xref: /netbsd-src/sys/arch/x86/include/pmap.h (revision 7f21db1c0118155e0dd40b75182e30c589d9f63e)
1 /*	$NetBSD: pmap.h,v 1.29 2010/02/09 22:51:13 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 #ifdef PAE
148 	paddr_t pm_pdirpa[PDP_SIZE];
149 #else
150 	paddr_t pm_pdirpa;		/* PA of PD (read-only after create) */
151 #endif
152 	struct vm_page *pm_ptphint[PTP_LEVELS-1];
153 					/* pointer to a PTP in our pmap */
154 	struct pmap_statistics pm_stats;  /* pmap stats (lck by object lock) */
155 
156 #if !defined(__x86_64__)
157 	vaddr_t pm_hiexec;		/* highest executable mapping */
158 #endif /* !defined(__x86_64__) */
159 	int pm_flags;			/* see below */
160 
161 	union descriptor *pm_ldt;	/* user-set LDT */
162 	size_t pm_ldt_len;		/* size of LDT in bytes */
163 	int pm_ldt_sel;			/* LDT selector */
164 	uint32_t pm_cpus;		/* mask of CPUs using pmap */
165 	uint32_t pm_kernel_cpus;	/* mask of CPUs using kernel part
166 					 of pmap */
167 };
168 
169 /* macro to access pm_pdirpa */
170 #ifdef PAE
171 #define pmap_pdirpa(pmap, index) \
172 	((pmap)->pm_pdirpa[l2tol3(index)] + l2tol2(index) * sizeof(pd_entry_t))
173 #else
174 #define pmap_pdirpa(pmap, index) \
175 	((pmap)->pm_pdirpa + (index) * sizeof(pd_entry_t))
176 #endif
177 
178 /*
179  * MD flags that we use for pmap_enter and pmap_kenter_pa:
180  */
181 #define PMAP_NOCACHE	0x01000000	/* set the non-cacheable bit */
182 
183 /*
184  * global kernel variables
185  */
186 
187 /* PDPpaddr: is the physical address of the kernel's PDP */
188 extern u_long PDPpaddr;
189 
190 extern int pmap_pg_g;			/* do we support PG_G? */
191 extern long nkptp[PTP_LEVELS];
192 
193 /*
194  * macros
195  */
196 
197 #define	pmap_resident_count(pmap)	((pmap)->pm_stats.resident_count)
198 #define	pmap_wired_count(pmap)		((pmap)->pm_stats.wired_count)
199 
200 #define pmap_clear_modify(pg)		pmap_clear_attrs(pg, PG_M)
201 #define pmap_clear_reference(pg)	pmap_clear_attrs(pg, PG_U)
202 #define pmap_copy(DP,SP,D,L,S)
203 #define pmap_is_modified(pg)		pmap_test_attrs(pg, PG_M)
204 #define pmap_is_referenced(pg)		pmap_test_attrs(pg, PG_U)
205 #define pmap_move(DP,SP,D,L,S)
206 #define pmap_phys_address(ppn)		x86_ptob(ppn)
207 #define pmap_valid_entry(E) 		((E) & PG_V) /* is PDE or PTE valid? */
208 
209 
210 /*
211  * prototypes
212  */
213 
214 void		pmap_activate(struct lwp *);
215 void		pmap_bootstrap(vaddr_t);
216 bool		pmap_clear_attrs(struct vm_page *, unsigned);
217 void		pmap_deactivate(struct lwp *);
218 void		pmap_page_remove (struct vm_page *);
219 void		pmap_remove(struct pmap *, vaddr_t, vaddr_t);
220 bool		pmap_test_attrs(struct vm_page *, unsigned);
221 void		pmap_write_protect(struct pmap *, vaddr_t, vaddr_t, vm_prot_t);
222 void		pmap_load(void);
223 paddr_t		pmap_init_tmp_pgtbl(paddr_t);
224 void		pmap_remove_all(struct pmap *);
225 void		pmap_ldt_sync(struct pmap *);
226 
227 void		pmap_emap_enter(vaddr_t, paddr_t, vm_prot_t);
228 void		pmap_emap_remove(vaddr_t, vsize_t);
229 void		pmap_emap_sync(bool);
230 
231 vaddr_t reserve_dumppages(vaddr_t); /* XXX: not a pmap fn */
232 
233 void	pmap_tlb_shootdown(pmap_t, vaddr_t, vaddr_t, pt_entry_t);
234 void	pmap_tlb_shootwait(void);
235 
236 #define	__HAVE_PMAP_EMAP
237 
238 #define PMAP_GROWKERNEL		/* turn on pmap_growkernel interface */
239 #define PMAP_FORK		/* turn on pmap_fork interface */
240 
241 /*
242  * Do idle page zero'ing uncached to avoid polluting the cache.
243  */
244 bool	pmap_pageidlezero(paddr_t);
245 #define	PMAP_PAGEIDLEZERO(pa)	pmap_pageidlezero((pa))
246 
247 /*
248  * inline functions
249  */
250 
251 /*
252  * pmap_update_pg: flush one page from the TLB (or flush the whole thing
253  *	if hardware doesn't support one-page flushing)
254  */
255 
256 __inline static void __unused
257 pmap_update_pg(vaddr_t va)
258 {
259 	invlpg(va);
260 }
261 
262 /*
263  * pmap_update_2pg: flush two pages from the TLB
264  */
265 
266 __inline static void __unused
267 pmap_update_2pg(vaddr_t va, vaddr_t vb)
268 {
269 	invlpg(va);
270 	invlpg(vb);
271 }
272 
273 /*
274  * pmap_page_protect: change the protection of all recorded mappings
275  *	of a managed page
276  *
277  * => this function is a frontend for pmap_page_remove/pmap_clear_attrs
278  * => we only have to worry about making the page more protected.
