On Thu, May 28, 2020 at 04:35:41PM +0530, Vignesh Raghavendra wrote:
>
>
> On 28/05/20 4:12 pm, Serge Semin wrote:
> [...]
> >>> +
> >>> +static map_word __xipram bt1_rom_dummy_read(struct map_info *map,
> >>> + unsigned long ofs)
> >>> +{
> >>> + map_word ret;
> >>> +
> >>> + ret.x[0] = 0xFF;
> >>> +
> >>> + return ret;
> >>> +}
> >> Why define dummy_io for "baikal,bt1-boot-rom"? I don't see any use of
> >> adding a driver that always reads 0xFFs
> > This is supposed to be temporary solution for the baikal,bt1-boot-rom ROM.
> > The Boot ROM mirror might reflect either an embedded firmware or the SPI
> > flash
> > directly mapped into the memory. In former case there is no problem, we can
> > freely read from the Boot ROM region. But in the later case the mirrored
> > region
> > (memory mapped SPI flash) is not always accessible. If normal Boot SPI
> > controller is enabled, then the SPI flash mapping isn't accessible (any
> > attempt
> > will cause a bus-error). In order to fix this we'd need a mutual exclusive
> > lock,
> > which would disable the SPI controller while the mirrored memory mapped SPI
> > flash
> > region needs to be accessed. Such mechanism isn't currently implemented,
> > but it
> > will in be in the framework of my patch created for the SPI subsystem.
> >
> > My idea was to just provide a dummy callback for now and replace it with
> > normal
> > IO-methods with mutual exclusive lock-unlocks when the corresponding SPI
> > driver
> > is accepted.
> >
>
> Yes I understand, but I don't see any advantage for end user by
> providing dummy calls... So please don't add the callbacks until they
> implement proper functionality
Agreed. I'll remove the Baikal-T1 Boot ROM support in the next version.
-Sergey
