| Message ID | 20220929112318.32393-3-zhengqi.arch@bytedance.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | use update_mmu_tlb() on the second thread | expand |
Hi, all, Should this patch go via mm tree or loongarch tree? If via mm tree, then Acked-by: Huacai Chen <chenhuacai@loongson.cn> On Thu, Sep 29, 2022 at 7:23 PM Qi Zheng <zhengqi.arch@bytedance.com> wrote: > > Currently, the implementation of update_mmu_tlb() is empty if > __HAVE_ARCH_UPDATE_MMU_TLB is not defined. Then if two threads > concurrently fault at the same page, the second thread that did > not win the race will give up and do nothing. In the LoongArch > architecture, this second thread will trigger another fault, > and only updates its local TLB. > > Instead of triggering another fault, it's better to implement > update_mmu_tlb() to directly update the local TLB of the second > thread. Just do it. > > Suggested-by: Bibo Mao <maobibo@loongson.cn> > Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com> > --- > arch/loongarch/include/asm/pgtable.h | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/arch/loongarch/include/asm/pgtable.h b/arch/loongarch/include/asm/pgtable.h > index 8ea57e2f0e04..946704bee599 100644 > --- a/arch/loongarch/include/asm/pgtable.h > +++ b/arch/loongarch/include/asm/pgtable.h > @@ -412,6 +412,9 @@ static inline void update_mmu_cache(struct vm_area_struct *vma, > __update_tlb(vma, address, ptep); > } > > +#define __HAVE_ARCH_UPDATE_MMU_TLB > +#define update_mmu_tlb update_mmu_cache > + > static inline void update_mmu_cache_pmd(struct vm_area_struct *vma, > unsigned long address, pmd_t *pmdp) > { > -- > 2.20.1 > >
On 2022/9/29 19:42, Huacai Chen wrote: > Hi, all, > > Should this patch go via mm tree or loongarch tree? If via mm tree, then Both are fine for me. Hi Andrew, can you help to apply this patch series? > > Acked-by: Huacai Chen <chenhuacai@loongson.cn> Thanks. :) > > On Thu, Sep 29, 2022 at 7:23 PM Qi Zheng <zhengqi.arch@bytedance.com> wrote: >> >> Currently, the implementation of update_mmu_tlb() is empty if >> __HAVE_ARCH_UPDATE_MMU_TLB is not defined. Then if two threads >> concurrently fault at the same page, the second thread that did >> not win the race will give up and do nothing. In the LoongArch >> architecture, this second thread will trigger another fault, >> and only updates its local TLB. >> >> Instead of triggering another fault, it's better to implement >> update_mmu_tlb() to directly update the local TLB of the second >> thread. Just do it. >> >> Suggested-by: Bibo Mao <maobibo@loongson.cn> >> Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com> >> --- >> arch/loongarch/include/asm/pgtable.h | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/arch/loongarch/include/asm/pgtable.h b/arch/loongarch/include/asm/pgtable.h >> index 8ea57e2f0e04..946704bee599 100644 >> --- a/arch/loongarch/include/asm/pgtable.h >> +++ b/arch/loongarch/include/asm/pgtable.h >> @@ -412,6 +412,9 @@ static inline void update_mmu_cache(struct vm_area_struct *vma, >> __update_tlb(vma, address, ptep); >> } >> >> +#define __HAVE_ARCH_UPDATE_MMU_TLB >> +#define update_mmu_tlb update_mmu_cache >> + >> static inline void update_mmu_cache_pmd(struct vm_area_struct *vma, >> unsigned long address, pmd_t *pmdp) >> { >> -- >> 2.20.1 >> >>
diff --git a/arch/loongarch/include/asm/pgtable.h b/arch/loongarch/include/asm/pgtable.h index 8ea57e2f0e04..946704bee599 100644 --- a/arch/loongarch/include/asm/pgtable.h +++ b/arch/loongarch/include/asm/pgtable.h @@ -412,6 +412,9 @@ static inline void update_mmu_cache(struct vm_area_struct *vma, __update_tlb(vma, address, ptep); } +#define __HAVE_ARCH_UPDATE_MMU_TLB +#define update_mmu_tlb update_mmu_cache + static inline void update_mmu_cache_pmd(struct vm_area_struct *vma, unsigned long address, pmd_t *pmdp) {
Currently, the implementation of update_mmu_tlb() is empty if __HAVE_ARCH_UPDATE_MMU_TLB is not defined. Then if two threads concurrently fault at the same page, the second thread that did not win the race will give up and do nothing. In the LoongArch architecture, this second thread will trigger another fault, and only updates its local TLB. Instead of triggering another fault, it's better to implement update_mmu_tlb() to directly update the local TLB of the second thread. Just do it. Suggested-by: Bibo Mao <maobibo@loongson.cn> Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com> --- arch/loongarch/include/asm/pgtable.h | 3 +++ 1 file changed, 3 insertions(+)