279  *	unprotecting a page is done on-demand at fault time.
280  */
281 
282 __inline static void __unused
283 pmap_page_protect(struct vm_page *pg, vm_prot_t prot)
284 {
285 	if ((prot & VM_PROT_WRITE) == 0) {
286 		if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) {
287 			(void) pmap_clear_attrs(pg, PG_RW);
288 		} else {
289 			pmap_page_remove(pg);
290 		}
291 	}
292 }
293 
294 /*
295  * pmap_protect: change the protection of pages in a pmap
296  *
297  * => this function is a frontend for pmap_remove/pmap_write_protect
298  * => we only have to worry about making the page more protected.
299  *	unprotecting a page is done on-demand at fault time.
300  */
301 
302 __inline static void __unused
303 pmap_protect(struct pmap *pmap, vaddr_t sva, vaddr_t eva, vm_prot_t prot)
304 {
305 	if ((prot & VM_PROT_WRITE) == 0) {
306 		if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) {
307 			pmap_write_protect(pmap, sva, eva, prot);
308 		} else {
309 			pmap_remove(pmap, sva, eva);
310 		}
311 	}
312 }
313 
314 /*
315  * various address inlines
316  *
317  *  vtopte: return a pointer to the PTE mapping a VA, works only for
318  *  user and PT addresses
319  *
320  *  kvtopte: return a pointer to the PTE mapping a kernel VA
321  */
322 
323 #include <lib/libkern/libkern.h>
324 
325 static __inline pt_entry_t * __unused
326 vtopte(vaddr_t va)
327 {
328 
329 	KASSERT(va < VM_MIN_KERNEL_ADDRESS);
330 
331 	return (PTE_BASE + pl1_i(va));
332 }
333 
334 static __inline pt_entry_t * __unused
335 kvtopte(vaddr_t va)
336 {
337 	pd_entry_t *pde;
338 
339 	KASSERT(va >= VM_MIN_KERNEL_ADDRESS);
340 
341 	pde = L2_BASE + pl2_i(va);
342 	if (*pde & PG_PS)
343 		return ((pt_entry_t *)pde);
344 
345 	return (PTE_BASE + pl1_i(va));
346 }
347 
348 paddr_t vtophys(vaddr_t);
349 vaddr_t	pmap_map(vaddr_t, paddr_t, paddr_t, vm_prot_t);
350 void	pmap_cpu_init_early(struct cpu_info *);
351 void	pmap_cpu_init_late(struct cpu_info *);
352 bool	sse2_idlezero_page(void *);
353 
354 
355 #ifdef XEN
356 
357 #define XPTE_MASK	L1_FRAME
358 /* XPTE_SHIFT = L1_SHIFT - log2(sizeof(pt_entry_t)) */
359 #if defined(__x86_64__) || defined(PAE)
360 #define XPTE_SHIFT	9
361 #else
362 #define XPTE_SHIFT	10
363 #endif
364 
365 /* PTE access inline fuctions */
366 
367 /*
368  * Get the machine address of the pointed pte
369  * We use hardware MMU to get value so works only for levels 1-3
370  */
371 
372 static __inline paddr_t
373 xpmap_ptetomach(pt_entry_t *pte)
374 {
375 	pt_entry_t *up_pte;
376 	vaddr_t va = (vaddr_t) pte;
377 
378 	va = ((va & XPTE_MASK) >> XPTE_SHIFT) | (vaddr_t) PTE_BASE;
379 	up_pte = (pt_entry_t *) va;
380 
381 	return (paddr_t) (((*up_pte) & PG_FRAME) + (((vaddr_t) pte) & (~PG_FRAME & ~VA_SIGN_MASK)));
382 }
383 
384 /*
385  * xpmap_update()
386  * Update an active pt entry with Xen
387  * Equivalent to *pte = npte
388  */
389 
390 static __inline void
391 xpmap_update (pt_entry_t *pte, pt_entry_t npte)
392 {
393         int s = splvm();
394 
395         xpq_queue_pte_update(xpmap_ptetomach(pte), npte);
396         xpq_flush_queue();
397         splx(s);
398 }
399 
400 
401 /* Xen helpers to change bits of a pte */
402 #define XPMAP_UPDATE_DIRECT	1	/* Update direct map entry flags too */
403 
404 /* pmap functions with machine addresses */
405 void	pmap_kenter_ma(vaddr_t, paddr_t, vm_prot_t, u_int);
406 int	pmap_enter_ma(struct pmap *, vaddr_t, paddr_t, paddr_t,
407 	    vm_prot_t, u_int, int);
408 bool	pmap_extract_ma(pmap_t, vaddr_t, paddr_t *);
409 
410 paddr_t	vtomach(vaddr_t);
411 #define vtomfn(va) (vtomach(va) >> PAGE_SHIFT)
412 
413 #endif	/* XEN */
414 
415 /*
416  * Hooks for the pool allocator.
417  */
418 #define	POOL_VTOPHYS(va)	vtophys((vaddr_t) (va))
419 
420 /*
421  * TLB shootdown mailbox.
422  */
423 
424 struct pmap_mbox {
425 	volatile void		*mb_pointer;
426 	volatile uintptr_t	mb_addr1;
427 	volatile uintptr_t	mb_addr2;
428 	volatile uintptr_t	mb_head;
429 	volatile uintptr_t	mb_tail;
430 	volatile uintptr_t	mb_global;
431 };
432 
433 #endif /* _KERNEL */
434 
435 #endif /* _X86_PMAP_H_ */
436