2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
26 #include "xfs_mount.h"
27 #include "xfs_inode.h"
28 #include "xfs_btree.h"
29 #include "xfs_bmap_btree.h"
31 #include "xfs_bmap_util.h"
32 #include "xfs_error.h"
33 #include "xfs_trans.h"
34 #include "xfs_trans_space.h"
35 #include "xfs_iomap.h"
36 #include "xfs_trace.h"
37 #include "xfs_icache.h"
38 #include "xfs_quota.h"
39 #include "xfs_dquot_item.h"
40 #include "xfs_dquot.h"
41 #include "xfs_dinode.h"
44 #define XFS_WRITEIO_ALIGN(mp,off) (((off) >> mp->m_writeio_log) \
46 #define XFS_WRITE_IMAPS XFS_BMAP_MAX_NMAP
49 xfs_iomap_eof_align_last_fsb(
53 xfs_fileoff_t
*last_fsb
)
55 xfs_fileoff_t new_last_fsb
= 0;
56 xfs_extlen_t align
= 0;
59 if (!XFS_IS_REALTIME_INODE(ip
)) {
61 * Round up the allocation request to a stripe unit
62 * (m_dalign) boundary if the file size is >= stripe unit
63 * size, and we are allocating past the allocation eof.
65 * If mounted with the "-o swalloc" option the alignment is
66 * increased from the strip unit size to the stripe width.
68 if (mp
->m_swidth
&& (mp
->m_flags
& XFS_MOUNT_SWALLOC
))
70 else if (mp
->m_dalign
)
73 if (align
&& XFS_ISIZE(ip
) >= XFS_FSB_TO_B(mp
, align
))
74 new_last_fsb
= roundup_64(*last_fsb
, align
);
78 * Always round up the allocation request to an extent boundary
79 * (when file on a real-time subvolume or has di_extsize hint).
83 align
= roundup_64(new_last_fsb
, extsize
);
86 new_last_fsb
= roundup_64(*last_fsb
, align
);
90 error
= xfs_bmap_eof(ip
, new_last_fsb
, XFS_DATA_FORK
, &eof
);
94 *last_fsb
= new_last_fsb
;
100 xfs_alert_fsblock_zero(
102 xfs_bmbt_irec_t
*imap
)
104 xfs_alert_tag(ip
->i_mount
, XFS_PTAG_FSBLOCK_ZERO
,
105 "Access to block zero in inode %llu "
106 "start_block: %llx start_off: %llx "
107 "blkcnt: %llx extent-state: %x",
108 (unsigned long long)ip
->i_ino
,
109 (unsigned long long)imap
->br_startblock
,
110 (unsigned long long)imap
->br_startoff
,
111 (unsigned long long)imap
->br_blockcount
,
117 xfs_iomap_write_direct(
121 xfs_bmbt_irec_t
*imap
,
124 xfs_mount_t
*mp
= ip
->i_mount
;
125 xfs_fileoff_t offset_fsb
;
126 xfs_fileoff_t last_fsb
;
127 xfs_filblks_t count_fsb
, resaligned
;
128 xfs_fsblock_t firstfsb
;
129 xfs_extlen_t extsz
, temp
;
135 xfs_bmap_free_t free_list
;
136 uint qblocks
, resblks
, resrtextents
;
140 error
= xfs_qm_dqattach(ip
, 0);
142 return XFS_ERROR(error
);
144 rt
= XFS_IS_REALTIME_INODE(ip
);
145 extsz
= xfs_get_extsz_hint(ip
);
147 offset_fsb
= XFS_B_TO_FSBT(mp
, offset
);
148 last_fsb
= XFS_B_TO_FSB(mp
, ((xfs_ufsize_t
)(offset
+ count
)));
149 if ((offset
+ count
) > XFS_ISIZE(ip
)) {
150 error
= xfs_iomap_eof_align_last_fsb(mp
, ip
, extsz
, &last_fsb
);
152 return XFS_ERROR(error
);
154 if (nmaps
&& (imap
->br_startblock
== HOLESTARTBLOCK
))
155 last_fsb
= MIN(last_fsb
, (xfs_fileoff_t
)
156 imap
->br_blockcount
+
159 count_fsb
= last_fsb
- offset_fsb
;
160 ASSERT(count_fsb
> 0);
162 resaligned
= count_fsb
;
163 if (unlikely(extsz
)) {
164 if ((temp
= do_mod(offset_fsb
, extsz
)))
166 if ((temp
= do_mod(resaligned
, extsz
)))
167 resaligned
+= extsz
- temp
;
171 resrtextents
= qblocks
= resaligned
;
172 resrtextents
/= mp
->m_sb
.sb_rextsize
;
173 resblks
= XFS_DIOSTRAT_SPACE_RES(mp
, 0);
174 quota_flag
= XFS_QMOPT_RES_RTBLKS
;
177 resblks
= qblocks
= XFS_DIOSTRAT_SPACE_RES(mp
, resaligned
);
178 quota_flag
= XFS_QMOPT_RES_REGBLKS
;
182 * Allocate and setup the transaction
184 tp
= xfs_trans_alloc(mp
, XFS_TRANS_DIOSTRAT
);
185 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_write
,
186 resblks
, resrtextents
);
188 * Check for running out of space, note: need lock to return
191 xfs_trans_cancel(tp
, 0);
192 return XFS_ERROR(error
);
195 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
197 error
= xfs_trans_reserve_quota_nblks(tp
, ip
, qblocks
, 0, quota_flag
);
199 goto out_trans_cancel
;
201 xfs_trans_ijoin(tp
, ip
, 0);
204 if (offset
< XFS_ISIZE(ip
) || extsz
)
205 bmapi_flag
|= XFS_BMAPI_PREALLOC
;
208 * From this point onwards we overwrite the imap pointer that the
211 xfs_bmap_init(&free_list
, &firstfsb
);
213 error
= xfs_bmapi_write(tp
, ip
, offset_fsb
, count_fsb
, bmapi_flag
,
214 &firstfsb
, 0, imap
, &nimaps
, &free_list
);
216 goto out_bmap_cancel
;
219 * Complete the transaction
221 error
= xfs_bmap_finish(&tp
, &free_list
, &committed
);
223 goto out_bmap_cancel
;
224 error
= xfs_trans_commit(tp
, XFS_TRANS_RELEASE_LOG_RES
);
229 * Copy any maps to caller's array and return any error.
232 error
= XFS_ERROR(ENOSPC
);
236 if (!(imap
->br_startblock
|| XFS_IS_REALTIME_INODE(ip
)))
237 error
= xfs_alert_fsblock_zero(ip
, imap
);
240 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
244 xfs_bmap_cancel(&free_list
);
245 xfs_trans_unreserve_quota_nblks(tp
, ip
, (long)qblocks
, 0, quota_flag
);
247 xfs_trans_cancel(tp
, XFS_TRANS_RELEASE_LOG_RES
| XFS_TRANS_ABORT
);
252 * If the caller is doing a write at the end of the file, then extend the
253 * allocation out to the file system's write iosize. We clean up any extra
254 * space left over when the file is closed in xfs_inactive().
256 * If we find we already have delalloc preallocation beyond EOF, don't do more
257 * preallocation as it it not needed.
260 xfs_iomap_eof_want_preallocate(
265 xfs_bmbt_irec_t
*imap
,
269 xfs_fileoff_t start_fsb
;
270 xfs_filblks_t count_fsb
;
271 xfs_fsblock_t firstblock
;
273 int found_delalloc
= 0;
276 if (offset
+ count
<= XFS_ISIZE(ip
))
280 * If the file is smaller than the minimum prealloc and we are using
281 * dynamic preallocation, don't do any preallocation at all as it is
282 * likely this is the only write to the file that is going to be done.
284 if (!(mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
) &&
285 XFS_ISIZE(ip
) < XFS_FSB_TO_B(mp
, mp
->m_writeio_blocks
))
289 * If there are any real blocks past eof, then don't
290 * do any speculative allocation.
