Message ID | 20230718092405.4124345-1-quic_ilial@quicinc.com (mailing list archive) |
---|---|
State | Changes Requested |
Delegated to: | Netdev Maintainers |
Headers | show |
Series | xfrm: Allow ESP over UDP in packet offload mode | expand |
On Tue, Jul 18, 2023 at 12:24:05PM +0300, Ilia Lin wrote: > The ESP encapsulation is not supported only in crypto mode. > In packet offload mode, the RX is bypassing the XFRM, > so we can enable the encapsulation. It is not accurate. RX is bypassed after XFRM validated packet to ensure that it was really handled by HW. However, this patch should come with relevant driver code which should support ESP over UDP. You can see it here: https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/log/?h=xfrm-next xfrm: Support UDP encapsulation in packet offload mode net/mlx5e: Support IPsec NAT-T functionality net/mlx5e: Check for IPsec NAT-T support net/mlx5: Add relevant capabilities bits to support NAT-T Thanks > > Signed-off-by: Ilia Lin <ilia.lin@kernel.org> > --- > net/xfrm/xfrm_device.c | 7 ++++--- > 1 file changed, 4 insertions(+), 3 deletions(-) > > diff --git a/net/xfrm/xfrm_device.c b/net/xfrm/xfrm_device.c > index 4aff76c6f12e0..3018468d97662 100644 > --- a/net/xfrm/xfrm_device.c > +++ b/net/xfrm/xfrm_device.c > @@ -246,8 +246,10 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, > return -EINVAL; > } > > - /* We don't yet support UDP encapsulation and TFC padding. */ > - if (x->encap || x->tfcpad) { > + is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; > + > + /* We don't yet support UDP encapsulation except full mode and TFC padding. */ > + if ((!is_packet_offload && x->encap) || x->tfcpad) { > NL_SET_ERR_MSG(extack, "Encapsulation and TFC padding can't be offloaded"); > return -EINVAL; > } > @@ -258,7 +260,6 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, > return -EINVAL; > } > > - is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; > dev = dev_get_by_index(net, xuo->ifindex); > if (!dev) { > if (!(xuo->flags & XFRM_OFFLOAD_INBOUND)) { > -- > >
Hi Leon, Indeed the policy check is checking the sec_path lags set after decapsulation, but this has nothing to do with UDP encapsulation, the driver will set them anyway. Regarding the driver support, each driver may restrict NAT-T support in their state_add callback, so in common code it may stay allowed. Thanks, Ilia BR, Ilia Lin On Tue, Jul 18, 2023 at 12:53 PM Leon Romanovsky <leonro@nvidia.com> wrote: > > On Tue, Jul 18, 2023 at 12:24:05PM +0300, Ilia Lin wrote: > > The ESP encapsulation is not supported only in crypto mode. > > In packet offload mode, the RX is bypassing the XFRM, > > so we can enable the encapsulation. > > It is not accurate. RX is bypassed after XFRM validated packet to ensure > that it was really handled by HW. > > However, this patch should come with relevant driver code which should > support ESP over UDP. You can see it here: > > https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/log/?h=xfrm-next > xfrm: Support UDP encapsulation in packet offload mode > net/mlx5e: Support IPsec NAT-T functionality > net/mlx5e: Check for IPsec NAT-T support > net/mlx5: Add relevant capabilities bits to support NAT-T > > Thanks > > > > > Signed-off-by: Ilia Lin <ilia.lin@kernel.org> > > --- > > net/xfrm/xfrm_device.c | 7 ++++--- > > 1 file changed, 4 insertions(+), 3 deletions(-) > > > > diff --git a/net/xfrm/xfrm_device.c b/net/xfrm/xfrm_device.c > > index 4aff76c6f12e0..3018468d97662 100644 > > --- a/net/xfrm/xfrm_device.c > > +++ b/net/xfrm/xfrm_device.c > > @@ -246,8 +246,10 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, > > return -EINVAL; > > } > > > > - /* We don't yet support UDP encapsulation and TFC padding. */ > > - if (x->encap || x->tfcpad) { > > + is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; > > + > > + /* We don't yet support UDP encapsulation except full mode and TFC padding. */ > > + if ((!is_packet_offload && x->encap) || x->tfcpad) { > > NL_SET_ERR_MSG(extack, "Encapsulation and TFC padding can't be offloaded"); > > return -EINVAL; > > } > > @@ -258,7 +260,6 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, > > return -EINVAL; > > } > > > > - is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; > > dev = dev_get_by_index(net, xuo->ifindex); > > if (!dev) { > > if (!(xuo->flags & XFRM_OFFLOAD_INBOUND)) { > > -- > > > >
On Tue, Jul 18, 2023 at 01:15:12PM +0300, Ilia Lin wrote: > Hi Leon, Please don't top-post your replies. > > Indeed the policy check is checking the sec_path lags set after > decapsulation, but this has nothing to do with UDP encapsulation, the > driver will set them anyway. It doesn't make commit message correct. "In packet offload mode, the RX is bypassing the XFRM, so we can enable the encapsulation." > Regarding the driver support, each driver may restrict NAT-T support > in their state_add callback, so in common code it may stay allowed. We don't support out-of-tree drivers. Please submit it together with relevant driver changes which exercise your newly opened path. If you want, you can even take my series and submit it. Thanks > > Thanks, > Ilia > > BR, > Ilia Lin > > > > On Tue, Jul 18, 2023 at 12:53 PM Leon Romanovsky <leonro@nvidia.com> wrote: > > > > On Tue, Jul 18, 2023 at 12:24:05PM +0300, Ilia Lin wrote: > > > The ESP encapsulation is not supported only in crypto mode. > > > In packet offload mode, the RX is bypassing the XFRM, > > > so we can enable the encapsulation. > > > > It is not accurate. RX is bypassed after XFRM validated packet to ensure > > that it was really handled by HW. > > > > However, this patch should come with relevant driver code which should > > support ESP over UDP. You can see it here: > > > > https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/log/?h=xfrm-next > > xfrm: Support UDP encapsulation in packet offload mode > > net/mlx5e: Support IPsec NAT-T functionality > > net/mlx5e: Check for IPsec NAT-T support > > net/mlx5: Add relevant capabilities bits to support NAT-T > > > > Thanks > > > > > > > > Signed-off-by: Ilia Lin <ilia.lin@kernel.org> > > > --- > > > net/xfrm/xfrm_device.c | 7 ++++--- > > > 1 file changed, 4 insertions(+), 3 deletions(-) > > > > > > diff --git a/net/xfrm/xfrm_device.c b/net/xfrm/xfrm_device.c > > > index 4aff76c6f12e0..3018468d97662 100644 > > > --- a/net/xfrm/xfrm_device.c > > > +++ b/net/xfrm/xfrm_device.c > > > @@ -246,8 +246,10 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, > > > return -EINVAL; > > > } > > > > > > - /* We don't yet support UDP encapsulation and TFC padding. */ > > > - if (x->encap || x->tfcpad) { > > > + is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; > > > + > > > + /* We don't yet support UDP encapsulation except full mode and TFC padding. */ > > > + if ((!is_packet_offload && x->encap) || x->tfcpad) { > > > NL_SET_ERR_MSG(extack, "Encapsulation and TFC padding can't be offloaded"); > > > return -EINVAL; > > > } > > > @@ -258,7 +260,6 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, > > > return -EINVAL; > > > } > > > > > > - is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; > > > dev = dev_get_by_index(net, xuo->ifindex); > > > if (!dev) { > > > if (!(xuo->flags & XFRM_OFFLOAD_INBOUND)) { > > > -- > > > > > > >
diff --git a/net/xfrm/xfrm_device.c b/net/xfrm/xfrm_device.c index 4aff76c6f12e0..3018468d97662 100644 --- a/net/xfrm/xfrm_device.c +++ b/net/xfrm/xfrm_device.c @@ -246,8 +246,10 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, return -EINVAL; } - /* We don't yet support UDP encapsulation and TFC padding. */ - if (x->encap || x->tfcpad) { + is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; + + /* We don't yet support UDP encapsulation except full mode and TFC padding. */ + if ((!is_packet_offload && x->encap) || x->tfcpad) { NL_SET_ERR_MSG(extack, "Encapsulation and TFC padding can't be offloaded"); return -EINVAL; } @@ -258,7 +260,6 @@ int xfrm_dev_state_add(struct net *net, struct xfrm_state *x, return -EINVAL; } - is_packet_offload = xuo->flags & XFRM_OFFLOAD_PACKET; dev = dev_get_by_index(net, xuo->ifindex); if (!dev) { if (!(xuo->flags & XFRM_OFFLOAD_INBOUND)) {
The ESP encapsulation is not supported only in crypto mode. In packet offload mode, the RX is bypassing the XFRM, so we can enable the encapsulation. Signed-off-by: Ilia Lin <ilia.lin@kernel.org> --- net/xfrm/xfrm_device.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) --