On Sat, 4 Feb 2023 at 22:09, Jassi Brar <jaswinder.si...@linaro.org> wrote: > > > > + fwu-mdata { > > + compatible = "u-boot,fwu-mdata-mtd"; > > + fwu-mdata-store = <&spi_flash>; > > + mdata-offsets = <0x500000 0x530000>; > > + }; > > > > which is based on DT going to location which is already labelled. > > > > 79 partition@500000 { > > 80 label = "Ex-OPTEE"; > > 81 reg = <0x500000 0x200000>; > > 82 }; > > > The 'ex-optee' is actually unused so we never faced any issue. But > yes, this is inconsistent. > > > I don't know what this space is used for but the whole code around is using > > MTD > > partitions and it's infrastructure and this is using RAW access without MTD. > > > > Why not to create separate partitions just for storing metadata? > > And also identify them like that. > > > > Or just switch it to complete RAW mode without MTD and then offsets can be > > used > > (but I expect with different dt description). > > > The design predates my involvement in fwu. I guess the reason was to > have a mechanism similar to GPT based implementation which uses a > similar fwu-mdata {} node structure. It does make sense to have > dedicated partitions for primary and secondary meta-data, identified > by uuid (like Banks) or standard labels. But may be Sughosh/Ilias know > why the current approach was chosen. > I changed the bindings to not require any changes to any non-FWU subsystem. Now every bit of info is contained in 'fwu-mdata' node. https://lore.kernel.org/u-boot/20230228005218.1635781-1-jassisinghb...@gmail.com/T/#u
Thanks.