*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2013, Intel Corporation.
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
*
* Client Lustre Page.
*
#define DEBUG_SUBSYSTEM S_CLASS
+#include <linux/list.h>
#include <libcfs/libcfs.h>
#include <obd_class.h>
#include <obd_support.h>
-#include <libcfs/list.h>
#include <cl_object.h>
#include "cl_internal.h"
-static void cl_page_delete0(const struct lu_env *env, struct cl_page *pg);
+static void __cl_page_delete(const struct lu_env *env, struct cl_page *pg);
+static DEFINE_MUTEX(cl_page_kmem_mutex);
#ifdef LIBCFS_DEBUG
# define PASSERT(env, page, expr) \
- do { \
- if (unlikely(!(expr))) { \
- CL_PAGE_DEBUG(D_ERROR, (env), (page), #expr "\n"); \
- LASSERT(0); \
- } \
- } while (0)
+do { \
+ if (unlikely(!(expr))) { \
+ CL_PAGE_DEBUG(D_ERROR, (env), (page), #expr "\n"); \
+ LASSERT(0); \
+ } \
+} while (0)
#else /* !LIBCFS_DEBUG */
-# define PASSERT(env, page, exp) \
- ((void)sizeof(env), (void)sizeof(page), (void)sizeof !!(exp))
+#define PASSERT(env, page, exp) \
+ ((void)sizeof(env), (void)sizeof(page), (void)sizeof !!(exp))
#endif /* !LIBCFS_DEBUG */
#ifdef CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK
# define PINVRNT(env, page, expr) \
- do { \
- if (unlikely(!(expr))) { \
- CL_PAGE_DEBUG(D_ERROR, (env), (page), #expr "\n"); \
- LINVRNT(0); \
- } \
- } while (0)
+do { \
+ if (unlikely(!(expr))) { \
+ CL_PAGE_DEBUG(D_ERROR, (env), (page), #expr "\n"); \
+ LINVRNT(0); \
+ } \
+} while (0)
#else /* !CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK */
# define PINVRNT(env, page, exp) \
((void)sizeof(env), (void)sizeof(page), (void)sizeof !!(exp))
#endif /* !CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK */
/* Disable page statistic by default due to huge performance penalty. */
+static void cs_page_inc(const struct cl_object *obj,
+ enum cache_stats_item item)
+{
+#ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
+ atomic_inc(&cl_object_site(obj)->cs_pages.cs_stats[item]);
+#endif
+}
+
+static void cs_page_dec(const struct cl_object *obj,
+ enum cache_stats_item item)
+{
+#ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
+ atomic_dec(&cl_object_site(obj)->cs_pages.cs_stats[item]);
+#endif
+}
+
+static void cs_pagestate_inc(const struct cl_object *obj,
+ enum cl_page_state state)
+{
+#ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
+ atomic_inc(&cl_object_site(obj)->cs_pages_state[state]);
+#endif
+}
+
+static void cs_pagestate_dec(const struct cl_object *obj,
+ enum cl_page_state state)
+{
#ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
-#define CS_PAGE_INC(o, item) \
- atomic_inc(&cl_object_site(o)->cs_pages.cs_stats[CS_##item])
-#define CS_PAGE_DEC(o, item) \
- atomic_dec(&cl_object_site(o)->cs_pages.cs_stats[CS_##item])
-#define CS_PAGESTATE_INC(o, state) \
- atomic_inc(&cl_object_site(o)->cs_pages_state[state])
-#define CS_PAGESTATE_DEC(o, state) \
- atomic_dec(&cl_object_site(o)->cs_pages_state[state])
-#else
-#define CS_PAGE_INC(o, item)
-#define CS_PAGE_DEC(o, item)
-#define CS_PAGESTATE_INC(o, state)
-#define CS_PAGESTATE_DEC(o, state)
+ atomic_dec(&cl_object_site(obj)->cs_pages_state[state]);
#endif
+}
/**
* Internal version of cl_page_get().
*/
static void cl_page_get_trust(struct cl_page *page)
{
- LASSERT(atomic_read(&page->cp_ref) > 0);
- atomic_inc(&page->cp_ref);
+ LASSERT(refcount_read(&page->cp_ref) > 0);
+ refcount_inc(&page->cp_ref);
}
-/**
- * Returns a slice within a page, corresponding to the given layer in the
- * device stack.
- *
- * \see cl_lock_at()
- */
-static const struct cl_page_slice *
-cl_page_at_trusted(const struct cl_page *page,
- const struct lu_device_type *dtype)
+static struct cl_page_slice *
+cl_page_slice_get(const struct cl_page *cl_page, int index)
{
- const struct cl_page_slice *slice;
- ENTRY;
+ if (index < 0 || index >= cl_page->cp_layer_count)
+ return NULL;
- list_for_each_entry(slice, &page->cp_layers, cpl_linkage) {
- if (slice->cpl_obj->co_lu.lo_dev->ld_type == dtype)
- RETURN(slice);
- }
- RETURN(NULL);
+ /* To get the cp_layer_offset values fit under 256 bytes, we
+ * use the offset beyond the end of struct cl_page.
+ */
+ return (struct cl_page_slice *)((char *)cl_page + sizeof(*cl_page) +
+ cl_page->cp_layer_offset[index]);
}
-static void cl_page_free(const struct lu_env *env, struct cl_page *page)
+#define cl_page_slice_for_each(cl_page, slice, i) \
+ for (i = 0, slice = cl_page_slice_get(cl_page, 0); \
+ i < (cl_page)->cp_layer_count; \
+ slice = cl_page_slice_get(cl_page, ++i))
+
+#define cl_page_slice_for_each_reverse(cl_page, slice, i) \
+ for (i = (cl_page)->cp_layer_count - 1, \
+ slice = cl_page_slice_get(cl_page, i); i >= 0; \
+ slice = cl_page_slice_get(cl_page, --i))
+
+static void __cl_page_free(struct cl_page *cl_page, unsigned short bufsize)
{
- struct cl_object *obj = page->cp_obj;
- int pagesize = cl_object_header(obj)->coh_page_bufsize;
+ int index = cl_page->cp_kmem_index;
+
+ if (index >= 0) {
+ LASSERT(index < ARRAY_SIZE(cl_page_kmem_array));
+ LASSERT(cl_page_kmem_size_array[index] == bufsize);
+ OBD_SLAB_FREE(cl_page, cl_page_kmem_array[index], bufsize);
+ } else {
+ OBD_FREE(cl_page, bufsize);
+ }
+}
- PASSERT(env, page, list_empty(&page->cp_batch));
- PASSERT(env, page, page->cp_owner == NULL);
- PASSERT(env, page, page->cp_req == NULL);
- PASSERT(env, page, page->cp_state == CPS_FREEING);
+static void cl_page_free(const struct lu_env *env, struct cl_page *cp,
+ struct folio_batch *fbatch)
+{
+ struct cl_object *obj = cp->cp_obj;
+ unsigned short bufsize = cl_object_header(obj)->coh_page_bufsize;
+ struct page *vmpage;
ENTRY;
- while (!list_empty(&page->cp_layers)) {
- struct cl_page_slice *slice;
-
- slice = list_entry(page->cp_layers.next,
- struct cl_page_slice, cpl_linkage);
- list_del_init(page->cp_layers.next);
- if (unlikely(slice->cpl_ops->cpo_fini != NULL))
- slice->cpl_ops->cpo_fini(env, slice);
+ PASSERT(env, cp, list_empty(&cp->cp_batch));
+ PASSERT(env, cp, cp->cp_owner == NULL);
+ if (cp->cp_type != CPT_TRANSIENT)
+ PASSERT(env, cp, cp->cp_state == CPS_FREEING);
+
+ if (cp->cp_type == CPT_CACHEABLE) {
+ /* vmpage->private was already cleared when page was
+ * moved into CPS_FREEING state.
