netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Leon Romanovsky <leon@kernel.org>
To: Feng Wang <wangfe@google.com>
Cc: netdev@vger.kernel.org, steffen.klassert@secunet.com,
	antony.antony@secunet.com, pabeni@redhat.com
Subject: Re: [PATCH v6] xfrm: add SA information to the offloaded packet when if_id is set
Date: Wed, 27 Nov 2024 11:00:32 +0200	[thread overview]
Message-ID: <20241127090032.GD1245331@unreal> (raw)
In-Reply-To: <CADsK2K_na0Ugwv2PPT_s4oHSAx2rtZvtYY58C4MQkjE6G5y4Ew@mail.gmail.com>

On Tue, Nov 26, 2024 at 01:54:20PM -0800, Feng Wang wrote:
> Please see embedded answers below. Thanks.
> 
> On Sun, Nov 24, 2024 at 12:15 AM Leon Romanovsky <leon@kernel.org> wrote:
> >
> > On Thu, Nov 21, 2024 at 09:52:58PM +0000, Feng Wang wrote:
> > > In packet offload mode, append Security Association (SA) information
> > > to each packet, replicating the crypto offload implementation.
> >
> > Please write explicitly WHY "replicating ..." is right thing to do and
> > why current code is not enough.
> >
> > The best thing will be to see how packet traversal in the netdev stack
> > till it gets to the wire.
> >
> I explained why doing this change in the third paragraph based on
> Steffen's suggestion, I can move it here.

Steffen didn't say "let's replicate ...". All that he said that he wants
to see API complete.

> 
> > >
> > > The XFRM_XMIT flag is set to enable packet to be returned immediately
> > > from the validate_xmit_xfrm function, thus aligning with the existing
> > > code path for packet offload mode.
> >
> > This chunk was dropped mysteriously. It doesn't exist in the current patch.
> > What type of testing did you perform? Can you please add it to the
> > commit message?
> >
> I didn't drop any code in the current patch,  I created a test and the
> patch will be upstreamed after this change is checked in.
> The link is https://lore.kernel.org/all/20241104233315.3387982-1-wangfe@google.com/
> Do I need to add the test details in this commit?

No, you need to create test that knows to handle case with and without
if_id properly. You should add test to the patchset as first patch and
implementation as second one.

> 
> > According to the strongswan documentation https://docs.strongswan.org/docs/5.9/features/routeBasedVpn.html,
> > "Traffic that’s routed to an XFRM interface, while no policies and SAs with matching interface ID exist,
> > will be dropped by the kernel."
> >
> > It looks like the current code doesn't handle this case, does it?
> >
> The current code will drop the packet if the interface ID is not matched.

How? Who will drop it?

> 
> > >
> > > This SA info helps HW offload match packets to their correct security
> > > policies. The XFRM interface ID is included, which is used in setups
> > > with multiple XFRM interfaces where source/destination addresses alone
> > > can't pinpoint the right policy.
> >
> > Please add an examples of iproute2 commands on how it is achieved,
> > together with tcprdump examples of before and after.
> >
> A test patch will be upstreamed.  There is no need for tcpdump because
> this change won't change packet content.

Of course it is needed, it will show that without if_id patch packets
are unencrypted, while after applying the patch, we will see encrypted
packets.

> 
> > >
> > > Enable packet offload mode on netdevsim and add code to check the XFRM
> > > interface ID.
> >
> > You still need to add checks in existing drivers to make sure that SAs
> > with if_id won't be offloaded.
> >
> There is no existing driver supporting packet offload mode except netdevsim.

It is not true, please check the code.

> 
> > IMHO, netdevsim implementation is not a replacement to support
> > out-of-tree code, but a way to help testing features without need to
> > have complex HW, but still for the code which is in-tree.
> >
> We discussed this before, and I followed your and Steffen's comment to
> add netdevsim simulation code to satisfy the requirement.

I didn't suggest netdevsim and always requested to upstream real driver.
Netdevsim is Steffen's suggestion to overcome kernel rule of no adding
code without in-kernel users.

Thanks

> 
> > Thanks
> >
> 
> > >
> > > Signed-off-by: wangfe <wangfe@google.com>
> > > ---
> > > v6: https://lore.kernel.org/all/20241119220411.2961121-1-wangfe@google.com/
> > >   - Fix code style to follow reverse x-mas tree order declaration.
> > > v5: https://lore.kernel.org/all/20241112192249.341515-1-wangfe@google.com/
> > >   - Add SA information only when XFRM interface ID is non-zero.
> > > v4: https://lore.kernel.org/all/20241104233251.3387719-1-wangfe@google.com/
> > >   - Add offload flag check and only doing check when XFRM interface
> > >     ID is non-zero.
> > > v3: https://lore.kernel.org/all/20240822200252.472298-1-wangfe@google.com/
> > >   - Add XFRM interface ID checking on netdevsim in the packet offload
> > >     mode.
> > > v2:
> > >   - Add why HW offload requires the SA info to the commit message
> > > v1: https://lore.kernel.org/all/20240812182317.1962756-1-wangfe@google.com/
> > > ---
> > > ---
> > >  drivers/net/netdevsim/ipsec.c     | 32 ++++++++++++++++++++++++++++++-
> > >  drivers/net/netdevsim/netdevsim.h |  1 +
> > >  net/xfrm/xfrm_output.c            | 22 +++++++++++++++++++++
> > >  3 files changed, 54 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/netdevsim/ipsec.c b/drivers/net/netdevsim/ipsec.c
> > > index 88187dd4eb2d..5677b2acf9c0 100644
> > > --- a/drivers/net/netdevsim/ipsec.c
> > > +++ b/drivers/net/netdevsim/ipsec.c
> > > @@ -153,7 +153,8 @@ static int nsim_ipsec_add_sa(struct xfrm_state *xs,
> > >               return -EINVAL;
> > >       }
> > >
> > > -     if (xs->xso.type != XFRM_DEV_OFFLOAD_CRYPTO) {
> > > +     if (xs->xso.type != XFRM_DEV_OFFLOAD_CRYPTO &&
> > > +         xs->xso.type != XFRM_DEV_OFFLOAD_PACKET) {
> > >               NL_SET_ERR_MSG_MOD(extack, "Unsupported ipsec offload type");
> > >               return -EINVAL;
> > >       }
> > > @@ -169,6 +170,7 @@ static int nsim_ipsec_add_sa(struct xfrm_state *xs,
> > >       memset(&sa, 0, sizeof(sa));
> > >       sa.used = true;
> > >       sa.xs = xs;
> > > +     sa.if_id = xs->if_id;
> > >
> > >       if (sa.xs->id.proto & IPPROTO_ESP)
> > >               sa.crypt = xs->ealg || xs->aead;
> > > @@ -227,16 +229,31 @@ static bool nsim_ipsec_offload_ok(struct sk_buff *skb, struct xfrm_state *xs)
> > >       return true;
> > >  }
> > >
> > > +static int nsim_ipsec_add_policy(struct xfrm_policy *policy,
> > > +                              struct netlink_ext_ack *extack)
> > > +{
> > > +     return 0;
> > > +}
> > > +
> > > +static void nsim_ipsec_del_policy(struct xfrm_policy *policy)
> > > +{
> > > +}
> > > +
> > >  static const struct xfrmdev_ops nsim_xfrmdev_ops = {
> > >       .xdo_dev_state_add      = nsim_ipsec_add_sa,
> > >       .xdo_dev_state_delete   = nsim_ipsec_del_sa,
> > >       .xdo_dev_offload_ok     = nsim_ipsec_offload_ok,
> > > +
> > > +     .xdo_dev_policy_add     = nsim_ipsec_add_policy,
> > > +     .xdo_dev_policy_delete  = nsim_ipsec_del_policy,
> > > +
> > >  };
> > >
> > >  bool nsim_ipsec_tx(struct netdevsim *ns, struct sk_buff *skb)
> > >  {
> > >       struct sec_path *sp = skb_sec_path(skb);
> > >       struct nsim_ipsec *ipsec = &ns->ipsec;
> > > +     struct xfrm_offload *xo;
> > >       struct xfrm_state *xs;
> > >       struct nsim_sa *tsa;
> > >       u32 sa_idx;
> > > @@ -275,6 +292,19 @@ bool nsim_ipsec_tx(struct netdevsim *ns, struct sk_buff *skb)
> > >               return false;
> > >       }
> > >
> > > +     if (xs->if_id) {
