1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2006 Peter Wemm
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 AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28 #include <sys/cdefs.h>
29 /*
30 * AMD64 machine dependent routines for kvm and minidumps.
31 */
32
33 #include <sys/param.h>
34 #include <sys/endian.h>
35 #include <stdint.h>
36 #include <stdlib.h>
37 #include <string.h>
38 #include <unistd.h>
39 #include <vm/vm.h>
40 #include <kvm.h>
41
42 #include "../../sys/amd64/include/minidump.h"
43
44 #include <limits.h>
45
46 #include "kvm_private.h"
47 #include "kvm_amd64.h"
48
49 #define amd64_round_page(x) roundup2((kvaddr_t)(x), AMD64_PAGE_SIZE)
50 #define VM_IS_V1(vm) (vm->hdr.version == 1)
51 #define VA_OFF(vm, va) \
52 (VM_IS_V1(vm) ? ((va) & (AMD64_PAGE_SIZE - 1)) : ((va) & AMD64_PAGE_MASK))
53
54 struct vmstate {
55 struct minidumphdr hdr;
56 };
57
58 static vm_prot_t
_amd64_entry_to_prot(uint64_t entry)59 _amd64_entry_to_prot(uint64_t entry)
60 {
61 vm_prot_t prot = VM_PROT_READ;
62
63 if ((entry & AMD64_PG_RW) != 0)
64 prot |= VM_PROT_WRITE;
65 if ((entry & AMD64_PG_NX) == 0)
66 prot |= VM_PROT_EXECUTE;
67 return prot;
68 }
69
70 /*
71 * Version 2 minidumps use page directory entries, while version 1 use page
72 * table entries.
73 */
74
75 static amd64_pde_t
_amd64_pde_get(kvm_t * kd,u_long pdeindex)76 _amd64_pde_get(kvm_t *kd, u_long pdeindex)
77 {
78 amd64_pde_t *pde = _kvm_pmap_get(kd, pdeindex, sizeof(*pde));
79
80 return le64toh(*pde);
81 }
82
83 static amd64_pte_t
_amd64_pte_get(kvm_t * kd,u_long pteindex)84 _amd64_pte_get(kvm_t *kd, u_long pteindex)
85 {
86 amd64_pte_t *pte = _kvm_pmap_get(kd, pteindex, sizeof(*pte));
87
88 return le64toh(*pte);
89 }
90
91 /* Get the first page table entry for a given page directory index. */
92 static amd64_pte_t *
_amd64_pde_first_pte(kvm_t * kd,u_long pdeindex)93 _amd64_pde_first_pte(kvm_t *kd, u_long pdeindex)
94 {
95 u_long *pa;
96
97 pa = _kvm_pmap_get(kd, pdeindex, sizeof(amd64_pde_t));
98 if (pa == NULL)
99 return NULL;
100 return _kvm_map_get(kd, *pa & AMD64_PG_FRAME, AMD64_PAGE_SIZE);
101 }
102
103 static int
_amd64_minidump_probe(kvm_t * kd)104 _amd64_minidump_probe(kvm_t *kd)
105 {
106
107 return (_kvm_probe_elf_kernel(kd, ELFCLASS64, EM_X86_64) &&
108 _kvm_is_minidump(kd));
109 }
110
111 static void
_amd64_minidump_freevtop(kvm_t * kd)112 _amd64_minidump_freevtop(kvm_t *kd)
113 {
114 struct vmstate *vm = kd->vmst;
115
116 free(vm);
117 kd->vmst = NULL;
118 }
119
120 static int
_amd64_minidump_initvtop(kvm_t * kd)121 _amd64_minidump_initvtop(kvm_t *kd)
122 {
123 struct vmstate *vmst;
124 off_t off, dump_avail_off, sparse_off;
125
126 vmst = _kvm_malloc(kd, sizeof(*vmst));
127 if (vmst == NULL) {
128 _kvm_err(kd, kd->program, "cannot allocate vm");
129 return (-1);
130 }
131 kd->vmst = vmst;
132 if (pread(kd->pmfd, &vmst->hdr, sizeof(vmst->hdr), 0) !=
133 sizeof(vmst->hdr)) {
134 _kvm_err(kd, kd->program, "cannot read dump header");
135 return (-1);
136 }
137 if (strncmp(MINIDUMP_MAGIC, vmst->hdr.magic, sizeof(vmst->hdr.magic)) != 0) {
138 _kvm_err(kd, kd->program, "not a minidump for this platform");
139 return (-1);
140 }
141
142 /*
143 * NB: amd64 minidump header is binary compatible between version 1
144 * and version 2; version 3 adds the dumpavailsize field
145 */
146 vmst->hdr.version = le32toh(vmst->hdr.version);
147 if (vmst->hdr.version > MINIDUMP_VERSION || vmst->hdr.version < 1) {
148 _kvm_err(kd, kd->program, "wrong minidump version. expected %d got %d",
149 MINIDUMP_VERSION, vmst->hdr.version);
150 return (-1);
151 }
152 vmst->hdr.msgbufsize = le32toh(vmst->hdr.msgbufsize);
153 vmst->hdr.bitmapsize = le32toh(vmst->hdr.bitmapsize);
154 vmst->hdr.pmapsize = le32toh(vmst->hdr.pmapsize);
155 vmst->hdr.kernbase = le64toh(vmst->hdr.kernbase);
156 vmst->hdr.dmapbase = le64toh(vmst->hdr.dmapbase);
157 vmst->hdr.dmapend = le64toh(vmst->hdr.dmapend);
158 vmst->hdr.dumpavailsize = vmst->hdr.version == MINIDUMP_VERSION ?
159 le32toh(vmst->hdr.dumpavailsize) : 0;
160
161 /* Skip header and msgbuf */
162 dump_avail_off = AMD64_PAGE_SIZE + amd64_round_page(vmst->hdr.msgbufsize);
163
164 /* Skip dump_avail */
165 off = dump_avail_off + amd64_round_page(vmst->hdr.dumpavailsize);
166
167 sparse_off = off + amd64_round_page(vmst->hdr.bitmapsize) +
168 amd64_round_page(vmst->hdr.pmapsize);
169 if (_kvm_pt_init(kd, vmst->hdr.dumpavailsize, dump_avail_off,
170 vmst->hdr.bitmapsize, off, sparse_off, AMD64_PAGE_SIZE) == -1) {
171 return (-1);
172 }
173 off += amd64_round_page(vmst->hdr.bitmapsize);
174
175 if (_kvm_pmap_init(kd, vmst->hdr.pmapsize, off) == -1) {
176 return (-1);
177 }
178 off += amd64_round_page(vmst->hdr.pmapsize);
179
180 return (0);
181 }
182
183 static int
_amd64_minidump_vatop_v1(kvm_t * kd,kvaddr_t va,off_t * pa)184 _amd64_minidump_vatop_v1(kvm_t *kd, kvaddr_t va, off_t *pa)
185 {
186 struct vmstate *vm;
187 amd64_physaddr_t offset;
188 amd64_pte_t pte;
189 kvaddr_t pteindex;
190 amd64_physaddr_t a;
191 off_t ofs;
192
193 vm = kd->vmst;
194 offset = va & AMD64_PAGE_MASK;
195
196 if (va >= vm->hdr.kernbase) {
197 pteindex = (va - vm->hdr.kernbase) >> AMD64_PAGE_SHIFT;
198 if (pteindex >= vm->hdr.pmapsize / sizeof(pte))
199 goto invalid;
200 pte = _amd64_pte_get(kd, pteindex);
201 if ((pte & AMD64_PG_V) == 0) {
202 _kvm_err(kd, kd->program,
203 "_amd64_minidump_vatop_v1: pte not valid");
204 goto invalid;
205 }
206 a = pte & AMD64_PG_FRAME;
207 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
208 if (ofs == -1) {
209 _kvm_err(kd, kd->program,
210 "_amd64_minidump_vatop_v1: physical address 0x%jx not in minidump",
211 (uintmax_t)a);
212 goto invalid;
213 }
214 *pa = ofs + offset;
215 return (AMD64_PAGE_SIZE - offset);
216 } else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
217 a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
218 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
219 if (ofs == -1) {
220 _kvm_err(kd, kd->program,
221 "_amd64_minidump_vatop_v1: direct map address 0x%jx not in minidump",
222 (uintmax_t)va);
223 goto invalid;
224 }
225 *pa = ofs + offset;
226 return (AMD64_PAGE_SIZE - offset);
