linux/tools/perf/util/map.c

873 lines
19 KiB
C
Raw Normal View History

#include "symbol.h"
#include <errno.h>
#include <inttypes.h>
#include <limits.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include <unistd.h>
#include "map.h"
#include "thread.h"
#include "strlist.h"
perf tools: Back [vdso] DSO with real data Storing data for VDSO shared object, because we need it for the post unwind processing. The VDSO shared object is same for all process on a running system, so it makes no difference when we store it inside the tracer - perf. When [vdso] map memory is hit, we retrieve [vdso] DSO image and store it into temporary file. During the build-id processing phase, the [vdso] DSO image is stored in build-id db, and build-id reference is made inside perf.data. The build-id vdso file object is called '[vdso]'. We don't use temporary file name which gets removed when record is finished. During report phase the vdso build-id object is treated as any other build-id DSO object. Adding following API for vdso object: bool is_vdso_map(const char *filename) - returns true if the filename matches vdso map name struct dso *vdso__dso_findnew(struct list_head *head) - find/create proper vdso DSO object vdso__exit(void) - removes temporary VDSO image if there's any This change makes backtrace dwarf post unwind possible from [vdso] maps. Following output is current report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00007fff3ace89af | --- 0x7fff3ace89af Following output is new report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00000000000009af | --- 0x7fff3ace89af main __libc_start_main _start Signed-off-by: Jiri Olsa <jolsa@redhat.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1347295819-23177-5-git-send-email-jolsa@redhat.com [ committer note: s/ALIGN/PERF_ALIGN/g to cope with the android build changes ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 00:50:19 +08:00
#include "vdso.h"
#include "build-id.h"
perf script: Add an option to print the source line number Add field 'srcline' that displays the source file name and line number associated with the sample ip. The information displayed is the same as from addr2line. $ perf script -f comm,tid,pid,time,ip,sym,dso,symoff,srcline grep 10701/10701 2497321.421013: ffffffff81043ffa native_write_msr_safe+0xa ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/arch/x86/include/asm/msr.h:95 grep 10701/10701 2497321.421984: ffffffff8165b6b3 _raw_spin_lock+0x13 ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/arch/x86/include/asm/spinlock.h:54 grep 10701/10701 2497321.421990: ffffffff810b64b3 tick_sched_timer+0x53 ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/kernel/time/tick-sched.c:840 grep 10701/10701 2497321.421992: ffffffff8106f63f run_timer_softirq+0x2f ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/kernel/timer.c:1372 Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1386315778-11633-3-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-12-06 15:42:57 +08:00
#include "util.h"
#include "debug.h"
#include "machine.h"
#include <linux/string.h>
static void __maps__insert(struct maps *maps, struct map *map);
const char *map_type__name[MAP__NR_TYPES] = {
[MAP__FUNCTION] = "Functions",
[MAP__VARIABLE] = "Variables",
};
static inline int is_anon_memory(const char *filename)
{
return !strcmp(filename, "//anon") ||
!strncmp(filename, "/dev/zero", sizeof("/dev/zero") - 1) ||
!strncmp(filename, "/anon_hugepage", sizeof("/anon_hugepage") - 1);
}
static inline int is_no_dso_memory(const char *filename)
{
perf tools: Fix detection of stack area Output of /proc/<pid>/maps contains helpful information to anonymous mappings like stack, heap, ... For the case of stack, it can show multiple stack area for each thread in the process: $ cat /proc/$(pidof gnome-shell)/maps | grep stack 7fe019946000-7fe01a146000 rw-p 00000000 00:00 0 [stack:1624] 7fe040e32000-7fe041632000 rw-p 00000000 00:00 0 [stack:1451] 7fe041643000-7fe041e43000 rw-p 00000000 00:00 0 [stack:1450] 7fe04204b000-7fe04284b000 rw-p 00000000 00:00 0 [stack:1449] 7fe042a7e000-7fe04327e000 rw-p 00000000 00:00 0 [stack:1446] 7fe0432ff000-7fe043aff000 rw-p 00000000 00:00 0 [stack:1445] 7fe043b00000-7fe044300000 rw-p 00000000 00:00 0 [stack:1444] 7fe044301000-7fe044b01000 rw-p 00000000 00:00 0 [stack:1443] 7fe044b02000-7fe045302000 rw-p 00000000 00:00 0 [stack:1442] 7fe045303000-7fe045b03000 rw-p 00000000 00:00 0 [stack:1441] 7fe045b04000-7fe046304000 rw-p 00000000 00:00 0 [stack:1440] 7fe046305000-7fe046b05000 rw-p 00000000 00:00 0 [stack:1439] 7fe046b06000-7fe047306000 rw-p 00000000 00:00 0 [stack:1438] 7fff4b16f000-7fff4b190000 rw-p 00000000 00:00 0 [stack] However perf only knew about the main thread's. Fix it. Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Andi Kleen <ak@linux.intel.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1352273234-28912-4-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-11-07 15:27:11 +08:00
return !strncmp(filename, "[stack", 6) ||
!strncmp(filename, "/SYSV",5) ||
!strcmp(filename, "[heap]");
}
static inline int is_android_lib(const char *filename)
{
return !strncmp(filename, "/data/app-lib", 13) ||
!strncmp(filename, "/system/lib", 11);
}
static inline bool replace_android_lib(const char *filename, char *newfilename)
{
const char *libname;
char *app_abi;
size_t app_abi_length, new_length;
size_t lib_length = 0;
libname = strrchr(filename, '/');
if (libname)
lib_length = strlen(libname);
app_abi = getenv("APP_ABI");
if (!app_abi)
return false;
app_abi_length = strlen(app_abi);
if (!strncmp(filename, "/data/app-lib", 13)) {
char *apk_path;
if (!app_abi_length)
return false;
new_length = 7 + app_abi_length + lib_length;
apk_path = getenv("APK_PATH");
if (apk_path) {
new_length += strlen(apk_path) + 1;
if (new_length > PATH_MAX)
return false;
snprintf(newfilename, new_length,
"%s/libs/%s/%s", apk_path, app_abi, libname);
} else {
if (new_length > PATH_MAX)
return false;
snprintf(newfilename, new_length,
"libs/%s/%s", app_abi, libname);
}
return true;
}
if (!strncmp(filename, "/system/lib/", 11)) {
char *ndk, *app;
const char *arch;
size_t ndk_length;
size_t app_length;
ndk = getenv("NDK_ROOT");
app = getenv("APP_PLATFORM");
if (!(ndk && app))
return false;
ndk_length = strlen(ndk);
app_length = strlen(app);
if (!(ndk_length && app_length && app_abi_length))
return false;
arch = !strncmp(app_abi, "arm", 3) ? "arm" :
!strncmp(app_abi, "mips", 4) ? "mips" :
!strncmp(app_abi, "x86", 3) ? "x86" : NULL;
if (!arch)
return false;
new_length = 27 + ndk_length +
app_length + lib_length
+ strlen(arch);
if (new_length > PATH_MAX)
return false;
snprintf(newfilename, new_length,
"%s/platforms/%s/arch-%s/usr/lib/%s",
ndk, app, arch, libname);
return true;
}
return false;
}
void map__init(struct map *map, enum map_type type,
u64 start, u64 end, u64 pgoff, struct dso *dso)
{
map->type = type;
map->start = start;
map->end = end;
map->pgoff = pgoff;
map->reloc = 0;
map->dso = dso__get(dso);
map->map_ip = map__map_ip;
map->unmap_ip = map__unmap_ip;
RB_CLEAR_NODE(&map->rb_node);
map->groups = NULL;
map->erange_warned = false;
atomic_set(&map->refcnt, 1);
}
struct map *map__new(struct machine *machine, u64 start, u64 len,
u64 pgoff, u32 pid, u32 d_maj, u32 d_min, u64 ino,
u64 ino_gen, u32 prot, u32 flags, char *filename,
enum map_type type, struct thread *thread)
{
struct map *map = malloc(sizeof(*map));
if (map != NULL) {
char newfilename[PATH_MAX];
struct dso *dso;
int anon, no_dso, vdso, android;
android = is_android_lib(filename);
anon = is_anon_memory(filename);
perf tools: Back [vdso] DSO with real data Storing data for VDSO shared object, because we need it for the post unwind processing. The VDSO shared object is same for all process on a running system, so it makes no difference when we store it inside the tracer - perf. When [vdso] map memory is hit, we retrieve [vdso] DSO image and store it into temporary file. During the build-id processing phase, the [vdso] DSO image is stored in build-id db, and build-id reference is made inside perf.data. The build-id vdso file object is called '[vdso]'. We don't use temporary file name which gets removed when record is finished. During report phase the vdso build-id object is treated as any other build-id DSO object. Adding following API for vdso object: bool is_vdso_map(const char *filename) - returns true if the filename matches vdso map name struct dso *vdso__dso_findnew(struct list_head *head) - find/create proper vdso DSO object vdso__exit(void) - removes temporary VDSO image if there's any This change makes backtrace dwarf post unwind possible from [vdso] maps. Following output is current report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00007fff3ace89af | --- 0x7fff3ace89af Following output is new report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00000000000009af | --- 0x7fff3ace89af main __libc_start_main _start Signed-off-by: Jiri Olsa <jolsa@redhat.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1347295819-23177-5-git-send-email-jolsa@redhat.com [ committer note: s/ALIGN/PERF_ALIGN/g to cope with the android build changes ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 00:50:19 +08:00
vdso = is_vdso_map(filename);
no_dso = is_no_dso_memory(filename);
map->maj = d_maj;
map->min = d_min;
map->ino = ino;
map->ino_generation = ino_gen;
map->prot = prot;
map->flags = flags;
if ((anon || no_dso) && type == MAP__FUNCTION) {
snprintf(newfilename, sizeof(newfilename), "/tmp/perf-%d.map", pid);
filename = newfilename;
}
if (android) {
if (replace_android_lib(filename, newfilename))
filename = newfilename;
}
perf tools: Back [vdso] DSO with real data Storing data for VDSO shared object, because we need it for the post unwind processing. The VDSO shared object is same for all process on a running system, so it makes no difference when we store it inside the tracer - perf. When [vdso] map memory is hit, we retrieve [vdso] DSO image and store it into temporary file. During the build-id processing phase, the [vdso] DSO image is stored in build-id db, and build-id reference is made inside perf.data. The build-id vdso file object is called '[vdso]'. We don't use temporary file name which gets removed when record is finished. During report phase the vdso build-id object is treated as any other build-id DSO object. Adding following API for vdso object: bool is_vdso_map(const char *filename) - returns true if the filename matches vdso map name struct dso *vdso__dso_findnew(struct list_head *head) - find/create proper vdso DSO object vdso__exit(void) - removes temporary VDSO image if there's any This change makes backtrace dwarf post unwind possible from [vdso] maps. Following output is current report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00007fff3ace89af | --- 0x7fff3ace89af Following output is new report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00000000000009af | --- 0x7fff3ace89af main __libc_start_main _start Signed-off-by: Jiri Olsa <jolsa@redhat.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1347295819-23177-5-git-send-email-jolsa@redhat.com [ committer note: s/ALIGN/PERF_ALIGN/g to cope with the android build changes ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 00:50:19 +08:00
if (vdso) {
pgoff = 0;
dso = machine__findnew_vdso(machine, thread);
perf tools: Back [vdso] DSO with real data Storing data for VDSO shared object, because we need it for the post unwind processing. The VDSO shared object is same for all process on a running system, so it makes no difference when we store it inside the tracer - perf. When [vdso] map memory is hit, we retrieve [vdso] DSO image and store it into temporary file. During the build-id processing phase, the [vdso] DSO image is stored in build-id db, and build-id reference is made inside perf.data. The build-id vdso file object is called '[vdso]'. We don't use temporary file name which gets removed when record is finished. During report phase the vdso build-id object is treated as any other build-id DSO object. Adding following API for vdso object: bool is_vdso_map(const char *filename) - returns true if the filename matches vdso map name struct dso *vdso__dso_findnew(struct list_head *head) - find/create proper vdso DSO object vdso__exit(void) - removes temporary VDSO image if there's any This change makes backtrace dwarf post unwind possible from [vdso] maps. Following output is current report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00007fff3ace89af | --- 0x7fff3ace89af Following output is new report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00000000000009af | --- 0x7fff3ace89af main __libc_start_main _start Signed-off-by: Jiri Olsa <jolsa@redhat.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1347295819-23177-5-git-send-email-jolsa@redhat.com [ committer note: s/ALIGN/PERF_ALIGN/g to cope with the android build changes ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 00:50:19 +08:00
} else
dso = machine__findnew_dso(machine, filename);
perf tools: Back [vdso] DSO with real data Storing data for VDSO shared object, because we need it for the post unwind processing. The VDSO shared object is same for all process on a running system, so it makes no difference when we store it inside the tracer - perf. When [vdso] map memory is hit, we retrieve [vdso] DSO image and store it into temporary file. During the build-id processing phase, the [vdso] DSO image is stored in build-id db, and build-id reference is made inside perf.data. The build-id vdso file object is called '[vdso]'. We don't use temporary file name which gets removed when record is finished. During report phase the vdso build-id object is treated as any other build-id DSO object. Adding following API for vdso object: bool is_vdso_map(const char *filename) - returns true if the filename matches vdso map name struct dso *vdso__dso_findnew(struct list_head *head) - find/create proper vdso DSO object vdso__exit(void) - removes temporary VDSO image if there's any This change makes backtrace dwarf post unwind possible from [vdso] maps. Following output is current report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00007fff3ace89af | --- 0x7fff3ace89af Following output is new report of [vdso] sample dwarf backtrace: # Overhead Command Shared Object Symbol # ........ ....... ................. ............................. # 99.52% ex [vdso] [.] 0x00000000000009af | --- 0x7fff3ace89af main __libc_start_main _start Signed-off-by: Jiri Olsa <jolsa@redhat.com> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1347295819-23177-5-git-send-email-jolsa@redhat.com [ committer note: s/ALIGN/PERF_ALIGN/g to cope with the android build changes ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 00:50:19 +08:00
if (dso == NULL)
goto out_delete;
map__init(map, type, start, start + len, pgoff, dso);
if (anon || no_dso) {
map->map_ip = map->unmap_ip = identity__map_ip;
/*
* Set memory without DSO as loaded. All map__find_*
* functions still return NULL, and we avoid the
* unnecessary map__load warning.
*/
if (type != MAP__FUNCTION)
dso__set_loaded(dso, map->type);
}
dso__put(dso);
}
return map;
out_delete:
free(map);
return NULL;
}
/*
* Constructor variant for modules (where we know from /proc/modules where
* they are loaded) and for vmlinux, where only after we load all the
* symbols we'll know where it starts and ends.
*/
struct map *map__new2(u64 start, struct dso *dso, enum map_type type)
{
struct map *map = calloc(1, (sizeof(*map) +
(dso->kernel ? sizeof(struct kmap) : 0)));
if (map != NULL) {
/*
* ->end will be filled after we load all the symbols
*/
map__init(map, type, start, 0, 0, dso);
}
return map;
}
/*
* Use this and __map__is_kmodule() for map instances that are in
* machine->kmaps, and thus have map->groups->machine all properly set, to
* disambiguate between the kernel and modules.
*
* When the need arises, introduce map__is_{kernel,kmodule)() that
* checks (map->groups != NULL && map->groups->machine != NULL &&
* map->dso->kernel) before calling __map__is_{kernel,kmodule}())
*/
bool __map__is_kernel(const struct map *map)
{
return __machine__kernel_map(map->groups->machine, map->type) == map;
}
static void map__exit(struct map *map)
{
BUG_ON(!RB_EMPTY_NODE(&map->rb_node));
dso__zput(map->dso);
}
void map__delete(struct map *map)
{
map__exit(map);
free(map);
}
void map__put(struct map *map)
{
if (map && atomic_dec_and_test(&map->refcnt))
map__delete(map);
}
void map__fixup_start(struct map *map)
{
struct rb_root *symbols = &map->dso->symbols[map->type];
struct rb_node *nd = rb_first(symbols);
if (nd != NULL) {
struct symbol *sym = rb_entry(nd, struct symbol, rb_node);
map->start = sym->start;
}
}
void map__fixup_end(struct map *map)
{
struct rb_root *symbols = &map->dso->symbols[map->type];
struct rb_node *nd = rb_last(symbols);
if (nd != NULL) {
struct symbol *sym = rb_entry(nd, struct symbol, rb_node);
map->end = sym->end;
}
}
#define DSO__DELETED "(deleted)"
int map__load(struct map *map, symbol_filter_t filter)
{
const char *name = map->dso->long_name;
int nr;
if (dso__loaded(map->dso, map->type))
return 0;
nr = dso__load(map->dso, map, filter);
if (nr < 0) {
if (map->dso->has_build_id) {
char sbuild_id[BUILD_ID_SIZE * 2 + 1];
build_id__sprintf(map->dso->build_id,
sizeof(map->dso->build_id),
sbuild_id);
pr_warning("%s with build id %s not found",
name, sbuild_id);
} else
pr_warning("Failed to open %s", name);
pr_warning(", continuing without symbols\n");
return -1;
} else if (nr == 0) {
#ifdef HAVE_LIBELF_SUPPORT
const size_t len = strlen(name);
const size_t real_len = len - sizeof(DSO__DELETED);
if (len > sizeof(DSO__DELETED) &&
strcmp(name + real_len + 1, DSO__DELETED) == 0) {
pr_warning("%.*s was updated (is prelink enabled?). "
"Restart the long running apps that use it!\n",
(int)real_len, name);
} else {
pr_warning("no symbols found in %s, maybe install "
"a debug package?\n", name);
}
#endif
return -1;
}
return 0;
}
int __weak arch__compare_symbol_names(const char *namea, const char *nameb)
{
return strcmp(namea, nameb);
}
struct symbol *map__find_symbol(struct map *map, u64 addr,
symbol_filter_t filter)
{
if (map__load(map, filter) < 0)
return NULL;
return dso__find_symbol(map->dso, map->type, addr);
}
struct symbol *map__find_symbol_by_name(struct map *map, const char *name,
symbol_filter_t filter)
{
if (map__load(map, filter) < 0)
return NULL;
if (!dso__sorted_by_name(map->dso, map->type))
dso__sort_by_name(map->dso, map->type);
return dso__find_symbol_by_name(map->dso, map->type, name);
}
struct map *map__clone(struct map *from)
{
struct map *map = memdup(from, sizeof(*map));
if (map != NULL) {
atomic_set(&map->refcnt, 1);
RB_CLEAR_NODE(&map->rb_node);
dso__get(map->dso);
map->groups = NULL;
}
return map;
}
int map__overlap(struct map *l, struct map *r)
{
if (l->start > r->start) {
struct map *t = l;
l = r;
r = t;
}
if (l->end > r->start)
return 1;
return 0;
}
size_t map__fprintf(struct map *map, FILE *fp)
{
return fprintf(fp, " %" PRIx64 "-%" PRIx64 " %" PRIx64 " %s\n",
map->start, map->end, map->pgoff, map->dso->name);
}
perf annotate: Fix it for non-prelinked *.so The problem was we were incorrectly calculating objdump addresses for sym->start and sym->end, look: For simple ET_DYN type DSO (*.so) with one function, objdump -dS output is something like this: 000004ac <my_strlen>: int my_strlen(const char *s) 4ac: 55 push %ebp 4ad: 89 e5 mov %esp,%ebp 4af: 83 ec 10 sub $0x10,%esp { i.e. we have relative-to-dso-mapping IPs (=RIP) there. For ET_EXEC type and probably for prelinked libs as well (sorry can't test - I don't use prelink) objdump outputs absolute IPs, e.g. 08048604 <zz_strlen>: extern "C" int zz_strlen(const char *s) 8048604: 55 push %ebp 8048605: 89 e5 mov %esp,%ebp 8048607: 83 ec 10 sub $0x10,%esp { So, if sym->start is always relative to dso mapping(*), we'll have to unmap it for ET_EXEC like cases, and leave as is for ET_DYN cases. (*) and it is - we've explicitely made it relative. Look for adjust_symbols handling in dso__load_sym() Previously we were always unmapping sym->start and for ET_DYN dsos resulting addresses were wrong, and so objdump output was empty. The end result was that perf annotate output for symbols from non-prelinked *.so had always 0.00% percents only, which is wrong. To fix it, let's introduce a helper for converting rip to objdump address, and also let's document what map_ip() and unmap_ip() do -- I had to study sources for several hours to understand it. Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1265223128-11786-8-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-02-04 02:52:07 +08:00
size_t map__fprintf_dsoname(struct map *map, FILE *fp)
{
const char *dsoname = "[unknown]";
if (map && map->dso && (map->dso->name || map->dso->long_name)) {
if (symbol_conf.show_kernel_path && map->dso->long_name)
dsoname = map->dso->long_name;
else if (map->dso->name)
dsoname = map->dso->name;
}
return fprintf(fp, "%s", dsoname);
}
perf script: Add an option to print the source line number Add field 'srcline' that displays the source file name and line number associated with the sample ip. The information displayed is the same as from addr2line. $ perf script -f comm,tid,pid,time,ip,sym,dso,symoff,srcline grep 10701/10701 2497321.421013: ffffffff81043ffa native_write_msr_safe+0xa ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/arch/x86/include/asm/msr.h:95 grep 10701/10701 2497321.421984: ffffffff8165b6b3 _raw_spin_lock+0x13 ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/arch/x86/include/asm/spinlock.h:54 grep 10701/10701 2497321.421990: ffffffff810b64b3 tick_sched_timer+0x53 ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/kernel/time/tick-sched.c:840 grep 10701/10701 2497321.421992: ffffffff8106f63f run_timer_softirq+0x2f ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/kernel/timer.c:1372 Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1386315778-11633-3-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-12-06 15:42:57 +08:00
int map__fprintf_srcline(struct map *map, u64 addr, const char *prefix,
FILE *fp)
{
char *srcline;
int ret = 0;
if (map && map->dso) {
srcline = get_srcline(map->dso,
map__rip_2objdump(map, addr), NULL, true);
perf script: Add an option to print the source line number Add field 'srcline' that displays the source file name and line number associated with the sample ip. The information displayed is the same as from addr2line. $ perf script -f comm,tid,pid,time,ip,sym,dso,symoff,srcline grep 10701/10701 2497321.421013: ffffffff81043ffa native_write_msr_safe+0xa ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/arch/x86/include/asm/msr.h:95 grep 10701/10701 2497321.421984: ffffffff8165b6b3 _raw_spin_lock+0x13 ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/arch/x86/include/asm/spinlock.h:54 grep 10701/10701 2497321.421990: ffffffff810b64b3 tick_sched_timer+0x53 ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/kernel/time/tick-sched.c:840 grep 10701/10701 2497321.421992: ffffffff8106f63f run_timer_softirq+0x2f ([kernel.kallsyms]) /usr/src/debug/kernel-3.9.fc17/linux-3.9.10-100.fc17.x86_64/kernel/timer.c:1372 Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1386315778-11633-3-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-12-06 15:42:57 +08:00
if (srcline != SRCLINE_UNKNOWN)
ret = fprintf(fp, "%s%s", prefix, srcline);
free_srcline(srcline);
}
return ret;
}
/**
* map__rip_2objdump - convert symbol start address to objdump address.
* @map: memory map
* @rip: symbol start address
*
perf annotate: Fix it for non-prelinked *.so The problem was we were incorrectly calculating objdump addresses for sym->start and sym->end, look: For simple ET_DYN type DSO (*.so) with one function, objdump -dS output is something like this: 000004ac <my_strlen>: int my_strlen(const char *s) 4ac: 55 push %ebp 4ad: 89 e5 mov %esp,%ebp 4af: 83 ec 10 sub $0x10,%esp { i.e. we have relative-to-dso-mapping IPs (=RIP) there. For ET_EXEC type and probably for prelinked libs as well (sorry can't test - I don't use prelink) objdump outputs absolute IPs, e.g. 08048604 <zz_strlen>: extern "C" int zz_strlen(const char *s) 8048604: 55 push %ebp 8048605: 89 e5 mov %esp,%ebp 8048607: 83 ec 10 sub $0x10,%esp { So, if sym->start is always relative to dso mapping(*), we'll have to unmap it for ET_EXEC like cases, and leave as is for ET_DYN cases. (*) and it is - we've explicitely made it relative. Look for adjust_symbols handling in dso__load_sym() Previously we were always unmapping sym->start and for ET_DYN dsos resulting addresses were wrong, and so objdump output was empty. The end result was that perf annotate output for symbols from non-prelinked *.so had always 0.00% percents only, which is wrong. To fix it, let's introduce a helper for converting rip to objdump address, and also let's document what map_ip() and unmap_ip() do -- I had to study sources for several hours to understand it. Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1265223128-11786-8-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-02-04 02:52:07 +08:00
* objdump wants/reports absolute IPs for ET_EXEC, and RIPs for ET_DYN.
* map->dso->adjust_symbols==1 for ET_EXEC-like cases except ET_REL which is
* relative to section start.
