[PATCH 2/2] ethdev: fix race on ports for telemetry commands

[David Marchand]

On Wed, Oct 2, 2024 at 9:09 PM Robin Jarry <rjarry at redhat.com> wrote:
> >> An alternative to this macro-fu, is to just define a single ethdev
> >> telemetry function, and within that, take the lock and then dispatch to the
> >> appropriate subfunction based upon the actual command coming in. The
> >> dispatch may be slightly slower due to the additional text matching (only
> >> from byte 8 onwards, so very short strings), but I think the code could be
> >> a simpler in C rather than in macros, and the perf impact for telemetry is
> >> likely to be negligible, compared to the overhead of the socket I/O etc.
> >
> > Hopefully, dispatching performance is not important here.
>
> I was going to suggest adding a rte_spinlock_t* parameter to a new
> telemetry register function that would need to be held while the
> callback is invoked. Or if we want to keep doors open to other kinds of
> lock, a wrapper callback.

Well, as you had experimented this approach, we know this does not
work: the ethdev lock is in dpdk shared memory which is not available
yet at the time RTE_INIT() is called.

A single callback is strange, I guess you mean pre/post callbacks then.


-- 
David Marchand