From 0aba691a7443a7541c1dc56423e0c92cc6ea7164 Mon Sep 17 00:00:00 2001 From: Alexandre Ghiti Date: Fri, 23 Jul 2021 15:01:24 +0200 Subject: riscv: Introduce va_kernel_pa_offset for 32-bit kernel va_kernel_pa_offset was only used for 64-bit as the kernel mapping lies in the linear mapping for 32-bit kernel and then only the offset between the PAGE_OFFSET and the kernel load address is needed. But this distinction complexifies the code with #ifdefs and especially with a separate definition of the address conversions macros. Simplify the code by defining this variable for both 32-bit and 64-bit. Signed-off-by: Alexandre Ghiti Signed-off-by: Palmer Dabbelt --- arch/riscv/mm/init.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 269fc648ef3d..263ae055e81a 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -552,11 +552,8 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) kernel_map.phys_addr = (uintptr_t)(&_start); kernel_map.size = (uintptr_t)(&_end) - kernel_map.phys_addr; #endif - kernel_map.va_pa_offset = PAGE_OFFSET - kernel_map.phys_addr; -#ifdef CONFIG_64BIT kernel_map.va_kernel_pa_offset = kernel_map.virt_addr - kernel_map.phys_addr; -#endif pfn_base = PFN_DOWN(kernel_map.phys_addr); -- cgit v1.2.3-70-g09d2 From 526f83df1d83b9c95e571ea5d4dff12be1b215ec Mon Sep 17 00:00:00 2001 From: Alexandre Ghiti Date: Fri, 23 Jul 2021 15:01:25 +0200 Subject: riscv: Get rid of map_size parameter to create_kernel_page_table The kernel must always be mapped using PMD_SIZE, and this is already the case, this just simplifies create_kernel_page_table. Signed-off-by: Alexandre Ghiti Signed-off-by: Palmer Dabbelt --- arch/riscv/mm/init.c | 30 +++++++++++------------------- 1 file changed, 11 insertions(+), 19 deletions(-) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 263ae055e81a..83d0ed915914 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -495,36 +495,35 @@ static __init pgprot_t pgprot_from_va(uintptr_t va) #endif #ifdef CONFIG_XIP_KERNEL -static void __init create_kernel_page_table(pgd_t *pgdir, uintptr_t map_size, +static void __init create_kernel_page_table(pgd_t *pgdir, __always_unused bool early) { uintptr_t va, end_va; /* Map the flash resident part */ end_va = kernel_map.virt_addr + kernel_map.xiprom_sz; - for (va = kernel_map.virt_addr; va < end_va; va += map_size) + for (va = kernel_map.virt_addr; va < end_va; va += PMD_SIZE) create_pgd_mapping(pgdir, va, kernel_map.xiprom + (va - kernel_map.virt_addr), - map_size, PAGE_KERNEL_EXEC); + PMD_SIZE, PAGE_KERNEL_EXEC); /* Map the data in RAM */ end_va = kernel_map.virt_addr + XIP_OFFSET + kernel_map.size; - for (va = kernel_map.virt_addr + XIP_OFFSET; va < end_va; va += map_size) + for (va = kernel_map.virt_addr + XIP_OFFSET; va < end_va; va += PMD_SIZE) create_pgd_mapping(pgdir, va, kernel_map.phys_addr + (va - (kernel_map.virt_addr + XIP_OFFSET)), - map_size, PAGE_KERNEL); + PMD_SIZE, PAGE_KERNEL); } #else -static void __init create_kernel_page_table(pgd_t *pgdir, uintptr_t map_size, - bool early) +static void __init create_kernel_page_table(pgd_t *pgdir, bool early) { uintptr_t va, end_va; end_va = kernel_map.virt_addr + kernel_map.size; - for (va = kernel_map.virt_addr; va < end_va; va += map_size) + for (va = kernel_map.virt_addr; va < end_va; va += PMD_SIZE) create_pgd_mapping(pgdir, va, kernel_map.phys_addr + (va - kernel_map.virt_addr), - map_size, + PMD_SIZE, early ? PAGE_KERNEL_EXEC : pgprot_from_va(va)); } @@ -533,7 +532,6 @@ static void __init create_kernel_page_table(pgd_t *pgdir, uintptr_t map_size, asmlinkage void __init setup_vm(uintptr_t dtb_pa) { uintptr_t __maybe_unused pa; - uintptr_t map_size; #ifndef __PAGETABLE_PMD_FOLDED pmd_t fix_bmap_spmd, fix_bmap_epmd; #endif @@ -557,15 +555,9 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) pfn_base = PFN_DOWN(kernel_map.phys_addr); - /* - * Enforce boot alignment requirements of RV32 and - * RV64 by only allowing PMD or PGD mappings. - */ - map_size = PMD_SIZE; - /* Sanity check alignment and size */ BUG_ON((PAGE_OFFSET % PGDIR_SIZE) != 0); - BUG_ON((kernel_map.phys_addr % map_size) != 0); + BUG_ON((kernel_map.phys_addr % PMD_SIZE) != 0); pt_ops.alloc_pte = alloc_pte_early; pt_ops.get_pte_virt = get_pte_virt_early; @@ -602,7 +594,7 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) * us to reach paging_init(). We map all memory banks later * in setup_vm_final() below. */ - create_kernel_page_table(early_pg_dir, map_size, true); + create_kernel_page_table(early_pg_dir, true); #ifndef __PAGETABLE_PMD_FOLDED /* Setup early PMD for DTB */ @@ -718,7 +710,7 @@ static void __init setup_vm_final(void) #ifdef CONFIG_64BIT /* Map the kernel */ - create_kernel_page_table(swapper_pg_dir, PMD_SIZE, false); + create_kernel_page_table(swapper_pg_dir, false); #endif /* Clear fixmap PTE and PMD mappings */ -- cgit v1.2.3-70-g09d2 From 6f3e5fd241c33d2407f7aaa930b141af6ad7c35b Mon Sep 17 00:00:00 2001 From: Alexandre Ghiti Date: Fri, 23 Jul 2021 15:01:26 +0200 Subject: riscv: Use __maybe_unused instead of #ifdefs around variable declarations This allows to simplify the code and make it more readable. Signed-off-by: Alexandre Ghiti Signed-off-by: Palmer Dabbelt --- arch/riscv/mm/init.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 83d0ed915914..a456b5a68d99 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -532,9 +532,7 @@ static void __init create_kernel_page_table(pgd_t *pgdir, bool early) asmlinkage void __init setup_vm(uintptr_t dtb_pa) { uintptr_t __maybe_unused pa; -#ifndef __PAGETABLE_PMD_FOLDED - pmd_t fix_bmap_spmd, fix_bmap_epmd; -#endif + pmd_t __maybe_unused fix_bmap_spmd, fix_bmap_epmd; kernel_map.virt_addr = KERNEL_LINK_ADDR; -- cgit v1.2.3-70-g09d2 From 977765ce319b98939205cf07aa1d76150713c69b Mon Sep 17 00:00:00 2001 From: Alexandre Ghiti Date: Fri, 23 Jul 2021 15:01:27 +0200 Subject: riscv: Simplify BUILTIN_DTB device tree mapping handling __PAGETABLE_PMD_FOLDED defines a 2-level page table that is only used in 32-bit kernel, so there is no need to check for CONFIG_64BIT in #ifndef __PAGETABLE_PMD_FOLDED and vice-versa. Signed-off-by: Alexandre Ghiti Signed-off-by: Palmer Dabbelt --- arch/riscv/mm/init.c | 12 ++---------- 1 file changed, 2 insertions(+), 10 deletions(-) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index a456b5a68d99..a93822df9d81 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -607,18 +607,14 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) 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(XIP_FIXUP(dtb_pa)); -#else - dtb_early_va = __va(dtb_pa); -#endif /* CONFIG_64BIT */ #endif /* CONFIG_BUILTIN_DTB */ -#else +#else /* __PAGETABLE_PMD_FOLDED */ #ifndef CONFIG_BUILTIN_DTB /* Create two consecutive PGD mappings for FDT early scan */ pa = dtb_pa & ~(PGDIR_SIZE - 1); @@ -628,13 +624,9 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) 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(XIP_FIXUP(dtb_pa)); -#else dtb_early_va = __va(dtb_pa); -#endif /* CONFIG_64BIT */ #endif /* CONFIG_BUILTIN_DTB */ -#endif +#endif /* __PAGETABLE_PMD_FOLDED */ dtb_early_pa = dtb_pa; /* -- cgit v1.2.3-70-g09d2 From fe45ffa4c505783637233609b677446020738b87 Mon Sep 17 00:00:00 2001 From: Alexandre Ghiti Date: Fri, 23 Jul 2021 15:01:28 +0200 Subject: riscv: Move early fdt mapping creation in its own function The code that handles the early fdt mapping is hard to read and does not create the same mapping size depending on the kernel: - for 64-bit, 2 PMD entries are used which amounts to a 4MB mapping - for 32-bit, 2 PGDIR entries are used which amounts to a 8MB mapping So keep using 2 PMD entries for 