2 * lcnalloc.c - Cluster (de)allocation code. Part of the Linux-NTFS project.
4 * Copyright (c) 2004-2005 Anton Altaparmakov
6 * This program/include file is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as published
8 * by the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program/include file is distributed in the hope that it will be
12 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
13 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program (in the main directory of the Linux-NTFS
18 * distribution in the file COPYING); if not, write to the Free Software
19 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24 #include <linux/pagemap.h>
37 * ntfs_cluster_free_from_rl_nolock - free clusters from runlist
38 * @vol: mounted ntfs volume on which to free the clusters
39 * @rl: runlist describing the clusters to free
41 * Free all the clusters described by the runlist @rl on the volume @vol. In
42 * the case of an error being returned, at least some of the clusters were not
45 * Return 0 on success and -errno on error.
47 * Locking: - The volume lcn bitmap must be locked for writing on entry and is
48 * left locked on return.
50 int ntfs_cluster_free_from_rl_nolock(ntfs_volume
*vol
,
51 const runlist_element
*rl
)
53 struct inode
*lcnbmp_vi
= vol
->lcnbmp_ino
;
56 ntfs_debug("Entering.");
59 for (; rl
->length
; rl
++) {
64 err
= ntfs_bitmap_clear_run(lcnbmp_vi
, rl
->lcn
, rl
->length
);
65 if (unlikely(err
&& (!ret
|| ret
== -ENOMEM
) && ret
!= err
))
73 * ntfs_cluster_alloc - allocate clusters on an ntfs volume
74 * @vol: mounted ntfs volume on which to allocate the clusters
75 * @start_vcn: vcn to use for the first allocated cluster
76 * @count: number of clusters to allocate
77 * @start_lcn: starting lcn at which to allocate the clusters (or -1 if none)
78 * @zone: zone from which to allocate the clusters
80 * Allocate @count clusters preferably starting at cluster @start_lcn or at the
81 * current allocator position if @start_lcn is -1, on the mounted ntfs volume
82 * @vol. @zone is either DATA_ZONE for allocation of normal clusters or
83 * MFT_ZONE for allocation of clusters for the master file table, i.e. the
84 * $MFT/$DATA attribute.
86 * @start_vcn specifies the vcn of the first allocated cluster. This makes
87 * merging the resulting runlist with the old runlist easier.
89 * You need to check the return value with IS_ERR(). If this is false, the
90 * function was successful and the return value is a runlist describing the
91 * allocated cluster(s). If IS_ERR() is true, the function failed and
92 * PTR_ERR() gives you the error code.
94 * Notes on the allocation algorithm
95 * =================================
97 * There are two data zones. First is the area between the end of the mft zone
98 * and the end of the volume, and second is the area between the start of the
99 * volume and the start of the mft zone. On unmodified/standard NTFS 1.x
100 * volumes, the second data zone does not exist due to the mft zone being
101 * expanded to cover the start of the volume in order to reserve space for the
102 * mft bitmap attribute.
104 * This is not the prettiest function but the complexity stems from the need of
105 * implementing the mft vs data zoned approach and from the fact that we have
106 * access to the lcn bitmap in portions of up to 8192 bytes at a time, so we
107 * need to cope with crossing over boundaries of two buffers. Further, the
108 * fact that the allocator allows for caller supplied hints as to the location
109 * of where allocation should begin and the fact that the allocator keeps track
110 * of where in the data zones the next natural allocation should occur,
111 * contribute to the complexity of the function. But it should all be
112 * worthwhile, because this allocator should: 1) be a full implementation of
113 * the MFT zone approach used by Windows NT, 2) cause reduction in
114 * fragmentation, and 3) be speedy in allocations (the code is not optimized
115 * for speed, but the algorithm is, so further speed improvements are probably
118 * FIXME: We should be monitoring cluster allocation and increment the MFT zone
119 * size dynamically but this is something for the future. We will just cause
120 * heavier fragmentation by not doing it and I am not even sure Windows would
121 * grow the MFT zone dynamically, so it might even be correct not to do this.
