struct dentry *to = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162
Al Virod4f4de52019-09-15 12:12:39 -0400163 inode_lock_shared(dentry->d_inode);
164
Al Viro26b6c982019-09-20 16:32:42 -0400165 if (offset > 2)
Al Viroda549bd2023-11-07 02:00:39 -0500166 to = scan_positives(cursor, &dentry->d_children.first,
Al Viro26b6c982019-09-20 16:32:42 -0400167 offset - 2, NULL);
168 spin_lock(&dentry->d_lock);
Al Viroda549bd2023-11-07 02:00:39 -0500169 hlist_del_init(&cursor->d_sib);
Al Viro26b6c982019-09-20 16:32:42 -0400170 if (to)
Al Viroda549bd2023-11-07 02:00:39 -0500171 hlist_add_behind(&cursor->d_sib, &to->d_sib);
Al Viro26b6c982019-09-20 16:32:42 -0400172 spin_unlock(&dentry->d_lock);
Al Virod4f4de52019-09-15 12:12:39 -0400173 dput(to);
174
Al Viro26b6c982019-09-20 16:32:42 -0400175 file->f_pos = offset;
176
Al Virod4f4de52019-09-15 12:12:39 -0400177 inode_unlock_shared(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 return offset;
180}
Al Viro12f38872013-09-15 21:20:49 -0400181EXPORT_SYMBOL(dcache_dir_lseek);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183/*
184 * Directory is locked and all positive dentries in it are safe, since
185 * for ramfs-type trees they can't go away without unlink() or rmdir(),
186 * both impossible due to the lock on directory.
187 */
188
Al Viro5f99f4e2013-05-15 20:23:06 -0400189int dcache_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
Al Viro5f99f4e2013-05-15 20:23:06 -0400191 struct dentry *dentry = file->f_path.dentry;
192 struct dentry *cursor = file->private_data;
Al Virod4f4de52019-09-15 12:12:39 -0400193 struct dentry *next = NULL;
Al Viroda549bd2023-11-07 02:00:39 -0500194 struct hlist_node **p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
Al Viro5f99f4e2013-05-15 20:23:06 -0400196 if (!dir_emit_dots(file, ctx))
197 return 0;
Al Viro4f42c1b2016-06-06 19:37:13 -0400198
Al Viro5f99f4e2013-05-15 20:23:06 -0400199 if (ctx->pos == 2)
Al Viroda549bd2023-11-07 02:00:39 -0500200 p = &dentry->d_children.first;
Al Viro26b6c982019-09-20 16:32:42 -0400201 else
Al Viroda549bd2023-11-07 02:00:39 -0500202 p = &cursor->d_sib.next;
Al Virod4f4de52019-09-15 12:12:39 -0400203
Al Viro26b6c982019-09-20 16:32:42 -0400204 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
Al Viro5f99f4e2013-05-15 20:23:06 -0400205 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
Jeff Layton364595a2023-03-30 06:41:43 -0400206 d_inode(next)->i_ino,
207 fs_umode_to_dtype(d_inode(next)->i_mode)))
Al Viro4f42c1b2016-06-06 19:37:13 -0400208 break;
Al Viro5f99f4e2013-05-15 20:23:06 -0400209 ctx->pos++;
Al Viroda549bd2023-11-07 02:00:39 -0500210 p = &next->d_sib.next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 }
Al Virod4f4de52019-09-15 12:12:39 -0400212 spin_lock(&dentry->d_lock);
Al Viroda549bd2023-11-07 02:00:39 -0500213 hlist_del_init(&cursor->d_sib);
Al Viro26b6c982019-09-20 16:32:42 -0400214 if (next)
Al Viroda549bd2023-11-07 02:00:39 -0500215 hlist_add_before(&cursor->d_sib, &next->d_sib);
Al Virod4f4de52019-09-15 12:12:39 -0400216 spin_unlock(&dentry->d_lock);
217 dput(next);
218
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 return 0;
220}
Al Viro12f38872013-09-15 21:20:49 -0400221EXPORT_SYMBOL(dcache_readdir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
223ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
224{
225 return -EISDIR;
226}
Al Viro12f38872013-09-15 21:20:49 -0400227EXPORT_SYMBOL(generic_read_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800229const struct file_operations simple_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 .open = dcache_dir_open,
231 .release = dcache_dir_close,
232 .llseek = dcache_dir_lseek,
233 .read = generic_read_dir,
Al Viro4e829012016-04-20 19:52:15 -0400234 .iterate_shared = dcache_readdir,
Christoph Hellwig1b061d92010-05-26 17:53:41 +0200235 .fsync = noop_fsync,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236};
Al Viro12f38872013-09-15 21:20:49 -0400237EXPORT_SYMBOL(simple_dir_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -0800239const struct inode_operations simple_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 .lookup = simple_lookup,
241};
Al Viro12f38872013-09-15 21:20:49 -0400242EXPORT_SYMBOL(simple_dir_inode_operations);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
Chuck Lever7beea722024-02-17 15:23:47 -0500244/* 0 is '.', 1 is '..', so always start with offset 2 or more */
245enum {
246 DIR_OFFSET_MIN = 2,
247};
248
Chuck Lever0e4a8622024-02-17 15:24:16 -0500249static void offset_set(struct dentry *dentry, long offset)
Chuck Lever6faddda692023-06-30 13:48:49 -0400250{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500251 dentry->d_fsdata = (void *)offset;
Chuck Lever6faddda692023-06-30 13:48:49 -0400252}
253
Chuck Lever0e4a8622024-02-17 15:24:16 -0500254static long dentry2offset(struct dentry *dentry)
Chuck Lever6faddda692023-06-30 13:48:49 -0400255{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500256 return (long)dentry->d_fsdata;
Chuck Lever6faddda692023-06-30 13:48:49 -0400257}
258
Chuck Lever0e4a8622024-02-17 15:24:16 -0500259static struct lock_class_key simple_offset_lock_class;
Chuck Leverbbaef792023-07-24 10:43:57 -0400260
Chuck Lever6faddda692023-06-30 13:48:49 -0400261/**
262 * simple_offset_init - initialize an offset_ctx
263 * @octx: directory offset map to be initialized
264 *
265 */
266void simple_offset_init(struct offset_ctx *octx)
267{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500268 mt_init_flags(&octx->mt, MT_FLAGS_ALLOC_RANGE);
269 lockdep_set_class(&octx->mt.ma_lock, &simple_offset_lock_class);
Chuck Lever7beea722024-02-17 15:23:47 -0500270 octx->next_offset = DIR_OFFSET_MIN;
Chuck Lever6faddda692023-06-30 13:48:49 -0400271}
272
273/**
274 * simple_offset_add - Add an entry to a directory's offset map
275 * @octx: directory offset ctx to be updated
276 * @dentry: new dentry being added
277 *
Chuck Lever0e4a8622024-02-17 15:24:16 -0500278 * Returns zero on success. @octx and the dentry's offset are updated.
Chuck Lever6faddda692023-06-30 13:48:49 -0400279 * Otherwise, a negative errno value is returned.
280 */
281int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry)
282{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500283 unsigned long offset;
Chuck Lever6faddda692023-06-30 13:48:49 -0400284 int ret;
285
286 if (dentry2offset(dentry) != 0)
287 return -EBUSY;
288
Chuck Lever0e4a8622024-02-17 15:24:16 -0500289 ret = mtree_alloc_cyclic(&octx->mt, &offset, dentry, DIR_OFFSET_MIN,
290 LONG_MAX, &octx->next_offset, GFP_KERNEL);
Chuck Lever6faddda692023-06-30 13:48:49 -0400291 if (ret < 0)
292 return ret;
293
294 offset_set(dentry, offset);
295 return 0;
296}
297
Chuck Lever23cdd0e2024-04-15 11:20:54 -0400298static int simple_offset_replace(struct offset_ctx *octx, struct dentry *dentry,
299 long offset)
300{
301 int ret;
302
303 ret = mtree_store(&octx->mt, offset, dentry, GFP_KERNEL);
304 if (ret)
305 return ret;
306 offset_set(dentry, offset);
307 return 0;
308}
309
Chuck Lever6faddda692023-06-30 13:48:49 -0400310/**
311 * simple_offset_remove - Remove an entry to a directory's offset map
312 * @octx: directory offset ctx to be updated
313 * @dentry: dentry being removed
314 *
315 */
316void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry)
317{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500318 long offset;
Chuck Lever6faddda692023-06-30 13:48:49 -0400319
320 offset = dentry2offset(dentry);
321 if (offset == 0)
322 return;
323
Chuck Lever0e4a8622024-02-17 15:24:16 -0500324 mtree_erase(&octx->mt, offset);
Chuck Lever6faddda692023-06-30 13:48:49 -0400325 offset_set(dentry, 0);
326}
327
328/**
Chuck Leverecba88a2024-02-17 15:23:54 -0500329 * simple_offset_empty - Check if a dentry can be unlinked
330 * @dentry: dentry to be tested
331 *
332 * Returns 0 if @dentry is a non-empty directory; otherwise returns 1.
333 */
334int 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(inode);
Chuck Lever0e4a8622024-02-17 15:24:16 -0500347 mt_for_each(&octx->mt, child, index, LONG_MAX) {
Chuck Leverecba88a2024-02-17 15:23:54 -0500348 spin_lock(&child->d_lock);
349 if (simple_positive(child)) {
350 spin_unlock(&child->d_lock);
351 ret = 0;
352 break;
353 }
354 spin_unlock(&child->d_lock);
355 }
356
357 return ret;
358}
359
360/**
Chuck Lever5a1a25b2024-04-15 11:20:55 -0400361 * simple_offset_rename - handle directory offsets for rename
362 * @old_dir: parent directory of source entry
363 * @old_dentry: dentry of source entry
364 * @new_dir: parent_directory of destination entry
365 * @new_dentry: dentry of destination
366 *
367 * Caller provides appropriate serialization.
368 *
Chuck Leverad191eb2024-04-15 11:20:56 -0400369 * User space expects the directory offset value of the replaced
370 * (new) directory entry to be unchanged after a rename.
371 *
Chuck Lever5a1a25b2024-04-15 11:20:55 -0400372 * Returns zero on success, a negative errno value on failure.
373 */
374int simple_offset_rename(struct inode *old_dir, struct dentry *old_dentry,
375 struct inode *new_dir, struct dentry *new_dentry)
376{
377 struct offset_ctx *old_ctx = old_dir->i_op->get_offset_ctx(old_dir);
378 struct offset_ctx *new_ctx = new_dir->i_op->get_offset_ctx(new_dir);
Chuck Leverad191eb2024-04-15 11:20:56 -0400379 long new_offset = dentry2offset(new_dentry);
Chuck Lever5a1a25b2024-04-15 11:20:55 -0400380
381 simple_offset_remove(old_ctx, old_dentry);
Chuck Leverad191eb2024-04-15 11:20:56 -0400382
383 if (new_offset) {
384 offset_set(new_dentry, 0);
385 return simple_offset_replace(new_ctx, old_dentry, new_offset);
386 }
Chuck Lever5a1a25b2024-04-15 11:20:55 -0400387 return simple_offset_add(new_ctx, old_dentry);
388}
389
390/**
Chuck Lever6faddda692023-06-30 13:48:49 -0400391 * simple_offset_rename_exchange - exchange rename with directory offsets
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 *
Chuck Lever23cdd0e2024-04-15 11:20:54 -0400397 * This API preserves the directory offset values. Caller provides
398 * appropriate serialization.
399 *
Chuck Lever6faddda692023-06-30 13:48:49 -0400400 * Returns zero on success. Otherwise a negative errno is returned and the
401 * rename is rolled back.
