一. 前言
前文分析了setup_vm函数，主要是几个配置页表的函数的实现。这一篇就进一步来看看setup_vm完之后，配置的页表到底将那些虚拟地址映射到了哪里。我们通过手算和GDB查看对照的方式来加深印象。
二. 分析过程
在如下位置打断点运行到该处，即setup_vm执行完后，relocate执行前。
hb *0x80201000
c
然后打印出各个页表的内容，和手动计算对比。
从源码可以看出构建了两个根页表，
一条是
trampoline_pg_dir->trampoline_pmd
该条是relocate时va-pa切换运行时临时使用，只映射了链接虚拟地址开始2MB到加载运行物理地址开始的2MB，
一条是
early_pg_dir->fixmap_pmd->fixmap_pte
->early_pmd
->early_dtb_pmd->
在reloacte通过trampoline_pg_dir切换到虚拟地址运行后，再切换到页表
2.1临时页表early_pg_dir
2.1.1trampoline_pg_dir
(gdb) p &trampoline_pg_dir$14 = (pgd_t (*)[512]) 0xffffffe001b7d000 <trampoline_pg_dir>(gdb)(gdb) p /x *(pgd_t (*)[512])0x81d7d000$15 = {{pgd = 0x0} <repeats 384 times>, {pgd = 0x2075ec01}, {pgd = 0x0} <repeats 127 times>}(gdb)
由以下代码配置
/* Setup trampoline PGD and PMD */create_pgd_mapping(trampoline_pg_dir, PAGE_OFFSET,(uintptr_t)trampoline_pmd, PGDIR_SIZE, PAGE_TABLE);
即配置pgd页表trampoline_pg_dir
Va=PAGE_OFFSET=0xffffffe0 00000000
Pa=trampoline_pmd=0x81d7b000
Sz=PGDIR_SIZE=1GB
Prot=PAGE_TABLE=1
先计算pgd_idx=(va>>30)&511=(0xffffffe0 00000000>>30)&511=0x180=384
即trampoline_pg_dir中一个条目对应1GB，这里需要对应到第384个条目去(右移30位，即按照1GB的颗粒度)，并且表只有一页即512个条目，所以按照512取余。
所以对应条目为trampoline_pg_dir[384]
然后计算条目的内容
pfn_pgd(PFN_DOWN(pa), prot);
其中pa为0x81d7b000，prot为1
所以PFN_DOWN(pa)为0x81d7b000>>12
所以值为
((0x81d7b000>>12 ) <<10)| 1=0x2075EC01
和前面的打印信息对应
(gdb) p /x *(pgd_t (*)[512])0x81d7d000$15 = {{pgd = 0x0} <repeats 384 times>, {pgd = 0x2075ec01}, {pgd = 0x0} <repeats 127 times>}(gdb)
2.1.2trampoline_pmd
(gdb) p &trampoline_pmd
$16 = (pmd_t (*)[512]) 0xffffffe001b7b000 <trampoline_pmd>
(gdb)
(gdb) p /x *(pmd_t (*)[512])0x81d7b000
$17 = {{pmd = 0x200800ef}, {pmd = 0x0} <repeats 511 times>}
(gdb)
由以下代码配置
create_pmd_mapping(trampoline_pmd, PAGE_OFFSET,load_pa, PMD_SIZE, PAGE_KERNEL_EXEC);
配置PMD页表 trampoline_pmd
Va=PAGE_OFFSET=0xffffffe0 00000000
Pa=load_pa=0x80200000
Sz=PMD_SIZE=2MB
Prot=PAGE_KERNEL_EXEC=0xEF
arch/riscv/include/asm/pgtable-bits.h
include/linux/pgtable.h
定义了PAGE_KERNEL_EXEC相关位
计算
pmd_idx = pmd_index(va); =( va>>21)&511=(0xffffffe0 00000000>>21)&511=0
即trampoline_pmd中一个条目对应2MB虚拟地址，这里需要对应到第0个条目去(右移21位，即按照2MB的颗粒度)，并且表只有一页即512个条目，所以按照512取余。
再计算
pfn_pmd(PFN_DOWN(pa), prot); =
pfn_pmd(PFN_DOWN(0x80200000), 0xEF); =
pfn_pmd(0x80200000>>12, 0xEF);=
((0x80200000>>12)<<10)|0xEF=0x200800EF
和GDB打印信息对应
(gdb) p /x *(pmd_t (*)[512])0x81d7b000
$17 = {{pmd = 0x200800ef}, {pmd = 0x0} <repeats 511 times>}
(gdb)
2.1.3 映射效果
