linux-kernel - Re: [PATCH] mm/mmap: remove the mm parameter in vma

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <7229eb6e-bf3e-fa97-6709-6c92bc72443b@linux.dev>
Date: Thu, 22 Feb 2024 18:26:18 +0800
From: Yajun Deng <yajun.deng@...ux.dev>
To: "Liam R. Howlett" <Liam.Howlett@...cle.com>
Cc: akpm@...ux-foundation.org, linux-mm@...ck.org,
 linux-kernel@...r.kernel.org, vbabka@...e.cz, lstoakes@...il.com
Subject: Re: [PATCH] mm/mmap: remove the mm parameter in vma_complete()

Adding Vlastimil and Lorenzo to discuss this patch.


On 2024/1/29 23:04, Liam R. Howlett wrote:
> * Yajun Deng <yajun.deng@...ux.dev> [240129 02:53]:
>> There are vma_merge() and do_brk_flags() pass mm to vma_complete(), others
>> would pass the vma->vm_mm. The following explains that the mm is the
>> vma->vm_mm in vma_merge() and do_brk_flags().
>>
>> All vma will point to the same mm struct if the vma_merge() is successful.
>> So the mm and the vma->mm are the same.
> Absolutely, they must be the same.  I don't think vma_merge() checks
> this, but it is true.
>
>> vm_brk_flags() and brk syscall will initialize vmi with current->mm,
>> so the vma->vm_mm and the current->mm are the same if vma exists in
>> do_brk_flags().
>>
>> Remove the mm parameter in vma_complete() and get mm from the vma in vp.
> You have added a dereference to the two paths that don't need it to
> reduce the argument list from 3 to 2.  It's the same number of lines as
> well.  vma_shrink() is only used on process creation, but brk is more
> common.  Note that this function is marked as inline.
>
> I'm not sure this change is worth making.

If we can make sure the mm is  vma->vm_mm, I don't think we need to pass 
the mm.

If we can't make sure that, this change is not worth it.

>> Signed-off-by: Yajun Deng <yajun.deng@...ux.dev>
>> ---
>>   mm/mmap.c | 16 ++++++++--------
>>   1 file changed, 8 insertions(+), 8 deletions(-)
>>
>> diff --git a/mm/mmap.c b/mm/mmap.c
>> index e97b9144c61a..9b968d1edf55 100644
>> --- a/mm/mmap.c
>> +++ b/mm/mmap.c
>> @@ -509,11 +509,11 @@ static inline void vma_prepare(struct vma_prepare *vp)
>>    *
>>    * @vp: The vma_prepare struct
>>    * @vmi: The vma iterator
>> - * @mm: The mm_struct
>>    */
>> -static inline void vma_complete(struct vma_prepare *vp,
>> -				struct vma_iterator *vmi, struct mm_struct *mm)
>> +static inline void vma_complete(struct vma_prepare *vp, struct vma_iterator *vmi)
>>   {
>> +	struct mm_struct *mm = vp->vma->vm_mm;
>> +
>>   	if (vp->file) {
>>   		if (vp->adj_next)
>>   			vma_interval_tree_insert(vp->adj_next,
>> @@ -666,7 +666,7 @@ int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
>>   	vma_set_range(vma, start, end, pgoff);
>>   	vma_iter_store(vmi, vma);
>>   
>> -	vma_complete(&vp, vmi, vma->vm_mm);
>> +	vma_complete(&vp, vmi);
>>   	return 0;
>>   
>>   nomem:
>> @@ -707,7 +707,7 @@ int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
>>   
>>   	vma_iter_clear(vmi);
>>   	vma_set_range(vma, start, end, pgoff);
>> -	vma_complete(&vp, vmi, vma->vm_mm);
>> +	vma_complete(&vp, vmi);
>>   	return 0;
>>   }
>>   
>> @@ -1030,7 +1030,7 @@ static struct vm_area_struct
>>   		}
>>   	}
>>   
>> -	vma_complete(&vp, vmi, mm);
>> +	vma_complete(&vp, vmi);
>>   	khugepaged_enter_vma(res, vm_flags);
>>   	return res;
>>   
>> @@ -2377,7 +2377,7 @@ static int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *vma,
>>   	}
>>   
>>   	/* vma_complete stores the new vma */
>> -	vma_complete(&vp, vmi, vma->vm_mm);
>> +	vma_complete(&vp, vmi);
>>   
>>   	/* Success. */
>>   	if (new_below)
>> @@ -3145,7 +3145,7 @@ static int do_brk_flags(struct vma_iterator *vmi, struct vm_area_struct *vma,
>>   		vm_flags_set(vma, VM_SOFTDIRTY);
>>   		vma_iter_store(vmi, vma);
>>   
>> -		vma_complete(&vp, vmi, mm);
>> +		vma_complete(&vp, vmi);
>>   		khugepaged_enter_vma(vma, flags);
>>   		goto out;
>>   	}
>> -- 
>> 2.25.1
>>
>>