402 */
403int simple_offset_rename_exchange(struct inode *old_dir,
404 struct dentry *old_dentry,
405 struct inode *new_dir,
406 struct dentry *new_dentry)
407{
408 struct offset_ctx *old_ctx = old_dir->i_op->get_offset_ctx(old_dir);
409 struct offset_ctx *new_ctx = new_dir->i_op->get_offset_ctx(new_dir);
Chuck Lever0e4a8622024-02-17 15:24:16 -0500410 long old_index = dentry2offset(old_dentry);
411 long new_index = dentry2offset(new_dentry);
Chuck Lever6faddda692023-06-30 13:48:49 -0400412 int ret;
413
414 simple_offset_remove(old_ctx, old_dentry);
415 simple_offset_remove(new_ctx, new_dentry);
416
Chuck Lever23cdd0e2024-04-15 11:20:54 -0400417 ret = simple_offset_replace(new_ctx, old_dentry, new_index);
Chuck Lever6faddda692023-06-30 13:48:49 -0400418 if (ret)
419 goto out_restore;
420
Chuck Lever23cdd0e2024-04-15 11:20:54 -0400421 ret = simple_offset_replace(old_ctx, new_dentry, old_index);
Chuck Lever6faddda692023-06-30 13:48:49 -0400422 if (ret) {
423 simple_offset_remove(new_ctx, old_dentry);
424 goto out_restore;
425 }
426
427 ret = simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
428 if (ret) {
429 simple_offset_remove(new_ctx, old_dentry);
430 simple_offset_remove(old_ctx, new_dentry);
431 goto out_restore;
432 }
433 return 0;
434
435out_restore:
Chuck Lever23cdd0e2024-04-15 11:20:54 -0400436 (void)simple_offset_replace(old_ctx, old_dentry, old_index);
437 (void)simple_offset_replace(new_ctx, new_dentry, new_index);
Chuck Lever6faddda692023-06-30 13:48:49 -0400438 return ret;
439}
440
441/**
442 * simple_offset_destroy - Release offset map
443 * @octx: directory offset ctx that is about to be destroyed
444 *
445 * During fs teardown (eg. umount), a directory's offset map might still
446 * contain entries. xa_destroy() cleans out anything that remains.
447 */
448void simple_offset_destroy(struct offset_ctx *octx)
449{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500450 mtree_destroy(&octx->mt);
Chuck Lever6faddda692023-06-30 13:48:49 -0400451}
452
453/**
454 * offset_dir_llseek - Advance the read position of a directory descriptor
455 * @file: an open directory whose position is to be updated
456 * @offset: a byte offset
457 * @whence: enumerator describing the starting position for this update
458 *
459 * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories.
460 *
461 * Returns the updated read position if successful; otherwise a
462 * negative errno is returned and the read position remains unchanged.
463 */
464static loff_t offset_dir_llseek(struct file *file, loff_t offset, int whence)
465{
466 switch (whence) {
467 case SEEK_CUR:
468 offset += file->f_pos;
469 fallthrough;
470 case SEEK_SET:
471 if (offset >= 0)
472 break;
473 fallthrough;
474 default:
475 return -EINVAL;
476 }
477
Chuck Lever796432e2023-11-19 18:56:17 -0500478 /* In this case, ->private_data is protected by f_pos_lock */
479 file->private_data = NULL;
Chuck Lever0e4a8622024-02-17 15:24:16 -0500480 return vfs_setpos(file, offset, LONG_MAX);
Chuck Lever6faddda692023-06-30 13:48:49 -0400481}
482
Chuck Lever3f6d8102024-02-17 15:23:40 -0500483static struct dentry *offset_find_next(struct offset_ctx *octx, loff_t offset)
Chuck Lever6faddda692023-06-30 13:48:49 -0400484{
Chuck Lever0e4a8622024-02-17 15:24:16 -0500485 MA_STATE(mas, &octx->mt, offset, offset);
Chuck Lever6faddda692023-06-30 13:48:49 -0400486 struct dentry *child, *found = NULL;
487
488 rcu_read_lock();
Chuck Lever0e4a8622024-02-17 15:24:16 -0500489 child = mas_find(&mas, LONG_MAX);
Chuck Lever6faddda692023-06-30 13:48:49 -0400490 if (!child)
491 goto out;
Chuck Lever2be4f052023-07-25 14:31:04 -0400492 spin_lock(&child->d_lock);
Chuck Lever6faddda692023-06-30 13:48:49 -0400493 if (simple_positive(child))
494 found = dget_dlock(child);
495 spin_unlock(&child->d_lock);
496out:
497 rcu_read_unlock();
498 return found;
499}
500
501static bool offset_dir_emit(struct dir_context *ctx, struct dentry *dentry)
502{
Chuck Lever6faddda692023-06-30 13:48:49 -0400503 struct inode *inode = d_inode(dentry);
Chuck Lever0e4a8622024-02-17 15:24:16 -0500504 long offset = dentry2offset(dentry);
Chuck Lever6faddda692023-06-30 13:48:49 -0400505
506 return ctx->actor(ctx, dentry->d_name.name, dentry->d_name.len, offset,
507 inode->i_ino, fs_umode_to_dtype(inode->i_mode));
508}
509
Chuck Lever796432e2023-11-19 18:56:17 -0500510static void *offset_iterate_dir(struct inode *inode, struct dir_context *ctx)
Chuck Lever6faddda692023-06-30 13:48:49 -0400511{
Chuck Lever3f6d8102024-02-17 15:23:40 -0500512 struct offset_ctx *octx = inode->i_op->get_offset_ctx(inode);
Chuck Lever6faddda692023-06-30 13:48:49 -0400513 struct dentry *dentry;
514
515 while (true) {
Chuck Lever3f6d8102024-02-17 15:23:40 -0500516 dentry = offset_find_next(octx, ctx->pos);
Chuck Lever6faddda692023-06-30 13:48:49 -0400517 if (!dentry)
Chuck Lever796432e2023-11-19 18:56:17 -0500518 return ERR_PTR(-ENOENT);
Chuck Lever6faddda692023-06-30 13:48:49 -0400519
520 if (!offset_dir_emit(ctx, dentry)) {
521 dput(dentry);
522 break;
523 }
524
Chuck Lever3f6d8102024-02-17 15:23:40 -0500525 ctx->pos = dentry2offset(dentry) + 1;
Chuck Lever6faddda692023-06-30 13:48:49 -0400526 dput(dentry);
Chuck Lever6faddda692023-06-30 13:48:49 -0400527 }
Chuck Lever796432e2023-11-19 18:56:17 -0500528 return NULL;
Chuck Lever6faddda692023-06-30 13:48:49 -0400529}
530
531/**
532 * offset_readdir - Emit entries starting at offset @ctx->pos
533 * @file: an open directory to iterate over
534 * @ctx: directory iteration context
535 *
536 * Caller must hold @file's i_rwsem to prevent insertion or removal of
537 * entries during this call.
538 *
539 * On entry, @ctx->pos contains an offset that represents the first entry
540 * to be read from the directory.
541 *
542 * The operation continues until there are no more entries to read, or
543 * until the ctx->actor indicates there is no more space in the caller's
544 * output buffer.
545 *
546 * On return, @ctx->pos contains an offset that will read the next entry
Chuck Lever2be4f052023-07-25 14:31:04 -0400547 * in this directory when offset_readdir() is called again with @ctx.
Chuck Lever6faddda692023-06-30 13:48:49 -0400548 *
549 * Return values:
550 * %0 - Complete
551 */
552static int offset_readdir(struct file *file, struct dir_context *ctx)
553{
554 struct dentry *dir = file->f_path.dentry;
555
556 lockdep_assert_held(&d_inode(dir)->i_rwsem);
557
558 if (!dir_emit_dots(file, ctx))
559 return 0;
560
Chuck Lever796432e2023-11-19 18:56:17 -0500561 /* In this case, ->private_data is protected by f_pos_lock */
Chuck Lever7beea722024-02-17 15:23:47 -0500562 if (ctx->pos == DIR_OFFSET_MIN)
Chuck Lever796432e2023-11-19 18:56:17 -0500563 file->private_data = NULL;
564 else if (file->private_data == ERR_PTR(-ENOENT))
565 return 0;
566 file->private_data = offset_iterate_dir(d_inode(dir), ctx);
Chuck Lever6faddda692023-06-30 13:48:49 -0400567 return 0;
568}
569
570const struct file_operations simple_offset_dir_operations = {
571 .llseek = offset_dir_llseek,
572 .iterate_shared = offset_readdir,
573 .read = generic_read_dir,
574 .fsync = noop_fsync,
575};
576
Al Viroa3d1e7e2019-11-18 09:43:10 -0500577static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
578{
Al Viroda549bd2023-11-07 02:00:39 -0500579 struct dentry *child = NULL, *d;
Al Viroa3d1e7e2019-11-18 09:43:10 -0500580
581 spin_lock(&parent->d_lock);
Al Viroda549bd2023-11-07 02:00:39 -0500582 d = prev ? d_next_sibling(prev) : d_first_child(parent);
583 hlist_for_each_entry_from(d, d_sib) {
Al Viroa3d1e7e2019-11-18 09:43:10 -0500584 if (simple_positive(d)) {
585 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
586 if (simple_positive(d))
587 child = dget_dlock(d);
588 spin_unlock(&d->d_lock);
589 if (likely(child))
590 break;
591 }
592 }
593 spin_unlock(&parent->d_lock);
594 dput(prev);
595 return child;
596}
597
598void simple_recursive_removal(struct dentry *dentry,
599 void (*callback)(struct dentry *))
600{
601 struct dentry *this = dget(dentry);
602 while (true) {
603 struct dentry *victim = NULL, *child;
604 struct inode *inode = this->d_inode;
605
606 inode_lock(inode);
607 if (d_is_dir(this))
608 inode->i_flags |= S_DEAD;
609 while ((child = find_next_child(this, victim)) == NULL) {
610 // kill and ascend
611 // update metadata while it's still locked
Jeff Laytonf7f43852023-07-05 15:01:21 -0400612 inode_set_ctime_current(inode);
Al Viroa3d1e7e2019-11-18 09:43:10 -0500613 clear_nlink(inode);
614 inode_unlock(inode);
615 victim = this;
616 this = this->d_parent;
617 inode = this->d_inode;
618 inode_lock(inode);
619 if (simple_positive(victim)) {
620 d_invalidate(victim); // avoid lost mounts
621 if (d_is_dir(victim))
622 fsnotify_rmdir(inode, victim);
623 else
624 fsnotify_unlink(inode, victim);
625 if (callback)
626 callback(victim);
627 dput(victim); // unpin it
628 }
629 if (victim == dentry) {
Jeff Layton077c2122023-10-04 14:52:37 -0400630 inode_set_mtime_to_ts(inode,
631 inode_set_ctime_current(inode));
Al Viroa3d1e7e2019-11-18 09:43:10 -0500632 if (d_is_dir(dentry))
633 drop_nlink(inode);
634 inode_unlock(inode);
635 dput(dentry);
636 return;
637 }
638 }
639 inode_unlock(inode);
640 this = child;
641 }
642}
643EXPORT_SYMBOL(simple_recursive_removal);
644
Hugh Dickins759b9772007-03-05 00:30:28 -0800645static const struct super_operations simple_super_operations = {
646 .statfs = simple_statfs,
647};
648
David Howellsdb2c2462019-03-25 16:38:26 +0000649static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650{
David Howells31d6d5c2019-03-25 16:38:23 +0000651 struct pseudo_fs_context *ctx = fc->fs_private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 struct inode *root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Jeff Layton89a4eb42009-08-18 14:11:08 -0700654 s->s_maxbytes = MAX_LFS_FILESIZE;
Alex Nixon3971e1a2008-07-29 22:33:03 -0700655 s->s_blocksize = PAGE_SIZE;
656 s->s_blocksize_bits = PAGE_SHIFT;
Al Viro8d9e46d2019-05-11 11:43:59 -0400657 s->s_magic = ctx->magic;
658 s->s_op = ctx->ops ?: &simple_super_operations;
659 s->s_xattr = ctx->xattr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 s->s_time_gran = 1;
661 root = new_inode(s);
662 if (!root)
David Howellsdb2c2462019-03-25 16:38:26 +0000663 return -ENOMEM;
664
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700665 /*
666 * since this is the first inode, make it number 1. New inodes created
667 * after this must take care not to collide with it (by passing
668 * max_reserved of 1 to iunique).
