Skip to contentSkip to conversation

xfrm: interface: Add unstable helpers for XFRM metadata

PATCH Superseded Eyal Birger 2 participants 12 messages Dec 2, 2022 – Dec 3, 2022

Subsystems (6)

v4 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v4 cover xfrm: interface: Add unstable helpers for XFRM metadata

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.

---

Series changes in v3:
- tag bpf-next tree instead of ipsec-next
- add IFLA_XFRM_COLLECT_METADATA sync patch

Eyal Birger (4):
xfrm: interface: rename xfrm_interface.c to xfrm_interface_core.c
xfrm: interface: Add unstable helpers for setting/getting XFRM
metadata from TC-BPF
tools: add IFLA_XFRM_COLLECT_METADATA to uapi/linux/if_link.h selftests/bpf: add xfrm_info tests

 include/net/dst_metadata.h                    |   1 +
 include/net/xfrm.h                            |  20 +
 net/core/dst.c                                |   8 +-
 net/xfrm/Makefile                             |   8 +
 net/xfrm/xfrm_interface_bpf.c                 | 123 ++++++
 ...xfrm_interface.c => xfrm_interface_core.c} |  15 +
 tools/include/uapi/linux/if_link.h            |   1 +
 tools/testing/selftests/bpf/DENYLIST.s390x    |   1 +
 tools/testing/selftests/bpf/config            |   2 +
 .../selftests/bpf/prog_tests/xfrm_info.c      | 365 ++++++++++++++++++
 .../selftests/bpf/progs/bpf_tracing_net.h     |   3 +
 tools/testing/selftests/bpf/progs/xfrm_info.c |  35 ++
 12 files changed, 580 insertions(+), 2 deletions(-)
 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/xfrm_info.c
 create mode 100644 tools/testing/selftests/bpf/progs/xfrm_info.c
Signature · 1 line
2.34.1
#1 lore raw

PATCH v4 1/4 xfrm: interface: rename xfrm_interface.c to xfrm_interface_core.c

This change allows adding additional files to the xfrm_interface module.

---
 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%)
net/xfrm/Makefile +2 -0
index 494aa744bfb9..08a2870fdd36 100644
@@ -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
net/xfrm/xfrm_interface_core.c renamed from net/xfrm/xfrm_interface.c +0 -0
similarity index 100%
rename from net/xfrm/xfrm_interface.c
rename to net/xfrm/xfrm_interface_core.c
Signature · 1 line
2.34.1
#2 lore raw

PATCH v4 2/4 xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

In reply to 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 availability 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 transmission logic - e.g. for MTU calculations.

---

v4: changes requested by Martin KaFai Lau:
- add kfunc documentation
- remove redundant memset
- minor coding

v3:
- remove redunant memset() as suggested by Martin KaFai Lau
- remove __exit annotation from cleanup() function as it's used from an __init function

v2: changed added following points raised by Martin KaFai Lau:
- make sure dst is refcounted prior to caching
- free dst_orig regardless of CONFIG_DST_CACHE
- call xfrm interface bpf cleanup in case of kfunc registration errors

---
 include/net/dst_metadata.h     |   1 +
 include/net/xfrm.h             |  20 ++++++
 net/core/dst.c                 |   8 ++-
 net/xfrm/Makefile              |   6 ++
 net/xfrm/xfrm_interface_bpf.c  | 123 +++++++++++++++++++++++++++++++++
 net/xfrm/xfrm_interface_core.c |  15 ++++
 6 files changed, 171 insertions(+), 2 deletions(-)
 create mode 100644 net/xfrm/xfrm_interface_bpf.c
include/net/dst_metadata.h +1 -0
index a454cf4327fe..1b7fae4c6b24 100644
@@ -26,6 +26,7 @@ struct macsec_info {
 struct xfrm_md_info {
 	u32 if_id;
 	int link;
+	struct dst_entry *dst_orig;
 };
 
 struct metadata_dst {
include/net/xfrm.h +20 -0
index e0cc6791c001..5e5fea3087b6 100644
@@ -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 */
net/core/dst.c +6 -2
index bc9c9be4e080..bb14a0392388 100644
@@ -316,6 +316,8 @@ void metadata_dst_free(struct metadata_dst *md_dst)
 	if (md_dst->type == METADATA_IP_TUNNEL)
 		dst_cache_destroy(&md_dst->u.tun_info.dst_cache);
 #endif
+	if (md_dst->type == METADATA_XFRM)
+		dst_release(md_dst->u.xfrm_info.dst_orig);
 	kfree(md_dst);
 }
 EXPORT_SYMBOL_GPL(metadata_dst_free);
@@ -340,16 +342,18 @@ EXPORT_SYMBOL_GPL(metadata_dst_alloc_percpu);
 
 void metadata_dst_free_percpu(struct metadata_dst __percpu *md_dst)
 {
-#ifdef CONFIG_DST_CACHE
 	int cpu;
 
 	for_each_possible_cpu(cpu) {
 		struct metadata_dst *one_md_dst = per_cpu_ptr(md_dst, cpu);
 
+#ifdef CONFIG_DST_CACHE
 		if (one_md_dst->type == METADATA_IP_TUNNEL)
 			dst_cache_destroy(&one_md_dst->u.tun_info.dst_cache);
-	}
 #endif
+		if (one_md_dst->type == METADATA_XFRM)
+			dst_release(one_md_dst->u.xfrm_info.dst_orig);
+	}
 	free_percpu(md_dst);
 }
 EXPORT_SYMBOL_GPL(metadata_dst_free_percpu);
net/xfrm/Makefile +6 -0
index 08a2870fdd36..cd47f88921f5 100644
@@ -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
net/xfrm/xfrm_interface_bpf.c new file +123 -0

123 lines not on this page · raw message

net/xfrm/xfrm_interface_core.c +15 -0
index 5a67b120c4db..1e1e8e965939 100644
@@ -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();
Signature · 1 line
2.34.1
#3 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Quoted 12 lines · Eyal Birger
net/xfrm/xfrm_interface_bpf.c new file +12 -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);

net/xfrm/xfrm_interface_bpf.c line 87

Quoted 10 lines
net/xfrm/xfrm_interface_bpf.c new file +10 -0
+
+	info = &md_dst->u.xfrm_info;
+
+	info->if_id = from->if_id;
+	info->link = from->link;
+	skb_dst_force(skb);
+	info->dst_orig = skb_dst(skb);
+
+	dst_hold((struct dst_entry *)md_dst);
+	skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

[ ... ]

Quoted 26 lines
net/xfrm/xfrm_interface_bpf.c new file +26 -0
+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)
+{
+	int err;
+
+	xfrm_md_dst = metadata_dst_alloc_percpu(0, METADATA_XFRM,
+						GFP_KERNEL);
+	if (!xfrm_md_dst)
+		return -ENOMEM;
+	err = register_btf_kfunc_id_set(BPF_PROG_TYPE_SCHED_CLS,
+					&xfrm_interface_kfunc_set);
+	if (err < 0) {
+		metadata_dst_free_percpu(xfrm_md_dst);
+		return err;
+	}
+	return 0;
+}
+
+void cleanup_xfrm_interface_bpf(void)
+{
+	metadata_dst_free_percpu(xfrm_md_dst);
+}
#4 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Hi Martin,

Quoted 28 lines · Martin KaFai Lau
Quoted 12 lines · Eyal Birger
net/xfrm/xfrm_interface_bpf.c new file +12 -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);

net/xfrm/xfrm_interface_bpf.c line 87

Quoted 10 lines
net/xfrm/xfrm_interface_bpf.c new file +10 -0
+
+     info = &md_dst->u.xfrm_info;
+
+     info->if_id = from->if_id;
+     info->link = from->link;
+     skb_dst_force(skb);
+     info->dst_orig = skb_dst(skb);
+
+     dst_hold((struct dst_entry *)md_dst);
+     skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

Quoted 3 lines

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

Oh I think you're right. I missed this.

In order to keep this implementation I suppose it means that the module would not be allowed to be removed upon use of this kfunc. but this could be seen as annoying from the configuration user experience.

Alternatively the metadata dsts can be separately allocated from the kfunc, which is probably the simplest approach to maintain, so I'll work on that approach.

Thanks for noticing this!
Eyal.

#5 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Quoted 36 lines · Eyal Birger

Hi Martin,

Quoted 28 lines · Martin KaFai Lau
Quoted 12 lines · Eyal Birger
net/xfrm/xfrm_interface_bpf.c new file +12 -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);

net/xfrm/xfrm_interface_bpf.c line 87

Quoted 10 lines
net/xfrm/xfrm_interface_bpf.c new file +10 -0
+
+     info = &md_dst->u.xfrm_info;
+
+     info->if_id = from->if_id;
+     info->link = from->link;
+     skb_dst_force(skb);
+     info->dst_orig = skb_dst(skb);
+
+     dst_hold((struct dst_entry *)md_dst);
+     skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

Quoted 3 lines

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

Oh I think you're right. I missed this.

Quoted 7 lines

In order to keep this implementation I suppose it means that the module would not be allowed to be removed upon use of this kfunc. but this could be seen as annoying from the configuration user experience.

Alternatively the metadata dsts can be separately allocated from the kfunc, which is probably the simplest approach to maintain, so I'll work on that approach.

If it means dst_alloc on every skb, it will not be cheap.

Another option is to metadata_dst_alloc_percpu() once during the very first bpf_skb_set_xfrm_info() call and the xfrm_md_dst memory will never be freed. It is a tradeoff but likely the correct one. You can take a look at bpf_get_skb_set_tunnel_proto().

#6 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Quoted 46 lines · Martin KaFai Lau
Quoted 36 lines · Eyal Birger

Hi Martin,

Quoted 28 lines · Martin KaFai Lau
Quoted 12 lines · Eyal Birger
net/xfrm/xfrm_interface_bpf.c new file +12 -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);

net/xfrm/xfrm_interface_bpf.c line 87

Quoted 10 lines
net/xfrm/xfrm_interface_bpf.c new file +10 -0
+
+     info = &md_dst->u.xfrm_info;
+
+     info->if_id = from->if_id;
+     info->link = from->link;
+     skb_dst_force(skb);
+     info->dst_orig = skb_dst(skb);
+
+     dst_hold((struct dst_entry *)md_dst);
+     skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

Quoted 3 lines

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

Oh I think you're right. I missed this.

Quoted 7 lines

In order to keep this implementation I suppose it means that the module would not be allowed to be removed upon use of this kfunc. but this could be seen as annoying from the configuration user experience.

Alternatively the metadata dsts can be separately allocated from the kfunc, which is probably the simplest approach to maintain, so I'll work on that approach.

Quoted 7 lines

If it means dst_alloc on every skb, it will not be cheap.

Another option is to metadata_dst_alloc_percpu() once during the very first bpf_skb_set_xfrm_info() call and the xfrm_md_dst memory will never be freed. It is a tradeoff but likely the correct one. You can take a look at bpf_get_skb_set_tunnel_proto().

Yes, I originally wrote this as a helper similar to the tunnel key helper which uses bpf_get_skb_set_tunnel_proto(), and when converting to kfuncs I kept the
percpu implementation.

However, the set tunnel key code is never unloaded. Whereas taking this approach here would mean that this memory would leak on each module reload iiuc.

Eyal.

#7 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Quoted 55 lines · Eyal Birger
Quoted 46 lines · Martin KaFai Lau
Quoted 36 lines · Eyal Birger

Hi Martin,

Quoted 28 lines · Martin KaFai Lau
Quoted 12 lines · Eyal Birger
net/xfrm/xfrm_interface_bpf.c new file +12 -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);

net/xfrm/xfrm_interface_bpf.c line 87

Quoted 10 lines
net/xfrm/xfrm_interface_bpf.c new file +10 -0
+
+     info = &md_dst->u.xfrm_info;
+
+     info->if_id = from->if_id;
+     info->link = from->link;
+     skb_dst_force(skb);
+     info->dst_orig = skb_dst(skb);
+
+     dst_hold((struct dst_entry *)md_dst);
+     skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

Quoted 3 lines

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

Oh I think you're right. I missed this.

Quoted 7 lines

In order to keep this implementation I suppose it means that the module would not be allowed to be removed upon use of this kfunc. but this could be seen as annoying from the configuration user experience.

Alternatively the metadata dsts can be separately allocated from the kfunc, which is probably the simplest approach to maintain, so I'll work on that approach.

Quoted 7 lines

If it means dst_alloc on every skb, it will not be cheap.

Another option is to metadata_dst_alloc_percpu() once during the very first bpf_skb_set_xfrm_info() call and the xfrm_md_dst memory will never be freed. It is a tradeoff but likely the correct one. You can take a look at bpf_get_skb_set_tunnel_proto().

Quoted 8 lines

Yes, I originally wrote this as a helper similar to the tunnel key helper which uses bpf_get_skb_set_tunnel_proto(), and when converting to kfuncs I kept the
percpu implementation.

However, the set tunnel key code is never unloaded. Whereas taking this approach here would mean that this memory would leak on each module reload iiuc.

'struct metadata_dst __percpu *xfrm_md_dst' cannot be in the xfrm module. filter.c could be an option.

#8 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Hi Martin,

Quoted 62 lines · Martin KaFai Lau
Quoted 60 lines · Eyal Birger
Quoted 46 lines · Martin KaFai Lau
Quoted 36 lines · Eyal Birger

Hi Martin,

Quoted 28 lines · Martin KaFai Lau
Quoted 12 lines · Eyal Birger
net/xfrm/xfrm_interface_bpf.c new file +12 -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);

net/xfrm/xfrm_interface_bpf.c line 87

Quoted 10 lines
net/xfrm/xfrm_interface_bpf.c new file +10 -0
+
+     info = &md_dst->u.xfrm_info;
+
+     info->if_id = from->if_id;
+     info->link = from->link;
+     skb_dst_force(skb);
+     info->dst_orig = skb_dst(skb);
+
+     dst_hold((struct dst_entry *)md_dst);
+     skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

Quoted 3 lines

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

Oh I think you're right. I missed this.

Quoted 7 lines

In order to keep this implementation I suppose it means that the module would not be allowed to be removed upon use of this kfunc. but this could be seen as annoying from the configuration user experience.

Alternatively the metadata dsts can be separately allocated from the kfunc, which is probably the simplest approach to maintain, so I'll work on that approach.

Quoted 7 lines

If it means dst_alloc on every skb, it will not be cheap.

Another option is to metadata_dst_alloc_percpu() once during the very first bpf_skb_set_xfrm_info() call and the xfrm_md_dst memory will never be freed. It is a tradeoff but likely the correct one. You can take a look at bpf_get_skb_set_tunnel_proto().

Yes, I originally wrote this as a helper similar to the tunnel key helper which uses bpf_get_skb_set_tunnel_proto(), and when converting to kfuncs I kept the
percpu implementation.

Quoted 6 lines
Quoted 3 lines

However, the set tunnel key code is never unloaded. Whereas taking this approach here would mean that this memory would leak on each module reload iiuc.

'struct metadata_dst __percpu *xfrm_md_dst' cannot be in the xfrm module. filter.c could be an option.

Looking at it some more, won't the module reference taken by the kfunc btf guarantee that the module can't be unloaded while the kfunc is used by a loaded program?

I tried this using a synthetic test attaching the program to a dummy interface and the module couldn't be unloaded while the program was loaded.

In such case, is it possible for the memory to be freed while there are in-use percpu metadata dsts?

Eyal.

#9 lore raw

Re: [PATCH bpf-next,v4 2/4] xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF

Quoted 77 lines · Eyal Birger

Hi Martin,

Quoted 62 lines · Martin KaFai Lau
Quoted 60 lines · Eyal Birger
Quoted 46 lines · Martin KaFai Lau
Quoted 36 lines · Eyal Birger

Hi Martin,

Quoted 28 lines · Martin KaFai Lau
Quoted 12 lines · Eyal Birger
(quoted diff) +12 -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);

Review

Quoted 10 lines
(quoted diff) +10 -0
+
+     info = &md_dst->u.xfrm_info;
+
+     info->if_id = from->if_id;
+     info->link = from->link;
+     skb_dst_force(skb);
+     info->dst_orig = skb_dst(skb);
+
+     dst_hold((struct dst_entry *)md_dst);
+     skb_dst_set(skb, (struct dst_entry *)md_dst);

I may be missed something obvious and this just came to my mind,

Quoted 3 lines

What stops cleanup_xfrm_interface_bpf() being run while skb is still holding the md_dst?

Oh I think you're right. I missed this.

Quoted 7 lines

In order to keep this implementation I suppose it means that the module would not be allowed to be removed upon use of this kfunc. but this could be seen as annoying from the configuration user experience.

Alternatively the metadata dsts can be separately allocated from the kfunc, which is probably the simplest approach to maintain, so I'll work on that approach.

Quoted 7 lines

If it means dst_alloc on every skb, it will not be cheap.

Another option is to metadata_dst_alloc_percpu() once during the very first bpf_skb_set_xfrm_info() call and the xfrm_md_dst memory will never be freed. It is a tradeoff but likely the correct one. You can take a look at bpf_get_skb_set_tunnel_proto().

Yes, I originally wrote this as a helper similar to the tunnel key helper which uses bpf_get_skb_set_tunnel_proto(), and when converting to kfuncs I kept the
percpu implementation.

Quoted 6 lines
Quoted 3 lines

However, the set tunnel key code is never unloaded. Whereas taking this approach here would mean that this memory would leak on each module reload iiuc.

'struct metadata_dst __percpu *xfrm_md_dst' cannot be in the xfrm module. filter.c could be an option.

Looking at it some more, won't the module reference taken by the kfunc btf guarantee that the module can't be unloaded while the kfunc is used by a loaded program?

Quoted 5 lines

I tried this using a synthetic test attaching the program to a dummy interface and the module couldn't be unloaded while the program was loaded.

In such case, is it possible for the memory to be freed while there are in-use percpu metadata dsts?

Decided to err on the side of caution and avoid the release of the percpu dsts. It seems unlikely that the module could be unloaded while there are in flight skbs pointing to the percpu memory, but it's safer not to rely on this, and the cost is rather minimal, so I agree this is the correct tradeoff.

Eyal.

#10 lore raw

PATCH v4 3/4 tools: add IFLA_XFRM_COLLECT_METADATA to uapi/linux/if_link.h

In reply to Eyal Birger

Needed for XFRM metadata tests.

---
 tools/include/uapi/linux/if_link.h | 1 +
 1 file changed, 1 insertion(+)
tools/include/uapi/linux/if_link.h +1 -0
index 0242f31e339c..901d98b865a1 100644
@@ -673,6 +673,7 @@ enum {
 	IFLA_XFRM_UNSPEC,
 	IFLA_XFRM_LINK,
 	IFLA_XFRM_IF_ID,
+	IFLA_XFRM_COLLECT_METADATA,
 	__IFLA_XFRM_MAX
 };
 
Signature · 1 line
2.34.1
#11 lore raw

PATCH v4 4/4 selftests/bpf: add xfrm_info tests

In reply to Eyal Birger

Test the xfrm_info kfunc helpers.

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 bpf global variable and validate that the return traffic had arrived from the desired NS.

---
v4: changes suggested by Martin KaFai Lau:
- minor coding and wording changes
- use vmlinux.h and bpf_tracing_net.h in test program
- avoid running on s390x in CI as kfuncs aren't supported there

v3: changes suggested by Martin KaFai Lau:
- rename test_xfrm_info{,_kern}.c -> xfrm_info.c
- avoid running in a separate thread
- simplify test setup
- verify underlay/overlay success
- create NS0 xfrmi external device using netlink to avoid dependency on newer iproute2
- remove use of bpf_trace_printk() in test programs
- add "preserve_access_index" attribute annotation in test program struct definition
- use bss variables access instead of a map for userspace/bpf interface

v2:
- use an lwt route in NS1 for testing that flow as well
- indendation fix

---
 tools/testing/selftests/bpf/DENYLIST.s390x    |   1 +
 tools/testing/selftests/bpf/config            |   2 +
 .../selftests/bpf/prog_tests/xfrm_info.c      | 365 ++++++++++++++++++
 .../selftests/bpf/progs/bpf_tracing_net.h     |   3 +
 tools/testing/selftests/bpf/progs/xfrm_info.c |  35 ++
 5 files changed, 406 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/xfrm_info.c
 create mode 100644 tools/testing/selftests/bpf/progs/xfrm_info.c
tools/testing/selftests/bpf/DENYLIST.s390x +1 -0
index 3481f3a5ea6f..585fcf73c731 100644
@@ -85,3 +85,4 @@ xdp_bonding                              # failed to auto-attach program 'trace_
 xdp_bpf2bpf                              # failed to auto-attach program 'trace_on_entry': -524                        (trampoline)
 xdp_do_redirect                          # prog_run_max_size unexpected error: -22 (errno 22)
 xdp_synproxy                             # JIT does not support calling kernel function                                (kfunc)
+xfrm_info                                # JIT does not support calling kernel function                                (kfunc)
tools/testing/selftests/bpf/config +2 -0
index f9034ea00bc9..3543c76cef56 100644
@@ -23,6 +23,7 @@ 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
@@ -74,3 +75,4 @@ CONFIG_TEST_BPF=y
 CONFIG_USERFAULTFD=y
 CONFIG_VXLAN=y
 CONFIG_XDP_SOCKETS=y
+CONFIG_XFRM_INTERFACE=y
tools/testing/selftests/bpf/prog_tests/xfrm_info.c new file +365 -0

365 lines not on this page · raw message

tools/testing/selftests/bpf/progs/bpf_tracing_net.h +3 -0
index adb087aecc9e..b394817126cf 100644
@@ -25,6 +25,9 @@
 #define IPV6_TCLASS		67
 #define IPV6_AUTOFLOWLABEL	70
 
+#define TC_ACT_UNSPEC		(-1)
+#define TC_ACT_SHOT		2
+
 #define SOL_TCP			6
 #define TCP_NODELAY		1
 #define TCP_MAXSEG		2
tools/testing/selftests/bpf/progs/xfrm_info.c new file +35 -0
new file mode 100644
index 000000000000..3acedcdd962d
@@ -0,0 +1,35 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "vmlinux.h"
+#include "bpf_tracing_net.h"
+#include <bpf/bpf_helpers.h>
+
+__u32 req_if_id;
+__u32 resp_if_id;
+
+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 = { .if_id = req_if_id };
+
+	return bpf_skb_set_xfrm_info(skb, &info) ? TC_ACT_SHOT : TC_ACT_UNSPEC;
+}
+
+SEC("tc")
+int get_xfrm_info(struct __sk_buff *skb)
+{
+	struct bpf_xfrm_info info = {};
+
+	if (bpf_skb_get_xfrm_info(skb, &info) < 0)
+		return TC_ACT_SHOT;
+
+	resp_if_id = info.if_id;
+
+	return TC_ACT_UNSPEC;
+}
+
+char _license[] SEC("license") = "GPL";
Signature · 1 line
2.34.1