所以以上两级页表，将链接虚拟地址0xffffffe000000000开始的2MB映射到了运行物理地址0x80200000开始的2MB。这样relocate时先将satp设置到trampoline_pg_dir时，访问这片链接地址实际就是访问到对应的物理地址实现无缝切换。这里仅仅作为中转，所以只映射了前面2MB即可，也就是relocate处的代码要在该范围内。
2.2 页表early_pg_dir
2.2.1 一级early_pg_dir
(gdb) p &early_pg_dir
$8 = (pgd_t (*)[512]) 0xffffffe00087b000 <early_pg_dir>
(gdb)
(gdb) p /x *(pgd_t (*)[512])0x80a7b000
$9 = {{pgd = 0x0}, {pgd = 0x2029e401}, {pgd = 0x0} <repeats 313 times>, {pgd = 0x2075e801}, {pgd = 0x0} <repeats 68 times>, {
pgd = 0x2029e801}, {pgd = 0x0} <repeats 127 times>}
(gdb)
可以看到页表 early_pg_dir有三个条目，分贝对应以下语句实现
一一来看
fixmap_pmd
create_pgd_mapping(early_pg_dir, FIXADDR_START,
(uintptr_t)fixmap_pgd_next, PGDIR_SIZE, PAGE_TABLE);
即配置pgd页表early_pg_dir
Va=FIXADDR_START=0xffffffcefee00000(这个地址上一篇文章已经分析过了)
Pa=fixmap_pgd_next即fixmap_pmd = 0x81d7a000
Sz=PGDIR_SIZE=1GB
Prot=PAGE_TABLE=1 下一级还是页表
先计算pgd_idx=(va>>30)&511=(0xffffffcefee00000>>30)&511=0x13b=315
即early_pg_dir中一个条目对应1GB，这里需要对应到第315个条目去(右移30位，即按照1GB的颗粒度)，并且表只有一页即512个条目，所以按照512取余。
所以对应条目为early_pg_dir[315]
然后计算条目的内容
pfn_pgd(PFN_DOWN(pa), prot);
其中pa为0x81d7a000，prot为1
所以PFN_DOWN(pa)为0x81d7a000>>12
所以值为
((0x81d7a000>>12 ) <<10)| 1=0x2075E801
和GDB打印的如下信息对应
(gdb) p /x *(pgd_t (*)[512])0x80a7b000$9 = {{pgd = 0x0}, {pgd = 0x2029e401}, {pgd = 0x0} <repeats 313 times>, {pgd = 0x2075e801}, {pgd = 0x0} <repeats 68 times>, {pgd = 0x2029e801}, {pgd = 0x0} <repeats 127 times>}(gdb)
early_dtb_pmd
create_pgd_mapping(early_pg_dir, DTB_EARLY_BASE_VA,
(uintptr_t)early_dtb_pmd, PGDIR_SIZE, PAGE_TABLE);
即配置pgd页表early_pg_dir
Va=DTB_EARLY_BASE_VA=0x40000000(#define DTB_EARLY_BASE_VA PGDIR_SIZE)
Pa=early_dtb_pmd= 0x80a79000
Sz=PGDIR_SIZE=1GB
Prot=PAGE_TABLE=1 下一级还是页表
先计算pgd_idx=(va>>30)&511=(0x40000000>>30)&511=1
即early_pg_dir中一个条目对应1GB，这里需要对应到第1个条目去(右移30位，即按照1GB的颗粒度)，并且表只有一页即512个条目，所以按照512取余。
所以对应条目为early_pg_dir[1]
然后计算条目的内容
pfn_pgd(PFN_DOWN(pa), prot);
其中pa为0x80a79000，prot为1
所以PFN_DOWN(pa)为0x80a79000>>12
所以值为
((0x80a79000>>12 ) <<10)| 1=0x2029E401
和GDB打印的如下信息对应
(gdb) p /x *(pgd_t (*)[512])0x80a7b000$9 = {{pgd = 0x0}, {pgd = 0x2029e401}, {pgd = 0x0} <repeats 313 times>, {pgd = 0x2075e801}, {pgd = 0x0} <repeats 68 times>, {pgd = 0x2029e801}, {pgd = 0x0} <repeats 127 times>}(gdb)
early_pmd
/** Setup early PGD covering entire kernel which will allows* us to reach paging_init(). We map all memory banks later* in setup_vm_final() below.*/end_va = PAGE_OFFSET + load_sz;for (va = PAGE_OFFSET; va < end_va; va += map_size)create_pgd_mapping(early_pg_dir, va,load_pa + (va - PAGE_OFFSET),map_size, PAGE_KERNEL_EXEC);
