@samuelolavo Yes, I do use NUT, but I don’t want to install any additional packet on XCP-ng. The way I found leverage the SSH authorized keys restrictions, and so you only need to change one file. Thanks for the link anyway

Does it work as you wanted now? If yes, probably something we can put in our official doc! Ping @thomas-dkmt

Hi, In those cases, my first bet is a memtest to check the RAM and CPU.

@TeddyAstie Sure thing: xl_dmesg.txt Dom0_dmesg.txt DomU_dmesg.txt

Also works with alt driver (broadcom-bnxt-en-alt-1.10.3_231.0.162.0-1.xcpng8.2.x86_64) in 8.2.1: # yum install broadcom-bnxt-en-alt # reboot # modinfo bnxt_en filename: /lib/modules/4.19.0+1/override/bnxt_en.ko version: 1.10.3-231.0.162.0 description: Broadcom NetXtreme-C/E/S network driver license: GPL srcversion: 533BB7E5866E52F63B9ACCB alias: pci:v000014E4d0000D800sv*sd*bc*sc*i* ... alias: pci:v000014E4d00001604sv*sd*bc*sc*i* depends: devlink,ipv6 retpoline: Y name: bnxt_en vermagic: 4.19.0+1 SMP mod_unload modversions # dmesg|grep bnxt_en [ 15.047312] Broadcom NetXtreme-C/E/S driver bnxt_en v1.10.3-231.0.162.0 [ 15.048046] bnxt_en 0001:5c:00.0 (unnamed net_device) (uninitialized): Device requests max timeout of 60 seconds, may trigger hung task watchdog [ 15.215453] bnxt_en 0001:5c:00.0 eth0: Broadcom BCM57508 NetXtreme-E 10Gb/25Gb/50Gb/100Gb/200Gb Ethernet found at mem 2affff010000, node addr 40:5b:7f:6e:22:00 [ 15.215469] bnxt_en 0001:5c:00.0: 252.048 Gb/s available PCIe bandwidth (16 GT/s x16 link) [ 15.216606] bnxt_en 0001:5c:00.1 (unnamed net_device) (uninitialized): Device requests max timeout of 60 seconds, may trigger hung task watchdog [ 15.265361] bnxt_en 0001:5c:00.1 eth1: Broadcom BCM57508 NetXtreme-E 10Gb/25Gb/50Gb/100Gb/200Gb Ethernet found at mem 2affff000000, node addr 40:5b:7f:6e:22:01 [ 15.265372] bnxt_en 0001:5c:00.1: 252.048 Gb/s available PCIe bandwidth (16 GT/s x16 link) [ 15.350702] bnxt_en 0001:5c:00.1 side-5183-eth1: renamed from eth1 [ 15.385121] bnxt_en 0001:5c:00.0 side-3345-eth0: renamed from eth0 [ 17.400142] bnxt_en 0001:5c:00.0 eth0: renamed from side-3345-eth0 [ 17.493640] bnxt_en 0001:5c:00.1 eth1: renamed from side-5183-eth1 [ 20.614447] bnxt_en 0001:5c:00.1 eth1: NIC Link is Up, 100000 Mbps (NRZ) full duplex, Flow control: none [ 20.614454] bnxt_en 0001:5c:00.1 eth1: FEC autoneg off encoding: None [ 21.113687] bnxt_en 0001:5c:00.0 eth0: NIC Link is Up, 100000 Mbps (NRZ) full duplex, Flow control: ON - receive & transmit [ 21.113694] bnxt_en 0001:5c:00.0 eth0: FEC autoneg off encoding: None

hardware_report_2025-12-12_17-20-47.txt lspci.txt mlx.txt pci.txt Diagnostic logs

@hoehnp we're aiming at sharing a very first public version before end of year, but don't hope for it to be anything complete or stable

@Joe_dev I have a MF unit with similar igpu. I was not able to get it working under any virtualization platform. There is something about amd igpu that makes passing it hard. I believe some have had it working under proxmox with custom kernel and video bios. I never bothered trying all that out so i just gave up on the unit.

Maybe that's my issue ? Investing this part # journalctl -k | grep -i nvidia [image: 1763219509282-aaed86b6-1384-434e-ac5d-bca510405f33-image.png]

Since GPU passthrough is involved I'd ping @Team-Hypervisor-Kernel

@ph7 huh, that is weird. I'll try to troubleshoot more over the weekend. Thanks for pointing it out. @olivierlambert said in Upgraded RAM, It get's detected but not used. Any ideas?: I would first upgrade the BIOS of the motherboard. Consumer grade hardware is pretty buggy in general when you are using "advanced" features, sadly. That's the first thing i tried, no luck unfortunately

Hi, Are you sure your GPU is correctly fed via the power connector?

What GPU do you have? I thought I read something about NVidia's newer stuff having vGPU without the expensive licensing and hardware. But it's possible I'm just conflating that with the increase in concurrent NVENC encodes. On the bright side, you can always get a cheap GPU and just do passthrough with it. I do that for my Jellyfin server with a 2060, it's very easy to do in XCP-ng now and works super reliably.

@andSmv Thanks for taking a look at it. It's not really needed for a modern platform.

Reboot seems to have fixed it and made it visible. Leaving this post up so others have all the commands in one place. [14:15 cloud ~]# ethtool -i eth5 driver: i40e version: 2.26.8 firmware-version: 9.50 0x8000f255 23.0.8 expansion-rom-version: bus-info: 0000:5e:00.1 supports-statistics: yes supports-test: yes supports-eeprom-access: yes supports-register-dump: yes supports-priv-flags: yes

@olivierlambert thank you very much, you are helpful

@gduperrey The thing is, the i40e driver works fine for me with the regular kernel. I was hoping to try kernel-alt to see whether it would resolve my erroneous CPU thermal shutdown issues (to narrow down whether it's a kernel issue or not), but then I ran into the driver issue. I can try that alt driver, though, and see if it works with kernel-alt for me.

@chicagomed there's no need to be editing grub for pass thru, I would use the standard commands and avoid corrupting grub in the process.

@olivierlambert said in HCP-NG 8.3 freezes and / or reboot on HGR-SAP3 by Ovh: You can edit your first post, which contains the title thanks a lot I'll wait until next week to disable deduplication (never on Friday )