xref: /netbsd-src/sys/arch/x86/include/pmap.h (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: pmap.h,v 1.33 2010/07/24 00:45:56 jym Exp $	*/
2 
3 /*
4  *
5  * Copyright (c) 1997 Charles D. Cranor and Washington University.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgment:
18  *      This product includes software developed by Charles D. Cranor and
19  *      Washington University.
20  * 4. The name of the author may not be used to endorse or promote products
21  *    derived from this software without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
24  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
27  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
28  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
32  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33  */
34 
35 /*
36  * Copyright (c) 2001 Wasabi Systems, Inc.
37  * All rights reserved.
38  *
39  * Written by Frank van der Linden for Wasabi Systems, Inc.
40  *
41  * Redistribution and use in source and binary forms, with or without
42  * modification, are permitted provided that the following conditions
43  * are met:
44  * 1. Redistributions of source code must retain the above copyright
45  *    notice, this list of conditions and the following disclaimer.
46  * 2. Redistributions in binary form must reproduce the above copyright
47  *    notice, this list of conditions and the following disclaimer in the
48  *    documentation and/or other materials provided with the distribution.
49  * 3. All advertising materials mentioning features or use of this software
50  *    must display the following acknowledgement:
51  *      This product includes software developed for the NetBSD Project by
52  *      Wasabi Systems, Inc.
53  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
54  *    or promote products derived from this software without specific prior
55  *    written permission.
56  *
57  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
58  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
59  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
60  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
61  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
62  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
63  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
64  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
65  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
66  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
67  * POSSIBILITY OF SUCH DAMAGE.
68  */
69 
70 /*
71  * pmap.h: see pmap.c for the history of this pmap module.
72  */
73 
74 #ifndef _X86_PMAP_H_
75 #define	_X86_PMAP_H_
76 
77 #define ptei(VA)	(((VA_SIGN_POS(VA)) & L1_MASK) >> L1_SHIFT)
78 
79 /*
80  * pl*_pi: index in the ptp page for a pde mapping a VA.
81  * (pl*_i below is the index in the virtual array of all pdes per level)
82  */
83 #define pl1_pi(VA)	(((VA_SIGN_POS(VA)) & L1_MASK) >> L1_SHIFT)
84 #define pl2_pi(VA)	(((VA_SIGN_POS(VA)) & L2_MASK) >> L2_SHIFT)
85 #define pl3_pi(VA)	(((VA_SIGN_POS(VA)) & L3_MASK) >> L3_SHIFT)
86 #define pl4_pi(VA)	(((VA_SIGN_POS(VA)) & L4_MASK) >> L4_SHIFT)
87 
88 /*
89  * pl*_i: generate index into pde/pte arrays in virtual space
90  */
91 #define pl1_i(VA)	(((VA_SIGN_POS(VA)) & L1_FRAME) >> L1_SHIFT)
92 #define pl2_i(VA)	(((VA_SIGN_POS(VA)) & L2_FRAME) >> L2_SHIFT)
93 #define pl3_i(VA)	(((VA_SIGN_POS(VA)) & L3_FRAME) >> L3_SHIFT)
94 #define pl4_i(VA)	(((VA_SIGN_POS(VA)) & L4_FRAME) >> L4_SHIFT)
95 #define pl_i(va, lvl) \
96         (((VA_SIGN_POS(va)) & ptp_masks[(lvl)-1]) >> ptp_shifts[(lvl)-1])
97 
98 #define	pl_i_roundup(va, lvl)	pl_i((va)+ ~ptp_masks[(lvl)-1], (lvl))
99 
100 /*
101  * PTP macros:
102  *   a PTP's index is the PD index of the PDE that points to it
103  *   a PTP's offset is the byte-offset in the PTE space that this PTP is at
104  *   a PTP's VA is the first VA mapped by that PTP
105  */
106 
107 #define ptp_va2o(va, lvl)	(pl_i(va, (lvl)+1) * PAGE_SIZE)
108 
109 /* size of a PDP: usually one page, except for PAE */
110 #ifdef PAE
111 #define PDP_SIZE 4
112 #else
113 #define PDP_SIZE 1
114 #endif
115 
116 
117 #if defined(_KERNEL)
118 /*
119  * pmap data structures: see pmap.c for details of locking.
