blob: 8131be8c0af3166aac865557baa9f0371564a397 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/file.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
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 *
Mingming Cao617ba132006-10-11 01:20:53 -070015 * ext4 fs regular file handling primitives
Dave Kleikampac27a0e2006-10-11 01:20:50 -070016 *
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>
Mingming Caodab291a2006-10-11 01:21:01 -070023#include <linux/jbd2.h>
Theodore Ts'obc0b0d62009-06-13 10:09:48 -040024#include <linux/mount.h>
25#include <linux/path.h>
Kent Overstreeta27bb332013-05-07 16:19:08 -070026#include <linux/aio.h>
Christoph Hellwig871a2932010-03-03 09:05:07 -050027#include <linux/quotaops.h>
Zheng Liuc8c0df22012-11-08 21:57:40 -050028#include <linux/pagevec.h>
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040029#include "ext4.h"
30#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include "xattr.h"
32#include "acl.h"
33
34/*
35 * Called when an inode is released. Note that this is different
Mingming Cao617ba132006-10-11 01:20:53 -070036 * from ext4_file_open: open gets called at every open, but release
Dave Kleikampac27a0e2006-10-11 01:20:50 -070037 * gets called only when /all/ the files are closed.
38 */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -040039static int ext4_release_file(struct inode *inode, struct file *filp)
Dave Kleikampac27a0e2006-10-11 01:20:50 -070040{
Theodore Ts'o19f5fb72010-01-24 14:34:07 -050041 if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE)) {
Theodore Ts'o7d8f9f72009-02-24 08:21:14 -050042 ext4_alloc_da_blocks(inode);
Theodore Ts'o19f5fb72010-01-24 14:34:07 -050043 ext4_clear_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
Theodore Ts'o7d8f9f72009-02-24 08:21:14 -050044 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -070045 /* if we are the last writer on the inode, drop the block reservation */
46 if ((filp->f_mode & FMODE_WRITE) &&
Aneesh Kumar K.Vd6014302009-03-27 22:36:43 -040047 (atomic_read(&inode->i_writecount) == 1) &&
48 !EXT4_I(inode)->i_reserved_data_blocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -070049 {
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -050050 down_write(&EXT4_I(inode)->i_data_sem);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -040051 ext4_discard_preallocations(inode);
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -050052 up_write(&EXT4_I(inode)->i_data_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070053 }
54 if (is_dx(inode) && filp->private_data)
Mingming Cao617ba132006-10-11 01:20:53 -070055 ext4_htree_free_dir_info(filp->private_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070056
57 return 0;
58}
59
Stephen Hemmingerc1978552014-05-12 10:50:23 -040060static void ext4_unwritten_wait(struct inode *inode)
Eric Sandeene9e3bce2011-02-12 08:17:34 -050061{
62 wait_queue_head_t *wq = ext4_ioend_wq(inode);
63
Dmitry Monakhove27f41e2012-09-28 23:24:52 -040064 wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_unwritten) == 0));
Eric Sandeene9e3bce2011-02-12 08:17:34 -050065}
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 */
76static int
Al Viro9b884162014-04-17 16:09:22 -040077ext4_unaligned_aio(struct inode *inode, struct iov_iter *from, loff_t pos)
Eric Sandeene9e3bce2011-02-12 08:17:34 -050078{
79 struct super_block *sb = inode->i_sb;
80 int blockmask = sb->s_blocksize - 1;
Eric Sandeene9e3bce2011-02-12 08:17:34 -050081
Theodore Ts'o6e6358f2014-04-12 12:45:25 -040082 if (pos >= i_size_read(inode))
Eric Sandeene9e3bce2011-02-12 08:17:34 -050083 return 0;
84
Al Viro9b884162014-04-17 16:09:22 -040085 if ((pos | iov_iter_alignment(from)) & blockmask)
Eric Sandeene9e3bce2011-02-12 08:17:34 -050086 return 1;
87
88 return 0;
89}
90
Dave Kleikampac27a0e2006-10-11 01:20:50 -070091static ssize_t
