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) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
33 * Author: Nikita Danilov <nikita.danilov@sun.com>
34 * Author: Jinshan Xiong <jinshan.xiong@intel.com>
37 #define DEBUG_SUBSYSTEM S_CLASS
39 #include <linux/list.h>
40 #include <libcfs/libcfs.h>
41 #include <obd_class.h>
42 #include <obd_support.h>
44 #include <cl_object.h>
45 #include "cl_internal.h"
47 static void cl_page_delete0(const struct lu_env *env, struct cl_page *pg);
48 static DEFINE_MUTEX(cl_page_kmem_mutex);
51 # define PASSERT(env, page, expr) \
53 if (unlikely(!(expr))) { \
54 CL_PAGE_DEBUG(D_ERROR, (env), (page), #expr "\n"); \
58 #else /* !LIBCFS_DEBUG */
59 # define PASSERT(env, page, exp) \
60 ((void)sizeof(env), (void)sizeof(page), (void)sizeof !!(exp))
61 #endif /* !LIBCFS_DEBUG */
63 #ifdef CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK
64 # define PINVRNT(env, page, expr) \
66 if (unlikely(!(expr))) { \
67 CL_PAGE_DEBUG(D_ERROR, (env), (page), #expr "\n"); \
71 #else /* !CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK */
72 # define PINVRNT(env, page, exp) \
73 ((void)sizeof(env), (void)sizeof(page), (void)sizeof !!(exp))
74 #endif /* !CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK */
76 /* Disable page statistic by default due to huge performance penalty. */
77 static void cs_page_inc(const struct cl_object *obj,
78 enum cache_stats_item item)
80 #ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
81 atomic_inc(&cl_object_site(obj)->cs_pages.cs_stats[item]);
85 static void cs_page_dec(const struct cl_object *obj,
86 enum cache_stats_item item)
88 #ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
89 atomic_dec(&cl_object_site(obj)->cs_pages.cs_stats[item]);
93 static void cs_pagestate_inc(const struct cl_object *obj,
94 enum cl_page_state state)
96 #ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
97 atomic_inc(&cl_object_site(obj)->cs_pages_state[state]);
101 static void cs_pagestate_dec(const struct cl_object *obj,
102 enum cl_page_state state)
104 #ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
105 atomic_dec(&cl_object_site(obj)->cs_pages_state[state]);
110 * Internal version of cl_page_get().
112 * This function can be used to obtain initial reference to previously
113 * unreferenced cached object. It can be called only if concurrent page
114 * reclamation is somehow prevented, e.g., by keeping a lock on a VM page,
115 * associated with \a page.
117 * Use with care! Not exported.
119 static void cl_page_get_trust(struct cl_page *page)
121 LASSERT(atomic_read(&page->cp_ref) > 0);
122 atomic_inc(&page->cp_ref);
125 static struct cl_page_slice *
126 cl_page_slice_get(const struct cl_page *cl_page, int index)
128 if (index < 0 || index >= cl_page->cp_layer_count)
131 /* To get the cp_layer_offset values fit under 256 bytes, we
132 * use the offset beyond the end of struct cl_page.
134 return (struct cl_page_slice *)((char *)cl_page + sizeof(*cl_page) +
135 cl_page->cp_layer_offset[index]);
138 #define cl_page_slice_for_each(cl_page, slice, i) \
139 for (i = 0, slice = cl_page_slice_get(cl_page, 0); \
140 i < (cl_page)->cp_layer_count; \
141 slice = cl_page_slice_get(cl_page, ++i))
143 #define cl_page_slice_for_each_reverse(cl_page, slice, i) \
144 for (i = (cl_page)->cp_layer_count - 1, \
145 slice = cl_page_slice_get(cl_page, i); i >= 0; \
146 slice = cl_page_slice_get(cl_page, --i))
149 * Returns a slice within a cl_page, corresponding to the given layer in the
154 static const struct cl_page_slice *
155 cl_page_at_trusted(const struct cl_page *cl_page,
156 const struct lu_device_type *dtype)
158 const struct cl_page_slice *slice;
