* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2013, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* super-class definitions.
*/
#include <lu_object.h>
-#include <lvfs.h>
#ifdef __KERNEL__
# include <linux/mutex.h>
# include <linux/radix-tree.h>
* to be used instead of newly created.
*/
int (*coo_page_init)(const struct lu_env *env, struct cl_object *obj,
- struct cl_page *page, cfs_page_t *vmpage);
+ struct cl_page *page, struct page *vmpage);
/**
* Initialize lock slice for this layer. Called top-to-bottom through
* every object layer when a new cl_lock is instantiated. Layer
*/
int (*coo_glimpse)(const struct lu_env *env,
const struct cl_object *obj, struct ost_lvb *lvb);
+ /**
+ * Object prune method. Called when the layout is going to change on
+ * this object, therefore each layer has to clean up their cache,
+ * mainly pages and locks.
+ */
+ int (*coo_prune)(const struct lu_env *env, struct cl_object *obj);
};
/**
* mostly useless otherwise.
*/
/** @{ */
- /** Lock protecting page tree. */
- spinlock_t coh_page_guard;
/** Lock protecting lock list. */
spinlock_t coh_lock_guard;
/** @} locks */
- /** Radix tree of cl_page's, cached for this object. */
- struct radix_tree_root coh_tree;
- /** # of pages in radix tree. */
- unsigned long coh_pages;
/** List of cl_lock's granted for this object. */
cfs_list_t coh_locks;
* corresponding radix tree at the corresponding logical offset.
*
* cl_page is associated with VM page of the hosting environment (struct
- * page in Linux kernel, for example), cfs_page_t. It is assumed, that this
+ * page in Linux kernel, for example), struct page. It is assumed, that this
* association is implemented by one of cl_page layers (top layer in the
* current design) that
*
* - translates state (page flag bits) and locking between lustre and
* environment.
*
- * The association between cl_page and cfs_page_t is immutable and
+ * The association between cl_page and struct page is immutable and
* established when cl_page is created.
*
* cl_page can be "owned" by a particular cl_io (see below), guaranteeing
* eviction of the page from the memory). Note, that in general cl_io
* cannot be identified with a particular thread, and page ownership is not
* exactly equal to the current thread holding a lock on the page. Layer
- * implementing association between cl_page and cfs_page_t has to implement
+ * implementing association between cl_page and struct page has to implement
* ownership on top of available synchronization mechanisms.
*
* While lustre client maintains the notion of an page ownership by io,
* - by doing a lookup in the cl_object radix tree, protected by the
* spin-lock;
*
- * - by starting from VM-locked cfs_page_t and following some
+ * - by starting from VM-locked struct page and following some
* hosting environment method (e.g., following ->private pointer in
* the case of Linux kernel), see cl_vmpage_page();
*
*
* Linux Kernel implementation.
*
- * Binding between cl_page and cfs_page_t (which is a typedef for
+ * Binding between cl_page and struct page (which is a typedef for
* struct page) is implemented in the vvp layer. cl_page is attached to the
* ->private pointer of the struct page, together with the setting of
* PG_private bit in page->flags, and acquiring additional reference on the
};
/**
- * Fields are protected by the lock on cfs_page_t, except for atomics and
+ * Fields are protected by the lock on struct page, except for atomics and
* immutables.
*
* \invariant Data type invariants are in cl_page_invariant(). Basically:
/**
* Debug information, the task is owning the page.
