linux/tools/objtool/check.c

2539 lines
60 KiB
C
Raw Normal View History

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 13 Based on 2 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not see http www gnu org licenses this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details [based] [from] [clk] [highbank] [c] you should have received a copy of the gnu general public license along with this program if not see http www gnu org licenses extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 355 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com> Reviewed-by: Steve Winslow <swinslow@gmail.com> Reviewed-by: Allison Randal <allison@lohutok.net> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190519154041.837383322@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-05-19 13:51:43 +00:00
// SPDX-License-Identifier: GPL-2.0-or-later
/*
* Copyright (C) 2015-2017 Josh Poimboeuf <jpoimboe@redhat.com>
*/
#include <string.h>
#include <stdlib.h>
#include "builtin.h"
#include "check.h"
#include "elf.h"
#include "special.h"
#include "arch.h"
#include "warn.h"
#include <linux/hashtable.h>
#include <linux/kernel.h>
#define FAKE_JUMP_OFFSET -1
#define C_JUMP_TABLE_SECTION ".rodata..c_jump_table"
struct alternative {
struct list_head list;
struct instruction *insn;
bool skip_orig;
};
const char *objname;
struct cfi_state initial_func_cfi;
struct instruction *find_insn(struct objtool_file *file,
struct section *sec, unsigned long offset)
{
struct instruction *insn;
hash_for_each_possible(file->insn_hash, insn, hash, offset)
if (insn->sec == sec && insn->offset == offset)
return insn;
return NULL;
}
static struct instruction *next_insn_same_sec(struct objtool_file *file,
struct instruction *insn)
{
struct instruction *next = list_next_entry(insn, list);
if (!next || &next->list == &file->insn_list || next->sec != insn->sec)
return NULL;
return next;
}
static struct instruction *next_insn_same_func(struct objtool_file *file,
struct instruction *insn)
{
struct instruction *next = list_next_entry(insn, list);
struct symbol *func = insn->func;
if (!func)
return NULL;
if (&next->list != &file->insn_list && next->func == func)
return next;
/* Check if we're already in the subfunction: */
if (func == func->cfunc)
return NULL;
/* Move to the subfunction: */
return find_insn(file, func->cfunc->sec, func->cfunc->offset);
}
#define func_for_each_insn(file, func, insn) \
for (insn = find_insn(file, func->sec, func->offset); \
insn; \
insn = next_insn_same_func(file, insn))
#define sym_for_each_insn(file, sym, insn) \
for (insn = find_insn(file, sym->sec, sym->offset); \
insn && &insn->list != &file->insn_list && \
insn->sec == sym->sec && \
insn->offset < sym->offset + sym->len; \
insn = list_next_entry(insn, list))
#define sym_for_each_insn_continue_reverse(file, sym, insn) \
for (insn = list_prev_entry(insn, list); \
&insn->list != &file->insn_list && \
insn->sec == sym->sec && insn->offset >= sym->offset; \
insn = list_prev_entry(insn, list))
#define sec_for_each_insn_from(file, insn) \
for (; insn; insn = next_insn_same_sec(file, insn))
#define sec_for_each_insn_continue(file, insn) \
for (insn = next_insn_same_sec(file, insn); insn; \
insn = next_insn_same_sec(file, insn))
static bool is_static_jump(struct instruction *insn)
{
return insn->type == INSN_JUMP_CONDITIONAL ||
insn->type == INSN_JUMP_UNCONDITIONAL;
}
static bool is_sibling_call(struct instruction *insn)
{
/* An indirect jump is either a sibling call or a jump to a table. */
if (insn->type == INSN_JUMP_DYNAMIC)
return list_empty(&insn->alts);
if (!is_static_jump(insn))
return false;
/* add_jump_destinations() sets insn->call_dest for sibling calls. */
return !!insn->call_dest;
}
/*
* This checks to see if the given function is a "noreturn" function.
*
* For global functions which are outside the scope of this object file, we
* have to keep a manual list of them.
*
* For local functions, we have to detect them manually by simply looking for
* the lack of a return instruction.
*/
static bool __dead_end_function(struct objtool_file *file, struct symbol *func,
int recursion)
{
int i;
struct instruction *insn;
bool empty = true;
/*
* Unfortunately these have to be hard coded because the noreturn
* attribute isn't provided in ELF data.
*/
static const char * const global_noreturns[] = {
"__stack_chk_fail",
"panic",
"do_exit",
"do_task_dead",
"__module_put_and_exit",
"complete_and_exit",
"__reiserfs_panic",
"lbug_with_loc",
"fortify_panic",
"usercopy_abort",
"machine_real_restart",
"rewind_stack_do_exit",
"kunit_try_catch_throw",
};
if (!func)
return false;
if (func->bind == STB_WEAK)
return false;
if (func->bind == STB_GLOBAL)
for (i = 0; i < ARRAY_SIZE(global_noreturns); i++)
if (!strcmp(func->name, global_noreturns[i]))
return true;
if (!func->len)
return false;
insn = find_insn(file, func->sec, func->offset);
if (!insn->func)
return false;
func_for_each_insn(file, func, insn) {
empty = false;
if (insn->type == INSN_RETURN)
return false;
}
if (empty)
return false;
/*
* A function can have a sibling call instead of a return. In that
* case, the function's dead-end status depends on whether the target
* of the sibling call returns.
*/
func_for_each_insn(file, func, insn) {
if (is_sibling_call(insn)) {
struct instruction *dest = insn->jump_dest;
if (!dest)
/* sibling call to another file */
return false;
/* local sibling call */
if (recursion == 5) {
/*
* Infinite recursion: two functions have
* sibling calls to each other. This is a very
* rare case. It means they aren't dead ends.
*/
return false;
}
return __dead_end_function(file, dest->func, recursion+1);
}
}
return true;
}
static bool dead_end_function(struct objtool_file *file, struct symbol *func)
{
return __dead_end_function(file, func, 0);
}
static void clear_insn_state(struct insn_state *state)
{
int i;
memset(state, 0, sizeof(*state));
state->cfa.base = CFI_UNDEFINED;
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
for (i = 0; i < CFI_NUM_REGS; i++) {
state->regs[i].base = CFI_UNDEFINED;
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
state->vals[i].base = CFI_UNDEFINED;
}
state->drap_reg = CFI_UNDEFINED;
state->drap_offset = -1;
}
/*
* Call the arch-specific instruction decoder for all the instructions and add
* them to the global instruction list.
*/
static int decode_instructions(struct objtool_file *file)
{
struct section *sec;
struct symbol *func;
unsigned long offset;
struct instruction *insn;
unsigned long nr_insns = 0;
int ret;
for_each_sec(file, sec) {
if (!(sec->sh.sh_flags & SHF_EXECINSTR))
continue;
if (strcmp(sec->name, ".altinstr_replacement") &&
strcmp(sec->name, ".altinstr_aux") &&
strncmp(sec->name, ".discard.", 9))
sec->text = true;
for (offset = 0; offset < sec->len; offset += insn->len) {
insn = malloc(sizeof(*insn));
if (!insn) {
WARN("malloc failed");
return -1;
}
memset(insn, 0, sizeof(*insn));
INIT_LIST_HEAD(&insn->alts);
clear_insn_state(&insn->state);
insn->sec = sec;
insn->offset = offset;
ret = arch_decode_instruction(file->elf, sec, offset,
sec->len - offset,
&insn->len, &insn->type,
&insn->immediate,
&insn->stack_op);
if (ret)
goto err;
hash_add(file->insn_hash, &insn->hash, insn->offset);
list_add_tail(&insn->list, &file->insn_list);
nr_insns++;
}
list_for_each_entry(func, &sec->symbol_list, list) {
if (func->type != STT_FUNC || func->alias != func)
continue;
if (!find_insn(file, sec, func->offset)) {
WARN("%s(): can't find starting instruction",
func->name);
return -1;
}
sym_for_each_insn(file, func, insn)
insn->func = func;
}
}
if (stats)
printf("nr_insns: %lu\n", nr_insns);
return 0;
err:
free(insn);
return ret;
}
/*
objtool: Assume unannotated UD2 instructions are dead ends Arnd reported some false positive warnings with GCC 7: drivers/hid/wacom_wac.o: warning: objtool: wacom_bpt3_touch()+0x2a5: stack state mismatch: cfa1=7+8 cfa2=6+16 drivers/iio/adc/vf610_adc.o: warning: objtool: vf610_adc_calculate_rates() falls through to next function vf610_adc_sample_set() drivers/pwm/pwm-hibvt.o: warning: objtool: hibvt_pwm_get_state() falls through to next function hibvt_pwm_remove() drivers/pwm/pwm-mediatek.o: warning: objtool: mtk_pwm_config() falls through to next function mtk_pwm_enable() drivers/spi/spi-bcm2835.o: warning: objtool: .text: unexpected end of section drivers/spi/spi-bcm2835aux.o: warning: objtool: .text: unexpected end of section drivers/watchdog/digicolor_wdt.o: warning: objtool: dc_wdt_get_timeleft() falls through to next function dc_wdt_restart() When GCC 7 detects a potential divide-by-zero condition, it sometimes inserts a UD2 instruction for the case where the divisor is zero, instead of letting the hardware trap on the divide instruction. Objtool doesn't consider UD2 to be fatal unless it's annotated with unreachable(). So it considers the GCC-generated UD2 to be non-fatal, and it tries to follow the control flow past the UD2 and gets confused. Previously, objtool *did* assume UD2 was always a dead end. That changed with the following commit: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") The motivation behind that change was that Peter was planning on using UD2 for __WARN(), which is *not* a dead end. However, it turns out that some emulators rely on UD2 being fatal, so he ended up using 'ud0' instead: 9a93848fe787 ("x86/debug: Implement __WARN() using UD0") For GCC 4.5+, it should be safe to go back to the previous assumption that UD2 is fatal, even when it's not annotated with unreachable(). But for pre-4.5 versions of GCC, the unreachable() macro isn't supported, so such cases of UD2 need to be explicitly annotated as reachable. Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") Link: http://lkml.kernel.org/r/e57fa9dfede25f79487da8126ee9cdf7b856db65.1501188854.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-07-27 20:56:53 +00:00
* Mark "ud2" instructions and manually annotated dead ends.
*/
static int add_dead_ends(struct objtool_file *file)
{
struct section *sec;
struct rela *rela;
struct instruction *insn;
bool found;
objtool: Assume unannotated UD2 instructions are dead ends Arnd reported some false positive warnings with GCC 7: drivers/hid/wacom_wac.o: warning: objtool: wacom_bpt3_touch()+0x2a5: stack state mismatch: cfa1=7+8 cfa2=6+16 drivers/iio/adc/vf610_adc.o: warning: objtool: vf610_adc_calculate_rates() falls through to next function vf610_adc_sample_set() drivers/pwm/pwm-hibvt.o: warning: objtool: hibvt_pwm_get_state() falls through to next function hibvt_pwm_remove() drivers/pwm/pwm-mediatek.o: warning: objtool: mtk_pwm_config() falls through to next function mtk_pwm_enable() drivers/spi/spi-bcm2835.o: warning: objtool: .text: unexpected end of section drivers/spi/spi-bcm2835aux.o: warning: objtool: .text: unexpected end of section drivers/watchdog/digicolor_wdt.o: warning: objtool: dc_wdt_get_timeleft() falls through to next function dc_wdt_restart() When GCC 7 detects a potential divide-by-zero condition, it sometimes inserts a UD2 instruction for the case where the divisor is zero, instead of letting the hardware trap on the divide instruction. Objtool doesn't consider UD2 to be fatal unless it's annotated with unreachable(). So it considers the GCC-generated UD2 to be non-fatal, and it tries to follow the control flow past the UD2 and gets confused. Previously, objtool *did* assume UD2 was always a dead end. That changed with the following commit: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") The motivation behind that change was that Peter was planning on using UD2 for __WARN(), which is *not* a dead end. However, it turns out that some emulators rely on UD2 being fatal, so he ended up using 'ud0' instead: 9a93848fe787 ("x86/debug: Implement __WARN() using UD0") For GCC 4.5+, it should be safe to go back to the previous assumption that UD2 is fatal, even when it's not annotated with unreachable(). But for pre-4.5 versions of GCC, the unreachable() macro isn't supported, so such cases of UD2 need to be explicitly annotated as reachable. Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") Link: http://lkml.kernel.org/r/e57fa9dfede25f79487da8126ee9cdf7b856db65.1501188854.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-07-27 20:56:53 +00:00
/*
* By default, "ud2" is a dead end unless otherwise annotated, because
* GCC 7 inserts it for certain divide-by-zero cases.
*/
for_each_insn(file, insn)
if (insn->type == INSN_BUG)
insn->dead_end = true;
/*
* Check for manually annotated dead ends.
