On Tue, Feb 18, 2020 at 09:18:57AM -0800, Guenter Roeck wrote:

Hi Greg,

please apply the following two patches to v5.4.y and v5.5.y to fix a problem when trying to boot various pxa machines from MMC.

d3a5bcb4a17f gpio: add gpiod_toggle_active_low() 9073d10b0989 mmc: core: Rework wp-gpio handling

The second patch fixes the problem, the first patch is necessary for the second patch to compile.

Background: Commit 9073d10b0989 claims "No functional changes intended". However, it does include the following functional code change.

--- a/drivers/mmc/host/pxamci.c +++ b/drivers/mmc/host/pxamci.c @@ -740,16 +740,16 @@ static int pxamci_probe(struct platform_device *pdev) goto out; }

•           if (!host->pdata->gpio_card_ro_invert)
  

•                   mmc->caps2 |= MMC_CAP2_RO_ACTIVE_HIGH;
  

•           ret = mmc_gpiod_request_ro(mmc, "wp", 0, 0, NULL);
            if (ret && ret != -ENOENT) {
                    dev_err(dev, "Failed requesting gpio_ro\n");
                    goto out;
            }
  

•           if (!ret) {
  

•           if (!ret)
                    host->use_ro_gpio = true;
  

•                   mmc->caps2 |= host->pdata->gpio_card_ro_invert ?
  

•                           0 : MMC_CAP2_RO_ACTIVE_HIGH;
  

•           }
  
  

This _is_ a functional change: Previously, if there was no "ro" gpio pin, caps2 was never updated to active-high. This can have the practical effect of making the the card read-only, thus preventing the system from booting if it was mounted (and expected to be mounted) read-write. This is seen when trying to boot "spitz" and similar qemu machines from mmc.

I bisected the problem to commit c914a27c92f91 ("mmc: pxamci: Support getting GPIO descs for RO and WP), affecting v5.0 and later kernels.

Looks like Sasha just beat me to this, thanks for letting us know about this.

greg k-h