]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blame - fs/sync.c
quota: Move quota syncing to ->sync_fs method
[mirror_ubuntu-eoan-kernel.git] / fs / sync.c
CommitLineData
f79e2abb
AM
1/*
2 * High-level sync()-related operations
3 */
4
5#include <linux/kernel.h>
6#include <linux/file.h>
7#include <linux/fs.h>
5a0e3ad6 8#include <linux/slab.h>
630d9c47 9#include <linux/export.h>
b7ed78f5 10#include <linux/namei.h>
914e2637 11#include <linux/sched.h>
f79e2abb
AM
12#include <linux/writeback.h>
13#include <linux/syscalls.h>
14#include <linux/linkage.h>
15#include <linux/pagemap.h>
cf9a2ae8 16#include <linux/quotaops.h>
5129a469 17#include <linux/backing-dev.h>
5a3e5cb8 18#include "internal.h"
f79e2abb
AM
19
20#define VALID_FLAGS (SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE| \
21 SYNC_FILE_RANGE_WAIT_AFTER)
22
c15c54f5 23/*
d8a8559c
JA
24 * Do the filesystem syncing work. For simple filesystems
25 * writeback_inodes_sb(sb) just dirties buffers with inodes so we have to
26 * submit IO for these buffers via __sync_blockdev(). This also speeds up the
27 * wait == 1 case since in that case write_inode() functions do
28 * sync_dirty_buffer() and thus effectively write one block at a time.
c15c54f5 29 */
60b0680f 30static int __sync_filesystem(struct super_block *sb, int wait)
c15c54f5 31{
5fb324ad 32 if (wait)
d8a8559c 33 sync_inodes_sb(sb);
5fb324ad 34 else
0e175a18 35 writeback_inodes_sb(sb, WB_REASON_SYNC);
5fb324ad 36
c15c54f5
JK
37 if (sb->s_op->sync_fs)
38 sb->s_op->sync_fs(sb, wait);
39 return __sync_blockdev(sb->s_bdev, wait);
40}
41
42/*
43 * Write out and wait upon all dirty data associated with this
44 * superblock. Filesystem data as well as the underlying block
45 * device. Takes the superblock lock.
46 */
60b0680f 47int sync_filesystem(struct super_block *sb)
c15c54f5
JK
48{
49 int ret;
50
5af7926f
CH
51 /*
52 * We need to be protected against the filesystem going from
53 * r/o to r/w or vice versa.
54 */
55 WARN_ON(!rwsem_is_locked(&sb->s_umount));
56
57 /*
58 * No point in syncing out anything if the filesystem is read-only.
59 */
60 if (sb->s_flags & MS_RDONLY)
61 return 0;
62
60b0680f 63 ret = __sync_filesystem(sb, 0);
c15c54f5
JK
64 if (ret < 0)
65 return ret;
60b0680f 66 return __sync_filesystem(sb, 1);
c15c54f5 67}
60b0680f 68EXPORT_SYMBOL_GPL(sync_filesystem);
c15c54f5 69
01a05b33
AV
70static void sync_one_sb(struct super_block *sb, void *arg)
71{
95f28604 72 if (!(sb->s_flags & MS_RDONLY))
01a05b33
AV
73 __sync_filesystem(sb, *(int *)arg);
74}
c15c54f5
JK
75/*
76 * Sync all the data for all the filesystems (called by sys_sync() and
77 * emergency sync)
c15c54f5
JK
78 */
79static void sync_filesystems(int wait)
80{
01a05b33 81 iterate_supers(sync_one_sb, &wait);
c15c54f5
JK
82}
83
3beab0b4
ZY
84/*
85 * sync everything. Start out by waking pdflush, because that writes back
86 * all queues in parallel.
87 */
5cee5815 88SYSCALL_DEFINE0(sync)
cf9a2ae8 89{
0e175a18 90 wakeup_flusher_threads(0, WB_REASON_SYNC);
5cee5815
JK
91 sync_filesystems(0);
92 sync_filesystems(1);
cf9a2ae8
DH
93 if (unlikely(laptop_mode))
94 laptop_sync_completion();
cf9a2ae8
DH
95 return 0;
96}
97
a2a9537a
JA
98static void do_sync_work(struct work_struct *work)
99{
5cee5815
JK
100 /*
101 * Sync twice to reduce the possibility we skipped some inodes / pages
102 * because they were temporarily locked
103 */
104 sync_filesystems(0);
105 sync_filesystems(0);
106 printk("Emergency Sync complete\n");
a2a9537a
JA
107 kfree(work);
108}
109
cf9a2ae8
DH
110void emergency_sync(void)
111{
a2a9537a
JA
112 struct work_struct *work;
113
114 work = kmalloc(sizeof(*work), GFP_ATOMIC);
115 if (work) {
116 INIT_WORK(work, do_sync_work);
117 schedule_work(work);
118 }
cf9a2ae8
DH
119}
120
b7ed78f5
SW
121/*
122 * sync a single super
123 */
124SYSCALL_DEFINE1(syncfs, int, fd)
125{
126 struct file *file;
127 struct super_block *sb;
128 int ret;
129 int fput_needed;
130
131 file = fget_light(fd, &fput_needed);
132 if (!file)
133 return -EBADF;
134 sb = file->f_dentry->d_sb;
135
136 down_read(&sb->s_umount);
137 ret = sync_filesystem(sb);
138 up_read(&sb->s_umount);
139
140 fput_light(file, fput_needed);
141 return ret;
142}
143
4c728ef5 144/**
148f948b 145 * vfs_fsync_range - helper to sync a range of data & metadata to disk
4c728ef5 146 * @file: file to sync
148f948b
JK
147 * @start: offset in bytes of the beginning of data range to sync
148 * @end: offset in bytes of the end of data range (inclusive)
149 * @datasync: perform only datasync
4c728ef5 150 *
148f948b
JK
151 * Write back data in range @start..@end and metadata for @file to disk. If
152 * @datasync is set only metadata needed to access modified file data is
153 * written.
4c728ef5 154 */
8018ab05 155int vfs_fsync_range(struct file *file, loff_t start, loff_t end, int datasync)
cf9a2ae8 156{
02c24a82
JB
157 if (!file->f_op || !file->f_op->fsync)
158 return -EINVAL;
159 return file->f_op->fsync(file, start, end, datasync);
cf9a2ae8 160}
148f948b
JK
161EXPORT_SYMBOL(vfs_fsync_range);
162
163/**
164 * vfs_fsync - perform a fsync or fdatasync on a file
165 * @file: file to sync
148f948b
JK
166 * @datasync: only perform a fdatasync operation
167 *
168 * Write back data and metadata for @file to disk. If @datasync is
169 * set only metadata needed to access modified file data is written.
148f948b 170 */
8018ab05 171int vfs_fsync(struct file *file, int datasync)
148f948b 172{
8018ab05 173 return vfs_fsync_range(file, 0, LLONG_MAX, datasync);
148f948b 174}
4c728ef5 175EXPORT_SYMBOL(vfs_fsync);
cf9a2ae8 176
4c728ef5 177static int do_fsync(unsigned int fd, int datasync)
cf9a2ae8
DH
178{
179 struct file *file;
180 int ret = -EBADF;
c2bd6c11 181 int fput_needed;
cf9a2ae8 182
c2bd6c11 183 file = fget_light(fd, &fput_needed);
cf9a2ae8 184 if (file) {
8018ab05 185 ret = vfs_fsync(file, datasync);
c2bd6c11 186 fput_light(file, fput_needed);
cf9a2ae8
DH
187 }
188 return ret;
189}
190
a5f8fa9e 191SYSCALL_DEFINE1(fsync, unsigned int, fd)
cf9a2ae8 192{
4c728ef5 193 return do_fsync(fd, 0);
cf9a2ae8
DH
194}
195
a5f8fa9e 196SYSCALL_DEFINE1(fdatasync, unsigned int, fd)
cf9a2ae8 197{
4c728ef5 198 return do_fsync(fd, 1);
cf9a2ae8
DH
199}
200
148f948b
JK
201/**
202 * generic_write_sync - perform syncing after a write if file / inode is sync
203 * @file: file to which the write happened
204 * @pos: offset where the write started
205 * @count: length of the write
206 *
207 * This is just a simple wrapper about our general syncing function.
208 */
209int generic_write_sync(struct file *file, loff_t pos, loff_t count)
210{
6b2f3d1f 211 if (!(file->f_flags & O_DSYNC) && !IS_SYNC(file->f_mapping->host))
148f948b 212 return 0;
8018ab05 213 return vfs_fsync_range(file, pos, pos + count - 1,
6b2f3d1f 214 (file->f_flags & __O_SYNC) ? 0 : 1);
148f948b
JK
215}
216EXPORT_SYMBOL(generic_write_sync);
217
f79e2abb
AM
218/*
219 * sys_sync_file_range() permits finely controlled syncing over a segment of
220 * a file in the range offset .. (offset+nbytes-1) inclusive. If nbytes is
221 * zero then sys_sync_file_range() will operate from offset out to EOF.
222 *
223 * The flag bits are:
224 *
225 * SYNC_FILE_RANGE_WAIT_BEFORE: wait upon writeout of all pages in the range
226 * before performing the write.
227 *
228 * SYNC_FILE_RANGE_WRITE: initiate writeout of all those dirty pages in the
cce77081
PM
229 * range which are not presently under writeback. Note that this may block for
230 * significant periods due to exhaustion of disk request structures.
f79e2abb
AM
231 *
232 * SYNC_FILE_RANGE_WAIT_AFTER: wait upon writeout of all pages in the range
233 * after performing the write.
234 *
235 * Useful combinations of the flag bits are:
236 *
237 * SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE: ensures that all pages
238 * in the range which were dirty on entry to sys_sync_file_range() are placed
239 * under writeout. This is a start-write-for-data-integrity operation.
240 *
241 * SYNC_FILE_RANGE_WRITE: start writeout of all dirty pages in the range which
242 * are not presently under writeout. This is an asynchronous flush-to-disk
243 * operation. Not suitable for data integrity operations.
