[apple/xnu.git] / bsd / vm / vnode_pager.c

/*
 * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_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 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.
 * 
 * @APPLE_LICENSE_HEADER_END@
 */
/* 
 * Mach Operating System
 * Copyright (c) 1987 Carnegie-Mellon University
 * All rights reserved.  The CMU software License Agreement specifies
 * the terms and conditions for use and redistribution.
 */
/*
 *	File:	vnode_pager.c
 *
 *	"Swap" pager that pages to/from vnodes.  Also
 *	handles demand paging from files.
 *
 */

#include <mach/boolean.h>
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/buf.h>
#include <sys/uio.h>
#include <sys/vnode.h>
#include <sys/namei.h>
#include <sys/mount.h>
#include <sys/ubc.h>
#include <sys/lock.h>

#include <mach/mach_types.h>
#include <mach/memory_object_types.h>

#include <vm/vm_map.h>
#include <vm/vm_kern.h>
#include <kern/parallel.h>
#include <kern/zalloc.h>
#include <kern/kalloc.h>
#include <libkern/libkern.h>

#include <vm/vnode_pager.h>
#include <vm/vm_pageout.h>

#include <kern/assert.h>

unsigned int vp_pagein=0;
unsigned int vp_pgodirty=0;
unsigned int vp_pgoclean=0;
unsigned int dp_pgouts=0;	/* Default pager pageouts */
unsigned int dp_pgins=0;	/* Default pager pageins */

vm_object_offset_t
vnode_pager_get_filesize(struct vnode *vp)
{
	if (UBCINVALID(vp)) {
		return (vm_object_offset_t) 0;
	}

	return (vm_object_offset_t) ubc_getsize(vp);
	
}

pager_return_t
vnode_pageout(struct vnode *vp,
	upl_t			upl,
	vm_offset_t		upl_offset,
	vm_object_offset_t	f_offset,
	vm_size_t		size,
	int			flags,
	int			*errorp)
{
	int		result = PAGER_SUCCESS;
	struct proc 	*p = current_proc();
	int		error = 0;
	int vp_size = 0;
	int blkno=0, s;
	int cnt, isize;
	int pg_index;
	int offset;
	struct buf *bp;
	boolean_t	funnel_state;
	int haveupl=0;
	upl_page_info_t *pl;
	upl_t vpupl = NULL;

	funnel_state = thread_funnel_set(kernel_flock, TRUE);

	if (upl != (upl_t)NULL) {
		haveupl = 1;
	}
	isize = (int)size;

	if (isize < 0)
		panic("-ve count in vnode_pageout");
	if (isize == 0)
		panic("vnode_pageout: size == 0\n");

	UBCINFOCHECK("vnode_pageout", vp);

	if (UBCINVALID(vp)) {
		result = PAGER_ERROR;
		error  = PAGER_ERROR;
		goto out;
	}
	if (haveupl) {
		/*
		 * This is a pageout form the Default pager,
		 * just go ahead and call VOP_PAGEOUT
		 */
		dp_pgouts++;
		if (error = VOP_PAGEOUT(vp, upl, upl_offset,
			 (off_t)f_offset,(size_t)size, p->p_ucred, flags)) {
			result = PAGER_ERROR;
			error  = PAGER_ERROR;
		}
		goto out;
	}
	ubc_create_upl( vp,
					f_offset,
					isize,
					&vpupl,
					&pl,
					UPL_COPYOUT_FROM);
	if (vpupl == (upl_t) 0)
		return PAGER_ABSENT;

