diff mbox series

[v2,2/7] perf/core: open access for CAP_SYS_PERFMON privileged process

Message ID fd6ffb43-ed43-14cd-b286-6ab4b199155b@linux.intel.com (mailing list archive)
State New, archived
Headers show
Series Introduce CAP_SYS_PERFMON to secure system performance monitoring and observability | expand

Commit Message

Alexey Budankov Dec. 16, 2019, 7:15 a.m. UTC 
Open access to perf_events monitoring for CAP_SYS_PERFMON privileged processes.
For backward compatibility reasons access to perf_events subsystem remains open
for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage for secure
perf_events monitoring is discouraged with respect to CAP_SYS_PERFMON capability.

Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
---
 include/linux/perf_event.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

Comments

Lubashev, Igor Dec. 16, 2019, 4:12 p.m. UTC | #1 
On Mon, Dec 16, 2019 at 2:15 AM, Alexey Budankov <alexey.budankov@linux.intel.com> wrote:
> 
> Open access to perf_events monitoring for CAP_SYS_PERFMON privileged
> processes.
> For backward compatibility reasons access to perf_events subsystem remains
> open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage
> for secure perf_events monitoring is discouraged with respect to
> CAP_SYS_PERFMON capability.
> 
> Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
> ---
>  include/linux/perf_event.h | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index
> 34c7c6910026..52313d2cc343 100644
> --- a/include/linux/perf_event.h
> +++ b/include/linux/perf_event.h
> @@ -1285,7 +1285,8 @@ static inline int perf_is_paranoid(void)
> 
>  static inline int perf_allow_kernel(struct perf_event_attr *attr)  {
> -	if (sysctl_perf_event_paranoid > 1 && !capable(CAP_SYS_ADMIN))
> +	if (sysctl_perf_event_paranoid > 1 &&
> +	   !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>  		return -EACCES;
> 
>  	return security_perf_event_open(attr, PERF_SECURITY_KERNEL); @@
> -1293,7 +1294,8 @@ static inline int perf_allow_kernel(struct
> perf_event_attr *attr)
> 
>  static inline int perf_allow_cpu(struct perf_event_attr *attr)  {
> -	if (sysctl_perf_event_paranoid > 0 && !capable(CAP_SYS_ADMIN))
> +	if (sysctl_perf_event_paranoid > 0 &&
> +	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>  		return -EACCES;
> 
>  	return security_perf_event_open(attr, PERF_SECURITY_CPU); @@ -
> 1301,7 +1303,8 @@ static inline int perf_allow_cpu(struct perf_event_attr
> *attr)
> 
>  static inline int perf_allow_tracepoint(struct perf_event_attr *attr)  {
> -	if (sysctl_perf_event_paranoid > -1 && !capable(CAP_SYS_ADMIN))
> +	if (sysctl_perf_event_paranoid > -1 &&
> +	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>  		return -EPERM;
> 
>  	return security_perf_event_open(attr, PERF_SECURITY_TRACEPOINT);
> --
> 2.20.1

Thanks.  I like the idea of CAP_SYS_PERFMON that does not require CAP_SYS_ADMIN.  It makes granting users ability to run perf a bit safer.

I see a lot of "(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)" constructs now.  Maybe wrapping it in an " inline bool perfmon_capable()" defined somewhere (like in /include/linux/capability.h)?

- Igor
Alexey Budankov Dec. 16, 2019, 4:33 p.m. UTC | #2 
On 16.12.2019 19:12, Lubashev, Igor wrote:
> On Mon, Dec 16, 2019 at 2:15 AM, Alexey Budankov <alexey.budankov@linux.intel.com> wrote:
>>
>> Open access to perf_events monitoring for CAP_SYS_PERFMON privileged
>> processes.
>> For backward compatibility reasons access to perf_events subsystem remains
>> open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage
>> for secure perf_events monitoring is discouraged with respect to
>> CAP_SYS_PERFMON capability.
>>
>> Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
>> ---
>>  include/linux/perf_event.h | 9 ++++++---
>>  1 file changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index
>> 34c7c6910026..52313d2cc343 100644
>> --- a/include/linux/perf_event.h
>> +++ b/include/linux/perf_event.h
>> @@ -1285,7 +1285,8 @@ static inline int perf_is_paranoid(void)
>>
>>  static inline int perf_allow_kernel(struct perf_event_attr *attr)  {
>> -	if (sysctl_perf_event_paranoid > 1 && !capable(CAP_SYS_ADMIN))
>> +	if (sysctl_perf_event_paranoid > 1 &&
>> +	   !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>>  		return -EACCES;
>>
>>  	return security_perf_event_open(attr, PERF_SECURITY_KERNEL); @@
>> -1293,7 +1294,8 @@ static inline int perf_allow_kernel(struct
>> perf_event_attr *attr)
>>
>>  static inline int perf_allow_cpu(struct perf_event_attr *attr)  {
>> -	if (sysctl_perf_event_paranoid > 0 && !capable(CAP_SYS_ADMIN))
>> +	if (sysctl_perf_event_paranoid > 0 &&
>> +	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>>  		return -EACCES;
>>
>>  	return security_perf_event_open(attr, PERF_SECURITY_CPU); @@ -
>> 1301,7 +1303,8 @@ static inline int perf_allow_cpu(struct perf_event_attr
>> *attr)
>>
>>  static inline int perf_allow_tracepoint(struct perf_event_attr *attr)  {
>> -	if (sysctl_perf_event_paranoid > -1 && !capable(CAP_SYS_ADMIN))
>> +	if (sysctl_perf_event_paranoid > -1 &&
>> +	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>>  		return -EPERM;
>>
>>  	return security_perf_event_open(attr, PERF_SECURITY_TRACEPOINT);
>> --
>> 2.20.1
> 
> Thanks.  I like the idea of CAP_SYS_PERFMON that does not require CAP_SYS_ADMIN.  It makes granting users ability to run perf a bit safer.
> 
> I see a lot of "(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)" constructs now.  Maybe wrapping it in an " inline bool perfmon_capable()" defined somewhere (like in /include/linux/capability.h)?

