Re: [Linux-ATM-General] Solos driver weirdness in bridged mode

[chas w. - C. <ch...@cm...>]

On Tue, 22 Mar 2011 13:13:05 -0700
Philip Prindeville <phi...@re...> wrote:

> On 3/22/11 12:55 PM, chas williams - CONTRACTOR wrote:
> > On Tue, 22 Mar 2011 12:36:25 -0700
> > Philip Prindeville<phi...@re...>  wrote:
> >
> >> So for users with the older FPGA firmware, that doesn't see the carrier state change messages, how should we handle this?
> > just update the carrier/signal status of the adapter.  initially
> > configure that state to SIG_UNKNOWN.  the logic to determine the value
> > of the carrier sysfs value will get the rest right.
> 
> Ok, but I thought we had agreed to change the initial state to LOST instead of UNKNOWN?

it is a little confusing/muddled.  when you create/allocate the driver
it is set to UNKNOWN for you.  really it should be set to FOUND on the
assumption that the driver does not handle carrier for you.  carrier is
assumed to be 1 in the FOUND or UNKNOWN state otherwise 0.  the
UNKNOWN state probably should just go away but we preserved it for
the time being.  of course, if your driver supports determining the
signal/carrier status, you should initialize to LOST.  but this
might/will yeild a transition from UNKNOWN to LOST.  part of the
problem here is that the device registration for the linux-atm stack is
monolithic.  you dont allocate the device, set it up and then register
otherwise you could avoid the 'false' transition.

> BTW:  Is there an easy way to detect which version of firmware we're talking to?  Or do we need to add a module parameter?

got me.  i would just handle it as above.  assume you have carrier. if
the firmware happens to tell you otherwise, update the carrier status.
this is what ethernet does and it seems to work well enough.