* [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL @ 2021-05-24 6:51 guoren 2021-05-24 6:51 ` [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h guoren ` (3 more replies) 0 siblings, 4 replies; 11+ messages in thread From: guoren @ 2021-05-24 6:51 UTC (permalink / raw) To: guoren, anup.patel, palmerdabbelt, arnd Cc: linux-riscv, linux-kernel, linux-arch, linux-sunxi, Guo Ren From: Guo Ren <guoren@linux.alibaba.com> Kernel virtual address translation should avoid care asid or it'll cause more TLB-miss and TLB-refill. Because the current asid in satp belongs to the current process, but the target kernel va TLB entry's asid still belongs to the previous process. Signed-off-by: Guo Ren <guoren@linux.alibaba.com> Cc: Anup Patel <anup.patel@wdc.com> Cc: Palmer Dabbelt <palmerdabbelt@google.com> --- arch/riscv/include/asm/pgtable.h | 1 + 1 file changed, 1 insertion(+) diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h index 78f2323..017da15 100644 --- a/arch/riscv/include/asm/pgtable.h +++ b/arch/riscv/include/asm/pgtable.h @@ -135,6 +135,7 @@ | _PAGE_PRESENT \ | _PAGE_ACCESSED \ | _PAGE_DIRTY \ + | _PAGE_GLOBAL \ | _PAGE_CACHE) #define PAGE_KERNEL __pgprot(_PAGE_KERNEL) -- 2.7.4 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h 2021-05-24 6:51 [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL guoren @ 2021-05-24 6:51 ` guoren 2021-05-25 6:34 ` Christoph Hellwig 2021-05-24 6:51 ` [PATCH 3/3] riscv: Use use_asid_allocator flush TLB guoren ` (2 subsequent siblings) 3 siblings, 1 reply; 11+ messages in thread From: guoren @ 2021-05-24 6:51 UTC (permalink / raw) To: guoren, anup.patel, palmerdabbelt, arnd Cc: linux-riscv, linux-kernel, linux-arch, linux-sunxi, Guo Ren From: Guo Ren <guoren@linux.alibaba.com> Current PAGE_UP implementation is wrong. PAGE_UP(0) should be 0x1000, but current implementation will give out 0. Although the current PAGE_UP isn't used, it will soon be used in tlb_flush optimization. Signed-off-by: Guo Ren <guoren@linux.alibaba.com> Cc: Anup Patel <anup.patel@wdc.com> Cc: Palmer Dabbelt <palmerdabbelt@google.com> --- arch/riscv/include/asm/page.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h index 6a7761c..c611b20 100644 --- a/arch/riscv/include/asm/page.h +++ b/arch/riscv/include/asm/page.h @@ -37,7 +37,7 @@ #ifndef __ASSEMBLY__ -#define PAGE_UP(addr) (((addr)+((PAGE_SIZE)-1))&(~((PAGE_SIZE)-1))) +#define PAGE_UP(addr) (((addr)+(PAGE_SIZE))&(~((PAGE_SIZE)-1))) #define PAGE_DOWN(addr) ((addr)&(~((PAGE_SIZE)-1))) /* align addr on a size boundary - adjust address up/down if needed */ -- 2.7.4 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h 2021-05-24 6:51 ` [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h guoren @ 2021-05-25 6:34 ` Christoph Hellwig 2021-05-25 9:28 ` Guo Ren 0 siblings, 1 reply; 11+ messages in thread From: Christoph Hellwig @ 2021-05-25 6:34 UTC (permalink / raw) To: guoren Cc: anup.patel, palmerdabbelt, arnd, linux-riscv, linux-kernel, linux-arch, linux-sunxi, Guo Ren On Mon, May 24, 2021 at 06:51:07AM +0000, guoren@kernel.org wrote: > From: Guo Ren <guoren@linux.alibaba.com> > > Current PAGE_UP implementation is wrong. PAGE_UP(0) should be > 0x1000, but current implementation will give out 0. > > Although the current PAGE_UP isn't used, it will soon be used in > tlb_flush optimization. Nak. Please just remove the PAGE_UP/PAGE_DOWN macros just like they have been removed from other architectures long ago and use the generic DIV_ROUND_UP macro for your new code like everyone else. ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h 2021-05-25 6:34 ` Christoph Hellwig @ 2021-05-25 9:28 ` Guo Ren 2021-05-25 9:34 ` Christoph Hellwig 0 siblings, 1 reply; 11+ messages in thread From: Guo Ren @ 2021-05-25 9:28 UTC (permalink / raw) To: Christoph Hellwig Cc: Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, Linux Kernel Mailing List, linux-arch, linux-sunxi, Guo Ren On Tue, May 25, 2021 at 2:34 PM Christoph Hellwig <hch@infradead.org> wrote: > > On Mon, May 24, 2021 at 06:51:07AM +0000, guoren@kernel.org wrote: > > From: Guo Ren <guoren@linux.alibaba.com> > > > > Current PAGE_UP implementation is wrong. PAGE_UP(0) should be > > 0x1000, but current implementation will give out 0. > > > > Although the current PAGE_UP isn't used, it will soon be used in > > tlb_flush optimization. > > Nak. Please just remove the PAGE_UP/PAGE_DOWN macros just like they > have been removed from other architectures long ago and use the > generic DIV_ROUND_UP macro for your new code like everyone else. This patch has been dropped because it's wrong, ref Anup's reply. Remove PAGE_UP/DOWN is okay for me. How about: static inline void local_flush_tlb_range_asid(unsigned long start, unsigned long size, unsigned long asid) { - unsigned long page_add = PAGE_DOWN(start); - unsigned long page_end = PAGE_UP(start + size); + unsigned long page_add = _ALIGN_DOWN(start, PAGE_SIZE); + unsigned long page_end = _ALIGN_UP(start + size, PAGE_SIZE); _ALIGN_XXX are also defined in arch/riscv/include/asm/page.h. -- Best Regards Guo Ren ML: https://lore.kernel.org/linux-csky/ ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h 2021-05-25 9:28 ` Guo Ren @ 2021-05-25 9:34 ` Christoph Hellwig 0 siblings, 0 replies; 11+ messages in thread From: Christoph Hellwig @ 2021-05-25 9:34 UTC (permalink / raw) To: Guo Ren Cc: Christoph Hellwig, Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, Linux Kernel Mailing List, linux-arch, linux-sunxi, Guo Ren On Tue, May 25, 2021 at 05:28:21PM +0800, Guo Ren wrote: > On Tue, May 25, 2021 at 2:34 PM Christoph Hellwig <hch@infradead.org> wrote: > > > > On Mon, May 24, 2021 at 06:51:07AM +0000, guoren@kernel.org wrote: > > > From: Guo Ren <guoren@linux.alibaba.com> > > > > > > Current PAGE_UP implementation is wrong. PAGE_UP(0) should be > > > 0x1000, but current implementation will give out 0. > > > > > > Although the current PAGE_UP isn't used, it will soon be used in > > > tlb_flush optimization. > > > > Nak. Please just remove the PAGE_UP/PAGE_DOWN macros just like they > > have been removed from other architectures long ago and use the > > generic DIV_ROUND_UP macro for your new code like everyone else. > > This patch has been dropped because it's wrong, ref Anup's reply. > > Remove PAGE_UP/DOWN is okay for me. How about: > static inline void local_flush_tlb_range_asid(unsigned long start, > unsigned long size, > unsigned long asid) > { > - unsigned long page_add = PAGE_DOWN(start); > - unsigned long page_end = PAGE_UP(start + size); > + unsigned long page_add = _ALIGN_DOWN(start, PAGE_SIZE); > + unsigned long page_end = _ALIGN_UP(start + size, PAGE_SIZE); > > _ALIGN_XXX are also defined in arch/riscv/include/asm/page.h. And these also are leftovers from days gone by and should be removed. I think this should simply be: unsigned long page_add = start & PAGE_MASK; unsigned long page_end = PAGE_ALIGN(start + size); (and page_add is a pretty horrible name as well..) ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 3/3] riscv: Use use_asid_allocator flush TLB 2021-05-24 6:51 [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL guoren 2021-05-24 6:51 ` [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h guoren @ 2021-05-24 6:51 ` guoren 2021-05-24 8:13 ` Anup Patel 2021-05-24 8:03 ` [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL Anup Patel 2021-05-24 10:42 ` Anup Patel 3 siblings, 1 reply; 11+ messages in thread From: guoren @ 2021-05-24 6:51 UTC (permalink / raw) To: guoren, anup.patel, palmerdabbelt, arnd Cc: linux-riscv, linux-kernel, linux-arch, linux-sunxi, Guo Ren From: Guo Ren <guoren@linux.alibaba.com> Use static_branch_unlikely(&use_asid_allocator) to keep the origin tlb flush style, so it's no effect on the existing machine. Here are the optimized functions: - flush_tlb_mm - flush_tlb_page - flush_tlb_range All above are based on the below new implement functions: - __sbi_tlb_flush_range_asid - local_flush_tlb_range_asid Signed-off-by: Guo Ren <guoren@linux.alibaba.com> Cc: Anup Patel <anup.patel@wdc.com> Cc: Palmer Dabbelt <palmerdabbelt@google.com> --- arch/riscv/include/asm/mmu_context.h | 2 ++ arch/riscv/include/asm/tlbflush.h | 22 ++++++++++++++++++++ arch/riscv/mm/context.c | 2 +- arch/riscv/mm/tlbflush.c | 40 +++++++++++++++++++++++++++++++++--- 4 files changed, 62 insertions(+), 4 deletions(-) diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h index b065941..7030837 100644 --- a/arch/riscv/include/asm/mmu_context.h +++ b/arch/riscv/include/asm/mmu_context.h @@ -33,6 +33,8 @@ static inline int init_new_context(struct task_struct *tsk, return 0; } +DECLARE_STATIC_KEY_FALSE(use_asid_allocator); + #include <asm-generic/mmu_context.h> #endif /* _ASM_RISCV_MMU_CONTEXT_H */ diff --git a/arch/riscv/include/asm/tlbflush.h b/arch/riscv/include/asm/tlbflush.h index c84218a..9390319 100644 --- a/arch/riscv/include/asm/tlbflush.h +++ b/arch/riscv/include/asm/tlbflush.h @@ -8,6 +8,7 @@ #define _ASM_RISCV_TLBFLUSH_H #include <linux/mm_types.h> +#include <asm/page.h> #include <asm/smp.h> #include <asm/errata_list.h> @@ -22,9 +23,30 @@ static inline void local_flush_tlb_page(unsigned long addr) { ALT_FLUSH_TLB_PAGE(__asm__ __volatile__ ("sfence.vma %0" : : "r" (addr) : "memory")); } + +static inline void local_flush_tlb_range_asid(unsigned long start, unsigned long size, + unsigned long asid) +{ + unsigned long page_add = PAGE_DOWN(start); + unsigned long page_end = PAGE_UP(start + size); + + if (size == -1) { + __asm__ __volatile__ ("sfence.vma x0, %0" : : "r" (asid) : "memory"); + return; + } + + while(page_add < page_end) { + __asm__ __volatile__ ("sfence.vma %0, %1" + : + : "r" (page_add), "r" (asid) + : "memory"); + page_add += PAGE_SIZE; + } +} #else /* CONFIG_MMU */ #define local_flush_tlb_all() do { } while (0) #define local_flush_tlb_page(addr) do { } while (0) +#define local_flush_tlb_range_asid(addr) do { } while (0) #endif /* CONFIG_MMU */ #if defined(CONFIG_SMP) && defined(CONFIG_MMU) diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c index 68aa312..45c1b04 100644 --- a/arch/riscv/mm/context.c +++ b/arch/riscv/mm/context.c @@ -18,7 +18,7 @@ #ifdef CONFIG_MMU -static DEFINE_STATIC_KEY_FALSE(use_asid_allocator); +DEFINE_STATIC_KEY_FALSE(use_asid_allocator); static unsigned long asid_bits; static unsigned long num_asids; diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c index 720b443..69588dc 100644 --- a/arch/riscv/mm/tlbflush.c +++ b/arch/riscv/mm/tlbflush.c @@ -4,6 +4,7 @@ #include <linux/smp.h> #include <linux/sched.h> #include <asm/sbi.h> +#include <asm/mmu_context.h> void flush_tlb_all(void) { @@ -39,18 +40,51 @@ static void __sbi_tlb_flush_range(struct cpumask *cmask, unsigned long start, put_cpu(); } +static void __sbi_tlb_flush_range_asid(struct cpumask *cmask, unsigned long start, + unsigned long size, unsigned long asid) +{ + struct cpumask hmask; + unsigned int cpuid; + + if (cpumask_empty(cmask)) + return; + + cpuid = get_cpu(); + + if (cpumask_any_but(cmask, cpuid) >= nr_cpu_ids) { + local_flush_tlb_range_asid(start, size, asid); + } else { + riscv_cpuid_to_hartid_mask(cmask, &hmask); + sbi_remote_sfence_vma_asid(cpumask_bits(&hmask), start, size, asid); + } + + put_cpu(); +} + void flush_tlb_mm(struct mm_struct *mm) { - __sbi_tlb_flush_range(mm_cpumask(mm), 0, -1); + if (static_branch_unlikely(&use_asid_allocator)) + __sbi_tlb_flush_range_asid(mm_cpumask(mm), 0, -1, + atomic_long_read(&mm->context.id)); + else + __sbi_tlb_flush_range(mm_cpumask(mm), 0, -1); } void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr) { - __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE); + if (static_branch_unlikely(&use_asid_allocator)) + __sbi_tlb_flush_range_asid(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE, + atomic_long_read(&vma->vm_mm->context.id)); + else + __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE); } void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned long end) { - __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), start, end - start); + if (static_branch_unlikely(&use_asid_allocator)) + __sbi_tlb_flush_range_asid(mm_cpumask(vma->vm_mm), start, end - start, + atomic_long_read(&vma->vm_mm->context.id)); + else + __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), start, end - start); } -- 2.7.4 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 3/3] riscv: Use use_asid_allocator flush TLB 2021-05-24 6:51 ` [PATCH 3/3] riscv: Use use_asid_allocator flush TLB guoren @ 2021-05-24 8:13 ` Anup Patel 0 siblings, 0 replies; 11+ messages in thread From: Anup Patel @ 2021-05-24 8:13 UTC (permalink / raw) To: Guo Ren Cc: Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, linux-kernel@vger.kernel.org List, linux-arch, linux-sunxi, Guo Ren On Mon, May 24, 2021 at 12:22 PM <guoren@kernel.org> wrote: > > From: Guo Ren <guoren@linux.alibaba.com> > > Use static_branch_unlikely(&use_asid_allocator) to keep the origin > tlb flush style, so it's no effect on the existing machine. Here > are the optimized functions: > - flush_tlb_mm > - flush_tlb_page > - flush_tlb_range > > All above are based on the below new implement functions: > - __sbi_tlb_flush_range_asid > - local_flush_tlb_range_asid > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com> > Cc: Anup Patel <anup.patel@wdc.com> > Cc: Palmer Dabbelt <palmerdabbelt@google.com> > --- > arch/riscv/include/asm/mmu_context.h | 2 ++ > arch/riscv/include/asm/tlbflush.h | 22 ++++++++++++++++++++ > arch/riscv/mm/context.c | 2 +- > arch/riscv/mm/tlbflush.c | 40 +++++++++++++++++++++++++++++++++--- > 4 files changed, 62 insertions(+), 4 deletions(-) > > diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h > index b065941..7030837 100644 > --- a/arch/riscv/include/asm/mmu_context.h > +++ b/arch/riscv/include/asm/mmu_context.h > @@ -33,6 +33,8 @@ static inline int init_new_context(struct task_struct *tsk, > return 0; > } > > +DECLARE_STATIC_KEY_FALSE(use_asid_allocator); > + > #include <asm-generic/mmu_context.h> > > #endif /* _ASM_RISCV_MMU_CONTEXT_H */ > diff --git a/arch/riscv/include/asm/tlbflush.h b/arch/riscv/include/asm/tlbflush.h > index c84218a..9390319 100644 > --- a/arch/riscv/include/asm/tlbflush.h > +++ b/arch/riscv/include/asm/tlbflush.h > @@ -8,6 +8,7 @@ > #define _ASM_RISCV_TLBFLUSH_H > > #include <linux/mm_types.h> > +#include <asm/page.h> > #include <asm/smp.h> > #include <asm/errata_list.h> > > @@ -22,9 +23,30 @@ static inline void local_flush_tlb_page(unsigned long addr) > { > ALT_FLUSH_TLB_PAGE(__asm__ __volatile__ ("sfence.vma %0" : : "r" (addr) : "memory")); > } > + > +static inline void local_flush_tlb_range_asid(unsigned long start, unsigned long size, > + unsigned long asid) > +{ > + unsigned long page_add = PAGE_DOWN(start); > + unsigned long page_end = PAGE_UP(start + size); Your PATCH2 is not correct because PAGE_UP(x) should in-fact return 0 when x == 0. In fact, if both "start" and "size" are zero then both page_add and page_end should be zero so that no "sfence.vma" is executed. If you want at least one TLB entry to be invalidated when size == 0 then you can simply set "size = 