]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/file.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/file.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
617ba13b | 15 | * ext4 fs regular file handling primitives |
ac27a0ec DK |
16 | * |
17 | * 64-bit file support on 64-bit platforms by Jakub Jelinek | |
18 | * (jj@sunsite.ms.mff.cuni.cz) | |
19 | */ | |
20 | ||
21 | #include <linux/time.h> | |
22 | #include <linux/fs.h> | |
bc0b0d6d TT |
23 | #include <linux/mount.h> |
24 | #include <linux/path.h> | |
c94c2acf | 25 | #include <linux/dax.h> |
871a2931 | 26 | #include <linux/quotaops.h> |
c8c0df24 | 27 | #include <linux/pagevec.h> |
e2e40f2c | 28 | #include <linux/uio.h> |
3dcf5451 CH |
29 | #include "ext4.h" |
30 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
31 | #include "xattr.h" |
32 | #include "acl.h" | |
33 | ||
34 | /* | |
35 | * Called when an inode is released. Note that this is different | |
617ba13b | 36 | * from ext4_file_open: open gets called at every open, but release |
ac27a0ec DK |
37 | * gets called only when /all/ the files are closed. |
38 | */ | |
af5bc92d | 39 | static int ext4_release_file(struct inode *inode, struct file *filp) |
ac27a0ec | 40 | { |
19f5fb7a | 41 | if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE)) { |
7d8f9f7d | 42 | ext4_alloc_da_blocks(inode); |
19f5fb7a | 43 | ext4_clear_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE); |
7d8f9f7d | 44 | } |
ac27a0ec DK |
45 | /* if we are the last writer on the inode, drop the block reservation */ |
46 | if ((filp->f_mode & FMODE_WRITE) && | |
d6014301 AK |
47 | (atomic_read(&inode->i_writecount) == 1) && |
48 | !EXT4_I(inode)->i_reserved_data_blocks) | |
ac27a0ec | 49 | { |
0e855ac8 | 50 | down_write(&EXT4_I(inode)->i_data_sem); |
c2ea3fde | 51 | ext4_discard_preallocations(inode); |
0e855ac8 | 52 | up_write(&EXT4_I(inode)->i_data_sem); |
ac27a0ec DK |
53 | } |
54 | if (is_dx(inode) && filp->private_data) | |
617ba13b | 55 | ext4_htree_free_dir_info(filp->private_data); |
ac27a0ec DK |
56 | |
57 | return 0; | |
58 | } | |
59 | ||
c197855e | 60 | static void ext4_unwritten_wait(struct inode *inode) |
e9e3bcec ES |
61 | { |
62 | wait_queue_head_t *wq = ext4_ioend_wq(inode); | |
63 | ||
e27f41e1 | 64 | wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_unwritten) == 0)); |
e9e3bcec ES |
65 | } |
66 | ||
67 | /* | |
68 | * This tests whether the IO in question is block-aligned or not. | |
69 | * Ext4 utilizes unwritten extents when hole-filling during direct IO, and they | |
70 | * are converted to written only after the IO is complete. Until they are | |
71 | * mapped, these blocks appear as holes, so dio_zero_block() will assume that | |
72 | * it needs to zero out portions of the start and/or end block. If 2 AIO | |
73 | * threads are at work on the same unwritten block, they must be synchronized | |
74 | * or one thread will zero the other's data, causing corruption. | |
75 | */ | |
76 | static int | |
9b884164 | 77 | ext4_unaligned_aio(struct inode *inode, struct iov_iter *from, loff_t pos) |
e9e3bcec ES |
78 | { |
79 | struct super_block *sb = inode->i_sb; | |
80 | int blockmask = sb->s_blocksize - 1; | |
e9e3bcec | 81 | |
6e6358fc | 82 | if (pos >= i_size_read(inode)) |
e9e3bcec ES |
83 | return 0; |
84 | ||
9b884164 | 85 | if ((pos | iov_iter_alignment(from)) & blockmask) |
e9e3bcec ES |
86 | return 1; |
87 | ||
88 | return 0; | |
89 | } | |
90 | ||
ac27a0ec | 91 | static ssize_t |
9b884164 | 92 | ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from) |
ac27a0ec | 93 | { |
8ad2850f | 94 | struct inode *inode = file_inode(iocb->ki_filp); |
2ba48ce5 | 95 | int o_direct = iocb->ki_flags & IOCB_DIRECT; |
e142d052 | 96 | int unaligned_aio = 0; |
4bd809db | 97 | int overwrite = 0; |
8563000d | 98 | ssize_t ret; |
7608e610 | 99 | |
e142d052 JK |
100 | inode_lock(inode); |
101 | ret = generic_write_checks(iocb, from); | |
102 | if (ret <= 0) | |
103 | goto out; | |
104 | ||
f5ccfe1d | 105 | /* |
e142d052 JK |
106 | * Unaligned direct AIO must be serialized among each other as zeroing |
107 | * of partial blocks of two competing unaligned AIOs can result in data | |
108 | * corruption. | |
f5ccfe1d | 109 | */ |
e142d052 | 110 | if (o_direct && ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) && |
f5ccfe1d | 111 | !is_sync_kiocb(iocb) && |
e142d052 JK |
112 | ext4_unaligned_aio(inode, from, iocb->ki_pos)) { |
113 | unaligned_aio = 1; | |
f5ccfe1d TT |
114 | ext4_unwritten_wait(inode); |
115 | } | |
116 | ||
e2b46574 ES |
117 | /* |
118 | * If we have encountered a bitmap-format file, the size limit | |
119 | * is smaller than s_maxbytes, which is for extent-mapped files. | |
120 | */ | |
12e9b892 | 121 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) { |
e2b46574 | 122 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
ac27a0ec | 123 | |
3309dd04 | 124 | if (iocb->ki_pos >= sbi->s_bitmap_maxbytes) { |
f5ccfe1d | 125 | ret = -EFBIG; |
e768d7ff | 126 | goto out; |
f5ccfe1d | 127 | } |
3309dd04 | 128 | iov_iter_truncate(from, sbi->s_bitmap_maxbytes - iocb->ki_pos); |
e2b46574 ES |
129 | } |
130 | ||
a41537e6 | 131 | iocb->private = &overwrite; |
7ed07ba8 | 132 | if (o_direct) { |
3309dd04 AV |
133 | size_t length = iov_iter_count(from); |
134 | loff_t pos = iocb->ki_pos; | |
8ad2850f | 135 | |
8ad2850f | 136 | /* check whether we do a DIO overwrite or not */ |
e142d052 | 137 | if (ext4_should_dioread_nolock(inode) && !unaligned_aio && |
4b0524aa | 138 | pos + length <= i_size_read(inode)) { |
8ad2850f TT |
139 | struct ext4_map_blocks map; |
140 | unsigned int blkbits = inode->i_blkbits; | |
141 | int err, len; | |
142 | ||
143 | map.m_lblk = pos >> blkbits; | |
518eaa63 | 144 | map.m_len = EXT4_MAX_BLOCKS(length, pos, blkbits); |
8ad2850f TT |
145 | len = map.m_len; |
146 | ||
147 | err = ext4_map_blocks(NULL, inode, &map, 0); | |
148 | /* | |
149 | * 'err==len' means that all of blocks has | |
150 | * been preallocated no matter they are | |
151 | * initialized or not. For excluding | |
152 | * unwritten extents, we need to check | |
153 | * m_flags. There are two conditions that | |
154 | * indicate for initialized extents. 1) If we | |
155 | * hit extent cache, EXT4_MAP_MAPPED flag is | |
156 | * returned; 2) If we do a real lookup, | |
157 | * non-flags are returned. So we should check | |
158 | * these two conditions. | |
159 | */ | |
160 | if (err == len && (map.m_flags & EXT4_MAP_MAPPED)) | |
161 | overwrite = 1; | |
162 | } | |
f5ccfe1d | 163 | } |
7608e610 | 164 | |
9b884164 | 165 | ret = __generic_file_write_iter(iocb, from); |
5955102c | 166 | inode_unlock(inode); |
7608e610 | 167 | |
e2592217 CH |
