[HN Gopher] Sipeed NanoKVM-PCIe
       ___________________________________________________________________
        
       Sipeed NanoKVM-PCIe
        
       Author : zdw
       Score  : 33 points
       Date   : 2024-12-24 02:48 UTC (20 hours ago)
        
 (HTM) web link (www.cnx-software.com)
 (TXT) w3m dump (www.cnx-software.com)
        
       | NetworkPerson wrote:
       | From the article "It would be laughable to argue the low-end
       | SG2002 AI SoC poses a threat to any country..."
       | 
       | I can see a great deal of trouble capable of coming from a
       | networked device capable of watching the screens 24x7 and
       | potentially intercepting passwords being entered. And those are
       | the legitimate functions for this device. Wouldn't take much to
       | throw a reverse shell for external access if you wanted to be
       | particularly nefarious.
       | 
       | Not saying there's any evidence this kvm is malicious. But I
       | probably wouldn't put it in anything more than one of my toy home
       | lab servers.
        
         | theamk wrote:
         | I was worrying about typical Chinese cloud you cannot turn off
         | (seems to be present on all cheap IP cameras), but this device
         | is actually pretty good.
         | 
         | For remote access, there is no cloud. But you can BYO tailscale
         | or FRP [0] (note: I really like the FRP idea, as it's trivial
         | to self-host)
         | 
         | For updating, there is a central server. But at least the
         | process seems to be manually-initiated [1].
         | 
         | I am not saying the firmware is backdoor-free, but at least it
         | would be feasible to monitor/block all outgoing network
         | connection attempts, and still have a functional device.
         | 
         | [0]
         | https://wiki.sipeed.com/hardware/en/kvm/NanoKVM/network/tail...
         | 
         | [1]
         | https://wiki.sipeed.com/hardware/en/kvm/NanoKVM/system/updat...
        
           | wkat4242 wrote:
           | What's FRP? Your source link speaks only of tailscale.
        
             | dzidol wrote:
             | Just open the link about tailscale, in the page it's one
             | tab below on the left.
        
           | stevefan1999 wrote:
           | For FRP do you mean https://github.com/fatedier/frp?
        
           | poisonborz wrote:
           | You can selfhost the control server, look at headscale, all
           | the clients support this.
        
       | mherkender wrote:
       | This is a great device but I can't imagine giving so much power
       | and control to a closed-source, self-updating device.
        
         | theamk wrote:
         | Hey, billions of people use Windows and Mac OS.
        
           | navigate8310 wrote:
           | But billions of people don't use Sipeed NanoKVM that gets an
           | OOBM access to critical infrastructure
        
             | wkat4242 wrote:
             | Yeah mine is on a non-internet-routed VLAN for that
             | purpose. I access it through my vpn only. It doesn't even
             | have outgoing internet access.
        
             | znpy wrote:
             | many more use closed source kvm solutions built-in into
             | servers, so...
             | 
             | as an homelabber, i'm using HP's iLO on my gen8 microserver
             | for example.
        
         | smcleod wrote:
         | They opened the standalone unit, assuming this will be also?
        
       | metadat wrote:
       | Are there feasible open alternatives to this closed-source blob?
       | The fundamental capabilities seem nice, on paper.
       | 
       | Also, is there Windows / Mac compatibility?
        
         | wkat4242 wrote:
         | They are open sourcing it apparently. At least they promised.
         | 
         | And yes it works fine on windows. I've got one. Haven't tried
         | it on Mac yet though.
        
       | jauntywundrkind wrote:
       | Worth mentioning Sophgo (CPU maker here) just got added to US
       | Sanction list for helping China dodge semiconductor sanctions.
       | 
       | Apparently it's the Bitmain cryptominer folk? Nice context from
       | Tom's.
       | 
       | https://www.tomshardware.com/tech-industry/artificial-intell...
        
         | Bluestein wrote:
         | Worth mentioning also, it apparently has non-configurable (to
         | off) root:root SSH on by default, according to the comments ...
        
       | toast0 wrote:
       | Looks like the pcie slot is just used for power?
       | 
       | I'd love to see something like this where the board had a basic
       | video card, so you could use it in a system without any video
       | output. Bonus if it also had a usb controller and a serial port,
       | so it didn't need to loop to plugs (although some of that could
       | happen on the internal side as well)
        
         | wkat4242 wrote:
         | Basically like Dell's old DRAC boards. They used to do exactly
         | that.
        
         | wolrah wrote:
         | Likewise, I have never been able to get a satisfying answer as
         | to why no one seems to be willing or able to put the same
         | ASpeed AST2x00 chips that it seems half the OEM
         | IPMI/iKVM/whatever solutions use on a standard PCIe card
         | instead of embedding it in the motherboard or using some
         | proprietary interface.
         | 
         | I have never been able to identify a technical barrier to doing
         | this, the important features most people actually care about
         | are implemented over a 1x PCIe link and USB, plus a couple of
         | GPIOs to twiddle the power/reset button connections. Most OEM
         | implementations also connect to the LPC bus and others on the
         | server board to allow more in depth diagnostics, voltage
         | logging, etc. but those are bonus features and not requirements
         | for a useful product. I do not see any technical reason a
         | useful generic PCIe implementation couldn't be produced, and as
         | a result I have a strong feeling that the lack of such products
         | is an intentional choice by one or more of the vendors involved
         | to increase margins by pushing users who want these features up
         | to entry level server boards instead of sticking a card in a
         | higher-end desktop board that might better fit their needs.
        
           | toast0 wrote:
           | M.2 A or E might be better for this actually. A lot of boarda
           | have slots for wifi/bluetooth with PCIe and USB. Would need a
           | cable to a panel mount network jack and to pull in the front
           | panel switches.
        
       | smcleod wrote:
       | I have the standalone unit and other than the painfully slow
       | 100mbit Ethernet that's too slow to upload ISOs and which also
       | doesn't work with many modern switches - it's really nice for the
       | price.
       | 
       | The problem with a pcie one for me is that modern motherboards
       | suffer from having hardly any PCIe ports - and when they do
       | they're mashed in close to each other essentially making one
       | useless if you have a decent GPU.
        
       | crest wrote:
       | One the one hand adding radios (WiFi, LTe) to KVM over IP device
       | sounds tempting on the other hand given the track record of KVM
       | over IP devices it sound terrifying to give them the ability of
       | bypass points of policy enforcement.
        
       ___________________________________________________________________
       (page generated 2024-12-24 23:02 UTC)