2017-06-05 00:40:13

by Li Bin

[permalink] [raw]
Subject: [PATCH] perf symbols: Fix plt entry calculation for ARM and AARCH64

On x86, the plt header size is as same as the plt entry size, and
can be identified from shdr's sh_entsize of the plt.
But we cann't assume that the sh_entsize of the plt shdr is always
the plt entry size in all architecture, and the plt header size may
be not as same as the plt entry size in some architecure.

On ARM, the plt header size is 20 bytes and the plt entry size is 12
bytes(don't consider the FOUR_WORD_PLT case)
that refer to the binutils implementation.
plt section is as follows:

Disassembly of section .plt:
000004a0 <__cxa_finalize@plt-0x14>:
4a0: e52de004 push {lr} ; (str lr, [sp, #-4]!)
4a4: e59fe004 ldr lr, [pc, #4] ; 4b0 <_init+0x1c>
4a8: e08fe00e add lr, pc, lr
4ac: e5bef008 ldr pc, [lr, #8]!
4b0: 00008424 .word 0x00008424

000004b4 <__cxa_finalize@plt>:
4b4: e28fc600 add ip, pc, #0, 12
4b8: e28cca08 add ip, ip, #8, 20 ; 0x8000
4bc: e5bcf424 ldr pc, [ip, #1060]! ; 0x424

000004c0 <printf@plt>:
4c0: e28fc600 add ip, pc, #0, 12
4c4: e28cca08 add ip, ip, #8, 20 ; 0x8000
4c8: e5bcf41c ldr pc, [ip, #1052]! ; 0x41c

On AARCH64, the plt header size is 32 bytes and the plt entry size is
16 bytes.
plt section is as follows:

Disassembly of section .plt:
0000000000000560 <__cxa_finalize@plt-0x20>:
560: a9bf7bf0 stp x16, x30, [sp,#-16]!
564: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
568: f944be11 ldr x17, [x16,#2424]
56c: 9125e210 add x16, x16, #0x978
570: d61f0220 br x17
574: d503201f nop
578: d503201f nop
57c: d503201f nop

0000000000000580 <__cxa_finalize@plt>:
580: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
584: f944c211 ldr x17, [x16,#2432]
588: 91260210 add x16, x16, #0x980
58c: d61f0220 br x17

0000000000000590 <__gmon_start__@plt>:
590: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
594: f944c611 ldr x17, [x16,#2440]
598: 91262210 add x16, x16, #0x988
59c: d61f0220 br x17

NOTES:
In addition to ARM and AARCH64, other architectures, such as
s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa also need to consider
this issue.

Signed-off-by: Li Bin <[email protected]>
---
tools/perf/util/symbol-elf.c | 27 ++++++++++++++++++++++-----
1 file changed, 22 insertions(+), 5 deletions(-)

diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c
index e7ee47f..1bccf71 100644
--- a/tools/perf/util/symbol-elf.c
+++ b/tools/perf/util/symbol-elf.c
@@ -259,7 +259,7 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
{
uint32_t nr_rel_entries, idx;
GElf_Sym sym;
- u64 plt_offset;
+ u64 plt_offset, plt_header_size, plt_entry_size;
GElf_Shdr shdr_plt;
struct symbol *f;
GElf_Shdr shdr_rel_plt, shdr_dynsym;
@@ -326,6 +326,23 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *

nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
plt_offset = shdr_plt.sh_offset;
+ switch (ehdr.e_machine) {
+ case EM_ARM:
+ plt_header_size = 20;
+ plt_entry_size = 12;
+ break;
+
+ case EM_AARCH64:
+ plt_header_size = 32;
+ plt_entry_size = 16;
+ break;
+
+ default: /* FIXME: s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa need to be checked */
+ plt_header_size = shdr_plt.sh_entsize;
+ plt_entry_size = shdr_plt.sh_entsize;
+ break;
+ }
+ plt_offset += plt_header_size;

if (shdr_rel_plt.sh_type == SHT_RELA) {
GElf_Rela pos_mem, *pos;
@@ -335,7 +352,6 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
const char *elf_name = NULL;
char *demangled = NULL;
symidx = GELF_R_SYM(pos->r_info);
- plt_offset += shdr_plt.sh_entsize;
gelf_getsym(syms, symidx, &sym);

elf_name = elf_sym__name(&sym, symstrs);
@@ -346,11 +362,12 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
"%s@plt", elf_name);
free(demangled);

- f = symbol__new(plt_offset, shdr_plt.sh_entsize,
+ f = symbol__new(plt_offset, plt_entry_size,
STB_GLOBAL, sympltname);
if (!f)
goto out_elf_end;

+ plt_offset += plt_entry_size;
symbols__insert(&dso->symbols[map->type], f);
++nr;
}
@@ -361,7 +378,6 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
const char *elf_name = NULL;
char *demangled = NULL;
symidx = GELF_R_SYM(pos->r_info);
- plt_offset += shdr_plt.sh_entsize;
gelf_getsym(syms, symidx, &sym);

elf_name = elf_sym__name(&sym, symstrs);
@@ -372,11 +388,12 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
"%s@plt", elf_name);
free(demangled);

- f = symbol__new(plt_offset, shdr_plt.sh_entsize,
+ f = symbol__new(plt_offset, plt_entry_size,
STB_GLOBAL, sympltname);
if (!f)
goto out_elf_end;

+ plt_offset += plt_entry_size;
symbols__insert(&dso->symbols[map->type], f);
++nr;
}
--
1.7.12.4


2017-06-06 01:47:50

by Namhyung Kim

[permalink] [raw]
Subject: Re: [PATCH] perf symbols: Fix plt entry calculation for ARM and AARCH64

On Mon, Jun 5, 2017 at 9:34 AM, Li Bin <[email protected]> wrote:
> On x86, the plt header size is as same as the plt entry size, and
> can be identified from shdr's sh_entsize of the plt.
> But we cann't assume that the sh_entsize of the plt shdr is always
> the plt entry size in all architecture, and the plt header size may
> be not as same as the plt entry size in some architecure.
>
> On ARM, the plt header size is 20 bytes and the plt entry size is 12
> bytes(don't consider the FOUR_WORD_PLT case)
> that refer to the binutils implementation.
> plt section is as follows:
>
> Disassembly of section .plt:
> 000004a0 <__cxa_finalize@plt-0x14>:
> 4a0: e52de004 push {lr} ; (str lr, [sp, #-4]!)
> 4a4: e59fe004 ldr lr, [pc, #4] ; 4b0 <_init+0x1c>
> 4a8: e08fe00e add lr, pc, lr
> 4ac: e5bef008 ldr pc, [lr, #8]!
> 4b0: 00008424 .word 0x00008424
>
> 000004b4 <__cxa_finalize@plt>:
> 4b4: e28fc600 add ip, pc, #0, 12
> 4b8: e28cca08 add ip, ip, #8, 20 ; 0x8000
> 4bc: e5bcf424 ldr pc, [ip, #1060]! ; 0x424
>
> 000004c0 <printf@plt>:
> 4c0: e28fc600 add ip, pc, #0, 12
> 4c4: e28cca08 add ip, ip, #8, 20 ; 0x8000
> 4c8: e5bcf41c ldr pc, [ip, #1052]! ; 0x41c
>
> On AARCH64, the plt header size is 32 bytes and the plt entry size is
> 16 bytes.
> plt section is as follows:
>
> Disassembly of section .plt:
> 0000000000000560 <__cxa_finalize@plt-0x20>:
> 560: a9bf7bf0 stp x16, x30, [sp,#-16]!
> 564: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
> 568: f944be11 ldr x17, [x16,#2424]
> 56c: 9125e210 add x16, x16, #0x978
> 570: d61f0220 br x17
> 574: d503201f nop
> 578: d503201f nop
> 57c: d503201f nop
>
> 0000000000000580 <__cxa_finalize@plt>:
> 580: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
> 584: f944c211 ldr x17, [x16,#2432]
> 588: 91260210 add x16, x16, #0x980
> 58c: d61f0220 br x17
>
> 0000000000000590 <__gmon_start__@plt>:
> 590: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
> 594: f944c611 ldr x17, [x16,#2440]
> 598: 91262210 add x16, x16, #0x988
> 59c: d61f0220 br x17
>
> NOTES:
> In addition to ARM and AARCH64, other architectures, such as
> s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa also need to consider
> this issue.
>
> Signed-off-by: Li Bin <[email protected]>