*/
sec = find_section_by_name(file->elf, ".rela.discard.unreachable");
if (!sec)
objtool: Assume unannotated UD2 instructions are dead ends Arnd reported some false positive warnings with GCC 7: drivers/hid/wacom_wac.o: warning: objtool: wacom_bpt3_touch()+0x2a5: stack state mismatch: cfa1=7+8 cfa2=6+16 drivers/iio/adc/vf610_adc.o: warning: objtool: vf610_adc_calculate_rates() falls through to next function vf610_adc_sample_set() drivers/pwm/pwm-hibvt.o: warning: objtool: hibvt_pwm_get_state() falls through to next function hibvt_pwm_remove() drivers/pwm/pwm-mediatek.o: warning: objtool: mtk_pwm_config() falls through to next function mtk_pwm_enable() drivers/spi/spi-bcm2835.o: warning: objtool: .text: unexpected end of section drivers/spi/spi-bcm2835aux.o: warning: objtool: .text: unexpected end of section drivers/watchdog/digicolor_wdt.o: warning: objtool: dc_wdt_get_timeleft() falls through to next function dc_wdt_restart() When GCC 7 detects a potential divide-by-zero condition, it sometimes inserts a UD2 instruction for the case where the divisor is zero, instead of letting the hardware trap on the divide instruction. Objtool doesn't consider UD2 to be fatal unless it's annotated with unreachable(). So it considers the GCC-generated UD2 to be non-fatal, and it tries to follow the control flow past the UD2 and gets confused. Previously, objtool *did* assume UD2 was always a dead end. That changed with the following commit: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") The motivation behind that change was that Peter was planning on using UD2 for __WARN(), which is *not* a dead end. However, it turns out that some emulators rely on UD2 being fatal, so he ended up using 'ud0' instead: 9a93848fe787 ("x86/debug: Implement __WARN() using UD0") For GCC 4.5+, it should be safe to go back to the previous assumption that UD2 is fatal, even when it's not annotated with unreachable(). But for pre-4.5 versions of GCC, the unreachable() macro isn't supported, so such cases of UD2 need to be explicitly annotated as reachable. Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") Link: http://lkml.kernel.org/r/e57fa9dfede25f79487da8126ee9cdf7b856db65.1501188854.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-07-27 20:56:53 +00:00
goto reachable;
list_for_each_entry(rela, &sec->rela_list, list) {
if (rela->sym->type != STT_SECTION) {
WARN("unexpected relocation symbol type in %s", sec->name);
return -1;
}
insn = find_insn(file, rela->sym->sec, rela->addend);
if (insn)
insn = list_prev_entry(insn, list);
else if (rela->addend == rela->sym->sec->len) {
found = false;
list_for_each_entry_reverse(insn, &file->insn_list, list) {
if (insn->sec == rela->sym->sec) {
found = true;
break;
}
}
if (!found) {
WARN("can't find unreachable insn at %s+0x%x",
rela->sym->sec->name, rela->addend);
return -1;
}
} else {
WARN("can't find unreachable insn at %s+0x%x",
rela->sym->sec->name, rela->addend);
return -1;
}
insn->dead_end = true;
}
objtool: Assume unannotated UD2 instructions are dead ends Arnd reported some false positive warnings with GCC 7: drivers/hid/wacom_wac.o: warning: objtool: wacom_bpt3_touch()+0x2a5: stack state mismatch: cfa1=7+8 cfa2=6+16 drivers/iio/adc/vf610_adc.o: warning: objtool: vf610_adc_calculate_rates() falls through to next function vf610_adc_sample_set() drivers/pwm/pwm-hibvt.o: warning: objtool: hibvt_pwm_get_state() falls through to next function hibvt_pwm_remove() drivers/pwm/pwm-mediatek.o: warning: objtool: mtk_pwm_config() falls through to next function mtk_pwm_enable() drivers/spi/spi-bcm2835.o: warning: objtool: .text: unexpected end of section drivers/spi/spi-bcm2835aux.o: warning: objtool: .text: unexpected end of section drivers/watchdog/digicolor_wdt.o: warning: objtool: dc_wdt_get_timeleft() falls through to next function dc_wdt_restart() When GCC 7 detects a potential divide-by-zero condition, it sometimes inserts a UD2 instruction for the case where the divisor is zero, instead of letting the hardware trap on the divide instruction. Objtool doesn't consider UD2 to be fatal unless it's annotated with unreachable(). So it considers the GCC-generated UD2 to be non-fatal, and it tries to follow the control flow past the UD2 and gets confused. Previously, objtool *did* assume UD2 was always a dead end. That changed with the following commit: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") The motivation behind that change was that Peter was planning on using UD2 for __WARN(), which is *not* a dead end. However, it turns out that some emulators rely on UD2 being fatal, so he ended up using 'ud0' instead: 9a93848fe787 ("x86/debug: Implement __WARN() using UD0") For GCC 4.5+, it should be safe to go back to the previous assumption that UD2 is fatal, even when it's not annotated with unreachable(). But for pre-4.5 versions of GCC, the unreachable() macro isn't supported, so such cases of UD2 need to be explicitly annotated as reachable. Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: d1091c7fa3d5 ("objtool: Improve detection of BUG() and other dead ends") Link: http://lkml.kernel.org/r/e57fa9dfede25f79487da8126ee9cdf7b856db65.1501188854.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-07-27 20:56:53 +00:00
reachable:
/*
* These manually annotated reachable checks are needed for GCC 4.4,
* where the Linux unreachable() macro isn't supported. In that case
* GCC doesn't know the "ud2" is fatal, so it generates code as if it's
* not a dead end.
*/
sec = find_section_by_name(file->elf, ".rela.discard.reachable");
if (!sec)
return 0;
list_for_each_entry(rela, &sec->rela_list, list) {
if (rela->sym->type != STT_SECTION) {
WARN("unexpected relocation symbol type in %s", sec->name);
return -1;
}
insn = find_insn(file, rela->sym->sec, rela->addend);
if (insn)
insn = list_prev_entry(insn, list);
else if (rela->addend == rela->sym->sec->len) {
found = false;
list_for_each_entry_reverse(insn, &file->insn_list, list) {
if (insn->sec == rela->sym->sec) {
found = true;
break;
}
}
if (!found) {
WARN("can't find reachable insn at %s+0x%x",
rela->sym->sec->name, rela->addend);
return -1;
}
} else {
WARN("can't find reachable insn at %s+0x%x",
rela->sym->sec->name, rela->addend);
return -1;
}
insn->dead_end = false;
}
return 0;
}
/*
* Warnings shouldn't be reported for ignored functions.
*/
static void add_ignores(struct objtool_file *file)
{
struct instruction *insn;
struct section *sec;
struct symbol *func;
struct rela *rela;
sec = find_section_by_name(file->elf, ".rela.discard.func_stack_frame_non_standard");
if (!sec)
return;
list_for_each_entry(rela, &sec->rela_list, list) {
switch (rela->sym->type) {
case STT_FUNC:
func = rela->sym;
break;
case STT_SECTION:
func = find_func_by_offset(rela->sym->sec, rela->addend);
if (!func)
continue;
break;
default:
WARN("unexpected relocation symbol type in %s: %d", sec->name, rela->sym->type);
continue;
}
func_for_each_insn(file, func, insn)
insn->ignore = true;
}
}
/*
* This is a whitelist of functions that is allowed to be called with AC set.
* The list is meant to be minimal and only contains compiler instrumentation
* ABI and a few functions used to implement *_{to,from}_user() functions.
*
* These functions must not directly change AC, but may PUSHF/POPF.
*/
static const char *uaccess_safe_builtin[] = {
/* KASAN */
"kasan_report",
"check_memory_region",
/* KASAN out-of-line */
"__asan_loadN_noabort",
"__asan_load1_noabort",
"__asan_load2_noabort",
"__asan_load4_noabort",
"__asan_load8_noabort",
"__asan_load16_noabort",
"__asan_storeN_noabort",
"__asan_store1_noabort",
"__asan_store2_noabort",
"__asan_store4_noabort",
"__asan_store8_noabort",
"__asan_store16_noabort",
/* KASAN in-line */
"__asan_report_load_n_noabort",
"__asan_report_load1_noabort",
"__asan_report_load2_noabort",
"__asan_report_load4_noabort",
"__asan_report_load8_noabort",
"__asan_report_load16_noabort",
"__asan_report_store_n_noabort",
"__asan_report_store1_noabort",
"__asan_report_store2_noabort",
"__asan_report_store4_noabort",
"__asan_report_store8_noabort",
"__asan_report_store16_noabort",
/* KCOV */
"write_comp_data",
"__sanitizer_cov_trace_pc",
"__sanitizer_cov_trace_const_cmp1",
"__sanitizer_cov_trace_const_cmp2",
"__sanitizer_cov_trace_const_cmp4",
"__sanitizer_cov_trace_const_cmp8",
"__sanitizer_cov_trace_cmp1",
"__sanitizer_cov_trace_cmp2",
"__sanitizer_cov_trace_cmp4",
"__sanitizer_cov_trace_cmp8",
/* UBSAN */
"ubsan_type_mismatch_common",
"__ubsan_handle_type_mismatch",
"__ubsan_handle_type_mismatch_v1",
"__ubsan_handle_shift_out_of_bounds",
/* misc */
"csum_partial_copy_generic",
"__memcpy_mcsafe",
"mcsafe_handle_tail",
"ftrace_likely_update", /* CONFIG_TRACE_BRANCH_PROFILING */
NULL
};
static void add_uaccess_safe(struct objtool_file *file)
{
struct symbol *func;
const char **name;
if (!uaccess)
return;
for (name = uaccess_safe_builtin; *name; name++) {
func = find_symbol_by_name(file->elf, *name);
if (!func)
continue;
func->uaccess_safe = true;
}
}
/*
* FIXME: For now, just ignore any alternatives which add retpolines. This is
* a temporary hack, as it doesn't allow ORC to unwind from inside a retpoline.
* But it at least allows objtool to understand the control flow *around* the
* retpoline.
*/
static int add_ignore_alternatives(struct objtool_file *file)
{
struct section *sec;
struct rela *rela;
struct instruction *insn;
sec = find_section_by_name(file->elf, ".rela.discard.ignore_alts");
if (!sec)
return 0;
list_for_each_entry(rela, &sec->rela_list, list) {
if (rela->sym->type != STT_SECTION) {
WARN("unexpected relocation symbol type in %s", sec->name);
return -1;
}
insn = find_insn(file, rela->sym->sec, rela->addend);
if (!insn) {
WARN("bad .discard.ignore_alts entry");
return -1;
}
insn->ignore_alts = true;
}
return 0;
}
/*
* Find the destination instructions for all jumps.
*/
static int add_jump_destinations(struct objtool_file *file)
{
struct instruction *insn;
struct rela *rela;
struct section *dest_sec;
unsigned long dest_off;
for_each_insn(file, insn) {
if (!is_static_jump(insn))
continue;
if (insn->ignore || insn->offset == FAKE_JUMP_OFFSET)
continue;
rela = find_rela_by_dest_range(file->elf, insn->sec,
insn->offset, insn->len);
if (!rela) {
dest_sec = insn->sec;
dest_off = insn->offset + insn->len + insn->immediate;
} else if (rela->sym->type == STT_SECTION) {
dest_sec = rela->sym->sec;
dest_off = rela->addend + 4;
} else if (rela->sym->sec->idx) {
dest_sec = rela->sym->sec;
dest_off = rela->sym->sym.st_value + rela->addend + 4;
} else if (strstr(rela->sym->name, "_indirect_thunk_")) {
/*
* Retpoline jumps are really dynamic jumps in
* disguise, so convert them accordingly.
*/
if (insn->type == INSN_JUMP_UNCONDITIONAL)
insn->type = INSN_JUMP_DYNAMIC;
else
insn->type = INSN_JUMP_DYNAMIC_CONDITIONAL;
insn->retpoline_safe = true;
continue;
} else {
/* external sibling call */
insn->call_dest = rela->sym;
continue;
}
insn->jump_dest = find_insn(file, dest_sec, dest_off);
if (!insn->jump_dest) {
/*
* This is a special case where an alt instruction
* jumps past the end of the section. These are
* handled later in handle_group_alt().
*/
if (!strcmp(insn->sec->name, ".altinstr_replacement"))
continue;
WARN_FUNC("can't find jump dest instruction at %s+0x%lx",
insn->sec, insn->offset, dest_sec->name,
dest_off);
return -1;
}
objtool: Fix GCC 8 cold subfunction detection for aliased functions The kbuild test robot reported the following issue: kernel/time/posix-stubs.o: warning: objtool: sys_ni_posix_timers.cold.1()+0x0: unreachable instruction This file creates symbol aliases for the sys_ni_posix_timers() function. So there are multiple ELF function symbols for the same function: 23: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __x64_sys_timer_create 24: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 sys_ni_posix_timers 25: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __ia32_sys_timer_create 26: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __x64_sys_timer_gettime Here's the corresponding cold subfunction: 11: 0000000000000000 45 FUNC LOCAL DEFAULT 6 sys_ni_posix_timers.cold.1 When analyzing overlapping functions, objtool only looks at the first one in the symbol list. The rest of the functions are basically ignored because they point to instructions which have already been analyzed. So in this case it analyzes the __x64_sys_timer_create() function, but then it fails to recognize that its cold subfunction is sys_ni_posix_timers.cold.1(), because the names are different. Make the subfunction detection a little smarter by associating each subfunction with the first function which jumps to it, since that's the one which will be analyzed. Unfortunately we still have to leave the original subfunction detection code in place, thanks to GCC switch tables. (See the comment for more details.) Fixes: 13810435b9a7 ("objtool: Support GCC 8's cold subfunctions") Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Link: https://lkml.kernel.org/r/d3ba52662cbc8e3a64a3b64d44b4efc5674fd9ab.1527855808.git.jpoimboe@redhat.com
2018-06-01 12:23:51 +00:00
/*
* Cross-function jump.
objtool: Fix GCC 8 cold subfunction detection for aliased functions The kbuild test robot reported the following issue: kernel/time/posix-stubs.o: warning: objtool: sys_ni_posix_timers.cold.1()+0x0: unreachable instruction This file creates symbol aliases for the sys_ni_posix_timers() function. So there are multiple ELF function symbols for the same function: 23: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __x64_sys_timer_create 24: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 sys_ni_posix_timers 25: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __ia32_sys_timer_create 26: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __x64_sys_timer_gettime Here's the corresponding cold subfunction: 11: 0000000000000000 45 FUNC LOCAL DEFAULT 6 sys_ni_posix_timers.cold.1 When analyzing overlapping functions, objtool only looks at the first one in the symbol list. The rest of the functions are basically ignored because they point to instructions which have already been analyzed. So in this case it analyzes the __x64_sys_timer_create() function, but then it fails to recognize that its cold subfunction is sys_ni_posix_timers.cold.1(), because the names are different. Make the subfunction detection a little smarter by associating each subfunction with the first function which jumps to it, since that's the one which will be analyzed. Unfortunately we still have to leave the original subfunction detection code in place, thanks to GCC switch tables. (See the comment for more details.) Fixes: 13810435b9a7 ("objtool: Support GCC 8's cold subfunctions") Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Link: https://lkml.kernel.org/r/d3ba52662cbc8e3a64a3b64d44b4efc5674fd9ab.1527855808.git.jpoimboe@redhat.com
2018-06-01 12:23:51 +00:00
*/
if (insn->func && insn->jump_dest->func &&
insn->func != insn->jump_dest->func) {
/*
* For GCC 8+, create parent/child links for any cold
* subfunctions. This is _mostly_ redundant with a
* similar initialization in read_symbols().
*
* If a function has aliases, we want the *first* such
* function in the symbol table to be the subfunction's
* parent. In that case we overwrite the
* initialization done in read_symbols().
