perf/core clang fixes
Subsystems (2)
Arnaldo Carvalho de Melo
maintainerHi Ingo,
Here is a bunch of changes to allow buildint tools/perf/ with CC=clang,
there are a few more to allow building the python code, but I already manage to finish the build without it.
Please consider pulling,
- Arnaldo
Test results at the end of this message, as usual.
The following changes since commit 277d6f1dcae09aed63cd4c7900a280b0e18cf2ca:
Merge tag 'perf-core-for-mingo-4.11-20170213' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core (2017-02-14 07:29:38 +0100)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-core-for-mingo-4.11-20170215
for you to fetch changes up to 34a0548f01626b01c9e98d9627812c3c9f6b6f7d:
perf tools: Add missing parse_events_error() prototype (2017-02-15 11:20:49 -0300)
----------------------------------------------------------------
perf/core clang fixes:
Changes to make tools/{perf,lib/{bpf,traceevent,api}} build with CC=clang, to, for instance, take advantage of warnings (Arnaldo Carvalho de Melo):
- Conditionally request some warning options not available on clang
- Set the maximum optimization level to -O3 when using CC=clang, leave the previous setting of -O6 otherwise.
- Make it an error to pass a signed value to OPTION_UINTEGER, so that we can remove abs(unsigned int) calls in 'perf bench futex'.
- Make sure dprintf() is not defined before using that name in 'perf bench numa'
- Avoid using field after variable sized type, its a GNU extension, use equivalent code.
- Fix some bugs where some variables could be used unitialized, something not caught by gcc.
- Fix some spots where we were testing struct->array[] members against NULL, it will always evaluate to 'true'.
- Add missing parse_events_error() prototype in the bison file.
There are still one problem when trying to build the python support, but this are the 'size' outputs for 'make -C tools/perf NO_LIBPYTHON' for gcc and clang builds:
DW_AT_producer: clang version 4.0.0 (http://llvm.org/git/clang.git f5be8ba13adc4ba1011a7ccd60c844bd60427c1c) (ht
$ size ~/bin/perf
text data bss dec hex filename
3447514 831320 23901696 28180530 1ae0032 /home/acme/bin/perf DW_AT_producer: GNU C99 6.3.1 20161221 (Red Hat 6.3.1-1) -mtune=generic -march=x86-64 -ggdb3 -O6 -std=gnu99
+-fno-omit-frame-pointer -funwind-tables -fstack-protector-all
$ size ~/bin/perf
text data bss dec hex filename
3671662 836480 23902752 28410894 1b1840e /home/acme/bin/perf----------------------------------------------------------------
Arnaldo Carvalho de Melo (14):
tools: Suppress request for warning options not existent in clang
tools: Set the maximum optimization level according to the compiler being used tools lib subcmd: Make it an error to pass a signed value to OPTION_UINTEGER Revert "perf bench futex: Sanitize numeric parameters"
perf bench numa: Make sure dprintf() is not defined
perf tests: Synthesize struct instead of using field after variable sized type perf record: Do not put a variable sized type not at the end of a struct perf tools: Do not put a variable sized type not at the end of a struct perf probe: Avoid accessing uninitialized 'map' variable
perf evsel: Do not put a variable sized type not at the end of a struct perf intel pt decoder: clang has no -Wno-override-init
perf tools: Be consistent on the type of map->symbols[] interator perf pmu: Fix check for unset alias->unit array
perf tools: Add missing parse_events_error() prototype
tools/lib/api/Makefile | 8 +++- tools/lib/subcmd/Makefile | 8 +++- tools/lib/subcmd/parse-options.c | 4 ++ tools/perf/Makefile.config | 4 ++ tools/perf/bench/futex-hash.c | 4 -- tools/perf/bench/futex-lock-pi.c | 3 -- tools/perf/bench/futex-requeue.c | 2 - tools/perf/bench/futex-wake-parallel.c | 4 -- tools/perf/bench/futex-wake.c | 3 -- tools/perf/bench/futex.h | 4 -- tools/perf/bench/numa.c | 1 + tools/perf/builtin-record.c | 17 +++++---- tools/perf/tests/parse-no-sample-id-all.c | 19 ++++------ tools/perf/util/evsel.c | 62 ++++++++++++++----------------- tools/perf/util/intel-pt-decoder/Build | 6 ++- tools/perf/util/machine.c | 4 +- tools/perf/util/parse-events.c | 15 +++----- tools/perf/util/parse-events.y | 2 + tools/perf/util/pmu.c | 4 +- tools/perf/util/probe-event.c | 2 +- tools/perf/util/session.c | 2 +- tools/scripts/Makefile.include | 5 ++- 22 files changed, 91 insertions(+), 92 deletions(-)
Test results:
The first ones are container (docker) based builds of tools/perf with and without libelf support, objtool where it is supported and samples/bpf/, ditto.
Several are cross builds, the ones with -x-ARCH, and the android one, and those may not have all the features built, due to lack of multi-arch devel packages, available and being used so far on just a few, like debian:experimental-x-{arm64,mipsel}.
The 'perf test' one will perform a variety of tests exercising tools/perf/util/, tools/lib/{bpf,traceevent,etc}, as well as run perf commands with a variety of command line event specifications to then intercept the sys_perf_event syscall to check that the perf_event_attr fields are set up as expected, among a variety of other unit tests.
