/*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
*
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
- * The contents of this file constitute Original Code as defined in and
- * are subject to the Apple Public Source License Version 1.1 (the
- * "License"). You may not use this file except in compliance with the
- * License. Please obtain a copy of the License at
- * http://www.apple.com/publicsource and read it before using this file.
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
*
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ *
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
- * License for the specific language governing rights and limitations
- * under the License.
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
*
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
/*
* @OSF_COPYRIGHT@
#ifndef _VM_VM_PAGE_H_
#define _VM_VM_PAGE_H_
+#include <debug.h>
+
#include <mach/boolean.h>
#include <mach/vm_prot.h>
#include <mach/vm_param.h>
#include <kern/lock.h>
#include <kern/macro_help.h>
+#include <libkern/OSAtomic.h>
+
/*
- * Each page entered on the inactive queue obtains a ticket from a
- * particular ticket roll. Pages granted tickets from a particular
- * roll generally flow through the queue as a group. In this way when a
- * page with a ticket from a particular roll is pulled from the top of the
- * queue it is extremely likely that the pages near the top will have tickets
- * from the same or adjacent rolls. In this way the proximity to the top
- * of the queue can be loosely ascertained by determining the identity of
- * the roll the pages ticket came from.
+ * VM_PAGE_MIN_SPECULATIVE_AGE_Q through VM_PAGE_MAX_SPECULATIVE_AGE_Q
+ * represents a set of aging bins that are 'protected'...
+ *
+ * VM_PAGE_SPECULATIVE_AGED_Q is a list of the speculative pages that have
+ * not yet been 'claimed' but have been aged out of the protective bins
+ * this occurs in vm_page_speculate when it advances to the next bin
+ * and discovers that it is still occupied... at that point, all of the
+ * pages in that bin are moved to the VM_PAGE_SPECULATIVE_AGED_Q. the pages
+ * in that bin are all guaranteed to have reached at least the maximum age
+ * we allow for a protected page... they can be older if there is no
+ * memory pressure to pull them from the bin, or there are no new speculative pages
+ * being generated to push them out.
+ * this list is the one that vm_pageout_scan will prefer when looking
+ * for pages to move to the underweight free list
+ *
+ * VM_PAGE_MAX_SPECULATIVE_AGE_Q * VM_PAGE_SPECULATIVE_Q_AGE_MS
+ * defines the amount of time a speculative page is normally
+ * allowed to live in the 'protected' state (i.e. not available
+ * to be stolen if vm_pageout_scan is running and looking for
+ * pages)... however, if the total number of speculative pages
+ * in the protected state exceeds our limit (defined in vm_pageout.c)
+ * and there are none available in VM_PAGE_SPECULATIVE_AGED_Q, then
+ * vm_pageout_scan is allowed to steal pages from the protected
+ * bucket even if they are underage.
+ *
+ * vm_pageout_scan is also allowed to pull pages from a protected
+ * bin if the bin has reached the "age of consent" we've set
*/
+#define VM_PAGE_MAX_SPECULATIVE_AGE_Q 10
+#define VM_PAGE_MIN_SPECULATIVE_AGE_Q 1
+#define VM_PAGE_SPECULATIVE_AGED_Q 0
+
+#define VM_PAGE_SPECULATIVE_Q_AGE_MS 500
+
+struct vm_speculative_age_q {
+ /*
+ * memory queue for speculative pages via clustered pageins
+ */
+ queue_head_t age_q;
+ mach_timespec_t age_ts;
+};
-extern int vm_page_ticket_roll;
-extern int vm_page_ticket;
+extern
+struct vm_speculative_age_q vm_page_queue_speculative[];
+
+extern int speculative_steal_index;
+extern int speculative_age_index;
+extern unsigned int vm_page_speculative_q_age_ms;
-#define VM_PAGE_TICKETS_IN_ROLL 512
-#define VM_PAGE_TICKET_ROLL_IDS 16
/*
* Management of resident (logical) pages.
*/
struct vm_page {
- queue_chain_t pageq; /* queue info for FIFO
- * queue or free list (P) */
+ queue_chain_t pageq; /* queue info for FIFO */
+ /* queue or free list (P) */
+
queue_chain_t listq; /* all pages in same object (O) */
struct vm_page *next; /* VP bucket link (O) */
vm_object_t object; /* which object am I in (O&P) */
vm_object_offset_t offset; /* offset into that object (O,P) */
+ /*
+ * The following word of flags is protected
+ * by the "page queues" lock.
