Joanne Koong says:
====================
This patchset adds a new kind of bpf map: the bloom filter map.
Bloom filters are a space-efficient probabilistic data structure
used to quickly test whether an element exists in a set.
For a brief overview about how bloom filters work,
https://en.wikipedia.org/wiki/Bloom_filter
may be helpful.
One example use-case is an application leveraging a bloom filter
map to determine whether a computationally expensive hashmap
lookup can be avoided. If the element was not found in the bloom
filter map, the hashmap lookup can be skipped.
This patchset includes benchmarks for testing the performance of
the bloom filter for different entry sizes and different number of
hash functions used, as well as comparisons for hashmap lookups
with vs. without the bloom filter.
A high level overview of this patchset is as follows:
1/5 - kernel changes for adding bloom filter map
2/5 - libbpf changes for adding map_extra flags
3/5 - tests for the bloom filter map
4/5 - benchmarks for bloom filter lookup/update throughput and false positive
rate
5/5 - benchmarks for how hashmap lookups perform with vs. without the bloom
filter
v5 -> v6:
* in 1/5: remove "inline" from the hash function, add check in syscall to
fail out in cases where map_extra is not 0 for non-bloom-filter maps,
fix alignment matching issues, move "map_extra flags" comments to inside
the bpf_attr struct, add bpf_map_info map_extra changes here, add map_extra
assignment in bpf_map_get_info_by_fd, change hash value_size to u32 instead of
a u64
* in 2/5: remove bpf_map_info map_extra changes, remove TODO comment about
extending BTF arrays to cover u64s, cast to unsigned long long for %llx when
printing out map_extra flags
* in 3/5: use __type(value, ...) instead of __uint(value_size, ...) for values
and keys
* in 4/5: fix wrong bounds for the index when iterating through random values,
update commit message to include update+lookup benchmark results for 8 byte
and 64-byte value sizes, remove explicit global bool initializaton to false
for hashmap_use_bloom and count_false_hits variables
v4 -> v5:
* Change the "bitset map with bloom filter capabilities" to a bloom filter map
with max_entries signifying the number of unique entries expected in the bloom
filter, remove bitset tests
* Reduce verbiage by changing "bloom_filter" to "bloom", and renaming progs to
more concise names.
* in 2/5: remove "map_extra" from struct definitions that are frozen, create a
"bpf_create_map_params" struct to propagate map_extra to the kernel at map
creation time, change map_extra to __u64
* in 4/5: check pthread condition variable in a loop when generating initial
map data, remove "err" checks where not pragmatic, generate random values
for the hashmap in the setup() instead of in the bpf program, add check_args()
for checking that there aren't more requested entries than possible unique
entries for the specified value size
* in 5/5: Update commit message with updated benchmark data
v3 -> v4:
* Generalize the bloom filter map to be a bitset map with bloom filter
capabilities
* Add map_extra flags; pass in nr_hash_funcs through lower 4 bits of map_extra
for the bitset map
* Add tests for the bitset map (non-bloom filter) functionality
* In the benchmarks, stats are computed only as monotonic increases, and place
stats in a struct instead of as a percpu_array bpf map
v2 -> v3:
* Add libbpf changes for supporting nr_hash_funcs, instead of passing the
number of hash functions through map_flags.
* Separate the hashing logic in kernel/bpf/bloom_filter.c into a helper
function
v1 -> v2:
* Remove libbpf changes, and pass the number of hash functions through
map_flags instead.
* Default to using 5 hash functions if no number of hash functions
is specified.
* Use set_bit instead of spinlocks in the bloom filter bitmap. This
improved the speed significantly. For example, using 5 hash functions
with 100k entries, there was roughly a 35% speed increase.
* Use jhash2 (instead of jhash) for u32-aligned value sizes. This
increased the speed by roughly 5 to 15%. When using jhash2 on value
sizes non-u32 aligned (truncating any remainder bits), there was not
a noticeable difference.
* Add test for using the bloom filter as an inner map.
* Reran the benchmarks, updated the commit messages to correspond to
the new results.
====================
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch adds benchmark tests for comparing the performance of hashmap
lookups without the bloom filter vs. hashmap lookups with the bloom filter.
Checking the bloom filter first for whether the element exists should
overall enable a higher throughput for hashmap lookups, since if the
element does not exist in the bloom filter, we can avoid a costly lookup in
the hashmap.
On average, using 5 hash functions in the bloom filter tended to perform
the best across the widest range of different entry sizes. The benchmark
results using 5 hash functions (running on 8 threads on a machine with one
numa node, and taking the average of 3 runs) were roughly as follows:
value_size = 4 bytes -
10k entries: 30% faster
50k entries: 40% faster
100k entries: 40% faster
500k entres: 70% faster
1 million entries: 90% faster
5 million entries: 140% faster
value_size = 8 bytes -
10k entries: 30% faster
50k entries: 40% faster
100k entries: 50% faster
500k entres: 80% faster
1 million entries: 100% faster
5 million entries: 150% faster
value_size = 16 bytes -
10k entries: 20% faster
50k entries: 30% faster
100k entries: 35% faster
500k entres: 65% faster
1 million entries: 85% faster
5 million entries: 110% faster
value_size = 40 bytes -
10k entries: 5% faster
50k entries: 15% faster
100k entries: 20% faster
500k entres: 65% faster
1 million entries: 75% faster
5 million entries: 120% faster
Signed-off-by: Joanne Koong <joannekoong@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211027234504.30744-6-joannekoong@fb.com
This patch adds benchmark tests for the throughput (for lookups + updates)
and the false positive rate of bloom filter lookups, as well as some
minor refactoring of the bash script for running the benchmarks.
These benchmarks show that as the number of hash functions increases,
the throughput and the false positive rate of the bloom filter decreases.
>From the benchmark data, the approximate average false-positive rates
are roughly as follows:
1 hash function = ~30%
2 hash functions = ~15%
3 hash functions = ~5%
4 hash functions = ~2.5%
5 hash functions = ~1%
6 hash functions = ~0.5%
7 hash functions = ~0.35%
8 hash functions = ~0.15%
9 hash functions = ~0.1%
10 hash functions = ~0%
For reference data, the benchmarks run on one thread on a machine
with one numa node for 1 to 5 hash functions for 8-byte and 64-byte
values are as follows:
1 hash function:
50k entries
8-byte value
Lookups - 51.1 M/s operations
Updates - 33.6 M/s operations
False positive rate: 24.15%
64-byte value
Lookups - 15.7 M/s operations
Updates - 15.1 M/s operations
False positive rate: 24.2%
100k entries
8-byte value
Lookups - 51.0 M/s operations
Updates - 33.4 M/s operations
False positive rate: 24.04%
64-byte value
Lookups - 15.6 M/s operations
Updates - 14.6 M/s operations
False positive rate: 24.06%
500k entries
8-byte value
Lookups - 50.5 M/s operations
Updates - 33.1 M/s operations
False positive rate: 27.45%
64-byte value
Lookups - 15.6 M/s operations
Updates - 14.2 M/s operations
False positive rate: 27.42%
1 mil entries
8-byte value
Lookups - 49.7 M/s operations
Updates - 32.9 M/s operations
False positive rate: 27.45%
64-byte value
Lookups - 15.4 M/s operations
Updates - 13.7 M/s operations
False positive rate: 27.58%
2.5 mil entries
8-byte value
Lookups - 47.2 M/s operations
Updates - 31.8 M/s operations
False positive rate: 30.94%
64-byte value
Lookups - 15.3 M/s operations
Updates - 13.2 M/s operations
False positive rate: 30.95%
5 mil entries
8-byte value
Lookups - 41.1 M/s operations
Updates - 28.1 M/s operations
False positive rate: 31.01%
64-byte value
Lookups - 13.3 M/s operations
Updates - 11.4 M/s operations
False positive rate: 30.98%
2 hash functions:
50k entries
8-byte value
Lookups - 34.1 M/s operations
Updates - 20.1 M/s operations
False positive rate: 9.13%
64-byte value
Lookups - 8.4 M/s operations
Updates - 7.9 M/s operations
False positive rate: 9.21%
100k entries
8-byte value
Lookups - 33.7 M/s operations
Updates - 18.9 M/s operations
False positive rate: 9.13%
64-byte value
Lookups - 8.4 M/s operations
Updates - 7.7 M/s operations
False positive rate: 9.19%
500k entries
8-byte value
Lookups - 32.7 M/s operations
Updates - 18.1 M/s operations
False positive rate: 12.61%
64-byte value
Lookups - 8.4 M/s operations
Updates - 7.5 M/s operations
False positive rate: 12.61%
1 mil entries
8-byte value
Lookups - 30.6 M/s operations
Updates - 18.9 M/s operations
False positive rate: 12.54%
64-byte value
Lookups - 8.0 M/s operations
Updates - 7.0 M/s operations
False positive rate: 12.52%
2.5 mil entries
8-byte value
Lookups - 25.3 M/s operations
Updates - 16.7 M/s operations
False positive rate: 16.77%
64-byte value
Lookups - 7.9 M/s operations
Updates - 6.5 M/s operations
False positive rate: 16.88%
5 mil entries
8-byte value
Lookups - 20.8 M/s operations
Updates - 14.7 M/s operations
False positive rate: 16.78%
64-byte value
Lookups - 7.0 M/s operations
Updates - 6.0 M/s operations
False positive rate: 16.78%
3 hash functions:
50k entries
8-byte value
Lookups - 25.1 M/s operations
Updates - 14.6 M/s operations
False positive rate: 7.65%
64-byte value
Lookups - 5.8 M/s operations
Updates - 5.5 M/s operations
False positive rate: 7.58%
100k entries
8-byte value
Lookups - 24.7 M/s operations
Updates - 14.1 M/s operations
False positive rate: 7.71%
64-byte value
Lookups - 5.8 M/s operations
Updates - 5.3 M/s operations
False positive rate: 7.62%
500k entries
8-byte value
Lookups - 22.9 M/s operations
Updates - 13.9 M/s operations
False positive rate: 2.62%
64-byte value
Lookups - 5.6 M/s operations
Updates - 4.8 M/s operations
False positive rate: 2.7%
1 mil entries
8-byte value
Lookups - 19.8 M/s operations
Updates - 12.6 M/s operations
False positive rate: 2.60%
64-byte value
Lookups - 5.3 M/s operations
Updates - 4.4 M/s operations
False positive rate: 2.69%
2.5 mil entries
8-byte value
Lookups - 16.2 M/s operations
Updates - 10.7 M/s operations
False positive rate: 4.49%
64-byte value
Lookups - 4.9 M/s operations
Updates - 4.1 M/s operations
False positive rate: 4.41%
5 mil entries
8-byte value
Lookups - 18.8 M/s operations
Updates - 9.2 M/s operations
False positive rate: 4.45%
64-byte value
Lookups - 5.2 M/s operations
Updates - 3.9 M/s operations
False positive rate: 4.54%
4 hash functions:
50k entries
8-byte value
Lookups - 19.7 M/s operations
Updates - 11.1 M/s operations
False positive rate: 1.01%
64-byte value
Lookups - 4.4 M/s operations
Updates - 4.0 M/s operations
False positive rate: 1.00%
100k entries
8-byte value
Lookups - 19.5 M/s operations
Updates - 10.9 M/s operations
False positive rate: 1.00%
64-byte value
Lookups - 4.3 M/s operations
Updates - 3.9 M/s operations
False positive rate: 0.97%
500k entries
8-byte value
Lookups - 18.2 M/s operations
Updates - 10.6 M/s operations
False positive rate: 2.05%
64-byte value
Lookups - 4.3 M/s operations
Updates - 3.7 M/s operations
False positive rate: 2.05%
1 mil entries
8-byte value
Lookups - 15.5 M/s operations
Updates - 9.6 M/s operations
False positive rate: 1.99%
64-byte value
Lookups - 4.0 M/s operations
Updates - 3.4 M/s operations
False positive rate: 1.99%
2.5 mil entries
8-byte value
Lookups - 13.8 M/s operations
Updates - 7.7 M/s operations
False positive rate: 3.91%
64-byte value
Lookups - 3.7 M/s operations
Updates - 3.6 M/s operations
False positive rate: 3.78%
5 mil entries
8-byte value
Lookups - 13.0 M/s operations
Updates - 6.9 M/s operations
False positive rate: 3.93%
64-byte value
Lookups - 3.5 M/s operations
Updates - 3.7 M/s operations
False positive rate: 3.39%
5 hash functions:
50k entries
8-byte value
Lookups - 16.4 M/s operations
Updates - 9.1 M/s operations
False positive rate: 0.78%
64-byte value
Lookups - 3.5 M/s operations
Updates - 3.2 M/s operations
False positive rate: 0.77%
100k entries
8-byte value
Lookups - 16.3 M/s operations
Updates - 9.0 M/s operations
False positive rate: 0.79%
64-byte value
Lookups - 3.5 M/s operations
Updates - 3.2 M/s operations
False positive rate: 0.78%
500k entries
8-byte value
Lookups - 15.1 M/s operations
Updates - 8.8 M/s operations
False positive rate: 1.82%
64-byte value
Lookups - 3.4 M/s operations
Updates - 3.0 M/s operations
False positive rate: 1.78%
1 mil entries
8-byte value
Lookups - 13.2 M/s operations
Updates - 7.8 M/s operations
False positive rate: 1.81%
64-byte value
Lookups - 3.2 M/s operations
Updates - 2.8 M/s operations
False positive rate: 1.80%
2.5 mil entries
8-byte value
Lookups - 10.5 M/s operations
Updates - 5.9 M/s operations
False positive rate: 0.29%
64-byte value
Lookups - 3.2 M/s operations
Updates - 2.4 M/s operations
False positive rate: 0.28%
5 mil entries
8-byte value
Lookups - 9.6 M/s operations
Updates - 5.7 M/s operations
False positive rate: 0.30%
64-byte value
Lookups - 3.2 M/s operations
Updates - 2.7 M/s operations
False positive rate: 0.30%
Signed-off-by: Joanne Koong <joannekoong@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211027234504.30744-5-joannekoong@fb.com
This patch adds test cases for bpf bloom filter maps. They include tests
checking against invalid operations by userspace, tests for using the
bloom filter map as an inner map, and a bpf program that queries the
bloom filter map for values added by a userspace program.
Signed-off-by: Joanne Koong <joannekoong@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211027234504.30744-4-joannekoong@fb.com
This patch adds the libbpf infrastructure for supporting a
per-map-type "map_extra" field, whose definition will be
idiosyncratic depending on map type.
For example, for the bloom filter map, the lower 4 bits of
map_extra is used to denote the number of hash functions.
Please note that until libbpf 1.0 is here, the
"bpf_create_map_params" struct is used as a temporary
means for propagating the map_extra field to the kernel.
Signed-off-by: Joanne Koong <joannekoong@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211027234504.30744-3-joannekoong@fb.com
This patch adds the kernel-side changes for the implementation of
a bpf bloom filter map.
