Comment # 9 on bug 1222429 from Wenruo Qu

Thanks for the info on Bug 1195894.

It really explains more than what I need.

So "norecovery" is to prevent log/journal replay, and btrfs' equivalent is
"rescue=nologreplay".

Personally speaking, let the installer just do a "ro" mount would be more than
enough. Unless the fs is already corrupted and log/journal replay could fail.
And even for a log replay failure case, unless it's something only fixed in
newer kernels, it doesn't make much difference.

That's why most fses replay their logs even in RO mount, because that's needed
to recover the not yet committed data.
And I do not think who is replaying the log (either the installer, or the
kernel of that partition on next boot) does any difference.

As a quick fix, just moving btrfs out of "NORECOVERY_FS" would be sufficient to
me.