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
79 * @is_extension: if 'true', this is an attribute extension
81 * Allocate @count clusters preferably starting at cluster @start_lcn or at the
82 * current allocator position if @start_lcn is -1, on the mounted ntfs volume
83 * @vol. @zone is either DATA_ZONE for allocation of normal clusters or
84 * MFT_ZONE for allocation of clusters for the master file table, i.e. the
85 * $MFT/$DATA attribute.
87 * @start_vcn specifies the vcn of the first allocated cluster. This makes
88 * merging the resulting runlist with the old runlist easier.
90 * If @is_extension is 'true', the caller is allocating clusters to extend an
91 * attribute and if it is 'false', the caller is allocating clusters to fill a
92 * hole in an attribute. Practically the difference is that if @is_extension
93 * is 'true' the returned runlist will be terminated with LCN_ENOENT and if
94 * @is_extension is 'false' the runlist will be terminated with
97 * You need to check the return value with IS_ERR(). If this is false, the
98 * function was successful and the return value is a runlist describing the
99 * allocated cluster(s). If IS_ERR() is true, the function failed and
100 * PTR_ERR() gives you the error code.
102 * Notes on the allocation algorithm
103 * =================================
105 * There are two data zones. First is the area between the end of the mft zone
106 * and the end of the volume, and second is the area between the start of the
107 * volume and the start of the mft zone. On unmodified/standard NTFS 1.x
108 * volumes, the second data zone does not exist due to the mft zone being
109 * expanded to cover the start of the volume in order to reserve space for the
110 * mft bitmap attribute.
112 * This is not the prettiest function but the complexity stems from the need of
113 * implementing the mft vs data zoned approach and from the fact that we have
114 * access to the lcn bitmap in portions of up to 8192 bytes at a time, so we
115 * need to cope with crossing over boundaries of two buffers. Further, the
116 * fact that the allocator allows for caller supplied hints as to the location
117 * of where allocation should begin and the fact that the allocator keeps track
118 * of where in the data zones the next natural allocation should occur,
119 * contribute to the complexity of the function. But it should all be
120 * worthwhile, because this allocator should: 1) be a full implementation of
121 * the MFT zone approach used by Windows NT, 2) cause reduction in
122 * fragmentation, and 3) be speedy in allocations (the code is not optimized
123 * for speed, but the algorithm is, so further speed improvements are probably
126 * FIXME: We should be monitoring cluster allocation and increment the MFT zone
127 * size dynamically but this is something for the future. We will just cause
128 * heavier fragmentation by not doing it and I am not even sure Windows would
129 * grow the MFT zone dynamically, so it might even be correct not to do this.
130 * The overhead in doing dynamic MFT zone expansion would be very large and
131 * unlikely worth the effort. (AIA)
133 * TODO: I have added in double the required zone position pointer wrap around
134 * logic which can be optimized to having only one of the two logic sets.
135 * However, having the double logic will work fine, but if we have only one of
136 * the sets and we get it wrong somewhere, then we get into trouble, so
137 * removing the duplicate logic requires _very_ careful consideration of _all_
138 * possible code paths. So at least for now, I am leaving the double logic -
139 * better safe than sorry... (AIA)
141 * Locking: - The volume lcn bitmap must be unlocked on entry and is unlocked
143 * - This function takes the volume lcn bitmap lock for writing and
144 * modifies the bitmap contents.