*
* However this code can't completely replace the
* read_symbols() code because this doesn't detect the
* case where the parent function's only reference to a
* subfunction is through a jump table.
*/
if (!strstr(insn->func->name, ".cold.") &&
strstr(insn->jump_dest->func->name, ".cold.")) {
insn->func->cfunc = insn->jump_dest->func;
insn->jump_dest->func->pfunc = insn->func;
} else if (insn->jump_dest->func->pfunc != insn->func->pfunc &&
insn->jump_dest->offset == insn->jump_dest->func->offset) {
/* internal sibling call */
insn->call_dest = insn->jump_dest->func;
}
objtool: Fix GCC 8 cold subfunction detection for aliased functions The kbuild test robot reported the following issue: kernel/time/posix-stubs.o: warning: objtool: sys_ni_posix_timers.cold.1()+0x0: unreachable instruction This file creates symbol aliases for the sys_ni_posix_timers() function. So there are multiple ELF function symbols for the same function: 23: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __x64_sys_timer_create 24: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 sys_ni_posix_timers 25: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __ia32_sys_timer_create 26: 0000000000000150 26 FUNC GLOBAL DEFAULT 1 __x64_sys_timer_gettime Here's the corresponding cold subfunction: 11: 0000000000000000 45 FUNC LOCAL DEFAULT 6 sys_ni_posix_timers.cold.1 When analyzing overlapping functions, objtool only looks at the first one in the symbol list. The rest of the functions are basically ignored because they point to instructions which have already been analyzed. So in this case it analyzes the __x64_sys_timer_create() function, but then it fails to recognize that its cold subfunction is sys_ni_posix_timers.cold.1(), because the names are different. Make the subfunction detection a little smarter by associating each subfunction with the first function which jumps to it, since that's the one which will be analyzed. Unfortunately we still have to leave the original subfunction detection code in place, thanks to GCC switch tables. (See the comment for more details.) Fixes: 13810435b9a7 ("objtool: Support GCC 8's cold subfunctions") Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Link: https://lkml.kernel.org/r/d3ba52662cbc8e3a64a3b64d44b4efc5674fd9ab.1527855808.git.jpoimboe@redhat.com
2018-06-01 12:23:51 +00:00
}
}
return 0;
}
/*
* Find the destination instructions for all calls.
*/
static int add_call_destinations(struct objtool_file *file)
{
struct instruction *insn;
unsigned long dest_off;
struct rela *rela;
for_each_insn(file, insn) {
if (insn->type != INSN_CALL)
continue;
rela = find_rela_by_dest_range(file->elf, insn->sec,
insn->offset, insn->len);
if (!rela) {
dest_off = insn->offset + insn->len + insn->immediate;
insn->call_dest = find_func_by_offset(insn->sec, dest_off);
if (!insn->call_dest)
insn->call_dest = find_symbol_by_offset(insn->sec, dest_off);
if (insn->ignore)
continue;
if (!insn->call_dest) {
WARN_FUNC("unsupported intra-function call",
insn->sec, insn->offset);
if (retpoline)
WARN("If this is a retpoline, please patch it in with alternatives and annotate it with ANNOTATE_NOSPEC_ALTERNATIVE.");
return -1;
}
if (insn->func && insn->call_dest->type != STT_FUNC) {
WARN_FUNC("unsupported call to non-function",
insn->sec, insn->offset);
return -1;
}
} else if (rela->sym->type == STT_SECTION) {
insn->call_dest = find_func_by_offset(rela->sym->sec,
rela->addend+4);
if (!insn->call_dest) {
WARN_FUNC("can't find call dest symbol at %s+0x%x",
insn->sec, insn->offset,
rela->sym->sec->name,
rela->addend + 4);
return -1;
}
} else
insn->call_dest = rela->sym;
}
return 0;
}
/*
* The .alternatives section requires some extra special care, over and above
* what other special sections require:
*
* 1. Because alternatives are patched in-place, we need to insert a fake jump
* instruction at the end so that validate_branch() skips all the original
* replaced instructions when validating the new instruction path.
*
* 2. An added wrinkle is that the new instruction length might be zero. In
* that case the old instructions are replaced with noops. We simulate that
* by creating a fake jump as the only new instruction.
*
* 3. In some cases, the alternative section includes an instruction which
* conditionally jumps to the _end_ of the entry. We have to modify these
* jumps' destinations to point back to .text rather than the end of the
* entry in .altinstr_replacement.
*/
static int handle_group_alt(struct objtool_file *file,
struct special_alt *special_alt,
struct instruction *orig_insn,
struct instruction **new_insn)
{
struct instruction *last_orig_insn, *last_new_insn, *insn, *fake_jump = NULL;
unsigned long dest_off;
last_orig_insn = NULL;
insn = orig_insn;
sec_for_each_insn_from(file, insn) {
if (insn->offset >= special_alt->orig_off + special_alt->orig_len)
break;
insn->alt_group = true;
last_orig_insn = insn;
}
if (next_insn_same_sec(file, last_orig_insn)) {
fake_jump = malloc(sizeof(*fake_jump));
if (!fake_jump) {
WARN("malloc failed");
return -1;
}
memset(fake_jump, 0, sizeof(*fake_jump));
INIT_LIST_HEAD(&fake_jump->alts);
clear_insn_state(&fake_jump->state);
fake_jump->sec = special_alt->new_sec;
fake_jump->offset = FAKE_JUMP_OFFSET;
fake_jump->type = INSN_JUMP_UNCONDITIONAL;
fake_jump->jump_dest = list_next_entry(last_orig_insn, list);
fake_jump->func = orig_insn->func;
}
if (!special_alt->new_len) {
if (!fake_jump) {
WARN("%s: empty alternative at end of section",
special_alt->orig_sec->name);
return -1;
}
*new_insn = fake_jump;
return 0;
}
last_new_insn = NULL;
insn = *new_insn;
sec_for_each_insn_from(file, insn) {
if (insn->offset >= special_alt->new_off + special_alt->new_len)
break;
last_new_insn = insn;
insn->ignore = orig_insn->ignore_alts;
insn->func = orig_insn->func;
/*
* Since alternative replacement code is copy/pasted by the
* kernel after applying relocations, generally such code can't
* have relative-address relocation references to outside the
* .altinstr_replacement section, unless the arch's
* alternatives code can adjust the relative offsets
* accordingly.
*
* The x86 alternatives code adjusts the offsets only when it
* encounters a branch instruction at the very beginning of the
* replacement group.
*/
if ((insn->offset != special_alt->new_off ||
(insn->type != INSN_CALL && !is_static_jump(insn))) &&
find_rela_by_dest_range(file->elf, insn->sec, insn->offset, insn->len)) {
WARN_FUNC("unsupported relocation in alternatives section",
insn->sec, insn->offset);
return -1;
}
if (!is_static_jump(insn))
continue;
if (!insn->immediate)
continue;
dest_off = insn->offset + insn->len + insn->immediate;
if (dest_off == special_alt->new_off + special_alt->new_len) {
if (!fake_jump) {
WARN("%s: alternative jump to end of section",
special_alt->orig_sec->name);
return -1;
}
insn->jump_dest = fake_jump;
}
if (!insn->jump_dest) {
WARN_FUNC("can't find alternative jump destination",
insn->sec, insn->offset);
return -1;
}
}
if (!last_new_insn) {
WARN_FUNC("can't find last new alternative instruction",
special_alt->new_sec, special_alt->new_off);
return -1;
}
if (fake_jump)
list_add(&fake_jump->list, &last_new_insn->list);
return 0;
}
/*
* A jump table entry can either convert a nop to a jump or a jump to a nop.
* If the original instruction is a jump, make the alt entry an effective nop
* by just skipping the original instruction.
*/
static int handle_jump_alt(struct objtool_file *file,
struct special_alt *special_alt,
struct instruction *orig_insn,
struct instruction **new_insn)
{
if (orig_insn->type == INSN_NOP)
return 0;
if (orig_insn->type != INSN_JUMP_UNCONDITIONAL) {
WARN_FUNC("unsupported instruction at jump label",
orig_insn->sec, orig_insn->offset);
return -1;
}
*new_insn = list_next_entry(orig_insn, list);
return 0;
}
/*
* Read all the special sections which have alternate instructions which can be
* patched in or redirected to at runtime. Each instruction having alternate
* instruction(s) has them added to its insn->alts list, which will be
* traversed in validate_branch().
*/
static int add_special_section_alts(struct objtool_file *file)
{
struct list_head special_alts;
struct instruction *orig_insn, *new_insn;
struct special_alt *special_alt, *tmp;
struct alternative *alt;
int ret;
ret = special_get_alts(file->elf, &special_alts);
if (ret)
return ret;
list_for_each_entry_safe(special_alt, tmp, &special_alts, list) {
orig_insn = find_insn(file, special_alt->orig_sec,
special_alt->orig_off);
if (!orig_insn) {
WARN_FUNC("special: can't find orig instruction",
special_alt->orig_sec, special_alt->orig_off);
ret = -1;
goto out;
}
new_insn = NULL;
if (!special_alt->group || special_alt->new_len) {
new_insn = find_insn(file, special_alt->new_sec,
special_alt->new_off);
if (!new_insn) {
WARN_FUNC("special: can't find new instruction",
special_alt->new_sec,
special_alt->new_off);
ret = -1;
goto out;
}
}
if (special_alt->group) {
ret = handle_group_alt(file, special_alt, orig_insn,
&new_insn);
if (ret)
goto out;
} else if (special_alt->jump_or_nop) {
ret = handle_jump_alt(file, special_alt, orig_insn,
&new_insn);
if (ret)
goto out;
}
alt = malloc(sizeof(*alt));
if (!alt) {
WARN("malloc failed");
ret = -1;
goto out;
}
alt->insn = new_insn;
alt->skip_orig = special_alt->skip_orig;
orig_insn->ignore_alts |= special_alt->skip_alt;
list_add_tail(&alt->list, &orig_insn->alts);
list_del(&special_alt->list);
free(special_alt);
}
out:
return ret;
}
static int add_jump_table(struct objtool_file *file, struct instruction *insn,
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
struct rela *table)
{
struct rela *rela = table;
struct instruction *dest_insn;
struct alternative *alt;
objtool: Support GCC 8 switch tables With GCC 8, some issues were found with the objtool switch table detection. 1) In the .rodata section, immediately after the switch table, there can be another object which contains a pointer to the function which had the switch statement. In this case objtool wrongly considers the function pointer to be part of the switch table. Fix it by: a) making sure there are no pointers to the beginning of the function; and b) making sure there are no gaps in the switch table. Only the former was needed, the latter adds additional protection for future optimizations. 2) In find_switch_table(), case 1 and case 2 are missing the check to ensure that the .rodata switch table data is anonymous, i.e. that it isn't already associated with an ELF symbol. Fix it by adding the same find_symbol_containing() check which is used for case 3. This fixes the following warnings with GCC 8: drivers/block/virtio_blk.o: warning: objtool: virtio_queue_rq()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+72 net/ipv6/icmp.o: warning: objtool: icmpv6_rcv()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 drivers/usb/core/quirks.o: warning: objtool: quirks_param_set()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+48 drivers/mtd/nand/raw/nand_hynix.o: warning: objtool: hynix_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+24 drivers/mtd/nand/raw/nand_samsung.o: warning: objtool: samsung_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+32 drivers/gpu/drm/nouveau/nvkm/subdev/top/gk104.o: warning: objtool: gk104_top_oneinit()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: David Laight <David.Laight@ACULAB.COM> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: damian <damian.tometzki@icloud.com> Link: http://lkml.kernel.org/r/20180510224849.xwi34d6tzheb5wgw@treble Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-10 22:48:49 +00:00
struct symbol *pfunc = insn->func->pfunc;
unsigned int prev_offset = 0;
/*
* Each @rela is a switch table relocation which points to the target
* instruction.
