[patch 0/5] early params NULL pointer deref fix

[patch 0/5] early params NULL pointer deref fix

[Cyrill Gorcunov]

 This series fix a possible NULL pointer deref in case if user
was stupid enough to pass boot parameter without a mandatory value

 Actually there is a number of such issues on other platforms too
(like powerpc, sh, s390) so I think it should be fixed too (any
volunteers?)

 Any comments are welcome.

		- Cyrill -
-- 

Re: [patch 0/5] early params NULL pointer deref fix

[Ingo Molnar]

* Cyrill Gorcunov <gorcunov@gmail.com> wrote:

>  This series fix a possible NULL pointer deref in case if user was 
> stupid enough to pass boot parameter without a mandatory value
> 
>  Actually there is a number of such issues on other platforms too 
> (like powerpc, sh, s390) so I think it should be fixed too (any 
> volunteers?)
> 
>  Any comments are welcome.

thanks Cyrill, i picked them up.

I'm wondering, would it be possible to filter NULL out on one higher 
level? Would it break any existing functionality? (i dont think it would 
or should)

	Ingo

Re: [patch 0/5] early params NULL pointer deref fix

[Cyrill Gorcunov]

[Ingo Molnar - Wed, Jul 09, 2008 at 01:57:48PM +0200]
| 
| * Cyrill Gorcunov <gorcunov@gmail.com> wrote:
| 
| >  This series fix a possible NULL pointer deref in case if user was 
| > stupid enough to pass boot parameter without a mandatory value
| > 
| >  Actually there is a number of such issues on other platforms too 
| > (like powerpc, sh, s390) so I think it should be fixed too (any 
| > volunteers?)
| > 
| >  Any comments are welcome.
| 
| thanks Cyrill, i picked them up.
| 
| I'm wondering, would it be possible to filter NULL out on one higher 
| level? Would it break any existing functionality? (i dont think it would 
| or should)
| 
| 	Ingo
| 

well, actually it would and should :)

Currently parser does a simple thing - just parses command line and split
key/param pairs. But kernel code uses it in three ways:

 1) as simple flag specified on booting
 2) as key and value
 3) combination of 1) and 2)

So for now I don't see any simple solution for this. We could use '='
param postfix to specify that a mandatory value required for param and
throw out any misconfigured boot options but it will make way 3) invalid
so kernel code will have to be fixed then. So some additional review
and conversation about this are needed.

		- Cyrill -