Jakub Sitnicki [Sun, 31 May 2020 08:28:41 +0000 (10:28 +0200)]
bpftool: Extract helpers for showing link attach type
Code for printing link attach_type is duplicated in a couple of places, and
likely will be duplicated for future link types as well. Create helpers to
prevent duplication.
Jakub Sitnicki [Sun, 31 May 2020 08:28:40 +0000 (10:28 +0200)]
libbpf: Add support for bpf_link-based netns attachment
Add bpf_program__attach_nets(), which uses LINK_CREATE subcommand to create
an FD-based kernel bpf_link, for attach types tied to network namespace,
that is BPF_FLOW_DISSECTOR for the moment.
Jakub Sitnicki [Sun, 31 May 2020 08:28:39 +0000 (10:28 +0200)]
bpf, cgroup: Return ENOLINK for auto-detached links on update
Failure to update a bpf_link because it has been auto-detached by a dying
cgroup currently results in EINVAL error, even though the arguments passed
to bpf() syscall are not wrong.
bpf_links attaching to netns in this case will return ENOLINK, which
carries the message that the link is no longer attached to anything.
Change cgroup bpf_links to do the same to keep the uAPI errors consistent.
Fixes: 0c991ebc8c69 ("bpf: Implement bpf_prog replacement for an active bpf_cgroup_link") Suggested-by: Lorenz Bauer <lmb@cloudflare.com> Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20200531082846.2117903-6-jakub@cloudflare.com
Jakub Sitnicki [Sun, 31 May 2020 08:28:38 +0000 (10:28 +0200)]
bpf: Add link-based BPF program attachment to network namespace
Extend bpf() syscall subcommands that operate on bpf_link, that is
LINK_CREATE, LINK_UPDATE, OBJ_GET_INFO, to accept attach types tied to
network namespaces (only flow dissector at the moment).
Link-based and prog-based attachment can be used interchangeably, but only
one can exist at a time. Attempts to attach a link when a prog is already
attached directly, and the other way around, will be met with -EEXIST.
Attempts to detach a program when link exists result in -EINVAL.
Attachment of multiple links of same attach type to one netns is not
supported with the intention to lift the restriction when a use-case
presents itself. Because of that link create returns -E2BIG when trying to
create another netns link, when one already exists.
Link-based attachments to netns don't keep a netns alive by holding a ref
to it. Instead links get auto-detached from netns when the latter is being
destroyed, using a pernet pre_exit callback.
When auto-detached, link lives in defunct state as long there are open FDs
for it. -ENOLINK is returned if a user tries to update a defunct link.
Because bpf_link to netns doesn't hold a ref to struct net, special care is
taken when releasing, updating, or filling link info. The netns might be
getting torn down when any of these link operations are in progress. That
is why auto-detach and update/release/fill_info are synchronized by the
same mutex. Also, link ops have to always check if auto-detach has not
happened yet and if netns is still alive (refcnt > 0).
Jakub Sitnicki [Sun, 31 May 2020 08:28:37 +0000 (10:28 +0200)]
flow_dissector: Move out netns_bpf prog callbacks
Move functions to manage BPF programs attached to netns that are not
specific to flow dissector to a dedicated module named
bpf/net_namespace.c.
The set of functions will grow with the addition of bpf_link support for
netns attached programs. This patch prepares ground by creating a place
for it.
This is a code move with no functional changes intended.
Jakub Sitnicki [Sun, 31 May 2020 08:28:36 +0000 (10:28 +0200)]
net: Introduce netns_bpf for BPF programs attached to netns
In order to:
(1) attach more than one BPF program type to netns, or
(2) support attaching BPF programs to netns with bpf_link, or
(3) support multi-prog attach points for netns
we will need to keep more state per netns than a single pointer like we
have now for BPF flow dissector program.
Prepare for the above by extracting netns_bpf that is part of struct net,
for storing all state related to BPF programs attached to netns.
Turn flow dissector callbacks for querying/attaching/detaching a program
into generic ones that operate on netns_bpf. Next patch will move the
generic callbacks into their own module.
This is similar to how it is organized for cgroup with cgroup_bpf.
Jakub Sitnicki [Sun, 31 May 2020 08:28:35 +0000 (10:28 +0200)]
flow_dissector: Pull locking up from prog attach callback
Split out the part of attach callback that happens with attach/detach lock
acquired. This structures the prog attach callback in a way that opens up
doors for moving the locking out of flow_dissector and into generic
callbacks for attaching/detaching progs to netns in subsequent patches.
Andrii Nakryiko [Mon, 1 Jun 2020 20:26:01 +0000 (13:26 -0700)]
libbpf: Add _GNU_SOURCE for reallocarray to ringbuf.c
On systems with recent enough glibc, reallocarray compat won't kick in, so
reallocarray() itself has to come from stdlib.h include. But _GNU_SOURCE is
necessary to enable it. So add it.
Fixes: bf99c936f947 ("libbpf: Add BPF ring buffer support") Signed-off-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200601202601.2139477-1-andriin@fb.com
Jiri Olsa [Sun, 31 May 2020 15:42:55 +0000 (17:42 +0200)]
bpf: Use tracing helpers for lsm programs
Currenty lsm uses bpf_tracing_func_proto helpers which do
not include stack trace or perf event output. It's useful
to have those for bpftrace lsm support [1].
Using tracing_prog_func_proto helpers for lsm programs.
Introduce xdp_convert_frame_to_buff utility routine to initialize xdp_buff
fields from xdp_frames ones. Rely on xdp_convert_frame_to_buff in veth xdp
code.
====================
This option makes it possible to programatically bind sockets
to netdevices. With the help of this option sockets
of VRF unaware applications could be distributed between
multiple VRFs with an eBPF program. This lets the applications
benefit from multiple possible routes.
v2:
- splitting up the patch to three parts
- lock_sk parameter for optional locking in sock_bindtoindex - Stanislav Fomichev
- testing the SO_BINDTODEVICE option - Andrii Nakryiko
====================
Ferenc Fejes [Sat, 30 May 2020 21:09:02 +0000 (23:09 +0200)]
selftests/bpf: Add test for SO_BINDTODEVICE opt of bpf_setsockopt
This test intended to verify if SO_BINDTODEVICE option works in
bpf_setsockopt. Because we already in the SOL_SOCKET level in this
connect bpf prog its safe to verify the sanity in the beginning of
the connect_v4_prog by calling the bind_to_device test helper.
The testing environment already created by the test_sock_addr.sh
script so this test assume that two netdevices already existing in
the system: veth pair with names test_sock_addr1 and test_sock_addr2.
The test will try to bind the socket to those devices first.
Then the test assume there are no netdevice with "nonexistent_dev"
name so the bpf_setsockopt will give use ENODEV error.
At the end the test remove the device binding from the socket
by binding it to an empty name.
Ferenc Fejes [Sat, 30 May 2020 21:09:01 +0000 (23:09 +0200)]
bpf: Allow SO_BINDTODEVICE opt in bpf_setsockopt
Extending the supported sockopts in bpf_setsockopt with
SO_BINDTODEVICE. We call sock_bindtoindex with parameter
lock_sk = false in this context because we already owning
the socket.
Ferenc Fejes [Sat, 30 May 2020 21:09:00 +0000 (23:09 +0200)]
net: Make locking in sock_bindtoindex optional
The sock_bindtoindex intended for kernel wide usage however
it will lock the socket regardless of the context. This modification
relax this behavior optionally: locking the socket will be optional
by calling the sock_bindtoindex with lock_sk = true.
The modification applied to all users of the sock_bindtoindex.
====================
If a socket is running a BPF_SK_SKB_SREAM_VERDICT program and KTLS is
enabled the data stream may be broken if both TLS stream parser and
BPF stream parser try to handle data. Fix this here by making KTLS
stream parser run first to ensure TLS messages are received correctly
and then calling the verdict program. This analogous to how we handle
a similar conflict on the TX side.
Note, this is a fix but it doesn't make sense to push this late to
bpf tree so targeting bpf-next and keeping fixes tags.
====================
John Fastabend [Fri, 29 May 2020 23:07:19 +0000 (16:07 -0700)]
bpf, selftests: Add test for ktls with skb bpf ingress policy
This adds a test for bpf ingress policy. To ensure data writes happen
as expected with extra TLS headers we run these tests with data
verification enabled by default. This will test receive packets have
"PASS" stamped into the front of the payload.
====================
Implementation of Daniel's proposal for allowing DEVMAP entries to be
a device index, program fd pair.
