Merge patch series "riscv: Use PUD/P4D/PGD pages for the linear mapping"

Alexandre Ghiti <alexghiti@rivosinc.com> says:

This patchset intends to improve tlb utilization by using hugepages for
the linear mapping.

As reported by Anup in v6, when STRICT_KERNEL_RWX is enabled, we must
take care of isolating the kernel text and rodata so that they are not
mapped with a PUD mapping which would then assign wrong permissions to
the whole region: it is achieved the same way as arm64 by using the
memblock nomap API which isolates those regions and re-merge them afterwards
thus avoiding any issue with the system resources tree creation.

arch/riscv/include/asm/page.h |  19 ++++++-
 arch/riscv/mm/init.c          | 102 ++++++++++++++++++++++++++--------
 arch/riscv/mm/physaddr.c      |  16 ++++++
 drivers/of/fdt.c              |  11 ++--
 4 files changed, 118 insertions(+), 30 deletions(-)

* b4-shazam-merge:
  riscv: Use PUD/P4D/PGD pages for the linear mapping
  riscv: Move the linear mapping creation in its own function
  riscv: Get rid of riscv_pfn_base variable

Link: https://lore.kernel.org/r/20230324155421.271544-1-alexghiti@rivosinc.comSigned-off-by: Palmer Dabbelt <palmer@rivosinc.com>

arch/riscv/include/asm/page.h

@@ -89,9 +89,16 @@ typedef struct page *pgtable_t;
 #define PTE_FMT "%08lx"
 #endif
 
+#ifdef CONFIG_64BIT
+/*
+ * We override this value as its generic definition uses __pa too early in
+ * the boot process (before kernel_map.va_pa_offset is set).
+ */
+#define MIN_MEMBLOCK_ADDR      0
+#endif
+
 #ifdef CONFIG_MMU
-extern unsigned long riscv_pfn_base;
-#define ARCH_PFN_OFFSET		(riscv_pfn_base)
+#define ARCH_PFN_OFFSET		(PFN_DOWN((unsigned long)phys_ram_base))
 #else
 #define ARCH_PFN_OFFSET		(PAGE_OFFSET >> PAGE_SHIFT)
 #endif /* CONFIG_MMU */
@@ -121,7 +128,11 @@ extern phys_addr_t phys_ram_base;
 #define is_linear_mapping(x)	\
 	((x) >= PAGE_OFFSET && (!IS_ENABLED(CONFIG_64BIT) || (x) < PAGE_OFFSET + KERN_VIRT_SIZE))
 
+#ifndef CONFIG_DEBUG_VIRTUAL
 #define linear_mapping_pa_to_va(x)	((void *)((unsigned long)(x) + kernel_map.va_pa_offset))
+#else
+void *linear_mapping_pa_to_va(unsigned long x);
+#endif
 #define kernel_mapping_pa_to_va(y)	({					\
 	unsigned long _y = (unsigned long)(y);					\
 	(IS_ENABLED(CONFIG_XIP_KERNEL) && _y < phys_ram_base) ?			\
@@ -130,7 +141,11 @@ extern phys_addr_t phys_ram_base;
 	})
 #define __pa_to_va_nodebug(x)		linear_mapping_pa_to_va(x)
 
+#ifndef CONFIG_DEBUG_VIRTUAL
 #define linear_mapping_va_to_pa(x)	((unsigned long)(x) - kernel_map.va_pa_offset)
+#else
+phys_addr_t linear_mapping_va_to_pa(unsigned long x);
+#endif
 #define kernel_mapping_va_to_pa(y) ({						\
 	unsigned long _y = (unsigned long)(y);					\
 	(IS_ENABLED(CONFIG_XIP_KERNEL) && _y < kernel_map.virt_addr + XIP_OFFSET) ? \

arch/riscv/mm/init.c

@@ -213,6 +213,14 @@ static void __init setup_bootmem(void)
 	phys_ram_end = memblock_end_of_DRAM();
 	if (!IS_ENABLED(CONFIG_XIP_KERNEL))
 		phys_ram_base = memblock_start_of_DRAM();
+
+	/*
+	 * In 64-bit, any use of __va/__pa before this point is wrong as we
+	 * did not know the start of DRAM before.
+	 */
+	if (IS_ENABLED(CONFIG_64BIT))
+		kernel_map.va_pa_offset = PAGE_OFFSET - phys_ram_base;
+
 	/*
 	 * memblock allocator is not aware of the fact that last 4K bytes of
 	 * the addressable memory can not be mapped because of IS_ERR_VALUE
@@ -271,9 +279,6 @@ static void __init setup_bootmem(void)
 #ifdef CONFIG_MMU
 struct pt_alloc_ops pt_ops __initdata;
 
