more fun with hitting the hard quota

[seth vidal <skvidal@phy.duke.edu>]

[-- Attachment #1: Type: text/plain, Size: 1118 bytes --]

Hi,
 So I've setup a couple of machines to test this in more detail.

The server is running rh's skipjack beta2. a 2.4.18 +a bunch of patches
kernel.

ext3 filesystem, quotas turned on.
I set the user's hard quota to 200MB

if I mount it over nfs and then run:


while `true`
do
cat /dev/zero >> /mnt/test/dumbfile
done

Then the nfs server does respond with error messages about being over
the quota and the current cat job will stop with this error:

cat: write error: Disk quota exceeded

But if I consistently make the request to write to that file (as in the
above loop) the load on the server goes up to about 4 or 5.

At which point the other clients access becomes almost unbearable.

Does it make sense that one client should be able to abuse the server in
this way? And why didn't this happen on 2.2.19? I had a lesser powered
machine doing the same job and this problem never occurred.

I can provide more detailed tcpdumps from the server or client side now
if that'd help more.

Thanks
-sv






-- 
GPG Public Key: http://www.phy.duke.edu/~skvidal/skvidal.gpg

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 232 bytes --]