Linker script .relocate section's first symbol, _srelocate, is incorrect (GCC Bug?)

Question

Problem My problem is that when I use the following script which is intended to place the Code into RAM the relocate section gets filled with bogus data. My question is: Why is the `_srelocate` symbol 4 bytes greater than the `_etext` symbol? Shouldn't they be the same? Also, if the answer to 1. is NO, shouldn't I be copying from `_etext + 4` to `_srelocate`? Background and relevant code I'm working with an Atmel ATSAM3N4X series processor (an ARM Cortex M3) and would like a little help with my linker script and `.relocate` section initialization. The reason for this is that the `_etext` symbol is 4 bytes less than the `_srelocate` symbol. The following linker script is the default script generated by Atmel Studio 6 (if you want the `ram`/`rom` symbol locations see the Apendix to the question). ``` /* ---------------------------------------------------------------------------- * SAM Software Package License * ----------------------------------------------------------------------------…

Accepted Answer

I had long ago discovered the answer to my own question but it seems that it is still a little unclear in my edits so allow me to formalize it here. The issue is that due to a bug in GCC the linker inserts an additional TMC_END symbol taking up 4 bytes in between my `_etext` and `_srelocate` symbols. This is incorrect because this symbol has nothing to do with the code that I'm writing and I'm not even using this functionality. I had discovered and explained this in my other SO question. Workaround: Atmel's sample code, rightly, assumes that if your code is running out of Flash memory your `_srelocate` symbol's address and `_etext` symbol's address will be different, one having an address starting with `0x004...` and the other with `0x20...`. If your code is running out of RAM, this means that you have a bootloader that copies the relevant sections into the correct memory location, hence the `_etext` symbol's address will also reside in RAM and be equal to the `_srelocate` symbol's address. Therefore Atmel is trying to make sure that the `_srelocate` section always resides in RAM. Due to the bug in GCC the solution was to change the code to check if the `_srelocate` symbol's address is already in RAM, like this: ``` // I can't remember what #define Atmel has for SRAM so I'm just going to // use this one for the purposes of my example. #define SRAM_START_ADDRESS 0x20000000 void Reset_Handler(void) { uint32_t *pSrc, *pDest, Size; /* Initialize the relocate segment */ pSrc = &_etext; // Could either be in Flash or SRAM depending on your configuration. pDest = &_srelocate; // Always in SRAM or there's something strange going on... if ((intptr_t)pSrc < (intptr_t)SRAM_START_ADDRESS) { // <<<< Changed code here <<<< // We enter here only if pSrc is pointing to a location in Flash. // If that's the case, we need to copy the memory from Flash to SRAM. for (; pDest < &_erelocate;) { *pDest++ = *pSrc++; } } /* Clear the zero segment */ for (pDest = &_szero; pDest < &_ezero;) { *pDest++ = 0; } /* Set the vector table base address */ pSrc = (uint32_t *) & _sfixed; SCB->VTOR = ((uint32_t) pSrc & SCB_VTOR_TBLOFF_Msk); if (((uint32_t) pSrc >= IRAM_ADDR) && ((uint32_t) pSrc < IRAM_ADDR + IRAM_SIZE)) { SCB->VTOR |= (1UL) << SCB_VTOR_TBLBASE_Pos; } /* Initialize the C library */ __libc_init_array(); /* Branch to main function */ main(); /* Infinite loop */ while (1); } ``` And that worked for my situation. From what I can see this bug was fixed and the fix has since been reverted so it is unlikely that you will have it fixed in your Atmel Studio.

Linker script .relocate section's first symbol, _srelocate, is incorrect (GCC Bug?)

Solution

Problem

Related problems