Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
[platform/kernel/linux-starfive.git] / fs / ntfs3 / inode.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *
4  * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
5  *
6  */
7
8 #include <linux/buffer_head.h>
9 #include <linux/fs.h>
10 #include <linux/mpage.h>
11 #include <linux/namei.h>
12 #include <linux/nls.h>
13 #include <linux/uio.h>
14 #include <linux/writeback.h>
15
16 #include "debug.h"
17 #include "ntfs.h"
18 #include "ntfs_fs.h"
19
20 /*
21  * ntfs_read_mft - Read record and parses MFT.
22  */
23 static struct inode *ntfs_read_mft(struct inode *inode,
24                                    const struct cpu_str *name,
25                                    const struct MFT_REF *ref)
26 {
27         int err = 0;
28         struct ntfs_inode *ni = ntfs_i(inode);
29         struct super_block *sb = inode->i_sb;
30         struct ntfs_sb_info *sbi = sb->s_fs_info;
31         mode_t mode = 0;
32         struct ATTR_STD_INFO5 *std5 = NULL;
33         struct ATTR_LIST_ENTRY *le;
34         struct ATTRIB *attr;
35         bool is_match = false;
36         bool is_root = false;
37         bool is_dir;
38         unsigned long ino = inode->i_ino;
39         u32 rp_fa = 0, asize, t32;
40         u16 roff, rsize, names = 0;
41         const struct ATTR_FILE_NAME *fname = NULL;
42         const struct INDEX_ROOT *root;
43         struct REPARSE_DATA_BUFFER rp; // 0x18 bytes
44         u64 t64;
45         struct MFT_REC *rec;
46         struct runs_tree *run;
47
48         inode->i_op = NULL;
49         /* Setup 'uid' and 'gid' */
50         inode->i_uid = sbi->options->fs_uid;
51         inode->i_gid = sbi->options->fs_gid;
52
53         err = mi_init(&ni->mi, sbi, ino);
54         if (err)
55                 goto out;
56
57         if (!sbi->mft.ni && ino == MFT_REC_MFT && !sb->s_root) {
58                 t64 = sbi->mft.lbo >> sbi->cluster_bits;
59                 t32 = bytes_to_cluster(sbi, MFT_REC_VOL * sbi->record_size);
60                 sbi->mft.ni = ni;
61                 init_rwsem(&ni->file.run_lock);
62
63                 if (!run_add_entry(&ni->file.run, 0, t64, t32, true)) {
64                         err = -ENOMEM;
65                         goto out;
66                 }
67         }
68
69         err = mi_read(&ni->mi, ino == MFT_REC_MFT);
70
71         if (err)
72                 goto out;
73
74         rec = ni->mi.mrec;
75
76         if (sbi->flags & NTFS_FLAGS_LOG_REPLAYING) {
77                 ;
78         } else if (ref->seq != rec->seq) {
79                 err = -EINVAL;
80                 ntfs_err(sb, "MFT: r=%lx, expect seq=%x instead of %x!", ino,
81                          le16_to_cpu(ref->seq), le16_to_cpu(rec->seq));
82                 goto out;
83         } else if (!is_rec_inuse(rec)) {
84                 err = -EINVAL;
85                 ntfs_err(sb, "Inode r=%x is not in use!", (u32)ino);
86                 goto out;
87         }
88
89         if (le32_to_cpu(rec->total) != sbi->record_size) {
90                 /* Bad inode? */
91                 err = -EINVAL;
92                 goto out;
93         }
94
95         if (!is_rec_base(rec))
96                 goto Ok;
97
98         /* Record should contain $I30 root. */
99         is_dir = rec->flags & RECORD_FLAG_DIR;
100
101         inode->i_generation = le16_to_cpu(rec->seq);
102
103         /* Enumerate all struct Attributes MFT. */
104         le = NULL;
105         attr = NULL;
106
107         /*
108          * To reduce tab pressure use goto instead of
109          * while( (attr = ni_enum_attr_ex(ni, attr, &le, NULL) ))
110          */
111 next_attr:
112         run = NULL;
113         err = -EINVAL;
114         attr = ni_enum_attr_ex(ni, attr, &le, NULL);
115         if (!attr)
116                 goto end_enum;
117
118         if (le && le->vcn) {
119                 /* This is non primary attribute segment. Ignore if not MFT. */
120                 if (ino != MFT_REC_MFT || attr->type != ATTR_DATA)
121                         goto next_attr;
122
123                 run = &ni->file.run;
124                 asize = le32_to_cpu(attr->size);
125                 goto attr_unpack_run;
126         }
127
128         roff = attr->non_res ? 0 : le16_to_cpu(attr->res.data_off);
129         rsize = attr->non_res ? 0 : le32_to_cpu(attr->res.data_size);
130         asize = le32_to_cpu(attr->size);
131
132         switch (attr->type) {
133         case ATTR_STD:
134                 if (attr->non_res ||
135                     asize < sizeof(struct ATTR_STD_INFO) + roff ||
136                     rsize < sizeof(struct ATTR_STD_INFO))
137                         goto out;
138
139                 if (std5)
140                         goto next_attr;
141
142                 std5 = Add2Ptr(attr, roff);
143
144 #ifdef STATX_BTIME
145                 nt2kernel(std5->cr_time, &ni->i_crtime);
146 #endif
147                 nt2kernel(std5->a_time, &inode->i_atime);
148                 nt2kernel(std5->c_time, &inode->i_ctime);
149                 nt2kernel(std5->m_time, &inode->i_mtime);
150
151                 ni->std_fa = std5->fa;
152
153                 if (asize >= sizeof(struct ATTR_STD_INFO5) + roff &&
154                     rsize >= sizeof(struct ATTR_STD_INFO5))
155                         ni->std_security_id = std5->security_id;
156                 goto next_attr;
157
158         case ATTR_LIST:
159                 if (attr->name_len || le || ino == MFT_REC_LOG)
160                         goto out;
161
162                 err = ntfs_load_attr_list(ni, attr);
163                 if (err)
164                         goto out;
165
166                 le = NULL;
167                 attr = NULL;
168                 goto next_attr;
169
170         case ATTR_NAME:
171                 if (attr->non_res || asize < SIZEOF_ATTRIBUTE_FILENAME + roff ||
172                     rsize < SIZEOF_ATTRIBUTE_FILENAME)
173                         goto out;
174
175                 fname = Add2Ptr(attr, roff);
176                 if (fname->type == FILE_NAME_DOS)
177                         goto next_attr;
178
179                 names += 1;
180                 if (name && name->len == fname->name_len &&
181                     !ntfs_cmp_names_cpu(name, (struct le_str *)&fname->name_len,
182                                         NULL, false))
183                         is_match = true;
184
185                 goto next_attr;
186
187         case ATTR_DATA:
188                 if (is_dir) {
189                         /* Ignore data attribute in dir record. */
190                         goto next_attr;
191                 }
192
193                 if (ino == MFT_REC_BADCLUST && !attr->non_res)
194                         goto next_attr;
195
196                 if (attr->name_len &&
197                     ((ino != MFT_REC_BADCLUST || !attr->non_res ||
198                       attr->name_len != ARRAY_SIZE(BAD_NAME) ||
199                       memcmp(attr_name(attr), BAD_NAME, sizeof(BAD_NAME))) &&
200                      (ino != MFT_REC_SECURE || !attr->non_res ||
201                       attr->name_len != ARRAY_SIZE(SDS_NAME) ||
202                       memcmp(attr_name(attr), SDS_NAME, sizeof(SDS_NAME))))) {
203                         /* File contains stream attribute. Ignore it. */
204                         goto next_attr;
205                 }
206
207                 if (is_attr_sparsed(attr))
208                         ni->std_fa |= FILE_ATTRIBUTE_SPARSE_FILE;
209                 else
210                         ni->std_fa &= ~FILE_ATTRIBUTE_SPARSE_FILE;
211
212                 if (is_attr_compressed(attr))
213                         ni->std_fa |= FILE_ATTRIBUTE_COMPRESSED;
214                 else
215                         ni->std_fa &= ~FILE_ATTRIBUTE_COMPRESSED;
216
217                 if (is_attr_encrypted(attr))
218                         ni->std_fa |= FILE_ATTRIBUTE_ENCRYPTED;
219                 else
220                         ni->std_fa &= ~FILE_ATTRIBUTE_ENCRYPTED;
221
222                 if (!attr->non_res) {
223                         ni->i_valid = inode->i_size = rsize;
224                         inode_set_bytes(inode, rsize);
225                 }
226
227                 mode = S_IFREG | (0777 & sbi->options->fs_fmask_inv);
228
229                 if (!attr->non_res) {
230                         ni->ni_flags |= NI_FLAG_RESIDENT;
231                         goto next_attr;
232                 }
233
234                 inode_set_bytes(inode, attr_ondisk_size(attr));
235
236                 ni->i_valid = le64_to_cpu(attr->nres.valid_size);
237                 inode->i_size = le64_to_cpu(attr->nres.data_size);
238                 if (!attr->nres.alloc_size)
239                         goto next_attr;
240
241                 run = ino == MFT_REC_BITMAP ? &sbi->used.bitmap.run
242                                             : &ni->file.run;
243                 break;
244
245         case ATTR_ROOT:
246                 if (attr->non_res)
247                         goto out;
248
249                 root = Add2Ptr(attr, roff);
250                 is_root = true;
251
252                 if (attr->name_len != ARRAY_SIZE(I30_NAME) ||
253                     memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
254                         goto next_attr;
255
256                 if (root->type != ATTR_NAME ||
257                     root->rule != NTFS_COLLATION_TYPE_FILENAME)
258                         goto out;
259
260                 if (!is_dir)
261                         goto next_attr;
262
263                 ni->ni_flags |= NI_FLAG_DIR;
264
265                 err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
266                 if (err)
267                         goto out;
268
269                 mode = sb->s_root
270                                ? (S_IFDIR | (0777 & sbi->options->fs_dmask_inv))
271                                : (S_IFDIR | 0777);
272                 goto next_attr;
273
274         case ATTR_ALLOC:
275                 if (!is_root || attr->name_len != ARRAY_SIZE(I30_NAME) ||
276                     memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
277                         goto next_attr;
278
279                 inode->i_size = le64_to_cpu(attr->nres.data_size);
280                 ni->i_valid = le64_to_cpu(attr->nres.valid_size);
281                 inode_set_bytes(inode, le64_to_cpu(attr->nres.alloc_size));
282
283                 run = &ni->dir.alloc_run;
284                 break;
285
286         case ATTR_BITMAP:
287                 if (ino == MFT_REC_MFT) {
288                         if (!attr->non_res)
289                                 goto out;
290 #ifndef CONFIG_NTFS3_64BIT_CLUSTER
291                         /* 0x20000000 = 2^32 / 8 */
292                         if (le64_to_cpu(attr->nres.alloc_size) >= 0x20000000)
293                                 goto out;
294 #endif
295                         run = &sbi->mft.bitmap.run;
296                         break;
297                 } else if (is_dir && attr->name_len == ARRAY_SIZE(I30_NAME) &&
298                            !memcmp(attr_name(attr), I30_NAME,
299                                    sizeof(I30_NAME)) &&
300                            attr->non_res) {
301                         run = &ni->dir.bitmap_run;
302                         break;
303                 }
304                 goto next_attr;
305
306         case ATTR_REPARSE:
307                 if (attr->name_len)
308                         goto next_attr;
309
310                 rp_fa = ni_parse_reparse(ni, attr, &rp);
311                 switch (rp_fa) {
312                 case REPARSE_LINK:
313                         /*
314                          * Normal symlink.
