releases/2.6.32.12/module-fix-__module_ref_addr.patch

   1 From mathieu.desnoyers@efficios.com  Wed Apr 21 15:42:43 2010
   2 From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
   3 Date: Tue, 20 Apr 2010 10:38:10 -0400
   4 Subject: module: fix __module_ref_addr()
   5 To: Greg KH <greg@kroah.com>
   6 Cc: Steven Rostedt <rostedt@goodmis.org>, Randy Dunlap <randy.dunlap@oracle.com>, Greg Kroah-Hartman <gregkh@suse.de>, Peter Zijlstra <a.p.zijlstra@chello.nl>, stable <stable@kernel.org>, Rusty Russell <rusty@rustcorp.com.au>, linux-kernel@vger.kernel.org, Eric Dumazet <dada1@cosmosbay.com>, Tejun Heo <tj@kernel.org>, Ingo Molnar <mingo@elte.hu>, Linus Torvalds <torvalds@linux-foundation.org>, Andrew Morton <akpm@linux-foundation.org>
   7 Message-ID: <20100420143810.GC14622@Krystal>
   8 Content-Disposition: inline
   9
  10 From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
  11
  12 The __module_ref_addr() problem disappears in 2.6.34-rc kernels because these
  13 percpu accesses were re-factored.
  14
  15 __module_ref_addr() should use per_cpu_ptr() to obfuscate the pointer
  16 (RELOC_HIDE is needed for per cpu pointers).
  17
  18 This non-standard per-cpu pointer use has been introduced by commit
  19 720eba31f47aeade8ec130ca7f4353223c49170f
  20
  21 It causes a NULL pointer exception on some configurations when CONFIG_TRACING is
  22 enabled on 2.6.33. This patch fixes the problem (acknowledged by Randy who
  23 reported the bug).
  24
  25 It did not appear to hurt previously because most of the accesses were done
  26 through local_inc, which probably obfuscated the access enough that no compiler
  27 optimizations were done. But with local_read() done when CONFIG_TRACING is
  28 active, this becomes a problem. Non-CONFIG_TRACING is probably affected as well
  29 (module.c contains local_set and local_read that use __module_ref_addr()), but I
  30 guess nobody noticed because we've been lucky enough that the compiler did not
  31 generate the inappropriate optimization pattern there.
  32
  33 This patch should be queued for the 2.6.29.x through 2.6.33.x stable branches.
  34 (tested on 2.6.33.1 x86_64)
  35
  36 Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
  37 Tested-by: Randy Dunlap <randy.dunlap@oracle.com>
  38 CC: Eric Dumazet <dada1@cosmosbay.com>
  39 CC: Rusty Russell <rusty@rustcorp.com.au>
  40 CC: Peter Zijlstra <a.p.zijlstra@chello.nl>
  41 CC: Tejun Heo <tj@kernel.org>
  42 CC: Ingo Molnar <mingo@elte.hu>
  43 CC: Andrew Morton <akpm@linux-foundation.org>
  44 CC: Linus Torvalds <torvalds@linux-foundation.org>
  45 CC: Greg Kroah-Hartman <gregkh@suse.de>
  46 CC: Steven Rostedt <rostedt@goodmis.org>
  47 Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
  48
  49 ---
  50  include/linux/module.h |    2 +-
  51  1 file changed, 1 insertion(+), 1 deletion(-)
  52
  53 --- a/include/linux/module.h
  54 +++ b/include/linux/module.h
  55 @@ -455,7 +455,7 @@ void symbol_put_addr(void *addr);
  56  static inline local_t *__module_ref_addr(struct module *mod, int cpu)
  57  {
  58  #ifdef CONFIG_SMP
  59 -       return (local_t *) (mod->refptr + per_cpu_offset(cpu));
  60 +       return (local_t *) per_cpu_ptr(mod->refptr, cpu);
  61  #else
  62         return &mod->ref;
  63  #endif