*/
/*
* This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
*
* Client Lustre Page.
*
#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))
}
static void cs_pagestate_dec(const struct cl_object *obj,
- enum cl_page_state state)
+ enum cl_page_state state)
{
#ifdef CONFIG_DEBUG_PAGESTATE_TRACKING
atomic_dec(&cl_object_site(obj)->cs_pages_state[state]);
*/
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);
}
static struct cl_page_slice *
slice = cl_page_slice_get(cl_page, i); i >= 0; \
slice = cl_page_slice_get(cl_page, --i))
-/**
- * Returns a slice within a cl_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 *cl_page,
- const struct lu_device_type *dtype)
-{
- const struct cl_page_slice *slice;
- int i;
-
- ENTRY;
-
- cl_page_slice_for_each(cl_page, slice, i) {
- if (slice->cpl_obj->co_lu.lo_dev->ld_type == dtype)
- RETURN(slice);
- }
-
- RETURN(NULL);
-}
-
static void __cl_page_free(struct cl_page *cl_page, unsigned short bufsize)
{
int index = cl_page->cp_kmem_index;
}
}
-static void cl_page_free(const struct lu_env *env, struct cl_page *cl_page,
- struct pagevec *pvec)
+static void cl_page_free(const struct lu_env *env, struct cl_page *cp,
+ struct folio_batch *fbatch)
{
- struct cl_object *obj = cl_page->cp_obj;
+ struct cl_object *obj = cp->cp_obj;
unsigned short bufsize = cl_object_header(obj)->coh_page_bufsize;
- struct cl_page_slice *slice;
- int i;
+ struct page *vmpage;
ENTRY;
- PASSERT(env, cl_page, list_empty(&cl_page->cp_batch));
- PASSERT(env, cl_page, cl_page->cp_owner == NULL);
- PASSERT(env, cl_page, cl_page->cp_state == CPS_FREEING);
+ PASSERT(env, cp, list_empty(&cp->cp_batch));
- cl_page_slice_for_each(cl_page, slice, i) {
- if (unlikely(slice->cpl_ops->cpo_fini != NULL))
- slice->cpl_ops->cpo_fini(env, slice, pvec);
+ if (cp->cp_type == CPT_CACHEABLE) {
+ PASSERT(env, cp, cp->cp_owner == NULL);
+ PASSERT(env, cp, cp->cp_state == CPS_FREEING);
+ /* 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);
+ }
}
- cl_page->cp_layer_count = 0;
+
+ cp->cp_layer_count = 0;
cs_page_dec(obj, CS_total);
- cs_pagestate_dec(obj, cl_page->cp_state);
- lu_object_ref_del_at(&obj->co_lu, &cl_page->cp_obj_ref,
- "cl_page", cl_page);
- cl_object_put(env, obj);
- lu_ref_fini(&cl_page->cp_reference);
- __cl_page_free(cl_page, bufsize);
+ 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;
}
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) {
+ 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)
mutex_unlock(&cl_page_kmem_mutex);
return NULL;
}
- smp_store_release(&cl_page_kmem_size_array[i],
- bufsize);
+ smp_store_release(&cl_page_kmem_size_array[i], bufsize);
mutex_unlock(&cl_page_kmem_mutex);
goto check;
} else {
enum cl_page_type type)
{
struct cl_page *cl_page;
- struct lu_object_header *head;
+ struct cl_object *head;
ENTRY;
if (cl_page != NULL) {
int result = 0;
- /*
- * Please fix cl_page:cp_state/type declaration if
+ /* 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 */
- atomic_set(&cl_page->cp_ref, 1);
+ refcount_set(&cl_page->cp_ref, 1);
cl_page->cp_obj = o;
- cl_object_get(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;
- cl_page->cp_state = CPS_CACHED;
+ if (cl_page->cp_type != CPT_TRANSIENT)
+ cl_page->cp_state = CPS_CACHED;
cl_page->cp_type = type;
- cl_page->cp_inode = NULL;
+ 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->co_lu.lo_header;
- list_for_each_entry(o, &head->loh_layers,
- co_lu.lo_linkage) {
+ 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,
cl_page, ind);
if (result != 0) {
- cl_page_delete0(env, cl_page);
+ __cl_page_delete(env, cl_page);
cl_page_free(env, cl_page, NULL);
cl_page = ERR_PTR(result);
break;
hdr = cl_object_header(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 is used to protect the child/parent
- * relationship */
+ 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);
+ /*
+ * 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, 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 *cl_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.
- */
+ /* Matrix of allowed state transitions [old][new] for sanity checking */
static const int allowed_transitions[CPS_NR][CPS_NR] = {
[CPS_CACHED] = {
[CPS_CACHED] = 0,
}
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, use the pagevec to release the pages
+ * Releases a reference to a page, use the folio_batch to release the pages
* in batch if provided.
*
- * Users need to do a final pagevec_release() to release any trailing pages.
+ * Users need to do a final folio_batch_release() to release any trailing pages.
*/
-void cl_pagevec_put(const struct lu_env *env, struct cl_page *page,
- struct pagevec *pvec)
+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);
+ PASSERT(env, page, page->cp_owner == NULL);
+ }
- LASSERT(atomic_read(&page->cp_ref) == 0);
- PASSERT(env, page, page->cp_owner == NULL);
+ LASSERT(refcount_read(&page->cp_ref) == 0);
PASSERT(env, page, list_empty(&page->cp_batch));
- /*
- * Page is no longer reachable by other threads. Tear
- * it down.
- */
- cl_page_free(env, page, pvec);
+ /* Page is no longer reachable by other threads. Tear it down */
+ cl_page_free(env, page, fbatch);
}
EXIT;
}
-EXPORT_SYMBOL(cl_pagevec_put);
+EXPORT_SYMBOL(cl_batch_put);
/**
- * Releases a reference to a page, wrapper to cl_pagevec_put
+ * 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
*/
void cl_page_put(const struct lu_env *env, struct cl_page *page)
{
- cl_pagevec_put(env, page, NULL);
+ cl_batch_put(env, page, NULL);
}
EXPORT_SYMBOL(cl_page_put);
-/**
- * Returns a cl_page associated with a VM page, and given cl_object.
- */
+/* 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;
}
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);
-
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 *cl_page)
+void __cl_page_disown(const struct lu_env *env, struct cl_page *cp)
{
- const struct cl_page_slice *slice;
+ struct page *vmpage;
enum cl_page_state state;
- int i;
-
- ENTRY;
- state = cl_page->cp_state;
- PINVRNT(env, cl_page, state == CPS_OWNED ||
- state == CPS_FREEING);
- PINVRNT(env, cl_page, cl_page_invariant(cl_page) ||
- state == CPS_FREEING);
- cl_page_owner_clear(cl_page);
- if (state == CPS_OWNED)
- cl_page_state_set(env, cl_page, 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_slice_for_each_reverse(cl_page, slice, i) {
- if (slice->cpl_ops->cpo_disown != NULL)
- (*slice->cpl_ops->cpo_disown)(env, slice, io);
+ ENTRY;
+ if (cp->cp_type == CPT_CACHEABLE) {
+ cl_page_owner_clear(cp);
+ 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);
* 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 *cl_page, int nonblock)
+static int __cl_page_own(const struct lu_env *env, struct cl_io *io,
+ struct cl_page *cl_page, int nonblock)
{
- const struct cl_page_slice *slice;
- int result = 0;
- int i;
+ struct page *vmpage = cl_page->cp_vmpage;
+ int result;
ENTRY;
PINVRNT(env, cl_page, !cl_page_is_owned(cl_page, io));
- io = cl_io_top(io);
- if (cl_page->cp_state == CPS_FREEING) {
+ LASSERT(cl_page->cp_type != CPT_TRANSIENT);
+
+ if (cl_page->cp_type != CPT_TRANSIENT &&
+ cl_page->cp_state == CPS_FREEING) {
result = -ENOENT;
goto out;
}
- cl_page_slice_for_each(cl_page, slice, i) {
- if (slice->cpl_ops->cpo_own)
- result = (*slice->cpl_ops->cpo_own)(env, slice,
- io, nonblock);
- if (result != 0)
- break;
- }
- if (result > 0)
- result = 0;
+ LASSERT(vmpage != NULL);
- if (result == 0) {
- 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_state != CPS_FREEING) {
- cl_page_state_set(env, cl_page, CPS_OWNED);
- } else {
- cl_page_disown0(env, io, cl_page);
- result = -ENOENT;
+ 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_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(cl_page, io)
- * \post cl_page_is_owned(cl_page, 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 *cl_page)
+ struct cl_io *io, struct cl_page *cp)
{
- const struct cl_page_slice *slice;
- int i;
+ struct page *vmpage;
ENTRY;
+ PINVRNT(env, cp, cl_object_same(cp->cp_obj, cl_io_top(io)->ci_obj));
- PINVRNT(env, cl_page,
- cl_object_same(cl_page->cp_obj, io->ci_obj));
- io = cl_io_top(io);
+ LASSERT(cp->cp_type != CPT_TRANSIENT);
+ PASSERT(env, cp, cp->cp_owner == NULL);
- cl_page_slice_for_each(cl_page, slice, i) {
- if (slice->cpl_ops->cpo_assume != NULL)
- (*slice->cpl_ops->cpo_assume)(env, slice, io);
- }
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT(PageLocked(vmpage));
+ wait_on_page_writeback(vmpage);
+ cp->cp_owner = cl_io_top(io);
+ cl_page_owner_set(cp);
+ cl_page_state_set(env, cp, CPS_OWNED);
- PASSERT(env, cl_page, cl_page->cp_owner == NULL);
- cl_page->cp_owner = cl_io_top(io);
- cl_page_owner_set(cl_page);
- cl_page_state_set(env, cl_page, CPS_OWNED);
EXIT;
}
EXPORT_SYMBOL(cl_page_assume);
* 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).
*
- * \pre cl_page_is_owned(cl_page, io)
- * \post !cl_page_is_owned(cl_page, io)
- *
- * \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 *cl_page)
+ struct cl_io *io, struct cl_page *cp)
{
- const struct cl_page_slice *slice;
- int i;
+ struct page *vmpage;
- ENTRY;
- PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
- PINVRNT(env, cl_page, cl_page_invariant(cl_page));
+ ENTRY;
- io = cl_io_top(io);
- cl_page_owner_clear(cl_page);
- cl_page_state_set(env, cl_page, CPS_CACHED);
+ LASSERT(cp->cp_type != CPT_TRANSIENT);
- cl_page_slice_for_each_reverse(cl_page, slice, i) {
- if (slice->cpl_ops->cpo_unassume != NULL)
- (*slice->cpl_ops->cpo_unassume)(env, slice, io);
- }
+ PINVRNT(env, cp, cl_page_is_owned(cp, io));
+ PINVRNT(env, cp, cl_page_invariant(cp));
+ cl_page_owner_clear(cp);
+ cl_page_state_set(env, cp, CPS_CACHED);
+ vmpage = cp->cp_vmpage;
+ LASSERT(vmpage != NULL);
+ LASSERT(PageLocked(vmpage));
EXIT;
}
* \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);
* \see cl_page_operations::cpo_discard()
*/
void cl_page_discard(const struct lu_env *env,
- struct cl_io *io, struct cl_page *cl_page)
+ struct cl_io *io, struct cl_page *cp)
{
+ struct page *vmpage;
const struct cl_page_slice *slice;
int i;
- 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) {
+ 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) {
+ PINVRNT(env, cp, cl_page_is_owned(cp, io));
+ PINVRNT(env, cp, cl_page_invariant(cp));
+ 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
- * cl_pages, e.g. in an error handling cl_page_find()->cl_page_delete0()
+ * 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 *cl_page)
+static void __cl_page_delete(const struct lu_env *env, struct cl_page *cp)
{
const struct cl_page_slice *slice;
int i;
- ENTRY;
-
- PASSERT(env, cl_page, cl_page->cp_state != CPS_FREEING);
+ ENTRY;
/*
* Severe all ways to obtain new pointers to @pg.
+ * Transient pages already can't be found because they're not in cache.
*/
- cl_page_owner_clear(cl_page);
- cl_page_state_set0(env, cl_page, CPS_FREEING);
+ if (cp->cp_type != CPT_TRANSIENT) {
+ PASSERT(env, cp, cp->cp_state != CPS_FREEING);
+ cl_page_owner_clear(cp);
+ __cl_page_state_set(env, cp, CPS_FREEING);
+ }
- cl_page_slice_for_each_reverse(cl_page, slice, i) {
+ cl_page_slice_for_each_reverse(cp, slice, i) {
if (slice->cpl_ops->cpo_delete != NULL)
(*slice->cpl_ops->cpo_delete)(env, slice);
}
{
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 *cl_page,
- int uptodate)
-{
- const struct cl_page_slice *slice;
- int i;
-
- PINVRNT(env, cl_page, cl_page_invariant(cl_page));
-
- cl_page_slice_for_each(cl_page, slice, i) {
- if (slice->cpl_ops->cpo_export != NULL)
- (*slice->cpl_ops->cpo_export)(env, slice, uptodate);
- }
-}
-EXPORT_SYMBOL(cl_page_export);
-
-/**
- * Returns true, if \a page 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 *cl_page)
-{
- const struct cl_page_slice *slice;
- int result;
-
- ENTRY;
- slice = cl_page_slice_get(cl_page, 0);
- PASSERT(env, cl_page, 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, cl_page, result == -EBUSY || result == -ENODATA);
-
- RETURN(result == -EBUSY);
-}
-EXPORT_SYMBOL(cl_page_is_vmlocked);
-
void cl_page_touch(const struct lu_env *env,
const struct cl_page *cl_page, size_t to)
{
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;
+ if (pg->cp_type != CPT_TRANSIENT) {
+ cl_page_owner_clear(pg);
+ 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 *cl_page, enum cl_req_type crt)
+ struct cl_page *cp, enum cl_req_type crt)
{
- const struct cl_page_slice *slice;
- int result = 0;
- int i;
-
- PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
- PINVRNT(env, cl_page, cl_page_invariant(cl_page));
- PINVRNT(env, cl_page, crt < CRT_NR);
-
- /*
- * 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;
+ struct page *vmpage = cp->cp_vmpage;
+ int rc;
+
+ if (cp->cp_type == CPT_TRANSIENT)
+ GOTO(start, rc = 0);
+ PASSERT(env, cp, crt < CRT_NR);
+ PINVRNT(env, cp, cl_page_is_owned(cp, io));
+ PINVRNT(env, cp, cl_page_invariant(cp));
+
+ if (crt == CRT_READ) {
+ if (PageUptodate(vmpage))
+ GOTO(out, rc = -EALREADY);
+ } else {
+ LASSERT(PageLocked(vmpage));
+ LASSERT(!PageDirty(vmpage));
- cl_page_slice_for_each(cl_page, slice, i) {
- if (slice->cpl_ops->cpo_own)
- result = (*slice->cpl_ops->io[crt].cpo_prep)(env,
- slice,
- io);
- if (result != 0)
- break;
+ /* 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);
}
+start:
- if (result >= 0) {
- result = 0;
- cl_page_io_start(env, cl_page, crt);
- }
+ cl_page_io_start(env, cp, crt);
+ rc = 0;
+out:
+ CL_PAGE_HEADER(D_TRACE, env, cp, "%d %d\n", crt, rc);
- CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", crt, result);
- return result;
+ return rc;
}
EXPORT_SYMBOL(cl_page_prep);
struct cl_sync_io *anchor = cl_page->cp_sync_io;
int i;
- ENTRY;
+ ENTRY;
PASSERT(env, cl_page, crt < CRT_NR);
- PASSERT(env, cl_page, cl_page->cp_state == cl_req_type_state(crt));
+ 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);
- cl_page_state_set(env, cl_page, CPS_CACHED);
+ if (cl_page->cp_type != CPT_TRANSIENT)
+ cl_page_state_set(env, cl_page, CPS_CACHED);
if (crt >= CRT_NR)
return;
*
* \pre cl_page->cp_state == CPS_CACHED
* \post cl_page->cp_state == CPS_PAGEIN || cl_page->cp_state == CPS_PAGEOUT
- *
- * \see cl_page_operations::cpo_make_ready()
*/
-int cl_page_make_ready(const struct lu_env *env, struct cl_page *cl_page,
- enum cl_req_type crt)
+int cl_page_make_ready(const struct lu_env *env, struct cl_page *cp,
+ enum cl_req_type crt)
{
- const struct cl_page_slice *slice;
- int result = 0;
- int i;
+ struct page *vmpage = cp->cp_vmpage;
+ bool unlock = false;