168 | if (ret > 0) |
169 | ret = generic_write_sync(iocb, ret); | |
e9e3bcec | 170 | |
e768d7ff AV |
171 | return ret; |
172 | ||
173 | out: | |
5955102c | 174 | inode_unlock(inode); |
e9e3bcec | 175 | return ret; |
ac27a0ec DK |
176 | } |
177 | ||
923ae0ff RZ |
178 | #ifdef CONFIG_FS_DAX |
179 | static int ext4_dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf) | |
180 | { | |
01a33b4a MW |
181 | int result; |
182 | handle_t *handle = NULL; | |
ea3d7209 JK |
183 | struct inode *inode = file_inode(vma->vm_file); |
184 | struct super_block *sb = inode->i_sb; | |
01a33b4a MW |
185 | bool write = vmf->flags & FAULT_FLAG_WRITE; |
186 | ||
187 | if (write) { | |
188 | sb_start_pagefault(sb); | |
189 | file_update_time(vma->vm_file); | |
ea3d7209 | 190 | down_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a MW |
191 | handle = ext4_journal_start_sb(sb, EXT4_HT_WRITE_PAGE, |
192 | EXT4_DATA_TRANS_BLOCKS(sb)); | |
ea3d7209 JK |
193 | } else |
194 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
195 | |
196 | if (IS_ERR(handle)) | |
197 | result = VM_FAULT_SIGBUS; | |
198 | else | |
6b524995 | 199 | result = dax_fault(vma, vmf, ext4_dax_get_block); |
01a33b4a MW |
200 | |
201 | if (write) { | |
202 | if (!IS_ERR(handle)) | |
203 | ext4_journal_stop(handle); | |
ea3d7209 | 204 | up_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a | 205 | sb_end_pagefault(sb); |
ea3d7209 JK |
206 | } else |
207 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
208 | |
209 | return result; | |
923ae0ff RZ |
210 | } |
211 | ||
11bd1a9e MW |
212 | static int ext4_dax_pmd_fault(struct vm_area_struct *vma, unsigned long addr, |
213 | pmd_t *pmd, unsigned int flags) | |
214 | { | |
01a33b4a MW |
215 | int result; |
216 | handle_t *handle = NULL; | |
217 | struct inode *inode = file_inode(vma->vm_file); | |
218 | struct super_block *sb = inode->i_sb; | |
219 | bool write = flags & FAULT_FLAG_WRITE; | |
220 | ||
221 | if (write) { | |
222 | sb_start_pagefault(sb); | |
223 | file_update_time(vma->vm_file); | |
ea3d7209 | 224 | down_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a MW |
225 | handle = ext4_journal_start_sb(sb, EXT4_HT_WRITE_PAGE, |
226 | ext4_chunk_trans_blocks(inode, | |
227 | PMD_SIZE / PAGE_SIZE)); | |
ea3d7209 JK |
228 | } else |
229 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
230 | |
231 | if (IS_ERR(handle)) | |
232 | result = VM_FAULT_SIGBUS; | |
233 | else | |
6b524995 | 234 | result = dax_pmd_fault(vma, addr, pmd, flags, |
02fbd139 | 235 | ext4_dax_get_block); |
01a33b4a MW |
236 | |
237 | if (write) { | |
238 | if (!IS_ERR(handle)) | |
239 | ext4_journal_stop(handle); | |
ea3d7209 | 240 | up_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a | 241 | sb_end_pagefault(sb); |
ea3d7209 JK |
242 | } else |
243 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
244 | |
245 | return result; | |
11bd1a9e MW |
246 | } |
247 | ||
ea3d7209 | 248 | /* |
1e9d180b | 249 | * Handle write fault for VM_MIXEDMAP mappings. Similarly to ext4_dax_fault() |
ea3d7209 JK |
250 | * handler we check for races agaist truncate. Note that since we cycle through |
251 | * i_mmap_sem, we are sure that also any hole punching that began before we | |
252 | * were called is finished by now and so if it included part of the file we | |
