Latest Update 5.13.19-4-pve broke my QEMU PCIe Sharing. Works with 5.13.19-3
- Thread starter bluepr0
- Start date
You dont even need the test repo, even according to the OP from PVE-Team:
The apt update and install should be enough.
Thank you Dunuin.
I added the test repository and could upgrade the kernel to 5.15.19-1-pve. However it looks like the upgrade caused additional issue, as the I am struck at
Code:
Loading Linux 5.15.19-1-pve
Loading initial ramdisk.....Now I see the below error messages in dmesg, which were not there earlier.
Code:
UBSAN: array-index-out-of-bounds in drivers/scsi/megaraid/megaraid_sas_fp.c:103:32
[ 1.672599] index 1 is out of range for type 'MR_LD_SPAN_MAP [1]'Screenshot attached.
Attachments
Sound like problems with the raid drivers of the 5.15 kernel:
https://www.mail-archive.com/linux-scsi@vger.kernel.org/msg80296.html
https://www.mail-archive.com/linux-scsi@vger.kernel.org/msg80296.html
You are very famous member and should be using Proxmox for few months. I am about to use Proxmox for some of my work related to Homelab, Home Automation, NAS, Media Server etc..
I am facing issues after one another in the last couple of week with Proxmox, getting VMs, setting up zfs, getting NAS server up, installing windows etc..
Is this how it usually is? I am reconsidering my decision to use Proxmox and wanted to evaluate any other similar products.
Can you please share your Proxmox experiences?
Thanks,
Proxmox is very powerful and allows you to do nearly anything as its a full Linux and not just an appliance. But as always with Linux, you need to know what you are doing and have years of experience to do it right. PVE is targeted at qualified administrators and not at consumers and you need to configure alot of stuff yourself using the CLI or even write your own scripts as not everything can be done using the GUI. And all this freedom and required knowlage doesn't make it easy for beginners.
But stuff like setting up ZFS for example is everywhere the same. Its not more difficult here than anywhere else. ZFS is just a complex enterprise grade storage solution where you need to learn first on how to use it. If you don't want to invest time to learn new stuff (like buying a book about ZFS and reading it first before setting up your ZFS pool) then other hypervisors might be a better choice.
But stuff like setting up ZFS for example is everywhere the same. Its not more difficult here than anywhere else. ZFS is just a complex enterprise grade storage solution where you need to learn first on how to use it. If you don't want to invest time to learn new stuff (like buying a book about ZFS and reading it first before setting up your ZFS pool) then other hypervisors might be a better choice.
Last edited:
Thanks for sharing your thoughts about Proxmox.
I am not afraid of the learning curve it requires, as I am looking at using the built features to start with. I am also happy with the community, who take time to respond and help others.
I will keep digging and learning more and hope to continue to use Proxmox.
I am not afraid of the learning curve it requires, as I am looking at using the built features to start with. I am also happy with the community, who take time to respond and help others.
I will keep digging and learning more and hope to continue to use Proxmox.
It's never like this. You're just seeing an unfortunate series of events as you're onboarding. I've been using Proxmox for 2+ years without issue. Major VE changes might occasionally have an issue (I do seem to recall an issue with PCI passthrough back during one of the major or minor PVE 6.x updates), but they're not usually breaking changes like this.
Don't let this scare you off. I still think Proxmox is the best tool for the job, if for no other reason than its backend stack of KVM/qemu. With other hypervisors, you start to run into the territory of closed source and a lack of community support.
Jup. Its hard to setup everything right but when its running it is rock solid...even if you just use the no-subscription repository. Community support is great and when bugs are reported they are usually fixed within a few days or even hours. But if your uptime is important its always a good idea to get a subscription to be able to access the enterprise repository. Its always some weeks behind and therefore better tested, because bugs hit the no-subscription repo users first, they complain about them and bugs get fixed before the update gets added to the enterprise repo.
I really like that everything is transparent. You can watch the mailing list, watch the progress in the bug tracker and the staff is also reporting back in the forums and point to the updates/patches after the bug is fixed.
