Ethernet dies after about 12 hours, nothing connected to USB


13 posts
by ausserirdischegesund » Sat Nov 03, 2012 11:45 am
I've just got 2 Raspis, and everything works fine for a while. I had no luck letting them run for longer times (like overnight) because Ethernet dies after a few hours (I had it running for close to 24 once, sometimes only 6 or so. 12 hours seems to be the average, so quite a bit longer than problems that show after a few minutes, but short enough to make it useless as a headless server.

I've tried both having nothing connected, as well as various peripherials on a powered hub. It does not seem to make much difference.

I've tried 2 different power supplies, and even though I've got another, "good" one on order, I doubt this is the root cause of the problem, because even using a 2.5 A specified powered USB hub as a power source does not help. I have not been able to check voltages yet, this is the next thing I will try.

The log gives me nice messages, as the Raspi continues to run, only with dead network (and USB stack?). This always looks similar, just one example:

Code: Select all
Nov  2 21:07:16 pi kernel: [44813.532421] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:07:21 pi kernel: [44818.532491] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000114
Nov  2 21:07:26 pi kernel: [44823.532560] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:07:31 pi kernel: [44828.532625] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000118
Nov  2 21:07:36 pi kernel: [44833.532694] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:07:41 pi kernel: [44838.532758] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000114
Nov  2 21:07:46 pi kernel: [44843.532818] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:07:51 pi kernel: [44848.532890] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000118
Nov  2 21:07:57 pi kernel: [44854.632970] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:08:02 pi kernel: [44859.633043] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000114
Nov  2 21:08:02 pi kernel: [44859.922962] ------------[ cut here ]------------
Nov  2 21:08:02 pi kernel: [44859.923024] WARNING: at net/sched/sch_generic.c:255 dev_watchdog+0x2a0/0x2c0()
Nov  2 21:08:02 pi kernel: [44859.923042] NETDEV WATCHDOG: eth0 (smsc95xx): transmit queue 0 timed out
Nov  2 21:08:02 pi kernel: [44859.923054] Modules linked in: evdev uinput btusb rfcomm bnep bluetooth snd_bcm2835 snd_pcm snd_seq snd_timer snd_seq_device snd snd_page_alloc
Nov  2 21:08:02 pi kernel: [44859.923159] [<c0013e1c>] (unwind_backtrace+0x0/0xf0) from [<c0028b24>] (warn_slowpath_common+0x4c/0x64)
Nov  2 21:08:02 pi kernel: [44859.923194] [<c0028b24>] (warn_slowpath_common+0x4c/0x64) from [<c0028bd0>] (warn_slowpath_fmt+0x30/0x40)
Nov  2 21:08:02 pi kernel: [44859.923226] [<c0028bd0>] (warn_slowpath_fmt+0x30/0x40) from [<c02eb38c>] (dev_watchdog+0x2a0/0x2c0)
Nov  2 21:08:02 pi kernel: [44859.923274] [<c02eb38c>] (dev_watchdog+0x2a0/0x2c0) from [<c0033fe8>] (run_timer_softirq+0x114/0x2a4)
Nov  2 21:08:02 pi kernel: [44859.923307] [<c0033fe8>] (run_timer_softirq+0x114/0x2a4) from [<c002e62c>] (__do_softirq+0x90/0x128)
Nov  2 21:08:02 pi kernel: [44859.923337] [<c002e62c>] (__do_softirq+0x90/0x128) from [<c002eaec>] (irq_exit+0x90/0x98)
Nov  2 21:08:02 pi kernel: [44859.923375] [<c002eaec>] (irq_exit+0x90/0x98) from [<c000e880>] (handle_IRQ+0x34/0x84)
Nov  2 21:08:02 pi kernel: [44859.923411] [<c000e880>] (handle_IRQ+0x34/0x84) from [<c037c974>] (__irq_svc+0x34/0xc8)
Nov  2 21:08:02 pi kernel: [44859.923443] [<c037c974>] (__irq_svc+0x34/0xc8) from [<c000e990>] (default_idle+0x24/0x28)
Nov  2 21:08:02 pi kernel: [44859.923473] [<c000e990>] (default_idle+0x24/0x28) from [<c000eb78>] (cpu_idle+0x9c/0xc4)
Nov  2 21:08:02 pi kernel: [44859.923513] [<c000eb78>] (cpu_idle+0x9c/0xc4) from [<c04c1718>] (start_kernel+0x280/0x2c8)
Nov  2 21:08:02 pi kernel: [44859.923532] ---[ end trace 14762c6030d9df86 ]---
Nov  2 21:08:07 pi kernel: [44864.633104] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:08:12 pi kernel: [44869.633172] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000118
Nov  2 21:08:17 pi kernel: [44874.633245] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:08:22 pi kernel: [44879.633312] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000114
Nov  2 21:08:27 pi kernel: [44884.633365] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Nov  2 21:08:32 pi kernel: [44889.633431] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000118
Nov  2 21:08:37 pi kernel: [44895.173506] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000014
Nov  2 21:08:37 pi kernel: [44895.173539] smsc95xx 1-1.1:1.0: eth0: Failed to write HW_CFG_LRST_ bit in HW_CFG register, ret = -1


This goes on for hours.

Any ideas except for trying another power supply (which I doubt is the problem, as it should give me enough juice, and I've got nothing plugged in into the Raspi's USB)?

I am using a current Raspbian, nothing overclocked, stock kernel, no X, just a few idle gettys on the console. And each of my two Raspis seems to act that way (so it is not only one defective one). Did I get two broken ones? Is this the general "Elephant in the Room" USB flakiness? Something else? Is there hope to get my Raspi running continuously as a headless server?

TIA
Ralph
Posts: 18
Joined: Wed Mar 07, 2012 7:52 am
by jamesh » Sat Nov 03, 2012 12:03 pm
Something like this has already been mentioned a few times in the forum - try a search form the front page of the website on some of the error messages.
Soon to be unemployed software engineer currently specialising in camera drivers and frameworks, but can put mind to most embedded tasks. Got a job in N.Cambridge or surroundings? I'm interested!
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 11509
Joined: Sat Jul 30, 2011 7:41 pm
by ausserirdischegesund » Sat Nov 03, 2012 12:16 pm
Thanks, jamesh! I've read through at least 10 or 20 of these threads, but found nothing to help me, as most solutions that were discribed as working were "I unplugged USB device X, and it worked", which does not help me as I've got nothing plugged in anyway. Other threads were quite inconclusive. I also tried the kernel commandline to slow down USB, but that did not work either (in fact my Pi would not even boot all the way with that kernel commandline).

Of course it still might be a power problem ... can't test that before next week, though.
Posts: 18
Joined: Wed Mar 07, 2012 7:52 am
by pluggy » Sat Nov 03, 2012 3:03 pm
I run all 3 of my Pi's headless on a long term basis. I have problems with the ones with USB wifi adapters sometimes (usually when my router glitches and loses the internet connection), but the wired one is rock solid. Uptime is presently reporting close on 5 days.

PS, I don't own a powered USB hub.
Don't judge Linux by the Pi.......
User avatar
Posts: 2275
Joined: Thu May 31, 2012 3:52 pm
Location: Barnoldswick, Lancashire,UK
by gsh » Sat Nov 03, 2012 9:31 pm
--
Gordon Hollingworth PhD
Raspberry Pi - Director of Software Engineering
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 759
Joined: Sat Sep 10, 2011 11:43 am
by ausserirdischegesund » Sun Nov 04, 2012 1:56 pm
Thanks pluggy&gsh, that sounds good (both that it works for some people and that there is a formal bugreport being worked on). I will certainly investigate this further as described in the bugreport, when it happens next time.

Strangely enough since the posting my pi Ethernet has not crashed. It might be because I've used yet another power supply (USB outlet of my other computer), or it is something else.
Posts: 18
Joined: Wed Mar 07, 2012 7:52 am
by SR-71 » Fri Nov 23, 2012 8:00 pm
Same problem here: I have an unattended system with one only device attached to the USB and of course the ethernet.
Nokia power supply as provided.
The system freezes every 6-24 hours.

uname -a
Linux raspberrypi 3.2.27+ #160 PREEMPT Mon Sep 17 23:18:42 BST 2012 armv6l GNU/Linux
Posts: 2
Joined: Fri Nov 23, 2012 7:56 pm
by SR-71 » Fri Nov 23, 2012 9:40 pm
I solved the problem using suggestions found in

https://github.com/raspberrypi/linux/issues/151

It seems that a rpi-update cured definitively that problem.
Posts: 2
Joined: Fri Nov 23, 2012 7:56 pm
by ewig » Fri Jan 11, 2013 3:22 pm
SR-71 wrote:https://github.com/raspberrypi/linux/issues/151

It seems that a rpi-update cured definitively that problem.


