xref: /netbsd-src/sys/arch/xen/x86/xen_pmap.c (revision ca453df649ce9db45b64d73678ba06cbccf9aa11)
1 /*	$NetBSD: xen_pmap.c,v 1.3 2011/06/12 03:35:50 rmind Exp $	*/
2 
3 /*
4  * Copyright (c) 2007 Manuel Bouyer.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25  *
26  */
27 
28 /*
29  * Copyright (c) 2006 Mathieu Ropert <mro@adviseo.fr>
30  *
31  * Permission to use, copy, modify, and distribute this software for any
32  * purpose with or without fee is hereby granted, provided that the above
33  * copyright notice and this permission notice appear in all copies.
34  *
35  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
36  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
37  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
38  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
39  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
40  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
41  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
42  */
43 
44 /*
45  * Copyright (c) 1997 Charles D. Cranor and Washington University.
46  * All rights reserved.
47  *
48  * Redistribution and use in source and binary forms, with or without
49  * modification, are permitted provided that the following conditions
50  * are met:
51  * 1. Redistributions of source code must retain the above copyright
52  *    notice, this list of conditions and the following disclaimer.
53  * 2. Redistributions in binary form must reproduce the above copyright
54  *    notice, this list of conditions and the following disclaimer in the
55  *    documentation and/or other materials provided with the distribution.
56  *
57  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
58  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
59  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
60  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
61  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
62  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
63  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
64  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
65  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
66  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright 2001 (c) Wasabi Systems, Inc.
71  * All rights reserved.
72  *
73  * Written by Frank van der Linden for Wasabi Systems, Inc.
74  *
75  * Redistribution and use in source and binary forms, with or without
76  * modification, are permitted provided that the following conditions
77  * are met:
78  * 1. Redistributions of source code must retain the above copyright
79  *    notice, this list of conditions and the following disclaimer.
80  * 2. Redistributions in binary form must reproduce the above copyright
81  *    notice, this list of conditions and the following disclaimer in the
82  *    documentation and/or other materials provided with the distribution.
83  * 3. All advertising materials mentioning features or use of this software
84  *    must display the following acknowledgement:
85  *      This product includes software developed for the NetBSD Project by
86  *      Wasabi Systems, Inc.
87  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
88  *    or promote products derived from this software without specific prior
89  *    written permission.
90  *
91  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
92  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
93  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
94  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
95  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
96  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
97  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
98  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
99  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
100  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
101  * POSSIBILITY OF SUCH DAMAGE.
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: xen_pmap.c,v 1.3 2011/06/12 03:35:50 rmind Exp $");
106 
107 #include "opt_user_ldt.h"
108 #include "opt_lockdebug.h"
109 #include "opt_multiprocessor.h"
110 #include "opt_xen.h"
111 #if !defined(__x86_64__)
112 #include "opt_kstack_dr0.h"
113 #endif /* !defined(__x86_64__) */
114 
115 #include <sys/param.h>
116 #include <sys/systm.h>
117 #include <sys/proc.h>
118 #include <sys/pool.h>
119 #include <sys/kernel.h>
120 #include <sys/atomic.h>
121 #include <sys/cpu.h>
122 #include <sys/intr.h>
123 #include <sys/xcall.h>
124 
125 #include <uvm/uvm.h>
126 
127 #include <dev/isa/isareg.h>
128 
129 #include <machine/specialreg.h>
130 #include <machine/gdt.h>
131 #include <machine/isa_machdep.h>
132 #include <machine/cpuvar.h>
133 
134 #include <x86/pmap.h>
135 #include <x86/pmap_pv.h>
136 
137 #include <x86/i82489reg.h>
138 #include <x86/i82489var.h>
139 
140 #ifdef XEN
141 #include <xen/xen3-public/xen.h>
142 #include <xen/hypervisor.h>
143 #endif
144 
145 /* flag to be used for kernel mappings: PG_u on Xen/amd64, 0 otherwise */
146 #if defined(XEN) && defined(__x86_64__)
147 #define PG_k PG_u
148 #else
149 #define PG_k 0
150 #endif
151 
152 #define COUNT(x)	/* nothing */
153 
154 static pd_entry_t * const alternate_pdes[] = APDES_INITIALIZER;
155 extern pd_entry_t * const normal_pdes[];
156 
157 extern paddr_t pmap_pa_start; /* PA of first physical page for this domain */
158 extern paddr_t pmap_pa_end;   /* PA of last physical page for this domain */
159 
160 void
161 pmap_apte_flush(struct pmap *pmap)
162 {
163 
164 	KASSERT(kpreempt_disabled());
165 
166 	/*
167 	 * Flush the APTE mapping from all other CPUs that
168 	 * are using the pmap we are using (who's APTE space
169 	 * is the one we've just modified).
170 	 *
171 	 * XXXthorpej -- find a way to defer the IPI.
