Messages in this thread |  | | From | Benjamin Herrenschmidt <> | Subject | Re: [PATCH] 2.5.15 IDE 62 | Date | Fri, 17 May 2002 03:27:49 +0100 |
| |
>1. The fact that there are some cases where the initialization code >doesn't necessarily go down to the host chip drivers right now.
Well, I'd rather see the init code be some kind of "lib" called by the host driver, but well...
>2. Most of the current code...
Sure ;) Note that the real point I missed is about broken controllers that can't grok dealing with simultaneous requests on both channels (like cmd640) in which case, you probably need the same request queue to serialize access to both of them. In practice, that means more or less dealing with both channels like the same way we do with slave vs. master.
My understanding though is that the block layer can't (yet ?) quite deal with a single request queue for several target devices, and it seems that the whole point of the old "busy" flag along with andre taskfile stuff was to perform some kind of fair arbitration between which channels/targets got a chance to process requests.
>BTW> The code will be much cleaner in the upcomming ide 65, since >the allocation of the structures shared between two channels will be >simple pushed down to the corresponding host chip drivers instead of >the "match search" done after the channles have been initialized.
Great ! We are slowly going toward a real host controller driver template finally ;)
>Since most of the host chip drivers are not reentrant anyway we will >be able to save quite a lot of allocation code as well.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
|  |