linux/tools/perf/util/unwind-libdw.c

266 lines
5.6 KiB
C
Raw Normal View History

#include <linux/compiler.h>
#include <elfutils/libdw.h>
#include <elfutils/libdwfl.h>
#include <inttypes.h>
#include <errno.h>
#include "debug.h"
#include "unwind.h"
#include "unwind-libdw.h"
#include "machine.h"
#include "thread.h"
#include <linux/types.h>
#include "event.h"
#include "perf_regs.h"
#include "callchain.h"
#include "util.h"
static char *debuginfo_path;
static const Dwfl_Callbacks offline_callbacks = {
.find_debuginfo = dwfl_standard_find_debuginfo,
.debuginfo_path = &debuginfo_path,
.section_address = dwfl_offline_section_address,
};
static int __report_module(struct addr_location *al, u64 ip,
struct unwind_info *ui)
{
Dwfl_Module *mod;
struct dso *dso = NULL;
thread__find_addr_location(ui->thread,
PERF_RECORD_MISC_USER,
MAP__FUNCTION, ip, al);
if (al->map)
dso = al->map->dso;
if (!dso)
return 0;
mod = dwfl_addrmodule(ui->dwfl, ip);
perf report: Ensure the perf DSO mapping matches what libdw sees In some situations the libdw unwinder stopped working properly. I.e. with libunwind we see: ~~~~~ heaptrack_gui 2228 135073.400112: 641314 cycles: e8ed _dl_fixup (/usr/lib/ld-2.25.so) 15f06 _dl_runtime_resolve_sse_vex (/usr/lib/ld-2.25.so) ed94c KDynamicJobTracker::KDynamicJobTracker (/home/milian/projects/compiled/kf5/lib64/libKF5KIOWidgets.so.5.35.0) 608f3 _GLOBAL__sub_I_kdynamicjobtracker.cpp (/home/milian/projects/compiled/kf5/lib64/libKF5KIOWidgets.so.5.35.0) f199 call_init.part.0 (/usr/lib/ld-2.25.so) f2a5 _dl_init (/usr/lib/ld-2.25.so) db9 _dl_start_user (/usr/lib/ld-2.25.so) ~~~~~ But with libdw and without this patch this sample is not properly unwound: ~~~~~ heaptrack_gui 2228 135073.400112: 641314 cycles: e8ed _dl_fixup (/usr/lib/ld-2.25.so) 15f06 _dl_runtime_resolve_sse_vex (/usr/lib/ld-2.25.so) ed94c KDynamicJobTracker::KDynamicJobTracker (/home/milian/projects/compiled/kf5/lib64/libKF5KIOWidgets.so.5.35.0) ~~~~~ Debug output showed me that libdw found a module for the last frame address, but it thinks it belongs to /usr/lib/ld-2.25.so. This patch double-checks what libdw sees and what perf knows. If the mappings mismatch, we now report the elf known to perf. This fixes the situation above, and the libdw unwinder produces the same stack as libunwind. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/20170602143753.16907-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-06-02 16:37:52 +02:00
if (mod) {
Dwarf_Addr s;
dwfl_module_info(mod, NULL, &s, NULL, NULL, NULL, NULL, NULL);
if (s != al->map->start)
mod = 0;
}
if (!mod)
mod = dwfl_report_elf(ui->dwfl, dso->short_name,
dso->long_name, -1, al->map->start,
false);
return mod && dwfl_addrmodule(ui->dwfl, ip) == mod ? 0 : -1;
}
static int report_module(u64 ip, struct unwind_info *ui)
{
struct addr_location al;
return __report_module(&al, ip, ui);
}
/*
* Store all entries within entries array,
* we will process it after we finish unwind.
