1 A large part of this code is from the generic VFS code in fs/ioctl.c in the
4 Index: linux-2.6.27.21-0.1/fs/ext4/ioctl.c
5 ===================================================================
6 --- linux-2.6.27.21-0.1.orig/fs/ext4/ioctl.c 2009-07-07 14:08:22.000000000 +0530
7 +++ linux-2.6.27.21-0.1/fs/ext4/ioctl.c 2009-07-07 14:38:12.000000000 +0530
14 +/* So that the fiemap access checks can't overflow on 32 bit machines. */
15 +#define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
18 + * fiemap_fill_next_extent - Fiemap helper function
19 + * @fieinfo: Fiemap context passed into ->fiemap
20 + * @logical: Extent logical start offset, in bytes
21 + * @phys: Extent physical start offset, in bytes
22 + * @len: Extent length, in bytes
23 + * @flags: FIEMAP_EXTENT flags that describe this extent
24 + * @lun: LUN on which this extent resides
26 + * Called from file system ->fiemap callback. Will populate extent
27 + * info as passed in via arguments and copy to user memory. On
28 + * success, extent count on fieinfo is incremented.
30 + * Returns 0 on success, -errno on error, 1 if this was the last
31 + * extent that will fit in user array.
33 +#define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
34 +#define SET_NO_DIRECT_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED \
36 +#define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
37 +#define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
38 +int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
39 + u64 phys, u64 len, u32 flags, dev_t dev)
41 + struct fiemap_extent extent = { 0 };
42 + struct fiemap_extent *dest = fieinfo->fi_extents_start;
44 + /* only count the extents */
45 + if (fieinfo->fi_extents_max == 0) {
46 + fieinfo->fi_extents_mapped++;
47 + return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
50 + if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
53 + if (flags & SET_UNKNOWN_FLAGS)
54 + flags |= FIEMAP_EXTENT_UNKNOWN;
55 + if (flags & SET_NO_DIRECT_FLAGS)
56 + flags |= FIEMAP_EXTENT_NO_DIRECT;
57 + if (flags & SET_NOT_ALIGNED_FLAGS)
58 + flags |= FIEMAP_EXTENT_NOT_ALIGNED;
59 + if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
60 + flags |= FIEMAP_EXTENT_ENCODED;
62 + extent.fe_logical = logical;
63 + extent.fe_physical = phys;
64 + extent.fe_length = len;
65 + extent.fe_flags = flags;
66 + extent.fe_device = new_encode_dev(dev);
68 + dest += fieinfo->fi_extents_mapped;
69 + if (copy_to_user(dest, &extent, sizeof(extent)))
72 + fieinfo->fi_extents_mapped++;
73 + if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
76 + return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
79 +static int fiemap_check_ranges(struct super_block *sb,
80 + u64 start, u64 len, u64 *new_len)
87 + if (start > sb->s_maxbytes)
91 + * Shrink request scope to what the fs can actually handle.
93 + if ((len > sb->s_maxbytes) ||
94 + (sb->s_maxbytes - len) < start)
95 + *new_len = sb->s_maxbytes - start;
101 + * fiemap_check_flags - check validity of requested flags for fiemap
102 + * @fieinfo: Fiemap context passed into ->fiemap
103 + * @fs_flags: Set of fiemap flags that the file system understands
105 + * Called from file system ->fiemap callback. This will compute the
106 + * intersection of valid fiemap flags and those that the fs supports. That
107 + * value is then compared against the user supplied flags. In case of bad user
108 + * flags, the invalid values will be written into the fieinfo structure, and
109 + * -EBADR is returned, which tells ioctl_fiemap() to return those values to
110 + * userspace. For this reason, a return code of -EBADR should be preserved.
112 + * Returns 0 on success, -EBADR on bad flags.
114 +int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
116 + u32 incompat_flags;
118 + incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
119 + if (incompat_flags) {
120 + fieinfo->fi_flags = incompat_flags;
127 +int ioctl_fiemap(struct inode *inode, struct file *filp, unsigned long arg)
129 + struct fiemap fiemap;
131 + struct fiemap_extent_info fieinfo = {0, };
132 + struct super_block *sb = inode->i_sb;
135 + if (copy_from_user(&fiemap, (struct fiemap __user *) arg,
136 + sizeof(struct fiemap)))
139 + if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
142 + error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
147 + fieinfo.fi_flags = fiemap.fm_flags;
148 + fieinfo.fi_extents_max = fiemap.fm_extent_count;
149 + fieinfo.fi_extents_start = (struct fiemap_extent *)(arg + sizeof(fiemap));
151 + if (fiemap.fm_extent_count != 0 &&
152 + !access_ok(VERIFY_WRITE, (void *)arg,
153 + offsetof(typeof(fiemap), fm_extents[fiemap.fm_extent_count])))
156 + if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
157 + filemap_write_and_wait(inode->i_mapping);
159 + error = ext4_fiemap(inode, &fieinfo, fiemap.fm_start, len);
160 + fiemap.fm_flags = fieinfo.fi_flags;
161 + fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
162 + if (copy_to_user((char *)arg, &fiemap, sizeof(fiemap)))
168 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
170 struct inode *inode = filp->f_dentry->d_inode;
175 + case EXT4_IOC_FIEMAP: {
176 + return ioctl_fiemap(inode, filp, arg);
182 Index: linux-2.6.27.21-0.1/fs/ext4/ext4.h
183 ===================================================================
184 --- linux-2.6.27.21-0.1.orig/fs/ext4/ext4.h 2009-07-07 14:36:58.000000000 +0530
185 +++ linux-2.6.27.21-0.1/fs/ext4/ext4.h 2009-07-07 14:46:12.000000000 +0530
187 #define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
188 #define EXT4_IOC_GROUP_ADD _IOW('f', 8, struct ext4_new_group_input)
189 #define EXT4_IOC_MIGRATE _IO('f', 9)
190 - /* note ioctl 11 reserved for filesystem-independent FIEMAP ioctl */
191 +#define EXT4_IOC_FIEMAP _IOWR('f', 11, struct fiemap)
195 * ioctl commands in 32 bit emulation
197 #define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
198 #define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
200 +/* FIEMAP flags supported by ext4 */
201 +#define EXT4_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC)
205 @@ -1138,6 +1141,9 @@
207 extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
208 extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
209 +struct fiemap_extent_info;
210 +extern int ext4_fiemap(struct inode *, struct fiemap_extent_info *, __u64,
214 extern int ext4_ext_migrate(struct inode *);
215 Index: linux-2.6.27.21-0.1/fs/ext4/ext4_extents.h
216 ===================================================================
217 --- linux-2.6.27.21-0.1.orig/fs/ext4/ext4_extents.h 2009-07-07 14:08:22.000000000 +0530
218 +++ linux-2.6.27.21-0.1/fs/ext4/ext4_extents.h 2009-07-07 14:46:11.000000000 +0530
220 #define EXT_MAX_BLOCK 0xffffffff
223 + * to be called by ext4_ext_walk_space()
224 + * negative retcode - error
225 + * positive retcode - signal for ext4_ext_walk_space(), see below
226 + * callback must return valid extent (passed or newly created)
228 +typedef int (*ext_prepare_callback)(struct inode *, struct ext4_ext_path *,
229 + struct ext4_ext_cache *,
230 + struct ext4_extent *, void *);
232 +#define HAVE_EXT_PREPARE_CB_EXTENT
234 +#define EXT_CONTINUE 0
236 +#define EXT_REPEAT 2
239 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
240 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
241 * MSB of ee_len field in the extent datastructure to signify if this
243 struct ext4_extent *);
244 extern unsigned int ext4_ext_check_overlap(struct inode *, struct ext4_extent *, struct ext4_ext_path *);
245 extern int ext4_ext_insert_extent(handle_t *, struct inode *, struct ext4_ext_path *, struct ext4_extent *);
246 +extern int ext4_ext_walk_space(struct inode *, ext4_lblk_t, ext4_lblk_t,
247 + ext_prepare_callback, void *);
248 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
249 struct ext4_ext_path *);
250 extern int ext4_ext_search_left(struct inode *, struct ext4_ext_path *,
251 Index: linux-2.6.27.21-0.1/fs/ext4/extents.c
252 ===================================================================
253 --- linux-2.6.27.21-0.1.orig/fs/ext4/extents.c 2009-07-07 14:08:22.000000000 +0530
254 +++ linux-2.6.27.21-0.1/fs/ext4/extents.c 2009-07-07 14:46:59.000000000 +0530
256 #include <asm/uaccess.h>
257 #include "ext4_jbd2.h"
258 #include "ext4_extents.h"
264 @@ -1622,6 +1622,114 @@
268 +int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block,
269 + ext4_lblk_t num, ext_prepare_callback func,
272 + struct ext4_ext_path *path = NULL;
273 + struct ext4_ext_cache cbex;
274 + struct ext4_extent *ex;
275 + ext4_lblk_t next, start = 0, end = 0;
276 + ext4_lblk_t last = block + num;
277 + int depth, exists, err = 0;
279 + BUG_ON(func == NULL);
280 + BUG_ON(inode == NULL);
282 + while (block < last && block != EXT_MAX_BLOCK) {
283 + num = last - block;
284 + /* find extent for this block */
285 + path = ext4_ext_find_extent(inode, block, path);
286 + if (IS_ERR(path)) {
287 + err = PTR_ERR(path);
292 + depth = ext_depth(inode);
293 + BUG_ON(path[depth].p_hdr == NULL);
294 + ex = path[depth].p_ext;
295 + next = ext4_ext_next_allocated_block(path);
299 + /* there is no extent yet, so try to allocate
300 + * all requested space */
303 + } else if (le32_to_cpu(ex->ee_block) > block) {
304 + /* need to allocate space before found extent */
306 + end = le32_to_cpu(ex->ee_block);
307 + if (block + num < end)
309 + } else if (block >= le32_to_cpu(ex->ee_block)
310 + + ext4_ext_get_actual_len(ex)) {
311 + /* need to allocate space after found extent */
316 + } else if (block >= le32_to_cpu(ex->ee_block)) {
318 + * some part of requested space is covered
322 + end = le32_to_cpu(ex->ee_block)
323 + + ext4_ext_get_actual_len(ex);
324 + if (block + num < end)
330 + BUG_ON(end <= start);
333 + cbex.ec_block = start;
334 + cbex.ec_len = end - start;
336 + cbex.ec_type = EXT4_EXT_CACHE_GAP;
338 + cbex.ec_block = le32_to_cpu(ex->ee_block);
339 + cbex.ec_len = ext4_ext_get_actual_len(ex);
340 + cbex.ec_start = ext_pblock(ex);
341 + cbex.ec_type = EXT4_EXT_CACHE_EXTENT;
344 + BUG_ON(cbex.ec_len == 0);
345 + err = func(inode, path, &cbex, ex, cbdata);
346 + ext4_ext_drop_refs(path);
351 + if (err == EXT_REPEAT)
353 + else if (err == EXT_BREAK) {
358 + if (ext_depth(inode) != depth) {
359 + /* depth was changed. we have to realloc path */
364 + block = cbex.ec_block + cbex.ec_len;
368 + ext4_ext_drop_refs(path);
374 +EXPORT_SYMBOL(ext4_ext_walk_space);
377 ext4_ext_put_in_cache(struct inode *inode, ext4_lblk_t block,
378 __u32 len, ext4_fsblk_t start, int type)
379 @@ -2966,3 +3074,100 @@
380 mutex_unlock(&inode->i_mutex);
381 return ret > 0 ? ret2 : ret;
385 + * Callback function called for each extent to gather FIEMAP information.
