Author Topic: Hybrid/Core suddenly refusing PCI NICs ... help!  (Read 4561 times)

asplodzor

  • Regular Poster
  • **
  • Posts: 17
    • View Profile
Hybrid/Core suddenly refusing PCI NICs ... help!
« on: July 24, 2008, 10:52:01 am »
OK, got another AMD 64 dvd image downloaded, and completed installation about seven hours ago. There were several minor hangups, but I'd been able to get past everything to the point where I thought I would experiment with net booting my laptop. DHCP found the boot location instantly and the laptop started downloading and installing (to ram?) the software needed, but got hung up after about five minutes. During this time, I had been trying to cure another problem by redoing the 'display res and sound' wizard on my hybrid/core which I had no idea would reload the core, let alone reboot the entire machine. The laptop seemed to take it in stride and tried to reconnect to the server, but it kept timing out. I let it sit for ten or so minutes before trying to jar it, but finally relented and hard powered it down. I had to run out the door on an errand for about an hour and briefly noticed that another computer on the internal network was showing a 'could not connect to server' page in firefox, but thought nothing of it.

Fast forward an hour later (five hours ago) and I tried net booting the laptop again, but this time it didn't find any DHCP. My first thought was a bad cord but I ruled that out, second thought was that my old wireless router (which I'd made static and was using as a switch) had locked up but I disproved that by connecting to its config page.

Five hours of HBOKing later, and here's where I'm at: The only NIC on the hybrid/core that is talking to any other machines at all now (old router, modem, internal computers etc) is the one integrated into the motherboard. I've tried three different PCI NICs, at least two of which were working just fine today even, and I can't get a peep out of them. This means that I have the choice of either an internal network, or an external network - not both. I've literally 'swapped interfaces' more than twenty times, coupled with every possible combination of powering down hardware in different orders and trying each configuration both ways (connecting integrated NIC to internal network and PCI to external, testing, then swapping them both without touching other settings). I've even isolated the internal network down to the hybrid/core, a crossover cable, and one other computer. The other computer can't even ping the PCI NICs! The oddest thing about this entire issue though, is that all three of the PCI NICs that I've tried have shown good link lights on themselves and and the other computer's NICs.

Here's a pretty common display in  Advanced | Network settings ... notice the lack of a MAC address on 'internal' (currently set to the PCI NIC)

EXTERNAL_IFACE     eth0
EXTERNAL_MAC    00:49:8D:50:B9:1E
EXTERNAL_IP    24.20.0.231
EXTERNAL_NETMASK    255.255.252.0
EXTERNAL_DHCP    1
INTERNAL_IFACE    eth1
INTERNAL_MAC    
INTERNAL_IP    192.168.1.1
INTERNAL_NETMASK    255.255.255.0
GATEWAY    24.20.0.1
DNS1    68.87.69.146
DNS2    68.87.78.130

The closest thing I have to a working theory about what's happened is that the 'display res and sound' wizard somehow has control of the NICs too. I'm guessing that my first PCI NIC (internal network) was old and failed when I was attempting to net boot, and the system ignored it when I started the wizard one or two minutes later, and treated the setup as one NIC with two IPs spoofed. If that is in fact the case, I have tried everything in my power short of wiping and reinstalling to correct the error - I've even exactly duplicated the events surrounding the wizard, but with a known good NIC, and reran the wizard. All this though, to no avail.

I'm going to give one more go at it tomorrow before completely reinstalling. Any ideas out there on anything else I should be looking into, or something that I'm not comprehending?

P.S. hardware problems and software problems can be annoying, but when they're both happening at the same time, it makes you want to throw you computer out the closest window ;-) (Though there still lies the possibility that this is entirely software related.)
« Last Edit: July 24, 2008, 10:55:14 am by asplodzor »

totallymaxed

  • LinuxMCE God
  • ****
  • Posts: 4660
  • Smart Home Consulting
    • View Profile
    • Dianemo - at home with technology
Re: Hybrid/Core suddenly refusing PCI NICs ... help!
« Reply #1 on: July 24, 2008, 12:11:44 pm »
OK, got another AMD 64 dvd image downloaded, and completed installation about seven hours ago. There were several minor hangups, but I'd been able to get past everything to the point where I thought I would experiment with net booting my laptop. DHCP found the boot location instantly and the laptop started downloading and installing (to ram?) the software needed, but got hung up after about five minutes. During this time, I had been trying to cure another problem by redoing the 'display res and sound' wizard on my hybrid/core which I had no idea would reload the core, let alone reboot the entire machine. The laptop seemed to take it in stride and tried to reconnect to the server, but it kept timing out. I let it sit for ten or so minutes before trying to jar it, but finally relented and hard powered it down. I had to run out the door on an errand for about an hour and briefly noticed that another computer on the internal network was showing a 'could not connect to server' page in firefox, but thought nothing of it.

Fast forward an hour later (five hours ago) and I tried net booting the laptop again, but this time it didn't find any DHCP. My first thought was a bad cord but I ruled that out, second thought was that my old wireless router (which I'd made static and was using as a switch) had locked up but I disproved that by connecting to its config page.

Five hours of HBOKing later, and here's where I'm at: The only NIC on the hybrid/core that is talking to any other machines at all now (old router, modem, internal computers etc) is the one integrated into the motherboard. I've tried three different PCI NICs, at least two of which were working just fine today even, and I can't get a peep out of them. This means that I have the choice of either an internal network, or an external network - not both. I've literally 'swapped interfaces' more than twenty times, coupled with every possible combination of powering down hardware in different orders and trying each configuration both ways (connecting integrated NIC to internal network and PCI to external, testing, then swapping them both without touching other settings). I've even isolated the internal network down to the hybrid/core, a crossover cable, and one other computer. The other computer can't even ping the PCI NICs! The oddest thing about this entire issue though, is that all three of the PCI NICs that I've tried have shown good link lights on themselves and and the other computer's NICs.

Here's a pretty common display in  Advanced | Network settings ... notice the lack of a MAC address on 'internal' (currently set to the PCI NIC)

EXTERNAL_IFACE     eth0
EXTERNAL_MAC    00:49:8D:50:B9:1E
EXTERNAL_IP    24.20.0.231
EXTERNAL_NETMASK    255.255.252.0
EXTERNAL_DHCP    1
INTERNAL_IFACE    eth1
INTERNAL_MAC    
INTERNAL_IP    192.168.1.1
INTERNAL_NETMASK    255.255.255.0
GATEWAY    24.20.0.1
DNS1    68.87.69.146
DNS2    68.87.78.130

The closest thing I have to a working theory about what's happened is that the 'display res and sound' wizard somehow has control of the NICs too. I'm guessing that my first PCI NIC (internal network) was old and failed when I was attempting to net boot, and the system ignored it when I started the wizard one or two minutes later, and treated the setup as one NIC with two IPs spoofed. If that is in fact the case, I have tried everything in my power short of wiping and reinstalling to correct the error - I've even exactly duplicated the events surrounding the wizard, but with a known good NIC, and reran the wizard. All this though, to no avail.

I'm going to give one more go at it tomorrow before completely reinstalling. Any ideas out there on anything else I should be looking into, or something that I'm not comprehending?

P.S. hardware problems and software problems can be annoying, but when they're both happening at the same time, it makes you want to throw you computer out the closest window ;-) (Though there still lies the possibility that this is entirely software related.)

Its unclear whether you had two nics in your core when you installed to it... but I would suggest the simplest solution is to reinstall a 2nd NIC (what chipsets do your NICs have?) and then reinstall from the DVD again. This will allow the installer to re-detect the NIcs and set them up correctly. Its possible to do this after installation too... but a reinstall is probably easier (you can use the retain /home directory option).

Unless your NICs are unsupported or the drivers are not present in the DVD then you should find this corrects your problems.

All the best

Andrew
Andy Herron,
CHT Ltd

For Dianemo/LinuxMCE consulting advice;
@herron on Twitter, totallymaxed+inquiries@gmail.com via email or PM me here.

Get Dianemo-Rpi2 ARM Licenses http://forum.linuxmce.org/index.php?topic=14026.0

Get RaspSqueeze-CEC or Raspbmc-CEC for Dianemo/LinuxMCE: http://wp.me/P4KgIc-5P

Facebook: https://www.facebook.com/pages/Dianemo-Home-Automation/226019387454465

http://www.dianemo.co.uk

Zaerc

  • Alumni
  • LinuxMCE God
  • *
  • Posts: 2256
  • Department of Redundancy Department.
    • View Profile
Re: Hybrid/Core suddenly refusing PCI NICs ... help!
« Reply #2 on: July 24, 2008, 05:03:00 pm »
Sounds like you need to edit /etc/udev/rules.d/70-persistent-net.rules to get rid of a bunch of cards you tried.  Maybe some corrections have to be made to /etc/network/interfaces as well.
« Last Edit: July 24, 2008, 05:20:30 pm by Zaerc »
"Change is inevitable. Progress is optional."
-- Anonymous


asplodzor

  • Regular Poster
  • **
  • Posts: 17
    • View Profile
Re: Hybrid/Core suddenly refusing PCI NICs ... help!
« Reply #3 on: July 25, 2008, 12:04:36 pm »
Its unclear whether you had two nics in your core when you installed to it... but I would suggest the simplest solution is to reinstall a 2nd NIC (what chipsets do your NICs have?) and then reinstall from the DVD again. This will allow the installer to re-detect the NIcs and set them up correctly. Its possible to do this after installation too... but a reinstall is probably easier (you can use the retain /home directory option).

Unless your NICs are unsupported or the drivers are not present in the DVD then you should find this corrects your problems.

All the best

Andrew

The core was never powered on with only one NIC, three of them were swapped, but only after I'd exhausted every was I was aware of testing eth1 which randomly stopped working. And yeah... reinstalling :-P

Sounds like you need to edit /etc/udev/rules.d/70-persistent-net.rules to get rid of a bunch of cards you tried.  Maybe some corrections have to be made to /etc/network/interfaces as well.

Unfortunately, I have almost zero knowledge so far about editing config files, so I just bit the bullet and reinstalled. Everything's working smoother than before, but I've ran into a new issue - on netboot, my laptop freezes on the orbiter screen. A picture background from flickr loads, and the overlayed (but not alpha-blended) GUI appears, but nothing responds at all. My laptop's a Toshiba A135-S7403 (the Best Buy one from last Black Friday) with a GMA 950 which I thought should be able to support OpenGL from here: http://forum.linuxmce.org/index.php?topic=5019.0 I deleted the device from the core and I'm currently setting up the netboot again with the lowest graphic settings, and I'll post how that works out.

I noticed something else while looking through the device configuration that I hadn't even thought about before - I'm using the 64bit install on my hybrid/core, but my laptop's i386. In the laptop's device configuration, it refers to it twice as amd64 - will this cause problems?

Thanks for the responses, both of you!
« Last Edit: July 25, 2008, 12:25:12 pm by asplodzor »

asplodzor

  • Regular Poster
  • **
  • Posts: 17
    • View Profile
Re: Hybrid/Core suddenly refusing PCI NICs ... help!
« Reply #4 on: July 25, 2008, 12:21:17 pm »
Ok, netbooting worked with the craptastic graphic settings, but now it just looks like a big mobile orbiter without a touchscreen. :-(

Think I'd have better luck wiping the 64bit core, installing the 32bit core, and attempting the medium graphic settings on the laptop? (OpenGL with overlay, but no alpha-blending)
« Last Edit: July 25, 2008, 12:24:32 pm by asplodzor »

Zaerc

  • Alumni
  • LinuxMCE God
  • *
  • Posts: 2256
  • Department of Redundancy Department.
    • View Profile
Re: Hybrid/Core suddenly refusing PCI NICs ... help!
« Reply #5 on: July 25, 2008, 06:19:42 pm »
To get the intel graphics chipset working properly you'll have to install the latest drivers from intel, most likely edit some config files, and probably change a few settings here and there.  Search the wiki and forum for more detailed information.
"Change is inevitable. Progress is optional."
-- Anonymous


niz23

  • Guru
  • ****
  • Posts: 361
    • View Profile
Re: Hybrid/Core suddenly refusing PCI NICs ... help!
« Reply #6 on: July 25, 2008, 07:58:00 pm »
Ok, netbooting worked with the craptastic graphic settings, but now it just looks like a big mobile orbiter without a touchscreen. :-(

Think I'd have better luck wiping the 64bit core, installing the 32bit core, and attempting the medium graphic settings on the laptop? (OpenGL with overlay, but no alpha-blending)

As Zaerc state. There are several problems with Intel Graphics and the driver shipped with ubuntu 7.10.
UI1 work in most cases. UI2 without alpha blending in many cases (no UI2 with alpha blend work), but in some cases UI2 come up after boot but suddenly freeze.
This problem is due to some bugs in the driver which have problems with how to handle hardware overlay which is needed by UI2.
The newest driver solve this problem. At least according to their bugfixes.
I´m very interested in getting the latest driver but haven´t had any time figure out how to compile one yet.
There seem to be noone that have done this earlier for ubuntu 7.10. Have searched many may places.

Seem like the only solution is to compile the driver which seem to depend on a lot of other things in XORG.

/niz23