[PATCH ipsec-next 0/3] xfrm: interface: Add unstable helpers for XFRM metadata

[PATCH ipsec-next 0/3] xfrm: interface: Add unstable helpers for XFRM metadata

[Eyal Birger]

This patch series adds xfrm metadata helpers using the unstable kfunc
call interface for the TC-BPF hooks.

This allows steering traffic towards different IPsec connections based
on logic implemented in bpf programs.

The helpers are integrated into the xfrm_interface module. For this
purpose the main functionality of this module is moved to
xfrm_interface_core.c.

Eyal Birger (3):
  xfrm: interface: rename xfrm_interface.c to xfrm_interface_core.c
  xfrm: interface: Add unstable helpers for setting/getting XFRM
    metadata from TC-BPF
  selftests/bpf: add xfrm_info tests

 include/net/dst_metadata.h                    |   1 +
 include/net/xfrm.h                            |  20 +
 net/core/dst.c                                |   4 +
 net/xfrm/Makefile                             |   8 +
 net/xfrm/xfrm_interface_bpf.c                 |  92 +++++
 ...xfrm_interface.c => xfrm_interface_core.c} |  15 +
 tools/testing/selftests/bpf/config            |   2 +
 .../selftests/bpf/prog_tests/test_xfrm_info.c | 342 ++++++++++++++++++
 .../selftests/bpf/progs/test_xfrm_info_kern.c |  74 ++++
 9 files changed, 558 insertions(+)
 create mode 100644 net/xfrm/xfrm_interface_bpf.c
 rename net/xfrm/{xfrm_interface.c => xfrm_interface_core.c} (98%)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/test_xfrm_info.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_xfrm_info_kern.c

-- 
2.34.1

[PATCH ipsec-next 1/3] xfrm: interface: rename xfrm_interface.c to xfrm_interface_core.c

[Eyal Birger]

This change allows adding additional files to the xfrm_interface module.

Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
---
 net/xfrm/Makefile                                    | 2 ++
 net/xfrm/{xfrm_interface.c => xfrm_interface_core.c} | 0
 2 files changed, 2 insertions(+)
 rename net/xfrm/{xfrm_interface.c => xfrm_interface_core.c} (100%)

diff --git a/net/xfrm/Makefile b/net/xfrm/Makefile
index 494aa744bfb9..08a2870fdd36 100644
--- a/net/xfrm/Makefile
+++ b/net/xfrm/Makefile
@@ -3,6 +3,8 @@
 # Makefile for the XFRM subsystem.
 #
 
+xfrm_interface-$(CONFIG_XFRM_INTERFACE) += xfrm_interface_core.o
+
 obj-$(CONFIG_XFRM) := xfrm_policy.o xfrm_state.o xfrm_hash.o \
 		      xfrm_input.o xfrm_output.o \
 		      xfrm_sysctl.o xfrm_replay.o xfrm_device.o
diff --git a/net/xfrm/xfrm_interface.c b/net/xfrm/xfrm_interface_core.c
similarity index 100%
rename from net/xfrm/xfrm_interface.c
rename to net/xfrm/xfrm_interface_core.c
-- 
2.34.1

[PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Eyal Birger]

This change adds xfrm metadata helpers using the unstable kfunc call
interface for the TC-BPF hooks. This allows steering traffic towards
different IPsec connections based on logic implemented in bpf programs.

This object is built based on the availabilty of BTF debug info.

The metadata percpu dsts used on TX take ownership of the original skb
dsts so that they may be used as part of the xfrm transmittion logic -
e.g.  for MTU calculations.

Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
---
 include/net/dst_metadata.h     |  1 +
 include/net/xfrm.h             | 20 ++++++++
 net/core/dst.c                 |  4 ++
 net/xfrm/Makefile              |  6 +++
 net/xfrm/xfrm_interface_bpf.c  | 92 ++++++++++++++++++++++++++++++++++
 net/xfrm/xfrm_interface_core.c | 15 ++++++
 6 files changed, 138 insertions(+)
 create mode 100644 net/xfrm/xfrm_interface_bpf.c

diff --git a/include/net/dst_metadata.h b/include/net/dst_metadata.h
index a454cf4327fe..1b7fae4c6b24 100644
--- a/include/net/dst_metadata.h
+++ b/include/net/dst_metadata.h
@@ -26,6 +26,7 @@ struct macsec_info {
 struct xfrm_md_info {
 	u32 if_id;
 	int link;
+	struct dst_entry *dst_orig;
 };
 
 struct metadata_dst {
diff --git a/include/net/xfrm.h b/include/net/xfrm.h
index e0cc6791c001..5e5fea3087b6 100644
--- a/include/net/xfrm.h
+++ b/include/net/xfrm.h
@@ -2086,4 +2086,24 @@ static inline bool xfrm6_local_dontfrag(const struct sock *sk)
 	return false;
 }
 #endif
+
+#if (IS_BUILTIN(CONFIG_XFRM_INTERFACE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF)) || \
+    (IS_MODULE(CONFIG_XFRM_INTERFACE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
+
+extern int register_xfrm_interface_bpf(void);
+extern void cleanup_xfrm_interface_bpf(void);
+
+#else
+
+static inline int register_xfrm_interface_bpf(void)
+{
+	return 0;
+}
+
+static inline void cleanup_xfrm_interface_bpf(void)
+{
+}
+
+#endif
+
 #endif	/* _NET_XFRM_H */
diff --git a/net/core/dst.c b/net/core/dst.c
index bc9c9be4e080..4c2eb7e56dab 100644
--- a/net/core/dst.c
+++ b/net/core/dst.c
@@ -315,6 +315,8 @@ void metadata_dst_free(struct metadata_dst *md_dst)
 #ifdef CONFIG_DST_CACHE
 	if (md_dst->type == METADATA_IP_TUNNEL)
 		dst_cache_destroy(&md_dst->u.tun_info.dst_cache);
+	else if (md_dst->type == METADATA_XFRM)
+		dst_release(md_dst->u.xfrm_info.dst_orig);
 #endif
 	kfree(md_dst);
 }
@@ -348,6 +350,8 @@ void metadata_dst_free_percpu(struct metadata_dst __percpu *md_dst)
 
 		if (one_md_dst->type == METADATA_IP_TUNNEL)
 			dst_cache_destroy(&one_md_dst->u.tun_info.dst_cache);
+		else if (one_md_dst->type == METADATA_XFRM)
+			dst_release(one_md_dst->u.xfrm_info.dst_orig);
 	}
 #endif
 	free_percpu(md_dst);
diff --git a/net/xfrm/Makefile b/net/xfrm/Makefile
index 08a2870fdd36..cd47f88921f5 100644
--- a/net/xfrm/Makefile
+++ b/net/xfrm/Makefile
@@ -5,6 +5,12 @@
 
 xfrm_interface-$(CONFIG_XFRM_INTERFACE) += xfrm_interface_core.o
 
+ifeq ($(CONFIG_XFRM_INTERFACE),m)
+xfrm_interface-$(CONFIG_DEBUG_INFO_BTF_MODULES) += xfrm_interface_bpf.o
+else ifeq ($(CONFIG_XFRM_INTERFACE),y)
+xfrm_interface-$(CONFIG_DEBUG_INFO_BTF) += xfrm_interface_bpf.o
+endif
+
 obj-$(CONFIG_XFRM) := xfrm_policy.o xfrm_state.o xfrm_hash.o \
 		      xfrm_input.o xfrm_output.o \
 		      xfrm_sysctl.o xfrm_replay.o xfrm_device.o
diff --git a/net/xfrm/xfrm_interface_bpf.c b/net/xfrm/xfrm_interface_bpf.c
new file mode 100644
index 000000000000..d3997ab7cc28
--- /dev/null
+++ b/net/xfrm/xfrm_interface_bpf.c
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Unstable XFRM Helpers for TC-BPF hook
+ *
+ * These are called from SCHED_CLS BPF programs. Note that it is
+ * allowed to break compatibility for these functions since the interface they
+ * are exposed through to BPF programs is explicitly unstable.
+ */
+
+#include <linux/bpf.h>
+#include <linux/btf_ids.h>
+
+#include <net/dst_metadata.h>
+#include <net/xfrm.h>
+
+struct bpf_xfrm_info {
+	u32 if_id;
+	int link;
+};
+
+static struct metadata_dst __percpu *xfrm_md_dst;
+__diag_push();
+__diag_ignore_all("-Wmissing-prototypes",
+		  "Global functions as their definitions will be in xfrm_interface BTF");
+
+__used noinline
+int bpf_skb_get_xfrm_info(struct __sk_buff *skb_ctx, struct bpf_xfrm_info *to)
+{
+	struct sk_buff *skb = (struct sk_buff *)skb_ctx;
+	struct xfrm_md_info *info;
+
+	memset(to, 0, sizeof(*to));
+
+	info = skb_xfrm_md_info(skb);
+	if (!info)
+		return -EINVAL;
+
+	to->if_id = info->if_id;
+	to->link = info->link;
+	return 0;
+}
+
+__used noinline
+int bpf_skb_set_xfrm_info(struct __sk_buff *skb_ctx,
+			  const struct bpf_xfrm_info *from)
+{
+	struct sk_buff *skb = (struct sk_buff *)skb_ctx;
+	struct metadata_dst *md_dst;
+	struct xfrm_md_info *info;
+
+	if (unlikely(skb_metadata_dst(skb)))
+		return -EINVAL;
+
+	md_dst = this_cpu_ptr(xfrm_md_dst);
+
+	info = &md_dst->u.xfrm_info;
+	memset(info, 0, sizeof(*info));
+
+	info->if_id = from->if_id;
+	info->link = from->link;
+	info->dst_orig = skb_dst(skb);
+
+	dst_hold((struct dst_entry *)md_dst);
+	skb_dst_set(skb, (struct dst_entry *)md_dst);
+	return 0;
+}
+
+__diag_pop()
+
+BTF_SET8_START(xfrm_ifc_kfunc_set)
+BTF_ID_FLAGS(func, bpf_skb_get_xfrm_info)
+BTF_ID_FLAGS(func, bpf_skb_set_xfrm_info)
+BTF_SET8_END(xfrm_ifc_kfunc_set)
+
+static const struct btf_kfunc_id_set xfrm_interface_kfunc_set = {
+	.owner = THIS_MODULE,
+	.set   = &xfrm_ifc_kfunc_set,
+};
+
+int __init register_xfrm_interface_bpf(void)
+{
+	xfrm_md_dst = metadata_dst_alloc_percpu(0, METADATA_XFRM,
+						GFP_KERNEL);
+	if (!xfrm_md_dst)
+		return -ENOMEM;
+	return register_btf_kfunc_id_set(BPF_PROG_TYPE_SCHED_CLS,
+					 &xfrm_interface_kfunc_set);
+}
+
+void __exit cleanup_xfrm_interface_bpf(void)
+{
+	metadata_dst_free_percpu(xfrm_md_dst);
+}
diff --git a/net/xfrm/xfrm_interface_core.c b/net/xfrm/xfrm_interface_core.c
index 5a67b120c4db..1e1e8e965939 100644
--- a/net/xfrm/xfrm_interface_core.c
+++ b/net/xfrm/xfrm_interface_core.c
@@ -396,6 +396,14 @@ xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
 
 		if_id = md_info->if_id;
 		fl->flowi_oif = md_info->link;
+		if (md_info->dst_orig) {
+			struct dst_entry *tmp_dst = dst;
+
+			dst = md_info->dst_orig;
+			skb_dst_set(skb, dst);
+			md_info->dst_orig = NULL;
+			dst_release(tmp_dst);
+		}
 	} else {
 		if_id = xi->p.if_id;
 	}
@@ -1162,12 +1170,18 @@ static int __init xfrmi_init(void)
 	if (err < 0)
 		goto rtnl_link_failed;
 
+	err = register_xfrm_interface_bpf();
+	if (err < 0)
+		goto kfunc_failed;
+
 	lwtunnel_encap_add_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
 
 	xfrm_if_register_cb(&xfrm_if_cb);
 
 	return err;
 
+kfunc_failed:
+	rtnl_link_unregister(&xfrmi_link_ops);
 rtnl_link_failed:
 	xfrmi6_fini();
 xfrmi6_failed:
@@ -1183,6 +1197,7 @@ static void __exit xfrmi_fini(void)
 {
 	xfrm_if_unregister_cb();
 	lwtunnel_encap_del_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
+	cleanup_xfrm_interface_bpf();
 	rtnl_link_unregister(&xfrmi_link_ops);
 	xfrmi4_fini();
 	xfrmi6_fini();
-- 
2.34.1

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Martin KaFai Lau]

On 11/28/22 8:05 AM, Eyal Birger wrote:
> This change adds xfrm metadata helpers using the unstable kfunc call
> interface for the TC-BPF hooks. This allows steering traffic towards
> different IPsec connections based on logic implemented in bpf programs.
> 
> This object is built based on the availabilty of BTF debug info.
> 
> The metadata percpu dsts used on TX take ownership of the original skb
> dsts so that they may be used as part of the xfrm transmittion logic -
> e.g.  for MTU calculations.

A few quick comments and questions:

> 
> Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
> ---
>   include/net/dst_metadata.h     |  1 +
>   include/net/xfrm.h             | 20 ++++++++
>   net/core/dst.c                 |  4 ++
>   net/xfrm/Makefile              |  6 +++
>   net/xfrm/xfrm_interface_bpf.c  | 92 ++++++++++++++++++++++++++++++++++

Please tag for bpf-next

>   net/xfrm/xfrm_interface_core.c | 15 ++++++
>   6 files changed, 138 insertions(+)
>   create mode 100644 net/xfrm/xfrm_interface_bpf.c
> 
> diff --git a/include/net/dst_metadata.h b/include/net/dst_metadata.h
> index a454cf4327fe..1b7fae4c6b24 100644
> --- a/include/net/dst_metadata.h
> +++ b/include/net/dst_metadata.h
> @@ -26,6 +26,7 @@ struct macsec_info {
>   struct xfrm_md_info {
>   	u32 if_id;
>   	int link;
> +	struct dst_entry *dst_orig;
>   };
>   
>   struct metadata_dst {

[ ... ]

> diff --git a/net/core/dst.c b/net/core/dst.c
> index bc9c9be4e080..4c2eb7e56dab 100644
> --- a/net/core/dst.c
> +++ b/net/core/dst.c
> @@ -315,6 +315,8 @@ void metadata_dst_free(struct metadata_dst *md_dst)
>   #ifdef CONFIG_DST_CACHE
>   	if (md_dst->type == METADATA_IP_TUNNEL)
>   		dst_cache_destroy(&md_dst->u.tun_info.dst_cache);
> +	else if (md_dst->type == METADATA_XFRM)
> +		dst_release(md_dst->u.xfrm_info.dst_orig);

Why only release dst_orig under CONFIG_DST_CACHE?

>   #endif
>   	kfree(md_dst);
>   }
> @@ -348,6 +350,8 @@ void metadata_dst_free_percpu(struct metadata_dst __percpu *md_dst)
>   
>   		if (one_md_dst->type == METADATA_IP_TUNNEL)
>   			dst_cache_destroy(&one_md_dst->u.tun_info.dst_cache);
> +		else if (one_md_dst->type == METADATA_XFRM)
> +			dst_release(one_md_dst->u.xfrm_info.dst_orig);

Same here.

[ ... ]

> diff --git a/net/xfrm/xfrm_interface_bpf.c b/net/xfrm/xfrm_interface_bpf.c
> new file mode 100644
> index 000000000000..d3997ab7cc28
> --- /dev/null
> +++ b/net/xfrm/xfrm_interface_bpf.c
> @@ -0,0 +1,92 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/* Unstable XFRM Helpers for TC-BPF hook
> + *
> + * These are called from SCHED_CLS BPF programs. Note that it is
> + * allowed to break compatibility for these functions since the interface they
> + * are exposed through to BPF programs is explicitly unstable.
> + */
> +
> +#include <linux/bpf.h>
> +#include <linux/btf_ids.h>
> +
> +#include <net/dst_metadata.h>
> +#include <net/xfrm.h>
> +
> +struct bpf_xfrm_info {
> +	u32 if_id;
> +	int link;
> +};
> +
> +static struct metadata_dst __percpu *xfrm_md_dst;
> +__diag_push();
> +__diag_ignore_all("-Wmissing-prototypes",
> +		  "Global functions as their definitions will be in xfrm_interface BTF");
> +
> +__used noinline
> +int bpf_skb_get_xfrm_info(struct __sk_buff *skb_ctx, struct bpf_xfrm_info *to)
> +{
> +	struct sk_buff *skb = (struct sk_buff *)skb_ctx;
> +	struct xfrm_md_info *info;
> +
> +	memset(to, 0, sizeof(*to));
> +
> +	info = skb_xfrm_md_info(skb);
> +	if (!info)
> +		return -EINVAL;
> +
> +	to->if_id = info->if_id;
> +	to->link = info->link;
> +	return 0;
> +}
> +
> +__used noinline
> +int bpf_skb_set_xfrm_info(struct __sk_buff *skb_ctx,
> +			  const struct bpf_xfrm_info *from)
> +{
> +	struct sk_buff *skb = (struct sk_buff *)skb_ctx;
> +	struct metadata_dst *md_dst;
> +	struct xfrm_md_info *info;
> +
> +	if (unlikely(skb_metadata_dst(skb)))
> +		return -EINVAL;
> +
> +	md_dst = this_cpu_ptr(xfrm_md_dst);
> +
> +	info = &md_dst->u.xfrm_info;
> +	memset(info, 0, sizeof(*info));
> +
> +	info->if_id = from->if_id;
> +	info->link = from->link;
> +	info->dst_orig = skb_dst(skb);
However, the dst_orig init is not done under CONFIG_DST_CACHE though...

Also, is it possible that skb->_skb_refdst has SKB_DST_NOREF set and later below 
... (contd)

> +
> +	dst_hold((struct dst_entry *)md_dst);
> +	skb_dst_set(skb, (struct dst_entry *)md_dst);
> +	return 0;
> +}
> +
> +__diag_pop()
> +
> +BTF_SET8_START(xfrm_ifc_kfunc_set)
> +BTF_ID_FLAGS(func, bpf_skb_get_xfrm_info)
> +BTF_ID_FLAGS(func, bpf_skb_set_xfrm_info)
> +BTF_SET8_END(xfrm_ifc_kfunc_set)
> +
> +static const struct btf_kfunc_id_set xfrm_interface_kfunc_set = {
> +	.owner = THIS_MODULE,
> +	.set   = &xfrm_ifc_kfunc_set,
> +};
> +
> +int __init register_xfrm_interface_bpf(void)
> +{
> +	xfrm_md_dst = metadata_dst_alloc_percpu(0, METADATA_XFRM,
> +						GFP_KERNEL);
> +	if (!xfrm_md_dst)
> +		return -ENOMEM;
> +	return register_btf_kfunc_id_set(BPF_PROG_TYPE_SCHED_CLS,
> +					 &xfrm_interface_kfunc_set);

Will cleanup_xfrm_interface_bpf() be called during error ?

> +}
> +
> +void __exit cleanup_xfrm_interface_bpf(void)
> +{
> +	metadata_dst_free_percpu(xfrm_md_dst);
> +}
> diff --git a/net/xfrm/xfrm_interface_core.c b/net/xfrm/xfrm_interface_core.c
> index 5a67b120c4db..1e1e8e965939 100644
> --- a/net/xfrm/xfrm_interface_core.c
> +++ b/net/xfrm/xfrm_interface_core.c
> @@ -396,6 +396,14 @@ xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
>   
>   		if_id = md_info->if_id;
>   		fl->flowi_oif = md_info->link;
> +		if (md_info->dst_orig) {
> +			struct dst_entry *tmp_dst = dst;
> +
> +			dst = md_info->dst_orig;
> +			skb_dst_set(skb, dst);

(contd) ... skb_dst_set() is always called here.  (considering there is 
skb_dst_set_noref()).


> +			md_info->dst_orig = NULL;
> +			dst_release(tmp_dst);
> +		}
>   	} else {
>   		if_id = xi->p.if_id;
>   	}
> @@ -1162,12 +1170,18 @@ static int __init xfrmi_init(void)
>   	if (err < 0)
>   		goto rtnl_link_failed;
>   
> +	err = register_xfrm_interface_bpf();
> +	if (err < 0)
> +		goto kfunc_failed;
> +
>   	lwtunnel_encap_add_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
>   
>   	xfrm_if_register_cb(&xfrm_if_cb);
>   
>   	return err;
>   
> +kfunc_failed:
> +	rtnl_link_unregister(&xfrmi_link_ops);
>   rtnl_link_failed:
>   	xfrmi6_fini();
>   xfrmi6_failed:
> @@ -1183,6 +1197,7 @@ static void __exit xfrmi_fini(void)
>   {
>   	xfrm_if_unregister_cb();
>   	lwtunnel_encap_del_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
> +	cleanup_xfrm_interface_bpf();
>   	rtnl_link_unregister(&xfrmi_link_ops);
>   	xfrmi4_fini();
>   	xfrmi6_fini();

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Eyal Birger]

(sent again in plain text, sorry for the noise).

Hi Martin.

On Tue, Nov 29, 2022 at 3:58 AM Martin KaFai Lau <martin.lau@linux.dev> wrote:
>
> On 11/28/22 8:05 AM, Eyal Birger wrote:
> > This change adds xfrm metadata helpers using the unstable kfunc call
> > interface for the TC-BPF hooks. This allows steering traffic towards
> > different IPsec connections based on logic implemented in bpf programs.
> >
> > This object is built based on the availabilty of BTF debug info.
> >
> > The metadata percpu dsts used on TX take ownership of the original skb
> > dsts so that they may be used as part of the xfrm transmittion logic -
> > e.g.  for MTU calculations.
>
> A few quick comments and questions:

Thanks for your comments!

>
> >
> > Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
> > ---
> >   include/net/dst_metadata.h     |  1 +
> >   include/net/xfrm.h             | 20 ++++++++
> >   net/core/dst.c                 |  4 ++
> >   net/xfrm/Makefile              |  6 +++
> >   net/xfrm/xfrm_interface_bpf.c  | 92 ++++++++++++++++++++++++++++++++++
>
> Please tag for bpf-next
Sure. I wasn't totally sure which tree this belongs to.

>
> >   net/xfrm/xfrm_interface_core.c | 15 ++++++
> >   6 files changed, 138 insertions(+)
> >   create mode 100644 net/xfrm/xfrm_interface_bpf.c
> >
> > diff --git a/include/net/dst_metadata.h b/include/net/dst_metadata.h
> > index a454cf4327fe..1b7fae4c6b24 100644
> > --- a/include/net/dst_metadata.h
> > +++ b/include/net/dst_metadata.h
> > @@ -26,6 +26,7 @@ struct macsec_info {
> >   struct xfrm_md_info {
> >       u32 if_id;
> >       int link;
> > +     struct dst_entry *dst_orig;
> >   };
> >
> >   struct metadata_dst {
>
> [ ... ]
>
> > diff --git a/net/core/dst.c b/net/core/dst.c
> > index bc9c9be4e080..4c2eb7e56dab 100644
> > --- a/net/core/dst.c
> > +++ b/net/core/dst.c
> > @@ -315,6 +315,8 @@ void metadata_dst_free(struct metadata_dst *md_dst)
> >   #ifdef CONFIG_DST_CACHE
> >       if (md_dst->type == METADATA_IP_TUNNEL)
> >               dst_cache_destroy(&md_dst->u.tun_info.dst_cache);
> > +     else if (md_dst->type == METADATA_XFRM)
> > +             dst_release(md_dst->u.xfrm_info.dst_orig);
>
> Why only release dst_orig under CONFIG_DST_CACHE?
It's a relic from a previous version where I'd used dst cache.
Will move out of this ifdef.

>
> >   #endif
> >       kfree(md_dst);
> >   }
> > @@ -348,6 +350,8 @@ void metadata_dst_free_percpu(struct metadata_dst __percpu *md_dst)
> >
> >               if (one_md_dst->type == METADATA_IP_TUNNEL)
> >                       dst_cache_destroy(&one_md_dst->u.tun_info.dst_cache);
> > +             else if (one_md_dst->type == METADATA_XFRM)
> > +                     dst_release(one_md_dst->u.xfrm_info.dst_orig);
>
> Same here.

Likewise.

>
> [ ... ]
>
> > diff --git a/net/xfrm/xfrm_interface_bpf.c b/net/xfrm/xfrm_interface_bpf.c
> > new file mode 100644
> > index 000000000000..d3997ab7cc28
> > --- /dev/null
> > +++ b/net/xfrm/xfrm_interface_bpf.c
> > @@ -0,0 +1,92 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/* Unstable XFRM Helpers for TC-BPF hook
> > + *
> > + * These are called from SCHED_CLS BPF programs. Note that it is
> > + * allowed to break compatibility for these functions since the interface they
> > + * are exposed through to BPF programs is explicitly unstable.
> > + */
> > +
> > +#include <linux/bpf.h>
> > +#include <linux/btf_ids.h>
> > +
> > +#include <net/dst_metadata.h>
> > +#include <net/xfrm.h>
> > +
> > +struct bpf_xfrm_info {
> > +     u32 if_id;
> > +     int link;
> > +};
> > +
> > +static struct metadata_dst __percpu *xfrm_md_dst;
> > +__diag_push();
> > +__diag_ignore_all("-Wmissing-prototypes",
> > +               "Global functions as their definitions will be in xfrm_interface BTF");
> > +
> > +__used noinline
> > +int bpf_skb_get_xfrm_info(struct __sk_buff *skb_ctx, struct bpf_xfrm_info *to)
> > +{
> > +     struct sk_buff *skb = (struct sk_buff *)skb_ctx;
> > +     struct xfrm_md_info *info;
> > +
> > +     memset(to, 0, sizeof(*to));
> > +
> > +     info = skb_xfrm_md_info(skb);
> > +     if (!info)
> > +             return -EINVAL;
> > +
> > +     to->if_id = info->if_id;
> > +     to->link = info->link;
> > +     return 0;
> > +}
> > +
> > +__used noinline
> > +int bpf_skb_set_xfrm_info(struct __sk_buff *skb_ctx,
> > +                       const struct bpf_xfrm_info *from)
> > +{
> > +     struct sk_buff *skb = (struct sk_buff *)skb_ctx;
> > +     struct metadata_dst *md_dst;
> > +     struct xfrm_md_info *info;
> > +
> > +     if (unlikely(skb_metadata_dst(skb)))
> > +             return -EINVAL;
> > +
> > +     md_dst = this_cpu_ptr(xfrm_md_dst);
> > +
> > +     info = &md_dst->u.xfrm_info;
> > +     memset(info, 0, sizeof(*info));
> > +
> > +     info->if_id = from->if_id;
> > +     info->link = from->link;
> > +     info->dst_orig = skb_dst(skb);
> However, the dst_orig init is not done under CONFIG_DST_CACHE though...
>
> Also, is it possible that skb->_skb_refdst has SKB_DST_NOREF set and later below
> ... (contd)
Nice catch! will force dst is refcounted.

>
> > +
> > +     dst_hold((struct dst_entry *)md_dst);
> > +     skb_dst_set(skb, (struct dst_entry *)md_dst);
> > +     return 0;
> > +}
> > +
> > +__diag_pop()
> > +
> > +BTF_SET8_START(xfrm_ifc_kfunc_set)
> > +BTF_ID_FLAGS(func, bpf_skb_get_xfrm_info)
> > +BTF_ID_FLAGS(func, bpf_skb_set_xfrm_info)
> > +BTF_SET8_END(xfrm_ifc_kfunc_set)
> > +
> > +static const struct btf_kfunc_id_set xfrm_interface_kfunc_set = {
> > +     .owner = THIS_MODULE,
> > +     .set   = &xfrm_ifc_kfunc_set,
> > +};
> > +
> > +int __init register_xfrm_interface_bpf(void)
> > +{
> > +     xfrm_md_dst = metadata_dst_alloc_percpu(0, METADATA_XFRM,
> > +                                             GFP_KERNEL);
> > +     if (!xfrm_md_dst)
> > +             return -ENOMEM;
> > +     return register_btf_kfunc_id_set(BPF_PROG_TYPE_SCHED_CLS,
> > +                                      &xfrm_interface_kfunc_set);
>
> Will cleanup_xfrm_interface_bpf() be called during error ?
No. Will fix in v2.

Thanks!
Eyal.

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Steffen Klassert]

On Mon, Nov 28, 2022 at 05:58:23PM -0800, Martin KaFai Lau wrote:
> On 11/28/22 8:05 AM, Eyal Birger wrote:
> > This change adds xfrm metadata helpers using the unstable kfunc call
> > interface for the TC-BPF hooks. This allows steering traffic towards
> > different IPsec connections based on logic implemented in bpf programs.
> > 
> > This object is built based on the availabilty of BTF debug info.
> > 
> > The metadata percpu dsts used on TX take ownership of the original skb
> > dsts so that they may be used as part of the xfrm transmittion logic -
> > e.g.  for MTU calculations.
> 
> A few quick comments and questions:
> 
> > 
> > Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
> > ---
> >   include/net/dst_metadata.h     |  1 +
> >   include/net/xfrm.h             | 20 ++++++++
> >   net/core/dst.c                 |  4 ++
> >   net/xfrm/Makefile              |  6 +++
> >   net/xfrm/xfrm_interface_bpf.c  | 92 ++++++++++++++++++++++++++++++++++
> 
> Please tag for bpf-next

This is a change to xfrm ipsec, so it should go
through the ipsec-next tree, unless there is
a good reason for handling that different.

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Jakub Kicinski]

On Tue, 29 Nov 2022 10:50:01 +0100 Steffen Klassert wrote:
> > Please tag for bpf-next  
> 
> This is a change to xfrm ipsec, so it should go
> through the ipsec-next tree, unless there is
> a good reason for handling that different.

Yeah, this is borderline. Do the patches apply cleanly to Linus's tree?
If so maybe they can be posted as a PR and both trees can pull them in,
avoiding any unnecessary back and forth...

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Steffen Klassert]

On Tue, Nov 29, 2022 at 08:15:10AM -0800, Jakub Kicinski wrote:
> On Tue, 29 Nov 2022 10:50:01 +0100 Steffen Klassert wrote:
> > > Please tag for bpf-next  
> > 
> > This is a change to xfrm ipsec, so it should go
> > through the ipsec-next tree, unless there is
> > a good reason for handling that different.
> 
> Yeah, this is borderline. Do the patches apply cleanly to Linus's tree?
> If so maybe they can be posted as a PR and both trees can pull them in,
> avoiding any unnecessary back and forth...

Now, after the last PR was merged, the ipsec-next tree is empty,
and we are close to the end of this development cycle. So I don't
expect conflicts if that patchset goes in before the merge window.
If the bpf-next tree wants to take the v2 version, just let me
know. Otherwise I consider taking it into ipsec-next.

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Martin KaFai Lau]

On 11/29/22 8:15 AM, Jakub Kicinski wrote:
> On Tue, 29 Nov 2022 10:50:01 +0100 Steffen Klassert wrote:
>>> Please tag for bpf-next
>>
>> This is a change to xfrm ipsec, so it should go
>> through the ipsec-next tree, unless there is
>> a good reason for handling that different.

The set is mostly depending on the bpf features.  Patch 2 is mostly depending on 
bpf and patch 3 is also a bpf selftest.  I assume the set should have been 
developed based on the bpf-next tree instead.  It is also good to have the test 
run in bpf CI sooner than later to bar on-going bpf changes that may break it. 
It is the reason I think bpf-next makes more sense.

If it is preferred to go through ipsec-next, the set should at least be tested 
against the bpf-next before posting.

https://patchwork.kernel.org/project/netdevbpf/patch/20221129132018.985887-4-eyal.birger@gmail.com/

Re: [PATCH ipsec-next 2/3] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

[Steffen Klassert]

On Wed, Nov 30, 2022 at 11:10:13AM -0800, Martin KaFai Lau wrote:
> On 11/29/22 8:15 AM, Jakub Kicinski wrote:
> > On Tue, 29 Nov 2022 10:50:01 +0100 Steffen Klassert wrote:
> > > > Please tag for bpf-next
> > > 
> > > This is a change to xfrm ipsec, so it should go
> > > through the ipsec-next tree, unless there is
> > > a good reason for handling that different.
> 
> The set is mostly depending on the bpf features.  Patch 2 is mostly
> depending on bpf and patch 3 is also a bpf selftest.  I assume the set
> should have been developed based on the bpf-next tree instead.  It is also
> good to have the test run in bpf CI sooner than later to bar on-going bpf
> changes that may break it. It is the reason I think bpf-next makes more
> sense.

As said, if there is a good reason, I'm ok with routing it
through bpf-next. Looks like there is a good readon, so
go with bpf-next.

[PATCH ipsec-next 3/3] selftests/bpf: add xfrm_info tests

[Eyal Birger]

Test the xfrm_info kfunc helpers.

Note: the tests require support for xfrmi "external" mode in iproute2.

The test setup creates three name spaces - NS0, NS1, NS2.

XFRM tunnels are setup between NS0 and the two other NSs.

The kfunc helpers are used to steer traffic from NS0 to the other
NSs based on a userspace populated map and validate that that
return traffic had arrived from the desired NS.

Signed-off-by: Eyal Birger <eyal.birger@gmail.com>
---
 tools/testing/selftests/bpf/config            |   2 +
 .../selftests/bpf/prog_tests/test_xfrm_info.c | 342 ++++++++++++++++++
 .../selftests/bpf/progs/test_xfrm_info_kern.c |  74 ++++
 3 files changed, 418 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/test_xfrm_info.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_xfrm_info_kern.c

diff --git a/tools/testing/selftests/bpf/config b/tools/testing/selftests/bpf/config
index 9213565c0311..9f39943d6ebd 100644
--- a/tools/testing/selftests/bpf/config
+++ b/tools/testing/selftests/bpf/config
@@ -20,6 +20,7 @@ CONFIG_IKCONFIG_PROC=y
 CONFIG_IMA=y
 CONFIG_IMA_READ_POLICY=y
 CONFIG_IMA_WRITE_POLICY=y
+CONFIG_INET_ESP=y
 CONFIG_IP_NF_FILTER=y
 CONFIG_IP_NF_RAW=y
 CONFIG_IP_NF_TARGET_SYNPROXY=y
@@ -71,3 +72,4 @@ CONFIG_TEST_BPF=y
 CONFIG_USERFAULTFD=y
 CONFIG_VXLAN=y
 CONFIG_XDP_SOCKETS=y
+CONFIG_XFRM_INTERFACE=y
diff --git a/tools/testing/selftests/bpf/prog_tests/test_xfrm_info.c b/tools/testing/selftests/bpf/prog_tests/test_xfrm_info.c
new file mode 100644
index 000000000000..3bd22ce6f00f
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/test_xfrm_info.c
@@ -0,0 +1,342 @@
+// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
+
+/*
+ * Topology:
+ * ---------
+ *   NS0 namespace         |   NS1 namespace        | NS2 namespace
+ *                         |                        |
+ *   +---------------+     |   +---------------+    |
+ *   |    ipsec0     |---------|    ipsec0     |    |
+ *   | 192.168.1.100 |     |   | 192.168.1.200 |    |
+ *   | if_id: bpf    |     |   +---------------+    |
+ *   +---------------+     |                        |
+ *           |             |                        |   +---------------+
+ *           |             |                        |   |    ipsec0     |
+ *           \------------------------------------------| 192.168.1.200 |
+ *                         |                        |   +---------------+
+ *                         |                        |
+ *                         |                        | (overlay network)
+ *      ------------------------------------------------------
+ *                         |                        | (underlay network)
+ *   +--------------+      |   +--------------+     |
+ *   |    veth01    |----------|    veth10    |     |
+ *   | 172.16.1.100 |      |   | 172.16.1.200 |     |
+ *   ---------------+      |   +--------------+     |
+ *                         |                        |
+ *   +--------------+      |                        |   +--------------+
+ *   |    veth02    |-----------------------------------|    veth20    |
+ *   | 172.16.2.100 |      |                        |   | 172.16.2.200 |
+ *   +--------------+      |                        |   +--------------+
+ *
+ *
+ * Test Packet flow
+ * -----------
+ *  The tests perform 'ping 192.168.1.200' from the NS0 namespace:
+ *  1) request is routed to NS0 ipsec0
+ *  2) NS0 ipsec0 tc egress BPF program is triggered and sets the if_id based
+ *     on a map value. This makes the ipsec0 device in external mode select the
+ *     destination tunnel
+ *  3) ping reaches the other namespace (NS1 or NS2 based on which if_id was
+ *     used) and response is sent
+ *  4) response is received on NS0 ipsec0, tc ingress program is triggered and
+ *     records the response if_id in the map
+ *  5) requested if_id is compared with received if_id
+ */
+
+#include <net/if.h>
+
+#include "test_progs.h"
+#include "network_helpers.h"
+#include "test_xfrm_info_kern.skel.h"
+
+#define NS0 "xfrm_test_ns0"
+#define NS1 "xfrm_test_ns1"
+#define NS2 "xfrm_test_ns2"
+
+#define IF_ID_0_TO_1 1
+#define IF_ID_0_TO_2 2
+#define IF_ID_1 3
+#define IF_ID_2 4
+
+#define IP4_ADDR_VETH01 "172.16.1.100"
+#define IP4_ADDR_VETH10 "172.16.1.200"
+#define IP4_ADDR_VETH02 "172.16.2.100"
+#define IP4_ADDR_VETH20 "172.16.2.200"
+
+#define ESP_DUMMY_PARAMS \
+    "proto esp aead 'rfc4106(gcm(aes))' " \
+    "0xe4d8f4b4da1df18a3510b3781496daa82488b713 128 mode tunnel "
+
+#define PING_ARGS "-i 0.01 -c 3 -w 10 -q"
+
+#define SYS(fmt, ...)						\
+	({							\
+		char cmd[1024];					\
+		snprintf(cmd, sizeof(cmd), fmt, ##__VA_ARGS__);	\
+		if (!ASSERT_OK(system(cmd), cmd))		\
+			goto fail;				\
+	})
+
+#define SYS_NOFAIL(fmt, ...)					\
+	({							\
+		char cmd[1024];					\
+		snprintf(cmd, sizeof(cmd), fmt, ##__VA_ARGS__);	\
+		system(cmd);					\
+	})
+
+static int attach_tc_prog(struct bpf_tc_hook *hook, int igr_fd, int egr_fd)
+{
+	DECLARE_LIBBPF_OPTS(bpf_tc_opts, opts1, .handle = 1,
+			    .priority = 1, .prog_fd = igr_fd);
+	DECLARE_LIBBPF_OPTS(bpf_tc_opts, opts2, .handle = 1,
+			    .priority = 1, .prog_fd = egr_fd);
+	int ret;
+
+	ret = bpf_tc_hook_create(hook);
+	if (!ASSERT_OK(ret, "create tc hook"))
+		return ret;
+
+	if (igr_fd >= 0) {
+		hook->attach_point = BPF_TC_INGRESS;
+		ret = bpf_tc_attach(hook, &opts1);
+		if (!ASSERT_OK(ret, "bpf_tc_attach")) {
+			bpf_tc_hook_destroy(hook);
+			return ret;
+		}
+	}
+
+	if (egr_fd >= 0) {
+		hook->attach_point = BPF_TC_EGRESS;
+		ret = bpf_tc_attach(hook, &opts2);
+		if (!ASSERT_OK(ret, "bpf_tc_attach")) {
+			bpf_tc_hook_destroy(hook);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static void cleanup(void)
+{
+	SYS_NOFAIL("test -f /var/run/netns/" NS0 " && ip netns delete " NS0);
+	SYS_NOFAIL("test -f /var/run/netns/" NS1 " && ip netns delete " NS1);
+	SYS_NOFAIL("test -f /var/run/netns/" NS2 " && ip netns delete " NS2);
+}
+
+static int config_underlay(void)
+{
+	SYS("ip netns add " NS0);
+	SYS("ip netns add " NS1);
+	SYS("ip netns add " NS2);
+
+	/* NS0 <-> NS1 [veth01 <-> veth10] */
+	SYS("ip link add veth01 netns " NS0 " type veth peer name veth10 netns " NS1);
+	SYS("ip -net " NS0 " addr add " IP4_ADDR_VETH01 "/24 dev veth01");
+	SYS("ip -net " NS0 " link set dev veth01 up");
+	SYS("ip -net " NS1 " addr add " IP4_ADDR_VETH10 "/24 dev veth10");
+	SYS("ip -net " NS1 " link set dev veth10 up");
+
+	/* NS0 <-> NS2 [veth02 <-> veth20] */
+	SYS("ip link add veth02 netns " NS0 " type veth peer name veth20 netns " NS2);
+	SYS("ip -net " NS0 " addr add " IP4_ADDR_VETH02 "/24 dev veth02");
+	SYS("ip -net " NS0 " link set dev veth02 up");
+	SYS("ip -net " NS2 " addr add " IP4_ADDR_VETH20 "/24 dev veth20");
+	SYS("ip -net " NS2 " link set dev veth20 up");
+
+	return 0;
+fail:
+	return -1;
+}
+
+static int setup_xfrm_tunnel_ns(const char *ns, const char *ipv4_local,
+				const char *ipv4_remote, int if_id)
+{
+	/* State: local -> remote */
+	SYS("ip -net %s xfrm state add src %s dst %s spi 1 "
+	    ESP_DUMMY_PARAMS "if_id %d", ns, ipv4_local, ipv4_remote, if_id);
+
+	/* State: local <- remote */
+	SYS("ip -net %s xfrm state add src %s dst %s spi 1 "
+	    ESP_DUMMY_PARAMS "if_id %d", ns, ipv4_remote, ipv4_local, if_id);
+
+	/* Policy: local -> remote */
+	SYS("ip -net %s xfrm policy add dir out src 0.0.0.0/0 dst 0.0.0.0/0 "
+	    "if_id %d tmpl src %s dst %s proto esp mode tunnel if_id %d", ns,
+	    if_id, ipv4_local, ipv4_remote, if_id);
+
+	/* Policy: local <- remote */
+	SYS("ip -net %s xfrm policy add dir in src 0.0.0.0/0 dst 0.0.0.0/0 "
+	    "if_id %d tmpl src %s dst %s proto esp mode tunnel if_id %d", ns,
+	    if_id, ipv4_remote, ipv4_local, if_id);
+
+	return 0;
+fail:
+	return -1;
+}
+
+static int setup_xfrm_tunnel(const char *ns_a, const char *ns_b,
+			     const char *ipv4_a, const char *ipv4_b,
+			     int if_id_a, int if_id_b)
+{
+	return setup_xfrm_tunnel_ns(ns_a, ipv4_a, ipv4_b, if_id_a) ||
+		setup_xfrm_tunnel_ns(ns_b, ipv4_b, ipv4_a, if_id_b);
+}
+
+static int config_overlay(void)
+{
+	if (setup_xfrm_tunnel(NS0, NS1, IP4_ADDR_VETH01, IP4_ADDR_VETH10,
+			      IF_ID_0_TO_1, IF_ID_1))
+		goto fail;
+	if (setup_xfrm_tunnel(NS0, NS2, IP4_ADDR_VETH02, IP4_ADDR_VETH20,
+			      IF_ID_0_TO_2, IF_ID_2))
+		goto fail;
+
+	SYS("ip -net " NS0 " link add ipsec0 type xfrm external");
+	SYS("ip -net " NS0 " addr add 192.168.1.100/24 dev ipsec0");
+	SYS("ip -net " NS0 " link set dev ipsec0 up");
+
+	SYS("ip -net " NS1 " link add ipsec0 type xfrm if_id %d", IF_ID_1);
+	SYS("ip -net " NS1 " addr add 192.168.1.200/24 dev ipsec0");
+	SYS("ip -net " NS1 " link set dev ipsec0 up");
+
+	SYS("ip -net " NS2 " link add ipsec0 type xfrm if_id %d", IF_ID_2);
+	SYS("ip -net " NS2 " addr add 192.168.1.200/24 dev ipsec0");
+	SYS("ip -net " NS2 " link set dev ipsec0 up");
+
+	return 0;
+fail:
+	return -1;
+}
+
+static int test_ping(int family, const char *addr)
+{
+	SYS("%s %s %s > /dev/null", ping_command(family), PING_ARGS, addr);
+	return 0;
+fail:
+	return -1;
+}
+
+static int test_xfrm_ping(int dst_if_id_map_fd, u32 if_id)
+{
+	u32 dst_if_id;
+	int key, err;
+
+	key = 0;
+	dst_if_id = if_id;
+	err = bpf_map_update_elem(dst_if_id_map_fd, &key, &dst_if_id, BPF_ANY);
+	if (!ASSERT_OK(err, "update bpf dst_if_id_map"))
+		return -1;
