上篇文章介绍了Android镜像的构成。我们来看看lk怎么将这个mtd.img大镜像进行解析和烧写到NandFlash的分区中的。我们的环境依然是基于SEP6200开发板的。首先我们来看看镜像在nandflash中的存放规则:
Dev | 地址空间 | Size | blocksize | Name | |
(预留) | 0x00000000 ~ 0x001FFFFF | 00200000 | 2MB | 00200000 | Lk |
Mtd0 | 0x00200000 ~ 0x003FFFFF | 00200000 | 2MB | 00200000 | Misc |
Mtd1 | 0x00400000 ~0x00DFFFFF | 00a00000 | 10MB | 00200000 | Recovery |
Mtd2 | 0x00E00000~ 0x018FFFFF | 00a00000 | 10MB | 00200000 | Boot |
Mtd3 | 0x01900000~ 0x338FFFFF | 32000000 | 800MB | 00200000 | System |
Mtd4 | 0x33900000~ 0x39CFFFFF | 06400000 | 100MB | 00200000 | Cache |
Mtd5 | 0x39D00000~ 0x411FFFFF | 06400000 | 100MB | 00200000 | Userdata |
Mtd6 | 0x41200000~ 0xC11FFFFF | 80000000 | 3GB | 00200000 | FTL |
解析这个文件的源码位于lk-release\platform\sep6200\cmd_mtd.c
首先定义了结构体,是我们需要解析文件的头部。 cmd_mtd.c就是对mtd.img进行解析的,而写入的所有的信息都已经存放在了每个文件的头部中,即stBunchHeader。而对所有的写入操作都是由一个函数完成的nandwrite函数
for(i=0; i<img_cnt; i++)
{
stBunchHeader = (tagDiskImageBunchHeaderType *)ptr;//每个镜像的头部
img_part = stBunchHeader->ullPart;
img_size = stBunchHeader->ullLength;
printf("|*************************************************|\n");
printf(" >img[%d]\n", i);
printf(" > ullPart = %d\n", img_part);
printf(" > ullLength = 0x%x\n", img_size);
ptr += 16; //sizeof(stBunchHeader);//跳过头部信息
nandwrite((u_char *)ptr, mtd_part[img_part], img_size);//写入
printf(" >Write img[%d] to nand OK!\n", i);
ptr += img_size;
}
那么,这个nandwrite函数究竟做了哪些事情呢?跟踪到nand_write_yaffs.c文件中发现其根据不同的文件类型采用了不同的写入方式,对于有文件系统的文件,如system.img是采用了yaffs文件系统,那么写入的时候就要以文件系统的方式写入到NandFlash中的。我们来看看nandwrite函数的实现:
int nandwrite(u_char *buf, const char *name, unsigned int len)
{
struct ptentry *ptn;
unsigned int i, found;
found = 0;
for(i = 0; i < sizeof(board_part_list); i++)
{
ptn = &board_part_list[i];
if (!strcmp(ptn->name, name))
{
found = 1;
break;
}
}
if(found == 0)
{
printf("error! partition not found!\n");
return -1;
}
switch(ptn->flags){
case(WRITE_LOADER):
printf("Write %s 0x%0x bytes from 0x%0x, MODE:WRITE_LOADER\n", ptn->name, len, buf);
burn_nandboot(buf);
return 0;
case(WRITE_RAW):
printf("Write %s 0x%0x bytes from 0x%0x, MODE:WRITE_RAW\n", ptn->name, len, buf);
write_nand(buf, ptn->start, len);
return 0;
case(WRITE_FS):
printf("Write %s 0x%0x bytes from 0x%0x, MODE:WRITE_FS\n", ptn->name, len, buf);
/*当涉及到文件系统的时候,我们需要擦除所有的相关块*/
write_nand_yaffs(buf, ptn->start, len, ptn->length);
return 0;
default:
printf("Error! Unknown write type for %s\n", ptn->name);
return -1;
}
}
board_part_list定义了我们文章开始的时候所列的表,即每个镜像在NandFlash中的存放位置。
这里我们知道了,Nand的写入是根据每个镜像的标签来识别的,WRITE_LOADER类型则是烧写bootloader的函数去完成;WRITE_RAW则采用直接烧写的方式;WRITE_FS则是烧写yaffs文件系统。
现在大概知道了lk的对于mtd.img的解析过程和烧写
过程。但是仔细跟代码发现lk对NandFlash的操作和u-boot中的方式并不一样。所以要么是将lk的nand驱动移植到u-boot中,要么是采用u-boot的方式去实现lk的烧写功能。权衡之后采用了后者,因为:如果修改u-boot的nand驱动则对其他的命令会造成不兼容。
下面为u-boot的mtd解析命令的移植过程:
1.在u-boot/common/目录下添加cmd_mtd.c文件,尤其注意添加u-boot的命令格式:
U_BOOT_CMD(
nandmtd, 2, 1, do_nandmtd,
"nandmtd [addr]",
"Resolve xxx_mtd.img at ddr address and write [boot.img system.img recovery.img] to nand."
);
int nand_erase_block(loff_t off, unsigned int len)
{ nand_erase_options_t opts;
nand_info_t *nand;
nand = &nand_info[0];
memset(&opts, 0, sizeof(opts));
opts.offset = off;
opts.length = len;
nand_erase_opts(nand, &opts);
return 0;
}
int write_nand(u_char *buf, unsigned start_blk, unsigned int len)
{ loff_t off;
nand_info_t *nand;
nand = &nand_info[0];
size_t rwsize;
rwsize = (size_t)len;
off = 2*(start_blk<<20);
printf("\n\nthe offset is : 0x%0x\n\n",off);
printf("\n\nthe rwsize is : 0x%0x\n\n",rwsize);
nand_erase_block(off,len);
nand_write_skip_bad(nand, off, &rwsize,buf,0);
return 0;
}
int write_nand_yaffs(u_char *buf, unsigned start_blk, unsigned int len, unsigned int blk_erase)
{
loff_t off;
nand_info_t *nand;
nand = &nand_info[0];
size_t rwsize;
rwsize = (size_t)len;
off = 2*(start_blk<<20);
printf("\n\nthe offset is : 0x%0x\n\n",off);
printf("\n\nthe rwsize is : 0x%0x\n\n",rwsize);
nand_erase_block(off,2*(blk_erase<<20));
nand_write_skip_bad(nand, off, &rwsize,buf,1);
return 0;
}
2.在u-boot/common/Makefile中添加需要编译的命令文件,COBJS-$(CONFIG_CMD_MTD) += cmd_mtd.o ,可以让cmd_mtd.c文件可以编译生成.o文件
3.在u-boot\include\config_cmd_all.h u-boot\include\config_cmd_default.h两个文件中加入预定义宏 #define CONFIG_CMD_MTD,使得uboot在编译命令时能加入该命令。
4.在u-boot\include\configs\SEP0611.h文件中同样加入宏定义#define CONFIG_CMD_MTD
5.最后编译即可