Renee Danson Sommerfeld wrote:
> On Wed, Nov 11, 2009 at 05:25:32PM +0000, Alan Maguire wrote:
>   
>> As part of the fini, we generate and consume a state
>> event directly, so we don't need to go through the process
>> of enqueueing then dequeueing (which as you say would
>> lead to things getting misordered). As a consequence, the
>> finis and state events are all processed and sent prior to
>> the NCP state event signifying the switch. I've expanded on
>> this in the comments.
>>     
>
> Sounds good, and the updated comments help.
>
>   
>>> ncu.c
>>> 1239: Do we need to do the ncu check on remove of a link, as well?
>>>
>>>       
>> Now that you mention it, I think in the removal case we do need
>> an NCU check, but I think there's a bit more work needed here.
>> If the Automatic NCP is active, there's nothing additional to do
>> since we simply remove the NCUs from storage and nwamd.
>> An NCU check is needed as you say though since removal
>> may have other consequences (a priority group may be invalid
>> post-removal etc).
>>
>> If the User NCP is active however, the device may be removed
>> but the link and interface NCUs remain and we need to get those
>> NCUS into the uninitialized/not found state. So I've added a
>> function to handle state changes for the User NCP in the case
>> that NCUs exist for the added/removed device.
>>
>> In the removal case, we transition to the uninitialized/not found
>> state (via the  online*/not found state in  cases where the NCU
>> is enabled).  In the case that the User NCP has an NCU for a
>> device that  was removed in the past but has been added again, we need
>> to transition from uninitialized to offline so that the NCU check
>> won't skip the uninitialized NCUs. I've tested all this with hotplug
>> and it works fine.
>>     
>
> Cool, the changes look good.  Just one small nit (and this isn't really
> part of your changes, but it's in that code): could you add a little
> more to the comment at line 1274 to make it clear that the NCUs are being
> removed from the *Automatic* NCP?  I was initially confused, thinkng that
> ncph was for the active NCP, not the Automatic one.
>
>   
done. thanks!

Alan

Reply via email to