riscv: Move kernel mapping outside of linear mapping

This is a preparatory patch for relocatable kernel and sv48 support.

The kernel used to be linked at PAGE_OFFSET address therefore we could use
the linear mapping for the kernel mapping. But the relocated kernel base
address will be different from PAGE_OFFSET and since in the linear mapping,
two different virtual addresses cannot point to the same physical address,
the kernel mapping needs to lie outside the linear mapping so that we don't
have to copy it at the same physical offset.

The kernel mapping is moved to the last 2GB of the address space, BPF
is now always after the kernel and modules use the 2GB memory range right
before the kernel, so BPF and modules regions do not overlap. KASLR
implementation will simply have to move the kernel in the last 2GB range
and just take care of leaving enough space for BPF.

In addition, by moving the kernel to the end of the address space, both
sv39 and sv48 kernels will be exactly the same without needing to be
relocated at runtime.

Suggested-by: Arnd Bergmann <[e-mail]>
Signed-off-by: Alexandre Ghiti <[e-mail]>
[Palmer: Squash the STRICT_RWX fix, and a !MMU fix]
Signed-off-by: Palmer Dabbelt <[e-mail]>

arch/riscv/boot/loader.lds.S

···
 /* SPDX-License-Identifier: GPL-2.0 */
 
 #include <asm/page.h>
+#include <asm/pgtable.h>
 
 OUTPUT_ARCH(riscv)
 ENTRY(_start)
 
 SECTIONS
 {
-	. = PAGE_OFFSET;
+	. = KERNEL_LINK_ADDR;
 
 	.payload : {
 		*(.payload)

arch/riscv/include/asm/page.h

···
 
 #ifdef CONFIG_MMU
 extern unsigned long va_pa_offset;
+#ifdef CONFIG_64BIT
+extern unsigned long va_kernel_pa_offset;
+#endif
 extern unsigned long pfn_base;
 #define ARCH_PFN_OFFSET		(pfn_base)
 #else
 #define va_pa_offset		0
+#ifdef CONFIG_64BIT
+#define va_kernel_pa_offset	0
+#endif
 #define ARCH_PFN_OFFSET		(PAGE_OFFSET >> PAGE_SHIFT)
 #endif /* CONFIG_MMU */
 
-#define __pa_to_va_nodebug(x)	((void *)((unsigned long) (x) + va_pa_offset))
-#define __va_to_pa_nodebug(x)	((unsigned long)(x) - va_pa_offset)
+#ifdef CONFIG_64BIT
+extern unsigned long kernel_virt_addr;
+
+#define linear_mapping_pa_to_va(x)	((void *)((unsigned long)(x) + va_pa_offset))
+#define kernel_mapping_pa_to_va(x)	((void *)((unsigned long)(x) + va_kernel_pa_offset))
+#define __pa_to_va_nodebug(x)		linear_mapping_pa_to_va(x)
+
+#define linear_mapping_va_to_pa(x)	((unsigned long)(x) - va_pa_offset)
+#define kernel_mapping_va_to_pa(x)	((unsigned long)(x) - va_kernel_pa_offset)
+#define __va_to_pa_nodebug(x)	({						\
+	unsigned long _x = x;							\
+	(_x < kernel_virt_addr) ?						\
+		linear_mapping_va_to_pa(_x) : kernel_mapping_va_to_pa(_x);	\
+	})
+#else
+#define __pa_to_va_nodebug(x)  ((void *)((unsigned long) (x) + va_pa_offset))
+#define __va_to_pa_nodebug(x)  ((unsigned long)(x) - va_pa_offset)
+#endif
 
 #ifdef CONFIG_DEBUG_VIRTUAL
 extern phys_addr_t __virt_to_phys(unsigned long x);

arch/riscv/include/asm/pgtable.h

···
 
 #include <asm/pgtable-bits.h>
 
-#ifndef __ASSEMBLY__
+#ifndef CONFIG_MMU
+#define KERNEL_LINK_ADDR	PAGE_OFFSET
+#else
 
-/* Page Upper Directory not used in RISC-V */
-#include <asm-generic/pgtable-nopud.h>
-#include <asm/page.h>
-#include <asm/tlbflush.h>
-#include <linux/mm_types.h>
+#define ADDRESS_SPACE_END	(UL(-1))
 
-#ifdef CONFIG_MMU
+#ifdef CONFIG_64BIT
+/* Leave 2GB for kernel and BPF at the end of the address space */
+#define KERNEL_LINK_ADDR	(ADDRESS_SPACE_END - SZ_2G + 1)
+#else
+#define KERNEL_LINK_ADDR	PAGE_OFFSET
+#endif
 
 #define VMALLOC_SIZE     (KERN_VIRT_SIZE >> 1)
 #define VMALLOC_END      (PAGE_OFFSET - 1)
 #define VMALLOC_START    (PAGE_OFFSET - VMALLOC_SIZE)
 
 #define BPF_JIT_REGION_SIZE	(SZ_128M)
+#ifdef CONFIG_64BIT
+/* KASLR should leave at least 128MB for BPF after the kernel */
+#define BPF_JIT_REGION_START	PFN_ALIGN((unsigned long)&_end)
+#define BPF_JIT_REGION_END	(BPF_JIT_REGION_START + BPF_JIT_REGION_SIZE)
+#else
 #define BPF_JIT_REGION_START	(PAGE_OFFSET - BPF_JIT_REGION_SIZE)
 #define BPF_JIT_REGION_END	(VMALLOC_END)
+#endif
+
+/* Modules always live before the kernel */
+#ifdef CONFIG_64BIT
+#define MODULES_VADDR	(PFN_ALIGN((unsigned long)&_end) - SZ_2G)
+#define MODULES_END	(PFN_ALIGN((unsigned long)&_start))
+#endif
 
 /*
  * Roughly size the vmemmap space to be large enough to fit enough
···
 #define FIXADDR_SIZE     PGDIR_SIZE
 #endif
 #define FIXADDR_START    (FIXADDR_TOP - FIXADDR_SIZE)
-
 #endif
+
+#ifndef __ASSEMBLY__
+
+/* Page Upper Directory not used in RISC-V */
+#include <asm-generic/pgtable-nopud.h>
+#include <asm/page.h>
+#include <asm/tlbflush.h>
+#include <linux/mm_types.h>
 
 #ifdef CONFIG_64BIT
 #include <asm/pgtable-64.h>
···
 
 #define kern_addr_valid(addr)   (1) /* FIXME */
 
+extern char _start[];
 extern void *dtb_early_va;
 extern uintptr_t dtb_early_pa;
 void setup_bootmem(void);

arch/riscv/include/asm/set_memory.h

···
 int set_memory_nx(unsigned long addr, int numpages);
 int set_memory_rw_nx(unsigned long addr, int numpages);
 void protect_kernel_text_data(void);
+void protect_kernel_linear_mapping_text_rodata(void);
 #else
 static inline int set_memory_ro(unsigned long addr, int numpages) { return 0; }
 static inline int set_memory_rw(unsigned long addr, int numpages) { return 0; }

arch/riscv/kernel/head.S

···
 #ifdef CONFIG_MMU
 relocate:
 	/* Relocate return address */
-	li a1, PAGE_OFFSET
+	la a1, kernel_virt_addr
+	REG_L a1, 0(a1)
 	la a2, _start
 	sub a1, a1, a2
 	add ra, ra, a1

arch/riscv/kernel/module.c

···
 }
 
 #if defined(CONFIG_MMU) && defined(CONFIG_64BIT)
-#define VMALLOC_MODULE_START \
-	 max(PFN_ALIGN((unsigned long)&_end - SZ_2G), VMALLOC_START)
 void *module_alloc(unsigned long size)
 {
-	return __vmalloc_node_range(size, 1, VMALLOC_MODULE_START,
-				    VMALLOC_END, GFP_KERNEL,
+	return __vmalloc_node_range(size, 1, MODULES_VADDR,
+				    MODULES_END, GFP_KERNEL,
 				    PAGE_KERNEL_EXEC, 0, NUMA_NO_NODE,
 				    __builtin_return_address(0));
 }

arch/riscv/kernel/setup.c

···
 
 	sbi_init();
 
-	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX))
+	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
 		protect_kernel_text_data();
+#if defined(CONFIG_64BIT) && defined(CONFIG_MMU)
+		protect_kernel_linear_mapping_text_rodata();
+#endif
+	}
+
 #ifdef CONFIG_SWIOTLB
 	swiotlb_init(1);
 #endif

arch/riscv/kernel/vmlinux.lds.S

···
  * Copyright (C) 2017 SiFive
  */
 
-#define LOAD_OFFSET PAGE_OFFSET
+#include <asm/pgtable.h>
+#define LOAD_OFFSET KERNEL_LINK_ADDR
 #include <asm/vmlinux.lds.h>
 #include <asm/page.h>
 #include <asm/cache.h>

arch/riscv/mm/fault.c

···
 		return;
 	}
 
+#ifdef CONFIG_64BIT
+	/*
+	 * Modules in 64bit kernels lie in their own virtual region which is not
+	 * in the vmalloc region, but dealing with page faults in this region
+	 * or the vmalloc region amounts to doing the same thing: checking that
+	 * the mapping exists in init_mm.pgd and updating user page table, so
+	 * just use vmalloc_fault.
+	 */
+	if (unlikely(addr >= MODULES_VADDR && addr < MODULES_END)) {
+		vmalloc_fault(regs, code, addr);
+		return;
+	}
+#endif
 	/* Enable interrupts if they were enabled in the parent context. */
 	if (likely(regs->status & SR_PIE))
 		local_irq_enable();

arch/riscv/mm/init.c

···
 
 #include "../kernel/head.h"
 
+unsigned long kernel_virt_addr = KERNEL_LINK_ADDR;
+EXPORT_SYMBOL(kernel_virt_addr);
+
 unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)]
 							__page_aligned_bss;
 EXPORT_SYMBOL(empty_zero_page);
···
 		  (unsigned long)VMALLOC_END);
 	print_mlm("lowmem", (unsigned long)PAGE_OFFSET,
 		  (unsigned long)high_memory);
+#ifdef CONFIG_64BIT
+	print_mlm("kernel", (unsigned long)KERNEL_LINK_ADDR,
+		  (unsigned long)ADDRESS_SPACE_END);
+#endif
 }
 #else
 static void print_vm_layout(void) { }
···
 	/* The maximal physical memory size is -PAGE_OFFSET. */
 	memblock_enforce_memory_limit(-PAGE_OFFSET);
 
-	/* Reserve from the start of the kernel to the end of the kernel */
-	memblock_reserve(vmlinux_start, vmlinux_end - vmlinux_start);
+	/*
+	 * Reserve from the start of the kernel to the end of the kernel
+	 * and make sure we align the reservation on PMD_SIZE since we will
+	 * map the kernel in the linear mapping as read-only: we do not want
+	 * any allocation to happen between _end and the next pmd aligned page.
+	 */
+	memblock_reserve(vmlinux_start, (vmlinux_end - vmlinux_start + PMD_SIZE - 1) & PMD_MASK);
 
 	/*
 	 * memblock allocator is not aware of the fact that last 4K bytes of
···
 #ifdef CONFIG_MMU
 static struct pt_alloc_ops pt_ops;
 
+/* Offset between linear mapping virtual address and kernel load address */
 unsigned long va_pa_offset;
 EXPORT_SYMBOL(va_pa_offset);
+#ifdef CONFIG_64BIT
+/* Offset between kernel mapping virtual address and kernel load address */
+unsigned long va_kernel_pa_offset;
+EXPORT_SYMBOL(va_kernel_pa_offset);
+#endif
 unsigned long pfn_base;
 EXPORT_SYMBOL(pfn_base);
 
···
 
 static phys_addr_t __init alloc_pmd_early(uintptr_t va)
 {
-	BUG_ON((va - PAGE_OFFSET) >> PGDIR_SHIFT);
+	BUG_ON((va - kernel_virt_addr) >> PGDIR_SHIFT);
 
 	return (uintptr_t)early_pmd;
 }
···
 #error "setup_vm() is called from head.S before relocate so it should not use absolute addressing."
 #endif
 
+uintptr_t load_pa, load_sz;
+
+static void __init create_kernel_page_table(pgd_t *pgdir, uintptr_t map_size)
+{
+	uintptr_t va, end_va;
+
+	end_va = kernel_virt_addr + load_sz;
+	for (va = kernel_virt_addr; va < end_va; va += map_size)
+		create_pgd_mapping(pgdir, va,
+				   load_pa + (va - kernel_virt_addr),
+				   map_size, PAGE_KERNEL_EXEC);
+}
+
 asmlinkage void __init setup_vm(uintptr_t dtb_pa)
 {
-	uintptr_t va, pa, end_va;
-	uintptr_t load_pa = (uintptr_t)(&_start);
-	uintptr_t load_sz = (uintptr_t)(&_end) - load_pa;
+	uintptr_t pa;
 	uintptr_t map_size;
 #ifndef __PAGETABLE_PMD_FOLDED
 	pmd_t fix_bmap_spmd, fix_bmap_epmd;
 #endif
+	load_pa = (uintptr_t)(&_start);
+	load_sz = (uintptr_t)(&_end) - load_pa;
 
 	va_pa_offset = PAGE_OFFSET - load_pa;
+#ifdef CONFIG_64BIT
+	va_kernel_pa_offset = kernel_virt_addr - load_pa;
+#endif
+
 	pfn_base = PFN_DOWN(load_pa);
 
 	/*
···
 	create_pmd_mapping(fixmap_pmd, FIXADDR_START,
 			   (uintptr_t)fixmap_pte, PMD_SIZE, PAGE_TABLE);
 	/* Setup trampoline PGD and PMD */
-	create_pgd_mapping(trampoline_pg_dir, PAGE_OFFSET,
+	create_pgd_mapping(trampoline_pg_dir, kernel_virt_addr,
 			   (uintptr_t)trampoline_pmd, PGDIR_SIZE, PAGE_TABLE);
-	create_pmd_mapping(trampoline_pmd, PAGE_OFFSET,
+	create_pmd_mapping(trampoline_pmd, kernel_virt_addr,
 			   load_pa, PMD_SIZE, PAGE_KERNEL_EXEC);
 #else
 	/* Setup trampoline PGD */
-	create_pgd_mapping(trampoline_pg_dir, PAGE_OFFSET,
+	create_pgd_mapping(trampoline_pg_dir, kernel_virt_addr,
 			   load_pa, PGDIR_SIZE, PAGE_KERNEL_EXEC);
 #endif
 
 	/*
-	 * Setup early PGD covering entire kernel which will allows
+	 * Setup early PGD covering entire kernel which will allow
 	 * 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);
+	create_kernel_page_table(early_pg_dir, map_size);
 
 #ifndef __PAGETABLE_PMD_FOLDED
 	/* Setup early PMD for DTB */
···
 			   pa + PMD_SIZE, PMD_SIZE, PAGE_KERNEL);
 	dtb_early_va = (void *)DTB_EARLY_BASE_VA + (dtb_pa & (PMD_SIZE - 1));
 #else /* CONFIG_BUILTIN_DTB */
+#ifdef CONFIG_64BIT
+	/*
+	 * __va can't be used since it would return a linear mapping address
+	 * whereas dtb_early_va will be used before setup_vm_final installs
+	 * the linear mapping.
+	 */
+	dtb_early_va = kernel_mapping_pa_to_va(dtb_pa);
+#else
 	dtb_early_va = __va(dtb_pa);
+#endif /* CONFIG_64BIT */
 #endif /* CONFIG_BUILTIN_DTB */
 #else
 #ifndef CONFIG_BUILTIN_DTB
···
 			   pa + PGDIR_SIZE, PGDIR_SIZE, PAGE_KERNEL);
 	dtb_early_va = (void *)DTB_EARLY_BASE_VA + (dtb_pa & (PGDIR_SIZE - 1));
 #else /* CONFIG_BUILTIN_DTB */
+#ifdef CONFIG_64BIT
+	dtb_early_va = kernel_mapping_pa_to_va(dtb_pa);
+#else
 	dtb_early_va = __va(dtb_pa);
+#endif /* CONFIG_64BIT */
 #endif /* CONFIG_BUILTIN_DTB */
 #endif
 	dtb_early_pa = dtb_pa;
···
 #endif
 }
 
+#ifdef CONFIG_64BIT
+void protect_kernel_linear_mapping_text_rodata(void)
+{
+	unsigned long text_start = (unsigned long)lm_alias(_start);
+	unsigned long init_text_start = (unsigned long)lm_alias(__init_text_begin);
+	unsigned long rodata_start = (unsigned long)lm_alias(__start_rodata);
+	unsigned long data_start = (unsigned long)lm_alias(_data);
+
+	set_memory_ro(text_start, (init_text_start - text_start) >> PAGE_SHIFT);
+	set_memory_nx(text_start, (init_text_start - text_start) >> PAGE_SHIFT);
+
+	set_memory_ro(rodata_start, (data_start - rodata_start) >> PAGE_SHIFT);
+	set_memory_nx(rodata_start, (data_start - rodata_start) >> PAGE_SHIFT);
+}
+#endif
+
 static void __init setup_vm_final(void)
 {
 	uintptr_t va, map_size;
···
 			   __pa_symbol(fixmap_pgd_next),
 			   PGDIR_SIZE, PAGE_TABLE);
 
-	/* Map all memory banks */
+	/* Map all memory banks in the linear mapping */
 	for_each_mem_range(i, &start, &end) {
 		if (start >= end)
 			break;
···
 		for (pa = start; pa < end; pa += map_size) {
 			va = (uintptr_t)__va(pa);
 			create_pgd_mapping(swapper_pg_dir, va, pa,
-					   map_size, PAGE_KERNEL_EXEC);
+					   map_size,
+#ifdef CONFIG_64BIT
+					   PAGE_KERNEL
+#else
+					   PAGE_KERNEL_EXEC
+#endif
+					);
+
 		}
 	}
+
+#ifdef CONFIG_64BIT
+	/* Map the kernel */
+	create_kernel_page_table(swapper_pg_dir, PMD_SIZE);
+#endif
 
 	/* Clear fixmap PTE and PMD mappings */
 	clear_fixmap(FIX_PTE);

arch/riscv/mm/kasan_init.c

···
 	phys_addr_t _start, _end;
 	u64 i;
 
+	/*
+	 * Populate all kernel virtual address space with kasan_early_shadow_page
+	 * except for the linear mapping and the modules/kernel/BPF mapping.
+	 */
 	kasan_populate_early_shadow((void *)KASAN_SHADOW_START,
 				    (void *)kasan_mem_to_shadow((void *)
 								VMEMMAP_END));
···
 			(void *)kasan_mem_to_shadow((void *)VMALLOC_START),
 			(void *)kasan_mem_to_shadow((void *)VMALLOC_END));
 
+	/* Populate the linear mapping */
 	for_each_mem_range(i, &_start, &_end) {
 		void *start = (void *)__va(_start);
 		void *end = (void *)__va(_end);
···
 
 		kasan_populate(kasan_mem_to_shadow(start), kasan_mem_to_shadow(end));
 	};
+
+	/* Populate kernel, BPF, modules mapping */
+	kasan_populate(kasan_mem_to_shadow((const void *)MODULES_VADDR),
+		       kasan_mem_to_shadow((const void *)BPF_JIT_REGION_END));
 
 	for (i = 0; i < PTRS_PER_PTE; i++)
 		set_pte(&kasan_early_shadow_pte[i],

arch/riscv/mm/physaddr.c

···
 
 phys_addr_t __phys_addr_symbol(unsigned long x)
 {
-	unsigned long kernel_start = (unsigned long)PAGE_OFFSET;
+	unsigned long kernel_start = (unsigned long)kernel_virt_addr;
 	unsigned long kernel_end = (unsigned long)_end;
 
 	/*