[bug report] mmc: sdhci: use WP GPIO in sdhci_check_ro()

[Dan Carpenter <dan.carpenter@xxxxxxxxxx>]

Hello Thomas Petazzoni,

The patch 6d5cd068ee59: "mmc: sdhci: use WP GPIO in sdhci_check_ro()"
from Feb 12, 2019, leads to the following static checker warning:

	drivers/gpio/gpiolib.c:3287 gpiod_get_value_cansleep()
	warn: sleeping in atomic context

drivers/mmc/host/sdhci.c
  2424  static int sdhci_check_ro(struct sdhci_host *host)
  2425  {
  2426          unsigned long flags;
  2427          int is_readonly;
  2428  
  2429          spin_lock_irqsave(&host->lock, flags);
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
We're holding a spin lock.

  2430  
  2431          if (host->flags & SDHCI_DEVICE_DEAD)
  2432                  is_readonly = 0;
  2433          else if (host->ops->get_ro)
  2434                  is_readonly = host->ops->get_ro(host);
  2435          else if (mmc_can_gpio_ro(host->mmc))
  2436                  is_readonly = mmc_gpio_get_ro(host->mmc);
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^
The mmc_gpio_get_ro() function calls gpiod_get_value_cansleep() which,
as the name suggests, can sleep.

  2437          else
  2438                  is_readonly = !(sdhci_readl(host, SDHCI_PRESENT_STATE)
  2439                                  & SDHCI_WRITE_PROTECT);
  2440  
  2441          spin_unlock_irqrestore(&host->lock, flags);
  2442  
  2443          /* This quirk needs to be replaced by a callback-function later */
  2444          return host->quirks & SDHCI_QUIRK_INVERTED_WRITE_PROTECT ?
  2445                  !is_readonly : is_readonly;
  2446  }

regards,
dan carpenter