On 24-04-18 19:17, Stijn Tintel wrote: > On 24-04-18 17:59, Christian Lamparter wrote: >> On Dienstag, 24. April 2018 13:29:17 CEST Stijn Tintel wrote: >>> On 27-03-18 20:42, Christian Lamparter wrote: >>>> This patch adds a image with squashfs as the root filesystem. >>>> A rootfs_data partition will be generated on the first boot >>>> and placed inside the rootfs partition (just after the squashfs >>>> image). >>>> >>>> advantages: >>>> - it is possible to migrate from an existing -ext4 >>>> installation and back via sysupgrade. >>>> - existing partition layout will not be lost. >>>> - slightly smaller image size. >>>> - support for attendedsysupgrade >>>> >>>> disadvantages: >>>> - needs f2fs + tools as well. This is because fs-tools decides on the >>>> blocksize of the sdcard. So either f2fs or ext4 can get choosen as >>>> the rootfs_data filesystem (depends on the size of the root partition). >>>> - rootfs_data is placed into the rootfs partition. This makes >>>> it difficult for tools that expect a /dev/mmc0pX device. >>>> It also makes it difficult for data recovery tools since they >>>> might not expect to find a embedded partition or will be >>>> confused. >>>> >>>> For people with existing build configurations: make sure to include mkf2fs >>>> and f2fsck package into the image... Otherwise the new -squashfs image will >>>> boot of a ram-overlay and won't keep the configurations after a reboot. >>>> >>> Tried to sysupgrade one of my RPi0W from ext4 to squashfs, and it >>> appeared to work at first, until I rebooted it again. Apparently the >>> overlay is completely missing, the SD card only has /boot (FAT32) and / >>> (squashfs). Can you please look into this? >> I think know the problem. The commit-message mentions it and the symptoms >> you are describing would fit. But just in case: can you please provide the >> image's diffconfig (or full-config ... and I guess a bootlog wouldn't >> hurt either - but it would be optional) to confirm? > Please find diffconfig attached. The machine was running headless so I > don't have a bootlog. >> (I also tested it again on an uptodate "OpenWrt SNAPSHOT, r6759+2-9e848f901c" >> with my rpi-b - which should be in the same family/subtarget as your RPi0W so >> I can try your generated images as well.) >> >>> If you can get it resolved I will merge the patch. >> The original issue [0] has been addressed by this patch [1]. And indeed >> the luci-app will now successfully generate a image (currently, only for the >> rpi-3 though!): >> "openwrt-dbbb09d2658d5c2-brcm2708-bcm2710-rpi-3-ext4-sdcard.img.gz" >> and install it. >> >> this patch is therefore no longer necessary since it the existing ext4 >> images work as well. :) ... Plus the ext4 images don't confuse tools and >> brains as much. > I still find it useful, I've had way too many corrupted ext4 filesystems > on my RPis, and I'm hoping squashfs + f2fs would be more reliable than > ext4 on SD cards... Read the commit message again, and I see what you mean. Must have misread CONFIG_DEFAULT_mkf2fs=y as CONFIG_PACKAGE_mkf2fs=y. Too bad make defconfig doesn't enable DEFAULT_PACKAGES. Sorry for the noise, I'll do some more testing and accept this patch.
Thanks, Stijn _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
