Failing to starta/create a VM when having a external GPU in

[Dennis H]

Hello

 

Tried to find a solution to my problem but can't find anything, even tried chatgpt with but it dient help much.

My server is a ProLiant ML350p Gen8 with an external Qudro4000 in, I am trying to patch through so that a NVR windows server can use the gpu but when I create the server I get an error and if I don't use gpun it goes through to vnc but refuses to connect and show the server and assume it dosent boot either.
Tried to check status and everything seems to be ok, my error is

internal error: qemu unexpectedly closed the monitor: 2023-03-24T13:34:24.023920Z qemu-system-x86_64: -device {"driver":"vfio-pci","host":"0000:0a:00.0","id":"hostdev0","bus":"pci.0","addr":"0x8"}: vfio 0000:0a:00.0: failed to setup container for group 31: Failed to set group container: Operation not permitted
qxl_send_events: spice-server bug: guest stopped, ignoring

status of quadro card ( think) i'm not used to linux unfortunately.

root@service:~# lspci -nnk | grep -i nvidia -A 2
0a:00.0 VGA compatible controller [0300]: NVIDIA Corporation GF100GL [Quadro 4000] [10de:06dd] (rev a3)
        Subsystem: Hewlett-Packard Company GF100GL [Quadro 4000] [103c:0780]
0a:00.1 Audio device [0403]: NVIDIA Corporation GF100 High Definition Audio Controller [10de:0be5] (rev a1)
        Subsystem: Hewlett-Packard Company GF100 High Definition Audio Controller [103c:0780]
1f:08.0 System peripheral [0880]: Intel Corporation Xeon E5/Core i7 QPI Link 0 [8086:3c80] (rev 07)
root@service:~# dmesg | grep -e DMAR -e IOMMU
[    0.010331] ACPI: DMAR 0x00000000BDDAD200 000414 (v01 HP     ProLiant 00000001 \xd2?   0000162E)
[    0.010402] ACPI: Reserving DMAR table memory at [mem 0xbddad200-0xbddad613]
[    0.458249] DMAR: Host address width 46
[    0.458316] DMAR: DRHD base: 0x000000e9ffe000 flags: 0x1
[    0.458392] DMAR: dmar0: reg_base_addr e9ffe000 ver 1:0 cap d2078c106f0462 ecap f020fe
[    0.458466] DMAR: RMRR base: 0x000000bdffd000 end: 0x000000bdffffff
[    0.458538] DMAR: RMRR base: 0x000000bdff6000 end: 0x000000bdffcfff
[    0.458609] DMAR: RMRR base: 0x000000bdf83000 end: 0x000000bdf84fff
[    0.458679] DMAR: RMRR base: 0x000000bdf7f000 end: 0x000000bdf82fff
[    0.458751] DMAR: RMRR base: 0x000000bdf6f000 end: 0x000000bdf7efff
[    0.458822] DMAR: RMRR base: 0x000000bdf6e000 end: 0x000000bdf6efff
[    0.458893] DMAR: RMRR base: 0x000000000f4000 end: 0x000000000f4fff
[    0.458964] DMAR: RMRR base: 0x000000000e8000 end: 0x000000000e8fff
[    0.459036] DMAR: RMRR base: 0x000000bddde000 end: 0x000000bdddefff
[    0.459107] DMAR: ATSR flags: 0x0
[    0.459176] DMAR-IR: IOAPIC id 8 under DRHD base  0xe9ffe000 IOMMU 0
[    0.459262] DMAR-IR: IOAPIC id 0 under DRHD base  0xe9ffe000 IOMMU 0
[    0.459333] DMAR-IR: HPET id 0 under DRHD base 0xe9ffe000
[    0.459401] DMAR-IR: x2apic is disabled because BIOS sets x2apic opt out bit.
[    0.459403] DMAR-IR: Use 'intremap=no_x2apic_optout' to override the BIOS setting.
[    0.459905] DMAR-IR: Enabled IRQ remapping in xapic mode
[    0.792056] DMAR: No SATC found
[    0.792209] DMAR: dmar0: Using Queued invalidation
[    0.805838] DMAR: Intel(R) Virtualization Technology for Directed I/O
[   20.207037] AMD-Vi: AMD IOMMUv2 functionality not available on this system - This is not a bug.
root@service:~# ls -l /dev/vfio/
total 0
crw-rw-rw- 1 root root 10, 196 Mar 24 06:27 vfio

any help would be appreciated, thanks!

Edited March 24, 2023 by Dennis H

[Reiject]

Hi - did you get anywhere resolving this? I also have picked up a ML350p Gen8 and am encountering very similar when trying to passthrough hardware

[Dennis H]

Hi Reiject

 

Wish I could say I did and give the solution, but I gave up and gutted the server and installed an unlocked motherboard. It was better anyway because of the built-in RAID card in ML350Ps, It was impossible to use JBOD. Sorry that i cant be much of a help

[escocrx]

On 8/8/2023 at 4:46 AM, Dennis H said:

Hi Reiject

 

Wish I could say I did and give the solution, but I gave up and gutted the server and installed an unlocked motherboard. It was better anyway because of the built-in RAID card in ML350Ps, It was impossible to use JBOD. Sorry that i cant be much of a help

 

Gave up because of JBOD? You can set the built in cards to HBA mode with the service pack disks. It won't work with the onboard program. I run ML350P gen 8. Too late now I guess but the file is P03093_001_spp-Gen8.1-SPPGen81.4.iso

 

You can find it for download. Load it into the physical or virtual drive and you can set them to HBA. Update firmware etc with it. 

 

I have tons of pass through issues but I don't use VM's but to play so I just give up when I do. I can get the GFX and any add on cards to pass through but onboard things like the NIC will not pass through.