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.sun.com/software/products/lustre/docs/GPLv2.pdf
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
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2012, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/fld/fld_cache.c
38 * FLD (Fids Location Database)
40 * Author: Pravin Shelar <pravin.shelar@sun.com>
41 * Author: Yury Umanets <umka@clusterfs.com>
44 #define DEBUG_SUBSYSTEM S_FLD
46 # include <linux/libcfs/libcfs.h>
47 # include <linux/module.h>
48 # include <asm/div64.h>
51 #include <obd_class.h>
52 #include <lustre_ver.h>
53 #include <obd_support.h>
54 #include <lprocfs_status.h>
56 #include <dt_object.h>
57 #include <md_object.h>
58 #include <lustre_req_layout.h>
59 #include <lustre_fld.h>
60 #include "fld_internal.h"
65 struct fld_cache *fld_cache_init(const char *name,
66 int cache_size, int cache_threshold)
68 struct fld_cache *cache;
70 LASSERT(name != NULL);
71 LASSERT(cache_threshold < cache_size);
75 return ERR_PTR(-ENOMEM);
77 INIT_LIST_HEAD(&cache->fci_entries_head);
78 INIT_LIST_HEAD(&cache->fci_lru);
80 cache->fci_cache_count = 0;
81 rwlock_init(&cache->fci_lock);
83 strlcpy(cache->fci_name, name,
84 sizeof(cache->fci_name));
86 cache->fci_cache_size = cache_size;
87 cache->fci_threshold = cache_threshold;
89 /* Init fld cache info. */
90 memset(&cache->fci_stat, 0, sizeof(cache->fci_stat));
92 CDEBUG(D_INFO, "%s: FLD cache - Size: %d, Threshold: %d\n",
93 cache->fci_name, cache_size, cache_threshold);
101 void fld_cache_fini(struct fld_cache *cache)
105 LASSERT(cache != NULL);
106 fld_cache_flush(cache);
108 if (cache->fci_stat.fst_count > 0) {
109 pct = cache->fci_stat.fst_cache * 100;
110 do_div(pct, cache->fci_stat.fst_count);
115 CDEBUG(D_INFO, "FLD cache statistics (%s):\n", cache->fci_name);
116 CDEBUG(D_INFO, " Total reqs: "LPU64"\n", cache->fci_stat.fst_count);
117 CDEBUG(D_INFO, " Cache reqs: "LPU64"\n", cache->fci_stat.fst_cache);
118 CDEBUG(D_INFO, " Cache hits: "LPU64"%%\n", pct);
124 * delete given node from list.
126 void fld_cache_entry_delete(struct fld_cache *cache,
127 struct fld_cache_entry *node)
129 list_del(&node->fce_list);
130 list_del(&node->fce_lru);
131 cache->fci_cache_count--;
136 * fix list by checking new entry with NEXT entry in order.
138 static void fld_fix_new_list(struct fld_cache *cache)
140 struct fld_cache_entry *f_curr;
141 struct fld_cache_entry *f_next;
142 struct lu_seq_range *c_range;
143 struct lu_seq_range *n_range;
144 struct list_head *head = &cache->fci_entries_head;
148 list_for_each_entry_safe(f_curr, f_next, head, fce_list) {
149 c_range = &f_curr->fce_range;
150 n_range = &f_next->fce_range;
152 LASSERT(range_is_sane(c_range));
153 if (&f_next->fce_list == head)
156 if (c_range->lsr_flags != n_range->lsr_flags)
159 LASSERTF(c_range->lsr_start <= n_range->lsr_start,
160 "cur lsr_start "DRANGE" next lsr_start "DRANGE"\n",
161 PRANGE(c_range), PRANGE(n_range));
163 /* check merge possibility with next range */
164 if (c_range->lsr_end == n_range->lsr_start) {
165 if (c_range->lsr_index != n_range->lsr_index)
167 n_range->lsr_start = c_range->lsr_start;
168 fld_cache_entry_delete(cache, f_curr);
172 /* check if current range overlaps with next range. */
173 if (n_range->lsr_start < c_range->lsr_end) {
174 if (c_range->lsr_index == n_range->lsr_index) {
175 n_range->lsr_start = c_range->lsr_start;
176 n_range->lsr_end = max(c_range->lsr_end,
178 fld_cache_entry_delete(cache, f_curr);
180 if (n_range->lsr_end <= c_range->lsr_end) {
182 fld_cache_entry_delete(cache, f_curr);
184 n_range->lsr_start = c_range->lsr_end;
187 /* we could have overlap over next
188 * range too. better restart. */
192 /* kill duplicates */
193 if (c_range->lsr_start == n_range->lsr_start &&
194 c_range->lsr_end == n_range->lsr_end)
195 fld_cache_entry_delete(cache, f_curr);
200 * add node to fld cache
202 static inline void fld_cache_entry_add(struct fld_cache *cache,
203 struct fld_cache_entry *f_new,
204 struct list_head *pos)
206 list_add(&f_new->fce_list, pos);
207 list_add(&f_new->fce_lru, &cache->fci_lru);
209 cache->fci_cache_count++;
210 fld_fix_new_list(cache);
214 * Check if cache needs to be shrunk. If so - do it.
