UBIFS does not mount after powerfail

[Richard Weinberger]

Manfred,

Am Dienstag, 5. Dezember 2017, 23:19:17 CET schrieb Richard Weinberger:
> Manfred,
> 
> Am Dienstag, 5. Dezember 2017, 21:06:28 CET schrieb Richard Weinberger:
> > Manfred,
> > 
> > Am Dienstag, 5. Dezember 2017, 20:11:20 CET schrieb Manfred Spraul:
> > > > So, which file do I need to download to get the image that contains
> > > > the
> > > > first error as detected by chk_fs?
> > > > 
> > > > The mode of operation of your tool should be:
> > > > 1. Mount UBIFS
> > > > 2. Do various filesystem operations
> > > > 3. Random power-cut (in software)
> > > > 4. Dump the image to a file X
> > > > 5. Set chk_fs to 1
> > > > 6. Mount it again
> > > > 7. If mount succeeds, drop X, umount and goto 1
> > > > 
> > > >     Else if, chk_fs finds a problem. Store file X.
> > > > 
> > > > I want file X. :-)
> > > 
> > > The tool is in release process, thus let's just wait a bit.
> > > There is no "goto 1" in my flow.
> > > 
> > > Part 1: (done 5x, i.e. with xattr, without xattr, ...)
> > > 
> > > 0. create UBI image, create empty UBIFS image.
> > > 1. Mount UBIFS
> > > 2. Do various filesystem operations.
> > > 
> > >     In the background, record nand operations and store it in a trace.
> > > 
> > > Part 2: (done ~100.000 times)
> > > 1. Create an image file X by replaying from the trace.
> > > 2. Set chk_fs to 1
> > > 3. Mount the image
> > > 7. If mount succeeds: ignore the file X
> > > 
> > >     If mount fails: remember file X
> > > 
> > > Thus every image is a first error. No image has seen multiple unclean
> > > unmounts.
> > 
> > Hmm, when does the power-cut occur in your setup?
> > i.e image-168167.bin and image-168168.bin show different errors and none
> > of
> > them fails the chf_fs test. They fail because of other reasons.
> 
> Short update, the issue in image-168168.bin seems to be a garbage collector
> bug which is not at all related to xattrs.
> Are you 100% sure that all your tests run fine when xattrs are not used?

Another question. Do you happen to have the full flight record file of that 
run? Such that I can replay on my own.

As it looks like we have a problem with the LEB property tree.
But no idea so far what the root cause is.

Thanks,
//richard