On Fri, Apr 28, 2017 at 05:38:23PM +0100, Will Deacon wrote:

Hi guys,

On Fri, Apr 28, 2017 at 01:46:24PM +0000, Jayachandran C wrote:

quoted

On Thu, Apr 27, 2017 at 06:37:59PM +0100, Will Deacon wrote:

quoted

quoted

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.

If my understanding is correct, when is_kernel_in_hyp_mode() is true,
the kernel is in EL2 and there is no real hypervisor with hvc calls
from kernel. Ignoring the exclude_hv would be correct.

When kernel is in EL1, it would be correct to consider exclude_hv to
skip events in EL2 (reached with hvc).

I don't see the issue, can you please give more detail on the config
with unexpected behavior?

This got me thinking, so I tried to look at the history of exclude_hv. It
turns out it was added in 0475f9ea8e2c ("perf_counters: allow users to
count user, kernel and/or hypervisor events") for PowerPC, not x86 (where
this doesn't seem to be supported).

Notably, it looks like it's always ignored for the x86 CPU PMU, and ignored
on PowerPC when a hypervisor is not present. I think that backs up your
suggestion that we should ignore it when is_kernel_in_hyp_mode() is true.

In which case, I withdraw my objection to ignoring exclude_hv when running
in hyp mode, but please add a comment explaining the rationale!

Thanks, we will send out an updated patch with a commit message summarizing
this disucssion.

JC