selftests/bpf: Properly enable hwtstamp in xdp_hw_metadata
authorStanislav Fomichev <sdf@google.com>
Thu, 26 Jan 2023 22:50:30 +0000 (14:50 -0800)
committerMartin KaFai Lau <martin.lau@kernel.org>
Fri, 27 Jan 2023 06:10:31 +0000 (22:10 -0800)
commita5f3a3f7c17273f825e96d1b3369164a39345a35
tree7816f3b6b05d5dde4286d1890f13cefc5775e0a8
parent0164776a2ce782a342b66b7980055e96efff746c
selftests/bpf: Properly enable hwtstamp in xdp_hw_metadata

The existing timestamping_enable() is a no-op because it applies
to the socket-related path that we are not verifying here
anymore. (but still leaving the code around hoping we can
have xdp->skb path verified here as well)

  poll: 1 (0)
  xsk_ring_cons__peek: 1
  0xf64788: rx_desc[0]->addr=100000000008000 addr=8100 comp_addr=8000
  rx_hash: 3697961069
  rx_timestamp:  1674657672142214773 (sec:1674657672.1422)
  XDP RX-time:   1674657709561774876 (sec:1674657709.5618) delta sec:37.4196
  AF_XDP time:   1674657709561871034 (sec:1674657709.5619) delta
sec:0.0001 (96.158 usec)
  0xf64788: complete idx=8 addr=8000

Also, maybe something to archive here, see [0] for Jesper's note
about NIC vs host clock delta.

0: https://lore.kernel.org/bpf/f3a116dc-1b14-3432-ad20-a36179ef0608@redhat.com/

v2:
- Restore original value (Martin)

Fixes: 297a3f124155 ("selftests/bpf: Simple program to dump XDP RX metadata")
Reported-by: Jesper Dangaard Brouer <jbrouer@redhat.com>
Tested-by: Jesper Dangaard Brouer <jbrouer@redhat.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/r/20230126225030.510629-1-sdf@google.com
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
tools/testing/selftests/bpf/xdp_hw_metadata.c