Merge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-2.6
[linux-2.6] / fs / ufs / balloc.c
1 /*
2  *  linux/fs/ufs/balloc.c
3  *
4  * Copyright (C) 1998
5  * Daniel Pirkl <daniel.pirkl@email.cz>
6  * Charles University, Faculty of Mathematics and Physics
7  *
8  * UFS2 write support Evgeniy Dushistov <dushistov@mail.ru>, 2007
9  */
10
11 #include <linux/fs.h>
12 #include <linux/stat.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/quotaops.h>
16 #include <linux/buffer_head.h>
17 #include <linux/capability.h>
18 #include <linux/bitops.h>
19 #include <asm/byteorder.h>
20
21 #include "ufs_fs.h"
22 #include "ufs.h"
23 #include "swab.h"
24 #include "util.h"
25
26 #define INVBLOCK ((u64)-1L)
27
28 static u64 ufs_add_fragments(struct inode *, u64, unsigned, unsigned, int *);
29 static u64 ufs_alloc_fragments(struct inode *, unsigned, u64, unsigned, int *);
30 static u64 ufs_alloccg_block(struct inode *, struct ufs_cg_private_info *, u64, int *);
31 static u64 ufs_bitmap_search (struct super_block *, struct ufs_cg_private_info *, u64, unsigned);
32 static unsigned char ufs_fragtable_8fpb[], ufs_fragtable_other[];
33 static void ufs_clusteracct(struct super_block *, struct ufs_cg_private_info *, unsigned, int);
34
35 /*
36  * Free 'count' fragments from fragment number 'fragment'
37  */
38 void ufs_free_fragments(struct inode *inode, u64 fragment, unsigned count)
39 {
40         struct super_block * sb;
41         struct ufs_sb_private_info * uspi;
42         struct ufs_super_block_first * usb1;
43         struct ufs_cg_private_info * ucpi;
44         struct ufs_cylinder_group * ucg;
45         unsigned cgno, bit, end_bit, bbase, blkmap, i;
46         u64 blkno;
47         
48         sb = inode->i_sb;
49         uspi = UFS_SB(sb)->s_uspi;
50         usb1 = ubh_get_usb_first(uspi);
51         
52         UFSD("ENTER, fragment %llu, count %u\n",
53              (unsigned long long)fragment, count);
54         
55         if (ufs_fragnum(fragment) + count > uspi->s_fpg)
56                 ufs_error (sb, "ufs_free_fragments", "internal error");
57         
58         lock_super(sb);
59         
60         cgno = ufs_dtog(uspi, fragment);
61         bit = ufs_dtogd(uspi, fragment);
62         if (cgno >= uspi->s_ncg) {
63                 ufs_panic (sb, "ufs_free_fragments", "freeing blocks are outside device");
64                 goto failed;
65         }
66                 
67         ucpi = ufs_load_cylinder (sb, cgno);
68         if (!ucpi) 
69                 goto failed;
70         ucg = ubh_get_ucg (UCPI_UBH(ucpi));
71         if (!ufs_cg_chkmagic(sb, ucg)) {
72                 ufs_panic (sb, "ufs_free_fragments", "internal error, bad magic number on cg %u", cgno);
73                 goto failed;
74         }
75
76         end_bit = bit + count;
77         bbase = ufs_blknum (bit);
78         blkmap = ubh_blkmap (UCPI_UBH(ucpi), ucpi->c_freeoff, bbase);
79         ufs_fragacct (sb, blkmap, ucg->cg_frsum, -1);
80         for (i = bit; i < end_bit; i++) {
81                 if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, i))
82                         ubh_setbit (UCPI_UBH(ucpi), ucpi->c_freeoff, i);
83                 else 
84                         ufs_error (sb, "ufs_free_fragments",
85                                    "bit already cleared for fragment %u", i);
86         }
87         
88         DQUOT_FREE_BLOCK (inode, count);
89
90         
91         fs32_add(sb, &ucg->cg_cs.cs_nffree, count);
92         uspi->cs_total.cs_nffree += count;
93         fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
94         blkmap = ubh_blkmap (UCPI_UBH(ucpi), ucpi->c_freeoff, bbase);
95         ufs_fragacct(sb, blkmap, ucg->cg_frsum, 1);
96
97         /*
98          * Trying to reassemble free fragments into block
99          */
100         blkno = ufs_fragstoblks (bbase);
101         if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno)) {
102                 fs32_sub(sb, &ucg->cg_cs.cs_nffree, uspi->s_fpb);
103                 uspi->cs_total.cs_nffree -= uspi->s_fpb;
104                 fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, uspi->s_fpb);
105                 if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
106                         ufs_clusteracct (sb, ucpi, blkno, 1);
107                 fs32_add(sb, &ucg->cg_cs.cs_nbfree, 1);
108                 uspi->cs_total.cs_nbfree++;
109                 fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nbfree, 1);
110                 if (uspi->fs_magic != UFS2_MAGIC) {
111                         unsigned cylno = ufs_cbtocylno (bbase);
112
113                         fs16_add(sb, &ubh_cg_blks(ucpi, cylno,
114                                                   ufs_cbtorpos(bbase)), 1);
115                         fs32_add(sb, &ubh_cg_blktot(ucpi, cylno), 1);
116                 }
