Adds support for calling sock_ops BPF program when there is a
retransmission. Three arguments are used; one for the sequence number,
another for the number of segments retransmitted, and the last one for
the return value of tcp_transmit_skb (0 => success).
Does not include syn-ack retransmissions.

New op: BPF_SOCK_OPS_RETRANS_CB.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Adds direct write access to sk_txhash and access to tclass for ipv6
flows through getsockopt and setsockopt. Sample usage for tclass:

  bpf_getsockopt(skops, SOL_IPV6, IPV6_TCLASS, &v, sizeof(v))

where skops is a pointer to the ctx (struct bpf_sock_ops).
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Add support for reading many more tcp_sock fields

  state,	same as sk->sk_state
  rtt_min	same as sk->rtt_min.s[0].v (current rtt_min)
  snd_ssthresh
  rcv_nxt
  snd_nxt
  snd_una
  mss_cache
  ecn_flags
  rate_delivered
  rate_interval_us
  packets_out
  retrans_out
  total_retrans
  segs_in
  data_segs_in
  segs_out
  data_segs_out
  lost_out
  sacked_out
  sk_txhash
  bytes_received (__u64)
  bytes_acked    (__u64)
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Adds an optional call to sock_ops BPF program based on whether the
BPF_SOCK_OPS_RTO_CB_FLAG is set in bpf_sock_ops_flags.
The BPF program is passed 2 arguments: icsk_retransmits and whether the
RTO has expired.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Adds field bpf_sock_ops_cb_flags to tcp_sock and bpf_sock_ops. Its primary
use is to determine if there should be calls to sock_ops bpf program at
various points in the TCP code. The field is initialized to zero,
disabling the calls. A sock_ops BPF program can set it, per connection and
as necessary, when the connection is established.

It also adds support for reading and writting the field within a
sock_ops BPF program. Reading is done by accessing the field directly.
However, writing is done through the helper function
bpf_sock_ops_cb_flags_set, in order to return an error if a BPF program
is trying to set a callback that is not supported in the current kernel
(i.e. running an older kernel). The helper function returns 0 if it was
able to set all of the bits set in the argument, a positive number
containing the bits that could not be set, or -EINVAL if the socket is
not a full TCP socket.

Examples of where one could call the bpf program:

1) When RTO fires
2) When a packet is retransmitted
3) When the connection terminates
4) When a packet is sent
5) When a packet is received
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Adds support for passing up to 4 arguments to sock_ops bpf functions. It
reusues the reply union, so the bpf_sock_ops structures are not
increased in size.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

This patch adds a macro, SOCK_OPS_SET_FIELD, for writing to
struct tcp_sock or struct sock fields. This required adding a new
field "temp" to struct bpf_sock_ops_kern for temporary storage that
is used by sock_ops_convert_ctx_access. It is used to store and recover
the contents of a register, so the register can be used to store the
address of the sk. Since we cannot overwrite the dst_reg because it
contains the pointer to ctx, nor the src_reg since it contains the value
we want to store, we need an extra register to contain the address
of the sk.

Also adds the macro SOCK_OPS_GET_OR_SET_FIELD that calls one of the
GET or SET macros depending on the value of the TYPE field.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Changed SOCK_OPS_GET_TCP to SOCK_OPS_GET_FIELD and added 2
arguments so now it can also work with struct sock fields.
The first argument is the name of the field in the bpf_sock_ops
struct, the 2nd argument is the name of the field in the OBJ struct.

Previous: SOCK_OPS_GET_TCP(FIELD_NAME)
New:      SOCK_OPS_GET_FIELD(BPF_FIELD, OBJ_FIELD, OBJ)

Where OBJ is either "struct tcp_sock" or "struct sock" (without
quotation). BPF_FIELD is the name of the field in the bpf_sock_ops
struct and OBJ_FIELD is the name of the field in the OBJ struct.

Although the field names are currently the same, the kernel struct names
could change in the future and this change makes it easier to support
that.

Note that adding access to tcp_sock fields in sock_ops programs does
not preclude the tcp_sock fields from being removed as long as we are
willing to do one of the following:

  1) Return a fixed value (e.x. 0 or 0xffffffff), or
  2) Make the verifier fail if that field is accessed (i.e. program
    fails to load) so the user will know that field is no longer
    supported.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Make SOCK_OPS_GET_TCP helper macro size independent (before only worked
with 4-byte fields.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Currently, a sock_ops BPF program can write the op field and all the
reply fields (reply and replylong). This is a bug. The op field should
not have been writeable and there is currently no way to use replylong
field for indices >= 1. This patch enforces that only the reply field
(which equals replylong[0]) is writeable.

Fixes: 40304b2a ("bpf: BPF support for sock_ops")
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

If we then OR this with 0x40, then the value of 6th bit (0th is first bit)
become known, so the right mask is 0xbf instead of 0xcf.
Signed-off-by: Wang YanQing <udknight@gmail.com>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

John Fastabend says:

====================
The sockmap sample is pretty simple at the moment. All it does is open
a few sockets attach BPF programs/sockmaps and sends a few packets.

However, for testing and debugging I wanted to have more control over
the sendmsg format and data than provided by tools like iperf3/netperf,
etc. The reason is for testing BPF programs and stream parser it is
helpful to be able submit multiple sendmsg calls with different msg
layouts. For example lots of 1B iovs or a single large MB of data, etc.

Additionally, my current test setup requires an entire orchestration
layer (cilium) to run. As well as lighttpd and http traffic generators
or for kafka testing brokers and clients. This makes it a bit more
difficult when doing performance optimizations to incrementally test
small changes and come up with performance delta's and perf numbers.

By adding a few more options and an additional few tests the sockmap
sample program can show a more complete example and do some of the
above. Because the sample program is self contained it doesn't require
additional infrastructure to run either.

This series, although still fairly crude, does provide some nice
additions. They are

  - a new sendmsg tests with a sender and recv threads
  - a new base tests so we can get metrics/data without BPF
  - multiple GBps of throughput on base and sendmsg tests
  - automatically set rlimit and common variables

That said the UI is still primitive, more features could be added,
more tests might be useful, the reporting is bare bones, etc. But,
IMO lets push this now rather than sit on it for weeks until I get
time to do the above improvements. Additional patches can address
the other limitations/issues. Another thing I am considering is
moving this into selftests, after a few more fixes so we avoid
false failures, so that we get more sockmap testing.

v2: removed bogus file added by patch 3/7
v3: 1/7 replace goto out with returns, remove sighandler update,
    2/7 free iov in error cases
    3/7 fix bogus makefile change, bail out early on errors
v4: add Martin's "nits" and ACKs along with fixes to 2/7 iov free
    also pointed out by Martin.

Thanks Daniel and Martin for the reviews!
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Avoid extra step of setting limit from cmdline and do it directly in
the program.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Put client sockets in blocking mode otherwise with sendmsg tests
its easy to overrun the socket buffers which results in the test
being aborted.

The original non-blocking was added to handle listen/accept with
a single thread the client/accepted sockets do not need to be
non-blocking.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Add a base test that does not use BPF hooks to test baseline case.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Report bytes/sec sent as well as total bytes. Useful to get rough
idea how different configurations and usage patterns perform with
sockmap.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Currently for SENDMSG tests first send completes then recv runs. This
does not work well for large data sizes and/or many iterations. So
fork the recv and send handler so that we run both send and recv. In
the future we can add a parameter to do more than a single fork of
tx/rx.

With this we can get many GBps of data which helps exercise the
sockmap code.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

When testing BPF programs using sockmap I often want to have more
control over how sendmsg is exercised. This becomes even more useful
as new sockmap program types are added.

This adds a test type option to select type of test to run. Currently,
only "ping" and "sendmsg" are supported, but more can be added as
needed.

The new help argument gives the following,

 Usage: ./sockmap --cgroup <cgroup_path>
 options:
 --help         -h
 --cgroup       -c
 --rate         -r
 --verbose      -v
 --iov_count    -i
 --length       -l
 --test         -t
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

sockmap sample program takes arguments from cmd line but it reads them
in using offsets into the array. Because we want to add more arguments
in the future lets do proper argument handling.

