On Mon, May 06, 2013 at 07:33:28AM -0700, Linus Torvalds wrote: > On Mon, May 6, 2013 at 1:55 AM, Peter Zijlstra <a.p.zijlstra@xxxxxxxxx> wrote: > > > > Doesn't i386 have all the funny per-cpu stuff too? So the only reason it still > > does the fugly stack based thing is because nobody could be arsed to do the > > work of converting it. > > Umm. That "fugly stack-based" thing is better than the per-cpu crap. > > The percpu stuff implies a memory load. The stack based thing gets > thread_info with pure register accesses. Much better. > > For "current()" the per-cpu thing may be better, but if you actually > need the thread-info (not the case here, but in other places), the > stack masking is superior when it works (ie when you don't have > multi-stack issues due to irq's etc) But you can do both right? Use per-cpu for current and stack frobbery for current_thread_info(). That said, ISTR some risky bits where the stack frobbery went awry due to irq-stacks which is the source for my feelings towards the stack frobbery. That and of course that i386 and x86-64 behave differently for no apparent reason. -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
