WIP: merge_config
[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         if (le16_to_cpu(attr->name_off) + attr->name_len > asize)
133                 goto out;
134
135         switch (attr->type) {
136         case ATTR_STD:
137                 if (attr->non_res ||
138                     asize < sizeof(struct ATTR_STD_INFO) + roff ||
139                     rsize < sizeof(struct ATTR_STD_INFO))
140                         goto out;
141
142                 if (std5)
143                         goto next_attr;
144
145                 std5 = Add2Ptr(attr, roff);
146
147 #ifdef STATX_BTIME
148                 nt2kernel(std5->cr_time, &ni->i_crtime);
149 #endif
150                 nt2kernel(std5->a_time, &inode->i_atime);
151                 nt2kernel(std5->c_time, &inode->i_ctime);
152                 nt2kernel(std5->m_time, &inode->i_mtime);
153
154                 ni->std_fa = std5->fa;
155
156                 if (asize >= sizeof(struct ATTR_STD_INFO5) + roff &&
157                     rsize >= sizeof(struct ATTR_STD_INFO5))
158                         ni->std_security_id = std5->security_id;
159                 goto next_attr;
160
161         case ATTR_LIST:
162                 if (attr->name_len || le || ino == MFT_REC_LOG)
163                         goto out;
164
165                 err = ntfs_load_attr_list(ni, attr);
166                 if (err)
167                         goto out;
168
169                 le = NULL;
170                 attr = NULL;
171                 goto next_attr;
172
173         case ATTR_NAME:
174                 if (attr->non_res || asize < SIZEOF_ATTRIBUTE_FILENAME + roff ||
175                     rsize < SIZEOF_ATTRIBUTE_FILENAME)
176                         goto out;
177
178                 fname = Add2Ptr(attr, roff);
179                 if (fname->type == FILE_NAME_DOS)
180                         goto next_attr;
181
182                 names += 1;
183                 if (name && name->len == fname->name_len &&
184                     !ntfs_cmp_names_cpu(name, (struct le_str *)&fname->name_len,
185                                         NULL, false))
186                         is_match = true;
187
188                 goto next_attr;
189
190         case ATTR_DATA:
191                 if (is_dir) {
192                         /* Ignore data attribute in dir record. */
193                         goto next_attr;
194                 }
195
196                 if (ino == MFT_REC_BADCLUST && !attr->non_res)
197                         goto next_attr;
198
199                 if (attr->name_len &&
200                     ((ino != MFT_REC_BADCLUST || !attr->non_res ||
201                       attr->name_len != ARRAY_SIZE(BAD_NAME) ||
202                       memcmp(attr_name(attr), BAD_NAME, sizeof(BAD_NAME))) &&
203                      (ino != MFT_REC_SECURE || !attr->non_res ||
204                       attr->name_len != ARRAY_SIZE(SDS_NAME) ||
205                       memcmp(attr_name(attr), SDS_NAME, sizeof(SDS_NAME))))) {
206                         /* File contains stream attribute. Ignore it. */
207                         goto next_attr;
208                 }
209
210                 if (is_attr_sparsed(attr))
211                         ni->std_fa |= FILE_ATTRIBUTE_SPARSE_FILE;
212                 else
213                         ni->std_fa &= ~FILE_ATTRIBUTE_SPARSE_FILE;
214
215                 if (is_attr_compressed(attr))
216                         ni->std_fa |= FILE_ATTRIBUTE_COMPRESSED;
217                 else
218                         ni->std_fa &= ~FILE_ATTRIBUTE_COMPRESSED;
219
220                 if (is_attr_encrypted(attr))
221                         ni->std_fa |= FILE_ATTRIBUTE_ENCRYPTED;
222                 else
223                         ni->std_fa &= ~FILE_ATTRIBUTE_ENCRYPTED;
224
225                 if (!attr->non_res) {
226                         ni->i_valid = inode->i_size = rsize;
227                         inode_set_bytes(inode, rsize);
228                 }
229
230                 mode = S_IFREG | (0777 & sbi->options->fs_fmask_inv);
231
232                 if (!attr->non_res) {
233                         ni->ni_flags |= NI_FLAG_RESIDENT;
234                         goto next_attr;
235                 }
236
237                 inode_set_bytes(inode, attr_ondisk_size(attr));
238
239                 ni->i_valid = le64_to_cpu(attr->nres.valid_size);
240                 inode->i_size = le64_to_cpu(attr->nres.data_size);
241                 if (!attr->nres.alloc_size)
242                         goto next_attr;
243
244                 run = ino == MFT_REC_BITMAP ? &sbi->used.bitmap.run
245                                             : &ni->file.run;
246                 break;
247
248         case ATTR_ROOT:
249                 if (attr->non_res)
250                         goto out;
251
252                 root = Add2Ptr(attr, roff);
253                 is_root = true;
254
255                 if (attr->name_len != ARRAY_SIZE(I30_NAME) ||
256                     memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
257                         goto next_attr;
258
259                 if (root->type != ATTR_NAME ||
260                     root->rule != NTFS_COLLATION_TYPE_FILENAME)
261                         goto out;
262
263                 if (!is_dir)
264                         goto next_attr;
265
266                 ni->ni_flags |= NI_FLAG_DIR;
267
268                 err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
269                 if (err)
270                         goto out;
271
272                 mode = sb->s_root
273                                ? (S_IFDIR | (0777 & sbi->options->fs_dmask_inv))
274                                : (S_IFDIR | 0777);
275                 goto next_attr;
276
277         case ATTR_ALLOC:
278                 if (!is_root || attr->name_len != ARRAY_SIZE(I30_NAME) ||
279                     memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
280                         goto next_attr;
281
282                 inode->i_size = le64_to_cpu(attr->nres.data_size);
283                 ni->i_valid = le64_to_cpu(attr->nres.valid_size);
284                 inode_set_bytes(inode, le64_to_cpu(attr->nres.alloc_size));
285
286                 run = &ni->dir.alloc_run;
287                 break;
288
289         case ATTR_BITMAP:
290                 if (ino == MFT_REC_MFT) {
291                         if (!attr->non_res)
292                                 goto out;
293 #ifndef CONFIG_NTFS3_64BIT_CLUSTER
294                         /* 0x20000000 = 2^32 / 8 */
295                         if (le64_to_cpu(attr->nres.alloc_size) >= 0x20000000)
296                                 goto out;
297 #endif
298                         run = &sbi->mft.bitmap.run;
299                         break;
300                 } else if (is_dir && attr->name_len == ARRAY_SIZE(I30_NAME) &&
301                            !memcmp(attr_name(attr), I30_NAME,
302                                    sizeof(I30_NAME)) &&
303                            attr->non_res) {
304                         run = &ni->dir.bitmap_run;
305                         break;
306                 }
307                 goto next_attr;
308
309         case ATTR_REPARSE:
310                 if (attr->name_len)
311                         goto next_attr;
312
313                 rp_fa = ni_parse_reparse(ni, attr, &rp);
314                 switch (rp_fa) {
315                 case REPARSE_LINK:
316                         /*
317                          * Normal symlink.
318                          * Assume one unicode symbol == one utf8.
319                          */
320                         inode->i_size = le16_to_cpu(rp.SymbolicLinkReparseBuffer
321                                                             .PrintNameLength) /
322                                         sizeof(u16);
323
324                         ni->i_valid = inode->i_size;
325
326                         /* Clear directory bit. */
327                         if (ni->ni_flags & NI_FLAG_DIR) {
328                                 indx_clear(&ni->dir);
329                                 memset(&ni->dir, 0, sizeof(ni->dir));
330                                 ni->ni_flags &= ~NI_FLAG_DIR;
331                         } else {
332                                 run_close(&ni->file.run);
333                         }
334                         mode = S_IFLNK | 0777;
335                         is_dir = false;
336                         if (attr->non_res) {
337                                 run = &ni->file.run;
338                                 goto attr_unpack_run; // Double break.