Programs are run after XDP_REDIRECT and have access to both Rx device
and Tx device.
v4
- moved struct bpf_devmap_val from uapi to devmap.c, named the union
and dropped the prefix from the elements - Jesper
- fixed 2 bugs in selftests
v3
- renamed struct to bpf_devmap_val
- used offsetofend to check for expected map size, modification of
Toke's comment
- check for explicit value sizes
- adjusted switch statement in dev_map_run_prog per Andrii's comment
- changed SEC shortcut to xdp_devmap
- changed selftests to use skeleton and new map declaration
v2
- moved dev_map_ext_val definition to uapi to formalize the API for devmap
extensions; add bpf_ prefix to the prog_fd and prog_id entries
- changed devmap code to handle struct in a way that it can support future
extensions
- fixed subject in libbpf patch
v1
- fixed prog put on invalid program - Toke
- changed write value from id to fd per Toke's comments about capabilities
- add test cases
====================
John Fastabend [Fri, 29 May 2020 23:06:59 +0000 (16:06 -0700)]
bpf: Fix running sk_skb program types with ktls
KTLS uses a stream parser to collect TLS messages and send them to
the upper layer tls receive handler. This ensures the tls receiver
has a full TLS header to parse when it is run. However, when a
socket has BPF_SK_SKB_STREAM_VERDICT program attached before KTLS
is enabled we end up with two stream parsers running on the same
socket.
The result is both try to run on the same socket. First the KTLS
stream parser runs and calls read_sock() which will tcp_read_sock
which in turn calls tcp_rcv_skb(). This dequeues the skb from the
sk_receive_queue. When this is done KTLS code then data_ready()
callback which because we stacked KTLS on top of the bpf stream
verdict program has been replaced with sk_psock_start_strp(). This
will in turn kick the stream parser again and eventually do the
same thing KTLS did above calling into tcp_rcv_skb() and dequeuing
a skb from the sk_receive_queue.
At this point the data stream is broke. Part of the stream was
handled by the KTLS side some other bytes may have been handled
by the BPF side. Generally this results in either missing data
or more likely a "Bad Message" complaint from the kTLS receive
handler as the BPF program steals some bytes meant to be in a
TLS header and/or the TLS header length is no longer correct.
We've already broke the idealized model where we can stack ULPs
in any order with generic callbacks on the TX side to handle this.
So in this patch we do the same thing but for RX side. We add
a sk_psock_strp_enabled() helper so TLS can learn a BPF verdict
program is running and add a tls_sw_has_ctx_rx() helper so BPF
side can learn there is a TLS ULP on the socket.
Then on BPF side we omit calling our stream parser to avoid
breaking the data stream for the KTLS receiver. Then on the
KTLS side we call BPF_SK_SKB_STREAM_VERDICT once the KTLS
receiver is done with the packet but before it posts the
msg to userspace. This gives us symmetry between the TX and
RX halfs and IMO makes it usable again. On the TX side we
process packets in this order BPF -> TLS -> TCP and on
the receive side in the reverse order TCP -> TLS -> BPF.
Discovered while testing OpenSSL 3.0 Alpha2.0 release.
David Ahern [Fri, 29 May 2020 22:07:16 +0000 (16:07 -0600)]
selftest: Add tests for XDP programs in devmap entries
Add tests to verify ability to add an XDP program to a
entry in a DEVMAP.
Add negative tests to show DEVMAP programs can not be
attached to devices as a normal XDP program, and accesses
to egress_ifindex require BPF_XDP_DEVMAP attach type.
John Fastabend [Fri, 29 May 2020 23:06:41 +0000 (16:06 -0700)]
bpf: Refactor sockmap redirect code so its easy to reuse
We will need this block of code called from tls context shortly
lets refactor the redirect logic so its easy to use. This also
cleans up the switch stmt so we have fewer fallthrough cases.
No logic changes are intended.
Fixes: d829e9c4112b5 ("tls: convert to generic sk_msg interface") Signed-off-by: John Fastabend <john.fastabend@gmail.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Reviewed-by: Jakub Sitnicki <jakub@cloudflare.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/159079360110.5745.7024009076049029819.stgit@john-Precision-5820-Tower Signed-off-by: Alexei Starovoitov <ast@kernel.org>
David Ahern [Fri, 29 May 2020 22:07:14 +0000 (16:07 -0600)]
xdp: Add xdp_txq_info to xdp_buff
Add xdp_txq_info as the Tx counterpart to xdp_rxq_info. At the
moment only the device is added. Other fields (queue_index)
can be added as use cases arise.
>From a UAPI perspective, add egress_ifindex to xdp context for
bpf programs to see the Tx device.
Update the verifier to only allow accesses to egress_ifindex by
XDP programs with BPF_XDP_DEVMAP expected attach type.
David Ahern [Fri, 29 May 2020 22:07:13 +0000 (16:07 -0600)]
bpf: Add support to attach bpf program to a devmap entry
Add BPF_XDP_DEVMAP attach type for use with programs associated with a
DEVMAP entry.
Allow DEVMAPs to associate a program with a device entry by adding
a bpf_prog.fd to 'struct bpf_devmap_val'. Values read show the program
id, so the fd and id are a union. bpf programs can get access to the
struct via vmlinux.h.
The program associated with the fd must have type XDP with expected
attach type BPF_XDP_DEVMAP. When a program is associated with a device
index, the program is run on an XDP_REDIRECT and before the buffer is
added to the per-cpu queue. At this point rxq data is still valid; the
next patch adds tx device information allowing the prorgam to see both
ingress and egress device indices.
XDP generic is skb based and XDP programs do not work with skb's. Block
the use case by walking maps used by a program that is to be attached
via xdpgeneric and fail if any of them are DEVMAP / DEVMAP_HASH with
Block attach of BPF_XDP_DEVMAP programs to devices.
Yonghong Song [Fri, 29 May 2020 00:48:10 +0000 (17:48 -0700)]
bpf: Use strncpy_from_unsafe_strict() in bpf_seq_printf() helper
In bpf_seq_printf() helper, when user specified a "%s" in the
format string, strncpy_from_unsafe() is used to read the actual string
to a buffer. The string could be a format string or a string in
the kernel data structure. It is really unlikely that the string
will reside in the user memory.
This is different from Commit b2a5212fb634 ("bpf: Restrict bpf_trace_printk()'s %s
usage and add %pks, %pus specifier") which still used
strncpy_from_unsafe() for "%s" to preserve the old behavior.
If in the future, bpf_seq_printf() indeed needs to read user
memory, we can implement "%pus" format string.
Based on discussion in [1], if the intent is to read kernel memory,
strncpy_from_unsafe_strict() should be used. So this patch
changed to use strncpy_from_unsafe_strict().
Amritha Nambiar [Wed, 27 May 2020 00:34:36 +0000 (17:34 -0700)]
bpf: Add rx_queue_mapping to bpf_sock
Add "rx_queue_mapping" to bpf_sock. This gives read access for the
existing field (sk_rx_queue_mapping) of struct sock from bpf_sock.
Semantics for the bpf_sock rx_queue_mapping access are similar to
sk_rx_queue_get(), i.e the value NO_QUEUE_MAPPING is not allowed
and -1 is returned in that case. This is useful for transmit queue
selection based on the received queue index which is cached in the
socket in the receive path.
v3: Addressed review comments to add usecase in patch description,
and fixed default value for rx_queue_mapping.
v2: fixed build error for CONFIG_XPS wrapping, reported by
kbuild test robot <lkp@intel.com>
Daniel Borkmann [Fri, 29 May 2020 15:11:09 +0000 (17:11 +0200)]
Merge branch 'bpf-ring-buffer'
Andrii Nakryiko says:
====================
Implement a new BPF ring buffer, as presented at BPF virtual conference ([0]).
It presents an alternative to perf buffer, following its semantics closely,
but allowing sharing same instance of ring buffer across multiple CPUs
efficiently.
Most patches have extensive commentary explaining various aspects, so I'll
keep cover letter short. Overall structure of the patch set:
- patch #1 adds BPF ring buffer implementation to kernel and necessary
verifier support;
- patch #2 adds libbpf consumer implementation for BPF ringbuf;
- patch #3 adds selftest, both for single BPF ring buf use case, as well as
using it with array/hash of maps;
- patch #4 adds extensive benchmarks and provide some analysis in commit
message, it builds upon selftests/bpf's bench runner.
- patch #5 adds most of patch #1 commit message as a doc under
Documentation/bpf/ringbuf.rst.
Litmus tests, validating consumer/producer protocols and memory orderings,
were moved out as discussed in [1] and are going to be posted against -rcu
tree and put under Documentation/litmus-tests/bpf-rb.
v3->v4:
- fix ringbuf freeing (vunmap, __free_page); verified with a trivial loop
creating and closing ringbuf map endlessly (Daniel);
v2->v3:
- dropped unnecessary smp_wmb() (Paul);
- verifier reference type enhancement patch was dropped (Alexei);
- better verifier message for various memory access checks (Alexei);
- clarified a bit roundup_len() bit shifting (Alexei);
- converted doc to .rst (Alexei);
- fixed warning on 32-bit arches regarding tautological ring area size check.
v1->v2:
- commit()/discard()/output() accept flags (NO_WAKEUP/FORCE_WAKEUP) (Stanislav);
- bpf_ringbuf_query() added, returning available data size, ringbuf size,
consumer/producer positions, needed to implement smarter notification policy
(Stanislav);
- added ringbuf UAPI constants to include/uapi/linux/bpf.h (Jonathan);
- fixed sample size check, added proper ringbuf size check (Jonathan, Alexei);
- wake_up_all() is done through irq_work (Alexei);
- consistent use of smp_load_acquire/smp_store_release, no
READ_ONCE/WRITE_ONCE (Alexei);
- added Documentation/bpf/ringbuf.txt (Stanislav);
- updated litmus test with smp_load_acquire/smp_store_release changes;
- added ring_buffer__consume() API to libbpf for busy-polling;
- ring_buffer__poll() on success returns number of records consumed;
- fixed EPOLL notifications, don't assume available data, done similarly to
perfbuf's implementation;
- both ringbuf and perfbuf now have --rb-sampled mode, instead of
pb-raw/pb-custom mode, updated benchmark results;
- extended ringbuf selftests to validate epoll logic/manual notification
logic, as well as bpf_ringbuf_query().
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Anton Protopopov [Wed, 27 May 2020 18:57:00 +0000 (18:57 +0000)]
selftests/bpf: Add tests for write-only stacks/queues
For write-only stacks and queues bpf_map_update_elem should be allowed, but
bpf_map_lookup_elem and bpf_map_lookup_and_delete_elem should fail with EPERM.
Andrii Nakryiko [Fri, 29 May 2020 07:54:24 +0000 (00:54 -0700)]
docs/bpf: Add BPF ring buffer design notes
Add commit description from patch #1 as a stand-alone documentation under
Documentation/bpf, as it might be more convenient format, in long term
perspective.
Suggested-by: Stanislav Fomichev <sdf@google.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/bpf/20200529075424.3139988-6-andriin@fb.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Andrii Nakryiko [Fri, 29 May 2020 07:54:23 +0000 (00:54 -0700)]
bpf: Add BPF ringbuf and perf buffer benchmarks
Extend bench framework with ability to have benchmark-provided child argument
parser for custom benchmark-specific parameters. This makes bench generic code
modular and independent from any specific benchmark.
Also implement a set of benchmarks for new BPF ring buffer and existing perf
buffer. 4 benchmarks were implemented: 2 variations for each of BPF ringbuf
and perfbuf:,
- rb-libbpf utilizes stock libbpf ring_buffer manager for reading data;
- rb-custom implements custom ring buffer setup and reading code, to
eliminate overheads inherent in generic libbpf code due to callback
functions and the need to update consumer position after each consumed
record, instead of batching updates (due to pessimistic assumption that
user callback might take long time and thus could unnecessarily hold ring
buffer space for too long);
- pb-libbpf uses stock libbpf perf_buffer code with all the default
settings, though uses higher-performance raw event callback to minimize
unnecessary overhead;
- pb-custom implements its own custom consumer code to minimize any possible
overhead of generic libbpf implementation and indirect function calls.
All of the test support default, no data notification skipped, mode, as well
as sampled mode (with --rb-sampled flag), which allows to trigger epoll
notification less frequently and reduce overhead. As will be shown, this mode
is especially critical for perf buffer, which suffers from high overhead of
wakeups in kernel.
Otherwise, all benchamrks implement similar way to generate a batch of records
by using fentry/sys_getpgid BPF program, which pushes a bunch of records in
a tight loop and records number of successful and dropped samples. Each record
is a small 8-byte integer, to minimize the effect of memory copying with
bpf_perf_event_output() and bpf_ringbuf_output().
Benchmarks that have only one producer implement optional back-to-back mode,
in which record production and consumption is alternating on the same CPU.
This is the highest-throughput happy case, showing ultimate performance
achievable with either BPF ringbuf or perfbuf.
All the below scenarios are implemented in a script in
benchs/run_bench_ringbufs.sh. Tests were performed on 28-core/56-thread
Intel Xeon CPU E5-2680 v4 @ 2.40GHz CPU.
Single producer on one CPU, consumer on another one, both running at full
speed. Curiously, rb-libbpf has higher throughput than objectively faster (due
to more lightweight consumer code path) rb-custom. It appears that faster
consumer causes kernel to send notifications more frequently, because consumer
appears to be caught up more frequently. Performance of perfbuf suffers from
default "no sampling" policy and huge overhead that causes.
In sampled mode, rb-custom is winning very significantly eliminating too
frequent in-kernel wakeups, the gain appears to be more than 2x.
Perf buffer achieves even more impressive wins, compared to stock perfbuf
settings, with 10x improvements in throughput with 1:500 sampling rate. The
trade-off is that with sampling, application might not get next X events until
X+1st arrives, which is not always acceptable. With steady influx of events,
though, this shouldn't be a problem.
Overall, single-producer performance of ring buffers seems to be better no
matter the sampled/non-sampled modes, but it especially beats ring buffer
without sampling due to its adaptive notification approach.
Here we test a back-to-back mode, which is arguably best-case scenario both
for BPF ringbuf and perfbuf, because there is no contention and for ringbuf
also no excessive notification, because consumer appears to be behind after
the first record. For ringbuf, custom consumer code clearly wins with 21.5 vs
16 million records per second exchanged between producer and consumer. Sampled
mode actually hurts a bit due to slightly slower producer logic (it needs to
fetch amount of data available to decide whether to skip or force notification).
Perfbuf with wakeup sampling gets 5.5x throughput increase, compared to
no-sampling version. There also doesn't seem to be noticeable overhead from
generic libbpf handling code.
This benchmark shows the effect of event sampling for perfbuf. Back-to-back
mode for highest throughput. Just doing every 5th record notification gives
3.5x speed up. 250-500 appears to be the point of diminishing return, with
almost 9x speed up. Most benchmarks use 500 as the default sampling for pb-raw
and pb-custom.
Similar benchmark for ring buffer also shows a great advantage (in terms of
throughput) of skipping notifications. Skipping every 5th one gives 4x boost.
Also similar to perfbuf case, 250-500 seems to be the point of diminishing
returns, giving roughly 20x better results.
Keep in mind, for this test, notifications are controlled manually with
BPF_RB_NO_WAKEUP and BPF_RB_FORCE_WAKEUP. As can be seen from previous
benchmarks, adaptive notifications based on consumer's positions provides same
(or even slightly better due to simpler load generator on BPF side) benefits in
favorable back-to-back scenario. Over zealous and fast consumer, which is
almost always caught up, will make thoughput numbers smaller. That's the case
when manual notification control might prove to be extremely beneficial.
BPF ringbuf supports two sets of APIs with various usability and performance
tradeoffs: bpf_ringbuf_reserve()+bpf_ringbuf_commit() vs bpf_ringbuf_output().
This benchmark clearly shows superiority of reserve+commit approach, despite
using a small 8-byte record size.
This benchmark shows one of the worst-case scenarios, in which producer and
consumer do not coordinate *and* fight for the same CPU. No batch count and
sampling settings were able to eliminate drops for ringbuffer, producer is
just too fast for consumer to keep up. But ringbuf and perfbuf still able to
pass through quite a lot of messages, which is more than enough for a lot of
applications.
Ringbuf uses a very short-duration spinlock during reservation phase, to check
few invariants, increment producer count and set record header. This is the
biggest point of contention for ringbuf implementation. This benchmark
evaluates the effect of multiple competing writers on overall throughput of
a single shared ringbuffer.
Overall throughput drops almost 2x when going from single to two
highly-contended producers, gradually dropping with additional competing
producers. Performance drop stabilizes at around 20 producers and hovers
around 2mln even with 50+ fighting producers, which is a 5x drop compared to
non-contended case. Good kernel implementation in kernel helps maintain decent
performance here.
Note, that in the intended real-world scenarios, it's not expected to get even
close to such a high levels of contention. But if contention will become
a problem, there is always an option of sharding few ring buffers across a set
of CPUs.
Andrii Nakryiko [Fri, 29 May 2020 07:54:22 +0000 (00:54 -0700)]
selftests/bpf: Add BPF ringbuf selftests
Both singleton BPF ringbuf and BPF ringbuf with map-in-map use cases are tested.
Also reserve+submit/discards and output variants of API are validated.
