blob: 1711b710b641f40a0c8c5690a94c9681f2b5fa92 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * lcnalloc.c - Cluster (de)allocation code. Part of the Linux-NTFS project.
3 *
Anton Altaparmakov43b01fd2005-03-09 15:18:43 +00004 * Copyright (c) 2004-2005 Anton Altaparmakov
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 *
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.
10 *
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.
15 *
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
20 */
21
22#ifdef NTFS_RW
23
24#include <linux/pagemap.h>
25
26#include "lcnalloc.h"
27#include "debug.h"
28#include "bitmap.h"
29#include "inode.h"
30#include "volume.h"
31#include "attrib.h"
32#include "malloc.h"
33#include "aops.h"
34#include "ntfs.h"
35
36/**
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
40 *
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
43 * freed.
44 *
45 * Return 0 on success and -errno on error.
46 *
47 * Locking: - The volume lcn bitmap must be locked for writing on entry and is
48 * left locked on return.
49 */
50int ntfs_cluster_free_from_rl_nolock(ntfs_volume *vol,
51 const runlist_element *rl)
52{
53 struct inode *lcnbmp_vi = vol->lcnbmp_ino;
54 int ret = 0;
55
56 ntfs_debug("Entering.");
Anton Altaparmakovbbf18132005-09-08 21:09:06 +010057 if (!rl)
58 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 for (; rl->length; rl++) {
60 int err;
61
62 if (rl->lcn < 0)
63 continue;
64 err = ntfs_bitmap_clear_run(lcnbmp_vi, rl->lcn, rl->length);
Anton Altaparmakov43b01fd2005-03-09 15:18:43 +000065 if (unlikely(err && (!ret || ret == -ENOMEM) && ret != err))
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 ret = err;
67 }
68 ntfs_debug("Done.");
69 return ret;
70}
71
72/**
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
Richard Knutssonc49c3112006-09-30 23:27:12 -070079 * @is_extension: if 'true', this is an attribute extension
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 *
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.
86 *
87 * @start_vcn specifies the vcn of the first allocated cluster. This makes
88 * merging the resulting runlist with the old runlist easier.
89 *
Richard Knutssonc49c3112006-09-30 23:27:12 -070090 * 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
Anton Altaparmakovfc0fa7d2005-10-04 14:36:56 +010092 * hole in an attribute. Practically the difference is that if @is_extension
Richard Knutssonc49c3112006-09-30 23:27:12 -070093 * is 'true' the returned runlist will be terminated with LCN_ENOENT and if
94 * @is_extension is 'false' the runlist will be terminated with
Anton Altaparmakovfc0fa7d2005-10-04 14:36:56 +010095 * LCN_RL_NOT_MAPPED.
96 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 * 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.
101 *
102 * Notes on the allocation algorithm
103 * =================================
104 *
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.
111 *
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
124 * possible).
125 *
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)
132 *
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)
140 *
141 * Locking: - The volume lcn bitmap must be unlocked on entry and is unlocked
142 * on return.
143 * - This function takes the volume lcn bitmap lock for writing and
144 * modifies the bitmap contents.
145 */
146runlist_element *ntfs_cluster_alloc(ntfs_volume *vol, const VCN start_vcn,
147 const s64 count, const LCN start_lcn,
Anton Altaparmakovfc0fa7d2005-10-04 14:36:56 +0100148 const NTFS_CLUSTER_ALLOCATION_ZONES zone,
Richard Knutssonc49c3112006-09-30 23:27:12 -0700149 const bool is_extension)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150{
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;
153 s64 clusters;
Anton Altaparmakovdb30d162004-11-11 12:42:00 +0000154 loff_t i_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 struct inode *lcnbmp_vi;
156 runlist_element *rl = NULL;
157 struct address_space *mapping;
158 struct page *page = NULL;
159 u8 *buf, *byte;
160 int err = 0, rlpos, rlsize, buf_size;
161 u8 pass, done_zones, search_zone, need_writeback = 0, bit;
162
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");
168 BUG_ON(!vol);
169 lcnbmp_vi = vol->lcnbmp_ino;
170 BUG_ON(!lcnbmp_vi);
171 BUG_ON(start_vcn < 0);
172 BUG_ON(count < 0);
173 BUG_ON(start_lcn < -1);
174 BUG_ON(zone < FIRST_ZONE);
175 BUG_ON(zone > LAST_ZONE);
176
Anton Altaparmakovbbf18132005-09-08 21:09:06 +0100177 /* Return NULL if @count is zero. */
178 if (!count)
179 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 /* Take the lcnbmp lock for writing. */
181 down_write(&vol->lcnbmp_lock);
182 /*
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.