292 start_fsb
= XFS_B_TO_FSBT(mp
, ((xfs_ufsize_t
)(offset
+ count
- 1)));
293 count_fsb
= XFS_B_TO_FSB(mp
, mp
->m_super
->s_maxbytes
);
294 while (count_fsb
> 0) {
296 firstblock
= NULLFSBLOCK
;
297 error
= xfs_bmapi_read(ip
, start_fsb
, count_fsb
, imap
, &imaps
,
301 for (n
= 0; n
< imaps
; n
++) {
302 if ((imap
[n
].br_startblock
!= HOLESTARTBLOCK
) &&
303 (imap
[n
].br_startblock
!= DELAYSTARTBLOCK
))
305 start_fsb
+= imap
[n
].br_blockcount
;
306 count_fsb
-= imap
[n
].br_blockcount
;
308 if (imap
[n
].br_startblock
== DELAYSTARTBLOCK
)
318 * Determine the initial size of the preallocation. We are beyond the current
319 * EOF here, but we need to take into account whether this is a sparse write or
320 * an extending write when determining the preallocation size. Hence we need to
321 * look up the extent that ends at the current write offset and use the result
322 * to determine the preallocation size.
324 * If the extent is a hole, then preallocation is essentially disabled.
325 * Otherwise we take the size of the preceeding data extent as the basis for the
326 * preallocation size. If the size of the extent is greater than half the
327 * maximum extent length, then use the current offset as the basis. This ensures
328 * that for large files the preallocation size always extends to MAXEXTLEN
329 * rather than falling short due to things like stripe unit/width alignment of
333 xfs_iomap_eof_prealloc_initial_size(
334 struct xfs_mount
*mp
,
335 struct xfs_inode
*ip
,
337 xfs_bmbt_irec_t
*imap
,
340 xfs_fileoff_t start_fsb
;
344 ASSERT(nimaps
>= imaps
);
346 /* if we are using a specific prealloc size, return now */
347 if (mp
->m_flags
& XFS_MOUNT_DFLT_IOSIZE
)
350 /* If the file is small, then use the minimum prealloc */
351 if (XFS_ISIZE(ip
) < XFS_FSB_TO_B(mp
, mp
->m_dalign
))
355 * As we write multiple pages, the offset will always align to the
356 * start of a page and hence point to a hole at EOF. i.e. if the size is
357 * 4096 bytes, we only have one block at FSB 0, but XFS_B_TO_FSB(4096)
358 * will return FSB 1. Hence if there are blocks in the file, we want to
359 * point to the block prior to the EOF block and not the hole that maps
360 * directly at @offset.
362 start_fsb
= XFS_B_TO_FSB(mp
, offset
);
365 error
= xfs_bmapi_read(ip
, start_fsb
, 1, imap
, &imaps
, XFS_BMAPI_ENTIRE
);
370 if (imap
[0].br_startblock
== HOLESTARTBLOCK
)
372 if (imap
[0].br_blockcount
<= (MAXEXTLEN
>> 1))
373 return imap
[0].br_blockcount
<< 1;
374 return XFS_B_TO_FSB(mp
, offset
);
378 xfs_quota_need_throttle(
379 struct xfs_inode
*ip
,
381 xfs_fsblock_t alloc_blocks
)
383 struct xfs_dquot
*dq
= xfs_inode_dquot(ip
, type
);
385 if (!dq
|| !xfs_this_quota_on(ip
->i_mount
, type
))
388 /* no hi watermark, no throttle */
389 if (!dq
->q_prealloc_hi_wmark
)
392 /* under the lo watermark, no throttle */
393 if (dq
->q_res_bcount
+ alloc_blocks
< dq
->q_prealloc_lo_wmark
)
400 xfs_quota_calc_throttle(
401 struct xfs_inode
*ip
,
403 xfs_fsblock_t
*qblocks
,
408 struct xfs_dquot
*dq
= xfs_inode_dquot(ip
, type
);
410 /* over hi wmark, squash the prealloc completely */
411 if (dq
->q_res_bcount
>= dq
->q_prealloc_hi_wmark
) {
416 freesp
= dq
->q_prealloc_hi_wmark
- dq
->q_res_bcount
;
417 if (freesp
< dq
->q_low_space
[XFS_QLOWSP_5_PCNT
]) {
419 if (freesp
< dq
->q_low_space
[XFS_QLOWSP_3_PCNT
])
421 if (freesp
< dq
->q_low_space
[XFS_QLOWSP_1_PCNT
])
425 /* only overwrite the throttle values if we are more aggressive */
426 if ((freesp
>> shift
) < (*qblocks
>> *qshift
)) {
433 * If we don't have a user specified preallocation size, dynamically increase
434 * the preallocation size as the size of the file grows. Cap the maximum size
435 * at a single extent or less if the filesystem is near full. The closer the
436 * filesystem is to full, the smaller the maximum prealocation.
439 xfs_iomap_prealloc_size(
440 struct xfs_mount
*mp
,
441 struct xfs_inode
*ip
,
443 struct xfs_bmbt_irec
*imap
,
446 xfs_fsblock_t alloc_blocks
= 0;
449 xfs_fsblock_t qblocks
;
452 alloc_blocks
= xfs_iomap_eof_prealloc_initial_size(mp
, ip
, offset
,
456 qblocks
= alloc_blocks
;
459 * MAXEXTLEN is not a power of two value but we round the prealloc down
460 * to the nearest power of two value after throttling. To prevent the
461 * round down from unconditionally reducing the maximum supported prealloc
462 * size, we round up first, apply appropriate throttling, round down and
463 * cap the value to MAXEXTLEN.
465 alloc_blocks
= XFS_FILEOFF_MIN(roundup_pow_of_two(MAXEXTLEN
),
468 xfs_icsb_sync_counters(mp
, XFS_ICSB_LAZY_COUNT
);
469 freesp
= mp
->m_sb
.sb_fdblocks
;
470 if (freesp
< mp
->m_low_space
[XFS_LOWSP_5_PCNT
]) {
472 if (freesp
< mp
->m_low_space
[XFS_LOWSP_4_PCNT
])
474 if (freesp
< mp
->m_low_space
[XFS_LOWSP_3_PCNT
])
476 if (freesp
< mp
->m_low_space
[XFS_LOWSP_2_PCNT
])
478 if (freesp
< mp
->m_low_space
[XFS_LOWSP_1_PCNT
])
483 * Check each quota to cap the prealloc size and provide a shift
484 * value to throttle with.
486 if (xfs_quota_need_throttle(ip
, XFS_DQ_USER
, alloc_blocks
))
487 xfs_quota_calc_throttle(ip
, XFS_DQ_USER
, &qblocks
, &qshift
);
488 if (xfs_quota_need_throttle(ip
, XFS_DQ_GROUP
, alloc_blocks
))
489 xfs_quota_calc_throttle(ip
, XFS_DQ_GROUP
, &qblocks
, &qshift
);
490 if (xfs_quota_need_throttle(ip
, XFS_DQ_PROJ
, alloc_blocks
))
491 xfs_quota_calc_throttle(ip
, XFS_DQ_PROJ
, &qblocks
, &qshift
);
494 * The final prealloc size is set to the minimum of free space available
495 * in each of the quotas and the overall filesystem.
497 * The shift throttle value is set to the maximum value as determined by
498 * the global low free space values and per-quota low free space values.
500 alloc_blocks
= MIN(alloc_blocks
, qblocks
);
501 shift
= MAX(shift
, qshift
);
504 alloc_blocks
>>= shift
;
506 * rounddown_pow_of_two() returns an undefined result if we pass in
510 alloc_blocks
= rounddown_pow_of_two(alloc_blocks
);
511 if (alloc_blocks
> MAXEXTLEN
)
512 alloc_blocks
= MAXEXTLEN
;
515 * If we are still trying to allocate more space than is
516 * available, squash the prealloc hard. This can happen if we
517 * have a large file on a small filesystem and the above
518 * lowspace thresholds are smaller than MAXEXTLEN.
520 while (alloc_blocks
&& alloc_blocks
>= freesp
)
524 if (alloc_blocks
< mp
->m_writeio_blocks
)
525 alloc_blocks
= mp
->m_writeio_blocks
;
527 trace_xfs_iomap_prealloc_size(ip
, alloc_blocks
, shift
,
528 mp
->m_writeio_blocks
);
534 xfs_iomap_write_delay(
538 xfs_bmbt_irec_t
*ret_imap
)
540 xfs_mount_t
*mp
= ip
->i_mount
;
541 xfs_fileoff_t offset_fsb
;
542 xfs_fileoff_t last_fsb
;
543 xfs_off_t aligned_offset
;
544 xfs_fileoff_t ioalign
;
547 xfs_bmbt_irec_t imap
[XFS_WRITE_IMAPS
];
551 ASSERT(xfs_isilocked(ip
, XFS_ILOCK_EXCL
));
554 * Make sure that the dquots are there. This doesn't hold
555 * the ilock across a disk read.
557 error
= xfs_qm_dqattach_locked(ip
, 0);
559 return XFS_ERROR(error
);
561 extsz
= xfs_get_extsz_hint(ip
);
562 offset_fsb
= XFS_B_TO_FSBT(mp
, offset
);
564 error
= xfs_iomap_eof_want_preallocate(mp
, ip
, offset
, count
,
565 imap
, XFS_WRITE_IMAPS
, &prealloc
);
571 xfs_fsblock_t alloc_blocks
;
573 alloc_blocks
= xfs_iomap_prealloc_size(mp
, ip
, offset
, imap
,
576 aligned_offset
= XFS_WRITEIO_ALIGN(mp
, (offset
+ count
- 1));
577 ioalign
= XFS_B_TO_FSBT(mp
, aligned_offset
);
578 last_fsb
= ioalign
+ alloc_blocks
;
580 last_fsb
= XFS_B_TO_FSB(mp
, ((xfs_ufsize_t
)(offset
+ count
)));
583 if (prealloc
|| extsz
) {
584 error
= xfs_iomap_eof_align_last_fsb(mp
, ip
, extsz
, &last_fsb
);
590 * Make sure preallocation does not create extents beyond the range we
591 * actually support in this filesystem.
593 if (last_fsb
> XFS_B_TO_FSB(mp
, mp
->m_super
->s_maxbytes
))
594 last_fsb
= XFS_B_TO_FSB(mp
, mp
->m_super
->s_maxbytes
);
596 ASSERT(last_fsb
> offset_fsb
);
598 nimaps
= XFS_WRITE_IMAPS
;
599 error
= xfs_bmapi_delay(ip
, offset_fsb
, last_fsb
- offset_fsb
,
600 imap
, &nimaps
, XFS_BMAPI_ENTIRE
);
607 return XFS_ERROR(error
);
611 * If bmapi returned us nothing, we got either ENOSPC or EDQUOT. Retry
612 * without EOF preallocation.
615 trace_xfs_delalloc_enospc(ip
, offset
, count
);
621 return XFS_ERROR(error
? error
: ENOSPC
);
624 if (!(imap
[0].br_startblock
|| XFS_IS_REALTIME_INODE(ip
)))
625 return xfs_alert_fsblock_zero(ip
, &imap
[0]);
628 * Tag the inode as speculatively preallocated so we can reclaim this
629 * space on demand, if necessary.
632 xfs_inode_set_eofblocks_tag(ip
);
639 * Pass in a delayed allocate extent, convert it to real extents;
640 * return to the caller the extent we create which maps on top of
641 * the originating callers request.
643 * Called without a lock on the inode.
645 * We no longer bother to look at the incoming map - all we have to
646 * guarantee is that whatever we allocate fills the required range.
649 xfs_iomap_write_allocate(
652 xfs_bmbt_irec_t
*imap
)
654 xfs_mount_t
*mp
= ip
->i_mount
;
655 xfs_fileoff_t offset_fsb
, last_block
;
656 xfs_fileoff_t end_fsb
, map_start_fsb
;
657 xfs_fsblock_t first_block
;
658 xfs_bmap_free_t free_list
;
659 xfs_filblks_t count_fsb
;
661 int nimaps
, committed
;
666 * Make sure that the dquots are there.
