softirq.c bug

I'm using 2.5.67, on a duron 1200 box , my keyboard lock up when a ppp connection goes down and also
pppd process won't die and kill, and the box don't event shutdown! I also found that my ethernet (rtl8139 NIC) don't work while CONFIG_ACPI=y it's response to other stations arp requests but don't add others to arp table

Apr 17 17:44:58 pars1 kernel: kernel BUG at kernel/softirq.c:105!
Apr 17 17:44:58 pars1 kernel: invalid operand: 0000 [#1]
Apr 17 17:44:58 pars1 kernel: CPU:    0
Apr 17 17:44:58 pars1 kernel: EIP:    0060:[<c011e575>]    Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
Apr 17 17:44:58 pars1 kernel: EFLAGS: 00010002
Apr 17 17:44:58 pars1 kernel: eax: 00000001   ebx: cdf30000   ecx: 00000001  
edx: 00000000
Apr 17 17:44:58 pars1 kernel: esi: 00000000   edi: 00000000   ebp: 00000001   
esp: cdf31ef4
Apr 17 17:44:58 pars1 kernel: ds: 007b   es: 007b   ss: 0068
Apr 17 17:44:58 pars1 kernel: Stack: cdc0d800 ce83ec02 c8f71000 00000000 
cdc0d894 00000049 c8f71000
00000049
Apr 17 17:44:58 pars1 kernel:        c8f71000 cdc0d800 c8f719e4 cdf30000 
ce83e501 cdc0d800 c8f71000
00000246
Apr 17 17:44:58 pars1 kernel:        ce927a31 c8f71000 c8f71000 c01c0c5d 
c8f71000 c8f810c0 00000000
00000002
Apr 17 17:44:58 pars1 kernel:  [<ce83ec02>] ppp_async_push+0xa2/0x190 
[ppp_async]
Apr 17 17:44:58 pars1 kernel:  [<ce83e501>] ppp_asynctty_wakeup+0x31/0x70 
[ppp_async]
Apr 17 17:44:58 pars1 kernel:  [<ce927a31>] uart_flush_buffer+0x71/0x80 [core]
Apr 17 17:44:58 pars1 kernel:  [<c01c0c5d>] do_tty_hangup+0x35d/0x3b0
Apr 17 17:44:58 pars1 kernel:  [<c0129033>] worker_thread+0x1e3/0x2d0
Apr 17 17:44:58 pars1 kernel:  [<c01c0900>] do_tty_hangup+0x0/0x3b0
Apr 17 17:44:58 pars1 kernel:  [<c0117700>] default_wake_function+0x0/0x20
Apr 17 17:44:58 pars1 kernel:  [<c010aa4e>] ret_from_fork+0x6/0x14
Apr 17 17:44:58 pars1 kernel:  [<c0117700>] default_wake_function+0x0/0x20
Apr 17 17:44:58 pars1 kernel:  [<c0128e50>] worker_thread+0x0/0x2d0
Apr 17 17:44:58 pars1 kernel:  [<c0108abd>] kernel_thread_helper+0x5/0x18
Apr 17 17:44:58 pars1 kernel: Code: 0f 0b 69 00 c4 f8 28 c0 eb c2 90 53 89 c1 

9c 5b fa b8 01 00

Code; c011e575 <local_bh_enable+55/60>
00000000 <_EIP>:
Code; c011e575 <local_bh_enable+55/60> <=====
0: 0f 0b ud2a <=====
Code; c011e577 <local_bh_enable+57/60>
2: 69 00 c4 f8 28 c0 imul $0xc028f8c4,(%eax),%eax
Code; c011e57d <local_bh_enable+5d/60>
8: eb c2 jmp ffffffcc <_EIP+0xffffffcc>
Code; c011e57f <local_bh_enable+5f/60>
a: 90 nop
Code; c011e580 <raise_softirq+0/40>
b: 53 push %ebx
Code; c011e581 <raise_softirq+1/40>
c: 89 c1 mov %eax,%ecx
Code; c011e583 <raise_softirq+3/40>
e: 9c pushf
Code; c011e584 <raise_softirq+4/40>
f: 5b pop %ebx
Code; c011e585 <raise_softirq+5/40>
10: fa cli
Code; c011e586 <raise_softirq+6/40>
11: b8 01 00 00 00 mov $0x1,%eax

I submitted the below (sad) patch yesterday:

This BUG_ON is triggering via ppp's line discipline flushing, due to brokenness in tty_io.c.

We need to fix tty. Meanwhile, let's not gratuitously nuke people's boxes.

 25-akpm/kernel/softirq.c |    2 +-
 1 files changed, 1 insertion(+), 1 deletion(-)

diff -puN kernel/softirq.c~local_bh_enable-niceness kernel/softirq.c
@@ -102,7 +102,7 @@ restart:
 void local_bh_enable(void)
 {
 	__local_bh_enable();
-	BUG_ON(irqs_disabled());
+	WARN_ON(irqs_disabled());
 	if (unlikely(!in_interrupt() &&
 		     local_softirq_pending()))
 		invoke_softirq();

Please log this in http://bugzilla.kernel.org