Re: [PATCH v9 12/21] mm: pagewalk: Allow walking without vma

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 28/07/2019 15:20, Anshuman Khandual wrote:
> 
> 
> On 07/22/2019 09:12 PM, Steven Price wrote:
>> Since 48684a65b4e3: "mm: pagewalk: fix misbehavior of walk_page_range
>> for vma(VM_PFNMAP)", page_table_walk() will report any kernel area as
>> a hole, because it lacks a vma.
>>
>> This means each arch has re-implemented page table walking when needed,
>> for example in the per-arch ptdump walker.
>>
>> Remove the requirement to have a vma except when trying to split huge
>> pages.
>>
>> Signed-off-by: Steven Price <steven.price@xxxxxxx>
>> ---
>>  mm/pagewalk.c | 25 +++++++++++++++++--------
>>  1 file changed, 17 insertions(+), 8 deletions(-)
>>
>> diff --git a/mm/pagewalk.c b/mm/pagewalk.c
>> index 98373a9f88b8..1cbef99e9258 100644
>> --- a/mm/pagewalk.c
>> +++ b/mm/pagewalk.c
>> @@ -36,7 +36,7 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
>>  	do {
>>  again:
>>  		next = pmd_addr_end(addr, end);
>> -		if (pmd_none(*pmd) || !walk->vma) {
>> +		if (pmd_none(*pmd)) {
>>  			if (walk->pte_hole)
>>  				err = walk->pte_hole(addr, next, walk);
>>  			if (err)
>> @@ -59,9 +59,14 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
>>  		if (!walk->pte_entry)
>>  			continue;
>>  
>> -		split_huge_pmd(walk->vma, pmd, addr);
>> -		if (pmd_trans_unstable(pmd))
>> -			goto again;
>> +		if (walk->vma) {
>> +			split_huge_pmd(walk->vma, pmd, addr);
> 
> Check for a PMD THP entry before attempting to split it ?

split_huge_pmd does the check for us:
> #define split_huge_pmd(__vma, __pmd, __address)				\
> 	do {								\
> 		pmd_t *____pmd = (__pmd);				\
> 		if (is_swap_pmd(*____pmd) || pmd_trans_huge(*____pmd)	\
> 					|| pmd_devmap(*____pmd))	\
> 			__split_huge_pmd(__vma, __pmd, __address,	\
> 						false, NULL);		\
> 	}  while (0)

And this isn't a change from the previous code - only that the entry is
no longer split when walk->vma==NULL.

>> +			if (pmd_trans_unstable(pmd))
>> +				goto again;
>> +		} else if (pmd_leaf(*pmd)) {
>> +			continue;
>> +		}
>> +
>>  		err = walk_pte_range(pmd, addr, next, walk);
>>  		if (err)
>>  			break;
>> @@ -81,7 +86,7 @@ static int walk_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end,
>>  	do {
>>   again:
>>  		next = pud_addr_end(addr, end);
>> -		if (pud_none(*pud) || !walk->vma) {
>> +		if (pud_none(*pud)) {
>>  			if (walk->pte_hole)
>>  				err = walk->pte_hole(addr, next, walk);
>>  			if (err)
>> @@ -95,9 +100,13 @@ static int walk_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end,
>>  				break;
>>  		}
>>  
>> -		split_huge_pud(walk->vma, pud, addr);
>> -		if (pud_none(*pud))
>> -			goto again;
>> +		if (walk->vma) {
>> +			split_huge_pud(walk->vma, pud, addr);
> 
> Check for a PUD THP entry before attempting to split it ?

Same as above.

>> +			if (pud_none(*pud))
>> +				goto again;
>> +		} else if (pud_leaf(*pud)) {
>> +			continue;
>> +		}
> 
> This is bit cryptic. walk->vma check should be inside a helper is_user_page_table()
> or similar to make things clear. p4d_leaf() check missing in walk_p4d_range() for
> kernel page table walk ? Wondering if p?d_leaf() test should be moved earlier while
> calling p?d_entry() for kernel page table walk.

I wasn't sure if it was worth putting p4d_leaf() and pgd_leaf() checks
in (yet). No architecture that I know of uses such large pages.

I'm not sure what you mean by moving the p?d_leaf() test earlier? Can
you explain with an example?

Thanks,

Steve




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux