[unix-history] / sys / kern / kern_physio.c

/*
 * Copyright (c) 1989, 1990, 1991, 1992 William F. Jolitz, TeleMuse
 * All rights reserved.
 *
 * 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 software is a component of "386BSD" developed by 
	William F. Jolitz, TeleMuse.
 * 4. Neither the name of the developer nor the name "386BSD"
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS A COMPONENT OF 386BSD DEVELOPED BY WILLIAM F. JOLITZ 
 * AND IS INTENDED FOR RESEARCH AND EDUCATIONAL PURPOSES ONLY. THIS 
 * SOFTWARE SHOULD NOT BE CONSIDERED TO BE A COMMERCIAL PRODUCT. 
 * THE DEVELOPER URGES THAT USERS WHO REQUIRE A COMMERCIAL PRODUCT 
 * NOT MAKE USE THIS WORK.
 *
 * FOR USERS WHO WISH TO UNDERSTAND THE 386BSD SYSTEM DEVELOPED
 * BY WILLIAM F. JOLITZ, WE RECOMMEND THE USER STUDY WRITTEN 
 * REFERENCES SUCH AS THE  "PORTING UNIX TO THE 386" SERIES 
 * (BEGINNING JANUARY 1991 "DR. DOBBS JOURNAL", USA AND BEGINNING 
 * JUNE 1991 "UNIX MAGAZIN", GERMANY) BY WILLIAM F. JOLITZ AND 
 * LYNNE GREER JOLITZ, AS WELL AS OTHER BOOKS ON UNIX AND THE 
 * ON-LINE 386BSD USER MANUAL BEFORE USE. A BOOK DISCUSSING THE INTERNALS 
 * OF 386BSD ENTITLED "386BSD FROM THE INSIDE OUT" WILL BE AVAILABLE LATE 1992.
 *
 * THIS SOFTWARE IS PROVIDED BY THE DEVELOPER ``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 DEVELOPER 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.
 *
 *	$Id: kern_physio.c,v 1.6 1994/04/14 07:54:54 davidg Exp $
 */

#include "param.h"
#include "systm.h"
#include "buf.h"
#include "conf.h"
#include "proc.h"
#include "malloc.h"
#include "vnode.h"
#include "vm/vm.h"
#include "vm/vm_page.h"
#include "specdev.h"

#define HOLD_WORKS_FOR_SHARING

/*
 * Driver interface to do "raw" I/O in the address space of a
 * user process directly for read and write operations..
 */

int
rawread(dev, uio)
	dev_t dev; struct uio *uio;
{
	return (uioapply(physio, (caddr_t) cdevsw[major(dev)].d_strategy,
				 (caddr_t) (u_long) dev, uio));
}

int
rawwrite(dev, uio)
	dev_t dev; struct uio *uio;
{
	return (uioapply(physio, (caddr_t) cdevsw[major(dev)].d_strategy,
				(caddr_t) (u_long) dev, uio));
}

static void
physwakeup(bp)
	struct buf *bp;
{
	wakeup((caddr_t) bp);
	bp->b_flags &= ~B_CALL;
}

int physio(strat, dev, bp, off, rw, base, len, p)
	d_strategy_t strat; 
	dev_t dev;
	struct buf *bp;
	int rw, off;
	caddr_t base;
	int *len;
	struct proc *p;
{
	int amttodo = *len;
	int error, amtdone;
	vm_prot_t ftype;
	vm_offset_t v, lastv, pa;
	caddr_t adr;
	int oldflags;
	int s;
	
	int bp_alloc = (bp == 0);

/*
 * keep the process from being swapped
 */
	oldflags = p->p_flag;
	p->p_flag |= SPHYSIO;

	rw = rw == UIO_READ ? B_READ : 0;

