1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1991 Regents of the University of California.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * the Systems Programming Group of the University of Utah Computer
9 * Science Department and William Jolitz of UUNET Technologies Inc.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 *    notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 *    notice, this list of conditions and the following disclaimer in the
18 *    documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its contributors
20 *    may be used to endorse or promote products derived from this software
21 *    without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * Derived from hp300 version by Mike Hibler, this version by William
36 * Jolitz uses a recursive map [a pde points to the page directory] to
37 * map the page tables using the pagetables themselves. This is done to
38 * reduce the impact on kernel virtual memory for lots of sparse address
39 * space, and to reduce the cost of memory to each process.
40 *
41 *	from: hp300: @(#)pmap.h	7.2 (Berkeley) 12/16/90
42 *	from: @(#)pmap.h	7.4 (Berkeley) 5/12/91
43 * $FreeBSD$
44 */
45
46#ifndef _MACHINE_PMAP_H_
47#define	_MACHINE_PMAP_H_
48
49/*
50 * Page-directory and page-table entries follow this format, with a few
51 * of the fields not present here and there, depending on a lot of things.
52 */
53				/* ---- Intel Nomenclature ---- */
54#define	PG_V		0x001	/* P	Valid			*/
55#define PG_RW		0x002	/* R/W	Read/Write		*/
56#define PG_U		0x004	/* U/S  User/Supervisor		*/
57#define	PG_NC_PWT	0x008	/* PWT	Write through		*/
58#define	PG_NC_PCD	0x010	/* PCD	Cache disable		*/
59#define PG_A		0x020	/* A	Accessed		*/
60#define	PG_M		0x040	/* D	Dirty			*/
61#define	PG_PS		0x080	/* PS	Page size (0=4k,1=4M)	*/
62#define	PG_PTE_PAT	0x080	/* PAT	PAT index		*/
63#define	PG_G		0x100	/* G	Global			*/
64#define	PG_AVAIL1	0x200	/*    /	Available for system	*/
65#define	PG_AVAIL2	0x400	/*   <	programmers use		*/
66#define	PG_AVAIL3	0x800	/*    \				*/
67#define	PG_PDE_PAT	0x1000	/* PAT	PAT index		*/
68#if defined(PAE) || defined(PAE_TABLES)
69#define	PG_NX		(1ull<<63) /* No-execute */
70#endif
71
72
73/* Our various interpretations of the above */
74#define PG_W		PG_AVAIL1	/* "Wired" pseudoflag */
75#define	PG_MANAGED	PG_AVAIL2
76#define	PG_PROMOTED	PG_AVAIL3	/* PDE only */
77#if defined(PAE) || defined(PAE_TABLES)
78#define	PG_FRAME	(0x000ffffffffff000ull)
79#define	PG_PS_FRAME	(0x000fffffffe00000ull)
80#else
81#define	PG_FRAME	(~PAGE_MASK)
82#define	PG_PS_FRAME	(0xffc00000)
83#endif
84#define	PG_PROT		(PG_RW|PG_U)	/* all protection bits . */
85#define PG_N		(PG_NC_PWT|PG_NC_PCD)	/* Non-cacheable */
86
87/* Page level cache control fields used to determine the PAT type */
88#define PG_PDE_CACHE	(PG_PDE_PAT | PG_NC_PWT | PG_NC_PCD)
89#define PG_PTE_CACHE	(PG_PTE_PAT | PG_NC_PWT | PG_NC_PCD)
90
91/*
92 * Promotion to a 2 or 4MB (PDE) page mapping requires that the corresponding
93 * 4KB (PTE) page mappings have identical settings for the following fields:
94 */
95#define PG_PTE_PROMOTE	(PG_MANAGED | PG_W | PG_G | PG_PTE_PAT | \
96	    PG_M | PG_A | PG_NC_PCD | PG_NC_PWT | PG_U | PG_RW | PG_V)
97
98/*
99 * Page Protection Exception bits
100 */
101
102#define PGEX_P		0x01	/* Protection violation vs. not present */
103#define PGEX_W		0x02	/* during a Write cycle */
104#define PGEX_U		0x04	/* access from User mode (UPL) */
105#define PGEX_RSV	0x08	/* reserved PTE field is non-zero */
106#define PGEX_I		0x10	/* during an instruction fetch */
107
108/*
109 * Size of Kernel address space.  This is the number of page table pages
110 * (4MB each) to use for the kernel.  256 pages == 1 Gigabyte.