227 } else {
228 _kvm_err(kd, kd->program,
229 "_amd64_minidump_vatop_v1: virtual address 0x%jx not minidumped",
230 (uintmax_t)va);
231 goto invalid;
232 }
233
234 invalid:
235 _kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
236 return (0);
237 }
238
239 static int
_amd64_minidump_vatop(kvm_t * kd,kvaddr_t va,off_t * pa)240 _amd64_minidump_vatop(kvm_t *kd, kvaddr_t va, off_t *pa)
241 {
242 amd64_pte_t pt[AMD64_NPTEPG];
243 struct vmstate *vm;
244 amd64_physaddr_t offset;
245 amd64_pde_t pde;
246 amd64_pte_t pte;
247 kvaddr_t pteindex;
248 kvaddr_t pdeindex;
249 amd64_physaddr_t a;
250 off_t ofs;
251
252 vm = kd->vmst;
253 offset = va & AMD64_PAGE_MASK;
254
255 if (va >= vm->hdr.kernbase) {
256 pdeindex = (va - vm->hdr.kernbase) >> AMD64_PDRSHIFT;
257 if (pdeindex >= vm->hdr.pmapsize / sizeof(pde))
258 goto invalid;
259 pde = _amd64_pde_get(kd, pdeindex);
260 if ((pde & AMD64_PG_V) == 0) {
261 _kvm_err(kd, kd->program,
262 "_amd64_minidump_vatop: pde not valid");
263 goto invalid;
264 }
265 if ((pde & AMD64_PG_PS) == 0) {
266 a = pde & AMD64_PG_FRAME;
267 /* TODO: Just read the single PTE */
268 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
269 if (ofs == -1) {
270 _kvm_err(kd, kd->program,
271 "cannot find page table entry for %ju",
272 (uintmax_t)a);
273 goto invalid;
274 }
275 if (pread(kd->pmfd, &pt, AMD64_PAGE_SIZE, ofs) !=
276 AMD64_PAGE_SIZE) {
277 _kvm_err(kd, kd->program,
278 "cannot read page table entry for %ju",
279 (uintmax_t)a);
280 goto invalid;
281 }
282 pteindex = (va >> AMD64_PAGE_SHIFT) &
283 (AMD64_NPTEPG - 1);
284 pte = le64toh(pt[pteindex]);
285 if ((pte & AMD64_PG_V) == 0) {
286 _kvm_err(kd, kd->program,
287 "_amd64_minidump_vatop: pte not valid");
288 goto invalid;
289 }
290 a = pte & AMD64_PG_FRAME;
291 } else {
292 a = pde & AMD64_PG_PS_FRAME;
293 a += (va & AMD64_PDRMASK) ^ offset;
294 }
295 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
296 if (ofs == -1) {
297 _kvm_err(kd, kd->program,
298 "_amd64_minidump_vatop: physical address 0x%jx not in minidump",
299 (uintmax_t)a);
300 goto invalid;
301 }
302 *pa = ofs + offset;
303 return (AMD64_PAGE_SIZE - offset);
304 } else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
305 a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
306 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
307 if (ofs == -1) {
308 _kvm_err(kd, kd->program,
309 "_amd64_minidump_vatop: direct map address 0x%jx not in minidump",
310 (uintmax_t)va);
311 goto invalid;
312 }
313 *pa = ofs + offset;
314 return (AMD64_PAGE_SIZE - offset);
315 } else {
316 _kvm_err(kd, kd->program,
317 "_amd64_minidump_vatop: virtual address 0x%jx not minidumped",
318 (uintmax_t)va);
319 goto invalid;
320 }
321
322 invalid:
323 _kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
324 return (0);
325 }
326
327 static int
_amd64_minidump_kvatop(kvm_t * kd,kvaddr_t va,off_t * pa)328 _amd64_minidump_kvatop(kvm_t *kd, kvaddr_t va, off_t *pa)
329 {
330
331 if (ISALIVE(kd)) {
332 _kvm_err(kd, 0,
333 "_amd64_minidump_kvatop called in live kernel!");
334 return (0);
335 }
336 if (((struct vmstate *)kd->vmst)->hdr.version == 1)
337 return (_amd64_minidump_vatop_v1(kd, va, pa));
338 else
339 return (_amd64_minidump_vatop(kd, va, pa));
340 }
341
342 static int
_amd64_minidump_walk_pages(kvm_t * kd,kvm_walk_pages_cb_t * cb,void * arg)343 _amd64_minidump_walk_pages(kvm_t *kd, kvm_walk_pages_cb_t *cb, void *arg)
344 {
345 struct vmstate *vm = kd->vmst;
346 u_long npdes = vm->hdr.pmapsize / sizeof(amd64_pde_t);
347 u_long bmindex, dva, pa, pdeindex, va;
348 struct kvm_bitmap bm;
349 int ret = 0;
350 vm_prot_t prot;
351 unsigned int pgsz = AMD64_PAGE_SIZE;
352
353 if (vm->hdr.version < 2)
354 return (0);
355
356 if (!_kvm_bitmap_init(&bm, vm->hdr.bitmapsize, &bmindex))
357 return (0);
358
359 for (pdeindex = 0; pdeindex < npdes; pdeindex++) {
360 amd64_pde_t pde = _amd64_pde_get(kd, pdeindex);
361 amd64_pte_t *ptes;
362 u_long i;
363
364 va = vm->hdr.kernbase + (pdeindex << AMD64_PDRSHIFT);
365 if ((pde & AMD64_PG_V) == 0)
366 continue;
367
368 if ((pde & AMD64_PG_PS) != 0) {
369 /*
370 * Large page. Iterate on each 4K page section
371 * within this page. This differs from 4K pages in
372 * that every page here uses the same PDE to
373 * generate permissions.
374 */
375 pa = (pde & AMD64_PG_PS_FRAME) +
376 ((va & AMD64_PDRMASK) ^ VA_OFF(vm, va));
377 dva = vm->hdr.dmapbase + pa;
378 _kvm_bitmap_set(&bm, _kvm_pa_bit_id(kd, pa, AMD64_PAGE_SIZE));
379 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva,
380 _amd64_entry_to_prot(pde), AMD64_NBPDR, pgsz)) {
381 goto out;
382 }
383 continue;
384 }
385
386 /* 4K pages: pde references another page of entries. */
387 ptes = _amd64_pde_first_pte(kd, pdeindex);
388 /* Ignore page directory pages that were not dumped. */
389 if (ptes == NULL)
390 continue;
391
392 for (i = 0; i < AMD64_NPTEPG; i++) {
393 amd64_pte_t pte = (u_long)ptes[i];
394
395 pa = pte & AMD64_PG_FRAME;
396 dva = vm->hdr.dmapbase + pa;
397 if ((pte & AMD64_PG_V) != 0) {
398 _kvm_bitmap_set(&bm,
399 _kvm_pa_bit_id(kd, pa, AMD64_PAGE_SIZE));
400 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva,
401 _amd64_entry_to_prot(pte), pgsz, 0)) {
402 goto out;
403 }
404 }
405 va += AMD64_PAGE_SIZE;
406 }
407 }
408
409 while (_kvm_bitmap_next(&bm, &bmindex)) {
410 pa = _kvm_bit_id_pa(kd, bmindex, AMD64_PAGE_SIZE);
411 if (pa == _KVM_PA_INVALID)
412 break;
413 dva = vm->hdr.dmapbase + pa;
414 if (vm->hdr.dmapend < (dva + pgsz))
415 break;
416 va = 0;
417 /* amd64/pmap.c: create_pagetables(): dmap always R|W. */
418 prot = VM_PROT_READ | VM_PROT_WRITE;
419 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva, prot, pgsz, 0)) {
420 goto out;
421 }
422 }
423
424 ret = 1;
425
426 out:
427 _kvm_bitmap_deinit(&bm);
428 return (ret);
429 }
430
431 static struct kvm_arch kvm_amd64_minidump = {
432 .ka_probe = _amd64_minidump_probe,
433 .ka_initvtop = _amd64_minidump_initvtop,
434 .ka_freevtop = _amd64_minidump_freevtop,
435 .ka_kvatop = _amd64_minidump_kvatop,
436 .ka_native = _amd64_native,
437 .ka_walk_pages = _amd64_minidump_walk_pages,
438 };
439
440 KVM_ARCH(kvm_amd64_minidump);
441