	vp_size = ubc_getsize(vp);
	if (vp_size == 0) {

		while (isize) {
			blkno = ubc_offtoblk(vp, (off_t)f_offset);
start0:
			if (bp = incore(vp, blkno)) {
				if (ISSET(bp->b_flags, B_BUSY)) {
					SET(bp->b_flags, B_WANTED);
					error = tsleep(bp, (PRIBIO + 1), "vnpgout", 0);
					goto start0;
				} else {
				        bremfree(bp);
					SET(bp->b_flags, (B_BUSY|B_INVAL));
				}
			}
			if (bp)
				brelse(bp);
			f_offset += PAGE_SIZE;
			isize    -= PAGE_SIZE;
		}
		ubc_upl_commit_range(vpupl, 0, size, UPL_COMMIT_FREE_ON_EMPTY);

		error = 0;
		goto out;
	}
	pg_index = 0;
	offset   = 0;

	while (isize) {
		int  xsize;
		int  num_of_pages;

		if ( !upl_valid_page(pl, pg_index)) {
			ubc_upl_abort_range(vpupl, offset, PAGE_SIZE,
					UPL_ABORT_FREE_ON_EMPTY);
	         
			offset += PAGE_SIZE;
			isize  -= PAGE_SIZE;
			pg_index++;

			continue;
		}
		if ( !upl_dirty_page(pl, pg_index)) {
			/*
			 * if the page is not dirty and reached here it is
			 * marked precious or it is due to invalidation in
			 * memory_object_lock request as part of truncation
			 * We also get here from vm_object_terminate()
			 * So all you need to do in these
			 * cases is to invalidate incore buffer if it is there
			 */
			blkno = ubc_offtoblk(vp, (off_t)(f_offset + offset));
			s = splbio();
			vp_pgoclean++;			
start:
			if (bp = incore(vp, blkno)) {
				if (ISSET(bp->b_flags, B_BUSY)) {
					SET(bp->b_flags, B_WANTED);
					error = tsleep(bp, (PRIBIO + 1), "vnpgout", 0);
					goto start;
				} else {
				        bremfree(bp);
					SET(bp->b_flags, (B_BUSY|B_INVAL));
				}
			}
			splx(s);
			if (bp)
				brelse(bp);

			ubc_upl_commit_range(vpupl, offset, PAGE_SIZE, 
					UPL_COMMIT_FREE_ON_EMPTY);

			offset += PAGE_SIZE;
			isize  -= PAGE_SIZE;
			pg_index++;

			continue;
		}
		vp_pgodirty++;

		num_of_pages = 1;
		xsize = isize - PAGE_SIZE;

		while (xsize) {
			if ( !upl_valid_page(pl, pg_index + num_of_pages))
				break;
			if ( !upl_dirty_page(pl, pg_index + num_of_pages))
				break;
			num_of_pages++;
			xsize -= PAGE_SIZE;
		}
		xsize = num_of_pages * PAGE_SIZE;

		/*  By defn callee will commit or abort upls */
		if (error = VOP_PAGEOUT(vp, vpupl, (vm_offset_t) offset,
					(off_t)(f_offset + offset),
					xsize, p->p_ucred, flags & ~UPL_NOCOMMIT)) {
			result = PAGER_ERROR;
			error  = PAGER_ERROR;
		}
		offset += xsize;
		isize  -= xsize;
		pg_index += num_of_pages;
	}
out:
	if (errorp)
		*errorp = result;

	thread_funnel_set(kernel_flock, funnel_state);

	return (error);
}


pager_return_t
vnode_pagein(
	struct vnode 		*vp,
	upl_t        		 pl,
	vm_offset_t  		 pl_offset,
	vm_object_offset_t	f_offset,
	vm_size_t     		size,
	int           		flags,
	int 			*errorp)
{
	int	result = PAGER_SUCCESS;
	struct proc	*p = current_proc();
	int		error = 0;
	int		xfer_size;
	boolean_t	funnel_state;
	int haveupl=0;
	upl_t vpupl = NULL;
	off_t	local_offset;
	unsigned int  ioaddr;

	funnel_state = thread_funnel_set(kernel_flock, TRUE);

#if 0
	if(pl->page_list.npages >1 )
		panic("vnode_pageout: Can't handle more than one page");
#endif /* 0 */

	if (pl != (upl_t)NULL) {
		haveupl = 1;
	}
	UBCINFOCHECK("vnode_pagein", vp);

	if (UBCINVALID(vp)) {
		result = PAGER_ERROR;
		error  = PAGER_ERROR;
		goto out;
	}

	if (haveupl) {
		dp_pgins++;
		if (error = VOP_PAGEIN(vp, pl, pl_offset, (off_t)f_offset,
				 size,p->p_ucred, flags)) {
			result = PAGER_ERROR;
		}
	} else {

		local_offset = 0;
		while (size) {
			if((size > 4096) && (vp->v_tag == VT_NFS)) {
				xfer_size =  4096;
				size = size - xfer_size;
			} else {
				xfer_size = size;
				size = 0;
			}
			ubc_create_upl(	vp,
							f_offset+local_offset,
							xfer_size,
							&vpupl,
							NULL,
							UPL_FLAGS_NONE);
			if (vpupl == (upl_t) 0) {
				result =  PAGER_ABSENT;
				error = PAGER_ABSENT;
				goto out;
			}

			vp_pagein++;

			/*  By defn callee will commit or abort upls */
			if (error = VOP_PAGEIN(vp, vpupl, (vm_offset_t) 0,
				(off_t)f_offset+local_offset, xfer_size,p->p_ucred, flags & ~UPL_NOCOMMIT)) {
				result = PAGER_ERROR;
				error  = PAGER_ERROR;
			}
			local_offset += PAGE_SIZE_64;
		}
	}
out:
	if (errorp)
	    *errorp = result;
	thread_funnel_set(kernel_flock, funnel_state);

	return (error);
}

void
vnode_pager_shutdown()
{
	int i;
	extern struct bs_map  bs_port_table[];
	struct vnode *vp;

	for(i = 0; i < MAX_BACKING_STORE; i++) {
		vp = (struct vnode *)(bs_port_table[i]).vp;
		if (vp) {
			(bs_port_table[i]).vp = 0;
			ubc_rele(vp);
			/* get rid of macx_swapon() namei() reference */
			vrele(vp);

			/* get rid of macx_swapon() "extra" reference */
			vrele(vp);
		}
	}
}


void *
upl_get_internal_page_list(upl_t upl)
{
  return(UPL_GET_INTERNAL_PAGE_LIST(upl));

