Re: [PATCH][2.5] BUG check in elevator.c:237

[Martin Dalecki <dalecki@evision-ventures.com>]

Zwane Mwaikambo wrote:
> Ok cool, regarding the ide-cd current problem, i still don't see how 
> moving it to a BH would help in this case, the only difference i can see 
> is that the ISR won't have to be so heavy and we can get more interrupt 
> handling done (assuming the BH processing keeps up). The problem isn't 
> that the read_intr is preempting other code and injecting already active 
> requests into the queue, but that ide-cd is allowing running requests to 
> be sent multiple times.

Yeep...

> If you can just explain marginally a  possible workaround other than the 
> kludge i diffed up i can try investigating further.

Actually I have right now no *precise* idea (please note the difference
between *no idea* and *no precise idea*) about the optimal fix.
However making things leaner in first step may give you direct insight
into the proper solution as a side effect (OK. I know this is a bit
philosophing, but this method works really frequently... it's called
deduction...).

And it may actually turn out that the diff could be shorter
as some wordish explanation ;-).