xref: /dragonfly/sys/platform/pc64/include/pmap.h (revision cf89a63b)
1 /*
2  * Copyright (c) 1991 Regents of the University of California.
3  * Copyright (c) 2003 Peter Wemm.
4  * Copyright (c) 2008 The DragonFly Project.
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. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  *
39  * Derived from hp300 version by Mike Hibler, this version by William
40  * Jolitz uses a recursive map [a pde points to the page directory] to
41  * map the page tables using the pagetables themselves. This is done to
42  * reduce the impact on kernel virtual memory for lots of sparse address
43  * space, and to reduce the cost of memory to each process.
44  *
45  * from: hp300: @(#)pmap.h	7.2 (Berkeley) 12/16/90
46  * from: @(#)pmap.h	7.4 (Berkeley) 5/12/91
47  * $FreeBSD: src/sys/i386/include/pmap.h,v 1.65.2.3 2001/10/03 07:15:37 peter Exp $
48  */
49 
50 #ifndef _MACHINE_PMAP_H_
51 #define	_MACHINE_PMAP_H_
52 
53 #include <cpu/pmap.h>
54 
55 /*
56  * Size of Kernel address space.  This is the number of page table pages
57  * (2GB each) to use for the kernel.  256 pages == 512 Gigabytes.
58  * This **MUST** be a multiple of 4 (eg: 252, 256, 260, etc).
59  */
60 #ifndef KVA_PAGES
61 #define KVA_PAGES	256
62 #endif
63 
64 /*
65  * Pte related macros.  This is complicated by having to deal with
66  * the sign extension of the 48th bit.
67  */
68 #define KVADDR(l4, l3, l2, l1) ( \
69 	((unsigned long)-1 << 47) | \
70 	((unsigned long)(l4) << PML4SHIFT) | \
71 	((unsigned long)(l3) << PDPSHIFT) | \
72 	((unsigned long)(l2) << PDRSHIFT) | \
73 	((unsigned long)(l1) << PAGE_SHIFT))
74 
75 #define UVADDR(l4, l3, l2, l1) ( \
76 	((unsigned long)(l4) << PML4SHIFT) | \
77 	((unsigned long)(l3) << PDPSHIFT) | \
78 	((unsigned long)(l2) << PDRSHIFT) | \
79 	((unsigned long)(l1) << PAGE_SHIFT))
80 
81 /*
82  * NOTE: We no longer hardwire NKPT, it is calculated in create_pagetables()
83  */
84 #define NKPML4E		1		/* number of kernel PML4 slots */
85 /* NKPDPE defined in vmparam.h */
86 
87 /*
88  * NUPDPs	512 (256 user)		number of PDPs in user page table
89  * NUPDs	512 * 512		number of PDs in user page table
90  * NUPTs	512 * 512 * 512		number of PTs in user page table
91  * NUPTEs	512 * 512 * 512 * 512	number of PTEs in user page table
92  *
93  * NUPDP_USER 	number of PDPs reserved for userland
94  * NUPTE_USER	number of PTEs reserved for userland (big number)
95  */
96 #define	NUPDP_USER	(NPML4EPG/2)
97 #define	NUPDP_TOTAL	(NPML4EPG)
98 #define	NUPD_TOTAL	(NPDPEPG * NUPDP_TOTAL)
99 #define	NUPT_TOTAL	(NPDEPG * NUPD_TOTAL)
100 #define NUPTE_TOTAL	((vm_pindex_t)NPTEPG * NUPT_TOTAL)
101 #define NUPTE_USER	((vm_pindex_t)NPTEPG * NPDEPG * NPDPEPG * NUPDP_USER)
102 
103 /*
104  * Number of 512G dmap PML4 slots (max ~254 or so but don't go over 64,
105  * which gives us 32TB of ram).  Because we cache free, empty pmaps the
106  * initialization overhead is minimal.
107  *
108  * It should be possible to bump this up to 255 (but not 256), which would
109  * be able to address a maximum of ~127TB of physical ram.