111 * This **MUST** be a multiple of 4 (eg: 252, 256, 260, etc).
112 * For PAE, the page table page unit size is 2MB.  This means that 512 pages
113 * is 1 Gigabyte.  Double everything.  It must be a multiple of 8 for PAE.
114 */
115#if defined(PAE) || defined(PAE_TABLES)
116#define KVA_PAGES	(512*4)
117#else
118#define KVA_PAGES	(256*4)
119#endif
120
121/*
122 * Pte related macros
123 */
124#define VADDR(pdi, pti) ((vm_offset_t)(((pdi)<<PDRSHIFT)|((pti)<<PAGE_SHIFT)))
125
126/*
127 * The initial number of kernel page table pages that are constructed
128 * by pmap_cold() must be sufficient to map vm_page_array[].  That number can
129 * be calculated as follows:
130 *     max_phys / PAGE_SIZE * sizeof(struct vm_page) / NBPDR
131 * PAE:      max_phys 16G, sizeof(vm_page) 76, NBPDR 2M, 152 page table pages.
132 * PAE_TABLES: max_phys 4G,  sizeof(vm_page) 68, NBPDR 2M, 36 page table pages.
133 * Non-PAE:  max_phys 4G,  sizeof(vm_page) 68, NBPDR 4M, 18 page table pages.
134 */
135#ifndef NKPT
136#if defined(PAE)
137#define	NKPT		240
138#elif defined(PAE_TABLES)
139#define	NKPT		60
140#else
141#define	NKPT		30
142#endif
143#endif
144
145#ifndef NKPDE
146#define NKPDE	(KVA_PAGES)	/* number of page tables/pde's */
147#endif
148
149/*
150 * The *PTDI values control the layout of virtual memory
151 */
152#define	KPTDI		0		/* start of kernel virtual pde's */
153#define	LOWPTDI		1		/* low memory map pde */
154#define	KERNPTDI	2		/* start of kernel text pde */
155#define	PTDPTDI		(NPDEPTD - 1 - NPGPTD)	/* ptd entry that points
156						   to ptd! */
157#define	TRPTDI		(NPDEPTD - 1)	/* u/k trampoline ptd */
158
159/*
160 * XXX doesn't really belong here I guess...
161 */
162#define ISA_HOLE_START    0xa0000
163#define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START)
164
165#ifndef LOCORE
166
167#include <sys/queue.h>
168#include <sys/_cpuset.h>
169#include <sys/_lock.h>
170#include <sys/_mutex.h>
171
172#include <vm/_vm_radix.h>
173
174#if defined(PAE) || defined(PAE_TABLES)
175
176typedef uint64_t pdpt_entry_t;
177typedef uint64_t pd_entry_t;
178typedef uint64_t pt_entry_t;
179
180#define	PTESHIFT	(3)
181#define	PDESHIFT	(3)
182
183#else
184
185typedef uint32_t pd_entry_t;
186typedef uint32_t pt_entry_t;
187
188#define	PTESHIFT	(2)
189#define	PDESHIFT	(2)
190
191#endif
192
193/*
194 * Address of current address space page table maps and directories.
195 */
196#ifdef _KERNEL
197#include <machine/atomic.h>
198
199extern pt_entry_t PTmap[];
200extern pd_entry_t PTD[];
201extern pd_entry_t PTDpde[];
202
203#if defined(PAE) || defined(PAE_TABLES)
204extern pdpt_entry_t *IdlePDPT;
205#endif
206extern pd_entry_t *IdlePTD;	/* physical address of "Idle" state directory */
207
208/*
209 * Translate a virtual address to the kernel virtual address of its page table
210 * entry (PTE).  This can be used recursively.  If the address of a PTE as
211 * previously returned by this macro is itself given as the argument, then the
212 * address of the page directory entry (PDE) that maps the PTE will be
213 * returned.
214 *
215 * This macro may be used before pmap_bootstrap() is called.
216 */
217#define	vtopte(va)	(PTmap + i386_btop(va))
218
219/*
220 * Translate a virtual address to its physical address.
221 *
222 * This macro may be used before pmap_bootstrap() is called.
223 */
224#define	vtophys(va)	pmap_kextract((vm_offset_t)(va))
225
226/*
227 * KPTmap is a linear mapping of the kernel page table.  It differs from the
228 * recursive mapping in two ways: (1) it only provides access to kernel page
229 * table pages, and not user page table pages, and (2) it provides access to
230 * a kernel page table page after the corresponding virtual addresses have
231 * been promoted to a 2/4MB page mapping.
