Merge branch 'tracing-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6] / fs / xfs / xfs_alloc_btree.c
1 /*
2  * Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_types.h"
21 #include "xfs_bit.h"
22 #include "xfs_log.h"
23 #include "xfs_inum.h"
24 #include "xfs_trans.h"
25 #include "xfs_sb.h"
26 #include "xfs_ag.h"
27 #include "xfs_dir2.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_btree.h"
38 #include "xfs_btree_trace.h"
39 #include "xfs_ialloc.h"
40 #include "xfs_alloc.h"
41 #include "xfs_error.h"
42
43
44 STATIC struct xfs_btree_cur *
45 xfs_allocbt_dup_cursor(
46         struct xfs_btree_cur    *cur)
47 {
48         return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
49                         cur->bc_private.a.agbp, cur->bc_private.a.agno,
50                         cur->bc_btnum);
51 }
52
53 STATIC void
54 xfs_allocbt_set_root(
55         struct xfs_btree_cur    *cur,
56         union xfs_btree_ptr     *ptr,
57         int                     inc)
58 {
59         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
60         struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
61         xfs_agnumber_t          seqno = be32_to_cpu(agf->agf_seqno);
62         int                     btnum = cur->bc_btnum;
63
64         ASSERT(ptr->s != 0);
65
66         agf->agf_roots[btnum] = ptr->s;
67         be32_add_cpu(&agf->agf_levels[btnum], inc);
68         cur->bc_mp->m_perag[seqno].pagf_levels[btnum] += inc;
69
70         xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
71 }
72
73 STATIC int
74 xfs_allocbt_alloc_block(
75         struct xfs_btree_cur    *cur,
76         union xfs_btree_ptr     *start,
77         union xfs_btree_ptr     *new,
78         int                     length,
79         int                     *stat)
80 {
81         int                     error;
82         xfs_agblock_t           bno;
83
84         XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
85
86         /* Allocate the new block from the freelist. If we can't, give up.  */
87         error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_private.a.agbp,
88                                        &bno, 1);
89         if (error) {
90                 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
91                 return error;
92         }
93
94         if (bno == NULLAGBLOCK) {
95                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
96                 *stat = 0;
97                 return 0;
98         }
99
100         xfs_trans_agbtree_delta(cur->bc_tp, 1);
101         new->s = cpu_to_be32(bno);
102
103         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
104         *stat = 1;
105         return 0;
106 }
107
108 STATIC int
109 xfs_allocbt_free_block(
110         struct xfs_btree_cur    *cur,
111         struct xfs_buf          *bp)
112 {
113         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
114         struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
115         xfs_agblock_t           bno;
116         int                     error;
117
118         bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
119         error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
120         if (error)
121                 return error;
122
123         /*
124          * Since blocks move to the free list without the coordination used in
125          * xfs_bmap_finish, we can't allow block to be available for
126          * reallocation and non-transaction writing (user data) until we know
127          * that the transaction that moved it to the free list is permanently
128          * on disk. We track the blocks by declaring these blocks as "busy";
129          * the busy list is maintained on a per-ag basis and each transaction
130          * records which entries should be removed when the iclog commits to
131          * disk. If a busy block is allocated, the iclog is pushed up to the
132          * LSN that freed the block.
133          */
134         xfs_alloc_mark_busy(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1);
135         xfs_trans_agbtree_delta(cur->bc_tp, -1);
136         return 0;
137 }
138
139 /*
140  * Update the longest extent in the AGF
141  */
142 STATIC void
143 xfs_allocbt_update_lastrec(
144         struct xfs_btree_cur    *cur,
145         struct xfs_btree_block  *block,
146         union xfs_btree_rec     *rec,
147         int                     ptr,
148         int                     reason)
149 {
150         struct xfs_agf          *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
151         xfs_agnumber_t          seqno = be32_to_cpu(agf->agf_seqno);
152         __be32                  len;
153         int                     numrecs;
154
155         ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
156
157         switch (reason) {
158         case LASTREC_UPDATE:
159                 /*
160                  * If this is the last leaf block and it's the last record,
161                  * then update the size of the longest extent in the AG.