122 * The overhead in doing dynamic MFT zone expansion would be very large and
123 * unlikely worth the effort. (AIA)
125 * TODO: I have added in double the required zone position pointer wrap around
126 * logic which can be optimized to having only one of the two logic sets.
127 * However, having the double logic will work fine, but if we have only one of
128 * the sets and we get it wrong somewhere, then we get into trouble, so
129 * removing the duplicate logic requires _very_ careful consideration of _all_
130 * possible code paths. So at least for now, I am leaving the double logic -
131 * better safe than sorry... (AIA)
133 * Locking: - The volume lcn bitmap must be unlocked on entry and is unlocked
135 * - This function takes the volume lcn bitmap lock for writing and
136 * modifies the bitmap contents.
138 runlist_element
*ntfs_cluster_alloc(ntfs_volume
*vol
, const VCN start_vcn
,
139 const s64 count
, const LCN start_lcn
,
140 const NTFS_CLUSTER_ALLOCATION_ZONES zone
)
142 LCN zone_start
, zone_end
, bmp_pos
, bmp_initial_pos
, last_read_pos
, lcn
;
143 LCN prev_lcn
= 0, prev_run_len
= 0, mft_zone_size
;
146 struct inode
*lcnbmp_vi
;
147 runlist_element
*rl
= NULL
;
148 struct address_space
*mapping
;
149 struct page
*page
= NULL
;
151 int err
= 0, rlpos
, rlsize
, buf_size
;
152 u8 pass
, done_zones
, search_zone
, need_writeback
= 0, bit
;
154 ntfs_debug("Entering for start_vcn 0x%llx, count 0x%llx, start_lcn "
155 "0x%llx, zone %s_ZONE.", (unsigned long long)start_vcn
,
156 (unsigned long long)count
,
157 (unsigned long long)start_lcn
,
158 zone
== MFT_ZONE
? "MFT" : "DATA");
160 lcnbmp_vi
= vol
->lcnbmp_ino
;
162 BUG_ON(start_vcn
< 0);
164 BUG_ON(start_lcn
< -1);
165 BUG_ON(zone
< FIRST_ZONE
);
166 BUG_ON(zone
> LAST_ZONE
);
168 /* Return NULL if @count is zero. */
171 /* Take the lcnbmp lock for writing. */
172 down_write(&vol
->lcnbmp_lock
);
174 * If no specific @start_lcn was requested, use the current data zone
175 * position, otherwise use the requested @start_lcn but make sure it
176 * lies outside the mft zone. Also set done_zones to 0 (no zones done)
177 * and pass depending on whether we are starting inside a zone (1) or
178 * at the beginning of a zone (2). If requesting from the MFT_ZONE,
179 * we either start at the current position within the mft zone or at
180 * the specified position. If the latter is out of bounds then we start
181 * at the beginning of the MFT_ZONE.
186 * zone_start and zone_end are the current search range. search_zone
187 * is 1 for mft zone, 2 for data zone 1 (end of mft zone till end of
188 * volume) and 4 for data zone 2 (start of volume till start of mft
191 zone_start
= start_lcn
;
192 if (zone_start
< 0) {
193 if (zone
== DATA_ZONE
)
194 zone_start
= vol
->data1_zone_pos
;
196 zone_start
= vol
->mft_zone_pos
;
199 * Zone starts at beginning of volume which means a
200 * single pass is sufficient.
204 } else if (zone
== DATA_ZONE
&& zone_start
>= vol
->mft_zone_start
&&
205 zone_start
< vol
->mft_zone_end
) {
206 zone_start
= vol
->mft_zone_end
;
208 * Starting at beginning of data1_zone which means a single
209 * pass in this zone is sufficient.
212 } else if (zone
== MFT_ZONE
&& (zone_start
< vol
->mft_zone_start
||
213 zone_start
>= vol
->mft_zone_end
)) {
214 zone_start
= vol
->mft_lcn
;
215 if (!vol
->mft_zone_end
)
218 * Starting at beginning of volume which means a single pass
223 if (zone
== MFT_ZONE
) {
224 zone_end
= vol
->mft_zone_end
;
226 } else /* if (zone == DATA_ZONE) */ {
227 /* Skip searching the mft zone. */
229 if (zone_start
>= vol
->mft_zone_end
) {
230 zone_end
= vol
->nr_clusters
;
233 zone_end
= vol
->mft_zone_start
;
238 * bmp_pos is the current bit position inside the bitmap. We use
239 * bmp_initial_pos to determine whether or not to do a zone switch.
241 bmp_pos
= bmp_initial_pos
= zone_start
;
243 /* Loop until all clusters are allocated, i.e. clusters == 0. */
246 mapping
= lcnbmp_vi
->i_mapping
;
247 i_size
= i_size_read(lcnbmp_vi
);
249 ntfs_debug("Start of outer while loop: done_zones 0x%x, "
250 "search_zone %i, pass %i, zone_start 0x%llx, "
251 "zone_end 0x%llx, bmp_initial_pos 0x%llx, "
252 "bmp_pos 0x%llx, rlpos %i, rlsize %i.",
253 done_zones
, search_zone
, pass
,
254 (unsigned long long)zone_start
,
255 (unsigned long long)zone_end
,
256 (unsigned long long)bmp_initial_pos
,
257 (unsigned long long)bmp_pos
, rlpos
, rlsize
);
258 /* Loop until we run out of free clusters. */
259 last_read_pos
= bmp_pos
>> 3;
260 ntfs_debug("last_read_pos 0x%llx.",
261 (unsigned long long)last_read_pos
);
262 if (last_read_pos
> i_size
) {
263 ntfs_debug("End of attribute reached. "
264 "Skipping to zone_pass_done.");
268 if (need_writeback
) {
269 ntfs_debug("Marking page dirty.");
270 flush_dcache_page(page
);
271 set_page_dirty(page
);
274 ntfs_unmap_page(page
);
276 page
= ntfs_map_page(mapping
, last_read_pos
>>
280 ntfs_error(vol
->sb
, "Failed to map page.");
283 buf_size
= last_read_pos
& ~PAGE_CACHE_MASK
;
284 buf
= page_address(page
) + buf_size
;
285 buf_size
= PAGE_CACHE_SIZE
- buf_size
;
286 if (unlikely(last_read_pos
+ buf_size
> i_size
))
287 buf_size
= i_size
- last_read_pos
;
291 ntfs_debug("Before inner while loop: buf_size %i, lcn 0x%llx, "
292 "bmp_pos 0x%llx, need_writeback %i.", buf_size
,
293 (unsigned long long)lcn
,
294 (unsigned long long)bmp_pos
, need_writeback
);
295 while (lcn
< buf_size
&& lcn
+ bmp_pos
< zone_end
) {
296 byte
= buf
+ (lcn
>> 3);
297 ntfs_debug("In inner while loop: buf_size %i, "
298 "lcn 0x%llx, bmp_pos 0x%llx, "
299 "need_writeback %i, byte ofs 0x%x, "
300 "*byte 0x%x.", buf_size
,
301 (unsigned long long)lcn
,
302 (unsigned long long)bmp_pos
,
304 (unsigned int)(lcn
>> 3),
305 (unsigned int)*byte
);
306 /* Skip full bytes. */
308 lcn
= (lcn
+ 8) & ~(LCN
)7;
309 ntfs_debug("Continuing while loop 1.");
312 bit
= 1 << (lcn
& 7);
313 ntfs_debug("bit %i.", bit
);
314 /* If the bit is already set, go onto the next one. */
317 ntfs_debug("Continuing while loop 2.");
321 * Allocate more memory if needed, including space for
322 * the terminator element.
323 * ntfs_malloc_nofs() operates on whole pages only.
