[PATCH v2 5/5] mm/hugetlb: avoid calculating fault_mutex_hash in truncate_op case

[PATCH v2 5/5] mm/hugetlb: avoid calculating fault_mutex_hash in truncate_op case

[Miaohe Lin]

The fault_mutex hashing overhead can be avoided in truncate_op case
because page faults can not race with truncation in this routine.  So
calculate hash for fault_mutex only in !truncate_op case to save some cpu
cycles.

Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
---
v1->v2:
remove unnecessary initialization for variable hash
collect Reviewed-by tag from Mike Kravetz
---
 fs/hugetlbfs/inode.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index c262566f7c5d..f7ec94bc7337 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -485,7 +485,6 @@ static void remove_inode_hugepages(struct inode *inode, loff_t lstart,
 			u32 hash;
 
 			index = page->index;
-			hash = hugetlb_fault_mutex_hash(mapping, index);
 			if (!truncate_op) {
 				/*
 				 * Only need to hold the fault mutex in the
@@ -493,6 +492,7 @@ static void remove_inode_hugepages(struct inode *inode, loff_t lstart,
 				 * page faults.  Races are not possible in the
 				 * case of truncation.
 				 */
+				hash = hugetlb_fault_mutex_hash(mapping, index);
 				mutex_lock(&hugetlb_fault_mutex_table[hash]);
 			}
 
-- 
2.19.1

Re: [PATCH v2 5/5] mm/hugetlb: avoid calculating fault_mutex_hash in truncate_op case

[Mike Kravetz]

On 3/15/21 7:27 PM, Miaohe Lin wrote:
> The fault_mutex hashing overhead can be avoided in truncate_op case
> because page faults can not race with truncation in this routine.  So
> calculate hash for fault_mutex only in !truncate_op case to save some cpu
> cycles.
> 
> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> Cc: Mike Kravetz <mike.kravetz@oracle.com>
> ---
> v1->v2:
> remove unnecessary initialization for variable hash
> collect Reviewed-by tag from Mike Kravetz

My apologies for not replying sooner and any misunderstanding from my
previous comments.

If the compiler is going to produce a warning because the variable is
not initialized, then we will need to keep the initialization.
Otherwise, this will show up as a build regression.  Ideally, there
would be a modifier which could be used to tell the compiler the
variable will used.  I do not know if such a modifier exists.

The patch can not produce a new warning.  So, if you need to initialize
the variable then do it.  My Reviewed-by still applies.
-- 
Mike Kravetz

> ---
>  fs/hugetlbfs/inode.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
> index c262566f7c5d..f7ec94bc7337 100644
> --- a/fs/hugetlbfs/inode.c
> +++ b/fs/hugetlbfs/inode.c
> @@ -485,7 +485,6 @@ static void remove_inode_hugepages(struct inode *inode, loff_t lstart,
>  			u32 hash;
>  
>  			index = page->index;
> -			hash = hugetlb_fault_mutex_hash(mapping, index);
>  			if (!truncate_op) {
>  				/*
>  				 * Only need to hold the fault mutex in the
> @@ -493,6 +492,7 @@ static void remove_inode_hugepages(struct inode *inode, loff_t lstart,
>  				 * page faults.  Races are not possible in the
>  				 * case of truncation.
>  				 */
> +				hash = hugetlb_fault_mutex_hash(mapping, index);
>  				mutex_lock(&hugetlb_fault_mutex_table[hash]);
>  			}
>  
> 

Re: [PATCH v2 5/5] mm/hugetlb: avoid calculating fault_mutex_hash in truncate_op case

[Miaohe Lin]

On 2021/3/16 11:07, Mike Kravetz wrote:
> On 3/15/21 7:27 PM, Miaohe Lin wrote:
>> The fault_mutex hashing overhead can be avoided in truncate_op case
>> because page faults can not race with truncation in this routine.  So
>> calculate hash for fault_mutex only in !truncate_op case to save some cpu
>> cycles.
>>
>> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>> Cc: Mike Kravetz <mike.kravetz@oracle.com>
>> ---
>> v1->v2:
>> remove unnecessary initialization for variable hash
>> collect Reviewed-by tag from Mike Kravetz
> 
> My apologies for not replying sooner and any misunderstanding from my
> previous comments.
> 

That's all right.

> If the compiler is going to produce a warning because the variable is
> not initialized, then we will need to keep the initialization.
> Otherwise, this will show up as a build regression.  Ideally, there
> would be a modifier which could be used to tell the compiler the
> variable will used.  I do not know if such a modifier exists.
> 

I do not know if such a modifier exists too. But maybe not all compilers are intelligent
enough to not produce a warning. It would be safe to keep the initialization...

> The patch can not produce a new warning.  So, if you need to initialize

So just drop this version of the patch? Or should I send a new version with your Reviewed-by tag and
keep the initialization?

> the variable then do it.  My Reviewed-by still applies.
>

Re: [PATCH v2 5/5] mm/hugetlb: avoid calculating fault_mutex_hash in truncate_op case

[Mike Kravetz]

On 3/15/21 11:49 PM, Miaohe Lin wrote:
> On 2021/3/16 11:07, Mike Kravetz wrote:
>> On 3/15/21 7:27 PM, Miaohe Lin wrote:
>>> The fault_mutex hashing overhead can be avoided in truncate_op case
>>> because page faults can not race with truncation in this routine.  So
>>> calculate hash for fault_mutex only in !truncate_op case to save some cpu
>>> cycles.
>>>
>>> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
>>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>>> Cc: Mike Kravetz <mike.kravetz@oracle.com>
>>> ---
>>> v1->v2:
>>> remove unnecessary initialization for variable hash
>>> collect Reviewed-by tag from Mike Kravetz
>>
>> My apologies for not replying sooner and any misunderstanding from my
>> previous comments.
>>
> 
> That's all right.
> 
>> If the compiler is going to produce a warning because the variable is
>> not initialized, then we will need to keep the initialization.
>> Otherwise, this will show up as a build regression.  Ideally, there
>> would be a modifier which could be used to tell the compiler the
>> variable will used.  I do not know if such a modifier exists.
>>
> 
> I do not know if such a modifier exists too. But maybe not all compilers are intelligent
> enough to not produce a warning. It would be safe to keep the initialization...
> 
>> The patch can not produce a new warning.  So, if you need to initialize
> 
> So just drop this version of the patch? Or should I send a new version with your Reviewed-by tag and
> keep the initialization?
> 

Yes, drop this version of the patch.  You can add my Reviewed-by to the
previous version that included the initialization and resend.

All the cleanup patches in this series should be good to go.
-- 
Mike Kravetz

Re: [PATCH v2 5/5] mm/hugetlb: avoid calculating fault_mutex_hash in truncate_op case

[Miaohe Lin]

On 2021/3/17 8:27, Mike Kravetz wrote:
> On 3/15/21 11:49 PM, Miaohe Lin wrote:
>> On 2021/3/16 11:07, Mike Kravetz wrote:
>>> On 3/15/21 7:27 PM, Miaohe Lin wrote:
>>>> The fault_mutex hashing overhead can be avoided in truncate_op case
>>>> because page faults can not race with truncation in this routine.  So
>>>> calculate hash for fault_mutex only in !truncate_op case to save some cpu
>>>> cycles.
>>>>
>>>> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
>>>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>>>> Cc: Mike Kravetz <mike.kravetz@oracle.com>
>>>> ---
>>>> v1->v2:
>>>> remove unnecessary initialization for variable hash
>>>> collect Reviewed-by tag from Mike Kravetz
>>>
>>> My apologies for not replying sooner and any misunderstanding from my
>>> previous comments.
>>>
>>
>> That's all right.
>>
>>> If the compiler is going to produce a warning because the variable is
>>> not initialized, then we will need to keep the initialization.
>>> Otherwise, this will show up as a build regression.  Ideally, there
>>> would be a modifier which could be used to tell the compiler the
>>> variable will used.  I do not know if such a modifier exists.
>>>
>>
>> I do not know if such a modifier exists too. But maybe not all compilers are intelligent
>> enough to not produce a warning. It would be safe to keep the initialization...
>>
>>> The patch can not produce a new warning.  So, if you need to initialize
>>
>> So just drop this version of the patch? Or should I send a new version with your Reviewed-by tag and
>> keep the initialization?
>>
> 
> Yes, drop this version of the patch.  You can add my Reviewed-by to the
> previous version that included the initialization and resend.
> 

Will do. Many thanks. :)

> All the cleanup patches in this series should be good to go.
>