On (04/22/08 11:48), Artem Kachitchkine wrote:
>>   - Don't we still need lines 209-218 for wifi drivers?
>
> Good question. If we have any kind of persistence for wifi drivers, it  
> seems to be working by accident. That is, it relies on drivers not being  
> unloaded, and the system is free to do so at any moment. And on DEBUG  
> builds, will do so every 60 seconds: so even with this line present,  
> when I set speed to 54, reboot and do dladm connect-wifi, then dladm  
> show-linkprop still shows 36. I guess the question boils down to, should  
> we preserve something that works by accident? If so, I can add an option  
> for dladm init-linkprop, like -w or -m wifi, to only initialize  
> properties for wireless links, and use this option in net-physical.

I think there may be some value in leaving it there, at least until
we fix more drivers (the br-pers only has fixes for bge, but nv
has Brussels support for nge, e1000g, nxge).

>> usr/src/uts/common/io/mac/mac.c
>>   - question: what happens if err at line 2947 is EBUSY (e.g., for
>>     mtu, where driver cannot apply the property just yet)? don't
>>     we want to do_update in this case, so that the driver can pick
>>     up the setting next time?
>
> We won't do update when err is EBUSY because of:
>
>       do_update = (err == 0);

I'm suggesting that we should do_update in this case, or provide
some way of letting the user update the persistent configuration 
so that the setting is picked up on the next restart.


--Sowmini

_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to