xtensa: cleanup MMU setup and kernel layout macros

Make kernel load address explicit, independent of the selected MMU
configuration and configurable from Kconfig. Do not restrict it to the
first 512MB of the physical address space.

Cleanup kernel memory layout macros:

- rename VECBASE_RESET_VADDR to VECBASE_VADDR, XC_VADDR to VECTOR_VADDR;
- drop VIRTUAL_MEMORY_ADDRESS and LOAD_MEMORY_ADDRESS;
- introduce PHYS_OFFSET and use it in __va and __pa definitions;
- synchronize MMU/noMMU vectors, drop unused NMI vector;
- replace hardcoded vectors offset of 0x3000 with Kconfig symbol.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

Documentation/xtensa/mmu.txt

@@ -3,15 +3,8 @@ MMUv3 initialization sequence.
 The code in the initialize_mmu macro sets up MMUv3 memory mapping
 identically to MMUv2 fixed memory mapping. Depending on
 CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX symbol this code is
-located in one of the following address ranges:
+located in addresses it was linked for (symbol undefined), or not
-
+(symbol defined), so it needs to be position-independent.
-    0xF0000000..0xFFFFFFFF (will keep same address in MMU v2 layout;
-    			 typically ROM)
-    0x00000000..0x07FFFFFF (system RAM; this code is actually linked
-    			 at 0xD0000000..0xD7FFFFFF [cached]
-    			 or 0xD8000000..0xDFFFFFFF [uncached];
-    			 in any case, initially runs elsewhere
-    			 than linked, so have to be careful)
 
 The code has the following assumptions:
   This code fragment is run only on an MMU v3.
@@ -28,24 +21,26 @@ TLB setup proceeds along the following steps.
     PA = physical address (two upper nibbles of it);
     pc = physical range that contains this code;
 
-After step 2, we jump to virtual address in 0x40000000..0x5fffffff
+After step 2, we jump to virtual address in the range 0x40000000..0x5fffffff
-that corresponds to next instruction to execute in this code.
+or 0x00000000..0x1fffffff, depending on whether the kernel was loaded below
-After step 4, we jump to intended (linked) address of this code.
+0x40000000 or above. That address corresponds to next instruction to execute
+in this code. After step 4, we jump to intended (linked) address of this code.
+The scheme below assumes that the kernel is loaded below 0x40000000.
 
         Step0  Step1  Step2  Step3          Step4  Step5
- ============  =====  ============  =====  ============  =====
+        =====  =====  =====  =====          =====  =====
-   VA      PA     PA    VA      PA     PA    VA      PA     PA
+   VA      PA     PA     PA     PA     VA      PA     PA
- ------    --     --  ------    --     --  ------    --     --
+ ------    --     --     --     --   ------    --     --
- E0..FF -> E0  -> E0  E0..FF -> E0         F0..FF -> F0  -> F0
+ E0..FF -> E0  -> E0  -> E0          F0..FF -> F0  -> F0
- C0..DF -> C0  -> C0  C0..DF -> C0         E0..EF -> F0  -> F0
+ C0..DF -> C0  -> C0  -> C0          E0..EF -> F0  -> F0
- A0..BF -> A0  -> A0  A0..BF -> A0         D8..DF -> 00  -> 00
+ A0..BF -> A0  -> A0  -> A0          D8..DF -> 00  -> 00
- 80..9F -> 80  -> 80  80..9F -> 80         D0..D7 -> 00  -> 00
+ 80..9F -> 80  -> 80  -> 80          D0..D7 -> 00  -> 00
- 60..7F -> 60  -> 60  60..7F -> 60
+ 60..7F -> 60  -> 60  -> 60
- 40..5F -> 40         40..5F -> pc  -> pc  40..5F -> pc
+ 40..5F -> 40         -> pc  -> pc   40..5F -> pc
- 20..3F -> 20  -> 20  20..3F -> 20
+ 20..3F -> 20  -> 20  -> 20
- 00..1F -> 00  -> 00  00..1F -> 00
+ 00..1F -> 00  -> 00  -> 00
 
-The default location of IO peripherals is above 0xf0000000. This may change
+The default location of IO peripherals is above 0xf0000000. This may be changed
 using a "ranges" property in a device tree simple-bus node. See ePAPR 1.1, §6.5
 for details on the syntax and semantic of simple-bus nodes. The following
 limitations apply:

arch/xtensa/Kconfig

@@ -249,6 +249,25 @@ config KSEG_PADDR
 
 	  If unsure, leave the default value here.
 
