[i2c] [patch 2.6.25-rc8] i2c: fix platform driver hotplug/coldplug

[David Brownell]

On Thursday 10 April 2008, Andrew Morton wrote:
> On Thu, 10 Apr 2008 09:49:09 -0700 David Brownell <david-b at pacbell.net> wrote:
> 
> > It was a regresssion as of 2.6.24, for 200+ drivers;
> > every platform_bus driver used have hotplug/coldplug
> > support, but that patch stopped that from working.
> > 
> > The thing is, that had only been working for a few
> > releases at that time, so not many folk were relying
> > on it yet.  They would just have been able to start
> > doing so when that mechanism stopped working.
> > 
> > So I can't see a rush to get this into 2.6.25 now.
> > I think the 200+ drivers will hotplug/coldplug again
> > in 2.6.26 ... I've pushed almost a hundred such driver
> > updates by now (sigh).
> > 
> 
> Oh.  I was thinking we should slam all these fixes into 2.6.25.

I would not object to that.  I can't see the patches I
signed off on could break anything, and I dislike the
notion of regression fixes getting deferred even longer.

(That includes I2C, RTC, SPI, USB, and Watchdog patches
I posted yesterday, and the earlier EHCI one ... fixing
about 90 drivers, and those entire subsystems.)

It's just that one kind of gets the feeling 2.6.25-final
was just around the corner... and there's still the 122
drivers from Kay's patch to review, plus the ones missing
from those patches.  (Another 15-20 percent if numbers
from the subsystems I looked at continue to hold.)


> If a couple fo them cause problems then we can patch things up in
> 2.6.25.x, but that's beter than having hundreds of broken drivers.
> 
> Or does your "as of" above mean that 2.6.24 is also busted?

Yes, that "platform:" change broke hotplug/coldplug for
most platform_bus drivers on 2.6.24 kernels too.

- Dave