On Thu, Oct 20, 2022 at 10:05:22AM -0500, Ian Pilcher wrote:
> On 10/19/22 01:44, Pablo Neira Ayuso wrote:
> > On Tue, Oct 18, 2022 at 11:45:28AM -0500, Ian Pilcher wrote:
> > > The res_id member of struct nfgenmsg is supposed to be in network
> > > byte order (big endian). Call htons() in __nftnl_nlmsg_build_hdr()
> > > to ensure that this is true on little endian systems.
> >
> > LGTM, this is zero all the time at this moment. But it might be useful
> > in the future to bump it.
>
> Actually it isn't always zero. I only noticed this because
> nftnl_batch_begin() and nftnl_batch_end() set res_id to
> NFNL_SUBSYS_NFTABLES (instead of putting it in the high 8 bits of
> nlmsg_type).
Indeed, nfnetlink batch uses this for begin and end message.
> It's entirely possible that this is also a bug, as the fact that the
> value isn't currently being byte-swapped doesn't seem to make any
> difference.
There is code to workaround this issue in the kernel, it was added in 4.3.
commit a9de9777d613500b089a7416f936bf3ae5f070d2
Author: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx>
Date: Fri Aug 28 21:01:43 2015 +0200
netfilter: nfnetlink: work around wrong endianess in res_id field
oldest stable kernel is 4.9.