117         }
118         
119         ubh_mark_buffer_dirty (USPI_UBH(uspi));
120         ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
121         if (sb->s_flags & MS_SYNCHRONOUS) {
122                 ubh_ll_rw_block(SWRITE, UCPI_UBH(ucpi));
123                 ubh_wait_on_buffer (UCPI_UBH(ucpi));
124         }
125         sb->s_dirt = 1;
126         
127         unlock_super (sb);
128         UFSD("EXIT\n");
129         return;
130
131 failed:
132         unlock_super (sb);
133         UFSD("EXIT (FAILED)\n");
134         return;
135 }
136
137 /*
138  * Free 'count' fragments from fragment number 'fragment' (free whole blocks)
139  */
140 void ufs_free_blocks(struct inode *inode, u64 fragment, unsigned count)
141 {
142         struct super_block * sb;
143         struct ufs_sb_private_info * uspi;
144         struct ufs_super_block_first * usb1;
145         struct ufs_cg_private_info * ucpi;
146         struct ufs_cylinder_group * ucg;
147         unsigned overflow, cgno, bit, end_bit, i;
148         u64 blkno;
149         
150         sb = inode->i_sb;
151         uspi = UFS_SB(sb)->s_uspi;
152         usb1 = ubh_get_usb_first(uspi);
153
154         UFSD("ENTER, fragment %llu, count %u\n",
155              (unsigned long long)fragment, count);
156         
157         if ((fragment & uspi->s_fpbmask) || (count & uspi->s_fpbmask)) {
158                 ufs_error (sb, "ufs_free_blocks", "internal error, "
159                            "fragment %llu, count %u\n",
160                            (unsigned long long)fragment, count);
161                 goto failed;
162         }
163
164         lock_super(sb);
165         
166 do_more:
167         overflow = 0;
168         cgno = ufs_dtog(uspi, fragment);
169         bit = ufs_dtogd(uspi, fragment);
170         if (cgno >= uspi->s_ncg) {
171                 ufs_panic (sb, "ufs_free_blocks", "freeing blocks are outside device");
172                 goto failed_unlock;
173         }
174         end_bit = bit + count;
175         if (end_bit > uspi->s_fpg) {
176                 overflow = bit + count - uspi->s_fpg;
177                 count -= overflow;
178                 end_bit -= overflow;
179         }
180
181         ucpi = ufs_load_cylinder (sb, cgno);
182         if (!ucpi) 
183                 goto failed_unlock;
184         ucg = ubh_get_ucg (UCPI_UBH(ucpi));
185         if (!ufs_cg_chkmagic(sb, ucg)) {
186                 ufs_panic (sb, "ufs_free_blocks", "internal error, bad magic number on cg %u", cgno);
187                 goto failed_unlock;
188         }
189
190         for (i = bit; i < end_bit; i += uspi->s_fpb) {
191                 blkno = ufs_fragstoblks(i);
192                 if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno)) {
193                         ufs_error(sb, "ufs_free_blocks", "freeing free fragment");
194                 }
195                 ubh_setblock(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno);
196                 if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
197                         ufs_clusteracct (sb, ucpi, blkno, 1);
198                 DQUOT_FREE_BLOCK(inode, uspi->s_fpb);
199
200                 fs32_add(sb, &ucg->cg_cs.cs_nbfree, 1);
201                 uspi->cs_total.cs_nbfree++;
202                 fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nbfree, 1);
203
204                 if (uspi->fs_magic != UFS2_MAGIC) {
205                         unsigned cylno = ufs_cbtocylno(i);
206
207                         fs16_add(sb, &ubh_cg_blks(ucpi, cylno,
208                                                   ufs_cbtorpos(i)), 1);
209                         fs32_add(sb, &ubh_cg_blktot(ucpi, cylno), 1);
210                 }
211         }
212
213         ubh_mark_buffer_dirty (USPI_UBH(uspi));
214         ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
215         if (sb->s_flags & MS_SYNCHRONOUS) {
216                 ubh_ll_rw_block(SWRITE, UCPI_UBH(ucpi));
217                 ubh_wait_on_buffer (UCPI_UBH(ucpi));
218         }
219
220         if (overflow) {
221                 fragment += count;
222                 count = overflow;
223                 goto do_more;
224         }
225
226         sb->s_dirt = 1;
227         unlock_super (sb);
228         UFSD("EXIT\n");
229         return;
230
231 failed_unlock:
232         unlock_super (sb);
233 failed:
234         UFSD("EXIT (FAILED)\n");
235         return;
236 }
237
238 /*
239  * Modify inode page cache in such way:
240  * have - blocks with b_blocknr equal to oldb...oldb+count-1
241  * get - blocks with b_blocknr equal to newb...newb+count-1
242  * also we suppose that oldb...oldb+count-1 blocks
243  * situated at the end of file.
244  *
245  * We can come here from ufs_writepage or ufs_prepare_write,
246  * locked_page is argument of these functions, so we already lock it.