*/
list_for_each_entry_from(rela, &table->sec->rela_list, list) {
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
/* Check for the end of the table: */
if (rela != table && rela->jump_table_start)
break;
/* Make sure the table entries are consecutive: */
objtool: Support GCC 8 switch tables With GCC 8, some issues were found with the objtool switch table detection. 1) In the .rodata section, immediately after the switch table, there can be another object which contains a pointer to the function which had the switch statement. In this case objtool wrongly considers the function pointer to be part of the switch table. Fix it by: a) making sure there are no pointers to the beginning of the function; and b) making sure there are no gaps in the switch table. Only the former was needed, the latter adds additional protection for future optimizations. 2) In find_switch_table(), case 1 and case 2 are missing the check to ensure that the .rodata switch table data is anonymous, i.e. that it isn't already associated with an ELF symbol. Fix it by adding the same find_symbol_containing() check which is used for case 3. This fixes the following warnings with GCC 8: drivers/block/virtio_blk.o: warning: objtool: virtio_queue_rq()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+72 net/ipv6/icmp.o: warning: objtool: icmpv6_rcv()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 drivers/usb/core/quirks.o: warning: objtool: quirks_param_set()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+48 drivers/mtd/nand/raw/nand_hynix.o: warning: objtool: hynix_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+24 drivers/mtd/nand/raw/nand_samsung.o: warning: objtool: samsung_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+32 drivers/gpu/drm/nouveau/nvkm/subdev/top/gk104.o: warning: objtool: gk104_top_oneinit()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: David Laight <David.Laight@ACULAB.COM> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: damian <damian.tometzki@icloud.com> Link: http://lkml.kernel.org/r/20180510224849.xwi34d6tzheb5wgw@treble Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-10 22:48:49 +00:00
if (prev_offset && rela->offset != prev_offset + 8)
break;
/* Detect function pointers from contiguous objects: */
if (rela->sym->sec == pfunc->sec &&
rela->addend == pfunc->offset)
break;
dest_insn = find_insn(file, rela->sym->sec, rela->addend);
if (!dest_insn)
break;
/* Make sure the destination is in the same function: */
objtool: Fix seg fault on bad switch table entry In one rare case, Clang generated the following code: 5ca: 83 e0 21 and $0x21,%eax 5cd: b9 04 00 00 00 mov $0x4,%ecx 5d2: ff 24 c5 00 00 00 00 jmpq *0x0(,%rax,8) 5d5: R_X86_64_32S .rodata+0x38 which uses the corresponding jump table relocations: 000000000038 000200000001 R_X86_64_64 0000000000000000 .text + 834 000000000040 000200000001 R_X86_64_64 0000000000000000 .text + 5d9 000000000048 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000050 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000058 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000060 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000068 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000070 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000078 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000080 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000088 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000090 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000098 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000a0 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000a8 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000b0 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000b8 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000c0 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000c8 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000d0 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000d8 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000e0 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000e8 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000f0 000200000001 R_X86_64_64 0000000000000000 .text + b96 0000000000f8 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000100 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000108 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000110 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000118 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000120 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000128 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000130 000200000001 R_X86_64_64 0000000000000000 .text + b96 000000000138 000200000001 R_X86_64_64 0000000000000000 .text + 82f 000000000140 000200000001 R_X86_64_64 0000000000000000 .text + 828 Since %eax was masked with 0x21, only the first two and the last two entries are possible. Objtool doesn't actually emulate all the code, so it isn't smart enough to know that all the middle entries aren't reachable. They point to the NOP padding area after the end of the function, so objtool seg faulted when it tried to dereference a NULL insn->func. After this fix, objtool still gives an "unreachable" error because it stops reading the jump table when it encounters the bad addresses: /home/jpoimboe/objtool-tests/adm1275.o: warning: objtool: adm1275_probe()+0x828: unreachable instruction While the above code is technically correct, it's very wasteful of memory -- it uses 34 jump table entries when only 4 are needed. It's also not possible for objtool to validate this type of switch table because the unused entries point outside the function and objtool has no way of determining if that's intentional. Hopefully the Clang folks can fix it. Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/a9db88eec4f1ca089e040989846961748238b6d8.1563413318.git.jpoimboe@redhat.com
2019-07-18 01:36:55 +00:00
if (!dest_insn->func || dest_insn->func->pfunc != pfunc)
break;
alt = malloc(sizeof(*alt));
if (!alt) {
WARN("malloc failed");
return -1;
}
alt->insn = dest_insn;
list_add_tail(&alt->list, &insn->alts);
objtool: Support GCC 8 switch tables With GCC 8, some issues were found with the objtool switch table detection. 1) In the .rodata section, immediately after the switch table, there can be another object which contains a pointer to the function which had the switch statement. In this case objtool wrongly considers the function pointer to be part of the switch table. Fix it by: a) making sure there are no pointers to the beginning of the function; and b) making sure there are no gaps in the switch table. Only the former was needed, the latter adds additional protection for future optimizations. 2) In find_switch_table(), case 1 and case 2 are missing the check to ensure that the .rodata switch table data is anonymous, i.e. that it isn't already associated with an ELF symbol. Fix it by adding the same find_symbol_containing() check which is used for case 3. This fixes the following warnings with GCC 8: drivers/block/virtio_blk.o: warning: objtool: virtio_queue_rq()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+72 net/ipv6/icmp.o: warning: objtool: icmpv6_rcv()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 drivers/usb/core/quirks.o: warning: objtool: quirks_param_set()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+48 drivers/mtd/nand/raw/nand_hynix.o: warning: objtool: hynix_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+24 drivers/mtd/nand/raw/nand_samsung.o: warning: objtool: samsung_nand_decode_id()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+32 drivers/gpu/drm/nouveau/nvkm/subdev/top/gk104.o: warning: objtool: gk104_top_oneinit()+0x0: stack state mismatch: cfa1=7+8 cfa2=7+64 Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: David Laight <David.Laight@ACULAB.COM> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: damian <damian.tometzki@icloud.com> Link: http://lkml.kernel.org/r/20180510224849.xwi34d6tzheb5wgw@treble Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-10 22:48:49 +00:00
prev_offset = rela->offset;
}
if (!prev_offset) {
WARN_FUNC("can't find switch jump table",
insn->sec, insn->offset);
return -1;
}
return 0;
}
/*
* find_jump_table() - Given a dynamic jump, find the switch jump table in
* .rodata associated with it.
*
* There are 3 basic patterns:
*
* 1. jmpq *[rodata addr](,%reg,8)
*
* This is the most common case by far. It jumps to an address in a simple
* jump table which is stored in .rodata.
*
* 2. jmpq *[rodata addr](%rip)
*
* This is caused by a rare GCC quirk, currently only seen in three driver
* functions in the kernel, only with certain obscure non-distro configs.
*
* As part of an optimization, GCC makes a copy of an existing switch jump
* table, modifies it, and then hard-codes the jump (albeit with an indirect
* jump) to use a single entry in the table. The rest of the jump table and
* some of its jump targets remain as dead code.
*
* In such a case we can just crudely ignore all unreachable instruction
* warnings for the entire object file. Ideally we would just ignore them
* for the function, but that would require redesigning the code quite a
* bit. And honestly that's just not worth doing: unreachable instruction
* warnings are of questionable value anyway, and this is such a rare issue.
*
* 3. mov [rodata addr],%reg1
* ... some instructions ...
* jmpq *(%reg1,%reg2,8)
*
* This is a fairly uncommon pattern which is new for GCC 6. As of this
* writing, there are 11 occurrences of it in the allmodconfig kernel.
*
* As of GCC 7 there are quite a few more of these and the 'in between' code
* is significant. Esp. with KASAN enabled some of the code between the mov
* and jmpq uses .rodata itself, which can confuse things.
*
* TODO: Once we have DWARF CFI and smarter instruction decoding logic,
* ensure the same register is used in the mov and jump instructions.
*
* NOTE: RETPOLINE made it harder still to decode dynamic jumps.
*/
static struct rela *find_jump_table(struct objtool_file *file,
struct symbol *func,
struct instruction *insn)
{
struct rela *text_rela, *table_rela;
struct instruction *dest_insn, *orig_insn = insn;
struct section *table_sec;
objtool: Detect RIP-relative switch table references Typically a switch table can be found by detecting a .rodata access followed an indirect jump: 1969: 4a 8b 0c e5 00 00 00 mov 0x0(,%r12,8),%rcx 1970: 00 196d: R_X86_64_32S .rodata+0x438 1971: e9 00 00 00 00 jmpq 1976 <dispc_runtime_suspend+0xb6a> 1972: R_X86_64_PC32 __x86_indirect_thunk_rcx-0x4 Randy Dunlap reported a case (seen with GCC 4.8) where the .rodata access uses RIP-relative addressing: 19bd: 48 8b 3d 00 00 00 00 mov 0x0(%rip),%rdi # 19c4 <dispc_runtime_suspend+0xbb8> 19c0: R_X86_64_PC32 .rodata+0x45c 19c4: e9 00 00 00 00 jmpq 19c9 <dispc_runtime_suspend+0xbbd> 19c5: R_X86_64_PC32 __x86_indirect_thunk_rdi-0x4 In this case the relocation addend needs to be adjusted accordingly in order to find the location of the switch table. The fix is for case 3 (as described in the comments), but also make the existing case 1 & 2 checks more precise by only adjusting the addend for R_X86_64_PC32 relocations. This fixes the following warnings: drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_suspend()+0xbb8: sibling call from callable instruction with modified stack frame drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_resume()+0xcc5: sibling call from callable instruction with modified stack frame Reported-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/b6098294fd67afb69af8c47c9883d7a68bf0f8ea.1526305958.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-14 13:53:24 +00:00
unsigned long table_offset;
/*
* Backward search using the @first_jump_src links, these help avoid
* much of the 'in between' code. Which avoids us getting confused by
* it.
*/
objtool: Detect RIP-relative switch table references, part 2 With the following commit: fd35c88b7417 ("objtool: Support GCC 8 switch tables") I added a "can't find switch jump table" warning, to stop covering up silent failures if add_switch_table() can't find anything. That warning found yet another bug in the objtool switch table detection logic. For cases 1 and 2 (as described in the comments of find_switch_table()), the find_symbol_containing() check doesn't adjust the offset for RIP-relative switch jumps. Incidentally, this bug was already fixed for case 3 with: 6f5ec2993b1f ("objtool: Detect RIP-relative switch table references") However, that commit missed the fix for cases 1 and 2. The different cases are now starting to look more and more alike. So fix the bug by consolidating them into a single case, by checking the original dynamic jump instruction in the case 3 loop. This also simplifies the code and makes it more robust against future switch table detection issues -- of which I'm sure there will be many... Switch table detection has been the most fragile area of objtool, by far. I long for the day when we'll have a GCC plugin for annotating switch tables. Linus asked me to delay such a plugin due to the flakiness of the plugin infrastructure in older versions of GCC, so this rickety code is what we're stuck with for now. At least the code is now a little simpler than it was. Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/f400541613d45689086329432f3095119ffbc328.1526674218.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-18 20:10:34 +00:00
for (;
&insn->list != &file->insn_list &&
insn->sec == func->sec &&
insn->offset >= func->offset;
insn = insn->first_jump_src ?: list_prev_entry(insn, list)) {
objtool: Detect RIP-relative switch table references, part 2 With the following commit: fd35c88b7417 ("objtool: Support GCC 8 switch tables") I added a "can't find switch jump table" warning, to stop covering up silent failures if add_switch_table() can't find anything. That warning found yet another bug in the objtool switch table detection logic. For cases 1 and 2 (as described in the comments of find_switch_table()), the find_symbol_containing() check doesn't adjust the offset for RIP-relative switch jumps. Incidentally, this bug was already fixed for case 3 with: 6f5ec2993b1f ("objtool: Detect RIP-relative switch table references") However, that commit missed the fix for cases 1 and 2. The different cases are now starting to look more and more alike. So fix the bug by consolidating them into a single case, by checking the original dynamic jump instruction in the case 3 loop. This also simplifies the code and makes it more robust against future switch table detection issues -- of which I'm sure there will be many... Switch table detection has been the most fragile area of objtool, by far. I long for the day when we'll have a GCC plugin for annotating switch tables. Linus asked me to delay such a plugin due to the flakiness of the plugin infrastructure in older versions of GCC, so this rickety code is what we're stuck with for now. At least the code is now a little simpler than it was. Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/f400541613d45689086329432f3095119ffbc328.1526674218.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-18 20:10:34 +00:00
if (insn != orig_insn && insn->type == INSN_JUMP_DYNAMIC)
break;
/* allow small jumps within the range */
if (insn->type == INSN_JUMP_UNCONDITIONAL &&
insn->jump_dest &&
(insn->jump_dest->offset <= insn->offset ||
insn->jump_dest->offset > orig_insn->offset))
break;
/* look for a relocation which references .rodata */
text_rela = find_rela_by_dest_range(file->elf, insn->sec,
insn->offset, insn->len);
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
if (!text_rela || text_rela->sym->type != STT_SECTION ||
!text_rela->sym->sec->rodata)
continue;
objtool: Detect RIP-relative switch table references Typically a switch table can be found by detecting a .rodata access followed an indirect jump: 1969: 4a 8b 0c e5 00 00 00 mov 0x0(,%r12,8),%rcx 1970: 00 196d: R_X86_64_32S .rodata+0x438 1971: e9 00 00 00 00 jmpq 1976 <dispc_runtime_suspend+0xb6a> 1972: R_X86_64_PC32 __x86_indirect_thunk_rcx-0x4 Randy Dunlap reported a case (seen with GCC 4.8) where the .rodata access uses RIP-relative addressing: 19bd: 48 8b 3d 00 00 00 00 mov 0x0(%rip),%rdi # 19c4 <dispc_runtime_suspend+0xbb8> 19c0: R_X86_64_PC32 .rodata+0x45c 19c4: e9 00 00 00 00 jmpq 19c9 <dispc_runtime_suspend+0xbbd> 19c5: R_X86_64_PC32 __x86_indirect_thunk_rdi-0x4 In this case the relocation addend needs to be adjusted accordingly in order to find the location of the switch table. The fix is for case 3 (as described in the comments), but also make the existing case 1 & 2 checks more precise by only adjusting the addend for R_X86_64_PC32 relocations. This fixes the following warnings: drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_suspend()+0xbb8: sibling call from callable instruction with modified stack frame drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_resume()+0xcc5: sibling call from callable instruction with modified stack frame Reported-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/b6098294fd67afb69af8c47c9883d7a68bf0f8ea.1526305958.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-14 13:53:24 +00:00
table_offset = text_rela->addend;
table_sec = text_rela->sym->sec;
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
objtool: Detect RIP-relative switch table references Typically a switch table can be found by detecting a .rodata access followed an indirect jump: 1969: 4a 8b 0c e5 00 00 00 mov 0x0(,%r12,8),%rcx 1970: 00 196d: R_X86_64_32S .rodata+0x438 1971: e9 00 00 00 00 jmpq 1976 <dispc_runtime_suspend+0xb6a> 1972: R_X86_64_PC32 __x86_indirect_thunk_rcx-0x4 Randy Dunlap reported a case (seen with GCC 4.8) where the .rodata access uses RIP-relative addressing: 19bd: 48 8b 3d 00 00 00 00 mov 0x0(%rip),%rdi # 19c4 <dispc_runtime_suspend+0xbb8> 19c0: R_X86_64_PC32 .rodata+0x45c 19c4: e9 00 00 00 00 jmpq 19c9 <dispc_runtime_suspend+0xbbd> 19c5: R_X86_64_PC32 __x86_indirect_thunk_rdi-0x4 In this case the relocation addend needs to be adjusted accordingly in order to find the location of the switch table. The fix is for case 3 (as described in the comments), but also make the existing case 1 & 2 checks more precise by only adjusting the addend for R_X86_64_PC32 relocations. This fixes the following warnings: drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_suspend()+0xbb8: sibling call from callable instruction with modified stack frame drivers/video/fbdev/omap2/omapfb/dss/dispc.o: warning: objtool: dispc_runtime_resume()+0xcc5: sibling call from callable instruction with modified stack frame Reported-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/b6098294fd67afb69af8c47c9883d7a68bf0f8ea.1526305958.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-14 13:53:24 +00:00
if (text_rela->type == R_X86_64_PC32)
table_offset += 4;
/*
* Make sure the .rodata address isn't associated with a
* symbol. GCC jump tables are anonymous data.
*
* Also support C jump tables which are in the same format as
* switch jump tables. For objtool to recognize them, they
* need to be placed in the C_JUMP_TABLE_SECTION section. They
* have symbols associated with them.
*/
if (find_symbol_containing(table_sec, table_offset) &&
strcmp(table_sec->name, C_JUMP_TABLE_SECTION))
continue;
/*
* Each table entry has a rela associated with it. The rela
* should reference text in the same function as the original
* instruction.
*/
table_rela = find_rela_by_dest(file->elf, table_sec, table_offset);
if (!table_rela)
continue;
dest_insn = find_insn(file, table_rela->sym->sec, table_rela->addend);
if (!dest_insn || !dest_insn->func || dest_insn->func->pfunc != func)
continue;
objtool: Detect RIP-relative switch table references, part 2 With the following commit: fd35c88b7417 ("objtool: Support GCC 8 switch tables") I added a "can't find switch jump table" warning, to stop covering up silent failures if add_switch_table() can't find anything. That warning found yet another bug in the objtool switch table detection logic. For cases 1 and 2 (as described in the comments of find_switch_table()), the find_symbol_containing() check doesn't adjust the offset for RIP-relative switch jumps. Incidentally, this bug was already fixed for case 3 with: 6f5ec2993b1f ("objtool: Detect RIP-relative switch table references") However, that commit missed the fix for cases 1 and 2. The different cases are now starting to look more and more alike. So fix the bug by consolidating them into a single case, by checking the original dynamic jump instruction in the case 3 loop. This also simplifies the code and makes it more robust against future switch table detection issues -- of which I'm sure there will be many... Switch table detection has been the most fragile area of objtool, by far. I long for the day when we'll have a GCC plugin for annotating switch tables. Linus asked me to delay such a plugin due to the flakiness of the plugin infrastructure in older versions of GCC, so this rickety code is what we're stuck with for now. At least the code is now a little simpler than it was. Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/f400541613d45689086329432f3095119ffbc328.1526674218.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-05-18 20:10:34 +00:00
/*
* Use of RIP-relative switch jumps is quite rare, and
* indicates a rare GCC quirk/bug which can leave dead code
* behind.
*/
if (text_rela->type == R_X86_64_PC32)
file->ignore_unreachables = true;
return table_rela;
}
return NULL;
}
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
/*
* First pass: Mark the head of each jump table so that in the next pass,
* we know when a given jump table ends and the next one starts.
*/
static void mark_func_jump_tables(struct objtool_file *file,
struct symbol *func)
{
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
struct instruction *insn, *last = NULL;
struct rela *rela;
func_for_each_insn(file, func, insn) {
if (!last)
last = insn;
/*
* Store back-pointers for unconditional forward jumps such
* that find_jump_table() can back-track using those and
* avoid some potentially confusing code.
*/
if (insn->type == INSN_JUMP_UNCONDITIONAL && insn->jump_dest &&
insn->offset > last->offset &&
insn->jump_dest->offset > insn->offset &&
!insn->jump_dest->first_jump_src) {
insn->jump_dest->first_jump_src = insn;
last = insn->jump_dest;
}
if (insn->type != INSN_JUMP_DYNAMIC)
continue;
rela = find_jump_table(file, func, insn);
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
if (rela) {
rela->jump_table_start = true;
insn->jump_table = rela;
}
}
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
}
static int add_func_jump_tables(struct objtool_file *file,
struct symbol *func)
{
struct instruction *insn;
int ret;
func_for_each_insn(file, func, insn) {
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
if (!insn->jump_table)
continue;
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
ret = add_jump_table(file, insn, insn->jump_table);
if (ret)
return ret;
}
return 0;
}
/*
* For some switch statements, gcc generates a jump table in the .rodata
* section which contains a list of addresses within the function to jump to.
* This finds these jump tables and adds them to the insn->alts lists.
*/
static int add_jump_table_alts(struct objtool_file *file)
{
struct section *sec;
struct symbol *func;
int ret;
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
if (!file->rodata)
return 0;
for_each_sec(file, sec) {
list_for_each_entry(func, &sec->symbol_list, list) {
if (func->type != STT_FUNC)
continue;
objtool: Support repeated uses of the same C jump table This fixes objtool for both a GCC issue and a Clang issue: 1) GCC issue: kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0x8d5: sibling call from callable instruction with modified stack frame With CONFIG_RETPOLINE=n, GCC is doing the following optimization in ___bpf_prog_run(). Before: select_insn: jmp *jumptable(,%rax,8) ... ALU64_ADD_X: ... jmp select_insn ALU_ADD_X: ... jmp select_insn After: select_insn: jmp *jumptable(, %rax, 8) ... ALU64_ADD_X: ... jmp *jumptable(, %rax, 8) ALU_ADD_X: ... jmp *jumptable(, %rax, 8) This confuses objtool. It has never seen multiple indirect jump sites which use the same jump table. For GCC switch tables, the only way of detecting the size of a table is by continuing to scan for more tables. The size of the previous table can only be determined after another switch table is found, or when the scan reaches the end of the function. That logic was reused for C jump tables, and was based on the assumption that each jump table only has a single jump site. The above optimization breaks that assumption. 2) Clang issue: drivers/usb/misc/sisusbvga/sisusb.o: warning: objtool: sisusb_write_mem_bulk()+0x588: can't find switch jump table With clang 9, code can be generated where a function contains two indirect jump instructions which use the same switch table. The fix is the same for both issues: split the jump table parsing into two passes. In the first pass, locate the heads of all switch tables for the function and mark their locations. In the second pass, parse the switch tables and add them. Fixes: e55a73251da3 ("bpf: Fix ORC unwinding in non-JIT BPF code") Reported-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/e995befaada9d4d8b2cf788ff3f566ba900d2b4d.1563413318.git.jpoimboe@redhat.com Co-developed-by: Josh Poimboeuf <jpoimboe@redhat.com>
2019-07-18 01:36:54 +00:00
mark_func_jump_tables(file, func);
ret = add_func_jump_tables(file, func);
if (ret)
return ret;
}
}
return 0;
}
static int read_unwind_hints(struct objtool_file *file)
{
struct section *sec, *relasec;
struct rela *rela;
struct unwind_hint *hint;
struct instruction *insn;
struct cfi_reg *cfa;
int i;
sec = find_section_by_name(file->elf, ".discard.unwind_hints");
if (!sec)
return 0;
relasec = sec->rela;
if (!relasec) {
WARN("missing .rela.discard.unwind_hints section");
return -1;
}
if (sec->len % sizeof(struct unwind_hint)) {
WARN("struct unwind_hint size mismatch");
return -1;
}
file->hints = true;
for (i = 0; i < sec->len / sizeof(struct unwind_hint); i++) {
hint = (struct unwind_hint *)sec->data->d_buf + i;
rela = find_rela_by_dest(file->elf, sec, i * sizeof(*hint));
if (!rela) {
WARN("can't find rela for unwind_hints[%d]", i);
return -1;
}
insn = find_insn(file, rela->sym->sec, rela->addend);
if (!insn) {
WARN("can't find insn for unwind_hints[%d]", i);
return -1;
}
cfa = &insn->state.cfa;
if (hint->type == UNWIND_HINT_TYPE_SAVE) {
insn->save = true;
continue;
} else if (hint->type == UNWIND_HINT_TYPE_RESTORE) {
insn->restore = true;
insn->hint = true;
continue;
}
insn->hint = true;
switch (hint->sp_reg) {
case ORC_REG_UNDEFINED:
cfa->base = CFI_UNDEFINED;
break;
case ORC_REG_SP:
cfa->base = CFI_SP;
break;
case ORC_REG_BP:
cfa->base = CFI_BP;
break;
case ORC_REG_SP_INDIRECT:
cfa->base = CFI_SP_INDIRECT;
break;
case ORC_REG_R10:
cfa->base = CFI_R10;
break;
case ORC_REG_R13:
cfa->base = CFI_R13;
break;
case ORC_REG_DI:
cfa->base = CFI_DI;
break;
case ORC_REG_DX:
cfa->base = CFI_DX;
break;
default:
WARN_FUNC("unsupported unwind_hint sp base reg %d",
insn->sec, insn->offset, hint->sp_reg);
return -1;
}
cfa->offset = hint->sp_offset;
insn->state.type = hint->type;
insn->state.end = hint->end;
}
return 0;
}
static int read_retpoline_hints(struct objtool_file *file)
{
struct section *sec;
struct instruction *insn;
struct rela *rela;
sec = find_section_by_name(file->elf, ".rela.discard.retpoline_safe");
if (!sec)
return 0;
list_for_each_entry(rela, &sec->rela_list, list) {
if (rela->sym->type != STT_SECTION) {
WARN("unexpected relocation symbol type in %s", sec->name);
return -1;
}
insn = find_insn(file, rela->sym->sec, rela->addend);
if (!insn) {
WARN("bad .discard.retpoline_safe entry");
return -1;
}
if (insn->type != INSN_JUMP_DYNAMIC &&
insn->type != INSN_CALL_DYNAMIC) {
WARN_FUNC("retpoline_safe hint not an indirect jump/call",
insn->sec, insn->offset);
return -1;
}
insn->retpoline_safe = true;
}
return 0;
}
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
static void mark_rodata(struct objtool_file *file)
{
struct section *sec;
bool found = false;
/*
* Search for the following rodata sections, each of which can
* potentially contain jump tables:
*
* - .rodata: can contain GCC switch tables
* - .rodata.<func>: same, if -fdata-sections is being used
* - .rodata..c_jump_table: contains C annotated jump tables
*
* .rodata.str1.* sections are ignored; they don't contain jump tables.
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
*/
for_each_sec(file, sec) {
if ((!strncmp(sec->name, ".rodata", 7) && !strstr(sec->name, ".str1.")) ||
!strcmp(sec->name, C_JUMP_TABLE_SECTION)) {
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
sec->rodata = true;
found = true;
}
}
file->rodata = found;
}
static int decode_sections(struct objtool_file *file)
{
int ret;
objtool: Support per-function rodata sections Add support for processing switch jump tables in objects with multiple .rodata sections, such as those created by '-ffunction-sections' and '-fdata-sections'. Currently, objtool always looks in .rodata for jump table information, which results in many "sibling call from callable instruction with modified stack frame" warnings with objects compiled using those flags. The fix is comprised of three parts: 1. Flagging all .rodata sections when importing ELF information for easier checking later. 2. Keeping a reference to the section each relocation is from in order to get the list_head for the other relocations in that section. 3. Finding jump tables by following relocations to .rodata sections, rather than always referencing a single global .rodata section. The patch has been tested without data sections enabled and no differences in the resulting orc unwind information were seen. Note that as objtool adds terminators to end of each .text section the unwind information generated between a function+data sections build and a normal build aren't directly comparable. Manual inspection suggests that objtool is now generating the correct information, or at least making more of an effort to do so than it did previously. Signed-off-by: Allan Xavier <allan.x.xavier@oracle.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/099bdc375195c490dda04db777ee0b95d566ded1.1536325914.git.jpoimboe@redhat.com
2018-09-07 13:12:01 +00:00
mark_rodata(file);
ret = decode_instructions(file);
if (ret)
return ret;
ret = add_dead_ends(file);
if (ret)
return ret;
add_ignores(file);
add_uaccess_safe(file);
ret = add_ignore_alternatives(file);
if (ret)
return ret;
ret = add_jump_destinations(file);
if (ret)
return ret;
ret = add_special_section_alts(file);
if (ret)
return ret;
ret = add_call_destinations(file);
if (ret)
return ret;
ret = add_jump_table_alts(file);
if (ret)
return ret;
ret = read_unwind_hints(file);
if (ret)
return ret;
ret = read_retpoline_hints(file);
if (ret)
return ret;
return 0;
}
static bool is_fentry_call(struct instruction *insn)
{
if (insn->type == INSN_CALL &&
insn->call_dest->type == STT_NOTYPE &&
!strcmp(insn->call_dest->name, "__fentry__"))
return true;
return false;
}
static bool has_modified_stack_frame(struct insn_state *state)
{
int i;
if (state->cfa.base != initial_func_cfi.cfa.base ||
state->cfa.offset != initial_func_cfi.cfa.offset ||
state->stack_size != initial_func_cfi.cfa.offset ||
state->drap)
return true;
for (i = 0; i < CFI_NUM_REGS; i++)
if (state->regs[i].base != initial_func_cfi.regs[i].base ||
state->regs[i].offset != initial_func_cfi.regs[i].offset)
return true;
return false;
}
static bool has_valid_stack_frame(struct insn_state *state)
{
if (state->cfa.base == CFI_BP && state->regs[CFI_BP].base == CFI_CFA &&
state->regs[CFI_BP].offset == -16)
return true;
if (state->drap && state->regs[CFI_BP].base == CFI_BP)
return true;
return false;
}
static int update_insn_state_regs(struct instruction *insn, struct insn_state *state)
{
struct cfi_reg *cfa = &state->cfa;
struct stack_op *op = &insn->stack_op;
if (cfa->base != CFI_SP)
return 0;
/* push */
if (op->dest.type == OP_DEST_PUSH || op->dest.type == OP_DEST_PUSHF)
cfa->offset += 8;
/* pop */
if (op->src.type == OP_SRC_POP || op->src.type == OP_SRC_POPF)
cfa->offset -= 8;
/* add immediate to sp */
if (op->dest.type == OP_DEST_REG && op->src.type == OP_SRC_ADD &&
op->dest.reg == CFI_SP && op->src.reg == CFI_SP)
cfa->offset -= op->src.offset;
return 0;
}
static void save_reg(struct insn_state *state, unsigned char reg, int base,
int offset)
{
if (arch_callee_saved_reg(reg) &&
state->regs[reg].base == CFI_UNDEFINED) {
state->regs[reg].base = base;
state->regs[reg].offset = offset;
}
}
static void restore_reg(struct insn_state *state, unsigned char reg)
{
state->regs[reg].base = CFI_UNDEFINED;
state->regs[reg].offset = 0;
}
/*
* A note about DRAP stack alignment:
*
* GCC has the concept of a DRAP register, which is used to help keep track of
* the stack pointer when aligning the stack. r10 or r13 is used as the DRAP
* register. The typical DRAP pattern is:
*
* 4c 8d 54 24 08 lea 0x8(%rsp),%r10
* 48 83 e4 c0 and $0xffffffffffffffc0,%rsp
* 41 ff 72 f8 pushq -0x8(%r10)
* 55 push %rbp
* 48 89 e5 mov %rsp,%rbp
* (more pushes)
* 41 52 push %r10
* ...