+ int rc;
- ENTRY;
- PINVRNT(env, cl_page, crt < CRT_NR);
- if (crt >= CRT_NR)
- RETURN(-EINVAL);
+ 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();
+ }
- cl_page_slice_for_each(cl_page, slice, i) {
- if (slice->cpl_ops->io[crt].cpo_make_ready != NULL)
- result = (*slice->cpl_ops->io[crt].cpo_make_ready)(env, slice);
- if (result != 0)
- break;
+out:
+ if (rc == 0) {
+ PASSERT(env, cp, cp->cp_state == CPS_CACHED);
+ cl_page_io_start(env, cp, crt);
}
- if (result >= 0) {
- result = 0;
- PASSERT(env, cl_page, cl_page->cp_state == CPS_CACHED);
- cl_page_io_start(env, cl_page, crt);
- }
- CL_PAGE_HEADER(D_TRACE, env, cl_page, "%d %d\n", crt, result);
+ if (unlock)
+ unlock_page(vmpage);
- RETURN(result);
+ CL_PAGE_HEADER(D_TRACE, env, cp, "%d %d\n", crt, rc);
+
+ return rc;
}
EXPORT_SYMBOL(cl_page_make_ready);
int i;
ENTRY;
+ LASSERT(cl_page->cp_type != CPT_TRANSIENT);
PINVRNT(env, cl_page, cl_page_is_owned(cl_page, io));
PINVRNT(env, cl_page, cl_page_invariant(cl_page));
* \see cl_page_operations::cpo_clip()
*/
void cl_page_clip(const struct lu_env *env, struct cl_page *cl_page,
- int from, int to)
+ int from, int to)
{
const struct cl_page_slice *slice;
int i;
}
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 %p]\n",
- pg, atomic_read(&pg->cp_ref), pg->cp_obj,
+ 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 cl_page 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 *cl_page)
+ lu_printer_t printer, const struct cl_page *cp)
{
+ struct page *vmpage = cp->cp_vmpage;
const struct cl_page_slice *slice;
int result = 0;
int i;
- cl_page_header_print(env, cookie, printer, cl_page);
- cl_page_slice_for_each(cl_page, slice, i) {
+ cl_page_header_print(env, cookie, printer, cp);
+
+ (*printer)(env, cookie, "vmpage @%p", vmpage);
+
+ 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-" : "");
+ }
+
+ (*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);
+ cookie, printer);
if (result != 0)
break;
}
- (*printer)(env, cookie, "end page@%p\n", cl_page);
-}
-EXPORT_SYMBOL(cl_page_print);
-
-/**
- * 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_SHIFT;
-}
-EXPORT_SYMBOL(cl_offset);
-/**
- * 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_SHIFT;
+ (*printer)(env, cookie, "end page@%p\n", cp);
}
-EXPORT_SYMBOL(cl_index);
-
-size_t cl_page_size(const struct cl_object *obj)
-{
- return 1UL << PAGE_SHIFT;
-}
-EXPORT_SYMBOL(cl_page_size);
+EXPORT_SYMBOL(cl_page_print);
/**
* Adds page slice to the compound page.
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_obj = obj;
slice->cpl_ops = ops;
slice->cpl_page = cl_page;
}
EXPORT_SYMBOL(cl_page_slice_add);
-/**
- * Allocate and initialize cl_cache, called by ll_init_sbi().
- */
+/* 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;
RETURN(NULL);
/* Initialize cache data */
- atomic_set(&cache->ccc_users, 1);
+ 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);
- /* turn unstable check off by default as it impacts performance */
- cache->ccc_unstable_check = 0;
+ cache->ccc_unstable_check = 1;
atomic_long_set(&cache->ccc_unstable_nr, 0);
- init_waitqueue_head(&cache->ccc_unstable_waitq);
mutex_init(&cache->ccc_max_cache_mb_lock);
RETURN(cache);
}
EXPORT_SYMBOL(cl_cache_init);
-/**
- * Increase cl_cache refcount
- */
+/* Increase cl_cache refcount */
void cl_cache_incref(struct cl_client_cache *cache)
{
- atomic_inc(&cache->ccc_users);
+ refcount_inc(&cache->ccc_users);
}
EXPORT_SYMBOL(cl_cache_incref);
*/
void cl_cache_decref(struct cl_client_cache *cache)
{
- if (atomic_dec_and_test(&cache->ccc_users))
+ if (refcount_dec_and_test(&cache->ccc_users))
OBD_FREE(cache, sizeof(*cache));
}
EXPORT_SYMBOL(cl_cache_decref);