| Message ID | 20240102220134.3229156-13-samuel.holland@sifive.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | riscv: ASID-related and UP-related TLB flush enhancements | expand |
On Tue, Jan 02, 2024 at 02:00:49PM -0800, Samuel Holland wrote: > Even if multiple ASIDs are not supported, using the single-ASID variant > of the sfence.vma instruction preserves TLB entries for global (kernel) > pages. So it is always more efficient to use the single-ASID code path. > > Signed-off-by: Samuel Holland <samuel.holland@sifive.com> > --- > > Changes in v4: > - There is now only one copy of __flush_tlb_range() > > Changes in v2: > - Update both copies of __flush_tlb_range() > > arch/riscv/include/asm/mmu_context.h | 2 -- > arch/riscv/mm/context.c | 3 +-- > arch/riscv/mm/tlbflush.c | 3 +-- > 3 files changed, 2 insertions(+), 6 deletions(-) > > diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h > index 7030837adc1a..b0659413a080 100644 > --- a/arch/riscv/include/asm/mmu_context.h > +++ b/arch/riscv/include/asm/mmu_context.h > @@ -33,8 +33,6 @@ 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/mm/context.c b/arch/riscv/mm/context.c > index 3ca9b653df7d..20057085ab8a 100644 > --- a/arch/riscv/mm/context.c > +++ b/arch/riscv/mm/context.c > @@ -18,8 +18,7 @@ > > #ifdef CONFIG_MMU > > -DEFINE_STATIC_KEY_FALSE(use_asid_allocator); > - > +static DEFINE_STATIC_KEY_FALSE(use_asid_allocator); One of my optimization "riscv: tlb: avoid tlb flushing if fullmm == 1" will make use of use_asid_allocator, so could we remove this modification? > static unsigned long num_asids; > > static atomic_long_t current_version; > diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c > index 5ec621545c69..39d80f56d292 100644 > --- a/arch/riscv/mm/tlbflush.c > +++ b/arch/riscv/mm/tlbflush.c > @@ -84,8 +84,7 @@ static void __flush_tlb_range(struct mm_struct *mm, unsigned long start, > if (cpumask_empty(cmask)) > return; > > - if (static_branch_unlikely(&use_asid_allocator)) > - asid = cntx2asid(atomic_long_read(&mm->context.id)); > + asid = cntx2asid(atomic_long_read(&mm->context.id)); > } else { > cmask = cpu_online_mask; > } > -- > 2.42.0 > > > _______________________________________________ > linux-riscv mailing list > linux-riscv@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-riscv
On Tue, Jan 2, 2024 at 11:01 PM Samuel Holland <samuel.holland@sifive.com> wrote: > > Even if multiple ASIDs are not supported, using the single-ASID variant > of the sfence.vma instruction preserves TLB entries for global (kernel) > pages. So it is always more efficient to use the single-ASID code path. > > Signed-off-by: Samuel Holland <samuel.holland@sifive.com> > --- > > Changes in v4: > - There is now only one copy of __flush_tlb_range() > > Changes in v2: > - Update both copies of __flush_tlb_range() > > arch/riscv/include/asm/mmu_context.h | 2 -- > arch/riscv/mm/context.c | 3 +-- > arch/riscv/mm/tlbflush.c | 3 +-- > 3 files changed, 2 insertions(+), 6 deletions(-) > > diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h > index 7030837adc1a..b0659413a080 100644 > --- a/arch/riscv/include/asm/mmu_context.h > +++ b/arch/riscv/include/asm/mmu_context.h > @@ -33,8 +33,6 @@ 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/mm/context.c b/arch/riscv/mm/context.c > index 3ca9b653df7d..20057085ab8a 100644 > --- a/arch/riscv/mm/context.c > +++ b/arch/riscv/mm/context.c > @@ -18,8 +18,7 @@ > > #ifdef CONFIG_MMU > > -DEFINE_STATIC_KEY_FALSE(use_asid_allocator); > - > +static DEFINE_STATIC_KEY_FALSE(use_asid_allocator); > static unsigned long num_asids; > > static atomic_long_t current_version; > diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c > index 5ec621545c69..39d80f56d292 100644 > --- a/arch/riscv/mm/tlbflush.c > +++ b/arch/riscv/mm/tlbflush.c > @@ -84,8 +84,7 @@ static void __flush_tlb_range(struct mm_struct *mm, unsigned long start, > if (cpumask_empty(cmask)) > return; > > - if (static_branch_unlikely(&use_asid_allocator)) > - asid = cntx2asid(atomic_long_read(&mm->context.id)); > + asid = cntx2asid(atomic_long_read(&mm->context.id)); > } else { > cmask = cpu_online_mask; > } > -- > 2.42.0 > You can add: Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com> Thanks! Alex
