Re: [RESEND RFC 2/3] nvmem: Add 'nvmem-blob' driver

From: Maxime Ripard
Date: Tue Mar 08 2016 - 17:28:35 EST


On Mon, Mar 07, 2016 at 08:07:41PM -0800, Andrey Smirnov wrote:
> >> >
> >> > Patch itself looks simple. Before we review it further could you provide
> >> > more details on the exact usecase or some background of this.
> >>
> >> The discussion on this topic originated on mailing list of Barebox
> >> project(which borrows very heavily from Linux designs). Barebox
> >> operates on two device tree blobs, one is used for its internal
> >> initialization, whereas second one is passed to Linux kernel when
> >> booting it. The problem I was trying to solve was to make possible to
> >> specify in the first DT blob what data would be used for MAC address
> >> fixup of the second DT blob(the one passed to Linux).
> >>
> >> My first approach was to implement a very limited DT code, however in
> >> discussing it the consensus was that porting 'nvmem' subsystem from
> >> the kernel and using for the same purpose would be a better approach.
> >> First pass adoption of that subsystem revealed that there were two
> >> use-cases that current design didn't allow us to handle:
> >>
> >> - Depending on the version i.MX SoC MAC address data stored in ROM
> >> would have different layout so as a possible solution to that I
> >> implemented "composite" driver(patch #3)
> >>
> >> - On i.MX28, part of the MAC address is hard-coded in
> >> arch/arm/mach-mxs/mach-mxs.c and only a portion of it is read from
> >> ROM, this patch in combination with the aforementioned one should
> >> allow us to encode all needed info in DT.
> >>
> >> Ideally, since all of the above is as applicable to Linux as it is to
> >> Barebox it would be good for BB not to invent its own custom 'nvmem'
> >> flavor, so hence me trying to start a conversation about adding this
> >> upstream.
> >
> > If the only use-case is to store the MAC address, why not using the
> > local-mac-address property directly?
> I don't think I understand what you mean, could you give me an example
> of how I'd use local-mac-address property for that use case? AFAIK,
> local-mac-address is just an array of bytes embedded into device tree,

Well, yeah, but the nvmem-blob is also just an array of bytes embedded
into the DT, right?

> how would it get populated with data from OTP memory of SoC?

In the bootloader, or Linux, read the OTP, patch the DT to add that
node, done.

It's already what U-Boot does here for example (the source of ethaddr
being dependent of the platform):;a=blob;f=common/fdt_support.c#l468


Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering

Attachment: signature.asc
Description: Digital signature