KVM switch causing FreeBSD 15-Release install to reset/reboot?

Jake0162

Jake0162

I've just upgraded from 14.3 --> 15.0 release. No issues with the upgrade process itself, but now I'm seeing issues when I'm switching PCs using my KVM switch(KCEVE KC-KVM204D3).
When I switch away from my FreeBSD workstation it'll cause it to reboot my workstation(FreeBSD PC).

This wasn't ever an issues before upgrading, so I'm assuming it might have something to do with moving from drm-61 to drm-66 and something breaking the gpu driver's hot-plugging.

The strangest thing is that there is nothing in '/var/log/messages', or from dmesg to indicate anything at all happening after this. It's like someone just smacked the reset switch and it's kinda driving me crazy as I often need to switch between my workstation PC and my other PCs for testing stuff bare metal.

If anyone has any insight into this or has had similar issues with their KVM switches that information would be super helpful, otherwise I'll try to update this post with my findings if I can solve the issue.

For some context my current workstation hardware:

CPU(s): 2x Intel Xeon E5-2698 v3
RAM: 256gb DDR4 ECC 2133MHz
NIC0: Broadcom NetExtreme II PCIe
NIC1: Realtek 8111H (Not in use).
GPU: RX480 8GB
USB: USB3.1 PCIe 5 port card.
PSU: SeaSonic Focus Gold 750W
Cooling: 2x 360mm rads + 8 Delta 120mm
 
Jake0162 said:
I've just upgraded from 14.3 --> 15.0 release. No issues with the upgrade process itself, but now I'm seeing issues when I'm switching PCs using my KVM switch(KCEVE KC-KVM204D3).
When I switch away from my FreeBSD workstation it'll cause it to reboot my workstation(FreeBSD PC).

This wasn't ever an issues before upgrading, so I'm assuming it might have something to do with moving from drm-61 to drm-66 and something breaking the gpu driver's hot-plugging.
Click to expand...
I think you are correct in your assumptions. Have you tested the KVW while freeBSD is in text console mode, or just while running a GUI?
 
kent_dorfman766 I've tested it both ways, in GUI, and in console mode. Same behavior either way.

I got a new KVM switch today with built in EDID support an S7241PH3 and tested it out, same exact behavior. So I don't think it's related at all to that.
 
Okay I've done some more looking and I'm starting to wonder if it could be an issue with back feeding across machines through the USB ports.

I cracked open the new KVM switch and found some pretty interesting things.

  • GL3510 USB HUB IC, actually a 3.1USB configured part not 3.0 so that good I guess.
  • PI3HDX ics, these are redriver ICs for HDMI2.1 and Dual mode Displayport level-shifters.
  • 3157 1CH SPDT analog switches for mostly the USBD+ and USBD- lines I think.
  • A34/A24 (FUSB340TMX?) I think these are the USB SS line switches as their rated for the correct speed.
  • MC51F03A4 (crappy 8051 micro-controller) hard to tell as the serial numbers are sanded off but I got that number from the TX/RX serial lines & matches pinout.
Okay for the not so great stuff:

The input or power supply for it supposed to be powered through the USB-C 6 pin port and it does run through it and a toggle switch leading to an LDO 3.3v voltage regulator the 1117-3.3 so that part of the design is fine and even has a decent solid state(poly/alum) 220uF capactor and some extra bypass caps there.

The big issue is the PC connection ports that are supposed to go from the PC1/2 <--> KVM as when I disconnect the USB-C power input connection the KVM is being powered by the PC1/2 connections.

That's less than ideal as I don't want this thing soaking up as much power as it wants from my motherboard.

The worse problem is that when I supply the KVM with the external 5V USB-C connection it can backfeed the power through the PC1/2 USB connections, I can read 4.98v on them.

Honestly the more I look at the PCBs the more I hate it, it doesn't have any ESD protection IC on the input, no TVS diodes, no load switch ICs for over-voltage, over-current and over-temp protection at all.

The USB hub IC also has a built in voltage regulator for going from 5V --> 3.3V but instead of using it they use the 1117-3.3 and the built in regulator.
 
Jake0162 said:
The big issue is the PC connection ports that are supposed to go from the PC1/2 <--> KVM as when I disconnect the USB-C power input connection the KVM is being powered by the PC1/2 connections.
Click to expand...
Every KVM I have ever owned does this. Currently a couple of ATENs.
 
tingo I wish but no not really.

trev Yes they can be powered from the PC connection on lots of them, but this not only does that but put's a voltage on the USB port itself.

When I say this I mean I can measure a voltage supplied from PC1 on the USB input connection meant for PC2. If the voltage is higher on PC1 than PC2 it will back-feed into PC2.

I think that's a big part of why it gets so toasty as well.

I initially assumed this wasn't going to be an issue because I saw the S14 barrier diode on the PCB for each PC connection, but it seems it's bypassed somewhere else, I'm assuming on an internal layer as it looks like a 4-layer PCB.

It could also be that it's just been fried from ESD at some point as it has no protection again that.

Andriy
As an update to the issue, yes I had my machine setup for generating core files and dumps, but as far as I can tell it's the equivalent of a hardware reset event.
 
You must log in or register to reply here.
Back
Top