+ *
+ * we use the 'wire_count' field to store the local
+ * queue id if local queues are enabled...
+ * see the comments at 'VM_PAGE_QUEUES_REMOVE' as to
+ * why this is safe to do
+ */
+#define local_id wire_count
unsigned int wire_count:16, /* how many wired down maps use me? (O&P) */
- page_ticket:4, /* age of the page on the */
- /* inactive queue. */
/* boolean_t */ inactive:1, /* page is in inactive list (P) */
+ zero_fill:1,
active:1, /* page is in active list (P) */
+ pageout_queue:1,/* page is on queue for pageout (P) */
+ speculative:1, /* page is on speculative list (P) */
laundry:1, /* page is being cleaned now (P)*/
free:1, /* page is on free list (P) */
reference:1, /* page has been used (P) */
- pageout:1, /* page wired & busy for pageout (P) */
gobbled:1, /* page used internally (P) */
private:1, /* Page should not be returned to
- * the free list (O) */
- zero_fill:1,
- :0;
+ * the free list (P) */
+ throttled:1, /* pager is not responding (P) */
+ local:1,
+ no_cache:1, /* page is not to be cached and should
+ * be reused ahead of other pages (P) */
+ __unused_pageq_bits:3; /* 3 bits available here */
+
+ ppnum_t phys_page; /* Physical address of page, passed
+ * to pmap_enter (read-only) */
+ /*
+ * The following word of flags is protected
+ * by the "VM object" lock.
+ */
unsigned int
- page_error:8, /* error from I/O operations */
/* boolean_t */ busy:1, /* page is in transit (O) */
wanted:1, /* someone is waiting for page (O) */
tabled:1, /* page is in VP table (O) */
fictitious:1, /* Physical page doesn't exist (O) */
- no_isync:1, /* page has not been instruction synced */
+ /*
+ * IMPORTANT: the "pmapped" bit can be turned on while holding the
+ * VM object "shared" lock. See vm_fault_enter().
+ * This is OK as long as it's the only bit in this bit field that
+ * can be updated without holding the VM object "exclusive" lock.
+ */
+ pmapped:1, /* page has been entered at some
+ * point into a pmap (O **shared**) */
+ wpmapped:1, /* page has been entered at some
+ * point into a pmap for write (O) */
+ pageout:1, /* page wired & busy for pageout (O) */
absent:1, /* Data has been requested, but is
* not yet available (O) */
error:1, /* Data manager was unable to provide
restart:1, /* Page was pushed higher in shadow
chain by copy_call-related pagers;
start again at top of chain */
- lock_supplied:1,/* protection supplied by pager (O) */
- /* vm_prot_t */ page_lock:3, /* Uses prohibited by pager (O) */
- /* vm_prot_t */ unlock_request:3,/* Outstanding unlock request (O) */
unusual:1, /* Page is absent, error, restart or
page locked */
- discard_request:1,/* a memory_object_discard_request()
- * has been sent */
+ encrypted:1, /* encrypted for secure swap (O) */
+ encrypted_cleaning:1, /* encrypting page */
list_req_pending:1, /* pagein/pageout alt mechanism */
/* allows creation of list */
/* requests on pages that are */
/* actively being paged. */
- dump_cleaning:1; /* set by the pageout daemon when */
+ dump_cleaning:1, /* set by the pageout daemon when */
/* a page being cleaned is */
/* encountered and targeted as */
/* a pageout candidate */
- /* we've used up all 32 bits */
-
- vm_offset_t phys_addr; /* Physical address of page, passed
- * to pmap_enter (read-only) */
+ cs_validated:1, /* code-signing: page was checked */
+ cs_tainted:1, /* code-signing: page is tainted */
+ reusable:1,
+ lopage:1,
+ slid:1,
+ __unused_object_bits:7; /* 7 bits available here */
+
