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> | |
dab291af | 23 | #include <linux/jbd2.h> |
bc0b0d6d TT |
24 | #include <linux/mount.h> |
25 | #include <linux/path.h> | |
a27bb332 | 26 | #include <linux/aio.h> |
871a2931 | 27 | #include <linux/quotaops.h> |
c8c0df24 | 28 | #include <linux/pagevec.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 | ||
c278531d | 60 | 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 | |
77 | ext4_unaligned_aio(struct inode *inode, const struct iovec *iov, | |
78 | unsigned long nr_segs, loff_t pos) | |
79 | { | |
80 | struct super_block *sb = inode->i_sb; | |
81 | int blockmask = sb->s_blocksize - 1; | |
82 | size_t count = iov_length(iov, nr_segs); | |
83 | loff_t final_size = pos + count; | |
84 | ||
fcc742c2 | 85 | if (pos >= i_size_read(inode)) |
e9e3bcec ES |
86 | return 0; |
87 | ||
88 | if ((pos & blockmask) || (final_size & blockmask)) | |
89 | return 1; | |
90 | ||
91 | return 0; | |
92 | } | |
93 | ||
ac27a0ec | 94 | static ssize_t |
fbe10494 ZL |
95 | ext4_file_dio_write(struct kiocb *iocb, const struct iovec *iov, |
96 | unsigned long nr_segs, loff_t pos) | |
ac27a0ec | 97 | { |
4bd809db ZL |
98 | struct file *file = iocb->ki_filp; |
99 | struct inode *inode = file->f_mapping->host; | |
100 | struct blk_plug plug; | |
e9e3bcec | 101 | int unaligned_aio = 0; |
8563000d | 102 | ssize_t ret; |
30d8c835 | 103 | int *overwrite = iocb->private; |
4bd809db | 104 | size_t length = iov_length(iov, nr_segs); |
ac27a0ec | 105 | |
fbe10494 ZL |
106 | if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) && |
107 | !is_sync_kiocb(iocb)) | |
e9e3bcec | 108 | unaligned_aio = ext4_unaligned_aio(inode, iov, nr_segs, pos); |
e2b46574 | 109 | |
e9e3bcec ES |
110 | /* Unaligned direct AIO must be serialized; see comment above */ |
111 | if (unaligned_aio) { | |
e9e3bcec | 112 | mutex_lock(ext4_aio_mutex(inode)); |
e27f41e1 | 113 | ext4_unwritten_wait(inode); |
e9e3bcec ES |
114 | } |
115 | ||
4bd809db ZL |
116 | BUG_ON(iocb->ki_pos != pos); |
117 | ||
118 | mutex_lock(&inode->i_mutex); | |
119 | blk_start_plug(&plug); | |
120 | ||
4bd809db ZL |
121 | /* check whether we do a DIO overwrite or not */ |
122 | if (ext4_should_dioread_nolock(inode) && !unaligned_aio && | |
123 | !file->f_mapping->nrpages && pos + length <= i_size_read(inode)) { | |
124 | struct ext4_map_blocks map; | |
125 | unsigned int blkbits = inode->i_blkbits; | |
126 | int err, len; | |
127 | ||
128 | map.m_lblk = pos >> blkbits; | |
129 | map.m_len = (EXT4_BLOCK_ALIGN(pos + length, blkbits) >> blkbits) | |
130 | - map.m_lblk; | |
131 | len = map.m_len; | |
132 | ||
133 | err = ext4_map_blocks(NULL, inode, &map, 0); | |
134 | /* | |
135 | * 'err==len' means that all of blocks has been preallocated no | |
136 | * matter they are initialized or not. For excluding | |
137 | * uninitialized extents, we need to check m_flags. There are | |
138 | * two conditions that indicate for initialized extents. | |
139 | * 1) If we hit extent cache, EXT4_MAP_MAPPED flag is returned; | |
140 | * 2) If we do a real lookup, non-flags are returned. | |
141 | * So we should check these two conditions. | |
142 | */ | |
143 | if (err == len && (map.m_flags & EXT4_MAP_MAPPED)) | |
30d8c835 | 144 | *overwrite = 1; |
4bd809db ZL |
145 | } |
146 | ||
147 | ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos); | |
148 | mutex_unlock(&inode->i_mutex); | |
149 | ||
150 | if (ret > 0 || ret == -EIOCBQUEUED) { | |
151 | ssize_t err; | |
152 | ||
153 | err = generic_write_sync(file, pos, ret); | |
154 | if (err < 0 && ret > 0) | |
155 | ret = err; | |
156 | } | |
157 | blk_finish_plug(&plug); | |
e9e3bcec ES |
158 | |
159 | if (unaligned_aio) | |
160 | mutex_unlock(ext4_aio_mutex(inode)); | |
161 | ||
162 | return ret; | |
ac27a0ec DK |
163 | } |
164 | ||
ac27a0ec | 165 | static ssize_t |
617ba13b | 166 | ext4_file_write(struct kiocb *iocb, const struct iovec *iov, |
ac27a0ec DK |
167 | unsigned long nr_segs, loff_t pos) |
168 | { | |
496ad9aa | 169 | struct inode *inode = file_inode(iocb->ki_filp); |
8563000d | 170 | ssize_t ret; |
30d8c835 | 171 | int overwrite = 0; |
ac27a0ec | 172 | |
e2b46574 ES |
173 | /* |
174 | * If we have encountered a bitmap-format file, the size limit | |
175 | * is smaller than s_maxbytes, which is for extent-mapped files. | |
176 | */ | |
177 | ||
12e9b892 | 178 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) { |
e2b46574 ES |
179 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
180 | size_t length = iov_length(iov, nr_segs); | |
ac27a0ec | 181 | |
d889dc83 TO |
182 | if ((pos > sbi->s_bitmap_maxbytes || |
183 | (pos == sbi->s_bitmap_maxbytes && length > 0))) | |
e2b46574 ES |
184 | return -EFBIG; |
185 | ||
186 | if (pos + length > sbi->s_bitmap_maxbytes) { | |
187 | nr_segs = iov_shorten((struct iovec *)iov, nr_segs, | |
188 | sbi->s_bitmap_maxbytes - pos); | |
189 | } | |
190 | } | |
191 | ||
30d8c835 | 192 | iocb->private = &overwrite; |
fbe10494 ZL |
193 | if (unlikely(iocb->ki_filp->f_flags & O_DIRECT)) |
194 | ret = ext4_file_dio_write(iocb, iov, nr_segs, pos); | |
195 | else | |
196 | ret = generic_file_aio_write(iocb, iov, nr_segs, pos); | |
e9e3bcec ES |
197 | |
198 | return ret; | |
ac27a0ec DK |
199 | } |
200 | ||
f0f37e2f | 201 | static const struct vm_operations_struct ext4_file_vm_ops = { |
2e9ee850 AK |
202 | .fault = filemap_fault, |
203 | .page_mkwrite = ext4_page_mkwrite, | |
0b173bc4 | 204 | .remap_pages = generic_file_remap_pages, |
2e9ee850 AK |
205 | }; |
206 | ||
207 | static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma) | |
208 | { | |
209 | struct address_space *mapping = file->f_mapping; | |
210 | ||
211 | if (!mapping->a_ops->readpage) | |
212 | return -ENOEXEC; | |
213 | file_accessed(file); | |
214 | vma->vm_ops = &ext4_file_vm_ops; | |
2e9ee850 AK |
215 | return 0; |
216 | } | |
217 | ||
bc0b0d6d TT |
218 | static int ext4_file_open(struct inode * inode, struct file * filp) |
219 | { | |
220 | struct super_block *sb = inode->i_sb; | |
221 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
8aefcd55 | 222 | struct ext4_inode_info *ei = EXT4_I(inode); |
bc0b0d6d TT |
223 | struct vfsmount *mnt = filp->f_path.mnt; |
224 | struct path path; | |
225 | char buf[64], *cp; | |
226 | ||
227 | if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) && | |
228 | !(sb->s_flags & MS_RDONLY))) { | |
229 | sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED; | |
230 | /* | |
231 | * Sample where the filesystem has been mounted and | |
232 | * store it in the superblock for sysadmin convenience | |
233 | * when trying to sort through large numbers of block | |
234 | * devices or filesystem images. | |
235 | */ | |
236 | memset(buf, 0, sizeof(buf)); | |
3899167d AV |
237 | path.mnt = mnt; |
238 | path.dentry = mnt->mnt_root; | |
bc0b0d6d | 239 | cp = d_path(&path, buf, sizeof(buf)); |
bc0b0d6d | 240 | if (!IS_ERR(cp)) { |
044ce47f JK |
241 | handle_t *handle; |
242 | int err; | |
243 | ||
9924a92a | 244 | handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1); |
044ce47f JK |
245 | if (IS_ERR(handle)) |
246 | return PTR_ERR(handle); | |
247 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); | |
248 | if (err) { | |
249 | ext4_journal_stop(handle); | |
250 | return err; | |
251 | } | |
cf803903 DW |
252 | strlcpy(sbi->s_es->s_last_mounted, cp, |
253 | sizeof(sbi->s_es->s_last_mounted)); | |
044ce47f JK |
254 | ext4_handle_dirty_super(handle, sb); |
255 | ext4_journal_stop(handle); | |
bc0b0d6d TT |
256 | } |
257 | } | |
8aefcd55 TT |
258 | /* |
259 | * Set up the jbd2_inode if we are opening the inode for | |
260 | * writing and the journal is present | |
261 | */ | |
262 | if (sbi->s_journal && !ei->jinode && (filp->f_mode & FMODE_WRITE)) { | |
263 | struct jbd2_inode *jinode = jbd2_alloc_inode(GFP_KERNEL); | |
264 | ||
265 | spin_lock(&inode->i_lock); | |
266 | if (!ei->jinode) { | |
267 | if (!jinode) { | |
268 | spin_unlock(&inode->i_lock); | |
269 | return -ENOMEM; | |
270 | } | |
271 | ei->jinode = jinode; | |
272 | jbd2_journal_init_jbd_inode(ei->jinode, inode); | |
273 | jinode = NULL; | |
274 | } | |
275 | spin_unlock(&inode->i_lock); | |
276 | if (unlikely(jinode != NULL)) | |
277 | jbd2_free_inode(jinode); | |
278 | } | |
907f4554 | 279 | return dquot_file_open(inode, filp); |
bc0b0d6d TT |
280 | } |
281 | ||
c8c0df24 ZL |
282 | /* |
283 | * Here we use ext4_map_blocks() to get a block mapping for a extent-based | |
284 | * file rather than ext4_ext_walk_space() because we can introduce | |
285 | * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same | |
286 | * function. When extent status tree has been fully implemented, it will | |
287 | * track all extent status for a file and we can directly use it to | |
288 | * retrieve the offset for SEEK_DATA/SEEK_HOLE. | |
289 | */ | |
290 | ||
291 | /* | |
292 | * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to | |
293 | * lookup page cache to check whether or not there has some data between | |
294 | * [startoff, endoff] because, if this range contains an unwritten extent, | |
295 | * we determine this extent as a data or a hole according to whether the | |
296 | * page cache has data or not. | |
297 | */ | |
298 | static int ext4_find_unwritten_pgoff(struct inode *inode, | |
965c8e59 | 299 | int whence, |
c8c0df24 ZL |
300 | struct ext4_map_blocks *map, |
301 | loff_t *offset) | |
302 | { | |
303 | struct pagevec pvec; | |
304 | unsigned int blkbits; | |
305 | pgoff_t index; | |
306 | pgoff_t end; | |
307 | loff_t endoff; | |
308 | loff_t startoff; | |
309 | loff_t lastoff; | |
310 | int found = 0; | |
311 | ||
312 | blkbits = inode->i_sb->s_blocksize_bits; | |
313 | startoff = *offset; | |
314 | lastoff = startoff; | |
80747f06 | 315 | endoff = (loff_t)(map->m_lblk + map->m_len) << blkbits; |
c8c0df24 ZL |
316 | |
317 | index = startoff >> PAGE_CACHE_SHIFT; | |
318 | end = endoff >> PAGE_CACHE_SHIFT; | |
319 | ||
320 | pagevec_init(&pvec, 0); | |
321 | do { | |
322 | int i, num; | |
323 | unsigned long nr_pages; | |
324 | ||
325 | num = min_t(pgoff_t, end - index, PAGEVEC_SIZE); | |
326 | nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index, | |
327 | (pgoff_t)num); | |
2f34aab1 | 328 | if (nr_pages == 0) |
c8c0df24 | 329 | break; |
c8c0df24 ZL |
330 | |
331 | for (i = 0; i < nr_pages; i++) { | |
332 | struct page *page = pvec.pages[i]; | |
333 | struct buffer_head *bh, *head; | |
334 | ||
335 | /* | |
2f34aab1 JK |
336 | * If current offset is smaller than the page offset, |
337 | * there is a hole at this offset. | |
c8c0df24 | 338 | */ |
2f34aab1 JK |
339 | if (whence == SEEK_HOLE && lastoff < endoff && |
340 | lastoff < page_offset(pvec.pages[i])) { | |
c8c0df24 ZL |
341 | found = 1; |
342 | *offset = lastoff; | |
343 | goto out; | |
344 | } | |
345 | ||
2f34aab1 JK |
346 | if (page->index > end) |
347 | goto out; | |
348 | ||
c8c0df24 ZL |
349 | lock_page(page); |
350 | ||
351 | if (unlikely(page->mapping != inode->i_mapping)) { | |
352 | unlock_page(page); | |
353 | continue; | |
354 | } | |
355 | ||
356 | if (!page_has_buffers(page)) { | |
357 | unlock_page(page); | |
358 | continue; | |
359 | } | |
360 | ||
361 | if (page_has_buffers(page)) { | |
362 | lastoff = page_offset(page); | |
363 | bh = head = page_buffers(page); | |
364 | do { | |
bdc5fbb3 JK |
365 | if (lastoff + bh->b_size <= startoff) |
366 | goto next; | |
c8c0df24 ZL |
367 | if (buffer_uptodate(bh) || |
368 | buffer_unwritten(bh)) { | |
965c8e59 | 369 | if (whence == SEEK_DATA) |
c8c0df24 ZL |
370 | found = 1; |
371 | } else { | |
965c8e59 | 372 | if (whence == SEEK_HOLE) |
c8c0df24 ZL |
373 | found = 1; |
374 | } | |
375 | if (found) { | |
376 | *offset = max_t(loff_t, | |
377 | startoff, lastoff); | |
378 | unlock_page(page); | |
379 | goto out; | |
380 | } | |
bdc5fbb3 | 381 | next: |
c8c0df24 ZL |
382 | lastoff += bh->b_size; |
383 | bh = bh->b_this_page; | |
384 | } while (bh != head); | |
385 | } | |
386 | ||
387 | lastoff = page_offset(page) + PAGE_SIZE; | |
388 | unlock_page(page); | |
389 | } | |
390 | ||
2f34aab1 JK |
391 | /* The no. of pages is less than our desired, we are done. */ |
392 | if (nr_pages < num) | |
c8c0df24 | 393 | break; |
c8c0df24 ZL |
394 | |
395 | index = pvec.pages[i - 1]->index + 1; | |
396 | pagevec_release(&pvec); | |
397 | } while (index <= end); | |
398 | ||
2f34aab1 JK |
399 | if (whence == SEEK_HOLE && lastoff < endoff) { |
400 | found = 1; | |
401 | *offset = lastoff; | |
402 | } | |
c8c0df24 ZL |
403 | out: |
404 | pagevec_release(&pvec); | |
405 | return found; | |
406 | } | |
407 | ||
408 | /* | |
409 | * ext4_seek_data() retrieves the offset for SEEK_DATA. | |
410 | */ | |
411 | static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize) | |
412 | { | |
413 | struct inode *inode = file->f_mapping->host; | |
414 | struct ext4_map_blocks map; | |
415 | struct extent_status es; | |
416 | ext4_lblk_t start, last, end; | |
417 | loff_t dataoff, isize; | |
418 | int blkbits; | |
419 | int ret = 0; | |
420 | ||
421 | mutex_lock(&inode->i_mutex); | |
422 | ||
423 | isize = i_size_read(inode); | |
ebb33aff | 424 | if (offset < 0 || offset >= isize) { |
c8c0df24 ZL |
425 | mutex_unlock(&inode->i_mutex); |
426 | return -ENXIO; | |
427 | } | |
428 | ||
429 | blkbits = inode->i_sb->s_blocksize_bits; | |
430 | start = offset >> blkbits; | |
431 | last = start; | |
432 | end = isize >> blkbits; | |
433 | dataoff = offset; | |
434 | ||
435 | do { | |
436 | map.m_lblk = last; | |
437 | map.m_len = end - last + 1; | |
438 | ret = ext4_map_blocks(NULL, inode, &map, 0); | |
439 | if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) { | |
440 | if (last != start) | |
80747f06 | 441 | dataoff = (loff_t)last << blkbits; |
c8c0df24 ZL |
442 | break; |
443 | } | |
444 | ||
445 | /* | |
446 | * If there is a delay extent at this offset, | |
447 | * it will be as a data. | |
448 | */ | |
e30b5dca | 449 | ext4_es_find_delayed_extent_range(inode, last, last, &es); |
06b0c886 | 450 | if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) { |
c8c0df24 | 451 | if (last != start) |
80747f06 | 452 | dataoff = (loff_t)last << blkbits; |
c8c0df24 ZL |
453 | break; |
454 | } | |
455 | ||
456 | /* | |
457 | * If there is a unwritten extent at this offset, | |
458 | * it will be as a data or a hole according to page | |
459 | * cache that has data or not. | |
460 | */ | |
461 | if (map.m_flags & EXT4_MAP_UNWRITTEN) { | |
462 | int unwritten; | |
463 | unwritten = ext4_find_unwritten_pgoff(inode, SEEK_DATA, | |
464 | &map, &dataoff); | |
465 | if (unwritten) | |
466 | break; | |
467 | } | |
468 | ||
469 | last++; | |
80747f06 | 470 | dataoff = (loff_t)last << blkbits; |
c8c0df24 ZL |
471 | } while (last <= end); |
472 | ||
473 | mutex_unlock(&inode->i_mutex); | |
474 | ||
475 | if (dataoff > isize) | |
476 | return -ENXIO; | |
477 | ||
478 | if (dataoff < 0 && !(file->f_mode & FMODE_UNSIGNED_OFFSET)) | |
479 | return -EINVAL; | |
480 | if (dataoff > maxsize) | |
481 | return -EINVAL; | |
482 | ||
483 | if (dataoff != file->f_pos) { | |
484 | file->f_pos = dataoff; | |
485 | file->f_version = 0; | |
486 | } | |
487 | ||
488 | return dataoff; | |
489 | } | |
490 | ||
491 | /* | |
492 | * ext4_seek_hole() retrieves the offset for SEEK_HOLE. | |
493 | */ | |
494 | static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize) | |
495 | { | |
496 | struct inode *inode = file->f_mapping->host; | |
497 | struct ext4_map_blocks map; | |
498 | struct extent_status es; | |
499 | ext4_lblk_t start, last, end; | |
500 | loff_t holeoff, isize; | |
501 | int blkbits; | |
502 | int ret = 0; | |
503 | ||
504 | mutex_lock(&inode->i_mutex); | |
505 | ||
506 | isize = i_size_read(inode); | |
ebb33aff | 507 | if (offset < 0 || offset >= isize) { |
c8c0df24 ZL |
508 | mutex_unlock(&inode->i_mutex); |
509 | return -ENXIO; | |
510 | } | |
511 | ||
512 | blkbits = inode->i_sb->s_blocksize_bits; | |
513 | start = offset >> blkbits; | |
514 | last = start; | |
515 | end = isize >> blkbits; | |
516 | holeoff = offset; | |
517 | ||
518 | do { | |
519 | map.m_lblk = last; | |
520 | map.m_len = end - last + 1; | |
521 | ret = ext4_map_blocks(NULL, inode, &map, 0); | |
522 | if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) { | |
523 | last += ret; | |
80747f06 | 524 | holeoff = (loff_t)last << blkbits; |
c8c0df24 ZL |
525 | continue; |
526 | } | |
527 | ||
528 | /* | |
529 | * If there is a delay extent at this offset, | |
530 | * we will skip this extent. | |
531 | */ | |
e30b5dca | 532 | ext4_es_find_delayed_extent_range(inode, last, last, &es); |
06b0c886 ZL |
533 | if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) { |
534 | last = es.es_lblk + es.es_len; | |
80747f06 | 535 | holeoff = (loff_t)last << blkbits; |
c8c0df24 ZL |
536 | continue; |
537 | } | |
538 | ||
539 | /* | |
540 | * If there is a unwritten extent at this offset, | |
541 | * it will be as a data or a hole according to page | |
542 | * cache that has data or not. | |
543 | */ | |
544 | if (map.m_flags & EXT4_MAP_UNWRITTEN) { | |
545 | int unwritten; | |
546 | unwritten = ext4_find_unwritten_pgoff(inode, SEEK_HOLE, | |
547 | &map, &holeoff); | |
548 | if (!unwritten) { | |
549 | last += ret; | |
80747f06 | 550 | holeoff = (loff_t)last << blkbits; |
c8c0df24 ZL |
551 | continue; |
552 | } | |
553 | } | |
554 | ||
555 | /* find a hole */ | |
556 | break; | |
557 | } while (last <= end); | |
558 | ||
559 | mutex_unlock(&inode->i_mutex); | |
560 | ||
561 | if (holeoff > isize) | |
562 | holeoff = isize; | |
563 | ||
564 | if (holeoff < 0 && !(file->f_mode & FMODE_UNSIGNED_OFFSET)) | |
565 | return -EINVAL; | |
566 | if (holeoff > maxsize) | |
567 | return -EINVAL; | |
568 | ||
569 | if (holeoff != file->f_pos) { | |
570 | file->f_pos = holeoff; | |
571 | file->f_version = 0; | |
572 | } | |
573 | ||
574 | return holeoff; | |
575 | } | |
576 | ||
e0d10bfa | 577 | /* |
ec7268ce ES |
578 | * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values |
579 | * by calling generic_file_llseek_size() with the appropriate maxbytes | |
580 | * value for each. | |
e0d10bfa | 581 | */ |
965c8e59 | 582 | loff_t ext4_llseek(struct file *file, loff_t offset, int whence) |
e0d10bfa TO |
583 | { |
584 | struct inode *inode = file->f_mapping->host; | |
585 | loff_t maxbytes; | |
586 | ||
587 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) | |
588 | maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes; | |
589 | else | |
590 | maxbytes = inode->i_sb->s_maxbytes; | |
e0d10bfa | 591 | |
965c8e59 | 592 | switch (whence) { |
c8c0df24 ZL |
593 | case SEEK_SET: |
594 | case SEEK_CUR: | |
595 | case SEEK_END: | |
965c8e59 | 596 | return generic_file_llseek_size(file, offset, whence, |
c8c0df24 ZL |
597 | maxbytes, i_size_read(inode)); |
598 | case SEEK_DATA: | |
599 | return ext4_seek_data(file, offset, maxbytes); | |
600 | case SEEK_HOLE: | |
601 | return ext4_seek_hole(file, offset, maxbytes); | |
602 | } | |
603 | ||
604 | return -EINVAL; | |
e0d10bfa TO |
605 | } |
606 | ||
617ba13b | 607 | const struct file_operations ext4_file_operations = { |
e0d10bfa | 608 | .llseek = ext4_llseek, |
ac27a0ec DK |
609 | .read = do_sync_read, |
610 | .write = do_sync_write, | |
611 | .aio_read = generic_file_aio_read, | |
617ba13b | 612 | .aio_write = ext4_file_write, |
5cdd7b2d | 613 | .unlocked_ioctl = ext4_ioctl, |
ac27a0ec | 614 | #ifdef CONFIG_COMPAT |
617ba13b | 615 | .compat_ioctl = ext4_compat_ioctl, |
ac27a0ec | 616 | #endif |
2e9ee850 | 617 | .mmap = ext4_file_mmap, |
bc0b0d6d | 618 | .open = ext4_file_open, |
617ba13b MC |
619 | .release = ext4_release_file, |
620 | .fsync = ext4_sync_file, | |
ac27a0ec DK |
621 | .splice_read = generic_file_splice_read, |
622 | .splice_write = generic_file_splice_write, | |
2fe17c10 | 623 | .fallocate = ext4_fallocate, |
ac27a0ec DK |
624 | }; |
625 | ||
754661f1 | 626 | const struct inode_operations ext4_file_inode_operations = { |
617ba13b | 627 | .setattr = ext4_setattr, |
3e3398a0 | 628 | .getattr = ext4_getattr, |
ac27a0ec DK |
629 | .setxattr = generic_setxattr, |
630 | .getxattr = generic_getxattr, | |
617ba13b | 631 | .listxattr = ext4_listxattr, |
ac27a0ec | 632 | .removexattr = generic_removexattr, |
4e34e719 | 633 | .get_acl = ext4_get_acl, |
6873fa0d | 634 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
635 | }; |
636 |