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 | { |
4bd809db | 94 | struct file *file = iocb->ki_filp; |
8ad2850f | 95 | struct inode *inode = file_inode(iocb->ki_filp); |
7ed07ba8 | 96 | struct mutex *aio_mutex = NULL; |
4bd809db | 97 | struct blk_plug plug; |
2ba48ce5 | 98 | int o_direct = iocb->ki_flags & IOCB_DIRECT; |
4bd809db | 99 | int overwrite = 0; |
8563000d | 100 | ssize_t ret; |
7608e610 | 101 | |
f5ccfe1d TT |
102 | /* |
103 | * Unaligned direct AIO must be serialized; see comment above | |
104 | * In the case of O_APPEND, assume that we must always serialize | |
105 | */ | |
106 | if (o_direct && | |
107 | ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) && | |
108 | !is_sync_kiocb(iocb) && | |
2ba48ce5 | 109 | (iocb->ki_flags & IOCB_APPEND || |
e768d7ff | 110 | ext4_unaligned_aio(inode, from, iocb->ki_pos))) { |
f5ccfe1d TT |
111 | aio_mutex = ext4_aio_mutex(inode); |
112 | mutex_lock(aio_mutex); | |
113 | ext4_unwritten_wait(inode); | |
114 | } | |
115 | ||
116 | mutex_lock(&inode->i_mutex); | |
3309dd04 AV |
117 | ret = generic_write_checks(iocb, from); |
118 | if (ret <= 0) | |
e768d7ff | 119 | goto out; |
f5ccfe1d | 120 | |
e2b46574 ES |
121 | /* |
122 | * If we have encountered a bitmap-format file, the size limit | |
123 | * is smaller than s_maxbytes, which is for extent-mapped files. | |
124 | */ | |
12e9b892 | 125 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) { |
e2b46574 | 126 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
ac27a0ec | 127 | |
3309dd04 | 128 | if (iocb->ki_pos >= sbi->s_bitmap_maxbytes) { |
f5ccfe1d | 129 | ret = -EFBIG; |
e768d7ff | 130 | goto out; |
f5ccfe1d | 131 | } |
3309dd04 | 132 | iov_iter_truncate(from, sbi->s_bitmap_maxbytes - iocb->ki_pos); |
e2b46574 ES |
133 | } |
134 | ||
a41537e6 | 135 | iocb->private = &overwrite; |
7ed07ba8 | 136 | if (o_direct) { |
3309dd04 AV |
137 | size_t length = iov_iter_count(from); |
138 | loff_t pos = iocb->ki_pos; | |
8ad2850f TT |
139 | blk_start_plug(&plug); |
140 | ||
8ad2850f | 141 | /* check whether we do a DIO overwrite or not */ |
7ed07ba8 | 142 | if (ext4_should_dioread_nolock(inode) && !aio_mutex && |
8ad2850f TT |
143 | !file->f_mapping->nrpages && pos + length <= i_size_read(inode)) { |
144 | struct ext4_map_blocks map; | |
145 | unsigned int blkbits = inode->i_blkbits; | |
146 | int err, len; | |
147 | ||
148 | map.m_lblk = pos >> blkbits; | |
149 | map.m_len = (EXT4_BLOCK_ALIGN(pos + length, blkbits) >> blkbits) | |
150 | - map.m_lblk; | |
151 | len = map.m_len; | |
152 | ||
153 | err = ext4_map_blocks(NULL, inode, &map, 0); | |
154 | /* | |
155 | * 'err==len' means that all of blocks has | |
156 | * been preallocated no matter they are | |
157 | * initialized or not. For excluding | |
158 | * unwritten extents, we need to check | |
159 | * m_flags. There are two conditions that | |
160 | * indicate for initialized extents. 1) If we | |
161 | * hit extent cache, EXT4_MAP_MAPPED flag is | |
162 | * returned; 2) If we do a real lookup, | |
163 | * non-flags are returned. So we should check | |
164 | * these two conditions. | |
165 | */ | |