	/* create and build a buffer header for a transfer */

	if (bp_alloc) {
		bp = (struct buf *)getpbuf();
		bzero((char *)bp, sizeof(*bp));			/* 09 Sep 92*/
	} else {
		s = splbio();
		while (bp->b_flags & B_BUSY) {
			bp->b_flags |= B_WANTED;
			tsleep((caddr_t)bp, PRIBIO, "physbw", 0);
		}
		bp->b_flags |= B_BUSY;
		splx(s);
	}

	bp->b_proc = p;
	bp->b_dev = dev;
	bp->b_error = 0;
	bp->b_blkno = off/DEV_BSIZE;
	amtdone = 0;

	/* iteratively do I/O on as large a chunk as possible */
	do {
		bp->b_flags = B_BUSY | B_PHYS | B_CALL | rw;
		bp->b_iodone = physwakeup;
		bp->b_un.b_addr = base;
		/*
		 * Notice that b_bufsize is more owned by the buffer
		 * allocating entity, while b_bcount might be modified
		 * by the called I/O routines.  So after I/O is complete
		 * the only thing guaranteed to be unchanged is
		 * b_bufsize.
		 */
		bp->b_bcount = min (256*1024, amttodo);
		bp->b_bufsize = bp->b_bcount;

		/* first, check if accessible */
		if (rw == B_READ && !useracc(base, bp->b_bufsize, B_WRITE)) {
			error = EFAULT;
			goto errrtn;
		}
		if (rw == B_WRITE && !useracc(base, bp->b_bufsize, B_READ)) {
			error = EFAULT;
			goto errrtn;
		}

		/* update referenced and dirty bits, handle copy objects */
		if (rw == B_READ)
			ftype = VM_PROT_READ | VM_PROT_WRITE;
		else
			ftype = VM_PROT_READ;

		lastv = 0;
		for (adr = (caddr_t)trunc_page(base); adr < base + bp->b_bufsize;
			adr += NBPG) {
	
/*
 * make sure that the pde is valid and held
 */
			v = trunc_page(((vm_offset_t)vtopte(adr)));
			if (v != lastv) {
				vm_fault_quick(v, VM_PROT_READ);
				pa = pmap_extract(&p->p_vmspace->vm_pmap, v);
				vm_page_hold(PHYS_TO_VM_PAGE(pa));
				lastv = v;
			}

/*
 * do the vm_fault if needed, do the copy-on-write thing when
 * reading stuff off device into memory.
 */
			vm_fault_quick(adr, ftype);
			pa = pmap_extract(&p->p_vmspace->vm_pmap, (vm_offset_t) adr);
/*
 * hold the data page
 */
			vm_page_hold(PHYS_TO_VM_PAGE(pa));
		}

		vmapbuf(bp);

		/* perform transfer */
		(*strat)(bp);

		/* pageout daemon doesn't wait for pushed pages */
		s = splbio();
		while ((bp->b_flags & B_DONE) == 0)
			tsleep((caddr_t)bp, PRIBIO, "physstr", 0);
		splx(s);

		vunmapbuf(bp);

/*
 * unhold the pde, and data pages
 */
		lastv = 0;
		for (adr = (caddr_t)trunc_page(base); adr < base + bp->b_bufsize;
			adr += NBPG) {
			v = trunc_page(((vm_offset_t)vtopte(adr)));
			if (v != lastv) {
				pa = pmap_extract(&p->p_vmspace->vm_pmap, v);
				vm_page_unhold(PHYS_TO_VM_PAGE(pa));
				lastv = v;
			}
			pa = pmap_extract(&p->p_vmspace->vm_pmap, (vm_offset_t) adr);
			vm_page_unhold(PHYS_TO_VM_PAGE(pa));
		}
			

		/*
		 * in this case, we need to use b_bcount instead of
		 * b_bufsize.
		 */
		amtdone = bp->b_bcount - bp->b_resid;
		amttodo -= amtdone;
		base += amtdone;
		bp->b_blkno += amtdone/DEV_BSIZE;
	} while (amttodo && (bp->b_flags & B_ERROR) == 0 && amtdone > 0);

	error = bp->b_error;
errrtn:
	if (bp_alloc) {
		relpbuf(bp);
	} else {
		bp->b_flags &= ~B_BUSY;
		wakeup((caddr_t)bp);
	}
	*len = amttodo;

/*
 * allow the process to be swapped
 */
	p->p_flag &= ~SPHYSIO;
	p->p_flag |= (oldflags & SPHYSIO);

	return (error);
}
Commit	Line	Data
dd403947 DG	1	/*
	2	* Copyright (c) 1989, 1990, 1991, 1992 William F. Jolitz, TeleMuse
	3	* All rights reserved.
	4	*
	5	* Redistribution and use in source and binary forms, with or without
	6	* modification, are permitted provided that the following conditions
	7	* are met:
	8	* 1. Redistributions of source code must retain the above copyright
	9	* notice, this list of conditions and the following disclaimer.
	10	* 2. Redistributions in binary form must reproduce the above copyright
	11	* notice, this list of conditions and the following disclaimer in the
	12	* documentation and/or other materials provided with the distribution.
	13	* 3. All advertising materials mentioning features or use of this software
	14	* must display the following acknowledgement:
	15	* This software is a component of "386BSD" developed by
	16	William F. Jolitz, TeleMuse.
	17	* 4. Neither the name of the developer nor the name "386BSD"
	18	* may be used to endorse or promote products derived from this software
	19	* without specific prior written permission.
	20	*
	21	* THIS SOFTWARE IS A COMPONENT OF 386BSD DEVELOPED BY WILLIAM F. JOLITZ
	22	* AND IS INTENDED FOR RESEARCH AND EDUCATIONAL PURPOSES ONLY. THIS
	23	* SOFTWARE SHOULD NOT BE CONSIDERED TO BE A COMMERCIAL PRODUCT.
	24	* THE DEVELOPER URGES THAT USERS WHO REQUIRE A COMMERCIAL PRODUCT
	25	* NOT MAKE USE THIS WORK.
	26	*
	27	* FOR USERS WHO WISH TO UNDERSTAND THE 386BSD SYSTEM DEVELOPED
	28	* BY WILLIAM F. JOLITZ, WE RECOMMEND THE USER STUDY WRITTEN
	29	* REFERENCES SUCH AS THE "PORTING UNIX TO THE 386" SERIES
	30	* (BEGINNING JANUARY 1991 "DR. DOBBS JOURNAL", USA AND BEGINNING
	31	* JUNE 1991 "UNIX MAGAZIN", GERMANY) BY WILLIAM F. JOLITZ AND
	32	* LYNNE GREER JOLITZ, AS WELL AS OTHER BOOKS ON UNIX AND THE
	33	* ON-LINE 386BSD USER MANUAL BEFORE USE. A BOOK DISCUSSING THE INTERNALS
	34	* OF 386BSD ENTITLED "386BSD FROM THE INSIDE OUT" WILL BE AVAILABLE LATE 1992.
	35	*
	36	* THIS SOFTWARE IS PROVIDED BY THE DEVELOPER ``AS IS'' AND
	37	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	38	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	39	* ARE DISCLAIMED. IN NO EVENT SHALL THE DEVELOPER BE LIABLE
	40	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	41	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	42	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	43	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	44	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	45	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	46	* SUCH DAMAGE.
	47	*
eeaaf165	48	* $Id: kern_physio.c,v 1.6 1994/04/14 07:54:54 davidg Exp $
dd403947 DG	49	*/
	50
	51	#include "param.h"