Al Viro9b884162014-04-17 16:09:22 -040092ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
Dave Kleikampac27a0e2006-10-11 01:20:50 -070093{
Theodore Ts'o7608e612014-04-21 14:26:28 -040094 struct file *file = iocb->ki_filp;
Al Viro496ad9a2013-01-23 17:07:38 -050095 struct inode *inode = file_inode(iocb->ki_filp);
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -040096 struct mutex *aio_mutex = NULL;
Theodore Ts'o8ad28502014-04-21 14:26:57 -040097 struct blk_plug plug;
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -040098 int o_direct = file->f_flags & O_DIRECT;
Theodore Ts'o8ad28502014-04-21 14:26:57 -040099 int overwrite = 0;
Al Viro9b884162014-04-17 16:09:22 -0400100 size_t length = iov_iter_count(from);
Zheng Liu85630002012-05-28 18:06:51 -0400101 ssize_t ret;
Al Viro9b884162014-04-17 16:09:22 -0400102 loff_t pos = iocb->ki_pos;
Theodore Ts'o7608e612014-04-21 14:26:28 -0400103
Eric Sandeene2b46572008-01-28 23:58:27 -0500104 /*
Theodore Ts'of5ccfe12014-04-21 14:37:52 -0400105 * Unaligned direct AIO must be serialized; see comment above
106 * In the case of O_APPEND, assume that we must always serialize
107 */
108 if (o_direct &&
109 ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) &&
110 !is_sync_kiocb(iocb) &&
111 (file->f_flags & O_APPEND ||
Al Viro9b884162014-04-17 16:09:22 -0400112 ext4_unaligned_aio(inode, from, pos))) {
Theodore Ts'of5ccfe12014-04-21 14:37:52 -0400113 aio_mutex = ext4_aio_mutex(inode);
114 mutex_lock(aio_mutex);
115 ext4_unwritten_wait(inode);
116 }
117
118 mutex_lock(&inode->i_mutex);
119 if (file->f_flags & O_APPEND)
120 iocb->ki_pos = pos = i_size_read(inode);
121
122 /*
Eric Sandeene2b46572008-01-28 23:58:27 -0500123 * If we have encountered a bitmap-format file, the size limit
124 * is smaller than s_maxbytes, which is for extent-mapped files.
125 */
Dmitry Monakhov12e9b892010-05-16 22:00:00 -0400126 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
Eric Sandeene2b46572008-01-28 23:58:27 -0500127 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
Eric Sandeene2b46572008-01-28 23:58:27 -0500128
Theodore Ts'of5ccfe12014-04-21 14:37:52 -0400129 if ((pos > sbi->s_bitmap_maxbytes) ||
130 (pos == sbi->s_bitmap_maxbytes && length > 0)) {
131 mutex_unlock(&inode->i_mutex);
132 ret = -EFBIG;
133 goto errout;
134 }
Eric Sandeene2b46572008-01-28 23:58:27 -0500135
Al Viro9b884162014-04-17 16:09:22 -0400136 if (pos + length > sbi->s_bitmap_maxbytes)
137 iov_iter_truncate(from, sbi->s_bitmap_maxbytes - pos);
Eric Sandeene2b46572008-01-28 23:58:27 -0500138 }
139
Dmitry Monakhova41537e2014-10-30 10:53:16 -0400140 iocb->private = &overwrite;
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -0400141 if (o_direct) {
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400142 blk_start_plug(&plug);
143
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400144
145 /* check whether we do a DIO overwrite or not */
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -0400146 if (ext4_should_dioread_nolock(inode) && !aio_mutex &&
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400147 !file->f_mapping->nrpages && pos + length <= i_size_read(inode)) {
148 struct ext4_map_blocks map;
149 unsigned int blkbits = inode->i_blkbits;
150 int err, len;
151
152 map.m_lblk = pos >> blkbits;
153 map.m_len = (EXT4_BLOCK_ALIGN(pos + length, blkbits) >> blkbits)
154 - map.m_lblk;
155 len = map.m_len;
156
157 err = ext4_map_blocks(NULL, inode, &map, 0);
158 /*
159 * 'err==len' means that all of blocks has
160 * been preallocated no matter they are
161 * initialized or not. For excluding
162 * unwritten extents, we need to check
163 * m_flags. There are two conditions that
164 * indicate for initialized extents. 1) If we
165 * hit extent cache, EXT4_MAP_MAPPED flag is
166 * returned; 2) If we do a real lookup,
167 * non-flags are returned. So we should check
168 * these two conditions.
