[PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[Zheng Yongjun]

mutex lock can be initialized automatically with DEFINE_MUTEX()
rather than explicitly calling mutex_init().

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
---
 drivers/md/bcache/super.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
index 78c08a8aece8..c124da6e646d 100644
--- a/drivers/md/bcache/super.c
+++ b/drivers/md/bcache/super.c
@@ -41,7 +41,7 @@ static const char invalid_uuid[] = {
 };
 
 static struct kobject *bcache_kobj;
-struct mutex bch_register_lock;
+DEFINE_MUTEX(bch_register_lock);
 bool bcache_is_reboot;
 LIST_HEAD(bch_cache_sets);
 static LIST_HEAD(uncached_devices);
@@ -2870,7 +2870,6 @@ static int __init bcache_init(void)
 
 	check_module_parameters();
 
-	mutex_init(&bch_register_lock);
 	init_waitqueue_head(&unregister_wait);
 	register_reboot_notifier(&reboot);
 

Re: [PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[Coly Li]

On 4/5/21 6:14 PM, Zheng Yongjun wrote:
> mutex lock can be initialized automatically with DEFINE_MUTEX()
> rather than explicitly calling mutex_init().
> 
> Reported-by: Hulk Robot <hulkci@huawei.com>
> Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>

NACK. This is not the first time people try to "fix" this location...

Using DEFINE_MUTEX() does not gain anything for us, it will generate
unnecessary extra size for the bcache.ko.
ines.

> ---
>  drivers/md/bcache/super.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
> index 78c08a8aece8..c124da6e646d 100644
> --- a/drivers/md/bcache/super.c
> +++ b/drivers/md/bcache/super.c
> @@ -41,7 +41,7 @@ static const char invalid_uuid[] = {
>  };
>  
>  static struct kobject *bcache_kobj;
> -struct mutex bch_register_lock;


Hmm, maybe if you compose a patch to add comments for bch_register_lock,
for something like: Don't initialize global variable here. It might be
helpful for noticing people not to fixing this in future.

Thanks.

Coly Li

> +DEFINE_MUTEX(bch_register_lock);
>  bool bcache_is_reboot;
>  LIST_HEAD(bch_cache_sets);
>  static LIST_HEAD(uncached_devices);
> @@ -2870,7 +2870,6 @@ static int __init bcache_init(void)
>  
>  	check_module_parameters();
>  
> -	mutex_init(&bch_register_lock);
>  	init_waitqueue_head(&unregister_wait);
>  	register_reboot_notifier(&reboot);
>  
> 

Re: [PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[Muhammad Usama Anjum]

On Mon, 2021-04-05 at 22:02 +0800, Coly Li wrote:
> On 4/5/21 6:14 PM, Zheng Yongjun wrote:
> > mutex lock can be initialized automatically with DEFINE_MUTEX()
> > rather than explicitly calling mutex_init().
> > 
> > Reported-by: Hulk Robot <hulkci@huawei.com>
> > Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
> 
> NACK. This is not the first time people try to "fix" this location...
> 
> Using DEFINE_MUTEX() does not gain anything for us, it will generate
> unnecessary extra size for the bcache.ko.
> ines.

How can the final binary have larger size by just static declaration?
By using DEFINE_MUTEX, the mutex is initialized at compile time. It'll
save initialization at run time and one line of code will be less also
from text section. 

#### with no change (dynamic initialization)
size drivers/md/bcache/bcache.ko
   text	   data	    bss	    dec	    hex	filename
 187792	  25310	    152	 213254	  34106	drivers/md/bcache/bcache.ko

#### with patch applied (static initialization)
   text	   data	    bss	    dec	    hex	filename
 187751	  25342	    120	 213213	  340dd	drivers/md/bcache/bcache.ko

Module's binary size has decreased by 41 bytes with the path applied
(x86_64 arch).

Re: [PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[Muhammad Usama Anjum]