232 *
233 * KPTmap is first initialized by pmap_cold() to support just NPKT page table
234 * pages.  Later, it is reinitialized by pmap_bootstrap() to allow for
235 * expansion of the kernel page table.
236 */
237extern pt_entry_t *KPTmap;
238
239#if (defined(PAE) || defined(PAE_TABLES))
240
241#define	pde_cmpset(pdep, old, new)	atomic_cmpset_64_i586(pdep, old, new)
242#define	pte_load_store(ptep, pte)	atomic_swap_64_i586(ptep, pte)
243#define	pte_load_clear(ptep)		atomic_swap_64_i586(ptep, 0)
244#define	pte_store(ptep, pte)		atomic_store_rel_64_i586(ptep, pte)
245#define	pte_load(ptep)			atomic_load_acq_64_i586(ptep)
246
247extern pt_entry_t pg_nx;
248
249#else /* !(PAE || PAE_TABLES) */
250
251#define	pde_cmpset(pdep, old, new)	atomic_cmpset_int(pdep, old, new)
252#define	pte_load_store(ptep, pte)	atomic_swap_int(ptep, pte)
253#define	pte_load_clear(ptep)		atomic_swap_int(ptep, 0)
254#define	pte_store(ptep, pte) do { \
255	*(u_int *)(ptep) = (u_int)(pte); \
256} while (0)
257#define	pte_load(ptep)			atomic_load_acq_int(ptep)
258
259#endif /* !(PAE || PAE_TABLES) */
260
261#define	pte_clear(ptep)			pte_store(ptep, 0)
262
263#define	pde_store(pdep, pde)		pte_store(pdep, pde)
264
265/*
266 * Extract from the kernel page table the physical address that is mapped by
267 * the given virtual address "va".
268 *
269 * This function may be used before pmap_bootstrap() is called.
270 */
271static __inline vm_paddr_t
272pmap_kextract(vm_offset_t va)
273{
274	vm_paddr_t pa;
275
276	if ((pa = pte_load(&PTD[va >> PDRSHIFT])) & PG_PS) {
277		pa = (pa & PG_PS_FRAME) | (va & PDRMASK);
278	} else {
279		/*
280		 * Beware of a concurrent promotion that changes the PDE at
281		 * this point!  For example, vtopte() must not be used to
282		 * access the PTE because it would use the new PDE.  It is,
283		 * however, safe to use the old PDE because the page table
284		 * page is preserved by the promotion.
285		 */
286		pa = KPTmap[i386_btop(va)];
287		pa = (pa & PG_FRAME) | (va & PAGE_MASK);
288	}
289	return (pa);
290}
291
292#endif /* _KERNEL */
293
294/*
295 * Pmap stuff
296 */
297struct	pv_entry;
298struct	pv_chunk;
299
300struct md_page {
301	TAILQ_HEAD(,pv_entry)	pv_list;
302	int			pat_mode;
303};
304
305struct pmap {
306	struct mtx		pm_mtx;
307	pd_entry_t		*pm_pdir;	/* KVA of page directory */
308	TAILQ_HEAD(,pv_chunk)	pm_pvchunk;	/* list of mappings in pmap */
309	cpuset_t		pm_active;	/* active on cpus */
310	struct pmap_statistics	pm_stats;	/* pmap statistics */
311	LIST_ENTRY(pmap) 	pm_list;	/* List of all pmaps */
312#if defined(PAE) || defined(PAE_TABLES)
313	pdpt_entry_t		*pm_pdpt;	/* KVA of page directory pointer
314						   table */
315#endif
316	struct vm_radix		pm_root;	/* spare page table pages */
317	vm_page_t		pm_ptdpg[NPGPTD];
318};
319
320typedef struct pmap	*pmap_t;
321
322#ifdef _KERNEL
323extern struct pmap	kernel_pmap_store;
324#define kernel_pmap	(&kernel_pmap_store)
325
326#define	PMAP_LOCK(pmap)		mtx_lock(&(pmap)->pm_mtx)
327#define	PMAP_LOCK_ASSERT(pmap, type) \
328				mtx_assert(&(pmap)->pm_mtx, (type))
329#define	PMAP_LOCK_DESTROY(pmap)	mtx_destroy(&(pmap)->pm_mtx)
330#define	PMAP_LOCK_INIT(pmap)	mtx_init(&(pmap)->pm_mtx, "pmap", \
331				    NULL, MTX_DEF | MTX_DUPOK)
332#define	PMAP_LOCKED(pmap)	mtx_owned(&(pmap)->pm_mtx)
333#define	PMAP_MTX(pmap)		(&(pmap)->pm_mtx)
334#define	PMAP_TRYLOCK(pmap)	mtx_trylock(&(pmap)->pm_mtx)
335#define	PMAP_UNLOCK(pmap)	mtx_unlock(&(pmap)->pm_mtx)
336#endif
337
338/*
339 * For each vm_page_t, there is a list of all currently valid virtual
340 * mappings of that page.  An entry is a pv_entry_t, the list is pv_list.
341 */
342typedef struct pv_entry {
343	vm_offset_t	pv_va;		/* virtual address for mapping */
344	TAILQ_ENTRY(pv_entry)	pv_next;
345} *pv_entry_t;
346
347/*
348 * pv_entries are allocated in chunks per-process.  This avoids the
349 * need to track per-pmap assignments.
350 */
351#define	_NPCM	11
352#define	_NPCPV	336
353struct pv_chunk {
354	pmap_t			pc_pmap;
355	TAILQ_ENTRY(pv_chunk)	pc_list;
356	uint32_t		pc_map[_NPCM];	/* bitmap; 1 = free */
357	TAILQ_ENTRY(pv_chunk)	pc_lru;
358	struct pv_entry		pc_pventry[_NPCPV];
359};
360
361#ifdef	_KERNEL
362
363extern caddr_t CADDR3;
364extern pt_entry_t *CMAP3;
365extern vm_paddr_t phys_avail[];
366extern vm_paddr_t dump_avail[];
367extern char *ptvmmap;		/* poor name! */
368extern vm_offset_t virtual_avail;
369extern vm_offset_t virtual_end;
370
371#define	pmap_page_get_memattr(m)	((vm_memattr_t)(m)->md.pat_mode)
372#define	pmap_page_is_write_mapped(m)	(((m)->aflags & PGA_WRITEABLE) != 0)
373#define	pmap_unmapbios(va, sz)	pmap_unmapdev((va), (sz))
374
375static inline int
376pmap_vmspace_copy(pmap_t dst_pmap __unused, pmap_t src_pmap __unused)
377{
378
379	return (0);
380}
381
382/*
383 * Only the following functions or macros may be used before pmap_bootstrap()
384 * is called: pmap_kenter(), pmap_kextract(), pmap_kremove(), vtophys(), and
385 * vtopte().
386 */
387void	pmap_activate_boot(pmap_t pmap);
388void	pmap_bootstrap(vm_paddr_t);
389int	pmap_cache_bits(pmap_t, int mode, boolean_t is_pde);
390int	pmap_change_attr(vm_offset_t, vm_size_t, int);
391void	pmap_init_pat(void);
392void	pmap_kenter(vm_offset_t va, vm_paddr_t pa);
393void	*pmap_kenter_temporary(vm_paddr_t pa, int i);
394void	pmap_kremove(vm_offset_t);
395void	*pmap_mapbios(vm_paddr_t, vm_size_t);
396void	*pmap_mapdev(vm_paddr_t, vm_size_t);
397void	*pmap_mapdev_attr(vm_paddr_t, vm_size_t, int);
398boolean_t pmap_page_is_mapped(vm_page_t m);
399void	pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma);
400bool	pmap_ps_enabled(pmap_t pmap);
401void	pmap_unmapdev(vm_offset_t, vm_size_t);
402pt_entry_t *pmap_pte(pmap_t, vm_offset_t) __pure2;
403void	pmap_invalidate_page(pmap_t, vm_offset_t);
404void	pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
405void	pmap_invalidate_all(pmap_t);
406void	pmap_invalidate_cache(void);
407void	pmap_invalidate_cache_pages(vm_page_t *pages, int count);
408void	pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
409void	pmap_force_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
410void	*pmap_trm_alloc(size_t size, int flags);
411void	pmap_trm_free(void *addr, size_t size);
412
413void	invltlb_glob(void);
414
415#endif /* _KERNEL */
416
417#endif /* !LOCORE */
418
419#endif /* !_MACHINE_PMAP_H_ */
420