Colibri T20 boot failure on power outage

We are using your Colibri T20 modul on top of the IRIS carrierboard as part of our test instruments. We need to provide a robust system, in case of power outages and emergency shutdowns. Recently we’ve switched to the Linux Image provided on your website, specifically version 2.6.1, and we’ve experienced boot failures due to some forced “hard resets” of the device, which is causes by the power supply being switched off. Do you have any suggestions on some specific changes that we can make to the Linux Image to avoid this, besides making the filesystem readonly, or suggestions to Uninterruptable Power Supply systems that are comptabile with your modules? In the attachment you can find the whole boot log, but we guess the most interesting lines are:

[    6.452639] UBIFS: background thread "ubifs_bgt0_2" started, PID 78
[    6.533247] UBIFS: recovery needed
[    6.771518] UBIFS error (pid 1): ubifs_recover_leb: corruption -3
[    6.785721] UBIFS error (pid 1): ubifs_scanned_corruption: corruption at LEB 954:3600
[    6.801680] UBIFS error (pid 1): ubifs_scanned_corruption: first 8192 bytes from LEB 954:3600
[    6.833808] UBIFS error (pid 1): ubifs_recover_leb: LEB 954 scanning failed
[    6.849641] UBIFS: background thread "ubifs_bgt0_2" stops
[    6.922412] VFS: Cannot open root device "ubi0:rootfs" or unknown-block(0,0)
[    6.939646] Please append a correct "root=" boot option; here are the available partitions:
[    6.958627] 1f00            2048 mtdblock0  (driver?)
[    6.974350] 1f01            1024 mtdblock1  (driver?)
[    6.989917] 1f02            1024 mtdblock2  (driver?)
[    7.005421] 1f03         1044480 mtdblock3  (driver?)
[    7.020661] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)][1]

Any comments on this would be more than welcome. Also, we’ve been using your Windows CE images for quite some time, and this issues do not arise.

I’m not sure why exactly one would pick an older long since unsupported V2.6.1b1 beta version. That said what you describe is a known issue as noted here and we recommend sticking to the latest stable version V2.3 for the time being.

Thank you for the answer. We’ll switch to the V2.3 stable version.