Magnus Karlsson says:

====================
Previously, the xsk code did not record which umem was bound to a
specific queue id. This was not required if all drivers were zero-copy
enabled as this had to be recorded in the driver anyway. So if a user
tried to bind two umems to the same queue, the driver would say
no. But if copy-mode was first enabled and then zero-copy mode (or the
reverse order), we mistakenly enabled both of them on the same umem
leading to buggy behavior. The main culprit for this is that we did
not store the association of umem to queue id in the copy case and
only relied on the driver reporting this. As this relation was not
stored in the driver for copy mode (it does not rely on the AF_XDP
NDOs), this obviously could not work.

This patch fixes the problem by always recording the umem to queue id
relationship in the netdev_queue and netdev_rx_queue structs. This way
we always know what kind of umem has been bound to a queue id and can
act appropriately at bind time. To make the bind semantics consistent
with ethtool queue manipulations and to facilitate the implementation
of drivers, we also forbid decreasing the number of queues/channels
with ethtool if there is an active AF_XDP socket in the set of queues
that are disabled.

Jakub, please take a look at your patches. The last one I had to
change slightly to make it fit with the new interface
xdp_get_umem_from_qid(). An added bonus with this function is that we,
in the future, can also use it from the driver to get a umem, thus
simplifying driver implementations (and later remove the umem from the
NDO completely). Björn will mail patches, at a later point in time,
using this in the i40e and ixgbe drivers, that removes a good chunk of
code from the ZC implementations. I also made your code aware of Tx
queues. If we create a socket that only has a Tx queue, then the queue
id will refer to a Tx queue id only and could be larger than the
available amount of Rx queues. Please take a look at it.

Differences against v1:
* Included patches from Jakub that forbids decreasing the number of active
  queues if a queue to be deactivated has an AF_XDP socket. These have
  been adapted somewhat to the new interfaces in patch 2.
* Removed redundant check against real_num_[rt]x_queue in xsk_bind
* Only need to test against real_num_[rt]x_queues in
  xdp_clear_umem_at_qid.

Patch 1: Introduces a umem reference in the netdev_rx_queue and
         netdev_queue structs.
Patch 2: Records which queue_id is bound to which umem and make sure
         that you cannot bind two different umems to the same queue_id.
Patch 3: Pre patch to ethtool_set_channels.
Patch 4: Forbid decreasing the number of active queues if a deactivated
         queue has an AF_XDP socket.
Patch 5: Simplify xdp_clear_umem_at_qid now when ethtool cannot deactivate
         the queue id we are running on.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

As we now do not allow ethtool to deactivate the queue id we are
running an AF_XDP socket on, we can simplify the implementation of
xdp_clear_umem_at_qid().
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

We already check the RSS indirection table does not use queues which
would be disabled by channel reconfiguration. Make sure user does not
try to disable queues which have a UMEM and zero-copy AF_XDP socket
installed.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

ethtool_set_channels() validates the config against driver's max
settings. It retrieves the current config and stores it in a
variable called max. This was okay when only max settings were
accessed but we will soon want to access current settings as
well, so calling the entire structure max makes the code less
readable.

While at it drop unnecessary parenthesis.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Previously, the xsk code did not record which umem was bound to a
specific queue id. This was not required if all drivers were zero-copy
enabled as this had to be recorded in the driver anyway. So if a user
tried to bind two umems to the same queue, the driver would say
no. But if copy-mode was first enabled and then zero-copy mode (or the
reverse order), we mistakenly enabled both of them on the same umem
leading to buggy behavior. The main culprit for this is that we did
not store the association of umem to queue id in the copy case and
only relied on the driver reporting this. As this relation was not
stored in the driver for copy mode (it does not rely on the AF_XDP
NDOs), this obviously could not work.

This patch fixes the problem by always recording the umem to queue id
relationship in the netdev_queue and netdev_rx_queue structs. This way
we always know what kind of umem has been bound to a queue id and can
act appropriately at bind time.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

These references to the umem will be used to store information
on what kind of AF_XDP umem that is bound to a queue id, if any.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Fix a simple typo: Completetion -> Completion
Signed-off-by: Konrad Djimeli <kdjimeli@igalia.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

There is a warning when compiling bpf sample programs in sample/bpf:

  make -C /home/foo/bpf/samples/bpf/../../tools/lib/bpf/ RM='rm -rf' LDFLAGS= srctree=/home/foo/bpf/samples/bpf/../../ O=
    HOSTCC  /home/foo/bpf/samples/bpf/tracex3_user.o
  /home/foo/bpf/samples/bpf/tracex3_user.c:20:0: warning: "ARRAY_SIZE" redefined
   #define ARRAY_SIZE(x) (sizeof(x) / sizeof(*(x)))

  In file included from /home/foo/bpf/samples/bpf/tracex3_user.c:18:0:
  ./tools/testing/selftests/bpf/bpf_util.h:48:0: note: this is the location of the previous definition
   # define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
Signed-off-by: Bo YU <tsu.yubo@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Andrey Ignatov says:

====================
This patch set renames a few interfaces in libbpf, mostly netlink related,
so that all symbols provided by the library have only three possible
prefixes:

% nm -D tools/lib/bpf/libbpf.so  | \
    awk '$2 == "T" {sub(/[_\(].*/, "", $3); if ($3) print $3}' | \
    sort | \
    uniq -c
     91 bpf
      8 btf
     14 libbpf

libbpf is used more and more outside kernel tree. That means the library
should follow good practices in library design and implementation to
play well with third party code that uses it.

One of such practices is to have a common prefix (or a few) for every
interface, function or data structure, library provides. It helps to
avoid name conflicts with other libraries and keeps API/ABI consistent.

Inconsistent names in libbpf already cause problems in real life. E.g.
an application can't use both libbpf and libnl due to conflicting
symbols (specifically nla_parse, nla_parse_nested and a few others).

Some of problematic global symbols are not part of ABI and can be
restricted from export with either visibility attribute/pragma or export
map (what is useful by itself and can be done in addition). That won't
solve the problem for those that are part of ABI though. Also export
restrictions would help only in DSO case. If third party application links
libbpf statically it won't help, and people do it (e.g. Facebook links
most of libraries statically, including libbpf).

libbpf already uses the following prefixes for its interfaces:
* bpf_ for bpf system call wrappers, program/map/elf-object
  abstractions and a few other things;
* btf_ for BTF related API;
* libbpf_ for everything else.

The patch adds libbpf_ prefix to interfaces that use none of mentioned
above prefixes and don't fit well into the first two categories.

Long term benefits of having common prefix should outweigh possible
inconvenience of changing API for those functions now.

Patches 2-4 add libbpf_ prefix to libbpf interfaces: separate patch per
header. Other patches are simple improvements in API.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Make bpf_program__load consistent with other interfaces: use __u32
instead of u32. That in turn fixes build of samples:

In file included from ./samples/bpf/trace_output_user.c:21:0:
./tools/lib/bpf/libbpf.h:132:9: error: unknown type name ‘u32’
         u32 kern_version);
         ^

Fixes: commit 29cd77f4 ("libbpf: Support loading individual progs")
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Rename include guards to have consistent names "__LIBBPF_<header_name>".
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

libbpf is used more and more outside kernel tree. That means the library
should follow good practices in library design and implementation to
play well with third party code that uses it.

One of such practices is to have a common prefix (or a few) for every
interface, function or data structure, library provides. I helps to
avoid name conflicts with other libraries and keeps API consistent.

Inconsistent names in libbpf already cause problems in real life. E.g.
an application can't use both libbpf and libnl due to conflicting
symbols.

Having common prefix will help to fix current and avoid future problems.

libbpf already uses the following prefixes for its interfaces:
* bpf_ for bpf system call wrappers, program/map/elf-object
  abstractions and a few other things;
* btf_ for BTF related API;
* libbpf_ for everything else.

The patch renames function in str_error.h to have libbpf_ prefix since it
misses one and doesn't fit well into the first two categories.
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

libbpf is used more and more outside kernel tree. That means the library
should follow good practices in library design and implementation to
play well with third party code that uses it.

One of such practices is to have a common prefix (or a few) for every
interface, function or data structure, library provides. I helps to
avoid name conflicts with other libraries and keeps API consistent.

Inconsistent names in libbpf already cause problems in real life. E.g.
an application can't use both libbpf and libnl due to conflicting
symbols.

Having common prefix will help to fix current and avoid future problems.

libbpf already uses the following prefixes for its interfaces:
* bpf_ for bpf system call wrappers, program/map/elf-object
  abstractions and a few other things;
* btf_ for BTF related API;
* libbpf_ for everything else.

The patch adds libbpf_ prefix to interfaces in nlattr.h that use none of
mentioned above prefixes and doesn't fit well into the first two
categories.

Since affected part of API is used in bpftool, the patch applies
corresponding change to bpftool as well. Having it in a separate patch
will cause a state of tree where bpftool is broken what may not be a
good idea.
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

libbpf is used more and more outside kernel tree. That means the library
should follow good practices in library design and implementation to
play well with third party code that uses it.

