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.
26 * $FreeBSD: src/sys/i386/i386/minidump_machdep.c,v 1.9 2009/05/29 21:27:12 jamie Exp $
29 #include <sys/param.h>
30 #include <sys/systm.h>
33 #include <sys/device.h>
34 #include <sys/globaldata.h>
35 #include <sys/kernel.h>
36 #include <sys/kerneldump.h>
37 #include <sys/msgbuf.h>
39 #include <vm/vm_kern.h>
41 #include <machine/atomic.h>
42 #include <machine/elf.h>
43 #include <machine/globaldata.h>
44 #include <machine/md_var.h>
45 #include <machine/vmparam.h>
46 #include <machine/minidump.h>
48 CTASSERT(sizeof(struct kerneldumpheader) == 512);
51 * Don't touch the first SIZEOF_METADATA bytes on the dump device. This
52 * is to protect us from metadata and to protect metadata from us.
54 #define SIZEOF_METADATA (64*1024)
56 #define MD_ALIGN(x) (((off_t)(x) + PAGE_MASK) & ~PAGE_MASK)
57 #define DEV_ALIGN(x) (((off_t)(x) + (DEV_BSIZE-1)) & ~(DEV_BSIZE-1))
59 uint32_t *vm_page_dump;
60 int vm_page_dump_size;
62 static struct kerneldumpheader kdh;
65 /* Handle chunked writes. */
68 static uint64_t counter, progress;
70 CTASSERT(sizeof(*vm_page_dump) == 4);
73 is_dumpable(vm_paddr_t pa)
77 for (i = 0; dump_avail[i] != 0 || dump_avail[i + 1] != 0; i += 2) {
78 if (pa >= dump_avail[i] && pa < dump_avail[i + 1])
84 #define PG2MB(pgs) (((pgs) + (1 << 8) - 1) >> 8)
87 blk_flush(struct dumperinfo *di)
94 error = dev_ddump(di->priv, dump_va, 0, dumplo, fragsz);
101 blk_write(struct dumperinfo *di, char *ptr, vm_paddr_t pa, size_t sz)
107 if ((sz % PAGE_SIZE) != 0) {
108 kprintf("size not page aligned\n");
111 if (ptr != NULL && pa != 0) {
112 kprintf("can't have both va and pa!\n");
115 if (pa != 0 && (((uintptr_t)ptr) % PAGE_SIZE) != 0) {
116 kprintf("address not page aligned\n");
120 /* If we're doing a virtual dump, flush any pre-existing pa pages */
121 error = blk_flush(di);
126 len = (MAXDUMPPGS * PAGE_SIZE) - fragsz;
132 kprintf(" %lld", PG2MB(progress >> PAGE_SHIFT));
133 counter &= (1<<24) - 1;
136 error = dev_ddump(di->priv, ptr, 0, dumplo, len);
143 for (i = 0; i < len; i += PAGE_SIZE) {
144 dump_va = pmap_kenter_temporary(pa + i,
145 (i + fragsz) >> PAGE_SHIFT);
151 if (fragsz == (MAXDUMPPGS * PAGE_SIZE)) {
152 error = blk_flush(di);
158 /* Check for user abort. */
163 kprintf(" (CTRL-C to abort) ");
169 /* A fake page table page, to avoid having to handle both 4K and 2M pages */
170 static pt_entry_t fakept[NPTEPG];
173 minidumpsys(struct dumperinfo *di)
178 vm_offset_t kern_end;
185 struct minidumphdr mdhdr;
186 struct mdglobaldata *md;
191 md = (struct mdglobaldata *)globaldata_find(0);
193 kern_end = kernel_vm_end;
194 if (kern_end < (vm_offset_t)&(md[ncpus]))
195 kern_end = (vm_offset_t)&(md[ncpus]);
197 kern_end = 0xFFFFF000;
200 /* Walk page table pages, set bits in vm_page_dump */
201 for (va = KERNBASE; va < kern_end; va += NBPDR) {
203 * We always write a page, even if it is zero. Each
204 * page written corresponds to 2MB of space
206 ptesize += PAGE_SIZE;
207 pd = (pd_entry_t *)((uintptr_t)IdlePTD + KERNBASE); /* always mapped! */
209 if ((pd[j] & (PG_PS | PG_V)) == (PG_PS | PG_V)) {
210 /* This is an entire 2M page. */
211 pa = pd[j] & PG_FRAME & ~PDRMASK;
212 for (k = 0; k < NPTEPG; k++) {
219 if ((pd[j] & PG_V) == PG_V) {
220 /* set bit for each valid page in this 2MB block */
221 pt = pmap_kenter_temporary(pd[j] & PG_FRAME, 0);
223 for (k = 0; k < NPTEPG; k++) {
224 if ((pt[k] & PG_V) == PG_V) {
225 pa = pt[k] & PG_FRAME;
231 /* nothing, we're going to dump a null page */
235 /* Calculate dump size. */
237 dumpsize += round_page(msgbufp->msg_size);
238 dumpsize += round_page(vm_page_dump_size);
239 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
240 bits = vm_page_dump[i];
243 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + bit) * PAGE_SIZE;
244 /* Clear out undumpable pages now if needed */
245 if (is_dumpable(pa)) {
246 dumpsize += PAGE_SIZE;
250 bits &= ~(1ul << bit);
253 dumpsize += PAGE_SIZE;
255 /* Determine dump offset on device. */
256 if (di->mediasize < SIZEOF_METADATA + dumpsize + sizeof(kdh) * 2) {
260 dumplo = di->mediaoffset + di->mediasize - dumpsize;
261 dumplo -= sizeof(kdh) * 2;
264 /* Initialize mdhdr */
265 bzero(&mdhdr, sizeof(mdhdr));
266 strcpy(mdhdr.magic, MINIDUMP_MAGIC);
267 mdhdr.version = MINIDUMP_VERSION;
268 mdhdr.msgbufsize = msgbufp->msg_size;
269 mdhdr.bitmapsize = vm_page_dump_size;
270 mdhdr.ptesize = ptesize;
271 mdhdr.kernbase = KERNBASE;
273 mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_I386_VERSION,
274 dumpsize, di->blocksize);
276 kprintf("Physical memory: %jd MB\n", (intmax_t)ptoa(physmem) / 1048576);
277 kprintf("Dumping %jd MB:", (intmax_t)dumpsize >> 20);
280 error = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
283 dumplo += sizeof(kdh);
286 bzero(&fakept, sizeof(fakept));
287 bcopy(&mdhdr, &fakept, sizeof(mdhdr));
288 error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
292 /* Dump msgbuf up front */
293 error = blk_write(di, (char *)msgbufp->msg_ptr, 0, round_page(msgbufp->msg_size));
298 error = blk_write(di, (char *)vm_page_dump, 0, round_page(vm_page_dump_size));
302 /* Dump kernel page table pages */
303 for (va = KERNBASE; va < kern_end; va += NBPDR) {
304 /* We always write a page, even if it is zero */
305 pd = (pd_entry_t *)((uintptr_t)IdlePTD + KERNBASE); /* always mapped! */
307 if ((pd[j] & (PG_PS | PG_V)) == (PG_PS | PG_V)) {
308 /* This is a single 2M block. Generate a fake PTP */
309 pa = pd[j] & PG_FRAME & ~PDRMASK;
310 for (k = 0; k < NPTEPG; k++) {
311 fakept[k] = (pa + (k * PAGE_SIZE)) | PG_V | PG_RW | PG_A | PG_M;
313 error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
316 /* flush, in case we reuse fakept in the same block */
317 error = blk_flush(di);
322 if ((pd[j] & PG_V) == PG_V) {
323 pa = pd[j] & PG_FRAME;
324 error = blk_write(di, 0, pa, PAGE_SIZE);
328 bzero(fakept, sizeof(fakept));
329 error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
332 /* flush, in case we reuse fakept in the same block */
333 error = blk_flush(di);
339 /* Dump memory chunks */
340 /* XXX cluster it up and use blk_dump() */
341 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
342 bits = vm_page_dump[i];
345 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + bit) * PAGE_SIZE;
346 error = blk_write(di, 0, pa, PAGE_SIZE);
349 bits &= ~(1ul << bit);
353 error = blk_flush(di);
358 error = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
361 dumplo += sizeof(kdh);
363 /* Signal completion, signoff and exit stage left. */
364 dev_ddump(di->priv, NULL, 0, 0, 0);
365 kprintf("\nDump complete\n");
372 if (error == ECANCELED)
373 kprintf("\nDump aborted\n");
374 else if (error == ENOSPC)
375 kprintf("\nDump failed. Partition too small.\n");
377 kprintf("\n** DUMP FAILED (ERROR %d) **\n", error);
381 dump_add_page(vm_paddr_t pa)
386 idx = pa >> 5; /* 2^5 = 32 */
388 atomic_set_int(&vm_page_dump[idx], 1ul << bit);
392 dump_drop_page(vm_paddr_t pa)
397 idx = pa >> 5; /* 2^5 = 32 */
399 atomic_clear_int(&vm_page_dump[idx], 1ul << bit);