[PATCH] btrfs: Simplify the allocation of slab caches in btrfs_delayed_inode_init

[PATCH] btrfs: Simplify the allocation of slab caches in btrfs_delayed_inode_init

[Kunwu Chan]

commit 0a31bd5f2bbb ("KMEM_CACHE(): simplify slab cache creation")
introduces a new macro.
Use the new KMEM_CACHE() macro instead of direct kmem_cache_create
to simplify the creation of SLAB caches.

Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
---
 fs/btrfs/delayed-inode.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index 08102883f560..8c748c6cdf6d 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -28,11 +28,7 @@ static struct kmem_cache *delayed_node_cache;
 
 int __init btrfs_delayed_inode_init(void)
 {
-	delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
-					sizeof(struct btrfs_delayed_node),
-					0,
-					SLAB_MEM_SPREAD,
-					NULL);
+	delayed_node_cache = KMEM_CACHE(btrfs_delayed_node, SLAB_MEM_SPREAD);
 	if (!delayed_node_cache)
 		return -ENOMEM;
 	return 0;
-- 
2.39.2

Re: [PATCH] btrfs: Simplify the allocation of slab caches in btrfs_delayed_inode_init

[Johannes Thumshirn]

On 31.01.24 07:20, Kunwu Chan wrote:
> commit 0a31bd5f2bbb ("KMEM_CACHE(): simplify slab cache creation")
> introduces a new macro.
> Use the new KMEM_CACHE() macro instead of direct kmem_cache_create

That commit is 17 years old. Why should we switch to it _now_? I 
wouldn't call it a new macro.

Don't get me wrong, I don't oppose the patch, but I'd prefer a better 
explanation why now and not 17 years ago when the macro got introduced.

> to simplify the creation of SLAB caches.
> 
> Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
> ---
>   fs/btrfs/delayed-inode.c | 6 +-----
>   1 file changed, 1 insertion(+), 5 deletions(-)
> 
> diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
> index 08102883f560..8c748c6cdf6d 100644
> --- a/fs/btrfs/delayed-inode.c
> +++ b/fs/btrfs/delayed-inode.c
> @@ -28,11 +28,7 @@ static struct kmem_cache *delayed_node_cache;
>   
>   int __init btrfs_delayed_inode_init(void)
>   {
> -	delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
> -					sizeof(struct btrfs_delayed_node),
> -					0,
> -					SLAB_MEM_SPREAD,
> -					NULL);
> +	delayed_node_cache = KMEM_CACHE(btrfs_delayed_node, SLAB_MEM_SPREAD);
>   	if (!delayed_node_cache)
>   		return -ENOMEM;
>   	return 0;

Re: [PATCH] btrfs: Simplify the allocation of slab caches in btrfs_delayed_inode_init

[David Sterba]

On Wed, Jan 31, 2024 at 10:20:35AM +0000, Johannes Thumshirn wrote:
> On 31.01.24 07:20, Kunwu Chan wrote:
> > commit 0a31bd5f2bbb ("KMEM_CACHE(): simplify slab cache creation")
> > introduces a new macro.
> > Use the new KMEM_CACHE() macro instead of direct kmem_cache_create
> 
> That commit is 17 years old. Why should we switch to it _now_? I 
> wouldn't call it a new macro.

I had the same reaction after checking the commit that added it.
> 
> Don't get me wrong, I don't oppose the patch, but I'd prefer a better 
> explanation why now and not 17 years ago when the macro got introduced.

We can add the macros where possible, at least it hides all the 0 or
NULL parameters, but yeah with a better changelog.

Re: [PATCH] btrfs: Simplify the allocation of slab caches in btrfs_delayed_inode_init

[Kunwu Chan]

On 2024/1/31 18:20, Johannes Thumshirn wrote:
> On 31.01.24 07:20, Kunwu Chan wrote:
>> commit 0a31bd5f2bbb ("KMEM_CACHE(): simplify slab cache creation")
>> introduces a new macro.
>> Use the new KMEM_CACHE() macro instead of direct kmem_cache_create
> 
> That commit is 17 years old. Why should we switch to it _now_? I
> wouldn't call it a new macro.
> 
> Don't get me wrong, I don't oppose the patch, but I'd prefer a better
> explanation why now and not 17 years ago when the macro got introduced.
> 
Thanks for your attention.
Like David say in 
https://lore.kernel.org/all/20240131183929.GP31555@twin.jikos.cz/#t.

The main reason is 'it hides all the 0 or NULL parameters', makes the 
code cleaner and more readable.

So i'll update the commit msg to this:

Use the new KMEM_CACHE() macro instead of direct kmem_cache_create
to simplify the creation of SLAB caches.
Make the code cleaner and more readable.

And resend a v2 patch.
Thanks again.
>> to simplify the creation of SLAB caches.
>>
>> Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
>> ---
>>    fs/btrfs/delayed-inode.c | 6 +-----
>>    1 file changed, 1 insertion(+), 5 deletions(-)
>>
>> diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
>> index 08102883f560..8c748c6cdf6d 100644
>> --- a/fs/btrfs/delayed-inode.c
>> +++ b/fs/btrfs/delayed-inode.c
>> @@ -28,11 +28,7 @@ static struct kmem_cache *delayed_node_cache;
>>    
>>    int __init btrfs_delayed_inode_init(void)
>>    {
>> -	delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
>> -					sizeof(struct btrfs_delayed_node),
>> -					0,
>> -					SLAB_MEM_SPREAD,
>> -					NULL);
>> +	delayed_node_cache = KMEM_CACHE(btrfs_delayed_node, SLAB_MEM_SPREAD);
>>    	if (!delayed_node_cache)
>>    		return -ENOMEM;
>>    	return 0;
> 
-- 
Thanks,
   Kunwu

Re: btrfs: Simplify the allocation of slab caches in btrfs_delayed_inode_init

[Markus Elfring]

…
> So i'll update the commit msg to this:
>
> Use the new KMEM_CACHE() macro instead of direct kmem_cache_create
> to simplify the creation of SLAB caches.
> Make the code cleaner and more readable.
…

* Please replace the word “new” by a reference to the commit 8eb8284b412906181357c2b0110d879d5af95e52
  ("usercopy: Prepare for usercopy whitelisting").

  See also related background information from 2017-06-10.

* How does your response fit to the repetition of improvable change descriptions?

  Example:
  [PATCH] btrfs: Simplify the allocation of slab caches in btrfs_transaction_init
  https://lore.kernel.org/lkml/20240201093554.208092-1-chentao@kylinos.cn/
  https://lkml.org/lkml/2024/2/1/387

* How do you think about to group similar source code transformations
  into patch series?


Regards,
Markus