4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2014, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/fld/fld_cache.c
34 * FLD (Fids Location Database)
36 * Author: Pravin Shelar <pravin.shelar@sun.com>
37 * Author: Yury Umanets <umka@clusterfs.com>
40 #define DEBUG_SUBSYSTEM S_FLD
42 #include <libcfs/libcfs.h>
43 #include <linux/module.h>
44 #include <linux/math64.h>
45 #include <obd_support.h>
46 #include <lustre_fld.h>
47 #include "fld_internal.h"
52 struct fld_cache *fld_cache_init(const char *name,
53 int cache_size, int cache_threshold)
55 struct fld_cache *cache;
58 LASSERT(name != NULL);
59 LASSERT(cache_threshold < cache_size);
63 RETURN(ERR_PTR(-ENOMEM));
65 INIT_LIST_HEAD(&cache->fci_entries_head);
66 INIT_LIST_HEAD(&cache->fci_lru);
68 cache->fci_cache_count = 0;
69 rwlock_init(&cache->fci_lock);
71 strlcpy(cache->fci_name, name,
72 sizeof(cache->fci_name));
74 cache->fci_cache_size = cache_size;
75 cache->fci_threshold = cache_threshold;
77 /* Init fld cache info. */
78 memset(&cache->fci_stat, 0, sizeof(cache->fci_stat));
80 CDEBUG(D_INFO, "%s: FLD cache - Size: %d, Threshold: %d\n",
81 cache->fci_name, cache_size, cache_threshold);
89 void fld_cache_fini(struct fld_cache *cache)
91 LASSERT(cache != NULL);
92 fld_cache_flush(cache);
94 CDEBUG(D_INFO, "FLD cache statistics (%s):\n", cache->fci_name);
95 CDEBUG(D_INFO, " Cache reqs: %llu\n", cache->fci_stat.fst_cache);
96 CDEBUG(D_INFO, " Total reqs: %llu\n", cache->fci_stat.fst_count);
102 * delete given node from list.
104 void fld_cache_entry_delete(struct fld_cache *cache,
105 struct fld_cache_entry *node)
107 list_del(&node->fce_list);
108 list_del(&node->fce_lru);
109 cache->fci_cache_count--;
114 * fix list by checking new entry with NEXT entry in order.
116 static void fld_fix_new_list(struct fld_cache *cache)
118 struct fld_cache_entry *f_curr;
119 struct fld_cache_entry *f_next;
120 struct lu_seq_range *c_range;
121 struct lu_seq_range *n_range;
122 struct list_head *head = &cache->fci_entries_head;
127 list_for_each_entry_safe(f_curr, f_next, head, fce_list) {
128 c_range = &f_curr->fce_range;
129 n_range = &f_next->fce_range;
131 LASSERT(lu_seq_range_is_sane(c_range));
132 if (&f_next->fce_list == head)
135 if (c_range->lsr_flags != n_range->lsr_flags)
138 LASSERTF(c_range->lsr_start <= n_range->lsr_start,
139 "cur lsr_start "DRANGE" next lsr_start "DRANGE"\n",
140 PRANGE(c_range), PRANGE(n_range));
142 /* check merge possibility with next range */
143 if (c_range->lsr_end == n_range->lsr_start) {
144 if (c_range->lsr_index != n_range->lsr_index)
146 n_range->lsr_start = c_range->lsr_start;
147 fld_cache_entry_delete(cache, f_curr);
151 /* check if current range overlaps with next range. */
152 if (n_range->lsr_start < c_range->lsr_end) {
153 if (c_range->lsr_index == n_range->lsr_index) {
154 n_range->lsr_start = c_range->lsr_start;
155 n_range->lsr_end = max(c_range->lsr_end,
157 fld_cache_entry_delete(cache, f_curr);
159 if (n_range->lsr_end <= c_range->lsr_end) {
161 fld_cache_entry_delete(cache, f_curr);
163 n_range->lsr_start = c_range->lsr_end;
166 /* we could have overlap over next
167 * range too. better restart. */
171 /* kill duplicates */
172 if (c_range->lsr_start == n_range->lsr_start &&
173 c_range->lsr_end == n_range->lsr_end)
174 fld_cache_entry_delete(cache, f_curr);
181 * add node to fld cache
183 static inline void fld_cache_entry_add(struct fld_cache *cache,
184 struct fld_cache_entry *f_new,
185 struct list_head *pos)
187 list_add(&f_new->fce_list, pos);
188 list_add(&f_new->fce_lru, &cache->fci_lru);
190 cache->fci_cache_count++;
191 fld_fix_new_list(cache);
195 * Check if cache needs to be shrunk. If so - do it.
