2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2006 Peter Wemm
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
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.
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
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
32 * AMD64 machine dependent routines for kvm and minidumps.
35 #include <sys/param.h>
36 #include <sys/endian.h>
44 #include "../../sys/amd64/include/minidump.h"
48 #include "kvm_private.h"
49 #include "kvm_amd64.h"
51 #define amd64_round_page(x) roundup2((kvaddr_t)(x), AMD64_PAGE_SIZE)
52 #define VM_IS_V1(vm) (vm->hdr.version == 1)
53 #define VA_OFF(vm, va) \
54 (VM_IS_V1(vm) ? ((va) & (AMD64_PAGE_SIZE - 1)) : ((va) & AMD64_PAGE_MASK))
57 struct minidumphdr hdr;
61 _amd64_entry_to_prot(uint64_t entry)
63 vm_prot_t prot = VM_PROT_READ;
65 if ((entry & AMD64_PG_RW) != 0)
66 prot |= VM_PROT_WRITE;
67 if ((entry & AMD64_PG_NX) == 0)
68 prot |= VM_PROT_EXECUTE;
73 * Version 2 minidumps use page directory entries, while version 1 use page
78 _amd64_pde_get(kvm_t *kd, u_long pdeindex)
80 amd64_pde_t *pde = _kvm_pmap_get(kd, pdeindex, sizeof(*pde));
86 _amd64_pte_get(kvm_t *kd, u_long pteindex)
88 amd64_pte_t *pte = _kvm_pmap_get(kd, pteindex, sizeof(*pte));
93 /* Get the first page table entry for a given page directory index. */
95 _amd64_pde_first_pte(kvm_t *kd, u_long pdeindex)
99 pa = _kvm_pmap_get(kd, pdeindex, sizeof(amd64_pde_t));
102 return _kvm_map_get(kd, *pa & AMD64_PG_FRAME, AMD64_PAGE_SIZE);
106 _amd64_minidump_probe(kvm_t *kd)
109 return (_kvm_probe_elf_kernel(kd, ELFCLASS64, EM_X86_64) &&
110 _kvm_is_minidump(kd));
114 _amd64_minidump_freevtop(kvm_t *kd)
116 struct vmstate *vm = kd->vmst;
123 _amd64_minidump_initvtop(kvm_t *kd)
125 struct vmstate *vmst;
126 off_t off, dump_avail_off, sparse_off;
128 vmst = _kvm_malloc(kd, sizeof(*vmst));
130 _kvm_err(kd, kd->program, "cannot allocate vm");
134 if (pread(kd->pmfd, &vmst->hdr, sizeof(vmst->hdr), 0) !=
136 _kvm_err(kd, kd->program, "cannot read dump header");
139 if (strncmp(MINIDUMP_MAGIC, vmst->hdr.magic, sizeof(vmst->hdr.magic)) != 0) {
140 _kvm_err(kd, kd->program, "not a minidump for this platform");
145 * NB: amd64 minidump header is binary compatible between version 1
146 * and version 2; version 3 adds the dumpavailsize field
148 vmst->hdr.version = le32toh(vmst->hdr.version);
149 if (vmst->hdr.version > MINIDUMP_VERSION || vmst->hdr.version < 1) {
150 _kvm_err(kd, kd->program, "wrong minidump version. expected %d got %d",
151 MINIDUMP_VERSION, vmst->hdr.version);
154 vmst->hdr.msgbufsize = le32toh(vmst->hdr.msgbufsize);
155 vmst->hdr.bitmapsize = le32toh(vmst->hdr.bitmapsize);
156 vmst->hdr.pmapsize = le32toh(vmst->hdr.pmapsize);
157 vmst->hdr.kernbase = le64toh(vmst->hdr.kernbase);
158 vmst->hdr.dmapbase = le64toh(vmst->hdr.dmapbase);
159 vmst->hdr.dmapend = le64toh(vmst->hdr.dmapend);
160 vmst->hdr.dumpavailsize = vmst->hdr.version == MINIDUMP_VERSION ?
161 le32toh(vmst->hdr.dumpavailsize) : 0;
163 /* Skip header and msgbuf */
164 dump_avail_off = AMD64_PAGE_SIZE + amd64_round_page(vmst->hdr.msgbufsize);
166 /* Skip dump_avail */
167 off = dump_avail_off + amd64_round_page(vmst->hdr.dumpavailsize);
169 sparse_off = off + amd64_round_page(vmst->hdr.bitmapsize) +
170 amd64_round_page(vmst->hdr.pmapsize);
171 if (_kvm_pt_init(kd, vmst->hdr.dumpavailsize, dump_avail_off,
172 vmst->hdr.bitmapsize, off, sparse_off, AMD64_PAGE_SIZE,
173 sizeof(uint64_t)) == -1) {
176 off += amd64_round_page(vmst->hdr.bitmapsize);
178 if (_kvm_pmap_init(kd, vmst->hdr.pmapsize, off) == -1) {
181 off += amd64_round_page(vmst->hdr.pmapsize);
187 _amd64_minidump_vatop_v1(kvm_t *kd, kvaddr_t va, off_t *pa)
190 amd64_physaddr_t offset;
197 offset = va & AMD64_PAGE_MASK;
199 if (va >= vm->hdr.kernbase) {
200 pteindex = (va - vm->hdr.kernbase) >> AMD64_PAGE_SHIFT;
201 if (pteindex >= vm->hdr.pmapsize / sizeof(pte))
203 pte = _amd64_pte_get(kd, pteindex);
204 if ((pte & AMD64_PG_V) == 0) {
205 _kvm_err(kd, kd->program,
206 "_amd64_minidump_vatop_v1: pte not valid");
209 a = pte & AMD64_PG_FRAME;
210 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
212 _kvm_err(kd, kd->program,
213 "_amd64_minidump_vatop_v1: physical address 0x%jx not in minidump",
218 return (AMD64_PAGE_SIZE - offset);
219 } else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
220 a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
221 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
223 _kvm_err(kd, kd->program,
224 "_amd64_minidump_vatop_v1: direct map address 0x%jx not in minidump",
229 return (AMD64_PAGE_SIZE - offset);
231 _kvm_err(kd, kd->program,
232 "_amd64_minidump_vatop_v1: virtual address 0x%jx not minidumped",
238 _kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
243 _amd64_minidump_vatop(kvm_t *kd, kvaddr_t va, off_t *pa)
245 amd64_pte_t pt[AMD64_NPTEPG];
247 amd64_physaddr_t offset;
256 offset = va & AMD64_PAGE_MASK;
258 if (va >= vm->hdr.kernbase) {
259 pdeindex = (va - vm->hdr.kernbase) >> AMD64_PDRSHIFT;
260 if (pdeindex >= vm->hdr.pmapsize / sizeof(pde))
262 pde = _amd64_pde_get(kd, pdeindex);
263 if ((pde & AMD64_PG_V) == 0) {
264 _kvm_err(kd, kd->program,
265 "_amd64_minidump_vatop: pde not valid");
268 if ((pde & AMD64_PG_PS) == 0) {
269 a = pde & AMD64_PG_FRAME;
270 /* TODO: Just read the single PTE */
271 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
273 _kvm_err(kd, kd->program,
274 "cannot find page table entry for %ju",
278 if (pread(kd->pmfd, &pt, AMD64_PAGE_SIZE, ofs) !=
280 _kvm_err(kd, kd->program,
281 "cannot read page table entry for %ju",
285 pteindex = (va >> AMD64_PAGE_SHIFT) &
287 pte = le64toh(pt[pteindex]);
288 if ((pte & AMD64_PG_V) == 0) {
289 _kvm_err(kd, kd->program,
290 "_amd64_minidump_vatop: pte not valid");
293 a = pte & AMD64_PG_FRAME;
295 a = pde & AMD64_PG_PS_FRAME;
296 a += (va & AMD64_PDRMASK) ^ offset;
298 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
300 _kvm_err(kd, kd->program,
301 "_amd64_minidump_vatop: physical address 0x%jx not in minidump",
306 return (AMD64_PAGE_SIZE - offset);
