Re: [PATCH v3 3/5] x86/mm: Optionally flush L1D on context switch

[Singh, Balbir]

On Fri, 2020-12-04 at 22:21 +0100, Thomas Gleixner wrote:
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.
> 
> 
> 
> On Fri, Nov 27 2020 at 17:59, Balbir Singh wrote:
> > 
> > +     /*
> > +      * Flush only if SMT is disabled as per the contract, which is checked
> > +      * when the feature is enabled.
> > +      */
> > +     if (sched_smt_active() && !this_cpu_read(cpu_info.smt_active) &&
> > +             (prev_mm & LAST_USER_MM_L1D_FLUSH))
> > +             l1d_flush_hw();
> 
> So if SMT is completely disabled then no flush? Shouldn't the logic be:
> 
>     if ((!sched_smt_active() || !this_cpu_read(cpu_info.smt_active) &&
>          (prev_mm & LAST_USER_MM_L1D_FLUSH))
> 
> Hmm?
> 
> But that's bad, because it's lot's of conditions to evaluate for every
> switch_mm where most of them are not interested in it at all.
> 
> Let me read through the rest of the pile.
>


We don't need this anymore with the new checks for preempting killing
of the task, so it can be removed

Balbir