191 */
192 done_zones = 0;
193 pass = 1;
194 /*
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
198 * zone).
199 */
200 zone_start = start_lcn;
201 if (zone_start < 0) {
202 if (zone == DATA_ZONE)
203 zone_start = vol->data1_zone_pos;
204 else
205 zone_start = vol->mft_zone_pos;
206 if (!zone_start) {
207 /*
208 * Zone starts at beginning of volume which means a
209 * single pass is sufficient.
210 */
211 pass = 2;
212 }
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;
216 /*
217 * Starting at beginning of data1_zone which means a single
218 * pass in this zone is sufficient.
219 */
220 pass = 2;
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)
225 zone_start = 0;
226 /*
227 * Starting at beginning of volume which means a single pass
228 * is sufficient.
229 */
230 pass = 2;
231 }
232 if (zone == MFT_ZONE) {
233 zone_end = vol->mft_zone_end;
234 search_zone = 1;
235 } else /* if (zone == DATA_ZONE) */ {
236 /* Skip searching the mft zone. */
237 done_zones |= 1;
238 if (zone_start >= vol->mft_zone_end) {
239 zone_end = vol->nr_clusters;
240 search_zone = 2;
241 } else {
242 zone_end = vol->mft_zone_start;
243 search_zone = 4;
244 }
245 }
246 /*
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.
249 */
250 bmp_pos = bmp_initial_pos = zone_start;
251
252 /* Loop until all clusters are allocated, i.e. clusters == 0. */
253 clusters = count;
254 rlpos = rlsize = 0;
255 mapping = lcnbmp_vi->i_mapping;
Anton Altaparmakovdb30d162004-11-11 12:42:00 +0000256 i_size = i_size_read(lcnbmp_vi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 while (1) {
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);
Anton Altaparmakovdb30d162004-11-11 12:42:00 +0000271 if (last_read_pos > i_size) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 ntfs_debug("End of attribute reached. "
273 "Skipping to zone_pass_done.");
274 goto zone_pass_done;
275 }
276 if (likely(page)) {
277 if (need_writeback) {
278 ntfs_debug("Marking page dirty.");
279 flush_dcache_page(page);
280 set_page_dirty(page);
281 need_writeback = 0;
282 }
283 ntfs_unmap_page(page);
284 }
285 page = ntfs_map_page(mapping, last_read_pos >>
286 PAGE_CACHE_SHIFT);
287 if (IS_ERR(page)) {
288 err = PTR_ERR(page);
289 ntfs_error(vol->sb, "Failed to map page.");
290 goto out;
291 }
292 buf_size = last_read_pos & ~PAGE_CACHE_MASK;
293 buf = page_address(page) + buf_size;
294 buf_size = PAGE_CACHE_SIZE - buf_size;
Anton Altaparmakovdb30d162004-11-11 12:42:00 +0000295 if (unlikely(last_read_pos + buf_size > i_size))
296 buf_size = i_size - last_read_pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 buf_size <<= 3;
298 lcn = bmp_pos & 7;
Anton Altaparmakov3bd1f4a2005-06-25 16:51:58 +0100299 bmp_pos &= ~(LCN)7;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 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,
312 need_writeback,
313 (unsigned int)(lcn >> 3),
314 (unsigned int)*byte);
315 /* Skip full bytes. */
316 if (*byte == 0xff) {
Anton Altaparmakov3bd1f4a2005-06-25 16:51:58 +0100317 lcn = (lcn + 8) & ~(LCN)7;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 ntfs_debug("Continuing while loop 1.");
319 continue;
320 }
321 bit = 1 << (lcn & 7);
Anton Altaparmakovfc0fa7d2005-10-04 14:36:56 +0100322 ntfs_debug("bit 0x%x.", bit);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 /* If the bit is already set, go onto the next one. */
324 if (*byte & bit) {
325 lcn++;
326 ntfs_debug("Continuing while loop 2.");
327 continue;
328 }
329 /*
330 * Allocate more memory if needed, including space for
331 * the terminator element.
