freebsd-nq/lib/libkvm/kvm_minidump_amd64.c
Will Andrews ffdeef3234 libkvm: Improve physical address lookup scaling.
Instead of using a hash table to convert physical page addresses to offsets
in the sparse page array, cache the number of bits set for each 4MB chunk of
physical pages.  Upon lookup, find the nearest cached population count, then
add/subtract the number of bits from that point to the page's PTE bit.
Then multiply by page size and add to the sparse page map's base offset.

This replaces O(n) worst-case lookup with O(1) (plus a small number of bits
to scan in the bitmap).  Also, for a 128GB system, a typical kernel core of
about 8GB will now only require ~4.5MB of RAM for this approach instead of
~48MB as with the hash table.

More concretely, /usr/sbin/crashinfo against the same core improves from a
max RSS of 188MB and wall time of 43.72s (33.25 user 2.94 sys) to 135MB and
9.43s (2.58 user 1.47 sys).  Running "thread apply all bt" in kgdb has a
similar RSS improvement, and wall time drops from 4.44s to 1.93s.

Reviewed by:	jhb
Sponsored by:	Backtrace I/O
2016-07-18 01:55:25 +00:00

309 lines
8.5 KiB
C

/*-
* Copyright (c) 2006 Peter Wemm
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
/*
* AMD64 machine dependent routines for kvm and minidumps.
*/
#include <sys/param.h>
#include <sys/endian.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <kvm.h>
#include "../../sys/amd64/include/minidump.h"
#include <limits.h>
#include "kvm_private.h"
#include "kvm_amd64.h"
#define amd64_round_page(x) roundup2((kvaddr_t)(x), AMD64_PAGE_SIZE)
struct vmstate {
struct minidumphdr hdr;
amd64_pte_t *page_map;
};
static int
_amd64_minidump_probe(kvm_t *kd)
{
return (_kvm_probe_elf_kernel(kd, ELFCLASS64, EM_X86_64) &&
_kvm_is_minidump(kd));
}
static void
_amd64_minidump_freevtop(kvm_t *kd)
{
struct vmstate *vm = kd->vmst;
free(vm->page_map);
free(vm);
kd->vmst = NULL;
}
static int
_amd64_minidump_initvtop(kvm_t *kd)
{
struct vmstate *vmst;
off_t off, sparse_off;
vmst = _kvm_malloc(kd, sizeof(*vmst));
if (vmst == NULL) {
_kvm_err(kd, kd->program, "cannot allocate vm");
return (-1);
}
kd->vmst = vmst;
if (pread(kd->pmfd, &vmst->hdr, sizeof(vmst->hdr), 0) !=
sizeof(vmst->hdr)) {
_kvm_err(kd, kd->program, "cannot read dump header");
return (-1);
}
if (strncmp(MINIDUMP_MAGIC, vmst->hdr.magic, sizeof(vmst->hdr.magic)) != 0) {
_kvm_err(kd, kd->program, "not a minidump for this platform");
return (-1);
}
/*
* NB: amd64 minidump header is binary compatible between version 1
* and version 2; this may not be the case for the future versions.
*/
vmst->hdr.version = le32toh(vmst->hdr.version);
if (vmst->hdr.version != MINIDUMP_VERSION && vmst->hdr.version != 1) {
_kvm_err(kd, kd->program, "wrong minidump version. expected %d got %d",
MINIDUMP_VERSION, vmst->hdr.version);
return (-1);
}
vmst->hdr.msgbufsize = le32toh(vmst->hdr.msgbufsize);
vmst->hdr.bitmapsize = le32toh(vmst->hdr.bitmapsize);
vmst->hdr.pmapsize = le32toh(vmst->hdr.pmapsize);
vmst->hdr.kernbase = le64toh(vmst->hdr.kernbase);
vmst->hdr.dmapbase = le64toh(vmst->hdr.dmapbase);
vmst->hdr.dmapend = le64toh(vmst->hdr.dmapend);
/* Skip header and msgbuf */
off = AMD64_PAGE_SIZE + amd64_round_page(vmst->hdr.msgbufsize);
sparse_off = off + amd64_round_page(vmst->hdr.bitmapsize) +
amd64_round_page(vmst->hdr.pmapsize);
if (_kvm_pt_init(kd, vmst->hdr.bitmapsize, off, sparse_off,
AMD64_PAGE_SIZE, sizeof(uint64_t)) == -1) {
_kvm_err(kd, kd->program, "cannot load core bitmap");
return (-1);
}
off += amd64_round_page(vmst->hdr.bitmapsize);
vmst->page_map = _kvm_malloc(kd, vmst->hdr.pmapsize);
if (vmst->page_map == NULL) {
_kvm_err(kd, kd->program, "cannot allocate %d bytes for page_map",
vmst->hdr.pmapsize);
return (-1);
}
if (pread(kd->pmfd, vmst->page_map, vmst->hdr.pmapsize, off) !=
(ssize_t)vmst->hdr.pmapsize) {
_kvm_err(kd, kd->program, "cannot read %d bytes for page_map",
vmst->hdr.pmapsize);
return (-1);
}
off += amd64_round_page(vmst->hdr.pmapsize);
return (0);
}
static int
_amd64_minidump_vatop_v1(kvm_t *kd, kvaddr_t va, off_t *pa)
{
struct vmstate *vm;
amd64_physaddr_t offset;
amd64_pte_t pte;
kvaddr_t pteindex;
amd64_physaddr_t a;
off_t ofs;
vm = kd->vmst;
offset = va & AMD64_PAGE_MASK;
if (va >= vm->hdr.kernbase) {
pteindex = (va - vm->hdr.kernbase) >> AMD64_PAGE_SHIFT;
if (pteindex >= vm->hdr.pmapsize / sizeof(*vm->page_map))
goto invalid;
pte = le64toh(vm->page_map[pteindex]);
if ((pte & AMD64_PG_V) == 0) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop_v1: pte not valid");
goto invalid;
}
a = pte & AMD64_PG_FRAME;
ofs = _kvm_pt_find(kd, a);
if (ofs == -1) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop_v1: physical address 0x%jx not in minidump",
(uintmax_t)a);
goto invalid;
}
*pa = ofs + offset;
return (AMD64_PAGE_SIZE - offset);
} else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
ofs = _kvm_pt_find(kd, a);
if (ofs == -1) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop_v1: direct map address 0x%jx not in minidump",
(uintmax_t)va);
goto invalid;
}
*pa = ofs + offset;
return (AMD64_PAGE_SIZE - offset);
} else {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop_v1: virtual address 0x%jx not minidumped",
(uintmax_t)va);
goto invalid;
}
invalid:
_kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
return (0);
}
static int
_amd64_minidump_vatop(kvm_t *kd, kvaddr_t va, off_t *pa)
{
amd64_pte_t pt[AMD64_NPTEPG];
struct vmstate *vm;
amd64_physaddr_t offset;
amd64_pde_t pde;
amd64_pte_t pte;
kvaddr_t pteindex;
kvaddr_t pdeindex;
amd64_physaddr_t a;
off_t ofs;
vm = kd->vmst;
offset = va & AMD64_PAGE_MASK;
if (va >= vm->hdr.kernbase) {
pdeindex = (va - vm->hdr.kernbase) >> AMD64_PDRSHIFT;
if (pdeindex >= vm->hdr.pmapsize / sizeof(*vm->page_map))
goto invalid;
pde = le64toh(vm->page_map[pdeindex]);
if ((pde & AMD64_PG_V) == 0) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop: pde not valid");
goto invalid;
}
if ((pde & AMD64_PG_PS) == 0) {
a = pde & AMD64_PG_FRAME;
/* TODO: Just read the single PTE */
ofs = _kvm_pt_find(kd, a);
if (ofs == -1) {
_kvm_err(kd, kd->program,
"cannot find page table entry for %ju",
(uintmax_t)a);
goto invalid;
}
if (pread(kd->pmfd, &pt, AMD64_PAGE_SIZE, ofs) !=
AMD64_PAGE_SIZE) {
_kvm_err(kd, kd->program,
"cannot read page table entry for %ju",
(uintmax_t)a);
goto invalid;
}
pteindex = (va >> AMD64_PAGE_SHIFT) &
(AMD64_NPTEPG - 1);
pte = le64toh(pt[pteindex]);
if ((pte & AMD64_PG_V) == 0) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop: pte not valid");
goto invalid;
}
a = pte & AMD64_PG_FRAME;
} else {
a = pde & AMD64_PG_PS_FRAME;
a += (va & AMD64_PDRMASK) ^ offset;
}
ofs = _kvm_pt_find(kd, a);
if (ofs == -1) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop: physical address 0x%jx not in minidump",
(uintmax_t)a);
goto invalid;
}
*pa = ofs + offset;
return (AMD64_PAGE_SIZE - offset);
} else if (va >= vm->hdr.dmapbase && va < vm->hdr.dmapend) {
a = (va - vm->hdr.dmapbase) & ~AMD64_PAGE_MASK;
ofs = _kvm_pt_find(kd, a);
if (ofs == -1) {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop: direct map address 0x%jx not in minidump",
(uintmax_t)va);
goto invalid;
}
*pa = ofs + offset;
return (AMD64_PAGE_SIZE - offset);
} else {
_kvm_err(kd, kd->program,
"_amd64_minidump_vatop: virtual address 0x%jx not minidumped",
(uintmax_t)va);
goto invalid;
}
invalid:
_kvm_err(kd, 0, "invalid address (0x%jx)", (uintmax_t)va);
return (0);
}
static int
_amd64_minidump_kvatop(kvm_t *kd, kvaddr_t va, off_t *pa)
{
if (ISALIVE(kd)) {
_kvm_err(kd, 0,
"_amd64_minidump_kvatop called in live kernel!");
return (0);
}
if (((struct vmstate *)kd->vmst)->hdr.version == 1)
return (_amd64_minidump_vatop_v1(kd, va, pa));
else
return (_amd64_minidump_vatop(kd, va, pa));
}
struct kvm_arch kvm_amd64_minidump = {
.ka_probe = _amd64_minidump_probe,
.ka_initvtop = _amd64_minidump_initvtop,
.ka_freevtop = _amd64_minidump_freevtop,
.ka_kvatop = _amd64_minidump_kvatop,
.ka_native = _amd64_native,
};
KVM_ARCH(kvm_amd64_minidump);