Sebastien Roy wrote:

>FYI, my usr/src/lib/libdlpi comments are coming shortly.
>
>  
>
Thank you.

>>>usr/src/cmd/cmd-inet/usr.sbin/ifconfig/ifconfig.c
>>>
>>>3649: why not use IF_NAMESIZE (the standard maximum interface name
>>>     constant)?
>>> 
>>>
>>>      
>>>
>>Using a standard constant makes more sense. I've changed it to be a 
>>IF_NAMESIZE.
>>    
>>
>
>On the other hand, if you'd rather have a libdlpi specific constant,
>that's fine, but instead of setting it to 32 in libdlpi.h, I'd set it to
>IF_NAMESIZE (or LIFNAMSIZ).
>
>  
>
I'll stick to having a libdlpi specific constant as Meem pointed out 
that "provider is a DLPI concept, and is passed to a function that 
expects a buffer of DLPI_LINKNAME_MAX in size. "
Setting DLPI_LINKNAME_MAX to IF_NAMESIZE rather than 32; wouldn't that 
be pegging the value to a upper layer constant?

>>>3659, 3774: A comment is warranted regarding why this needs to be done.
>>>
>>>      
>>>
>>Before adding comments, I wanted to get opinions about making 
>>dlpi_detach a flag to dlpi_open() instead of providing it as a 
>>Consolidation private interface. 
>>    
>>
>
>You mean like a DLPI_NOATTACH flag to dlpi_open()?
>
>  
>
Yes.

>That sound fine to me, but in reality, I'd like to get to the bottom of
>why the ip module needs to drive this attach.  I think it would be fine
>if dlpi_open() did the attach and the ip module were modified to expect
>an already attached DLPI style-2 device...  Anyway, that part can be
>fixed as a separate RFE or bug fix.
>
>  
>
Alright, I'll leave that to be a RFE or bug fix.

-Thanks,

Sagun

Reply via email to