669 */
670 root->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
Jeff Layton077c2122023-10-04 14:52:37 -0400672 simple_inode_init_ts(root);
Al Viro8d9e46d2019-05-11 11:43:59 -0400673 s->s_root = d_make_root(root);
674 if (!s->s_root)
David Howellsdb2c2462019-03-25 16:38:26 +0000675 return -ENOMEM;
Al Viro8d9e46d2019-05-11 11:43:59 -0400676 s->s_d_op = ctx->dops;
David Howells31d6d5c2019-03-25 16:38:23 +0000677 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678}
Al Viro8d9e46d2019-05-11 11:43:59 -0400679
David Howellsdb2c2462019-03-25 16:38:26 +0000680static int pseudo_fs_get_tree(struct fs_context *fc)
681{
Al Viro2ac295d2019-06-01 20:48:55 -0400682 return get_tree_nodev(fc, pseudo_fs_fill_super);
David Howells31d6d5c2019-03-25 16:38:23 +0000683}
684
685static void pseudo_fs_free(struct fs_context *fc)
686{
687 kfree(fc->fs_private);
688}
689
690static const struct fs_context_operations pseudo_fs_context_ops = {
691 .free = pseudo_fs_free,
692 .get_tree = pseudo_fs_get_tree,
693};
694
695/*
696 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
697 * will never be mountable)
698 */
699struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
700 unsigned long magic)
701{
702 struct pseudo_fs_context *ctx;
703
704 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
705 if (likely(ctx)) {
706 ctx->magic = magic;
707 fc->fs_private = ctx;
708 fc->ops = &pseudo_fs_context_ops;
David Howellsdb2c2462019-03-25 16:38:26 +0000709 fc->sb_flags |= SB_NOUSER;
710 fc->global = true;
David Howells31d6d5c2019-03-25 16:38:23 +0000711 }
712 return ctx;
713}
714EXPORT_SYMBOL(init_pseudo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Stephen Boyd20955e82012-04-05 14:25:09 -0700716int simple_open(struct inode *inode, struct file *file)
717{
718 if (inode->i_private)
719 file->private_data = inode->i_private;
720 return 0;
721}
Al Viro12f38872013-09-15 21:20:49 -0400722EXPORT_SYMBOL(simple_open);
Stephen Boyd20955e82012-04-05 14:25:09 -0700723
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
725{
David Howellsdea655c2015-03-17 22:26:15 +0000726 struct inode *inode = d_inode(old_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
Jeff Layton077c2122023-10-04 14:52:37 -0400728 inode_set_mtime_to_ts(dir,
729 inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode)));
Dave Hansend8c76e62006-09-30 23:29:04 -0700730 inc_nlink(inode);
Al Viro7de9c6ee2010-10-23 11:11:40 -0400731 ihold(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 dget(dentry);
733 d_instantiate(dentry, inode);
734 return 0;
735}
Al Viro12f38872013-09-15 21:20:49 -0400736EXPORT_SYMBOL(simple_link);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738int simple_empty(struct dentry *dentry)
739{
740 struct dentry *child;
741 int ret = 0;
742
Nick Piggin2fd6b7f2011-01-07 17:49:34 +1100743 spin_lock(&dentry->d_lock);
Al Viroda549bd2023-11-07 02:00:39 -0500744 hlist_for_each_entry(child, &dentry->d_children, d_sib) {
Nick Pigginda502952011-01-07 17:49:33 +1100745 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
746 if (simple_positive(child)) {
747 spin_unlock(&child->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto out;
Nick Pigginda502952011-01-07 17:49:33 +1100749 }
750 spin_unlock(&child->d_lock);
751 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 ret = 1;
753out:
Nick Piggin2fd6b7f2011-01-07 17:49:34 +1100754 spin_unlock(&dentry->d_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 return ret;
756}
Al Viro12f38872013-09-15 21:20:49 -0400757EXPORT_SYMBOL(simple_empty);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
759int simple_unlink(struct inode *dir, struct dentry *dentry)
760{
David Howellsdea655c2015-03-17 22:26:15 +0000761 struct inode *inode = d_inode(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
Jeff Layton077c2122023-10-04 14:52:37 -0400763 inode_set_mtime_to_ts(dir,
764 inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode)));
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700765 drop_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 dput(dentry);
767 return 0;
768}
Al Viro12f38872013-09-15 21:20:49 -0400769EXPORT_SYMBOL(simple_unlink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
771int simple_rmdir(struct inode *dir, struct dentry *dentry)
772{
773 if (!simple_empty(dentry))
774 return -ENOTEMPTY;
775
David Howellsdea655c2015-03-17 22:26:15 +0000776 drop_nlink(d_inode(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 simple_unlink(dir, dentry);
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700778 drop_nlink(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 return 0;
780}
Al Viro12f38872013-09-15 21:20:49 -0400781EXPORT_SYMBOL(simple_rmdir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782
Jeff Layton0c476792023-07-05 14:58:11 -0400783/**
784 * simple_rename_timestamp - update the various inode timestamps for rename
785 * @old_dir: old parent directory
786 * @old_dentry: dentry that is being renamed
787 * @new_dir: new parent directory
788 * @new_dentry: target for rename
789 *
790 * POSIX mandates that the old and new parent directories have their ctime and
791 * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have
792 * their ctime updated.
793 */
794void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry,
795 struct inode *new_dir, struct dentry *new_dentry)
796{
797 struct inode *newino = d_inode(new_dentry);
798
Jeff Layton077c2122023-10-04 14:52:37 -0400799 inode_set_mtime_to_ts(old_dir, inode_set_ctime_current(old_dir));
Jeff Layton0c476792023-07-05 14:58:11 -0400800 if (new_dir != old_dir)
Jeff Layton077c2122023-10-04 14:52:37 -0400801 inode_set_mtime_to_ts(new_dir,
802 inode_set_ctime_current(new_dir));
Jeff Layton0c476792023-07-05 14:58:11 -0400803 inode_set_ctime_current(d_inode(old_dentry));
804 if (newino)
805 inode_set_ctime_current(newino);
806}
807EXPORT_SYMBOL_GPL(simple_rename_timestamp);
808
Lorenz Bauer6429e462021-10-28 10:47:21 +0100809int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
810 struct inode *new_dir, struct dentry *new_dentry)
811{
812 bool old_is_dir = d_is_dir(old_dentry);
813 bool new_is_dir = d_is_dir(new_dentry);
814
815 if (old_dir != new_dir && old_is_dir != new_is_dir) {
816 if (old_is_dir) {
817 drop_nlink(old_dir);
818 inc_nlink(new_dir);
819 } else {
820 drop_nlink(new_dir);
821 inc_nlink(old_dir);
822 }
823 }
Jeff Layton0c476792023-07-05 14:58:11 -0400824 simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry);
Lorenz Bauer6429e462021-10-28 10:47:21 +0100825 return 0;
826}
827EXPORT_SYMBOL_GPL(simple_rename_exchange);
828
Christian Braunere18275a2023-01-13 12:49:17 +0100829int simple_rename(struct mnt_idmap *idmap, struct inode *old_dir,
Christian Brauner549c7292021-01-21 14:19:43 +0100830 struct dentry *old_dentry, struct inode *new_dir,
831 struct dentry *new_dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832{
David Howellse36cb0b2015-01-29 12:02:35 +0000833 int they_are_dirs = d_is_dir(old_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
Lorenz Bauer3871cb82021-10-28 10:47:22 +0100835 if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
Miklos Szeredie0e0be82016-09-27 11:03:57 +0200836 return -EINVAL;
837
Lorenz Bauer3871cb82021-10-28 10:47:22 +0100838 if (flags & RENAME_EXCHANGE)
839 return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
840
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 if (!simple_empty(new_dentry))
842 return -ENOTEMPTY;
843
David Howellsdea655c2015-03-17 22:26:15 +0000844 if (d_really_is_positive(new_dentry)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 simple_unlink(new_dir, new_dentry);
Al Viro841590c2011-07-21 15:49:09 -0400846 if (they_are_dirs) {
David Howellsdea655c2015-03-17 22:26:15 +0000847 drop_nlink(d_inode(new_dentry));
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700848 drop_nlink(old_dir);
Al Viro841590c2011-07-21 15:49:09 -0400849 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 } else if (they_are_dirs) {
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700851 drop_nlink(old_dir);
Dave Hansend8c76e62006-09-30 23:29:04 -0700852 inc_nlink(new_dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 }
854
Jeff Layton0c476792023-07-05 14:58:11 -0400855 simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 return 0;
857}
Al Viro12f38872013-09-15 21:20:49 -0400858EXPORT_SYMBOL(simple_rename);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000860/**
Christoph Hellwigeef23802010-06-04 11:30:01 +0200861 * simple_setattr - setattr for simple filesystem
Christian Braunerc1632a02023-01-13 12:49:11 +0100862 * @idmap: idmap of the target mount
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000863 * @dentry: dentry
864 * @iattr: iattr structure
865 *
866 * Returns 0 on success, -error on failure.
867 *
Christoph Hellwigeef23802010-06-04 11:30:01 +0200868 * simple_setattr is a simple ->setattr implementation without a proper
869 * implementation of size changes.
870 *
871 * It can either be used for in-memory filesystems or special files
872 * on simple regular filesystems. Anything that needs to change on-disk
873 * or wire state on size changes needs its own setattr method.
