On Thu, Jun 10, 2021 at 7:12 AM Lorenz Bauer [off-list ref] wrote:

quoted

quoted

The idea of basing this on unique fields in types is neat, the
downside I see is that we encode the logic in the BPF bitstream. If in
the future struct pt_regs is changed, code breaks and we can't do much

If pt_regs fields are renamed all PT_REGS-related stuff, provided by
libbpf in bpf_tracing.h will break as well and will require
re-compilation of BPF application.

I'm thinking more along the lines of, if a PT_REGS definition changes
so that the unique field isn't unique anymore. The BPF is still valid,
but the logic that determines the platform isn't.

struct pt_regs is uapi on every arch.
They cannot change. New registers can be added :) but the chance is
close to zero.