Re: e1000: eth0: e1000_clean_tx_irq: Detected Tx Unit Hang

[Auke Kok <auke-jan.h.kok@intel.com>]

Sami Farin wrote:
> On Tue, Jan 09, 2007 at 15:59:30 -0800, Auke Kok wrote:
> > Sami Farin wrote:
> ...
> > > I do "ethtool -K eth0 tso off" now and check if I get the hang again. =)
> > I'm unsure whether v7.2.x already automatically disables TSO for 100mbit 
> > speed link, probably not. It should.
>
> It disabled it but I enabled it just for fun.
>  
> > Please try our updated driver from http://e1000.sf.net/ (7.3.20) against 
> > the same kernel. There are some changes with regard to the ich8/TSO driver 
> > that might affect this, so re-testing is worth it for us.
>
> I now run 7.3.20-NAPI.
>
> BTW. the Makefile is buggy: it does not get CC from kernel's Makefile.
> Using wrong compiler can cause for example a reboot when loading the module.
> (At least that's what happened with gcc-2.95.3 vs 3.x.x some years ago...)

I'll look into that, do you have any suggestions?

> > also, please always include the full dmesg output. Feel free to CC 
> > e1000-devel@lists.sourceforge.net on this.
>
> I enabled TSO again.  I write again if TSO causes problems.

There are known problems with that configuration, that's why the newer drivers disable 
TSO for 10/100 speeds.

do you really think that you can see the performance gain fro musing TSO at those speeds 
anyway? we don't ;). In any case you should keep TSO off for 10/100 speeds.

> Why shouldn't it work with 100 Mbps?  Not that it would help a lot,
> but I ask this on principle.
>
>   /* disable TSO for pcie and 10/100 speeds, to avoid
>    * some hardware issues */
>
> Issues on the motherboard or the NIC?

we (the e1000 team) don't write drivers for the motherboard, but only for the NIC 
component, so I hope that answers your question.

> 2007-01-10 02:39:51.889908500 <6>ACPI: PCI interrupt for device 0000:00:19.0 disabled
> 2007-01-10 02:39:54.545194500 <6>Intel(R) PRO/1000 Network Driver - version 7.3.20-NAPI
> 2007-01-10 02:39:54.545198500 <6>Copyright (c) 1999-2006 Intel Corporation.
> 2007-01-10 02:39:54.545395500 <6>ACPI: PCI Interrupt 0000:00:19.0[A] -> GSI 20 (level, low) -> IRQ 22
> 2007-01-10 02:39:54.545435500 <7>PCI: Setting latency timer of device 0000:00:19.0 to 64
> 2007-01-10 02:39:54.562905500 <6>e1000: 0000:00:19.0: e1000_probe: (PCI Express:2.5Gb/s:Width x1) 00:19:d1:00:5f:01
> 2007-01-10 02:39:54.638093500 <6>e1000: eth0: e1000_probe: Intel(R) PRO/1000 Network Connection
> 2007-01-10 02:40:07.513619500 <6>ADDRCONF(NETDEV_UP): eth0: link is not ready
> 2007-01-10 02:40:07.614768500 <6>e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full Duplex, Flow Control: None
> 2007-01-10 02:40:07.614770500 <6>e1000: eth0: e1000_watchdog: 10/100 speed: disabling TSO
> 2007-01-10 02:40:07.614771500 <6>ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
> 2007-01-10 02:40:09.271631500 <3>e1000: eth0: e1000_reset: Hardware Error
> 2007-01-10 02:40:10.930000500 <6>e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full Duplex, Flow Control: None
> 2007-01-10 02:40:10.930049500 <6>e1000: eth0: e1000_watchdog: 10/100 speed: disabling TSO
>
> PS. please do not delete Mail-Followup-To header field.

I hit "reply-all" and I have no control over which field thunderbird removes or adds. I 
have to manually add your e-mail address too? Maybe your mail client is broken instead? 
Don't you want to receive replies?

Cheers,

Auke
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/