The new sys-kernel/hardened-sources-2.6.27-r5 dies immediately at the beginning with amd64 (core2). No such problem with x86 or with 2.6.27-r4. Instead of the kernel name it writes at the bottom of the page: Kernel alive Kernel really alive PANIC: early exception 0e rip 10:ffffffff80353e79 error 0 cr2 ffffffffff300000 Then it waits for a reset...
Post your emerge --info and kernel config please. Can you reproduce this with gentoo-sources-2.6.27-r8? There was only one minor change to the PaX patch in 2.6.27-r5 vs. 2.6.27-r4 besides the fix for bugs 253733 and 254843 (which is not related or the cause of this). Might be interested in these: https://bugs.gentoo.org/show_bug.cgi?id=250695 http://bugzilla.kernel.org/show_bug.cgi?id=12236 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/54020
Created attachment 179594 [details] .config (most comments deleted)
Created attachment 179595 [details] Output of env -i emerge --info (some unrelated variables removed)
With gentoo-sources-2.6.27-r8 the problem does not occur. Since also with hardened-sources-2.6.27-r4 the problem did not occur, I doubt that the above links are related either (although the error message is similar).
Fixed in 2.6.27-r6. Thanks for reporting and thanks to Sadako on IRC for also reproducing the bug and testing patches.
Thanks. I can confirm that -r6 fixes the problem. And sorry for attaching the wrong .config version (with comments not removed).
can you guys test the latest PaX test patches as well please?
I had this error with the new 2.6.27-r8 hardened kernel on a single core amd64.
In the same boat as Agoston, just tried to use the newly-unmasked 2.6.27-r8 on amd64 and got this: Kernel alive Kernel really alive PANIC: early exception 0e rip 10:ffffffff80394c59 error 0 cr2 ffffffffff300000
The bug is not fixed! sys-kernel/hardened-sources-2.6.27-r8 Kernel alive Kernel really alive PANIC: early exception 0e rip 10: ... Same problem with dual Xeon
This was fixed in sys-kernel/hardened-sources-2.6.27-r7 as I dropped the offending patch. The bug was mistakenly re-introduced in sys-kernel/hardened-sources-2.6.27-r8 as I added all the latest PaX test patches (we believed the issue to be resolved in later PaX patches). I have removed the amd64 keyword for 2.6.27-r8. PaX Team: grsecurity patch used in 2.6.27-r8: http://sources.gentoo.org/viewcvs.py/linux-patches/hardened/2.6/tags/2.6.27-9/4420_grsec-2.1.12-2.6.27.19-200902231736.patch?rev=1504&view=log description of patch included here: http://sources.gentoo.org/viewcvs.py/*checkout*/linux-patches/hardened/2.6/tags/2.6.27-9/0000_README?rev=1504
Could someone who can reproduce this issue test the latest PaX patch again a vanilla 2.6.27.20 kernel? Link: http://www.grsecurity.net/~paxguy1/pax-linux-2.6.27.20-test34.patch
All 2.6.27 series hardened-sources have been removed, none of the remaining kernels in the tree display this issue. Closing bug. If someone is willing to test, I may add back a new 2.6.27 kernel.
