Re: [PATCH v2 3/4] powerpc/mm/radix: Avoid flushing the PWC on every flush_tlb_range

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

Benjamin Herrenschmidt <benh@kernel.crashing.org> writes:

> We do that because it's used by THP pmd collapsing, so use
> instead a dedicated flush function.

Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
>
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
>
> v2. Add missing #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> ---
>  .../powerpc/include/asm/book3s/64/tlbflush-radix.h |  1 +
>  arch/powerpc/mm/pgtable-radix.c                    |  5 ++-
>  arch/powerpc/mm/tlb-radix.c                        | 43 +++++++++++++++++++---
>  3 files changed, 43 insertions(+), 6 deletions(-)
>
> diff --git a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h b/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
> index 7196999cdc82..9b433a624bf3 100644
> --- a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
> +++ b/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
> @@ -36,6 +36,7 @@ extern void radix__flush_tlb_page_psize(struct mm_struct *mm, unsigned long vmad
>  #define radix__flush_tlb_page_psize(mm,addr,p) radix__local_flush_tlb_page_psize(mm,addr,p)
>  #endif
>  extern void radix__flush_tlb_pwc(struct mmu_gather *tlb, unsigned long addr);
> +extern void radix__flush_tlb_collapsed_pmd(struct mm_struct *mm, unsigned long addr);
>  extern void radix__flush_tlb_lpid_va(unsigned long lpid, unsigned long gpa,
>  				     unsigned long page_size);
>  extern void radix__flush_tlb_lpid(unsigned long lpid);
> diff --git a/arch/powerpc/mm/pgtable-radix.c b/arch/powerpc/mm/pgtable-radix.c
> index 8c13e4282308..02134c95a956 100644
> --- a/arch/powerpc/mm/pgtable-radix.c
> +++ b/arch/powerpc/mm/pgtable-radix.c
> @@ -786,9 +786,12 @@ pmd_t radix__pmdp_collapse_flush(struct vm_area_struct *vma, unsigned long addre
>  	 */
>  	pmd = *pmdp;
>  	pmd_clear(pmdp);
> +
>  	/*FIXME!!  Verify whether we need this kick below */
>  	kick_all_cpus_sync();
> -	flush_tlb_range(vma, address, address + HPAGE_PMD_SIZE);
> +
> +	radix__flush_tlb_collapsed_pmd(vma->vm_mm, address);
> +
>  	return pmd;
>  }
>
> diff --git a/arch/powerpc/mm/tlb-radix.c b/arch/powerpc/mm/tlb-radix.c
> index 28f339cdd836..18151e9ad694 100644
> --- a/arch/powerpc/mm/tlb-radix.c
> +++ b/arch/powerpc/mm/tlb-radix.c
> @@ -272,11 +272,7 @@ void radix__flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
>  {
>  	struct mm_struct *mm = vma->vm_mm;
>
> -	/*
> -	 * This is currently used when collapsing THPs so we need to
> -	 * flush the PWC. We should fix this.
> -	 */
> -	radix__flush_all_mm(mm);
> +	radix__flush_tlb_mm(mm);
>  }
>  EXPORT_SYMBOL(radix__flush_tlb_range);
>
> @@ -355,6 +351,43 @@ void radix__flush_tlb_range_psize(struct mm_struct *mm, unsigned long start,
>  	preempt_enable();
>  }
>
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +void radix__flush_tlb_collapsed_pmd(struct mm_struct *mm, unsigned long addr)
> +{
> +	int local = mm_is_thread_local(mm);
> +	unsigned long ap = mmu_get_ap(mmu_virtual_psize);
> +	unsigned long pid, end;
> +
> +
> +	pid = mm ? mm->context.id : 0;
> +	if (unlikely(pid == MMU_NO_CONTEXT))
> +		goto no_context;
> +
> +	/* 4k page size, just blow the world */
> +	if (PAGE_SIZE == 0x1000) {
> +		radix__flush_all_mm(mm);
> +		return;
> +	}
> +
> +	/* Otherwise first do the PWC */
> +	if (local)
> +		_tlbiel_pid(pid, RIC_FLUSH_PWC);
> +	else
> +		_tlbie_pid(pid, RIC_FLUSH_PWC);
> +
> +	/* Then iterate the pages */
> +	end = addr + HPAGE_PMD_SIZE;
> +	for (; addr < end; addr += PAGE_SIZE) {
> +		if (local)
> +			_tlbiel_va(addr, pid, ap, RIC_FLUSH_TLB);
> +		else
> +			_tlbie_va(addr, pid, ap, RIC_FLUSH_TLB);
> +	}
> +no_context:
> +	preempt_enable();
> +}
> +#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> +
>  void radix__flush_tlb_lpid_va(unsigned long lpid, unsigned long gpa,
>  			      unsigned long page_size)
>  {
> -- 
> 2.13.3