166 | if (err == len && (map.m_flags & EXT4_MAP_MAPPED)) | |
167 | overwrite = 1; | |
168 | } | |
f5ccfe1d | 169 | } |
7608e610 | 170 | |
9b884164 | 171 | ret = __generic_file_write_iter(iocb, from); |
7ed07ba8 | 172 | mutex_unlock(&inode->i_mutex); |
7608e610 | 173 | |
7ed07ba8 TT |
174 | if (ret > 0) { |
175 | ssize_t err; | |
176 | ||
177 | err = generic_write_sync(file, iocb->ki_pos - ret, ret); | |
178 | if (err < 0) | |
179 | ret = err; | |
7608e610 | 180 | } |
7ed07ba8 TT |
181 | if (o_direct) |
182 | blk_finish_plug(&plug); | |
e9e3bcec | 183 | |
e768d7ff AV |
184 | if (aio_mutex) |
185 | mutex_unlock(aio_mutex); | |
186 | return ret; | |
187 | ||
188 | out: | |
189 | mutex_unlock(&inode->i_mutex); | |
7ed07ba8 TT |
190 | if (aio_mutex) |
191 | mutex_unlock(aio_mutex); | |
e9e3bcec | 192 | return ret; |
ac27a0ec DK |
193 | } |
194 | ||
923ae0ff | 195 | #ifdef CONFIG_FS_DAX |
e842f290 DC |
196 | static void ext4_end_io_unwritten(struct buffer_head *bh, int uptodate) |
197 | { | |
198 | struct inode *inode = bh->b_assoc_map->host; | |
e676a4c1 | 199 | /* XXX: breaks on 32-bit > 16TB. Is that even supported? */ |
e842f290 DC |
200 | loff_t offset = (loff_t)(uintptr_t)bh->b_private << inode->i_blkbits; |
201 | int err; | |
202 | if (!uptodate) | |
203 | return; | |
204 | WARN_ON(!buffer_unwritten(bh)); | |
205 | err = ext4_convert_unwritten_extents(NULL, inode, offset, bh->b_size); | |
206 | } | |
207 | ||
923ae0ff RZ |
208 | static int ext4_dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
209 | { | |
01a33b4a MW |
210 | int result; |
211 | handle_t *handle = NULL; | |
0b680de4 JK |
212 | struct inode *inode = file_inode(vma->vm_file); |
213 | struct super_block *sb = inode->i_sb; | |
01a33b4a MW |
214 | bool write = vmf->flags & FAULT_FLAG_WRITE; |
215 | ||
216 | if (write) { | |
217 | sb_start_pagefault(sb); | |
218 | file_update_time(vma->vm_file); | |
0b680de4 | 219 | down_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a MW |
220 | handle = ext4_journal_start_sb(sb, EXT4_HT_WRITE_PAGE, |
221 | EXT4_DATA_TRANS_BLOCKS(sb)); | |
0b680de4 JK |
222 | } else |
223 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
224 | |
225 | if (IS_ERR(handle)) | |
226 | result = VM_FAULT_SIGBUS; | |
227 | else | |
228 | result = __dax_fault(vma, vmf, ext4_get_block_dax, | |
229 | ext4_end_io_unwritten); | |
230 | ||
231 | if (write) { | |
232 | if (!IS_ERR(handle)) | |
233 | ext4_journal_stop(handle); | |
0b680de4 | 234 | up_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a | 235 | sb_end_pagefault(sb); |
0b680de4 JK |
236 | } else |
237 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
238 | |
239 | return result; | |
923ae0ff RZ |
240 | } |
241 | ||
11bd1a9e MW |
242 | static int ext4_dax_pmd_fault(struct vm_area_struct *vma, unsigned long addr, |
243 | pmd_t *pmd, unsigned int flags) | |
244 | { | |
01a33b4a MW |
245 | int result; |
246 | handle_t *handle = NULL; | |
247 | struct inode *inode = file_inode(vma->vm_file); | |
248 | struct super_block *sb = inode->i_sb; | |
249 | bool write = flags & FAULT_FLAG_WRITE; | |
250 | ||
251 | if (write) { | |
252 | sb_start_pagefault(sb); | |
