1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
38 * Cache of triples - object, lock, extent
42 # define EXPORT_SYMTAB
44 #define DEBUG_SUBSYSTEM S_OSC
47 # include <linux/version.h>
48 # include <linux/module.h>
49 # include <linux/list.h>
50 #else /* __KERNEL__ */
51 # include <liblustre.h>
54 #include <lustre_dlm.h>
55 #include <lustre_cache.h>
57 #include <lustre_debug.h>
59 #include "osc_internal.h"
61 /* Adding @lock to the @cache */
62 int cache_add_lock(struct lustre_cache *cache, struct lustre_handle *lockh)
64 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
66 if (!lock) // Lock disappeared under us.
69 spin_lock(&cache->lc_locks_list_lock);
70 list_add_tail(&lock->l_cache_locks_list, &cache->lc_locks_list);
71 spin_unlock(&cache->lc_locks_list_lock);
78 /* Tries to add @extent to lock represented by @lockh if non-NULL, otherwise
79 just tries to match some suitable lock by resource and data contained in
81 /* Should be called with oap->lock held (except on initial addition, see
82 comment in osc_request.c*/
83 int cache_add_extent(struct lustre_cache *cache, struct ldlm_res_id *res,
84 struct osc_async_page *extent, struct lustre_handle *lockh)
86 struct lustre_handle tmplockh;
87 ldlm_policy_data_t tmpex;
88 struct ldlm_lock *lock = NULL;
92 /* Don't add anything second time */
93 if (!list_empty(&extent->oap_page_list)) {
98 if (lockh && lustre_handle_is_used(lockh)) {
99 lock = ldlm_handle2lock(lockh);
103 if(lock->l_policy_data.l_extent.start > extent->oap_obj_off ||
104 extent->oap_obj_off + CFS_PAGE_SIZE - 1 >
105 lock->l_policy_data.l_extent.end) {
106 CDEBUG(D_CACHE, "Got wrong lock [" LPU64 "," LPU64 "] "
107 "for page with offset " LPU64 "\n",
108 lock->l_policy_data.l_extent.start,
109 lock->l_policy_data.l_extent.end,
110 extent->oap_obj_off);
115 /* Real extent width calculation here once we have real
118 tmpex.l_extent.start = extent->oap_obj_off;
119 tmpex.l_extent.end = tmpex.l_extent.start + CFS_PAGE_SIZE - 1;
121 /* XXX find lock from extent or something like that */
122 /* The lock mode does not matter. If this is dirty page - then
123 * there could be only one PW lock. If the page is clean,
124 * any PR lock is good
126 mode = ldlm_lock_match(cache->lc_obd->obd_namespace,
127 LDLM_FL_BLOCK_GRANTED |
128 LDLM_FL_CBPENDING, res, LDLM_EXTENT,
129 &tmpex, LCK_PW | LCK_PR, &tmplockh);
132 CDEBUG(D_CACHE, "No lock to attach " LPU64 "->" LPU64
133 " extent to!\n", tmpex.l_extent.start,
135 RETURN((mode < 0) ? mode : -ENOLCK);
138 lock = ldlm_handle2lock(&tmplockh);
139 if (!lock) { // Race - lock disappeared under us (eviction?)
140 CDEBUG(D_CACHE, "Newly matched lock just disappeared "
145 /* XXX Note! if the caller passed a unused lock handle,
146 * it expects us to return the lockh of the lock we matched,
147 * reference(LCK_PR) of the lock is increased here to assure
148 * its validity, and the caller should decrease the reference
149 * when it isn't used any more. */
150 if (lockh && !lustre_handle_is_used(lockh)) {
151 ldlm_lock_addref(&tmplockh, LCK_PR);
152 lustre_handle_copy(lockh, &tmplockh);
156 spin_lock(&lock->l_extents_list_lock);
157 list_add_tail(&extent->oap_page_list, &lock->l_extents_list);
158 spin_unlock(&lock->l_extents_list_lock);
159 extent->oap_ldlm_lock = lock;
160 LASSERTF(!(lock->l_flags & LDLM_FL_CANCEL), "Adding a page to already "
161 "cancelled lock %p", lock);
163 ldlm_lock_decref(&tmplockh, mode);
169 static void cache_extent_removal_get(struct page_removal_cb_element *element)
171 atomic_inc(&element->prce_refcnt);
174 static void cache_extent_removal_put(struct page_removal_cb_element *element)
176 if(atomic_dec_and_test(&element->prce_refcnt))
177 OBD_FREE_PTR(element);
180 static int cache_extent_removal_event(struct lustre_cache *cache,
181 void *data, int discard)
183 struct page *page = data;
184 struct list_head *iter;
185 struct page_removal_cb_element *element;
187 read_lock(&cache->lc_page_removal_cb_lock);
188 iter = cache->lc_page_removal_callback_list.next;
189 while(iter != &cache->lc_page_removal_callback_list) {
190 element = list_entry(iter, struct page_removal_cb_element, prce_list);
191 cache_extent_removal_get(element);
192 read_unlock(&cache->lc_page_removal_cb_lock);
194 element->prce_callback(page, discard);
196 read_lock(&cache->lc_page_removal_cb_lock);
198 cache_extent_removal_put(element);
200 read_unlock(&cache->lc_page_removal_cb_lock);
205 /* Registers set of pin/remove callbacks for extents. Current limitation is
206 there could be only one pin_cb per cache.
