[RFC/PATCH] perf: Expand definition of sysfs format attribute

[RFC/PATCH] perf: Expand definition of sysfs format attribute

[Michael Ellerman]

Make it explicit that the format attributes may define overlapping bit
ranges. Unfortunately this was left unspecified originally, and all the
examples show non-overlapping ranges. I don't believe this is an ABI
change, as we are defining something that was previously undefined, but
others may disagree.

The POWER8 PMU would like to define overlapping ranges, as bit ranges in
the event code have different meanings for certain events. It will also
allow us to define an overarching "event" field, that encompasses all
others.

As far as I can see perf is comfortable with this change, however I am
not sure if there are any other users of the interface.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
---
 Documentation/ABI/testing/sysfs-bus-event_source-devices-format |    6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
index 079afc7..77f47ff 100644
--- a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
+++ b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
@@ -9,6 +9,12 @@ Description:
 		we want to export, so that userspace can deal with sane
 		name/value pairs.
 
+		Userspace must be prepared for the possibility that attributes
+		define overlapping bit ranges. For example:
+			attr1 = 'config:0-23'
+			attr2 = 'config:0-7'
+			attr3 = 'config:12-35'
+
 		Example: 'config1:1,6-10,44'
 		Defines contents of attribute that occupies bits 1,6-10,44 of
 		perf_event_attr::config1.
-- 
1.7.10.4

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Michael Ellerman]

On Mon, Mar 04, 2013 at 03:21:05PM +1100, Michael Ellerman wrote:
> Make it explicit that the format attributes may define overlapping bit
> ranges. Unfortunately this was left unspecified originally, and all the
> examples show non-overlapping ranges. I don't believe this is an ABI
> change, as we are defining something that was previously undefined, but
> others may disagree.
> 
> The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> the event code have different meanings for certain events. It will also
> allow us to define an overarching "event" field, that encompasses all
> others.
> 
> As far as I can see perf is comfortable with this change, however I am
> not sure if there are any other users of the interface.

Any comments on this one?

cheers

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Michael Ellerman]

On Mon, Apr 08, 2013 at 05:24:22PM +1000, Michael Ellerman wrote:
> On Mon, Mar 04, 2013 at 03:21:05PM +1100, Michael Ellerman wrote:
> > Make it explicit that the format attributes may define overlapping bit
> > ranges. Unfortunately this was left unspecified originally, and all the
> > examples show non-overlapping ranges. I don't believe this is an ABI
> > change, as we are defining something that was previously undefined, but
> > others may disagree.
> > 
> > The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> > the event code have different meanings for certain events. It will also
> > allow us to define an overarching "event" field, that encompasses all
> > others.
> > 
> > As far as I can see perf is comfortable with this change, however I am
> > not sure if there are any other users of the interface.
> 
> Any comments on this one?

Hi Arnaldo,

I think the consensus from Peter and Jiri was that this was OK, are you
happy to merge it?

cheers

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Arnaldo Carvalho de Melo]

Em Mon, May 06, 2013 at 11:08:30AM +1000, Michael Ellerman escreveu:
> On Mon, Apr 08, 2013 at 05:24:22PM +1000, Michael Ellerman wrote:
> > On Mon, Mar 04, 2013 at 03:21:05PM +1100, Michael Ellerman wrote:
> > > Make it explicit that the format attributes may define overlapping bit
> > > ranges. Unfortunately this was left unspecified originally, and all the
> > > examples show non-overlapping ranges. I don't believe this is an ABI
> > > change, as we are defining something that was previously undefined, but
> > > others may disagree.
> > > 
> > > The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> > > the event code have different meanings for certain events. It will also
> > > allow us to define an overarching "event" field, that encompasses all
> > > others.
> > > 
> > > As far as I can see perf is comfortable with this change, however I am
> > > not sure if there are any other users of the interface.
> > 
> > Any comments on this one?

> I think the consensus from Peter and Jiri was that this was OK, are you
> happy to merge it?

I'll do it eventually, but now I'm on vacations, so perhaps Jiri can
process it and push to Ingo? Jiri?

- Arnaldo

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Jiri Olsa]

On Mon, May 06, 2013 at 05:20:08AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Mon, May 06, 2013 at 11:08:30AM +1000, Michael Ellerman escreveu:
> > On Mon, Apr 08, 2013 at 05:24:22PM +1000, Michael Ellerman wrote:
> > > On Mon, Mar 04, 2013 at 03:21:05PM +1100, Michael Ellerman wrote:
> > > > Make it explicit that the format attributes may define overlapping bit
> > > > ranges. Unfortunately this was left unspecified originally, and all the
> > > > examples show non-overlapping ranges. I don't believe this is an ABI
> > > > change, as we are defining something that was previously undefined, but
> > > > others may disagree.
> > > > 
> > > > The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> > > > the event code have different meanings for certain events. It will also
> > > > allow us to define an overarching "event" field, that encompasses all
> > > > others.
> > > > 
> > > > As far as I can see perf is comfortable with this change, however I am
> > > > not sure if there are any other users of the interface.
> > > 
> > > Any comments on this one?
> 
> > I think the consensus from Peter and Jiri was that this was OK, are you
> > happy to merge it?
> 
> I'll do it eventually, but now I'm on vacations, so perhaps Jiri can
> process it and push to Ingo? Jiri?

ok, I'll try to do this soon ;)

jirka

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Peter Zijlstra]

