NDK Telnet Stack shutting down on its own

Tom Farmer

Other Parts Discussed in Thread: TM4C1294KCPDT, SYSBIOS

I have a proprietary board with a TM4C1294KCPDT processor. I am using ndk_2_24_03_35 with tirtos_tivac_2_14_00_10. I am using XCONF to set up a Telnet server in the TIVA chip. The ethernet connection is 10BASET, DHCP. It is communicating to a proprietary PC Windows application that works correctly with an earlier Stellaris implementation that used lwip.

The board seems to boot, acquire its IP address, and establish a Telnet connection to the application. Everything works fine until the following (as seen from wireshark):

1. PC sends a command via a Telnet packet.

2. TIVA returns a TCP ACK packet.

3. TIVA returns a response to the command via a Telnet packet.

4. The PC does not ACK the response immediately. Aprox 102 milliseconds after the original response packet was sent, the TIVA returns the same response via a TCP retransmission packet.

5. The PC sends a TCP ACK packet.

6. The TIVA exits the NetScheduler function in netctrl.c (line 207). {Don't think it should do this}

7. From then on things fall apart. If I remove the breakpoint I placed immediately after line 207 in netctrl.c, the TIVA almost always sends a TCP ACK+FIN packet.

8. The PC sends a TCP ACK packet.

9. The TIVA eventually gets to the ti_ndk_config_Global_NetworkIPAddr function in app_pem4f.c (line 3724) and executes the xdc_runtime_System_printf("Network Removed: "); path.

10. Ultimately the entire TCP/IP stack stops executing and will not even respond to a ping.

The console usually displays that a TCP retransmission occurred immediately before the Network Removed... stuff.

Suggestions on how to diagnose what is happening?

Tom

over 9 years ago

0 ToddMullanix over 9 years ago

TI__Guru* 96960 points

Can you please attach a simple CCS project that demonstrates your issue and instructions on how to reproduce it? I'm assuming you just need a telnet client on the host side.

You can also check the following in CCS, Tools->ROV->BIOS->Scan for Errors. This will show you any blown stacks or errors in the kernel.

Todd

0 Tom Farmer over 9 years ago in reply to ToddMullanix

Intellectual 765 points

Hi Todd,

I have been able to dig back into things and found the following:

NC_NetStop in netctrl.c (line 258) is being called with an argument of (-1) by DbgPrintf in ossys.c (line 59):

if(Level >= DBG_ABORT_LEVEL) {
NC_NetStop(-1);

DbgPrintf in ossy.c is being called with arguments (Level = 1, TcpTimeoutRexmt:.Retransmit.Timeout) by TcpTimeoutRexmt in tcptime.c (line 164)

I have not been able to dig back into where DBG_ABORT_LEVEL is ultimately defined, but it is declared in osif.h.

Is there some kind of a setting that is not set properly in the app.cfg file, project properties, or elsewhere?

-------------------------------------------------------------------------------------------------------------------------------------------------------------------
To respond to the points you made:

The only error that shows up in ROV is in the ti.sysbios.family.arm.lm4.Timer module:
Caught exception in view init code: "C:/ti/xdctools_3_31_01_33_core/packages/xdc/rov/StructureDecoder.xs", line 547: java.lang.Exception: Target memory read failed at address: 0x40036000, length: 80

I suspect this is not correlated with the problem above because ROV reports the same error at seven address following a system reset.

Providing a CCS project would be difficult. I did not have this problem when using the NDK console.c module on the target and the command line telnet utility on windows, so I suspect it is an interaction issue between the PC code, some of my application's tasks, and the NDK.
----------------------------------------------------------------------------------------------------------------------------------------------------------------

Tom

Processors

Processors forum

NDK Telnet Stack shutting down on its own