Re: [PATCH] /proc/PID/smaps: Add PMD migration entry parsing

[Konstantin Khlebnikov <khlebnikov@yandex-team.ru>]

On 01/04/2020 05.31, Huang, Ying wrote:
> Konstantin Khlebnikov <khlebnikov@yandex-team.ru> writes:
> 
>> On 31/03/2020 11.56, Huang, Ying wrote:
>>> From: Huang Ying <ying.huang@intel.com>
>>>
>>> Now, when read /proc/PID/smaps, the PMD migration entry in page table is simply
>>> ignored.  To improve the accuracy of /proc/PID/smaps, its parsing and processing
>>> is added.
>>>
>>> Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
>>> Cc: Andrea Arcangeli <aarcange@redhat.com>
>>> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>>> Cc: Zi Yan <ziy@nvidia.com>
>>> Cc: Vlastimil Babka <vbabka@suse.cz>
>>> Cc: Alexey Dobriyan <adobriyan@gmail.com>
>>> Cc: Michal Hocko <mhocko@suse.com>
>>> Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
>>> Cc: "Jérôme Glisse" <jglisse@redhat.com>
>>> Cc: Yang Shi <yang.shi@linux.alibaba.com>
>>> ---
>>>    fs/proc/task_mmu.c | 16 ++++++++++++----
>>>    1 file changed, 12 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
>>> index 8d382d4ec067..b5b3aef8cb3b 100644
>>> --- a/fs/proc/task_mmu.c
>>> +++ b/fs/proc/task_mmu.c
>>> @@ -548,8 +548,17 @@ static void smaps_pmd_entry(pmd_t *pmd, unsigned long addr,
>>>    	bool locked = !!(vma->vm_flags & VM_LOCKED);
>>>    	struct page *page;
>>
>>          struct page *page = NULL;
> 
> Looks good.  Will do this in the next version.
> 
>>>    -	/* FOLL_DUMP will return -EFAULT on huge zero page */
>>> -	page = follow_trans_huge_pmd(vma, addr, pmd, FOLL_DUMP);
>>> +	if (pmd_present(*pmd)) {
>>> +		/* FOLL_DUMP will return -EFAULT on huge zero page */
>>> +		page = follow_trans_huge_pmd(vma, addr, pmd, FOLL_DUMP);
>>> +	} else if (unlikely(is_swap_pmd(*pmd))) {
>>> +		swp_entry_t entry = pmd_to_swp_entry(*pmd);
>>> +
>>> +		VM_BUG_ON(!is_migration_entry(entry));
>>> +		page = migration_entry_to_page(entry);
>>
>>                  if (is_migration_entry(entry))
>>                          page = migration_entry_to_page(entry);
>>
>> Seems safer and doesn't add much code.
> 
> With this, we lose an opportunity to capture some bugs during debugging.
> Right?

You can keep VM_BUG_ON or VM_WARN_ON_ONCE

Off-by-page in statistics isn't a big deal and not a good reason to crash (even debug) kernel.
But for normal build should use safe behaviour if this isn't hard.

> 
> Best Regards,
> Huang, Ying
> 
>>> +	} else {
>>> +		return;
>>> +	}
>>>    	if (IS_ERR_OR_NULL(page))
>>>    		return;
>>>    	if (PageAnon(page))
>>> @@ -578,8 +587,7 @@ static int smaps_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
>>>      	ptl = pmd_trans_huge_lock(pmd, vma);
>>>    	if (ptl) {
>>> -		if (pmd_present(*pmd))
>>> -			smaps_pmd_entry(pmd, addr, walk);
>>> +		smaps_pmd_entry(pmd, addr, walk);
>>>    		spin_unlock(ptl);
>>>    		goto out;
>>>    	}
>>>