Kernel BUG (Oops) issue with USB LAN dongles (DM9601 driver)


12 posts
by FTrevorGowen » Fri May 31, 2013 3:29 pm
Background: Some years ago I acquired a couple of USB LAN dongles to give my netbook an additional ethernet port to enable direct connection to a NSLU2 box onto which I was about to install Debian (Lenny IIRC) - during its "redboot" sequence it operated on a different sub-net to that of the household. Fast-forward to last autumn when I acquired my first Pi (B1 256Mb) - tried them out briefly - seemed to be working - could ping or login via ssh etc. Didn't get around to trying them with a "B2 512Mb" Pi but next tried with a Model A - this is when the issues first "came to light" and were noted in this thread: viewtopic.php?f=36&t=35959&p=302445#p302445
I've now tried six devices in total and only one of them (the oldest) does not generate/trigger the Kernel BUG.
Some details (pictures etc. at http://www.cpmspectrepi.webspace.virgin ... pters.html):
Working device - Pluscom U1EC, lsusb reports ID 0a46:9000 Davicom Semiconductor, Inc. DM9000E Fast Ethernet Adapter
Problem devices - variants of JP108 : 030818 (all physically different in board layout & some chips), lsusb reports ID 0fe6:9700 Kontron (Industrial Computer Source / ICS Advent) DM9601 Fast Ethernet Adapter in all but one case where its ID 0fe6:8101.
Since the "Oops" locks up the Pi (model A or B2) I can only attach low-res "physical" screenshots, but I hope there's some useful info. visible. The other thing the problem devices have in common is that their "LAN LED" flickers at 1 - 2 times a second - the working device may "occasionally" turn off and back on briefly.

Now for the questions:
1) Has anyone else come across this or a similar problem?
2) I think there are other devices that use a different chipset/driver - any suggestions?
3) Has anyone purchased a recent version of the Pluscom U1EC device and found it to work?
4) Given that the Pi gets "locked" by the bug are there any logs or similar I could check on w/o triggering the bug? (ie. by not generating "heavy" network traffic through the device) and report back with details?
5) Would a member of the group working on USB issues be interested in the "loan" of a device or two to use for testing I don't have the knowledge and/or equipment to do myself? (They only cost a few pounds anyway).

DSCN0944css.JPG
Model A "Oops"
DSCN0944css.JPG (52.43 KiB) Viewed 3130 times

DSCN0945css.JPG
Model B2 pre-oops
DSCN0945css.JPG (56.07 KiB) Viewed 3130 times

DSCN0946css.JPG
Model B2 post-oops
DSCN0946css.JPG (53.2 KiB) Viewed 3130 times


TIA, Trev.
Mainly running "Wheezy Raspbian" on five Pi's (an A, B1, 2xB2 & a B+). Notes about them begin at http://www.cpmspectrepi.webspace.virginmedia.com/raspberry_pi/raspiidx.htm
User avatar
Forum Moderator
Forum Moderator
Posts: 1720
Joined: Mon Mar 04, 2013 6:12 pm
Location: Bristol, U.K.
by FTrevorGowen » Sat Jun 01, 2013 9:30 pm
Update/"Bump": I've added a bit more info. to my webpage (link in original post). Also I didn't mention that the (most recent) "failing device" works fine with my Ubuntu/XP netbook . From other threads it seems likely that a new release of "Wheezy" is due soon - with luck maybe the "needed fix" will have already happened. (Better not "hold my breath" though :) ).
Trev.
Mainly running "Wheezy Raspbian" on five Pi's (an A, B1, 2xB2 & a B+). Notes about them begin at http://www.cpmspectrepi.webspace.virginmedia.com/raspberry_pi/raspiidx.htm
User avatar
Forum Moderator
Forum Moderator
Posts: 1720
Joined: Mon Mar 04, 2013 6:12 pm
Location: Bristol, U.K.
by FTrevorGowen » Mon Jun 03, 2013 3:39 pm
Update: Some more info. (that may hold a clue?)
Fairly confident that I've ruled out power issues - usb-devices output suggested that the problem device(s) should require less current (120mA cf. 144mA). Checks on the Pi's supply voltage confirm this: 4.9V cf. 4.8V (problem cf. working) - measured during "ping testing" which doesn't lead to lock-up.
However the "ping behaviour" is different:
Good, "Pluscom" USB LAN dongle - rtt min/avg/max/mdev 0.736/1.202/2.011/0.329 ms (40 pings)
Problem, "JP1082" USB LAN dongle - rtt min/avg/max/mdev 0.845/1.745/12.835/1.808 ms (40 pings)
I suspect that the latter's max of 12.835 ms is probably also associated with the observed "LED flicker" for the "problem devices".
Trev.
Mainly running "Wheezy Raspbian" on five Pi's (an A, B1, 2xB2 & a B+). Notes about them begin at http://www.cpmspectrepi.webspace.virginmedia.com/raspberry_pi/raspiidx.htm
User avatar
Forum Moderator
Forum Moderator
Posts: 1720
Joined: Mon Mar 04, 2013 6:12 pm
Location: Bristol, U.K.
by rpdom » Mon Jun 03, 2013 3:55 pm
Yes, the DM9601 devices do cause an oops when under any significant loading. I have a couple of them for my model A's. It is annoying :(

