]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - drivers/md/persistent-data/dm-space-map-metadata.c
Merge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[karo-tx-linux.git] / drivers / md / persistent-data / dm-space-map-metadata.c
1 /*
2  * Copyright (C) 2011 Red Hat, Inc.
3  *
4  * This file is released under the GPL.
5  */
6
7 #include "dm-space-map.h"
8 #include "dm-space-map-common.h"
9 #include "dm-space-map-metadata.h"
10
11 #include <linux/list.h>
12 #include <linux/slab.h>
13 #include <linux/device-mapper.h>
14
15 #define DM_MSG_PREFIX "space map metadata"
16
17 /*----------------------------------------------------------------*/
18
19 /*
20  * An edge triggered threshold.
21  */
22 struct threshold {
23         bool threshold_set;
24         bool value_set;
25         dm_block_t threshold;
26         dm_block_t current_value;
27         dm_sm_threshold_fn fn;
28         void *context;
29 };
30
31 static void threshold_init(struct threshold *t)
32 {
33         t->threshold_set = false;
34         t->value_set = false;
35 }
36
37 static void set_threshold(struct threshold *t, dm_block_t value,
38                           dm_sm_threshold_fn fn, void *context)
39 {
40         t->threshold_set = true;
41         t->threshold = value;
42         t->fn = fn;
43         t->context = context;
44 }
45
46 static bool below_threshold(struct threshold *t, dm_block_t value)
47 {
48         return t->threshold_set && value <= t->threshold;
49 }
50
51 static bool threshold_already_triggered(struct threshold *t)
52 {
53         return t->value_set && below_threshold(t, t->current_value);
54 }
55
56 static void check_threshold(struct threshold *t, dm_block_t value)
57 {
58         if (below_threshold(t, value) &&
59             !threshold_already_triggered(t))
60                 t->fn(t->context);
61
62         t->value_set = true;
63         t->current_value = value;
64 }
65
66 /*----------------------------------------------------------------*/
67
68 /*
69  * Space map interface.
70  *
71  * The low level disk format is written using the standard btree and
72  * transaction manager.  This means that performing disk operations may
73  * cause us to recurse into the space map in order to allocate new blocks.
74  * For this reason we have a pool of pre-allocated blocks large enough to
75  * service any metadata_ll_disk operation.
76  */
77
78 /*
79  * FIXME: we should calculate this based on the size of the device.
80  * Only the metadata space map needs this functionality.
81  */
82 #define MAX_RECURSIVE_ALLOCATIONS 1024
83
84 enum block_op_type {
85         BOP_INC,
86         BOP_DEC
87 };
88
89 struct block_op {
90         enum block_op_type type;
91         dm_block_t block;
92 };
93
94 struct bop_ring_buffer {
95         unsigned begin;
96         unsigned end;
97         struct block_op bops[MAX_RECURSIVE_ALLOCATIONS + 1];
98 };
99
100 static void brb_init(struct bop_ring_buffer *brb)
101 {
102         brb->begin = 0;
103         brb->end = 0;
104 }
105
106 static bool brb_empty(struct bop_ring_buffer *brb)
107 {
108         return brb->begin == brb->end;
109 }
110
111 static unsigned brb_next(struct bop_ring_buffer *brb, unsigned old)
112 {
113         unsigned r = old + 1;
114         return (r >= (sizeof(brb->bops) / sizeof(*brb->bops))) ? 0 : r;
115 }
116
117 static int brb_push(struct bop_ring_buffer *brb,
118                     enum block_op_type type, dm_block_t b)
119 {
120         struct block_op *bop;
121         unsigned next = brb_next(brb, brb->end);
122
123         /*
124          * We don't allow the last bop to be filled, this way we can
125          * differentiate between full and empty.
