On Thu, 2019-06-06 at 15:13 -0700, Jakub Kicinski wrote:

On Thu, 6 Jun 2019 11:54:27 -0700, Matt Mullins wrote:

quoted

BPF_PROG_TYPE_RAW_TRACEPOINTs can be executed nested on the same CPU, as
they do not increment bpf_prog_active while executing.

This enables three levels of nesting, to support
  - a kprobe or raw tp or perf event,
  - another one of the above that irq context happens to call, and
  - another one in nmi context
(at most one of which may be a kprobe or perf event).

Fixes: 20b9d7ac4852 ("bpf: avoid excessive stack usage for perf_sample_data")

No comment on the code, but you're definitely missing a sign-off.

Oops, I totally am.  I'll give it some more time for opinions to roll
in, and I'll fix that before I resubmit :)

quoted

---
This is more lines of code, but possibly less intrusive than the
per-array-element approach.

I don't necessarily like that I duplicated the nest_level logic in two
places, but I don't see a way to unify them:
  - kprobes' bpf_perf_event_output doesn't use bpf_raw_tp_regs, and does
    use the perf_sample_data,
  - raw tracepoints' bpf_get_stackid uses bpf_raw_tp_regs, but not
    the perf_sample_data, and
  - raw tracepoints' bpf_perf_event_output uses both...