110  */
111 #define	NDMPML4E	64
112 
113 /*
114  * The *PML4I values control the layout of virtual memory.  Each PML4
115  * entry represents 512G.
116  */
117 #define	PML4PML4I	(NPML4EPG/2)	/* Index of recursive pml4 mapping */
118 
119 #define	KPML4I		(NPML4EPG-1)	/* Top 512GB for KVM */
120 #define	DMPML4I		(KPML4I-NDMPML4E) /* Next 512GBxN down for dmap */
121 
122 /*
123  * The location of KERNBASE in the last PD of the kernel's KVM (KPML4I)
124  * space.  Each PD represents 1GB.  The kernel must be placed here
125  * for the compile/link options to work properly so absolute 32-bit
126  * addressing can be used to access stuff.
127  */
128 #define	KPDPI		(NPDPEPG-2)	/* kernbase at -2GB */
129 
130 /*
131  * per-CPU data assume ~64K x SMP_MAXCPU, say up to 256 cpus
132  * in the future or 16MB of space.  Each PD represents 2MB so
133  * use NPDEPG-8 to place the per-CPU data.
134  */
135 #define	MPPML4I		KPML4I
136 #define	MPPDPI		KPDPI
137 #define	MPPTDI		(NPDEPG-8)
138 
139 /*
140  * XXX doesn't really belong here I guess...
141  */
142 #define ISA_HOLE_START    0xa0000
143 #define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START)
144 
145 #ifndef LOCORE
146 
147 #ifndef _SYS_TYPES_H_
148 #include <sys/types.h>
149 #endif
150 #ifndef _SYS_QUEUE_H_
151 #include <sys/queue.h>
152 #endif
153 #ifndef _SYS_TREE_H_
154 #include <sys/tree.h>
155 #endif
156 #ifndef _SYS_SPINLOCK_H_
157 #include <sys/spinlock.h>
158 #endif
159 #ifndef _SYS_THREAD_H_
160 #include <sys/thread.h>
161 #endif
162 #ifndef _MACHINE_TYPES_H_
163 #include <machine/types.h>
164 #endif
165 #ifndef _MACHINE_PARAM_H_
166 #include <machine/param.h>
167 #endif
168 
169 /*
170  * Address of current and alternate address space page table maps
171  * and directories.
172  */
173 #ifdef _KERNEL
174 #define	addr_PTmap	(KVADDR(PML4PML4I, 0, 0, 0))
175 #define	addr_PDmap	(KVADDR(PML4PML4I, PML4PML4I, 0, 0))
176 #define	addr_PDPmap	(KVADDR(PML4PML4I, PML4PML4I, PML4PML4I, 0))
177 #define	addr_PML4map	(KVADDR(PML4PML4I, PML4PML4I, PML4PML4I, PML4PML4I))
178 #define	addr_PML4pml4e	(addr_PML4map + (PML4PML4I * sizeof(pml4_entry_t)))
179 #define	PTmap		((pt_entry_t *)(addr_PTmap))
180 #define	PDmap		((pd_entry_t *)(addr_PDmap))
181 #define	PDPmap		((pd_entry_t *)(addr_PDPmap))
182 #define	PML4map		((pd_entry_t *)(addr_PML4map))
183 #define	PML4pml4e	((pd_entry_t *)(addr_PML4pml4e))
184 
185 extern u_int64_t KPML4phys;	/* physical address of kernel level 4 */
186 #endif
187 
188 #ifdef _KERNEL
189 
190 /*
191  * XXX
192  */
193 #define	vtophys(va)	pmap_kextract(((vm_offset_t)(va)))
194 #define	vtophys_pte(va)	((pt_entry_t)pmap_kextract(((vm_offset_t)(va))))
195 
196 #endif
197 
198 #define	pte_load_clear(pte)	atomic_readandclear_long(pte)
199 
200 static __inline void
201 pte_store(pt_entry_t *ptep, pt_entry_t pte)
202 {
203 	*ptep = pte;
204 }
205 
206 #define	pde_store(pdep, pde)	pte_store((pdep), (pde))
207 
208 /*
209  * Pmap stuff
210  */
211 struct pmap;
212 struct pv_entry;
213 struct vm_page;
214 struct vm_object;
215 struct vmspace;
216 
217 TAILQ_HEAD(md_page_pv_list, pv_entry);
218 /*
219  * vm_page structures embed a list of related pv_entry's
220  */
221 struct md_page {
222 	struct md_page_pv_list	pv_list;
223 };
224 
225 /*
226  * vm_object's representing large mappings can contain embedded pmaps
227  * to organize sharing at higher page table levels for PROT_READ and
228  * PROT_READ|PROT_WRITE maps.
