Flash写入错误

我试图通过覆盖一个程序指令来在DSSPICEV128GM102上为Bootloader编写一个修补程序。该指令目前是一个分支到无限循环和（0x3FFFF），而我只是试图将其更改为0x3FF8C。不知何故，总是有一个位是不正确的。当我告诉它写0x37 FF8C时，我读了它，新的值是0x33 FF8C。在尝试了第一个字节的许多不同的值之后，我发现如果我试图为命令字节写0x3f，我得到了想要的0x37，但是当我这样做时，跳转地址变为0xFF9C。例行程序在调试配置时工作正常，有人能帮助我吗？请参见下面的代码。*定义环地址（Uti32），0x2E6U定义FungPy值（unxFu8t）（0xFF8C）空隙FLASH PATCH55（空隙）{/*检查在PC位置0x02e6的Flash中的相对分支地址。*/Unt1616t SaveDeIpl；UTI32×t ADR= Loopl地址；StIsIa和SaveSePcPuiIPL（SaveDay-IPL，7）；TBLPAG＝0U；如果有一个循环指令，则假定这是1.1.55＊/yz f IFDEF *调试（（LVAL＝0xFFFF）和AMP；（HVAL＝0x00 FF））。T值*/Unt1616t偏移＝（Unt16-t）（（ADR）& 0xFFFFU）；UTI1616T页=（UTI1616T）（（ADR & GT；16U）和0xFFFFU）；/*设置指针到闪存位置*/NVMADRU=页；NVMADR＝偏移；/*设置TBLPAG指向数据锁存寄存器*/TBLPAG＝0xFA；/*设置NVM控制寄存器对于写操作*/NVMCON＝Studio Word；/*写入数据到锁存器//OxBuffTiNIN TBLWTL（0，0xFFFF）；γ-Bug TynIn TBLWTH（0，0xFFFF）；γ-Bug TynIn TBLWTL（2，JUMPPY值）；γ-Bug TynIn TBLWTH（2，0x00 37）；/*开始写操作*/Y-BuuthTiNoWrnWiMe（）；/*等待写入完成* /（NVMCONBIT.W）CUPYIPIL（SaveDayIpl）；}空隙FlldRead EdWord（UTIN 32×T ADR，UTIN 16LT LSW1，UTI1616T MSW1，UTI1616T LSW2，UIT1616 T MSW2）{UIT88T SaveDePiIPL；/*获取页面和偏移值*/Unt1616t偏移＝（UTI1616T）（（ADR）& 0xFFFFU）；UTI1616T页＝（UTIN 16t）（（ADR & GT；16U）和0；R=＝1U）{}}XFFFUU）；StIGH和SaveSub CPUIL IPL（SaveDay-IPL，7）；/*保存IRQ优先级和禁用*中断，用于Flash写入*//*安装指针到闪存位置*/NVMADRU=页面；NVMADR＝偏移；/*设置TBLPAG指向数据锁存寄存器*/TBLPAG＝0xFA；/*设置NVM控制寄存器用于WRITE操作*/NVMCON＝Studio Word；/*写入数据到锁存器//OxBuiTiNIN TBLWTL（0，LSW1）；γ-Bug TynIn TBLWTH（0，MSW1）；γ-Bug TynIn TBLWTL（2，LSW2）；γ-Bug TynIn TBLWTH（2，MSW2）；/*开始写操作*/Y-BuuthTiNoWrnWiMe（）；/*等待写入完成* /（NVMCONBITW.W=＝1U）{RESTE.CPUIIPL（SaveDayIpl）；/*恢复IRQ优先级级别*/}



      以下为原文

    I am trying to write a patch for a bootloader on a dsPIC33EV128GM102 by overwriting a single program instruction.  The instruction is currently a branch to an infinite loop and (0x37FFFF) and I am just trying to change it to 0x37FF8C.  Somehow there is always a single bit that is incorrect.  When I tell it to write 0x37FF8C I read it back and the new value is 0x33FF8C.  After trying a number of different values for the first byte I figured out that if I tried to write 0x3F for the command byte I got the desired 0x37, but when I did that the jump address changed to 0xFF9C.  To top it off the routine works as expected when in debug configuration.

Can someone please help me?  See the code below.

#define LOOP_ADDRESS (uint32_t)0x2E6u
#define JUMP_VALUE (int16_t)(0xFF8C)
void FLASH_patch55(void)
{
/* Check relative branch address at PC location 0x02E6 in flash.*/
uint16_t saved_ipl;
uint32_t adr = LOOP_ADDRESS;

SET_AND_SAVE_CPU_IPL(saved_ipl, 7);

TBLPAG = 0u;
int16_t lval = __builtin_tblrdl(adr);
int16_t hval = __builtin_tblrdh(adr);

/* If there is a loop instruction there assume this is version 1.1.55 */
#ifdef __DEBUG
if ((lval == 0xFFFF) && (hval == 0x00FF))
#else
if ((lval == 0xFFFF) && (hval == 0x0037u))
#endif /* __DEBUG */
{
/* Get page and offset values */
uint16_t offset = (uint16_t)((adr) & 0xFFFFu);
uint16_t page = (uint16_t)((adr >> 16u) & 0xFFFFu);

/* Setup pointers to flash memory location*/
NVMADRU = page;
NVMADR = offset;

/* Set TBLPAG to point to data latch registers */
TBLPAG = 0xFA;

/* Set NVM Control register for write operation */
NVMCON = PROGRAM_WORD;

/* Write data to latches */
__builtin_tblwtl(0, 0xFFFF);
__builtin_tblwth(0, 0xFFFF);
__builtin_tblwtl(2, JUMP_VALUE);
__builtin_tblwth(2, 0x0037);

/* Start write operation*/
__builtin_write_NVM();

/* Wait for write to complete */
while (NVMCONbits.WR == 1u)
{}

}
RESTORE_CPU_IPL(saved_ipl);
}
void FLASH_WriteDWord(uint32_t adr, uint16_t lsw1, uint16_t msw1, uint16_t lsw2, uint16_t msw2)
{
uint8_t saved_ipl;

/* Get page and offset values */
uint16_t offset = (uint16_t)((adr) & 0xFFFFu);
uint16_t page = (uint16_t)((adr >> 16u) & 0xFFFFu);

SET_AND_SAVE_CPU_IPL(saved_ipl, 7); /* Save irq priority level and disable
* interrupts for flash write */

/* Setup pointers to flash memory location*/
NVMADRU = page;
NVMADR = offset;

/* Set TBLPAG to point to data latch registers */
TBLPAG = 0xFA;

/* Set NVM Control register for write operation */
NVMCON = PROGRAM_WORD;

/* Write data to latches */
__builtin_tblwtl(0, lsw1);
__builtin_tblwth(0, msw1);
__builtin_tblwtl(2, lsw2);
__builtin_tblwth(2, msw2);

/* Start write operation*/
__builtin_write_NVM();

/* Wait for write to complete */
while (NVMCONbits.WR == 1u)
{}

RESTORE_CPU_IPL(saved_ipl); /* Restore irq priority level */
}