Bug 112372

Summary:	Get a Stale NFS handle whith the firewall on
Product:	[Fedora] Fedora	Reporter:	Need Real Name <e.d.grootjans>
Component:	kernel	Assignee:	Dave Jones <davej>
Status:	CLOSED WONTFIX	QA Contact:	Brian Brock <bbrock>
Severity:	medium	Docs Contact:
Priority:	medium
Version:	1	CC:	pfrields, wtogami
Target Milestone:	---
Target Release:	---
Hardware:	i686
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2005-02-08 01:32:42 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Need Real Name 2003-12-18 15:50:36 UTC

From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows 98; WMM; .NET 
CLR 1.1.4322)

Description of problem:
When I used NFS with REDHAT 9. I could ENABLE the firewall with WWW, 
SSH as trusted services and as a trusted device eth0 (i don't have 
more eth devices).
When i upgraded from REDHAT 9 to Fedora Core 1 i did make a NFS 
connection.
>  mount enterprise:/dos-e /mnt/smb
I could do cd to the directory i want to go to.
I could do a ls. I did get the ls information.
when i did a:
> rpm -Uvh *.rpm  (update packages)
I get a:
error open of (filename) failed: Stale NFS file handle.

When i disable the firewall it works fine.

I have a few questions:
1) Why did it work on REDHAT 9 and not on Fedora Core 1 ?
2) How do i configure my firewall that it enables nfs-connections, 
www and ssh and not the rest? 
3) Why don't i have the same kind of problem when i make an smb 
connection. In that case the firewall could be enabled and still smb 
works.




Version-Release number of selected component (if applicable):


How reproducible:
Always

Steps to Reproduce:
1.Firewall on Enable (WWW, SSH and eth0 trusted)
2.mount a nfs directory on a mounting point
3.try to install a file with rpm
    

Actual Results:  --> Error Stale-handle.

Expected Results:  Just an installation of a rpm.

Additional info:

Comment 1 Need Real Name 2003-12-28 22:23:10 UTC

When i tryed later to update packages with nfs i dit get again a 
stale handle. I tryed the packages gnu_cash and ltpf. I think for 
some little packages it works fine, but for bigger packages i get a 
stale handle. Can anybody say my how this is possible.
When i mount the same directory using smb i don't have any problems.

Comment 2 Brent Fox 2004-01-08 22:15:33 UTC

I don't believe that the firewall should have any affect on creating
stale NFS file handles.

Comment 3 Need Real Name 2004-01-15 09:17:03 UTC

I have tree pc's. The enterprise is the hostmachine, where the rpm's 
are. I log in on all tree machine's as root because i have to update.
I do a mount on the first machine, the Britanica.
I mount the enterprise:/dos-e on /mnt/smb on the britanica. Dos-e is 
a fat32 partition (Win98).
After the mount i go to the directory:
cd /mnt/smb
cd download/new/linux/redhat/fedora/errata/packages/to_install
ls
till sofar it always works. Sometimes i can do a rpm -Uvh *.rpm or a 
rpm -Fvh *.rpm and it works but sometimes i get a stale handle.


I always get a stale handle when:
If i try to mount the directory on another machine, and go to the 
install-directory on the mounted dos-e-dir i get a stale handle on 
the first machine.

When i have a stale handle on say the Britanica, i can mount again 
and go again to the install-directory. A do first everything i have 
to do on this machine and then go to the other machine and start with 
mounting the directory.

I thought it must be possible to have more than one PC mounted on the 
same directory at the same time.

Comment 4 Thomas Woerner 2005-01-26 17:57:43 UTC

This is not an iptable problem, whichis the user-land tool.
This might be a kernel problem.

Assigning to kernel.

Comment 5 Dave Jones 2005-01-26 22:30:34 UTC

do you still see this problem with the current Fedora releases ?
FC1 reached end of life a while ago, so unless its still occuring, this
should be closed.