This patch is based on a patch made by John Fastabend.
It adds support for offloading cls_flower.
when NETIF_F_HW_TC is on:
  flags = 0       => Rule will be processed twice - by hardware, and if
                     still relevant, by software.
  flags = SKIP_HW => Rull will be processed by software only

If hardware fail/not capabale to apply the rule, operation will NOT
fail. Filter will be processed by SW only.

Acked-by: Jiri Pirko <jiri@mellanox.com>
Suggested-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Amir Vadai <amir@vadai.me>
Signed-off-by: David S. Miller <davem@davemloft.net>

In the initial implementation the only way to stop a rule from being
inserted into the hardware table was via the device feature flag.
However this doesn't work well when working on an end host system
where packets are expect to hit both the hardware and software
datapaths.

For example we can imagine a rule that will match an IP address and
increment a field. If we install this rule in both hardware and
software we may increment the field twice. To date we have only
added support for the drop action so we have been able to ignore
these cases. But as we extend the action support we will hit this
example plus more such cases. Arguably these are not even corner
cases in many working systems these cases will be common.

To avoid forcing the driver to always abort (i.e. the above example)
this patch adds a flag to add a rule in software only. A careful
user can use this flag to build software and hardware datapaths
that work together. One example we have found particularly useful
is to use hardware resources to set the skb->mark on the skb when
the match may be expensive to run in software but a mark lookup
in a hash table is cheap. The idea here is hardware can do in one
lookup what the u32 classifier may need to traverse multiple lists
and hash tables to compute. The flag is only passed down on inserts.
On deletion to avoid stale references in hardware we always try
to remove a rule if it exists.

The flags field is part of the classifier specific options. Although
it is tempting to lift this into the generic structure doing this
proves difficult do to how the tc netlink attributes are implemented
along with how the dump/change routines are called. There is also
precedence for putting seemingly generic pieces in the specific
classifier options such as TCA_U32_POLICE, TCA_U32_ACT, etc. So
although not ideal I've left FLAGS in the u32 options as well as it
simplifies the code greatly and user space has already learned how
to manage these bits ala 'tc' tool.

Another thing if trying to update a rule we require the flags to
be unchanged. This is to force user space, software u32 and
the hardware u32 to keep in sync. Thanks to Simon Horman for
catching this case.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Existing bpf_clone_redirect() helper clones skb before redirecting
it to RX or TX of destination netdev.
Introduce bpf_redirect() helper that does that without cloning.

Benchmarked with two hosts using 10G ixgbe NICs.
One host is doing line rate pktgen.
Another host is configured as:
$ tc qdisc add dev $dev ingress
$ tc filter add dev $dev root pref 10 u32 match u32 0 0 flowid 1:2 \
   action bpf run object-file tcbpf1_kern.o section clone_redirect_xmit drop
so it receives the packet on $dev and immediately xmits it on $dev + 1
The section 'clone_redirect_xmit' in tcbpf1_kern.o file has the program
that does bpf_clone_redirect() and performance is 2.0 Mpps

$ tc filter add dev $dev root pref 10 u32 match u32 0 0 flowid 1:2 \
   action bpf run object-file tcbpf1_kern.o section redirect_xmit drop
which is using bpf_redirect() - 2.4 Mpps

and using cls_bpf with integrated actions as:
$ tc filter add dev $dev root pref 10 \
  bpf run object-file tcbpf1_kern.o section redirect_xmit integ_act classid 1
performance is 2.5 Mpps

To summarize:
u32+act_bpf using clone_redirect - 2.0 Mpps
u32+act_bpf using redirect - 2.4 Mpps
cls_bpf using redirect - 2.5 Mpps

For comparison linux bridge in this setup is doing 2.1 Mpps
and ixgbe rx + drop in ip_rcv - 7.8 Mpps

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Often cls_bpf classifier is used with single action drop attached.
Optimize this use case and let cls_bpf return both classid and action.
For backwards compatibility reasons enable this feature under
TCA_BPF_FLAG_ACT_DIRECT flag.

Then more interesting programs like the following are easier to write:
int cls_bpf_prog(struct __sk_buff *skb)
{
  /* classify arp, ip, ipv6 into different traffic classes
   * and drop all other packets
   */
  switch (skb->protocol) {
  case htons(ETH_P_ARP):
    skb->tc_classid = 1;
    break;
  case htons(ETH_P_IP):
    skb->tc_classid = 2;
    break;
  case htons(ETH_P_IPV6):
    skb->tc_classid = 3;
    break;
  default:
    return TC_ACT_SHOT;
  }

  return TC_ACT_OK;
}

Joint work with Daniel Borkmann.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jamal points out that this header also contains kernel internal magic that
cannot be used from userspace for anything meaningful.

Lets remove what the kernel doesn't use anymore and wrap remainder with
__KERNEL__.

