Re: [PATCH] parisc: Use the VMA iterator in the cache handling

[Helge Deller <deller@gmx.de>]

On 2/18/22 21:34, Liam Howlett wrote:
> This patch is on top of the parisc branch for-next.
>
> Helge Deller identified a conflict between the maple tree next branch
> and the parisc next branch.

Actually, Stephen noticed and reported it with the linux-next branch...

> This patch removes the vma linked list
> usage in the parisc next branch identified by Helge as a conflict to
> maple tree "parisc: Remove mmap linked list from cache handling" and the
> parisc "parisc: Add vDSO support" patch.  I do not expect this to apply
> cleanly after the maple tree next merge, but the resolution should be
> made clear by this patch.
>
> This is an untested conflict resolution for linux-next.

Thanks for doing this, Liam!

I'd propose to keep for now (for the linux-next tree) the conflict resolution
fix which Stephen has in place already, and then figure out when the
merge window opens how to resolve it - either you or me or fixing afterwards
with this patch.

Would that be ok?

Helge

> Signed-off-by: Liam R. Howlett <Liam.Howlett@oracle.com>
> ---
>  arch/parisc/kernel/cache.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/arch/parisc/kernel/cache.c b/arch/parisc/kernel/cache.c
> index e7b8e74dad8e..05b4498a1244 100644
> --- a/arch/parisc/kernel/cache.c
> +++ b/arch/parisc/kernel/cache.c
> @@ -515,12 +515,12 @@ void flush_cache_all(void)
>  	on_each_cpu(cacheflush_h_tmp_function, NULL, 1);
>  }
>
> -static inline unsigned long mm_total_size(struct mm_struct *mm)
> +static inline unsigned long mm_total_size(struct vma_iterator *vmi)
>  {
>  	struct vm_area_struct *vma;
>  	unsigned long usize = 0;
>
> -	for (vma = mm->mmap; vma; vma = vma->vm_next)
> +	for_each_vma(*vmi, vma)
>  		usize += vma->vm_end - vma->vm_start;
>  	return usize;
>  }
> @@ -570,11 +570,12 @@ static void flush_user_cache_tlb(struct vm_area_struct *vma,
>  void flush_cache_mm(struct mm_struct *mm)
>  {
>  	struct vm_area_struct *vma;
> +	VMA_ITERATOR(vmi, mm, 0);
>
>  	/* Flushing the whole cache on each cpu takes forever on
>  	   rp3440, etc.  So, avoid it if the mm isn't too big.  */
>  	if ((!IS_ENABLED(CONFIG_SMP) || !arch_irqs_disabled()) &&
> -	    mm_total_size(mm) >= parisc_cache_flush_threshold) {
> +	    mm_total_size(&vmi) >= parisc_cache_flush_threshold) {
>  		if (mm->context.space_id)
>  			flush_tlb_all();
>  		flush_cache_all();
> @@ -583,13 +584,13 @@ void flush_cache_mm(struct mm_struct *mm)
>
>  	preempt_disable();
>  	if (mm->context.space_id == mfsp(3)) {
> -		for (vma = mm->mmap; vma; vma = vma->vm_next)
> +		for_each_vma(vmi, vma)
>  			flush_user_cache_tlb(vma, vma->vm_start, vma->vm_end);
>  		preempt_enable();
>  		return;
>  	}
>
> -	for (vma = mm->mmap; vma; vma = vma->vm_next)
> +	for_each_vma(vmi, vma)
>  		flush_cache_pages(vma, mm, vma->vm_start, vma->vm_end);
>  	preempt_enable();
>  }