Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / xfs / libxfs / 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_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_sb.h"
25 #include "xfs_mount.h"
26 #include "xfs_btree.h"
27 #include "xfs_alloc_btree.h"
28 #include "xfs_alloc.h"
29 #include "xfs_extent_busy.h"
30 #include "xfs_error.h"
31 #include "xfs_trace.h"
32 #include "xfs_cksum.h"
33 #include "xfs_trans.h"
34
35
36 STATIC struct xfs_btree_cur *
37 xfs_allocbt_dup_cursor(
38         struct xfs_btree_cur    *cur)
39 {
40         return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
41                         cur->bc_private.a.agbp, cur->bc_private.a.agno,
42                         cur->bc_btnum);
43 }
44
45 STATIC void
46 xfs_allocbt_set_root(
47         struct xfs_btree_cur    *cur,
48         union xfs_btree_ptr     *ptr,
49         int                     inc)
50 {
51         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
52         struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
53         xfs_agnumber_t          seqno = be32_to_cpu(agf->agf_seqno);
54         int                     btnum = cur->bc_btnum;
55         struct xfs_perag        *pag = xfs_perag_get(cur->bc_mp, seqno);
56
57         ASSERT(ptr->s != 0);
58
59         agf->agf_roots[btnum] = ptr->s;
60         be32_add_cpu(&agf->agf_levels[btnum], inc);
61         pag->pagf_levels[btnum] += inc;
62         xfs_perag_put(pag);
63
64         xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
65 }
66
67 STATIC int
68 xfs_allocbt_alloc_block(
69         struct xfs_btree_cur    *cur,
70         union xfs_btree_ptr     *start,
71         union xfs_btree_ptr     *new,
72         int                     *stat)
73 {
74         int                     error;
75         xfs_agblock_t           bno;
76
77         XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
78
79         /* Allocate the new block from the freelist. If we can't, give up.  */
80         error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_private.a.agbp,
81                                        &bno, 1);
82         if (error) {
83                 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
84                 return error;
85         }
86
87         if (bno == NULLAGBLOCK) {
88                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
89                 *stat = 0;
90                 return 0;
91         }
92
93         xfs_extent_busy_reuse(cur->bc_mp, cur->bc_private.a.agno, bno, 1, false);
94
95         xfs_trans_agbtree_delta(cur->bc_tp, 1);
96         new->s = cpu_to_be32(bno);
97
98         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
99         *stat = 1;
100         return 0;
101 }
102
103 STATIC int
104 xfs_allocbt_free_block(
105         struct xfs_btree_cur    *cur,
106         struct xfs_buf          *bp)
107 {
108         struct xfs_buf          *agbp = cur->bc_private.a.agbp;
109         struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
110         xfs_agblock_t           bno;
111         int                     error;
112
113         bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
114         error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
115         if (error)
116                 return error;
117
118         xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
119                               XFS_EXTENT_BUSY_SKIP_DISCARD);
120         xfs_trans_agbtree_delta(cur->bc_tp, -1);
121
122         xfs_trans_binval(cur->bc_tp, bp);
123         return 0;
124 }
125
126 /*
127  * Update the longest extent in the AGF
128  */
129 STATIC void
130 xfs_allocbt_update_lastrec(
131         struct xfs_btree_cur    *cur,
132         struct xfs_btree_block  *block,
133         union xfs_btree_rec     *rec,
134         int                     ptr,
135         int                     reason)
136 {
137         struct xfs_agf          *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
138         xfs_agnumber_t          seqno = be32_to_cpu(agf->agf_seqno);
139         struct xfs_perag        *pag;
140         __be32                  len;
141         int                     numrecs;
142
143         ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
144
145         switch (reason) {
146         case LASTREC_UPDATE:
147                 /*
148                  * If this is the last leaf block and it's the last record,
149                  * then update the size of the longest extent in the AG.
