Libdw/dwarf build clean up
Subsystems (4)
Ian Rogers
reviewercover Libdw/dwarf build clean up
There patches are on top of:
https://lore.kernel.org/lkml/[email protected]/
where it was pointed out that a lot of the libdw conditional compilation was due to features that have now been standard for 10 or 15 years. The patches remove the conditional compilation assuming the features are in libdw where the feature test is expanded to check there pressence.
In the Makefile code dwarf tends to mean unwind or libdw support for dwarf things. To make it clearer when dwarf really just means libdw numerous build variables and defines are renamed.
There is some tech debt in the changes as perf check still reports the values using the old name and for features that are no longer tested.
Ian Rogers (11):
perf build: Rename NO_DWARF to NO_LIBDW
perf build: Remove defined but never used variable perf build: Rename test-dwarf to test-libdw
perf build: Combine libdw-dwarf-unwind into libdw feature tests perf build: Combine test-dwarf-getlocations into test-libdw perf build: Combine test-dwarf-getcfi into test-libdw perf probe: Move elfutils support check to libdw check perf libdw: Remove unnecessary defines
perf build: Rename HAVE_DWARF_SUPPORT to HAVE_LIBDW_SUPPORT perf build: Rename CONFIG_DWARF to CONFIG_LIBDW
perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
tools/build/Makefile.feature | 11 +--- tools/build/feature/Makefile | 24 ++------ tools/build/feature/test-all.c | 16 +----- tools/build/feature/test-dwarf.c | 11 ---- tools/build/feature/test-dwarf_getcfi.c | 9 --- tools/build/feature/test-dwarf_getlocations.c | 13 ----- tools/build/feature/test-libdw-dwarf-unwind.c | 14 ----- tools/build/feature/test-libdw.c | 56 ++++++++++++++++++ tools/perf/Documentation/perf-check.txt | 6 +- tools/perf/Makefile.config | 57 ++++++------------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 4 +- tools/perf/arch/arm/util/Build | 2 +- tools/perf/arch/arm64/Makefile | 4 +- tools/perf/arch/arm64/util/Build | 2 +- tools/perf/arch/csky/Makefile | 4 +- tools/perf/arch/csky/util/Build | 2 +- tools/perf/arch/loongarch/Makefile | 4 +- tools/perf/arch/loongarch/util/Build | 2 +- tools/perf/arch/mips/Makefile | 4 +- tools/perf/arch/mips/util/Build | 2 +- tools/perf/arch/powerpc/Makefile | 4 +- .../perf/arch/powerpc/annotate/instructions.c | 4 +- tools/perf/arch/powerpc/util/Build | 4 +- tools/perf/arch/riscv/Makefile | 4 +- tools/perf/arch/riscv/util/Build | 2 +- tools/perf/arch/s390/Makefile | 4 +- tools/perf/arch/s390/util/Build | 2 +- tools/perf/arch/sh/Makefile | 4 +- tools/perf/arch/sh/util/Build | 2 +- tools/perf/arch/sparc/Makefile | 4 +- tools/perf/arch/sparc/util/Build | 2 +- tools/perf/arch/x86/Makefile | 4 +- tools/perf/arch/x86/annotate/instructions.c | 2 +- tools/perf/arch/x86/util/Build | 2 +- tools/perf/arch/xtensa/Makefile | 4 +- tools/perf/arch/xtensa/util/Build | 2 +- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 6 +- tools/perf/builtin-probe.c | 14 ++--- tools/perf/builtin-report.c | 4 +- tools/perf/util/Build | 12 ++-- tools/perf/util/annotate-data.h | 8 +-- tools/perf/util/debuginfo.h | 6 +- tools/perf/util/disasm.c | 4 +- tools/perf/util/disasm.h | 4 +- tools/perf/util/dwarf-aux.c | 6 -- tools/perf/util/dwarf-aux.h | 54 ------------------ tools/perf/util/genelf.c | 4 +- tools/perf/util/genelf.h | 2 +- tools/perf/util/include/dwarf-regs.h | 6 +- tools/perf/util/probe-event.c | 4 +- tools/perf/util/probe-finder.c | 6 -- tools/perf/util/probe-finder.h | 6 +- 54 files changed, 168 insertions(+), 279 deletions(-) delete mode 100644 tools/build/feature/test-dwarf.c delete mode 100644 tools/build/feature/test-dwarf_getcfi.c delete mode 100644 tools/build/feature/test-dwarf_getlocations.c delete mode 100644 tools/build/feature/test-libdw-dwarf-unwind.c create mode 100644 tools/build/feature/test-libdw.c
Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerPATCH 1/11 perf build: Rename NO_DWARF to NO_LIBDW
NO_DWARF could mean more than NO_LIBDW support, in particular no libunwind support. Rename to be more intention revealing.
--- tools/perf/Makefile.config | 16 ++++++++-------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- tools/perf/builtin-probe.c | 2 +- 15 files changed, 22 insertions(+), 22 deletions(-)
tools/perf/Makefile.config +8 -8
@@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1) endif ifdef NO_LIBELF - NO_DWARF := 1 + NO_LIBDW := 1 NO_LIBUNWIND := 1 NO_LIBDW_DWARF_UNWIND := 1 NO_LIBBPF := 1
@@ -471,9 +471,9 @@ else endif endif ifneq ($(feature-dwarf), 1) - ifndef NO_DWARF + ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) - NO_DWARF := 1 + NO_LIBDW := 1 endif else ifneq ($(feature-dwarf_getlocations), 1)
@@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1) endif endif -ifdef NO_DWARF +ifdef NO_LIBDW NO_LIBDW_DWARF_UNWIND := 1 endif
@@ -574,17 +574,17 @@ ifndef NO_LIBELF endif endif - ifndef NO_DWARF + ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) - NO_DWARF := 1 + NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGS - endif # NO_DWARF + endif # NO_LIBDW ifndef NO_LIBBPF ifeq ($(feature-bpf), 1)
@@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP endif ifeq ($(SRCARCH),powerpc) - ifndef NO_DWARF + ifndef NO_LIBDW CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX endif endif
tools/perf/Makefile.perf +1 -1
@@ -40,7 +40,7 @@ include ../scripts/utilities.mak # # Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS. # -# Define NO_DWARF if you do not want debug-info analysis feature at all. +# Define NO_LIBDW if you do not want debug-info analysis feature at all. # # Define WERROR=0 to disable treating any warnings as errors. #
tools/perf/arch/arm/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1
tools/perf/arch/arm64/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1
tools/perf/arch/csky/Makefile +1 -1
@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif
tools/perf/arch/loongarch/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1
tools/perf/arch/mips/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif
tools/perf/arch/powerpc/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif
tools/perf/arch/riscv/Makefile +1 -1
@@ -1,4 +1,4 @@ -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1
tools/perf/arch/s390/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1
tools/perf/arch/sh/Makefile +1 -1
@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif
tools/perf/arch/sparc/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif
tools/perf/arch/x86/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1
tools/perf/arch/xtensa/Makefile +1 -1
@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif
tools/perf/builtin-probe.c +1 -1
@@ -616,7 +616,7 @@ __cmd_probe(int argc, const char **argv)set_option_flag (options,'L' ,"line" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'V' ,"vars" ,PARSE_OPT_EXCLUSIVE );#else -# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_DWARF=1", c) +# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_LIBDW=1", c)set_nobuild ('L' ,"line" , false);set_nobuild ('V' ,"vars" , false);set_nobuild ('\0' ,"externs" , false);
Signature · 1 line
2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 01/11] perf build: Rename NO_DWARF to NO_LIBDW
Quoted 91 lines · Ian Rogers
NO_DWARFcould mean more thanNO_LIBDWsupport, in particular no libunwind support. Rename to be more intention revealing.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Makefile.config | 16 ++++++++-------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- tools/perf/builtin-probe.c | 2 +- 15 files changed, 22 insertions(+), 22 deletions(-)tools/perf/Makefile.config +8
-8@@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1) endif ifdef NO_LIBELF - NO_DWARF := 1 + NO_LIBDW := 1 NO_LIBUNWIND := 1 NO_LIBDW_DWARF_UNWIND := 1 NO_LIBBPF := 1@@ -471,9 +471,9 @@ else endif endif ifneq ($(feature-dwarf), 1) - ifndef NO_DWARF + ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) - NO_DWARF := 1 + NO_LIBDW := 1 endif else ifneq ($(feature-dwarf_getlocations), 1)@@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1) endif endif -ifdef NO_DWARF +ifdef NO_LIBDW NO_LIBDW_DWARF_UNWIND := 1 endif@@ -574,17 +574,17 @@ ifndef NO_LIBELF endif endif - ifndef NO_DWARF + ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) - NO_DWARF := 1 + NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGS - endif # NO_DWARF + endif # NO_LIBDW ifndef NO_LIBBPF ifeq ($(feature-bpf), 1)@@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP endif ifeq ($(SRCARCH),powerpc) - ifndef NO_DWARF + ifndef NO_LIBDW CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX endif endiftools/perf/Makefile.perf +0
-0
tools/perf/Makefile.perf include ../scripts/utilities.mak · line 43
Quoted 5 lines
tools/perf/Makefile.perf +1
-1@@ -40,7 +40,7 @@ include ../scripts/utilities.mak # # Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS. # -# Define NO_DWARF if you do not want debug-info analysis feature at all. +# Define NO_LIBDW if you do not want debug-info analysis feature at all.
Can we keep NO_DWARF for compatibility and set NO_LIBDW=1 internally? I think it's fine to change it here to advertise NO_LIBDW over NO_DWARF but still want to support NO_DWARF as well.
Also it seem we don't have an entry in the tests/make for no-dwarf build. Can you please add one too?
Thanks,
Namhyung
Quoted 147 lines
tools/perf/Makefile.perf +0
-0# # Define WERROR=0 to disable treating any warnings as errors. #tools/perf/arch/arm/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1tools/perf/arch/arm64/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1tools/perf/arch/csky/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/loongarch/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1tools/perf/arch/mips/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/powerpc/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/riscv/Makefile +1
-1@@ -1,4 +1,4 @@ -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1tools/perf/arch/s390/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1tools/perf/arch/sh/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/sparc/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/x86/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1tools/perf/arch/xtensa/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/builtin-probe.c +1
-1@@ -616,7 +616,7 @@ __cmd_probe(int argc, const char **argv)set_option_flag (options,'L' ,"line" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'V' ,"vars" ,PARSE_OPT_EXCLUSIVE );#else -# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_DWARF=1", c) +# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_LIBDW=1", c)set_nobuild ('L' ,"line" , false);set_nobuild ('V' ,"vars" , false);set_nobuild ('\0' ,"externs" , false);Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerRe: [PATCH v1 01/11] perf build: Rename NO_DWARF to NO_LIBDW
Quoted 94 lines · Namhyung Kim
Quoted 92 lines · Ian Rogers
NO_DWARFcould mean more thanNO_LIBDWsupport, in particular no libunwind support. Rename to be more intention revealing.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Makefile.config | 16 ++++++++-------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- tools/perf/builtin-probe.c | 2 +- 15 files changed, 22 insertions(+), 22 deletions(-)tools/perf/Makefile.config +8
-8@@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1) endif ifdef NO_LIBELF - NO_DWARF := 1 + NO_LIBDW := 1 NO_LIBUNWIND := 1 NO_LIBDW_DWARF_UNWIND := 1 NO_LIBBPF := 1@@ -471,9 +471,9 @@ else endif endif ifneq ($(feature-dwarf), 1) - ifndef NO_DWARF + ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) - NO_DWARF := 1 + NO_LIBDW := 1 endif else ifneq ($(feature-dwarf_getlocations), 1)@@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1) endif endif -ifdef NO_DWARF +ifdef NO_LIBDW NO_LIBDW_DWARF_UNWIND := 1 endif@@ -574,17 +574,17 @@ ifndef NO_LIBELF endif endif - ifndef NO_DWARF + ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) - NO_DWARF := 1 + NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGS - endif # NO_DWARF + endif # NO_LIBDW ifndef NO_LIBBPF ifeq ($(feature-bpf), 1)@@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP endif ifeq ($(SRCARCH),powerpc) - ifndef NO_DWARF + ifndef NO_LIBDW CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX endif endiftools/perf/Makefile.perf +0
-0@@ -40,7 +40,7 @@ include ../scripts/utilities.mak #
Quoted 8 lines
tools/perf/Makefile.perf include ../scripts/utilities.mak · line 43
Quoted 4 lines
tools/perf/Makefile.perf +1
-1# Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS. # -# Define NO_DWARF if you do not want debug-info analysis feature at all. +# Define NO_LIBDW if you do not want debug-info analysis feature at all.Can we keep
NO_DWARFfor compatibility and set NO_LIBDW=1 internally? I think it's fine to change it here to advertiseNO_LIBDWoverNO_DWARFbut still want to supportNO_DWARFas well.
This would be actively confusing. NO_DWARF should really imply NO_LIBUNWIND, it is like removing the ability for "perf record --call-graph dwarf" to work - dwarf meaning libdw or libunwind in the tool. Having NO_DWARF control NO_LIBDW, I'm not sure what I'm being asked to do. The point of this clean up is to make the names mean what they should mean and not something potentially confusingly broader, but you're asking for the confusingly broader case to mean just libdw. I don't follow and I don't see how it serves the purpose of the cleanup.
Quoted 2 lines
Also it seem we don't have an entry in the tests/make for no-dwarf build. Can you please add one too?
Sure.
Thanks,
Ian
Quoted 151 lines
Thanks,
NamhyungQuoted 147 lines
tools/perf/Makefile.perf +0
-0# # Define WERROR=0 to disable treating any warnings as errors. #tools/perf/arch/arm/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1tools/perf/arch/arm64/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1tools/perf/arch/csky/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/loongarch/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1tools/perf/arch/mips/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/powerpc/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/riscv/Makefile +1
-1@@ -1,4 +1,4 @@ -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1tools/perf/arch/s390/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1tools/perf/arch/sh/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/sparc/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/x86/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1tools/perf/arch/xtensa/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/builtin-probe.c +1
-2@@ -616,7 +616,7 @@ __cmd_probe(int argc, const char **argv)set_option_flag (options,'L' ,"line" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'V' ,"vars" ,PARSE_OPT_EXCLUSIVE );#else -# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_DWARF=1", c) +# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_LIBDW=1", c)set_nobuild ('L' ,"line" , false);set_nobuild ('V' ,"vars" , false);set_nobuild ('\0' ,"externs" , false); --2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 01/11] perf build: Rename NO_DWARF to NO_LIBDW
Quoted 104 lines · Ian Rogers
Quoted 94 lines · Namhyung Kim
Quoted 92 lines · Ian Rogers
NO_DWARFcould mean more thanNO_LIBDWsupport, in particular no libunwind support. Rename to be more intention revealing.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Makefile.config | 16 ++++++++-------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- tools/perf/builtin-probe.c | 2 +- 15 files changed, 22 insertions(+), 22 deletions(-)tools/perf/Makefile.config +8
-8@@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1) endif ifdef NO_LIBELF - NO_DWARF := 1 + NO_LIBDW := 1 NO_LIBUNWIND := 1 NO_LIBDW_DWARF_UNWIND := 1 NO_LIBBPF := 1@@ -471,9 +471,9 @@ else endif endif ifneq ($(feature-dwarf), 1) - ifndef NO_DWARF + ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) - NO_DWARF := 1 + NO_LIBDW := 1 endif else ifneq ($(feature-dwarf_getlocations), 1)@@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1) endif endif -ifdef NO_DWARF +ifdef NO_LIBDW NO_LIBDW_DWARF_UNWIND := 1 endif@@ -574,17 +574,17 @@ ifndef NO_LIBELF endif endif - ifndef NO_DWARF + ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) - NO_DWARF := 1 + NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGS - endif # NO_DWARF + endif # NO_LIBDW ifndef NO_LIBBPF ifeq ($(feature-bpf), 1)@@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP endif ifeq ($(SRCARCH),powerpc) - ifndef NO_DWARF + ifndef NO_LIBDW CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX endif endiftools/perf/Makefile.perf +0
-0@@ -40,7 +40,7 @@ include ../scripts/utilities.mak #Quoted 8 lines
tools/perf/Makefile.perf include ../scripts/utilities.mak · line 43
Quoted 4 lines
tools/perf/Makefile.perf +1
-1# Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS. # -# Define NO_DWARF if you do not want debug-info analysis feature at all. +# Define NO_LIBDW if you do not want debug-info analysis feature at all.Can we keep
NO_DWARFfor compatibility and set NO_LIBDW=1 internally? I think it's fine to change it here to advertiseNO_LIBDWoverNO_DWARFbut still want to supportNO_DWARFas well.
Quoted 9 lines
This would be actively confusing.
NO_DWARFshould really implyNO_LIBUNWIND, it is like removing the ability for "perf record --call-graph dwarf" to work - dwarf meaning libdw or libunwind in the tool. HavingNO_DWARFcontrolNO_LIBDW, I'm not sure what I'm being asked to do. The point of this clean up is to make the names mean what they should mean and not something potentially confusingly broader, but you're asking for the confusingly broader case to mean just libdw. I don't follow and I don't see how it serves the purpose of the cleanup.
I have no idea how many people use NO_DWARF actually. We can probably ignore such users and convert it to NO_LIBDW.
What I actually thought was
ifeq ($(NO_DWARF),1)
NO_LIBDW := 1
NO_LIBUNWIND := 1
endifbut it'd break the compatibility too. :(
Thanks,
Namhyung
Re: [PATCH v1 01/11] perf build: Rename NO_DWARF to NO_LIBDW
Quoted 118 lines · On Thu, 26 Sep 2024 12:28:09 -0700 Namhyung Kim <[email protected]> wrote:
Quoted 104 lines · Ian Rogers
Quoted 94 lines · Namhyung Kim
Quoted 92 lines · Ian Rogers
NO_DWARFcould mean more thanNO_LIBDWsupport, in particular no libunwind support. Rename to be more intention revealing.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Makefile.config | 16 ++++++++-------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- tools/perf/builtin-probe.c | 2 +- 15 files changed, 22 insertions(+), 22 deletions(-)tools/perf/Makefile.config +8
-8@@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1) endif ifdef NO_LIBELF - NO_DWARF := 1 + NO_LIBDW := 1 NO_LIBUNWIND := 1 NO_LIBDW_DWARF_UNWIND := 1 NO_LIBBPF := 1@@ -471,9 +471,9 @@ else endif endif ifneq ($(feature-dwarf), 1) - ifndef NO_DWARF + ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) - NO_DWARF := 1 + NO_LIBDW := 1 endif else ifneq ($(feature-dwarf_getlocations), 1)@@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1) endif endif -ifdef NO_DWARF +ifdef NO_LIBDW NO_LIBDW_DWARF_UNWIND := 1 endif@@ -574,17 +574,17 @@ ifndef NO_LIBELF endif endif - ifndef NO_DWARF + ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) - NO_DWARF := 1 + NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGS - endif # NO_DWARF + endif # NO_LIBDW ifndef NO_LIBBPF ifeq ($(feature-bpf), 1)@@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP endif ifeq ($(SRCARCH),powerpc) - ifndef NO_DWARF + ifndef NO_LIBDW CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX endif endiftools/perf/Makefile.perf +0
-0@@ -40,7 +40,7 @@ include ../scripts/utilities.mak #Quoted 8 lines
tools/perf/Makefile.perf include ../scripts/utilities.mak · line 43
Quoted 4 lines
tools/perf/Makefile.perf +1
-1# Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS. # -# Define NO_DWARF if you do not want debug-info analysis feature at all. +# Define NO_LIBDW if you do not want debug-info analysis feature at all.Can we keep
NO_DWARFfor compatibility and set NO_LIBDW=1 internally? I think it's fine to change it here to advertiseNO_LIBDWoverNO_DWARFbut still want to supportNO_DWARFas well.Quoted 9 lines
This would be actively confusing.
NO_DWARFshould really implyNO_LIBUNWIND, it is like removing the ability for "perf record --call-graph dwarf" to work - dwarf meaning libdw or libunwind in the tool. HavingNO_DWARFcontrolNO_LIBDW, I'm not sure what I'm being asked to do. The point of this clean up is to make the names mean what they should mean and not something potentially confusingly broader, but you're asking for the confusingly broader case to mean just libdw. I don't follow and I don't see how it serves the purpose of the cleanup.I have no idea how many people use
NO_DWARFactually. We can probably ignore such users and convert it toNO_LIBDW.
Quoted 6 lines
What I actually thought was
ifeq ($(NO_DWARF),1) NO_LIBDW := 1 NO_LIBUNWIND := 1 endif
Maybe we can add another patch to do this for new compile option?
Thanks,
Quoted 10 lines
but it'd break the compatibility too. :(
Thanks,
Namhyung
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Re: [PATCH v1 01/11] perf build: Rename NO_DWARF to NO_LIBDW
In reply to Ian RogersQuoted 3 lines · On Tue, 24 Sep 2024 09:04:08 -0700 Ian Rogers <[email protected]> wrote:
NO_DWARFcould mean more thanNO_LIBDWsupport, in particular no libunwind support. Rename to be more intention revealing.
Looks good to me.
