tequilaOS/platform_system_core
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
platform_system_core/libcorkscrew/backtrace.c

306 lines
11 KiB
C
Raw Normal View History

Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
/*
* Copyright (C) 2011 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#define LOG_TAG "Corkscrew"
//#define LOG_NDEBUG 0
#include "backtrace-arch.h"
#include "backtrace-helper.h"
#include "ptrace-arch.h"
#include <corkscrew/map_info.h>
#include <corkscrew/symbol_table.h>
#include <corkscrew/ptrace.h>
#include <corkscrew/demangle.h>
#include <unistd.h>
#include <signal.h>
#include <pthread.h>
#include <unwind.h>
#include <sys/exec_elf.h>
#include <cutils/log.h>
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
#include <cutils/atomic.h>
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
#if HAVE_DLADDR
#include <dlfcn.h>
#endif
typedef struct {
backtrace_frame_t* backtrace;
size_t ignore_depth;
size_t max_depth;
size_t ignored_frames;
size_t returned_frames;
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
memory_t memory;
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
} backtrace_state_t;
static _Unwind_Reason_Code unwind_backtrace_callback(struct _Unwind_Context* context, void* arg) {
backtrace_state_t* state = (backtrace_state_t*)arg;
uintptr_t pc = _Unwind_GetIP(context);
if (pc) {
// TODO: Get information about the stack layout from the _Unwind_Context.
// This will require a new architecture-specific function to query
// the appropriate registers. Current callers of unwind_backtrace
// don't need this information, so we won't bother collecting it just yet.
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
add_backtrace_entry(rewind_pc_arch(&state->memory, pc), state->backtrace,
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
state->ignore_depth, state->max_depth,
&state->ignored_frames, &state->returned_frames);
}
return state->returned_frames < state->max_depth ? _URC_NO_REASON : _URC_END_OF_STACK;
}
ssize_t unwind_backtrace(backtrace_frame_t* backtrace, size_t ignore_depth, size_t max_depth) {
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
ALOGV("Unwinding current thread %d.", gettid());
map_info_t* milist = acquire_my_map_info_list();
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
backtrace_state_t state;
state.backtrace = backtrace;
state.ignore_depth = ignore_depth;
state.max_depth = max_depth;
state.ignored_frames = 0;
state.returned_frames = 0;
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
init_memory(&state.memory, milist);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
_Unwind_Reason_Code rc =_Unwind_Backtrace(unwind_backtrace_callback, &state);
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
release_my_map_info_list(milist);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
if (state.returned_frames) {
return state.returned_frames;
}
return rc == _URC_END_OF_STACK ? 0 : -1;
}
#ifdef CORKSCREW_HAVE_ARCH
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
static const int32_t STATE_DUMPING = -1;
static const int32_t STATE_DONE = -2;
static const int32_t STATE_CANCEL = -3;
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
static pthread_mutex_t g_unwind_signal_mutex = PTHREAD_MUTEX_INITIALIZER;
static volatile struct {
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
int32_t tid_state;
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
const map_info_t* map_info_list;
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
backtrace_frame_t* backtrace;
size_t ignore_depth;
size_t max_depth;
size_t returned_frames;
} g_unwind_signal_state;
static void unwind_backtrace_thread_signal_handler(int n, siginfo_t* siginfo, void* sigcontext) {
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
if (!android_atomic_acquire_cas(gettid(), STATE_DUMPING, &g_unwind_signal_state.tid_state)) {
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
g_unwind_signal_state.returned_frames = unwind_backtrace_signal_arch(
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
siginfo, sigcontext,
g_unwind_signal_state.map_info_list,
g_unwind_signal_state.backtrace,
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
g_unwind_signal_state.ignore_depth,
g_unwind_signal_state.max_depth);
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
android_atomic_release_store(STATE_DONE, &g_unwind_signal_state.tid_state);
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
} else {
ALOGV("Received spurious SIGURG on thread %d that was intended for thread %d.",
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
gettid(), android_atomic_acquire_load(&g_unwind_signal_state.tid_state));
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
}
}
#endif
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
extern int tgkill(int tgid, int tid, int sig);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
ssize_t unwind_backtrace_thread(pid_t tid, backtrace_frame_t* backtrace,
size_t ignore_depth, size_t max_depth) {
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
if (tid == gettid()) {
return unwind_backtrace(backtrace, ignore_depth + 1, max_depth);
}
ALOGV("Unwinding thread %d from thread %d.", tid, gettid());
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
#ifdef CORKSCREW_HAVE_ARCH
struct sigaction act;
struct sigaction oact;
memset(&act, 0, sizeof(act));
act.sa_sigaction = unwind_backtrace_thread_signal_handler;
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
act.sa_flags = SA_RESTART | SA_SIGINFO | SA_ONSTACK;
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
sigemptyset(&act.sa_mask);
pthread_mutex_lock(&g_unwind_signal_mutex);
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
map_info_t* milist = acquire_my_map_info_list();
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
ssize_t frames = -1;
if (!sigaction(SIGURG, &act, &oact)) {
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
g_unwind_signal_state.map_info_list = milist;
g_unwind_signal_state.backtrace = backtrace;
g_unwind_signal_state.ignore_depth = ignore_depth;
g_unwind_signal_state.max_depth = max_depth;
g_unwind_signal_state.returned_frames = 0;
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
android_atomic_release_store(tid, &g_unwind_signal_state.tid_state);
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
// Signal the specific thread that we want to dump.