215 * Remove one entry in list and so on until cache is shrunk enough.
217 static int fld_cache_shrink(struct fld_cache *cache)
219 struct fld_cache_entry *flde;
220 struct list_head *curr;
223 LASSERT(cache != NULL);
225 if (cache->fci_cache_count < cache->fci_cache_size)
228 curr = cache->fci_lru.prev;
230 while (cache->fci_cache_count + cache->fci_threshold >
231 cache->fci_cache_size && curr != &cache->fci_lru) {
233 flde = list_entry(curr, struct fld_cache_entry, fce_lru);
235 fld_cache_entry_delete(cache, flde);
239 CDEBUG(D_INFO, "%s: FLD cache - Shrunk by "
240 "%d entries\n", cache->fci_name, num);
246 * kill all fld cache entries.
248 void fld_cache_flush(struct fld_cache *cache)
250 write_lock(&cache->fci_lock);
251 cache->fci_cache_size = 0;
252 fld_cache_shrink(cache);
253 write_unlock(&cache->fci_lock);
257 * punch hole in existing range. divide this range and add new
261 void fld_cache_punch_hole(struct fld_cache *cache,
262 struct fld_cache_entry *f_curr,
263 struct fld_cache_entry *f_new)
265 const struct lu_seq_range *range = &f_new->fce_range;
266 const seqno_t new_start = range->lsr_start;
267 const seqno_t new_end = range->lsr_end;
268 struct fld_cache_entry *fldt;
270 OBD_ALLOC_GFP(fldt, sizeof(*fldt), GFP_ATOMIC);
273 /* overlap is not allowed, so dont mess up list. */
276 /* break f_curr RANGE into three RANGES:
277 * f_curr, f_new , fldt
283 fldt->fce_range.lsr_start = new_end;
284 fldt->fce_range.lsr_end = f_curr->fce_range.lsr_end;
285 fldt->fce_range.lsr_index = f_curr->fce_range.lsr_index;
288 f_curr->fce_range.lsr_end = new_start;
290 /* add these two entries to list */
291 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
292 fld_cache_entry_add(cache, fldt, &f_new->fce_list);
294 /* no need to fixup */
298 * handle range overlap in fld cache.
300 static void fld_cache_overlap_handle(struct fld_cache *cache,
301 struct fld_cache_entry *f_curr,
302 struct fld_cache_entry *f_new)
304 const struct lu_seq_range *range = &f_new->fce_range;
305 const seqno_t new_start = range->lsr_start;
306 const seqno_t new_end = range->lsr_end;
307 const mdsno_t mdt = range->lsr_index;
309 /* this is overlap case, these case are checking overlapping with
310 * prev range only. fixup will handle overlaping with next range. */
312 if (f_curr->fce_range.lsr_index == mdt) {
313 f_curr->fce_range.lsr_start = min(f_curr->fce_range.lsr_start,
316 f_curr->fce_range.lsr_end = max(f_curr->fce_range.lsr_end,
320 fld_fix_new_list(cache);
322 } else if (new_start <= f_curr->fce_range.lsr_start &&
323 f_curr->fce_range.lsr_end <= new_end) {
324 /* case 1: new range completely overshadowed existing range.
325 * e.g. whole range migrated. update fld cache entry */
327 f_curr->fce_range = *range;
329 fld_fix_new_list(cache);
331 } else if (f_curr->fce_range.lsr_start < new_start &&
332 new_end < f_curr->fce_range.lsr_end) {
333 /* case 2: new range fit within existing range. */
335 fld_cache_punch_hole(cache, f_curr, f_new);
337 } else if (new_end <= f_curr->fce_range.lsr_end) {
339 * [new_start [c_start new_end) c_end)
342 LASSERT(new_start <= f_curr->fce_range.lsr_start);
344 f_curr->fce_range.lsr_start = new_end;
345 fld_cache_entry_add(cache, f_new, f_curr->fce_list.prev);
347 } else if (f_curr->fce_range.lsr_start <= new_start) {
349 * [c_start [new_start c_end) new_end)
352 LASSERT(f_curr->fce_range.lsr_end <= new_end);
354 f_curr->fce_range.lsr_end = new_start;
355 fld_cache_entry_add(cache, f_new, &f_curr->fce_list);
357 CERROR("NEW range ="DRANGE" curr = "DRANGE"\n",
358 PRANGE(range),PRANGE(&f_curr->fce_range));
361 struct fld_cache_entry
362 *fld_cache_entry_create(const struct lu_seq_range *range)
364 struct fld_cache_entry *f_new;
366 LASSERT(range_is_sane(range));
368 OBD_ALLOC_PTR(f_new);
370 return ERR_PTR(-ENOMEM);
372 f_new->fce_range = *range;
377 * Insert FLD entry in FLD cache.