163 cl_page_slice_for_each(cl_page, slice, i) {
164 if (slice->cpl_obj->co_lu.lo_dev->ld_type == dtype)
171 static void __cl_page_free(struct cl_page *cl_page, unsigned short bufsize)
173 int index = cl_page->cp_kmem_index;
176 LASSERT(index < ARRAY_SIZE(cl_page_kmem_array));
177 LASSERT(cl_page_kmem_size_array[index] == bufsize);
178 OBD_SLAB_FREE(cl_page, cl_page_kmem_array[index], bufsize);
180 OBD_FREE(cl_page, bufsize);
184 static void cl_page_free(const struct lu_env *env, struct cl_page *cl_page,
185 struct pagevec *pvec)
187 struct cl_object *obj = cl_page->cp_obj;
188 unsigned short bufsize = cl_object_header(obj)->coh_page_bufsize;
189 struct cl_page_slice *slice;
193 PASSERT(env, cl_page, list_empty(&cl_page->cp_batch));
194 PASSERT(env, cl_page, cl_page->cp_owner == NULL);
195 PASSERT(env, cl_page, cl_page->cp_state == CPS_FREEING);
197 cl_page_slice_for_each(cl_page, slice, i) {
198 if (unlikely(slice->cpl_ops->cpo_fini != NULL))
199 slice->cpl_ops->cpo_fini(env, slice, pvec);
201 cl_page->cp_layer_count = 0;
202 cs_page_dec(obj, CS_total);
203 cs_pagestate_dec(obj, cl_page->cp_state);
204 lu_object_ref_del_at(&obj->co_lu, &cl_page->cp_obj_ref,
206 cl_object_put(env, obj);
207 lu_ref_fini(&cl_page->cp_reference);
208 __cl_page_free(cl_page, bufsize);
212 static struct cl_page *__cl_page_alloc(struct cl_object *o)
215 struct cl_page *cl_page = NULL;
216 unsigned short bufsize = cl_object_header(o)->coh_page_bufsize;
219 /* the number of entries in cl_page_kmem_array is expected to
220 * only be 2-3 entries, so the lookup overhead should be low.
222 for ( ; i < ARRAY_SIZE(cl_page_kmem_array); i++) {
223 if (smp_load_acquire(&cl_page_kmem_size_array[i])
225 OBD_SLAB_ALLOC_GFP(cl_page, cl_page_kmem_array[i],
228 cl_page->cp_kmem_index = i;
231 if (cl_page_kmem_size_array[i] == 0)
235 if (i < ARRAY_SIZE(cl_page_kmem_array)) {
238 mutex_lock(&cl_page_kmem_mutex);
239 if (cl_page_kmem_size_array[i]) {
240 mutex_unlock(&cl_page_kmem_mutex);
243 snprintf(cache_name, sizeof(cache_name),
244 "cl_page_kmem-%u", bufsize);
245 cl_page_kmem_array[i] =
246 kmem_cache_create(cache_name, bufsize,
248 if (cl_page_kmem_array[i] == NULL) {
249 mutex_unlock(&cl_page_kmem_mutex);
252 smp_store_release(&cl_page_kmem_size_array[i],
254 mutex_unlock(&cl_page_kmem_mutex);
257 OBD_ALLOC_GFP(cl_page, bufsize, GFP_NOFS);
259 cl_page->cp_kmem_index = -1;
265 struct cl_page *cl_page_alloc(const struct lu_env *env, struct cl_object *o,
266 pgoff_t ind, struct page *vmpage,
267 enum cl_page_type type)
269 struct cl_page *cl_page;
270 struct cl_object *head;
274 cl_page = __cl_page_alloc(o);
275 if (cl_page != NULL) {
279 * Please fix cl_page:cp_state/type declaration if
280 * these assertions fail in the future.
282 BUILD_BUG_ON((1 << CP_STATE_BITS) < CPS_NR); /* cp_state */
283 BUILD_BUG_ON((1 << CP_TYPE_BITS) < CPT_NR); /* cp_type */
284 atomic_set(&cl_page->cp_ref, 1);
287 lu_object_ref_add_at(&o->co_lu, &cl_page->cp_obj_ref,
289 cl_page->cp_vmpage = vmpage;
290 cl_page->cp_state = CPS_CACHED;
291 cl_page->cp_type = type;
292 cl_page->cp_inode = NULL;
293 INIT_LIST_HEAD(&cl_page->cp_batch);
294 lu_ref_init(&cl_page->cp_reference);
296 cl_object_for_each(o, head) {
297 if (o->co_ops->coo_page_init != NULL) {
298 result = o->co_ops->coo_page_init(env, o,
301 cl_page_delete0(env, cl_page);
302 cl_page_free(env, cl_page, NULL);
303 cl_page = ERR_PTR(result);
309 cs_page_inc(o, CS_total);
310 cs_page_inc(o, CS_create);
311 cs_pagestate_dec(o, CPS_CACHED);
314 cl_page = ERR_PTR(-ENOMEM);
320 * Returns a cl_page with index \a idx at the object \a o, and associated with
321 * the VM page \a vmpage.
