User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

smsc95xx errors

Fri Jul 20, 2012 11:24 pm

Hi,

With the Raspbian image on the main downloads page, I am getting quite a lot of

rasberrypi kernel: [.....] smsc95xx 1-1.1:1.0: eth0: Failed to write register index ...
rasberrypi kernel: [.....] smsc95xx 1-1.1:1.0: eth0: Failed to read register index ...

It looks like it might be a power issue, perhaps too much noise? Has anyone else seen this?

Thanks and best regards, Will

User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

Re: smsc95xx errors

Fri Jul 20, 2012 11:35 pm

Hi,

It is repeatable and from electronic noise. If the powered up (Trust 7 port compact hub) and usb charger are plugged into the same power socket, the LAN connection on the Pi does not work. It is probably the usb charger which is to blame. If the Pi is connected to the USB charger and not the powered hub at all, then all is well and good. If only there was a GND pin on one of these items.

Best regards, Will

User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

Re: smsc95xx errors

Sat Jul 21, 2012 2:11 pm

Hi,

The same problem is observed with another Raspberry Pi, different brand of mini-hub and power supply: it appears to be the grounding. The kernel slows down and the smsc95xx errors can be seen by leaving a terminal running with

sudo tail -f /var/log/messages

This does not happen when the PI is connected directly to the keyboard and mouse without the mini-hub. In fact, the power from one hub leaks back into the Pi to such an extent that it causes it to boot, without the micro-USB power adapter being connected. It looks like a ground/signal isolation circuit is needed on the USB line. Plugging in some speakers into the audio, noise can be heard from the HDMI signal. For example, the audio noise matches the text printed onto the console.

Best regards, Will

blackarchon
Posts: 20
Joined: Sat Jun 23, 2012 5:39 pm

Re: smsc95xx errors

Sat Jul 21, 2012 6:26 pm

I had similar problems and opened an issue: https://github.com/raspberrypi/linux/issues/60

User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

Re: smsc95xx errors

Sat Jul 21, 2012 9:50 pm

blackarchon wrote:I had similar problems and opened an issue: https://github.com/raspberrypi/linux/issues/60
Thanks blackarchon. I have added more details to your issue report. Regards, Will

blackarchon
Posts: 20
Joined: Sat Jun 23, 2012 5:39 pm

Re: smsc95xx errors

Sun Jul 22, 2012 11:49 am

So do you think it may be a problem with the hardware rather than the software?

marosg
Posts: 25
Joined: Sat Jul 07, 2012 2:41 pm

Re: smsc95xx errors

Sun Jul 22, 2012 11:57 am

Exactly this happened to me yesterday after I moved my whole raspi/hub/HDD/cables setup from one wooden box to another. I was receiving smsc95xx errors and system was very slow. I am not using eth0 and LAN port is not connected. So first I checked if something is not touching etc. Also some cables (especially USB cable to wifi stuff) went just over the Pi board. So I disconnected and connected all again and made sure no cables go over the board. After the boot, everything was ok. So it was really some hardware interference type of stuff. I rebooted couple of times since then and all is fine.
HTH

marosg
Posts: 25
Joined: Sat Jul 07, 2012 2:41 pm

Re: smsc95xx errors

Sun Jul 22, 2012 12:18 pm

And to add - I have powered USB hub and I power my Pi from the HUB. I use one USB on Pi which goes to the HUB and I have HDD, keyboard and wifi connected to the HUB. The second USB port on Pi is not connected as well as Rj45 port.

User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

Re: smsc95xx errors

Sun Jul 22, 2012 6:13 pm

blackarchon wrote:So do you think it may be a problem with the hardware rather than the software?
I have not looked at the schematics or layout of the electronics and therefore cannot be sure what is going on. It is suspicious that the noise level on the USB connection is high at the same time as the errors, but this might not be correlated with the errors in /var/log/messages. The HDMI chip firmware had to be tuned to get the PI through the CE tests. Therefore, I am not surprised that I can hear HDMI imaging noise. There are several others on the forum who have a reasonable or good knowledge of the layout and may be able to give a better diagnosis.

User avatar
redhawk
Posts: 3465
Joined: Sun Mar 04, 2012 2:13 pm
Location: ::1

Re: smsc95xx errors

Sun Jul 22, 2012 6:26 pm

If you have a voltmeter can you measure TP1 - TP2 if it drops below 4.7 volts the PI will boot but tends to crap itself communicating with hardware.

Richard S.

User avatar
mahjongg
Forum Moderator
Forum Moderator
Posts: 11301
Joined: Sun Mar 11, 2012 12:19 am
Location: South Holland, The Netherlands

Re: smsc95xx errors

Sun Jul 22, 2012 7:39 pm

williamhbell wrote:
blackarchon wrote:So do you think it may be a problem with the hardware rather than the software?
I have not looked at the schematics or layout of the electronics and therefore cannot be sure what is going on. It is suspicious that the noise level on the USB connection is high at the same time as the errors, but this might not be correlated with the errors in /var/log/messages. The HDMI chip firmware had to be tuned to get the PI through the CE tests. Therefore, I am not surprised that I can hear HDMI imaging noise. There are several others on the forum who have a reasonable or good knowledge of the layout and may be able to give a better diagnosis.
What do you mean by "I can hear HDMI imaging noises"?
Do you pick it up on a radio, or what?
The firmware tweak needed for EMC certification probably were only to do with a lower default drive level. These signals are notorious for EMC in that they are in the range that is tested for and are almost "monotone", and thus score high peaks on a spectrum analyser, especially if used with badly shielded cables.

