Re: [PATCH] libata queue updated

[Tejun <htejun@gmail.com>]

Hello, Ingo.

Ingo Oeser wrote:
> Hi Tejun,
> 
> On Tuesday 31 January 2006 01:04, Tejun wrote:
> 
>>Ingo Oeser wrote:
>>
>>>What about putting the information directly into "ap->device[INDEX].class" 
>>>in the sole caller (ata_drive_probe_reset) so far?
>>>
>>
>>Not altering ->class directly in lldd driver is one major point of this 
>>whole patchset such that higher level driving logic has a say on whether 
>>a device is online or not, not the low level driver.  Primarily this is 
>>useful for sharing low-level codes with hot plugging / EH but it's also 
>>possible to retry some of the operations during probing in limited cases.
> 
> 
> Ok, with this argument, I finally get it. Now I know why you do it this
> way. You let the lld driver suggest a class for it's devices and verify
> these suggestions by high level code. 
> 
> The only way to get to this classification data is by resetting the ATA
> device.
> 
> It might be technically possible to set ->class directly and 
> fix it up in high level logic, as needed.

Yeap, that's right.  I actually considered that too but it was kind of 
messy that way - storing the current value, invoke callbacks, restoring 
values if something doesn't look right.  Just using temporary variable 
is much more straight-forward, it seemed.

> 
> Your explicit design decision was NOT to do this but to put this
> suggestions from low level driver into a temporary on stack structure
> from the higher level API.
> 
> And since the maintainer is happy already, I couldn't care less.
> 
> Thanks for your patience :-)

Great that we could reach an agreement.

Thanks. :-)

-- 
tejun