pings return two at a time, every two seconds?

Issues related to configuring your network
gaylon
Posts: 21
Joined: 2005/03/23 16:35:33
Location: New Mexico, USA

Re: pings return two at a time, every two seconds?

Post by gaylon » 2007/04/27 18:41:26

JMicron DISABLED (no network problem):
[code]
[root]# cat /proc/interrupts
CPU0 CPU1
0: 138800 99860 IO-APIC-edge timer
1: 42 16 IO-APIC-edge i8042
8: 3 0 IO-APIC-edge rtc
9: 0 0 IO-APIC-level acpi
14: 34910 20164 IO-APIC-edge ide0
15: 0 0 IO-APIC-edge libata
50: 266 0 IO-APIC-level HDA Intel
169: 5268 5235 IO-APIC-level uhci_hcd:usb1, i915@pci:0000:00:02.0
177: 143 0 IO-APIC-level uhci_hcd:usb2, eth0
217: 740 0 IO-APIC-level uhci_hcd:usb3, ehci_hcd:usb7
225: 0 0 IO-APIC-level uhci_hcd:usb4, libata
233: 0 0 IO-APIC-level uhci_hcd:usb5, ehci_hcd:usb6
NMI: 0 0
LOC: 220278 231030
ERR: 0
MIS: 0
[root]#
[/code]

JMicron ENABLED (has two-second network problem):
[code]
[root]# cat /proc/interrupts
CPU0 CPU1
0: 340998 290057 IO-APIC-edge timer
1: 57 2 IO-APIC-edge i8042
8: 3 0 IO-APIC-edge rtc
9: 0 0 IO-APIC-level acpi
14: 33087 21706 IO-APIC-edge ide0
15: 0 0 IO-APIC-edge libata
50: 0 0 IO-APIC-level uhci_hcd:usb5, ehci_hcd:usb6
58: 266 0 IO-APIC-level HDA Intel
169: 20318 19196 IO-APIC-level uhci_hcd:usb1, libata, i915@pci:0000:00:02.0
177: 4999 0 IO-APIC-level ide2, uhci_hcd:usb2, eth0
225: 344 0 IO-APIC-level uhci_hcd:usb3, ehci_hcd:usb7
233: 0 0 IO-APIC-level uhci_hcd:usb4, libata
NMI: 0 0
LOC: 613965 622100
ERR: 0
MIS: 0
[root]#
[/code]

I'm very happy to provide whatever info to anyone with knowledge to figure out exactly what's going on here, so it can be fixed. There are zillions of these P5B-VM mobos out there, so it's well worth some of my time, to improve linux support for all of them. Though I can get by for now, without the working JMicron chip.

Motherboard BIOS is ver 0613 (2007/03/01), which is the latest at Asus.
Realtek driver is also latest: v 1.05 (2006/11/27)
I didn't do anything about a driver for the JMicron chip, since the CentOS 5 install disks did their thing without complaint.

Useless background info:
Now a few puzzle pieces start to come together, in the tapioca pudding I call a brain... When I first installed CentOS-5 on this box (I'd had WinXP working, to test the hardware), the ethernet wouldn't link. See my [url=http://www.centos.org/modules/newbb/viewtopic.php?topic_id=7838]earlier thread[/url] for the historical record.
In the course of debugging that, I flashed the bios to the latest (0613), which reset all bios settings to default, and required me to re-tweak the settings (disable the floppy, set audio output mode, etc etc etc). And the network mysteriously began working. But I didn't think to test how *well* it was working. I just assumed that 'yum update' took 14 hours to download 281MB (I have 1.5Mbps ISP) because CentOS-5 had just been released the day before and all the repos were overloaded. I now suspect that the problem was that the 8111B on this box was hobbled (perhaps only allowing a few pkts in a burst, every 2 seconds). I'd just not opened Firefox more than once or twice, and didn't notice that the network was super slow. I set this box aside and began installing CentOS-5 on the server. Eventually got to the setting up NFS exports, and came back to this box and booted it up as a client, to config & test. At some point I did a ping, and noticed the weirdness.

Lenard
Posts: 2283
Joined: 2005/11/29 02:35:25
Location: Indiana

Re: pings return two at a time, every two seconds?

Post by Lenard » 2007/04/28 12:07:17

Yep, you do seem to have an IRQ conflict issue;

JMicron ENABLED (has two-second network problem):
177: 4999 0 IO-APIC-level ide2, uhci_hcd:usb2, eth0

JMicron DISABLED (no network problem):
177: 143 0 IO-APIC-level uhci_hcd:usb2, eth0

Notice how ide2 is sharing the same IRQ (177) as eth0, this is not good.

My first suggestion is to build your own custom kernel using as little of the ATA/ATAPI/MFM/RLL support as possible. Enable a much of the Serial ATA (prod) and Parallel ATA (experimental) drivers as possible. I have never liked how Red Hat kernels assign IRQ address along with a few other tings myself (it is a personal thing) and use a kernel built from the vanilla kernel sources instead. The 2.6.21 kernel source has just been released and seems so far to be well worth the time to build.

You can also try downgrading the BIOS back to the older version if possible and check if the issue still exists, if not then keep this older BIOS and enjoy. If the issue is still present or if you cannot downgrade the BIOS then as my suggestion try using the current Fedora Core 7 kernel, they seem a bit better and have the Serial ATA and Parallel ATA configured over the older ATA/ATAPI/MFM/RLL support. This just may resolve the issue also instead of building your own kernel. You can install this kernel and keep your current CentOS5 kernel(s) installed for testing.

gaylon
Posts: 21
Joined: 2005/03/23 16:35:33
Location: New Mexico, USA

Re: pings return two at a time, every two seconds?

Post by gaylon » 2007/04/28 17:09:12

Thanks to Lenard for the advice. I may have a go at fixing this at some point, but for now, can just live without the JMicron chip.
Should I report this behaviour to someone(s) able to improve things for future versions?
What would be the/an appropriate mechanism for doing so?
I'm not sure whether this issue would be most accurately attributed to the Realtek driver, or to the kernel itself, or what?

Lenard
Posts: 2283
Joined: 2005/11/29 02:35:25
Location: Indiana

Re: pings return two at a time, every two seconds?

Post by Lenard » 2007/04/28 17:29:22

I'm not quite sure about if this a kernel or Realtek driver or BIOS issue myself. As a suggestion you could start by posting a bug report here; http://bugs.centos.org/main_page.php

Or check the bugs reports here (and maybe submit a bug report also); https://bugzilla.redhat.com/bugzilla/index.cgi

As part of the bug reporting provide a link to this thread.

Post Reply