[SOLVED] no network at startup, but network after restart

rtmistler · 04-24-2015, 07:19 AM

To me there's a point, not shown in your logs there (because it is the same in both logs) where the Ethernet card is initialized and the networking stack begins to work with it. During cold boot, that Ethernet card is not ready for use for some reason, however you see no failures, in fact it appears as if you do not see that card at all during cold boot. Are there any instances of "e1000e" in the cold boot log at all?

The problem is 16-18 seconds after boot ... that device should be all set electrically and also should have been initialized properly by device discovery.

If you can determine the driver which is for the Ethernet NIC, then attempt to load or reload that driver after a cold boot and see if you can bring the network stack up manually. If that is possible, then I'd try something odd like forcibly loading that driver much earlier in the boot, say at runlevel2 and see if that fixes the cold boot issues.

And if the driver is already loaded, but things didn't work, unload the driver and reload it and see if that gets it working. Likely upon a forcible load or reload you'd see the e1000e device being discovered in your system log.

floppy_stuttgart · 04-24-2015, 10:53 AM

Quote:

Originally Posted by arindamsingh

1. Differences between lsmod-cold-boot and lsmod-hot-boot not found in contents.
But the sequencing of items are diferent.

a) post the whole files as attachment (if possible, dmesg too).. because something is perhaps different
b) or post dmesg at the point where it becomes different. You see from the IRQ for the nvidia the PC load differently the PCIs bbetween cold and hot boot (the e1000e is loaded after nvidia in the cold-boot? we dont see it)
As per indication of "rtmistler", I would remove the e1000e driver everytime, and load it after automatically. Look at "manual page modprobe.conf" how to do it or here
http://www.linuxquestions.org/questi...es-4175536656/
or google "change module loading sequence linux".
= "rmmod e1000e"
= "modprobe e1000e"
or perhaps a waitdev=60 bootparameter? see http://distro.ibiblio.org/fatdog/web...t-options.html

arindamsingh · 04-30-2015, 10:05 PM

Some experiments:
1. I booted from mint 17 installation live disk.
Network card not detected.
Restarted the system from the live cd.
Network is up.
Booting from ubuntu 14.04 shows the same behaviour.
2. Booting from ubuntu 12.04 live disk shows the same behaviour.
3. Tried to boot from Arch linux latest version in usb.
The system did not boot, asked to reboot.
When rebooted from the hard disk, network is up.
4. Suspected some thing else is happening. Shutdown the system.
Put an unbootable usb stick and booted the system.
System reported the disk is not bootable and prompted to reboot.
On reboot from the hard disk, network is up.

Does it mean there is some problem with the bios?

floppy_stuttgart · 05-01-2015, 04:56 AM

Quote:

Originally Posted by arindamsingh

Some experiments:
1. I booted from mint 17 installation live disk.
Network card not detected.
Restarted the system from the live cd.
Network is up.
Booting from ubuntu 14.04 shows the same behaviour.
2. Booting from ubuntu 12.04 live disk shows the same behaviour.
3. Tried to boot from Arch linux latest version in usb.
The system did not boot, asked to reboot.
When rebooted from the hard disk, network is up.
4. Suspected some thing else is happening. Shutdown the system.
Put an unbootable usb stick and booted the system.
System reported the disk is not bootable and prompted to reboot.
On reboot from the hard disk, network is up.

Does it mean there is some problem with the bios?

Funny tries. But the new question has nothing to do with the original question.
Please stick to the original question and proposed solutions, else I would imagine people here will loose theier willingness to help you.

rtmistler · 05-01-2015, 07:08 AM

The testing was all over the map.

So what do you wish to do?

Try distributions and find something which works and leave it at that?

MINT and Ubuntu are both Ubuntu unless you tried MINT Debian specifically. MATE has nothing to do with the base of the distribution, that's just a desktop choice. So try something Debian based. I've not used Arch, but I see/hear that it's a rolling distribution and thus very edge, so perhaps not the best choice if you want common behavior and stability all the time.

Your question/conclusion is that this could be BIOS is not based on any level of comprehensive analysis.

