08:37 <linusw__> arnd: are you pursuing https://lore.kernel.org/netdev/19ca470e-8219-5ba9-3de6-f4560278f87b@linaro.org/T/

08:38 <linusw__> arnd: asking since I am cleaning a bit of wireless drivers from GPIO numbers, if this will be fixed I do not need to look into it.

08:42 <arnd> linusw__: I should just send what I have here, I think I addressed the comments a while ago and never sent out the next version, but need to look again to be suire

08:42 <linusw__> arnd: thanks :) the series look very nice, you can add my Reviewed-by on anything I failed to ACK.

11:01 <ardb> robher: so do you have any suggestions on how to deal with this issue? as i see it, the principle that a single DT is used for all boot phases simply does not apply to DT being used to hand over state from one boot phase to the next

11:02 <ardb> if the early boot code wants to describe some UEFI related properties to a UEFI firmware payload that runs next, it should do so in a way that is not observeable by the OS, and so those properties should not be in global DT namespace

11:02 <ardb> IMHO

13:47 <robher> ardb: Things that don't apply to the next stage should just be removed.

14:26 <arnd> mraynal: I ran into a problem with your commit 0088cbc19276 ("nvmem: core: Expose cells through sysfs") on a Mac Studio, which fails to register the nvmem device after trying to create duplicate sysfs files

14:27 <arnd> the problem seems to be "attrs[i].attr.name = devm_kasprintf(&nvmem->dev, GFP_KERNEL, %s@%x", entry->name, entry->offset);"

14:27 <arnd> where I have multiple cells with the same name and offset

14:29 <arnd> https://github.com/AsahiLinux/linux/blob/bd0a1a7d465fcb60685a2360565ed424bafff354/arch/arm64/boot/dts/apple/t600x-dieX.dtsi#L156

14:29 <arnd> [ 0.620001] sysfs: cannot create duplicate filename '/devices/platform/soc@2200000000/22922bc000.efuse/apple_efuses_nvmem1/cells/efuse@a10'

14:32 <geertu> arnd: Ah, it should take into account the first value of the bits property, too

14:34 <arnd> geertu: probably. I'm not sure about what the best way to do it is, but I think it needs to be fixed for 6.8 since this introduces an ABI

14:42 <geertu> arnd: Interestingly, this does not match the bindings, which don't have the bits property, not the first bit as part of the unit address

14:42 <geertu> s/not/nor/

14:43 <geertu> Might have been better to merge reg and bits, and use #{address,size}-cells = <2>

14:45 <geertu> Oh, the devm_kasprintf() is in generic nvmem code

15:00 <arnd> geertu: I did notice that the apple efuse driver uses the deprecated dt binding, and the DT changes are not actually upstream yet, so it's possible that changing the driver would still work, but I'm not actually sure it would result in a different sysfs representation if it used the new binding

15:02 <arnd> ie. Documentation/devicetree/bindings/nvmem/nvmem-deprecated-cells.yaml vs Documentation/devicetree/bindings/nvmem/layouts/fixed-cell.yaml

22:45 <palmer> I don't really understand what's going on with <https://lore.kernel.org/all/be959a4bb660466faba5ade7976485c8@AcuMS.aculab.com/>, but LMK if I should do something

23:39 <palmer> never mind