int32_t tid_state = tid;
if (tgkill(getpid(), tid, SIGURG)) {
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
ALOGV("Failed to send SIGURG to thread %d.", tid);
} else {
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
// Wait for the other thread to start dumping the stack, or time out.
int wait_millis = 250;
for (;;) {
tid_state = android_atomic_acquire_load(&g_unwind_signal_state.tid_state);
if (tid_state != tid) {
break;
}
if (wait_millis--) {
ALOGV("Waiting for thread %d to start dumping the stack...", tid);
usleep(1000);
} else {
ALOGV("Timed out waiting for thread %d to start dumping the stack.", tid);
break;
}
}
}
// Try to cancel the dump if it has not started yet.
if (tid_state == tid) {
if (!android_atomic_acquire_cas(tid, STATE_CANCEL, &g_unwind_signal_state.tid_state)) {
ALOGV("Canceled thread %d stack dump.", tid);
tid_state = STATE_CANCEL;
} else {
tid_state = android_atomic_acquire_load(&g_unwind_signal_state.tid_state);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
}
Minor robustness tweaks. Use the alternate stack specified by sigaltstack (if there is one). Use tgkill instead of kill to ensure that we send the signal to precisely the right thread. Added timeout logic. Change-Id: I72922137d1bc0c8becff98244926bde09fe3990e
2011-11-19 00:34:35 +01:00
}
// Wait indefinitely for the dump to finish or be canceled.
// We cannot apply a timeout here because the other thread is accessing state that
// is owned by this thread, such as milist. It should not take very
// long to take the dump once started.
while (tid_state == STATE_DUMPING) {
ALOGV("Waiting for thread %d to finish dumping the stack...", tid);
usleep(1000);
tid_state = android_atomic_acquire_load(&g_unwind_signal_state.tid_state);
}
if (tid_state == STATE_DONE) {
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
frames = g_unwind_signal_state.returned_frames;
}
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
sigaction(SIGURG, &oact, NULL);
}
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
release_my_map_info_list(milist);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
pthread_mutex_unlock(&g_unwind_signal_mutex);
return frames;
#else
return -1;
#endif
}
ssize_t unwind_backtrace_ptrace(pid_t tid, const ptrace_context_t* context,
backtrace_frame_t* backtrace, size_t ignore_depth, size_t max_depth) {
#ifdef CORKSCREW_HAVE_ARCH
return unwind_backtrace_ptrace_arch(tid, context, backtrace, ignore_depth, max_depth);
#else
return -1;
#endif
}
static void init_backtrace_symbol(backtrace_symbol_t* symbol, uintptr_t pc) {
symbol->relative_pc = pc;
Improve backtrace formatting. Show the symbol offset, when available. Centralized formatting of native stack traces in libcorkscrew. It's handy for automated tools if all stacks look the same. Since we already made them all look them same, we might as well do the formatting in just one place. Do not strip the Thumb bit on ARM. This fixes an off-by-one issue that could happen when resolving a PC that was at the very beginning of a Thumb function, since the symbol table would have the Thumb bit set but since we stripped the bit from our PC, we would be looking for an address one byte before the one listed in the symbol table. It's also quite useful to see whether a given function is executing in Thumb mode just by glancing at the PC. Change-Id: Icaa29add85ce0bcafe24d5ce2098e138d809e2ab
2011-11-22 06:10:00 +01:00
symbol->relative_symbol_addr = 0;
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
symbol->map_name = NULL;
Improve backtrace formatting. Show the symbol offset, when available. Centralized formatting of native stack traces in libcorkscrew. It's handy for automated tools if all stacks look the same. Since we already made them all look them same, we might as well do the formatting in just one place. Do not strip the Thumb bit on ARM. This fixes an off-by-one issue that could happen when resolving a PC that was at the very beginning of a Thumb function, since the symbol table would have the Thumb bit set but since we stripped the bit from our PC, we would be looking for an address one byte before the one listed in the symbol table. It's also quite useful to see whether a given function is executing in Thumb mode just by glancing at the PC. Change-Id: Icaa29add85ce0bcafe24d5ce2098e138d809e2ab
2011-11-22 06:10:00 +01:00
symbol->symbol_name = NULL;
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
symbol->demangled_name = NULL;
}
void get_backtrace_symbols(const backtrace_frame_t* backtrace, size_t frames,
backtrace_symbol_t* backtrace_symbols) {
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
map_info_t* milist = acquire_my_map_info_list();
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
for (size_t i = 0; i < frames; i++) {
const backtrace_frame_t* frame = &backtrace[i];
backtrace_symbol_t* symbol = &backtrace_symbols[i];
init_backtrace_symbol(symbol, frame->absolute_pc);
const map_info_t* mi = find_map_info(milist, frame->absolute_pc);
if (mi) {
symbol->relative_pc = frame->absolute_pc - mi->start;
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
if (mi->name[0]) {
symbol->map_name = strdup(mi->name);
}
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
#if HAVE_DLADDR
Dl_info info;
if (dladdr((const void*)frame->absolute_pc, &info) && info.dli_sname) {
Improve backtrace formatting. Show the symbol offset, when available. Centralized formatting of native stack traces in libcorkscrew. It's handy for automated tools if all stacks look the same. Since we already made them all look them same, we might as well do the formatting in just one place. Do not strip the Thumb bit on ARM. This fixes an off-by-one issue that could happen when resolving a PC that was at the very beginning of a Thumb function, since the symbol table would have the Thumb bit set but since we stripped the bit from our PC, we would be looking for an address one byte before the one listed in the symbol table. It's also quite useful to see whether a given function is executing in Thumb mode just by glancing at the PC. Change-Id: Icaa29add85ce0bcafe24d5ce2098e138d809e2ab
2011-11-22 06:10:00 +01:00
symbol->relative_symbol_addr = (uintptr_t)info.dli_saddr
- (uintptr_t)info.dli_fbase;
symbol->symbol_name = strdup(info.dli_sname);
symbol->demangled_name = demangle_symbol_name(symbol->symbol_name);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
}
#endif
}
}
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
release_my_map_info_list(milist);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
}
void get_backtrace_symbols_ptrace(const ptrace_context_t* context,
const backtrace_frame_t* backtrace, size_t frames,
backtrace_symbol_t* backtrace_symbols) {
for (size_t i = 0; i < frames; i++) {
const backtrace_frame_t* frame = &backtrace[i];
backtrace_symbol_t* symbol = &backtrace_symbols[i];
init_backtrace_symbol(symbol, frame->absolute_pc);
const map_info_t* mi;
const symbol_t* s;
find_symbol_ptrace(context, frame->absolute_pc, &mi, &s);
if (mi) {
symbol->relative_pc = frame->absolute_pc - mi->start;
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
if (mi->name[0]) {
symbol->map_name = strdup(mi->name);
}
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
}
if (s) {
Improve backtrace formatting. Show the symbol offset, when available. Centralized formatting of native stack traces in libcorkscrew. It's handy for automated tools if all stacks look the same. Since we already made them all look them same, we might as well do the formatting in just one place. Do not strip the Thumb bit on ARM. This fixes an off-by-one issue that could happen when resolving a PC that was at the very beginning of a Thumb function, since the symbol table would have the Thumb bit set but since we stripped the bit from our PC, we would be looking for an address one byte before the one listed in the symbol table. It's also quite useful to see whether a given function is executing in Thumb mode just by glancing at the PC. Change-Id: Icaa29add85ce0bcafe24d5ce2098e138d809e2ab
2011-11-22 06:10:00 +01:00
symbol->relative_symbol_addr = s->start;
symbol->symbol_name = strdup(s->name);
symbol->demangled_name = demangle_symbol_name(symbol->symbol_name);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
}
}
}
void free_backtrace_symbols(backtrace_symbol_t* backtrace_symbols, size_t frames) {
for (size_t i = 0; i < frames; i++) {
backtrace_symbol_t* symbol = &backtrace_symbols[i];