172 	 */
173 	pmap_tlb_shootdown(pmap, (vaddr_t)-1LL, 0, TLBSHOOT_APTE);
174 	pmap_tlb_shootnow();
175 }
176 
177 /*
178  * Unmap the content of APDP PDEs
179  */
180 void
181 pmap_unmap_apdp(void)
182 {
183 	int i;
184 
185 	for (i = 0; i < PDP_SIZE; i++) {
186 		pmap_pte_set(APDP_PDE+i, 0);
187 #if defined (PAE)
188 		/* clear shadow entries too */
189 		pmap_pte_set(APDP_PDE_SHADOW+i, 0);
190 #endif
191 	}
192 }
193 
194 /*
195  * pmap_map_ptes: map a pmap's PTEs into KVM and lock them in
196  *
197  * => we lock enough pmaps to keep things locked in
198  * => must be undone with pmap_unmap_ptes before returning
199  */
200 
201 void
202 pmap_map_ptes(struct pmap *pmap, struct pmap **pmap2,
203 	      pd_entry_t **ptepp, pd_entry_t * const **pdeppp)
204 {
205 	pd_entry_t opde, npde;
206 	struct pmap *ourpmap;
207 	struct cpu_info *ci;
208 	struct lwp *l;
209 	bool iscurrent;
210 	uint64_t ncsw;
211 	int s;
212 
213 	/* the kernel's pmap is always accessible */
214 	if (pmap == pmap_kernel()) {
215 		*pmap2 = NULL;
216 		*ptepp = PTE_BASE;
217 		*pdeppp = normal_pdes;
218 		return;
219 	}
220 	KASSERT(kpreempt_disabled());
221 
222  retry:
223 	l = curlwp;
224 	ncsw = l->l_ncsw;
225  	ourpmap = NULL;
226 	ci = curcpu();
227 #if defined(__x86_64__)
228 	/*
229 	 * curmap can only be pmap_kernel so at this point
230 	 * pmap_is_curpmap is always false
231 	 */
232 	iscurrent = 0;
233 	ourpmap = pmap_kernel();
234 #else /* __x86_64__*/
235 	if (ci->ci_want_pmapload &&
236 	    vm_map_pmap(&l->l_proc->p_vmspace->vm_map) == pmap) {
237 		pmap_load();
238 		if (l->l_ncsw != ncsw)
239 			goto retry;
240 	}
241 	iscurrent = pmap_is_curpmap(pmap);
242 	/* if curpmap then we are always mapped */
243 	if (iscurrent) {
244 		mutex_enter(pmap->pm_lock);
245 		*pmap2 = NULL;
246 		*ptepp = PTE_BASE;
247 		*pdeppp = normal_pdes;
248 		goto out;
249 	}
250 	ourpmap = ci->ci_pmap;
251 #endif /* __x86_64__ */
252 
253 	/* need to lock both curpmap and pmap: use ordered locking */
254 	pmap_reference(ourpmap);
255 	if ((uintptr_t) pmap < (uintptr_t) ourpmap) {
256 		mutex_enter(pmap->pm_lock);
257 		mutex_enter(ourpmap->pm_lock);
258 	} else {
259 		mutex_enter(ourpmap->pm_lock);
260 		mutex_enter(pmap->pm_lock);
261 	}
262 
263 	if (l->l_ncsw != ncsw)
264 		goto unlock_and_retry;
265 
266 	/* need to load a new alternate pt space into curpmap? */
267 	COUNT(apdp_pde_map);
268 	opde = *APDP_PDE;
269 	if (!pmap_valid_entry(opde) ||
270 	    pmap_pte2pa(opde) != pmap_pdirpa(pmap, 0)) {
271 		int i;
272 		s = splvm();
273 		/* Make recursive entry usable in user PGD */
274 		for (i = 0; i < PDP_SIZE; i++) {
275 			npde = pmap_pa2pte(
276 			    pmap_pdirpa(pmap, i * NPDPG)) | PG_k | PG_V;
277 			xpq_queue_pte_update(
278 			    xpmap_ptom(pmap_pdirpa(pmap, PDIR_SLOT_PTE + i)),
279 			    npde);
280 			xpq_queue_pte_update(xpmap_ptetomach(&APDP_PDE[i]),
281 			    npde);
282 #ifdef PAE
283 			/* update shadow entry too */
284 			xpq_queue_pte_update(
285 			    xpmap_ptetomach(&APDP_PDE_SHADOW[i]), npde);
286 #endif /* PAE */
287 			xpq_queue_invlpg(
288 			    (vaddr_t)&pmap->pm_pdir[PDIR_SLOT_PTE + i]);
289 		}
290 		if (pmap_valid_entry(opde))
291 			pmap_apte_flush(ourpmap);
292 		splx(s);
293 	}
294 	*pmap2 = ourpmap;
295 	*ptepp = APTE_BASE;
296 	*pdeppp = alternate_pdes;
297 	KASSERT(l->l_ncsw == ncsw);
298 #if !defined(__x86_64__)
299  out:
300 #endif
301  	/*
302  	 * might have blocked, need to retry?
