[PATCH 0/4] perf_counter: fixes and an optimization

[PATCH 0/4] perf_counter: fixes and an optimization

[Peter Zijlstra]

 - fixes an off-by-one in string handling
 - fixes the mmap tracing to provide the full path
 - allows unpriviledged use of cpu counters
 - optimize mmap/comm tracing

-- 

[PATCH 1/4] perf_counter: fix off-by-one on the string lengths

[Peter Zijlstra]

[-- Attachment #1: perf_counter-fix-strings.patch --]
[-- Type: text/plain, Size: 891 bytes --]

strlen() does _not_ include the terminating \0.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/perf_counter.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Index: linux-2.6/kernel/perf_counter.c
===================================================================
--- linux-2.6.orig/kernel/perf_counter.c
+++ linux-2.6/kernel/perf_counter.c
@@ -1989,7 +1989,7 @@ static void perf_counter_comm_event(stru
 	unsigned int size;
 	char *comm = comm_event->task->comm;
 
-	size = ALIGN(strlen(comm), sizeof(u64));
+	size = ALIGN(strlen(comm)+1, sizeof(u64));
 
 	comm_event->comm = comm;
 	comm_event->comm_size = size;
@@ -2109,7 +2109,7 @@ static void perf_counter_mmap_event(stru
 	}
 
 got_name:
-	size = ALIGN(strlen(name), sizeof(u64));
+	size = ALIGN(strlen(name)+1, sizeof(u64));
 
 	mmap_event->file_name = name;
 	mmap_event->file_size = size;

-- 

[PATCH 2/4] perf_counter: optimize mmap/comm tracking

[Peter Zijlstra]

[-- Attachment #1: perf_counter-opt-tracing.patch --]
[-- Type: text/plain, Size: 2339 bytes --]

The mmap/comm tracking code does quite a lot of work before it discovers
there's no interest in it, avoid that by keeping a counter.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/perf_counter.c |   32 ++++++++++++++++++++++++++++++--
 1 file changed, 30 insertions(+), 2 deletions(-)

Index: linux-2.6/kernel/perf_counter.c
===================================================================
--- linux-2.6.orig/kernel/perf_counter.c
+++ linux-2.6/kernel/perf_counter.c
@@ -38,6 +38,10 @@ int perf_max_counters __read_mostly = 1;
 static int perf_reserved_percpu __read_mostly;
 static int perf_overcommit __read_mostly = 1;
 
+static atomic_t nr_mmap_tracking __read_mostly;
+static atomic_t nr_munmap_tracking __read_mostly;
+static atomic_t nr_comm_tracking __read_mostly;
+
 /*
  * Mutex for (sysadmin-configurable) counter reservations:
  */
@@ -1186,6 +1190,13 @@ static void free_counter(struct perf_cou
 {
 	perf_pending_sync(counter);
 
+	if (counter->hw_event.mmap)
+		atomic_dec(&nr_mmap_tracking);
+	if (counter->hw_event.munmap)
+		atomic_dec(&nr_munmap_tracking);
+	if (counter->hw_event.comm)
+		atomic_dec(&nr_comm_tracking);
+
 	if (counter->destroy)
 		counter->destroy(counter);
 
@@ -2128,7 +2139,12 @@ got_name:
 void perf_counter_mmap(unsigned long addr, unsigned long len,
 		       unsigned long pgoff, struct file *file)
 {
-	struct perf_mmap_event mmap_event = {
+	struct perf_mmap_event mmap_event;
+
+	if (!atomic_read(&nr_mmap_tracking))
+		return;
+
+	mmap_event = (struct perf_mmap_event){
 		.file   = file,
 		.event  = {
 			.header = { .type = PERF_EVENT_MMAP, },
@@ -2146,7 +2162,12 @@ void perf_counter_mmap(unsigned long add
 void perf_counter_munmap(unsigned long addr, unsigned long len,
 			 unsigned long pgoff, struct file *file)
 {
-	struct perf_mmap_event mmap_event = {
+	struct perf_mmap_event mmap_event;
+
+	if (!atomic_read(&nr_munmap_tracking))
+		return;
+
+	mmap_event = (struct perf_mmap_event){
 		.file   = file,
 		.event  = {
 			.header = { .type = PERF_EVENT_MUNMAP, },
@@ -2725,6 +2746,13 @@ done:
 
 	counter->hw_ops = hw_ops;
 
+	if (counter->hw_event.mmap)
+		atomic_inc(&nr_mmap_tracking);
+	if (counter->hw_event.munmap)
+		atomic_inc(&nr_munmap_tracking);
+	if (counter->hw_event.comm)
+		atomic_inc(&nr_comm_tracking);
+
 	return counter;
 }
 

-- 

[PATCH 3/4] perf_counter: sysctl for system wide perf counters

[Peter Zijlstra]

[-- Attachment #1: pref_counter-sysctl-global.patch --]
[-- Type: text/plain, Size: 2470 bytes --]

Allow the use of system wide perf counters to everybody, but provide
a sysctl to disable it for the paranoid security minded.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 include/linux/perf_counter.h |    2 ++
 kernel/perf_counter.c        |    4 +++-
 kernel/sysctl.c              |   11 +++++++++++
 3 files changed, 16 insertions(+), 1 deletion(-)

Index: linux-2.6/include/linux/perf_counter.h
===================================================================
--- linux-2.6.orig/include/linux/perf_counter.h
+++ linux-2.6/include/linux/perf_counter.h
@@ -568,6 +568,8 @@ struct perf_callchain_entry {
 
 extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs);
 
+extern int sysctl_perf_counter_priv;
+
 #else
 static inline void
 perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
Index: linux-2.6/kernel/perf_counter.c
===================================================================
--- linux-2.6.orig/kernel/perf_counter.c
+++ linux-2.6/kernel/perf_counter.c
@@ -42,6 +42,8 @@ static atomic_t nr_mmap_tracking __read_
 static atomic_t nr_munmap_tracking __read_mostly;
 static atomic_t nr_comm_tracking __read_mostly;
 
+int sysctl_perf_counter_priv; /* do we need to be privileged */
+
 /*
  * Mutex for (sysadmin-configurable) counter reservations:
  */
@@ -1132,7 +1134,7 @@ static struct perf_counter_context *find
 	 */
 	if (cpu != -1) {
 		/* Must be root to operate on a CPU counter: */
-		if (!capable(CAP_SYS_ADMIN))
+		if (sysctl_perf_counter_priv && !capable(CAP_SYS_ADMIN))
 			return ERR_PTR(-EACCES);
 
 		if (cpu < 0 || cpu > num_possible_cpus())
Index: linux-2.6/kernel/sysctl.c
===================================================================
--- linux-2.6.orig/kernel/sysctl.c
+++ linux-2.6/kernel/sysctl.c
@@ -50,6 +50,7 @@
 #include <linux/reboot.h>
 #include <linux/ftrace.h>
 #include <linux/slow-work.h>
+#include <linux/perf_counter.h>
 
 #include <asm/uaccess.h>
 #include <asm/processor.h>
@@ -931,6 +932,16 @@ static struct ctl_table kern_table[] = {
 		.child		= slow_work_sysctls,
 	},
 #endif
+#ifdef CONFIG_PERF_COUNTERS
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "perf_counter_privileged",
+		.data		= &sysctl_perf_counter_priv,
+		.maxlen		= sizeof(sysctl_perf_counter_priv),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+#endif
 /*
  * NOTE: do not add new entries to this table unless you have read
  * Documentation/sysctl/ctl_unnumbered.txt

-- 

[PATCH 4/4] perf_counter: log full path names

[Peter Zijlstra]

[-- Attachment #1: perf_counter-fix-pathnames.patch --]
[-- Type: text/plain, Size: 780 bytes --]

dentry_path() only provide path-names up to the mount root, which is
unsuited for out purpose, use d_path() instead.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/perf_counter.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux-2.6/kernel/perf_counter.c
===================================================================
--- linux-2.6.orig/kernel/perf_counter.c
+++ linux-2.6/kernel/perf_counter.c
@@ -2111,7 +2111,7 @@ static void perf_counter_mmap_event(stru
 			name = strncpy(tmp, "//enomem", sizeof(tmp));
 			goto got_name;
 		}
-		name = dentry_path(file->f_dentry, buf, PATH_MAX);
+		name = d_path(&file->f_path, buf, PATH_MAX);
 		if (IS_ERR(name)) {
 			name = strncpy(tmp, "//toolong", sizeof(tmp));
 			goto got_name;

-- 

Re: [PATCH 0/4] perf_counter: fixes and an optimization

[Ingo Molnar]

* Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

>  - fixes an off-by-one in string handling
>  - fixes the mmap tracing to provide the full path
>  - allows unpriviledged use of cpu counters

i made the variable here __read_mostly.

>  - optimize mmap/comm tracing

Applied, thanks!

	Ingo

[tip:perfcounters/core] perf_counter: optimize mmap/comm tracking

[Peter Zijlstra]

Commit-ID:  e37f8d14c1f56e274b3e827912f23f0b2fb7a7bb
Gitweb:     http://git.kernel.org/tip/e37f8d14c1f56e274b3e827912f23f0b2fb7a7bb
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Thu, 9 Apr 2009 10:53:44 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 9 Apr 2009 11:00:53 +0200

perf_counter: optimize mmap/comm tracking

Impact: performance optimization

The mmap/comm tracking code does quite a lot of work before it discovers
there's no interest in it, avoid that by keeping a counter.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090409085524.427173196@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 kernel/perf_counter.c |   32 ++++++++++++++++++++++++++++++--
 1 files changed, 30 insertions(+), 2 deletions(-)

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index b07195b..af9fd89 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -38,6 +38,10 @@ int perf_max_counters __read_mostly = 1;
 static int perf_reserved_percpu __read_mostly;
 static int perf_overcommit __read_mostly = 1;
 
+static atomic_t nr_mmap_tracking __read_mostly;
+static atomic_t nr_munmap_tracking __read_mostly;
+static atomic_t nr_comm_tracking __read_mostly;
+
 /*
  * Mutex for (sysadmin-configurable) counter reservations:
  */
@@ -1186,6 +1190,13 @@ static void free_counter(struct perf_counter *counter)
 {
 	perf_pending_sync(counter);
 
+	if (counter->hw_event.mmap)
+		atomic_dec(&nr_mmap_tracking);
+	if (counter->hw_event.munmap)
+		atomic_dec(&nr_munmap_tracking);
+	if (counter->hw_event.comm)
+		atomic_dec(&nr_comm_tracking);
+
 	if (counter->destroy)
 		counter->destroy(counter);
 
@@ -2128,7 +2139,12 @@ got_name:
 void perf_counter_mmap(unsigned long addr, unsigned long len,
 		       unsigned long pgoff, struct file *file)
 {
-	struct perf_mmap_event mmap_event = {
+	struct perf_mmap_event mmap_event;
+
+	if (!atomic_read(&nr_mmap_tracking))
+		return;
+
+	mmap_event = (struct perf_mmap_event){
 		.file   = file,
 		.event  = {
 			.header = { .type = PERF_EVENT_MMAP, },
@@ -2146,7 +2162,12 @@ void perf_counter_mmap(unsigned long addr, unsigned long len,
 void perf_counter_munmap(unsigned long addr, unsigned long len,
 			 unsigned long pgoff, struct file *file)
 {
-	struct perf_mmap_event mmap_event = {
+	struct perf_mmap_event mmap_event;
+
+	if (!atomic_read(&nr_munmap_tracking))
+		return;
+
+	mmap_event = (struct perf_mmap_event){
 		.file   = file,
 		.event  = {
 			.header = { .type = PERF_EVENT_MUNMAP, },
@@ -2725,6 +2746,13 @@ done:
 
 	counter->hw_ops = hw_ops;
 
+	if (counter->hw_event.mmap)
+		atomic_inc(&nr_mmap_tracking);
+	if (counter->hw_event.munmap)
+		atomic_inc(&nr_munmap_tracking);
+	if (counter->hw_event.comm)
+		atomic_inc(&nr_comm_tracking);
+
 	return counter;
 }
 

[tip:perfcounters/core] perf_counter: sysctl for system wide perf counters

[Peter Zijlstra]

Commit-ID:  79bbb0bfff331422caf8a353be68c4764bc1e47f
Gitweb:     http://git.kernel.org/tip/79bbb0bfff331422caf8a353be68c4764bc1e47f
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Thu, 9 Apr 2009 10:53:45 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 9 Apr 2009 11:00:54 +0200

perf_counter: sysctl for system wide perf counters

Impact: add sysctl for paranoid/relaxed perfcounters policy

Allow the use of system wide perf counters to everybody, but provide
a sysctl to disable it for the paranoid security minded.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090409085524.514046352@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 include/linux/perf_counter.h |    2 ++
 kernel/perf_counter.c        |    4 +++-
 kernel/sysctl.c              |   11 +++++++++++
 3 files changed, 16 insertions(+), 1 deletions(-)

diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index c22363a..9814328 100644
--- a/include/linux/perf_counter.h
+++ b/include/linux/perf_counter.h
@@ -568,6 +568,8 @@ struct perf_callchain_entry {
 
 extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs);
 
+extern int sysctl_perf_counter_priv;
+
 #else
 static inline void
 perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index af9fd89..a2fa5bc 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -42,6 +42,8 @@ static atomic_t nr_mmap_tracking __read_mostly;
 static atomic_t nr_munmap_tracking __read_mostly;
 static atomic_t nr_comm_tracking __read_mostly;
 
+int sysctl_perf_counter_priv __read_mostly; /* do we need to be privileged */
+
 /*
  * Mutex for (sysadmin-configurable) counter reservations:
  */
@@ -1132,7 +1134,7 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu)
 	 */
 	if (cpu != -1) {
 		/* Must be root to operate on a CPU counter: */
-		if (!capable(CAP_SYS_ADMIN))
+		if (sysctl_perf_counter_priv && !capable(CAP_SYS_ADMIN))
 			return ERR_PTR(-EACCES);
 
 		if (cpu < 0 || cpu > num_possible_cpus())
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 4286b62..8ba4578 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -49,6 +49,7 @@
 #include <linux/reboot.h>
 #include <linux/ftrace.h>
 #include <linux/slow-work.h>
+#include <linux/perf_counter.h>
 
 #include <asm/uaccess.h>
 #include <asm/processor.h>
@@ -920,6 +921,16 @@ static struct ctl_table kern_table[] = {
 		.child		= slow_work_sysctls,
 	},
 #endif
+#ifdef CONFIG_PERF_COUNTERS
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "perf_counter_privileged",
+		.data		= &sysctl_perf_counter_priv,
+		.maxlen		= sizeof(sysctl_perf_counter_priv),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+#endif
 /*
  * NOTE: do not add new entries to this table unless you have read
  * Documentation/sysctl/ctl_unnumbered.txt

[tip:perfcounters/core] perf_counter: log full path names

[Peter Zijlstra]

Commit-ID:  0a24487bebbee88609004bc5a5ae1b375b292897
Gitweb:     http://git.kernel.org/tip/0a24487bebbee88609004bc5a5ae1b375b292897
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Thu, 9 Apr 2009 10:53:46 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 9 Apr 2009 11:00:54 +0200

perf_counter: log full path names

Impact: fix perf-report output for /home mounted binaries, etc.

dentry_path() only provide path-names up to the mount root, which is
unsuited for out purpose, use d_path() instead.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090409085524.601794134@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 kernel/perf_counter.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index a2fa5bc..0d646ac 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -2111,7 +2111,7 @@ static void perf_counter_mmap_event(struct perf_mmap_event *mmap_event)
 			name = strncpy(tmp, "//enomem", sizeof(tmp));
 			goto got_name;
 		}
-		name = dentry_path(file->f_dentry, buf, PATH_MAX);
+		name = d_path(&file->f_path, buf, PATH_MAX);
 		if (IS_ERR(name)) {
 			name = strncpy(tmp, "//toolong", sizeof(tmp));
 			goto got_name;

Re: [PATCH 2/4] perf_counter: optimize mmap/comm tracking

[Ingo Molnar]

* Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> +static atomic_t nr_comm_tracking __read_mostly;

hm, nr_comm_tracking is not actually used by any fastpath - just 
increased and then decreased.

	Ingo

Re: [PATCH 2/4] perf_counter: optimize mmap/comm tracking

[Peter Zijlstra]

On Thu, 2009-04-09 at 11:16 +0200, Ingo Molnar wrote:
> * Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > +static atomic_t nr_comm_tracking __read_mostly;
> 
> hm, nr_comm_tracking is not actually used by any fastpath - just 
> increased and then decreased.

Index: linux-2.6/kernel/perf_counter.c
===================================================================
--- linux-2.6.orig/kernel/perf_counter.c
+++ linux-2.6/kernel/perf_counter.c
@@ -2018,7 +2018,12 @@ static void perf_counter_comm_event(stru
 
 void perf_counter_comm(struct task_struct *task)
 {
-	struct perf_comm_event comm_event = {
+	struct perf_comm_event comm_event;
+
+	if (!atomic_read(&nr_comm_tracking))
+		return;
+       
+	comm_event = (struct perf_comm_event){
 		.task	= task,
 		.event  = {
 			.header = { .type = PERF_EVENT_COMM, },

[tip:perfcounters/core] perf_counter: optimize mmap/comm tracking

[Peter Zijlstra]

Commit-ID:  9ee318a7825929bc3734110b83ae8e20e53d9de3
Gitweb:     http://git.kernel.org/tip/9ee318a7825929bc3734110b83ae8e20e53d9de3
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Thu, 9 Apr 2009 10:53:44 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 9 Apr 2009 11:50:43 +0200

perf_counter: optimize mmap/comm tracking

Impact: performance optimization

The mmap/comm tracking code does quite a lot of work before it discovers
there's no interest in it, avoid that by keeping a counter.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090409085524.427173196@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 kernel/perf_counter.c |   39 ++++++++++++++++++++++++++++++++++++---
 1 files changed, 36 insertions(+), 3 deletions(-)

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index b07195b..76376ec 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -38,6 +38,10 @@ int perf_max_counters __read_mostly = 1;
 static int perf_reserved_percpu __read_mostly;
 static int perf_overcommit __read_mostly = 1;
 
+static atomic_t nr_mmap_tracking __read_mostly;
+static atomic_t nr_munmap_tracking __read_mostly;
+static atomic_t nr_comm_tracking __read_mostly;
+
 /*
  * Mutex for (sysadmin-configurable) counter reservations:
  */
@@ -1186,6 +1190,13 @@ static void free_counter(struct perf_counter *counter)
 {
 	perf_pending_sync(counter);
 
+	if (counter->hw_event.mmap)
+		atomic_dec(&nr_mmap_tracking);
+	if (counter->hw_event.munmap)
+		atomic_dec(&nr_munmap_tracking);
+	if (counter->hw_event.comm)
+		atomic_dec(&nr_comm_tracking);
+
 	if (counter->destroy)
 		counter->destroy(counter);
 
@@ -2005,7 +2016,12 @@ static void perf_counter_comm_event(struct perf_comm_event *comm_event)
 
 void perf_counter_comm(struct task_struct *task)
 {
-	struct perf_comm_event comm_event = {
+	struct perf_comm_event comm_event;
+
+	if (!atomic_read(&nr_comm_tracking))
+		return;
+       
+	comm_event = (struct perf_comm_event){
 		.task	= task,
 		.event  = {
 			.header = { .type = PERF_EVENT_COMM, },
@@ -2128,7 +2144,12 @@ got_name:
 void perf_counter_mmap(unsigned long addr, unsigned long len,
 		       unsigned long pgoff, struct file *file)
 {
-	struct perf_mmap_event mmap_event = {
+	struct perf_mmap_event mmap_event;
+
+	if (!atomic_read(&nr_mmap_tracking))
+		return;
+
+	mmap_event = (struct perf_mmap_event){
 		.file   = file,
 		.event  = {
 			.header = { .type = PERF_EVENT_MMAP, },
@@ -2146,7 +2167,12 @@ void perf_counter_mmap(unsigned long addr, unsigned long len,
 void perf_counter_munmap(unsigned long addr, unsigned long len,
 			 unsigned long pgoff, struct file *file)
 {
-	struct perf_mmap_event mmap_event = {
+	struct perf_mmap_event mmap_event;
+
+	if (!atomic_read(&nr_munmap_tracking))
+		return;
+
+	mmap_event = (struct perf_mmap_event){
 		.file   = file,
 		.event  = {
 			.header = { .type = PERF_EVENT_MUNMAP, },
@@ -2725,6 +2751,13 @@ done:
 
 	counter->hw_ops = hw_ops;
 
+	if (counter->hw_event.mmap)
+		atomic_inc(&nr_mmap_tracking);
+	if (counter->hw_event.munmap)
+		atomic_inc(&nr_munmap_tracking);
+	if (counter->hw_event.comm)
+		atomic_inc(&nr_comm_tracking);
+
 	return counter;
 }
 