253 | file_update_time(vma->vm_file); | |
0b680de4 | 254 | down_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a MW |
255 | handle = ext4_journal_start_sb(sb, EXT4_HT_WRITE_PAGE, |
256 | ext4_chunk_trans_blocks(inode, | |
257 | PMD_SIZE / PAGE_SIZE)); | |
0b680de4 JK |
258 | } else |
259 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
260 | |
261 | if (IS_ERR(handle)) | |
262 | result = VM_FAULT_SIGBUS; | |
263 | else | |
264 | result = __dax_pmd_fault(vma, addr, pmd, flags, | |
265 | ext4_get_block_dax, ext4_end_io_unwritten); | |
266 | ||
267 | if (write) { | |
268 | if (!IS_ERR(handle)) | |
269 | ext4_journal_stop(handle); | |
0b680de4 | 270 | up_read(&EXT4_I(inode)->i_mmap_sem); |
01a33b4a | 271 | sb_end_pagefault(sb); |
0b680de4 JK |
272 | } else |
273 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
01a33b4a MW |
274 | |
275 | return result; | |
11bd1a9e MW |
276 | } |
277 | ||
923ae0ff RZ |
278 | static int ext4_dax_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) |
279 | { | |
0b680de4 JK |
280 | int err; |
281 | struct inode *inode = file_inode(vma->vm_file); | |
282 | ||
283 | sb_start_pagefault(inode->i_sb); | |
284 | file_update_time(vma->vm_file); | |
285 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
286 | err = __dax_mkwrite(vma, vmf, ext4_get_block_dax, | |
287 | ext4_end_io_unwritten); | |
288 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
289 | sb_end_pagefault(inode->i_sb); | |
290 | ||
291 | return err; | |
292 | } | |
293 | ||
294 | /* | |
295 | * Handle write fault for VM_MIXEDMAP mappings. Similarly to ext4_dax_mkwrite() | |
296 | * handler we check for races agaist truncate. Note that since we cycle through | |
297 | * i_mmap_sem, we are sure that also any hole punching that began before we | |
298 | * were called is finished by now and so if it included part of the file we | |
299 | * are working on, our pte will get unmapped and the check for pte_same() in | |
300 | * wp_pfn_shared() fails. Thus fault gets retried and things work out as | |
301 | * desired. | |
302 | */ | |
303 | static int ext4_dax_pfn_mkwrite(struct vm_area_struct *vma, | |
304 | struct vm_fault *vmf) | |
305 | { | |
306 | struct inode *inode = file_inode(vma->vm_file); | |
307 | struct super_block *sb = inode->i_sb; | |
308 | int ret = VM_FAULT_NOPAGE; | |
309 | loff_t size; | |
310 | ||
311 | sb_start_pagefault(sb); | |
312 | file_update_time(vma->vm_file); | |
313 | down_read(&EXT4_I(inode)->i_mmap_sem); | |
314 | size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
315 | if (vmf->pgoff >= size) | |
316 | ret = VM_FAULT_SIGBUS; | |
317 | up_read(&EXT4_I(inode)->i_mmap_sem); | |
318 | sb_end_pagefault(sb); | |
319 | ||
320 | return ret; | |
923ae0ff RZ |
321 | } |
322 | ||
323 | static const struct vm_operations_struct ext4_dax_vm_ops = { | |
324 | .fault = ext4_dax_fault, | |
11bd1a9e | 325 | .pmd_fault = ext4_dax_pmd_fault, |
923ae0ff | 326 | .page_mkwrite = ext4_dax_mkwrite, |
0b680de4 | 327 | .pfn_mkwrite = ext4_dax_pfn_mkwrite, |
923ae0ff RZ |
328 | }; |
329 | #else | |
330 | #define ext4_dax_vm_ops ext4_file_vm_ops | |
