memory leak in NFS if a sever goes away?

memory leak in NFS if a sever goes away?

Post by Florian Weime » Thu, 02 Aug 2001 01:10:11



We've received a hard disk for post-mortem analysis because of a
strange hole of several months in the system logs, and it seems the
system's syslogd was killed by the VM subsystem during an OOM
situation.

The problems seem to begin at the following syslog event:

   kernel: nfs: server sun not responding, still trying

Periodically (every quarter of an hour), the following messages appear:

   kernel: nfs: task 111 can't get a request slot

(The task number is monotonically increasing.)

This goes on for about two days, after which the VM subsystem starts
killing processes (kdm first, then several times the X server, and
finally syslogd itself).

Are there some known issues with 2.2.13, for example, a memory leak in
the NFS code which is triggered in this specific situation?

(The kernel seems to have some SuSE-specific patches, for example, the
X server is sent the TERM signal, not the KILL signal on OOM.  Perhaps
I should ask the SuSE folks if there were any NFS peculiarities in
their 2.2.13 version. :-/)

--

University of Stuttgart           http://cert.uni-stuttgart.de/
RUS-CERT                          +49-711-685-5973/fax +49-711-685-5898
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in

More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

 
 
 

memory leak in NFS if a sever goes away?

Post by Matthias Andre » Thu, 02 Aug 2001 06:30:10


Florian Weimer schrieb am Dienstag, den 31. Juli 2001:

Quote:> We've received a hard disk for post-mortem analysis because of a
> strange hole of several months in the system logs, and it seems the
> system's syslogd was killed by the VM subsystem during an OOM
> situation.

If it's crucial, list it in inittab.

Quote:> Are there some known issues with 2.2.13, for example, a memory leak in
> the NFS code which is triggered in this specific situation?

There are known security threats in 2.2.x which have been fixed in
2.2.16, and there have been VM fixes in a later version, and further
minor but numerable security fixes in 2.2.19.

Update to 2.2.19. Don't bother analyzing 2.2.13, no-one fixes that
nowadays unless the problem persists in 2.2.19.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in

More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

 
 
 

1. Mandrake 6.1 nfs client mountpoints "go away" after a while

More info, found this in /var/log/messages today:

Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 1 still busy, i_count=1
Nov  9 04:03:42 rex last message repeated 4 times
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 2 still busy, i_count=1
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 3 still busy, i_count=1
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 4 still busy, i_count=1
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 5 still busy, i_count=1
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 7 still busy, i_count=1
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 9 still busy, i_count=1
Nov  9 04:03:42 rex kernel: __nfs_fhget: inode 16 still busy, i_count=2
Nov  9 04:03:42 rex kernel: __nfs_fhget: killing ///
filehandle                

This is on the client side.
Could this be when it "died"?
What does it mean?

-Bruce.

--

Sattel Global Networks          818.709.6201 x122

2. Problem compiling VNC to work in LPPC 1999

3. NFS Mounts go away!

4. Need ELM for Solaris 2.3

5. nfs memory leak

6. Reading is much more interesting than TV (1575/1708)

7. memory leaks by shared memory and fork

8. Western Digital Caviar HD and 2048 Cylinders in Linux

9. Sun workshop memory monitor does not detect all memory leak

10. 64bit solaris 2.8 memory name resolution memory leak?

11. How can I recover memory from a memory leak

12. mm/memory.c, 2.4.1 : memory leak with swap cache (updated)

13. How to find out the memory size used by a process or check its memory leaking?