}
Commit	Line	Data
1c79356b A	1	/*
	2	* Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
	3	*
	4	* @APPLE_LICENSE_HEADER_START@
	5	*
	6	* The contents of this file constitute Original Code as defined in and
	7	* are subject to the Apple Public Source License Version 1.1 (the
	8	* "License"). You may not use this file except in compliance with the
	9	* License. Please obtain a copy of the License at
	10	* http://www.apple.com/publicsource and read it before using this file.
	11	*
	12	* This Original Code and all software distributed under the License are
	13	* distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
	14	* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
	15	* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
	16	* FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
	17	* License for the specific language governing rights and limitations
	18	* under the License.
	19	*
	20	* @APPLE_LICENSE_HEADER_END@
	21	*/
	22	/*
	23	* Mach Operating System
	24	* Copyright (c) 1987 Carnegie-Mellon University
	25	* All rights reserved. The CMU software License Agreement specifies
	26	* the terms and conditions for use and redistribution.
	27	*/
	28	/*
	29	* File: vnode_pager.c
	30	*
	31	* "Swap" pager that pages to/from vnodes. Also
	32	* handles demand paging from files.
	33	*
	34	*/
	35
	36	#include <mach/boolean.h>
	37	#include <sys/param.h>
	38	#include <sys/systm.h>
	39	#include <sys/proc.h>
	40	#include <sys/buf.h>
	41	#include <sys/uio.h>
	42	#include <sys/vnode.h>
	43	#include <sys/namei.h>
	44	#include <sys/mount.h>
	45	#include <sys/ubc.h>
	46	#include <sys/lock.h>
	47
	48	#include <mach/mach_types.h>
	49	#include <mach/memory_object_types.h>
	50
	51	#include <vm/vm_map.h>
	52	#include <vm/vm_kern.h>
	53	#include <kern/parallel.h>
	54	#include <kern/zalloc.h>
	55	#include <kern/kalloc.h>
	56	#include <libkern/libkern.h>
	57
	58	#include <vm/vnode_pager.h>
	59	#include <vm/vm_pageout.h>
	60
	61	#include <kern/assert.h>
	62
	63	unsigned int vp_pagein=0;
	64	unsigned int vp_pgodirty=0;
65	unsigned int vp_pgoclean=0;
66	unsigned int dp_pgouts=0; /* Default pager pageouts */
67	unsigned int dp_pgins=0; /* Default pager pageins */
68
0b4e3aa0 A	69	vm_object_offset_t
	70	vnode_pager_get_filesize(struct vnode *vp)
	71	{
	72	if (UBCINVALID(vp)) {
	73	return (vm_object_offset_t) 0;
	74	}
	75
	76	return (vm_object_offset_t) ubc_getsize(vp);
	77
	78	}
	79
1c79356b A	80	pager_return_t
	81	vnode_pageout(struct vnode *vp,
	82	upl_t upl,
	83	vm_offset_t upl_offset,
	84	vm_object_offset_t f_offset,
	85	vm_size_t size,
	86	int flags,
	87	int *errorp)
	88	{
	89	int result = PAGER_SUCCESS;
	90	struct proc *p = current_proc();
	91	int error = 0;
	92	int vp_size = 0;
	93	int blkno=0, s;
	94	int cnt, isize;
	95	int pg_index;
	96	int offset;
	97	struct buf *bp;
	98	boolean_t funnel_state;
	99	int haveupl=0;
1c79356b	100	upl_page_info_t *pl;
0b4e3aa0	101	upl_t vpupl = NULL;
1c79356b A	102
	103	funnel_state = thread_funnel_set(kernel_flock, TRUE);
	104
	105	if (upl != (upl_t)NULL) {
	106	haveupl = 1;
	107	}
	108	isize = (int)size;
	109
	110	if (isize < 0)
	111	panic("-ve count in vnode_pageout");
