Following the (so far not really explicable) observation in [1], plus
the fact that mca_rdmsr() isn't simply an RDMSR, patch 1 looked like
something we may want to have. While putting that together, another
inefficiency popped up, which patch 2 covers.

1: reduce number of MSR reads/lookups
2: drop redundant code from mcheck_mca_logout()

Jan

[1] https://lists.xen.org/archives/html/xen-devel/2026-10/msg01825.html