247  */
248 static void ufs_change_blocknr(struct inode *inode, sector_t beg,
249                                unsigned int count, sector_t oldb,
250                                sector_t newb, struct page *locked_page)
251 {
252         const unsigned blks_per_page =
253                 1 << (PAGE_CACHE_SHIFT - inode->i_blkbits);
254         const unsigned mask = blks_per_page - 1;
255         struct address_space * const mapping = inode->i_mapping;
256         pgoff_t index, cur_index, last_index;
257         unsigned pos, j, lblock;
258         sector_t end, i;
259         struct page *page;
260         struct buffer_head *head, *bh;
261
262         UFSD("ENTER, ino %lu, count %u, oldb %llu, newb %llu\n",
263               inode->i_ino, count,
264              (unsigned long long)oldb, (unsigned long long)newb);
265
266         BUG_ON(!locked_page);
267         BUG_ON(!PageLocked(locked_page));
268
269         cur_index = locked_page->index;
270         end = count + beg;
271         last_index = end >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
272         for (i = beg; i < end; i = (i | mask) + 1) {
273                 index = i >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
274
275                 if (likely(cur_index != index)) {
276                         page = ufs_get_locked_page(mapping, index);
277                         if (!page)/* it was truncated */
278                                 continue;
279                         if (IS_ERR(page)) {/* or EIO */
280                                 ufs_error(inode->i_sb, __FUNCTION__,
281                                           "read of page %llu failed\n",
282                                           (unsigned long long)index);
283                                 continue;
284                         }
285                 } else
286                         page = locked_page;
287
288                 head = page_buffers(page);
289                 bh = head;
290                 pos = i & mask;
291                 for (j = 0; j < pos; ++j)
292                         bh = bh->b_this_page;
293
294
295                 if (unlikely(index == last_index))
296                         lblock = end & mask;
297                 else
298                         lblock = blks_per_page;
299
300                 do {
301                         if (j >= lblock)
302                                 break;
303                         pos = (i - beg) + j;
304
305                         if (!buffer_mapped(bh))
306                                         map_bh(bh, inode->i_sb, oldb + pos);
307                         if (!buffer_uptodate(bh)) {
308                                 ll_rw_block(READ, 1, &bh);
309                                 wait_on_buffer(bh);
310                                 if (!buffer_uptodate(bh)) {
311                                         ufs_error(inode->i_sb, __FUNCTION__,
312                                                   "read of block failed\n");
313                                         break;
314                                 }
315                         }
316
317                         UFSD(" change from %llu to %llu, pos %u\n",
318                              (unsigned long long)(pos + oldb),
319                              (unsigned long long)(pos + newb), pos);
320
321                         bh->b_blocknr = newb + pos;
322                         unmap_underlying_metadata(bh->b_bdev,
323                                                   bh->b_blocknr);
324                         mark_buffer_dirty(bh);
325                         ++j;
326                         bh = bh->b_this_page;
327                 } while (bh != head);
328
329                 if (likely(cur_index != index))
330                         ufs_put_locked_page(page);
331         }
332         UFSD("EXIT\n");
333 }
334
335 static void ufs_clear_frags(struct inode *inode, sector_t beg, unsigned int n,
336                             int sync)
337 {
338         struct buffer_head *bh;
339         sector_t end = beg + n;
340
341         for (; beg < end; ++beg) {
342                 bh = sb_getblk(inode->i_sb, beg);
343                 lock_buffer(bh);
344                 memset(bh->b_data, 0, inode->i_sb->s_blocksize);
345                 set_buffer_uptodate(bh);
346                 mark_buffer_dirty(bh);
347                 unlock_buffer(bh);
348                 if (IS_SYNC(inode) || sync)
349                         sync_dirty_buffer(bh);
350                 brelse(bh);
351         }
352 }
353
354 u64 ufs_new_fragments(struct inode *inode, void *p, u64 fragment,
355                            u64 goal, unsigned count, int *err,
356                            struct page *locked_page)
357 {
358         struct super_block * sb;
359         struct ufs_sb_private_info * uspi;
360         struct ufs_super_block_first * usb1;
361         unsigned cgno, oldcount, newcount;
362         u64 tmp, request, result;
363         
364         UFSD("ENTER, ino %lu, fragment %llu, goal %llu, count %u\n",
365              inode->i_ino, (unsigned long long)fragment,
366              (unsigned long long)goal, count);
367         
368         sb = inode->i_sb;
369         uspi = UFS_SB(sb)->s_uspi;
370         usb1 = ubh_get_usb_first(uspi);
371         *err = -ENOSPC;
372
373         lock_super (sb);
374         tmp = ufs_data_ptr_to_cpu(sb, p);
375
376         if (count + ufs_fragnum(fragment) > uspi->s_fpb) {
377                 ufs_warning(sb, "ufs_new_fragments", "internal warning"
378                             " fragment %llu, count %u",
379                             (unsigned long long)fragment, count);
380                 count = uspi->s_fpb - ufs_fragnum(fragment); 
381         }
382         oldcount = ufs_fragnum (fragment);
383         newcount = oldcount + count;
384
385         /*
386          * Somebody else has just allocated our fragments
387          */
388         if (oldcount) {
389                 if (!tmp) {
390                         ufs_error(sb, "ufs_new_fragments", "internal error, "
391                                   "fragment %llu, tmp %llu\n",
392                                   (unsigned long long)fragment,
393                                   (unsigned long long)tmp);
394                         unlock_super(sb);
395                         return INVBLOCK;
396                 }
397                 if (fragment < UFS_I(inode)->i_lastfrag) {
398                         UFSD("EXIT (ALREADY ALLOCATED)\n");
399                         unlock_super (sb);
400                         return 0;
401                 }
402         }
403         else {
404                 if (tmp) {
405                         UFSD("EXIT (ALREADY ALLOCATED)\n");
406                         unlock_super(sb);
407                         return 0;
408                 }
409         }
410
411         /*
412          * There is not enough space for user on the device
413          */
414         if (!capable(CAP_SYS_RESOURCE) && ufs_freespace(uspi, UFS_MINFREE) <= 0) {
415                 unlock_super (sb);
416                 UFSD("EXIT (FAILED)\n");
417                 return 0;
418         }
419
420         if (goal >= uspi->s_size) 
421                 goal = 0;
422         if (goal == 0) 
423                 cgno = ufs_inotocg (inode->i_ino);
424         else
425                 cgno = ufs_dtog(uspi, goal);
426          
427         /*
428          * allocate new fragment
429          */
430         if (oldcount == 0) {
431                 result = ufs_alloc_fragments (inode, cgno, goal, count, err);
432                 if (result) {
433                         ufs_cpu_to_data_ptr(sb, p, result);
434                         *err = 0;
435                         UFS_I(inode)->i_lastfrag =
436                                 max_t(u32, UFS_I(inode)->i_lastfrag,
437                                       fragment + count);
438                         ufs_clear_frags(inode, result + oldcount,
439                                         newcount - oldcount, locked_page != NULL);
440                 }
441                 unlock_super(sb);
442                 UFSD("EXIT, result %llu\n", (unsigned long long)result);
443                 return result;
444         }
445
446         /*
447          * resize block
448          */
449         result = ufs_add_fragments (inode, tmp, oldcount, newcount, err);
450         if (result) {
451                 *err = 0;
452                 UFS_I(inode)->i_lastfrag = max_t(u32, UFS_I(inode)->i_lastfrag, fragment + count);
453                 ufs_clear_frags(inode, result + oldcount, newcount - oldcount,
454                                 locked_page != NULL);
455                 unlock_super(sb);
456                 UFSD("EXIT, result %llu\n", (unsigned long long)result);
457                 return result;
458         }
459
460         /*
461          * allocate new block and move data
462          */
463         switch (fs32_to_cpu(sb, usb1->fs_optim)) {
464             case UFS_OPTSPACE:
465                 request = newcount;
466                 if (uspi->s_minfree < 5 || uspi->cs_total.cs_nffree
467                     > uspi->s_dsize * uspi->s_minfree / (2 * 100))
468                         break;
469                 usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
470                 break;
471             default:
472                 usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
473         
474             case UFS_OPTTIME:
475                 request = uspi->s_fpb;
476                 if (uspi->cs_total.cs_nffree < uspi->s_dsize *
477                     (uspi->s_minfree - 2) / 100)
478                         break;
479                 usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
480                 break;
481         }
482         result = ufs_alloc_fragments (inode, cgno, goal, request, err);
483         if (result) {
484                 ufs_clear_frags(inode, result + oldcount, newcount - oldcount,
485                                 locked_page != NULL);
486                 ufs_change_blocknr(inode, fragment - oldcount, oldcount,
487                                    uspi->s_sbbase + tmp,
488                                    uspi->s_sbbase + result, locked_page);
489                 ufs_cpu_to_data_ptr(sb, p, result);
490                 *err = 0;
491                 UFS_I(inode)->i_lastfrag = max_t(u32, UFS_I(inode)->i_lastfrag, fragment + count);
492                 unlock_super(sb);
493                 if (newcount < request)
494                         ufs_free_fragments (inode, result + newcount, request - newcount);
495                 ufs_free_fragments (inode, tmp, oldcount);
496                 UFSD("EXIT, result %llu\n", (unsigned long long)result);
497                 return result;
498         }
499
500         unlock_super(sb);
501         UFSD("EXIT (FAILED)\n");
502         return 0;
503 }               
504
505 static u64 ufs_add_fragments(struct inode *inode, u64 fragment,
506                              unsigned oldcount, unsigned newcount, int *err)
507 {
508         struct super_block * sb;
509         struct ufs_sb_private_info * uspi;
510         struct ufs_super_block_first * usb1;
511         struct ufs_cg_private_info * ucpi;
