]>
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 | ||
364443cb JK |
34 | #ifdef CONFIG_FS_DAX |
35 | static ssize_t ext4_dax_read_iter(struct kiocb *iocb, struct iov_iter *to) | |
36 | { | |
37 | struct inode *inode = file_inode(iocb->ki_filp); | |
38 | ssize_t ret; | |
39 | ||
728fbc0e GR |
40 | if (!inode_trylock_shared(inode)) { |
41 | if (iocb->ki_flags & IOCB_NOWAIT) | |
42 | return -EAGAIN; | |
43 | inode_lock_shared(inode); | |
44 | } | |
364443cb JK |
45 | /* |
46 | * Recheck under inode lock - at this point we are sure it cannot | |
47 | * change anymore | |
48 | */ | |
49 | if (!IS_DAX(inode)) { | |
50 | inode_unlock_shared(inode); | |
51 | /* Fallback to buffered IO in case we cannot support DAX */ | |
52 | return generic_file_read_iter(iocb, to); | |
53 | } | |
54 | ret = dax_iomap_rw(iocb, to, &ext4_iomap_ops); | |
55 | inode_unlock_shared(inode); | |
56 | ||
57 | file_accessed(iocb->ki_filp); | |
58 | return ret; | |
59 | } | |
60 | #endif | |
61 | ||
62 | static ssize_t ext4_file_read_iter(struct kiocb *iocb, struct iov_iter *to) | |
63 | { | |
0db1ff22 TT |
64 | if (unlikely(ext4_forced_shutdown(EXT4_SB(file_inode(iocb->ki_filp)->i_sb)))) |
65 | return -EIO; | |
66 | ||
364443cb JK |
67 | if (!iov_iter_count(to)) |
68 | return 0; /* skip atime */ | |
69 | ||
70 | #ifdef CONFIG_FS_DAX | |
71 | if (IS_DAX(file_inode(iocb->ki_filp))) | |
72 | return ext4_dax_read_iter(iocb, to); | |
73 | #endif | |
74 | return generic_file_read_iter(iocb, to); | |
75 | } | |
76 | ||
ac27a0ec DK |
77 | /* |
78 | * Called when an inode is released. Note that this is different | |
617ba13b | 79 | * from ext4_file_open: open gets called at every open, but release |
ac27a0ec DK |
80 | * gets called only when /all/ the files are closed. |
81 | */ | |
af5bc92d | 82 | static int ext4_release_file(struct inode *inode, struct file *filp) |
ac27a0ec | 83 | { |
19f5fb7a | 84 | if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE)) { |
7d8f9f7d | 85 | ext4_alloc_da_blocks(inode); |
19f5fb7a | 86 | ext4_clear_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE); |
7d8f9f7d | 87 | } |
ac27a0ec DK |
88 | /* if we are the last writer on the inode, drop the block reservation */ |
89 | if ((filp->f_mode & FMODE_WRITE) && | |
d6014301 AK |
90 | (atomic_read(&inode->i_writecount) == 1) && |
91 | !EXT4_I(inode)->i_reserved_data_blocks) | |
ac27a0ec | 92 | { |
0e855ac8 | 93 | down_write(&EXT4_I(inode)->i_data_sem); |
c2ea3fde | 94 | ext4_discard_preallocations(inode); |
0e855ac8 | 95 | up_write(&EXT4_I(inode)->i_data_sem); |
ac27a0ec DK |
96 | } |
97 | if (is_dx(inode) && filp->private_data) | |
617ba13b | 98 | ext4_htree_free_dir_info(filp->private_data); |
ac27a0ec DK |
99 | |
100 | return 0; | |
101 | } | |
102 | ||
c197855e | 103 | static void ext4_unwritten_wait(struct inode *inode) |
e9e3bcec ES |
104 | { |
105 | wait_queue_head_t *wq = ext4_ioend_wq(inode); | |
106 | ||
e27f41e1 | 107 | wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_unwritten) == 0)); |
e9e3bcec ES |
108 | } |
109 | ||
110 | /* | |
111 | * This tests whether the IO in question is block-aligned or not. | |
112 | * Ext4 utilizes unwritten extents when hole-filling during direct IO, and they | |
113 | * are converted to written only after the IO is complete. Until they are | |
114 | * mapped, these blocks appear as holes, so dio_zero_block() will assume that | |
115 | * it needs to zero out portions of the start and/or end block. If 2 AIO | |
116 | * threads are at work on the same unwritten block, they must be synchronized | |
117 | * or one thread will zero the other's data, causing corruption. | |
118 | */ | |
119 | static int | |
9b884164 | 120 | ext4_unaligned_aio(struct inode *inode, struct iov_iter *from, loff_t pos) |
e9e3bcec ES |
121 | { |
122 | struct super_block *sb = inode->i_sb; | |
123 | int blockmask = sb->s_blocksize - 1; | |
e9e3bcec | 124 | |
6e6358fc | 125 | if (pos >= i_size_read(inode)) |
e9e3bcec ES |
126 | return 0; |
127 | ||
9b884164 | 128 | if ((pos | iov_iter_alignment(from)) & blockmask) |
e9e3bcec ES |
129 | return 1; |
130 | ||
131 | return 0; | |
132 | } | |
133 | ||
213bcd9c JK |
134 | /* Is IO overwriting allocated and initialized blocks? */ |
135 | static bool ext4_overwrite_io(struct inode *inode, loff_t pos, loff_t len) | |
136 | { | |
137 | struct ext4_map_blocks map; | |
138 | unsigned int blkbits = inode->i_blkbits; | |
139 | int err, blklen; | |
140 | ||
141 | if (pos + len > i_size_read(inode)) | |
142 | return false; | |
143 | ||
144 | map.m_lblk = pos >> blkbits; | |
145 | map.m_len = EXT4_MAX_BLOCKS(len, pos, blkbits); | |
146 | blklen = map.m_len; | |
147 | ||
148 | err = ext4_map_blocks(NULL, inode, &map, 0); | |
149 | /* | |
150 | * 'err==len' means that all of the blocks have been preallocated, | |
151 | * regardless of whether they have been initialized or not. To exclude | |
152 | * unwritten extents, we need to check m_flags. | |
153 | */ | |
154 | return err == blklen && (map.m_flags & EXT4_MAP_MAPPED); | |
155 | } | |
156 | ||
157 | static ssize_t ext4_write_checks(struct kiocb *iocb, struct iov_iter *from) | |
158 | { | |
159 | struct inode *inode = file_inode(iocb->ki_filp); | |
160 | ssize_t ret; | |
161 | ||
162 | ret = generic_write_checks(iocb, from); | |
163 | if (ret <= 0) | |
164 | return ret; | |
165 | /* | |
166 | * If we have encountered a bitmap-format file, the size limit | |
167 | * is smaller than s_maxbytes, which is for extent-mapped files. | |
168 | */ | |
169 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) { | |
170 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
171 | ||
172 | if (iocb->ki_pos >= sbi->s_bitmap_maxbytes) | |
173 | return -EFBIG; | |
174 | iov_iter_truncate(from, sbi->s_bitmap_maxbytes - iocb->ki_pos); | |
175 | } | |
176 | return iov_iter_count(from); | |
177 | } | |
178 | ||
776722e8 JK |
179 | #ifdef CONFIG_FS_DAX |
180 | static ssize_t | |
181 | ext4_dax_write_iter(struct kiocb *iocb, struct iov_iter *from) | |
182 | { | |
183 | struct inode *inode = file_inode(iocb->ki_filp); | |
184 | ssize_t ret; | |
776722e8 | 185 | |
728fbc0e GR |
186 | if (!inode_trylock(inode)) { |
187 | if (iocb->ki_flags & IOCB_NOWAIT) | |
188 | return -EAGAIN; | |
189 | inode_lock(inode); | |
190 | } | |
776722e8 JK |
191 | ret = ext4_write_checks(iocb, from); |
192 | if (ret <= 0) | |
193 | goto out; | |
194 | ret = file_remove_privs(iocb->ki_filp); | |
195 | if (ret) | |
196 | goto out; | |
197 | ret = file_update_time(iocb->ki_filp); | |
198 | if (ret) | |