Then there is the 'make -C tools/perf build-test' ones, that build tools/perf/ with a variety of feature sets, exercising the build with an incomplete set of features as well as with a complete one. It is planned to have it run on each of the containers mentioned above, using some container orchestration infrastructure. Get in contact if interested in helping having this in place.
# time dm
1 alpine:3.4: Ok
2 android-ndk:r12b-arm: Ok
3 archlinux:latest: Ok
4 centos:5: Ok
5 centos:6: Ok
6 centos:7: Ok
7 debian:7: Ok
8 debian:8: Ok
9 debian:experimental: Ok
10 debian:experimental-x-arm64: Ok
11 debian:experimental-x-mips: Ok
12 debian:experimental-x-mips64: Ok
13 debian:experimental-x-mipsel: Ok
14 fedora:20: Ok
15 fedora:21: Ok
16 fedora:22: Ok
17 fedora:23: Ok
18 fedora:24: Ok
19 fedora:24-x-ARC-uClibc: Ok
20 fedora:25: Ok
21 fedora:rawhide: Ok
22 mageia:5: Ok
23 opensuse:13.2: Ok
24 opensuse:42.1: Ok
25 opensuse:tumbleweed: Ok
26 ubuntu:12.04.5: Ok
27 ubuntu:14.04.4-x-linaro-arm64: Ok
28 ubuntu:15.10: Ok
29 ubuntu:16.04: Ok
30 ubuntu:16.04-x-arm: Ok
31 ubuntu:16.04-x-arm64: Ok
32 ubuntu:16.04-x-powerpc: Ok
33 ubuntu:16.04-x-powerpc64: Ok
34 ubuntu:16.04-x-powerpc64el: Ok
35 ubuntu:16.04-x-s390: Ok
36 ubuntu:16.10: Ok
#
# uname -a
Linux jouet 4.10.0-rc8 #2 SMP Wed Feb 15 15:26:36 BRT 2017 x86_64 x86_64 x86_64 GNU/Linux
# perf test
1: vmlinux symtab matches kallsyms : Ok
2: Detect openat syscall event : Ok
3: Detect openat syscall event on all cpus : Ok
4: Read samples using the mmap interface : Ok
5: Parse event definition strings : Ok
6: PERF_RECORD_* events & perf_sample fields : Ok
7: Parse perf pmu format : Ok
8: DSO data read : Ok
9: DSO data cache : Ok
10: DSO data reopen : Ok
11: Roundtrip evsel->name : Ok
12: Parse sched tracepoints fields : Ok
13: syscalls:sys_enter_openat event fields : Ok
14: Setup struct perf_event_attr : Ok
15: Match and link multiple hists : Ok
16: 'import perf' in python : Ok
17: Breakpoint overflow signal handler : Ok
18: Breakpoint overflow sampling : Ok
19: Number of exit events of a simple workload : Ok
20: Software clock events period values : Ok
21: Object code reading : Ok
22: Sample parsing : Ok
23: Use a dummy software event to keep tracking: Ok
24: Parse with no sample_id_all bit set : Ok
25: Filter hist entries : Ok
26: Lookup mmap thread : Ok
27: Share thread mg : Ok
28: Sort output of hist entries : Ok
29: Cumulate child hist entries : Ok
30: Track with sched_switch : Ok
31: Filter fds with revents mask in a fdarray : Ok
32: Add fd to a fdarray, making it autogrow : Ok
33: kmod_path__parse : Ok
34: Thread map : Ok
35: LLVM search and compile :
35.1: Basic BPF llvm compile : Ok
35.2: kbuild searching : Ok
35.3: Compile source for BPF prologue generation: Ok
35.4: Compile source for BPF relocation : Ok
36: Session topology : Ok
37: BPF filter :
37.1: Basic BPF filtering : Ok
37.2: BPF pinning : Ok
37.3: BPF prologue generation : Ok
37.4: BPF relocation checker : Ok
38: Synthesize thread map : Ok
39: Remove thread map : Ok
40: Synthesize cpu map : Ok
41: Synthesize stat config : Ok
42: Synthesize stat : Ok
43: Synthesize stat round : Ok
44: Synthesize attr update : Ok
45: Event times : Ok
46: Read backward ring buffer : Ok
47: Print cpu map : Ok
48: Probe SDT events : Ok
49: is_printable_array : Ok
50: Print bitmap : Ok
51: perf hooks : Ok
52: builtin clang support : Skip (not compiled in)
53: unit_number__scnprintf : Ok
54: x86 rdpmc : Ok
55: Convert perf time to TSC : Ok
56: DWARF unwind : Ok
57: x86 instruction decoder - new instructions : Ok
58: Intel cqm nmi context read : Skip
#
$ make -C tools/perf build-test
make: Entering directory '/home/acme/git/linux/tools/perf'
- tarpkg: ./tests/perf-targz-src-pkg .
make_install_prefix_O: make install prefix=/tmp/krava
make_no_slang_O: make NO_SLANG=1
make_help_O: make help
make_no_libaudit_O: make NO_LIBAUDIT=1
make_no_demangle_O: make NO_DEMANGLE=1
make_no_gtk2_O: make NO_GTK2=1
make_no_libdw_dwarf_unwind_O: make NO_LIBDW_DWARF_UNWIND=1
make_minimal_O: make NO_LIBPERL=1 NO_LIBPYTHON=1 NO_NEWT=1 NO_GTK2=1 NO_DEMANGLE=1 NO_LIBELF=1 NO_LIBUNWIND=1 NO_BACKTRACE=1 NO_LIBNUMA=1 NO_LIBAUDIT=1 NO_LIBBIONIC=1 NO_LIBDW_DWARF_UNWIND=1 NO_AUXTRACE=1 NO_LIBBPF=1 NO_LIBCRYPTO=1 NO_SDT=1 NO_JVMTI=1
make_util_pmu_bison_o_O: make util/pmu-bison.o
make_install_prefix_slash_O: make install prefix=/tmp/krava/
make_util_map_o_O: make util/map.o
make_perf_o_O: make perf.o
make_with_clangllvm_O: make LIBCLANGLLVM=1
make_no_libunwind_O: make NO_LIBUNWIND=1
make_doc_O: make doc
make_pure_O: make
make_debug_O: make DEBUG=1
make_no_libnuma_O: make NO_LIBNUMA=1
make_no_libpython_O: make NO_LIBPYTHON=1
make FEATURE_DUMP_COPY=/home/acme/git/linux/tools/perf/BUILD_TEST_FEATURE_DUMP_STATIC LDFLAGS='-static' feature-dump
make_static_O: make LDFLAGS=-static
make_no_backtrace_O: make NO_BACKTRACE=1
make_with_babeltrace_O: make LIBBABELTRACE=1
make_install_O: make install
make_no_auxtrace_O: make NO_AUXTRACE=1
make_no_libperl_O: make NO_LIBPERL=1
make_install_bin_O: make install-bin
make_no_ui_O: make NO_NEWT=1 NO_SLANG=1 NO_GTK2=1
make_no_scripts_O: make NO_LIBPYTHON=1 NO_LIBPERL=1
make_no_libelf_O: make NO_LIBELF=1
make_no_newt_O: make NO_NEWT=1
make_clean_all_O: make clean all
make_tags_O: make tags
make_no_libbpf_O: make NO_LIBBPF=1
make_no_libbionic_O: make NO_LIBBIONIC=1
OK
make: Leaving directory '/home/acme/git/linux/tools/perf'
$ perf stat make -C tools/perf build-testArnaldo Carvalho de Melo
maintainerPATCH 1/14 tools: Suppress request for warning options not existent in clang
From: Arnaldo Carvalho de Melo <[email protected]>
To allow building with clang, avoiding:
error: unknown warning option '-Wstrict-aliasing=3'; did you mean '-Wstring-plus-int'? [-Werror,-Wunknown-warning-option]
--- tools/scripts/Makefile.include | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
tools/scripts/Makefile.include +4 -1
@@ -32,7 +32,6 @@ EXTRA_WARNINGS += -Wold-style-definition EXTRA_WARNINGS += -Wpacked EXTRA_WARNINGS += -Wredundant-decls EXTRA_WARNINGS += -Wshadow -EXTRA_WARNINGS += -Wstrict-aliasing=3 EXTRA_WARNINGS += -Wstrict-prototypes EXTRA_WARNINGS += -Wswitch-default EXTRA_WARNINGS += -Wswitch-enum
@@ -40,6 +39,10 @@ EXTRA_WARNINGS += -Wundef EXTRA_WARNINGS += -Wwrite-strings EXTRA_WARNINGS += -Wformat +ifneq ($(CC), clang) +EXTRA_WARNINGS += -Wstrict-aliasing=3 +endif + ifneq ($(findstring $(MAKEFLAGS), w),w) PRINT_DIR = --no-print-directory else
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 2/14 tools: Set the maximum optimization level according to the compiler being used
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
To avoid this when using clang:
warning: optimization level '-O6' is not supported; using '-O3' instead
--- tools/lib/api/Makefile | 8 +++++++- tools/lib/subcmd/Makefile | 8 +++++++- tools/perf/Makefile.config | 4 ++++ 3 files changed, 18 insertions(+), 2 deletions(-)
tools/lib/api/Makefile +7 -1
@@ -17,7 +17,13 @@ MAKEFLAGS += --no-print-directory LIBFILE = $(OUTPUT)libapi.a CFLAGS := $(EXTRA_WARNINGS) $(EXTRA_CFLAGS) -CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -O6 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC +CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC + +ifeq ($(CC), clang) + CFLAGS += -O3 +else + CFLAGS += -O6 +endif # Treat warnings as errors unless directed not to ifneq ($(WERROR),0)
tools/lib/subcmd/Makefile +7 -1
@@ -19,7 +19,13 @@ MAKEFLAGS += --no-print-directory LIBFILE = $(OUTPUT)libsubcmd.a CFLAGS := $(EXTRA_WARNINGS) $(EXTRA_CFLAGS) -CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -O6 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC +CFLAGS += -ggdb3 -Wall -Wextra -std=gnu99 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fPIC + +ifeq ($(CC), clang) + CFLAGS += -O3 +else + CFLAGS += -O6 +endif # Treat warnings as errors unless directed not to ifneq ($(WERROR),0)
tools/perf/Makefile.config +4 -0
@@ -144,8 +144,12 @@ ifndef DEBUG endif ifeq ($(DEBUG),0) +ifeq ($(CC), clang) + CFLAGS += -O3 +else CFLAGS += -O6 endif +endif ifdef PARSER_DEBUG PARSER_DEBUG_BISON := -t
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 3/14 tools lib subcmd: Make it an error to pass a signed value to OPTION_UINTEGER
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
Options marked OPTION_UINTEGER or OPTION_U64 clearly indicates that an unsigned value is expected, so just error out when a negative value is passed, instead of returning something undesired to the tool.
E.g.:
# perf bench futex hash -t -4 # Running 'futex/hash' benchmark: Error: switch `t' expects an unsigned numerical value Usage: perf bench futex hash <options>
-t, --threads <n> Specify amount of threads
#
--- tools/lib/subcmd/parse-options.c | 4 ++++ 1 file changed, 4 insertions(+)
tools/lib/subcmd/parse-options.c +4 -0
@@ -270,6 +270,8 @@ static int get_value(struct parse_opt_ctx_t *p, }if (get_arg (p, opt, flags, &arg))return -1 ; +if (arg[0 ] =='-' ) +return opterror (opt,"expects an unsigned numerical value" , flags); *(unsigned int *)opt->value =strtol (arg, (char **)&s,10 );if (*s)return opterror (opt,"expects a numerical value" , flags);
@@ -302,6 +304,8 @@ static int get_value(struct parse_opt_ctx_t *p, }if (get_arg (p, opt, flags, &arg))return -1 ; +if (arg[0 ] =='-' ) +return opterror (opt,"expects an unsigned numerical value" , flags); *(u64 *)opt->value =strtoull (arg, (char **)&s,10 );if (*s)return opterror (opt,"expects a numerical value" , flags);
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 4/14 Revert "perf bench futex: Sanitize numeric parameters"
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
This reverts commit 60758d6668b3e2fa8e5fd143d24d0425203d007e.
Now that libsubcmd makes sure that OPT_UINTEGER options will not return negative values, we can revert this patch while addressing the problem it solved:
# perf bench futex hash -t -4 # Running 'futex/hash' benchmark: Error: switch `t' expects an unsigned numerical value Usage: perf bench futex hash <options>
-t, --threads <n> Specify amount of threads # perf bench futex hash -t-4 # Running 'futex/hash' benchmark: Error: switch `t' expects an unsigned numerical value Usage: perf bench futex hash <options>
-t, --threads <n> Specify amount of threads
#
IMO it is more reasonable to flat out refuse to process a negative number than to silently turn it into an absolute value.
This also helps in silencing clang's complaint about asking for an absolute value of an unsigned integer:
bench/futex-hash.c:133:10: error: taking the absolute value of unsigned type 'unsigned int' has no effect [-Werror,-Wabsolute-value]
nsecs = futexbench_sanitize_numeric(nsecs);
^
bench/futex.h:104:42: note: expanded from macro 'futexbench_sanitize_numeric'
#define futexbench_sanitize_numeric(__n) abs((__n))
^ bench/futex-hash.c:133:10: note: remove the call to 'abs' since unsigned values cannot be negative
--- tools/perf/bench/futex-hash.c | 4 ---- tools/perf/bench/futex-lock-pi.c | 3 --- tools/perf/bench/futex-requeue.c | 2 -- tools/perf/bench/futex-wake-parallel.c | 4 ---- tools/perf/bench/futex-wake.c | 3 --- tools/perf/bench/futex.h | 4 ---- 6 files changed, 20 deletions(-)
tools/perf/bench/futex-hash.c +0 -4
@@ -130,8 +130,6 @@ int bench_futex_hash(int argc, const char **argv, } ncpus =sysconf (_SC_NPROCESSORS_ONLN); - nsecs =futexbench_sanitize_numeric (nsecs); - nfutexes =futexbench_sanitize_numeric (nfutexes);sigfillset (&act.sa_mask); act.sa_sigaction = toggle_done;
@@ -139,8 +137,6 @@ int bench_futex_hash(int argc, const char **argv,if (!nthreads)/* default to the number of CPUs */ nthreads = ncpus; -else - nthreads =futexbench_sanitize_numeric (nthreads); worker =calloc (nthreads,sizeof (*worker));if (!worker)
tools/perf/bench/futex-lock-pi.c +0 -3
@@ -152,7 +152,6 @@ int bench_futex_lock_pi(int argc, const char **argv, goto err; ncpus =sysconf (_SC_NPROCESSORS_ONLN); - nsecs =futexbench_sanitize_numeric (nsecs);sigfillset (&act.sa_mask); act.sa_sigaction = toggle_done;
@@ -160,8 +159,6 @@ int bench_futex_lock_pi(int argc, const char **argv,if (!nthreads) nthreads = ncpus; -else - nthreads =futexbench_sanitize_numeric (nthreads); worker =calloc (nthreads,sizeof (*worker));if (!worker)
tools/perf/bench/futex-requeue.c +0 -2
@@ -128,8 +128,6 @@ int bench_futex_requeue(int argc, const char **argv,if (!nthreads) nthreads = ncpus; -else - nthreads =futexbench_sanitize_numeric (nthreads); worker =calloc (nthreads,sizeof (*worker));if (!worker)
tools/perf/bench/futex-wake-parallel.c +0 -4
@@ -217,12 +217,8 @@ int bench_futex_wake_parallel(int argc, const char **argv,sigaction (SIGINT , &act,NULL ); ncpus =sysconf (_SC_NPROCESSORS_ONLN); - nwaking_threads =futexbench_sanitize_numeric (nwaking_threads); -if (!nblocked_threads) nblocked_threads = ncpus; -else - nblocked_threads =futexbench_sanitize_numeric (nblocked_threads);/* some sanity checks */ if (nwaking_threads > nblocked_threads || !nwaking_threads)
tools/perf/bench/futex-wake.c +0 -3
@@ -129,7 +129,6 @@ int bench_futex_wake(int argc, const char **argv, } ncpus =sysconf (_SC_NPROCESSORS_ONLN); - nwakes =futexbench_sanitize_numeric (nwakes);sigfillset (&act.sa_mask); act.sa_sigaction = toggle_done;
@@ -137,8 +136,6 @@ int bench_futex_wake(int argc, const char **argv,if (!nthreads) nthreads = ncpus; -else - nthreads =futexbench_sanitize_numeric (nthreads); worker =calloc (nthreads,sizeof (*worker));if (!worker)
tools/perf/bench/futex.h +0 -4
@@ -7,7 +7,6 @@#ifndef _FUTEX_H#define _FUTEX_H -#include <stdlib.h> #include <unistd.h> #include <sys/syscall.h> #include <sys/types.h>
@@ -100,7 +99,4 @@ static inline int pthread_attr_setaffinity_np(pthread_attr_t *attr, }#endif -/* User input sanitation */ -#define futexbench_sanitize_numeric (__n) abs((__n)) -#endif /* _FUTEX_H */
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 5/14 perf bench numa: Make sure dprintf() is not defined
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
When building with clang we get this error:
bench/numa.c:46:9: error: 'dprintf' macro redefined [-Werror,-Wmacro-redefined]
#define dprintf(x...) do { if (g && g->p.show_details >= 1) printf(x); } while (0)
^
/usr/include/bits/stdio2.h:145:12: note: previous definition is here
# define dprintf(fd, ...) \
^
CC /tmp/build/perf/tests/parse-no-sample-id-all.o1 error generated.
So, make sure it is undefined before using that name.
--- tools/perf/bench/numa.c | 1 + 1 file changed, 1 insertion(+)
tools/perf/bench/numa.c +1 -0
@@ -43,6 +43,7 @@/* * Debug printf: */ +#undef dprintf #define dprintf(x...) do { if (g && g->p.show_details >= 1) printf(x); } while (0)struct thread_data {
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 6/14 perf tests: Synthesize struct instead of using field after variable sized type
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
End result is the same, its an ABI, so the struct won't change, avoid using a GNU extension, so that we can catch other cases that may be bugs.
Caught when building with clang:
tests/parse-no-sample-id-all.c:53:20: error: field 'attr' with variable sized type 'struct attr_event' not at the end of a struct or class is a GNU extension
[-Werror,-Wgnu-variable-sized-type-not-at-end]
struct attr_event attr;
^1 error generated.
Testing it:
# perf test sample_id 24: Parse with no sample_id_all bit set : Ok #
--- tools/perf/tests/parse-no-sample-id-all.c | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-)
tools/perf/tests/parse-no-sample-id-all.c +8 -11
@@ -50,7 +50,8 @@ static int process_events(union perf_event **events, size_t count) }struct test_attr_event { -struct attr_event attr; +struct perf_event_header header; +struct perf_event_attr attr;u64 id; };
@@ -71,20 +72,16 @@ int test__parse_no_sample_id_all(int subtest __maybe_unused)int err;struct test_attr_event event1 = { - .attr = { - .header = { - .type =PERF_RECORD_HEADER_ATTR , - .size =sizeof (struct test_attr_event ), - }, + .header = { + .type =PERF_RECORD_HEADER_ATTR , + .size =sizeof (struct test_attr_event ), }, .id =1 , };struct test_attr_event event2 = { - .attr = { - .header = { - .type =PERF_RECORD_HEADER_ATTR , - .size =sizeof (struct test_attr_event ), - }, + .header = { + .type =PERF_RECORD_HEADER_ATTR , + .size =sizeof (struct test_attr_event ), }, .id =2 , };
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 7/14 perf record: Do not put a variable sized type not at the end of a struct
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
As this is a GNU extension and while harmless in this case, we can do the same thing in a more clearer way by using an existing thread_map constructor.
With this we avoid this while compiling with clang:
builtin-record.c:659:21: error: field 'map' with variable sized type 'struct thread_map' not at the end of a struct or class is a GNU extension
[-Werror,-Wgnu-variable-sized-type-not-at-end]
struct thread_map map;
^1 error generated.
--- tools/perf/builtin-record.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-)
tools/perf/builtin-record.c +9 -8
@@ -655,22 +655,23 @@ record__finish_output(struct record *rec)static int record__synthesize_workload (struct record *rec,bool tail) { -struct { -struct thread_map map; -struct thread_map_data map_data; - } thread_map; +int err; +struct thread_map *thread_map;if (rec->opts.tail_synthesize != tail)return 0 ; - thread_map.map.nr =1 ; - thread_map.map.map[0 ].pid = rec->evlist->workload.pid; - thread_map.map.map[0 ].comm =NULL ; -return perf_event__synthesize_thread_map (&rec->tool, &thread_map.map, + thread_map =thread_map__new_by_tid (rec->evlist->workload.pid); +if (thread_map ==NULL ) +return -1 ; + + err =perf_event__synthesize_thread_map (&rec->tool, thread_map, process_synthesized_event, &rec->session->machines.host, rec->opts.sample_address, rec->opts.proc_map_timeout); +thread_map__put (thread_map); +return err; }static int record__synthesize (struct record *rec,bool tail);
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 8/14 perf tools: Do not put a variable sized type not at the end of a struct
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
As this is a GNU extension and while harmless in this case, we can do the same thing in a more clearer way by using an existing thread_map constructor.
