grsecurity forums

I have an eeepc with Intel Atom N270. Using grsecurity-2.2.2-2.6.32.41 the kernel boots fine. The next I tried was grsecurity-2.2.2-2.6.32.51. This one fails to boot right in the first part of the bootprocess, printing unreadable lines of text on the screen indefinetely. The same Kernel boots on a core2 just fine.

I compiled with gcc-4.5.3, trying both atom and 386 as processor type in menuconfig.

An non grsec patched Kernel boots fine on the Atom.

Unfortunately I have not enough knowledge of the bootprocess to look into the changes from 2.6.32.41 to 2.6.32.51 myself.

Thanks for hints,

thorsten

fly_a320 wrote:I have an eeepc with Intel Atom N270. Using grsecurity-2.2.2-2.6.32.41 the kernel boots fine. The next I tried was grsecurity-2.2.2-2.6.32.51. This one fails to boot right in the first part of the bootprocess, printing unreadable lines of text on the screen indefinetely. The same Kernel boots on a core2 just fine.

i'll need the .config and vmlinux and bzImage files (we have a wiki page on how to submit bugs btw ;P). if you have time to play with various configurations, you could start by disabling all of PaX and see which feature causes the problem (it's possible that even with all options disabled you get the problem, some changes are not under config control). next, capturing the boot messages would be really helpful, maybe you can try netconsole?

Thanks for having a look at this issue.

I will send an email with a download link for the kernel.


I will play with Kernel Options and try netconsole, however the kernel dies very early, I do not think the networking stack will be ready at this stage.

the first update:

Trying a GRSEC patched Kernel with all GRSEC and PAX features disabled also hangs at the same early stage in the bootprocess. The kernel just prints the usual uncompressing and booting the kernel, right after that only scrolling of text indefinetely.

The 3.2.10-201201192026 patch works fine on a N270 with gcc 4.4.5 (debian stable).
I'm just upgrading my old EEE to see what the influence of the compiler is, since of course the 4.4.5-version does not support the plugins.

I had no trouble booting up a 2.6.32.54-kernel with the 201201192026-grsecurity patch...

However: running 2.6.32.54 I could not run lilo (I first suspected severe LILO/UUID problems):

Code: Select all

Fatal: raid_setup: stat("/dev/sda")

After upgrading debian stable to debian testing + 2.6.32.54 there was no /dev/disk tree and the old way of using /dev/sda did not work.
Rebooting to 3.1.10 + 201201192026-grsecurity patch (earlier debian stable, now testing) I could run lilo and install the 2.6.32.54-kernel.
The problems were the same with both the gcc-4.4.5 compiled version as with the gcc-4.6.2 compiled version.

Could this be a similar problem as in thread viewtopic.php?f=3&t=2651&p=10781&hilit=lilo#p10781 ?
I don't have any information about the problem the original poster ran into, but it might not be related to my problem.

PS since about half a year I changed my lilo-configuration to stick to /dev/sda instead of UUID addressing of hard drives.

This boot failure should be fixed in the latest patch just released. It only affected the 2.6.32 series.

Thanks,
-Brad

Thanks for fixing this issue!