Re: [PATCH 3/5] net/netfilter/ipvs/ip_vs_ctl.c: drop argument range check just before the check for equality

[Julian Anastasov <ja@xxxxxx>]

        Hello,

On Tue, 22 Jul 2014, Dan Carpenter wrote:

> On Mon, Jul 21, 2014 at 11:01:56PM +0300, Julian Anastasov wrote:
> > @@ -2333,13 +2339,12 @@ do_ip_vs_set_ctl(struct sock *sk, int cmd, void 
> > __user *user, unsigned int len)
> >     struct ip_vs_dest_user_kern udest;
> >     struct netns_ipvs *ipvs = net_ipvs(net);
> >  
> > +   BUILD_BUG_ON(sizeof(arg) > 256);
> 
> 256 is off-by-one because u8 ranges from 0-255 so we are never able to
> copy 256 bytes into the "arg" buffer.

        I'll change it to >= 256, to catch that we can
not hold such big size in unsigned char [gs]et_arglen and
also as an indication that we have to use allocated buffer
instead of stack.

> > -   if (copylen > 128)
> > +   if (*len < (int) copylen || *len < 0) {
> > +           pr_err("get_ctl: len %d < %u\n", *len, copylen);
> 
> Don't let users flood dmesg.  Just return an error.  (This can be
> triggered by non-root as well).

        For now both set and get are privileged operations,
so we can keep it, it can catch if something wrong happens
with the structure sizes.

Regards

--
Julian Anastasov <ja@xxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe lvs-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html