146 runlist_element
*ntfs_cluster_alloc(ntfs_volume
*vol
, const VCN start_vcn
,
147 const s64 count
, const LCN start_lcn
,
148 const NTFS_CLUSTER_ALLOCATION_ZONES zone
,
149 const bool is_extension
)
151 LCN zone_start
, zone_end
, bmp_pos
, bmp_initial_pos
, last_read_pos
, lcn
;
152 LCN prev_lcn
= 0, prev_run_len
= 0, mft_zone_size
;
155 struct inode
*lcnbmp_vi
;
156 runlist_element
*rl
= NULL
;
157 struct address_space
*mapping
;
158 struct page
*page
= NULL
;
160 int err
= 0, rlpos
, rlsize
, buf_size
;
161 u8 pass
, done_zones
, search_zone
, need_writeback
= 0, bit
;
163 ntfs_debug("Entering for start_vcn 0x%llx, count 0x%llx, start_lcn "
164 "0x%llx, zone %s_ZONE.", (unsigned long long)start_vcn
,
165 (unsigned long long)count
,
166 (unsigned long long)start_lcn
,
167 zone
== MFT_ZONE
? "MFT" : "DATA");
169 lcnbmp_vi
= vol
->lcnbmp_ino
;
171 BUG_ON(start_vcn
< 0);
173 BUG_ON(start_lcn
< -1);
174 BUG_ON(zone
< FIRST_ZONE
);
175 BUG_ON(zone
> LAST_ZONE
);
177 /* Return NULL if @count is zero. */
180 /* Take the lcnbmp lock for writing. */
181 down_write(&vol
->lcnbmp_lock
);
183 * If no specific @start_lcn was requested, use the current data zone
184 * position, otherwise use the requested @start_lcn but make sure it
185 * lies outside the mft zone. Also set done_zones to 0 (no zones done)
186 * and pass depending on whether we are starting inside a zone (1) or
187 * at the beginning of a zone (2). If requesting from the MFT_ZONE,
188 * we either start at the current position within the mft zone or at
189 * the specified position. If the latter is out of bounds then we start
190 * at the beginning of the MFT_ZONE.
195 * zone_start and zone_end are the current search range. search_zone
196 * is 1 for mft zone, 2 for data zone 1 (end of mft zone till end of
197 * volume) and 4 for data zone 2 (start of volume till start of mft
200 zone_start
= start_lcn
;
201 if (zone_start
< 0) {
202 if (zone
== DATA_ZONE
)
203 zone_start
= vol
->data1_zone_pos
;
205 zone_start
= vol
->mft_zone_pos
;
208 * Zone starts at beginning of volume which means a
209 * single pass is sufficient.
213 } else if (zone
== DATA_ZONE
&& zone_start
>= vol
->mft_zone_start
&&
214 zone_start
< vol
->mft_zone_end
) {
215 zone_start
= vol
->mft_zone_end
;
217 * Starting at beginning of data1_zone which means a single
218 * pass in this zone is sufficient.
221 } else if (zone
== MFT_ZONE
&& (zone_start
< vol
->mft_zone_start
||
222 zone_start
>= vol
->mft_zone_end
)) {
223 zone_start
= vol
->mft_lcn
;
224 if (!vol
->mft_zone_end
)
227 * Starting at beginning of volume which means a single pass
232 if (zone
== MFT_ZONE
) {
233 zone_end
= vol
->mft_zone_end
;
235 } else /* if (zone == DATA_ZONE) */ {
236 /* Skip searching the mft zone. */
238 if (zone_start
>= vol
->mft_zone_end
) {
239 zone_end
= vol
->nr_clusters
;
242 zone_end
= vol
->mft_zone_start
;
247 * bmp_pos is the current bit position inside the bitmap. We use
248 * bmp_initial_pos to determine whether or not to do a zone switch.
250 bmp_pos
= bmp_initial_pos
= zone_start
;
252 /* Loop until all clusters are allocated, i.e. clusters == 0. */
255 mapping
= lcnbmp_vi
->i_mapping
;
256 i_size
= i_size_read(lcnbmp_vi
);
258 ntfs_debug("Start of outer while loop: done_zones 0x%x, "
259 "search_zone %i, pass %i, zone_start 0x%llx, "
260 "zone_end 0x%llx, bmp_initial_pos 0x%llx, "
261 "bmp_pos 0x%llx, rlpos %i, rlsize %i.",
262 done_zones
, search_zone
, pass
,
263 (unsigned long long)zone_start
,
264 (unsigned long long)zone_end
,
265 (unsigned long long)bmp_initial_pos
,
266 (unsigned long long)bmp_pos
, rlpos
, rlsize
);
267 /* Loop until we run out of free clusters. */
268 last_read_pos
= bmp_pos
>> 3;
269 ntfs_debug("last_read_pos 0x%llx.",
270 (unsigned long long)last_read_pos
);
271 if (last_read_pos
> i_size
) {
272 ntfs_debug("End of attribute reached. "
273 "Skipping to zone_pass_done.");
277 if (need_writeback
) {
278 ntfs_debug("Marking page dirty.");
279 flush_dcache_page(page
);
280 set_page_dirty(page
);
283 ntfs_unmap_page(page
);
285 page
= ntfs_map_page(mapping
, last_read_pos
>>
289 ntfs_error(vol
->sb
, "Failed to map page.");
292 buf_size
= last_read_pos
& ~PAGE_CACHE_MASK
;
293 buf
= page_address(page
) + buf_size
;
294 buf_size
= PAGE_CACHE_SIZE
- buf_size
;
295 if (unlikely(last_read_pos
+ buf_size
> i_size
))
296 buf_size
= i_size
- last_read_pos
;
300 ntfs_debug("Before inner while loop: buf_size %i, lcn 0x%llx, "
301 "bmp_pos 0x%llx, need_writeback %i.", buf_size
,
302 (unsigned long long)lcn
,
303 (unsigned long long)bmp_pos
, need_writeback
);
304 while (lcn
< buf_size
&& lcn
+ bmp_pos
< zone_end
) {
305 byte
= buf
+ (lcn
>> 3);
306 ntfs_debug("In inner while loop: buf_size %i, "
307 "lcn 0x%llx, bmp_pos 0x%llx, "
308 "need_writeback %i, byte ofs 0x%x, "
309 "*byte 0x%x.", buf_size
,
310 (unsigned long long)lcn
,
311 (unsigned long long)bmp_pos
,
313 (unsigned int)(lcn
>> 3),
314 (unsigned int)*byte
);
315 /* Skip full bytes. */
317 lcn
= (lcn
+ 8) & ~(LCN
)7;
318 ntfs_debug("Continuing while loop 1.");
321 bit
= 1 << (lcn
& 7);
322 ntfs_debug("bit 0x%x.", bit
);
323 /* If the bit is already set, go onto the next one. */
326 ntfs_debug("Continuing while loop 2.");
330 * Allocate more memory if needed, including space for
331 * the terminator element.
332 * ntfs_malloc_nofs() operates on whole pages only.
334 if ((rlpos
+ 2) * sizeof(*rl
) > rlsize
) {
335 runlist_element
*rl2
;
337 ntfs_debug("Reallocating memory.");
339 ntfs_debug("First free bit is at LCN "
343 rl2
= ntfs_malloc_nofs(rlsize
+ (int)PAGE_SIZE
);
344 if (unlikely(!rl2
)) {
346 ntfs_error(vol
->sb
, "Failed to "
350 memcpy(rl2
, rl
, rlsize
);
354 ntfs_debug("Reallocated memory, rlsize 0x%x.",
357 /* Allocate the bitmap bit. */
359 /* We need to write this bitmap page to disk. */
361 ntfs_debug("*byte 0x%x, need_writeback is set.",
362 (unsigned int)*byte
);
364 * Coalesce with previous run if adjacent LCNs.
365 * Otherwise, append a new run.
