On Thu, Apr 10, 2003 at 03:01:36PM +0200, Jaromir Dolecek wrote:
> According to directory contents, 'my' file used inode #196.  The
> directory entry was apparently correct, including correct file type
> (DT_REG).  I'm attaching the directory dump (cat /) to this mail,
> it's file /.profile.

Thanks for looking into it; I can't reproduce this one myself so far.
I even read through a context diff of my entire UFS2 commit, but
I haven't found anything yet.

The directory itself won't reveal many problems. The issue is
with the bitmaps and/or inode blocks, I think. It's like the
kernel somehow selects the wrong inode block for some inodes
to read from. This causes a bad vnode type. To catch this
early, I'll add a DIAGNOSTIC check to ufs_vinit shortly.

'dup alloc' is the same problem. It reads a wrong dinode from
the disk; it should have a 0 mode field, but doesn't, so it
thinks it was already allocated.

- Frank