307 } else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
308 a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
309 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
311 _kvm_err(kd, kd->program,
312 "_amd64_minidump_vatop: direct map address 0x%jx not in minidump",
317 return (AMD64_PAGE_SIZE - offset);
319 _kvm_err(kd, kd->program,
320 "_amd64_minidump_vatop: virtual address 0x%jx not minidumped",
326 _kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
331 _amd64_minidump_kvatop(kvm_t *kd, kvaddr_t va, off_t *pa)
336 "_amd64_minidump_kvatop called in live kernel!");
339 if (((struct vmstate *)kd->vmst)->hdr.version == 1)
340 return (_amd64_minidump_vatop_v1(kd, va, pa));
342 return (_amd64_minidump_vatop(kd, va, pa));
346 _amd64_minidump_walk_pages(kvm_t *kd, kvm_walk_pages_cb_t *cb, void *arg)
348 struct vmstate *vm = kd->vmst;
349 u_long npdes = vm->hdr.pmapsize / sizeof(amd64_pde_t);
350 u_long bmindex, dva, pa, pdeindex, va;
351 struct kvm_bitmap bm;
354 unsigned int pgsz = AMD64_PAGE_SIZE;
356 if (vm->hdr.version < 2)
359 if (!_kvm_bitmap_init(&bm, vm->hdr.bitmapsize, &bmindex))
362 for (pdeindex = 0; pdeindex < npdes; pdeindex++) {
363 amd64_pde_t pde = _amd64_pde_get(kd, pdeindex);
367 va = vm->hdr.kernbase + (pdeindex << AMD64_PDRSHIFT);
368 if ((pde & AMD64_PG_V) == 0)
371 if ((pde & AMD64_PG_PS) != 0) {
373 * Large page. Iterate on each 4K page section
374 * within this page. This differs from 4K pages in
375 * that every page here uses the same PDE to
376 * generate permissions.
378 pa = (pde & AMD64_PG_PS_FRAME) +
379 ((va & AMD64_PDRMASK) ^ VA_OFF(vm, va));
380 dva = vm->hdr.dmapbase + pa;
381 _kvm_bitmap_set(&bm, _kvm_pa_bit_id(kd, pa, AMD64_PAGE_SIZE));
382 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva,
383 _amd64_entry_to_prot(pde), AMD64_NBPDR, pgsz)) {
389 /* 4K pages: pde references another page of entries. */
390 ptes = _amd64_pde_first_pte(kd, pdeindex);
391 /* Ignore page directory pages that were not dumped. */
395 for (i = 0; i < AMD64_NPTEPG; i++) {
396 amd64_pte_t pte = (u_long)ptes[i];
398 pa = pte & AMD64_PG_FRAME;
399 dva = vm->hdr.dmapbase + pa;
400 if ((pte & AMD64_PG_V) != 0) {
402 _kvm_pa_bit_id(kd, pa, AMD64_PAGE_SIZE));
403 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva,
404 _amd64_entry_to_prot(pte), pgsz, 0)) {
408 va += AMD64_PAGE_SIZE;
412 while (_kvm_bitmap_next(&bm, &bmindex)) {
413 pa = _kvm_bit_id_pa(kd, bmindex, AMD64_PAGE_SIZE);
414 if (pa == _KVM_PA_INVALID)
416 dva = vm->hdr.dmapbase + pa;
417 if (vm->hdr.dmapend < (dva + pgsz))
420 /* amd64/pmap.c: create_pagetables(): dmap always R|W. */
421 prot = VM_PROT_READ | VM_PROT_WRITE;
422 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva, prot, pgsz, 0)) {
430 _kvm_bitmap_deinit(&bm);
434 static struct kvm_arch kvm_amd64_minidump = {
435 .ka_probe = _amd64_minidump_probe,
436 .ka_initvtop = _amd64_minidump_initvtop,
437 .ka_freevtop = _amd64_minidump_freevtop,
438 .ka_kvatop = _amd64_minidump_kvatop,
439 .ka_native = _amd64_native,
440 .ka_walk_pages = _amd64_minidump_walk_pages,
443 KVM_ARCH(kvm_amd64_minidump);