[RFC,nf-next,0/9] netfilter: bpf base hook program generator

Message

Florian Westphal Oct. 14, 2021, 12:10 p.m. UTC

This series adds a bpf program generator for netfilter base hooks.

Currently netfilter hooks are invoked via nf_hook_slow, which walks
an array of function_addr:arg pairs:

for i in hooks[]; do
  verdict = hooks[i]->addr(hooks->[i].arg, skb, state);
  switch (verdict) { ....

The autogenerator unrolls this loop and builds a bpf program
that does:

state->priv = hooks->[0].hook_arg;
v = firstfunction(state);
if (v != ACCEPT) goto out;
state->priv = hooks->[1].hook_arg;
v = secondfunction(state); ...
if (v != ACCEPT) goto out;

... and so on.

Indirections are converted to direct calls. Invocation of the
autogenerated programs is done via bpf dispatcher from nf_hook().

As long as NF_QUEUE is not used, normal data path will not call
nf_hook_slow "interpreter" anymore.

Purpose of this is to eventually add a 'netfilter prog type' to bpf and
permit attachment of (userspace generated) bpf programs to the netfilter
machinery, e.g.  'attach bpf prog id 1234 to ipv6 PREROUTING at prio -300'.
The autogenerator would be adjusted so that these userspace-bpf programs
are invoked just like native c functions.

This will require to expose the context structure (program argument,
'struct __nf_hook_state *' and rewrite read-accesses to it to match internal
nf_hook_state layout plus new verifier checks on permitted return values
(e.g. a plain 'return NF_STOLEN' results in a skb leak).

Known problems:
- I did not convert all hooks to the new scheme, e.g. ILA won't compile ATM.
- checkpatch complains about a few indendation issues, line lengths etc.

Future work:
add support for NAT hooks, they still use indirect calls, but those
are less of a problem because these get called only once per connection.

Could annotate ops struct as to what kind of verdicts the
C function can return.  This would allow to elide retval
check when hook can only return NF_ACCEPT.

Could add extra support for INGRESS hook to move more code from
inline functions to the autogenerated program.

Initial tests show roughly 8% performance improvement in a netns-to-netns
UDP forward test with conntrack enabled in the 'forward' net namespace.

I'm looking for feedback on the chosen approach.

Thanks,
Florian

Florian Westphal (9):
  netfilter: nf_queue: carry index in hook state
  netfilter: nat: split nat hook iteration into a helper
  netfilter: remove hook index from nf_hook_slow arguments
  netfilter: make hook functions accept only one argument
  netfilter: reduce allowed hook count to 32
  netfilter: add bpf base hook program generator
  netfilter: core: do not rebuild bpf program on dying netns
  netfilter: ingress: switch to invocation via bpf
  netfilter: hook_jit: add prog cache

 drivers/net/ipvlan/ipvlan_l3s.c            |   4 +-
 include/linux/netfilter.h                  |  72 ++-
 include/linux/netfilter_ingress.h          |  17 +-
 include/net/netfilter/br_netfilter.h       |   7 +-
 include/net/netfilter/nf_flow_table.h      |   6 +-
 include/net/netfilter/nf_hook_bpf.h        |  14 +
 include/net/netfilter/nf_queue.h           |   3 +-
 include/net/netfilter/nf_synproxy.h        |   6 +-
 net/bridge/br_input.c                      |   3 +-
 net/bridge/br_netfilter_hooks.c            |  30 +-
 net/bridge/br_netfilter_ipv6.c             |   5 +-
 net/bridge/netfilter/ebtable_broute.c      |   8 +-
 net/bridge/netfilter/ebtable_filter.c      |   5 +-
 net/bridge/netfilter/ebtable_nat.c         |   5 +-
 net/bridge/netfilter/nf_conntrack_bridge.c |   8 +-
 net/ipv4/netfilter/arptable_filter.c       |   5 +-
 net/ipv4/netfilter/ipt_CLUSTERIP.c         |   6 +-
 net/ipv4/netfilter/iptable_filter.c        |   5 +-
 net/ipv4/netfilter/iptable_mangle.c        |   7 +-
 net/ipv4/netfilter/iptable_nat.c           |   6 +-
 net/ipv4/netfilter/iptable_raw.c           |   5 +-
 net/ipv4/netfilter/iptable_security.c      |   5 +-
 net/ipv4/netfilter/nf_defrag_ipv4.c        |   5 +-
 net/ipv6/netfilter/ip6table_filter.c       |   5 +-
 net/ipv6/netfilter/ip6table_mangle.c       |   6 +-
 net/ipv6/netfilter/ip6table_nat.c          |   6 +-
 net/ipv6/netfilter/ip6table_raw.c          |   5 +-
 net/ipv6/netfilter/ip6table_security.c     |   5 +-
 net/ipv6/netfilter/nf_defrag_ipv6_hooks.c  |   5 +-
 net/netfilter/Kconfig                      |  10 +
 net/netfilter/Makefile                     |   1 +
 net/netfilter/core.c                       | 103 +++-
 net/netfilter/ipvs/ip_vs_core.c            |  48 +-
 net/netfilter/nf_conntrack_proto.c         |  34 +-
 net/netfilter/nf_flow_table_inet.c         |   9 +-
 net/netfilter/nf_flow_table_ip.c           |  12 +-
 net/netfilter/nf_hook_bpf.c                | 569 +++++++++++++++++++++
 net/netfilter/nf_nat_core.c                |  50 +-
 net/netfilter/nf_nat_proto.c               |  56 +-
 net/netfilter/nf_queue.c                   |  12 +-
 net/netfilter/nf_synproxy_core.c           |   8 +-
 net/netfilter/nft_chain_filter.c           |  48 +-
 net/netfilter/nft_chain_nat.c              |   7 +-
 net/netfilter/nft_chain_route.c            |  22 +-
 security/selinux/hooks.c                   |  58 +--
 45 files changed, 1001 insertions(+), 315 deletions(-)
 create mode 100644 include/net/netfilter/nf_hook_bpf.h
 create mode 100644 net/netfilter/nf_hook_bpf.c