120  */
121 
122 /*
123  * we maintain a list of all non-kernel pmaps
124  */
125 
126 LIST_HEAD(pmap_head, pmap); /* struct pmap_head: head of a pmap list */
127 
128 /*
129  * the pmap structure
130  *
131  * note that the pm_obj contains the simple_lock, the reference count,
132  * page list, and number of PTPs within the pmap.
133  *
134  * pm_lock is the same as the spinlock for vm object 0. Changes to
135  * the other objects may only be made if that lock has been taken
136  * (the other object locks are only used when uvm_pagealloc is called)
137  *
138  * XXX If we ever support processor numbers higher than 31, we'll have
139  * XXX to rethink the CPU mask.
140  */
141 
142 struct pmap {
143 	struct uvm_object pm_obj[PTP_LEVELS-1]; /* objects for lvl >= 1) */
144 #define	pm_lock	pm_obj[0].vmobjlock
145 	LIST_ENTRY(pmap) pm_list;	/* list (lck by pm_list lock) */
146 	pd_entry_t *pm_pdir;		/* VA of PD (lck by object lock) */
147 	paddr_t pm_pdirpa[PDP_SIZE];	/* PA of PDs (read-only after create) */
148 	struct vm_page *pm_ptphint[PTP_LEVELS-1];
149 					/* pointer to a PTP in our pmap */
150 	struct pmap_statistics pm_stats;  /* pmap stats (lck by object lock) */
151 
152 #if !defined(__x86_64__)
153 	vaddr_t pm_hiexec;		/* highest executable mapping */
154 #endif /* !defined(__x86_64__) */
155 	int pm_flags;			/* see below */
156 
157 	union descriptor *pm_ldt;	/* user-set LDT */
158 	size_t pm_ldt_len;		/* size of LDT in bytes */
159 	int pm_ldt_sel;			/* LDT selector */
160 	uint32_t pm_cpus;		/* mask of CPUs using pmap */
161 	uint32_t pm_kernel_cpus;	/* mask of CPUs using kernel part
162 					 of pmap */
163 };
164 
165 /* macro to access pm_pdirpa slots */
166 #ifdef PAE
167 #define pmap_pdirpa(pmap, index) \
168 	((pmap)->pm_pdirpa[l2tol3(index)] + l2tol2(index) * sizeof(pd_entry_t))
169 #else
170 #define pmap_pdirpa(pmap, index) \
171 	((pmap)->pm_pdirpa[0] + (index) * sizeof(pd_entry_t))
172 #endif
173 
174 /*
175  * MD flags that we use for pmap_enter and pmap_kenter_pa:
176  */
177 
178 /*
179  * global kernel variables
180  */
181 
182 /*
183  * PDPpaddr is the physical address of the kernel's PDP.
184  * - i386 non-PAE and amd64: PDPpaddr corresponds directly to the %cr3
185  * value associated to the kernel process, proc0.
186  * - i386 PAE: it still represents the PA of the kernel's PDP (L2). Due to
187  * the L3 PD, it cannot be considered as the equivalent of a %cr3 any more.
188  * - Xen: it corresponds to the PFN of the kernel's PDP.