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000874 */
Christian Braunerc1632a02023-01-13 12:49:11 +0100875int simple_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
Christian Brauner549c7292021-01-21 14:19:43 +0100876 struct iattr *iattr)
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000877{
David Howellsdea655c2015-03-17 22:26:15 +0000878 struct inode *inode = d_inode(dentry);
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000879 int error;
880
Christian Braunerc1632a02023-01-13 12:49:11 +0100881 error = setattr_prepare(idmap, dentry, iattr);
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000882 if (error)
883 return error;
884
Christoph Hellwig2c27c652010-06-04 11:30:04 +0200885 if (iattr->ia_valid & ATTR_SIZE)
886 truncate_setsize(inode, iattr->ia_size);
Christian Braunerc1632a02023-01-13 12:49:11 +0100887 setattr_copy(idmap, inode, iattr);
Christoph Hellwigeef23802010-06-04 11:30:01 +0200888 mark_inode_dirty(inode);
889 return 0;
npiggin@suse.de7bb46a62010-05-27 01:05:33 +1000890}
891EXPORT_SYMBOL(simple_setattr);
892
Matthew Wilcox (Oracle)a77f5802022-04-29 11:49:41 -0400893static int simple_read_folio(struct file *file, struct folio *folio)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894{
Matthew Wilcox (Oracle)a77f5802022-04-29 11:49:41 -0400895 folio_zero_range(folio, 0, folio_size(folio));
896 flush_dcache_folio(folio);
897 folio_mark_uptodate(folio);
898 folio_unlock(folio);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 return 0;
900}
901
Nick Pigginafddba42007-10-16 01:25:01 -0700902int simple_write_begin(struct file *file, struct address_space *mapping,
Matthew Wilcox (Oracle)9d6b0cd2022-02-22 14:31:43 -0500903 loff_t pos, unsigned len,
Nick Pigginafddba42007-10-16 01:25:01 -0700904 struct page **pagep, void **fsdata)
905{
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100906 struct folio *folio;
Nick Pigginafddba42007-10-16 01:25:01 -0700907
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100908 folio = __filemap_get_folio(mapping, pos / PAGE_SIZE, FGP_WRITEBEGIN,
909 mapping_gfp_mask(mapping));
910 if (IS_ERR(folio))
911 return PTR_ERR(folio);
Nick Pigginafddba42007-10-16 01:25:01 -0700912
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100913 *pagep = &folio->page;
Nick Pigginafddba42007-10-16 01:25:01 -0700914
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100915 if (!folio_test_uptodate(folio) && (len != folio_size(folio))) {
916 size_t from = offset_in_folio(folio, pos);
Nick Pigginafddba42007-10-16 01:25:01 -0700917
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100918 folio_zero_segments(folio, 0, from,
919 from + len, folio_size(folio));
Boaz Harrosh193cf4b2010-01-12 16:18:08 +0200920 }
921 return 0;
Nick Pigginafddba42007-10-16 01:25:01 -0700922}
Al Viro12f38872013-09-15 21:20:49 -0400923EXPORT_SYMBOL(simple_write_begin);
Nick Pigginafddba42007-10-16 01:25:01 -0700924
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200925/**
926 * simple_write_end - .write_end helper for non-block-device FSes
Randy Dunlap8e88bfb2019-10-14 14:12:14 -0700927 * @file: See .write_end of address_space_operations
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200928 * @mapping: "
929 * @pos: "
930 * @len: "
931 * @copied: "
932 * @page: "
933 * @fsdata: "
934 *
935 * simple_write_end does the minimum needed for updating a page after writing is
936 * done. It has the same API signature as the .write_end of
937 * address_space_operations vector. So it can just be set onto .write_end for
938 * FSes that don't need any other processing. i_mutex is assumed to be held.
939 * Block based filesystems should use generic_write_end().
940 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
941 * is not called, so a filesystem that actually does store data in .write_inode
942 * should extend on what's done here with a call to mark_inode_dirty() in the
943 * case that i_size has changed.
Al Viro04fff642016-08-29 22:39:56 -0400944 *
Matthew Wilcox (Oracle)a77f5802022-04-29 11:49:41 -0400945 * Use *ONLY* with simple_read_folio()
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200946 */
Christoph Hellwigc1e3dbe2021-06-28 19:36:09 -0700947static int simple_write_end(struct file *file, struct address_space *mapping,
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200948 loff_t pos, unsigned len, unsigned copied,
949 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950{
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100951 struct folio *folio = page_folio(page);
952 struct inode *inode = folio->mapping->host;
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200953 loff_t last_pos = pos + copied;
954
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100955 /* zero the stale part of the folio if we did a short copy */
956 if (!folio_test_uptodate(folio)) {
Al Viro04fff642016-08-29 22:39:56 -0400957 if (copied < len) {
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100958 size_t from = offset_in_folio(folio, pos);
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200959
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100960 folio_zero_range(folio, from + copied, len - copied);
Al Viro04fff642016-08-29 22:39:56 -0400961 }
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100962 folio_mark_uptodate(folio);
Al Viro04fff642016-08-29 22:39:56 -0400963 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 /*
965 * No need to use i_size_read() here, the i_size
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800966 * cannot change under us because we hold the i_mutex.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 */
Boaz Harroshad2a722f2010-01-12 15:13:47 +0200968 if (last_pos > inode->i_size)
969 i_size_write(inode, last_pos);
970
Matthew Wilcox (Oracle)5522d9f2023-08-21 15:13:22 +0100971 folio_mark_dirty(folio);
972 folio_unlock(folio);
973 folio_put(folio);
Nick Pigginafddba42007-10-16 01:25:01 -0700974
975 return copied;
976}
Christoph Hellwigc1e3dbe2021-06-28 19:36:09 -0700977
978/*
979 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
980 */
981const struct address_space_operations ram_aops = {
Matthew Wilcox (Oracle)a77f5802022-04-29 11:49:41 -0400982 .read_folio = simple_read_folio,
Christoph Hellwigc1e3dbe2021-06-28 19:36:09 -0700983 .write_begin = simple_write_begin,
984 .write_end = simple_write_end,
Matthew Wilcox (Oracle)46de8b972022-02-09 20:22:13 +0000985 .dirty_folio = noop_dirty_folio,
Christoph Hellwigc1e3dbe2021-06-28 19:36:09 -0700986};
987EXPORT_SYMBOL(ram_aops);
Nick Pigginafddba42007-10-16 01:25:01 -0700988
Jeff Layton1a1c9bb2007-05-08 00:32:31 -0700989/*
990 * the inodes created here are not hashed. If you use iunique to generate
991 * unique inode values later for this filesystem, then you must take care
992 * to pass it an appropriate max_reserved value to avoid collisions.
993 */
Roberto Sassu7d683a02010-06-03 11:58:28 +0200994int simple_fill_super(struct super_block *s, unsigned long magic,
Eric Biggerscda37122017-03-25 21:15:37 -0700995 const struct tree_descr *files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 struct dentry *dentry;
999 int i;
1000
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001001 s->s_blocksize = PAGE_SIZE;
1002 s->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 s->s_magic = magic;
Hugh Dickins759b9772007-03-05 00:30:28 -08001004 s->s_op = &simple_super_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 s->s_time_gran = 1;
1006
1007 inode = new_inode(s);
1008 if (!inode)
1009 return -ENOMEM;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -07001010 /*
1011 * because the root inode is 1, the files array must not contain an
1012 * entry at index 1
1013 */
1014 inode->i_ino = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 inode->i_mode = S_IFDIR | 0755;
Jeff Layton077c2122023-10-04 14:52:37 -04001016 simple_inode_init_ts(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 inode->i_op = &simple_dir_inode_operations;
1018 inode->i_fop = &simple_dir_operations;
Miklos Szeredibfe86842011-10-28 14:13:29 +02001019 set_nlink(inode, 2);
Al Viro715cd662023-11-11 15:56:55 -05001020 s->s_root = d_make_root(inode);
1021 if (!s->s_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 for (i = 0; !files->name || files->name[0]; i++, files++) {
1024 if (!files->name)
1025 continue;
Jeff Layton1a1c9bb2007-05-08 00:32:31 -07001026
1027 /* warn if it tries to conflict with the root inode */
1028 if (unlikely(i == 1))
1029 printk(KERN_WARNING "%s: %s passed in a files array"
1030 "with an index of 1!\n", __func__,
1031 s->s_type->name);
1032
Al Viro715cd662023-11-11 15:56:55 -05001033 dentry = d_alloc_name(s->s_root, files->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 if (!dentry)
Al Viro715cd662023-11-11 15:56:55 -05001035 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 inode = new_inode(s);
Konstantin Khlebnikov32096ea2011-11-01 16:12:33 +03001037 if (!inode) {
1038 dput(dentry);
Al Viro715cd662023-11-11 15:56:55 -05001039 return -ENOMEM;
Konstantin Khlebnikov32096ea2011-11-01 16:12:33 +03001040 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 inode->i_mode = S_IFREG | files->mode;
Jeff Layton077c2122023-10-04 14:52:37 -04001042 simple_inode_init_ts(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 inode->i_fop = files->ops;
1044 inode->i_ino = i;
1045 d_add(dentry, inode);
1046 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048}
Al Viro12f38872013-09-15 21:20:49 -04001049EXPORT_SYMBOL(simple_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
1051static DEFINE_SPINLOCK(pin_fs_lock);
1052
Trond Myklebust1f5ce9e2006-06-09 09:34:16 -04001053int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054{
1055 struct vfsmount *mnt = NULL;
1056 spin_lock(&pin_fs_lock);
1057 if (unlikely(!*mount)) {
1058 spin_unlock(&pin_fs_lock);
Linus Torvalds1751e8a2017-11-27 13:05:09 -08001059 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 if (IS_ERR(mnt))
1061 return PTR_ERR(mnt);
1062 spin_lock(&pin_fs_lock);
1063 if (!*mount)
1064 *mount = mnt;
1065 }
1066 mntget(*mount);
1067 ++*count;
1068 spin_unlock(&pin_fs_lock);
1069 mntput(mnt);
1070 return 0;
1071}
Al Viro12f38872013-09-15 21:20:49 -04001072EXPORT_SYMBOL(simple_pin_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073
1074void simple_release_fs(struct vfsmount **mount, int *count)
1075{
1076 struct vfsmount *mnt;
1077 spin_lock(&pin_fs_lock);
1078 mnt = *mount;
1079 if (!--*count)
1080 *mount = NULL;
1081 spin_unlock(&pin_fs_lock);
1082 mntput(mnt);
1083}
Al Viro12f38872013-09-15 21:20:49 -04001084EXPORT_SYMBOL(simple_release_fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085
Akinobu Mita6d1029b2008-07-04 09:59:51 -07001086/**
1087 * simple_read_from_buffer - copy data from the buffer to user space
1088 * @to: the user space buffer to read to
1089 * @count: the maximum number of bytes to read
1090 * @ppos: the current position in the buffer
1091 * @from: the buffer to read from
1092 * @available: the size of the buffer
1093 *
1094 * The simple_read_from_buffer() function reads up to @count bytes from the
1095 * buffer @from at offset @ppos into the user space address starting at @to.
1096 *
1097 * On success, the number of bytes read is returned and the offset @ppos is
1098 * advanced by this number, or negative value is returned on error.
1099 **/
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
1101 const void *from, size_t available)
1102{
1103 loff_t pos = *ppos;
Steven Rostedt14be2742009-09-18 13:05:42 -07001104 size_t ret;
1105
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 if (pos < 0)
1107 return -EINVAL;
Steven Rostedt14be2742009-09-18 13:05:42 -07001108 if (pos >= available || !count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 return 0;
1110 if (count > available - pos)
1111 count = available - pos;
Steven Rostedt14be2742009-09-18 13:05:42 -07001112 ret = copy_to_user(to, from + pos, count);
1113 if (ret == count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 return -EFAULT;
Steven Rostedt14be2742009-09-18 13:05:42 -07001115 count -= ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 *ppos = pos + count;
1117 return count;
1118}
Al Viro12f38872013-09-15 21:20:49 -04001119EXPORT_SYMBOL(simple_read_from_buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120
Akinobu Mita6d1029b2008-07-04 09:59:51 -07001121/**
Jiri Slaby6a727b42010-05-01 23:51:22 +02001122 * simple_write_to_buffer - copy data from user space to the buffer
1123 * @to: the buffer to write to
1124 * @available: the size of the buffer
1125 * @ppos: the current position in the buffer
1126 * @from: the user space buffer to read from
1127 * @count: the maximum number of bytes to read
1128 *
1129 * The simple_write_to_buffer() function reads up to @count bytes from the user
1130 * space address starting at @from into the buffer @to at offset @ppos.
1131 *
1132 * On success, the number of bytes written is returned and the offset @ppos is
1133 * advanced by this number, or negative value is returned on error.