+#if __LP64__
+ unsigned int __unused_padding; /* Pad structure explicitly
+ * to 8-byte multiple for LP64 */
+#endif
};
+#define DEBUG_ENCRYPTED_SWAP 1
+#if DEBUG_ENCRYPTED_SWAP
+#define ASSERT_PAGE_DECRYPTED(page) \
+ MACRO_BEGIN \
+ if ((page)->encrypted) { \
+ panic("VM page %p should not be encrypted here\n", \
+ (page)); \
+ } \
+ MACRO_END
+#else /* DEBUG_ENCRYPTED_SWAP */
+#define ASSERT_PAGE_DECRYPTED(page) assert(!(page)->encrypted)
+#endif /* DEBUG_ENCRYPTED_SWAP */
+
typedef struct vm_page *vm_page_t;
+
+typedef struct vm_locks_array {
+ char pad __attribute__ ((aligned (64)));
+ lck_mtx_t vm_page_queue_lock2 __attribute__ ((aligned (64)));
+ lck_mtx_t vm_page_queue_free_lock2 __attribute__ ((aligned (64)));
+ char pad2 __attribute__ ((aligned (64)));
+} vm_locks_array_t;
+
+
+#define VM_PAGE_WIRED(m) ((!(m)->local && (m)->wire_count))
#define VM_PAGE_NULL ((vm_page_t) 0)
-#define NEXT_PAGE(m) ((vm_page_t) (m)->pageq.next)
+#define NEXT_PAGE(m) ((vm_page_t) (m)->pageq.next)
+#define NEXT_PAGE_PTR(m) ((vm_page_t *) &(m)->pageq.next)
/*
* XXX The unusual bit should not be necessary. Most of the bit
* some useful check on a page structure.
*/
-#define VM_PAGE_CHECK(mem)
+#define VM_PAGE_CHECK(mem) \
+ MACRO_BEGIN \
+ VM_PAGE_QUEUES_ASSERT(mem, 1); \
+ MACRO_END
+
+/* Page coloring:
+ *
+ * The free page list is actually n lists, one per color,
+ * where the number of colors is a function of the machine's
+ * cache geometry set at system initialization. To disable
+ * coloring, set vm_colors to 1 and vm_color_mask to 0.
+ * The boot-arg "colors" may be used to override vm_colors.
+ * Note that there is little harm in having more colors than needed.
+ */
+
+#define MAX_COLORS 128
+#define DEFAULT_COLORS 32
+
+extern
+unsigned int vm_colors; /* must be in range 1..MAX_COLORS */
+extern
+unsigned int vm_color_mask; /* must be (vm_colors-1) */
+extern
+unsigned int vm_cache_geometry_colors; /* optimal #colors based on cache geometry */
+
+/*
+ * Wired memory is a very limited resource and we can't let users exhaust it
+ * and deadlock the entire system. We enforce the following limits:
+ *
+ * vm_user_wire_limit (default: all memory minus vm_global_no_user_wire_amount)
+ * how much memory can be user-wired in one user task
+ *
+ * vm_global_user_wire_limit (default: same as vm_user_wire_limit)
+ * how much memory can be user-wired in all user tasks
+ *
+ * vm_global_no_user_wire_amount (default: VM_NOT_USER_WIREABLE)
+ * how much memory must remain user-unwired at any time
+ */
+#define VM_NOT_USER_WIREABLE (64*1024*1024) /* 64MB */
+extern
+vm_map_size_t vm_user_wire_limit;
+extern
+vm_map_size_t vm_global_user_wire_limit;
+extern
+vm_map_size_t vm_global_no_user_wire_amount;
/*
* Each pageable resident page falls into one of three lists:
*
* free
- * Available for allocation now.
+ * Available for allocation now. The free list is
+ * actually an array of lists, one per color.
* inactive
* Not referenced in any map, but still has an
* object/offset-page mapping, and may be dirty.
* This is the list of pages that should be
- * paged out next.
+ * paged out next. There are actually two
+ * inactive lists, one for pages brought in from
+ * disk or other backing store, and another
+ * for "zero-filled" pages. See vm_pageout_scan()
+ * for the distinction and usage.
* active
* A list of pages which have been placed in
* at least one physical map. This list is
* ordered, in LRU-like fashion.