253 | * are working on, our pte will get unmapped and the check for pte_same() in | |
254 | * wp_pfn_shared() fails. Thus fault gets retried and things work out as | |
255 | * desired. | |
256 | */ | |
257 | static int ext4_dax_pfn_mkwrite(struct vm_area_struct *vma, | |
258 | struct vm_fault *vmf) | |
259 | { | |
260 | struct inode *inode = file_inode(vma->vm_file); | |
261 | struct super_block *sb = inode->i_sb; | |
ea3d7209 | 262 | loff_t size; |
d5be7a03 | 263 | int ret; |
ea3d7209 JK |
264 | |
265 | sb_start_pagefault(sb); | |
266 | file_update_time(vma->vm_file); | |
267 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
268 | size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
269 | if (vmf->pgoff >= size) | |
270 | ret = VM_FAULT_SIGBUS; | |
d5be7a03 RZ |
271 | else |
272 | ret = dax_pfn_mkwrite(vma, vmf); | |
ea3d7209 JK |
273 | up_read(&EXT4_I(inode)->i_mmap_sem); |
274 | sb_end_pagefault(sb); | |
275 | ||
276 | return ret; | |
923ae0ff RZ |
277 | } |
278 | ||
279 | static const struct vm_operations_struct ext4_dax_vm_ops = { | |
280 | .fault = ext4_dax_fault, | |
11bd1a9e | 281 | .pmd_fault = ext4_dax_pmd_fault, |
1e9d180b | 282 | .page_mkwrite = ext4_dax_fault, |
ea3d7209 | 283 | .pfn_mkwrite = ext4_dax_pfn_mkwrite, |
923ae0ff RZ |
284 | }; |
285 | #else | |
286 | #define ext4_dax_vm_ops ext4_file_vm_ops | |
287 | #endif | |
288 | ||
f0f37e2f | 289 | static const struct vm_operations_struct ext4_file_vm_ops = { |
ea3d7209 | 290 | .fault = ext4_filemap_fault, |
f1820361 | 291 | .map_pages = filemap_map_pages, |
2e9ee850 AK |
292 | .page_mkwrite = ext4_page_mkwrite, |
293 | }; | |
294 | ||
295 | static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma) | |
296 | { | |
c9c7429c MH |
297 | struct inode *inode = file->f_mapping->host; |
298 | ||
299 | if (ext4_encrypted_inode(inode)) { | |
a7550b30 | 300 | int err = fscrypt_get_encryption_info(inode); |
c9c7429c MH |
301 | if (err) |
302 | return 0; | |
a7550b30 | 303 | if (!fscrypt_has_encryption_key(inode)) |
abdd438b | 304 | return -ENOKEY; |
c9c7429c | 305 | } |
2e9ee850 | 306 | file_accessed(file); |
923ae0ff RZ |
307 | if (IS_DAX(file_inode(file))) { |
308 | vma->vm_ops = &ext4_dax_vm_ops; | |
11bd1a9e | 309 | vma->vm_flags |= VM_MIXEDMAP | VM_HUGEPAGE; |
923ae0ff RZ |
310 | } else { |
311 | vma->vm_ops = &ext4_file_vm_ops; | |
312 | } | |
2e9ee850 AK |
313 | return 0; |
314 | } | |
315 | ||
bc0b0d6d TT |
316 | static int ext4_file_open(struct inode * inode, struct file * filp) |
317 | { | |
318 | struct super_block *sb = inode->i_sb; | |
319 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
320 | struct vfsmount *mnt = filp->f_path.mnt; | |
9dd78d8c | 321 | struct dentry *dir; |
bc0b0d6d TT |
322 | struct path path; |
323 | char buf[64], *cp; | |
c9c7429c | 324 | int ret; |
bc0b0d6d TT |
325 | |
326 | if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) && | |
327 | !(sb->s_flags & MS_RDONLY))) { | |
328 | sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED; | |
329 | /* | |
330 | * Sample where the filesystem has been mounted and | |
331 | * store it in the superblock for sysadmin convenience | |
332 | * when trying to sort through large numbers of block | |
333 | * devices or filesystem images. | |
334 | */ | |
335 | memset(buf, 0, sizeof(buf)); | |
3899167d AV |
336 | path.mnt = mnt; |
337 | path.dentry = mnt->mnt_root; | |
