On Sun, Feb 24, 2019 at 07:37:13AM -0800, Andi Kleen wrote:
> From: Andi Kleen <a...@linux.intel.com>
>
> perf script -F +insn was only working for PT traces because
> the PT instruction decoder was filling in the insn/insn_len
> sample attributes. Support it for non PT samples too on x86
> using the existing x86 instruction decoder.
>
> % perf record -a sleep 1
> % perf script -F ip,sym,insn --xed
> ffffffff811704c9 remote_function movl %eax, 0x18(%rbx)
> ffffffff8100bb50 intel_bts_enable_local retq
> ffffffff81048612 native_apic_mem_write movl %esi,
> -0xa04000(%rdi)
> ffffffff81048612 native_apic_mem_write movl %esi,
> -0xa04000(%rdi)
> ffffffff81048612 native_apic_mem_write movl %esi,
> -0xa04000(%rdi)
> ffffffff810f1f79 generic_exec_single xor %eax, %eax
> ffffffff811704c9 remote_function movl %eax, 0x18(%rbx)
> ffffffff8100bb34 intel_bts_enable_local movl 0x2000(%rax),
> %edx
> ffffffff81048610 native_apic_mem_write mov %edi, %edi
> ...
>
> Signed-off-by: Andi Kleen <a...@linux.intel.com>
> ---
> tools/perf/arch/x86/util/Build | 1 +
> tools/perf/arch/x86/util/archinsn.c | 41 +++++++++++++++++++++++++++++
> tools/perf/builtin-script.c | 10 +++++++
> tools/perf/util/archinsn.h | 12 +++++++++
> 4 files changed, 64 insertions(+)
> create mode 100644 tools/perf/arch/x86/util/archinsn.c
> create mode 100644 tools/perf/util/archinsn.h
>
> diff --git a/tools/perf/arch/x86/util/Build b/tools/perf/arch/x86/util/Build
> index 7aab0be5fc5f..7b8e69bbbdfe 100644
> --- a/tools/perf/arch/x86/util/Build
> +++ b/tools/perf/arch/x86/util/Build
> @@ -6,6 +6,7 @@ perf-y += perf_regs.o
> perf-y += group.o
> perf-y += machine.o
> perf-y += event.o
> +perf-y += archinsn.o
>
> perf-$(CONFIG_DWARF) += dwarf-regs.o
> perf-$(CONFIG_BPF_PROLOGUE) += dwarf-regs.o
> diff --git a/tools/perf/arch/x86/util/archinsn.c
> b/tools/perf/arch/x86/util/archinsn.c
> new file mode 100644
> index 000000000000..9e3b0828b018
> --- /dev/null
> +++ b/tools/perf/arch/x86/util/archinsn.c
> @@ -0,0 +1,41 @@
> +// SPDX-License-Identifier: GPL-2.0
> +#include "perf.h"
> +#include "archinsn.h"
> +#include "util/intel-pt-decoder/insn.h"
> +#include "machine.h"
> +#include "thread.h"
> +#include "symbol.h"
> +#include "map.h"
> +
> +void arch_fetch_insn(struct perf_sample *sample,
> + struct thread *thread,
> + struct machine *machine)
> +{
> + struct addr_location al;
> + u8 cpumode;
> + long offset;
> + struct insn insn;
> + int len;
> +
> + if (!sample->ip)
> + return;
> +
> + if (machine__kernel_ip(machine, sample->ip))
> + cpumode = PERF_RECORD_MISC_KERNEL;
> + else
> + cpumode = PERF_RECORD_MISC_USER;
> + if (!thread__find_map(thread, cpumode, sample->ip, &al) || !al.map->dso)
> + return;
> + if (al.map->dso->data.status == DSO_DATA_STATUS_ERROR)
> + return;
> + map__load(al.map);
> + offset = al.map->map_ip(al.map, sample->ip);
> + len = dso__data_read_offset(al.map->dso, machine, offset, (u8
> *)sample->insn,
> + sizeof(sample->insn));
> + if (len <= 0)
> + return;
> + insn_init(&insn, sample->insn, len, al.map->dso->is_64_bit);
> + insn_get_length(&insn);
> + if (insn_complete(&insn) && insn.length <= len)
> + sample->insn_len = insn.length;
> +}
I saw this code around multiple times.. I think bts and pt are using
same code to fetch instructions.. could we synchronize and have just
one function to do this?
thanks,
jirka