*/
static int entry(u64 ip, struct unwind_info *ui)
{
struct unwind_entry *e = &ui->entries[ui->idx++];
struct addr_location al;
if (__report_module(&al, ip, ui))
return -1;
perf unwind: Use addr_location::addr instead of ip for entries This fixes the srcline translation for call chains of user space applications. Before we got: perf report --stdio --no-children -s sym,srcline -g address 8.92% [.] main mandelbrot.h:41 | |--3.70%--main +8390240 | __libc_start_main +139950056726769 | _start +8388650 | |--2.74%--main +8390189 | --2.08%--main +8390296 __libc_start_main +139950056726769 _start +8388650 7.59% [.] main complex:1326 | |--4.79%--main +8390203 | __libc_start_main +139950056726769 | _start +8388650 | --2.80%--main +8390219 7.12% [.] __muldc3 libgcc2.c:1945 | |--3.76%--__muldc3 +139950060519490 | main +8390224 | __libc_start_main +139950056726769 | _start +8388650 | --3.32%--__muldc3 +139950060519512 main +8390224 With this patch applied, we instead get: perf report --stdio --no-children -s sym,srcline -g address 8.92% [.] main mandelbrot.h:41 | |--3.70%--main mandelbrot.h:41 | __libc_start_main +241 | _start +4194346 | |--2.74%--main mandelbrot.h:41 | --2.08%--main mandelbrot.h:41 __libc_start_main +241 _start +4194346 7.59% [.] main complex:1326 | |--4.79%--main complex:1326 | __libc_start_main +241 | _start +4194346 | --2.80%--main complex:1326 7.12% [.] __muldc3 libgcc2.c:1945 | |--3.76%--__muldc3 libgcc2.c:1945 | main mandelbrot.h:39 | __libc_start_main +241 | _start +4194346 | --3.32%--__muldc3 libgcc2.c:1945 main mandelbrot.h:39 Suggested-and-Acked-by: Namhyung Kim <namhyung@kernel.org> Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> LPU-Reference: 20160816153926.11288-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-08-16 17:39:26 +02:00
e->ip = al.addr;
e->map = al.map;
e->sym = al.sym;
pr_debug("unwind: %s:ip = 0x%" PRIx64 " (0x%" PRIx64 ")\n",
al.sym ? al.sym->name : "''",
ip,
al.map ? al.map->map_ip(al.map, ip) : (u64) 0);
return 0;
}
static pid_t next_thread(Dwfl *dwfl, void *arg, void **thread_argp)
{
/* We want only single thread to be processed. */
if (*thread_argp != NULL)
return 0;
*thread_argp = arg;
return dwfl_pid(dwfl);
}
static int access_dso_mem(struct unwind_info *ui, Dwarf_Addr addr,
Dwarf_Word *data)
{
struct addr_location al;
ssize_t size;
thread__find_addr_map(ui->thread, PERF_RECORD_MISC_USER,
MAP__FUNCTION, addr, &al);
if (!al.map) {
/*
* We've seen cases (softice) where DWARF unwinder went
* through non executable mmaps, which we need to lookup
* in MAP__VARIABLE tree.
*/
thread__find_addr_map(ui->thread, PERF_RECORD_MISC_USER,
MAP__VARIABLE, addr, &al);
}
if (!al.map) {
pr_debug("unwind: no map for %lx\n", (unsigned long)addr);
return -1;
}
if (!al.map->dso)
return -1;
size = dso__data_read_addr(al.map->dso, al.map, ui->machine,
addr, (u8 *) data, sizeof(*data));
return !(size == sizeof(*data));
}
static bool memory_read(Dwfl *dwfl __maybe_unused, Dwarf_Addr addr, Dwarf_Word *result,
void *arg)
{
struct unwind_info *ui = arg;
struct stack_dump *stack = &ui->sample->user_stack;
u64 start, end;
int offset;
int ret;
ret = perf_reg_value(&start, &ui->sample->user_regs, PERF_REG_SP);
if (ret)
return false;
end = start + stack->size;
/* Check overflow. */
if (addr + sizeof(Dwarf_Word) < addr)
return false;
if (addr < start || addr + sizeof(Dwarf_Word) > end) {
ret = access_dso_mem(ui, addr, result);
