Re: [BUGFIX] NULL pointer dereference in __vm_enough_memory()

[Cyrill Gorcunov <gorcunov@gmail.com>]

[WU Fengguang - Mon, Aug 13, 2007 at 08:23:42AM +0800]
| On Sun, Aug 12, 2007 at 08:21:43PM +0400, Cyrill Gorcunov wrote:
| > [Alan Cox - Sun, Aug 12, 2007 at 04:17:44PM +0100]
| > | Try this (it compiles but isnt tested). Its a weekend here, the sun is
| > | shining, the beach is a short walk, and I have more interesting things to
| > | do right now 8)
| > | 
| > | 
| [...]
| > | -int __vm_enough_memory(long pages, int cap_sys_admin)
| > | +int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin)
| > |  {
| > |  	unsigned long free, allowed;
| > |  
| > | @@ -166,7 +166,7 @@
| > |  
| > |  	/* Don't let a single process grow too big:
| > |  	   leave 3% of the size of this process for other processes */
| > | -	allowed -= current->mm->total_vm / 32;
| > | +	allowed -= mm->total_vm / 32;
| > 
| > So mm->total_vm is 0 for __bprm_mm_init case. Is that ok? Or I miss
| > something?
| 
| Yeah, Alan adds mm to the interfaces and leaves us the question of
| "what mm to pass in when current->mm == NULL?" ;)
| 

Well, as I see, it seems the Alan's patch is correct. We pass
newly created mm to security_vm_enough_memory_mm() and get no errors
here even for overcommit = 2. But my question was that mm->total_vm
= 0 for this case and that is probably valid too I think. What about
the thing you pointed about? Well I think security_vm_enough_memory
should never be called from kernel thread (we have secrurity_vm_enough_memory_mm
for this). But I will check it more closely. Dont get me wrong - I'm not
VMM expert and may do errors ;)

		Cyrill