If you want to experiment with "noise" on the USB (I'm guessing you must mean noise on the datalines), and power feedback over the uplink cable to your hub, then I suggest using a tiny piece of cellotape to cover pin 1 of the USB plug (see wikipedia for a description which pin is pin 1).
Another test is to measure with a multimeter if there is a voltage between the ground of the PI, and earth there is an AC voltage. Ground of the raspberry PI, is TP2. For earth, I suggest you use either the ground pin of your mains, or perhaps a waterpipe (not the Turkish variant mind you) but something electrically tied to earth.
If there is a lot of voltage there, something like half the mains voltage, its probable that there is a PSU used in the system that has two small capacitors tied between the mains lead, and the "ground pin" of the PSU, if that ground isn't really connected to earth it means that half the mains voltage is on GND of the system, which may well lead to strange problems.
be careful with tying TP2 to earth in this case, as you do not know how "hard" the voltage might be, I suggest you start with a 10K resistor in series just in case, then see if the voltage disappears.
Multimeters (good ones) typically have an impedance in the AC and DC voltage settings of about 11.11 Meg Ohm, (for 1:10 probes consisting a 100M resistor) so they wont act as shorts to ground during measurements like this.

User avatar
mahjongg
Forum Moderator
Forum Moderator
Posts: 11301
Joined: Sun Mar 11, 2012 12:19 am
Location: South Holland, The Netherlands

Re: smsc95xx errors

Sun Jul 22, 2012 7:53 pm

P.S. my favorite theory why there could be many repeating error messages, especially if there are a hundred a second or so, is maybe ripple on the 5V supply, causing the working voltage of whatever causes the problem dropping below its minimum working voltage a hundred times a second (or maybe 50 depending on the construction of the PSU), each time causing a fault condition.

If you simply measure with the DC setting of your multimeter the voltage between TP1 and TP2 you might find an acceptable level, say 5.1 volt, but you should actually also measure with AC, and if you find anything above a few hundred millivolt there it is bad news!
If you have one, I would suggest using an oscilloscope to try to find if there is any ripple dipping below say 4.9 volt.

I can imagine that developers of phone chargers do not care much about a bit of ripple on the output, and if you use smaller (high voltage) elco's on the input of your switching supply, that might be a big saving post.

In fact if you are really serious about researching Raspberry PI power issues, I would use a laboratory 5V supply, so you are certain there is no ripple.

A cheaper temporary solution might be to take a big (10.000uF 16V) axial elco, and solder it between TP1 and TP2 (obviously you should regard the right polarity) and see if that improves things.

User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

Re: smsc95xx errors

Sun Jul 22, 2012 8:43 pm

Hi mahjongg, Thanks for the long and detailed post. I will try to try out your suggestions to reduce the noise.
mahjongg wrote: What do you mean by "I can hear HDMI imaging noises"?
Do you pick it up on a radio, or what?
No, on USB powered speakers. When the speakers are powered by the same hub which is connected to the PI and connected to the audio out of the PI, one can hear noise which matches the text printed on the console. So when the PI boots one can turn the screen off and listen to the text being printed via the hdmi out. If the PI is disconnected from the usb hub, then the audio is almost completely noise free. The noise signal is also different if the hdmi is connected or not, since the only GND line to the PI is through the monitor to hdmi shield.

Regards, Will

User avatar
williamhbell
Posts: 291
Joined: Mon Dec 26, 2011 5:13 pm
Contact: Website Twitter

Re: smsc95xx errors

Sun Jul 22, 2012 8:46 pm

mahjongg wrote:P.S. my favorite theory why there could be many repeating error messages, especially if there are a hundred a second or so, is maybe ripple on the 5V supply, causing the working voltage of whatever causes the problem dropping below its minimum working voltage a hundred times a second (or maybe 50 depending on the construction of the PSU), each time causing a fault condition.
The rate of error messages is at the level of one or two per second, but the kernel is visibly struggling.

medols
Posts: 1
Joined: Mon Aug 06, 2012 6:15 pm

Re: smsc95xx errors

Mon Aug 06, 2012 6:40 pm

We had similar problems with a device attached to a powered usb hub.

The "lsusb -t" listed the device through the hub.

The usb driver loaded by default was cdc_acm.

Nevertheless, "smsc95xx" kernel errors appeared when we accessed the device.

As a result, the device could not be used at all.

We know this is not a solution but it may help to progress on this issue:

- Start Raspbian in text mode.

- Start a remote "ssh" session through the ethernet on another computer.

- Disconnect all USB devices (including mouse and keyboard).

- Disconnect the HDMI display cable.

- Connect the powered usb-hub and the usb device.

- Now the "powered-through-usb-hub" device works properly at the remote session.

Return to “Raspbian”