| Message ID | 20191001003623.255186-1-irogers@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v3] perf tools: avoid sample_reg_masks being const + weak | expand |
On Mon, Sep 30, 2019 at 5:36 PM 'Ian Rogers' via Clang Built Linux <clang-built-linux@googlegroups.com> wrote: > > Being const + weak breaks with some compilers that constant-propagate > from the weak symbol. This behavior is outside of the specification, but > in LLVM is chosen to match GCC's behavior. > > LLVM's implementation was set in this patch: > https://github.com/llvm/llvm-project/commit/f49573d1eedcf1e44893d5a062ac1b72c8419646 > A const + weak symbol is set to be weak_odr: > https://llvm.org/docs/LangRef.html > ODR is one definition rule, and given there is one constant definition > constant-propagation is possible. It is possible to get this code to > miscompile with LLVM when applying link time optimization. As compilers > become more aggressive, this is likely to break in more instances. > > Move the definition of sample_reg_masks to the conditional part of > perf_regs.h and guard usage with HAVE_PERF_REGS_SUPPORT. This avoids the > weak symbol. > > Fix an issue when HAVE_PERF_REGS_SUPPORT isn't defined from patch v1. > In v3, add perf_regs.c for architectures that HAVE_PERF_REGS_SUPPORT but > don't declare sample_regs_masks. s/sample_regs_masks/sample_reg_masks/ (otherwise I thought for a second that my grep was broken) So powerpc and x86 set `NO_PERF_REGS := 0` AND declare `const struct sample_reg sample_reg_masks[]`. From what I can tell, it makes the below architectures match the way x86 and powerpc are structured. Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> > > Signed-off-by: Ian Rogers <irogers@google.com> > --- > tools/perf/arch/arm/util/Build | 2 ++ > tools/perf/arch/arm/util/perf_regs.c | 6 ++++++ > tools/perf/arch/arm64/util/Build | 1 + > tools/perf/arch/arm64/util/perf_regs.c | 6 ++++++ > tools/perf/arch/csky/util/Build | 2 ++ > tools/perf/arch/csky/util/perf_regs.c | 6 ++++++ > tools/perf/arch/riscv/util/Build | 2 ++ > tools/perf/arch/riscv/util/perf_regs.c | 6 ++++++ > tools/perf/arch/s390/util/Build | 1 + > tools/perf/arch/s390/util/perf_regs.c | 6 ++++++ > tools/perf/util/parse-regs-options.c | 8 ++++++-- > tools/perf/util/perf_regs.c | 4 ---- > tools/perf/util/perf_regs.h | 4 ++-- > 13 files changed, 46 insertions(+), 8 deletions(-) > create mode 100644 tools/perf/arch/arm/util/perf_regs.c > create mode 100644 tools/perf/arch/arm64/util/perf_regs.c > create mode 100644 tools/perf/arch/csky/util/perf_regs.c > create mode 100644 tools/perf/arch/riscv/util/perf_regs.c > create mode 100644 tools/perf/arch/s390/util/perf_regs.c > > diff --git a/tools/perf/arch/arm/util/Build b/tools/perf/arch/arm/util/Build > index 296f0eac5e18..37fc63708966 100644 > --- a/tools/perf/arch/arm/util/Build > +++ b/tools/perf/arch/arm/util/Build > @@ -1,3 +1,5 @@ > +perf-y += perf_regs.o > + > perf-$(CONFIG_DWARF) += dwarf-regs.o > > perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o > diff --git a/tools/perf/arch/arm/util/perf_regs.c b/tools/perf/arch/arm/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/arm/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/arm64/util/Build b/tools/perf/arch/arm64/util/Build > index 3cde540d2fcf..0a7782c61209 100644 > --- a/tools/perf/arch/arm64/util/Build > +++ b/tools/perf/arch/arm64/util/Build > @@ -1,4 +1,5 @@ > perf-y += header.o > +perf-y += perf_regs.o > perf-y += sym-handling.o > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o > diff --git a/tools/perf/arch/arm64/util/perf_regs.c b/tools/perf/arch/arm64/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/arm64/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/csky/util/Build b/tools/perf/arch/csky/util/Build > index 1160bb2332ba..7d3050134ae0 100644 > --- a/tools/perf/arch/csky/util/Build > +++ b/tools/perf/arch/csky/util/Build > @@ -1,2 +1,4 @@ > +perf-y += perf_regs.o > + > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > diff --git a/tools/perf/arch/csky/util/perf_regs.c b/tools/perf/arch/csky/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/csky/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/riscv/util/Build b/tools/perf/arch/riscv/util/Build > index 1160bb2332ba..7d3050134ae0 100644 > --- a/tools/perf/arch/riscv/util/Build > +++ b/tools/perf/arch/riscv/util/Build > @@ -1,2 +1,4 @@ > +perf-y += perf_regs.o > + > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > diff --git a/tools/perf/arch/riscv/util/perf_regs.c b/tools/perf/arch/riscv/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/riscv/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/s390/util/Build b/tools/perf/arch/s390/util/Build > index 22797f043b84..3d9d0f4f72ca 100644 > --- a/tools/perf/arch/s390/util/Build > +++ b/tools/perf/arch/s390/util/Build > @@ -1,5 +1,6 @@ > perf-y += header.o > perf-y += kvm-stat.o > +perf-y += perf_regs.o > > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > diff --git a/tools/perf/arch/s390/util/perf_regs.c b/tools/perf/arch/s390/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/s390/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/util/parse-regs-options.c b/tools/perf/util/parse-regs-options.c > index ef46c2848808..e687497b3aac 100644 > --- a/tools/perf/util/parse-regs-options.c > +++ b/tools/perf/util/parse-regs-options.c > @@ -13,7 +13,7 @@ static int > __parse_regs(const struct option *opt, const char *str, int unset, bool intr) > { > uint64_t *mode = (uint64_t *)opt->value; > - const struct sample_reg *r; > + const struct sample_reg *r = NULL; > char *s, *os = NULL, *p; > int ret = -1; > uint64_t mask; > @@ -46,19 +46,23 @@ __parse_regs(const struct option *opt, const char *str, int unset, bool intr) > > if (!strcmp(s, "?")) { > fprintf(stderr, "available registers: "); > +#ifdef HAVE_PERF_REGS_SUPPORT > for (r = sample_reg_masks; r->name; r++) { > if (r->mask & mask) > fprintf(stderr, "%s ", r->name); > } > +#endif > fputc('\n', stderr); > /* just printing available regs */ > return -1; > } > +#ifdef HAVE_PERF_REGS_SUPPORT > for (r = sample_reg_masks; r->name; r++) { > if ((r->mask & mask) && !strcasecmp(s, r->name)) > break; > } > - if (!r->name) { > +#endif > + if (!r || !r->name) { > ui__warning("Unknown register \"%s\", check man page or run \"perf record %s?\"\n", > s, intr ? "-I" : "--user-regs="); > goto error; > diff --git a/tools/perf/util/perf_regs.c b/tools/perf/util/perf_regs.c > index 2774cec1f15f..5ee47ae1509c 100644 > --- a/tools/perf/util/perf_regs.c > +++ b/tools/perf/util/perf_regs.c > @@ -3,10 +3,6 @@ > #include "perf_regs.h" > #include "event.h" > > -const struct sample_reg __weak sample_reg_masks[] = { > - SMPL_REG_END > -}; > - > int __weak arch_sdt_arg_parse_op(char *old_op __maybe_unused, > char **new_op __maybe_unused) > { > diff --git a/tools/perf/util/perf_regs.h b/tools/perf/util/perf_regs.h > index 47fe34e5f7d5..e014c2c038f4 100644 > --- a/tools/perf/util/perf_regs.h > +++ b/tools/perf/util/perf_regs.h > @@ -15,8 +15,6 @@ struct sample_reg { > #define SMPL_REG2(n, b) { .name = #n, .mask = 3ULL << (b) } > #define SMPL_REG_END { .name = NULL } > > -extern const struct sample_reg sample_reg_masks[]; > - > enum { > SDT_ARG_VALID = 0, > SDT_ARG_SKIP, > @@ -27,6 +25,8 @@ uint64_t arch__intr_reg_mask(void); > uint64_t arch__user_reg_mask(void); > > #ifdef HAVE_PERF_REGS_SUPPORT > +extern const struct sample_reg sample_reg_masks[]; > + > #include <perf_regs.h> > > #define DWARF_MINIMAL_REGS ((1ULL << PERF_REG_IP) | (1ULL << PERF_REG_SP)) > -- > 2.23.0.444.g18eeb5a265-goog > > --
On Mon, Sep 30, 2019 at 05:36:23PM -0700, Ian Rogers wrote: > Being const + weak breaks with some compilers that constant-propagate > from the weak symbol. This behavior is outside of the specification, but > in LLVM is chosen to match GCC's behavior. > > LLVM's implementation was set in this patch: > https://github.com/llvm/llvm-project/commit/f49573d1eedcf1e44893d5a062ac1b72c8419646 > A const + weak symbol is set to be weak_odr: > https://llvm.org/docs/LangRef.html > ODR is one definition rule, and given there is one constant definition > constant-propagation is possible. It is possible to get this code to > miscompile with LLVM when applying link time optimization. As compilers > become more aggressive, this is likely to break in more instances. is this just aprecaution or you actualy saw some breakage? > > Move the definition of sample_reg_masks to the conditional part of > perf_regs.h and guard usage with HAVE_PERF_REGS_SUPPORT. This avoids the > weak symbol. > > Fix an issue when HAVE_PERF_REGS_SUPPORT isn't defined from patch v1. > In v3, add perf_regs.c for architectures that HAVE_PERF_REGS_SUPPORT but > don't declare sample_regs_masks. looks good to me (again ;-)), let's see if it passes Arnaldo's farm thanks, jirka > > Signed-off-by: Ian Rogers <irogers@google.com> > --- > tools/perf/arch/arm/util/Build | 2 ++ > tools/perf/arch/arm/util/perf_regs.c | 6 ++++++ > tools/perf/arch/arm64/util/Build | 1 + > tools/perf/arch/arm64/util/perf_regs.c | 6 ++++++ > tools/perf/arch/csky/util/Build | 2 ++ > tools/perf/arch/csky/util/perf_regs.c | 6 ++++++ > tools/perf/arch/riscv/util/Build | 2 ++ > tools/perf/arch/riscv/util/perf_regs.c | 6 ++++++ > tools/perf/arch/s390/util/Build | 1 + > tools/perf/arch/s390/util/perf_regs.c | 6 ++++++ > tools/perf/util/parse-regs-options.c | 8 ++++++-- > tools/perf/util/perf_regs.c | 4 ---- > tools/perf/util/perf_regs.h | 4 ++-- > 13 files changed, 46 insertions(+), 8 deletions(-) > create mode 100644 tools/perf/arch/arm/util/perf_regs.c > create mode 100644 tools/perf/arch/arm64/util/perf_regs.c > create mode 100644 tools/perf/arch/csky/util/perf_regs.c > create mode 100644 tools/perf/arch/riscv/util/perf_regs.c > create mode 100644 tools/perf/arch/s390/util/perf_regs.c > > diff --git a/tools/perf/arch/arm/util/Build b/tools/perf/arch/arm/util/Build > index 296f0eac5e18..37fc63708966 100644 > --- a/tools/perf/arch/arm/util/Build > +++ b/tools/perf/arch/arm/util/Build > @@ -1,3 +1,5 @@ > +perf-y += perf_regs.o > + > perf-$(CONFIG_DWARF) += dwarf-regs.o > > perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o > diff --git a/tools/perf/arch/arm/util/perf_regs.c b/tools/perf/arch/arm/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/arm/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/arm64/util/Build b/tools/perf/arch/arm64/util/Build > index 3cde540d2fcf..0a7782c61209 100644 > --- a/tools/perf/arch/arm64/util/Build > +++ b/tools/perf/arch/arm64/util/Build > @@ -1,4 +1,5 @@ > perf-y += header.o > +perf-y += perf_regs.o > perf-y += sym-handling.o > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o > diff --git a/tools/perf/arch/arm64/util/perf_regs.c b/tools/perf/arch/arm64/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/arm64/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/csky/util/Build b/tools/perf/arch/csky/util/Build > index 1160bb2332ba..7d3050134ae0 100644 > --- a/tools/perf/arch/csky/util/Build > +++ b/tools/perf/arch/csky/util/Build > @@ -1,2 +1,4 @@ > +perf-y += perf_regs.o > + > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > diff --git a/tools/perf/arch/csky/util/perf_regs.c b/tools/perf/arch/csky/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/csky/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/riscv/util/Build b/tools/perf/arch/riscv/util/Build > index 1160bb2332ba..7d3050134ae0 100644 > --- a/tools/perf/arch/riscv/util/Build > +++ b/tools/perf/arch/riscv/util/Build > @@ -1,2 +1,4 @@ > +perf-y += perf_regs.o > + > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > diff --git a/tools/perf/arch/riscv/util/perf_regs.c b/tools/perf/arch/riscv/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/riscv/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/arch/s390/util/Build b/tools/perf/arch/s390/util/Build > index 22797f043b84..3d9d0f4f72ca 100644 > --- a/tools/perf/arch/s390/util/Build > +++ b/tools/perf/arch/s390/util/Build > @@ -1,5 +1,6 @@ > perf-y += header.o > perf-y += kvm-stat.o > +perf-y += perf_regs.o > > perf-$(CONFIG_DWARF) += dwarf-regs.o > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > diff --git a/tools/perf/arch/s390/util/perf_regs.c b/tools/perf/arch/s390/util/perf_regs.c > new file mode 100644 > index 000000000000..2864e2e3776d > --- /dev/null > +++ b/tools/perf/arch/s390/util/perf_regs.c > @@ -0,0 +1,6 @@ > +// SPDX-License-Identifier: GPL-2.0 > +#include "../../util/perf_regs.h" > + > +const struct sample_reg sample_reg_masks[] = { > + SMPL_REG_END > +}; > diff --git a/tools/perf/util/parse-regs-options.c b/tools/perf/util/parse-regs-options.c > index ef46c2848808..e687497b3aac 100644 > --- a/tools/perf/util/parse-regs-options.c > +++ b/tools/perf/util/parse-regs-options.c > @@ -13,7 +13,7 @@ static int > __parse_regs(const struct option *opt, const char *str, int unset, bool intr) > { > uint64_t *mode = (uint64_t *)opt->value; > - const struct sample_reg *r; > + const struct sample_reg *r = NULL; > char *s, *os = NULL, *p; > int ret = -1; > uint64_t mask; > @@ -46,19 +46,23 @@ __parse_regs(const struct option *opt, const char *str, int unset, bool intr) > > if (!strcmp(s, "?")) { > fprintf(stderr, "available registers: "); > +#ifdef HAVE_PERF_REGS_SUPPORT > for (r = sample_reg_masks; r->name; r++) { > if (r->mask & mask) > fprintf(stderr, "%s ", r->name); > } > +#endif > fputc('\n', stderr); > /* just printing available regs */ > return -1; > } > +#ifdef HAVE_PERF_REGS_SUPPORT > for (r = sample_reg_masks; r->name; r++) { > if ((r->mask & mask) && !strcasecmp(s, r->name)) > break; > } > - if (!r->name) { > +#endif > + if (!r || !r->name) { > ui__warning("Unknown register \"%s\", check man page or run \"perf record %s?\"\n", > s, intr ? "-I" : "--user-regs="); > goto error; > diff --git a/tools/perf/util/perf_regs.c b/tools/perf/util/perf_regs.c > index 2774cec1f15f..5ee47ae1509c 100644 > --- a/tools/perf/util/perf_regs.c > +++ b/tools/perf/util/perf_regs.c > @@ -3,10 +3,6 @@ > #include "perf_regs.h" > #include "event.h" > > -const struct sample_reg __weak sample_reg_masks[] = { > - SMPL_REG_END > -}; > - > int __weak arch_sdt_arg_parse_op(char *old_op __maybe_unused, > char **new_op __maybe_unused) > { > diff --git a/tools/perf/util/perf_regs.h b/tools/perf/util/perf_regs.h > index 47fe34e5f7d5..e014c2c038f4 100644 > --- a/tools/perf/util/perf_regs.h > +++ b/tools/perf/util/perf_regs.h > @@ -15,8 +15,6 @@ struct sample_reg { > #define SMPL_REG2(n, b) { .name = #n, .mask = 3ULL << (b) } > #define SMPL_REG_END { .name = NULL } > > -extern const struct sample_reg sample_reg_masks[]; > - > enum { > SDT_ARG_VALID = 0, > SDT_ARG_SKIP, > @@ -27,6 +25,8 @@ uint64_t arch__intr_reg_mask(void); > uint64_t arch__user_reg_mask(void); > > #ifdef HAVE_PERF_REGS_SUPPORT > +extern const struct sample_reg sample_reg_masks[]; > + > #include <perf_regs.h> > > #define DWARF_MINIMAL_REGS ((1ULL << PERF_REG_IP) | (1ULL << PERF_REG_SP)) > -- > 2.23.0.444.g18eeb5a265-goog >
On Tue, Oct 8, 2019 at 5:31 AM Jiri Olsa <jolsa@redhat.com> wrote: > > On Mon, Sep 30, 2019 at 05:36:23PM -0700, Ian Rogers wrote: > > Being const + weak breaks with some compilers that constant-propagate > > from the weak symbol. This behavior is outside of the specification, but > > in LLVM is chosen to match GCC's behavior. > > > > LLVM's implementation was set in this patch: > > https://github.com/llvm/llvm-project/commit/f49573d1eedcf1e44893d5a062ac1b72c8419646 > > A const + weak symbol is set to be weak_odr: > > https://llvm.org/docs/LangRef.html > > ODR is one definition rule, and given there is one constant definition > > constant-propagation is possible. It is possible to get this code to > > miscompile with LLVM when applying link time optimization. As compilers > > become more aggressive, this is likely to break in more instances. > > is this just aprecaution or you actualy saw some breakage? We saw a breakage with clang with thinlto enabled for linking. Our compiler team had recently seen, and were surprised by, a similar issue and were able to dig out the weak ODR issue. Thanks, Ian > > Move the definition of sample_reg_masks to the conditional part of > > perf_regs.h and guard usage with HAVE_PERF_REGS_SUPPORT. This avoids the > > weak symbol. > > > > Fix an issue when HAVE_PERF_REGS_SUPPORT isn't defined from patch v1. > > In v3, add perf_regs.c for architectures that HAVE_PERF_REGS_SUPPORT but > > don't declare sample_regs_masks. > > looks good to me (again ;-)), let's see if it passes Arnaldo's farm > > thanks, > jirka > > > > > Signed-off-by: Ian Rogers <irogers@google.com> > > --- > > tools/perf/arch/arm/util/Build | 2 ++ > > tools/perf/arch/arm/util/perf_regs.c | 6 ++++++ > > tools/perf/arch/arm64/util/Build | 1 + > > tools/perf/arch/arm64/util/perf_regs.c | 6 ++++++ > > tools/perf/arch/csky/util/Build | 2 ++ > > tools/perf/arch/csky/util/perf_regs.c | 6 ++++++ > > tools/perf/arch/riscv/util/Build | 2 ++ > > tools/perf/arch/riscv/util/perf_regs.c | 6 ++++++ > > tools/perf/arch/s390/util/Build | 1 + > > tools/perf/arch/s390/util/perf_regs.c | 6 ++++++ > > tools/perf/util/parse-regs-options.c | 8 ++++++-- > > tools/perf/util/perf_regs.c | 4 ---- > > tools/perf/util/perf_regs.h | 4 ++-- > > 13 files changed, 46 insertions(+), 8 deletions(-) > > create mode 100644 tools/perf/arch/arm/util/perf_regs.c > > create mode 100644 tools/perf/arch/arm64/util/perf_regs.c > > create mode 100644 tools/perf/arch/csky/util/perf_regs.c > > create mode 100644 tools/perf/arch/riscv/util/perf_regs.c > > create mode 100644 tools/perf/arch/s390/util/perf_regs.c > > > > diff --git a/tools/perf/arch/arm/util/Build b/tools/perf/arch/arm/util/Build > > index 296f0eac5e18..37fc63708966 100644 > > --- a/tools/perf/arch/arm/util/Build > > +++ b/tools/perf/arch/arm/util/Build > > @@ -1,3 +1,5 @@ > > +perf-y += perf_regs.o > > + > > perf-$(CONFIG_DWARF) += dwarf-regs.o > > > > perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o > > diff --git a/tools/perf/arch/arm/util/perf_regs.c b/tools/perf/arch/arm/util/perf_regs.c > > new file mode 100644 > > index 000000000000..2864e2e3776d > > --- /dev/null > > +++ b/tools/perf/arch/arm/util/perf_regs.c > > @@ -0,0 +1,6 @@ > > +// SPDX-License-Identifier: GPL-2.0 > > +#include "../../util/perf_regs.h" > > + > > +const struct sample_reg sample_reg_masks[] = { > > + SMPL_REG_END > > +}; > > diff --git a/tools/perf/arch/arm64/util/Build b/tools/perf/arch/arm64/util/Build > > index 3cde540d2fcf..0a7782c61209 100644 > > --- a/tools/perf/arch/arm64/util/Build > > +++ b/tools/perf/arch/arm64/util/Build > > @@ -1,4 +1,5 @@ > > perf-y += header.o > > +perf-y += perf_regs.o > > perf-y += sym-handling.o > > perf-$(CONFIG_DWARF) += dwarf-regs.o > > perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o > > diff --git a/tools/perf/arch/arm64/util/perf_regs.c b/tools/perf/arch/arm64/util/perf_regs.c > > new file mode 100644 > > index 000000000000..2864e2e3776d > > --- /dev/null > > +++ b/tools/perf/arch/arm64/util/perf_regs.c > > @@ -0,0 +1,6 @@ > > +// SPDX-License-Identifier: GPL-2.0 > > +#include "../../util/perf_regs.h" > > + > > +const struct sample_reg sample_reg_masks[] = { > > + SMPL_REG_END > > +}; > > diff --git a/tools/perf/arch/csky/util/Build b/tools/perf/arch/csky/util/Build > > index 1160bb2332ba..7d3050134ae0 100644 > > --- a/tools/perf/arch/csky/util/Build > > +++ b/tools/perf/arch/csky/util/Build > > @@ -1,2 +1,4 @@ > > +perf-y += perf_regs.o > > + > > perf-$(CONFIG_DWARF) += dwarf-regs.o > > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > > diff --git a/tools/perf/arch/csky/util/perf_regs.c b/tools/perf/arch/csky/util/perf_regs.c > > new file mode 100644 > > index 000000000000..2864e2e3776d > > --- /dev/null > > +++ b/tools/perf/arch/csky/util/perf_regs.c > > @@ -0,0 +1,6 @@ > > +// SPDX-License-Identifier: GPL-2.0 > > +#include "../../util/perf_regs.h" > > + > > +const struct sample_reg sample_reg_masks[] = { > > + SMPL_REG_END > > +}; > > diff --git a/tools/perf/arch/riscv/util/Build b/tools/perf/arch/riscv/util/Build > > index 1160bb2332ba..7d3050134ae0 100644 > > --- a/tools/perf/arch/riscv/util/Build > > +++ b/tools/perf/arch/riscv/util/Build > > @@ -1,2 +1,4 @@ > > +perf-y += perf_regs.o > > + > > perf-$(CONFIG_DWARF) += dwarf-regs.o > > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > > diff --git a/tools/perf/arch/riscv/util/perf_regs.c b/tools/perf/arch/riscv/util/perf_regs.c > > new file mode 100644 > > index 000000000000..2864e2e3776d > > --- /dev/null > > +++ b/tools/perf/arch/riscv/util/perf_regs.c > > @@ -0,0 +1,6 @@ > > +// SPDX-License-Identifier: GPL-2.0 > > +#include "../../util/perf_regs.h" > > + > > +const struct sample_reg sample_reg_masks[] = { > > + SMPL_REG_END > > +}; > > diff --git a/tools/perf/arch/s390/util/Build b/tools/perf/arch/s390/util/Build > > index 22797f043b84..3d9d0f4f72ca 100644 > > --- a/tools/perf/arch/s390/util/Build > > +++ b/tools/perf/arch/s390/util/Build > > @@ -1,5 +1,6 @@ > > perf-y += header.o > > perf-y += kvm-stat.o > > +perf-y += perf_regs.o > > > > perf-$(CONFIG_DWARF) += dwarf-regs.o > > perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o > > diff --git a/tools/perf/arch/s390/util/perf_regs.c b/tools/perf/arch/s390/util/perf_regs.c > > new file mode 100644 > > index 000000000000..2864e2e3776d > > --- /dev/null > > +++ b/tools/perf/arch/s390/util/perf_regs.c > > @@ -0,0 +1,6 @@ > > +// SPDX-License-Identifier: GPL-2.0 > > +#include "../../util/perf_regs.h" > > + > > +const struct sample_reg sample_reg_masks[] = { > > + SMPL_REG_END > > +}; > > diff --git a/tools/perf/util/parse-regs-options.c b/tools/perf/util/parse-regs-options.c > > index ef46c2848808..e687497b3aac 100644 > > --- a/tools/perf/util/parse-regs-options.c > > +++ b/tools/perf/util/parse-regs-options.c > > @@ -13,7 +13,7 @@ static int > > __parse_regs(const struct option *opt, const char *str, int unset, bool intr) > > { > > uint64_t *mode = (uint64_t *)opt->value; > > - const struct sample_reg *r; > > + const struct sample_reg *r = NULL; > > char *s, *os = NULL, *p; > > int ret = -1; > > uint64_t mask; > > @@ -46,19 +46,23 @@ __parse_regs(const struct option *opt, const char *str, int unset, bool intr) > > > > if (!strcmp(s, "?")) { > > fprintf(stderr, "available registers: "); > > +#ifdef HAVE_PERF_REGS_SUPPORT > > for (r = sample_reg_masks; r->name; r++) { > > if (r->mask & mask) > > fprintf(stderr, "%s ", r->name); > > } > > +#endif > > fputc('\n', stderr); > > /* just printing available regs */ > > return -1; > > } > > +#ifdef HAVE_PERF_REGS_SUPPORT > > for (r = sample_reg_masks; r->name; r++) { > > if ((r->mask & mask) && !strcasecmp(s, r->name)) > > break; > > } > > - if (!r->name) { > > +#endif > > + if (!r || !r->name) { > > ui__warning("Unknown register \"%s\", check man page or run \"perf record %s?\"\n", > > s, intr ? "-I" : "--user-regs="); > > goto error; > > diff --git a/tools/perf/util/perf_regs.c b/tools/perf/util/perf_regs.c > > index 2774cec1f15f..5ee47ae1509c 100644 > > --- a/tools/perf/util/perf_regs.c > > +++ b/tools/perf/util/perf_regs.c > > @@ -3,10 +3,6 @@ > > #include "perf_regs.h" > > #include "event.h" > > > > -const struct sample_reg __weak sample_reg_masks[] = { > > - SMPL_REG_END > > -}; > > - > > int __weak arch_sdt_arg_parse_op(char *old_op __maybe_unused, > > char **new_op __maybe_unused) > > { > > diff --git a/tools/perf/util/perf_regs.h b/tools/perf/util/perf_regs.h > > index 47fe34e5f7d5..e014c2c038f4 100644 > > --- a/tools/perf/util/perf_regs.h > > +++ b/tools/perf/util/perf_regs.h > > @@ -15,8 +15,6 @@ struct sample_reg { > > #define SMPL_REG2(n, b) { .name = #n, .mask = 3ULL << (b) } > > #define SMPL_REG_END { .name = NULL } > > > > -extern const struct sample_reg sample_reg_masks[]; > > - > > enum { > > SDT_ARG_VALID = 0, > > SDT_ARG_SKIP, > > @@ -27,6 +25,8 @@ uint64_t arch__intr_reg_mask(void); > > uint64_t arch__user_reg_mask(void); > > > > #ifdef HAVE_PERF_REGS_SUPPORT > > +extern const struct sample_reg sample_reg_masks[]; > > + > > #include <perf_regs.h> > > > > #define DWARF_MINIMAL_REGS ((1ULL << PERF_REG_IP) | (1ULL << PERF_REG_SP)) > > -- > > 2.23.0.444.g18eeb5a265-goog > >
Em Wed, Oct 09, 2019 at 04:07:37PM -0700, Ian Rogers escreveu: > On Tue, Oct 8, 2019 at 5:31 AM Jiri Olsa <jolsa@redhat.com> wrote: > > On Mon, Sep 30, 2019 at 05:36:23PM -0700, Ian Rogers wrote: > > > Being const + weak breaks with some compilers that constant-propagate > > > from the weak symbol. This behavior is outside of the specification, but > > > in LLVM is chosen to match GCC's behavior. > > > > > > LLVM's implementation was set in this patch: > > > https://github.com/llvm/llvm-project/commit/f49573d1eedcf1e44893d5a062ac1b72c8419646 > > > A const + weak symbol is set to be weak_odr: > > > https://llvm.org/docs/LangRef.html > > > ODR is one definition rule, and given there is one constant definition > > > constant-propagation is possible. It is possible to get this code to > > > miscompile with LLVM when applying link time optimization. As compilers > > > become more aggressive, this is likely to break in more instances. > > is this just aprecaution or you actualy saw some breakage? > We saw a breakage with clang with thinlto enabled for linking. Our > compiler team had recently seen, and were surprised by, a similar > issue and were able to dig out the weak ODR issue. This is useful info, I'll add it to the commit log message. > > > Move the definition of sample_reg_masks to the conditional part of > > > perf_regs.h and guard usage with HAVE_PERF_REGS_SUPPORT. This avoids the > > > weak symbol. > > > Fix an issue when HAVE_PERF_REGS_SUPPORT isn't defined from patch v1. > > > In v3, add perf_regs.c for architectures that HAVE_PERF_REGS_SUPPORT but > > > don't declare sample_regs_masks. > > looks good to me (again ;-)), let's see if it passes Arnaldo's farm It passed a few of the usual places where things like this break, I'll submit it to a full set of build environments soon, together with what is sitting in acme/perf/core. Thanks, - Arnaldo
diff --git a/tools/perf/arch/arm/util/Build b/tools/perf/arch/arm/util/Build index 296f0eac5e18..37fc63708966 100644 --- a/tools/perf/arch/arm/util/Build +++ b/tools/perf/arch/arm/util/Build @@ -1,3 +1,5 @@ +perf-y += perf_regs.o + perf-$(CONFIG_DWARF) += dwarf-regs.o perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o diff --git a/tools/perf/arch/arm/util/perf_regs.c b/tools/perf/arch/arm/util/perf_regs.c new file mode 100644 index 000000000000..2864e2e3776d --- /dev/null +++ b/tools/perf/arch/arm/util/perf_regs.c @@ -0,0 +1,6 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "../../util/perf_regs.h" + +const struct sample_reg sample_reg_masks[] = { + SMPL_REG_END +}; diff --git a/tools/perf/arch/arm64/util/Build b/tools/perf/arch/arm64/util/Build index 3cde540d2fcf..0a7782c61209 100644 --- a/tools/perf/arch/arm64/util/Build +++ b/tools/perf/arch/arm64/util/Build @@ -1,4 +1,5 @@ perf-y += header.o +perf-y += perf_regs.o perf-y += sym-handling.o perf-$(CONFIG_DWARF) += dwarf-regs.o perf-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o diff --git a/tools/perf/arch/arm64/util/perf_regs.c b/tools/perf/arch/arm64/util/perf_regs.c new file mode 100644 index 000000000000..2864e2e3776d --- /dev/null +++ b/tools/perf/arch/arm64/util/perf_regs.c @@ -0,0 +1,6 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "../../util/perf_regs.h" + +const struct sample_reg sample_reg_masks[] = { + SMPL_REG_END +}; diff --git a/tools/perf/arch/csky/util/Build b/tools/perf/arch/csky/util/Build index 1160bb2332ba..7d3050134ae0 100644 --- a/tools/perf/arch/csky/util/Build +++ b/tools/perf/arch/csky/util/Build @@ -1,2 +1,4 @@ +perf-y += perf_regs.o + perf-$(CONFIG_DWARF) += dwarf-regs.o perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o diff --git a/tools/perf/arch/csky/util/perf_regs.c b/tools/perf/arch/csky/util/perf_regs.c new file mode 100644 index 000000000000..2864e2e3776d --- /dev/null +++ b/tools/perf/arch/csky/util/perf_regs.c @@ -0,0 +1,6 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "../../util/perf_regs.h" + +const struct sample_reg sample_reg_masks[] = { + SMPL_REG_END +}; diff --git a/tools/perf/arch/riscv/util/Build b/tools/perf/arch/riscv/util/Build index 1160bb2332ba..7d3050134ae0 100644 --- a/tools/perf/arch/riscv/util/Build +++ b/tools/perf/arch/riscv/util/Build @@ -1,2 +1,4 @@ +perf-y += perf_regs.o + perf-$(CONFIG_DWARF) += dwarf-regs.o perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o diff --git a/tools/perf/arch/riscv/util/perf_regs.c b/tools/perf/arch/riscv/util/perf_regs.c new file mode 100644 index 000000000000..2864e2e3776d --- /dev/null +++ b/tools/perf/arch/riscv/util/perf_regs.c @@ -0,0 +1,6 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "../../util/perf_regs.h" + +const struct sample_reg sample_reg_masks[] = { + SMPL_REG_END +}; diff --git a/tools/perf/arch/s390/util/Build b/tools/perf/arch/s390/util/Build index 22797f043b84..3d9d0f4f72ca 100644 --- a/tools/perf/arch/s390/util/Build +++ b/tools/perf/arch/s390/util/Build @@ -1,5 +1,6 @@ perf-y += header.o perf-y += kvm-stat.o +perf-y += perf_regs.o perf-$(CONFIG_DWARF) += dwarf-regs.o perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o diff --git a/tools/perf/arch/s390/util/perf_regs.c b/tools/perf/arch/s390/util/perf_regs.c new file mode 100644 index 000000000000..2864e2e3776d --- /dev/null +++ b/tools/perf/arch/s390/util/perf_regs.c @@ -0,0 +1,6 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "../../util/perf_regs.h" + +const struct sample_reg sample_reg_masks[] = { + SMPL_REG_END +}; diff --git a/tools/perf/util/parse-regs-options.c b/tools/perf/util/parse-regs-options.c index ef46c2848808..e687497b3aac 100644 --- a/tools/perf/util/parse-regs-options.c +++ b/tools/perf/util/parse-regs-options.c @@ -13,7 +13,7 @@ static int __parse_regs(const struct option *opt, const char *str, int unset, bool intr) { uint64_t *mode = (uint64_t *)opt->value; - const struct sample_reg *r; + const struct sample_reg *r = NULL; char *s, *os = NULL, *p; int ret = -1; uint64_t mask; @@ -46,19 +46,23 @@ __parse_regs(const struct option *opt, const char *str, int unset, bool intr) if (!strcmp(s, "?")) { fprintf(stderr, "available registers: "); +#ifdef HAVE_PERF_REGS_SUPPORT for (r = sample_reg_masks; r->name; r++) { if (r->mask & mask) fprintf(stderr, "%s ", r->name); } +#endif fputc('\n', stderr); /* just printing available regs */ return -1; } +#ifdef HAVE_PERF_REGS_SUPPORT for (r = sample_reg_masks; r->name; r++) { if ((r->mask & mask) && !strcasecmp(s, r->name)) break; } - if (!r->name) { +#endif + if (!r || !r->name) { ui__warning("Unknown register \"%s\", check man page or run \"perf record %s?\"\n", s, intr ? "-I" : "--user-regs="); goto error; diff --git a/tools/perf/util/perf_regs.c b/tools/perf/util/perf_regs.c index 2774cec1f15f..5ee47ae1509c 100644 --- a/tools/perf/util/perf_regs.c +++ b/tools/perf/util/perf_regs.c @@ -3,10 +3,6 @@ #include "perf_regs.h" #include "event.h" -const struct sample_reg __weak sample_reg_masks[] = { - SMPL_REG_END -}; - int __weak arch_sdt_arg_parse_op(char *old_op __maybe_unused, char **new_op __maybe_unused) { diff --git a/tools/perf/util/perf_regs.h b/tools/perf/util/perf_regs.h index 47fe34e5f7d5..e014c2c038f4 100644 --- a/tools/perf/util/perf_regs.h +++ b/tools/perf/util/perf_regs.h @@ -15,8 +15,6 @@ struct sample_reg { #define SMPL_REG2(n, b) { .name = #n, .mask = 3ULL << (b) } #define SMPL_REG_END { .name = NULL } -extern const struct sample_reg sample_reg_masks[]; - enum { SDT_ARG_VALID = 0, SDT_ARG_SKIP, @@ -27,6 +25,8 @@ uint64_t arch__intr_reg_mask(void); uint64_t arch__user_reg_mask(void); #ifdef HAVE_PERF_REGS_SUPPORT +extern const struct sample_reg sample_reg_masks[]; + #include <perf_regs.h> #define DWARF_MINIMAL_REGS ((1ULL << PERF_REG_IP) | (1ULL << PERF_REG_SP))
Being const + weak breaks with some compilers that constant-propagate from the weak symbol. This behavior is outside of the specification, but in LLVM is chosen to match GCC's behavior. LLVM's implementation was set in this patch: https://github.com/llvm/llvm-project/commit/f49573d1eedcf1e44893d5a062ac1b72c8419646 A const + weak symbol is set to be weak_odr: https://llvm.org/docs/LangRef.html ODR is one definition rule, and given there is one constant definition constant-propagation is possible. It is possible to get this code to miscompile with LLVM when applying link time optimization. As compilers become more aggressive, this is likely to break in more instances. Move the definition of sample_reg_masks to the conditional part of perf_regs.h and guard usage with HAVE_PERF_REGS_SUPPORT. This avoids the weak symbol. Fix an issue when HAVE_PERF_REGS_SUPPORT isn't defined from patch v1. In v3, add perf_regs.c for architectures that HAVE_PERF_REGS_SUPPORT but don't declare sample_regs_masks. Signed-off-by: Ian Rogers <irogers@google.com> --- tools/perf/arch/arm/util/Build | 2 ++ tools/perf/arch/arm/util/perf_regs.c | 6 ++++++ tools/perf/arch/arm64/util/Build | 1 + tools/perf/arch/arm64/util/perf_regs.c | 6 ++++++ tools/perf/arch/csky/util/Build | 2 ++ tools/perf/arch/csky/util/perf_regs.c | 6 ++++++ tools/perf/arch/riscv/util/Build | 2 ++ tools/perf/arch/riscv/util/perf_regs.c | 6 ++++++ tools/perf/arch/s390/util/Build | 1 + tools/perf/arch/s390/util/perf_regs.c | 6 ++++++ tools/perf/util/parse-regs-options.c | 8 ++++++-- tools/perf/util/perf_regs.c | 4 ---- tools/perf/util/perf_regs.h | 4 ++-- 13 files changed, 46 insertions(+), 8 deletions(-) create mode 100644 tools/perf/arch/arm/util/perf_regs.c create mode 100644 tools/perf/arch/arm64/util/perf_regs.c create mode 100644 tools/perf/arch/csky/util/perf_regs.c create mode 100644 tools/perf/arch/riscv/util/perf_regs.c create mode 100644 tools/perf/arch/s390/util/perf_regs.c