Re: [PATCH v7 3/3] ARM: dts: renesas: r9a06g032: add EDAC node
From: Paul Louvel
Date: Wed Nov 18 2026 - 04:39:49 EST
Hello,
On Tue Nov 17, 2026 at 10:09 PM CET, Wolfram Sang wrote:
> Hi Paul,
>
>> > On second thought, any reason we don't want this enabled by default?
>>
>> Please see my answer to Geert [1].
>> After reconsideration, maybe it is advisable to enable it by default. Enabling
>> ECC is done by the bootloader, and it might be better if no further actions are
>> required to enable it kernel side.
>> If ECC is disabled, no errors will be generated, and the capabilties will just be
>> set to EDAC_CAP_NONE.
>
> I agree that enabling ECC belongs to the bootloader. With the current DT
> bindings, the Linux driver continues to work on whatever it has been set
> up before (which may be nothing), or? From that point of view, I think
> it makes sense to enable it by default. But again, my knowledge is
> limited here...
The EDAC_CAP_NONE reported in the struct mem_ctl_info structure if ECC is
disabled is just used once in the EDAC subsystem [1] and is purely
informational. If ECC is disabled, no interrupts will be generated in case of an
error and cdns_mc_err_handler().
So there is no risk of any side effects if the driver is enabled and ECC is
disabled by the bootloader.
I will set it to enabled by default on the next iteration.
[1]: https://elixir.bootlin.com/linux/v7.3.1/source/drivers/edac/edac_mc.c#L125
>
> Thanks and happy hacking,
>
> Wolfram
Thanks.
Best regards,
--
Paul Louvel, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com