On Wed, Nov 30, 2016 at 10:16:50AM -0800, David Ahern wrote:
Add examples preventing a process in a cgroup from opening a socket
based family, protocol and type.
Signed-off-by: David Ahern <redacted>
...
quoted hunk ↗ jump to hunk
+++ b/samples/bpf/sock_flags_kern.c
@@ -0,0 +1,37 @@
+#include <uapi/linux/bpf.h>
+#include <linux/socket.h>
+#include "bpf_helpers.h"
+
+SEC("cgroup/sock1")
+int bpf_prog1(struct bpf_sock *sk)
+{
+ char fmt[] = "socket: family %d type %d protocol %d\n";
+
+ bpf_trace_printk(fmt, sizeof(fmt), sk->family, sk->type, sk->protocol);
+
+ /* block PF_INET6, SOCK_RAW, IPPROTO_ICMPV6 sockets
+ * ie., make ping6 fail
+ */
+ if (sk->family == PF_INET6 && sk->type == 3 && sk->protocol == 58)
+ return 0;
why not to use SOCK_RAW and IPPROTO_ICMPV6 instead of constants?
Thanks for the test!