Yes, it makes sense.

Thanks,
Alexey

> 
> - Igor
>
Alexey Budankov Dec. 16, 2019, 5:12 p.m. UTC | #3 
On 16.12.2019 19:12, Lubashev, Igor wrote:
> On Mon, Dec 16, 2019 at 2:15 AM, Alexey Budankov <alexey.budankov@linux.intel.com> wrote:
>>
>> Open access to perf_events monitoring for CAP_SYS_PERFMON privileged
>> processes.
>> For backward compatibility reasons access to perf_events subsystem remains
>> open for CAP_SYS_ADMIN privileged processes but CAP_SYS_ADMIN usage
>> for secure perf_events monitoring is discouraged with respect to
>> CAP_SYS_PERFMON capability.
>>
>> Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
>> ---
>>  include/linux/perf_event.h | 9 ++++++---
>>  1 file changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index
>> 34c7c6910026..52313d2cc343 100644
>> --- a/include/linux/perf_event.h
>> +++ b/include/linux/perf_event.h
>> @@ -1285,7 +1285,8 @@ static inline int perf_is_paranoid(void)
>>
>>  static inline int perf_allow_kernel(struct perf_event_attr *attr)  {
>> -	if (sysctl_perf_event_paranoid > 1 && !capable(CAP_SYS_ADMIN))
>> +	if (sysctl_perf_event_paranoid > 1 &&
>> +	   !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>>  		return -EACCES;
>>
>>  	return security_perf_event_open(attr, PERF_SECURITY_KERNEL); @@
>> -1293,7 +1294,8 @@ static inline int perf_allow_kernel(struct
>> perf_event_attr *attr)
>>
>>  static inline int perf_allow_cpu(struct perf_event_attr *attr)  {
>> -	if (sysctl_perf_event_paranoid > 0 && !capable(CAP_SYS_ADMIN))
>> +	if (sysctl_perf_event_paranoid > 0 &&
>> +	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>>  		return -EACCES;
>>
>>  	return security_perf_event_open(attr, PERF_SECURITY_CPU); @@ -
>> 1301,7 +1303,8 @@ static inline int perf_allow_cpu(struct perf_event_attr
>> *attr)
>>
>>  static inline int perf_allow_tracepoint(struct perf_event_attr *attr)  {
>> -	if (sysctl_perf_event_paranoid > -1 && !capable(CAP_SYS_ADMIN))
>> +	if (sysctl_perf_event_paranoid > -1 &&
>> +	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
>>  		return -EPERM;
>>
>>  	return security_perf_event_open(attr, PERF_SECURITY_TRACEPOINT);
>> --
>> 2.20.1
> 
> Thanks.  I like the idea of CAP_SYS_PERFMON that does not require CAP_SYS_ADMIN.  It makes granting users ability to run perf a bit safer.
> 
> I see a lot of "(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)" constructs now.  Maybe wrapping it in an " inline bool perfmon_capable()" defined somewhere (like in /include/linux/capability.h)?

Sounds reasonable, thanks!

~Alexey

> 
> - Igor
>
diff mbox series

Patch

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 34c7c6910026..52313d2cc343 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -1285,7 +1285,8 @@  static inline int perf_is_paranoid(void)
 
 static inline int perf_allow_kernel(struct perf_event_attr *attr)
 {
-	if (sysctl_perf_event_paranoid > 1 && !capable(CAP_SYS_ADMIN))
+	if (sysctl_perf_event_paranoid > 1 &&
+	   !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
 		return -EACCES;
 
 	return security_perf_event_open(attr, PERF_SECURITY_KERNEL);
@@ -1293,7 +1294,8 @@  static inline int perf_allow_kernel(struct perf_event_attr *attr)
 
 static inline int perf_allow_cpu(struct perf_event_attr *attr)
 {
-	if (sysctl_perf_event_paranoid > 0 && !capable(CAP_SYS_ADMIN))
+	if (sysctl_perf_event_paranoid > 0 &&
+	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
 		return -EACCES;
 
 	return security_perf_event_open(attr, PERF_SECURITY_CPU);
@@ -1301,7 +1303,8 @@  static inline int perf_allow_cpu(struct perf_event_attr *attr)
 
 static inline int perf_allow_tracepoint(struct perf_event_attr *attr)
 {
-	if (sysctl_perf_event_paranoid > -1 && !capable(CAP_SYS_ADMIN))
+	if (sysctl_perf_event_paranoid > -1 &&
+	    !(capable(CAP_SYS_PERFMON) || capable(CAP_SYS_ADMIN)))
 		return -EPERM;
 
 	return security_perf_event_open(attr, PERF_SECURITY_TRACEPOINT);