On 2024-01-03 9:02 AM, Jisheng Zhang wrote: > On Tue, Jan 02, 2024 at 02:00:49PM -0800, Samuel Holland wrote: >> Even if multiple ASIDs are not supported, using the single-ASID variant >> of the sfence.vma instruction preserves TLB entries for global (kernel) >> pages. So it is always more efficient to use the single-ASID code path. >> >> Signed-off-by: Samuel Holland <samuel.holland@sifive.com> >> --- >> >> Changes in v4: >> - There is now only one copy of __flush_tlb_range() >> >> Changes in v2: >> - Update both copies of __flush_tlb_range() >> >> arch/riscv/include/asm/mmu_context.h | 2 -- >> arch/riscv/mm/context.c | 3 +-- >> arch/riscv/mm/tlbflush.c | 3 +-- >> 3 files changed, 2 insertions(+), 6 deletions(-) >> >> diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h >> index 7030837adc1a..b0659413a080 100644 >> --- a/arch/riscv/include/asm/mmu_context.h >> +++ b/arch/riscv/include/asm/mmu_context.h >> @@ -33,8 +33,6 @@ 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/mm/context.c b/arch/riscv/mm/context.c >> index 3ca9b653df7d..20057085ab8a 100644 >> --- a/arch/riscv/mm/context.c >> +++ b/arch/riscv/mm/context.c >> @@ -18,8 +18,7 @@ >> >> #ifdef CONFIG_MMU >> >> -DEFINE_STATIC_KEY_FALSE(use_asid_allocator); >> - >> +static DEFINE_STATIC_KEY_FALSE(use_asid_allocator); > > One of my optimization "riscv: tlb: avoid tlb flushing if fullmm == 1" > will make use of use_asid_allocator, so could we remove this modification? Yes, I can leave the global declaration alone for now. >> static unsigned long num_asids; >> >> static atomic_long_t current_version; >> diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c >> index 5ec621545c69..39d80f56d292 100644 >> --- a/arch/riscv/mm/tlbflush.c >> +++ b/arch/riscv/mm/tlbflush.c >> @@ -84,8 +84,7 @@ static void __flush_tlb_range(struct mm_struct *mm, unsigned long start, >> if (cpumask_empty(cmask)) >> return; >> >> - if (static_branch_unlikely(&use_asid_allocator)) >> - asid = cntx2asid(atomic_long_read(&mm->context.id)); >> + asid = cntx2asid(atomic_long_read(&mm->context.id)); >> } else { >> cmask = cpu_online_mask; >> } >> -- >> 2.42.0 >> >> >> _______________________________________________ >> linux-riscv mailing list >> linux-riscv@lists.infradead.org >> http://lists.infradead.org/mailman/listinfo/linux-riscv
diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h index 7030837adc1a..b0659413a080 100644 --- a/arch/riscv/include/asm/mmu_context.h +++ b/arch/riscv/include/asm/mmu_context.h @@ -33,8 +33,6 @@ 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/mm/context.c b/arch/riscv/mm/context.c index 3ca9b653df7d..20057085ab8a 100644 --- a/arch/riscv/mm/context.c +++ b/arch/riscv/mm/context.c @@ -18,8 +18,7 @@ #ifdef CONFIG_MMU -DEFINE_STATIC_KEY_FALSE(use_asid_allocator); - +static DEFINE_STATIC_KEY_FALSE(use_asid_allocator); static unsigned long num_asids; static atomic_long_t current_version; diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c index 5ec621545c69..39d80f56d292 100644 --- a/arch/riscv/mm/tlbflush.c +++ b/arch/riscv/mm/tlbflush.c @@ -84,8 +84,7 @@ static void __flush_tlb_range(struct mm_struct *mm, unsigned long start, if (cpumask_empty(cmask)) return; - if (static_branch_unlikely(&use_asid_allocator)) - asid = cntx2asid(atomic_long_read(&mm->context.id)); + asid = cntx2asid(atomic_long_read(&mm->context.id)); } else { cmask = cpu_online_mask; }
Even if multiple ASIDs are not supported, using the single-ASID variant of the sfence.vma instruction preserves TLB entries for global (kernel) pages. So it is always more efficient to use the single-ASID code path. Signed-off-by: Samuel Holland <samuel.holland@sifive.com> --- Changes in v4: - There is now only one copy of __flush_tlb_range() Changes in v2: - Update both copies of __flush_tlb_range() arch/riscv/include/asm/mmu_context.h | 2 -- arch/riscv/mm/context.c | 3 +-- arch/riscv/mm/tlbflush.c | 3 +-- 3 files changed, 2 insertions(+), 6 deletions(-)