* 41 5a pop %r10
* (more pops)
* 5d pop %rbp
* 49 8d 62 f8 lea -0x8(%r10),%rsp
* c3 retq
*
* There are some variations in the epilogues, like:
*
* 5b pop %rbx
* 41 5a pop %r10
* 41 5c pop %r12
* 41 5d pop %r13
* 41 5e pop %r14
* c9 leaveq
* 49 8d 62 f8 lea -0x8(%r10),%rsp
* c3 retq
*
* and:
*
* 4c 8b 55 e8 mov -0x18(%rbp),%r10
* 48 8b 5d e0 mov -0x20(%rbp),%rbx
* 4c 8b 65 f0 mov -0x10(%rbp),%r12
* 4c 8b 6d f8 mov -0x8(%rbp),%r13
* c9 leaveq
* 49 8d 62 f8 lea -0x8(%r10),%rsp
* c3 retq
*
* Sometimes r13 is used as the DRAP register, in which case it's saved and
* restored beforehand:
*
* 41 55 push %r13
* 4c 8d 6c 24 10 lea 0x10(%rsp),%r13
* 48 83 e4 f0 and $0xfffffffffffffff0,%rsp
* ...
* 49 8d 65 f0 lea -0x10(%r13),%rsp
* 41 5d pop %r13
* c3 retq
*/
static int update_insn_state(struct instruction *insn, struct insn_state *state)
{
struct stack_op *op = &insn->stack_op;
struct cfi_reg *cfa = &state->cfa;
struct cfi_reg *regs = state->regs;
/* stack operations don't make sense with an undefined CFA */
if (cfa->base == CFI_UNDEFINED) {
if (insn->func) {
WARN_FUNC("undefined stack state", insn->sec, insn->offset);
return -1;
}
return 0;
}
if (state->type == ORC_TYPE_REGS || state->type == ORC_TYPE_REGS_IRET)
return update_insn_state_regs(insn, state);
switch (op->dest.type) {
case OP_DEST_REG:
switch (op->src.type) {
case OP_SRC_REG:
objtool: Handle another GCC stack pointer adjustment bug The kbuild bot reported the following warning with GCC 4.4 and a randconfig: net/socket.o: warning: objtool: compat_sock_ioctl()+0x1083: stack state mismatch: cfa1=7+160 cfa2=-1+0 This is caused by another GCC non-optimization, where it backs up and restores the stack pointer for no apparent reason: 2f91: 48 89 e0 mov %rsp,%rax 2f94: 4c 89 e7 mov %r12,%rdi 2f97: 4c 89 f6 mov %r14,%rsi 2f9a: ba 20 00 00 00 mov $0x20,%edx 2f9f: 48 89 c4 mov %rax,%rsp This issue would have been happily ignored before the following commit: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") But now that objtool is paying attention to such stack pointer writes to/from a register, it needs to understand them properly. In this case that means recognizing that the "mov %rsp, %rax" instruction is potentially a backup of the stack pointer. Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Dmitriy Vyukov <dvyukov@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matthias Kaehlcke <mka@chromium.org> Cc: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") Link: http://lkml.kernel.org/r/8c7aa8e9a36fbbb6655d9d8e7cea58958c912da8.1505942196.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-09-20 21:24:32 +00:00
if (op->src.reg == CFI_SP && op->dest.reg == CFI_BP &&
cfa->base == CFI_SP &&
regs[CFI_BP].base == CFI_CFA &&
regs[CFI_BP].offset == -cfa->offset) {
/* mov %rsp, %rbp */
cfa->base = op->dest.reg;
state->bp_scratch = false;
}
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
objtool: Handle another GCC stack pointer adjustment bug The kbuild bot reported the following warning with GCC 4.4 and a randconfig: net/socket.o: warning: objtool: compat_sock_ioctl()+0x1083: stack state mismatch: cfa1=7+160 cfa2=-1+0 This is caused by another GCC non-optimization, where it backs up and restores the stack pointer for no apparent reason: 2f91: 48 89 e0 mov %rsp,%rax 2f94: 4c 89 e7 mov %r12,%rdi 2f97: 4c 89 f6 mov %r14,%rsi 2f9a: ba 20 00 00 00 mov $0x20,%edx 2f9f: 48 89 c4 mov %rax,%rsp This issue would have been happily ignored before the following commit: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") But now that objtool is paying attention to such stack pointer writes to/from a register, it needs to understand them properly. In this case that means recognizing that the "mov %rsp, %rax" instruction is potentially a backup of the stack pointer. Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Dmitriy Vyukov <dvyukov@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matthias Kaehlcke <mka@chromium.org> Cc: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") Link: http://lkml.kernel.org/r/8c7aa8e9a36fbbb6655d9d8e7cea58958c912da8.1505942196.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-09-20 21:24:32 +00:00
else if (op->src.reg == CFI_SP &&
op->dest.reg == CFI_BP && state->drap) {
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
objtool: Handle another GCC stack pointer adjustment bug The kbuild bot reported the following warning with GCC 4.4 and a randconfig: net/socket.o: warning: objtool: compat_sock_ioctl()+0x1083: stack state mismatch: cfa1=7+160 cfa2=-1+0 This is caused by another GCC non-optimization, where it backs up and restores the stack pointer for no apparent reason: 2f91: 48 89 e0 mov %rsp,%rax 2f94: 4c 89 e7 mov %r12,%rdi 2f97: 4c 89 f6 mov %r14,%rsi 2f9a: ba 20 00 00 00 mov $0x20,%edx 2f9f: 48 89 c4 mov %rax,%rsp This issue would have been happily ignored before the following commit: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") But now that objtool is paying attention to such stack pointer writes to/from a register, it needs to understand them properly. In this case that means recognizing that the "mov %rsp, %rax" instruction is potentially a backup of the stack pointer. Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Dmitriy Vyukov <dvyukov@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matthias Kaehlcke <mka@chromium.org> Cc: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") Link: http://lkml.kernel.org/r/8c7aa8e9a36fbbb6655d9d8e7cea58958c912da8.1505942196.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-09-20 21:24:32 +00:00
/* drap: mov %rsp, %rbp */
regs[CFI_BP].base = CFI_BP;
regs[CFI_BP].offset = -state->stack_size;
state->bp_scratch = false;
}
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
objtool: Handle another GCC stack pointer adjustment bug The kbuild bot reported the following warning with GCC 4.4 and a randconfig: net/socket.o: warning: objtool: compat_sock_ioctl()+0x1083: stack state mismatch: cfa1=7+160 cfa2=-1+0 This is caused by another GCC non-optimization, where it backs up and restores the stack pointer for no apparent reason: 2f91: 48 89 e0 mov %rsp,%rax 2f94: 4c 89 e7 mov %r12,%rdi 2f97: 4c 89 f6 mov %r14,%rsi 2f9a: ba 20 00 00 00 mov $0x20,%edx 2f9f: 48 89 c4 mov %rax,%rsp This issue would have been happily ignored before the following commit: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") But now that objtool is paying attention to such stack pointer writes to/from a register, it needs to understand them properly. In this case that means recognizing that the "mov %rsp, %rax" instruction is potentially a backup of the stack pointer. Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Dmitriy Vyukov <dvyukov@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matthias Kaehlcke <mka@chromium.org> Cc: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: dd88a0a0c861 ("objtool: Handle GCC stack pointer adjustment bug") Link: http://lkml.kernel.org/r/8c7aa8e9a36fbbb6655d9d8e7cea58958c912da8.1505942196.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-09-20 21:24:32 +00:00
else if (op->src.reg == CFI_SP && cfa->base == CFI_SP) {
/*
* mov %rsp, %reg
*
* This is needed for the rare case where GCC
* does:
*
* mov %rsp, %rax
* ...
* mov %rax, %rsp
*/
state->vals[op->dest.reg].base = CFI_CFA;
state->vals[op->dest.reg].offset = -state->stack_size;
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
}
else if (op->src.reg == CFI_BP && op->dest.reg == CFI_SP &&
cfa->base == CFI_BP) {
/*
* mov %rbp, %rsp
*
* Restore the original stack pointer (Clang).
*/
state->stack_size = -state->regs[CFI_BP].offset;
}
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
else if (op->dest.reg == cfa->base) {
/* mov %reg, %rsp */
if (cfa->base == CFI_SP &&
state->vals[op->src.reg].base == CFI_CFA) {
/*
* This is needed for the rare case
* where GCC does something dumb like:
*
* lea 0x8(%rsp), %rcx
* ...
* mov %rcx, %rsp
*/
cfa->offset = -state->vals[op->src.reg].offset;
state->stack_size = cfa->offset;
} else {
cfa->base = CFI_UNDEFINED;
cfa->offset = 0;
}
}
break;
case OP_SRC_ADD:
if (op->dest.reg == CFI_SP && op->src.reg == CFI_SP) {
/* add imm, %rsp */
state->stack_size -= op->src.offset;
if (cfa->base == CFI_SP)
cfa->offset -= op->src.offset;
break;
}
if (op->dest.reg == CFI_SP && op->src.reg == CFI_BP) {
/* lea disp(%rbp), %rsp */
state->stack_size = -(op->src.offset + regs[CFI_BP].offset);
break;
}
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
if (op->src.reg == CFI_SP && cfa->base == CFI_SP) {
/* drap: lea disp(%rsp), %drap */
state->drap_reg = op->dest.reg;
objtool: Handle GCC stack pointer adjustment bug Arnd Bergmann reported the following warning with GCC 7.1.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x139: stack state mismatch: cfa1=7+88 cfa2=7+96 And the kbuild robot reported the following warnings with GCC 5.4.1: fs/fs_pin.o: warning: objtool: pin_kill()+0x182: return with modified stack frame fs/quota/dquot.o: warning: objtool: dquot_alloc_inode()+0x140: stack state mismatch: cfa1=7+120 cfa2=7+128 fs/quota/dquot.o: warning: objtool: dquot_free_inode()+0x11a: stack state mismatch: cfa1=7+112 cfa2=7+120 Those warnings are caused by an unusual GCC non-optimization where it uses an intermediate register to adjust the stack pointer. It does: lea 0x8(%rsp), %rcx ... mov %rcx, %rsp Instead of the obvious: add $0x8, %rsp It makes no sense to use an intermediate register, so I opened a GCC bug to track it: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81813 But it's not exactly a high-priority bug and it looks like we'll be stuck with this issue for a while. So for now we have to track register values when they're loaded with stack pointer offsets. This is kind of a big workaround for a tiny problem, but c'est la vie. I hope to eventually create a GCC plugin to implement a big chunk of objtool's functionality. Hopefully at that point we'll be able to remove of a lot of these GCC-isms from the objtool code. Reported-by: Arnd Bergmann <arnd@arndb.de> Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/6a41a96884c725e7f05413bb7df40cfe824b2444.1504028945.git.jpoimboe@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-08-29 17:51:03 +00:00
/*
* lea disp(%rsp), %reg
*
* This is needed for the rare case where GCC
* does something dumb like:
*
* lea 0x8(%rsp), %rcx
* ...
* mov %rcx, %rsp
*/
state->vals[op->dest.reg].base = CFI_CFA;
state->vals[op->dest.reg].offset = \
-state->stack_size + op->src.offset;
break;
}
if (state->drap && op->dest.reg == CFI_SP &&
op->src.reg == state->drap_reg) {
/* drap: lea disp(%drap), %rsp */
cfa->base = CFI_SP;
cfa->offset = state->stack_size = -op->src.offset;
state->drap_reg = CFI_UNDEFINED;
state->drap = false;
break;
}
if (op->dest.reg == state->cfa.base) {
WARN_FUNC("unsupported stack register modification",
insn->sec, insn->offset);
return -1;
}
break;
case OP_SRC_AND:
if (op->dest.reg != CFI_SP ||
(state->drap_reg != CFI_UNDEFINED && cfa->base != CFI_SP) ||
(state->drap_reg == CFI_UNDEFINED && cfa->base != CFI_BP)) {
WARN_FUNC("unsupported stack pointer realignment",
insn->sec, insn->offset);
return -1;
}
if (state->drap_reg != CFI_UNDEFINED) {
/* drap: and imm, %rsp */
cfa->base = state->drap_reg;
cfa->offset = state->stack_size = 0;
state->drap = true;
}
/*
* Older versions of GCC (4.8ish) realign the stack
* without DRAP, with a frame pointer.
