umount error: device is busy after nfs + lock

umount error: device is busy after nfs + lock

[wang dengyi]

Hello,

When I ran ltp test case:tlocklfs, I met a old problem
with the kernel:2.6.9. I can not umount a file system
after nfs and lock using it. Here is the steps.
 1) mount /dev/sda3 /xyz
 2) export /xyz
 3) restart nfs
 4) mount 127.0.0.1:/xyz /mnt/nfs_xyz
 5) run the special lock test program: "tlocklfs -t 7
/mnt/nfs_xyz
 6) umount /mnt/nfs_xyz
 7) stop nfs
 8) umount /xyz
Then I got the error: "umount /xyz: device is busy".
And the error displays twice. 

If I run the different lock test case for the step 5,
the umount works fine. It only failed on the test case
7 from tlocklfs. 

There are 2 processes in this special test case.
First, the parent process "setlk" for a file from nfs
file system. Then the child process "setlkw". At the
end, the parent and the child process unlock the file.


With some debug line in the code, I found the file
system's "mnt_count" is 3 instead of 2 before
umounting it. I traced back the problem and I'm lost
in linux/net/sunrpc/sched.c. Could anyone give me some
hint? Thank you very much.

Best regard

Dengyi Wang
  

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

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

Re: umount error: device is busy after nfs + lock

[Trond Myklebust]

On Mon, 2006-05-15 at 11:59 -0700, wang dengyi wrote:
> Hello,
> 
> When I ran ltp test case:tlocklfs, I met a old problem
> with the kernel:2.6.9. I can not umount a file system
> after nfs and lock using it. Here is the steps.

Kernel 2.6.9 is almost 2 years old, and the NFS locking has changed a
_lot_ since then. Can you reproduce this with a recent kernel?

Cheers,
  Trond

>  1) mount /dev/sda3 /xyz
>  2) export /xyz
>  3) restart nfs
>  4) mount 127.0.0.1:/xyz /mnt/nfs_xyz
>  5) run the special lock test program: "tlocklfs -t 7
> /mnt/nfs_xyz
>  6) umount /mnt/nfs_xyz
>  7) stop nfs
>  8) umount /xyz
> Then I got the error: "umount /xyz: device is busy".
> And the error displays twice. 
> 
> If I run the different lock test case for the step 5,
> the umount works fine. It only failed on the test case
> 7 from tlocklfs. 
> 
> There are 2 processes in this special test case.
> First, the parent process "setlk" for a file from nfs
> file system. Then the child process "setlkw". At the
> end, the parent and the child process unlock the file.
> 
> 
> With some debug line in the code, I found the file
> system's "mnt_count" is 3 instead of 2 before
> umounting it. I traced back the problem and I'm lost
> in linux/net/sunrpc/sched.c. Could anyone give me some
> hint? Thank you very much.
> 
> Best regard
> 
> Dengyi Wang
>   
> 
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam
> protection around 
> http://mail.yahoo.com 
> 
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around 
> http://mail.yahoo.com 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/