It may be something to do with the fact that the interfaces I've got are only USB 1.1 (they say they are compatible with USB 2.0 and they are because 2.0 can handle 1.1 devices quite happily, just slowly). I'm giving up on them and getting some proper USB2.0 interfaces.
Posts: 3092
Joined: Sun May 06, 2012 5:17 am
Location: Essex, UK
by FTrevorGowen » Mon Jun 03, 2013 4:21 pm
rpdom wrote:Yes, the DM9601 devices do cause an oops when under any significant loading. I have a couple of them for my model A's. It is annoying :(

It may be something to do with the fact that the interfaces I've got are only USB 1.1 (they say they are compatible with USB 2.0 and they are because 2.0 can handle 1.1 devices quite happily, just slowly). I'm giving up on them and getting some proper USB2.0 interfaces.


Thank you - at least I now know it's not "just me" that's had the problem. Would you mind checking the 'photos at http://www.cpmspectrepi.webspace.virgin ... pters.html and reporting back if your devices are "of the same type". (If different it might be useful to have the details too.)
TIA, Trev.
Mainly running "Wheezy Raspbian" on five Pi's (an A, B1, 2xB2 & a B+). Notes about them begin at http://www.cpmspectrepi.webspace.virginmedia.com/raspberry_pi/raspiidx.htm
User avatar
Forum Moderator
Forum Moderator
Posts: 1720
Joined: Mon Mar 04, 2013 6:12 pm
Location: Bristol, U.K.
by rpdom » Mon Jun 03, 2013 4:46 pm
The board inside mine appears to be identical to the one labelled KY-RS9600 in your pictures, although it is labelled as SR9600 NO: 030818 MADE IN CHINA. The USB ident is "Kontron (Industrial Computer Source / ICS Advent) DM9601 Fast Ethernet Adapter". I have one in black plastic and one in blue.

I'm disappointed to see the "PowerSynch Tech. LAN adapter & Hub" on your list, as I've just ordered something that looks very like that in the hope that it would have a different chipset and support USB2.0 properly :(

I do have a different interface somewhere, which does use a different chipset and does support USB2.0, but I can't find it after a recent bout of "tidying up".
Posts: 3092
Joined: Sun May 06, 2012 5:17 am
Location: Essex, UK
by FTrevorGowen » Mon Jun 03, 2013 6:10 pm
rpdom wrote:The board inside mine appears to be identical to the one labelled KY-RS9600 in your pictures, although it is labelled as SR9600 NO: 030818 MADE IN CHINA. The USB ident is "Kontron (Industrial Computer Source / ICS Advent) DM9601 Fast Ethernet Adapter". I have one in black plastic and one in blue.

I'm disappointed to see the "PowerSynch Tech. LAN adapter & Hub" on your list, as I've just ordered something that looks very like that in the hope that it would have a different chipset and support USB2.0 properly :(

I do have a different interface somewhere, which does use a different chipset and does support USB2.0, but I can't find it after a recent bout of "tidying up".

