Re: [PATCH net 1/2] ipv4: mcast: getsockopt: do not overwrite past optlen

[Breno Leitao <leitao@debian.org>]

Hello Jakub,

On Tue, Aug 11, 2026 at 08:15:43AM -0700, Jakub Kicinski wrote:
> On Tue, 11 Aug 2026 05:19:17 -0700 Breno Leitao wrote:
> > So my question to you: can you point me to actual software that
> > passes a "small" optlen and expects the kernel to write past it? That
> > would help to decide about the two options above.
> 
> If you are very confident that no such SW exists - we can try to queue
> this up for -next. (TBH I'm not, mcast specifically may be full of
> strange one off manually written user space (as opposed to common libraries)).

> If we decide to change the behavior- we will probably have to wait
> until this makes it to an LTS release + some time for people to deploy.
> It can't be a fix.
> 
> So practically speaking it may be more expedient to add some hacks to
> cater to this case in the conversion, and then remove the hack. That'd
> be easier to revert if someone pipes up later that we broke their SW.

Thanks. I was mostly trying to find out whether this is a real problem
before creating this quirk, and your intuition is helpful here. I will
proceed with the quirk then.

For the sockopt_t conversion I can expand the iterator mid-air for the
few options that size their reply from a count in the header, so the
conversion stays a no-op for userspace.

Expanding is only safe for a user buffer, though - a kernel-backed
optval might be too risky, I would say. And it will not break anything,
given we don't have in-kernel users for this yet (given the __user
params)


I might come up with a helper like this one:
        static inline void sockopt_expand_out(sockopt_t *opt, size_t len)
        {
                if (len < iov_iter_count(&opt->iter_out))
			// No quirk required
			return;
			
                /* Only a user buffer can be larger than the caller
                 * declared; a kernel-backed optval has nothing behind it.
                 */
                if (WARN_ON_ONCE(!iter_is_ubuf(&opt->iter_out)))
                        return;

		iov_iter_reexpand(&opt->iter_out, len);
        }

and then, at the call site, right after the header is read and before
anything is written back:

        /* The reply is sized by gf_numsrc, not by optlen, which only has
         * to cover the fixed header. Grow the output iterator to what the
         * header declares, or the source list gets truncated.
         */
        sockopt_expand_out(opt, size_add(size0,
                                         size_mul(gsf.gf_numsrc,
                                                  sizeof(gsf.gf_slist_flex[0]))));


(All completely untested so far.)

Thanks for the suggestion,
--breno