Dan Groves wrote:

>
>
> Cathy Zhou wrote:
>
>> Hi,
>>
>> I am thinking about the issue we discussed yesterday, and feel that 
>> we should change the current libdladm implementation before you do 
>> any more work on this.
>>
>> One problem of Nemo's alternative path configuration today, is that 
>> when "-R" is used, libdladm still tries to change the *active* 
>> configuration of this current system, and when that succeeds, it 
>> tries to persist the configuration in the specified alternative path. 
>> But as "-R" is tends to be used to change the configuration of 
>> another system, the current processing is not correct.
>>
>
> Ouch... bug.
>
>> So I think we should first make that right. Also, as in the current 
>> system, we are not able to get any <linkname, linkid> mapping 
>> information of another system, or even validate the command. 
>> Probably, instead of tries to update the configuration in a 
>> alternative path, we should just record all the dladm subcommands in 
>> that alternation path, and when that system boot, run those commands 
>> one by one.
>>
>> What do you think?
>>
>
> Yes, let's fix the bug first.
>
> Putting the dladm commands in the alternate root is very similar to 
> what Alan Maguire suggested to me when I talked to him about his 
> experiences using SMF.  He put routeadm commands in 
> <altroot>/var/svc/profile/upgrade.  I think we could put dladm 
> commands there.


That filename is project private to SMF so I'm not sure what the 
requirements
are going to be for you to use it in a patch - you may need to execute a 
contract
between the original project (for which this patch is against) and SMF 
in order
to use it.

Darren


Reply via email to