>-----Original Message----- >From: Jeffrey Marshall [mailto:jmarshl...@gmail.com] >Sent: Thursday, February 26, 2015 12:25 AM >Subject: [E1000-devel] Interface Suddenly Started Dropping Packet > >Hi, > >I believe that I am encountering a similar issue with a that was reported >to this mailing list earlier, without a resolution: >https://sourceforge.net/p/e1000/mailman/e1000->devel/thread/9b4a1b1917080e46b64f07f2989dadd65347f...@orsmsx114.amr.corp.intel.com/
Even if you think it is the same problem (and could very well be) I'd still suggest you open your own ticket as this makes it easier for us to follow the issue and reduces the possibility of confusion if the root cause turns out to be different. >We are using a Intel X520 DP 10Gb DA/SFP+ NIC, and we were successfully >capturing packets up until we installed PF_RING and rebooted the server. >Immediately after reboot, we started dropping 100% of packets, so we rolled >back all of our changes, and unfortunately were not able to re-establish a >link to capture packets. I'm not familiar with pf_ring and it's not something we test our drivers with, so I can't say for sure how it is related, but if your setup worked prior to installing pf_ring I'd guess it is related somehow. Have you tried re-installing the OS and doing a cold reboot (in case pf_ring caused some issue with the driver and the HW)? >We posted some additional details here: >https://communities.intel.com/message/279791. As I mentioned earlier we usually track issues through e1000.sf.net, so I'd suggest you open a bug there. >Could someone please give us some recommendations on how to further >investigate this? We have experimented with various MTU sizes, and tried >disabling auto-negotiate and manually set the connection speed. I'm guessing you are seeing the packets being dropped and the rx_missed_errors counter incrementing simultaneously? There are couple things to note here: 1. The rx_missed_errors counter comes from the MPC register which usually increments when there are no sufficient buffers to receive the packets, or bus bandwidth. 2. rx_pkts_nic increment, but rx_packets doesn't which means that the packets reached the HW, but were never processed by the driver. So there seems to be something suffocating the Rx buffer, or the HW is in a weird state which is causing it to drop them. So basically try to reset the adapter and do a clean OS install (if you haven't already). I'll see if we can reproduce this issue with pf_ring as well. Thanks, Emil ------------------------------------------------------------------------------ Dive into the World of Parallel Programming The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/ _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
