linux/tools/objtool/special.c
Peter Zijlstra 8b5fa6bc32 objtool: Optimize read_sections()
Perf showed that __hash_init() is a significant portion of
read_sections(), so instead of doing a per section rela_hash, use an
elf-wide rela_hash.

Statistics show us there are about 1.1 million relas, so size it
accordingly.

This reduces the objtool on vmlinux.o runtime to a third, from 15 to 5
seconds.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Miroslav Benes <mbenes@suse.cz>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Link: https://lkml.kernel.org/r/20200324160924.739153726@infradead.org
2020-03-25 18:28:30 +01:00

200 lines
4.5 KiB
C

// SPDX-License-Identifier: GPL-2.0-or-later
/*
* Copyright (C) 2015 Josh Poimboeuf <jpoimboe@redhat.com>
*/
/*
* This file reads all the special sections which have alternate instructions
* which can be patched in or redirected to at runtime.
*/
#include <stdlib.h>
#include <string.h>
#include "builtin.h"
#include "special.h"
#include "warn.h"
#define EX_ENTRY_SIZE 12
#define EX_ORIG_OFFSET 0
#define EX_NEW_OFFSET 4
#define JUMP_ENTRY_SIZE 16
#define JUMP_ORIG_OFFSET 0
#define JUMP_NEW_OFFSET 4
#define ALT_ENTRY_SIZE 13
#define ALT_ORIG_OFFSET 0
#define ALT_NEW_OFFSET 4
#define ALT_FEATURE_OFFSET 8
#define ALT_ORIG_LEN_OFFSET 10
#define ALT_NEW_LEN_OFFSET 11
#define X86_FEATURE_POPCNT (4*32+23)
#define X86_FEATURE_SMAP (9*32+20)
struct special_entry {
const char *sec;
bool group, jump_or_nop;
unsigned char size, orig, new;
unsigned char orig_len, new_len; /* group only */
unsigned char feature; /* ALTERNATIVE macro CPU feature */
};
struct special_entry entries[] = {
{
.sec = ".altinstructions",
.group = true,
.size = ALT_ENTRY_SIZE,
.orig = ALT_ORIG_OFFSET,
.orig_len = ALT_ORIG_LEN_OFFSET,
.new = ALT_NEW_OFFSET,
.new_len = ALT_NEW_LEN_OFFSET,
.feature = ALT_FEATURE_OFFSET,
},
{
.sec = "__jump_table",
.jump_or_nop = true,
.size = JUMP_ENTRY_SIZE,
.orig = JUMP_ORIG_OFFSET,
.new = JUMP_NEW_OFFSET,
},
{
.sec = "__ex_table",
.size = EX_ENTRY_SIZE,
.orig = EX_ORIG_OFFSET,
.new = EX_NEW_OFFSET,
},
{},
};
static int get_alt_entry(struct elf *elf, struct special_entry *entry,
struct section *sec, int idx,
struct special_alt *alt)
{
struct rela *orig_rela, *new_rela;
unsigned long offset;
offset = idx * entry->size;
alt->group = entry->group;
alt->jump_or_nop = entry->jump_or_nop;
if (alt->group) {
alt->orig_len = *(unsigned char *)(sec->data->d_buf + offset +
entry->orig_len);
alt->new_len = *(unsigned char *)(sec->data->d_buf + offset +
entry->new_len);
}
if (entry->feature) {
unsigned short feature;
feature = *(unsigned short *)(sec->data->d_buf + offset +
entry->feature);
/*
* It has been requested that we don't validate the !POPCNT
* feature path which is a "very very small percentage of
* machines".
*/
if (feature == X86_FEATURE_POPCNT)
alt->skip_orig = true;
/*
* If UACCESS validation is enabled; force that alternative;
* otherwise force it the other way.
*
* What we want to avoid is having both the original and the
* alternative code flow at the same time, in that case we can
* find paths that see the STAC but take the NOP instead of
* CLAC and the other way around.
*/
if (feature == X86_FEATURE_SMAP) {
if (uaccess)
alt->skip_orig = true;
else
alt->skip_alt = true;
}
}
orig_rela = find_rela_by_dest(elf, sec, offset + entry->orig);
if (!orig_rela) {
WARN_FUNC("can't find orig rela", sec, offset + entry->orig);
return -1;
}
if (orig_rela->sym->type != STT_SECTION) {
WARN_FUNC("don't know how to handle non-section rela symbol %s",
sec, offset + entry->orig, orig_rela->sym->name);
return -1;
}
alt->orig_sec = orig_rela->sym->sec;
alt->orig_off = orig_rela->addend;
if (!entry->group || alt->new_len) {
new_rela = find_rela_by_dest(elf, sec, offset + entry->new);
if (!new_rela) {
WARN_FUNC("can't find new rela",
sec, offset + entry->new);
return -1;
}
alt->new_sec = new_rela->sym->sec;
alt->new_off = (unsigned int)new_rela->addend;
/* _ASM_EXTABLE_EX hack */
if (alt->new_off >= 0x7ffffff0)
alt->new_off -= 0x7ffffff0;
}
return 0;
}
/*
* Read all the special sections and create a list of special_alt structs which
* describe all the alternate instructions which can be patched in or
* redirected to at runtime.
*/
int special_get_alts(struct elf *elf, struct list_head *alts)
{
struct special_entry *entry;
struct section *sec;
unsigned int nr_entries;
struct special_alt *alt;
int idx, ret;
INIT_LIST_HEAD(alts);
for (entry = entries; entry->sec; entry++) {
sec = find_section_by_name(elf, entry->sec);
if (!sec)
continue;
if (sec->len % entry->size != 0) {
WARN("%s size not a multiple of %d",
sec->name, entry->size);
return -1;
}
nr_entries = sec->len / entry->size;
for (idx = 0; idx < nr_entries; idx++) {
alt = malloc(sizeof(*alt));
if (!alt) {
WARN("malloc failed");
return -1;
}
memset(alt, 0, sizeof(*alt));
ret = get_alt_entry(elf, entry, sec, idx, alt);
if (ret)
return ret;
list_add_tail(&alt->list, alts);
}
}
return 0;
}