331 | #endif | |
332 | ||
f0f37e2f | 333 | static const struct vm_operations_struct ext4_file_vm_ops = { |
0b680de4 | 334 | .fault = ext4_filemap_fault, |
f1820361 | 335 | .map_pages = filemap_map_pages, |
2e9ee850 AK |
336 | .page_mkwrite = ext4_page_mkwrite, |
337 | }; | |
338 | ||
339 | static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma) | |
340 | { | |
c9c7429c MH |
341 | struct inode *inode = file->f_mapping->host; |
342 | ||
343 | if (ext4_encrypted_inode(inode)) { | |
b7236e21 | 344 | int err = ext4_get_encryption_info(inode); |
c9c7429c MH |
345 | if (err) |
346 | return 0; | |
abdd438b TT |
347 | if (ext4_encryption_info(inode) == NULL) |
348 | return -ENOKEY; | |
c9c7429c | 349 | } |
2e9ee850 | 350 | file_accessed(file); |
923ae0ff RZ |
351 | if (IS_DAX(file_inode(file))) { |
352 | vma->vm_ops = &ext4_dax_vm_ops; | |
11bd1a9e | 353 | vma->vm_flags |= VM_MIXEDMAP | VM_HUGEPAGE; |
923ae0ff RZ |
354 | } else { |
355 | vma->vm_ops = &ext4_file_vm_ops; | |
356 | } | |
2e9ee850 AK |
357 | return 0; |
358 | } | |
359 | ||
bc0b0d6d TT |
360 | static int ext4_file_open(struct inode * inode, struct file * filp) |
361 | { | |
362 | struct super_block *sb = inode->i_sb; | |
363 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
364 | struct vfsmount *mnt = filp->f_path.mnt; | |
1cac41cb | 365 | struct dentry *dir; |
bc0b0d6d TT |
366 | struct path path; |
367 | char buf[64], *cp; | |
c9c7429c | 368 | int ret; |
bc0b0d6d TT |
369 | |
370 | if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) && | |
371 | !(sb->s_flags & MS_RDONLY))) { | |
372 | sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED; | |
373 | /* | |
374 | * Sample where the filesystem has been mounted and | |
375 | * store it in the superblock for sysadmin convenience | |
376 | * when trying to sort through large numbers of block | |
377 | * devices or filesystem images. | |
378 | */ | |
379 | memset(buf, 0, sizeof(buf)); | |
3899167d AV |
380 | path.mnt = mnt; |
381 | path.dentry = mnt->mnt_root; | |
bc0b0d6d | 382 | cp = d_path(&path, buf, sizeof(buf)); |
bc0b0d6d | 383 | if (!IS_ERR(cp)) { |
044ce47f JK |
384 | handle_t *handle; |
385 | int err; | |
386 | ||
9924a92a | 387 | handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1); |
044ce47f JK |
388 | if (IS_ERR(handle)) |
389 | return PTR_ERR(handle); | |
5d601255 | 390 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
044ce47f JK |
391 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); |
392 | if (err) { | |
393 | ext4_journal_stop(handle); | |
394 | return err; | |
395 | } | |
cf803903 DW |
396 | strlcpy(sbi->s_es->s_last_mounted, cp, |
397 | sizeof(sbi->s_es->s_last_mounted)); | |
044ce47f JK |
398 | ext4_handle_dirty_super(handle, sb); |
399 | ext4_journal_stop(handle); | |
bc0b0d6d TT |
400 | } |
401 | } | |
abdd438b TT |
402 | if (ext4_encrypted_inode(inode)) { |
403 | ret = ext4_get_encryption_info(inode); | |
404 | if (ret) | |
405 | return -EACCES; | |
406 | if (ext4_encryption_info(inode) == NULL) | |
407 | return -ENOKEY; | |
408 | } | |
1cac41cb MB |
409 | |
410 | dir = dget_parent(filp->f_path.dentry); | |
