> On Tue, 2014-06-17 at 09:31 +0200, Vlad Ungureanu wrote: > > Hello, > > > > I recently started to develop a USB kernel driver for a GoogleSummer Of Code > > project and I'm running in some small issues. The source code is available at > > https://github.com/ungureanuvladvictor/ADK-Driver/blob/master/adk-driver.c . > > The debugging printks in your open method (around line 104) are > obviously wrong. You test for NULL but then use the NULL pointer > for debugging. The same bug in adk_release(). > > adk_delete() has a memory leak. You don't free the buffer. > In adk_probe() please use the testing macros for endpoint > type and direction. > This is just a simple bad sketch, will fix all the issues for final tests. > > > The driver matches the VID/PID of a LG Nexus5 and tries in the probe function > > to switch it to ADK Mode *Accessory Development Kit Mode*. Afterwards the same > > driver should deal with the device in ADK Mode and should find it's bulk IN/OUT > > EPs. > > > > The problem is that my probe callback is called twice for each mode of the > > device, normal mode and ADK mode. > > It is called for each interface. Found out that already from the answers. :) > > > If you look at this Gist: > > https://gist.github.com/ungureanuvladvictor/44bc345037916bbd0bf1 lines 19 and > > 25 is where the probe function is called for the normal device. Lines 58 and 66 > > are for the device in ADK mode. As I read from the documentation the probe > > callback is called when a device that matches my device_table is found by the > > usb core. In the log I see that the usb core calls probe for each interface it > > has: lines 16, 22 and 55, 63. > > USB drivers are drivers for interfaces, not whole devices. Each > interface of a device is probed separately in turn. If you have > a device that exposes two mutually exclusive access methods in > the same configuration, you need to claim the second interface > for the undesired method in the probe method and refuse to bind > to the undesired interface in the probe method, as you cannot > reliably predict which interface is probed first. > Look at cdc-acm for an example of claiming other interfaces > from within probe() Thank you for pointing this hint out. > > HTH > Oliver > > -- Vlad Ungureanu
Attachment:
signature.asc
Description: Digital signature
