Hi Moritz,
On Thu, 26 May 2016 10:28:48 -0700
Moritz Fischer <moritz.fischer@xxxxxxxxx> wrote:
> Hi Boris,
>
> On Thu, May 26, 2016 at 1:04 AM, Boris Brezillon
> <boris.brezillon@xxxxxxxxxxxxxxxxxx> wrote:
>
> > I think the MTD partition -> nvmem connection could benefit to non-OTP
> > partitions too.
>
> Yeah, I thought about that, too. Would you use the _read, and _write
> callbacks in that case?
Yep, this means you'll need a flag to decide what kind of partition is
exposed (OTP or normal), and then adjust the partition
->_read()/->_write() implementations.
Actually, I'm not sure what will happen if you really expose OTP
partitions, since these ones shouldn't implement norma
->_read()/_write(), or they should wrap those calls around their
master ->_read/write_prot_xx().
I must admit I haven't looked at patch 2 yet, so maybe that's what
you're doing.
>
> > So, how about defining the nvmem regions under the partition nodes,
> > like that:
> >
> > flash@0 {
> > partitions {
> > compatible = "fixed-partitions";
> > #address-cells = <1>;
> > #size-cells = <1>;
> >
> > partition@0 {
> > label = "uboot-spl";
> > reg = <0x0 0xe0000>;
> > };
> >
> > /* ... */
> >
> > partition@X{
> > label = "factory-data-part";
> > reg = <0x200000 0x100000>;
> > #address-cells = <1>;
> > #size-cells = <1>;
> >
> > product: nvmem@0 {
> > reg = <0x0 0x2>;
> > };
> >
> > revision: nvmem@3 {
> > reg = <0x3 0x2>;
> > };
> > };
> > };
> >
> > otp-partitions {
> > compatible = "fixed-partitions";
> > #address-cells = <1>;
> > #size-cells = <1>;
> >
> > partition@X{
> > label = "factory-data-part";
> > reg = <0x0 0x40>;
> > #address-cells = <1>;
> > #size-cells = <1>;
> >
> > product: nvmem@0 {
> > reg = <0x0 0x2>;
> > };
> >
> > revision: nvmem@3 {
> > reg = <0x3 0x2>;
> > };
> > };
> > };
> > };
> >
> > I know this requires changing the implementation to select the
> > appropriate nvmem wrapper to use depending on whether we're interfacing
> > with an OTP area or a regular one, but that should be doable.
>
> The implementation still needs work anyways, so I might as well add
> this to my list ...
> Would you do the nvmem mapping always, or conditionalize on a flag in
> the dt node like 'nvmem-export'?
Well, I asked myself the same question when I proposed the binding. You
actually don't need the property since you can detect the presence of
the #address/size-cells properties or the presence of child nodes, but
maybe it's safer to make it dependent on an extra property (or use
compatible = "mtd-to-nvmem").
I honestly don't know what's the best solution. Rob, any suggestion?
Regards,
Boris
--
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