207 @pin_cb is called when we have the page locked to pin it in memory so that
208 it does not disappear after we release page lock (which we need to do
210 @func_cb is removal callback that is called after page and all spinlocks are
211 released, and is supposed to clean the page and remove it from all
212 (vfs) caches it might be in */
213 int cache_add_extent_removal_cb(struct lustre_cache *cache,
214 obd_page_removal_cb_t func_cb,
215 obd_pin_extent_cb pin_cb)
217 struct page_removal_cb_element *element;
222 OBD_ALLOC_PTR(element);
225 element->prce_callback = func_cb;
226 atomic_set(&element->prce_refcnt, 1);
228 write_lock(&cache->lc_page_removal_cb_lock);
229 list_add_tail(&element->prce_list,
230 &cache->lc_page_removal_callback_list);
231 write_unlock(&cache->lc_page_removal_cb_lock);
233 cache->lc_pin_extent_cb = pin_cb;
236 EXPORT_SYMBOL(cache_add_extent_removal_cb);
238 /* Unregister exntent removal callback registered earlier. If the list of
239 registered removal callbacks becomes empty, we also clear pin callback
240 since it could only be one */
241 int cache_del_extent_removal_cb(struct lustre_cache *cache,
242 obd_page_removal_cb_t func_cb)
245 struct page_removal_cb_element *element, *t;
248 write_lock(&cache->lc_page_removal_cb_lock);
249 list_for_each_entry_safe(element, t,
250 &cache->lc_page_removal_callback_list,
252 if (element->prce_callback == func_cb) {
253 list_del(&element->prce_list);
254 write_unlock(&cache->lc_page_removal_cb_lock);
256 cache_extent_removal_put(element);
257 write_lock(&cache->lc_page_removal_cb_lock);
258 /* We continue iterating the list in case this function
259 was registered more than once */
262 write_unlock(&cache->lc_page_removal_cb_lock);
264 if (list_empty(&cache->lc_page_removal_callback_list))
265 cache->lc_pin_extent_cb = NULL;
269 EXPORT_SYMBOL(cache_del_extent_removal_cb);
271 static int cache_remove_extent_nolock(struct lustre_cache *cache,
272 struct osc_async_page *extent)
274 int have_lock = !!extent->oap_ldlm_lock;
275 /* We used to check oap_ldlm_lock for non NULL here, but it might be
276 NULL, in fact, due to parallel page eviction clearing it and waiting
277 on a lock's page list lock */
278 extent->oap_ldlm_lock = NULL;
280 if (!list_empty(&extent->oap_page_list))
281 list_del_init(&extent->oap_page_list);
286 /* Request the @extent to be removed from cache and locks it belongs to. */
287 void cache_remove_extent(struct lustre_cache *cache,
288 struct osc_async_page *extent)
290 struct ldlm_lock *lock;
292 spin_lock(&extent->oap_lock);
293 lock = extent->oap_ldlm_lock;
295 extent->oap_ldlm_lock = NULL;
296 spin_unlock(&extent->oap_lock);
298 /* No lock - means this extent is not in any list */
302 spin_lock(&lock->l_extents_list_lock);
303 if (!list_empty(&extent->oap_page_list))
304 list_del_init(&extent->oap_page_list);
305 spin_unlock(&lock->l_extents_list_lock);
308 /* Iterate through list of extents in given lock identified by @lockh,
309 calling @cb_func for every such extent. Also passed @data to every call.
310 Stops iterating prematurely if @cb_func returns nonzero. */
311 int cache_iterate_extents(struct lustre_cache *cache,
312 struct lustre_handle *lockh,
313 cache_iterate_extents_cb_t cb_func, void *data)
315 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
316 struct osc_async_page *extent, *t;
318 if (!lock) // Lock disappeared
320 /* Parallel page removal from mem pressure can race with us */
321 spin_lock(&lock->l_extents_list_lock);
322 list_for_each_entry_safe(extent, t, &lock->l_extents_list,
324 if (cb_func(cache, lockh, extent, data))
327 spin_unlock(&lock->l_extents_list_lock);
333 static int cache_remove_extents_from_lock(struct lustre_cache *cache,
334 struct ldlm_lock *lock, void *data)
336 struct osc_async_page *extent;
341 spin_lock(&lock->l_extents_list_lock);
342 while (!list_empty(&lock->l_extents_list)) {
343 extent = list_entry(lock->l_extents_list.next,
344 struct osc_async_page, oap_page_list);
346 spin_lock(&extent->oap_lock);
347 /* If there is no lock referenced from this oap, it means
348 there is parallel page-removal process waiting to free that
349 page on l_extents_list_lock and it holds page lock.
350 We need this page to completely go away and for that to
351 happen we will just try to truncate it here too.
352 Serialisation on page lock will achieve that goal for us. */
353 /* Try to add extent back to the cache first, but only if we
354 * cancel read lock, write locks cannot have other overlapping
355 * locks. If adding is not possible (or canceling pw lock),
356 * then remove extent from cache */
357 if (!cache_remove_extent_nolock(cache, extent) ||
358 (lock->l_granted_mode == LCK_PW) ||
359 cache_add_extent(cache, &lock->l_resource->lr_name, extent,
361 /* We need to remember this oap_page value now,
362 once we release spinlocks, extent struct
363 might be freed and we endup requesting
364 page with address 0x5a5a5a5a in
365 cache_extent_removal_event */
366 ext_data = extent->oap_page;
367 LASSERT(cache->lc_pin_extent_cb != NULL);
368 cache->lc_pin_extent_cb(extent->oap_page);
370 if (lock->l_flags & LDLM_FL_BL_AST)
371 extent->oap_async_flags |= ASYNC_HP;
372 spin_unlock(&extent->oap_lock);
373 spin_unlock(&lock->l_extents_list_lock);
374 cache_extent_removal_event(cache, ext_data,
377 LDLM_FL_DISCARD_DATA);
378 spin_lock(&lock->l_extents_list_lock);
380 spin_unlock(&extent->oap_lock);
383 spin_unlock(&lock->l_extents_list_lock);
388 /* Remoes @lock from cache after necessary checks. */
389 int cache_remove_lock(struct lustre_cache *cache, struct lustre_handle *lockh)
391 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
393 if (!lock) // The lock was removed by somebody just now, nothing to do
396 cache_remove_extents_from_lock(cache, lock, NULL /*data */ );
398 spin_lock(&cache->lc_locks_list_lock);
399 list_del_init(&lock->l_cache_locks_list);
400 spin_unlock(&cache->lc_locks_list_lock);
407 /* Supposed to iterate through all locks in the cache for given resource.
408 Not implemented atthe moment. */
409 int cache_iterate_locks(struct lustre_cache *cache, struct ldlm_res_id *res,
410 cache_iterate_locks_cb_t cb_fun, void *data)
415 /* Create lustre cache and attach it to @obd */
416 struct lustre_cache *cache_create(struct obd_device *obd)
418 struct lustre_cache *cache;
420 OBD_ALLOC(cache, sizeof(*cache));
424 spin_lock_init(&cache->lc_locks_list_lock);
425 CFS_INIT_LIST_HEAD(&cache->lc_locks_list);
426 CFS_INIT_LIST_HEAD(&cache->lc_page_removal_callback_list);
427 rwlock_init(&cache->lc_page_removal_cb_lock);
434 /* Destroy @cache and free its memory */
435 int cache_destroy(struct lustre_cache *cache)
440 spin_lock(&cache->lc_locks_list_lock);
441 if (!list_empty(&cache->lc_locks_list)) {
442 struct ldlm_lock *lock, *tmp;
443 CERROR("still have locks in the list on cleanup:\n");
445 list_for_each_entry_safe(lock, tmp,
446 &cache->lc_locks_list,
447 l_cache_locks_list) {
448 list_del_init(&lock->l_cache_locks_list);
449 /* XXX: Of course natural idea would be to print
450 * offending locks here, but if we use
451 * e.g. LDLM_ERROR, we will likely crash here,
452 * as LDLM error tries to access e.g.
453 * nonexisting namespace. Normally this kind of
454 * case could only happen when somebody did not
455 * release lock reference and we have other ways
457 /* Make sure there are no pages left under the
459 LASSERT(list_empty(&lock->l_extents_list));
462 spin_unlock(&cache->lc_locks_list_lock);
463 LASSERT(list_empty(&cache->lc_page_removal_callback_list));
465 OBD_FREE(cache, sizeof(*cache));