	52	#include "systm.h"
	53	#include "buf.h"
	54	#include "conf.h"
	55	#include "proc.h"
	56	#include "malloc.h"
	57	#include "vnode.h"
	58	#include "vm/vm.h"
41aefbec	59	#include "vm/vm_page.h"
dd403947 DG	60	#include "specdev.h"
dd403947 DG	61
4d107077 DG	62	#define HOLD_WORKS_FOR_SHARING
4d107077 DG	63
dd403947 DG	64	/*
	65	* Driver interface to do "raw" I/O in the address space of a
	66	* user process directly for read and write operations..
	67	*/
	68
	69	int
	70	rawread(dev, uio)
	71	dev_t dev; struct uio *uio;
	72	{
	73	return (uioapply(physio, (caddr_t) cdevsw[major(dev)].d_strategy,
	74	(caddr_t) (u_long) dev, uio));
	75	}
	76
	77	int
	78	rawwrite(dev, uio)
	79	dev_t dev; struct uio *uio;
	80	{
	81	return (uioapply(physio, (caddr_t) cdevsw[major(dev)].d_strategy,
	82	(caddr_t) (u_long) dev, uio));
	83	}
	84
82b95dad DG	85	static void
	86	physwakeup(bp)
	87	struct buf *bp;
	88	{
	89	wakeup((caddr_t) bp);
	90	bp->b_flags &= ~B_CALL;
	91	}
dd403947 DG	92
	93	int physio(strat, dev, bp, off, rw, base, len, p)
	94	d_strategy_t strat;
	95	dev_t dev;
	96	struct buf *bp;
	97	int rw, off;
	98	caddr_t base;
	99	int *len;
	100	struct proc *p;
	101	{
	102	int amttodo = *len;
	103	int error, amtdone;
	104	vm_prot_t ftype;
41aefbec	105	vm_offset_t v, lastv, pa;
dd403947 DG	106	caddr_t adr;
	107	int oldflags;
	108	int s;
	109
	110	int bp_alloc = (bp == 0);
	111
	112	/*
	113	* keep the process from being swapped
	114	*/
	115	oldflags = p->p_flag;
	116	p->p_flag \|= SPHYSIO;
	117
	118	rw = rw == UIO_READ ? B_READ : 0;
	119
	120	/* create and build a buffer header for a transfer */
	121
	122	if (bp_alloc) {
	123	bp = (struct buf *)getpbuf();
	124	bzero((char )bp, sizeof(bp)); /* 09 Sep 92*/
	125	} else {
	126	s = splbio();
	127	while (bp->b_flags & B_BUSY) {
	128	bp->b_flags \|= B_WANTED;
	129	tsleep((caddr_t)bp, PRIBIO, "physbw", 0);
	130	}
	131	bp->b_flags \|= B_BUSY;
	132	splx(s);
	133	}
	134
dd403947 DG	135	bp->b_proc = p;
	136	bp->b_dev = dev;
	137	bp->b_error = 0;
	138	bp->b_blkno = off/DEV_BSIZE;
	139	amtdone = 0;
	140
	141	/* iteratively do I/O on as large a chunk as possible */
	142	do {
82b95dad DG	143	bp->b_flags = B_BUSY \| B_PHYS \| B_CALL \| rw;
82b95dad DG	144	bp->b_iodone = physwakeup;
dd403947	145	bp->b_un.b_addr = base;
4d107077 DG	146	/*
	147	* Notice that b_bufsize is more owned by the buffer
	148	* allocating entity, while b_bcount might be modified
	149	* by the called I/O routines. So after I/O is complete
	150	* the only thing guaranteed to be unchanged is
	151	* b_bufsize.
	152	*/
dd403947	153	bp->b_bcount = min (256*1024, amttodo);
4d107077	154	bp->b_bufsize = bp->b_bcount;
dd403947 DG	155
dd403947 DG	156	/* first, check if accessible */
4d107077	157	if (rw == B_READ && !useracc(base, bp->b_bufsize, B_WRITE)) {
dd403947 DG	158	error = EFAULT;
