1/*-
2 * Copyright (c) 1991 Regents of the University of California.
3 * All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * the Systems Programming Group of the University of Utah Computer
7 * Science Department and William Jolitz of UUNET Technologies Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * Derived from hp300 version by Mike Hibler, this version by William
34 * Jolitz uses a recursive map [a pde points to the page directory] to
35 * map the page tables using the pagetables themselves. This is done to
36 * reduce the impact on kernel virtual memory for lots of sparse address
37 * space, and to reduce the cost of memory to each process.
38 *
39 *	from: hp300: @(#)pmap.h	7.2 (Berkeley) 12/16/90
40 *	from: @(#)pmap.h	7.4 (Berkeley) 5/12/91
41 * $FreeBSD: stable/10/sys/i386/include/pmap.h 320432 2017-06-28 04:23:20Z alc $
42 */
43
44#ifndef _MACHINE_PMAP_H_
45#define	_MACHINE_PMAP_H_
46
47/*
48 * Page-directory and page-table entries follow this format, with a few
49 * of the fields not present here and there, depending on a lot of things.
50 */
51				/* ---- Intel Nomenclature ---- */
52#define	PG_V		0x001	/* P	Valid			*/
53#define PG_RW		0x002	/* R/W	Read/Write		*/
54#define PG_U		0x004	/* U/S  User/Supervisor		*/
55#define	PG_NC_PWT	0x008	/* PWT	Write through		*/
56#define	PG_NC_PCD	0x010	/* PCD	Cache disable		*/
57#define PG_A		0x020	/* A	Accessed		*/
58#define	PG_M		0x040	/* D	Dirty			*/
59#define	PG_PS		0x080	/* PS	Page size (0=4k,1=4M)	*/
60#define	PG_PTE_PAT	0x080	/* PAT	PAT index		*/
61#define	PG_G		0x100	/* G	Global			*/
62#define	PG_AVAIL1	0x200	/*    /	Available for system	*/
63#define	PG_AVAIL2	0x400	/*   <	programmers use		*/
64#define	PG_AVAIL3	0x800	/*    \				*/
65#define	PG_PDE_PAT	0x1000	/* PAT	PAT index		*/
66#if defined(PAE) || defined(PAE_TABLES)
67#define	PG_NX		(1ull<<63) /* No-execute */
68#endif
69
70
71/* Our various interpretations of the above */
72#define PG_W		PG_AVAIL1	/* "Wired" pseudoflag */
73#define	PG_MANAGED	PG_AVAIL2
74#define	PG_PROMOTED	PG_AVAIL3	/* PDE only */
75#if defined(PAE) || defined(PAE_TABLES)
76#define	PG_FRAME	(0x000ffffffffff000ull)
77#define	PG_PS_FRAME	(0x000fffffffe00000ull)
78#else
79#define	PG_FRAME	(~PAGE_MASK)
80#define	PG_PS_FRAME	(0xffc00000)
81#endif
82#define	PG_PROT		(PG_RW|PG_U)	/* all protection bits . */
83#define PG_N		(PG_NC_PWT|PG_NC_PCD)	/* Non-cacheable */
84
85/* Page level cache control fields used to determine the PAT type */
86#define PG_PDE_CACHE	(PG_PDE_PAT | PG_NC_PWT | PG_NC_PCD)
87#define PG_PTE_CACHE	(PG_PTE_PAT | PG_NC_PWT | PG_NC_PCD)
88
89/*
90 * Promotion to a 2 or 4MB (PDE) page mapping requires that the corresponding
91 * 4KB (PTE) page mappings have identical settings for the following fields:
92 */
93#define PG_PTE_PROMOTE	(PG_MANAGED | PG_W | PG_G | PG_PTE_PAT | \
94	    PG_M | PG_A | PG_NC_PCD | PG_NC_PWT | PG_U | PG_RW | PG_V)
95
96/*
97 * Page Protection Exception bits
98 */
99
100#define PGEX_P		0x01	/* Protection violation vs. not present */
101#define PGEX_W		0x02	/* during a Write cycle */
102#define PGEX_U		0x04	/* access from User mode (UPL) */
103#define PGEX_RSV	0x08	/* reserved PTE field is non-zero */
104#define PGEX_I		0x10	/* during an instruction fetch */
105
106/*
107 * Size of Kernel address space.  This is the number of page table pages
108 * (4MB each) to use for the kernel.  256 pages == 1 Gigabyte.
109 * This **MUST** be a multiple of 4 (eg: 252, 256, 260, etc).
110 * For PAE, the page table page unit size is 2MB.  This means that 512 pages
111 * is 1 Gigabyte.  Double everything.  It must be a multiple of 8 for PAE.