229  */
230 struct md_object {
231 	struct pmap *pmap_rw;
232 	struct pmap *pmap_ro;
233 };
234 
235 /*
236  * Each machine dependent implementation is expected to
237  * keep certain statistics.  They may do this anyway they
238  * so choose, but are expected to return the statistics
239  * in the following structure.
240  *
241  * NOTE: We try to match the size of the pc32 pmap with the vkernel pmap
242  * so the same utilities (like 'ps') can be used on both.
243  */
244 struct pmap_statistics {
245 	long resident_count;    /* # of pages mapped (total) */
246 	long wired_count;       /* # of pages wired */
247 };
248 typedef struct pmap_statistics *pmap_statistics_t;
249 
250 struct pv_entry_rb_tree;
251 RB_PROTOTYPE2(pv_entry_rb_tree, pv_entry, pv_entry,
252 	      pv_entry_compare, vm_pindex_t);
253 
254 /* Types of PMAP (regular, EPT Intel, NPT Amd) */
255 #define	REGULAR_PMAP		0
256 #define	EPT_PMAP		1
257 
258 /* Bits indexes in pmap_bits */
259 #define	TYPE_IDX		0
260 #define	PG_V_IDX		1
261 #define	PG_RW_IDX		2
262 #define	PG_U_IDX		3
263 #define	PG_A_IDX		4
264 #define	PG_M_IDX		5
265 #define	PG_PS_IDX		6
266 #define	PG_G_IDX		7
267 #define	PG_W_IDX		8
268 #define	PG_MANAGED_IDX		9
269 #define	PG_DEVICE_IDX		10
270 #define	PG_N_IDX		11
271 #define	PG_BITS_SIZE		12
272 
273 #define PROTECTION_CODES_SIZE	8
274 #define PAT_INDEX_SIZE  8
275 
276 struct pmap {
277 	pml4_entry_t		*pm_pml4;	/* KVA of level 4 page table */
278 	struct pv_entry		*pm_pmlpv;	/* PV entry for pml4 */
279 	TAILQ_ENTRY(pmap)	pm_pmnode;	/* list of pmaps */
280 	RB_HEAD(pv_entry_rb_tree, pv_entry) pm_pvroot;
281 	int			pm_count;	/* reference count */
282 	cpumask_t		pm_active;	/* active on cpus */
283 	int			pm_flags;
284 	struct pmap_statistics	pm_stats;	/* pmap statistics */
285 	struct pv_entry		*pm_pvhint;	/* pv_entry lookup hint */
286 	int			pm_generation;	/* detect pvlist deletions */
287 	struct spinlock		pm_spin;
288 	struct lwkt_token	pm_token;
289 	long			pm_invgen;
290 	uint64_t		pmap_bits[PG_BITS_SIZE];
291 	int			protection_codes[PROTECTION_CODES_SIZE];
292 	pt_entry_t		pmap_cache_bits[PAT_INDEX_SIZE];
293 	pt_entry_t		pmap_cache_mask;
294 	int (*copyinstr)(const void *, void *, size_t, size_t *);
295 	int (*copyin)(const void *, void *, size_t);
296 	int (*copyout)(const void *, void *, size_t);
297 	int (*fubyte)(const void *);
298 	int (*subyte)(void *, int);
299 	long (*fuword)(const void *);
300 	int (*suword)(void *, long);
301 	int (*suword32)(void *, int);
302 };
303 
304 #define CPUMASK_LOCK		CPUMASK(SMP_MAXCPU)
305 #define CPUMASK_BIT		SMP_MAXCPU	/* for 1LLU << SMP_MAXCPU */
306 
307 #define PMAP_FLAG_SIMPLE	0x00000001
308 #define PMAP_EMULATE_AD_BITS	0x00000002
309 
310 #define pmap_resident_count(pmap) (pmap)->pm_stats.resident_count
311 
312 typedef struct pmap	*pmap_t;
313 
314 #ifdef _KERNEL
315 extern struct pmap	kernel_pmap;
316 #endif
317 
318 /*
319  * For each vm_page_t, there is a list of all currently valid virtual
320  * mappings of that page.  An entry is a pv_entry_t, the list is pv_table.
321  */
322 typedef struct pv_entry {
323 	pmap_t		pv_pmap;	/* pmap where mapping lies */
324 	vm_pindex_t	pv_pindex;	/* PTE, PT, PD, PDP, or PML4 */
325 	TAILQ_ENTRY(pv_entry)	pv_list;
326 	RB_ENTRY(pv_entry)	pv_entry;
327 	struct vm_page	*pv_m;		/* page being mapped */
328 	u_int		pv_hold;	/* interlock action */
329 	u_int		pv_flags;
330 #ifdef PMAP_DEBUG
331 	const char	*pv_func;
332 	int		pv_line;
333 #endif
334 } *pv_entry_t;
335 
336 #define PV_HOLD_LOCKED		0x80000000U
337 #define PV_HOLD_WAITING		0x40000000U
338 #define PV_HOLD_UNUSED2000	0x20000000U
339 #define PV_HOLD_MASK		0x1FFFFFFFU
340 
341 #define PV_FLAG_VMOBJECT	0x00000001U	/* shared pt in VM obj */
342 
343 #ifdef	_KERNEL
344 
345 extern caddr_t	CADDR1;
346 extern pt_entry_t *CMAP1;
347 extern vm_paddr_t dump_avail[];
348 extern vm_paddr_t avail_end;
349 extern vm_paddr_t avail_start;
350 extern vm_offset_t clean_eva;
351 extern vm_offset_t clean_sva;
352 extern char *ptvmmap;		/* poor name! */
353 
354 typedef struct vm_page *vm_page_t;
355 typedef char vm_memattr_t;
356 
357 void	pmap_release(struct pmap *pmap);
358 void	pmap_interlock_wait (struct vmspace *);
359 void	pmap_bootstrap (vm_paddr_t *);
360 void	*pmap_mapbios(vm_paddr_t, vm_size_t);
361 void	*pmap_mapdev (vm_paddr_t, vm_size_t);
362 void	*pmap_mapdev_attr(vm_paddr_t, vm_size_t, int);
363 void	*pmap_mapdev_uncacheable(vm_paddr_t, vm_size_t);
364 void	pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma);
365 void	pmap_unmapdev (vm_offset_t, vm_size_t);
366 struct vm_page *pmap_use_pt (pmap_t, vm_offset_t);
367 void	pmap_set_opt (void);
368 void	pmap_init_pat(void);
369 vm_paddr_t pmap_kextract(vm_offset_t);
370 void	pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t);
371 void	pmap_invalidate_cache_pages(vm_page_t *pages, int count);
372 void	pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva);
373 
374 static __inline int
375 pmap_emulate_ad_bits(pmap_t pmap) {
376 	return pmap->pm_flags & PMAP_EMULATE_AD_BITS;
377 }
378 
379 #endif /* _KERNEL */
380 
381 #endif /* !LOCORE */
382 
383 #endif /* !_MACHINE_PMAP_H_ */
384