*/
+
+#define VPL_LOCK_SPIN 1
+
+struct vpl {
+ unsigned int vpl_count;
+ queue_head_t vpl_queue;
+#ifdef VPL_LOCK_SPIN
+ lck_spin_t vpl_lock;
+#else
+ lck_mtx_t vpl_lock;
+ lck_mtx_ext_t vpl_lock_ext;
+#endif
+};
+
+struct vplq {
+ union {
+ char cache_line_pad[128];
+ struct vpl vpl;
+ } vpl_un;
+};
extern
-vm_page_t vm_page_queue_free; /* memory free queue */
+unsigned int vm_page_local_q_count;
+extern
+struct vplq *vm_page_local_q;
+extern
+unsigned int vm_page_local_q_soft_limit;
+extern
+unsigned int vm_page_local_q_hard_limit;
+extern
+vm_locks_array_t vm_page_locks;
+
extern
-vm_page_t vm_page_queue_fictitious; /* fictitious free queue */
+queue_head_t vm_page_queue_free[MAX_COLORS]; /* memory free queue */
+extern
+queue_head_t vm_lopage_queue_free; /* low memory free queue */
extern
queue_head_t vm_page_queue_active; /* active memory queue */
extern
-queue_head_t vm_page_queue_inactive; /* inactive memory queue */
+queue_head_t vm_page_queue_inactive; /* inactive memory queue for normal pages */
+extern
queue_head_t vm_page_queue_zf; /* inactive memory queue for zero fill */
+extern
+queue_head_t vm_page_queue_throttled; /* memory queue for throttled pageout pages */
extern
vm_offset_t first_phys_addr; /* physical address for first_page */
vm_offset_t last_phys_addr; /* physical address for last_page */
extern
-int vm_page_free_count; /* How many pages are free? */
+unsigned int vm_page_free_count; /* How many pages are free? (sum of all colors) */
+extern
+unsigned int vm_page_fictitious_count;/* How many fictitious pages are free? */
+extern
+unsigned int vm_page_active_count; /* How many pages are active? */
+extern
+unsigned int vm_page_inactive_count; /* How many pages are inactive? */
+extern
+unsigned int vm_page_throttled_count;/* How many inactives are throttled */
+extern
+unsigned int vm_page_speculative_count; /* How many speculative pages are unclaimed? */
+extern
+unsigned int vm_page_wire_count; /* How many pages are wired? */
extern
-int vm_page_fictitious_count;/* How many fictitious pages are free? */
+unsigned int vm_page_free_target; /* How many do we want free? */
extern
-int vm_page_active_count; /* How many pages are active? */
+unsigned int vm_page_free_min; /* When to wakeup pageout */
extern
-int vm_page_inactive_count; /* How many pages are inactive? */
+unsigned int vm_page_throttle_limit; /* When to throttle new page creation */
extern
-int vm_page_wire_count; /* How many pages are wired? */
+uint32_t vm_page_creation_throttle; /* When to throttle new page creation */
extern
-int vm_page_free_target; /* How many do we want free? */
+unsigned int vm_page_inactive_target;/* How many do we want inactive? */
extern
-int vm_page_free_min; /* When to wakeup pageout */
+unsigned int vm_page_inactive_min; /* When do wakeup pageout */
extern
-int vm_page_inactive_target;/* How many do we want inactive? */
+unsigned int vm_page_free_reserved; /* How many pages reserved to do pageout */
extern
-int vm_page_free_reserved; /* How many pages reserved to do pageout */
+unsigned int vm_page_throttle_count; /* Count of page allocations throttled */
extern
-int vm_page_laundry_count; /* How many pages being laundered? */
+unsigned int vm_page_gobble_count;
-decl_mutex_data(,vm_page_queue_lock)
- /* lock on active and inactive page queues */
-decl_mutex_data(,vm_page_queue_free_lock)
- /* lock on free page queue */
-decl_simple_lock_data(extern,vm_page_preppin_lock) /* lock for prep/pin */
-decl_mutex_data(,vm_page_zero_fill_lock)
+#if DEVELOPMENT || DEBUG
+extern
+unsigned int vm_page_speculative_used;
+#endif
+
+extern
+unsigned int vm_page_purgeable_count;/* How many pages are purgeable now ? */
+extern
+unsigned int vm_page_purgeable_wired_count;/* How many purgeable pages are wired now ? */
+extern
+uint64_t vm_page_purged_count; /* How many pages got purged so far ? */
extern unsigned int vm_page_free_wanted;
/* how many threads are waiting for memory */
-extern vm_offset_t vm_page_fictitious_addr;
+extern unsigned int vm_page_free_wanted_privileged;
+ /* how many VM privileged threads are waiting for memory */
+
+extern ppnum_t vm_page_fictitious_addr;
/* (fake) phys_addr of fictitious pages */
+extern ppnum_t vm_page_guard_addr;
+ /* (fake) phys_addr of guard pages */
+
+
+extern boolean_t vm_page_deactivate_hint;
+
+/*
+ 0 = all pages avail ( default. )
+ 1 = disable high mem ( cap max pages to 4G)
+ 2 = prefer himem
+*/
+extern int vm_himemory_mode;
+
+extern boolean_t vm_lopage_needed;
+extern uint32_t vm_lopage_free_count;
+extern uint32_t vm_lopage_free_limit;
+extern uint32_t vm_lopage_lowater;
+extern boolean_t vm_lopage_refill;
+extern uint64_t max_valid_dma_address;
+extern ppnum_t max_valid_low_ppnum;
+
/*
* Prototypes for functions exported by this module.