It looks as if the "key id" (to avoid, perhaps) is "03818" which appears on yours and three of mine. Ironically it my "oldest" device that works for me (the Pluscom U1EC) which may still be USB1.1 (purchased early 2010?). I have found the "Pluscom website" and a similarily labelled device is available (IIRC) but at 2-3 times the cost (w/o delivery) cf. that of the others tried. I've also come across a device similar to the "PowerSynch Tech. LAN adapter & Hub" which appears to use a different (Apple compatible/device "clone") chipset, but, again at ~2x the cost. One of my aims was to find a device, that could be used with Model A's, with lower power requirements and cheaper than a nano WiFi adapter (see also an older thread of mine here: viewtopic.php?t=35959).
Thanks again, Trev.
Mainly running "Wheezy Raspbian" on five Pi's (an A, B1, 2xB2 & a B+). Notes about them begin at http://www.cpmspectrepi.webspace.virginmedia.com/raspberry_pi/raspiidx.htm
User avatar
Forum Moderator
Forum Moderator
Posts: 1720
Joined: Mon Mar 04, 2013 6:12 pm
Location: Bristol, U.K.
by FTrevorGowen » Thu Jun 06, 2013 3:38 pm
Update: I've now received and tried out a new dongle, a Pluscom U2E-ADM8515 which is based upon a different chipset and uses the "pegasus" driver. Like the old Pluscom DM9000E-based device this, so far, appears to be working fine :D and is a USB 2.0 device. It's downside is a higher current (224mA) leading to a lower voltage of 4.74V (cf. 4.8 for the older device). It's also about twice the cost** of the other devices tested.
Ping stats. are much better: rtt min/avg/max/mdev 0.384/0.470/0.891/0.080 ms (40 pings) ~2x faster and more stable!
All I need to find now is a similar device combined with a hub (cf. the non-working PowerSynch Tech. device) with a similar cost-break**** - one possibility is that I'm asking about in this thread: viewtopic.php?f=46&t=46084 .
(Of course, it would be nice if the issues with the {USB 1.1?} devices could be fixed)
Trev.
** ~ 6 pounds.
**** < 14 pounds.
Mainly running "Wheezy Raspbian" on five Pi's (an A, B1, 2xB2 & a B+). Notes about them begin at http://www.cpmspectrepi.webspace.virginmedia.com/raspberry_pi/raspiidx.htm
User avatar
Forum Moderator
Forum Moderator
Posts: 1720
Joined: Mon Mar 04, 2013 6:12 pm
Location: Bristol, U.K.
by iansexton » Mon Sep 02, 2013 4:35 pm
I'm joining this thread very late, however...
I bought a model A some time ago. To set it up I took the SD card from a model B, I plugged in a PASSIVE usb hub into my model A to which I connected a keyboard, mouse & wifi adapter.
This all worked fine and using this setup I configured the model A as I wanted it. My intention was to ditch the hub and connect my USB/wifi dongle as the only USB device.
The WiFi dongle was a cheap unbranded device bought on ebay. Note that it worked perfectly on two model Bs as well as on my A through a hub.I think you can guess what's coming - the wifi dongle doesn't work when connected directly.
I had the same problem with a similar wifi device with a different chipset.
I tried a very old (and big) D-Link device & it worked fine through the hub & plugged directly to the A's USB.
Unfortunately I didn't document this as I didn't have time to pursue the problem/solution.

Curiously, I've seen exactly the same problem using openwrt on a TP-Link 703 router where the usual explanation is that the USB host does not support USB1.1 Is this the case with rpi A too?
Posts: 4
Joined: Mon Jan 14, 2013 5:40 pm
by jdb » Mon Sep 02, 2013 5:58 pm
There are very few differences when using USB2.0 devices on a hub (which a model B has) versus on the model A port directly. There was a specific bug that was fixed ages ago relating to certain types of devices plugged directly into the model A port that would become unresponsive after a while.

If the issue with the wifi dongle still exists after updating to latest kernel (from rpi-update) then start another thread about it.

USB1.1 devices connected directly to the model A port usually work, but certain ones (low-speed hubs especially) don't seem to work.
Keeper of the Proboscidean features
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 656
Joined: Thu Jul 11, 2013 2:37 pm
by iansexton » Tue Sep 03, 2013 5:12 pm
Could you elaborate?
There was a specific bug that was fixed ages ago relating to certain types of devices plugged directly into the model A port that would become unresponsive after a while.


What bug? what devices? When was it fixed? Can you point me to something explaining this?

Cheers
Ian
Posts: 4
Joined: Mon Jan 14, 2013 5:40 pm
by jdb » Tue Sep 03, 2013 5:32 pm
iansexton wrote:Could you elaborate?


There was a specific bug that was fixed ages ago relating to certain types of devices plugged directly into the model A port that would become unresponsive after a while.


What bug? what devices? When was it fixed? Can you point me to something explaining this?

Cheers
Ian


https://github.com/raspberrypi/linux/co ... 4d22b74723

This was exhibited if there were no USB devices with periodic endpoints attached to the Pi. In the case of a model B, there is always a periodic endpoint (the hub event endpoint) therefore this bug was only seen with single devices (again with no periodic endpoints) attached to the model A.

The presence of a periodic endpoint on a device (usually an Interrupt endpoint for reporting status) is implementation-specific. Manufacturers may choose to omit this, but they will then always keep a request queued on a non-periodic endpoint that blocks until data or status changes are received.
Keeper of the Proboscidean features
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 656
Joined: Thu Jul 11, 2013 2:37 pm