Re: [PATCH] 2.5.15 IDE 62

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

Uz.ytkownik benh@kernel.crashing.org napisa?:
>>>So I think we should have per channel locks on this level
>>>right? This is anyway our unit for serialization.
>>>(I'm just surprised that blk_init_queue() doesn't
>>>provide queue specific locking and relies on exported
>>>locks from the drivers...)
>>
>>Sure go ahead and fine grain it, I had no time to go that much into
>>detail when ripping out io_request_lock. A drive->lock passed to
>>blk_init_queue would do nicely.
>>
>>But beware that ide locking is a lot nastier than you think. I saw other
>>irq changes earlier, I just want to make sure that you are _absolutely_
>>certain that these changes are safe??
> 
> 
> You'll probably need a per-host lock (but that one can be safely
> hidden in the host controller driver I beleive) since some hosts
> share some registers for their 2 channels (timings can be bitfields
> in a single register controlling 2 channels, I'm not too sure about
> legacy DMA).

Just to clarify it... From the host view it's not the chipset
it's a channel we have to deal with. And there are typically two
channels on a host. For the serialized parts, we have to
possiblities:

1. Preserve the current behaviour of using additionally a global
lock.

2. "Cheat" and reuse the lock from the primary channel during
the initialization of the secondary channel.

Hmmm.... Thinking a bit about it I'm now conviced that 2. is more
elegant then 1. And finally this will
just allow us to make the hwgroup_t go entierly away.


The only thing that worries me are the checks for hwgroupt_t's
only remaining member -> handler use to determine whatever some
IRQ is pending or not.