On Sun, Feb 06, 2005 at 02:31:10PM -0800, David S. Miller wrote: > This is kind of strange to be arguing about, given that we just > put 4-level page tables into the tree, right? That regressed > everybody performance wise, even people not using the full
Hmm, are you sure? I benchmarked i386 and there was no measurable difference. x86-64 is regressed right now for fork etc., but that is to be expected because it walks twice as many page tables. And there is a plan to fix it. Text size has increased a little bit for all people, but not too bad. My benchmarking was with the PML4 patchkit though, I haven't retried it with PUD. > And it's not just a sparc64 issue. Sparc64 hardware traps the > access, but it's a bug regardless of platform to try to do user > accesses whilst get_fs()==KERNEL_DS. All the user has to do is On x86-64 it's not a bug and we have code in the x86-64 specific compat layer that relies on this. Also BTW your proposed check wouldn't catch all cases anyways, it would only handle the case where the access_ok() check is done inside the KERNEL_DS. -Andi
