According to http://linux-sunxi.org/H2%2B#VariantsI should be able to differentiate between the H2+ and the H3 via the CPU SID.  I have an Orange Pi Zero board with this output from /proc/cpuinfo

 

Hardware        : Allwinner sun8i Family
Revision        : 0000
Serial          : 02004620161dea8a
 

According to that site "the last byte of the first word" should be either 0x42 or 0x83 since the Orange Pi Zero has an H2+.  Yet I am not finding 0x42 or 0x83 *anywhere* in "02004620161dea8a".

 

Anyone know how to correctly identify the CPU variant?

Serial is not same as SID. SID is special storage for one time programmable data (OTP) also known as eFuse (you can change it too). Many different things can be stored there, calibration factors among the others... I'm not sure if this data is exposed via procfs or sysfs.

SID != serial number displayed in /proc/cpuinfo

 

For getting SID in the legacy kernel you need to use devmem utility, please check the details here: https://linux-sunxi.org/SID_Register_Guide#Currently_known_SID.27s

SID != serial number displayed in /proc/cpuinfo

 

For getting SID in the legacy kernel you need to use devmem utility, please check the details here: https://linux-sunxi.org/SID_Register_Guide#Currently_known_SID.27s

Well, the first address read by devmem2 matches the output from /proc/cpuinfo:

 

root@orangepizero:~# devmem2 0x01c14200 w|grep Value|sed 's/^.*: 0x/ /'

2004620

root@orangepizero:~# cat /proc/cpuinfo | grep Serial

Serial          : 02004620161dea8a

 

Of course, that's where it ends.  Here's the next region:

 

root@orangepizero:~# devmem2 0x01c14204 w|grep Value|sed 's/^.*: 0x/ /'

3435C614

 

Regardless, the "last byte of the first word" still doesn't match.  So I'm guessing that sun8i.c is just plain wrong since 0x42 and 0x83 can't be found in any of these numbers.

On that wiki page you can see that 02004620 was already pulled from some other board, but it was the wrong value. Please look here for the sunxi-tools fix and some background info: https://github.com/linux-sunxi/sunxi-tools/pull/90

Unfortunately the legacy kernel will need to be fixed too and I'm not sure anybody will want to do this.