64-bit and use only one PGD entry for 32-bit needed to cover 4MB. Move that into a new function called create_fdt_early_page_table which, using the same naming as create_kernel_page_table. Signed-off-by: Alexandre Ghiti Signed-off-by: Palmer Dabbelt --- arch/riscv/mm/init.c | 76 +++++++++++++++++++++++++++------------------------- 1 file changed, 40 insertions(+), 36 deletions(-) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index a93822df9d81..fdf093d01c6f 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -222,6 +222,7 @@ pgd_t trampoline_pg_dir[PTRS_PER_PGD] __page_aligned_bss; static pte_t fixmap_pte[PTRS_PER_PTE] __page_aligned_bss; pgd_t early_pg_dir[PTRS_PER_PGD] __initdata __aligned(PAGE_SIZE); +static pmd_t __maybe_unused early_dtb_pmd[PTRS_PER_PMD] __initdata __aligned(PAGE_SIZE); #ifdef CONFIG_XIP_KERNEL #define trampoline_pg_dir ((pgd_t *)XIP_FIXUP(trampoline_pg_dir)) @@ -302,7 +303,6 @@ static void __init create_pte_mapping(pte_t *ptep, static pmd_t trampoline_pmd[PTRS_PER_PMD] __page_aligned_bss; static pmd_t fixmap_pmd[PTRS_PER_PMD] __page_aligned_bss; static pmd_t early_pmd[PTRS_PER_PMD] __initdata __aligned(PAGE_SIZE); -static pmd_t early_dtb_pmd[PTRS_PER_PMD] __initdata __aligned(PAGE_SIZE); #ifdef CONFIG_XIP_KERNEL #define trampoline_pmd ((pmd_t *)XIP_FIXUP(trampoline_pmd)) @@ -388,6 +388,7 @@ static void __init create_pmd_mapping(pmd_t *pmdp, #define create_pgd_next_mapping(__nextp, __va, __pa, __sz, __prot) \ create_pte_mapping(__nextp, __va, __pa, __sz, __prot) #define fixmap_pgd_next fixmap_pte +#define create_pmd_mapping(__pmdp, __va, __pa, __sz, __prot) #endif void __init create_pgd_mapping(pgd_t *pgdp, @@ -529,9 +530,44 @@ static void __init create_kernel_page_table(pgd_t *pgdir, bool early) } #endif +/* + * Setup a 4MB mapping that encompasses the device tree: for 64-bit kernel, + * this means 2 PMD entries whereas for 32-bit kernel, this is only 1 PGDIR + * entry. + */ +static void __init create_fdt_early_page_table(pgd_t *pgdir, uintptr_t dtb_pa) +{ +#ifndef CONFIG_BUILTIN_DTB + uintptr_t pa = dtb_pa & ~(PMD_SIZE - 1); + + create_pgd_mapping(early_pg_dir, DTB_EARLY_BASE_VA, + IS_ENABLED(CONFIG_64BIT) ? (uintptr_t)early_dtb_pmd : pa, + PGDIR_SIZE, + IS_ENABLED(CONFIG_64BIT) ? PAGE_TABLE : PAGE_KERNEL); + + if (IS_ENABLED(CONFIG_64BIT)) { + create_pmd_mapping(early_dtb_pmd, DTB_EARLY_BASE_VA, + pa, PMD_SIZE, PAGE_KERNEL); + create_pmd_mapping(early_dtb_pmd, DTB_EARLY_BASE_VA + PMD_SIZE, + pa + PMD_SIZE, PMD_SIZE, PAGE_KERNEL); + } + + dtb_early_va = (void *)DTB_EARLY_BASE_VA + (dtb_pa & (PMD_SIZE - 1)); +#else + /* + * For 64-bit kernel, __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. For 32-bit kernel, as the + * kernel is mapped in the linear mapping, that makes no difference. + */ + dtb_early_va = kernel_mapping_pa_to_va(XIP_FIXUP(dtb_pa)); +#endif + + dtb_early_pa = dtb_pa; +} + asmlinkage void __init setup_vm(uintptr_t dtb_pa) { - uintptr_t __maybe_unused pa; pmd_t __maybe_unused fix_bmap_spmd, fix_bmap_epmd; kernel_map.virt_addr = KERNEL_LINK_ADDR; @@ -594,40 +630,8 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) */ create_kernel_page_table(early_pg_dir, true); -#ifndef __PAGETABLE_PMD_FOLDED - /* Setup early PMD for DTB */ - create_pgd_mapping(early_pg_dir, DTB_EARLY_BASE_VA, - (uintptr_t)early_dtb_pmd, PGDIR_SIZE, PAGE_TABLE); -#ifndef CONFIG_BUILTIN_DTB - /* Create two consecutive PMD mappings for FDT early scan */ - pa = dtb_pa & ~(PMD_SIZE - 1); - create_pmd_mapping(early_dtb_pmd, DTB_EARLY_BASE_VA, - pa, PMD_SIZE, PAGE_KERNEL); - create_pmd_mapping(early_dtb_pmd, DTB_EARLY_BASE_VA + PMD_SIZE, - pa + PMD_SIZE, PMD_SIZE, PAGE_KERNEL); - dtb_early_va = (void *)DTB_EARLY_BASE_VA + (dtb_pa & (PMD_SIZE - 1)); -#else /* CONFIG_BUILTIN_DTB */ - /* - * __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(XIP_FIXUP(dtb_pa)); -#endif /* CONFIG_BUILTIN_DTB */ -#else /* __PAGETABLE_PMD_FOLDED */ -#ifndef CONFIG_BUILTIN_DTB - /* Create two consecutive PGD mappings for FDT early scan */ - pa = dtb_pa & ~(PGDIR_SIZE - 1); - create_pgd_mapping(early_pg_dir, DTB_EARLY_BASE_VA, - pa, PGDIR_SIZE, PAGE_KERNEL); - create_pgd_mapping(early_pg_dir, DTB_EARLY_BASE_VA + PGDIR_SIZE, - pa + PGDIR_SIZE, PGDIR_SIZE, PAGE_KERNEL); - dtb_early_va = (void *)DTB_EARLY_BASE_VA + (dtb_pa & (PGDIR_SIZE - 1)); -#else /* CONFIG_BUILTIN_DTB */ - dtb_early_va = __va(dtb_pa); -#endif /* CONFIG_BUILTIN_DTB */ -#endif /* __PAGETABLE_PMD_FOLDED */ - dtb_early_pa = dtb_pa; + /* Setup early mapping for FDT early scan */ + create_fdt_early_page_table(early_pg_dir, dtb_pa); /* * Bootime fixmap only can handle PMD_SIZE mapping. Thus, boot-ioremap -- cgit v1.2.3-70-g09d2 From fb31f0a499332a053477ed57312b214e42476e6d Mon Sep 17 00:00:00 2001 From: Kenneth Lee Date: Wed, 28 Jul 2021 15:15:57 +0800 Subject: riscv: fix the global name pfn_base confliction error RISCV uses a global variable pfn_base for page/pfn translation. But this is a common name and will be used elsewhere. In those cases, the page-pfn macros which refer to this name will be referred to the local/input variable instead. (such as in vfio_pin_pages_remote). This make everything wrong. This patch changes the name from pfn_base to riscv_pfn_base to fix this problem. Signed-off-by: Kenneth Lee Signed-off-by: Palmer Dabbelt --- arch/riscv/include/asm/page.h | 4 ++-- arch/riscv/mm/init.c | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h index 4da92cf28a19..d34327be7574 100644 --- a/arch/riscv/include/asm/page.h +++ b/arch/riscv/include/asm/page.h @@ -79,8 +79,8 @@ typedef struct page *pgtable_t; #endif #ifdef CONFIG_MMU -extern unsigned long pfn_base; -#define ARCH_PFN_OFFSET (pfn_base) +extern unsigned long riscv_pfn_base; +#define ARCH_PFN_OFFSET (riscv_pfn_base) #else #define ARCH_PFN_OFFSET (PAGE_OFFSET >> PAGE_SHIFT) #endif /* CONFIG_MMU */ diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index fdf093d01c6f..888dff9530b8 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -214,8 +214,8 @@ static struct pt_alloc_ops _pt_ops __initdata; #define pt_ops _pt_ops #endif -unsigned long pfn_base __ro_after_init; -EXPORT_SYMBOL(pfn_base); +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; @@ -587,7 +587,7 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) kernel_map.va_pa_offset = PAGE_OFFSET - kernel_map.phys_addr; kernel_map.va_kernel_pa_offset = kernel_map.virt_addr - kernel_map.phys_addr; - pfn_base = PFN_DOWN(kernel_map.phys_addr); + riscv_pfn_base = PFN_DOWN(kernel_map.phys_addr); /* Sanity check alignment and size */ BUG_ON((PAGE_OFFSET % PGDIR_SIZE) != 0); -- cgit v1.2.3-70-g09d2 From 8ba1a8b77ba1eb3aef441ed2caf28ab2b1261f5f Mon Sep 17 00:00:00 2001 From: Kefeng Wang Date: Fri, 30 Jul 2021 20:48:41 +0800 Subject: riscv: Support allocating gigantic hugepages using CMA This patch adds support to allocate gigantic hugepages using CMA by specifying the hugetlb_cma= kernel parameter. This is only supported on RV64. Reviewed-by: Alexandre Ghiti Signed-off-by: Kefeng Wang Signed-off-by: Palmer Dabbelt --- arch/riscv/mm/init.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch/riscv/mm') diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 888dff9530b8..248f9ec393c1 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -19,6 +19,7 @@ #include #include #include +#include #include #include @@ -202,6 +203,8 @@ static void __init setup_bootmem(void) early_init_fdt_scan_reserved_mem(); dma_contiguous_reserve(dma32_phys_limit); + if (IS_ENABLED(CONFIG_64BIT)) + hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); memblock_allow_resize(); } -- cgit v1.2.3-70-g09d2