if (ret) {
pr_debug("unwind: access_mem 0x%" PRIx64 " not inside range"
" 0x%" PRIx64 "-0x%" PRIx64 "\n",
addr, start, end);
return false;
}
return true;
}
offset = addr - start;
*result = *(Dwarf_Word *)&stack->data[offset];
pr_debug("unwind: access_mem addr 0x%" PRIx64 ", val %lx, offset %d\n",
addr, (unsigned long)*result, offset);
return true;
}
static const Dwfl_Thread_Callbacks callbacks = {
.next_thread = next_thread,
.memory_read = memory_read,
.set_initial_registers = libdw__arch_set_initial_registers,
};
static int
frame_callback(Dwfl_Frame *state, void *arg)
{
struct unwind_info *ui = arg;
Dwarf_Addr pc;
perf report: Fix off-by-one for non-activation frames As the documentation for dwfl_frame_pc says, frames that are no activation frames need to have their program counter decremented by one to properly find the function of the caller. This fixes many cases where perf report currently attributes the cost to the next line. I.e. I have code like this: ~~~~~~~~~~~~~~~ #include <thread> #include <chrono> using namespace std; int main() { this_thread::sleep_for(chrono::milliseconds(1000)); this_thread::sleep_for(chrono::milliseconds(100)); this_thread::sleep_for(chrono::milliseconds(10)); return 0; } ~~~~~~~~~~~~~~~ Now compile and record it: ~~~~~~~~~~~~~~~ g++ -std=c++11 -g -O2 test.cpp echo 1 | sudo tee /proc/sys/kernel/sched_schedstats perf record \ --event sched:sched_stat_sleep \ --event sched:sched_process_exit \ --event sched:sched_switch --call-graph=dwarf \ --output perf.data.raw \ ./a.out echo 0 | sudo tee /proc/sys/kernel/sched_schedstats perf inject --sched-stat --input perf.data.raw --output perf.data ~~~~~~~~~~~~~~~ Before this patch, the report clearly shows the off-by-one issue. Most notably, the last sleep invocation is incorrectly attributed to the "return 0;" line: ~~~~~~~~~~~~~~~ Overhead Source:Line ........ ........... 100.00% core.c:0 | ---__schedule core.c:0 schedule do_nanosleep hrtimer.c:0 hrtimer_nanosleep sys_nanosleep entry_SYSCALL_64_fastpath .tmp_entry_64.o:0 __nanosleep_nocancel .:0 std::this_thread::sleep_for<long, std::ratio<1l, 1000l> > thread:323 | |--90.08%--main test.cpp:9 | __libc_start_main | _start | |--9.01%--main test.cpp:10 | __libc_start_main | _start | --0.91%--main test.cpp:13 __libc_start_main _start ~~~~~~~~~~~~~~~ With this patch here applied, the issue is fixed. The report becomes much more usable: ~~~~~~~~~~~~~~~ Overhead Source:Line ........ ........... 100.00% core.c:0 | ---__schedule core.c:0 schedule do_nanosleep hrtimer.c:0 hrtimer_nanosleep sys_nanosleep entry_SYSCALL_64_fastpath .tmp_entry_64.o:0 __nanosleep_nocancel .:0 std::this_thread::sleep_for<long, std::ratio<1l, 1000l> > thread:323 | |--90.08%--main test.cpp:8 | __libc_start_main | _start | |--9.01%--main test.cpp:9 | __libc_start_main | _start | --0.91%--main test.cpp:10 __libc_start_main _start ~~~~~~~~~~~~~~~ Similarly it works for signal frames: ~~~~~~~~~~~~~~~ __noinline void bar(void) { volatile long cnt = 0; for (cnt = 0; cnt < 100000000; cnt++); } __noinline void foo(void) { bar(); } void sig_handler(int sig) { foo(); } int main(void) { signal(SIGUSR1, sig_handler); raise(SIGUSR1); foo(); return 0; } ~~~~~~~~~~~~~~~~ Before, the report wrongly points to `signal.c:29` after raise(): ~~~~~~~~~~~~~~~~ $ perf report --stdio --no-children -g srcline -s srcline ... 100.00% signal.c:11 | ---bar signal.c:11 | |--50.49%--main signal.c:29 | __libc_start_main | _start | --49.51%--0x33a8f raise .:0 main signal.c:29 __libc_start_main _start ~~~~~~~~~~~~~~~~ With this patch in, the issue is fixed and we instead get: ~~~~~~~~~~~~~~~~ 100.00% signal signal [.] bar | ---bar signal.c:11 | |--50.49%--main signal.c:29 | __libc_start_main | _start | --49.51%--0x33a8f raise .:0 main signal.c:27 __libc_start_main _start ~~~~~~~~~~~~~~~~ Note how this patch fixes this issue for both unwinding methods, i.e. both dwfl and libunwind. The former case is straight-forward thanks to dwfl_frame_pc(). For libunwind, we replace the functionality via unw_is_signal_frame() for any but the very first frame. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Yao Jin <yao.jin@linux.intel.com> Cc: kernel-team@lge.com Link: http://lkml.kernel.org/r/20170524062129.32529-4-namhyung@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-05-24 08:21:25 +02:00
bool isactivation;
perf unwind: Report module before querying isactivation in dwfl unwind The PC returned by dwfl_frame_pc() may map into a not-yet-reported module. We have to report it before we continue unwinding. But when we query for the isactivation flag in dwfl_frame_pc, libdw will actually do one more unwinding step internally which can then break and lead to missed frames or broken stacks. With libunwind we get e.g.: ~~~~~ heaptrack_gui 2228 135073.400474: 613969 cycles: 108c8e [unknown] (/usr/lib/libQt5Core.so.5.8.0) 1093bc [unknown] (/usr/lib/libQt5Core.so.5.8.0) 109e7b QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0) 1470ff [unknown] (/usr/lib/libQt5Core.so.5.8.0) 147f67 QSystemLocale::query (/usr/lib/libQt5Core.so.5.8.0) 109fbf QLocalePrivate::updateSystemPrivate (/usr/lib/libQt5Core.so.5.8.0) 10aa27 QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0) 1e02c3 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 2113bb [unknown] (/usr/lib/libQt5Core.so.5.8.0) 211505 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 1b5df0 QFileInfo::exists (/usr/lib/libQt5Core.so.5.8.0) 92eb2 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 93423 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 93d2a QLibraryInfo::location (/usr/lib/libQt5Core.so.5.8.0) 2170af [unknown] (/usr/lib/libQt5Core.so.5.8.0) 297c53 QCoreApplicationPrivate::init (/usr/lib/libQt5Core.so.5.8.0) f7cde QGuiApplicationPrivate::init (/usr/lib/libQt5Gui.so.5.8.0) 1589e8 QApplicationPrivate::init (/usr/lib/libQt5Widgets.so.5.8.0) 78622 main (/home/milian/projects/compiled/other/bin/heaptrack_gui) 20439 __libc_start_main (/usr/lib/libc-2.25.so) 78299 _start (/home/milian/projects/compiled/other/bin/heaptrack_gui) heaptrack_gui 2228 135073.401156: 569521 cycles: 131633 QString::endsWith (/usr/lib/libQt5Core.so.5.8.0) 1a0701 QDir::cleanPath (/usr/lib/libQt5Core.so.5.8.0) 21b82d [unknown] (/usr/lib/libQt5Core.so.5.8.0) 1b3727 QFileInfo::canonicalFilePath (/usr/lib/libQt5Core.so.5.8.0) 2780c7 QFactoryLoader::update (/usr/lib/libQt5Core.so.5.8.0) 279525 QFactoryLoader::QFactoryLoader (/usr/lib/libQt5Core.so.5.8.0) e5bd0 QPlatformIntegrationFactory::create (/usr/lib/libQt5Gui.so.5.8.0) f5a1c QGuiApplicationPrivate::createPlatformIntegration (/usr/lib/libQt5Gui.so.5.8.0) f650c QGuiApplicationPrivate::createEventDispatcher (/usr/lib/libQt5Gui.so.5.8.0) 298524 QCoreApplicationPrivate::init (/usr/lib/libQt5Core.so.5.8.0) f7cde QGuiApplicationPrivate::init (/usr/lib/libQt5Gui.so.5.8.0) 1589e8 QApplicationPrivate::init (/usr/lib/libQt5Widgets.so.5.8.0) 78622 main (/home/milian/projects/compiled/other/bin/heaptrack_gui) 20439 __libc_start_main (/usr/lib/libc-2.25.so) 78299 _start (/home/milian/projects/compiled/other/bin/heaptrack_gui) ~~~~~ Note the two frames 1589e8 and 78622 in the first sample. These are missing when unwinding with libdw. The second sample's breakage is more obvious: ~~~~~ heaptrack_gui 2228 135073.400474: 613969 cycles: 108c8e [unknown] (/usr/lib/libQt5Core.so.5.8.0) 1093bc [unknown] (/usr/lib/libQt5Core.so.5.8.0) 109e7b QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0) 1470ff [unknown] (/usr/lib/libQt5Core.so.5.8.0) 147f67 QSystemLocale::query (/usr/lib/libQt5Core.so.5.8.0) 109fbf QLocalePrivate::updateSystemPrivate (/usr/lib/libQt5Core.so.5.8.0) 10aa27 QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0) 1e02c3 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 2113bb [unknown] (/usr/lib/libQt5Core.so.5.8.0) 211505 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 1b5df0 QFileInfo::exists (/usr/lib/libQt5Core.so.5.8.0) 92eb2 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 93423 [unknown] (/usr/lib/libQt5Core.so.5.8.0) 93d2a QLibraryInfo::location (/usr/lib/libQt5Core.so.5.8.0) 2170af [unknown] (/usr/lib/libQt5Core.so.5.8.0) 297c53 QCoreApplicationPrivate::init (/usr/lib/libQt5Core.so.5.8.0) f7cde QGuiApplicationPrivate::init (/usr/lib/libQt5Gui.so.5.8.0) 20439 __libc_start_main (/usr/lib/libc-2.25.so) 78299 _start (/home/milian/projects/compiled/other/bin/heaptrack_gui) heaptrack_gui 2228 135073.401156: 569521 cycles: 131633 QString::endsWith (/usr/lib/libQt5Core.so.5.8.0) 1a0701 QDir::cleanPath (/usr/lib/libQt5Core.so.5.8.0) 21b82d [unknown] (/usr/lib/libQt5Core.so.5.8.0) 1b3727 QFileInfo::canonicalFilePath (/usr/lib/libQt5Core.so.5.8.0) 2780c7 QFactoryLoader::update (/usr/lib/libQt5Core.so.5.8.0) 279525 QFactoryLoader::QFactoryLoader (/usr/lib/libQt5Core.so.5.8.0) e5bd0 QPlatformIntegrationFactory::create (/usr/lib/libQt5Gui.so.5.8.0) 723dbf [unknown] ([unknown]) ~~~~~ This patch fixes this issue and the libdw unwinder mimicks the libunwind behavior more closely. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Acked-by: Jan Kratochvil <jan.kratochvil@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/20170602143753.16907-2-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-06-02 16:37:53 +02:00
if (!dwfl_frame_pc(state, &pc, NULL)) {
pr_err("%s", dwfl_errmsg(-1));
return DWARF_CB_ABORT;
}
// report the module before we query for isactivation
report_module(pc, ui);
perf report: Fix off-by-one for non-activation frames As the documentation for dwfl_frame_pc says, frames that are no activation frames need to have their program counter decremented by one to properly find the function of the caller. This fixes many cases where perf report currently attributes the cost to the next line. I.e. I have code like this: ~~~~~~~~~~~~~~~ #include <thread> #include <chrono> using namespace std; int main() { this_thread::sleep_for(chrono::milliseconds(1000)); this_thread::sleep_for(chrono::milliseconds(100)); this_thread::sleep_for(chrono::milliseconds(10)); return 0; } ~~~~~~~~~~~~~~~ Now compile and record it: ~~~~~~~~~~~~~~~ g++ -std=c++11 -g -O2 test.cpp echo 1 | sudo tee /proc/sys/kernel/sched_schedstats perf record \ --event sched:sched_stat_sleep \ --event sched:sched_process_exit \ --event sched:sched_switch --call-graph=dwarf \ --output perf.data.raw \ ./a.out echo 0 | sudo tee /proc/sys/kernel/sched_schedstats perf inject --sched-stat --input perf.data.raw --output perf.data ~~~~~~~~~~~~~~~ Before this patch, the report clearly shows the off-by-one issue. Most notably, the last sleep invocation is incorrectly attributed to the "return 0;" line: ~~~~~~~~~~~~~~~ Overhead Source:Line ........ ........... 100.00% core.c:0 | ---__schedule core.c:0 schedule do_nanosleep hrtimer.c:0 hrtimer_nanosleep sys_nanosleep entry_SYSCALL_64_fastpath .tmp_entry_64.o:0 __nanosleep_nocancel .:0 std::this_thread::sleep_for<long, std::ratio<1l, 1000l> > thread:323 | |--90.08%--main test.cpp:9 | __libc_start_main | _start | |--9.01%--main test.cpp:10 | __libc_start_main | _start | --0.91%--main test.cpp:13 __libc_start_main _start ~~~~~~~~~~~~~~~ With this patch here applied, the issue is fixed. The report becomes much more usable: ~~~~~~~~~~~~~~~ Overhead Source:Line ........ ........... 100.00% core.c:0 | ---__schedule core.c:0 schedule do_nanosleep hrtimer.c:0 hrtimer_nanosleep sys_nanosleep entry_SYSCALL_64_fastpath .tmp_entry_64.o:0 __nanosleep_nocancel .:0 std::this_thread::sleep_for<long, std::ratio<1l, 1000l> > thread:323 | |--90.08%--main test.cpp:8 | __libc_start_main | _start | |--9.01%--main test.cpp:9 | __libc_start_main | _start | --0.91%--main test.cpp:10 __libc_start_main _start ~~~~~~~~~~~~~~~ Similarly it works for signal frames: ~~~~~~~~~~~~~~~ __noinline void bar(void) { volatile long cnt = 0; for (cnt = 0; cnt < 100000000; cnt++); } __noinline void foo(void) { bar(); } void sig_handler(int sig) { foo(); } int main(void) { signal(SIGUSR1, sig_handler); raise(SIGUSR1); foo(); return 0; } ~~~~~~~~~~~~~~~~ Before, the report wrongly points to `signal.c:29` after raise(): ~~~~~~~~~~~~~~~~ $ perf report --stdio --no-children -g srcline -s srcline ... 100.00% signal.c:11 | ---bar signal.c:11 | |--50.49%--main signal.c:29 | __libc_start_main | _start | --49.51%--0x33a8f raise .:0 main signal.c:29 __libc_start_main _start ~~~~~~~~~~~~~~~~ With this patch in, the issue is fixed and we instead get: ~~~~~~~~~~~~~~~~ 100.00% signal signal [.] bar | ---bar signal.c:11 | |--50.49%--main signal.c:29 | __libc_start_main | _start | --49.51%--0x33a8f raise .:0 main signal.c:27 __libc_start_main _start ~~~~~~~~~~~~~~~~ Note how this patch fixes this issue for both unwinding methods, i.e. both dwfl and libunwind. The former case is straight-forward thanks to dwfl_frame_pc(). For libunwind, we replace the functionality via unw_is_signal_frame() for any but the very first frame. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Yao Jin <yao.jin@linux.intel.com> Cc: kernel-team@lge.com Link: http://lkml.kernel.org/r/20170524062129.32529-4-namhyung@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-05-24 08:21:25 +02:00
if (!dwfl_frame_pc(state, &pc, &isactivation)) {
pr_err("%s", dwfl_errmsg(-1));
return DWARF_CB_ABORT;
}
