Proxmox user base seems rather thin?

keeka said:
As a newcomer to Linux and Linux-based virtualisation, your time might be better spent putting this network card issue aside for a while and sticking with PVE8. I have a similar network card in my setup, so was interested in your report. It's a common card so should surface again soon as more users upgrade and experienced users take an interest in it. But if it is associated with a specific and old hardware combination (motherboard?) you may be out of luck. I have winged it for a long time on old hardware but PVE9 may present some issues for me! I may be OK in this instance as I have not enables passthrough and your problem looks to be connected to IOMMU.
I have not come across another community forum where you'll find the same level engagement as you do from the staff here. This forum is an amazing resource. My only problem is using google these days to find the posts that help me.
Click to expand...

I didn't specifically enable passthrough, and I'm just a baby newcomer to Linux, so I don't know if passthrough is something that happens during install or not. Yet :). What is super weird is that the onboard NICs (two of the four anyway - the other two are x540, which is an issue I'll get to after this one is resolved) are recognized and functional, and I have to assume that they are being assigned address space, but the T4 is unable to reserve address space.

Don't feel alone in regard to google. It works, of course, but takes a lot of sifting. I've been using Grok and GPT-4 for searches, which returns source and synopsis.
 
tinus_p said:
Proxmox VE is built on Debian, chances are if there is a problem with the kernel and these network cards, it also appears on a stock Debian installation. Did you happen to check if they work if you run Debian 13 (could just try the live image)?

If not, you might be more successful in getting this fixed by reporting the issue to the Debian developers.
Click to expand...

Yep, they sure did work on Debian, but I'm going to reinstall it on another drive to test it out today. I'll post back what I find. I did see during my searching that it did crop up in Debian installs, but that was usually when someone was trying to passthrough to a VM. In this case, in this particular install, I haven't added any VMs.
 
leesteken said:
Proxmox's Linux kernel (6.14) is based on Ubuntu instead of Debian and since drivers come with the kernel, maybe try an Ubuntu (installer without installing it) with the same kernel version (25.04).
EDIT: The user-space is indeed based on Debian stable (13).
Click to expand...

I've also installed Ubuntu and Ubuntu server. Both saw all the NICs except the two x540 (also Intel and supposedly the same driver). Maybe I'll try all three again today.
 
mgibson said:
I don't have an intel NIC, but I see there are a number of posts on the forum that are about this card (here, here and here). Are you able to modify grub as in this post?
Click to expand...

I did read all of those a couple days ago. They seemed similar, but different in their own ways, with the possible exception of the Grub edit (which I haven't checked into yet). I do have them saved in my resource bookmarks so I can refer back to them though.
 
Isn't this a problem caused by a Bios glitch?

Isn't this a case where the range was allocated because it was recognized as usable, but in fact could not be used?

There was a patch that ignored the Bios defect and forced it to be usable, but of course it was not reflected in the normal kernel.

Is this different from this?

Don't rant just because you can't find a solution.

https://bugzilla.kernel.org/show_bug.cgi?id=218050
 
Last edited:
  • Like
Reactions: mike the newb and fba
uzumo said:
Isn't this a problem caused by a Bios glitch?

Isn't this a case where the range was allocated because it was recognized as usable, but in fact could not be used?

There was a patch that ignored the Bios defect and forced it to be usable, but of course it was not reflected in the normal kernel.

Is this different from this?

Don't rant just because you can't find a solution.

https://bugzilla.kernel.org/show_bug.cgi?id=218050
Click to expand...

The ixgbe is a different beast than the igb (i350), but interesting find that I hadn't seen yet. And it does relate to memory address space, but I don't think it's the issue here. On this motherboard, as on many, there is configurable memory allocation space set aside for pci. I installed a USB card, which was also addressed space in the same range. It did not overlap the i350, and worked fine. I pulled the i350 out and put the USB card in the same slot (just for grins) and it pulled the same address space the i350 had (since that was the beginning of pci space available), and it worked. So, the bios seems to be doing its job correctly.

As a further test, I pulled both cards, rebooted with them out, shut down, put in the i350, and it pulled the same address space as before. It didn't work and shows identical failures to before.