Suggested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Suggested-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch introduces a flow-based filter. So far, the very essential
packet fields are supported.

This patch is only the first step. There is a lot of potential performance
improvements possible to implement. Also a lot of features are missing
now. They will be addressed in follow-up patches.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Seems all we want here is to avoid endless 'goto reclassify' loop.
tc_classify_compat even resets this counter when something other
than TC_ACT_RECLASSIFY is returned, so this skb-counter doesn't
break hypothetical loops induced by something other than perpetual
TC_ACT_RECLASSIFY return values.

skb_act_clone is now identical to skb_clone, so just use that.

Tested with following (bogus) filter:
tc filter add dev eth0 parent ffff: \
 protocol ip u32 match u32 0 0 police rate 10Kbit burst \
 64000 mtu 1500 action reclassify

Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

improves ingress+u32 performance from 22.4 Mpps to 22.9 Mpps

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Florian Westphal <fw@strlen.de>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Not used.

pedit sets TC_MUNGED when packet content was altered, but all the core
does is unset MUNGED again and then set OK2MUNGE.

And the latter isn't tested anywhere. So lets remove both
TC_MUNGED and TC_OK2MUNGE.

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This work extends the "classic" BPF programmable tc classifier by
extending its scope also to native eBPF code!

This allows for user space to implement own custom, 'safe' C like
classifiers (or whatever other frontend language LLVM et al may
provide in future), that can then be compiled with the LLVM eBPF
backend to an eBPF elf file. The result of this can be loaded into
the kernel via iproute2's tc. In the kernel, they can be JITed on
major archs and thus run in native performance.

Simple, minimal toy example to demonstrate the workflow:

  #include <linux/ip.h>
  #include <linux/if_ether.h>
  #include <linux/bpf.h>

  #include "tc_bpf_api.h"

  __section("classify")
  int cls_main(struct sk_buff *skb)
  {
    return (0x800 << 16) | load_byte(skb, ETH_HLEN + __builtin_offsetof(struct iphdr, tos));
  }

  char __license[] __section("license") = "GPL";

The classifier can then be compiled into eBPF opcodes and loaded
via tc, for example:

  clang -O2 -emit-llvm -c cls.c -o - | llc -march=bpf -filetype=obj -o cls.o
  tc filter add dev em1 parent 1: bpf cls.o [...]

As it has been demonstrated, the scope can even reach up to a fully
fledged flow dissector (similarly as in samples/bpf/sockex2_kern.c).

For tc, maps are allowed to be used, but from kernel context only,
in other words, eBPF code can keep state across filter invocations.
In future, we perhaps may reattach from a different application to
those maps e.g., to read out collected statistics/state.

Similarly as in socket filters, we may extend functionality for eBPF
classifiers over time depending on the use cases. For that purpose,
cls_bpf programs are using BPF_PROG_TYPE_SCHED_CLS program type, so
we can allow additional functions/accessors (e.g. an ABI compatible
offset translation to skb fields/metadata). For an initial cls_bpf
support, we allow the same set of helper functions as eBPF socket
filters, but we could diverge at some point in time w/o problem.

I was wondering whether cls_bpf and act_bpf could share C programs,
I can imagine that at some point, we introduce i) further common
handlers for both (or even beyond their scope), and/or if truly needed
ii) some restricted function space for each of them. Both can be
abstracted easily through struct bpf_verifier_ops in future.

The context of cls_bpf versus act_bpf is slightly different though:
a cls_bpf program will return a specific classid whereas act_bpf a
drop/non-drop return code, latter may also in future mangle skbs.
That said, we can surely have a "classify" and "action" section in
a single object file, or considered mentioned constraint add a
possibility of a shared section.

The workflow for getting native eBPF running from tc [1] is as
follows: for f_bpf, I've added a slightly modified ELF parser code
from Alexei's kernel sample, which reads out the LLVM compiled
object, sets up maps (and dynamically fixes up map fds) if any, and
loads the eBPF instructions all centrally through the bpf syscall.

The resulting fd from the loaded program itself is being passed down
to cls_bpf, which looks up struct bpf_prog from the fd store, and
holds reference, so that it stays available also after tc program
lifetime. On tc filter destruction, it will then drop its reference.

Moreover, I've also added the optional possibility to annotate an
eBPF filter with a name (e.g. path to object file, or something
else if preferred) so that when tc dumps currently installed filters,
some more context can be given to an admin for a given instance (as
opposed to just the file descriptor number).

Last but not least, bpf_prog_get() and bpf_prog_put() needed to be
exported, so that eBPF can be used from cls_bpf built as a module.
Thanks to 60a3b2253c41 ("net: bpf: make eBPF interpreter images
read-only") I think this is of no concern since anything wanting to
alter eBPF opcode after verification stage would crash the kernel.

  [1] http://git.breakpoint.cc/cgit/dborkman/iproute2.git/log/?h=ebpf

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>