Andrii Nakryiko [Fri, 29 May 2020 07:54:21 +0000 (00:54 -0700)]
libbpf: Add BPF ring buffer support
Declaring and instantiating BPF ring buffer doesn't require any changes to
libbpf, as it's just another type of maps. So using existing BTF-defined maps
syntax with __uint(type, BPF_MAP_TYPE_RINGBUF) and __uint(max_elements,
<size-of-ring-buf>) is all that's necessary to create and use BPF ring buffer.
This patch adds BPF ring buffer consumer to libbpf. It is very similar to
perf_buffer implementation in terms of API, but also attempts to fix some
minor problems and inconveniences with existing perf_buffer API.
ring_buffer support both single ring buffer use case (with just using
ring_buffer__new()), as well as allows to add more ring buffers, each with its
own callback and context. This allows to efficiently poll and consume
multiple, potentially completely independent, ring buffers, using single
epoll instance.
The latter is actually a problem in practice for applications
that are using multiple sets of perf buffers. They have to create multiple
instances for struct perf_buffer and poll them independently or in a loop,
each approach having its own problems (e.g., inability to use a common poll
timeout). struct ring_buffer eliminates this problem by aggregating many
independent ring buffer instances under the single "ring buffer manager".
Second, perf_buffer's callback can't return error, so applications that need
to stop polling due to error in data or data signalling the end, have to use
extra mechanisms to signal that polling has to stop. ring_buffer's callback
can return error, which will be passed through back to user code and can be
acted upon appropariately.
Two APIs allow to consume ring buffer data:
- ring_buffer__poll(), which will wait for data availability notification
and will consume data only from reported ring buffer(s); this API allows
to efficiently use resources by reading data only when it becomes
available;
- ring_buffer__consume(), will attempt to read new records regardless of
data availablity notification sub-system. This API is useful for cases
when lowest latency is required, in expense of burning CPU resources.
Andrii Nakryiko [Fri, 29 May 2020 07:54:20 +0000 (00:54 -0700)]
bpf: Implement BPF ring buffer and verifier support for it
This commit adds a new MPSC ring buffer implementation into BPF ecosystem,
which allows multiple CPUs to submit data to a single shared ring buffer. On
the consumption side, only single consumer is assumed.
Motivation
----------
There are two distinctive motivators for this work, which are not satisfied by
existing perf buffer, which prompted creation of a new ring buffer
implementation.
- more efficient memory utilization by sharing ring buffer across CPUs;
- preserving ordering of events that happen sequentially in time, even
across multiple CPUs (e.g., fork/exec/exit events for a task).
These two problems are independent, but perf buffer fails to satisfy both.
Both are a result of a choice to have per-CPU perf ring buffer. Both can be
also solved by having an MPSC implementation of ring buffer. The ordering
problem could technically be solved for perf buffer with some in-kernel
counting, but given the first one requires an MPSC buffer, the same solution
would solve the second problem automatically.
Semantics and APIs
------------------
Single ring buffer is presented to BPF programs as an instance of BPF map of
type BPF_MAP_TYPE_RINGBUF. Two other alternatives considered, but ultimately
rejected.
One way would be to, similar to BPF_MAP_TYPE_PERF_EVENT_ARRAY, make
BPF_MAP_TYPE_RINGBUF could represent an array of ring buffers, but not enforce
"same CPU only" rule. This would be more familiar interface compatible with
existing perf buffer use in BPF, but would fail if application needed more
advanced logic to lookup ring buffer by arbitrary key. HASH_OF_MAPS addresses
this with current approach. Additionally, given the performance of BPF
ringbuf, many use cases would just opt into a simple single ring buffer shared
among all CPUs, for which current approach would be an overkill.
Another approach could introduce a new concept, alongside BPF map, to
represent generic "container" object, which doesn't necessarily have key/value
interface with lookup/update/delete operations. This approach would add a lot
of extra infrastructure that has to be built for observability and verifier
support. It would also add another concept that BPF developers would have to
familiarize themselves with, new syntax in libbpf, etc. But then would really
provide no additional benefits over the approach of using a map.
BPF_MAP_TYPE_RINGBUF doesn't support lookup/update/delete operations, but so
doesn't few other map types (e.g., queue and stack; array doesn't support
delete, etc).
The approach chosen has an advantage of re-using existing BPF map
infrastructure (introspection APIs in kernel, libbpf support, etc), being
familiar concept (no need to teach users a new type of object in BPF program),
and utilizing existing tooling (bpftool). For common scenario of using
a single ring buffer for all CPUs, it's as simple and straightforward, as
would be with a dedicated "container" object. On the other hand, by being
a map, it can be combined with ARRAY_OF_MAPS and HASH_OF_MAPS map-in-maps to
implement a wide variety of topologies, from one ring buffer for each CPU
(e.g., as a replacement for perf buffer use cases), to a complicated
application hashing/sharding of ring buffers (e.g., having a small pool of
ring buffers with hashed task's tgid being a look up key to preserve order,
but reduce contention).
Key and value sizes are enforced to be zero. max_entries is used to specify
the size of ring buffer and has to be a power of 2 value.
There are a bunch of similarities between perf buffer
(BPF_MAP_TYPE_PERF_EVENT_ARRAY) and new BPF ring buffer semantics:
- variable-length records;
- if there is no more space left in ring buffer, reservation fails, no
blocking;
- memory-mappable data area for user-space applications for ease of
consumption and high performance;
- epoll notifications for new incoming data;
- but still the ability to do busy polling for new data to achieve the
lowest latency, if necessary.
BPF ringbuf provides two sets of APIs to BPF programs:
- bpf_ringbuf_output() allows to *copy* data from one place to a ring
buffer, similarly to bpf_perf_event_output();
- bpf_ringbuf_reserve()/bpf_ringbuf_commit()/bpf_ringbuf_discard() APIs
split the whole process into two steps. First, a fixed amount of space is
reserved. If successful, a pointer to a data inside ring buffer data area
is returned, which BPF programs can use similarly to a data inside
array/hash maps. Once ready, this piece of memory is either committed or
discarded. Discard is similar to commit, but makes consumer ignore the
record.
bpf_ringbuf_output() has disadvantage of incurring extra memory copy, because
record has to be prepared in some other place first. But it allows to submit
records of the length that's not known to verifier beforehand. It also closely
matches bpf_perf_event_output(), so will simplify migration significantly.
bpf_ringbuf_reserve() avoids the extra copy of memory by providing a memory
pointer directly to ring buffer memory. In a lot of cases records are larger
than BPF stack space allows, so many programs have use extra per-CPU array as
a temporary heap for preparing sample. bpf_ringbuf_reserve() avoid this needs
completely. But in exchange, it only allows a known constant size of memory to
be reserved, such that verifier can verify that BPF program can't access
memory outside its reserved record space. bpf_ringbuf_output(), while slightly
slower due to extra memory copy, covers some use cases that are not suitable
for bpf_ringbuf_reserve().
The difference between commit and discard is very small. Discard just marks
a record as discarded, and such records are supposed to be ignored by consumer
code. Discard is useful for some advanced use-cases, such as ensuring
all-or-nothing multi-record submission, or emulating temporary malloc()/free()
within single BPF program invocation.
Each reserved record is tracked by verifier through existing
reference-tracking logic, similar to socket ref-tracking. It is thus
impossible to reserve a record, but forget to submit (or discard) it.
bpf_ringbuf_query() helper allows to query various properties of ring buffer.
Currently 4 are supported:
- BPF_RB_AVAIL_DATA returns amount of unconsumed data in ring buffer;
- BPF_RB_RING_SIZE returns the size of ring buffer;
- BPF_RB_CONS_POS/BPF_RB_PROD_POS returns current logical possition of
consumer/producer, respectively.
Returned values are momentarily snapshots of ring buffer state and could be
off by the time helper returns, so this should be used only for
debugging/reporting reasons or for implementing various heuristics, that take
into account highly-changeable nature of some of those characteristics.
One such heuristic might involve more fine-grained control over poll/epoll
notifications about new data availability in ring buffer. Together with
BPF_RB_NO_WAKEUP/BPF_RB_FORCE_WAKEUP flags for output/commit/discard helpers,
it allows BPF program a high degree of control and, e.g., more efficient
batched notifications. Default self-balancing strategy, though, should be
adequate for most applications and will work reliable and efficiently already.
Design and implementation
-------------------------
This reserve/commit schema allows a natural way for multiple producers, either
on different CPUs or even on the same CPU/in the same BPF program, to reserve
independent records and work with them without blocking other producers. This
means that if BPF program was interruped by another BPF program sharing the
same ring buffer, they will both get a record reserved (provided there is
enough space left) and can work with it and submit it independently. This
applies to NMI context as well, except that due to using a spinlock during
reservation, in NMI context, bpf_ringbuf_reserve() might fail to get a lock,
in which case reservation will fail even if ring buffer is not full.
The ring buffer itself internally is implemented as a power-of-2 sized
circular buffer, with two logical and ever-increasing counters (which might
wrap around on 32-bit architectures, that's not a problem):
- consumer counter shows up to which logical position consumer consumed the
data;
- producer counter denotes amount of data reserved by all producers.
Each time a record is reserved, producer that "owns" the record will
successfully advance producer counter. At that point, data is still not yet
ready to be consumed, though. Each record has 8 byte header, which contains
the length of reserved record, as well as two extra bits: busy bit to denote
that record is still being worked on, and discard bit, which might be set at
commit time if record is discarded. In the latter case, consumer is supposed
to skip the record and move on to the next one. Record header also encodes
record's relative offset from the beginning of ring buffer data area (in
pages). This allows bpf_ringbuf_commit()/bpf_ringbuf_discard() to accept only
the pointer to the record itself, without requiring also the pointer to ring
buffer itself. Ring buffer memory location will be restored from record
metadata header. This significantly simplifies verifier, as well as improving
API usability.
Producer counter increments are serialized under spinlock, so there is
a strict ordering between reservations. Commits, on the other hand, are
completely lockless and independent. All records become available to consumer
in the order of reservations, but only after all previous records where
already committed. It is thus possible for slow producers to temporarily hold
off submitted records, that were reserved later.
Reservation/commit/consumer protocol is verified by litmus tests in
Documentation/litmus-test/bpf-rb.
One interesting implementation bit, that significantly simplifies (and thus
speeds up as well) implementation of both producers and consumers is how data
area is mapped twice contiguously back-to-back in the virtual memory. This
allows to not take any special measures for samples that have to wrap around
at the end of the circular buffer data area, because the next page after the
last data page would be first data page again, and thus the sample will still
appear completely contiguous in virtual memory. See comment and a simple ASCII
diagram showing this visually in bpf_ringbuf_area_alloc().
Another feature that distinguishes BPF ringbuf from perf ring buffer is
a self-pacing notifications of new data being availability.
bpf_ringbuf_commit() implementation will send a notification of new record
being available after commit only if consumer has already caught up right up
to the record being committed. If not, consumer still has to catch up and thus
will see new data anyways without needing an extra poll notification.
Benchmarks (see tools/testing/selftests/bpf/benchs/bench_ringbuf.c) show that
this allows to achieve a very high throughput without having to resort to
tricks like "notify only every Nth sample", which are necessary with perf
buffer. For extreme cases, when BPF program wants more manual control of
notifications, commit/discard/output helpers accept BPF_RB_NO_WAKEUP and
BPF_RB_FORCE_WAKEUP flags, which give full control over notifications of data
availability, but require extra caution and diligence in using this API.
Comparison to alternatives
--------------------------
Before considering implementing BPF ring buffer from scratch existing
alternatives in kernel were evaluated, but didn't seem to meet the needs. They
largely fell into few categores:
- per-CPU buffers (perf, ftrace, etc), which don't satisfy two motivations
outlined above (ordering and memory consumption);
- linked list-based implementations; while some were multi-producer designs,
consuming these from user-space would be very complicated and most
probably not performant; memory-mapping contiguous piece of memory is
simpler and more performant for user-space consumers;
- io_uring is SPSC, but also requires fixed-sized elements. Naively turning
SPSC queue into MPSC w/ lock would have subpar performance compared to
locked reserve + lockless commit, as with BPF ring buffer. Fixed sized
elements would be too limiting for BPF programs, given existing BPF
programs heavily rely on variable-sized perf buffer already;
- specialized implementations (like a new printk ring buffer, [0]) with lots
of printk-specific limitations and implications, that didn't seem to fit
well for intended use with BPF programs.
Anton Protopopov [Wed, 27 May 2020 18:56:59 +0000 (18:56 +0000)]
bpf: Fix map permissions check
The map_lookup_and_delete_elem() function should check for both FMODE_CAN_WRITE
and FMODE_CAN_READ permissions because it returns a map element to user space.
Fixes: bd513cd08f10 ("bpf: add MAP_LOOKUP_AND_DELETE_ELEM syscall") Signed-off-by: Anton Protopopov <a.s.protopopov@gmail.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/bpf/20200527185700.14658-5-a.s.protopopov@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org>
John Fastabend [Sun, 24 May 2020 16:51:57 +0000 (09:51 -0700)]
bpf, selftests: Test probe_* helpers from SCHED_CLS
Lets test using probe* in SCHED_CLS network programs as well just
to be sure these keep working. Its cheap to add the extra test
and provides a second context to test outside of sk_msg after
we generalized probe* helpers to all networking types.
John Fastabend [Sun, 24 May 2020 16:51:36 +0000 (09:51 -0700)]
bpf, selftests: Add sk_msg helpers load and attach test
The test itself is not particularly useful but it encodes a common
pattern we have.
Namely do a sk storage lookup then depending on data here decide if
we need to do more work or alternatively allow packet to PASS. Then
if we need to do more work consult task_struct for more information
about the running task. Finally based on this additional information
drop or pass the data. In this case the suspicious check is not so
realisitic but it encodes the general pattern and uses the helpers
so we test the workflow.
This is a load test to ensure verifier correctly handles this case.
John Fastabend [Sun, 24 May 2020 16:50:55 +0000 (09:50 -0700)]
bpf: Extend bpf_base_func_proto helpers with probe_* and *current_task*
Often it is useful when applying policy to know something about the
task. If the administrator has CAP_SYS_ADMIN rights then they can
use kprobe + networking hook and link the two programs together to
accomplish this. However, this is a bit clunky and also means we have
to call both the network program and kprobe program when we could just
use a single program and avoid passing metadata through sk_msg/skb->cb,
socket, maps, etc.
To accomplish this add probe_* helpers to bpf_base_func_proto programs
guarded by a perfmon_capable() check. New supported helpers are the
following,
Eelco Chaudron [Tue, 26 May 2020 09:21:42 +0000 (11:21 +0200)]
libbpf: Add API to consume the perf ring buffer content
This new API, perf_buffer__consume, can be used as follows:
- When you have a perf ring where wakeup_events is higher than 1,
and you have remaining data in the rings you would like to pull
out on exit (or maybe based on a timeout).
- For low latency cases where you burn a CPU that constantly polls
the queues.
Jakub Sitnicki [Mon, 25 May 2020 12:29:28 +0000 (14:29 +0200)]
bpf: Fix returned error sign when link doesn't support updates
System calls encode returned errors as negative values. Fix a typo that
breaks this convention for bpf(LINK_UPDATE) when bpf_link doesn't support
update operation.
Tobias Klauser [Mon, 25 May 2020 13:54:21 +0000 (15:54 +0200)]
tools, bpftool: Print correct error message when failing to load BTF
btf__parse_raw and btf__parse_elf return negative error numbers wrapped
in an ERR_PTR, so the extracted value needs to be negated before passing
them to strerror which expects a positive error number.
Before:
Error: failed to load BTF from .../vmlinux: Unknown error -2
After:
Error: failed to load BTF from .../vmlinux: No such file or directory
Quentin Monnet [Sat, 23 May 2020 01:02:47 +0000 (02:02 +0100)]
tools, bpftool: Make capability check account for new BPF caps
Following the introduction of CAP_BPF, and the switch from CAP_SYS_ADMIN
to other capabilities for various BPF features, update the capability
checks (and potentially, drops) in bpftool for feature probes. Because
bpftool and/or the system might not know of CAP_BPF yet, some caution is
necessary:
- If compiled and run on a system with CAP_BPF, check CAP_BPF,
CAP_SYS_ADMIN, CAP_PERFMON, CAP_NET_ADMIN.
- Guard against CAP_BPF being undefined, to allow compiling bpftool from
latest sources on older systems. If the system where feature probes
are run does not know of CAP_BPF, stop checking after CAP_SYS_ADMIN,
as this should be the only capability required for all the BPF
probing.
- If compiled from latest sources on a system without CAP_BPF, but later
executed on a newer system with CAP_BPF knowledge, then we only test
CAP_SYS_ADMIN. Some probes may fail if the bpftool process has
CAP_SYS_ADMIN but misses the other capabilities. The alternative would
be to redefine the value for CAP_BPF in bpftool, but this does not
look clean, and the case sounds relatively rare anyway.
Note that libcap offers a cap_to_name() function to retrieve the name of
a given capability (e.g. "cap_sys_admin"). We do not use it because
deriving the names from the macros looks simpler than using
cap_to_name() (doing a strdup() on the string) + cap_free() + handling
the case of failed allocations, when we just want to use the name of the
capability in an error message.