315                          * Assume one unicode symbol == one utf8.
316                          */
317                         inode->i_size = le16_to_cpu(rp.SymbolicLinkReparseBuffer
318                                                             .PrintNameLength) /
319                                         sizeof(u16);
320
321                         ni->i_valid = inode->i_size;
322
323                         /* Clear directory bit. */
324                         if (ni->ni_flags & NI_FLAG_DIR) {
325                                 indx_clear(&ni->dir);
326                                 memset(&ni->dir, 0, sizeof(ni->dir));
327                                 ni->ni_flags &= ~NI_FLAG_DIR;
328                         } else {
329                                 run_close(&ni->file.run);
330                         }
331                         mode = S_IFLNK | 0777;
332                         is_dir = false;
333                         if (attr->non_res) {
334                                 run = &ni->file.run;
335                                 goto attr_unpack_run; // Double break.
336                         }
337                         break;
338
339                 case REPARSE_COMPRESSED:
340                         break;
341
342                 case REPARSE_DEDUPLICATED:
343                         break;
344                 }
345                 goto next_attr;
346
347         case ATTR_EA_INFO:
348                 if (!attr->name_len &&
349                     resident_data_ex(attr, sizeof(struct EA_INFO))) {
350                         ni->ni_flags |= NI_FLAG_EA;
351                         /*
352                          * ntfs_get_wsl_perm updates inode->i_uid, inode->i_gid, inode->i_mode
353                          */
354                         inode->i_mode = mode;
355                         ntfs_get_wsl_perm(inode);
356                         mode = inode->i_mode;
357                 }
358                 goto next_attr;
359
360         default:
361                 goto next_attr;
362         }
363
364 attr_unpack_run:
365         roff = le16_to_cpu(attr->nres.run_off);
366
367         t64 = le64_to_cpu(attr->nres.svcn);
368         err = run_unpack_ex(run, sbi, ino, t64, le64_to_cpu(attr->nres.evcn),
369                             t64, Add2Ptr(attr, roff), asize - roff);
370         if (err < 0)
371                 goto out;
372         err = 0;
373         goto next_attr;
374
375 end_enum:
376
377         if (!std5)
378                 goto out;
379
380         if (!is_match && name) {
381                 /* Reuse rec as buffer for ascii name. */
382                 err = -ENOENT;
383                 goto out;
384         }
385
386         if (std5->fa & FILE_ATTRIBUTE_READONLY)
387                 mode &= ~0222;
388
389         if (!names) {
390                 err = -EINVAL;
391                 goto out;
392         }
393
394         if (names != le16_to_cpu(rec->hard_links)) {
395                 /* Correct minor error on the fly. Do not mark inode as dirty. */
396                 rec->hard_links = cpu_to_le16(names);
397                 ni->mi.dirty = true;
398         }
399
400         set_nlink(inode, names);
401
402         if (S_ISDIR(mode)) {
403                 ni->std_fa |= FILE_ATTRIBUTE_DIRECTORY;
404
405                 /*
406                  * Dot and dot-dot should be included in count but was not
407                  * included in enumeration.
408                  * Usually a hard links to directories are disabled.
409                  */
410                 inode->i_op = &ntfs_dir_inode_operations;
411                 inode->i_fop = &ntfs_dir_operations;
412                 ni->i_valid = 0;
413         } else if (S_ISLNK(mode)) {
414                 ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
415                 inode->i_op = &ntfs_link_inode_operations;
416                 inode->i_fop = NULL;
417                 inode_nohighmem(inode);
418         } else if (S_ISREG(mode)) {
419                 ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
420                 inode->i_op = &ntfs_file_inode_operations;
421                 inode->i_fop = &ntfs_file_operations;
422                 inode->i_mapping->a_ops =
423                         is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
424                 if (ino != MFT_REC_MFT)
425                         init_rwsem(&ni->file.run_lock);
426         } else if (S_ISCHR(mode) || S_ISBLK(mode) || S_ISFIFO(mode) ||
427                    S_ISSOCK(mode)) {
428                 inode->i_op = &ntfs_special_inode_operations;
429                 init_special_inode(inode, mode, inode->i_rdev);
430         } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
431                    fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
432                 /* Records in $Extend are not a files or general directories. */
433                 inode->i_op = &ntfs_file_inode_operations;
434         } else {
435                 err = -EINVAL;
436                 goto out;
437         }
438
439         if ((sbi->options->sys_immutable &&
440              (std5->fa & FILE_ATTRIBUTE_SYSTEM)) &&
441             !S_ISFIFO(mode) && !S_ISSOCK(mode) && !S_ISLNK(mode)) {
442                 inode->i_flags |= S_IMMUTABLE;
443         } else {
444                 inode->i_flags &= ~S_IMMUTABLE;
445         }
446
447         inode->i_mode = mode;
448         if (!(ni->ni_flags & NI_FLAG_EA)) {
449                 /* If no xattr then no security (stored in xattr). */
450                 inode->i_flags |= S_NOSEC;
451         }
452
453 Ok:
454         if (ino == MFT_REC_MFT && !sb->s_root)
455                 sbi->mft.ni = NULL;
456
457         unlock_new_inode(inode);
458
459         return inode;
460
461 out:
462         if (ino == MFT_REC_MFT && !sb->s_root)
463                 sbi->mft.ni = NULL;
464
465         iget_failed(inode);
466         return ERR_PTR(err);
467 }
468
469 /*
470  * ntfs_test_inode
471  *
472  * Return: 1 if match.