+config KERNEL_LOAD_ADDRESS
+	hex "Kernel load address"
+	default 0x00003000
+	help
+	  This is the address where the kernel is loaded.
+	  It is virtual address for MMUv2 configurations and physical address
+	  for all other configurations.
+
+	  If unsure, leave the default value here.
+
+config VECTORS_OFFSET
+	hex "Kernel vectors offset"
+	default 0x00003000
+	help
+	  This is the offset of the kernel image from the relocatable vectors
+	  base.
+
+	  If unsure, leave the default value here.
+
 choice
 	prompt "KSEG layout"
 	depends on MMU
@@ -487,12 +506,7 @@ config DEFAULT_MEM_START
 	  used when no physical memory size is passed through DTB or through
 	  boot parameter from bootloader.
 
-	  In noMMU configuration the following parameters are derived from it:
+	  It's also used for TASK_SIZE calculation in noMMU configuration.
-	  - kernel load address;
-	  - kernel entry point address;
-	  - relocatable vectors base address;
-	  - uBoot load address;
-	  - TASK_SIZE.
 
 	  If unsure, leave the default value here.
 

arch/xtensa/boot/boot-elf/boot.lds.S

@@ -23,7 +23,7 @@ SECTIONS
 		*(.ResetVector.text)
 	}
 
-	.image KERNELOFFSET: AT (LOAD_MEMORY_ADDRESS)
+	.image KERNELOFFSET: AT (CONFIG_KERNEL_LOAD_ADDRESS)
 	{
 		_image_start = .;
 		*(image)

arch/xtensa/boot/boot-elf/bootstrap.S

@@ -35,7 +35,12 @@ _ResetVector:
 
 	.align 4
 RomInitAddr:
-	.word	LOAD_MEMORY_ADDRESS
+#if defined(CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX) && \
+	XCHAL_HAVE_PTP_MMU && XCHAL_HAVE_SPANNING_WAY
+	.word	CONFIG_KERNEL_LOAD_ADDRESS
+#else
+	.word	KERNELOFFSET
+#endif
 RomBootParam:
 	.word _bootparam
 _bootparam:

arch/xtensa/boot/boot-uboot/Makefile

@@ -4,15 +4,7 @@
 # for more details.
 #
 
-ifdef CONFIG_MMU
+UIMAGE_LOADADDR = $(CONFIG_KERNEL_LOAD_ADDRESS)
-ifdef CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX
-UIMAGE_LOADADDR = 0x00003000
-else
-UIMAGE_LOADADDR = 0xd0003000
-endif
-else
-UIMAGE_LOADADDR = $(shell printf "0x%x" $$(( ${CONFIG_DEFAULT_MEM_START} + 0x3000 )) )
-endif
 UIMAGE_COMPRESSION = gzip
 
 $(obj)/../uImage: vmlinux.bin.gz FORCE

arch/xtensa/include/asm/initialize_mmu.h

@@ -77,13 +77,16 @@
 
 	.align	4
 1:	movi	a2, 0x10000000
-	movi	a3, 0x18000000
+
-	add	a2, a2, a0
+#if CONFIG_KERNEL_LOAD_ADDRESS < 0x40000000ul
-9:	bgeu	a2, a3, 9b	/* PC is out of the expected range */
+#define TEMP_MAPPING_VADDR 0x40000000
+#else
+#define TEMP_MAPPING_VADDR 0x00000000
+#endif
 
 	/* Step 1: invalidate mapping at 0x40000000..0x5FFFFFFF. */
 
-	movi	a2, 0x40000000 | XCHAL_SPANNING_WAY
+	movi	a2, TEMP_MAPPING_VADDR | XCHAL_SPANNING_WAY
 	idtlb	a2
 	iitlb	a2
 	isync
@@ -95,14 +98,14 @@
 	srli	a3, a0, 27
 	slli	a3, a3, 27
 	addi	a3, a3, CA_BYPASS
-	addi	a7, a2, -1
+	addi	a7, a2, 5 - XCHAL_SPANNING_WAY
 	wdtlb	a3, a7
 	witlb	a3, a7
 	isync
 
 	slli	a4, a0, 5
 	srli	a4, a4, 5
-	addi	a5, a2, -6
+	addi	a5, a2, -XCHAL_SPANNING_WAY
 	add	a4, a4, a5
 	jx	a4
 
@@ -145,19 +148,19 @@
 	witlb	a4, a5
 #endif
 
-	movi	a5, XCHAL_KIO_CACHED_VADDR + 6
+	movi	a5, XCHAL_KIO_CACHED_VADDR + XCHAL_KIO_TLB_WAY
 	movi	a4, XCHAL_KIO_DEFAULT_PADDR + CA_WRITEBACK
 	wdtlb	a4, a5
 	witlb	a4, a5
 