*/
- cfs_task_t *cp_task;
+ struct task_struct *cp_task;
/**
* Owning IO request in cl_page_state::CPS_PAGEOUT and
* cl_page_state::CPS_PAGEIN states. This field is maintained only in
struct cl_req *cp_req;
/** List of references to this page, for debugging. */
struct lu_ref cp_reference;
- /** Link to an object, for debugging. */
- struct lu_ref_link *cp_obj_ref;
- /** Link to a queue, for debugging. */
- struct lu_ref_link *cp_queue_ref;
- /** Per-page flags from enum cl_page_flags. Protected by a VM lock. */
- unsigned cp_flags;
- /** Assigned if doing a sync_io */
- struct cl_sync_io *cp_sync_io;
+ /** Link to an object, for debugging. */
+ struct lu_ref_link cp_obj_ref;
+ /** Link to a queue, for debugging. */
+ struct lu_ref_link cp_queue_ref;
+ /** Per-page flags from enum cl_page_flags. Protected by a VM lock. */
+ unsigned cp_flags;
+ /** Assigned if doing a sync_io */
+ struct cl_sync_io *cp_sync_io;
};
/**
*/
struct cl_page_operations {
/**
- * cl_page<->cfs_page_t methods. Only one layer in the stack has to
+ * cl_page<->struct page methods. Only one layer in the stack has to
* implement these. Current code assumes that this functionality is
* provided by the topmost layer, see cl_page_disown0() as an example.
*/
/**
* \return the underlying VM page. Optional.
*/
- cfs_page_t *(*cpo_vmpage)(const struct lu_env *env,
+ struct page *(*cpo_vmpage)(const struct lu_env *env,
const struct cl_page_slice *slice);
/**
* Called when \a io acquires this page into the exclusive
void (*cpo_export)(const struct lu_env *env,
const struct cl_page_slice *slice, int uptodate);
/**
- * Unmaps page from the user space (if it is mapped).
- *
- * \see cl_page_unmap()
- * \see vvp_page_unmap()
- */
- int (*cpo_unmap)(const struct lu_env *env,
- const struct cl_page_slice *slice, struct cl_io *io);
- /**
* Checks whether underlying VM page is locked (in the suitable
* sense). Used for assertions.
*
*/
int (*cpo_make_ready)(const struct lu_env *env,
const struct cl_page_slice *slice);
- /**
- * Announce that this page is to be written out
- * opportunistically, that is, page is dirty, it is not
- * necessary to start write-out transfer right now, but
- * eventually page has to be written out.
- *
- * Main caller of this is the write path (see
- * vvp_io_commit_write()), using this method to build a
- * "transfer cache" from which large transfers are then
- * constructed by the req-formation engine.
- *
- * \todo XXX it would make sense to add page-age tracking
- * semantics here, and to oblige the req-formation engine to
- * send the page out not later than it is too old.
- *
- * \see cl_page_cache_add()
- */
- int (*cpo_cache_add)(const struct lu_env *env,
- const struct cl_page_slice *slice,
- struct cl_io *io);
} io[CRT_NR];
/**
* Tell transfer engine that only [to, from] part of a page should be
*/
#define CL_PAGE_DEBUG(mask, env, page, format, ...) \
do { \
- LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, mask, NULL); \
- \
if (cfs_cdebug_show(mask, DEBUG_SUBSYSTEM)) { \
+ LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, mask, NULL); \
cl_page_print(env, &msgdata, lu_cdebug_printer, page); \
CDEBUG(mask, format , ## __VA_ARGS__); \
} \
*/
#define CL_PAGE_HEADER(mask, env, page, format, ...) \
do { \
- LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, mask, NULL); \
- \
if (cfs_cdebug_show(mask, DEBUG_SUBSYSTEM)) { \
+ LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, mask, NULL); \
cl_page_header_print(env, &msgdata, lu_cdebug_printer, page); \
CDEBUG(mask, format , ## __VA_ARGS__); \
} \
*/
struct cl_lock_descr cll_descr;
/** Protected by cl_lock::cll_guard. */
- enum cl_lock_state cll_state;
- /** signals state changes. */
- cfs_waitq_t cll_wq;
- /**
- * Recursive lock, most fields in cl_lock{} are protected by this.
- *
- * Locking rules: this mutex is never held across network
- * communication, except when lock is being canceled.
- *
- * Lock ordering: a mutex of a sub-lock is taken first, then a mutex
- * on a top-lock. Other direction is implemented through a
- * try-lock-repeat loop. Mutices of unrelated locks can be taken only
- * by try-locking.
- *
- * \see osc_lock_enqueue_wait(), lov_lock_cancel(), lov_sublock_wait().
