On Tue, Apr 03, 2018 at 10:50:59AM +0200, Pavel Machek wrote: > > gcc already does some nice optimisations around free(). For example, it > > can eliminate dead stores: > > Are we comfortable with that optimalization for kernel? > > us: "Hey, let's remove those encryption keys before freeing memory." > gcc: :-). > > us: "Hey, we want to erase lock magic values not to cause confusion > later." > gcc: "I like confusion!" > > Yes, these probably can be fixed by strategic "volatile" and/or > barriers, but...
Exactly, we should mark those sites explicitly with volatile so that they aren't dead stores.
