2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * Copyright (c) 2013 Red Hat, Inc.
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as
8 * published by the Free Software Foundation.
10 * This program is distributed in the hope that it would be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
27 #include "xfs_mount.h"
28 #include "xfs_da_format.h"
29 #include "xfs_da_btree.h"
30 #include "xfs_inode.h"
31 #include "xfs_alloc.h"
32 #include "xfs_trans.h"
33 #include "xfs_inode_item.h"
35 #include "xfs_bmap_util.h"
37 #include "xfs_attr_leaf.h"
38 #include "xfs_attr_remote.h"
39 #include "xfs_trans_space.h"
40 #include "xfs_trace.h"
41 #include "xfs_cksum.h"
42 #include "xfs_buf_item.h"
43 #include "xfs_error.h"
45 #define ATTR_RMTVALUE_MAPSIZE 1 /* # of map entries at once */
48 * Each contiguous block has a header, so it is not just a simple attribute
49 * length to FSB conversion.
56 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
57 int buflen
= XFS_ATTR3_RMT_BUF_SPACE(mp
, mp
->m_sb
.sb_blocksize
);
58 return (attrlen
+ buflen
- 1) / buflen
;
60 return XFS_B_TO_FSB(mp
, attrlen
);
64 * Checking of the remote attribute header is split into two parts. The verifier
65 * does CRC, location and bounds checking, the unpacking function checks the
66 * attribute parameters and owner.
77 struct xfs_attr3_rmt_hdr
*rmt
= ptr
;
79 if (bno
!= be64_to_cpu(rmt
->rm_blkno
))
81 if (offset
!= be32_to_cpu(rmt
->rm_offset
))
83 if (size
!= be32_to_cpu(rmt
->rm_bytes
))
85 if (ino
!= be64_to_cpu(rmt
->rm_owner
))
99 struct xfs_attr3_rmt_hdr
*rmt
= ptr
;
101 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
103 if (rmt
->rm_magic
!= cpu_to_be32(XFS_ATTR3_RMT_MAGIC
))
105 if (!uuid_equal(&rmt
->rm_uuid
, &mp
->m_sb
.sb_uuid
))
107 if (be64_to_cpu(rmt
->rm_blkno
) != bno
)
109 if (be32_to_cpu(rmt
->rm_bytes
) > fsbsize
- sizeof(*rmt
))
111 if (be32_to_cpu(rmt
->rm_offset
) +
112 be32_to_cpu(rmt
->rm_bytes
) >= XATTR_SIZE_MAX
)
114 if (rmt
->rm_owner
== 0)
121 xfs_attr3_rmt_read_verify(
124 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
127 bool corrupt
= false;
130 /* no verification of non-crc buffers */
131 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
136 len
= BBTOB(bp
->b_length
);
137 ASSERT(len
>= XFS_LBSIZE(mp
));
140 if (!xfs_verify_cksum(ptr
, XFS_LBSIZE(mp
),
141 XFS_ATTR3_RMT_CRC_OFF
)) {
145 if (!xfs_attr3_rmt_verify(mp
, ptr
, XFS_LBSIZE(mp
), bno
)) {
149 len
-= XFS_LBSIZE(mp
);
150 ptr
+= XFS_LBSIZE(mp
);
155 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
, mp
, bp
->b_addr
);
156 xfs_buf_ioerror(bp
, EFSCORRUPTED
);
162 xfs_attr3_rmt_write_verify(
165 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
166 struct xfs_buf_log_item
*bip
= bp
->b_fspriv
;
171 /* no verification of non-crc buffers */
172 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
177 len
= BBTOB(bp
->b_length
);
178 ASSERT(len
>= XFS_LBSIZE(mp
));
181 if (!xfs_attr3_rmt_verify(mp
, ptr
, XFS_LBSIZE(mp