Also refactor code to pull apart sock init and ping/pong test. This
allows us to add new tests in the future.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

mostly revert the previous workaround and make
'dubious pointer arithmetic' test useful again.
Use (ptr - ptr) << const instead of ptr << const to generate large scalar.
The rest stays as before commit 2b36047e.

Fixes: 2b36047e ("selftests/bpf: fix test_align")
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Bug: BPF programs and maps related to sockmaps test exist
in memory even after test_maps ends.

This patch fixes it as a short term workaround (sockmap
kernel side needs real fixing) by empyting sockmaps when
test ends.

Fixes: 6f6d33f3 ("bpf: selftests add sockmap tests")
Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
[ daniel: Note on workaround. ]
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

The test incorrectly doing
mkdir /mnt/cgroup-test-work-dirtest-bpf-based-device-cgroup
instead of
mkdir /mnt/cgroup-test-work-dir/test-bpf-based-device-cgroup

somehow such mkdir succeeds and new directory appears:
/mnt/cgroup-test-work-dir/cgroup-test-work-dirtest-bpf-based-device-cgroup

Later cleanup via nftw("/mnt/cgroup-test-work-dir", ...);
doesn't walk this directory.
"rmdir /mnt/cgroup-test-work-dir" succeeds, but bpf program and
dangling cgroup stays in memory.
That's a separate issue on a cgroup side.
For now fix the test.

Fixes: 37f1ba09 ("selftests/bpf: add a test for device cgroup controller")
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

test_hashmap_walk takes very long time on debug kernel with kasan on.
Reduce the number of iterations in this test without sacrificing
test coverage.
Also add printfs as progress indicator.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Commit 111e6b45 ("selftests/bpf: make test_verifier run most programs")
enables tools/testing/selftests/bpf/test_verifier unit cases to run
via bpf_prog_test_run command. With the latest code base,
test_verifier had one test case failure:

  ...
  #473/p check deducing bounds from const, 2 FAIL retval 1 != 0
  0: (b7) r0 = 1
  1: (75) if r0 s>= 0x1 goto pc+1
   R0=inv1 R1=ctx(id=0,off=0,imm=0) R10=fp0,call_-1
  2: (95) exit

  from 1 to 3: R0=inv1 R1=ctx(id=0,off=0,imm=0) R10=fp0,call_-1
  3: (d5) if r0 s<= 0x1 goto pc+1
   R0=inv1 R1=ctx(id=0,off=0,imm=0) R10=fp0,call_-1
  4: (95) exit

  from 3 to 5: R0=inv1 R1=ctx(id=0,off=0,imm=0) R10=fp0,call_-1
  5: (1f) r1 -= r0
  6: (95) exit
  processed 7 insns (limit 131072), stack depth 0
  ...

The test case does not set return value in the test
structure and hence the return value from the prog run
is assumed to be 0. However, the actual return value is 1.
As a result, the test failed. The fix is to correctly set
the return value in the test structure.

Fixes: 111e6b45 ("selftests/bpf: make test_verifier run most programs")
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

In commit b471f2f1 ("bpf: implement MAP_GET_NEXT_KEY command for LPM_TRIE map"),
the implemented MAP_GET_NEXT_KEY callback function is guarded with rcu read lock.
In the function body, "kmalloc(size, GFP_USER | __GFP_NOWARN)" is used which may
sleep and violate rcu read lock region requirements. This patch fixed the issue
by using GFP_ATOMIC instead to avoid blocking kmalloc. Tested with
CONFIG_DEBUG_ATOMIC_SLEEP=y as suggested by Eric Dumazet.

Fixes: b471f2f1 ("bpf: implement MAP_GET_NEXT_KEY command for LPM_TRIE map")
Signed-off-by: Yonghong Song <yhs@fb.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Fixes the following sparse warning:

drivers/net/ethernet/aquantia/atlantic/aq_pci_func.c:50:34: warning:
 symbol 'hw_atl_boards' was not declared. Should it be static?

Fixes: 4948293f ("net: aquantia: Introduce new AQC devices and capabilities")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix to return error code -ENOMEM from the aq_ndev_alloc() error
handling case instead of 0, as done elsewhere in this function.

