On Tue Feb 20, 2024 at 11:10 PM UTC, Jarkko Sakkinen wrote > On Tue Feb 20, 2024 at 10:57 PM UTC, wrote: > > On 2/20/24 2:26 PM, Jarkko Sakkinen wrote: > > > On Tue Feb 20, 2024 at 8:54 PM UTC, Lino Sanfilippo wrote: > > >> for (i = 0; i <= MAX_LOCALITY; i++) > > >> __tpm_tis_relinquish_locality(priv, i); > > > > > > I'm pretty unfamiliar with Intel TXT so asking a dummy question: > > > if Intel TXT uses locality 2 I suppose we should not try to > > > relinquish it, or? > > > > The TPM has five localities (0 - 4). Localities 1 - 4 are for DRTM > > support. For TXT, locality 4 is hard wired to the CPU - nothing else can > > Locality 4 is familiar because it comes across from time to time. > > If I recall correctly DRTM should use only localities 3-4 and > localities 0-2 should be reserved for the OS use. > > So this does not match what I recall unfortunately but I'm not > really expert with this stuff. > > The patches has zero explanations SINIT ACM's behaviour on > locality use and without that this cannot move forward because > there's neither way to reproduce any of this. > > Actually there's zero effort on anything related to SINIT. To put short we need a clearer sequence what Intel TXT does causing leaving localities does. If that is nailed to less abstract description then we can review this. If we know this blackbox model then there's a chance to make simulation of it e.g. with QEMU. Alternatively we need a more trivial scenario to reproduce a bug in locality handling than Intel TXT. There's no enough beef ATM to really make good decisions what sort of code change would be best for Linux. BR, Jarkko
