Exception Analysis on CC2650

Hi Manuel,
Can you use ROV to verify that the Task stacks have not overrun?
Thanks,
Janet

Hi Janet,

this is what ROV shows for the task when the exception occurred:

I increased the stack size before to see if it solves the issue (that's why it's that large), but it still remains.

Best regards,

Manuel

Hi Manuel,
It looks like the task with the large stack is preempted, so I wonder what was actually running. From the exception dump, it appears that the bus fault is caused by an access to the address 0xcc007f65, which doesn't look like a valid address. Can you look at the addresses LR to see what was calling the code that caused the exception? And PC is always pointing to an address related to Hwi_construct? You could try looking at the disassembly code near where the exception occurred and set a breakpoint.
Best regards,
Janet

Hi,

I checked the address in LR after the exception, I'm not exactly sure where it leads to, the function is labelled with $C$L607. Is there a way to see all the commands that call this function?

Using another CC2650 and SmartRF06 doesn't solve the problem.

I went through the code in disassembly to find the functions that cause the interrupt. I had a closer look at the sin() function as this is where I always get the exception. This is the code:

106         float x = sin(28);

000053e2:   A032               adr       r0, #0xc8

000053e4:   C803               ldm       r0, {r0, r1}

000053e6:   F002FE0F           bl         #0x8008

000053ea:   F001FA20           bl         #0x682e

000053ee:   9044               str       r0, [sp, #0x110]

The first bl leads to:

         $Tramp$TT$L$PI$$ti_sysbios_family_arm_m3_Hwi_construct():

00008000:   F8DFF000           ldr.w     pc, [pc, #0]

00008004:   AF29               add       r7, sp, #0xa4

...

But it doesn't lead to the start of the function, is it intended to be this way?

Is the HWI used for function calls or is this already the position where the exception is caused?

The other sin() functions in my code also have this bl command to the same position.

Going further from this point somehow leads to this without returning to the sin() function:

...

1001bbd4:   6812               ldr       r2, [r2]

1001bbd6:   2A00               cmp       r2, #0

1001bbd8:   D0FD               beq       #0x1001bbd6

1001bbda:   4790               blx       r2

R2 at that point:

R2           0x000062E1        General Purpose Register 2 [Core]         …

 Which leads to the exception handler:

         ti_sysbios_family_arm_m3_Hwi_excHandlerMin__I():

000062e0:   B530               push       {r4, r5, lr}

…

If the sin() is replaced with sinf(), there is no call of the  HWI constructor:

float x = sinf(28);

0000642e:   4830               ldr       r0, [pc, #0xc0]

00006430:   F7FAFFB6            bl         #0x13a0

00006434:   9044               str       r0, [sp, #0x110]

...

 Here bl leads to:

         sinf():

000013a0:   49C3               ldr       r1, [pc, #0x30c]

000013a2:   E92D4FF0           push.w     {r4, r5, r6, r7, r8, r9, r10, r11, lr}

...

Should the sin() functions contain a call of the HWI constructor function? If no, how can they be removed? And if yes, what's their purpose? Do they cause the exception?

Edit: Some more information: I made some changes to the .cfg file (mostly stack/heap sizes) and the project properties (compiler optimization, there could be some more changes that I don't remember) before the problem occurred. Could any of this have caused this problem? Is creating a new project resetting the .cfg file completely? When doing this, some of the stack/heap values still remain.

Best regards and thanks for your help,

Manuel

Hi Manuel,
The sin() function should not branch to the Hwi constructor. Do you have a simple test case project you could send so we can try to debug this?
Thanks,
Janet

Hi Janet,

I observed that the error only occurs when most of my software components are copied to the project directory. If there is only the component in which the exception is caused is in the project folder, it works fine, no exception. However, if I copy more components at some point it will start to crash. The files don't even have to be included in the code for this to happen, they just need to be in the project folder in the workspace. I can't make sense of this.

I'm not allowed to upload most parts of the code so sadly I can't provide a test case project. Is there any other way we can find the cause?

Best regards,

Manuel

Hi Manuel,
Can you tell us what compiler/linker your are using (TI, GCC, or IAR), and what version? Also, what are the compiler and linker options you are using?
Thanks,
Janet

Hi Janet,

I'm using the TI v15.12.1.LTS compiler.

Compiler options are:

-mv7M3 --code_state=16 -me -Ooff --opt_for_speed=0 --include_path="C:/ti/tirtos_simplelink_2_13_00_06/products/cc26xxware_2_21_01_15600" --include_path="C:/Users/<username>/workspace_v6_1_3/<project name>" --include_path="C:/ti/ccsv6/tools/compiler/ti-cgt-arm_15.12.1.LTS/include" -g --gcc --diag_wrap=off --diag_warning=225 --diag_warning=255 --display_error_number --gen_func_subsections=on --abi=eabi

Linker options:

-mv7M3 --code_state=16 -me -Ooff --opt_for_speed=0 -g --gcc --diag_wrap=off --diag_warning=225 --diag_warning=255 --display_error_number --gen_func_subsections=on --abi=eabi -z -m"<project name>.map" --stack_size=256 --heap_size=0 -i"C:/ti/ccsv6/tools/compiler/ti-cgt-arm_15.12.1.LTS/lib" -i"C:/ti/ccsv6/tools/compiler/ti-cgt-arm_15.12.1.LTS/include" --reread_libs --display_error_number --warn_sections --diag_wrap=off --xml_link_info="<project name>_linkInfo.xml" --rom_model

Thanks and best regards,

Manuel

Hi Manuel,

We need to verify that a trampoline is being used in your call to sin().  Can you check that there is something similar to the following in your .map file:

FAR CALL TRAMPOLINES

 

callee name               trampoline name

   callee addr  tramp addr   call addr  call info

--------------  -----------  ---------  ----------------

sin                       $Tramp$TT$L$PI$$sin

   00000725     f0000018     f0000002   try1.obj (.text)

Thanks,

Janet

Hi Janet,

I checked the map files of several projects, some with sin(), some only with sinf(), but none of those contain any entries for sin() or sinf() in the trampoline section (not even the working ones with sin() functions).

Best regards,

Manuel

Hi Manuel,

Can you post a screen shot of the disassembly window in CCS showing the address in the first branch (0x8000, if 0x8008 is still the address)?  If this is a trampoline, you should be able to figure out where it's supposed to go by piecing together the two following words.  For example, if you look at the trampoline for Task_sleep() below, combine the two highlighted words to get 1001A96C, which in my program is the address of Task_sleep().

Also, what is at the address of the the second branch (0x682e).  I'm guessing it is something like __eabi_d2f.

So far, I haven't been able to reproduce your failure with calls to sin() and sinf().  Are these functions from rtsv7M3_T_le_eabi.lib, or do you have your own?

Thanks,

Janet

Hi Janet,

thanks for asking about which sin() functions I'm using, I guess I found the cause. It was a pretty stupid mistake I totally forgot about: before the error occurred,  I tried implementing my own sine function, and for that I created a lookup table called sin[ ] (without thinking about the function sin()). Now that I went back to trying to access the sin() function in math.h, the array caused an interference as soon as the containing file was copied into the project folder. Now that I renamed the array, the code works fine.

Sorry for making work, and thank you very much for helping me find the cause of the problem!

Best regards,

Manuel