On Tue, 2026-09-22 at 07:35 -0700, Adrián Moreno wrote:
> > 
> > IIUC, this change only affects the printing and doesn't affect debugging
> > tools like retis that attempt to surface the drop reasons.  But, maybe
> > Adrian and Antoine (CCed) may want to have a glance as well.
> > 
> 
> I don't think this will affect retis in a bad way. We collect the list
> of drop reasons by inspecting the BTF enums. It generally works but we
> need to keep track of new subsystems enums being added

Right, that wouldn't really change.

> so a dynamically-generated full-list of drop reasons is interesting.

> IIUC, the only way for userspace to read this list is by looking at the
> event format and parse the printf line, right? Any ideas to make this
> more machine-friendly?

It's intended to be machine friendly "enough", trace-cmd/perf (via
libtraceevent or so I think?) parse and use this? But it's also only
available when tracing is enabled.

I think if you really wanted to have a full list dynamically exposed at
runtime across all kernels, then perhaps dropmonitor or some other
netlink interface could expose it? Or maybe another debugfs/sysfs file?
I don't know what you need it for, but I think it's out of scope for
this patchset, and could even be done without it?

johannes

Reply via email to