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, int cache_size,
55 struct fld_cache *cache;
59 LASSERT(name != NULL);
60 LASSERT(cache_threshold < cache_size);
64 RETURN(ERR_PTR(-ENOMEM));
66 INIT_LIST_HEAD(&cache->fci_entries_head);
67 INIT_LIST_HEAD(&cache->fci_lru);
69 cache->fci_cache_count = 0;
70 rwlock_init(&cache->fci_lock);
72 strlcpy(cache->fci_name, name, 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 static 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;
128 list_for_each_entry_safe(f_curr, f_next, head, fce_list) {
129 c_range = &f_curr->fce_range;
130 n_range = &f_next->fce_range;
132 LASSERT(lu_seq_range_is_sane(c_range));
133 if (&f_next->fce_list == head)
136 if (c_range->lsr_flags != n_range->lsr_flags)
139 LASSERTF(c_range->lsr_start <= n_range->lsr_start,
140 "cur lsr_start "DRANGE" next lsr_start "DRANGE"\n",
141 PRANGE(c_range), PRANGE(n_range));
143 /* check merge possibility with next range */
144 if (c_range->lsr_end == n_range->lsr_start) {
145 if (c_range->lsr_index != n_range->lsr_index)
147 n_range->lsr_start = c_range->lsr_start;
148 fld_cache_entry_delete(cache, f_curr);
152 /* check if current range overlaps with next range. */
153 if (n_range->lsr_start < c_range->lsr_end) {
154 if (c_range->lsr_index == n_range->lsr_index) {
155 n_range->lsr_start = c_range->lsr_start;
156 n_range->lsr_end = max(c_range->lsr_end,
158 fld_cache_entry_delete(cache, f_curr);
160 if (n_range->lsr_end <= c_range->lsr_end) {
162 fld_cache_entry_delete(cache, f_curr);
164 n_range->lsr_start = c_range->lsr_end;
167 /* we could have overlap over next
168 * range too. better restart.
173 /* kill duplicates */
174 if (c_range->lsr_start == n_range->lsr_start &&
175 c_range->lsr_end == n_range->lsr_end)
176 fld_cache_entry_delete(cache, f_curr);
183 * add node to fld cache
185 static inline void fld_cache_entry_add(struct fld_cache *cache,
186 struct fld_cache_entry *f_new,
187 struct list_head *pos)
189 list_add(&f_new->fce_list, pos);
190 list_add(&f_new->fce_lru, &cache->fci_lru);
192 cache->fci_cache_count++;
193 fld_fix_new_list(cache);
197 * Check if cache needs to be shrunk. If so - do it.
198 * Remove one entry in list and so on until cache is shrunk enough.
200 static int fld_cache_shrink(struct fld_cache *cache)
206 LASSERT(cache != NULL);
208 if (cache->fci_cache_count < cache->fci_cache_size)
211 while (cache->fci_cache_count + cache->fci_threshold >
212 cache->fci_cache_size &&
213 !list_empty(&cache->fci_lru)) {
214 struct fld_cache_entry *flde =
215 list_last_entry(&cache->fci_lru, struct fld_cache_entry,
218 fld_cache_entry_delete(cache, flde);
222 CDEBUG(D_INFO, "%s: FLD cache - Shrunk by %d entries\n",
223 cache->fci_name, num);
229 * kill all fld cache entries.
231 void fld_cache_flush(struct fld_cache *cache)
235 write_lock(&cache->fci_lock);
236 cache->fci_cache_size = 0;
237 fld_cache_shrink(cache);
238 write_unlock(&cache->fci_lock);
244 * punch hole in existing range. divide this range and add new
248 static void fld_cache_punch_hole(struct fld_cache *cache,
249 struct fld_cache_entry *f_curr,
250 struct fld_cache_entry *f_new)
252 const struct lu_seq_range *range = &f_new->fce_range;
253 const u64 new_start = range->lsr_start;
254 const u64 new_end = range->lsr_end;
255 struct fld_cache_entry *fldt;
258 OBD_ALLOC_GFP(fldt, sizeof(*fldt), GFP_ATOMIC);
262 /* overlap is not allowed, so dont mess up list. */
265 /* break f_curr RANGE into three RANGES:
266 * f_curr, f_new , fldt
270 fldt->fce_range.lsr_start = new_end;
271 fldt->fce_range.lsr_end = f_curr->fce_range.lsr_end;
272 fldt->fce_range.lsr_index = f_curr->fce_range.lsr_index;
275 f_curr->fce_range.lsr_end = new_start;
277 /* add these two entries to list */
278 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
279 fld_cache_entry_add(cache, fldt, &f_new->fce_list);
281 /* no need to fixup */
286 * handle range overlap in fld cache.
