Re: [RFC PATCH 0/8] rqbased-dm: request-based device-mapper

[Jens Axboe <jens.axboe@oracle.com>]

On Tue, Dec 19 2006, Kiyoshi Ueda wrote:
> For block layer maintainer and developers:
>   This patch set has 2 block layer changes below.
>     - Changed blk_get_request() to allow calls from interrupt context
>       so that queue's request_fn can use it.  (PATCH 1/8)
>       (*) The behaviour of CFQ (or other scheduler which depends on
>           "current") may be affected when blk_get_request() is called
>           in interrupt context, because "current" is not the process
>           which issue the original request.

This we already covered.

>     - Added new "end_io_first" hook to __end_that_request_first()
>       and struct request.  (PATCH 2/8)

This I still need an explanation for.

>   And I'm thinking about:
>     - Moving blk_clone_rq() to ll_rw_blk.c from dm.c. (PATCH 7/8)

That definitely should be core code. I see no problem in that.

-- 
Jens Axboe