[Openib-windows] RE: [PATCH] use of event log for link up/down msg

[Yossi Leybovich]

> -----Original Message-----
> From: Fab Tillier [mailto:ftillier at silverstorm.com] 
> Sent: Monday, December 12, 2005 9:52 PM
> To: Yossi Leybovich
> Cc: openib-windows at openib.org
> Subject: RE: [PATCH] use of event log for link up/down msg
> 
> Hi Yossi,
> 
> > From: Yossi Leybovich [mailto:sleybo at mellanox.co.il]
> > Sent: Sunday, December 11, 2005 8:23 AM
> > 
> > Fab
> > 
> > I create patch that enable to log port up\down events to 
> the log viewer.
> > I add msg file to the repository.
> > In the near future I will send patch that will add more massages to 
> > the log viewer(I will also use the NDIS_ERROR_CODE  defines from 
> > ndis.h)
> 
> I don't see much value in reporting events that are reported 
> through the link state of the adapter - the adapter icon will 
> show whether the link is up or not via the cable unplugged 
> icon.  The error log should be used only for errors.
> 

I follow the behavior of regular NICs.
For example:(from my local machine after I pull the cable out)
Error Event
Adapter Intel(R) PRO/100 S Desktop Adapter: Adapter Link Down

Info Event
Adapter Intel(R) PRO/100 S Desktop Adapter: Adapter Link Up

I also saw this behavior in other Nics (broadcom ....), you can try it
yourself.
We also got comment from customers that they would like to see these
events.(They claim that this is a requirment for NICs drivers)
Moreover I think that it good that these event will be logged so we can
track the IPoIB behavior.
So I would appriciate if we keep reporting these events .

> That said, I think you've started something very valuable 
> here by laying the foundation for error reporting in the 
> IPoIB driver.  This should be extended to report various 
> IPoIB specific errors:
> 
> - Failure to join multicast group (record port and MC group 
> link width/speed).
> - Failure of SA path queries (record destination Ethernet and 
> IPoIB MAC).
> - Failure of ATS registrations (record and SA status).

I agree,SM problem should be looged. I had this on my plan. 

> - Any other error that might prevent proper operation.
> 

I also plan to looged any init error ( like resource allocation...)

> What do you think?
> 
> - Fab
>