*/
break;
case OP_SRC_POP:
case OP_SRC_POPF:
if (!state->drap && op->dest.type == OP_DEST_REG &&
op->dest.reg == cfa->base) {
/* pop %rbp */
cfa->base = CFI_SP;
}
if (state->drap && cfa->base == CFI_BP_INDIRECT &&
op->dest.type == OP_DEST_REG &&
op->dest.reg == state->drap_reg &&
state->drap_offset == -state->stack_size) {
/* drap: pop %drap */
cfa->base = state->drap_reg;
cfa->offset = 0;
state->drap_offset = -1;
} else if (regs[op->dest.reg].offset == -state->stack_size) {
/* pop %reg */
restore_reg(state, op->dest.reg);
}
state->stack_size -= 8;
if (cfa->base == CFI_SP)
cfa->offset -= 8;
break;
case OP_SRC_REG_INDIRECT:
if (state->drap && op->src.reg == CFI_BP &&
op->src.offset == state->drap_offset) {
/* drap: mov disp(%rbp), %drap */
cfa->base = state->drap_reg;
cfa->offset = 0;
state->drap_offset = -1;
}
if (state->drap && op->src.reg == CFI_BP &&
op->src.offset == regs[op->dest.reg].offset) {
/* drap: mov disp(%rbp), %reg */
restore_reg(state, op->dest.reg);
} else if (op->src.reg == cfa->base &&
op->src.offset == regs[op->dest.reg].offset + cfa->offset) {
/* mov disp(%rbp), %reg */
/* mov disp(%rsp), %reg */
restore_reg(state, op->dest.reg);
}
break;
default:
WARN_FUNC("unknown stack-related instruction",
insn->sec, insn->offset);
return -1;
}
break;
case OP_DEST_PUSH:
case OP_DEST_PUSHF:
state->stack_size += 8;
if (cfa->base == CFI_SP)
cfa->offset += 8;
if (op->src.type != OP_SRC_REG)
break;
if (state->drap) {
if (op->src.reg == cfa->base && op->src.reg == state->drap_reg) {
/* drap: push %drap */
cfa->base = CFI_BP_INDIRECT;
cfa->offset = -state->stack_size;
/* save drap so we know when to restore it */
state->drap_offset = -state->stack_size;
} else if (op->src.reg == CFI_BP && cfa->base == state->drap_reg) {
/* drap: push %rbp */
state->stack_size = 0;
} else if (regs[op->src.reg].base == CFI_UNDEFINED) {
/* drap: push %reg */
save_reg(state, op->src.reg, CFI_BP, -state->stack_size);
}
} else {
/* push %reg */
save_reg(state, op->src.reg, CFI_CFA, -state->stack_size);
}
/* detect when asm code uses rbp as a scratch register */
if (!no_fp && insn->func && op->src.reg == CFI_BP &&
cfa->base != CFI_BP)
state->bp_scratch = true;
break;
case OP_DEST_REG_INDIRECT:
if (state->drap) {
if (op->src.reg == cfa->base && op->src.reg == state->drap_reg) {
/* drap: mov %drap, disp(%rbp) */
cfa->base = CFI_BP_INDIRECT;
cfa->offset = op->dest.offset;
/* save drap offset so we know when to restore it */
state->drap_offset = op->dest.offset;
}
else if (regs[op->src.reg].base == CFI_UNDEFINED) {
/* drap: mov reg, disp(%rbp) */
save_reg(state, op->src.reg, CFI_BP, op->dest.offset);
}
} else if (op->dest.reg == cfa->base) {
/* mov reg, disp(%rbp) */
/* mov reg, disp(%rsp) */
save_reg(state, op->src.reg, CFI_CFA,
op->dest.offset - state->cfa.offset);
}
break;
case OP_DEST_LEAVE:
if ((!state->drap && cfa->base != CFI_BP) ||
(state->drap && cfa->base != state->drap_reg)) {
WARN_FUNC("leave instruction with modified stack frame",
insn->sec, insn->offset);
return -1;
}
/* leave (mov %rbp, %rsp; pop %rbp) */
state->stack_size = -state->regs[CFI_BP].offset - 8;
restore_reg(state, CFI_BP);
if (!state->drap) {
cfa->base = CFI_SP;
cfa->offset -= 8;
}
break;
case OP_DEST_MEM:
if (op->src.type != OP_SRC_POP && op->src.type != OP_SRC_POPF) {
WARN_FUNC("unknown stack-related memory operation",
insn->sec, insn->offset);
return -1;
}
/* pop mem */
state->stack_size -= 8;
if (cfa->base == CFI_SP)
cfa->offset -= 8;
break;
default:
WARN_FUNC("unknown stack-related instruction",
insn->sec, insn->offset);
return -1;
}
return 0;
}
static bool insn_state_match(struct instruction *insn, struct insn_state *state)
{
struct insn_state *state1 = &insn->state, *state2 = state;
int i;
if (memcmp(&state1->cfa, &state2->cfa, sizeof(state1->cfa))) {
WARN_FUNC("stack state mismatch: cfa1=%d%+d cfa2=%d%+d",
insn->sec, insn->offset,
state1->cfa.base, state1->cfa.offset,
state2->cfa.base, state2->cfa.offset);
} else if (memcmp(&state1->regs, &state2->regs, sizeof(state1->regs))) {
for (i = 0; i < CFI_NUM_REGS; i++) {
if (!memcmp(&state1->regs[i], &state2->regs[i],
sizeof(struct cfi_reg)))
continue;
WARN_FUNC("stack state mismatch: reg1[%d]=%d%+d reg2[%d]=%d%+d",
insn->sec, insn->offset,
i, state1->regs[i].base, state1->regs[i].offset,
i, state2->regs[i].base, state2->regs[i].offset);
break;
}
} else if (state1->type != state2->type) {
WARN_FUNC("stack state mismatch: type1=%d type2=%d",
insn->sec, insn->offset, state1->type, state2->type);
} else if (state1->drap != state2->drap ||
(state1->drap && state1->drap_reg != state2->drap_reg) ||
(state1->drap && state1->drap_offset != state2->drap_offset)) {
WARN_FUNC("stack state mismatch: drap1=%d(%d,%d) drap2=%d(%d,%d)",
insn->sec, insn->offset,
state1->drap, state1->drap_reg, state1->drap_offset,
state2->drap, state2->drap_reg, state2->drap_offset);
} else
return true;
return false;
}
static inline bool func_uaccess_safe(struct symbol *func)
{
if (func)
return func->uaccess_safe;
return false;
}
static inline const char *call_dest_name(struct instruction *insn)
{
if (insn->call_dest)
return insn->call_dest->name;
return "{dynamic}";
}
static int validate_call(struct instruction *insn, struct insn_state *state)
{
if (state->uaccess && !func_uaccess_safe(insn->call_dest)) {
WARN_FUNC("call to %s() with UACCESS enabled",
insn->sec, insn->offset, call_dest_name(insn));
return 1;
}
if (state->df) {
WARN_FUNC("call to %s() with DF set",
insn->sec, insn->offset, call_dest_name(insn));
return 1;
}
return 0;
}
static int validate_sibling_call(struct instruction *insn, struct insn_state *state)
{
if (has_modified_stack_frame(state)) {
WARN_FUNC("sibling call from callable instruction with modified stack frame",
insn->sec, insn->offset);
return 1;
}
return validate_call(insn, state);
}
static int validate_return(struct symbol *func, struct instruction *insn, struct insn_state *state)
{
if (state->uaccess && !func_uaccess_safe(func)) {
WARN_FUNC("return with UACCESS enabled",
insn->sec, insn->offset);
return 1;
}
if (!state->uaccess && func_uaccess_safe(func)) {
WARN_FUNC("return with UACCESS disabled from a UACCESS-safe function",
insn->sec, insn->offset);
return 1;
}
if (state->df) {
WARN_FUNC("return with DF set",
insn->sec, insn->offset);
return 1;
}
if (func && has_modified_stack_frame(state)) {
WARN_FUNC("return with modified stack frame",
insn->sec, insn->offset);
return 1;
}
if (state->bp_scratch) {
WARN("%s uses BP as a scratch register",
func->name);
return 1;
}
return 0;
}
/*
* Follow the branch starting at the given instruction, and recursively follow
* any other branches (jumps). Meanwhile, track the frame pointer state at
* each instruction and validate all the rules described in
* tools/objtool/Documentation/stack-validation.txt.
*/
static int validate_branch(struct objtool_file *file, struct symbol *func,
struct instruction *first, struct insn_state state)
{
struct alternative *alt;
struct instruction *insn, *next_insn;
struct section *sec;
objtool: Improve UACCESS coverage A clang build reported an (obvious) double CLAC while a GCC build did not; it turns out that objtool only re-visits instructions if the first visit was with AC=0. If OTOH the first visit was with AC=1, it completely ignores any subsequent visit, even when it has AC=0. Fix this by using a visited mask instead of a boolean, and (explicitly) mark the AC state. $ ./objtool check -b --no-fp --retpoline --uaccess drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0x22: redundant UACCESS disable drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xea: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xd9: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xb2: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x39: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x0: <=== (func) Reported-by: Josh Poimboeuf <jpoimboe@redhat.com> Reported-by: Thomas Gleixner <tglx@linutronix.de> Reported-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nathan Chancellor <natechancellor@gmail.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Link: https://github.com/ClangBuiltLinux/linux/issues/617 Link: https://lkml.kernel.org/r/5359166aad2d53f3145cd442d83d0e5115e0cd17.1564007838.git.jpoimboe@redhat.com
2019-07-24 22:47:26 +00:00
u8 visited;
int ret;
insn = first;
sec = insn->sec;
if (insn->alt_group && list_empty(&insn->alts)) {
WARN_FUNC("don't know how to handle branch to middle of alternative instruction group",
sec, insn->offset);
return 1;
}
while (1) {
next_insn = next_insn_same_sec(file, insn);
if (file->c_file && func && insn->func && func != insn->func->pfunc) {
WARN("%s() falls through to next function %s()",
func->name, insn->func->name);
return 1;
}
if (func && insn->ignore) {
WARN_FUNC("BUG: why am I validating an ignored function?",
sec, insn->offset);
return 1;
}
objtool: Improve UACCESS coverage A clang build reported an (obvious) double CLAC while a GCC build did not; it turns out that objtool only re-visits instructions if the first visit was with AC=0. If OTOH the first visit was with AC=1, it completely ignores any subsequent visit, even when it has AC=0. Fix this by using a visited mask instead of a boolean, and (explicitly) mark the AC state. $ ./objtool check -b --no-fp --retpoline --uaccess drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0x22: redundant UACCESS disable drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xea: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xd9: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xb2: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x39: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x0: <=== (func) Reported-by: Josh Poimboeuf <jpoimboe@redhat.com> Reported-by: Thomas Gleixner <tglx@linutronix.de> Reported-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nathan Chancellor <natechancellor@gmail.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Link: https://github.com/ClangBuiltLinux/linux/issues/617 Link: https://lkml.kernel.org/r/5359166aad2d53f3145cd442d83d0e5115e0cd17.1564007838.git.jpoimboe@redhat.com
2019-07-24 22:47:26 +00:00
visited = 1 << state.uaccess;
if (insn->visited) {
if (!insn->hint && !insn_state_match(insn, &state))
return 1;
objtool: Improve UACCESS coverage A clang build reported an (obvious) double CLAC while a GCC build did not; it turns out that objtool only re-visits instructions if the first visit was with AC=0. If OTOH the first visit was with AC=1, it completely ignores any subsequent visit, even when it has AC=0. Fix this by using a visited mask instead of a boolean, and (explicitly) mark the AC state. $ ./objtool check -b --no-fp --retpoline --uaccess drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0x22: redundant UACCESS disable drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xea: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xd9: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xb2: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x39: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x0: <=== (func) Reported-by: Josh Poimboeuf <jpoimboe@redhat.com> Reported-by: Thomas Gleixner <tglx@linutronix.de> Reported-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nathan Chancellor <natechancellor@gmail.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Link: https://github.com/ClangBuiltLinux/linux/issues/617 Link: https://lkml.kernel.org/r/5359166aad2d53f3145cd442d83d0e5115e0cd17.1564007838.git.jpoimboe@redhat.com
2019-07-24 22:47:26 +00:00
if (insn->visited & visited)
return 0;
}
if (insn->hint) {
if (insn->restore) {
struct instruction *save_insn, *i;
i = insn;
save_insn = NULL;
sym_for_each_insn_continue_reverse(file, func, i) {
if (i->save) {
save_insn = i;
break;
}
}
if (!save_insn) {
WARN_FUNC("no corresponding CFI save for CFI restore",
sec, insn->offset);
return 1;
}
if (!save_insn->visited) {
/*
* Oops, no state to copy yet.
* Hopefully we can reach this
* instruction from another branch
* after the save insn has been
* visited.