即配置pgd页表early_pg_dir
Va=PAGE_OFFSET开始按照map_size(前一篇文章分析的大小是2MB)为单位递进
Pa=load_pa 开始按照2MB递进
Sz=2MB
Prot=PAGE_KERNEL_EXEC 下一级叶子pte
这里希望直接映射
PAGE_OFFSET开始的虚拟地址到load_pa处，以2MB为单位映射整个镜像。
我们看create_pgd_mapping的实现
这里sz不为PGDIR_SIZE,且之前只映射了
DTB_EARLY_BASE_VA 0x40000000
FIXADDR_START 0xffffffcefee00000
所以此时映射0xffffffe0 00000000
按照1GB的颗粒度early_pg_dir中肯定是没有这个条目的，
所以会走下面红色框中代码，先在early_pg_dir中创建一个条目，而其
下一级是alloc_pgd_next
pt_ops.alloc_pmd(__va)即
pt_ops.alloc_pmd = alloc_pmd_early;
即
&early_pmd[pmd_num * PTRS_PER_PMD];
MAX_EARLY_MAPPING_SIZE 小于PGDIR_SIZE的话
early_pmd只有一个页大小
这里MAX_EARLY_MAPPING_SIZE是128M小于PGDIR_SIZE
所以
next_phys = alloc_pgd_next(va);pgdp[pgd_idx] = pfn_pgd(PFN_DOWN(next_phys), PAGE_TABLE);nextp = get_pgd_next_virt(next_phys);memset(nextp, 0, PAGE_SIZE);
执行后
next_phys=&pmd_t early_pmd = 0x80A7A000
Pgd_idx=(va>>30)&511=(0xffffffe000000000>>30)&511=0x180=384
pfn_pgd(PFN_DOWN(next_phys), prot);
其中pa为0x80a7a000，prot为PAGE_TABLE=1
所以PFN_DOWN(pa)为0x80a7a000>>12
所以值为
((0x80a7a000>>12 ) <<10)| 1=0x2029E801
所以
early_pg_dir[384]=0x2029E801
和如下打印对应
(gdb) p /x *(pgd_t (*)[512])0x80a7b000$9 = {{pgd = 0x0}, {pgd = 0x2029e401}, {pgd = 0x0} <repeats 313 times>, {pgd = 0x2075e801}, {pgd = 0x0} <repeats 68 times>, {pgd = 0x2029e801}, {pgd = 0x0} <repeats 127 times>}(gdb)
一级PGD效果
至此上面early_pg_dir下的3个条目，映射了3个GB块。
此时还只映射到下一级PMD
对应如下黄色部分的3块虚拟地址起点
后续还要继续构建下一级PMD到物理地址
2.2.2 二级pmd
在上述三个pmd页表下继续映射到最终的物理块。
分别对应以下语句
fixmap_pmd
(gdb) p &fixmap_pmd$10 = (pmd_t (*)[512]) 0xffffffe001b7a000 <fixmap_pmd>(gdb)(gdb) p /x *(pmd_t (*)[512])0x81d7a000$11 = {{pmd = 0x0} <repeats 503 times>, {pmd = 0x2075f001}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}}(gdb)
对应代码如下
/* Setup fixmap PMD */create_pmd_mapping(fixmap_pmd, FIXADDR_START,(uintptr_t)fixmap_pte, PMD_SIZE, PAGE_TABLE);
Va= FIXADDR_START=0xffffffcefee00000
Pa=fixmap_pte=0x81d7c000
Sz=2MB
Prot=PAGE_TABLE = 1表示下一级还是页表。
pmd_idx=(0xffffffcefee00000>>21)&511=0x1F7=503
pfn_pmd(PFN_DOWN(pa), prot);=((0x81d7c000>>12)<<10) | 1=2075F001
所以
fixmap_pmd[503]=0x2075F001
和打印以下对应
(gdb) p /x *(pmd_t (*)[512])0x81d7a000$11 = {{pmd = 0x0} <repeats 503 times>, {pmd = 0x2075f001}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}, {pmd = 0x0}}(gdb)
fixmap_pte
此时fixmap_pte还没有映射叶子pte物理地址。
(gdb) p &fixmap_pte$12 = (pte_t (*)[512]) 0xffffffe001b7c000 <fixmap_pte>(gdb)(gdb) p /x *(pte_t (*)[512])0x81d7c000$13 = {{pte = 0x0} <repeats 512 times>}(gdb)
页表对应如下
这一部分还未映射物理地址
