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

[Qu Wenruo <wqu@suse.com>]

On 2022/10/13 17:29, Nikolay Borisov wrote:
> 
> 
> On 13.10.22 г. 12:22 ч., Qu Wenruo wrote:
>>> ?
>>
>> For adding a new sequence, one has to understand the dependency (if any).
>>
>> If no dependency (which I believe is the most common case), then the
>> generic idea is just to add it before the selftest.
>>
>>
>> The question would be more critical for open_ctree(), in fact
>> open_ctree() has a lot of cases that something can only be initialized
>> after its dependency.
>>
>> In that case, your concern is correct, one has to go through the init
>> functions to find a proper location.
>> And unlike the original code, it's one extra level of indirection.
>>
>> But I'd say, for most part, the init function names should explain
>> themselves, thus I hope it won't cause too much hassles in the future.
> 
> 
> In this case I'd say open_ctree shouldn't be switched to this mechanism.

The truth is, open_ctree() has the worst mismatch in its init sequence, 
almost half of the error labels are not properly matched.

And for the several init functions which needs dependency, it's not that 
hard to grasp and put into proper order.

Thanks,
Qu