From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Ryan O'Hara" <rohara@redhat.com>
Subject: Re: [PATCH] ipvsadm: Fix svc->pe_name conditional
Date: Mon, 19 Aug 2013 15:40:44 -0500
Message-ID: <20130819204043.GB14857@redhat.com>
References: <1376844063-4298-1-git-send-email-rohara@redhat.com>
 <alpine.LFD.2.00.1308182035550.3222@ja.ssi.bg>
 <20130819145217.GA12716@redhat.com>
 <alpine.LFD.2.00.1308191817200.1606@ja.ssi.bg>
Mime-Version: 1.0
Return-path: <lvs-devel-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <alpine.LFD.2.00.1308191817200.1606@ja.ssi.bg>
Sender: lvs-devel-owner@vger.kernel.org
List-ID: <lvs-devel.vger.kernel.org>
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: Julian Anastasov <ja@ssi.bg>
Cc: Simon Horman <horms@verge.net.au>, lvs-devel@vger.kernel.org

On Mon, Aug 19, 2013 at 06:27:21PM +0300, Julian Anastasov wrote:
> 
> 	Hello,
> 
> On Mon, 19 Aug 2013, Ryan O'Hara wrote:
> 
> > On Sun, Aug 18, 2013 at 09:13:39PM +0300, Julian Anastasov wrote:
> > 
> > > - CHECK_PE and CHECK_IPV4 as part of CHECK_COMPAT_SVC use
> > > 'return' and we can leak svc in ipvs_get_service()
> > >
> > > - in ipvs_get_service() CHECK_PE does not need to be called,
> > > it is already part of CHECK_COMPAT_SVC
> > 
> > Yes. I'm tempted to remove these macros due to the 'return'
> > issue. Thoughts?
> 
> 	May be they can use 'goto' instead of 'return':
> 
> 	if (condition) {
> 		errno = XXX;
> 		goto out_err;
> 	}
> 
> 	and funcs can have such exit point:
> 
> out_err:
> 	free stuff on error
> 	return ret;

OK. It looks like ipvs_get_service() is the only function with a
potential leak. The other functions that call any of the CHECK_*
macros will just get an 'out_err: return -1'.

I also noticed that CHECK_COMPAT_DEST can be removed. This macro
simply calls CHECK_IPV4, which is also called by CHECK_COMPAT_SVC. I
noticed that CHECK_COMPAT_DEST is only called after CHECK_COMPAT_SVC,
so this macro can be removed entirely. Agree?

> > > - ipvs_get_service uses malloc but later calls CHECK_PE, may be
> > > malloc should be changed with calloc
> > 
> > Aside from the fact that CHECK_PE could return without freeing memory,
> > I don't see the problem here.
> 
> 	malloc does not initialize the memory and checking of
> svc->pe_name[0] in CHECK_PE would give random results.

I see. Thanks.

Ryan