I don't know other archs, but this change looks ok to me..

Acked-by: Namhyung Kim <[email protected]>

Thanks,
Namhyung


> ---
> tools/perf/util/symbol-elf.c | 27 ++++++++++++++++++++++-----
> 1 file changed, 22 insertions(+), 5 deletions(-)
>
> diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c
> index e7ee47f..1bccf71 100644
> --- a/tools/perf/util/symbol-elf.c
> +++ b/tools/perf/util/symbol-elf.c
> @@ -259,7 +259,7 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
> {
> uint32_t nr_rel_entries, idx;
> GElf_Sym sym;
> - u64 plt_offset;
> + u64 plt_offset, plt_header_size, plt_entry_size;
> GElf_Shdr shdr_plt;
> struct symbol *f;
> GElf_Shdr shdr_rel_plt, shdr_dynsym;
> @@ -326,6 +326,23 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
>
> nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
> plt_offset = shdr_plt.sh_offset;
> + switch (ehdr.e_machine) {
> + case EM_ARM:
> + plt_header_size = 20;
> + plt_entry_size = 12;
> + break;
> +
> + case EM_AARCH64:
> + plt_header_size = 32;
> + plt_entry_size = 16;
> + break;
> +
> + default: /* FIXME: s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa need to be checked */
> + plt_header_size = shdr_plt.sh_entsize;
> + plt_entry_size = shdr_plt.sh_entsize;
> + break;
> + }
> + plt_offset += plt_header_size;
>
> if (shdr_rel_plt.sh_type == SHT_RELA) {
> GElf_Rela pos_mem, *pos;
> @@ -335,7 +352,6 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
> const char *elf_name = NULL;
> char *demangled = NULL;
> symidx = GELF_R_SYM(pos->r_info);
> - plt_offset += shdr_plt.sh_entsize;
> gelf_getsym(syms, symidx, &sym);
>
> elf_name = elf_sym__name(&sym, symstrs);
> @@ -346,11 +362,12 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
> "%s@plt", elf_name);
> free(demangled);
>
> - f = symbol__new(plt_offset, shdr_plt.sh_entsize,
> + f = symbol__new(plt_offset, plt_entry_size,
> STB_GLOBAL, sympltname);
> if (!f)
> goto out_elf_end;
>
> + plt_offset += plt_entry_size;
> symbols__insert(&dso->symbols[map->type], f);
> ++nr;
> }
> @@ -361,7 +378,6 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
> const char *elf_name = NULL;
> char *demangled = NULL;
> symidx = GELF_R_SYM(pos->r_info);
> - plt_offset += shdr_plt.sh_entsize;
> gelf_getsym(syms, symidx, &sym);
>
> elf_name = elf_sym__name(&sym, symstrs);
> @@ -372,11 +388,12 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
> "%s@plt", elf_name);
> free(demangled);
>
> - f = symbol__new(plt_offset, shdr_plt.sh_entsize,
> + f = symbol__new(plt_offset, plt_entry_size,
> STB_GLOBAL, sympltname);
> if (!f)
> goto out_elf_end;
>
> + plt_offset += plt_entry_size;
> symbols__insert(&dso->symbols[map->type], f);
> ++nr;
> }
> --
> 1.7.12.4
>

2017-08-29 14:49:59

by Arnaldo Carvalho de Melo

[permalink] [raw]
Subject: Re: [PATCH] perf symbols: Fix plt entry calculation for ARM and AARCH64

Em Tue, Jun 06, 2017 at 10:47:26AM +0900, Namhyung Kim escreveu:
> On Mon, Jun 5, 2017 at 9:34 AM, Li Bin <[email protected]> wrote:
> > On x86, the plt header size is as same as the plt entry size, and
> > can be identified from shdr's sh_entsize of the plt.
> > But we cann't assume that the sh_entsize of the plt shdr is always
> > the plt entry size in all architecture, and the plt header size may
> > be not as same as the plt entry size in some architecure.
> >
> > On ARM, the plt header size is 20 bytes and the plt entry size is 12
> > bytes(don't consider the FOUR_WORD_PLT case)
> > that refer to the binutils implementation.
<SNIP>
> > NOTES:
> > In addition to ARM and AARCH64, other architectures, such as
> > s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa also need to consider
> > this issue.
> >
> > Signed-off-by: Li Bin <[email protected]>
>
> I don't know other archs, but this change looks ok to me..
>
> Acked-by: Namhyung Kim <[email protected]>

Thanks, and sorry for the delay in processing this one.

- Arnaldo

Subject: [tip:perf/core] perf symbols: Fix plt entry calculation for ARM and AARCH64

Commit-ID: b2f7605076d6cdd68162c42c34caadafbbe4c69f
Gitweb: http://git.kernel.org/tip/b2f7605076d6cdd68162c42c34caadafbbe4c69f
Author: Li Bin <[email protected]>
AuthorDate: Mon, 5 Jun 2017 08:34:09 +0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 29 Aug 2017 11:41:27 -0300

perf symbols: Fix plt entry calculation for ARM and AARCH64

On x86, the plt header size is as same as the plt entry size, and can be
identified from shdr's sh_entsize of the plt.

But we can't assume that the sh_entsize of the plt shdr is always the
plt entry size in all architecture, and the plt header size may be not
as same as the plt entry size in some architecure.

On ARM, the plt header size is 20 bytes and the plt entry size is 12
bytes (don't consider the FOUR_WORD_PLT case) that refer to the binutils
implementation. The plt section is as follows:

Disassembly of section .plt:
000004a0 <__cxa_finalize@plt-0x14>:
4a0: e52de004 push {lr} ; (str lr, [sp, #-4]!)
4a4: e59fe004 ldr lr, [pc, #4] ; 4b0 <_init+0x1c>
4a8: e08fe00e add lr, pc, lr
4ac: e5bef008 ldr pc, [lr, #8]!
4b0: 00008424 .word 0x00008424

000004b4 <__cxa_finalize@plt>:
4b4: e28fc600 add ip, pc, #0, 12
4b8: e28cca08 add ip, ip, #8, 20 ; 0x8000
4bc: e5bcf424 ldr pc, [ip, #1060]! ; 0x424

000004c0 <printf@plt>:
4c0: e28fc600 add ip, pc, #0, 12
4c4: e28cca08 add ip, ip, #8, 20 ; 0x8000
4c8: e5bcf41c ldr pc, [ip, #1052]! ; 0x41c

On AARCH64, the plt header size is 32 bytes and the plt entry size is 16
bytes. The plt section is as follows:

Disassembly of section .plt:
0000000000000560 <__cxa_finalize@plt-0x20>:
560: a9bf7bf0 stp x16, x30, [sp,#-16]!
564: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
568: f944be11 ldr x17, [x16,#2424]
56c: 9125e210 add x16, x16, #0x978
570: d61f0220 br x17
574: d503201f nop
578: d503201f nop
57c: d503201f nop

