On Mon, Aug 20, 2018 at 12:12:47PM +1000, Benjamin Herrenschmidt wrote:
> On Sun, 2018-08-19 at 09:46 -0700, Linus Torvalds wrote:
> > On Sun, Aug 19, 2018 at 6:51 AM Sedat Dilek <sedat.di...@credativ.de> wrote:
> > >
> > > This can be dropped with commit 771c035372a036f83353eef46dbb829780330234
> > > ("deprecate the '__deprecated' attribute warnings entirely and for good")
> > > now in upstream.
> >
> > Could we please just remove the __deprecated use in i2c entirely?
> >
> > As fat as I can tell, it's used for one thing, which is the
> > "attach_adapter" function pointer member in 'struct i2c_driver'.
> >
> > And I think there is _one_ remaining driver using it, namely the
> > powermac "therm_windtunnel" driver.
> >
> > Could somebody who knows the i2c probing just a bit convert that *one*
> > remaining driver, then remove the "attach_adapter" entirely, and also
> > remove this thing?
> >
> > Alternatively, perhaps the driver should be removed? The last actual
> > powermac change to that driver seems to be from August 2009.
> > Everything since has been just cleanup unrelated to that driver code
> > itself (ie spelling fixes, automated coccinelle scripts etc).
> >
> > Added a few people either because they're still officially maintainers
> > of that file, or because they touched the code ten years ago.
>
> Some people still use those old machines, I'll give a try at fixing the
> driver.
I did this here a while ago [1] ("macintosh: therm_windtunnel: drop
using attach adapter")...
> Paul: Do you still have one of these machines ?
... but this was the main problem. I couldn't find anyone to test this.
Even asked on PPC lists of various distributions.
Note that we did apply the same approach of patch [1] already in commit
ac397c80de89 ("ALSA: ppc: keywest: drop using attach adapter"). This has
been tested on HW. I also have a hack to make this driver probe on one
of my ARM boards. Of course, no real HW present there, but shows that
the new probing works at least there.
This is very interesting timing, too. Because I wanted to use the
tentative version jump to 5.0 to finally get rid of this deprecated
feature independently of the discussion here. Which would have meant
resending the patch and arguing we have enough indications that the
patch [1] is sane. And then I could finally remove this callback.
Thanks,
Wolfram
[1] http://patchwork.ozlabs.org/patch/461759/
signature.asc
Description: PGP signature
