kjournald hang on ext3 to ext3 copy

[John Marconi]

All,

I am running into a situation in which one of my ext3 filesystems is 
getting hung during normal usage.  There are three ext3 filesystems on a 
CompactFLASH.  One is mounted as / and one as /tmp.  In my test, I am 
copying a 100 MB file from /root to /tmp repeatedly.  While doing this 
test, I eventually see the copying stop, and any attempts to access /tmp 
fail - if I even do ls /tmp the command will hang.

I suspect kjournald because of the following ps output:
PID      PPID   WCHAN:20      PCPU  %MEM  PSR  COMM
 8847    99 start_this_handle        1.1  0.0  28     pdflush
 8853    99 schedule_timeout       0.2  0.0   7     pdflush
  188     1 kswapd                       0.0  0.0  19   kswapd0
 8051     1 mtd_blktrans_thread   0.0  0.0  22   mtdblockd
 8243     1 kjournald                    0.0  0.0   0   kjournald
 8305     1 schedule_timeout        0.0  0.0   2   udevd
 8378     1 kjournald                    0.0  0.0   0   kjournald
 8379     1 journal_commit_trans 16.6  0.0   0   kjournald
 8437     1 schedule_timeout       0.0  0.0   0   evlogd
 8527     1 syslog                        0.0  0.0   1   klogd
 8534     1 schedule_timeout       0.0  0.0   0   portmap
 8569     1 schedule_timeout       0.0  0.0   0   rngd
 8639     1 schedule_timeout       0.1  0.0  24   sshd
 8741  8639 schedule_timeout    0.0  0.0   0     sshd
 8743  8741 wait                        0.0  0.0   9       bash
 8857  8743 schedule_timeout    4.9  0.0   7         cp
 8664     1 schedule_timeout       0.0  0.0   0   xinetd
 8679     1 schedule_timeout       0.0  0.0   0   evlnotifyd
 8689     1 schedule_timeout       0.0  0.0   0   evlactiond
 8704     1 wait                           0.0  0.0   1   bash
 8882  8704 -                            0.0  0.0   2     ps

If I run ps repeatedly, I always see process 8379 in 
journal_commit_transaction, and it is always taking between 12% and 20% 
of processor 0 up.  This process never completes.  I also see process 
8847 in start_this_handle forever as well - so I believe they are related. 

This system is using a 2.6.14 kernel.

Has anyone seen this type of behaviour before?  Note, if I change /tmp 
to ext2 I never see this issue - it is only when /tmp is mounted as ext3.

Thank you,
John