Re: [DVB patch 51/54] ttpci: av7110: RC5+ remote control support

[Nish Aravamudan <nish.aravamudan@gmail.com>]

On 9/4/05, Johannes Stezenbach <js@linuxtv.org> wrote:
> On Sun, Sep 04, 2005 Nish Aravamudan wrote:
> > On 9/4/05, Johannes Stezenbach <js@linuxtv.org> wrote:
> > > On Sun, Sep 04, 2005 Nish Aravamudan wrote:
> > > > On 9/4/05, Johannes Stezenbach <js@linuxtv.org> wrote:
> > > > >
> > > > > -#define UP_TIMEOUT (HZ/4)
> > > > > +#define UP_TIMEOUT (HZ*7/25)
> > > >
> > > > #define UP_TIMEOUT msecs_to_jiffies(280)
> > > > #define UP_TIMEOUT (7*msecs_to_jiffies(40)
> > >
> > > I agree it's nicer to read, but AFAIK not required for correctness?
> > > If so, then we'll fix those up in linuxtv.org CVS and submit
> > > cleanup patches later.
> >
> > Yeah, it's correct with the three current values of HZ (100, 250 and
> > 1000), but if you try a not-so-clean value (like Con did with 864, or
> > something), you might run into rounding issues. msecs_to_jiffies()
> > should take care of them (or will be a single point to do so
> > eventually).
> 
> Well, if msecs_to_jiffies() is the new way of specifying timeouts
> we'd have a lot more to fix up in our tree. But something like
> a remote control key-up timeout doesn't need much precision.
> Generally I see nothing wrong with HZ/4, but something like
> HZ*20/1000 could be problematic with small or odd HZ values.

This discussion comes up every time ;) HZ/4 rounds incorrectly
(depends on your perspective, I guess) with HZ=250.

> Agreed? Or is it desired that people generally use msecs_to_jiffies()?

I agree, generally it's ok, as timeouts aren't meant to be precise. I,
personally, am converting code over to msecs_to_jiffies() as I come to
it, just so that the conversions are consistent (a number in the tree
were not, not that yours are) and basically self-commented. And, like
I mentioned, they are automatically rounded correctly with "strange"
values of HZ.

The changes I mentioned in your patchset are generally not critical,
but just comments and my own preferences.

Thanks,
Nish
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/