*/
extern void vm_page_bootstrap(
vm_offset_t *startp,
- vm_offset_t *endp);
+ vm_offset_t *endp) __attribute__((section("__TEXT, initcode")));
-extern void vm_page_module_init(void);
+extern void vm_page_module_init(void) __attribute__((section("__TEXT, initcode")));
+
+extern void vm_page_init_local_q(void);
extern void vm_page_create(
- vm_offset_t start,
- vm_offset_t end);
+ ppnum_t start,
+ ppnum_t end);
extern vm_page_t vm_page_lookup(
vm_object_t object,
extern vm_page_t vm_page_grab_fictitious(void);
+extern vm_page_t vm_page_grab_guard(void);
+
extern void vm_page_release_fictitious(
vm_page_t page);
-extern boolean_t vm_page_convert(
- vm_page_t page);
-
extern void vm_page_more_fictitious(void);
extern int vm_pool_low(void);
extern vm_page_t vm_page_grab(void);
-extern void vm_page_release(
- vm_page_t page);
+extern vm_page_t vm_page_grablo(void);
-extern void vm_page_release_limbo(
+extern void vm_page_release(
vm_page_t page);
-extern void vm_page_limbo_exchange(
- vm_page_t limbo_m,
- vm_page_t new_m);
-
extern boolean_t vm_page_wait(
int interruptible );
vm_object_t object,
vm_object_offset_t offset);
+extern vm_page_t vm_page_alloclo(
+ vm_object_t object,
+ vm_object_offset_t offset);
+
+extern vm_page_t vm_page_alloc_guard(
+ vm_object_t object,
+ vm_object_offset_t offset);
+
extern void vm_page_init(
vm_page_t page,
- vm_offset_t phys_addr);
+ ppnum_t phys_page,
+ boolean_t lopage);
extern void vm_page_free(
- vm_page_t page);
+ vm_page_t page);
+
+extern void vm_page_free_unlocked(
+ vm_page_t page,
+ boolean_t remove_from_hash);
extern void vm_page_activate(
vm_page_t page);
extern void vm_page_deactivate(
vm_page_t page);
+extern void vm_page_deactivate_internal(
+ vm_page_t page,
+ boolean_t clear_hw_reference);
+
+extern void vm_page_lru(
+ vm_page_t page);
+
+extern void vm_page_speculate(
+ vm_page_t page,
+ boolean_t new);
+
+extern void vm_page_speculate_ageit(
+ struct vm_speculative_age_q *aq);
+
+extern void vm_page_reactivate_all_throttled(void);
+
+extern void vm_page_reactivate_local(uint32_t lid, boolean_t force, boolean_t nolocks);
+
extern void vm_page_rename(
vm_page_t page,
vm_object_t new_object,
- vm_object_offset_t new_offset);
+ vm_object_offset_t new_offset,
+ boolean_t encrypted_ok);
extern void vm_page_insert(
vm_page_t page,
vm_object_t object,
vm_object_offset_t offset);
+extern void vm_page_insert_internal(
+ vm_page_t page,
+ vm_object_t object,
+ vm_object_offset_t offset,
+ boolean_t queues_lock_held,
+ boolean_t insert_in_hash);
+
extern void vm_page_replace(
vm_page_t mem,
vm_object_t object,
vm_object_offset_t offset);
extern void vm_page_remove(
- vm_page_t page);
+ vm_page_t page,
+ boolean_t remove_from_hash);
extern void vm_page_zero_fill(
vm_page_t page);
vm_page_t page);
extern void vm_page_unwire(
- vm_page_t page);
+ vm_page_t page,
+ boolean_t queueit);
extern void vm_set_page_size(void);
extern void vm_page_gobble(
vm_page_t page);
+extern void vm_page_validate_cs(vm_page_t page);
+extern void vm_page_validate_cs_mapped(
+ vm_page_t page,
+ const void *kaddr);
+
+extern void vm_page_free_prepare_queues(
+ vm_page_t page);
+
+extern void vm_page_free_prepare_object(
+ vm_page_t page,
+ boolean_t remove_from_hash);
+
+extern void vm_check_memorystatus(void);
+
+
/*
* Functions implemented as macros. m->wanted and m->busy are
* protected by the object lock.
#define VM_PAGE_FREE(p) \
MACRO_BEGIN \
- vm_page_lock_queues(); \
- vm_page_free(p); \
- vm_page_unlock_queues(); \
+ vm_page_free_unlocked(p, TRUE); \
MACRO_END
#define VM_PAGE_GRAB_FICTITIOUS(M) \
vm_page_more_fictitious(); \
MACRO_END
-#define VM_PAGE_THROTTLED() \
- (vm_page_free_count < (vm_page_free_target - \
- ((vm_page_free_target-vm_page_free_min)>>2)))
-
#define VM_PAGE_WAIT() ((void)vm_page_wait(THREAD_UNINT))
-#define vm_page_lock_queues() mutex_lock(&vm_page_queue_lock)
-#define vm_page_unlock_queues() mutex_unlock(&vm_page_queue_lock)
-#define vm_page_pin_lock() simple_lock(&vm_page_preppin_lock)
-#define vm_page_pin_unlock() simple_unlock(&vm_page_preppin_lock)
+#define vm_page_queue_lock (vm_page_locks.vm_page_queue_lock2)
+#define vm_page_queue_free_lock (vm_page_locks.vm_page_queue_free_lock2)
+
+#define vm_page_lock_queues() lck_mtx_lock(&vm_page_queue_lock)
+#define vm_page_unlock_queues() lck_mtx_unlock(&vm_page_queue_lock)
+
+#define vm_page_lockspin_queues() lck_mtx_lock_spin(&vm_page_queue_lock)
+#define vm_page_trylockspin_queues() lck_mtx_try_lock_spin(&vm_page_queue_lock)
+#define vm_page_lockconvert_queues() lck_mtx_convert_spin(&vm_page_queue_lock)
+#ifdef VPL_LOCK_SPIN
+#define VPL_LOCK_INIT(vlq, vpl_grp, vpl_attr) lck_spin_init(&vlq->vpl_lock, vpl_grp, vpl_attr)
+#define VPL_LOCK(vpl) lck_spin_lock(vpl)
+#define VPL_UNLOCK(vpl) lck_spin_unlock(vpl)
+#else
+#define VPL_LOCK_INIT(vlq, vpl_grp, vpl_attr) lck_mtx_init_ext(&vlq->vpl_lock, &vlq->vpl_lock_ext, vpl_grp, vpl_attr)
+#define VPL_LOCK(vpl) lck_mtx_lock_spin(vpl)
+#define VPL_UNLOCK(vpl) lck_mtx_unlock(vpl)
+#endif
+
+#if MACH_ASSERT
+extern void vm_page_queues_assert(vm_page_t mem, int val);
+#define VM_PAGE_QUEUES_ASSERT(mem, val) vm_page_queues_assert((mem), (val))
+#else
+#define VM_PAGE_QUEUES_ASSERT(mem, val)
+#endif
+
+
+/*
+ * 'vm_fault_enter' will place newly created pages (zero-fill and COW) onto the
+ * local queues if they exist... its the only spot in the system where we add pages
+ * to those queues... once on those queues, those pages can only move to one of the
+ * global page queues or the free queues... they NEVER move from local q to local q.
+ * the 'local' state is stable when VM_PAGE_QUEUES_REMOVE is called since we're behind
+ * the global vm_page_queue_lock at this point... we still need to take the local lock
+ * in case this operation is being run on a different CPU then the local queue's identity,
+ * but we don't have to worry about the page moving to a global queue or becoming wired
+ * while we're grabbing the local lock since those operations would require the global
+ * vm_page_queue_lock to be held, and we already own it.
+ *
+ * this is why its safe to utilze the wire_count field in the vm_page_t as the local_id...
+ * 'wired' and local are ALWAYS mutually exclusive conditions.
+ */
#define VM_PAGE_QUEUES_REMOVE(mem) \
MACRO_BEGIN \
- if (mem->active) { \
- assert(!mem->inactive); \
+ VM_PAGE_QUEUES_ASSERT(mem, 1); \
+ assert(!mem->laundry); \
+ assert(!mem->pageout_queue); \
+ if (mem->local) { \
+ struct vpl *lq; \
+ assert(mem->object != kernel_object); \
+ assert(!mem->inactive && !mem->speculative); \
+ assert(!mem->active && !mem->throttled); \
+ assert(!mem->fictitious); \
+ lq = &vm_page_local_q[mem->local_id].vpl_un.vpl; \
+ VPL_LOCK(&lq->vpl_lock); \
+ queue_remove(&lq->vpl_queue, \
+ mem, vm_page_t, pageq); \
+ mem->local = FALSE; \
+ mem->local_id = 0; \
+ lq->vpl_count--; \
+ VPL_UNLOCK(&lq->vpl_lock); \
+ } \
+ \
+ else if (mem->active) { \
+ assert(mem->object != kernel_object); \
+ assert(!mem->inactive && !mem->speculative); \
+ assert(!mem->throttled); \
+ assert(!mem->fictitious); \
queue_remove(&vm_page_queue_active, \
mem, vm_page_t, pageq); \
mem->active = FALSE; \
- if (!mem->fictitious) \
- vm_page_active_count--; \
+ vm_page_active_count--; \
} \
\
- if (mem->inactive) { \
- assert(!mem->active); \
+ else if (mem->inactive) { \
+ assert(mem->object != kernel_object); \
+ assert(!mem->active && !mem->speculative); \
+ assert(!mem->throttled); \
+ assert(!mem->fictitious); \
if (mem->zero_fill) { \
queue_remove(&vm_page_queue_zf, \
mem, vm_page_t, pageq); \
+ vm_zf_queue_count--; \
} else { \
queue_remove(&vm_page_queue_inactive, \
mem, vm_page_t, pageq); \
} \
mem->inactive = FALSE; \
- if (!mem->fictitious) \
- vm_page_inactive_count--; \
+ vm_page_inactive_count--; \
+ vm_purgeable_q_advance_all(); \
+ } \
+ \
+ else if (mem->throttled) { \
+ assert(!mem->active && !mem->inactive); \
+ assert(!mem->speculative); \
+ assert(!mem->fictitious); \
+ queue_remove(&vm_page_queue_throttled, \
+ mem, vm_page_t, pageq); \
+ mem->throttled = FALSE; \
+ vm_page_throttled_count--; \
+ } \
+ \
+ else if (mem->speculative) { \
+ assert(!mem->active && !mem->inactive); \
+ assert(!mem->throttled); \
+ assert(!mem->fictitious); \
+ remque(&mem->pageq); \
+ mem->speculative = FALSE; \
+ vm_page_speculative_count--; \
+ } \
+ \
+ else if (mem->pageq.next || mem->pageq.prev) \
+ panic("VM_PAGE_QUEUES_REMOVE: unmarked page on Q"); \
+ mem->pageq.next = NULL; \
+ mem->pageq.prev = NULL; \
+ VM_PAGE_QUEUES_ASSERT(mem, 0); \
+ MACRO_END
+
+
+#define VM_PAGE_ENQUEUE_INACTIVE(mem, first) \
+ MACRO_BEGIN \
+ VM_PAGE_QUEUES_ASSERT(mem, 0); \
+ assert(!mem->fictitious); \
+ assert(!mem->laundry); \