332 * ntfs_malloc_nofs() operates on whole pages only.
333 */
334 if ((rlpos + 2) * sizeof(*rl) > rlsize) {
335 runlist_element *rl2;
336
337 ntfs_debug("Reallocating memory.");
338 if (!rl)
339 ntfs_debug("First free bit is at LCN "
340 "0x%llx.",
341 (unsigned long long)
342 (lcn + bmp_pos));
343 rl2 = ntfs_malloc_nofs(rlsize + (int)PAGE_SIZE);
344 if (unlikely(!rl2)) {
345 err = -ENOMEM;
346 ntfs_error(vol->sb, "Failed to "
347 "allocate memory.");
348 goto out;
349 }
350 memcpy(rl2, rl, rlsize);
351 ntfs_free(rl);
352 rl = rl2;
353 rlsize += PAGE_SIZE;
354 ntfs_debug("Reallocated memory, rlsize 0x%x.",
355 rlsize);
356 }
357 /* Allocate the bitmap bit. */
358 *byte |= bit;
359 /* We need to write this bitmap page to disk. */
360 need_writeback = 1;
361 ntfs_debug("*byte 0x%x, need_writeback is set.",
362 (unsigned int)*byte);
363 /*
364 * Coalesce with previous run if adjacent LCNs.
365 * Otherwise, append a new run.
366 */
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, "
370 "rlpos %i.",
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,
376 rlpos);
377 if (prev_lcn == lcn + bmp_pos - prev_run_len && rlpos) {
378 ntfs_debug("Coalescing to run (lcn 0x%llx, "
379 "len 0x%llx).",
380 (unsigned long long)
381 rl[rlpos - 1].lcn,
382 (unsigned long long)
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.",
387 (unsigned long long)
388 rl[rlpos - 1].lcn,
389 (unsigned long long)
390 rl[rlpos - 1].length,
391 (unsigned long long)
392 prev_run_len);
393 } else {
394 if (likely(rlpos)) {
395 ntfs_debug("Adding new run, (previous "
396 "run lcn 0x%llx, "
397 "len 0x%llx).",
398 (unsigned long long)
399 rl[rlpos - 1].lcn,
400 (unsigned long long)
401 rl[rlpos - 1].length);
402 rl[rlpos].vcn = rl[rlpos - 1].vcn +
403 prev_run_len;
404 } else {
405 ntfs_debug("Adding new run, is first "
406 "run.");
407 rl[rlpos].vcn = start_vcn;
408 }
409 rl[rlpos].lcn = prev_lcn = lcn + bmp_pos;
410 rl[rlpos].length = prev_run_len = 1;
411 rlpos++;
412 }
413 /* Done? */
414 if (!--clusters) {
415 LCN tc;
416 /*
417 * Update the current zone position. Positions
418 * of already scanned zones have been updated
419 * during the respective zone switches.
420 */
421 tc = lcn + bmp_pos + 1;
422 ntfs_debug("Done. Updating current zone "
423 "position, tc 0x%llx, "
424 "search_zone %i.",
425 (unsigned long long)tc,
426 search_zone);
427 switch (search_zone) {
428 case 1:
429 ntfs_debug("Before checks, "
430 "vol->mft_zone_pos "
431 "0x%llx.",
432 (unsigned long long)
433 vol->mft_zone_pos);
434 if (tc >= vol->mft_zone_end) {
435 vol->mft_zone_pos =
436 vol->mft_lcn;
437 if (!vol->mft_zone_end)
438 vol->mft_zone_pos = 0;
439 } else if ((bmp_initial_pos >=
440 vol->mft_zone_pos ||
441 tc > vol->mft_zone_pos)
442 && tc >= vol->mft_lcn)
443 vol->mft_zone_pos = tc;
444 ntfs_debug("After checks, "
445 "vol->mft_zone_pos "
446 "0x%llx.",
447 (unsigned long long)
448 vol->mft_zone_pos);
449 break;
450 case 2:
451 ntfs_debug("Before checks, "
452 "vol->data1_zone_pos "
453 "0x%llx.",
454 (unsigned long long)
455 vol->data1_zone_pos);
456 if (tc >= vol->nr_clusters)
457 vol->data1_zone_pos =
458 vol->mft_zone_end;
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 "
466 "0x%llx.",
467 (unsigned long long)
468 vol->data1_zone_pos);
469 break;
470 case 4:
471 ntfs_debug("Before checks, "
472 "vol->data2_zone_pos "
473 "0x%llx.",
474 (unsigned long long)
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 "
484 "0x%llx.",
485 (unsigned long long)
486 vol->data2_zone_pos);
487 break;
488 default:
489 BUG();
490 }
491 ntfs_debug("Finished. Going to out.");
492 goto out;
493 }
494 lcn++;
495 }
496 bmp_pos += buf_size;
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);
506 continue;
507 }
508zone_pass_done: /* Finished with the current zone pass. */
509 ntfs_debug("At zone_pass_done, pass %i.", pass);
510 if (pass == 1) {
511 /*
512 * Now do pass 2, scanning the first part of the zone
513 * we omitted in pass 1.
514 */
515 pass = 2;
516 zone_end = zone_start;
517 switch (search_zone) {
518 case 1: /* mft_zone */
519 zone_start = vol->mft_zone_start;
520 break;
521 case 2: /* data1_zone */
522 zone_start = vol->mft_zone_end;
523 break;
524 case 4: /* data2_zone */
525 zone_start = 0;
526 break;
527 default:
528 BUG();
529 }
530 /* Sanity check. */
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, "
536 "bmp_pos 0x%llx.",
537 (unsigned long long)zone_start,
538 (unsigned long long)zone_end,
539 (unsigned long long)bmp_pos);
540 continue;
541 } /* pass == 2 */
542done_zones_check:
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. */
551 pass = 1;
552 switch (search_zone) {
553 case 1:
554 ntfs_debug("Switching from mft zone to data1 "
555 "zone.");
556 /* Update mft zone position. */
557 if (rlpos) {
558 LCN tc;
559
560 ntfs_debug("Before checks, "
561 "vol->mft_zone_pos "
562 "0x%llx.",
563 (unsigned long long)
564 vol->mft_zone_pos);
565 tc = rl[rlpos - 1].lcn +
566 rl[rlpos - 1].length;
567 if (tc >= vol->mft_zone_end) {
568 vol->mft_zone_pos =
569 vol->mft_lcn;
570 if (!vol->mft_zone_end)
571 vol->mft_zone_pos = 0;
572 } else if ((bmp_initial_pos >=
573 vol->mft_zone_pos ||
574 tc > vol->mft_zone_pos)
575 && tc >= vol->mft_lcn)
576 vol->mft_zone_pos = tc;
577 ntfs_debug("After checks, "
578 "vol->mft_zone_pos "
579 "0x%llx.",
580 (unsigned long long)
581 vol->mft_zone_pos);
582 }
583 /* Switch from mft zone to data1 zone. */
584switch_to_data1_zone: search_zone = 2;
585 zone_start = bmp_initial_pos =
586 vol->data1_zone_pos;
587 zone_end = vol->nr_clusters;
588 if (zone_start == vol->mft_zone_end)
589 pass = 2;
590 if (zone_start >= zone_end) {
591 vol->data1_zone_pos = zone_start =
592 vol->mft_zone_end;
593 pass = 2;
594 }
595 break;
596 case 2:
597 ntfs_debug("Switching from data1 zone to "
598 "data2 zone.");
599 /* Update data1 zone position. */
600 if (rlpos) {
601 LCN tc;
602
603 ntfs_debug("Before checks, "
604 "vol->data1_zone_pos "
605 "0x%llx.",
606 (unsigned long long)
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 =
612 vol->mft_zone_end;
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 "
620 "0x%llx.",
621 (unsigned long long)
622 vol->data1_zone_pos);
623 }
624 /* Switch from data1 zone to data2 zone. */
625 search_zone = 4;
626 zone_start = bmp_initial_pos =
627 vol->data2_zone_pos;
628 zone_end = vol->mft_zone_start;
629 if (!zone_start)
630 pass = 2;
631 if (zone_start >= zone_end) {
632 vol->data2_zone_pos = zone_start =
633 bmp_initial_pos = 0;
634 pass = 2;
635 }
636 break;
