bpf, net: Rework cookie generator as per-cpu one

With its use in BPF, the cookie generator can be called very frequently
in particular when used out of cgroup v2 hooks (e.g. connect / sendmsg)
and attached to the root cgroup, for example, when used in v1/v2 mixed
environments. In particular, when there's a high churn on sockets in the
system there can be many parallel requests to the bpf_get_socket_cookie()
and bpf_get_netns_cookie() helpers which then cause contention on the
atomic counter.

As similarly done in f991bd2e1421 ("fs: introduce a per-cpu last_ino
allocator"), add a small helper library that both can use for the 64 bit
counters. Given this can be called from different contexts, we also need
to deal with potential nested calls even though in practice they are
considered extremely rare. One idea as suggested by Eric Dumazet was
to use a reverse counter for this situation since we don't expect 64 bit
overflows anyways; that way, we can avoid bigger gaps in the 64 bit
counter space compared to just batch-wise increase. Even on machines
with small number of cores (e.g. 4) the cookie generation shrinks from
min/max/med/avg (ns) of 22/50/40/38.9 down to 10/35/14/17.3 when run
in parallel from multiple CPUs.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Link: https://lore.kernel.org/bpf/8a80b8d27d3c49f9a14e1d5213c19d8be87d1dc8.1601477936.git.daniel@iogearbox.net

3 files changed, 65 insertions, 2 deletions

diff --git a/include/linux/cookie.h b/include/linux/cookie.h
new file mode 100644
index 000000000000..0c159f585109
--- /dev/null
+++ b/include/linux/cookie.h
@@ -0,0 +1,51 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __LINUX_COOKIE_H
+#define __LINUX_COOKIE_H
+
+#include <linux/atomic.h>
+#include <linux/percpu.h>
+#include <asm/local.h>
+
+struct pcpu_gen_cookie {
+	local_t nesting;
+	u64 last;
+} __aligned(16);
+
+struct gen_cookie {
+	struct pcpu_gen_cookie __percpu *local;
+	atomic64_t forward_last ____cacheline_aligned_in_smp;
+	atomic64_t reverse_last;
+};
+
+#define COOKIE_LOCAL_BATCH	4096
+
+#define DEFINE_COOKIE(name)						\
+	static DEFINE_PER_CPU(struct pcpu_gen_cookie, __##name);	\
+	static struct gen_cookie name = {				\
+		.local		= &__##name,				\
+		.forward_last	= ATOMIC64_INIT(0),			\
+		.reverse_last	= ATOMIC64_INIT(0),			\
+	}
+
+static __always_inline u64 gen_cookie_next(struct gen_cookie *gc)
+{
+	struct pcpu_gen_cookie *local = this_cpu_ptr(gc->local);
+	u64 val;
+
+	if (likely(local_inc_return(&local->nesting) == 1)) {
+		val = local->last;
+		if (__is_defined(CONFIG_SMP) &&
+		    unlikely((val & (COOKIE_LOCAL_BATCH - 1)) == 0)) {
+			s64 next = atomic64_add_return(COOKIE_LOCAL_BATCH,
+						       &gc->forward_last);
+			val = next - COOKIE_LOCAL_BATCH;
+		}
+		local->last = ++val;
+	} else {
+		val = atomic64_dec_return(&gc->reverse_last);
+	}
+	local_dec(&local->nesting);
+	return val;
+}
+
+#endif /* __LINUX_COOKIE_H */
diff --git a/include/linux/sock_diag.h b/include/linux/sock_diag.h
index 15fe980a27ea..0b9ecd8cf979 100644
--- a/include/linux/sock_diag.h
+++ b/include/linux/sock_diag.h
@@ -25,7 +25,19 @@ void sock_diag_unregister(const struct sock_diag_handler *h);
 void sock_diag_register_inet_compat(int (*fn)(struct sk_buff *skb, struct nlmsghdr *nlh));
 void sock_diag_unregister_inet_compat(int (*fn)(struct sk_buff *skb, struct nlmsghdr *nlh));
 
-u64 sock_gen_cookie(struct sock *sk);
+u64 __sock_gen_cookie(struct sock *sk);
+
+static inline u64 sock_gen_cookie(struct sock *sk)
+{
+	u64 cookie;
+
+	preempt_disable();
+	cookie = __sock_gen_cookie(sk);
+	preempt_enable();
+
+	return cookie;
+}
+
 int sock_diag_check_cookie(struct sock *sk, const __u32 *cookie);
 void sock_diag_save_cookie(struct sock *sk, __u32 *cookie);
 
diff --git a/include/net/net_namespace.h b/include/net/net_namespace.h
index 2ee5901bec7a..22bc07f4b043 100644
--- a/include/net/net_namespace.h
+++ b/include/net/net_namespace.h
@@ -230,7 +230,7 @@ extern struct list_head net_namespace_list;
 struct net *get_net_ns_by_pid(pid_t pid);
 struct net *get_net_ns_by_fd(int fd);
 
-u64 net_gen_cookie(struct net *net);
+u64 __net_gen_cookie(struct net *net);
 
 #ifdef CONFIG_SYSCTL
 void ipx_register_sysctl(void);