语句“mrc p15, 0, r9, c0, c0”将协处理器寄存器CP15读入寄存器r9,此寄存器保存CPUID;之后调用函数__lookup_processor_type查找处理器类型;
__lookup_processor_type函数:
__lookup_processor_type:
ARM( adr r3, 3f )
ARM( ldmda r3, {r5 - r7} )
THUMB( adr r3, 3f+4 )
THUMB( ldmdb r3, {r5 - r7} )
THUMB( sub r3, r3, #4 )
sub r3, r3, r7 @ get offset between virt&phys
add r5, r5, r3 @ convert virt addresses to
add r6, r6, r3 @ physical address space
1: ldmia r5, {r3, r4} @ value, mask
and r4, r4, r9 @ mask wanted bits
teq r3, r4
beq 2f
add r5, r5, #PROC_INFO_SZ @ sizeof(proc_info_list)
cmp r5, r6
blo 1b
mov r5, #0 @ unknown processor
2: mov pc, lr
ENDPROC(__lookup_processor_type)
/*
* Look in and arch/arm/kernel/arch.[ch] for
* more information about the __proc_info and __arch_info structures.
*/
.long __proc_info_begin
.long __proc_info_end
3: .long .
.long __arch_info_begin
.long __arch_info_end
语句“adr r3, 3f”向前寻找到标号3的地址,赋给寄存器r3;
语句“ldmda r3, {r5 - r7}”将寄存器r3所指地址处的数据分别读入r5,r6,r7,由于ldmda是“过后减少装载”方式,因此r6与r7分别读入的应是__proc_info_end和__proc_info_begin;
函数返回后,有一个操作:
movs r10, r5
将寄存器r5中的proc_info入口地址保存在寄存器r10中,这是为了后来调用__create_page_tables准备的
__lookup_machine_type函数:
__lookup_machine_type:
adr r3, 3b
ldmia r3, {r4, r5, r6}
sub r3, r3, r4 @ get offset between virt&phys
add r5, r5, r3 @ convert virt addresses to
add r6, r6, r3 @ physical address space
1: ldr r3, [r5, #MACHINFO_TYPE] @ get machine type
teq r3, r1 @ matches loader number?
beq 2f @ found
add r5, r5, #SIZEOF_MACHINE_DESC @ next machine_desc
cmp r5, r6
blo 1b
mov r5, #0 @ unknown machine
2: mov pc, lr
ENDPROC(__lookup_machine_type)
此函数紧挨着__lookup_processor_type,与__lookup_processor_type共用标号3,不同的是语句“adr r3, 3b”为向后查找标号3,将标号3的地址赋给寄存器r3;
接下来的语句“ldmia r3, {r4, r5, r6}”采用了过后增加装载指令查表,将__arch_info_begin、__arch_info_end地地址分别装入寄存器r5、r6;
回顾一下标号3:
/*
* Look in and arch/arm/kernel/arch.[ch] for
* more information about the __proc_info and __arch_info structures.
*/
.long __proc_info_begin
.long __proc_info_end
3: .long .
.long __arch_info_begin
.long __arch_info_end
可见,__lookup_processor_type函数通过ldmda指令查找__proc_info_表格,__lookup_machine_type函数则通过ldmia指令查找__arch_info_表格。
__arch_info_表格的生成,可见arch/arm/include/asm/mach/arch.h文件中定义的宏MACHINE_START:
/*
* Set of macros to define architecture features. This is built into
* a table by the linker.