bc0b0d6d | 338 | cp = d_path(&path, buf, sizeof(buf)); |
bc0b0d6d | 339 | if (!IS_ERR(cp)) { |
044ce47f JK |
340 | handle_t *handle; |
341 | int err; | |
342 | ||
9924a92a | 343 | handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1); |
044ce47f JK |
344 | if (IS_ERR(handle)) |
345 | return PTR_ERR(handle); | |
5d601255 | 346 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
044ce47f JK |
347 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); |
348 | if (err) { | |
349 | ext4_journal_stop(handle); | |
350 | return err; | |
351 | } | |
cf803903 DW |
352 | strlcpy(sbi->s_es->s_last_mounted, cp, |
353 | sizeof(sbi->s_es->s_last_mounted)); | |
044ce47f JK |
354 | ext4_handle_dirty_super(handle, sb); |
355 | ext4_journal_stop(handle); | |
bc0b0d6d TT |
356 | } |
357 | } | |
abdd438b | 358 | if (ext4_encrypted_inode(inode)) { |
a7550b30 | 359 | ret = fscrypt_get_encryption_info(inode); |
abdd438b TT |
360 | if (ret) |
361 | return -EACCES; | |
a7550b30 | 362 | if (!fscrypt_has_encryption_key(inode)) |
abdd438b TT |
363 | return -ENOKEY; |
364 | } | |
9dd78d8c | 365 | |
c0a37d48 | 366 | dir = dget_parent(file_dentry(filp)); |
9dd78d8c | 367 | if (ext4_encrypted_inode(d_inode(dir)) && |
a7550b30 | 368 | !fscrypt_has_permitted_context(d_inode(dir), inode)) { |
ff978b09 | 369 | ext4_warning(inode->i_sb, |
8d2ae1cb | 370 | "Inconsistent encryption contexts: %lu/%lu", |
9dd78d8c | 371 | (unsigned long) d_inode(dir)->i_ino, |
ff978b09 | 372 | (unsigned long) inode->i_ino); |
9dd78d8c | 373 | dput(dir); |
ff978b09 TT |
374 | return -EPERM; |
375 | } | |
9dd78d8c | 376 | dput(dir); |
8aefcd55 TT |
377 | /* |
378 | * Set up the jbd2_inode if we are opening the inode for | |
379 | * writing and the journal is present | |
380 | */ | |
a361293f | 381 | if (filp->f_mode & FMODE_WRITE) { |
c9c7429c | 382 | ret = ext4_inode_attach_jinode(inode); |
a361293f JK |
383 | if (ret < 0) |
384 | return ret; | |
8aefcd55 | 385 | } |
abdd438b | 386 | return dquot_file_open(inode, filp); |
bc0b0d6d TT |
387 | } |
388 | ||
c8c0df24 ZL |
389 | /* |
390 | * Here we use ext4_map_blocks() to get a block mapping for a extent-based | |
391 | * file rather than ext4_ext_walk_space() because we can introduce | |
392 | * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same | |
393 | * function. When extent status tree has been fully implemented, it will | |
394 | * track all extent status for a file and we can directly use it to | |
395 | * retrieve the offset for SEEK_DATA/SEEK_HOLE. | |
396 | */ | |
397 | ||
398 | /* | |
399 | * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to | |
400 | * lookup page cache to check whether or not there has some data between | |
401 | * [startoff, endoff] because, if this range contains an unwritten extent, | |
402 | * we determine this extent as a data or a hole according to whether the | |
403 | * page cache has data or not. | |
404 | */ | |
ad7fefb1 TT |
405 | static int ext4_find_unwritten_pgoff(struct inode *inode, |
406 | int whence, | |
2d90c160 | 407 | ext4_lblk_t end_blk, |
ad7fefb1 | 408 | loff_t *offset) |
c8c0df24 ZL |
409 | { |
410 | struct pagevec pvec; | |
ad7fefb1 | 411 | unsigned int blkbits; |
c8c0df24 ZL |
412 | pgoff_t index; |
413 | pgoff_t end; | |
ad7fefb1 | 414 | loff_t endoff; |
c8c0df24 ZL |
415 | loff_t startoff; |
416 | loff_t lastoff; | |
417 | int found = 0; | |
418 | ||
ad7fefb1 | 419 | blkbits = inode->i_sb->s_blocksize_bits; |
c8c0df24 ZL |
420 | startoff = *offset; |
421 | lastoff = startoff; | |
2d90c160 | 422 | endoff = (loff_t)end_blk << blkbits; |
c8c0df24 | 423 | |
09cbfeaf KS |
424 | index = startoff >> PAGE_SHIFT; |
425 | end = endoff >> PAGE_SHIFT; | |
c8c0df24 ZL |
426 | |
427 | pagevec_init(&pvec, 0); | |
428 | do { | |
429 | int i, num; | |
430 | unsigned long nr_pages; | |
431 | ||
432 | num = min_t(pgoff_t, end - index, PAGEVEC_SIZE); | |
433 | nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index, | |
434 | (pgoff_t)num); | |
435 | if (nr_pages == 0) { | |
965c8e59 | 436 | if (whence == SEEK_DATA) |
c8c0df24 ZL |
437 | break; |
438 | ||
965c8e59 | 439 | BUG_ON(whence != SEEK_HOLE); |
c8c0df24 ZL |
440 | /* |
441 | * If this is the first time to go into the loop and | |
442 | * offset is not beyond the end offset, it will be a | |
443 | * hole at this offset | |
444 | */ | |
445 | if (lastoff == startoff || lastoff < endoff) | |
446 | found = 1; | |
447 | break; | |
448 | } | |
449 | ||
450 | /* | |
451 | * If this is the first time to go into the loop and | |
452 | * offset is smaller than the first page offset, it will be a | |
453 | * hole at this offset. | |
454 | */ | |
965c8e59 | 455 | if (lastoff == startoff && whence == SEEK_HOLE && |
c8c0df24 ZL |
456 | lastoff < page_offset(pvec.pages[0])) { |
457 | found = 1; | |
458 | break; | |
459 | } | |
460 | ||
461 | for (i = 0; i < nr_pages; i++) { | |
462 | struct page *page = pvec.pages[i]; | |
463 | struct buffer_head *bh, *head; | |
464 | ||
465 | /* | |
466 | * If the current offset is not beyond the end of given | |
467 | * range, it will be a hole. | |
468 | */ | |
965c8e59 | 469 | if (lastoff < endoff && whence == SEEK_HOLE && |
c8c0df24 ZL |
470 | page->index > end) { |
471 | found = 1; | |
472 | *offset = lastoff; | |
473 | goto out; | |
474 | } | |
475 | ||
476 | lock_page(page); | |
477 | ||
478 | if (unlikely(page->mapping != inode->i_mapping)) { | |
479 | unlock_page(page); | |
480 | continue; | |
481 | } | |
482 | ||
483 | if (!page_has_buffers(page)) { | |
484 | unlock_page(page); | |
485 | continue; | |
486 | } | |
487 | ||
488 | if (page_has_buffers(page)) { | |
489 | lastoff = page_offset(page); | |
490 | bh = head = page_buffers(page); | |
491 | do { | |
492 | if (buffer_uptodate(bh) || | |
493 | buffer_unwritten(bh)) { | |
965c8e59 | 494 | if (whence == SEEK_DATA) |
c8c0df24 ZL |
495 | found = 1; |
496 | } else { | |
965c8e59 | 497 | if (whence == SEEK_HOLE) |
c8c0df24 ZL |
498 | found = 1; |
499 | } | |
500 | if (found) { | |
501 | *offset = max_t(loff_t, | |
502 | startoff, lastoff); | |
503 | unlock_page(page); | |
504 | goto out; | |
505 | } | |
506 | lastoff += bh->b_size; | |
507 | bh = bh->b_this_page; | |
508 | } while (bh != head); | |
509 | } | |
510 | ||
511 | lastoff = page_offset(page) + PAGE_SIZE; | |
512 | unlock_page(page); | |
513 | } | |
514 | ||
515 | /* | |
516 | * The no. of pages is less than our desired, that would be a | |
517 | * hole in there. | |
518 | */ | |
965c8e59 | 519 | if (nr_pages < num && whence == SEEK_HOLE) { |
c8c0df24 ZL |
520 | found = 1; |
521 | *offset = lastoff; | |
522 | break; | |
523 | } | |
524 | ||
525 | index = pvec.pages[i - 1]->index + 1; | |
526 | pagevec_release(&pvec); | |
527 | } while (index <= end); | |
528 | ||
529 | out: | |
530 | pagevec_release(&pvec); | |
531 | return found; | |
532 | } | |
533 | ||
534 | /* | |
535 | * ext4_seek_data() retrieves the offset for SEEK_DATA. | |
536 | */ | |
537 | static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize) | |
538 | { | |
539 | struct inode *inode = file->f_mapping->host; | |
ad7fefb1 TT |
540 | struct extent_status es; |
541 | ext4_lblk_t start, last, end; | |
542 | loff_t dataoff, isize; | |
543 | int blkbits; | |
2d90c160 | 544 | int ret; |
c8c0df24 | 545 | |
5955102c | 546 | inode_lock(inode); |
ad7fefb1 TT |
547 | |
548 | isize = i_size_read(inode); | |
549 | if (offset >= isize) { | |
5955102c | 550 | inode_unlock(inode); |
c8c0df24 ZL |
551 | return -ENXIO; |
552 | } | |
ad7fefb1 TT |
553 | |
554 | blkbits = inode->i_sb->s_blocksize_bits; | |
555 | start = offset >> blkbits; | |
556 | last = start; | |
557 | end = isize >> blkbits; | |
558 | dataoff = offset; | |
559 | ||
560 | do { | |
2d90c160 JK |
561 | ret = ext4_get_next_extent(inode, last, end - last + 1, &es); |
562 | if (ret <= 0) { | |
563 | /* No extent found -> no data */ | |
564 | if (ret == 0) | |
565 | ret = -ENXIO; | |
566 | inode_unlock(inode); | |
567 | return ret; | |
ad7fefb1 | 568 | } |
c8c0df24 | 569 | |
2d90c160 JK |
570 | last = es.es_lblk; |
571 | if (last != start) | |
572 | dataoff = (loff_t)last << blkbits; | |
573 | if (!ext4_es_is_unwritten(&es)) | |
c8c0df24 | 574 | break; |
c8c0df24 | 575 | |
ad7fefb1 TT |
576 | /* |
577 | * If there is a unwritten extent at this offset, | |
578 | * it will be as a data or a hole according to page | |
579 | * cache that has data or not. | |
580 | */ | |
2d90c160 JK |
581 | if (ext4_find_unwritten_pgoff(inode, SEEK_DATA, |
582 | es.es_lblk + es.es_len, &dataoff)) | |
583 | break; | |
584 | last += es.es_len; | |
ad7fefb1 | 585 | dataoff = (loff_t)last << blkbits; |
2d90c160 | 586 | cond_resched(); |
ad7fefb1 | 587 | } while (last <= end); |
c8c0df24 | 588 | |
5955102c | 589 | inode_unlock(inode); |
c8c0df24 | 590 | |
ad7fefb1 TT |
591 | if (dataoff > isize) |
592 | return -ENXIO; | |
593 | ||
594 | return vfs_setpos(file, dataoff, maxsize); | |
c8c0df24 ZL |
595 | } |
596 | ||
597 | /* | |
ad7fefb1 | 598 | * ext4_seek_hole() retrieves the offset for SEEK_HOLE. |
c8c0df24 ZL |
599 | */ |
600 | static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize) | |
601 | { | |
602 | struct inode *inode = file->f_mapping->host; | |
ad7fefb1 TT |
603 | struct extent_status es; |
604 | ext4_lblk_t start, last, end; | |
605 | loff_t holeoff, isize; | |
606 | int blkbits; | |
2d90c160 | 607 | int ret; |
c8c0df24 | 608 | |
5955102c | 609 | inode_lock(inode); |
ad7fefb1 TT |
610 | |
611 | isize = i_size_read(inode); | |
612 | if (offset >= isize) { | |
5955102c | 613 | inode_unlock(inode); |
c8c0df24 ZL |
614 | return -ENXIO; |
615 | } | |
616 | ||
ad7fefb1 TT |
617 | blkbits = inode->i_sb->s_blocksize_bits; |
618 | start = offset >> blkbits; | |
619 | last = start; | |
620 | end = isize >> blkbits; | |
621 | holeoff = offset; | |
c8c0df24 | 622 | |
ad7fefb1 | 623 | do { |
2d90c160 JK |
624 | ret = ext4_get_next_extent(inode, last, end - last + 1, &es); |
625 | if (ret < 0) { | |
626 | inode_unlock(inode); | |
627 | return ret; | |
ad7fefb1 | 628 | } |
