[PATCH v5 11/21] perf x86: Define arch_fetch_insn in NO_AUXTRACE builds

Ian Rogers posted 21 patches 3 weeks, 4 days ago
There is a newer version of this series
[PATCH v5 11/21] perf x86: Define arch_fetch_insn in NO_AUXTRACE builds
Posted by Ian Rogers 3 weeks, 4 days ago
archinsn.c containing arch_fetch_insn was only enabled with
CONFIG_AUXTRACE, but this meant that a NO_AUXTRACE build on x86 would
use the empty weak version of arch_fetch_insn - weak symbols are a
frequent source of errors like this and are outside of the C
specification. Change it so that archinsn.c is always built on x86 and
make the weak symbol empty version of arch_fetch_insn a strong one
guarded by ifdefs.

arch_fetch_insn on x86 depends on insn_decode which is a function
included then built into
intel-pt-insn-decoder.c. intel-pt-insn-decoder.c isn't built in a
NO_AUXTRACE=1 build. Separate the insn_decode function from
intel-pt-insn-decoder.c by just directly compiling the relevant
file. Guard this compilation to be for either always on x86 (because
of the use in arch_fetch_insn) or when auxtrace is enabled. Apply the
CFLAGS overrides as necessary, reducing the amount of code where
warnings are disabled.

Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/arch/x86/util/Build                 |  2 +-
 tools/perf/util/Build                          |  2 +-
 tools/perf/util/intel-pt-decoder/Build         | 18 ++++++++++++++----
 .../intel-pt-decoder/intel-pt-insn-decoder.c   |  3 ---
 tools/perf/util/trace-event-scripting.c        |  4 +++-
 5 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/tools/perf/arch/x86/util/Build b/tools/perf/arch/x86/util/Build
index bc56a8e70f34..c5df4a2cf180 100644
--- a/tools/perf/arch/x86/util/Build
+++ b/tools/perf/arch/x86/util/Build
@@ -18,6 +18,6 @@ perf-util-$(CONFIG_LOCAL_LIBUNWIND)    += unwind-libunwind.o
 perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
 
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
-perf-util-$(CONFIG_AUXTRACE) += archinsn.o
+perf-util-y += archinsn.o
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
 perf-util-$(CONFIG_AUXTRACE) += intel-bts.o
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 650974413849..340544a6f5ec 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -122,7 +122,7 @@ perf-util-y += iostat.o
 perf-util-y += stream.o
 perf-util-y += kvm-stat.o
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
-perf-util-$(CONFIG_AUXTRACE) += intel-pt-decoder/
+perf-util-y += intel-pt-decoder/
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
 perf-util-$(CONFIG_AUXTRACE) += intel-bts.o
 perf-util-$(CONFIG_AUXTRACE) += arm-spe.o
diff --git a/tools/perf/util/intel-pt-decoder/Build b/tools/perf/util/intel-pt-decoder/Build
index 30793d08c6d4..f99d150059b9 100644
--- a/tools/perf/util/intel-pt-decoder/Build
+++ b/tools/perf/util/intel-pt-decoder/Build
@@ -9,14 +9,24 @@ $(OUTPUT)util/intel-pt-decoder/inat-tables.c: $(inat_tables_script) $(inat_table
 
 # Busybox's diff doesn't have -I, avoid warning in the case
 
-$(OUTPUT)util/intel-pt-decoder/intel-pt-insn-decoder.o: util/intel-pt-decoder/intel-pt-insn-decoder.c $(OUTPUT)util/intel-pt-decoder/inat-tables.c
+ifeq ($(SRCARCH),x86)
+  perf-util-y += inat.o insn.o
+else
+  perf-util-$(CONFIG_AUXTRACE) += inat.o insn.o
+endif
+
+$(OUTPUT)util/intel-pt-decoder/inat.o: $(srctree)/tools/arch/x86/lib/inat.c $(OUTPUT)util/intel-pt-decoder/inat-tables.c
 	$(call rule_mkdir)
 	$(call if_changed_dep,cc_o_c)
 
-CFLAGS_intel-pt-insn-decoder.o += -I$(OUTPUT)util/intel-pt-decoder
+CFLAGS_inat.o += -I$(OUTPUT)util/intel-pt-decoder
+
+$(OUTPUT)util/intel-pt-decoder/insn.o: $(srctree)/tools/arch/x86/lib/insn.c
+	$(call rule_mkdir)
+	$(call if_changed_dep,cc_o_c)
 
 ifeq ($(CC_NO_CLANG), 1)
-  CFLAGS_intel-pt-insn-decoder.o += -Wno-override-init
+  CFLAGS_insn.o += -Wno-override-init
 endif
 
-CFLAGS_intel-pt-insn-decoder.o += -Wno-packed
+CFLAGS_insn.o += -Wno-packed
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
index 47cf35799a4d..8fabddc1c0da 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
@@ -11,9 +11,6 @@
 #include <byteswap.h>
 #include "../../../arch/x86/include/asm/insn.h"
 
-#include "../../../arch/x86/lib/inat.c"
-#include "../../../arch/x86/lib/insn.c"
-
 #include "event.h"
 
 #include "intel-pt-insn-decoder.h"
diff --git a/tools/perf/util/trace-event-scripting.c b/tools/perf/util/trace-event-scripting.c
index 2e9da0b089ef..8d71998a9ecd 100644
--- a/tools/perf/util/trace-event-scripting.c
+++ b/tools/perf/util/trace-event-scripting.c
@@ -272,11 +272,13 @@ void setup_perl_scripting(void)
 #endif
 #endif
 
-__weak void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
+#if !defined(__i386__) && !defined(__x86_64__)
+void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
 		     struct thread *thread __maybe_unused,
 		     struct machine *machine __maybe_unused)
 {
 }
+#endif
 
 void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
 		       struct machine *machine, bool native_arch)
-- 
2.47.0.163.g1226f6d8fa-goog
Re: [PATCH v5 11/21] perf x86: Define arch_fetch_insn in NO_AUXTRACE builds
Posted by Adrian Hunter 3 weeks, 3 days ago
On 31/10/24 03:42, Ian Rogers wrote:
> archinsn.c containing arch_fetch_insn was only enabled with
> CONFIG_AUXTRACE, but this meant that a NO_AUXTRACE build on x86 would
> use the empty weak version of arch_fetch_insn - weak symbols are a
> frequent source of errors like this and are outside of the C
> specification. Change it so that archinsn.c is always built on x86 and
> make the weak symbol empty version of arch_fetch_insn a strong one
> guarded by ifdefs.
> 
> arch_fetch_insn on x86 depends on insn_decode which is a function
> included then built into
> intel-pt-insn-decoder.c. intel-pt-insn-decoder.c isn't built in a
> NO_AUXTRACE=1 build. Separate the insn_decode function from
> intel-pt-insn-decoder.c by just directly compiling the relevant
> file. Guard this compilation to be for either always on x86 (because
> of the use in arch_fetch_insn) or when auxtrace is enabled. Apply the
> CFLAGS overrides as necessary, reducing the amount of code where
> warnings are disabled.
> 
> Signed-off-by: Ian Rogers <irogers@google.com>

