Forums

I assume this is with respect to OMAP3530 since your previous post was related to OMAP3530.  It's a good practice to always include that info in a new thread.

I've seen similar behavior even with later kernels.  Can you do something like this:

• insmod your gadget driver. 
• Now the USB is "ready" for when something gets plugged in.
• Device gets plugged in
• Device gets removed
• rmmod/insmod to "re-arm" the USB

Are you building the kernel with OTG support turned on?  The behavior I saw was definitely with OTG support enabled.  It may have behaved different when I just built it as a straight-up host or device (sorry, it was several months ago!).

Do you need to support both host and device modes?  If not, which one do you need?  Can you attach your kernel config (please don't copy/paste into the editor!).

Brad

Duh, sorry, I just noticed that the subject says OMAP3530!

In the sequence you listed, I only get hotplug events when insmod g_ether module (and musb_hdrc module which is a dependency for g_ether) and rmmod g_ether & musb_hdrc modules.  I do NOT get any hotplug event when plugging device into or removing from host PC via USB cable).

• OTG is enabled
• Only using in device mode (connecting to host PC and using Windows RNDIS driver)
• Hotplug events only when g_ether/musb_hdrc modules installed or removed
• Needing hotplug to detect device connection and removal from host PC to start/stop DHCP client and TFTP server via mdev; mdev part is working; just can't get hotplug events from kernel
• For debugging, I've created a bash script that just dumps the env vars to a log file in tmp when a hotplug event occurs 'echo /usr/bin/hpdbg.sh > /proc/sys/kernel/hotplug'

Attached is the

• kernel cfg
• bash shell script that is run when a hotplug event occurs
• log files for hotplug events when installing/removing g_ether/musb_hdrc driver modules

4213.hpdbg_rmmod.log

Attached is an archive of all the files mentioned in the previous post.  I didn't realize forum only allows one file per post.

3660.hotplugging.zip

Found this post indicating that hot plug is not functional in musb driver in kernel ver 2.6.31, but may be fixed in 2.6.33

http://gumstix.8.n6.nabble.com/USB-OTG-not-working-as-host-td662224.html

On Thu, Jan 7, 2010 at 12:46 PM, zachlac <[hidden email]> wrote: 
> 
> I'd like to have two free USB ports, the normal Host port and the OTG port as 
> a second host port.  The first host port works properly, but when I plug a 
> slave device into the OTG port, nothing happens.  No messages on dmesg. 
> It's configured in the kernel as Host through the linux-omap3-2.6.31.bb 
> recipe.  Is there a module I need to load?  I'm using the Overo Water with 
> the Tobi board. 

On the musb (OTG) port devices must be plugged in at boot time to be 
recognized -- hot plug is not yet functional. 

I believe that this should be fixed in the 2.6.33 musb driver, but I 
haven't verified that yet. 

Steve 

That's a good find.  I'm going to move this thread over to our Linux forum so we can get some commentary from the USB developers.

FYI, I'm still trying to get this in front of the right person.

Thanks for the response Ajay.  My only concern is I don't get a hot plug event when the device is initially connected.  I only get hot plug events when the driver is installed and uninstalled.  Does this work around cause a hot plug event for initial device connection?

Ajay,

One of my customers is facing a similar issue, summary of the problem is captured below:

---

We need to get hot-plug event with AM37x EVM as USB Gadget. There is requirement to mount/unmount encrypted file system (dm-crypt) in USB Mass Storage mode on the USB Plugin/out events.

We do NOT get any hotplug event when plugging device into or removing from host PC via USB cable).

• Peripheral mode is enabled
• Only using in device mode (connecting to host PC via HUSB_OTG port and using Windows Mass Storage driver)
• Hotplug events only when g_file_strorage or g_mass_storage modules installed or removed
• Needing hotplug to detect device connection and removal from host PC to mount/unmount via cryptsetup; cryptsetup part is working; just can't get hotplug events from kernel
• For debugging, there is a bash script that just dumps the env vars to a log file when a hotplug event occurs 'echo /usr/hotplug/hpdbg.sh > /proc/sys/kernel/hotplug'

Attached is the:

• kernel cfg
• bash shell script that is run when a hotplug event occurs
• log files for hotplug events when installing/removing g_file_strorage or g_mass_storage driver
• dmesg log

Hi,

I have tried a patch for the usb connection detection and it is working in android.

The following sysfs entry is changed whenever a host/peripheral connected

/sys/devices/platform/musb-omap2430.0/musb-hdrc.0/mode

In kernel, whenever the sysfs entry is changed we pass a uevent with this patch

diff --git a/drivers/usb/musb/musb_core.c b/drivers/usb/musb/musb_core.c
index efef5ff..2c240e1 100644
--- a/drivers/usb/musb/musb_core.c
+++ b/drivers/usb/musb/musb_core.c
@@ -1894,6 +1894,7 @@ static void musb_irq_work(struct work_struct *data)
if (musb->xceiv->state != old_state) {
old_state = musb->xceiv->state;
sysfs_notify(&musb->controller->kobj, NULL, "mode");
+ kobject_uevent(&musb->controller->kobj, KOBJ_CHANGE);
}
}

The android vold (volume deamon) service listen to this uevent notification and does the required things. The android patch is available at 

http://gitorious.org/rowboat/system-vold/commit/ea55061d003609eecafd26e62ea4e8786f3866df