- */
+ enum cl_lock_state cll_state;
+ /** signals state changes. */
+ wait_queue_head_t cll_wq;
+ /**
+ * Recursive lock, most fields in cl_lock{} are protected by this.
+ *
+ * Locking rules: this mutex is never held across network
+ * communication, except when lock is being canceled.
+ *
+ * Lock ordering: a mutex of a sub-lock is taken first, then a mutex
+ * on a top-lock. Other direction is implemented through a
+ * try-lock-repeat loop. Mutices of unrelated locks can be taken only
+ * by try-locking.
+ *
+ * \see osc_lock_enqueue_wait(), lov_lock_cancel(), lov_sublock_wait().
+ */
struct mutex cll_guard;
- cfs_task_t *cll_guarder;
+ struct task_struct *cll_guarder;
int cll_depth;
/**
* the owner for INTRANSIT state
*/
- cfs_task_t *cll_intransit_owner;
+ struct task_struct *cll_intransit_owner;
int cll_error;
/**
* Number of holds on a lock. A hold prevents a lock from being
* A list of holds on this lock, for debugging.
*/
struct lu_ref cll_holders;
- /**
- * A reference for cl_lock::cll_descr::cld_obj. For debugging.
- */
- struct lu_ref_link *cll_obj_ref;
+ /**
+ * A reference for cl_lock::cll_descr::cld_obj. For debugging.
+ */
+ struct lu_ref_link cll_obj_ref;
#ifdef CONFIG_LOCKDEP
- /* "dep_map" name is assumed by lockdep.h macros. */
- struct lockdep_map dep_map;
+ /* "dep_map" name is assumed by lockdep.h macros. */
+ struct lockdep_map dep_map;
#endif
};
#define CL_LOCK_DEBUG(mask, env, lock, format, ...) \
do { \
- LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, mask, NULL); \
- \
if (cfs_cdebug_show(mask, DEBUG_SUBSYSTEM)) { \
+ LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, mask, NULL); \
cl_lock_print(env, &msgdata, lu_cdebug_printer, lock); \
CDEBUG(mask, format , ## __VA_ARGS__); \
} \
* @{
*/
struct cl_page_list {
- unsigned pl_nr;
- cfs_list_t pl_pages;
- cfs_task_t *pl_owner;
+ unsigned pl_nr;
+ cfs_list_t pl_pages;
+ struct task_struct *pl_owner;
};
/**
cfs_list_t cis_linkage;
};
+typedef void (*cl_commit_cbt)(const struct lu_env *, struct cl_io *,
+ struct cl_page *);
/**
* Per-layer io operations.
void (*cio_fini) (const struct lu_env *env,
const struct cl_io_slice *slice);
} op[CIT_OP_NR];
- struct {
- /**
- * Submit pages from \a queue->c2_qin for IO, and move
- * successfully submitted pages into \a queue->c2_qout. Return
- * non-zero if failed to submit even the single page. If
- * submission failed after some pages were moved into \a
- * queue->c2_qout, completion callback with non-zero ioret is
- * executed on them.
- */
- int (*cio_submit)(const struct lu_env *env,
- const struct cl_io_slice *slice,
- enum cl_req_type crt,
- struct cl_2queue *queue);
- } req_op[CRT_NR];
+
+ /**
+ * Submit pages from \a queue->c2_qin for IO, and move
+ * successfully submitted pages into \a queue->c2_qout. Return
+ * non-zero if failed to submit even the single page. If
+ * submission failed after some pages were moved into \a
+ * queue->c2_qout, completion callback with non-zero ioret is
+ * executed on them.
+ */
+ int (*cio_submit)(const struct lu_env *env,
+ const struct cl_io_slice *slice,
+ enum cl_req_type crt,
+ struct cl_2queue *queue);
+ /**
+ * Queue async page for write.
+ * The difference between cio_submit and cio_queue is that
+ * cio_submit is for urgent request.
+ */
+ int (*cio_commit_async)(const struct lu_env *env,
+ const struct cl_io_slice *slice,
+ struct cl_page_list *queue, int from, int to,
+ cl_commit_cbt cb);
/**
* Read missing page.
*
const struct cl_io_slice *slice,
const struct cl_page_slice *page);
/**
- * Prepare write of a \a page. Called bottom-to-top by a top-level
- * cl_io_operations::op[CIT_WRITE]::cio_start() to prepare page for
- * get data from user-level buffer.
- *
- * \pre io->ci_type == CIT_WRITE
- *
- * \see vvp_io_prepare_write(), lov_io_prepare_write(),
- * osc_io_prepare_write().
- */
- int (*cio_prepare_write)(const struct lu_env *env,
- const struct cl_io_slice *slice,
- const struct cl_page_slice *page,
- unsigned from, unsigned to);
- /**
- *
- * \pre io->ci_type == CIT_WRITE
- *
- * \see vvp_io_commit_write(), lov_io_commit_write(),
- * osc_io_commit_write().
- */
- int (*cio_commit_write)(const struct lu_env *env,
- const struct cl_io_slice *slice,
- const struct cl_page_slice *page,
- unsigned from, unsigned to);
- /**
* Optional debugging helper. Print given io slice.
*/
int (*cio_print)(const struct lu_env *env, void *cookie,
* Right now, only two opertaions need to verify layout: glimpse
* and setattr.
*/
- ci_verify_layout:1;
- /**
- * Number of pages owned by this IO. For invariant checking.
- */
- unsigned ci_owned_nr;
+ ci_verify_layout:1,
+ /**
+ * file is released, restore has to to be triggered by vvp layer
+ */
+ ci_restore_needed:1,
+ /**
+ * O_NOATIME
+ */
+ ci_noatime:1;
+ /**
+ * Number of pages owned by this IO. For invariant checking.
+ */
+ unsigned ci_owned_nr;
};
/** @} cl_io */
* A per-object state that (potentially multi-object) transfer request keeps.
*/
struct cl_req_obj {
- /** object itself */
- struct cl_object *ro_obj;
- /** reference to cl_req_obj::ro_obj. For debugging. */
- struct lu_ref_link *ro_obj_ref;
- /* something else? Number of pages for a given object? */
+ /** object itself */
+ struct cl_object *ro_obj;
+ /** reference to cl_req_obj::ro_obj. For debugging. */
+ struct lu_ref_link ro_obj_ref;
+ /* something else? Number of pages for a given object? */
};
/**
return (void *)((char *)page + clob->co_slice_off);
}
+/**
+ * Return refcount of cl_object.
+ */
+static inline int cl_object_refc(struct cl_object *clob)
+{
+ struct lu_object_header *header = clob->co_lu.lo_header;
+ return cfs_atomic_read(&header->loh_ref);
+}
+
/** @} cl_object */
/** \defgroup cl_page cl_page
CLP_GANG_AGAIN,
CLP_GANG_ABORT
};
-
/* callback of cl_page_gang_lookup() */
-typedef int (*cl_page_gang_cb_t) (const struct lu_env *, struct cl_io *,
- struct cl_page *, void *);
-int cl_page_gang_lookup (const struct lu_env *env,
- struct cl_object *obj,
- struct cl_io *io,
- pgoff_t start, pgoff_t end,
- cl_page_gang_cb_t cb, void *cbdata);
-struct cl_page *cl_page_lookup (struct cl_object_header *hdr,
- pgoff_t index);
+
struct cl_page *cl_page_find (const struct lu_env *env,
struct cl_object *obj,
pgoff_t idx, struct page *vmpage,
enum cl_page_type type);
-struct cl_page *cl_page_find_sub (const struct lu_env *env,
- struct cl_object *obj,
- pgoff_t idx, struct page *vmpage,
- struct cl_page *parent);
+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);
void cl_page_get (struct cl_page *page);
void cl_page_put (const struct lu_env *env,
struct cl_page *page);
void cl_page_header_print(const struct lu_env *env, void *cookie,
