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, 4 insertions, 42 deletions

diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
index 279ebbed75a5..098632fdbc45 100644
--- a/kernel/bpf/cgroup.c
+++ b/kernel/bpf/cgroup.c
@@ -1384,20 +1384,6 @@ int __cgroup_bpf_run_filter_sysctl(struct ctl_table_header *head,
 }
 
 #ifdef CONFIG_NET
-static bool __cgroup_bpf_prog_array_is_empty(struct cgroup *cgrp,
-					     enum cgroup_bpf_attach_type attach_type)
-{
-	struct bpf_prog_array *prog_array;
-	bool empty;
-
-	rcu_read_lock();
-	prog_array = rcu_dereference(cgrp->bpf.effective[attach_type]);
-	empty = bpf_prog_array_is_empty(prog_array);
-	rcu_read_unlock();
-
-	return empty;
-}
-
 static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen,
 			     struct bpf_sockopt_buf *buf)
 {
@@ -1456,19 +1442,11 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level,
 	};
 	int ret, max_optlen;
 
-	/* Opportunistic check to see whether we have any BPF program
-	 * attached to the hook so we don't waste time allocating
-	 * memory and locking the socket.
-	 */
-	if (__cgroup_bpf_prog_array_is_empty(cgrp, CGROUP_SETSOCKOPT))
-		return 0;
-
 	/* Allocate a bit more than the initial user buffer for
 	 * BPF program. The canonical use case is overriding
 	 * TCP_CONGESTION(nv) to TCP_CONGESTION(cubic).
 	 */
 	max_optlen = max_t(int, 16, *optlen);
-
 	max_optlen = sockopt_alloc_buf(&ctx, max_optlen, &buf);
 	if (max_optlen < 0)
 		return max_optlen;
@@ -1550,15 +1528,7 @@ int __cgroup_bpf_run_filter_getsockopt(struct sock *sk, int level,
 	};
 	int ret;
 
-	/* Opportunistic check to see whether we have any BPF program
-	 * attached to the hook so we don't waste time allocating
-	 * memory and locking the socket.
-	 */
-	if (__cgroup_bpf_prog_array_is_empty(cgrp, CGROUP_GETSOCKOPT))
-		return retval;
-
 	ctx.optlen = max_optlen;
-
 	max_optlen = sockopt_alloc_buf(&ctx, max_optlen, &buf);
 	if (max_optlen < 0)
 		return max_optlen;
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 0a1cfd8544b9..04a8d5bea552 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -1968,18 +1968,10 @@ static struct bpf_prog_dummy {
 	},
 };
 
-/* to avoid allocating empty bpf_prog_array for cgroups that
- * don't have bpf program attached use one global '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()
- */
-static struct {
-	struct bpf_prog_array hdr;
-	struct bpf_prog *null_prog;
-} empty_prog_array = {
+struct bpf_empty_prog_array bpf_empty_prog_array = {
 	.null_prog = NULL,
 };
+EXPORT_SYMBOL(bpf_empty_prog_array);
 
 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags)
 {
@@ -1989,12 +1981,12 @@ struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags)
 			       (prog_cnt + 1),
 			       flags);
 
-	return &empty_prog_array.hdr;
+	return &bpf_empty_prog_array.hdr;
 }
 
 void bpf_prog_array_free(struct bpf_prog_array *progs)
 {
-	if (!progs || progs == &empty_prog_array.hdr)
+	if (!progs || progs == &bpf_empty_prog_array.hdr)
 		return;
 	kfree_rcu(progs, rcu);
 }