stable v5.15, v5.10 and v5.4: fix perf build errors
Subsystems (3)
cover stable v5.15, v5.10 and v5.4: fix perf build errors
Hi,
I would like to see these patches backported. They are needed so perf
can be cross compiled with gcc on v5.15, v5.10 and v5.4.
I built it with tuxmake [1] here are two example commandlines:
tuxmake --runtime podman --target-arch arm64 --toolchain gcc-12 --kconfig defconfig perf tuxmake --runtime podman --target-arch x86_64 --toolchain gcc-12 --kconfig defconfig perf
Tried to build perf with both gcc-11 and gcc-12.
Patch 'tools perf: Fix compilation error with new binutils' and 'tools build: Add feature test for init_disassemble_info API changes' didn't apply cleanly, thats why I send these in a patchset.
When apply 'tools build: Add feature test for init_disassemble_info API changes' to 5.4 it will be a minor merge conflict, do you want me to send this patch in two separate patches one for 5.4 and another for v5.10?
The sha for these two patches in mainline are.cfd59ca91467 tools build: Add feature test for init_disassemble_info API changes 83aa0120487e tools perf: Fix compilation error with new binutils
The above patches solves these:util/annotate.c: In function 'symbol__disassemble_bpf':util/annotate.c:1729:9: error: too few arguments to function 'init_disassemble_info'
1729 | init_disassemble_info(&info, s,
| ^~~~~~~~~~~~~~~~~~~~~Please apply these to v5.10 and v5.4a45b3d692623 tools include: add dis-asm-compat.h to handle version differences d08c84e01afa perf sched: Cast PTHREAD_STACK_MIN to int as it may turn into sysconf(__SC_THREAD_STACK>
The above patches solves these:/home/anders/src/kernel/stable-5.10/tools/include/linux/kernel.h:43:24: error: comparison of distinct pointer types lacks a cast [-Werror]
43 | (void) (&_max1 == &_max2); \
| ^~
builtin-sched.c:673:34: note: in expansion of macro 'max'
673 | (size_t) max(16 * 1024, PTHREAD_STACK_MIN));
| ^~~
Please apply these to v5.15, v5.10 and v5.48e8bf60a6754 perf build: Fixup disabling of -Wdeprecated-declarations for the python scripting engine 4ee3c4da8b1b perf scripting python: Do not build fail on deprecation warnings 63a4354ae75c perf scripting perl: Ignore some warnings to keep building with perl headers
Build error that the above 3 patches solves are:/usr/lib/x86_64-linux-gnu/perl/5.36/CORE/handy.h:125:23: error: cast from function call of type 'STRLEN' {aka 'long unsigned int'} to non-matching type '_Bool' [-Werror=bad-function-cast]
125 | #define cBOOL(cbool) ((bool) (cbool))
| ^
Cheers,
Anders
[1] https://tuxmake.org/
Andres Freund (2):
tools perf: Fix compilation error with new binutils tools build: Add feature test for init_disassemble_info API changes
tools/build/Makefile.feature | 1 + tools/build/feature/Makefile | 4 ++++ tools/build/feature/test-all.c | 4 ++++ tools/build/feature/test-disassembler-init-styled.c | 13 +++++++++++++ tools/perf/Makefile.config | 8 ++++++++ tools/perf/util/annotate.c | 7 ++++--- 6 files changed, 34 insertions(+), 3 deletions(-) create mode 100644 tools/build/feature/test-disassembler-init-styled.c
--
2.39.2
PATCH 1/2 tools perf: Fix compilation error with new binutils
From: Andres Freund <[email protected]>
binutils changed the signature of init_disassemble_info(), which now causes compilation failures for tools/perf/util/annotate.c, e.g. on debian unstable.
Relevant binutils commit:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=60a3da00bd5407f07
Wire up the feature test and switch to init_disassemble_info_compat(), which were introduced in prior commits, fixing the compilation failure.