The checks when compiling without libcap (i.e. root versus non-root) are
unchanged.
v2:
- Do not allocate cap_list dynamically.
- Drop BPF-related capabilities when running with "unprivileged", even
if we didn't have the full set in the first place (in v1, we would
skip dropping them in that case).
- Keep track of what capabilities we have, print the names of the
missing ones for privileged probing.
- Attempt to drop only the capabilities we actually have.
- Rename a couple variables.
Quentin Monnet [Sat, 23 May 2020 01:07:51 +0000 (02:07 +0100)]
tools, bpftool: Clean subcommand help messages
This is a clean-up for the formatting of the do_help functions for
bpftool's subcommands. The following fixes are included:
- Do not use argv[-2] for "iter" help message, as the help is shown by
default if no "iter" action is selected, resulting in messages looking
like "./bpftool bpftool pin...".
- Do not print unused HELP_SPEC_PROGRAM in help message for "bpftool
link".
- Andrii used argument indexing to avoid having multiple occurrences of
bin_name and argv[-2] in the fprintf() for the help message, for
"bpftool gen" and "bpftool link". Let's reuse this for all other help
functions. We can remove up to thirty arguments for the "bpftool map"
help message.
- Harmonise all functions, e.g. use ending quotes-comma on a separate
line.
Vladimir Oltean [Sun, 31 May 2020 18:25:51 +0000 (21:25 +0300)]
net: dsa: sja1105: suppress -Wmissing-prototypes in sja1105_vl.c
Newer C compilers are complaining about the fact that there are no
function prototypes in sja1105_vl.c for the non-static functions.
Give them what they want.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 1 Jun 2020 19:09:08 +0000 (12:09 -0700)]
Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue
Jeff Kirsher says:
====================
100GbE Intel Wired LAN Driver Updates 2020-05-31
This series contains updates to the ice driver only.
Brett modifies the driver to allow users to clear a VF's
administratively set MAC address on the PF. Fixes the driver to
recognize an existing VLAN tag when DMAC/SMAC is enabled in a packet.
Fixes an issue, so that VF's are reset after any VF port VLAN
modifications are made on the PF. Made sure the register QRXFLXP_CNTXT
is cleared before writing a new value to ensure the previous value is
not passed forward. Updates the PF to allow the VF to request a reset
as soon as it has been initialized. Fixes an issue to ensure when a VSI
is created, it uses the current coalesce value, not the default value.
Paul allows untrusted VF's to add 16 filters.
Dan increases the timeout needed after a PFR to allow ample time for
package download.
Chinh adjust the define value for the number of PHY speeds we currently
support. Changes the driver to ignore EMODE error when configuring the
PHY.
Jesse fixes an issue which was preventing a user from configuring the
interface before bringing it up.
Henry fixes the logic for adding back perfect flows after flow director
filter does a deletion.
Bruce fixes line wrappings to make it more consistent.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Roopa Prabhu [Sun, 31 May 2020 05:17:20 +0000 (22:17 -0700)]
vxlan: fix dereference of nexthop group in nexthop update path
fix dereference of nexthop group in fdb nexthop group
update validation path.
Fixes: 1274e1cc4226 ("vxlan: ecmp support for mac fdb entries") Reported-by: Ido Schimmel <idosch@idosch.org> Suggested-by: Ido Schimmel <idosch@idosch.org> Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 1 Jun 2020 19:04:32 +0000 (12:04 -0700)]
Merge branch 'dpaa2-eth-add-PFC-support'
Ioana Ciornei says:
====================
dpaa2-eth: add PFC support
This patch set adds support for Priority Flow Control in DPAA2 Ethernet
devices.
The first patch make the necessary changes so that multiple
traffic classes are configured. The dequeue priority
of the maximum 8 traffic classes is configured to be equal.
The second patch adds a static distribution to said traffic
classes based on the VLAN PCP field. In the future, this could be
extended through the .setapp() DCB callback for dynamic configuration.
Also, add support for the congestion group taildrop mechanism that
allows us to control the number of frames that can accumulate on a group
of Rx frame queues belonging to the same traffic class.
The basic subset of the DCB ops is implemented so that the user can
query the number of PFC capable traffic classes, their state and
reconfigure them if necessary.
Changes in v3:
- add patches 6-7 which add the PFC functionality
- patch 2/7: revert to explicitly cast mask to u16 * to not get into
sparse warnings
Changes in v4:
- really fix the sparse warnings in 2/7
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Ciornei [Sat, 30 May 2020 21:08:14 +0000 (00:08 +0300)]
dpaa2-eth: Keep congestion group taildrop enabled when PFC on
Leave congestion group taildrop enabled for all traffic classes
when PFC is enabled. Notification threshold is low enough such
that it will be hit first and this also ensures that FQs on
traffic classes which are not PFC enabled won't drain the buffer
pool.
FQ taildrop threshold is kept disabled as long as any form of
flow control is on. Since FQ taildrop works with bytes, not number
of frames, we can't guarantee it will not interfere with the
congestion notification mechanism for all frame sizes.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Ciornei [Sat, 30 May 2020 21:08:13 +0000 (00:08 +0300)]
dpaa2-eth: Add PFC support through DCB ops
Add support in dpaa2-eth for PFC (Priority Flow Control)
through the DCB ops.
Instruct the hardware to respond to received PFC frames.
Current firmware doesn't allow us to selectively enable PFC
on the Rx side for some priorities only, so we will react to
all incoming PFC frames (and stop transmitting on the traffic
classes specified in the frame).
Also, configure the hardware to generate PFC frames based on Rx
congestion notifications. When a certain number of frames accumulate in
the ingress queues corresponding to a traffic class, priority flow
control frames are generated for that TC.
The number of PFC traffic classes available can be queried through
lldptool. Also, which of those traffic classes have PFC enabled is also
controlled through the same dcbnl_rtnl_ops callbacks.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Radulescu [Sat, 30 May 2020 21:08:12 +0000 (00:08 +0300)]
dpaa2-eth: Update FQ taildrop threshold and buffer pool count
Now that we have congestion group taildrop configured at all
times, we can afford to increase the frame queue taildrop
threshold; this will ensure a better response when receiving
bursts of large-sized frames.
Also decouple the buffer pool count from the Rx FQ taildrop
threshold, as above change would increase it too much. Instead,
keep the old count as a hardcoded value.
With the new limits, we try to ensure that:
* we allow enough leeway for large frame bursts (by buffering
enough of them in queues to avoid heavy dropping in case of
bursty traffic, but when overall ingress bandwidth is manageable)
* allow pending frames to be evenly spread between ingress FQs,
regardless of frame size
* avoid dropping frames due to the buffer pool being empty; this
is not a bad behaviour per se, but system overall response is
more linear and predictable when frames are dropped at frame
queue/group level.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Radulescu [Sat, 30 May 2020 21:08:11 +0000 (00:08 +0300)]
dpaa2-eth: Add congestion group taildrop
The increase in number of ingress frame queues means we now risk
depleting the buffer pool before the FQ taildrop kicks in.
Congestion group taildrop allows us to control the number of frames that
can accumulate on a group of Rx frame queues belonging to the same
traffic class. This setting coexists with the frame queue based
taildrop: whichever limit gets hit first triggers the frame drop.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Radulescu [Sat, 30 May 2020 21:08:10 +0000 (00:08 +0300)]
dpaa2-eth: Add helper functions
Add convenient helper functions that determines whether Rx/Tx pause
frames are enabled based on link state flags received from firmware.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Radulescu [Sat, 30 May 2020 21:08:09 +0000 (00:08 +0300)]
dpaa2-eth: Distribute ingress frames based on VLAN prio
Configure static ingress classification based on VLAN PCP field.
If the DPNI doesn't have enough traffic classes to accommodate all
priority levels, the lowest ones end up on TC 0 (default on miss).
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ioana Radulescu [Sat, 30 May 2020 21:08:08 +0000 (00:08 +0300)]
dpaa2-eth: Add support for Rx traffic classes
The firmware reserves for each DPNI a number of RX frame queues
equal to the number of configured flows x number of configured
traffic classes.
Current driver configuration directs all incoming traffic to
FQs corresponding to TC0, leaving all other priority levels unused.
Start adding support for multiple ingress traffic classes, by
configuring the FQs associated with all priority levels, not just
TC0. All settings that are per-TC, such as those related to
hashing and flow steering, are also updated.
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Don't export __bcm_phy_enable_rdb_access() and
__bcm_phy_enable_legacy_access() functions. They aren't used outside this
module and it was forgotten to provide a prototype for these functions.
Just make them static for now.
