Hey Paul,
On 16-03-15 23:09, Paul Kocialkowski wrote:
Le lundi 16 mars 2015 à 09:12 +0100, Hans de Goede a écrit :
Hi,
On 15-03-15 22:38, Paul Kocialkowski wrote:
As far as I know, there is currently no mechanism in place to give each
device a unique serial number, that would be shown in
e.g. /proc/cpuinfo. This can seem not very useful, but it would be
nice to have that for USB gadget drivers that make use of it (e.g. ADB on
Android, fastboot, etc). This would allow having multiple sunxi devices
connected at the same time.
Normally, the serial number has to be figured out by the bootloader and
passed to the kernel through ATAGs. In U-Boot, it is defined using
get_board_serial, that is not yet implemented on sunxi.
I'm interested in figuring out the best way to do this. As far as I can
see, Allwinner SoCs have some e-fuses, known as SID (Security ID):
http://linux-sunxi.org/SID_Register_Guide for which Oliver has written a
mainline Linux driver. Those should have device-specific values, except
when they are set to 0.
Do you have any better idea to do this, or is there some reference that
I am missing?
AFAIK the only way to do a serial number on sunxi devices is using the SID,
note that Oliver's driver will only work on A10, A13 and A20, with the
A31, A23 and others the SID is burned into the pmic rather then the SoC
itself, but it should be easy to implement get_board_serial() in u-boot
as we already have a sunxi_get_sid() helper to do that in a SoC
independent way (this is currently used to set the ethernet MAC address).
Note that sunxi_get_sid() can fail, so you need to error check it,
and currently it will happily return an all 0 SID, you should probably
make it return an error in that case, as that check really belongs
inside sunxi_get_sid().
Why exactly is it a problem if the SID is 0 here? I don't see a big
difference between providing a zero serial and not providing any serial
at all (in any case, Linux will keep it at 0 when none is provided).
U-Boot doesn't check the return code of setup_serial_tag either and
isn't designed to not include the tag if it somewhat failed to grab the
serial number.
Hence my point is, let's not bother whether it's zero or not: it will
just improve the situation on the vast majority of devices that have (at
least) the first 4 32bit words of the SID populated and leave others
with a zero serial, as it was until now.
Yeah I agree with you there, To bad we don't know the meaning of the SID
fields, so we can determine which bytes are more important than others ;)
--
You received this message because you are subscribed to the Google Groups
"linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
