Re: [PATCH v3 08/16] net/avp: device initialization

-----Original Message-----
From: Legacy, Allain

quoted

From: Chas Williams [mailto:3chas3@gmail.com]
I don't see the other side of this to unregister the callback.  It's also a bit
confusing with this here and the other parts in part 15.  It looks like you
enable the interrupts on .dev_create but disable on .dev_stop?
If that's the case, you likely want to just do the setup here and the enable

in

quoted

.dev_start.

Agreed.  This is not symmetric.  I will setup in create, enable in start, disable
in stop.

I was mistaken.  The interrupts need to be setup and enabled when the device 
is created, and then only disabled if the device is closed in preparation for 
removal.  The start/stop functions should not be involved.

So:

avp_dev_create(), will do:
    avp_dev_setup_interrupts()
    avp_enable_interrupts()

and

avp_dev_close(), will do:
    avp_disable_interrupts().

Currently, avp_dev_create() and avp_dev_close() are in separate patches.  
 I'll see how much trouble it would be to move the dev_close() to be in the 
same patch as the dev_create() but there is little value in doing that since 
functionally there is no net effect.  If the effort is too high I will be inclined 
to leave it as is.