On Wed, Apr 18, 2018 at 5:38 PM, Andi Kleen <ak@xxxxxxxxxxxxxxx> wrote: > On Wed, Apr 18, 2018 at 05:02:02PM -0700, Linus Torvalds wrote: >> On Wed, Apr 18, 2018 at 4:26 PM, Andi Kleen <ak@xxxxxxxxxxxxxxx> wrote: >> > >> > Seems like a hack. Why can't that be stored in a per cpu variable? >> >> It *is* a percpu variable - the whole x86_tss structure is percpu. >> >> I guess it could be a different (separate) percpu variable, but might >> as well use the space we already have allocated. > > Would be better/cleaner to use a separate variable instead of reusing > x86 structures like this. Who knows what subtle side effects that > may have eventually. This variable is extremely hot, and it’s used under the same circumstances as sp0, so sharing a cache line makes sense. And x86_64 works this way. > > It will be also easier to understand in the code. I suppose it could go right before the TSS, but then we have potential alignment issues. We could also muck with unions to give the field an alternative, clearer name, I suppose. But this patch should go in regardless and any cleanups should be done on x86_32 and x86_64 simultaneously, I think.
