[U-Boot] [PATCH] mmc: mv_sdhci: Fix host version read for Armada100

[Marek Vasut <marek.vasut@gmail.com>]

> ----- "Marek Vasut" <marek.vasut@gmail.com> wrote:
> > > sdhci_readw does not work for host version read in Armada100 series
> > > SoCs. This patch fix this issue by making a sdhci_readl call to get
> > 
> > host
> > 
> > > version.
> > > 
> > > Signed-off-by: Ajay Bhargav <ajay.bhargav@einfochips.com>
> > > ---
> > > 
> > >  drivers/mmc/mv_sdhci.c |    6 ++++++
> > >  1 files changed, 6 insertions(+), 0 deletions(-)
> > > 
> > > diff --git a/drivers/mmc/mv_sdhci.c b/drivers/mmc/mv_sdhci.c
> > > index f92caeb..c7fd287 100644
> > > --- a/drivers/mmc/mv_sdhci.c
> > > +++ b/drivers/mmc/mv_sdhci.c
> > > @@ -30,6 +30,7 @@ static inline void mv_sdhci_writeb(struct
> > 
> > sdhci_host
> > 
> > > *host, u8 val, int reg) #endif /* CONFIG_MMC_SDHCI_IO_ACCESSORS */
> > > 
> > >  static char *MVSDH_NAME = "mv_sdh";
> > > 
> > > +
> > > 
> > >  int mv_sdh_init(u32 regbase, u32 max_clk, u32 min_clk, u32 quirks)
> > >  {
> > >  
> > >  	struct sdhci_host *host = NULL;
> > > 
> > > @@ -48,7 +49,12 @@ int mv_sdh_init(u32 regbase, u32 max_clk, u32
> > 
> > min_clk,
> > 
> > > u32 quirks) mv_ops.write_b = mv_sdhci_writeb;
> > > 
> > >  	host->ops = &mv_ops;
> > >  
> > >  #endif
> > > 
> > > +#ifdef CONFIG_ARMADA100
> > > +	/* SDHCI host version read workaround for Armada100 series */
> > > +	host->version = sdhci_readl(host, SDHCI_HOST_VERSION - 2) >> 16;
> > > +#else
> > > 
> > >  	host->version = sdhci_readw(host, SDHCI_HOST_VERSION);
> > > 
> > > +#endif
> > > 
> > >  	add_sdhci(host, max_clk, min_clk);
> > >  	return 0;
> > >  
> > >  }
> > 
> > Basically armada100 can't do 16bit access to that register. Ok, but
> > why does it
> > work for other registers? Also, can you make this a config option like
> > 
> > CONFIG_SDHCI_REG32_QUIRK (invent some better name ;-) ).
> > 
> > Cheers
> 
> Hi Marek,
> 
> Armada100 SOC has this bug for particularly that register. This is
> particularly for that register so Do you think a config is required for
> this?

Well such crippled hardware might appear somewhere else too, so yes. But still, 
it's surprising it's only one register, really. Can you point me to a 
datasheet/errata stating it?

Thanks!