Fixes: 23ee07ad ("net: aquantia: Cleanup pci functions module")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix to return error code -EINVAL instead of 0 when num_vfs above
limit_vfs, as done elsewhere in this function.

Fixes: 0dc78621 ("nfp: handle SR-IOV already enabled when driver is probing")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix bug that causes _absolute_ rtsym sizes of > 8 bytes (as per symbol
table) to result in incorrect space used during a TLV-based debug dump.

Detail: The size calculation stage calculates the correct size (size of
the rtsym address field == 8), while the dump uses the size in the table
to calculate the TLV size to reserve. Symbols with size <= 8 are handled
OK due to aligning sizes to 8, but including any absolute symbol with
listed size > 8 leads to an ENOSPC error during the dump.

Fixes: da762863 ("nfp: fix absolute rtsym handling in debug dump")
Signed-off-by: Carl Heymann <carl.heymann@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Similarly to tx ring, xdp_rxq_info is only registered
when !tfile->detached, so we need to avoid calling
xdp_rxq_info_unreg() twice too. The helper tun_cleanup_tx_ring()
already checks for this properly, so it is correct to put
xdp_rxq_info_unreg() just inside there.

Reported-by: syzbot+1c788d7ce0f0888f1d7f@syzkaller.appspotmail.com
Fixes: 8565d26b ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net")
Cc: Jason Wang <jasowang@redhat.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jakub Kicinski says:

====================
net: sched: add extack support for cls offloads

I've dropped the tests from the series because test_offloads.py changes
will conflict with bpf-next patches.  I will send four more patches with
tests once bpf-next is merged back, hopefully still making it into 4.16 :)

v4:
 - rebase on top of Alex's changes.

---

Quentin says:

This series tries to improve user experience when eBPF hardware offload
hits error paths at load time. In particular, it introduces netlink
extended ack support in the nfp driver.

To that aim, transmission of the pointer to the extack object is piped
through the `change()` operation of the existing classifiers (patch 1 to
6). Then it is used for TC offload in the nfp driver (patch 8) and in
netdevsim (patch 9, selftest in patch 10). Patch 7 adds a helper to handle
extack messages in the core when TC offload is disabled on the net device.

For completeness extack is propagated for classifiers other than cls_bpf,
but it's up to the drivers to make use of it.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Use the recently added extack support for eBPF offload in the driver.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pass a pointer to an extack object to nfp_app_xdp_offload() in order to
prepare for extack usage in the nfp driver. Next step will be to forward
this extack pointer to nfp_net_bpf_offload(), once this function is able
to use it for printing error messages.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Create a wrapper around tc_can_offload() that takes an additional
extack pointer argument in order to output an error message if TC
offload is disabled on the device.

In this way, the error message is handled by the core and can be the
same for all drivers.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add extack support for hardware offload of classifiers. In order
to achieve this, a pointer to a struct netlink_ext_ack is added to the
struct tc_cls_common_offload that is passed to the callback for setting
up the classifier. Function tc_cls_common_offload_init() is updated to
support initialization of this new attribute.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pass the extack pointer obtained in the `->change()` filter operation to
cls_bpf_offload() and then to cls_bpf_offload_cmd(). This makes it
possible to use this extack pointer in drivers offloading BPF programs
in a future patch.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Propagate the extack pointer from the `->change()` classifier operation
to the function used for filter replacement in cls_u32. This makes it
possible to use netlink extack messages in the future at replacement
time for this filter, although it is not used at this point.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Propagate the extack pointer from the `->change()` classifier operation
to the function used for filter replacement in cls_matchall. This makes
it possible to use netlink extack messages in the future at replacement
time for this filter, although it is not used at this point.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Propagate the extack pointer from the `->change()` classifier operation
to the function used for filter replacement in cls_flower. This makes it
possible to use netlink extack messages in the future at replacement
time for this filter, although it is not used at this point.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since we no longer localize channel/CPU affiliation within one NUMA
node, num_online_cpus() is used as the number of channel cap, instead of
the number of processors in a NUMA node.

This patch allows a bigger range for tuning the number of channels.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>