Skip to contentSkip to conversation

How to trigger sdio bus probing from user space

DISCUSSION Bojan Prtvar 2 participants 3 messages Sep 27, 2013 – Sep 30, 2013 MMC, SD and SDIO

lore
Share & tools
lore
#0 lore raw

Hi,

Is there a clean way to trigger sdio bus probing from user space?

I have a non removable sdio WiFi/BT module on embedded system that needs to be tested and calibrated on production line.
In order to speed up production process, I need to be able to replace the WiFi/BT modules in run time without reboot.
Modules are connected to the host with a ribbon cable, not with a sdio connector.

I came up with fallowing two-part solution:
a)
I extended WiFi driver's /proc interface with a call to mmc_detect_change()
b)
I also needed to let the sdio host think module is removable by commenting fallowing in sdhci-mv.c:
sdhci_mv_probe_dt(struct platform_device *pdev)

{
    struct device_node *np = pdev->dev.of_node;
    struct sdhci_host *host = platform_get_drvdata(pdev);

    if (of_get_property(np, "mrvl,card-wired", NULL)) {
        /*
         * RT-RK
         * We need to comment out these in order to be able to
         * trigger sdio bus probing on the fly
         */
        //host->quirks |= SDHCI_QUIRK_BROKEN_CARD_DETECTION;
        //host->mmc->caps |= MMC_CAP_NONREMOVABLE;
    }

This works in situations when working modules are inserted. Unfortunately, it does not work in case when broken module is inserted on production line.
In this case production board needs to be rebooted, because the WiFi driver's /proc interface is gone, and I can not force a call to mmc_detect_change() any more.

Is there cleaner and better approach?

Best regards,
Bojan

#1 lore raw
Quoted 32 lines · Bojan Prtvar

Is there a clean way to trigger sdio bus probing from user space?

I have a non removable sdio WiFi/BT module on embedded system that needs to be tested and calibrated on production line.
In order to speed up production process, I need to be able to replace the WiFi/BT modules in run time without reboot.
Modules are connected to the host with a ribbon cable, not with a sdio connector.

I came up with fallowing two-part solution:
a)
I extended WiFi driver's /proc interface with a call to mmc_detect_change()
b)
I also needed to let the sdio host think module is removable by commenting fallowing in sdhci-mv.c:
sdhci_mv_probe_dt(struct platform_device *pdev)

{
    struct device_node *np = pdev->dev.of_node;
    struct sdhci_host *host = platform_get_drvdata(pdev);

    if (of_get_property(np, "mrvl,card-wired", NULL)) {
        /*
         * RT-RK
         * We need to comment out these in order to be able to
         * trigger sdio bus probing on the fly
         */
        //host->quirks |= SDHCI_QUIRK_BROKEN_CARD_DETECTION;
        //host->mmc->caps |= MMC_CAP_NONREMOVABLE;
    }
Quoted 8 lines

This works in situations when working modules are inserted. Unfortunately, it does not work in case when broken module is inserted on production line.
In this case production board needs to be rebooted, because the WiFi driver's /proc interface is gone, and I can not force a call to mmc_detect_change() any more.

Is there cleaner and better approach?

For internal use I have added a 'rescan' sysfs entry to the mmc controller, instead of the sdio device itself. It is quite similar to the way it is done in the scsi subsystem, and uses the same 'mmc_detect_change' function as you do .

Updated to use the new attribute groups syntax for 3.12, I obtain the following patch - compile-tested only, nothing more, as I do not have a 3.12 test system available.

---

drivers/mmc/core/host.c +17 -0
index 49bc403..b5b0811 100644
@@ -39,9 +39,26 @@ static void mmc_host_classdev_release(struct device *dev)
        kfree(host);
 }
 
+static ssize_t rescan_store(struct device *dev, struct device_attribute *attr,
+                            const char *buf, size_t count)
+{
+       struct mmc_host *host = cls_dev_to_mmc_host(dev);
+       mmc_detect_change(host, 0);
+       return count;
+}
+
+static DEVICE_ATTR_WO(rescan);
+
+static struct attribute *mmc_host_attrs[] = {
+       &dev_attr_rescan.attr,
+       NULL,
+};
+ATTRIBUTE_GROUPS(mmc_host);
+
 static struct class mmc_host_class = {
        .name           = "mmc_host",
        .dev_release    = mmc_host_classdev_release,
+       .dev_groups     = mmc_host_groups,
 };
 
 int mmc_register_host_class(void)
Signature · 1 line
Romain Izard
#2 lore raw
Quoted 9 lines · Romain Izard <romain.izard.pro <at> gmail.com> writes:

For internal use I have added a 'rescan' sysfs entry to the mmc controller, instead of the sdio device itself. It is quite similar to the way it is done in the scsi subsystem, and uses the same 'mmc_detect_change' function as you do .

Updated to use the new attribute groups syntax for 3.12, I obtain the following patch - compile-tested only, nothing more, as I do not have a 3.12 test system available.

Hi Romain,

Than you. This is exactly what I was looking for. My kernel version is 3.4.19 so I had back port your patch a little. At the and it is working OK.

Regards,
Bojan