amd64: Don't probe MSR_IRPERF when running under bhyve

[markj]

Summary

vmm doesn't implement it, and bhyve simply aborts. I fixed this in
commit c83e075d72f6, but it'll still be impossible to boot newer FreeBSD
guests on older hosts without this workaround.

Fixes: 2272fac218a8 ("hwpmc: Add support for AMD-only IRPERF in hwpmc PERF class")

[adrian]

oh, good catch! Is this going to be true forever, or just at the present moment?

[markj]

In D60730#1392841, @adrian wrote:

oh, good catch! Is this going to be true forever, or just at the present moment?

What do you mean by "this"? That vmm won't implement performance counters? Probably not. Maybe I should add a tunable to override this.

[adrian]

In D60730#1392856, @markj wrote:

In D60730#1392841, @adrian wrote:

oh, good catch! Is this going to be true forever, or just at the present moment?

What do you mean by "this"? That vmm won't implement performance counters? Probably not. Maybe I should add a tunable to override this.

right, i wonder if this is better expressed as a tunable for now until it grows another guest extension capability flag at some point?

[markj]

In D60730#1392857, @adrian wrote:

In D60730#1392856, @markj wrote:

In D60730#1392841, @adrian wrote:

oh, good catch! Is this going to be true forever, or just at the present moment?

What do you mean by "this"? That vmm won't implement performance counters? Probably not. Maybe I should add a tunable to override this.

right, i wonder if this is better expressed as a tunable for now until it grows another guest extension capability flag at some point?

Well, the disabled code is just doing some sanity checking of the MSR interface. It doesn't hurt anything to skip it.

[markj]

I plan to commit this soon, as the problem breaks my dev environment, i.e., I need this patch on every branch I'm working on.

[adrian]

ship it! it needs to be not immediately broken!