Lawrence Brakmo says:
====================
Host Bandwidth Manager is a framework for limiting the bandwidth used
by v2 cgroups. It consists of 1 BPF helper, a sample BPF program to
limit egress bandwdith as well as a sample user program and script to
simplify HBM testing.
The sample HBM BPF program is not meant to be production quality, it is
provided as proof of concept. A lot more information, including sample
runs in some cases, are provided in the commit messages of the individual
patches.
A future patch will add support for reducing TCP's cwnd (we are evaluating
alternatives). Another patch will add support for fair queueing's Earliest
Departure Time. Until then, HBM is better suited for flows supporitng ECN.
In addition, A BPF program to limit ingress bandwidth will be provided in
an upcomming patchset.
Changes from v1 to v2:
* bpf_tcp_enter_cwr can only be called from a cgroup skb egress BPF
program (otherwise load or attach will fail) where we already hold
the sk lock. Also only applies for ESTABLISHED state.
* bpf_skb_ecn_set_ce uses INET_ECN_set_ce()
* bpf_tcp_check_probe_timer now uses tcp_reset_xmit_timer. Can only be
used by egress cgroup skb programs.
* removed load_cg_skb user program.
* nrm bpf egress program checks packet header in skb to determine
ECN value. Now also works for ECN enabled UDP packets.
Using ECN_ defines instead of integers.
* NRM script test program now uses bpftool instead of load_cg_skb
Changes from v2 to v3:
* Changed name from NRM (Network Resource Manager) to HBM (Host
Bandwdith Manager)
* The bpf helper to set ECN ce now checks that the header is writeable
* Removed helper bpf functions that modified TCP state due to a concern
about whether the socket is locked by the current thread.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Script for testing HBM (Host Bandwidth Manager) framework.
It creates a cgroup to use for testing and load a BPF program to limit
egress bandwidht. It then uses iperf3 or netperf to create
loads. The output is the goodput in Mbps (unless -D is used).
It can work on a single host using loopback or among two hosts (with netperf).
When using loopback, it is recommended to also introduce a delay of at least
1ms (-d=1), otherwise the assigned bandwidth is likely to be underutilized.
USAGE: $name [out] [-b=<prog>|--bpf=<prog>] [-c=<cc>|--cc=<cc>] [-D]
[-d=<delay>|--delay=<delay>] [--debug] [-E]
[-f=<#flows>|--flows=<#flows>] [-h] [-i=<id>|--id=<id >] [-l]
[-N] [-p=<port>|--port=<port>] [-P] [-q=<qdisc>]
[-R] [-s=<server>|--server=<server] [--stats]
[-t=<time>|--time=<time>] [-w] [cubic|dctcp]
Where:
out Egress (default egress)
-b or --bpf BPF program filename to load and attach.
Default is nrm_out_kern.o for egress,
-c or -cc TCP congestion control (cubic or dctcp)
-d or --delay Add a delay in ms using netem
-D In addition to the goodput in Mbps, it also outputs
other detailed information. This information is
test dependent (i.e. iperf3 or netperf).
--debug Print BPF trace buffer
-E Enable ECN (not required for dctcp)
-f or --flows Number of concurrent flows (default=1)
-i or --id cgroup id (an integer, default is 1)
-l Do not limit flows using loopback
-N Use netperf instead of iperf3
-h Help
-p or --port iperf3 port (default is 5201)
-P Use an iperf3 instance for each flow
-q Use the specified qdisc.
-r or --rate Rate in Mbps (default 1s 1Gbps)
-R Use TCP_RR for netperf. 1st flow has req
size of 10KB, rest of 1MB. Reply in all
cases is 1 byte.
More detailed output for each flow can be found
in the files netperf.<cg>.<flow>, where <cg> is the
cgroup id as specified with the -i flag, and <flow>
is the flow id starting at 1 and increasing by 1 for
flow (as specified by -f).
-s or --server hostname of netperf server. Used to create netperf
test traffic between to hosts (default is within host)
netserver must be running on the host.
--stats Get HBM stats (marked, dropped, etc.)
-t or --time duration of iperf3 in seconds (default=5)
-w Work conserving flag. cgroup can increase its
bandwidth beyond the rate limit specified
while there is available bandwidth. Current
implementation assumes there is only one NIC
(eth0), but can be extended to support multiple
NICs. This is just a proof of concept.
cubic or dctcp specify TCP CC to use
Examples:
./do_hbm_test.sh -l -d=1 -D --stats
Runs a 5 second test, using a single iperf3 flow and with the default
rate limit of 1Gbps and a delay of 1ms (using netem) using the default
TCP congestion control on the loopback device (hence we use "-l" to
enforce bandwidth limit on loopback device). Since no direction is
specified, it defaults to egress. Since no TCP CC algorithm is
specified it uses the system default (Cubic for this test).
With no -D flag, only the value of the AGGREGATE OUTPUT would show.
id refers to the cgroup id and is useful when running multi cgroup
tests (supported by a future patch).
This patchset does not support calling TCP's congesion window
reduction, even when packets are dropped by the BPF program, resulting
in a large number of packets dropped. It is recommended that the current
HBM implemenation only be used with ECN enabled flows. A future patch
will add support for reducing TCP's cwnd and will increase the
performance of non-ECN enabled flows.
Output:
Details for HBM in cgroup 1
id:1
rate_mbps:493
duration:4.8 secs
packets:11355
bytes_MB:590
pkts_dropped:4497
bytes_dropped_MB:292
pkts_marked_percent: 39.60
bytes_marked_percent: 49.49
pkts_dropped_percent: 39.60
bytes_dropped_percent: 49.49
PING AVG DELAY:2.075
AGGREGATE_GOODPUT:505
./do_nrm_test.sh -l -d=1 -D --stats dctcp
Same as above but using dctcp. Note that fewer bytes are dropped
(0.01% vs. 49%).
Output:
Details for HBM in cgroup 1
id:1
rate_mbps:945
duration:4.9 secs
packets:16859
bytes_MB:578
pkts_dropped:1
bytes_dropped_MB:0
pkts_marked_percent: 28.74
bytes_marked_percent: 45.15
pkts_dropped_percent: 0.01
bytes_dropped_percent: 0.01
PING AVG DELAY:2.083
AGGREGATE_GOODPUT:965
./do_nrm_test.sh -d=1 -D --stats
As first example, but without limiting loopback device (i.e. no
"-l" flag). Since there is no bandwidth limiting, no details for
HBM are printed out.
Output:
Details for HBM in cgroup 1
PING AVG DELAY:2.019
AGGREGATE_GOODPUT:42655
./do_hbm.sh -l -d=1 -D --stats -f=2
Uses iper3 and does 2 flows
./do_hbm.sh -l -d=1 -D --stats -f=4 -P
Uses iperf3 and does 4 flows, each flow as a separate process.
./do_hbm.sh -l -d=1 -D --stats -f=4 -N
Uses netperf, 4 flows
./do_hbm.sh -f=1 -r=2000 -t=5 -N -D --stats dctcp -s=<server-name>
Uses netperf between two hosts. The remote host name is specified
with -s= and you need to start the program netserver manually on
the remote host. It will use 1 flow, a rate limit of 2Gbps and dctcp.
./do_hbm.sh -f=1 -r=2000 -t=5 -N -D --stats -w dctcp \
-s=<server-name>
As previous, but allows use of extra bandwidth. For this test the
rate is 8Gbps vs. 1Gbps of the previous test.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The program nrm creates a cgroup and attaches a BPF program to the
cgroup for testing HBM (Host Bandwidth Manager) for egress traffic.
One still needs to create network traffic. This can be done through
netesto, netperf or iperf3.
A follow-up patch contains a script to create traffic.
USAGE: hbm [-d] [-l] [-n <id>] [-r <rate>] [-s] [-t <secs>]
[-w] [-h] [prog]
Where:
-d Print BPF trace debug buffer
-l Also limit flows doing loopback
-n <#> To create cgroup "/hbm#" and attach prog. Default is /nrm1
This is convenient when testing HBM in more than 1 cgroup
-r <rate> Rate limit in Mbps
-s Get HBM stats (marked, dropped, etc.)
-t <time> Exit after specified seconds (deault is 0)
-w Work conserving flag. cgroup can increase its bandwidth
beyond the rate limit specified while there is available
bandwidth. Current implementation assumes there is only
NIC (eth0), but can be extended to support multiple NICs.
Currrently only supported for egress. Note, this is just
a proof of concept.
-h Print this info
prog BPF program file name. Name defaults to hbm_out_kern.o
More information about HBM can be found in the paper "BPF Host Resource
Management" presented at the 2018 Linux Plumbers Conference, Networking Track
(http://vger.kernel.org/lpc_net2018_talks/LPC%20BPF%20Network%20Resource%20Paper.pdf)
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
A cgroup skb BPF program to limit cgroup output bandwidth.
It uses a modified virtual token bucket queue to limit average
egress bandwidth. The implementation uses credits instead of tokens.
Negative credits imply that queueing would have happened (this is
a virtual queue, so no queueing is done by it. However, queueing may
occur at the actual qdisc (which is not used for rate limiting).
This implementation uses 3 thresholds, one to start marking packets and
the other two to drop packets:
CREDIT
- <--------------------------|------------------------> +
| | | 0
| Large pkt |
| drop thresh |
Small pkt drop Mark threshold
thresh
The effect of marking depends on the type of packet:
a) If the packet is ECN enabled, then the packet is ECN ce marked.
The current mark threshold is tuned for DCTCP.
c) Else, it is dropped if it is a large packet.
If the credit is below the drop threshold, the packet is dropped.
Note that dropping a packet through the BPF program does not trigger CWR
(Congestion Window Reduction) in TCP packets. A future patch will add
support for triggering CWR.
This BPF program actually uses 2 drop thresholds, one threshold
for larger packets (>= 120 bytes) and another for smaller packets. This
protects smaller packets such as SYNs, ACKs, etc.
The default bandwidth limit is set at 1Gbps but this can be changed by
a user program through a shared BPF map. In addition, by default this BPF
program does not limit connections using loopback. This behavior can be
overwritten by the user program. There is also an option to calculate
some statistics, such as percent of packets marked or dropped, which
the user program can access.
A latter patch provides such a program (hbm.c)
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch syncs the uapi bpf.h to tools/ and also updates
bpf_herlpers.h in tools/
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch adds a new bpf helper BPF_FUNC_skb_ecn_set_ce
"int bpf_skb_ecn_set_ce(struct sk_buff *skb)". It is added to
BPF_PROG_TYPE_CGROUP_SKB typed bpf_prog which currently can
be attached to the ingress and egress path. The helper is needed
because his type of bpf_prog cannot modify the skb directly.
This helper is used to set the ECN field of ECN capable IP packets to ce
(congestion encountered) in the IPv6 or IPv4 header of the skb. It can be
used by a bpf_prog to manage egress or ingress network bandwdith limit
per cgroupv2 by inducing an ECN response in the TCP sender.
This works best when using DCTCP.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Compiling xdpsock_user.c with 4.8.5, I hit the following
compilation warning:
HOSTCC samples/bpf/xdpsock_user.o
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c: In function ‘main’:
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c:449:6: warning: ‘idx_cq’ may be used unini
tialized in this function [-Wmaybe-uninitialized]
u32 idx_cq, idx_fq;
^
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c:606:7: warning: ‘idx_rx’ may be used unini
tialized in this function [-Wmaybe-uninitialized]
u32 idx_rx, idx_tx = 0;
^
/data/users/yhs/work/net-next/samples/bpf/xdpsock_user.c:506:6: warning: ‘idx_rx’ may be used unini
tialized in this function [-Wmaybe-uninitialized]
u32 idx_rx, idx_fq = 0;
As an example, the code pattern looks like:
u32 idx_cq;
...
ret = xsk_ring_prod__reserve(&xsk->umem->fq, rcvd, &idx_fq);
if (ret) {
...
}
... idx_fq ...
The compiler warns since it does not know whether &idx_fq is assigned
or not inside the library function xsk_ring_prod__reserve().
Let us assign an initial value 0 to such auto variables to silence
compiler warning.
Fixes: 248c7f9c0e ("samples/bpf: convert xdpsock to use libbpf for AF_XDP access")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This is to avoid permission denied error. A lot of systems
may have a much lower number, e.g., 64KB, for RLIMIT_MEMLOCK,
which may not be sufficient for the test to run successfully.
Fixes: e0b27b3f97 ("bpf: Add test_sock_fields for skb->sk and bpf_tcp_sock")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Andrii Nakryiko says:
====================
A bunch of BPF-related docs typo, wording and formatting fixes.
v1->v2:
- split off non-documentation changes into separate patchset
====================
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reflow paragraphs to more fully and evenly fill 78 character lines.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Fix various typos, some of the formatting and wording for
Documentation/btf.rst.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
We need to iterate through all possible cpus.
Fixes: 492ecee892 ("bpf: enable program stats")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Andrii Nakryiko says:
====================
This patchset fixes a bug in btf_dedup() algorithm, which under specific
hash collision causes infinite loop. It also exposes ability to tune BTF
deduplication table size, with double purpose of allowing applications to
adjust size according to the size of BTF data, as well as allowing a simple
way to force hash collisions by setting table size to 1.
- Patch #1 fixes bug in btf_dedup testing code that's checking strings
- Patch #2 fixes pointer arg formatting in btf.h
- Patch #3 adds option to specify custom dedup table size
- Patch #4 fixes aforementioned bug in btf_dedup
- Patch #5 adds test that validates the fix
v1->v2:
- remove "Fixes" from formatting change patch
- extract roundup_pow2_max func for dedup table size
- btf_equal_struct -> btf_shallow_equal_struct
- explain in comment why we can't rely on just btf_dedup_is_equiv
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch adds a btf_dedup test exercising logic of STRUCT<->FWD
resolution and validating that STRUCT is not resolved to a FWD. It also
forces hash collisions, forcing both FWD and STRUCT to be candidates for
each other. Previously this condition caused infinite loop due to FWD
pointing to STRUCT and STRUCT pointing to its FWD.
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When checking available canonical candidates for struct/union algorithm
utilizes btf_dedup_is_equiv to determine if candidate is suitable. This
check is not enough when candidate is corresponding FWD for that
struct/union, because according to equivalence logic they are
equivalent. When it so happens that FWD and STRUCT/UNION end in hashing
to the same bucket, it's possible to create remapping loop from FWD to
STRUCT and STRUCT to same FWD, which will cause btf_dedup() to loop
forever.
This patch fixes the issue by additionally checking that type and
canonical candidate are strictly equal (utilizing btf_equal_struct).
Fixes: d5caef5b56 ("btf: add BTF types deduplication algorithm")
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Default size of dedup table (16k) is good enough for most binaries, even
typical vmlinux images. But there are cases of binaries with huge amount
of BTF types (e.g., allyesconfig variants of kernel), which benefit from
having bigger dedup table size to lower amount of unnecessary hash
collisions. Tools like pahole, thus, can tune this parameter to reach
optimal performance.
This change also serves double purpose of allowing tests to force hash
collisions to test some corner cases, used in follow up patch.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Fix invalid formatting of pointer arg.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
btf_dedup testing code doesn't account for length of struct btf_header
when calculating the start of a string section. This patch fixes this
problem.
Fixes: 49b57e0d01 ("tools/bpf: remove btf__get_strings() superseded by raw data API")
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The "ref_type_id" variable needs to be signed for the error handling
to work.
Fixes: d5caef5b56 ("btf: add BTF types deduplication algorithm")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Jakub Kicinski says:
====================
This set is next part of a quest to get rid of the bpf_load
ELF loader. It fixes some minor issues with the samples and
starts the conversion.
First patch fixes ping invocations, ping localhost defaults
to IPv6 on modern setups. Next load_sock_ops sample is removed
and users are directed towards using bpftool directly.
Patch 4 removes the use of bpf_load from samples which don't
need the auto-attachment functionality at all.
Patch 5 improves symbol counting in libbpf, it's not currently
an issue but it will be when anyone adds a symbol with a long
name. Let's make sure that person doesn't have to spend time
scratching their head and wondering why .a and .so symbol
counts don't match.
v2: - specify prog_type where possible (Andrii).
====================
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
readelf truncates its output by default to attempt to make it more
readable. This can lead to function names getting aliased if they
differ late in the string. Use --wide parameter to avoid
truncation.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Some samples don't really need the magic of bpf_load,
switch them to libbpf.
v2: - specify program types.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For historical reasons the helper to loop over maps in an object
is called bpf_map__for_each while it really should be called
bpf_object__for_each_map. Rename and add a correctly named
define for backward compatibility.
Switch all in-tree users to the correct name (Quentin).
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
bpftool can do all the things load_sock_ops used to do, and more.
Point users to bpftool instead of maintaining this sample utility.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
ping localhost may default of IPv6 on modern systems, but
samples are trying to only parse IPv4. Force IPv4.
samples/bpf/tracex1_user.c doesn't interpret the packet so
we don't care which IP version will be used there.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Older GCC (<4.8) isn't smart enough to optimize !__builtin_constant_p()
branch in bpf_htons.