-	movi	a5, XCHAL_KIO_BYPASS_VADDR + 6
+	movi	a5, XCHAL_KIO_BYPASS_VADDR + XCHAL_KIO_TLB_WAY
 	movi	a4, XCHAL_KIO_DEFAULT_PADDR + CA_BYPASS
 	wdtlb	a4, a5
 	witlb	a4, a5
 
 	isync
 
-	/* Jump to self, using MMU v2 mappings. */
+	/* Jump to self, using final mappings. */
 	movi	a4, 1f
 	jx	a4
 

arch/xtensa/include/asm/kmem_layout.h

@@ -29,6 +29,7 @@
 #define XCHAL_KSEG_SIZE		__XTENSA_UL_CONST(0x08000000)
 #define XCHAL_KSEG_ALIGNMENT	__XTENSA_UL_CONST(0x08000000)
 #define XCHAL_KSEG_TLB_WAY	5
+#define XCHAL_KIO_TLB_WAY	6
 
 #elif defined(CONFIG_XTENSA_KSEG_256M)
 
@@ -37,6 +38,7 @@
 #define XCHAL_KSEG_SIZE		__XTENSA_UL_CONST(0x10000000)
 #define XCHAL_KSEG_ALIGNMENT	__XTENSA_UL_CONST(0x10000000)
 #define XCHAL_KSEG_TLB_WAY	6
+#define XCHAL_KIO_TLB_WAY	6
 
 #elif defined(CONFIG_XTENSA_KSEG_512M)
 
@@ -45,6 +47,7 @@
 #define XCHAL_KSEG_SIZE		__XTENSA_UL_CONST(0x20000000)
 #define XCHAL_KSEG_ALIGNMENT	__XTENSA_UL_CONST(0x10000000)
 #define XCHAL_KSEG_TLB_WAY	6
+#define XCHAL_KIO_TLB_WAY	6
 
 #else
 #error Unsupported KSEG configuration

arch/xtensa/include/asm/page.h

@@ -27,10 +27,12 @@
 
 #ifdef CONFIG_MMU
 #define PAGE_OFFSET	XCHAL_KSEG_CACHED_VADDR
+#define PHYS_OFFSET	XCHAL_KSEG_PADDR
 #define MAX_LOW_PFN	(PHYS_PFN(XCHAL_KSEG_PADDR) + \
 			 PHYS_PFN(XCHAL_KSEG_SIZE))
 #else
 #define PAGE_OFFSET	__XTENSA_UL_CONST(0)
+#define PHYS_OFFSET	__XTENSA_UL_CONST(0)
 #define MAX_LOW_PFN	(PHYS_PFN(PLATFORM_DEFAULT_MEM_START) + \
 			 PHYS_PFN(PLATFORM_DEFAULT_MEM_SIZE))
 #endif
@@ -163,8 +165,10 @@ void copy_user_highpage(struct page *to, struct page *from,
 
 #define ARCH_PFN_OFFSET		(PLATFORM_DEFAULT_MEM_START >> PAGE_SHIFT)
 
-#define __pa(x)			((unsigned long) (x) - PAGE_OFFSET)
+#define __pa(x)	\
-#define __va(x)			((void *)((unsigned long) (x) + PAGE_OFFSET))
+	((unsigned long) (x) - PAGE_OFFSET + PHYS_OFFSET)
+#define __va(x)	\
+	((void *)((unsigned long) (x) - PHYS_OFFSET + PAGE_OFFSET))
 #define pfn_valid(pfn) \
 	((pfn) >= ARCH_PFN_OFFSET && ((pfn) - ARCH_PFN_OFFSET) < max_mapnr)
 

arch/xtensa/include/asm/vectors.h

@@ -48,61 +48,42 @@ static inline unsigned long xtensa_get_kio_paddr(void)
 
 #if defined(CONFIG_MMU)
 
