Re: [PATCH 3/6] x86/dumpstack: make printk_stack_address() more generally useful

From: Joe Perches <joe@perches.com>
To: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@kernel.org>, "H . Peter Anvin" <hpa@zytor.com>,
	x86@kernel.org, linux-kernel@vger.kernel.org,
	Andy Lutomirski <luto@amacapital.net>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Steven Rostedt <rostedt@goodmis.org>,
	Brian Gerst <brgerst@gmail.com>,
	Kees Cook <keescook@chromium.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Frederic Weisbecker <fweisbec@gmail.com>,
	Byungchul Park <byungchul.park@lge.com>,
	Nilay Vaish <nilayvaish@gmail.com>
Subject: Re: [PATCH 3/6] x86/dumpstack: make printk_stack_address() more generally useful
Date: Wed, 24 Aug 2016 12:57:16 -0700	[thread overview]
Message-ID: <1472068636.3746.170.camel@perches.com> (raw)
In-Reply-To: <20160824192445.tiftvwceupuixays@treble>

On Wed, 2016-08-24 at 14:24 -0500, Josh Poimboeuf wrote:
> On Wed, Aug 24, 2016 at 12:07:06PM -0700, Joe Perches wrote:
> > On Wed, 2016-08-24 at 13:43 -0500, Josh Poimboeuf wrote:
> > > On Wed, Aug 24, 2016 at 10:28:38AM -0700, Joe Perches wrote:
> > > > On Wed, 2016-08-24 at 11:50 -0500, Josh Poimboeuf wrote:
> > > > > Change printk_stack_address() to be useful when called by an unwinder
> > > > > outside the context of dump_trace().
> > > > > 
> > > > > Specifically:
> > > > > 
> > > > > - printk_stack_address()'s 'data' argument is always used as the log
> > > > >   level string.  Make that explicit.
> > []
> > > > If this is true, and I'm not sure it is as I believe
> > > > there are static strings emitted like EOE and IRQ,
> > > > shouldn't this bubble up through the calling tree?
> > > []
> > > This function needs to keep its 'void *data' argument because it's a
> > > callback for stacktrace_ops, so it has to conform to the callback
> > > interface.  'data' is used for passing a pointer to an opaque data
> > > structure to the callback.
> > > 
> > > Also this is the only caller of printk_stack_address(), so there's
> > > nowhere else to bubble it up to.
> > And that shows that print_stack_address(data is not always a log level.
> > ie: walk_stack uses it to print a string not a log level.
> Hm, can you be more specific?  As far as I can tell, here's the only
> possible call path to print_trace_address() and printk_stack_address():
> 
> show_trace_log_lvl()
>   dump_trace()				// ops is print_trace_op
>     print_context_stack()		// ops->walk_stack
>       print_trace_address()		// ops->address
>           printk_stack_address()
> 
> So 'data' is a sneaky way to pass 'log_lvl' from show_trace_log_lvl() to
> print_trace_address(), without dump_trace() and print_context_stack()
> knowing what it is, because they're used in other places where 'data'
> means something else.

hmm, perhaps I got lost in a twisting maze of little callbacks.
I'll drop some stuff and see where I am next time.