0000000000000580 <__cxa_finalize@plt>:
580: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
584: f944c211 ldr x17, [x16,#2432]
588: 91260210 add x16, x16, #0x980
58c: d61f0220 br x17

0000000000000590 <__gmon_start__@plt>:
590: 90000090 adrp x16, 10000 <__FRAME_END__+0xf8a8>
594: f944c611 ldr x17, [x16,#2440]
598: 91262210 add x16, x16, #0x988
59c: d61f0220 br x17

NOTES:

In addition to ARM and AARCH64, other architectures, such as
s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa also need to consider
this issue.

Signed-off-by: Li Bin <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Alexis Berlemont <[email protected]>
Cc: David Tolnay <[email protected]>
Cc: Hanjun Guo <[email protected]>
Cc: Hemant Kumar <[email protected]>
Cc: Masami Hiramatsu <[email protected]>
Cc: Milian Wolff <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/symbol-elf.c | 27 ++++++++++++++++++++++-----
1 file changed, 22 insertions(+), 5 deletions(-)

diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c
index a704790..5c39f42 100644
--- a/tools/perf/util/symbol-elf.c
+++ b/tools/perf/util/symbol-elf.c
@@ -259,7 +259,7 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
{
uint32_t nr_rel_entries, idx;
GElf_Sym sym;
- u64 plt_offset;
+ u64 plt_offset, plt_header_size, plt_entry_size;
GElf_Shdr shdr_plt;
struct symbol *f;
GElf_Shdr shdr_rel_plt, shdr_dynsym;
@@ -326,6 +326,23 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *

nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
plt_offset = shdr_plt.sh_offset;
+ switch (ehdr.e_machine) {
+ case EM_ARM:
+ plt_header_size = 20;
+ plt_entry_size = 12;
+ break;
+
+ case EM_AARCH64:
+ plt_header_size = 32;
+ plt_entry_size = 16;
+ break;
+
+ default: /* FIXME: s390/alpha/mips/parisc/poperpc/sh/sparc/xtensa need to be checked */
+ plt_header_size = shdr_plt.sh_entsize;
+ plt_entry_size = shdr_plt.sh_entsize;
+ break;
+ }
+ plt_offset += plt_header_size;

if (shdr_rel_plt.sh_type == SHT_RELA) {
GElf_Rela pos_mem, *pos;
@@ -335,7 +352,6 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
const char *elf_name = NULL;
char *demangled = NULL;
symidx = GELF_R_SYM(pos->r_info);
- plt_offset += shdr_plt.sh_entsize;
gelf_getsym(syms, symidx, &sym);

elf_name = elf_sym__name(&sym, symstrs);
@@ -346,11 +362,12 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
"%s@plt", elf_name);
free(demangled);

- f = symbol__new(plt_offset, shdr_plt.sh_entsize,
+ f = symbol__new(plt_offset, plt_entry_size,
STB_GLOBAL, sympltname);
if (!f)
goto out_elf_end;

+ plt_offset += plt_entry_size;
symbols__insert(&dso->symbols[map->type], f);
++nr;
}
@@ -361,7 +378,6 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
const char *elf_name = NULL;
char *demangled = NULL;
symidx = GELF_R_SYM(pos->r_info);
- plt_offset += shdr_plt.sh_entsize;
gelf_getsym(syms, symidx, &sym);

elf_name = elf_sym__name(&sym, symstrs);
@@ -372,11 +388,12 @@ int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *
"%s@plt", elf_name);
free(demangled);

- f = symbol__new(plt_offset, shdr_plt.sh_entsize,
+ f = symbol__new(plt_offset, plt_entry_size,
STB_GLOBAL, sympltname);
if (!f)
goto out_elf_end;

+ plt_offset += plt_entry_size;
symbols__insert(&dso->symbols[map->type], f);
++nr;
}