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_shared.h"
22 #include "xfs_format.h"
23 #include "xfs_log_format.h"
24 #include "xfs_trans_resv.h"
28 #include "xfs_mount.h"
29 #include "xfs_da_format.h"
30 #include "xfs_da_btree.h"
31 #include "xfs_inode.h"
32 #include "xfs_alloc.h"
33 #include "xfs_trans.h"
34 #include "xfs_inode_item.h"
36 #include "xfs_bmap_util.h"
38 #include "xfs_attr_leaf.h"
39 #include "xfs_attr_remote.h"
40 #include "xfs_trans_space.h"
41 #include "xfs_trace.h"
42 #include "xfs_cksum.h"
43 #include "xfs_buf_item.h"
44 #include "xfs_error.h"
46 #define ATTR_RMTVALUE_MAPSIZE 1 /* # of map entries at once */
49 * Each contiguous block has a header, so it is not just a simple attribute
50 * length to FSB conversion.
57 if (xfs_sb_version_hascrc(&mp->m_sb)) {
58 int buflen = XFS_ATTR3_RMT_BUF_SPACE(mp, mp->m_sb.sb_blocksize);
59 return (attrlen + buflen - 1) / buflen;
61 return XFS_B_TO_FSB(mp, attrlen);
65 * Checking of the remote attribute header is split into two parts. The verifier
66 * does CRC, location and bounds checking, the unpacking function checks the
67 * 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;
128 int blksize = mp->m_attr_geo->blksize;
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 >= blksize);
140 if (!xfs_verify_cksum(ptr, blksize, XFS_ATTR3_RMT_CRC_OFF)) {
141 xfs_buf_ioerror(bp, EFSBADCRC);
144 if (!xfs_attr3_rmt_verify(mp, ptr, blksize, bno)) {
145 xfs_buf_ioerror(bp, EFSCORRUPTED);
150 bno += BTOBB(blksize);
154 xfs_verifier_error(bp);
160 xfs_attr3_rmt_write_verify(
163 struct xfs_mount *mp = bp->b_target->bt_mount;
164 struct xfs_buf_log_item *bip = bp->b_fspriv;
168 int blksize = mp->m_attr_geo->blksize;
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 >= blksize);
180 if (!xfs_attr3_rmt_verify(mp, ptr, blksize, bno)) {
181 xfs_buf_ioerror(bp, EFSCORRUPTED);
182 xfs_verifier_error(bp);
186 struct xfs_attr3_rmt_hdr *rmt;
188 rmt = (struct xfs_attr3_rmt_hdr *)ptr;
189 rmt->rm_lsn = cpu_to_be64(bip->bli_item.li_lsn);
191 xfs_update_cksum(ptr, blksize, XFS_ATTR3_RMT_CRC_OFF);
195 bno += BTOBB(blksize);
200 const struct xfs_buf_ops xfs_attr3_rmt_buf_ops = {
201 .verify_read = xfs_attr3_rmt_read_verify,
202 .verify_write = xfs_attr3_rmt_write_verify,
206 xfs_attr3_rmt_hdr_set(
207 struct xfs_mount *mp,
214 struct xfs_attr3_rmt_hdr *rmt = ptr;
216 if (!xfs_sb_version_hascrc(&mp->m_sb))
219 rmt->rm_magic = cpu_to_be32(XFS_ATTR3_RMT_MAGIC);
220 rmt->rm_offset = cpu_to_be32(offset);
221 rmt->rm_bytes = cpu_to_be32(size);
222 uuid_copy(&rmt->rm_uuid, &mp->m_sb.sb_uuid);
223 rmt->rm_owner = cpu_to_be64(ino);
224 rmt->rm_blkno = cpu_to_be64(bno);
226 return sizeof(struct xfs_attr3_rmt_hdr);
230 * Helper functions to copy attribute data in and out of the one disk extents
233 xfs_attr_rmtval_copyout(
234 struct xfs_mount *mp,
241 char *src = bp->b_addr;
242 xfs_daddr_t bno = bp->b_bn;
243 int len = BBTOB(bp->b_length);
244 int blksize = mp->m_attr_geo->blksize;
246 ASSERT(len >= blksize);
248 while (len > 0 && *valuelen > 0) {
250 int byte_cnt = XFS_ATTR3_RMT_BUF_SPACE(mp, blksize);
252 byte_cnt = min(*valuelen, byte_cnt);
254 if (xfs_sb_version_hascrc(&mp->m_sb)) {
255 if (!xfs_attr3_rmt_hdr_ok(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 */
270 bno += BTOBB(blksize);
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);
292 int blksize = mp->m_attr_geo->blksize;
294 ASSERT(len >= blksize);
296 while (len > 0 && *valuelen > 0) {
298 int byte_cnt = XFS_ATTR3_RMT_BUF_SPACE(mp, blksize);
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 < blksize) {
311 ASSERT(*valuelen - byte_cnt == 0);
312 ASSERT(len == blksize);
313 memset(dst + hdr_size + byte_cnt, 0,
314 blksize - hdr_size - byte_cnt);
317 /* roll buffer forwards */
320 bno += BTOBB(blksize);
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;
345 int blkcnt = args->rmtblkcnt;
349 trace_xfs_attr_rmtval_get(args);
351 ASSERT(!(args->flags & ATTR_KERNOVAL));
352 ASSERT(args->rmtvaluelen == args->valuelen);
354 valuelen = args->rmtvaluelen;
355 while (valuelen > 0) {
356 nmap = ATTR_RMTVALUE_MAPSIZE;
357 error = xfs_bmapi_read(args->dp, (xfs_fileoff_t)lblkno,
364 for (i = 0; (i < nmap) && (valuelen > 0); i++) {
368 ASSERT((map[i].br_startblock != DELAYSTARTBLOCK) &&
369 (map[i].br_startblock != HOLESTARTBLOCK));
370 dblkno = XFS_FSB_TO_DADDR(mp, map[i].br_startblock);
371 dblkcnt = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
372 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
373 dblkno, dblkcnt, 0, &bp,
374 &xfs_attr3_rmt_buf_ops);
378 error = xfs_attr_rmtval_copyout(mp, bp, args->dp->i_ino,
385 /* roll attribute extent map forwards */
386 lblkno += map[i].br_blockcount;
387 blkcnt -= map[i].br_blockcount;
390 ASSERT(valuelen == 0);
395 * Write the value associated with an attribute into the out-of-line buffer
396 * that we have defined for it.
400 struct xfs_da_args *args)
402 struct xfs_inode *dp = args->dp;
403 struct xfs_mount *mp = dp->i_mount;
404 struct xfs_bmbt_irec map;
406 xfs_fileoff_t lfileoff = 0;
407 __uint8_t *src = args->value;
414 trace_xfs_attr_rmtval_set(args);
417 * Find a "hole" in the attribute address space large enough for
418 * us to drop the new attribute's value into. Because CRC enable
419 * attributes have headers, we can't just do a straight byte to FSB
420 * conversion and have to take the header space into account.
422 blkcnt = xfs_attr3_rmt_blocks(mp, args->rmtvaluelen);
423 error = xfs_bmap_first_unused(args->trans, args->dp, blkcnt, &lfileoff,
428 args->rmtblkno = lblkno = (xfs_dablk_t)lfileoff;
429 args->rmtblkcnt = blkcnt;
432 * Roll through the "value", allocating blocks on disk as required.
438 * Allocate a single extent, up to the size of the value.
440 xfs_bmap_init(args->flist, args->firstblock);
442 error = xfs_bmapi_write(args->trans, dp, (xfs_fileoff_t)lblkno,
444 XFS_BMAPI_ATTRFORK | XFS_BMAPI_METADATA,
445 args->firstblock, args->total, &map, &nmap,
448 error = xfs_bmap_finish(&args->trans, args->flist,
454 xfs_bmap_cancel(args->flist);
459 * bmap_finish() may have committed the last trans and started
460 * a new one. We need the inode to be in all transactions.