*/
#define MACHINE_START(_type,_name) /
static const struct machine_desc __mach_desc_##_type /
__used /
__attribute__((__section__(".arch.info.init"))) = { /
.nr = MACH_TYPE_##_type, /
.name = _name,
#define MACHINE_END /
};
以TI OMAP3 参考板为例,MACHINE_START表格的具体填写,在arch/arm/mach-omap2/board-zoom2.c文件中:
MACHINE_START(OMAP_ZOOM2, "OMAP ZOOM2 board")
.phys_io = 0x48000000,
.io_pg_offst = ((0xd8000000) >> 18) & 0xfffc,
.boot_params = 0x80000100,
.map_io = omap_ldp_map_io,
.init_irq = omap_ldp_init_irq,
.init_machine = omap_ldp_init,
.timer = &omap_timer,
MACHINE_END
回到__lookup_machine_type函数,寄存器r1在kernel入口时已被赋予了type的数值,从u-boot代码include/asm-arm/mach-types.h文件中定义:
#define MACH_TYPE_OMAP_ZOOM2 1967
可知,boot向Kernel传入的数据应为1967,kernel中,OMAP_ZOOM2对应的type值由脚本文件arch/arm/tools/mach-types指定,该脚本中,行
omap_zoom2 MACH_OMAP_ZOOM2 OMAP_ZOOM2 1967
指定OMAP_ZOOM2对应的type值为1967,因此,函数__lookup_machine_type的查表过程成功,寄存器r5中返回ZOOM2对应信息的数据结构首地址;函数返回后,有一个操作:
movs r8, r5
将寄存器r5中的__arch_info入口地址保存在寄存器r8中,这也是为了后来调用__create_page_tables准备的
需要关注的数据结构machine_desc(在文件arch/arm/include/asm/mach/arch.h中定义):
struct machine_desc {
/*
* Note! The first four elements are used
* by assembler code in head.S, head-common.S
*/
unsigned int nr; /* architecture number */
unsigned int phys_io; /* start of physical io */
unsigned int io_pg_offst; /* byte offset for io
* page tabe entry */
const char *name; /* architecture name */
unsigned long boot_params; /* tagged list */
unsigned int video_start; /* start of video RAM */
unsigned int video_end; /* end of video RAM */
unsigned int reserve_lp0 :1; /* never has lp0 */
unsigned int reserve_lp1 :1; /* never has lp1 */
unsigned int reserve_lp2 :1; /* never has lp2 */
unsigned int soft_reboot :1; /* soft reboot */
void (*fixup)(struct machine_desc *,
struct tag *, char **,
struct meminfo *);
void (*map_io)(void);/* IO mapping function */
void (*init_irq)(void);
struct sys_timer *timer; /* system tick timer */
void (*init_machine)(void);
};
有许多事,需要回到u-boot才能说清楚……
include/configs/XXXX.h中(XXXX视具体平台而定),一般会作类似如下定义:
#define CONFIG_BOOTCOMMAND "mmcinit; fatload mmc 0 0x81c00000 uImage; bootm 0x81c00000"
编译时该宏CONFIG_BOOTCOMMAND传递给一个ENV项bootcmd,而在common/main.c中,函数main_loop取出了该env项,作为boot的过程开始启动kernel:
s = getenv ("bootcmd");
debug ("### main_loop: bootcmd=/"%s/"/n", s ? s : "");
if (bootdelay >= 0 && s && !abortboot (bootdelay)) {
# ifdef CONFIG_AUTOBOOT_KEYED
int prev = disable_ctrlc(1); /* disable Control C checking */
# endif
# ifndef CFG_HUSH_PARSER
run_command (s, 0);
我们关心的是最后一条命令bootm,在文件common/cmd_bootm.c中已经定义:
U_BOOT_CMD(
bootm, CFG_MAXARGS, 1, do_bootm,
"bootm - boot application image from memory/n",
"[addr [arg ...]]/n - boot application image stored in memory/n"
"/tpassing arguments 'arg ...'; when booting a Linux kernel,/n"
"/t'arg' can be the address of an initrd image/n"
);
即,真正要运行的程序是do_bootm。
一些关键的数据结构:
1、gd_t
在文件include/asm-arm/global_data.h中定义,在u-boot中广泛使用,通过宏:
DECLARE_GLOBAL_DATA_PTR;
将该数据结构的指针放入寄存器r8。
gd这个数据结构,在lib-arm/board.c文件中的start_armboot函数中进行初始化,详细分析以后再看。
2、bd_t
在文件include/asm-arm/u-boot.h中定义,作为u-boot的接口。
3、image_header_t
在文件include/image.h中定义,是u-boot即将加载的镜像文件头信息,u-boot将根据该头信息确定镜像文件是否正确,校验,镜像文件的起始地址以及入口地址等信息。
typedef struct image_header {
uint32_t ih_magic; /* Image Header Magic Number*/
uint32_t ih_hcrc; /* Image Header CRC Checksum*/
uint32_t ih_time; /* Image Creation Timestamp*/
uint32_t ih_size; /* Image Data Size
*/
uint32_t ih_load; /* Data Load Address*/
uint32_t ih_ep; /* Entry Point Address
*/
uint32_t ih_dcrc; /* Image Data CRC Checksum*/
uint8_t ih_os; /* Operating System */
uint8_t ih_arch; /* CPU architecture
*/
uint8_t ih_type; /* Image Type */
uint8_t ih_comp; /* Compression Type
*/
uint8_t ih_name[IH_NMLEN]; /* Image Name
*/
} image_header_t;
do_bootm函数会读出即将加载的镜像文件头,填入由该数据结构声明的全局变量header中,do_bootm函数通过对header的分析,确定是一个Linux镜像,则会调用do_bootm_linux函数,开始启动Linux的Kernel。
do_bootm_linux函数(在文件lib_arm/armlinux.c中),目前只关心最后一句:
theKernel (0, bd->bi_arch_number, bd->bi_boot_params);
theKernel在前面被定义为:
theKernel = (void (*)(int, int, uint))ntohl(hdr->ih_ep);
hdr->ih_ep使用的是do_bootm函数从image头中读取的入口,对于我们的板子来说,就是0x80008000;
bd->bi_arch_number, bd->bi_boot_params两个入口参数在的board_init函数中定义,bd->bi_arch_number赋值为板子的ID,bd->bi_boot_params赋值为0x80000100,该地址保存着do_bootm_linux函数创建的一系列ATAG。
参数的传递要看APCS(ARM过程调用标准),前四个参数会依次放入a1~a4,这是APCS为寄存器起的别名,对应于r0~r3,即,调用到kernel时,寄存器r0为0,r1为1967,r2为0x80000100。
u-boot的事情还远未说清楚,但是与目前有关的,大致差不多了。
欢迎回到Kernel中……(看了2.6.29内核,做了些修订,关于it指令的)
继续说__vet_atags函数,这个函数仍旧定义在arch/arm/kernel/head-common.s文件中:
/* Determine validity of the r2 atags pointer. The heuristic requires
* that the pointer be aligned, in the first 16k of physical RAM and
* that the ATAG_CORE marker is first and present. Future revisions
* of this function may be more lenient with the physical address and
* may also be able to move the ATAGS block if necessary.
*
* r8 = machinfo
*
* Returns:
* r2 either valid atags pointer, or zero
* r5, r6 corrupted
*/
__vet_atags:
tstr2, #0x3@ aligned?
bne1f
ldrr5, [r2, #0]@ is first tag ATAG_CORE?
subsr5, r5, #ATAG_CORE_SIZE
bne1f
ldrr5, [r2, #4]
ldrr6, =ATAG_CORE
cmpr5, r6
bne1f
movpc, lr@ atag pointer is ok
1:movr2, #0
movpc, lr
ENDPROC(__vet_atags)
由TI的芯片手册知,OMAP3430的SDRAM地址空间自0x80000000起,计1GB空间。函数开始的注释明确要求ATAG表需要在RAM物理地址前16KB,即0x80000000-0x80004000范围内,r2的0x80000100满足此要求。简单的判断了一下ATAG表的起始地址是否对齐,是否以ATAG_CORE作为开头标志,该函数返回。
__create_page_tables函数,这是要分析的重点函数,该函数就在arch/arm/kernel/head.s文件末尾。
先来看一个宏定义:
.macro pgtbl, rd
ldr /rd, =(KERNEL_RAM_PADDR - 0x4000)
.endm
KERNEL_RAM_PADDR的定义为:
#define KERNEL_RAM_PADDR (PHYS_OFFSET + TEXT_OFFSET)
PHYS_OFFSET在文件arch/arm/plat-omap/include/mach/memory.h中(怎么找到这里的,其实是arch/arm/kernel/head.s中包含了arch/arm/include/asm/memory.h中又包含了arch/arm/plat-omap/include/mach/memory.h的缘故……)定义为0x80000000,而TEXT_OFFSET则需要在Makefile中寻找到答案。arch/arm/Makefile中定义
……
textofs-y := 0x00008000
……
TEXT_OFFSET := $(textofs-y)
可见,TEXT_OFFSET的值为0x00008000,再回到宏pgtbl,可计算出KERNEL_RAM_PADDR的值为0x80000000+0x00008000=0x80008000最后赋给寄存器值为0x80008000-0x4000=0x80004000。
__create_page_tables函数的第一条语句为:
pgtbl r4 @ page table address
根据前面计算,寄存器r4赋值0x80004000,该地址是页表的起始地址,在kernel的入口地址(0x80008000)之前16KB。
之后,对该页表进行初始化操作。
第一步、将这16KB空间清零:
/*
* Clear the 16K level 1 swapper page table
*/
mov r0, r4
mov r3, #0
add r6, r0, #0x4000
1: str r3, [r0], #4
str r3, [r0], #4
str r3, [r0], #4
str r3, [r0], #4
teq r0, r6
bne 1b
第二步、填写内核所在页的页目录项
ldr r7, [r10, #PROCINFO_MM_MMUFLAGS] @ mm_mmuflags
这里需要再来回顾一下proc_info_list数据结构(在arch/arm/include/asm/procinfo.h文件中定义):
struct proc_info_list {
unsigned int cpu_val;
unsigned int cpu_mask;
unsigned long __cpu_mm_mmu_flags; /* used by head.S */
unsigned long __cpu_io_mmu_flags; /* used by head.S */
unsigned long __cpu_flush; /* used by head.S */
const char *arch_name;
const char *elf_name;
unsigned int elf_hwcap;
const char *cpu_name;
struct processor *proc;
struct cpu_tlb_fns *tlb;
struct cpu_user_fns *user;
struct cpu_cache_fns *cache;
};
指令ldr r7, [r10, #PROCINFO_MM_MMUFLAGS]取到了结构中的数据__cpu_mm_mmu_flags,送入寄存器r7,该参数在arch/arm/mm/proc-v7.s文件中定义如下:
.long PMD_TYPE_SECT | /
PMD_SECT_BUFFERABLE | /
PMD_SECT_CACHEABLE | /
PMD_SECT_AP_WRITE | /
PMD_SECT_AP_READ
计算下来,应该是:0b0000110000001110,即0x0c0e。
参照如下表格:arm采用的是段式页表,按照VMSA(Virtual Memory System Architecture)规定,每个表项描述1MB空间,16KB可存放4K个表项,覆盖4GB虚拟地址空间。关于页表的详细说明,可参阅ARM Architecture Reference Manual DDI0406B的B3章节。
mov r6, pc, lsr #20 @ start of kernel section
当前pc的高12位,即内核所在物理地址的索引放入寄存器r6;
orr r3, r7, r6, lsl #20 @ flags + kernel base
索引号与寄存器r7中的标志字节相或,合并后的描述符放入寄存器r3;
str r3, [r4, r6, lsl #2] @ identity mapping
每个描述符占4字节,所以内核的起始页目录项地址应该在[r4]+[r6]x4处,将r3的描述符内容送入其中;
第三步、根据当前内核所划分的虚拟地址空间,建立内核页表项,Linux默认内核在最高1G,因此内核起始虚拟地址一般为0xc0000000。
add r0, r4, #(KERNEL_START & 0xff000000) >> 18
KERNEL_START的定义如下:
#define KERNEL_START KERNEL_RAM_VADDR
KERNEL_RAM_VADDR的定义如下:
#define KERNEL_RAM_VADDR (PAGE_OFFSET + TEXT_OFFSET)
PAGE_OFFSET的定义,如果别处没有关怀过,那就应该是arch/arm/include/asm/memory.h文件中的如下定义了:
/*
* Page offset: 3GB
*/
#ifndef PAGE_OFFSET
#define PAGE_OFFSET UL(0xc0000000)
#endif
计算所得,的值应为0xc0008000,这是Kernel起始的虚拟地址,截取高14位,取高8位有效与表头地址相加,作为基址送入寄存器r0;这里之所以截取高14位,是由于section页表,可记录4K个表项,正好取虚拟地址的高12位作为页表的索引,每个表项占用4字节,则索引号需乘4,为14位。
str r3, [r0, #(KERNEL_START & 0x00f00000) >> 18]!
寄存器r3内容,是前面已经计算过的包含物理地址索引的页描述符,将其存放入寄存器r0增加偏移虚拟地址高12位中的低4位所指表项中,并更新r0的内容。
ldr r6, =(KERNEL_END - 1)
寄存器r6保留内核代码长度;
add r0, r0, #4
寄存器r0指向下一页表项;
add r6, r4, r6, lsr #18
根据内核长度计算出内核页表的最后入口;
1: cmp r0, r6
比较内核页表是否已填写完成;
add r3, r3, #1 << 20
物理地址索引加一;
it ls
小于等于判断,it指令是if-then块,具体内容,请参阅RealView编译工具《汇编器指南》(发现2.6.29内核把这句去掉了,草!);
strls r3, [r0], #4
如果小于等于成立,填写该表项;
bls 1b
如果小于等于则跳回标号1;
第四步、填写内存开始时的1MB段,因为这里有boot传来的内核启动参数。
/*
* Then map first 1MB of ram in case it contains our boot params.
*/
add r0, r4, #PAGE_OFFSET >> 18
orr r6, r7, #(PHYS_OFFSET & 0xff000000)
.if (PHYS_OFFSET & 0x00f00000)
orr r6, r6, #(PHYS_OFFSET & 0x00f00000)
.endif
str r6, [r0]
刚好一页,不用判断了。
最后,返回了……
mov pc, lr
回到arch/arm/kernel/head.s文件中填充完页表后的初始化程序:
/*
* The following calls CPU specific code in a position independent
* manner. See arch/arm/mm/proc-*.S for details. r10 = base of
* xxx_proc_info structure selected by __lookup_machine_type
* above. On return, the CPU will be ready for the MMU to be
* turned on, and r0 will hold the CPU control register value.
*/
ldr r13, __switch_data @ address to jump to after
@ mmu has been enabled
badr lr, __enable_mmu @ return (PIC) address
准备好了返回地址,调用如下:
ARM( add pc, r10, #PROCINFO_INITFUNC
)
前面的初始化保证,寄存器r10指向procinfo,在文件arch/arm/kernel/asm-offsets.c文件中宏PROCINFO_INITFUNC的定义如下:
DEFINE(PROCINFO_INITFUNC, offsetof(struct proc_info_list, __cpu_flush));
到这里,是该轻易莲步关注arch/arm/include/asm/procinfo.h文件中的数据结构proc_info_list的时候了:
struct proc_info_list {
unsigned int cpu_val;
unsigned int cpu_mask;
unsigned long __cpu_mm_mmu_flags; /* used by head.S */
unsigned long __cpu_io_mmu_flags; /* used by head.S */
unsigned long __cpu_flush; /* used by head.S */
const char *arch_name;
const char *elf_name;
unsigned int elf_hwcap;
const char *cpu_name;
struct processor *proc;
struct cpu_tlb_fns *tlb;
struct cpu_user_fns *user;
struct cpu_cache_fns *cache;
};
此数据结构中,__cpu_flush一项,在arch/arm/mm/proc-v7.s中,定义为一条调用指令:b __v7_setup
因此add pc, r10, #PROCINFO_INITFUNC这一句辗转调用了__v7_setup函数。
/*
* __v7_setup
*
* Initialise TLB, Caches, and MMU state ready to switch the MMU
* on. Return in r0 the new CP15 C1 control register setting.
*
* We automatically detect if we have a Harvard cache, and use the
* Harvard cache control instructions insead of the unified cache
* control instructions.
*
* This should be able to cover all ARMv7 cores.
*
* It is assumed that:
* - cache type register is implemented
*/
__v7_setup:
adr r12, __v7_setup_stack @ the local stack
stmia r12, {r0-r5, r7, r9, r11, lr}
寄存器入栈__v7_setup_stack是个11个word的局部栈,就在本函数后面定义,stmia表明此处用的是升序栈;
bl v7_flush_dcache_all
清除数据缓存;这个函数定义在arch/arm/mm/cache-v7.s文件中(这里为什么不调用v7_flush_cache_all()将数据、指令缓冲区全部清除,而只清除数据缓冲区呢?);
ldmia r12, {r0-r5, r7, r9, r11, lr}
寄存器出栈;
/*
* On OMAP3 devices the auxilary control register can be accessed
* only is secure mode using SMI /PPA. The IBE bit is enabled at the
* u-boot level using SMI service. So no need to set that bit again.
*/
#ifndef CONFIG_ARCH_OMAP3430
#ifdef CONFIG_ARM_ERRATA_430973
mrc p15, 0, r10, c1, c0, 1 @ read aux control register
orr r10, r10, #(1 << 6) @ set IBE to 1
mcr p15, 0, r10, c1, c0, 1 @ write aux control register
#endif
#endif
清除分支预测缓冲区,对OMAP3来京不需要。
mov r10, #0
#ifdef HARVARD_CACHE
mcr p15, 0, r10, c7, c5, 0 @ I+BTB cache invalidate
#endif
哈佛结构的cache,v7不支持。
dsb
数据同步屏障是一种特殊的内存屏障。 只有当此指令执行完毕后,才会执行程序中位于此指令后的指令。 当满足以下条件时,此指令才会完成:
位于此指令前的所有显式内存访问均完成。
位于此指令前的所有高速缓存、跳转预测和 TLB 维护操作全部完成。
#ifdef CONFIG_MMU
mcr p15, 0, r10, c8, c7, 0 @ invalidate I + D TLBs
置指令和数据TLB无效(寄存器r10在前面已清零);
mcr p15, 0, r10, c2, c0, 2 @ TTB control register
TLB控制寄存器清零;
orr r4, r4, #TTB_RGN_OC_WB @ mark PTWs outer cacheable, WB
寄存器r4仍然保存着页表起始地址0x80004000,TTB_RGN_OC_WB被定义为3<<3,对应于转换表基址寄存器的RNG位,此配置为使能TLB缓存,写回模式,写时不分配空间;
mcr p15, 0, r4, c2, c0, 0 @ load TTB0
mcr p15, 0, r4, c2, c0, 1 @ load TTB1
将该参数写入转换表基址寄存器0和1;
mov r10, #0x1f @ domains 0, 1 = manager
mcr p15, 0, r10, c3, c0, 0 @ load domain access register
区域访问许可控制器配置为,D0、D1区域的访问不与TLB中的访问许可位校验,D3区域的访问会与TLB中的访问许可位校验(为什么这样配置,我现在还不清楚);
#endif
#if defined(CONFIG_ARCH_OMAP3)
@ OMAP3: L2EN bit accessed in nonsecure mode
@ L2 cache is enabled in the aux control register
mrc p15, 0, r0, c1, c0, 1
#ifdef CONFIG_CPU_L2CACHE_DISABLE
bic r0, r0, #0x2 @ disable L2 Cache
#else
orr r0, r0, #0x2 @ enable L2 Cache
#endif
mcr p15, 0, r0, c1, c0, 1 @ Enable the L2EN banked bit as well
#endif
使能L2 Cache;
#ifdef CONFIG_ARCH_OMAP34XX
#ifdef CONFIG_CPU_LOCKDOWN_TO_64K_L2
mov r10, #0xfc
mcr p15, 1, r10, c9, c0, 0
#endif
#ifdef CONFIG_CPU_LOCKDOWN_TO_128K_L2
mov r10, #0xf0
mcr p15, 1, r10, c9, c0, 0
#endif
#ifdef CONFIG_CPU_LOCKDOWN_TO_256K_L2
mov r10, #0x00
mcr p15, 1, r10, c9, c0, 0
#endif
以上是OMAP自定义的寄存器,配置了OMAP的L2 Cache大小,OMAP3430是256K;
#ifdef CONFIG_CPU_USER_L2_PLE_ACCESS
mov r10, #0x3
mcr p15, 0, r10, c11, c1, 0
使能应用程序访问PLE(preloading engine)寄存器通道0、1
#endif
#endif
adr r5, v7_crval
ldmia r5, {r5, r6}
mrc p15, 0, r0, c1, c0, 0 @ read control register
bic r0, r0, r5 @ clear bits them
orr r0, r0, r6 @ set them
mov pc, lr @ return to head.S:__ret
见如下定义:
v7_crval:
ARM( crval clear=0x0120c302, mmuset=0x00c0387d, ucset=0x00c0187c)
THUMB( crval clear=0x0120c302, mmuset=0x40c0387d, ucset=0x40c0187c)
控制寄存器值读入寄存器r0,清了一些位,又置了一些位,关键是,寄存器r0的值,已经使能了mmu,但是很显然,这里还没有写入CP15,这个激动人心的操作将由__enable_mmu函数来完成。
ENDPROC(__v7_setup)
打完收工!
结案陈词,这段程序关闭了指令和数据缓冲区,配置TLB寄存器及相应的控制寄存器,以及,计算好了mmu控制寄存器的初始化参数,就等着打开mmu,进入一个虚幻的世界了。
从__v7_setup函数返回来,进入到__enable_mmu过程,做起飞前的最后确认:
__enable_mmu:
#ifdef CONFIG_ALIGNMENT_TRAP
orr r0, r0, #CR_A
#else
bic r0, r0, #CR_A
#endif
根据配置选项,决定是否打开数据对齐检查,一般是要打开的;
#ifdef CONFIG_CPU_DCACHE_DISABLE
bic r0, r0, #CR_C
#endif
根据配置选项,决定是否关闭开数据缓冲区,一般是不需要关闭的;
#ifdef CONFIG_CPU_BPREDICT_DISABLE
bic r0, r0, #CR_Z
#endif
根据配置选项,决定是否要关闭分支预测功能,一般是不需要关闭的;
#ifdef CONFIG_CPU_ICACHE_DISABLE
bic r0, r0, #CR_I
#endif
根据配置选项,决定是否关闭开指令缓冲区,一般是不需要关闭的;
mov r5, #(domain_val(DOMAIN_USER, DOMAIN_MANAGER) | /
domain_val(DOMAIN_KERNEL, DOMAIN_MANAGER) | /
domain_val(DOMAIN_TABLE, DOMAIN_MANAGER) | /
domain_val(DOMAIN_IO, DOMAIN_CLIENT))
结合数据手册,此段是将D0、D1两个域,D1对应于USER域,D0对应于KERNEL和TABLE域,配置为管理模式,对这两个域的访问和执行不受TLB中相应标志位影响,D2域,对应于IO域,配置为客户模式,对此域的访问受TLB中相应标志位的检查;
mcr p15, 0, r5, c3, c0, 0 @ load domain access register
写入域访问控制寄存器,其实和__v7_setup函数中写入的值相同;
mcr p15, 0, r4, c2, c0, 0 @ load page table pointer
__v7_setup函数中也已经写过这个寄存器了;
b __turn_mmu_on
__turn_mmu_on就在后面紧跟着,不明白为何还要b一下;
ENDPROC(__enable_mmu)
再看看紧接着的__turn_mmu_on函数:
/*
* Enable the MMU. This completely changes the structure of the visible
* memory space. You will not be able to trace execution through this.
* If you have an enquiry about this, *please* check the linux-arm-kernel
* mailing list archives BEFORE sending another post to the list.
*
* r0 = cp#15 control register
* r13 = *virtual* address to jump to upon completion
*
* other registers depend on the function called upon completion
*/
.align 5
原来是因为有这个,所以要b过来,但为什么是以5对齐?在网上搜索了一下,原来.align n的语法,有按照n对齐的,也有按照2的n次方对齐的,而arm-linux是按照2的n次方对齐的,即,是以20字节位置对齐的,详见某篇博文:http://www.eetop.cn/blog/html/45/11145-1211.html
__turn_mmu_on:
mov r0, r0
暂时认为这是常规的nop语句……
mcr p15, 0, r0, c1, c0, 0 @ write control reg
终于将寄存器r0的值写入到了mmu控制寄存器中,此时的CPU终于可以“内牛满面”!
mrc p15, 0, r3, c0, c0, 0 @ read id reg
对于Cortex A8处理器来说,此时寄存器r3的值也许是:0x413fc082字样,其中41代表ARM,c08就是Cortex A8;
mov r3, r3
mov r3, r3
继续认为这是nop语句,或者说,在这里是清除流水线,那么,之后的CPU,已经平稳过渡到了虚拟地址空间的环境?(我不敢确认)……
mov pc, r13
无需废话了,很久以前,寄存器r13就准备好了__switch_data的入口地址;
ENDPROC(__turn_mmu_on)