Skip to contentSkip to conversation

tools/perf: Add loongarch cpuinfo and fix build warning

PATCH Superseded Yanteng Si 3 participants 11 messages Aug 22, 2023 – Aug 24, 2023

Subsystems (2)

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Yanteng Si

author
#0 lore raw

cover tools/perf: Add loongarch cpuinfo and fix build warning

* LoongArch and riscv don't have tools/arch/xxxxx/include/uapi/asm/mman.h, let's modify the detection rules.
* Allow to use cpuinfo on loongarch.

Yanteng Si (2):
perf/trace: Fix mmap_flags for archs use generic mman.h perf/tools: Allow to use cpuinfo on loongarch

 tools/perf/trace/beauty/mmap_flags.sh | 2 +-
 tools/perf/util/header.c              | 2 ++
 tools/perf/util/svghelper.c           | 7 ++++++-
 3 files changed, 9 insertions(+), 2 deletions(-)
Signature · 1 line
2.31.4
author
#1 lore raw

PATCH 1/2 perf/trace: Fix mmap_flags for archs use generic mman.h

Silence this warning:

grep: /root/linux-next/tools/arch/xxxxx/include/uapi/asm//mman.h:
No such file or directory

---
 tools/perf/trace/beauty/mmap_flags.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/trace/beauty/mmap_flags.sh +1 -1
index 3022597c8c17..d035b29fc5ec 100755
@@ -19,7 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
 
 printf "static const char *mmap_flags[] = {\n"
 regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
-grep -E -q $regex ${arch_mman} && \
+([ -f ${arch_mman} ] && grep -E -q $regex ${arch_mman}) && \
 (grep -E $regex ${arch_mman} | \
 	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
 	xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n")
Signature · 1 line
2.31.4
#2 lore raw

Re: [PATCH v1 1/2] perf/trace: Fix mmap_flags for archs use generic mman.h

Em Tue, Aug 22, 2023 at 05:32:06PM +0800, Yanteng Si escreveu:

Quoted 16 lines

Silence this warning:

grep: /root/linux-next/tools/arch/xxxxx/include/uapi/asm//mman.h:
No such file or directory

---
 tools/perf/trace/beauty/mmap_flags.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c17..d035b29fc5ec 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 25

Quoted 8 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -1
@@ -19,7 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
 
 printf "static const char *mmap_flags[] = {\n"
 regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
-grep -E -q $regex ${arch_mman} && \
+([ -f ${arch_mman} ] && grep -E -q $regex ${arch_mman}) && \
 (grep -E $regex ${arch_mman} | \
 	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
 	xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n")

Humm, wouldn't this be simpler without that extra subshell? Like this:

tools/perf/trace/beauty/mmap_flags.sh +1 -0
index 3022597c8c178f21..0ac1886a83b5aceb 100755
@@ -19,6 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
 
 printf "static const char *mmap_flags[] = {\n"
 regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
+test -f ${arch_mman} && \
 grep -E -q $regex ${arch_mman} && \
 (grep -E $regex ${arch_mman} | \
 	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
author
#3 lore raw

Re: [PATCH v1 1/2] perf/trace: Fix mmap_flags for archs use generic mman.h

Quoted 31 lines · Arnaldo Carvalho de Melo

Em Tue, Aug 22, 2023 at 05:32:06PM +0800, Yanteng Si escreveu:

Quoted 15 lines

Silence this warning:

grep: /root/linux-next/tools/arch/xxxxx/include/uapi/asm//mman.h:
No such file or directory

---
  tools/perf/trace/beauty/mmap_flags.sh | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c17..d035b29fc5ec 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 25

Quoted 8 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -1
@@ -19,7 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
 
  printf "static const char *mmap_flags[] = {\n"
  regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
-grep -E -q $regex ${arch_mman} && \
+([ -f ${arch_mman} ] && grep -E -q $regex ${arch_mman}) && \
  (grep -E $regex ${arch_mman} | \
  	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
  	xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n")

Humm, wouldn't this be simpler without that extra subshell? Like this:

tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c178f21..0ac1886a83b5aceb 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 22

Quoted 4 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -0
@@ -19,6 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
 
  printf "static const char *mmap_flags[] = {\n"
  regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
+test -f ${arch_mman} && \

The following lines have already used the [ ! -f ${arch_mman}, so let's stay consistent?

Thanks,

Yanteng

Quoted 3 lines
tools/perf/trace/beauty/mmap_flags.sh +0 -0
  grep -E -q $regex ${arch_mman} && \
  (grep -E $regex ${arch_mman} | \
  	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
#4 lore raw

Re: [PATCH v1 1/2] perf/trace: Fix mmap_flags for archs use generic mman.h

Em Wed, Aug 23, 2023 at 03:11:01PM +0800, Yanteng Si escreveu:

Quoted 26 lines · Arnaldo Carvalho de Melo
Quoted 24 lines

Em Tue, Aug 22, 2023 at 05:32:06PM +0800, Yanteng Si escreveu:

Quoted 14 lines

Silence this warning:

grep: /root/linux-next/tools/arch/xxxxx/include/uapi/asm//mman.h:
No such file or directory

---
  tools/perf/trace/beauty/mmap_flags.sh | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c17..d035b29fc5ec 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 24

Quoted 7 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -1
@@ -19,7 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
  printf "static const char *mmap_flags[] = {\n"
  regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
-grep -E -q $regex ${arch_mman} && \
+([ -f ${arch_mman} ] && grep -E -q $regex ${arch_mman}) && \
  (grep -E $regex ${arch_mman} | \
  	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
  	xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n")

Humm, wouldn't this be simpler without that extra subshell? Like this:

Quoted 11 lines

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 21

Quoted 8 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -0
index 3022597c8c178f21..0ac1886a83b5aceb 100755
@@ -19,6 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
  printf "static const char *mmap_flags[] = {\n"
  regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
+test -f ${arch_mman} && \

The following lines have already used the [ ! -f ${arch_mman}, so let's stay consistent?

Well, we can make it consistent by simplifying the other places where the long, equivalent test is being done, which I think is the way to go, agreed?

- Arnaldo

Quoted 9 lines

Thanks,

Yanteng

Quoted 3 lines
tools/perf/trace/beauty/mmap_flags.sh +0 -0
  grep -E -q $regex ${arch_mman} && \
  (grep -E $regex ${arch_mman} | \
  	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
Signature · 1 line
- Arnaldo
author
#5 lore raw

Re: [PATCH v1 1/2] perf/trace: Fix mmap_flags for archs use generic mman.h

Quoted 36 lines · Arnaldo Carvalho de Melo

Em Wed, Aug 23, 2023 at 03:11:01PM +0800, Yanteng Si escreveu:

Quoted 25 lines · Arnaldo Carvalho de Melo
Quoted 24 lines

Em Tue, Aug 22, 2023 at 05:32:06PM +0800, Yanteng Si escreveu:

Quoted 14 lines

Silence this warning:

grep: /root/linux-next/tools/arch/xxxxx/include/uapi/asm//mman.h:
No such file or directory

---
   tools/perf/trace/beauty/mmap_flags.sh | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c17..d035b29fc5ec 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 24

Quoted 7 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -1
@@ -19,7 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
   printf "static const char *mmap_flags[] = {\n"
   regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
-grep -E -q $regex ${arch_mman} && \
+([ -f ${arch_mman} ] && grep -E -q $regex ${arch_mman}) && \
   (grep -E $regex ${arch_mman} | \
   	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
   	xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n")

Humm, wouldn't this be simpler without that extra subshell? Like this:

Quoted 10 lines

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 21

Quoted 8 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -0
index 3022597c8c178f21..0ac1886a83b5aceb 100755
@@ -19,6 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
   printf "static const char *mmap_flags[] = {\n"
   regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
+test -f ${arch_mman} && \

The following lines have already used the [ ! -f ${arch_mman}, so let's stay consistent?

Quoted 3 lines

Well, we can make it consistent by simplifying the other places where the long, equivalent test is being done, which I think is the way to go, agreed?

I agree, but this is not suitable to be done in the fix patch, I will split it out and make a new patch.

Thanks

Yanteng

Quoted 10 lines

- Arnaldo

Quoted 7 lines

Thanks,

Yanteng

Quoted 3 lines
tools/perf/trace/beauty/mmap_flags.sh +0 -0
   grep -E -q $regex ${arch_mman} && \
   (grep -E $regex ${arch_mman} | \
   	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
#6 lore raw

Re: [PATCH v1 1/2] perf/trace: Fix mmap_flags for archs use generic mman.h

Em Thu, Aug 24, 2023 at 10:22:57AM +0800, Yanteng Si escreveu:

Quoted 38 lines · Arnaldo Carvalho de Melo
Quoted 36 lines

Em Wed, Aug 23, 2023 at 03:11:01PM +0800, Yanteng Si escreveu:

Quoted 35 lines · Arnaldo Carvalho de Melo
Quoted 29 lines

Em Tue, Aug 22, 2023 at 05:32:06PM +0800, Yanteng Si escreveu:

Quoted 14 lines

Silence this warning:

grep: /root/linux-next/tools/arch/xxxxx/include/uapi/asm//mman.h:
No such file or directory