Fixes: 11ecf8c55b91 ("net: phy: broadcom: add cable test support") Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Michael Walle <michael@walle.cc> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Willem de Bruijn [Sat, 30 May 2020 19:41:31 +0000 (15:41 -0400)]
tun: correct header offsets in napi frags mode
Tun in IFF_NAPI_FRAGS mode calls napi_gro_frags. Unlike netif_rx and
netif_gro_receive, this expects skb->data to point to the mac layer.
But skb_probe_transport_header, __skb_get_hash_symmetric, and
xdp_do_generic in tun_get_user need skb->data to point to the network
header. Flow dissection also needs skb->protocol set, so
eth_type_trans has to be called.
Ensure the link layer header lies in linear as eth_type_trans pulls
ETH_HLEN. Then take the same code paths for frags as for not frags.
Push the link layer header back just before calling napi_gro_frags.
By pulling up to ETH_HLEN from frag0 into linear, this disables the
frag0 optimization in the special case when IFF_NAPI_FRAGS is used
with zero length iov[0] (and thus empty skb->linear).
Fixes: 90e33d459407 ("tun: enable napi_gro_frags() for TUN/TAP driver") Signed-off-by: Willem de Bruijn <willemb@google.com> Acked-by: Petar Penkov <ppenkov@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Guillaume Nault [Sat, 30 May 2020 18:49:56 +0000 (20:49 +0200)]
cls_flower: remove mpls_opts_policy
Compiling with W=1 gives the following warning:
net/sched/cls_flower.c:731:1: warning: ‘mpls_opts_policy’ defined but not used [-Wunused-const-variable=]
The TCA_FLOWER_KEY_MPLS_OPTS contains a list of
TCA_FLOWER_KEY_MPLS_OPTS_LSE. Therefore, the attributes all have the
same type and we can't parse the list with nla_parse*() and have the
attributes validated automatically using an nla_policy.
fl_set_key_mpls_opts() properly verifies that all attributes in the
list are TCA_FLOWER_KEY_MPLS_OPTS_LSE. Then fl_set_key_mpls_lse()
uses nla_parse_nested() on all these attributes, thus verifying that
they have the NLA_F_NESTED flag. So we can safely drop the
mpls_opts_policy.
Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Guillaume Nault <gnault@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
bridge: mrp: Add support for MRA role
This patch series extends the MRP with the MRA role.
A node that has the MRA role can behave as a MRM or as a MRC. In case there are
multiple nodes in the topology that has the MRA role then only one node can
behave as MRM and all the others need to be have as MRC. The node that has the
higher priority(lower value) will behave as MRM.
A node that has the MRA role and behaves as MRC, it just needs to forward the
MRP_Test frames between the ring ports but also it needs to detect in case it
stops receiving MRP_Test frames. In that case it would try to behave as MRM.
v2:
- add new patch that fixes sparse warnings
- fix parsing of prio attribute
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Horatiu Vultur [Sat, 30 May 2020 18:09:48 +0000 (18:09 +0000)]
bridge: mrp: Add support for role MRA
A node that has the MRA role, it can behave as MRM or MRC.
Initially it starts as MRM and sends MRP_Test frames on both ring ports.
If it detects that there are MRP_Test send by another MRM, then it
checks if these frames have a lower priority than itself. In this case
it would send MRP_Nack frames to notify the other node that it needs to
stop sending MRP_Test frames.
If it receives a MRP_Nack frame then it stops sending MRP_Test frames
and starts to behave as a MRC but it would continue to monitor the
MRP_Test frames send by MRM. If at a point the MRM stops to send
MRP_Test frames it would get the MRM role and start to send MRP_Test
frames.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Horatiu Vultur [Sat, 30 May 2020 18:09:47 +0000 (18:09 +0000)]
bridge: mrp: Set the priority of MRP instance
Each MRP instance has a priority, a lower value means a higher priority.
The priority of MRP instance is stored in MRP_Test frame in this way
all the MRP nodes in the ring can see other nodes priority.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Jia-Ju Bai [Sat, 30 May 2020 02:41:50 +0000 (10:41 +0800)]
net: vmxnet3: fix possible buffer overflow caused by bad DMA value in vmxnet3_get_rss()
The value adapter->rss_conf is stored in DMA memory, and it is assigned
to rssConf, so rssConf->indTableSize can be modified at anytime by
malicious hardware. Because rssConf->indTableSize is assigned to n,
buffer overflow may occur when the code "rssConf->indTable[n]" is
executed.
To fix this possible bug, n is checked after being used.
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Arnd Bergmann [Fri, 29 May 2020 20:13:58 +0000 (22:13 +0200)]
flow_dissector: work around stack frame size warning
The fl_flow_key structure is around 500 bytes, so having two of them
on the stack in one function now exceeds the warning limit after an
otherwise correct change:
net/sched/cls_flower.c:298:12: error: stack frame size of 1056 bytes in function 'fl_classify' [-Werror,-Wframe-larger-than=]
I suspect the fl_classify function could be reworked to only have one
of them on the stack and modify it in place, but I could not work out
how to do that.
As a somewhat hacky workaround, move one of them into an out-of-line
function to reduce its scope. This does not necessarily reduce the stack
usage of the outer function, but at least the second copy is removed
from the stack during most of it and does not add up to whatever is
called from there.
I now see 552 bytes of stack usage for fl_classify(), plus 528 bytes
for fl_mask_lookup().
Fixes: 58cff782cc55 ("flow_dissector: Parse multiple MPLS Label Stack Entries") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Cong Wang <xiyou.wangcong@gmail.com> Acked-by: Guillaume Nault <gnault@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Roelof Berg [Fri, 29 May 2020 19:30:02 +0000 (21:30 +0200)]
lan743x: Added fixed link and RGMII support
Microchip lan7431 is frequently connected to a phy. However, it
can also be directly connected to a MII remote peer without
any phy in between. For supporting such a phyless hardware setup
in Linux we utilized phylib, which supports a fixed-link
configuration via the device tree. And we added support for
defining the connection type R/GMII in the device tree.
New behavior:
-------------
. The automatic speed and duplex detection of the lan743x silicon
between mac and phy is disabled. Instead phylib is used like in
other typical Linux drivers. The usage of phylib allows to
specify fixed-link parameters in the device tree.
. The device tree entry phy-connection-type is supported now with
the modes RGMII or (G)MII (default).
Development state:
------------------
. Tested with fixed-phy configurations. Not yet tested in normal
configurations with phy. Microchip kindly offered testing
as soon as the Corona measures allow this.
====================
devlink: Add support for control packet traps
So far device drivers were only able to register drop and exception
packet traps with devlink. These traps are used for packets that were
either dropped by the underlying device or encountered an exception
(e.g., missing neighbour entry) during forwarding.
However, in the steady state, the majority of the packets being trapped
to the CPU are packets that are required for the correct functioning of
the control plane. For example, ARP request and IGMP query packets.
This patch set allows device drivers to register such control traps with
devlink and expose their default control plane policy to user space.
User space can then tune the packet trap policer settings according to
its needs, as with existing packet traps.
In a similar fashion to exception traps, the action associated with such
traps cannot be changed as it can easily break the control plane. Unlike
drop and exception traps, packets trapped via control traps are not
reported to the kernel's drop monitor as they are not indicative of any
problem.
Patch set overview:
Patches #1-#3 break out layer 3 exceptions to a different group to
provide better granularity. A future patch set will make this completely
configurable.
Patch #4 adds a new trap action ('mirror') that is used for packets that
are forwarded by the device and sent to the CPU. Such packets are marked
by device drivers with 'skb->offload_fwd_mark = 1' in order to prevent
the kernel from forwarding them again.
Patch #5 adds the new trap type, 'control'.
Patches #6-#8 gradually add various control traps to devlink with proper
documentation.
Patch #9 adds a few control traps to netdevsim, which are automatically
exercised by existing devlink-trap selftest.
Patches #10 performs small refactoring in mlxsw.
Patches #11-#13 change mlxsw to register its existing control traps with
devlink.
Patch #14 adds a selftest over mlxsw that exercises all the registered
control traps.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:45 +0000 (21:36 +0300)]
mlxsw: spectrum_trap: Factor out common Rx listener function
We currently have an Rx listener function for exception traps that marks
received skbs with 'offload_fwd_mark' and injects them to the kernel's
Rx path. The marking is done because all these exceptions occur during
L3 forwarding, after the packets were potentially flooded at L2.