+ */
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT((struct cl_page *)vmpage->private != cp);
+
+ if (fbatch != NULL) {
+ if (!folio_batch_add_page(fbatch, vmpage))
+ folio_batch_release(fbatch);
+ } else {
+ put_page(vmpage);
+ }
}
- CS_PAGE_DEC(obj, total);
- CS_PAGESTATE_DEC(obj, page->cp_state);
- lu_object_ref_del_at(&obj->co_lu, &page->cp_obj_ref, "cl_page", page);
- cl_object_put(env, obj);
- lu_ref_fini(&page->cp_reference);
- OBD_FREE(page, pagesize);
+
+ cp->cp_layer_count = 0;
+ cs_page_dec(obj, CS_total);
+ if (cp->cp_type != CPT_TRANSIENT)
+ cs_pagestate_dec(obj, cp->cp_state);
+ lu_object_ref_del_at(&obj->co_lu, &cp->cp_obj_ref, "cl_page", cp);
+ if (cp->cp_type != CPT_TRANSIENT)
+ cl_object_put(env, obj);
+ lu_ref_fini(&cp->cp_reference);
+ __cl_page_free(cp, bufsize);
EXIT;
}
-/**
- * Helper function updating page state. This is the only place in the code
- * where cl_page::cp_state field is mutated.
- */
-static inline void cl_page_state_set_trust(struct cl_page *page,
- enum cl_page_state state)
+static struct cl_page *__cl_page_alloc(struct cl_object *o)
{
- /* bypass const. */
- *(enum cl_page_state *)&page->cp_state = state;
+ int i = 0;
+ struct cl_page *cl_page = NULL;
+ unsigned short bufsize = cl_object_header(o)->coh_page_bufsize;
+
+ if (CFS_FAIL_CHECK(OBD_FAIL_LLITE_PAGE_ALLOC))
+ return NULL;
+
+check:
+ /* the number of entries in cl_page_kmem_array is expected to
+ * only be 2-3 entries, so the lookup overhead should be low.
+ */
+ for ( ; i < ARRAY_SIZE(cl_page_kmem_array); i++) {
+ if (smp_load_acquire(&cl_page_kmem_size_array[i]) == bufsize) {
+ OBD_SLAB_ALLOC_GFP(cl_page, cl_page_kmem_array[i],
+ bufsize, GFP_NOFS);
+ if (cl_page)
+ cl_page->cp_kmem_index = i;
+ return cl_page;
+ }
+ if (cl_page_kmem_size_array[i] == 0)
+ break;
+ }
+
+ if (i < ARRAY_SIZE(cl_page_kmem_array)) {
+ char cache_name[32];
+
+ mutex_lock(&cl_page_kmem_mutex);
+ if (cl_page_kmem_size_array[i]) {
+ mutex_unlock(&cl_page_kmem_mutex);
+ goto check;
+ }
+ snprintf(cache_name, sizeof(cache_name),
+ "cl_page_kmem-%u", bufsize);
+ cl_page_kmem_array[i] =
+ kmem_cache_create(cache_name, bufsize,
+ 0, 0, NULL);
+ if (cl_page_kmem_array[i] == NULL) {
+ mutex_unlock(&cl_page_kmem_mutex);
+ return NULL;
+ }
+ smp_store_release(&cl_page_kmem_size_array[i], bufsize);
+ mutex_unlock(&cl_page_kmem_mutex);
+ goto check;
+ } else {
+ OBD_ALLOC_GFP(cl_page, bufsize, GFP_NOFS);
+ if (cl_page)
+ cl_page->cp_kmem_index = -1;
+ }
+
+ return cl_page;
}
-struct cl_page *cl_page_alloc(const struct lu_env *env,
- struct cl_object *o, pgoff_t ind, struct page *vmpage,
- enum cl_page_type type)
+struct cl_page *cl_page_alloc(const struct lu_env *env, struct cl_object *o,
+ pgoff_t ind, struct page *vmpage,
+ enum cl_page_type type)
{
- struct cl_page *page;
- struct lu_object_header *head;
+ struct cl_page *cl_page;
+ struct cl_object *head;
ENTRY;
- OBD_ALLOC_GFP(page, cl_object_header(o)->coh_page_bufsize,
- GFP_NOFS);
- if (page != NULL) {
+
+ cl_page = __cl_page_alloc(o);
+ if (cl_page != NULL) {
int result = 0;
- atomic_set(&page->cp_ref, 1);
- page->cp_obj = o;
- cl_object_get(o);
- lu_object_ref_add_at(&o->co_lu, &page->cp_obj_ref, "cl_page",
- page);
- page->cp_vmpage = vmpage;
- cl_page_state_set_trust(page, CPS_CACHED);
- page->cp_type = type;
- INIT_LIST_HEAD(&page->cp_layers);
- INIT_LIST_HEAD(&page->cp_batch);
- INIT_LIST_HEAD(&page->cp_flight);
- lu_ref_init(&page->cp_reference);
- head = o->co_lu.lo_header;
- list_for_each_entry(o, &head->loh_layers,
- co_lu.lo_linkage) {
+
+ /* Please fix cl_page:cp_state/type declaration if
+ * these assertions fail in the future.
+ */
+ BUILD_BUG_ON((1 << CP_STATE_BITS) < CPS_NR); /* cp_state */
+ BUILD_BUG_ON((1 << CP_TYPE_BITS) < CPT_NR); /* cp_type */
+ refcount_set(&cl_page->cp_ref, 1);
+ cl_page->cp_obj = o;
+ if (type != CPT_TRANSIENT)
+ cl_object_get(o);
+ lu_object_ref_add_at(&o->co_lu, &cl_page->cp_obj_ref,
+ "cl_page", cl_page);
+ cl_page->cp_vmpage = vmpage;
+ if (cl_page->cp_type != CPT_TRANSIENT)
+ cl_page->cp_state = CPS_CACHED;
+ cl_page->cp_type = type;
+ if (type == CPT_TRANSIENT)
+ /* correct inode to be added in ll_direct_rw_pages */
+ cl_page->cp_inode = NULL;
+ else
+ cl_page->cp_inode = page2inode(vmpage);
+ INIT_LIST_HEAD(&cl_page->cp_batch);
+ lu_ref_init(&cl_page->cp_reference);
+ head = o;
+ cl_page->cp_page_index = ind;
+ cl_object_for_each(o, head) {
if (o->co_ops->coo_page_init != NULL) {
- result = o->co_ops->coo_page_init(env, o, page,
- ind);
+ result = o->co_ops->coo_page_init(env, o,
+ cl_page, ind);
if (result != 0) {
- cl_page_delete0(env, page);
- cl_page_free(env, page);
- page = ERR_PTR(result);
+ __cl_page_delete(env, cl_page);
+ cl_page_free(env, cl_page, NULL);
+ cl_page = ERR_PTR(result);
break;
}
}
}
if (result == 0) {
- CS_PAGE_INC(o, total);
- CS_PAGE_INC(o, create);
- CS_PAGESTATE_DEC(o, CPS_CACHED);
+ cs_page_inc(o, CS_total);
+ cs_page_inc(o, CS_create);
+ cs_pagestate_dec(o, CPS_CACHED);
}
} else {
- page = ERR_PTR(-ENOMEM);
+ cl_page = ERR_PTR(-ENOMEM);
}
- RETURN(page);
+ RETURN(cl_page);
}
-EXPORT_SYMBOL(cl_page_alloc);
/**
* Returns a cl_page with index \a idx at the object \a o, and associated with
ENTRY;
hdr = cl_object_header(o);
- CS_PAGE_INC(o, lookup);
-
- CDEBUG(D_PAGE, "%lu@"DFID" %p %lx %d\n",
- idx, PFID(&hdr->coh_lu.loh_fid), vmpage, vmpage->private, type);
- /* fast path. */
- if (type == CPT_CACHEABLE) {
- /* vmpage lock is used to protect the child/parent
- * relationship */
- KLASSERT(PageLocked(vmpage));
- /*
- * cl_vmpage_page() can be called here without any locks as
- *
- * - "vmpage" is locked (which prevents ->private from
- * concurrent updates), and
- *
- * - "o" cannot be destroyed while current thread holds a
- * reference on it.
- */
- page = cl_vmpage_page(vmpage, o);
+ cs_page_inc(o, CS_lookup);
+
+ CDEBUG(D_PAGE, "%lu@"DFID" %p %lx %d\n",
+ idx, PFID(&hdr->coh_lu.loh_fid), vmpage, vmpage->private, type);
+ /* fast path. */
+ if (type == CPT_CACHEABLE) {
+ /* vmpage lock used to protect the child/parent relationship */
+ LASSERT(PageLocked(vmpage));
+ /*
+ * cl_vmpage_page() can be called here without any locks as
+ *
+ * - "vmpage" is locked (which prevents ->private from
+ * concurrent updates), and
+ *
+ * - "o" cannot be destroyed while current thread holds a
+ * reference on it.
+ */
+ page = cl_vmpage_page(vmpage, o);
if (page != NULL) {
- CS_PAGE_INC(o, hit);
+ cs_page_inc(o, CS_hit);
RETURN(page);
}
- }
+ }
- /* allocate and initialize cl_page */
- page = cl_page_alloc(env, o, idx, vmpage, type);
+ /* allocate and initialize cl_page */
+ page = cl_page_alloc(env, o, idx, vmpage, type);
RETURN(page);
}
EXPORT_SYMBOL(cl_page_find);
return cl_page_in_use_noref(pg);
}
-static void cl_page_state_set0(const struct lu_env *env,
- struct cl_page *page, enum cl_page_state state)
+static void __cl_page_state_set(const struct lu_env *env,
+ struct cl_page *cl_page,
+ enum cl_page_state state)
{
- enum cl_page_state old;
-
- /*
- * Matrix of allowed state transitions [old][new], for sanity
- * checking.
- */
- static const int allowed_transitions[CPS_NR][CPS_NR] = {
- [CPS_CACHED] = {
- [CPS_CACHED] = 0,
- [CPS_OWNED] = 1, /* io finds existing cached page */
- [CPS_PAGEIN] = 0,
- [CPS_PAGEOUT] = 1, /* write-out from the cache */
- [CPS_FREEING] = 1, /* eviction on the memory pressure */
- },
- [CPS_OWNED] = {
- [CPS_CACHED] = 1, /* release to the cache */
- [CPS_OWNED] = 0,
- [CPS_PAGEIN] = 1, /* start read immediately */
- [CPS_PAGEOUT] = 1, /* start write immediately */
- [CPS_FREEING] = 1, /* lock invalidation or truncate */
- },
- [CPS_PAGEIN] = {
- [CPS_CACHED] = 1, /* io completion */
- [CPS_OWNED] = 0,
- [CPS_PAGEIN] = 0,
- [CPS_PAGEOUT] = 0,
- [CPS_FREEING] = 0,
- },
- [CPS_PAGEOUT] = {
- [CPS_CACHED] = 1, /* io completion */
- [CPS_OWNED] = 0,
- [CPS_PAGEIN] = 0,
- [CPS_PAGEOUT] = 0,
- [CPS_FREEING] = 0,
- },
- [CPS_FREEING] = {
- [CPS_CACHED] = 0,
- [CPS_OWNED] = 0,
- [CPS_PAGEIN] = 0,
- [CPS_PAGEOUT] = 0,
- [CPS_FREEING] = 0,
- }
- };
-
- ENTRY;
- old = page->cp_state;
- PASSERT(env, page, allowed_transitions[old][state]);
- CL_PAGE_HEADER(D_TRACE, env, page, "%d -> %d\n", old, state);
- PASSERT(env, page, page->cp_state == old);
- PASSERT(env, page, equi(state == CPS_OWNED, page->cp_owner != NULL));
-
- CS_PAGESTATE_DEC(page->cp_obj, page->cp_state);
- CS_PAGESTATE_INC(page->cp_obj, state);
- cl_page_state_set_trust(page, state);
+ enum cl_page_state old;
+
+ /* Matrix of allowed state transitions [old][new] for sanity checking */
+ static const int allowed_transitions[CPS_NR][CPS_NR] = {
+ [CPS_CACHED] = {
+ [CPS_CACHED] = 0,
+ [CPS_OWNED] = 1, /* io finds existing cached page */
+ [CPS_PAGEIN] = 0,
+ [CPS_PAGEOUT] = 1, /* write-out from the cache */
+ [CPS_FREEING] = 1, /* eviction on the memory pressure */
+ },
+ [CPS_OWNED] = {
+ [CPS_CACHED] = 1, /* release to the cache */
+ [CPS_OWNED] = 0,
+ [CPS_PAGEIN] = 1, /* start read immediately */
+ [CPS_PAGEOUT] = 1, /* start write immediately */
+ [CPS_FREEING] = 1, /* lock invalidation or truncate */
+ },
+ [CPS_PAGEIN] = {
+ [CPS_CACHED] = 1, /* io completion */
+ [CPS_OWNED] = 0,
+ [CPS_PAGEIN] = 0,
+ [CPS_PAGEOUT] = 0,
+ [CPS_FREEING] = 0,
+ },
+ [CPS_PAGEOUT] = {
+ [CPS_CACHED] = 1, /* io completion */
+ [CPS_OWNED] = 0,
+ [CPS_PAGEIN] = 0,
+ [CPS_PAGEOUT] = 0,
+ [CPS_FREEING] = 0,
+ },
+ [CPS_FREEING] = {
+ [CPS_CACHED] = 0,
+ [CPS_OWNED] = 0,
+ [CPS_PAGEIN] = 0,
+ [CPS_PAGEOUT] = 0,
+ [CPS_FREEING] = 0,
+ }
+ };
+
+ ENTRY;
+ old = cl_page->cp_state;
+ PASSERT(env, cl_page, allowed_transitions[old][state]);
+ CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d -> %d\n", old, state);
+ PASSERT(env, cl_page, cl_page->cp_state == old);
+ PASSERT(env, cl_page, equi(state == CPS_OWNED,
+ cl_page->cp_owner != NULL));
+
+ cs_pagestate_dec(cl_page->cp_obj, cl_page->cp_state);
+ cs_pagestate_inc(cl_page->cp_obj, state);
+ cl_page->cp_state = state;
EXIT;
}
static void cl_page_state_set(const struct lu_env *env,
- struct cl_page *page, enum cl_page_state state)
+ struct cl_page *page, enum cl_page_state state)
{
- cl_page_state_set0(env, page, state);
+ LASSERT(page->cp_type != CPT_TRANSIENT);
+ __cl_page_state_set(env, page, state);
}
/**
EXPORT_SYMBOL(cl_page_get);
/**
- * Releases a reference to a page.
- *
- * When last reference is released, page is returned to the cache, unless it
- * is in cl_page_state::CPS_FREEING state, in which case it is immediately
- * destroyed.
+ * Releases a reference to a page, use the folio_batch to release the pages
+ * in batch if provided.
*
- * \see cl_object_put(), cl_lock_put().
+ * Users need to do a final folio_batch_release() to release any trailing pages.
*/
-void cl_page_put(const struct lu_env *env, struct cl_page *page)
+void cl_batch_put(const struct lu_env *env, struct cl_page *page,
+ struct folio_batch *fbatch)
{
- ENTRY;
- CL_PAGE_HEADER(D_TRACE, env, page, "%d\n",
- atomic_read(&page->cp_ref));
+ ENTRY;
+ CL_PAGE_HEADER(D_TRACE, env, page, "%d\n",
+ refcount_read(&page->cp_ref));
- if (atomic_dec_and_test(&page->cp_ref)) {
- LASSERT(page->cp_state == CPS_FREEING);
+ if (refcount_dec_and_test(&page->cp_ref)) {
+ if (page->cp_type != CPT_TRANSIENT)
+ LASSERT(page->cp_state == CPS_FREEING);
- LASSERT(atomic_read(&page->cp_ref) == 0);
+ LASSERT(refcount_read(&page->cp_ref) == 0);
PASSERT(env, page, page->cp_owner == NULL);
PASSERT(env, page, list_empty(&page->cp_batch));
- /*
- * Page is no longer reachable by other threads. Tear
- * it down.
- */
- cl_page_free(env, page);
+ /* Page is no longer reachable by other threads. Tear it down */
+ cl_page_free(env, page, fbatch);
}
EXIT;
}
-EXPORT_SYMBOL(cl_page_put);
+EXPORT_SYMBOL(cl_batch_put);
/**
- * Returns a cl_page associated with a VM page, and given cl_object.
+ * Releases a reference to a page, wrapper to cl_batch_put
+ *
+ * When last reference is released, page is returned to the cache, unless it
+ * is in cl_page_state::CPS_FREEING state, in which case it is immediately
+ * destroyed.
+ *
+ * \see cl_object_put(), cl_lock_put().
*/
+void cl_page_put(const struct lu_env *env, struct cl_page *page)
+{
+ cl_batch_put(env, page, NULL);
+}
+EXPORT_SYMBOL(cl_page_put);
+
+/* Returns a cl_page associated with a VM page, and given cl_object. */
struct cl_page *cl_vmpage_page(struct page *vmpage, struct cl_object *obj)
{
struct cl_page *page;
ENTRY;
- KLASSERT(PageLocked(vmpage));
+ LASSERT(PageLocked(vmpage));
/*
* NOTE: absence of races and liveness of data are guaranteed by page
}
EXPORT_SYMBOL(cl_vmpage_page);
-const struct cl_page_slice *cl_page_at(const struct cl_page *page,
- const struct lu_device_type *dtype)
-{
- return cl_page_at_trusted(page, dtype);
-}
-EXPORT_SYMBOL(cl_page_at);
-
-#define CL_PAGE_OP(opname) offsetof(struct cl_page_operations, opname)
-
-#define CL_PAGE_INVOKE(_env, _page, _op, _proto, ...) \
-({ \
- const struct lu_env *__env = (_env); \
- struct cl_page *__page = (_page); \
- const struct cl_page_slice *__scan; \
- int __result; \
- ptrdiff_t __op = (_op); \
- int (*__method)_proto; \
- \
- __result = 0; \
- list_for_each_entry(__scan, &__page->cp_layers, cpl_linkage) { \
- __method = *(void **)((char *)__scan->cpl_ops + __op); \
- if (__method != NULL) { \
- __result = (*__method)(__env, __scan, ## __VA_ARGS__); \
- if (__result != 0) \
- break; \
- } \
- } \
- if (__result > 0) \
- __result = 0; \
- __result; \
-})
-
-#define CL_PAGE_INVOKE_REVERSE(_env, _page, _op, _proto, ...) \
-({ \
- const struct lu_env *__env = (_env); \
- struct cl_page *__page = (_page); \
- const struct cl_page_slice *__scan; \
- int __result; \
- ptrdiff_t __op = (_op); \
- int (*__method)_proto; \
- \
- __result = 0; \
- list_for_each_entry_reverse(__scan, &__page->cp_layers, \
- cpl_linkage) { \
- __method = *(void **)((char *)__scan->cpl_ops + __op); \
- if (__method != NULL) { \
- __result = (*__method)(__env, __scan, ## __VA_ARGS__); \
- if (__result != 0) \
- break; \
- } \
- } \
- if (__result > 0) \
- __result = 0; \
- __result; \
-})
-
-#define CL_PAGE_INVOID(_env, _page, _op, _proto, ...) \
-do { \
- const struct lu_env *__env = (_env); \
- struct cl_page *__page = (_page); \
- const struct cl_page_slice *__scan; \
- ptrdiff_t __op = (_op); \
- void (*__method)_proto; \
- \
- list_for_each_entry(__scan, &__page->cp_layers, cpl_linkage) { \
- __method = *(void **)((char *)__scan->cpl_ops + __op); \
- if (__method != NULL) \
- (*__method)(__env, __scan, ## __VA_ARGS__); \
- } \
-} while (0)
-
-#define CL_PAGE_INVOID_REVERSE(_env, _page, _op, _proto, ...) \
-do { \
- const struct lu_env *__env = (_env); \
- struct cl_page *__page = (_page); \
- const struct cl_page_slice *__scan; \
- ptrdiff_t __op = (_op); \
- void (*__method)_proto; \
- \
- /* get to the bottom page. */ \
- list_for_each_entry_reverse(__scan, &__page->cp_layers, \
- cpl_linkage) { \
- __method = *(void **)((char *)__scan->cpl_ops + __op); \
- if (__method != NULL) \
- (*__method)(__env, __scan, ## __VA_ARGS__); \
- } \
-} while (0)
-
-static int cl_page_invoke(const struct lu_env *env,
- struct cl_io *io, struct cl_page *page, ptrdiff_t op)
-
-{
- PINVRNT(env, page, cl_object_same(page->cp_obj, io->ci_obj));
- ENTRY;
- RETURN(CL_PAGE_INVOKE(env, page, op,
- (const struct lu_env *,
- const struct cl_page_slice *, struct cl_io *),
- io));
-}
-
-static void cl_page_invoid(const struct lu_env *env,
- struct cl_io *io, struct cl_page *page, ptrdiff_t op)
-
-{
- PINVRNT(env, page, cl_object_same(page->cp_obj, io->ci_obj));
- ENTRY;
- CL_PAGE_INVOID(env, page, op,
- (const struct lu_env *,
- const struct cl_page_slice *, struct cl_io *), io);
- EXIT;
-}
-
static void cl_page_owner_clear(struct cl_page *page)
{
ENTRY;
EXIT;
}
-void cl_page_disown0(const struct lu_env *env,
- struct cl_io *io, struct cl_page *pg)
+void __cl_page_disown(const struct lu_env *env, struct cl_page *cp)
{
- enum cl_page_state state;
-
- ENTRY;
- state = pg->cp_state;
- PINVRNT(env, pg, state == CPS_OWNED || state == CPS_FREEING);
- PINVRNT(env, pg, cl_page_invariant(pg) || state == CPS_FREEING);
- cl_page_owner_clear(pg);
-
- if (state == CPS_OWNED)
- cl_page_state_set(env, pg, CPS_CACHED);
- /*
- * Completion call-backs are executed in the bottom-up order, so that
- * uppermost layer (llite), responsible for VFS/VM interaction runs
- * last and can release locks safely.
- */
- CL_PAGE_INVOID_REVERSE(env, pg, CL_PAGE_OP(cpo_disown),
- (const struct lu_env *,
- const struct cl_page_slice *, struct cl_io *),
- io);
- EXIT;
+ struct page *vmpage;
+ enum cl_page_state state;
+
+ ENTRY;
+ cl_page_owner_clear(cp);
+
+ if (cp->cp_type == CPT_CACHEABLE) {
+ state = cp->cp_state;
+ PINVRNT(env, cp, state == CPS_OWNED || state == CPS_FREEING);
+ PINVRNT(env, cp, cl_page_invariant(cp) || state == CPS_FREEING);
+ if (state == CPS_OWNED)
+ cl_page_state_set(env, cp, CPS_CACHED);
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT(PageLocked(vmpage));
+ unlock_page(vmpage);
+ }
+
+ EXIT;
}
-/**
- * returns true, iff page is owned by the given io.
- */
+/* returns true, iff page is owned by the given io. */
int cl_page_is_owned(const struct cl_page *pg, const struct cl_io *io)
{
struct cl_io *top = cl_io_top((struct cl_io *)io);
- LINVRNT(cl_object_same(pg->cp_obj, io->ci_obj));
+
+ LINVRNT(cl_object_same(pg->cp_obj, top->ci_obj));
ENTRY;
- RETURN(pg->cp_state == CPS_OWNED && pg->cp_owner == top);
+ if (pg->cp_type != CPT_TRANSIENT)
+ RETURN(pg->cp_state == CPS_OWNED && pg->cp_owner == top);
+ else
+ RETURN(pg->cp_owner == top);
}
EXPORT_SYMBOL(cl_page_is_owned);
* Waits until page is in cl_page_state::CPS_CACHED state, and then switch it
* into cl_page_state::CPS_OWNED state.
*
- * \pre !cl_page_is_owned(pg, io)
- * \post result == 0 iff cl_page_is_owned(pg, io)
+ * \pre !cl_page_is_owned(cl_page, io)
+ * \post result == 0 iff cl_page_is_owned(cl_page, io)
*
* \retval 0 success
*
- * \retval -ve failure, e.g., page was destroyed (and landed in
+ * \retval -ve failure, e.g., cl_page was destroyed (and landed in
* cl_page_state::CPS_FREEING instead of cl_page_state::CPS_CACHED).
* or, page was owned by another thread, or in IO.
*
* \see cl_page_disown()
- * \see cl_page_operations::cpo_own()
* \see cl_page_own_try()
* \see cl_page_own
*/
-static int cl_page_own0(const struct lu_env *env, struct cl_io *io,
- struct cl_page *pg, int nonblock)
+static int __cl_page_own(const struct lu_env *env, struct cl_io *io,
+ struct cl_page *cl_page, int nonblock)
{
- int result;
-
- PINVRNT(env, pg, !cl_page_is_owned(pg, io));
-
- ENTRY;
- io = cl_io_top(io);
-
- if (pg->cp_state == CPS_FREEING) {
- result = -ENOENT;
- } else {
- result = CL_PAGE_INVOKE(env, pg, CL_PAGE_OP(cpo_own),
- (const struct lu_env *,
- const struct cl_page_slice *,
- struct cl_io *, int),
- io, nonblock);
- if (result == 0) {
- PASSERT(env, pg, pg->cp_owner == NULL);
- PASSERT(env, pg, pg->cp_req == NULL);
- pg->cp_owner = cl_io_top(io);;
- cl_page_owner_set(pg);
- if (pg->cp_state != CPS_FREEING) {
- cl_page_state_set(env, pg, CPS_OWNED);
- } else {
- cl_page_disown0(env, io, pg);
- result = -ENOENT;
- }
- }
- }
- PINVRNT(env, pg, ergo(result == 0, cl_page_invariant(pg)));
- RETURN(result);
+ struct page *vmpage = cl_page->cp_vmpage;
+ int result;
+
+ ENTRY;
+ PINVRNT(env, cl_page, !cl_page_is_owned(cl_page, io));
+
+ if (cl_page->cp_type != CPT_TRANSIENT &&
+ cl_page->cp_state == CPS_FREEING) {
+ result = -ENOENT;
+ goto out;
+ }
+
+ LASSERT(vmpage != NULL);
+
+ if (cl_page->cp_type == CPT_TRANSIENT) {
+ /* OK */
+ } else if (nonblock) {
+ if (!trylock_page(vmpage)) {
+ result = -EAGAIN;
+ goto out;
+ }
+
+ if (unlikely(PageWriteback(vmpage))) {
+ unlock_page(vmpage);
+ result = -EAGAIN;
+ goto out;
+ }
+ } else {
+ lock_page(vmpage);
+ wait_on_page_writeback(vmpage);
+ }
+
+ PASSERT(env, cl_page, cl_page->cp_owner == NULL);
+ cl_page->cp_owner = cl_io_top(io);
+ cl_page_owner_set(cl_page);
+
+ if (cl_page->cp_type != CPT_TRANSIENT) {
+ if (cl_page->cp_state == CPS_FREEING) {
+ __cl_page_disown(env, cl_page);
+ result = -ENOENT;
+ goto out;
+ }
+
+ cl_page_state_set(env, cl_page, CPS_OWNED);
+ }
+
+ result = 0;
+out:
+ CDEBUG(D_INFO, "res %d\n", result);
+ PINVRNT(env, cl_page, ergo(result == 0,
+ cl_page_invariant(cl_page)));
+ RETURN(result);
}
-/**
- * Own a page, might be blocked.
- *
- * \see cl_page_own0()
- */
+/* Own a page, might be blocked. (see __cl_page_own()) */
int cl_page_own(const struct lu_env *env, struct cl_io *io, struct cl_page *pg)
{
- return cl_page_own0(env, io, pg, 0);
+ return __cl_page_own(env, io, pg, 0);
}
EXPORT_SYMBOL(cl_page_own);
-/**
- * Nonblock version of cl_page_own().
- *
- * \see cl_page_own0()
- */
+/* Nonblock version of cl_page_own(). (see __cl_page_own()) */
int cl_page_own_try(const struct lu_env *env, struct cl_io *io,
- struct cl_page *pg)
+ struct cl_page *pg)
{
- return cl_page_own0(env, io, pg, 1);
+ return __cl_page_own(env, io, pg, 1);
}
EXPORT_SYMBOL(cl_page_own_try);
*
* Called when page is already locked by the hosting VM.
*
- * \pre !cl_page_is_owned(pg, io)
- * \post cl_page_is_owned(pg, io)
- *
- * \see cl_page_operations::cpo_assume()
+ * \pre !cl_page_is_owned(cp, io)
+ * \post cl_page_is_owned(cp, io)
*/
void cl_page_assume(const struct lu_env *env,
- struct cl_io *io, struct cl_page *pg)
+ struct cl_io *io, struct cl_page *cp)
{
- PINVRNT(env, pg, cl_object_same(pg->cp_obj, io->ci_obj));
+ struct page *vmpage;
ENTRY;
- io = cl_io_top(io);
+ PINVRNT(env, cp, cl_object_same(cp->cp_obj, cl_io_top(io)->ci_obj));
+
+ if (cp->cp_type == CPT_CACHEABLE) {
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT(PageLocked(vmpage));
+ wait_on_page_writeback(vmpage);
+ }
- cl_page_invoid(env, io, pg, CL_PAGE_OP(cpo_assume));
- PASSERT(env, pg, pg->cp_owner == NULL);
- pg->cp_owner = cl_io_top(io);
- cl_page_owner_set(pg);
- cl_page_state_set(env, pg, CPS_OWNED);
+ PASSERT(env, cp, cp->cp_owner == NULL);
+ cp->cp_owner = cl_io_top(io);
+ cl_page_owner_set(cp);
+ if (cp->cp_type != CPT_TRANSIENT)
+ cl_page_state_set(env, cp, CPS_OWNED);
EXIT;
}
EXPORT_SYMBOL(cl_page_assume);
/**
* Releases page ownership without unlocking the page.
*
- * Moves page into cl_page_state::CPS_CACHED without releasing a lock on the
- * underlying VM page (as VM is supposed to do this itself).
- *
- * \pre cl_page_is_owned(pg, io)
- * \post !cl_page_is_owned(pg, io)
+ * Moves cl_page into cl_page_state::CPS_CACHED without releasing a lock
+ * on the underlying VM page (as VM is supposed to do this itself).
*
- * \see cl_page_assume()
+ * \pre cl_page_is_owned(cp, io)
+ * \post !cl_page_is_owned(cp, io)
*/
void cl_page_unassume(const struct lu_env *env,
- struct cl_io *io, struct cl_page *pg)
+ struct cl_io *io, struct cl_page *cp)
{
- PINVRNT(env, pg, cl_page_is_owned(pg, io));
- PINVRNT(env, pg, cl_page_invariant(pg));
-
- ENTRY;
- io = cl_io_top(io);
- cl_page_owner_clear(pg);
- cl_page_state_set(env, pg, CPS_CACHED);
- CL_PAGE_INVOID_REVERSE(env, pg, CL_PAGE_OP(cpo_unassume),
- (const struct lu_env *,
- const struct cl_page_slice *, struct cl_io *),
- io);
- EXIT;
+ struct page *vmpage;
+
+ ENTRY;
+ PINVRNT(env, cp, cl_page_is_owned(cp, io));
+ PINVRNT(env, cp, cl_page_invariant(cp));
+
+ cl_page_owner_clear(cp);
+ if (cp->cp_type != CPT_TRANSIENT)
+ cl_page_state_set(env, cp, CPS_CACHED);
+
+ if (cp->cp_type == CPT_CACHEABLE) {
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT(PageLocked(vmpage));
+ }
+
+ EXIT;
}
EXPORT_SYMBOL(cl_page_unassume);
* \post !cl_page_is_owned(pg, io)
*
* \see cl_page_own()
- * \see cl_page_operations::cpo_disown()
*/
void cl_page_disown(const struct lu_env *env,
- struct cl_io *io, struct cl_page *pg)
+ struct cl_io *io, struct cl_page *pg)
{
- PINVRNT(env, pg, cl_page_is_owned(pg, io) ||
- pg->cp_state == CPS_FREEING);
+ if (pg->cp_type != CPT_TRANSIENT) {
+ PINVRNT(env, pg, cl_page_is_owned(pg, cl_io_top(io)) ||
+ pg->cp_state == CPS_FREEING);
+ }
- ENTRY;
- io = cl_io_top(io);
- cl_page_disown0(env, io, pg);
- EXIT;
+ __cl_page_disown(env, pg);
}
EXPORT_SYMBOL(cl_page_disown);
/**
- * Called when page is to be removed from the object, e.g., as a result of
- * truncate.
+ * Called when cl_page is to be removed from the object, e.g.,
+ * as a result of truncate.
*
* Calls cl_page_operations::cpo_discard() top-to-bottom.
*
- * \pre cl_page_is_owned(pg, io)
+ * \pre cl_page_is_owned(cl_page, io)
*
* \see cl_page_operations::cpo_discard()
*/
void cl_page_discard(const struct lu_env *env,
- struct cl_io *io, struct cl_page *pg)
+ struct cl_io *io, struct cl_page *cp)
{
- PINVRNT(env, pg, cl_page_is_owned(pg, io));
- PINVRNT(env, pg, cl_page_invariant(pg));
+ struct page *vmpage;
+ const struct cl_page_slice *slice;
+ int i;
- cl_page_invoid(env, io, pg, CL_PAGE_OP(cpo_discard));
+ PINVRNT(env, cp, cl_page_is_owned(cp, io));
+ PINVRNT(env, cp, cl_page_invariant(cp));
+
+ cl_page_slice_for_each(cp, slice, i) {
+ if (slice->cpl_ops->cpo_discard != NULL)
+ (*slice->cpl_ops->cpo_discard)(env, slice, io);
+ }
+
+ if (cp->cp_type == CPT_CACHEABLE) {
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT(PageLocked(vmpage));
+ generic_error_remove_folio(vmpage->mapping, page_folio(vmpage));
+ } else {
+ cl_page_delete(env, cp);
+ }
}
EXPORT_SYMBOL(cl_page_discard);
/**
* Version of cl_page_delete() that can be called for not fully constructed
- * pages, e.g,. in a error handling cl_page_find()->cl_page_delete0()
- * path. Doesn't check page invariant.
+ * cl_pages, e.g. in an error handling cl_page_find()->__cl_page_delete()
+ * path. Doesn't check cl_page invariant.
*/
-static void cl_page_delete0(const struct lu_env *env, struct cl_page *pg)
+static void __cl_page_delete(const struct lu_env *env, struct cl_page *cp)
{
- ENTRY;
-
- PASSERT(env, pg, pg->cp_state != CPS_FREEING);
+ const struct cl_page_slice *slice;
+ int i;
- /*
- * Severe all ways to obtain new pointers to @pg.
- */
- cl_page_owner_clear(pg);
+ ENTRY;
+ if (cp->cp_type != CPT_TRANSIENT)
+ PASSERT(env, cp, cp->cp_state != CPS_FREEING);
- cl_page_state_set0(env, pg, CPS_FREEING);
+ /* Severe all ways to obtain new pointers to @pg. */
+ cl_page_owner_clear(cp);
+ if (cp->cp_type != CPT_TRANSIENT)
+ __cl_page_state_set(env, cp, CPS_FREEING);
- CL_PAGE_INVOID_REVERSE(env, pg, CL_PAGE_OP(cpo_delete),
- (const struct lu_env *, const struct cl_page_slice *));
+ cl_page_slice_for_each_reverse(cp, slice, i) {
+ if (slice->cpl_ops->cpo_delete != NULL)
+ (*slice->cpl_ops->cpo_delete)(env, slice);
+ }
- EXIT;
+ EXIT;
}
/**
{
PINVRNT(env, pg, cl_page_invariant(pg));
ENTRY;
- cl_page_delete0(env, pg);
+ __cl_page_delete(env, pg);
EXIT;
}
EXPORT_SYMBOL(cl_page_delete);
-/**
- * Marks page up-to-date.
- *
- * Call cl_page_operations::cpo_export() through all layers top-to-bottom. The
- * layer responsible for VM interaction has to mark/clear page as up-to-date
- * by the \a uptodate argument.
- *
- * \see cl_page_operations::cpo_export()
- */
-void cl_page_export(const struct lu_env *env, struct cl_page *pg, int uptodate)
+void cl_page_touch(const struct lu_env *env,
+ const struct cl_page *cl_page, size_t to)
{
- PINVRNT(env, pg, cl_page_invariant(pg));
- CL_PAGE_INVOID(env, pg, CL_PAGE_OP(cpo_export),
- (const struct lu_env *,
- const struct cl_page_slice *, int), uptodate);
-}
-EXPORT_SYMBOL(cl_page_export);
+ const struct cl_page_slice *slice;
+ int i;
-/**
- * Returns true, iff \a pg is VM locked in a suitable sense by the calling
- * thread.
- */
-int cl_page_is_vmlocked(const struct lu_env *env, const struct cl_page *pg)
-{
- int result;
- const struct cl_page_slice *slice;
-
- ENTRY;
- slice = container_of(pg->cp_layers.next,
- const struct cl_page_slice, cpl_linkage);
- PASSERT(env, pg, slice->cpl_ops->cpo_is_vmlocked != NULL);
- /*
- * Call ->cpo_is_vmlocked() directly instead of going through
- * CL_PAGE_INVOKE(), because cl_page_is_vmlocked() is used by
- * cl_page_invariant().
- */
- result = slice->cpl_ops->cpo_is_vmlocked(env, slice);
- PASSERT(env, pg, result == -EBUSY || result == -ENODATA);
- RETURN(result == -EBUSY);
+ ENTRY;
+
+ cl_page_slice_for_each(cl_page, slice, i) {
+ if (slice->cpl_ops->cpo_page_touch != NULL)
+ (*slice->cpl_ops->cpo_page_touch)(env, slice, to);
+ }
+
+ EXIT;
}
-EXPORT_SYMBOL(cl_page_is_vmlocked);
+EXPORT_SYMBOL(cl_page_touch);
static enum cl_page_state cl_req_type_state(enum cl_req_type crt)
{
- ENTRY;
- RETURN(crt == CRT_WRITE ? CPS_PAGEOUT : CPS_PAGEIN);
+ ENTRY;
+ RETURN(crt == CRT_WRITE ? CPS_PAGEOUT : CPS_PAGEIN);
}
static void cl_page_io_start(const struct lu_env *env,
- struct cl_page *pg, enum cl_req_type crt)
+ struct cl_page *pg, enum cl_req_type crt)
{
- /*
- * Page is queued for IO, change its state.
- */
- ENTRY;
- cl_page_owner_clear(pg);
- cl_page_state_set(env, pg, cl_req_type_state(crt));
- EXIT;
+ /* Page is queued for IO, change its state. */
+ ENTRY;
+ cl_page_owner_clear(pg);
+ if (pg->cp_type != CPT_TRANSIENT)
+ cl_page_state_set(env, pg, cl_req_type_state(crt));
+ EXIT;
}
/**
- * Prepares page for immediate transfer. cl_page_operations::cpo_prep() is
- * called top-to-bottom. Every layer either agrees to submit this page (by
- * returning 0), or requests to omit this page (by returning -EALREADY). Layer
- * handling interactions with the VM also has to inform VM that page is under
- * transfer now.
+ * Prepares page for immediate transfer. Return -EALREADY if this page
+ * should be omitted from transfer.
*/
int cl_page_prep(const struct lu_env *env, struct cl_io *io,
- struct cl_page *pg, enum cl_req_type crt)
+ struct cl_page *cp, enum cl_req_type crt)
{
- int result;
+ struct page *vmpage = cp->cp_vmpage;
+ int rc;
- PINVRNT(env, pg, cl_page_is_owned(pg, io));
- PINVRNT(env, pg, cl_page_invariant(pg));
- PINVRNT(env, pg, crt < CRT_NR);
+ PASSERT(env, cp, crt < CRT_NR);
+ PINVRNT(env, cp, cl_page_is_owned(cp, io));
+ PINVRNT(env, cp, cl_page_invariant(cp));
- /*
- * XXX this has to be called bottom-to-top, so that llite can set up
- * PG_writeback without risking other layers deciding to skip this
- * page.
- */
- if (crt >= CRT_NR)
- return -EINVAL;
- result = cl_page_invoke(env, io, pg, CL_PAGE_OP(io[crt].cpo_prep));
- if (result == 0)
- cl_page_io_start(env, pg, crt);
-
- KLASSERT(ergo(crt == CRT_WRITE && pg->cp_type == CPT_CACHEABLE,
- equi(result == 0,
- PageWriteback(cl_page_vmpage(pg)))));
- CL_PAGE_HEADER(D_TRACE, env, pg, "%d %d\n", crt, result);
- return result;
+ if (cp->cp_type == CPT_TRANSIENT) {
+ /* Nothing to do. */
+ } else if (crt == CRT_READ) {
+ if (PageUptodate(vmpage))
+ GOTO(out, rc = -EALREADY);
+ } else {
+ LASSERT(PageLocked(vmpage));
+ LASSERT(!PageDirty(vmpage));
+
+ /* ll_writepage path is not a sync write, so need to
+ * set page writeback flag
+ */
+ if (cp->cp_sync_io == NULL)
+ set_page_writeback(vmpage);
+ }
+
+ cl_page_io_start(env, cp, crt);
+ rc = 0;
+out:
+ CL_PAGE_HEADER(D_TRACE, env, cp, "%d %d\n", crt, rc);
+
+ return rc;
}
EXPORT_SYMBOL(cl_page_prep);
* uppermost layer (llite), responsible for the VFS/VM interaction runs last
* and can release locks safely.
*
- * \pre pg->cp_state == CPS_PAGEIN || pg->cp_state == CPS_PAGEOUT
- * \post pg->cp_state == CPS_CACHED
+ * \pre cl_page->cp_state == CPS_PAGEIN || cl_page->cp_state == CPS_PAGEOUT
+ * \post cl_page->cl_page_state == CPS_CACHED
*
* \see cl_page_operations::cpo_completion()
*/
void cl_page_completion(const struct lu_env *env,
- struct cl_page *pg, enum cl_req_type crt, int ioret)
+ struct cl_page *cl_page, enum cl_req_type crt,
+ int ioret)
{
- struct cl_sync_io *anchor = pg->cp_sync_io;
-
- PASSERT(env, pg, crt < CRT_NR);
- /* cl_page::cp_req already cleared by the caller (osc_completion()) */
- PASSERT(env, pg, pg->cp_req == NULL);
- PASSERT(env, pg, pg->cp_state == cl_req_type_state(crt));
+ const struct cl_page_slice *slice;
+ struct cl_sync_io *anchor = cl_page->cp_sync_io;
+ int i;
- ENTRY;
- CL_PAGE_HEADER(D_TRACE, env, pg, "%d %d\n", crt, ioret);
- cl_page_state_set(env, pg, CPS_CACHED);
+ ENTRY;
+ PASSERT(env, cl_page, crt < CRT_NR);
+ if (cl_page->cp_type != CPT_TRANSIENT)
+ PASSERT(env, cl_page,
+ cl_page->cp_state == cl_req_type_state(crt));
+
+ CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", crt, ioret);
+ if (cl_page->cp_type != CPT_TRANSIENT)
+ cl_page_state_set(env, cl_page, CPS_CACHED);
if (crt >= CRT_NR)
return;
- CL_PAGE_INVOID_REVERSE(env, pg, CL_PAGE_OP(io[crt].cpo_completion),
- (const struct lu_env *,
- const struct cl_page_slice *, int), ioret);
- if (anchor) {
- LASSERT(pg->cp_sync_io == anchor);
- pg->cp_sync_io = NULL;
- }
- /*
- * As page->cp_obj is pinned by a reference from page->cp_req, it is
- * safe to call cl_page_put() without risking object destruction in a
- * non-blocking context.
- */
- cl_page_put(env, pg);
- if (anchor)
- cl_sync_io_note(anchor, ioret);
+ cl_page_slice_for_each_reverse(cl_page, slice, i) {
+ if (slice->cpl_ops->io[crt].cpo_completion != NULL)
+ (*slice->cpl_ops->io[crt].cpo_completion)(env, slice,
+ ioret);
+ }
- EXIT;
+ if (anchor != NULL) {
+ LASSERT(cl_page->cp_sync_io == anchor);
+ cl_page->cp_sync_io = NULL;
+ cl_sync_io_note(env, anchor, ioret);
+ }
+ EXIT;
}
EXPORT_SYMBOL(cl_page_completion);
* Notify layers that transfer formation engine decided to yank this page from
* the cache and to make it a part of a transfer.
*
- * \pre pg->cp_state == CPS_CACHED
- * \post pg->cp_state == CPS_PAGEIN || pg->cp_state == CPS_PAGEOUT
- *
- * \see cl_page_operations::cpo_make_ready()
+ * \pre cl_page->cp_state == CPS_CACHED
+ * \post cl_page->cp_state == CPS_PAGEIN || cl_page->cp_state == CPS_PAGEOUT
*/
-int cl_page_make_ready(const struct lu_env *env, struct cl_page *pg,
- enum cl_req_type crt)
+int cl_page_make_ready(const struct lu_env *env, struct cl_page *cp,
+ enum cl_req_type crt)
{
- int result;
+ struct page *vmpage = cp->cp_vmpage;
+ bool unlock = false;
+ int rc;
- PINVRNT(env, pg, crt < CRT_NR);
+ ENTRY;
+ PASSERT(env, cp, crt == CRT_WRITE);
+
+ if (cp->cp_type == CPT_TRANSIENT)
+ GOTO(out, rc = 0);
+
+ lock_page(vmpage);
+ PASSERT(env, cp, PageUptodate(vmpage));
+ unlock = true;
+
+ if (clear_page_dirty_for_io(vmpage)) {
+ LASSERT(cp->cp_state == CPS_CACHED);
+ /* This actually clears the dirty bit in the radix tree */
+ set_page_writeback(vmpage);
+ CL_PAGE_HEADER(D_PAGE, env, cp, "readied\n");
+ rc = 0;
+ } else if (cp->cp_state == CPS_PAGEOUT) {
+ /* is it possible for osc_flush_async_page()
+ * to already make it ready?
+ */
+ rc = -EALREADY;
+ } else {
+ CL_PAGE_DEBUG(D_ERROR, env, cp,
+ "unexpecting page state %d\n",
+ cp->cp_state);
+ LBUG();
+ }
- ENTRY;
- if (crt >= CRT_NR)
- RETURN(-EINVAL);
- result = CL_PAGE_INVOKE(env, pg, CL_PAGE_OP(io[crt].cpo_make_ready),
- (const struct lu_env *,
- const struct cl_page_slice *));
- if (result == 0) {
- PASSERT(env, pg, pg->cp_state == CPS_CACHED);
- cl_page_io_start(env, pg, crt);
- }
- CL_PAGE_HEADER(D_TRACE, env, pg, "%d %d\n", crt, result);
- RETURN(result);
+out:
+ if (rc == 0) {
+ PASSERT(env, cp, cp->cp_state == CPS_CACHED);
+ cl_page_io_start(env, cp, crt);
+ }
+
+ if (unlock)
+ unlock_page(vmpage);
+
+ CL_PAGE_HEADER(D_TRACE, env, cp, "%d %d\n", crt, rc);
+
+ return rc;
}
EXPORT_SYMBOL(cl_page_make_ready);
/**
- * Called if a pge is being written back by kernel's intention.
+ * Called if a page is being written back by kernel's intention.
*
- * \pre cl_page_is_owned(pg, io)
- * \post ergo(result == 0, pg->cp_state == CPS_PAGEOUT)
+ * \pre cl_page_is_owned(cl_page, io)
+ * \post ergo(result == 0, cl_page->cp_state == CPS_PAGEOUT)
*
* \see cl_page_operations::cpo_flush()
*/
int cl_page_flush(const struct lu_env *env, struct cl_io *io,
- struct cl_page *pg)
+ struct cl_page *cl_page)
{
- int result;
-
- PINVRNT(env, pg, cl_page_is_owned(pg, io));
- PINVRNT(env, pg, cl_page_invariant(pg));
+ const struct cl_page_slice *slice;
+ int result = 0;
+ int i;
ENTRY;
+ PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
+ PINVRNT(env, cl_page, cl_page_invariant(cl_page));
+
+ cl_page_slice_for_each(cl_page, slice, i) {
+ if (slice->cpl_ops->cpo_flush != NULL)
+ result = (*slice->cpl_ops->cpo_flush)(env, slice, io);
+ if (result != 0)
+ break;
+ }
+ if (result > 0)
+ result = 0;
- result = cl_page_invoke(env, io, pg, CL_PAGE_OP(cpo_flush));
-
- CL_PAGE_HEADER(D_TRACE, env, pg, "%d\n", result);
+ CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d\n", result);
RETURN(result);
}
EXPORT_SYMBOL(cl_page_flush);
/**
- * Checks whether page is protected by any extent lock is at least required
- * mode.
- *
- * \return the same as in cl_page_operations::cpo_is_under_lock() method.
- * \see cl_page_operations::cpo_is_under_lock()
- */
-int cl_page_is_under_lock(const struct lu_env *env, struct cl_io *io,
- struct cl_page *page, pgoff_t *max_index)
-{
- int rc;
-
- PINVRNT(env, page, cl_page_invariant(page));
-
- ENTRY;
- rc = CL_PAGE_INVOKE_REVERSE(env, page, CL_PAGE_OP(cpo_is_under_lock),
- (const struct lu_env *,
- const struct cl_page_slice *,
- struct cl_io *, pgoff_t *),
- io, max_index);
- RETURN(rc);
-}
-EXPORT_SYMBOL(cl_page_is_under_lock);
-
-/**
* Tells transfer engine that only part of a page is to be transmitted.
*
* \see cl_page_operations::cpo_clip()
*/
-void cl_page_clip(const struct lu_env *env, struct cl_page *pg,
- int from, int to)
+void cl_page_clip(const struct lu_env *env, struct cl_page *cl_page,
+ int from, int to)
{
- PINVRNT(env, pg, cl_page_invariant(pg));
+ const struct cl_page_slice *slice;
+ int i;
- CL_PAGE_HEADER(D_TRACE, env, pg, "%d %d\n", from, to);
- CL_PAGE_INVOID(env, pg, CL_PAGE_OP(cpo_clip),
- (const struct lu_env *,
- const struct cl_page_slice *,int, int),
- from, to);
+ PINVRNT(env, cl_page, cl_page_invariant(cl_page));
+
+ CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", from, to);
+ cl_page_slice_for_each(cl_page, slice, i) {
+ if (slice->cpl_ops->cpo_clip != NULL)
+ (*slice->cpl_ops->cpo_clip)(env, slice, from, to);
+ }
}
EXPORT_SYMBOL(cl_page_clip);
-/**
- * Prints human readable representation of \a pg to the \a f.
- */
+/* Prints human readable representation of \a pg to the \a f. */
void cl_page_header_print(const struct lu_env *env, void *cookie,
- lu_printer_t printer, const struct cl_page *pg)
+ lu_printer_t printer, const struct cl_page *pg)
{
(*printer)(env, cookie,
- "page@%p[%d %p %d %d %d %p %p]\n",
- pg, atomic_read(&pg->cp_ref), pg->cp_obj,
- pg->cp_state, pg->cp_error, pg->cp_type,
- pg->cp_owner, pg->cp_req);
+ "page@%p[%d %p %d %d %p]\n",
+ pg, refcount_read(&pg->cp_ref), pg->cp_obj,
+ pg->cp_state, pg->cp_type,
+ pg->cp_owner);
}
EXPORT_SYMBOL(cl_page_header_print);
-/**
- * Prints human readable representation of \a pg to the \a f.
- */
+/* Prints human readable representation of \a cl_page to the \a f. */
void cl_page_print(const struct lu_env *env, void *cookie,
- lu_printer_t printer, const struct cl_page *pg)
+ lu_printer_t printer, const struct cl_page *cp)
{
- cl_page_header_print(env, cookie, printer, pg);
- CL_PAGE_INVOKE(env, (struct cl_page *)pg, CL_PAGE_OP(cpo_print),
- (const struct lu_env *env,
- const struct cl_page_slice *slice,
- void *cookie, lu_printer_t p), cookie, printer);
- (*printer)(env, cookie, "end page@%p\n", pg);
-}
-EXPORT_SYMBOL(cl_page_print);
+ struct page *vmpage = cp->cp_vmpage;
+ const struct cl_page_slice *slice;
+ int result = 0;
+ int i;
-/**
- * Cancel a page which is still in a transfer.
- */
-int cl_page_cancel(const struct lu_env *env, struct cl_page *page)
-{
- return CL_PAGE_INVOKE(env, page, CL_PAGE_OP(cpo_cancel),
- (const struct lu_env *,
- const struct cl_page_slice *));
-}
-EXPORT_SYMBOL(cl_page_cancel);
+ cl_page_header_print(env, cookie, printer, cp);
-/**
- * Converts a byte offset within object \a obj into a page index.
- */
-loff_t cl_offset(const struct cl_object *obj, pgoff_t idx)
-{
- return (loff_t)idx << PAGE_CACHE_SHIFT;
-}
-EXPORT_SYMBOL(cl_offset);
+ (*printer)(env, cookie, "vmpage @%p", vmpage);
-/**
- * Converts a page index into a byte offset within object \a obj.
- */
-pgoff_t cl_index(const struct cl_object *obj, loff_t offset)
-{
- return offset >> PAGE_CACHE_SHIFT;
-}
-EXPORT_SYMBOL(cl_index);
+ if (vmpage != NULL) {
+ (*printer)(env, cookie, " %lx %d:%d %lx %lu %slru",
+ (long)vmpage->flags, page_count(vmpage),
+ page_mapcount(vmpage), vmpage->private,
+ page_index(vmpage),
+ list_empty(&vmpage->lru) ? "not-" : "");
+ }
-int cl_page_size(const struct cl_object *obj)
-{
- return 1 << PAGE_CACHE_SHIFT;
+ (*printer)(env, cookie, "\n");
+
+ cl_page_slice_for_each(cp, slice, i) {
+ if (slice->cpl_ops->cpo_print != NULL)
+ result = (*slice->cpl_ops->cpo_print)(env, slice,
+ cookie, printer);
+ if (result != 0)
+ break;
+ }
+
+ (*printer)(env, cookie, "end page@%p\n", cp);
}
-EXPORT_SYMBOL(cl_page_size);
+EXPORT_SYMBOL(cl_page_print);
/**
* Adds page slice to the compound page.
*
* \see cl_lock_slice_add(), cl_req_slice_add(), cl_io_slice_add()
*/
-void cl_page_slice_add(struct cl_page *page, struct cl_page_slice *slice,
- struct cl_object *obj, pgoff_t index,
+void cl_page_slice_add(struct cl_page *cl_page, struct cl_page_slice *slice,
+ struct cl_object *obj,
const struct cl_page_operations *ops)
{
+ unsigned int offset = (char *)slice -
+ ((char *)cl_page + sizeof(*cl_page));
+
ENTRY;
- list_add_tail(&slice->cpl_linkage, &page->cp_layers);
- slice->cpl_obj = obj;
- slice->cpl_index = index;
+ LASSERT(cl_page->cp_layer_count < CP_MAX_LAYER);
+ LASSERT(offset < (1 << sizeof(cl_page->cp_layer_offset[0]) * 8));
+ cl_page->cp_layer_offset[cl_page->cp_layer_count++] = offset;
slice->cpl_ops = ops;
- slice->cpl_page = page;
+ slice->cpl_page = cl_page;
+
EXIT;
}
EXPORT_SYMBOL(cl_page_slice_add);
-int cl_page_init(void)
+/* Allocate and initialize cl_cache, called by ll_init_sbi(). */
+struct cl_client_cache *cl_cache_init(unsigned long lru_page_max)
+{
+ struct cl_client_cache *cache = NULL;
+
+ ENTRY;
+ OBD_ALLOC(cache, sizeof(*cache));
+ if (cache == NULL)
+ RETURN(NULL);
+
+ /* Initialize cache data */
+ refcount_set(&cache->ccc_users, 1);
+ cache->ccc_lru_max = lru_page_max;
+ atomic_long_set(&cache->ccc_lru_left, lru_page_max);
+ spin_lock_init(&cache->ccc_lru_lock);
+ INIT_LIST_HEAD(&cache->ccc_lru);
+
+ cache->ccc_unstable_check = 1;
+ atomic_long_set(&cache->ccc_unstable_nr, 0);
+ mutex_init(&cache->ccc_max_cache_mb_lock);
+
+ RETURN(cache);
+}
+EXPORT_SYMBOL(cl_cache_init);
+
+/* Increase cl_cache refcount */
+void cl_cache_incref(struct cl_client_cache *cache)
{
- return 0;
+ refcount_inc(&cache->ccc_users);
}
+EXPORT_SYMBOL(cl_cache_incref);
-void cl_page_fini(void)
+/**
+ * Decrease cl_cache refcount and free the cache if refcount=0.
+ * Since llite, lov and osc all hold cl_cache refcount,
+ * the free will not cause race. (LU-6173)
+ */
+void cl_cache_decref(struct cl_client_cache *cache)
{
+ if (refcount_dec_and_test(&cache->ccc_users))
+ OBD_FREE(cache, sizeof(*cache));
}
+EXPORT_SYMBOL(cl_cache_decref);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff