Re: [PATCH 1/2] x86/kmsan: Fix CPU entry area metadata lookup

[Peter Zijlstra]

On Tue, Oct 27, 2026 at 04:28:36PM +0100, Alexander Potapenko wrote:
> On Tue, Oct 6, 2026 at 8:10 PM Viorel Cernateanu via B4 Relay
> <devnull+vrilutza.gmail.com@xxxxxxxxxx> wrote:
> >
> > From: Viorel Cernateanu <vrilutza@xxxxxxxxx>
> >
> > arch_kmsan_get_meta_or_null() works out which CPU owns an address in the
> > CPU entry area as (addr - CPU_ENTRY_AREA_BASE) / CPU_ENTRY_AREA_SIZE.
> > That is wrong in two ways:
> >
> >  - With KASLR, init_cea_offsets() puts each CPU's area in a random slot,
> >    so the result is a slot number, usually far above NR_CPUS, and
> >    get_cpu_entry_area() reads past the end of __per_cpu_offset[].
> 
> Thanks for noticing this!
> I find the O(n) lookup per each memory access to the cpu_entry_area
> excessive. Could we disable the randomization in
> arch/x86/mm/cpu_entry_area.c
> 
>         if (!kaslr_enabled() || IS_ENABLED(CONFIG_KMSAN)) {
>                 for_each_possible_cpu(i)
>                         per_cpu(_cea_offset, i) = i;
>                 return;
>         }
> 
> (there's a similar precedent for KASAN in arch/x86/include/asm/setup.h already)
> 
> Peter, what do you think?

I suppose that works; KMSAN is a debug kernel per definition and I don't
think we really care all that much about kaslr on debug kernels (I
certainly have it disabled on all my dev boxes because its a pain).