OSDN Git Service

bpftool: Probe for instruction set extensions
authorPaul Chaignon <paul@isovalent.com>
Tue, 4 Jan 2022 18:00:13 +0000 (19:00 +0100)
committerDaniel Borkmann <daniel@iogearbox.net>
Wed, 5 Jan 2022 12:31:40 +0000 (13:31 +0100)
commit0fd800b2456cf90ed738a1260b53acaa8843b5ae
tree8359e6ca88de5515d3fdbf0673c4d1cc53446213
parentc04fb2b0bd9275969be3b0a95f9c3ef76b1bfb73
bpftool: Probe for instruction set extensions

This patch introduces new probes to check whether the kernel supports
instruction set extensions v2 and v3. The first introduced eBPF
instructions BPF_J{LT,LE,SLT,SLE} in commit 92b31a9af73b ("bpf: add
BPF_J{LT,LE,SLT,SLE} instructions"). The second introduces 32-bit
variants of all jump instructions in commit 092ed0968bb6 ("bpf:
verifier support JMP32").

These probes are useful for userspace BPF projects that want to use newer
instruction set extensions on newer kernels, to reduce the programs'
sizes or their complexity. LLVM already provides an mcpu=probe option to
automatically probe the kernel and select the newest-supported
instruction set extension. That is however not flexible enough for all
use cases. For example, in Cilium, we only want to use the v3
instruction set extension on v5.10+, even though it is supported on all
kernels v5.1+.

Signed-off-by: Paul Chaignon <paul@isovalent.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/3bfedcd9898c1f41ac67ca61f144fec84c6c3a92.1641314075.git.paul@isovalent.com
tools/bpf/bpftool/feature.c