[PATCH 3/3] h1940: implement mmc_power function

[marek.vasut@gmail.com (Marek Vasut)]

Dne ?t 24. srpna 2010 16:56:06 Vasily Khoruzhick napsal(a):
> ? ????????? ?? 24 ??????? 2010 17:41:57 ????? Marek Vasut ???????:
> > Dne ?t 24. srpna 2010 16:21:01 Vasily Khoruzhick napsal(a):
> > > ? ????????? ?? 24 ??????? 2010 17:01:12 ????? Marek Vasut ???????:
> > > > pr_err() and don't probe the device for which you couldn't request
> > > > GPIO.
> > > 
> > > Well, handling those errors in machine init functions doesn't make
> > > sense. PDA is not usable without basic hardware.
> > 
> > You can use PDA without LCD for example (there are such people, trust
> > me). So, this comment really isn't too valid. You can also use PDA
> > without MMC support, you still have onboard flash, etc. etc.
> 
> But error handling here just increases code size, and gpio_request in init
> function should always succeed. It means something's going really wrong
> (kernel bug or misconfiguration) if gpio_request fails. Btw, Ben suggested
> to use WARN_ON in machine init, look through init functions of qt2410,
> n30, rx1950, vr1000, etc.

See my previous response. It'd increase code size, but if you mark that code 
__init, it'll be freed after that function finishes so it doesn't matter.

If something wrong goes on, you can still debug it without for example MMC card 
(though if you continued probing the MMC driver, it might cause additional 
unwanted noise and make the debugging harder). That's the whole point.

Cheers
> 
> Regards
> Vasily