339                         }
340                         break;
341
342                 case REPARSE_COMPRESSED:
343                         break;
344
345                 case REPARSE_DEDUPLICATED:
346                         break;
347                 }
348                 goto next_attr;
349
350         case ATTR_EA_INFO:
351                 if (!attr->name_len &&
352                     resident_data_ex(attr, sizeof(struct EA_INFO))) {
353                         ni->ni_flags |= NI_FLAG_EA;
354                         /*
355                          * ntfs_get_wsl_perm updates inode->i_uid, inode->i_gid, inode->i_mode
356                          */
357                         inode->i_mode = mode;
358                         ntfs_get_wsl_perm(inode);
359                         mode = inode->i_mode;
360                 }
361                 goto next_attr;
362
363         default:
364                 goto next_attr;
365         }
366
367 attr_unpack_run:
368         roff = le16_to_cpu(attr->nres.run_off);
369
370         if (roff > asize) {
371                 err = -EINVAL;
372                 goto out;
373         }
374
375         t64 = le64_to_cpu(attr->nres.svcn);
376
377         err = run_unpack_ex(run, sbi, ino, t64, le64_to_cpu(attr->nres.evcn),
378                             t64, Add2Ptr(attr, roff), asize - roff);
379         if (err < 0)
380                 goto out;
381         err = 0;
382         goto next_attr;
383
384 end_enum:
385
386         if (!std5)
387                 goto out;
388
389         if (!is_match && name) {
390                 /* Reuse rec as buffer for ascii name. */
391                 err = -ENOENT;
392                 goto out;
393         }
394
395         if (std5->fa & FILE_ATTRIBUTE_READONLY)
396                 mode &= ~0222;
397
398         if (!names) {
399                 err = -EINVAL;
400                 goto out;
401         }
402
403         if (names != le16_to_cpu(rec->hard_links)) {
404                 /* Correct minor error on the fly. Do not mark inode as dirty. */
405                 rec->hard_links = cpu_to_le16(names);
406                 ni->mi.dirty = true;
407         }
408
409         set_nlink(inode, names);
410
411         if (S_ISDIR(mode)) {
412                 ni->std_fa |= FILE_ATTRIBUTE_DIRECTORY;
413
414                 /*
415                  * Dot and dot-dot should be included in count but was not
416                  * included in enumeration.
417                  * Usually a hard links to directories are disabled.
418                  */
419                 inode->i_op = &ntfs_dir_inode_operations;
420                 inode->i_fop = &ntfs_dir_operations;
421                 ni->i_valid = 0;
422         } else if (S_ISLNK(mode)) {
423                 ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
424                 inode->i_op = &ntfs_link_inode_operations;
425                 inode->i_fop = NULL;
426                 inode_nohighmem(inode);
427         } else if (S_ISREG(mode)) {
428                 ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
429                 inode->i_op = &ntfs_file_inode_operations;
430                 inode->i_fop = &ntfs_file_operations;
431                 inode->i_mapping->a_ops =
432                         is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
433                 if (ino != MFT_REC_MFT)
434                         init_rwsem(&ni->file.run_lock);
435         } else if (S_ISCHR(mode) || S_ISBLK(mode) || S_ISFIFO(mode) ||
436                    S_ISSOCK(mode)) {
437                 inode->i_op = &ntfs_special_inode_operations;
438                 init_special_inode(inode, mode, inode->i_rdev);
439         } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
440                    fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
441                 /* Records in $Extend are not a files or general directories. */
442                 inode->i_op = &ntfs_file_inode_operations;
443         } else {
444                 err = -EINVAL;
445                 goto out;
446         }
447
448         if ((sbi->options->sys_immutable &&
449              (std5->fa & FILE_ATTRIBUTE_SYSTEM)) &&
450             !S_ISFIFO(mode) && !S_ISSOCK(mode) && !S_ISLNK(mode)) {
451                 inode->i_flags |= S_IMMUTABLE;
452         } else {
453                 inode->i_flags &= ~S_IMMUTABLE;
454         }
455
456         inode->i_mode = mode;
457         if (!(ni->ni_flags & NI_FLAG_EA)) {
458                 /* If no xattr then no security (stored in xattr). */
459                 inode->i_flags |= S_NOSEC;
460         }
461
462 Ok:
463         if (ino == MFT_REC_MFT && !sb->s_root)
464                 sbi->mft.ni = NULL;
465
466         unlock_new_inode(inode);
467
468         return inode;
469
470 out:
471         if (ino == MFT_REC_MFT && !sb->s_root)
472                 sbi->mft.ni = NULL;
473
474         iget_failed(inode);
475         return ERR_PTR(err);
476 }
477
478 /*
479  * ntfs_test_inode
480  *
481  * Return: 1 if match.
482  */
483 static int ntfs_test_inode(struct inode *inode, void *data)
484 {
485         struct MFT_REF *ref = data;
486
487         return ino_get(ref) == inode->i_ino;
488 }
489
490 static int ntfs_set_inode(struct inode *inode, void *data)
491 {
492         const struct MFT_REF *ref = data;
493
494         inode->i_ino = ino_get(ref);
495         return 0;
496 }
497
498 struct inode *ntfs_iget5(struct super_block *sb, const struct MFT_REF *ref,
499                          const struct cpu_str *name)
500 {
501         struct inode *inode;
502
503         inode = iget5_locked(sb, ino_get(ref), ntfs_test_inode, ntfs_set_inode,
504                              (void *)ref);
505         if (unlikely(!inode))
506                 return ERR_PTR(-ENOMEM);
507
508         /* If this is a freshly allocated inode, need to read it now. */
509         if (inode->i_state & I_NEW)
510                 inode = ntfs_read_mft(inode, name, ref);
511         else if (ref->seq != ntfs_i(inode)->mi.mrec->seq) {
512                 /* Inode overlaps? */
513                 _ntfs_bad_inode(inode);
514         }
515
516         return inode;
517 }
518
519 enum get_block_ctx {
520         GET_BLOCK_GENERAL = 0,
521         GET_BLOCK_WRITE_BEGIN = 1,
522         GET_BLOCK_DIRECT_IO_R = 2,
523         GET_BLOCK_DIRECT_IO_W = 3,
524         GET_BLOCK_BMAP = 4,
525 };
526
527 static noinline int ntfs_get_block_vbo(struct inode *inode, u64 vbo,
528                                        struct buffer_head *bh, int create,
529                                        enum get_block_ctx ctx)
530 {
531         struct super_block *sb = inode->i_sb;
532         struct ntfs_sb_info *sbi = sb->s_fs_info;
533         struct ntfs_inode *ni = ntfs_i(inode);
534         struct page *page = bh->b_page;
535         u8 cluster_bits = sbi->cluster_bits;
536         u32 block_size = sb->s_blocksize;
537         u64 bytes, lbo, valid;
538         u32 off;
539         int err;
540         CLST vcn, lcn, len;
541         bool new;
542
543         /* Clear previous state. */
544         clear_buffer_new(bh);
545         clear_buffer_uptodate(bh);
546
547         /* Direct write uses 'create=0'. */
548         if (!create && vbo >= ni->i_valid) {
549                 /* Out of valid. */
550                 return 0;
551         }
552
553         if (vbo >= inode->i_size) {
554                 /* Out of size. */
555                 return 0;
556         }
557
558         if (is_resident(ni)) {
559                 ni_lock(ni);
560                 err = attr_data_read_resident(ni, page);
561                 ni_unlock(ni);
562
563                 if (!err)
564                         set_buffer_uptodate(bh);
565                 bh->b_size = block_size;
566                 return err;
567         }
568
569         vcn = vbo >> cluster_bits;
570         off = vbo & sbi->cluster_mask;
571         new = false;
572
573         err = attr_data_get_block(ni, vcn, 1, &lcn, &len, create ? &new : NULL);
574         if (err)
575                 goto out;
576
577         if (!len)
578                 return 0;
579
580         bytes = ((u64)len << cluster_bits) - off;
581
582         if (lcn == SPARSE_LCN) {
583                 if (!create) {
584                         if (bh->b_size > bytes)
585                                 bh->b_size = bytes;
586                         return 0;
587                 }
588                 WARN_ON(1);
589         }
590
591         if (new) {
592                 set_buffer_new(bh);
593                 if ((len << cluster_bits) > block_size)
594                         ntfs_sparse_cluster(inode, page, vcn, len);
595         }
596
597         lbo = ((u64)lcn << cluster_bits) + off;
598
599         set_buffer_mapped(bh);
600         bh->b_bdev = sb->s_bdev;
601         bh->b_blocknr = lbo >> sb->s_blocksize_bits;
602
603         valid = ni->i_valid;
604
605         if (ctx == GET_BLOCK_DIRECT_IO_W) {
606                 /* ntfs_direct_IO will update ni->i_valid. */
607                 if (vbo >= valid)
608                         set_buffer_new(bh);
609         } else if (create) {
610                 /* Normal write. */
611                 if (bytes > bh->b_size)
612                         bytes = bh->b_size;
613
614                 if (vbo >= valid)
615                         set_buffer_new(bh);
616
617                 if (vbo + bytes > valid) {
618                         ni->i_valid = vbo + bytes;
619                         mark_inode_dirty(inode);
620                 }
621         } else if (vbo >= valid) {
622                 /* Read out of valid data. */
623                 /* Should never be here 'cause already checked. */
624                 clear_buffer_mapped(bh);
625         } else if (vbo + bytes <= valid) {
626                 /* Normal read. */
627         } else if (vbo + block_size <= valid) {
628                 /* Normal short read. */
629                 bytes = block_size;
630         } else {
631                 /*
632                  * Read across valid size: vbo < valid && valid < vbo + block_size
633                  */
634                 bytes = block_size;
635
636                 if (page) {
637                         u32 voff = valid - vbo;
638
639                         bh->b_size = block_size;
640                         off = vbo & (PAGE_SIZE - 1);
641                         set_bh_page(bh, page, off);
642                         err = bh_read(bh, 0);
643                         if (err < 0)
644                                 goto out;
645                         zero_user_segment(page, off + voff, off + block_size);
646                 }
647         }
648
649         if (bh->b_size > bytes)
650                 bh->b_size = bytes;
651
652 #ifndef __LP64__
653         if (ctx == GET_BLOCK_DIRECT_IO_W || ctx == GET_BLOCK_DIRECT_IO_R) {
654                 static_assert(sizeof(size_t) < sizeof(loff_t));
655                 if (bytes > 0x40000000u)
656                         bh->b_size = 0x40000000u;
657         }
658 #endif
659
660         return 0;
661
662 out:
663         return err;
664 }
665
666 int ntfs_get_block(struct inode *inode, sector_t vbn,
667                    struct buffer_head *bh_result, int create)
668 {
669         return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
670                                   bh_result, create, GET_BLOCK_GENERAL);
671 }
672
673 static int ntfs_get_block_bmap(struct inode *inode, sector_t vsn,
674                                struct buffer_head *bh_result, int create)
675 {
676         return ntfs_get_block_vbo(inode,
677                                   (u64)vsn << inode->i_sb->s_blocksize_bits,
678                                   bh_result, create, GET_BLOCK_BMAP);
679 }
680
681 static sector_t ntfs_bmap(struct address_space *mapping, sector_t block)
682 {
683         return generic_block_bmap(mapping, block, ntfs_get_block_bmap);
684 }
685
686 static int ntfs_read_folio(struct file *file, struct folio *folio)
687 {
688         struct page *page = &folio->page;
689         int err;
690         struct address_space *mapping = page->mapping;
691         struct inode *inode = mapping->host;
692         struct ntfs_inode *ni = ntfs_i(inode);
693
694         if (is_resident(ni)) {
695                 ni_lock(ni);
696                 err = attr_data_read_resident(ni, page);
697                 ni_unlock(ni);
698                 if (err != E_NTFS_NONRESIDENT) {
699                         unlock_page(page);
700                         return err;
701                 }
702         }
703
704         if (is_compressed(ni)) {
705                 ni_lock(ni);
706                 err = ni_readpage_cmpr(ni, page);
707                 ni_unlock(ni);
708                 return err;
709         }
710
711         /* Normal + sparse files. */
712         return mpage_read_folio(folio, ntfs_get_block);
713 }
714
715 static void ntfs_readahead(struct readahead_control *rac)
716 {
717         struct address_space *mapping = rac->mapping;
718         struct inode *inode = mapping->host;
719         struct ntfs_inode *ni = ntfs_i(inode);
720         u64 valid;
721         loff_t pos;
722
723         if (is_resident(ni)) {
724                 /* No readahead for resident. */
725                 return;
726         }
727
728         if (is_compressed(ni)) {
729                 /* No readahead for compressed. */
730                 return;
731         }
732
733         valid = ni->i_valid;
734         pos = readahead_pos(rac);
735
736         if (valid < i_size_read(inode) && pos <= valid &&
737             valid < pos + readahead_length(rac)) {
738                 /* Range cross 'valid'. Read it page by page. */
739                 return;
740         }
741
742         mpage_readahead(rac, ntfs_get_block);
743 }
744
745 static int ntfs_get_block_direct_IO_R(struct inode *inode, sector_t iblock,
746                                       struct buffer_head *bh_result, int create)
747 {
748         return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
749                                   bh_result, create, GET_BLOCK_DIRECT_IO_R);
750 }
751
752 static int ntfs_get_block_direct_IO_W(struct inode *inode, sector_t iblock,
753                                       struct buffer_head *bh_result, int create)
754 {
755         return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
756                                   bh_result, create, GET_BLOCK_DIRECT_IO_W);
757 }
758
759 static ssize_t ntfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
760 {
761         struct file *file = iocb->ki_filp;
762         struct address_space *mapping = file->f_mapping;
763         struct inode *inode = mapping->host;
764         struct ntfs_inode *ni = ntfs_i(inode);
765         loff_t vbo = iocb->ki_pos;
766         loff_t end;
767         int wr = iov_iter_rw(iter) & WRITE;
768         size_t iter_count = iov_iter_count(iter);
769         loff_t valid;
770         ssize_t ret;
771
772         if (is_resident(ni)) {
773                 /* Switch to buffered write. */
774                 ret = 0;
775                 goto out;
776         }
777
778         ret = blockdev_direct_IO(iocb, inode, iter,
779                                  wr ? ntfs_get_block_direct_IO_W
780                                     : ntfs_get_block_direct_IO_R);
781
782         if (ret > 0)
783                 end = vbo + ret;
784         else if (wr && ret == -EIOCBQUEUED)
785                 end = vbo + iter_count;
786         else
787                 goto out;
788
789         valid = ni->i_valid;
790         if (wr) {
791                 if (end > valid && !S_ISBLK(inode->i_mode)) {
792                         ni->i_valid = end;
793                         mark_inode_dirty(inode);
794                 }
795         } else if (vbo < valid && valid < end) {
796                 /* Fix page. */
797                 iov_iter_revert(iter, end - valid);
798                 iov_iter_zero(end - valid, iter);
799         }
800
801 out:
802         return ret;
803 }
804
805 int ntfs_set_size(struct inode *inode, u64 new_size)
806 {
807         struct super_block *sb = inode->i_sb;
808         struct ntfs_sb_info *sbi = sb->s_fs_info;
809         struct ntfs_inode *ni = ntfs_i(inode);
810         int err;
811
812         /* Check for maximum file size. */
813         if (is_sparsed(ni) || is_compressed(ni)) {
814                 if (new_size > sbi->maxbytes_sparse) {
815                         err = -EFBIG;
816                         goto out;
817                 }
818         } else if (new_size > sbi->maxbytes) {
819                 err = -EFBIG;
820                 goto out;
821         }
822
823         ni_lock(ni);
824         down_write(&ni->file.run_lock);
825
826         err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run, new_size,
827                             &ni->i_valid, true, NULL);
828
829         up_write(&ni->file.run_lock);
830         ni_unlock(ni);
831
832         mark_inode_dirty(inode);
833
834 out:
835         return err;
836 }
837
838 static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
839 {
840         struct address_space *mapping = page->mapping;
841         struct inode *inode = mapping->host;
842         struct ntfs_inode *ni = ntfs_i(inode);
843         int err;
844
845         if (is_resident(ni)) {
846                 ni_lock(ni);
847                 err = attr_data_write_resident(ni, page);
848                 ni_unlock(ni);
849                 if (err != E_NTFS_NONRESIDENT) {
850                         unlock_page(page);
851                         return err;
852                 }
853         }
854
855         return block_write_full_page(page, ntfs_get_block, wbc);
856 }
857
858 static int ntfs_writepages(struct address_space *mapping,
859                            struct writeback_control *wbc)
860 {
861         /* Redirect call to 'ntfs_writepage' for resident files. */
862         if (is_resident(ntfs_i(mapping->host)))
863                 return generic_writepages(mapping, wbc);
864         return mpage_writepages(mapping, wbc, ntfs_get_block);
865 }
866
867 static int ntfs_get_block_write_begin(struct inode *inode, sector_t vbn,
868                                       struct buffer_head *bh_result, int create)
869 {
870         return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
871                                   bh_result, create, GET_BLOCK_WRITE_BEGIN);
872 }
873
874 int ntfs_write_begin(struct file *file, struct address_space *mapping,
875                      loff_t pos, u32 len, struct page **pagep, void **fsdata)
876 {
877         int err;
878         struct inode *inode = mapping->host;
879         struct ntfs_inode *ni = ntfs_i(inode);
880
881         *pagep = NULL;
882         if (is_resident(ni)) {
883                 struct page *page = grab_cache_page_write_begin(
884                         mapping, pos >> PAGE_SHIFT);
885
886                 if (!page) {
887                         err = -ENOMEM;
888                         goto out;
889                 }
890
891                 ni_lock(ni);
892                 err = attr_data_read_resident(ni, page);
893                 ni_unlock(ni);
894
895                 if (!err) {
896                         *pagep = page;
897                         goto out;
898                 }
899                 unlock_page(page);
900                 put_page(page);
901
902                 if (err != E_NTFS_NONRESIDENT)
903                         goto out;
904         }
905
906         err = block_write_begin(mapping, pos, len, pagep,
907                                 ntfs_get_block_write_begin);
908
909 out:
910         return err;
911 }
912
913 /*
914  * ntfs_write_end - Address_space_operations::write_end.
915  */
916 int ntfs_write_end(struct file *file, struct address_space *mapping,
917                    loff_t pos, u32 len, u32 copied, struct page *page,
918                    void *fsdata)
919 {
920         struct inode *inode = mapping->host;
921         struct ntfs_inode *ni = ntfs_i(inode);
922         u64 valid = ni->i_valid;
923         bool dirty = false;
924         int err;
925
926         if (is_resident(ni)) {
927                 ni_lock(ni);
928                 err = attr_data_write_resident(ni, page);
929                 ni_unlock(ni);
930                 if (!err) {
931                         dirty = true;
932                         /* Clear any buffers in page. */
933                         if (page_has_buffers(page)) {
934                                 struct buffer_head *head, *bh;
935
936                                 bh = head = page_buffers(page);
937                                 do {
938                                         clear_buffer_dirty(bh);
939                                         clear_buffer_mapped(bh);
940                                         set_buffer_uptodate(bh);
941                                 } while (head != (bh = bh->b_this_page));
942                         }
943                         SetPageUptodate(page);
944                         err = copied;
945                 }
946                 unlock_page(page);
947                 put_page(page);
948         } else {
949                 err = generic_write_end(file, mapping, pos, len, copied, page,
950                                         fsdata);
951         }
952
953         if (err >= 0) {
954                 if (!(ni->std_fa & FILE_ATTRIBUTE_ARCHIVE)) {
955                         inode->i_ctime = inode->i_mtime = current_time(inode);
956                         ni->std_fa |= FILE_ATTRIBUTE_ARCHIVE;
957                         dirty = true;
958                 }
959
960                 if (valid != ni->i_valid) {
961                         /* ni->i_valid is changed in ntfs_get_block_vbo. */
962                         dirty = true;
963                 }
964
965                 if (dirty)
966                         mark_inode_dirty(inode);
967         }
968
969         return err;
970 }
971
972 int reset_log_file(struct inode *inode)
973 {
974         int err;
975         loff_t pos = 0;
976         u32 log_size = inode->i_size;
977         struct address_space *mapping = inode->i_mapping;
978
979         for (;;) {
980                 u32 len;
981                 void *kaddr;
982                 struct page *page;
983
984                 len = pos + PAGE_SIZE > log_size ? (log_size - pos) : PAGE_SIZE;
985
986                 err = block_write_begin(mapping, pos, len, &page,
987                                         ntfs_get_block_write_begin);
988                 if (err)
989                         goto out;
990
991                 kaddr = kmap_atomic(page);
992                 memset(kaddr, -1, len);
993                 kunmap_atomic(kaddr);
994                 flush_dcache_page(page);
995
996                 err = block_write_end(NULL, mapping, pos, len, len, page, NULL);
997                 if (err < 0)
998                         goto out;
999                 pos += len;
1000
1001                 if (pos >= log_size)
1002                         break;
1003                 balance_dirty_pages_ratelimited(mapping);
1004         }
1005 out:
1006         mark_inode_dirty_sync(inode);
1007
1008         return err;
1009 }
1010
1011 int ntfs3_write_inode(struct inode *inode, struct writeback_control *wbc)
1012 {
1013         return _ni_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
1014 }
1015
1016 int ntfs_sync_inode(struct inode *inode)
1017 {
1018         return _ni_write_inode(inode, 1);
1019 }
1020
1021 /*
1022  * writeback_inode - Helper function for ntfs_flush_inodes().
1023  *
1024  * This writes both the inode and the file data blocks, waiting
1025  * for in flight data blocks before the start of the call.  It
1026  * does not wait for any io started during the call.
1027  */
1028 static int writeback_inode(struct inode *inode)
1029 {
1030         int ret = sync_inode_metadata(inode, 0);
1031
1032         if (!ret)
1033                 ret = filemap_fdatawrite(inode->i_mapping);
1034         return ret;
1035 }
1036
1037 /*
1038  * ntfs_flush_inodes
1039  *
1040  * Write data and metadata corresponding to i1 and i2.  The io is
1041  * started but we do not wait for any of it to finish.
1042  *
1043  * filemap_flush() is used for the block device, so if there is a dirty
1044  * page for a block already in flight, we will not wait and start the
1045  * io over again.
1046  */
1047 int ntfs_flush_inodes(struct super_block *sb, struct inode *i1,
1048                       struct inode *i2)
1049 {
1050         int ret = 0;
1051
1052         if (i1)
1053                 ret = writeback_inode(i1);
1054         if (!ret && i2)
1055                 ret = writeback_inode(i2);
1056         if (!ret)
1057                 ret = sync_blockdev_nowait(sb->s_bdev);
1058         return ret;
1059 }
1060
1061 int inode_write_data(struct inode *inode, const void *data, size_t bytes)
1062 {
1063         pgoff_t idx;
1064
1065         /* Write non resident data. */
1066         for (idx = 0; bytes; idx++) {
1067                 size_t op = bytes > PAGE_SIZE ? PAGE_SIZE : bytes;
1068                 struct page *page = ntfs_map_page(inode->i_mapping, idx);
1069
1070                 if (IS_ERR(page))
1071                         return PTR_ERR(page);
1072
1073                 lock_page(page);
1074                 WARN_ON(!PageUptodate(page));
1075                 ClearPageUptodate(page);
1076
1077                 memcpy(page_address(page), data, op);
1078
1079                 flush_dcache_page(page);
1080                 SetPageUptodate(page);
1081                 unlock_page(page);
1082
1083                 ntfs_unmap_page(page);
1084
1085                 bytes -= op;
1086                 data = Add2Ptr(data, PAGE_SIZE);
1087         }
1088         return 0;
1089 }
1090
1091 /*
1092  * ntfs_reparse_bytes
1093  *
1094  * Number of bytes for REPARSE_DATA_BUFFER(IO_REPARSE_TAG_SYMLINK)
1095  * for unicode string of @uni_len length.
1096  */
1097 static inline u32 ntfs_reparse_bytes(u32 uni_len)
1098 {
1099         /* Header + unicode string + decorated unicode string. */
1100         return sizeof(short) * (2 * uni_len + 4) +
1101                offsetof(struct REPARSE_DATA_BUFFER,
1102                         SymbolicLinkReparseBuffer.PathBuffer);
1103 }
1104
1105 static struct REPARSE_DATA_BUFFER *
1106 ntfs_create_reparse_buffer(struct ntfs_sb_info *sbi, const char *symname,
1107                            u32 size, u16 *nsize)
1108 {
1109         int i, err;
1110         struct REPARSE_DATA_BUFFER *rp;
1111         __le16 *rp_name;
1112         typeof(rp->SymbolicLinkReparseBuffer) *rs;
1113
1114         rp = kzalloc(ntfs_reparse_bytes(2 * size + 2), GFP_NOFS);
1115         if (!rp)
1116                 return ERR_PTR(-ENOMEM);
1117
1118         rs = &rp->SymbolicLinkReparseBuffer;
1119         rp_name = rs->PathBuffer;
1120
1121         /* Convert link name to UTF-16. */
1122         err = ntfs_nls_to_utf16(sbi, symname, size,
1123                                 (struct cpu_str *)(rp_name - 1), 2 * size,
1124                                 UTF16_LITTLE_ENDIAN);
1125         if (err < 0)
1126                 goto out;
1127
1128         /* err = the length of unicode name of symlink. */
1129         *nsize = ntfs_reparse_bytes(err);
1130
1131         if (*nsize > sbi->reparse.max_size) {
1132                 err = -EFBIG;
1133                 goto out;
1134         }
1135
1136         /* Translate Linux '/' into Windows '\'. */
1137         for (i = 0; i < err; i++) {
1138                 if (rp_name[i] == cpu_to_le16('/'))
1139                         rp_name[i] = cpu_to_le16('\\');
1140         }
1141
1142         rp->ReparseTag = IO_REPARSE_TAG_SYMLINK;
1143         rp->ReparseDataLength =
1144                 cpu_to_le16(*nsize - offsetof(struct REPARSE_DATA_BUFFER,
1145                                               SymbolicLinkReparseBuffer));
1146
1147         /* PrintName + SubstituteName. */
1148         rs->SubstituteNameOffset = cpu_to_le16(sizeof(short) * err);
1149         rs->SubstituteNameLength = cpu_to_le16(sizeof(short) * err + 8);
1150         rs->PrintNameLength = rs->SubstituteNameOffset;
1151
1152         /*
1153          * TODO: Use relative path if possible to allow Windows to
1154          * parse this path.