199 | goto out; | |
200 | ||
776722e8 JK |
201 | ret = dax_iomap_rw(iocb, from, &ext4_iomap_ops); |
202 | out: | |
ff5462e3 | 203 | inode_unlock(inode); |
776722e8 JK |
204 | if (ret > 0) |
205 | ret = generic_write_sync(iocb, ret); | |
206 | return ret; | |
207 | } | |
208 | #endif | |
209 | ||
ac27a0ec | 210 | static ssize_t |
9b884164 | 211 | ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from) |
ac27a0ec | 212 | { |
8ad2850f | 213 | struct inode *inode = file_inode(iocb->ki_filp); |
2ba48ce5 | 214 | int o_direct = iocb->ki_flags & IOCB_DIRECT; |
e142d052 | 215 | int unaligned_aio = 0; |
4bd809db | 216 | int overwrite = 0; |
8563000d | 217 | ssize_t ret; |
7608e610 | 218 | |
0db1ff22 TT |
219 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
220 | return -EIO; | |
221 | ||
776722e8 JK |
222 | #ifdef CONFIG_FS_DAX |
223 | if (IS_DAX(inode)) | |
224 | return ext4_dax_write_iter(iocb, from); | |
225 | #endif | |
226 | ||
728fbc0e GR |
227 | if (!inode_trylock(inode)) { |
228 | if (iocb->ki_flags & IOCB_NOWAIT) | |
229 | return -EAGAIN; | |
230 | inode_lock(inode); | |
231 | } | |
232 | ||
213bcd9c | 233 | ret = ext4_write_checks(iocb, from); |
e142d052 JK |
234 | if (ret <= 0) |
235 | goto out; | |
236 | ||
f5ccfe1d | 237 | /* |
e142d052 JK |
238 | * Unaligned direct AIO must be serialized among each other as zeroing |
239 | * of partial blocks of two competing unaligned AIOs can result in data | |
240 | * corruption. | |
f5ccfe1d | 241 | */ |
e142d052 | 242 | if (o_direct && ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) && |
f5ccfe1d | 243 | !is_sync_kiocb(iocb) && |
e142d052 JK |
244 | ext4_unaligned_aio(inode, from, iocb->ki_pos)) { |
245 | unaligned_aio = 1; | |
f5ccfe1d TT |
246 | ext4_unwritten_wait(inode); |
247 | } | |
248 | ||
a41537e6 | 249 | iocb->private = &overwrite; |
213bcd9c | 250 | /* Check whether we do a DIO overwrite or not */ |
728fbc0e GR |
251 | if (o_direct && !unaligned_aio) { |
252 | if (ext4_overwrite_io(inode, iocb->ki_pos, iov_iter_count(from))) { | |
253 | if (ext4_should_dioread_nolock(inode)) | |
254 | overwrite = 1; | |
255 | } else if (iocb->ki_flags & IOCB_NOWAIT) { | |
256 | ret = -EAGAIN; | |
257 | goto out; | |
258 | } | |
259 | } | |
7608e610 | 260 | |
9b884164 | 261 | ret = __generic_file_write_iter(iocb, from); |
5955102c | 262 | inode_unlock(inode); |
7608e610 | 263 | |
e2592217 CH |
264 | if (ret > 0) |
265 | ret = generic_write_sync(iocb, ret); | |
e9e3bcec | 266 | |
e768d7ff AV |
267 | return ret; |
268 | ||
269 | out: | |
5955102c | 270 | inode_unlock(inode); |
e9e3bcec | 271 | return ret; |
ac27a0ec DK |
272 | } |
273 | ||
923ae0ff | 274 | #ifdef CONFIG_FS_DAX |
c791ace1 DJ |
275 | static int ext4_dax_huge_fault(struct vm_fault *vmf, |
276 | enum page_entry_size pe_size) | |
923ae0ff | 277 | { |
01a33b4a | 278 | int result; |
fb26a1cb | 279 | handle_t *handle = NULL; |
11bac800 | 280 | struct inode *inode = file_inode(vmf->vma->vm_file); |
ea3d7209 | 281 | struct super_block *sb = inode->i_sb; |
01a33b4a MW |
282 | bool write = vmf->flags & FAULT_FLAG_WRITE; |
283 | ||
284 | if (write) { | |
285 | sb_start_pagefault(sb); | |
11bac800 | 286 | file_update_time(vmf->vma->vm_file); |
fb26a1cb JK |
287 | down_read(&EXT4_I(inode)->i_mmap_sem); |
