Take the 2-minute tour ×
Unix & Linux Stack Exchange is a question and answer site for users of Linux, FreeBSD and other Un*x-like operating systems.. It's 100% free, no registration required.

I have an IBM BladeCenter blade running Fedora 15. Everything works fine with 2.6.38 kernels, but when I boot with a 2.6.40 kernel, I lose connectivity after a few seconds. Removing and then re-inserting the tg3 module restores connectivity for about five seconds, after which I lose connectivity again.

These are dual-Xeon blades. This one has two 2.8GHz Xeons with HT and 2GB of RAM. The blade is an 8832-L1X running BIOS version 1.13. The output of lspci is:

00:00.0 Host bridge: Broadcom CMIC-LE Host Bridge (GC-LE chipset) (rev 33)    
00:00.1 Host bridge: Broadcom CMIC-LE Host Bridge (GC-LE chipset)    
00:00.2 Host bridge: Broadcom CMIC-LE Host Bridge (GC-LE chipset)    
00:01.0 VGA compatible controller: ATI Technologies Inc Rage XL (rev 27)
00:0f.0 Host bridge: Broadcom CSB6 South Bridge (rev b0)
00:0f.1 IDE interface: Broadcom CSB6 RAID/IDE Controller (rev b0)
00:0f.2 USB Controller: Broadcom CSB6 OHCI USB Controller (rev 05)
00:0f.3 ISA bridge: Broadcom GCLE-2 Host Bridge
00:10.0 Host bridge: Broadcom CIOB-E I/O Bridge with Gigabit Ethernet (rev 12)
00:10.2 Host bridge: Broadcom CIOB-E I/O Bridge with Gigabit Ethernet (rev 12)
01:00.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5704S Gigabit Ethernet (rev 02)    
01:00.1 Ethernet controller: Broadcom Corporation NetXtreme BCM5704S Gigabit Ethernet (rev 02)

Here's a dmesg excerpt with kernel-PAE-2.6.38.8-32.fc15.i686 (working):

[11.545123] tg3.c:v3.116 (December 3, 2010)
[11.545152] tg3 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[11.599499] tg3 0000:01:00.0: eth0: Tigon3 [partno(BCM95704A41) rev 2002] (PCIX:133MHz:64-bit) MAC address
[11.599510] tg3 0000:01:00.0: eth0: attached PHY is serdes (1000Base-SX Ethernet) (WireSpeed[0])
[11.599518] tg3 0000:01:00.0: eth0: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[1] TSOcap[0]
[11.599525] tg3 0000:01:00.0: eth0: dma_rwctrl[769f4000] dma_mask[64-bit]
[11.599577] tg3 0000:01:00.1: PCI INT B -> GSI 17 (level, low) -> IRQ 17
[11.627997] tg3 0000:01:00.1: eth1: Tigon3 [partno(BCM95704A41) rev 2002] (PCIX:133MHz:64-bit) MAC address
[11.628066] tg3 0000:01:00.1: eth1: attached PHY is serdes (1000Base-SX Ethernet) (WireSpeed[0])
[11.628074] tg3 0000:01:00.1: eth1: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[0] TSOcap[1]
[11.628082] tg3 0000:01:00.1: eth1: dma_rwctrl[769f4000] dma_mask[64-bit]
[22.000286] tg3 0000:01:00.0: eth0: Link is up at 1000 Mbps, full duplex
[22.000294] tg3 0000:01:00.0: eth0: Flow control is off for TX and off for RX

And here's one from kernel-PAE-2.6.40.6-0.fc15.i686 (not working):

[10.262141] tg3.c:v3.119 (May 18, 2011)
[10.262177] tg3 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[10.309325] tg3 0000:01:00.0: eth0: Tigon3 [partno(BCM95704A41) rev 2002] (PCIX:133MHz:64-bit) MAC address         
[10.309336] tg3 0000:01:00.0: eth0: attached PHY is serdes (1000Base-SX Ethernet) (WireSpeed[0], EEE[0])
[10.309344] tg3 0000:01:00.0: eth0: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[1] TSOcap[1]
[10.309351] tg3 0000:01:00.0: eth0: dma_rwctrl[769f4000] dma_mask[64-bit]
[10.309431] tg3 0000:01:00.1: PCI INT B -> GSI 17 (level, low) -> IRQ 17
[10.361613] tg3 0000:01:00.1: eth1: Tigon3 [partno(BCM95704A41) rev 2002] (PCIX:133MHz:64-bit) MAC address         
[10.361624] tg3 0000:01:00.1: eth1: attached PHY is serdes (1000Base-SX Ethernet) (WireSpeed[0], EEE[0])
[10.361633] tg3 0000:01:00.1: eth1: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[0] TSOcap[1]
[10.361640] tg3 0000:01:00.1: eth1: dma_rwctrl[769f4000] dma_mask[64-bit]
[21.054276] tg3 0000:01:00.0: eth0: Link is up at 1000 Mbps, full duplex
[21.054284] tg3 0000:01:00.0: eth0: Flow control is off for TX and off for RX

Again, the symptom is that networking works fine for a few seconds but then completely stops. Doing an rmmod tg3 followed by a modprobe tg3 restores connectivity for a few seconds. Nothing unusual appears in any logs.

I've tried a few kernel option workarounds. I don't remember exactly which, but I know I tried acpi=noirq, acpi=ht, irqpoll, and noapic.

modinfo, minus the many alias lines shows:

filename: /lib/modules/2.6.40.6-0.fc15.i686.PAE/kernel/drivers/net/tg3.ko
firmware:       tigon/tg3_tso5.bin
firmware:       tigon/tg3_tso.bin
firmware:       tigon/tg3.bin
version:        3.119
license:        GPL
description:    Broadcom Tigon3 ethernet driver
author:         David S. Miller (davem@redhat.com) and Jeff Garzik (jgarzik@pobox.com)
srcversion:     389C3BA89E4ECF8460A74C0
depends:    
vermagic:       2.6.40.6-0.fc15.i686.PAE SMP mod_unload 686
parm:           tg3_debug:Tigon3 bitmapped debugging message enable value (int)

Fedora reports the kernel at 2.6.40 for "compatibility with older userspace".

Update: Doing ifconfig eth0 down followed by a new ifconfig to bring it up and adding the appropriate routes makes everything work indefinitely. Doing an rmmod tg3 followed by a modprobe tg3 makes it work for a few seconds and then break again. So at least I have a workaround for now -- adding /etc/rc.d/init.d network restart into rc.local. I'd still like to know what's wrong and whether there's a proper fix.

share|improve this question
    
I wonder how come Fedora invented a 2.6.40 kernel. There is no such thing... Maybe they used some old pre-3.0 scripts to generate their system configs/paths and this results in unrecognized version strings... –  rozcietrzewiacz Oct 16 '11 at 21:11
    
What does modinfo tg3 show (with the so-called 2.6.40 kernel)? The vermagic is mostly interesting. –  rozcietrzewiacz Oct 16 '11 at 21:15
    
Could you try booting a livecd from other distro, with 3.0 kernel? (For example Ubuntu 11.10) This would clarify whether the problem is Fedora-specific. –  rozcietrzewiacz Oct 27 '11 at 13:12
    
Fedora "invented" 2.6.40 to keep up the numbering scheme (some braindead software was looking for the kernel version, and freaked out when seeing 3.0 instead of 2.6.xx). –  vonbrand Jan 16 '13 at 12:46
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.