嵌入式Linux2.6内核启动流程苍松书屋

1、linux内核构成(国嵌)linux/arch/arm/boot/compressed/head.s1解压缩2初始化3启动应用程序1 arch/arm/boot/compressed/makefile arch/arm/boot/compressed/vmlinux.lds2. arch/arm/kernel/vmlinux.ldslinux内核启动流程（国嵌）arch/arm/boot/compressed/start.s（head.s负责解压缩）start: .type start,#function .rept 8 mov r0, r0 .endr b 1f .word 0x016f28

2、18 magic numbers to help the loader .word start absolute load/run zimage address .word _edata zimage end address1: mov r7, r1 save architecture id mov r8, r2 save atags pointer这也标志着u-boot将系统完全的交给了os，bootloader生命终止。之后代码在133行会读取cpsr并判断是否处理器处于supervisor模式从u-boot进入kernel，系统已经处于svc32模式；而利用angel进入则处于user模

3、式，还需要额外两条指令。之后是再次确认中断关闭，并完成cpsr写入 mrs r2, cpsr get current mode tst r2, #3 not user? bne not_angel mov r0, #0x17 angel_swireason_entersvc swi 0x123456 angel_swi_armnot_angel: mrs r2, cpsr turn off interrupts to orr r2, r2, #0xc0 prevent angel from running msr cpsr_c, r2 然后在lc0地址处将分段信息导入r0-r6、ip、sp等寄

4、存器，并检查代码是否运行在与链接时相同的目标地址，以决定是否进行处理。由于现在很少有人不使用loader和tags，将zimage烧写到rom直接从0x0位置执行，所以这个处理是必须的（但是zimage的头现在也保留了不用loader也可启动的能力）。arm架构下自解压头一般是链接在0x0地址而被加载到0x30008000运行，所以要修正这个变化。涉及到r5寄存器存放的zimage基地址 r6和r12（即ip寄存器）存放的got（global offset table） r2和r3存放的bss段起止地址 sp栈指针地址 很简单，这些寄存器统统被加上一个你也能猜到的偏移地址 0x30008000

5、。该地址是s3c2410相关的，其他的arm处理器可以参考下表pxa2xx是0xa0008000 ixp2x00和ixp4xx是0x00008000 freescale i.mx31/37是0x80008000 ti davinci dm64xx是0x80008000 ti omap系列是0x80008000 at91rm/sam92xx系列是0x20008000 cirrus ep93xx是0x00008000 这些操作发生在代码172行开始的地方，下面只粘贴一部分 add r5, r5, r0 add r6, r6, r0 add ip, ip, r0后面在211行进行bss段的清零工作n

6、ot_relocated: mov r0, #01: str r0, r2, #4 clear bss str r0, r2, #4 str r0, r2, #4 str r0, r2, #4 cmp r2, r3 blo 1b 然后224行，打开cache，并为后面解压缩设置64kb的临时malloc空间 bl cache_on mov r1, sp malloc space above stack add r2, sp, #0x10000 64k max 接下来238行进行检查，确定内核解压缩后的image目标地址是否会覆盖到zimage头，如果是则准备将zimage头转移到解压出来的内核

7、后面 cmp r4, r2 bhs wont_overwrite sub r3, sp, r5 > compressed kernel size add r0, r4, r3, lsl #2 allow for 4x expansion cmp r0, r5 bls wont_overwrite mov r5, r2 decompress after malloc space mov r0, r5 mov r3, r7 bl decompress_kernel真实情况在大多数的应用中，内核编译都会把压缩的zimage和非压缩的image链接到同样的地址，s3c2410平台下即是0x300

8、08000。这样做的好处是，人们不用关心内核是image还是zimage，放到这个位置执行就ok，所以在解压缩后zimage头必须为真正的内核让路。在250行解压完毕，内核长度返回值存放在r0寄存器里。在内核末尾空出128字节的栈空间用，并且使其长度128字节对齐。 add r0, r0, #127 + 128 alignment + stack bic r0, r0, #127 align the kernel length算出搬移代码的参数：计算内核末尾地址并存放于r1寄存器，需要搬移代码原来地址放在r2，需要搬移的长度放在r3。然后执行搬移，并设置好sp指针指向新的栈（原来的栈也会被内核

9、覆盖掉） add r1, r5, r0 end of decompressed kernel adr r2, reloc_start ldr r3, lc1 add r3, r2, r31: ldmia r2!, r9 - r14 copy relocation code stmia r1!, r9 - r14 ldmia r2!, r9 - r14 stmia r1!, r9 - r14 cmp r2, r3 blo 1b add sp, r1, #128 relocate the stack搬移完成后刷新cache，因为代码地址变化了不能让cache再命中被内核覆盖的老地址。然后跳转到新的

10、地址继续执行 bl cache_clean_flush add pc, r5, r0 call relocation code注意zimage在解压后的搬移和跳转会给gdb调试内核带来麻烦。因为用来调试的符号表是在编译是生成的，并不知道以后会被搬移到何处去，只有在内核解压缩完成之后，根据计算出来的参数“告诉”调试器这个变化。以撰写本文时使用的zimage为例，内核自解压头重定向后，reloc_start地址由0x30008360变为0x30533e60。故我们要把vmlinux的符号表也相应的从0x30008000后移到0x30533b00开始，这样gdb就可以正确的对应源代码和机器指令。随着

11、头部代码移动到新的位置，不会再和内核的目标地址冲突，可以开始内核自身的搬移了。此时r0寄存器存放的是内核长度（严格的说是长度外加128byte的栈），r4存放的是内核的目的地址0x30008000，r5是目前内核存放地址，r6是cpu id，r7是machine id，r8是atags地址。代码从501行开始reloc_start: add r9, r5, r0 sub r9, r9, #128 do not copy the stack debug_reloc_start mov r1, r41: .rept 4 ldmia r5!, r0, r2, r3, r10 - r14 reloca

12、te kernel stmia r1!, r0, r2, r3, r10 - r14 .endr cmp r5, r9 blo 1b add sp, r1, #128 relocate the stack接下来在516行清除并关闭cache，清零r0，将machine id存入r1，atags指针存入r2，再跳入0x30008000执行真正的内核imagecall_kernel: bl cache_clean_flush bl cache_off mov r0, #0 must be zero mov r1, r7 restore architecture number mov r2, r8

13、restore atags pointer mov pc, r4 call kernel内核代码入口在arch/arm/kernel/head.s文件的83行。首先进入svc32模式，并查询cpu id，检查合法性 msr cpsr_c, #psr_f_bit | psr_i_bit | svc_mode ensure svc mode and irqs disabled mrc p15, 0, r9, c0, c0 get processor id bl _lookup_processor_type r5=procinfo r9=cpuid movs r10, r5 invalid proc

14、essor (r5=0)? beq _error_p yes, error 'p'接着在87行进一步查询machine id并检查合法性 bl _lookup_machine_type r5=machinfo movs r8, r5 invalid machine (r5=0)? beq _error_a yes, error 'a'其中_lookup_processor_type在linux-2.6.24-moko-linuxbj/arch/arm/kernel/head-common.s文件的149行，该函数首将标号3的实际地址加载到r3，然后将编译时生成的

15、_proc_info_begin虚拟地址载入到r5，_proc_info_end虚拟地址载入到r6，标号3的虚拟地址载入到r7。由于adr伪指令和标号3的使用，以及_proc_info_begin等符号在linux-2.6.24-moko-linuxbj/arch/arm/kernel/vmlinux.lds而不是代码中被定义，此处代码不是非常直观，想弄清楚代码缘由的读者请耐心阅读这两个文件和adr伪指令的说明。r3和r7分别存储的是同一位置标号3的物理地址（由于没有启用mmu，所以当前肯定是物理地址）和虚拟地址，所以儿者相减即得到虚拟地址和物理地址之间的offset。利用此offset，将r

16、5和r6中保存的虚拟地址转变为物理地址_lookup_processor_type: adr r3, 3f ldmda r3, r5 - r7 sub r3, r3, r7 get offset between virt&phys add r5, r5, r3 convert virt addresses to add r6, r6, r3 physical address space然后从proc_info中读出内核编译时写入的processor id和之前从cpsr中读到的processor id对比，查看代码和cpu硬件是否匹配（想在arm920t上运行为cortex-a8编译的

17、内核？不让！）。如果编译了多种处理器支持，如versatile板，则会循环每种type依次检验，如果硬件读出的id在内核中找不到匹配，则r5置0返回1:ldmiar5, r3, r4 value, maskandr4, r4, r9 mask wanted bitsteqr3, r4beq2faddr5, r5, #proc_info_sz sizeof(proc_info_list)cmpr5, r6blo1bmovr5, #0 unknown processor2:movpc, lr _lookup_machine_type在linux-2.6.24-moko-linuxbj/arch/a

18、rm/kernel/head-common.s文件的197行，编码方法与检查processor id完全一样，请参考前段_lookup_machine_type:adrr3, 3bldmiar3, r4, r5, r6subr3, r3, r4 get offset between virt&physaddr5, r5, r3 convert virt addresses toaddr6, r6, r3 physical address space1:ldrr3, r5, #machinfo_type get machine typeteqr3, r1 matches loader n

19、umber?beq2f foundaddr5, r5, #sizeof_machine_desc next machine_desccmpr5, r6blo1bmovr5, #0 unknown machine2:movpc, lr代码回到head.s第92行，检查atags合法性，然后创建初始页表bl_vet_atagsbl_create_page_tables 创建页表的代码在218行，首先将内核起始地址-0x4000到内核起始地址之间的16k存储器清0_create_page_tables:pgtblr4 page table address/* * clear the 16k leve

20、l 1 swapper page table */movr0, r4movr3, #0addr6, r0, #0x40001:strr3, r0, #4strr3, r0, #4strr3, r0, #4strr3, r0, #4teqr0, r6bne1b 然后在234行将proc_info中的mmu_flags加载到r7ldrr7, r10, #procinfo_mm_mmuflags mm_mmuflags在242行将pc指针右移20位，得到内核第一个1mb空间的段地址存入r6，在s3c2410平台该值是0x300。接着根据此值存入映射标识movr6, pc, lsr #20 start

21、 of kernel sectionorrr3, r7, r6, lsl #20 flags + kernel basestrr3, r4, r6, lsl #2 identity mapping完成页表设置后回到102行，为打开虚拟地址映射作准备。设置sp指针，函数返回地址lr指向_enable_mmu，并跳转到linux-2.6.24-moko-linuxbj/arch/arm/mm/proc-arm920.s的386行，清除i-cache、d-cache、write buffer和tlb_arm920_setup:movr0, #0mcrp15, 0, r0, c7, c7 invali

22、date i,d caches on v4mcrp15, 0, r0, c7, c10, 4 drain write buffer on v4#ifdef config_mmumcrp15, 0, r0, c8, c7 invalidate i,d tlbs on v4#endif然后返回head.s的158行，加载domain和页表，跳转到_turn_mmu_on_enable_mmu:#ifdef config_alignment_traporrr0, r0, #cr_a#elsebicr0, r0, #cr_a#endif#ifdef config_cpu_dcache_disableb

23、icr0, r0, #cr_c#endif#ifdef config_cpu_bpredict_disablebicr0, r0, #cr_z#endif#ifdef config_cpu_icache_disablebicr0, r0, #cr_i#endifmovr5, #(domain_val(domain_user, domain_manager) | domain_val(domain_kernel, domain_manager) | domain_val(domain_table, domain_manager) | domain_val(domain_io, domain_cl

24、ient)mcrp15, 0, r5, c3, c0, 0 load domain access registermcrp15, 0, r4, c2, c0, 0 load page table pointerb_turn_mmu_on在194行把mmu使能位写入mmu，激活虚拟地址。然后将原来保存在sp中的地址载入pc，跳转到head-common.s的_mmap_switched，至此代码进入虚拟地址的世界movr0, r0mcrp15, 0, r0, c1, c0, 0 write control regmrcp15, 0, r3, c0, c0, 0 read id regmovr3,

25、 r3movr3, r3movpc, r13在head-common.s的37行开始清除内核bss段，processor id保存在r9，machine id报存在r1，atags地址保存在r2，并将控制寄存器保存到r7定义的内存地址。接下来跳入linux-2.6.24-moko-linuxbj/init/main.c的507行，start_kernel函数。这里只粘贴部分代码（第一个c语言函数，作一系列的初始化）_mmap_switched:adrr3, _switch_data + 4ldmiar3!, r4, r5, r6, r7cmpr4, r5 copy data segment i

26、f needed1:cmpner5, r6ldrnefp, r4, #4strnefp, r5, #4bne1basmlinkage void _init start_kernel(void)char * command_line;extern struct kernel_param _start_param, _stop_param;smp_setup_processor_id();/* * need to run as early as possible, to initialize the * lockdep hash: */lockdep_init();debug_objects_ea

27、rly_init();cgroup_init_early();local_irq_disable();early_boot_irqs_off();early_init_irq_lock_class();/* * interrupts are still disabled. do necessary setups, then * enable them */lock_kernel();tick_init();boot_cpu_init();page_address_init();printk(kern_notice);printk(linux_banner);setup_arch(&co

28、mmand_line);mm_init_owner(&init_mm, &init_task);setup_command_line(command_line);setup_per_cpu_areas();setup_nr_cpu_ids();smp_prepare_boot_cpu();/* arch-specific boot-cpu hooks */* * set up the scheduler prior starting any interrupts (such as the * timer interrupt). full topology setup happe

29、ns at smp_init() * time - but meanwhile we still have a functioning scheduler. */sched_init();/* * disable preemption - early bootup scheduling is extremely * fragile until we cpu_idle() for the first time. */preempt_disable();build_all_zonelists();page_alloc_init();printk(kern_notice "kernel c

30、ommand line: %sn", boot_command_line);parse_early_param();parse_args("booting kernel", static_command_line, _start_param, _stop_param - _start_param, &unknown_bootoption);if (!irqs_disabled() printk(kern_warning "start_kernel(): bug: interrupts were ""enabled *very*

31、 early, fixing itn");local_irq_disable();sort_main_extable();trap_init();rcu_init();/* init some links before init_isa_irqs() */early_irq_init();init_irq();pidhash_init();init_timers();hrtimers_init();softirq_init();timekeeping_init();time_init();sched_clock_init();profile_init();if (!irqs_disa

32、bled()printk(kern_crit "start_kernel(): bug: interrupts were " "enabled earlyn");early_boot_irqs_on();local_irq_enable();/* * hack alert! this is early. we're enabling the console before * we've done pci setups etc, and console_init() must be aware of * this. but we do wa

33、nt output early, in case something goes wrong. */console_init();if (panic_later)panic(panic_later, panic_param);lockdep_info();/* * need to run this when irqs are enabled, because it wants * to self-test hard/soft-irqs on/off lock inversion bugs * too: */locking_selftest();#ifdef config_blk_dev_init

34、rdif (initrd_start && !initrd_below_start_ok && page_to_pfn(virt_to_page(void *)initrd_start) < min_low_pfn) printk(kern_crit "initrd overwritten (0x%08lx < 0x%08lx) - " "disabling it.n", page_to_pfn(virt_to_page(void *)initrd_start), min_low_pfn);initrd_sta

35、rt = 0;#endifvmalloc_init();vfs_caches_init_early();cpuset_init_early();page_cgroup_init();mem_init();enable_debug_pagealloc();cpu_hotplug_init();kmem_cache_init();debug_objects_mem_init();idr_init_cache();setup_per_cpu_pageset();numa_policy_init();if (late_time_init)late_time_init();calibrate_delay

36、();pidmap_init();pgtable_cache_init();prio_tree_init();anon_vma_init();#ifdef config_x86if (efi_enabled)efi_enter_virtual_mode();#endifthread_info_cache_init();cred_init();fork_init(num_physpages);proc_caches_init();buffer_init();key_init();security_init();vfs_caches_init(num_physpages);radix_tree_i

37、nit();signals_init();/* rootfs populating might need page-writeback */page_writeback_init();#ifdef config_proc_fsproc_root_init();#endifcgroup_init();cpuset_init();taskstats_init_early();delayacct_init();check_bugs();acpi_early_init(); /* before lapic and smp init */ftrace_init();/* do the rest non-

38、_init'ed, we're now alive */rest_init();tatic noinline void _init_refok rest_init(void)_releases(kernel_lock)int pid;kernel_thread(kernel_init, null, clone_fs | clone_sighand);numa_default_policy();pid = kernel_thread(kthreadd, null, clone_fs | clone_files);kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);unlock_kernel();/* * the boot idle thread must execute schedule() * at least once to get things moving: */init_idle_bootup_task(current);rcu_scheduler_starting()