Hi,
> Le 15 mars 2020 à 14:20, Roger Pueyo Centelles | Guifi.net
> <roger.pu...@guifi.net> a écrit :
>
> Hi,
>
>> I believe this is a waste of resources and a very suboptimal approach. I’m
>> not sure I’m interested in spending time on this :P
> Probably it is. How would you approach it? Some devices that are the same
> hardware with just a different name are already supported like this:
> https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=ac36cca012dd1bbeea0fc4c2dc7a00941de34b52
>
> <https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=ac36cca012dd1bbeea0fc4c2dc7a00941de34b52>
Yes, except in this case the resulting image name isn’t changed and the
difference in naming is very subtle. In the case I quote below, one device is
called RB 911L, the other RB SXT 2nD r3. The average user is never going to
know they’re one and the same :P
That’s why I’d prefer maintaining the one-image for all devices approach, which
has benefits both for the openwrt infrastructure (it scales and consumes less
ressources) and for the users (« you have a mikrotik SPI NOR device? You can’t
get it wrong, the image works on all of those we support »).
Considering routerboot’s lack of support for DTS, I suspect the only way to
tackle this is via an intermediary loader, unless there is a specific mechanism
in the kernel we could use (I’m not aware of any, but I know very little about
the implementation details of DTS).
>> Some devices share the exact same hardware and differ only in (marketing)
>> name, as evidenced by:
>> https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=5ac974f2145c770431a6eb7e006dd086b70224b1
>>
>> <https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=5ac974f2145c770431a6eb7e006dd086b70224b1>
>>
>> (this device uses the 911L platform)
>>
>>> Just have a look at how the few ath79 devices are implemented, but note
>>> that they will be moved to a mikrotik subtarget soon as indicated by Roger
>>> already.
>>
>> I’ve offered in this thread a couple patches to align the ath79
>> implementation on the existing ramips one wrt mtd partitioning and naming.
> To me they're OK, I have no preference for having the partitions nested or
> not. What are the benefits and drawbacks?
>
>
As was once discussed and eventually accepted (when renaming RBMxxG
partitions), this is in line with the canonical way to define partitions in
DTS, as documented in Documentation/devicetree/bindings/mtd/partition.txt
This method is apparently used in all bcm targets, including ath79, ipq and
lantiq. The aforementioned documentation says:
For backwards compatibility partitions as direct subnodes of the flash
device are
supported. This use is discouraged.
Cheers,
Thibaut_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel
