Sorry but yesterday was a busy day for me (I recovered a copule of terabyte lost due to a server crash
 ) I'm back at home now 4:20 AM.Yesterday morning I built a kernel with the USB debug options enabled, I haven't it tried yet, but tomorrow, I'll see if we can get anything useful from it. I'm still hoping on a further reply from Carl Nordbeck, with a bit of more detailed infos on the device. And now some sleep... I suspect we are doing things right with the switching, but the device will reset itself after five seconds.
Any news about that 18 sec. delay that udevmonitor shows? Some more ideas to try: - the "power suspend" trick (see Troubleshooting) - not calling manually but entering a udev rule; very short time from plugging to switching (if you don't do this already)
Sorry for the late reply, but I had no time due to my work.
I tested the kernel with the debug options enabled: the dmesg logs in this way are becoming simply huge, and impossible to understand due the lack of timing infos, for the moment I abandoned the idea. The most important thing that I discovered is this tread http://www.gossamer-threads.com/lists/l ... el/1013988 The kernel where I did most of my terst is just a 2.6.28, and the fedora 10 original kernel is a 2.6.27.9 probably very close to the 2.6.28 one. Carl Nordbeck is back from his vacations and has replied to the my mail, and still says that the correct way is to switch is the udev rule method. He also told that the udev rule is working on his EC400 which is an expresscard on an ubuntu distro. The HW Ec400 must be very close to the md400 one. So, I'm going to install a different distro in the 32 bit flavor to see if something changes and what changes, and then I'll report here what happened.
Ok, I did some tests with the my fedora x64 with a couple of older kernels, I tested also the modem on a kubuntu 8.10 x86 and on my asus wl500gp router (which runs on a mipsel 32 architecture) and nothing changes, same results with usb_modeswitch, same results with the custom udev rule, same results ecoing directly the configuration value.
At the same time Carl at Ericsson has tried his EC400 express card on different fedora's kernels, an the udev rule works fine. So the only big difference is the fw revision (my MD400 has a newer release). Now I'm short of ideas. Maybe is worth to wait that a second user on the planet buy this device, and test it on his linux machine.
Good news from Ericsson. Carl has contacted His colleagues at SE, and they confirmed that there are some differences between EC400 and MD400, they also will ship a MD400 to sweden, so Carl can check what's going on. I'll keep You informed. Unfortunately, I can only recommend to try everything possible with the Sony MD400. I don't have it here to play with (someone send me one?) so you are sort of on your own ...
At least the behaviour of the device seems to be consistent as opposed to erratic. The setting of the configuration (and why it fails) might be one track to follow.
After 1.5 days of brute force trial and error I'm happy to say that the MD 400 seems to stay in modem mode now.
On the bad side, we have to wait for the device to return after the first switch before issuing the next command; this may take more than ten seconds. I'll update the proc, maybe Sunday evening. Here is a test version 0.9.7gamma: source file
Use it with: "-v 0x0fce -p 0xd0e1 -O 1". I found one entry in the Windows logs where after switching an AltSetting was done on one interface. If you want to reproduce that, use these parameters: "-v 0x0fce -p 0xd0e1 -O 1 -i 8 -a 2" Details: The behaviour of the MD400 on Windows is still fundamentally different. After the switching command the device returns immediately with the new configuration. On Linux, it's usually gone for 19 seconds before returning. On the other hand, doing "ResetUSB" on the switched device makes it switch back immediately, with the USB kernel message "device firmware changed" (from hub.c). I'll investigate further if I can provoke this state during the first switch. But I suspect this is a kernel/system thing and not possible to fix without a patch. To make the device stay in the switched mode the switch command has to be issued again to the switched device (after it shows up on the system again). You can do this even with the previous version of USB_ModeSwitch and some kind of timer. In the test version, there is a wait and check loop which addresses the switched device once it returns. Your choice. But please do give some feedback ... |
