On Mon, 24 Aug 2015 13:52:30 -0400 (EDT), Mirko Parthey wrote: > > On Mon, Aug 24, 2015 at 06:45:10AM -0400, Stephen Powell wrote: >> Furtherfore, a non-PAE kernel is >> useful even on PAE-capable hardware. The main purpose of PAE is to >> address memory above 4G. But if the machine has less than 4G of >> memory, what does a PAE-capable kernel buy you? PAE-capable kernels >> tend to be a bit bigger, all other things being equal, than non-PAE >> kernels, which chews up more precious memory with no obvious benefit. > > One such benefit is that the NX bit (non-executable memory pages) > is only available with 64 bit page table entries, which in turn depend on > PAE mode. This could be an argument for preferring a PAE kernel on > PAE-capable hardware.
Wow, did I really say, "Furtherfore"? Mercy! I need to do a better job of proofreading. Obviously, that was supposed to be "Furthermore". Anyway, to your point: PAE may be a necessary condition for NX, but it is not a sufficient condition. I am presently using three 32-bit computers: one is a Pentium M (2G), one is a Pentium 4 (2G), and one is a Xeon (4G). All three are PAE-capable, and all three are presently running PAE Linux kernels. And all three display the following message during boot, according to "dmesg|less": Notice: NX (Execute Disable) protection missing in CPU! So, for my hardware, that argument doesn't seem to hold up. Am I missing something? -- .''`. Stephen Powell <zlinux...@wowway.com> : :' : `. `'` `-
