linux/tools/perf/util/map.c

599 lines
13 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"
#include <linux/string.h>
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") ||
!strcmp(filename, "/dev/zero (deleted)") ||
!strcmp(filename, "/anon_hugepage (deleted)");
}
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) ||
!strcmp(filename, "[heap]");
}
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->dso = dso;
map->map_ip = map__map_ip;
map->unmap_ip = map__unmap_ip;
RB_CLEAR_NODE(&map->rb_node);
map->groups = NULL;
map->referenced = false;
map->erange_warned = false;
}
struct map *map__new(struct list_head *dsos__list, u64 start, u64 len,
u64 pgoff, u32 pid, u32 d_maj, u32 d_min, u64 ino,
u64 ino_gen, char *filename,
enum map_type type)
{
struct map *map = malloc(sizeof(*map));
if (map != NULL) {
char newfilename[PATH_MAX];
struct dso *dso;
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
int anon, no_dso, vdso;
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;
if (anon) {
snprintf(newfilename, sizeof(newfilename), "/tmp/perf-%d.map", pid);
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 = vdso__dso_findnew(dsos__list);
} else
dso = __dsos__findnew(dsos__list, filename);
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 (no_dso)
dso__set_loaded(dso, map->type);
}
}
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;
}
void map__delete(struct map *map)
{
free(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;
}
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 *map)
{
return memdup(map, sizeof(*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 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.
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;
return map->unmap_ip(map, rip);
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
void map_groups__init(struct map_groups *mg)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i) {
mg->maps[i] = RB_ROOT;
INIT_LIST_HEAD(&mg->removed_maps[i]);
}
mg->machine = NULL;
}
static void maps__delete(struct rb_root *maps)
{
struct rb_node *next = rb_first(maps);
while (next) {
struct map *pos = rb_entry(next, struct map, rb_node);
next = rb_next(&pos->rb_node);
rb_erase(&pos->rb_node, maps);
map__delete(pos);
}
}
static void maps__delete_removed(struct list_head *maps)
{
struct map *pos, *n;
list_for_each_entry_safe(pos, n, maps, node) {
list_del(&pos->node);
map__delete(pos);
}
}
void map_groups__exit(struct map_groups *mg)
{
int i;
for (i = 0; i < MAP__NR_TYPES; ++i) {
maps__delete(&mg->maps[i]);
maps__delete_removed(&mg->removed_maps[i]);
}
}
void map_groups__flush(struct map_groups *mg)
{
int type;
for (type = 0; type < MAP__NR_TYPES; type++) {
struct rb_root *root = &mg->maps[type];
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(&pos->rb_node, root);
/*
* We may have references to this map, for
* instance in some hist_entry instances, so
* just move them to a separate list.
*/
list_add_tail(&pos->node, &mg->removed_maps[pos->type]);
}
}
}
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);
if (map != NULL) {
if (mapp != NULL)
*mapp = map;
return map__find_symbol(map, map->map_ip(map, addr), filter);
}
return NULL;
}
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 rb_node *nd;
for (nd = rb_first(&mg->maps[type]); nd; nd = rb_next(nd)) {
struct map *pos = rb_entry(nd, struct map, rb_node);
struct symbol *sym = map__find_symbol_by_name(pos, name, filter);
if (sym == NULL)
continue;
if (mapp != NULL)
*mapp = pos;
return sym;
}
return NULL;
}
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;
}
size_t __map_groups__fprintf_maps(struct map_groups *mg,
enum map_type type, int verbose, FILE *fp)
{
size_t printed = fprintf(fp, "%s:\n", map_type__name[type]);
struct rb_node *nd;
for (nd = rb_first(&mg->maps[type]); 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, type, fp);
printed += fprintf(fp, "--\n");
}
}
return printed;
}
size_t map_groups__fprintf_maps(struct map_groups *mg, int verbose, FILE *fp)
{
size_t printed = 0, i;
for (i = 0; i < MAP__NR_TYPES; ++i)
printed += __map_groups__fprintf_maps(mg, i, verbose, fp);
return printed;
}
static size_t __map_groups__fprintf_removed_maps(struct map_groups *mg,
enum map_type type,
int verbose, FILE *fp)
{
struct map *pos;
size_t printed = 0;
list_for_each_entry(pos, &mg->removed_maps[type], node) {
printed += fprintf(fp, "Map:");
printed += map__fprintf(pos, fp);
if (verbose > 1) {
printed += dso__fprintf(pos->dso, type, fp);
printed += fprintf(fp, "--\n");
}
}
return printed;
}
static size_t map_groups__fprintf_removed_maps(struct map_groups *mg,
int verbose, FILE *fp)
{
size_t printed = 0, i;
for (i = 0; i < MAP__NR_TYPES; ++i)
printed += __map_groups__fprintf_removed_maps(mg, i, verbose, fp);
return printed;
}
size_t map_groups__fprintf(struct map_groups *mg, int verbose, FILE *fp)
{
size_t printed = map_groups__fprintf_maps(mg, verbose, fp);
printed += fprintf(fp, "Removed maps:\n");
return printed + map_groups__fprintf_removed_maps(mg, verbose, fp);
}
int map_groups__fixup_overlappings(struct map_groups *mg, struct map *map,
int verbose, FILE *fp)
{
struct rb_root *root = &mg->maps[map->type];
struct rb_node *next = rb_first(root);
int err = 0;
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(&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 move_map;
}
before->end = map->start - 1;
map_groups__insert(mg, before);
if (verbose >= 2)
map__fprintf(before, fp);
}
if (map->end < pos->end) {
struct map *after = map__clone(pos);
if (after == NULL) {
err = -ENOMEM;
goto move_map;
}
after->start = map->end + 1;
map_groups__insert(mg, after);
if (verbose >= 2)
map__fprintf(after, fp);
}
move_map:
/*
* If we have references, just move them to a separate list.
*/
if (pos->referenced)
list_add_tail(&pos->node, &mg->removed_maps[map->type]);
else
map__delete(pos);
if (err)
return err;
}
return 0;
}
/*
* 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)
{
struct rb_node *nd;
for (nd = rb_first(&parent->maps[type]); nd; nd = rb_next(nd)) {
struct map *map = rb_entry(nd, struct map, rb_node);
struct map *new = map__clone(map);
if (new == NULL)
return -ENOMEM;
map_groups__insert(mg, new);
}
return 0;
}
void maps__insert(struct rb_root *maps, struct map *map)
{
struct rb_node **p = &maps->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);
}
void maps__remove(struct rb_root *maps, struct map *map)
{
rb_erase(&map->rb_node, maps);
}
struct map *maps__find(struct rb_root *maps, u64 ip)
{
struct rb_node **p = &maps->rb_node;
struct rb_node *parent = NULL;
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 if (ip > m->end)
p = &(*p)->rb_right;
else
return m;
}
return NULL;
}
struct map *maps__first(struct rb_root *maps)
{
struct rb_node *first = rb_first(maps);
if (first)
return rb_entry(first, struct map, rb_node);
return NULL;
}
struct map *maps__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;
}