Fun with df

I've seen this problem on both RELENG_6 from Mar 27 19:37 UTC and HEAD:

                                                               vnodes

      Filesystem   Size  Used Avail Capacity  Mounted on   num  want free
16:51 /dev/ad0s1e  1.9G  -68G   70G -3900%    /tmp         480 17495 132
16:52 /dev/ad0s1e  1.9G  -68G   70G -3900%    /tmp         479 17495 132
16:53 /dev/ad0s1e  1.9G  -68G   70G -3900%    /tmp         479 17495 132
16:54 /dev/ad0s1e  1.9G  -68G   70G -3900%    /tmp         17885 17495 17495
16:55 /dev/ad0s1e  1.9G  -68G   70G -3899%    /tmp         17875 17495 17496
16:56 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         17855 17495 17496
16:57 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         17856 17495 5594
16:58 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         356 17495 5
16:59 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         356 17495 5
17:00 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         364 17495 13
17:01 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         364 17495 13
17:02 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         364 17495 13
17:03 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         370 17495 18
17:04 /dev/ad0s1e  1.9G  -70G   72G -4013%    /tmp         380 17495 20
17:05 /dev/ad0s1e  1.9G  226M  1.5G    13%    /tmp         393 17495 21

This is what "fixed" the problem (the fsck is not really needed):
:
$ umount -f /tmp; fsck -y /tmp; mount /tmp
** /dev/ad0s1e
** Last Mounted on /tmp
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
26939 files, 115910 used, 875489 free (633 frags, 109357 blocks, 0.1% fragmentation)

A test with "cd /tmp; umount /tmp) did not seem to have any effect. 

-- 
Peter Holm