[rsbac] Newbie: 2.4.18+1.2.1 doesn't start, complains of invalid device

Amon Ott rsbac@rsbac.org
Mon Oct 7 09:32:01 2002

On Saturday, 5. October 2002 06:16, Tomasz Korycki wrote:
> A newbie here, but I'll try to make it brief: RSBAC kernel never starts, 
> complains about invalid device until syslog fills /var. Here's how I got 

That looks fine - except the version: I never made a 1.2.1-final patch for 

> 03:12 is the root partition (/dev/hda12, reiserfs 3.x, reiserfs is compiled 
> into the kernel), /dev/hda1 is /boot, ext2
> OK, I'm stuck. Here is (heavily pruned) copy of the last /var/log/messages:
> ----------------------------- start -------------------------
> Oct  4 17:03:17 switch syslogd 1.4-0: restart.
> Oct  4 17:03:17 switch kernel: klogd 1.4-0, log source = /proc/kmsg started.
> Oct  4 17:03:17 switch kernel: Inspecting /boot/System.map
> Oct  4 17:03:17 switch kernel: Loaded 5 symbols from 1 module.
> Oct  4 17:03:17 switch kernel: INVALIDDEV!
> Oct  4 17:03:17 switch kernel: rsbac_get_attr(): Could not lookup device 
> Oct  4 17:03:17 switch kernel: rsbac_adf_request(): rsbac_get_attr() for 
> internal returned EINVALIDDEV!
> ------------------- end --------------

Somehow the RSBAC initialization either failed to add the root device to its 
internal device list, or some strange umount stuff happened.

It might be related to reiserfs, which is weird in many respects.

> ANd that's how it continues. Every once in a while there are some other 
> messages, either indicating another rc script completion, or other RSBAC 
> complaints like:
> -------------------- start -------------
> Oct  4 17:03:17 switch kernel: rsbac_adf_set_attr_auth(): rsbac_get_attr() 
> returned error!
> Oct  4 17:03:17 switch kernel: rsbac_adf_set_attr(): request EXECUTE, 
> caller_pid 1426, target-type FILE, tid Device 03:12 Inode 13299 Path 
> /sbin/consoletype, new_target-type NONE, new_tid NONE, attr none, value 0, 
> error -1003
> Oct  4 17:03:17 switch kernel: do_execve() [sys_execve]: 
> rsbac_adf_set_attr() returned error
> -------------------- end ---------------

When a module cannot get its attributes, access is denied and notification 
fails. This is the expected behaviour.

> Questions:
> 1. Has anyone seen this? I went through the archive, but couldn't find 
> anything like this - which month/year did I skip?
> 2. Is there something I am missing, something bleeding obvious?
> 3. Is there anything more I should do that would tell me (and You) where 
> does the problem originate?
> 4. Finally: is it a problem with software, or just with my lack of patience 
> and /var space?

Without softmode enabled, please do the following:

1. check, if your root partition gets added in rsbac_do_init() (see boot 
2. add the kernel parameter rsbac_debug_adf and check, whether the root fs 
gets umounted at some place.
3. If not, please also add rsbac_debug_stack, take one boot log until the 
first INVALIDDEV messages appear and send it to me. I need the init messages 
up to the part you included.

To get you up and running, another root fs, e.g. ext3, might help.