2 * Copyright (c) 2006 Peter Wemm
3 * Copyright (c) 2015 The FreeBSD Foundation
6 * This software was developed by Andrew Turner under
7 * sponsorship from the FreeBSD Foundation.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_watchdog.h"
37 #include "opt_watchdog.h"
39 #include <sys/param.h>
40 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/kerneldump.h>
45 #include <sys/msgbuf.h>
46 #include <sys/watchdog.h>
47 #include <sys/vmmeter.h>
50 #include <vm/vm_param.h>
51 #include <vm/vm_page.h>
52 #include <vm/vm_phys.h>
55 #include <machine/md_var.h>
56 #include <machine/pte.h>
57 #include <machine/minidump.h>
59 CTASSERT(sizeof(struct kerneldumpheader) == 512);
62 * Don't touch the first SIZEOF_METADATA bytes on the dump device. This
63 * is to protect us from metadata and to protect metadata from us.
65 #define SIZEOF_METADATA (64*1024)
67 uint64_t *vm_page_dump;
68 int vm_page_dump_size;
70 static struct kerneldumpheader kdh;
73 /* Handle chunked writes. */
76 static size_t counter, progress, dumpsize;
78 static uint64_t tmpbuffer[PAGE_SIZE / sizeof(uint64_t)];
80 CTASSERT(sizeof(*vm_page_dump) == 8);
83 is_dumpable(vm_paddr_t pa)
88 if ((m = vm_phys_paddr_to_vm_page(pa)) != NULL)
89 return ((m->flags & PG_NODUMP) == 0);
90 for (i = 0; dump_avail[i] != 0 || dump_avail[i + 1] != 0; i += 2) {
91 if (pa >= dump_avail[i] && pa < dump_avail[i + 1])
98 blk_flush(struct dumperinfo *di)
105 error = dump_write(di, dump_va, 0, dumplo, fragsz);
115 } progress_track[10] = {
129 report_progress(size_t progress, size_t dumpsize)
133 sofar = 100 - ((progress * 100) / dumpsize);
134 for (i = 0; i < nitems(progress_track); i++) {
135 if (sofar < progress_track[i].min_per ||
136 sofar > progress_track[i].max_per)
138 if (progress_track[i].visited)
140 progress_track[i].visited = 1;
141 printf("..%d%%", sofar);
147 blk_write(struct dumperinfo *di, char *ptr, vm_paddr_t pa, size_t sz)
153 maxdumpsz = min(di->maxiosize, MAXDUMPPGS * PAGE_SIZE);
154 if (maxdumpsz == 0) /* seatbelt */
155 maxdumpsz = PAGE_SIZE;
157 if ((sz % PAGE_SIZE) != 0) {
158 printf("size not page aligned\n");
161 if (ptr != NULL && pa != 0) {
162 printf("cant have both va and pa!\n");
165 if ((((uintptr_t)pa) % PAGE_SIZE) != 0) {
166 printf("address not page aligned %p\n", ptr);
171 * If we're doing a virtual dump, flush any
172 * pre-existing pa pages.
174 error = blk_flush(di);
179 len = maxdumpsz - fragsz;
185 report_progress(progress, dumpsize);
186 counter &= (1 << 22) - 1;
189 wdog_kern_pat(WD_LASTVAL);
192 error = dump_write(di, ptr, 0, dumplo, len);
199 dump_va = (void *)PHYS_TO_DMAP(pa);
203 error = blk_flush(di);
208 /* Check for user abort. */
213 printf(" (CTRL-C to abort) ");
220 minidumpsys(struct dumperinfo *di)
222 pd_entry_t *l0, *l1, *l2;
231 struct minidumphdr mdhdr;
238 for (va = VM_MIN_KERNEL_ADDRESS; va < kernel_vm_end; va += L2_SIZE) {
239 pmapsize += PAGE_SIZE;
240 if (!pmap_get_tables(pmap_kernel(), va, &l0, &l1, &l2, &l3))
243 /* We should always be using the l2 table for kvm */
247 if ((*l2 & ATTR_DESCR_MASK) == L2_BLOCK) {
248 pa = *l2 & ~ATTR_MASK;
249 for (i = 0; i < Ln_ENTRIES; i++, pa += PAGE_SIZE) {
253 } else if ((*l2 & ATTR_DESCR_MASK) == L2_TABLE) {
254 for (i = 0; i < Ln_ENTRIES; i++) {
255 if ((l3[i] & ATTR_DESCR_MASK) != L3_PAGE)
257 pa = l3[i] & ~ATTR_MASK;
264 /* Calculate dump size. */
266 dumpsize += round_page(msgbufp->msg_size);
267 dumpsize += round_page(vm_page_dump_size);
268 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
269 bits = vm_page_dump[i];
271 bit = ffsl(bits) - 1;
272 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) +
274 /* Clear out undumpable pages now if needed */
276 dumpsize += PAGE_SIZE;
279 bits &= ~(1ul << bit);
282 dumpsize += PAGE_SIZE;
284 /* Determine dump offset on device. */
285 if (di->mediasize < SIZEOF_METADATA + dumpsize + di->blocksize * 2 +
286 kerneldumpcrypto_dumpkeysize(di->kdc)) {
290 dumplo = di->mediaoffset + di->mediasize - dumpsize;
291 dumplo -= di->blocksize * 2;
292 dumplo -= kerneldumpcrypto_dumpkeysize(di->kdc);
