2 * Copyright (c) 2006 Peter Wemm
5 * Redistribution and use in source and binary forms, with or without
6 * 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 ``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.
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include <sys/param.h>
31 #include <sys/systm.h>
34 #include <sys/kernel.h>
35 #include <sys/kerneldump.h>
36 #include <sys/msgbuf.h>
37 #include <sys/vimage.h>
40 #include <machine/atomic.h>
41 #include <machine/elf.h>
42 #include <machine/md_var.h>
43 #include <machine/vmparam.h>
44 #include <machine/minidump.h>
46 CTASSERT(sizeof(struct kerneldumpheader) == 512);
49 * Don't touch the first SIZEOF_METADATA bytes on the dump device. This
50 * is to protect us from metadata and to protect metadata from us.
52 #define SIZEOF_METADATA (64*1024)
54 #define MD_ALIGN(x) (((off_t)(x) + PAGE_MASK) & ~PAGE_MASK)
55 #define DEV_ALIGN(x) (((off_t)(x) + (DEV_BSIZE-1)) & ~(DEV_BSIZE-1))
57 uint32_t *vm_page_dump;
58 int vm_page_dump_size;
60 static struct kerneldumpheader kdh;
63 /* Handle chunked writes. */
66 static uint64_t counter, progress;
68 CTASSERT(sizeof(*vm_page_dump) == 4);
71 is_dumpable(vm_paddr_t pa)
75 for (i = 0; dump_avail[i] != 0 || dump_avail[i + 1] != 0; i += 2) {
76 if (pa >= dump_avail[i] && pa < dump_avail[i + 1])
82 #define PG2MB(pgs) (((pgs) + (1 << 8) - 1) >> 8)
85 blk_flush(struct dumperinfo *di)
92 error = dump_write(di, dump_va, 0, dumplo, fragsz);
99 blk_write(struct dumperinfo *di, char *ptr, vm_paddr_t pa, size_t sz)
105 maxdumpsz = di->maxiosize;
106 if (maxdumpsz == 0) /* seatbelt */
107 maxdumpsz = PAGE_SIZE;
109 if ((sz % PAGE_SIZE) != 0) {
110 printf("size not page aligned\n");
113 if (ptr != NULL && pa != 0) {
114 printf("cant have both va and pa!\n");
117 if (pa != 0 && (((uintptr_t)ptr) % PAGE_SIZE) != 0) {
118 printf("address not page aligned\n");
122 /* If we're doing a virtual dump, flush any pre-existing pa pages */
123 error = blk_flush(di);
128 len = maxdumpsz - fragsz;
134 printf(" %lld", PG2MB(progress >> PAGE_SHIFT));
135 counter &= (1<<24) - 1;
138 error = dump_write(di, ptr, 0, dumplo, len);
145 for (i = 0; i < len; i += PAGE_SIZE)
146 dump_va = pmap_kenter_temporary(pa + i, (i + fragsz) >> PAGE_SHIFT);
150 if (fragsz == maxdumpsz) {
151 error = blk_flush(di);
157 /* Check for user abort. */
162 printf(" (CTRL-C to abort) ");
168 /* A fake page table page, to avoid having to handle both 4K and 2M pages */
169 static pt_entry_t fakept[NPTEPG];
172 minidumpsys(struct dumperinfo *di)
183 struct minidumphdr mdhdr;
186 /* Walk page table pages, set bits in vm_page_dump */
188 for (va = KERNBASE; va < kernel_vm_end; va += NBPDR) {
190 * We always write a page, even if it is zero. Each
191 * page written corresponds to 2MB of space
193 ptesize += PAGE_SIZE;
194 pd = (pd_entry_t *)((uintptr_t)IdlePTD + KERNBASE); /* always mapped! */
196 if ((pd[j] & (PG_PS | PG_V)) == (PG_PS | PG_V)) {
197 /* This is an entire 2M page. */
198 pa = pd[j] & PG_PS_FRAME;
199 for (k = 0; k < NPTEPG; k++) {
206 if ((pd[j] & PG_V) == PG_V) {
207 /* set bit for each valid page in this 2MB block */
208 pt = pmap_kenter_temporary(pd[j] & PG_FRAME, 0);
209 for (k = 0; k < NPTEPG; k++) {
210 if ((pt[k] & PG_V) == PG_V) {
211 pa = pt[k] & PG_FRAME;
217 /* nothing, we're going to dump a null page */
221 /* Calculate dump size. */
223 dumpsize += round_page(msgbufp->msg_size);
