[Intel-wired-lan] is there a way other than PCI IDs to distinguish E810 from E825/E830?

[Intel-wired-lan] is there a way other than PCI IDs to distinguish E810 from E825/E830?

[Chris Friesen via Intel-wired-lan]

[-- Attachment #1: Type: text/plain, Size: 431 bytes --]

Hi,

I've got an odd issue.  We've got a request to use the in-tree ice 
driver for the "legacy" NICs like the E810, and the out-of-tree ice 
driver for the newer NICs associated with the Granite Rapids-D (E825/E830).

Is there any way to distinguish between these other than the PCI device 
IDs?  I'd rather not need to maintain a list of devices and need to 
update them every time a new NIC variant comes out.

Thanks,

Chris

[-- Attachment #2: Type: text/html, Size: 669 bytes --]

Re: [Intel-wired-lan] is there a way other than PCI IDs to distinguish E810 from E825/E830?

[Przemek Kitszel]

On 3/9/26 19:37, Chris Friesen wrote:
> Hi,
> 
> I've got an odd issue.  We've got a request to use the in-tree ice 
> driver for the "legacy" NICs like the E810, and the out-of-tree ice 
> driver for the newer NICs associated with the Granite Rapids-D (E825/E830).
> 
> Is there any way to distinguish between these other than the PCI device 
> IDs?  I'd rather not need to maintain a list of devices and need to 
> update them every time a new NIC variant comes out.
> 
> Thanks,
> 
> Chris
> 

you are asking this question on the public mailing list about in-tree
drivers, here recommendation is to use in-tree driver, always

as "also OOT" developer, I would say it it sometime warranted to use
OOT, but then don't mix and either stick to all OOT or all upstream

Re: [Intel-wired-lan] is there a way other than PCI IDs to distinguish E810 from E825/E830?

[Dawid Osuchowski]

On 2026-03-09 7:37 PM, Chris Friesen via Intel-wired-lan wrote:
> Hi,

Hello Chris,

Thank you for your message. Given I'm from the more customer facing wing 
I'll try to answer / guide you towards a path satisfactory for you. As 
Przemek pointed out this is a mailing list for the upstream / in-tree 
development side of things.

> I've got an odd issue.  We've got a request to use the in-tree ice 
> driver for the "legacy" NICs like the E810, and the out-of-tree ice 
> driver for the newer NICs associated with the Granite Rapids-D (E825/E830).

As Przemek noted, our general recommendation is not to mix drivers 
within the same system. That would mean using all in-tree drivers or all 
OOT drivers.

The following scenarios:

	1. all in-tree drivers (in-tree ice, in-tree iavf, in-tree irdma),
	2. all OOT drivers (OOT ice, OOT iavf, OOT irdma).

are supported and validated.

A configuration like (and any combination similar to it):
	* OOT ice, in-tree iavf, in-tree irdma

is not validated by Intel and certainly not recommended or supported. It 
might work, but we are unable to help in case you run into issues using 
such a setup.

> Is there any way to distinguish between these other than the PCI device 
> IDs?  I'd rather not need to maintain a list of devices and need to 
> update them every time a new NIC variant comes out.

I think I know what you are trying to do here, please correct me if I'm 
wrong.

You would like to pin a specific device like e.g. E810 to the in-tree 
driver and the E825/E830 to the OOT driver.

If that's the case, then I think the PCI device ID is the only option 
I'm afraid.

One question that popped into my mind is: what happens if e.g. you have 
both an E810 and an E825 present in the system? You cannot load two ice 
drivers (one in-tree, another OOT) at the same time.

> Thanks,
> 
> Chris

Side-note: Be aware that customer support folk do not usually monitor 
this mailing list (myself being an exception rather than the rule). If 
you want / need prompt updates and something you can share with your 
team / manager for tracking, please open a support thread using the 
established support processes (e.g. IPS [1]).

Best regards
Dawid

[1] 
https://www.intel.com/content/www/us/en/support/articles/000057045/ethernet-products.html