Re: [PATCH v2 10/10] binfmt_flat: allow compressed flat binary format to work on MMU systems

[Nicolas Pitre]

On Mon, 18 Jul 2016, One Thousand Gnomes wrote:

> On Mon, 18 Jul 2016 11:45:53 -0400 (EDT)
> Nicolas Pitre <nicolas.pitre@xxxxxxxxxx> wrote:
> 
> > On Mon, 18 Jul 2016, One Thousand Gnomes wrote:
> > 
> > > On Sun, 17 Jul 2016 23:31:56 -0400
> > > Nicolas Pitre <nicolas.pitre@xxxxxxxxxx> wrote:
> > >   
> > > > Let's take the simple and obvious approach by decompressing the binary
> > > > into a kernel buffer and then copying it to user space.  Those who are
> > > > looking for more performance on a MMU system are unlikely to choose this
> > > > executable format anyway.  
> > > 
> > > The flat loader takes a very casual attitude to overruns and corrupted
> > > binaries. It's after all MMUless so has no real security model. If you
> > > enable flat for an MMU system then IMHO those all need to be fixed
> > > including all the missing overflow checks on the maths on textlen and the
> > > like.  
> > 
> > What about the following patch?  This with existing user accessors and 
> > allocation error checks should cover it all.
> > 
> > ----- >8  
> > commit cc1051c9c57202772568600e96b75229a2a7cf19
> > Author: Nicolas Pitre <nicolas.pitre@xxxxxxxxxx>
> > Date:   Mon Jul 18 11:28:57 2016 -0400
> > 
> >     binfmt_flat: prevent kernel dammage from corrupted executable headers
> >     
> >     Signed-off-by: Nicolas Pitre <nico@xxxxxxxxxx>
> > 
> > diff --git a/fs/binfmt_flat.c b/fs/binfmt_flat.c
> > index 24deae4dcb..fa0054c1c3 100644
> > --- a/fs/binfmt_flat.c
> > +++ b/fs/binfmt_flat.c
> > @@ -498,6 +498,17 @@ static int load_flat_file(struct linux_binprm * bprm,
> >  	}
> >  
> >  	/*
> > +	 * Make sure the header params are sane.
> > +	 * 28 bits (256 MB) is way more than reasonable in this case.
> > +	 * If some top bits are set we have probable binary corruption.
> > +	*/
> > +	if ((text_len | data_len | bss_len | stack_len | full_data) >> 28) {
> > +		printk("BINFMT_FLAT: bad header\n");
> 
> Apart from the printk that looks good for the header but I think the rest
> could do with a fair bit more review (eg relocations in range checks).

Given that they all go through put_user() now, the worst that could 
happen is an executable that craps onto itself.  I don't think there is 
much we can do here besides letting the user task crash.


Nicolas