Re: [PATCH v2] btrfs: make btrfs module init/exit match their sequence

[Anand Jain <anand.jain@oracle.com>]

>>>>> +static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
>>>>
>>>>   Why not move bool mod_init_result into the (non-const) struct
>>>> init_sequence?
>>
>> Any comment on this suggestion?
> 
> Why you want to change the init_sequence array into non-const then?

   We can remove an isolated array mod_init_result to contain the result.
   Instead struct init_sequence can have results as a member.


>>>>> +    /*
>>>>> +     * If we call exit_btrfs_fs() it would cause section mismatch.
>>>>> +     * As init_btrfs_fs() belongs to .init.text, while 
>>>>> exit_btrfs_fs()
>>>>> +     * belongs to .exit.text.
>>>>> +     */
>>>>   Why not move it into a helper that can be called at both exit and
>>>> init?
>>>
>>> IIRC the last time I went the helper path, it caused section mismatch
>>> again, as all __init/__exit functions can only call functions inside
>>> .init/.exit.text.
>>>
>>> Thus the helper way won't solve it.
>>
>> Really? Maybe it was something else because, I see it as working.
>> As below.
> 
> You removed __exit, which removed the section type check.

  Yeah. We don't need __exit return type in the helper function. It does 
not make sense in the helper function.