323 * This is the main entry point into the cl_page caching interface. First, a
324 * cache (implemented as a per-object radix tree) is consulted. If page is
325 * found there, it is returned immediately. Otherwise new page is allocated
326 * and returned. In any case, additional reference to page is acquired.
328 * \see cl_object_find(), cl_lock_find()
330 struct cl_page *cl_page_find(const struct lu_env *env,
332 pgoff_t idx, struct page *vmpage,
333 enum cl_page_type type)
335 struct cl_page *page = NULL;
336 struct cl_object_header *hdr;
338 LASSERT(type == CPT_CACHEABLE || type == CPT_TRANSIENT);
343 hdr = cl_object_header(o);
344 cs_page_inc(o, CS_lookup);
346 CDEBUG(D_PAGE, "%lu@"DFID" %p %lx %d\n",
347 idx, PFID(&hdr->coh_lu.loh_fid), vmpage, vmpage->private, type);
349 if (type == CPT_CACHEABLE) {
350 /* vmpage lock is used to protect the child/parent
352 LASSERT(PageLocked(vmpage));
354 * cl_vmpage_page() can be called here without any locks as
356 * - "vmpage" is locked (which prevents ->private from
357 * concurrent updates), and
359 * - "o" cannot be destroyed while current thread holds a
362 page = cl_vmpage_page(vmpage, o);
364 cs_page_inc(o, CS_hit);
369 /* allocate and initialize cl_page */
370 page = cl_page_alloc(env, o, idx, vmpage, type);
373 EXPORT_SYMBOL(cl_page_find);
375 static inline int cl_page_invariant(const struct cl_page *pg)
377 return cl_page_in_use_noref(pg);
380 static void cl_page_state_set0(const struct lu_env *env,
381 struct cl_page *cl_page,
382 enum cl_page_state state)
384 enum cl_page_state old;
387 * Matrix of allowed state transitions [old][new], for sanity
390 static const int allowed_transitions[CPS_NR][CPS_NR] = {
393 [CPS_OWNED] = 1, /* io finds existing cached page */
395 [CPS_PAGEOUT] = 1, /* write-out from the cache */
396 [CPS_FREEING] = 1, /* eviction on the memory pressure */
399 [CPS_CACHED] = 1, /* release to the cache */
401 [CPS_PAGEIN] = 1, /* start read immediately */
402 [CPS_PAGEOUT] = 1, /* start write immediately */
403 [CPS_FREEING] = 1, /* lock invalidation or truncate */
406 [CPS_CACHED] = 1, /* io completion */
413 [CPS_CACHED] = 1, /* io completion */
429 old = cl_page->cp_state;
430 PASSERT(env, cl_page, allowed_transitions[old][state]);
431 CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d -> %d\n", old, state);
432 PASSERT(env, cl_page, cl_page->cp_state == old);
433 PASSERT(env, cl_page, equi(state == CPS_OWNED,
434 cl_page->cp_owner != NULL));
436 cs_pagestate_dec(cl_page->cp_obj, cl_page->cp_state);
437 cs_pagestate_inc(cl_page->cp_obj, state);
438 cl_page->cp_state = state;
442 static void cl_page_state_set(const struct lu_env *env,
443 struct cl_page *page, enum cl_page_state state)
445 cl_page_state_set0(env, page, state);
449 * Acquires an additional reference to a page.
451 * This can be called only by caller already possessing a reference to \a
454 * \see cl_object_get(), cl_lock_get().
456 void cl_page_get(struct cl_page *page)
459 cl_page_get_trust(page);
462 EXPORT_SYMBOL(cl_page_get);
465 * Releases a reference to a page, use the pagevec to release the pages
466 * in batch if provided.
468 * Users need to do a final pagevec_release() to release any trailing pages.
470 void cl_pagevec_put(const struct lu_env *env, struct cl_page *page,
471 struct pagevec *pvec)
474 CL_PAGE_HEADER(D_TRACE, env, page, "%d\n",
475 atomic_read(&page->cp_ref));
477 if (atomic_dec_and_test(&page->cp_ref)) {
478 LASSERT(page->cp_state == CPS_FREEING);
480 LASSERT(atomic_read(&page->cp_ref) == 0);
481 PASSERT(env, page, page->cp_owner == NULL);
482 PASSERT(env, page, list_empty(&page->cp_batch));
484 * Page is no longer reachable by other threads. Tear
487 cl_page_free(env, page, pvec);
492 EXPORT_SYMBOL(cl_pagevec_put);
495 * Releases a reference to a page, wrapper to cl_pagevec_put
497 * When last reference is released, page is returned to the cache, unless it
498 * is in cl_page_state::CPS_FREEING state, in which case it is immediately
501 * \see cl_object_put(), cl_lock_put().