1134 **/
1135ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
1136 const void __user *from, size_t count)
1137{
1138 loff_t pos = *ppos;
1139 size_t res;
1140
1141 if (pos < 0)
1142 return -EINVAL;
1143 if (pos >= available || !count)
1144 return 0;
1145 if (count > available - pos)
1146 count = available - pos;
1147 res = copy_from_user(to + pos, from, count);
1148 if (res == count)
1149 return -EFAULT;
1150 count -= res;
1151 *ppos = pos + count;
1152 return count;
1153}
Al Viro12f38872013-09-15 21:20:49 -04001154EXPORT_SYMBOL(simple_write_to_buffer);
Jiri Slaby6a727b42010-05-01 23:51:22 +02001155
1156/**
Akinobu Mita6d1029b2008-07-04 09:59:51 -07001157 * memory_read_from_buffer - copy data from the buffer
1158 * @to: the kernel space buffer to read to
1159 * @count: the maximum number of bytes to read
1160 * @ppos: the current position in the buffer
1161 * @from: the buffer to read from
1162 * @available: the size of the buffer
1163 *
1164 * The memory_read_from_buffer() function reads up to @count bytes from the
1165 * buffer @from at offset @ppos into the kernel space address starting at @to.
1166 *
1167 * On success, the number of bytes read is returned and the offset @ppos is
1168 * advanced by this number, or negative value is returned on error.
1169 **/
Akinobu Mita93b07112008-06-05 22:46:21 -07001170ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
1171 const void *from, size_t available)
1172{
1173 loff_t pos = *ppos;
1174
1175 if (pos < 0)
1176 return -EINVAL;
1177 if (pos >= available)
1178 return 0;
1179 if (count > available - pos)
1180 count = available - pos;
1181 memcpy(to, from + pos, count);
1182 *ppos = pos + count;
1183
1184 return count;
1185}
Al Viro12f38872013-09-15 21:20:49 -04001186EXPORT_SYMBOL(memory_read_from_buffer);
Akinobu Mita93b07112008-06-05 22:46:21 -07001187
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188/*
1189 * Transaction based IO.
1190 * The file expects a single write which triggers the transaction, and then
1191 * possibly a read which collects the result - which is stored in a
1192 * file-local buffer.
1193 */
Ingo Molnar76791ab2009-03-25 16:48:35 +01001194
1195void simple_transaction_set(struct file *file, size_t n)
1196{
1197 struct simple_transaction_argresp *ar = file->private_data;
1198
1199 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
1200
1201 /*
1202 * The barrier ensures that ar->size will really remain zero until
1203 * ar->data is ready for reading.
1204 */
1205 smp_mb();
1206 ar->size = n;
1207}
Al Viro12f38872013-09-15 21:20:49 -04001208EXPORT_SYMBOL(simple_transaction_set);
Ingo Molnar76791ab2009-03-25 16:48:35 +01001209
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
1211{
1212 struct simple_transaction_argresp *ar;
1213 static DEFINE_SPINLOCK(simple_transaction_lock);
1214
1215 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
1216 return ERR_PTR(-EFBIG);
1217
1218 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
1219 if (!ar)
1220 return ERR_PTR(-ENOMEM);
1221
1222 spin_lock(&simple_transaction_lock);
1223
1224 /* only one write allowed per open */
1225 if (file->private_data) {
1226 spin_unlock(&simple_transaction_lock);
1227 free_page((unsigned long)ar);
1228 return ERR_PTR(-EBUSY);
1229 }
1230
1231 file->private_data = ar;
1232
1233 spin_unlock(&simple_transaction_lock);
1234
1235 if (copy_from_user(ar->data, buf, size))
1236 return ERR_PTR(-EFAULT);
1237
1238 return ar->data;
1239}
Al Viro12f38872013-09-15 21:20:49 -04001240EXPORT_SYMBOL(simple_transaction_get);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
1242ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
1243{
1244 struct simple_transaction_argresp *ar = file->private_data;
1245
1246 if (!ar)
1247 return 0;
1248 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
1249}
Al Viro12f38872013-09-15 21:20:49 -04001250EXPORT_SYMBOL(simple_transaction_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
1252int simple_transaction_release(struct inode *inode, struct file *file)
1253{
1254 free_page((unsigned long)file->private_data);
1255 return 0;
1256}
Al Viro12f38872013-09-15 21:20:49 -04001257EXPORT_SYMBOL(simple_transaction_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001259/* Simple attribute files */
1260
1261struct simple_attr {
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001262 int (*get)(void *, u64 *);
1263 int (*set)(void *, u64);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001264 char get_buf[24]; /* enough to store a u64 and "\n\0" */
1265 char set_buf[24];
1266 void *data;
1267 const char *fmt; /* format for read operation */
Ingo Molnar7cf34c72006-03-23 03:00:36 -08001268 struct mutex mutex; /* protects access to these buffers */
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001269};
1270
1271/* simple_attr_open is called by an actual attribute open file operation
1272 * to set the attribute specific access operations. */
1273int simple_attr_open(struct inode *inode, struct file *file,
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001274 int (*get)(void *, u64 *), int (*set)(void *, u64),
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001275 const char *fmt)
1276{
1277 struct simple_attr *attr;
1278
Eric Biggersa65cab72020-03-07 18:38:49 -08001279 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001280 if (!attr)
1281 return -ENOMEM;
1282
1283 attr->get = get;
1284 attr->set = set;
Theodore Ts'o8e18e292006-09-27 01:50:46 -07001285 attr->data = inode->i_private;
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001286 attr->fmt = fmt;
Ingo Molnar7cf34c72006-03-23 03:00:36 -08001287 mutex_init(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001288
1289 file->private_data = attr;
1290
1291 return nonseekable_open(inode, file);
1292}
Al Viro12f38872013-09-15 21:20:49 -04001293EXPORT_SYMBOL_GPL(simple_attr_open);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001294
Christoph Hellwig74bedc42008-02-08 04:20:28 -08001295int simple_attr_release(struct inode *inode, struct file *file)
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001296{
1297 kfree(file->private_data);
1298 return 0;
1299}
Al Viro12f38872013-09-15 21:20:49 -04001300EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001301
1302/* read from the buffer that is filled with the get function */
1303ssize_t simple_attr_read(struct file *file, char __user *buf,
1304 size_t len, loff_t *ppos)
1305{
1306 struct simple_attr *attr;
1307 size_t size;
1308 ssize_t ret;
1309
1310 attr = file->private_data;
1311
1312 if (!attr->get)
1313 return -EACCES;
1314
Christoph Hellwig92613032008-02-08 04:20:27 -08001315 ret = mutex_lock_interruptible(&attr->mutex);
1316 if (ret)
1317 return ret;
1318
Eric Biggersa65cab72020-03-07 18:38:49 -08001319 if (*ppos && attr->get_buf[0]) {
1320 /* continued read */
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001321 size = strlen(attr->get_buf);
Eric Biggersa65cab72020-03-07 18:38:49 -08001322 } else {
1323 /* first read */
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001324 u64 val;
1325 ret = attr->get(attr->data, &val);
1326 if (ret)
1327 goto out;
1328
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001329 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001330 attr->fmt, (unsigned long long)val);
1331 }
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001332
1333 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001334out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -08001335 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001336 return ret;
1337}
Al Viro12f38872013-09-15 21:20:49 -04001338EXPORT_SYMBOL_GPL(simple_attr_read);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001339
1340/* interpret the buffer as a number to call the set function with */
Akinobu Mita2e41f272022-09-20 02:24:16 +09001341static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf,
1342 size_t len, loff_t *ppos, bool is_signed)
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001343{
1344 struct simple_attr *attr;
Yicong Yang488dac02020-11-21 22:17:19 -08001345 unsigned long long val;
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001346 size_t size;
1347 ssize_t ret;
1348
1349 attr = file->private_data;
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001350 if (!attr->set)
1351 return -EACCES;
1352
Christoph Hellwig92613032008-02-08 04:20:27 -08001353 ret = mutex_lock_interruptible(&attr->mutex);
1354 if (ret)
1355 return ret;
1356
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001357 ret = -EFAULT;
1358 size = min(sizeof(attr->set_buf) - 1, len);
1359 if (copy_from_user(attr->set_buf, buf, size))
1360 goto out;
1361
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001362 attr->set_buf[size] = '\0';
Akinobu Mita2e41f272022-09-20 02:24:16 +09001363 if (is_signed)
1364 ret = kstrtoll(attr->set_buf, 0, &val);
1365 else
1366 ret = kstrtoull(attr->set_buf, 0, &val);
Yicong Yang488dac02020-11-21 22:17:19 -08001367 if (ret)
1368 goto out;
Wu Fengguang05cc0ce2009-09-18 13:06:03 -07001369 ret = attr->set(attr->data, val);
1370 if (ret == 0)
1371 ret = len; /* on success, claim we got the whole input */
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001372out:
Ingo Molnar7cf34c72006-03-23 03:00:36 -08001373 mutex_unlock(&attr->mutex);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001374 return ret;
1375}
Akinobu Mita2e41f272022-09-20 02:24:16 +09001376
1377ssize_t simple_attr_write(struct file *file, const char __user *buf,
1378 size_t len, loff_t *ppos)
1379{
1380 return simple_attr_write_xsigned(file, buf, len, ppos, false);
1381}
Al Viro12f38872013-09-15 21:20:49 -04001382EXPORT_SYMBOL_GPL(simple_attr_write);
Arnd Bergmannacaefc22005-05-18 14:40:59 +02001383
Akinobu Mita2e41f272022-09-20 02:24:16 +09001384ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
1385 size_t len, loff_t *ppos)
1386{
1387 return simple_attr_write_xsigned(file, buf, len, ppos, true);
1388}
1389EXPORT_SYMBOL_GPL(simple_attr_write_signed);
1390
Christoph Hellwig25961102007-10-21 16:42:05 -07001391/**
Amir Goldsteind9e5d922023-10-26 23:45:40 +03001392 * generic_encode_ino32_fh - generic export_operations->encode_fh function
1393 * @inode: the object to encode
1394 * @fh: where to store the file handle fragment
1395 * @max_len: maximum length to store there (in 4 byte units)
1396 * @parent: parent directory inode, if wanted
1397 *
1398 * This generic encode_fh function assumes that the 32 inode number
1399 * is suitable for locating an inode, and that the generation number
1400 * can be used to check that it is still valid. It places them in the
1401 * filehandle fragment where export_decode_fh expects to find them.
1402 */
1403int generic_encode_ino32_fh(struct inode *inode, __u32 *fh, int *max_len,
1404 struct inode *parent)
1405{
1406 struct fid *fid = (void *)fh;
1407 int len = *max_len;
1408 int type = FILEID_INO32_GEN;
1409
1410 if (parent && (len < 4)) {
1411 *max_len = 4;
1412 return FILEID_INVALID;
1413 } else if (len < 2) {
1414 *max_len = 2;
1415 return FILEID_INVALID;
1416 }
1417
1418 len = 2;
1419 fid->i32.ino = inode->i_ino;
1420 fid->i32.gen = inode->i_generation;
1421 if (parent) {
1422 fid->i32.parent_ino = parent->i_ino;
1423 fid->i32.parent_gen = parent->i_generation;
1424 len = 4;
1425 type = FILEID_INO32_GEN_PARENT;
1426 }
1427 *max_len = len;
1428 return type;
1429}
1430EXPORT_SYMBOL_GPL(generic_encode_ino32_fh);
1431
1432/**
Christoph Hellwig25961102007-10-21 16:42:05 -07001433 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1434 * @sb: filesystem to do the file handle conversion on
1435 * @fid: file handle to convert
1436 * @fh_len: length of the file handle in bytes
1437 * @fh_type: type of file handle
1438 * @get_inode: filesystem callback to retrieve inode
1439 *
1440 * This function decodes @fid as long as it has one of the well-known
1441 * Linux filehandle types and calls @get_inode on it to retrieve the
1442 * inode for the object specified in the file handle.