303  	 */
304 	if (l->l_ncsw != ncsw) {
305  unlock_and_retry:
306 	    	if (ourpmap != NULL) {
307 			mutex_exit(ourpmap->pm_lock);
308 			pmap_destroy(ourpmap);
309 		}
310 		mutex_exit(pmap->pm_lock);
311 		goto retry;
312 	}
313 }
314 
315 /*
316  * pmap_unmap_ptes: unlock the PTE mapping of "pmap"
317  */
318 
319 void
320 pmap_unmap_ptes(struct pmap *pmap, struct pmap *pmap2)
321 {
322 
323 	if (pmap == pmap_kernel()) {
324 		return;
325 	}
326 	KASSERT(kpreempt_disabled());
327 	if (pmap2 == NULL) {
328 		mutex_exit(pmap->pm_lock);
329 	} else {
330 #if defined(__x86_64__)
331 		KASSERT(pmap2 == pmap_kernel());
332 #else
333 		KASSERT(curcpu()->ci_pmap == pmap2);
334 #endif
335 #if defined(MULTIPROCESSOR)
336 		pmap_unmap_apdp();
337 		pmap_pte_flush();
338 		pmap_apte_flush(pmap2);
339 #endif /* MULTIPROCESSOR */
340 		COUNT(apdp_pde_unmap);
341 		mutex_exit(pmap->pm_lock);
342 		mutex_exit(pmap2->pm_lock);
343 		pmap_destroy(pmap2);
344 	}
345 }
346 
347 int
348 pmap_enter(struct pmap *pmap, vaddr_t va, paddr_t pa, vm_prot_t prot, u_int flags)
349 {
350         paddr_t ma;
351 
352 	if (__predict_false(pa < pmap_pa_start || pmap_pa_end <= pa)) {
353 		ma = pa; /* XXX hack */
354 	} else {
355 		ma = xpmap_ptom(pa);
356 	}
357 
358 	return pmap_enter_ma(pmap, va, ma, pa, prot, flags, DOMID_SELF);
359 }
360 
361 /*
362  * pmap_kenter_ma: enter a kernel mapping without R/M (pv_entry) tracking
363  *
364  * => no need to lock anything, assume va is already allocated
365  * => should be faster than normal pmap enter function
366  * => we expect a MACHINE address
367  */
368 
369 void
370 pmap_kenter_ma(vaddr_t va, paddr_t ma, vm_prot_t prot, u_int flags)
371 {
372 	pt_entry_t *pte, opte, npte;
373 
374 	if (va < VM_MIN_KERNEL_ADDRESS)
375 		pte = vtopte(va);
376 	else
377 		pte = kvtopte(va);
378 
379 	npte = ma | ((prot & VM_PROT_WRITE) ? PG_RW : PG_RO) |
380 	     PG_V | PG_k;
381 	if (flags & PMAP_NOCACHE)
382 		npte |= PG_N;
383 
384 	if ((cpu_feature[2] & CPUID_NOX) && !(prot & VM_PROT_EXECUTE))
385 		npte |= PG_NX;
386 
387 	opte = pmap_pte_testset (pte, npte); /* zap! */
388 
389 	if (pmap_valid_entry(opte)) {
390 #if defined(MULTIPROCESSOR)
391 		kpreempt_disable();
392 		pmap_tlb_shootdown(pmap_kernel(), va, opte, TLBSHOOT_KENTER);
393 		kpreempt_enable();
394 #else
395 		/* Don't bother deferring in the single CPU case. */
396 		pmap_update_pg(va);
397 #endif
398 	}
399 }
400 
401 /*
402  * pmap_extract_ma: extract a MA for the given VA
403  */
404 
405 bool
406 pmap_extract_ma(struct pmap *pmap, vaddr_t va, paddr_t *pap)
407 {
408 	pt_entry_t *ptes, pte;
409 	pd_entry_t pde;
410 	pd_entry_t * const *pdes;
411 	struct pmap *pmap2;
412 
413 	kpreempt_disable();
414 	pmap_map_ptes(pmap, &pmap2, &ptes, &pdes);
415 	if (!pmap_pdes_valid(va, pdes, &pde)) {
416 		pmap_unmap_ptes(pmap, pmap2);
417 		kpreempt_enable();
418 		return false;
419 	}
420 
421 	pte = ptes[pl1_i(va)];
422 	pmap_unmap_ptes(pmap, pmap2);
423 	kpreempt_enable();
424 
425 	if (__predict_true((pte & PG_V) != 0)) {
426 		if (pap != NULL)
427 			*pap = (pte & PG_FRAME) | (va & (NBPD_L1 - 1));
428 		return true;
429 	}
430 
431 	return false;
432 }
433