| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Tue, 1 Mar 2016 13:08:35 +0000 (UTC) From: Bernhard Schmidt <berni@...kenwald.de> To: netdev@...r.kernel.org Subject: vmxnet3 LROv6 performance issues Hi, this started as a Debian bug (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=816377), but since it's affecting SLES as well I'm hoping to get some help here. Back in 2014 we migrated our VMware farm from old HP blade servers to new ones Old: HP BL490c Gen6, Flex10 something, BCM57711E New: HP BL460c Gen8, HP FlexFabric 630FLB, BCM57840 The new network chipset apparently sports IPV6 LRO support in hardware. Unfortunately that support was broken with in-tree vmxnet3 kernel modules back then, TCPv6 connections were stalling all over the place. Disabling LRO or using the vmxnet3 module provided by VMware fixed the issue. We are now seeing the issue again. Not as prominent as before, but still affecting some workloads pretty badly (for example a simple iperf3 benchmark towards an affected VM). client% iperf3 -c ping.lrz.de -t 30 Connecting to host ping.lrz.de, port 5201 [ 4] local 2001:4ca0:0:f000:bf47:886b:a813:df4f port 60174 connected to 2001:4ca0:0:101::81bb:a11 port 5201 [ ID] Interval Transfer Bandwidth Retr Cwnd [ 4] 0.00-1.00 sec 305 KBytes 2.50 Mbits/sec 34 8.37 KBytes [ 4] 1.00-2.00 sec 0.00 Bytes 0.00 bits/sec 24 2.79 KBytes [ 4] 2.00-3.00 sec 251 KBytes 2.06 Mbits/sec 75 2.79 KBytes [ 4] 3.00-4.00 sec 251 KBytes 2.06 Mbits/sec 124 2.79 KBytes [ 4] 4.00-5.00 sec 88.7 MBytes 744 Mbits/sec 24 332 KBytes [ 4] 5.00-6.00 sec 111 MBytes 931 Mbits/sec 0 476 KBytes [ 4] 6.00-7.00 sec 110 MBytes 921 Mbits/sec 0 478 KBytes [ 4] 7.00-8.00 sec 110 MBytes 923 Mbits/sec 0 481 KBytes [ 4] 8.00-9.00 sec 110 MBytes 921 Mbits/sec 0 502 KBytes [ 4] 9.00-10.00 sec 19.8 MBytes 166 Mbits/sec 27 2.79 KBytes [ 4] 10.00-11.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 11.00-12.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 12.00-13.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 13.00-14.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 14.00-15.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 15.00-16.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 16.00-17.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 17.00-18.00 sec 0.00 Bytes 0.00 bits/sec 16 2.79 KBytes [ 4] 18.00-19.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 19.00-20.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 20.00-21.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 21.00-22.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 22.00-23.00 sec 0.00 Bytes 0.00 bits/sec 22 2.79 KBytes [ 4] 23.00-24.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 24.00-25.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 25.00-26.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 26.00-27.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 27.00-28.00 sec 0.00 Bytes 0.00 bits/sec 16 2.79 KBytes [ 4] 28.00-29.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes [ 4] 29.00-30.00 sec 0.00 Bytes 0.00 bits/sec 20 2.79 KBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-30.00 sec 550 MBytes 154 Mbits/sec 702 sender [ 4] 0.00-30.00 sec 547 MBytes 153 Mbits/sec receiver iperf Done. server# ethtool -K eth0 lro off client% iperf3 -c ping.lrz.de -t 30 Connecting to host ping.lrz.de, port 5201 [ 4] local 2001:4ca0:0:f000:bf47:886b:a813:df4f port 60228 connected to 2001:4ca0:0:101::81bb:a11 port 5201 [ ID] Interval Transfer Bandwidth Retr Cwnd [ 4] 0.00-1.00 sec 112 MBytes 942 Mbits/sec 0 477 KBytes [ 4] 1.00-2.00 sec 110 MBytes 921 Mbits/sec 0 499 KBytes [ 4] 2.00-3.00 sec 110 MBytes 924 Mbits/sec 0 499 KBytes [ 4] 3.00-4.00 sec 109 MBytes 918 Mbits/sec 0 499 KBytes [ 4] 4.00-5.00 sec 110 MBytes 919 Mbits/sec 0 523 KBytes [ 4] 5.00-6.00 sec 110 MBytes 926 Mbits/sec 0 523 KBytes [ 4] 6.00-7.00 sec 110 MBytes 919 Mbits/sec 0 547 KBytes [ 4] 7.00-8.00 sec 110 MBytes 927 Mbits/sec 0 547 KBytes [ 4] 8.00-9.00 sec 110 MBytes 924 Mbits/sec 0 629 KBytes [ 4] 9.00-10.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 10.00-11.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 11.00-12.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 12.00-13.00 sec 109 MBytes 912 Mbits/sec 0 629 KBytes [ 4] 13.00-14.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 14.00-15.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 15.00-16.00 sec 110 MBytes 922 Mbits/sec 0 629 KBytes [ 4] 16.00-17.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 17.00-18.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 18.00-19.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 19.00-20.00 sec 110 MBytes 922 Mbits/sec 0 629 KBytes [ 4] 20.00-21.00 sec 110 MBytes 922 Mbits/sec 0 629 KBytes [ 4] 21.00-22.00 sec 109 MBytes 912 Mbits/sec 0 629 KBytes [ 4] 22.00-23.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 23.00-24.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 24.00-25.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 25.00-26.00 sec 110 MBytes 922 Mbits/sec 0 629 KBytes [ 4] 26.00-27.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 27.00-28.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 28.00-29.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes [ 4] 29.00-30.00 sec 110 MBytes 923 Mbits/sec 0 629 KBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-30.00 sec 3.22 GBytes 922 Mbits/sec 0 sender [ 4] 0.00-30.00 sec 3.22 GBytes 922 Mbits/sec receiver We see this in Debian Jessie 3.16.7-ckt20-1+deb8u3 1.2.0.0-k Debian Jessie+bpo 4.3.3-7~bpo8+1 1.4.2.0-k SLES11SP4 3.0.101-68-default 1.4.2.0-k SLES12SP1 3.12.53-60.30-default 1.4.2.0-k We do _not_ see this issue when using the "official" vmxnet3 kernel module from the VMware tools on SLES11SP4, which has the same version. This was the only way to get LROv6 working back then as well. SLES11SP4+VMW 3.0.101-68-default 1.4.2.0 Compiling our own vmxnet3 is not supported by VMware on Debian Jessie and SLES12SP1 anymore, you are supposed to use the in-kernel driver. The host runs ESXi 5.5.0 U3b (Build 3343343) with the latest firmware and bnx2x driver supported by HP for VMware. Disabling HW LRO on the host works as well (setting Net.Vmxnet3HwLRO=0) as described in https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2055140 You need to migrate the machine off the host and back on to activate it. I'm a bit unsure where to search here. It is definitely hardware/firmware and/or ESXi related. OTOH the VMware-Tools vmxnet3 driver seems to do something very different from the in-kernel driver. Best Regards, Bernhard
Powered by blists - more mailing lists