-unsigned long riscv_pfn_base __ro_after_init;
-EXPORT_SYMBOL(riscv_pfn_base);
-
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_bss;
 pgd_t trampoline_pg_dir[PTRS_PER_PGD] __page_aligned_bss;
 static pte_t fixmap_pte[PTRS_PER_PTE] __page_aligned_bss;
@@ -285,7 +290,6 @@ static pmd_t __maybe_unused early_dtb_pmd[PTRS_PER_PMD] __initdata __aligned(PAG
 
 #ifdef CONFIG_XIP_KERNEL
 #define pt_ops			(*(struct pt_alloc_ops *)XIP_FIXUP(&pt_ops))
-#define riscv_pfn_base         (*(unsigned long  *)XIP_FIXUP(&riscv_pfn_base))
 #define trampoline_pg_dir      ((pgd_t *)XIP_FIXUP(trampoline_pg_dir))
 #define fixmap_pte             ((pte_t *)XIP_FIXUP(fixmap_pte))
 #define early_pg_dir           ((pgd_t *)XIP_FIXUP(early_pg_dir))
@@ -671,9 +675,16 @@ void __init create_pgd_mapping(pgd_t *pgdp,
 
 static uintptr_t __init best_map_size(phys_addr_t base, phys_addr_t size)
 {
-	/* Upgrade to PMD_SIZE mappings whenever possible */
-	base &= PMD_SIZE - 1;
-	if (!base && size >= PMD_SIZE)
+	if (!(base & (PGDIR_SIZE - 1)) && size >= PGDIR_SIZE)
+		return PGDIR_SIZE;
+
+	if (!(base & (P4D_SIZE - 1)) && size >= P4D_SIZE)
+		return P4D_SIZE;
+
+	if (!(base & (PUD_SIZE - 1)) && size >= PUD_SIZE)
+		return PUD_SIZE;
+
+	if (!(base & (PMD_SIZE - 1)) && size >= PMD_SIZE)
 		return PMD_SIZE;
 
 	return PAGE_SIZE;
@@ -982,11 +993,22 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa)
 	set_satp_mode();
 #endif
 
-	kernel_map.va_pa_offset = PAGE_OFFSET - kernel_map.phys_addr;
+	/*
+	 * In 64-bit, we defer the setup of va_pa_offset to setup_bootmem,
+	 * where we have the system memory layout: this allows us to align
+	 * the physical and virtual mappings and then make use of PUD/P4D/PGD
+	 * for the linear mapping. This is only possible because the kernel
+	 * mapping lies outside the linear mapping.
+	 * In 32-bit however, as the kernel resides in the linear mapping,
+	 * setup_vm_final can not change the mapping established here,
+	 * otherwise the same kernel addresses would get mapped to different
+	 * physical addresses (if the start of dram is different from the
+	 * kernel physical address start).
+	 */
+	kernel_map.va_pa_offset = IS_ENABLED(CONFIG_64BIT) ?
+				0UL : PAGE_OFFSET - kernel_map.phys_addr;
 	kernel_map.va_kernel_pa_offset = kernel_map.virt_addr - kernel_map.phys_addr;
 
-	riscv_pfn_base = PFN_DOWN(kernel_map.phys_addr);
-
 	/*
 	 * The default maximal physical memory size is KERN_VIRT_SIZE for 32-bit
 	 * kernel, whereas for 64-bit kernel, the end of the virtual address
@@ -1090,16 +1112,36 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa)
 	pt_ops_set_fixmap();
 }
 
-static void __init setup_vm_final(void)
+static void __init create_linear_mapping_range(phys_addr_t start,
+					       phys_addr_t end)
 {
+	phys_addr_t pa;
 	uintptr_t va, map_size;
-	phys_addr_t pa, start, end;
+
+	for (pa = start; pa < end; pa += map_size) {
+		va = (uintptr_t)__va(pa);
+		map_size = best_map_size(pa, end - pa);
+
+		create_pgd_mapping(swapper_pg_dir, va, pa, map_size,
+				   pgprot_from_va(va));
+	}
+}
+
+static void __init create_linear_mapping_page_table(void)
+{
+	phys_addr_t start, end;
 	u64 i;
 
