linux-kernel - Re: Please revert "debug: Make CONFIG_EXPERT select CONFIG_DEBUG

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20111012083801.GA27853@elte.hu>
Date:	Wed, 12 Oct 2011 10:38:01 +0200
From:	Ingo Molnar <mingo@...e.hu>
To:	Adrian Bunk <bunk@...sta.de>
Cc:	Arjan van de Ven <arjan@...radead.org>,
	Josh Triplett <josh@...htriplett.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	linux-kernel@...r.kernel.org,
	Frederic Weisbecker <fweisbec@...il.com>,
	Sam Ravnborg <sam@...nborg.org>
Subject: Re: Please revert "debug: Make CONFIG_EXPERT select
 CONFIG_DEBUG_KERNEL to unhide debug options"


* Adrian Bunk <bunk@...sta.de> wrote:

> On Mon, Oct 10, 2011 at 12:21:21PM +0200, Ingo Molnar wrote:
> > 
> > * Adrian Bunk <bunk@...sta.de> wrote:
> > 
> > > On Mon, Oct 10, 2011 at 09:29:48AM +0200, Ingo Molnar wrote:
> > > > * Adrian Bunk <bunk@...sta.de> wrote:
> > > >...
> > > > I think you are wrong not just about that detail but about the whole 
> > > > premise of your complaint as well:
> > > > 
> > > > >  config DEBUG_BUGVERBOSE
> > > > > -	bool "Verbose BUG() reporting (adds 70K)" if DEBUG_KERNEL && EXPERT
> > > > > +	bool "Verbose BUG() reporting (adds 70K)" if EXPERT
> > > > > 
> > > > > This part of the patch would have been the correct and complete 
> > > > > solution for DEBUG_BUGVERBOSE.
> > > > 
> > > > Not really - it's a debugging-only option and when i turn on 
> > > > CONFIG_DEBUG_KERNEL I expect to have full config control over all 
> > > > debug options - with sane defaults provided.
> > > 
> > > Then you would have to remove the dependency on EXPERT from the prompt, 
> > > and allow unsetting DEBUG_BUGVERBOSE with EXPERT=n, DEBUG_KERNEL=y.
> > > 
> > > Note that this is completely unrelated to the commit we are discussing,
> > > since commit f505c553 has no effect in the EXPERT=n case you are 
> > > discussing here.
> > > 
> > > > I definitely don't want a debugging option to depend on 
> > > > CONFIG_EXPERT.
> > > 
> > > DEBUG_BUGVERBOSE does not depend on EXPERT.
> > > 
> > > But EXPERT is currently required for disabling it.
> > 
> > Correct - that's a further variation. In the case of debug options 
> > that we *really* don't want normal users to disable we do something 
> > like this:
> > 
> >  config DEBUG_BUGVERBOSE
> >          bool "Verbose BUG() reporting (adds 70K)" if DEBUG_KERNEL && EXPERT
> > 
> > Commit f505c553 ("debug: Make CONFIG_EXPERT select 
> > CONFIG_DEBUG_KERNEL to unhide debug options") allows this line to be 
> > further simplified into:
> > 
> >         bool "Verbose BUG() reporting (adds 70K)" if EXPERT
> > 
> > ... but this was not the main purpose of the commit - nor is this 
> > simplification strictly necessary.
> 
> You do not need commit f505c553 for that, the dependency of this 
> prompt on DEBUG_KERNEL should be removed in any case.

An opt-out model is more maintainable here than an opt-in method. 
People will get debug dependencies right most of the time - but 
getting debug *and* CONFIG_EXPERT interactions right is on the 
backburner generally.

So we are better off if CONFIG_EXPERT simply implies (selects) 
CONFIG_KERNEL_DEBUG - makes CONFIG_EXPERT an invariant as far as 
debugging features are concerned and reduces/eliminates the trickle 
of avoidable CONFIG_EXPERT tweaking patches in lib/Kconfig.debug.

> Why do you want to make life harder for people with EXPERT=y by not 
> allowing them to turn this off if they want to?

It's simpler to have one flat CONFIG_EXPERT=y option to gain broad 
expert-configurability of core debug functionality of the kernel.

It should arguably not explode the options to *all* drivers of the 
kernel:

> When configuring his kernel, a user set MISC_FILESYSTEMS=n.
> 
> Now he sets EXPERT=y and runs "make oldconfig".
>
> Why would it make sense that he is now asked for each of these 
> filesystems whether he wants to enable it?

I agree with you that filesystems are more like drivers here and 
should probably not be selected by CONFIG_EXPERT. It's up to the VFS 
folks whether they consider experts to be frequent requestors. 
(probably not)

But the important case here is the situation outlined in the 
changelog, that a default-y core kernel option such as BUGVERBOSE 
unconditionally *adds* code. Core kernel debugging code is an area 
that by its nature has and is bound to have such options - 
MISC_FILESYSTEMS probably not.

Thanks,

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/