[dragonfly.git] / sys / vm / vm_pager.c

/*
 * (MPSAFE)
 *
 * Copyright (c) 1991, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * The Mach Operating System project at Carnegie-Mellon University.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	from: @(#)vm_pager.c	8.6 (Berkeley) 1/12/94
 *
 *
 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
 * All rights reserved.
 *
 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
 *
 * Permission to use, copy, modify and distribute this software and
 * its documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 *
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 *
 * Carnegie Mellon requests users of this software to return to
 *
 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 *
 * any improvements or extensions that they make and grant Carnegie the
 * rights to redistribute these changes.
 *
 * $FreeBSD: src/sys/vm/vm_pager.c,v 1.54.2.2 2001/11/18 07:11:00 dillon Exp $
 */

/*
 *	Paging space routine stubs.  Emulates a matchmaker-like interface
 *	for builtin pagers.
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/vnode.h>
#include <sys/buf.h>
#include <sys/ucred.h>
#include <sys/malloc.h>
#include <sys/dsched.h>
#include <sys/proc.h>
#include <sys/sysctl.h>
#include <sys/thread2.h>

#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/vm_kern.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/vm_pager.h>
#include <vm/vm_extern.h>

#include <sys/buf2.h>

MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data");

extern struct pagerops defaultpagerops;
extern struct pagerops swappagerops;
extern struct pagerops vnodepagerops;
extern struct pagerops devicepagerops;
extern struct pagerops physpagerops;

int cluster_pbuf_freecnt = -1;	/* unlimited to begin with */

static int dead_pager_getpage (vm_object_t, vm_page_t *, int);
static void dead_pager_putpages (vm_object_t, vm_page_t *, int, int, int *);
static boolean_t dead_pager_haspage (vm_object_t, vm_pindex_t);
static void dead_pager_dealloc (vm_object_t);

/*
 * No requirements.
 */
static int
dead_pager_getpage(vm_object_t obj, vm_page_t *mpp, int seqaccess)
{
	return VM_PAGER_FAIL;
}

/*
 * No requirements.
 */
static void
dead_pager_putpages(vm_object_t object, vm_page_t *m, int count, int flags,
		    int *rtvals)
{
	int i;

	for (i = 0; i < count; i++) {
		rtvals[i] = VM_PAGER_AGAIN;
	}
}

/*
 * No requirements.
 */
static int
dead_pager_haspage(vm_object_t object, vm_pindex_t pindex)
{
	return FALSE;
}

/*
 * No requirements.
 */
static void
dead_pager_dealloc(vm_object_t object)
{
	KKASSERT(object->swblock_count == 0);
	return;
}

static struct pagerops deadpagerops = {
	dead_pager_dealloc,
	dead_pager_getpage,
	dead_pager_putpages,
	dead_pager_haspage
};

struct pagerops *pagertab[] = {
	&defaultpagerops,	/* OBJT_DEFAULT */
	&swappagerops,		/* OBJT_SWAP */
	&vnodepagerops,		/* OBJT_VNODE */
	&devicepagerops,	/* OBJT_DEVICE */
	&physpagerops,		/* OBJT_PHYS */
	&deadpagerops		/* OBJT_DEAD */
};

int npagers = NELEM(pagertab);

/*
 * Kernel address space for mapping pages.
 * Used by pagers where KVAs are needed for IO.
 *
 * XXX needs to be large enough to support the number of pending async
 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
 * (MAXPHYS == 64k) if you want to get the most efficiency.
 */
#define PAGER_MAP_SIZE	(8 * 1024 * 1024)

TAILQ_HEAD(swqueue, buf);

int pager_map_size = PAGER_MAP_SIZE;
struct vm_map pager_map;

static int bswneeded_raw;
static int bswneeded_kva;
static int nswbuf_raw;
static struct buf *swbuf_raw;
static vm_offset_t swapbkva;		/* swap buffers kva */
static struct swqueue bswlist_raw;	/* without kva */
static struct swqueue bswlist_kva;	/* with kva */
static struct spinlock bswspin = SPINLOCK_INITIALIZER(&bswspin);
static int pbuf_raw_count;
static int pbuf_kva_count;

SYSCTL_INT(_vfs, OID_AUTO, pbuf_raw_count, CTLFLAG_RD, &pbuf_raw_count, 0,
    "Kernel virtual address space reservations");
SYSCTL_INT(_vfs, OID_AUTO, pbuf_kva_count, CTLFLAG_RD, &pbuf_kva_count, 0,
    "Kernel raw address space reservations");

/*
 * Initialize the swap buffer list.
 *
 * Called from the low level boot code only.
 */
static void
vm_pager_init(void *arg __unused)
{
	TAILQ_INIT(&bswlist_raw);
	TAILQ_INIT(&bswlist_kva);
}
SYSINIT(vm_mem, SI_BOOT1_VM, SI_ORDER_SECOND, vm_pager_init, NULL)

/*
 * Called from the low level boot code only.
 */
void
vm_pager_bufferinit(void)
{
	struct buf *bp;
	int i;

	/*
	 * Reserve KVM space for pbuf data.
	 */
	swapbkva = kmem_alloc_pageable(&pager_map, nswbuf * MAXPHYS);
	if (!swapbkva)
		panic("Not enough pager_map VM space for physical buffers");

	/*
	 * Initial pbuf setup.  These pbufs have KVA reservations.
	 */
	bp = swbuf;
	for (i = 0; i < nswbuf; ++i, ++bp) {
		bp->b_kvabase = (caddr_t)((intptr_t)i * MAXPHYS) + swapbkva;
		bp->b_kvasize = MAXPHYS;
		BUF_LOCKINIT(bp);
		buf_dep_init(bp);
		TAILQ_INSERT_HEAD(&bswlist_kva, bp, b_freelist);
		++pbuf_kva_count;
	}

	/*
	 * Initial pbuf setup.  These pbufs do not have KVA reservations,
	 * so we can have a lot more of them.  These are typically used
	 * to massage low level buf/bio requests.
	 */
	nswbuf_raw = nbuf * 2;
	swbuf_raw = (void *)kmem_alloc(&kernel_map,
				round_page(nswbuf_raw * sizeof(struct buf)));
	bp = swbuf_raw;
	for (i = 0; i < nswbuf_raw; ++i, ++bp) {
		BUF_LOCKINIT(bp);
		buf_dep_init(bp);
		TAILQ_INSERT_HEAD(&bswlist_raw, bp, b_freelist);
		++pbuf_raw_count;
	}

	/*
	 * Allow the clustering code to use half of our pbufs.
	 */
	cluster_pbuf_freecnt = nswbuf / 2;
}

/*
 * No requirements.
 */
void
vm_pager_deallocate(vm_object_t object)
{
	(*pagertab[object->type]->pgo_dealloc) (object);
}

/*
 * vm_pager_get_pages() - inline, see vm/vm_pager.h
 * vm_pager_put_pages() - inline, see vm/vm_pager.h
 * vm_pager_has_page() - inline, see vm/vm_pager.h
 * vm_pager_page_inserted() - inline, see vm/vm_pager.h
 * vm_pager_page_removed() - inline, see vm/vm_pager.h
 */

#if 0
/*
 *	vm_pager_sync:
 *
 *	Called by pageout daemon before going back to sleep.
 *	Gives pagers a chance to clean up any completed async pageing 
 *	operations.
 */
void
vm_pager_sync(void)
{
	struct pagerops **pgops;

	for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
		if (pgops && ((*pgops)->pgo_sync != NULL))
			(*(*pgops)->pgo_sync) ();
}

#endif

/*
 * Initialize a physical buffer.
 *
 * No requirements.
 */
static void
initpbuf(struct buf *bp)
{
	bp->b_qindex = 0;		/* BQUEUE_NONE */
	bp->b_data = bp->b_kvabase;	/* NULL if pbuf sans kva */
	bp->b_flags = B_PAGING;
	bp->b_cmd = BUF_CMD_DONE;
	bp->b_error = 0;
	bp->b_bcount = 0;
	bp->b_bufsize = MAXPHYS;
	initbufbio(bp);
	xio_init(&bp->b_xio);
	BUF_LOCK(bp, LK_EXCLUSIVE);
}

/*
 * Allocate a physical buffer
 *
 *	There are a limited number (nswbuf) of physical buffers.  We need
 *	to make sure that no single subsystem is able to hog all of them,
 *	so each subsystem implements a counter which is typically initialized
 *	to 1/2 nswbuf.  getpbuf() decrements this counter in allocation and
 *	increments it on release, and blocks if the counter hits zero.  A
 *	subsystem may initialize the counter to -1 to disable the feature,
 *	but it must still be sure to match up all uses of getpbuf() with 
 *	relpbuf() using the same variable.
 *
 *	NOTE: pfreecnt can be NULL, but this 'feature' will be removed
 *	relatively soon when the rest of the subsystems get smart about it. XXX
 *
 *	Physical buffers can be with or without KVA space reserved.  There
 *	are severe limitations on the ones with KVA reserved, and fewer
 *	limitations on the ones without.  getpbuf() gets one without,
 *	getpbuf_kva() gets one with.
 *
 * No requirements.
 */
struct buf *
getpbuf(int *pfreecnt)
{
	struct buf *bp;

	spin_lock(&bswspin);

	for (;;) {
		if (pfreecnt) {
			while (*pfreecnt == 0)
				ssleep(pfreecnt, &bswspin, 0, "wswbuf0", 0);
		}