+ assert(!mem->pageout_queue); \
+ if (mem->zero_fill) { \
+ if (first == TRUE) \
+ queue_enter_first(&vm_page_queue_zf, mem, vm_page_t, pageq); \
+ else \
+ queue_enter(&vm_page_queue_zf, mem, vm_page_t, pageq); \
+ vm_zf_queue_count++; \
+ } else { \
+ if (first == TRUE) \
+ queue_enter_first(&vm_page_queue_inactive, mem, vm_page_t, pageq); \
+ else \
+ queue_enter(&vm_page_queue_inactive, mem, vm_page_t, pageq); \
+ } \
+ mem->inactive = TRUE; \
+ vm_page_inactive_count++; \
+ token_new_pagecount++; \
+ MACRO_END
+
+
+#if DEVELOPMENT || DEBUG
+#define VM_PAGE_SPECULATIVE_USED_ADD() \
+ MACRO_BEGIN \
+ OSAddAtomic(1, &vm_page_speculative_used); \
+ MACRO_END
+#else
+#define VM_PAGE_SPECULATIVE_USED_ADD()
+#endif
+
+
+#define VM_PAGE_CONSUME_CLUSTERED(mem) \
+ MACRO_BEGIN \
+ if (mem->clustered) { \
+ assert(mem->object); \
+ mem->object->pages_used++; \
+ mem->clustered = FALSE; \
+ VM_PAGE_SPECULATIVE_USED_ADD(); \
} \
MACRO_END
+
+
+#define DW_vm_page_unwire 0x01
+#define DW_vm_page_wire 0x02
+#define DW_vm_page_free 0x04
+#define DW_vm_page_activate 0x08
+#define DW_vm_page_deactivate_internal 0x10
+#define DW_vm_page_speculate 0x20
+#define DW_vm_page_lru 0x40
+#define DW_vm_pageout_throttle_up 0x80
+#define DW_PAGE_WAKEUP 0x100
+#define DW_clear_busy 0x200
+#define DW_clear_reference 0x400
+#define DW_set_reference 0x800
+#define DW_move_page 0x1000
+#define DW_VM_PAGE_QUEUES_REMOVE 0x2000
+#define DW_set_list_req_pending 0x4000
+
+struct vm_page_delayed_work {
+ vm_page_t dw_m;
+ int dw_mask;
+};
+
+void vm_page_do_delayed_work(vm_object_t object, struct vm_page_delayed_work *dwp, int dw_count);
+
+extern unsigned int vm_max_delayed_work_limit;
+
+#define DEFAULT_DELAYED_WORK_LIMIT 32
+
+#define DELAYED_WORK_LIMIT(max) ((vm_max_delayed_work_limit >= max ? max : vm_max_delayed_work_limit))
+
+/*
+ * vm_page_do_delayed_work may need to drop the object lock...
+ * if it does, we need the pages it's looking at to
+ * be held stable via the busy bit, so if busy isn't already
+ * set, we need to set it and ask vm_page_do_delayed_work
+ * to clear it and wakeup anyone that might have blocked on
+ * it once we're done processing the page.
+ *
+ * additionally, we can't call vm_page_do_delayed_work with
+ * list_req_pending == TRUE since it may need to
+ * drop the object lock before dealing
+ * with this page and because list_req_pending == TRUE,
+ * busy == TRUE will NOT protect this page from being stolen
+ * so clear list_req_pending and ask vm_page_do_delayed_work
+ * to re-set it once it holds both the pageq and object locks
+ */
+
+#define VM_PAGE_ADD_DELAYED_WORK(dwp, mem, dw_cnt) \
+ MACRO_BEGIN \
+ if (mem->busy == FALSE) { \
+ mem->busy = TRUE; \
+ if ( !(dwp->dw_mask & DW_vm_page_free)) \
+ dwp->dw_mask |= (DW_clear_busy | DW_PAGE_WAKEUP); \
+ } \
+ if (mem->list_req_pending) { \
+ mem->list_req_pending = FALSE; \
+ dwp->dw_mask |= DW_set_list_req_pending; \
+ } \
+ dwp->dw_m = mem; \
+ dwp++; \
+ dw_count++; \
+ MACRO_END
+
+extern vm_page_t vm_object_page_grab(vm_object_t);
+
+
#endif /* _VM_VM_PAGE_H_ */
projects / apple / xnu.git / blobdiff