288 | handle = ext4_journal_start_sb(sb, EXT4_HT_WRITE_PAGE, | |
289 | EXT4_DATA_TRANS_BLOCKS(sb)); | |
290 | } else { | |
291 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
1db17542 | 292 | } |
fb26a1cb JK |
293 | if (!IS_ERR(handle)) |
294 | result = dax_iomap_fault(vmf, pe_size, &ext4_iomap_ops); | |
295 | else | |
296 | result = VM_FAULT_SIGBUS; | |
297 | if (write) { | |
298 | if (!IS_ERR(handle)) | |
299 | ext4_journal_stop(handle); | |
300 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a | 301 | sb_end_pagefault(sb); |
fb26a1cb JK |
302 | } else { |
303 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
304 | } | |
01a33b4a MW |
305 | |
306 | return result; | |
923ae0ff RZ |
307 | } |
308 | ||
c791ace1 DJ |
309 | static int ext4_dax_fault(struct vm_fault *vmf) |
310 | { | |
311 | return ext4_dax_huge_fault(vmf, PE_SIZE_PTE); | |
312 | } | |
313 | ||
923ae0ff RZ |
314 | static const struct vm_operations_struct ext4_dax_vm_ops = { |
315 | .fault = ext4_dax_fault, | |
c791ace1 | 316 | .huge_fault = ext4_dax_huge_fault, |
1e9d180b | 317 | .page_mkwrite = ext4_dax_fault, |
54585e35 | 318 | .pfn_mkwrite = ext4_dax_fault, |
923ae0ff RZ |
319 | }; |
320 | #else | |
321 | #define ext4_dax_vm_ops ext4_file_vm_ops | |
322 | #endif | |
323 | ||
f0f37e2f | 324 | static const struct vm_operations_struct ext4_file_vm_ops = { |
ea3d7209 | 325 | .fault = ext4_filemap_fault, |
f1820361 | 326 | .map_pages = filemap_map_pages, |
2e9ee850 AK |
327 | .page_mkwrite = ext4_page_mkwrite, |
328 | }; | |
329 | ||
330 | static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma) | |
331 | { | |
c9c7429c MH |
332 | struct inode *inode = file->f_mapping->host; |
333 | ||
0db1ff22 TT |
334 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
335 | return -EIO; | |
336 | ||
2e9ee850 | 337 | file_accessed(file); |
923ae0ff RZ |
338 | if (IS_DAX(file_inode(file))) { |
339 | vma->vm_ops = &ext4_dax_vm_ops; | |
11bd1a9e | 340 | vma->vm_flags |= VM_MIXEDMAP | VM_HUGEPAGE; |
923ae0ff RZ |
341 | } else { |
342 | vma->vm_ops = &ext4_file_vm_ops; | |
343 | } | |
2e9ee850 AK |
344 | return 0; |
345 | } | |
346 | ||
bc0b0d6d TT |
347 | static int ext4_file_open(struct inode * inode, struct file * filp) |
348 | { | |
349 | struct super_block *sb = inode->i_sb; | |
350 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
351 | struct vfsmount *mnt = filp->f_path.mnt; | |
9dd78d8c | 352 | struct dentry *dir; |
bc0b0d6d TT |
353 | struct path path; |
354 | char buf[64], *cp; | |
c9c7429c | 355 | int ret; |
bc0b0d6d | 356 | |
0db1ff22 TT |
357 | if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) |
358 | return -EIO; | |
359 | ||
bc0b0d6d TT |
360 | if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) && |
361 | !(sb->s_flags & MS_RDONLY))) { | |
362 | sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED; | |
363 | /* | |
364 | * Sample where the filesystem has been mounted and | |
365 | * store it in the superblock for sysadmin convenience | |
366 | * when trying to sort through large numbers of block | |
367 | * devices or filesystem images. | |
368 | */ | |
369 | memset(buf, 0, sizeof(buf)); | |
3899167d AV |
370 | path.mnt = mnt; |
371 | path.dentry = mnt->mnt_root; | |