1155          * 0-absolute path 1- relative path (SYMLINK_FLAG_RELATIVE).
1156          */
1157         rs->Flags = 0;
1158
1159         memmove(rp_name + err + 4, rp_name, sizeof(short) * err);
1160
1161         /* Decorate SubstituteName. */
1162         rp_name += err;
1163         rp_name[0] = cpu_to_le16('\\');
1164         rp_name[1] = cpu_to_le16('?');
1165         rp_name[2] = cpu_to_le16('?');
1166         rp_name[3] = cpu_to_le16('\\');
1167
1168         return rp;
1169 out:
1170         kfree(rp);
1171         return ERR_PTR(err);
1172 }
1173
1174 struct inode *ntfs_create_inode(struct user_namespace *mnt_userns,
1175                                 struct inode *dir, struct dentry *dentry,
1176                                 const struct cpu_str *uni, umode_t mode,
1177                                 dev_t dev, const char *symname, u32 size,
1178                                 struct ntfs_fnd *fnd)
1179 {
1180         int err;
1181         struct super_block *sb = dir->i_sb;
1182         struct ntfs_sb_info *sbi = sb->s_fs_info;
1183         const struct qstr *name = &dentry->d_name;
1184         CLST ino = 0;
1185         struct ntfs_inode *dir_ni = ntfs_i(dir);
1186         struct ntfs_inode *ni = NULL;
1187         struct inode *inode = NULL;
1188         struct ATTRIB *attr;
1189         struct ATTR_STD_INFO5 *std5;
1190         struct ATTR_FILE_NAME *fname;
1191         struct MFT_REC *rec;
1192         u32 asize, dsize, sd_size;
1193         enum FILE_ATTRIBUTE fa;
1194         __le32 security_id = SECURITY_ID_INVALID;
1195         CLST vcn;
1196         const void *sd;
1197         u16 t16, nsize = 0, aid = 0;
1198         struct INDEX_ROOT *root, *dir_root;
1199         struct NTFS_DE *e, *new_de = NULL;
1200         struct REPARSE_DATA_BUFFER *rp = NULL;
1201         bool rp_inserted = false;
1202
1203         ni_lock_dir(dir_ni);
1204
1205         dir_root = indx_get_root(&dir_ni->dir, dir_ni, NULL, NULL);
1206         if (!dir_root) {
1207                 err = -EINVAL;
1208                 goto out1;
1209         }
1210
1211         if (S_ISDIR(mode)) {
1212                 /* Use parent's directory attributes. */
1213                 fa = dir_ni->std_fa | FILE_ATTRIBUTE_DIRECTORY |
1214                      FILE_ATTRIBUTE_ARCHIVE;
1215                 /*
1216                  * By default child directory inherits parent attributes.
1217                  * Root directory is hidden + system.
1218                  * Make an exception for children in root.
1219                  */
1220                 if (dir->i_ino == MFT_REC_ROOT)
1221                         fa &= ~(FILE_ATTRIBUTE_HIDDEN | FILE_ATTRIBUTE_SYSTEM);
1222         } else if (S_ISLNK(mode)) {
1223                 /* It is good idea that link should be the same type (file/dir) as target */
1224                 fa = FILE_ATTRIBUTE_REPARSE_POINT;
1225
1226                 /*
1227                  * Linux: there are dir/file/symlink and so on.
1228                  * NTFS: symlinks are "dir + reparse" or "file + reparse"
1229                  * It is good idea to create:
1230                  * dir + reparse if 'symname' points to directory
1231                  * or
1232                  * file + reparse if 'symname' points to file
1233                  * Unfortunately kern_path hangs if symname contains 'dir'.
1234                  */
1235
1236                 /*
1237                  *      struct path path;
1238                  *
1239                  *      if (!kern_path(symname, LOOKUP_FOLLOW, &path)){
1240                  *              struct inode *target = d_inode(path.dentry);
1241                  *
1242                  *              if (S_ISDIR(target->i_mode))
1243                  *                      fa |= FILE_ATTRIBUTE_DIRECTORY;
1244                  *              // if ( target->i_sb == sb ){
1245                  *              //      use relative path?
1246                  *              // }
1247                  *              path_put(&path);
1248                  *      }
1249                  */
1250         } else if (S_ISREG(mode)) {
1251                 if (sbi->options->sparse) {
1252                         /* Sparsed regular file, cause option 'sparse'. */
1253                         fa = FILE_ATTRIBUTE_SPARSE_FILE |
1254                              FILE_ATTRIBUTE_ARCHIVE;
1255                 } else if (dir_ni->std_fa & FILE_ATTRIBUTE_COMPRESSED) {
1256                         /* Compressed regular file, if parent is compressed. */
1257                         fa = FILE_ATTRIBUTE_COMPRESSED | FILE_ATTRIBUTE_ARCHIVE;
1258                 } else {
1259                         /* Regular file, default attributes. */
1260                         fa = FILE_ATTRIBUTE_ARCHIVE;
1261                 }
1262         } else {
1263                 fa = FILE_ATTRIBUTE_ARCHIVE;
1264         }
1265
1266         if (!(mode & 0222))
1267                 fa |= FILE_ATTRIBUTE_READONLY;
1268
1269         /* Allocate PATH_MAX bytes. */
1270         new_de = __getname();
1271         if (!new_de) {
1272                 err = -ENOMEM;
1273                 goto out1;
1274         }
1275
1276         /* Mark rw ntfs as dirty. it will be cleared at umount. */
1277         ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
1278
1279         /* Step 1: allocate and fill new mft record. */
1280         err = ntfs_look_free_mft(sbi, &ino, false, NULL, NULL);
1281         if (err)
1282                 goto out2;
1283
1284         ni = ntfs_new_inode(sbi, ino, fa & FILE_ATTRIBUTE_DIRECTORY);
1285         if (IS_ERR(ni)) {
1286                 err = PTR_ERR(ni);
1287                 ni = NULL;
1288                 goto out3;
1289         }
1290         inode = &ni->vfs_inode;
1291         inode_init_owner(mnt_userns, inode, dir, mode);
1292         mode = inode->i_mode;
1293
1294         inode->i_atime = inode->i_mtime = inode->i_ctime = ni->i_crtime =
1295                 current_time(inode);
1296
1297         rec = ni->mi.mrec;
1298         rec->hard_links = cpu_to_le16(1);
1299         attr = Add2Ptr(rec, le16_to_cpu(rec->attr_off));
1300
1301         /* Get default security id. */
1302         sd = s_default_security;
1303         sd_size = sizeof(s_default_security);
1304
1305         if (is_ntfs3(sbi)) {
1306                 security_id = dir_ni->std_security_id;
1307                 if (le32_to_cpu(security_id) < SECURITY_ID_FIRST) {
1308                         security_id = sbi->security.def_security_id;
1309
1310                         if (security_id == SECURITY_ID_INVALID &&
1311                             !ntfs_insert_security(sbi, sd, sd_size,
1312                                                   &security_id, NULL))
1313                                 sbi->security.def_security_id = security_id;
1314                 }
1315         }
1316
1317         /* Insert standard info. */
1318         std5 = Add2Ptr(attr, SIZEOF_RESIDENT);
1319
1320         if (security_id == SECURITY_ID_INVALID) {
1321                 dsize = sizeof(struct ATTR_STD_INFO);
1322         } else {
1323                 dsize = sizeof(struct ATTR_STD_INFO5);
1324                 std5->security_id = security_id;
1325                 ni->std_security_id = security_id;
1326         }
1327         asize = SIZEOF_RESIDENT + dsize;
1328
1329         attr->type = ATTR_STD;
1330         attr->size = cpu_to_le32(asize);
1331         attr->id = cpu_to_le16(aid++);
1332         attr->res.data_off = SIZEOF_RESIDENT_LE;
1333         attr->res.data_size = cpu_to_le32(dsize);
1334
1335         std5->cr_time = std5->m_time = std5->c_time = std5->a_time =
1336                 kernel2nt(&inode->i_atime);
1337
1338         ni->std_fa = fa;
1339         std5->fa = fa;
1340
1341         attr = Add2Ptr(attr, asize);
1342
1343         /* Insert file name. */
1344         err = fill_name_de(sbi, new_de, name, uni);
1345         if (err)
1346                 goto out4;
1347
1348         mi_get_ref(&ni->mi, &new_de->ref);
1349
1350         fname = (struct ATTR_FILE_NAME *)(new_de + 1);
1351         mi_get_ref(&dir_ni->mi, &fname->home);
1352         fname->dup.cr_time = fname->dup.m_time = fname->dup.c_time =
1353                 fname->dup.a_time = std5->cr_time;
1354         fname->dup.alloc_size = fname->dup.data_size = 0;
1355         fname->dup.fa = std5->fa;
1356         fname->dup.ea_size = fname->dup.reparse = 0;
1357
1358         dsize = le16_to_cpu(new_de->key_size);
1359         asize = ALIGN(SIZEOF_RESIDENT + dsize, 8);
1360
1361         attr->type = ATTR_NAME;
1362         attr->size = cpu_to_le32(asize);
1363         attr->res.data_off = SIZEOF_RESIDENT_LE;
1364         attr->res.flags = RESIDENT_FLAG_INDEXED;
1365         attr->id = cpu_to_le16(aid++);
1366         attr->res.data_size = cpu_to_le32(dsize);
1367         memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), fname, dsize);
1368
1369         attr = Add2Ptr(attr, asize);
1370
1371         if (security_id == SECURITY_ID_INVALID) {
1372                 /* Insert security attribute. */
1373                 asize = SIZEOF_RESIDENT + ALIGN(sd_size, 8);
1374
1375                 attr->type = ATTR_SECURE;
1376                 attr->size = cpu_to_le32(asize);
1377                 attr->id = cpu_to_le16(aid++);
1378                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1379                 attr->res.data_size = cpu_to_le32(sd_size);
1380                 memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), sd, sd_size);
1381
1382                 attr = Add2Ptr(attr, asize);
1383         }
1384
1385         attr->id = cpu_to_le16(aid++);
1386         if (fa & FILE_ATTRIBUTE_DIRECTORY) {
1387                 /*
1388                  * Regular directory or symlink to directory.
