Re: [PATCH] stacktrace: Provide stack_trace_save_tsk() stub in the !CONFIG_STACKTRACE case too

[Peter Zijlstra]

On Fri, Oct 22, 2021 at 09:43:50AM +0200, Ingo Molnar wrote:
> 
> * Qi Zheng <zhengqi.arch@xxxxxxxxxxxxx> wrote:
> 
> > 
> > 
> > On 10/18/21 2:23 PM, Stephen Rothwell wrote:
> > > Hi all,
> > > 
> > > After merging the tip tree, today's linux-next build (x86_64 allnoconfig)
> > > failed like this:
> > > 
> > > arch/x86/kernel/process.c: In function '__get_wchan':
> > > arch/x86/kernel/process.c:950:2: error: implicit declaration of function 'stack_trace_save_tsk' [-Werror=implicit-function-declaration]
> > >    950 |  stack_trace_save_tsk(p, &entry, 1, 0);
> > >        |  ^~~~~~~~~~~~~~~~~~~~
> > > cc1: some warnings being treated as errors
> > > 
> > > Caused by commit
> > > 
> > >    bc9bbb81730e ("x86: Fix get_wchan() to support the ORC unwinder")
> > > 
> > > stack_trace_save_tsk() requires CONFIG_STACKTRACE which is not set for
> > > this build.
> > 
> > Maybe get_wchan() can be updated to:
> > 
> > unsigned long get_wchan(struct task_struct *p)
> > {
> > #ifdef CONFIG_STACKTRACE
> > 	unsigned long entry = 0;
> > 
> > 	stack_trace_save_tsk(p, &entry, 1, 0);
> > 	return entry;
> > #else /* CONFIG_STACKTRACE */
> > 	return 0;
> > #endif
> > }
> 
> And repeat the same ugliness in every single function that happens to use 
> the stack_trace_save_tsk() API??
> 
> The correct solution is to define stack_trace_save_tsk() in the 
> !CONFIG_STACKTRACE case too, as the patch below does.

That doesn't make sense for x86. We have an unconditional unwinder
present.

I've got these, meant to post them later today:

  https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=sched/wchan