Re: [PATCH][RFC] kthread abstraction

[Adam Keys <akeys@post.cis.smu.edu>]

On February 01, 2002 09:38, Christoph Hellwig wrote:
> Currently the startup of custom kernel threads contains lots
> of duplicated code (and bugs!).

Very interesting and approachable patch.  Am I correct this replaces the API 
for starting threads like bdflush and kswapd?  Is this just an API 
cleanliness thing or is there a performance motivation?

Just one question....

>
> I'd like to propose the following interface to get rid of these
> code waste:
>
>     int kthread_start(struct kthread *kth)
>
> 	Startup a new kernel thread as described by 'kth' (details
> 	below).  Wait until it has finished initialization.
>
>
>     void kthread_stop(struct kthread *kth)

Do you think you could get by just passing struct task_struct here?  I 
realize that would make it less pleasant for calling functions.  However, it 
would also prevent you from changing something else in the kthread in a later 
version and catching a caller by surprise.

Otherwise, looks nice to my naive eyes :)

-- 
akk~