2 * Copyright (c) 2006 Peter Wemm
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
30 * AMD64 machine dependent routines for kvm and minidumps.
33 #include <sys/param.h>
34 #include <sys/endian.h>
41 #include "../../sys/amd64/include/minidump.h"
45 #include "kvm_private.h"
46 #include "kvm_amd64.h"
48 #define amd64_round_page(x) roundup2((kvaddr_t)(x), AMD64_PAGE_SIZE)
49 #define VM_IS_V1(vm) (vm->hdr.version == 1)
50 #define VA_OFF(vm, va) \
51 (VM_IS_V1(vm) ? ((va) & (AMD64_PAGE_SIZE - 1)) : ((va) & AMD64_PAGE_MASK))
54 struct minidumphdr hdr;
58 _amd64_entry_to_prot(uint64_t entry)
60 vm_prot_t prot = VM_PROT_READ;
62 if ((entry & PG_RW) != 0)
63 prot |= VM_PROT_WRITE;
64 if ((entry & PG_NX) == 0)
65 prot |= VM_PROT_EXECUTE;
70 * Version 2 minidumps use page directory entries, while version 1 use page
75 _amd64_pde_get(kvm_t *kd, u_long pdeindex)
77 amd64_pde_t *pde = _kvm_pmap_get(kd, pdeindex, sizeof(*pde));
83 _amd64_pte_get(kvm_t *kd, u_long pteindex)
85 amd64_pte_t *pte = _kvm_pmap_get(kd, pteindex, sizeof(*pte));
90 /* Get the first page table entry for a given page directory index. */
92 _amd64_pde_first_pte(kvm_t *kd, u_long pdeindex)
96 pa = _kvm_pmap_get(kd, pdeindex, sizeof(amd64_pde_t));
99 return _kvm_map_get(kd, *pa & AMD64_PG_FRAME, AMD64_PAGE_SIZE);
103 _amd64_minidump_probe(kvm_t *kd)
106 return (_kvm_probe_elf_kernel(kd, ELFCLASS64, EM_X86_64) &&
107 _kvm_is_minidump(kd));
111 _amd64_minidump_freevtop(kvm_t *kd)
113 struct vmstate *vm = kd->vmst;
120 _amd64_minidump_initvtop(kvm_t *kd)
122 struct vmstate *vmst;
123 off_t off, sparse_off;
125 vmst = _kvm_malloc(kd, sizeof(*vmst));
127 _kvm_err(kd, kd->program, "cannot allocate vm");
131 if (pread(kd->pmfd, &vmst->hdr, sizeof(vmst->hdr), 0) !=
133 _kvm_err(kd, kd->program, "cannot read dump header");
136 if (strncmp(MINIDUMP_MAGIC, vmst->hdr.magic, sizeof(vmst->hdr.magic)) != 0) {
137 _kvm_err(kd, kd->program, "not a minidump for this platform");
142 * NB: amd64 minidump header is binary compatible between version 1
143 * and version 2; this may not be the case for the future versions.
145 vmst->hdr.version = le32toh(vmst->hdr.version);
146 if (vmst->hdr.version != MINIDUMP_VERSION && vmst->hdr.version != 1) {
147 _kvm_err(kd, kd->program, "wrong minidump version. expected %d got %d",
148 MINIDUMP_VERSION, vmst->hdr.version);
151 vmst->hdr.msgbufsize = le32toh(vmst->hdr.msgbufsize);
152 vmst->hdr.bitmapsize = le32toh(vmst->hdr.bitmapsize);
153 vmst->hdr.pmapsize = le32toh(vmst->hdr.pmapsize);
154 vmst->hdr.kernbase = le64toh(vmst->hdr.kernbase);
155 vmst->hdr.dmapbase = le64toh(vmst->hdr.dmapbase);
156 vmst->hdr.dmapend = le64toh(vmst->hdr.dmapend);
158 /* Skip header and msgbuf */
159 off = AMD64_PAGE_SIZE + amd64_round_page(vmst->hdr.msgbufsize);
161 sparse_off = off + amd64_round_page(vmst->hdr.bitmapsize) +
162 amd64_round_page(vmst->hdr.pmapsize);
163 if (_kvm_pt_init(kd, vmst->hdr.bitmapsize, off, sparse_off,
164 AMD64_PAGE_SIZE, sizeof(uint64_t)) == -1) {
167 off += amd64_round_page(vmst->hdr.bitmapsize);
169 if (_kvm_pmap_init(kd, vmst->hdr.pmapsize, off) == -1) {
172 off += amd64_round_page(vmst->hdr.pmapsize);
178 _amd64_minidump_vatop_v1(kvm_t *kd, kvaddr_t va, off_t *pa)
181 amd64_physaddr_t offset;
188 offset = va & AMD64_PAGE_MASK;
190 if (va >= vm->hdr.kernbase) {
191 pteindex = (va - vm->hdr.kernbase) >> AMD64_PAGE_SHIFT;
192 if (pteindex >= vm->hdr.pmapsize / sizeof(pte))
194 pte = _amd64_pte_get(kd, pteindex);
195 if ((pte & AMD64_PG_V) == 0) {
196 _kvm_err(kd, kd->program,
197 "_amd64_minidump_vatop_v1: pte not valid");
200 a = pte & AMD64_PG_FRAME;
201 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
203 _kvm_err(kd, kd->program,
204 "_amd64_minidump_vatop_v1: physical address 0x%jx not in minidump",
209 return (AMD64_PAGE_SIZE - offset);
210 } else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
211 a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
212 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
214 _kvm_err(kd, kd->program,
215 "_amd64_minidump_vatop_v1: direct map address 0x%jx not in minidump",
220 return (AMD64_PAGE_SIZE - offset);
222 _kvm_err(kd, kd->program,
223 "_amd64_minidump_vatop_v1: virtual address 0x%jx not minidumped",
229 _kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
234 _amd64_minidump_vatop(kvm_t *kd, kvaddr_t va, off_t *pa)
236 amd64_pte_t pt[AMD64_NPTEPG];
238 amd64_physaddr_t offset;
247 offset = va & AMD64_PAGE_MASK;
249 if (va >= vm->hdr.kernbase) {
250 pdeindex = (va - vm->hdr.kernbase) >> AMD64_PDRSHIFT;
251 if (pdeindex >= vm->hdr.pmapsize / sizeof(pde))
253 pde = _amd64_pde_get(kd, pdeindex);
254 if ((pde & AMD64_PG_V) == 0) {
255 _kvm_err(kd, kd->program,
256 "_amd64_minidump_vatop: pde not valid");
259 if ((pde & AMD64_PG_PS) == 0) {
260 a = pde & AMD64_PG_FRAME;
261 /* TODO: Just read the single PTE */
262 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
264 _kvm_err(kd, kd->program,
265 "cannot find page table entry for %ju",
269 if (pread(kd->pmfd, &pt, AMD64_PAGE_SIZE, ofs) !=
271 _kvm_err(kd, kd->program,
272 "cannot read page table entry for %ju",
276 pteindex = (va >> AMD64_PAGE_SHIFT) &
278 pte = le64toh(pt[pteindex]);
279 if ((pte & AMD64_PG_V) == 0) {
280 _kvm_err(kd, kd->program,
281 "_amd64_minidump_vatop: pte not valid");
284 a = pte & AMD64_PG_FRAME;
286 a = pde & AMD64_PG_PS_FRAME;
287 a += (va & AMD64_PDRMASK) ^ offset;
289 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
291 _kvm_err(kd, kd->program,
292 "_amd64_minidump_vatop: physical address 0x%jx not in minidump",
297 return (AMD64_PAGE_SIZE - offset);
298 } else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
299 a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
300 ofs = _kvm_pt_find(kd, a, AMD64_PAGE_SIZE);
302 _kvm_err(kd, kd->program,
303 "_amd64_minidump_vatop: direct map address 0x%jx not in minidump",
308 return (AMD64_PAGE_SIZE - offset);
310 _kvm_err(kd, kd->program,
311 "_amd64_minidump_vatop: virtual address 0x%jx not minidumped",
317 _kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
322 _amd64_minidump_kvatop(kvm_t *kd, kvaddr_t va, off_t *pa)
327 "_amd64_minidump_kvatop called in live kernel!");
330 if (((struct vmstate *)kd->vmst)->hdr.version == 1)
331 return (_amd64_minidump_vatop_v1(kd, va, pa));
333 return (_amd64_minidump_vatop(kd, va, pa));
337 _amd64_minidump_walk_pages(kvm_t *kd, kvm_walk_pages_cb_t *cb, void *arg)
339 struct vmstate *vm = kd->vmst;
340 u_long npdes = vm->hdr.pmapsize / sizeof(amd64_pde_t);
341 u_long bmindex, dva, pa, pdeindex, va;
342 struct kvm_bitmap bm;
345 unsigned int pgsz = AMD64_PAGE_SIZE;
347 if (vm->hdr.version < 2)
350 if (!_kvm_bitmap_init(&bm, vm->hdr.bitmapsize, &bmindex))
353 for (pdeindex = 0; pdeindex < npdes; pdeindex++) {
354 pd_entry_t pde = _amd64_pde_get(kd, pdeindex);
358 va = vm->hdr.kernbase + (pdeindex << AMD64_PDRSHIFT);
359 if ((pde & PG_V) == 0)
362 if ((pde & AMD64_PG_PS) != 0) {
364 * Large page. Iterate on each 4K page section
365 * within this page. This differs from 4K pages in
366 * that every page here uses the same PDE to
367 * generate permissions.
369 pa = pde & AMD64_PG_PS_FRAME +
370 ((va & AMD64_PDRMASK) ^ VA_OFF(vm, va));
371 dva = vm->hdr.dmapbase + pa;
372 _kvm_bitmap_set(&bm, pa, AMD64_PAGE_SIZE);
373 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva,
374 _amd64_entry_to_prot(pde), AMD64_NBPDR, pgsz)) {
380 /* 4K pages: pde references another page of entries. */
381 ptes = _amd64_pde_first_pte(kd, pdeindex);
382 /* Ignore page directory pages that were not dumped. */
386 for (i = 0; i < NPTEPG; i++) {
387 pt_entry_t pte = (u_long)ptes[i];
389 pa = pte & AMD64_PG_FRAME;
390 dva = vm->hdr.dmapbase + pa;
391 if ((pte & PG_V) != 0) {
392 _kvm_bitmap_set(&bm, pa, AMD64_PAGE_SIZE);
393 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva,
394 _amd64_entry_to_prot(pte), pgsz, 0)) {
398 va += AMD64_PAGE_SIZE;
402 while (_kvm_bitmap_next(&bm, &bmindex)) {
403 pa = bmindex * AMD64_PAGE_SIZE;
404 dva = vm->hdr.dmapbase + pa;
405 if (vm->hdr.dmapend < (dva + pgsz))
408 /* amd64/pmap.c: create_pagetables(): dmap always R|W. */
409 prot = VM_PROT_READ | VM_PROT_WRITE;
410 if (!_kvm_visit_cb(kd, cb, arg, pa, va, dva, prot, pgsz, 0)) {
418 _kvm_bitmap_deinit(&bm);
422 static struct kvm_arch kvm_amd64_minidump = {
423 .ka_probe = _amd64_minidump_probe,
424 .ka_initvtop = _amd64_minidump_initvtop,
425 .ka_freevtop = _amd64_minidump_freevtop,
426 .ka_kvatop = _amd64_minidump_kvatop,
427 .ka_native = _amd64_native,
428 .ka_walk_pages = _amd64_minidump_walk_pages,
431 KVM_ARCH(kvm_amd64_minidump);