2006-06-08 04:24:54

by mehta kiran

[permalink] [raw]
Subject: query about kernel 2.4.21-37.ELsmp

Hi,
From where can i get information about all the
bugs
which were there in redhat kernel 2.4.21-37.ELsmp
?
I want to know if there existed a bug which
did not allow unmounting of device which was once
nfs exported and accessed by clients

Thanks,
kiran



__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com


_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs


2006-06-08 06:24:39

by Wendy Cheng

[permalink] [raw]
Subject: Re: query about kernel 2.4.21-37.ELsmp

On Wed, 2006-06-07 at 21:24 -0700, mehta kiran wrote:
> Hi,
> From where can i get information about all the
> bugs
> which were there in redhat kernel 2.4.21-37.ELsmp
> ?
> I want to know if there existed a bug which
> did not allow unmounting of device which was once
> nfs exported and accessed by clients
>

I'm personally not aware of any RHEL 3 bugzilla that turns out to be a
real bug to prevent NFS exported filesytem getting un-mounted. However,
I'm not frequently involved with RHEL NFS bugs either so I could be
wrong.

On the other hand, things to check before you convince yourself there is
a bug somewhere:

1. All the local file accesses (on the server) have been stopped
(processes killed) ?
2.1 If possible, "service nfs stop" invoked ? This will clean up all the
states associated with NFS accesses; Or
2.2.a Manually unexport filesystem. This is a tricky one. Note that if
kernel still hold the vfsmount reference for the export entry, umount
will fail. The "exportfs -u hostname:/export_path" doesn't necessarily
go into the kernel to delete the export entry (to drop vfsmount
reference count). You have to use either "exportfs -ua" or "exportfs
*:/export_path" to explicitly remove the export entry from RHEL 3
kernel.
2.2.b Is NFS client holding a posix lock via NLM without releasing it ?
If yes, a quick and dirty way is to send lockd a signal ("kill -9
lockd"). This will signal lockd to drop the NLM locks (and subsequently
close the file).

-- Wendy






_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2006-06-08 07:54:16

by Wendy Cheng

[permalink] [raw]
Subject: Re: query about kernel 2.4.21-37.ELsmp

On Thu, 2006-06-08 at 02:35 -0400, Wendy Cheng wrote:


> 2.2.a Manually unexport filesystem. This is a tricky one. Note that if
> kernel still hold the vfsmount reference for the export entry, umount
> will fail. The "exportfs -u hostname:/export_path" doesn't necessarily
> go into the kernel to delete the export entry (to drop vfsmount
> reference count). You have to use either "exportfs -ua" or "exportfs
> *:/export_path" to explicitly remove the export entry from RHEL 3
> kernel.

I'm famous for my typo :) ... should use "exportfs -u *:/export_path".

> 2.2.b Is NFS client holding a posix lock via NLM without releasing it ?
> If yes, a quick and dirty way is to send lockd a signal ("kill -9
> lockd"). This will signal lockd to drop the NLM locks (and subsequently
> close the file).

-- Wendy



_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2006-06-08 10:44:19

by mehta kiran

[permalink] [raw]
Subject: Re: query about kernel 2.4.21-37.ELsmp

I tried all the steps you talked
about mount remains busy. The only
thing effective which you mentioned
is killing nfsd. This causes kernel to
release all reference to exported fs.

Steps to reproduce this problem :

1. mount <device> <mountpoint>
2. export it
3. unexport it
4. try umounting it. it wont



Thanks,
kiran

--- Wendy Cheng <[email protected]> wrote:

> On Thu, 2006-06-08 at 02:35 -0400, Wendy Cheng
> wrote:
>
>
> > 2.2.a Manually unexport filesystem. This is a
> tricky one. Note that if
> > kernel still hold the vfsmount reference for the
> export entry, umount
> > will fail. The "exportfs -u hostname:/export_path"
> doesn't necessarily
> > go into the kernel to delete the export entry (to
> drop vfsmount
> > reference count). You have to use either "exportfs
> -ua" or "exportfs
> > *:/export_path" to explicitly remove the export
> entry from RHEL 3
> > kernel.
>
> I'm famous for my typo :) ... should use "exportfs
> -u *:/export_path".
>
> > 2.2.b Is NFS client holding a posix lock via NLM
> without releasing it ?
> > If yes, a quick and dirty way is to send lockd a
> signal ("kill -9
> > lockd"). This will signal lockd to drop the NLM
> locks (and subsequently
> > close the file).
>
> -- Wendy
>
>
>
> _______________________________________________
> NFS maillist - [email protected]
> https://lists.sourceforge.net/lists/listinfo/nfs
>


__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com


_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2006-06-08 13:58:43

by Wendy Cheng

[permalink] [raw]
Subject: Re: query about kernel 2.4.21-37.ELsmp

mehta kiran wrote:

>I tried all the steps you talked
>about mount remains busy. The only
>thing effective which you mentioned
>is killing nfsd. This causes kernel to
>release all reference to exported fs.
>
>Steps to reproduce this problem :
>
>1. mount <device> <mountpoint>
>2. export it
>3. unexport it
>4. try umounting it. it wont
>
>
>
>
>
I don't seem to be able to create it ..... ok, two more things:

* nfs-utils updated ?
* do a "cat /proc/fs/nfs/exports" when umount fails and passes back the
info.
* give me the "exact" command you used for unexport.

-- Wendy


_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

2006-07-02 15:12:02

by Jeff Layton

[permalink] [raw]
Subject: Re: query about kernel 2.4.21-37.ELsmp

On Wed, 2006-06-07 at 21:24 -0700, mehta kiran wrote:
> Hi,
> From where can i get information about all the
> bugs
> which were there in redhat kernel 2.4.21-37.ELsmp
> ?
> I want to know if there existed a bug which
> did not allow unmounting of device which was once
> nfs exported and accessed by clients
>
> Thanks,
> kiran

2.4.21-45.EL and greater (and hence U8) should have a fix for this
problem. See:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=167636

for more info.

-- Jeff



Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs