[linux-cifs-client] RE: zombie cifsd kernel threads

Nick Millington Nick.Millington at rinconnetworks.com
Tue May 25 21:30:44 GMT 2004


I think the cause might actually be more mundane: I am performing the
mount operations using a direct invocation of the "mount" system call
from a long-running process.  My long-running process is listed as the
parent PID of the zombies by ps.  Should the cifs_demultiplex_thread
entry point call "reparent_to_init" after "daemonize" in order to ensure
that someone reaps the thread when it exits?
 
-N
 
-----Original Message-----
From: Steven French [mailto:sfrench at us.ibm.com] 
Sent: Tuesday, May 25, 2004 12:53 PM
To: linux-cifs-client at lists.samba.org
Cc: sfrench at samba.org; Nick Millington
Subject: Re: zombie cifsd kernel threads
 

My guess about seeing a zombie cifsd thread that goes away on unmount -
in that version of the 2.4 cifs code, cifsd had cases in which it would
prematurely exit due to what turned out to be "expected" network errors.
I have reworked the "cifs_demultiplex_thread" function (which is running
in the cifsd background process, one per server you are connected to,
and is used to receive all smb/cifs replies from the server and dish
them out to the right process on the client) in 2.6 but have not
backported it to Linux 2.4 yet. 

If you can still do operations to the server, then cifsd is working of
course no matter what its process state displays as on your distro, but
if you are seeing requests hang and cifsd in zombie state then I expect
that it would be fixed by a backport of the fixes in
cifs_demultiplex_thread code in fs/cifs/connect.c from 2.6 to 2.4    I
know there are cases I have fixed in which cifsd could prematurely exit
in that older version of the cifs vfs. 




Steve French
Senior Software Engineer
Linux Technology Center - IBM Austin
phone: 512-838-2294
email: sfrench at-sign us dot ibm dot com
-------------- next part --------------
HTML attachment scrubbed and removed


More information about the linux-cifs-client mailing list