512         struct ufs_cylinder_group * ucg;
513         unsigned cgno, fragno, fragoff, count, fragsize, i;
514         
515         UFSD("ENTER, fragment %llu, oldcount %u, newcount %u\n",
516              (unsigned long long)fragment, oldcount, newcount);
517         
518         sb = inode->i_sb;
519         uspi = UFS_SB(sb)->s_uspi;
520         usb1 = ubh_get_usb_first (uspi);
521         count = newcount - oldcount;
522         
523         cgno = ufs_dtog(uspi, fragment);
524         if (fs32_to_cpu(sb, UFS_SB(sb)->fs_cs(cgno).cs_nffree) < count)
525                 return 0;
526         if ((ufs_fragnum (fragment) + newcount) > uspi->s_fpb)
527                 return 0;
528         ucpi = ufs_load_cylinder (sb, cgno);
529         if (!ucpi)
530                 return 0;
531         ucg = ubh_get_ucg (UCPI_UBH(ucpi));
532         if (!ufs_cg_chkmagic(sb, ucg)) {
533                 ufs_panic (sb, "ufs_add_fragments",
534                         "internal error, bad magic number on cg %u", cgno);
535                 return 0;
536         }
537
538         fragno = ufs_dtogd(uspi, fragment);
539         fragoff = ufs_fragnum (fragno);
540         for (i = oldcount; i < newcount; i++)
541                 if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, fragno + i))
542                         return 0;
543         /*
544          * Block can be extended
545          */
546         ucg->cg_time = cpu_to_fs32(sb, get_seconds());
547         for (i = newcount; i < (uspi->s_fpb - fragoff); i++)
548                 if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, fragno + i))
549                         break;
550         fragsize = i - oldcount;
551         if (!fs32_to_cpu(sb, ucg->cg_frsum[fragsize]))
552                 ufs_panic (sb, "ufs_add_fragments",
553                         "internal error or corrupted bitmap on cg %u", cgno);
554         fs32_sub(sb, &ucg->cg_frsum[fragsize], 1);
555         if (fragsize != count)
556                 fs32_add(sb, &ucg->cg_frsum[fragsize - count], 1);
557         for (i = oldcount; i < newcount; i++)
558                 ubh_clrbit (UCPI_UBH(ucpi), ucpi->c_freeoff, fragno + i);
559         if(DQUOT_ALLOC_BLOCK(inode, count)) {
560                 *err = -EDQUOT;
561                 return 0;
562         }
563
564         fs32_sub(sb, &ucg->cg_cs.cs_nffree, count);
565         fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
566         uspi->cs_total.cs_nffree -= count;
567         
568         ubh_mark_buffer_dirty (USPI_UBH(uspi));
569         ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
570         if (sb->s_flags & MS_SYNCHRONOUS) {
571                 ubh_ll_rw_block(SWRITE, UCPI_UBH(ucpi));
572                 ubh_wait_on_buffer (UCPI_UBH(ucpi));
573         }
574         sb->s_dirt = 1;
575
576         UFSD("EXIT, fragment %llu\n", (unsigned long long)fragment);
577         
578         return fragment;
579 }
580
581 #define UFS_TEST_FREE_SPACE_CG \
582         ucg = (struct ufs_cylinder_group *) UFS_SB(sb)->s_ucg[cgno]->b_data; \
583         if (fs32_to_cpu(sb, ucg->cg_cs.cs_nbfree)) \
584                 goto cg_found; \
585         for (k = count; k < uspi->s_fpb; k++) \
586                 if (fs32_to_cpu(sb, ucg->cg_frsum[k])) \
587                         goto cg_found; 
588
589 static u64 ufs_alloc_fragments(struct inode *inode, unsigned cgno,
590                                u64 goal, unsigned count, int *err)
591 {
592         struct super_block * sb;
593         struct ufs_sb_private_info * uspi;
594         struct ufs_super_block_first * usb1;
595         struct ufs_cg_private_info * ucpi;
596         struct ufs_cylinder_group * ucg;
597         unsigned oldcg, i, j, k, allocsize;
598         u64 result;
599         
600         UFSD("ENTER, ino %lu, cgno %u, goal %llu, count %u\n",
601              inode->i_ino, cgno, (unsigned long long)goal, count);
602
603         sb = inode->i_sb;
604         uspi = UFS_SB(sb)->s_uspi;
605         usb1 = ubh_get_usb_first(uspi);
606         oldcg = cgno;
607         
608         /*
609          * 1. searching on preferred cylinder group
610          */
611         UFS_TEST_FREE_SPACE_CG
612
613         /*
614          * 2. quadratic rehash
615          */
616         for (j = 1; j < uspi->s_ncg; j *= 2) {
617                 cgno += j;
618                 if (cgno >= uspi->s_ncg) 
619                         cgno -= uspi->s_ncg;
620                 UFS_TEST_FREE_SPACE_CG
621         }
622
623         /*
624          * 3. brute force search
625          * We start at i = 2 ( 0 is checked at 1.step, 1 at 2.step )
626          */
627         cgno = (oldcg + 1) % uspi->s_ncg;
628         for (j = 2; j < uspi->s_ncg; j++) {
629                 cgno++;
630                 if (cgno >= uspi->s_ncg)
631                         cgno = 0;
632                 UFS_TEST_FREE_SPACE_CG
633         }
634         
635         UFSD("EXIT (FAILED)\n");
636         return 0;
637
638 cg_found:
639         ucpi = ufs_load_cylinder (sb, cgno);
640         if (!ucpi)
641                 return 0;
642         ucg = ubh_get_ucg (UCPI_UBH(ucpi));
643         if (!ufs_cg_chkmagic(sb, ucg)) 
644                 ufs_panic (sb, "ufs_alloc_fragments",
645                         "internal error, bad magic number on cg %u", cgno);
646         ucg->cg_time = cpu_to_fs32(sb, get_seconds());
647
648         if (count == uspi->s_fpb) {
649                 result = ufs_alloccg_block (inode, ucpi, goal, err);
650                 if (result == INVBLOCK)
651                         return 0;
652                 goto succed;
653         }
654
655         for (allocsize = count; allocsize < uspi->s_fpb; allocsize++)
656                 if (fs32_to_cpu(sb, ucg->cg_frsum[allocsize]) != 0)
657                         break;
658         
659         if (allocsize == uspi->s_fpb) {
660                 result = ufs_alloccg_block (inode, ucpi, goal, err);
661                 if (result == INVBLOCK)
662                         return 0;
663                 goal = ufs_dtogd(uspi, result);
664                 for (i = count; i < uspi->s_fpb; i++)
665                         ubh_setbit (UCPI_UBH(ucpi), ucpi->c_freeoff, goal + i);
666                 i = uspi->s_fpb - count;
667                 DQUOT_FREE_BLOCK(inode, i);
668
669                 fs32_add(sb, &ucg->cg_cs.cs_nffree, i);
670                 uspi->cs_total.cs_nffree += i;
671                 fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, i);
672                 fs32_add(sb, &ucg->cg_frsum[i], 1);
673                 goto succed;
674         }
675
676         result = ufs_bitmap_search (sb, ucpi, goal, allocsize);
677         if (result == INVBLOCK)
678                 return 0;
679         if(DQUOT_ALLOC_BLOCK(inode, count)) {
680                 *err = -EDQUOT;
681                 return 0;
682         }
683         for (i = 0; i < count; i++)
684                 ubh_clrbit (UCPI_UBH(ucpi), ucpi->c_freeoff, result + i);
685         
686         fs32_sub(sb, &ucg->cg_cs.cs_nffree, count);
687         uspi->cs_total.cs_nffree -= count;
688         fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
689         fs32_sub(sb, &ucg->cg_frsum[allocsize], 1);
690
691         if (count != allocsize)
692                 fs32_add(sb, &ucg->cg_frsum[allocsize - count], 1);
693
694 succed:
695         ubh_mark_buffer_dirty (USPI_UBH(uspi));
696         ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
697         if (sb->s_flags & MS_SYNCHRONOUS) {
698                 ubh_ll_rw_block(SWRITE, UCPI_UBH(ucpi));
699                 ubh_wait_on_buffer (UCPI_UBH(ucpi));
700         }
701         sb->s_dirt = 1;
702
703         result += cgno * uspi->s_fpg;
704         UFSD("EXIT3, result %llu\n", (unsigned long long)result);
705         return result;
706 }
707
708 static u64 ufs_alloccg_block(struct inode *inode,
709                              struct ufs_cg_private_info *ucpi,
710                              u64 goal, int *err)
711 {
712         struct super_block * sb;
713         struct ufs_sb_private_info * uspi;
714         struct ufs_super_block_first * usb1;
715         struct ufs_cylinder_group * ucg;
716         u64 result, blkno;
717
718         UFSD("ENTER, goal %llu\n", (unsigned long long)goal);
719
720         sb = inode->i_sb;
721         uspi = UFS_SB(sb)->s_uspi;
722         usb1 = ubh_get_usb_first(uspi);
723         ucg = ubh_get_ucg(UCPI_UBH(ucpi));
724
725         if (goal == 0) {
726                 goal = ucpi->c_rotor;
727                 goto norot;
728         }
729         goal = ufs_blknum (goal);
730         goal = ufs_dtogd(uspi, goal);
731         
732         /*
733          * If the requested block is available, use it.
734          */
735         if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, ufs_fragstoblks(goal))) {
736                 result = goal;
737                 goto gotit;
738         }
739         
740 norot:  
741         result = ufs_bitmap_search (sb, ucpi, goal, uspi->s_fpb);
742         if (result == INVBLOCK)
743                 return INVBLOCK;
744         ucpi->c_rotor = result;
745 gotit:
746         blkno = ufs_fragstoblks(result);
747         ubh_clrblock (UCPI_UBH(ucpi), ucpi->c_freeoff, blkno);
748         if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
749                 ufs_clusteracct (sb, ucpi, blkno, -1);
750         if(DQUOT_ALLOC_BLOCK(inode, uspi->s_fpb)) {
751                 *err = -EDQUOT;
752                 return INVBLOCK;
753         }
754
755         fs32_sub(sb, &ucg->cg_cs.cs_nbfree, 1);
756         uspi->cs_total.cs_nbfree--;
757         fs32_sub(sb, &UFS_SB(sb)->fs_cs(ucpi->c_cgx).cs_nbfree, 1);
758
759         if (uspi->fs_magic != UFS2_MAGIC) {
760                 unsigned cylno = ufs_cbtocylno((unsigned)result);
761
762                 fs16_sub(sb, &ubh_cg_blks(ucpi, cylno,
763                                           ufs_cbtorpos((unsigned)result)), 1);
764                 fs32_sub(sb, &ubh_cg_blktot(ucpi, cylno), 1);
765         }
766         
767         UFSD("EXIT, result %llu\n", (unsigned long long)result);
768
769         return result;
770 }
771
772 static unsigned ubh_scanc(struct ufs_sb_private_info *uspi,
773                           struct ufs_buffer_head *ubh,
774                           unsigned begin, unsigned size,
775                           unsigned char *table, unsigned char mask)
776 {
777         unsigned rest, offset;
778         unsigned char *cp;
779         
780
781         offset = begin & ~uspi->s_fmask;
782         begin >>= uspi->s_fshift;
783         for (;;) {
784                 if ((offset + size) < uspi->s_fsize)
785                         rest = size;
786                 else
787                         rest = uspi->s_fsize - offset;
788                 size -= rest;
789                 cp = ubh->bh[begin]->b_data + offset;
790                 while ((table[*cp++] & mask) == 0 && --rest)
791                         ;
792                 if (rest || !size)
793                         break;
794                 begin++;
795                 offset = 0;
796         }
797         return (size + rest);
798 }
799
800 /*
801  * Find a block of the specified size in the specified cylinder group.