112 */
113#ifndef KVA_PAGES
114#if defined(PAE) || defined(PAE_TABLES)
115#define KVA_PAGES	512
116#else
117#define KVA_PAGES	256
118#endif
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 locore 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 * XXX This works for now, but I am not real happy with it, I'll fix it
153 * right after I fix locore.s and the magic 28K hole
154 */
155#define	KPTDI		(NPDEPTD-NKPDE)	/* start of kernel virtual pde's */
156#define	PTDPTDI		(KPTDI-NPGPTD)	/* ptd entry that points to ptd! */
157
158/*
159 * XXX doesn't really belong here I guess...
160 */
161#define ISA_HOLE_START    0xa0000
162#define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START)
163
164#ifndef LOCORE
165
166#include <sys/queue.h>
167#include <sys/_cpuset.h>
168#include <sys/_lock.h>
169#include <sys/_mutex.h>
170
171#include <vm/_vm_radix.h>
172
173#if defined(PAE) || defined(PAE_TABLES)
174
175typedef uint64_t pdpt_entry_t;
176typedef uint64_t pd_entry_t;
177typedef uint64_t pt_entry_t;
178
179#define	PTESHIFT	(3)
180#define	PDESHIFT	(3)
181
182#else
183
184typedef uint32_t pd_entry_t;
185typedef uint32_t pt_entry_t;
186
187#define	PTESHIFT	(2)
188#define	PDESHIFT	(2)
189
190#endif
191
192/*
193 * Address of current address space page table maps and directories.
194 */
195#ifdef _KERNEL
196extern pt_entry_t PTmap[];
197extern pd_entry_t PTD[];
198extern pd_entry_t PTDpde[];
199
200#if defined(PAE) || defined(PAE_TABLES)
201extern pdpt_entry_t *IdlePDPT;
202#endif
203extern pd_entry_t *IdlePTD;	/* physical address of "Idle" state directory */
204
205/*
206 * Translate a virtual address to the kernel virtual address of its page table
207 * entry (PTE).  This can be used recursively.  If the address of a PTE as
208 * previously returned by this macro is itself given as the argument, then the
209 * address of the page directory entry (PDE) that maps the PTE will be
210 * returned.
211 *
212 * This macro may be used before pmap_bootstrap() is called.
213 */
214#define	vtopte(va)	(PTmap + i386_btop(va))
215
216/*
217 * Translate a virtual address to its physical address.
218 *
219 * This macro may be used before pmap_bootstrap() is called.
220 */
221#define	vtophys(va)	pmap_kextract((vm_offset_t)(va))
222
223#if defined(XEN)
224#include <sys/param.h>
225
226#include <xen/xen-os.h>
227
228#include <machine/xen/xenvar.h>
229#include <machine/xen/xenpmap.h>
230
231extern pt_entry_t pg_nx;
232
233#define PG_KERNEL  (PG_V | PG_A | PG_RW | PG_M)
234
235#define MACH_TO_VM_PAGE(ma) PHYS_TO_VM_PAGE(xpmap_mtop((ma)))
236#define VM_PAGE_TO_MACH(m) xpmap_ptom(VM_PAGE_TO_PHYS((m)))
237
238#define VTOM(va) xpmap_ptom(VTOP(va))
239
240static __inline vm_paddr_t
241pmap_kextract_ma(vm_offset_t va)
242{
243        vm_paddr_t ma;
244        if ((ma = PTD[va >> PDRSHIFT]) & PG_PS) {
245                ma = (ma & ~(NBPDR - 1)) | (va & (NBPDR - 1));
246        } else {
247                ma = (*vtopte(va) & PG_FRAME) | (va & PAGE_MASK);
248        }
249        return ma;
250}
251
252static __inline vm_paddr_t
253pmap_kextract(vm_offset_t va)
254{
255        return xpmap_mtop(pmap_kextract_ma(va));
256}
257#define vtomach(va)     pmap_kextract_ma(((vm_offset_t) (va)))
258
259vm_paddr_t pmap_extract_ma(struct pmap *pmap, vm_offset_t va);
260
261void    pmap_kenter_ma(vm_offset_t va, vm_paddr_t pa);
262void    pmap_map_readonly(struct pmap *pmap, vm_offset_t va, int len);
263void    pmap_map_readwrite(struct pmap *pmap, vm_offset_t va, int len);
264
265static __inline pt_entry_t
266pte_load_store(pt_entry_t *ptep, pt_entry_t v)
267{
268	pt_entry_t r;
269
270	r = *ptep;
271	PT_SET_VA(ptep, v, TRUE);
272	return (r);
273}
274
275static __inline pt_entry_t
276pte_load_store_ma(pt_entry_t *ptep, pt_entry_t v)
277{
278	pt_entry_t r;
279
280	r = *ptep;
281	PT_SET_VA_MA(ptep, v, TRUE);
282	return (r);
283}
284
285#define	pte_load_clear(ptep)	pte_load_store((ptep), (pt_entry_t)0ULL)
286
287#define	pte_store(ptep, pte)	pte_load_store((ptep), (pt_entry_t)pte)
288#define	pte_store_ma(ptep, pte)	pte_load_store_ma((ptep), (pt_entry_t)pte)
289#define	pde_store_ma(ptep, pte)	pte_load_store_ma((ptep), (pt_entry_t)pte)
290
291#elif !defined(XEN)
292
293/*
294 * KPTmap is a linear mapping of the kernel page table.  It differs from the
295 * recursive mapping in two ways: (1) it only provides access to kernel page
296 * table pages, and not user page table pages, and (2) it provides access to
297 * a kernel page table page after the corresponding virtual addresses have
298 * been promoted to a 2/4MB page mapping.