	112	if (isize == 0)
	113	panic("vnode_pageout: size == 0\n");
	114
	115	UBCINFOCHECK("vnode_pageout", vp);
	116
	117	if (UBCINVALID(vp)) {
	118	result = PAGER_ERROR;
	119	error = PAGER_ERROR;
	120	goto out;
	121	}
	122	if (haveupl) {
	123	/*
	124	* This is a pageout form the Default pager,
	125	* just go ahead and call VOP_PAGEOUT
	126	*/
	127	dp_pgouts++;
	128	if (error = VOP_PAGEOUT(vp, upl, upl_offset,
	129	(off_t)f_offset,(size_t)size, p->p_ucred, flags)) {
	130	result = PAGER_ERROR;
	131	error = PAGER_ERROR;
	132	}
	133	goto out;
	134	}
0b4e3aa0 A	135	ubc_create_upl( vp,
	136	f_offset,
	137	isize,
	138	&vpupl,
	139	&pl,
	140	UPL_COPYOUT_FROM);
1c79356b	141	if (vpupl == (upl_t) 0)
0b4e3aa0	142	return PAGER_ABSENT;
1c79356b	143
0b4e3aa0	144	vp_size = ubc_getsize(vp);
1c79356b A	145	if (vp_size == 0) {
	146
	147	while (isize) {
	148	blkno = ubc_offtoblk(vp, (off_t)f_offset);
	149	start0:
	150	if (bp = incore(vp, blkno)) {
	151	if (ISSET(bp->b_flags, B_BUSY)) {
	152	SET(bp->b_flags, B_WANTED);
	153	error = tsleep(bp, (PRIBIO + 1), "vnpgout", 0);
	154	goto start0;
	155	} else {
	156	bremfree(bp);
	157	SET(bp->b_flags, (B_BUSY\|B_INVAL));
	158	}
	159	}
	160	if (bp)
	161	brelse(bp);
	162	f_offset += PAGE_SIZE;
	163	isize -= PAGE_SIZE;
	164	}
0b4e3aa0	165	ubc_upl_commit_range(vpupl, 0, size, UPL_COMMIT_FREE_ON_EMPTY);
1c79356b A	166
	167	error = 0;
	168	goto out;
	169	}
	170	pg_index = 0;
	171	offset = 0;
	172
	173	while (isize) {
	174	int xsize;
	175	int num_of_pages;
	176
	177	if ( !upl_valid_page(pl, pg_index)) {
0b4e3aa0 A	178	ubc_upl_abort_range(vpupl, offset, PAGE_SIZE,
0b4e3aa0 A	179	UPL_ABORT_FREE_ON_EMPTY);
1c79356b A	180
	181	offset += PAGE_SIZE;
	182	isize -= PAGE_SIZE;
	183	pg_index++;
	184
	185	continue;
	186	}
	187	if ( !upl_dirty_page(pl, pg_index)) {
	188	/*
	189	* if the page is not dirty and reached here it is
	190	* marked precious or it is due to invalidation in
	191	* memory_object_lock request as part of truncation
	192	* We also get here from vm_object_terminate()
	193	* So all you need to do in these
	194	* cases is to invalidate incore buffer if it is there
	195	*/
	196	blkno = ubc_offtoblk(vp, (off_t)(f_offset + offset));
	197	s = splbio();
	198	vp_pgoclean++;
	199	start:
	200	if (bp = incore(vp, blkno)) {
	201	if (ISSET(bp->b_flags, B_BUSY)) {
	202	SET(bp->b_flags, B_WANTED);
	203	error = tsleep(bp, (PRIBIO + 1), "vnpgout", 0);
	204	goto start;
	205	} else {
	206	bremfree(bp);
	207	SET(bp->b_flags, (B_BUSY\|B_INVAL));
	208	}
	209	}
	210	splx(s);
	211	if (bp)
	212	brelse(bp);
	213
0b4e3aa0 A	214	ubc_upl_commit_range(vpupl, offset, PAGE_SIZE,
0b4e3aa0 A	215	UPL_COMMIT_FREE_ON_EMPTY);
1c79356b A	216
	217	offset += PAGE_SIZE;
	218	isize -= PAGE_SIZE;
	219	pg_index++;
	220
	221	continue;
	222	}
	223	vp_pgodirty++;
	224
	225	num_of_pages = 1;
	226	xsize = isize - PAGE_SIZE;
	227
	228	while (xsize) {
	229	if ( !upl_valid_page(pl, pg_index + num_of_pages))
	230	break;
	231	if ( !upl_dirty_page(pl, pg_index + num_of_pages))
	232	break;
	233	num_of_pages++;
	234	xsize -= PAGE_SIZE;
	235	}
	236	xsize = num_of_pages * PAGE_SIZE;
	237
	238	/* By defn callee will commit or abort upls */
	239	if (error = VOP_PAGEOUT(vp, vpupl, (vm_offset_t) offset,
	240	(off_t)(f_offset + offset),
	241	xsize, p->p_ucred, flags & ~UPL_NOCOMMIT)) {
	242	result = PAGER_ERROR;
	243	error = PAGER_ERROR;
	244	}
	245	offset += xsize;
	246	isize -= xsize;
	247	pg_index += num_of_pages;
	248	}
	249	out:
	250	if (errorp)
	251	*errorp = result;
	252
	253	thread_funnel_set(kernel_flock, funnel_state);
	254
	255	return (error);
	256	}
	257
	258
	259	pager_return_t
	260	vnode_pagein(
	261	struct vnode *vp,
	262	upl_t pl,
	263	vm_offset_t pl_offset,
	264	vm_object_offset_t f_offset,
	265	vm_size_t size,
	266	int flags,
	267	int *errorp)
	268	{
	269	int result = PAGER_SUCCESS;
	270	struct proc *p = current_proc();
	271	int error = 0;
0b4e3aa0	272	int xfer_size;
1c79356b A	273	boolean_t funnel_state;
1c79356b A	274	int haveupl=0;
0b4e3aa0 A	275	upl_t vpupl = NULL;
0b4e3aa0 A	276	off_t local_offset;
1c79356b A	277	unsigned int ioaddr;
	278
	279	funnel_state = thread_funnel_set(kernel_flock, TRUE);
	280
	281	#if 0
	282	if(pl->page_list.npages >1 )
	283	panic("vnode_pageout: Can't handle more than one page");
	284	#endif /* 0 */
	285
	286	if (pl != (upl_t)NULL) {
	287	haveupl = 1;
	288	}
	289	UBCINFOCHECK("vnode_pagein", vp);
	290
	291	if (UBCINVALID(vp)) {
	292	result = PAGER_ERROR;
	293	error = PAGER_ERROR;
	294	goto out;
	295	}
	296
	297	if (haveupl) {
	298	dp_pgins++;
	299	if (error = VOP_PAGEIN(vp, pl, pl_offset, (off_t)f_offset,
	300	size,p->p_ucred, flags)) {
	301	result = PAGER_ERROR;
	302	}
	303	} else {
	304
0b4e3aa0 A	305	local_offset = 0;
	306	while (size) {
	307	if((size > 4096) && (vp->v_tag == VT_NFS)) {
	308	xfer_size = 4096;
	309	size = size - xfer_size;
	310	} else {
	311	xfer_size = size;
	312	size = 0;
	313	}
	314	ubc_create_upl( vp,
	315	f_offset+local_offset,
	316	xfer_size,
	317	&vpupl,
	318	NULL,
	319	UPL_FLAGS_NONE);
	320	if (vpupl == (upl_t) 0) {
	321	result = PAGER_ABSENT;
	322	error = PAGER_ABSENT;
	323	goto out;
	324	}
1c79356b	325
0b4e3aa0	326	vp_pagein++;
1c79356b	327
0b4e3aa0 A	328	/* By defn callee will commit or abort upls */
	329	if (error = VOP_PAGEIN(vp, vpupl, (vm_offset_t) 0,
	330	(off_t)f_offset+local_offset, xfer_size,p->p_ucred, flags & ~UPL_NOCOMMIT)) {
	331	result = PAGER_ERROR;
	332	error = PAGER_ERROR;
	333	}
	334	local_offset += PAGE_SIZE_64;
1c79356b A	335	}
	336	}
	337	out:
	338	if (errorp)
	339	*errorp = result;
	340	thread_funnel_set(kernel_flock, funnel_state);
	341
	342	return (error);
	343	}
	344
	345	void
	346	vnode_pager_shutdown()
	347	{
	348	int i;
	349	extern struct bs_map bs_port_table[];
	350	struct vnode *vp;
	351
	352	for(i = 0; i < MAX_BACKING_STORE; i++) {
	353	vp = (struct vnode *)(bs_port_table[i]).vp;
	354	if (vp) {
	355	(bs_port_table[i]).vp = 0;
	356	ubc_rele(vp);
	357	/* get rid of macx_swapon() namei() reference */
	358	vrele(vp);
	359
	360	/* get rid of macx_swapon() "extra" reference */
	361	vrele(vp);
	362	}
	363	}
	364	}
	365
0b4e3aa0 A	366
0b4e3aa0 A	367	void *
1c79356b A	368	upl_get_internal_page_list(upl_t upl)
1c79356b A	369	{
0b4e3aa0	370	return(UPL_GET_INTERNAL_PAGE_LIST(upl));
1c79356b A	371
1c79356b A	372	}