On Tue, 2021-04-06 at 02:17 +0500, Muhammad Usama Anjum wrote:
> On Mon, 2021-04-05 at 22:02 +0800, Coly Li wrote:
> > On 4/5/21 6:14 PM, Zheng Yongjun wrote:
> > > mutex lock can be initialized automatically with DEFINE_MUTEX()
> > > rather than explicitly calling mutex_init().
> > > 
> > > Reported-by: Hulk Robot <hulkci@huawei.com>
> > > Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
> > 
> > NACK. This is not the first time people try to "fix" this location...
> > 
> > Using DEFINE_MUTEX() does not gain anything for us, it will generate
> > unnecessary extra size for the bcache.ko.
> > ines.
> 
> How can the final binary have larger size by just static declaration?
> By using DEFINE_MUTEX, the mutex is initialized at compile time. It'll
> save initialization at run time and one line of code will be less also
> from text section. 
> 
> #### with no change (dynamic initialization)
> size drivers/md/bcache/bcache.ko
>    text	   data	    bss	    dec	    hex	filename
>  187792	  25310	    152	 213254	  34106	drivers/md/bcache/bcache.ko
> 
> #### with patch applied (static initialization)
>    text	   data	    bss	    dec	    hex	filename
>  187751	  25342	    120	 213213	  340dd	drivers/md/bcache/bcache.ko
> 
> Module's binary size has decreased by 41 bytes with the path applied
> (x86_64 arch).
> 
Anybody has any thoughts on it?

Re: [PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[Dan Carpenter]

On Wed, Apr 28, 2021 at 09:19:26PM +0500, Muhammad Usama Anjum wrote:
> On Tue, 2021-04-06 at 02:17 +0500, Muhammad Usama Anjum wrote:
> > On Mon, 2021-04-05 at 22:02 +0800, Coly Li wrote:
> > > On 4/5/21 6:14 PM, Zheng Yongjun wrote:
> > > > mutex lock can be initialized automatically with DEFINE_MUTEX()
> > > > rather than explicitly calling mutex_init().
> > > > 
> > > > Reported-by: Hulk Robot <hulkci@huawei.com>
> > > > Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
> > > 
> > > NACK. This is not the first time people try to "fix" this location...
> > > 
> > > Using DEFINE_MUTEX() does not gain anything for us, it will generate
> > > unnecessary extra size for the bcache.ko.
> > > ines.
> > 
> > How can the final binary have larger size by just static declaration?
> > By using DEFINE_MUTEX, the mutex is initialized at compile time. It'll
> > save initialization at run time and one line of code will be less also
> > from text section. 
> > 
> > #### with no change (dynamic initialization)
> > size drivers/md/bcache/bcache.ko
> >    text	   data	    bss	    dec	    hex	filename
> >  187792	  25310	    152	 213254	  34106	drivers/md/bcache/bcache.ko
> > 
> > #### with patch applied (static initialization)
> >    text	   data	    bss	    dec	    hex	filename
> >  187751	  25342	    120	 213213	  340dd	drivers/md/bcache/bcache.ko
> > 
> > Module's binary size has decreased by 41 bytes with the path applied
> > (x86_64 arch).
> > 
> Anybody has any thoughts on it?
>

I think you're right and the response is puzzling.  But who cares?  It's
a small thing.  Leave it and move on.

regards,
dan carpenter

Re: [PATCH -next] bcache: use DEFINE_MUTEX() for mutex lock

[Coly Li]

On 4/29/21 12:19 AM, Muhammad Usama Anjum wrote:
> On Tue, 2021-04-06 at 02:17 +0500, Muhammad Usama Anjum wrote:
>> On Mon, 2021-04-05 at 22:02 +0800, Coly Li wrote:
>>> On 4/5/21 6:14 PM, Zheng Yongjun wrote:
>>>> mutex lock can be initialized automatically with DEFINE_MUTEX()
>>>> rather than explicitly calling mutex_init().
>>>>
>>>> Reported-by: Hulk Robot <hulkci@huawei.com>
>>>> Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
>>>
>>> NACK. This is not the first time people try to "fix" this location...
>>>
>>> Using DEFINE_MUTEX() does not gain anything for us, it will generate
>>> unnecessary extra size for the bcache.ko.
>>> ines.
>>
>> How can the final binary have larger size by just static declaration?
>> By using DEFINE_MUTEX, the mutex is initialized at compile time. It'll
>> save initialization at run time and one line of code will be less also
>> from text section. 
>>
>> #### with no change (dynamic initialization)
>> size drivers/md/bcache/bcache.ko
>>    text	   data	    bss	    dec	    hex	filename
>>  187792	  25310	    152	 213254	  34106	drivers/md/bcache/bcache.ko
>>
>> #### with patch applied (static initialization)
>>    text	   data	    bss	    dec	    hex	filename
>>  187751	  25342	    120	 213213	  340dd	drivers/md/bcache/bcache.ko
>>
>> Module's binary size has decreased by 41 bytes with the path applied
>> (x86_64 arch).
>>
> Anybody has any thoughts on it?
> 
> 


I am waiting for Yongjun's v4 patch to update the commit log, which was
suggested by Pavel Goran.

Coly Li