Improve stack unwinder robustness. Keep track of whether memory maps are readable. Use the information in try_get_word to try to avoid accidentally dereferencing an invalid pointer within the current process. (Note that I haven't ever seen that happen during normal unwinding, but it pays to be a little more careful.) Refactored try_get_word a little to make it easier to pass it the needed state for validation checks by way of a little memory_t struct. Improved how the memory map for the current process is cached. This is important because we need up to date information about readable maps. Use a 5 second cache expiration. Improved the PC -> LR fallback logic in the unwinder so we can eke out an extra frame sometimes. Fixed a bug reading ELF program headers. The phnum & phentsize fields are half-words. We were incorrectly interpreting phnum as a whole word. Used android_atomic_* operations carefully in the unwinder to prevent possible memory races between the dumper and the dumpee. This was highly unlikely (or even impossible due to the presence of other barriers along the way) but the code is clearer now about its invariants. Fixed a bug in debuggerd where the pid was being passed to have its stack dump taken instead of the tid, resulting in short stacks because ptrace couldn't read the data if pid != tid. Did a full sweep to ensure that we use pid / tid correctly everywhere. Ported old code from debuggerd to rewind the program counter back one instruction so that it points to the branch instruction itself instead of the return address. Change-Id: Icc4eb08320052975a4ae7f0f5f0ac9308a2d33d7
2011-11-04 01:58:44 +01:00
free(symbol->map_name);
Improve backtrace formatting. Show the symbol offset, when available. Centralized formatting of native stack traces in libcorkscrew. It's handy for automated tools if all stacks look the same. Since we already made them all look them same, we might as well do the formatting in just one place. Do not strip the Thumb bit on ARM. This fixes an off-by-one issue that could happen when resolving a PC that was at the very beginning of a Thumb function, since the symbol table would have the Thumb bit set but since we stripped the bit from our PC, we would be looking for an address one byte before the one listed in the symbol table. It's also quite useful to see whether a given function is executing in Thumb mode just by glancing at the PC. Change-Id: Icaa29add85ce0bcafe24d5ce2098e138d809e2ab
2011-11-22 06:10:00 +01:00
free(symbol->symbol_name);
Add a new library for collecting stack traces. Supports collecting the stack trace of the current thread, another thread in the same process, or a thread in a different process (using ptrace). Change-Id: Ica2594e4436edde4ceb7bcc3d78e6c31a7902cbf
2011-10-20 05:35:35 +02:00
free(symbol->demangled_name);
init_backtrace_symbol(symbol, 0);
}
}
Improve backtrace formatting. Show the symbol offset, when available. Centralized formatting of native stack traces in libcorkscrew. It's handy for automated tools if all stacks look the same. Since we already made them all look them same, we might as well do the formatting in just one place. Do not strip the Thumb bit on ARM. This fixes an off-by-one issue that could happen when resolving a PC that was at the very beginning of a Thumb function, since the symbol table would have the Thumb bit set but since we stripped the bit from our PC, we would be looking for an address one byte before the one listed in the symbol table. It's also quite useful to see whether a given function is executing in Thumb mode just by glancing at the PC. Change-Id: Icaa29add85ce0bcafe24d5ce2098e138d809e2ab
2011-11-22 06:10:00 +01:00
void format_backtrace_line(unsigned frameNumber, const backtrace_frame_t* frame,
const backtrace_symbol_t* symbol, char* buffer, size_t bufferSize) {
const char* mapName = symbol->map_name ? symbol->map_name : "<unknown>";
const char* symbolName = symbol->demangled_name ? symbol->demangled_name : symbol->symbol_name;
size_t fieldWidth = (bufferSize - 80) / 2;
if (symbolName) {
uint32_t pc_offset = symbol->relative_pc - symbol->relative_symbol_addr;
if (pc_offset) {
snprintf(buffer, bufferSize, "#%02d pc %08x %.*s (%.*s+%u)",
frameNumber, symbol->relative_pc, fieldWidth, mapName,
fieldWidth, symbolName, pc_offset);
} else {
snprintf(buffer, bufferSize, "#%02d pc %08x %.*s (%.*s)",
frameNumber, symbol->relative_pc, fieldWidth, mapName,
fieldWidth, symbolName);
}
} else {
snprintf(buffer, bufferSize, "#%02d pc %08x %.*s",
frameNumber, symbol->relative_pc, fieldWidth, mapName);
}
}
Reference in a new issue Copy permalink