668 error
= xfs_qm_dqattach(ip
, 0);
670 return XFS_ERROR(error
);
672 offset_fsb
= XFS_B_TO_FSBT(mp
, offset
);
673 count_fsb
= imap
->br_blockcount
;
674 map_start_fsb
= imap
->br_startoff
;
676 XFS_STATS_ADD(xs_xstrat_bytes
, XFS_FSB_TO_B(mp
, count_fsb
));
678 while (count_fsb
!= 0) {
680 * Set up a transaction with which to allocate the
681 * backing store for the file. Do allocations in a
682 * loop until we get some space in the range we are
683 * interested in. The other space that might be allocated
684 * is in the delayed allocation extent on which we sit
685 * but before our buffer starts.
689 while (nimaps
== 0) {
690 tp
= xfs_trans_alloc(mp
, XFS_TRANS_STRAT_WRITE
);
691 tp
->t_flags
|= XFS_TRANS_RESERVE
;
692 nres
= XFS_EXTENTADD_SPACE_RES(mp
, XFS_DATA_FORK
);
693 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_write
,
696 xfs_trans_cancel(tp
, 0);
697 return XFS_ERROR(error
);
699 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
700 xfs_trans_ijoin(tp
, ip
, 0);
702 xfs_bmap_init(&free_list
, &first_block
);
705 * it is possible that the extents have changed since
706 * we did the read call as we dropped the ilock for a
707 * while. We have to be careful about truncates or hole
708 * punchs here - we are not allowed to allocate
709 * non-delalloc blocks here.
711 * The only protection against truncation is the pages
712 * for the range we are being asked to convert are
713 * locked and hence a truncate will block on them
716 * As a result, if we go beyond the range we really
717 * need and hit an delalloc extent boundary followed by
718 * a hole while we have excess blocks in the map, we
719 * will fill the hole incorrectly and overrun the
720 * transaction reservation.
722 * Using a single map prevents this as we are forced to
723 * check each map we look for overlap with the desired
724 * range and abort as soon as we find it. Also, given
725 * that we only return a single map, having one beyond
726 * what we can return is probably a bit silly.
728 * We also need to check that we don't go beyond EOF;
729 * this is a truncate optimisation as a truncate sets
730 * the new file size before block on the pages we
731 * currently have locked under writeback. Because they
732 * are about to be tossed, we don't need to write them
736 end_fsb
= XFS_B_TO_FSB(mp
, XFS_ISIZE(ip
));
737 error
= xfs_bmap_last_offset(NULL
, ip
, &last_block
,
742 last_block
= XFS_FILEOFF_MAX(last_block
, end_fsb
);
743 if ((map_start_fsb
+ count_fsb
) > last_block
) {
744 count_fsb
= last_block
- map_start_fsb
;
745 if (count_fsb
== 0) {
752 * From this point onwards we overwrite the imap
753 * pointer that the caller gave to us.
755 error
= xfs_bmapi_write(tp
, ip
, map_start_fsb
,
757 XFS_BMAPI_STACK_SWITCH
,
759 imap
, &nimaps
, &free_list
);
763 error
= xfs_bmap_finish(&tp
, &free_list
, &committed
);
767 error
= xfs_trans_commit(tp
, XFS_TRANS_RELEASE_LOG_RES
);
771 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
775 * See if we were able to allocate an extent that
776 * covers at least part of the callers request
778 if (!(imap
->br_startblock
|| XFS_IS_REALTIME_INODE(ip
)))
779 return xfs_alert_fsblock_zero(ip
, imap
);
781 if ((offset_fsb
>= imap
->br_startoff
) &&
782 (offset_fsb
< (imap
->br_startoff
+
783 imap
->br_blockcount
))) {
784 XFS_STATS_INC(xs_xstrat_quick
);
789 * So far we have not mapped the requested part of the
790 * file, just surrounding data, try again.
