Re: [RFC PATCH v2 01/10] perf workqueue: threadpool creation and destruction

From: Riccardo Mancini
Date: Wed Aug 11 2021 - 13:55:34 EST


On Tue, 2021-08-10 at 17:24 -0300, Arnaldo Carvalho de Melo wrote:
> Em Tue, Aug 10, 2021 at 11:54:19AM -0700, Namhyung Kim escreveu:
> > On Mon, Aug 9, 2021 at 3:30 AM Riccardo Mancini <rickyman7@xxxxxxxxx> wrote:
> > > On Fri, 2021-08-06 at 19:24 -0700, Namhyung Kim wrote:
> > > > > +/**
> > > > > + * threadpool__strerror - print message regarding given @err in @pool
> > > > > + *
> > > > > + * Buffer size should be at least THREADPOOL_STRERR_BUFSIZE bytes.
> > > > > + */
> > > > > +int threadpool__strerror(struct threadpool *pool __maybe_unused, int
> > > > > err,
> > > > > char *buf, size_t size)
> > > > > +{
> > > > > +       char sbuf[STRERR_BUFSIZE], *emsg;
> > > > > +
> > > > > +       emsg = str_error_r(err, sbuf, sizeof(sbuf));
> > > > > +       return scnprintf(buf, size, "Error: %s.\n", emsg);
> > > > > +}
> > > > > +
> > > > > +/**
> > > > > + * threadpool__new_strerror - print message regarding @err_ptr
> > > > > + *
> > > > > + * Buffer size should be at least THREADPOOL_STRERR_BUFSIZE bytes.
> > > > > + */
> > > > > +int threadpool__new_strerror(struct threadpool *err_ptr, char *buf,
> > > > > size_t
> > > > > size)
> > > > > +{
> > > > > +       return threadpool__strerror(err_ptr, PTR_ERR(err_ptr), buf,
> > > > > size);
> > > > > +}
>
> > > > Why two different functions?
>
> > > Since when new fails you don't have a err number, just an err_ptr so it's
> > > not
> > > very clear how to call threadpool__strerror. Therefore I made a wrapper to
> > > remove any ambiguity.
> >
> > What do you mean by "when new fails"?
>
> I think 'new' is 'constructor', i.e. something__new() returns a newly
> created object and this not an error number, so he uses ERR_PTR() and
> then he needs to pass it to the 'strerror' specific to the
> threadpool__new, which will use PTR_ERR() to get an integer, and then
> map that to a proper error string, right?

Correct.
threadpool__new_strerror is not really needed since one could use
threadpool__strerror directly, but then I would need to handle all possible ways
it could be called (e.g. (NULL, PTR_ERR(err_ptr)), (err_ptr, 0), (err_ptr,
PTR_ERR(err_ptr)), so I thought it was better to just provide a strerror that
only took the err_ptr.
Maybe I can provide it as an inline in the header, or as a macro.

Riccardo

>
> - Arnaldo