224 dumpsize += round_page(vm_page_dump_size);
225 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
226 bits = vm_page_dump[i];
229 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + bit) * PAGE_SIZE;
230 /* Clear out undumpable pages now if needed */
231 if (is_dumpable(pa)) {
232 dumpsize += PAGE_SIZE;
236 bits &= ~(1ul << bit);
239 dumpsize += PAGE_SIZE;
241 /* Determine dump offset on device. */
242 if (di->mediasize < SIZEOF_METADATA + dumpsize + sizeof(kdh) * 2) {
246 dumplo = di->mediaoffset + di->mediasize - dumpsize;
247 dumplo -= sizeof(kdh) * 2;
250 /* Initialize mdhdr */
251 bzero(&mdhdr, sizeof(mdhdr));
252 strcpy(mdhdr.magic, MINIDUMP_MAGIC);
253 mdhdr.version = MINIDUMP_VERSION;
254 mdhdr.msgbufsize = msgbufp->msg_size;
255 mdhdr.bitmapsize = vm_page_dump_size;
256 mdhdr.ptesize = ptesize;
257 mdhdr.kernbase = KERNBASE;
262 mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_I386_VERSION, dumpsize, di->blocksize);
264 printf("Physical memory: %ju MB\n", ptoa((uintmax_t)physmem) / 1048576);
265 printf("Dumping %llu MB:", (long long)dumpsize >> 20);
268 error = dump_write(di, &kdh, 0, dumplo, sizeof(kdh));
271 dumplo += sizeof(kdh);
274 bzero(&fakept, sizeof(fakept));
275 bcopy(&mdhdr, &fakept, sizeof(mdhdr));
276 error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
280 /* Dump msgbuf up front */
281 error = blk_write(di, (char *)msgbufp->msg_ptr, 0, round_page(msgbufp->msg_size));
286 error = blk_write(di, (char *)vm_page_dump, 0, round_page(vm_page_dump_size));
290 /* Dump kernel page table pages */
291 for (va = KERNBASE; va < kernel_vm_end; va += NBPDR) {
292 /* We always write a page, even if it is zero */
293 pd = (pd_entry_t *)((uintptr_t)IdlePTD + KERNBASE); /* always mapped! */
295 if ((pd[j] & (PG_PS | PG_V)) == (PG_PS | PG_V)) {
296 /* This is a single 2M block. Generate a fake PTP */
297 pa = pd[j] & PG_PS_FRAME;
298 for (k = 0; k < NPTEPG; k++) {
299 fakept[k] = (pa + (k * PAGE_SIZE)) | PG_V | PG_RW | PG_A | PG_M;
301 error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
304 /* flush, in case we reuse fakept in the same block */
305 error = blk_flush(di);
310 if ((pd[j] & PG_V) == PG_V) {
311 pa = pd[j] & PG_FRAME;
312 error = blk_write(di, 0, pa, PAGE_SIZE);
316 bzero(fakept, sizeof(fakept));
317 error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
320 /* flush, in case we reuse fakept in the same block */
321 error = blk_flush(di);
327 /* Dump memory chunks */
328 /* XXX cluster it up and use blk_dump() */
329 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
330 bits = vm_page_dump[i];
333 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + bit) * PAGE_SIZE;
334 error = blk_write(di, 0, pa, PAGE_SIZE);
337 bits &= ~(1ul << bit);
341 error = blk_flush(di);
346 error = dump_write(di, &kdh, 0, dumplo, sizeof(kdh));
349 dumplo += sizeof(kdh);
351 /* Signal completion, signoff and exit stage left. */
352 dump_write(di, NULL, 0, 0, 0);
353 printf("\nDump complete\n");
360 if (error == ECANCELED)
361 printf("\nDump aborted\n");
362 else if (error == ENOSPC)
363 printf("\nDump failed. Partition too small.\n");
365 printf("\n** DUMP FAILED (ERROR %d) **\n", error);
369 dump_add_page(vm_paddr_t pa)
374 idx = pa >> 5; /* 2^5 = 32 */
376 atomic_set_int(&vm_page_dump[idx], 1ul << bit);
380 dump_drop_page(vm_paddr_t pa)
385 idx = pa >> 5; /* 2^5 = 32 */
387 atomic_clear_int(&vm_page_dump[idx], 1ul << bit);