[U-Boot] mmc: sdhci: Add a quirk to add delay during completion of sdhci_send_cmd

From: Lukasz Majewski <l.majewski@samsung.com>
To: u-boot@lists.denx.de
Subject: [U-Boot] mmc: sdhci: Add a quirk to add delay during completion of sdhci_send_cmd
Date: Mon, 08 Apr 2013 09:35:55 +0200	[thread overview]
Message-ID: <20130408093555.35ec3fa0@amdc308.digital.local> (raw)
In-Reply-To: <5162429C.7040400@samsung.com>

Dear All,

> Hi,
> 
> On 04/08/2013 12:52 PM, Tushar Behera wrote:
> > On 04/07/2013 10:27 AM, Jagan Teki wrote:
> >> Hi,
> >>
> >> I saw that you have been added the SDHCI_QUIRK_WAIT_SEND_CMD on
> >> below commit
> >> http://git.denx.de/?p=u-boot.git;a=commitdiff;h=13243f2eafc4292917178051fe1bb5aab2774dca
> >>
> >> I need few quires regarding the QUIRK delay.
> >> 1. Why the delay is 1000
> >> +       if (host->quirks & SDHCI_QUIRK_WAIT_SEND_CMD)
> >> +               udelay(1000);
> >> +
> >> 2. is this delay specific to s5p_sdhci  controller?
> > 
> > This was specific to s5p_sdhci driver so as to replicate the
> > behavior of s5p_mmc driver.
> It didn't be denoted to use the udelay(1000) in spec.
> But To ensure the completing previous request, it has added the
> udelay(1000).

But with some memory chips/workloads this udelay is indeed too short
and "Controller never released inhibit bit(s)" appears.

> > 
> >> 3. I have an issue "Controller never released inhibit bit(s)"
> >>    but when I enable this quirk on my driver with udelay(10000),
> >> it's working.
> udelay(10000)? It's too long delay time for completing the request.

When I recently has been experimenting with ext4 it turned out, that for
some load this value is not enough. Moreover I'm not quite sure if
after several such errors we can still count that sent data are really
stored on the chip.

> Which SoC do you use?
> 
> If we can use the get_timer() for busy-waiting, i think that we can
> remove the udelay().

I think that we shall try to improve the sdhci framework instead. 

I've described this problem some time ago, but no relevant solution was
devised:

http://article.gmane.org/gmane.comp.boot-loaders.u-boot/150636/match=rfc+mmc+fix+increase+timeout+value+sdhci_send_command

Maybe now we can come up with some solution.

> 
> Best Regards,
> Jaehoon Chung
> > 
> > I am not sure about the exact delay as per the spec. I have added
> > Jaehoon to CC who might have additional information about this.
> > 
> >>
> >> Could you please help me.
> >>
> >> Thanks,
> >> Jagan.
> >>
> > 
> > 
> 
> _______________________________________________
> U-Boot mailing list
> U-Boot at lists.denx.de
> http://lists.denx.de/mailman/listinfo/u-boot

-- 
Best regards,

Lukasz Majewski

Samsung R&D Poland (SRPOL) | Linux Platform Group