* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
[not found] <20201109072930.14048-1-nusiddiq@redhat.com>
@ 2020-11-09 19:54 ` Jakub Kicinski
2020-11-10 8:39 ` Numan Siddique
[not found] ` <20201109213557.GE23619@breakpoint.cc>
1 sibling, 1 reply; 7+ messages in thread
From: Jakub Kicinski @ 2020-11-09 19:54 UTC (permalink / raw)
To: nusiddiq
Cc: dev, netdev, Pravin B Shelar, Florian Westphal,
Pablo Neira Ayuso, netfilter-devel
On Mon, 9 Nov 2020 12:59:30 +0530 nusiddiq@redhat.com wrote:
> From: Numan Siddique <nusiddiq@redhat.com>
>
> Before calling nf_conntrack_in(), caller can set this flag in the
> connection template for a tcp packet and any errors in the
> tcp_in_window() will be ignored.
>
> A helper function - nf_ct_set_tcp_be_liberal(nf_conn) is added which
> sets this flag for both the directions of the nf_conn.
>
> openvswitch makes use of this feature so that any out of window tcp
> packets are not marked invalid. Prior to this there was no easy way
> to distinguish if conntracked packet is marked invalid because of
> tcp_in_window() check error or because it doesn't belong to an
> existing connection.
>
> An earlier attempt (see the link) tried to solve this problem for
> openvswitch in a different way. Florian Westphal instead suggested
> to be liberal in openvswitch for tcp packets.
>
> Link: https://patchwork.ozlabs.org/project/netdev/patch/20201006083355.121018-1-nusiddiq@redhat.com/
>
> Suggested-by: Florian Westphal <fw@strlen.de>
> Signed-off-by: Numan Siddique <nusiddiq@redhat.com>
Please repost Ccing Pablo & netfilter-devel.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
2020-11-09 19:54 ` [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis Jakub Kicinski
@ 2020-11-10 8:39 ` Numan Siddique
0 siblings, 0 replies; 7+ messages in thread
From: Numan Siddique @ 2020-11-10 8:39 UTC (permalink / raw)
To: Jakub Kicinski
Cc: ovs dev, netdev, Pravin B Shelar, Florian Westphal,
Pablo Neira Ayuso, netfilter-devel
On Tue, Nov 10, 2020 at 1:25 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Mon, 9 Nov 2020 12:59:30 +0530 nusiddiq@redhat.com wrote:
> > From: Numan Siddique <nusiddiq@redhat.com>
> >
> > Before calling nf_conntrack_in(), caller can set this flag in the
> > connection template for a tcp packet and any errors in the
> > tcp_in_window() will be ignored.
> >
> > A helper function - nf_ct_set_tcp_be_liberal(nf_conn) is added which
> > sets this flag for both the directions of the nf_conn.
> >
> > openvswitch makes use of this feature so that any out of window tcp
> > packets are not marked invalid. Prior to this there was no easy way
> > to distinguish if conntracked packet is marked invalid because of
> > tcp_in_window() check error or because it doesn't belong to an
> > existing connection.
> >
> > An earlier attempt (see the link) tried to solve this problem for
> > openvswitch in a different way. Florian Westphal instead suggested
> > to be liberal in openvswitch for tcp packets.
> >
> > Link: https://patchwork.ozlabs.org/project/netdev/patch/20201006083355.121018-1-nusiddiq@redhat.com/
> >
> > Suggested-by: Florian Westphal <fw@strlen.de>
> > Signed-off-by: Numan Siddique <nusiddiq@redhat.com>
>
> Please repost Ccing Pablo & netfilter-devel.
Thanks. I will repost.
Numan
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
[not found] ` <20201109213557.GE23619@breakpoint.cc>
@ 2020-11-10 8:47 ` Numan Siddique
2020-11-10 12:25 ` Florian Westphal
0 siblings, 1 reply; 7+ messages in thread
From: Numan Siddique @ 2020-11-10 8:47 UTC (permalink / raw)
To: Florian Westphal
Cc: ovs dev, netdev, Pravin B Shelar, Pablo Neira Ayuso, netfilter-devel
On Tue, Nov 10, 2020 at 3:06 AM Florian Westphal <fw@strlen.de> wrote:
>
> nusiddiq@redhat.com <nusiddiq@redhat.com> wrote:
> > From: Numan Siddique <nusiddiq@redhat.com>
> >
> > Before calling nf_conntrack_in(), caller can set this flag in the
> > connection template for a tcp packet and any errors in the
> > tcp_in_window() will be ignored.
> >
> > A helper function - nf_ct_set_tcp_be_liberal(nf_conn) is added which
> > sets this flag for both the directions of the nf_conn.
> >
> > openvswitch makes use of this feature so that any out of window tcp
> > packets are not marked invalid. Prior to this there was no easy way
> > to distinguish if conntracked packet is marked invalid because of
> > tcp_in_window() check error or because it doesn't belong to an
> > existing connection.
> >
> > An earlier attempt (see the link) tried to solve this problem for
> > openvswitch in a different way. Florian Westphal instead suggested
> > to be liberal in openvswitch for tcp packets.
> >
> > Link: https://patchwork.ozlabs.org/project/netdev/patch/20201006083355.121018-1-nusiddiq@redhat.com/
> >
> > Suggested-by: Florian Westphal <fw@strlen.de>
> > Signed-off-by: Numan Siddique <nusiddiq@redhat.com>
> > ---
> > include/net/netfilter/nf_conntrack_l4proto.h | 6 ++++++
> > net/netfilter/nf_conntrack_core.c | 13 +++++++++++--
> > net/openvswitch/conntrack.c | 1 +
> > 3 files changed, 18 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/net/netfilter/nf_conntrack_l4proto.h b/include/net/netfilter/nf_conntrack_l4proto.h
> > index 88186b95b3c2..572ae8d2a622 100644
> > --- a/include/net/netfilter/nf_conntrack_l4proto.h
> > +++ b/include/net/netfilter/nf_conntrack_l4proto.h
> > @@ -203,6 +203,12 @@ static inline struct nf_icmp_net *nf_icmpv6_pernet(struct net *net)
> > {
> > return &net->ct.nf_ct_proto.icmpv6;
> > }
> > +
> > +static inline void nf_ct_set_tcp_be_liberal(struct nf_conn *ct)
> > +{
> > + ct->proto.tcp.seen[0].flags |= IP_CT_TCP_FLAG_BE_LIBERAL;
> > + ct->proto.tcp.seen[1].flags |= IP_CT_TCP_FLAG_BE_LIBERAL;
> > +}
> > #endif
> >
> > #ifdef CONFIG_NF_CT_PROTO_DCCP
> > diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
> > index 234b7cab37c3..8290c5b04e88 100644
> > --- a/net/netfilter/nf_conntrack_core.c
> > +++ b/net/netfilter/nf_conntrack_core.c
> > @@ -1748,10 +1748,18 @@ static int nf_conntrack_handle_packet(struct nf_conn *ct,
> > struct sk_buff *skb,
> > unsigned int dataoff,
> > enum ip_conntrack_info ctinfo,
> > - const struct nf_hook_state *state)
> > + const struct nf_hook_state *state,
> > + union nf_conntrack_proto *tmpl_proto)
>
> I would prefer if we could avoid adding yet another function argument.
>
> AFAICS its enough to call the new nf_ct_set_tcp_be_liberal() helper
> before nf_conntrack_confirm() in ovs_ct_commit(), e.g.:
Thanks for the comments. I actually tried this approach first, but it
doesn't seem to work.
I noticed that for the committed connections, the ct tcp flag -
IP_CT_TCP_FLAG_BE_LIBERAL is
not set when nf_conntrack_in() calls resolve_normal_ct().
Would you expect that the tcp ct flags should have been preserved once
the connection is committed ?
Thanks
Numan
>
> diff --git a/net/openvswitch/conntrack.c b/net/openvswitch/conntrack.c
> --- a/net/openvswitch/conntrack.c
> +++ b/net/openvswitch/conntrack.c
> @@ -1235,10 +1235,13 @@ static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
> if (!nf_ct_is_confirmed(ct)) {
> err = ovs_ct_init_labels(ct, key, &info->labels.value,
> &info->labels.mask);
> if (err)
> return err;
> +
> + if (nf_ct_protonum(ct) == IPPROTO_TCP)
> + nf_ct_set_tcp_be_liberal(ct);
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
2020-11-10 8:47 ` Numan Siddique
@ 2020-11-10 12:25 ` Florian Westphal
2020-11-10 12:58 ` Numan Siddique
0 siblings, 1 reply; 7+ messages in thread
From: Florian Westphal @ 2020-11-10 12:25 UTC (permalink / raw)
To: Numan Siddique
Cc: Florian Westphal, ovs dev, netdev, Pravin B Shelar,
Pablo Neira Ayuso, netfilter-devel
Numan Siddique <nusiddiq@redhat.com> wrote:
> On Tue, Nov 10, 2020 at 3:06 AM Florian Westphal <fw@strlen.de> wrote:
> Thanks for the comments. I actually tried this approach first, but it
> doesn't seem to work.
> I noticed that for the committed connections, the ct tcp flag -
> IP_CT_TCP_FLAG_BE_LIBERAL is
> not set when nf_conntrack_in() calls resolve_normal_ct().
Yes, it won't be set during nf_conntrack_in, thats why I suggested
to do it before confirming the connection.
> Would you expect that the tcp ct flags should have been preserved once
> the connection is committed ?
Yes, they are preserved when you set them after nf_conntrack_in(), else
we would already have trouble with hw flow offloading which needs to
turn off window checks as well.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
2020-11-10 12:25 ` Florian Westphal
@ 2020-11-10 12:58 ` Numan Siddique
2020-11-10 13:11 ` Florian Westphal
0 siblings, 1 reply; 7+ messages in thread
From: Numan Siddique @ 2020-11-10 12:58 UTC (permalink / raw)
To: Florian Westphal
Cc: ovs dev, netdev, Pravin B Shelar, Pablo Neira Ayuso, netfilter-devel
On Tue, Nov 10, 2020 at 5:55 PM Florian Westphal <fw@strlen.de> wrote:
>
> Numan Siddique <nusiddiq@redhat.com> wrote:
> > On Tue, Nov 10, 2020 at 3:06 AM Florian Westphal <fw@strlen.de> wrote:
> > Thanks for the comments. I actually tried this approach first, but it
> > doesn't seem to work.
> > I noticed that for the committed connections, the ct tcp flag -
> > IP_CT_TCP_FLAG_BE_LIBERAL is
> > not set when nf_conntrack_in() calls resolve_normal_ct().
>
> Yes, it won't be set during nf_conntrack_in, thats why I suggested
> to do it before confirming the connection.
Sorry for the confusion. What I mean is - I tested your suggestion - i.e called
nf_ct_set_tcp_be_liberal() before calling nf_conntrack_confirm().
Once the connection is established, for subsequent packets, openvswitch
calls nf_conntrack_in() [1] to see if the packet is part of the
existing connection or not (i.e ct.new or ct.est )
and if the packet happens to be out-of-window then skb->_nfct is set
to NULL. And the tcp
be flags set during confirmation are not reflected when
nf_conntrack_in() calls resolve_normal_ct().
>
> > Would you expect that the tcp ct flags should have been preserved once
> > the connection is committed ?
>
> Yes, they are preserved when you set them after nf_conntrack_in(), else
> we would already have trouble with hw flow offloading which needs to
> turn off window checks as well.
Looks they are not preserved for the openvswitch case. Probably
openvswitch is doing something wrong.
I will debug further and see what is going on.
Please let me know if you have any further comments.
Thanks
Numan
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
2020-11-10 12:58 ` Numan Siddique
@ 2020-11-10 13:11 ` Florian Westphal
2020-11-16 13:06 ` Numan Siddique
0 siblings, 1 reply; 7+ messages in thread
From: Florian Westphal @ 2020-11-10 13:11 UTC (permalink / raw)
To: Numan Siddique
Cc: Florian Westphal, ovs dev, netdev, Pravin B Shelar,
Pablo Neira Ayuso, netfilter-devel
Numan Siddique <nusiddiq@redhat.com> wrote:
> On Tue, Nov 10, 2020 at 5:55 PM Florian Westphal <fw@strlen.de> wrote:
> >
> > Numan Siddique <nusiddiq@redhat.com> wrote:
> > > On Tue, Nov 10, 2020 at 3:06 AM Florian Westphal <fw@strlen.de> wrote:
> > > Thanks for the comments. I actually tried this approach first, but it
> > > doesn't seem to work.
> > > I noticed that for the committed connections, the ct tcp flag -
> > > IP_CT_TCP_FLAG_BE_LIBERAL is
> > > not set when nf_conntrack_in() calls resolve_normal_ct().
> >
> > Yes, it won't be set during nf_conntrack_in, thats why I suggested
> > to do it before confirming the connection.
>
> Sorry for the confusion. What I mean is - I tested your suggestion - i.e called
> nf_ct_set_tcp_be_liberal() before calling nf_conntrack_confirm().
