On Thu, 5 Feb 2009, Erin Teran wrote: > <postme1@...> writes: > > > I'm working on an embedded system with a PHYTEC IMX31 board and run a patched > kernel based on 2.6.22.6. > > > > I use the OTG USB interface to connect as client. > > This works fine using usb.c and the gadgetfs which is loaded as > > module above the arcotg_udc driver > > > > During operation I need to disconnect from the host and connect > > again afterwards. > > I try to do the disconnect by closing all file descriptors associated > > with the connection (ep_in, ep_out > > and ep0) and the close is always successful. > > In most cases this works fine. > > Sometimes, however, I can not reopen the ep0 file descriptor > > /dev/gadgetfs/arc_udc again. > > Instead of opening the device I get a -EBUSY. > > > > So I added some printk output to the inode.c and arcotg_udc.c and > > recompiled the modules to see what happens. > > It seems that in the error case the release function is not called > > even through the file descriptors for ep0, > > ep_in and ep_out where closed. > > > Hi, > > I also experience the same Problem, when I try to disconnect the demo > application usb.c from gadgetfs. > > Normally I would assume that the usb device driver is released when I close all > three endpoints. But the printk() show that in some cases (possible gadgetfs is > busy?) the closing of the file descriptors does not lead to a dev_release() call. > > The error looks identical to the error, postme1 posted in the message above. > > Maybe the most experienced gadgetfs developers (David Brownell, Alan Stern, ...) > are able to help us? > > Please, I'm completely stuck on this one... Closing the device file should always cause dev_release() to be called. If it doesn't then there must still be an open reference to the file. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
