From mboxrd@z Thu Jan  1 00:00:00 1970
From: linux@arm.linux.org.uk (Russell King - ARM Linux)
Date: Fri, 23 Nov 2012 12:55:07 +0000
Subject: [PATCH 3/3] MMC: sdhci-dove: allow GPIOs to be used for card
 detection on Dove
In-Reply-To: <20121123121422.GA26168@S2100-06.ap.freescale.net>
References: <20121122235426.GT3332@n2100.arm.linux.org.uk>
 <E1TbgcZ-0007Kt-35@rmk-PC.arm.linux.org.uk>
 <20121123083117.GA25754@S2100-06.ap.freescale.net>
 <20121123085726.GK5764@n2100.arm.linux.org.uk>
 <20121123121422.GA26168@S2100-06.ap.freescale.net>
Message-ID: <20121123125507.GO5764@n2100.arm.linux.org.uk>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Fri, Nov 23, 2012 at 08:14:27PM +0800, Shawn Guo wrote:
> On Fri, Nov 23, 2012 at 08:57:26AM +0000, Russell King - ARM Linux wrote:
> > Does this work with the sdhci stuff?
> 
> Honestly, I'm not sure, but I guess it does, since I have seen
> sdhci-pxav3 driver using the helpers.  Anyway, I'm going to adopt
> the helpers for sdhci-esdhc-imx driver to find it out.

The thing that worries me is this:

static void sdhci_tasklet_card(unsigned long param)
{
...
        /* Check host->mrq first in case we are runtime suspended */
        if (host->mrq &&
            !(sdhci_readl(host, SDHCI_PRESENT_STATE) & SDHCI_CARD_PRESENT)) {
                pr_err("%s: Card removed during transfer!\n",
                        mmc_hostname(host->mmc));
                pr_err("%s: Resetting controller.\n",
                        mmc_hostname(host->mmc));

                sdhci_reset(host, SDHCI_RESET_CMD);
                sdhci_reset(host, SDHCI_RESET_DATA);

                host->mrq->cmd->error = -ENOMEDIUM;
                tasklet_schedule(&host->finish_tasklet);
        }
...
        mmc_detect_change(host->mmc, msecs_to_jiffies(200));
}


That gets called whenever a card is inserted/removed by the SDHCI code (if
the SDHCI card detect is wired), or in my case by the interrupt routine
the code in my patch adds.

The slot-gpio.c stuff directly calls into mmc_detect_change (a) with a
shorter delay, and (b) completely omits the above handling and resetting.
My guess from the above code is that it'll work fine 90% of the time
(because you'll remove the card without an active request), but the above
code looks like it's addressing a corner case which will be omitted with
the "generic" slot-gpio.c solution.

So I don't think it's a good idea to use slot-gpio.c in this case.