>
> Once the connection is established, for subsequent packets, openvswitch
> calls nf_conntrack_in() [1] to see if the packet is part of the
> existing connection or not (i.e ct.new or ct.est )
> and if the packet happens to be out-of-window then skb->_nfct is set
> to NULL. And the tcp
> be flags set during confirmation are not reflected when
> nf_conntrack_in() calls resolve_normal_ct().
Can you debug where this happens? This looks very very wrong.
resolve_normal_ct() has no business to check any of those flags
(and I don't see where it uses them, it should only deal with the
tuples).
The flags come into play when nf_conntrack_handle_packet() gets called
after resolve_normal_ct has found an entry, since that will end up
calling the tcp conntrack part.
The entry found/returned by resolve_normal_ct should be the same
nf_conn entry that was confirmed earlier, i.e. it should be in "liberal"
mode.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis.
2020-11-10 13:11 ` Florian Westphal
@ 2020-11-16 13:06 ` Numan Siddique
0 siblings, 0 replies; 7+ messages in thread
From: Numan Siddique @ 2020-11-16 13:06 UTC (permalink / raw)
To: Florian Westphal
Cc: ovs dev, netdev, Pravin B Shelar, Pablo Neira Ayuso, netfilter-devel
On Tue, Nov 10, 2020 at 6:41 PM Florian Westphal <fw@strlen.de> wrote:
>
> Numan Siddique <nusiddiq@redhat.com> wrote:
> > On Tue, Nov 10, 2020 at 5:55 PM Florian Westphal <fw@strlen.de> wrote:
> > >
> > > Numan Siddique <nusiddiq@redhat.com> wrote:
> > > > On Tue, Nov 10, 2020 at 3:06 AM Florian Westphal <fw@strlen.de> wrote:
> > > > Thanks for the comments. I actually tried this approach first, but it
> > > > doesn't seem to work.
> > > > I noticed that for the committed connections, the ct tcp flag -
> > > > IP_CT_TCP_FLAG_BE_LIBERAL is
> > > > not set when nf_conntrack_in() calls resolve_normal_ct().
> > >
> > > Yes, it won't be set during nf_conntrack_in, thats why I suggested
> > > to do it before confirming the connection.
> >
> > Sorry for the confusion. What I mean is - I tested your suggestion - i.e called
> > nf_ct_set_tcp_be_liberal() before calling nf_conntrack_confirm().
> >
> > Once the connection is established, for subsequent packets, openvswitch
> > calls nf_conntrack_in() [1] to see if the packet is part of the
> > existing connection or not (i.e ct.new or ct.est )
> > and if the packet happens to be out-of-window then skb->_nfct is set
> > to NULL. And the tcp
> > be flags set during confirmation are not reflected when
> > nf_conntrack_in() calls resolve_normal_ct().
>
> Can you debug where this happens? This looks very very wrong.
> resolve_normal_ct() has no business to check any of those flags
> (and I don't see where it uses them, it should only deal with the
> tuples).
>
> The flags come into play when nf_conntrack_handle_packet() gets called
> after resolve_normal_ct has found an entry, since that will end up
> calling the tcp conntrack part.
>
> The entry found/returned by resolve_normal_ct should be the same
> nf_conn entry that was confirmed earlier, i.e. it should be in "liberal"
> mode.
I debugged a bit. Calling nf_ct_set_tcp_be_liberal() in ct_commit
doesn't work because
- the first SYN packet during connection establishment is committed
to the contract.
- but tcp_in_window() calls tcp_options() which clears the tcp ct
flags for the SYN and SYN-ACK packets.
And hence the flags get cleared.
So I think it should be enough if openvswitch calls
nf_ct_set_tcp_be_liberal() once the connection is established.
I posted v2. Request to take a look.
Thanks
Numan
>
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2020-11-16 13:07 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <20201109072930.14048-1-nusiddiq@redhat.com>
2020-11-09 19:54 ` [net-next] netfiler: conntrack: Add the option to set ct tcp flag - BE_LIBERAL per-ct basis Jakub Kicinski
2020-11-10 8:39 ` Numan Siddique
[not found] ` <20201109213557.GE23619@breakpoint.cc>
2020-11-10 8:47 ` Numan Siddique
2020-11-10 12:25 ` Florian Westphal
2020-11-10 12:58 ` Numan Siddique
2020-11-10 13:11 ` Florian Westphal
2020-11-16 13:06 ` Numan Siddique
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).