On 02/13, Alan Jowett wrote:
BPF,
The existing bpf_test_run_opts structure exposes additional fields
including "flags" and "cpu". I propose extending the bpftool prog run to
accept options so set these additional fields.
Use case:
Some BPF programs access state that is potentially shared between
programs running on different CPUs. This can impact the performance of a
BPF program. To permit users to more accurately assess how their BPF
program will perform when being invoked on multiple CPUs in parallel I am
proposing adding the ability for bpftool to set the cpu field in
bpf_test_run_opts struct when calling the syscall.
If no one else is working on this or there are no objections, I will
submit a patch using bpf-next https://github.com/xdp-project/bpf-next/.
Patches are welcome. I don't see anything controversial with extending
bpftool with those extra fields.
PTAL at [0] on how and where to submit the patches.
0: https://www.kernel.org/doc/Documentation/bpf/bpf_devel_QA.rst
Regards,
Alan Jowett