325 if ((rlpos
+ 2) * sizeof(*rl
) > rlsize
) {
326 runlist_element
*rl2
;
328 ntfs_debug("Reallocating memory.");
330 ntfs_debug("First free bit is at LCN "
334 rl2
= ntfs_malloc_nofs(rlsize
+ (int)PAGE_SIZE
);
335 if (unlikely(!rl2
)) {
337 ntfs_error(vol
->sb
, "Failed to "
341 memcpy(rl2
, rl
, rlsize
);
345 ntfs_debug("Reallocated memory, rlsize 0x%x.",
348 /* Allocate the bitmap bit. */
350 /* We need to write this bitmap page to disk. */
352 ntfs_debug("*byte 0x%x, need_writeback is set.",
353 (unsigned int)*byte
);
355 * Coalesce with previous run if adjacent LCNs.
356 * Otherwise, append a new run.
358 ntfs_debug("Adding run (lcn 0x%llx, len 0x%llx), "
359 "prev_lcn 0x%llx, lcn 0x%llx, "
360 "bmp_pos 0x%llx, prev_run_len 0x%llx, "
362 (unsigned long long)(lcn
+ bmp_pos
),
363 1ULL, (unsigned long long)prev_lcn
,
364 (unsigned long long)lcn
,
365 (unsigned long long)bmp_pos
,
366 (unsigned long long)prev_run_len
,
368 if (prev_lcn
== lcn
+ bmp_pos
- prev_run_len
&& rlpos
) {
369 ntfs_debug("Coalescing to run (lcn 0x%llx, "
374 rl
[rlpos
- 1].length
);
375 rl
[rlpos
- 1].length
= ++prev_run_len
;
376 ntfs_debug("Run now (lcn 0x%llx, len 0x%llx), "
377 "prev_run_len 0x%llx.",
381 rl
[rlpos
- 1].length
,
386 ntfs_debug("Adding new run, (previous "
392 rl
[rlpos
- 1].length
);
393 rl
[rlpos
].vcn
= rl
[rlpos
- 1].vcn
+
396 ntfs_debug("Adding new run, is first "
398 rl
[rlpos
].vcn
= start_vcn
;
400 rl
[rlpos
].lcn
= prev_lcn
= lcn
+ bmp_pos
;
401 rl
[rlpos
].length
= prev_run_len
= 1;
408 * Update the current zone position. Positions
409 * of already scanned zones have been updated
410 * during the respective zone switches.
412 tc
= lcn
+ bmp_pos
+ 1;
413 ntfs_debug("Done. Updating current zone "
414 "position, tc 0x%llx, "
416 (unsigned long long)tc
,
418 switch (search_zone
) {
420 ntfs_debug("Before checks, "
425 if (tc
>= vol
->mft_zone_end
) {
428 if (!vol
->mft_zone_end
)
429 vol
->mft_zone_pos
= 0;
430 } else if ((bmp_initial_pos
>=
432 tc
> vol
->mft_zone_pos
)
433 && tc
>= vol
->mft_lcn
)
434 vol
->mft_zone_pos
= tc
;
435 ntfs_debug("After checks, "
442 ntfs_debug("Before checks, "
443 "vol->data1_zone_pos "
446 vol
->data1_zone_pos
);
447 if (tc
>= vol
->nr_clusters
)
448 vol
->data1_zone_pos
=
450 else if ((bmp_initial_pos
>=
451 vol
->data1_zone_pos
||
452 tc
> vol
->data1_zone_pos
)
453 && tc
>= vol
->mft_zone_end
)
454 vol
->data1_zone_pos
= tc
;
455 ntfs_debug("After checks, "
456 "vol->data1_zone_pos "
459 vol
->data1_zone_pos
);
462 ntfs_debug("Before checks, "
463 "vol->data2_zone_pos "
466 vol
->data2_zone_pos
);
467 if (tc
>= vol
->mft_zone_start
)
468 vol
->data2_zone_pos
= 0;
469 else if (bmp_initial_pos
>=
470 vol
->data2_zone_pos
||
471 tc
> vol
->data2_zone_pos
)
472 vol
->data2_zone_pos
= tc
;
473 ntfs_debug("After checks, "
474 "vol->data2_zone_pos "
477 vol
->data2_zone_pos
);
482 ntfs_debug("Finished. Going to out.");
488 ntfs_debug("After inner while loop: buf_size 0x%x, lcn "
489 "0x%llx, bmp_pos 0x%llx, need_writeback %i.",
490 buf_size
, (unsigned long long)lcn
,
491 (unsigned long long)bmp_pos
, need_writeback
);
492 if (bmp_pos
< zone_end
) {
493 ntfs_debug("Continuing outer while loop, "
494 "bmp_pos 0x%llx, zone_end 0x%llx.",
495 (unsigned long long)bmp_pos
,
496 (unsigned long long)zone_end
);
499 zone_pass_done
: /* Finished with the current zone pass. */
500 ntfs_debug("At zone_pass_done, pass %i.", pass
);
503 * Now do pass 2, scanning the first part of the zone
504 * we omitted in pass 1.