387 +int ext4_ext_fiemap_cb(struct inode *inode, struct ext4_ext_path *path,
388 + struct ext4_ext_cache *newex, struct ext4_extent *ex,
391 + struct fiemap_extent_info *fieinfo = data;
392 + unsigned char blksize_bits = inode->i_sb->s_blocksize_bits;
399 + logical = (__u64)newex->ec_block << blksize_bits;
401 + if (newex->ec_type == EXT4_EXT_CACHE_GAP) {
404 + struct buffer_head *bh = NULL;
406 + offset = logical >> PAGE_SHIFT;
407 + page = find_get_page(inode->i_mapping, offset);
408 + if (!page || !page_has_buffers(page))
409 + return EXT_CONTINUE;
411 + bh = page_buffers(page);
414 + return EXT_CONTINUE;
416 + if (buffer_delay(bh)) {
417 + flags |= FIEMAP_EXTENT_DELALLOC;
418 + page_cache_release(page);
420 + page_cache_release(page);
421 + return EXT_CONTINUE;
425 + physical = (__u64)newex->ec_start << blksize_bits;
426 + length = (__u64)newex->ec_len << blksize_bits;
428 + if (ex && ext4_ext_is_uninitialized(ex))
429 + flags |= FIEMAP_EXTENT_UNWRITTEN;
432 + * If this extent reaches EXT_MAX_BLOCK, it must be last.
434 + * Or if ext4_ext_next_allocated_block is EXT_MAX_BLOCK,
435 + * this indicates no more allocated blocks.
437 + * XXX this might miss a single-block extent at EXT_MAX_BLOCK
439 + if (ext4_ext_next_allocated_block(path) == EXT_MAX_BLOCK ||
440 + newex->ec_block + newex->ec_len - 1 == EXT_MAX_BLOCK)
441 + flags |= FIEMAP_EXTENT_LAST;
443 + error = fiemap_fill_next_extent(fieinfo, logical, physical,
444 + length, flags, inode->i_sb->s_dev);
450 + return EXT_CONTINUE;
453 +int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
454 + __u64 start, __u64 len)
456 + ext4_fsblk_t start_blk;
457 + ext4_fsblk_t len_blks;
460 + if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
461 + return -EOPNOTSUPP;
463 + if (fiemap_check_flags(fieinfo, EXT4_FIEMAP_FLAGS_COMPAT))
466 + start_blk = start >> inode->i_sb->s_blocksize_bits;
467 + len_blks = (len + inode->i_sb->s_blocksize - 1) >> inode->i_sb->s_blocksize_bits;
470 + * Walk the extent tree gathering extent information.
471 + * ext4_ext_fiemap_cb will push extents back to user.
473 + down_write(&EXT4_I(inode)->i_data_sem);
474 + error = ext4_ext_walk_space(inode, start_blk, len_blks,
475 + ext4_ext_fiemap_cb, fieinfo);
476 + up_write(&EXT4_I(inode)->i_data_sem);
480 Index: linux-2.6.27.21-0.1/fs/ext4/fiemap.h
481 ===================================================================
482 --- /dev/null 1970-01-01 00:00:00.000000000 +0000
483 +++ linux-2.6.27.21-0.1/fs/ext4/fiemap.h 2009-07-07 14:38:12.000000000 +0530
486 + * FIEMAP ioctl infrastructure.
