[Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Chen Gang]

Hello All:

I searched 'arch/*' and 'drivers/*' sub-directory, all of them are 'obey
this rule', even in device_unregister() itself, it also firstly calls
device_remove_file(), then call kobject_del().

But after read the related code (fs/sysfs/*, drivers/base/core.c), it
seems kobject_del() -> sysfs_remove_dir() which will release all related
things (can instead of device_remove_file()).

So in fact, we need not call device_remove_file() before call
device_unregister(), is it correct ?


Welcome any member to give a suggestion or completion about it, thanks.


Thanks.
-- 
Chen Gang

Asianux Corporation

Re: [Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Ming Lei]

Hi,

On Fri, May 17, 2013 at 1:43 PM, Chen Gang <gang.chen@asianux.com> wrote:
> Hello All:
>
> I searched 'arch/*' and 'drivers/*' sub-directory, all of them are 'obey
> this rule', even in device_unregister() itself, it also firstly calls
> device_remove_file(), then call kobject_del().
>
> But after read the related code (fs/sysfs/*, drivers/base/core.c), it
> seems kobject_del() -> sysfs_remove_dir() which will release all related
> things (can instead of device_remove_file()).
>
> So in fact, we need not call device_remove_file() before call
> device_unregister(), is it correct ?

Looks it is correct but it is a bit implicit.

Thanks,
-- 
Ming Lei

Re: [Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Chen Gang]

On 05/18/2013 07:06 PM, Ming Lei wrote:
> Hi,
> 
> On Fri, May 17, 2013 at 1:43 PM, Chen Gang <gang.chen@asianux.com> wrote:
>> Hello All:
>>
>> I searched 'arch/*' and 'drivers/*' sub-directory, all of them are 'obey
>> this rule', even in device_unregister() itself, it also firstly calls
>> device_remove_file(), then call kobject_del().
>>
>> But after read the related code (fs/sysfs/*, drivers/base/core.c), it
>> seems kobject_del() -> sysfs_remove_dir() which will release all related
>> things (can instead of device_remove_file()).
>>
>> So in fact, we need not call device_remove_file() before call
>> device_unregister(), is it correct ?
> 
> Looks it is correct but it is a bit implicit.
> 

If really no other members reply within a week, we should treat your
opinion (or suggestion) as the final result conclusion within
linux-kernel@vger.kernel.org.  :-)


Thanks.
-- 
Chen Gang

Asianux Corporation

Re: [Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Greg Kroah-Hartman]

On Mon, May 20, 2013 at 09:03:27AM +0800, Chen Gang wrote:
> On 05/18/2013 07:06 PM, Ming Lei wrote:
> > Hi,
> > 
> > On Fri, May 17, 2013 at 1:43 PM, Chen Gang <gang.chen@asianux.com> wrote:
> >> Hello All:
> >>
> >> I searched 'arch/*' and 'drivers/*' sub-directory, all of them are 'obey
> >> this rule', even in device_unregister() itself, it also firstly calls
> >> device_remove_file(), then call kobject_del().
> >>
> >> But after read the related code (fs/sysfs/*, drivers/base/core.c), it
> >> seems kobject_del() -> sysfs_remove_dir() which will release all related
> >> things (can instead of device_remove_file()).
> >>
> >> So in fact, we need not call device_remove_file() before call
> >> device_unregister(), is it correct ?
> > 
> > Looks it is correct but it is a bit implicit.
> > 
> 
> If really no other members reply within a week, we should treat your
> opinion (or suggestion) as the final result conclusion within
> linux-kernel@vger.kernel.org.  :-)

I have no idea what you mean by this.

Re: [Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Chen Gang]

