Daniel Xu <dxu@xxxxxxxxx> wrote: > On Mon, Aug 15, 2022, at 4:40 PM, Florian Westphal wrote: > > Toke Høiland-Jørgensen <toke@xxxxxxxxxx> wrote: > >> > Support direct writes to nf_conn:mark from TC and XDP prog types. This > >> > is useful when applications want to store per-connection metadata. This > >> > is also particularly useful for applications that run both bpf and > >> > iptables/nftables because the latter can trivially access this metadata. > >> > > >> > One example use case would be if a bpf prog is responsible for advanced > >> > packet classification and iptables/nftables is later used for routing > >> > due to pre-existing/legacy code. > >> > > >> > Signed-off-by: Daniel Xu <dxu@xxxxxxxxx> > >> > >> Didn't we agree the last time around that all field access should be > >> using helper kfuncs instead of allowing direct writes to struct nf_conn? > > > > I don't see why ct->mark needs special handling. > > > > It might be possible we need to change accesses on nf/tc side to use > > READ/WRITE_ONCE though. > > I reviewed some of the LKMM literature and I would concur that > READ/WRITE_ONCE() is necessary. Especially after this patchset. > > However, it's unclear to me if this is a latent issue. IOW: is reading > ct->mark protected by a lock? I only briefly looked but it doesn't > seem like it. No, its not protected by a lock. READ/WRITE_ONCE is unrelated to your patchset, this is a pre-existing "bug".
