Re: mmc: tmio: why enable/disable SDIO irq on every transaction with IOMOD?

[Yasushi SHOJI <yashi@atmark-techno.com>]

Hi Wolfram,

On Thu, 01 Dec 2016 19:46:22 +0900,
Wolfram Sang wrote:
> 
> > > So my question is that "What is the reason behind to disable IRQ with
> > > SDIO_MODE?  Is there any situation which masking with SDIO_INFO1_MASK
> > > is not enough?
> 
> The code was introduced with 845ecd20239c28 ("mmc: tmio_mmc: implement
> SDIO IRQ support") which was in 2010. I don't have that old datasheets
> to check if the SCLKDIVEN restriction was already present in the SDHI
> cores which were available back then.

Is Arnd Hannemann, the author, around these days by any chance?

Or does anyone on the list have a shareable old datasheet?  The one we
have is the one for R-Mobile A1 and has CONFIDENTIAL on it.

> My assumption is that it was not, or it was overlooked. So, it might be
> just for completeness that not only the individual IRQs have been
> disabled but also the big master switch (IOMOD) was turned off.
> 
> My further assumption is that it is very likely good enough to disable
> the individual IRQs. If we can't guarantee the conditions to set IOMOD,
> it seems okay to me to just leave it.

Yes that's what we think and what we see with our test code, which is
just the original without IOMOD removed.

> Do you have a patch which works for you?

We don't have it yet.  A question we have is that where should we
enable / disable the IOMOD?

Do we have to keep it disabled while the controller is in non-SDIO?
The spec says that we don't get SDIO irq while in non-SDIO mode.  But
should we trust?

Or, should we disable it when we switch back to non-SDIO mode and
enable when we detect SDIO?

Please enlighten us this area.  We'll submit the proper fix once we
know how we should do it.

Thanks,
-- 
            yashi