Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Feedback-ID: i6a694271:Fastmail
From:   Daniel Xu <dxu@dxuuu.xyz>
To:     "David S. Miller" <davem@davemloft.net>,
        Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>,
        David Ahern <dsahern@kernel.org>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>
Cc:     ppenkov@aviatrix.com, dbird@aviatrix.com,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        bpf@vger.kernel.org
Subject: [PATCH bpf-next 3/6] bpf, net, frags: Add bpf_ip_check_defrag() kfunc
Date:   Wed, 14 Dec 2022 16:25:30 -0700
Message-Id: <1f48a340a898c4d22d65e0e445dbf15f72081b9a.1671049840.git.dxu@dxuuu.xyz>
In-Reply-To: <cover.1671049840.git.dxu@dxuuu.xyz>
References: <cover.1671049840.git.dxu@dxuuu.xyz>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

This kfunc is used to defragment IPv4 packets. The idea is that if you
see a fragmented packet, you call this kfunc. If the kfunc returns 0,
then the skb has been updated to contain the entire reassembled packet.

If the kfunc returns an error (most likely -EINPROGRESS), then it means
the skb is part of a yet-incomplete original packet. A reasonable
response to -EINPROGRESS is to drop the packet, as the ip defrag
infrastructure is already hanging onto the frag for future reassembly.

Care has been taken to ensure the prog skb remains valid no matter what
the underlying ip_check_defrag() call does. This is in contrast to
ip_defrag(), which may consume the skb if the skb is part of a
yet-incomplete original packet.

So far this kfunc is only callable from TC clsact progs.

Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
---
 include/net/ip.h           | 11 +++++
 net/ipv4/Makefile          |  1 +
 net/ipv4/ip_fragment.c     |  2 +
 net/ipv4/ip_fragment_bpf.c | 98 ++++++++++++++++++++++++++++++++++++++
 4 files changed, 112 insertions(+)
 create mode 100644 net/ipv4/ip_fragment_bpf.c

diff --git a/include/net/ip.h b/include/net/ip.h
index 144bdfbb25af..14f1e69a6523 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -679,6 +679,7 @@ enum ip_defrag_users {
 	IP_DEFRAG_VS_FWD,
 	IP_DEFRAG_AF_PACKET,
 	IP_DEFRAG_MACVLAN,
+	IP_DEFRAG_BPF,
 };
 
 /* Return true if the value of 'user' is between 'lower_bond'
@@ -692,6 +693,16 @@ static inline bool ip_defrag_user_in_between(u32 user,
 }
 
 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
+
+#ifdef CONFIG_DEBUG_INFO_BTF
+int register_ip_frag_bpf(void);
+#else
+static inline int register_ip_frag_bpf(void)
+{
+	return 0;
+}
+#endif
+
 #ifdef CONFIG_INET
 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
 #else
diff --git a/net/ipv4/Makefile b/net/ipv4/Makefile
index af7d2cf490fb..749da1599933 100644
--- a/net/ipv4/Makefile
+++ b/net/ipv4/Makefile
@@ -64,6 +64,7 @@ obj-$(CONFIG_TCP_CONG_ILLINOIS) += tcp_illinois.o
 obj-$(CONFIG_NET_SOCK_MSG) += tcp_bpf.o
 obj-$(CONFIG_BPF_SYSCALL) += udp_bpf.o
 obj-$(CONFIG_NETLABEL) += cipso_ipv4.o
+obj-$(CONFIG_DEBUG_INFO_BTF) += ip_fragment_bpf.o
 
 obj-$(CONFIG_XFRM) += xfrm4_policy.o xfrm4_state.o xfrm4_input.o \
 		      xfrm4_output.o xfrm4_protocol.o
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index 7406c6b6376d..467aa8ace9fb 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -757,5 +757,7 @@ void __init ipfrag_init(void)
 	if (inet_frags_init(&ip4_frags))
 		panic("IP: failed to allocate ip4_frags cache\n");
 	ip4_frags_ctl_register();
+	if (register_ip_frag_bpf())
+		panic("IP: bpf: failed to register ip_frag_bpf\n");
 	register_pernet_subsys(&ip4_frags_ops);
 }
diff --git a/net/ipv4/ip_fragment_bpf.c b/net/ipv4/ip_fragment_bpf.c
new file mode 100644
index 000000000000..a9e5908ed216
--- /dev/null
+++ b/net/ipv4/ip_fragment_bpf.c
@@ -0,0 +1,98 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Unstable ipv4 fragmentation 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 <linux/ip.h>
+#include <linux/filter.h>
+#include <linux/netdevice.h>
+#include <net/ip.h>
+#include <net/sock.h>
+
+__diag_push();
+__diag_ignore_all("-Wmissing-prototypes",
+		  "Global functions as their definitions will be in ip_fragment BTF");
+
+/* bpf_ip_check_defrag - Defragment an ipv4 packet
+ *
+ * This helper takes an skb as input. If this skb successfully reassembles
+ * the original packet, the skb is updated to contain the original, reassembled
+ * packet.
+ *
+ * Otherwise (on error or incomplete reassembly), the input skb remains
+ * unmodified.
+ *
+ * Parameters:
+ * @ctx		- Pointer to program context (skb)
+ * @netns	- Child network namespace id. If value is a negative signed
+ *		  32-bit integer, the netns of the device in the skb is used.
+ *
+ * Return:
+ * 0 on successfully reassembly or non-fragmented packet. Negative value on
+ * error or incomplete reassembly.
+ */
+int bpf_ip_check_defrag(struct __sk_buff *ctx, u64 netns)
+{
+	struct sk_buff *skb = (struct sk_buff *)ctx;
+	struct sk_buff *skb_cpy, *skb_out;
+	struct net *caller_net;
+	struct net *net;
+	int mac_len;
+	void *mac;
+
+	if (unlikely(!((s32)netns < 0 || netns <= S32_MAX)))
+		return -EINVAL;
+
+	caller_net = skb->dev ? dev_net(skb->dev) : sock_net(skb->sk);
+	if ((s32)netns < 0) {
+		net = caller_net;
+	} else {
+		net = get_net_ns_by_id(caller_net, netns);
+		if (unlikely(!net))
+			return -EINVAL;
+	}
+
+	mac_len = skb->mac_len;
+	skb_cpy = skb_copy(skb, GFP_ATOMIC);
+	if (!skb_cpy)
+		return -ENOMEM;
+
+	skb_out = ip_check_defrag(net, skb_cpy, IP_DEFRAG_BPF);
+	if (IS_ERR(skb_out))
+		return PTR_ERR(skb_out);
+
+	skb_morph(skb, skb_out);
+	kfree_skb(skb_out);
+
+	/* ip_check_defrag() does not maintain mac header, so push empty header
+	 * in so prog sees the correct layout. The empty mac header will be
+	 * later pulled from cls_bpf.
+	 */
+	mac = skb_push(skb, mac_len);
+	memset(mac, 0, mac_len);
+	bpf_compute_data_pointers(skb);
+
+	return 0;
+}
+
+__diag_pop()
+
+BTF_SET8_START(ip_frag_kfunc_set)
+BTF_ID_FLAGS(func, bpf_ip_check_defrag, KF_CHANGES_PKT)
+BTF_SET8_END(ip_frag_kfunc_set)
+
+static const struct btf_kfunc_id_set ip_frag_bpf_kfunc_set = {
+	.owner = THIS_MODULE,
+	.set   = &ip_frag_kfunc_set,
+};
+
+int register_ip_frag_bpf(void)
+{
+	return register_btf_kfunc_id_set(BPF_PROG_TYPE_SCHED_CLS,
+					 &ip_frag_bpf_kfunc_set);
+}
-- 
2.39.0