463 xfs_trans_ijoin(args->trans, dp, 0);
466 ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
467 (map.br_startblock != HOLESTARTBLOCK));
468 lblkno += map.br_blockcount;
469 blkcnt -= map.br_blockcount;
472 * Start the next trans in the chain.
474 error = xfs_trans_roll(&args->trans, dp);
480 * Roll through the "value", copying the attribute value to the
481 * already-allocated blocks. Blocks are written synchronously
482 * so that we can know they are all on disk before we turn off
483 * the INCOMPLETE flag.
485 lblkno = args->rmtblkno;
486 blkcnt = args->rmtblkcnt;
487 valuelen = args->rmtvaluelen;
488 while (valuelen > 0) {
495 xfs_bmap_init(args->flist, args->firstblock);
497 error = xfs_bmapi_read(dp, (xfs_fileoff_t)lblkno,
503 ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
504 (map.br_startblock != HOLESTARTBLOCK));
506 dblkno = XFS_FSB_TO_DADDR(mp, map.br_startblock),
507 dblkcnt = XFS_FSB_TO_BB(mp, map.br_blockcount);
509 bp = xfs_buf_get(mp->m_ddev_targp, dblkno, dblkcnt, 0);
512 bp->b_ops = &xfs_attr3_rmt_buf_ops;
514 xfs_attr_rmtval_copyin(mp, bp, args->dp->i_ino, &offset,
517 error = xfs_bwrite(bp); /* GROT: NOTE: synchronous write */
523 /* roll attribute extent map forwards */
524 lblkno += map.br_blockcount;
525 blkcnt -= map.br_blockcount;
527 ASSERT(valuelen == 0);
532 * Remove the value associated with an attribute by deleting the
533 * out-of-line buffer that it is stored on.
536 xfs_attr_rmtval_remove(
537 struct xfs_da_args *args)
539 struct xfs_mount *mp = args->dp->i_mount;
545 trace_xfs_attr_rmtval_remove(args);
548 * Roll through the "value", invalidating the attribute value's blocks.
550 lblkno = args->rmtblkno;
551 blkcnt = args->rmtblkcnt;
553 struct xfs_bmbt_irec map;
560 * Try to remember where we decided to put the value.
563 error = xfs_bmapi_read(args->dp, (xfs_fileoff_t)lblkno,
564 blkcnt, &map, &nmap, XFS_BMAPI_ATTRFORK);
568 ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
569 (map.br_startblock != HOLESTARTBLOCK));
571 dblkno = XFS_FSB_TO_DADDR(mp, map.br_startblock),
572 dblkcnt = XFS_FSB_TO_BB(mp, map.br_blockcount);
575 * If the "remote" value is in the cache, remove it.
577 bp = xfs_incore(mp->m_ddev_targp, dblkno, dblkcnt, XBF_TRYLOCK);
584 lblkno += map.br_blockcount;
585 blkcnt -= map.br_blockcount;
589 * Keep de-allocating extents until the remote-value region is gone.
591 lblkno = args->rmtblkno;
592 blkcnt = args->rmtblkcnt;
597 xfs_bmap_init(args->flist, args->firstblock);
598 error = xfs_bunmapi(args->trans, args->dp, lblkno, blkcnt,
599 XFS_BMAPI_ATTRFORK | XFS_BMAPI_METADATA,
600 1, args->firstblock, args->flist,
603 error = xfs_bmap_finish(&args->trans, args->flist,
609 xfs_bmap_cancel(args->flist);
614 * bmap_finish() may have committed the last trans and started
615 * a new one. We need the inode to be in all transactions.
618 xfs_trans_ijoin(args->trans, args->dp, 0);
621 * Close out trans and start the next one in the chain.
623 error = xfs_trans_roll(&args->trans, args->dp);