Eoin as soon as I find some time I will check the issue you reported. Stay tuned
Luca Eoin Miller wrote: > Hey all, > > We have been attempting to troubleshoot an issue with the interface > hanging when using PF_RING with the TNAPI driver. > > Everything in /var/log/messages appears as normal: > > ---SNIP--- > kernel: [ 2113.940087] igb 0000:06:00.1: PCI INTB disabled > kernel: [ 2114.030088] igb 0000:06:00.0: PCI INTA disabled > Kernel: [ 2124.021606] Intel(R) Gigabit Ethernet Network Driver - > version 1.3.19.3 > kernel: [ 2124.021610] Copyright (c) 2007-2008 Intel Corporation. > kernel: [ 2124.021664] igb 0000:06:00.0: PCI INT A -> GSI 32 (level, > low) -> IRQ 32 > kernel: [ 2124.021861] igb: 0000:06:00.0: igb_validate_option: > Interrupt Mode set to 3 > kernel: [ 2124.253167] igb: eth1: igb_probe: Intel(R) Gigabit Ethernet > Network Connection > kernel: [ 2124.253172] igb: eth1: igb_probe: (PCIe:2.5Gb/s:Width x4) > 00:1b:21:37:a7:74 > kernel: [ 2124.253176] igb: eth1: igb_probe: Using MSI-X interrupts. 4 > rx queue(s), 4 tx queue(s) > kernel: [ 2124.253196] igb 0000:06:00.1: PCI INT B -> GSI 32 (level, > low) -> IRQ 32 > kernel: [ 2124.483177] igb: eth2: igb_probe: Intel(R) Gigabit Ethernet > Network Connection > kernel: [ 2124.483181] igb: eth2: igb_probe: (PCIe:2.5Gb/s:Width x4) > 00:1b:21:37:a7:75 > kernel: [ 2124.483185] igb: eth2: igb_probe: Using MSI-X interrupts. 1 > rx queue(s), 1 tx queue(s) > kernel: [ 2124.483212] igb-TNAPI - (C) 2006-09 ntop.org > kernel: [ 2289.222725] TNAPI: configure_tnapi > kernel: [ 2289.222729] TNAPI: init_tnapi(eth1) > kernel: [ 2289.222732] ---- TNAPI: eth1 added at device slot 0 > kernel: [ 2289.222733] TNAPI: init_tnapi num_rx_queues=4 > kernel: [ 2289.222815] TNAPI: spawn thread [eth1], 0 > kernel: [ 2289.222859] TNAPI: spawn thread [eth1], 1 > kernel: [ 2289.222901] TNAPI: spawn thread [eth1], 2 > kernel: [ 2289.222938] TNAPI: spawn thread [eth1], 3 > kernel: [ 2289.227089] ADDRCONF(NETDEV_UP): eth1: link is not ready > kernel: [ 2289.234889] igb: eth1: igb_watchdog_task: NIC Link is Up > 1000 Mbps Full Duplex, Flow Control: None > kernel: [ 2289.237885] ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready > kernel: [ 2342.011141] [PF_RING] successfully allocated 1454080 bytes > at 0xffffc200040e2000 > kernel: [ 2342.011146] [PF_RING] allocated 4106 slots > [slot_len=354][tot_mem=1454080] > kernel: [ 2342.011629] device eth1 entered promiscuous mode > kernel: [ 2352.873183] device eth1 left promiscuous mode > kernel: [ 2352.901278] [PF_RING] removed /proc/net/pf_ring/5125-eth1.29 > ---SNIP--- > > However, when we are running any libpcap application it only captures > traffic for a few seconds and then the interface locks and sees no > further traffic until the network driver (IGB) is unloaded and > reloaded through rmmod and insmod. Below is an example of the > interface hanging with the supplied pcount utility: > > [r...@danzig]# /usr/src/PF_RING/userland/examples/pcount -i e...@1 > Capturing from e...@1 > ========================= > Absolute Stats: [1204 pkts rcvd][0 pkts dropped] > Total Pkts=1204/Dropped=0.0 % > 1204 pkts [1204.3 pkt/sec] - 531703 bytes [4.25 Mbit/sec] > ========================= > ========================= > Absolute Stats: [2408 pkts rcvd][0 pkts dropped] > Total Pkts=2408/Dropped=0.0 % > 1204 pkts [602.0 pkt/sec] - 531703 bytes [2.13 Mbit/sec] > ========================= > Actual Stats: 2408 pkts [1000.1 ms][2407.7 pkt/sec] > ========================= > ========================= > Absolute Stats: [3612 pkts rcvd][0 pkts dropped] > Total Pkts=3612/Dropped=0.0 % > 1204 pkts [401.3 pkt/sec] - 531703 bytes [1.42 Mbit/sec] > ========================= > Actual Stats: 1204 pkts [1000.0 ms][1203.9 pkt/sec] > ========================= > ========================= > Absolute Stats: [4816 pkts rcvd][0 pkts dropped] > Total Pkts=4816/Dropped=0.0 % > 1204 pkts [301.0 pkt/sec] - 531703 bytes [1.06 Mbit/sec] > ========================= > Actual Stats: 1204 pkts [1000.0 ms][1203.9 pkt/sec] > ========================= > ========================= > Absolute Stats: [6020 pkts rcvd][0 pkts dropped] > Total Pkts=6020/Dropped=0.0 % > 1204 pkts [240.8 pkt/sec] - 531703 bytes [0.85 Mbit/sec] > ========================= > Actual Stats: 1204 pkts [1000.0 ms][1203.9 pkt/sec] > ========================= > <control-c> > > You may note that the "Absolute Stats" and "Total pkts" continues to > increment, but the last line does not for number of packets or bytes. > It also does appear to be odd that this happens immediately after > receiving 1024 packets so some buffer may not be getting flushed? This > appears to happen for any libpcap application so we are thinking the > issue may reside within the igb TNAPI driver or PF_RING. Any insight > would be much appreciated. > > -- Eoin > > > _______________________________________________ > Ntop-dev mailing list > [email protected] > http://listgateway.unipi.it/mailman/listinfo/ntop-dev _______________________________________________ Ntop-dev mailing list [email protected] http://listgateway.unipi.it/mailman/listinfo/ntop-dev
