Michael Ellerman wrote:
"Naveen N. Rao" <naveen.n....@linux.vnet.ibm.com> writes:
My earlier assumption was that we have other scenarios when we are in
realmode (specifically with MSR_RI unset) where we won't be able to
recover from a trap, during function tracing (*). I did a set of
experiments yesterday to verify that, but I was not able to uncover
any such scenarios with my brief testing. So, we seem to be
functioning just fine while tracing realmode C code, except for KVM.
Hmm. If MSR_RI is clear then that should indicate that you can't recover
from an interrupt, typically because you'd lose state in SRR0/1. So I
would expect things to go badly in that case.
Yes, so it looks like we aren't calling into any C code (that isn't
already annotated with 'notrace') with MSR_RI unset. At least, with the
usual interrupt handling on powernv. I tested this by putting a 'trap'
in the function tracer callback after the recursion test. This forces a
trap for each function that we trace non-recursively.
There may be other paths where we do so, but it isn't as pervasive as I
previously thought. So, we should be able to exclude those paths using
the paca field, as and when we find them.
- Naveen
