On Sun, 28 Jul 2013, Russell King - ARM Linux wrote: > We don't want to make the asm() itself volatile, because that means the > asm() statement can't be optimised away. > > Adding a memory clobber to that asm seems to place it more appropriately, > but again, that's not particularly desirable. > > Another solution would be to make both tlb_ops_need_broadcast and > cache_ops_need_broadcast both be a flag test, but that introduces a > memory load in all those paths no matter if we're running on SMP_ON_UP > or not. I considered adding a HWCAP_* bit under the theory that elf_hwcap may already be cache-resident. But thought that the loss of the flag space and possible memory access vs. the additional mrc wasn't worth it. Anyway, if either of you have a preference between these three, I'd be happy to update the patch. Or feel free to take it over from here if either of you prefer. - Paul -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