-/* Will Become VECBASE */
+#if XCHAL_HAVE_PTP_MMU && XCHAL_HAVE_SPANNING_WAY
-#define VIRTUAL_MEMORY_ADDRESS		XCHAL_KSEG_CACHED_VADDR
-
 /* Image Virtual Start Address */
-#define KERNELOFFSET			(XCHAL_KSEG_CACHED_VADDR + 0x3000)
+#define KERNELOFFSET			(XCHAL_KSEG_CACHED_VADDR + \
-
+					 CONFIG_KERNEL_LOAD_ADDRESS - \
-#if defined(XCHAL_HAVE_PTP_MMU) && XCHAL_HAVE_PTP_MMU && XCHAL_HAVE_SPANNING_WAY
+					 XCHAL_KSEG_PADDR)
-  /* MMU v3  - XCHAL_HAVE_PTP_MMU  == 1 */
-  #define LOAD_MEMORY_ADDRESS		0x00003000
 #else
-  /* MMU V2 -  XCHAL_HAVE_PTP_MMU  == 0 */
+#define KERNELOFFSET			CONFIG_KERNEL_LOAD_ADDRESS
-  #define LOAD_MEMORY_ADDRESS		0xD0003000
 #endif
 
-#define RESET_VECTOR1_VADDR		(VIRTUAL_MEMORY_ADDRESS + \
-					 XCHAL_RESET_VECTOR1_PADDR)
-
 #else /* !defined(CONFIG_MMU) */
   /* MMU Not being used - Virtual == Physical */
 
-  /* VECBASE */
-  #define VIRTUAL_MEMORY_ADDRESS	(PLATFORM_DEFAULT_MEM_START + 0x2000)
-
 /* Location of the start of the kernel text, _start */
-  #define KERNELOFFSET			(PLATFORM_DEFAULT_MEM_START + 0x3000)
+#define KERNELOFFSET			CONFIG_KERNEL_LOAD_ADDRESS
 
-  /* Loaded just above possibly live vectors */
-  #define LOAD_MEMORY_ADDRESS		(PLATFORM_DEFAULT_MEM_START + 0x3000)
-
-#define RESET_VECTOR1_VADDR		(XCHAL_RESET_VECTOR1_VADDR)
 
 #endif /* CONFIG_MMU */
 
-#define XC_VADDR(offset)		(VIRTUAL_MEMORY_ADDRESS  + offset)
+#define RESET_VECTOR1_VADDR		(XCHAL_RESET_VECTOR1_VADDR)
-
+#define VECBASE_VADDR			(KERNELOFFSET - CONFIG_VECTORS_OFFSET)
-/* Used to set VECBASE register */
-#define VECBASE_RESET_VADDR		VIRTUAL_MEMORY_ADDRESS
 
 #if defined(XCHAL_HAVE_VECBASE) && XCHAL_HAVE_VECBASE
 
-#define USER_VECTOR_VADDR		XC_VADDR(XCHAL_USER_VECOFS)
+#define VECTOR_VADDR(offset)		(VECBASE_VADDR + offset)
-#define KERNEL_VECTOR_VADDR		XC_VADDR(XCHAL_KERNEL_VECOFS)
-#define DOUBLEEXC_VECTOR_VADDR		XC_VADDR(XCHAL_DOUBLEEXC_VECOFS)
-#define WINDOW_VECTORS_VADDR		XC_VADDR(XCHAL_WINDOW_OF4_VECOFS)
-#define INTLEVEL2_VECTOR_VADDR		XC_VADDR(XCHAL_INTLEVEL2_VECOFS)
-#define INTLEVEL3_VECTOR_VADDR		XC_VADDR(XCHAL_INTLEVEL3_VECOFS)
-#define INTLEVEL4_VECTOR_VADDR		XC_VADDR(XCHAL_INTLEVEL4_VECOFS)
-#define INTLEVEL5_VECTOR_VADDR		XC_VADDR(XCHAL_INTLEVEL5_VECOFS)
-#define INTLEVEL6_VECTOR_VADDR		XC_VADDR(XCHAL_INTLEVEL6_VECOFS)
 
-#define DEBUG_VECTOR_VADDR		XC_VADDR(XCHAL_DEBUG_VECOFS)
+#define USER_VECTOR_VADDR		VECTOR_VADDR(XCHAL_USER_VECOFS)
-
+#define KERNEL_VECTOR_VADDR		VECTOR_VADDR(XCHAL_KERNEL_VECOFS)
-#define NMI_VECTOR_VADDR		XC_VADDR(XCHAL_NMI_VECOFS)
+#define DOUBLEEXC_VECTOR_VADDR		VECTOR_VADDR(XCHAL_DOUBLEEXC_VECOFS)
-
+#define WINDOW_VECTORS_VADDR		VECTOR_VADDR(XCHAL_WINDOW_OF4_VECOFS)
-#define INTLEVEL7_VECTOR_VADDR		XC_VADDR(XCHAL_INTLEVEL7_VECOFS)
+#define INTLEVEL2_VECTOR_VADDR		VECTOR_VADDR(XCHAL_INTLEVEL2_VECOFS)
+#define INTLEVEL3_VECTOR_VADDR		VECTOR_VADDR(XCHAL_INTLEVEL3_VECOFS)
+#define INTLEVEL4_VECTOR_VADDR		VECTOR_VADDR(XCHAL_INTLEVEL4_VECOFS)
+#define INTLEVEL5_VECTOR_VADDR		VECTOR_VADDR(XCHAL_INTLEVEL5_VECOFS)
+#define INTLEVEL6_VECTOR_VADDR		VECTOR_VADDR(XCHAL_INTLEVEL6_VECOFS)
+#define INTLEVEL7_VECTOR_VADDR		VECTOR_VADDR(XCHAL_INTLEVEL7_VECOFS)
+#define DEBUG_VECTOR_VADDR		VECTOR_VADDR(XCHAL_DEBUG_VECOFS)
 