488 + * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
490 + * Author: Kalpak Shah <kalpak.shah@sun.com>
491 + * Andreas Dilger <adilger@sun.com>
494 +#ifndef _LINUX_EXT4_FIEMAP_H
495 +#define _LINUX_EXT4_FIEMAP_H
497 +struct fiemap_extent {
498 + __u64 fe_logical; /* logical offset in bytes for the start of
499 + * the extent from the beginning of the file */
500 + __u64 fe_physical; /* physical offset in bytes for the start
501 + * of the extent from the beginning of the disk */
502 + __u64 fe_length; /* length in bytes for this extent */
503 + __u64 fe_reserved64[2];
504 + __u32 fe_flags; /* FIEMAP_EXTENT_* flags for this extent */
505 + __u32 fe_device; /* device number for this extent */
506 + __u32 fe_reserved[2];
510 + __u64 fm_start; /* logical offset (inclusive) at
511 + * which to start mapping (in) */
512 + __u64 fm_length; /* logical length of mapping which
513 + * userspace wants (in) */
514 + __u32 fm_flags; /* FIEMAP_FLAG_* flags for request (in/out) */
515 + __u32 fm_mapped_extents;/* number of extents that were mapped (out) */
516 + __u32 fm_extent_count; /* size of fm_extents array (in) */
518 + struct fiemap_extent fm_extents[0]; /* array of mapped extents (out) */
522 + * FIEMAP helper definition.
524 +struct fiemap_extent_info {
525 + unsigned int fi_flags; /* Flags as passed from user */
526 + unsigned int fi_extents_mapped; /* Number of mapped extents */
527 + unsigned int fi_extents_max; /* Size of fiemap_extent array*/
528 + struct fiemap_extent *fi_extents_start; /* Start of fiemap_extent array */
531 +int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags);
532 +int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical,
533 + u64 phys, u64 len, u32 flags, u32 lun);
535 +#define FIEMAP_MAX_OFFSET (~0ULL)
537 +#define FIEMAP_FLAG_SYNC 0x00000001 /* sync file data before map */
538 +#define FIEMAP_FLAG_XATTR 0x00000002 /* map extended attribute tree */
540 +/* ldiskfs only supports FLAG_SYNC flag currently */
541 +#define FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR)
543 +#define FIEMAP_EXTENT_LAST 0x00000001 /* Last extent in file. */
544 +#define FIEMAP_EXTENT_UNKNOWN 0x00000002 /* Data location unknown. */
545 +#define FIEMAP_EXTENT_DELALLOC 0x00000004 /* Location still pending.
546 + * Sets EXTENT_UNKNOWN. */
547 +#define FIEMAP_EXTENT_ENCODED 0x00000008 /* Data can not be read
548 + * while fs is unmounted */
549 +#define FIEMAP_EXTENT_DATA_ENCRYPTED 0x00000080 /* Data is encrypted by fs.
550 + * Sets EXTENT_NO_DIRECT. */
551 +#define FIEMAP_EXTENT_NOT_ALIGNED 0x00000100 /* Extent offsets may not be
552 + * block aligned. */
553 +#define FIEMAP_EXTENT_DATA_INLINE 0x00000200 /* Data mixed with metadata.
554 + * Sets EXTENT_NOT_ALIGNED.*/
555 +#define FIEMAP_EXTENT_DATA_TAIL 0x00000400 /* Multiple files in block.
556 + * Sets EXTENT_NOT_ALIGNED.*/
557 +#define FIEMAP_EXTENT_UNWRITTEN 0x00000800 /* Space allocated, but
558 + * no data (i.e. zero). */
559 +#define FIEMAP_EXTENT_MERGED 0x00001000 /* File does not natively
560 + * support extents. Result
561 + * merged for efficiency. */
563 +/* Lustre specific flags - use a high bit, don't conflict with upstream flag */
564 +#define FIEMAP_EXTENT_NO_DIRECT 0x40000000 /* Data mapping undefined */
565 +#define FIEMAP_EXTENT_NET 0x80000000 /* Data stored remotely.
566 + * Sets NO_DIRECT flag */
568 +#endif /* _LINUX_EXT4_FIEMAP_H */