Hi,
I'm hoping that I've uncovered a du bug and not a problem with the RAID
patches but....
I'm using du that came with fileutils-3.16 as part of RedHat Linux 5.2.
I upgraded to the 2.2.6 kernel today, along with the latest raidtools and
patches. I've been putting some CD's onto a RAID partition. I was playing
around, checking to see how much space was left and toggling between
issuing "du" and "du -h". After I began copying the eighth CD, I noticed
that "du -h" produced some strange output, such as:
.6M ./MASFTE1_EBS/eei
451M ./MASFTE1_EBS/epdata
6.0M ./MASFTE1_EBS/epwincd
879K ./MASFTE1_EBS/img
[stuff deleted]
-319025686.0 .
Then, at the point where I reached 4 Gigs in that directory, "du -h"
reported results as if they were from zero, so now, the directory, which
contains 4477Mbytes shows up as containing 381Mbytes (4477-4096=381).
du alone reports the correct number of blocks used so thats what leads me
to believe that there is a bug with the human option of du.
Is there anything funky that RAID would be doing to generate these kinds
of reports? In looking through the du.c code, it looks like it just does
a conversion of BLOCK_SIZE to a human readable form and it also looks like
it uses a 64bit value for the largest human readable string, which is
bigger than my 4Gb.
Any ideas?
Kevin
--
~ Kevin M. Myer
. . Network/System Administrator
/V\ ELANCO School District
// \
/( )\
^`~'^