I recently fixed it for pkt_v4 and pkt_v6 in commit a0517a0f7e
("selftests/bpf: use __bpf_constant_htons in test_prog.c"), but
later added another bunch of bpf_htons in commit bf0f0fd939
("selftests/bpf: add simple BPF_PROG_TEST_RUN examples for flow
dissector").
Fixes: bf0f0fd939 ("selftests/bpf: add simple BPF_PROG_TEST_RUN examples for flow dissector")
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This header defines the BPF functions enumerated in uapi/linux.bpf.h
in a callable format. Expand to include all registered functions.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
wrap bpf_stats_enabled sysctl with #ifdef
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 492ecee892 ("bpf: enable program stats")
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Yonghong Song says:
====================
The inner_map_meta->spin_lock_off is not set correctly during
map creation for BPF_MAP_TYPE_ARRAY_OF_MAPS and BPF_MAP_TYPE_HASH_OF_MAPS.
This may lead verifier error due to misinformation.
This patch set fixed the issue with Patch #1 for the kernel change
and Patch #2 for enhanced selftest test_maps.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit d83525ca62 ("bpf: introduce bpf_spin_lock")
introduced bpf_spin_lock and the field spin_lock_off
in kernel internal structure bpf_map has the following
meaning:
>=0 valid offset, <0 error
For every map created, the kernel will ensure
spin_lock_off has correct value.
Currently, bpf_map->spin_lock_off is not copied
from the inner map to the map_in_map inner_map_meta
during a map_in_map type map creation, so
inner_map_meta->spin_lock_off = 0.
This will give verifier wrong information that
inner_map has bpf_spin_lock and the bpf_spin_lock
is defined at offset 0. An access to offset 0
of a value pointer will trigger the following error:
bpf_spin_lock cannot be accessed directly by load/store
This patch fixed the issue by copy inner map's spin_lock_off
value to inner_map_meta->spin_lock_off.
Fixes: d83525ca62 ("bpf: introduce bpf_spin_lock")
Signed-off-by: Yonghong Song <yhs@fb.com>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, running sample "task_fd_query" and "tracex3" occurs the
following error. On kernel v5.0-rc* this sample will be unavailable
due to the removal of function 'blk_start_request' at commit "a1ce35f".
(function removed, as "Single Queue IO scheduler" no longer exists)
$ sudo ./task_fd_query
failed to create kprobe 'blk_start_request' error 'No such file or
directory'
This commit will change the function 'blk_start_request' to
'blk_mq_start_request' to fix the broken sample.
Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Alexei Starovoitov says:
====================
Introduce per program stats to monitor the usage BPF.
v2->v3:
- rename to run_time_ns/run_cnt everywhere
v1->v2:
- fixed u64 stats on 32-bit archs. Thanks Eric
- use more verbose run_time_ns in json output as suggested by Andrii
- refactored prog_alloc and clarified behavior of stats in subprogs
====================
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Return bpf program run_time_ns and run_cnt via bpf_prog_info
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
JITed BPF programs are indistinguishable from kernel functions, but unlike
kernel code BPF code can be changed often.
Typical approach of "perf record" + "perf report" profiling and tuning of
kernel code works just as well for BPF programs, but kernel code doesn't
need to be monitored whereas BPF programs do.
Users load and run large amount of BPF programs.
These BPF stats allow tools monitor the usage of BPF on the server.
The monitoring tools will turn sysctl kernel.bpf_stats_enabled
on and off for few seconds to sample average cost of the programs.
Aggregated data over hours and days will provide an insight into cost of BPF
and alarms can trigger in case given program suddenly gets more expensive.
The cost of two sched_clock() per program invocation adds ~20 nsec.
Fast BPF progs (like selftests/bpf/progs/test_pkt_access.c) will slow down
from ~10 nsec to ~30 nsec.
static_key minimizes the cost of the stats collection.
There is no measurable difference before/after this patch
with kernel.bpf_stats_enabled=0
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Magnus Karlsson says:
====================
This patch proposes to add AF_XDP support to libbpf. The main reason
for this is to facilitate writing applications that use AF_XDP by
offering higher-level APIs that hide many of the details of the AF_XDP
uapi. This is in the same vein as libbpf facilitates XDP adoption by
offering easy-to-use higher level interfaces of XDP
functionality. Hopefully this will facilitate adoption of AF_XDP, make
applications using it simpler and smaller, and finally also make it
possible for applications to benefit from optimizations in the AF_XDP
user space access code. Previously, people just copied and pasted the
code from the sample application into their application, which is not
desirable.
The proposed interface is composed of two parts:
* Low-level access interface to the four rings and the packet
* High-level control plane interface for creating and setting up umems
and AF_XDP sockets. This interface also loads a simple XDP program
that routes all traffic on a queue up to the AF_XDP socket.
The sample program has been updated to use this new interface and in
that process it lost roughly 300 lines of code. I cannot detect any
performance degradations due to the use of this library instead of the
previous functions that were inlined in the sample application. But I
did measure this on a slower machine and not the Broadwell that we
normally use.
The rings are now called xsk_ring and when a producer operates on
it. It is xsk_ring_prod and for a consumer it is xsk_ring_cons. This
way we can get some compile time error checking that the rings are
used correctly.
Comments and contenplations:
* The current behaviour is that the library loads an XDP program (if
requested to do so) but the clean up of this program is left to the
application. It would be possible to implement this cleanup in the
library, but it would require state to be kept on netdev level,
which there is none at the moment, and the synchronization of this
between processes. All this adding complexity. But when we get an
XDP program per queue id, then it becomes trivial to also remove the
XDP program when the application exits. This proposal from Jesper,
Björn and others will also improve the performance of libbpf, since
most of the XDP program code can be removed when that feature is
supported.
* In a future release, I am planning on adding a higher level data
plane interface too. This will be based around recvmsg and sendmsg
with the use of struct iovec for batching, without the user having
to know anything about the underlying four rings of an AF_XDP
socket. There will be one semantic difference though from the
standard recvmsg and that is that the kernel will fill in the iovecs
instead of the application. But the rest should be the same as the
libc versions so that application writers feel at home.
Patch 1: adds AF_XDP support in libbpf
Patch 2: updates the xdpsock sample application to use the libbpf functions
Patch 3: Documentation update to help first time users
Changes v5 to v6:
* Fixed prog_fd bug found by Xiaolong Ye. Thanks!
Changes v4 to v5:
* Added a FAQ to the documentation
* Removed xsk_umem__get_data and renamed xsk_umem__get_dat_raw to
xsk_umem__get_data
* Replaced the netlink code with bpf_get_link_xdp_id()
* Dynamic allocation of the map sizes. They are now sized after
the max number of queueus on the netdev in question.
Changes v3 to v4:
* Dropped the pr_*() patch in favor of Yonghong Song's patch set
* Addressed the review comments of Daniel Borkmann, mainly leaking
of file descriptors at clean up and making the data plane APIs
all static inline (with the exception of xsk_umem__get_data that
uses an internal structure I do not want to expose).
* Fixed the netlink callback as suggested by Maciej Fijalkowski.
* Removed an unecessary include in the sample program as spotted by
Ilia Fillipov.
Changes v2 to v3:
* Added automatic loading of a simple XDP program that routes all
traffic on a queue up to the AF_XDP socket. This program loading
can be disabled.
* Updated function names to be consistent with the libbpf naming
convention
* Moved all code to xsk.[ch]
* Removed all the XDP program loading code from the sample since
this is now done by libbpf
* The initialization functions now return a handle as suggested by
Alexei
* const statements added in the API where applicable.
Changes v1 to v2:
* Fixed cleanup of library state on error.
* Moved API to initial version
* Prefixed all public functions by xsk__ instead of xsk_
* Added comment about changed default ring sizes, batch size and umem
size in the sample application commit message
* The library now only creates an Rx or Tx ring if the respective
parameter is != NULL
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Added an FAQ section in Documentation/networking/af_xdp.rst to help
first time users with common problems. As problems are getting
identified, entries will be added to the FAQ.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit converts the xdpsock sample application to use the AF_XDP
functions present in libbpf. This cuts down the size of it by nearly
300 lines of code.
The default ring sizes plus the batch size has been increased and the
size of the umem area has decreased. This so that the sample application
will provide higher throughput. Note also that the shared umem code
has been removed from the sample as this is not supported by libbpf
at this point in time.
Tested-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This commit adds AF_XDP support to libbpf. The main reason for this is
to facilitate writing applications that use AF_XDP by offering
higher-level APIs that hide many of the details of the AF_XDP
uapi. This is in the same vein as libbpf facilitates XDP adoption by
offering easy-to-use higher level interfaces of XDP
functionality. Hopefully this will facilitate adoption of AF_XDP, make
applications using it simpler and smaller, and finally also make it
possible for applications to benefit from optimizations in the AF_XDP
user space access code. Previously, people just copied and pasted the
code from the sample application into their application, which is not
desirable.
The interface is composed of two parts:
* Low-level access interface to the four rings and the packet
* High-level control plane interface for creating and setting
up umems and af_xdp sockets as well as a simple XDP program.
Tested-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Simple test that I used to reproduce the issue in the previous commit:
Do BPF_PROG_TEST_RUN with max iterations, each program is 4096 simple
move instructions. File alarm in 0.1 second and check that
bpf_prog_test_run is interrupted (i.e. test doesn't hang).
Note: reposting this for bpf-next to avoid linux-next conflict. In this
version I test both BPF_PROG_TYPE_SOCKET_FILTER (which uses generic
bpf_test_run implementation) and BPF_PROG_TYPE_FLOW_DISSECTOR (which has
it own loop with preempt handling in bpf_prog_test_run_flow_dissector).
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Syzbot found out that running BPF_PROG_TEST_RUN with repeat=0xffffffff
makes process unkillable. The problem is that when CONFIG_PREEMPT is
enabled, we never see need_resched() return true. This is due to the
fact that preempt_enable() (which we do in bpf_test_run_one on each
iteration) now handles resched if it's needed.
Let's disable preemption for the whole run, not per test. In this case
we can properly see whether resched is needed.
Let's also properly return -EINTR to the userspace in case of a signal
interrupt.
This is a follow up for a recently fixed issue in bpf_test_run, see
commit df1a2cb7c7 ("bpf/test_run: fix unkillable
BPF_PROG_TEST_RUN").
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The xdp_redirect and xdp_redirect_map sample programs both load a dummy
program onto the egress interfaces. However, the unload code checks these
programs against the wrong id number, and thus refuses to unload them. Fix
the comparison to avoid this.
Fixes: 3b7a8ec2de ("samples/bpf: Check the prog id before exiting")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
All BPF programs must be called with preemption disabled.
Fixes: 568f196756 ("bpf: check that BPF programs run with preemption disabled")
Reported-by: syzbot+8bf19ee2aa580de7a2a7@syzkaller.appspotmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The skb->queue_mapping already have read access, via __sk_buff->queue_mapping.
This patch allow BPF tc qdisc clsact write access to the queue_mapping via
tc_cls_act_is_valid_access. Also handle that the value NO_QUEUE_MAPPING
is not allowed.
It is already possible to change this via TC filter action skbedit
tc-skbedit(8). Due to the lack of TC examples, lets show one:
# tc qdisc add dev ixgbe1 clsact
# tc filter add dev ixgbe1 ingress matchall action skbedit queue_mapping 5
# tc filter list dev ixgbe1 ingress
The most common mistake is that XPS (Transmit Packet Steering) takes
precedence over setting skb->queue_mapping. XPS is configured per DEVICE
via /sys/class/net/DEVICE/queues/tx-*/xps_cpus via a CPU hex mask. To
disable set mask=00.
The purpose of changing skb->queue_mapping is to influence the selection of
the net_device "txq" (struct netdev_queue), which influence selection of
the qdisc "root_lock" (via txq->qdisc->q.lock) and txq->_xmit_lock. When
using the MQ qdisc the txq->qdisc points to different qdiscs and associated
locks, and HARD_TX_LOCK (txq->_xmit_lock), allowing for CPU scalability.
Due to lack of TC examples, lets show howto attach clsact BPF programs:
# tc qdisc add dev ixgbe2 clsact
# tc filter add dev ixgbe2 egress bpf da obj XXX_kern.o sec tc_qmap2cpu
# tc filter list dev ixgbe2 egress
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Introduce cant_sleep() macro for annotation of functions that
cannot sleep.
Use it in BPF_PROG_RUN to catch execution of BPF programs in
preemptable context.
Suggested-by: Jann Horn <jannh@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
bpftool has support for attach types "stream_verdict" and
"stream_parser" but the documentation was referring to them as
"skb_verdict" and "skb_parse". The inconsistency comes from commit
b7d3826c2e ("bpf: bpftool, add support for attaching programs to
maps").
This patch changes the documentation to match the implementation:
- "bpftool prog help"
- man pages
- bash completion
Signed-off-by: Alban Crequy <alban@kinvolk.io>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
