RE: [PATCH 1/3] mtd: spi-nor: Add die_cnt field and flags

["Krzeminski, Marcin (Nokia - PL/Wroclaw)" <marcin.krzeminski@nokia.com>]

Cyrille,

....
> >
> > If think you should first check whether nor->erase() is implemented by
> > the SPI controller driver and call it instead of nor->write_reg().
> >
> > For instance, some SPI controllers are mapped to the system memory to
> > speed up Fast Read operations. In which case, they could also enable
> > data cache to fasten even more the read operations.
> >
> > So when the spi-nor framework calls nor->erase() or nor->write(),
> > those handlers might need to invalidate the cache before performing
> > the next read.
> >
> > This is just an example but generally speaking, don't forget the
> > nor->erase() handler! :)
> >
> Thanks for extensive clarification, I will include this in v2.
> 
> Regards,
> Marcin
> 
I've just sent v2 with added check for nor->erase() handler.
As we talked on IRC, with the example you gave above there is
a bug in case of chip erase command, because the driver will not
know that whole flash has been erased.
Maybe there is a need to have a handler in spi-nor framework
for memory invalidation?

Thanks,
Marcin