On Thu, Mar 02 2023, Peter Maydell <peter.maydell@xxxxxxxxxx> wrote: > On Thu, 2 Mar 2023 at 14:29, Cornelia Huck <cohuck@xxxxxxxxxx> wrote: >> >> On Thu, Mar 02 2023, Peter Maydell <peter.maydell@xxxxxxxxxx> wrote: >> > I think having MTE in the specific case of KVM behave differently >> > to how we've done all these existing properties and how we've >> > done MTE for TCG would be confusing. The simplest thing is to just >> > follow the existing UI for TCG MTE. >> > >> > The underlying reason for this is that MTE in general is not a feature >> > only of the CPU, but also of the whole system design. It happens >> > that KVM gives us tagged RAM "for free" but that's an oddity >> > of the KVM implementation -- in real hardware there needs to >> > be system level support for tagging. >> >> Hm... the Linux kernel actually seems to consider MTE to be a cpu >> feature (at least, it lists it in the cpu features). >> >> So, is your suggestion to use the 'mte' prop of the virt machine to mean >> "enable all prereqs for MTE, i.e. allocate tag memory for TCG and enable >> MTE in the kernel for KVM"? For TCG, we'll get MTE for the max cpu >> model; for KVM, we'd get MTE for host (== max), but I'm wondering what >> should happen if we get named cpu models and the user specifies one >> where we won't have MTE (i.e. some pre-8.5 one)? > > I think we can probably cross that bridge when we get to it, > but I imagine the semantics would be "cortex-foo plus MTE" > (in the same way that -cpu cortex-foo,+x,-y can add and > subtract features from a baseline). While implementing this, I realized another thing that I had managed to miss before: With tcg, we'll start out with mte=3 and downgrade to mte=0 if we don't have tag memory. With kvm, enabling mte can at most give us the mte version that the host exposes, so setting mte=on for the machine might give as well mte=2 in the end [which I still need to implement by querying the host support, I guess.] This means we have slightly different semantics for tcg and kvm; but more importantly, we need to implement something for compat handling. The Linux kernel distinguishes between 'mte' and 'mte3', and KVM exposes the MTE cap if mte >=2. Do we need two props as well? If yes, what about tcg?