*
* Return: Address suitable for passing to "objdump --start-address="
perf annotate: Fix it for non-prelinked *.so The problem was we were incorrectly calculating objdump addresses for sym->start and sym->end, look: For simple ET_DYN type DSO (*.so) with one function, objdump -dS output is something like this: 000004ac <my_strlen>: int my_strlen(const char *s) 4ac: 55 push %ebp 4ad: 89 e5 mov %esp,%ebp 4af: 83 ec 10 sub $0x10,%esp { i.e. we have relative-to-dso-mapping IPs (=RIP) there. For ET_EXEC type and probably for prelinked libs as well (sorry can't test - I don't use prelink) objdump outputs absolute IPs, e.g. 08048604 <zz_strlen>: extern "C" int zz_strlen(const char *s) 8048604: 55 push %ebp 8048605: 89 e5 mov %esp,%ebp 8048607: 83 ec 10 sub $0x10,%esp { So, if sym->start is always relative to dso mapping(*), we'll have to unmap it for ET_EXEC like cases, and leave as is for ET_DYN cases. (*) and it is - we've explicitely made it relative. Look for adjust_symbols handling in dso__load_sym() Previously we were always unmapping sym->start and for ET_DYN dsos resulting addresses were wrong, and so objdump output was empty. The end result was that perf annotate output for symbols from non-prelinked *.so had always 0.00% percents only, which is wrong. To fix it, let's introduce a helper for converting rip to objdump address, and also let's document what map_ip() and unmap_ip() do -- I had to study sources for several hours to understand it. Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1265223128-11786-8-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-02-04 02:52:07 +08:00
*/
u64 map__rip_2objdump(struct map *map, u64 rip)
{
if (!map->dso->adjust_symbols)
return rip;
if (map->dso->rel)
return rip - map->pgoff;
/*
* kernel modules also have DSO_TYPE_USER in dso->kernel,
* but all kernel modules are ET_REL, so won't get here.
*/
if (map->dso->kernel == DSO_TYPE_USER)
return rip + map->dso->text_offset;
return map->unmap_ip(map, rip) - map->reloc;
perf annotate: Fix it for non-prelinked *.so The problem was we were incorrectly calculating objdump addresses for sym->start and sym->end, look: For simple ET_DYN type DSO (*.so) with one function, objdump -dS output is something like this: 000004ac <my_strlen>: int my_strlen(const char *s) 4ac: 55 push %ebp 4ad: 89 e5 mov %esp,%ebp 4af: 83 ec 10 sub $0x10,%esp { i.e. we have relative-to-dso-mapping IPs (=RIP) there. For ET_EXEC type and probably for prelinked libs as well (sorry can't test - I don't use prelink) objdump outputs absolute IPs, e.g. 08048604 <zz_strlen>: extern "C" int zz_strlen(const char *s) 8048604: 55 push %ebp 8048605: 89 e5 mov %esp,%ebp 8048607: 83 ec 10 sub $0x10,%esp { So, if sym->start is always relative to dso mapping(*), we'll have to unmap it for ET_EXEC like cases, and leave as is for ET_DYN cases. (*) and it is - we've explicitely made it relative. Look for adjust_symbols handling in dso__load_sym() Previously we were always unmapping sym->start and for ET_DYN dsos resulting addresses were wrong, and so objdump output was empty. The end result was that perf annotate output for symbols from non-prelinked *.so had always 0.00% percents only, which is wrong. To fix it, let's introduce a helper for converting rip to objdump address, and also let's document what map_ip() and unmap_ip() do -- I had to study sources for several hours to understand it. Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1265223128-11786-8-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-02-04 02:52:07 +08:00
}
perf top: Fix annotate for userspace First, for programs and prelinked libraries, annotate code was fooled by objdump output IPs (src->eip in the code) being wrongly converted to absolute IPs. In such case there were no conversion needed, but in src->eip = strtoull(src->line, NULL, 16); src->eip = map->unmap_ip(map, src->eip); // = eip + map->start - map->pgoff we were reading absolute address from objdump (e.g. 8048604) and then almost doubling it, because eip & map->start are approximately close for small programs. Needless to say, that later, in record_precise_ip() there was no matching with real runtime IPs. And second, like with `perf annotate` the problem with non-prelinked *.so was that we were doing rip -> objdump address conversion wrong. Also, because unlike `perf annotate`, `perf top` code does annotation based on absolute IPs for performance reasons(*), new helper for mapping objdump addresse to IP is introduced. (*) we get samples info in absolute IPs, and since we do lots of hit-testing on absolute IPs at runtime in record_precise_ip(), it's better to convert objdump addresses to IPs once and do no conversion at runtime. I also had to fix how objdump output is parsed (with hardcoded 8/16 characters format, which was inappropriate for ET_DYN dsos with small addresses like '4ac') Also note, that not all objdump output lines has associtated IPs, e.g. look at source lines here: 000004ac <my_strlen>: extern "C" int my_strlen(const char *s) 4ac: 55 push %ebp 4ad: 89 e5 mov %esp,%ebp 4af: 83 ec 10 sub $0x10,%esp { int len = 0; 4b2: c7 45 fc 00 00 00 00 movl $0x0,-0x4(%ebp) 4b9: eb 08 jmp 4c3 <my_strlen+0x17> while (*s) { ++len; 4bb: 83 45 fc 01 addl $0x1,-0x4(%ebp) ++s; 4bf: 83 45 08 01 addl $0x1,0x8(%ebp) So we mark them with eip=0, and ignore such lines in annotate lookup code. Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> [ Note: one hunk of this patch was applied by Mike in 57d8188 ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <1265550376-12665-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-02-07 21:46:15 +08:00
/**
* map__objdump_2mem - convert objdump address to a memory address.
* @map: memory map
* @ip: objdump address
*
* Closely related to map__rip_2objdump(), this function takes an address from
* objdump and converts it to a memory address. Note this assumes that @map
* contains the address. To be sure the result is valid, check it forwards
* e.g. map__rip_2objdump(map->map_ip(map, map__objdump_2mem(map, ip))) == ip
*
* Return: Memory address.
*/
u64 map__objdump_2mem(struct map *map, u64 ip)
{
if (!map->dso->adjust_symbols)
return map->unmap_ip(map, ip);
if (map->dso->rel)
return map->unmap_ip(map, ip + map->pgoff);
/*
* kernel modules also have DSO_TYPE_USER in dso->kernel,
* but all kernel modules are ET_REL, so won't get here.
*/
if (map->dso->kernel == DSO_TYPE_USER)
return map->unmap_ip(map, ip - map->dso->text_offset);
return ip + map->reloc;
}
static void maps__init(struct maps *maps)
{
maps->entries = RB_ROOT;
pthread_rwlock_init(&maps->lock, NULL);
}
void map_groups__init(struct map_groups *mg, struct machine *machine)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i) {
maps__init(&mg->maps[i]);
}
mg->machine = machine;
atomic_set(&mg->refcnt, 1);
}
static void __maps__purge(struct maps *maps)
{
struct rb_root *root = &maps->entries;
struct rb_node *next = rb_first(root);
while (next) {
struct map *pos = rb_entry(next, struct map, rb_node);
next = rb_next(&pos->rb_node);
rb_erase_init(&pos->rb_node, root);
map__put(pos);
}
}
static void maps__exit(struct maps *maps)
{
pthread_rwlock_wrlock(&maps->lock);
__maps__purge(maps);
pthread_rwlock_unlock(&maps->lock);
}
void map_groups__exit(struct map_groups *mg)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i)
maps__exit(&mg->maps[i]);
}
bool map_groups__empty(struct map_groups *mg)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i) {
if (maps__first(&mg->maps[i]))
return false;
}
return true;
}
struct map_groups *map_groups__new(struct machine *machine)
{
struct map_groups *mg = malloc(sizeof(*mg));
if (mg != NULL)
map_groups__init(mg, machine);
return mg;
}
void map_groups__delete(struct map_groups *mg)
{
map_groups__exit(mg);
free(mg);
}
void map_groups__put(struct map_groups *mg)
{
if (mg && atomic_dec_and_test(&mg->refcnt))
map_groups__delete(mg);
}
struct symbol *map_groups__find_symbol(struct map_groups *mg,
enum map_type type, u64 addr,
struct map **mapp,
symbol_filter_t filter)
{
struct map *map = map_groups__find(mg, type, addr);
/* Ensure map is loaded before using map->map_ip */
if (map != NULL && map__load(map, filter) >= 0) {
if (mapp != NULL)
*mapp = map;
return map__find_symbol(map, map->map_ip(map, addr), filter);
}
return NULL;
}
struct symbol *maps__find_symbol_by_name(struct maps *maps, const char *name,
struct map **mapp, symbol_filter_t filter)
{
struct symbol *sym;
struct rb_node *nd;
pthread_rwlock_rdlock(&maps->lock);
for (nd = rb_first(&maps->entries); nd; nd = rb_next(nd)) {
struct map *pos = rb_entry(nd, struct map, rb_node);
sym = map__find_symbol_by_name(pos, name, filter);
if (sym == NULL)
continue;
if (mapp != NULL)
*mapp = pos;
goto out;
}
sym = NULL;
out:
pthread_rwlock_unlock(&maps->lock);
return sym;
}
struct symbol *map_groups__find_symbol_by_name(struct map_groups *mg,
enum map_type type,
const char *name,
struct map **mapp,
symbol_filter_t filter)
{
struct symbol *sym = maps__find_symbol_by_name(&mg->maps[type], name, mapp, filter);
return sym;
}
int map_groups__find_ams(struct addr_map_symbol *ams, symbol_filter_t filter)
{
if (ams->addr < ams->map->start || ams->addr >= ams->map->end) {
if (ams->map->groups == NULL)
return -1;
ams->map = map_groups__find(ams->map->groups, ams->map->type,
ams->addr);
if (ams->map == NULL)
return -1;
}
ams->al_addr = ams->map->map_ip(ams->map, ams->addr);
ams->sym = map__find_symbol(ams->map, ams->al_addr, filter);
return ams->sym ? 0 : -1;
}
static size_t maps__fprintf(struct maps *maps, FILE *fp)
{
size_t printed = 0;
struct rb_node *nd;
pthread_rwlock_rdlock(&maps->lock);
for (nd = rb_first(&maps->entries); nd; nd = rb_next(nd)) {
struct map *pos = rb_entry(nd, struct map, rb_node);
printed += fprintf(fp, "Map:");
printed += map__fprintf(pos, fp);
if (verbose > 2) {
printed += dso__fprintf(pos->dso, pos->type, fp);
printed += fprintf(fp, "--\n");
}
}
pthread_rwlock_unlock(&maps->lock);
return printed;
}
size_t __map_groups__fprintf_maps(struct map_groups *mg, enum map_type type,
FILE *fp)
{
size_t printed = fprintf(fp, "%s:\n", map_type__name[type]);
return printed += maps__fprintf(&mg->maps[type], fp);
}
size_t map_groups__fprintf(struct map_groups *mg, FILE *fp)
{
size_t printed = 0, i;
for (i = 0; i < MAP__NR_TYPES; ++i)
printed += __map_groups__fprintf_maps(mg, i, fp);
return printed;
}
perf tools: Insert split maps correctly into origin group When new maps are cloned out of split map they are added into origin map's group, but their groups pointer is not updated. This could lead to a segfault, because map->groups is expected to be always set as reported by Markus: __map__is_kernel (map=map@entry=0x1abb7a0) at util/map.c:238 238 return __machine__kernel_map(map->groups->machine, map->type) = (gdb) bt #0 __map__is_kernel (map=map@entry=0x1abb7a0) at util/map.c:238 #1 0x00000000004393e4 in symbol_filter (map=map@entry=0x1abb7a0, sym=sym@entry #2 0x00000000004fcd4d in dso__load_sym (dso=dso@entry=0x166dae0, map=map@entry #3 0x00000000004a64e0 in dso__load (dso=0x166dae0, map=map@entry=0x1abb7a0, fi #4 0x00000000004b941f in map__load (filter=0x4393c0 <symbol_filter>, map=<opti #5 map__find_symbol (map=0x1abb7a0, addr=40188, filter=0x4393c0 <symbol_filter ... Adding __map_groups__insert function to add map into groups together with map->groups pointer update. It takes no lock as opposed to existing map_groups__insert, as maps__fixup_overlappings(), where it is being called, already has the necessary lock held. Using __map_groups__insert to add new maps after map split. Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de> Signed-off-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Markus Trippelsdorf <markus@trippelsdorf.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20151104140811.GA32664@krava.brq.redhat.com Fixes: cfc5acd4c80b ("perf top: Filter symbols based on __map__is_kernel(map)") Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-04 22:08:11 +08:00
static void __map_groups__insert(struct map_groups *mg, struct map *map)
{
__maps__insert(&mg->maps[map->type], map);
map->groups = mg;
}
static int maps__fixup_overlappings(struct maps *maps, struct map *map, FILE *fp)
{
struct rb_root *root;
struct rb_node *next;
int err = 0;
pthread_rwlock_wrlock(&maps->lock);
root = &maps->entries;
next = rb_first(root);
while (next) {
struct map *pos = rb_entry(next, struct map, rb_node);
next = rb_next(&pos->rb_node);
if (!map__overlap(pos, map))
continue;
if (verbose >= 2) {
fputs("overlapping maps:\n", fp);
map__fprintf(map, fp);
map__fprintf(pos, fp);
}
rb_erase_init(&pos->rb_node, root);
/*
* Now check if we need to create new maps for areas not
* overlapped by the new map:
*/
if (map->start > pos->start) {
struct map *before = map__clone(pos);
if (before == NULL) {
err = -ENOMEM;
goto put_map;
}
before->end = map->start;