early_dtb_pmd
(gdb) p &early_dtb_pmd$20 = (pmd_t (*)[512]) 0xffffffe000879000 <early_dtb_pmd>(gdb)(gdb) p /x *(pmd_t (*)[512])0x80a79000$23 = {{pmd = 0x20f800e7}, {pmd = 0x210000e7}, {pmd = 0x0} <repeats 510 times>}(gdb)
该部分映射2个2MB的设备树块。
对应代码如下
其中dtb_pa由参数传入，kernel启动时由寄存器r1传入
我这里是
gdbinit.txt中设置
(gdb) p /x $dtb_addr
$25 = 0x83f00000
(gdb)
这里先按2MB对齐
pa = dtb_pa & ~(PMD_SIZE - 1);
所以pa=0x83e00000
后面映射了两个连续的2MB的块，所以分析一个即可。
create_pmd_mapping(early_dtb_pmd, DTB_EARLY_BASE_VA,pa, PMD_SIZE, PAGE_KERNEL);
Va= DTB_EARLY_BASE_VA=0x40000000
Pa=0x83e00000
Sz=2MB
Prot=PAGE_KERNEL = 0xE7表示没有下一级，直接索引物理地址了。
先计算
pmd_idx = pmd_index(va);=(va>>21)&511, 即按照2MB块分配，索引对512取余。
pmd_idx=(0x40000000>>21)&511=0
pfn_pmd(PFN_DOWN(pa), prot);=((0x83e00000>>12)<<10) | 0xE7=0x20F800E7
所以
early_dtb_pmd[0]=0x20F800E7
和打印以下对应
(gdb) p /x *(pmd_t (*)[512])0x80a79000$23 = {{pmd = 0x20f800e7}, {pmd = 0x210000e7}, {pmd = 0x0} <repeats 510 times>}(gdb)
再来看
create_pmd_mapping(early_dtb_pmd, DTB_EARLY_BASE_VA + PMD_SIZE,pa + PMD_SIZE, PMD_SIZE, PAGE_KERNEL);
Va= DTB_EARLY_BASE_VA=0x40200000
Pa=0x84000000
Sz=2MB
Prot=PAGE_KERNEL = 0xE7表示没有下一级，直接索引物理地址了。
pmd_idx=(0x40200000>>21)&511=1
pfn_pmd(PFN_DOWN(pa), prot);=((0x84000000>>12)<<10) | 0xE7=0x210000E7
所以
early_dtb_pmd[1]=0x210000E7
和打印以下对应
(gdb) p /x *(pmd_t (*)[512])0x80a79000$23 = {{pmd = 0x20f800e7}, {pmd = 0x210000e7}, {pmd = 0x0} <repeats 510 times>}(gdb)
对应的映射如下
即将0x40000000开始的2个连续的2MB映射到了0x83e00000开始的两个连续的2MB。
采用大页2MB为单位映射。
early_pmd
(gdb) p &early_pmd$2 = (pmd_t (*)[512]) 0xffffffe00087a000 <early_pmd>(gdb) p /x * (pmd_t (*)[512])0x80A7A000$5 = {{pmd = 0x200800ef}, {pmd = 0x201000ef}, {pmd = 0x201800ef}, {pmd = 0x202000ef}, {pmd = 0x202800ef}, {pmd = 0x203000ef}, {pmd = 0x203800ef}, {pmd = 0x204000ef}, {pmd = 0x204800ef}, {pmd = 0x205000ef}, {pmd = 0x205800ef}, {pmd = 0x206000ef}, {pmd = 0x206800ef}, {pmd = 0x207000ef}, {pmd = 0x0} <repeats 498 times>}(gdb)
这里对应代码如下
/** Setup early PGD covering entire kernel which will allows* us to reach paging_init(). We map all memory banks later* in setup_vm_final() below.*/end_va = PAGE_OFFSET + load_sz;for (va = PAGE_OFFSET; va < end_va; va += map_size)create_pgd_mapping(early_pg_dir, va,load_pa + (va - PAGE_OFFSET),map_size, PAGE_KERNEL_EXEC);
这里是对kernel镜像大小，从
PAGE_OFFSET处0xffffffe000000000开始，按照2MB进行映射到load_pa即0x80200000处。
我们前面一篇分析了load_sz大小为29151232=27.8M，所以按照2MB为单位，要创建14个条目，从上面gdb的打印p /x * (pmd_t (*)[512])0x80A7A000也可看到下有14个条目。
