samples: bpf: Add BPF support for devmap_xmit tracepoint
authorKumar Kartikeya Dwivedi <memxor@gmail.com>
Sat, 21 Aug 2021 00:19:58 +0000 (05:49 +0530)
committerAlexei Starovoitov <ast@kernel.org>
Tue, 24 Aug 2021 21:48:41 +0000 (14:48 -0700)
This adds support for the devmap_xmit tracepoint, and its multi device
variant that can be used to obtain streams for each individual
net_device to net_device redirection. This is useful for decomposing
total xmit stats in xdp_monitor.

Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210821002010.845777-11-memxor@gmail.com
samples/bpf/xdp_sample.bpf.c

index f01a5529751c1fa347ab06657dda333b4694f230..0eb7e1dcae22295c8ea75aeed60fc02a8e9ef8c5 100644 (file)
@@ -11,6 +11,14 @@ array_map redir_err_cnt SEC(".maps");
 array_map cpumap_enqueue_cnt SEC(".maps");
 array_map cpumap_kthread_cnt SEC(".maps");
 array_map exception_cnt SEC(".maps");
+array_map devmap_xmit_cnt SEC(".maps");
+
+struct {
+       __uint(type, BPF_MAP_TYPE_PERCPU_HASH);
+       __uint(max_entries, 32 * 32);
+       __type(key, u64);
+       __type(value, struct datarec);
+} devmap_xmit_cnt_multi SEC(".maps");
 
 const volatile int nr_cpus = 0;
 
@@ -193,3 +201,66 @@ int BPF_PROG(tp_xdp_exception, const struct net_device *dev,
 
        return 0;
 }
+
+SEC("tp_btf/xdp_devmap_xmit")
+int BPF_PROG(tp_xdp_devmap_xmit, const struct net_device *from_dev,
+            const struct net_device *to_dev, int sent, int drops, int err)
+{
+       struct datarec *rec;
+       int idx_in, idx_out;
+       u32 cpu;
+
+       idx_in = from_dev->ifindex;
+       idx_out = to_dev->ifindex;
+
+       if (!IN_SET(from_match, idx_in))
+               return 0;
+       if (!IN_SET(to_match, idx_out))
+               return 0;
+
+       cpu = bpf_get_smp_processor_id();
+       rec = bpf_map_lookup_elem(&devmap_xmit_cnt, &cpu);
+       if (!rec)
+               return 0;
+       NO_TEAR_ADD(rec->processed, sent);
+       NO_TEAR_ADD(rec->dropped, drops);
+       /* Record bulk events, then userspace can calc average bulk size */
+       NO_TEAR_INC(rec->info);
+       /* Record error cases, where no frame were sent */
+       /* Catch API error of drv ndo_xdp_xmit sent more than count */
+       if (err || drops < 0)
+               NO_TEAR_INC(rec->issue);
+       return 0;
+}
+
+SEC("tp_btf/xdp_devmap_xmit")
+int BPF_PROG(tp_xdp_devmap_xmit_multi, const struct net_device *from_dev,
+            const struct net_device *to_dev, int sent, int drops, int err)
+{
+       struct datarec empty = {};
+       struct datarec *rec;
+       int idx_in, idx_out;
+       u64 idx;
+
+       idx_in = from_dev->ifindex;
+       idx_out = to_dev->ifindex;
+       idx = idx_in;
+       idx = idx << 32 | idx_out;
+
+       if (!IN_SET(from_match, idx_in))
+               return 0;
+       if (!IN_SET(to_match, idx_out))
+               return 0;
+
+       bpf_map_update_elem(&devmap_xmit_cnt_multi, &idx, &empty, BPF_NOEXIST);
+       rec = bpf_map_lookup_elem(&devmap_xmit_cnt_multi, &idx);
+       if (!rec)
+               return 0;
+
+       NO_TEAR_ADD(rec->processed, sent);
+       NO_TEAR_ADD(rec->dropped, drops);
+       NO_TEAR_INC(rec->info);
+       if (err || drops < 0)
+               NO_TEAR_INC(rec->issue);
+       return 0;
+}