		/* get a bp from the swap buffer header pool */
		if ((bp = TAILQ_FIRST(&bswlist_raw)) != NULL)
			break;
		bswneeded_raw = 1;
		ssleep(&bswneeded_raw, &bswspin, 0, "wswbuf1", 0);
		/* loop in case someone else grabbed one */
	}
	TAILQ_REMOVE(&bswlist_raw, bp, b_freelist);
	--pbuf_raw_count;
	if (pfreecnt)
		--*pfreecnt;

	spin_unlock(&bswspin);

	initpbuf(bp);
	KKASSERT(dsched_is_clear_buf_priv(bp));

	return (bp);
}

struct buf *
getpbuf_kva(int *pfreecnt)
{
	struct buf *bp;

	spin_lock(&bswspin);

	for (;;) {
		if (pfreecnt) {
			while (*pfreecnt == 0)
				ssleep(pfreecnt, &bswspin, 0, "wswbuf0", 0);
		}

		/* get a bp from the swap buffer header pool */
		if ((bp = TAILQ_FIRST(&bswlist_kva)) != NULL)
			break;
		bswneeded_kva = 1;
		ssleep(&bswneeded_kva, &bswspin, 0, "wswbuf1", 0);
		/* loop in case someone else grabbed one */
	}
	TAILQ_REMOVE(&bswlist_kva, bp, b_freelist);
	--pbuf_kva_count;
	if (pfreecnt)
		--*pfreecnt;

	spin_unlock(&bswspin);

	initpbuf(bp);
	KKASSERT(dsched_is_clear_buf_priv(bp));

	return (bp);
}

/*
 * Allocate a physical buffer, if one is available.
 *
 *	Note that there is no NULL hack here - all subsystems using this
 *	call understand how to use pfreecnt.
 *
 * No requirements.
 */
struct buf *
trypbuf(int *pfreecnt)
{
	struct buf *bp;

	spin_lock(&bswspin);

	if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist_raw)) == NULL) {
		spin_unlock(&bswspin);
		return NULL;
	}
	TAILQ_REMOVE(&bswlist_raw, bp, b_freelist);
	--pbuf_raw_count;
	--*pfreecnt;

	spin_unlock(&bswspin);

	initpbuf(bp);

	return bp;
}

struct buf *
trypbuf_kva(int *pfreecnt)
{
	struct buf *bp;

	spin_lock(&bswspin);

	if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist_kva)) == NULL) {
		spin_unlock(&bswspin);
		return NULL;
	}
	TAILQ_REMOVE(&bswlist_kva, bp, b_freelist);
	--pbuf_kva_count;
	--*pfreecnt;

	spin_unlock(&bswspin);

	initpbuf(bp);

	return bp;
}

/*
 * Release a physical buffer
 *
 *	NOTE: pfreecnt can be NULL, but this 'feature' will be removed
 *	relatively soon when the rest of the subsystems get smart about it. XXX
 *
 * No requirements.
 */
void
relpbuf(struct buf *bp, int *pfreecnt)
{
	int wake_bsw_kva = 0;
	int wake_bsw_raw = 0;
	int wake_freecnt = 0;

	KKASSERT(bp->b_flags & B_PAGING);
	dsched_exit_buf(bp);

	BUF_UNLOCK(bp);

	spin_lock(&bswspin);
	if (bp->b_kvabase) {
		TAILQ_INSERT_HEAD(&bswlist_kva, bp, b_freelist);
		++pbuf_kva_count;
	} else {
		TAILQ_INSERT_HEAD(&bswlist_raw, bp, b_freelist);
		++pbuf_raw_count;
	}
	if (bswneeded_kva) {
		bswneeded_kva = 0;
		wake_bsw_kva = 1;
	}
	if (bswneeded_raw) {
		bswneeded_raw = 0;
		wake_bsw_raw = 1;
	}
	if (pfreecnt) {
		if (++*pfreecnt == 1)
			wake_freecnt = 1;
	}
	spin_unlock(&bswspin);

