Re: [PATCH] Cosmetic:Partially remove deprecated __initcall() and change to

["Justin P. mattock" <justinmattock@gmail.com>]

On 03/21/2010 05:33 PM, Randy Dunlap wrote:
> On 03/19/10 16:24, Justin P. Mattock wrote:
>> On 03/19/2010 03:56 PM, Randy Dunlap wrote:
>>> On 03/19/10 12:51, Justin P. Mattock wrote:
>>>> After doing some things on a small issue,
>>>> I noticed through web surfing, that there were patches
>>>> submitted pertaining that __initcall is deprecated,
>>>> and device_initcall should be used.
>>>
>>> Where was this discussion?  Do you have any pointers to it?
>>>
>>
>> The best info on this is scripts/checkpatch.pl
>> line #2664
>>
>> when I found this I just did a quick search of __initcall
>> (which gives hits here and there)
>> https://patchwork.kernel.org/patch/23344/
>> (also found others at around 2008 or so)
>
> Thanks.  IMO there should be something in the kernel source tree
> that says explicitly that __initcall is deprecated and should be
> replaced by using<whatever should be used>.  That's missing.
>

agree..

I should of searched the source tree for something
that says deprecated and so forth before doing anything(the 
checkpatch.pl was something I noticed down the line but doesn't say 
deprecated(say's explicitly).


>
>>> I don't see any mention of __initcall being deprecated in
>>> Linus' mainline git tree, or in linux-next, or in mmotm.
>>> Where are those patches?
>>>
>>
>> I don't know(I'm out of the social pipeline when it comes to Linux news,
>> and updates)..
>> like in the explanation part of the patch
>> I was looking into something else, then ran into this,
>> so as a break(from what I was originally doing)
>> decided to do this and submit.
>>
>>>
>>>> So as a change of subject(since what I was looking at
>>>> was frustrating me),I decided to grep the whole tree
>>>> and make the change(partially).
>>>>
>>>> Currently I'm running this patch on my system, kernel compiles
>>>> without any errors or warnings.(thought there would be a speed increase
>>>> but didn't see much(if any)).
>>>
>>> No, __initcall(x) is just a shorthand version of typing
>>> device_initcall(x).  They do the same thing.
>>>
>>
>> yep, that's what I found out as well(reason for the cosmetic
>> in the subject line).
>>
>>>> Biggest problem I have though is testing this on other hardware types
>>>> (I only have a macbook,and an iMac).
>>>> So please if you have the access to other arch/hardware types please
>>>> test.
>>>>
>>>> Now what I mean by partially is the __initcall function is still
>>>> there, so(if any) userspace apps/libs depend on this it's there
>>>> so they dont break and/or any other subsystem, that needs time
>>>> to make the changes.
>>>
>>> The only thing that might be affected is building out-of-tree drivers,
>>> but those are easy to fix.
>>>
>>
>> alright..main concern is making sure things don't break in the
>> kernel(even though things do).
>>
>> I can have a go at the header files, submit
>> then if it's something people agree they want to do, then they
>> can go from there(if not it's fine as well).
>>
>>>> Note:
>>>> the remaining files that still have __initcall in them are:
>>>> (according to grep)
>>>>
>>>> arch/um/include/shared/init.h
>>>> include/linux/init.h
>>>> scripts/checkpatch.pl
>>>>
>>>> either I or somebody else, can change this(although a bit
>>>> concerned about breaking things).
>>>>
>>>> Signed-off-by: Justin P. Mattock<justinmattock@gmail.com>
>>>> ---
>
>


In any case I'll leave this to you guys to decide.
The patch is in cyberspace now, so if/when this ever
is decided it's there(patch), then can be used then.

Justin P. Mattock