	159	goto errrtn;
	160	}
4d107077	161	if (rw == B_WRITE && !useracc(base, bp->b_bufsize, B_READ)) {
dd403947 DG	162	error = EFAULT;
	163	goto errrtn;
	164	}
	165
	166	/* update referenced and dirty bits, handle copy objects */
	167	if (rw == B_READ)
	168	ftype = VM_PROT_READ \| VM_PROT_WRITE;
	169	else
	170	ftype = VM_PROT_READ;
	171
	172	lastv = 0;
4d107077	173	for (adr = (caddr_t)trunc_page(base); adr < base + bp->b_bufsize;
dd403947 DG	174	adr += NBPG) {
	175
	176	/*
41aefbec	177	* make sure that the pde is valid and held
dd403947 DG	178	*/
	179	v = trunc_page(((vm_offset_t)vtopte(adr)));
	180	if (v != lastv) {
eeaaf165	181	vm_fault_quick(v, VM_PROT_READ);
41aefbec DG	182	pa = pmap_extract(&p->p_vmspace->vm_pmap, v);
41aefbec DG	183	vm_page_hold(PHYS_TO_VM_PAGE(pa));
dd403947 DG	184	lastv = v;
	185	}
	186
	187	/*
	188	* do the vm_fault if needed, do the copy-on-write thing when
	189	* reading stuff off device into memory.
	190	*/
eeaaf165	191	vm_fault_quick(adr, ftype);
41aefbec	192	pa = pmap_extract(&p->p_vmspace->vm_pmap, (vm_offset_t) adr);
dd403947	193	/*
41aefbec	194	* hold the data page
dd403947	195	*/
41aefbec DG	196	vm_page_hold(PHYS_TO_VM_PAGE(pa));
41aefbec DG	197	}
dd403947	198
4d107077 DG	199	vmapbuf(bp);
4d107077 DG	200
dd403947	201	/* perform transfer */
4d107077 DG	202	(*strat)(bp);
	203
	204	/* pageout daemon doesn't wait for pushed pages */
	205	s = splbio();
	206	while ((bp->b_flags & B_DONE) == 0)
	207	tsleep((caddr_t)bp, PRIBIO, "physstr", 0);
	208	splx(s);
	209
	210	vunmapbuf(bp);
	211
dd403947	212	/*
41aefbec	213	* unhold the pde, and data pages
dd403947	214	*/
41aefbec	215	lastv = 0;
4d107077	216	for (adr = (caddr_t)trunc_page(base); adr < base + bp->b_bufsize;
dd403947 DG	217	adr += NBPG) {
	218	v = trunc_page(((vm_offset_t)vtopte(adr)));
	219	if (v != lastv) {
41aefbec DG	220	pa = pmap_extract(&p->p_vmspace->vm_pmap, v);
41aefbec DG	221	vm_page_unhold(PHYS_TO_VM_PAGE(pa));
dd403947 DG	222	lastv = v;
dd403947 DG	223	}
41aefbec DG	224	pa = pmap_extract(&p->p_vmspace->vm_pmap, (vm_offset_t) adr);
41aefbec DG	225	vm_page_unhold(PHYS_TO_VM_PAGE(pa));
dd403947 DG	226	}
	227
	228
4d107077 DG	229	/*
	230	* in this case, we need to use b_bcount instead of
	231	* b_bufsize.
	232	*/
dd403947 DG	233	amtdone = bp->b_bcount - bp->b_resid;
	234	amttodo -= amtdone;
	235	base += amtdone;
	236	bp->b_blkno += amtdone/DEV_BSIZE;
	237	} while (amttodo && (bp->b_flags & B_ERROR) == 0 && amtdone > 0);
	238
	239	error = bp->b_error;
	240	errrtn:
	241	if (bp_alloc) {
	242	relpbuf(bp);
	243	} else {
	244	bp->b_flags &= ~B_BUSY;
	245	wakeup((caddr_t)bp);
	246	}
	247	*len = amttodo;
	248
	249	/*
	250	* allow the process to be swapped
	251	*/
	252	p->p_flag &= ~SPHYSIO;
	253	p->p_flag \|= (oldflags & SPHYSIO);
	254
	255	return (error);
	256	}