'Badness in local_bh_enable' error with grsecurity enabled

Discuss usability issues, general maintenance, and general support issues for a grsecurity-enabled system.

'Badness in local_bh_enable' error with grsecurity enabled

Postby altruist » Tue May 09, 2006 12:50 am

I get some weird errors with grsecurity enabled.. any idea what's causing it, or what option I need to disable? Everything seems fine, except these messages appear in syslog/dmesg. When I disable grsecurity and leave pax enabled, the errors disappear. I've tried several kernels.. I'm on my ~15th compile now, I'm still not sure what causes it.

Badness in local_bh_enable at kernel/softirq.c:140
[<c013fef1>]
[<c014419c>]
[<c013cafd>]
[<c013db37>]
[<c013defd>]
[<c013df56>]
[<c0127759>]
Badness in local_bh_enable at kernel/softirq.c:140
[<c013fef1>]
[<c014419c>]
[<c013cafd>]
[<c013db37>]
[<c013defd>]
[<c013df56>]
[<c0127759>]
[<c0127771>]

(both same kernel compile)

From another compile:
Badness in local_bh_enable at kernel/softirq.c:140
[<c013fa75>] local_bh_enable+0x2a/0x78
[<c0143d20>] __exit_signal+0xbf/0x191
[<c013c72d>] release_task+0x5f/0x126
[<c013d767>] exit_notify+0x28d/0x295
[<c013da7f>] do_exit+0x310/0x346
[<c013dad8>] next_thread+0x0/0xc
[<c0127769>] syscall_call+0x7/0xb
[<c01269eb>] pax_randomize_kstack+0x5/0x3a
[<c0127781>] restore_all+0x0/0x18
altruist
 
Posts: 2
Joined: Tue May 09, 2006 12:48 am

Re: 'Badness in local_bh_enable' error with grsecurity enabl

Postby PaX Team » Tue May 09, 2006 5:53 am

altruist wrote:Badness in local_bh_enable at kernel/softirq.c:140
[<c013fef1>]
[<c014419c>]
[<c013cafd>]
[<c013db37>]
[<c013defd>]
[<c013df56>]
[<c0127759>]
Badness in local_bh_enable at kernel/softirq.c:140
[<c013fef1>]
[<c014419c>]
[<c013cafd>]
[<c013db37>]
[<c013defd>]
[<c013df56>]
[<c0127759>]
[<c0127771>]
can you decode these as well please (if you still have the corresponding System.map file)?
Badness in local_bh_enable at kernel/softirq.c:140
[<c013fa75>] local_bh_enable+0x2a/0x78
[<c0143d20>] __exit_signal+0xbf/0x191
[<c013c72d>] release_task+0x5f/0x126
[<c013d767>] exit_notify+0x28d/0x295
[<c013da7f>] do_exit+0x310/0x346
[<c013dad8>] next_thread+0x0/0xc
[<c0127769>] syscall_call+0x7/0xb
[<c01269eb>] pax_randomize_kstack+0x5/0x3a
[<c0127781>] restore_all+0x0/0x18
in grsecurity/grsec_sock.c:gr_del_task_from_ip_table() remove _bh from the spin_lock_bh/spin_unlock_bh calls and it should work (and be still correct).
PaX Team
 
Posts: 2310
Joined: Mon Mar 18, 2002 4:35 pm

Postby altruist » Tue May 09, 2006 11:05 am

I reinstalled one of the kernels that produced the errors without details. I tried to decode them for you, but oddly enough, only one of them appear in the corresponding /boot/System.map for the kernel. (Assuming /boot is where I'm supposed to look?)

c013df56 T next_thread

Edit: I did the source code modifications and recompiled, the errors seemed to be surpressed now even with the highest security settings.
altruist
 
Posts: 2
Joined: Tue May 09, 2006 12:48 am

Also happening here

Postby stilgar » Sun Jun 25, 2006 7:37 pm

I'm compiling on a debian-unstable system, 2.6.16.14 with grsec 2.6.16 patch and gcc 4.1.2 20060613, getting this error:
Badness in local_bh_enable at kernel/softirq.c:140
[<c0146ad1>] local_bh_enable+0x61/0x70
[<c014c197>] __exit_signal+0x57/0x180
[<c0143cbe>] release_task+0x3e/0x100
[<c0144cc0>] do_exit+0x460/0x760
[<c014504e>] sys_exit+0xe/0x10
[<c012bca9>] syscall_call+0x7/0xb
stilgar
 
Posts: 1
Joined: Sun Jun 25, 2006 7:35 pm

Re: Also happening here

Postby PaX Team » Mon Jul 03, 2006 3:26 pm

stilgar wrote:I'm compiling on a debian-unstable system, 2.6.16.14
try a recent grsecurity patch.
PaX Team
 
Posts: 2310
Joined: Mon Mar 18, 2002 4:35 pm


Return to grsecurity support

cron