379 * This function handles all cases of merging and breaking up of
382 int fld_cache_insert_nolock(struct fld_cache *cache,
383 struct fld_cache_entry *f_new)
385 struct fld_cache_entry *f_curr;
386 struct fld_cache_entry *n;
387 struct list_head *head;
388 struct list_head *prev = NULL;
389 const seqno_t new_start = f_new->fce_range.lsr_start;
390 const seqno_t new_end = f_new->fce_range.lsr_end;
391 __u32 new_flags = f_new->fce_range.lsr_flags;
394 * Duplicate entries are eliminated in insert op.
395 * So we don't need to search new entry before starting
399 if (!cache->fci_no_shrink)
400 fld_cache_shrink(cache);
402 head = &cache->fci_entries_head;
404 list_for_each_entry_safe(f_curr, n, head, fce_list) {
405 /* add list if next is end of list */
406 if (new_end < f_curr->fce_range.lsr_start ||
407 (new_end == f_curr->fce_range.lsr_start &&
408 new_flags != f_curr->fce_range.lsr_flags))
411 prev = &f_curr->fce_list;
412 /* check if this range is to left of new range. */
413 if (new_start < f_curr->fce_range.lsr_end &&
414 new_flags == f_curr->fce_range.lsr_flags) {
415 fld_cache_overlap_handle(cache, f_curr, f_new);
423 CDEBUG(D_INFO, "insert range "DRANGE"\n", PRANGE(&f_new->fce_range));
424 /* Add new entry to cache and lru list. */
425 fld_cache_entry_add(cache, f_new, prev);
430 int fld_cache_insert(struct fld_cache *cache,
431 const struct lu_seq_range *range)
433 struct fld_cache_entry *flde;
436 flde = fld_cache_entry_create(range);
438 return PTR_ERR(flde);
440 write_lock(&cache->fci_lock);
441 rc = fld_cache_insert_nolock(cache, flde);
442 write_unlock(&cache->fci_lock);
449 void fld_cache_delete_nolock(struct fld_cache *cache,
450 const struct lu_seq_range *range)
452 struct fld_cache_entry *flde;
453 struct fld_cache_entry *tmp;
454 struct list_head *head;
456 head = &cache->fci_entries_head;
457 list_for_each_entry_safe(flde, tmp, head, fce_list) {
458 /* add list if next is end of list */
459 if (range->lsr_start == flde->fce_range.lsr_start ||
460 (range->lsr_end == flde->fce_range.lsr_end &&
461 range->lsr_flags == flde->fce_range.lsr_flags)) {
462 fld_cache_entry_delete(cache, flde);
469 * Delete FLD entry in FLD cache.
472 void fld_cache_delete(struct fld_cache *cache,
473 const struct lu_seq_range *range)
475 write_lock(&cache->fci_lock);
476 fld_cache_delete_nolock(cache, range);
477 write_unlock(&cache->fci_lock);
480 struct fld_cache_entry
481 *fld_cache_entry_lookup_nolock(struct fld_cache *cache,
482 struct lu_seq_range *range)
484 struct fld_cache_entry *flde;
485 struct fld_cache_entry *got = NULL;
486 struct list_head *head;
488 head = &cache->fci_entries_head;
489 list_for_each_entry(flde, head, fce_list) {
490 if (range->lsr_start == flde->fce_range.lsr_start ||
491 (range->lsr_end == flde->fce_range.lsr_end &&
492 range->lsr_flags == flde->fce_range.lsr_flags)) {
502 * lookup \a seq sequence for range in fld cache.
504 struct fld_cache_entry
505 *fld_cache_entry_lookup(struct fld_cache *cache, struct lu_seq_range *range)
507 struct fld_cache_entry *got = NULL;
509 read_lock(&cache->fci_lock);
510 got = fld_cache_entry_lookup_nolock(cache, range);
511 read_unlock(&cache->fci_lock);
516 * lookup \a seq sequence for range in fld cache.
518 int fld_cache_lookup(struct fld_cache *cache,
519 const seqno_t seq, struct lu_seq_range *range)
521 struct fld_cache_entry *flde;
522 struct fld_cache_entry *prev = NULL;
523 struct list_head *head;
525 read_lock(&cache->fci_lock);
526 head = &cache->fci_entries_head;
528 cache->fci_stat.fst_count++;
529 list_for_each_entry(flde, head, fce_list) {
530 if (flde->fce_range.lsr_start > seq) {
532 *range = prev->fce_range;
537 if (range_within(&flde->fce_range, seq)) {
538 *range = flde->fce_range;
540 cache->fci_stat.fst_cache++;
541 read_unlock(&cache->fci_lock);
545 read_unlock(&cache->fci_lock);