CVE-2026-98729

In the Linux kernel, the following vulnerability has been resolved:

x86/mm: Don't apply va align to hugetlb mappings on AMD F15h

get align mask() returns huge page mask align() for hugetlbfs, but get align bits() adds va align.bits regardless, so vm unmapped area() returns an address off the huge page boundary and unmap hugepage range() hits BUG ON(start & ~huge page mask(h)) at teardown.

This can be triggered on Carrizo and FX-8370E, both hstates.

Pass the file to get align bits() and skip the randomisation for hugetlbfs.

[ bp: Massage commit message. ]