503 void cl_page_put(const struct lu_env *env, struct cl_page *page)
505 cl_pagevec_put(env, page, NULL);
507 EXPORT_SYMBOL(cl_page_put);
510 * Returns a cl_page associated with a VM page, and given cl_object.
512 struct cl_page *cl_vmpage_page(struct page *vmpage, struct cl_object *obj)
514 struct cl_page *page;
517 LASSERT(PageLocked(vmpage));
520 * NOTE: absence of races and liveness of data are guaranteed by page
521 * lock on a "vmpage". That works because object destruction has
522 * bottom-to-top pass.
525 page = (struct cl_page *)vmpage->private;
527 cl_page_get_trust(page);
528 LASSERT(page->cp_type == CPT_CACHEABLE);
532 EXPORT_SYMBOL(cl_vmpage_page);
534 const struct cl_page_slice *cl_page_at(const struct cl_page *page,
535 const struct lu_device_type *dtype)
537 return cl_page_at_trusted(page, dtype);
539 EXPORT_SYMBOL(cl_page_at);
541 static void cl_page_owner_clear(struct cl_page *page)
544 if (page->cp_owner != NULL) {
545 LASSERT(page->cp_owner->ci_owned_nr > 0);
546 page->cp_owner->ci_owned_nr--;
547 page->cp_owner = NULL;
552 static void cl_page_owner_set(struct cl_page *page)
555 LASSERT(page->cp_owner != NULL);
556 page->cp_owner->ci_owned_nr++;
560 void cl_page_disown0(const struct lu_env *env,
561 struct cl_io *io, struct cl_page *cl_page)
563 const struct cl_page_slice *slice;
564 enum cl_page_state state;
568 state = cl_page->cp_state;
569 PINVRNT(env, cl_page, state == CPS_OWNED ||
570 state == CPS_FREEING);
571 PINVRNT(env, cl_page, cl_page_invariant(cl_page) ||
572 state == CPS_FREEING);
573 cl_page_owner_clear(cl_page);
575 if (state == CPS_OWNED)
576 cl_page_state_set(env, cl_page, CPS_CACHED);
578 * Completion call-backs are executed in the bottom-up order, so that
579 * uppermost layer (llite), responsible for VFS/VM interaction runs
580 * last and can release locks safely.
582 cl_page_slice_for_each_reverse(cl_page, slice, i) {
583 if (slice->cpl_ops->cpo_disown != NULL)
584 (*slice->cpl_ops->cpo_disown)(env, slice, io);
591 * returns true, iff page is owned by the given io.
593 int cl_page_is_owned(const struct cl_page *pg, const struct cl_io *io)
595 struct cl_io *top = cl_io_top((struct cl_io *)io);
596 LINVRNT(cl_object_same(pg->cp_obj, io->ci_obj));
598 RETURN(pg->cp_state == CPS_OWNED && pg->cp_owner == top);
600 EXPORT_SYMBOL(cl_page_is_owned);
603 * Try to own a page by IO.
605 * Waits until page is in cl_page_state::CPS_CACHED state, and then switch it
606 * into cl_page_state::CPS_OWNED state.
608 * \pre !cl_page_is_owned(cl_page, io)
609 * \post result == 0 iff cl_page_is_owned(cl_page, io)
613 * \retval -ve failure, e.g., cl_page was destroyed (and landed in
614 * cl_page_state::CPS_FREEING instead of cl_page_state::CPS_CACHED).
615 * or, page was owned by another thread, or in IO.
617 * \see cl_page_disown()
618 * \see cl_page_operations::cpo_own()
619 * \see cl_page_own_try()
622 static int cl_page_own0(const struct lu_env *env, struct cl_io *io,
623 struct cl_page *cl_page, int nonblock)
625 const struct cl_page_slice *slice;
630 PINVRNT(env, cl_page, !cl_page_is_owned(cl_page, io));
633 if (cl_page->cp_state == CPS_FREEING) {
638 cl_page_slice_for_each(cl_page, slice, i) {
639 if (slice->cpl_ops->cpo_own)
640 result = (*slice->cpl_ops->cpo_own)(env, slice,
649 PASSERT(env, cl_page, cl_page->cp_owner == NULL);
650 cl_page->cp_owner = cl_io_top(io);
651 cl_page_owner_set(cl_page);
652 if (cl_page->cp_state != CPS_FREEING) {
653 cl_page_state_set(env, cl_page, CPS_OWNED);
655 cl_page_disown0(env, io, cl_page);
661 PINVRNT(env, cl_page, ergo(result == 0,
662 cl_page_invariant(cl_page)));
667 * Own a page, might be blocked.
