On Fri, Oct 16, 2020 at 01:12:26PM +0200, Peter Zijlstra wrote: > On Tue, Oct 13, 2020 at 11:31:45AM -0700, Dave Hansen wrote: > > > +/** > > > + * It should also be noted that the underlying WRMSR(MSR_IA32_PKRS) is not > > > + * serializing but still maintains ordering properties similar to WRPKRU. > > > + * The current SDM section on PKRS needs updating but should be the same as > > > + * that of WRPKRU. So to quote from the WRPKRU text: > > > + * > > > + * WRPKRU will never execute transiently. Memory accesses > > > + * affected by PKRU register will not execute (even transiently) > > > + * until all prior executions of WRPKRU have completed execution > > > + * and updated the PKRU register. > > > + */ > > > +void write_pkrs(u32 new_pkrs) > > > +{ > > > + u32 *pkrs; > > > + > > > + if (!static_cpu_has(X86_FEATURE_PKS)) > > > + return; > > > + > > > + pkrs = get_cpu_ptr(&pkrs_cache); > > > + if (*pkrs != new_pkrs) { > > > + *pkrs = new_pkrs; > > > + wrmsrl(MSR_IA32_PKRS, new_pkrs); > > > + } > > > + put_cpu_ptr(pkrs); > > > +} > > > > > > > It bugs me a *bit* that this is being called in a preempt-disabled > > region, but we still bother with the get/put_cpu jazz. Are there other > > future call-sites for this that aren't in preempt-disabled regions? > > So the previous version had a useful comment that got lost. Ok Looking back I see what happened... This comment... /* * PKRS is only temporarily changed during specific code paths. * Only a preemption during these windows away from the default * value would require updating the MSR. */ ... was added to pks_sched_in() but that got simplified down because cleaning up write_pkrs() made the code there obsolete. > This stuff > needs to fundamentally be preempt disabled, I agree, the update to the percpu cache value and MSR can not be torn. > so it either needs to > explicitly do so, or have an assertion that preemption is indeed > disabled. However, I don't think I understand clearly. Doesn't [get|put]_cpu_ptr() handle the preempt_disable() for us? Is it not sufficient to rely on that? Dave's comment seems to be the opposite where we need to eliminate preempt disable before calling write_pkrs(). FWIW I think I'm mistaken in my response to Dave regarding the preempt_disable() in pks_update_protection(). Ira