Mpc5200B network communication failure

[JoachimSchaefer]

Hi,

we have some problems with the Mpc5200B network communication.

an aplication which periodically sends huge data over ftp (64 MB Files)

still hangs every 10 - 30 times and the CPU seems to be dead.

If i read in this case some registers using a Jtag Debugger , the CPU Core (603)registers are readable, but all registers adressed over MBAR are showing a 0x0.

The Mbar Register self is set by the bootloader to 0x0000f000 and

this value is still set. our suspicion is it could be a problem with the XLB Bus.

The same Application on a mpc5200 still works!

I think also a silicon Bug is possible!?

Did anyone have an idea or tips to localize or solve the problem?

thanks

Joachim

[CarlHa]

Hi,

we have exactly the same symptoms, which are "communication failure", "TEA error" and as result a machine check exception.. At our application the XLB-Pipelining is always disabled. We also had test with enabled and disabled XLB bus snooping, but we didn’t have any success.

Can it be that there is a problem with the different bus clocks bus clocks (XLB and IPB)? Any synchronisation between this two clock domains?

Have you got any new cognitions to this problem?

Thanks

Carl

[Superberti]

Hi,

I don't know if this really helps but I'm using here a MPC5200B on a TQM5200S module with a linux 2.4.25 kernel. I'm transfering very large (gigabytes) amounts of TCP/IP-data and I achieve very good data rates with it (about 11.5 MBytes/s). I have never discovered any kind of problem in my tests so it may help to have a look into the linux kernel sources. You can download the latest 2.4 kernel sources at www.denx.de and maybe you find some special code/setup for the B revision in the BestComm or the network drivers.

Bye,

[toe5]

Hi Joachim,

we seem to have the same problem. On heavy load of the XLB-Bus (Bestcomm-Tasks and Core) the MPC5200B crashes due to a TEA-Error (Data Tenure Error) and throws an machine check exception.

Till the Data-Tenure-Timeout watchdog is activated every transmission on the XLB-Bus is replied by 0x00000000.

We still try to get rid of the problem with changing some configurations... but the only working way for us, was to disable the XLB-Pipelining during long and heavy loads of the XLB. Which is a worse work around due to the suffering performance.

We can not rule out that this is a silicon bug, because with the MPC5200 A-rev we never had that problem.

Sébastien