126          */
127         if (next == brb->begin)
128                 return -ENOMEM;
129
130         bop = brb->bops + brb->end;
131         bop->type = type;
132         bop->block = b;
133
134         brb->end = next;
135
136         return 0;
137 }
138
139 static int brb_peek(struct bop_ring_buffer *brb, struct block_op *result)
140 {
141         struct block_op *bop;
142
143         if (brb_empty(brb))
144                 return -ENODATA;
145
146         bop = brb->bops + brb->begin;
147         result->type = bop->type;
148         result->block = bop->block;
149
150         return 0;
151 }
152
153 static int brb_pop(struct bop_ring_buffer *brb)
154 {
155         struct block_op *bop;
156
157         if (brb_empty(brb))
158                 return -ENODATA;
159
160         bop = brb->bops + brb->begin;
161         brb->begin = brb_next(brb, brb->begin);
162
163         return 0;
164 }
165
166 /*----------------------------------------------------------------*/
167
168 struct sm_metadata {
169         struct dm_space_map sm;
170
171         struct ll_disk ll;
172         struct ll_disk old_ll;
173
174         dm_block_t begin;
175
176         unsigned recursion_count;
177         unsigned allocated_this_transaction;
178         struct bop_ring_buffer uncommitted;
179
180         struct threshold threshold;
181 };
182
183 static int add_bop(struct sm_metadata *smm, enum block_op_type type, dm_block_t b)
184 {
185         int r = brb_push(&smm->uncommitted, type, b);
186
187         if (r) {
188                 DMERR("too many recursive allocations");
189                 return -ENOMEM;
190         }
191
192         return 0;
193 }
194
195 static int commit_bop(struct sm_metadata *smm, struct block_op *op)
196 {
197         int r = 0;
198         enum allocation_event ev;
199
200         switch (op->type) {
201         case BOP_INC:
202                 r = sm_ll_inc(&smm->ll, op->block, &ev);
203                 break;
204
205         case BOP_DEC:
206                 r = sm_ll_dec(&smm->ll, op->block, &ev);
207                 break;
208         }
209
210         return r;
211 }
212
213 static void in(struct sm_metadata *smm)
214 {
215         smm->recursion_count++;
216 }
217
218 static int apply_bops(struct sm_metadata *smm)
219 {
220         int r = 0;
221
222         while (!brb_empty(&smm->uncommitted)) {
223                 struct block_op bop;
224
225                 r = brb_peek(&smm->uncommitted, &bop);
226                 if (r) {
227                         DMERR("bug in bop ring buffer");
228                         break;
229                 }
230
231                 r = commit_bop(smm, &bop);
232                 if (r)
233                         break;
234
235                 brb_pop(&smm->uncommitted);
236         }
237
238         return r;
239 }
240
241 static int out(struct sm_metadata *smm)
242 {
243         int r = 0;
244
245         /*
246          * If we're not recursing then very bad things are happening.
247          */
248         if (!smm->recursion_count) {
249                 DMERR("lost track of recursion depth");
250                 return -ENOMEM;
251         }
252
253         if (smm->recursion_count == 1)
254                 apply_bops(smm);
255
256         smm->recursion_count--;
257
258         return r;
259 }
260
261 /*
262  * When using the out() function above, we often want to combine an error
263  * code for the operation run in the recursive context with that from
264  * out().
265  */
266 static int combine_errors(int r1, int r2)
267 {
268         return r1 ? r1 : r2;
269 }
270
271 static int recursing(struct sm_metadata *smm)
272 {
273         return smm->recursion_count;
274 }
275
276 static void sm_metadata_destroy(struct dm_space_map *sm)
277 {
278         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
279
280         kfree(smm);
281 }
282
283 static int sm_metadata_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
284 {
285         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
286
287         *count = smm->ll.nr_blocks;
288
289         return 0;
290 }
291
292 static int sm_metadata_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
293 {
294         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
295
296         *count = smm->old_ll.nr_blocks - smm->old_ll.nr_allocated -
297                  smm->allocated_this_transaction;
298
299         return 0;
300 }
301
302 static int sm_metadata_get_count(struct dm_space_map *sm, dm_block_t b,
303                                  uint32_t *result)
304 {
305         int r;
306         unsigned i;
307         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
308         unsigned adjustment = 0;
309
310         /*
311          * We may have some uncommitted adjustments to add.  This list
312          * should always be really short.
313          */
314         for (i = smm->uncommitted.begin;
315              i != smm->uncommitted.end;
316              i = brb_next(&smm->uncommitted, i)) {
317                 struct block_op *op = smm->uncommitted.bops + i;
318
319                 if (op->block != b)
320                         continue;
321
322                 switch (op->type) {
323                 case BOP_INC:
324                         adjustment++;
325                         break;
326
327                 case BOP_DEC:
328                         adjustment--;
329                         break;
330                 }
331         }
332
333         r = sm_ll_lookup(&smm->ll, b, result);
334         if (r)
335                 return r;
336
337         *result += adjustment;
338
339         return 0;
340 }
341
342 static int sm_metadata_count_is_more_than_one(struct dm_space_map *sm,
343                                               dm_block_t b, int *result)
344 {
345         int r, adjustment = 0;
346         unsigned i;
347         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
348         uint32_t rc;
349
350         /*
351          * We may have some uncommitted adjustments to add.  This list
352          * should always be really short.
