nfp: bpf: move to new BPF program offload infrastructure
authorJakub Kicinski <jakub.kicinski@netronome.com>
Fri, 3 Nov 2017 20:56:29 +0000 (13:56 -0700)
committerDavid S. Miller <davem@davemloft.net>
Sun, 5 Nov 2017 13:26:19 +0000 (22:26 +0900)
commitc6c580d7bc390f864488c66153a487057e76d9d8
treef7e849ee2bb639f764bdb448c3e2b1704d3ccf7a
parent9314c442d7ddf749d29c09ab48ffa5333d2bf48e
nfp: bpf: move to new BPF program offload infrastructure

Following steps are taken in the driver to offload an XDP program:

XDP_SETUP_PROG:
 * prepare:
   - allocate program state;
   - run verifier (bpf_analyzer());
   - run translation;
 * load:
   - stop old program if needed;
   - load program;
   - enable BPF if not enabled;
 * clean up:
   - free program image.

With new infrastructure the flow will look like this:

BPF_OFFLOAD_VERIFIER_PREP:
  - allocate program state;
BPF_OFFLOAD_TRANSLATE:
   - run translation;
XDP_SETUP_PROG:
   - stop old program if needed;
   - load program;
   - enable BPF if not enabled;
BPF_OFFLOAD_DESTROY:
   - free program image.

Take advantage of the new infrastructure.  Allocation of driver
metadata has to be moved from jit.c to offload.c since it's now
done at a different stage.  Since there is no separate driver
private data for verification step, move temporary nfp_meta
pointer into nfp_prog.  We will now use user space context
offsets.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/netronome/nfp/bpf/jit.c
drivers/net/ethernet/netronome/nfp/bpf/main.c
drivers/net/ethernet/netronome/nfp/bpf/main.h
drivers/net/ethernet/netronome/nfp/bpf/offload.c
drivers/net/ethernet/netronome/nfp/bpf/verifier.c
drivers/net/ethernet/netronome/nfp/nfp_app.h
drivers/net/ethernet/netronome/nfp/nfp_net_common.c