Values not getting applied

[Peskedor]

I followed this guide here:
intel/gvt-linux#131 (comment)
I tested this on two different mainboards, and both resulted in the same outcome: the EFI tool shows that the values have been written successfully. After restarting, they remain at whatever I set them to; however, no change was noticed in the OS regarding the Aperture Size.
One of my mainboards supports changing it directly in the BIOS. I tried setting it with setup_var to see if the value changes in the BIOS, but it stays at the default value. Should the BIOS reflect that change? Are there any known issues with certain mainboards where values are "written" but won't get applied?
Currently, I'm facing this issue on an MSI Z370.

[datasone]

Reading after writing (and rebooting) is sufficient for verifying the value is correctly written. For this specific issue, it may be relevant to if the value being written (varstore name and/or offset) corresponds to the correct setting item.

As you are trying with DIY motherboards, they usually don't limit user from accessing settings, and thus won't bother restricting third party tools. And if the correct value are written, the BIOS setting UI would change as well, because it read from the same storage you have written to.

[Peskedor]

Reading after writing (and rebooting) is sufficient for verifying the value is correctly written. For this specific issue, it may be relevant to if the value being written (varstore name and/or offset) corresponds to the correct setting item.

As you are trying with DIY motherboards, they usually don't limit user from accessing settings, and thus won't bother restricting third party tools. And if the correct value are written, the BIOS setting UI would change as well, because it read from the same storage you have written to.

I changed the variables for the iGPU memory, restarted the system, and checked again in the shell that the written values are indeed still what i set them to, but when checking in the BIOS, it will remain on the original values / it won't reflect those changes. When booted into the OS same thing there, still running with the old values.
I downloaded the BIOS from the MSI website, dumped it and extracted the corresponding offsets. Before writing any values, I verified that the current values at those offsets matched what was set in the BIOS UI, which they did. So I'm fairly confident I'm using the correct varstore names and offsets.
Could there be another mechanism that prevents these specific settings from being applied, even though they appear to be written successfully?

[datasone]

Have you checked which varstore it belongs/written to? Later generations of BIOS often split setting values into various varstores, not all into the default one named Setup.

[Peskedor]

I appreciate your help with this!

Steps I took:
Downloaded BIOS from MSI website and extracted Setup module using UEFITool. I searched for "Aperture Size" as Unicode text in UEFITool, which led me to the Setup module. I then extracted the PE32 image section body from the Setup module.
Used ifrextractor to convert the binary to readable format:
./ifrextractor Setup.bin verbose > setup_readable.txt
Found the Aperture Size configuration by searching for "Aperture Size" in the output:
OneOf Prompt: "Aperture Size", VarStoreId: 0x1, VarOffset: 0x7F4
OneOfOption Option: "128MB" Value: 0
OneOfOption Option: "256MB" Value: 1, Default, MfgDefault
OneOfOption Option: "512MB" Value: 3
OneOfOption Option: "1024MB" Value: 7
OneOfOption Option: "2048MB" Value: 15
To find the correct VarStore name, I searched for "VarStoreId: 0x1" and found:
VarStore Guid: EC87D643-EBA4-4BB5-A1E5-3F3E36B20DA9, VarStoreId: 0x1, Size: 0x14DC, Name: "Setup"

I had to specify the ID:
FS1:> setup_var.efi Setup(1):0x7F4
Current value: 0x01 (256MB)
FS1:> setup_var.efi Setup(1):0x7F4=0x7
Successfully set to 0x07 (1024MB)

After rebooting in the BIOS and OS it will stay at 256MB - reading the value again will return 0x07

[datasone]

Yes, this seems right, which is quite strange. Maybe the setting is controlled by another module in the firmware (very unsure on this though)?

BTW, have you tried setup_var.efi Setup:0x7F4=0x7? A little side note is that Setup(1), here the VAR_ID "(1)" is not the VarStoreId in ifr extracted information. It is used for distinguishing different varstores with same name (yes they can have same name). If there are multiple varstores with same name, setup_var.efi will prompt user to try again with the id it generates in output.

However, if the firmware only have one varstore named Setup, this ID will be ignored, so I don't think this is the problem here. As the firmware usually won't have two varstores with same name, and with large enough size (here it is larger than 0x7F4).