> > > +             if (xs->if_id != tsa->if_id) {
> > > +                     netdev_err(ns->netdev, "unmatched if_id %d %d\n",
> > > +                                xs->if_id, tsa->if_id);
> > > +                     return false;
> > > +             }
> > > +             xo = xfrm_offload(skb);
> > > +             if (!xo || !(xo->flags & XFRM_XMIT)) {
> > > +                     netdev_err(ns->netdev, "offload flag missing or wrong\n");
> > > +                     return false;
> > > +             }
> > > +     }
> > > +
> > >       ipsec->tx++;
> > >
> > >       return true;
> > > diff --git a/drivers/net/netdevsim/netdevsim.h b/drivers/net/netdevsim/netdevsim.h
> > > index bf02efa10956..4941b6e46d0a 100644
> > > --- a/drivers/net/netdevsim/netdevsim.h
> > > +++ b/drivers/net/netdevsim/netdevsim.h
> > > @@ -41,6 +41,7 @@ struct nsim_sa {
> > >       __be32 ipaddr[4];
> > >       u32 key[4];
> > >       u32 salt;
> > > +     u32 if_id;
> > >       bool used;
> > >       bool crypt;
> > >       bool rx;
> > > diff --git a/net/xfrm/xfrm_output.c b/net/xfrm/xfrm_output.c
> > > index e5722c95b8bb..3e9a1b17f37e 100644
> > > --- a/net/xfrm/xfrm_output.c
> > > +++ b/net/xfrm/xfrm_output.c
> > > @@ -704,6 +704,8 @@ int xfrm_output(struct sock *sk, struct sk_buff *skb)
> > >  {
> > >       struct net *net = dev_net(skb_dst(skb)->dev);
> > >       struct xfrm_state *x = skb_dst(skb)->xfrm;
> > > +     struct xfrm_offload *xo;
> > > +     struct sec_path *sp;
> > >       int family;
> > >       int err;
> > >
> > > @@ -728,7 +730,27 @@ int xfrm_output(struct sock *sk, struct sk_buff *skb)
> > >                       kfree_skb(skb);
> > >                       return -EHOSTUNREACH;
> > >               }
> > > +             if (x->if_id) {
> > > +                     sp = secpath_set(skb);
> > > +                     if (!sp) {
> > > +                             XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
> > > +                             kfree_skb(skb);
> > > +                             return -ENOMEM;
> > > +                     }
> > > +
> > > +                     sp->olen++;
> > > +                     sp->xvec[sp->len++] = x;
> > > +                     xfrm_state_hold(x);
> > >
> > > +                     xo = xfrm_offload(skb);
> > > +                     if (!xo) {
> > > +                             secpath_reset(skb);
> > > +                             XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
> > > +                             kfree_skb(skb);
> > > +                             return -EINVAL;
> > > +                     }
> > > +                     xo->flags |= XFRM_XMIT;
> > > +             }
> > >               return xfrm_output_resume(sk, skb, 0);
> > >       }
> > >
> > > --
> > > 2.47.0.371.ga323438b13-goog
> > >
> > >

  reply	other threads:[~2024-11-27  9:00 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-11-21 21:52 [PATCH v6] xfrm: add SA information to the offloaded packet when if_id is set Feng Wang
2024-11-24  8:15 ` Leon Romanovsky
2024-11-26 21:54   ` Feng Wang
2024-11-27  9:00     ` Leon Romanovsky [this message]
2024-12-02 20:41       ` Feng Wang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20241127090032.GD1245331@unreal \
    --to=leon@kernel.org \
    --cc=antony.antony@secunet.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=steffen.klassert@secunet.com \
    --cc=wangfe@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).