1389                  * Create root attribute.
1390                  */
1391                 dsize = sizeof(struct INDEX_ROOT) + sizeof(struct NTFS_DE);
1392                 asize = sizeof(I30_NAME) + SIZEOF_RESIDENT + dsize;
1393
1394                 attr->type = ATTR_ROOT;
1395                 attr->size = cpu_to_le32(asize);
1396
1397                 attr->name_len = ARRAY_SIZE(I30_NAME);
1398                 attr->name_off = SIZEOF_RESIDENT_LE;
1399                 attr->res.data_off =
1400                         cpu_to_le16(sizeof(I30_NAME) + SIZEOF_RESIDENT);
1401                 attr->res.data_size = cpu_to_le32(dsize);
1402                 memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), I30_NAME,
1403                        sizeof(I30_NAME));
1404
1405                 root = Add2Ptr(attr, sizeof(I30_NAME) + SIZEOF_RESIDENT);
1406                 memcpy(root, dir_root, offsetof(struct INDEX_ROOT, ihdr));
1407                 root->ihdr.de_off =
1408                         cpu_to_le32(sizeof(struct INDEX_HDR)); // 0x10
1409                 root->ihdr.used = cpu_to_le32(sizeof(struct INDEX_HDR) +
1410                                               sizeof(struct NTFS_DE));
1411                 root->ihdr.total = root->ihdr.used;
1412
1413                 e = Add2Ptr(root, sizeof(struct INDEX_ROOT));
1414                 e->size = cpu_to_le16(sizeof(struct NTFS_DE));
1415                 e->flags = NTFS_IE_LAST;
1416         } else if (S_ISLNK(mode)) {
1417                 /*
1418                  * Symlink to file.
1419                  * Create empty resident data attribute.
1420                  */
1421                 asize = SIZEOF_RESIDENT;
1422
1423                 /* Insert empty ATTR_DATA */
1424                 attr->type = ATTR_DATA;
1425                 attr->size = cpu_to_le32(SIZEOF_RESIDENT);
1426                 attr->name_off = SIZEOF_RESIDENT_LE;
1427                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1428         } else if (S_ISREG(mode)) {
1429                 /*
1430                  * Regular file. Create empty non resident data attribute.
1431                  */
1432                 attr->type = ATTR_DATA;
1433                 attr->non_res = 1;
1434                 attr->nres.evcn = cpu_to_le64(-1ll);
1435                 if (fa & FILE_ATTRIBUTE_SPARSE_FILE) {
1436                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
1437                         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
1438                         attr->flags = ATTR_FLAG_SPARSED;
1439                         asize = SIZEOF_NONRESIDENT_EX + 8;
1440                 } else if (fa & FILE_ATTRIBUTE_COMPRESSED) {
1441                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
1442                         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
1443                         attr->flags = ATTR_FLAG_COMPRESSED;
1444                         attr->nres.c_unit = COMPRESSION_UNIT;
1445                         asize = SIZEOF_NONRESIDENT_EX + 8;
1446                 } else {
1447                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT + 8);
1448                         attr->name_off = SIZEOF_NONRESIDENT_LE;
1449                         asize = SIZEOF_NONRESIDENT + 8;
1450                 }
1451                 attr->nres.run_off = attr->name_off;
1452         } else {
1453                 /*
1454                  * Node. Create empty resident data attribute.
1455                  */
1456                 attr->type = ATTR_DATA;
1457                 attr->size = cpu_to_le32(SIZEOF_RESIDENT);
1458                 attr->name_off = SIZEOF_RESIDENT_LE;
1459                 if (fa & FILE_ATTRIBUTE_SPARSE_FILE)
1460                         attr->flags = ATTR_FLAG_SPARSED;
1461                 else if (fa & FILE_ATTRIBUTE_COMPRESSED)
1462                         attr->flags = ATTR_FLAG_COMPRESSED;
1463                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1464                 asize = SIZEOF_RESIDENT;
1465                 ni->ni_flags |= NI_FLAG_RESIDENT;
1466         }
1467
1468         if (S_ISDIR(mode)) {
1469                 ni->ni_flags |= NI_FLAG_DIR;
1470                 err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
1471                 if (err)
1472                         goto out4;
1473         } else if (S_ISLNK(mode)) {
1474                 rp = ntfs_create_reparse_buffer(sbi, symname, size, &nsize);
1475
1476                 if (IS_ERR(rp)) {
1477                         err = PTR_ERR(rp);
1478                         rp = NULL;
1479                         goto out4;
1480                 }
1481
1482                 /*
1483                  * Insert ATTR_REPARSE.
1484                  */
1485                 attr = Add2Ptr(attr, asize);
1486                 attr->type = ATTR_REPARSE;
1487                 attr->id = cpu_to_le16(aid++);
1488
1489                 /* Resident or non resident? */
1490                 asize = ALIGN(SIZEOF_RESIDENT + nsize, 8);
1491                 t16 = PtrOffset(rec, attr);
1492
1493                 /*
1494                  * Below function 'ntfs_save_wsl_perm' requires 0x78 bytes.
1495                  * It is good idea to keep extened attributes resident.