The bloom filter map supports peek (determining whether an element
is present in the map) and push (adding an element to the map)
operations.These operations are exposed to userspace applications
through the already existing syscalls in the following way:
BPF_MAP_LOOKUP_ELEM -> peek
BPF_MAP_UPDATE_ELEM -> push
The bloom filter map does not have keys, only values. In light of
this, the bloom filter map's API matches that of queue stack maps:
user applications use BPF_MAP_LOOKUP_ELEM/BPF_MAP_UPDATE_ELEM
which correspond internally to bpf_map_peek_elem/bpf_map_push_elem,
and bpf programs must use the bpf_map_peek_elem and bpf_map_push_elem
APIs to query or add an element to the bloom filter map. When the
bloom filter map is created, it must be created with a key_size of 0.
For updates, the user will pass in the element to add to the map
as the value, with a NULL key. For lookups, the user will pass in the
element to query in the map as the value, with a NULL key. In the
verifier layer, this requires us to modify the argument type of
a bloom filter's BPF_FUNC_map_peek_elem call to ARG_PTR_TO_MAP_VALUE;
as well, in the syscall layer, we need to copy over the user value
so that in bpf_map_peek_elem, we know which specific value to query.
A few things to please take note of:
* If there are any concurrent lookups + updates, the user is
responsible for synchronizing this to ensure no false negative lookups
occur.
* The number of hashes to use for the bloom filter is configurable from
userspace. If no number is specified, the default used will be 5 hash
functions. The benchmarks later in this patchset can help compare the
performance of using different number of hashes on different entry
sizes. In general, using more hashes decreases both the false positive
rate and the speed of a lookup.
* Deleting an element in the bloom filter map is not supported.
* The bloom filter map may be used as an inner map.
* The "max_entries" size that is specified at map creation time is used
to approximate a reasonable bitmap size for the bloom filter, and is not
otherwise strictly enforced. If the user wishes to insert more entries
into the bloom filter than "max_entries", they may do so but they should
be aware that this may lead to a higher false positive rate.
Signed-off-by: Joanne Koong <joannekoong@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211027234504.30744-2-joannekoong@fb.com
After commit 9298e63eaf ("bpf/tests: Add exhaustive tests of ALU
operand magnitudes"), when modprobe test_bpf.ko with JIT on mips64,
there exists segment fault due to the following reason:
[...]
ALU64_MOV_X: all register value magnitudes jited:1
Break instruction in kernel code[#1]
[...]
It seems that the related JIT implementations of some test cases
in test_bpf() have problems. At this moment, I do not care about
the segment fault while I just want to verify the test cases of
tail calls.
Based on the above background and motivation, add the following
module parameter test_suite to the test_bpf.ko:
test_suite=<string>: only the specified test suite will be run, the
string can be "test_bpf", "test_tail_calls" or "test_skb_segment".
If test_suite is not specified, but test_id, test_name or test_range
is specified, set 'test_bpf' as the default test suite. This is useful
to only test the corresponding test suite when specifying the valid
test_suite string.
Any invalid test suite will result in -EINVAL being returned and no
tests being run. If the test_suite is not specified or specified as
empty string, it does not change the current logic, all of the test
cases will be run.
Here are some test results:
# dmesg -c
# modprobe test_bpf
# dmesg | grep Summary
test_bpf: Summary: 1009 PASSED, 0 FAILED, [0/997 JIT'ed]
test_bpf: test_tail_calls: Summary: 8 PASSED, 0 FAILED, [0/8 JIT'ed]
test_bpf: test_skb_segment: Summary: 2 PASSED, 0 FAILED
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_suite=test_bpf
# dmesg | tail -1
test_bpf: Summary: 1009 PASSED, 0 FAILED, [0/997 JIT'ed]
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_suite=test_tail_calls
# dmesg
test_bpf: #0 Tail call leaf jited:0 21 PASS
[...]
test_bpf: #7 Tail call error path, index out of range jited:0 32 PASS
test_bpf: test_tail_calls: Summary: 8 PASSED, 0 FAILED, [0/8 JIT'ed]
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_suite=test_skb_segment
# dmesg
test_bpf: #0 gso_with_rx_frags PASS
test_bpf: #1 gso_linear_no_head_frag PASS
test_bpf: test_skb_segment: Summary: 2 PASSED, 0 FAILED
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_id=1
# dmesg
test_bpf: test_bpf: set 'test_bpf' as the default test_suite.
test_bpf: #1 TXA jited:0 54 51 50 PASS
test_bpf: Summary: 1 PASSED, 0 FAILED, [0/1 JIT'ed]
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_suite=test_bpf test_name=TXA
# dmesg
test_bpf: #1 TXA jited:0 54 50 51 PASS
test_bpf: Summary: 1 PASSED, 0 FAILED, [0/1 JIT'ed]
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_suite=test_tail_calls test_range=6,7
# dmesg
test_bpf: #6 Tail call error path, NULL target jited:0 41 PASS
test_bpf: #7 Tail call error path, index out of range jited:0 32 PASS
test_bpf: test_tail_calls: Summary: 2 PASSED, 0 FAILED, [0/2 JIT'ed]
# rmmod test_bpf
# dmesg -c
# modprobe test_bpf test_suite=test_skb_segment test_id=1
# dmesg
test_bpf: #1 gso_linear_no_head_frag PASS
test_bpf: test_skb_segment: Summary: 1 PASSED, 0 FAILED
By the way, the above segment fault has been fixed in the latest bpf-next
tree which contains the mips64 JIT rework.
