2 * High-level sync()-related operations
5 #include <linux/kernel.h>
6 #include <linux/file.h>
8 #include <linux/slab.h>
9 #include <linux/export.h>
10 #include <linux/namei.h>
11 #include <linux/sched.h>
12 #include <linux/writeback.h>
13 #include <linux/syscalls.h>
14 #include <linux/linkage.h>
15 #include <linux/pagemap.h>
16 #include <linux/quotaops.h>
17 #include <linux/backing-dev.h>
20 #define VALID_FLAGS (SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE| \
21 SYNC_FILE_RANGE_WAIT_AFTER)
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.
30 static int __sync_filesystem(struct super_block *sb, int wait)
32 if (sb->s_qcop && sb->s_qcop->quota_sync)
33 sb->s_qcop->quota_sync(sb, -1, wait);
38 writeback_inodes_sb(sb, WB_REASON_SYNC);
40 if (sb->s_op->sync_fs)
41 sb->s_op->sync_fs(sb, wait);
42 return __sync_blockdev(sb->s_bdev, wait);
46 * Write out and wait upon all dirty data associated with this
47 * superblock. Filesystem data as well as the underlying block
48 * device. Takes the superblock lock.
50 int sync_filesystem(struct super_block *sb)
55 * We need to be protected against the filesystem going from
56 * r/o to r/w or vice versa.
58 WARN_ON(!rwsem_is_locked(&sb->s_umount));
61 * No point in syncing out anything if the filesystem is read-only.
63 if (sb->s_flags & MS_RDONLY)
66 ret = __sync_filesystem(sb, 0);
69 return __sync_filesystem(sb, 1);
71 EXPORT_SYMBOL_GPL(sync_filesystem);
73 static void sync_one_sb(struct super_block *sb, void *arg)
75 if (!(sb->s_flags & MS_RDONLY))
76 __sync_filesystem(sb, *(int *)arg);
79 * Sync all the data for all the filesystems (called by sys_sync() and
82 static void sync_filesystems(int wait)
84 iterate_supers(sync_one_sb, &wait);
88 * sync everything. Start out by waking pdflush, because that writes back
89 * all queues in parallel.
93 wakeup_flusher_threads(0, WB_REASON_SYNC);
96 if (unlikely(laptop_mode))
97 laptop_sync_completion();
101 static void do_sync_work(struct work_struct *work)
104 * Sync twice to reduce the possibility we skipped some inodes / pages
105 * because they were temporarily locked
109 printk("Emergency Sync complete\n");
113 void emergency_sync(void)
115 struct work_struct *work;
117 work = kmalloc(sizeof(*work), GFP_ATOMIC);
119 INIT_WORK(work, do_sync_work);
125 * sync a single super
127 SYSCALL_DEFINE1(syncfs, int, fd)
130 struct super_block *sb;
134 file = fget_light(fd, &fput_needed);
137 sb = file->f_dentry->d_sb;
139 down_read(&sb->s_umount);
140 ret = sync_filesystem(sb);
141 up_read(&sb->s_umount);
143 fput_light(file, fput_needed);
148 * vfs_fsync_range - helper to sync a range of data & metadata to disk
149 * @file: file to sync
150 * @start: offset in bytes of the beginning of data range to sync
151 * @end: offset in bytes of the end of data range (inclusive)
152 * @datasync: perform only datasync
154 * Write back data in range @start..@end and metadata for @file to disk. If
155 * @datasync is set only metadata needed to access modified file data is
158 int vfs_fsync_range(struct file *file, loff_t start, loff_t end, int datasync)
160 if (!file->f_op || !file->f_op->fsync)
162 return file->f_op->fsync(file, start, end, datasync);
164 EXPORT_SYMBOL(vfs_fsync_range);
167 * vfs_fsync - perform a fsync or fdatasync on a file
168 * @file: file to sync
169 * @datasync: only perform a fdatasync operation
171 * Write back data and metadata for @file to disk. If @datasync is
172 * set only metadata needed to access modified file data is written.
174 int vfs_fsync(struct file *file, int datasync)
176 return vfs_fsync_range(file, 0, LLONG_MAX, datasync);
178 EXPORT_SYMBOL(vfs_fsync);
180 static int do_fsync(unsigned int fd, int datasync)
186 file = fget_light(fd, &fput_needed);
188 ret = vfs_fsync(file, datasync);
189 fput_light(file, fput_needed);
194 SYSCALL_DEFINE1(fsync, unsigned int, fd)
196 return do_fsync(fd, 0);
199 SYSCALL_DEFINE1(fdatasync, unsigned int, fd)
201 return do_fsync(fd, 1);
205 * generic_write_sync - perform syncing after a write if file / inode is sync
206 * @file: file to which the write happened
207 * @pos: offset where the write started
208 * @count: length of the write
210 * This is just a simple wrapper about our general syncing function.
212 int generic_write_sync(struct file *file, loff_t pos, loff_t count)
214 if (!(file->f_flags & O_DSYNC) && !IS_SYNC(file->f_mapping->host))
216 return vfs_fsync_range(file, pos, pos + count - 1,
217 (file->f_flags & __O_SYNC) ? 0 : 1);
219 EXPORT_SYMBOL(generic_write_sync);
222 * sys_sync_file_range() permits finely controlled syncing over a segment of
223 * a file in the range offset .. (offset+nbytes-1) inclusive. If nbytes is
224 * zero then sys_sync_file_range() will operate from offset out to EOF.
228 * SYNC_FILE_RANGE_WAIT_BEFORE: wait upon writeout of all pages in the range
229 * before performing the write.
231 * SYNC_FILE_RANGE_WRITE: initiate writeout of all those dirty pages in the
232 * range which are not presently under writeback. Note that this may block for
233 * significant periods due to exhaustion of disk request structures.
235 * SYNC_FILE_RANGE_WAIT_AFTER: wait upon writeout of all pages in the range
236 * after performing the write.
238 * Useful combinations of the flag bits are:
240 * SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE: ensures that all pages
241 * in the range which were dirty on entry to sys_sync_file_range() are placed
242 * under writeout. This is a start-write-for-data-integrity operation.
244 * SYNC_FILE_RANGE_WRITE: start writeout of all dirty pages in the range which
245 * are not presently under writeout. This is an asynchronous flush-to-disk
246 * operation. Not suitable for data integrity operations.
248 * SYNC_FILE_RANGE_WAIT_BEFORE (or SYNC_FILE_RANGE_WAIT_AFTER): wait for
249 * completion of writeout of all pages in the range. This will be used after an
250 * earlier SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE operation to wait
251 * for that operation to complete and to return the result.
253 * SYNC_FILE_RANGE_WAIT_BEFORE|SYNC_FILE_RANGE_WRITE|SYNC_FILE_RANGE_WAIT_AFTER:
254 * a traditional sync() operation. This is a write-for-data-integrity operation
255 * which will ensure that all pages in the range which were dirty on entry to
256 * sys_sync_file_range() are committed to disk.
259 * SYNC_FILE_RANGE_WAIT_BEFORE and SYNC_FILE_RANGE_WAIT_AFTER will detect any
260 * I/O errors or ENOSPC conditions and will return those to the caller, after
261 * clearing the EIO and ENOSPC flags in the address_space.
263 * It should be noted that none of these operations write out the file's
264 * metadata. So unless the application is strictly performing overwrites of
265 * already-instantiated disk blocks, there are no guarantees here that the data
266 * will be available after a crash.
268 SYSCALL_DEFINE(sync_file_range)(int fd, loff_t offset, loff_t nbytes,
273 struct address_space *mapping;
274 loff_t endbyte; /* inclusive */
279 if (flags & ~VALID_FLAGS)
282 endbyte = offset + nbytes;
286 if ((s64)endbyte < 0)
288 if (endbyte < offset)
291 if (sizeof(pgoff_t) == 4) {
292 if (offset >= (0x100000000ULL << PAGE_CACHE_SHIFT)) {
294 * The range starts outside a 32 bit machine's
295 * pagecache addressing capabilities. Let it "succeed"
300 if (endbyte >= (0x100000000ULL << PAGE_CACHE_SHIFT)) {
311 endbyte--; /* inclusive */
314 file = fget_light(fd, &fput_needed);
318 i_mode = file->f_path.dentry->d_inode->i_mode;
320 if (!S_ISREG(i_mode) && !S_ISBLK(i_mode) && !S_ISDIR(i_mode) &&
324 mapping = file->f_mapping;
331 if (flags & SYNC_FILE_RANGE_WAIT_BEFORE) {
332 ret = filemap_fdatawait_range(mapping, offset, endbyte);
337 if (flags & SYNC_FILE_RANGE_WRITE) {
338 ret = filemap_fdatawrite_range(mapping, offset, endbyte);
343 if (flags & SYNC_FILE_RANGE_WAIT_AFTER)
344 ret = filemap_fdatawait_range(mapping, offset, endbyte);
347 fput_light(file, fput_needed);
351 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
352 asmlinkage long SyS_sync_file_range(long fd, loff_t offset, loff_t nbytes,
355 return SYSC_sync_file_range((int) fd, offset, nbytes,
356 (unsigned int) flags);
358 SYSCALL_ALIAS(sys_sync_file_range, SyS_sync_file_range);
361 /* It would be nice if people remember that not all the world's an i386
362 when they introduce new system calls */
363 SYSCALL_DEFINE(sync_file_range2)(int fd, unsigned int flags,
364 loff_t offset, loff_t nbytes)
366 return sys_sync_file_range(fd, offset, nbytes, flags);
368 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
369 asmlinkage long SyS_sync_file_range2(long fd, long flags,
370 loff_t offset, loff_t nbytes)
372 return SYSC_sync_file_range2((int) fd, (unsigned int) flags,
375 SYSCALL_ALIAS(sys_sync_file_range2, SyS_sync_file_range2);