Hi Prabhakar,
Thanks for your feedback!
-----Original Message----- From: Prabhakar Kushwaha Sent: 2016年1月19日 21:54 To: Zhiqiang Hou Zhiqiang.Hou@freescale.com; u-boot@lists.denx.de; albert.u.boot@aribaud.net; Mingkai.hu@freescale.com; yorksun@freescale.com Cc: leoli@freescale.com; prabhakar@freescale.com; bhupesh.sharma@freescale.com; sjg@chromium.org; bmeng.cn@gmail.com; hs@denx.de; joe.hershberger@ni.com; marex@denx.de; Zhiqiang Hou zhiqiang.hou@nxp.com; Hou Zhiqiang B48286@freescale.com Subject: RE: [PATCH 1/3] ARMv8/layerscape: Add mmu_init API
-----Original Message----- From: Zhiqiang Hou [mailto:Zhiqiang.Hou@freescale.com] Sent: Tuesday, January 19, 2016 6:10 PM To: u-boot@lists.denx.de; albert.u.boot@aribaud.net; Mingkai.hu@freescale.com; yorksun@freescale.com Cc: leoli@freescale.com; prabhakar@freescale.com; bhupesh.sharma@freescale.com; sjg@chromium.org; bmeng.cn@gmail.com; hs@denx.de; joe.hershberger@ni.com; marex@denx.de; Zhiqiang Hou zhiqiang.hou@nxp.com; Hou Zhiqiang B48286@freescale.com Subject: [PATCH 1/3] ARMv8/layerscape: Add mmu_init API
From: Hou Zhiqiang B48286@freescale.com
Expose this API to make it reuseable when u-boot turn into other EL from EL3.
Signed-off-by: Hou Zhiqiang B48286@freescale.com
arch/arm/cpu/armv8/fsl-layerscape/cpu.c | 24 ++++++++++++++++++++++++ include/common.h | 1 + 2 files changed, 25 insertions(+)
diff --git a/arch/arm/cpu/armv8/fsl-layerscape/cpu.c b/arch/arm/cpu/armv8/fsl-layerscape/cpu.c index 6ea28ed..df5670f 100644 --- a/arch/arm/cpu/armv8/fsl-layerscape/cpu.c +++ b/arch/arm/cpu/armv8/fsl-layerscape/cpu.c @@ -403,6 +403,30 @@ void enable_caches(void) final_mmu_setup(); __asm_invalidate_tlb_all(); }
+static void mmu_disable(void) +{
- if (get_sctlr() & CR_M)
set_sctlr(get_sctlr() & ~CR_M);+}
+static void mmu_enable(void) +{
- if (!(get_sctlr() & CR_M))
set_sctlr(get_sctlr() | CR_M);+}
+void mmu_init(void)
Name of function is not mapping about what it is doing. This function assume MMU is already enabled with early_table and it will setup final_mmu_setup.
This function map basically with what it is doing except some I-cache operations that will be remove from this func, and not assume the early_mmu_setup. It assumes the relocation has been done.
+{
- mmu_disable();
- dcache_disable();
- icache_disable();
- final_mmu_setup();
- __asm_invalidate_tlb_all();
- mmu_enable();
- icache_enable();
- set_sctlr(get_sctlr() | CR_C);
+} #endif
If I am correct board_init_r deploy final_mmu_setup via enable_caches(). Why cannot existing framework be used.
This patch aims to make final_mmu_setup() can be called flexibly in case as When it return from PPA and execute at EL2, the MMU must be initialized for the current EL. Yes, the enable_caches() will call the __weak__ mmu_setup() if MMU wasn't enabled. But for fsl layerscape platforms, the MMU has been enabled by early mmu setup, so the __weak__ mmu_setup won't be called.
I am not know cache and mmu so much, and have some question: For ARM: Why there isn't a isolate API for mmu_setup, but invoke it from dcache_enable()? If data cache won't be used, the MMU also cannot be used?
static inline u32 initiator_type(u32 cluster, int init_id) diff --git a/include/common.h b/include/common.h index 75c78d5..57a9b30 100644 --- a/include/common.h +++ b/include/common.h @@ -757,6 +757,7 @@ void flush_dcache_range(unsigned long start, unsigned long stop); void invalidate_dcache_range(unsigned long start, unsigned long stop); void invalidate_dcache_all(void); void invalidate_icache_all(void); +void mmu_init(void);
enum { /* Disable caches (else flush caches but leave them active) */ -- 2.1.0.27.g96db324
Thanks, Zhiqiang