367 ntfs_debug("Adding run (lcn 0x%llx, len 0x%llx), "
368 "prev_lcn 0x%llx, lcn 0x%llx, "
369 "bmp_pos 0x%llx, prev_run_len 0x%llx, "
371 (unsigned long long)(lcn
+ bmp_pos
),
372 1ULL, (unsigned long long)prev_lcn
,
373 (unsigned long long)lcn
,
374 (unsigned long long)bmp_pos
,
375 (unsigned long long)prev_run_len
,
377 if (prev_lcn
== lcn
+ bmp_pos
- prev_run_len
&& rlpos
) {
378 ntfs_debug("Coalescing to run (lcn 0x%llx, "
383 rl
[rlpos
- 1].length
);
384 rl
[rlpos
- 1].length
= ++prev_run_len
;
385 ntfs_debug("Run now (lcn 0x%llx, len 0x%llx), "
386 "prev_run_len 0x%llx.",
390 rl
[rlpos
- 1].length
,
395 ntfs_debug("Adding new run, (previous "
401 rl
[rlpos
- 1].length
);
402 rl
[rlpos
].vcn
= rl
[rlpos
- 1].vcn
+
405 ntfs_debug("Adding new run, is first "
407 rl
[rlpos
].vcn
= start_vcn
;
409 rl
[rlpos
].lcn
= prev_lcn
= lcn
+ bmp_pos
;
410 rl
[rlpos
].length
= prev_run_len
= 1;
417 * Update the current zone position. Positions
418 * of already scanned zones have been updated
419 * during the respective zone switches.
421 tc
= lcn
+ bmp_pos
+ 1;
422 ntfs_debug("Done. Updating current zone "
423 "position, tc 0x%llx, "
425 (unsigned long long)tc
,
427 switch (search_zone
) {
429 ntfs_debug("Before checks, "
434 if (tc
>= vol
->mft_zone_end
) {
437 if (!vol
->mft_zone_end
)
438 vol
->mft_zone_pos
= 0;
439 } else if ((bmp_initial_pos
>=
441 tc
> vol
->mft_zone_pos
)
442 && tc
>= vol
->mft_lcn
)
443 vol
->mft_zone_pos
= tc
;
444 ntfs_debug("After checks, "
451 ntfs_debug("Before checks, "
452 "vol->data1_zone_pos "
455 vol
->data1_zone_pos
);
456 if (tc
>= vol
->nr_clusters
)
457 vol
->data1_zone_pos
=
459 else if ((bmp_initial_pos
>=
460 vol
->data1_zone_pos
||
461 tc
> vol
->data1_zone_pos
)
462 && tc
>= vol
->mft_zone_end
)
463 vol
->data1_zone_pos
= tc
;
464 ntfs_debug("After checks, "
465 "vol->data1_zone_pos "
468 vol
->data1_zone_pos
);
471 ntfs_debug("Before checks, "
472 "vol->data2_zone_pos "
475 vol
->data2_zone_pos
);
476 if (tc
>= vol
->mft_zone_start
)
477 vol
->data2_zone_pos
= 0;
478 else if (bmp_initial_pos
>=
479 vol
->data2_zone_pos
||
480 tc
> vol
->data2_zone_pos
)
481 vol
->data2_zone_pos
= tc
;
482 ntfs_debug("After checks, "
483 "vol->data2_zone_pos "
486 vol
->data2_zone_pos
);
491 ntfs_debug("Finished. Going to out.");
497 ntfs_debug("After inner while loop: buf_size 0x%x, lcn "
498 "0x%llx, bmp_pos 0x%llx, need_writeback %i.",
499 buf_size
, (unsigned long long)lcn
,
500 (unsigned long long)bmp_pos
, need_writeback
);
501 if (bmp_pos
< zone_end
) {
502 ntfs_debug("Continuing outer while loop, "
503 "bmp_pos 0x%llx, zone_end 0x%llx.",
504 (unsigned long long)bmp_pos
,
505 (unsigned long long)zone_end
);
508 zone_pass_done
: /* Finished with the current zone pass. */
509 ntfs_debug("At zone_pass_done, pass %i.", pass
);
512 * Now do pass 2, scanning the first part of the zone
513 * we omitted in pass 1.
