Michael Ellerman [off-list ref] writes:

Fabiano Rosas [off-list ref] writes:

quoted

Nicholas Piggin [off-list ref] writes:

quoted

The TIDR SPR only exists on POWER9. Avoid accessing it when the
feature bit for it is not set.

Not related to this patch, but how does this work with compat mode? A P9
compat mode guest would get an invalid instruction when trying to access
this SPR?

Good question.

I assume you're talking about P9 compat mode on P10.

In general compat mode only applies to userspace, because it's
implemented by setting the PCR which only (mostly?) applies to PR=1.

I don't think there's any special casing in the ISA for the TIDR, so I
think it just falls into the unimplemented SPR case for mt/fspr.

That's documented in Book III section 5.4.4, in particular on page 1171
it says:

  Execution of this instruction specifying an SPR number
  that is undefined for the implementation causes one of
  the following.
  • if spr[0]=0:
    - if MSR[PR]=1: Hypervisor Emulation Assistance interrupt
    - if MSR[PR]=0: Hypervisor Emulation Assistance interrupt for SPR
      0,4,5, and 6, and no operation (i.e., the instruction is treated
      as a no-op) when LPCR[EVIRT]=0 and Hypervisor Emulation Assistance
      interrupt when LPCR[EVIRT]=1 for all other SPRs

I knew this must have been somewhere in there but had no idea how to
find it. Thanks.

Linux doesn't set EVIRT, and I assume neither does phyp, so it behaves
like a nop.

We actually use that behaviour in xmon to detect that an SPR is not
implemented, by noticing that the mfspr has no effect on the target
register, see dump_one_spr().

We should really write some docs on compat mode in the linuxppc wiki
and/or Documentation ;)

Hmm I was not aware we had a wiki. I'll see if I can contribute
something. I need to go learn all this stuff first, though =D.

cheers