Skip to contentSkip to conversation

perf tools: fix incorrect header string

PATCH In review Wang Nan 2 participants 5 messages Oct 15, 2014 – Oct 21, 2014 perf events

lore
Share & tools
lore

Wang Nan

author
#0 lore raw

PATCH 1/1 perf tools: fix incorrect header string

Commit fbe96f29 (perf tools: Make perf.data more self-descriptive) read '/proc/cpuinfo' to form cpu descriptor. For ARM, it finds 'Processor' field. It is correct when the patch merged, but due to commit b4b8f770 (ARM: kernel: update cpuinfo to print all online CPUs features), the corresponding information becomes 'model name' field.

This patch simply corrects it.

---
 tools/perf/perf-sys.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/perf-sys.h +1 -1
index 937e432..4293970 100644
@@ -113,7 +113,7 @@
 #define mb()		((void(*)(void))0xffff0fa0)()
 #define wmb()		((void(*)(void))0xffff0fa0)()
 #define rmb()		((void(*)(void))0xffff0fa0)()
-#define CPUINFO_PROC	"Processor"
+#define CPUINFO_PROC	"model name"
 #endif
 
 #ifdef __aarch64__
Signature · 1 line
1.8.4
#1 lore raw

Em Wed, Oct 15, 2014 at 11:28:53AM +0800, Wang Nan escreveu:

Quoted 7 lines

Commit fbe96f29 (perf tools: Make perf.data more self-descriptive) read '/proc/cpuinfo' to form cpu descriptor. For ARM, it finds 'Processor' field. It is correct when the patch merged, but due to commit b4b8f770 (ARM: kernel: update cpuinfo to print all online CPUs features), the corresponding information becomes 'model name' field.

This patch simply corrects it.

It doesn't :-)

It makes it work with kernels >= b4b8f770, and fail with older kernels.

We need to somehow figure out where the right information is regardless of the kernel.

- Arnaldo

Quoted 20 lines
---
 tools/perf/perf-sys.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/perf-sys.h +1 -1
index 937e432..4293970 100644
@@ -113,7 +113,7 @@
 #define mb()		((void(*)(void))0xffff0fa0)()
 #define wmb()		((void(*)(void))0xffff0fa0)()
 #define rmb()		((void(*)(void))0xffff0fa0)()
-#define CPUINFO_PROC	"Processor"
+#define CPUINFO_PROC	"model name"
 #endif
 
 #ifdef __aarch64__
Signature · 1 line
1.8.4
author
#2 lore raw
Quoted 9 lines · Arnaldo Carvalho de Melo

Em Wed, Oct 15, 2014 at 11:28:53AM +0800, Wang Nan escreveu:

Quoted 7 lines

Commit fbe96f29 (perf tools: Make perf.data more self-descriptive) read '/proc/cpuinfo' to form cpu descriptor. For ARM, it finds 'Processor' field. It is correct when the patch merged, but due to commit b4b8f770 (ARM: kernel: update cpuinfo to print all online CPUs features), the corresponding information becomes 'model name' field.

This patch simply corrects it.

Quoted 9 lines

It doesn't :-)

It makes it work with kernels >= b4b8f770, and fail with older kernels.

We need to somehow figure out where the right information is regardless of the kernel.

- Arnaldo

However it is a real problem. Look at output on arm platform at the bottom of this mail, especially "cpudesc", "total memory" and "cmdline" field.

By further debugging I found that the real problem resides in return value checking when write header error. Please see my other patches:

https://lkml.org/lkml/2014/10/15/612
https://lkml.org/lkml/2014/10/15/611

I think at least one of them is required.

Thanks.

Quoted 17 lines
Quoted 17 lines
---
 tools/perf/perf-sys.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/perf-sys.h +1 -1
index 937e432..4293970 100644
@@ -113,7 +113,7 @@
 #define mb()		((void(*)(void))0xffff0fa0)()
 #define wmb()		((void(*)(void))0xffff0fa0)()
 #define rmb()		((void(*)(void))0xffff0fa0)()
-#define CPUINFO_PROC	"Processor"
+#define CPUINFO_PROC	"model name"
 #endif
 
Quoted 3 lines
Quoted 3 lines
tools/perf/perf-sys.h +0 -0
 #ifdef __aarch64__
Signature · 1 line
1.8.4

bash-4.2# perf record ls
...
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.001 MB perf.data (~36 samples) ] bash-4.2# perf report --stdio --header
Error:
The perf.data file has no samples!
# ========

# captured on: Fri Sep 12 10:09:10 2014
# hostname : arma15el
# os release : 3.17.0+
# perf version : 3.10.53
# arch : armv7l
# nrcpus online : 4
# nrcpus avail : 1
# cpudesc : (null)
# total memory : 0 kB
# cmdline :
# event : name = cycles, type = 0, config = 0x0, config1 = 0x0, config2 = 0x0, excl_usr = 0, excl_kern = 0, excl_host = 0, excl_guest = 1, precise_ip = 0
# pmu mappings: not available
# ========
#
#3 lore raw

Em Thu, Oct 16, 2014 at 11:21:13AM +0800, Wang Nan escreveu:

Quoted 14 lines · Arnaldo Carvalho de Melo
Quoted 13 lines

Em Wed, Oct 15, 2014 at 11:28:53AM +0800, Wang Nan escreveu:

Quoted 7 lines

Commit fbe96f29 (perf tools: Make perf.data more self-descriptive) read '/proc/cpuinfo' to form cpu descriptor. For ARM, it finds 'Processor' field. It is correct when the patch merged, but due to commit b4b8f770 (ARM: kernel: update cpuinfo to print all online CPUs features), the corresponding information becomes 'model name' field.

This patch simply corrects it.

It doesn't :-)

It makes it work with kernels >= b4b8f770, and fail with older kernels.

Quoted 8 lines
Quoted 5 lines

We need to somehow figure out where the right information is regardless of the kernel.

- Arnaldo

However it is a real problem. Look at output on arm platform at the bottom of this mail, especially "cpudesc", "total memory" and "cmdline" field.

Yeah, I haven't said it wasn't a problem, just that the proposed fix wasn't enough.

- Arnaldo

Quoted 57 lines

By further debugging I found that the real problem resides in return value checking when write header error. Please see my other patches:

https://lkml.org/lkml/2014/10/15/612
https://lkml.org/lkml/2014/10/15/611

I think at least one of them is required.

Thanks.

Quoted 17 lines
Quoted 17 lines
---
 tools/perf/perf-sys.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/perf-sys.h +1 -1
index 937e432..4293970 100644
@@ -113,7 +113,7 @@
 #define mb()		((void(*)(void))0xffff0fa0)()
 #define wmb()		((void(*)(void))0xffff0fa0)()
 #define rmb()		((void(*)(void))0xffff0fa0)()
-#define CPUINFO_PROC	"Processor"
+#define CPUINFO_PROC	"model name"
 #endif
 
Quoted 3 lines
Quoted 3 lines
tools/perf/perf-sys.h +0 -0
 #ifdef __aarch64__
Signature · 1 line
1.8.4

bash-4.2# perf record ls
...
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.001 MB perf.data (~36 samples) ] bash-4.2# perf report --stdio --header
Error:
The perf.data file has no samples!
# ========

# captured on: Fri Sep 12 10:09:10 2014
# hostname : arma15el
# os release : 3.17.0+
# perf version : 3.10.53
# arch : armv7l
# nrcpus online : 4
# nrcpus avail : 1
# cpudesc : (null)
# total memory : 0 kB
# cmdline :
# event : name = cycles, type = 0, config = 0x0, config1 = 0x0, config2 = 0x0, excl_usr = 0, excl_kern = 0, excl_host = 0, excl_guest = 1, precise_ip = 0
# pmu mappings: not available
# ========
#
author
#4 lore raw
Quoted 21 lines · Arnaldo Carvalho de Melo

Em Thu, Oct 16, 2014 at 11:21:13AM +0800, Wang Nan escreveu:

Quoted 20 lines · Arnaldo Carvalho de Melo
Quoted 18 lines

Em Wed, Oct 15, 2014 at 11:28:53AM +0800, Wang Nan escreveu:

Quoted 7 lines

Commit fbe96f29 (perf tools: Make perf.data more self-descriptive) read '/proc/cpuinfo' to form cpu descriptor. For ARM, it finds 'Processor' field. It is correct when the patch merged, but due to commit b4b8f770 (ARM: kernel: update cpuinfo to print all online CPUs features), the corresponding information becomes 'model name' field.

This patch simply corrects it.

It doesn't :-)

It makes it work with kernels >= b4b8f770, and fail with older kernels.

We need to somehow figure out where the right information is regardless of the kernel.

- Arnaldo

Quoted 13 lines
Quoted 2 lines

However it is a real problem. Look at output on arm platform at the bottom of this mail, especially "cpudesc", "total memory" and "cmdline" field.

Yeah, I haven't said it wasn't a problem, just that the proposed fix wasn't enough.

- Arnaldo

Quoted 5 lines

By further debugging I found that the real problem resides in return value checking when write header error. Please see my other patches:

https://lkml.org/lkml/2014/10/15/612
https://lkml.org/lkml/2014/10/15/611

So what's your opinion on these two patches?

Quoted 52 lines
Quoted 52 lines

I think at least one of them is required.

Thanks.

Quoted 17 lines
Quoted 17 lines
---
 tools/perf/perf-sys.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/perf-sys.h +1 -1
index 937e432..4293970 100644
@@ -113,7 +113,7 @@
 #define mb()		((void(*)(void))0xffff0fa0)()
 #define wmb()		((void(*)(void))0xffff0fa0)()
 #define rmb()		((void(*)(void))0xffff0fa0)()
-#define CPUINFO_PROC	"Processor"
+#define CPUINFO_PROC	"model name"
 #endif
 
Quoted 3 lines
Quoted 3 lines
tools/perf/perf-sys.h +0 -0
 #ifdef __aarch64__
Signature · 1 line
1.8.4

bash-4.2# perf record ls
...
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.001 MB perf.data (~36 samples) ] bash-4.2# perf report --stdio --header
Error:
The perf.data file has no samples!
# ========

# captured on: Fri Sep 12 10:09:10 2014
# hostname : arma15el
# os release : 3.17.0+
# perf version : 3.10.53
# arch : armv7l
# nrcpus online : 4
# nrcpus avail : 1
# cpudesc : (null)
# total memory : 0 kB
# cmdline :
# event : name = cycles, type = 0, config = 0x0, config1 = 0x0, config2 = 0x0, excl_usr = 0, excl_kern = 0, excl_host = 0, excl_guest = 1, precise_ip = 0
# pmu mappings: not available
# ========
#