>Number:         8177
>Category:       install
>Synopsis:       mkfs can make a file system that passes fsck, but df reports bogus info
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    install-manager (NetBSD system installation bug manager)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Aug  8 13:05:00 1999
>Last-Modified:
>Originator:     Nathan Raymond
>Organization:
>Release:        1.4 mac68k
>Environment:
NetBSD speedbump 1.4 NetBSD 1.4 (GENERIC) #0: Sat May  8 20:54:59 PDT 1999     root2@c610:/usr/src/sys/arch/mac68k/compile/GENERIC mac68k
>Description:
I formatted a 2GB partition on a 3GB drive successfully with Mkfs 1.45 and noticed that df reported the empty partition was 37% full (about the size of the 1GB partition!), yet it was an empty partition:

Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
/dev/sd1a     3038071  1014244  1720019    37%    /pub

The output of "du /pub":
8       /pub/lost+found
9       /pub

Unmounting the drive and running fsck -f returned no errors.  After rebooting it was still there.  Someone recommended changing the fslevel, so I changed it from a level 1 to a level 2 with:

fsck -T ffs:"-c 2" /pub

Remounted, problem still persisted.  Then changed it to a level 3, still there.  Finally gave up, and since it was an empty partition, ran newfs on it in netbsd.  The output of df is now accurate:

Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
/dev/sd1a     2023869        1  1922674     0%    /pub

Here is the top output of dumpfs on the Mkfs formatted fslevel 3 file system and the newfs formatted file system, for comaprison:

Before with Mkfs and fsck -T ffs:"-c 3" /pub:

Endian big-endian magic 11954   time    Sat Aug  7 09:01:53 1999
cylgrp  dynamic inodes  4.4BSD  fslevel 3
nbfree  252977  ndir    2       nifree  505339  nffree  11
ncg     329     ncyl    5264    size    2092290 blocks  3038071
bsize   8192    shift   13      mask    0xffffe000
fsize   1024    shift   10      mask    0xfffffc00
frag    8       shift   3       fsbtodb 1
cpg     16      bpg     795     fpg     6360    ipg     1536
minfree 10%     optim   space   maxcontig 1     maxbpg  2048
rotdelay 4ms    headswitch 0us  trackseek 0us   rps     60
ntrak   5       nsect   159     npsect  159     spc     795
symlinklen 60   trackskew 0     interleave 1    contigsumsize 0
nindir  2048    inopb   64      nspf    2
sblkno  16      cblkno  24      iblkno  32      dblkno  224
sbsize  3072    cgsize  2048    cgoffset 80     cgmask  0xfffffff8
csaddr  224     cssize  6144    shift   9       mask    0xfffffe00
cgrotor 0       fmod    0       ronly   0       clean   0x01

After newfs:

Endian big-endian magic 11954   time    Sat Aug  7 19:42:45 1999
cylgrp  dynamic inodes  4.4BSD  fslevel 3
nbfree  252982  ndir    1       nifree  510717  nffree  12
ncg     285     ncyl    4549    size    2092290 blocks  2023869
bsize   8192    shift   13      mask    0xffffe000
fsize   1024    shift   10      mask    0xfffffc00
frag    8       shift   3       fsbtodb 1
cpg     16      bpg     920     fpg     7360    ipg     1792
minfree 5%      optim   time    maxcontig 8     maxbpg  2048
rotdelay 0ms    headswitch 0us  trackseek 0us   rps     60
ntrak   5       nsect   184     npsect  184     spc     920
symlinklen 60   trackskew 0     interleave 1    contigsumsize 8
nindir  2048    inopb   64      nspf    2
sblkno  16      cblkno  24      iblkno  32      dblkno  256
sbsize  2048    cgsize  2048    cgoffset 96     cgmask  0xfffffff8
csaddr  256     cssize  5120    shift   9       mask    0xfffffe00
cgrotor 0       fmod    0       ronly   0       clean   0x01
>How-To-Repeat:
Seems to have to do with formatting two partitions on a drive (maybe a large drive?) with Mkfs, though its very strange that the bogus info in one of those partitions is coming up as clean in fsck.  I guess this could be a bug in Mkfs, fsck, and/or the kernel.
>Fix:
No real fix, just a workaround (use newfs if possible).  I have no idea where the real problem lies.
>Audit-Trail:
>Unformatted: