USBProxy on Kernel 4.14.71-ti-r80

[MrSliphe]

Hello, I'm currently trying to get USBProxy running on the BeagleBone Black using the latest image release from beagleboard.org. The kernel version of this is 4.14.71-ti-r80. I'm using this https://github.com/RobertCNelson/ti-linux-kernel-dev/releases/tag/4.14.71-ti-r80 to integrate the patches, but since the source code has changed, I tried doing it by hand. As far as I understand, I would just need to apply the two changes in the inode.c.patch. Rebuilding using the rebuild.sh also works (at least it throws no errors).

Using USBProxy on the Beaglebone though makes me think, that the patches are not applied correctly, as it throws this error when using it:

Loading plugins from /usr/local/lib/USBProxy/
vendorId=ffffffff
productId=ffffffff
cleaning up /tmp
removing 1
Made directory /tmp/gadget-dzayif for gadget
Error mounting gadgetfs from [/tmp/gadget-dzayif].
Error code from mount is: [No such device]
Printing Config data
    Strings: 2
        DeviceProxy: DeviceProxy_LibUSB
        HostProxy: HostProxy_GadgetFS
    Vectors: 0
Pointer: 0
Device: 12 01 10 01 00 00 00 08 8a 25 01 00 00 02 01 02 00 01
  Manufacturer: SINO WEALTH
  Product:      USB KEYBOARD
    *Config(1): 09 02 3b 00 02 01 00 a0 32
        Interface(0):
            *Alt(0): 09 04 00 00 01 03 01 01 00
                HID: 09 21 10 01 00 01 22 41 00
                EP(81): 07 05 81 03 08 00 0a
        Interface(1):
            *Alt(0): 09 04 01 00 01 03 00 00 00
                HID: 09 21 10 01 00 01 22 9e 00
                EP(82): 07 05 82 03 08 00 0a
old bInterval: 0A
calculated new bInterval: 07
old bInterval: 0A
calculated new bInterval: 07
searching in [/tmp/gadget-dzayif]
/tmp/gadget-dzayif device file not found.
Error, unable to find gadget file.
Fail on open 11 Resource temporarily unavailable
GadgetFS not connected.
done

lsmod | grep "gadgetfs" shows the module is loaded, so the problem must be with how I applied the patch though I'm not sure what I did wrong. I just looked for the corresponding part in the inode.c (they are approx. 100 lines further up than the inode.c.patch tries to insert them) and inserted the spin_unlock_irq and spin_lock_irq lines. The second block appeared to be already applied in the newer inode.c so I only added the first block around line 559.

I tested using the same USB-device with the image provided in releases and there it worked. Is it possible to use USBProxy with this image or any image using a 4.X kernel or is it not compatible?

[mweal-ed]

I am guessing here as I had the similar error message with the pi zero ( issue #76 ). Is it possible they changed the filename for the gadgetfs usb in the sys tree?

Here is what I did:

• check for a file in the format *.usb in the /sys/devices/platform/soc/ (or is it /sys/devices/platform/ocp/ on the beagbone). (or print entry->d_name in the while loop of src/Plugins/Hosts/GadgetFS_helpers.c find_gadget_filename())
• check if this file name is in *devices[] in src/Plugins/Hosts/GadgetFS_helpers.c find_gadget_filename().
• If not add it and recompile/install.

hope this helps

[MrSliphe]

Sorry, I just had time to test your tips. Sadly it did not work. I checked in both /sys/devices/platform/soc/ and /sys/devices/platform/ocp/ and found the folder 47400000.usb in ocp. Adding it to devices[] in find_gadget_filename() just before the NULL did not help. The error is still the same after recompiling. In the 47400000.usb folder there are more folders with .usb maybe adding those would help? I will try later.

Regarding the print of entry->d_name: When I added a fprintf line to print entry-d_name I could see, that the while loop is looped through 3 times. The first time the print was empty, the second time it was .. and the third time . (so just a dot). Maybe this is an indicator for something else?

Thanks regardless.

[mweal-ed]

I'm working on a pi zero w, and don't have access to a bbb right now, so I can't try it, but I get the following in the while loop on the Pi (no empty line at the beginning). Sorry I can't be of any more help.

.
..
20980000.usb (your .usb will be different)

"sudo usb-mitm -l" ?

[MrSliphe]

The empty line was me not thinking correctly, I added the print

fprintf(stderr, "%s\n", entry->d_name);

before the

if (readdir_r(dir, entry, &result)) break;

Although I still do not see my .usb if I add it after that line. Using "usb-mitm -l" as root the output is now:

..
.
.

What did you do to apply the inode.c.patch? Just adjust the two parts manually and what did you change? I still hope it just has something to do with me patching wrong (and maybe adding the right part to devices[]).

[mweal-ed]

I have not loaded the patch. I originally thought it wasn't needed anymore. You got me looking into it further and now I'm questioning that decision.

[MrSliphe]

Okay I have now fixed it, the issue was not with the patch or USBProxy. Thus I did not need to add to the device list, as it was "musb-hdrc" for me, which is already included. The error came, because with the image I used, libcomposite was loaded at boot, which conflicts with the use of gadgetfs. By editing the use_libcomposite () in

opt/scripts/boot/am335x_evm.sh

and removing (or commenting) the following lines

echo "${log} modprobe libcomposite"
modprobe libcomposite || true
if [ -d /sys/module/libcomposite ] ; then
        run_libcomposite
else
        if [ -f /sbin/depmod ] ; then
                /sbin/depmod -a
        fi
        echo "${log} ERROR: [libcomposite didn't load]"
fi

the conflicting module wasn't loaded and I could freely use gadgetfs.

Thank you very much @mweal-ed for your help! Even if the solution was something else you gave me the idea that it may be something specific with the BeagleBone Black and gadgetfs.

[palesius]

For what it's worth, (And it has been a LONG LONG time, since I have touched the current source). If the inode patch was not working it would totally lock up the device, I think it gets stuck indefinitely in the gadgetfs driver. You would not just get a simple error message (unless things have changed since then).

On Wed, Feb 27, 2019 at 12:47 PM MrSliphe notifications@github.com wrote:

Closed #78 https://github.com/dominicgs/USBProxy/issues/78.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/dominicgs/USBProxy/issues/78#event-2168748866, or mute the thread https://github.com/notifications/unsubscribe-auth/AFjlHqRdufGmn6GwT_epbTtEd1jZKsTjks5vRsS1gaJpZM4anyCL .

[mweal-ed]

yes, @palesius that agrees with what I found on the history of the patch at https://www.spinics.net/lists/linux-usb/msg99032.html