189  */
190 extern u_long PDPpaddr;
191 
192 extern int pmap_pg_g;			/* do we support PG_G? */
193 extern long nkptp[PTP_LEVELS];
194 
195 /*
196  * macros
197  */
198 
199 #define	pmap_resident_count(pmap)	((pmap)->pm_stats.resident_count)
200 #define	pmap_wired_count(pmap)		((pmap)->pm_stats.wired_count)
201 
202 #define pmap_clear_modify(pg)		pmap_clear_attrs(pg, PG_M)
203 #define pmap_clear_reference(pg)	pmap_clear_attrs(pg, PG_U)
204 #define pmap_copy(DP,SP,D,L,S)
205 #define pmap_is_modified(pg)		pmap_test_attrs(pg, PG_M)
206 #define pmap_is_referenced(pg)		pmap_test_attrs(pg, PG_U)
207 #define pmap_move(DP,SP,D,L,S)
208 #define pmap_phys_address(ppn)		x86_ptob(ppn)
209 #define pmap_valid_entry(E) 		((E) & PG_V) /* is PDE or PTE valid? */
210 
211 
212 /*
213  * prototypes
214  */
215 
216 void		pmap_activate(struct lwp *);
217 void		pmap_bootstrap(vaddr_t);
218 bool		pmap_clear_attrs(struct vm_page *, unsigned);
219 void		pmap_deactivate(struct lwp *);
220 void		pmap_page_remove (struct vm_page *);
221 void		pmap_remove(struct pmap *, vaddr_t, vaddr_t);
222 bool		pmap_test_attrs(struct vm_page *, unsigned);
223 void		pmap_write_protect(struct pmap *, vaddr_t, vaddr_t, vm_prot_t);
224 void		pmap_load(void);
225 paddr_t		pmap_init_tmp_pgtbl(paddr_t);
226 void		pmap_remove_all(struct pmap *);
227 void		pmap_ldt_sync(struct pmap *);
228 
229 void		pmap_emap_enter(vaddr_t, paddr_t, vm_prot_t);
230 void		pmap_emap_remove(vaddr_t, vsize_t);
231 void		pmap_emap_sync(bool);
232 
233 void		pmap_map_ptes(struct pmap *, struct pmap **, pd_entry_t **,
234 		    pd_entry_t * const **);
235 void		pmap_unmap_ptes(struct pmap *, struct pmap *);
236 
237 int		pmap_pdes_invalid(vaddr_t, pd_entry_t * const *, pd_entry_t *);
238 
239 vaddr_t reserve_dumppages(vaddr_t); /* XXX: not a pmap fn */
240 
241 void	pmap_tlb_shootdown(pmap_t, vaddr_t, vaddr_t, pt_entry_t);
242 void	pmap_tlb_shootwait(void);
243 
244 #define	__HAVE_PMAP_EMAP
245 
246 #define PMAP_GROWKERNEL		/* turn on pmap_growkernel interface */
247 #define PMAP_FORK		/* turn on pmap_fork interface */
248 
249 /*
250  * Do idle page zero'ing uncached to avoid polluting the cache.
251  */
252 bool	pmap_pageidlezero(paddr_t);
253 #define	PMAP_PAGEIDLEZERO(pa)	pmap_pageidlezero((pa))
254 
255 /*
256  * inline functions
257  */
258 
259 __inline static bool __unused
260 pmap_pdes_valid(vaddr_t va, pd_entry_t * const *pdes, pd_entry_t *lastpde)
261 {
262 	return pmap_pdes_invalid(va, pdes, lastpde) == 0;
263 }
264 
265 /*
266  * pmap_update_pg: flush one page from the TLB (or flush the whole thing
267  *	if hardware doesn't support one-page flushing)
268  */
269 
270 __inline static void __unused
271 pmap_update_pg(vaddr_t va)
272 {
273 	invlpg(va);
274 }
275 
276 /*
277  * pmap_update_2pg: flush two pages from the TLB
278  */
279 
280 __inline static void __unused
281 pmap_update_2pg(vaddr_t va, vaddr_t vb)
282 {
283 	invlpg(va);
284 	invlpg(vb);
285 }
286 
287 /*
288  * pmap_page_protect: change the protection of all recorded mappings
289  *	of a managed page
290  *
291  * => this function is a frontend for pmap_page_remove/pmap_clear_attrs
292  * => we only have to worry about making the page more protected.
293  *	unprotecting a page is done on-demand at fault time.
294  */
295 
296 __inline static void __unused
297 pmap_page_protect(struct vm_page *pg, vm_prot_t prot)
298 {
299 	if ((prot & VM_PROT_WRITE) == 0) {
300 		if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) {
301 			(void) pmap_clear_attrs(pg, PG_RW);
302 		} else {
303 			pmap_page_remove(pg);
304 		}
305 	}
306 }
307 
308 /*
309  * pmap_protect: change the protection of pages in a pmap
310  *
311  * => this function is a frontend for pmap_remove/pmap_write_protect
312  * => we only have to worry about making the page more protected.
313  *	unprotecting a page is done on-demand at fault time.