669 * \see cl_page_own0()
671 int cl_page_own(const struct lu_env *env, struct cl_io *io, struct cl_page *pg)
673 return cl_page_own0(env, io, pg, 0);
675 EXPORT_SYMBOL(cl_page_own);
678 * Nonblock version of cl_page_own().
680 * \see cl_page_own0()
682 int cl_page_own_try(const struct lu_env *env, struct cl_io *io,
685 return cl_page_own0(env, io, pg, 1);
687 EXPORT_SYMBOL(cl_page_own_try);
691 * Assume page ownership.
693 * Called when page is already locked by the hosting VM.
695 * \pre !cl_page_is_owned(cl_page, io)
696 * \post cl_page_is_owned(cl_page, io)
698 * \see cl_page_operations::cpo_assume()
700 void cl_page_assume(const struct lu_env *env,
701 struct cl_io *io, struct cl_page *cl_page)
703 const struct cl_page_slice *slice;
708 PINVRNT(env, cl_page,
709 cl_object_same(cl_page->cp_obj, io->ci_obj));
712 cl_page_slice_for_each(cl_page, slice, i) {
713 if (slice->cpl_ops->cpo_assume != NULL)
714 (*slice->cpl_ops->cpo_assume)(env, slice, io);
717 PASSERT(env, cl_page, cl_page->cp_owner == NULL);
718 cl_page->cp_owner = cl_io_top(io);
719 cl_page_owner_set(cl_page);
720 cl_page_state_set(env, cl_page, CPS_OWNED);
723 EXPORT_SYMBOL(cl_page_assume);
726 * Releases page ownership without unlocking the page.
728 * Moves cl_page into cl_page_state::CPS_CACHED without releasing a lock
729 * on the underlying VM page (as VM is supposed to do this itself).
731 * \pre cl_page_is_owned(cl_page, io)
732 * \post !cl_page_is_owned(cl_page, io)
734 * \see cl_page_assume()
736 void cl_page_unassume(const struct lu_env *env,
737 struct cl_io *io, struct cl_page *cl_page)
739 const struct cl_page_slice *slice;
743 PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
744 PINVRNT(env, cl_page, cl_page_invariant(cl_page));
747 cl_page_owner_clear(cl_page);
748 cl_page_state_set(env, cl_page, CPS_CACHED);
750 cl_page_slice_for_each_reverse(cl_page, slice, i) {
751 if (slice->cpl_ops->cpo_unassume != NULL)
752 (*slice->cpl_ops->cpo_unassume)(env, slice, io);
757 EXPORT_SYMBOL(cl_page_unassume);
760 * Releases page ownership.
762 * Moves page into cl_page_state::CPS_CACHED.
764 * \pre cl_page_is_owned(pg, io)
765 * \post !cl_page_is_owned(pg, io)
768 * \see cl_page_operations::cpo_disown()
770 void cl_page_disown(const struct lu_env *env,
771 struct cl_io *io, struct cl_page *pg)
773 PINVRNT(env, pg, cl_page_is_owned(pg, io) ||
774 pg->cp_state == CPS_FREEING);
778 cl_page_disown0(env, io, pg);
781 EXPORT_SYMBOL(cl_page_disown);
784 * Called when cl_page is to be removed from the object, e.g.,
785 * as a result of truncate.
787 * Calls cl_page_operations::cpo_discard() top-to-bottom.
789 * \pre cl_page_is_owned(cl_page, io)
791 * \see cl_page_operations::cpo_discard()
793 void cl_page_discard(const struct lu_env *env,
794 struct cl_io *io, struct cl_page *cl_page)
796 const struct cl_page_slice *slice;
799 PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
800 PINVRNT(env, cl_page, cl_page_invariant(cl_page));
802 cl_page_slice_for_each(cl_page, slice, i) {
803 if (slice->cpl_ops->cpo_discard != NULL)
804 (*slice->cpl_ops->cpo_discard)(env, slice, io);
807 EXPORT_SYMBOL(cl_page_discard);
810 * Version of cl_page_delete() that can be called for not fully constructed
811 * cl_pages, e.g. in an error handling cl_page_find()->cl_page_delete0()
812 * path. Doesn't check cl_page invariant.
814 static void cl_page_delete0(const struct lu_env *env,
815 struct cl_page *cl_page)
817 const struct cl_page_slice *slice;
822 PASSERT(env, cl_page, cl_page->cp_state != CPS_FREEING);
825 * Severe all ways to obtain new pointers to @pg.
