Re: Fwd: Re: df shows wrong size for smbfs/cifs/nfs/davfs mounted directories

[Paul Eggert]

Toralf Förster <address@hidden> writes:

> n22 # df -m /mnt/ramdisk/dav/ /mnt/dav_n22/
> Filesystem           1M-blocks      Used Available Use% Mounted on
> tmpfs                      660         1       660   1% /mnt/ramdisk
> http://n22/address@hidden/
>                           8790         0      8790   0% /mnt/dav_n22
> n22 # strace -e statfs64 df -m /mnt/ramdisk/dav/ /mnt/dav_n22/
> statfs64("/mnt/ramdisk", 84, {f_type=0x1021994, f_bsize=4096, 
> f_blocks=168960, 
> f_bfree=168956, f_bavail=168956, f_files=129449, f_ffree=129441, f_fsid={0, 
> 0}, f_namelen=255, f_frsize=4096}) = 0
> statfs64("/mnt/dav_n22", 84, {f_type="CODA_SUPER_MAGIC", f_bsize=1024, 
> f_blocks=9000000, f_bfree=9000000, f_bavail=9000000, f_files=9000000, 
> f_ffree=9000000, f_fsid={0, 0}, f_namelen=255, f_frsize=1024}) = 0


statfs64 is claiming that /mnt/dav_n22 has 9000000 blocks, each of
size 1024 bytes.  That adds up to 8789.0625 MiB, and so the "8790" is
correct (as df always rounds up) if statfs64 is correct.

You mentioned that the file system actually has only 660 MB, so it
sounds like statfs64 is reporting the wrong numbers.  Hence it's not a
coreutils bug; it's a bug in your kernel or drivers or something like
that.

Most likely the other bugs are similar.

Those "9000000" values all look very suspicious.