On Mon, 2013-03-04 at 15:21 +1100, Michael Ellerman wrote:
> Make it explicit that the format attributes may define overlapping bit
> ranges. Unfortunately this was left unspecified originally, and all the
> examples show non-overlapping ranges. I don't believe this is an ABI
> change, as we are defining something that was previously undefined, but
> others may disagree.
> 
> The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> the event code have different meanings for certain events. It will also
> allow us to define an overarching "event" field, that encompasses all
> others.
> 
> As far as I can see perf is comfortable with this change, however I am
> not sure if there are any other users of the interface.
> 
> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> ---
>  Documentation/ABI/testing/sysfs-bus-event_source-devices-format |    6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> index 079afc7..77f47ff 100644
> --- a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> +++ b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> @@ -9,6 +9,12 @@ Description:
>  		we want to export, so that userspace can deal with sane
>  		name/value pairs.
>  
> +		Userspace must be prepared for the possibility that attributes
> +		define overlapping bit ranges. For example:
> +			attr1 = 'config:0-23'
> +			attr2 = 'config:0-7'
> +			attr3 = 'config:12-35'
> +
>  		Example: 'config1:1,6-10,44'
>  		Defines contents of attribute that occupies bits 1,6-10,44 of
>  		perf_event_attr::config1.

ISTR discussing this with Jiri at some point.. I think we ended up with
being fine with overlapping ranges but having perf issue a warning (not
an error) when attributes of a single event have overlap.

I'm not sure the latter was ever implemented in the userspace side.

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Michael Ellerman]

On Mon, Apr 08, 2013 at 11:05:15AM +0200, Peter Zijlstra wrote:
> On Mon, 2013-03-04 at 15:21 +1100, Michael Ellerman wrote:
> > Make it explicit that the format attributes may define overlapping bit
> > ranges. Unfortunately this was left unspecified originally, and all the
> > examples show non-overlapping ranges. I don't believe this is an ABI
> > change, as we are defining something that was previously undefined, but
> > others may disagree.
> > 
> > The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> > the event code have different meanings for certain events. It will also
> > allow us to define an overarching "event" field, that encompasses all
> > others.
> > 
> > As far as I can see perf is comfortable with this change, however I am
> > not sure if there are any other users of the interface.
> > 
> > Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
> > ---
> >  Documentation/ABI/testing/sysfs-bus-event_source-devices-format |    6 ++++++
> >  1 file changed, 6 insertions(+)
> > 
> > diff --git a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> > index 079afc7..77f47ff 100644
> > --- a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> > +++ b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> > @@ -9,6 +9,12 @@ Description:
> >  		we want to export, so that userspace can deal with sane
> >  		name/value pairs.
> >  
> > +		Userspace must be prepared for the possibility that attributes
> > +		define overlapping bit ranges. For example:
> > +			attr1 = 'config:0-23'
> > +			attr2 = 'config:0-7'
> > +			attr3 = 'config:12-35'
> > +
> >  		Example: 'config1:1,6-10,44'
> >  		Defines contents of attribute that occupies bits 1,6-10,44 of
> >  		perf_event_attr::config1.
> 
> ISTR discussing this with Jiri at some point.. I think we ended up with
> being fine with overlapping ranges but having perf issue a warning (not
> an error) when attributes of a single event have overlap.

OK great.

> I'm not sure the latter was ever implemented in the userspace side.

Yeah not that I can see with a quick look. It's fairly advanced usage
IMHO so a warning would be nice, but not essential.

cheers

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Jiri Olsa]

On Mon, Apr 08, 2013 at 09:57:50PM +1000, Michael Ellerman wrote:

SNIP

> > 
> > ISTR discussing this with Jiri at some point.. I think we ended up with
> > being fine with overlapping ranges but having perf issue a warning (not
> > an error) when attributes of a single event have overlap.
> 
> OK great.
> 
> > I'm not sure the latter was ever implemented in the userspace side.
> 
> Yeah not that I can see with a quick look. It's fairly advanced usage
> IMHO so a warning would be nice, but not essential.

oops, missed this email.. sry

it should be ok to overlap.. we don't check those ranges in perf
tool now.. just take whatever is defined in format file

adding the warning to my todo list ;-)

jirka

Re: [RFC/PATCH] perf: Expand definition of sysfs format attribute

[Peter Zijlstra]

On Mon, Mar 04, 2013 at 03:21:05PM +1100, Michael Ellerman wrote:
> Make it explicit that the format attributes may define overlapping bit
> ranges. Unfortunately this was left unspecified originally, and all the
> examples show non-overlapping ranges. I don't believe this is an ABI
> change, as we are defining something that was previously undefined, but
> others may disagree.
> 
> The POWER8 PMU would like to define overlapping ranges, as bit ranges in
> the event code have different meanings for certain events. It will also
> allow us to define an overarching "event" field, that encompasses all
> others.
> 
> As far as I can see perf is comfortable with this change, however I am
> not sure if there are any other users of the interface.
> 
> Signed-off-by: Michael Ellerman <michael@ellerman.id.au>

Jiri was going to collect to send to Ingo..

Acked-by: Peter Zijlstra <peterz@infradead.org>

> ---
>  Documentation/ABI/testing/sysfs-bus-event_source-devices-format |    6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> index 079afc7..77f47ff 100644
> --- a/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> +++ b/Documentation/ABI/testing/sysfs-bus-event_source-devices-format
> @@ -9,6 +9,12 @@ Description:
>  		we want to export, so that userspace can deal with sane
>  		name/value pairs.
>  
> +		Userspace must be prepared for the possibility that attributes
> +		define overlapping bit ranges. For example:
> +			attr1 = 'config:0-23'
> +			attr2 = 'config:0-7'
> +			attr3 = 'config:12-35'
> +
>  		Example: 'config1:1,6-10,44'
>  		Defines contents of attribute that occupies bits 1,6-10,44 of
>  		perf_event_attr::config1.
> -- 
> 1.7.10.4
>