411 | if (ext4_encrypted_inode(d_inode(dir)) && | |
412 | !ext4_is_child_context_consistent_with_parent(d_inode(dir), inode)) { | |
413 | ext4_warning(inode->i_sb, | |
414 | "Inconsistent encryption contexts: %lu/%lu\n", | |
415 | (unsigned long) d_inode(dir)->i_ino, | |
416 | (unsigned long) inode->i_ino); | |
417 | dput(dir); | |
418 | return -EPERM; | |
419 | } | |
420 | dput(dir); | |
8aefcd55 TT |
421 | /* |
422 | * Set up the jbd2_inode if we are opening the inode for | |
423 | * writing and the journal is present | |
424 | */ | |
a361293f | 425 | if (filp->f_mode & FMODE_WRITE) { |
c9c7429c | 426 | ret = ext4_inode_attach_jinode(inode); |
a361293f JK |
427 | if (ret < 0) |
428 | return ret; | |
8aefcd55 | 429 | } |
abdd438b | 430 | return dquot_file_open(inode, filp); |
bc0b0d6d TT |
431 | } |
432 | ||
c8c0df24 ZL |
433 | /* |
434 | * Here we use ext4_map_blocks() to get a block mapping for a extent-based | |
435 | * file rather than ext4_ext_walk_space() because we can introduce | |
436 | * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same | |
437 | * function. When extent status tree has been fully implemented, it will | |
438 | * track all extent status for a file and we can directly use it to | |
439 | * retrieve the offset for SEEK_DATA/SEEK_HOLE. | |
440 | */ | |
441 | ||
442 | /* | |
443 | * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to | |
444 | * lookup page cache to check whether or not there has some data between | |
445 | * [startoff, endoff] because, if this range contains an unwritten extent, | |
446 | * we determine this extent as a data or a hole according to whether the | |
447 | * page cache has data or not. | |
448 | */ | |
ad7fefb1 TT |
449 | static int ext4_find_unwritten_pgoff(struct inode *inode, |
450 | int whence, | |
451 | struct ext4_map_blocks *map, | |
452 | loff_t *offset) | |
c8c0df24 ZL |
453 | { |
454 | struct pagevec pvec; | |
ad7fefb1 | 455 | unsigned int blkbits; |
c8c0df24 ZL |
456 | pgoff_t index; |
457 | pgoff_t end; | |
ad7fefb1 | 458 | loff_t endoff; |
c8c0df24 ZL |
459 | loff_t startoff; |
460 | loff_t lastoff; | |
461 | int found = 0; | |
462 | ||
ad7fefb1 | 463 | blkbits = inode->i_sb->s_blocksize_bits; |
c8c0df24 ZL |
464 | startoff = *offset; |
465 | lastoff = startoff; | |
ad7fefb1 | 466 | endoff = (loff_t)(map->m_lblk + map->m_len) << blkbits; |
c8c0df24 ZL |
467 | |
468 | index = startoff >> PAGE_CACHE_SHIFT; | |
469 | end = endoff >> PAGE_CACHE_SHIFT; | |
470 | ||
471 | pagevec_init(&pvec, 0); | |
472 | do { | |
473 | int i, num; | |
474 | unsigned long nr_pages; | |
475 | ||
476 | num = min_t(pgoff_t, end - index, PAGEVEC_SIZE); | |
477 | nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index, | |
478 | (pgoff_t)num); | |
08dc390b | 479 | if (nr_pages == 0) |
c8c0df24 | 480 | break; |
c8c0df24 ZL |
481 | |
482 | for (i = 0; i < nr_pages; i++) { | |
483 | struct page *page = pvec.pages[i]; | |
484 | struct buffer_head *bh, *head; | |
485 | ||
486 | /* | |
08dc390b JK |
487 | * If current offset is smaller than the page offset, |
488 | * there is a hole at this offset. | |
c8c0df24 | 489 | */ |
08dc390b JK |
