oprofile: ARM backtraces and few minors

This patchset re-enables ARM backtraces with non-PMU capable CPUs.

In V2 I have tried to address the comments received for the V1 patches. arch/sh/oprofile/common.c is untouched but should work as earlier.

Compile tested with arm/omap2plus_defconfig (with and without CONFIG_HW_PERF_EVENTS).

Ari Kauppi (3):
oprofile: Use CONFIG_HW_PERF_EVENTS for oprofile_perf_init and
friends
oprofile: Provide stubs for oprofile_perf_{init,exit}
ARM: oprofile: Always allow backtraces

 arch/arm/oprofile/common.c |   12 ++----------
 include/linux/oprofile.h   |   13 +++++++++++--
 2 files changed, 13 insertions(+), 12 deletions(-)

1.7.4.rc2.3.g60a2e

The implementations are flagged in Makefile with CONFIG_HW_PERF_EVENTS instead of CONFIG_PERF_EVENTS.

---
 include/linux/oprofile.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

index 32fb812..54c8382 100644
@@ -186,10 +186,10 @@ int oprofile_add_data(struct op_entry *entry, unsigned long val);
 int oprofile_add_data64(struct op_entry *entry, u64 val);
 int oprofile_write_commit(struct op_entry *entry);
 
-#ifdef CONFIG_PERF_EVENTS
+#ifdef CONFIG_HW_PERF_EVENTS
 int __init oprofile_perf_init(struct oprofile_operations *ops);
 void oprofile_perf_exit(void);
 char *op_name_from_perf_id(void);
-#endif /* CONFIG_PERF_EVENTS */
+#endif /* CONFIG_HW_PERF_EVENTS */
 
 #endif /* OPROFILE_H */

1.7.4.rc2.3.g60a2e

If CONFIG_HW_PERF_EVENTS is not defined, provide stubs for oprofile_perf_init and oprofile_perf_exit. Allows cleaning of architecture specific implementations.

---
 include/linux/oprofile.h |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

index 54c8382..1ca6411 100644
@@ -16,6 +16,8 @@
 #include <linux/types.h>
 #include <linux/spinlock.h>
 #include <linux/init.h>
+#include <linux/errno.h>
+#include <linux/printk.h>
 #include <asm/atomic.h>
  
 /* Each escaped entry is prefixed by ESCAPE_CODE

@@ -190,6 +192,13 @@ int oprofile_write_commit(struct op_entry *entry);
 int __init oprofile_perf_init(struct oprofile_operations *ops);
 void oprofile_perf_exit(void);
 char *op_name_from_perf_id(void);
+#else
+static inline int __init oprofile_perf_init(struct oprofile_operations *ops)
+{
+	pr_info("oprofile: hardware counters not available\n");
+	return -ENODEV;
+}
+static inline void oprofile_perf_exit(void) { }
 #endif /* CONFIG_HW_PERF_EVENTS */
 
 #endif /* OPROFILE_H */

1.7.4.rc2.3.g60a2e

Always allow backtrace when using oprofile on ARM, even if a PMU isn't present. Implemented by removing code and using stubs provided by <linux/oprofile.h>.

Restores functionality originally introduced in commit 1b7b56982fdcd9d85effd76f3928cf5d6eb26155 ("oprofile: Always allow backtraces on ARM") by Richard Purdie.

---
 arch/arm/oprofile/common.c |   12 ++----------
 1 files changed, 2 insertions(+), 10 deletions(-)

index 8aa9744..2b66391 100644
@@ -10,8 +10,6 @@
  */
 
 #include <linux/cpumask.h>
-#include <linux/err.h>
-#include <linux/errno.h>
 #include <linux/init.h>
 #include <linux/mutex.h>
 #include <linux/oprofile.h>

@@ -46,6 +44,7 @@ char *op_name_from_perf_id(void)
 		return NULL;
 	}
 }
+#endif
 
 static int report_trace(struct stackframe *frame, void *d)
 {

@@ -111,6 +110,7 @@ static void arm_backtrace(struct pt_regs * const regs, unsigned int depth)
 
 int __init oprofile_arch_init(struct oprofile_operations *ops)
 {
+	/* provide backtrace support also in timer mode: */
 	ops->backtrace		= arm_backtrace;
 
 	return oprofile_perf_init(ops);

@@ -120,11 +120,3 @@ void __exit oprofile_arch_exit(void)
 {
 	oprofile_perf_exit();
 }
-#else
-int __init oprofile_arch_init(struct oprofile_operations *ops)
-{
-	pr_info("oprofile: hardware counters not available\n");
-	return -ENODEV;
-}
-void __exit oprofile_arch_exit(void) {}
-#endif /* CONFIG_HW_PERF_EVENTS */

1.7.4.rc2.3.g60a2e

Hi Ari,

[...]

Providing Robert is happy with the associated changes to linux/oprofile.h then this patch looks good to me.

Will

Hi Ari,

[...]

Providing Robert is happy with the associated changes to linux/oprofile.h then this patch looks good to me.

Will

Applied to oprofile/urgent, thanks Ari.

I squashed patch 2 and 3 and modified the commit messages (see below).

Added to .37-stable.

-Robert

commit d14dd7e20d5e526557f5d3cfef4046a642f80924

Author: Ari Kauppi <[email protected]>
Date:   Thu Jan 20 13:57:19 2011 -0500

ARM: oprofile: Fix backtraces in timer mode

Always allow backtraces when using oprofile on ARM, even if a PMU isn't present. Restores functionality originally introduced in commit 1b7b56982fdcd9d85effd76f3928cf5d6eb26155 ("oprofile: Always allow backtraces on ARM") by Richard Purdie.

It is not that obvious, but there is now only one oprofile_arch_init() function. So the .backtrace callback is available also in timer mode.

Implemented by removing code and using stubs for oprofile_perf_{init, exit} provided by <linux/oprofile.h>. This allows cleaning of other architecture specific implementations too.

commit 1ea1bdf7faa4d0b5293e605f2e1ef1c2c59f6b53

Author: Ari Kauppi <[email protected]>
Date:   Thu Jan 20 13:57:18 2011 -0500

oprofile: Fix usage of CONFIG_HW_PERF_EVENTS for oprofile_perf_init and friends

The implementations are flagged in Makefile with CONFIG_HW_PERF_EVENTS instead of CONFIG_PERF_EVENTS.

Advanced Micro Devices, Inc.
Operating System Research Center