Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / staging / lustre / lustre / fld / fld_cache.c
1 /*
2  * GPL HEADER START
3  *
4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License version 2 only,
8  * as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope that it will be useful, but
11  * WITHOUT ANY WARRANTY; without even the implied warranty of
12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13  * General Public License version 2 for more details (a copy is included
14  * in the LICENSE file that accompanied this code).
15  *
16  * You should have received a copy of the GNU General Public License
17  * version 2 along with this program; If not, see
18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
19  *
20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21  * CA 95054 USA or visit www.sun.com if you need additional information or
22  * have any questions.
23  *
24  * GPL HEADER END
25  */
26 /*
27  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28  * Use is subject to license terms.
29  *
30  * Copyright (c) 2012, 2013, Intel Corporation.
31  */
32 /*
33  * This file is part of Lustre, http://www.lustre.org/
34  * Lustre is a trademark of Sun Microsystems, Inc.
35  *
36  * lustre/fld/fld_cache.c
37  *
38  * FLD (Fids Location Database)
39  *
40  * Author: Pravin Shelar <pravin.shelar@sun.com>
41  * Author: Yury Umanets <umka@clusterfs.com>
42  */
43
44 #define DEBUG_SUBSYSTEM S_FLD
45
46 #include "../../include/linux/libcfs/libcfs.h"
47 #include <linux/module.h>
48 #include <asm/div64.h>
49
50 #include "../include/obd.h"
51 #include "../include/obd_class.h"
52 #include "../include/lustre_ver.h"
53 #include "../include/obd_support.h"
54 #include "../include/lprocfs_status.h"
55
56 #include "../include/dt_object.h"
57 #include "../include/lustre_req_layout.h"
58 #include "../include/lustre_fld.h"
59 #include "fld_internal.h"
60
61 /**
62  * create fld cache.
63  */
64 struct fld_cache *fld_cache_init(const char *name,
65                                  int cache_size, int cache_threshold)
66 {
67         struct fld_cache *cache;
68
69         LASSERT(name != NULL);
70         LASSERT(cache_threshold < cache_size);
71
72         OBD_ALLOC_PTR(cache);
73         if (cache == NULL)
74                 return ERR_PTR(-ENOMEM);
75
76         INIT_LIST_HEAD(&cache->fci_entries_head);
77         INIT_LIST_HEAD(&cache->fci_lru);
78
79         cache->fci_cache_count = 0;
80         rwlock_init(&cache->fci_lock);
81
82         strlcpy(cache->fci_name, name,
83                 sizeof(cache->fci_name));
84
85         cache->fci_cache_size = cache_size;
86         cache->fci_threshold = cache_threshold;
87
88         /* Init fld cache info. */
89         memset(&cache->fci_stat, 0, sizeof(cache->fci_stat));
90
91         CDEBUG(D_INFO, "%s: FLD cache - Size: %d, Threshold: %d\n",
92                cache->fci_name, cache_size, cache_threshold);
93
94         return cache;
95 }
96
97 /**
98  * destroy fld cache.
99  */
100 void fld_cache_fini(struct fld_cache *cache)
101 {
102         __u64 pct;
103
104         LASSERT(cache != NULL);
105         fld_cache_flush(cache);
106
107         if (cache->fci_stat.fst_count > 0) {
108                 pct = cache->fci_stat.fst_cache * 100;
109                 do_div(pct, cache->fci_stat.fst_count);
110         } else {
111                 pct = 0;
112         }
113
114         CDEBUG(D_INFO, "FLD cache statistics (%s):\n", cache->fci_name);
115         CDEBUG(D_INFO, "  Total reqs: %llu\n", cache->fci_stat.fst_count);
116         CDEBUG(D_INFO, "  Cache reqs: %llu\n", cache->fci_stat.fst_cache);
117         CDEBUG(D_INFO, "  Cache hits: %llu%%\n", pct);
118
119         OBD_FREE_PTR(cache);
120 }
121
122 /**
123  * delete given node from list.
124  */
125 void fld_cache_entry_delete(struct fld_cache *cache,
126                             struct fld_cache_entry *node)
127 {
128         list_del(&node->fce_list);
129         list_del(&node->fce_lru);
130         cache->fci_cache_count--;
131         OBD_FREE_PTR(node);
132 }
133
134 /**
135  * fix list by checking new entry with NEXT entry in order.