314  */
315 
316 __inline static void __unused
317 pmap_protect(struct pmap *pmap, vaddr_t sva, vaddr_t eva, vm_prot_t prot)
318 {
319 	if ((prot & VM_PROT_WRITE) == 0) {
320 		if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) {
321 			pmap_write_protect(pmap, sva, eva, prot);
322 		} else {
323 			pmap_remove(pmap, sva, eva);
324 		}
325 	}
326 }
327 
328 /*
329  * various address inlines
330  *
331  *  vtopte: return a pointer to the PTE mapping a VA, works only for
332  *  user and PT addresses
333  *
334  *  kvtopte: return a pointer to the PTE mapping a kernel VA
335  */
336 
337 #include <lib/libkern/libkern.h>
338 
339 static __inline pt_entry_t * __unused
340 vtopte(vaddr_t va)
341 {
342 
343 	KASSERT(va < VM_MIN_KERNEL_ADDRESS);
344 
345 	return (PTE_BASE + pl1_i(va));
346 }
347 
348 static __inline pt_entry_t * __unused
349 kvtopte(vaddr_t va)
350 {
351 	pd_entry_t *pde;
352 
353 	KASSERT(va >= VM_MIN_KERNEL_ADDRESS);
354 
355 	pde = L2_BASE + pl2_i(va);
356 	if (*pde & PG_PS)
357 		return ((pt_entry_t *)pde);
358 
359 	return (PTE_BASE + pl1_i(va));
360 }
361 
362 paddr_t vtophys(vaddr_t);
363 vaddr_t	pmap_map(vaddr_t, paddr_t, paddr_t, vm_prot_t);
364 void	pmap_cpu_init_early(struct cpu_info *);
365 void	pmap_cpu_init_late(struct cpu_info *);
366 bool	sse2_idlezero_page(void *);
367 
368 
369 #ifdef XEN
370 
371 #define XPTE_MASK	L1_FRAME
372 /* XPTE_SHIFT = L1_SHIFT - log2(sizeof(pt_entry_t)) */
373 #if defined(__x86_64__) || defined(PAE)
374 #define XPTE_SHIFT	9
375 #else
376 #define XPTE_SHIFT	10
377 #endif
378 
379 /* PTE access inline fuctions */
380 
381 /*
382  * Get the machine address of the pointed pte
383  * We use hardware MMU to get value so works only for levels 1-3
384  */
385 
386 static __inline paddr_t
387 xpmap_ptetomach(pt_entry_t *pte)
388 {
389 	pt_entry_t *up_pte;
390 	vaddr_t va = (vaddr_t) pte;
391 
392 	va = ((va & XPTE_MASK) >> XPTE_SHIFT) | (vaddr_t) PTE_BASE;
393 	up_pte = (pt_entry_t *) va;
394 
395 	return (paddr_t) (((*up_pte) & PG_FRAME) + (((vaddr_t) pte) & (~PG_FRAME & ~VA_SIGN_MASK)));
396 }
397 
398 /*
399  * xpmap_update()
400  * Update an active pt entry with Xen
401  * Equivalent to *pte = npte
402  */
403 
404 static __inline void
405 xpmap_update (pt_entry_t *pte, pt_entry_t npte)
406 {
407         int s = splvm();
408 
409         xpq_queue_pte_update(xpmap_ptetomach(pte), npte);
410         xpq_flush_queue();
411         splx(s);
412 }
413 
414 
415 /* Xen helpers to change bits of a pte */
416 #define XPMAP_UPDATE_DIRECT	1	/* Update direct map entry flags too */
417 
418 paddr_t	vtomach(vaddr_t);
419 #define vtomfn(va) (vtomach(va) >> PAGE_SHIFT)
420 
421 #endif	/* XEN */
422 
423 /* pmap functions with machine addresses */
424 void	pmap_kenter_ma(vaddr_t, paddr_t, vm_prot_t, u_int);
425 int	pmap_enter_ma(struct pmap *, vaddr_t, paddr_t, paddr_t,
426 	    vm_prot_t, u_int, int);
427 bool	pmap_extract_ma(pmap_t, vaddr_t, paddr_t *);
428 
429 /*
430  * Hooks for the pool allocator.
431  */
432 #define	POOL_VTOPHYS(va)	vtophys((vaddr_t) (va))
433 
434 /*
435  * TLB shootdown mailbox.
436  */
437 
438 struct pmap_mbox {
439 	volatile void		*mb_pointer;
440 	volatile uintptr_t	mb_addr1;
441 	volatile uintptr_t	mb_addr2;
442 	volatile uintptr_t	mb_head;
443 	volatile uintptr_t	mb_tail;
444 	volatile uintptr_t	mb_global;
445 };
446 
447 #endif /* _KERNEL */
448 
449 #endif /* _X86_PMAP_H_ */
450