我们以第一个为例分析，后面的类似。
这里create_pgd_mapping由于传入sz为2MB不是1GB
且之前已经创建了early_pg_dir的条目。所以这里会走如下红色框部分，即调用create_pmd_mapping
Va= PAGE_OFFSET=0xffffffe000000000
Pa=0x80200000
Sz=2MB
Prot=PAGE_KERNEL_EXEC = 0xEF表示没有下一级，直接索引物理地址了。
pmd_idx=(0xffffffe000000000>>21)&511=0
pfn_pmd(PFN_DOWN(pa), prot);=((0x80200000>>12)<<10) | 0xEF=200800EF
所以
early_pmd[0]=200800EF
和打印以下对应
(gdb) p /x * (pmd_t (*)[512])0x80A7A000$5 = {{pmd = 0x200800ef}, {pmd = 0x201000ef}, {pmd = 0x201800ef}, {pmd = 0x202000ef}, {pmd = 0x202800ef}, {pmd = 0x203000ef}, {pmd = 0x203800ef}, {pmd = 0x204000ef}, {pmd = 0x204800ef}, {pmd = 0x205000ef}, {pmd = 0x205800ef}, {pmd = 0x206000ef}, {pmd = 0x206800ef}, {pmd = 0x207000ef}, {pmd = 0x0} <repeats 498 times>}(gdb)
如果对应最后一个则是
Va= PAGE_OFFSET=0xffffffe000000000+13*2MB=0xffffffe001A00000
Pa=0x80200000+13x2MB=0x81C00000
Sz=2MB
Prot=PAGE_KERNEL_EXEC = 0xEF表示没有下一级，直接索引物理地址了。
pmd_idx=(0xffffffe001A00000>>21)&511=13
pfn_pmd(PFN_DOWN(pa), prot);=((0x81C00000>>12)<<10) | 0xEF=0x207000EF
所以
early_pmd[13]=0x207000EF
和打印以下对应
(gdb) p /x * (pmd_t (*)[512])0x80A7A000$5 = {{pmd = 0x200800ef}, {pmd = 0x201000ef}, {pmd = 0x201800ef}, {pmd = 0x202000ef}, {pmd = 0x202800ef}, {pmd = 0x203000ef}, {pmd = 0x203800ef}, {pmd = 0x204000ef}, {pmd = 0x204800ef}, {pmd = 0x205000ef}, {pmd = 0x205800ef}, {pmd = 0x206000ef}, {pmd = 0x206800ef}, {pmd = 0x207000ef}, {pmd = 0x0} <repeats 498 times>}(gdb)
映射完后对应如下
2.3 swapper_pg_dir
(gdb) p &swapper_pg_dir$18 = (pgd_t (*)[512]) 0xffffffe001b7e000 <swapper_pg_dir>(gdb)(gdb) p /x * (pgd_t (*)[512]) 0x81D7E000$19 = {{pgd = 0x0} <repeats 512 times>}(gdb)
swapper_pg_dir在setup_vm_final中设置,此时还未设置。
swapper_pg_dir下一篇再分析。
2.4 .对fixmap_pmd的检查
setup_vm函数分析时，还有如下部分未分析，我们来继续分析下
/** Bootime fixmap only can handle PMD_SIZE mapping. Thus, boot-ioremap* range can not span multiple pmds.*/BUILD_BUG_ON((__fix_to_virt(FIX_BTMAP_BEGIN) >> PMD_SHIFT)!= (__fix_to_virt(FIX_BTMAP_END) >> PMD_SHIFT));/** Early ioremap fixmap is already created as it lies within first 2MB* of fixmap region. We always map PMD_SIZE. Thus, both FIX_BTMAP_END* FIX_BTMAP_BEGIN should lie in the same pmd. Verify that and warn* the user if not.*/fix_bmap_spmd = fixmap_pmd[pmd_index(__fix_to_virt(FIX_BTMAP_BEGIN))];fix_bmap_epmd = fixmap_pmd[pmd_index(__fix_to_virt(FIX_BTMAP_END))];if (pmd_val(fix_bmap_spmd) != pmd_val(fix_bmap_epmd)) {WARN_ON(1);pr_warn("fixmap btmap start [%08lx] != end [%08lx]\n",pmd_val(fix_bmap_spmd), pmd_val(fix_bmap_epmd));pr_warn("fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n",fix_to_virt(FIX_BTMAP_BEGIN));pr_warn("fix_to_virt(FIX_BTMAP_END): %08lx\n",fix_to_virt(FIX_BTMAP_END));pr_warn("FIX_BTMAP_END: %d\n", FIX_BTMAP_END);pr_warn("FIX_BTMAP_BEGIN: %d\n", FIX_BTMAP_BEGIN);}