-	/* Setup swapper PGD for fixmap */
-	create_pgd_mapping(swapper_pg_dir, FIXADDR_START,
-			   __pa_symbol(fixmap_pgd_next),
-			   PGDIR_SIZE, PAGE_TABLE);
+#ifdef CONFIG_STRICT_KERNEL_RWX
+	phys_addr_t ktext_start = __pa_symbol(_start);
+	phys_addr_t ktext_size = __init_data_begin - _start;
+	phys_addr_t krodata_start = __pa_symbol(__start_rodata);
+	phys_addr_t krodata_size = _data - __start_rodata;
+
+	/* Isolate kernel text and rodata so they don't get mapped with a PUD */
+	memblock_mark_nomap(ktext_start,  ktext_size);
+	memblock_mark_nomap(krodata_start, krodata_size);
+#endif
 
 	/* Map all memory banks in the linear mapping */
 	for_each_mem_range(i, &start, &end) {
@@ -1111,15 +1153,29 @@ static void __init setup_vm_final(void)
 		if (end >= __pa(PAGE_OFFSET) + memory_limit)
 			end = __pa(PAGE_OFFSET) + memory_limit;
 
-		for (pa = start; pa < end; pa += map_size) {
-			va = (uintptr_t)__va(pa);
-			map_size = best_map_size(pa, end - pa);
-
-			create_pgd_mapping(swapper_pg_dir, va, pa, map_size,
-					   pgprot_from_va(va));
-		}
+		create_linear_mapping_range(start, end);
 	}
 
+#ifdef CONFIG_STRICT_KERNEL_RWX
+	create_linear_mapping_range(ktext_start, ktext_start + ktext_size);
+	create_linear_mapping_range(krodata_start,
+				    krodata_start + krodata_size);
+
+	memblock_clear_nomap(ktext_start,  ktext_size);
+	memblock_clear_nomap(krodata_start, krodata_size);
+#endif
+}
+
+static void __init setup_vm_final(void)
+{
+	/* Setup swapper PGD for fixmap */
+	create_pgd_mapping(swapper_pg_dir, FIXADDR_START,
+			   __pa_symbol(fixmap_pgd_next),
+			   PGDIR_SIZE, PAGE_TABLE);
+
+	/* Map the linear mapping */
+	create_linear_mapping_page_table();
+
 	/* Map the kernel */
 	if (IS_ENABLED(CONFIG_64BIT))
 		create_kernel_page_table(swapper_pg_dir, false);

arch/riscv/mm/physaddr.c

@@ -33,3 +33,19 @@ phys_addr_t __phys_addr_symbol(unsigned long x)
 	return __va_to_pa_nodebug(x);
 }
 EXPORT_SYMBOL(__phys_addr_symbol);
+
+phys_addr_t linear_mapping_va_to_pa(unsigned long x)
+{
+	BUG_ON(!kernel_map.va_pa_offset);
+
+	return ((unsigned long)(x) - kernel_map.va_pa_offset);
+}
+EXPORT_SYMBOL(linear_mapping_va_to_pa);
+
+void *linear_mapping_pa_to_va(unsigned long x)
+{
+	BUG_ON(!kernel_map.va_pa_offset);
+
+	return ((void *)((unsigned long)(x) + kernel_map.va_pa_offset));
+}
+EXPORT_SYMBOL(linear_mapping_pa_to_va);

drivers/of/fdt.c

@@ -887,12 +887,13 @@ const void * __init of_flat_dt_match_machine(const void *default_match,
 static void __early_init_dt_declare_initrd(unsigned long start,
 					   unsigned long end)
 {
-	/* ARM64 would cause a BUG to occur here when CONFIG_DEBUG_VM is
-	 * enabled since __va() is called too early. ARM64 does make use
-	 * of phys_initrd_start/phys_initrd_size so we can skip this
-	 * conversion.
+	/*
+	 * __va() is not yet available this early on some platforms. In that
+	 * case, the platform uses phys_initrd_start/phys_initrd_size instead
+	 * and does the VA conversion itself.
 	 */
-	if (!IS_ENABLED(CONFIG_ARM64)) {
+	if (!IS_ENABLED(CONFIG_ARM64) &&
+	    !(IS_ENABLED(CONFIG_RISCV) && IS_ENABLED(CONFIG_64BIT))) {
 		initrd_start = (unsigned long)__va(start);
 		initrd_end = (unsigned long)__va(end);
 		initrd_below_start_ok = 1;