169 */
170 if (err == len && (map.m_flags & EXT4_MAP_MAPPED))
171 overwrite = 1;
172 }
Theodore Ts'of5ccfe12014-04-21 14:37:52 -0400173 }
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400174
Al Viro9b884162014-04-17 16:09:22 -0400175 ret = __generic_file_write_iter(iocb, from);
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -0400176 mutex_unlock(&inode->i_mutex);
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400177
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -0400178 if (ret > 0) {
179 ssize_t err;
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400180
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -0400181 err = generic_write_sync(file, iocb->ki_pos - ret, ret);
182 if (err < 0)
183 ret = err;
184 }
185 if (o_direct)
Theodore Ts'o8ad28502014-04-21 14:26:57 -0400186 blk_finish_plug(&plug);
187
Theodore Ts'of5ccfe12014-04-21 14:37:52 -0400188errout:
Theodore Ts'o7ed07ba2014-04-21 14:36:30 -0400189 if (aio_mutex)
190 mutex_unlock(aio_mutex);
Eric Sandeene9e3bce2011-02-12 08:17:34 -0500191 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700192}
193
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +0400194static const struct vm_operations_struct ext4_file_vm_ops = {
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400195 .fault = filemap_fault,
Kirill A. Shutemovf1820362014-04-07 15:37:19 -0700196 .map_pages = filemap_map_pages,
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400197 .page_mkwrite = ext4_page_mkwrite,
Konstantin Khlebnikov0b173bc2012-10-08 16:28:46 -0700198 .remap_pages = generic_file_remap_pages,
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400199};
200
201static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma)
202{
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400203 file_accessed(file);
204 vma->vm_ops = &ext4_file_vm_ops;
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400205 return 0;
206}
207
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400208static int ext4_file_open(struct inode * inode, struct file * filp)
209{
210 struct super_block *sb = inode->i_sb;
211 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
212 struct vfsmount *mnt = filp->f_path.mnt;
213 struct path path;
214 char buf[64], *cp;
215
216 if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) &&
217 !(sb->s_flags & MS_RDONLY))) {
218 sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED;
219 /*
220 * Sample where the filesystem has been mounted and
221 * store it in the superblock for sysadmin convenience
222 * when trying to sort through large numbers of block
223 * devices or filesystem images.
224 */
225 memset(buf, 0, sizeof(buf));
Al Viro38991672010-01-23 20:10:29 -0500226 path.mnt = mnt;
227 path.dentry = mnt->mnt_root;
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400228 cp = d_path(&path, buf, sizeof(buf));
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400229 if (!IS_ERR(cp)) {
Jan Kara044ce472012-07-22 20:31:31 -0400230 handle_t *handle;
231 int err;
232
Theodore Ts'o9924a922013-02-08 21:59:22 -0500233 handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
Jan Kara044ce472012-07-22 20:31:31 -0400234 if (IS_ERR(handle))
235 return PTR_ERR(handle);
liang xie5d601252014-05-12 22:06:43 -0400236 BUFFER_TRACE(sbi->s_sbh, "get_write_access");
Jan Kara044ce472012-07-22 20:31:31 -0400237 err = ext4_journal_get_write_access(handle, sbi->s_sbh);
238 if (err) {
239 ext4_journal_stop(handle);
240 return err;
241 }
Darrick J. Wongcf803902011-10-25 09:18:41 -0400242 strlcpy(sbi->s_es->s_last_mounted, cp,
243 sizeof(sbi->s_es->s_last_mounted));
Jan Kara044ce472012-07-22 20:31:31 -0400244 ext4_handle_dirty_super(handle, sb);
245 ext4_journal_stop(handle);
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400246 }
247 }
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500248 /*
249 * Set up the jbd2_inode if we are opening the inode for
250 * writing and the journal is present
251 */
Jan Karaa3612932013-08-16 21:19:41 -0400252 if (filp->f_mode & FMODE_WRITE) {
253 int ret = ext4_inode_attach_jinode(inode);
254 if (ret < 0)
255 return ret;
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500256 }
Christoph Hellwig907f4552010-03-03 09:05:06 -0500257 return dquot_file_open(inode, filp);
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400258}
259
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400260/*
Zheng Liuc8c0df22012-11-08 21:57:40 -0500261 * Here we use ext4_map_blocks() to get a block mapping for a extent-based
262 * file rather than ext4_ext_walk_space() because we can introduce
263 * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same
264 * function. When extent status tree has been fully implemented, it will
265 * track all extent status for a file and we can directly use it to
266 * retrieve the offset for SEEK_DATA/SEEK_HOLE.
267 */
268
269/*
270 * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to
271 * lookup page cache to check whether or not there has some data between
272 * [startoff, endoff] because, if this range contains an unwritten extent,
273 * we determine this extent as a data or a hole according to whether the
274 * page cache has data or not.
