Hi, * drEagle (drea...@doukki.net) wrote: > I have started to add sysupgrade to guruplug server in kirkwood platform. > I get stuck in the understanding in my analysis of OpenWrt implementation of > Ugrade process. > > So, > On the GuruPlugServerPlus are Kirkwood CPU. > The available NAND (512Mo) is ok with UBI. > > What are best solutions of MTD partitionning for all OpenWrt longtimes use > cases ? > For factory reset, upgrade, fault, others ?
It seems that most NAND platforms are using UBI with squashfs + a R/W UBIFS overlay. I think this is a good format because you retain a "good" base image in /rom, should you wish to reset settings and packages. > What are differences between squashfs, ubifs, ubi, ubinized ? AFAIK: squashfs is a RO filesystem not suitable to writing directly on NAND (doesn't handle bad blocks, etc.) UBI is a container format that takes care of things like bad blocks, wear leveling, etc. As far as I understand it, ubinizing is encapsulating squashfs (or another image) such that it can live inside an UBI block volume. UBIFS is another filesystem running on top of UBI. > What is the most simple implementation of sysupgrade ? > I cannot get it to work with ubifs, the root is blocking my ubidetach try. > If I get all unmount with /etc/init/umount stop, then my /mnt and images are > no more available. I'm sitting on some patches that implement the aforementioned squashfs+ubifs format on Kirkwood, rather than the straight UBIFS image it is now. I'll try to clean them up in the next few days and submit. I don't remember the details, but even before switching to squashfs it was mostly a matter of creating a platform_check_image() block in /lib/upgrade/platform.sh, and calling "nand_do_platform_check" for the board. It "just worked" after that. -Claudio _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
