Slobodan Matic wrote: > Hi Jan, > >> How long did you run the ping test? Note that it "only" sends out one >> packet per second, while the RTmac TDMA discipline transmits one packet >> per cycle. So, if you are leaking rtskbs, the latter will stumble much >> earlier. >> > > I ran rtping tests that lasted more than 15min, i.e., about 1000 packets. No
You could also run rtping at a higher rate (-i ...), something that is equivalent to the TDMA cycle. > error and no packet loss occured, and I did that from both machines. > The only strange thing I once noticed on one machine is that rtt times had > negative values, like -89us, -85us, -96us... I guess this does not have to > do with my main problem (tdma rtmac) because typically with rtping I would > have +70-100us range of values. > > >> Did I ask this already? Check /proc/interrupts if the NIC on the master >> is assigned to Linux (in order to detect potential IRQ conflicts), also >> check /proc/rtai/<don't-know-its-name: RT IRQs> if there is progress >> /wrt handled NIC IRQs when RTnet is running. >> > > Well, I think there are no IRQ conflicts because my card is at IRQ #6 and > after I do './rtnet start' I have the folowing: OK. > > ##### /proc/interrupts: > CPU0 CPU1 > 0: 2398005 0 XT-PIC-XT timer > 1: 3007 0 XT-PIC-XT i8042 > 2: 0 0 XT-PIC-XT cascade > 8: 0 0 XT-PIC-XT rtc > 10: 91059 0 XT-PIC-XT yenta, ahci, [EMAIL PROTECTED] > :0000:00:02.0 > 11: 146 0 XT-PIC-XT sdhci:slot0, HDA Intel > 12: 41822 0 XT-PIC-XT i8042 > NMI: 0 0 > LOC: 2397695 2397606 > ERR: 0 Mmh, SMP... Could you give this also a try with maxcpus=1 (or without CONFIG_SMP)? Just to check if we see some SMP-related race. > > ##### /proc/rtai/hal > ** RTAI/x86: > APIC Frequency: 12468000 > APIC Latency: 3944 ns > APIC Setup: 1000 ns > ** Real-time IRQs used by RTAI: > #6 at ffffffff8830d1ff > #325 at ffffffff882e13ec > #328 at ffffffff882e0d0b > ** RTAI extension traps: > SYSREQ=0xf6 > ** RTAI SYSREQs in use: #1 #2 > > I was not able to find anything that would give me the IRQ count in order to > see the progress. I know /proc/xenomai/irq works under xenomai, but I do not > know the rtai equivalent. Then some tracing will be required next, in order to gain overview what happens around the out-of-rtskb situation. What kernel are you on, what hal (==ipipe) version? > > Some other output that might be of interest: > > ##### /proc/rtai/rtnet/* > Index Name Flags > 1 rteth0 UP BROADCAST > 2 rtlo UP LOOPBACK > Statistics Current Maximum > rtskb pools 6 6 > rtskbs 184 184 > rtskb memory need 329728 329728 > RTnet 0.9.10 - built on Nov 3 2008 10:28:22 > RTcap: yes > rtnetproxy: no > bug checks: no > > ##### /proc/rtai/rtnet/ipv4/host_route > Hash Destination HW Address Device > 00 0.0.0.0 00:00:00:00:00:00 rtlo > 01 127.0.0.1 00:00:00:00:00:00 rtlo > 02 10.0.0.2 00:00:00:00:00:00 rtlo > 3F 10.0.0.255 FF:FF:FF:FF:FF:FF rteth0 > > ##### /proc/rtai/rtnet/host_route (SLAVE) > Hash Destination HW Address Device > 00 0.0.0.0 00:00:00:00:00:00 rtlo > 01 10.0.0.1 00:00:00:00:00:00 rtlo > 01 127.0.0.1 00:00:00:00:00:00 rtlo > 02 10.0.0.2 00:16:D3:3E:8C:B5 rteth0 > 3F 10.255.255.255 FF:FF:FF:FF:FF:FF rteth0 > > So, it seems that the error occurs after the slave has the correct master > MAC, but before the master have the correct slave MAC address. > > That hang may either be the reason or a result of the actual problem, >> still don't know. Can you check with a different NIC? Even a different >> e1000 revision may be interesting. What happens if master and slave swap >> roles? >> > > Yeah, I also have no idea whether the hang is a cause or a consequence of > the TDMA failure. > I am working with a laptop, so I can't check with other NIC. What e1000 > revision you exactly had in mind? > Finally, when master and slave swap roles the outcome remains the same, > i.e., I again get the TDMA failure and hang (on the master). The NICs are > different though (Intel 82566M and 80003ES2LAN gigabit controllers). OK, this reduces the probability of a hardware quirk significantly. Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