473  */
474 static int ntfs_test_inode(struct inode *inode, void *data)
475 {
476         struct MFT_REF *ref = data;
477
478         return ino_get(ref) == inode->i_ino;
479 }
480
481 static int ntfs_set_inode(struct inode *inode, void *data)
482 {
483         const struct MFT_REF *ref = data;
484
485         inode->i_ino = ino_get(ref);
486         return 0;
487 }
488
489 struct inode *ntfs_iget5(struct super_block *sb, const struct MFT_REF *ref,
490                          const struct cpu_str *name)
491 {
492         struct inode *inode;
493
494         inode = iget5_locked(sb, ino_get(ref), ntfs_test_inode, ntfs_set_inode,
495                              (void *)ref);
496         if (unlikely(!inode))
497                 return ERR_PTR(-ENOMEM);
498
499         /* If this is a freshly allocated inode, need to read it now. */
500         if (inode->i_state & I_NEW)
501                 inode = ntfs_read_mft(inode, name, ref);
502         else if (ref->seq != ntfs_i(inode)->mi.mrec->seq) {
503                 /* Inode overlaps? */
504                 _ntfs_bad_inode(inode);
505         }
506
507         return inode;
508 }
509
510 enum get_block_ctx {
511         GET_BLOCK_GENERAL = 0,
512         GET_BLOCK_WRITE_BEGIN = 1,
513         GET_BLOCK_DIRECT_IO_R = 2,
514         GET_BLOCK_DIRECT_IO_W = 3,
515         GET_BLOCK_BMAP = 4,
516 };
517
518 static noinline int ntfs_get_block_vbo(struct inode *inode, u64 vbo,
519                                        struct buffer_head *bh, int create,
520                                        enum get_block_ctx ctx)
521 {
522         struct super_block *sb = inode->i_sb;
523         struct ntfs_sb_info *sbi = sb->s_fs_info;
524         struct ntfs_inode *ni = ntfs_i(inode);
525         struct page *page = bh->b_page;
526         u8 cluster_bits = sbi->cluster_bits;
527         u32 block_size = sb->s_blocksize;
528         u64 bytes, lbo, valid;
529         u32 off;
530         int err;
531         CLST vcn, lcn, len;
532         bool new;
533
534         /* Clear previous state. */
535         clear_buffer_new(bh);
536         clear_buffer_uptodate(bh);
537
538         /* Direct write uses 'create=0'. */
539         if (!create && vbo >= ni->i_valid) {
540                 /* Out of valid. */
541                 return 0;
542         }
543
544         if (vbo >= inode->i_size) {
545                 /* Out of size. */
546                 return 0;
547         }
548
549         if (is_resident(ni)) {
550                 ni_lock(ni);
551                 err = attr_data_read_resident(ni, page);
552                 ni_unlock(ni);
553
554                 if (!err)
555                         set_buffer_uptodate(bh);
556                 bh->b_size = block_size;
557                 return err;
558         }
559
560         vcn = vbo >> cluster_bits;
561         off = vbo & sbi->cluster_mask;
562         new = false;
563
564         err = attr_data_get_block(ni, vcn, 1, &lcn, &len, create ? &new : NULL);
565         if (err)
566                 goto out;
567
568         if (!len)
569                 return 0;
570
571         bytes = ((u64)len << cluster_bits) - off;
572
573         if (lcn == SPARSE_LCN) {
574                 if (!create) {
575                         if (bh->b_size > bytes)
576                                 bh->b_size = bytes;
577                         return 0;
578                 }
579                 WARN_ON(1);
580         }
581
582         if (new) {
583                 set_buffer_new(bh);
584                 if ((len << cluster_bits) > block_size)
585                         ntfs_sparse_cluster(inode, page, vcn, len);
586         }
587
588         lbo = ((u64)lcn << cluster_bits) + off;
589
590         set_buffer_mapped(bh);
591         bh->b_bdev = sb->s_bdev;
592         bh->b_blocknr = lbo >> sb->s_blocksize_bits;
593
594         valid = ni->i_valid;
595
596         if (ctx == GET_BLOCK_DIRECT_IO_W) {
597                 /* ntfs_direct_IO will update ni->i_valid. */
598                 if (vbo >= valid)
599                         set_buffer_new(bh);
600         } else if (create) {
601                 /* Normal write. */
602                 if (bytes > bh->b_size)
603                         bytes = bh->b_size;
604
605                 if (vbo >= valid)
606                         set_buffer_new(bh);
607
608                 if (vbo + bytes > valid) {
609                         ni->i_valid = vbo + bytes;
610                         mark_inode_dirty(inode);
611                 }
612         } else if (vbo >= valid) {
613                 /* Read out of valid data. */
614                 /* Should never be here 'cause already checked. */
615                 clear_buffer_mapped(bh);
616         } else if (vbo + bytes <= valid) {
617                 /* Normal read. */
618         } else if (vbo + block_size <= valid) {
619                 /* Normal short read. */
620                 bytes = block_size;
621         } else {
622                 /*
623                  * Read across valid size: vbo < valid && valid < vbo + block_size
624                  */
625                 bytes = block_size;
626
627                 if (page) {
628                         u32 voff = valid - vbo;
629
630                         bh->b_size = block_size;
631                         off = vbo & (PAGE_SIZE - 1);
632                         set_bh_page(bh, page, off);
633                         ll_rw_block(REQ_OP_READ, 1, &bh);
634                         wait_on_buffer(bh);
635                         if (!buffer_uptodate(bh)) {
636                                 err = -EIO;
637                                 goto out;
638                         }
639                         zero_user_segment(page, off + voff, off + block_size);
640                 }
641         }
642
643         if (bh->b_size > bytes)
644                 bh->b_size = bytes;
645
646 #ifndef __LP64__
647         if (ctx == GET_BLOCK_DIRECT_IO_W || ctx == GET_BLOCK_DIRECT_IO_R) {
648                 static_assert(sizeof(size_t) < sizeof(loff_t));
649                 if (bytes > 0x40000000u)
650                         bh->b_size = 0x40000000u;
651         }
652 #endif
653
654         return 0;
655
656 out:
657         return err;
658 }
659
660 int ntfs_get_block(struct inode *inode, sector_t vbn,
661                    struct buffer_head *bh_result, int create)
662 {
663         return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
664                                   bh_result, create, GET_BLOCK_GENERAL);
665 }
666
667 static int ntfs_get_block_bmap(struct inode *inode, sector_t vsn,
668                                struct buffer_head *bh_result, int create)
669 {
670         return ntfs_get_block_vbo(inode,
671                                   (u64)vsn << inode->i_sb->s_blocksize_bits,
672                                   bh_result, create, GET_BLOCK_BMAP);
673 }
674
675 static sector_t ntfs_bmap(struct address_space *mapping, sector_t block)
676 {
677         return generic_block_bmap(mapping, block, ntfs_get_block_bmap);
678 }
679
680 static int ntfs_read_folio(struct file *file, struct folio *folio)
681 {
682         struct page *page = &folio->page;
683         int err;
684         struct address_space *mapping = page->mapping;
685         struct inode *inode = mapping->host;
686         struct ntfs_inode *ni = ntfs_i(inode);
687
688         if (is_resident(ni)) {
689                 ni_lock(ni);
690                 err = attr_data_read_resident(ni, page);
691                 ni_unlock(ni);
692                 if (err != E_NTFS_NONRESIDENT) {
693                         unlock_page(page);
694                         return err;
695                 }
696         }
697
698         if (is_compressed(ni)) {
699                 ni_lock(ni);
700                 err = ni_readpage_cmpr(ni, page);
701                 ni_unlock(ni);
702                 return err;
703         }
704
705         /* Normal + sparse files. */
706         return mpage_read_folio(folio, ntfs_get_block);
707 }
708
709 static void ntfs_readahead(struct readahead_control *rac)
710 {
711         struct address_space *mapping = rac->mapping;
712         struct inode *inode = mapping->host;
713         struct ntfs_inode *ni = ntfs_i(inode);
714         u64 valid;
715         loff_t pos;
716
717         if (is_resident(ni)) {
718                 /* No readahead for resident. */
719                 return;
720         }
721
722         if (is_compressed(ni)) {
723                 /* No readahead for compressed. */
724                 return;
725         }
726
727         valid = ni->i_valid;
728         pos = readahead_pos(rac);
729
730         if (valid < i_size_read(inode) && pos <= valid &&
731             valid < pos + readahead_length(rac)) {
732                 /* Range cross 'valid'. Read it page by page. */
733                 return;
734         }
735
736         mpage_readahead(rac, ntfs_get_block);
737 }
738
739 static int ntfs_get_block_direct_IO_R(struct inode *inode, sector_t iblock,
740                                       struct buffer_head *bh_result, int create)
741 {
742         return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
743                                   bh_result, create, GET_BLOCK_DIRECT_IO_R);
744 }
745
746 static int ntfs_get_block_direct_IO_W(struct inode *inode, sector_t iblock,
747                                       struct buffer_head *bh_result, int create)
748 {
749         return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
750                                   bh_result, create, GET_BLOCK_DIRECT_IO_W);
751 }
752
753 static ssize_t ntfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
754 {
755         struct file *file = iocb->ki_filp;
756         struct address_space *mapping = file->f_mapping;
757         struct inode *inode = mapping->host;
758         struct ntfs_inode *ni = ntfs_i(inode);
759         loff_t vbo = iocb->ki_pos;
760         loff_t end;
761         int wr = iov_iter_rw(iter) & WRITE;
762         size_t iter_count = iov_iter_count(iter);
763         loff_t valid;
764         ssize_t ret;
765
766         if (is_resident(ni)) {
767                 /* Switch to buffered write. */
768                 ret = 0;
769                 goto out;
770         }
771
772         ret = blockdev_direct_IO(iocb, inode, iter,
773                                  wr ? ntfs_get_block_direct_IO_W
774                                     : ntfs_get_block_direct_IO_R);
775
776         if (ret > 0)
777                 end = vbo + ret;
778         else if (wr && ret == -EIOCBQUEUED)
779                 end = vbo + iter_count;
780         else
781                 goto out;
782
783         valid = ni->i_valid;
784         if (wr) {
785                 if (end > valid && !S_ISBLK(inode->i_mode)) {
786                         ni->i_valid = end;
787                         mark_inode_dirty(inode);
788                 }
789         } else if (vbo < valid && valid < end) {
790                 /* Fix page. */
791                 iov_iter_revert(iter, end - valid);
792                 iov_iter_zero(end - valid, iter);
793         }
794
795 out:
796         return ret;
797 }
798
799 int ntfs_set_size(struct inode *inode, u64 new_size)
800 {
801         struct super_block *sb = inode->i_sb;
802         struct ntfs_sb_info *sbi = sb->s_fs_info;
803         struct ntfs_inode *ni = ntfs_i(inode);
804         int err;
805
806         /* Check for maximum file size. */
807         if (is_sparsed(ni) || is_compressed(ni)) {
808                 if (new_size > sbi->maxbytes_sparse) {
809                         err = -EFBIG;
810                         goto out;
811                 }
812         } else if (new_size > sbi->maxbytes) {
813                 err = -EFBIG;
814                 goto out;
815         }
816
817         ni_lock(ni);
818         down_write(&ni->file.run_lock);
819
820         err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run, new_size,
821                             &ni->i_valid, true, NULL);
822
823         up_write(&ni->file.run_lock);
824         ni_unlock(ni);
825
826         mark_inode_dirty(inode);
827
828 out:
829         return err;
830 }
831
832 static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
833 {
834         struct address_space *mapping = page->mapping;
835         struct inode *inode = mapping->host;
836         struct ntfs_inode *ni = ntfs_i(inode);
837         int err;
838
839         if (is_resident(ni)) {
840                 ni_lock(ni);
841                 err = attr_data_write_resident(ni, page);
842                 ni_unlock(ni);
843                 if (err != E_NTFS_NONRESIDENT) {
844                         unlock_page(page);
845                         return err;
846                 }
847         }
848
849         return block_write_full_page(page, ntfs_get_block, wbc);
850 }
851
852 static int ntfs_writepages(struct address_space *mapping,
853                            struct writeback_control *wbc)
854 {
855         /* Redirect call to 'ntfs_writepage' for resident files. */
856         if (is_resident(ntfs_i(mapping->host)))
857                 return generic_writepages(mapping, wbc);
858         return mpage_writepages(mapping, wbc, ntfs_get_block);
859 }
860
861 static int ntfs_get_block_write_begin(struct inode *inode, sector_t vbn,
862                                       struct buffer_head *bh_result, int create)
863 {
864         return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
865                                   bh_result, create, GET_BLOCK_WRITE_BEGIN);
866 }
867
868 int ntfs_write_begin(struct file *file, struct address_space *mapping,
869                      loff_t pos, u32 len, struct page **pagep, void **fsdata)
870 {
871         int err;
872         struct inode *inode = mapping->host;
873         struct ntfs_inode *ni = ntfs_i(inode);
874
875         *pagep = NULL;
876         if (is_resident(ni)) {
877                 struct page *page = grab_cache_page_write_begin(
878                         mapping, pos >> PAGE_SHIFT);
879
880                 if (!page) {
881                         err = -ENOMEM;
882                         goto out;
883                 }
884
885                 ni_lock(ni);
886                 err = attr_data_read_resident(ni, page);
887                 ni_unlock(ni);
888
889                 if (!err) {
890                         *pagep = page;
891                         goto out;
892                 }
893                 unlock_page(page);
894                 put_page(page);
895
896                 if (err != E_NTFS_NONRESIDENT)
897                         goto out;
898         }
899
900         err = block_write_begin(mapping, pos, len, pagep,
901                                 ntfs_get_block_write_begin);
902
903 out:
904         return err;
905 }
906
907 /*
908  * ntfs_write_end - Address_space_operations::write_end.
909  */
910 int ntfs_write_end(struct file *file, struct address_space *mapping,
911                    loff_t pos, u32 len, u32 copied, struct page *page,
912                    void *fsdata)
913 {
914         struct inode *inode = mapping->host;
915         struct ntfs_inode *ni = ntfs_i(inode);
916         u64 valid = ni->i_valid;
917         bool dirty = false;
918         int err;
919
920         if (is_resident(ni)) {
921                 ni_lock(ni);
922                 err = attr_data_write_resident(ni, page);
923                 ni_unlock(ni);
924                 if (!err) {
925                         dirty = true;
926                         /* Clear any buffers in page. */
927                         if (page_has_buffers(page)) {
928                                 struct buffer_head *head, *bh;
929
930                                 bh = head = page_buffers(page);
931                                 do {
932                                         clear_buffer_dirty(bh);
933                                         clear_buffer_mapped(bh);
934                                         set_buffer_uptodate(bh);
935                                 } while (head != (bh = bh->b_this_page));
936                         }
937                         SetPageUptodate(page);
938                         err = copied;
939                 }
940                 unlock_page(page);
941                 put_page(page);
942         } else {
943                 err = generic_write_end(file, mapping, pos, len, copied, page,
944                                         fsdata);
945         }
946
947         if (err >= 0) {
948                 if (!(ni->std_fa & FILE_ATTRIBUTE_ARCHIVE)) {
949                         inode->i_ctime = inode->i_mtime = current_time(inode);
950                         ni->std_fa |= FILE_ATTRIBUTE_ARCHIVE;
951                         dirty = true;
952                 }
953
954                 if (valid != ni->i_valid) {
955                         /* ni->i_valid is changed in ntfs_get_block_vbo. */
956                         dirty = true;
957                 }
958
959                 if (dirty)
960                         mark_inode_dirty(inode);
961         }
962
963         return err;
964 }
965
966 int reset_log_file(struct inode *inode)
967 {
968         int err;
969         loff_t pos = 0;
970         u32 log_size = inode->i_size;
971         struct address_space *mapping = inode->i_mapping;
972
973         for (;;) {
974                 u32 len;
975                 void *kaddr;
976                 struct page *page;
977
978                 len = pos + PAGE_SIZE > log_size ? (log_size - pos) : PAGE_SIZE;
979
980                 err = block_write_begin(mapping, pos, len, &page,
981                                         ntfs_get_block_write_begin);
982                 if (err)
983                         goto out;
984
985                 kaddr = kmap_atomic(page);
986                 memset(kaddr, -1, len);
987                 kunmap_atomic(kaddr);
988                 flush_dcache_page(page);
989
990                 err = block_write_end(NULL, mapping, pos, len, len, page, NULL);
991                 if (err < 0)
992                         goto out;
993                 pos += len;
994
995                 if (pos >= log_size)
996                         break;
997                 balance_dirty_pages_ratelimited(mapping);
998         }
999 out:
1000         mark_inode_dirty_sync(inode);
1001
1002         return err;
1003 }
1004
1005 int ntfs3_write_inode(struct inode *inode, struct writeback_control *wbc)
1006 {
1007         return _ni_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
1008 }
1009
1010 int ntfs_sync_inode(struct inode *inode)
1011 {
1012         return _ni_write_inode(inode, 1);
1013 }
1014
1015 /*
1016  * writeback_inode - Helper function for ntfs_flush_inodes().
1017  *
1018  * This writes both the inode and the file data blocks, waiting
1019  * for in flight data blocks before the start of the call.  It
1020  * does not wait for any io started during the call.
1021  */
1022 static int writeback_inode(struct inode *inode)
1023 {
1024         int ret = sync_inode_metadata(inode, 0);
1025
1026         if (!ret)
1027                 ret = filemap_fdatawrite(inode->i_mapping);
1028         return ret;
1029 }
1030
1031 /*
1032  * ntfs_flush_inodes
1033  *
1034  * Write data and metadata corresponding to i1 and i2.  The io is
1035  * started but we do not wait for any of it to finish.
1036  *
1037  * filemap_flush() is used for the block device, so if there is a dirty
1038  * page for a block already in flight, we will not wait and start the
1039  * io over again.
1040  */
1041 int ntfs_flush_inodes(struct super_block *sb, struct inode *i1,
1042                       struct inode *i2)
1043 {
1044         int ret = 0;
1045
1046         if (i1)
1047                 ret = writeback_inode(i1);
1048         if (!ret && i2)
1049                 ret = writeback_inode(i2);
1050         if (!ret)
1051                 ret = sync_blockdev_nowait(sb->s_bdev);
1052         return ret;
1053 }
1054
1055 int inode_write_data(struct inode *inode, const void *data, size_t bytes)
1056 {
1057         pgoff_t idx;
1058
1059         /* Write non resident data. */
1060         for (idx = 0; bytes; idx++) {
1061                 size_t op = bytes > PAGE_SIZE ? PAGE_SIZE : bytes;
1062                 struct page *page = ntfs_map_page(inode->i_mapping, idx);
1063
1064                 if (IS_ERR(page))
1065                         return PTR_ERR(page);
1066
1067                 lock_page(page);
1068                 WARN_ON(!PageUptodate(page));
1069                 ClearPageUptodate(page);
1070
1071                 memcpy(page_address(page), data, op);
1072
1073                 flush_dcache_page(page);
1074                 SetPageUptodate(page);
1075                 unlock_page(page);
1076
1077                 ntfs_unmap_page(page);
1078
1079                 bytes -= op;
1080                 data = Add2Ptr(data, PAGE_SIZE);
1081         }
1082         return 0;
1083 }
1084
1085 /*
1086  * ntfs_reparse_bytes
1087  *
1088  * Number of bytes for REPARSE_DATA_BUFFER(IO_REPARSE_TAG_SYMLINK)
1089  * for unicode string of @uni_len length.
1090  */
1091 static inline u32 ntfs_reparse_bytes(u32 uni_len)
1092 {
1093         /* Header + unicode string + decorated unicode string. */
1094         return sizeof(short) * (2 * uni_len + 4) +
1095                offsetof(struct REPARSE_DATA_BUFFER,
1096                         SymbolicLinkReparseBuffer.PathBuffer);
1097 }
1098
1099 static struct REPARSE_DATA_BUFFER *
1100 ntfs_create_reparse_buffer(struct ntfs_sb_info *sbi, const char *symname,
1101                            u32 size, u16 *nsize)
1102 {
1103         int i, err;
1104         struct REPARSE_DATA_BUFFER *rp;
1105         __le16 *rp_name;
1106         typeof(rp->SymbolicLinkReparseBuffer) *rs;
1107
1108         rp = kzalloc(ntfs_reparse_bytes(2 * size + 2), GFP_NOFS);
1109         if (!rp)
1110                 return ERR_PTR(-ENOMEM);
1111
1112         rs = &rp->SymbolicLinkReparseBuffer;
1113         rp_name = rs->PathBuffer;
1114
1115         /* Convert link name to UTF-16. */
1116         err = ntfs_nls_to_utf16(sbi, symname, size,
1117                                 (struct cpu_str *)(rp_name - 1), 2 * size,
1118                                 UTF16_LITTLE_ENDIAN);
1119         if (err < 0)
1120                 goto out;
1121
1122         /* err = the length of unicode name of symlink. */
1123         *nsize = ntfs_reparse_bytes(err);
1124
1125         if (*nsize > sbi->reparse.max_size) {
1126                 err = -EFBIG;
1127                 goto out;
1128         }
1129
1130         /* Translate Linux '/' into Windows '\'. */
1131         for (i = 0; i < err; i++) {
1132                 if (rp_name[i] == cpu_to_le16('/'))
1133                         rp_name[i] = cpu_to_le16('\\');
1134         }
1135
1136         rp->ReparseTag = IO_REPARSE_TAG_SYMLINK;
1137         rp->ReparseDataLength =
1138                 cpu_to_le16(*nsize - offsetof(struct REPARSE_DATA_BUFFER,
1139                                               SymbolicLinkReparseBuffer));
1140
1141         /* PrintName + SubstituteName. */
1142         rs->SubstituteNameOffset = cpu_to_le16(sizeof(short) * err);
1143         rs->SubstituteNameLength = cpu_to_le16(sizeof(short) * err + 8);
1144         rs->PrintNameLength = rs->SubstituteNameOffset;
1145
1146         /*
1147          * TODO: Use relative path if possible to allow Windows to
1148          * parse this path.
1149          * 0-absolute path 1- relative path (SYMLINK_FLAG_RELATIVE).
1150          */
1151         rs->Flags = 0;
1152
1153         memmove(rp_name + err + 4, rp_name, sizeof(short) * err);
1154
1155         /* Decorate SubstituteName. */
1156         rp_name += err;
1157         rp_name[0] = cpu_to_le16('\\');
1158         rp_name[1] = cpu_to_le16('?');
1159         rp_name[2] = cpu_to_le16('?');
1160         rp_name[3] = cpu_to_le16('\\');
1161
1162         return rp;
1163 out:
1164         kfree(rp);
1165         return ERR_PTR(err);
1166 }
1167
1168 struct inode *ntfs_create_inode(struct user_namespace *mnt_userns,
1169                                 struct inode *dir, struct dentry *dentry,
1170                                 const struct cpu_str *uni, umode_t mode,
1171                                 dev_t dev, const char *symname, u32 size,
1172                                 struct ntfs_fnd *fnd)
1173 {
1174         int err;
1175         struct super_block *sb = dir->i_sb;
1176         struct ntfs_sb_info *sbi = sb->s_fs_info;
1177         const struct qstr *name = &dentry->d_name;
1178         CLST ino = 0;
1179         struct ntfs_inode *dir_ni = ntfs_i(dir);
1180         struct ntfs_inode *ni = NULL;
1181         struct inode *inode = NULL;
1182         struct ATTRIB *attr;
1183         struct ATTR_STD_INFO5 *std5;
1184         struct ATTR_FILE_NAME *fname;
1185         struct MFT_REC *rec;
1186         u32 asize, dsize, sd_size;
1187         enum FILE_ATTRIBUTE fa;
1188         __le32 security_id = SECURITY_ID_INVALID;
1189         CLST vcn;
1190         const void *sd;
1191         u16 t16, nsize = 0, aid = 0;
1192         struct INDEX_ROOT *root, *dir_root;
1193         struct NTFS_DE *e, *new_de = NULL;
1194         struct REPARSE_DATA_BUFFER *rp = NULL;
1195         bool rp_inserted = false;
1196
1197         ni_lock_dir(dir_ni);
1198
1199         dir_root = indx_get_root(&dir_ni->dir, dir_ni, NULL, NULL);
1200         if (!dir_root) {
1201                 err = -EINVAL;
1202                 goto out1;
1203         }
1204
1205         if (S_ISDIR(mode)) {
1206                 /* Use parent's directory attributes. */
1207                 fa = dir_ni->std_fa | FILE_ATTRIBUTE_DIRECTORY |
1208                      FILE_ATTRIBUTE_ARCHIVE;
1209                 /*
1210                  * By default child directory inherits parent attributes.
1211                  * Root directory is hidden + system.
1212                  * Make an exception for children in root.
1213                  */
1214                 if (dir->i_ino == MFT_REC_ROOT)
1215                         fa &= ~(FILE_ATTRIBUTE_HIDDEN | FILE_ATTRIBUTE_SYSTEM);
1216         } else if (S_ISLNK(mode)) {
1217                 /* It is good idea that link should be the same type (file/dir) as target */
1218                 fa = FILE_ATTRIBUTE_REPARSE_POINT;
1219
1220                 /*
1221                  * Linux: there are dir/file/symlink and so on.
1222                  * NTFS: symlinks are "dir + reparse" or "file + reparse"
1223                  * It is good idea to create:
1224                  * dir + reparse if 'symname' points to directory
1225                  * or
1226                  * file + reparse if 'symname' points to file
1227                  * Unfortunately kern_path hangs if symname contains 'dir'.
1228                  */
1229
1230                 /*
1231                  *      struct path path;
1232                  *
1233                  *      if (!kern_path(symname, LOOKUP_FOLLOW, &path)){
1234                  *              struct inode *target = d_inode(path.dentry);
1235                  *
1236                  *              if (S_ISDIR(target->i_mode))
1237                  *                      fa |= FILE_ATTRIBUTE_DIRECTORY;
1238                  *              // if ( target->i_sb == sb ){
1239                  *              //      use relative path?
1240                  *              // }
1241                  *              path_put(&path);
1242                  *      }
1243                  */
1244         } else if (S_ISREG(mode)) {
1245                 if (sbi->options->sparse) {
1246                         /* Sparsed regular file, cause option 'sparse'. */
1247                         fa = FILE_ATTRIBUTE_SPARSE_FILE |
1248                              FILE_ATTRIBUTE_ARCHIVE;
1249                 } else if (dir_ni->std_fa & FILE_ATTRIBUTE_COMPRESSED) {
1250                         /* Compressed regular file, if parent is compressed. */
1251                         fa = FILE_ATTRIBUTE_COMPRESSED | FILE_ATTRIBUTE_ARCHIVE;
1252                 } else {
1253                         /* Regular file, default attributes. */
1254                         fa = FILE_ATTRIBUTE_ARCHIVE;
1255                 }
1256         } else {
1257                 fa = FILE_ATTRIBUTE_ARCHIVE;
1258         }
1259
1260         if (!(mode & 0222))
1261                 fa |= FILE_ATTRIBUTE_READONLY;
1262
1263         /* Allocate PATH_MAX bytes. */
1264         new_de = __getname();
1265         if (!new_de) {
1266                 err = -ENOMEM;
1267                 goto out1;
1268         }
1269
1270         /* Mark rw ntfs as dirty. it will be cleared at umount. */
1271         ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
1272
1273         /* Step 1: allocate and fill new mft record. */
1274         err = ntfs_look_free_mft(sbi, &ino, false, NULL, NULL);
1275         if (err)
1276                 goto out2;
1277
1278         ni = ntfs_new_inode(sbi, ino, fa & FILE_ATTRIBUTE_DIRECTORY);
1279         if (IS_ERR(ni)) {
1280                 err = PTR_ERR(ni);
1281                 ni = NULL;
1282                 goto out3;
1283         }
1284         inode = &ni->vfs_inode;
1285         inode_init_owner(mnt_userns, inode, dir, mode);
1286         mode = inode->i_mode;
1287
1288         inode->i_atime = inode->i_mtime = inode->i_ctime = ni->i_crtime =
1289                 current_time(inode);
1290
1291         rec = ni->mi.mrec;
1292         rec->hard_links = cpu_to_le16(1);
1293         attr = Add2Ptr(rec, le16_to_cpu(rec->attr_off));
1294
1295         /* Get default security id. */
1296         sd = s_default_security;
1297         sd_size = sizeof(s_default_security);
1298
1299         if (is_ntfs3(sbi)) {
1300                 security_id = dir_ni->std_security_id;
1301                 if (le32_to_cpu(security_id) < SECURITY_ID_FIRST) {
1302                         security_id = sbi->security.def_security_id;
1303
1304                         if (security_id == SECURITY_ID_INVALID &&
1305                             !ntfs_insert_security(sbi, sd, sd_size,
1306                                                   &security_id, NULL))
1307                                 sbi->security.def_security_id = security_id;
1308                 }
1309         }
1310
1311         /* Insert standard info. */
1312         std5 = Add2Ptr(attr, SIZEOF_RESIDENT);
1313
1314         if (security_id == SECURITY_ID_INVALID) {
1315                 dsize = sizeof(struct ATTR_STD_INFO);
1316         } else {
1317                 dsize = sizeof(struct ATTR_STD_INFO5);
1318                 std5->security_id = security_id;
1319                 ni->std_security_id = security_id;
1320         }
1321         asize = SIZEOF_RESIDENT + dsize;
1322
1323         attr->type = ATTR_STD;
1324         attr->size = cpu_to_le32(asize);
1325         attr->id = cpu_to_le16(aid++);
1326         attr->res.data_off = SIZEOF_RESIDENT_LE;
1327         attr->res.data_size = cpu_to_le32(dsize);
1328
1329         std5->cr_time = std5->m_time = std5->c_time = std5->a_time =
1330                 kernel2nt(&inode->i_atime);
1331
1332         ni->std_fa = fa;
1333         std5->fa = fa;
1334
1335         attr = Add2Ptr(attr, asize);
1336
1337         /* Insert file name. */
1338         err = fill_name_de(sbi, new_de, name, uni);
1339         if (err)
1340                 goto out4;
1341
1342         mi_get_ref(&ni->mi, &new_de->ref);
1343
1344         fname = (struct ATTR_FILE_NAME *)(new_de + 1);
1345         mi_get_ref(&dir_ni->mi, &fname->home);
1346         fname->dup.cr_time = fname->dup.m_time = fname->dup.c_time =
1347                 fname->dup.a_time = std5->cr_time;
1348         fname->dup.alloc_size = fname->dup.data_size = 0;
1349         fname->dup.fa = std5->fa;
1350         fname->dup.ea_size = fname->dup.reparse = 0;
1351
1352         dsize = le16_to_cpu(new_de->key_size);
1353         asize = ALIGN(SIZEOF_RESIDENT + dsize, 8);
1354
1355         attr->type = ATTR_NAME;
1356         attr->size = cpu_to_le32(asize);
1357         attr->res.data_off = SIZEOF_RESIDENT_LE;
1358         attr->res.flags = RESIDENT_FLAG_INDEXED;
1359         attr->id = cpu_to_le16(aid++);
1360         attr->res.data_size = cpu_to_le32(dsize);
1361         memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), fname, dsize);
1362
1363         attr = Add2Ptr(attr, asize);
1364
1365         if (security_id == SECURITY_ID_INVALID) {
1366                 /* Insert security attribute. */
1367                 asize = SIZEOF_RESIDENT + ALIGN(sd_size, 8);
1368
1369                 attr->type = ATTR_SECURE;
1370                 attr->size = cpu_to_le32(asize);
1371                 attr->id = cpu_to_le16(aid++);
1372                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1373                 attr->res.data_size = cpu_to_le32(sd_size);
1374                 memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), sd, sd_size);
1375
1376                 attr = Add2Ptr(attr, asize);
1377         }
1378
1379         attr->id = cpu_to_le16(aid++);
1380         if (fa & FILE_ATTRIBUTE_DIRECTORY) {
1381                 /*
1382                  * Regular directory or symlink to directory.
1383                  * Create root attribute.
1384                  */
1385                 dsize = sizeof(struct INDEX_ROOT) + sizeof(struct NTFS_DE);
1386                 asize = sizeof(I30_NAME) + SIZEOF_RESIDENT + dsize;
1387
1388                 attr->type = ATTR_ROOT;
1389                 attr->size = cpu_to_le32(asize);
1390
1391                 attr->name_len = ARRAY_SIZE(I30_NAME);
1392                 attr->name_off = SIZEOF_RESIDENT_LE;
1393                 attr->res.data_off =
1394                         cpu_to_le16(sizeof(I30_NAME) + SIZEOF_RESIDENT);
1395                 attr->res.data_size = cpu_to_le32(dsize);
1396                 memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), I30_NAME,
1397                        sizeof(I30_NAME));
1398
1399                 root = Add2Ptr(attr, sizeof(I30_NAME) + SIZEOF_RESIDENT);
1400                 memcpy(root, dir_root, offsetof(struct INDEX_ROOT, ihdr));
1401                 root->ihdr.de_off =
1402                         cpu_to_le32(sizeof(struct INDEX_HDR)); // 0x10
1403                 root->ihdr.used = cpu_to_le32(sizeof(struct INDEX_HDR) +
1404                                               sizeof(struct NTFS_DE));
1405                 root->ihdr.total = root->ihdr.used;
1406
1407                 e = Add2Ptr(root, sizeof(struct INDEX_ROOT));
1408                 e->size = cpu_to_le16(sizeof(struct NTFS_DE));
1409                 e->flags = NTFS_IE_LAST;
1410         } else if (S_ISLNK(mode)) {
1411                 /*
1412                  * Symlink to file.
1413                  * Create empty resident data attribute.
1414                  */
1415                 asize = SIZEOF_RESIDENT;
1416
1417                 /* Insert empty ATTR_DATA */
1418                 attr->type = ATTR_DATA;
1419                 attr->size = cpu_to_le32(SIZEOF_RESIDENT);
1420                 attr->name_off = SIZEOF_RESIDENT_LE;
1421                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1422         } else if (S_ISREG(mode)) {
1423                 /*
1424                  * Regular file. Create empty non resident data attribute.
1425                  */
1426                 attr->type = ATTR_DATA;
1427                 attr->non_res = 1;
1428                 attr->nres.evcn = cpu_to_le64(-1ll);
1429                 if (fa & FILE_ATTRIBUTE_SPARSE_FILE) {
1430                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
1431                         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
1432                         attr->flags = ATTR_FLAG_SPARSED;
1433                         asize = SIZEOF_NONRESIDENT_EX + 8;
1434                 } else if (fa & FILE_ATTRIBUTE_COMPRESSED) {
1435                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
1436                         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
1437                         attr->flags = ATTR_FLAG_COMPRESSED;
1438                         attr->nres.c_unit = COMPRESSION_UNIT;
1439                         asize = SIZEOF_NONRESIDENT_EX + 8;
1440                 } else {
1441                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT + 8);
1442                         attr->name_off = SIZEOF_NONRESIDENT_LE;
1443                         asize = SIZEOF_NONRESIDENT + 8;
1444                 }
1445                 attr->nres.run_off = attr->name_off;
1446         } else {
1447                 /*
1448                  * Node. Create empty resident data attribute.
1449                  */
1450                 attr->type = ATTR_DATA;
1451                 attr->size = cpu_to_le32(SIZEOF_RESIDENT);
1452                 attr->name_off = SIZEOF_RESIDENT_LE;
1453                 if (fa & FILE_ATTRIBUTE_SPARSE_FILE)
1454                         attr->flags = ATTR_FLAG_SPARSED;
1455                 else if (fa & FILE_ATTRIBUTE_COMPRESSED)
1456                         attr->flags = ATTR_FLAG_COMPRESSED;
1457                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1458                 asize = SIZEOF_RESIDENT;
1459                 ni->ni_flags |= NI_FLAG_RESIDENT;
1460         }
1461
1462         if (S_ISDIR(mode)) {
1463                 ni->ni_flags |= NI_FLAG_DIR;
1464                 err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
1465                 if (err)
1466                         goto out4;
1467         } else if (S_ISLNK(mode)) {
1468                 rp = ntfs_create_reparse_buffer(sbi, symname, size, &nsize);
1469
1470                 if (IS_ERR(rp)) {
1471                         err = PTR_ERR(rp);
1472                         rp = NULL;
1473                         goto out4;
1474                 }
1475
1476                 /*
1477                  * Insert ATTR_REPARSE.
1478                  */
1479                 attr = Add2Ptr(attr, asize);
1480                 attr->type = ATTR_REPARSE;
1481                 attr->id = cpu_to_le16(aid++);
1482
1483                 /* Resident or non resident? */
1484                 asize = ALIGN(SIZEOF_RESIDENT + nsize, 8);
1485                 t16 = PtrOffset(rec, attr);
1486
1487                 /*
1488                  * Below function 'ntfs_save_wsl_perm' requires 0x78 bytes.
1489                  * It is good idea to keep extened attributes resident.
1490                  */
1491                 if (asize + t16 + 0x78 + 8 > sbi->record_size) {
1492                         CLST alen;
1493                         CLST clst = bytes_to_cluster(sbi, nsize);
1494
1495                         /* Bytes per runs. */
1496                         t16 = sbi->record_size - t16 - SIZEOF_NONRESIDENT;
1497
1498                         attr->non_res = 1;
1499                         attr->nres.evcn = cpu_to_le64(clst - 1);
1500                         attr->name_off = SIZEOF_NONRESIDENT_LE;
1501                         attr->nres.run_off = attr->name_off;
1502                         attr->nres.data_size = cpu_to_le64(nsize);
1503                         attr->nres.valid_size = attr->nres.data_size;
1504                         attr->nres.alloc_size =
1505                                 cpu_to_le64(ntfs_up_cluster(sbi, nsize));
1506
1507                         err = attr_allocate_clusters(sbi, &ni->file.run, 0, 0,
1508                                                      clst, NULL, 0, &alen, 0,
1509                                                      NULL);
1510                         if (err)
1511                                 goto out5;
1512
1513                         err = run_pack(&ni->file.run, 0, clst,
1514                                        Add2Ptr(attr, SIZEOF_NONRESIDENT), t16,
1515                                        &vcn);
1516                         if (err < 0)
1517                                 goto out5;
1518
1519                         if (vcn != clst) {
1520                                 err = -EINVAL;
1521                                 goto out5;
1522                         }
1523
1524                         asize = SIZEOF_NONRESIDENT + ALIGN(err, 8);
1525                 } else {
1526                         attr->res.data_off = SIZEOF_RESIDENT_LE;
1527                         attr->res.data_size = cpu_to_le32(nsize);
1528                         memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), rp, nsize);
1529                         nsize = 0;
1530                 }
1531                 /* Size of symlink equals the length of input string. */
1532                 inode->i_size = size;
1533
1534                 attr->size = cpu_to_le32(asize);
1535
1536                 err = ntfs_insert_reparse(sbi, IO_REPARSE_TAG_SYMLINK,
1537                                           &new_de->ref);
1538                 if (err)
1539                         goto out5;
1540
1541                 rp_inserted = true;
1542         }
1543
1544         attr = Add2Ptr(attr, asize);
1545         attr->type = ATTR_END;
1546
1547         rec->used = cpu_to_le32(PtrOffset(rec, attr) + 8);
1548         rec->next_attr_id = cpu_to_le16(aid);
1549
1550         /* Step 2: Add new name in index. */
1551         err = indx_insert_entry(&dir_ni->dir, dir_ni, new_de, sbi, fnd, 0);
1552         if (err)
1553                 goto out6;
1554
1555         /* Unlock parent directory before ntfs_init_acl. */
1556         ni_unlock(dir_ni);
1557
1558         inode->i_generation = le16_to_cpu(rec->seq);
1559
1560         dir->i_mtime = dir->i_ctime = inode->i_atime;
1561
1562         if (S_ISDIR(mode)) {
1563                 inode->i_op = &ntfs_dir_inode_operations;
1564                 inode->i_fop = &ntfs_dir_operations;
1565         } else if (S_ISLNK(mode)) {
1566                 inode->i_op = &ntfs_link_inode_operations;
1567                 inode->i_fop = NULL;
1568                 inode->i_mapping->a_ops = &ntfs_aops;
1569                 inode->i_size = size;
1570                 inode_nohighmem(inode);
1571         } else if (S_ISREG(mode)) {
1572                 inode->i_op = &ntfs_file_inode_operations;
1573                 inode->i_fop = &ntfs_file_operations;
1574                 inode->i_mapping->a_ops =
1575                         is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
1576                 init_rwsem(&ni->file.run_lock);
1577         } else {
1578                 inode->i_op = &ntfs_special_inode_operations;
1579                 init_special_inode(inode, mode, dev);
1580         }
1581
1582 #ifdef CONFIG_NTFS3_FS_POSIX_ACL
1583         if (!S_ISLNK(mode) && (sb->s_flags & SB_POSIXACL)) {
1584                 err = ntfs_init_acl(mnt_userns, inode, dir);
1585                 if (err)
1586                         goto out7;
1587         } else
1588 #endif
1589         {
1590                 inode->i_flags |= S_NOSEC;
1591         }
1592
1593         /* Write non resident data. */
1594         if (nsize) {
1595                 err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rp, nsize, 0);
1596                 if (err)
1597                         goto out7;
1598         }
1599
1600         /*
1601          * Call 'd_instantiate' after inode->i_op is set
1602          * but before finish_open.
1603          */
1604         d_instantiate(dentry, inode);
1605
1606         ntfs_save_wsl_perm(inode);
1607         mark_inode_dirty(dir);
1608         mark_inode_dirty(inode);
1609
1610         /* Normal exit. */
1611         goto out2;
1612
1613 out7:
1614
1615         /* Undo 'indx_insert_entry'. */
1616         ni_lock_dir(dir_ni);
1617         indx_delete_entry(&dir_ni->dir, dir_ni, new_de + 1,
1618                           le16_to_cpu(new_de->key_size), sbi);
1619         /* ni_unlock(dir_ni); will be called later. */
1620 out6:
1621         if (rp_inserted)
1622                 ntfs_remove_reparse(sbi, IO_REPARSE_TAG_SYMLINK, &new_de->ref);
1623
1624 out5:
1625         if (S_ISDIR(mode) || run_is_empty(&ni->file.run))
1626                 goto out4;
1627
1628         run_deallocate(sbi, &ni->file.run, false);
1629
1630 out4:
1631         clear_rec_inuse(rec);
1632         clear_nlink(inode);
1633         ni->mi.dirty = false;
1634         discard_new_inode(inode);
1635 out3:
1636         ntfs_mark_rec_free(sbi, ino, false);
1637
1638 out2:
1639         __putname(new_de);
1640         kfree(rp);
1641
1642 out1:
1643         if (err) {
1644                 ni_unlock(dir_ni);
1645                 return ERR_PTR(err);
1646         }
1647
1648         unlock_new_inode(inode);
1649
1650         return inode;
1651 }
1652
1653 int ntfs_link_inode(struct inode *inode, struct dentry *dentry)
1654 {
1655         int err;
1656         struct ntfs_inode *ni = ntfs_i(inode);
1657         struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
1658         struct NTFS_DE *de;
1659
1660         /* Allocate PATH_MAX bytes. */
1661         de = __getname();
1662         if (!de)
1663                 return -ENOMEM;
1664
1665         /* Mark rw ntfs as dirty. It will be cleared at umount. */
1666         ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
1667
1668         /* Construct 'de'. */
1669         err = fill_name_de(sbi, de, &dentry->d_name, NULL);
1670         if (err)
1671                 goto out;
1672
1673         err = ni_add_name(ntfs_i(d_inode(dentry->d_parent)), ni, de);
1674 out:
1675         __putname(de);
1676         return err;
1677 }
1678
1679 /*
1680  * ntfs_unlink_inode
1681  *
1682  * inode_operations::unlink
1683  * inode_operations::rmdir
1684  */
1685 int ntfs_unlink_inode(struct inode *dir, const struct dentry *dentry)
1686 {
1687         int err;
1688         struct ntfs_sb_info *sbi = dir->i_sb->s_fs_info;
1689         struct inode *inode = d_inode(dentry);
1690         struct ntfs_inode *ni = ntfs_i(inode);
1691         struct ntfs_inode *dir_ni = ntfs_i(dir);
1692         struct NTFS_DE *de, *de2 = NULL;
1693         int undo_remove;
1694
1695         if (ntfs_is_meta_file(sbi, ni->mi.rno))
1696                 return -EINVAL;
1697
1698         /* Allocate PATH_MAX bytes. */
1699         de = __getname();
1700         if (!de)
1701                 return -ENOMEM;
1702
1703         ni_lock(ni);
1704
1705         if (S_ISDIR(inode->i_mode) && !dir_is_empty(inode)) {
1706                 err = -ENOTEMPTY;
1707                 goto out;
1708         }
1709
1710         err = fill_name_de(sbi, de, &dentry->d_name, NULL);
1711         if (err < 0)
1712                 goto out;
1713
1714         undo_remove = 0;
1715         err = ni_remove_name(dir_ni, ni, de, &de2, &undo_remove);
1716
1717         if (!err) {
1718                 drop_nlink(inode);
1719                 dir->i_mtime = dir->i_ctime = current_time(dir);
1720                 mark_inode_dirty(dir);
1721                 inode->i_ctime = dir->i_ctime;
1722                 if (inode->i_nlink)
1723                         mark_inode_dirty(inode);
1724         } else if (!ni_remove_name_undo(dir_ni, ni, de, de2, undo_remove)) {