I know it may seem tedious, but a suggestion is to save the dmesg after a cold boot, save the dmesg after a warm boot. Sounds like you already have those, right?

Edit them in emacs, make the window huge, enter into the command prompt "split-window-horizontally" and this will split the buffer into two side by side halves. Then get one log file into the left and the other log file into the right. These should be near identical. You can view them line by line.

You also could choose to use ediff-files as well and use the PIPE "|" to cause the compare buffers to be side by side. Either case, it's worth doing and understanding the sequencing.

Or as floppy_stuttgart suggests, post them as an attachment. Zip them and post that as an attachment. We'll do the comparisons.

If you're just going to half-haphazardly try different distributions, then claim "BIOS", this is all I can do for you.

Well I can suggest this thread about BIOS for your fun reading.

arindamsingh · 05-07-2015, 06:16 AM

Sorry the delay; I had to be away for a week.

The files are attached.
lsmod on hot boot, lsmod on cold boot, lsmod-hot-cold-diff, dmesg on cold boot and dmesg on hot boot.
lsmod output is a smaller file, so I merged both lsmod_hot and lsmod_cold in one file with a line and some tabs gap.
I inserted a heading to make it readable, with # marks.
For comparing of dmesg-hot and dmesg-cold, I used sdiff command.
This is not attached, since the system allows only five.
If you ask, I will attach it in my next mail. But you can also get it from those two.

Sorry to overload you; for I do not know how to figure out the meanings of the differences.

rtmistler · 05-07-2015, 07:26 AM

My conclusion is that you try what floppy_stuttgart suggested in post # 17 to remove the module, reload it, and see if this will allow the cold boot case to work because it seems that during cold boot the driver does not discover or configure the device properly. When it gets engaged again, because it hasn't been powered down, warm boot case, it then decides to respond properly.

I'd also check Intel and see if they have an updated e1000 driver for Linux. A quick check brought me to this link where I see a version of the driver for Linux at about 8.0.35 which to me is definitely far beyond version 2.3.2-k.

Some Analysis of What I See in Your Logs:

Between lines 668 and 683 in the HOT boot log, non-relevant lines removed:

Code:

[    0.708465] e1000e: Intel(R) PRO/1000 Network Driver - 2.3.2-k
[    0.708466] e1000e: Copyright(c) 1999 - 2013 Intel Corporation.
[    0.716476] e1000e 0000:00:19.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
[    0.716498] e1000e 0000:00:19.0: irq 44 for MSI/MSI-X

Between lines 667 and 683 for the COLD boot log, same non-relevant lines removed:

Code:

[    0.893562] e1000e: Intel(R) PRO/1000 Network Driver - 2.3.2-k
[    0.893564] e1000e: Copyright(c) 1999 - 2013 Intel Corporation.
[    0.897124] e1000e 0000:00:19.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
[    0.897152] e1000e 0000:00:19.0: irq 44 for MSI/MSI-X

Lines 685-689 in the HOT boot log:

Code:

[    0.980746] e1000e 0000:00:19.0 eth0: registered PHC clock
[    0.980749] e1000e 0000:00:19.0 eth0: (PCI Express:2.5GT/s:Width x1) 90:b1:1c:a1:a5:01
[    0.980750] e1000e 0000:00:19.0 eth0: Intel(R) PRO/1000 Network Connection
[    0.980799] e1000e 0000:00:19.0 eth0: MAC: 10, PHY: 11, PBA No: 1011FF-0FF

Lines 828-835 in the HOT boot log:

Code:

[   18.322614] e1000e 0000:00:19.0: irq 44 for MSI/MSI-X
[   18.424383] e1000e 0000:00:19.0: irq 44 for MSI/MSI-X
[   18.424494] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[   18.424705] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[   18.822149] init: samba-ad-dc main process (873) terminated with status 1
[   20.007244] e1000e: eth0 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None
[   20.007248] e1000e 0000:00:19.0 eth0: 10/100 speed: disabling TSO

Line 746 in the COLD boot log:

Code:

[    1.913476] e1000e: probe of 0000:00:19.0 failed with error -3

arindamsingh · 05-10-2015, 11:02 PM

This did not solve the problem.
What I did:
Downloaded the latest driver e1000e-3.1.0.2.tar.gz from intel.
$ sudo rmmod e1000e
$ locate e1000e.ko
$ sudo rm -r /lib/modules/3.13.0-37-generic/kernel/drivers/net/ethernet/intel/e1000e/
$ sudo cp e1000e-3.1.0.2.tar.gz /usr/local/src/
$ sudo cd /usr/local/src/
$ sudo tar zxf e1000e-3.1.0.2.tar.gz
$ sudo cd e1000e-3.1.0.2/src/
$ sudo make install
$ sudo modprobe e1000e
$ sudo update-initramfs -u -k all
$ sudo poweroff

Now, I hard-boot the system.
No network after start-up.
I restart. Network is up, as earlier.

Since I am new to linux, there is every possibility I have not doen what you intended me to do.
Please let me know. I will try again.
Thanks

floppy_stuttgart · 05-10-2015, 11:49 PM

Quote:

Originally Posted by arindamsingh

This did not solve the problem.
Thanks

What did not solve the problem?
Please describe the commands you have made which "did not solve the problem".

rtmistler · 05-11-2015, 06:20 AM

Quote:

Originally Posted by arindamsingh

This did not solve the problem.
What I did:
Downloaded the latest driver e1000e-3.1.0.2.tar.gz from intel.
$ sudo rmmod e1000e
$ locate e1000e.ko
$ sudo rm -r /lib/modules/3.13.0-37-generic/kernel/drivers/net/ethernet/intel/e1000e/
$ sudo cp e1000e-3.1.0.2.tar.gz /usr/local/src/
$ sudo cd /usr/local/src/
$ sudo tar zxf e1000e-3.1.0.2.tar.gz
$ sudo cd e1000e-3.1.0.2/src/
$ sudo make install
$ sudo modprobe e1000e
$ sudo update-initramfs -u -k all
$ sudo poweroff

Now, I hard-boot the system.
No network after start-up.
I restart. Network is up, as earlier.

Since I am new to linux, there is every possibility I have not doen what you intended me to do.
Please let me know. I will try again.
Thanks

When you look at your system log, does it now show that the newer driver version is running?

arindamsingh · 05-11-2015, 07:06 AM

Looks like, the new driver is running.
The old version of e1000e was 2.3.2.
The new one has version 3.1.0.2.

Attached the relevant lines from system log on cold boot.
Thanks

rtmistler · 05-11-2015, 07:12 AM

Now once the system is up with a cold boot and the network is not working.

Can you next rmmod and insmod and then try to raise your network and have it work?

Because if so, then you can write a script to do these actions and run that later in the boot time, after the driver has loaded the first time.

That's a work-around I know, but if it ends up working...

JeremyBoden · 05-11-2015, 07:33 AM

Is this a power supply problem perhaps?

From a cold boot, when GRUB brings up a "select kernel selection" press CTRL+ALT+DLT.
Then allow to proceed when the PC restarts.

If this works, then I think you have a hardware problem of some kind.

arindamsingh · 05-11-2015, 11:55 PM

After a cold boot, I tried the following as suggested by rtmistler:

$ sudo service network-manager stop
$ sudo rmmod e1000e
$ sudo insmod e1000e
(It gave some error)
$ sudo modprobe e1000e
(nothing reported)
$ sudo service network-manager start
(nothing reported)

But the network is still down.

Next, tried JeremyBoden's suggestion:

At the grub screen, pressed Ctrl+Alt+Delete.
Allowed the system to boot. (It is really a restart.)
The network is up.

You think, there is some hardware problem.
Any guess? So that I can ask the DELL technicians to see it, with some pointer.

Thanks

JeremyBoden · 05-12-2015, 05:42 AM

At least it gives you a rapid reboot.
Note that GRUB isn't looking at your network connection etc...

I've seen this before - on an old machine.
It eventually fixed itself, although I left the power connected to the PC 24/7.
So I never actually found the real cause.

Does your BIOS retain the correct date/time between cold boots?
If not, then replace the BIOS battery.