Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Local Fault/Remote faults





Bipin,

Bipin Dama wrote:
> 
> Ben,
> 
> I have a question about generating Local Fault in TX
> direction. It is embeded in this email.
> 
> Thanks
> 
> -bipin d. dama
>  bipin_dama@xxxxxxxxxxxxxx
>  (610) 289 5218
> 
>         Suppose that we have XAUI retimers on both ends of the fiber.
> 
>         MAC1->RS1_TX->PCS1_TX->XAUI-RETIMER1_TX ->  ---------- FIBRE ---------> XAUI-RETIMER2_RX->RS2_RX->MAC2
>         MAC1<-RS1_RX<-PCS1_RX<-XAUI-RETIMER1_RX<- ----------- FIBRE ---------<-XAUI-RETIMER2_TX<-RS2_TX<-MAC2
> 
>         What happens when the XAUI-RETIMER2_TX losses sync ? Is it suppose to generate Local Fault ? If the answer is
>         yes, then what additional value does it bring as compared to doing nothing ? If it does nothing then the XAUI-RETIMER1_RX
>         will loose sync and will generate Local Fault due to receiving junk.
>         As generation of this Local Fault does not help in fault isolation, I suggest we dont require that a Local Fault be generated
>         towards the Fibre.
> 

If it is a true retimer, I doubt it has the capability to
actually insert an LF or any other value into the data stream.
If it is actually a PHY XGXS backed up against a 10GBASE-X PCS
then it does indeed have the capability to insert the LF.

I think this does indeed help with fault isolation. When the
RS1_RX device in your (now somewhat scrambled) picture above
detects local fault, STA1 knows there is a problem somewhere
between RS1_RX and the RX2_TX device. As STA1 polls its own 
devices, they will all report that everything is operating
normally. It will then alert the STA2 in charge of MAC2 and
PHY2 with what it has seen. The STA2 will poll its devices
and, lo and behold, XAUI-RETIMER2_TX will announce that it
isn't in sync. The problem will be found.

If the local fault was not generated by XAUI-RETIMER2_TX, it
has to generate something. Should that be IDLEs? This is bad
because then no one would know anything was wrong. Should it
be all zeros? This would mean the XAUI-RETIMER1_RX wouldn't
get sync (or perhaps even signal_detect?). Then it would be
hard to know if the problem was with XAUI-RETIMER1_RX or if
it was with the FIBRE plant.

Regards,
Ben

-- 
-----------------------------------------
Benjamin Brown
AMCC
2 Commerce Park West
Suite 104 
Bedford NH 03110
603-641-9837 - Work
603-491-0296 - Cell
603-626-7455 - Fax
603-798-4115 - Home Office
bbrown@xxxxxxxx
-----------------------------------------