[TTM] Buffer eviction failed

[tanolis]

I am using following Proxmox version.

# pveversion --verboseproxmox-ve: 8.2.0 (running kernel: 6.8.12-1-pve)
pve-manager: 8.2.4 (running version: 8.2.4)
proxmox-kernel-helper: 8.1.0pve-kernel-5.15: 7.4-13
proxmox-kernel-6.8: 6.8.12-1
proxmox-kernel-6.8.12-1-pve-signed: 6.8.12-1

In my System Log, I keep getting these errors;
Aug 12 19:07:27 pvetower kernel: [TTM] Buffer eviction failed
Aug 12 19:07:27 pvetower kernel: qxl 0000:00:1e.0: object_init failed for (4096, 0x00000001)
Aug 12 19:07:27 pvetower kernel: [drm:qxl_alloc_bo_reserved [qxl]] *ERROR* failed to allocate VRAM BO

For debugging and resolution, Can someone point me to the right direction?

 

[THX1138]

I get exactly the same error (except for the precise address, which I expect will be machine specific). In my case, it's not actually ProxMox, but rather a qemu/kvm virtual machine (on Linux). ProxMox uses the kvm hypervisor, suggesting that it's a kvm problem. QXL is the default VM video driver for qemu/kvm, and the error includes reference to 'VRAM", so I guess it's related to that. I'm also running the 6.8 series kernel (6.8.0-40).

This issue report from RedHat seems to support the theory that the problem is in kvm or one of its drivers:
https://access.redhat.com/solutions/7004385

RedHat says the status is Solution in Progress.

 

[jebbam]

I have seen a similar error occassionally for years... My setup and how I see it:

The KVMs are running X with the qxl driver: "SPICE (qxl, memory=128)" is selected under the "Hardware" tab. I'll also note that the max "memory=" I can get it to boot with is 128. Anything larger than that and it doesn't even get to the Proxmox BIOS when powered on. So I always run them with "memory=128".

I heavily use many SPICE KVMs, all day long.

It almost always happens when I am actively using the SPICE exported display on my workstation. In other words, I don't come back to see a crash, it seems to happen more when I'm switching back and forth between windows or just created a new Firefox window or something like that. It isn't happening because of a sleep or suspend or anything like that (which I've seen mentioned elsewhere).

I'll have KVMs run for days, weeks, months without a problem, then it crashes X. Note, the KVM itself keeps running, and I can log into it, run `sync` etc ok. But doing a `poweroff` will never shut down the KVM. I have to do a hard "Stop" of the KVM from Proxmox web gui.

Example crash:

2024-08-30T14:49:08.981384-06:00 muh-kvm kernel: [191594.195388] [TTM] Buffer eviction failed
2024-08-30T14:49:08.981415-06:00 muh-kvm kernel: [191594.195435] qxl 0000:00:02.0: object_init failed for (262144, 0x00000001)
2024-08-30T14:49:08.981417-06:00 muh-kvm kernel: [191594.195440] [drm:qxl_gem_object_create [qxl]] *ERROR* Failed to allocate GEM object (260868, 1, 4096, -12)
2024-08-30T14:49:08.981432-06:00 muh-kvm kernel: [191594.195465] [drm:qxl_alloc_ioctl [qxl]] *ERROR* qxl_alloc_ioctl: failed to create gem ret=-12

I'll also note, I usually enable allowing multiple SPICE connections, but I think the error happens whether I do that or not. I don't think it is related, but I mention it, just in case. I edit `/usr/share/perl5/PVE/Tools.pm` and add this line so I can have SPICE connections to the same KVM from multiple workstations (which is awesome, btw):

local $ENV{SPICE_DEBUG_ALLOW_MC} = "1";

I run Debian on the KVMs. It has happened with Debian bullseye, bookworm, trixie, and sid...

I would love to see a solution to this, it has bugged me for years!

 

[Valombre]

Hi,
same problem with some of my VMs ubuntu mate 24.04 LTS, and with graphical crash, i have to reconnect.
It happens a lot with firefox.
Here some logs :

2024-09-02T15:07:55.319113+02:00 vmtest kernel: [TTM] Buffer eviction failed
2024-09-02T15:07:55.319130+02:00 vmtest kernel: qxl 0000:00:02.0: object_init failed for (258048, 0x00000001)
2024-09-02T15:07:55.319131+02:00 vmtest kernel: [drm:qxl_gem_object_create [qxl]] *ERROR* Failed to allocate GEM object (256020, 1, 4096, -12)
2024-09-02T15:07:55.319132+02:00 vmtest kernel: [drm:qxl_alloc_ioctl [qxl]] *ERROR* qxl_alloc_ioctl: failed to create gem ret=-12
2024-09-02T15:08:00.179099+02:00 vmtest kernel: Lockdown: Xorg: raw io port access is restricted; see man kernel_lockdown.7
2024-09-02T15:08:00.941091+02:00 vmtest kernel: Lockdown: systemd-logind: hibernation is restricted; see man kernel_lockdown.7
2024-09-02T15:08:00.951091+02:00 vmtest kernel: Lockdown: systemd-logind: hibernation is restricted; see man kernel_lockdown.7
2024-09-02T15:08:01.986153+02:00 vmtest kernel: workqueue: page_reporting_process hogged CPU for >10000us 4 times, consider switching to WQ_UNBOUND

Regards

 

[jebbam]

It happens a lot with firefox.

Yes, I see it most often with Firefox too, especially when creating a new window or switching windows. It also seems to happen more when I'm "going fast", bouncing between windows, etc. I just had it happen with Thunderbird when creating a new task (which creates a new window).

In the past, I saw it more and I was blaming urxvt, which was my main terminal back in the day. I switched to using lxterminal as my main terminal, and it seems to happen less with that, but that's just a guess.