anchao opened a new pull request, #6056:
URL: https://github.com/apache/incubator-nuttx/pull/6056
## Summary
arch/stack_color: correct the end address of stack color
The different optimization of compilers will cause ambiguity in
obtaining sp through up_getsp() in *_stack_color(), if compile
with clang and enable the optimization flag (-Ofast), up_getsp()
call will be earlier than push {r0-r9,lr}, the end address of color
stack will overlap with saved registers.
Compile line:
```
clang --target=arm-none-eabi -c "-Ofast" -fno-builtin
-march=armv8.1-m.main+mve.fp+fp.dp \
-mtune=cortex-m55 -mthumb -mfpu=fpv5-d16 -mfloat-abi=hard -D__NuttX__
-common/arm_checkstack.c -o arm_checkstack.o
```
Assembler code:
`llvm-objdump -aS arm_checkstack.o`
```
------------------------------------
|00000000 <arm_stack_color>:
|; start = INT32_ALIGN_UP((uintptr_t)stackbase);
| 0: c2 1c adds r2, r0, #3
| 2: 22 f0 03 02 bic r2, r2, #3
|; end = nbytes ? INT32_ALIGN_DOWN((uintptr_t)stackbase + nbytes) :
| 6: 19 b1 cbz r1, 0x10 <arm_stack_color+0x10> @ imm = #6
| 8: 08 44 add r0, r1
| a: 20 f0 03 00 bic r0, r0, #3
| e: 00 e0 b 0x12 <arm_stack_color+0x12> @ imm = #0
|; __asm__
| 10: 68 46 mov r0, sp <--- fetch
the sp before push {r7 lr}
| 12: 80 b5 push {r7, lr} <--- sp
changed
|; nwords = (end - start) >> 2;
| 14: 80 1a subs r0, r0, r2
| 16: 80 08 lsrs r0, r0, #2
|; }
| 18: 08 bf it eq
| 1a: 80 bd popeq {r7, pc}
| 1c: 4b f6 ef 63 movw r3, #48879
| 20: cd f6 ad 63 movt r3, #57005
| 24: a0 ee 10 3b vdup.32 q0, r3
|; while (nwords-- > 0)
| 28: 20 f0 01 e0 dlstp.32 lr, r0
|; *ptr++ = STACK_COLOR; <---
overwrite
| 2c: a2 ec 04 1f vstrw.32 q0, [r2], #16
| 30: 1f f0 05 c0 letp lr, 0x2c <arm_stack_color+0x2c> @ imm = #-8
|; }
| 34: 80 bd pop {r7, pc}
------------------------------------
```
Signed-off-by: chao.an <[email protected]>
## Impact
colorize of running stack
## Testing
Cortex-m55, Clang + -Ofast
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
