[dpdk-dev] rte_log simplification using constructor scheme

[Jerin Jacob]

On Tue, Jun 9, 2020 at 1:55 AM Thomas Monjalon <thomas at monjalon.net> wrote:
>
> 27/05/2020 18:05, Jerin Kollanukkaran:
> > +#define RTE_LOG_REGISTER(type, name, level)                            \
> > +int type;                                                              \
> > +RTE_INIT(__##type)                                                     \
> > +{                                                                      \
> > +       type = rte_log_register(RTE_STR(name));                         \
> > +       if (type >= 0)                                                  \
> > +               rte_log_set_level(type, RTE_LOG_##level);               \
> > +}
>
> We should consider using rte_log_register_type_and_pick_level()
> which works for drivers loaded later in the init sequence.

I sent the first version[1] based on the current driver usage to avoid
any behavior change. (Major work is changing the drivers, we can
decide the RTE_LOG_REGISTER definition in next versions)
[1]
http://patches.dpdk.org/patch/71650/
I will send the v2 based on the consensus.
Please comment on http://patches.dpdk.org/patch/71650/.

> I feel rte_log_register() could be deprecated.

No strong opinion on this. @Olivier Matz , Thoughts?

> The other question around log level is about the default level.
> Do we want to allow having different default log levels per log type?

Currently, drivers are using various levels per log type. I can send
the v2 based on
the consensus.


>
>