Hi Berenger, Thanks for the report!
There are two issues here. First of all, echo_server/client apps are testing apps that can misbehave is not properly configured. I just tried master with 4 workers for both and with the following configuration: server: test echo server fifo-size 4096 uri udp://6.0.1.1/1234 no-echo client: test echo client fifo-size 4096 test-timeout 100 no-return mbytes 10000 uri udp://6.0.1.1/1234 <udp://6.0.1.1/1234> DBGvpp# test echo client fifo-size 4096 test-timeout 100 no-return mbytes 10000 uri udp://6.0.1.1/1234 1 three-way handshakes in 0.00 seconds 1398.78/s Test started at 4.588836 Test finished at 12.819988 10485760000 bytes (10000 mbytes, 9 gbytes) in 8.23 seconds 1273911466.12 bytes/second half-duplex 10.1913 gbit/second half-duplex Note that this is with a debug image, release images should be able to do 40Gbps or more but with tx drops. I suspect you've used the server/client commands without no-echo/no-return options and you’ve hit an assert in the echo-server put in place for tcp. TCP pins connections to a worker, udp does not anymore. The reason why I only run with no-echo/no-return is because udp is not reliable and the echo apps, as programmed now, fail on any packet drop. The echo apps could be improved to avoid this kind of surprises but at this time this is not very high on my list. Still, in light of your report, I’ve updated the wiki page here [1] so hopefully this can be avoided in the future. The other issue is IPSec + udp. Unfortunately, I’ve never run this combination. However, from your report, it would seem that packets do hit ipX-local, in which case a checksum error should be properly reported, as opposed to “unknown ip protocol”. Could it be that your udp server is not turned on? For more clues, I would run vpp from gdb and take a closer look at what happens in ipX_local_inline. To check if your udp server is running: sh app server Regards, Florin [1] https://wiki.fd.io/view/VPP/HostStack/EchoClientServer <https://wiki.fd.io/view/VPP/HostStack/EchoClientServer> > On Jul 13, 2018, at 1:04 AM, Foucher Berenger > <berenger.fouc...@stagiaires.ssi.gouv.fr> wrote: > > Hi, > > I recently tried to use the UDP echo server that is included in VPP. It seems > to work fine when used by itself, but when I tried to use it on an IPsec > interface (and my configuration looks good because ICMP is working with > IPSec), an “unknown ip protocol” error occurs when the packet is decapsulated > (I can see it with show errors). Packet is dropped by the receiver. In my > opinion a checksum error could occur after decapsulating the packet. > > Plus I get the following error when I try to use it WITHOUT IPsec on multiple > cores. > > 1: > /home/bfoucher/vpp_up_to_date/vpp/build-data/../src/vnet/session-apps/echo_server.c:153 > (echo_server_rx_callback) assertion `s→thread_index == thread_index' fails > > These issues are quite annoying when it comes to benchmarking VPP. Do you > have any suggestion in order to make these features work together? > > Thanks, > > Berenger > > > -=-=-=-=-=-=-=-=-=-=-=- > Links: You receive all messages sent to this group. > > View/Reply Online (#9832): https://lists.fd.io/g/vpp-dev/message/9832 > <https://lists.fd.io/g/vpp-dev/message/9832> > Mute This Topic: https://lists.fd.io/mt/23376157/675152 > <https://lists.fd.io/mt/23376157/675152> > Group Owner: vpp-dev+ow...@lists.fd.io <mailto:vpp-dev+ow...@lists.fd.io> > Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub > <https://lists.fd.io/g/vpp-dev/unsub> [fcoras.li...@gmail.com > <mailto:fcoras.li...@gmail.com>] > -=-=-=-=-=-=-=-=-=-=-=-
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#9843): https://lists.fd.io/g/vpp-dev/message/9843 Mute This Topic: https://lists.fd.io/mt/23376157/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