802  * @sp: pointer to super block
803  * @ucpi: pointer to cylinder group info
804  * @goal: near which block we want find new one
805  * @count: specified size
806  */
807 static u64 ufs_bitmap_search(struct super_block *sb,
808                              struct ufs_cg_private_info *ucpi,
809                              u64 goal, unsigned count)
810 {
811         /*
812          * Bit patterns for identifying fragments in the block map
813          * used as ((map & mask_arr) == want_arr)
814          */
815         static const int mask_arr[9] = {
816                 0x3, 0x7, 0xf, 0x1f, 0x3f, 0x7f, 0xff, 0x1ff, 0x3ff
817         };
818         static const int want_arr[9] = {
819                 0x0, 0x2, 0x6, 0xe, 0x1e, 0x3e, 0x7e, 0xfe, 0x1fe
820         };
821         struct ufs_sb_private_info *uspi = UFS_SB(sb)->s_uspi;
822         struct ufs_super_block_first *usb1;
823         struct ufs_cylinder_group *ucg;
824         unsigned start, length, loc;
825         unsigned pos, want, blockmap, mask, end;
826         u64 result;
827
828         UFSD("ENTER, cg %u, goal %llu, count %u\n", ucpi->c_cgx,
829              (unsigned long long)goal, count);
830
831         usb1 = ubh_get_usb_first (uspi);
832         ucg = ubh_get_ucg(UCPI_UBH(ucpi));
833
834         if (goal)
835                 start = ufs_dtogd(uspi, goal) >> 3;
836         else
837                 start = ucpi->c_frotor >> 3;
838                 
839         length = ((uspi->s_fpg + 7) >> 3) - start;
840         loc = ubh_scanc(uspi, UCPI_UBH(ucpi), ucpi->c_freeoff + start, length,
841                 (uspi->s_fpb == 8) ? ufs_fragtable_8fpb : ufs_fragtable_other,
842                 1 << (count - 1 + (uspi->s_fpb & 7))); 
843         if (loc == 0) {
844                 length = start + 1;
845                 loc = ubh_scanc(uspi, UCPI_UBH(ucpi), ucpi->c_freeoff, length,
846                                 (uspi->s_fpb == 8) ? ufs_fragtable_8fpb :
847                                 ufs_fragtable_other,
848                                 1 << (count - 1 + (uspi->s_fpb & 7)));
849                 if (loc == 0) {
850                         ufs_error(sb, "ufs_bitmap_search",
851                                   "bitmap corrupted on cg %u, start %u,"
852                                   " length %u, count %u, freeoff %u\n",
853                                   ucpi->c_cgx, start, length, count,
854                                   ucpi->c_freeoff);
855                         return INVBLOCK;
856                 }
857                 start = 0;
858         }
859         result = (start + length - loc) << 3;
860         ucpi->c_frotor = result;
861
862         /*
863          * found the byte in the map
864          */
865
866         for (end = result + 8; result < end; result += uspi->s_fpb) {
867                 blockmap = ubh_blkmap(UCPI_UBH(ucpi), ucpi->c_freeoff, result);
868                 blockmap <<= 1;
869                 mask = mask_arr[count];
870                 want = want_arr[count];
871                 for (pos = 0; pos <= uspi->s_fpb - count; pos++) {
872                         if ((blockmap & mask) == want) {
873                                 UFSD("EXIT, result %llu\n",
874                                      (unsigned long long)result);
875                                 return result + pos;
876                         }
877                         mask <<= 1;
878                         want <<= 1;
879                 }
880         }
881
882         ufs_error(sb, "ufs_bitmap_search", "block not in map on cg %u\n",
883                   ucpi->c_cgx);
884         UFSD("EXIT (FAILED)\n");
885         return INVBLOCK;
886 }
887
888 static void ufs_clusteracct(struct super_block * sb,
889         struct ufs_cg_private_info * ucpi, unsigned blkno, int cnt)
890 {
891         struct ufs_sb_private_info * uspi;
892         int i, start, end, forw, back;
893         
894         uspi = UFS_SB(sb)->s_uspi;
895         if (uspi->s_contigsumsize <= 0)
896                 return;
897
898         if (cnt > 0)
899                 ubh_setbit(UCPI_UBH(ucpi), ucpi->c_clusteroff, blkno);
900         else
901                 ubh_clrbit(UCPI_UBH(ucpi), ucpi->c_clusteroff, blkno);
902
903         /*
904          * Find the size of the cluster going forward.
905          */
906         start = blkno + 1;
907         end = start + uspi->s_contigsumsize;
908         if ( end >= ucpi->c_nclusterblks)
909                 end = ucpi->c_nclusterblks;
910         i = ubh_find_next_zero_bit (UCPI_UBH(ucpi), ucpi->c_clusteroff, end, start);
911         if (i > end)
912                 i = end;
913         forw = i - start;
914         
915         /*
916          * Find the size of the cluster going backward.
917          */
918         start = blkno - 1;
919         end = start - uspi->s_contigsumsize;
920         if (end < 0 ) 
921                 end = -1;
922         i = ubh_find_last_zero_bit (UCPI_UBH(ucpi), ucpi->c_clusteroff, start, end);
923         if ( i < end) 
924                 i = end;
925         back = start - i;
926         
927         /*
928          * Account for old cluster and the possibly new forward and
929          * back clusters.
930          */
931         i = back + forw + 1;
932         if (i > uspi->s_contigsumsize)
933                 i = uspi->s_contigsumsize;
934         fs32_add(sb, (__fs32*)ubh_get_addr(UCPI_UBH(ucpi), ucpi->c_clustersumoff + (i << 2)), cnt);
935         if (back > 0)
936                 fs32_sub(sb, (__fs32*)ubh_get_addr(UCPI_UBH(ucpi), ucpi->c_clustersumoff + (back << 2)), cnt);
937         if (forw > 0)
938                 fs32_sub(sb, (__fs32*)ubh_get_addr(UCPI_UBH(ucpi), ucpi->c_clustersumoff + (forw << 2)), cnt);
939 }
940
941
942 static unsigned char ufs_fragtable_8fpb[] = {
943         0x00, 0x01, 0x01, 0x02, 0x01, 0x01, 0x02, 0x04, 0x01, 0x01, 0x01, 0x03, 0x02, 0x03, 0x04, 0x08,
944         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x02, 0x03, 0x03, 0x02, 0x04, 0x05, 0x08, 0x10,
945         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09,
946         0x02, 0x03, 0x03, 0x02, 0x03, 0x03, 0x02, 0x06, 0x04, 0x05, 0x05, 0x06, 0x08, 0x09, 0x10, 0x20,
947         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09, 
948         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x03, 0x03, 0x03, 0x03, 0x05, 0x05, 0x09, 0x11,
949         0x02, 0x03, 0x03, 0x02, 0x03, 0x03, 0x02, 0x06, 0x03, 0x03, 0x03, 0x03, 0x02, 0x03, 0x06, 0x0A,
950         0x04, 0x05, 0x05, 0x06, 0x05, 0x05, 0x06, 0x04, 0x08, 0x09, 0x09, 0x0A, 0x10, 0x11, 0x20, 0x40,
951         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09,
952         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x03, 0x03, 0x03, 0x03, 0x05, 0x05, 0x09, 0x11,
953         0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09,
954         0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x07, 0x05, 0x05, 0x05, 0x07, 0x09, 0x09, 0x11, 0x21,
955         0x02, 0x03, 0x03, 0x02, 0x03, 0x03, 0x02, 0x06, 0x03, 0x03, 0x03, 0x03, 0x02, 0x03, 0x06, 0x0A,
956         0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x07, 0x02, 0x03, 0x03, 0x02, 0x06, 0x07, 0x0A, 0x12,
957         0x04, 0x05, 0x05, 0x06, 0x05, 0x05, 0x06, 0x04, 0x05, 0x05, 0x05, 0x07, 0x06, 0x07, 0x04, 0x0C,
958         0x08, 0x09, 0x09, 0x0A, 0x09, 0x09, 0x0A, 0x0C, 0x10, 0x11, 0x11, 0x12, 0x20, 0x21, 0x40, 0x80,
959 };
960
961 static unsigned char ufs_fragtable_other[] = {
962         0x00, 0x16, 0x16, 0x2A, 0x16, 0x16, 0x26, 0x4E, 0x16, 0x16, 0x16, 0x3E, 0x2A, 0x3E, 0x4E, 0x8A,
963         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
964         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
965         0x2A, 0x3E, 0x3E, 0x2A, 0x3E, 0x3E, 0x2E, 0x6E, 0x3E, 0x3E, 0x3E, 0x3E, 0x2A, 0x3E, 0x6E, 0xAA,
966         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
967         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
968         0x26, 0x36, 0x36, 0x2E, 0x36, 0x36, 0x26, 0x6E, 0x36, 0x36, 0x36, 0x3E, 0x2E, 0x3E, 0x6E, 0xAE,
969         0x4E, 0x5E, 0x5E, 0x6E, 0x5E, 0x5E, 0x6E, 0x4E, 0x5E, 0x5E, 0x5E, 0x7E, 0x6E, 0x7E, 0x4E, 0xCE,
970         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
971         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
972         0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
973         0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0xBE,
974         0x2A, 0x3E, 0x3E, 0x2A, 0x3E, 0x3E, 0x2E, 0x6E, 0x3E, 0x3E, 0x3E, 0x3E, 0x2A, 0x3E, 0x6E, 0xAA,
975         0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0xBE,
976         0x4E, 0x5E, 0x5E, 0x6E, 0x5E, 0x5E, 0x6E, 0x4E, 0x5E, 0x5E, 0x5E, 0x7E, 0x6E, 0x7E, 0x4E, 0xCE,
977         0x8A, 0x9E, 0x9E, 0xAA, 0x9E, 0x9E, 0xAE, 0xCE, 0x9E, 0x9E, 0x9E, 0xBE, 0xAA, 0xBE, 0xCE, 0x8A,
978 };