275 */
276static int ext4_find_unwritten_pgoff(struct inode *inode,
Andrew Morton965c8e52012-12-17 15:59:39 -0800277 int whence,
Zheng Liuc8c0df22012-11-08 21:57:40 -0500278 struct ext4_map_blocks *map,
279 loff_t *offset)
280{
281 struct pagevec pvec;
282 unsigned int blkbits;
283 pgoff_t index;
284 pgoff_t end;
285 loff_t endoff;
286 loff_t startoff;
287 loff_t lastoff;
288 int found = 0;
289
290 blkbits = inode->i_sb->s_blocksize_bits;
291 startoff = *offset;
292 lastoff = startoff;
Jan Karae7293fd2013-05-31 19:37:56 -0400293 endoff = (loff_t)(map->m_lblk + map->m_len) << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500294
295 index = startoff >> PAGE_CACHE_SHIFT;
296 end = endoff >> PAGE_CACHE_SHIFT;
297
298 pagevec_init(&pvec, 0);
299 do {
300 int i, num;
301 unsigned long nr_pages;
302
303 num = min_t(pgoff_t, end - index, PAGEVEC_SIZE);
304 nr_pages = pagevec_lookup(&pvec, inode->i_mapping, index,
305 (pgoff_t)num);
306 if (nr_pages == 0) {
Andrew Morton965c8e52012-12-17 15:59:39 -0800307 if (whence == SEEK_DATA)
Zheng Liuc8c0df22012-11-08 21:57:40 -0500308 break;
309
Andrew Morton965c8e52012-12-17 15:59:39 -0800310 BUG_ON(whence != SEEK_HOLE);
Zheng Liuc8c0df22012-11-08 21:57:40 -0500311 /*
312 * If this is the first time to go into the loop and
313 * offset is not beyond the end offset, it will be a
314 * hole at this offset
315 */
316 if (lastoff == startoff || lastoff < endoff)
317 found = 1;
318 break;
319 }
320
321 /*
322 * If this is the first time to go into the loop and
323 * offset is smaller than the first page offset, it will be a
324 * hole at this offset.
325 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800326 if (lastoff == startoff && whence == SEEK_HOLE &&
Zheng Liuc8c0df22012-11-08 21:57:40 -0500327 lastoff < page_offset(pvec.pages[0])) {
328 found = 1;
329 break;
330 }
331
332 for (i = 0; i < nr_pages; i++) {
333 struct page *page = pvec.pages[i];
334 struct buffer_head *bh, *head;
335
336 /*
337 * If the current offset is not beyond the end of given
338 * range, it will be a hole.
339 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800340 if (lastoff < endoff && whence == SEEK_HOLE &&
Zheng Liuc8c0df22012-11-08 21:57:40 -0500341 page->index > end) {
342 found = 1;
343 *offset = lastoff;
344 goto out;
345 }
346
347 lock_page(page);
348
349 if (unlikely(page->mapping != inode->i_mapping)) {
350 unlock_page(page);
351 continue;
352 }
353
354 if (!page_has_buffers(page)) {
355 unlock_page(page);
356 continue;
357 }
358
359 if (page_has_buffers(page)) {
360 lastoff = page_offset(page);
361 bh = head = page_buffers(page);
362 do {
363 if (buffer_uptodate(bh) ||
364 buffer_unwritten(bh)) {
Andrew Morton965c8e52012-12-17 15:59:39 -0800365 if (whence == SEEK_DATA)
Zheng Liuc8c0df22012-11-08 21:57:40 -0500366 found = 1;
367 } else {
Andrew Morton965c8e52012-12-17 15:59:39 -0800368 if (whence == SEEK_HOLE)
Zheng Liuc8c0df22012-11-08 21:57:40 -0500369 found = 1;
370 }
371 if (found) {
372 *offset = max_t(loff_t,
373 startoff, lastoff);
374 unlock_page(page);
375 goto out;
376 }
377 lastoff += bh->b_size;
378 bh = bh->b_this_page;
379 } while (bh != head);
380 }
381
382 lastoff = page_offset(page) + PAGE_SIZE;
383 unlock_page(page);
384 }
385
386 /*
387 * The no. of pages is less than our desired, that would be a
388 * hole in there.
389 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800390 if (nr_pages < num && whence == SEEK_HOLE) {
Zheng Liuc8c0df22012-11-08 21:57:40 -0500391 found = 1;
392 *offset = lastoff;
393 break;
394 }
395
396 index = pvec.pages[i - 1]->index + 1;
397 pagevec_release(&pvec);
398 } while (index <= end);
399
400out:
401 pagevec_release(&pvec);
402 return found;
403}
404
405/*
406 * ext4_seek_data() retrieves the offset for SEEK_DATA.
407 */
408static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
409{
410 struct inode *inode = file->f_mapping->host;
411 struct ext4_map_blocks map;
412 struct extent_status es;
413 ext4_lblk_t start, last, end;
414 loff_t dataoff, isize;
415 int blkbits;
416 int ret = 0;
417
418 mutex_lock(&inode->i_mutex);
419
420 isize = i_size_read(inode);
421 if (offset >= isize) {
422 mutex_unlock(&inode->i_mutex);
423 return -ENXIO;
424 }
425
426 blkbits = inode->i_sb->s_blocksize_bits;
427 start = offset >> blkbits;
428 last = start;
429 end = isize >> blkbits;
430 dataoff = offset;
431
432 do {
433 map.m_lblk = last;
434 map.m_len = end - last + 1;
435 ret = ext4_map_blocks(NULL, inode, &map, 0);
436 if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) {
437 if (last != start)
Jan Karae7293fd2013-05-31 19:37:56 -0400438 dataoff = (loff_t)last << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500439 break;
440 }
441
442 /*
443 * If there is a delay extent at this offset,
444 * it will be as a data.
445 */
Yan, Zhenge30b5dc2013-05-03 02:15:52 -0400446 ext4_es_find_delayed_extent_range(inode, last, last, &es);
Zheng Liu06b0c882013-02-18 00:26:51 -0500447 if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) {
Zheng Liuc8c0df22012-11-08 21:57:40 -0500448 if (last != start)
Jan Karae7293fd2013-05-31 19:37:56 -0400449 dataoff = (loff_t)last << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500450 break;
451 }
452
453 /*
454 * If there is a unwritten extent at this offset,
455 * it will be as a data or a hole according to page
456 * cache that has data or not.
457 */
458 if (map.m_flags & EXT4_MAP_UNWRITTEN) {
459 int unwritten;
460 unwritten = ext4_find_unwritten_pgoff(inode, SEEK_DATA,
461 &map, &dataoff);
462 if (unwritten)
463 break;
464 }
465
466 last++;
Jan Karae7293fd2013-05-31 19:37:56 -0400467 dataoff = (loff_t)last << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500468 } while (last <= end);
469
470 mutex_unlock(&inode->i_mutex);
471
472 if (dataoff > isize)
473 return -ENXIO;
474
Jie Liu46a1c2c2013-06-25 12:02:13 +0800475 return vfs_setpos(file, dataoff, maxsize);
Zheng Liuc8c0df22012-11-08 21:57:40 -0500476}
477
478/*
479 * ext4_seek_hole() retrieves the offset for SEEK_HOLE.
480 */
481static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize)
482{
483 struct inode *inode = file->f_mapping->host;
484 struct ext4_map_blocks map;
485 struct extent_status es;
486 ext4_lblk_t start, last, end;
487 loff_t holeoff, isize;
488 int blkbits;
489 int ret = 0;
490
491 mutex_lock(&inode->i_mutex);
492
493 isize = i_size_read(inode);
494 if (offset >= isize) {
495 mutex_unlock(&inode->i_mutex);
496 return -ENXIO;
497 }
498
499 blkbits = inode->i_sb->s_blocksize_bits;
500 start = offset >> blkbits;
501 last = start;
502 end = isize >> blkbits;
503 holeoff = offset;
504
505 do {
506 map.m_lblk = last;
507 map.m_len = end - last + 1;
508 ret = ext4_map_blocks(NULL, inode, &map, 0);
509 if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) {
510 last += ret;
Jan Karae7293fd2013-05-31 19:37:56 -0400511 holeoff = (loff_t)last << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500512 continue;
513 }
514
515 /*
516 * If there is a delay extent at this offset,
517 * we will skip this extent.
518 */
Yan, Zhenge30b5dc2013-05-03 02:15:52 -0400519 ext4_es_find_delayed_extent_range(inode, last, last, &es);
Zheng Liu06b0c882013-02-18 00:26:51 -0500520 if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) {
521 last = es.es_lblk + es.es_len;
Jan Karae7293fd2013-05-31 19:37:56 -0400522 holeoff = (loff_t)last << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500523 continue;
524 }
525
526 /*
527 * If there is a unwritten extent at this offset,
528 * it will be as a data or a hole according to page
529 * cache that has data or not.
530 */
531 if (map.m_flags & EXT4_MAP_UNWRITTEN) {
532 int unwritten;
533 unwritten = ext4_find_unwritten_pgoff(inode, SEEK_HOLE,
534 &map, &holeoff);
535 if (!unwritten) {
536 last += ret;
Jan Karae7293fd2013-05-31 19:37:56 -0400537 holeoff = (loff_t)last << blkbits;
Zheng Liuc8c0df22012-11-08 21:57:40 -0500538 continue;
539 }
540 }
541
542 /* find a hole */
543 break;
544 } while (last <= end);
545
546 mutex_unlock(&inode->i_mutex);
547
548 if (holeoff > isize)
549 holeoff = isize;
550
Jie Liu46a1c2c2013-06-25 12:02:13 +0800551 return vfs_setpos(file, holeoff, maxsize);
Zheng Liuc8c0df22012-11-08 21:57:40 -0500552}
553
554/*
Eric Sandeenec7268c2012-04-30 13:14:03 -0500555 * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values
556 * by calling generic_file_llseek_size() with the appropriate maxbytes
557 * value for each.
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400558 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800559loff_t ext4_llseek(struct file *file, loff_t offset, int whence)
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400560{
561 struct inode *inode = file->f_mapping->host;
562 loff_t maxbytes;
563
564 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
565 maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes;
566 else
567 maxbytes = inode->i_sb->s_maxbytes;
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400568
Andrew Morton965c8e52012-12-17 15:59:39 -0800569 switch (whence) {
Zheng Liuc8c0df22012-11-08 21:57:40 -0500570 case SEEK_SET:
571 case SEEK_CUR:
572 case SEEK_END:
Andrew Morton965c8e52012-12-17 15:59:39 -0800573 return generic_file_llseek_size(file, offset, whence,
Zheng Liuc8c0df22012-11-08 21:57:40 -0500574 maxbytes, i_size_read(inode));
575 case SEEK_DATA:
576 return ext4_seek_data(file, offset, maxbytes);
577 case SEEK_HOLE:
578 return ext4_seek_hole(file, offset, maxbytes);
579 }
580
581 return -EINVAL;
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400582}
583
Mingming Cao617ba132006-10-11 01:20:53 -0700584const struct file_operations ext4_file_operations = {
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400585 .llseek = ext4_llseek,
Al Viroaad4f8b2014-04-02 14:33:16 -0400586 .read = new_sync_read,
Al Viro9b884162014-04-17 16:09:22 -0400587 .write = new_sync_write,
Al Viroaad4f8b2014-04-02 14:33:16 -0400588 .read_iter = generic_file_read_iter,
Al Viro9b884162014-04-17 16:09:22 -0400589 .write_iter = ext4_file_write_iter,
Andi Kleen5cdd7b22008-04-29 22:03:54 -0400590 .unlocked_ioctl = ext4_ioctl,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700591#ifdef CONFIG_COMPAT
Mingming Cao617ba132006-10-11 01:20:53 -0700592 .compat_ioctl = ext4_compat_ioctl,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700593#endif
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400594 .mmap = ext4_file_mmap,
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400595 .open = ext4_file_open,
Mingming Cao617ba132006-10-11 01:20:53 -0700596 .release = ext4_release_file,
597 .fsync = ext4_sync_file,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700598 .splice_read = generic_file_splice_read,
Al Viro8d020762014-04-05 04:27:08 -0400599 .splice_write = iter_file_splice_write,
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100600 .fallocate = ext4_fallocate,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601};
602
Arjan van de Ven754661f2007-02-12 00:55:38 -0800603const struct inode_operations ext4_file_inode_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -0700604 .setattr = ext4_setattr,
Mingming Cao3e3398a2008-07-11 19:27:31 -0400605 .getattr = ext4_getattr,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700606 .setxattr = generic_setxattr,
607 .getxattr = generic_getxattr,
Mingming Cao617ba132006-10-11 01:20:53 -0700608 .listxattr = ext4_listxattr,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700609 .removexattr = generic_removexattr,
Christoph Hellwig4e34e712011-07-23 17:37:31 +0200610 .get_acl = ext4_get_acl,
Christoph Hellwig64e178a2013-12-20 05:16:44 -0800611 .set_acl = ext4_set_acl,
Eric Sandeen6873fa02008-10-07 00:46:36 -0400612 .fiemap = ext4_fiemap,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700613};
614