For Intel PT

Tested-by: Adrian Hunter <adrian.hunter@intel.com>

> ---
>  tools/perf/arch/x86/util/Build                 |  2 +-
>  tools/perf/util/Build                          |  2 +-
>  tools/perf/util/intel-pt-decoder/Build         | 18 ++++++++++++++----
>  .../intel-pt-decoder/intel-pt-insn-decoder.c   |  3 ---
>  tools/perf/util/trace-event-scripting.c        |  4 +++-
>  5 files changed, 19 insertions(+), 10 deletions(-)
> 
> diff --git a/tools/perf/arch/x86/util/Build b/tools/perf/arch/x86/util/Build
> index bc56a8e70f34..c5df4a2cf180 100644
> --- a/tools/perf/arch/x86/util/Build
> +++ b/tools/perf/arch/x86/util/Build
> @@ -18,6 +18,6 @@ perf-util-$(CONFIG_LOCAL_LIBUNWIND)    += unwind-libunwind.o
>  perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
>  
>  perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
> -perf-util-$(CONFIG_AUXTRACE) += archinsn.o
> +perf-util-y += archinsn.o
>  perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
>  perf-util-$(CONFIG_AUXTRACE) += intel-bts.o
> diff --git a/tools/perf/util/Build b/tools/perf/util/Build
> index 650974413849..340544a6f5ec 100644
> --- a/tools/perf/util/Build
> +++ b/tools/perf/util/Build
> @@ -122,7 +122,7 @@ perf-util-y += iostat.o
>  perf-util-y += stream.o
>  perf-util-y += kvm-stat.o
>  perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
> -perf-util-$(CONFIG_AUXTRACE) += intel-pt-decoder/
> +perf-util-y += intel-pt-decoder/
>  perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
>  perf-util-$(CONFIG_AUXTRACE) += intel-bts.o
>  perf-util-$(CONFIG_AUXTRACE) += arm-spe.o
> diff --git a/tools/perf/util/intel-pt-decoder/Build b/tools/perf/util/intel-pt-decoder/Build
> index 30793d08c6d4..f99d150059b9 100644
> --- a/tools/perf/util/intel-pt-decoder/Build
> +++ b/tools/perf/util/intel-pt-decoder/Build
> @@ -9,14 +9,24 @@ $(OUTPUT)util/intel-pt-decoder/inat-tables.c: $(inat_tables_script) $(inat_table
>  
>  # Busybox's diff doesn't have -I, avoid warning in the case
>  
> -$(OUTPUT)util/intel-pt-decoder/intel-pt-insn-decoder.o: util/intel-pt-decoder/intel-pt-insn-decoder.c $(OUTPUT)util/intel-pt-decoder/inat-tables.c
> +ifeq ($(SRCARCH),x86)
> +  perf-util-y += inat.o insn.o
> +else
> +  perf-util-$(CONFIG_AUXTRACE) += inat.o insn.o
> +endif
> +
> +$(OUTPUT)util/intel-pt-decoder/inat.o: $(srctree)/tools/arch/x86/lib/inat.c $(OUTPUT)util/intel-pt-decoder/inat-tables.c
>  	$(call rule_mkdir)
>  	$(call if_changed_dep,cc_o_c)
>  
> -CFLAGS_intel-pt-insn-decoder.o += -I$(OUTPUT)util/intel-pt-decoder
> +CFLAGS_inat.o += -I$(OUTPUT)util/intel-pt-decoder
> +
> +$(OUTPUT)util/intel-pt-decoder/insn.o: $(srctree)/tools/arch/x86/lib/insn.c
> +	$(call rule_mkdir)
> +	$(call if_changed_dep,cc_o_c)
>  
>  ifeq ($(CC_NO_CLANG), 1)
> -  CFLAGS_intel-pt-insn-decoder.o += -Wno-override-init
> +  CFLAGS_insn.o += -Wno-override-init
>  endif
>  
> -CFLAGS_intel-pt-insn-decoder.o += -Wno-packed
> +CFLAGS_insn.o += -Wno-packed
> diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
> index 47cf35799a4d..8fabddc1c0da 100644
> --- a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
> +++ b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
> @@ -11,9 +11,6 @@
>  #include <byteswap.h>
>  #include "../../../arch/x86/include/asm/insn.h"
>  
> -#include "../../../arch/x86/lib/inat.c"
> -#include "../../../arch/x86/lib/insn.c"
> -
>  #include "event.h"
>  
>  #include "intel-pt-insn-decoder.h"
> diff --git a/tools/perf/util/trace-event-scripting.c b/tools/perf/util/trace-event-scripting.c
> index 2e9da0b089ef..8d71998a9ecd 100644
> --- a/tools/perf/util/trace-event-scripting.c
> +++ b/tools/perf/util/trace-event-scripting.c
> @@ -272,11 +272,13 @@ void setup_perl_scripting(void)
>  #endif
>  #endif
>  
> -__weak void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
> +#if !defined(__i386__) && !defined(__x86_64__)
> +void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
>  		     struct thread *thread __maybe_unused,
>  		     struct machine *machine __maybe_unused)
>  {
>  }
> +#endif
>  
>  void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
>  		       struct machine *machine, bool native_arch)