1" when size is zero which will force one TLB invalidation. Please drop your PATCH2 and the rest of the things look good to me. Reviewed-by: Anup Patel <anup@brainfault.org> Regards, Anup > + > + if (size == -1) { > + __asm__ __volatile__ ("sfence.vma x0, %0" : : "r" (asid) : "memory"); > + return; > + } > + > + while(page_add < page_end) { > + __asm__ __volatile__ ("sfence.vma %0, %1" > + : > + : "r" (page_add), "r" (asid) > + : "memory"); > + page_add += PAGE_SIZE; > + } > +} > #else /* CONFIG_MMU */ > #define local_flush_tlb_all() do { } while (0) > #define local_flush_tlb_page(addr) do { } while (0) > +#define local_flush_tlb_range_asid(addr) do { } while (0) > #endif /* CONFIG_MMU */ > > #if defined(CONFIG_SMP) && defined(CONFIG_MMU) > diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c > index 68aa312..45c1b04 100644 > --- a/arch/riscv/mm/context.c > +++ b/arch/riscv/mm/context.c > @@ -18,7 +18,7 @@ > > #ifdef CONFIG_MMU > > -static DEFINE_STATIC_KEY_FALSE(use_asid_allocator); > +DEFINE_STATIC_KEY_FALSE(use_asid_allocator); > > static unsigned long asid_bits; > static unsigned long num_asids; > diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c > index 720b443..69588dc 100644 > --- a/arch/riscv/mm/tlbflush.c > +++ b/arch/riscv/mm/tlbflush.c > @@ -4,6 +4,7 @@ > #include <linux/smp.h> > #include <linux/sched.h> > #include <asm/sbi.h> > +#include <asm/mmu_context.h> > > void flush_tlb_all(void) > { > @@ -39,18 +40,51 @@ static void __sbi_tlb_flush_range(struct cpumask *cmask, unsigned long start, > put_cpu(); > } > > +static void __sbi_tlb_flush_range_asid(struct cpumask *cmask, unsigned long start, > + unsigned long size, unsigned long asid) > +{ > + struct cpumask hmask; > + unsigned int cpuid; > + > + if (cpumask_empty(cmask)) > + return; > + > + cpuid = get_cpu(); > + > + if (cpumask_any_but(cmask, cpuid) >= nr_cpu_ids) { > + local_flush_tlb_range_asid(start, size, asid); > + } else { > + riscv_cpuid_to_hartid_mask(cmask, &hmask); > + sbi_remote_sfence_vma_asid(cpumask_bits(&hmask), start, size, asid); > + } > + > + put_cpu(); > +} > + > void flush_tlb_mm(struct mm_struct *mm) > { > - __sbi_tlb_flush_range(mm_cpumask(mm), 0, -1); > + if (static_branch_unlikely(&use_asid_allocator)) > + __sbi_tlb_flush_range_asid(mm_cpumask(mm), 0, -1, > + atomic_long_read(&mm->context.id)); > + else > + __sbi_tlb_flush_range(mm_cpumask(mm), 0, -1); > } > > void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr) > { > - __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE); > + if (static_branch_unlikely(&use_asid_allocator)) > + __sbi_tlb_flush_range_asid(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE, > + atomic_long_read(&vma->vm_mm->context.id)); > + else > + __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE); > } > > void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, > unsigned long end) > { > - __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), start, end - start); > + if (static_branch_unlikely(&use_asid_allocator)) > + __sbi_tlb_flush_range_asid(mm_cpumask(vma->vm_mm), start, end - start, > + atomic_long_read(&vma->vm_mm->context.id)); > + else > + __sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), start, end - start); > } > -- > 2.7.4 > ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL 2021-05-24 6:51 [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL guoren 2021-05-24 6:51 ` [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h guoren 2021-05-24 6:51 ` [PATCH 3/3] riscv: Use use_asid_allocator flush TLB guoren @ 2021-05-24 8:03 ` Anup Patel 2021-05-24 8:22 ` Guo Ren 2021-05-24 10:42 ` Anup Patel 3 siblings, 1 reply; 11+ messages in thread From: Anup Patel @ 2021-05-24 8:03 UTC (permalink / raw) To: Guo Ren Cc: Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, linux-kernel@vger.kernel.org List, linux-arch, linux-sunxi, Guo Ren On Mon, May 24, 2021 at 12:22 PM <guoren@kernel.org> wrote: > > From: Guo Ren <guoren@linux.alibaba.com> > > Kernel virtual address translation should avoid care asid or it'll > cause more TLB-miss and TLB-refill. Because the current asid in satp > belongs to the current process, but the target kernel va TLB entry's > asid still belongs to the previous process. > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com> > Cc: Anup Patel <anup.patel@wdc.com> > Cc: Palmer Dabbelt <palmerdabbelt@google.com> First of all thanks for doing this series, I had similar changes in mind as follow-up to the ASID allocator. I went through all three patches and at least I don't see any obvious issue but I think we should try testing it more on a few existing platforms. Reviewed-by: Anup Patel <anup@brainfault.org> Regards, Anup > --- > arch/riscv/include/asm/pgtable.h | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h > index 78f2323..017da15 100644 > --- a/arch/riscv/include/asm/pgtable.h > +++ b/arch/riscv/include/asm/pgtable.h > @@ -135,6 +135,7 @@ > | _PAGE_PRESENT \ > | _PAGE_ACCESSED \ > | _PAGE_DIRTY \ > + | _PAGE_GLOBAL \ > | _PAGE_CACHE) > > #define PAGE_KERNEL __pgprot(_PAGE_KERNEL) > -- > 2.7.4 > ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL 2021-05-24 8:03 ` [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL Anup Patel @ 2021-05-24 8:22 ` Guo Ren 0 siblings, 0 replies; 11+ messages in thread From: Guo Ren @ 2021-05-24 8:22 UTC (permalink / raw) To: Anup Patel Cc: Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, linux-kernel@vger.kernel.org List, linux-arch, linux-sunxi, Guo Ren Hi Anup, On Mon, May 24, 2021 at 4:03 PM Anup Patel <anup@brainfault.org> wrote: > > On Mon, May 24, 2021 at 12:22 PM <guoren@kernel.org> wrote: > > > > From: Guo Ren <guoren@linux.alibaba.com> > > > > Kernel virtual address translation should avoid care asid or it'll > > cause more TLB-miss and TLB-refill. Because the current asid in satp > > belongs to the current process, but the target kernel va TLB entry's > > asid still belongs to the previous process. > > > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com> > > Cc: Anup Patel <anup.patel@wdc.com> > > Cc: Palmer Dabbelt <palmerdabbelt@google.com> > > First of all thanks for doing this series, I had similar changes in mind > as follow-up to the ASID allocator. > > I went through all three patches and at least I don't see any > obvious issue but I think we should try testing it more on a few > existing platforms. We've tested it on Allwinner D1 C906 and C910 SMP*4 for a long time, just hope it won't affect u540. (In fact, C910 has used ASID allocator for more than two years with our own kernel-tree. I remember we've talked about it in 2019 plumber.) > > Reviewed-by: Anup Patel <anup@brainfault.org> > > Regards, > Anup > > > --- > > arch/riscv/include/asm/pgtable.h | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h > > index 78f2323..017da15 100644 > > --- a/arch/riscv/include/asm/pgtable.h > > +++ b/arch/riscv/include/asm/pgtable.h > > @@ -135,6 +135,7 @@ > > | _PAGE_PRESENT \ > > | _PAGE_ACCESSED \ > > | _PAGE_DIRTY \ > > + | _PAGE_GLOBAL \ > > | _PAGE_CACHE) > > > > #define PAGE_KERNEL __pgprot(_PAGE_KERNEL) > > -- > > 2.7.4 > > -- Best Regards Guo Ren ML: https://lore.kernel.org/linux-csky/ ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL 2021-05-24 6:51 [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL guoren ` (2 preceding siblings ...) 2021-05-24 8:03 ` [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL Anup Patel @ 2021-05-24 10:42 ` Anup Patel 2021-05-24 12:01 ` Guo Ren 3 siblings, 1 reply; 11+ messages in thread From: Anup Patel @ 2021-05-24 10:42 UTC (permalink / raw) To: Guo Ren Cc: Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, linux-kernel@vger.kernel.org List, linux-arch, linux-sunxi, Guo Ren On Mon, May 24, 2021 at 12:22 PM <guoren@kernel.org> wrote: > > From: Guo Ren <guoren@linux.alibaba.com> > > Kernel virtual address translation should avoid care asid or it'll > cause more TLB-miss and TLB-refill. Because the current asid in satp > belongs to the current process, but the target kernel va TLB entry's > asid still belongs to the previous process. > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com> > Cc: Anup Patel <anup.patel@wdc.com> > Cc: Palmer Dabbelt <palmerdabbelt@google.com> > --- > arch/riscv/include/asm/pgtable.h | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h > index 78f2323..017da15 100644 > --- a/arch/riscv/include/asm/pgtable.h > +++ b/arch/riscv/include/asm/pgtable.h > @@ -135,6 +135,7 @@ > | _PAGE_PRESENT \ > | _PAGE_ACCESSED \ > | _PAGE_DIRTY \ > + | _PAGE_GLOBAL \ > | _PAGE_CACHE) It seems this patch is not based on the upstream kernel. The _PAGE_CACHE seems to be from your other patch series. Please rebase these patches on the latest upstream kernel without dependency on any other patch series. Regards, Anup > > #define PAGE_KERNEL __pgprot(_PAGE_KERNEL) > -- > 2.7.4 > ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL 2021-05-24 10:42 ` Anup Patel @ 2021-05-24 12:01 ` Guo Ren 0 siblings, 0 replies; 11+ messages in thread From: Guo Ren @ 2021-05-24 12:01 UTC (permalink / raw) To: Anup Patel Cc: Anup Patel, Palmer Dabbelt, Arnd Bergmann, linux-riscv, linux-kernel@vger.kernel.org List, linux-arch, linux-sunxi, Guo Ren On Mon, May 24, 2021 at 6:42 PM Anup Patel <anup@brainfault.org> wrote: > > On Mon, May 24, 2021 at 12:22 PM <guoren@kernel.org> wrote: > > > > From: Guo Ren <guoren@linux.alibaba.com> > > > > Kernel virtual address translation should avoid care asid or it'll > > cause more TLB-miss and TLB-refill. Because the current asid in satp > > belongs to the current process, but the target kernel va TLB entry's > > asid still belongs to the previous process. > > > > Signed-off-by: Guo Ren <guoren@linux.alibaba.com> > > Cc: Anup Patel <anup.patel@wdc.com> > > Cc: Palmer Dabbelt <palmerdabbelt@google.com> > > --- > > arch/riscv/include/asm/pgtable.h | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h > > index 78f2323..017da15 100644 > > --- a/arch/riscv/include/asm/pgtable.h > > +++ b/arch/riscv/include/asm/pgtable.h > > @@ -135,6 +135,7 @@ > > | _PAGE_PRESENT \ > > | _PAGE_ACCESSED \ > > | _PAGE_DIRTY \ > > + | _PAGE_GLOBAL \ > > | _PAGE_CACHE) > > It seems this patch is not based on the upstream kernel. The > _PAGE_CACHE seems to be from your other patch series. > > Please rebase these patches on the latest upstream kernel without > dependency on any other patch series. Yes, it based on DMA_COHERENT. I'll rebase in PATCH V2, thx. > > Regards, > Anup > > > > > #define PAGE_KERNEL __pgprot(_PAGE_KERNEL) > > -- > > 2.7.4 > > -- Best Regards Guo Ren ML: https://lore.kernel.org/linux-csky/ ^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2021-05-25 9:35 UTC | newest] Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2021-05-24 6:51 [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL guoren 2021-05-24 6:51 ` [PATCH 2/3] riscv: Fixup PAGE_UP in asm/page.h guoren 2021-05-25 6:34 ` Christoph Hellwig 2021-05-25 9:28 ` Guo Ren 2021-05-25 9:34 ` Christoph Hellwig 2021-05-24 6:51 ` [PATCH 3/3] riscv: Use use_asid_allocator flush TLB guoren 2021-05-24 8:13 ` Anup Patel 2021-05-24 8:03 ` [PATCH 1/3] riscv: Fixup _PAGE_GLOBAL in _PAGE_KERNEL Anup Patel 2021-05-24 8:22 ` Guo Ren 2021-05-24 10:42 ` Anup Patel 2021-05-24 12:01 ` Guo Ren
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).