[vortex] Re: Case 254686: GX150 autonegotiate problem

[Bill Cattey]

Status report:

I have charged the my Linux Platform lead, Andrew Boardman, with doing
the work to get the candidate 3c59x.o modules compiled for testing.  (I
tried myself but missed some bit of lore and got no working module.  So
I'll wait for Andrew.)  Andrew has a couple of other tasks to complete
before doing this.  (Tasks that affect even larger numbers of machines.)
 So while I wait, I have continued testing.  This note contains
questions to help us move forward in understanding.

Testing:

I have looked at several specific systems, one of which nearly always
powers up in a network-unhappy state.  According to vortex-diag,
ALL the GX150's both happy and sad ones have Enet chipset manufacture
dates of 6/10/2000.

Note that my phrase "powers up" is significant.

If I unplug, then replug the network cable, it renegotiates, and every
time I've tried it, the negotiation results in happy network state.

But power cycling the machine results almost always in autonegotiation
of an unhappy state.

Right now, mii-diag -v says that 100baseTx-FD was negotiated, but the
full duplex light on my Netgear unmanaged 10/100 switch is OUT!  (It
does light up when the machine is in happy network state.)

It seems to me likely that we're talking about some chipset power-on
initialization that is flaky.

QUESTION:

Does the code for which I've received diffs still seem like the relevant
code to be poking at given these new observations?

-wdc