	if (wake_bsw_kva)
		wakeup(&bswneeded_kva);
	if (wake_bsw_raw)
		wakeup(&bswneeded_raw);
	if (wake_freecnt)
		wakeup(pfreecnt);
}

void
pbuf_adjcount(int *pfreecnt, int n)
{
	if (n) {
		spin_lock(&bswspin);
		*pfreecnt += n;
		spin_unlock(&bswspin);
		wakeup(pfreecnt);
	}
}
Commit	Line	Data
984263bc	1	/*
99ad9bc4 MD	2	* (MPSAFE)
99ad9bc4 MD	3	*
984263bc MD	4	* Copyright (c) 1991, 1993
	5	* The Regents of the University of California. All rights reserved.
	6	*
	7	* This code is derived from software contributed to Berkeley by
	8	* The Mach Operating System project at Carnegie-Mellon University.
	9	*
	10	* Redistribution and use in source and binary forms, with or without
	11	* modification, are permitted provided that the following conditions
	12	* are met:
	13	* 1. Redistributions of source code must retain the above copyright
	14	* notice, this list of conditions and the following disclaimer.
	15	* 2. Redistributions in binary form must reproduce the above copyright
	16	* notice, this list of conditions and the following disclaimer in the
	17	* documentation and/or other materials provided with the distribution.
	18	* 3. All advertising materials mentioning features or use of this software
	19	* must display the following acknowledgement:
	20	* This product includes software developed by the University of
	21	* California, Berkeley and its contributors.
	22	* 4. Neither the name of the University nor the names of its contributors
	23	* may be used to endorse or promote products derived from this software
	24	* without specific prior written permission.
	25	*
	26	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
	27	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	28	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	29	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
	30	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	31	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	32	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	33	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	34	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	35	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	36	* SUCH DAMAGE.
	37	*
	38	* from: @(#)vm_pager.c 8.6 (Berkeley) 1/12/94
	39	*
	40	*
	41	* Copyright (c) 1987, 1990 Carnegie-Mellon University.
	42	* All rights reserved.
	43	*
	44	* Authors: Avadis Tevanian, Jr., Michael Wayne Young
	45	*
	46	* Permission to use, copy, modify and distribute this software and
	47	* its documentation is hereby granted, provided that both the copyright
	48	* notice and this permission notice appear in all copies of the
	49	* software, derivative works or modified versions, and any portions
	50	* thereof, and that both notices appear in supporting documentation.
	51	*
	52	* CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
	53	* CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
	54	* FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
	55	*
	56	* Carnegie Mellon requests users of this software to return to
	57	*
	58	* Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
	59	* School of Computer Science
	60	* Carnegie Mellon University
	61	* Pittsburgh PA 15213-3890
	62	*
	63	* any improvements or extensions that they make and grant Carnegie the
	64	* rights to redistribute these changes.
	65	*
	66	* $FreeBSD: src/sys/vm/vm_pager.c,v 1.54.2.2 2001/11/18 07:11:00 dillon Exp $
	67	*/
68
69	/*
70	* Paging space routine stubs. Emulates a matchmaker-like interface
71	* for builtin pagers.
72	*/
73
74	#include <sys/param.h>
75	#include <sys/systm.h>
76	#include <sys/kernel.h>
77	#include <sys/vnode.h>
78	#include <sys/buf.h>
79	#include <sys/ucred.h>
80	#include <sys/malloc.h>
8c72e3d5	81	#include <sys/dsched.h>
984263bc	82	#include <sys/proc.h>
9a82e536	83	#include <sys/sysctl.h>
cdd46d2e	84	#include <sys/thread2.h>
984263bc MD	85
	86	#include <vm/vm.h>
	87	#include <vm/vm_param.h>
9a82e536	88	#include <vm/vm_kern.h>
984263bc MD	89	#include <vm/vm_object.h>
	90	#include <vm/vm_page.h>
	91	#include <vm/vm_pager.h>
	92	#include <vm/vm_extern.h>
	93
3020e3be MD	94	#include <sys/buf2.h>
3020e3be MD	95
984263bc MD	96	MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data");
	97
	98	extern struct pagerops defaultpagerops;
	99	extern struct pagerops swappagerops;
	100	extern struct pagerops vnodepagerops;
	101	extern struct pagerops devicepagerops;
	102	extern struct pagerops physpagerops;
	103
	104	int cluster_pbuf_freecnt = -1; /* unlimited to begin with */
	105
1b9d3514	106	static int dead_pager_getpage (vm_object_t, vm_page_t *, int);
1388df65	107	static void dead_pager_putpages (vm_object_t, vm_page_t , int, int, int );
1b9d3514	108	static boolean_t dead_pager_haspage (vm_object_t, vm_pindex_t);
1388df65	109	static void dead_pager_dealloc (vm_object_t);
984263bc	110
99ad9bc4 MD	111	/*
	112	* No requirements.
	113	*/
984263bc	114	static int
1b9d3514	115	dead_pager_getpage(vm_object_t obj, vm_page_t *mpp, int seqaccess)
984263bc MD	116	{
	117	return VM_PAGER_FAIL;
	118	}
	119
99ad9bc4 MD	120	/*
	121	* No requirements.
	122	*/
984263bc	123	static void
57e43348	124	dead_pager_putpages(vm_object_t object, vm_page_t *m, int count, int flags,
5a648714	125	int *rtvals)
984263bc MD	126	{
	127	int i;
	128
	129	for (i = 0; i < count; i++) {
	130	rtvals[i] = VM_PAGER_AGAIN;
	131	}
	132	}
	133
99ad9bc4 MD	134	/*
	135	* No requirements.
	136	*/
984263bc	137	static int
1b9d3514	138	dead_pager_haspage(vm_object_t object, vm_pindex_t pindex)
984263bc	139	{
984263bc MD	140	return FALSE;
	141	}
	142
99ad9bc4 MD	143	/*
	144	* No requirements.
	145	*/
984263bc	146	static void
57e43348	147	dead_pager_dealloc(vm_object_t object)
984263bc	148	{
8d292090	149	KKASSERT(object->swblock_count == 0);
984263bc MD	150	return;
	151	}
	152
	153	static struct pagerops deadpagerops = {
984263bc	154	dead_pager_dealloc,
1b9d3514	155	dead_pager_getpage,
984263bc	156	dead_pager_putpages,
107e9bcc	157	dead_pager_haspage
984263bc MD	158	};
	159
	160	struct pagerops *pagertab[] = {
	161	&defaultpagerops, /* OBJT_DEFAULT */
	162	&swappagerops, /* OBJT_SWAP */
	163	&vnodepagerops, /* OBJT_VNODE */
	164	&devicepagerops, /* OBJT_DEVICE */
	165	&physpagerops, /* OBJT_PHYS */
	166	&deadpagerops /* OBJT_DEAD */
	167	};
	168
c157ff7a	169	int npagers = NELEM(pagertab);
984263bc MD	170
	171	/*
	172	* Kernel address space for mapping pages.
	173	* Used by pagers where KVAs are needed for IO.
	174	*
	175	* XXX needs to be large enough to support the number of pending async
	176	* cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
	177	* (MAXPHYS == 64k) if you want to get the most efficiency.
	178	*/
	179	#define PAGER_MAP_SIZE (8 * 1024 * 1024)
	180
9a82e536 MD	181	TAILQ_HEAD(swqueue, buf);
9a82e536 MD	182
984263bc	183	int pager_map_size = PAGER_MAP_SIZE;
e4846942 MD	184	struct vm_map pager_map;