On 05/20/2013 09:45 AM, Greg Kroah-Hartman wrote:
> On Mon, May 20, 2013 at 09:03:27AM +0800, Chen Gang wrote:
>> > On 05/18/2013 07:06 PM, Ming Lei wrote:
>>> > > Hi,
>>> > > 
>>> > > On Fri, May 17, 2013 at 1:43 PM, Chen Gang <gang.chen@asianux.com> wrote:
>>>> > >> Hello All:
>>>> > >>
>>>> > >> I searched 'arch/*' and 'drivers/*' sub-directory, all of them are 'obey
>>>> > >> this rule', even in device_unregister() itself, it also firstly calls
>>>> > >> device_remove_file(), then call kobject_del().
>>>> > >>
>>>> > >> But after read the related code (fs/sysfs/*, drivers/base/core.c), it
>>>> > >> seems kobject_del() -> sysfs_remove_dir() which will release all related
>>>> > >> things (can instead of device_remove_file()).
>>>> > >>
>>>> > >> So in fact, we need not call device_remove_file() before call
>>>> > >> device_unregister(), is it correct ?
>>> > > 
>>> > > Looks it is correct but it is a bit implicit.
>>> > > 
>> > 
>> > If really no other members reply within a week, we should treat your
>> > opinion (or suggestion) as the final result conclusion within
>> > linux-kernel@vger.kernel.org.  :-)
> I have no idea what you mean by this.
> 

I mean that if no reply by any other members within a week, I will know
it is correct that "we need not call device_remove_file() firstly before
call device_unregister()" (at least, one member's reply supports this
conclusion).

I find this 'question' when discussing a patch with another members in
linux-kernel@vger.kernel.org, I have read the related code and also have
searched with google, but can not find the result, so I want to consult
it in linux-kernel@vger.kernel.org.

And I think, this 'question' is not suitable to give a test firstly,
because the test plan need be discussed firstly (or the test result
means nothing).


This time, I send the 'consult' mail to 'All', not to specific members.


Thanks.
-- 
Chen Gang

Asianux Corporation

Re: [Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Greg Kroah-Hartman]

On Mon, May 20, 2013 at 10:12:41AM +0800, Chen Gang wrote:
> I mean that if no reply by any other members within a week, I will know
> it is correct that "we need not call device_remove_file() firstly before
> call device_unregister()" (at least, one member's reply supports this
> conclusion).
> 
> I find this 'question' when discussing a patch with another members in
> linux-kernel@vger.kernel.org, I have read the related code and also have
> searched with google, but can not find the result, so I want to consult
> it in linux-kernel@vger.kernel.org.

Asking random questions on lkml, and relying on the fact that no one
else happens to say anything, is not any judge as to if the answer is
correct at all.

In fact, just asking questions on lkml has a very low chance of ever
getting a correct answer, given that the people that usually do know the
answer to these types of things are usually:
  1) not reading lkml because they are busy doing real work
  2) annoyed by questions that are easily answered by themselves by
     either:
       a) reading the code
       b) writing a simple example module and testing it out yourself


Good luck,

greg k-h

Re: [Consult] Why need we call device_remove_file() firstly before call device_unregister() ?

[Chen Gang]

On 05/20/2013 10:20 AM, Greg Kroah-Hartman wrote:
> On Mon, May 20, 2013 at 10:12:41AM +0800, Chen Gang wrote:
>> > I mean that if no reply by any other members within a week, I will know
>> > it is correct that "we need not call device_remove_file() firstly before
>> > call device_unregister()" (at least, one member's reply supports this
>> > conclusion).
>> > 
>> > I find this 'question' when discussing a patch with another members in
>> > linux-kernel@vger.kernel.org, I have read the related code and also have
>> > searched with google, but can not find the result, so I want to consult
>> > it in linux-kernel@vger.kernel.org.
> Asking random questions on lkml, and relying on the fact that no one
> else happens to say anything, is not any judge as to if the answer is
> correct at all.
> 

OK, I can understand, now, thank you for your reply.

And I wish, we can provide the confirmation of all related questions
about Linux kernel, in the future,


> In fact, just asking questions on lkml has a very low chance of ever
> getting a correct answer, given that the people that usually do know the
> answer to these types of things are usually:
>   1) not reading lkml because they are busy doing real work

I should understand, they have no duty to have to reply the related
mail, especially every members already have their own work (and
normally, they are really busy).


>   2) annoyed by questions that are easily answered by themselves by
>      either:
>        a) reading the code

I have done, so I need not worry about this item. :-)

>        b) writing a simple example module and testing it out yourself
> 

Precisely, I did not do it firstly. It seems I should do it firstly
(although, at least now, I do not think it will get any valuable result
for our this case)

> 
> Good luck,
> 

OK, 'Lucky' is really the first important !!

I should continue to analyze this question, independent this 'consult' mail.


Thanks.
-- 
Chen Gang

Asianux Corporation