I'd say, it can be a "permanent" fix. The problem is, the worse the degradation on the LPC_CLOCKOUT port has gotten, the lower value resistor you need to use for the pull-up. Lower the pull-up resistance, higher the load on the low side switch, which can also break.
Thanks for details. I looked over some forum threads for Cisco routers affected by this bug and the fix they used was 150Ohm (≤1%) (SMD 18A).
After installing the sonic, use dpkg -i to install the platform modules.
Where did you find the platform modules file? I have the main file used for all Broadcom based switches but I couldn't find the rest...After installing the sonic, use dpkg -i to install the platform modules.
Leds work now too
https://raeb.ae/sonic-broadcom.bin
https://raeb.ae/platform-modules-d4040_0.9_amd64.deb
I made it
The transceiver hot-plugging is working too.
show interfaces transceiver eeprom -d
Ethernet96: SFP EEPROM detected
Connector: MPOx12
Encoding: 64B66B
Extended Identifier: Power Class 1(1.5W max), CLEI present
Extended RateSelect Compliance: QSFP+ Rate Select Version 1
Identifier: QSFP+ or later
Length OM3(2m): 50
Nominal Bit Rate(100Mbs): 103
Specification compliance:
10/40G Ethernet Compliance Code: 40GBASE-SR4
Fibre Channel link length/Transmitter Technology: Short distance (S)
Fibre Channel transmission media: Shielded Twisted Pair (TP)
Vendor Date Code(YYYY-MM-DD Lot): 2015-05-11
Vendor Name: FINISAR CORP
Vendor OUI: 00-90-65
Vendor PN: FTL410QE2C
Vendor Rev: A
Vendor SN: XTG04EH
ChannelMonitorValues:
RX1Power: -infdBm
RX2Power: -infdBm
RX3Power: -infdBm
RX4Power: -infdBm
TX1Bias: 0.0000mA
TX2Bias: 0.0000mA
TX3Bias: 0.0000mA
TX4Bias: 0.0000mA
ModuleMonitorValues:
Temperature: 31.6836C
Vcc: 3.2948Voltssort of, the "left side" 2.54 pins are for power, pcie can be traced coming off the cpu. So there's LPC & pcie + uart + ethernet coming on to those two "dense" connectors. It also crossed my mind, to replace the whole atom board with some itx x99 + 2600 v4 xeon
I'm heading that direction. I've got 2 dead D4040s. They both have different model main boards. One of them I had working on a Friday, but when Monday rolled around it too had a dead console. Frustrating...sort of, the "left side" 2.54 pins are for power, pcie can be traced coming off the cpu. So there's LPC & pcie + uart + ethernet coming on to those two "dense" connectors. It also crossed my mind, to replace the whole atom board with some itx x99 + 2600 v4 xeon
One other option would be, to use the xo (~14MHz, if I remember correctly) near the cpu as a clock source for a fnpll board to derive the 25MHz clock for the lpc.
You're significantly more knowledgeable than I am here, and I'm greatly appreciative of it. I was going to just build myself a breakout board for the B2B connector and run it to Molex iPass receptacle. I've got some old PCIe cables and a few spare One Stop Systems Max Express cards to play with.
It's a very tempting idea. It'd be fast enough to do routing/nat etc on the cpu.
The SMT connectors are less than $2.50/ea. I'll have 10 of them in a couple days.
I did already install the switch in it's place but I'll try to get it done.The SMT connectors are less than $2.50/ea. I'll have 10 of them in a couple days.
I'll be out of town this weekend, but I should have a rough PCB design by next week. Printing and cutting will take a couple weeks, though.
Any chance you could trace the PCIe pins? Again, you're much more knowledgeable here than I am. If so I can get that into the PCB design, rather than just doing a simple breakout. And If you're willing, I'd love to send you some test boards.