288 static void fld_cache_overlap_handle(struct fld_cache *cache,
289 struct fld_cache_entry *f_curr,
290 struct fld_cache_entry *f_new)
292 const struct lu_seq_range *range = &f_new->fce_range;
293 const u64 new_start = range->lsr_start;
294 const u64 new_end = range->lsr_end;
295 const u32 mdt = range->lsr_index;
297 /* this is overlap case, these case are checking overlapping with
298 * 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
317 f_curr->fce_range = *range;
319 fld_fix_new_list(cache);
321 } else if (f_curr->fce_range.lsr_start < new_start &&
322 new_end < f_curr->fce_range.lsr_end) {
323 /* case 2: new range fit within existing range. */
325 fld_cache_punch_hole(cache, f_curr, f_new);
327 } else if (new_end <= f_curr->fce_range.lsr_end) {
329 * [new_start [c_start new_end) c_end)
332 LASSERT(new_start <= f_curr->fce_range.lsr_start);
334 f_curr->fce_range.lsr_start = new_end;
335 fld_cache_entry_add(cache, f_new, f_curr->fce_list.prev);
337 } else if (f_curr->fce_range.lsr_start <= new_start) {
339 * [c_start [new_start c_end) new_end)
342 LASSERT(f_curr->fce_range.lsr_end <= new_end);
344 f_curr->fce_range.lsr_end = new_start;
345 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
347 CERROR("NEW range ="DRANGE" curr = "DRANGE"\n",
348 PRANGE(range), PRANGE(&f_curr->fce_range));
351 struct fld_cache_entry
352 *fld_cache_entry_create(const struct lu_seq_range *range)
354 struct fld_cache_entry *f_new;
356 LASSERT(lu_seq_range_is_sane(range));
358 OBD_ALLOC_PTR(f_new);
360 RETURN(ERR_PTR(-ENOMEM));
362 f_new->fce_range = *range;
367 * Insert FLD entry in FLD cache.
369 * This function handles all cases of merging and breaking up of
372 int fld_cache_insert_nolock(struct fld_cache *cache,
373 struct fld_cache_entry *f_new)
375 struct fld_cache_entry *f_curr;
376 struct fld_cache_entry *n;
377 struct list_head *head;
378 struct list_head *prev = NULL;
379 const u64 new_start = f_new->fce_range.lsr_start;
380 const u64 new_end = f_new->fce_range.lsr_end;
381 __u32 new_flags = f_new->fce_range.lsr_flags;
386 * Duplicate entries are eliminated in insert op.
387 * So we don't need to search new entry before starting
391 fld_cache_shrink(cache);
393 head = &cache->fci_entries_head;
395 list_for_each_entry_safe(f_curr, n, head, fce_list) {
396 /* add list if next is end of list */
397 if (new_end < f_curr->fce_range.lsr_start ||
398 (new_end == f_curr->fce_range.lsr_start &&
399 new_flags != f_curr->fce_range.lsr_flags))
402 prev = &f_curr->fce_list;
403 /* check if this range is to left of new range. */
404 if (new_start < f_curr->fce_range.lsr_end &&
405 new_flags == f_curr->fce_range.lsr_flags) {
406 fld_cache_overlap_handle(cache, f_curr, f_new);
414 CDEBUG(D_INFO, "insert range "DRANGE"\n", PRANGE(&f_new->fce_range));
415 /* Add new entry to cache and lru list. */
416 fld_cache_entry_add(cache, f_new, prev);
421 int fld_cache_insert(struct fld_cache *cache,
422 const struct lu_seq_range *range)
424 struct fld_cache_entry *flde;
427 flde = fld_cache_entry_create(range);
429 RETURN(PTR_ERR(flde));
431 write_lock(&cache->fci_lock);
432 rc = fld_cache_insert_nolock(cache, flde);
433 write_unlock(&cache->fci_lock);
440 void fld_cache_delete_nolock(struct fld_cache *cache,
441 const struct lu_seq_range *range)
443 struct fld_cache_entry *flde;
444 struct fld_cache_entry *tmp;
445 struct list_head *head;
447 head = &cache->fci_entries_head;
448 list_for_each_entry_safe(flde, tmp, head, fce_list) {
449 /* add list if next is end of list */
450 if (range->lsr_start == flde->fce_range.lsr_start ||
451 (range->lsr_end == flde->fce_range.lsr_end &&
452 range->lsr_flags == flde->fce_range.lsr_flags)) {
453 fld_cache_entry_delete(cache, flde);
460 * lookup \a seq sequence for range in fld cache.
462 int fld_cache_lookup(struct fld_cache *cache,
463 const u64 seq, struct lu_seq_range *range)
465 struct fld_cache_entry *flde;
466 struct fld_cache_entry *prev = NULL;
467 struct list_head *head;
471 read_lock(&cache->fci_lock);
472 head = &cache->fci_entries_head;
474 cache->fci_stat.fst_count++;
475 list_for_each_entry(flde, head, fce_list) {
476 if (flde->fce_range.lsr_start > seq) {
478 *range = prev->fce_range;
483 if (lu_seq_range_within(&flde->fce_range, seq)) {
484 *range = flde->fce_range;
486 cache->fci_stat.fst_cache++;
487 read_unlock(&cache->fci_lock);
491 read_unlock(&cache->fci_lock);