353          */
354         for (i = smm->uncommitted.begin;
355              i != smm->uncommitted.end;
356              i = brb_next(&smm->uncommitted, i)) {
357
358                 struct block_op *op = smm->uncommitted.bops + i;
359
360                 if (op->block != b)
361                         continue;
362
363                 switch (op->type) {
364                 case BOP_INC:
365                         adjustment++;
366                         break;
367
368                 case BOP_DEC:
369                         adjustment--;
370                         break;
371                 }
372         }
373
374         if (adjustment > 1) {
375                 *result = 1;
376                 return 0;
377         }
378
379         r = sm_ll_lookup_bitmap(&smm->ll, b, &rc);
380         if (r)
381                 return r;
382
383         if (rc == 3)
384                 /*
385                  * We err on the side of caution, and always return true.
386                  */
387                 *result = 1;
388         else
389                 *result = rc + adjustment > 1;
390
391         return 0;
392 }
393
394 static int sm_metadata_set_count(struct dm_space_map *sm, dm_block_t b,
395                                  uint32_t count)
396 {
397         int r, r2;
398         enum allocation_event ev;
399         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
400
401         if (smm->recursion_count) {
402                 DMERR("cannot recurse set_count()");
403                 return -EINVAL;
404         }
405
406         in(smm);
407         r = sm_ll_insert(&smm->ll, b, count, &ev);
408         r2 = out(smm);
409
410         return combine_errors(r, r2);
411 }
412
413 static int sm_metadata_inc_block(struct dm_space_map *sm, dm_block_t b)
414 {
415         int r, r2 = 0;
416         enum allocation_event ev;
417         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
418
419         if (recursing(smm))
420                 r = add_bop(smm, BOP_INC, b);
421         else {
422                 in(smm);
423                 r = sm_ll_inc(&smm->ll, b, &ev);
424                 r2 = out(smm);
425         }
426
427         return combine_errors(r, r2);
428 }
429
430 static int sm_metadata_dec_block(struct dm_space_map *sm, dm_block_t b)
431 {
432         int r, r2 = 0;
433         enum allocation_event ev;
434         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
435
436         if (recursing(smm))
437                 r = add_bop(smm, BOP_DEC, b);
438         else {
439                 in(smm);
440                 r = sm_ll_dec(&smm->ll, b, &ev);
441                 r2 = out(smm);
442         }
443
444         return combine_errors(r, r2);
445 }
446
447 static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b)
448 {
449         int r, r2 = 0;
450         enum allocation_event ev;
451         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
452
453         r = sm_ll_find_free_block(&smm->old_ll, smm->begin, smm->old_ll.nr_blocks, b);
454         if (r)
455                 return r;
456
457         smm->begin = *b + 1;
458
459         if (recursing(smm))
460                 r = add_bop(smm, BOP_INC, *b);
461         else {
462                 in(smm);
463                 r = sm_ll_inc(&smm->ll, *b, &ev);
464                 r2 = out(smm);
465         }
466
467         if (!r)
468                 smm->allocated_this_transaction++;
469
470         return combine_errors(r, r2);
471 }
472
473 static int sm_metadata_new_block(struct dm_space_map *sm, dm_block_t *b)
474 {
475         dm_block_t count;
476         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
477
478         int r = sm_metadata_new_block_(sm, b);
479         if (r) {
480                 DMERR_LIMIT("unable to allocate new metadata block");
481                 return r;
482         }
483
484         r = sm_metadata_get_nr_free(sm, &count);
485         if (r) {
486                 DMERR_LIMIT("couldn't get free block count");
487                 return r;
488         }
489
490         check_threshold(&smm->threshold, count);
491
492         return r;
493 }
494
495 static int sm_metadata_commit(struct dm_space_map *sm)
496 {
497         int r;
498         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
499
500         r = sm_ll_commit(&smm->ll);
501         if (r)
502                 return r;
503
504         memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));
505         smm->begin = 0;
506         smm->allocated_this_transaction = 0;
507
508         return 0;
509 }
510
511 static int sm_metadata_register_threshold_callback(struct dm_space_map *sm,
512                                                    dm_block_t threshold,
513                                                    dm_sm_threshold_fn fn,
514                                                    void *context)
515 {
516         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
517
518         set_threshold(&smm->threshold, threshold, fn, context);
519
520         return 0;
521 }
522
523 static int sm_metadata_root_size(struct dm_space_map *sm, size_t *result)
524 {
525         *result = sizeof(struct disk_sm_root);
526
527         return 0;
528 }
529
530 static int sm_metadata_copy_root(struct dm_space_map *sm, void *where_le, size_t max)
531 {
532         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
533         struct disk_sm_root root_le;
534
535         root_le.nr_blocks = cpu_to_le64(smm->ll.nr_blocks);
536         root_le.nr_allocated = cpu_to_le64(smm->ll.nr_allocated);
537         root_le.bitmap_root = cpu_to_le64(smm->ll.bitmap_root);
538         root_le.ref_count_root = cpu_to_le64(smm->ll.ref_count_root);
539
540         if (max < sizeof(root_le))
541                 return -ENOSPC;
542
543         memcpy(where_le, &root_le, sizeof(root_le));
544
545         return 0;
546 }
547
548 static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks);
549
550 static struct dm_space_map ops = {
551         .destroy = sm_metadata_destroy,
552         .extend = sm_metadata_extend,
553         .get_nr_blocks = sm_metadata_get_nr_blocks,
554         .get_nr_free = sm_metadata_get_nr_free,
555         .get_count = sm_metadata_get_count,
556         .count_is_more_than_one = sm_metadata_count_is_more_than_one,
557         .set_count = sm_metadata_set_count,
558         .inc_block = sm_metadata_inc_block,
559         .dec_block = sm_metadata_dec_block,
560         .new_block = sm_metadata_new_block,
561         .commit = sm_metadata_commit,
562         .root_size = sm_metadata_root_size,
563         .copy_root = sm_metadata_copy_root,
564         .register_threshold_callback = sm_metadata_register_threshold_callback
565 };
566
567 /*----------------------------------------------------------------*/
568
569 /*
570  * When a new space map is created that manages its own space.  We use
571  * this tiny bootstrap allocator.
572  */
573 static void sm_bootstrap_destroy(struct dm_space_map *sm)
574 {
575 }
576
577 static int sm_bootstrap_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
578 {
579         DMERR("bootstrap doesn't support extend");
580
581         return -EINVAL;
582 }
583
584 static int sm_bootstrap_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
585 {
586         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
587
588         *count = smm->ll.nr_blocks;
589
590         return 0;
591 }
592
593 static int sm_bootstrap_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
594 {
595         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
596
597         *count = smm->ll.nr_blocks - smm->begin;
598
599         return 0;
600 }
601
602 static int sm_bootstrap_get_count(struct dm_space_map *sm, dm_block_t b,
603                                   uint32_t *result)
604 {
605         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
606
607         *result = (b < smm->begin) ? 1 : 0;
608
609         return 0;
610 }
611
612 static int sm_bootstrap_count_is_more_than_one(struct dm_space_map *sm,
613                                                dm_block_t b, int *result)
614 {
615         *result = 0;
616
617         return 0;
618 }
619
620 static int sm_bootstrap_set_count(struct dm_space_map *sm, dm_block_t b,
621                                   uint32_t count)
622 {
623         DMERR("bootstrap doesn't support set_count");
624
625         return -EINVAL;
626 }
627
628 static int sm_bootstrap_new_block(struct dm_space_map *sm, dm_block_t *b)
629 {
630         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
631
632         /*
633          * We know the entire device is unused.
634          */
635         if (smm->begin == smm->ll.nr_blocks)
636                 return -ENOSPC;
637
638         *b = smm->begin++;
639
640         return 0;
641 }
642
643 static int sm_bootstrap_inc_block(struct dm_space_map *sm, dm_block_t b)
644 {
645         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
646
647         return add_bop(smm, BOP_INC, b);
648 }
649
650 static int sm_bootstrap_dec_block(struct dm_space_map *sm, dm_block_t b)
651 {
652         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
653
654         return add_bop(smm, BOP_DEC, b);
655 }
656
657 static int sm_bootstrap_commit(struct dm_space_map *sm)
658 {
659         return 0;
660 }
661
662 static int sm_bootstrap_root_size(struct dm_space_map *sm, size_t *result)
663 {
664         DMERR("bootstrap doesn't support root_size");
665
666         return -EINVAL;
667 }
668
669 static int sm_bootstrap_copy_root(struct dm_space_map *sm, void *where,
670                                   size_t max)
671 {
672         DMERR("bootstrap doesn't support copy_root");
673
674         return -EINVAL;
675 }
676
677 static struct dm_space_map bootstrap_ops = {
678         .destroy = sm_bootstrap_destroy,
679         .extend = sm_bootstrap_extend,
680         .get_nr_blocks = sm_bootstrap_get_nr_blocks,
681         .get_nr_free = sm_bootstrap_get_nr_free,
682         .get_count = sm_bootstrap_get_count,
683         .count_is_more_than_one = sm_bootstrap_count_is_more_than_one,
684         .set_count = sm_bootstrap_set_count,
685         .inc_block = sm_bootstrap_inc_block,
686         .dec_block = sm_bootstrap_dec_block,
687         .new_block = sm_bootstrap_new_block,
688         .commit = sm_bootstrap_commit,
689         .root_size = sm_bootstrap_root_size,
690         .copy_root = sm_bootstrap_copy_root,
691         .register_threshold_callback = NULL
692 };
693
694 /*----------------------------------------------------------------*/
695
696 static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
697 {
698         int r, i;
699         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
700         dm_block_t old_len = smm->ll.nr_blocks;
701
702         /*
703          * Flick into a mode where all blocks get allocated in the new area.
704          */
705         smm->begin = old_len;
706         memcpy(sm, &bootstrap_ops, sizeof(*sm));
707
708         /*
709          * Extend.
710          */
711         r = sm_ll_extend(&smm->ll, extra_blocks);
712         if (r)
713                 goto out;
714
715         /*
716          * We repeatedly increment then commit until the commit doesn't
717          * allocate any new blocks.
718          */
719         do {
720                 for (i = old_len; !r && i < smm->begin; i++)
721                         r = add_bop(smm, BOP_INC, i);
722
723                 if (r)
724                         goto out;
725
726                 old_len = smm->begin;
727
728                 r = apply_bops(smm);
729                 if (r) {
730                         DMERR("%s: apply_bops failed", __func__);
731                         goto out;
732                 }
733
734                 r = sm_ll_commit(&smm->ll);
735                 if (r)
736                         goto out;
737
738         } while (old_len != smm->begin);
739
740 out:
741         /*
742          * Switch back to normal behaviour.
743          */
744         memcpy(sm, &ops, sizeof(*sm));
745         return r;
746 }
747
748 /*----------------------------------------------------------------*/
749
750 struct dm_space_map *dm_sm_metadata_init(void)
751 {
752         struct sm_metadata *smm;
753
754         smm = kmalloc(sizeof(*smm), GFP_KERNEL);
755         if (!smm)
756                 return ERR_PTR(-ENOMEM);
757
758         memcpy(&smm->sm, &ops, sizeof(smm->sm));
759
760         return &smm->sm;
761 }
762
763 int dm_sm_metadata_create(struct dm_space_map *sm,
764                           struct dm_transaction_manager *tm,
765                           dm_block_t nr_blocks,
766                           dm_block_t superblock)
767 {
768         int r;
769         dm_block_t i;
770         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
771
772         smm->begin = superblock + 1;
773         smm->recursion_count = 0;
774         smm->allocated_this_transaction = 0;
775         brb_init(&smm->uncommitted);
776         threshold_init(&smm->threshold);
777
778         memcpy(&smm->sm, &bootstrap_ops, sizeof(smm->sm));
779
780         r = sm_ll_new_metadata(&smm->ll, tm);
781         if (r)
782                 return r;
783
784         if (nr_blocks > DM_SM_METADATA_MAX_BLOCKS)
785                 nr_blocks = DM_SM_METADATA_MAX_BLOCKS;
786         r = sm_ll_extend(&smm->ll, nr_blocks);
787         if (r)
788                 return r;
789
790         memcpy(&smm->sm, &ops, sizeof(smm->sm));
791
792         /*
793          * Now we need to update the newly created data structures with the
794          * allocated blocks that they were built from.
795          */
796         for (i = superblock; !r && i < smm->begin; i++)
797                 r = add_bop(smm, BOP_INC, i);
798
799         if (r)
800                 return r;
801
802         r = apply_bops(smm);
803         if (r) {
804                 DMERR("%s: apply_bops failed", __func__);
805                 return r;
806         }
807
808         return sm_metadata_commit(sm);
809 }
810
811 int dm_sm_metadata_open(struct dm_space_map *sm,
812                         struct dm_transaction_manager *tm,
813                         void *root_le, size_t len)
814 {
815         int r;
816         struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
817
818         r = sm_ll_open_metadata(&smm->ll, tm, root_le, len);
819         if (r)
820                 return r;
821
822         smm->begin = 0;
823         smm->recursion_count = 0;
824         smm->allocated_this_transaction = 0;
825         brb_init(&smm->uncommitted);
826         threshold_init(&smm->threshold);
827
828         memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));
829         return 0;
830 }