Re: [RFC v1] copy_{to,from}_user(): only inline when !__CHECKER__

[Tycho Andersen]

On Sun, Dec 09, 2018 at 09:46:00PM +0000, Al Viro wrote:
> On Sun, Dec 09, 2018 at 02:25:23PM -0700, Tycho Andersen wrote:
> 
> > > Which sparse checks do not trigger?  Explain, please - as it is, I had been
> > > unable to guess what could "specifically looks for a call instruction" refer
> > > to.
> > 
> > In sparse.c there's check_call_instruction(), which is triggered when
> > there's an instruction of OP_CALL type in the basic block. This simply
> > compares against the name of the call target to determine whether or
> > not to call check_ctu().
> 
> Oh, that Linus' experiment with "look for huge constant size argument
> to memcpy() et.al."?  Frankly, it's not only the wrong place to put the
> checks, but breaking inlining loses the _real_ "known constant size"
> checks in there.
> 
> I don't know if the check_ctu thing has ever caught a bug...  What kind of
> checks do you want to add?  Because this place is almost certainly wrong
> for anything useful...

Yeah, agreed that the static size check doesn't seem particularly
useful. I linked to these in the other mail, but the top two patches
here are what I was playing with:

https://github.com/tych0/sparse/commits/check-as-infoleaks

> If anything, I would suggest simulating this behaviour with something like
> 	if (__builtin_constant_p(size) && size > something)
> 		/* something that would trigger a warning */
> _inside_ copy_from_user()/copy_to_user() and to hell with name-recognizing
> magic...

Hmm. I wonder if we couldn't do some size checking with the argument
like this instead. Thanks for the idea, I'll play around with it.

Tycho