bpf: Optimize recursion detection for arm64 #6287
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Author
|
Upstream branch: 11369e6 |
kernel-patches-daemon-bpf-rc
bot
force-pushed
the
bpf-next_base
branch
from
fe98223 to
6fe2c31
Compare
Author
|
Upstream branch: efa4756 |
kernel-patches-daemon-bpf-rc
bot
force-pushed
the
series/1019553=>bpf-next
branch
from
59edb82 to
8f77b32
Compare
kernel-patches-daemon-bpf-rc
bot
force-pushed
the
bpf-next_base
branch
from
6fe2c31 to
634398c
Compare
Author
|
Upstream branch: b3387b3 |
kernel-patches-daemon-bpf-rc
bot
force-pushed
the
series/1019553=>bpf-next
branch
from
8f77b32 to
1c28618
Compare
kernel-patches-daemon-bpf-rc
bot
force-pushed
the
bpf-next_base
branch
from
634398c to
c95b89c
Compare
Author
|
Upstream branch: 4cb4897 |
BPF programs detect recursion by a per-cpu active flag in struct bpf_prog. This flag is set/unset in the trampoline using atomic operations to prevent inter-context recursion. Some arm64 platforms have slow per-CPU atomic operations, for example, the Neoverse V2. This commit therefore changes the recursion detection mechanism to allow four levels of recursion (normal -> softirq -> hardirq -> NMI). With allowing limited recursion, we can now stop using atomic operations. This approach is similar to get_recursion_context() in perf. Change active to a per-cpu array of four u8 values, one for each context and use non-atomic increment/decrement on them. This improves the performance on ARM64 (64-CPU Neoverse-N1): +----------------+-------------------+-------------------+---------+ | Benchmark | Base run | Patched run | Δ (%) | +----------------+-------------------+-------------------+---------+ | fentry | 3.694 ± 0.003M/s | 3.828 ± 0.007M/s | +3.63% | | fexit | 1.389 ± 0.006M/s | 1.406 ± 0.003M/s | +1.22% | | fmodret | 1.366 ± 0.011M/s | 1.398 ± 0.002M/s | +2.34% | | rawtp | 3.453 ± 0.026M/s | 3.714 ± 0.003M/s | +7.56% | | tp | 2.596 ± 0.005M/s | 2.699 ± 0.006M/s | +3.97% | +----------------+-------------------+-------------------+---------+ Benchmarked using: tools/testing/selftests/bpf/benchs/run_bench_trigger.sh Signed-off-by: Puranjay Mohan <[email protected]>
kernel-patches-daemon-bpf-rc
bot
force-pushed
the
series/1019553=>bpf-next
branch
from
1c28618 to
db00232
Compare
Author
|
At least one diff in series https://patchwork.kernel.org/project/netdevbpf/list/?series=1019553 expired. Closing PR. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Pull request for series with
subject: bpf: Optimize recursion detection for arm64
version: 1
url: https://patchwork.kernel.org/project/netdevbpf/list/?series=1019553