516 zone_end
= zone_start
;
517 switch (search_zone
) {
518 case 1: /* mft_zone */
519 zone_start
= vol
->mft_zone_start
;
521 case 2: /* data1_zone */
522 zone_start
= vol
->mft_zone_end
;
524 case 4: /* data2_zone */
531 if (zone_end
< zone_start
)
532 zone_end
= zone_start
;
533 bmp_pos
= zone_start
;
534 ntfs_debug("Continuing outer while loop, pass 2, "
535 "zone_start 0x%llx, zone_end 0x%llx, "
537 (unsigned long long)zone_start
,
538 (unsigned long long)zone_end
,
539 (unsigned long long)bmp_pos
);
543 ntfs_debug("At done_zones_check, search_zone %i, done_zones "
544 "before 0x%x, done_zones after 0x%x.",
545 search_zone
, done_zones
,
546 done_zones
| search_zone
);
547 done_zones
|= search_zone
;
548 if (done_zones
< 7) {
549 ntfs_debug("Switching zone.");
550 /* Now switch to the next zone we haven't done yet. */
552 switch (search_zone
) {
554 ntfs_debug("Switching from mft zone to data1 "
556 /* Update mft zone position. */
560 ntfs_debug("Before checks, "
565 tc
= rl
[rlpos
- 1].lcn
+
566 rl
[rlpos
- 1].length
;
567 if (tc
>= vol
->mft_zone_end
) {
570 if (!vol
->mft_zone_end
)
571 vol
->mft_zone_pos
= 0;
572 } else if ((bmp_initial_pos
>=
574 tc
> vol
->mft_zone_pos
)
575 && tc
>= vol
->mft_lcn
)
576 vol
->mft_zone_pos
= tc
;
577 ntfs_debug("After checks, "
583 /* Switch from mft zone to data1 zone. */
584 switch_to_data1_zone
: search_zone
= 2;
585 zone_start
= bmp_initial_pos
=
587 zone_end
= vol
->nr_clusters
;
588 if (zone_start
== vol
->mft_zone_end
)
590 if (zone_start
>= zone_end
) {
591 vol
->data1_zone_pos
= zone_start
=
597 ntfs_debug("Switching from data1 zone to "
599 /* Update data1 zone position. */
603 ntfs_debug("Before checks, "
604 "vol->data1_zone_pos "
607 vol
->data1_zone_pos
);
608 tc
= rl
[rlpos
- 1].lcn
+
609 rl
[rlpos
- 1].length
;
610 if (tc
>= vol
->nr_clusters
)
611 vol
->data1_zone_pos
=
613 else if ((bmp_initial_pos
>=
614 vol
->data1_zone_pos
||
615 tc
> vol
->data1_zone_pos
)
616 && tc
>= vol
->mft_zone_end
)
617 vol
->data1_zone_pos
= tc
;
618 ntfs_debug("After checks, "
619 "vol->data1_zone_pos "
622 vol
->data1_zone_pos
);
624 /* Switch from data1 zone to data2 zone. */
626 zone_start
= bmp_initial_pos
=
628 zone_end
= vol
->mft_zone_start
;
631 if (zone_start
>= zone_end
) {
632 vol
->data2_zone_pos
= zone_start
=
638 ntfs_debug("Switching from data2 zone to "
640 /* Update data2 zone position. */
644 ntfs_debug("Before checks, "
645 "vol->data2_zone_pos "
648 vol
->data2_zone_pos
);
649 tc
= rl
[rlpos
- 1].lcn
+
650 rl
[rlpos
- 1].length
;
651 if (tc
>= vol
->mft_zone_start
)
652 vol
->data2_zone_pos
= 0;
653 else if (bmp_initial_pos
>=
654 vol
->data2_zone_pos
||
655 tc
> vol
->data2_zone_pos
)
656 vol
->data2_zone_pos
= tc
;
657 ntfs_debug("After checks, "
658 "vol->data2_zone_pos "
661 vol
->data2_zone_pos
);
663 /* Switch from data2 zone to data1 zone. */
664 goto switch_to_data1_zone
;
668 ntfs_debug("After zone switch, search_zone %i, "
669 "pass %i, bmp_initial_pos 0x%llx, "
670 "zone_start 0x%llx, zone_end 0x%llx.",
672 (unsigned long long)bmp_initial_pos
,
673 (unsigned long long)zone_start
,
674 (unsigned long long)zone_end
);
675 bmp_pos
= zone_start
;
676 if (zone_start
== zone_end
) {
677 ntfs_debug("Empty zone, going to "
678 "done_zones_check.");
679 /* Empty zone. Don't bother searching it. */
680 goto done_zones_check
;
682 ntfs_debug("Continuing outer while loop.");
684 } /* done_zones == 7 */
685 ntfs_debug("All zones are finished.");
687 * All zones are finished! If DATA_ZONE, shrink mft zone. If
688 * MFT_ZONE, we have really run out of space.
