Here's part of an ongoing discussion on icb.  Bill Studenmund suggested
I fsck the disk.  It found some problems:

<wrstuden> My thought is that fsck will be indicating whatever write didn't
+make it out.
<wrstuden> Also, we may need to bug Darrin.
<wrstuden> Or fvdl
<dyoung> ok, i just created /permanent/etc/guysmiley, rebooting...
<dyoung> is fsck -fy what you want?
<wrstuden> Probably.
<dyoung> i just got luke's joke.
<perry> We need an fvck and an fvdl command.
<dyoung> ok, guy, guysmiely both 0.  fsck'ing...
<rpaulo> perry: yeah, since we have wizd :)
<wrstuden> Problem is that I'm not going to be 100% on what's wrong with what
+fsck will fix. :-|
<dyoung> ** Phase 5 - Check Cyl groups
<dyoung> FREE BLK COUNT(S) WRONG IN SUPERBLK
<dyoung> SALVAGE? yes
<dyoung> SUMMARY INFORMATION BAD
<dyoung> SALVAGE? yes
<dyoung> BLK(S) MISSING IN BIT MAPS
<dyoung> SALVAGE? yes
<wrstuden> Please log this.
<wrstuden> So it basically seems it allocated the blocks but didn't update the
+inode.
<uep> that sounds consistent with earlier speculation, good to confirm

Dave

-- 
David Young             OJC Technologies
dyoung@ojctech.com      Urbana, IL * (217) 278-3933