e4846942 MD	185
9a82e536 MD	186	static int bswneeded_raw;
	187	static int bswneeded_kva;
	188	static int nswbuf_raw;
	189	static struct buf *swbuf_raw;
984263bc	190	static vm_offset_t swapbkva; /* swap buffers kva */
9a82e536 MD	191	static struct swqueue bswlist_raw; /* without kva */
9a82e536 MD	192	static struct swqueue bswlist_kva; /* with kva */
a50fe973	193	static struct spinlock bswspin = SPINLOCK_INITIALIZER(&bswspin);
9a82e536 MD	194	static int pbuf_raw_count;
	195	static int pbuf_kva_count;
	196
093e85dc SG	197	SYSCTL_INT(_vfs, OID_AUTO, pbuf_raw_count, CTLFLAG_RD, &pbuf_raw_count, 0,
	198	"Kernel virtual address space reservations");
	199	SYSCTL_INT(_vfs, OID_AUTO, pbuf_kva_count, CTLFLAG_RD, &pbuf_kva_count, 0,
	200	"Kernel raw address space reservations");
984263bc	201
99ad9bc4 MD	202	/*
	203	* Initialize the swap buffer list.
	204	*
	205	* Called from the low level boot code only.
	206	*/
107e9bcc MD	207	static void
107e9bcc MD	208	vm_pager_init(void *arg __unused)
984263bc	209	{
9a82e536 MD	210	TAILQ_INIT(&bswlist_raw);
9a82e536 MD	211	TAILQ_INIT(&bswlist_kva);
984263bc	212	}
107e9bcc	213	SYSINIT(vm_mem, SI_BOOT1_VM, SI_ORDER_SECOND, vm_pager_init, NULL)
984263bc	214
99ad9bc4 MD	215	/*
	216	* Called from the low level boot code only.
	217	*/
984263bc	218	void
57e43348	219	vm_pager_bufferinit(void)
984263bc MD	220	{
	221	struct buf *bp;
	222	int i;
	223
984263bc	224	/*
312dcd01 MD	225	* Reserve KVM space for pbuf data.
312dcd01 MD	226	*/
e4846942	227	swapbkva = kmem_alloc_pageable(&pager_map, nswbuf * MAXPHYS);
312dcd01 MD	228	if (!swapbkva)
	229	panic("Not enough pager_map VM space for physical buffers");
	230
	231	/*
9a82e536	232	* Initial pbuf setup. These pbufs have KVA reservations.
984263bc	233	*/
312dcd01 MD	234	bp = swbuf;
312dcd01 MD	235	for (i = 0; i < nswbuf; ++i, ++bp) {
973c11b9	236	bp->b_kvabase = (caddr_t)((intptr_t)i * MAXPHYS) + swapbkva;
312dcd01	237	bp->b_kvasize = MAXPHYS;
984263bc	238	BUF_LOCKINIT(bp);
408357d8	239	buf_dep_init(bp);
9a82e536 MD	240	TAILQ_INSERT_HEAD(&bswlist_kva, bp, b_freelist);
	241	++pbuf_kva_count;
	242	}
	243
	244	/*
	245	* Initial pbuf setup. These pbufs do not have KVA reservations,
	246	* so we can have a lot more of them. These are typically used
	247	* to massage low level buf/bio requests.
	248	*/
	249	nswbuf_raw = nbuf * 2;
	250	swbuf_raw = (void *)kmem_alloc(&kernel_map,
	251	round_page(nswbuf_raw * sizeof(struct buf)));
	252	bp = swbuf_raw;
	253	for (i = 0; i < nswbuf_raw; ++i, ++bp) {
	254	BUF_LOCKINIT(bp);
	255	buf_dep_init(bp);
	256	TAILQ_INSERT_HEAD(&bswlist_raw, bp, b_freelist);
	257	++pbuf_raw_count;
984263bc MD	258	}
984263bc MD	259
312dcd01 MD	260	/*
	261	* Allow the clustering code to use half of our pbufs.
	262	*/
984263bc	263	cluster_pbuf_freecnt = nswbuf / 2;
984263bc MD	264	}
984263bc MD	265
99ad9bc4 MD	266	/*
	267	* No requirements.
	268	*/
984263bc	269	void
57e43348	270	vm_pager_deallocate(vm_object_t object)
984263bc MD	271	{
	272	(*pagertab[object->type]->pgo_dealloc) (object);
	273	}
	274
	275	/*
984263bc MD	276	* vm_pager_get_pages() - inline, see vm/vm_pager.h
	277	* vm_pager_put_pages() - inline, see vm/vm_pager.h
	278	* vm_pager_has_page() - inline, see vm/vm_pager.h
	279	* vm_pager_page_inserted() - inline, see vm/vm_pager.h
	280	* vm_pager_page_removed() - inline, see vm/vm_pager.h
	281	*/
	282
	283	#if 0
	284	/*
	285	* vm_pager_sync:
	286	*
	287	* Called by pageout daemon before going back to sleep.
	288	* Gives pagers a chance to clean up any completed async pageing
	289	* operations.
	290	*/
	291	void
57e43348	292	vm_pager_sync(void)
984263bc MD	293	{
	294	struct pagerops **pgops;
	295
	296	for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
	297	if (pgops && ((*pgops)->pgo_sync != NULL))
	298	((pgops)->pgo_sync) ();
	299	}
	300
	301	#endif
	302
984263bc	303	/*
312dcd01	304	* Initialize a physical buffer.
99ad9bc4 MD	305	*
99ad9bc4 MD	306	* No requirements.
984263bc	307	*/
984263bc MD	308	static void
	309	initpbuf(struct buf *bp)
	310	{
9a82e536 MD	311	bp->b_qindex = 0; /* BQUEUE_NONE */
9a82e536 MD	312	bp->b_data = bp->b_kvabase; /* NULL if pbuf sans kva */
a8f169e2	313	bp->b_flags = B_PAGING;
10f3fee5	314	bp->b_cmd = BUF_CMD_DONE;
984263bc	315	bp->b_error = 0;
9a71d53f MD	316	bp->b_bcount = 0;
9a71d53f MD	317	bp->b_bufsize = MAXPHYS;
81b5c339	318	initbufbio(bp);
54f51aeb	319	xio_init(&bp->b_xio);
984263bc MD	320	BUF_LOCK(bp, LK_EXCLUSIVE);
	321	}
	322
	323	/*
a50fe973	324	* Allocate a physical buffer
984263bc MD	325	*
	326	* There are a limited number (nswbuf) of physical buffers. We need
	327	* to make sure that no single subsystem is able to hog all of them,
	328	* so each subsystem implements a counter which is typically initialized
	329	* to 1/2 nswbuf. getpbuf() decrements this counter in allocation and
	330	* increments it on release, and blocks if the counter hits zero. A
	331	* subsystem may initialize the counter to -1 to disable the feature,
	332	* but it must still be sure to match up all uses of getpbuf() with
	333	* relpbuf() using the same variable.
	334	*
	335	* NOTE: pfreecnt can be NULL, but this 'feature' will be removed
	336	* relatively soon when the rest of the subsystems get smart about it. XXX
a50fe973	337	*
9a82e536 MD	338	* Physical buffers can be with or without KVA space reserved. There
	339	* are severe limitations on the ones with KVA reserved, and fewer
	340	* limitations on the ones without. getpbuf() gets one without,
	341	* getpbuf_kva() gets one with.
	342	*
99ad9bc4	343	* No requirements.
984263bc MD	344	*/
984263bc MD	345	struct buf *
57e43348	346	getpbuf(int *pfreecnt)
984263bc	347	{
984263bc MD	348	struct buf *bp;
984263bc MD	349
287a8577	350	spin_lock(&bswspin);
984263bc MD	351
	352	for (;;) {
	353	if (pfreecnt) {
a50fe973	354	while (*pfreecnt == 0)
e590ee86	355	ssleep(pfreecnt, &bswspin, 0, "wswbuf0", 0);
984263bc MD	356	}
	357
	358	/* get a bp from the swap buffer header pool */
9a82e536	359	if ((bp = TAILQ_FIRST(&bswlist_raw)) != NULL)
984263bc	360	break;
9a82e536 MD	361	bswneeded_raw = 1;
9a82e536 MD	362	ssleep(&bswneeded_raw, &bswspin, 0, "wswbuf1", 0);
984263bc MD	363	/* loop in case someone else grabbed one */
984263bc MD	364	}
9a82e536 MD	365	TAILQ_REMOVE(&bswlist_raw, bp, b_freelist);
9a82e536 MD	366	--pbuf_raw_count;
984263bc MD	367	if (pfreecnt)
984263bc MD	368	--*pfreecnt;
a50fe973	369
287a8577	370	spin_unlock(&bswspin);
984263bc MD	371
984263bc MD	372	initpbuf(bp);
8c72e3d5	373	KKASSERT(dsched_is_clear_buf_priv(bp));
9a82e536 MD	374
	375	return (bp);
	376	}
	377
	378	struct buf *
	379	getpbuf_kva(int *pfreecnt)
	380	{
	381	struct buf *bp;
	382
287a8577	383	spin_lock(&bswspin);
9a82e536 MD	384
	385	for (;;) {
	386	if (pfreecnt) {
	387	while (*pfreecnt == 0)
	388	ssleep(pfreecnt, &bswspin, 0, "wswbuf0", 0);
	389	}
	390
	391	/* get a bp from the swap buffer header pool */
	392	if ((bp = TAILQ_FIRST(&bswlist_kva)) != NULL)
	393	break;
	394	bswneeded_kva = 1;
	395	ssleep(&bswneeded_kva, &bswspin, 0, "wswbuf1", 0);
	396	/* loop in case someone else grabbed one */
	397	}
	398	TAILQ_REMOVE(&bswlist_kva, bp, b_freelist);
	399	--pbuf_kva_count;
	400	if (pfreecnt)
	401	--*pfreecnt;
	402
287a8577	403	spin_unlock(&bswspin);
9a82e536 MD	404
	405	initpbuf(bp);
	406	KKASSERT(dsched_is_clear_buf_priv(bp));
	407
	408	return (bp);
984263bc MD	409	}
	410
	411	/*
a50fe973	412	* Allocate a physical buffer, if one is available.
984263bc MD	413	*
	414	* Note that there is no NULL hack here - all subsystems using this
	415	* call understand how to use pfreecnt.
a50fe973	416	*
99ad9bc4	417	* No requirements.
984263bc MD	418	*/
984263bc MD	419	struct buf *
57e43348	420	trypbuf(int *pfreecnt)
984263bc	421	{
984263bc MD	422	struct buf *bp;
984263bc MD	423
287a8577	424	spin_lock(&bswspin);
a50fe973	425
9a82e536	426	if (*pfreecnt == 0 \|\| (bp = TAILQ_FIRST(&bswlist_raw)) == NULL) {
287a8577	427	spin_unlock(&bswspin);
9a82e536 MD	428	return NULL;
	429	}
	430	TAILQ_REMOVE(&bswlist_raw, bp, b_freelist);
	431	--pbuf_raw_count;
	432	--*pfreecnt;
	433
287a8577	434	spin_unlock(&bswspin);
9a82e536 MD	435
	436	initpbuf(bp);
	437
	438	return bp;
	439	}
	440
	441	struct buf *
	442	trypbuf_kva(int *pfreecnt)
	443	{
	444	struct buf *bp;
	445
287a8577	446	spin_lock(&bswspin);
9a82e536 MD	447
9a82e536 MD	448	if (*pfreecnt == 0 \|\| (bp = TAILQ_FIRST(&bswlist_kva)) == NULL) {
287a8577	449	spin_unlock(&bswspin);
984263bc MD	450	return NULL;
984263bc MD	451	}
9a82e536 MD	452	TAILQ_REMOVE(&bswlist_kva, bp, b_freelist);
9a82e536 MD	453	--pbuf_kva_count;
984263bc MD	454	--*pfreecnt;
984263bc MD	455
287a8577	456	spin_unlock(&bswspin);
984263bc MD	457
	458	initpbuf(bp);
	459
	460	return bp;
	461	}
	462
	463	/*
a50fe973	464	* Release a physical buffer
984263bc MD	465	*
	466	* NOTE: pfreecnt can be NULL, but this 'feature' will be removed
	467	* relatively soon when the rest of the subsystems get smart about it. XXX
a50fe973	468	*
99ad9bc4	469	* No requirements.
984263bc MD	470	*/
984263bc MD	471	void
57e43348	472	relpbuf(struct buf bp, int pfreecnt)
984263bc	473	{
9a82e536 MD	474	int wake_bsw_kva = 0;
9a82e536 MD	475	int wake_bsw_raw = 0;
a50fe973	476	int wake_freecnt = 0;
984263bc	477
a8f169e2	478	KKASSERT(bp->b_flags & B_PAGING);
aa166ad1	479	dsched_exit_buf(bp);
984263bc	480
a50fe973	481	BUF_UNLOCK(bp);
7757d1c2	482
287a8577	483	spin_lock(&bswspin);
9a82e536 MD	484	if (bp->b_kvabase) {
	485	TAILQ_INSERT_HEAD(&bswlist_kva, bp, b_freelist);
	486	++pbuf_kva_count;
	487	} else {
	488	TAILQ_INSERT_HEAD(&bswlist_raw, bp, b_freelist);
	489	++pbuf_raw_count;
	490	}
	491	if (bswneeded_kva) {
	492	bswneeded_kva = 0;
	493	wake_bsw_kva = 1;
	494	}
	495	if (bswneeded_raw) {
	496	bswneeded_raw = 0;
	497	wake_bsw_raw = 1;
984263bc MD	498	}
	499	if (pfreecnt) {
	500	if (++*pfreecnt == 1)
a50fe973	501	wake_freecnt = 1;
984263bc	502	}
287a8577	503	spin_unlock(&bswspin);
a50fe973	504
9a82e536 MD	505	if (wake_bsw_kva)
	506	wakeup(&bswneeded_kva);
	507	if (wake_bsw_raw)
	508	wakeup(&bswneeded_raw);
a50fe973 MD	509	if (wake_freecnt)
a50fe973 MD	510	wakeup(pfreecnt);
984263bc	511	}
7f86d367 MD	512
	513	void
	514	pbuf_adjcount(int *pfreecnt, int n)
	515	{
	516	if (n) {
	517	spin_lock(&bswspin);
	518	*pfreecnt += n;
	519	spin_unlock(&bswspin);
	520	wakeup(pfreecnt);
	521	}
	522	}