Re: Warning and BUG with btrfs and corrupted image

From: Pavel Machek <pavel@suse.cz>
To: Chris Mason <chris.mason@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>,
	Eric Sesterhenn <snakebyte@gmx.de>,
	linux-kernel@vger.kernel.org, linux-btrfs@vger.kernel.org
Subject: Re: Warning and BUG with btrfs and corrupted image
Date: Thu, 5 Feb 2009 14:50:11 +0100	[thread overview]
Message-ID: <20090205135011.GA2294@elf.ucw.cz> (raw)
In-Reply-To: <1233838959.15948.6.camel@think.oraclecorp.com>

On Thu 2009-02-05 08:02:39, Chris Mason wrote:
> On Thu, 2009-02-05 at 10:02 +0100, Pavel Machek wrote:
> > > > > CONFIG_*_DEBUG means include *debug* code there to help developers,
> > > > > including adding additional failure tests into the kernel. Besides,
> > > > > which bit of "don't turn it on unless you are an XFS developer"
> > > > > don't you understand?
> > > > 
> > > > Yes, but DEBUG code is normally to help debugging, not to crash
> > > > kernels.
> > > 
> > > Crashing the kernel at exactly the point a problem is detected
> > > is often the simplest way of debugging the problem.
> > > 
> > > e.g. CONFIG_VM_DEBUG=y turns on VM_BUG_ON() which crashes the kernel
> > > whenever it detects something wrong. Do I turn it on? Yes. Do i
> > 
> > That's different. User is not supposed to be able to trigger
> > VM_BUG_ON().
> > 
> > > complain about it when I hit a VM_BUG_ON()? No, I report the
> > > bug and move on. If you turn on a DEBUG option, then you are
> > > asking the system to behave in a way useful to a developer,
> > > not an end user. That includes panicing when something wrong
> > > is detected.
> > 
> > Imagine vm going panic() on mkdir("/lost+found")...
> 
> It is up to the XFS developers to decide what their debugging options
> do.  

Eh?

> The whole point of panicing is so that you can collect important
> information about the system at the time of the error condition.  When
> this option is compiled on, panic on mkdir is exactly what they are
> asking for.
> 
> If you don't want it, don't compile it in.  The Kconfig text is very
> clear.

No, I'd not expect that option to panic systems. That's why I
suggested:

diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig
index 29228f5..b7ac847 100644
--- a/fs/xfs/Kconfig
+++ b/fs/xfs/Kconfig
@@ -77,4 +77,7 @@ config XFS_DEBUG
 	  Note that the resulting code will be HUGE and SLOW, and probably
 	  not useful unless you are debugging a particular problem.
 
+	  Turning this option on will result in kernel panicking any time
+	  it detects on-disk corruption.
+
 	  Say N unless you are an XFS developer, or you play one on TV.


-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