 /*
  * These XCHAL_* #defines from varian/core.h
@@ -110,7 +91,6 @@ static inline unsigned long xtensa_get_kio_paddr(void)
  * constants are defined above and should be used.
  */
 #undef  XCHAL_VECBASE_RESET_VADDR
-#undef  XCHAL_RESET_VECTOR0_VADDR
 #undef  XCHAL_USER_VECTOR_VADDR
 #undef  XCHAL_KERNEL_VECTOR_VADDR
 #undef  XCHAL_DOUBLEEXC_VECTOR_VADDR
@@ -120,9 +100,8 @@ static inline unsigned long xtensa_get_kio_paddr(void)
 #undef  XCHAL_INTLEVEL4_VECTOR_VADDR
 #undef  XCHAL_INTLEVEL5_VECTOR_VADDR
 #undef  XCHAL_INTLEVEL6_VECTOR_VADDR
-#undef  XCHAL_DEBUG_VECTOR_VADDR
-#undef  XCHAL_NMI_VECTOR_VADDR
 #undef  XCHAL_INTLEVEL7_VECTOR_VADDR
+#undef  XCHAL_DEBUG_VECTOR_VADDR
 
 #else
 
@@ -135,6 +114,7 @@ static inline unsigned long xtensa_get_kio_paddr(void)
 #define INTLEVEL4_VECTOR_VADDR		XCHAL_INTLEVEL4_VECTOR_VADDR
 #define INTLEVEL5_VECTOR_VADDR		XCHAL_INTLEVEL5_VECTOR_VADDR
 #define INTLEVEL6_VECTOR_VADDR		XCHAL_INTLEVEL6_VECTOR_VADDR
+#define INTLEVEL7_VECTOR_VADDR		XCHAL_INTLEVEL6_VECTOR_VADDR
 #define DEBUG_VECTOR_VADDR		XCHAL_DEBUG_VECTOR_VADDR
 
 #endif

arch/xtensa/kernel/entry.S

@@ -1632,10 +1632,11 @@ ENTRY(fast_second_level_miss)
 	 * The messy computation for 'pteval' above really simplifies
 	 * into the following:
 	 *
-	 * pteval = ((pmdval - PAGE_OFFSET) & PAGE_MASK) | PAGE_DIRECTORY
+	 * pteval = ((pmdval - PAGE_OFFSET + PHYS_OFFSET) & PAGE_MASK)
+	 *                 | PAGE_DIRECTORY
 	 */
 
-	movi	a1, (-PAGE_OFFSET) & 0xffffffff
+	movi	a1, (PHYS_OFFSET - PAGE_OFFSET) & 0xffffffff
 	add	a0, a0, a1		# pmdval - PAGE_OFFSET
 	extui	a1, a0, 0, PAGE_SHIFT	# ... & PAGE_MASK
 	xor	a0, a0, a1

arch/xtensa/kernel/head.S

@@ -113,7 +113,7 @@ ENTRY(_startup)
 	movi	a0, 0
 
 #if XCHAL_HAVE_VECBASE
-	movi    a2, VECBASE_RESET_VADDR
+	movi    a2, VECBASE_VADDR
 	wsr	a2, vecbase
 #endif
 

arch/xtensa/kernel/vmlinux.lds.S

@@ -30,10 +30,6 @@ jiffies = jiffies_64 + 4;
 jiffies = jiffies_64;
 #endif
 
-#ifndef KERNELOFFSET
-#define KERNELOFFSET 0xd0003000
-#endif
-
 /* Note: In the following macros, it would be nice to specify only the
    vector name and section kind and construct "sym" and "section" using
    CPP concatenation, but that does not work reliably.  Concatenating a