---
   tools/perf/trace/beauty/mmap_flags.sh | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c17..d035b29fc5ec 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 24

Quoted 7 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -1
@@ -19,7 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
   printf "static const char *mmap_flags[] = {\n"
   regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
-grep -E -q $regex ${arch_mman} && \
+([ -f ${arch_mman} ] && grep -E -q $regex ${arch_mman}) && \
   (grep -E $regex ${arch_mman} | \
   	sed -r "s/$regex/\2 \1 \1 \1 \2/g"	| \
   	xargs printf "\t[ilog2(%s) + 1] = \"%s\",\n#ifndef MAP_%s\n#define MAP_%s %s\n#endif\n")

Humm, wouldn't this be simpler without that extra subshell? Like this:

tools/perf/trace/beauty/mmap_flags.sh +0 -0
index 3022597c8c178f21..0ac1886a83b5aceb 100755

tools/perf/trace/beauty/mmap_flags.sh arch_mman=${arch_header_dir}/mman.h · line 21

Quoted 3 lines
tools/perf/trace/beauty/mmap_flags.sh +1 -0
@@ -19,6 +19,7 @@ arch_mman=${arch_header_dir}/mman.h
   printf "static const char *mmap_flags[] = {\n"
   regex='^[[:space:]]*#[[:space:]]*define[[:space:]]+MAP_([[:alnum:]_]+)[[:space:]]+(0x[[:xdigit:]]+)[[:space:]]*.*'
+test -f ${arch_mman} && \

The following lines have already used the [ ! -f ${arch_mman}, so let's stay consistent?

Quoted 6 lines
Quoted 3 lines

Well, we can make it consistent by simplifying the other places where the long, equivalent test is being done, which I think is the way to go, agreed?

I agree, but this is not suitable to be done in the fix patch, I will split it out and make a new patch.

Sure, send the one with the fix, already using the simple form, then convert the rest to the simpler form.

Thanks,

- Arnaldo

author
#7 lore raw

PATCH 2/2 perf/tools: Allow to use cpuinfo on loongarch

In reply to Yanteng Si

Define these macros so that the CPU name can be displayed when running perf report and perf timechart.

---
 tools/perf/util/header.c    | 2 ++
 tools/perf/util/svghelper.c | 7 ++++++-
 2 files changed, 8 insertions(+), 1 deletion(-)
tools/perf/util/header.c +2 -0
index 52fbf526fe74..c6e19192bd28 100644
@@ -456,6 +456,8 @@ static int write_cpudesc(struct feat_fd *ff,
 #define CPUINFO_PROC	{ "Processor", }
 #elif defined(__xtensa__)
 #define CPUINFO_PROC	{ "core ID", }
+#elif defined(__loongarch__)
+#define CPUINFO_PROC	{ "Model Name", }
 #else
 #define CPUINFO_PROC	{ "model name", }
 #endif
tools/perf/util/svghelper.c +6 -1
index 5c62d3118c41..0259715ad6a6 100644
@@ -322,6 +322,11 @@ void svg_waiting(int Yslot, int cpu, u64 start, u64 end, const char *backtrace)
 
 static char *cpu_model(void)
 {
+#if defined(__loongarch__)
+#define CPU_TYPE	"Model Name"
+#else
+#define CPU_TYPE	"model name"
+#endif
 	static char cpu_m[255];
 	char buf[256];
 	FILE *file;
@@ -331,7 +336,7 @@ static char *cpu_model(void)
 	file = fopen("/proc/cpuinfo", "r");
 	if (file) {
 		while (fgets(buf, 255, file)) {
-			if (strstr(buf, "model name")) {
+			if (strstr(buf, CPU_TYPE)) {
 				strlcpy(cpu_m, &buf[13], 255);
 				break;
 			}
Signature · 1 line
2.31.4
#8 lore raw

Re: [PATCH v1 2/2] perf/tools: Allow to use cpuinfo on loongarch

Em Tue, Aug 22, 2023 at 05:32:07PM +0800, Yanteng Si escreveu:

Quoted 39 lines

Define these macros so that the CPU name can be displayed when running perf report and perf timechart.

