On Fri, Jan 18, 2019 at 11:28:24PM +1100, Michael Ellerman wrote: > Segher Boessenkool <seg...@kernel.crashing.org> writes: > > > On Thu, Jan 17, 2019 at 11:13:26PM +1100, Michael Ellerman wrote: > >> The recent rewrite of the SLB code into C included the assumption that > >> all CPUs we run on have at least 32 SLB entries. This is currently > >> true but a bit fragile as the SLB size is actually defined by the > >> device tree and so could theoretically change at any time. > > > > It also is guaranteed by the architecture, since at least 2.02, FWIW. > > True. Actually 2.00 says at least 32. > > Unfortunately we don't live in a world where "the architecture > guarantees it" has any bearing on reality :)
It's a pretty strong hint. I don't remember any hardware where it is not true, either. (That might be selective memory ;-) ) > But given it *should* always be at least 32 maybe I should optimise for > that case. We could use a static key to skip the U32_MAX comparison and > go down the else path. Ah that sounds like a good idea :-) Segher