perf tools: Insert split maps correctly into origin group When new maps are cloned out of split map they are added into origin map's group, but their groups pointer is not updated. This could lead to a segfault, because map->groups is expected to be always set as reported by Markus: __map__is_kernel (map=map@entry=0x1abb7a0) at util/map.c:238 238 return __machine__kernel_map(map->groups->machine, map->type) = (gdb) bt #0 __map__is_kernel (map=map@entry=0x1abb7a0) at util/map.c:238 #1 0x00000000004393e4 in symbol_filter (map=map@entry=0x1abb7a0, sym=sym@entry #2 0x00000000004fcd4d in dso__load_sym (dso=dso@entry=0x166dae0, map=map@entry #3 0x00000000004a64e0 in dso__load (dso=0x166dae0, map=map@entry=0x1abb7a0, fi #4 0x00000000004b941f in map__load (filter=0x4393c0 <symbol_filter>, map=<opti #5 map__find_symbol (map=0x1abb7a0, addr=40188, filter=0x4393c0 <symbol_filter ... Adding __map_groups__insert function to add map into groups together with map->groups pointer update. It takes no lock as opposed to existing map_groups__insert, as maps__fixup_overlappings(), where it is being called, already has the necessary lock held. Using __map_groups__insert to add new maps after map split. Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de> Signed-off-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Markus Trippelsdorf <markus@trippelsdorf.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20151104140811.GA32664@krava.brq.redhat.com Fixes: cfc5acd4c80b ("perf top: Filter symbols based on __map__is_kernel(map)") Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-04 22:08:11 +08:00
__map_groups__insert(pos->groups, before);
if (verbose >= 2)
map__fprintf(before, fp);
perf tools: Fix maps__fixup_overlappings to put used maps Since the __map_groups__insert got the given map, we don't need to keep it. So put the maps. Refcnt debugger shows that map_groups__fixup_overlappings() got a map twice but the group released it just once. This pattern usually indicates the leak happens in caller site. ---- ==== [0] ==== Unreclaimed map@0x39d3ae0 Refcount +1 => 1 at ./perf(map_groups__fixup_overlappings+0x335) [0x4c1865] ./perf(thread__insert_map+0x30) [0x4c8e00] ./perf(machine__process_mmap2_event+0x106) [0x4bd876] ./perf() [0x4c378e] ./perf() [0x4c4393] ./perf(perf_session__process_events+0x38a) [0x4c654a] ./perf(cmd_record+0xe24) [0x42fc94] ./perf() [0x47b745] ./perf(main+0x617) [0x422547] /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f2eca2deaf5] ./perf() [0x4226bd] Refcount +1 => 2 at ./perf(map_groups__fixup_overlappings+0x3c5) [0x4c18f5] ./perf(thread__insert_map+0x30) [0x4c8e00] ./perf(machine__process_mmap2_event+0x106) [0x4bd876] ./perf() [0x4c378e] ./perf() [0x4c4393] ./perf(perf_session__process_events+0x38a) [0x4c654a] ./perf(cmd_record+0xe24) [0x42fc94] ./perf() [0x47b745] ./perf(main+0x617) [0x422547] /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f2eca2deaf5] ./perf() [0x4226bd] Refcount -1 => 1 at ./perf(map_groups__exit+0x92) [0x4c0962] ./perf(map_groups__put+0x60) [0x4c0bc0] ./perf(thread__put+0x90) [0x4c8a40] ./perf(machine__delete_threads+0x7e) [0x4bad9e] ./perf(perf_session__delete+0x4f) [0x4c499f] ./perf(cmd_record+0xb6d) [0x42f9dd] ./perf() [0x47b745] ./perf(main+0x617) [0x422547] /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f2eca2deaf5] ./perf() [0x4226bd] ---- Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20151209021131.10245.41485.stgit@localhost.localdomain Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-09 10:11:31 +08:00
map__put(before);
}
if (map->end < pos->end) {
struct map *after = map__clone(pos);
if (after == NULL) {
err = -ENOMEM;
goto put_map;
}
after->start = map->end;
perf tools: Insert split maps correctly into origin group When new maps are cloned out of split map they are added into origin map's group, but their groups pointer is not updated. This could lead to a segfault, because map->groups is expected to be always set as reported by Markus: __map__is_kernel (map=map@entry=0x1abb7a0) at util/map.c:238 238 return __machine__kernel_map(map->groups->machine, map->type) = (gdb) bt #0 __map__is_kernel (map=map@entry=0x1abb7a0) at util/map.c:238 #1 0x00000000004393e4 in symbol_filter (map=map@entry=0x1abb7a0, sym=sym@entry #2 0x00000000004fcd4d in dso__load_sym (dso=dso@entry=0x166dae0, map=map@entry #3 0x00000000004a64e0 in dso__load (dso=0x166dae0, map=map@entry=0x1abb7a0, fi #4 0x00000000004b941f in map__load (filter=0x4393c0 <symbol_filter>, map=<opti #5 map__find_symbol (map=0x1abb7a0, addr=40188, filter=0x4393c0 <symbol_filter ... Adding __map_groups__insert function to add map into groups together with map->groups pointer update. It takes no lock as opposed to existing map_groups__insert, as maps__fixup_overlappings(), where it is being called, already has the necessary lock held. Using __map_groups__insert to add new maps after map split. Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de> Signed-off-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Markus Trippelsdorf <markus@trippelsdorf.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20151104140811.GA32664@krava.brq.redhat.com Fixes: cfc5acd4c80b ("perf top: Filter symbols based on __map__is_kernel(map)") Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-04 22:08:11 +08:00
__map_groups__insert(pos->groups, after);
if (verbose >= 2)
map__fprintf(after, fp);
perf tools: Fix maps__fixup_overlappings to put used maps Since the __map_groups__insert got the given map, we don't need to keep it. So put the maps. Refcnt debugger shows that map_groups__fixup_overlappings() got a map twice but the group released it just once. This pattern usually indicates the leak happens in caller site. ---- ==== [0] ==== Unreclaimed map@0x39d3ae0 Refcount +1 => 1 at ./perf(map_groups__fixup_overlappings+0x335) [0x4c1865] ./perf(thread__insert_map+0x30) [0x4c8e00] ./perf(machine__process_mmap2_event+0x106) [0x4bd876] ./perf() [0x4c378e] ./perf() [0x4c4393] ./perf(perf_session__process_events+0x38a) [0x4c654a] ./perf(cmd_record+0xe24) [0x42fc94] ./perf() [0x47b745] ./perf(main+0x617) [0x422547] /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f2eca2deaf5] ./perf() [0x4226bd] Refcount +1 => 2 at ./perf(map_groups__fixup_overlappings+0x3c5) [0x4c18f5] ./perf(thread__insert_map+0x30) [0x4c8e00] ./perf(machine__process_mmap2_event+0x106) [0x4bd876] ./perf() [0x4c378e] ./perf() [0x4c4393] ./perf(perf_session__process_events+0x38a) [0x4c654a] ./perf(cmd_record+0xe24) [0x42fc94] ./perf() [0x47b745] ./perf(main+0x617) [0x422547] /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f2eca2deaf5] ./perf() [0x4226bd] Refcount -1 => 1 at ./perf(map_groups__exit+0x92) [0x4c0962] ./perf(map_groups__put+0x60) [0x4c0bc0] ./perf(thread__put+0x90) [0x4c8a40] ./perf(machine__delete_threads+0x7e) [0x4bad9e] ./perf(perf_session__delete+0x4f) [0x4c499f] ./perf(cmd_record+0xb6d) [0x42f9dd] ./perf() [0x47b745] ./perf(main+0x617) [0x422547] /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f2eca2deaf5] ./perf() [0x4226bd] ---- Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20151209021131.10245.41485.stgit@localhost.localdomain Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-09 10:11:31 +08:00
map__put(after);
}
put_map:
map__put(pos);
if (err)
goto out;
}
err = 0;
out:
pthread_rwlock_unlock(&maps->lock);
return err;
}
int map_groups__fixup_overlappings(struct map_groups *mg, struct map *map,
FILE *fp)
{
return maps__fixup_overlappings(&mg->maps[map->type], map, fp);
}
/*
* XXX This should not really _copy_ te maps, but refcount them.
*/
int map_groups__clone(struct map_groups *mg,
struct map_groups *parent, enum map_type type)
{
int err = -ENOMEM;
struct map *map;
struct maps *maps = &parent->maps[type];
pthread_rwlock_rdlock(&maps->lock);
for (map = maps__first(maps); map; map = map__next(map)) {
struct map *new = map__clone(map);
if (new == NULL)
goto out_unlock;
map_groups__insert(mg, new);
perf tools: Fix map_groups__clone to put cloned map Fix map_groups__clone to put cloned map after inserting it to the map_groups. Refcnt debugger shows: ---- ==== [0] ==== Unreclaimed map: 0x2a27ee0 Refcount +1 => 1 at ./perf(map_groups__clone+0x8d) [0x4bb7ed] ./perf(thread__fork+0xbe) [0x4c1f9e] ./perf(machine__process_fork_event+0x216) [0x4b79a6] ./perf(perf_event__synthesize_threads+0x38b) [0x48135b] ./perf(cmd_top+0xdc6) [0x43cb76] ./perf() [0x477223] ./perf(main+0x617) [0x422077] /lib64/libc.so.6(__libc_start_main+0xf0) [0x7ff806af8fe0] ./perf() [0x4221ed] Refcount +1 => 2 at ./perf(map_groups__clone+0x128) [0x4bb888] ./perf(thread__fork+0xbe) [0x4c1f9e] ./perf(machine__process_fork_event+0x216) [0x4b79a6] ./perf(perf_event__synthesize_threads+0x38b) [0x48135b] ./perf(cmd_top+0xdc6) [0x43cb76] ./perf() [0x477223] ./perf(main+0x617) [0x422077] /lib64/libc.so.6(__libc_start_main+0xf0) [0x7ff806af8fe0] ./perf() [0x4221ed] Refcount -1 => 1 at ./perf(map_groups__exit+0x87) [0x4ba757] ./perf(map_groups__put+0x68) [0x4ba9a8] ./perf(thread__put+0x8b) [0x4c1aeb] ./perf(machine__delete_threads+0x81) [0x4b48f1] ./perf(perf_session__delete+0x4f) [0x4be63f] ./perf(cmd_top+0x1094) [0x43ce44] ./perf() [0x477223] ./perf(main+0x617) [0x422077] /lib64/libc.so.6(__libc_start_main+0xf0) [0x7ff806af8fe0] ./perf() [0x4221ed] ---- This shows map_groups__clone get the map twice and put it when map_groups__exit. Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20151209021120.10245.95388.stgit@localhost.localdomain Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-09 10:11:20 +08:00
map__put(new);
}
err = 0;
out_unlock:
pthread_rwlock_unlock(&maps->lock);
return err;
}
static void __maps__insert(struct maps *maps, struct map *map)
{
struct rb_node **p = &maps->entries.rb_node;
struct rb_node *parent = NULL;
const u64 ip = map->start;
struct map *m;
while (*p != NULL) {
parent = *p;
m = rb_entry(parent, struct map, rb_node);
if (ip < m->start)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&map->rb_node, parent, p);
rb_insert_color(&map->rb_node, &maps->entries);
map__get(map);
}
void maps__insert(struct maps *maps, struct map *map)
{
pthread_rwlock_wrlock(&maps->lock);
__maps__insert(maps, map);
pthread_rwlock_unlock(&maps->lock);
}
static void __maps__remove(struct maps *maps, struct map *map)
{
rb_erase_init(&map->rb_node, &maps->entries);
map__put(map);
}
void maps__remove(struct maps *maps, struct map *map)
{
pthread_rwlock_wrlock(&maps->lock);
__maps__remove(maps, map);
pthread_rwlock_unlock(&maps->lock);
}
struct map *maps__find(struct maps *maps, u64 ip)
{
struct rb_node **p, *parent = NULL;
struct map *m;
pthread_rwlock_rdlock(&maps->lock);
p = &maps->entries.rb_node;
while (*p != NULL) {
parent = *p;
m = rb_entry(parent, struct map, rb_node);
if (ip < m->start)
p = &(*p)->rb_left;
else if (ip >= m->end)
p = &(*p)->rb_right;
else
goto out;
}
m = NULL;
out:
pthread_rwlock_unlock(&maps->lock);
return m;
}
struct map *maps__first(struct maps *maps)
{
struct rb_node *first = rb_first(&maps->entries);
if (first)
return rb_entry(first, struct map, rb_node);
return NULL;
}
struct map *map__next(struct map *map)
{
struct rb_node *next = rb_next(&map->rb_node);
if (next)
return rb_entry(next, struct map, rb_node);
return NULL;
}
struct kmap *map__kmap(struct map *map)
{
if (!map->dso || !map->dso->kernel) {
pr_err("Internal error: map__kmap with a non-kernel map\n");
return NULL;
}
return (struct kmap *)(map + 1);
}
struct map_groups *map__kmaps(struct map *map)
{
struct kmap *kmap = map__kmap(map);
if (!kmap || !kmap->kmaps) {
pr_err("Internal error: map__kmaps with a non-kernel map\n");
return NULL;
}
return kmap->kmaps;
}