This shows the onboard NIC and the first NIC of the T4:

Code: 
root@PVE:/# lshw -C network
  *-network:0               
       description: Ethernet interface
       product: I350 Gigabit Network Connection
       vendor: Intel Corporation
       physical id: 0
       bus info: pci@0000:04:00.0
       logical name: eno1
       version: 01
       serial: 00:25:90:7c:2e:ea
       size: 1Gbit/s
       capacity: 1Gbit/s
       width: 32 bits
       clock: 33MHz
       capabilities: pm msi msix pciexpress bus_master cap_list ethernet physical tp 10bt 10bt-fd 100bt 100bt-fd 1000bt-fd autonegotiation
       configuration: autonegotiation=on broadcast=yes driver=igb driverversion=6.14.11-1-pve duplex=full firmware=1.52.0 latency=0 link=yes multicast=yes port=twisted pair speed=1Gbit/s
       resources: irq:33 memory:fb920000-fb93ffff ioport:d020(size=32) memory:fb944000-fb947fff memory:fb948000-fb967fff memory:fb968000-fb987fff
      
      
      
       *-network:0 UNCLAIMED
       description: Ethernet controller
       product: I350 Gigabit Network Connection
       vendor: Intel Corporation
       physical id: 0
       bus info: pci@0000:07:00.0
       version: 01
       width: 32 bits
       clock: 33MHz
       capabilities: pm msi msix pciexpress cap_list
       configuration: latency=0
       resources: memory:30000000-300fffff memory:30600000-30603fff memory:30400000-3047ffff memory:30604000-30623fff memory:30624000-30643fff
 
I have these NICs running for ages and atm. on 6.14.8-2-bpo12-pve

Flash firmware: https://calvin.me/how-to-update-intel-nic-firmware/ https://www.reddit.com/r/homelab/comments/eda5qp/updating_firmware_on_i350_t4/
20.4.1 mentioned there is old, very old. I do not know if firmware for the I350 is still available in the latest package, I think it went EOL some versions back from v30, but I don't know for sure. https://www.intel.com/content/www/u...t-utility-preboot-images-and-efi-drivers.html
Dig through the packages or readme if there is support/firmware. Choose then the latest for flashing and the "combo"-option. That means flashing parts for legacy machines and newer ones with UEFI. You need to flash every single port, so 4x.
With compatible UEFI-initialization your bar errors should be gone, regardless of the previous behaviour with older kernels.
If this doesn't help after flashing, try to switch off rebar/resizeable bar in BIOS or boot in CSM/legacy mode.
 
  • Like
Reactions: mike the newb
mr44er said:
I have these NICs running for ages and atm. on 6.14.8-2-bpo12-pve

Flash firmware: https://calvin.me/how-to-update-intel-nic-firmware/ https://www.reddit.com/r/homelab/comments/eda5qp/updating_firmware_on_i350_t4/
20.4.1 mentioned there is old, very old. I do not know if firmware for the I350 is still available in the latest package, I think it went EOL some versions back from v30, but I don't know for sure. https://www.intel.com/content/www/u...t-utility-preboot-images-and-efi-drivers.html
Dig through the packages or readme if there is support/firmware. Choose then the latest for flashing and the "combo"-option. That means flashing parts for legacy machines and newer ones with UEFI. You need to flash every single port, so 4x.
With compatible UEFI-initialization your bar errors should be gone, regardless of the previous behaviour with older kernels.
If this doesn't help after flashing, try to switch off rebar/resizeable bar in BIOS or boot in CSM/legacy mode.
Click to expand...

I'll try to work my way through that but, by way of analogy, you're a rocket scientist and I'm paddling a canoe :)
However, what is still very odd is that the onboard NICs are also i350, and they work fine. I'll post back after I follow your guidance here.
Thank you, BTW.
 
mike the newb said:
I'll try to work my way through that but, by way of analogy, you're a rocket scientist and I'm paddling a canoe :)
However, what is still very odd is that the onboard NICs are also i350, and they work fine. I'll post back after I follow your guidance here.
Thank you, BTW.
Click to expand...

Well, no luck trying to flash. I can't even get the EFI shell to work properly. in my case, this is the built in shell. If I try to map, I can only see half of the output, but the USB drive I'm using comes up as fs2, which I can't do anything with. It's a DOS bootable disk with the necessary files on it. As I said, I'm paddling a canoe.
 
Create a bootable FreeDOS flash drive with https://rufus.ie/en/
Switch your mobo-BIOS to boot legacy/CSM with it. (hammering F12 after start opens up boot override on some mainboards, could also be F9 or F11. if you see your drive 2x, then you choose the non-UEFI entry). If it boots to DOS-prompt, you're golden.

mike the newb said:
It's a DOS bootable disk with the necessary files on it
Click to expand...
Copy these files additionally onto the flash drive, then boot it. Ignore *.efi, flash the binary with the .exe
If it does not work, try it with an old spare PC, best if this doesn't have at all UEFI-Bios...but regardless of this, flash "combo".
 
You must log in or register to reply here.
Top Bottom