507 zone_end
= zone_start
;
508 switch (search_zone
) {
509 case 1: /* mft_zone */
510 zone_start
= vol
->mft_zone_start
;
512 case 2: /* data1_zone */
513 zone_start
= vol
->mft_zone_end
;
515 case 4: /* data2_zone */
522 if (zone_end
< zone_start
)
523 zone_end
= zone_start
;
524 bmp_pos
= zone_start
;
525 ntfs_debug("Continuing outer while loop, pass 2, "
526 "zone_start 0x%llx, zone_end 0x%llx, "
528 (unsigned long long)zone_start
,
529 (unsigned long long)zone_end
,
530 (unsigned long long)bmp_pos
);
534 ntfs_debug("At done_zones_check, search_zone %i, done_zones "
535 "before 0x%x, done_zones after 0x%x.",
536 search_zone
, done_zones
,
537 done_zones
| search_zone
);
538 done_zones
|= search_zone
;
539 if (done_zones
< 7) {
540 ntfs_debug("Switching zone.");
541 /* Now switch to the next zone we haven't done yet. */
543 switch (search_zone
) {
545 ntfs_debug("Switching from mft zone to data1 "
547 /* Update mft zone position. */
551 ntfs_debug("Before checks, "
556 tc
= rl
[rlpos
- 1].lcn
+
557 rl
[rlpos
- 1].length
;
558 if (tc
>= vol
->mft_zone_end
) {
561 if (!vol
->mft_zone_end
)
562 vol
->mft_zone_pos
= 0;
563 } else if ((bmp_initial_pos
>=
565 tc
> vol
->mft_zone_pos
)
566 && tc
>= vol
->mft_lcn
)
567 vol
->mft_zone_pos
= tc
;
568 ntfs_debug("After checks, "
574 /* Switch from mft zone to data1 zone. */
575 switch_to_data1_zone
: search_zone
= 2;
576 zone_start
= bmp_initial_pos
=
578 zone_end
= vol
->nr_clusters
;
579 if (zone_start
== vol
->mft_zone_end
)
581 if (zone_start
>= zone_end
) {
582 vol
->data1_zone_pos
= zone_start
=
588 ntfs_debug("Switching from data1 zone to "
590 /* Update data1 zone position. */
594 ntfs_debug("Before checks, "
595 "vol->data1_zone_pos "
598 vol
->data1_zone_pos
);
599 tc
= rl
[rlpos
- 1].lcn
+
600 rl
[rlpos
- 1].length
;
601 if (tc
>= vol
->nr_clusters
)
602 vol
->data1_zone_pos
=
604 else if ((bmp_initial_pos
>=
605 vol
->data1_zone_pos
||
606 tc
> vol
->data1_zone_pos
)
607 && tc
>= vol
->mft_zone_end
)
608 vol
->data1_zone_pos
= tc
;
609 ntfs_debug("After checks, "
610 "vol->data1_zone_pos "
613 vol
->data1_zone_pos
);
615 /* Switch from data1 zone to data2 zone. */
617 zone_start
= bmp_initial_pos
=
619 zone_end
= vol
->mft_zone_start
;
622 if (zone_start
>= zone_end
) {
623 vol
->data2_zone_pos
= zone_start
=
629 ntfs_debug("Switching from data2 zone to "
631 /* Update data2 zone position. */
635 ntfs_debug("Before checks, "
636 "vol->data2_zone_pos "
639 vol
->data2_zone_pos
);
640 tc
= rl
[rlpos
- 1].lcn
+
641 rl
[rlpos
- 1].length
;
642 if (tc
>= vol
->mft_zone_start
)
643 vol
->data2_zone_pos
= 0;
644 else if (bmp_initial_pos
>=
645 vol
->data2_zone_pos
||
646 tc
> vol
->data2_zone_pos
)
647 vol
->data2_zone_pos
= tc
;
648 ntfs_debug("After checks, "
649 "vol->data2_zone_pos "
652 vol
->data2_zone_pos
);
654 /* Switch from data2 zone to data1 zone. */
655 goto switch_to_data1_zone
;
659 ntfs_debug("After zone switch, search_zone %i, "