---
 tools/perf/util/header.c    | 2 ++
 tools/perf/util/svghelper.c | 7 ++++++-
 2 files changed, 8 insertions(+), 1 deletion(-)
tools/perf/util/header.c +2 -0
index 52fbf526fe74..c6e19192bd28 100644
@@ -456,6 +456,8 @@ static int write_cpudesc(struct feat_fd *ff,
 #define CPUINFO_PROC	{ "Processor", }
 #elif defined(__xtensa__)
 #define CPUINFO_PROC	{ "core ID", }
+#elif defined(__loongarch__)
+#define CPUINFO_PROC	{ "Model Name", }
 #else
 #define CPUINFO_PROC	{ "model name", }
 #endif
tools/perf/util/svghelper.c +5 -0
index 5c62d3118c41..0259715ad6a6 100644
@@ -322,6 +322,11 @@ void svg_waiting(int Yslot, int cpu, u64 start, u64 end, const char *backtrace)
 
 static char *cpu_model(void)
 {
+#if defined(__loongarch__)
+#define CPU_TYPE	"Model Name"
+#else
+#define CPU_TYPE	"model name"
+#endif
 	static char cpu_m[255];
 	char buf[256];
 	FILE *file;

tools/perf/util/svghelper.c static char *cpu_model(void) · line 339

Quoted 5 lines
tools/perf/util/svghelper.c +1 -1
@@ -331,7 +336,7 @@ static char *cpu_model(void)
 	file = fopen("/proc/cpuinfo", "r");
 	if (file) {
 		while (fgets(buf, 255, file)) {
-			if (strstr(buf, "model name")) {
+			if (strstr(buf, CPU_TYPE)) {

Isn't it better to just switch to strcasestr()?

- Arnaldo

Quoted 6 lines
tools/perf/util/svghelper.c +0 -0
 				strlcpy(cpu_m, &buf[13], 255);
 				break;
 			}
Signature · 1 line
2.31.4
Signature · 1 line
- Arnaldo
author
#9 lore raw

Re: [PATCH v1 2/2] perf/tools: Allow to use cpuinfo on loongarch

Quoted 38 lines · Arnaldo Carvalho de Melo

Em Tue, Aug 22, 2023 at 05:32:07PM +0800, Yanteng Si escreveu:

Quoted 37 lines

Define these macros so that the CPU name can be displayed when running perf report and perf timechart.

---
  tools/perf/util/header.c    | 2 ++
  tools/perf/util/svghelper.c | 7 ++++++-
  2 files changed, 8 insertions(+), 1 deletion(-)
tools/perf/util/header.c +2 -0
index 52fbf526fe74..c6e19192bd28 100644
@@ -456,6 +456,8 @@ static int write_cpudesc(struct feat_fd *ff,
  #define CPUINFO_PROC	{ "Processor", }
  #elif defined(__xtensa__)
  #define CPUINFO_PROC	{ "core ID", }
+#elif defined(__loongarch__)
+#define CPUINFO_PROC	{ "Model Name", }
  #else
  #define CPUINFO_PROC	{ "model name", }
  #endif
tools/perf/util/svghelper.c +5 -0
index 5c62d3118c41..0259715ad6a6 100644
@@ -322,6 +322,11 @@ void svg_waiting(int Yslot, int cpu, u64 start, u64 end, const char *backtrace)
 
  static char *cpu_model(void)
  {
+#if defined(__loongarch__)
+#define CPU_TYPE	"Model Name"
+#else
+#define CPU_TYPE	"model name"
+#endif
  	static char cpu_m[255];
  	char buf[256];
Quoted 8 lines

tools/perf/util/svghelper.c static char *cpu_model(void) · line 339

Quoted 7 lines
tools/perf/util/svghelper.c +1 -1
  	FILE *file;
@@ -331,7 +336,7 @@ static char *cpu_model(void)
  	file = fopen("/proc/cpuinfo", "r");
  	if (file) {
  		while (fgets(buf, 255, file)) {
-			if (strstr(buf, "model name")) {
+			if (strstr(buf, CPU_TYPE)) {

Isn't it better to just switch to strcasestr()?

OK!

Thanks,

Yanteng

Quoted 9 lines

- Arnaldo

Quoted 6 lines
tools/perf/util/svghelper.c +0 -0
  				strlcpy(cpu_m, &buf[13], 255);
  				break;
  			}
Signature · 1 line
2.31.4
maintainer Acked-by on the v1 series
#10 lore raw
In reply to Yanteng Si

For this series:

Quoted 18 lines · Yanteng Si

* LoongArch and riscv don't have tools/arch/xxxxx/include/uapi/asm/mman.h, let's modify the detection rules.
* Allow to use cpuinfo on loongarch.

Yanteng Si (2):
perf/trace: Fix mmap_flags for archs use generic mman.h perf/tools: Allow to use cpuinfo on loongarch

 tools/perf/trace/beauty/mmap_flags.sh | 2 +-
 tools/perf/util/header.c              | 2 ++
 tools/perf/util/svghelper.c           | 7 ++++++-
 3 files changed, 9 insertions(+), 2 deletions(-)

--
2.31.4