1 A large part of this code is from the generic VFS code in fs/ioctl.c in the
4 Index: linux-2.6.22.14/fs/ext3/ioctl.c
5 ===================================================================
6 --- linux-2.6.22.14.orig/fs/ext3/ioctl.c
7 +++ linux-2.6.22.14/fs/ext3/ioctl.c
9 #include <linux/smp_lock.h>
10 #include <asm/uaccess.h>
11 #include <linux/namei.h>
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 = ext3_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 int ext3_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
170 @@ -274,7 +429,9 @@ flags_err:
175 + case EXT3_IOC_FIEMAP: {
176 + return ioctl_fiemap(inode, filp, arg);
181 Index: linux-2.6.22.14/include/linux/ext3_fs.h
182 ===================================================================
183 --- linux-2.6.22.14.orig/include/linux/ext3_fs.h
184 +++ linux-2.6.22.14/include/linux/ext3_fs.h
185 @@ -255,15 +255,16 @@ struct ext3_new_group_data {
186 #define EXT3_IOC_SETFLAGS FS_IOC_SETFLAGS
187 #define EXT3_IOC_GETVERSION _IOR('f', 3, long)
188 #define EXT3_IOC_SETVERSION _IOW('f', 4, long)
189 +#define EXT3_IOC_GETRSVSZ _IOR('f', 5, long)
190 +#define EXT3_IOC_SETRSVSZ _IOW('f', 6, long)
191 #define EXT3_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
192 #define EXT3_IOC_GROUP_ADD _IOW('f', 8,struct ext3_new_group_input)
193 +#define EXT3_IOC_FIEMAP _IOWR('f', 11, struct fiemap)
194 #define EXT3_IOC_GETVERSION_OLD FS_IOC_GETVERSION
195 #define EXT3_IOC_SETVERSION_OLD FS_IOC_SETVERSION
196 #ifdef CONFIG_JBD_DEBUG
197 #define EXT3_IOC_WAIT_FOR_READONLY _IOR('f', 99, long)
199 -#define EXT3_IOC_GETRSVSZ _IOR('f', 5, long)
200 -#define EXT3_IOC_SETRSVSZ _IOW('f', 6, long)
203 * ioctl commands in 32 bit emulation
204 @@ -281,6 +282,8 @@ struct ext3_new_group_data {
205 #define EXT3_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
206 #define EXT3_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
208 +/* FIEMAP flags supported by ext3 */
209 +#define EXT3_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC)
213 @@ -1057,6 +1060,9 @@ extern int ext3_block_truncate_page(hand
214 extern int ext3_ioctl (struct inode *, struct file *, unsigned int,
216 extern long ext3_compat_ioctl (struct file *, unsigned int, unsigned long);
217 +struct fiemap_extent_info;
218 +extern int ext3_fiemap(struct inode *, struct fiemap_extent_info *, __u64,
222 extern int ext3_orphan_add(handle_t *, struct inode *);
223 @@ -1134,7 +1140,6 @@ ext3_get_blocks_wrap(handle_t *handle, s
228 #endif /* __KERNEL__ */
230 /* EXT3_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */
231 Index: linux-2.6.22.14/include/linux/ext3_extents.h
232 ===================================================================
233 --- linux-2.6.22.14.orig/include/linux/ext3_extents.h
234 +++ linux-2.6.22.14/include/linux/ext3_extents.h
235 @@ -142,8 +142,10 @@ struct ext3_ext_path {
236 * callback must return valid extent (passed or newly created)
238 typedef int (*ext_prepare_callback)(struct inode *, struct ext3_ext_path *,
239 - struct ext3_ext_cache *,
241 + struct ext3_ext_cache *,
242 + struct ext3_extent *, void *);
244 +#define HAVE_EXT_PREPARE_CB_EXTENT
246 #define EXT_CONTINUE 0
248 @@ -152,6 +154,26 @@ typedef int (*ext_prepare_callback)(stru
250 #define EXT_MAX_BLOCK 0xffffffff
253 + * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
254 + * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
255 + * MSB of ee_len field in the extent datastructure to signify if this
256 + * particular extent is an initialized extent or an uninitialized (i.e.
258 + * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
259 + * uninitialized extent.
260 + * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
261 + * uninitialized one. In other words, if MSB of ee_len is set, it is an
262 + * uninitialized extent with only one special scenario when ee_len = 0x8000.
263 + * In this case we can not have an uninitialized extent of zero length and
264 + * thus we make it as a special case of initialized extent with 0x8000 length.
265 + * This way we get better extent-to-group alignment for initialized extents.
266 + * Hence, the maximum number of blocks we can have in an *initialized*
267 + * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
269 +#define EXT_INIT_MAX_LEN (1UL << 15)
270 +#define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
272 #define EXT_FLAGS_CLR_UNKNOWN 0x7 /* Flags cleared on modification */
273 #define EXT_HDR_GEN_BITS 24
274 #define EXT_HDR_GEN_MASK ((1 << EXT_HDR_GEN_BITS) - 1)
275 @@ -219,6 +241,12 @@ ext3_ext_invalidate_cache(struct inode *
276 EXT3_I(inode)->i_cached_extent.ec_type = EXT3_EXT_CACHE_NO;
279 +static inline int ext3_ext_is_uninitialized(struct ext3_extent *ext)
281 + /* Extent with ee_len of 0x8000 is treated as an initialized extent */
282 + return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
285 extern int ext3_ext_search_left(struct inode *, struct ext3_ext_path *, unsigned long *, unsigned long *);
286 extern int ext3_ext_search_right(struct inode *, struct ext3_ext_path *, unsigned long *, unsigned long *);
287 extern int ext3_extent_tree_init(handle_t *, struct inode *);
288 Index: linux-2.6.22.14/fs/ext3/extents.c
289 ===================================================================
290 --- linux-2.6.22.14.orig/fs/ext3/extents.c
291 +++ linux-2.6.22.14/fs/ext3/extents.c
293 #include <linux/string.h>
294 #include <linux/slab.h>
295 #include <linux/ext3_extents.h>
297 #include <asm/uaccess.h>
300 @@ -1481,7 +1482,7 @@ int ext3_ext_walk_space(struct inode *in
303 BUG_ON(cbex.ec_len == 0);
304 - err = func(inode, path, &cbex, cbdata);
305 + err = func(inode, path, &cbex, ex, cbdata);
306 ext3_ext_drop_refs(path);
309 @@ -2296,6 +2297,103 @@ int ext3_ext_writepage_trans_blocks(stru
314 + * Callback function called for each extent to gather FIEMAP information.
316 +int ext3_ext_fiemap_cb(struct inode *inode, struct ext3_ext_path *path,
317 + struct ext3_ext_cache *newex, struct ext3_extent *ex,
320 + struct fiemap_extent_info *fieinfo = data;
321 + unsigned long blksize_bits = inode->i_sb->s_blocksize_bits;
328 + logical = (__u64)newex->ec_block << blksize_bits;
330 + if (newex->ec_type == EXT3_EXT_CACHE_GAP) {
333 + struct buffer_head *bh = NULL;
335 + offset = logical >> PAGE_SHIFT;
336 + page = find_get_page(inode->i_mapping, offset);
337 + if (!page || !page_has_buffers(page))
338 + return EXT_CONTINUE;
340 + bh = page_buffers(page);
343 + return EXT_CONTINUE;
345 + if (buffer_delay(bh)) {
346 + flags |= FIEMAP_EXTENT_DELALLOC;
347 + page_cache_release(page);
349 + page_cache_release(page);
350 + return EXT_CONTINUE;
354 + physical = (__u64)newex->ec_start << blksize_bits;
355 + length = (__u64)newex->ec_len << blksize_bits;
357 + if (ex && ext3_ext_is_uninitialized(ex))
358 + flags |= FIEMAP_EXTENT_UNWRITTEN;
361 + * If this extent reaches EXT_MAX_BLOCK, it must be last.
363 + * Or if ext3_ext_next_allocated_block is EXT_MAX_BLOCK,
364 + * this indicates no more allocated blocks.
366 + * XXX this might miss a single-block extent at EXT_MAX_BLOCK
368 + if (logical + length - 1 == EXT_MAX_BLOCK ||
369 + ext3_ext_next_allocated_block(path) == EXT_MAX_BLOCK)
370 + flags |= FIEMAP_EXTENT_LAST;
372 + error = fiemap_fill_next_extent(fieinfo, logical, physical,
373 + length, flags, inode->i_sb->s_dev);
379 + return EXT_CONTINUE;
382 +int ext3_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
383 + __u64 start, __u64 len)
385 + ext3_fsblk_t start_blk;
386 + ext3_fsblk_t len_blks;
389 + if (!(EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL))
390 + return -EOPNOTSUPP;
392 + if (fiemap_check_flags(fieinfo, EXT3_FIEMAP_FLAGS_COMPAT))
395 + start_blk = start >> inode->i_sb->s_blocksize_bits;
396 + len_blks = (len + inode->i_sb->s_blocksize - 1) >> inode->i_sb->s_blocksize_bits;
399 + * Walk the extent tree gathering extent information.
400 + * ext3_ext_fiemap_cb will push extents back to user.