244 *
245 * SYNC_FILE_RANGE_WAIT_BEFORE (or SYNC_FILE_RANGE_WAIT_AFTER): wait for
246 * completion of writeout of all pages in the range. This will be used after an
247 * earlier SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE operation to wait
248 * for that operation to complete and to return the result.
249 *
250 * SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE|SYNC_FILE_RANGE_WAIT_AFTER:
251 * a traditional sync() operation. This is a write-for-data-integrity operation
252 * which will ensure that all pages in the range which were dirty on entry to
253 * sys_sync_file_range() are committed to disk.
254 *
255 *
256 * SYNC_FILE_RANGE_WAIT_BEFORE and SYNC_FILE_RANGE_WAIT_AFTER will detect any
257 * I/O errors or ENOSPC conditions and will return those to the caller, after
258 * clearing the EIO and ENOSPC flags in the address_space.
259 *
260 * It should be noted that none of these operations write out the file's
261 * metadata. So unless the application is strictly performing overwrites of
262 * already-instantiated disk blocks, there are no guarantees here that the data
263 * will be available after a crash.
264 */
6673e0c3
HC
265SYSCALL_DEFINE(sync_file_range)(int fd, loff_t offset, loff_t nbytes,
266 unsigned int flags)
f79e2abb
AM
267{
268 int ret;
269 struct file *file;
7a0ad10c 270 struct address_space *mapping;
f79e2abb
AM
271 loff_t endbyte; /* inclusive */
272 int fput_needed;
273 umode_t i_mode;
274
275 ret = -EINVAL;
276 if (flags & ~VALID_FLAGS)
277 goto out;
278
279 endbyte = offset + nbytes;
280
281 if ((s64)offset < 0)
282 goto out;
283 if ((s64)endbyte < 0)
284 goto out;
285 if (endbyte < offset)
286 goto out;
287
288 if (sizeof(pgoff_t) == 4) {
289 if (offset >= (0x100000000ULL << PAGE_CACHE_SHIFT)) {
290 /*
291 * The range starts outside a 32 bit machine's
292 * pagecache addressing capabilities. Let it "succeed"
293 */
294 ret = 0;
295 goto out;
296 }
297 if (endbyte >= (0x100000000ULL << PAGE_CACHE_SHIFT)) {
298 /*
299 * Out to EOF
300 */
301 nbytes = 0;
302 }
303 }
304
305 if (nbytes == 0)
111ebb6e 306 endbyte = LLONG_MAX;
f79e2abb
AM
307 else
308 endbyte--; /* inclusive */
309
310 ret = -EBADF;
311 file = fget_light(fd, &fput_needed);
312 if (!file)
313 goto out;
314
0f7fc9e4 315 i_mode = file->f_path.dentry->d_inode->i_mode;
f79e2abb
AM
316 ret = -ESPIPE;
317 if (!S_ISREG(i_mode) && !S_ISBLK(i_mode) && !S_ISDIR(i_mode) &&
318 !S_ISLNK(i_mode))
319 goto out_put;
320
7a0ad10c
CH
321 mapping = file->f_mapping;
322 if (!mapping) {
323 ret = -EINVAL;
324 goto out_put;
325 }
326
327 ret = 0;
328 if (flags & SYNC_FILE_RANGE_WAIT_BEFORE) {
329 ret = filemap_fdatawait_range(mapping, offset, endbyte);
330 if (ret < 0)
331 goto out_put;
332 }
333
334 if (flags & SYNC_FILE_RANGE_WRITE) {
335 ret = filemap_fdatawrite_range(mapping, offset, endbyte);
336 if (ret < 0)
337 goto out_put;
338 }
339
340 if (flags & SYNC_FILE_RANGE_WAIT_AFTER)
341 ret = filemap_fdatawait_range(mapping, offset, endbyte);
342
f79e2abb
AM
343out_put:
344 fput_light(file, fput_needed);
345out:
346 return ret;
347}
6673e0c3
HC
348#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
349asmlinkage long SyS_sync_file_range(long fd, loff_t offset, loff_t nbytes,
350 long flags)
351{
352 return SYSC_sync_file_range((int) fd, offset, nbytes,
353 (unsigned int) flags);
354}
355SYSCALL_ALIAS(sys_sync_file_range, SyS_sync_file_range);
356#endif
f79e2abb 357
edd5cd4a
DW
358/* It would be nice if people remember that not all the world's an i386
359 when they introduce new system calls */
6673e0c3
HC
360SYSCALL_DEFINE(sync_file_range2)(int fd, unsigned int flags,
361 loff_t offset, loff_t nbytes)
edd5cd4a
DW
362{
363 return sys_sync_file_range(fd, offset, nbytes, flags);
364}
6673e0c3
HC
365#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
366asmlinkage long SyS_sync_file_range2(long fd, long flags,
367 loff_t offset, loff_t nbytes)
368{
369 return SYSC_sync_file_range2((int) fd, (unsigned int) flags,
370 offset, nbytes);
371}
372SYSCALL_ALIAS(sys_sync_file_range2, SyS_sync_file_range2);
373#endif