1443 */
1444struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1445 int fh_len, int fh_type, struct inode *(*get_inode)
1446 (struct super_block *sb, u64 ino, u32 gen))
1447{
1448 struct inode *inode = NULL;
1449
1450 if (fh_len < 2)
1451 return NULL;
1452
1453 switch (fh_type) {
1454 case FILEID_INO32_GEN:
1455 case FILEID_INO32_GEN_PARENT:
1456 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1457 break;
1458 }
1459
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +02001460 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -07001461}
1462EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1463
1464/**
Yanchuan Nianca186832012-09-05 16:31:29 +08001465 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
Christoph Hellwig25961102007-10-21 16:42:05 -07001466 * @sb: filesystem to do the file handle conversion on
1467 * @fid: file handle to convert
1468 * @fh_len: length of the file handle in bytes
1469 * @fh_type: type of file handle
1470 * @get_inode: filesystem callback to retrieve inode
1471 *
1472 * This function decodes @fid as long as it has one of the well-known
1473 * Linux filehandle types and calls @get_inode on it to retrieve the
1474 * inode for the _parent_ object specified in the file handle if it
1475 * is specified in the file handle, or NULL otherwise.
1476 */
1477struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1478 int fh_len, int fh_type, struct inode *(*get_inode)
1479 (struct super_block *sb, u64 ino, u32 gen))
1480{
1481 struct inode *inode = NULL;
1482
1483 if (fh_len <= 2)
1484 return NULL;
1485
1486 switch (fh_type) {
1487 case FILEID_INO32_GEN_PARENT:
1488 inode = get_inode(sb, fid->i32.parent_ino,
1489 (fh_len > 3 ? fid->i32.parent_gen : 0));
1490 break;
1491 }
1492
Christoph Hellwig4ea3ada2008-08-11 15:48:57 +02001493 return d_obtain_alias(inode);
Christoph Hellwig25961102007-10-21 16:42:05 -07001494}
1495EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1496
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001497/**
Fabian Frederickac13a822014-06-04 16:06:27 -07001498 * __generic_file_fsync - generic fsync implementation for simple filesystems
1499 *
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001500 * @file: file to synchronize
Fabian Frederickac13a822014-06-04 16:06:27 -07001501 * @start: start offset in bytes
1502 * @end: end offset in bytes (inclusive)
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001503 * @datasync: only synchronize essential metadata if true
1504 *
1505 * This is a generic implementation of the fsync method for simple
1506 * filesystems which track all non-inode metadata in the buffers list
1507 * hanging off the address_space structure.
1508 */
Fabian Frederickac13a822014-06-04 16:06:27 -07001509int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1510 int datasync)
Al Virod5aacad2009-06-07 14:56:44 -04001511{
Christoph Hellwig7ea80852010-05-26 17:53:25 +02001512 struct inode *inode = file->f_mapping->host;
Al Virod5aacad2009-06-07 14:56:44 -04001513 int err;
1514 int ret;
1515
Jeff Layton383aa542017-07-06 07:02:29 -04001516 err = file_write_and_wait_range(file, start, end);
Josef Bacik02c24a82011-07-16 20:44:56 -04001517 if (err)
1518 return err;
1519
Al Viro59551022016-01-22 15:40:57 -05001520 inode_lock(inode);
Al Virod5aacad2009-06-07 14:56:44 -04001521 ret = sync_mapping_buffers(inode->i_mapping);
Theodore Ts'o0ae45f62015-02-02 00:37:00 -05001522 if (!(inode->i_state & I_DIRTY_ALL))
Josef Bacik02c24a82011-07-16 20:44:56 -04001523 goto out;
Al Virod5aacad2009-06-07 14:56:44 -04001524 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
Josef Bacik02c24a82011-07-16 20:44:56 -04001525 goto out;
Al Virod5aacad2009-06-07 14:56:44 -04001526
Christoph Hellwigc37650162010-10-06 10:48:20 +02001527 err = sync_inode_metadata(inode, 1);
Al Virod5aacad2009-06-07 14:56:44 -04001528 if (ret == 0)
1529 ret = err;
Fabian Frederickac13a822014-06-04 16:06:27 -07001530
Josef Bacik02c24a82011-07-16 20:44:56 -04001531out:
Al Viro59551022016-01-22 15:40:57 -05001532 inode_unlock(inode);
Jeff Layton383aa542017-07-06 07:02:29 -04001533 /* check and advance again to catch errors after syncing out buffers */
1534 err = file_check_and_advance_wb_err(file);
1535 if (ret == 0)
1536 ret = err;
1537 return ret;
Al Virod5aacad2009-06-07 14:56:44 -04001538}
Fabian Frederickac13a822014-06-04 16:06:27 -07001539EXPORT_SYMBOL(__generic_file_fsync);
1540
1541/**
1542 * generic_file_fsync - generic fsync implementation for simple filesystems
1543 * with flush
1544 * @file: file to synchronize
1545 * @start: start offset in bytes
1546 * @end: end offset in bytes (inclusive)
1547 * @datasync: only synchronize essential metadata if true
1548 *
1549 */
1550
1551int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1552 int datasync)
1553{
1554 struct inode *inode = file->f_mapping->host;
1555 int err;
1556
1557 err = __generic_file_fsync(file, start, end, datasync);
1558 if (err)
1559 return err;
Christoph Hellwigc6bf3f02021-01-26 15:52:35 +01001560 return blkdev_issue_flush(inode->i_sb->s_bdev);
Fabian Frederickac13a822014-06-04 16:06:27 -07001561}
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001562EXPORT_SYMBOL(generic_file_fsync);
1563
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001564/**
1565 * generic_check_addressable - Check addressability of file system
1566 * @blocksize_bits: log of file system block size
1567 * @num_blocks: number of blocks in file system
1568 *
1569 * Determine whether a file system with @num_blocks blocks (and a
1570 * block size of 2**@blocksize_bits) is addressable by the sector_t
1571 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1572 */
1573int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1574{
1575 u64 last_fs_block = num_blocks - 1;
Joel Beckera33f13e2010-08-16 12:10:17 -07001576 u64 last_fs_page =
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001577 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001578
1579 if (unlikely(num_blocks == 0))
1580 return 0;
1581
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001582 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001583 return -EINVAL;
1584
Joel Beckera33f13e2010-08-16 12:10:17 -07001585 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1586 (last_fs_page > (pgoff_t)(~0ULL))) {
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07001587 return -EFBIG;
1588 }
1589 return 0;
1590}
1591EXPORT_SYMBOL(generic_check_addressable);
1592
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001593/*
1594 * No-op implementation of ->fsync for in-memory filesystems.
1595 */
Josef Bacik02c24a82011-07-16 20:44:56 -04001596int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001597{
1598 return 0;
1599}
Christoph Hellwig1b061d92010-05-26 17:53:41 +02001600EXPORT_SYMBOL(noop_fsync);
Al Viro87dc8002013-09-16 10:30:04 -04001601
Dan Williamsf44c7762018-03-07 15:26:44 -08001602ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1603{
1604 /*
1605 * iomap based filesystems support direct I/O without need for
1606 * this callback. However, it still needs to be set in
1607 * inode->a_ops so that open/fcntl know that direct I/O is
1608 * generally supported.
1609 */
1610 return -EINVAL;
1611}
1612EXPORT_SYMBOL_GPL(noop_direct_IO);
1613
Al Virofceef392015-12-29 15:58:39 -05001614/* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1615void kfree_link(void *p)
Al Viro87dc8002013-09-16 10:30:04 -04001616{
Al Virofceef392015-12-29 15:58:39 -05001617 kfree(p);
Al Viro87dc8002013-09-16 10:30:04 -04001618}
Al Virofceef392015-12-29 15:58:39 -05001619EXPORT_SYMBOL(kfree_link);
Al Viro69878432013-10-02 22:35:11 -04001620
Al Viro69878432013-10-02 22:35:11 -04001621struct inode *alloc_anon_inode(struct super_block *s)
1622{
1623 static const struct address_space_operations anon_aops = {
Matthew Wilcox (Oracle)46de8b972022-02-09 20:22:13 +00001624 .dirty_folio = noop_dirty_folio,
Al Viro69878432013-10-02 22:35:11 -04001625 };
1626 struct inode *inode = new_inode_pseudo(s);
1627
1628 if (!inode)
1629 return ERR_PTR(-ENOMEM);
1630
1631 inode->i_ino = get_next_ino();
1632 inode->i_mapping->a_ops = &anon_aops;
1633
1634 /*
1635 * Mark the inode dirty from the very beginning,
1636 * that way it will never be moved to the dirty
1637 * list because mark_inode_dirty() will think
1638 * that it already _is_ on the dirty list.
1639 */
1640 inode->i_state = I_DIRTY;
1641 inode->i_mode = S_IRUSR | S_IWUSR;
1642 inode->i_uid = current_fsuid();
1643 inode->i_gid = current_fsgid();
1644 inode->i_flags |= S_PRIVATE;
Jeff Layton077c2122023-10-04 14:52:37 -04001645 simple_inode_init_ts(inode);
Al Viro69878432013-10-02 22:35:11 -04001646 return inode;
1647}
1648EXPORT_SYMBOL(alloc_anon_inode);
Jeff Layton1c994a02014-08-27 06:49:41 -04001649
1650/**
1651 * simple_nosetlease - generic helper for prohibiting leases
1652 * @filp: file pointer
1653 * @arg: type of lease to obtain
1654 * @flp: new lease supplied for insertion
Jeff Laytone6f5c782014-08-22 10:40:25 -04001655 * @priv: private data for lm_setup operation
Jeff Layton1c994a02014-08-27 06:49:41 -04001656 *
1657 * Generic helper for filesystems that do not wish to allow leases to be set.
1658 * All arguments are ignored and it just returns -EINVAL.
1659 */
1660int
Jeff Laytonc69ff402024-01-31 18:02:28 -05001661simple_nosetlease(struct file *filp, int arg, struct file_lease **flp,
Jeff Laytone6f5c782014-08-22 10:40:25 -04001662 void **priv)
Jeff Layton1c994a02014-08-27 06:49:41 -04001663{
1664 return -EINVAL;
1665}
1666EXPORT_SYMBOL(simple_nosetlease);
Al Viro61ba64f2015-05-02 09:54:06 -04001667
Eric Biggers6ee97062019-04-11 16:16:30 -07001668/**
1669 * simple_get_link - generic helper to get the target of "fast" symlinks
1670 * @dentry: not used here
1671 * @inode: the symlink inode
1672 * @done: not used here
1673 *
1674 * Generic helper for filesystems to use for symlink inodes where a pointer to
1675 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1676 * since as an optimization the path lookup code uses any non-NULL ->i_link
1677 * directly, without calling ->get_link(). But ->get_link() still must be set,
1678 * to mark the inode_operations as being for a symlink.