One of such practices is to have a common prefix (or a few) for every
interface, function or data structure, library provides. I helps to
avoid name conflicts with other libraries and keeps API consistent.

Inconsistent names in libbpf already cause problems in real life. E.g.
an application can't use both libbpf and libnl due to conflicting
symbols.

Having common prefix will help to fix current and avoid future problems.

libbpf already uses the following prefixes for its interfaces:
* bpf_ for bpf system call wrappers, program/map/elf-object
  abstractions and a few other things;
* btf_ for BTF related API;
* libbpf_ for everything else.

The patch adds libbpf_ prefix to functions and typedef in libbpf.h that
use none of mentioned above prefixes and doesn't fit well into the first
two categories.

Since affected part of API is used in bpftool, the patch applies
corresponding change to bpftool as well. Having it in a separate patch
will cause a state of tree where bpftool is broken what may not be a
good idea.
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This typedef is used only by implementation in netlink.c. Nothing uses
it in public API. Move it to netlink.c.
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Stephen Rothwell reports the following link failure with IPv6 as module:

  x86_64-linux-gnu-ld: net/core/filter.o: in function `sk_lookup':
  (.text+0x19219): undefined reference to `__udp6_lib_lookup'

Fix the build by only enabling the IPv6 socket lookup if IPv6 support is
compiled into the kernel.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Joe Stringer says:

====================
This series proposes a new helper for the BPF API which allows BPF programs to
perform lookups for sockets in a network namespace. This would allow programs
to determine early on in processing whether the stack is expecting to receive
the packet, and perform some action (eg drop, forward somewhere) based on this
information.

The series is structured roughly into:
* Misc refactor
* Add the socket pointer type
* Add reference tracking to ensure that socket references are freed
* Extend the BPF API to add sk_lookup_xxx() / sk_release() functions
* Add tests/documentation

The helper proposed in this series includes a parameter for a tuple which must
be filled in by the caller to determine the socket to look up. The simplest
case would be filling with the contents of the packet, ie mapping the packet's
5-tuple into the parameter. In common cases, it may alternatively be useful to
reverse the direction of the tuple and perform a lookup, to find the socket
that initiates this connection; and if the BPF program ever performs a form of
IP address translation, it may further be useful to be able to look up
arbitrary tuples that are not based upon the packet, but instead based on state
held in BPF maps or hardcoded in the BPF program.

Currently, access into the socket's fields are limited to those which are
otherwise already accessible, and are restricted to read-only access.

Changes since v3:
* New patch: "bpf: Reuse canonical string formatter for ctx errs"
* Add PTR_TO_SOCKET to is_ctx_reg().
* Add a few new checks to prevent mixing of socket/non-socket pointers.
* Swap order of checks in sock_filter_is_valid_access().
* Prefix register spill macros with "bpf_".
* Add acks from previous round
* Rebase

Changes since v2:
* New patch: "selftests/bpf: Generalize dummy program types".
  This enables adding verifier tests for socket lookup with tail calls.
* Define the semantics of the new helpers more clearly in uAPI header.
* Fix release of caller_net when netns is not specified.
* Use skb->sk to find caller net when skb->dev is unavailable.
* Fix build with !CONFIG_NET.
* Replace ptr_id defensive coding when releasing reference state with an
  internal error (-EFAULT).
* Remove flags argument to sk_release().
* Add several new assembly tests suggested by Daniel.
* Add a few new C tests.
* Fix typo in verifier error message.

Changes since v1:
* Limit netns_id field to 32 bits
* Reuse reg_type_mismatch() in more places
* Reduce the number of passes at convert_ctx_access()
* Replace ptr_id defensive coding when releasing reference state with an
  internal error (-EFAULT)
* Rework 'struct bpf_sock_tuple' to allow passing a packet pointer
* Allow direct packet access from helper
* Fix compile error with CONFIG_IPV6 enabled
* Improve commit messages

Changes since RFC:
* Split up sk_lookup() into sk_lookup_tcp(), sk_lookup_udp().
* Only take references on the socket when necessary.
  * Make sk_release() only free the socket reference in this case.
* Fix some runtime reference leaks:
  * Disallow BPF_LD_[ABS|IND] instructions while holding a reference.
  * Disallow bpf_tail_call() while holding a reference.
* Prevent the same instruction being used for reference and other
  pointer type.
* Simplify locating copies of a reference during helper calls by caching
  the pointer id from the caller.
* Fix kbuild compilation warnings with particular configs.
* Improve code comments describing the new verifier pieces.
* Tested by Nitin
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Document the new pointer types in the verifier and how the pointer ID
tracking works to ensure that references which are taken are later
released.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Add some tests that demonstrate and test the balanced lookup/free
nature of socket lookup. Section names that start with "fail" represent
programs that are expected to fail verification; all others should
succeed.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Allow the individual program load to be invoked. This will help with
testing, where a single ELF may contain several sections, some of which
denote subprograms that are expected to fail verification, along with
some which are expected to pass verification. By allowing programs to be
iterated and individually loaded, each program can be independently
checked against its expected verification result.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

reference tracking: leak potential reference
reference tracking: leak potential reference on stack
reference tracking: leak potential reference on stack 2
reference tracking: zero potential reference
reference tracking: copy and zero potential references
reference tracking: release reference without check
reference tracking: release reference
reference tracking: release reference twice
reference tracking: release reference twice inside branch
reference tracking: alloc, check, free in one subbranch
reference tracking: alloc, check, free in both subbranches
reference tracking in call: free reference in subprog
reference tracking in call: free reference in subprog and outside
reference tracking in call: alloc & leak reference in subprog
reference tracking in call: alloc in subprog, release outside
reference tracking in call: sk_ptr leak into caller stack
reference tracking in call: sk_ptr spill into caller stack
reference tracking: allow LD_ABS
reference tracking: forbid LD_ABS while holding reference
reference tracking: allow LD_IND
reference tracking: forbid LD_IND while holding reference
reference tracking: check reference or tail call
reference tracking: release reference then tail call
reference tracking: leak possible reference over tail call
reference tracking: leak checked reference over tail call
reference tracking: mangle and release sock_or_null
reference tracking: mangle and release sock
reference tracking: access member
reference tracking: write to member
reference tracking: invalid 64-bit access of member
reference tracking: access after release
reference tracking: direct access for lookup
unpriv: spill/fill of different pointers stx - ctx and sock
unpriv: spill/fill of different pointers stx - leak sock
unpriv: spill/fill of different pointers stx - sock and ctx (read)
unpriv: spill/fill of different pointers stx - sock and ctx (write)
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Don't hardcode the dummy program types to SOCKET_FILTER type, as this
prevents testing bpf_tail_call in conjunction with other program types.
Instead, use the program type specified in the test case.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This patch adds new BPF helper functions, bpf_sk_lookup_tcp() and
bpf_sk_lookup_udp() which allows BPF programs to find out if there is a
socket listening on this host, and returns a socket pointer which the
BPF program can then access to determine, for instance, whether to
forward or drop traffic. bpf_sk_lookup_xxx() may take a reference on the
socket, so when a BPF program makes use of this function, it must
subsequently pass the returned pointer into the newly added sk_release()
to return the reference.

By way of example, the following pseudocode would filter inbound
connections at XDP if there is no corresponding service listening for
the traffic:

  struct bpf_sock_tuple tuple;
  struct bpf_sock_ops *sk;

  populate_tuple(ctx, &tuple); // Extract the 5tuple from the packet
  sk = bpf_sk_lookup_tcp(ctx, &tuple, sizeof tuple, netns, 0);
  if (!sk) {
    // Couldn't find a socket listening for this traffic. Drop.
    return TC_ACT_SHOT;
  }
  bpf_sk_release(sk, 0);
  return TC_ACT_OK;
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Allow helper functions to acquire a reference and return it into a
register. Specific pointer types such as the PTR_TO_SOCKET will
implicitly represent such a reference. The verifier must ensure that
these references are released exactly once in each path through the
program.

To achieve this, this commit assigns an id to the pointer and tracks it
in the 'bpf_func_state', then when the function or program exits,
verifies that all of the acquired references have been freed. When the
pointer is passed to a function that frees the reference, it is removed
from the 'bpf_func_state` and all existing copies of the pointer in
registers are marked invalid.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

An upcoming commit will need very similar copy/realloc boilerplate, so
refactor the existing stack copy/realloc functions into macros to
simplify it.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Teach the verifier a little bit about a new type of pointer, a
PTR_TO_SOCKET. This pointer type is accessed from BPF through the
'struct bpf_sock' structure.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This check will be reused by an upcoming commit for conditional jump
checks for sockets. Refactor it a bit to simplify the later commit.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

The array "reg_type_str" provides canonical formatting of register
types, however a couple of places would previously check whether a
register represented the context and write the name "context" directly.
An upcoming commit will add another pointer type to these statements, so
to provide more accurate error messages in the verifier, update these
error messages to use "reg_type_str" instead.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

An upcoming commit will add another two pointer types that need very
similar behaviour, so generalise this function now.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Add this iterator for spilled registers, it concentrates the details of
how to get the current frame's spilled registers into a single macro
while clarifying the intention of the code which is calling the macro.
Signed-off-by: Joe Stringer <joe@wand.net.nz>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Jakub Kicinski says:

====================
This series makes the control message parsing for interacting
with BPF maps more flexible.  Up until now we had a hard limit
in the ABI for key and value size to be 64B at most.  Using
TLV capability allows us to support large map entries.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

In current ABI the size of the messages carrying map elements was
statically defined to at most 16 words of key and 16 words of value
(NFP word is 4 bytes).  We should not make this assumption and use
the max key and value sizes from the BPF capability instead.

