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_types.h"
24 #include "xfs_trans.h"
27 #include "xfs_mount.h"
28 #include "xfs_error.h"
29 #include "xfs_da_btree.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_dinode.h"
32 #include "xfs_inode.h"
33 #include "xfs_alloc.h"
34 #include "xfs_inode_item.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"
44 #define ATTR_RMTVALUE_MAPSIZE 1 /* # of map entries at once */
47 * Each contiguous block has a header, so it is not just a simple attribute
48 * length to FSB conversion.
55 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
56 int buflen
= XFS_ATTR3_RMT_BUF_SPACE(mp
, mp
->m_sb
.sb_blocksize
);
57 return (attrlen
+ buflen
- 1) / buflen
;
59 return XFS_B_TO_FSB(mp
, attrlen
);
63 * Checking of the remote attribute header is split into two parts. The verifier
64 * does CRC, location and bounds checking, the unpacking function checks the
65 * attribute parameters and owner.
76 struct xfs_attr3_rmt_hdr
*rmt
= ptr
;
78 if (bno
!= be64_to_cpu(rmt
->rm_blkno
))
80 if (offset
!= be32_to_cpu(rmt
->rm_offset
))
82 if (size
!= be32_to_cpu(rmt
->rm_bytes
))
84 if (ino
!= be64_to_cpu(rmt
->rm_owner
))
98 struct xfs_attr3_rmt_hdr
*rmt
= ptr
;
100 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
102 if (rmt
->rm_magic
!= cpu_to_be32(XFS_ATTR3_RMT_MAGIC
))
104 if (!uuid_equal(&rmt
->rm_uuid
, &mp
->m_sb
.sb_uuid
))
106 if (be64_to_cpu(rmt
->rm_blkno
) != bno
)
108 if (be32_to_cpu(rmt
->rm_bytes
) > fsbsize
- sizeof(*rmt
))
110 if (be32_to_cpu(rmt
->rm_offset
) +
111 be32_to_cpu(rmt
->rm_bytes
) >= XATTR_SIZE_MAX
)
113 if (rmt
->rm_owner
== 0)
120 xfs_attr3_rmt_read_verify(
123 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
126 bool corrupt
= false;
129 /* no verification of non-crc buffers */
130 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
135 len
= BBTOB(bp
->b_length
);
136 ASSERT(len
>= XFS_LBSIZE(mp
));
139 if (!xfs_verify_cksum(ptr
, XFS_LBSIZE(mp
),
140 XFS_ATTR3_RMT_CRC_OFF
)) {
144 if (!xfs_attr3_rmt_verify(mp
, ptr
, XFS_LBSIZE(mp
), bno
)) {
148 len
-= XFS_LBSIZE(mp
);
149 ptr
+= XFS_LBSIZE(mp
);
154 XFS_CORRUPTION_ERROR(__func__
, XFS_ERRLEVEL_LOW
, mp
, bp
->b_addr
);
155 xfs_buf_ioerror(bp
, EFSCORRUPTED
);
161 xfs_attr3_rmt_write_verify(
164 struct xfs_mount
*mp
= bp
->b_target
->bt_mount
;
165 struct xfs_buf_log_item
*bip
= bp
->b_fspriv
;
170 /* no verification of non-crc buffers */
171 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
176 len
= BBTOB(bp
->b_length
);
177 ASSERT(len
>= XFS_LBSIZE(mp
));
180 if (!xfs_attr3_rmt_verify(mp
, ptr
, XFS_LBSIZE(mp
), bno
)) {
181 XFS_CORRUPTION_ERROR(__func__
,
182 XFS_ERRLEVEL_LOW
, mp
, bp
->b_addr
);
183 xfs_buf_ioerror(bp
, EFSCORRUPTED
);
187 struct xfs_attr3_rmt_hdr
*rmt
;
189 rmt
= (struct xfs_attr3_rmt_hdr
*)ptr
;
190 rmt
->rm_lsn
= cpu_to_be64(bip
->bli_item
.li_lsn
);
192 xfs_update_cksum(ptr
, XFS_LBSIZE(mp
), XFS_ATTR3_RMT_CRC_OFF
);
194 len
-= XFS_LBSIZE(mp
);
195 ptr
+= XFS_LBSIZE(mp
);
201 const struct xfs_buf_ops xfs_attr3_rmt_buf_ops
= {
202 .verify_read
= xfs_attr3_rmt_read_verify
,
203 .verify_write
= xfs_attr3_rmt_write_verify
,
207 xfs_attr3_rmt_hdr_set(
208 struct xfs_mount
*mp
,
215 struct xfs_attr3_rmt_hdr
*rmt
= ptr
;
217 if (!xfs_sb_version_hascrc(&mp
->m_sb
))
220 rmt
->rm_magic
= cpu_to_be32(XFS_ATTR3_RMT_MAGIC
);
221 rmt
->rm_offset
= cpu_to_be32(offset
);
222 rmt
->rm_bytes
= cpu_to_be32(size
);
223 uuid_copy(&rmt
->rm_uuid
, &mp
->m_sb
.sb_uuid
);
224 rmt
->rm_owner
= cpu_to_be64(ino
);
225 rmt
->rm_blkno
= cpu_to_be64(bno
);
227 return sizeof(struct xfs_attr3_rmt_hdr
);
231 * Helper functions to copy attribute data in and out of the one disk extents
234 xfs_attr_rmtval_copyout(
235 struct xfs_mount
*mp
,
242 char *src
= bp
->b_addr
;
243 xfs_daddr_t bno
= bp
->b_bn
;
244 int len
= BBTOB(bp
->b_length
);
246 ASSERT(len
>= XFS_LBSIZE(mp
));
248 while (len
> 0 && *valuelen
> 0) {
250 int byte_cnt
= XFS_ATTR3_RMT_BUF_SPACE(mp
, XFS_LBSIZE(mp
));
252 byte_cnt
= min_t(int, *valuelen
, byte_cnt
);
254 if (xfs_sb_version_hascrc(&mp
->m_sb
)) {
255 if (!xfs_attr3_rmt_hdr_ok(mp
, src
, ino
, *offset
,
258 "remote attribute header mismatch bno/off/len/owner (0x%llx/0x%x/Ox%x/0x%llx)",
259 bno
, *offset
, byte_cnt
, ino
);
262 hdr_size
= sizeof(struct xfs_attr3_rmt_hdr
);
265 memcpy(*dst
, src
+ hdr_size
, byte_cnt
);
267 /* roll buffer forwards */
268 len
-= XFS_LBSIZE(mp
);
269 src
+= XFS_LBSIZE(mp
);
272 /* roll attribute data forwards */
273 *valuelen
-= byte_cnt
;
281 xfs_attr_rmtval_copyin(
282 struct xfs_mount
*mp
,
289 char *dst
= bp
->b_addr
;
290 xfs_daddr_t bno
= bp
->b_bn
;
291 int len
= BBTOB(bp
->b_length
);
293 ASSERT(len
>= XFS_LBSIZE(mp
));
295 while (len
> 0 && *valuelen
> 0) {
297 int byte_cnt
= XFS_ATTR3_RMT_BUF_SPACE(mp
, XFS_LBSIZE(mp
));
299 byte_cnt
= min(*valuelen
, byte_cnt
);
300 hdr_size
= xfs_attr3_rmt_hdr_set(mp
, dst
, ino
, *offset
,
303 memcpy(dst
+ hdr_size
, *src
, byte_cnt
);
306 * If this is the last block, zero the remainder of it.
307 * Check that we are actually the last block, too.
309 if (byte_cnt
+ hdr_size
< XFS_LBSIZE(mp
)) {
310 ASSERT(*valuelen
- byte_cnt
== 0);
311 ASSERT(len
== XFS_LBSIZE(mp
));
312 memset(dst
+ hdr_size
+ byte_cnt
, 0,
313 XFS_LBSIZE(mp
) - hdr_size
- byte_cnt
);
316 /* roll buffer forwards */
317 len
-= XFS_LBSIZE(mp
);
318 dst
+= XFS_LBSIZE(mp
);
321 /* roll attribute data forwards */
322 *valuelen
-= byte_cnt
;
329 * Read the value associated with an attribute from the out-of-line buffer
330 * that we stored it in.
334 struct xfs_da_args
*args
)
336 struct xfs_bmbt_irec map
[ATTR_RMTVALUE_MAPSIZE
];
337 struct xfs_mount
*mp
= args
->dp
->i_mount
;
339 xfs_dablk_t lblkno
= args
->rmtblkno
;
340 char *dst
= args
->value
;
341 int valuelen
= args
->valuelen
;
344 int blkcnt
= args
->rmtblkcnt
;
348 trace_xfs_attr_rmtval_get(args
);
350 ASSERT(!(args
->flags
& ATTR_KERNOVAL
));
352 while (valuelen
> 0) {
353 nmap
= ATTR_RMTVALUE_MAPSIZE
;
354 error
= xfs_bmapi_read(args
->dp
, (xfs_fileoff_t
)lblkno
,
361 for (i
= 0; (i
< nmap
) && (valuelen
> 0); i
++) {
365 ASSERT((map
[i
].br_startblock
!= DELAYSTARTBLOCK
) &&
366 (map
[i
].br_startblock
!= HOLESTARTBLOCK
));
367 dblkno
= XFS_FSB_TO_DADDR(mp
, map
[i
].br_startblock
);
368 dblkcnt
= XFS_FSB_TO_BB(mp
, map
[i
].br_blockcount
);
369 error
= xfs_trans_read_buf(mp
, NULL
, mp
->m_ddev_targp
,
370 dblkno
, dblkcnt
, 0, &bp
,
371 &xfs_attr3_rmt_buf_ops
);
375 error
= xfs_attr_rmtval_copyout(mp
, bp
, args
->dp
->i_ino
,
382 /* roll attribute extent map forwards */
383 lblkno
+= map
[i
].br_blockcount
;
384 blkcnt
-= map
[i
].br_blockcount
;
387 ASSERT(valuelen
== 0);
392 * Write the value associated with an attribute into the out-of-line buffer
393 * that we have defined for it.
397 struct xfs_da_args
*args
)
399 struct xfs_inode
*dp
= args
->dp
;
400 struct xfs_mount
*mp
= dp
->i_mount
;
401 struct xfs_bmbt_irec map
;
403 xfs_fileoff_t lfileoff
= 0;
404 char *src
= args
->value
;
411 trace_xfs_attr_rmtval_set(args
);
414 * Find a "hole" in the attribute address space large enough for
415 * us to drop the new attribute's value into. Because CRC enable
416 * attributes have headers, we can't just do a straight byte to FSB
417 * conversion and have to take the header space into account.
419 blkcnt
= xfs_attr3_rmt_blocks(mp
, args
->valuelen
);
420 error
= xfs_bmap_first_unused(args
->trans
, args
->dp
, blkcnt
, &lfileoff
,
425 args
->rmtblkno
= lblkno
= (xfs_dablk_t
)lfileoff
;
426 args
->rmtblkcnt
= blkcnt
;
429 * Roll through the "value", allocating blocks on disk as required.
435 * Allocate a single extent, up to the size of the value.
437 xfs_bmap_init(args
->flist
, args
->firstblock
);
439 error
= xfs_bmapi_write(args
->trans
, dp
, (xfs_fileoff_t
)lblkno
,
441 XFS_BMAPI_ATTRFORK
| XFS_BMAPI_METADATA
,
442 args
->firstblock
, args
->total
, &map
, &nmap
,
445 error
= xfs_bmap_finish(&args
->trans
, args
->flist
,
451 xfs_bmap_cancel(args
->flist
);
456 * bmap_finish() may have committed the last trans and started
457 * a new one. We need the inode to be in all transactions.
460 xfs_trans_ijoin(args
->trans
, dp
, 0);
463 ASSERT((map
.br_startblock
!= DELAYSTARTBLOCK
) &&
464 (map
.br_startblock
!= HOLESTARTBLOCK
));
465 lblkno
+= map
.br_blockcount
;
466 blkcnt
-= map
.br_blockcount
;
469 * Start the next trans in the chain.
471 error
= xfs_trans_roll(&args
->trans
, dp
);
477 * Roll through the "value", copying the attribute value to the
478 * already-allocated blocks. Blocks are written synchronously
479 * so that we can know they are all on disk before we turn off
480 * the INCOMPLETE flag.
482 lblkno
= args
->rmtblkno
;
483 blkcnt
= args
->rmtblkcnt
;
484 valuelen
= args
->valuelen
;
485 while (valuelen
> 0) {
492 xfs_bmap_init(args
->flist
, args
->firstblock
);
494 error
= xfs_bmapi_read(dp
, (xfs_fileoff_t
)lblkno
,
500 ASSERT((map
.br_startblock
!= DELAYSTARTBLOCK
) &&
501 (map
.br_startblock
!= HOLESTARTBLOCK
));
503 dblkno
= XFS_FSB_TO_DADDR(mp
, map
.br_startblock
),
504 dblkcnt
= XFS_FSB_TO_BB(mp
, map
.br_blockcount
);
506 bp
= xfs_buf_get(mp
->m_ddev_targp
, dblkno
, dblkcnt
, 0);
509 bp
->b_ops
= &xfs_attr3_rmt_buf_ops
;
511 xfs_attr_rmtval_copyin(mp
, bp
, args
->dp
->i_ino
, &offset
,
514 error
= xfs_bwrite(bp
); /* GROT: NOTE: synchronous write */
520 /* roll attribute extent map forwards */
521 lblkno
+= map
.br_blockcount
;
522 blkcnt
-= map
.br_blockcount
;
524 ASSERT(valuelen
== 0);
529 * Remove the value associated with an attribute by deleting the
530 * out-of-line buffer that it is stored on.
533 xfs_attr_rmtval_remove(
534 struct xfs_da_args
*args
)
536 struct xfs_mount
*mp
= args
->dp
->i_mount
;
542 trace_xfs_attr_rmtval_remove(args
);
545 * Roll through the "value", invalidating the attribute value's blocks.
546 * Note that args->rmtblkcnt is the minimum number of data blocks we'll
547 * see for a CRC enabled remote attribute. Each extent will have a
548 * header, and so we may have more blocks than we realise here. If we
549 * fail to map the blocks correctly, we'll have problems with the buffer
552 lblkno
= args
->rmtblkno
;
553 blkcnt
= args
->rmtblkcnt
;
555 struct xfs_bmbt_irec map
;
562 * Try to remember where we decided to put the value.
565 error
= xfs_bmapi_read(args
->dp
, (xfs_fileoff_t
)lblkno
,
566 blkcnt
, &map
, &nmap
, XFS_BMAPI_ATTRFORK
);
570 ASSERT((map
.br_startblock
!= DELAYSTARTBLOCK
) &&
571 (map
.br_startblock
!= HOLESTARTBLOCK
));
573 dblkno
= XFS_FSB_TO_DADDR(mp
, map
.br_startblock
),
574 dblkcnt
= XFS_FSB_TO_BB(mp
, map
.br_blockcount
);
577 * If the "remote" value is in the cache, remove it.
579 bp
= xfs_incore(mp
->m_ddev_targp
, dblkno
, dblkcnt
, XBF_TRYLOCK
);
586 lblkno
+= map
.br_blockcount
;
587 blkcnt
-= map
.br_blockcount
;
591 * Keep de-allocating extents until the remote-value region is gone.
593 lblkno
= args
->rmtblkno
;
594 blkcnt
= args
->rmtblkcnt
;
599 xfs_bmap_init(args
->flist
, args
->firstblock
);
600 error
= xfs_bunmapi(args
->trans
, args
->dp
, lblkno
, blkcnt
,
601 XFS_BMAPI_ATTRFORK
| XFS_BMAPI_METADATA
,
602 1, args
->firstblock
, args
->flist
,
605 error
= xfs_bmap_finish(&args
->trans
, args
->flist
,
611 xfs_bmap_cancel(args
->flist
);
616 * bmap_finish() may have committed the last trans and started
617 * a new one. We need the inode to be in all transactions.
620 xfs_trans_ijoin(args
->trans
, args
->dp
, 0);
623 * Close out trans and start the next one in the chain.
625 error
= xfs_trans_roll(&args
->trans
, args
->dp
);