827 cl_page_owner_clear(cl_page);
828 cl_page_state_set0(env, cl_page, CPS_FREEING);
830 cl_page_slice_for_each_reverse(cl_page, slice, i) {
831 if (slice->cpl_ops->cpo_delete != NULL)
832 (*slice->cpl_ops->cpo_delete)(env, slice);
839 * Called when a decision is made to throw page out of memory.
841 * Notifies all layers about page destruction by calling
842 * cl_page_operations::cpo_delete() method top-to-bottom.
844 * Moves page into cl_page_state::CPS_FREEING state (this is the only place
845 * where transition to this state happens).
847 * Eliminates all venues through which new references to the page can be
850 * - removes page from the radix trees,
852 * - breaks linkage from VM page to cl_page.
854 * Once page reaches cl_page_state::CPS_FREEING, all remaining references will
855 * drain after some time, at which point page will be recycled.
857 * \pre VM page is locked
858 * \post pg->cp_state == CPS_FREEING
860 * \see cl_page_operations::cpo_delete()
862 void cl_page_delete(const struct lu_env *env, struct cl_page *pg)
864 PINVRNT(env, pg, cl_page_invariant(pg));
866 cl_page_delete0(env, pg);
869 EXPORT_SYMBOL(cl_page_delete);
872 * Marks page up-to-date.
874 * Call cl_page_operations::cpo_export() through all layers top-to-bottom. The
875 * layer responsible for VM interaction has to mark/clear page as up-to-date
876 * by the \a uptodate argument.
878 * \see cl_page_operations::cpo_export()
880 void cl_page_export(const struct lu_env *env, struct cl_page *cl_page,
883 const struct cl_page_slice *slice;
886 PINVRNT(env, cl_page, cl_page_invariant(cl_page));
888 cl_page_slice_for_each(cl_page, slice, i) {
889 if (slice->cpl_ops->cpo_export != NULL)
890 (*slice->cpl_ops->cpo_export)(env, slice, uptodate);
893 EXPORT_SYMBOL(cl_page_export);
896 * Returns true, if \a page is VM locked in a suitable sense by the calling
899 int cl_page_is_vmlocked(const struct lu_env *env,
900 const struct cl_page *cl_page)
902 const struct cl_page_slice *slice;
906 slice = cl_page_slice_get(cl_page, 0);
907 PASSERT(env, cl_page, slice->cpl_ops->cpo_is_vmlocked != NULL);
909 * Call ->cpo_is_vmlocked() directly instead of going through
910 * CL_PAGE_INVOKE(), because cl_page_is_vmlocked() is used by
911 * cl_page_invariant().
913 result = slice->cpl_ops->cpo_is_vmlocked(env, slice);
914 PASSERT(env, cl_page, result == -EBUSY || result == -ENODATA);
916 RETURN(result == -EBUSY);
918 EXPORT_SYMBOL(cl_page_is_vmlocked);
920 void cl_page_touch(const struct lu_env *env,
921 const struct cl_page *cl_page, size_t to)
923 const struct cl_page_slice *slice;
928 cl_page_slice_for_each(cl_page, slice, i) {
929 if (slice->cpl_ops->cpo_page_touch != NULL)
930 (*slice->cpl_ops->cpo_page_touch)(env, slice, to);
935 EXPORT_SYMBOL(cl_page_touch);
937 static enum cl_page_state cl_req_type_state(enum cl_req_type crt)
940 RETURN(crt == CRT_WRITE ? CPS_PAGEOUT : CPS_PAGEIN);
943 static void cl_page_io_start(const struct lu_env *env,
944 struct cl_page *pg, enum cl_req_type crt)
947 * Page is queued for IO, change its state.
950 cl_page_owner_clear(pg);
951 cl_page_state_set(env, pg, cl_req_type_state(crt));
956 * Prepares page for immediate transfer. cl_page_operations::cpo_prep() is
957 * called top-to-bottom. Every layer either agrees to submit this page (by
958 * returning 0), or requests to omit this page (by returning -EALREADY). Layer
959 * handling interactions with the VM also has to inform VM that page is under
962 int cl_page_prep(const struct lu_env *env, struct cl_io *io,
963 struct cl_page *cl_page, enum cl_req_type crt)
965 const struct cl_page_slice *slice;
969 PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
970 PINVRNT(env, cl_page, cl_page_invariant(cl_page));
971 PINVRNT(env, cl_page, crt < CRT_NR);
974 * this has to be called bottom-to-top, so that llite can set up
975 * PG_writeback without risking other layers deciding to skip this
981 cl_page_slice_for_each(cl_page, slice, i) {
982 if (slice->cpl_ops->cpo_own)
983 result = (*slice->cpl_ops->io[crt].cpo_prep)(env,
992 cl_page_io_start(env, cl_page, crt);
995 CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", crt, result);
998 EXPORT_SYMBOL(cl_page_prep);
1001 * Notify layers about transfer completion.