*/
if (insn == first)
return 0;
WARN_FUNC("objtool isn't smart enough to handle this CFI save/restore combo",
sec, insn->offset);
return 1;
}
insn->state = save_insn->state;
}
state = insn->state;
} else
insn->state = state;
objtool: Improve UACCESS coverage A clang build reported an (obvious) double CLAC while a GCC build did not; it turns out that objtool only re-visits instructions if the first visit was with AC=0. If OTOH the first visit was with AC=1, it completely ignores any subsequent visit, even when it has AC=0. Fix this by using a visited mask instead of a boolean, and (explicitly) mark the AC state. $ ./objtool check -b --no-fp --retpoline --uaccess drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0x22: redundant UACCESS disable drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xea: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xd9: (alt) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0xb2: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x39: (branch) drivers/gpu/drm/i915/gem/i915_gem_execbuffer.o: warning: objtool: eb_copy_relocations.isra.34()+0x0: <=== (func) Reported-by: Josh Poimboeuf <jpoimboe@redhat.com> Reported-by: Thomas Gleixner <tglx@linutronix.de> Reported-by: Sedat Dilek <sedat.dilek@gmail.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Nathan Chancellor <natechancellor@gmail.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> Link: https://github.com/ClangBuiltLinux/linux/issues/617 Link: https://lkml.kernel.org/r/5359166aad2d53f3145cd442d83d0e5115e0cd17.1564007838.git.jpoimboe@redhat.com
2019-07-24 22:47:26 +00:00
insn->visited |= visited;
if (!insn->ignore_alts) {
bool skip_orig = false;
list_for_each_entry(alt, &insn->alts, list) {
if (alt->skip_orig)
skip_orig = true;
ret = validate_branch(file, func, alt->insn, state);
objtool: Add --backtrace support For when you want to know the path that reached your fail state: $ ./objtool check --no-fp --backtrace arch/x86/lib/usercopy_64.o arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0x3: UACCESS disable without MEMOPs: __clear_user() arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x3a: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x2e: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x18: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x5: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x0: <=== (func) 0000000000000000 <__clear_user>: 0: e8 00 00 00 00 callq 5 <__clear_user+0x5> 1: R_X86_64_PLT32 __fentry__-0x4 5: 90 nop 6: 90 nop 7: 90 nop 8: 48 89 f0 mov %rsi,%rax b: 48 c1 ee 03 shr $0x3,%rsi f: 83 e0 07 and $0x7,%eax 12: 48 89 f1 mov %rsi,%rcx 15: 48 85 c9 test %rcx,%rcx 18: 74 0f je 29 <__clear_user+0x29> 1a: 48 c7 07 00 00 00 00 movq $0x0,(%rdi) 21: 48 83 c7 08 add $0x8,%rdi 25: ff c9 dec %ecx 27: 75 f1 jne 1a <__clear_user+0x1a> 29: 48 89 c1 mov %rax,%rcx 2c: 85 c9 test %ecx,%ecx 2e: 74 0a je 3a <__clear_user+0x3a> 30: c6 07 00 movb $0x0,(%rdi) 33: 48 ff c7 inc %rdi 36: ff c9 dec %ecx 38: 75 f6 jne 30 <__clear_user+0x30> 3a: 90 nop 3b: 90 nop 3c: 90 nop 3d: 48 89 c8 mov %rcx,%rax 40: c3 retq Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-03-01 10:15:49 +00:00
if (ret) {
if (backtrace)
BT_FUNC("(alt)", insn);
return ret;
}
}
if (skip_orig)
return 0;
}
switch (insn->type) {
case INSN_RETURN:
return validate_return(func, insn, &state);
case INSN_CALL:
case INSN_CALL_DYNAMIC:
ret = validate_call(insn, &state);
if (ret)
return ret;
if (!no_fp && func && !is_fentry_call(insn) &&
!has_valid_stack_frame(&state)) {
WARN_FUNC("call without frame pointer save/setup",
sec, insn->offset);
return 1;
}
if (dead_end_function(file, insn->call_dest))
return 0;
break;
case INSN_JUMP_CONDITIONAL:
case INSN_JUMP_UNCONDITIONAL:
if (func && is_sibling_call(insn)) {
ret = validate_sibling_call(insn, &state);
if (ret)
return ret;
} else if (insn->jump_dest) {
ret = validate_branch(file, func,
insn->jump_dest, state);
objtool: Add --backtrace support For when you want to know the path that reached your fail state: $ ./objtool check --no-fp --backtrace arch/x86/lib/usercopy_64.o arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0x3: UACCESS disable without MEMOPs: __clear_user() arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x3a: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x2e: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x18: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x5: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x0: <=== (func) 0000000000000000 <__clear_user>: 0: e8 00 00 00 00 callq 5 <__clear_user+0x5> 1: R_X86_64_PLT32 __fentry__-0x4 5: 90 nop 6: 90 nop 7: 90 nop 8: 48 89 f0 mov %rsi,%rax b: 48 c1 ee 03 shr $0x3,%rsi f: 83 e0 07 and $0x7,%eax 12: 48 89 f1 mov %rsi,%rcx 15: 48 85 c9 test %rcx,%rcx 18: 74 0f je 29 <__clear_user+0x29> 1a: 48 c7 07 00 00 00 00 movq $0x0,(%rdi) 21: 48 83 c7 08 add $0x8,%rdi 25: ff c9 dec %ecx 27: 75 f1 jne 1a <__clear_user+0x1a> 29: 48 89 c1 mov %rax,%rcx 2c: 85 c9 test %ecx,%ecx 2e: 74 0a je 3a <__clear_user+0x3a> 30: c6 07 00 movb $0x0,(%rdi) 33: 48 ff c7 inc %rdi 36: ff c9 dec %ecx 38: 75 f6 jne 30 <__clear_user+0x30> 3a: 90 nop 3b: 90 nop 3c: 90 nop 3d: 48 89 c8 mov %rcx,%rax 40: c3 retq Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-03-01 10:15:49 +00:00
if (ret) {
if (backtrace)
BT_FUNC("(branch)", insn);
return ret;
}
}
if (insn->type == INSN_JUMP_UNCONDITIONAL)
return 0;
break;
case INSN_JUMP_DYNAMIC:
case INSN_JUMP_DYNAMIC_CONDITIONAL:
if (func && is_sibling_call(insn)) {
ret = validate_sibling_call(insn, &state);
if (ret)
return ret;
}
if (insn->type == INSN_JUMP_DYNAMIC)
return 0;
break;
case INSN_CONTEXT_SWITCH:
if (func && (!next_insn || !next_insn->hint)) {
WARN_FUNC("unsupported instruction in callable function",
sec, insn->offset);
return 1;
}
return 0;
case INSN_STACK:
if (update_insn_state(insn, &state))
return 1;
if (insn->stack_op.dest.type == OP_DEST_PUSHF) {
if (!state.uaccess_stack) {
state.uaccess_stack = 1;
} else if (state.uaccess_stack >> 31) {
WARN_FUNC("PUSHF stack exhausted", sec, insn->offset);
return 1;
}
state.uaccess_stack <<= 1;
state.uaccess_stack |= state.uaccess;
}
if (insn->stack_op.src.type == OP_SRC_POPF) {
if (state.uaccess_stack) {
state.uaccess = state.uaccess_stack & 1;
state.uaccess_stack >>= 1;
if (state.uaccess_stack == 1)
state.uaccess_stack = 0;
}
}
break;
case INSN_STAC:
if (state.uaccess) {
WARN_FUNC("recursive UACCESS enable", sec, insn->offset);
return 1;
}
state.uaccess = true;
break;
case INSN_CLAC:
if (!state.uaccess && func) {
WARN_FUNC("redundant UACCESS disable", sec, insn->offset);
return 1;
}
if (func_uaccess_safe(func) && !state.uaccess_stack) {
WARN_FUNC("UACCESS-safe disables UACCESS", sec, insn->offset);
return 1;
}
state.uaccess = false;
break;
case INSN_STD:
if (state.df)
WARN_FUNC("recursive STD", sec, insn->offset);
state.df = true;
break;
case INSN_CLD:
if (!state.df && func)
WARN_FUNC("redundant CLD", sec, insn->offset);
state.df = false;
break;
default:
break;
}
if (insn->dead_end)
return 0;
if (!next_insn) {
if (state.cfa.base == CFI_UNDEFINED)
return 0;
WARN("%s: unexpected end of section", sec->name);
return 1;
}
insn = next_insn;
}
return 0;
}
static int validate_unwind_hints(struct objtool_file *file)
{
struct instruction *insn;
int ret, warnings = 0;
struct insn_state state;
if (!file->hints)
return 0;
clear_insn_state(&state);
for_each_insn(file, insn) {
if (insn->hint && !insn->visited) {
ret = validate_branch(file, insn->func, insn, state);
objtool: Add --backtrace support For when you want to know the path that reached your fail state: $ ./objtool check --no-fp --backtrace arch/x86/lib/usercopy_64.o arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0x3: UACCESS disable without MEMOPs: __clear_user() arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x3a: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x2e: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x18: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x5: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x0: <=== (func) 0000000000000000 <__clear_user>: 0: e8 00 00 00 00 callq 5 <__clear_user+0x5> 1: R_X86_64_PLT32 __fentry__-0x4 5: 90 nop 6: 90 nop 7: 90 nop 8: 48 89 f0 mov %rsi,%rax b: 48 c1 ee 03 shr $0x3,%rsi f: 83 e0 07 and $0x7,%eax 12: 48 89 f1 mov %rsi,%rcx 15: 48 85 c9 test %rcx,%rcx 18: 74 0f je 29 <__clear_user+0x29> 1a: 48 c7 07 00 00 00 00 movq $0x0,(%rdi) 21: 48 83 c7 08 add $0x8,%rdi 25: ff c9 dec %ecx 27: 75 f1 jne 1a <__clear_user+0x1a> 29: 48 89 c1 mov %rax,%rcx 2c: 85 c9 test %ecx,%ecx 2e: 74 0a je 3a <__clear_user+0x3a> 30: c6 07 00 movb $0x0,(%rdi) 33: 48 ff c7 inc %rdi 36: ff c9 dec %ecx 38: 75 f6 jne 30 <__clear_user+0x30> 3a: 90 nop 3b: 90 nop 3c: 90 nop 3d: 48 89 c8 mov %rcx,%rax 40: c3 retq Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-03-01 10:15:49 +00:00
if (ret && backtrace)
BT_FUNC("<=== (hint)", insn);
warnings += ret;
}
}
return warnings;
}
static int validate_retpoline(struct objtool_file *file)
{
struct instruction *insn;
int warnings = 0;
for_each_insn(file, insn) {
if (insn->type != INSN_JUMP_DYNAMIC &&
insn->type != INSN_CALL_DYNAMIC)
continue;
if (insn->retpoline_safe)
continue;
/*
* .init.text code is ran before userspace and thus doesn't
* strictly need retpolines, except for modules which are
* loaded late, they very much do need retpoline in their
* .init.text
*/
if (!strcmp(insn->sec->name, ".init.text") && !module)
continue;
WARN_FUNC("indirect %s found in RETPOLINE build",
insn->sec, insn->offset,
insn->type == INSN_JUMP_DYNAMIC ? "jump" : "call");
warnings++;
}
return warnings;
}
static bool is_kasan_insn(struct instruction *insn)
{
return (insn->type == INSN_CALL &&
!strcmp(insn->call_dest->name, "__asan_handle_no_return"));
}
static bool is_ubsan_insn(struct instruction *insn)
{
return (insn->type == INSN_CALL &&
!strcmp(insn->call_dest->name,
"__ubsan_handle_builtin_unreachable"));
}
static bool ignore_unreachable_insn(struct instruction *insn)
{
int i;
if (insn->ignore || insn->type == INSN_NOP)
return true;
/*
* Ignore any unused exceptions. This can happen when a whitelisted
* function has an exception table entry.
*
* Also ignore alternative replacement instructions. This can happen
* when a whitelisted function uses one of the ALTERNATIVE macros.
*/
if (!strcmp(insn->sec->name, ".fixup") ||
!strcmp(insn->sec->name, ".altinstr_replacement") ||
!strcmp(insn->sec->name, ".altinstr_aux"))
return true;
/*
* Check if this (or a subsequent) instruction is related to
* CONFIG_UBSAN or CONFIG_KASAN.
*
* End the search at 5 instructions to avoid going into the weeds.
*/
if (!insn->func)
return false;
for (i = 0; i < 5; i++) {
if (is_kasan_insn(insn) || is_ubsan_insn(insn))
return true;
if (insn->type == INSN_JUMP_UNCONDITIONAL) {
if (insn->jump_dest &&
insn->jump_dest->func == insn->func) {
insn = insn->jump_dest;
continue;
}
break;
}
if (insn->offset + insn->len >= insn->func->offset + insn->func->len)
break;
insn = list_next_entry(insn, list);
}
return false;
}
static int validate_functions(struct objtool_file *file)
{
struct section *sec;
struct symbol *func;
struct instruction *insn;
struct insn_state state;
int ret, warnings = 0;
clear_insn_state(&state);
state.cfa = initial_func_cfi.cfa;
memcpy(&state.regs, &initial_func_cfi.regs,
CFI_NUM_REGS * sizeof(struct cfi_reg));
state.stack_size = initial_func_cfi.cfa.offset;
for_each_sec(file, sec) {
list_for_each_entry(func, &sec->symbol_list, list) {
if (func->type != STT_FUNC)
continue;
if (!func->len) {
WARN("%s() is missing an ELF size annotation",
func->name);
warnings++;
}
if (func->pfunc != func || func->alias != func)
continue;
insn = find_insn(file, sec, func->offset);
if (!insn || insn->ignore || insn->visited)
continue;
state.uaccess = func->uaccess_safe;
ret = validate_branch(file, func, insn, state);
objtool: Add --backtrace support For when you want to know the path that reached your fail state: $ ./objtool check --no-fp --backtrace arch/x86/lib/usercopy_64.o arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0x3: UACCESS disable without MEMOPs: __clear_user() arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x3a: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x2e: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x18: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: .altinstr_replacement+0xffffffffffffffff: (branch) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x5: (alt) arch/x86/lib/usercopy_64.o: warning: objtool: __clear_user()+0x0: <=== (func) 0000000000000000 <__clear_user>: 0: e8 00 00 00 00 callq 5 <__clear_user+0x5> 1: R_X86_64_PLT32 __fentry__-0x4 5: 90 nop 6: 90 nop 7: 90 nop 8: 48 89 f0 mov %rsi,%rax b: 48 c1 ee 03 shr $0x3,%rsi f: 83 e0 07 and $0x7,%eax 12: 48 89 f1 mov %rsi,%rcx 15: 48 85 c9 test %rcx,%rcx 18: 74 0f je 29 <__clear_user+0x29> 1a: 48 c7 07 00 00 00 00 movq $0x0,(%rdi) 21: 48 83 c7 08 add $0x8,%rdi 25: ff c9 dec %ecx 27: 75 f1 jne 1a <__clear_user+0x1a> 29: 48 89 c1 mov %rax,%rcx 2c: 85 c9 test %ecx,%ecx 2e: 74 0a je 3a <__clear_user+0x3a> 30: c6 07 00 movb $0x0,(%rdi) 33: 48 ff c7 inc %rdi 36: ff c9 dec %ecx 38: 75 f6 jne 30 <__clear_user+0x30> 3a: 90 nop 3b: 90 nop 3c: 90 nop 3d: 48 89 c8 mov %rcx,%rax 40: c3 retq Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@kernel.org>
2019-03-01 10:15:49 +00:00
if (ret && backtrace)
BT_FUNC("<=== (func)", insn);
warnings += ret;
}
}
return warnings;
}
static int validate_reachable_instructions(struct objtool_file *file)
{
struct instruction *insn;
if (file->ignore_unreachables)
return 0;
for_each_insn(file, insn) {
if (insn->visited || ignore_unreachable_insn(insn))
continue;
WARN_FUNC("unreachable instruction", insn->sec, insn->offset);
return 1;
}
return 0;
}
static struct objtool_file file;
int check(const char *_objname, bool orc)
{
int ret, warnings = 0;
objname = _objname;
file.elf = elf_read(objname, orc ? O_RDWR : O_RDONLY);
if (!file.elf)
return 1;
INIT_LIST_HEAD(&file.insn_list);
hash_init(file.insn_hash);
file.c_file = find_section_by_name(file.elf, ".comment");
file.ignore_unreachables = no_unreachable;
file.hints = false;
arch_initial_func_cfi_state(&initial_func_cfi);
ret = decode_sections(&file);
if (ret < 0)
goto out;
warnings += ret;
if (list_empty(&file.insn_list))
goto out;
if (retpoline) {
ret = validate_retpoline(&file);
if (ret < 0)
return ret;
warnings += ret;
}
ret = validate_functions(&file);
if (ret < 0)
goto out;
warnings += ret;
ret = validate_unwind_hints(&file);
if (ret < 0)
goto out;
warnings += ret;
if (!warnings) {
ret = validate_reachable_instructions(&file);
if (ret < 0)
goto out;
warnings += ret;
}
if (orc) {
ret = create_orc(&file);
if (ret < 0)
goto out;
ret = create_orc_sections(&file);
if (ret < 0)
goto out;
ret = elf_write(file.elf);
if (ret < 0)
goto out;
}
out:
if (ret < 0) {
/*
* Fatal error. The binary is corrupt or otherwise broken in
* some way, or objtool itself is broken. Fail the kernel
* build.
*/
return ret;
}
return 0;
}