[tip:perfcounters/core] perf_counter: sysctl for system wide perf counters

[Peter Zijlstra]

Commit-ID:  1ccd15497869f3ed83b5225d410df53a96e52757
Gitweb:     http://git.kernel.org/tip/1ccd15497869f3ed83b5225d410df53a96e52757
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Thu, 9 Apr 2009 10:53:45 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 9 Apr 2009 11:50:52 +0200

perf_counter: sysctl for system wide perf counters

Impact: add sysctl for paranoid/relaxed perfcounters policy

Allow the use of system wide perf counters to everybody, but provide
a sysctl to disable it for the paranoid security minded.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090409085524.514046352@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 include/linux/perf_counter.h |    2 ++
 kernel/perf_counter.c        |    4 +++-
 kernel/sysctl.c              |   11 +++++++++++
 3 files changed, 16 insertions(+), 1 deletions(-)

diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index c22363a..9814328 100644
--- a/include/linux/perf_counter.h
+++ b/include/linux/perf_counter.h
@@ -568,6 +568,8 @@ struct perf_callchain_entry {
 
 extern struct perf_callchain_entry *perf_callchain(struct pt_regs *regs);
 
+extern int sysctl_perf_counter_priv;
+
 #else
 static inline void
 perf_counter_task_sched_in(struct task_struct *task, int cpu)		{ }
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 76376ec..7efb7eb 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -42,6 +42,8 @@ static atomic_t nr_mmap_tracking __read_mostly;
 static atomic_t nr_munmap_tracking __read_mostly;
 static atomic_t nr_comm_tracking __read_mostly;
 
+int sysctl_perf_counter_priv __read_mostly; /* do we need to be privileged */
+
 /*
  * Mutex for (sysadmin-configurable) counter reservations:
  */
@@ -1132,7 +1134,7 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu)
 	 */
 	if (cpu != -1) {
 		/* Must be root to operate on a CPU counter: */
-		if (!capable(CAP_SYS_ADMIN))
+		if (sysctl_perf_counter_priv && !capable(CAP_SYS_ADMIN))
 			return ERR_PTR(-EACCES);
 
 		if (cpu < 0 || cpu > num_possible_cpus())
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 4286b62..8ba4578 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -49,6 +49,7 @@
 #include <linux/reboot.h>
 #include <linux/ftrace.h>
 #include <linux/slow-work.h>
+#include <linux/perf_counter.h>
 
 #include <asm/uaccess.h>
 #include <asm/processor.h>
@@ -920,6 +921,16 @@ static struct ctl_table kern_table[] = {
 		.child		= slow_work_sysctls,
 	},
 #endif
+#ifdef CONFIG_PERF_COUNTERS
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "perf_counter_privileged",
+		.data		= &sysctl_perf_counter_priv,
+		.maxlen		= sizeof(sysctl_perf_counter_priv),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
+#endif
 /*
  * NOTE: do not add new entries to this table unless you have read
  * Documentation/sysctl/ctl_unnumbered.txt

[tip:perfcounters/core] perf_counter: log full path names

[Peter Zijlstra]

Commit-ID:  d3d21c412d8525eb2e208d990ab5eee5fb0fe03d
Gitweb:     http://git.kernel.org/tip/d3d21c412d8525eb2e208d990ab5eee5fb0fe03d
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Thu, 9 Apr 2009 10:53:46 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 9 Apr 2009 11:50:54 +0200

perf_counter: log full path names

Impact: fix perf-report output for /home mounted binaries, etc.

dentry_path() only provide path-names up to the mount root, which is
unsuited for out purpose, use d_path() instead.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090409085524.601794134@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 kernel/perf_counter.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 7efb7eb..7f9521c 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -2116,7 +2116,7 @@ static void perf_counter_mmap_event(struct perf_mmap_event *mmap_event)
 			name = strncpy(tmp, "//enomem", sizeof(tmp));
 			goto got_name;
 		}
-		name = dentry_path(file->f_dentry, buf, PATH_MAX);
+		name = d_path(&file->f_path, buf, PATH_MAX);
 		if (IS_ERR(name)) {
 			name = strncpy(tmp, "//toolong", sizeof(tmp));
 			goto got_name;