xref: /netbsd-src/sys/arch/xen/x86/xen_pmap.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: xen_pmap.c,v 1.26 2017/03/23 18:08:06 maxv 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  * Copyright (c) 2006 Mathieu Ropert <mro@adviseo.fr>
29  *
30  * Permission to use, copy, modify, and distribute this software for any
31  * purpose with or without fee is hereby granted, provided that the above
32  * copyright notice and this permission notice appear in all copies.
33  *
34  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
35  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
36  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
37  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
38  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
39  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
40  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
41  */
42 
43 /*
44  * Copyright (c) 1997 Charles D. Cranor and Washington University.
45  * All rights reserved.
46  *
47  * Redistribution and use in source and binary forms, with or without
48  * modification, are permitted provided that the following conditions
49  * are met:
50  * 1. Redistributions of source code must retain the above copyright
51  *    notice, this list of conditions and the following disclaimer.
52  * 2. Redistributions in binary form must reproduce the above copyright
53  *    notice, this list of conditions and the following disclaimer in the
54  *    documentation and/or other materials provided with the distribution.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
57  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
58  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
59  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
60  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
61  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
62  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
63  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
64  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
65  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
66  */
67 
68 /*
69  * Copyright 2001 (c) Wasabi Systems, Inc.
70  * All rights reserved.
71  *
72  * Written by Frank van der Linden for Wasabi Systems, Inc.
73  *
74  * Redistribution and use in source and binary forms, with or without
75  * modification, are permitted provided that the following conditions
76  * are met:
77  * 1. Redistributions of source code must retain the above copyright
78  *    notice, this list of conditions and the following disclaimer.
79  * 2. Redistributions in binary form must reproduce the above copyright
80  *    notice, this list of conditions and the following disclaimer in the
81  *    documentation and/or other materials provided with the distribution.
82  * 3. All advertising materials mentioning features or use of this software
83  *    must display the following acknowledgement:
84  *      This product includes software developed for the NetBSD Project by
85  *      Wasabi Systems, Inc.
86  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
87  *    or promote products derived from this software without specific prior
88  *    written permission.
89  *
90  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
91  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
92  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
93  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
94  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
95  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
96  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
97  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
98  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
99  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
100  * POSSIBILITY OF SUCH DAMAGE.
101  */
102 
103 #include <sys/cdefs.h>
104 __KERNEL_RCSID(0, "$NetBSD: xen_pmap.c,v 1.26 2017/03/23 18:08:06 maxv Exp $");
105 
106 #include "opt_user_ldt.h"
107 #include "opt_lockdebug.h"
108 #include "opt_multiprocessor.h"
109 #include "opt_xen.h"
110 
111 #include <sys/param.h>
112 #include <sys/systm.h>
113 #include <sys/proc.h>
114 #include <sys/pool.h>
115 #include <sys/kernel.h>
116 #include <sys/atomic.h>
117 #include <sys/cpu.h>
118 #include <sys/intr.h>
119 
120 #include <uvm/uvm.h>
121 
122 #include <dev/isa/isareg.h>
123 
124 #include <machine/specialreg.h>
125 #include <machine/gdt.h>
126 #include <machine/isa_machdep.h>
127 #include <machine/cpuvar.h>
128 
129 #include <x86/pmap.h>
130 #include <x86/pmap_pv.h>
131 
132 #include <x86/i82489reg.h>
133 #include <x86/i82489var.h>
134 
135 #include <xen/xen-public/xen.h>
136 #include <xen/hypervisor.h>
137 #include <xen/xenpmap.h>
138 
139 #define COUNT(x)	/* nothing */
140 
141 extern pd_entry_t * const normal_pdes[];
142 
143 extern paddr_t pmap_pa_start; /* PA of first physical page for this domain */
144 extern paddr_t pmap_pa_end;   /* PA of last physical page for this domain */
145 
146 int
147 pmap_enter(struct pmap *pmap, vaddr_t va, paddr_t pa, vm_prot_t prot, u_int flags)
148 {
149 	paddr_t ma;
150 
151 	if (__predict_false(pa < pmap_pa_start || pmap_pa_end <= pa)) {
152 		ma = pa; /* XXX hack */
153 	} else {
154 		ma = xpmap_ptom(pa);
155 	}
156 
157 	return pmap_enter_ma(pmap, va, ma, pa, prot, flags, DOMID_SELF);
158 }
159 
160 /*
161  * pmap_kenter_ma: enter a kernel mapping without R/M (pv_entry) tracking
162  *
163  * => no need to lock anything, assume va is already allocated
164  * => should be faster than normal pmap enter function
165  * => we expect a MACHINE address
166  */
167 
168 void
169 pmap_kenter_ma(vaddr_t va, paddr_t ma, vm_prot_t prot, u_int flags)
170 {
171 	pt_entry_t *pte, opte, npte;
172 
173 	if (va < VM_MIN_KERNEL_ADDRESS)
174 		pte = vtopte(va);
175 	else
176 		pte = kvtopte(va);
177 
178 	npte = ma | ((prot & VM_PROT_WRITE) ? PG_RW : PG_RO) | PG_V;
179 	if (flags & PMAP_NOCACHE)
180 		npte |= PG_N;
181 
182 	if ((cpu_feature[2] & CPUID_NOX) && !(prot & VM_PROT_EXECUTE))
183 		npte |= PG_NX;
184 
185 	opte = pmap_pte_testset(pte, npte); /* zap! */
186 
187 	if (pmap_valid_entry(opte)) {
188 #if defined(MULTIPROCESSOR)
189 		if (__predict_false(x86_mp_online == false)) {
190 			pmap_update_pg(va);
191 		} else {
192 			kpreempt_disable();
193 			pmap_tlb_shootdown(pmap_kernel(), va, opte,
194 			    TLBSHOOT_KENTER);
195 			kpreempt_enable();
196 		}
197 #else
198 		/* Don't bother deferring in the single CPU case. */
199 		pmap_update_pg(va);
200 #endif
201 	}
202 }
203 
204 /*
205  * pmap_extract_ma: extract a MA for the given VA
206  */
207 
208 bool
209 pmap_extract_ma(struct pmap *pmap, vaddr_t va, paddr_t *pap)
210 {
211 	pt_entry_t *ptes, pte;
212 	pd_entry_t pde;
213 	pd_entry_t * const *pdes;
214 	struct pmap *pmap2;
215 
216 	kpreempt_disable();
217 	pmap_map_ptes(pmap, &pmap2, &ptes, &pdes);
218 	if (!pmap_pdes_valid(va, pdes, &pde)) {
219 		pmap_unmap_ptes(pmap, pmap2);
220 		kpreempt_enable();
221 		return false;
222 	}
223 
224 	pte = ptes[pl1_i(va)];
225 	pmap_unmap_ptes(pmap, pmap2);
226 	kpreempt_enable();
227 
228 	if (__predict_true((pte & PG_V) != 0)) {
229 		if (pap != NULL)
230 			*pap = (pte & PG_FRAME) | (va & (NBPD_L1 - 1));
231 		return true;
232 	}
233 
234 	return false;
235 }
236 
237 /*
238  * Xen pmap's handlers for save/restore
239  */
240 void
241 pmap_xen_suspend(void)
242 {
243 	pmap_unmap_recursive_entries();
244 
245 	xpq_flush_queue();
246 }
247 
248 void
249 pmap_xen_resume(void)
250 {
251 	pmap_map_recursive_entries();
252 
253 	xpq_flush_queue();
254 }
255 
256 /*
257  * NetBSD uses L2 shadow pages to support PAE with Xen. However, Xen does not
258  * handle them correctly during save/restore, leading to incorrect page
259  * tracking and pinning during restore.
260  * For save/restore to succeed, two functions are introduced:
261  * - pmap_map_recursive_entries(), used by resume code to set the recursive
262  *   mapping entries to their correct value
263  * - pmap_unmap_recursive_entries(), used by suspend code to clear all
264  *   PDIR_SLOT_PTE entries
265  */
266 void
267 pmap_map_recursive_entries(void)
268 {
269 	int i;
270 	struct pmap *pm;
271 
272 	mutex_enter(&pmaps_lock);
273 	LIST_FOREACH(pm, &pmaps, pm_list) {
274 		for (i = 0; i < PDP_SIZE; i++) {
275 			xpq_queue_pte_update(
276 			    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_PTE + i)),
277 			    xpmap_ptom((pm)->pm_pdirpa[i]) | PG_V);
278 		}
279 	}
280 	mutex_exit(&pmaps_lock);
281 
282 	for (i = 0; i < PDP_SIZE; i++) {
283 		xpq_queue_pte_update(
284 		    xpmap_ptom(pmap_pdirpa(pmap_kernel(), PDIR_SLOT_PTE + i)),
285 		    xpmap_ptom(pmap_kernel()->pm_pdirpa[i]) | PG_V);
286 	}
287 }
288 
289 /*
290  * Unmap recursive entries found in pmaps. Required during Xen
291  * save/restore operations, as Xen does not handle recursive mappings
292  * properly.
293  */
294 void
295 pmap_unmap_recursive_entries(void)
296 {
297 	int i;
298 	struct pmap *pm;
299 
300 	/*
301 	 * Invalidate pmap_pdp_cache as it contains L2-pinned objects with
302 	 * recursive entries.
303 	 * XXX jym@ : find a way to drain per-CPU caches to. pool_cache_inv
304 	 * does not do that.
305 	 */
306 	pool_cache_invalidate(&pmap_pdp_cache);
307 
308 	mutex_enter(&pmaps_lock);
309 	LIST_FOREACH(pm, &pmaps, pm_list) {
310 		for (i = 0; i < PDP_SIZE; i++) {
311 			xpq_queue_pte_update(
312 			    xpmap_ptom(pmap_pdirpa(pm, PDIR_SLOT_PTE + i)), 0);
313 		}
314 	}
315 	mutex_exit(&pmaps_lock);
316 
317 	/* do it for pmap_kernel() too! */
318 	for (i = 0; i < PDP_SIZE; i++) {
319 		xpq_queue_pte_update(
320 		    xpmap_ptom(pmap_pdirpa(pmap_kernel(), PDIR_SLOT_PTE + i)),
321 		    0);
322 	}
323 }
324 
325 #if defined(PAE) || defined(__x86_64__)
326 
327 static __inline void
328 pmap_kpm_setpte(struct cpu_info *ci, struct pmap *pmap, int index)
329 {
330 	KASSERT(mutex_owned(pmap->pm_lock));
331 	KASSERT(mutex_owned(&ci->ci_kpm_mtx));
332 	if (pmap == pmap_kernel()) {
333 		KASSERT(index >= PDIR_SLOT_KERN);
334 	}
335 #ifdef PAE
336 	xpq_queue_pte_update(
337 	    xpmap_ptetomach(&ci->ci_kpm_pdir[l2tol2(index)]),
338 	    pmap->pm_pdir[index]);
339 #elif defined(__x86_64__)
340 	xpq_queue_pte_update(
341 	    xpmap_ptetomach(&ci->ci_kpm_pdir[index]),
342 	    pmap->pm_pdir[index]);
343 #endif
344 	xpq_flush_queue();
345 }
346 
347 /*
348  * Synchronise shadow pdir with the pmap on all cpus on which it is
349  * loaded.
350  */
351 void
352 xen_kpm_sync(struct pmap *pmap, int index)
353 {
354 	CPU_INFO_ITERATOR cii;
355 	struct cpu_info *ci;
356 
357 	KASSERT(pmap != NULL);
358 	KASSERT(kpreempt_disabled());
359 
360 	pmap_pte_flush();
361 
362 	for (CPU_INFO_FOREACH(cii, ci)) {
363 		if (ci == NULL) {
364 			continue;
365 		}
366 		cpuid_t cid = cpu_index(ci);
367 		if (pmap != pmap_kernel() &&
368 		    !kcpuset_isset(pmap->pm_xen_ptp_cpus, cid))
369 			continue;
370 
371 		/* take the lock and check again */
372 		mutex_enter(&ci->ci_kpm_mtx);
373 		if (pmap == pmap_kernel() ||
374 		    kcpuset_isset(pmap->pm_xen_ptp_cpus, cid)) {
375 			pmap_kpm_setpte(ci, pmap, index);
376 		}
377 		mutex_exit(&ci->ci_kpm_mtx);
378 	}
379 }
380 
381 #endif /* PAE || __x86_64__ */
382