690 mft_zone_size
= vol
->mft_zone_end
- vol
->mft_zone_start
;
691 ntfs_debug("vol->mft_zone_start 0x%llx, vol->mft_zone_end "
692 "0x%llx, mft_zone_size 0x%llx.",
693 (unsigned long long)vol
->mft_zone_start
,
694 (unsigned long long)vol
->mft_zone_end
,
695 (unsigned long long)mft_zone_size
);
696 if (zone
== MFT_ZONE
|| mft_zone_size
<= 0) {
697 ntfs_debug("No free clusters left, going to out.");
698 /* Really no more space left on device. */
701 } /* zone == DATA_ZONE && mft_zone_size > 0 */
702 ntfs_debug("Shrinking mft zone.");
703 zone_end
= vol
->mft_zone_end
;
705 if (mft_zone_size
> 0)
706 vol
->mft_zone_end
= vol
->mft_zone_start
+ mft_zone_size
;
707 else /* mft zone and data2 zone no longer exist. */
708 vol
->data2_zone_pos
= vol
->mft_zone_start
=
709 vol
->mft_zone_end
= 0;
710 if (vol
->mft_zone_pos
>= vol
->mft_zone_end
) {
711 vol
->mft_zone_pos
= vol
->mft_lcn
;
712 if (!vol
->mft_zone_end
)
713 vol
->mft_zone_pos
= 0;
715 bmp_pos
= zone_start
= bmp_initial_pos
=
716 vol
->data1_zone_pos
= vol
->mft_zone_end
;
720 ntfs_debug("After shrinking mft zone, mft_zone_size 0x%llx, "
721 "vol->mft_zone_start 0x%llx, "
722 "vol->mft_zone_end 0x%llx, "
723 "vol->mft_zone_pos 0x%llx, search_zone 2, "
724 "pass 2, dones_zones 0x%x, zone_start 0x%llx, "
725 "zone_end 0x%llx, vol->data1_zone_pos 0x%llx, "
726 "continuing outer while loop.",
727 (unsigned long long)mft_zone_size
,
728 (unsigned long long)vol
->mft_zone_start
,
729 (unsigned long long)vol
->mft_zone_end
,
730 (unsigned long long)vol
->mft_zone_pos
,
731 done_zones
, (unsigned long long)zone_start
,
732 (unsigned long long)zone_end
,
733 (unsigned long long)vol
->data1_zone_pos
);
735 ntfs_debug("After outer while loop.");
737 ntfs_debug("At out.");
738 /* Add runlist terminator element. */
740 rl
[rlpos
].vcn
= rl
[rlpos
- 1].vcn
+ rl
[rlpos
- 1].length
;
741 rl
[rlpos
].lcn
= is_extension
? LCN_ENOENT
: LCN_RL_NOT_MAPPED
;
742 rl
[rlpos
].length
= 0;
744 if (likely(page
&& !IS_ERR(page
))) {
745 if (need_writeback
) {
746 ntfs_debug("Marking page dirty.");
747 flush_dcache_page(page
);
748 set_page_dirty(page
);
751 ntfs_unmap_page(page
);
754 up_write(&vol
->lcnbmp_lock
);
758 ntfs_error(vol
->sb
, "Failed to allocate clusters, aborting "
764 ntfs_debug("Not enough space to complete allocation, "
765 "err -ENOSPC, first free lcn 0x%llx, "
766 "could allocate up to 0x%llx "
768 (unsigned long long)rl
[0].lcn
,
769 (unsigned long long)(count
- clusters
));
770 /* Deallocate all allocated clusters. */
771 ntfs_debug("Attempting rollback...");
772 err2
= ntfs_cluster_free_from_rl_nolock(vol
, rl
);
774 ntfs_error(vol
->sb
, "Failed to rollback (error %i). "
775 "Leaving inconsistent metadata! "
776 "Unmount and run chkdsk.", err2
);
779 /* Free the runlist. */
781 } else if (err
== -ENOSPC
)
782 ntfs_debug("No space left at all, err = -ENOSPC, first free "
784 (long long)vol
->data1_zone_pos
);
785 up_write(&vol
->lcnbmp_lock
);
790 * __ntfs_cluster_free - free clusters on an ntfs volume
791 * @ni: ntfs inode whose runlist describes the clusters to free
792 * @start_vcn: vcn in the runlist of @ni at which to start freeing clusters
793 * @count: number of clusters to free or -1 for all clusters
794 * @ctx: active attribute search context if present or NULL if not
795 * @is_rollback: true if this is a rollback operation
797 * Free @count clusters starting at the cluster @start_vcn in the runlist
798 * described by the vfs inode @ni.
