mirror of
https://github.com/ArduPilot/ardupilot.git
synced 2026-10-02 10:23:25 +08:00
Thomas Watson
b68348e5d4
Necessary to work around a bug in newlib. This applies to all hardfloat targets, including Cortex-M4. The jmp_buf is now 12 bytes larger, totaling 104 bytes. Info from newlib commit 15ad816dddf836def06cd0330ec0efa9ce50e5bf.
30 lines
1.1 KiB
C
30 lines
1.1 KiB
C
#include <AP_HAL/ap_setjmp.h>
|
|
|
|
#if AP_SETJMP_CORTEX_ENABLE
|
|
|
|
int __attribute__((naked, noinline)) ap_setjmp(ap_jmp_buf env) {
|
|
__asm__(
|
|
"mov ip, sp\n\t" // ip must be free on a function call; can't store sp
|
|
"stmia r0!, {r4-r10, fp, ip, lr}\n\t" // store 10 regular registers
|
|
"vstm r0, {d8-d15}\n\t" // store 16 FP registers. use 8 byte store for
|
|
// speed as our buffer is 8 byte aligned
|
|
"movs r0, #0\n\t" // return 0
|
|
"bx lr\n\t" // do return
|
|
);
|
|
}
|
|
|
|
void __attribute__((noreturn, naked, noinline)) ap_longjmp(ap_jmp_buf env, int val) {
|
|
__asm__(
|
|
"ldmia r0!, {r4-r10, fp, ip, lr}\n\t" // load 10 regular registers
|
|
"mov sp, ip\n\t" // restore stack pointer from scratch register
|
|
"vldm r0, {d8-d15}\n\t" // load 16 FP registers. use 8 byte load for
|
|
// speed as our buffer is 8 byte aligned
|
|
"movs r0, r1\n\t" // prepare return of value
|
|
"it eq\n\t" // but if value is zero
|
|
"moveq r0, #1\n\t" // return one
|
|
"bx lr\n\t" // do return
|
|
);
|
|
}
|
|
|
|
#endif
|