Samba shared folder size

Hi everyone,

Apologies if these two (somewhat related) questions have been dealt with before. Haven't found them anywhere...

I find myself in the position of managing a 32TB Gentoo file server. It was previously set up with dozens of LVM2 logical volumes, all under a couple Samba shares. The main share is called Graphics. There is, in fact, a logical volume mounted at /mnt/graphics; it is 20GB.

# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/mapper/vg-graphics_base
                       20G   12M   20G   1% /mnt/graphics

Other logical volumes are mounted beneath this location, like at /mnt/graphics/whatever. Since each logical volume has a fixed size, it would be nice if the users could see that size and how much space is available. When they, in Windows, map a network drive to any logical volume beneath the Graphics share, they see that drive as being 19.9GB with 19.9GB free. They're seeing the size of the logical volume mounted at /mnt/graphics, not /mnt/graphics/whatever. They get the same thing when they map just the Graphics share itself - which is correct.

Fine, I thought, I'll get rid of that nested structure and they'll see the proper size, right? Wrong. Another share, called Dragon, does not have its own logical volume, just some at /mnt/dragon/raid01, etc.

# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/mapper/vg-raid1
                      170G  155G   16G  92% /mnt/dragon/raid01

When mapping a drive to the raid01 folder, I thought I'd get an accurate size. But no, I get a total size of 34.6GB with 27.6GB free. That's the size of the root partition on the server's system hard drive.

So my users have no idea how much room is in each logical volume. I'm cleaning up the nested structure anyway and drastically reducing the number of logical volumes, but is there a way for Windows users to get an accurate read of how much space is available?

That leads to my second question, because it's been the case several times that users unknowingly fill up a logical volume. This tends to crash the server with some smbd-related kernel panic. Seems like that shouldn't happen; I've read about full root partitions bringing down a server, but just any logical volume? Is that expected behavior? Can it be prevented?

Some selected system info follows; ask if you want more!

# uname -a
Linux godzilla 2.6.24-gentoo-r5 #1 SMP Thu Jun 26 18:36:15 UTC 2008 i686 Intel(R) Xeon(TM) CPU 2.80GHz GenuineIntel GNU/Linux

# lvs --version
  LVM version:     2.02.36 (2008-04-29)
  Library version: 1.02.24 (2007-12-20)
  Driver version:  4.12.0

# smbd -V
Version 3.0.28a

smb.conf excerpt:

[Graphics]
   comment = Materiel for graphics department
   path = /mnt/graphics
   public = yes
   writable = yes

[Dragon]
   comment = Old Dragon RAIDs
   path = /mnt/dragon
   public = yes
   writable = yes

A partial answer!

Apparently, Windows will show the correct size IF the logical volume is mounted at the Samba share mount point. So the 20GB logical volume at /mnt/graphics shows up correctly because it's mounted at the same place in the directory structure as the [Graphics] Samba share. If I un-nest all these logical volumes and mount them all at the same points as Samba shares, their sizes will display correctly on Windows clients. That works for me.

Anyone know why a full logical volume would kernel panic the system?