Re: [PATCH 1/2] userfaultfd: remove one unnecessary warn_on in __mcopy_atomic_hugetlb

[Wei Yang <richardw.yang@linux.intel.com>]

On Wed, Sep 25, 2019 at 07:10:46PM -0700, Mike Kravetz wrote:
>On 9/25/19 5:35 PM, Wei Yang wrote:
>> On Wed, Sep 25, 2019 at 10:44:58AM -0700, Mike Kravetz wrote:
>>> On 9/25/19 5:18 AM, Wei Yang wrote:
>>>> The warning here is to make sure address(dst_addr) and length(len -
>>>> copied) are huge page size aligned.
>>>>
>>>> While this is ensured by:
>>>>
>>>>     dst_start and len is huge page size aligned
>>>>     dst_addr equals to dst_start and increase huge page size each time
>>>>     copied increase huge page size each time
>>>
>>> Can we also remove the following for the same reasons?
>>>
>>> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
>>> index 640ff2bd9a69..f82d5ec698d8 100644
>>> --- a/mm/userfaultfd.c
>>> +++ b/mm/userfaultfd.c
>>> @@ -262,7 +262,6 @@ static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
>>> 		pte_t dst_pteval;
>>>
>>> 		BUG_ON(dst_addr >= dst_start + len);
>>> -		VM_BUG_ON(dst_addr & ~huge_page_mask(h));
>>>
>> 
>> Thanks for your comment.
>> 
>> It looks good, while I lack some knowledge between vma_hpagesize and
>> huge_page_mask().
>
>vma_hpagesize is just a local variable used so that repeated calls to
>vma_kernel_pagesize() or huge_page_size() are not necessary.
>

Thanks for your confirmation. If this is the case, we can remove this BUG_ON
safely.

>> If they are the same, why not use the same interface for all those checks in
>> this function?
>
>If we remove the VM_BUG_ON, that is the only use of huge_page_mask() in
>the function.
>
>We can can also eliminate a call to huge_page_size() by making this change.
>
>@@ -273,7 +272,7 @@ static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
> 		mutex_lock(&hugetlb_fault_mutex_table[hash]);
> 
> 		err = -ENOMEM;
>-		dst_pte = huge_pte_alloc(dst_mm, dst_addr, huge_page_size(h));
>+		dst_pte = huge_pte_alloc(dst_mm, dst_addr, vma_hpagesize);
> 		if (!dst_pte) {
> 			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
> 			goto out_unlock;

Agree, and also with this I think

diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index c153344774c7..74363f0a0dd0 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -315,7 +315,7 @@ static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
 
                        err = copy_huge_page_from_user(page,
                                                (const void __user *)src_addr,
-                                               pages_per_huge_page(h), true);
+                                               vma_hpagesize / PAGE_SIZE, true);
                        if (unlikely(err)) {
                                err = -EFAULT;
                                goto out;

After these cleanup, we use vma_pagesize to deal with all page size related
calculation in this function, which looks more consistent to me.

Does it looks good to you?

>-- 
>Mike Kravetz

-- 
Wei Yang
Help you, Help me