Re: [Intel-wired-lan] [PATCH] e1000e: disregard NVM checksum on tgp when valid checksum mask is not set

[Lifshits, Vitaly]

On 5/15/2025 10:07 PM, Vlad URSU wrote:
> On 15.05.2025 07:39, Lifshits, Vitaly wrote:
> > Since the checksum word is 0xFFFF which is peculiar, can you dump the 
> > whole NVM and share with us?
>
> Sure, here's a dump of my NVM
>
> Offset        Values
> ------        ------
> 0x0000:        d0 8e 79 07 78 c8 01 08 ff ff 44 00 01 00 6c 00
> 0x0010:        ff ff ff ff c9 10 54 0a 28 10 f9 15 00 00 00 00
> 0x0020:        00 00 00 00 00 80 05 a7 30 30 00 16 00 00 00 0c
> 0x0030:        f3 08 00 0a 43 08 13 01 f9 15 ad ba f9 15 fa 15
> 0x0040:        ad ba f9 15 ad ba f9 15 00 00 80 80 00 4e 86 08

You're right — I see that the SW compatibility bit is set and the 
checksum appears to be incorrect.

Since the NVM is part of the system firmware and typically managed by 
the system manufacturer, I recommend checking whether a firmware update 
is available for your system as a first step.

If no update is available, perhaps we can consider ignoring the checksum 
on TGP systems if one of the following conditions is met:
1. SW compatibility bit is not set (current Jacek's approach)
2. The checksum word at offset 0x3F retains its factory default value of 
0xFFFF.