Re: [PATCH 01/30] blk_end_request: add new request completion interface (take 4)

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Tue, 2007-12-11 at 17:40 -0500, Kiyoshi Ueda wrote:
> This patch adds 2 new interfaces for request completion:
>   o blk_end_request()   : called without queue lock
>   o __blk_end_request() : called with queue lock held
> 
> blk_end_request takes 'error' as an argument instead of 'uptodate',
> which current end_that_request_* take.
> The meanings of values are below and the value is used when bio is
> completed.
>     0 : success
>   < 0 : error
> 
> Some device drivers call some generic functions below between
> end_that_request_{first/chunk} and end_that_request_last().
>   o add_disk_randomness()
>   o blk_queue_end_tag()
>   o blkdev_dequeue_request()

If we can roll the whole thing together, that would be nice.  However,
the way you're doing it with this patch, we now have an asymmetrical
interface:  The request routine must explicitly start the tag, but now
doesn't have to end it.

We really need symmetry.  Either go back to start tag/end tag, or absorb
the whole lot into the block infrastructure.

The original reason for the explicit start/end is that there are some
requests on a tagged device that aren't able to be tagged by the block
layer (some devices reserve tag numbers for specific meanings).
However, I don't think there's any driver that actually implemented this
feature.

James