A subsequent patch will add support for control traps. Packets received
via some of these control traps need different handling:
1. Packets might not need to be marked with 'offload_fwd_mark'. For
example, if packet was trapped before L2 forwarding
2. Packets might not need to be injected to the kernel's Rx path. For
example, sampled packets are reported to user space via the psample
module
Factor out a common Rx listener function that only reports trapped
packets to devlink. Call it from mlxsw_sp_rx_no_mark_listener() and
mlxsw_sp_rx_mark_listener() that will inject the packets to the kernel's
Rx path, without and with the marking, respectively.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:44 +0000 (21:36 +0300)]
netdevsim: Register control traps
Register two control traps with devlink. The existing selftest at
tools/testing/selftests/drivers/net/netdevsim/devlink_trap.sh iterates
over all registered traps and checks that the action of non-drop traps
cannot be changed. Up until now only exception traps were tested, now
control traps will be tested as well.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:43 +0000 (21:36 +0300)]
devlink: Add ACL control packet traps
Add packet traps for packets that are sampled / trapped by ACLs, so that
capable drivers could register them with devlink. Add documentation for
every added packet trap and packet trap group.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:42 +0000 (21:36 +0300)]
devlink: Add layer 3 control packet traps
Add layer 3 control packet traps such as ARP and DHCP, so that capable
device drivers could register them with devlink. Add documentation for
every added packet trap and packet trap group.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:41 +0000 (21:36 +0300)]
devlink: Add layer 2 control packet traps
Add layer 2 control packet traps such as STP and IGMP query, so that
capable device drivers could register them with devlink. Add
documentation for every added packet trap and packet trap group.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:39 +0000 (21:36 +0300)]
devlink: Add 'mirror' trap action
The action is used by control traps such as IGMP query. The packet is
flooded by the device, but also trapped to the CPU in order for the
software bridge to mark the receiving port as a multicast router port.
Such packets are marked with 'skb->offload_fwd_mark = 1' in order to
prevent the software bridge from flooding them again.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Fri, 29 May 2020 18:36:36 +0000 (21:36 +0300)]
devlink: Create dedicated trap group for layer 3 exceptions
Packets that hit exceptions during layer 3 forwarding must be trapped to
the CPU for the control plane to function properly. Create a dedicated
group for them, so that user space could choose to assign a different
policer for them.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Reviewed-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
The removal of mips_swiotlb_ops exposed a problem in octeon_mgmt Ethernet
driver. mips_swiotlb_ops had an mb() after most of the operations and the
removal of the ops had broken the receive functionality of the driver.
My code inspection has shown no other places except
octeon_mgmt_rx_fill_ring() where an explicit barrier would be obviously
missing. The latter function however has to make sure that "ringing the
bell" doesn't happen before RX ring entry is really written.
The patch has been successfully tested on Octeon II.
Fixes: a999933db9ed ("MIPS: remove mips_swiotlb_ops") Cc: stable@vger.kernel.org Signed-off-by: Alexander Sverdlin <alexander.sverdlin@nokia.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
the indirect flow_block infrastructure, revisited
This series fixes b5140a36da78 ("netfilter: flowtable: add indr block
setup support") that adds support for the indirect block for the
flowtable. This patch crashes the kernel with the TC CT action.
A workaround patch to cure this crash has been proposed. However, there
is another problem: The indirect flow_block still does not work for the
new TC CT action. The problem is that the existing flow_indr_block_entry
callback assumes you can look up for the flowtable from the netdevice to
get the flow_block. This flow_block allows you to offload the flows via
TC_SETUP_CLSFLOWER. Unfortunately, it is not possible to get the
flow_block from the TC CT flowtables because they are _not_ bound to any
specific netdevice.
= What is the indirect flow_block infrastructure?
The indirect flow_block infrastructure allows drivers to offload
tc/netfilter rules that belong to software tunnel netdevices, e.g.
vxlan.
This indirect flow_block infrastructure relates tunnel netdevices with
drivers because there is no obvious way to relate these two things
from the control plane.
= How does the indirect flow_block work before this patchset?
Front-ends register the indirect block callback through
flow_indr_add_block_cb() if they support for offloading tunnel
netdevices.
== Setting up an indirect block
1) Drivers track tunnel netdevices via NETDEV_{REGISTER,UNREGISTER} events.
If there is a new tunnel netdevice that the driver can offload, then the
driver invokes __flow_indr_block_cb_register() with the new tunnel
netdevice and the driver callback. The __flow_indr_block_cb_register()
call iterates over the list of the front-end callbacks.
2) The front-end callback sets up the flow_block_offload structure and it
invokes the driver callback to set up the flow_block.
3) The driver callback now registers the flow_block structure and it
returns the flow_block back to the front-end.
4) The front-end gets the flow_block object and it is now ready to
offload rules for this tunnel netdevice.
There are two possibilities, either tunnel netdevice is removed or
a netdevice (port representor) is removed.
=== Tunnel netdevice is removed
Driver waits for the NETDEV_UNREGISTER event that announces the tunnel
netdevice removal. Then, it calls __flow_indr_block_cb_unregister() to
remove the flow_block and rules. Callgraph is very similar to the one
described above.
=== Netdevice is removed (port representor)
Driver calls __flow_indr_block_cb_unregister() to remove the existing
netfilter/tc rule that belong to the tunnel netdevice.
= How does the indirect flow_block work after this patchset?
Drivers register the indirect flow_block setup callback through
flow_indr_dev_register() if they support for offloading tunnel
netdevices.
== Setting up an indirect flow_block
1) Frontends check if dev->netdev_ops->ndo_setup_tc is unset. If so,
frontends call flow_indr_dev_setup_offload(). This call invokes
the drivers' indirect flow_block setup callback.
2) The indirect flow_block setup callback sets up a flow_block structure
which relates the tunnel netdevice and the driver.
3) The front-end uses flow_block and offload the rules.
Note that the operational to set up (non-indirect) flow_block is very
similar.
== Releasing the indirect flow_block
=== Tunnel netdevice is removed
This calls flow_indr_dev_setup_offload() to set down the flow_block and
remove the offloaded rules. This alternate path is exercised if
dev->netdev_ops->ndo_setup_tc is unset.
=== Netdevice is removed (port representor)
If a netdevice is removed, then it might need to to clean up the
offloaded tc/netfilter rules that belongs to the tunnel netdevice:
1) The driver invokes flow_indr_dev_unregister() when a netdevice is
removed.
2) This call iterates over the existing indirect flow_blocks
and it invokes the cleanup callback to let the front-end remove the
tc/netfilter rules. The cleanup callback already provides the
flow_block that the front-end needs to clean up.
Front-end Driver
|
flow_indr_dev_unregister(...)
|
iterate over list of indirect flow_block
and invoke cleanup callback
|
.-----------------------------
|
.
frontend_flow_block_cleanup(flow_block)
.
|
\/
remove rules to flow_block
TC_SETUP_CLSFLOWER
= About this patchset
This patchset aims to address the existing TC CT problem while
simplifying the indirect flow_block infrastructure. Saving 300 LoC in
the flow_offload core and the drivers. The operational gets aligned with
the (non-indirect) flow_blocks logic. Patchset is composed of:
Patch #1 add nf_flow_table_gc_cleanup() which is required by the
netfilter's flowtable new indirect flow_block approach.
Patch #2 adds the flow_block_indr object which is actually part of
of the flow_block object. This stores the indirect flow_block
metadata such as the tunnel netdevice owner and the cleanup
callback (in case the tunnel netdevice goes away).
This patch adds flow_indr_dev_{un}register() to allow drivers
to offer netdevice tunnel hardware offload to the front-ends.
Then, front-ends call flow_indr_dev_setup_offload() to invoke
the drivers to set up the (indirect) flow_block.
Patch #3 add the tcf_block_offload_init() helper function, this is
a preparation patch to adapt the tc front-end to use this
new indirect flow_block infrastructure.
Patch #4 updates the tc and netfilter front-ends to use the new
indirect flow_block infrastructure.
Patch #5 updates the mlx5 driver to use the new indirect flow_block
infrastructure.
Patch #6 updates the nfp driver to use the new indirect flow_block
infrastructure.
Patch #7 updates the bnxt driver to use the new indirect flow_block
infrastructure.
Patch #8 removes the indirect flow_block infrastructure version 1,
now that frontends and drivers have been translated to
version 2 (coming in this patchset).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Drivers do not register to netdev events to set up indirect blocks
anymore. Remove __flow_indr_block_cb_register() and
__flow_indr_block_cb_unregister().
The frontends set up the callbacks through flow_indr_dev_setup_block()
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Register ndo callback via flow_indr_dev_register() and
flow_indr_dev_unregister().
No need for mlx5e_rep_indr_clean_block_privs() since flow_block_cb_free()
already releases the internal mapping via ->release callback, which in
this case is mlx5e_rep_indr_tc_block_unbind().
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: David S. Miller <davem@davemloft.net>
projects / mirror_ubuntu-kernels.git / log