cgroup/bpf: fast path skb BPF filtering

Even though there is a static key protecting from overhead from
cgroup-bpf skb filtering when there is nothing attached, in many cases
it's not enough as registering a filter for one type will ruin the fast
path for all others. It's observed in production servers I've looked
at but also in laptops, where registration is done during init by
systemd or something else.

Add a per-socket fast path check guarding from such overhead. This
affects both receive and transmit paths of TCP, UDP and other
protocols. It showed ~1% tx/s improvement in small payload UDP
send benchmarks using a real NIC and in a server environment and the
number jumps to 2-3% for preemtible kernels.

Reviewed-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/r/d8c58857113185a764927a46f4b5a058d36d3ec3.1643292455.git.asml.silence@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

2 files changed, 33 insertions, 4 deletions

diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h
index b525d8cdc25b..88a51b242adc 100644
--- a/include/linux/bpf-cgroup.h
+++ b/include/linux/bpf-cgroup.h
@@ -8,6 +8,7 @@
 #include <linux/jump_label.h>
 #include <linux/percpu.h>
 #include <linux/rbtree.h>
+#include <net/sock.h>
 #include <uapi/linux/bpf.h>
 
 struct sock;
@@ -165,11 +166,23 @@ int bpf_percpu_cgroup_storage_copy(struct bpf_map *map, void *key, void *value);
 int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
 				     void *value, u64 flags);
 
+/* Opportunistic check to see whether we have any BPF program attached*/
+static inline bool cgroup_bpf_sock_enabled(struct sock *sk,
+					   enum cgroup_bpf_attach_type type)
+{
+	struct cgroup *cgrp = sock_cgroup_ptr(&sk->sk_cgrp_data);
+	struct bpf_prog_array *array;
+
+	array = rcu_access_pointer(cgrp->bpf.effective[type]);
+	return array != &bpf_empty_prog_array.hdr;
+}
+
 /* Wrappers for __cgroup_bpf_run_filter_skb() guarded by cgroup_bpf_enabled. */
 #define BPF_CGROUP_RUN_PROG_INET_INGRESS(sk, skb)			      \
 ({									      \
 	int __ret = 0;							      \
-	if (cgroup_bpf_enabled(CGROUP_INET_INGRESS))		      \
+	if (cgroup_bpf_enabled(CGROUP_INET_INGRESS) &&			      \
+	    cgroup_bpf_sock_enabled(sk, CGROUP_INET_INGRESS))		      \
 		__ret = __cgroup_bpf_run_filter_skb(sk, skb,		      \
 						    CGROUP_INET_INGRESS); \
 									      \
@@ -181,7 +194,8 @@ int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
 	int __ret = 0;							       \
 	if (cgroup_bpf_enabled(CGROUP_INET_EGRESS) && sk && sk == skb->sk) { \
 		typeof(sk) __sk = sk_to_full_sk(sk);			       \
-		if (sk_fullsock(__sk))					       \
+		if (sk_fullsock(__sk) &&				       \
+		    cgroup_bpf_sock_enabled(__sk, CGROUP_INET_EGRESS))	       \
 			__ret = __cgroup_bpf_run_filter_skb(__sk, skb,	       \
 						      CGROUP_INET_EGRESS); \
 	}								       \
@@ -347,7 +361,8 @@ int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
 				       kernel_optval)			       \
 ({									       \
 	int __ret = 0;							       \
-	if (cgroup_bpf_enabled(CGROUP_SETSOCKOPT))			       \
+	if (cgroup_bpf_enabled(CGROUP_SETSOCKOPT) &&			       \
+	    cgroup_bpf_sock_enabled(sock, CGROUP_SETSOCKOPT))		       \
 		__ret = __cgroup_bpf_run_filter_setsockopt(sock, level,	       \
 							   optname, optval,    \
 							   optlen,	       \
@@ -367,7 +382,8 @@ int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
 				       max_optlen, retval)		       \
 ({									       \
 	int __ret = retval;						       \
-	if (cgroup_bpf_enabled(CGROUP_GETSOCKOPT))			       \
+	if (cgroup_bpf_enabled(CGROUP_GETSOCKOPT) &&			       \
+	    cgroup_bpf_sock_enabled(sock, CGROUP_GETSOCKOPT))		       \
 		if (!(sock)->sk_prot->bpf_bypass_getsockopt ||		       \
 		    !INDIRECT_CALL_INET_1((sock)->sk_prot->bpf_bypass_getsockopt, \
 					tcp_bpf_bypass_getsockopt,	       \
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 2344f793c4dc..e3b82ce51445 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -1233,6 +1233,19 @@ struct bpf_prog_array {
 	struct bpf_prog_array_item items[];
 };
 
+struct bpf_empty_prog_array {
+	struct bpf_prog_array hdr;
+	struct bpf_prog *null_prog;
+};
+
+/* to avoid allocating empty bpf_prog_array for cgroups that
+ * don't have bpf program attached use one global 'bpf_empty_prog_array'
+ * It will not be modified the caller of bpf_prog_array_alloc()
+ * (since caller requested prog_cnt == 0)
+ * that pointer should be 'freed' by bpf_prog_array_free()
+ */
+extern struct bpf_empty_prog_array bpf_empty_prog_array;
+
 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
 void bpf_prog_array_free(struct bpf_prog_array *progs);
 int bpf_prog_array_length(struct bpf_prog_array *progs);