), bno
)) {
182 XFS_CORRUPTION_ERROR(__func__
,
183 XFS_ERRLEVEL_LOW
, mp
, bp
->b_addr
);
184 xfs_buf_ioerror(bp
, EFSCORRUPTED
);
188 struct xfs_attr3_rmt_hdr
*rmt
;
190 rmt
= (struct xfs_attr3_rmt_hdr
*)ptr
;
191 rmt
->rm_lsn
= cpu_to_be64(bip
->bli_item
.li_lsn
);
193 xfs_update_cksum(ptr
, XFS_LBSIZE(mp
), XFS_ATTR3_RMT_CRC_OFF
);
195 len
-= XFS_LBSIZE(mp
);
196 ptr
+= XFS_LBSIZE(mp
);
202 const struct xfs_buf_ops xfs_attr3_rmt_buf_ops
= {
203 .verify_read
= xfs_attr3_rmt_read_verify
,
204 .verify_write
= xfs_attr3_rmt_write_verify
,
208 xfs_attr3_rmt_hdr_set(
209 struct xfs_mount
*mp
,
216 struct xfs_attr3_rmt_hdr
*rmt
= ptr
;
218 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
221 rmt
->rm_magic
= cpu_to_be32(XFS_ATTR3_RMT_MAGIC
);
222 rmt
->rm_offset
= cpu_to_be32(offset
);
223 rmt
->rm_bytes
= cpu_to_be32(size
);
224 uuid_copy(&rmt
->rm_uuid
, &mp
->m_sb
.sb_uuid
);
225 rmt
->rm_owner
= cpu_to_be64(ino
);
226 rmt
->rm_blkno
= cpu_to_be64(bno
);
228 return sizeof(struct xfs_attr3_rmt_hdr
);
232 * Helper functions to copy attribute data in and out of the one disk extents
235 xfs_attr_rmtval_copyout(
236 struct xfs_mount
*mp
,
243 char *src
= bp
->b_addr
;
244 xfs_daddr_t bno
= bp
->b_bn
;
245 int len
= BBTOB(bp
->b_length
);
247 ASSERT(len
>= XFS_LBSIZE(mp
));
249 while (len
> 0 && *valuelen
> 0) {
251 int byte_cnt
= XFS_ATTR3_RMT_BUF_SPACE(mp
, XFS_LBSIZE(mp
));
253 byte_cnt
= min(*valuelen
, byte_cnt
);
255 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
256 if (!xfs_attr3_rmt_hdr_ok(mp
, src
, ino
, *offset
,
259 "remote attribute header mismatch bno/off/len/owner (0x%llx/0x%x/Ox%x/0x%llx)",
260 bno
, *offset
, byte_cnt
, ino
);
263 hdr_size
= sizeof(struct xfs_attr3_rmt_hdr
);
266 memcpy(*dst
, src
+ hdr_size
, byte_cnt
);
268 /* roll buffer forwards */
269 len
-= XFS_LBSIZE(mp
);
270 src
+= XFS_LBSIZE(mp
);
273 /* roll attribute data forwards */
274 *valuelen
-= byte_cnt
;
282 xfs_attr_rmtval_copyin(
283 struct xfs_mount
*mp
,
290 char *dst
= bp
->b_addr
;
291 xfs_daddr_t bno
= bp
->b_bn
;
292 int len
= BBTOB(bp
->b_length
);
294 ASSERT(len
>= XFS_LBSIZE(mp
));
296 while (len
> 0 && *valuelen
> 0) {
298 int byte_cnt
= XFS_ATTR3_RMT_BUF_SPACE(mp
, XFS_LBSIZE(mp
));
300 byte_cnt
= min(*valuelen
, byte_cnt
);
301 hdr_size
= xfs_attr3_rmt_hdr_set(mp
, dst
, ino
, *offset
,
304 memcpy(dst
+ hdr_size
, *src
, byte_cnt
);
307 * If this is the last block, zero the remainder of it.
308 * Check that we are actually the last block, too.
310 if (byte_cnt
+ hdr_size
< XFS_LBSIZE(mp
)) {
311 ASSERT(*valuelen
- byte_cnt
== 0);
312 ASSERT(len
== XFS_LBSIZE(mp
));
313 memset(dst
+ hdr_size
+ byte_cnt
, 0,
314 XFS_LBSIZE(mp
) - hdr_size
- byte_cnt
);
317 /* roll buffer forwards */
318 len
-= XFS_LBSIZE(mp
);
319 dst
+= XFS_LBSIZE(mp
);
322 /* roll attribute data forwards */
323 *valuelen
-= byte_cnt
;
330 * Read the value associated with an attribute from the out-of-line buffer
331 * that we stored it in.
335 struct xfs_da_args
*args
)
337 struct xfs_bmbt_irec map
[ATTR_RMTVALUE_MAPSIZE
];
338 struct xfs_mount
*mp
= args
->dp
->i_mount
;
340 xfs_dablk_t lblkno
= args
->rmtblkno
;
341 __uint8_t
*dst
= args
->value
;
342 int valuelen
= args
->valuelen
;
345 int blkcnt
= args
->rmtblkcnt
;
349 trace_xfs_attr_rmtval_get(args
);
351 ASSERT(!(args
->flags
& ATTR_KERNOVAL
));
353 while (valuelen
> 0) {
354 nmap
= ATTR_RMTVALUE_MAPSIZE
;
355 error
= xfs_bmapi_read(args
->dp
, (xfs_fileoff_t
)lblkno
,
362 for (i
= 0; (i
< nmap
) && (valuelen
> 0); i
++) {
366 ASSERT((map
[i
].br_startblock
!= DELAYSTARTBLOCK
) &&
367 (map
[i
].br_startblock
!= HOLESTARTBLOCK
));
368 dblkno
= XFS_FSB_TO_DADDR(mp
, map
[i
].br_startblock
);
369 dblkcnt
= XFS_FSB_TO_BB(mp
, map
[i
].br_blockcount
);
370 error
= xfs_trans_read_buf(mp
, NULL
, mp
->m_ddev_targp
,
371 dblkno
, dblkcnt
, 0, &bp
,
372 &xfs_attr3_rmt_buf_ops
);
376 error
= xfs_attr_rmtval_copyout(mp
, bp
, args
->dp
->i_ino
,
383 /* roll attribute extent map forwards */
384 lblkno
+= map
[i
].br_blockcount
;
385 blkcnt
-= map
[i
].br_blockcount
;
388 ASSERT(valuelen
== 0);
393 * Write the value associated with an attribute into the out-of-line buffer
394 * that we have defined for it.
398 struct xfs_da_args
*args
)
400 struct xfs_inode
*dp
= args
->dp
;
401 struct xfs_mount
*mp
= dp
->i_mount
;
402 struct xfs_bmbt_irec map
;
404 xfs_fileoff_t lfileoff
= 0;
405 __uint8_t
*src
= args
->value
;
412 trace_xfs_attr_rmtval_set(args
);
415 * Find a "hole" in the attribute address space large enough for
416 * us to drop the new attribute's value into. Because CRC enable
417 * attributes have headers, we can't just do a straight byte to FSB
418 * conversion and have to take the header space into account.
420 blkcnt
= xfs_attr3_rmt_blocks(mp
, args
->valuelen
);
421 error
= xfs_bmap_first_unused(args
->trans
, args
->dp
, blkcnt
, &lfileoff
,
426 args
->rmtblkno
= lblkno
= (xfs_dablk_t
)lfileoff
;
427 args
->rmtblkcnt
= blkcnt
;
430 * Roll through the "value", allocating blocks on disk as required.
436 * Allocate a single extent, up to the size of the value.
438 xfs_bmap_init(args
->flist
, args
->firstblock
);
440 error
= xfs_bmapi_write(args
->trans
, dp
, (xfs_fileoff_t
)lblkno
,
442 XFS_BMAPI_ATTRFORK
| XFS_BMAPI_METADATA
,
443 args
->firstblock
, args
->total
, &map
, &nmap
,
446 error
= xfs_bmap_finish(&args
->trans
, args
->flist
,
452 xfs_bmap_cancel(args
->flist
);
457 * bmap_finish() may have committed the last trans and started
458 * a new one. We need the inode to be in all transactions.
461 xfs_trans_ijoin(args
->trans
, dp
, 0);
464 ASSERT((map
.br_startblock
!= DELAYSTARTBLOCK
) &&
465 (map
.br_startblock
!= HOLESTARTBLOCK
));
466 lblkno
+= map
.br_blockcount
;
467 blkcnt
-= map
.br_blockcount
;
470 * Start the next trans in the chain.
472 error
= xfs_trans_roll(&args
->trans
, dp
);
478 * Roll through the "value", copying the attribute value to the
479 * already-allocated blocks. Blocks are written synchronously
480 * so that we can know they are all on disk before we turn off
481 * the INCOMPLETE flag.
483 lblkno
= args
->rmtblkno
;
484 blkcnt
= args
->rmtblkcnt
;
485 valuelen
= args
->valuelen
;
486 while (valuelen
> 0) {
493 xfs_bmap_init(args
->flist
, args
->firstblock
);
495 error
= xfs_bmapi_read(dp
, (xfs_fileoff_t
)lblkno
,
501 ASSERT((map
.br_startblock
!= DELAYSTARTBLOCK
) &&
502 (map
.br_startblock
!= HOLESTARTBLOCK
));
504 dblkno
= XFS_FSB_TO_DADDR(mp
, map
.br_startblock
),
505 dblkcnt
= XFS_FSB_TO_BB(mp
, map
.br_blockcount
);
507 bp
= xfs_buf_get(mp
->m_ddev_targp
, dblkno
, dblkcnt
, 0);
510 bp
->b_ops
= &xfs_attr3_rmt_buf_ops
;
512 xfs_attr_rmtval_copyin(mp
, bp
, args
->dp
->i_ino
, &offset
,
515 error
= xfs_bwrite(bp
); /* GROT: NOTE: synchronous write */
521 /* roll attribute extent map forwards */
522 lblkno
+= map
.br_blockcount
;
523 blkcnt
-= map
.br_blockcount
;
525 ASSERT(valuelen
== 0);
530 * Remove the value associated with an attribute by deleting the
531 * out-of-line buffer that it is stored on.
534 xfs_attr_rmtval_remove(
535 struct xfs_da_args
*args
)
537 struct xfs_mount
*mp
= args
->dp
->i_mount
;
543 trace_xfs_attr_rmtval_remove(args
);
546 * Roll through the "value", invalidating the attribute value's blocks.
548 lblkno
= args
->rmtblkno
;
549 blkcnt
= args
->rmtblkcnt
;
551 struct xfs_bmbt_irec map
;
558 * Try to remember where we decided to put the value.
561 error
= xfs_bmapi_read(args
->dp
, (xfs_fileoff_t
)lblkno
,
562 blkcnt
, &map
, &nmap
, XFS_BMAPI_ATTRFORK
);
566 ASSERT((map
.br_startblock
!= DELAYSTARTBLOCK
) &&
567 (map
.br_startblock
!= HOLESTARTBLOCK
));
569 dblkno
= XFS_FSB_TO_DADDR(mp
, map
.br_startblock
),
570 dblkcnt
= XFS_FSB_TO_BB(mp
, map
.br_blockcount
);
573 * If the "remote" value is in the cache, remove it.
575 bp
= xfs_incore(mp
->m_ddev_targp
, dblkno
, dblkcnt
, XBF_TRYLOCK
);
582 lblkno
+= map
.br_blockcount
;
583 blkcnt
-= map
.br_blockcount
;
587 * Keep de-allocating extents until the remote-value region is gone.
589 lblkno
= args
->rmtblkno
;
590 blkcnt
= args
->rmtblkcnt
;
595 xfs_bmap_init(args
->flist
, args
->firstblock
);
596 error
= xfs_bunmapi(args
->trans
, args
->dp
, lblkno
, blkcnt
,
597 XFS_BMAPI_ATTRFORK
| XFS_BMAPI_METADATA
,
598 1, args
->firstblock
, args
->flist
,
601 error
= xfs_bmap_finish(&args
->trans
, args
->flist
,
607 xfs_bmap_cancel(args
->flist
);
612 * bmap_finish() may have committed the last trans and started
613 * a new one. We need the inode to be in all transactions.
616 xfs_trans_ijoin(args
->trans
, args
->dp
, 0);
619 * Close out trans and start the next one in the chain.
621 error
= xfs_trans_roll(&args
->trans
, args
->dp
);