1496                  */
1497                 if (asize + t16 + 0x78 + 8 > sbi->record_size) {
1498                         CLST alen;
1499                         CLST clst = bytes_to_cluster(sbi, nsize);
1500
1501                         /* Bytes per runs. */
1502                         t16 = sbi->record_size - t16 - SIZEOF_NONRESIDENT;
1503
1504                         attr->non_res = 1;
1505                         attr->nres.evcn = cpu_to_le64(clst - 1);
1506                         attr->name_off = SIZEOF_NONRESIDENT_LE;
1507                         attr->nres.run_off = attr->name_off;
1508                         attr->nres.data_size = cpu_to_le64(nsize);
1509                         attr->nres.valid_size = attr->nres.data_size;
1510                         attr->nres.alloc_size =
1511                                 cpu_to_le64(ntfs_up_cluster(sbi, nsize));
1512
1513                         err = attr_allocate_clusters(sbi, &ni->file.run, 0, 0,
1514                                                      clst, NULL, 0, &alen, 0,
1515                                                      NULL);
1516                         if (err)
1517                                 goto out5;
1518
1519                         err = run_pack(&ni->file.run, 0, clst,
1520                                        Add2Ptr(attr, SIZEOF_NONRESIDENT), t16,
1521                                        &vcn);
1522                         if (err < 0)
1523                                 goto out5;
1524
1525                         if (vcn != clst) {
1526                                 err = -EINVAL;
1527                                 goto out5;
1528                         }
1529
1530                         asize = SIZEOF_NONRESIDENT + ALIGN(err, 8);
1531                 } else {
1532                         attr->res.data_off = SIZEOF_RESIDENT_LE;
1533                         attr->res.data_size = cpu_to_le32(nsize);
1534                         memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), rp, nsize);
1535                         nsize = 0;
1536                 }
1537                 /* Size of symlink equals the length of input string. */
1538                 inode->i_size = size;
1539
1540                 attr->size = cpu_to_le32(asize);
1541
1542                 err = ntfs_insert_reparse(sbi, IO_REPARSE_TAG_SYMLINK,
1543                                           &new_de->ref);
1544                 if (err)
1545                         goto out5;
1546
1547                 rp_inserted = true;
1548         }
1549
1550         attr = Add2Ptr(attr, asize);
1551         attr->type = ATTR_END;
1552
1553         rec->used = cpu_to_le32(PtrOffset(rec, attr) + 8);
1554         rec->next_attr_id = cpu_to_le16(aid);
1555
1556         /* Step 2: Add new name in index. */
1557         err = indx_insert_entry(&dir_ni->dir, dir_ni, new_de, sbi, fnd, 0);
1558         if (err)
1559                 goto out6;
1560
1561         /* Unlock parent directory before ntfs_init_acl. */
1562         ni_unlock(dir_ni);
1563
1564         inode->i_generation = le16_to_cpu(rec->seq);
1565
1566         dir->i_mtime = dir->i_ctime = inode->i_atime;
1567
1568         if (S_ISDIR(mode)) {
1569                 inode->i_op = &ntfs_dir_inode_operations;
1570                 inode->i_fop = &ntfs_dir_operations;
1571         } else if (S_ISLNK(mode)) {
1572                 inode->i_op = &ntfs_link_inode_operations;
1573                 inode->i_fop = NULL;
1574                 inode->i_mapping->a_ops = &ntfs_aops;
1575                 inode->i_size = size;
1576                 inode_nohighmem(inode);
1577         } else if (S_ISREG(mode)) {
1578                 inode->i_op = &ntfs_file_inode_operations;
1579                 inode->i_fop = &ntfs_file_operations;
1580                 inode->i_mapping->a_ops =
1581                         is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
1582                 init_rwsem(&ni->file.run_lock);
1583         } else {
1584                 inode->i_op = &ntfs_special_inode_operations;
1585                 init_special_inode(inode, mode, dev);
1586         }
1587
1588 #ifdef CONFIG_NTFS3_FS_POSIX_ACL
1589         if (!S_ISLNK(mode) && (sb->s_flags & SB_POSIXACL)) {
1590                 err = ntfs_init_acl(mnt_userns, inode, dir);
1591                 if (err)
1592                         goto out7;
1593         } else
1594 #endif
1595         {
1596                 inode->i_flags |= S_NOSEC;
1597         }
1598
1599         /* Write non resident data. */
1600         if (nsize) {
1601                 err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rp, nsize, 0);
1602                 if (err)
1603                         goto out7;
1604         }
1605
1606         /*
1607          * Call 'd_instantiate' after inode->i_op is set
1608          * but before finish_open.
1609          */
1610         d_instantiate(dentry, inode);
1611
1612         ntfs_save_wsl_perm(inode);
1613         mark_inode_dirty(dir);
1614         mark_inode_dirty(inode);
1615
1616         /* Normal exit. */
1617         goto out2;
1618
1619 out7:
1620
1621         /* Undo 'indx_insert_entry'. */
1622         ni_lock_dir(dir_ni);
1623         indx_delete_entry(&dir_ni->dir, dir_ni, new_de + 1,
1624                           le16_to_cpu(new_de->key_size), sbi);
1625         /* ni_unlock(dir_ni); will be called later. */
1626 out6:
1627         if (rp_inserted)
1628                 ntfs_remove_reparse(sbi, IO_REPARSE_TAG_SYMLINK, &new_de->ref);
1629
1630 out5:
1631         if (S_ISDIR(mode) || run_is_empty(&ni->file.run))
1632                 goto out4;
1633
1634         run_deallocate(sbi, &ni->file.run, false);
1635
1636 out4:
1637         clear_rec_inuse(rec);
1638         clear_nlink(inode);
1639         ni->mi.dirty = false;
1640         discard_new_inode(inode);
1641 out3:
1642         ntfs_mark_rec_free(sbi, ino, false);
1643
1644 out2:
1645         __putname(new_de);
1646         kfree(rp);
1647
1648 out1:
1649         if (err) {
1650                 ni_unlock(dir_ni);
1651                 return ERR_PTR(err);
1652         }
1653
1654         unlock_new_inode(inode);
1655
1656         return inode;
1657 }
1658
1659 int ntfs_link_inode(struct inode *inode, struct dentry *dentry)
1660 {
1661         int err;
1662         struct ntfs_inode *ni = ntfs_i(inode);
1663         struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
1664         struct NTFS_DE *de;
1665
1666         /* Allocate PATH_MAX bytes. */
1667         de = __getname();
1668         if (!de)
1669                 return -ENOMEM;
1670
1671         /* Mark rw ntfs as dirty. It will be cleared at umount. */
1672         ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
1673
1674         /* Construct 'de'. */
1675         err = fill_name_de(sbi, de, &dentry->d_name, NULL);
1676         if (err)
1677                 goto out;
1678
1679         err = ni_add_name(ntfs_i(d_inode(dentry->d_parent)), ni, de);
1680 out:
1681         __putname(de);
1682         return err;
1683 }
1684
1685 /*
1686  * ntfs_unlink_inode
1687  *
1688  * inode_operations::unlink
1689  * inode_operations::rmdir
1690  */
1691 int ntfs_unlink_inode(struct inode *dir, const struct dentry *dentry)
1692 {
1693         int err;
1694         struct ntfs_sb_info *sbi = dir->i_sb->s_fs_info;
1695         struct inode *inode = d_inode(dentry);
1696         struct ntfs_inode *ni = ntfs_i(inode);
1697         struct ntfs_inode *dir_ni = ntfs_i(dir);
1698         struct NTFS_DE *de, *de2 = NULL;
1699         int undo_remove;
1700
1701         if (ntfs_is_meta_file(sbi, ni->mi.rno))
1702                 return -EINVAL;
1703
1704         /* Allocate PATH_MAX bytes. */
1705         de = __getname();
1706         if (!de)
1707                 return -ENOMEM;
1708
1709         ni_lock(ni);
1710
1711         if (S_ISDIR(inode->i_mode) && !dir_is_empty(inode)) {
1712                 err = -ENOTEMPTY;
1713                 goto out;
1714         }
1715
1716         err = fill_name_de(sbi, de, &dentry->d_name, NULL);
1717         if (err < 0)
1718                 goto out;
1719
1720         undo_remove = 0;
1721         err = ni_remove_name(dir_ni, ni, de, &de2, &undo_remove);
1722
1723         if (!err) {
1724                 drop_nlink(inode);
1725                 dir->i_mtime = dir->i_ctime = current_time(dir);
1726                 mark_inode_dirty(dir);
1727                 inode->i_ctime = dir->i_ctime;
1728                 if (inode->i_nlink)
1729                         mark_inode_dirty(inode);
1730         } else if (!ni_remove_name_undo(dir_ni, ni, de, de2, undo_remove)) {
1731                 _ntfs_bad_inode(inode);
1732         } else {
1733                 if (ni_is_dirty(dir))
1734                         mark_inode_dirty(dir);
1735                 if (ni_is_dirty(inode))
1736                         mark_inode_dirty(inode);
1737         }
1738
1739 out:
1740         ni_unlock(ni);
1741         __putname(de);
1742         return err;
1743 }
1744
1745 void ntfs_evict_inode(struct inode *inode)
1746 {
1747         truncate_inode_pages_final(&inode->i_data);
1748
1749         if (inode->i_nlink)
1750                 _ni_write_inode(inode, inode_needs_sync(inode));
1751
1752         invalidate_inode_buffers(inode);
1753         clear_inode(inode);
1754
1755         ni_clear(ntfs_i(inode));
1756 }
1757
1758 static noinline int ntfs_readlink_hlp(struct inode *inode, char *buffer,
1759                                       int buflen)
1760 {
1761         int i, err = -EINVAL;
1762         struct ntfs_inode *ni = ntfs_i(inode);
1763         struct super_block *sb = inode->i_sb;
1764         struct ntfs_sb_info *sbi = sb->s_fs_info;
1765         u64 size;
1766         u16 ulen = 0;
1767         void *to_free = NULL;
1768         struct REPARSE_DATA_BUFFER *rp;
1769         const __le16 *uname;
1770         struct ATTRIB *attr;
1771
1772         /* Reparse data present. Try to parse it. */
1773         static_assert(!offsetof(struct REPARSE_DATA_BUFFER, ReparseTag));
1774         static_assert(sizeof(u32) == sizeof(rp->ReparseTag));
1775
1776         *buffer = 0;
1777
1778         attr = ni_find_attr(ni, NULL, NULL, ATTR_REPARSE, NULL, 0, NULL, NULL);
1779         if (!attr)
1780                 goto out;
1781
1782         if (!attr->non_res) {
1783                 rp = resident_data_ex(attr, sizeof(struct REPARSE_DATA_BUFFER));
1784                 if (!rp)
1785                         goto out;
1786                 size = le32_to_cpu(attr->res.data_size);
1787         } else {
1788                 size = le64_to_cpu(attr->nres.data_size);
1789                 rp = NULL;
1790         }
1791
1792         if (size > sbi->reparse.max_size || size <= sizeof(u32))
1793                 goto out;
1794
1795         if (!rp) {
1796                 rp = kmalloc(size, GFP_NOFS);
1797                 if (!rp) {
1798                         err = -ENOMEM;
1799                         goto out;
1800                 }
1801                 to_free = rp;
1802                 /* Read into temporal buffer. */
1803                 err = ntfs_read_run_nb(sbi, &ni->file.run, 0, rp, size, NULL);
1804                 if (err)
1805                         goto out;
1806         }
1807
1808         /* Microsoft Tag. */
1809         switch (rp->ReparseTag) {
1810         case IO_REPARSE_TAG_MOUNT_POINT:
1811                 /* Mount points and junctions. */
1812                 /* Can we use 'Rp->MountPointReparseBuffer.PrintNameLength'? */
1813                 if (size <= offsetof(struct REPARSE_DATA_BUFFER,
1814                                      MountPointReparseBuffer.PathBuffer))
1815                         goto out;
1816                 uname = Add2Ptr(rp,
1817                                 offsetof(struct REPARSE_DATA_BUFFER,
1818                                          MountPointReparseBuffer.PathBuffer) +
1819                                         le16_to_cpu(rp->MountPointReparseBuffer
1820                                                             .PrintNameOffset));
1821                 ulen = le16_to_cpu(rp->MountPointReparseBuffer.PrintNameLength);
1822                 break;
1823
1824         case IO_REPARSE_TAG_SYMLINK:
1825                 /* FolderSymbolicLink */
1826                 /* Can we use 'Rp->SymbolicLinkReparseBuffer.PrintNameLength'? */
1827                 if (size <= offsetof(struct REPARSE_DATA_BUFFER,
1828                                      SymbolicLinkReparseBuffer.PathBuffer))
1829                         goto out;
1830                 uname = Add2Ptr(
1831                         rp, offsetof(struct REPARSE_DATA_BUFFER,
1832                                      SymbolicLinkReparseBuffer.PathBuffer) +
1833                                     le16_to_cpu(rp->SymbolicLinkReparseBuffer
1834                                                         .PrintNameOffset));
1835                 ulen = le16_to_cpu(
1836                         rp->SymbolicLinkReparseBuffer.PrintNameLength);
1837                 break;
1838
1839         case IO_REPARSE_TAG_CLOUD:
1840         case IO_REPARSE_TAG_CLOUD_1:
1841         case IO_REPARSE_TAG_CLOUD_2:
1842         case IO_REPARSE_TAG_CLOUD_3:
1843         case IO_REPARSE_TAG_CLOUD_4:
1844         case IO_REPARSE_TAG_CLOUD_5:
1845         case IO_REPARSE_TAG_CLOUD_6:
1846         case IO_REPARSE_TAG_CLOUD_7:
1847         case IO_REPARSE_TAG_CLOUD_8:
1848         case IO_REPARSE_TAG_CLOUD_9:
1849         case IO_REPARSE_TAG_CLOUD_A:
1850         case IO_REPARSE_TAG_CLOUD_B:
1851         case IO_REPARSE_TAG_CLOUD_C:
1852         case IO_REPARSE_TAG_CLOUD_D:
1853         case IO_REPARSE_TAG_CLOUD_E:
1854         case IO_REPARSE_TAG_CLOUD_F:
1855                 err = sizeof("OneDrive") - 1;
1856                 if (err > buflen)
1857                         err = buflen;
1858                 memcpy(buffer, "OneDrive", err);
1859                 goto out;
1860
1861         default:
1862                 if (IsReparseTagMicrosoft(rp->ReparseTag)) {
1863                         /* Unknown Microsoft Tag. */
1864                         goto out;
1865                 }
1866                 if (!IsReparseTagNameSurrogate(rp->ReparseTag) ||
1867                     size <= sizeof(struct REPARSE_POINT)) {
1868                         goto out;
1869                 }
1870
1871                 /* Users tag. */
1872                 uname = Add2Ptr(rp, sizeof(struct REPARSE_POINT));
1873                 ulen = le16_to_cpu(rp->ReparseDataLength) -
1874                        sizeof(struct REPARSE_POINT);
1875         }
1876
1877         /* Convert nlen from bytes to UNICODE chars. */
1878         ulen >>= 1;
1879
1880         /* Check that name is available. */
1881         if (!ulen || uname + ulen > (__le16 *)Add2Ptr(rp, size))
1882                 goto out;
1883
1884         /* If name is already zero terminated then truncate it now. */
1885         if (!uname[ulen - 1])
1886                 ulen -= 1;
1887
1888         err = ntfs_utf16_to_nls(sbi, uname, ulen, buffer, buflen);
1889
1890         if (err < 0)
1891                 goto out;
1892
1893         /* Translate Windows '\' into Linux '/'. */
1894         for (i = 0; i < err; i++) {
1895                 if (buffer[i] == '\\')
1896                         buffer[i] = '/';
1897         }
1898
1899         /* Always set last zero. */
1900         buffer[err] = 0;
1901 out:
1902         kfree(to_free);
1903         return err;
1904 }
1905
1906 static const char *ntfs_get_link(struct dentry *de, struct inode *inode,
1907                                  struct delayed_call *done)
1908 {
1909         int err;
1910         char *ret;
1911
1912         if (!de)
1913                 return ERR_PTR(-ECHILD);
1914
1915         ret = kmalloc(PAGE_SIZE, GFP_NOFS);
1916         if (!ret)
1917                 return ERR_PTR(-ENOMEM);
1918
1919         err = ntfs_readlink_hlp(inode, ret, PAGE_SIZE);
1920         if (err < 0) {
1921                 kfree(ret);
1922                 return ERR_PTR(err);
1923         }
1924
1925         set_delayed_call(done, kfree_link, ret);
1926
1927         return ret;
1928 }
1929
1930 // clang-format off
1931 const struct inode_operations ntfs_link_inode_operations = {
1932         .get_link       = ntfs_get_link,
1933         .setattr        = ntfs3_setattr,
1934         .listxattr      = ntfs_listxattr,
1935         .permission     = ntfs_permission,
1936 };
1937
1938 const struct address_space_operations ntfs_aops = {
1939         .read_folio     = ntfs_read_folio,
1940         .readahead      = ntfs_readahead,
1941         .writepage      = ntfs_writepage,
1942         .writepages     = ntfs_writepages,
1943         .write_begin    = ntfs_write_begin,
1944         .write_end      = ntfs_write_end,
1945         .direct_IO      = ntfs_direct_IO,
1946         .bmap           = ntfs_bmap,
1947         .dirty_folio    = block_dirty_folio,
1948         .invalidate_folio = block_invalidate_folio,
1949 };
1950
1951 const struct address_space_operations ntfs_aops_cmpr = {
1952         .read_folio     = ntfs_read_folio,
1953         .readahead      = ntfs_readahead,
1954 };
1955 // clang-format on