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Johan Almbladh <johan.almbladh@anyfinetworks.com>
Acked-by: Johan Almbladh <johan.almbladh@anyfinetworks.com>
Link: https://lore.kernel.org/bpf/1635384321-28128-1-git-send-email-yangtiezhu@loongson.cn
When a tracing BPF program attempts to read memory without using the
bpf_probe_read() helper, the verifier marks the load instruction with
the BPF_PROBE_MEM flag. Since the riscv JIT does not currently recognize
this flag it falls back to the interpreter.
Add support for BPF_PROBE_MEM, by appending an exception table to the
BPF program. If the load instruction causes a data abort, the fixup
infrastructure finds the exception table and fixes up the fault, by
clearing the destination register and jumping over the faulting
instruction.
A more generic solution would add a "handler" field to the table entry,
like on x86 and s390. The same issue in ARM64 is fixed in 8008342853
("bpf, arm64: Add BPF exception tables").
Signed-off-by: Tong Tiangen <tongtiangen@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Pu Lehui <pulehui@huawei.com>
Tested-by: Björn Töpel <bjorn@kernel.org>
Acked-by: Björn Töpel <bjorn@kernel.org>
Link: https://lore.kernel.org/bpf/20211027111822.3801679-1-tongtiangen@huawei.com
Yucong Sun says:
====================
Several patches to improve parallel execution mode, updating vmtest.sh
and fixed two previously dropped patches according to feedback.
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
This patch delete ns_src/ns_dst/ns_redir namespaces before recreating
them, making the test more robust.
Signed-off-by: Yucong Sun <sunyucong@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211025223345.2136168-5-fallentree@fb.com
This patch makes attach_probe uses its own method as attach point,
avoiding conflict with other tests like bpf_cookie.
Signed-off-by: Yucong Sun <sunyucong@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211025223345.2136168-4-fallentree@fb.com
Increase memory to 4G, 8 SMP core with host cpu passthrough. This
make it run faster in parallel mode and more likely to succeed.
Signed-off-by: Yucong Sun <sunyucong@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211025223345.2136168-2-fallentree@fb.com
Eric Dumazet says:
====================
From: Eric Dumazet <edumazet@google.com>
Two first patches fix bugs added in 5.1 and 5.5
Third patch replaces the u64 fields in struct bpf_prog_stats
with u64_stats_t ones to avoid possible sampling errors,
in case of load/store stearing.
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 316580b69d ("u64_stats: provide u64_stats_t type")
fixed possible load/store tearing on 64bit arches.
For instance the following C code
stats->nsecs += sched_clock() - start;
Could be rightfully implemented like this by a compiler,
confusing concurrent readers a lot:
stats->nsecs += sched_clock();
// arbitrary delay
stats->nsecs -= start;
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211026214133.3114279-4-eric.dumazet@gmail.com
It seems update_prog_stats() suffers from same issue fixed
in the prior patch:
As it can run while interrupts are enabled, it could
be re-entered and the u64_stats syncp could be mangled.
Fixes: fec56f5890 ("bpf: Introduce BPF trampoline")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211026214133.3114279-3-eric.dumazet@gmail.com
Add a flag to `enum libbpf_strict_mode' to disable the global
`bpf_objects_list', preventing race conditions when concurrent threads
call bpf_object__open() or bpf_object__close().
bpf_object__next() will return NULL if this option is set.
Callers may achieve the same workflow by tracking bpf_objects in
application code.
[0] Closes: https://github.com/libbpf/libbpf/issues/293
Signed-off-by: Joe Burton <jevburton@google.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026223528.413950-1-jevburton.kernel@gmail.com
Function in modules could appear in /proc/kallsyms in random order.
ffffffffa02608a0 t bpf_testmod_loop_test
ffffffffa02600c0 t __traceiter_bpf_testmod_test_writable_bare
ffffffffa0263b60 d __tracepoint_bpf_testmod_test_write_bare
ffffffffa02608c0 T bpf_testmod_test_read
ffffffffa0260d08 t __SCT__tp_func_bpf_testmod_test_writable_bare
ffffffffa0263300 d __SCK__tp_func_bpf_testmod_test_read
ffffffffa0260680 T bpf_testmod_test_write
ffffffffa0260860 t bpf_testmod_test_mod_kfunc
Therefore, we cannot reliably use kallsyms_find_next() to find the end of
a function. Replace it with a simple guess (start + 128). This is good
enough for this test.
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211022234814.318457-1-songliubraving@fb.com
Skipping the second half of the test is not enough to silent the warning
in dmesg. Skip the whole test before we can either properly silent the
warning in kernel, or fix LBR snapshot for VM.
Fixes: 025bd7c753 ("selftests/bpf: Add test for bpf_get_branch_snapshot")
Fixes: aa67fdb464 ("selftests/bpf: Skip the second half of get_branch_snapshot in vm")
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026000733.477714-1-songliubraving@fb.com
Ilya Leoshkevich says:
====================
v2: https://lore.kernel.org/bpf/20211025131214.731972-1-iii@linux.ibm.com/
v2 -> v3: Split the fix from the cleanup (Daniel).
v1: https://lore.kernel.org/bpf/20211021234653.643302-1-iii@linux.ibm.com/
v1 -> v2: Drop bpf_core_calc_field_relo() restructuring, split the
__BYTE_ORDER__ change (Andrii).
Hi,
this series fixes test failures in core_reloc on s390.
Patch 1 fixes an endianness bug with __BYTE_ORDER vs __BYTE_ORDER__.
Patches 2-5 make the rest of the code consistent in that respect.
Patch 6 fixes an endianness issue in test_core_reloc_mods.
Best regards,
Ilya
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Use the compiler-defined __BYTE_ORDER__ instead of the libc-defined
__BYTE_ORDER for consistency.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026010831.748682-6-iii@linux.ibm.com
Use the compiler-defined __BYTE_ORDER__ instead of the libc-defined
__BYTE_ORDER for consistency.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026010831.748682-5-iii@linux.ibm.com
Use the compiler-defined __BYTE_ORDER__ instead of the libc-defined
__BYTE_ORDER for consistency.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026010831.748682-4-iii@linux.ibm.com
Use the compiler-defined __BYTE_ORDER__ instead of the libc-defined
__BYTE_ORDER for consistency.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026010831.748682-3-iii@linux.ibm.com
__BYTE_ORDER is supposed to be defined by a libc, and __BYTE_ORDER__ -
by a compiler. bpf_core_read.h checks __BYTE_ORDER == __LITTLE_ENDIAN,
which is true if neither are defined, leading to incorrect behavior on
big-endian hosts if libc headers are not included, which is often the
case.
Fixes: ee26dade0e ("libbpf: Add support for relocatable bitfields")
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211026010831.748682-2-iii@linux.ibm.com
Andrii Nakryiko says:
====================
Add libbpf APIs to access BPF program instructions. Both before and after
libbpf processing (before and after bpf_object__load()). This allows to
inspect what's going on with BPF program assembly instructions as libbpf
performs its processing magic.
But in more practical terms, this allows to do a no-brainer BPF program
cloning, which is something you need when working with fentry/fexit BPF
programs to be able to attach the same BPF program code to multiple kernel
functions. Currently, kernel needs multiple copies of BPF programs, each
loaded with its own target BTF ID. retsnoop is one such example that
previously had to rely on bpf_program__set_prep() API to hijack program
instructions ([0] for before and after).
Speaking of bpf_program__set_prep() API and the whole concept of
multiple-instance BPF programs in libbpf, all that is scheduled for
deprecation in v0.7. It doesn't work well, it's cumbersome, and it will become
more broken as libbpf adds more functionality. So deprecate and remove it in
libbpf 1.0. It doesn't seem to be used by anyone anyways (except for that
retsnoop hack, which is now much cleaner with new APIs as can be seen in [0]).
[0] https://github.com/anakryiko/retsnoop/pull/1
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The name of the API doesn't convey clearly that this size is in number
of bytes (there needed to be a separate comment to make this clear in
libbpf.h). Further, measuring the size of BPF program in bytes is not
exactly the best fit, because BPF programs always consist of 8-byte
instructions. As such, bpf_program__insn_cnt() is a better alternative
in pretty much any imaginable case.
So schedule bpf_program__size() deprecation starting from v0.7 and it
will be removed in libbpf 1.0.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211025224531.1088894-5-andrii@kernel.org
Schedule deprecation of a set of APIs that are related to multi-instance
bpf_programs:
- bpf_program__set_prep() ([0]);
- bpf_program__{set,unset}_instance() ([1]);
- bpf_program__nth_fd().
These APIs are obscure, very niche, and don't seem to be used much in
practice. bpf_program__set_prep() is pretty useless for anything but the
simplest BPF programs, as it doesn't allow to adjust BPF program load
attributes, among other things. In short, it already bitrotted and will
bitrot some more if not removed.
With bpf_program__insns() API, which gives access to post-processed BPF
program instructions of any given entry-point BPF program, it's now
possible to do whatever necessary adjustments were possible with
set_prep() API before, but also more. Given any such use case is
automatically an advanced use case, requiring users to stick to
low-level bpf_prog_load() APIs and managing their own prog FDs is
reasonable.
[0] Closes: https://github.com/libbpf/libbpf/issues/299
[1] Closes: https://github.com/libbpf/libbpf/issues/300
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211025224531.1088894-4-andrii@kernel.org
Add APIs providing read-only access to bpf_program BPF instructions ([0]).
This is useful for diagnostics purposes, but it also allows a cleaner
support for cloning BPF programs after libbpf did all the FD resolution
and CO-RE relocations, subprog instructions appending, etc. Currently,
cloning BPF program is possible only through hijacking a half-broken
bpf_program__set_prep() API, which doesn't really work well for anything
but most primitive programs. For instance, set_prep() API doesn't allow
adjusting BPF program load parameters which are necessary for loading
fentry/fexit BPF programs (the case where BPF program cloning is
a necessity if doing some sort of mass-attachment functionality).
Given bpf_program__set_prep() API is set to be deprecated, having
a cleaner alternative is a must. libbpf internally already keeps track
of linear array of struct bpf_insn, so it's not hard to expose it. The
only gotcha is that libbpf previously freed instructions array during
bpf_object load time, which would make this API much less useful overall,
because in between bpf_object__open() and bpf_object__load() a lot of
changes to instructions are done by libbpf.
So this patch makes libbpf hold onto prog->insns array even after BPF
program loading. I think this is a small price for added functionality
and improved introspection of BPF program code.
See retsnoop PR ([1]) for how it can be used in practice and code
savings compared to relying on bpf_program__set_prep().
[0] Closes: https://github.com/libbpf/libbpf/issues/298
[1] https://github.com/anakryiko/retsnoop/pull/1
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211025224531.1088894-3-andrii@kernel.org
Quentin Monnet says:
====================
When listing BPF objects, bpftool can print a number of properties about
items holding references to these objects. For example, it can show pinned
paths for BPF programs, maps, and links; or programs and maps using a given
BTF object; or the names and PIDs of processes referencing BPF objects. To
collect this information, bpftool uses hash maps (to be clear: the data
structures, inside bpftool - we are not talking of BPF maps). It uses the
implementation available from the kernel, and picks it up from
tools/include/linux/hashtable.h.
This patchset converts bpftool's hash maps to a distinct implementation
instead, the one coming with libbpf. The main motivation for this change is
that it should ease the path towards a potential out-of-tree mirror for
bpftool, like the one libbpf already has. Although it's not perfect to
depend on libbpf's internal components, bpftool is intimately tied with the
library anyway, and this looks better than depending too much on (non-UAPI)
kernel headers.
The first two patches contain preparatory work on the Makefile and on the
initialisation of the hash maps for collecting pinned paths for objects.
Then the transition is split into several steps, one for each kind of
properties for which the collection is backed by hash maps.
v2:
- Move hashmap cleanup for pinned paths for links from do_detach() to
do_show().
- Handle errors on hashmap__append() (in three of the patches).
- Rename bpftool_hash_fn() and bpftool_equal_fn() as hash_fn_for_key_id()
and equal_fn_for_key_id(), respectively.
- Add curly braces for hashmap__for_each_key_entry() { } in
show_btf_plain() and show_btf_json(), where the flow was difficult to
read.
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
In order to show PIDs and names for processes holding references to BPF
programs, maps, links, or BTF objects, bpftool creates hash maps to
store all relevant information. This commit is part of a set that
transitions from the kernel's hash map implementation to the one coming
with libbpf.
The motivation is to make bpftool less dependent of kernel headers, to
ease the path to a potential out-of-tree mirror, like libbpf has.
This is the third and final step of the transition, in which we convert
the hash maps used for storing the information about the processes
holding references to BPF objects (programs, maps, links, BTF), and at
last we drop the inclusion of tools/include/linux/hashtable.h.
Note: Checkpatch complains about the use of __weak declarations, and the
missing empty lines after the bunch of empty function declarations when
compiling without the BPF skeletons (none of these were introduced in
this patch). We want to keep things as they are, and the reports should
be safe to ignore.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211023205154.6710-6-quentin@isovalent.com
In order to show BPF programs and maps using BTF objects when the latter
are being listed, bpftool creates hash maps to store all relevant items.
This commit is part of a set that transitions from the kernel's hash map
implementation to the one coming with libbpf.
The motivation is to make bpftool less dependent of kernel headers, to
ease the path to a potential out-of-tree mirror, like libbpf has.
This commit focuses on the two hash maps used by bpftool when listing
BTF objects to store references to programs and maps, and convert them
to the libbpf's implementation.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211023205154.6710-5-quentin@isovalent.com
In order to show pinned paths for BPF programs, maps, or links when
listing them with the "-f" option, bpftool creates hash maps to store
all relevant paths under the bpffs. So far, it would rely on the
kernel implementation (from tools/include/linux/hashtable.h).
We can make bpftool rely on libbpf's implementation instead. The
motivation is to make bpftool less dependent of kernel headers, to ease
the path to a potential out-of-tree mirror, like libbpf has.
This commit is the first step of the conversion: the hash maps for
pinned paths for programs, maps, and links are converted to libbpf's
hashmap.{c,h}. Other hash maps used for the PIDs of process holding
references to BPF objects are left unchanged for now. On the build side,
this requires adding a dependency to a second header internal to libbpf,
and making it a dependency for the bootstrap bpftool version as well.
The rest of the changes are a rather straightforward conversion.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211023205154.6710-4-quentin@isovalent.com
BPF programs, maps, and links, can all be listed with their pinned paths
by bpftool, when the "-f" option is provided. To do so, bpftool builds
hash maps containing all pinned paths for each kind of objects.
These three hash maps are always initialised in main.c, and exposed
through main.h. There appear to be no particular reason to do so: we can
just as well make them static to the files that need them (prog.c,
map.c, and link.c respectively), and initialise them only when we want
to show objects and the "-f" switch is provided.
This may prevent unnecessary memory allocations if the implementation of
the hash maps was to change in the future.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211023205154.6710-3-quentin@isovalent.com
The dependency is only useful to make sure that the $(LIBBPF_HDRS_DIR)
directory is created before we try to install locally the required
libbpf internal header. Let's create this directory properly instead.
This is in preparation of making $(LIBBPF_INTERNAL_HDRS) a dependency to
the bootstrap bpftool version, in which case we want no dependency on
$(LIBBPF).
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20211023205154.6710-2-quentin@isovalent.com
Andrii Nakryiko says:
====================
Reduce amount of waiting time when running test_progs in parallel mode (-j) by
splitting bpf_verif_scale selftests into multiple tests. Previously it was
structured as a test with multiple subtests, but subtests are not easily
parallelizable with test_progs' infra. Also in practice each scale subtest is
really an independent test with nothing shared across all substest.
This patch set changes how test_progs test discovery works. Now it is possible
to define multiple tests within a single source code file. One of the patches
also marks tc_redirect selftests as serial, because it's extremely harmful to
the test system when run in parallel mode.
====================
Acked-by: Yucong Sun <sunyucong@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Instead of using subtests in bpf_verif_scale selftest, turn each scale
sub-test into its own test. Each subtest is compltely independent and
just reuses a bit of common test running logic, so the conversion is
trivial. For convenience, keep all of BPF verifier scale tests in one
file.
This conversion shaves off a significant amount of time when running
test_progs in parallel mode. E.g., just running scale tests (-t verif_scale):
BEFORE
======
Summary: 24/0 PASSED, 0 SKIPPED, 0 FAILED
real 0m22.894s
user 0m0.012s
sys 0m22.797s
AFTER
=====
Summary: 24/0 PASSED, 0 SKIPPED, 0 FAILED
real 0m12.044s
user 0m0.024s
sys 0m27.869s
Ten second saving right there. test_progs -j is not yet ready to be
turned on by default, unfortunately, and some tests fail almost every
time, but this is a good improvement nevertheless. Ignoring few
failures, here is sequential vs parallel run times when running all
tests now:
SEQUENTIAL
==========
Summary: 206/953 PASSED, 4 SKIPPED, 0 FAILED
real 1m5.625s
user 0m4.211s
sys 0m31.650s
PARALLEL
========
Summary: 204/952 PASSED, 4 SKIPPED, 2 FAILED
real 0m35.550s
user 0m4.998s
sys 0m39.890s
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211022223228.99920-5-andrii@kernel.org
It seems to cause a lot of harm to kprobe/tracepoint selftests. Yucong
mentioned before that it does manipulate sysfs, which might be the
reason. So let's mark it as serial, though ideally it would be less
intrusive on the system at test.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211022223228.99920-4-andrii@kernel.org
Revamp how test discovery works for test_progs and allow multiple test
entries per file. Any global void function with no arguments and
serial_test_ or test_ prefix is considered a test.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211022223228.99920-3-andrii@kernel.org
Ensure that all test entry points are global void functions with no
input arguments. Mark few subtest entry points as static.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211022223228.99920-2-andrii@kernel.org
Original code assumed fixed and correct BTF header length. That's not
always the case, though, so fix this bug with a proper additional check.
And use actual header length instead of sizeof(struct btf_header) in
sanity checks.
Fixes: 8a138aed4a ("bpf: btf: Add BTF support to libbpf")
Reported-by: Evgeny Vereshchagin <evvers@ya.ru>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211023003157.726961-2-andrii@kernel.org
btf_header's str_off+str_len or type_off+type_len can overflow as they
are u32s. This will lead to bypassing the sanity checks during BTF
parsing, resulting in crashes afterwards. Fix by using 64-bit signed
integers for comparison.
Fixes: d812362450 ("libbpf: Fix BTF data layout checks and allow empty BTF")
Reported-by: Evgeny Vereshchagin <evvers@ya.ru>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211023003157.726961-1-andrii@kernel.org
Yonghong Song says:
====================
Latest upstream llvm-project added support for btf_decl_tag attributes
for typedef declarations ([1], [2]). Similar to other btf_decl_tag cases,
func/func_param/global_var/struct/union/field, btf_decl_tag with typedef
declaration can carry information from kernel source to clang compiler
and then to dwarf/BTF, for bpf verification or other use cases.
This patch set added kernel support for BTF_KIND_DECL_TAG to typedef
declaration (Patch 1). Additional selftests are added to cover
unit testing, dedup, or bpf program usage of btf_decl_tag with typedef.
(Patches 2, 3 and 4). The btf documentation is updated to include
BTF_KIND_DECL_TAG typedef (Patch 5).
[1] https://reviews.llvm.org/D110127
[2] https://reviews.llvm.org/D112259
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add unit tests for deduplication of BTF_KIND_DECL_TAG to typedef types.
Also changed a few comments from "tag" to "decl_tag" to match
BTF_KIND_DECL_TAG enum value name.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211021195638.4019770-1-yhs@fb.com
