Re: BUG: spinlock recursion (sys_chdir, user_path_at, do_path_lookup ...)

[Thomas Gleixner <tglx@linutronix.de>]

On Wed, 12 Jan 2011, Russell King - ARM Linux wrote:
> On Wed, Jan 12, 2011 at 12:35:08PM +0000, Russell King - ARM Linux wrote:
> > ARM doesn't implement save_stack_trace_regs() nor save_stack_trace_bp()
> > so if the compiler referenced these, you'd have a kernel which doesn't
> > link.  The only places that this symbol appears is:
> > 
> > arch/x86/kernel/stacktrace.c:void save_stack_trace_regs(struct stack_trace *trac
> > arch/x86/mm/kmemcheck/error.c:  save_stack_trace_regs(&e->trace, regs);
> > include/linux/stacktrace.h:extern void save_stack_trace_regs(struct stack_trace
> > 
> > So, if this is where your bisect decided was the problem, your bisect
> > was faulty.
> 
> BTW, a useful thing to do after a bisect is to return to the point in
> the history where you first noticed the regression (so Linus' tip,
> your tip, or whatever).  Then try reverting the commit which git bisect
> _thinks_ is the cause of your problem and re-test that.
> 
> If the problem is fixed, you have greater confidence that the commit is
> the problem.
> 
> If it made no difference, then you know that something else (maybe in
> combination) is causing the problem.
> 
> If you couldn't revert it because of other dependencies then you have
> to rely on analysis (such as what I did) and maybe try again with a
> slightly different strategy - maybe the problem only _occasionally_
> occurs, making the 'git bisect good' points unreliable, so maybe you
> need to do more testing when the problem doesn't immediately appear?
> 
> Lastly, it is worth bearing in mind that GCC is really finicky with its
> optimization.  It may be hard to believe, but unrelated function
> definitions in headers can (and do) affect the code generation in
> completely unrelated functions causing them to be optimized
> differently [*].  Maybe this applies to prototypes too?

Yes, it does. Also adding an inline or define can change the
behaviour.

> So it _could_ be that the prototype change in include/linux/stacktrace.h
> is tickling a GCC code generation bug.
> 
> * - ISTR, this behaviour was raised as a bug with GCC folk, which I
> believe was closed down as wontfix as its a result of the way the
> optimizer works.

Right, they just fixed the problem where this effect generated buggy
code on x86 in some cases.

Thanks,

	tglx