150                  */
151                 if (ptr != xfs_btree_get_numrecs(block))
152                         return;
153                 len = rec->alloc.ar_blockcount;
154                 break;
155         case LASTREC_INSREC:
156                 if (be32_to_cpu(rec->alloc.ar_blockcount) <=
157                     be32_to_cpu(agf->agf_longest))
158                         return;
159                 len = rec->alloc.ar_blockcount;
160                 break;
161         case LASTREC_DELREC:
162                 numrecs = xfs_btree_get_numrecs(block);
163                 if (ptr <= numrecs)
164                         return;
165                 ASSERT(ptr == numrecs + 1);
166
167                 if (numrecs) {
168                         xfs_alloc_rec_t *rrp;
169
170                         rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
171                         len = rrp->ar_blockcount;
172                 } else {
173                         len = 0;
174                 }
175
176                 break;
177         default:
178                 ASSERT(0);
179                 return;
180         }
181
182         agf->agf_longest = len;
183         pag = xfs_perag_get(cur->bc_mp, seqno);
184         pag->pagf_longest = be32_to_cpu(len);
185         xfs_perag_put(pag);
186         xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp, XFS_AGF_LONGEST);
187 }
188
189 STATIC int
190 xfs_allocbt_get_minrecs(
191         struct xfs_btree_cur    *cur,
192         int                     level)
193 {
194         return cur->bc_mp->m_alloc_mnr[level != 0];
195 }
196
197 STATIC int
198 xfs_allocbt_get_maxrecs(
199         struct xfs_btree_cur    *cur,
200         int                     level)
201 {
202         return cur->bc_mp->m_alloc_mxr[level != 0];
203 }
204
205 STATIC void
206 xfs_allocbt_init_key_from_rec(
207         union xfs_btree_key     *key,
208         union xfs_btree_rec     *rec)
209 {
210         ASSERT(rec->alloc.ar_startblock != 0);
211
212         key->alloc.ar_startblock = rec->alloc.ar_startblock;
213         key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
214 }
215
216 STATIC void
217 xfs_allocbt_init_rec_from_key(
218         union xfs_btree_key     *key,
219         union xfs_btree_rec     *rec)
220 {
221         ASSERT(key->alloc.ar_startblock != 0);
222
223         rec->alloc.ar_startblock = key->alloc.ar_startblock;
224         rec->alloc.ar_blockcount = key->alloc.ar_blockcount;
225 }
226
227 STATIC void
228 xfs_allocbt_init_rec_from_cur(
229         struct xfs_btree_cur    *cur,
230         union xfs_btree_rec     *rec)
231 {
232         ASSERT(cur->bc_rec.a.ar_startblock != 0);
233
234         rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
235         rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
236 }
237
238 STATIC void
239 xfs_allocbt_init_ptr_from_cur(
240         struct xfs_btree_cur    *cur,
241         union xfs_btree_ptr     *ptr)
242 {
243         struct xfs_agf          *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
244
245         ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
246         ASSERT(agf->agf_roots[cur->bc_btnum] != 0);
247
248         ptr->s = agf->agf_roots[cur->bc_btnum];
249 }
250
251 STATIC __int64_t
252 xfs_allocbt_key_diff(
253         struct xfs_btree_cur    *cur,
254         union xfs_btree_key     *key)
255 {
256         xfs_alloc_rec_incore_t  *rec = &cur->bc_rec.a;
257         xfs_alloc_key_t         *kp = &key->alloc;
258         __int64_t               diff;
259
260         if (cur->bc_btnum == XFS_BTNUM_BNO) {
261                 return (__int64_t)be32_to_cpu(kp->ar_startblock) -
262                                 rec->ar_startblock;
263         }
264
265         diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
266         if (diff)
267                 return diff;
268
269         return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
270 }
271
272 static bool
273 xfs_allocbt_verify(
274         struct xfs_buf          *bp)
275 {
276         struct xfs_mount        *mp = bp->b_target->bt_mount;
277         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
278         struct xfs_perag        *pag = bp->b_pag;
279         unsigned int            level;
280
281         /*
282          * magic number and level verification
283          *
284          * During growfs operations, we can't verify the exact level or owner as
285          * the perag is not fully initialised and hence not attached to the
286          * buffer.  In this case, check against the maximum tree depth.
287          *
288          * Similarly, during log recovery we will have a perag structure
289          * attached, but the agf information will not yet have been initialised
290          * from the on disk AGF. Again, we can only check against maximum limits
291          * in this case.
292          */
293         level = be16_to_cpu(block->bb_level);
294         switch (block->bb_magic) {
295         case cpu_to_be32(XFS_ABTB_CRC_MAGIC):
296                 if (!xfs_sb_version_hascrc(&mp->m_sb))
297                         return false;
298                 if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_uuid))
299                         return false;
300                 if (block->bb_u.s.bb_blkno != cpu_to_be64(bp->b_bn))
301                         return false;
302                 if (pag &&
303                     be32_to_cpu(block->bb_u.s.bb_owner) != pag->pag_agno)
304                         return false;
305                 /* fall through */
306         case cpu_to_be32(XFS_ABTB_MAGIC):
307                 if (pag && pag->pagf_init) {
308                         if (level >= pag->pagf_levels[XFS_BTNUM_BNOi])
309                                 return false;
310                 } else if (level >= mp->m_ag_maxlevels)
311                         return false;
312                 break;
313         case cpu_to_be32(XFS_ABTC_CRC_MAGIC):
314                 if (!xfs_sb_version_hascrc(&mp->m_sb))
315                         return false;
316                 if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_uuid))
317                         return false;
318                 if (block->bb_u.s.bb_blkno != cpu_to_be64(bp->b_bn))
319                         return false;
320                 if (pag &&
321                     be32_to_cpu(block->bb_u.s.bb_owner) != pag->pag_agno)
322                         return false;
323                 /* fall through */
324         case cpu_to_be32(XFS_ABTC_MAGIC):
325                 if (pag && pag->pagf_init) {
326                         if (level >= pag->pagf_levels[XFS_BTNUM_CNTi])
327                                 return false;
328                 } else if (level >= mp->m_ag_maxlevels)
329                         return false;
330                 break;
331         default:
332                 return false;
333         }
334
335         /* numrecs verification */
336         if (be16_to_cpu(block->bb_numrecs) > mp->m_alloc_mxr[level != 0])
337                 return false;
338
339         /* sibling pointer verification */
340         if (!block->bb_u.s.bb_leftsib ||
341             (be32_to_cpu(block->bb_u.s.bb_leftsib) >= mp->m_sb.sb_agblocks &&
342              block->bb_u.s.bb_leftsib != cpu_to_be32(NULLAGBLOCK)))
343                 return false;
344         if (!block->bb_u.s.bb_rightsib ||
345             (be32_to_cpu(block->bb_u.s.bb_rightsib) >= mp->m_sb.sb_agblocks &&
346              block->bb_u.s.bb_rightsib != cpu_to_be32(NULLAGBLOCK)))
347                 return false;
348
349         return true;
350 }
351
352 static void
353 xfs_allocbt_read_verify(
354         struct xfs_buf  *bp)
355 {
356         if (!xfs_btree_sblock_verify_crc(bp))
357                 xfs_buf_ioerror(bp, -EFSBADCRC);
358         else if (!xfs_allocbt_verify(bp))
359                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
360
361         if (bp->b_error) {
362                 trace_xfs_btree_corrupt(bp, _RET_IP_);
363                 xfs_verifier_error(bp);
364         }
365 }
366
367 static void
368 xfs_allocbt_write_verify(
369         struct xfs_buf  *bp)
370 {
371         if (!xfs_allocbt_verify(bp)) {
372                 trace_xfs_btree_corrupt(bp, _RET_IP_);
373                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
374                 xfs_verifier_error(bp);
375                 return;
376         }
377         xfs_btree_sblock_calc_crc(bp);
378
379 }
380
381 const struct xfs_buf_ops xfs_allocbt_buf_ops = {
382         .verify_read = xfs_allocbt_read_verify,
383         .verify_write = xfs_allocbt_write_verify,
384 };
385
386
387 #if defined(DEBUG) || defined(XFS_WARN)
388 STATIC int
389 xfs_allocbt_keys_inorder(
390         struct xfs_btree_cur    *cur,
391         union xfs_btree_key     *k1,
392         union xfs_btree_key     *k2)
393 {
394         if (cur->bc_btnum == XFS_BTNUM_BNO) {
395                 return be32_to_cpu(k1->alloc.ar_startblock) <
396                        be32_to_cpu(k2->alloc.ar_startblock);
397         } else {
398                 return be32_to_cpu(k1->alloc.ar_blockcount) <
399                         be32_to_cpu(k2->alloc.ar_blockcount) ||
400                         (k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
401                          be32_to_cpu(k1->alloc.ar_startblock) <
402                          be32_to_cpu(k2->alloc.ar_startblock));
403         }
404 }
405
406 STATIC int
407 xfs_allocbt_recs_inorder(
408         struct xfs_btree_cur    *cur,
409         union xfs_btree_rec     *r1,
410         union xfs_btree_rec     *r2)
411 {
412         if (cur->bc_btnum == XFS_BTNUM_BNO) {
413                 return be32_to_cpu(r1->alloc.ar_startblock) +
414                         be32_to_cpu(r1->alloc.ar_blockcount) <=
415                         be32_to_cpu(r2->alloc.ar_startblock);
416         } else {
417                 return be32_to_cpu(r1->alloc.ar_blockcount) <
418                         be32_to_cpu(r2->alloc.ar_blockcount) ||
419                         (r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
420                          be32_to_cpu(r1->alloc.ar_startblock) <
421                          be32_to_cpu(r2->alloc.ar_startblock));
422         }
423 }
424 #endif  /* DEBUG */
425
426 static const struct xfs_btree_ops xfs_allocbt_ops = {
427         .rec_len                = sizeof(xfs_alloc_rec_t),
428         .key_len                = sizeof(xfs_alloc_key_t),
429
430         .dup_cursor             = xfs_allocbt_dup_cursor,
431         .set_root               = xfs_allocbt_set_root,
432         .alloc_block            = xfs_allocbt_alloc_block,
433         .free_block             = xfs_allocbt_free_block,
434         .update_lastrec         = xfs_allocbt_update_lastrec,
435         .get_minrecs            = xfs_allocbt_get_minrecs,
436         .get_maxrecs            = xfs_allocbt_get_maxrecs,
437         .init_key_from_rec      = xfs_allocbt_init_key_from_rec,
438         .init_rec_from_key      = xfs_allocbt_init_rec_from_key,
439         .init_rec_from_cur      = xfs_allocbt_init_rec_from_cur,
440         .init_ptr_from_cur      = xfs_allocbt_init_ptr_from_cur,
441         .key_diff               = xfs_allocbt_key_diff,
442         .buf_ops                = &xfs_allocbt_buf_ops,
443 #if defined(DEBUG) || defined(XFS_WARN)
444         .keys_inorder           = xfs_allocbt_keys_inorder,
445         .recs_inorder           = xfs_allocbt_recs_inorder,
446 #endif
447 };
448
449 /*
450  * Allocate a new allocation btree cursor.
451  */
452 struct xfs_btree_cur *                  /* new alloc btree cursor */
453 xfs_allocbt_init_cursor(
454         struct xfs_mount        *mp,            /* file system mount point */
455         struct xfs_trans        *tp,            /* transaction pointer */
456         struct xfs_buf          *agbp,          /* buffer for agf structure */
457         xfs_agnumber_t          agno,           /* allocation group number */
458         xfs_btnum_t             btnum)          /* btree identifier */
459 {
460         struct xfs_agf          *agf = XFS_BUF_TO_AGF(agbp);
461         struct xfs_btree_cur    *cur;
462
463         ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);
464
465         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
466
467         cur->bc_tp = tp;
468         cur->bc_mp = mp;
469         cur->bc_btnum = btnum;
470         cur->bc_blocklog = mp->m_sb.sb_blocklog;
471         cur->bc_ops = &xfs_allocbt_ops;
472
473         if (btnum == XFS_BTNUM_CNT) {
474                 cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
475                 cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
476         } else {
477                 cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]);
478         }
479
480         cur->bc_private.a.agbp = agbp;
481         cur->bc_private.a.agno = agno;
482
483         if (xfs_sb_version_hascrc(&mp->m_sb))
484                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
485
486         return cur;
487 }
488
489 /*
490  * Calculate number of records in an alloc btree block.
491  */
492 int
493 xfs_allocbt_maxrecs(
494         struct xfs_mount        *mp,
495         int                     blocklen,
496         int                     leaf)
497 {
498         blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
499
500         if (leaf)
501                 return blocklen / sizeof(xfs_alloc_rec_t);
502         return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
503 }