637 case 4:
638 ntfs_debug("Switching from data2 zone to "
639 "data1 zone.");
640 /* Update data2 zone position. */
641 if (rlpos) {
642 LCN tc;
643
644 ntfs_debug("Before checks, "
645 "vol->data2_zone_pos "
646 "0x%llx.",
647 (unsigned long long)
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 "
659 "0x%llx.",
660 (unsigned long long)
661 vol->data2_zone_pos);
662 }
663 /* Switch from data2 zone to data1 zone. */
664 goto switch_to_data1_zone;
665 default:
666 BUG();
667 }
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.",
671 search_zone, pass,
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;
681 }
682 ntfs_debug("Continuing outer while loop.");
683 continue;
684 } /* done_zones == 7 */
685 ntfs_debug("All zones are finished.");
686 /*
687 * All zones are finished! If DATA_ZONE, shrink mft zone. If
688 * MFT_ZONE, we have really run out of space.
689 */
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. */
Anton Altaparmakov43b01fd2005-03-09 15:18:43 +0000699 err = -ENOSPC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 goto out;
701 } /* zone == DATA_ZONE && mft_zone_size > 0 */
702 ntfs_debug("Shrinking mft zone.");
703 zone_end = vol->mft_zone_end;
704 mft_zone_size >>= 1;
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;
714 }
715 bmp_pos = zone_start = bmp_initial_pos =
716 vol->data1_zone_pos = vol->mft_zone_end;
717 search_zone = 2;
718 pass = 2;
719 done_zones &= ~2;
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);
734 }
735 ntfs_debug("After outer while loop.");
736out:
737 ntfs_debug("At out.");
738 /* Add runlist terminator element. */
739 if (likely(rl)) {
740 rl[rlpos].vcn = rl[rlpos - 1].vcn + rl[rlpos - 1].length;
Anton Altaparmakovfc0fa7d2005-10-04 14:36:56 +0100741 rl[rlpos].lcn = is_extension ? LCN_ENOENT : LCN_RL_NOT_MAPPED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 rl[rlpos].length = 0;
743 }
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);
749 need_writeback = 0;
750 }
751 ntfs_unmap_page(page);
752 }
753 if (likely(!err)) {
754 up_write(&vol->lcnbmp_lock);
755 ntfs_debug("Done.");
756 return rl;
757 }
758 ntfs_error(vol->sb, "Failed to allocate clusters, aborting "
759 "(error %i).", err);
760 if (rl) {
761 int err2;
762
Anton Altaparmakov43b01fd2005-03-09 15:18:43 +0000763 if (err == -ENOSPC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 ntfs_debug("Not enough space to complete allocation, "
Anton Altaparmakov43b01fd2005-03-09 15:18:43 +0000765 "err -ENOSPC, first free lcn 0x%llx, "
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 "could allocate up to 0x%llx "
767 "clusters.",
768 (unsigned long long)rl[0].lcn,
Randy Dunlap89075472005-03-03 11:19:53 +0000769 (unsigned long long)(count - clusters));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 /* Deallocate all allocated clusters. */
771 ntfs_debug("Attempting rollback...");
772 err2 = ntfs_cluster_free_from_rl_nolock(vol, rl);
773 if (err2) {
774 ntfs_error(vol->sb, "Failed to rollback (error %i). "
775 "Leaving inconsistent metadata! "
776 "Unmount and run chkdsk.", err2);
777 NVolSetErrors(vol);
778 }
779 /* Free the runlist. */
780 ntfs_free(rl);
Anton Altaparmakov43b01fd2005-03-09 15:18:43 +0000781 } else if (err == -ENOSPC)
782 ntfs_debug("No space left at all, err = -ENOSPC, first free "
783 "lcn = 0x%llx.",
784 (long long)vol->data1_zone_pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 up_write(&vol->lcnbmp_lock);
786 return ERR_PTR(err);
787}
788
789/**
790 * __ntfs_cluster_free - free clusters on an ntfs volume
Anton Altaparmakov715dc632005-09-23 11:24:28 +0100791 * @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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 * @count: number of clusters to free or -1 for all clusters
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100794 * @ctx: active attribute search context if present or NULL if not
Anton Altaparmakovbbf18132005-09-08 21:09:06 +0100795 * @is_rollback: true if this is a rollback operation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 *
797 * Free @count clusters starting at the cluster @start_vcn in the runlist
Anton Altaparmakov715dc632005-09-23 11:24:28 +0100798 * described by the vfs inode @ni.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 *
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.
