Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Kuniyuki Iwashima <kuniyu@amazon.com>
To:     <leitao@debian.org>
CC:     <alexander@mihalicyn.com>, <ast@kernel.org>, <davem@davemloft.net>,
        <dhowells@redhat.com>, <edumazet@google.com>,
        <kernelxing@tencent.com>, <kuba@kernel.org>, <kuniyu@amazon.com>,
        <leit@meta.com>, <linux-kernel@vger.kernel.org>,
        <lucien.xin@gmail.com>, <martin.lau@kernel.org>,
        <netdev@vger.kernel.org>, <pabeni@redhat.com>
Subject: Re: [PATCH net-next] net: Use _K_SS_MAXSIZE instead of absolute value
Date:   Wed, 19 Jul 2023 10:30:17 -0700
Message-ID: <20230719173017.33951-1-kuniyu@amazon.com>
In-Reply-To: <ZLga+cBUKkN5Fnn7@gmail.com>
References: <ZLga+cBUKkN5Fnn7@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain
Precedence: bulk

From: Breno Leitao <leitao@debian.org>
Date: Wed, 19 Jul 2023 10:18:49 -0700
> On Wed, Jul 19, 2023 at 10:04:45AM -0700, Kuniyuki Iwashima wrote:
> > From: Breno Leitao <leitao@debian.org>
> > Date: Wed, 19 Jul 2023 01:44:12 -0700
> > > Looking at sk_getsockopt function, it is unclear why 128 is a magical
> > > number.
> > > 
> > > Use the proper macro, so it becomes clear to understand what the value
> > > mean, and get a reference where it is coming from (user-exported API).
> > > 
> > > Signed-off-by: Breno Leitao <leitao@debian.org>
> > > ---
> > >  net/core/sock.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/net/core/sock.c b/net/core/sock.c
> > > index 9370fd50aa2c..58b6f00197d6 100644
> > > --- a/net/core/sock.c
> > > +++ b/net/core/sock.c
> > > @@ -1815,7 +1815,7 @@ int sk_getsockopt(struct sock *sk, int level, int optname,
> > >  
> > >  	case SO_PEERNAME:
> > >  	{
> > > -		char address[128];
> > > +		char address[_K_SS_MAXSIZE];
> > 
> > I guess you saw a bug caught by the fortified memcpy(), but this
> > doesn't fix it properly.
> 
> Not really, in fact. I was reading this code, and I found this
> discussion a while ago, where I got the idea:
> 
> https://lore.kernel.org/lkml/20140930.005925.995989898229686123.davem@davemloft.net/

I got it, but I prefer using struct sockaddr_storage as done in
other places.

  $ grep -rn sockaddr_storage net/

Also, there would be some situations where we must cast each
family-specific address back to sockaddr_storage for fortified
library.

Then, it makes more sense to use sockaddr_storage rather than
_K_SS_MAXSIZE.