2d90c160 JK |
629 | /* Found a hole? */ |
630 | if (ret == 0 || es.es_lblk > last) { | |
631 | if (last != start) | |
632 | holeoff = (loff_t)last << blkbits; | |
633 | break; | |
ad7fefb1 | 634 | } |
ad7fefb1 TT |
635 | /* |
636 | * If there is a unwritten extent at this offset, | |
637 | * it will be as a data or a hole according to page | |
638 | * cache that has data or not. | |
639 | */ | |
2d90c160 JK |
640 | if (ext4_es_is_unwritten(&es) && |
641 | ext4_find_unwritten_pgoff(inode, SEEK_HOLE, | |
642 | last + es.es_len, &holeoff)) | |
643 | break; | |
ad7fefb1 | 644 | |
2d90c160 JK |
645 | last += es.es_len; |
646 | holeoff = (loff_t)last << blkbits; | |
647 | cond_resched(); | |
ad7fefb1 TT |
648 | } while (last <= end); |
649 | ||
5955102c | 650 | inode_unlock(inode); |
c8c0df24 | 651 | |
ad7fefb1 TT |
652 | if (holeoff > isize) |
653 | holeoff = isize; | |
654 | ||
655 | return vfs_setpos(file, holeoff, maxsize); | |
c8c0df24 ZL |
656 | } |
657 | ||
e0d10bfa | 658 | /* |
ec7268ce ES |
659 | * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values |
660 | * by calling generic_file_llseek_size() with the appropriate maxbytes | |
661 | * value for each. | |
e0d10bfa | 662 | */ |
965c8e59 | 663 | loff_t ext4_llseek(struct file *file, loff_t offset, int whence) |
e0d10bfa TO |
664 | { |
665 | struct inode *inode = file->f_mapping->host; | |
666 | loff_t maxbytes; | |
667 | ||
668 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) | |
669 | maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes; | |
670 | else | |
671 | maxbytes = inode->i_sb->s_maxbytes; | |
e0d10bfa | 672 | |
965c8e59 | 673 | switch (whence) { |
c8c0df24 ZL |
674 | case SEEK_SET: |
675 | case SEEK_CUR: | |
676 | case SEEK_END: | |
965c8e59 | 677 | return generic_file_llseek_size(file, offset, whence, |
c8c0df24 ZL |
678 | maxbytes, i_size_read(inode)); |
679 | case SEEK_DATA: | |
680 | return ext4_seek_data(file, offset, maxbytes); | |
681 | case SEEK_HOLE: | |
682 | return ext4_seek_hole(file, offset, maxbytes); | |
683 | } | |
684 | ||
685 | return -EINVAL; | |
e0d10bfa TO |
686 | } |
687 | ||
617ba13b | 688 | const struct file_operations ext4_file_operations = { |
e0d10bfa | 689 | .llseek = ext4_llseek, |
aad4f8bb | 690 | .read_iter = generic_file_read_iter, |
9b884164 | 691 | .write_iter = ext4_file_write_iter, |
5cdd7b2d | 692 | .unlocked_ioctl = ext4_ioctl, |
ac27a0ec | 693 | #ifdef CONFIG_COMPAT |
617ba13b | 694 | .compat_ioctl = ext4_compat_ioctl, |
ac27a0ec | 695 | #endif |
2e9ee850 | 696 | .mmap = ext4_file_mmap, |
bc0b0d6d | 697 | .open = ext4_file_open, |
617ba13b MC |
698 | .release = ext4_release_file, |
699 | .fsync = ext4_sync_file, | |
dbe6ec81 | 700 | .get_unmapped_area = thp_get_unmapped_area, |
ac27a0ec | 701 | .splice_read = generic_file_splice_read, |
8d020765 | 702 | .splice_write = iter_file_splice_write, |
2fe17c10 | 703 | .fallocate = ext4_fallocate, |
ac27a0ec DK |
704 | }; |
705 | ||
754661f1 | 706 | const struct inode_operations ext4_file_inode_operations = { |
617ba13b | 707 | .setattr = ext4_setattr, |
3e3398a0 | 708 | .getattr = ext4_getattr, |
617ba13b | 709 | .listxattr = ext4_listxattr, |
4e34e719 | 710 | .get_acl = ext4_get_acl, |
64e178a7 | 711 | .set_acl = ext4_set_acl, |
6873fa0d | 712 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
713 | }; |
714 |