>
> As suggested, I checked out the swap and memory space on my system. My
> setup is a 700MHz cpu, 128Mb of ram and equal swap space. The 'free' command
>
> shows, that there are approximately 100Mb in use, and an equal amount of swap
>
> space unused.
>
> I built the kernel in the following manner....
>
> Untar the linux 2.4.3 kernel tree... enter that directory.
> zcat <path>/linux-2.4.3-core.patch.gz | patch -p1
> zcat <path>/linux-xfs-patch.gz | patch -p1
> make xconfig
> {
> enabled entities are: athlon, ne2k network card, parallel port, usb,
> busmouse,nfs v3, server v3,automount,ppp async/sync,i2c,video,
> ntfs (read only),udf,ufs,reiser fs,xfs,advanced partitions,xfs partition,
> sound,acl support
> }
> make dep
> make bzImage
> { put the arch/i386/boot/bzImage into the /boot and edit lilo etc :-) }
> cp System.map /boot/System.map-<kernel> and symlink.
> reboot. into the new kernel.
>
> side note on compile: GCC that comes with RedHat 7.1, does not work
> for compiling the XFS kernel, as it will
> break on compiling the XFS code for anything
> except i386. The setup also uses kgcc to
> compile the kernel, which is provided with
> compat-egcs, which compiles the kernel ok.
>
> Other kernel modules are added, such as Alsa and Nvidia, but I don't think
> they effect this. The same thing as above, was done for the redhat 2.4.2-2
> kernel. But that time, the xfs core file for that RH kernel was used, along
> with the linux-xfs patch, to alter the kernel. This kernel behaves the same
> way as the vanilla 2.4.3 one.
>
> This time, I'm using the 1.0.1 tree of XFS on kernel 2.4.5 (vanilla), and
> without ACL compiled into the kernel. It works well, no problems and the
> kernel is far more stable than the 1.0 one. No hangups yet, whatsoever. But
>
> the differences in sizes of the filesystem after the data has been untarred
> onto it, persists. But the sizes are not constant this time, rather vary
> each time I do it. After I unmount and remount the drive, the size is fixed
> to normal (well, a little larger than ext2fs, but that is due to difference
> in block sizes,etc).
>
Hmm, I tried untaring the linux kernel (I am pretty sure Linus is not
using xfs to create these...) the end result was as expected, the files
come out the correct size, du reports sizes as expected. df reports free
space as expected.
Can you provide an example ls -lsR of part of the tree from ext2 and xfs,
can you also run xfs_bmap -v -v on one of the files which appears larger
than it should be.
Also, do you have quotas enabled and are you accessing the filesystem via NFS
at any point?
Steve
>
>
> Orn
|