Description of problem:
When smbd is under heavy load the RHEL4.5 kernel (and likely all current RHEL
kernels, RHEL5 included) will eventually hit a race that causes free_uid to NULL
pointer.

Version-Release number of selected component (if applicable):
2.6.9-55.0.12.ELsmp

How reproducible:
In production environments that make heavy use of smbd; the issue has hit 4
times in the past 2 weeks.

Steps to Reproduce:
1. run RHEL4.5 kernel
2. put heavy load on samba with many users
3. eventually you'll lose this race
  
Actual results:
Unable to handle kernel paging request at 0000000000100108 RIP: 
<ffffffff801411f7>{free_uid+45}
...
Process smbd (pid: 2227, threadinfo 0000010038b72000, task 0000010087814030)
Stack: 0000000000000000 0000000000000002 0000010237e066e8 ffffffff801419c9 
       0000000000000000 0000010038b73e78 0000010087814030 0000010087814708 
       0000010038b73f58 ffffffff80141a7e 
Call Trace:<ffffffff801419c9>{__dequeue_signal+347}
           <ffffffff80141a7e>{dequeue_signal+58} 
           <ffffffff801435ca>{get_signal_to_deliver+338}
           <ffffffff8010f6fb>{do_signal+131} 
           <ffffffff8030c8f6>{thread_return+88}
           <ffffffff801102f3>{sysret_signal+28} 
           <ffffffff801105df>{ptregscall_common+103} 

Code: 48 89 50 08 48 89 02 48 c7 41 08 00 02 20 00 48 8b 7b 38 48 
RIP <ffffffff801411f7>{free_uid+45} RSP <0000010038b73d98>
CR2: 0000000000100108

Expected results:
No NULL pointer.

Additional info:
Linus Torvalds fixed the issue upstream in 2.6.19-rc4:
http://lkml.org/lkml/2006/11/4/45
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=45c18b0

RHEL5 also doesn't have this fix and it should.