Received-SPF: pass (google.com: domain of linux-wireless+bounces-197-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Message-ID: <tencent_6E9CA7EF46E452C90650899D2ECFEEFE4C06@qq.com>
From: Edward Adam Davis <eadavis@qq.com>
To: johannes@sipsolutions.net
Cc: davem@davemloft.net,
	eadavis@qq.com,
	edumazet@google.com,
	kuba@kernel.org,
	linux-kernel@vger.kernel.org,
	linux-wireless@vger.kernel.org,
	llvm@lists.linux.dev,
	nathan@kernel.org,
	ndesaulniers@google.com,
	netdev@vger.kernel.org,
	pabeni@redhat.com,
	syzbot+62d7eef57b09bfebcd84@syzkaller.appspotmail.com,
	syzkaller-bugs@googlegroups.com,
	trix@redhat.com
Subject: Re: [PATCH] wifi: mac80211: sband's null check should precede params
Date: Wed, 29 Nov 2023 16:48:46 +0800
In-Reply-To: <56d8b3b9099d3935a7b53e976fa998f06fbfd9a5.camel@sipsolutions.net>
References: <56d8b3b9099d3935a7b53e976fa998f06fbfd9a5.camel@sipsolutions.net>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

On Wed, 29 Nov 2023 09:33:23 +0100, Johannes Berg wrote:
> > > > [Analysis]
> > > > When ieee80211_get_link_sband() fails to find a valid sband and first checks
> > > > for params in sta_link_apply_parameters(), it will return 0 due to new_link
> > > > being 0, which will lead to an incorrect process after sta_apply_parameters().
> > > > 
> > > > [Fix]
> > > > First obtain sband and perform a non null check before checking the params.
> > > 
> > > Not sure I can even disagree with that analysis, it seems right, but ...
> > > 
> > > > +	if (!link || !link_sta)
> > > > +		return -EINVAL;
> > > > +
> > > > +	sband = ieee80211_get_link_sband(link);
> > > > +	if (!sband)
> > > > +		return -EINVAL;
> > > > +
> > > >  	/*
> > > >  	 * If there are no changes, then accept a link that doesn't exist,
> > > >  	 * unless it's a new link.
> > > 
> > > There's a comment here which is clearly not true after this change,
> > > since you've already returned for !link_sta?
> > No, after applying my patch, it will return due to !sband.
> > 
> 
> Right, OK, but the way I read the comment (now) is that it wanted to
> accept it in that case?
> 
> That said, I just threw the patch into our internal testing machinery
> quickly (probably has more MLO tests than upstream hostap for now), and
> it worked just fine ...
> 
> Maybe we should just remove the comment?
Do you mean to delete the comments below?
   3         /*
   2          * If there are no changes, then accept a link that doesn't exist,
   1          * unless it's a new link.
1800          */

Edward