299 *
300 * KPTmap is first initialized by locore to support just NPKT page table
301 * pages.  Later, it is reinitialized by pmap_bootstrap() to allow for
302 * expansion of the kernel page table.
303 */
304extern pt_entry_t *KPTmap;
305
306/*
307 * Extract from the kernel page table the physical address that is mapped by
308 * the given virtual address "va".
309 *
310 * This function may be used before pmap_bootstrap() is called.
311 */
312static __inline vm_paddr_t
313pmap_kextract(vm_offset_t va)
314{
315	vm_paddr_t pa;
316
317	if ((pa = PTD[va >> PDRSHIFT]) & PG_PS) {
318		pa = (pa & PG_PS_FRAME) | (va & PDRMASK);
319	} else {
320		/*
321		 * Beware of a concurrent promotion that changes the PDE at
322		 * this point!  For example, vtopte() must not be used to
323		 * access the PTE because it would use the new PDE.  It is,
324		 * however, safe to use the old PDE because the page table
325		 * page is preserved by the promotion.
326		 */
327		pa = KPTmap[i386_btop(va)];
328		pa = (pa & PG_FRAME) | (va & PAGE_MASK);
329	}
330	return (pa);
331}
332#endif
333
334#if !defined(XEN)
335#define PT_UPDATES_FLUSH()
336#endif
337
338#if (defined(PAE) || defined(PAE_TABLES)) && !defined(XEN)
339
340#define	pde_cmpset(pdep, old, new)	atomic_cmpset_64_i586(pdep, old, new)
341#define	pte_load_store(ptep, pte)	atomic_swap_64_i586(ptep, pte)
342#define	pte_load_clear(ptep)		atomic_swap_64_i586(ptep, 0)
343#define	pte_store(ptep, pte)		atomic_store_rel_64_i586(ptep, pte)
344
345extern pt_entry_t pg_nx;
346
347#elif !defined(PAE) && !defined(PAE_TABLES) && !defined(XEN)
348
349#define	pde_cmpset(pdep, old, new)	atomic_cmpset_int(pdep, old, new)
350#define	pte_load_store(ptep, pte)	atomic_swap_int(ptep, pte)
351#define	pte_load_clear(ptep)		atomic_swap_int(ptep, 0)
352#define	pte_store(ptep, pte) do { \
353	*(u_int *)(ptep) = (u_int)(pte); \
354} while (0)
355
356#endif /* PAE */
357
358#define	pte_clear(ptep)			pte_store(ptep, 0)
359
360#define	pde_store(pdep, pde)		pte_store(pdep, pde)
361
362#endif /* _KERNEL */
363
364/*
365 * Pmap stuff
366 */
367struct	pv_entry;
368struct	pv_chunk;
369
370struct md_page {
371	TAILQ_HEAD(,pv_entry)	pv_list;
372	int			pat_mode;
373};
374
375struct pmap {
376	struct mtx		pm_mtx;
377	pd_entry_t		*pm_pdir;	/* KVA of page directory */
378	TAILQ_HEAD(,pv_chunk)	pm_pvchunk;	/* list of mappings in pmap */
379	cpuset_t		pm_active;	/* active on cpus */
380	struct pmap_statistics	pm_stats;	/* pmap statistics */
381	LIST_ENTRY(pmap) 	pm_list;	/* List of all pmaps */
382#if defined(PAE) || defined(PAE_TABLES)
383	pdpt_entry_t		*pm_pdpt;	/* KVA of page directory pointer
384						   table */
385#endif
386	struct vm_radix		pm_root;	/* spare page table pages */
387};
388
389typedef struct pmap	*pmap_t;
390
391#ifdef _KERNEL
392extern struct pmap	kernel_pmap_store;
393#define kernel_pmap	(&kernel_pmap_store)
394
395#define	PMAP_LOCK(pmap)		mtx_lock(&(pmap)->pm_mtx)
396#define	PMAP_LOCK_ASSERT(pmap, type) \
397				mtx_assert(&(pmap)->pm_mtx, (type))
398#define	PMAP_LOCK_DESTROY(pmap)	mtx_destroy(&(pmap)->pm_mtx)
399#define	PMAP_LOCK_INIT(pmap)	mtx_init(&(pmap)->pm_mtx, "pmap", \
400				    NULL, MTX_DEF | MTX_DUPOK)
401#define	PMAP_LOCKED(pmap)	mtx_owned(&(pmap)->pm_mtx)
402#define	PMAP_MTX(pmap)		(&(pmap)->pm_mtx)
403#define	PMAP_TRYLOCK(pmap)	mtx_trylock(&(pmap)->pm_mtx)
404#define	PMAP_UNLOCK(pmap)	mtx_unlock(&(pmap)->pm_mtx)
405#endif
406
407/*
408 * For each vm_page_t, there is a list of all currently valid virtual
409 * mappings of that page.  An entry is a pv_entry_t, the list is pv_list.
410 */
411typedef struct pv_entry {
412	vm_offset_t	pv_va;		/* virtual address for mapping */
413	TAILQ_ENTRY(pv_entry)	pv_next;
414} *pv_entry_t;
415
416/*
417 * pv_entries are allocated in chunks per-process.  This avoids the
418 * need to track per-pmap assignments.
419 */
420#define	_NPCM	11
421#define	_NPCPV	336
422struct pv_chunk {
423	pmap_t			pc_pmap;
424	TAILQ_ENTRY(pv_chunk)	pc_list;
425	uint32_t		pc_map[_NPCM];	/* bitmap; 1 = free */
426	TAILQ_ENTRY(pv_chunk)	pc_lru;
427	struct pv_entry		pc_pventry[_NPCPV];
428};
429
430#ifdef	_KERNEL
431
432extern caddr_t	CADDR3;
433extern pt_entry_t *CMAP3;
434extern vm_paddr_t phys_avail[];
435extern vm_paddr_t dump_avail[];
436extern int pseflag;
437extern int pgeflag;
438extern char *ptvmmap;		/* poor name! */
439extern vm_offset_t virtual_avail;
440extern vm_offset_t virtual_end;
441
442#define	pmap_page_get_memattr(m)	((vm_memattr_t)(m)->md.pat_mode)
443#define	pmap_page_is_write_mapped(m)	(((m)->aflags & PGA_WRITEABLE) != 0)
444#define	pmap_unmapbios(va, sz)	pmap_unmapdev((va), (sz))
445
446/*
447 * Only the following functions or macros may be used before pmap_bootstrap()
448 * is called: pmap_kenter(), pmap_kextract(), pmap_kremove(), vtophys(), and
449 * vtopte().
450 */
451void	pmap_bootstrap(vm_paddr_t);
452int	pmap_cache_bits(int mode, boolean_t is_pde);
453int	pmap_change_attr(vm_offset_t, vm_size_t, int);
454void	pmap_init_pat(void);
455void	pmap_kenter(vm_offset_t va, vm_paddr_t pa);
456void	*pmap_kenter_temporary(vm_paddr_t pa, int i);
457void	pmap_kremove(vm_offset_t);
458void	*pmap_mapbios(vm_paddr_t, vm_size_t);
459void	*pmap_mapdev(vm_paddr_t, vm_size_t);
460void	*pmap_mapdev_attr(vm_paddr_t, vm_size_t, int);
461boolean_t pmap_page_is_mapped(vm_page_t m);
462void	pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma);
463void	pmap_unmapdev(vm_offset_t, vm_size_t);
464pt_entry_t *pmap_pte(pmap_t, vm_offset_t) __pure2;
465void	pmap_invalidate_page(pmap_t, vm_offset_t);
466void	pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
467void	pmap_invalidate_all(pmap_t);
468void	pmap_invalidate_cache(void);
469void	pmap_invalidate_cache_pages(vm_page_t *pages, int count);
470void	pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva,
471	    boolean_t force);
472
473#endif /* _KERNEL */
474
475#endif /* !LOCORE */
476
477#endif /* !_MACHINE_PMAP_H_ */
478