To make sure old kernels don't get surprised with larger (or smaller)
messages bump the FW ABI version to 3 when key/value size is different
than 16 words.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Some apps may want to have higher MTU on the control vNIC/queue.
Allow them to set the requested MTU at init time.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Up until now we only had per-vNIC BPF ABI version capabilities,
which are slightly awkward to use because bulk of the resources
and configuration does not relate to any particular vNIC.  Add
a new capability for global ABI version and check the per-vNIC
version are equal to it.  Assume the ABI version 2 if no explicit
version capability is present.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Roman Gushchin says:

====================
This patchset implements per-cpu cgroup local storage and provides
an example how per-cpu and shared cgroup local storage can be used
for efficient accounting of network traffic.

v4->v3:
  1) incorporated Alexei's feedback

v3->v2:
  1) incorporated Song's feedback
  2) rebased on top of current bpf-next

v2->v1:
  1) added a selftest implementing network counters
  2) added a missing free() in cgroup local storage selftest
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This commit adds a bpf kselftest, which demonstrates how percpu
and shared cgroup local storage can be used for efficient lookup-free
network accounting.

Cgroup local storage provides generic memory area with a very efficient
lookup free access. To avoid expensive atomic operations for each
packet, per-cpu cgroup local storage is used. Each packet is initially
charged to a per-cpu counter, and only if the counter reaches certain
value (32 in this case), the charge is moved into the global atomic
counter. This allows to amortize atomic operations, keeping reasonable
accuracy.

The test also implements a naive network traffic throttling, mostly to
demonstrate the possibility of bpf cgroup--based network bandwidth
control.

Expected output:
  ./test_netcnt
  test_netcnt:PASS
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This commit extends the test_cgrp2_attach2 test to cover per-cpu
cgroup storage. Bpf program will use shared and per-cpu cgroup
storages simultaneously, so a better coverage of corresponding
core code will be achieved.

Expected output:
  $ ./test_cgrp2_attach2
  Attached DROP prog. This ping in cgroup /foo should fail...
  ping: sendmsg: Operation not permitted
  Attached DROP prog. This ping in cgroup /foo/bar should fail...
  ping: sendmsg: Operation not permitted
  Attached PASS prog. This ping in cgroup /foo/bar should pass...
  Detached PASS from /foo/bar while DROP is attached to /foo.
  This ping in cgroup /foo/bar should fail...
  ping: sendmsg: Operation not permitted
  Attached PASS from /foo/bar and detached DROP from /foo.
  This ping in cgroup /foo/bar should pass...
  ### override:PASS
  ### multi:PASS
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This test extends the cgroup storage test to use per-cpu flavor
of the cgroup storage as well.

The test initializes a per-cpu cgroup storage to some non-zero initial
value (1000), and then simple bumps a per-cpu counter each time
the shared counter is atomically incremented. Then it reads all
per-cpu areas from the userspace side, and checks that the sum
of values adds to the expected sum.

Expected output:
  $ ./test_cgroup_storage
  test_cgroup_storage:PASS
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This commits adds verifier tests covering per-cpu cgroup storage
functionality. There are 6 new tests, which are exactly the same
as for shared cgroup storage, but do use per-cpu cgroup storage
map.

Expected output:
  $ ./test_verifier
  #0/u add+sub+mul OK
  #0/p add+sub+mul OK
  ...
  #286/p invalid cgroup storage access 6 OK
  #287/p valid per-cpu cgroup storage access OK
  #288/p invalid per-cpu cgroup storage access 1 OK
  #289/p invalid per-cpu cgroup storage access 2 OK
  #290/p invalid per-cpu cgroup storage access 3 OK
  #291/p invalid per-cpu cgroup storage access 4 OK
  #292/p invalid per-cpu cgroup storage access 5 OK
  #293/p invalid per-cpu cgroup storage access 6 OK
  #294/p multiple registers share map_lookup_elem result OK
  ...
  #662/p mov64 src == dst OK
  #663/p mov64 src != dst OK
  Summary: 914 PASSED, 0 SKIPPED, 0 FAILED
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This commit adds support for BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
map type.
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Song Liu <songliubraving@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>