How can I tell if I have this fix on my system? I'm seeing network fails after about 12 hours uptime with identical messages.

Code: Select all
$ uname -a
Linux raspberrypi 3.2.27+ #250 PREEMPT Thu Oct 18 19:03:02 BST 2012 armv6l GNU/Linux

and
Code: Select all
Jan 11 14:58:10 raspberrypi kernel: [ 2088.329951] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000114
Jan 11 14:58:15 raspberrypi kernel: [ 2093.330015] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000114
Jan 11 14:58:20 raspberrypi kernel: [ 2098.330087] smsc95xx 1-1.1:1.0: eth0: Failed to read register index 0x00000118
Jan 11 14:58:25 raspberrypi kernel: [ 2103.900156] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000014
Jan 11 14:58:25 raspberrypi kernel: [ 2103.900189] smsc95xx 1-1.1:1.0: eth0: Failed to write HW_CFG_LRST_ bit in HW_CFG register, ret = -110
Jan 11 14:58:31 raspberrypi kernel: [ 2109.900248] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000014
Jan 11 14:58:31 raspberrypi kernel: [ 2109.900282] smsc95xx 1-1.1:1.0: eth0: Failed to write HW_CFG_LRST_ bit in HW_CFG register, ret = -110
Jan 11 14:58:37 raspberrypi kernel: [ 2115.900318] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000014
Jan 11 14:58:37 raspberrypi kernel: [ 2115.900352] smsc95xx 1-1.1:1.0: eth0: Failed to write HW_CFG_LRST_ bit in HW_CFG register, ret = -110
Jan 11 14:58:43 raspberrypi kernel: [ 2121.900411] smsc95xx 1-1.1:1.0: eth0: Failed to write register index 0x00000014


If this problem is supposedly fixed how can I tell if my system has the patch?

Thanks
ewig
Posts: 10
Joined: Fri Jan 11, 2013 3:11 pm
by bgirardot » Fri Jan 11, 2013 3:54 pm
It does not look like you have the development firmware installed from your uname -a output.

If you want the development firmware and kernel to see if it fixes your problem go to this page:

https://github.com/Hexxeh/rpi-update

and install the rpi-update tool and then run it.

I never had to do the 'preparation' steps, just the install and update.

There is no easy way to downgrade from the development stuff, so you might want to back up your SD card first so you can easily go back to it if the development code causes you any issues, not likely, but certainly possible.
Posts: 517
Joined: Wed Oct 10, 2012 6:20 am
Location: Switzerland
by ewig » Fri Jan 11, 2013 3:59 pm
bgirardot wrote:It does not look like you have the development firmware installed from your uname -a output.

If you want the development firmware and kernel to see if it fixes your problem go to this page:

https://github.com/Hexxeh/rpi-update

and install the rpi-update tool and then run it.

I never had to do the 'preparation' steps, just the install and update.

There is no easy way to downgrade from the development stuff, so you might want to back up your SD card first so you can easily go back to it if the development code causes you any issues, not likely, but certainly possible.


Cool, thank you.
ewig
Posts: 10
Joined: Fri Jan 11, 2013 3:11 pm
by ewig » Sun Jan 13, 2013 9:21 pm
ewig wrote:
bgirardot wrote:...
There is no easy way to downgrade from the development stuff, so you might want to back up your SD card first so you can easily go back to it if the development code causes you any issues, not likely, but certainly possible.


Cool, thank you.
ewig


My Pi has 48 hours uptime since applying this update. Could be the fix is working for me.

ewig
Posts: 10
Joined: Fri Jan 11, 2013 3:11 pm
by JasonNaughton » Sat Jan 26, 2013 1:32 am
Hi ewig,

Can you do me a favour and post the following:

uname -a

after you did the firmware update? I noticed this morning the same log entry in my syslog. Now my pi is still operational but I was wondering if I was heading into issues to come. I'm using the same kernel version that you were using. My uname -a contains:

Linux XXXXXX 3.2.27+ #250 PREEMPT Thu Oct 18 19:03:02 BST 2012 armv6l GNU/Linux

Yet someone said you weren't using the development firmware from your kernel release. So I'm now curious what you have now after the update. Yet when I cat /etc/issue I have:

# cat /etc/issue
Debian GNU/Linux 7.0 \n \l

I've installed rasbian wheezy and have done a number of updates over the past few weeks.

Cheers

Jason
Posts: 5
Joined: Tue Jan 15, 2013 10:46 pm