660 "pass %i, bmp_initial_pos 0x%llx, "
661 "zone_start 0x%llx, zone_end 0x%llx.",
663 (unsigned long long)bmp_initial_pos
,
664 (unsigned long long)zone_start
,
665 (unsigned long long)zone_end
);
666 bmp_pos
= zone_start
;
667 if (zone_start
== zone_end
) {
668 ntfs_debug("Empty zone, going to "
669 "done_zones_check.");
670 /* Empty zone. Don't bother searching it. */
671 goto done_zones_check
;
673 ntfs_debug("Continuing outer while loop.");
675 } /* done_zones == 7 */
676 ntfs_debug("All zones are finished.");
678 * All zones are finished! If DATA_ZONE, shrink mft zone. If
679 * MFT_ZONE, we have really run out of space.
681 mft_zone_size
= vol
->mft_zone_end
- vol
->mft_zone_start
;
682 ntfs_debug("vol->mft_zone_start 0x%llx, vol->mft_zone_end "
683 "0x%llx, mft_zone_size 0x%llx.",
684 (unsigned long long)vol
->mft_zone_start
,
685 (unsigned long long)vol
->mft_zone_end
,
686 (unsigned long long)mft_zone_size
);
687 if (zone
== MFT_ZONE
|| mft_zone_size
<= 0) {
688 ntfs_debug("No free clusters left, going to out.");
689 /* Really no more space left on device. */
692 } /* zone == DATA_ZONE && mft_zone_size > 0 */
693 ntfs_debug("Shrinking mft zone.");
694 zone_end
= vol
->mft_zone_end
;
696 if (mft_zone_size
> 0)
697 vol
->mft_zone_end
= vol
->mft_zone_start
+ mft_zone_size
;
698 else /* mft zone and data2 zone no longer exist. */
699 vol
->data2_zone_pos
= vol
->mft_zone_start
=
700 vol
->mft_zone_end
= 0;
701 if (vol
->mft_zone_pos
>= vol
->mft_zone_end
) {
702 vol
->mft_zone_pos
= vol
->mft_lcn
;
703 if (!vol
->mft_zone_end
)
704 vol
->mft_zone_pos
= 0;
706 bmp_pos
= zone_start
= bmp_initial_pos
=
707 vol
->data1_zone_pos
= vol
->mft_zone_end
;
711 ntfs_debug("After shrinking mft zone, mft_zone_size 0x%llx, "
712 "vol->mft_zone_start 0x%llx, "
713 "vol->mft_zone_end 0x%llx, "
714 "vol->mft_zone_pos 0x%llx, search_zone 2, "
715 "pass 2, dones_zones 0x%x, zone_start 0x%llx, "
716 "zone_end 0x%llx, vol->data1_zone_pos 0x%llx, "
717 "continuing outer while loop.",
718 (unsigned long long)mft_zone_size
,
719 (unsigned long long)vol
->mft_zone_start
,
720 (unsigned long long)vol
->mft_zone_end
,
721 (unsigned long long)vol
->mft_zone_pos
,
722 done_zones
, (unsigned long long)zone_start
,
723 (unsigned long long)zone_end
,
724 (unsigned long long)vol
->data1_zone_pos
);
726 ntfs_debug("After outer while loop.");
728 ntfs_debug("At out.");
729 /* Add runlist terminator element. */
731 rl
[rlpos
].vcn
= rl
[rlpos
- 1].vcn
+ rl
[rlpos
- 1].length
;
732 rl
[rlpos
].lcn
= LCN_RL_NOT_MAPPED
;
733 rl
[rlpos
].length
= 0;
735 if (likely(page
&& !IS_ERR(page
))) {
736 if (need_writeback
) {
737 ntfs_debug("Marking page dirty.");
738 flush_dcache_page(page
);
739 set_page_dirty(page
);
742 ntfs_unmap_page(page
);
745 up_write(&vol
->lcnbmp_lock
);
749 ntfs_error(vol
->sb
, "Failed to allocate clusters, aborting "
755 ntfs_debug("Not enough space to complete allocation, "
756 "err -ENOSPC, first free lcn 0x%llx, "
757 "could allocate up to 0x%llx "
759 (unsigned long long)rl
[0].lcn
,
760 (unsigned long long)(count
- clusters
));
761 /* Deallocate all allocated clusters. */
762 ntfs_debug("Attempting rollback...");
763 err2
= ntfs_cluster_free_from_rl_nolock(vol
, rl
);
765 ntfs_error(vol
->sb
, "Failed to rollback (error %i). "
766 "Leaving inconsistent metadata! "
767 "Unmount and run chkdsk.", err2
);
770 /* Free the runlist. */
772 } else if (err
== -ENOSPC
)
773 ntfs_debug("No space left at all, err = -ENOSPC, first free "
775 (long long)vol
->data1_zone_pos
);
776 up_write(&vol
->lcnbmp_lock
);
781 * __ntfs_cluster_free - free clusters on an ntfs volume
782 * @vi: vfs inode whose runlist describes the clusters to free
783 * @start_vcn: vcn in the runlist of @vi at which to start freeing clusters
784 * @count: number of clusters to free or -1 for all clusters
785 * @write_locked: true if the runlist is locked for writing
786 * @is_rollback: true if this is a rollback operation
788 * Free @count clusters starting at the cluster @start_vcn in the runlist
789 * described by the vfs inode @vi.
791 * If @count is -1, all clusters from @start_vcn to the end of the runlist are
792 * deallocated. Thus, to completely free all clusters in a runlist, use
793 * @start_vcn = 0 and @count = -1.
795 * @is_rollback should always be FALSE, it is for internal use to rollback
796 * errors. You probably want to use ntfs_cluster_free() instead.
798 * Note, ntfs_cluster_free() does not modify the runlist at all, so the caller
799 * has to deal with it later.
801 * Return the number of deallocated clusters (not counting sparse ones) on
802 * success and -errno on error.
804 * Locking: - The runlist described by @vi must be locked on entry and is
805 * locked on return. Note if the runlist is locked for reading the
806 * lock may be dropped and reacquired. Note the runlist may be
807 * modified when needed runlist fragments need to be mapped.
808 * - The volume lcn bitmap must be unlocked on entry and is unlocked
810 * - This function takes the volume lcn bitmap lock for writing and
811 * modifies the bitmap contents.
813 s64
__ntfs_cluster_free(struct inode
*vi
, const VCN start_vcn
, s64 count
,
814 const BOOL write_locked
, const BOOL is_rollback
)
816 s64 delta
, to_free
, total_freed
, real_freed
;
819 struct inode
*lcnbmp_vi
;
824 ntfs_debug("Entering for i_ino 0x%lx, start_vcn 0x%llx, count "
825 "0x%llx.%s", vi
->i_ino
, (unsigned long long)start_vcn
,
826 (unsigned long long)count
,
827 is_rollback
? " (rollback)" : "");
830 lcnbmp_vi
= vol
->lcnbmp_ino
;
832 BUG_ON(start_vcn
< 0);
835 * Lock the lcn bitmap for writing but only if not rolling back. We
836 * must hold the lock all the way including through rollback otherwise
837 * rollback is not possible because once we have cleared a bit and
838 * dropped the lock, anyone could have set the bit again, thus
839 * allocating the cluster for another use.
841 if (likely(!is_rollback
))
842 down_write(&vol
->lcnbmp_lock
);
844 total_freed
= real_freed
= 0;
846 rl
= ntfs_attr_find_vcn_nolock(ni
, start_vcn
, write_locked
);
849 ntfs_error(vol
->sb
, "Failed to find first runlist "
850 "element (error %li), aborting.",
855 if (unlikely(rl
->lcn
< LCN_HOLE
)) {
857 ntfs_error(vol
->sb
, "First runlist element has "
858 "invalid lcn, aborting.");
862 /* Find the starting cluster inside the run that needs freeing. */
863 delta
= start_vcn
- rl
->vcn
;
865 /* The number of clusters in this run that need freeing. */
866 to_free
= rl
->length
- delta
;
867 if (count
>= 0 && to_free
> count
)
870 if (likely(rl
->lcn
>= 0)) {
871 /* Do the actual freeing of the clusters in this run. */
872 err
= ntfs_bitmap_set_bits_in_run(lcnbmp_vi
, rl
->lcn
+ delta
,
873 to_free
, likely(!is_rollback
) ? 0 : 1);
876 ntfs_error(vol
->sb
, "Failed to clear first run "
877 "(error %i), aborting.", err
);
880 /* We have freed @to_free real clusters. */
881 real_freed
= to_free
;
883 /* Go to the next run and adjust the number of clusters left to free. */
888 /* Keep track of the total "freed" clusters, including sparse ones. */
889 total_freed
= to_free
;
891 * Loop over the remaining runs, using @count as a capping value, and
894 for (; rl
->length
&& count
!= 0; ++rl
) {
895 if (unlikely(rl
->lcn
< LCN_HOLE
)) {
898 /* Attempt to map runlist. */
900 rl
= ntfs_attr_find_vcn_nolock(ni
, vcn
, write_locked
);
904 ntfs_error(vol
->sb
, "Failed to map "
905 "runlist fragment or "
907 "subsequent runlist "
911 if (unlikely(rl
->lcn
< LCN_HOLE
)) {
913 ntfs_error(vol
->sb
, "Runlist element "
922 /* The number of clusters in this run that need freeing. */
923 to_free
= rl
->length
;
924 if (count
>= 0 && to_free
> count
)
927 if (likely(rl
->lcn
>= 0)) {
928 /* Do the actual freeing of the clusters in the run. */
929 err
= ntfs_bitmap_set_bits_in_run(lcnbmp_vi
, rl
->lcn
,
930 to_free
, likely(!is_rollback
) ? 0 : 1);
933 ntfs_error(vol
->sb
, "Failed to clear "
937 /* We have freed @to_free real clusters. */
938 real_freed
+= to_free
;
940 /* Adjust the number of clusters left to free. */
944 /* Update the total done clusters. */
945 total_freed
+= to_free
;
947 if (likely(!is_rollback
))
948 up_write(&vol
->lcnbmp_lock
);
952 /* We are done. Return the number of actually freed clusters. */
958 /* If no real clusters were freed, no need to rollback. */
960 up_write(&vol
->lcnbmp_lock
);
964 * Attempt to rollback and if that succeeds just return the error code.
965 * If rollback fails, set the volume errors flag, emit an error
966 * message, and return the error code.
968 delta
= __ntfs_cluster_free(vi
, start_vcn
, total_freed
, write_locked
,
971 ntfs_error(vol
->sb
, "Failed to rollback (error %i). Leaving "
972 "inconsistent metadata! Unmount and run "
973 "chkdsk.", (int)delta
);
976 up_write(&vol
->lcnbmp_lock
);
977 ntfs_error(vol
->sb
, "Aborting (error %i).", err
);