792 count_fsb
-= imap
->br_blockcount
;
793 map_start_fsb
= imap
->br_startoff
+ imap
->br_blockcount
;
797 xfs_bmap_cancel(&free_list
);
798 xfs_trans_cancel(tp
, XFS_TRANS_RELEASE_LOG_RES
| XFS_TRANS_ABORT
);
800 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
801 return XFS_ERROR(error
);
805 xfs_iomap_write_unwritten(
810 xfs_mount_t
*mp
= ip
->i_mount
;
811 xfs_fileoff_t offset_fsb
;
812 xfs_filblks_t count_fsb
;
813 xfs_filblks_t numblks_fsb
;
814 xfs_fsblock_t firstfsb
;
817 xfs_bmbt_irec_t imap
;
818 xfs_bmap_free_t free_list
;
824 trace_xfs_unwritten_convert(ip
, offset
, count
);
826 offset_fsb
= XFS_B_TO_FSBT(mp
, offset
);
827 count_fsb
= XFS_B_TO_FSB(mp
, (xfs_ufsize_t
)offset
+ count
);
828 count_fsb
= (xfs_filblks_t
)(count_fsb
- offset_fsb
);
831 * Reserve enough blocks in this transaction for two complete extent
832 * btree splits. We may be converting the middle part of an unwritten
833 * extent and in this case we will insert two new extents in the btree
834 * each of which could cause a full split.
836 * This reservation amount will be used in the first call to
837 * xfs_bmbt_split() to select an AG with enough space to satisfy the
838 * rest of the operation.
840 resblks
= XFS_DIOSTRAT_SPACE_RES(mp
, 0) << 1;
844 * set up a transaction to convert the range of extents
845 * from unwritten to real. Do allocations in a loop until
846 * we have covered the range passed in.
848 * Note that we open code the transaction allocation here
849 * to pass KM_NOFS--we can't risk to recursing back into
850 * the filesystem here as we might be asked to write out
851 * the same inode that we complete here and might deadlock
854 sb_start_intwrite(mp
->m_super
);
855 tp
= _xfs_trans_alloc(mp
, XFS_TRANS_STRAT_WRITE
, KM_NOFS
);
856 tp
->t_flags
|= XFS_TRANS_RESERVE
| XFS_TRANS_FREEZE_PROT
;
857 error
= xfs_trans_reserve(tp
, &M_RES(mp
)->tr_write
,
860 xfs_trans_cancel(tp
, 0);
861 return XFS_ERROR(error
);
864 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
865 xfs_trans_ijoin(tp
, ip
, 0);
868 * Modify the unwritten extent state of the buffer.
870 xfs_bmap_init(&free_list
, &firstfsb
);
872 error
= xfs_bmapi_write(tp
, ip
, offset_fsb
, count_fsb
,
873 XFS_BMAPI_CONVERT
, &firstfsb
,
874 1, &imap
, &nimaps
, &free_list
);
876 goto error_on_bmapi_transaction
;
879 * Log the updated inode size as we go. We have to be careful
880 * to only log it up to the actual write offset if it is
881 * halfway into a block.
883 i_size
= XFS_FSB_TO_B(mp
, offset_fsb
+ count_fsb
);
884 if (i_size
> offset
+ count
)
885 i_size
= offset
+ count
;
887 i_size
= xfs_new_eof(ip
, i_size
);
889 ip
->i_d
.di_size
= i_size
;
890 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_CORE
);
893 error
= xfs_bmap_finish(&tp
, &free_list
, &committed
);
895 goto error_on_bmapi_transaction
;
897 error
= xfs_trans_commit(tp
, XFS_TRANS_RELEASE_LOG_RES
);
898 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
900 return XFS_ERROR(error
);
902 if (!(imap
.br_startblock
|| XFS_IS_REALTIME_INODE(ip
)))
903 return xfs_alert_fsblock_zero(ip
, &imap
);
905 if ((numblks_fsb
= imap
.br_blockcount
) == 0) {
907 * The numblks_fsb value should always get
908 * smaller, otherwise the loop is stuck.
910 ASSERT(imap
.br_blockcount
);
913 offset_fsb
+= numblks_fsb
;
914 count_fsb
-= numblks_fsb
;
915 } while (count_fsb
> 0);
919 error_on_bmapi_transaction
:
920 xfs_bmap_cancel(&free_list
);
921 xfs_trans_cancel(tp
, (XFS_TRANS_RELEASE_LOG_RES
| XFS_TRANS_ABORT
));
922 xfs_iunlock(ip
, XFS_ILOCK_EXCL
);
923 return XFS_ERROR(error
);