Hello!
I'm facing a problem with out code crashing. I'm using TMS320F28335 with DSP/BIOS 5.41.13.42 on a custom board and CCS 5.1.1.00031.
We have connected an SPI eeprom to the DSP along with other SPI devices. The SPI bus has its own task handling the I/O. Other code puts requests in a queue which the SPI task reads and performs the requests. The completion of a read is indicated by a semaphore.
The crashes occur when a lot of data is cleared in the eeprom (lots of SPI activity). The command to clear the eeprom is given trough a custom debugger interface which make heavy use of SYS_sprintf.
When the DSP crashes it goes to the illegal instruction trap. It seems to have to do with the semaphores. I put a breakpoint in the illegal instruction trap at took a look at the RPC register to get a idea what was the last thing that was called.
The RPC contained the address 0x3159E9, which is in the SEM_postEnterKnl routine. The instruction at 0x3159E9 is a "LCR 0x318bbe". Here comes the interesting part. That address does not contain any data (0xFFFF) which is the ITRAP1 instruction. Thus the jump to the illegal instruction trap.
Is this a bug in the DSP/BIOS or what could it be?
Relating to this problem we have situations where the watchdog doesn't seem to work. In case of a illegal trap, the ISR function should do a shutdown and then wait for a watchdog reset. The watchdog reset didn't always work. That is to say the device hang for 5-10minutes or forever and a power cycle was needed to reset it. I tried to check with the debugger what was going on, but it coudn't get contact with the 28335, or said that the execution was at 0x0.