1679 *
1680 * Return: the symlink target
1681 */
Al Viro6b255392015-11-17 10:20:54 -05001682const char *simple_get_link(struct dentry *dentry, struct inode *inode,
Al Virofceef392015-12-29 15:58:39 -05001683 struct delayed_call *done)
Al Viro61ba64f2015-05-02 09:54:06 -04001684{
Al Viro6b255392015-11-17 10:20:54 -05001685 return inode->i_link;
Al Viro61ba64f2015-05-02 09:54:06 -04001686}
Al Viro6b255392015-11-17 10:20:54 -05001687EXPORT_SYMBOL(simple_get_link);
Al Viro61ba64f2015-05-02 09:54:06 -04001688
1689const struct inode_operations simple_symlink_inode_operations = {
Al Viro6b255392015-11-17 10:20:54 -05001690 .get_link = simple_get_link,
Al Viro61ba64f2015-05-02 09:54:06 -04001691};
1692EXPORT_SYMBOL(simple_symlink_inode_operations);
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001693
1694/*
1695 * Operations for a permanently empty directory.
1696 */
1697static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1698{
1699 return ERR_PTR(-ENOENT);
1700}
1701
Christian Braunerb74d24f2023-01-13 12:49:12 +01001702static int empty_dir_getattr(struct mnt_idmap *idmap,
Christian Brauner549c7292021-01-21 14:19:43 +01001703 const struct path *path, struct kstat *stat,
David Howellsa528d352017-01-31 16:46:22 +00001704 u32 request_mask, unsigned int query_flags)
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001705{
David Howellsa528d352017-01-31 16:46:22 +00001706 struct inode *inode = d_inode(path->dentry);
Jeff Layton0d72b9282023-08-07 15:38:33 -04001707 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001708 return 0;
1709}
1710
Christian Braunerc1632a02023-01-13 12:49:11 +01001711static int empty_dir_setattr(struct mnt_idmap *idmap,
Christian Brauner549c7292021-01-21 14:19:43 +01001712 struct dentry *dentry, struct iattr *attr)
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001713{
1714 return -EPERM;
1715}
1716
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001717static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1718{
1719 return -EOPNOTSUPP;
1720}
1721
1722static const struct inode_operations empty_dir_inode_operations = {
1723 .lookup = empty_dir_lookup,
1724 .permission = generic_permission,
1725 .setattr = empty_dir_setattr,
1726 .getattr = empty_dir_getattr,
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001727 .listxattr = empty_dir_listxattr,
1728};
1729
1730static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1731{
1732 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1733 return generic_file_llseek_size(file, offset, whence, 2, 2);
1734}
1735
1736static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1737{
1738 dir_emit_dots(file, ctx);
1739 return 0;
1740}
1741
1742static const struct file_operations empty_dir_operations = {
1743 .llseek = empty_dir_llseek,
1744 .read = generic_read_dir,
Al Viroc51da202016-04-30 22:37:34 -04001745 .iterate_shared = empty_dir_readdir,
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001746 .fsync = noop_fsync,
1747};
1748
1749
1750void make_empty_dir_inode(struct inode *inode)
1751{
1752 set_nlink(inode, 2);
1753 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1754 inode->i_uid = GLOBAL_ROOT_UID;
1755 inode->i_gid = GLOBAL_ROOT_GID;
1756 inode->i_rdev = 0;
Eric W. Biederman4b75de862015-08-12 15:00:12 -05001757 inode->i_size = 0;
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001758 inode->i_blkbits = PAGE_SHIFT;
1759 inode->i_blocks = 0;
1760
1761 inode->i_op = &empty_dir_inode_operations;
Andreas Gruenbacherf5c24432016-09-29 17:48:41 +02001762 inode->i_opflags &= ~IOP_XATTR;
Eric W. Biedermanfbabfd02015-05-09 15:54:49 -05001763 inode->i_fop = &empty_dir_operations;
1764}
1765
1766bool is_empty_dir_inode(struct inode *inode)
1767{
1768 return (inode->i_fop == &empty_dir_operations) &&
1769 (inode->i_op == &empty_dir_inode_operations);
1770}
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001771
Christoph Hellwig5298d4b2022-01-18 07:56:14 +01001772#if IS_ENABLED(CONFIG_UNICODE)
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001773/**
1774 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1775 * @dentry: dentry whose name we are checking against
1776 * @len: len of name of dentry
1777 * @str: str pointer to name of dentry
1778 * @name: Name to compare against
1779 *
1780 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1781 */
Eric Biggers794c43f2020-12-28 15:25:29 -08001782static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1783 const char *str, const struct qstr *name)
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001784{
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001785 const struct dentry *parent;
1786 const struct inode *dir;
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001787 char strbuf[DNAME_INLINE_LEN];
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001788 struct qstr qstr;
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001789
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001790 /*
1791 * Attempt a case-sensitive match first. It is cheaper and
1792 * should cover most lookups, including all the sane
1793 * applications that expect a case-sensitive filesystem.
1794 *
1795 * This comparison is safe under RCU because the caller
1796 * guarantees the consistency between str and len. See
1797 * __d_lookup_rcu_op_compare() for details.
1798 */
1799 if (len == name->len && !memcmp(str, name->name, len))
1800 return 0;
1801
1802 parent = READ_ONCE(dentry->d_parent);
1803 dir = READ_ONCE(parent->d_inode);
Eric Biggersaf494af2023-08-14 11:29:03 -07001804 if (!dir || !IS_CASEFOLDED(dir))
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001805 return 1;
1806
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001807 /*
1808 * If the dentry name is stored in-line, then it may be concurrently
1809 * modified by a rename. If this happens, the VFS will eventually retry
1810 * the lookup, so it doesn't matter what ->d_compare() returns.
1811 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1812 * string. Therefore, we have to copy the name into a temporary buffer.
1813 */
1814 if (len <= DNAME_INLINE_LEN - 1) {
1815 memcpy(strbuf, str, len);
1816 strbuf[len] = 0;
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001817 str = strbuf;
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001818 /* prevent compiler from optimizing out the temporary buffer */
1819 barrier();
1820 }
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001821 qstr.len = len;
1822 qstr.name = str;
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001823
Gabriel Krisman Bertazi0906fbb2024-01-24 15:13:40 -03001824 return utf8_strncasecmp(dentry->d_sb->s_encoding, name, &qstr);
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001825}
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001826
1827/**
1828 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1829 * @dentry: dentry of the parent directory
1830 * @str: qstr of name whose hash we should fill in
1831 *
1832 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1833 */
Eric Biggers794c43f2020-12-28 15:25:29 -08001834static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001835{
1836 const struct inode *dir = READ_ONCE(dentry->d_inode);
1837 struct super_block *sb = dentry->d_sb;
1838 const struct unicode_map *um = sb->s_encoding;
Li zemingbae8bc42024-02-20 14:20:30 +08001839 int ret;
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001840
Eric Biggersaf494af2023-08-14 11:29:03 -07001841 if (!dir || !IS_CASEFOLDED(dir))
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001842 return 0;
1843
1844 ret = utf8_casefold_hash(um, dentry, str);
1845 if (ret < 0 && sb_has_strict_encoding(sb))
1846 return -EINVAL;
1847 return 0;
1848}
Daniel Rosenberg608af702020-11-19 06:09:02 +00001849
1850static const struct dentry_operations generic_ci_dentry_ops = {
1851 .d_hash = generic_ci_d_hash,
1852 .d_compare = generic_ci_d_compare,
Gabriel Krisman Bertazie6ca2882024-02-21 12:14:07 -05001853#ifdef CONFIG_FS_ENCRYPTION
1854 .d_revalidate = fscrypt_d_revalidate,
1855#endif
Daniel Rosenberg608af702020-11-19 06:09:02 +00001856};
Daniel Rosenbergc8438432020-07-08 02:12:35 -07001857#endif
Daniel Rosenberg608af702020-11-19 06:09:02 +00001858
1859#ifdef CONFIG_FS_ENCRYPTION
1860static const struct dentry_operations generic_encrypted_dentry_ops = {
1861 .d_revalidate = fscrypt_d_revalidate,
1862};
1863#endif
1864
Daniel Rosenberg608af702020-11-19 06:09:02 +00001865/**
Gabriel Krisman Bertazi70dfe3f2024-02-21 12:14:08 -05001866 * generic_set_sb_d_ops - helper for choosing the set of
1867 * filesystem-wide dentry operations for the enabled features
1868 * @sb: superblock to be configured
Daniel Rosenberg608af702020-11-19 06:09:02 +00001869 *
Gabriel Krisman Bertazi70dfe3f2024-02-21 12:14:08 -05001870 * Filesystems supporting casefolding and/or fscrypt can call this
1871 * helper at mount-time to configure sb->s_d_op to best set of dentry
1872 * operations required for the enabled features. The helper must be
1873 * called after these have been configured, but before the root dentry
1874 * is created.
Daniel Rosenberg608af702020-11-19 06:09:02 +00001875 */
Gabriel Krisman Bertazi70dfe3f2024-02-21 12:14:08 -05001876void generic_set_sb_d_ops(struct super_block *sb)
Daniel Rosenberg608af702020-11-19 06:09:02 +00001877{
Christoph Hellwig5298d4b2022-01-18 07:56:14 +01001878#if IS_ENABLED(CONFIG_UNICODE)
Gabriel Krisman Bertazi70dfe3f2024-02-21 12:14:08 -05001879 if (sb->s_encoding) {
1880 sb->s_d_op = &generic_ci_dentry_ops;
Daniel Rosenberg608af702020-11-19 06:09:02 +00001881 return;
1882 }
1883#endif
1884#ifdef CONFIG_FS_ENCRYPTION
Gabriel Krisman Bertazi70dfe3f2024-02-21 12:14:08 -05001885 if (sb->s_cop) {
1886 sb->s_d_op = &generic_encrypted_dentry_ops;
Daniel Rosenberg608af702020-11-19 06:09:02 +00001887 return;
1888 }
1889#endif
1890}
Gabriel Krisman Bertazi70dfe3f2024-02-21 12:14:08 -05001891EXPORT_SYMBOL(generic_set_sb_d_ops);
Andrew Morton5ca14832022-09-09 13:57:41 -07001892
1893/**
1894 * inode_maybe_inc_iversion - increments i_version
1895 * @inode: inode with the i_version that should be updated
1896 * @force: increment the counter even if it's not necessary?
1897 *
1898 * Every time the inode is modified, the i_version field must be seen to have
1899 * changed by any observer.
1900 *
1901 * If "force" is set or the QUERIED flag is set, then ensure that we increment
1902 * the value, and clear the queried flag.
1903 *
1904 * In the common case where neither is set, then we can return "false" without
1905 * updating i_version.
1906 *
1907 * If this function returns false, and no other metadata has changed, then we
1908 * can avoid logging the metadata.
1909 */
1910bool inode_maybe_inc_iversion(struct inode *inode, bool force)
1911{
1912 u64 cur, new;
1913
1914 /*
1915 * The i_version field is not strictly ordered with any other inode
1916 * information, but the legacy inode_inc_iversion code used a spinlock
1917 * to serialize increments.
1918 *
1919 * Here, we add full memory barriers to ensure that any de-facto
1920 * ordering with other info is preserved.
1921 *
1922 * This barrier pairs with the barrier in inode_query_iversion()
1923 */
1924 smp_mb();
1925 cur = inode_peek_iversion_raw(inode);
1926 do {
1927 /* If flag is clear then we needn't do anything */
1928 if (!force && !(cur & I_VERSION_QUERIED))
1929 return false;
1930
1931 /* Since lowest bit is flag, add 2 to avoid it */
1932 new = (cur & ~I_VERSION_QUERIED) + I_VERSION_INCREMENT;
1933 } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1934 return true;
1935}
1936EXPORT_SYMBOL(inode_maybe_inc_iversion);
Jeff Laytonc5bc1b32022-09-16 09:37:51 -04001937
1938/**
1939 * inode_query_iversion - read i_version for later use
1940 * @inode: inode from which i_version should be read
1941 *
1942 * Read the inode i_version counter. This should be used by callers that wish
1943 * to store the returned i_version for later comparison. This will guarantee
1944 * that a later query of the i_version will result in a different value if
1945 * anything has changed.
1946 *
1947 * In this implementation, we fetch the current value, set the QUERIED flag and
1948 * then try to swap it into place with a cmpxchg, if it wasn't already set. If
1949 * that fails, we try again with the newly fetched value from the cmpxchg.
1950 */
1951u64 inode_query_iversion(struct inode *inode)
1952{
1953 u64 cur, new;
1954
1955 cur = inode_peek_iversion_raw(inode);
1956 do {
1957 /* If flag is already set, then no need to swap */
1958 if (cur & I_VERSION_QUERIED) {
1959 /*
1960 * This barrier (and the implicit barrier in the
1961 * cmpxchg below) pairs with the barrier in
1962 * inode_maybe_inc_iversion().
1963 */
1964 smp_mb();
1965 break;
1966 }
1967
1968 new = cur | I_VERSION_QUERIED;
1969 } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new));
1970 return cur >> I_VERSION_QUERIED_SHIFT;
1971}
1972EXPORT_SYMBOL(inode_query_iversion);
Christoph Hellwig44fff0f2023-06-01 16:59:01 +02001973
1974ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
1975 ssize_t direct_written, ssize_t buffered_written)
1976{
1977 struct address_space *mapping = iocb->ki_filp->f_mapping;
1978 loff_t pos = iocb->ki_pos - buffered_written;
1979 loff_t end = iocb->ki_pos - 1;
1980 int err;
1981
1982 /*
1983 * If the buffered write fallback returned an error, we want to return
1984 * the number of bytes which were written by direct I/O, or the error
1985 * code if that was zero.
1986 *
1987 * Note that this differs from normal direct-io semantics, which will
1988 * return -EFOO even if some bytes were written.
1989 */
1990 if (unlikely(buffered_written < 0)) {
1991 if (direct_written)
1992 return direct_written;
1993 return buffered_written;
1994 }
1995
1996 /*
1997 * We need to ensure that the page cache pages are written to disk and
1998 * invalidated to preserve the expected O_DIRECT semantics.
1999 */
2000 err = filemap_write_and_wait_range(mapping, pos, end);
2001 if (err < 0) {
2002 /*
2003 * We don't know how much we wrote, so just return the number of
2004 * bytes which were direct-written
2005 */
Al Viro82874742023-09-13 18:28:15 +02002006 iocb->ki_pos -= buffered_written;
Christoph Hellwig44fff0f2023-06-01 16:59:01 +02002007 if (direct_written)
2008 return direct_written;
2009 return err;
2010 }
2011 invalidate_mapping_pages(mapping, pos >> PAGE_SHIFT, end >> PAGE_SHIFT);
2012 return direct_written + buffered_written;
2013}
2014EXPORT_SYMBOL_GPL(direct_write_fallback);
Jeff Layton077c2122023-10-04 14:52:37 -04002015
2016/**
2017 * simple_inode_init_ts - initialize the timestamps for a new inode
2018 * @inode: inode to be initialized
2019 *
2020 * When a new inode is created, most filesystems set the timestamps to the
2021 * current time. Add a helper to do this.
2022 */
2023struct timespec64 simple_inode_init_ts(struct inode *inode)
2024{
2025 struct timespec64 ts = inode_set_ctime_current(inode);
2026
2027 inode_set_atime_to_ts(inode, ts);
2028 inode_set_mtime_to_ts(inode, ts);
2029 return ts;
2030}
2031EXPORT_SYMBOL(simple_inode_init_ts);
Christian Brauner07fd7c32024-02-18 14:50:13 +01002032
2033static inline struct dentry *get_stashed_dentry(struct dentry *stashed)
2034{
2035 struct dentry *dentry;
2036
2037 guard(rcu)();
2038 dentry = READ_ONCE(stashed);
2039 if (!dentry)
2040 return NULL;
2041 if (!lockref_get_not_dead(&dentry->d_lockref))
2042 return NULL;
2043 return dentry;
2044}
2045
Christian Brauner2558e3b2024-02-21 09:59:51 +01002046static struct dentry *prepare_anon_dentry(struct dentry **stashed,
Christian Brauner159a0d92024-02-18 14:52:24 +01002047 struct super_block *sb,
Christian Brauner159a0d92024-02-18 14:52:24 +01002048 void *data)
Christian Brauner07fd7c32024-02-18 14:50:13 +01002049{
2050 struct dentry *dentry;
2051 struct inode *inode;
Christian Braunere9c52632024-03-01 10:26:03 +01002052 const struct stashed_operations *sops = sb->s_fs_info;
Christian Brauner9d9539d2024-03-12 10:39:44 +01002053 int ret;
Christian Brauner07fd7c32024-02-18 14:50:13 +01002054
2055 inode = new_inode_pseudo(sb);
2056 if (!inode) {
Christian Brauner9d9539d2024-03-12 10:39:44 +01002057 sops->put_data(data);
Christian Brauner07fd7c32024-02-18 14:50:13 +01002058 return ERR_PTR(-ENOMEM);
2059 }
2060
Christian Brauner07fd7c32024-02-18 14:50:13 +01002061 inode->i_flags |= S_IMMUTABLE;
Christian Braunere9c52632024-03-01 10:26:03 +01002062 inode->i_mode = S_IFREG;
Christian Brauner07fd7c32024-02-18 14:50:13 +01002063 simple_inode_init_ts(inode);
Christian Brauner9d9539d2024-03-12 10:39:44 +01002064
2065 ret = sops->init_inode(inode, data);
2066 if (ret < 0) {
2067 iput(inode);
2068 return ERR_PTR(ret);
2069 }
Christian Braunere9c52632024-03-01 10:26:03 +01002070
2071 /* Notice when this is changed. */
2072 WARN_ON_ONCE(!S_ISREG(inode->i_mode));
2073 WARN_ON_ONCE(!IS_IMMUTABLE(inode));
Christian Brauner07fd7c32024-02-18 14:50:13 +01002074
Christian Brauner9d9539d2024-03-12 10:39:44 +01002075 dentry = d_alloc_anon(sb);
2076 if (!dentry) {
2077 iput(inode);
2078 return ERR_PTR(-ENOMEM);
2079 }
2080
Christian Brauner2558e3b2024-02-21 09:59:51 +01002081 /* Store address of location where dentry's supposed to be stashed. */
2082 dentry->d_fsdata = stashed;
2083
Christian Brauner07fd7c32024-02-18 14:50:13 +01002084 /* @data is now owned by the fs */
2085 d_instantiate(dentry, inode);
Christian Brauner07fd7c32024-02-18 14:50:13 +01002086 return dentry;
2087}
2088
Christian Brauner159a0d92024-02-18 14:52:24 +01002089static struct dentry *stash_dentry(struct dentry **stashed,
2090 struct dentry *dentry)
2091{
2092 guard(rcu)();
2093 for (;;) {
2094 struct dentry *old;
2095
2096 /* Assume any old dentry was cleared out. */
2097 old = cmpxchg(stashed, NULL, dentry);
2098 if (likely(!old))
2099 return dentry;
2100
2101 /* Check if somebody else installed a reusable dentry. */
2102 if (lockref_get_not_dead(&old->d_lockref))
2103 return old;
2104
2105 /* There's an old dead dentry there, try to take it over. */
2106 if (likely(try_cmpxchg(stashed, &old, dentry)))
2107 return dentry;
2108 }
2109}
2110
Christian Brauner07fd7c32024-02-18 14:50:13 +01002111/**
2112 * path_from_stashed - create path from stashed or new dentry
2113 * @stashed: where to retrieve or stash dentry
Christian Brauner07fd7c32024-02-18 14:50:13 +01002114 * @mnt: mnt of the filesystems to use
Christian Brauner07fd7c32024-02-18 14:50:13 +01002115 * @data: data to store in inode->i_private
2116 * @path: path to create
2117 *
2118 * The function tries to retrieve a stashed dentry from @stashed. If the dentry
2119 * is still valid then it will be reused. If the dentry isn't able the function
Christian Brauner159a0d92024-02-18 14:52:24 +01002120 * will allocate a new dentry and inode. It will then check again whether it
2121 * can reuse an existing dentry in case one has been added in the meantime or
2122 * update @stashed with the newly added dentry.
Christian Brauner07fd7c32024-02-18 14:50:13 +01002123 *
2124 * Special-purpose helper for nsfs and pidfs.
2125 *
Christian Braunere9c52632024-03-01 10:26:03 +01002126 * Return: On success zero and on failure a negative error is returned.
Christian Brauner07fd7c32024-02-18 14:50:13 +01002127 */
Christian Brauner9d9539d2024-03-12 10:39:44 +01002128int path_from_stashed(struct dentry **stashed, struct vfsmount *mnt, void *data,
2129 struct path *path)
Christian Brauner07fd7c32024-02-18 14:50:13 +01002130{
2131 struct dentry *dentry;
Christian Braunere9c52632024-03-01 10:26:03 +01002132 const struct stashed_operations *sops = mnt->mnt_sb->s_fs_info;
Christian Brauner07fd7c32024-02-18 14:50:13 +01002133
Christian Brauner159a0d92024-02-18 14:52:24 +01002134 /* See if dentry can be reused. */
2135 path->dentry = get_stashed_dentry(*stashed);
Christian Braunere9c52632024-03-01 10:26:03 +01002136 if (path->dentry) {
2137 sops->put_data(data);
Christian Brauner07fd7c32024-02-18 14:50:13 +01002138 goto out_path;
Christian Braunere9c52632024-03-01 10:26:03 +01002139 }
Christian Brauner07fd7c32024-02-18 14:50:13 +01002140
Christian Brauner159a0d92024-02-18 14:52:24 +01002141 /* Allocate a new dentry. */
Christian Brauner9d9539d2024-03-12 10:39:44 +01002142 dentry = prepare_anon_dentry(stashed, mnt->mnt_sb, data);
2143 if (IS_ERR(dentry))
Christian Brauner07fd7c32024-02-18 14:50:13 +01002144 return PTR_ERR(dentry);
Christian Brauner159a0d92024-02-18 14:52:24 +01002145
2146 /* Added a new dentry. @data is now owned by the filesystem. */
2147 path->dentry = stash_dentry(stashed, dentry);
2148 if (path->dentry != dentry)
2149 dput(dentry);
Christian Brauner07fd7c32024-02-18 14:50:13 +01002150
2151out_path:
Christian Brauner2558e3b2024-02-21 09:59:51 +01002152 WARN_ON_ONCE(path->dentry->d_fsdata != stashed);
2153 WARN_ON_ONCE(d_inode(path->dentry)->i_private != data);
Christian Brauner07fd7c32024-02-18 14:50:13 +01002154 path->mnt = mntget(mnt);
Christian Braunere9c52632024-03-01 10:26:03 +01002155 return 0;
Christian Brauner07fd7c32024-02-18 14:50:13 +01002156}
Christian Brauner2558e3b2024-02-21 09:59:51 +01002157
2158void stashed_dentry_prune(struct dentry *dentry)
2159{
2160 struct dentry **stashed = dentry->d_fsdata;
2161 struct inode *inode = d_inode(dentry);
2162
2163 if (WARN_ON_ONCE(!stashed))
2164 return;
2165
2166 if (!inode)
2167 return;
2168
2169 /*
2170 * Only replace our own @dentry as someone else might've
2171 * already cleared out @dentry and stashed their own
2172 * dentry in there.
2173 */
2174 cmpxchg(stashed, dentry, NULL);
2175}