kdump: crash_kexec()-smp_send_stop() race in panic

[Michael Holzheu]

On Tue, 2011-10-25 at 05:04 -0700, Eric W. Biederman wrote:
> Michael Holzheu <holzheu at linux.vnet.ibm.com> writes:
> 
> > Hello Eric,
> >
> > On Mon, 2011-10-24 at 10:07 -0700, Eric W. Biederman wrote:
> >
> > [snip]
> >
> >> So my second thought is to introduce another atomic variable
> >> panic_in_progress, visible only in panic.  The cpu that sets
> >> increments panic_in_progress can call smp_send_stop.  The rest of
> >> the cpus can just go into a busy wait.  That should stop nasty
> >> fights about who is going to come out of smp_send_stop first.
> >
> > So this is a spinlock, no? What about the following patch:
> Do we want both panic printks?

Ok, good point. We proably should not change that.

> We really only need the mutual exclusion starting just before
> smp_send_stop so that is where I would be inclined to put it.

I think to fix the race, at least we have the get the lock before we
call crash_kexec(). 

Is the following patch ok for you?
---
 kernel/panic.c |    8 ++++++++
 1 file changed, 8 insertions(+)

--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -59,6 +59,7 @@ EXPORT_SYMBOL(panic_blink);
  */
 NORET_TYPE void panic(const char * fmt, ...)
 {
+	static DEFINE_SPINLOCK(panic_lock);
 	static char buf[1024];
 	va_list args;
 	long i, i_next = 0;
@@ -82,6 +83,13 @@ NORET_TYPE void panic(const char * fmt,
 #endif
 
 	/*
+	 * Only one CPU is allowed to execute the panic code from here. For
+	 * multiple parallel invocations of panic all other CPUs will wait on
+	 * the panic_lock. They are stopped afterwards by smp_send_stop().
+	 */
+	spin_lock(&panic_lock);
+
+	/*
 	 * If we have crashed and we have a crash kernel loaded let it handle
 	 * everything else.
 	 * Do we want to call this before we try to display a message?