xfs
[Top] [All Lists]

Re: wbee (sample_hsm) dumped core

To: Takayuki Sasaki <sasaki@xxxxxxxxxxxxxxxxxx>
Subject: Re: wbee (sample_hsm) dumped core
From: Dean Roehrich <roehrich@xxxxxxx>
Date: Tue, 06 Nov 2001 10:15:01 -0600
Cc: linux-xfs@xxxxxxxxxxx
Sender: owner-linux-xfs@xxxxxxxxxxx
>From:  Takayuki Sasaki <sasaki@xxxxxxxxxxxxxxxxxx>
>Hi,
>
>Thank you for your time, Dean.
>
>Dean Roehrich <roehrich@xxxxxxx>  wrote:
>> 
>> >Takayuki Sasaki <sasaki@xxxxxxxxxxxxxxxxxx>  wrote:
>> >
>> >> migin daemon in sample_hsm started with the patch which I
>> >> posted, but if I try to read the migrated file then it
>> >> stalled. It was caused by
>> >> linux-2.4-xfs/cmd/xfstests/dmapi/src/sample_hsm/wbee which was
>> >> dispatched by migin dumped core.
>> >
>> >In the above situation, I killed the stalled command ( cp ) and
>> >migin by pressing Ctrl + c to find out what is wrong.  Then,
>> >unmount the XFS file system, the following console messages
>> >appeared:
>> >
>> >  XFS unmount got error 16
>> >  linvfs_put_super: vfsp/0xc2acb38c left dangling!
>> >  VFS: Busy inodes after unmount. Self-destruct in 5 seconds.  Have a nice 
>day

I didn't forget about you :)

The problem happens because invisible I/O was not invisible, causing DMAPI to
deadlock on itself.  Then you end up with busy vnodes and a mess.

The VOP_READ/VOP_WRITE calls in xfs_dm_rdwr() were not sending the O_INVISIBLE
flag.  I just checked in a fix, and it should make its way to CVS soon.

Dean


<Prev in Thread] Current Thread [Next in Thread>