Re: [PATCH 6/8] KVM: arm64: Add sys_reg_descs for remaining MTE CMO instructions

[Leonardo Bras <leo.bras@arm.com>]

On Mon, Nov 09, 2026 at 09:48:21AM -0800, Oliver Upton wrote:
> On Mon, Nov 09, 2026 at 05:28:20PM +0000, Leonardo Bras wrote:
> > On Wed, Nov 04, 2026 at 11:21:22AM -0800, Oliver Upton wrote:
> > > Filtering unsupported cache maintenance instructions from a guest relies
> > > on either having a suitable FGU bit for the feature or a descriptor with
> > > a visibility() hook.
> > > 
> > > MTE CMO instructions can be trapped by HCR_EL2.TPCP and lack a suitable
> > > FGU. Add descriptors for the CMO instructions that UNDEF if MTE isn't
> > > implemented for the VM.
> > > 
> > 
> > I think I get the idea of using MTE_CMO_INSN(): having the visibility test 
> > for MTE defined in the previous patch, which makes sense with the above 
> > commit message: if !mte -> hidden, then on access it injects undefined.
> > 
> > Please help me understand the other way: if we have mte, then it calls the 
> > 'access' callback, which is bad_cmo_trap(), which also inject undefined. 
> > What is the goal here?
> 
> This is consistent with how we handle other unexpected traps, such as
> registers that redirect into the VNCR. If we take a trap for one of
> these instructions then either we UNDEF it early in triage or redirect
> the trap to the L1 hypervisor.
> 
> Under no circumstance do we expect to handle the trap locally, i.e.
> reach the ->access() function.
> 
> Thanks,
> Oliver
>  

Ah, I see your point there :)

FWIW:
Reviewed-by: Leonardo Bras <leo.bras@arm.com>

Thanks!
Leo