iflib: Use CPU_FFS correctly

[cem]

Summary

Add some asserts and invariants checking to be sure we do not index with
a bogus qid.

PR: 212418

[cem]

Fix assert message to match pasted-to function.

[nwhitehorn]

sys/net/iflib.c

4306–4308

Check CPU_ABSENT(cpuid) here, too, I think.

4370

This would be another good place to assert !CPU_ABSENT(cpu)

[cem]

sys/net/iflib.c

4306–4308

ifc_cpus comes from bus_get_cpus(dev, INTR_CPUS, sizeof(ctx->ifc_cpus), &ctx->ifc_cpus), unless that call fails, and then it's just used uninitialized (but M_ZERO from iflib_device_register).

So... does bus_get_cpus(..., INTR_CPUS, ...) return absent CPUs?

[nwhitehorn]

sys/net/iflib.c

4306–4308

It shouldn't, but it seems good to check out of paranoia in some KASSERT -- we've had so many bugs like that in other parts of the kernel. I don't think it's something you need to check for beyond an assertion.

[cem]

Add !CPU_ABSENT KASSERTs as well.

[cem]

So, I don't think the CPU_FFS stuff is actually causing the issue from the PR, unfortunately.

[cem]

Log the gtask cpu index we failed to find a taskgroup for on panic.

[cem]

@sbruno committed something similar in r307560, r307561, r307566, r307567, obsoleting this review.