Re: [PATCH nf] netfilter: conntrack: avoid integer overflow when resizing

[Florian Westphal <fw@strlen.de>]

Pablo Neira Ayuso <pablo@netfilter.org> wrote:

[ Sorry for late reply ]

> On Sun, Apr 24, 2016 at 01:18:21AM +0200, Florian Westphal wrote:
> > Can overflow so we might allocate very small table when bucket count is
> > high on a 32bit platform.
> > 
> > Note: resize is only possible from init_netns.
> > 
> > Signed-off-by: Florian Westphal <fw@strlen.de>
> > ---
> >  net/netfilter/nf_conntrack_core.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> > 
> > diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
> > index 2bbb962..11daca5 100644
> > --- a/net/netfilter/nf_conntrack_core.c
> > +++ b/net/netfilter/nf_conntrack_core.c
> > @@ -1563,8 +1563,15 @@ void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls)
> >  	unsigned int nr_slots, i;
> >  	size_t sz;
> >  
> > +	if (*sizep > (UINT_MAX / sizeof(struct hlist_nulls_head)))
> > +		return NULL;
> 
> *sizep gets initially set to the number of buckets.

Yes.

> >  	BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head));
> >  	nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head));
> 
> Then, this value is divided by the number of hlist heads that fit into
> a page.

No, its rounded up to a multiple of PAGE/hlist heads, so for very large
values of *sizep nr_slots would be 0.

> > +	if (nr_slots > (UINT_MAX / sizeof(struct hlist_nulls_head)))
> > +		return NULL;

Alternative would be to change this to:

if (nr_slots == 0 || nr_slots > (UINT_MAX / sizeof(struct hlist_nulls_head)))
	return NULL;

Or, add this check:

if (nr_slots < roundup(1, PAGE_SIZE / sizeof(struct hlist_nulls_head))
    nr_slots = roundup(1, PAGE_SIZE / sizeof(struct hlist_nulls_head)))

I wasn't sure if its better to fail or if we should just pretend a sane
value was given.

Let me know what you think and I'll submit a v2.

[ Its not a big deal, but eventually I'd like to make the sysctl
  writeable so users can just increase that, no need to use this obscure
  module parameter/sysfs param ... ]

Cheers,
Florian