[U-Boot] [v2] armv8: ls1012a: enable/disable eSDHC1 through hwconfig for RDB

[Y.b. Lu]

Hi York,

Thank you for your suggestion.
I sent out v3 patch which is backward-compatible. Please check.

Thanks a lot.


Best regards,
Yangbo Lu

> -----Original Message-----
> From: York Sun
> Sent: 2017年12月1日 1:59
> To: Y.b. Lu <yangbo.lu at nxp.com>; u-boot at lists.denx.de
> Cc: Xiaobo Xie <xiaobo.xie at nxp.com>
> Subject: Re: [v2] armv8: ls1012a: enable/disable eSDHC1 through hwconfig for
> RDB
> 
> On 11/29/2017 08:44 PM, Y.b. Lu wrote:
> > Hi York,
> >
> > I copied hardware team Kinjalk's explain here.
> >
> > "Enabling SDHC2 on ‘00’ is not correct on revision D and later boards as the
> sd wifi is not on there on these revs.
> > The IO expander was designed to override the dip switch values. So, the DIP
> switch values are driven through low strength drivers. The I2C reads of DIP
> switch settings may not be correct and reliable as the noise margin is very low.
> If the IO expander is driving CFG than the situation is different and reads will
> always be reliable."
> >
> > It seemed ls1012ardb RevD reworked the SDHC2 circuit. The SDIO wifi device
> was removed and I2C reading of DIP switch setting was not reliable.
> > This changes is causing many kernel error messages on RevD boards.
> > "mmc1: Controller never released inhibit bit(s)."
> > It's safe to disable SDHC2 in default and enable it manually when hardware
> configuration(DIP switch) is correct for SDHC2 since software doesn’t have
> way to check.
> >
> > Do you think this should be fixed by hardware? If so, I will try to suggest that.
> 
> Yangbo,
> 
> If the hardware can be fixed to operate the same way, please push for a fix. If
> rev D board (and later) removes the feature to read DIP switches, and you can
> make the code backward-compatible, then we can accept this change.
> 
> Your change is to use hwconfig to determine if eSDHC1 is available. This will
> cause all previous boards losing the ability to automatically select eSDHC1. Is
> that acceptable to existing users, and clearly documented?
> 
> York