Stephen Hemminger wrote: > Same thing with KVM (kernel 2.6.28) and e1000, the driver thinks it is up > (per dmesg) but the carrier is not asserted.
I believe what you are seeing on KVM and vmware are unrelated. In the case of vmware it looks like it keeps enabling the the carrier after the TX hangs, whereas in the case of KVM the carrier is off as per ethtool so you don't have it sending any traffic. In the KVM case you may want to check to verify the interface is actually up in ifconfig. The only way I could reproduce the ethtool and dmesg results you had seen in KVM was to bring up the interface and then bring it down via ifconfig. My reasoning on this is the link up message is always associated with netif_carrier_on, and only 3 spots call netif_carrier_off, and of those only e1000_down could do so without dumping some sort of message into the dmesg log and that would only be called if the interface was brought back down. I've seen something like this before as a result of Network Manager running in the background. As for the vmware issue I'm wondering if the emulation layer implemented the disconnect feature in the emulated e1000 layer correctly. What is the ethtool output after you have disconnected the emulated interface from the host? Also could you send us an ethtool -d register dump from the interface after the disconnect? Thanks, Alex ------------------------------------------------------------------------------ Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel
