[Skiboot] [PATCH 3/6] flash: Fix 39d9e14 fail to erase flash before write

[Cyril Bur]

On Wed, 27 Jul 2016 18:34:40 +1000
Stewart Smith <stewart at linux.vnet.ibm.com> wrote:

> Cyril Bur <cyrilbur at gmail.com> writes:
> > On Tue, 26 Jul 2016 13:49:32 +1000
> > Cyril Bur <cyril.bur at au1.ibm.com> wrote:
> >  
> >> This causes NVRAM writes to flash on OpenPOWER platforms to be
> >> unreliable.
> >> 
> >> Fixes: 39d9e14 ("libflash: start using the blocklevel interface.")
> >> Signed-off-by: Cyril Bur <cyril.bur at au1.ibm.com>  
> >
> > Hi, I was reviewing the 64bit blocklevel series from Mikey and I've realised
> > that libflash.c registers flash_smart_write() as the handler for
> > blocklevel_write(). This actually means there is no bug here,
> > flash_smart_write() does the same thing as blocklevel_smart_write().
> >
> > I'm pretty sure this is the case because 39d9e14 ("libflash: start using the
> > blocklevel interface.") predates blocklevel being smart.
> >
> > Is it time to move 'smarts' out of the backends and trust blocklevel to do both
> > ECC and pre write erasing?
> >
> > Sorry for the noise  
> 
> okay, so i'll ignore this particular patch in the series.

For now yes but I am of the opinion we might want this at some stage (as part
of a larger series probably) to move smarts out of backends and into blocklevel.

> 
> others good to go though?

Absolutely!

>