netdev - Re: [PATCH] netpoll: cleanup sparse warnings

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Fri, 8 Feb 2013 10:47:27 -0500
From:	Neil Horman <nhorman@...driver.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	netdev@...r.kernel.org, fengguang.wu@...el.com,
	David Miller <davem@...emloft.net>
Subject: Re: [PATCH] netpoll: cleanup sparse warnings

On Thu, Feb 07, 2013 at 07:25:33PM -0800, Eric Dumazet wrote:
> On Thu, 2013-02-07 at 13:37 -0500, Neil Horman wrote:
> > On Thu, Feb 07, 2013 at 07:52:56AM -0800, Eric Dumazet wrote:
> > > On Thu, 2013-02-07 at 09:56 -0500, Neil Horman wrote:
> > > > With my recent commit I introduced two sparse warnings.  Looking closer there
> > > > were a few more in the same file, so I fixed them all up.  Basic rcu pointer
> > > > dereferencing suff
> > > 
> > > > -	npinfo = np->dev->npinfo;
> > > > +	/* rtnl_dereference would be preferable here but
> > > > +	 * rcu_cleanup_netpoll path can put us in here safely without
> > > > +	 * holding the rtnl, so plain rcu_dereference it is
> > > > +	 */
> > > > +	npinfo = rcu_dereference(np->dev->npinfo);
> > > >  	if (!npinfo)
> > > >  		return;
> > > >  
> > > 
> > > Are you sure it wont trigger a LOCKDEP complain (CONFIG_PROVE_RCU=y) ?
> > > 
> > Hm, looking at it, you're probably right.  We're not holding the rcu_read_lock,
> > and I'd forgotten that rcu_dereference implicitly checks that rcu_read_lock is
> > held.  I guess, since the only paths that we get here on are in a bh rcu
> > quiescence point or with the rtnl held we should probably make this:
> > 
> > rcu_dereference_protected(np->dev->npinfo, rtnl_locked() || in_interrupt());
> > 
> > Although, thinking about this further somewhat begs the question as to how we
> > prevent one context from calling __netpoll_cleanup in a path holding rtnl, while
> > in parallel calling __netpoll_cleanup from the rcu callback.  That might not be
> > a huge deal as __netpoll_cleanup uses spinlocks to do list modification, and an
> > atomic_dec_and_test to gate the free, but it still seems ugly.
> > 
> > What do you think?
> 
> I think you could use 
> 
>  rcu_dereference_check(p, lockdep_rtnl_is_held() || something)
> 
Actually, I think all we need to do is take the rcu_read_lock and use
rcu_dereference directly.  __netpoll_cleanup doesn't actually change the
dev->npinfo pointer itself, thats handled by the caller, always under protection
of the rtnl, so here we just need to ensure that no one changes npifo while
we're using it
Neil

> 
> 
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html