首先
arch/riscv/include/asm/fixmap.h中
计算
__fix_to_virt(FIX_BTMAP_BEGIN)和__fix_to_virt(FIX_BTMAP_END)要在一个2MB内。
其中
FIX_BTMAP_BEGIN和(FIX_BTMAP_END计算如下
/** Here we define all the compile-time 'special' virtual addresses.* The point is to have a constant address at compile time, but to* set the physical address only in the boot process.** These 'compile-time allocated' memory buffers are page-sized. Use* set_fixmap(idx,phys) to associate physical memory with fixmap indices.*/enum fixed_addresses {FIX_HOLE,FIX_PTE,FIX_PMD,FIX_TEXT_POKE1,FIX_TEXT_POKE0,FIX_EARLYCON_MEM_BASE,__end_of_permanent_fixed_addresses,/** Temporary boot-time mappings, used by early_ioremap(),* before ioremap() is functional.*/FIX_BTMAP_END = __end_of_permanent_fixed_addresses,FIX_BTMAP_BEGIN = FIX_BTMAP_END + TOTAL_FIX_BTMAPS - 1,__end_of_fixed_addresses};
得到FIX_BTMAP_END=6
FIX_BTMAP_END_BEGIN=6+7*64-1=453
__fix_to_virt
include/asm-generic/fixmap.h中
#define __fix_to_virt(x) (FIXADDR_TOP - ((x) << PAGE_SHIFT))
#define __virt_to_fix(x) ((FIXADDR_TOP - ((x)&PAGE_MASK)) >> PAGE_SHIFT)
即先<<PAGE_SHIFT即成衣4K
然后计算FIXADDR_TOP减去该值，得到BEGIN和END处的虚拟地址。
FIXADDR_TOP在arch/riscv/include/asm/pgtable.h中定义
对应如下空间，即如下7*64个4k空间要在同一个2MB内
然后pmd_index虚拟地址转pmd_idx
fix_bmap_spmd = fixmap_pmd[pmd_index(__fix_to_virt(FIX_BTMAP_BEGIN))];fix_bmap_epmd = fixmap_pmd[pmd_index(__fix_to_virt(FIX_BTMAP_END))];
找到该idx对应的fixmap_pmd条目
对应我们前面分析的这个索引503
然后判断对应的pmd条目内容要一样，我们这里是在同一个2MB内是一样的。
if (pmd_val(fix_bmap_spmd) != pmd_val(fix_bmap_epmd)) {WARN_ON(1);pr_warn("fixmap btmap start [%08lx] != end [%08lx]\n",pmd_val(fix_bmap_spmd), pmd_val(fix_bmap_epmd));pr_warn("fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n",fix_to_virt(FIX_BTMAP_BEGIN));pr_warn("fix_to_virt(FIX_BTMAP_END): %08lx\n",fix_to_virt(FIX_BTMAP_END));pr_warn("FIX_BTMAP_END: %d\n", FIX_BTMAP_END);pr_warn("FIX_BTMAP_BEGIN: %d\n", FIX_BTMAP_BEGIN);}
三. 总结
setup_vm后，总的页表映射关系如下
两个根页表在relocate中使用，其中先临时使用trampoline_pg_dir跳转下，再使用
early_pg_dir。 这两个页表都映射了PAGE_OFFSET:0xffffffe000000000到load_pa:0x80200000，都包括了relocate处的代码，所以才能无缝切换。
前者只映射了2MB，后者映射了整个Image的大小，都是按照大页2MB为单位映射。
此时的虚拟地址和物理地址对应如下
配置了
CONFIG_MMU和CONFIG_DEBUG_VM
函数
print_vm_layout可打印上述信息