490 | if (whence == SEEK_HOLE && lastoff < endoff && |
491 | lastoff < page_offset(pvec.pages[i])) { | |
c8c0df24 ZL |
492 | found = 1; |
493 | *offset = lastoff; | |
494 | goto out; | |
495 | } | |
496 | ||
08dc390b JK |
497 | if (page->index > end) |
498 | goto out; | |
499 | ||
c8c0df24 ZL |
500 | lock_page(page); |
501 | ||
502 | if (unlikely(page->mapping != inode->i_mapping)) { | |
503 | unlock_page(page); | |
504 | continue; | |
505 | } | |
506 | ||
507 | if (!page_has_buffers(page)) { | |
508 | unlock_page(page); | |
509 | continue; | |
510 | } | |
511 | ||
512 | if (page_has_buffers(page)) { | |
513 | lastoff = page_offset(page); | |
514 | bh = head = page_buffers(page); | |
515 | do { | |
bad9f614 JK |
516 | if (lastoff + bh->b_size <= startoff) |
517 | goto next; | |
c8c0df24 ZL |
518 | if (buffer_uptodate(bh) || |
519 | buffer_unwritten(bh)) { | |
965c8e59 | 520 | if (whence == SEEK_DATA) |
c8c0df24 ZL |
521 | found = 1; |
522 | } else { | |
965c8e59 | 523 | if (whence == SEEK_HOLE) |
c8c0df24 ZL |
524 | found = 1; |
525 | } | |
526 | if (found) { | |
527 | *offset = max_t(loff_t, | |
528 | startoff, lastoff); | |
529 | unlock_page(page); | |
530 | goto out; | |
531 | } | |
bad9f614 | 532 | next: |
c8c0df24 ZL |
533 | lastoff += bh->b_size; |
534 | bh = bh->b_this_page; | |
535 | } while (bh != head); | |
536 | } | |
537 | ||
538 | lastoff = page_offset(page) + PAGE_SIZE; | |
539 | unlock_page(page); | |
540 | } | |
541 | ||
08dc390b JK |
542 | /* The no. of pages is less than our desired, we are done. */ |
543 | if (nr_pages < num) | |
c8c0df24 | 544 | break; |
c8c0df24 ZL |
545 | |
546 | index = pvec.pages[i - 1]->index + 1; | |
547 | pagevec_release(&pvec); | |
548 | } while (index <= end); | |
549 | ||
08dc390b JK |
550 | if (whence == SEEK_HOLE && lastoff < endoff) { |
551 | found = 1; | |
552 | *offset = lastoff; | |
553 | } | |
c8c0df24 ZL |
554 | out: |
555 | pagevec_release(&pvec); | |
556 | return found; | |
557 | } | |
558 | ||
559 | /* | |
560 | * ext4_seek_data() retrieves the offset for SEEK_DATA. | |
561 | */ | |
562 | static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize) | |
563 | { | |
564 | struct inode *inode = file->f_mapping->host; | |
ad7fefb1 TT |
565 | struct ext4_map_blocks map; |
566 | struct extent_status es; | |
567 | ext4_lblk_t start, last, end; | |
568 | loff_t dataoff, isize; | |
569 | int blkbits; | |
570 | int ret = 0; | |
c8c0df24 ZL |
571 | |
572 | mutex_lock(&inode->i_mutex); | |
ad7fefb1 TT |
573 | |
574 | isize = i_size_read(inode); | |
bd368269 | 575 | if (offset < 0 || offset >= isize) { |
c8c0df24 ZL |
576 | mutex_unlock(&inode->i_mutex); |
577 | return -ENXIO; | |
578 | } | |
ad7fefb1 TT |
579 | |
580 | blkbits = inode->i_sb->s_blocksize_bits; | |
581 | start = offset >> blkbits; | |
582 | last = start; | |
583 | end = isize >> blkbits; | |
584 | dataoff = offset; | |
585 | ||
586 | do { | |
587 | map.m_lblk = last; | |
588 | map.m_len = end - last + 1; | |
589 | ret = ext4_map_blocks(NULL, inode, &map, 0); | |
590 | if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) { | |
591 | if (last != start) | |
