Re: [PATCH] reworked IDE/general tagged command queueing

[Martin Dalecki]

Użytkownik Jens Axboe napisał:
> Hi,
> 
> I've rewritten parts of the IDE TCQ stuff to be, well, a lot better in
> my oppinion. I had to accept that the ata_request and rq->special usage
> sucked, it was just one big mess.
> 
> So following a suggestion from Martin and Linus, I implemented some
> basic tagged command queueing back end in the block layer. This is what
> the new IDE TCQ core is build on, and what potentially others can use as
> well. I'll start by describing the new API:

Looking at the IDE part we can now see that pushing the
generic functions one level up the impact on the code flow
on the IDE side is now:

1. Low (most of stuff is due to the ugly /proc special-ide-interface.

2. Nicely isolated.

Great work Jens! (just my humble opinnion).
However I see a note about the need
to unify the DMA parts, so I will se what can be done on this
side becouse I have always planned to get rid of the
silly switch(ide_dma_function_t) on the dmaproc-path.

May I ask you as well to just call ide-tcq.c simple tcq.c?
The ide- is entierly redundant and I see no need to stick
to the previous "convention" here. It is just a leftover from
the days where the IDE stuff didn't sit in his own directory.
In general  I rather prefer the prefix ata_ instead of ide_ becouse
we are on the command level and on the host here -
ide resides on the disk and the whole world
outside linux calls it ata_. Finally ata_ is far better
grep-able overall becouse the ide letter combination is very
common :-) But that's a minor nit of course.

My convention is to prefix functions with the module specific prefix
only if they are exported, which means:

1. They are directly external.

2. They are hiddenly exported by setting some methods in structs to them.

Otherwise I stick to the most convenient semantically related
name without the fear that it could sound too generic...
so queue_data() is meant to be local for example and it doesn't
clash with the generic bio_queue_data() or whatever.

For me this convention turned out to help narrowing the focus
during reading code...