402 + mutex_lock(&EXT3_I(inode)->truncate_mutex);
403 + error = ext3_ext_walk_space(inode, start_blk, len_blks,
404 + ext3_ext_fiemap_cb, fieinfo);
405 + mutex_unlock(&EXT3_I(inode)->truncate_mutex);
410 EXPORT_SYMBOL(ext3_mark_inode_dirty);
411 EXPORT_SYMBOL(ext3_ext_invalidate_cache);
412 EXPORT_SYMBOL(ext3_ext_insert_extent);
413 Index: linux-2.6.22.14/fs/ext3/fiemap.h
414 ===================================================================
416 +++ linux-2.6.22.14/fs/ext3/fiemap.h
419 + * FIEMAP ioctl infrastructure.
421 + * Copyright 2008 Sun Microsystems, Inc.
423 + * Author: Kalpak Shah <kalpak.shah@sun.com>
424 + * Andreas Dilger <adilger@sun.com>
427 +#ifndef _LINUX_EXT3_FIEMAP_H
428 +#define _LINUX_EXT3_FIEMAP_H
430 +struct fiemap_extent {
431 + __u64 fe_logical; /* logical offset in bytes for the start of
432 + * the extent from the beginning of the file */
433 + __u64 fe_physical; /* physical offset in bytes for the start
434 + * of the extent from the beginning of the disk */
435 + __u64 fe_length; /* length in bytes for this extent */
436 + __u64 fe_reserved64[2];
437 + __u32 fe_flags; /* FIEMAP_EXTENT_* flags for this extent */
438 + __u32 fe_device; /* device number for this extent */
439 + __u32 fe_reserved[2];
443 + __u64 fm_start; /* logical offset (inclusive) at
444 + * which to start mapping (in) */
445 + __u64 fm_length; /* logical length of mapping which
446 + * userspace wants (in) */
447 + __u32 fm_flags; /* FIEMAP_FLAG_* flags for request (in/out) */
448 + __u32 fm_mapped_extents;/* number of extents that were mapped (out) */
449 + __u32 fm_extent_count; /* size of fm_extents array (in) */
451 + struct fiemap_extent fm_extents[0]; /* array of mapped extents (out) */
455 + * FIEMAP helper definition.
457 +struct fiemap_extent_info {
458 + unsigned int fi_flags; /* Flags as passed from user */
459 + unsigned int fi_extents_mapped; /* Number of mapped extents */
460 + unsigned int fi_extents_max; /* Size of fiemap_extent array*/
461 + struct fiemap_extent *fi_extents_start; /* Start of fiemap_extent array */
464 +int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags);
465 +int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical,
466 + u64 phys, u64 len, u32 flags, u32 lun);
468 +#define FIEMAP_MAX_OFFSET (~0ULL)
470 +#define FIEMAP_FLAG_SYNC 0x00000001 /* sync file data before map */
471 +#define FIEMAP_FLAG_XATTR 0x00000002 /* map extended attribute tree */
473 +/* ldiskfs only supports FLAG_SYNC flag currently */
474 +#define FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR)
476 +#define FIEMAP_EXTENT_LAST 0x00000001 /* Last extent in file. */
477 +#define FIEMAP_EXTENT_UNKNOWN 0x00000002 /* Data location unknown. */
478 +#define FIEMAP_EXTENT_DELALLOC 0x00000004 /* Location still pending.
479 + * Sets EXTENT_UNKNOWN. */
480 +#define FIEMAP_EXTENT_ENCODED 0x00000008 /* Data can not be read
481 + * while fs is unmounted */
482 +#define FIEMAP_EXTENT_DATA_ENCRYPTED 0x00000080 /* Data is encrypted by fs.
483 + * Sets EXTENT_NO_DIRECT. */
484 +#define FIEMAP_EXTENT_NOT_ALIGNED 0x00000100 /* Extent offsets may not be
485 + * block aligned. */
486 +#define FIEMAP_EXTENT_DATA_INLINE 0x00000200 /* Data mixed with metadata.
487 + * Sets EXTENT_NOT_ALIGNED.*/
488 +#define FIEMAP_EXTENT_DATA_TAIL 0x00000400 /* Multiple files in block.
489 + * Sets EXTENT_NOT_ALIGNED.*/
490 +#define FIEMAP_EXTENT_UNWRITTEN 0x00000800 /* Space allocated, but
491 + * no data (i.e. zero). */
492 +#define FIEMAP_EXTENT_MERGED 0x00001000 /* File does not natively
493 + * support extents. Result
494 + * merged for efficiency. */
496 +/* Lustre specific flags - use a high bit, don't conflict with upstream flag */
497 +#define FIEMAP_EXTENT_NO_DIRECT 0x40000000 /* Data mapping undefined */
498 +#define FIEMAP_EXTENT_NET 0x80000000 /* Data stored remotely.
499 + * Sets NO_DIRECT flag */
501 +#endif /* _LINUX_EXT3_FIEMAP_H */