From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] hfsplus: fix warnings in fs/hfsplus/bfind.c: In
 function 'hfs_find_1st_rec_by_cnid'
Date: Fri, 8 Mar 2013 14:29:29 -0800
Message-ID: <20130308142929.fd24d05b529d420fd7beaf1e@linux-foundation.org>
References: <1362324385.29435.31.camel@slavad-ubuntu-12.04>
	<20130307111832.ba5844ce385220fdd57723c0@linux-foundation.org>
	<0AC1048B-F0AB-47BF-A1E9-16125F127D60@dubeyko.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: linux-fsdevel@vger.kernel.org,
	Christoph Hellwig <hch@infradead.org>,
	Al Viro <viro@zeniv.linux.org.uk>,
	Hin-Tak Leung <htl10@users.sourceforge.net>,
	Fengguang Wu <fengguang.wu@intel.com>
To: Vyacheslav Dubeyko <slava@dubeyko.com>
Return-path: <linux-fsdevel-owner@vger.kernel.org>
Received: from mail.linuxfoundation.org ([140.211.169.12]:50782 "EHLO
	mail.linuxfoundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S932158Ab3CHW3b (ORCPT
	<rfc822;linux-fsdevel@vger.kernel.org>);
	Fri, 8 Mar 2013 17:29:31 -0500
In-Reply-To: <0AC1048B-F0AB-47BF-A1E9-16125F127D60@dubeyko.com>
Sender: linux-fsdevel-owner@vger.kernel.org
List-ID: <linux-fsdevel.vger.kernel.org>

On Fri, 8 Mar 2013 16:15:48 +0300 Vyacheslav Dubeyko <slava@dubeyko.com> wrote:

> > I don't see that warning.  Whatever gcc you're using appears to be
> > mishandling unreachable().  Please provide details?
> > 
> 
> The warning is reproducible with special kernel configuration (please, see attached configuration file).

# CONFIG_BUG is not set

Erk yes, CONFIG_BUG=n causes a ton of problems.  But in the case of
x86, it should still work OK:

#define BUG()							\
do {								\
	asm volatile("ud2");					\
	unreachable();						\
} while (0)

But the problem is that the above code is dependent upon both
CONFIG_BUG and CONFIG_DEBUG_BUGVERBOSE.  If both these are unset then
we get the stupid implementation, from include/asm-generic/bug.h:

#ifndef HAVE_ARCH_BUG
#define BUG() do {} while(0)
#endif

ie: the kernel silently locks up.

I think it would be better if x86 were to provide the "ud2" version of
BUG() when CONFIG_BUG=n, as well as when CONFIG_DEBUG_BUGVERBOSE=n. 
That would probably generate less code than the daft
include/asm-generic/bug.h implementation anyway.


As for why you got those warnings: gcc bug, IMO.  If the compiler sees

	int x;

	do { } while (0);

	wibbles(x);

then it should have brains to not tell us that x is used-unintialised.
Sigh.

I'm not particularly enthused about fixing CONFIG_BUG=n stuff.  It's a
nuisance.