I verified that perf can still disassemble bpf programs by using bpftrace under load, recording a perf trace, and then annotating the bpf "function" with and without the changes. With old binutils there's no change in output before/after this patch. When comparing the output from old binutils (2.35) to new bintuils with the patch (upstream snapshot) there are a few output differences, but they are unrelated to this patch. An example hunk is:
1.15 : 55:mov %rbp,%rdx
0.00 : 58:add $0xfffffffffffffff8,%rdx
0.00 : 5c:xor %ecx,%ecx
- 1.03 : 5e:callq 0xffffffffe12aca3c
+ 1.03 : 5e:call 0xffffffffe12aca3c
0.00 : 63:xor %eax,%eax
- 2.18 : 65:leaveq
- 2.82 : 66:retq
+ 2.18 : 65:leave
+ 2.82 : 66:ret--- tools/perf/Makefile.config | 8 ++++++++ tools/perf/util/annotate.c | 7 ++++--- 2 files changed, 12 insertions(+), 3 deletions(-)
tools/perf/Makefile.config +8 -0
@@ -281,6 +281,7 @@ FEATURE_CHECK_LDFLAGS-libpython := $(PYTHON_EMBED_LDOPTS) FEATURE_CHECK_LDFLAGS-libaio = -lrt FEATURE_CHECK_LDFLAGS-disassembler-four-args = -lbfd -lopcodes -ldl +FEATURE_CHECK_LDFLAGS-disassembler-init-styled = -lbfd -lopcodes -ldl CORE_CFLAGS += -fno-omit-frame-pointer CORE_CFLAGS += -ggdb3
@@ -838,13 +839,16 @@ else ifeq ($(feature-libbfd-liberty), 1) EXTLIBS += -lbfd -lopcodes -liberty FEATURE_CHECK_LDFLAGS-disassembler-four-args += -liberty -ldl + FEATURE_CHECK_LDFLAGS-disassembler-init-styled += -liberty -ldl else ifeq ($(feature-libbfd-liberty-z), 1) EXTLIBS += -lbfd -lopcodes -liberty -lz FEATURE_CHECK_LDFLAGS-disassembler-four-args += -liberty -lz -ldl + FEATURE_CHECK_LDFLAGS-disassembler-init-styled += -liberty -ldl endif endif $(call feature_check,disassembler-four-args) + $(call feature_check,disassembler-init-styled) endif ifeq ($(feature-libbfd-buildid), 1)
@@ -957,6 +961,10 @@ ifeq ($(feature-disassembler-four-args), 1) CFLAGS += -DDISASM_FOUR_ARGS_SIGNATURE endif +ifeq ($(feature-disassembler-init-styled), 1) + CFLAGS += -DDISASM_INIT_STYLED +endif + ifeq (${IS_64_BIT}, 1) ifndef NO_PERF_READ_VDSO32 $(call feature_check,compile-32)
tools/perf/util/annotate.c +4 -3
@@ -1684,6 +1684,7 @@ static int dso__disassemble_filename(struct dso *dso, char *filename, size_t fil#define PACKAGE "perf"#include <bfd.h> #include <dis-asm.h> +#include <tools/dis-asm-compat.h> static int symbol__disassemble_bpf (struct symbol *sym,struct annotate_args *args)
@@ -1726,9 +1727,9 @@ static int symbol__disassemble_bpf(struct symbol *sym, ret = errno; goto out; } -init_disassemble_info (&info, s, - (fprintf_ftype ) fprintf); - +init_disassemble_info_compat (&info, s, + (fprintf_ftype ) fprintf, + fprintf_styled); info.arch =bfd_get_arch (bfdf); info.mach =bfd_get_mach (bfdf);
Signature · 1 line
2.39.2
kernel test robot
build botRe: [backport PATCH 1/2] tools perf: Fix compilation error with new binutils
FYI: kernel test robot notices the stable kernel rule is not satisfied.
Hi,
Thanks for your patch.
FYI: kernel test robot notices the stable kernel rule is not satisfied.