803 *
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100804 * 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.
809 *
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:
816 * m = ctx->mrec;
817 * a = ctx->attr;
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 *.
820 *
Richard Knutssonc49c3112006-09-30 23:27:12 -0700821 * @is_rollback should always be 'false', it is for internal use to rollback
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 * errors. You probably want to use ntfs_cluster_free() instead.
823 *
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100824 * 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 *
827 * Return the number of deallocated clusters (not counting sparse ones) on
828 * success and -errno on error.
829 *
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100830 * WARNING: If @ctx is supplied, regardless of whether success or failure is
Richard Knutssonc49c3112006-09-30 23:27:12 -0700831 * returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100832 * 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.
836 *
Anton Altaparmakov715dc632005-09-23 11:24:28 +0100837 * 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 * - The volume lcn bitmap must be unlocked on entry and is unlocked
841 * on return.
842 * - This function takes the volume lcn bitmap lock for writing and
843 * modifies the bitmap contents.
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100844 * - 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 */
Anton Altaparmakov715dc632005-09-23 11:24:28 +0100849s64 __ntfs_cluster_free(ntfs_inode *ni, const VCN start_vcn, s64 count,
Richard Knutssonc49c3112006-09-30 23:27:12 -0700850 ntfs_attr_search_ctx *ctx, const bool is_rollback)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851{
852 s64 delta, to_free, total_freed, real_freed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 ntfs_volume *vol;
854 struct inode *lcnbmp_vi;
855 runlist_element *rl;
856 int err;
857
Anton Altaparmakov715dc632005-09-23 11:24:28 +0100858 BUG_ON(!ni);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 ntfs_debug("Entering for i_ino 0x%lx, start_vcn 0x%llx, count "
Anton Altaparmakov715dc632005-09-23 11:24:28 +0100860 "0x%llx.%s", ni->mft_no, (unsigned long long)start_vcn,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 (unsigned long long)count,
862 is_rollback ? " (rollback)" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 vol = ni->vol;
864 lcnbmp_vi = vol->lcnbmp_ino;
865 BUG_ON(!lcnbmp_vi);
866 BUG_ON(start_vcn < 0);
867 BUG_ON(count < -1);
868 /*
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.
874 */
875 if (likely(!is_rollback))
876 down_write(&vol->lcnbmp_lock);
877
878 total_freed = real_freed = 0;
879
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100880 rl = ntfs_attr_find_vcn_nolock(ni, start_vcn, ctx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 if (IS_ERR(rl)) {
882 if (!is_rollback)
883 ntfs_error(vol->sb, "Failed to find first runlist "
884 "element (error %li), aborting.",
885 PTR_ERR(rl));
886 err = PTR_ERR(rl);
887 goto err_out;
888 }
889 if (unlikely(rl->lcn < LCN_HOLE)) {
890 if (!is_rollback)
891 ntfs_error(vol->sb, "First runlist element has "
892 "invalid lcn, aborting.");
893 err = -EIO;
Anton Altaparmakovb6ad6c52005-02-15 10:08:43 +0000894 goto err_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 }
896 /* Find the starting cluster inside the run that needs freeing. */
897 delta = start_vcn - rl->vcn;
898
899 /* The number of clusters in this run that need freeing. */
900 to_free = rl->length - delta;
901 if (count >= 0 && to_free > count)
902 to_free = count;
903
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);
908 if (unlikely(err)) {
909 if (!is_rollback)
910 ntfs_error(vol->sb, "Failed to clear first run "
911 "(error %i), aborting.", err);
Anton Altaparmakovb6ad6c52005-02-15 10:08:43 +0000912 goto err_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 }
914 /* We have freed @to_free real clusters. */
915 real_freed = to_free;
916 };
917 /* Go to the next run and adjust the number of clusters left to free. */
918 ++rl;
919 if (count >= 0)
920 count -= to_free;
921
922 /* Keep track of the total "freed" clusters, including sparse ones. */
923 total_freed = to_free;
924 /*
925 * Loop over the remaining runs, using @count as a capping value, and
926 * free them.
927 */
928 for (; rl->length && count != 0; ++rl) {
929 if (unlikely(rl->lcn < LCN_HOLE)) {
930 VCN vcn;
931
Anton Altaparmakovb6ad6c52005-02-15 10:08:43 +0000932 /* Attempt to map runlist. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 vcn = rl->vcn;
Anton Altaparmakov511bea52005-10-04 14:24:21 +0100934 rl = ntfs_attr_find_vcn_nolock(ni, vcn, ctx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 if (IS_ERR(rl)) {
936 err = PTR_ERR(rl);
937 if (!is_rollback)
Anton Altaparmakovb6ad6c52005-02-15 10:08:43 +0000938 ntfs_error(vol->sb, "Failed to map "
939 "runlist fragment or "
940 "failed to find "
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 "subsequent runlist "
942 "element.");
943 goto err_out;
944 }
945 if (unlikely(rl->lcn < LCN_HOLE)) {
946 if (!is_rollback)
947 ntfs_error(vol->sb, "Runlist element "
948 "has invalid lcn "
949 "(0x%llx).",
950 (unsigned long long)
951 rl->lcn);
952 err = -EIO;
Anton Altaparmakovb6ad6c52005-02-15 10:08:43 +0000953 goto err_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 }
955 }
956 /* The number of clusters in this run that need freeing. */
957 to_free = rl->length;
958 if (count >= 0 && to_free > count)
959 to_free = count;
960
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);
965 if (unlikely(err)) {
966 if (!is_rollback)
967 ntfs_error(vol->sb, "Failed to clear "
968 "subsequent run.");
Anton Altaparmakovb6ad6c52005-02-15 10:08:43 +0000969 goto err_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 }
971 /* We have freed @to_free real clusters. */
972 real_freed += to_free;
973 }
974 /* Adjust the number of clusters left to free. */
975 if (count >= 0)
976 count -= to_free;
977
978 /* Update the total done clusters. */
979 total_freed += to_free;
980 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 if (likely(!is_rollback))
982 up_write(&vol->lcnbmp_lock);
983
984 BUG_ON(count > 0);
985
986 /* We are done. Return the number of actually freed clusters. */
987 ntfs_debug("Done.");
988 return real_freed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989err_out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 if (is_rollback)
991 return err;
992 /* If no real clusters were freed, no need to rollback. */
993 if (!real_freed) {
994 up_write(&vol->lcnbmp_lock);
995 return err;
996 }
997 /*
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.
1001 */
Richard Knutssonc49c3112006-09-30 23:27:12 -07001002 delta = __ntfs_cluster_free(ni, start_vcn, total_freed, ctx, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 if (delta < 0) {
1004 ntfs_error(vol->sb, "Failed to rollback (error %i). Leaving "
1005 "inconsistent metadata! Unmount and run "
1006 "chkdsk.", (int)delta);
1007 NVolSetErrors(vol);
1008 }
1009 up_write(&vol->lcnbmp_lock);
1010 ntfs_error(vol->sb, "Aborting (error %i).", err);
1011 return err;
1012}
1013
1014#endif /* NTFS_RW */