pmap.c

早期freebsd实现

/* 
 * Copyright (c) 1992, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * the Systems Programming Group of the University of Utah Computer
 * Science Department and Ralph Campbell.
 *
 * 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.
 *
 *	@(#)pmap.c	8.4 (Berkeley) 1/26/94
 */

/*
 *	Manages physical address maps.
 *
 *	In addition to hardware address maps, this
 *	module is called upon to provide software-use-only
 *	maps which may or may not be stored in the same
 *	form as hardware maps.  These pseudo-maps are
 *	used to store intermediate results from copy
 *	operations to and from address spaces.
 *
 *	Since the information managed by this module is
 *	also stored by the logical address mapping module,
 *	this module may throw away valid virtual-to-physical
 *	mappings at almost any time.  However, invalidations
 *	of virtual-to-physical mappings must be done as
 *	requested.
 *
 *	In order to cope with hardware architectures which
 *	make virtual-to-physical map invalidates expensive,
 *	this module may delay invalidate or reduced protection
 *	operations until such time as they are actually
 *	necessary.  This module is given full information as
 *	to which processors are currently using which maps,
 *	and to when physical maps must be made correct.
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/malloc.h>
#include <sys/user.h>
#include <sys/buf.h>
#ifdef SYSVSHM
#include <sys/shm.h>
#endif

#include <vm/vm_kern.h>
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>

#include <machine/machConst.h>
#include <machine/pte.h>

extern vm_page_t vm_page_alloc1 __P((void));
extern void vm_page_free1 __P((vm_page_t));

/*
 * For each vm_page_t, there is a list of all currently valid virtual
 * mappings of that page.  An entry is a pv_entry_t, the list is pv_table.
 * XXX really should do this as a part of the higher level code.
 */
typedef struct pv_entry {
	struct pv_entry	*pv_next;	/* next pv_entry */
	struct pmap	*pv_pmap;	/* pmap where mapping lies */
	vm_offset_t	pv_va;		/* virtual address for mapping */
} *pv_entry_t;

pv_entry_t	pv_table;	/* array of entries, one per page */
extern void	pmap_remove_pv();

#define pa_index(pa)		atop((pa) - first_phys_addr)
#define pa_to_pvh(pa)		(&pv_table[pa_index(pa)])

#ifdef DEBUG
struct {
	int kernel;	/* entering kernel mapping */
	int user;	/* entering user mapping */
	int ptpneeded;	/* needed to allocate a PT page */
	int pwchange;	/* no mapping change, just wiring or protection */
	int wchange;	/* no mapping change, just wiring */
	int mchange;	/* was mapped but mapping to different page */
	int managed;	/* a managed page */
	int firstpv;	/* first mapping for this PA */
	int secondpv;	/* second mapping for this PA */
	int ci;		/* cache inhibited */
	int unmanaged;	/* not a managed page */
	int flushes;	/* cache flushes */
	int cachehit;	/* new entry forced valid entry out */
} enter_stats;
struct {
	int calls;
	int removes;
	int flushes;
	int pidflushes;	/* HW pid stolen */
	int pvfirst;
	int pvsearch;
} remove_stats;

int pmapdebug;
#define PDB_FOLLOW	0x0001
#define PDB_INIT	0x0002
#define PDB_ENTER	0x0004
#define PDB_REMOVE	0x0008
#define PDB_CREATE	0x0010
#define PDB_PTPAGE	0x0020
#define PDB_PVENTRY	0x0040
#define PDB_BITS	0x0080
#define PDB_COLLECT	0x0100
#define PDB_PROTECT	0x0200
#define PDB_TLBPID	0x0400
#define PDB_PARANOIA	0x2000
#define PDB_WIRING	0x4000
#define PDB_PVDUMP	0x8000

#endif /* DEBUG */

struct pmap	kernel_pmap_store;

vm_offset_t    	avail_start;	/* PA of first available physical page */
vm_offset_t	avail_end;	/* PA of last available physical page */
vm_size_t	mem_size;	/* memory size in bytes */
vm_offset_t	virtual_avail;  /* VA of first avail page (after kernel bss)*/
vm_offset_t	virtual_end;	/* VA of last avail page (end of kernel AS) */
int		pmaxpagesperpage;	/* PAGE_SIZE / NBPG */
#ifdef ATTR
char		*pmap_attributes;	/* reference and modify bits */
#endif
struct segtab	*free_segtab;		/* free list kept locally */
u_int		tlbpid_gen = 1;		/* TLB PID generation count */
int		tlbpid_cnt = 2;		/* next available TLB PID */
pt_entry_t	*Sysmap;		/* kernel pte table */
u_int		Sysmapsize;		/* number of pte's in Sysmap */

/*
 *	Bootstrap the system enough to run with virtual memory.
 *	firstaddr is the first unused kseg0 address (not page aligned).
 */
void
pmap_bootstrap(firstaddr)
	vm_offset_t firstaddr;
{
	register int i;
	vm_offset_t start = firstaddr;
	extern int maxmem, physmem;

#define	valloc(name, type, num) \
	    (name) = (type *)firstaddr; firstaddr = (vm_offset_t)((name)+(num))
	/*
	 * Allocate a PTE table for the kernel.
	 * The '1024' comes from PAGER_MAP_SIZE in vm_pager_init().
	 * This should be kept in sync.
	 * We also reserve space for kmem_alloc_pageable() for vm_fork().
	 */
	Sysmapsize = (VM_KMEM_SIZE + VM_MBUF_SIZE + VM_PHYS_SIZE +
		nbuf * MAXBSIZE + 16 * NCARGS) / NBPG + 1024 + 256;
#ifdef SYSVSHM
	Sysmapsize += shminfo.shmall;
#endif
	valloc(Sysmap, pt_entry_t, Sysmapsize);
#ifdef ATTR
	valloc(pmap_attributes, char, physmem);
#endif
	/*
	 * Allocate memory for pv_table.
	 * This will allocate more entries than we really need.
	 * We could do this in pmap_init when we know the actual
	 * phys_start and phys_end but its better to use kseg0 addresses
	 * rather than kernel virtual addresses mapped through the TLB.
	 */
	i = maxmem - pmax_btop(MACH_CACHED_TO_PHYS(firstaddr));
	valloc(pv_table, struct pv_entry, i);

	/*
	 * Clear allocated memory.
	 */
	firstaddr = pmax_round_page(firstaddr);
	bzero((caddr_t)start, firstaddr - start);

	avail_start = MACH_CACHED_TO_PHYS(firstaddr);
	avail_end = pmax_ptob(maxmem);
	mem_size = avail_end - avail_start;

	virtual_avail = VM_MIN_KERNEL_ADDRESS;
	virtual_end = VM_MIN_KERNEL_ADDRESS + Sysmapsize * NBPG;
	/* XXX need to decide how to set cnt.v_page_size */
	pmaxpagesperpage = 1;

	simple_lock_init(&kernel_pmap_store.pm_lock);
	kernel_pmap_store.pm_count = 1;
}

/*
 * Bootstrap memory allocator. This function allows for early dynamic
 * memory allocation until the virtual memory system has been bootstrapped.
 * After that point, either kmem_alloc or malloc should be used. This
 * function works by stealing pages from the (to be) managed page pool,
 * stealing virtual address space, then mapping the pages and zeroing them.
 *
 * It should be used from pmap_bootstrap till vm_page_startup, afterwards
 * it cannot be used, and will generate a panic if tried. Note that this
 * memory will never be freed, and in essence it is wired down.
 */
void *
pmap_bootstrap_alloc(size)
	int size;
{
	vm_offset_t val;
	extern boolean_t vm_page_startup_initialized;

	if (vm_page_startup_initialized)
		panic("pmap_bootstrap_alloc: called after startup initialized");

	val = MACH_PHYS_TO_CACHED(avail_start);
	size = round_page(size);
	avail_start += size;

	blkclr((caddr_t)val, size);
	return ((void *)val);
}

/*
 *	Initialize the pmap module.
 *	Called by vm_init, to initialize any structures that the pmap
 *	system needs to map virtual memory.
 */
void
pmap_init(phys_start, phys_end)
	vm_offset_t phys_start, phys_end;
{

#ifdef DEBUG
	if (pmapdebug & (PDB_FOLLOW|PDB_INIT))
		printf("pmap_init(%x, %x)\n", phys_start, phys_end);
#endif
}

/*
 *	Create and return a physical map.
 *
 *	If the size specified for the map
 *	is zero, the map is an actual physical
 *	map, and may be referenced by the
 *	hardware.
 *
 *	If the size specified is non-zero,
 *	the map will be used in software only, and
 *	is bounded by that size.
 */
pmap_t
pmap_create(size)
	vm_size_t size;
{
	register pmap_t pmap;

#ifdef DEBUG
	if (pmapdebug & (PDB_FOLLOW|PDB_CREATE))
		printf("pmap_create(%x)\n", size);
#endif
	/*
	 * Software use map does not need a pmap
	 */
	if (size)
		return (NULL);

	/* XXX: is it ok to wait here? */
	pmap = (pmap_t) malloc(sizeof *pmap, M_VMPMAP, M_WAITOK);
#ifdef notifwewait
	if (pmap == NULL)
		panic("pmap_create: cannot allocate a pmap");
#endif
	bzero(pmap, sizeof(*pmap));
	pmap_pinit(pmap);
	return (pmap);
}

/*
 * Initialize a preallocated and zeroed pmap structure,
 * such as one in a vmspace structure.
 */
void
pmap_pinit(pmap)
	register struct pmap *pmap;
{
	register int i;
	int s;
	extern struct vmspace vmspace0;
	extern struct user *proc0paddr;