lu_printer_t printer,
const struct cl_page *pg);
-cfs_page_t *cl_page_vmpage (const struct lu_env *env,
+struct page *cl_page_vmpage (const struct lu_env *env,
struct cl_page *page);
-struct cl_page *cl_vmpage_page (cfs_page_t *vmpage, struct cl_object *obj);
+struct cl_page *cl_vmpage_page (struct page *vmpage, struct cl_object *obj);
struct cl_page *cl_page_top (struct cl_page *page);
const struct cl_page_slice *cl_page_at(const struct cl_page *page,
void cl_page_discard (const struct lu_env *env, struct cl_io *io,
struct cl_page *pg);
void cl_page_delete (const struct lu_env *env, struct cl_page *pg);
-int cl_page_unmap (const struct lu_env *env, struct cl_io *io,
- struct cl_page *pg);
int cl_page_is_vmlocked (const struct lu_env *env,
const struct cl_page *pg);
void cl_page_export (const struct lu_env *env,
struct cl_lock_descr *descr);
int cl_io_read_page (const struct lu_env *env, struct cl_io *io,
struct cl_page *page);
-int cl_io_prepare_write(const struct lu_env *env, struct cl_io *io,
- struct cl_page *page, unsigned from, unsigned to);
-int cl_io_commit_write (const struct lu_env *env, struct cl_io *io,
- struct cl_page *page, unsigned from, unsigned to);
int cl_io_submit_rw (const struct lu_env *env, struct cl_io *io,
enum cl_req_type iot, struct cl_2queue *queue);
int cl_io_submit_sync (const struct lu_env *env, struct cl_io *io,
enum cl_req_type iot, struct cl_2queue *queue,
long timeout);
+int cl_io_commit_async (const struct lu_env *env, struct cl_io *io,
+ struct cl_page_list *queue, int from, int to,
+ cl_commit_cbt cb);
void cl_io_rw_advance (const struct lu_env *env, struct cl_io *io,
size_t nob);
int cl_io_cancel (const struct lu_env *env, struct cl_io *io,
return cfs_list_entry(plist->pl_pages.prev, struct cl_page, cp_batch);
}
+static inline struct cl_page *cl_page_list_first(struct cl_page_list *plist)
+{
+ LASSERT(plist->pl_nr > 0);
+ return cfs_list_entry(plist->pl_pages.next, struct cl_page, cp_batch);
+}
+
/**
* Iterate over pages in a page list.
*/
void cl_page_list_add (struct cl_page_list *plist, struct cl_page *page);
void cl_page_list_move (struct cl_page_list *dst, struct cl_page_list *src,
struct cl_page *page);
+void cl_page_list_move_head(struct cl_page_list *dst, struct cl_page_list *src,
+ struct cl_page *page);
void cl_page_list_splice (struct cl_page_list *list,
struct cl_page_list *head);
void cl_page_list_del (const struct lu_env *env,
struct cl_io *io, struct cl_page_list *plist);
void cl_page_list_discard(const struct lu_env *env,
struct cl_io *io, struct cl_page_list *plist);
-int cl_page_list_unmap (const struct lu_env *env,
- struct cl_io *io, struct cl_page_list *plist);
void cl_page_list_fini (const struct lu_env *env, struct cl_page_list *plist);
void cl_2queue_init (struct cl_2queue *queue);
/** barrier of destroy this structure */
cfs_atomic_t csi_barrier;
/** completion to be signaled when transfer is complete. */
- cfs_waitq_t csi_waitq;
+ wait_queue_head_t csi_waitq;
};
void cl_sync_io_init(struct cl_sync_io *anchor, int nrpages);
void cl_env_reexit (void *cookie);
void cl_env_implant (struct lu_env *env, int *refcheck);
void cl_env_unplant (struct lu_env *env, int *refcheck);
+unsigned cl_env_cache_purge(unsigned nr);
+struct lu_env *cl_env_percpu_get (void);
+void cl_env_percpu_put (struct lu_env *env);
/** @} cl_env */
git://git.whamcloud.com / fs / lustre-release.git / blobdiff