+
+	if (test_ping(AF_INET, "192.168.1.200"))
+		return -1;
+
+	key = 1;
+	dst_if_id = 0;
+	err = bpf_map_lookup_elem(dst_if_id_map_fd, &key, &dst_if_id);
+	if (!ASSERT_OK(err, "lookup bpf dst_if_id_map"))
+		return -1;
+
+	if (!ASSERT_EQ(dst_if_id, if_id, "if_id"))
+		return -1;
+
+	return 0;
+}
+
+static void test_xfrm_info(void)
+{
+	int get_xfrm_info_prog_fd, set_xfrm_info_prog_fd;
+	struct test_xfrm_info_kern *skel = NULL;
+	struct nstoken *nstoken = NULL;
+	int dst_if_id_map_fd = -1;
+	int ifindex = -1;
+	DECLARE_LIBBPF_OPTS(bpf_tc_hook, tc_hook,
+			    .attach_point = BPF_TC_INGRESS);
+
+	/* load and attach bpf progs to ipsec dev tc hook point */
+	skel = test_xfrm_info_kern__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "test_xfrm_info_kern__open_and_load"))
+		goto done;
+	nstoken = open_netns(NS0);
+	ifindex = if_nametoindex("ipsec0");
+	if (!ASSERT_NEQ(ifindex, 0, "ipsec0 ifindex"))
+		goto done;
+	tc_hook.ifindex = ifindex;
+	set_xfrm_info_prog_fd = bpf_program__fd(skel->progs.set_xfrm_info);
+	get_xfrm_info_prog_fd = bpf_program__fd(skel->progs.get_xfrm_info);
+	if (!ASSERT_GE(set_xfrm_info_prog_fd, 0, "bpf_program__fd"))
+		goto done;
+	if (!ASSERT_GE(get_xfrm_info_prog_fd, 0, "bpf_program__fd"))
+		goto done;
+	if (attach_tc_prog(&tc_hook, get_xfrm_info_prog_fd,
+	    set_xfrm_info_prog_fd))
+		goto done;
+	dst_if_id_map_fd = bpf_map__fd(skel->maps.dst_if_id_map);
+	if (!ASSERT_GE(dst_if_id_map_fd, 0, "bpf_map__fd"))
+		goto done;
+
+	if (!ASSERT_EQ(test_xfrm_ping(dst_if_id_map_fd, IF_ID_0_TO_1), 0,
+		       "ping " NS1))
+		goto done;
+	if (!ASSERT_EQ(test_xfrm_ping(dst_if_id_map_fd, IF_ID_0_TO_2), 0,
+		       "ping " NS2))
+		goto done;
+
+done:
+	if (nstoken)
+		close_netns(nstoken);
+	if (dst_if_id_map_fd >= 0)
+		close(dst_if_id_map_fd);
+	if (skel)
+		test_xfrm_info_kern__destroy(skel);
+}
+
+#define RUN_TEST(name)							\
+	({								\
+		if (test__start_subtest(#name)) {			\
+			test_ ## name();				\
+		}							\
+	})
+
+static void *test_xfrm_info_run_tests(void *arg)
+{
+	cleanup();
+
+	config_underlay();
+	config_overlay();
+
+	RUN_TEST(xfrm_info);
+
+	cleanup();
+
+	return NULL;
+}
+
+static int probe_iproute2(void)
+{
+	if (SYS_NOFAIL("ip link add type xfrm help 2>&1 | "
+		       "grep external > /dev/null")) {
+		fprintf(stdout, "%s:SKIP: iproute2 with xfrm external support needed for this test\n", __func__);
+		return -1;
+	}
+	return 0;
+}
+
+void serial_test_xfrm_info(void)
+{
+	pthread_t test_thread;
+	int err;
+
+	if (probe_iproute2()) {
+		test__skip();
+		return;
+	}
+
+	/* Run the tests in their own thread to isolate the namespace changes
+	 * so they do not affect the environment of other tests.
+	 * (specifically needed because of unshare(CLONE_NEWNS) in open_netns())
+	 */
+	err = pthread_create(&test_thread, NULL, &test_xfrm_info_run_tests, NULL);
+	if (ASSERT_OK(err, "pthread_create"))
+		ASSERT_OK(pthread_join(test_thread, NULL), "pthread_join");
+}
diff --git a/tools/testing/selftests/bpf/progs/test_xfrm_info_kern.c b/tools/testing/selftests/bpf/progs/test_xfrm_info_kern.c
new file mode 100644
index 000000000000..98991a83c1e9
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/test_xfrm_info_kern.c
@@ -0,0 +1,74 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/bpf.h>
+#include <linux/pkt_cls.h>
+#include <bpf/bpf_helpers.h>
+
+#define log_err(__ret) bpf_printk("ERROR line:%d ret:%d\n", __LINE__, __ret)
+
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__uint(max_entries, 2);
+	__type(key, __u32);
+	__type(value, __u32);
+} dst_if_id_map SEC(".maps");
+
+struct bpf_xfrm_info {
+	__u32 if_id;
+	int link;
+};
+
+int bpf_skb_set_xfrm_info(struct __sk_buff *skb_ctx,
+			  const struct bpf_xfrm_info *from) __ksym;
+int bpf_skb_get_xfrm_info(struct __sk_buff *skb_ctx,
+			  struct bpf_xfrm_info *to) __ksym;
+
+SEC("tc")
+int set_xfrm_info(struct __sk_buff *skb)
+{
+	struct bpf_xfrm_info info = {};
+	__u32 *if_id = NULL;
+	__u32 index = 0;
+	int ret = -1;
+
+	if_id = bpf_map_lookup_elem(&dst_if_id_map, &index);
+	if (!if_id) {
+		log_err(ret);
+		return TC_ACT_SHOT;
+	}
+
+	info.if_id = *if_id;
+	ret = bpf_skb_set_xfrm_info(skb, &info);
+	if (ret < 0) {
+		log_err(ret);
+		return TC_ACT_SHOT;
+	}
+
+	return TC_ACT_UNSPEC;
+}
+
+SEC("tc")
+int get_xfrm_info(struct __sk_buff *skb)
+{
+	struct bpf_xfrm_info info = {};
+	__u32 *if_id = NULL;
+	__u32 index = 1;
+	int ret = -1;
+
+	if_id = bpf_map_lookup_elem(&dst_if_id_map, &index);
+	if (!if_id) {
+		log_err(ret);
+		return TC_ACT_SHOT;
+	}
+
+	ret = bpf_skb_get_xfrm_info(skb, &info);
+	if (ret < 0) {
+		log_err(ret);
+		return TC_ACT_SHOT;
+	}
+
+	*if_id = info.if_id;
+
+	return TC_ACT_UNSPEC;
+}
+
+char _license[] SEC("license") = "GPL";
-- 
2.34.1