Thanks,
Quoted 241 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Makefile.config | 16 ++++++++-------- tools/perf/Makefile.perf | 2 +- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- tools/perf/builtin-probe.c | 2 +- 15 files changed, 22 insertions(+), 22 deletions(-)tools/perf/Makefile.config +8
-8@@ -429,7 +429,7 @@ ifeq ($(feature-file-handle), 1) endif ifdef NO_LIBELF - NO_DWARF := 1 + NO_LIBDW := 1 NO_LIBUNWIND := 1 NO_LIBDW_DWARF_UNWIND := 1 NO_LIBBPF := 1@@ -471,9 +471,9 @@ else endif endif ifneq ($(feature-dwarf), 1) - ifndef NO_DWARF + ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) - NO_DWARF := 1 + NO_LIBDW := 1 endif else ifneq ($(feature-dwarf_getlocations), 1)@@ -496,7 +496,7 @@ ifeq ($(feature-libaio), 1) endif endif -ifdef NO_DWARF +ifdef NO_LIBDW NO_LIBDW_DWARF_UNWIND := 1 endif@@ -574,17 +574,17 @@ ifndef NO_LIBELF endif endif - ifndef NO_DWARF + ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) - NO_DWARF := 1 + NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGS - endif # NO_DWARF + endif # NO_LIBDW ifndef NO_LIBBPF ifeq ($(feature-bpf), 1)@@ -633,7 +633,7 @@ ifdef PERF_HAVE_JITDUMP endif ifeq ($(SRCARCH),powerpc) - ifndef NO_DWARF + ifndef NO_LIBDW CFLAGS += -DHAVE_SKIP_CALLCHAIN_IDX endif endiftools/perf/Makefile.perf +1
-1@@ -40,7 +40,7 @@ include ../scripts/utilities.mak # # Define EXTRA_PERFLIBS to pass extra libraries to PERFLIBS. # -# Define NO_DWARF if you do not want debug-info analysis feature at all. +# Define NO_LIBDW if you do not want debug-info analysis feature at all. # # Define WERROR=0 to disable treating any warnings as errors. #tools/perf/arch/arm/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1tools/perf/arch/arm64/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_JITDUMP := 1tools/perf/arch/csky/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/loongarch/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1tools/perf/arch/mips/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/powerpc/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/riscv/Makefile +1
-1@@ -1,4 +1,4 @@ -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1tools/perf/arch/s390/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1tools/perf/arch/sh/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/sparc/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/arch/x86/Makefile +1
-1@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1tools/perf/arch/xtensa/Makefile +1
-1@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only -ifndef NO_DWARF +ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endiftools/perf/builtin-probe.c +1
-1@@ -616,7 +616,7 @@ __cmd_probe(int argc, const char **argv)set_option_flag (options,'L' ,"line" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'V' ,"vars" ,PARSE_OPT_EXCLUSIVE );#else -# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_DWARF=1", c) +# define set_nobuild (s, l, c) set_option_nobuild(options, s, l, "NO_LIBDW=1", c)set_nobuild ('L' ,"line" , false);set_nobuild ('V' ,"vars" , false);set_nobuild ('\0' ,"externs" , false);Signature · 1 line
2.46.0.792.g87dc391469-goog
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerPATCH 2/11 perf build: Remove defined but never used variable
In reply to Ian RogersPreviously NO_DWARF_UNWIND was part of conditional compilation but it is now unused so remove.
--- tools/perf/Makefile.config | 2 -- 1 file changed, 2 deletions(-)
tools/perf/Makefile.config +0 -2
@@ -753,8 +753,6 @@ endif ifeq ($(dwarf-post-unwind),1) CFLAGS += -DHAVE_DWARF_UNWIND_SUPPORT $(call detected,CONFIG_DWARF_UNWIND) -else - NO_DWARF_UNWIND := 1 endif ifndef NO_LOCAL_LIBUNWIND
Signature · 1 line
2.46.0.792.g87dc391469-goog
Re: [PATCH v1 02/11] perf build: Remove defined but never used variable
Quoted 3 lines · On Tue, 24 Sep 2024 09:04:09 -0700 Ian Rogers <[email protected]> wrote:
Previously
NO_DWARF_UNWINDwas part of conditional compilation but it is now unused so remove.
Thank you!
Quoted 21 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Makefile.config | 2 -- 1 file changed, 2 deletions(-)tools/perf/Makefile.config +0
-2@@ -753,8 +753,6 @@ endif ifeq ($(dwarf-post-unwind),1) CFLAGS += -DHAVE_DWARF_UNWIND_SUPPORT $(call detected,CONFIG_DWARF_UNWIND) -else - NO_DWARF_UNWIND := 1 endif ifndef NO_LOCAL_LIBUNWINDSignature · 1 line
2.46.0.792.g87dc391469-goog
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerPATCH 3/11 perf build: Rename test-dwarf to test-libdw
In reply to Ian RogersBe more intention revealing that the dwarf test is actually testing for libdw support.
---
tools/build/Makefile.feature | 6 +++---
tools/build/feature/Makefile | 16 ++++++++--------
tools/build/feature/test-all.c | 6 +++---
.../build/feature/{test-dwarf.c => test-libdw.c} | 0
tools/perf/Makefile.config | 6 +++---
5 files changed, 17 insertions(+), 17 deletions(-)
rename tools/build/feature/{test-dwarf.c => test-libdw.c} (100%)tools/build/Makefile.feature +3 -3
@@ -30,7 +30,7 @@ endef # FEATURE_TESTS_BASIC := \ backtrace \ - dwarf \ + libdw \ dwarf_getlocations \ dwarf_getcfi \ eventfd \
@@ -120,7 +120,7 @@ ifeq ($(FEATURE_TESTS),all) endif FEATURE_DISPLAY ?= \ - dwarf \ + libdw \ dwarf_getlocations \ glibc \ libbfd \
@@ -233,7 +233,7 @@ endef # # generates feature value assignment for name, like: -# $(call feature_assign,dwarf) == feature-dwarf=1 +# $(call feature_assign,libdw) == feature-libdw=1 # feature_assign = feature-$(1)=$(feature-$(1))
tools/build/feature/Makefile +8 -8
@@ -5,7 +5,7 @@ FILES= \ test-all.bin \ test-backtrace.bin \ test-bionic.bin \ - test-dwarf.bin \ + test-libdw.bin \ test-dwarf_getlocations.bin \ test-dwarf_getcfi.bin \ test-eventfd.bin \
@@ -168,9 +168,9 @@ $(OUTPUT)test-libopencsd.bin: $(BUILD) # -lopencsd_c_api -lopencsd provided by # $(FEATURE_CHECK_LDFLAGS-libopencsd) -DWARFLIBS := -ldw +DWLIBS := -ldw ifeq ($(findstring -static,${LDFLAGS}),-static) - DWARFLIBS += -lelf -lz -llzma -lbz2 -lzstd + DWLIBS += -lelf -lz -llzma -lbz2 -lzstd LIBDW_VERSION := $(shell $(PKG_CONFIG) --modversion libdw) LIBDW_VERSION_1 := $(word 1, $(subst ., ,$(LIBDW_VERSION)))
@@ -179,18 +179,18 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) # Elfutils merged libebl.a into libdw.a starting from version 0.177, # Link libebl.a only if libdw is older than this version. ifeq ($(shell test $(LIBDW_VERSION_2) -lt 177; echo $$?),0) - DWARFLIBS += -lebl + DWLIBS += -lebl endif endif -$(OUTPUT)test-dwarf.bin: - $(BUILD) $(DWARFLIBS) +$(OUTPUT)test-libdw.bin: + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getlocations.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getcfi.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-libelf-getphdrnum.bin: $(BUILD) -lelf
tools/build/feature/test-all.c +3 -3
@@ -38,8 +38,8 @@# include "test-glibc.c" #undef main -#define main main_test_dwarf -# include "test-dwarf.c" +#define main main_test_libdw +# include "test-libdw.c" #undef main#define main main_test_dwarf_getlocations
@@ -187,7 +187,7 @@ int main(int argc, char *argv[])main_test_get_current_dir_name ();main_test_gettid ();main_test_glibc (); -main_test_dwarf (); +main_test_libdw ();main_test_dwarf_getlocations ();main_test_eventfd ();main_test_libelf_getphdrnum ();
tools/build/feature/test-libdw.c renamed from tools/build/feature/test-dwarf.c +0 -0
tools/perf/Makefile.config +3 -3
@@ -159,8 +159,8 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) DWARFLIBS += -lebl endif endif -FEATURE_CHECK_CFLAGS-dwarf := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf := $(LIBDW_LDFLAGS) $(DWARFLIBS) +FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) +FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-libdw-dwarf-unwind := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS)
@@ -470,7 +470,7 @@ else $(warning No libdw DWARF unwind found, Please install elfutils-devel/libdw-dev >= 0.158 and/or set LIBDW_DIR) endif endif - ifneq ($(feature-dwarf), 1) + ifneq ($(feature-libdw), 1) ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1
Signature · 1 line
2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 03/11] perf build: Rename test-dwarf to test-libdw
Quoted 126 lines · Ian Rogers
Be more intention revealing that the dwarf test is actually testing for libdw support.
Signed-off-by Ian Rogers <[email protected]>--- tools/build/Makefile.feature | 6 +++--- tools/build/feature/Makefile | 16 ++++++++-------- tools/build/feature/test-all.c | 6 +++--- .../build/feature/{test-dwarf.c => test-libdw.c} | 0 tools/perf/Makefile.config | 6 +++--- 5 files changed, 17 insertions(+), 17 deletions(-) rename tools/build/feature/{test-dwarf.c => test-libdw.c} (100%)tools/build/Makefile.feature +3
-3@@ -30,7 +30,7 @@ endef # FEATURE_TESTS_BASIC := \ backtrace \ - dwarf \ + libdw \ dwarf_getlocations \ dwarf_getcfi \ eventfd \@@ -120,7 +120,7 @@ ifeq ($(FEATURE_TESTS),all) endif FEATURE_DISPLAY ?= \ - dwarf \ + libdw \ dwarf_getlocations \ glibc \ libbfd \@@ -233,7 +233,7 @@ endef # # generates feature value assignment for name, like: -# $(call feature_assign,dwarf) == feature-dwarf=1 +# $(call feature_assign,libdw) == feature-libdw=1 # feature_assign = feature-$(1)=$(feature-$(1))tools/build/feature/Makefile +8
-8@@ -5,7 +5,7 @@ FILES= \ test-all.bin \ test-backtrace.bin \ test-bionic.bin \ - test-dwarf.bin \ + test-libdw.bin \ test-dwarf_getlocations.bin \ test-dwarf_getcfi.bin \ test-eventfd.bin \@@ -168,9 +168,9 @@ $(OUTPUT)test-libopencsd.bin: $(BUILD) # -lopencsd_c_api -lopencsd provided by # $(FEATURE_CHECK_LDFLAGS-libopencsd) -DWARFLIBS := -ldw +DWLIBS := -ldw ifeq ($(findstring -static,${LDFLAGS}),-static) - DWARFLIBS += -lelf -lz -llzma -lbz2 -lzstd + DWLIBS += -lelf -lz -llzma -lbz2 -lzstd LIBDW_VERSION := $(shell $(PKG_CONFIG) --modversion libdw) LIBDW_VERSION_1 := $(word 1, $(subst ., ,$(LIBDW_VERSION)))@@ -179,18 +179,18 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) # Elfutils merged libebl.a into libdw.a starting from version 0.177, # Link libebl.a only if libdw is older than this version. ifeq ($(shell test $(LIBDW_VERSION_2) -lt 177; echo $$?),0) - DWARFLIBS += -lebl + DWLIBS += -lebl endif endif -$(OUTPUT)test-dwarf.bin: - $(BUILD) $(DWARFLIBS) +$(OUTPUT)test-libdw.bin: + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getlocations.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getcfi.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-libelf-getphdrnum.bin: $(BUILD) -lelftools/build/feature/test-all.c +3
-3@@ -38,8 +38,8 @@# include "test-glibc.c" #undef main -#define main main_test_dwarf -# include "test-dwarf.c" +#define main main_test_libdw +# include "test-libdw.c" #undef main#define main main_test_dwarf_getlocations@@ -187,7 +187,7 @@ int main(int argc, char *argv[])main_test_get_current_dir_name ();main_test_gettid ();main_test_glibc (); -main_test_dwarf (); +main_test_libdw ();main_test_dwarf_getlocations ();main_test_eventfd ();main_test_libelf_getphdrnum ();tools/build/feature/test-libdw.c renamed from tools/build/feature/test-dwarf.c +0
-0tools/perf/Makefile.config +0
-0
tools/perf/Makefile.config ifeq ($(findstring -static,${LDFLAGS}),-static) · line 163
Quoted 7 lines
tools/perf/Makefile.config +2
-2@@ -159,8 +159,8 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) DWARFLIBS += -lebl endif endif -FEATURE_CHECK_CFLAGS-dwarf := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf := $(LIBDW_LDFLAGS) $(DWARFLIBS) +FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) +FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS)
s/DWARFLIBS/DWLIBS/ ?
tools/perf/Makefile.config ifeq ($(findstring -static,${LDFLAGS}),-static) · line 165
Quoted 2 lines
tools/perf/Makefile.config +0
-0FEATURE_CHECK_CFLAGS-libdw-dwarf-unwind := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind := $(LIBDW_LDFLAGS) $(DWARFLIBS)
Ditto.
Thanks,
Namhyung
Quoted 13 lines
tools/perf/Makefile.config +1
-1FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS)@@ -470,7 +470,7 @@ else $(warning No libdw DWARF unwind found, Please install elfutils-devel/libdw-dev >= 0.158 and/or set LIBDW_DIR) endif endif - ifneq ($(feature-dwarf), 1) + ifneq ($(feature-libdw), 1) ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerRe: [PATCH v1 03/11] perf build: Rename test-dwarf to test-libdw
Quoted 129 lines · Namhyung Kim
Quoted 127 lines · Ian Rogers
Be more intention revealing that the dwarf test is actually testing for libdw support.
Signed-off-by Ian Rogers <[email protected]>--- tools/build/Makefile.feature | 6 +++--- tools/build/feature/Makefile | 16 ++++++++-------- tools/build/feature/test-all.c | 6 +++--- .../build/feature/{test-dwarf.c => test-libdw.c} | 0 tools/perf/Makefile.config | 6 +++--- 5 files changed, 17 insertions(+), 17 deletions(-) rename tools/build/feature/{test-dwarf.c => test-libdw.c} (100%)tools/build/Makefile.feature +3
-3@@ -30,7 +30,7 @@ endef # FEATURE_TESTS_BASIC := \ backtrace \ - dwarf \ + libdw \ dwarf_getlocations \ dwarf_getcfi \ eventfd \@@ -120,7 +120,7 @@ ifeq ($(FEATURE_TESTS),all) endif FEATURE_DISPLAY ?= \ - dwarf \ + libdw \ dwarf_getlocations \ glibc \ libbfd \@@ -233,7 +233,7 @@ endef # # generates feature value assignment for name, like: -# $(call feature_assign,dwarf) == feature-dwarf=1 +# $(call feature_assign,libdw) == feature-libdw=1 # feature_assign = feature-$(1)=$(feature-$(1))tools/build/feature/Makefile +8
-8@@ -5,7 +5,7 @@ FILES= \ test-all.bin \ test-backtrace.bin \ test-bionic.bin \ - test-dwarf.bin \ + test-libdw.bin \ test-dwarf_getlocations.bin \ test-dwarf_getcfi.bin \ test-eventfd.bin \@@ -168,9 +168,9 @@ $(OUTPUT)test-libopencsd.bin: $(BUILD) # -lopencsd_c_api -lopencsd provided by # $(FEATURE_CHECK_LDFLAGS-libopencsd) -DWARFLIBS := -ldw +DWLIBS := -ldw ifeq ($(findstring -static,${LDFLAGS}),-static) - DWARFLIBS += -lelf -lz -llzma -lbz2 -lzstd + DWLIBS += -lelf -lz -llzma -lbz2 -lzstd LIBDW_VERSION := $(shell $(PKG_CONFIG) --modversion libdw) LIBDW_VERSION_1 := $(word 1, $(subst ., ,$(LIBDW_VERSION)))@@ -179,18 +179,18 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) # Elfutils merged libebl.a into libdw.a starting from version 0.177, # Link libebl.a only if libdw is older than this version. ifeq ($(shell test $(LIBDW_VERSION_2) -lt 177; echo $$?),0) - DWARFLIBS += -lebl + DWLIBS += -lebl endif endif -$(OUTPUT)test-dwarf.bin: - $(BUILD) $(DWARFLIBS) +$(OUTPUT)test-libdw.bin: + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getlocations.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getcfi.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-libelf-getphdrnum.bin: $(BUILD) -lelftools/build/feature/test-all.c +3
-3@@ -38,8 +38,8 @@# include "test-glibc.c" #undef main -#define main main_test_dwarf -# include "test-dwarf.c" +#define main main_test_libdw +# include "test-libdw.c" #undef main#define main main_test_dwarf_getlocations@@ -187,7 +187,7 @@ int main(int argc, char *argv[])main_test_get_current_dir_name ();main_test_gettid ();main_test_glibc (); -main_test_dwarf (); +main_test_libdw ();main_test_dwarf_getlocations ();main_test_eventfd ();main_test_libelf_getphdrnum ();tools/build/feature/test-libdw.c renamed from tools/build/feature/test-dwarf.c +0
-0tools/perf/Makefile.config +0
-0@@ -159,8 +159,8 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) DWARFLIBS += -lebl
Quoted 8 lines
tools/perf/Makefile.config ifeq ($(findstring -static,${LDFLAGS}),-static) · line 163
Quoted 6 lines
tools/perf/Makefile.config +2
-2endif endif -FEATURE_CHECK_CFLAGS-dwarf := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf := $(LIBDW_LDFLAGS) $(DWARFLIBS) +FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) +FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS)s/DWARFLIBS/DWLIBS/ ?
The patch is changing the name of a feature test. It is possible to change this Makefile.config variable in the same patch but I think they are distinct things and the variable rename is follow up work.
Thanks,
Ian
Quoted 23 lines
tools/perf/Makefile.config ifeq ($(findstring -static,${LDFLAGS}),-static) · line 165
Quoted 3 lines
tools/perf/Makefile.config +0
-0FEATURE_CHECK_CFLAGS-libdw-dwarf-unwind := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind := $(LIBDW_LDFLAGS) $(DWARFLIBS)Ditto.
Thanks,
NamhyungQuoted 13 lines
tools/perf/Makefile.config +1
-2FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS)@@ -470,7 +470,7 @@ else $(warning No libdw DWARF unwind found, Please install elfutils-devel/libdw-dev >= 0.158 and/or set LIBDW_DIR) endif endif - ifneq ($(feature-dwarf), 1) + ifneq ($(feature-libdw), 1) ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1 --2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 03/11] perf build: Rename test-dwarf to test-libdw
Quoted 64 lines · Ian Rogers
Quoted 63 lines · Namhyung Kim
Quoted 61 lines · Ian Rogers
Be more intention revealing that the dwarf test is actually testing for libdw support.
Signed-off-by Ian Rogers <[email protected]>--- tools/build/Makefile.feature | 6 +++--- tools/build/feature/Makefile | 16 ++++++++-------- tools/build/feature/test-all.c | 6 +++--- .../build/feature/{test-dwarf.c => test-libdw.c} | 0 tools/perf/Makefile.config | 6 +++--- 5 files changed, 17 insertions(+), 17 deletions(-) rename tools/build/feature/{test-dwarf.c => test-libdw.c} (100%)tools/build/Makefile.feature +3
-3@@ -30,7 +30,7 @@ endef # FEATURE_TESTS_BASIC := \ backtrace \ - dwarf \ + libdw \ dwarf_getlocations \ dwarf_getcfi \ eventfd \@@ -120,7 +120,7 @@ ifeq ($(FEATURE_TESTS),all) endif FEATURE_DISPLAY ?= \ - dwarf \ + libdw \ dwarf_getlocations \ glibc \ libbfd \@@ -233,7 +233,7 @@ endef # # generates feature value assignment for name, like: -# $(call feature_assign,dwarf) == feature-dwarf=1 +# $(call feature_assign,libdw) == feature-libdw=1 # feature_assign = feature-$(1)=$(feature-$(1))tools/build/feature/Makefile +1
-1@@ -5,7 +5,7 @@ FILES= \ test-all.bin \ test-backtrace.bin \ test-bionic.bin \ - test-dwarf.bin \ + test-libdw.bin \ test-dwarf_getlocations.bin \ test-dwarf_getcfi.bin \ test-eventfd.bin \@@ -168,9 +168,9 @@ $(OUTPUT)test-libopencsd.bin: $(BUILD) # -lopencsd_c_api -lopencsd provided by # $(FEATURE_CHECK_LDFLAGS-libopencsd)
Quoted 2 lines
Quoted 2 lines
tools/build/feature/Makefile $(OUTPUT)test-libopencsd.bin: · line 171
Quoted 2 lines
tools/build/feature/Makefile +1
-1-DWARFLIBS := -ldw +DWLIBS := -ldw
See here and the below changes.
Quoted 73 lines
Quoted 64 lines
Quoted 64 lines
tools/build/feature/Makefile +6
-6ifeq ($(findstring -static,${LDFLAGS}),-static) - DWARFLIBS += -lelf -lz -llzma -lbz2 -lzstd + DWLIBS += -lelf -lz -llzma -lbz2 -lzstd LIBDW_VERSION := $(shell $(PKG_CONFIG) --modversion libdw) LIBDW_VERSION_1 := $(word 1, $(subst ., ,$(LIBDW_VERSION)))@@ -179,18 +179,18 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) # Elfutils merged libebl.a into libdw.a starting from version 0.177, # Link libebl.a only if libdw is older than this version. ifeq ($(shell test $(LIBDW_VERSION_2) -lt 177; echo $$?),0) - DWARFLIBS += -lebl + DWLIBS += -lebl endif endif -$(OUTPUT)test-dwarf.bin: - $(BUILD) $(DWARFLIBS) +$(OUTPUT)test-libdw.bin: + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getlocations.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getcfi.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-libelf-getphdrnum.bin: $(BUILD) -lelftools/build/feature/test-all.c +3
-3@@ -38,8 +38,8 @@# include "test-glibc.c" #undef main -#define main main_test_dwarf -# include "test-dwarf.c" +#define main main_test_libdw +# include "test-libdw.c" #undef main#define main main_test_dwarf_getlocations@@ -187,7 +187,7 @@ int main(int argc, char *argv[])main_test_get_current_dir_name ();main_test_gettid ();main_test_glibc (); -main_test_dwarf (); +main_test_libdw ();main_test_dwarf_getlocations ();main_test_eventfd ();main_test_libelf_getphdrnum ();tools/build/feature/test-libdw.c renamed from tools/build/feature/test-dwarf.c +0
-0tools/perf/Makefile.config +0
-0@@ -159,8 +159,8 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) DWARFLIBS += -leblQuoted 8 lines
tools/perf/Makefile.config ifeq ($(findstring -static,${LDFLAGS}),-static) · line 163
Quoted 6 lines
tools/perf/Makefile.config +2
-2endif endif -FEATURE_CHECK_CFLAGS-dwarf := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf := $(LIBDW_LDFLAGS) $(DWARFLIBS) +FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) +FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS)s/DWARFLIBS/DWLIBS/ ?
Quoted 3 lines
The patch is changing the name of a feature test. It is possible to change this Makefile.config variable in the same patch but I think they are distinct things and the variable rename is follow up work.
Oh.. you changed it in tools/build/feature/Makefile above but didn't in tools/perf/Makefile.config. That confused me.. ;-)
Thanks,
Namhyung
Quoted 23 lines
Quoted 23 lines
tools/perf/Makefile.config ifeq ($(findstring -static,${LDFLAGS}),-static) · line 165
Quoted 3 lines
tools/perf/Makefile.config +0
-0FEATURE_CHECK_CFLAGS-libdw-dwarf-unwind := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind := $(LIBDW_LDFLAGS) $(DWARFLIBS)Ditto.
Thanks,
NamhyungQuoted 13 lines
tools/perf/Makefile.config +1
-2FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS)@@ -470,7 +470,7 @@ else $(warning No libdw DWARF unwind found, Please install elfutils-devel/libdw-dev >= 0.158 and/or set LIBDW_DIR) endif endif - ifneq ($(feature-dwarf), 1) + ifneq ($(feature-libdw), 1) ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1 --2.46.0.792.g87dc391469-goog
Re: [PATCH v1 03/11] perf build: Rename test-dwarf to test-libdw
In reply to Ian RogersQuoted 3 lines · On Tue, 24 Sep 2024 09:04:10 -0700 Ian Rogers <[email protected]> wrote:
Be more intention revealing that the dwarf test is actually testing for libdw support.
Looks good to me.
Thanks!
Quoted 146 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/build/Makefile.feature | 6 +++--- tools/build/feature/Makefile | 16 ++++++++-------- tools/build/feature/test-all.c | 6 +++--- .../build/feature/{test-dwarf.c => test-libdw.c} | 0 tools/perf/Makefile.config | 6 +++--- 5 files changed, 17 insertions(+), 17 deletions(-) rename tools/build/feature/{test-dwarf.c => test-libdw.c} (100%)tools/build/Makefile.feature +3
-3@@ -30,7 +30,7 @@ endef # FEATURE_TESTS_BASIC := \ backtrace \ - dwarf \ + libdw \ dwarf_getlocations \ dwarf_getcfi \ eventfd \@@ -120,7 +120,7 @@ ifeq ($(FEATURE_TESTS),all) endif FEATURE_DISPLAY ?= \ - dwarf \ + libdw \ dwarf_getlocations \ glibc \ libbfd \@@ -233,7 +233,7 @@ endef # # generates feature value assignment for name, like: -# $(call feature_assign,dwarf) == feature-dwarf=1 +# $(call feature_assign,libdw) == feature-libdw=1 # feature_assign = feature-$(1)=$(feature-$(1))tools/build/feature/Makefile +8
-8@@ -5,7 +5,7 @@ FILES= \ test-all.bin \ test-backtrace.bin \ test-bionic.bin \ - test-dwarf.bin \ + test-libdw.bin \ test-dwarf_getlocations.bin \ test-dwarf_getcfi.bin \ test-eventfd.bin \@@ -168,9 +168,9 @@ $(OUTPUT)test-libopencsd.bin: $(BUILD) # -lopencsd_c_api -lopencsd provided by # $(FEATURE_CHECK_LDFLAGS-libopencsd) -DWARFLIBS := -ldw +DWLIBS := -ldw ifeq ($(findstring -static,${LDFLAGS}),-static) - DWARFLIBS += -lelf -lz -llzma -lbz2 -lzstd + DWLIBS += -lelf -lz -llzma -lbz2 -lzstd LIBDW_VERSION := $(shell $(PKG_CONFIG) --modversion libdw) LIBDW_VERSION_1 := $(word 1, $(subst ., ,$(LIBDW_VERSION)))@@ -179,18 +179,18 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) # Elfutils merged libebl.a into libdw.a starting from version 0.177, # Link libebl.a only if libdw is older than this version. ifeq ($(shell test $(LIBDW_VERSION_2) -lt 177; echo $$?),0) - DWARFLIBS += -lebl + DWLIBS += -lebl endif endif -$(OUTPUT)test-dwarf.bin: - $(BUILD) $(DWARFLIBS) +$(OUTPUT)test-libdw.bin: + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getlocations.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-dwarf_getcfi.bin: - $(BUILD) $(DWARFLIBS) + $(BUILD) $(DWLIBS) $(OUTPUT)test-libelf-getphdrnum.bin: $(BUILD) -lelftools/build/feature/test-all.c +3
-3@@ -38,8 +38,8 @@# include "test-glibc.c" #undef main -#define main main_test_dwarf -# include "test-dwarf.c" +#define main main_test_libdw +# include "test-libdw.c" #undef main#define main main_test_dwarf_getlocations@@ -187,7 +187,7 @@ int main(int argc, char *argv[])main_test_get_current_dir_name ();main_test_gettid ();main_test_glibc (); -main_test_dwarf (); +main_test_libdw ();main_test_dwarf_getlocations ();main_test_eventfd ();main_test_libelf_getphdrnum ();tools/build/feature/test-libdw.c renamed from tools/build/feature/test-dwarf.c +0
-0tools/perf/Makefile.config +3
-3@@ -159,8 +159,8 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) DWARFLIBS += -lebl endif endif -FEATURE_CHECK_CFLAGS-dwarf := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf := $(LIBDW_LDFLAGS) $(DWARFLIBS) +FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) +FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-libdw-dwarf-unwind := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS)@@ -470,7 +470,7 @@ else $(warning No libdw DWARF unwind found, Please install elfutils-devel/libdw-dev >= 0.158 and/or set LIBDW_DIR) endif endif - ifneq ($(feature-dwarf), 1) + ifneq ($(feature-libdw), 1) ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1Signature · 1 line
2.46.0.792.g87dc391469-goog
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerPATCH 4/11 perf build: Combine libdw-dwarf-unwind into libdw feature tests
In reply to Ian RogersSupport in libdw has been present for 10 years so let's simplify the build logic with a single feature test.
--- tools/build/Makefile.feature | 2 -- tools/build/feature/Makefile | 4 ---- tools/build/feature/test-all.c | 5 ----- tools/build/feature/test-libdw-dwarf-unwind.c | 14 -------------- tools/build/feature/test-libdw.c | 19 ++++++++++++++++++- tools/perf/Makefile.config | 8 -------- 6 files changed, 18 insertions(+), 34 deletions(-) delete mode 100644 tools/build/feature/test-libdw-dwarf-unwind.c
tools/build/Makefile.feature +0 -2
@@ -60,7 +60,6 @@ FEATURE_TESTS_BASIC := \ reallocarray \ stackprotector-all \ timerfd \ - libdw-dwarf-unwind \ zlib \ lzma \ get_cpuid \
@@ -133,7 +132,6 @@ FEATURE_DISPLAY ?= \ libpython \ libcrypto \ libunwind \ - libdw-dwarf-unwind \ libcapstone \ llvm-perf \ zlib \
tools/build/feature/Makefile +0 -4
@@ -52,7 +52,6 @@ FILES= \ test-pthread-barrier.bin \ test-stackprotector-all.bin \ test-timerfd.bin \ - test-libdw-dwarf-unwind.bin \ test-libbabeltrace.bin \ test-libcapstone.bin \ test-compile-32.bin \
@@ -314,9 +313,6 @@ $(OUTPUT)test-backtrace.bin: $(OUTPUT)test-timerfd.bin: $(BUILD) -$(OUTPUT)test-libdw-dwarf-unwind.bin: - $(BUILD) # -ldw provided by $(FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind) - $(OUTPUT)test-libbabeltrace.bin: $(BUILD) # -lbabeltrace provided by $(FEATURE_CHECK_LDFLAGS-libbabeltrace)
tools/build/feature/test-all.c +0 -5
@@ -98,10 +98,6 @@# include "test-stackprotector-all.c" #undef main -#define main main_test_libdw_dwarf_unwind -# include "test-libdw-dwarf-unwind.c" -#undef main -#define main main_test_zlib# include "test-zlib.c" #undef main
@@ -202,7 +198,6 @@ int main(int argc, char *argv[])main_test_numa_num_possible_cpus ();main_test_timerfd ();main_test_stackprotector_all (); -main_test_libdw_dwarf_unwind ();main_test_zlib ();main_test_pthread_attr_setaffinity_np ();main_test_pthread_barrier ();
tools/build/feature/test-libdw-dwarf-unwind.c deleted +0 -14
@@ -1,14 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 - -#include <elfutils/libdwfl.h> - -int main (void ) -{ -/* -* This function is guarded via: __nonnull_attribute__ (1, 2). -* Passing '1' as arguments value. This code is never executed, -* only compiled. -*/ -dwfl_thread_getframes ((void *)1 , (void *)1 ,NULL ); -return 0 ; -}
tools/build/feature/test-libdw.c +18 -1
@@ -1,11 +1,28 @@// SPDX-License-Identifier: GPL-2.0 #include <dwarf.h> #include <elfutils/libdw.h> +#include <elfutils/libdwfl.h> #include <elfutils/version.h> -int main (void ) +int test_libdw (void ) {Dwarf *dbg =dwarf_begin (0 ,DWARF_C_READ );return (long )dbg; } + +int test_libdw_unwind (void ) +{ +/* +* This function is guarded via: __nonnull_attribute__ (1, 2). +* Passing '1' as arguments value. This code is never executed, +* only compiled. +*/ +dwfl_thread_getframes ((void *)1 , (void *)1 ,NULL ); +return 0 ; +} + +int main (void ) +{ +return test_libdw () +test_libdw_unwind (); +}
tools/perf/Makefile.config +0 -8
@@ -161,8 +161,6 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) endif FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS) -FEATURE_CHECK_CFLAGS-libdw-dwarf-unwind := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-libdw-dwarf-unwind := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-dwarf_getlocations := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-dwarf_getcfi := $(LIBDW_CFLAGS)
@@ -464,12 +462,6 @@ else endif endif else - ifndef NO_LIBDW_DWARF_UNWIND - ifneq ($(feature-libdw-dwarf-unwind),1) - NO_LIBDW_DWARF_UNWIND := 1 - $(warning No libdw DWARF unwind found, Please install elfutils-devel/libdw-dev >= 0.158 and/or set LIBDW_DIR) - endif - endif ifneq ($(feature-libdw), 1) ifndef NO_LIBDW $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev)
Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerPATCH 5/11 perf build: Combine test-dwarf-getlocations into test-libdw
In reply to Ian Rogersdwarf_getlocations support in libdw is more than 10 years old. Make libdw imply dwarf_getlocations support and simplify build logic.
--- tools/build/Makefile.feature | 2 -- tools/build/feature/Makefile | 4 ---- tools/build/feature/test-all.c | 5 ----- tools/build/feature/test-dwarf_getlocations.c | 13 ------------- tools/build/feature/test-libdw.c | 14 +++++++++++++- tools/perf/Makefile.config | 10 ++-------- 6 files changed, 15 insertions(+), 33 deletions(-) delete mode 100644 tools/build/feature/test-dwarf_getlocations.c
tools/build/Makefile.feature +0 -2
@@ -31,7 +31,6 @@ endef FEATURE_TESTS_BASIC := \ backtrace \ libdw \ - dwarf_getlocations \ dwarf_getcfi \ eventfd \ fortify-source \
@@ -120,7 +119,6 @@ endif FEATURE_DISPLAY ?= \ libdw \ - dwarf_getlocations \ glibc \ libbfd \ libbfd-buildid \
tools/build/feature/Makefile +0 -4
@@ -6,7 +6,6 @@ FILES= \ test-backtrace.bin \ test-bionic.bin \ test-libdw.bin \ - test-dwarf_getlocations.bin \ test-dwarf_getcfi.bin \ test-eventfd.bin \ test-fortify-source.bin \
@@ -185,9 +184,6 @@ endif $(OUTPUT)test-libdw.bin: $(BUILD) $(DWLIBS) -$(OUTPUT)test-dwarf_getlocations.bin: - $(BUILD) $(DWLIBS) - $(OUTPUT)test-dwarf_getcfi.bin: $(BUILD) $(DWLIBS)
tools/build/feature/test-all.c +0 -5
@@ -42,10 +42,6 @@# include "test-libdw.c" #undef main -#define main main_test_dwarf_getlocations -# include "test-dwarf_getlocations.c" -#undef main -#define main main_test_eventfd# include "test-eventfd.c" #undef main
@@ -184,7 +180,6 @@ int main(int argc, char *argv[])main_test_gettid ();main_test_glibc ();main_test_libdw (); -main_test_dwarf_getlocations ();main_test_eventfd ();main_test_libelf_getphdrnum ();main_test_libelf_gelf_getnote ();
tools/build/feature/test-dwarf_getlocations.c deleted +0 -13
@@ -1,13 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -#include <stdlib.h> -#include <elfutils/libdw.h> - -int main (void ) -{ -Dwarf_Addr base, start, end; -Dwarf_Attribute attr; -Dwarf_Op *op; -size_t nops; -ptrdiff_t offset =0 ; -return (int )dwarf_getlocations (&attr, offset, &base, &start, &end, &op, &nops); -}
tools/build/feature/test-libdw.c +13 -1
@@ -1,4 +1,5 @@// SPDX-License-Identifier: GPL-2.0 +#include <stdlib.h> #include <dwarf.h> #include <elfutils/libdw.h> #include <elfutils/libdwfl.h>
@@ -22,7 +23,18 @@ int test_libdw_unwind(void)return 0 ; } +int test_libdw_getlocations (void ) +{ +Dwarf_Addr base, start, end; +Dwarf_Attribute attr; +Dwarf_Op *op; +size_t nops; +ptrdiff_t offset =0 ; + +return (int )dwarf_getlocations (&attr, offset, &base, &start, &end, &op, &nops); +} +int main (void ) { -return test_libdw () +test_libdw_unwind (); +return test_libdw () +test_libdw_unwind () +test_libdw_getlocations (); }
tools/perf/Makefile.config +2 -8
@@ -161,8 +161,6 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) endif FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS) -FEATURE_CHECK_CFLAGS-dwarf_getlocations := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf_getlocations := $(LIBDW_LDFLAGS) $(DWARFLIBS) FEATURE_CHECK_CFLAGS-dwarf_getcfi := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-dwarf_getcfi := $(LIBDW_LDFLAGS) $(DWARFLIBS)
@@ -464,15 +462,11 @@ else else ifneq ($(feature-libdw), 1) ifndef NO_LIBDW - $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev) + $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.157, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1 endif else - ifneq ($(feature-dwarf_getlocations), 1) - $(warning Old libdw.h, finding variables at given 'perf probe' point will not work, install elfutils-devel/libdw-dev >= 0.157) - else - CFLAGS += -DHAVE_DWARF_GETLOCATIONS_SUPPORT - endif # dwarf_getlocations + CFLAGS += -DHAVE_DWARF_GETLOCATIONS_SUPPORT ifneq ($(feature-dwarf_getcfi), 1) $(warning Old libdw.h, finding variables at given 'perf probe' point will not work, install elfutils-devel/libdw-dev >= 0.142) else
Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerPATCH 6/11 perf build: Combine test-dwarf-getcfi into test-libdw
In reply to Ian Rogersdwarf_getcfi support in libdw is 15 years old. Make libdw imply dwarf_getcfi support and simplify build logic.
--- tools/build/Makefile.feature | 1 - tools/build/feature/Makefile | 4 ---- tools/build/feature/test-dwarf_getcfi.c | 9 --------- tools/build/feature/test-libdw.c | 10 +++++++++- tools/perf/Makefile.config | 8 +------- 5 files changed, 10 insertions(+), 22 deletions(-) delete mode 100644 tools/build/feature/test-dwarf_getcfi.c
tools/build/Makefile.feature +0 -1
@@ -31,7 +31,6 @@ endef FEATURE_TESTS_BASIC := \ backtrace \ libdw \ - dwarf_getcfi \ eventfd \ fortify-source \ get_current_dir_name \
tools/build/feature/Makefile +0 -4
@@ -6,7 +6,6 @@ FILES= \ test-backtrace.bin \ test-bionic.bin \ test-libdw.bin \ - test-dwarf_getcfi.bin \ test-eventfd.bin \ test-fortify-source.bin \ test-get_current_dir_name.bin \
@@ -184,9 +183,6 @@ endif $(OUTPUT)test-libdw.bin: $(BUILD) $(DWLIBS) -$(OUTPUT)test-dwarf_getcfi.bin: - $(BUILD) $(DWLIBS) - $(OUTPUT)test-libelf-getphdrnum.bin: $(BUILD) -lelf
tools/build/feature/test-dwarf_getcfi.c deleted +0 -9
@@ -1,9 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -#include <stdio.h> -#include <elfutils/libdw.h> - -int main (void ) -{ -Dwarf *dwarf =NULL ; -return dwarf_getcfi (dwarf) ==NULL ; -}
tools/build/feature/test-libdw.c +9 -1
@@ -34,7 +34,15 @@ int test_libdw_getlocations(void)return (int )dwarf_getlocations (&attr, offset, &base, &start, &end, &op, &nops); } +int test_libdw_getcfi (void ) +{ +Dwarf *dwarf =NULL ; + +return dwarf_getcfi (dwarf) ==NULL ; +} +int main (void ) { -return test_libdw () +test_libdw_unwind () +test_libdw_getlocations (); +return test_libdw () +test_libdw_unwind () +test_libdw_getlocations () + +test_libdw_getcfi (); }
tools/perf/Makefile.config +1 -7
@@ -161,8 +161,6 @@ ifeq ($(findstring -static,${LDFLAGS}),-static) endif FEATURE_CHECK_CFLAGS-libdw := $(LIBDW_CFLAGS) FEATURE_CHECK_LDFLAGS-libdw := $(LIBDW_LDFLAGS) $(DWARFLIBS) -FEATURE_CHECK_CFLAGS-dwarf_getcfi := $(LIBDW_CFLAGS) -FEATURE_CHECK_LDFLAGS-dwarf_getcfi := $(LIBDW_LDFLAGS) $(DWARFLIBS) # for linking with debug library, run like: # make DEBUG=1 LIBBABELTRACE_DIR=/opt/libbabeltrace/
@@ -467,11 +465,7 @@ else endif else CFLAGS += -DHAVE_DWARF_GETLOCATIONS_SUPPORT - ifneq ($(feature-dwarf_getcfi), 1) - $(warning Old libdw.h, finding variables at given 'perf probe' point will not work, install elfutils-devel/libdw-dev >= 0.142) - else - CFLAGS += -DHAVE_DWARF_CFI_SUPPORT - endif # dwarf_getcfi + CFLAGS += -DHAVE_DWARF_CFI_SUPPORT endif # Dwarf support endif # libelf support endif # NO_LIBELF
Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerPATCH 7/11 perf probe: Move elfutils support check to libdw check
In reply to Ian RogersThe test _ELFUTILS_PREREQ(0, 142) is false for elfutils before 2009-06-13, but that is 15 years ago and very unlikely. Add a test to test-libdw.c and assume the libdw version is at least 0.142 to simplify the build logic.
--- tools/build/feature/test-libdw.c | 10 +++++++++- tools/perf/util/probe-finder.c | 2 -- tools/perf/util/probe-finder.h | 2 -- 3 files changed, 9 insertions(+), 5 deletions(-)
tools/build/feature/test-libdw.c +9 -1
@@ -41,8 +41,16 @@ int test_libdw_getcfi(void)return dwarf_getcfi (dwarf) ==NULL ; } +int test_elfutils (void ) +{ +Dwarf_CFI *cfi =NULL ; + +dwarf_cfi_end (cfi); +return 0 ; +} +int main (void ) {return test_libdw () +test_libdw_unwind () +test_libdw_getlocations () + -test_libdw_getcfi (); +test_libdw_getcfi () +test_elfutils (); }
tools/perf/util/probe-finder.c +0 -2
@@ -1379,10 +1379,8 @@ int debuginfo__find_trace_events(struct debuginfo *dbg,if (ret >=0 && tf.pf.skip_empty_arg) ret =fill_empty_trace_arg (pev, tf.tevs, tf.ntevs); -#if _ELFUTILS_PREREQ (0 ,142 )dwarf_cfi_end (tf.pf.cfi_eh);dwarf_cfi_end (tf.pf.cfi_dbg); -#endif if (ret <0 || tf.ntevs ==0 ) {for (i =0 ; i < tf.ntevs; i++)
tools/perf/util/probe-finder.h +0 -2
@@ -63,12 +63,10 @@ struct probe_finder {struct intlist *lcache;/* Line cache for lazy match */ /* For variable searching */ -#if _ELFUTILS_PREREQ (0 ,142 )/* Call Frame Information from .eh_frame */ Dwarf_CFI *cfi_eh;/* Call Frame Information from .debug_frame */ Dwarf_CFI *cfi_dbg; -#endif Dwarf_Op *fb_ops;/* Frame base attribute */ unsigned int machine;/* Target machine arch */ struct perf_probe_arg *pvar;/* Current target variable */
Signature · 1 line
2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 07/11] perf probe: Move elfutils support check to libdw check
Quoted 17 lines · Ian Rogers
The test
_ELFUTILS_PREREQ(0, 142)is false for elfutils before 2009-06-13, but that is 15 years ago and very unlikely. Add a test totest-libdw.cand assume the libdw version is at least 0.142 to simplify the build logic.Signed-off-by Ian Rogers <[email protected]>--- tools/build/feature/test-libdw.c | 10 +++++++++- tools/perf/util/probe-finder.c | 2 -- tools/perf/util/probe-finder.h | 2 -- 3 files changed, 9 insertions(+), 5 deletions(-)tools/build/feature/test-libdw.c +0
-0
tools/build/feature/test-libdw.c int test_libdw_getcfi(void) · line 50
Quoted 10 lines
tools/build/feature/test-libdw.c +7
-0@@ -41,8 +41,16 @@ int test_libdw_getcfi(void)return dwarf_getcfi (dwarf) ==NULL ; } +int test_elfutils (void ) +{ +Dwarf_CFI *cfi =NULL ; + +dwarf_cfi_end (cfi); +return 0 ; +}
I think it's the same as test_libdw_getcfi() and let's get rid of it.
Thanks,
Namhyung
Quoted 42 lines
tools/build/feature/test-libdw.c +2
-1+int main (void ) {return test_libdw () +test_libdw_unwind () +test_libdw_getlocations () + -test_libdw_getcfi (); +test_libdw_getcfi () +test_elfutils (); }tools/perf/util/probe-finder.c +0
-2@@ -1379,10 +1379,8 @@ int debuginfo__find_trace_events(struct debuginfo *dbg,if (ret >=0 && tf.pf.skip_empty_arg) ret =fill_empty_trace_arg (pev, tf.tevs, tf.ntevs); -#if _ELFUTILS_PREREQ (0 ,142 )dwarf_cfi_end (tf.pf.cfi_eh);dwarf_cfi_end (tf.pf.cfi_dbg); -#endif if (ret <0 || tf.ntevs ==0 ) {for (i =0 ; i < tf.ntevs; i++)tools/perf/util/probe-finder.h +0
-2@@ -63,12 +63,10 @@ struct probe_finder {struct intlist *lcache;/* Line cache for lazy match */ /* For variable searching */ -#if _ELFUTILS_PREREQ (0 ,142 )/* Call Frame Information from .eh_frame */ Dwarf_CFI *cfi_eh;/* Call Frame Information from .debug_frame */ Dwarf_CFI *cfi_dbg; -#endif Dwarf_Op *fb_ops;/* Frame base attribute */ unsigned int machine;/* Target machine arch */ struct perf_probe_arg *pvar;/* Current target variable */ Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerRe: [PATCH v1 07/11] perf probe: Move elfutils support check to libdw check
Quoted 22 lines · Namhyung Kim
Quoted 20 lines · Ian Rogers
The test
_ELFUTILS_PREREQ(0, 142)is false for elfutils before 2009-06-13, but that is 15 years ago and very unlikely. Add a test totest-libdw.cand assume the libdw version is at least 0.142 to simplify the build logic.Signed-off-by Ian Rogers <[email protected]>--- tools/build/feature/test-libdw.c | 10 +++++++++- tools/perf/util/probe-finder.c | 2 -- tools/perf/util/probe-finder.h | 2 -- 3 files changed, 9 insertions(+), 5 deletions(-)tools/build/feature/test-libdw.c +0
-0@@ -41,8 +41,16 @@ int test_libdw_getcfi(void)return dwarf_getcfi (dwarf) ==NULL ; }
Quoted 9 lines
tools/build/feature/test-libdw.c int test_libdw_getcfi(void) · line 50
Quoted 7 lines
tools/build/feature/test-libdw.c +7
-0+int test_elfutils (void ) +{ +Dwarf_CFI *cfi =NULL ; + +dwarf_cfi_end (cfi); +return 0 ; +}I think it's the same as
test_libdw_getcfi()and let's get rid of it.
The point of doing the change this way, and I think what's being ignored, is that I am trying to simply replace an #if with a clearly equivalent feature test - clearly equivalent as the #if guards use of dwarf_cfi_end in the code. Merging test_elfutils with test_libdw_getcfi is another cognitive step and I think it is fine as follow up cleanup. Combining that step with the change here fails the minimal meaningful change and tbh I think this is just bike shedding.
Thanks,
Ian
Quoted 45 lines
Thanks,
NamhyungQuoted 42 lines
tools/build/feature/test-libdw.c +2
-1+int main (void ) {return test_libdw () +test_libdw_unwind () +test_libdw_getlocations () + -test_libdw_getcfi (); +test_libdw_getcfi () +test_elfutils (); }tools/perf/util/probe-finder.c +0
-2@@ -1379,10 +1379,8 @@ int debuginfo__find_trace_events(struct debuginfo *dbg,if (ret >=0 && tf.pf.skip_empty_arg) ret =fill_empty_trace_arg (pev, tf.tevs, tf.ntevs); -#if _ELFUTILS_PREREQ (0 ,142 )dwarf_cfi_end (tf.pf.cfi_eh);dwarf_cfi_end (tf.pf.cfi_dbg); -#endif if (ret <0 || tf.ntevs ==0 ) {for (i =0 ; i < tf.ntevs; i++)tools/perf/util/probe-finder.h +0
-3@@ -63,12 +63,10 @@ struct probe_finder {struct intlist *lcache;/* Line cache for lazy match */ /* For variable searching */ -#if _ELFUTILS_PREREQ (0 ,142 )/* Call Frame Information from .eh_frame */ Dwarf_CFI *cfi_eh;/* Call Frame Information from .debug_frame */ Dwarf_CFI *cfi_dbg; -#endif Dwarf_Op *fb_ops;/* Frame base attribute */ unsigned int machine;/* Target machine arch */ struct perf_probe_arg *pvar;/* Current target variable */ --2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 07/11] perf probe: Move elfutils support check to libdw check
Quoted 33 lines · Ian Rogers
Quoted 22 lines · Namhyung Kim
Quoted 20 lines · Ian Rogers
The test
_ELFUTILS_PREREQ(0, 142)is false for elfutils before 2009-06-13, but that is 15 years ago and very unlikely. Add a test totest-libdw.cand assume the libdw version is at least 0.142 to simplify the build logic.Signed-off-by Ian Rogers <[email protected]>--- tools/build/feature/test-libdw.c | 10 +++++++++- tools/perf/util/probe-finder.c | 2 -- tools/perf/util/probe-finder.h | 2 -- 3 files changed, 9 insertions(+), 5 deletions(-)tools/build/feature/test-libdw.c +0
-0@@ -41,8 +41,16 @@ int test_libdw_getcfi(void)return dwarf_getcfi (dwarf) ==NULL ; }Quoted 9 lines
tools/build/feature/test-libdw.c int test_libdw_getcfi(void) · line 50
Quoted 7 lines
tools/build/feature/test-libdw.c +7
-0+int test_elfutils (void ) +{ +Dwarf_CFI *cfi =NULL ; + +dwarf_cfi_end (cfi); +return 0 ; +}I think it's the same as
test_libdw_getcfi()and let's get rid of it.
Quoted 7 lines
The point of doing the change this way, and I think what's being ignored, is that I am trying to simply replace an #if with a clearly equivalent feature test - clearly equivalent as the #if guards use of
dwarf_cfi_endin the code. Mergingtest_elfutilswithtest_libdw_getcfiis another cognitive step and I think it is fine as follow up cleanup. Combining that step with the change here fails the minimal meaningful change and tbh I think this is just bike shedding.
Ok, I think it's ok to have a separate commit to combine equivalent tests if you really want to. What I want in the end is not to have unnecessary or duplicate code.
In order to have a clearly equivalent feature test, I think you can add
#if !_ELFUTILS_PREREQ(0, 142) #error "elfutils version is too old" #endif
Thanks,
Namhyung
Quoted 43 lines
Quoted 43 lines
Quoted 43 lines
tools/build/feature/test-libdw.c +2
-1+int main (void ) {return test_libdw () +test_libdw_unwind () +test_libdw_getlocations () + -test_libdw_getcfi (); +test_libdw_getcfi () +test_elfutils (); }tools/perf/util/probe-finder.c +0
-2@@ -1379,10 +1379,8 @@ int debuginfo__find_trace_events(struct debuginfo *dbg,if (ret >=0 && tf.pf.skip_empty_arg) ret =fill_empty_trace_arg (pev, tf.tevs, tf.ntevs); -#if _ELFUTILS_PREREQ (0 ,142 )dwarf_cfi_end (tf.pf.cfi_eh);dwarf_cfi_end (tf.pf.cfi_dbg); -#endif if (ret <0 || tf.ntevs ==0 ) {for (i =0 ; i < tf.ntevs; i++)tools/perf/util/probe-finder.h +0
-3@@ -63,12 +63,10 @@ struct probe_finder {struct intlist *lcache;/* Line cache for lazy match */ /* For variable searching */ -#if _ELFUTILS_PREREQ (0 ,142 )/* Call Frame Information from .eh_frame */ Dwarf_CFI *cfi_eh;/* Call Frame Information from .debug_frame */ Dwarf_CFI *cfi_dbg; -#endif Dwarf_Op *fb_ops;/* Frame base attribute */ unsigned int machine;/* Target machine arch */ struct perf_probe_arg *pvar;/* Current target variable */ --2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerPATCH 8/11 perf libdw: Remove unnecessary defines
In reply to Ian RogersAs HAVE_DWARF_GETLOCATIONS_SUPPORT and HAVE_DWARF_CFI_SUPPORT always match HAVE_DWARF_SUPPORT remove the macros and use HAVE_DWARF_SUPPORT. If building the file is guarded by CONFIG_DWARF then remove all ifs.
--- tools/perf/Documentation/perf-check.txt | 2 +- tools/perf/Makefile.config | 3 -- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 2 +- tools/perf/builtin-report.c | 2 +- tools/perf/util/dwarf-aux.c | 6 --- tools/perf/util/dwarf-aux.h | 54 ------------------------- tools/perf/util/probe-finder.c | 4 -- 8 files changed, 4 insertions(+), 71 deletions(-)
tools/perf/Documentation/perf-check.txt +1 -1
@@ -48,7 +48,7 @@ feature:: bpf_skeletons / HAVE_BPF_SKEL debuginfod / HAVE_DEBUGINFOD_SUPPORT dwarf / HAVE_DWARF_SUPPORT - dwarf_getlocations / HAVE_DWARF_GETLOCATIONS_SUPPORT + dwarf_getlocations / HAVE_DWARF_SUPPORT dwarf-unwind / HAVE_DWARF_UNWIND_SUPPORT auxtrace / HAVE_AUXTRACE_SUPPORT libaudit / HAVE_LIBAUDIT_SUPPORT
tools/perf/Makefile.config +0 -3
@@ -463,9 +463,6 @@ else $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.157, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1 endif - else - CFLAGS += -DHAVE_DWARF_GETLOCATIONS_SUPPORT - CFLAGS += -DHAVE_DWARF_CFI_SUPPORT endif # Dwarf support endif # libelf support endif # NO_LIBELF
tools/perf/builtin-annotate.c +1 -1
@@ -840,7 +840,7 @@ int cmd_annotate(int argc, const char **argv) }#endif -#ifndef HAVE_DWARF_GETLOCATIONS_SUPPORT +#ifndef HAVE_DWARF_SUPPORT if (annotate.data_type) {pr_err ("Error: Data type profiling is disabled due to missing DWARF support\n" );return -ENOTSUP ;
tools/perf/builtin-check.c +1 -1
@@ -28,7 +28,7 @@ struct feature_status supported_features[] = {FEATURE_STATUS ("bpf_skeletons" ,HAVE_BPF_SKEL ),FEATURE_STATUS ("debuginfod" ,HAVE_DEBUGINFOD_SUPPORT ),FEATURE_STATUS ("dwarf" ,HAVE_DWARF_SUPPORT ), -FEATURE_STATUS ("dwarf_getlocations" ,HAVE_DWARF_GETLOCATIONS_SUPPORT ), +FEATURE_STATUS ("dwarf_getlocations" ,HAVE_DWARF_SUPPORT ),FEATURE_STATUS ("dwarf-unwind" ,HAVE_DWARF_UNWIND_SUPPORT ),FEATURE_STATUS ("auxtrace" ,HAVE_AUXTRACE_SUPPORT ), FEATURE_STATUS("libaudit", HAVE_LIBAUDIT_SUPPORT),
tools/perf/builtin-report.c +1 -1
@@ -1701,7 +1701,7 @@ int cmd_report(int argc, const char **argv) report.data_type = true; annotate_opts.annotate_src = false; -#ifndef HAVE_DWARF_GETLOCATIONS_SUPPORT +#ifndef HAVE_DWARF_SUPPORT pr_err ("Error: Data type profiling is disabled due to missing DWARF support\n" ); goto error;#endif
tools/perf/util/dwarf-aux.c +0 -6
@@ -1182,7 +1182,6 @@ int die_get_varname(Dwarf_Die *vr_die, struct strbuf *buf)return ret <0 ? ret :strbuf_addf (buf,"\t%s" ,dwarf_diename (vr_die)); } -#if defined(HAVE_DWARF_GETLOCATIONS_SUPPORT ) || defined(HAVE_DWARF_CFI_SUPPORT )static int reg_from_dwarf_op (Dwarf_Op *op) {switch (op->atom) {
@@ -1245,9 +1244,7 @@ static bool check_allowed_ops(Dwarf_Op *ops, size_t nops) }return true; } -#endif /* HAVE_DWARF_GETLOCATIONS_SUPPORT || HAVE_DWARF_CFI_SUPPORT */ -#ifdef HAVE_DWARF_GETLOCATIONS_SUPPORT /** * die_get_var_innermost_scope - Get innermost scope range of given variable DIE * @sp_die: a subprogram DIE
@@ -1697,9 +1694,7 @@ void die_collect_global_vars(Dwarf_Die *cu_die, struct die_var_type **var_types)die_find_child (cu_die, __die_collect_global_vars_cb, (void *)var_types, &die_mem); } -#endif /* HAVE_DWARF_GETLOCATIONS_SUPPORT */ -#ifdef HAVE_DWARF_CFI_SUPPORT /** * die_get_cfa - Get frame base information * @dwarf: a Dwarf info
@@ -1732,7 +1727,6 @@ int die_get_cfa(Dwarf *dwarf, u64 pc, int *preg, int *poffset) }return -1 ; } -#endif /* HAVE_DWARF_CFI_SUPPORT */ /* * die_has_loclist - Check if DW_AT_location of @vr_die is a location list
tools/perf/util/dwarf-aux.h +0 -54
67 lines not on this page · raw message
tools/perf/util/probe-finder.c +0 -4
@@ -602,7 +602,6 @@ static int call_probe_finder(Dwarf_Die *sc_die, struct probe_finder *pf) ret =dwarf_getlocation_addr (&fb_attr, pf->addr, &pf->fb_ops, &nops,1 );if (ret <=0 || nops ==0 ) { pf->fb_ops =NULL ; -#ifdef HAVE_DWARF_CFI_SUPPORT }else if (nops ==1 && pf->fb_ops[0 ].atom == DW_OP_call_frame_cfa && (pf->cfi_eh !=NULL || pf->cfi_dbg !=NULL )) { if ((dwarf_cfi_addrframe(pf->cfi_eh, pf->addr, &frame) != 0 &&
@@ -613,7 +612,6 @@ static int call_probe_finder(Dwarf_Die *sc_die, struct probe_finder *pf)free (frame);return -ENOENT ; } -#endif /* HAVE_DWARF_CFI_SUPPORT */ }/* Call finder's callback handler */
@@ -1138,7 +1136,6 @@ static int debuginfo__find_probes(struct debuginfo *dbg, pf->machine = ehdr.e_machine; -#ifdef HAVE_DWARF_CFI_SUPPORT do { GElf_Shdr shdr;
@@ -1148,7 +1145,6 @@ static int debuginfo__find_probes(struct debuginfo *dbg, pf->cfi_dbg =dwarf_getcfi (dbg->dbg); }while (0 ); -#endif /* HAVE_DWARF_CFI_SUPPORT */ ret =debuginfo__find_probe_location (dbg, pf);return ret;
Signature · 1 line
2.46.0.792.g87dc391469-goog
Re: [PATCH v1 08/11] perf libdw: Remove unnecessary defines
Quoted 4 lines · On Tue, 24 Sep 2024 09:04:15 -0700 Ian Rogers <[email protected]> wrote:
As
HAVE_DWARF_GETLOCATIONS_SUPPORTandHAVE_DWARF_CFI_SUPPORTalways matchHAVE_DWARF_SUPPORTremove the macros and useHAVE_DWARF_SUPPORT. If building the file is guarded byCONFIG_DWARFthen remove all ifs.
Thanks for the cleanups!
Quoted 18 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-check.txt | 2 +- tools/perf/Makefile.config | 3 -- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 2 +- tools/perf/builtin-report.c | 2 +- tools/perf/util/dwarf-aux.c | 6 --- tools/perf/util/dwarf-aux.h | 54 ------------------------- tools/perf/util/probe-finder.c | 4 -- 8 files changed, 4 insertions(+), 71 deletions(-)tools/perf/Documentation/perf-check.txt +0
-0
tools/perf/Documentation/perf-check.txt feature:: · line 51
Quoted 5 lines
tools/perf/Documentation/perf-check.txt +1
-1@@ -48,7 +48,7 @@ feature:: bpf_skeletons / HAVE_BPF_SKEL debuginfod / HAVE_DEBUGINFOD_SUPPORT dwarf / HAVE_DWARF_SUPPORT - dwarf_getlocations / HAVE_DWARF_GETLOCATIONS_SUPPORT + dwarf_getlocations / HAVE_DWARF_SUPPORT
nit: BTW, I think we can remove this feature itself now.
Thank you,
Quoted 209 lines
tools/perf/Documentation/perf-check.txt +0
-0dwarf-unwind / HAVE_DWARF_UNWIND_SUPPORT auxtrace / HAVE_AUXTRACE_SUPPORT libaudit / HAVE_LIBAUDIT_SUPPORTtools/perf/Makefile.config +0
-3@@ -463,9 +463,6 @@ else $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.157, disables dwarf support. Please install new elfutils-devel/libdw-dev) NO_LIBDW := 1 endif - else - CFLAGS += -DHAVE_DWARF_GETLOCATIONS_SUPPORT - CFLAGS += -DHAVE_DWARF_CFI_SUPPORT endif # Dwarf support endif # libelf support endif # NO_LIBELFtools/perf/builtin-annotate.c +1
-1@@ -840,7 +840,7 @@ int cmd_annotate(int argc, const char **argv) }#endif -#ifndef HAVE_DWARF_GETLOCATIONS_SUPPORT +#ifndef HAVE_DWARF_SUPPORT if (annotate.data_type) {pr_err ("Error: Data type profiling is disabled due to missing DWARF support\n" );return -ENOTSUP ;tools/perf/builtin-check.c +1
-1@@ -28,7 +28,7 @@ struct feature_status supported_features[] = {FEATURE_STATUS ("bpf_skeletons" ,HAVE_BPF_SKEL ),FEATURE_STATUS ("debuginfod" ,HAVE_DEBUGINFOD_SUPPORT ),FEATURE_STATUS ("dwarf" ,HAVE_DWARF_SUPPORT ), -FEATURE_STATUS ("dwarf_getlocations" ,HAVE_DWARF_GETLOCATIONS_SUPPORT ), +FEATURE_STATUS ("dwarf_getlocations" ,HAVE_DWARF_SUPPORT ),FEATURE_STATUS ("dwarf-unwind" ,HAVE_DWARF_UNWIND_SUPPORT ),FEATURE_STATUS ("auxtrace" ,HAVE_AUXTRACE_SUPPORT ), FEATURE_STATUS("libaudit", HAVE_LIBAUDIT_SUPPORT),tools/perf/builtin-report.c +1
-1@@ -1701,7 +1701,7 @@ int cmd_report(int argc, const char **argv) report.data_type = true; annotate_opts.annotate_src = false; -#ifndef HAVE_DWARF_GETLOCATIONS_SUPPORT +#ifndef HAVE_DWARF_SUPPORT pr_err ("Error: Data type profiling is disabled due to missing DWARF support\n" ); goto error;#endif tools/perf/util/dwarf-aux.c +0
-6@@ -1182,7 +1182,6 @@ int die_get_varname(Dwarf_Die *vr_die, struct strbuf *buf)return ret <0 ? ret :strbuf_addf (buf,"\t%s" ,dwarf_diename (vr_die)); } -#if defined(HAVE_DWARF_GETLOCATIONS_SUPPORT ) || defined(HAVE_DWARF_CFI_SUPPORT )static int reg_from_dwarf_op (Dwarf_Op *op) {switch (op->atom) {@@ -1245,9 +1244,7 @@ static bool check_allowed_ops(Dwarf_Op *ops, size_t nops) }return true; } -#endif /* HAVE_DWARF_GETLOCATIONS_SUPPORT || HAVE_DWARF_CFI_SUPPORT */ -#ifdef HAVE_DWARF_GETLOCATIONS_SUPPORT /** * die_get_var_innermost_scope - Get innermost scope range of given variable DIE * @sp_die: a subprogram DIE @@ -1697,9 +1694,7 @@ void die_collect_global_vars(Dwarf_Die *cu_die, struct die_var_type **var_types)die_find_child (cu_die, __die_collect_global_vars_cb, (void *)var_types, &die_mem); } -#endif /* HAVE_DWARF_GETLOCATIONS_SUPPORT */ -#ifdef HAVE_DWARF_CFI_SUPPORT /** * die_get_cfa - Get frame base information * @dwarf: a Dwarf info @@ -1732,7 +1727,6 @@ int die_get_cfa(Dwarf *dwarf, u64 pc, int *preg, int *poffset) }return -1 ; } -#endif /* HAVE_DWARF_CFI_SUPPORT */ /* * die_has_loclist - Check if DW_AT_location of @vr_die is a location list tools/perf/util/dwarf-aux.h +0
-54@@ -156,8 +156,6 @@ Dwarf_Die *die_get_member_type(Dwarf_Die *type_die, int offset, Dwarf_Die *die_m/* Return type info where the pointer and offset point to */ Dwarf_Die *die_deref_ptr_type (Dwarf_Die *ptr_die,int offset,Dwarf_Die *die_mem); -#ifdef HAVE_DWARF_GETLOCATIONS_SUPPORT -/* Get byte offset range of given variable DIE */ int die_get_var_range (Dwarf_Die *sp_die,Dwarf_Die *vr_die,struct strbuf *buf);@@ -176,59 +174,7 @@ void die_collect_vars(Dwarf_Die *sc_die, struct die_var_type **var_types);/* Save all global variables in this CU */ void die_collect_global_vars (Dwarf_Die *cu_die,struct die_var_type **var_types); -#else /* HAVE_DWARF_GETLOCATIONS_SUPPORT */ -#include <errno.h> - -static inline int die_get_var_range (Dwarf_Die *sp_die__maybe_unused , -Dwarf_Die *vr_die__maybe_unused , -struct strbuf *buf__maybe_unused ) -{ -return -ENOTSUP ; -} - -static inline Dwarf_Die *die_find_variable_by_reg (Dwarf_Die *sc_die__maybe_unused , -Dwarf_Addr pc__maybe_unused , -int reg__maybe_unused , -int *poffset__maybe_unused , -bool is_fbreg__maybe_unused , -Dwarf_Die *die_mem__maybe_unused ) -{ -return NULL ; -} - -static inline Dwarf_Die *die_find_variable_by_addr (Dwarf_Die *sc_die__maybe_unused , -Dwarf_Addr addr__maybe_unused , -Dwarf_Die *die_mem__maybe_unused , -int *offset__maybe_unused ) -{ -return NULL ; -} - -static inline void die_collect_vars (Dwarf_Die *sc_die__maybe_unused , -struct die_var_type **var_types__maybe_unused ) -{ -} - -static inline void die_collect_global_vars (Dwarf_Die *cu_die__maybe_unused , -struct die_var_type **var_types__maybe_unused ) -{ -} - -#endif /* HAVE_DWARF_GETLOCATIONS_SUPPORT */ - -#ifdef HAVE_DWARF_CFI_SUPPORT -/* Get the frame base information from CFA */ int die_get_cfa (Dwarf *dwarf,u64 pc,int *preg,int *poffset); -#else /* HAVE_DWARF_CFI_SUPPORT */ - -static inline int die_get_cfa (Dwarf *dwarf__maybe_unused ,u64 pc__maybe_unused , -int *preg__maybe_unused ,int *poffset__maybe_unused ) -{ -return -1 ; -} - -#endif /* HAVE_DWARF_CFI_SUPPORT */ -#endif /* _DWARF_AUX_H */ tools/perf/util/probe-finder.c +0
-4@@ -602,7 +602,6 @@ static int call_probe_finder(Dwarf_Die *sc_die, struct probe_finder *pf) ret =dwarf_getlocation_addr (&fb_attr, pf->addr, &pf->fb_ops, &nops,1 );if (ret <=0 || nops ==0 ) { pf->fb_ops =NULL ; -#ifdef HAVE_DWARF_CFI_SUPPORT }else if (nops ==1 && pf->fb_ops[0 ].atom == DW_OP_call_frame_cfa && (pf->cfi_eh !=NULL || pf->cfi_dbg !=NULL )) { if ((dwarf_cfi_addrframe(pf->cfi_eh, pf->addr, &frame) != 0 &&@@ -613,7 +612,6 @@ static int call_probe_finder(Dwarf_Die *sc_die, struct probe_finder *pf)free (frame);return -ENOENT ; } -#endif /* HAVE_DWARF_CFI_SUPPORT */ }/* Call finder's callback handler */ @@ -1138,7 +1136,6 @@ static int debuginfo__find_probes(struct debuginfo *dbg, pf->machine = ehdr.e_machine; -#ifdef HAVE_DWARF_CFI_SUPPORT do { GElf_Shdr shdr;@@ -1148,7 +1145,6 @@ static int debuginfo__find_probes(struct debuginfo *dbg, pf->cfi_dbg =dwarf_getcfi (dbg->dbg); }while (0 ); -#endif /* HAVE_DWARF_CFI_SUPPORT */ ret =debuginfo__find_probe_location (dbg, pf);return ret;Signature · 1 line
2.46.0.792.g87dc391469-goog
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerPATCH 9/11 perf build: Rename HAVE_DWARF_SUPPORT to HAVE_LIBDW_SUPPORT
In reply to Ian RogersIn Makefile.config for unwinding the name dwarf implies either libunwind or libdw. Make it clearer that HAVE_DWARF_SUPPORT is really just defined when libdw is present by renaming to HAVE_LIBDW_SUPPORT.
--- tools/perf/Documentation/perf-check.txt | 6 +++--- tools/perf/Makefile.config | 2 +- tools/perf/arch/powerpc/annotate/instructions.c | 4 ++-- tools/perf/arch/x86/annotate/instructions.c | 2 +- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 6 +++--- tools/perf/builtin-probe.c | 12 ++++++------ tools/perf/builtin-report.c | 4 ++-- tools/perf/util/annotate-data.h | 8 ++++---- tools/perf/util/debuginfo.h | 6 +++--- tools/perf/util/disasm.c | 4 ++-- tools/perf/util/disasm.h | 4 ++-- tools/perf/util/genelf.c | 4 ++-- tools/perf/util/genelf.h | 2 +- tools/perf/util/include/dwarf-regs.h | 6 +++--- tools/perf/util/probe-event.c | 4 ++-- tools/perf/util/probe-finder.h | 4 ++-- 17 files changed, 40 insertions(+), 40 deletions(-)
tools/perf/Documentation/perf-check.txt +3 -3
@@ -47,15 +47,15 @@ feature:: bpf / HAVE_LIBBPF_SUPPORT bpf_skeletons / HAVE_BPF_SKEL debuginfod / HAVE_DEBUGINFOD_SUPPORT - dwarf / HAVE_DWARF_SUPPORT - dwarf_getlocations / HAVE_DWARF_SUPPORT + dwarf / HAVE_LIBDW_SUPPORT + dwarf_getlocations / HAVE_LIBDW_SUPPORT dwarf-unwind / HAVE_DWARF_UNWIND_SUPPORT auxtrace / HAVE_AUXTRACE_SUPPORT libaudit / HAVE_LIBAUDIT_SUPPORT libbfd / HAVE_LIBBFD_SUPPORT libcapstone / HAVE_LIBCAPSTONE_SUPPORT libcrypto / HAVE_LIBCRYPTO_SUPPORT - libdw-dwarf-unwind / HAVE_DWARF_SUPPORT + libdw-dwarf-unwind / HAVE_LIBDW_SUPPORT libelf / HAVE_LIBELF_SUPPORT libnuma / HAVE_LIBNUMA_SUPPORT libopencsd / HAVE_CSTRACE_SUPPORT
tools/perf/Makefile.config +1 -1
@@ -556,7 +556,7 @@ ifndef NO_LIBELF $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) NO_LIBDW := 1 else - CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) + CFLAGS += -DHAVE_LIBDW_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF)
tools/perf/arch/powerpc/annotate/instructions.c +2 -2
@@ -255,7 +255,7 @@ static struct ins_ops *check_ppc_insn(struct disasm_line *dl)* is moved to r31. update_insn_state_powerpc tracks these state * changes */ -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT static void update_insn_state_powerpc (struct type_state *state,struct data_loc_info *dloc,Dwarf_Die * cu_die__maybe_unused ,struct disasm_line *dl)
@@ -300,7 +300,7 @@ static void update_insn_state_powerpc(struct type_state *state, insn_offset, src->reg1, dst->reg1); pr_debug_type_name(&tsr->type, tsr->kind); } -#endif /* HAVE_DWARF_SUPPORT */ +#endif /* HAVE_LIBDW_SUPPORT */ static int powerpc__annotate_init (struct arch *arch,char *cpuid__maybe_unused ) {
tools/perf/arch/x86/annotate/instructions.c +1 -1
@@ -207,7 +207,7 @@ static int x86__annotate_init(struct arch *arch, char *cpuid)return err; } -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT static void update_insn_state_x86 (struct type_state *state,struct data_loc_info *dloc,Dwarf_Die *cu_die,struct disasm_line *dl)
tools/perf/builtin-annotate.c +1 -1
@@ -840,7 +840,7 @@ int cmd_annotate(int argc, const char **argv) }#endif -#ifndef HAVE_DWARF_SUPPORT +#ifndef HAVE_LIBDW_SUPPORT if (annotate.data_type) {pr_err ("Error: Data type profiling is disabled due to missing DWARF support\n" );return -ENOTSUP ;
tools/perf/builtin-check.c +3 -3
@@ -27,15 +27,15 @@ struct feature_status supported_features[] = {FEATURE_STATUS ("bpf" ,HAVE_LIBBPF_SUPPORT ),FEATURE_STATUS ("bpf_skeletons" ,HAVE_BPF_SKEL ),FEATURE_STATUS ("debuginfod" ,HAVE_DEBUGINFOD_SUPPORT ), -FEATURE_STATUS ("dwarf" ,HAVE_DWARF_SUPPORT ), -FEATURE_STATUS ("dwarf_getlocations" ,HAVE_DWARF_SUPPORT ), +FEATURE_STATUS ("dwarf" ,HAVE_LIBDW_SUPPORT ), +FEATURE_STATUS ("dwarf_getlocations" ,HAVE_LIBDW_SUPPORT ),FEATURE_STATUS ("dwarf-unwind" ,HAVE_DWARF_UNWIND_SUPPORT ),FEATURE_STATUS ("auxtrace" ,HAVE_AUXTRACE_SUPPORT ),FEATURE_STATUS ("libaudit" ,HAVE_LIBAUDIT_SUPPORT ),FEATURE_STATUS ("libbfd" ,HAVE_LIBBFD_SUPPORT ),FEATURE_STATUS ("libcapstone" ,HAVE_LIBCAPSTONE_SUPPORT ),FEATURE_STATUS ("libcrypto" ,HAVE_LIBCRYPTO_SUPPORT ), -FEATURE_STATUS ("libdw-dwarf-unwind" ,HAVE_DWARF_SUPPORT ), +FEATURE_STATUS ("libdw-dwarf-unwind" ,HAVE_LIBDW_SUPPORT ),FEATURE_STATUS ("libelf" ,HAVE_LIBELF_SUPPORT ),FEATURE_STATUS ("libnuma" ,HAVE_LIBNUMA_SUPPORT ), FEATURE_STATUS("libopencsd", HAVE_CSTRACE_SUPPORT),
tools/perf/builtin-probe.c +6 -6
@@ -229,7 +229,7 @@ static int opt_set_target_ns(const struct option *opt __maybe_unused,/* Command option callbacks */ -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT static int opt_show_lines (const struct option *opt,const char *str,int unset__maybe_unused ) {
@@ -505,7 +505,7 @@ static int perf_del_probe_events(struct strfilter *filter)return ret; } -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT #define PROBEDEF_STR \ "[EVENT=]FUNC[@SRC][+OFF|%return|:RL|;PT]|SRC:AL|SRC;PT [[NAME=]ARG ...]"#else
@@ -521,7 +521,7 @@ __cmd_probe(int argc, const char **argv)"perf probe [<options>] --add 'PROBEDEF' [--add 'PROBEDEF' ...]" ,"perf probe [<options>] --del '[GROUP:]EVENT' ..." ,"perf probe --list [GROUP:]EVENT ..." , -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT"perf probe [<options>] --line 'LINEDESC'" , "perf probe [<options>] --vars 'PROBEPOINT'", #endif
@@ -545,7 +545,7 @@ __cmd_probe(int argc, const char **argv)"\t\tFUNC:\tFunction name\n" "\t\tOFF:\tOffset from function entry (in byte)\n" "\t\t%return:\tPut the probe at function return\n" -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT"\t\tSRC:\tSource code path\n" "\t\tRL:\tRelative line number from function entry.\n" "\t\tAL:\tAbsolute line number in file.\n"
@@ -612,7 +612,7 @@ __cmd_probe(int argc, const char **argv)set_option_flag (options,'d' ,"del" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'D' ,"definition" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'l' ,"list" ,PARSE_OPT_EXCLUSIVE ); -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT set_option_flag (options,'L' ,"line" ,PARSE_OPT_EXCLUSIVE );set_option_flag (options,'V' ,"vars" ,PARSE_OPT_EXCLUSIVE );#else
@@ -694,7 +694,7 @@ __cmd_probe(int argc, const char **argv)if (ret <0 )pr_err_with_code (" Error: Failed to show functions." , ret);return ret; -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT case 'L' : ret =show_line_range (¶ms->line_range, params->target, params->nsi, params->uprobes);
tools/perf/builtin-report.c +2 -2
@@ -455,7 +455,7 @@ static int report__setup_sample_type(struct report *rep)if (!(evlist__combined_branch_type (session->evlist) &PERF_SAMPLE_BRANCH_ANY )) rep->nonany_branch_mode = true; -#if !defined(HAVE_LIBUNWIND_SUPPORT ) && !defined(HAVE_DWARF_SUPPORT ) +#if !defined(HAVE_LIBUNWIND_SUPPORT ) && !defined(HAVE_LIBDW_SUPPORT )if (dwarf_callchain_users) {ui__warning ("Please install libunwind or libdw " "development packages during the perf build.\n" );
@@ -1701,7 +1701,7 @@ int cmd_report(int argc, const char **argv) report.data_type = true; annotate_opts.annotate_src = false; -#ifndef HAVE_DWARF_SUPPORT +#ifndef HAVE_LIBDW_SUPPORT pr_err ("Error: Data type profiling is disabled due to missing DWARF support\n" ); goto error;#endif
tools/perf/util/annotate-data.h +4 -4
@@ -9,7 +9,7 @@#include "dwarf-regs.h" #include "annotate.h" -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT #include "debuginfo.h" #endif
@@ -165,7 +165,7 @@ struct annotated_data_stat { };extern struct annotated_data_stat ann_data_stat; -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT /* * Type information in a register, valid when @ok is true. * The @caller_saved registers are invalidated after a function call.
@@ -244,7 +244,7 @@ bool get_global_var_info(struct data_loc_info *dloc, u64 addr, const char **var_name, int *var_offset);void pr_debug_type_name (Dwarf_Die *die,enum type_state_kind kind); -#else /* HAVE_DWARF_SUPPORT */ +#else /* HAVE_LIBDW_SUPPORT */ static inline struct annotated_data_type *find_data_type (struct data_loc_info *dloc__maybe_unused )
@@ -276,7 +276,7 @@ static inline int hist_entry__annotate_data_tty(struct hist_entry *he __maybe_unreturn -1 ; } -#endif /* HAVE_DWARF_SUPPORT */ +#endif /* HAVE_LIBDW_SUPPORT */ #ifdef HAVE_SLANG_SUPPORT int hist_entry__annotate_data_tui(struct hist_entry *he, struct evsel *evsel,
tools/perf/util/debuginfo.h +3 -3
@@ -5,7 +5,7 @@#include <errno.h> #include <linux/compiler.h> -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT #include "dwarf-aux.h"
@@ -25,7 +25,7 @@ void debuginfo__delete(struct debuginfo *dbg);int debuginfo__get_text_offset (struct debuginfo *dbg,Dwarf_Addr *offs,bool adjust_offset); -#else /* HAVE_DWARF_SUPPORT */ +#else /* HAVE_LIBDW_SUPPORT */ /* dummy debug information structure */ struct debuginfo {
@@ -49,7 +49,7 @@ static inline int debuginfo__get_text_offset(struct debuginfo *dbg __maybe_unusereturn -EINVAL ; } -#endif /* HAVE_DWARF_SUPPORT */ +#endif /* HAVE_LIBDW_SUPPORT */ #ifdef HAVE_DEBUGINFOD_SUPPORT int get_source_from_debuginfod(const char *raw_path, const char *sbuild_id,
tools/perf/util/disasm.c +2 -2
@@ -151,14 +151,14 @@ static struct arch architectures[] = { .memory_ref_char = '(', .imm_char = '$', }, -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT .update_insn_state = update_insn_state_x86,#endif }, { .name = "powerpc", .init = powerpc__annotate_init, -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT .update_insn_state = update_insn_state_powerpc,#endif },
tools/perf/util/disasm.h +2 -2
@@ -4,7 +4,7 @@#include "map_symbol.h" -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT #include "dwarf-aux.h" #endif
@@ -39,7 +39,7 @@ struct arch {char memory_ref_char;char imm_char; } objdump; -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT void (*update_insn_state)(struct type_state *state,struct data_loc_info *dloc,Dwarf_Die *cu_die,struct disasm_line *dl);
tools/perf/util/genelf.c +2 -2
@@ -16,7 +16,7 @@#include <inttypes.h> #include <fcntl.h> #include <err.h> -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT #include <dwarf.h> #endif
@@ -499,7 +499,7 @@ jit_write_elf(int fd, uint64_t load_addr, const char *sym, shdr->sh_size =sizeof (bnote); shdr->sh_entsize =0 ; -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT if (debug && nr_debug_entries) { retval =jit_add_debug_info (e, load_addr, debug, nr_debug_entries);if (retval)
tools/perf/util/genelf.h +1 -1
@@ -8,7 +8,7 @@int jit_write_elf (int fd,uint64_t code_addr,const char *sym,const void *code,int csize,void *debug,int nr_debug_entries,void *unwinding,uint64_t unwinding_header_size,uint64_t unwinding_size); -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT /* genelf_debug.c */ int jit_add_debug_info (Elf *e,uint64_t code_addr,void *debug,int nr_debug_entries);#endif
tools/perf/util/include/dwarf-regs.h +3 -3
@@ -6,7 +6,7 @@#define DWARF_REG_PC 0xd3af9c/* random number */ #define DWARF_REG_FB 0xd3affb/* random number */ -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT const char *get_arch_regstr (unsigned int n);/* * get_dwarf_regstr - Returns ftrace register string from DWARF regnum
@@ -23,7 +23,7 @@ int get_arch_regnum(const char *name);*/ int get_dwarf_regnum (const char *name,unsigned int machine); -#else /* HAVE_DWARF_SUPPORT */ +#else /* HAVE_LIBDW_SUPPORT */ static inline int get_dwarf_regnum (const char *name__maybe_unused ,unsigned int machine__maybe_unused )
@@ -32,7 +32,7 @@ static inline int get_dwarf_regnum(const char *name __maybe_unused, }#endif -#if !defined(__powerpc__) || !defined(HAVE_DWARF_SUPPORT ) +#if !defined(__powerpc__) || !defined(HAVE_LIBDW_SUPPORT )static inline void get_powerpc_regs (u32 raw_insn__maybe_unused ,int is_source__maybe_unused ,struct annotated_op_loc *op_loc__maybe_unused ) {
tools/perf/util/probe-event.c +2 -2
@@ -342,7 +342,7 @@ static char *find_module_name(const char *module)return mod_name; } -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT static int kernel_get_module_dso (const char *module,struct dso **pdso) {
@@ -1250,7 +1250,7 @@ int show_available_vars(struct perf_probe_event *pevs, int npevs, return ret; } -#else /* !HAVE_DWARF_SUPPORT */ +#else /* !HAVE_LIBDW_SUPPORT */ static void debuginfo_cache__exit (void ) {
tools/perf/util/probe-finder.h +2 -2
@@ -21,7 +21,7 @@ static inline int is_c_varname(const char *name)return isalpha (name[0 ]) || name[0 ] =='_' ; } -#ifdef HAVE_DWARF_SUPPORT +#ifdef HAVE_LIBDW_SUPPORT #include "dwarf-aux.h" #include "debuginfo.h"
@@ -102,6 +102,6 @@ struct line_finder {int found; }; -#endif /* HAVE_DWARF_SUPPORT */ +#endif /* HAVE_LIBDW_SUPPORT */ #endif /*_PROBE_FINDER_H */
Signature · 1 line
2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 09/11] perf build: Rename HAVE_DWARF_SUPPORT to HAVE_LIBDW_SUPPORT
Quoted 30 lines · Ian Rogers
In Makefile.config for unwinding the name dwarf implies either libunwind or libdw. Make it clearer that
HAVE_DWARF_SUPPORTis really just defined when libdw is present by renaming toHAVE_LIBDW_SUPPORT.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-check.txt | 6 +++--- tools/perf/Makefile.config | 2 +- tools/perf/arch/powerpc/annotate/instructions.c | 4 ++-- tools/perf/arch/x86/annotate/instructions.c | 2 +- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 6 +++--- tools/perf/builtin-probe.c | 12 ++++++------ tools/perf/builtin-report.c | 4 ++-- tools/perf/util/annotate-data.h | 8 ++++---- tools/perf/util/debuginfo.h | 6 +++--- tools/perf/util/disasm.c | 4 ++-- tools/perf/util/disasm.h | 4 ++-- tools/perf/util/genelf.c | 4 ++-- tools/perf/util/genelf.h | 2 +- tools/perf/util/include/dwarf-regs.h | 6 +++--- tools/perf/util/probe-event.c | 4 ++-- tools/perf/util/probe-finder.h | 4 ++-- 17 files changed, 40 insertions(+), 40 deletions(-)tools/perf/Documentation/perf-check.txt +0
-0
tools/perf/Documentation/perf-check.txt feature:: · line 51
Quoted 7 lines
tools/perf/Documentation/perf-check.txt +2
-2@@ -47,15 +47,15 @@ feature:: bpf / HAVE_LIBBPF_SUPPORT bpf_skeletons / HAVE_BPF_SKEL debuginfod / HAVE_DEBUGINFOD_SUPPORT - dwarf / HAVE_DWARF_SUPPORT - dwarf_getlocations / HAVE_DWARF_SUPPORT + dwarf / HAVE_LIBDW_SUPPORT + dwarf_getlocations / HAVE_LIBDW_SUPPORT
I'm not sure if we really want to display dwarf_getlocatiosn as it's too implementation detail IMHO. Maybe just 'dwarf' or 'libdw' is enough.
Thanks,
Namhyung
Quoted 11 lines
tools/perf/Documentation/perf-check.txt +1
-1dwarf-unwind / HAVE_DWARF_UNWIND_SUPPORT auxtrace / HAVE_AUXTRACE_SUPPORT libaudit / HAVE_LIBAUDIT_SUPPORT libbfd / HAVE_LIBBFD_SUPPORT libcapstone / HAVE_LIBCAPSTONE_SUPPORT libcrypto / HAVE_LIBCRYPTO_SUPPORT - libdw-dwarf-unwind / HAVE_DWARF_SUPPORT + libdw-dwarf-unwind / HAVE_LIBDW_SUPPORT libelf / HAVE_LIBELF_SUPPORT libnuma / HAVE_LIBNUMA_SUPPORT libopencsd / HAVE_CSTRACE_SUPPORT
Ian Rogers
reviewerRe: [PATCH v1 09/11] perf build: Rename HAVE_DWARF_SUPPORT to HAVE_LIBDW_SUPPORT
Quoted 27 lines · Namhyung Kim
Quoted 25 lines · Ian Rogers
In Makefile.config for unwinding the name dwarf implies either libunwind or libdw. Make it clearer that
HAVE_DWARF_SUPPORTis really just defined when libdw is present by renaming toHAVE_LIBDW_SUPPORT.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-check.txt | 6 +++--- tools/perf/Makefile.config | 2 +- tools/perf/arch/powerpc/annotate/instructions.c | 4 ++-- tools/perf/arch/x86/annotate/instructions.c | 2 +- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 6 +++--- tools/perf/builtin-probe.c | 12 ++++++------ tools/perf/builtin-report.c | 4 ++-- tools/perf/util/annotate-data.h | 8 ++++---- tools/perf/util/debuginfo.h | 6 +++--- tools/perf/util/disasm.c | 4 ++-- tools/perf/util/disasm.h | 4 ++-- tools/perf/util/genelf.c | 4 ++-- tools/perf/util/genelf.h | 2 +- tools/perf/util/include/dwarf-regs.h | 6 +++--- tools/perf/util/probe-event.c | 4 ++-- tools/perf/util/probe-finder.h | 4 ++-- 17 files changed, 40 insertions(+), 40 deletions(-)
Quoted 15 lines
tools/perf/Documentation/perf-check.txt feature:: · line 51
Quoted 12 lines
tools/perf/Documentation/perf-check.txt +2
-2@@ -47,15 +47,15 @@ feature:: bpf / HAVE_LIBBPF_SUPPORT bpf_skeletons / HAVE_BPF_SKEL debuginfod / HAVE_DEBUGINFOD_SUPPORT - dwarf / HAVE_DWARF_SUPPORT - dwarf_getlocations / HAVE_DWARF_SUPPORT + dwarf / HAVE_LIBDW_SUPPORT + dwarf_getlocations / HAVE_LIBDW_SUPPORTI'm not sure if we really want to display
dwarf_getlocatiosnas it's too implementation detail IMHO. Maybe just 'dwarf' or 'libdw' is enough.
I think that's follow up work. I think we should rename dwarf to libdw given the confusion using dwarf here can create.
Thanks,
Ian
Quoted 15 lines
Thanks,
NamhyungQuoted 11 lines
tools/perf/Documentation/perf-check.txt +1
-1dwarf-unwind / HAVE_DWARF_UNWIND_SUPPORT auxtrace / HAVE_AUXTRACE_SUPPORT libaudit / HAVE_LIBAUDIT_SUPPORT libbfd / HAVE_LIBBFD_SUPPORT libcapstone / HAVE_LIBCAPSTONE_SUPPORT libcrypto / HAVE_LIBCRYPTO_SUPPORT - libdw-dwarf-unwind / HAVE_DWARF_SUPPORT + libdw-dwarf-unwind / HAVE_LIBDW_SUPPORT libelf / HAVE_LIBELF_SUPPORT libnuma / HAVE_LIBNUMA_SUPPORT libopencsd / HAVE_CSTRACE_SUPPORT
Namhyung Kim
maintainerRe: [PATCH v1 09/11] perf build: Rename HAVE_DWARF_SUPPORT to HAVE_LIBDW_SUPPORT
Quoted 44 lines · Ian Rogers
Quoted 27 lines · Namhyung Kim
Quoted 25 lines · Ian Rogers
In Makefile.config for unwinding the name dwarf implies either libunwind or libdw. Make it clearer that
HAVE_DWARF_SUPPORTis really just defined when libdw is present by renaming toHAVE_LIBDW_SUPPORT.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-check.txt | 6 +++--- tools/perf/Makefile.config | 2 +- tools/perf/arch/powerpc/annotate/instructions.c | 4 ++-- tools/perf/arch/x86/annotate/instructions.c | 2 +- tools/perf/builtin-annotate.c | 2 +- tools/perf/builtin-check.c | 6 +++--- tools/perf/builtin-probe.c | 12 ++++++------ tools/perf/builtin-report.c | 4 ++-- tools/perf/util/annotate-data.h | 8 ++++---- tools/perf/util/debuginfo.h | 6 +++--- tools/perf/util/disasm.c | 4 ++-- tools/perf/util/disasm.h | 4 ++-- tools/perf/util/genelf.c | 4 ++-- tools/perf/util/genelf.h | 2 +- tools/perf/util/include/dwarf-regs.h | 6 +++--- tools/perf/util/probe-event.c | 4 ++-- tools/perf/util/probe-finder.h | 4 ++-- 17 files changed, 40 insertions(+), 40 deletions(-)Quoted 15 lines
tools/perf/Documentation/perf-check.txt feature:: · line 51
Quoted 12 lines
tools/perf/Documentation/perf-check.txt +2
-2@@ -47,15 +47,15 @@ feature:: bpf / HAVE_LIBBPF_SUPPORT bpf_skeletons / HAVE_BPF_SKEL debuginfod / HAVE_DEBUGINFOD_SUPPORT - dwarf / HAVE_DWARF_SUPPORT - dwarf_getlocations / HAVE_DWARF_SUPPORT + dwarf / HAVE_LIBDW_SUPPORT + dwarf_getlocations / HAVE_LIBDW_SUPPORTI'm not sure if we really want to display
dwarf_getlocatiosnas it's too implementation detail IMHO. Maybe just 'dwarf' or 'libdw' is enough.
Quoted 2 lines
I think that's follow up work. I think we should rename dwarf to libdw given the confusion using dwarf here can create.
Ok, agreed.
Thanks,
Namhyung
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
tools/perf/Documentation/perf-check.txt +1
-1dwarf-unwind / HAVE_DWARF_UNWIND_SUPPORT auxtrace / HAVE_AUXTRACE_SUPPORT libaudit / HAVE_LIBAUDIT_SUPPORT libbfd / HAVE_LIBBFD_SUPPORT libcapstone / HAVE_LIBCAPSTONE_SUPPORT libcrypto / HAVE_LIBCRYPTO_SUPPORT - libdw-dwarf-unwind / HAVE_DWARF_SUPPORT + libdw-dwarf-unwind / HAVE_LIBDW_SUPPORT libelf / HAVE_LIBELF_SUPPORT libnuma / HAVE_LIBNUMA_SUPPORT libopencsd / HAVE_CSTRACE_SUPPORT
Ian Rogers
reviewerPATCH 10/11 perf build: Rename CONFIG_DWARF to CONFIG_LIBDW
In reply to Ian RogersIn Makefile.config for unwinding the name dwarf implies either libunwind or libdw. Make it clearer that CONFIG_DWARF is really just defined when libdw is present by renaming to CONFIG_LIBDW.
--- tools/perf/Makefile.config | 2 +- tools/perf/arch/arm/util/Build | 2 +- tools/perf/arch/arm64/util/Build | 2 +- tools/perf/arch/csky/util/Build | 2 +- tools/perf/arch/loongarch/util/Build | 2 +- tools/perf/arch/mips/util/Build | 2 +- tools/perf/arch/powerpc/util/Build | 4 ++-- tools/perf/arch/riscv/util/Build | 2 +- tools/perf/arch/s390/util/Build | 2 +- tools/perf/arch/sh/util/Build | 2 +- tools/perf/arch/sparc/util/Build | 2 +- tools/perf/arch/x86/util/Build | 2 +- tools/perf/arch/xtensa/util/Build | 2 +- tools/perf/util/Build | 12 ++++++------ 14 files changed, 20 insertions(+), 20 deletions(-)
tools/perf/Makefile.config +1 -1
@@ -559,7 +559,7 @@ ifndef NO_LIBELF CFLAGS += -DHAVE_LIBDW_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} - $(call detected,CONFIG_DWARF) + $(call detected,CONFIG_LIBDW) endif # PERF_HAVE_DWARF_REGS endif # NO_LIBDW
tools/perf/arch/arm/util/Build +1 -1
@@ -1,6 +1,6 @@ perf-util-y += perf_regs.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
tools/perf/arch/arm64/util/Build +1 -1
@@ -4,7 +4,7 @@ perf-util-y += perf_regs.o perf-util-y += tsc.o perf-util-y += pmu.o perf-util-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
tools/perf/arch/csky/util/Build +1 -1
@@ -1,4 +1,4 @@ perf-util-y += perf_regs.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
tools/perf/arch/loongarch/util/Build +1 -1
@@ -1,7 +1,7 @@ perf-util-y += header.o perf-util-y += perf_regs.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o perf-util-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o
tools/perf/arch/mips/util/Build +1 -1
@@ -1,3 +1,3 @@ perf-util-y += perf_regs.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o
tools/perf/arch/powerpc/util/Build +2 -2
@@ -7,8 +7,8 @@ perf-util-y += sym-handling.o perf-util-y += evsel.o perf-util-y += event.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o -perf-util-$(CONFIG_DWARF) += skip-callchain-idx.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += skip-callchain-idx.o perf-util-$(CONFIG_LIBUNWIND) += unwind-libunwind.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
tools/perf/arch/riscv/util/Build +1 -1
@@ -2,5 +2,5 @@ perf-util-y += perf_regs.o perf-util-y += header.o perf-util-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
tools/perf/arch/s390/util/Build +1 -1
@@ -2,7 +2,7 @@ perf-util-y += header.o perf-util-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o perf-util-y += perf_regs.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o perf-util-y += machine.o
tools/perf/arch/sh/util/Build +1 -1
@@ -1 +1 @@ -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o
tools/perf/arch/sparc/util/Build +1 -1
@@ -1 +1 @@ -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o
tools/perf/arch/x86/util/Build +1 -1
@@ -12,7 +12,7 @@ perf-util-y += evsel.o perf-util-y += iostat.o perf-util-y += env.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o perf-util-$(CONFIG_BPF_PROLOGUE) += dwarf-regs.o perf-util-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind.o
tools/perf/arch/xtensa/util/Build +1 -1
@@ -1 +1 @@ -perf-$(CONFIG_DWARF) += dwarf-regs.o +perf-$(CONFIG_LIBDW) += dwarf-regs.o
tools/perf/util/Build +6 -6
@@ -199,11 +199,11 @@ ifndef CONFIG_SETNS perf-util-y += setns.o endif -perf-util-$(CONFIG_DWARF) += probe-finder.o -perf-util-$(CONFIG_DWARF) += dwarf-aux.o -perf-util-$(CONFIG_DWARF) += dwarf-regs.o -perf-util-$(CONFIG_DWARF) += debuginfo.o -perf-util-$(CONFIG_DWARF) += annotate-data.o +perf-util-$(CONFIG_LIBDW) += probe-finder.o +perf-util-$(CONFIG_LIBDW) += dwarf-aux.o +perf-util-$(CONFIG_LIBDW) += dwarf-regs.o +perf-util-$(CONFIG_LIBDW) += debuginfo.o +perf-util-$(CONFIG_LIBDW) += annotate-data.o perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o perf-util-$(CONFIG_LOCAL_LIBUNWIND) += unwind-libunwind-local.o
@@ -234,7 +234,7 @@ perf-util-$(CONFIG_LIBLLVM) += llvm-c-helpers.o ifdef CONFIG_JITDUMP perf-util-$(CONFIG_LIBELF) += jitdump.o perf-util-$(CONFIG_LIBELF) += genelf.o -perf-util-$(CONFIG_DWARF) += genelf_debug.o +perf-util-$(CONFIG_LIBDW) += genelf_debug.o endif perf-util-y += perf-hooks.o
Signature · 1 line
2.46.0.792.g87dc391469-goog
Ian Rogers
reviewerPATCH 11/11 perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
In reply to Ian RogersThe name dwarf can imply libunwind support, whereas PERF_HAVE_DWARF_REGS is only enabled with libdw support. Rename to make it clearer there is a libdw connection.
--- tools/perf/Makefile.config | 4 ++-- tools/perf/arch/arm/Makefile | 2 +- tools/perf/arch/arm64/Makefile | 2 +- tools/perf/arch/csky/Makefile | 2 +- tools/perf/arch/loongarch/Makefile | 2 +- tools/perf/arch/mips/Makefile | 2 +- tools/perf/arch/powerpc/Makefile | 2 +- tools/perf/arch/riscv/Makefile | 2 +- tools/perf/arch/s390/Makefile | 2 +- tools/perf/arch/sh/Makefile | 2 +- tools/perf/arch/sparc/Makefile | 2 +- tools/perf/arch/x86/Makefile | 2 +- tools/perf/arch/xtensa/Makefile | 2 +- 13 files changed, 14 insertions(+), 14 deletions(-)
tools/perf/Makefile.config +2 -2
@@ -552,7 +552,7 @@ ifndef NO_LIBELF endif ifndef NO_LIBDW - ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) + ifeq ($(origin PERF_HAVE_LIBDW_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled) NO_LIBDW := 1 else
@@ -560,7 +560,7 @@ ifndef NO_LIBELF LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_LIBDW) - endif # PERF_HAVE_DWARF_REGS + endif # PERF_HAVE_LIBDW_REGS endif # NO_LIBDW ifndef NO_LIBBPF
tools/perf/arch/arm/Makefile +1 -1
@@ -1,5 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif PERF_HAVE_JITDUMP := 1
tools/perf/arch/arm64/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0 ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif PERF_HAVE_JITDUMP := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1
tools/perf/arch/csky/Makefile +1 -1
@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif
tools/perf/arch/loongarch/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0 ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1 PERF_HAVE_JITDUMP := 1
tools/perf/arch/mips/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0 ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif # Syscall table generation for perf
tools/perf/arch/powerpc/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0 ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1
tools/perf/arch/riscv/Makefile +1 -1
@@ -1,5 +1,5 @@ ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1 PERF_HAVE_JITDUMP := 1
tools/perf/arch/s390/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1
tools/perf/arch/sh/Makefile +1 -1
@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif
tools/perf/arch/sparc/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif PERF_HAVE_JITDUMP := 1
tools/perf/arch/x86/Makefile +1 -1
@@ -1,6 +1,6 @@ # SPDX-License-Identifier: GPL-2.0 ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif HAVE_KVM_STAT_SUPPORT := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1
tools/perf/arch/xtensa/Makefile +1 -1
@@ -1,4 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW -PERF_HAVE_DWARF_REGS := 1 +PERF_HAVE_LIBDW_REGS := 1 endif
Signature · 1 line
2.46.0.792.g87dc391469-goog
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
Thanks,
Namhyung
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 9 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
Thanks,
Ian
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
Thanks,
Namhyung
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.
Thanks,
Ian
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 19 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
Quoted 7 lines
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.
I understand your point but calling it LIBDW_REGS looks unnatural to me.
Thanks,
Namhyung
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 29 lines · Namhyung Kim
Quoted 19 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
Quoted 7 lines
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.
I don't follow. Wouldn't it be unnatural to see PERF_HAVE_DWARF_REGS in libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the name PERF_HAVE_LIBDW_REGS better captures how the code is, makes the code more intention revealing and so readable, etc.
Thanks,
Ian
Re: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.
Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as DWARF_REGS.
Thank you,
Quoted 4 lines
Thanks,
Ian
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.
Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next
``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```
So the Makefile says that PERF_HAVE_DWARF_REGS is dependent on having NO_LIBDW, that is having libdw implies PERF_HAVE_DWARF_REGS is defined for csky.
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then NO_LIBDW will be set and PERF_HAVE_DWARF_REGS won't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.
In the code base right now PERF_HAVE_DWARF_REGS isn't a dwarf dependent thing, it is a libdw dependent thing, this is why PERF_HAVE_LIBDW_REGS is a more intention revealing name as it makes the connection explicit.
We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```
but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
Ian
Quoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```
Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.
I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers DWARF_REGS. :)
Thanks,
Namhyung
Quoted 54 lines
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then
NO_LIBDWwill be set andPERF_HAVE_DWARF_REGSwon't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.In the code base right now
PERF_HAVE_DWARF_REGSisn't a dwarf dependent thing, it is a libdw dependent thing, this is whyPERF_HAVE_LIBDW_REGSis a more intention revealing name as it makes the connection explicit.We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
IanQuoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 69 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.
Quoted 5 lines
I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)
The define is called PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.
Thanks,
Ian
Quoted 57 lines
Thanks,
NamhyungQuoted 54 lines
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then
NO_LIBDWwill be set andPERF_HAVE_DWARF_REGSwon't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.In the code base right now
PERF_HAVE_DWARF_REGSisn't a dwarf dependent thing, it is a libdw dependent thing, this is whyPERF_HAVE_LIBDW_REGSis a more intention revealing name as it makes the connection explicit.We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
IanQuoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>
Re: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 73 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 72 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
Quoted 6 lines
Quoted 2 lines
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.
Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
Thank you,
Quoted 62 lines
Thanks,
IanQuoted 57 lines
Thanks,
NamhyungQuoted 54 lines
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then
NO_LIBDWwill be set andPERF_HAVE_DWARF_REGSwon't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.In the code base right now
PERF_HAVE_DWARF_REGSisn't a dwarf dependent thing, it is a libdw dependent thing, this is whyPERF_HAVE_LIBDW_REGSis a more intention revealing name as it makes the connection explicit.We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
IanQuoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 80 lines · Masami Hiramatsu
Quoted 76 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 74 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGSis only enabled with libdw support. Rename to make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)
Quoted 6 lines
Quoted 3 lines
The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
So you are saying the #define is guarding a libdw feature? perf record/report --call-graph=dwarf is supported with either libdw or libunwind. The dwarf support in the tool may come from more sources hence wanting in this patch set to be clear what variable is guarding what. PERF_HAVE_DWARF_REGS is set to 1 for a specific set of architectures and only when libdw is present. The variable is saying that libdw supports the notion of registers needed for the #define HAVE_DWARF_SUPPORT that patch 9 in the series renamed to HAVE_LIBDW_SUPPORT. So I want the makefile variable PERF_HAVE_LIBDW_REGS to guard the #define HAVE_LIBDW_SUPPORT, rather than what is being argued by yourself and Namhyung that the #define HAVE_LIBDW_SUPPORT be guarded by a variable called PERF_HAVE_DWARF_REGS and that is only set when NO_LIBDW isn't set. We've made a digression into the name dwarf for a reason I can't fathom, at best it is inconsistent. Having dwarf registers is like having a bright sun or numeric numbers, it is a truism (playing devils advocate maybe if there were an ELF file format for postscript we could have a dwarf specification without registers). Anyway, I'm trying to connect the dots that libdw support controls the libdw type variables and defines hence not wanting 10 out of 11 patches applied.
Thanks,
Ian
Quoted 68 lines
Thank you,
Quoted 62 lines
Thanks,
IanQuoted 57 lines
Thanks,
NamhyungQuoted 54 lines
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then
NO_LIBDWwill be set andPERF_HAVE_DWARF_REGSwon't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.In the code base right now
PERF_HAVE_DWARF_REGSisn't a dwarf dependent thing, it is a libdw dependent thing, this is whyPERF_HAVE_LIBDW_REGSis a more intention revealing name as it makes the connection explicit.We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
IanQuoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>--
Masami Hiramatsu (Google) <[email protected]>
Re: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 88 lines · On Tue, 1 Oct 2024 18:31:43 -0700 Ian Rogers <[email protected]> wrote:
Quoted 80 lines · Masami Hiramatsu
Quoted 76 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 74 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
Quoted 3 lines · Ian Rogers
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGS is only enabled with libdw support. Rename to
make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether the arch has register mappings defined in DWARF standard. So I think it's better to keep the name for this case.
How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)Quoted 6 lines
Quoted 3 lines
The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
Quoted 13 lines
So you are saying the #define is guarding a libdw feature? perf record/report --call-graph=dwarf is supported with either libdw or libunwind. The dwarf support in the tool may come from more sources hence wanting in this patch set to be clear what variable is guarding what.
PERF_HAVE_DWARF_REGSis set to 1 for a specific set of architectures and only when libdw is present. The variable is saying that libdw supports the notion of registers needed for the #defineHAVE_DWARF_SUPPORTthat patch 9 in the series renamed toHAVE_LIBDW_SUPPORT. So I want the makefile variablePERF_HAVE_LIBDW_REGSto guard the #defineHAVE_LIBDW_SUPPORT, rather than what is being argued by yourself and Namhyung that the #defineHAVE_LIBDW_SUPPORTbe guarded by a variable calledPERF_HAVE_DWARF_REGSand that is only set whenNO_LIBDWisn't set.
It will be only used with the libdw, but I don't care. "HAVE_DWARF_REG" (internal config, just indicates the arch implemented feature) simply means there is arch/XXX/util/dwarf-regs.c. Also the APIs provided by the dwarf-regs.c are still based on DWARF standard, which defines registers by number like DW_OP_reg[0-31]. So the mapping of these suffix number and actual register must be defined for each architecture.
That is why I had introduced dwarf-regs.c and call it "dwarf"-regs. Even if the implementation depends on libdw, this dwarf-regs.c is still based on DWARF standard.
Quoted 7 lines
We've made a digression into the name dwarf for a reason I can't fathom, at best it is inconsistent. Having dwarf registers is like having a bright sun or numeric numbers, it is a truism (playing devils advocate maybe if there were an ELF file format for postscript we could have a dwarf specification without registers). Anyway, I'm trying to connect the dots that libdw support controls the libdw type variables and defines hence not wanting 10 out of 11 patches applied.
Oh, wait, I think we can apply other patches. I just don't like this patch. I think the other patches are good. But this is
Thank you,
Quoted 73 lines
Thanks,
IanQuoted 68 lines
Thank you,
Quoted 62 lines
Thanks,
IanQuoted 57 lines
Thanks,
NamhyungQuoted 54 lines
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then
NO_LIBDWwill be set andPERF_HAVE_DWARF_REGSwon't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.In the code base right now
PERF_HAVE_DWARF_REGSisn't a dwarf dependent thing, it is a libdw dependent thing, this is whyPERF_HAVE_LIBDW_REGSis a more intention revealing name as it makes the connection explicit.We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
IanQuoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>--
Masami Hiramatsu (Google) <[email protected]>
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 114 lines · Masami Hiramatsu
Quoted 88 lines · On Tue, 1 Oct 2024 18:31:43 -0700 Ian Rogers <[email protected]> wrote:
Quoted 80 lines · Masami Hiramatsu
Quoted 76 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 74 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
Quoted 6 lines · Namhyung Kim
On Tue, Sep 24, 2024 at 09:04:18AM -0700, Ian Rogers wrote:
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGS is only enabled with libdw support. Rename to
make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
While it only covers libdw, I think the idea of this macro is whether
the arch has register mappings defined in DWARF standard. So I think
it's better to keep the name for this case.How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)Quoted 6 lines
Quoted 3 lines
The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
Quoted 13 lines
So you are saying the #define is guarding a libdw feature? perf record/report --call-graph=dwarf is supported with either libdw or libunwind. The dwarf support in the tool may come from more sources hence wanting in this patch set to be clear what variable is guarding what.
PERF_HAVE_DWARF_REGSis set to 1 for a specific set of architectures and only when libdw is present. The variable is saying that libdw supports the notion of registers needed for the #defineHAVE_DWARF_SUPPORTthat patch 9 in the series renamed toHAVE_LIBDW_SUPPORT. So I want the makefile variablePERF_HAVE_LIBDW_REGSto guard the #defineHAVE_LIBDW_SUPPORT, rather than what is being argued by yourself and Namhyung that the #defineHAVE_LIBDW_SUPPORTbe guarded by a variable calledPERF_HAVE_DWARF_REGSand that is only set whenNO_LIBDWisn't set.It will be only used with the libdw, but I don't care. "
HAVE_DWARF_REG" (internal config, just indicates the arch implemented feature) simply means there isarch/XXX/util/dwarf-regs.c. Also the APIs provided by thedwarf-regs.care still based on DWARF standard, which defines registers by number like DW_OP_reg[0-31]. So the mapping of these suffix number and actual register must be defined for each architecture.
Quoted 3 lines
That is why I had introduced
dwarf-regs.cand call it "dwarf"-regs. Even if the implementation depends on libdw, thisdwarf-regs.cis still based on DWARF standard.
You seem to be missing the point of the series which is to clean up inconsistencies where dwarf is used to mean libdw. Here we have libdw guarding a #define with DWARF in the name, it should have libdw in the name as the patch cleans up. This is a coding thing and not a dwarf specificatin thing.
Quoted 10 lines
Quoted 7 lines
We've made a digression into the name dwarf for a reason I can't fathom, at best it is inconsistent. Having dwarf registers is like having a bright sun or numeric numbers, it is a truism (playing devils advocate maybe if there were an ELF file format for postscript we could have a dwarf specification without registers). Anyway, I'm trying to connect the dots that libdw support controls the libdw type variables and defines hence not wanting 10 out of 11 patches applied.
Oh, wait, I think we can apply other patches. I just don't like this patch. I think the other patches are good. But this is
Then we are intentionally aiming to be inconsistent, with libdw meaning dwarf with a #define that just states a truism. Arguably the code is better with none of the series applied.
Thanks,
Ian
Quoted 79 lines
Thank you,
Quoted 73 lines
Thanks,
IanQuoted 68 lines
Thank you,
Quoted 62 lines
Thanks,
IanQuoted 57 lines
Thanks,
NamhyungQuoted 54 lines
Dwarf in the code base implies libdw, libunwind and potentially other dwarf capable things like llvm. If we don't have libdw then
NO_LIBDWwill be set andPERF_HAVE_DWARF_REGSwon't be set. That is the more general dwarf thing will not be set because of missing libdw. This goes contrary to wanting this to be true whenever a dwarf thing is present - something that reflecting what the standard says would achieve.In the code base right now
PERF_HAVE_DWARF_REGSisn't a dwarf dependent thing, it is a libdw dependent thing, this is whyPERF_HAVE_LIBDW_REGSis a more intention revealing name as it makes the connection explicit.We could change the code so that in Makefile.config we set something like:
``` ... ifndef NO_LIBDW PERF_HAVE_DWARF := 1 ... ``` and in the arch/.../Makefiles change them to be: ``` if PERF_HAVE_DWARF PERF_HAVE_DWARF_REGS := 1 endif ```but this is going beyond the clean up this patch series was trying to achieve. I also don't know of an architecture where dwarf is present but registers are not, so having a definition for this case feels redundant.
Thanks,
IanQuoted 19 lines
Actually, there is a histrical reason I had called it is DWARF. I used to use "libdwarf", and switched to "libdw" for required features. So I know there are more than 1 implementation of DWARF library, and the libdwarf also uses the same operation number because it depends on the same standard.
https://github.com/davea42/libdwarf-code/blob/main/src/lib/libdwarf/dwarf.h#L809
So I think we'd better keep it call as
DWARF_REGS.Thank you,
Quoted 4 lines
Thanks,
Ian--
Masami Hiramatsu (Google) <[email protected]>--
Masami Hiramatsu (Google) <[email protected]>--
Masami Hiramatsu (Google) <[email protected]>
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 133 lines · Ian Rogers
Quoted 114 lines · Masami Hiramatsu
Quoted 88 lines · On Tue, 1 Oct 2024 18:31:43 -0700 Ian Rogers <[email protected]> wrote:
Quoted 80 lines · Masami Hiramatsu
Quoted 76 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 74 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
On Wed, Sep 25, 2024 at 8:27 PM Namhyung Kim <[email protected]> wrote:
On Tue, Sep 24, 2024 at 09:04:18AM -0700, Ian Rogers wrote:
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGS is only enabled with libdw support. Rename to
make it clearer there is a libdw connection.Quoted 8 lines
Quoted 5 lines
While it only covers libdw, I think the idea of this macro is whether
the arch has register mappings defined in DWARF standard. So I think
it's better to keep the name for this case.How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping, not the DWARF standard itself.
But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)Quoted 6 lines
Quoted 3 lines
The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
Quoted 13 lines
So you are saying the #define is guarding a libdw feature? perf record/report --call-graph=dwarf is supported with either libdw or libunwind. The dwarf support in the tool may come from more sources hence wanting in this patch set to be clear what variable is guarding what.
PERF_HAVE_DWARF_REGSis set to 1 for a specific set of architectures and only when libdw is present. The variable is saying that libdw supports the notion of registers needed for the #defineHAVE_DWARF_SUPPORTthat patch 9 in the series renamed toHAVE_LIBDW_SUPPORT. So I want the makefile variablePERF_HAVE_LIBDW_REGSto guard the #defineHAVE_LIBDW_SUPPORT, rather than what is being argued by yourself and Namhyung that the #defineHAVE_LIBDW_SUPPORTbe guarded by a variable calledPERF_HAVE_DWARF_REGSand that is only set whenNO_LIBDWisn't set.It will be only used with the libdw, but I don't care. "
HAVE_DWARF_REG" (internal config, just indicates the arch implemented feature) simply means there isarch/XXX/util/dwarf-regs.c. Also the APIs provided by thedwarf-regs.care still based on DWARF standard, which defines registers by number like DW_OP_reg[0-31]. So the mapping of these suffix number and actual register must be defined for each architecture.Quoted 3 lines
That is why I had introduced
dwarf-regs.cand call it "dwarf"-regs. Even if the implementation depends on libdw, thisdwarf-regs.cis still based on DWARF standard.You seem to be missing the point of the series which is to clean up inconsistencies where dwarf is used to mean libdw. Here we have libdw guarding a #define with DWARF in the name, it should have libdw in the name as the patch cleans up. This is a coding thing and not a dwarf specificatin thing.
Quoted 8 lines
Quoted 7 lines
We've made a digression into the name dwarf for a reason I can't fathom, at best it is inconsistent. Having dwarf registers is like having a bright sun or numeric numbers, it is a truism (playing devils advocate maybe if there were an ELF file format for postscript we could have a dwarf specification without registers). Anyway, I'm trying to connect the dots that libdw support controls the libdw type variables and defines hence not wanting 10 out of 11 patches applied.
Quoted 6 lines
Quoted 2 lines
Oh, wait, I think we can apply other patches. I just don't like this patch. I think the other patches are good. But this is
Then we are intentionally aiming to be inconsistent, with libdw meaning dwarf with a #define that just states a truism. Arguably the code is better with none of the series applied.
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.
Thanks,
Namhyung
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 138 lines · Namhyung Kim
Quoted 136 lines · Ian Rogers
Quoted 114 lines · Masami Hiramatsu
Quoted 88 lines · On Tue, 1 Oct 2024 18:31:43 -0700 Ian Rogers <[email protected]> wrote:
Quoted 80 lines · Masami Hiramatsu
Quoted 76 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 74 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
Quoted 17 lines · Namhyung Kim
On Thu, Sep 26, 2024 at 05:47:16AM -0700, Ian Rogers wrote:
On Wed, Sep 25, 2024 at 8:27 PM Namhyung Kim <[email protected]> wrote:On Tue, Sep 24, 2024 at 09:04:18AM -0700, Ian Rogers wrote:
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGS is only enabled with libdw support. Rename to
make it clearer there is a libdw connection.While it only covers libdw, I think the idea of this macro is whether
the arch has register mappings defined in DWARF standard. So I think
it's better to keep the name for this case.How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping,
not the DWARF standard itself.But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use for libunwind/libdw. Hence wanting to be specific that they are just a libdw and not a dwarf thing. Trying to use the regs in libunwind code would be broken. That could change but I wanted to make the code clear for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)Quoted 6 lines
Quoted 3 lines
The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
Quoted 13 lines
So you are saying the #define is guarding a libdw feature? perf record/report --call-graph=dwarf is supported with either libdw or libunwind. The dwarf support in the tool may come from more sources hence wanting in this patch set to be clear what variable is guarding what.
PERF_HAVE_DWARF_REGSis set to 1 for a specific set of architectures and only when libdw is present. The variable is saying that libdw supports the notion of registers needed for the #defineHAVE_DWARF_SUPPORTthat patch 9 in the series renamed toHAVE_LIBDW_SUPPORT. So I want the makefile variablePERF_HAVE_LIBDW_REGSto guard the #defineHAVE_LIBDW_SUPPORT, rather than what is being argued by yourself and Namhyung that the #defineHAVE_LIBDW_SUPPORTbe guarded by a variable calledPERF_HAVE_DWARF_REGSand that is only set whenNO_LIBDWisn't set.It will be only used with the libdw, but I don't care. "
HAVE_DWARF_REG" (internal config, just indicates the arch implemented feature) simply means there isarch/XXX/util/dwarf-regs.c. Also the APIs provided by thedwarf-regs.care still based on DWARF standard, which defines registers by number like DW_OP_reg[0-31]. So the mapping of these suffix number and actual register must be defined for each architecture.Quoted 3 lines
That is why I had introduced
dwarf-regs.cand call it "dwarf"-regs. Even if the implementation depends on libdw, thisdwarf-regs.cis still based on DWARF standard.You seem to be missing the point of the series which is to clean up inconsistencies where dwarf is used to mean libdw. Here we have libdw guarding a #define with DWARF in the name, it should have libdw in the name as the patch cleans up. This is a coding thing and not a dwarf specificatin thing.
Quoted 10 lines
Quoted 7 lines
We've made a digression into the name dwarf for a reason I can't fathom, at best it is inconsistent. Having dwarf registers is like having a bright sun or numeric numbers, it is a truism (playing devils advocate maybe if there were an ELF file format for postscript we could have a dwarf specification without registers). Anyway, I'm trying to connect the dots that libdw support controls the libdw type variables and defines hence not wanting 10 out of 11 patches applied.
Oh, wait, I think we can apply other patches. I just don't like this patch. I think the other patches are good. But this is
Quoted 7 lines
Quoted 3 lines
Then we are intentionally aiming to be inconsistent, with libdw meaning dwarf with a #define that just states a truism. Arguably the code is better with none of the series applied.
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.
So your objection is that we have files called:tools/perf/arch/*/util/dwarf-regs.c
and PERF_HAVE_DRWARF_REGS is an indication that this file exists. This file declares a single get_arch_regnum function. The building of the file after this series is:
perf-util-$(CONFIG_LIBDW) += dwarf-regs.o
My objection is that PERF_HAVE_DWARF_REGS is controlling the #define HAVE_LIBDW_SUPPORT, so dwarf (that can mean libunwind, libdw, etc.) is guarding having libdw which is backward and part of what this series has been trying to clean up.
If we rename tools/perf/arch/*/util/dwarf-regs.c to tools/perf/arch/*/util/libdw-helpers.c the PERF_HAVE_DWARF_REGS can be renamed to PERF_HAVE_LIBDW_HELPERS to align. Then PERF_HAVE_LIBDW_HELPERS guarding the #define PERF_HAVE_LIBDW makes sense to me and I think we achieve the filename alignment you are looking for.
Yes get_arch_regnum could make sense out of libdw and needn't just be a helper for it, but let's worry about that when there's a need. What's confusing at the moment is does libdw provide dwarf support, which I'd say is expected, or does dwarf provide libdw support?
Thanks,
Ian
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 11 lines · Ian Rogers
Quoted 3 lines · Namhyung Kim
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.So your objection is that we have files called:
tools/perf/arch/*/util/dwarf-regs.c
andPERF_HAVE_DRWARF_REGSis an indication that this file exists. This file declares a singleget_arch_regnumfunction. The building of the file after this series is:
perf-util-$(CONFIG_LIBDW) += dwarf-regs.o
Well.. I think we can even make it
perf-util-y += dwarf-regs.o
since it doesn't have any dependency on libdw. But it'd be inefficent to ship the dead code and data. Anyway we may remove the condition to define the PERF_HAVE_DWARF_REGS like below.
tools/perf/arch/x86/Makefile +0 -2
@@ -1,7 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 -endif HAVE_KVM_STAT_SUPPORT := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1 PERF_HAVE_JITDUMP := 1
Quoted 5 lines
My objection is that
PERF_HAVE_DWARF_REGSis controlling the #defineHAVE_LIBDW_SUPPORT, so dwarf (that can mean libunwind, libdw, etc.) is guarding having libdw which is backward and part of what this series has been trying to clean up.
Why not? If the arch doesn't define DWARF registers, it can refuse libdw support because it won't work well.
Quoted 7 lines
If we rename
tools/perf/arch/*/util/dwarf-regs.ctotools/perf/arch/*/util/libdw-helpers.cthePERF_HAVE_DWARF_REGScan be renamed toPERF_HAVE_LIBDW_HELPERSto align. ThenPERF_HAVE_LIBDW_HELPERSguarding the #definePERF_HAVE_LIBDWmakes sense to me and I think we achieve the filename alignment you are looking for.
I don't think it's a good idea. The logic is not specific to libdw.
Quoted 5 lines
Yes
get_arch_regnumcould make sense out of libdw and needn't just be a helper for it, but let's worry about that when there's a need. What's confusing at the moment is does libdw provide dwarf support, which I'd say is expected, or does dwarf provide libdw support?
As I said, it's about refusing libdw.
ifndef NO_LIBDW
ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined)
$(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled)
NO_LIBDW := 1
else
CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS)
LDFLAGS += $(LIBDW_LDFLAGS)
EXTLIBS += ${DWARFLIBS}
$(call detected,CONFIG_DWARF)
endif # PERF_HAVE_DWARF_REGS
endif # NO_LIBDWThanks,
Namhyung
Re: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 35 lines · On Thu, 3 Oct 2024 22:12:25 -0700 Namhyung Kim <[email protected]> wrote:
Quoted 11 lines · Ian Rogers
Quoted 3 lines · Namhyung Kim
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.So your objection is that we have files called:
tools/perf/arch/*/util/dwarf-regs.c
andPERF_HAVE_DRWARF_REGSis an indication that this file exists. This file declares a singleget_arch_regnumfunction. The building of the file after this series is:
perf-util-$(CONFIG_LIBDW) += dwarf-regs.oWell.. I think we can even make it
perf-util-y += dwarf-regs.o
since it doesn't have any dependency on libdw. But it'd be inefficent to ship the dead code and data. Anyway we may remove the condition to define the
PERF_HAVE_DWARF_REGSlike below.tools/perf/arch/x86/Makefile +0
-2@@ -1,7 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 -endif HAVE_KVM_STAT_SUPPORT := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1 PERF_HAVE_JITDUMP := 1Quoted 2 lines
Quoted 7 lines
Quoted 4 lines
My objection is that
PERF_HAVE_DWARF_REGSis controlling the #defineHAVE_LIBDW_SUPPORT, so dwarf (that can mean libunwind, libdw, etc.) is guarding having libdw which is backward and part of what this series has been trying to clean up.Why not? If the arch doesn't define DWARF registers, it can refuse libdw support because it won't work well.
It actually does not DWARF registers, but just "dwarf-regs.c" file since arch should define DWARF registers if the compiler generates the DWARF.
Here the flag means only "we implemented dwarf-regs.c file for this arch." So if it is called as "libdw-helper.c" then we can rename the flag as PERF_HAVE_ARCH_LIBDW_HELPER simply.
Quoted 8 lines
Quoted 6 lines
If we rename
tools/perf/arch/*/util/dwarf-regs.ctotools/perf/arch/*/util/libdw-helpers.cthePERF_HAVE_DWARF_REGScan be renamed toPERF_HAVE_LIBDW_HELPERSto align. ThenPERF_HAVE_LIBDW_HELPERSguarding the #definePERF_HAVE_LIBDWmakes sense to me and I think we achieve the filename alignment you are looking for.I don't think it's a good idea. The logic is not specific to libdw.
Yes, the logic (DWARF register mapping to the ISA register name) is not libdw. But I think we can also implement it in "libdw-helper.c". (In fact, this implementation does not depend only on Dwarf, but
rather on the convenience of ftrace.)
Quoted 7 lines
Quoted 5 lines
Yes
get_arch_regnumcould make sense out of libdw and needn't just be a helper for it, but let's worry about that when there's a need. What's confusing at the moment is does libdw provide dwarf support, which I'd say is expected, or does dwarf provide libdw support?As I said, it's about refusing libdw.
I think Ian pointed this part, even if libdw is available, dwarf-regs.c controls its usage, but libunwind is not.
