Re: [PATCH][ATM] use rtnl_{lock,unlock} during device operations (take 2)

[Werner Almesberger <wa@almesberger.net>]

Mitchell Blank Jr wrote:
> so basically if atmsigd goes nuts in can easily stomp all
> over the kernel's memory.

Naw, it isn't supposed to do that :-) I wonder if anyone
actually made functional changes to atmsigd (or qgen ;-) since
I last touched it ...

But yes, with a unified VCC table, it certainly makes sense to
add a hash to validate those pointers. I still think that using
pointers per se is a good idea, because they're naturally
unique numbers. Given that a VCC can be in all kinds of states,
it would be pretty hard to use anything in struct atm_vcc else
as a unique key. Also, it's not very common to have atmsigd
talk ISP (Internal Signaling Protocol - the thing used between
atmsigd and the kernel) to a different box.

> the ATMSIGD_CTRL ioctl so at least there's no security hole but it's still
> damn gross (no offense, Werner :-)

It could probably be used to leverage other security holes in
atmsigd. (Not that I'm aware of any, but I'd be surprised if
there were none.)

- Werner

-- 
  _________________________________________________________________________
 / Werner Almesberger, Buenos Aires, Argentina         wa@almesberger.net /
/_http://www.almesberger.net/____________________________________________/