Rule: 'Cc: [email protected]' or 'commit <sha1> upstream.'
Subject: [backport PATCH 1/2] tools perf: Fix compilation error with new binutils
The check is based on https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests
Ian Rogers
reviewerRe: [backport PATCH 1/2] tools perf: Fix compilation error with new binutils
In reply to Anders RoxellQuoted 6 lines · Anders Roxell
From: Andres Freund <[email protected]>
binutils changed the signature of
init_disassemble_info(), which now causes compilation failures fortools/perf/util/annotate.c, e.g. on debian unstable.
Thanks, I believe the compilation issue may well be resolved by:
https://lore.kernel.org/lkml/[email protected]/
where binutils is made opt-in rather than opt-out.
Quoted 6 lines
Relevant binutils commit:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=60a3da00bd5407f07
Wire up the feature test and switch to
init_disassemble_info_compat(), which were introduced in prior commits, fixing the compilation failure.
I was kind of surprised to see no version check ifdef. Is init_disassemble_info_compat is supported in older binutils?
Thanks,
Ian
Quoted 102 lines
I verified that perf can still disassemble bpf programs by using bpftrace under load, recording a perf trace, and then annotating the bpf "function" with and without the changes. With old binutils there's no change in output before/after this patch. When comparing the output from old binutils (2.35) to new bintuils with the patch (upstream snapshot) there are a few output differences, but they are unrelated to this patch. An example hunk is:
1.15 : 55:mov %rbp,%rdx 0.00 : 58:add $0xfffffffffffffff8,%rdx 0.00 : 5c:xor %ecx,%ecx - 1.03 : 5e:callq 0xffffffffe12aca3c + 1.03 : 5e:call 0xffffffffe12aca3c 0.00 : 63:xor %eax,%eax - 2.18 : 65:leaveq - 2.82 : 66:retq + 2.18 : 65:leave + 2.82 : 66:retSigned-off-by Andres Freund <[email protected]>Acked-by Quentin Monnet <[email protected]>Cc Alexei Starovoitov <[email protected]>Cc Ben Hutchings <[email protected]>Cc Jiri Olsa <[email protected]>Cc Sedat Dilek <[email protected]>Signed-off-by Arnaldo Carvalho de Melo <[email protected]>Signed-off-by Anders Roxell <[email protected]>--- tools/perf/Makefile.config | 8 ++++++++ tools/perf/util/annotate.c | 7 ++++--- 2 files changed, 12 insertions(+), 3 deletions(-)tools/perf/Makefile.config +8
-0@@ -281,6 +281,7 @@ FEATURE_CHECK_LDFLAGS-libpython := $(PYTHON_EMBED_LDOPTS) FEATURE_CHECK_LDFLAGS-libaio = -lrt FEATURE_CHECK_LDFLAGS-disassembler-four-args = -lbfd -lopcodes -ldl +FEATURE_CHECK_LDFLAGS-disassembler-init-styled = -lbfd -lopcodes -ldl CORE_CFLAGS += -fno-omit-frame-pointer CORE_CFLAGS += -ggdb3@@ -838,13 +839,16 @@ else ifeq ($(feature-libbfd-liberty), 1) EXTLIBS += -lbfd -lopcodes -liberty FEATURE_CHECK_LDFLAGS-disassembler-four-args += -liberty -ldl + FEATURE_CHECK_LDFLAGS-disassembler-init-styled += -liberty -ldl else ifeq ($(feature-libbfd-liberty-z), 1) EXTLIBS += -lbfd -lopcodes -liberty -lz FEATURE_CHECK_LDFLAGS-disassembler-four-args += -liberty -lz -ldl + FEATURE_CHECK_LDFLAGS-disassembler-init-styled += -liberty -ldl endif endif $(call feature_check,disassembler-four-args) + $(call feature_check,disassembler-init-styled) endif ifeq ($(feature-libbfd-buildid), 1)@@ -957,6 +961,10 @@ ifeq ($(feature-disassembler-four-args), 1) CFLAGS += -DDISASM_FOUR_ARGS_SIGNATURE endif +ifeq ($(feature-disassembler-init-styled), 1) + CFLAGS += -DDISASM_INIT_STYLED +endif + ifeq (${IS_64_BIT}, 1) ifndef NO_PERF_READ_VDSO32 $(call feature_check,compile-32)tools/perf/util/annotate.c +4
