Comment 25 for bug 1559723

As described in my previous comment, the fs was available in read-write mode again.

I added another disk to the btrfs and with the additional space I was able to rerun and successfully finish the btrfs balance operation (as described in https://www.slicewise.net/debian/balancierung-eines-vollen-btrfs-dateisystems/ ):

btrfs balance start x
Done, had to relocate 2797 out of 2797 chunks

So my file system is completely rescued now.

To sum it up:

There is no way I could reproduce this error again with a recent kernel.

All I can say is that Kernel 4.2.0-27.32~14.04.1-generic 4.2.8-ckt1 ran into unexpected behaviour when doing a btrfs balance:

Expected Result: Error "enospc errors during balance" (at least)

Actual Result: Kernel bug in dmesg, Filesystem goes read-only, Input/Output errors occur, many directories are not readable even in read-only mode, even ls does not work. It seems as if the system totally screwed it up.

Suggested Fix: All of these problems should be avoidable because the later reboot "healed" them (Error "enospc errors during balance" reported, fs mounted read-write, but btrfs balance still fails, which is totally fine at this point)