Leopold Palomo-Avellaneda wrote: > A Dilluns 26 Març 2007 18:15, Jan Kiszka va escriure: >> Leopold Palomo-Avellaneda wrote: >>> Hi, >>> >>> after follow the recomendations of Mr. Kiszka I have configured my system >>> without tdma. >>> >>> Adding the other node in the net with rtroute solicit I have found that >>> doing a rtping between a rtai+rtnet with a VxWorks box with a cross cable >>> an unbelieved numbers: >>> >>> >>> # ./rtping -s 1200 192.168.1.4 >>> Real-time PING 192.168.1.4 1200(1228) bytes of data. >>> 1208 bytes from 192.168.1.4: icmp_seq=1 time=0.0 us >>> 1208 bytes from 192.168.1.4: icmp_seq=2 time=0.0 us >>> >>> >>> this is possible? >> Nope, it's too good to be true. > > Sure !!!!! >> When using RTnet with RTAI, you have to >> keep in mind that RTAI's system timer requires manual start-up, either >> in your main application (rt_timer_start or so) or via one RTnet module >> that needs it too (tdma.ko or rtcfg.ko - but you don't need any of them >> for your scenario). > > Yes, ok I have load the rtcfng module. Now the numbers are a bit different. > In > a loopback mode I have this: > > # ./rtping -s 1200 192.168.1.3 > Real-time PING 192.168.1.3 1200(1228) bytes of data. > 1208 bytes from 127.0.0.1: icmp_seq=1 time=31.9 us > 1208 bytes from 127.0.0.1: icmp_seq=2 time=22.7 us > 1208 bytes from 127.0.0.1: icmp_seq=3 time=19.8 us > 1208 bytes from 127.0.0.1: icmp_seq=4 time=19.3 us > 1208 bytes from 127.0.0.1: icmp_seq=5 time=20.0 us > > --- 192.168.1.3 rtping statistics --- > 5 packets transmitted, 5 received, 0% packet loss > worst case rtt = 31.9 us > > and with the VxWorks box: > > ulises:/usr/local/rtnet/sbin# ./rtping -s 1200 192.168.1.4 > Real-time PING 192.168.1.4 1200(1228) bytes of data. > 1208 bytes from 192.168.1.4: icmp_seq=1 time=1509.9 us > 1208 bytes from 192.168.1.4: icmp_seq=2 time=690.3 us > 1208 bytes from 192.168.1.4: icmp_seq=3 time=672.3 us > 1208 bytes from 192.168.1.4: icmp_seq=4 time=665.4 us > 1208 bytes from 192.168.1.4: icmp_seq=5 time=666.8 us > 1208 bytes from 192.168.1.4: icmp_seq=6 time=672.9 us > 1208 bytes from 192.168.1.4: icmp_seq=7 time=680.8 us > 1208 bytes from 192.168.1.4: icmp_seq=8 time=661.2 us > 1208 bytes from 192.168.1.4: icmp_seq=9 time=674.0 us > > --- 192.168.1.4 rtping statistics --- > 9 packets transmitted, 9 received, 0% packet loss > worst case rtt = 1509.9 us > > it's curious because the number are not very different from a normal net. > Also, I have a peak of ~1ms every 16 iteration .. > > Could I have a better numbers or are so good?
Well, you are also measuring a black box - VxWorks. I don't know how that OS internally realised its networking stack and what priority it assigns to ICMP (ping) packets. You should have a look at the network traffic first, ideally by sniffing with a third node that is sharing the net via a *hub*, or by listening on the RTnet node via RTcap (see README.rtcap). You may also do your measurements via UDP, using the protocol you actually want to apply on the control job. Maybe UDP shows a better latency in VxWorks. > > > Regards, > > Leo > > Pd looking on the dmesg I found this message, it's important? > > RTnet: initialising real-time networking > PCI: Found IRQ 11 for device 0000:00:0e.0 > rteth0: 0000:00:0e.0, 00:A0:C9:1D:52:B9, IRQ 11. > *** Receiver lock-up bug detected *** > Your device may not work reliably! > RTnet: registered rteth0 > initializing loopback... Take it as a safety note: Your Intel PRO/100 NIC /may/ lock-up in its receiver state machine. Normal NIC drivers have a timer that can recover from this, but for RT, this is generally already far too to late, deadlines have been missed then. I don't know the concrete boundary conditions for this hardware bug, but I also failed to watch it with our hardware in real life over the last 5 years or so. Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
