Re: [PATCH v6 01/12] tools lib traceevent: Convert remaining %p[fF] users to %p[sS]

From: Steven Rostedt
Date: Tue Sep 10 2019 - 15:03:13 EST


On Tue, 10 Sep 2019 11:42:06 -0700
Joe Perches <joe@xxxxxxxxxxx> wrote:

> On Tue, 2019-09-10 at 14:26 -0400, Steven Rostedt wrote:
> > On Tue, 10 Sep 2019 10:18:44 -0700
> > Joe Perches <joe@xxxxxxxxxxx> wrote:
> >
> > > > It's not just for the lastest kernel. We must maintain backward
> > > > compatibility here too. If there use to be a usage of this, then we
> > > > must keep it until the kernels are no longer used (perhaps 7 years?)
> > >
> > > That argues for not using "%pfw" at all for some number of years.
> > >
> > > Perhaps the '%pfw' should be '%pnfw' for 'name' and 'fwnode'
> >
> > -ENOCOMPREHENSION
>
> Perhaps you were not copied on the whole series.
>
> https://lore.kernel.org/lkml/20190910084707.18380-1-sakari.ailus@xxxxxxxxxxxxxxx/

Thanks for the link.

>
> As I understand it, Sakair Ailus is proposing to
> obsolete the current vsprintf "%p[Ff]" extension
> and replace the usage with a new "%pfw" extension
> which would emit the name of a pointer to "struct fwnode {}".
>
> https://lore.kernel.org/lkml/20190910084707.18380-10-sakari.ailus@xxxxxxxxxxxxxxx/
>
> If reusing "%pf<foo>" is a problem, then instead
> it might be reasonable to have a new "%pn<foo>" for
> that use instead.
>
> btw:
>
> Is there kernel version information available in
> trace output files?

Not really. This is just a library that parses the trace event formats,
there's not kernel versions passed in, but we do use variations in
formats and such to determine what is supported.

>
> If so, it might be reasonable to change the tooling
> there instead.
>

Actually, I think we could just look to see if "%pfw" is used and fall
to that, otherwise consider it an older kernel and do it the original
way.

-- Steve