With this we avoid this while compiling with clang:
util/parse-events.c:2024:21: error: field 'map' with variable sized type 'struct thread_map' not at the end of a struct or class is a GNU extension
[-Werror,-Wgnu-variable-sized-type-not-at-end]
struct thread_map map;
^1 error generated.
--- tools/perf/util/parse-events.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-)
tools/perf/util/parse-events.c +6 -9
@@ -2020,17 +2020,14 @@ static bool is_event_supported(u8 type, unsigned config) .config = config, .disabled = 1, }; -struct { -struct thread_map map; -int threads[1 ]; - } tmap = { - .map.nr =1 , - .threads = {0 }, - }; +struct thread_map *tmap =thread_map__new_by_tid (0 ); + +if (tmap ==NULL ) +return false; evsel =perf_evsel__new (&attr);if (evsel) { - open_return =perf_evsel__open (evsel,NULL , &tmap.map); + open_return =perf_evsel__open (evsel,NULL , tmap); ret = open_return >=0 ;if (open_return == -EACCES ) {
@@ -2042,7 +2039,7 @@ static bool is_event_supported(u8 type, unsigned config)* */ evsel->attr.exclude_kernel =1 ; - ret =perf_evsel__open (evsel,NULL , &tmap.map) >=0 ; + ret =perf_evsel__open (evsel,NULL , tmap) >=0 ; }perf_evsel__delete (evsel); }
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 9/14 perf probe: Avoid accessing uninitialized 'map' variable
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
Genuine problem detected with clang, the warnings are spot on:
util/probe-event.c:2079:7: error: variable 'map' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
if (addr) {
^~~~
util/probe-event.c:2094:6: note: uninitialized use occurs here
if (map && !is_kprobe) {
^~~
util/probe-event.c:2079:3: note: remove the 'if' if its condition is always true
if (addr) {
^~~~~~~~~~
util/probe-event.c:2075:8: error: variable 'map' is used uninitialized whenever 'if' condition is true [-Werror,-Wsometimes-uninitialized]
if (kernel_get_symbol_address_by_name(tp->symbol,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
util/probe-event.c:2094:6: note: uninitialized use occurs here
if (map && !is_kprobe) {
^~~
util/probe-event.c:2075:4: note: remove the 'if' if its condition is always false
if (kernel_get_symbol_address_by_name(tp->symbol,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
util/probe-event.c:2064:17: note: initialize the variable 'map' to silence this warning
struct map *map;
^
= NULL--- tools/perf/util/probe-event.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/probe-event.c +1 -1
@@ -2061,7 +2061,7 @@ static int find_perf_probe_point_from_map(struct probe_trace_point *tp, bool is_kprobe) {struct symbol *sym =NULL ; -struct map *map; +struct map *map =NULL ;u64 addr = tp->address;int ret = -ENOENT ;
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 10/14 perf evsel: Do not put a variable sized type not at the end of a struct
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
As this is a GNU extension and while harmless in this case, we can do the same thing in a more clearer way by using a existing thread_map and cpu_map constructors:
With this we avoid this while compiling with clang:
util/evsel.c:1659:17: error: field 'map' with variable sized type 'struct cpu_map' not at the end of a struct or class is a GNU extension
[-Werror,-Wgnu-variable-sized-type-not-at-end]
struct cpu_map map;
^
util/evsel.c:1667:20: error: field 'map' with variable sized type 'struct thread_map' not at the end of a struct or class is a GNU extension
[-Werror,-Wgnu-variable-sized-type-not-at-end]
struct thread_map map;
^--- tools/perf/util/evsel.c | 62 ++++++++++++++++++++++--------------------------- 1 file changed, 28 insertions(+), 34 deletions(-)
tools/perf/util/evsel.c +28 -34
@@ -1448,8 +1448,8 @@ static bool ignore_missing_thread(struct perf_evsel *evsel,return true; } -static int __perf_evsel__open (struct perf_evsel *evsel,struct cpu_map *cpus, -struct thread_map *threads) +int perf_evsel__open (struct perf_evsel *evsel,struct cpu_map *cpus, +struct thread_map *threads) {int cpu, thread, nthreads;unsigned long flags =PERF_FLAG_FD_CLOEXEC ;
@@ -1459,6 +1459,30 @@ static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,if (perf_missing_features.write_backward && evsel->attr.write_backward)return -EINVAL ; +if (cpus ==NULL ) { +static struct cpu_map *empty_cpu_map; + +if (empty_cpu_map ==NULL ) { + empty_cpu_map =cpu_map__dummy_new (); +if (empty_cpu_map ==NULL ) +return -ENOMEM ; + } + + cpus = empty_cpu_map; + } + +if (threads ==NULL ) { +static struct thread_map *empty_thread_map; + +if (empty_thread_map ==NULL ) { + empty_thread_map =thread_map__new_by_tid (-1 ); +if (empty_thread_map ==NULL ) +return -ENOMEM ; + } + + threads = empty_thread_map; + } +if (evsel->system_wide) nthreads =1 ;else
@@ -1655,46 +1679,16 @@ void perf_evsel__close(struct perf_evsel *evsel, int ncpus, int nthreads)perf_evsel__free_fd (evsel); } -static struct { -struct cpu_map map; -int cpus[1 ]; -} empty_cpu_map = { - .map.nr =1 , - .cpus = {-1 , }, -}; - -static struct { -struct thread_map map; -int threads[1 ]; -} empty_thread_map = { - .map.nr =1 , - .threads = {-1 , }, -}; - -int perf_evsel__open (struct perf_evsel *evsel,struct cpu_map *cpus, -struct thread_map *threads) -{ -if (cpus ==NULL ) { -/* Work around old compiler warnings about strict aliasing */ - cpus = &empty_cpu_map.map; - } - -if (threads ==NULL ) - threads = &empty_thread_map.map; - -return __perf_evsel__open (evsel, cpus, threads); -} -int perf_evsel__open_per_cpu (struct perf_evsel *evsel,struct cpu_map *cpus) { -return __perf_evsel__open (evsel, cpus, &empty_thread_map.map); +return perf_evsel__open (evsel, cpus,NULL ); }int perf_evsel__open_per_thread (struct perf_evsel *evsel,struct thread_map *threads) { -return __perf_evsel__open (evsel, &empty_cpu_map.map, threads); +return perf_evsel__open (evsel,NULL , threads); } static int perf_evsel__parse_id_sample(const struct perf_evsel *evsel,
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 11/14 perf intel pt decoder: clang has no -Wno-override-init
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
So set it only for other compilers, allowing us to overcome yet another build failure due to an inexistent clang -W option:
error: unknown warning option '-Wno-override-init'; did you mean '-Wno-override-module'? [-Werror,-Wunknown-warning-option]
--- tools/perf/util/intel-pt-decoder/Build | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
tools/perf/util/intel-pt-decoder/Build +5 -1
@@ -23,4 +23,8 @@ $(OUTPUT)util/intel-pt-decoder/intel-pt-insn-decoder.o: util/intel-pt-decoder/in $(call rule_mkdir) $(call if_changed_dep,cc_o_c) -CFLAGS_intel-pt-insn-decoder.o += -I$(OUTPUT)util/intel-pt-decoder -Wno-override-init +CFLAGS_intel-pt-insn-decoder.o += -I$(OUTPUT)util/intel-pt-decoder + +ifneq ($(CC), clang) + CFLAGS_intel-pt-insn-decoder.o += -Wno-override-init +endif
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 12/14 perf tools: Be consistent on the type of map->symbols[] interator
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
In a few cases we were using 'enum map_type' and that triggered this warning when using clang:
util/session.c:1923:16: error: comparison of constant 2 with expression of type 'enum map_type' is always true
[-Werror,-Wtautological-constant-out-of-range-compare]
for (i = 0; i < MAP__NR_TYPES; ++i) {--- tools/perf/util/machine.c | 4 ++-- tools/perf/util/session.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-)
tools/perf/util/machine.c +2 -2
@@ -782,7 +782,7 @@ static u64 machine__get_running_kernel_start(struct machine *machine,int __machine__create_kernel_maps (struct machine *machine,struct dso *kernel) { -enum map_type type; +int type;u64 start =machine__get_running_kernel_start (machine,NULL );/* In case of renewal the kernel map, destroy previous one */
@@ -813,7 +813,7 @@ int __machine__create_kernel_maps(struct machine *machine, struct dso *kernel)void machine__destroy_kernel_maps (struct machine *machine) { -enum map_type type; +int type;for (type =0 ; type <MAP__NR_TYPES ; ++type) {struct kmap *kmap;
tools/perf/util/session.c +1 -1
@@ -1901,7 +1901,7 @@ int maps__set_kallsyms_ref_reloc_sym(struct map **maps, const char *symbol_name, u64 addr) { char *bracket; -enum map_type i; +int i;struct ref_reloc_sym *ref; ref =zalloc (sizeof (struct ref_reloc_sym ));
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 13/14 perf pmu: Fix check for unset alias->unit array
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
The alias->unit field is an array, so to check that it is not set we should see if it is an empty string, i.e. alias->unit[0], instead of checking alias->unit != NULL, as this will _always_ evaluate to 'true'.
Pointed out by clang.
--- tools/perf/util/pmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/util/pmu.c +2 -2
@@ -945,12 +945,12 @@ static int check_info_data(struct perf_pmu_alias *alias,* define unit, scale and snapshot, fail * if there's more than one. */ -if ((info->unit && alias->unit) || +if ((info->unit && alias->unit[0 ]) || (info->scale && alias->scale) || (info->snapshot && alias->snapshot))return -EINVAL ; -if (alias->unit) +if (alias->unit[0 ]) info->unit = alias->unit;if (alias->scale)
Signature · 1 line
2.9.3
Arnaldo Carvalho de Melo
maintainerPATCH 14/14 perf tools: Add missing parse_events_error() prototype
In reply to Arnaldo Carvalho de MeloFrom: Arnaldo Carvalho de Melo <[email protected]>
As pointed out by clang, we were not providing a prototype for a function before using it:
util/parse-events.y:699:6: error: conflicting types for 'parse_events_error' void parse_events_error(YYLTYPE *loc, void *data,
^
/tmp/build/perf/util/parse-events-bison.c:2224:7: note: previous implicit declaration is here
yyerror (&yylloc, _data, scanner, YY_("syntax error"));
^ /tmp/build/perf/util/parse-events-bison.c:65:25: note: expanded from macro 'yyerror'
#define yyerror parse_events_error
1 error generated.
One line fix it.
--- tools/perf/util/parse-events.y | 2 ++ 1 file changed, 2 insertions(+)
tools/perf/util/parse-events.y +2 -0
@@ -17,6 +17,8 @@ #include "parse-events.h" #include "parse-events-bison.h" +void parse_events_error(YYLTYPE *loc, void *data, void *scanner, char const *msg); + #define ABORT_ON(val) \ do { \ if (val) \
Signature · 1 line
2.9.3
Quoted 102 lines · Arnaldo Carvalho de Melo
Hi Ingo,
Here is a bunch of changes to allow buildint
tools/perf/with CC=clang,
there are a few more to allow building the python code, but I already manage to finish the build without it.Please consider pulling,
- Arnaldo
Test results at the end of this message, as usual.
The following changes since commit
277d6f1dcae09aed63cd4c7900a280b0e18cf2ca:Merge tag 'perf-core-for-mingo-4.11-20170213' of
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linuxinto perf/core (2017-02-14 07:29:38 +0100)are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.gittags/perf-core-for-mingo-4.11-20170215for you to fetch changes up to
34a0548f01626b01c9e98d9627812c3c9f6b6f7d:perf tools: Add missing
parse_events_error()prototype (2017-02-15 11:20:49 -0300)----------------------------------------------------------------
perf/core clang fixes:Changes to make
tools/{perf,lib/{bpf,traceevent,api}}build with CC=clang, to, for instance, take advantage of warnings (Arnaldo Carvalho de Melo):- Conditionally request some warning options not available on clang
- Set the maximum optimization level to -O3 when using CC=clang, leave the previous setting of -O6 otherwise.
- Make it an error to pass a signed value to
OPTION_UINTEGER, so that we can remove abs(unsigned int) calls in 'perf bench futex'.- Make sure
dprintf()is not defined before using that name in 'perf bench numa'- Avoid using field after variable sized type, its a GNU extension, use equivalent code.
- Fix some bugs where some variables could be used unitialized, something not caught by gcc.
- Fix some spots where we were testing struct->array[] members against NULL, it will always evaluate to 'true'.
- Add missing
parse_events_error()prototype in the bison file.There are still one problem when trying to build the python support, but this are the 'size' outputs for 'make -C
tools/perfNO_LIBPYTHON' for gcc and clang builds:
DW_AT_producer: clang version 4.0.0 (http://llvm.org/git/clang.gitf5be8ba13adc4ba1011a7ccd60c844bd60427c1c) (ht$ size ~/bin/perf text data bss dec hex filename 3447514 831320 23901696 28180530 1ae0032 /home/acme/bin/perf
DW_AT_producer: GNU C99 6.3.1 20161221 (Red Hat 6.3.1-1) -mtune=generic -march=x86-64 -ggdb3 -O6 -std=gnu99
+-fno-omit-frame-pointer -funwind-tables -fstack-protector-all$ size ~/bin/perf text data bss dec hex filename 3671662 836480 23902752 28410894 1b1840e /home/acme/bin/perfSigned-off-by Arnaldo Carvalho de Melo <[email protected]>----------------------------------------------------------------
Arnaldo Carvalho de Melo (14):
tools: Suppress request for warning options not existent in clang
tools: Set the maximum optimization level according to the compiler being used tools lib subcmd: Make it an error to pass a signed value toOPTION_UINTEGERRevert "perf bench futex: Sanitize numeric parameters"
perf bench numa: Make suredprintf()is not defined
perf tests: Synthesize struct instead of using field after variable sized type perf record: Do not put a variable sized type not at the end of a struct perf tools: Do not put a variable sized type not at the end of a struct perf probe: Avoid accessing uninitialized 'map' variable
perf evsel: Do not put a variable sized type not at the end of a struct perf intel pt decoder: clang has no -Wno-override-init
perf tools: Be consistent on the type of map->symbols[] interator perf pmu: Fix check for unsetalias->unitarray
perf tools: Add missingparse_events_error()prototypetools/lib/api/Makefile | 8 +++- tools/lib/subcmd/Makefile | 8 +++- tools/lib/subcmd/parse-options.c | 4 ++ tools/perf/Makefile.config | 4 ++ tools/perf/bench/futex-hash.c | 4 -- tools/perf/bench/futex-lock-pi.c | 3 -- tools/perf/bench/futex-requeue.c | 2 - tools/perf/bench/futex-wake-parallel.c | 4 -- tools/perf/bench/futex-wake.c | 3 -- tools/perf/bench/futex.h | 4 -- tools/perf/bench/numa.c | 1 + tools/perf/builtin-record.c | 17 +++++---- tools/perf/tests/parse-no-sample-id-all.c | 19 ++++------ tools/perf/util/evsel.c | 62 ++++++++++++++----------------- tools/perf/util/intel-pt-decoder/Build | 6 ++-
Quoted 8 lines
tools/perf/util/machine.c | 4 +- tools/perf/util/parse-events.c | 15 +++----- tools/perf/util/parse-events.y | 2 + tools/perf/util/pmu.c | 4 +- tools/perf/util/probe-event.c | 2 +- tools/perf/util/session.c | 2 +- tools/scripts/Makefile.include | 5 ++- 22 files changed, 91 insertions(+), 92 deletions(-)
Pulled, thanks a lot Arnaldo!
Ingo