592 | dataoff = (loff_t)last << blkbits; | |
c8c0df24 | 593 | break; |
ad7fefb1 | 594 | } |
c8c0df24 | 595 | |
ad7fefb1 TT |
596 | /* |
597 | * If there is a delay extent at this offset, | |
598 | * it will be as a data. | |
599 | */ | |
600 | ext4_es_find_delayed_extent_range(inode, last, last, &es); | |
601 | if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) { | |
602 | if (last != start) | |
603 | dataoff = (loff_t)last << blkbits; | |
c8c0df24 ZL |
604 | break; |
605 | } | |
606 | ||
ad7fefb1 TT |
607 | /* |
608 | * If there is a unwritten extent at this offset, | |
609 | * it will be as a data or a hole according to page | |
610 | * cache that has data or not. | |
611 | */ | |
612 | if (map.m_flags & EXT4_MAP_UNWRITTEN) { | |
613 | int unwritten; | |
614 | unwritten = ext4_find_unwritten_pgoff(inode, SEEK_DATA, | |
615 | &map, &dataoff); | |
616 | if (unwritten) | |
617 | break; | |
618 | } | |
c8c0df24 | 619 | |
ad7fefb1 TT |
620 | last++; |
621 | dataoff = (loff_t)last << blkbits; | |
622 | } while (last <= end); | |
c8c0df24 ZL |
623 | |
624 | mutex_unlock(&inode->i_mutex); | |
625 | ||
ad7fefb1 TT |
626 | if (dataoff > isize) |
627 | return -ENXIO; | |
628 | ||
629 | return vfs_setpos(file, dataoff, maxsize); | |
c8c0df24 ZL |
630 | } |
631 | ||
632 | /* | |
ad7fefb1 | 633 | * ext4_seek_hole() retrieves the offset for SEEK_HOLE. |
c8c0df24 ZL |
634 | */ |
635 | static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize) | |
636 | { | |
637 | struct inode *inode = file->f_mapping->host; | |
ad7fefb1 TT |
638 | struct ext4_map_blocks map; |
639 | struct extent_status es; | |
640 | ext4_lblk_t start, last, end; | |
641 | loff_t holeoff, isize; | |
642 | int blkbits; | |
643 | int ret = 0; | |
c8c0df24 ZL |
644 | |
645 | mutex_lock(&inode->i_mutex); | |
ad7fefb1 TT |
646 | |
647 | isize = i_size_read(inode); | |
bd368269 | 648 | if (offset < 0 || offset >= isize) { |
c8c0df24 ZL |
649 | mutex_unlock(&inode->i_mutex); |
650 | return -ENXIO; | |
651 | } | |
652 | ||
ad7fefb1 TT |
653 | blkbits = inode->i_sb->s_blocksize_bits; |
654 | start = offset >> blkbits; | |
655 | last = start; | |
656 | end = isize >> blkbits; | |
657 | holeoff = offset; | |
c8c0df24 | 658 | |
ad7fefb1 TT |
659 | do { |
660 | map.m_lblk = last; | |
661 | map.m_len = end - last + 1; | |
662 | ret = ext4_map_blocks(NULL, inode, &map, 0); | |
663 | if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) { | |
664 | last += ret; | |
665 | holeoff = (loff_t)last << blkbits; | |
666 | continue; | |
667 | } | |
c8c0df24 | 668 | |
ad7fefb1 TT |
669 | /* |
670 | * If there is a delay extent at this offset, | |
671 | * we will skip this extent. | |
672 | */ | |
673 | ext4_es_find_delayed_extent_range(inode, last, last, &es); | |
674 | if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) { | |
675 | last = es.es_lblk + es.es_len; | |
676 | holeoff = (loff_t)last << blkbits; | |
677 | continue; | |
678 | } | |
14516bb7 | 679 | |
ad7fefb1 TT |
680 | /* |
681 | * If there is a unwritten extent at this offset, | |
682 | * it will be as a data or a hole according to page | |
683 | * cache that has data or not. | |
684 | */ | |
685 | if (map.m_flags & EXT4_MAP_UNWRITTEN) { | |
686 | int unwritten; | |
687 | unwritten = ext4_find_unwritten_pgoff(inode, SEEK_HOLE, | |
688 | &map, &holeoff); | |
689 | if (!unwritten) { | |
690 | last += ret; | |
691 | holeoff = (loff_t)last << blkbits; | |
c8c0df24 ZL |
692 | continue; |
693 | } | |
14516bb7 | 694 | } |
ad7fefb1 TT |
695 | |
696 | /* find a hole */ | |
697 | break; | |
698 | } while (last <= end); | |
699 | ||
c8c0df24 ZL |
700 | mutex_unlock(&inode->i_mutex); |
701 | ||
ad7fefb1 TT |
702 | if (holeoff > isize) |
703 | holeoff = isize; | |
704 | ||
705 | return vfs_setpos(file, holeoff, maxsize); | |
c8c0df24 ZL |
706 | } |
707 | ||
e0d10bfa | 708 | /* |
ec7268ce ES |
709 | * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values |
710 | * by calling generic_file_llseek_size() with the appropriate maxbytes | |
711 | * value for each. | |
e0d10bfa | 712 | */ |
965c8e59 | 713 | loff_t ext4_llseek(struct file *file, loff_t offset, int whence) |
e0d10bfa TO |
714 | { |
715 | struct inode *inode = file->f_mapping->host; | |
716 | loff_t maxbytes; | |
717 | ||
718 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) | |
719 | maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes; | |
720 | else | |
721 | maxbytes = inode->i_sb->s_maxbytes; | |
e0d10bfa | 722 | |
965c8e59 | 723 | switch (whence) { |
c8c0df24 ZL |
724 | case SEEK_SET: |
725 | case SEEK_CUR: | |
726 | case SEEK_END: | |
965c8e59 | 727 | return generic_file_llseek_size(file, offset, whence, |
c8c0df24 ZL |
728 | maxbytes, i_size_read(inode)); |
729 | case SEEK_DATA: | |
730 | return ext4_seek_data(file, offset, maxbytes); | |
731 | case SEEK_HOLE: | |
732 | return ext4_seek_hole(file, offset, maxbytes); | |
733 | } | |
734 | ||
735 | return -EINVAL; | |
e0d10bfa TO |
736 | } |
737 | ||
617ba13b | 738 | const struct file_operations ext4_file_operations = { |
e0d10bfa | 739 | .llseek = ext4_llseek, |
aad4f8bb | 740 | .read_iter = generic_file_read_iter, |
9b884164 | 741 | .write_iter = ext4_file_write_iter, |
5cdd7b2d | 742 | .unlocked_ioctl = ext4_ioctl, |
ac27a0ec | 743 | #ifdef CONFIG_COMPAT |
617ba13b | 744 | .compat_ioctl = ext4_compat_ioctl, |
ac27a0ec | 745 | #endif |
2e9ee850 | 746 | .mmap = ext4_file_mmap, |
bc0b0d6d | 747 | .open = ext4_file_open, |
617ba13b MC |
748 | .release = ext4_release_file, |
749 | .fsync = ext4_sync_file, | |
ac27a0ec | 750 | .splice_read = generic_file_splice_read, |
8d020765 | 751 | .splice_write = iter_file_splice_write, |
2fe17c10 | 752 | .fallocate = ext4_fallocate, |
ac27a0ec DK |
753 | }; |
754 | ||
754661f1 | 755 | const struct inode_operations ext4_file_inode_operations = { |
617ba13b | 756 | .setattr = ext4_setattr, |
3e3398a0 | 757 | .getattr = ext4_getattr, |
ac27a0ec DK |
758 | .setxattr = generic_setxattr, |
759 | .getxattr = generic_getxattr, | |
617ba13b | 760 | .listxattr = ext4_listxattr, |
ac27a0ec | 761 | .removexattr = generic_removexattr, |
4e34e719 | 762 | .get_acl = ext4_get_acl, |
64e178a7 | 763 | .set_acl = ext4_set_acl, |
6873fa0d | 764 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
765 | }; |
766 |