From: Nanyong Sun <sunnanyong@huawei.com> To: <paul.walmsley@sifive.com>, <palmer@dabbelt.com>, <aou@eecs.berkeley.edu>, <anup.patel@wdc.com> Cc: <linux-riscv@lists.infradead.org>, <linux-kernel@vger.kernel.org>, <palmerdabbelt@google.com>, <atish.patra@wdc.com>, <wangkefeng.wang@huawei.com>, <sunnanyong@huawei.com> Subject: [PATCH -next] riscv: mm: remove redundant trampoline PGD for 64bit Date: Thu, 27 May 2021 22:48:19 +0800 [thread overview] Message-ID: <20210527144819.12101-1-sunnanyong@huawei.com> (raw) Remove redundant trampoline PGD for 64bit and add more comment for why 32bit systems need trampoline PGD. There was a patch and discussion similar to this,refer to the link [1][2]. The trampoline PGD is redundant for 64bit systems because: 1. The early PGD covers the entire kernel mapping. Directly loading early PGD can achieve the result in boot stage. A more trampoline PGD makes code hard to understand. 2. Directly loading early PGD is safe in 64bit systems since the kernel virtual address starts as 0xFFFFxxxxxxxxxxxx, which has a very big gap with RAM address.It won't fall into the corner case that 32bit system worrys. 3. Remove redundant trampoline PGD can benefit to code maintaince, because 64bit systems have more page table levels.For example: If we want to support SV48 which has 4 page table levels, we have to add a trampoline_pud and insert it before trampoline_pmd. Reference link: [1]https://lore.kernel.org/linux-riscv/20190325092234.5451-4-anup.patel@wdc.com/ [2]https://lkml.org/lkml/2019/3/28/147 Signed-off-by: Nanyong Sun <sunnanyong@huawei.com> --- arch/riscv/kernel/head.S | 13 +++++++++++-- arch/riscv/mm/init.c | 21 +++++++-------------- 2 files changed, 18 insertions(+), 16 deletions(-) diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S index 89cc58ab52b4..1897b17c5fcc 100644 --- a/arch/riscv/kernel/head.S +++ b/arch/riscv/kernel/head.S @@ -93,12 +93,18 @@ relocate: add a2, a2, a1 csrw CSR_TVEC, a2 - /* Compute satp for kernel page tables, but don't load it yet */ + /* Compute satp for kernel page tables */ srl a2, a0, PAGE_SHIFT li a1, SATP_MODE or a2, a2, a1 - +#ifdef CONFIG_64BIT + /* Load kernel page directory */ + sfence.vma + csrw CSR_SATP, a2 +#else /* + * 32bit system need a trampoline to handle a corner case where + * load address range overlaps kernel virtual address range. * Load trampoline page directory, which will cause us to trap to * stvec if VA != PA, or simply fall through if VA == PA. We need a * full fence here because setup_vm() just wrote these PTEs and we need @@ -110,6 +116,7 @@ relocate: or a0, a0, a1 sfence.vma csrw CSR_SATP, a0 +#endif /* CONFIG_64BIT */ .align 2 1: /* Set trap vector to spin forever to help debug */ @@ -122,6 +129,7 @@ relocate: la gp, __global_pointer$ .option pop +#ifdef CONFIG_32BIT /* * Switch to kernel page tables. A full fence is necessary in order to * avoid using the trampoline translations, which are only correct for @@ -130,6 +138,7 @@ relocate: */ csrw CSR_SATP, a2 sfence.vma +#endif /* CONFIG_32BIT */ ret #endif /* CONFIG_MMU */ diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 11b61bea0c4d..b7226ac2d04f 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -210,13 +210,17 @@ unsigned long pfn_base __ro_after_init; EXPORT_SYMBOL(pfn_base); pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_bss; +#ifdef CONFIG_32BIT pgd_t trampoline_pg_dir[PTRS_PER_PGD] __page_aligned_bss; +#endif /* CONFIG_32BIT */ pte_t fixmap_pte[PTRS_PER_PTE] __page_aligned_bss; pgd_t early_pg_dir[PTRS_PER_PGD] __initdata __aligned(PAGE_SIZE); #ifdef CONFIG_XIP_KERNEL +#ifdef CONFIG_32BIT #define trampoline_pg_dir ((pgd_t *)XIP_FIXUP(trampoline_pg_dir)) +#endif /* CONFIG_32BIT */ #define fixmap_pte ((pte_t *)XIP_FIXUP(fixmap_pte)) #define early_pg_dir ((pgd_t *)XIP_FIXUP(early_pg_dir)) #endif /* CONFIG_XIP_KERNEL */ @@ -291,13 +295,11 @@ static void __init create_pte_mapping(pte_t *ptep, #ifndef __PAGETABLE_PMD_FOLDED -pmd_t trampoline_pmd[PTRS_PER_PMD] __page_aligned_bss; pmd_t fixmap_pmd[PTRS_PER_PMD] __page_aligned_bss; pmd_t early_pmd[PTRS_PER_PMD] __initdata __aligned(PAGE_SIZE); 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)) #define fixmap_pmd ((pmd_t *)XIP_FIXUP(fixmap_pmd)) #define early_pmd ((pmd_t *)XIP_FIXUP(early_pmd)) #endif /* CONFIG_XIP_KERNEL */ @@ -543,21 +545,12 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) /* Setup fixmap PMD */ 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, kernel_virt_addr, - (uintptr_t)trampoline_pmd, PGDIR_SIZE, PAGE_TABLE); -#ifdef CONFIG_XIP_KERNEL - create_pmd_mapping(trampoline_pmd, kernel_virt_addr, - xiprom, PMD_SIZE, PAGE_KERNEL_EXEC); -#else - create_pmd_mapping(trampoline_pmd, kernel_virt_addr, - load_pa, PMD_SIZE, PAGE_KERNEL_EXEC); -#endif -#else +#endif /* __PAGETABLE_PMD_FOLDED */ +#ifdef CONFIG_32BIT /* Setup trampoline PGD */ create_pgd_mapping(trampoline_pg_dir, kernel_virt_addr, load_pa, PGDIR_SIZE, PAGE_KERNEL_EXEC); -#endif +#endif /* CONFIG_32BIT */ /* * Setup early PGD covering entire kernel which will allow -- 2.18.0.huawei.25
WARNING: multiple messages have this Message-ID (diff)
From: Nanyong Sun <sunnanyong@huawei.com> To: <paul.walmsley@sifive.com>, <palmer@dabbelt.com>, <aou@eecs.berkeley.edu>, <anup.patel@wdc.com> Cc: <linux-riscv@lists.infradead.org>, <linux-kernel@vger.kernel.org>, <palmerdabbelt@google.com>, <atish.patra@wdc.com>, <wangkefeng.wang@huawei.com>, <sunnanyong@huawei.com> Subject: [PATCH -next] riscv: mm: remove redundant trampoline PGD for 64bit Date: Thu, 27 May 2021 22:48:19 +0800 [thread overview] Message-ID: <20210527144819.12101-1-sunnanyong@huawei.com> (raw) Remove redundant trampoline PGD for 64bit and add more comment for why 32bit systems need trampoline PGD. There was a patch and discussion similar to this,refer to the link [1][2]. The trampoline PGD is redundant for 64bit systems because: 1. The early PGD covers the entire kernel mapping. Directly loading early PGD can achieve the result in boot stage. A more trampoline PGD makes code hard to understand. 2. Directly loading early PGD is safe in 64bit systems since the kernel virtual address starts as 0xFFFFxxxxxxxxxxxx, which has a very big gap with RAM address.It won't fall into the corner case that 32bit system worrys. 3. Remove redundant trampoline PGD can benefit to code maintaince, because 64bit systems have more page table levels.For example: If we want to support SV48 which has 4 page table levels, we have to add a trampoline_pud and insert it before trampoline_pmd. Reference link: [1]https://lore.kernel.org/linux-riscv/20190325092234.5451-4-anup.patel@wdc.com/ [2]https://lkml.org/lkml/2019/3/28/147 Signed-off-by: Nanyong Sun <sunnanyong@huawei.com> --- arch/riscv/kernel/head.S | 13 +++++++++++-- arch/riscv/mm/init.c | 21 +++++++-------------- 2 files changed, 18 insertions(+), 16 deletions(-) diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S index 89cc58ab52b4..1897b17c5fcc 100644 --- a/arch/riscv/kernel/head.S +++ b/arch/riscv/kernel/head.S @@ -93,12 +93,18 @@ relocate: add a2, a2, a1 csrw CSR_TVEC, a2 - /* Compute satp for kernel page tables, but don't load it yet */ + /* Compute satp for kernel page tables */ srl a2, a0, PAGE_SHIFT li a1, SATP_MODE or a2, a2, a1 - +#ifdef CONFIG_64BIT + /* Load kernel page directory */ + sfence.vma + csrw CSR_SATP, a2 +#else /* + * 32bit system need a trampoline to handle a corner case where + * load address range overlaps kernel virtual address range. * Load trampoline page directory, which will cause us to trap to * stvec if VA != PA, or simply fall through if VA == PA. We need a * full fence here because setup_vm() just wrote these PTEs and we need @@ -110,6 +116,7 @@ relocate: or a0, a0, a1 sfence.vma csrw CSR_SATP, a0 +#endif /* CONFIG_64BIT */ .align 2 1: /* Set trap vector to spin forever to help debug */ @@ -122,6 +129,7 @@ relocate: la gp, __global_pointer$ .option pop +#ifdef CONFIG_32BIT /* * Switch to kernel page tables. A full fence is necessary in order to * avoid using the trampoline translations, which are only correct for @@ -130,6 +138,7 @@ relocate: */ csrw CSR_SATP, a2 sfence.vma +#endif /* CONFIG_32BIT */ ret #endif /* CONFIG_MMU */ diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 11b61bea0c4d..b7226ac2d04f 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -210,13 +210,17 @@ unsigned long pfn_base __ro_after_init; EXPORT_SYMBOL(pfn_base); pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_bss; +#ifdef CONFIG_32BIT pgd_t trampoline_pg_dir[PTRS_PER_PGD] __page_aligned_bss; +#endif /* CONFIG_32BIT */ pte_t fixmap_pte[PTRS_PER_PTE] __page_aligned_bss; pgd_t early_pg_dir[PTRS_PER_PGD] __initdata __aligned(PAGE_SIZE); #ifdef CONFIG_XIP_KERNEL +#ifdef CONFIG_32BIT #define trampoline_pg_dir ((pgd_t *)XIP_FIXUP(trampoline_pg_dir)) +#endif /* CONFIG_32BIT */ #define fixmap_pte ((pte_t *)XIP_FIXUP(fixmap_pte)) #define early_pg_dir ((pgd_t *)XIP_FIXUP(early_pg_dir)) #endif /* CONFIG_XIP_KERNEL */ @@ -291,13 +295,11 @@ static void __init create_pte_mapping(pte_t *ptep, #ifndef __PAGETABLE_PMD_FOLDED -pmd_t trampoline_pmd[PTRS_PER_PMD] __page_aligned_bss; pmd_t fixmap_pmd[PTRS_PER_PMD] __page_aligned_bss; pmd_t early_pmd[PTRS_PER_PMD] __initdata __aligned(PAGE_SIZE); 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)) #define fixmap_pmd ((pmd_t *)XIP_FIXUP(fixmap_pmd)) #define early_pmd ((pmd_t *)XIP_FIXUP(early_pmd)) #endif /* CONFIG_XIP_KERNEL */ @@ -543,21 +545,12 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa) /* Setup fixmap PMD */ 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, kernel_virt_addr, - (uintptr_t)trampoline_pmd, PGDIR_SIZE, PAGE_TABLE); -#ifdef CONFIG_XIP_KERNEL - create_pmd_mapping(trampoline_pmd, kernel_virt_addr, - xiprom, PMD_SIZE, PAGE_KERNEL_EXEC); -#else - create_pmd_mapping(trampoline_pmd, kernel_virt_addr, - load_pa, PMD_SIZE, PAGE_KERNEL_EXEC); -#endif -#else +#endif /* __PAGETABLE_PMD_FOLDED */ +#ifdef CONFIG_32BIT /* Setup trampoline PGD */ create_pgd_mapping(trampoline_pg_dir, kernel_virt_addr, load_pa, PGDIR_SIZE, PAGE_KERNEL_EXEC); -#endif +#endif /* CONFIG_32BIT */ /* * Setup early PGD covering entire kernel which will allow -- 2.18.0.huawei.25 _______________________________________________ linux-riscv mailing list linux-riscv@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-riscv
next reply other threads:[~2021-05-27 14:15 UTC|newest] Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-05-27 14:48 Nanyong Sun [this message] 2021-05-27 14:48 ` [PATCH -next] riscv: mm: remove redundant trampoline PGD for 64bit Nanyong Sun 2021-06-17 9:38 ` Sunnanyong (Nanyong Sun, Intelligent Computing Solution Development Dep) 2021-06-17 9:38 ` Sunnanyong (Nanyong Sun, Intelligent Computing Solution Development Dep) 2021-06-29 13:01 ` Alex Ghiti 2021-06-29 13:01 ` Alex Ghiti 2021-07-28 3:09 ` Sunnanyong (Nanyong Sun, Intelligent Computing Solution Development Dep) 2021-07-28 3:09 ` Sunnanyong (Nanyong Sun, Intelligent Computing Solution Development Dep) 2021-07-28 6:25 ` Alex Ghiti 2021-07-28 6:25 ` Alex Ghiti
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20210527144819.12101-1-sunnanyong@huawei.com \ --to=sunnanyong@huawei.com \ --cc=anup.patel@wdc.com \ --cc=aou@eecs.berkeley.edu \ --cc=atish.patra@wdc.com \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-riscv@lists.infradead.org \ --cc=palmer@dabbelt.com \ --cc=palmerdabbelt@google.com \ --cc=paul.walmsley@sifive.com \ --cc=wangkefeng.wang@huawei.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.