162                  */
163                 if (ptr != xfs_btree_get_numrecs(block))
164                         return;
165                 len = rec->alloc.ar_blockcount;
166                 break;
167         case LASTREC_INSREC:
168                 if (be32_to_cpu(rec->alloc.ar_blockcount) <=
169                     be32_to_cpu(agf->agf_longest))
170                         return;
171                 len = rec->alloc.ar_blockcount;
172                 break;
173         case LASTREC_DELREC:
174                 numrecs = xfs_btree_get_numrecs(block);
175                 if (ptr <= numrecs)
176                         return;
177                 ASSERT(ptr == numrecs + 1);
178
179                 if (numrecs) {
180                         xfs_alloc_rec_t *rrp;
181
182                         rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
183                         len = rrp->ar_blockcount;
184                 } else {
185                         len = 0;
186                 }
187
188                 break;
189         default:
190                 ASSERT(0);
191                 return;
192         }
193
194         agf->agf_longest = len;
195         cur->bc_mp->m_perag[seqno].pagf_longest = be32_to_cpu(len);
196         xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp, XFS_AGF_LONGEST);
197 }
198
199 STATIC int
200 xfs_allocbt_get_minrecs(
201         struct xfs_btree_cur    *cur,
202         int                     level)
203 {
204         return cur->bc_mp->m_alloc_mnr[level != 0];
205 }
206
207 STATIC int
208 xfs_allocbt_get_maxrecs(
209         struct xfs_btree_cur    *cur,
210         int                     level)
211 {
212         return cur->bc_mp->m_alloc_mxr[level != 0];
213 }
214
215 STATIC void
216 xfs_allocbt_init_key_from_rec(
217         union xfs_btree_key     *key,
218         union xfs_btree_rec     *rec)
219 {
220         ASSERT(rec->alloc.ar_startblock != 0);
221
222         key->alloc.ar_startblock = rec->alloc.ar_startblock;
223         key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
224 }
225
226 STATIC void
227 xfs_allocbt_init_rec_from_key(
228         union xfs_btree_key     *key,
229         union xfs_btree_rec     *rec)
230 {
231         ASSERT(key->alloc.ar_startblock != 0);
232
233         rec->alloc.ar_startblock = key->alloc.ar_startblock;
234         rec->alloc.ar_blockcount = key->alloc.ar_blockcount;
235 }
236
237 STATIC void
238 xfs_allocbt_init_rec_from_cur(
239         struct xfs_btree_cur    *cur,
240         union xfs_btree_rec     *rec)
241 {
242         ASSERT(cur->bc_rec.a.ar_startblock != 0);
243
244         rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
245         rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
246 }
247
248 STATIC void
249 xfs_allocbt_init_ptr_from_cur(
250         struct xfs_btree_cur    *cur,
251         union xfs_btree_ptr     *ptr)
252 {
253         struct xfs_agf          *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
254
255         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
256         ASSERT(agf->agf_roots[cur->bc_btnum] != 0);
257
258         ptr->s = agf->agf_roots[cur->bc_btnum];
259 }
260
261 STATIC __int64_t
262 xfs_allocbt_key_diff(
263         struct xfs_btree_cur    *cur,
264         union xfs_btree_key     *key)
265 {
266         xfs_alloc_rec_incore_t  *rec = &cur->bc_rec.a;
267         xfs_alloc_key_t         *kp = &key->alloc;
268         __int64_t               diff;
269
270         if (cur->bc_btnum == XFS_BTNUM_BNO) {
271                 return (__int64_t)be32_to_cpu(kp->ar_startblock) -
272                                 rec->ar_startblock;
273         }
274
275         diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
276         if (diff)
277                 return diff;
278
279         return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
280 }
281
282 STATIC int
283 xfs_allocbt_kill_root(
284         struct xfs_btree_cur    *cur,
285         struct xfs_buf          *bp,
286         int                     level,
287         union xfs_btree_ptr     *newroot)
288 {
289         int                     error;
290
291         XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
292         XFS_BTREE_STATS_INC(cur, killroot);
293
294         /*
295          * Update the root pointer, decreasing the level by 1 and then
296          * free the old root.
297          */
298         xfs_allocbt_set_root(cur, newroot, -1);
299         error = xfs_allocbt_free_block(cur, bp);
300         if (error) {
301                 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
302                 return error;
303         }
304
305         XFS_BTREE_STATS_INC(cur, free);
306
307         xfs_btree_setbuf(cur, level, NULL);
308         cur->bc_nlevels--;
309
310         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
311         return 0;
312 }
313
314 #ifdef DEBUG
315 STATIC int
316 xfs_allocbt_keys_inorder(
317         struct xfs_btree_cur    *cur,
318         union xfs_btree_key     *k1,
319         union xfs_btree_key     *k2)
320 {
321         if (cur->bc_btnum == XFS_BTNUM_BNO) {
322                 return be32_to_cpu(k1->alloc.ar_startblock) <
323                        be32_to_cpu(k2->alloc.ar_startblock);
324         } else {
325                 return be32_to_cpu(k1->alloc.ar_blockcount) <
326                         be32_to_cpu(k2->alloc.ar_blockcount) ||
327                         (k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
328                          be32_to_cpu(k1->alloc.ar_startblock) <
329                          be32_to_cpu(k2->alloc.ar_startblock));
330         }
331 }
332
333 STATIC int
334 xfs_allocbt_recs_inorder(
335         struct xfs_btree_cur    *cur,
336         union xfs_btree_rec     *r1,
337         union xfs_btree_rec     *r2)
338 {
339         if (cur->bc_btnum == XFS_BTNUM_BNO) {
340                 return be32_to_cpu(r1->alloc.ar_startblock) +
341                         be32_to_cpu(r1->alloc.ar_blockcount) <=
342                         be32_to_cpu(r2->alloc.ar_startblock);
343         } else {
344                 return be32_to_cpu(r1->alloc.ar_blockcount) <
345                         be32_to_cpu(r2->alloc.ar_blockcount) ||
346                         (r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
347                          be32_to_cpu(r1->alloc.ar_startblock) <
348                          be32_to_cpu(r2->alloc.ar_startblock));
349         }
350 }
351 #endif  /* DEBUG */
352
353 #ifdef XFS_BTREE_TRACE
354 ktrace_t        *xfs_allocbt_trace_buf;
355
356 STATIC void
357 xfs_allocbt_trace_enter(
358         struct xfs_btree_cur    *cur,
359         const char              *func,
360         char                    *s,
361         int                     type,
362         int                     line,
363         __psunsigned_t          a0,
364         __psunsigned_t          a1,
365         __psunsigned_t          a2,
366         __psunsigned_t          a3,
367         __psunsigned_t          a4,
368         __psunsigned_t          a5,
369         __psunsigned_t          a6,
370         __psunsigned_t          a7,
371         __psunsigned_t          a8,
372         __psunsigned_t          a9,
373         __psunsigned_t          a10)
374 {
375         ktrace_enter(xfs_allocbt_trace_buf, (void *)(__psint_t)type,
376                 (void *)func, (void *)s, NULL, (void *)cur,
377                 (void *)a0, (void *)a1, (void *)a2, (void *)a3,
378                 (void *)a4, (void *)a5, (void *)a6, (void *)a7,
379                 (void *)a8, (void *)a9, (void *)a10);
380 }
381
382 STATIC void
383 xfs_allocbt_trace_cursor(
384         struct xfs_btree_cur    *cur,
385         __uint32_t              *s0,
386         __uint64_t              *l0,
387         __uint64_t              *l1)
388 {
389         *s0 = cur->bc_private.a.agno;
390         *l0 = cur->bc_rec.a.ar_startblock;
391         *l1 = cur->bc_rec.a.ar_blockcount;
392 }
393
394 STATIC void
395 xfs_allocbt_trace_key(
396         struct xfs_btree_cur    *cur,
397         union xfs_btree_key     *key,
398         __uint64_t              *l0,
399         __uint64_t              *l1)
400 {
401         *l0 = be32_to_cpu(key->alloc.ar_startblock);
402         *l1 = be32_to_cpu(key->alloc.ar_blockcount);
403 }
404
405 STATIC void
406 xfs_allocbt_trace_record(
407         struct xfs_btree_cur    *cur,
408         union xfs_btree_rec     *rec,
409         __uint64_t              *l0,
410         __uint64_t              *l1,
411         __uint64_t              *l2)
412 {
413         *l0 = be32_to_cpu(rec->alloc.ar_startblock);
414         *l1 = be32_to_cpu(rec->alloc.ar_blockcount);
415         *l2 = 0;
416 }
417 #endif /* XFS_BTREE_TRACE */
418
419 static const struct xfs_btree_ops xfs_allocbt_ops = {
420         .rec_len                = sizeof(xfs_alloc_rec_t),
421         .key_len                = sizeof(xfs_alloc_key_t),
422
423         .dup_cursor             = xfs_allocbt_dup_cursor,
424         .set_root               = xfs_allocbt_set_root,
425         .kill_root              = xfs_allocbt_kill_root,
426         .alloc_block            = xfs_allocbt_alloc_block,
427         .free_block             = xfs_allocbt_free_block,
428         .update_lastrec         = xfs_allocbt_update_lastrec,
429         .get_minrecs            = xfs_allocbt_get_minrecs,
430         .get_maxrecs            = xfs_allocbt_get_maxrecs,
431         .init_key_from_rec      = xfs_allocbt_init_key_from_rec,
432         .init_rec_from_key      = xfs_allocbt_init_rec_from_key,
433         .init_rec_from_cur      = xfs_allocbt_init_rec_from_cur,
434         .init_ptr_from_cur      = xfs_allocbt_init_ptr_from_cur,
435         .key_diff               = xfs_allocbt_key_diff,
436
437 #ifdef DEBUG
438         .keys_inorder           = xfs_allocbt_keys_inorder,
439         .recs_inorder           = xfs_allocbt_recs_inorder,
440 #endif
441
442 #ifdef XFS_BTREE_TRACE
443         .trace_enter            = xfs_allocbt_trace_enter,
444         .trace_cursor           = xfs_allocbt_trace_cursor,
445         .trace_key              = xfs_allocbt_trace_key,
446         .trace_record           = xfs_allocbt_trace_record,
447 #endif
448 };
449
450 /*
451  * Allocate a new allocation btree cursor.
452  */
453 struct xfs_btree_cur *                  /* new alloc btree cursor */
454 xfs_allocbt_init_cursor(
455         struct xfs_mount        *mp,            /* file system mount point */
456         struct xfs_trans        *tp,            /* transaction pointer */
457         struct xfs_buf          *agbp,          /* buffer for agf structure */
458         xfs_agnumber_t          agno,           /* allocation group number */
459         xfs_btnum_t             btnum)          /* btree identifier */
460 {
461         struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
462         struct xfs_btree_cur    *cur;
463
464         ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);
465
466         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
467
468         cur->bc_tp = tp;
469         cur->bc_mp = mp;
470         cur->bc_nlevels = be32_to_cpu(agf->agf_levels[btnum]);
471         cur->bc_btnum = btnum;
472         cur->bc_blocklog = mp->m_sb.sb_blocklog;
473
474         cur->bc_ops = &xfs_allocbt_ops;
475         if (btnum == XFS_BTNUM_CNT)
476                 cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
477
478         cur->bc_private.a.agbp = agbp;
479         cur->bc_private.a.agno = agno;
480
481         return cur;
482 }
483
484 /*
485  * Calculate number of records in an alloc btree block.
486  */
487 int
488 xfs_allocbt_maxrecs(
489         struct xfs_mount        *mp,
490         int                     blocklen,
491         int                     leaf)
492 {
493         blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
494
495         if (leaf)
496                 return blocklen / sizeof(xfs_alloc_rec_t);
497         return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
498 }