Hi Chris
On Mon, 17 Sep 2012, Chris Ball wrote:
> Hi Guennadi,
>
> I'm having trouble using a cd-gpio. After I request it, I get the
> IRQ and mmc_rescan() triggers successfully, but mmc_attach_sd()
> fails because every command returns -123 (ENOMEDIUM), due to:
>
> sdhci.c:
> /* If polling, assume that the card is always present. */
> if (host->quirks & SDHCI_QUIRK_BROKEN_CARD_DETECTION)
> present = true;
> else
> present = sdhci_readl(host, SDHCI_PRESENT_STATE) &
> SDHCI_CARD_PRESENT;
>
> if (!present || host->flags & SDHCI_DEVICE_DEAD) {
> host->mrq->cmd->error = -ENOMEDIUM;
> tasklet_schedule(&host->finish_tasklet);
> } else {
> ...
>
> "present" is false because SDHCI_PRESENT_STATE doesn't have the presence
> bit sent (if it did, we wouldn't need a cd-gpio, right?) and the command
> fails.
>
> And we can't just set SDHCI_QUIRK_BROKEN_CARD_DETECTION, because that
> would turn on polling, which we don't want. How's this supposed to work?
Sorry, I'm not very familiar with the SDHCI hardware or the driver, so,
I'll be speculating here.
cd-gpio can function in two modes: polling and IRQ. IRQ is enabled if a
number of conditions is satisfied: gpio_to_irq() returns a non-negative
number, the user doesn't force polling by setting MMC_CAP_NEEDS_POLL and
request(_threaded)_irq() is successful. If either of the three conditions
fail, GPIO polling will be used.
Now, concerning SDHCI. cd-gpio doesn't magically make all MMC drivers
capable of using GPIO card detection, sorry:-) To use it you should first
make sure your driver can take card-present or card-detect information
from a GPIO. Then the slot-gpio module shall be used to add this GPIO CD
functionality to your driver instead of cooking it up by yourself.
>From the above it seems to me, that your SDHCI instance is not yet quite
trained to take GPIO input for card-detection:-) Looking through various
SDHCI implementations I see several drivers already implementing and using
various GPIO slot services internally: tegra, spear, s3c, pci, esdhc-imx.
They all seem to work with their own GPIO CD implementations, which are
very similar to the slot-gpio one and it shouldn't be too difficult to
replace them with the latter.
Concerning how specifically it should work in your case - I'm not sure,
sorry. What platform are you working with? Does it already have GPIO
support? If it did, it should have been pretty simple to replace it with
the generic one from slot-gpio. Looking at the code, it seems to me, that
even on SDHCI systems, where a GPIO is used for card detection, the
controller is sometimes still able to recognise the card's presence. This
seems to be the case on tegra, pci (except cafe),... However, e.g., s3c
uses a dirty hack of setting and clearing
SDHCI_QUIRK_BROKEN_CARD_DETECTION in their CD routine... In general, I
think, it should suffice to do in sdhci_request() (and others) something
like
/* If polling, assume that the card is always present. */
if (host->quirks & SDHCI_QUIRK_BROKEN_CARD_DETECTION)
present = true;
else
present = sdhci_readl(host, SDHCI_PRESENT_STATE) &
SDHCI_CARD_PRESENT;
if (!present) {
int ret = mmc_gpio_get_cd(host->mmc);
if (ret > 0)
present = true;
}
Also, once a card has been detected, shouldn't it be possible to use
mmc_card_present() to check its presence?
Thanks
Guennadi
> Maybe your system was differently broken, and we need some new way to
> communicate that neither the PRESENT bit nor polling should be used?
>
> Thanks,
>
> - Chris.
> --
> Chris Ball <cjb@xxxxxxxxxx> <http://printf.net/>
> One Laptop Per Child
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