800 * If @count is -1, all clusters from @start_vcn to the end of the runlist are
801 * deallocated. Thus, to completely free all clusters in a runlist, use
802 * @start_vcn = 0 and @count = -1.
804 * If @ctx is specified, it is an active search context of @ni and its base mft
805 * record. This is needed when __ntfs_cluster_free() encounters unmapped
806 * runlist fragments and allows their mapping. If you do not have the mft
807 * record mapped, you can specify @ctx as NULL and __ntfs_cluster_free() will
808 * perform the necessary mapping and unmapping.
810 * Note, __ntfs_cluster_free() saves the state of @ctx on entry and restores it
811 * before returning. Thus, @ctx will be left pointing to the same attribute on
812 * return as on entry. However, the actual pointers in @ctx may point to
813 * different memory locations on return, so you must remember to reset any
814 * cached pointers from the @ctx, i.e. after the call to __ntfs_cluster_free(),
815 * you will probably want to do:
818 * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
819 * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
821 * @is_rollback should always be 'false', it is for internal use to rollback
822 * errors. You probably want to use ntfs_cluster_free() instead.
824 * Note, __ntfs_cluster_free() does not modify the runlist, so you have to
825 * remove from the runlist or mark sparse the freed runs later.
827 * Return the number of deallocated clusters (not counting sparse ones) on
828 * success and -errno on error.
830 * WARNING: If @ctx is supplied, regardless of whether success or failure is
831 * returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
832 * is no longer valid, i.e. you need to either call
833 * ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
834 * In that case PTR_ERR(@ctx->mrec) will give you the error code for
835 * why the mapping of the old inode failed.
837 * Locking: - The runlist described by @ni must be locked for writing on entry
838 * and is locked on return. Note the runlist may be modified when
839 * needed runlist fragments need to be mapped.
840 * - The volume lcn bitmap must be unlocked on entry and is unlocked
842 * - This function takes the volume lcn bitmap lock for writing and
843 * modifies the bitmap contents.
844 * - If @ctx is NULL, the base mft record of @ni must not be mapped on
845 * entry and it will be left unmapped on return.
846 * - If @ctx is not NULL, the base mft record must be mapped on entry
847 * and it will be left mapped on return.
849 s64
__ntfs_cluster_free(ntfs_inode
*ni
, const VCN start_vcn
, s64 count
,
850 ntfs_attr_search_ctx
*ctx
, const bool is_rollback
)
852 s64 delta
, to_free
, total_freed
, real_freed
;
854 struct inode
*lcnbmp_vi
;
859 ntfs_debug("Entering for i_ino 0x%lx, start_vcn 0x%llx, count "
860 "0x%llx.%s", ni
->mft_no
, (unsigned long long)start_vcn
,
861 (unsigned long long)count
,
862 is_rollback
? " (rollback)" : "");
864 lcnbmp_vi
= vol
->lcnbmp_ino
;
866 BUG_ON(start_vcn
< 0);
869 * Lock the lcn bitmap for writing but only if not rolling back. We
870 * must hold the lock all the way including through rollback otherwise
871 * rollback is not possible because once we have cleared a bit and
872 * dropped the lock, anyone could have set the bit again, thus
873 * allocating the cluster for another use.
