🎯 [libcpu/arm/cortex-a]修改arm/cortex-a启动方式，解决cortex-a开启SMP工作异常问题 #8517
Conversation
|
Good Job! |
这里也帮忙做个调整吧，原来的代码有些不太规范 |
已修改 rt_mpr_start = (void *)((rt_size_t)0 - rt_mpr_size); |
mysterywolf
added
the
v5.1.0
BernardXiong
changed the title
|
你好 这份代码先挂在这里 等#8537 合并之后再麻烦您重新rebase和测试一下~ |
好的，这份代码和刚才合入的有一些冲突，我只是重新修改提交一下 |
|
#8537 已经合并 麻烦重新rebase到最新基线并测试一下把~ |
bsp/phytium/board/board.c
Outdated
| @@ -189,27 +189,28 @@ void rt_hw_board_aarch64_init(void) | |||
| } | |||
| #else | |||
|
|
|||
|
|
|||
| extern void rt_hw_mmu_init(void) ; | |||
There was a problem hiding this comment.
最好是放到头文件里面。否则后续原型修改很容易不一致，而且发现不了。
| @@ -64,12 +64,16 @@ void idle_wfi(void) | |||
| asm volatile ("wfi"); | |||
| } | |||
|
|
|||
|
|
|||
| extern void rt_hw_mmu_init(void); | |||
libcpu/arm/cortex-a/start_gcc.S
Outdated
| wfe | ||
| b _halt | ||
| .weak rt_asm_cpu_id | ||
| rt_asm_cpu_id: |
libcpu/arm/cortex-a/start_gcc.S
Outdated
| .macro get_phy, reg, symbol | ||
| ldr \reg, =\symbol | ||
| add \reg, pv_off |
There was a problem hiding this comment.
这个 pv_off 最好是个参数，否则用这个宏的时候，就依赖于 pv_off 这个寄存器的初始化。后面加代码，或者维护起来容易疏忽。
| .macro get_phy, reg, symbol | |
| ldr \reg, =\symbol | |
| add \reg, pv_off | |
| .macro get_phy, reg, symbol, _pvoff | |
| ldr \reg, =\symbol | |
| add \reg, \_pvoff |
There was a problem hiding this comment.
.macro get_phy, reg, symbol, _pvoff
ldr \reg, =\symbol
add \reg, _pvoff
.endm
/Calculate the offset between the physical address and the virtual address of the "_reset"./
.macro get_pvoff, tmp, out
ldr \tmp, =_reset
adr \out, _reset
sub \out, \out, \tmp
.endm
libcpu/arm/cortex-a/start_gcc.S
Outdated
| ldr r0, =_reset | ||
| adr pv_off, _reset | ||
| sub pv_off, pv_off, r0 |
There was a problem hiding this comment.
如果修改了，这个就可以用 get_pvoff 宏替换。
libcpu/arm/cortex-a/start_gcc.S
Outdated
| .macro get_pvoff, tmp, out | ||
| ldr \tmp, =_reset | ||
| get_phy \out, _reset | ||
| sub \out, \out, \tmp | ||
| .endm |
There was a problem hiding this comment.
而且，按照现在的参数设计， get_pvoff 依赖了 get_phy，get_phy 依赖了 pv_off。虽然 pv_off 可以单独先初始化，结果上也对。但是逻辑上很奇怪。
There was a problem hiding this comment.
.macro get_pvoff, tmp, out
ldr \tmp, =_reset
adr \out, _reset
sub \out, \out, \tmp
.endm
libcpu/arm/cortex-a/start_gcc.S
Outdated
| get_phy r0, init_mtbl | ||
| get_pvoff r2,r1 |
There was a problem hiding this comment.
这里就是 get_phy 的参数设计很容易造成混乱的例子。get_phy 依赖于 pvoff 的，但是这里又先计算了 get_phy，然后计算一个新的 pvoff。看的时候就很别扭。
如果给 get_phy 加入 pvoff 作为第三个参数，修改后个人感觉是更清晰可读的。
| get_phy r0, init_mtbl | |
| get_pvoff r2,r1 | |
| get_phy r0, init_mtbl, pv_off | |
| mov r1, pv_off |
拉取/合并请求描述：(PR description)
[
为什么提交这份PR (why to submit this PR)
#8469 解决issue
你的解决方案是什么 (what is your solution)
对cortex-a cpu的启动步骤仿照aarch64的风格进行了整理，使逻辑更加清晰，并解决了上述问题。已在qemu-vexpress-a9，raspi4-32，phytium开发板上进行测试，目前可以正常启动
请提供验证的bsp和config (provide the config and bsp)
]
当前拉取/合并请求的状态 Intent for your PR
必须选择一项 Choose one (Mandatory):
代码质量 Code Quality：
我在这个拉取/合并请求中已经考虑了 As part of this pull request, I've considered the following:
#if 0代码，不包含已经被注释了的代码 All redundant code is removed and cleaned up