Re: [PATCH 2/2] x86: speed cpu_up by quirking cpu_init_udelay

[Ingo Molnar <mingo@kernel.org>]

(Resending my reply with more dyn-debug folks Cc:-ed)

* Len Brown <lenb@kernel.org> wrote:

> On Fri, May 8, 2015 at 4:32 AM, Borislav Petkov <bp@alien8.de> wrote:
> 
> >> +     pr_debug("cpu_init_udelay quirk to %d, was %d", new_udelay, init_udelay);
> >
> > Can we make this printk(KERN_DEBUG please?
> >
> > I'd like to be able to slap "debug" on the command line and not 
> > recompile the kernel. And no, dyndbg="file smpboot.c +p" or 
> > whatever the syntax is, simply doesn't scale if I want to see all 
> > debug messages from early boot.

Ugh, so I see we have grown this gem some time ago:

 #if defined(CONFIG_DYNAMIC_DEBUG)
 /* dynamic_pr_debug() uses pr_fmt() internally so we don't need it here */
 #define pr_debug(fmt, ...) \
         dynamic_pr_debug(fmt, ##__VA_ARGS__)

I didn't even realize it's there and it happend 6 years ago, in a very 
unintuitively titled commit:

  346e15beb534 driver core: basic infrastructure for per-module dynamic debug messages

So in what way does that title tell us that all pr_debug() calls are 
redirected away if CONFIG_DYNAMIC_DEBUG is enabled (which distros do)?

So could we instead either add a dyndbg=all variant, or make 'debug' 
trigger all dynamic_pr_debug() messages?

Because this redirection breaks the whole pr_*() abstraction rather 
fundamentally, dyndebug stealing pr_debug() and hiding debug messages 
when the user specifically asked for them via 'debug' is pretty nasty 
IMHO ...

Thanks,

	Ingo