1725                 _ntfs_bad_inode(inode);
1726         } else {
1727                 if (ni_is_dirty(dir))
1728                         mark_inode_dirty(dir);
1729                 if (ni_is_dirty(inode))
1730                         mark_inode_dirty(inode);
1731         }
1732
1733 out:
1734         ni_unlock(ni);
1735         __putname(de);
1736         return err;
1737 }
1738
1739 void ntfs_evict_inode(struct inode *inode)
1740 {
1741         truncate_inode_pages_final(&inode->i_data);
1742
1743         if (inode->i_nlink)
1744                 _ni_write_inode(inode, inode_needs_sync(inode));
1745
1746         invalidate_inode_buffers(inode);
1747         clear_inode(inode);
1748
1749         ni_clear(ntfs_i(inode));
1750 }
1751
1752 static noinline int ntfs_readlink_hlp(struct inode *inode, char *buffer,
1753                                       int buflen)
1754 {
1755         int i, err = -EINVAL;
1756         struct ntfs_inode *ni = ntfs_i(inode);
1757         struct super_block *sb = inode->i_sb;
1758         struct ntfs_sb_info *sbi = sb->s_fs_info;
1759         u64 size;
1760         u16 ulen = 0;
1761         void *to_free = NULL;
1762         struct REPARSE_DATA_BUFFER *rp;
1763         const __le16 *uname;
1764         struct ATTRIB *attr;
1765
1766         /* Reparse data present. Try to parse it. */
1767         static_assert(!offsetof(struct REPARSE_DATA_BUFFER, ReparseTag));
1768         static_assert(sizeof(u32) == sizeof(rp->ReparseTag));
1769
1770         *buffer = 0;
1771
1772         attr = ni_find_attr(ni, NULL, NULL, ATTR_REPARSE, NULL, 0, NULL, NULL);
1773         if (!attr)
1774                 goto out;
1775
1776         if (!attr->non_res) {
1777                 rp = resident_data_ex(attr, sizeof(struct REPARSE_DATA_BUFFER));
1778                 if (!rp)
1779                         goto out;
1780                 size = le32_to_cpu(attr->res.data_size);
1781         } else {
1782                 size = le64_to_cpu(attr->nres.data_size);
1783                 rp = NULL;
1784         }
1785
1786         if (size > sbi->reparse.max_size || size <= sizeof(u32))
1787                 goto out;
1788
1789         if (!rp) {
1790                 rp = kmalloc(size, GFP_NOFS);
1791                 if (!rp) {
1792                         err = -ENOMEM;
1793                         goto out;
1794                 }
1795                 to_free = rp;
1796                 /* Read into temporal buffer. */
1797                 err = ntfs_read_run_nb(sbi, &ni->file.run, 0, rp, size, NULL);
1798                 if (err)
1799                         goto out;
1800         }
1801
1802         /* Microsoft Tag. */
1803         switch (rp->ReparseTag) {
1804         case IO_REPARSE_TAG_MOUNT_POINT:
1805                 /* Mount points and junctions. */
1806                 /* Can we use 'Rp->MountPointReparseBuffer.PrintNameLength'? */
1807                 if (size <= offsetof(struct REPARSE_DATA_BUFFER,
1808                                      MountPointReparseBuffer.PathBuffer))
1809                         goto out;
1810                 uname = Add2Ptr(rp,
1811                                 offsetof(struct REPARSE_DATA_BUFFER,
1812                                          MountPointReparseBuffer.PathBuffer) +
1813                                         le16_to_cpu(rp->MountPointReparseBuffer
1814                                                             .PrintNameOffset));
1815                 ulen = le16_to_cpu(rp->MountPointReparseBuffer.PrintNameLength);
1816                 break;
1817
1818         case IO_REPARSE_TAG_SYMLINK:
1819                 /* FolderSymbolicLink */
1820                 /* Can we use 'Rp->SymbolicLinkReparseBuffer.PrintNameLength'? */
1821                 if (size <= offsetof(struct REPARSE_DATA_BUFFER,
1822                                      SymbolicLinkReparseBuffer.PathBuffer))
1823                         goto out;
1824                 uname = Add2Ptr(
1825                         rp, offsetof(struct REPARSE_DATA_BUFFER,
1826                                      SymbolicLinkReparseBuffer.PathBuffer) +
1827                                     le16_to_cpu(rp->SymbolicLinkReparseBuffer
1828                                                         .PrintNameOffset));
1829                 ulen = le16_to_cpu(
1830                         rp->SymbolicLinkReparseBuffer.PrintNameLength);
1831                 break;
1832
1833         case IO_REPARSE_TAG_CLOUD:
1834         case IO_REPARSE_TAG_CLOUD_1:
1835         case IO_REPARSE_TAG_CLOUD_2:
1836         case IO_REPARSE_TAG_CLOUD_3:
1837         case IO_REPARSE_TAG_CLOUD_4:
1838         case IO_REPARSE_TAG_CLOUD_5:
1839         case IO_REPARSE_TAG_CLOUD_6:
1840         case IO_REPARSE_TAG_CLOUD_7:
1841         case IO_REPARSE_TAG_CLOUD_8:
1842         case IO_REPARSE_TAG_CLOUD_9:
1843         case IO_REPARSE_TAG_CLOUD_A:
1844         case IO_REPARSE_TAG_CLOUD_B:
1845         case IO_REPARSE_TAG_CLOUD_C:
1846         case IO_REPARSE_TAG_CLOUD_D:
1847         case IO_REPARSE_TAG_CLOUD_E:
1848         case IO_REPARSE_TAG_CLOUD_F:
1849                 err = sizeof("OneDrive") - 1;
1850                 if (err > buflen)
1851                         err = buflen;
1852                 memcpy(buffer, "OneDrive", err);
1853                 goto out;
1854
1855         default:
1856                 if (IsReparseTagMicrosoft(rp->ReparseTag)) {
1857                         /* Unknown Microsoft Tag. */
1858                         goto out;
1859                 }
1860                 if (!IsReparseTagNameSurrogate(rp->ReparseTag) ||
1861                     size <= sizeof(struct REPARSE_POINT)) {
1862                         goto out;
1863                 }
1864
1865                 /* Users tag. */
1866                 uname = Add2Ptr(rp, sizeof(struct REPARSE_POINT));
1867                 ulen = le16_to_cpu(rp->ReparseDataLength) -
1868                        sizeof(struct REPARSE_POINT);
1869         }
1870
1871         /* Convert nlen from bytes to UNICODE chars. */
1872         ulen >>= 1;
1873
1874         /* Check that name is available. */
1875         if (!ulen || uname + ulen > (__le16 *)Add2Ptr(rp, size))
1876                 goto out;
1877
1878         /* If name is already zero terminated then truncate it now. */
1879         if (!uname[ulen - 1])
1880                 ulen -= 1;
1881
1882         err = ntfs_utf16_to_nls(sbi, uname, ulen, buffer, buflen);
1883
1884         if (err < 0)
1885                 goto out;
1886
1887         /* Translate Windows '\' into Linux '/'. */
1888         for (i = 0; i < err; i++) {
1889                 if (buffer[i] == '\\')
1890                         buffer[i] = '/';
1891         }
1892
1893         /* Always set last zero. */
1894         buffer[err] = 0;
1895 out:
1896         kfree(to_free);
1897         return err;
1898 }
1899
1900 static const char *ntfs_get_link(struct dentry *de, struct inode *inode,
1901                                  struct delayed_call *done)
1902 {
1903         int err;
1904         char *ret;
1905
1906         if (!de)
1907                 return ERR_PTR(-ECHILD);
1908
1909         ret = kmalloc(PAGE_SIZE, GFP_NOFS);
1910         if (!ret)
1911                 return ERR_PTR(-ENOMEM);
1912
1913         err = ntfs_readlink_hlp(inode, ret, PAGE_SIZE);
1914         if (err < 0) {
1915                 kfree(ret);
1916                 return ERR_PTR(err);
1917         }
1918
1919         set_delayed_call(done, kfree_link, ret);
1920
1921         return ret;
1922 }
1923
1924 // clang-format off
1925 const struct inode_operations ntfs_link_inode_operations = {
1926         .get_link       = ntfs_get_link,
1927         .setattr        = ntfs3_setattr,
1928         .listxattr      = ntfs_listxattr,
1929         .permission     = ntfs_permission,
1930         .get_acl        = ntfs_get_acl,
1931         .set_acl        = ntfs_set_acl,
1932 };
1933
1934 const struct address_space_operations ntfs_aops = {
1935         .read_folio     = ntfs_read_folio,
1936         .readahead      = ntfs_readahead,
1937         .writepage      = ntfs_writepage,
1938         .writepages     = ntfs_writepages,
1939         .write_begin    = ntfs_write_begin,
1940         .write_end      = ntfs_write_end,
1941         .direct_IO      = ntfs_direct_IO,
1942         .bmap           = ntfs_bmap,
1943         .dirty_folio    = block_dirty_folio,
1944         .invalidate_folio = block_invalidate_folio,
1945 };
1946
1947 const struct address_space_operations ntfs_aops_cmpr = {
1948         .read_folio     = ntfs_read_folio,
1949         .readahead      = ntfs_readahead,
1950 };
1951 // clang-format on