Re: [PATCH v11 14/22] mm: pagewalk: Add 'depth' parameter to pte_hole

[Steven Price]

On 07/10/2019 17:10, Jason Gunthorpe wrote:
> On Mon, Oct 07, 2019 at 04:38:14PM +0100, Steven Price wrote:
>> diff --git a/mm/hmm.c b/mm/hmm.c
>> index 902f5fa6bf93..34fe904dd417 100644
>> +++ b/mm/hmm.c
>> @@ -376,7 +376,7 @@ static void hmm_range_need_fault(const struct hmm_vma_walk *hmm_vma_walk,
>>  }
>>  
>>  static int hmm_vma_walk_hole(unsigned long addr, unsigned long end,
>> -			     struct mm_walk *walk)
>> +			     __always_unused int depth, struct mm_walk *walk)
> 
> It this __always_unused on function arguments something we are doing
> now?

$ git grep __always_unused | wc -l
191

It's elsewhere in the kernel tree. It seems like a good way of both
documenting and silencing compiler warnings. But I'm open to other
suggestions.

> Can we have negative depth? Should it be unsigned?

As per the documentation added in this patch:

 * @pte_hole:	if set, called for each hole at all levels,
 *		depth is -1 if not known, 0:PGD, 1:P4D, 2:PUD, 3:PMD
 *		4:PTE. Any folded depths (where PTRS_PER_P?D is equal
 *		to 1) are skipped.

So it's signed to allow "-1" in the cases where pte_hole is called
without knowing the actual depth. This is used in the function
walk_page_test() because it don't walk the actual page tables, but is
called on a VMA instead. This means that there may not be a single depth
for the range provided.

Steve

>>  {
>>  	struct hmm_vma_walk *hmm_vma_walk = walk->private;
>>  	struct hmm_range *range = hmm_vma_walk->range;
>> @@ -564,7 +564,7 @@ static int hmm_vma_walk_pmd(pmd_t *pmdp,
>>  again:
>>  	pmd = READ_ONCE(*pmdp);
>>  	if (pmd_none(pmd))
>> -		return hmm_vma_walk_hole(start, end, walk);
>> +		return hmm_vma_walk_hole(start, end, 0, walk);
>>  
>>  	if (thp_migration_supported() && is_pmd_migration_entry(pmd)) {
>>  		bool fault, write_fault;
>> @@ -666,7 +666,7 @@ static int hmm_vma_walk_pud(pud_t *pudp, unsigned long start, unsigned long end,
>>  again:
>>  	pud = READ_ONCE(*pudp);
>>  	if (pud_none(pud))
>> -		return hmm_vma_walk_hole(start, end, walk);
>> +		return hmm_vma_walk_hole(start, end, 0, walk);
>>  
>>  	if (pud_huge(pud) && pud_devmap(pud)) {
>>  		unsigned long i, npages, pfn;
>> @@ -674,7 +674,7 @@ static int hmm_vma_walk_pud(pud_t *pudp, unsigned long start, unsigned long end,
>>  		bool fault, write_fault;
>>  
>>  		if (!pud_present(pud))
>> -			return hmm_vma_walk_hole(start, end, walk);
>> +			return hmm_vma_walk_hole(start, end, 0, walk);
>>  
>>  		i = (addr - range->start) >> PAGE_SHIFT;
>>  		npages = (end - addr) >> PAGE_SHIFT;
> 
> Otherwise this mechanical change to hmm.c looks OK to me
> 
> Jason
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>