perf report: Fix off-by-one for non-activation frames As the documentation for dwfl_frame_pc says, frames that are no activation frames need to have their program counter decremented by one to properly find the function of the caller. This fixes many cases where perf report currently attributes the cost to the next line. I.e. I have code like this: ~~~~~~~~~~~~~~~ #include <thread> #include <chrono> using namespace std; int main() { this_thread::sleep_for(chrono::milliseconds(1000)); this_thread::sleep_for(chrono::milliseconds(100)); this_thread::sleep_for(chrono::milliseconds(10)); return 0; } ~~~~~~~~~~~~~~~ Now compile and record it: ~~~~~~~~~~~~~~~ g++ -std=c++11 -g -O2 test.cpp echo 1 | sudo tee /proc/sys/kernel/sched_schedstats perf record \ --event sched:sched_stat_sleep \ --event sched:sched_process_exit \ --event sched:sched_switch --call-graph=dwarf \ --output perf.data.raw \ ./a.out echo 0 | sudo tee /proc/sys/kernel/sched_schedstats perf inject --sched-stat --input perf.data.raw --output perf.data ~~~~~~~~~~~~~~~ Before this patch, the report clearly shows the off-by-one issue. Most notably, the last sleep invocation is incorrectly attributed to the "return 0;" line: ~~~~~~~~~~~~~~~ Overhead Source:Line ........ ........... 100.00% core.c:0 | ---__schedule core.c:0 schedule do_nanosleep hrtimer.c:0 hrtimer_nanosleep sys_nanosleep entry_SYSCALL_64_fastpath .tmp_entry_64.o:0 __nanosleep_nocancel .:0 std::this_thread::sleep_for<long, std::ratio<1l, 1000l> > thread:323 | |--90.08%--main test.cpp:9 | __libc_start_main | _start | |--9.01%--main test.cpp:10 | __libc_start_main | _start | --0.91%--main test.cpp:13 __libc_start_main _start ~~~~~~~~~~~~~~~ With this patch here applied, the issue is fixed. The report becomes much more usable: ~~~~~~~~~~~~~~~ Overhead Source:Line ........ ........... 100.00% core.c:0 | ---__schedule core.c:0 schedule do_nanosleep hrtimer.c:0 hrtimer_nanosleep sys_nanosleep entry_SYSCALL_64_fastpath .tmp_entry_64.o:0 __nanosleep_nocancel .:0 std::this_thread::sleep_for<long, std::ratio<1l, 1000l> > thread:323 | |--90.08%--main test.cpp:8 | __libc_start_main | _start | |--9.01%--main test.cpp:9 | __libc_start_main | _start | --0.91%--main test.cpp:10 __libc_start_main _start ~~~~~~~~~~~~~~~ Similarly it works for signal frames: ~~~~~~~~~~~~~~~ __noinline void bar(void) { volatile long cnt = 0; for (cnt = 0; cnt < 100000000; cnt++); } __noinline void foo(void) { bar(); } void sig_handler(int sig) { foo(); } int main(void) { signal(SIGUSR1, sig_handler); raise(SIGUSR1); foo(); return 0; } ~~~~~~~~~~~~~~~~ Before, the report wrongly points to `signal.c:29` after raise(): ~~~~~~~~~~~~~~~~ $ perf report --stdio --no-children -g srcline -s srcline ... 100.00% signal.c:11 | ---bar signal.c:11 | |--50.49%--main signal.c:29 | __libc_start_main | _start | --49.51%--0x33a8f raise .:0 main signal.c:29 __libc_start_main _start ~~~~~~~~~~~~~~~~ With this patch in, the issue is fixed and we instead get: ~~~~~~~~~~~~~~~~ 100.00% signal signal [.] bar | ---bar signal.c:11 | |--50.49%--main signal.c:29 | __libc_start_main | _start | --49.51%--0x33a8f raise .:0 main signal.c:27 __libc_start_main _start ~~~~~~~~~~~~~~~~ Note how this patch fixes this issue for both unwinding methods, i.e. both dwfl and libunwind. The former case is straight-forward thanks to dwfl_frame_pc(). For libunwind, we replace the functionality via unw_is_signal_frame() for any but the very first frame. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Yao Jin <yao.jin@linux.intel.com> Cc: kernel-team@lge.com Link: http://lkml.kernel.org/r/20170524062129.32529-4-namhyung@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-05-24 08:21:25 +02:00
if (!isactivation)
--pc;
return entry(pc, ui) || !(--ui->max_stack) ?
DWARF_CB_ABORT : DWARF_CB_OK;
}
int unwind__get_entries(unwind_entry_cb_t cb, void *arg,
struct thread *thread,
struct perf_sample *data,
int max_stack)
{
struct unwind_info *ui, ui_buf = {
.sample = data,
.thread = thread,
.machine = thread->mg->machine,
.cb = cb,
.arg = arg,
.max_stack = max_stack,
};
Dwarf_Word ip;
int err = -EINVAL, i;
if (!data->user_regs.regs)
return -EINVAL;
ui = zalloc(sizeof(ui_buf) + sizeof(ui_buf.entries[0]) * max_stack);
if (!ui)
return -ENOMEM;
*ui = ui_buf;
ui->dwfl = dwfl_begin(&offline_callbacks);
if (!ui->dwfl)
goto out;
err = perf_reg_value(&ip, &data->user_regs, PERF_REG_IP);
if (err)
goto out;
err = report_module(ip, ui);
if (err)
goto out;
if (!dwfl_attach_state(ui->dwfl, EM_NONE, thread->tid, &callbacks, ui))
goto out;
err = dwfl_getthread_frames(ui->dwfl, thread->tid, frame_callback, ui);
perf report: Include partial stacks unwound with libdw So far the whole stack was thrown away when any error occurred before the maximum stack depth was unwound. This is actually a very common scenario though. The stacks that got unwound so far are still interesting. This removes a large chunk of differences when comparing perf script output for libunwind and libdw perf unwinding. E.g. with libunwind: ~~~~~ heaptrack_gui 2228 135073.388524: 479408 cycles: ffffffff811749ed perf_iterate_ctx ([kernel.kallsyms]) ffffffff81181662 perf_event_mmap ([kernel.kallsyms]) ffffffff811cf5ed mmap_region ([kernel.kallsyms]) ffffffff811cfe6b do_mmap ([kernel.kallsyms]) ffffffff811b0dca vm_mmap_pgoff ([kernel.kallsyms]) ffffffff811cdb0c sys_mmap_pgoff ([kernel.kallsyms]) ffffffff81033acb sys_mmap ([kernel.kallsyms]) ffffffff81631d37 entry_SYSCALL_64_fastpath ([kernel.kallsyms]) 192ca mmap64 (/usr/lib/ld-2.25.so) 59a9 _dl_map_object_from_fd (/usr/lib/ld-2.25.so) 83d0 _dl_map_object (/usr/lib/ld-2.25.so) cda1 openaux (/usr/lib/ld-2.25.so) 1834f _dl_catch_error (/usr/lib/ld-2.25.so) cfe2 _dl_map_object_deps (/usr/lib/ld-2.25.so) 3481 dl_main (/usr/lib/ld-2.25.so) 17387 _dl_sysdep_start (/usr/lib/ld-2.25.so) 4d37 _dl_start (/usr/lib/ld-2.25.so) d87 _start (/usr/lib/ld-2.25.so) heaptrack_gui 2228 135073.388677: 611329 cycles: 1a3e0 strcmp (/usr/lib/ld-2.25.so) 82b2 _dl_map_object (/usr/lib/ld-2.25.so) cda1 openaux (/usr/lib/ld-2.25.so) 1834f _dl_catch_error (/usr/lib/ld-2.25.so) cfe2 _dl_map_object_deps (/usr/lib/ld-2.25.so) 3481 dl_main (/usr/lib/ld-2.25.so) 17387 _dl_sysdep_start (/usr/lib/ld-2.25.so) 4d37 _dl_start (/usr/lib/ld-2.25.so) d87 _start (/usr/lib/ld-2.25.so) ~~~~~ With libdw without this patch: ~~~~~ heaptrack_gui 2228 135073.388524: 479408 cycles: ffffffff811749ed perf_iterate_ctx ([kernel.kallsyms]) ffffffff81181662 perf_event_mmap ([kernel.kallsyms]) ffffffff811cf5ed mmap_region ([kernel.kallsyms]) ffffffff811cfe6b do_mmap ([kernel.kallsyms]) ffffffff811b0dca vm_mmap_pgoff ([kernel.kallsyms]) ffffffff811cdb0c sys_mmap_pgoff ([kernel.kallsyms]) ffffffff81033acb sys_mmap ([kernel.kallsyms]) ffffffff81631d37 entry_SYSCALL_64_fastpath ([kernel.kallsyms]) heaptrack_gui 2228 135073.388677: 611329 cycles: ~~~~~ With this patch applied, the libdw unwinder will produce the same output as the libunwind unwinder. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/20170601210021.20046-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-06-01 23:00:21 +02:00
if (err && ui->max_stack != max_stack)
err = 0;
/*
* Display what we got based on the order setup.
*/
for (i = 0; i < ui->idx && !err; i++) {
int j = i;
if (callchain_param.order == ORDER_CALLER)
j = ui->idx - i - 1;
err = ui->entries[j].ip ? ui->cb(&ui->entries[j], ui->arg) : 0;
}
out:
if (err)
pr_debug("unwind: failed with '%s'\n", dwfl_errmsg(-1));
dwfl_end(ui->dwfl);
free(ui);
return 0;
}