-4@@ -1684,6 +1684,7 @@ static int dso__disassemble_filename(struct dso *dso, char *filename, size_t fil#define PACKAGE "perf"#include <bfd.h> #include <dis-asm.h> +#include <tools/dis-asm-compat.h> static int symbol__disassemble_bpf (struct symbol *sym,struct annotate_args *args)@@ -1726,9 +1727,9 @@ static int symbol__disassemble_bpf(struct symbol *sym, ret = errno; goto out; } -init_disassemble_info (&info, s, - (fprintf_ftype ) fprintf); - +init_disassemble_info_compat (&info, s, + (fprintf_ftype ) fprintf, + fprintf_styled); info.arch =bfd_get_arch (bfdf); info.mach =bfd_get_mach (bfdf); --2.39.2
Re: [backport PATCH 1/2] tools perf: Fix compilation error with new binutils
2023-04-17 10:14 UTC-0700 ~ Ian Rogers <[email protected]>
Quoted 11 lines · Anders Roxell
Quoted 6 lines
From: Andres Freund <[email protected]>
binutils changed the signature of
init_disassemble_info(), which now causes compilation failures fortools/perf/util/annotate.c, e.g. on debian unstable.Thanks, I believe the compilation issue may well be resolved by:
https://lore.kernel.org/lkml/[email protected]/
where binutils is made opt-in rather than opt-out.
Hi,
These commits are to make it possible to build against recent binutils, without having to leave it out at compile time, so as I understand they address a different issue?
Quoted 10 lines
Quoted 7 lines
Relevant binutils commit:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=60a3da00bd5407f07
Wire up the feature test and switch to
init_disassemble_info_compat(), which were introduced in prior commits, fixing the compilation failure.I was kind of surprised to see no version check ifdef. Is
init_disassemble_info_compatis supported in older binutils?
It is not part of binutils, it was introduced in commit a45b3d692623 ("tools include: add dis-asm-compat.h to handle version differences"), which should likely be backported alongside these ones if it hasn't been already. Possibly the others from the same series [0], as well?
I think all 5 patches from Andres' series were backported to 5.15 [1].
[0]
https://lore.kernel.org/all/[email protected]/t/#m999a44663894e235b523ffc41ce87e956019ea72
[1]
https://lore.kernel.org/all/[email protected]/t/#u
Best regards,
Quentin
Ian Rogers
reviewerRe: [backport PATCH 1/2] tools perf: Fix compilation error with new binutils
Quoted 10 lines · Quentin Monnet
2023-04-17 10:14 UTC-0700 ~ Ian Rogers <[email protected]>
Quoted 8 lines · Anders Roxell
Quoted 6 lines
From: Andres Freund <[email protected]>
binutils changed the signature of
init_disassemble_info(), which now causes compilation failures fortools/perf/util/annotate.c, e.g. on debian unstable.
Quoted 8 lines
Quoted 3 lines
Thanks, I believe the compilation issue may well be resolved by:
https://lore.kernel.org/lkml/[email protected]/
where binutils is made opt-in rather than opt-out.Hi,
These commits are to make it possible to build against recent binutils, without having to leave it out at compile time, so as I understand they address a different issue?
Kind of. We don't want the Linux perf build to break. Previously if binutils were installed then Linux perf would default to linking with it and break your build were binutils to change its API. That is no longer the case as we don't default to linking against binutils. This was motivated by distributions not being able to link Linux perf with binutils due to the license incompatibilities. I don't see a problem supporting linking against newer and older binutils if people want to on non-distributed binaries. We'll probably need more build tests/containers to cover the possibilities of this. I'm not sure what's motivating binutils support other than personal experimentation though.
Thanks,
Ian
Quoted 25 lines
Quoted 10 lines
Quoted 7 lines
Relevant binutils commit:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=60a3da00bd5407f07
Wire up the feature test and switch to
init_disassemble_info_compat(), which were introduced in prior commits, fixing the compilation failure.I was kind of surprised to see no version check ifdef. Is
init_disassemble_info_compatis supported in older binutils?It is not part of binutils, it was introduced in commit
a45b3d692623("tools include: adddis-asm-compat.hto handle version differences"), which should likely be backported alongside these ones if it hasn't been already. Possibly the others from the same series [0], as well?I think all 5 patches from Andres' series were backported to 5.15 [1].
[0]
https://lore.kernel.org/all/[email protected]/t/#m999a44663894e235b523ffc41ce87e956019ea72
[1]
https://lore.kernel.org/all/[email protected]/t/#uBest regards,
Quentin
PATCH 2/2 tools build: Add feature test for init_disassemble_info API changes
In reply to Anders RoxellFrom: Andres Freund <[email protected]>
binutils changed the signature of init_disassemble_info(), which now causes compilation failures for tools/{perf,bpf}, e.g. on debian unstable.
Relevant binutils commit:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=60a3da00bd5407f07
This commit adds a feature test to detect the new signature. Subsequent commits will use it to fix the build failures.
--- tools/build/Makefile.feature | 1 + tools/build/feature/Makefile | 4 ++++ tools/build/feature/test-all.c | 4 ++++ tools/build/feature/test-disassembler-init-styled.c | 13 +++++++++++++ 4 files changed, 22 insertions(+) create mode 100644 tools/build/feature/test-disassembler-init-styled.c
tools/build/Makefile.feature +1 -0
@@ -69,6 +69,7 @@ FEATURE_TESTS_BASIC := \ libaio \ libzstd \ disassembler-four-args \ + disassembler-init-styled \ file-handle # FEATURE_TESTS_BASIC + FEATURE_TESTS_EXTRA is the complete list
tools/build/feature/Makefile +4 -0
@@ -17,6 +17,7 @@ FILES= \ test-libbfd.bin \ test-libbfd-buildid.bin \ test-disassembler-four-args.bin \ + test-disassembler-init-styled.bin \ test-reallocarray.bin \ test-libbfd-liberty.bin \ test-libbfd-liberty-z.bin \
@@ -235,6 +236,9 @@ $(OUTPUT)test-libbfd-buildid.bin: $(OUTPUT)test-disassembler-four-args.bin: $(BUILD) -DPACKAGE='"perf"' -lbfd -lopcodes +$(OUTPUT)test-disassembler-init-styled.bin: + $(BUILD) -DPACKAGE='"perf"' -lbfd -lopcodes + $(OUTPUT)test-reallocarray.bin: $(BUILD)
tools/build/feature/test-all.c +4 -0
@@ -170,6 +170,10 @@# include "test-disassembler-four-args.c" #undef main +#define main main_test_disassembler_init_styled +# include "test-disassembler-init-styled.c" +#undef main +#define main main_test_libzstd# include "test-libzstd.c" #undef main
tools/build/feature/test-disassembler-init-styled.c new file +13 -0
@@ -0,0 +1,13 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <stdio.h> +#include <dis-asm.h> + +int main (void ) +{ +struct disassemble_info info; + +init_disassemble_info (&info, stdout, +NULL ,NULL ); + +return 0 ; +}
Signature · 1 line
2.39.2
Ian Rogers
reviewerQuoted 5 lines · Anders Roxell
Hi,
I would like to see these patches backported. They are needed so perf can be cross compiled with gcc on v5.15, v5.10 and v5.4.
Thanks Anders, why not just use the latest perf for these kernels? Perhaps this discussion is relevant:
https://lore.kernel.org/linux-perf-users/[email protected]/
Thanks,
Ian
Quoted 69 lines
I built it with tuxmake [1] here are two example commandlines:
tuxmake --runtime podman --target-arch arm64 --toolchain gcc-12 --kconfig defconfig perf tuxmake --runtime podman --target-archx86_64--toolchain gcc-12 --kconfig defconfig perfTried to build perf with both gcc-11 and gcc-12.
Patch 'tools perf: Fix compilation error with new binutils' and 'tools build: Add feature test for
init_disassemble_infoAPI changes' didn't apply cleanly, thats why I send these in a patchset.When apply 'tools build: Add feature test for
init_disassemble_infoAPI changes' to 5.4 it will be a minor merge conflict, do you want me to send this patch in two separate patches one for 5.4 and another for v5.10?The sha for these two patches in mainline are.
cfd59ca91467tools build: Add feature test forinit_disassemble_infoAPI changes83aa0120487etools perf: Fix compilation error with new binutilsThe above patches solves these:
util/annotate.c: In function 'symbol__disassemble_bpf':util/annotate.c:1729:9: error: too few arguments to function 'init_disassemble_info'1729 | init_disassemble_info(&info, s, | ^~~~~~~~~~~~~~~~~~~~~Please apply these to v5.10 and v5.4
a45b3d692623tools include: adddis-asm-compat.hto handle version differencesd08c84e01afaperf sched: CastPTHREAD_STACK_MINto int as it may turn into sysconf(__SC_THREAD_STACK>The above patches solves these:
/home/anders/src/kernel/stable-5.10/tools/include/linux/kernel.h:43:24: error: comparison of distinct pointer types lacks a cast [-Werror]
43 | (void) (&_max1==&_max2); \
| ^~
builtin-sched.c:673:34: note: in expansion of macro 'max'
673 | (size_t)max(16 * 1024, PTHREAD_STACK_MIN));
| ^~~Please apply these to v5.15, v5.10 and v5.4
8e8bf60a6754perf build: Fixup disabling of -Wdeprecated-declarations for the python scripting engine4ee3c4da8b1bperf scripting python: Do not build fail on deprecation warnings63a4354ae75cperf scripting perl: Ignore some warnings to keep building with perl headersBuild error that the above 3 patches solves are:
/usr/lib/x86_64-linux-gnu/perl/5.36/CORE/handy.h:125:23: error: cast from function call of type 'STRLEN' {aka 'long unsigned int'} to non-matching type '_Bool' [-Werror=bad-function-cast]
125 | #define cBOOL(cbool) ((bool) (cbool))
| ^Cheers,
Anders
[1] https://tuxmake.org/Andres Freund (2):
tools perf: Fix compilation error with new binutils tools build: Add feature test forinit_disassemble_infoAPI changestools/build/Makefile.feature | 1 + tools/build/feature/Makefile | 4 ++++ tools/build/feature/test-all.c | 4 ++++ tools/build/feature/test-disassembler-init-styled.c | 13 +++++++++++++ tools/perf/Makefile.config | 8 ++++++++ tools/perf/util/annotate.c | 7 ++++--- 6 files changed, 34 insertions(+), 3 deletions(-) create mode 100644 tools/build/feature/test-disassembler-init-styled.c--
2.39.2
Greg KH
maintainerQuoted 43 lines · Anders Roxell
Hi,
I would like to see these patches backported. They are needed so perf
can be cross compiled with gcc on v5.15, v5.10 and v5.4.
I built it with tuxmake [1] here are two example commandlines:
tuxmake --runtime podman --target-arch arm64 --toolchain gcc-12 --kconfig defconfig perf tuxmake --runtime podman --target-archx86_64--toolchain gcc-12 --kconfig defconfig perfTried to build perf with both gcc-11 and gcc-12.
Patch 'tools perf: Fix compilation error with new binutils' and 'tools build: Add feature test for
init_disassemble_infoAPI changes' didn't apply cleanly, thats why I send these in a patchset.When apply 'tools build: Add feature test for
init_disassemble_infoAPI changes' to 5.4 it will be a minor merge conflict, do you want me to send this patch in two separate patches one for 5.4 and another for v5.10?The sha for these two patches in mainline are.
cfd59ca91467tools build: Add feature test forinit_disassemble_infoAPI changes83aa0120487etools perf: Fix compilation error with new binutilsThe above patches solves these:
util/annotate.c: In function 'symbol__disassemble_bpf':util/annotate.c:1729:9: error: too few arguments to function 'init_disassemble_info'1729 | init_disassemble_info(&info, s, | ^~~~~~~~~~~~~~~~~~~~~Please apply these to v5.10 and v5.4
a45b3d692623tools include: adddis-asm-compat.hto handle version differencesd08c84e01afaperf sched: CastPTHREAD_STACK_MINto int as it may turn into sysconf(__SC_THREAD_STACK>The above patches solves these:
/home/anders/src/kernel/stable-5.10/tools/include/linux/kernel.h:43:24: error: comparison of distinct pointer types lacks a cast [-Werror]
43 | (void) (&_max1==&_max2); \
| ^~
builtin-sched.c:673:34: note: in expansion of macro 'max'
673 | (size_t)max(16 * 1024, PTHREAD_STACK_MIN));
| ^~~
Quoted 4 lines
Please apply these to v5.15, v5.10 and v5.4
8e8bf60a6754perf build: Fixup disabling of -Wdeprecated-declarations for the python scripting engine4ee3c4da8b1bperf scripting python: Do not build fail on deprecation warnings63a4354ae75cperf scripting perl: Ignore some warnings to keep building with perl headers
Can you please provide patch series of these upstream commits backported to the relevant branchs that you wish to see them in? You have 2 patches in this series without git commit ids, and I have no idea where to apply them, or not apply them...
Or better yet, just use the latest version of perf as was pointed out, on these old kernel releases.
thanks,
greg k-h
Quoted 50 lines · Greg KH
Quoted 43 lines · Anders Roxell
Hi,
I would like to see these patches backported. They are needed so perf
can be cross compiled with gcc on v5.15, v5.10 and v5.4.
I built it with tuxmake [1] here are two example commandlines:
tuxmake --runtime podman --target-arch arm64 --toolchain gcc-12 --kconfig defconfig perf tuxmake --runtime podman --target-archx86_64--toolchain gcc-12 --kconfig defconfig perfTried to build perf with both gcc-11 and gcc-12.
Patch 'tools perf: Fix compilation error with new binutils' and 'tools build: Add feature test for
init_disassemble_infoAPI changes' didn't apply cleanly, thats why I send these in a patchset.When apply 'tools build: Add feature test for
init_disassemble_infoAPI changes' to 5.4 it will be a minor merge conflict, do you want me to send this patch in two separate patches one for 5.4 and another for v5.10?The sha for these two patches in mainline are.
cfd59ca91467tools build: Add feature test forinit_disassemble_infoAPI changes83aa0120487etools perf: Fix compilation error with new binutilsThe above patches solves these:
util/annotate.c: In function 'symbol__disassemble_bpf':util/annotate.c:1729:9: error: too few arguments to function 'init_disassemble_info'1729 | init_disassemble_info(&info, s, | ^~~~~~~~~~~~~~~~~~~~~Please apply these to v5.10 and v5.4
a45b3d692623tools include: adddis-asm-compat.hto handle version differencesd08c84e01afaperf sched: CastPTHREAD_STACK_MINto int as it may turn into sysconf(__SC_THREAD_STACK>The above patches solves these:
/home/anders/src/kernel/stable-5.10/tools/include/linux/kernel.h:43:24: error: comparison of distinct pointer types lacks a cast [-Werror]
43 | (void) (&_max1==&_max2); \
| ^~
builtin-sched.c:673:34: note: in expansion of macro 'max'
673 | (size_t)max(16 * 1024, PTHREAD_STACK_MIN));
| ^~~Quoted 4 lines
Please apply these to v5.15, v5.10 and v5.4
8e8bf60a6754perf build: Fixup disabling of -Wdeprecated-declarations for the python scripting engine4ee3c4da8b1bperf scripting python: Do not build fail on deprecation warnings63a4354ae75cperf scripting perl: Ignore some warnings to keep building with perl headers
Quoted 4 lines
Can you please provide patch series of these upstream commits backported to the relevant branchs that you wish to see them in? You have 2 patches in this series without git commit ids, and I have no idea where to apply them, or not apply them...
Yes, apologies, I will get that fixed up.
Quoted 3 lines
Or better yet, just use the latest version of perf as was pointed out, on these old kernel releases.
Makes sense, we can do this. Is this the preferred way going forward?
Cheers,
Anders
Ian Rogers
reviewerQuoted 57 lines · Anders Roxell
Quoted 50 lines · Greg KH
Quoted 43 lines · Anders Roxell
Hi,
I would like to see these patches backported. They are needed so perf
can be cross compiled with gcc on v5.15, v5.10 and v5.4.
I built it with tuxmake [1] here are two example commandlines:
tuxmake --runtime podman --target-arch arm64 --toolchain gcc-12 --kconfig defconfig perf tuxmake --runtime podman --target-archx86_64--toolchain gcc-12 --kconfig defconfig perfTried to build perf with both gcc-11 and gcc-12.
Patch 'tools perf: Fix compilation error with new binutils' and 'tools build: Add feature test for
init_disassemble_infoAPI changes' didn't apply cleanly, thats why I send these in a patchset.When apply 'tools build: Add feature test for
init_disassemble_infoAPI changes' to 5.4 it will be a minor merge conflict, do you want me to send this patch in two separate patches one for 5.4 and another for v5.10?The sha for these two patches in mainline are.
cfd59ca91467tools build: Add feature test forinit_disassemble_infoAPI changes83aa0120487etools perf: Fix compilation error with new binutilsThe above patches solves these:
util/annotate.c: In function 'symbol__disassemble_bpf':util/annotate.c:1729:9: error: too few arguments to function 'init_disassemble_info'1729 | init_disassemble_info(&info, s, | ^~~~~~~~~~~~~~~~~~~~~Please apply these to v5.10 and v5.4
a45b3d692623tools include: adddis-asm-compat.hto handle version differencesd08c84e01afaperf sched: CastPTHREAD_STACK_MINto int as it may turn into sysconf(__SC_THREAD_STACK>The above patches solves these:
/home/anders/src/kernel/stable-5.10/tools/include/linux/kernel.h:43:24: error: comparison of distinct pointer types lacks a cast [-Werror]
43 | (void) (&_max1==&_max2); \
| ^~
builtin-sched.c:673:34: note: in expansion of macro 'max'
673 | (size_t)max(16 * 1024, PTHREAD_STACK_MIN));
| ^~~Quoted 4 lines
Please apply these to v5.15, v5.10 and v5.4
8e8bf60a6754perf build: Fixup disabling of -Wdeprecated-declarations for the python scripting engine4ee3c4da8b1bperf scripting python: Do not build fail on deprecation warnings63a4354ae75cperf scripting perl: Ignore some warnings to keep building with perl headersQuoted 4 lines
Can you please provide patch series of these upstream commits backported to the relevant branchs that you wish to see them in? You have 2 patches in this series without git commit ids, and I have no idea where to apply them, or not apply them...
Quoted 7 lines
Yes, apologies, I will get that fixed up.
Quoted 3 lines
Or better yet, just use the latest version of perf as was pointed out, on these old kernel releases.
Makes sense, we can do this. Is this the preferred way going forward?
Fwiw, it definitely has my vote.
Thanks,
Ian
Quoted 2 lines
Cheers,
Anders