196 * Remove one entry in list and so on until cache is shrunk enough.
198 static int fld_cache_shrink(struct fld_cache *cache)
200 struct fld_cache_entry *flde;
201 struct list_head *curr;
205 LASSERT(cache != NULL);
207 if (cache->fci_cache_count < cache->fci_cache_size)
210 curr = cache->fci_lru.prev;
212 while (cache->fci_cache_count + cache->fci_threshold >
213 cache->fci_cache_size && curr != &cache->fci_lru) {
215 flde = list_entry(curr, struct fld_cache_entry, fce_lru);
217 fld_cache_entry_delete(cache, flde);
221 CDEBUG(D_INFO, "%s: FLD cache - Shrunk by "
222 "%d entries\n", cache->fci_name, num);
228 * kill all fld cache entries.
230 void fld_cache_flush(struct fld_cache *cache)
234 write_lock(&cache->fci_lock);
235 cache->fci_cache_size = 0;
236 fld_cache_shrink(cache);
237 write_unlock(&cache->fci_lock);
243 * punch hole in existing range. divide this range and add new
247 static void fld_cache_punch_hole(struct fld_cache *cache,
248 struct fld_cache_entry *f_curr,
249 struct fld_cache_entry *f_new)
251 const struct lu_seq_range *range = &f_new->fce_range;
252 const u64 new_start = range->lsr_start;
253 const u64 new_end = range->lsr_end;
254 struct fld_cache_entry *fldt;
257 OBD_ALLOC_GFP(fldt, sizeof *fldt, GFP_ATOMIC);
261 /* overlap is not allowed, so dont mess up list. */
264 /* break f_curr RANGE into three RANGES:
265 * f_curr, f_new , fldt
271 fldt->fce_range.lsr_start = new_end;
272 fldt->fce_range.lsr_end = f_curr->fce_range.lsr_end;
273 fldt->fce_range.lsr_index = f_curr->fce_range.lsr_index;
276 f_curr->fce_range.lsr_end = new_start;
278 /* add these two entries to list */
279 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
280 fld_cache_entry_add(cache, fldt, &f_new->fce_list);
282 /* no need to fixup */
287 * handle range overlap in fld cache.
289 static void fld_cache_overlap_handle(struct fld_cache *cache,
290 struct fld_cache_entry *f_curr,
291 struct fld_cache_entry *f_new)
293 const struct lu_seq_range *range = &f_new->fce_range;
294 const u64 new_start = range->lsr_start;
295 const u64 new_end = range->lsr_end;
296 const u32 mdt = range->lsr_index;
298 /* this is overlap case, these case are checking overlapping with
299 * prev range only. fixup will handle overlaping with next range. */
301 if (f_curr->fce_range.lsr_index == mdt) {
302 f_curr->fce_range.lsr_start = min(f_curr->fce_range.lsr_start,
305 f_curr->fce_range.lsr_end = max(f_curr->fce_range.lsr_end,
309 fld_fix_new_list(cache);
311 } else if (new_start <= f_curr->fce_range.lsr_start &&
312 f_curr->fce_range.lsr_end <= new_end) {
313 /* case 1: new range completely overshadowed existing range.
314 * e.g. whole range migrated. update fld cache entry */
316 f_curr->fce_range = *range;
318 fld_fix_new_list(cache);
320 } else if (f_curr->fce_range.lsr_start < new_start &&
321 new_end < f_curr->fce_range.lsr_end) {
322 /* case 2: new range fit within existing range. */
324 fld_cache_punch_hole(cache, f_curr, f_new);
326 } else if (new_end <= f_curr->fce_range.lsr_end) {
328 * [new_start [c_start new_end) c_end)
331 LASSERT(new_start <= f_curr->fce_range.lsr_start);
333 f_curr->fce_range.lsr_start = new_end;
334 fld_cache_entry_add(cache, f_new, f_curr->fce_list.prev);
336 } else if (f_curr->fce_range.lsr_start <= new_start) {
338 * [c_start [new_start c_end) new_end)
341 LASSERT(f_curr->fce_range.lsr_end <= new_end);
343 f_curr->fce_range.lsr_end = new_start;
344 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
346 CERROR("NEW range ="DRANGE" curr = "DRANGE"\n",
347 PRANGE(range),PRANGE(&f_curr->fce_range));
350 struct fld_cache_entry
351 *fld_cache_entry_create(const struct lu_seq_range *range)
353 struct fld_cache_entry *f_new;
355 LASSERT(lu_seq_range_is_sane(range));
357 OBD_ALLOC_PTR(f_new);
359 RETURN(ERR_PTR(-ENOMEM));
361 f_new->fce_range = *range;
366 * Insert FLD entry in FLD cache.
368 * This function handles all cases of merging and breaking up of
371 int fld_cache_insert_nolock(struct fld_cache *cache,
372 struct fld_cache_entry *f_new)
374 struct fld_cache_entry *f_curr;
375 struct fld_cache_entry *n;
376 struct list_head *head;
377 struct list_head *prev = NULL;
378 const u64 new_start = f_new->fce_range.lsr_start;
379 const u64 new_end = f_new->fce_range.lsr_end;
380 __u32 new_flags = f_new->fce_range.lsr_flags;
384 * Duplicate entries are eliminated in insert op.
385 * So we don't need to search new entry before starting
389 fld_cache_shrink(cache);
391 head = &cache->fci_entries_head;
393 list_for_each_entry_safe(f_curr, n, head, fce_list) {
394 /* add list if next is end of list */
395 if (new_end < f_curr->fce_range.lsr_start ||
396 (new_end == f_curr->fce_range.lsr_start &&
397 new_flags != f_curr->fce_range.lsr_flags))
400 prev = &f_curr->fce_list;
401 /* check if this range is to left of new range. */
402 if (new_start < f_curr->fce_range.lsr_end &&
403 new_flags == f_curr->fce_range.lsr_flags) {
404 fld_cache_overlap_handle(cache, f_curr, f_new);
412 CDEBUG(D_INFO, "insert range "DRANGE"\n", PRANGE(&f_new->fce_range));
413 /* Add new entry to cache and lru list. */
414 fld_cache_entry_add(cache, f_new, prev);
419 int fld_cache_insert(struct fld_cache *cache,
420 const struct lu_seq_range *range)
422 struct fld_cache_entry *flde;
425 flde = fld_cache_entry_create(range);
427 RETURN(PTR_ERR(flde));
429 write_lock(&cache->fci_lock);
430 rc = fld_cache_insert_nolock(cache, flde);
431 write_unlock(&cache->fci_lock);
438 void fld_cache_delete_nolock(struct fld_cache *cache,
439 const struct lu_seq_range *range)
441 struct fld_cache_entry *flde;
442 struct fld_cache_entry *tmp;
443 struct list_head *head;
445 head = &cache->fci_entries_head;
446 list_for_each_entry_safe(flde, tmp, head, fce_list) {
447 /* add list if next is end of list */
448 if (range->lsr_start == flde->fce_range.lsr_start ||
449 (range->lsr_end == flde->fce_range.lsr_end &&
450 range->lsr_flags == flde->fce_range.lsr_flags)) {
451 fld_cache_entry_delete(cache, flde);
458 * Delete FLD entry in FLD cache.
461 void fld_cache_delete(struct fld_cache *cache,
462 const struct lu_seq_range *range)
464 write_lock(&cache->fci_lock);
465 fld_cache_delete_nolock(cache, range);
466 write_unlock(&cache->fci_lock);
469 struct fld_cache_entry *
470 fld_cache_entry_lookup_nolock(struct fld_cache *cache,
471 const struct lu_seq_range *range)
473 struct fld_cache_entry *flde;
474 struct fld_cache_entry *got = NULL;
475 struct list_head *head;
477 head = &cache->fci_entries_head;
478 list_for_each_entry(flde, head, fce_list) {
479 if (range->lsr_start == flde->fce_range.lsr_start ||
480 (range->lsr_end == flde->fce_range.lsr_end &&
481 range->lsr_flags == flde->fce_range.lsr_flags)) {
491 * lookup \a seq sequence for range in fld cache.
493 struct fld_cache_entry *
494 fld_cache_entry_lookup(struct fld_cache *cache,
495 const struct lu_seq_range *range)
497 struct fld_cache_entry *got = NULL;
500 read_lock(&cache->fci_lock);
501 got = fld_cache_entry_lookup_nolock(cache, range);
502 read_unlock(&cache->fci_lock);
508 * lookup \a seq sequence for range in fld cache.
510 int fld_cache_lookup(struct fld_cache *cache,
511 const u64 seq, struct lu_seq_range *range)
513 struct fld_cache_entry *flde;
514 struct fld_cache_entry *prev = NULL;
515 struct list_head *head;
518 read_lock(&cache->fci_lock);
519 head = &cache->fci_entries_head;
521 cache->fci_stat.fst_count++;
522 list_for_each_entry(flde, head, fce_list) {
523 if (flde->fce_range.lsr_start > seq) {
525 *range = prev->fce_range;
530 if (lu_seq_range_within(&flde->fce_range, seq)) {
531 *range = flde->fce_range;
533 cache->fci_stat.fst_cache++;
534 read_unlock(&cache->fci_lock);
538 read_unlock(&cache->fci_lock);