#ifdef DEBUG
	if (pmapdebug & (PDB_FOLLOW|PDB_CREATE))
		printf("pmap_pinit(%x)\n", pmap);
#endif
	simple_lock_init(&pmap->pm_lock);
	pmap->pm_count = 1;
	if (free_segtab) {
		s = splimp();
		pmap->pm_segtab = free_segtab;
		free_segtab = *(struct segtab **)free_segtab;
		pmap->pm_segtab->seg_tab[0] = NULL;
		splx(s);
	} else {
		register struct segtab *stp;
		vm_page_t mem;

		mem = vm_page_alloc1();
		pmap_zero_page(VM_PAGE_TO_PHYS(mem));
		pmap->pm_segtab = stp = (struct segtab *)
			MACH_PHYS_TO_CACHED(VM_PAGE_TO_PHYS(mem));
		i = pmaxpagesperpage * (NBPG / sizeof(struct segtab));
		s = splimp();
		while (--i != 0) {
			stp++;
			*(struct segtab **)stp = free_segtab;
			free_segtab = stp;
		}
		splx(s);
	}
#ifdef DIAGNOSTIC
	for (i = 0; i < PMAP_SEGTABSIZE; i++)
		if (pmap->pm_segtab->seg_tab[i] != 0)
			panic("pmap_pinit: pm_segtab != 0");
#endif
	if (pmap == &vmspace0.vm_pmap) {
		/*
		 * The initial process has already been allocated a TLBPID
		 * in mach_init().
		 */
		pmap->pm_tlbpid = 1;
		pmap->pm_tlbgen = tlbpid_gen;
		proc0paddr->u_pcb.pcb_segtab = (void *)pmap->pm_segtab;
	} else {
		pmap->pm_tlbpid = 0;
		pmap->pm_tlbgen = 0;
	}
}

/*
 *	Retire the given physical map from service.
 *	Should only be called if the map contains
 *	no valid mappings.
 */
void
pmap_destroy(pmap)
	register pmap_t pmap;
{
	int count;

#ifdef DEBUG
	if (pmapdebug & (PDB_FOLLOW|PDB_CREATE))
		printf("pmap_destroy(%x)\n", pmap);
#endif
	if (pmap == NULL)
		return;

	simple_lock(&pmap->pm_lock);
	count = --pmap->pm_count;
	simple_unlock(&pmap->pm_lock);
	if (count == 0) {
		pmap_release(pmap);
		free((caddr_t)pmap, M_VMPMAP);
	}
}

/*
 * Release any resources held by the given physical map.
 * Called when a pmap initialized by pmap_pinit is being released.
 * Should only be called if the map contains no valid mappings.
 */
void
pmap_release(pmap)
	register pmap_t pmap;
{

#ifdef DEBUG
	if (pmapdebug & (PDB_FOLLOW|PDB_CREATE))
		printf("pmap_release(%x)\n", pmap);
#endif

	if (pmap->pm_segtab) {
		register pt_entry_t *pte;
		register int i;
		int s;
#ifdef DIAGNOSTIC
		register int j;
#endif

		for (i = 0; i < PMAP_SEGTABSIZE; i++) {
			/* get pointer to segment map */
			pte = pmap->pm_segtab->seg_tab[i];
			if (!pte)
				continue;
			vm_page_free1(
				PHYS_TO_VM_PAGE(MACH_CACHED_TO_PHYS(pte)));
#ifdef DIAGNOSTIC
			for (j = 0; j < NPTEPG; j++) {
				if (pte->pt_entry)
					panic("pmap_release: segmap not empty");
			}
#endif
			pmap->pm_segtab->seg_tab[i] = NULL;
		}
		s = splimp();
		*(struct segtab **)pmap->pm_segtab = free_segtab;
		free_segtab = pmap->pm_segtab;
		splx(s);
		pmap->pm_segtab = NULL;
	}
}

/*
 *	Add a reference to the specified pmap.
 */
void
pmap_reference(pmap)
	pmap_t pmap;
{

#ifdef DEBUG
	if (pmapdebug & PDB_FOLLOW)
		printf("pmap_reference(%x)\n", pmap);
#endif
	if (pmap != NULL) {
		simple_lock(&pmap->pm_lock);
		pmap->pm_count++;
		simple_unlock(&pmap->pm_lock);
	}
}

/*
 *	Remove the given range of addresses from the specified map.
 *
 *	It is assumed that the start and end are properly
 *	rounded to the page size.
 */
void
pmap_remove(pmap, sva, eva)
	register pmap_t pmap;
	vm_offset_t sva, eva;
{
	register vm_offset_t nssva;
	register pt_entry_t *pte;
	unsigned entry;

#ifdef DEBUG
	if (pmapdebug & (PDB_FOLLOW|PDB_REMOVE|PDB_PROTECT))
		printf("pmap_remove(%x, %x, %x)\n", pmap, sva, eva);
	remove_stats.calls++;
#endif
	if (pmap == NULL)
		return;

	if (!pmap->pm_segtab) {
		register pt_entry_t *pte;

		/* remove entries from kernel pmap */
#ifdef DIAGNOSTIC
		if (sva < VM_MIN_KERNEL_ADDRESS || eva > virtual_end)
			panic("pmap_remove: kva not in range");
#endif
		pte = kvtopte(sva);
		for (; sva < eva; sva += NBPG, pte++) {
			entry = pte->pt_entry;
			if (!(entry & PG_V))
				continue;
			if (entry & PG_WIRED)
				pmap->pm_stats.wired_count--;