2.4.23 AMD k6-2 init fails with paging based non-executable

[Razathorn]

Paging based non-executable pages makes my machine unable to boot. It's an old clunker k6-2 350mhz box. Using the segmentation based and turning off paging based caused it to boot fine. It looks like it fails when it is about to run init.

Debian Woody
Linux version 2.4.23-grsec (root@eggs) (gcc version 2.95.4 20011002 (Debian prerelease)) #5 Mon Dec 8 15:24:04 CST 2003

I don't think I have done anything crazy to my debian install to get the version of gcc I have so I would imagine it's the default one for stable debian.

I wasn't able to find anything that would give me the impression that it would break a k6-2. Perhaps this is just a fact of life for this cpu -- if so, a tidbit in the help section or something in the config that won't let you pick that option if the config is set on a k6-2 amd.

Thoughts?

[Sleight of Mind]

what do you mean by 'unable to boot'? Where do things get stuck?
Please give us some more information.
Afaik a k6-2 should run just fine with any of the grsec options enabled.

[PaX Team]

I wasn't able to find anything that would give me the impression that it would break a k6-2. Perhaps this is just a fact of life for this cpu -- if so, a tidbit in the help section or something in the config that won't let you pick that option if the config is set on a k6-2 amd.

indeed, this is the case (the TLB of the k6 is just not good enough) and if you were using the PaX patch itself, you could not select PAGEEXEC there. for some reason spender simplified the config.in safeguards in grsec, this is one part of the 'collateral damage' only, 386/486 and RANDKSTACK is not a winning combination either (he already knows about the issue, so hopefully it will be fixed soon).