295 /* Initialize kernel dump crypto. */
296 error = kerneldumpcrypto_init(di->kdc);
300 /* Initialize mdhdr */
301 bzero(&mdhdr, sizeof(mdhdr));
302 strcpy(mdhdr.magic, MINIDUMP_MAGIC);
303 mdhdr.version = MINIDUMP_VERSION;
304 mdhdr.msgbufsize = msgbufp->msg_size;
305 mdhdr.bitmapsize = vm_page_dump_size;
306 mdhdr.pmapsize = pmapsize;
307 mdhdr.kernbase = VM_MIN_KERNEL_ADDRESS;
308 mdhdr.dmapphys = DMAP_MIN_PHYSADDR;
309 mdhdr.dmapbase = DMAP_MIN_ADDRESS;
310 mdhdr.dmapend = DMAP_MAX_ADDRESS;
312 mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_AARCH64_VERSION,
313 dumpsize, kerneldumpcrypto_dumpkeysize(di->kdc), di->blocksize);
315 printf("Dumping %llu out of %ju MB:", (long long)dumpsize >> 20,
316 ptoa((uintmax_t)physmem) / 1048576);
319 error = dump_write_header(di, &kdh, 0, dumplo);
322 dumplo += di->blocksize;
325 error = dump_write_key(di, 0, dumplo);
328 dumplo += kerneldumpcrypto_dumpkeysize(di->kdc);
331 bzero(&tmpbuffer, sizeof(tmpbuffer));
332 bcopy(&mdhdr, &tmpbuffer, sizeof(mdhdr));
333 error = blk_write(di, (char *)&tmpbuffer, 0, PAGE_SIZE);
337 /* Dump msgbuf up front */
338 error = blk_write(di, (char *)msgbufp->msg_ptr, 0,
339 round_page(msgbufp->msg_size));
344 error = blk_write(di, (char *)vm_page_dump, 0,
345 round_page(vm_page_dump_size));
349 /* Dump kernel page directory pages */
350 bzero(&tmpbuffer, sizeof(tmpbuffer));
351 for (va = VM_MIN_KERNEL_ADDRESS; va < kernel_vm_end; va += L2_SIZE) {
352 if (!pmap_get_tables(pmap_kernel(), va, &l0, &l1, &l2, &l3)) {
353 /* We always write a page, even if it is zero */
354 error = blk_write(di, (char *)&tmpbuffer, 0, PAGE_SIZE);
357 /* flush, in case we reuse tmpbuffer in the same block*/
358 error = blk_flush(di);
361 } else if (l2 == NULL) {
362 pa = (*l1 & ~ATTR_MASK) | (va & L1_OFFSET);
364 /* Generate fake l3 entries based upon the l1 entry */
365 for (i = 0; i < Ln_ENTRIES; i++) {
366 tmpbuffer[i] = pa + (i * PAGE_SIZE) |
367 ATTR_DEFAULT | L3_PAGE;
369 /* We always write a page, even if it is zero */
370 error = blk_write(di, (char *)&tmpbuffer, 0, PAGE_SIZE);
373 /* flush, in case we reuse tmpbuffer in the same block*/
374 error = blk_flush(di);
377 bzero(&tmpbuffer, sizeof(tmpbuffer));
378 } else if ((*l2 & ATTR_DESCR_MASK) == L2_BLOCK) {
379 /* TODO: Handle an invalid L2 entry */
380 pa = (*l2 & ~ATTR_MASK) | (va & L2_OFFSET);
382 /* Generate fake l3 entries based upon the l1 entry */
383 for (i = 0; i < Ln_ENTRIES; i++) {
384 tmpbuffer[i] = pa + (i * PAGE_SIZE) |
385 ATTR_DEFAULT | L3_PAGE;
387 /* We always write a page, even if it is zero */
388 error = blk_write(di, (char *)&tmpbuffer, 0, PAGE_SIZE);
391 /* flush, in case we reuse fakepd in the same block */
392 error = blk_flush(di);
395 bzero(&tmpbuffer, sizeof(tmpbuffer));
398 pa = *l2 & ~ATTR_MASK;
400 /* We always write a page, even if it is zero */
401 error = blk_write(di, NULL, pa, PAGE_SIZE);
407 /* Dump memory chunks */
408 /* XXX cluster it up and use blk_dump() */
409 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
410 bits = vm_page_dump[i];
412 bit = ffsl(bits) - 1;
413 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) +
415 error = blk_write(di, 0, pa, PAGE_SIZE);
418 bits &= ~(1ul << bit);
422 error = blk_flush(di);
427 error = dump_write_header(di, &kdh, 0, dumplo);
430 dumplo += di->blocksize;
432 /* Signal completion, signoff and exit stage left. */
433 dump_write(di, NULL, 0, 0, 0);
434 printf("\nDump complete\n");
442 if (error == ENOSPC) {
443 printf("Dump map grown while dumping. ");
444 if (retry_count < 5) {
445 printf("Retrying...\n");
448 printf("Dump failed.\n");
450 else if (error == ECANCELED)
451 printf("Dump aborted\n");
452 else if (error == E2BIG)
453 printf("Dump failed. Partition too small.\n");
455 printf("** DUMP FAILED (ERROR %d) **\n", error);
460 dump_add_page(vm_paddr_t pa)
465 idx = pa >> 6; /* 2^6 = 64 */
467 atomic_set_long(&vm_page_dump[idx], 1ul << bit);
471 dump_drop_page(vm_paddr_t pa)
476 idx = pa >> 6; /* 2^6 = 64 */
478 atomic_clear_long(&vm_page_dump[idx], 1ul << bit);