2.5.40 bug related to virtual consoles

2.5.40 bug related to virtual consoles

[David Ashley]

Please cc me on replies.

I'm running 2.5.40 on a cutting edge via embedded C3 motherboard. When I
turn off CONFIG_FB and CONFIG_VGA_CONSOLE I get this during bootup:
Unable to handle kernel NULL pointer dereference at virtual address 00000004    
 printing eip:                                                                  
c01e5a1d                                                                        
*pde = 00000000                                                                 
Oops: 0000                                                                      
                                                                                
CPU:    0                                                                       
EIP:    0060:[<c01e5a1d>]    Not tainted                                        
EFLAGS: 00010246                                                                
eax: c6ee4a00   ebx: 00000000   ecx: 00000001   edx: 00000000                   
esi: 00000000   edi: 00000000   ebp: c0399dc0   esp: c116fe70                   
ds: 0068   es: 0068   ss: 0068                                                  
Process init (pid: 1, threadinfo=c116e000 task=c116c040)                        
Stack: c6ee4a00 00000001 c6ee4a00 00000000 c01e5af8 00000000 00000001 00000000  
       c6427000 00000000 00000401 c116fec0 c01e924f 00000000 c6427000 00000001  
       c01d8b58 c6427000 c6ed9120 01024180 c6427000 c6bba2d0 00000006 c116ff70  
Call Trace: [<c01e5af8>] [<c01e924f>] [<c01d8b58>] [<c014813d>] [<c013de06>] [<]
Code: ff 52 04 8b 83 c0 9d 39 c0 5a 59 66 83 78 22 00 75 17 ba 00               
 <0>Kernel panic: Attempted to kill init!                                       

This is inside drivers/char/vt.c in vc_allocate just after the visual_init
call.

Related configs are:
CONFIG_VT=y
CONFIG_VT_CONSOLE=y
CONFIG_HW_CONSOLE=y
CONFIG_SERIAL_8250_CONSOLE=y

If I turn either CONFIG_FB or CONFIG_VGA_CONSOLE back on it boots fine.
With CONFIG_FB I don't actually have any other FB related settings on and
I never actually use the frame buffer device.

Let me know if more information is needed. Thanks!
-Dave

Re: 2.5.40 bug related to virtual consoles

[Randy.Dunlap]

Regarding "2.5.40 bug related to virtual consoles":
http://marc.theaimsgroup.com/?l=linux-kernel&m=103403138113853&w=2

Hi David,

Did you have some (private) replies to this?
If so, is it solved?

There are console & framebuffer updates pending, but I
don't know if they would solve this problem (which is
quite easy to reproduce).

The vc (virtual console) code appears to expect (eh, require)
some console (data struct) to be present.  For now, if you
enable Frame Buffer support but don't select/enable any
FB devices, you'll just get the dummy console driver, and
the vc code will be happy (at least it booted without an
error for me on 2.5.44).  In my system (x86), this added
about 60 KB of code... :(

-- 
~Randy