bc0b0d6d | 372 | cp = d_path(&path, buf, sizeof(buf)); |
bc0b0d6d | 373 | if (!IS_ERR(cp)) { |
044ce47f JK |
374 | handle_t *handle; |
375 | int err; | |
376 | ||
9924a92a | 377 | handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1); |
044ce47f JK |
378 | if (IS_ERR(handle)) |
379 | return PTR_ERR(handle); | |
5d601255 | 380 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
044ce47f JK |
381 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); |
382 | if (err) { | |
383 | ext4_journal_stop(handle); | |
384 | return err; | |
385 | } | |
cf803903 DW |
386 | strlcpy(sbi->s_es->s_last_mounted, cp, |
387 | sizeof(sbi->s_es->s_last_mounted)); | |
044ce47f JK |
388 | ext4_handle_dirty_super(handle, sb); |
389 | ext4_journal_stop(handle); | |
bc0b0d6d TT |
390 | } |
391 | } | |
abdd438b | 392 | if (ext4_encrypted_inode(inode)) { |
a7550b30 | 393 | ret = fscrypt_get_encryption_info(inode); |
abdd438b TT |
394 | if (ret) |
395 | return -EACCES; | |
a7550b30 | 396 | if (!fscrypt_has_encryption_key(inode)) |
abdd438b TT |
397 | return -ENOKEY; |
398 | } | |
9dd78d8c | 399 | |
c0a37d48 | 400 | dir = dget_parent(file_dentry(filp)); |
9dd78d8c | 401 | if (ext4_encrypted_inode(d_inode(dir)) && |
a7550b30 | 402 | !fscrypt_has_permitted_context(d_inode(dir), inode)) { |
ff978b09 | 403 | ext4_warning(inode->i_sb, |
8d2ae1cb | 404 | "Inconsistent encryption contexts: %lu/%lu", |
9dd78d8c | 405 | (unsigned long) d_inode(dir)->i_ino, |
ff978b09 | 406 | (unsigned long) inode->i_ino); |
9dd78d8c | 407 | dput(dir); |
ff978b09 TT |
408 | return -EPERM; |
409 | } | |
9dd78d8c | 410 | dput(dir); |
8aefcd55 TT |
411 | /* |
412 | * Set up the jbd2_inode if we are opening the inode for | |
413 | * writing and the journal is present | |
414 | */ | |
a361293f | 415 | if (filp->f_mode & FMODE_WRITE) { |
c9c7429c | 416 | ret = ext4_inode_attach_jinode(inode); |
a361293f JK |
417 | if (ret < 0) |
418 | return ret; | |
8aefcd55 | 419 | } |
728fbc0e GR |
420 | |
421 | /* Set the flags to support nowait AIO */ | |
422 | filp->f_mode |= FMODE_AIO_NOWAIT; | |
423 | ||
abdd438b | 424 | return dquot_file_open(inode, filp); |
bc0b0d6d TT |
425 | } |
426 | ||
c8c0df24 ZL |
427 | /* |
428 | * Here we use ext4_map_blocks() to get a block mapping for a extent-based | |
429 | * file rather than ext4_ext_walk_space() because we can introduce | |
430 | * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same | |
431 | * function. When extent status tree has been fully implemented, it will | |
432 | * track all extent status for a file and we can directly use it to | |
433 | * retrieve the offset for SEEK_DATA/SEEK_HOLE. | |
434 | */ | |
435 | ||
436 | /* | |
437 | * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to | |
438 | * lookup page cache to check whether or not there has some data between | |
439 | * [startoff, endoff] because, if this range contains an unwritten extent, | |
440 | * we determine this extent as a data or a hole according to whether the | |
441 | * page cache has data or not. | |
442 | */ | |
ad7fefb1 TT |
443 | static int ext4_find_unwritten_pgoff(struct inode *inode, |
444 | int whence, | |
2d90c160 | 445 | ext4_lblk_t end_blk, |
ad7fefb1 | 446 | loff_t *offset) |
c8c0df24 ZL |
447 | { |
448 | struct pagevec pvec; | |
ad7fefb1 | 449 | unsigned int blkbits; |
c8c0df24 ZL |
450 | pgoff_t index; |
451 | pgoff_t end; | |
ad7fefb1 | 452 | loff_t endoff; |
c8c0df24 ZL |
453 | loff_t startoff; |
454 | loff_t lastoff; | |
455 | int found = 0; | |
456 | ||
ad7fefb1 | 457 | blkbits = inode->i_sb->s_blocksize_bits; |
c8c0df24 ZL |
458 | startoff = *offset; |
459 | lastoff = startoff; | |
2d90c160 | 460 | endoff = (loff_t)end_blk << blkbits; |
c8c0df24 | 461 | |
09cbfeaf | 462 | index = startoff >> PAGE_SHIFT; |
3f1d5bad | 463 | end = (endoff - 1) >> PAGE_SHIFT; |
c8c0df24 ZL |
464 | |
465 | pagevec_init(&pvec, 0); | |
466 | do { | |
467 | int i, num; | |
468 | unsigned long nr_pages; | |
469 | ||
624327f8 | 470 | num = min_t(pgoff_t, end - index, PAGEVEC_SIZE - 1) + 1; |
c8c0df24 ZL |
471 | nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index, |
472 | (pgoff_t)num); | |
7d95eddf | 473 | if (nr_pages == 0) |
c8c0df24 | 474 | break; |
c8c0df24 ZL |
475 | |
476 | for (i = 0; i < nr_pages; i++) { | |
477 | struct page *page = pvec.pages[i]; | |
478 | struct buffer_head *bh, *head; | |
479 | ||
480 | /* | |
7d95eddf JK |
481 | * If current offset is smaller than the page offset, |
482 | * there is a hole at this offset. | |
c8c0df24 | 483 | */ |
7d95eddf JK |
484 | if (whence == SEEK_HOLE && lastoff < endoff && |
485 | lastoff < page_offset(pvec.pages[i])) { | |
c8c0df24 ZL |
486 | found = 1; |
487 | *offset = lastoff; | |
488 | goto out; | |
489 | } | |
490 | ||
7d95eddf JK |
491 | if (page->index > end) |
492 | goto out; | |
493 | ||
c8c0df24 ZL |
494 | lock_page(page); |
495 | ||
496 | if (unlikely(page->mapping != inode->i_mapping)) { | |
497 | unlock_page(page); | |
498 | continue; | |
499 | } | |
500 | ||
501 | if (!page_has_buffers(page)) { | |
502 | unlock_page(page); | |
503 | continue; | |
504 | } | |
505 | ||
506 | if (page_has_buffers(page)) { | |
507 | lastoff = page_offset(page); | |
508 | bh = head = page_buffers(page); | |
509 | do { | |
fcf5ea10 JK |
510 | if (lastoff + bh->b_size <= startoff) |
511 | goto next; | |
c8c0df24 ZL |
512 | if (buffer_uptodate(bh) || |
513 | buffer_unwritten(bh)) { | |
965c8e59 | 514 | if (whence == SEEK_DATA) |
c8c0df24 ZL |
515 | found = 1; |
516 | } else { | |
965c8e59 | 517 | if (whence == SEEK_HOLE) |
c8c0df24 ZL |
518 | found = 1; |
519 | } | |
520 | if (found) { | |
521 | *offset = max_t(loff_t, | |
522 | startoff, lastoff); | |
523 | unlock_page(page); | |
524 | goto out; | |
525 | } | |
fcf5ea10 | 526 | next: |
c8c0df24 ZL |
527 | lastoff += bh->b_size; |
528 | bh = bh->b_this_page; | |
529 | } while (bh != head); | |
530 | } | |
531 | ||
532 | lastoff = page_offset(page) + PAGE_SIZE; | |
533 | unlock_page(page); | |
534 | } | |
535 | ||
7d95eddf JK |
536 | /* The no. of pages is less than our desired, we are done. */ |
537 | if (nr_pages < num) | |
c8c0df24 | 538 | break; |
c8c0df24 ZL |
539 | |
540 | index = pvec.pages[i - 1]->index + 1; | |
541 | pagevec_release(&pvec); | |
542 | } while (index <= end); | |
543 | ||
7d95eddf JK |
544 | if (whence == SEEK_HOLE && lastoff < endoff) { |
545 | found = 1; | |
546 | *offset = lastoff; | |
547 | } | |
c8c0df24 ZL |
548 | out: |
549 | pagevec_release(&pvec); | |
550 | return found; | |
551 | } | |
552 | ||
553 | /* | |
554 | * ext4_seek_data() retrieves the offset for SEEK_DATA. | |
555 | */ | |
556 | static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize) | |
557 | { | |
558 | struct inode *inode = file->f_mapping->host; | |
ad7fefb1 TT |
559 | struct extent_status es; |
560 | ext4_lblk_t start, last, end; | |
561 | loff_t dataoff, isize; | |
562 | int blkbits; | |
2d90c160 | 563 | int ret; |
c8c0df24 | 564 | |
5955102c | 565 | inode_lock(inode); |
ad7fefb1 TT |
566 | |
567 | isize = i_size_read(inode); | |
f289f8c0 | 568 | if (offset < 0 || offset >= isize) { |
5955102c | 569 | inode_unlock(inode); |
c8c0df24 ZL |
570 | return -ENXIO; |
571 | } | |
ad7fefb1 TT |
572 | |
573 | blkbits = inode->i_sb->s_blocksize_bits; | |
574 | start = offset >> blkbits; | |
575 | last = start; | |
576 | end = isize >> blkbits; | |
577 | dataoff = offset; | |
578 | ||
579 | do { | |
2d90c160 JK |
580 | ret = ext4_get_next_extent(inode, last, end - last + 1, &es); |
581 | if (ret <= 0) { | |
582 | /* No extent found -> no data */ | |
583 | if (ret == 0) | |
584 | ret = -ENXIO; | |
585 | inode_unlock(inode); | |
586 | return ret; | |
ad7fefb1 | 587 | } |
c8c0df24 | 588 | |
2d90c160 JK |
589 | last = es.es_lblk; |
590 | if (last != start) | |
591 | dataoff = (loff_t)last << blkbits; | |
592 | if (!ext4_es_is_unwritten(&es)) | |
c8c0df24 | 593 | break; |
c8c0df24 | 594 | |
ad7fefb1 TT |
595 | /* |
596 | * If there is a unwritten extent at this offset, | |
597 | * it will be as a data or a hole according to page | |
598 | * cache that has data or not. | |
599 | */ | |
2d90c160 JK |
600 | if (ext4_find_unwritten_pgoff(inode, SEEK_DATA, |
601 | es.es_lblk + es.es_len, &dataoff)) | |
602 | break; | |
603 | last += es.es_len; | |
ad7fefb1 | 604 | dataoff = (loff_t)last << blkbits; |
2d90c160 | 605 | cond_resched(); |
ad7fefb1 | 606 | } while (last <= end); |
c8c0df24 | 607 | |
5955102c | 608 | inode_unlock(inode); |
c8c0df24 | 609 | |
ad7fefb1 TT |
610 | if (dataoff > isize) |
611 | return -ENXIO; | |
612 | ||
613 | return vfs_setpos(file, dataoff, maxsize); | |
c8c0df24 ZL |
614 | } |
615 | ||
616 | /* | |
ad7fefb1 | 617 | * ext4_seek_hole() retrieves the offset for SEEK_HOLE. |
c8c0df24 ZL |
618 | */ |
619 | static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize) | |
620 | { | |
621 | struct inode *inode = file->f_mapping->host; | |
ad7fefb1 TT |
622 | struct extent_status es; |
623 | ext4_lblk_t start, last, end; | |
624 | loff_t holeoff, isize; | |
625 | int blkbits; | |
2d90c160 | 626 | int ret; |
c8c0df24 | 627 | |
5955102c | 628 | inode_lock(inode); |
ad7fefb1 TT |
629 | |
630 | isize = i_size_read(inode); | |
f289f8c0 | 631 | if (offset < 0 || offset >= isize) { |
5955102c | 632 | inode_unlock(inode); |
c8c0df24 ZL |
633 | return -ENXIO; |
634 | } | |
635 | ||
ad7fefb1 TT |
636 | blkbits = inode->i_sb->s_blocksize_bits; |
637 | start = offset >> blkbits; | |
638 | last = start; | |
639 | end = isize >> blkbits; | |
640 | holeoff = offset; | |
c8c0df24 | 641 | |
ad7fefb1 | 642 | do { |
2d90c160 JK |
643 | ret = ext4_get_next_extent(inode, last, end - last + 1, &es); |
644 | if (ret < 0) { | |
645 | inode_unlock(inode); | |
646 | return ret; | |
ad7fefb1 | 647 | } |
2d90c160 JK |
648 | /* Found a hole? */ |
649 | if (ret == 0 || es.es_lblk > last) { | |
650 | if (last != start) | |
651 | holeoff = (loff_t)last << blkbits; | |
652 | break; | |
ad7fefb1 | 653 | } |
ad7fefb1 TT |
654 | /* |
655 | * If there is a unwritten extent at this offset, | |
656 | * it will be as a data or a hole according to page | |
657 | * cache that has data or not. | |
658 | */ | |
2d90c160 JK |
659 | if (ext4_es_is_unwritten(&es) && |
660 | ext4_find_unwritten_pgoff(inode, SEEK_HOLE, | |
661 | last + es.es_len, &holeoff)) | |
662 | break; | |
ad7fefb1 | 663 | |
2d90c160 JK |
664 | last += es.es_len; |
665 | holeoff = (loff_t)last << blkbits; | |
666 | cond_resched(); | |
ad7fefb1 TT |
667 | } while (last <= end); |
668 | ||
5955102c | 669 | inode_unlock(inode); |
c8c0df24 | 670 | |
ad7fefb1 TT |
671 | if (holeoff > isize) |
672 | holeoff = isize; | |
673 | ||
674 | return vfs_setpos(file, holeoff, maxsize); | |
c8c0df24 ZL |
675 | } |
676 | ||
e0d10bfa | 677 | /* |
ec7268ce ES |
678 | * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values |
679 | * by calling generic_file_llseek_size() with the appropriate maxbytes | |
680 | * value for each. | |
e0d10bfa | 681 | */ |
965c8e59 | 682 | loff_t ext4_llseek(struct file *file, loff_t offset, int whence) |
e0d10bfa TO |
683 | { |
684 | struct inode *inode = file->f_mapping->host; | |
685 | loff_t maxbytes; | |
686 | ||
687 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) | |
688 | maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes; | |
689 | else | |
690 | maxbytes = inode->i_sb->s_maxbytes; | |
e0d10bfa | 691 | |
965c8e59 | 692 | switch (whence) { |
c8c0df24 ZL |
693 | case SEEK_SET: |
694 | case SEEK_CUR: | |
695 | case SEEK_END: | |
965c8e59 | 696 | return generic_file_llseek_size(file, offset, whence, |
c8c0df24 ZL |
697 | maxbytes, i_size_read(inode)); |
698 | case SEEK_DATA: | |
699 | return ext4_seek_data(file, offset, maxbytes); | |
700 | case SEEK_HOLE: | |
701 | return ext4_seek_hole(file, offset, maxbytes); | |
702 | } | |
703 | ||
704 | return -EINVAL; | |
e0d10bfa TO |
705 | } |
706 | ||
617ba13b | 707 | const struct file_operations ext4_file_operations = { |
e0d10bfa | 708 | .llseek = ext4_llseek, |
364443cb | 709 | .read_iter = ext4_file_read_iter, |
9b884164 | 710 | .write_iter = ext4_file_write_iter, |
5cdd7b2d | 711 | .unlocked_ioctl = ext4_ioctl, |
ac27a0ec | 712 | #ifdef CONFIG_COMPAT |
617ba13b | 713 | .compat_ioctl = ext4_compat_ioctl, |
ac27a0ec | 714 | #endif |
2e9ee850 | 715 | .mmap = ext4_file_mmap, |
bc0b0d6d | 716 | .open = ext4_file_open, |
617ba13b MC |
717 | .release = ext4_release_file, |
718 | .fsync = ext4_sync_file, | |
dbe6ec81 | 719 | .get_unmapped_area = thp_get_unmapped_area, |
ac27a0ec | 720 | .splice_read = generic_file_splice_read, |
8d020765 | 721 | .splice_write = iter_file_splice_write, |
2fe17c10 | 722 | .fallocate = ext4_fallocate, |
ac27a0ec DK |
723 | }; |
724 | ||
754661f1 | 725 | const struct inode_operations ext4_file_inode_operations = { |
617ba13b | 726 | .setattr = ext4_setattr, |
99652ea5 | 727 | .getattr = ext4_file_getattr, |
617ba13b | 728 | .listxattr = ext4_listxattr, |
4e34e719 | 729 | .get_acl = ext4_get_acl, |
64e178a7 | 730 | .set_acl = ext4_set_acl, |
6873fa0d | 731 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
732 | }; |
733 |