875 if (likely(!is_rollback
))
876 down_write(&vol
->lcnbmp_lock
);
878 total_freed
= real_freed
= 0;
880 rl
= ntfs_attr_find_vcn_nolock(ni
, start_vcn
, ctx
);
883 ntfs_error(vol
->sb
, "Failed to find first runlist "
884 "element (error %li), aborting.",
889 if (unlikely(rl
->lcn
< LCN_HOLE
)) {
891 ntfs_error(vol
->sb
, "First runlist element has "
892 "invalid lcn, aborting.");
896 /* Find the starting cluster inside the run that needs freeing. */
897 delta
= start_vcn
- rl
->vcn
;
899 /* The number of clusters in this run that need freeing. */
900 to_free
= rl
->length
- delta
;
901 if (count
>= 0 && to_free
> count
)
904 if (likely(rl
->lcn
>= 0)) {
905 /* Do the actual freeing of the clusters in this run. */
906 err
= ntfs_bitmap_set_bits_in_run(lcnbmp_vi
, rl
->lcn
+ delta
,
907 to_free
, likely(!is_rollback
) ? 0 : 1);
910 ntfs_error(vol
->sb
, "Failed to clear first run "
911 "(error %i), aborting.", err
);
914 /* We have freed @to_free real clusters. */
915 real_freed
= to_free
;
917 /* Go to the next run and adjust the number of clusters left to free. */
922 /* Keep track of the total "freed" clusters, including sparse ones. */
923 total_freed
= to_free
;
925 * Loop over the remaining runs, using @count as a capping value, and
928 for (; rl
->length
&& count
!= 0; ++rl
) {
929 if (unlikely(rl
->lcn
< LCN_HOLE
)) {
932 /* Attempt to map runlist. */
934 rl
= ntfs_attr_find_vcn_nolock(ni
, vcn
, ctx
);
938 ntfs_error(vol
->sb
, "Failed to map "
939 "runlist fragment or "
941 "subsequent runlist "
945 if (unlikely(rl
->lcn
< LCN_HOLE
)) {
947 ntfs_error(vol
->sb
, "Runlist element "
956 /* The number of clusters in this run that need freeing. */
957 to_free
= rl
->length
;
958 if (count
>= 0 && to_free
> count
)
961 if (likely(rl
->lcn
>= 0)) {
962 /* Do the actual freeing of the clusters in the run. */
963 err
= ntfs_bitmap_set_bits_in_run(lcnbmp_vi
, rl
->lcn
,
964 to_free
, likely(!is_rollback
) ? 0 : 1);
967 ntfs_error(vol
->sb
, "Failed to clear "
971 /* We have freed @to_free real clusters. */
972 real_freed
+= to_free
;
974 /* Adjust the number of clusters left to free. */
978 /* Update the total done clusters. */
979 total_freed
+= to_free
;
981 if (likely(!is_rollback
))
982 up_write(&vol
->lcnbmp_lock
);
986 /* We are done. Return the number of actually freed clusters. */
992 /* If no real clusters were freed, no need to rollback. */
994 up_write(&vol
->lcnbmp_lock
);
998 * Attempt to rollback and if that succeeds just return the error code.
999 * If rollback fails, set the volume errors flag, emit an error
1000 * message, and return the error code.
1002 delta
= __ntfs_cluster_free(ni
, start_vcn
, total_freed
, ctx
, true);
1004 ntfs_error(vol
->sb
, "Failed to rollback (error %i). Leaving "
1005 "inconsistent metadata! Unmount and run "
1006 "chkdsk.", (int)delta
);
1009 up_write(&vol
->lcnbmp_lock
);
1010 ntfs_error(vol
->sb
, "Aborting (error %i).", err
);
1014 #endif /* NTFS_RW */