Re: [PATCH v2] arm64: perf: Use only exclude_kernel attribute when kernel is running in HYP

[Will Deacon <will.deacon@arm.com>]

On Wed, Apr 26, 2017 at 01:41:42PM +0000, Jayachandran C wrote:
> On Wed, Apr 26, 2017 at 11:10:21AM +0100, Will Deacon wrote:
> > On Wed, Apr 26, 2017 at 07:22:46AM +0000, Pinski, Andrew wrote:
> > > On 4/25/2017 11:53 PM, Jayachandran C. wrote:
> > > > On Tue, Apr 25, 2017 at 10:23 PM, Will Deacon <will.deacon@arm.com> wrote:
> > > >> On Tue, Apr 25, 2017 at 09:13:40AM +0530, Ganapatrao Kulkarni wrote:
> > > >>> On Mon, Apr 24, 2017 at 9:15 PM, Will Deacon <will.deacon@arm.com> wrote:
> > > >>>> On Thu, Apr 20, 2017 at 02:56:50PM +0530, Ganapatrao Kulkarni wrote:
> > > >>>>> OK, if you are ok with sysfs part, i can send next version with that
> > > >>>>> change only?.
> > > >>>> I think the sysfs part is still a little dodgy, since you still expose the
> > > >>>> "exclude_hv" file with a value of 0 when not running at EL2, which would
> > > >>>> imply that exclude_hv is forced to zero. I don't think that's correct.
> > > >>> okay, i can make exclude_hv visible only when kernel booted in EL2.
> > > >>> is it ok to have empty directory "attr" when kernel booted to EL1?
> > > >>> attr can be place holder for any other miscellaneous attributes, that
> > > >>> can be added in future.
> > > >> Sounds good to me, although I'll seek comment from the other perf folks
> > > >> before merging anything with ABI implications.
> > > > Do you really think this is the solution given:
> > > > - this is an arm64 specific sysfs interface that is tied to the perf API
> > 
> > That's why I want feedback from others. The intention would be that this can
> > be used by other PMUs as well, since it's not uncommon that parts of the
> > sizeable perf_event_attr structure are not used by a given PMU.
> > 
> > > > - the perf API documentation has to be updated for this
> > 
> > So? If having to update documentation means we shouldn't change the kernel,
> > then we may as well all find new jobs.
> > 
> > > > - All the applications that use the perf API have to be modified to
> > > > check this sysfs interface
> > > > - If the application fails to do so, a very narrow corner case
> > > > (exclude_hv != exclude_kernel and VHE enabled) fails.
> > 
> > See below, but apparently people care about it.
> > 
> > > > Any application that really cares can already do see if exclude_hv !=
> > > > exclude_kernel case works by calling perf_open_event() with those
> > > > options and checking the return value.
> > 
> > That's a good point: there is *something* userspace can do, although that
> > would be arm64-specific and doesn't really help with the state-space
> > explosion you get with combinations of invalid/unused perf_event_attr
> > fields.
> > 
> > > An example of an application which needs to changed is HHVM. Currently 
> > > it sets exclude_hv to true but exclude_kernel to false as it does not 
> > > care about the hypervisor associated perf events associated with the 
> > > code, only the kernel and userspace associated evnts.
> > > Yes we could submit a patch to use the sysfs interface to check but it 
> > > would look funny and the facebook folks might reject the patch as it is 
> > > ARM64 specific in generic code.  Note this is how all of this discussion 
> > > started was HHVM's call to perf_open_event was failing.
> > 
> > Hmm, if you're saying that HHVM won't be changed to use the sysfs stuff,
> > then why are we bothering?
> > 
> > Not sure where this leaves us.
> 
> If my understanding is correct, the sysfs suggestion above is going to
> add API complexity without solving the issue. Ignoring the exclude_hv if
> it cannot be honored would be a better solution.

Better for HHVM, sure, but I don't think it's better in general. It means
that we silently do the opposite of what the user has requested in some
configurations.

Will