xfs full when not

[Curtis Doty]

I had a lone 795G xfs "fill up" prematurely at 89% blocks and 28% inodes. 
It was built/running under 32-bit OS, and I rebuilt OS 64-bit but the 
problem persisted.

Just for fun, I moved off and back on about 50K files with old ctimes from 
early in the fs life. This actually kinda worked. But I was only able to 
populate and addtional 10K new files before it hit the roof again.

An xfs_repair produced no change in behavior.

meta-data=/dev/cciss/c1d0p3      isize=2048   agcount=33, agsize=6509200 blks
          =                       sectsz=512   attr=1
data     =                       bsize=4096   blocks=208294773, imaxpct=50
          =                       sunit=16     swidth=16 blks, unwritten=1
naming   =version 2              bsize=4096
log      =internal               bsize=4096   blocks=8192, version=1
          =                       sectsz=512   sunit=0 blks, lazy-count=0
realtime =none                   extsz=65536  blocks=0, rtextents=0

As you can see, it was setup with 2K inodes. The imaxpct was originally 
at 25% but I raised it and the problem still occurs at 50%.

I can't even get it to reproduce the behavior Eric reported previously: 
http://oss.sgi.com/archives/xfs/2005-02/msg00325.html

../C