Last edited:
That's very encouraging to hear. We learn from the experiences and when the community helps from their learnings, its really good.
Today morning one of the user creating a ticket and within few minutes, one of the staff member acknowledging the PCI passthrough issues. Thanks for helping out.
FYI: That's a backport request from 2019, that patch is in our kernels since ages, i.e., here the link to the commit in our 5.13/impish kernel mirror:
https://git.proxmox.com/?p=mirror_u...ff;h=ad669505c4e9db9af9faeb5c51aa399326a80d91
Last edited:
Today there was an update and now the kernel it's working correctly. I think you can safely update now. Here's the log to check the versions
Code:
root@pve:~# apt update
Get:1 http://download.proxmox.com/debian/pve bullseye InRelease [3,053 B]
Hit:2 http://security.debian.org bullseye-security InRelease
Hit:3 http://ftp.es.debian.org/debian bullseye InRelease
Hit:4 http://ftp.es.debian.org/debian bullseye-updates InRelease
Get:5 http://download.proxmox.com/debian/pve bullseye/pve-no-subscription amd64 Packages [213 kB]
Fetched 216 kB in 0s (451 kB/s)
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
5 packages can be upgraded. Run 'apt list --upgradable' to see them.
root@pve:~# apt upgrade
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
Calculating upgrade... Done
The following packages will be upgraded:
libpve-rs-perl libpve-storage-perl pve-firmware pve-kernel-5.13.19-4-pve pve-kernel-helper
5 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
Need to get 126 MB of archives.
After this operation, 10.5 MB of additional disk space will be used.
Do you want to continue? [Y/n] Y
Get:1 http://download.proxmox.com/debian/pve bullseye/pve-no-subscription amd64 libpve-rs-perl amd64 0.5.1 [1,301 kB]
Get:2 http://download.proxmox.com/debian/pve bullseye/pve-no-subscription amd64 libpve-storage-perl all 7.1-1 [129 kB]
Get:3 http://download.proxmox.com/debian/pve bullseye/pve-no-subscription amd64 pve-firmware all 3.3-5 [57.5 MB]
Get:4 http://download.proxmox.com/debian/pve bullseye/pve-no-subscription amd64 pve-kernel-5.13.19-4-pve amd64 5.13.19-9 [66.9 MB]
Get:5 http://download.proxmox.com/debian/pve bullseye/pve-no-subscription amd64 pve-kernel-helper all 7.1-10 [12.4 kB]Yeah, that was a 5.13 based one fixing an regression in the block device (disks) part of the kernel:
https://forum.proxmox.com/threads/k...g-vm-with-hba-passthrough.104380/#post-449697
IMO it'd slightly weird for that issue to cause more fallout with PCIe/USB stack in general, but not out of the possible.
What sticks out from your update command: you got a newer pve-firmware package too, so maybe the updated FW versions helped. Anyhow, happy to hear if it works now for you!
Thank you for fixing the issue. PVE is booting up nicely with all vm's starting. All the passthroughs of HBAs, Disks, controllers and USBs are going smoothly now too.
@ t.lamprecht and the rest of the proxmox team: good job guy´s 
Here is working passthrough fine again.
Here is working passthrough fine again.
Hi, everyone;
Yesterday 02/26/22, I upgraded to kernel 5.13.19-4 pve, It caused one of my nodes (they are in a cluster) failed to boot, this node has X5675 cpu, when I reboot it, I do not see any grub screen, just go into a reboot loop. Wondering anybody has similar issue with this older hardware?
This cpu was released on 2010 something like that.
many thanks
Yesterday 02/26/22, I upgraded to kernel 5.13.19-4 pve, It caused one of my nodes (they are in a cluster) failed to boot, this node has X5675 cpu, when I reboot it, I do not see any grub screen, just go into a reboot loop. Wondering anybody has similar issue with this older hardware?
This cpu was released on 2010 something like that.
many thanks