fsck.gfs2: reports master/root dinodes as unused and fixes the bitmap
The problem is that the block map is not set at the point of
time when the master and root dinodes are checked. So fsck
sees the discrepancy between the on-disk bitmap (which is
correct) and the in-core blockmap (which hasn't been set
yet for these two system dinodes). So it complains improperly.
This is a simple fix. Long ago, I wrote a function called
resuscitate_metalist whose purpose was to make sure dinodes
get marked "in use" for each system dinode. It's called by
function check_system_inode, and by way of check_metalist.
The function is called for root and master, and therefore
all the system dinodes _within_ those directories are
marked properly in the block map, but nothing ever bothers
to mark the directories themselves. (Side note: the root
directory doesn't have system dinodes like master, so that's
moot).
The solution (this patch) is to set the block map
accordingly, prior to the other checks. Since those two
dinodes have no parent, nothing else is going to do it for
them.
rhbz#642797