1003 * Invoked by transfer sub-system (which is a part of osc) to notify layers
1004 * that a transfer, of which this page is a part of has completed.
1006 * Completion call-backs are executed in the bottom-up order, so that
1007 * uppermost layer (llite), responsible for the VFS/VM interaction runs last
1008 * and can release locks safely.
1010 * \pre cl_page->cp_state == CPS_PAGEIN || cl_page->cp_state == CPS_PAGEOUT
1011 * \post cl_page->cl_page_state == CPS_CACHED
1013 * \see cl_page_operations::cpo_completion()
1015 void cl_page_completion(const struct lu_env *env,
1016 struct cl_page *cl_page, enum cl_req_type crt,
1019 const struct cl_page_slice *slice;
1020 struct cl_sync_io *anchor = cl_page->cp_sync_io;
1024 PASSERT(env, cl_page, crt < CRT_NR);
1025 PASSERT(env, cl_page, cl_page->cp_state == cl_req_type_state(crt));
1027 CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", crt, ioret);
1028 cl_page_state_set(env, cl_page, CPS_CACHED);
1032 cl_page_slice_for_each_reverse(cl_page, slice, i) {
1033 if (slice->cpl_ops->io[crt].cpo_completion != NULL)
1034 (*slice->cpl_ops->io[crt].cpo_completion)(env, slice,
1038 if (anchor != NULL) {
1039 LASSERT(cl_page->cp_sync_io == anchor);
1040 cl_page->cp_sync_io = NULL;
1041 cl_sync_io_note(env, anchor, ioret);
1045 EXPORT_SYMBOL(cl_page_completion);
1048 * Notify layers that transfer formation engine decided to yank this page from
1049 * the cache and to make it a part of a transfer.
1051 * \pre cl_page->cp_state == CPS_CACHED
1052 * \post cl_page->cp_state == CPS_PAGEIN || cl_page->cp_state == CPS_PAGEOUT
1054 * \see cl_page_operations::cpo_make_ready()
1056 int cl_page_make_ready(const struct lu_env *env, struct cl_page *cl_page,
1057 enum cl_req_type crt)
1059 const struct cl_page_slice *slice;
1064 PINVRNT(env, cl_page, crt < CRT_NR);
1068 cl_page_slice_for_each(cl_page, slice, i) {
1069 if (slice->cpl_ops->io[crt].cpo_make_ready != NULL)
1070 result = (*slice->cpl_ops->io[crt].cpo_make_ready)(env, slice);
1077 PASSERT(env, cl_page, cl_page->cp_state == CPS_CACHED);
1078 cl_page_io_start(env, cl_page, crt);
1080 CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", crt, result);
1084 EXPORT_SYMBOL(cl_page_make_ready);
1087 * Called if a page is being written back by kernel's intention.
1089 * \pre cl_page_is_owned(cl_page, io)
1090 * \post ergo(result == 0, cl_page->cp_state == CPS_PAGEOUT)
1092 * \see cl_page_operations::cpo_flush()
1094 int cl_page_flush(const struct lu_env *env, struct cl_io *io,
1095 struct cl_page *cl_page)
1097 const struct cl_page_slice *slice;
1102 PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
1103 PINVRNT(env, cl_page, cl_page_invariant(cl_page));
1105 cl_page_slice_for_each(cl_page, slice, i) {
1106 if (slice->cpl_ops->cpo_flush != NULL)
1107 result = (*slice->cpl_ops->cpo_flush)(env, slice, io);
1114 CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d\n", result);
1117 EXPORT_SYMBOL(cl_page_flush);
1120 * Tells transfer engine that only part of a page is to be transmitted.
1122 * \see cl_page_operations::cpo_clip()
1124 void cl_page_clip(const struct lu_env *env, struct cl_page *cl_page,
1127 const struct cl_page_slice *slice;
1130 PINVRNT(env, cl_page, cl_page_invariant(cl_page));
1132 CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", from, to);
1133 cl_page_slice_for_each(cl_page, slice, i) {
1134 if (slice->cpl_ops->cpo_clip != NULL)
1135 (*slice->cpl_ops->cpo_clip)(env, slice, from, to);
1138 EXPORT_SYMBOL(cl_page_clip);
1141 * Prints human readable representation of \a pg to the \a f.
1143 void cl_page_header_print(const struct lu_env *env, void *cookie,
1144 lu_printer_t printer, const struct cl_page *pg)
1146 (*printer)(env, cookie,
1147 "page@%p[%d %p %d %d %p]\n",
1148 pg, atomic_read(&pg->cp_ref), pg->cp_obj,
1149 pg->cp_state, pg->cp_type,
1152 EXPORT_SYMBOL(cl_page_header_print);
1155 * Prints human readable representation of \a cl_page to the \a f.
1157 void cl_page_print(const struct lu_env *env, void *cookie,
1158 lu_printer_t printer, const struct cl_page *cl_page)
1160 const struct cl_page_slice *slice;
1164 cl_page_header_print(env, cookie, printer, cl_page);
1165 cl_page_slice_for_each(cl_page, slice, i) {
1166 if (slice->cpl_ops->cpo_print != NULL)
1167 result = (*slice->cpl_ops->cpo_print)(env, slice,
1172 (*printer)(env, cookie, "end page@%p\n", cl_page);
1174 EXPORT_SYMBOL(cl_page_print);
1177 * Converts a byte offset within object \a obj into a page index.
1179 loff_t cl_offset(const struct cl_object *obj, pgoff_t idx)
1181 return (loff_t)idx << PAGE_SHIFT;
1183 EXPORT_SYMBOL(cl_offset);
1186 * Converts a page index into a byte offset within object \a obj.
1188 pgoff_t cl_index(const struct cl_object *obj, loff_t offset)
1190 return offset >> PAGE_SHIFT;
1192 EXPORT_SYMBOL(cl_index);
1194 size_t cl_page_size(const struct cl_object *obj)
1196 return 1UL << PAGE_SHIFT;
1198 EXPORT_SYMBOL(cl_page_size);
1201 * Adds page slice to the compound page.
1203 * This is called by cl_object_operations::coo_page_init() methods to add a
1204 * per-layer state to the page. New state is added at the end of
1205 * cl_page::cp_layers list, that is, it is at the bottom of the stack.
1207 * \see cl_lock_slice_add(), cl_req_slice_add(), cl_io_slice_add()
1209 void cl_page_slice_add(struct cl_page *cl_page, struct cl_page_slice *slice,
1210 struct cl_object *obj,
1211 const struct cl_page_operations *ops)
1213 unsigned int offset = (char *)slice -
1214 ((char *)cl_page + sizeof(*cl_page));
1217 LASSERT(cl_page->cp_layer_count < CP_MAX_LAYER);
1218 LASSERT(offset < (1 << sizeof(cl_page->cp_layer_offset[0]) * 8));
1219 cl_page->cp_layer_offset[cl_page->cp_layer_count++] = offset;
1220 slice->cpl_obj = obj;
1221 slice->cpl_ops = ops;
1222 slice->cpl_page = cl_page;
1226 EXPORT_SYMBOL(cl_page_slice_add);
1229 * Allocate and initialize cl_cache, called by ll_init_sbi().
1231 struct cl_client_cache *cl_cache_init(unsigned long lru_page_max)
1233 struct cl_client_cache *cache = NULL;
1236 OBD_ALLOC(cache, sizeof(*cache));
1240 /* Initialize cache data */
1241 atomic_set(&cache->ccc_users, 1);
1242 cache->ccc_lru_max = lru_page_max;
1243 atomic_long_set(&cache->ccc_lru_left, lru_page_max);
1244 spin_lock_init(&cache->ccc_lru_lock);
1245 INIT_LIST_HEAD(&cache->ccc_lru);
1247 /* turn unstable check off by default as it impacts performance */
1248 cache->ccc_unstable_check = 0;
1249 atomic_long_set(&cache->ccc_unstable_nr, 0);
1250 init_waitqueue_head(&cache->ccc_unstable_waitq);
1251 mutex_init(&cache->ccc_max_cache_mb_lock);
1255 EXPORT_SYMBOL(cl_cache_init);
1258 * Increase cl_cache refcount
1260 void cl_cache_incref(struct cl_client_cache *cache)
1262 atomic_inc(&cache->ccc_users);
1264 EXPORT_SYMBOL(cl_cache_incref);
1267 * Decrease cl_cache refcount and free the cache if refcount=0.
1268 * Since llite, lov and osc all hold cl_cache refcount,
1269 * the free will not cause race. (LU-6173)
1271 void cl_cache_decref(struct cl_client_cache *cache)
1273 if (atomic_dec_and_test(&cache->ccc_users))
1274 OBD_FREE(cache, sizeof(*cache));
1276 EXPORT_SYMBOL(cl_cache_decref);