Quoted 4 lines
ifndef NO_LIBDW
ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined)
$(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled)I think *this message* is the root cause of this discussion. It should be changed to
"DWARF register mappings have not been defined for architecture $(SRCARCH), libdw support disabled."
or (if changed to libdw-helper)
"libdw-helper.c is not implemented for architecture $(SRCARCH), libdw support disabled."
Thank you,
Quoted 12 lines
NO_LIBDW := 1 else CFLAGS += -DHAVE_DWARF_SUPPORT $(LIBDW_CFLAGS) LDFLAGS += $(LIBDW_LDFLAGS) EXTLIBS += ${DWARFLIBS} $(call detected,CONFIG_DWARF) endif # PERF_HAVE_DWARF_REGSendif #
NO_LIBDWThanks,
Namhyung
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Ian Rogers
reviewerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 84 lines · Masami Hiramatsu
Quoted 35 lines · On Thu, 3 Oct 2024 22:12:25 -0700 Namhyung Kim <[email protected]> wrote:
Quoted 11 lines · Ian Rogers
Quoted 3 lines · Namhyung Kim
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.So your objection is that we have files called:
tools/perf/arch/*/util/dwarf-regs.c
andPERF_HAVE_DRWARF_REGSis an indication that this file exists. This file declares a singleget_arch_regnumfunction. The building of the file after this series is:
perf-util-$(CONFIG_LIBDW) += dwarf-regs.oWell.. I think we can even make it
perf-util-y += dwarf-regs.o
since it doesn't have any dependency on libdw. But it'd be inefficent to ship the dead code and data. Anyway we may remove the condition to define the
PERF_HAVE_DWARF_REGSlike below.tools/perf/arch/x86/Makefile +0
-2@@ -1,7 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 -endif HAVE_KVM_STAT_SUPPORT := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1 PERF_HAVE_JITDUMP := 1Quoted 2 lines
Quoted 7 lines
Quoted 4 lines
My objection is that
PERF_HAVE_DWARF_REGSis controlling the #defineHAVE_LIBDW_SUPPORT, so dwarf (that can mean libunwind, libdw, etc.) is guarding having libdw which is backward and part of what this series has been trying to clean up.Why not? If the arch doesn't define DWARF registers, it can refuse libdw support because it won't work well.
It actually does not DWARF registers, but just "
dwarf-regs.c" file since arch should define DWARF registers if the compiler generates the DWARF.
Here the flag means only "we implementeddwarf-regs.cfile for this arch." So if it is called as "libdw-helper.c" then we can rename the flag asPERF_HAVE_ARCH_LIBDW_HELPERsimply.Quoted 8 lines
Quoted 6 lines
If we rename
tools/perf/arch/*/util/dwarf-regs.ctotools/perf/arch/*/util/libdw-helpers.cthePERF_HAVE_DWARF_REGScan be renamed toPERF_HAVE_LIBDW_HELPERSto align. ThenPERF_HAVE_LIBDW_HELPERSguarding the #definePERF_HAVE_LIBDWmakes sense to me and I think we achieve the filename alignment you are looking for.I don't think it's a good idea. The logic is not specific to libdw.
Yes, the logic (DWARF register mapping to the ISA register name) is not libdw. But I think we can also implement it in "
libdw-helper.c". (In fact, this implementation does not depend only on Dwarf, but
rather on the convenience of ftrace.)Quoted 7 lines
Quoted 5 lines
Yes
get_arch_regnumcould make sense out of libdw and needn't just be a helper for it, but let's worry about that when there's a need. What's confusing at the moment is does libdw provide dwarf support, which I'd say is expected, or does dwarf provide libdw support?As I said, it's about refusing libdw.
I think Ian pointed this part, even if libdw is available,
dwarf-regs.ccontrols its usage, but libunwind is not.Quoted 4 lines
ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled)
Quoted 8 lines
I think *this message* is the root cause of this discussion. It should be changed to
"DWARF register mappings have not been defined for architecture $(SRCARCH), libdw support disabled."
or (if changed to libdw-helper)
"
libdw-helper.cis not implemented for architecture $(SRCARCH), libdw support disabled."
So looking at the code I think the whole thing looks wrong. The get_arch_regnum function is used by get_dwarf_regnum which is used in 2 places in annotate.c:
```
static int extract_reg_offset(struct arch *arch, const char *str,
struct annotated_op_loc *op_loc)
...
int annotate_get_insn_location(struct arch *arch, struct disasm_line *dl,
struct annotated_insn_loc *loc)
```
So these functions are passing in an architecture. In get_dwarf_regnum:
```
/* Return DWARF register number from architecture register name */
int get_dwarf_regnum(const char *name, unsigned int machine)
{
char *regname = strdup(name);
int reg = -1;
char *p;
if (regname == NULL)
return -EINVAL;
/* For convenience, remove trailing characters */
p = strpbrk(regname, " ,)");
if (p)
*p = '\0';
switch (machine) {
case EM_NONE: /* Generic arch - use host arch */
reg = get_arch_regnum(regname);
break;
default:
pr_err("ELF MACHINE %x is not supported.\n", machine);
}
free(regname);
return reg;
}
```
But why, if the machine is EM_X86_64 and I'm on an x86-64, can't I
call get_arch_regnum? The code should be something like:
```
if (machine == EM_NONE) {
#ifdef __x86_64__
machine = EM_X86_64;
#elf...
```
Once we have an architecture specific machine then instead of
get_arch_regnum it should call get_x86_64_regnum or
get_aarch64_regnum.
```
switch(machine) }
case EM_X86_64:
reg = get_x86_64_regnum(regname);
break;
...
```Is this better? Yes, it means that the annotate logic can work if, say, annotating/disassembling an ARM binary on an x86-64.
So we need to pull all the tools/perf/arch/<arch>/util/dwarf-regs.c files into tools/perf/util/dwarf-regs-<arch>.c files. We need to rename the get_arch_regnum to reflect the <arch> in the file name. The Makefile logic can include all of this unconditionally and PERF_HAVE_DWARF_REGS can just be removed. In the process the ability to annotate binaries from one architecture on another is improved. It needn't be the case that we have dwarf regs for the architecture perf is being run on as we may be annotating an x86-64 binary where there is support.
What's strange is that get_dwarf_regstr in the common code already does things pretty much this way. This whole Makefile, arch, weak function, PERF_HAVE... logic just looks like a mistake that is making the tool worse than it needs to be. I think this is frequently the case with code in arch/, a lot of the functionality there can be moved into pmu.c and doing things conditional on the pmu, which is inherently architecture dependent. This can fix unusual cases of say running the perf tool on user land qemu, where we may have an ARM perf binary but see the PMUs of an x86.
I can work to put this into a v2 so please scream if my reasoning doesn't make sense.
Thanks,
Ian
Namhyung Kim
maintainerRe: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
Quoted 173 lines · Ian Rogers
Quoted 84 lines · Masami Hiramatsu
Quoted 35 lines · On Thu, 3 Oct 2024 22:12:25 -0700 Namhyung Kim <[email protected]> wrote:
Quoted 11 lines · Ian Rogers
Quoted 3 lines · Namhyung Kim
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.So your objection is that we have files called:
tools/perf/arch/*/util/dwarf-regs.c
andPERF_HAVE_DRWARF_REGSis an indication that this file exists. This file declares a singleget_arch_regnumfunction. The building of the file after this series is:
perf-util-$(CONFIG_LIBDW) += dwarf-regs.oWell.. I think we can even make it
perf-util-y += dwarf-regs.o
since it doesn't have any dependency on libdw. But it'd be inefficent to ship the dead code and data. Anyway we may remove the condition to define the
PERF_HAVE_DWARF_REGSlike below.tools/perf/arch/x86/Makefile +0
-2@@ -1,7 +1,5 @@ # SPDX-License-Identifier: GPL-2.0 -ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 -endif HAVE_KVM_STAT_SUPPORT := 1 PERF_HAVE_ARCH_REGS_QUERY_REGISTER_OFFSET := 1 PERF_HAVE_JITDUMP := 1Quoted 2 lines
Quoted 7 lines
Quoted 4 lines
My objection is that
PERF_HAVE_DWARF_REGSis controlling the #defineHAVE_LIBDW_SUPPORT, so dwarf (that can mean libunwind, libdw, etc.) is guarding having libdw which is backward and part of what this series has been trying to clean up.Why not? If the arch doesn't define DWARF registers, it can refuse libdw support because it won't work well.
It actually does not DWARF registers, but just "
dwarf-regs.c" file since arch should define DWARF registers if the compiler generates the DWARF.
Here the flag means only "we implementeddwarf-regs.cfile for this arch." So if it is called as "libdw-helper.c" then we can rename the flag asPERF_HAVE_ARCH_LIBDW_HELPERsimply.Quoted 8 lines
Quoted 6 lines
If we rename
tools/perf/arch/*/util/dwarf-regs.ctotools/perf/arch/*/util/libdw-helpers.cthePERF_HAVE_DWARF_REGScan be renamed toPERF_HAVE_LIBDW_HELPERSto align. ThenPERF_HAVE_LIBDW_HELPERSguarding the #definePERF_HAVE_LIBDWmakes sense to me and I think we achieve the filename alignment you are looking for.I don't think it's a good idea. The logic is not specific to libdw.
Yes, the logic (DWARF register mapping to the ISA register name) is not libdw. But I think we can also implement it in "
libdw-helper.c". (In fact, this implementation does not depend only on Dwarf, but
rather on the convenience of ftrace.)Quoted 7 lines
Quoted 5 lines
Yes
get_arch_regnumcould make sense out of libdw and needn't just be a helper for it, but let's worry about that when there's a need. What's confusing at the moment is does libdw provide dwarf support, which I'd say is expected, or does dwarf provide libdw support?As I said, it's about refusing libdw.
I think Ian pointed this part, even if libdw is available,
dwarf-regs.ccontrols its usage, but libunwind is not.Quoted 4 lines
ifndef NO_LIBDW ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) $(warning DWARF register mappings have not been defined for architecture $(SRCARCH), DWARF support disabled)Quoted 8 lines
I think *this message* is the root cause of this discussion. It should be changed to
"DWARF register mappings have not been defined for architecture $(SRCARCH), libdw support disabled."
or (if changed to libdw-helper)
"
libdw-helper.cis not implemented for architecture $(SRCARCH), libdw support disabled."So looking at the code I think the whole thing looks wrong. The
get_arch_regnumfunction is used byget_dwarf_regnumwhich is used in 2 places inannotate.c:``` static int extract_reg_offset(struct arch *arch, const char *str, struct annotated_op_loc *op_loc) ... int annotate_get_insn_location(struct arch *arch, struct disasm_line *dl, struct annotated_insn_loc *loc) ``` So these functions are passing in an architecture. In get_dwarf_regnum: ``` /* Return DWARF register number from architecture register name */ int get_dwarf_regnum(const char *name, unsigned int machine) { char *regname = strdup(name); int reg = -1; char *p; if (regname == NULL) return -EINVAL; /* For convenience, remove trailing characters */ p = strpbrk(regname, " ,)"); if (p) *p = '\0'; switch (machine) { case EM_NONE: /* Generic arch - use host arch */ reg = get_arch_regnum(regname); break; default: pr_err("ELF MACHINE %x is not supported.\n", machine); } free(regname); return reg; } ``` But why, if the machine is EM_X86_64 and I'm on an x86-64, can't I call get_arch_regnum? The code should be something like: ``` if (machine == EM_NONE) { #ifdef __x86_64__ machine = EM_X86_64; #elf... ``` Once we have an architecture specific machine then instead of get_arch_regnum it should call get_x86_64_regnum or get_aarch64_regnum. ``` switch(machine) } case EM_X86_64: reg = get_x86_64_regnum(regname); break; ... ```Is this better? Yes, it means that the annotate logic can work if, say, annotating/disassembling an ARM binary on an x86-64.
So we need to pull all the tools/perf/arch/<arch>/util/dwarf-regs.c files into tools/perf/util/dwarf-regs-<arch>.c files. We need to rename the
get_arch_regnumto reflect the <arch> in the file name. TheMakefilelogic can include all of this unconditionally andPERF_HAVE_DWARF_REGScan just be removed. In the process the ability to annotate binaries from one architecture on another is improved. It needn't be the case that we have dwarf regs for the architecture perf is being run on as we may be annotating an x86-64 binary where there is support.What's strange is that
get_dwarf_regstrin the common code already does things pretty much this way. This wholeMakefile, arch, weak function,PERF_HAVE... logic just looks like a mistake that is making the tool worse than it needs to be. I think this is frequently the case with code inarch/, a lot of the functionality there can be moved intopmu.cand doing things conditional on the pmu, which is inherently architecture dependent. This can fix unusual cases of say running the perf tool on user land qemu, where we may have an ARM perf binary but see the PMUs of an x86.
Quoted 2 lines
I can work to put this into a v2 so please scream if my reasoning doesn't make sense.
Sounds good, it'd be easier if we merge patch 1-10 first and you would work on the register thing separately.
Thanks,
Namhyung
Re: [PATCH v1 11/11] perf build: Rename PERF_HAVE_DWARF_REGS to PERF_HAVE_LIBDW_REGS
In reply to Ian RogersQuoted 154 lines · On Thu, 3 Oct 2024 17:58:13 -0700 Ian Rogers <[email protected]> wrote:
Quoted 138 lines · Namhyung Kim
Quoted 136 lines · Ian Rogers
Quoted 114 lines · Masami Hiramatsu
Quoted 88 lines · On Tue, 1 Oct 2024 18:31:43 -0700 Ian Rogers <[email protected]> wrote:
Quoted 80 lines · Masami Hiramatsu
Quoted 76 lines · On Tue, 1 Oct 2024 16:17:34 -0700 Ian Rogers <[email protected]> wrote:
Quoted 74 lines · Namhyung Kim
Quoted 63 lines · Ian Rogers
Quoted 43 lines · Masami Hiramatsu
Quoted 30 lines · On Fri, 27 Sep 2024 11:15:21 -0700 Ian Rogers <[email protected]> wrote:
Quoted 29 lines · Namhyung Kim
Quoted 26 lines · Ian Rogers
On Thu, Sep 26, 2024 at 12:40 PM Namhyung Kim <[email protected]> wrote:
On Thu, Sep 26, 2024 at 05:47:16AM -0700, Ian Rogers wrote:
On Wed, Sep 25, 2024 at 8:27 PM Namhyung Kim <[email protected]> wrote:On Tue, Sep 24, 2024 at 09:04:18AM -0700, Ian Rogers wrote:
The name dwarf can imply libunwind support, whereas
PERF_HAVE_DWARF_REGS is only enabled with libdw support. Rename to
make it clearer there is a libdw connection.While it only covers libdw, I think the idea of this macro is whether
the arch has register mappings defined in DWARF standard. So I think
it's better to keep the name for this case.How can the dwarf standard exist for an arch but not define registers?
I meant it's about the arch code in the perf tools to have the mapping,
not the DWARF standard itself.But we guard those definitions behind having libdw:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/Makefile?h=perf-tools-next#n3
So we only have the regs if libdw is present, not if dwarf is in use
for libunwind/libdw. Hence wanting to be specific that they are just a
libdw and not a dwarf thing. Trying to use the regs in libunwind code
would be broken. That could change but I wanted to make the code clear
for the way things are at the moment.Quoted 8 lines
Quoted 1 line
I understand your point but calling it
LIBDW_REGSlooks unnatural to me.I don't follow. Wouldn't it be unnatural to see
PERF_HAVE_DWARF_REGSin libunwind code but you are to some how know that the code only had meaning if libdw was present? I don't like the implication that DWARF means LIBDW as throughout the code it doesn't. I think the namePERF_HAVE_LIBDW_REGSbetter captures how the code is, makes the code more intention revealing and so readable, etc.Quoted 3 lines
I agree with Namhyung this point. dwarf-regs is defined only by the DWARF standard, not libdw only. The standard encode registers by a digit number and the dwarf-regs decode the number to actual register name.
The code is not making a statement about the DWARF standard, take
arch/csky:
https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/csky/Makefile?h=perf-tools-next``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_DWARF PERF_HAVE_DWARF_REGS := 1 endif ``` in the patch series NO_DWARF becomes NO_LIBDW, so it is now: ``` # SPDX-License-Identifier: GPL-2.0-only ifndef NO_LIBDW PERF_HAVE_DWARF_REGS := 1 endif ```Quoted 3 lines
So the
Makefilesays thatPERF_HAVE_DWARF_REGSis dependent on havingNO_LIBDW, that is having libdw impliesPERF_HAVE_DWARF_REGSis defined for csky.I think this is totally fine and we can change the condition later if needed.
After all, I don't think it's a big deal. Let's just call DWARF registers
DWARF_REGS. :)Quoted 6 lines
Quoted 3 lines
The define is called
PERF_HAVE_DWARF_REGS, notice the HAVE, but we're not setting it while supporting call-graph dwarf with libunwind. It is actively confusing.Does libunwind requires the dwarf regs? I think the dwarf regs information is only needed for analyzing dwarf register assignment, not stack unwinding.
Quoted 13 lines
So you are saying the #define is guarding a libdw feature? perf record/report --call-graph=dwarf is supported with either libdw or libunwind. The dwarf support in the tool may come from more sources hence wanting in this patch set to be clear what variable is guarding what.
PERF_HAVE_DWARF_REGSis set to 1 for a specific set of architectures and only when libdw is present. The variable is saying that libdw supports the notion of registers needed for the #defineHAVE_DWARF_SUPPORTthat patch 9 in the series renamed toHAVE_LIBDW_SUPPORT. So I want the makefile variablePERF_HAVE_LIBDW_REGSto guard the #defineHAVE_LIBDW_SUPPORT, rather than what is being argued by yourself and Namhyung that the #defineHAVE_LIBDW_SUPPORTbe guarded by a variable calledPERF_HAVE_DWARF_REGSand that is only set whenNO_LIBDWisn't set.It will be only used with the libdw, but I don't care. "
HAVE_DWARF_REG" (internal config, just indicates the arch implemented feature) simply means there isarch/XXX/util/dwarf-regs.c. Also the APIs provided by thedwarf-regs.care still based on DWARF standard, which defines registers by number like DW_OP_reg[0-31]. So the mapping of these suffix number and actual register must be defined for each architecture.Quoted 3 lines
That is why I had introduced
dwarf-regs.cand call it "dwarf"-regs. Even if the implementation depends on libdw, thisdwarf-regs.cis still based on DWARF standard.You seem to be missing the point of the series which is to clean up inconsistencies where dwarf is used to mean libdw. Here we have libdw guarding a #define with DWARF in the name, it should have libdw in the name as the patch cleans up. This is a coding thing and not a dwarf specificatin thing.
Quoted 10 lines
Quoted 7 lines
We've made a digression into the name dwarf for a reason I can't fathom, at best it is inconsistent. Having dwarf registers is like having a bright sun or numeric numbers, it is a truism (playing devils advocate maybe if there were an ELF file format for postscript we could have a dwarf specification without registers). Anyway, I'm trying to connect the dots that libdw support controls the libdw type variables and defines hence not wanting 10 out of 11 patches applied.
Oh, wait, I think we can apply other patches. I just don't like this patch. I think the other patches are good. But this is
Quoted 7 lines
Quoted 3 lines
Then we are intentionally aiming to be inconsistent, with libdw meaning dwarf with a #define that just states a truism. Arguably the code is better with none of the series applied.
I agree renaming libdw-specific parts. But the register is for DWARF,
not libdw even if it's currently used by libdw only. So I don't want
to rename it.So your objection is that we have files called:
tools/perf/arch/*/util/dwarf-regs.c
andPERF_HAVE_DRWARF_REGSis an indication that this file exists. This file declares a singleget_arch_regnumfunction. The building of the file after this series is:
perf-util-$(CONFIG_LIBDW) += dwarf-regs.o
Quoted 4 lines
My objection is that
PERF_HAVE_DWARF_REGSis controlling the #defineHAVE_LIBDW_SUPPORT, so dwarf (that can mean libunwind, libdw, etc.) is guarding having libdw which is backward and part of what this series has been trying to clean up.
OK.
Quoted 7 lines
If we rename
tools/perf/arch/*/util/dwarf-regs.ctotools/perf/arch/*/util/libdw-helpers.cthePERF_HAVE_DWARF_REGScan be renamed toPERF_HAVE_LIBDW_HELPERSto align. ThenPERF_HAVE_LIBDW_HELPERSguarding the #definePERF_HAVE_LIBDWmakes sense to me and I think we achieve the filename alignment you are looking for.
Yeah, I think that is OK to me.
Quoted 4 lines
Yes
get_arch_regnumcould make sense out of libdw and needn't just be a helper for it, but let's worry about that when there's a need. What's confusing at the moment is does libdw provide dwarf support, which I'd say is expected, or does dwarf provide libdw support?
You missed to cut the word, PERF_HAVE_DWARF_REGS means "perf have 'dwarf-regs.c'". If dwarf-regs.c is not there, we can not enable libdw support because it causes linker error.
So, if the file name simply changed to libdw-helper.c, I think we can change it.
Thank you,
Quoted 4 lines
Thanks,
Ian
Signature · 1 line
Masami Hiramatsu (Google) <[email protected]>
Quoted 10 lines · Ian Rogers
The patches remove the conditional compilation assuming the features are in libdw where the feature test is expanded to check there pressence.
In the
Makefilecode dwarf tends to mean unwind or libdw support for dwarf things. To make it clearer when dwarf really just means libdw numerous build variables and defines are renamed.There is some tech debt in the changes as perf check still reports the values using the old name and for features that are no longer tested.
I built for Arm64 target with static and dynamic linkage, both can build successfully in Ubuntu 24.04 and Debian bookworm. FWIW: