| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAP-5=fVxwV0rKrHgDN_j-SsHnAvrJK-WgDuvi-KAYkqNO6iTJQ@mail.gmail.com>
Date: Fri, 16 Jan 2026 21:42:04 -0800
From: Ian Rogers <irogers@...gle.com>
To: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
Arnaldo Carvalho de Melo <acme@...nel.org>, Namhyung Kim <namhyung@...nel.org>, Jiri Olsa <jolsa@...nel.org>,
Ian Rogers <irogers@...gle.com>, Adrian Hunter <adrian.hunter@...el.com>,
James Clark <james.clark@...aro.org>, John Garry <john.g.garry@...cle.com>,
Will Deacon <will@...nel.org>, Leo Yan <leo.yan@...ux.dev>, Guo Ren <guoren@...nel.org>,
Paul Walmsley <pjw@...nel.org>, Palmer Dabbelt <palmer@...belt.com>, Albert Ou <aou@...s.berkeley.edu>,
Alexandre Ghiti <alex@...ti.fr>, Shimin Guo <shimin.guo@...dio.com>,
Athira Rajeev <atrajeev@...ux.ibm.com>, Stephen Brennan <stephen.s.brennan@...cle.com>,
Howard Chu <howardchu95@...il.com>, Thomas Falcon <thomas.falcon@...el.com>,
Andi Kleen <ak@...ux.intel.com>, "Dr. David Alan Gilbert" <linux@...blig.org>,
Dmitry Vyukov <dvyukov@...gle.com>,
Krzysztof Łopatowski <krzysztof.m.lopatowski@...il.com>,
Chun-Tse Shao <ctshao@...gle.com>, Aditya Bodkhe <aditya.b1@...ux.ibm.com>,
Haibo Xu <haibo1.xu@...el.com>, Sergei Trofimovich <slyich@...il.com>, linux-kernel@...r.kernel.org,
linux-perf-users@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-csky@...r.kernel.org, linux-riscv@...ts.infradead.org,
Mark Wielaard <mark@...mp.org>
Subject: Re: [PATCH v1 12/23] perf dwarf-regs: Add get_dwarf_regnum_for_perf_regnum
and use for x86 unwinding
On Fri, Jan 16, 2026 at 9:29 PM Ian Rogers <irogers@...gle.com> wrote:
>
> Add a utility to map a perf register number to a dwarf register number
> for a particular ELF machine type.
>
> Create a generic unwind-libdw initial register initialization routine
> that uses this function and thereby avoids arch specific
> initialization. The unwind-libdw code does:
> 1) compute the maximum dwarf register from the set of sampled user registers,
> 2) allocates a set of dwarf registers,
> 3) copies the sample registers into the appropriate dwarf registers.
>
> This generic solution is initially implemented for use with x86 as
> only get_dwarf_regnum_for_perf_regnum support for x86 is currently present.
>
> Signed-off-by: Ian Rogers <irogers@...gle.com>
> ---
> .../util/dwarf-regs-arch/dwarf-regs-x86.c | 95 +++++++++++++++++++
> tools/perf/util/dwarf-regs.c | 55 +++++++++++
> tools/perf/util/include/dwarf-regs.h | 8 ++
> tools/perf/util/unwind-libdw-arch/Build | 1 -
> .../util/unwind-libdw-arch/unwind-libdw-x86.c | 54 -----------
> tools/perf/util/unwind-libdw.c | 70 ++++++++++++--
> tools/perf/util/unwind-libdw.h | 2 +-
> 7 files changed, 222 insertions(+), 63 deletions(-)
> delete mode 100644 tools/perf/util/unwind-libdw-arch/unwind-libdw-x86.c
>
> diff --git a/tools/perf/util/dwarf-regs-arch/dwarf-regs-x86.c b/tools/perf/util/dwarf-regs-arch/dwarf-regs-x86.c
> index f0c42e4d7423..cadef120aeb4 100644
> --- a/tools/perf/util/dwarf-regs-arch/dwarf-regs-x86.c
> +++ b/tools/perf/util/dwarf-regs-arch/dwarf-regs-x86.c
> @@ -10,6 +10,7 @@
> #include <string.h> /* for strcmp */
> #include <linux/kernel.h> /* for ARRAY_SIZE */
> #include <dwarf-regs.h>
> +#include "../../../arch/x86/include/uapi/asm/perf_regs.h"
>
> struct dwarf_regs_idx {
> const char *name;
> @@ -163,3 +164,97 @@ int __get_dwarf_regnum_x86_64(const char *name)
> {
> return get_regnum(x86_64_regidx_table, ARRAY_SIZE(x86_64_regidx_table), name);
> }
> +
> +int __get_dwarf_regnum_for_perf_regnum_i386(int perf_regnum)
> +{
> + static const int dwarf_i386_regnums[] = {
> + [PERF_REG_X86_AX] = 0,
> + [PERF_REG_X86_BX] = 3,
> + [PERF_REG_X86_CX] = 1,
> + [PERF_REG_X86_DX] = 2,
> + [PERF_REG_X86_SI] = 6,
> + [PERF_REG_X86_DI] = 7,
> + [PERF_REG_X86_BP] = 5,
> + [PERF_REG_X86_SP] = 4,
> + [PERF_REG_X86_IP] = 8,
> + [PERF_REG_X86_FLAGS] = 9,
> + [PERF_REG_X86_CS] = 41,
> + [PERF_REG_X86_SS] = 42,
> + [PERF_REG_X86_DS] = 43,
> + [PERF_REG_X86_ES] = 40,
> + [PERF_REG_X86_FS] = 44,
> + [PERF_REG_X86_GS] = 45,
> + [PERF_REG_X86_XMM0] = 21,
> + [PERF_REG_X86_XMM1] = 22,
> + [PERF_REG_X86_XMM2] = 23,
> + [PERF_REG_X86_XMM3] = 24,
> + [PERF_REG_X86_XMM4] = 25,
> + [PERF_REG_X86_XMM5] = 26,
> + [PERF_REG_X86_XMM6] = 27,
> + [PERF_REG_X86_XMM7] = 28,
> + };
> +
> + if (perf_regnum == 0)
> + return 0;
> +
> + if (perf_regnum < 0 || perf_regnum > (int)ARRAY_SIZE(dwarf_i386_regnums) ||
> + dwarf_i386_regnums[perf_regnum] == 0)
> + return -ENOENT;
> +
> + return dwarf_i386_regnums[perf_regnum];
> +}
> +
> +int __get_dwarf_regnum_for_perf_regnum_x86_64(int perf_regnum)
> +{
> + static const int dwarf_x86_64_regnums[] = {
> + [PERF_REG_X86_AX] = 0,
> + [PERF_REG_X86_BX] = 3,
> + [PERF_REG_X86_CX] = 2,
> + [PERF_REG_X86_DX] = 1,
> + [PERF_REG_X86_SI] = 4,
> + [PERF_REG_X86_DI] = 5,
> + [PERF_REG_X86_BP] = 6,
> + [PERF_REG_X86_SP] = 7,
> + [PERF_REG_X86_IP] = 16,
> + [PERF_REG_X86_FLAGS] = 49,
> + [PERF_REG_X86_CS] = 51,
> + [PERF_REG_X86_SS] = 52,
> + [PERF_REG_X86_DS] = 53,
> + [PERF_REG_X86_ES] = 50,
> + [PERF_REG_X86_FS] = 54,
> + [PERF_REG_X86_GS] = 55,
> + [PERF_REG_X86_R8] = 8,
> + [PERF_REG_X86_R9] = 9,
> + [PERF_REG_X86_R10] = 10,
> + [PERF_REG_X86_R11] = 11,
> + [PERF_REG_X86_R12] = 12,
> + [PERF_REG_X86_R13] = 13,
> + [PERF_REG_X86_R14] = 14,
> + [PERF_REG_X86_R15] = 15,
> + [PERF_REG_X86_XMM0] = 17,
> + [PERF_REG_X86_XMM1] = 18,
> + [PERF_REG_X86_XMM2] = 19,
> + [PERF_REG_X86_XMM3] = 20,
> + [PERF_REG_X86_XMM4] = 21,
> + [PERF_REG_X86_XMM5] = 22,
> + [PERF_REG_X86_XMM6] = 23,
> + [PERF_REG_X86_XMM7] = 24,
> + [PERF_REG_X86_XMM8] = 25,
> + [PERF_REG_X86_XMM9] = 26,
> + [PERF_REG_X86_XMM10] = 27,
> + [PERF_REG_X86_XMM11] = 28,
> + [PERF_REG_X86_XMM12] = 29,
> + [PERF_REG_X86_XMM13] = 30,
> + [PERF_REG_X86_XMM14] = 31,
> + [PERF_REG_X86_XMM15] = 32,
> + };
> +
> + if (perf_regnum == 0)
> + return 0;
> +
> + if (perf_regnum < 0 || perf_regnum > (int)ARRAY_SIZE(dwarf_x86_64_regnums) ||
> + dwarf_x86_64_regnums[perf_regnum] == 0)
> + return -ENOENT;
> +
> + return dwarf_x86_64_regnums[perf_regnum];
> +}
> diff --git a/tools/perf/util/dwarf-regs.c b/tools/perf/util/dwarf-regs.c
> index ef249dd589e3..1f7d892612df 100644
> --- a/tools/perf/util/dwarf-regs.c
> +++ b/tools/perf/util/dwarf-regs.c
> @@ -103,3 +103,58 @@ int get_dwarf_regnum(const char *name, unsigned int machine, unsigned int flags
> free(regname);
> return reg;
> }
> +
> +static int get_libdw_frame_nregs(unsigned int machine, unsigned int flags __maybe_unused)
> +{
> + switch (machine) {
> + case EM_X86_64:
> + return 17;
> + case EM_386:
> + return 9;
> + case EM_ARM:
> + return 16;
> + case EM_AARCH64:
> + return 97;
> + case EM_CSKY:
> + return 38;
> + case EM_S390:
> + return 32;
> + case EM_PPC:
> + case EM_PPC64:
> + return 145;
> + case EM_RISCV:
> + return 66;
> + case EM_SPARC:
> + case EM_SPARCV9:
> + return 103;
> + case EM_LOONGARCH:
> + return 74;
> + default:
> + return 0;
> + }
> +}
> +
> +int get_dwarf_regnum_for_perf_regnum(int perf_regnum, unsigned int machine,
> + unsigned int flags, bool only_libdw_supported)
> +{
> + int reg;
> +
> + switch (machine) {
> + case EM_X86_64:
> + reg = __get_dwarf_regnum_for_perf_regnum_x86_64(perf_regnum);
> + break;
> + case EM_386:
> + reg = __get_dwarf_regnum_for_perf_regnum_i386(perf_regnum);
> + break;
> + default:
> + pr_err("ELF MACHINE %x is not supported.\n", machine);
> + return -ENOENT;
> + }
> + if (reg >= 0 && only_libdw_supported) {
> + int nregs = get_libdw_frame_nregs(machine, flags);
> +
> + if (reg >= nregs)
> + reg = -ENOENT;
> + }
> + return reg;
> +}
> diff --git a/tools/perf/util/include/dwarf-regs.h b/tools/perf/util/include/dwarf-regs.h
> index bb5413b0fee4..00881f1d45d6 100644
> --- a/tools/perf/util/include/dwarf-regs.h
> +++ b/tools/perf/util/include/dwarf-regs.h
> @@ -101,6 +101,8 @@ const char *get_dwarf_regstr(unsigned int n, unsigned int machine, unsigned int
>
> int __get_dwarf_regnum_i386(const char *name);
> int __get_dwarf_regnum_x86_64(const char *name);
> +int __get_dwarf_regnum_for_perf_regnum_i386(int perf_regnum);
> +int __get_dwarf_regnum_for_perf_regnum_x86_64(int perf_regnum);
>
> /*
> * get_dwarf_regnum - Returns DWARF regnum from register name
> @@ -109,6 +111,12 @@ int __get_dwarf_regnum_x86_64(const char *name);
> */
> int get_dwarf_regnum(const char *name, unsigned int machine, unsigned int flags);
>
> +/*
> + * get_dwarf_regnum - Returns DWARF regnum from perf register number.
> + */
> +int get_dwarf_regnum_for_perf_regnum(int perf_regnum, unsigned int machine, unsigned int flags,
> + bool only_libdw_supported);
> +
> void get_powerpc_regs(u32 raw_insn, int is_source, struct annotated_op_loc *op_loc);
>
> #else /* HAVE_LIBDW_SUPPORT */
> diff --git a/tools/perf/util/unwind-libdw-arch/Build b/tools/perf/util/unwind-libdw-arch/Build
> index ef17a83a7813..5b5682029953 100644
> --- a/tools/perf/util/unwind-libdw-arch/Build
> +++ b/tools/perf/util/unwind-libdw-arch/Build
> @@ -1,4 +1,3 @@
> -perf-util-y += unwind-libdw-x86.o
> perf-util-y += unwind-libdw-arm.o
> perf-util-y += unwind-libdw-arm64.o
> perf-util-y += unwind-libdw-csky.o
> diff --git a/tools/perf/util/unwind-libdw-arch/unwind-libdw-x86.c b/tools/perf/util/unwind-libdw-arch/unwind-libdw-x86.c
> deleted file mode 100644
> index dd27545a4a68..000000000000
> --- a/tools/perf/util/unwind-libdw-arch/unwind-libdw-x86.c
> +++ /dev/null
> @@ -1,54 +0,0 @@
> -// SPDX-License-Identifier: GPL-2.0
> -#include <elfutils/libdwfl.h>
> -#include "../arch/x86/include/uapi/asm/perf_regs.h"
> -#include "util/unwind-libdw.h"
> -#include "util/perf_regs.h"
> -#include "util/sample.h"
> -
> -bool libdw_set_initial_registers_x86(Dwfl_Thread *thread, void *arg)
> -{
> - struct unwind_info *ui = arg;
> - struct regs_dump *user_regs = perf_sample__user_regs(ui->sample);
> - Dwarf_Word dwarf_regs[17];
> - unsigned nregs;
> -
> -#define REG(r) ({ \
> - Dwarf_Word val = 0; \
> - perf_reg_value(&val, user_regs, PERF_REG_X86_##r); \
> - val; \
> -})
> -
> - if (user_regs->abi == PERF_SAMPLE_REGS_ABI_32) {
> - dwarf_regs[0] = REG(AX);
> - dwarf_regs[1] = REG(CX);
> - dwarf_regs[2] = REG(DX);
> - dwarf_regs[3] = REG(BX);
> - dwarf_regs[4] = REG(SP);
> - dwarf_regs[5] = REG(BP);
> - dwarf_regs[6] = REG(SI);
> - dwarf_regs[7] = REG(DI);
> - dwarf_regs[8] = REG(IP);
> - nregs = 9;
> - } else {
> - dwarf_regs[0] = REG(AX);
> - dwarf_regs[1] = REG(DX);
> - dwarf_regs[2] = REG(CX);
> - dwarf_regs[3] = REG(BX);
> - dwarf_regs[4] = REG(SI);
> - dwarf_regs[5] = REG(DI);
> - dwarf_regs[6] = REG(BP);
> - dwarf_regs[7] = REG(SP);
> - dwarf_regs[8] = REG(R8);
> - dwarf_regs[9] = REG(R9);
> - dwarf_regs[10] = REG(R10);
> - dwarf_regs[11] = REG(R11);
> - dwarf_regs[12] = REG(R12);
> - dwarf_regs[13] = REG(R13);
> - dwarf_regs[14] = REG(R14);
> - dwarf_regs[15] = REG(R15);
> - dwarf_regs[16] = REG(IP);
> - nregs = 17;
> - }
> -
> - return dwfl_thread_state_registers(thread, 0, nregs, dwarf_regs);
> -}
> diff --git a/tools/perf/util/unwind-libdw.c b/tools/perf/util/unwind-libdw.c
> index c25190cdceb4..055dab921442 100644
> --- a/tools/perf/util/unwind-libdw.c
> +++ b/tools/perf/util/unwind-libdw.c
> @@ -6,6 +6,7 @@
> #include <errno.h>
> #include "debug.h"
> #include "dso.h"
> +#include <dwarf-regs.h>
> #include "unwind.h"
> #include "unwind-libdw.h"
> #include "machine.h"
> @@ -225,6 +226,59 @@ static bool memory_read(Dwfl *dwfl __maybe_unused, Dwarf_Addr addr, Dwarf_Word *
> return true;
> }
>
> +static bool libdw_set_initial_registers_generic(Dwfl_Thread *thread, void *arg)
> +{
> + struct unwind_info *ui = arg;
> + struct regs_dump *user_regs = perf_sample__user_regs(ui->sample);
> + Dwarf_Word *dwarf_regs;
> + int max_dwarf_reg = 0;
> + bool ret;
> + uint16_t e_machine = ui->e_machine;
> + int e_flags = 0;
> + uint64_t ip_perf_reg = perf_arch_reg_ip(e_machine);
> + Dwarf_Word val = 0;
> +
> +
> + /*
> + * For every possible perf register in the bitmap determine the dwarf
> + * register and use to compute the max.
> + */
> + for (int perf_reg = 0; perf_reg < 64; perf_reg++) {
> + if (user_regs->mask & (1ULL << perf_reg)) {
> + int dwarf_reg =
> + get_dwarf_regnum_for_perf_regnum(perf_reg, e_machine,
> + e_flags,
> + /*only_libdw_supported=*/true);
> + if (dwarf_reg > max_dwarf_reg)
> + max_dwarf_reg = dwarf_reg;
> + }
> + }
> +
> + dwarf_regs = calloc(max_dwarf_reg + 1, sizeof(*dwarf_regs));
> + if (!dwarf_regs)
> + return false;
> +
> + for (int perf_reg = 0; perf_reg < 64; perf_reg++) {
> + if (user_regs->mask & (1ULL << perf_reg)) {
> + int dwarf_reg =
> + get_dwarf_regnum_for_perf_regnum(perf_reg, e_machine,
> + e_flags,
> + /*only_libdw_supported=*/true);
> + if (dwarf_reg >= 0) {
> + val = 0;
> + if (perf_reg_value(&val, user_regs, perf_reg) == 0)
> + dwarf_regs[dwarf_reg] = val;
> + }
> + }
> + }
> + if (perf_reg_value(&val, user_regs, ip_perf_reg) == 0)
> + dwfl_thread_state_register_pc(thread, val);
> +
> + ret = dwfl_thread_state_registers(thread, 0, max_dwarf_reg + 1, dwarf_regs);
I always forget something. There's possibly a TODO/alternative here
where rather than computing the max dwarf register, copying the
registers and calling dwfl_thread_state_registers, the code could read
the register then call dwfl_thread_state_registers just for that 1
register. The code as it is here is closest to the code as it was
before the change, but removing a memory allocation/free may have some
benefit.
Thanks,
Ian
> + free(dwarf_regs);
> + return ret;
> +}
> +
> #define DEFINE_DWFL_THREAD_CALLBACKS(arch) \
> static const Dwfl_Thread_Callbacks callbacks_##arch = { \
> .next_thread = next_thread, \
> @@ -232,7 +286,12 @@ static const Dwfl_Thread_Callbacks callbacks_##arch = { \
> .set_initial_registers = libdw_set_initial_registers_##arch, \
> }
>
> -DEFINE_DWFL_THREAD_CALLBACKS(x86);
> +static const Dwfl_Thread_Callbacks callbacks_generic = {
> + .next_thread = next_thread,
> + .memory_read = memory_read,
> + .set_initial_registers = libdw_set_initial_registers_generic,
> +};
> +
> DEFINE_DWFL_THREAD_CALLBACKS(arm);
> DEFINE_DWFL_THREAD_CALLBACKS(arm64);
> DEFINE_DWFL_THREAD_CALLBACKS(csky);
> @@ -257,12 +316,8 @@ static const Dwfl_Thread_Callbacks *get_thread_callbacks(const char *arch)
> return &callbacks_riscv;
> else if (!strcmp(arch, "s390"))
> return &callbacks_s390;
> - else if (!strcmp(arch, "x86"))
> - return &callbacks_x86;
>
> - pr_err("Fail to get thread callbacks for arch %s, returns NULL\n",
> - arch);
> - return NULL;
> + return &callbacks_generic;
> }
>
> static int
> @@ -301,6 +356,7 @@ int unwind__get_entries(unwind_entry_cb_t cb, void *arg,
> bool best_effort)
> {
> struct machine *machine = maps__machine(thread__maps(thread));
> + uint16_t e_machine = thread__e_machine(thread, machine);
> struct unwind_info *ui, ui_buf = {
> .sample = data,
> .thread = thread,
> @@ -308,9 +364,9 @@ int unwind__get_entries(unwind_entry_cb_t cb, void *arg,
> .cb = cb,
> .arg = arg,
> .max_stack = max_stack,
> + .e_machine = e_machine,
> .best_effort = best_effort
> };
> - uint16_t e_machine = thread__e_machine(thread, machine);
> const char *arch = perf_env__arch(machine->env);
> Dwarf_Word ip;
> int err = -EINVAL, i;
> diff --git a/tools/perf/util/unwind-libdw.h b/tools/perf/util/unwind-libdw.h
> index 574b29848cce..496e5898e7ef 100644
> --- a/tools/perf/util/unwind-libdw.h
> +++ b/tools/perf/util/unwind-libdw.h
> @@ -9,7 +9,6 @@ struct machine;
> struct perf_sample;
> struct thread;
>
> -bool libdw_set_initial_registers_x86(Dwfl_Thread *thread, void *arg);
> bool libdw_set_initial_registers_arm(Dwfl_Thread *thread, void *arg);
> bool libdw_set_initial_registers_arm64(Dwfl_Thread *thread, void *arg);
> bool libdw_set_initial_registers_csky(Dwfl_Thread *thread, void *arg);
> @@ -28,6 +27,7 @@ struct unwind_info {
> void *arg;
> int max_stack;
> int idx;
> + uint16_t e_machine;
> bool best_effort;
> struct unwind_entry entries[];
> };
> --
> 2.52.0.457.g6b5491de43-goog
>
Powered by blists - more mailing lists