136  */
137 static void fld_fix_new_list(struct fld_cache *cache)
138 {
139         struct fld_cache_entry *f_curr;
140         struct fld_cache_entry *f_next;
141         struct lu_seq_range *c_range;
142         struct lu_seq_range *n_range;
143         struct list_head *head = &cache->fci_entries_head;
144
145 restart_fixup:
146
147         list_for_each_entry_safe(f_curr, f_next, head, fce_list) {
148                 c_range = &f_curr->fce_range;
149                 n_range = &f_next->fce_range;
150
151                 LASSERT(range_is_sane(c_range));
152                 if (&f_next->fce_list == head)
153                         break;
154
155                 if (c_range->lsr_flags != n_range->lsr_flags)
156                         continue;
157
158                 LASSERTF(c_range->lsr_start <= n_range->lsr_start,
159                          "cur lsr_start "DRANGE" next lsr_start "DRANGE"\n",
160                          PRANGE(c_range), PRANGE(n_range));
161
162                 /* check merge possibility with next range */
163                 if (c_range->lsr_end == n_range->lsr_start) {
164                         if (c_range->lsr_index != n_range->lsr_index)
165                                 continue;
166                         n_range->lsr_start = c_range->lsr_start;
167                         fld_cache_entry_delete(cache, f_curr);
168                         continue;
169                 }
170
171                 /* check if current range overlaps with next range. */
172                 if (n_range->lsr_start < c_range->lsr_end) {
173                         if (c_range->lsr_index == n_range->lsr_index) {
174                                 n_range->lsr_start = c_range->lsr_start;
175                                 n_range->lsr_end = max(c_range->lsr_end,
176                                                        n_range->lsr_end);
177                                 fld_cache_entry_delete(cache, f_curr);
178                         } else {
179                                 if (n_range->lsr_end <= c_range->lsr_end) {
180                                         *n_range = *c_range;
181                                         fld_cache_entry_delete(cache, f_curr);
182                                 } else
183                                         n_range->lsr_start = c_range->lsr_end;
184                         }
185
186                         /* we could have overlap over next
187                          * range too. better restart. */
188                         goto restart_fixup;
189                 }
190
191                 /* kill duplicates */
192                 if (c_range->lsr_start == n_range->lsr_start &&
193                     c_range->lsr_end == n_range->lsr_end)
194                         fld_cache_entry_delete(cache, f_curr);
195         }
196 }
197
198 /**
199  * add node to fld cache
200  */
201 static inline void fld_cache_entry_add(struct fld_cache *cache,
202                                        struct fld_cache_entry *f_new,
203                                        struct list_head *pos)
204 {
205         list_add(&f_new->fce_list, pos);
206         list_add(&f_new->fce_lru, &cache->fci_lru);
207
208         cache->fci_cache_count++;
209         fld_fix_new_list(cache);
210 }
211
212 /**
213  * Check if cache needs to be shrunk. If so - do it.
214  * Remove one entry in list and so on until cache is shrunk enough.
215  */
216 static int fld_cache_shrink(struct fld_cache *cache)
217 {
218         struct fld_cache_entry *flde;
219         struct list_head *curr;
220         int num = 0;
221
222         LASSERT(cache != NULL);
223
224         if (cache->fci_cache_count < cache->fci_cache_size)
225                 return 0;
226
227         curr = cache->fci_lru.prev;
228
229         while (cache->fci_cache_count + cache->fci_threshold >
230                cache->fci_cache_size && curr != &cache->fci_lru) {
231
232                 flde = list_entry(curr, struct fld_cache_entry, fce_lru);
233                 curr = curr->prev;
234                 fld_cache_entry_delete(cache, flde);
235                 num++;
236         }
237
238         CDEBUG(D_INFO, "%s: FLD cache - Shrunk by %d entries\n",
239                         cache->fci_name, num);
240
241         return 0;
242 }
243
244 /**
245  * kill all fld cache entries.
246  */
247 void fld_cache_flush(struct fld_cache *cache)
248 {
249         write_lock(&cache->fci_lock);
250         cache->fci_cache_size = 0;
251         fld_cache_shrink(cache);
252         write_unlock(&cache->fci_lock);
253 }
254
255 /**
256  * punch hole in existing range. divide this range and add new
257  * entry accordingly.
258  */
259
260 static void fld_cache_punch_hole(struct fld_cache *cache,
261                                  struct fld_cache_entry *f_curr,
262                                  struct fld_cache_entry *f_new)
263 {
264         const struct lu_seq_range *range = &f_new->fce_range;
265         const u64 new_start  = range->lsr_start;
266         const u64 new_end  = range->lsr_end;
267         struct fld_cache_entry *fldt;
268
269         OBD_ALLOC_GFP(fldt, sizeof(*fldt), GFP_ATOMIC);
270         if (!fldt) {
271                 OBD_FREE_PTR(f_new);
272                 /* overlap is not allowed, so dont mess up list. */
273                 return;
274         }
275         /*  break f_curr RANGE into three RANGES:
276          *      f_curr, f_new , fldt
277          */
278
279         /* f_new = *range */
280
281         /* fldt */
282         fldt->fce_range.lsr_start = new_end;
283         fldt->fce_range.lsr_end = f_curr->fce_range.lsr_end;
284         fldt->fce_range.lsr_index = f_curr->fce_range.lsr_index;
285
286         /* f_curr */
287         f_curr->fce_range.lsr_end = new_start;
288
289         /* add these two entries to list */
290         fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
291         fld_cache_entry_add(cache, fldt, &f_new->fce_list);
292
293         /* no need to fixup */
294 }
295
296 /**
297  * handle range overlap in fld cache.
298  */
299 static void fld_cache_overlap_handle(struct fld_cache *cache,
300                                 struct fld_cache_entry *f_curr,
301                                 struct fld_cache_entry *f_new)
302 {
303         const struct lu_seq_range *range = &f_new->fce_range;
304         const u64 new_start  = range->lsr_start;
305         const u64 new_end  = range->lsr_end;
306         const u32 mdt = range->lsr_index;
307
308         /* this is overlap case, these case are checking overlapping with
309          * prev range only. fixup will handle overlapping with next range. */
310
311         if (f_curr->fce_range.lsr_index == mdt) {
312                 f_curr->fce_range.lsr_start = min(f_curr->fce_range.lsr_start,
313                                                   new_start);
314
315                 f_curr->fce_range.lsr_end = max(f_curr->fce_range.lsr_end,
316                                                 new_end);
317
318                 OBD_FREE_PTR(f_new);
319                 fld_fix_new_list(cache);
320
321         } else if (new_start <= f_curr->fce_range.lsr_start &&
322                         f_curr->fce_range.lsr_end <= new_end) {
323                 /* case 1: new range completely overshadowed existing range.
324                  *       e.g. whole range migrated. update fld cache entry */
325
326                 f_curr->fce_range = *range;
327                 OBD_FREE_PTR(f_new);
328                 fld_fix_new_list(cache);
329
330         } else if (f_curr->fce_range.lsr_start < new_start &&
331                         new_end < f_curr->fce_range.lsr_end) {
332                 /* case 2: new range fit within existing range. */
333
334                 fld_cache_punch_hole(cache, f_curr, f_new);
335
336         } else  if (new_end <= f_curr->fce_range.lsr_end) {
337                 /* case 3: overlap:
338                  *       [new_start [c_start  new_end)  c_end)
339                  */
340
341                 LASSERT(new_start <= f_curr->fce_range.lsr_start);
342
343                 f_curr->fce_range.lsr_start = new_end;
344                 fld_cache_entry_add(cache, f_new, f_curr->fce_list.prev);
345
346         } else if (f_curr->fce_range.lsr_start <= new_start) {
347                 /* case 4: overlap:
348                  *       [c_start [new_start c_end) new_end)
349                  */
350
351                 LASSERT(f_curr->fce_range.lsr_end <= new_end);
352
353                 f_curr->fce_range.lsr_end = new_start;
354                 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
355         } else
356                 CERROR("NEW range ="DRANGE" curr = "DRANGE"\n",
357                        PRANGE(range), PRANGE(&f_curr->fce_range));
358 }
359
360 struct fld_cache_entry
361 *fld_cache_entry_create(const struct lu_seq_range *range)
362 {
363         struct fld_cache_entry *f_new;
364
365         LASSERT(range_is_sane(range));
366
367         OBD_ALLOC_PTR(f_new);
368         if (!f_new)
369                 return ERR_PTR(-ENOMEM);
370
371         f_new->fce_range = *range;
372         return f_new;
373 }
374
375 /**
376  * Insert FLD entry in FLD cache.
377  *
378  * This function handles all cases of merging and breaking up of
379  * ranges.
380  */
381 int fld_cache_insert_nolock(struct fld_cache *cache,
382                             struct fld_cache_entry *f_new)
383 {
384         struct fld_cache_entry *f_curr;
385         struct fld_cache_entry *n;
386         struct list_head *head;
387         struct list_head *prev = NULL;
388         const u64 new_start  = f_new->fce_range.lsr_start;
389         const u64 new_end  = f_new->fce_range.lsr_end;
390         __u32 new_flags  = f_new->fce_range.lsr_flags;
391
392         /*
393          * Duplicate entries are eliminated in insert op.
394          * So we don't need to search new entry before starting
395          * insertion loop.
396          */
397
398         if (!cache->fci_no_shrink)
399                 fld_cache_shrink(cache);
400
401         head = &cache->fci_entries_head;
402
403         list_for_each_entry_safe(f_curr, n, head, fce_list) {
404                 /* add list if next is end of list */
405                 if (new_end < f_curr->fce_range.lsr_start ||
406                    (new_end == f_curr->fce_range.lsr_start &&
407                     new_flags != f_curr->fce_range.lsr_flags))
408                         break;
409
410                 prev = &f_curr->fce_list;
411                 /* check if this range is to left of new range. */
412                 if (new_start < f_curr->fce_range.lsr_end &&
413                     new_flags == f_curr->fce_range.lsr_flags) {
414                         fld_cache_overlap_handle(cache, f_curr, f_new);
415                         goto out;
416                 }
417         }
418
419         if (prev == NULL)
420                 prev = head;
421
422         CDEBUG(D_INFO, "insert range "DRANGE"\n", PRANGE(&f_new->fce_range));
423         /* Add new entry to cache and lru list. */
424         fld_cache_entry_add(cache, f_new, prev);
425 out:
426         return 0;
427 }
428
429 int fld_cache_insert(struct fld_cache *cache,
430                      const struct lu_seq_range *range)
431 {
432         struct fld_cache_entry  *flde;
433         int rc;
434
435         flde = fld_cache_entry_create(range);
436         if (IS_ERR(flde))
437                 return PTR_ERR(flde);
438
439         write_lock(&cache->fci_lock);
440         rc = fld_cache_insert_nolock(cache, flde);
441         write_unlock(&cache->fci_lock);
442         if (rc)
443                 OBD_FREE_PTR(flde);
444
445         return rc;
446 }
447
448 void fld_cache_delete_nolock(struct fld_cache *cache,
449                       const struct lu_seq_range *range)
450 {
451         struct fld_cache_entry *flde;
452         struct fld_cache_entry *tmp;
453         struct list_head *head;
454
455         head = &cache->fci_entries_head;
456         list_for_each_entry_safe(flde, tmp, head, fce_list) {
457                 /* add list if next is end of list */
458                 if (range->lsr_start == flde->fce_range.lsr_start ||
459                    (range->lsr_end == flde->fce_range.lsr_end &&
460                     range->lsr_flags == flde->fce_range.lsr_flags)) {
461                         fld_cache_entry_delete(cache, flde);
462                         break;
463                 }
464         }
465 }
466
467 /**
468  * Delete FLD entry in FLD cache.
469  *
470  */
471 void fld_cache_delete(struct fld_cache *cache,
472                       const struct lu_seq_range *range)
473 {
474         write_lock(&cache->fci_lock);
475         fld_cache_delete_nolock(cache, range);
476         write_unlock(&cache->fci_lock);
477 }
478
479 struct fld_cache_entry
480 *fld_cache_entry_lookup_nolock(struct fld_cache *cache,
481                               struct lu_seq_range *range)
482 {
483         struct fld_cache_entry *flde;
484         struct fld_cache_entry *got = NULL;
485         struct list_head *head;
486
487         head = &cache->fci_entries_head;
488         list_for_each_entry(flde, head, fce_list) {
489                 if (range->lsr_start == flde->fce_range.lsr_start ||
490                    (range->lsr_end == flde->fce_range.lsr_end &&
491                     range->lsr_flags == flde->fce_range.lsr_flags)) {
492                         got = flde;
493                         break;
494                 }
495         }
496
497         return got;
498 }
499
500 /**
501  * lookup \a seq sequence for range in fld cache.
502  */
503 struct fld_cache_entry
504 *fld_cache_entry_lookup(struct fld_cache *cache, struct lu_seq_range *range)
505 {
506         struct fld_cache_entry *got = NULL;
507
508         read_lock(&cache->fci_lock);
509         got = fld_cache_entry_lookup_nolock(cache, range);
510         read_unlock(&cache->fci_lock);
511         return got;
512 }
513
514 /**
515  * lookup \a seq sequence for range in fld cache.
516  */
517 int fld_cache_lookup(struct fld_cache *cache,
518                      const u64 seq, struct lu_seq_range *range)
519 {
520         struct fld_cache_entry *flde;
521         struct fld_cache_entry *prev = NULL;
522         struct list_head *head;
523
524         read_lock(&cache->fci_lock);
525         head = &cache->fci_entries_head;
526
527         cache->fci_stat.fst_count++;
528         list_for_each_entry(flde, head, fce_list) {
529                 if (flde->fce_range.lsr_start > seq) {
530                         if (prev != NULL)
531                                 *range = prev->fce_range;
532                         break;
533                 }
534
535                 prev = flde;
536                 if (range_within(&flde->fce_range, seq)) {
537                         *range = flde->fce_range;
538
539                         cache->fci_stat.fst_cache++;
540                         read_unlock(&cache->fci_lock);
541                         return 0;
542                 }
543         }
544         read_unlock(&cache->fci_lock);
545         return -ENOENT;
546 }