linux-kernel - Re: [PATCH] net: mac80211: rx.c: Use built-in RCU list checking

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:   Sat, 22 Feb 2020 19:09:28 +0530
From:   Madhuparna Bhowmik <madhuparnabhowmik10@...il.com>
To:     Johannes Berg <johannes@...solutions.net>
Cc:     madhuparnabhowmik10@...il.com, davem@...emloft.net,
        linux-wireless@...r.kernel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, joel@...lfernandes.org,
        frextrite@...il.com,
        linux-kernel-mentees@...ts.linuxfoundation.org, paulmck@...nel.org
Subject: Re: [PATCH] net: mac80211: rx.c: Use built-in RCU list checking

On Sat, Feb 22, 2020 at 01:53:25PM +0100, Johannes Berg wrote:
> On Sat, 2020-02-22 at 15:48 +0530, madhuparnabhowmik10@...il.com wrote:
> > From: Madhuparna Bhowmik <madhuparnabhowmik10@...il.com>
> > 
> > list_for_each_entry_rcu() has built-in RCU and lock checking.
> > 
> > Pass cond argument to list_for_each_entry_rcu() to silence
> > false lockdep warning when CONFIG_PROVE_RCU_LIST is enabled
> > by default.
> 
> Umm. What warning?
>
If list_for_each_entry_rcu() is called from non rcu protection
i.e without holding rcu_read_lock, but under the protection of
a different lock then we can pass that as the condition for lockdep checking
because otherwise lockdep will complain if list_for_each_entry_rcu()
is used without rcu protection. So, if we do not pass this argument
(cond) it may lead to false lockdep warnings.

> > +++ b/net/mac80211/rx.c
> > @@ -3547,7 +3547,8 @@ static void ieee80211_rx_cooked_monitor(struct ieee80211_rx_data *rx,
> >  	skb->pkt_type = PACKET_OTHERHOST;
> >  	skb->protocol = htons(ETH_P_802_2);
> >  
> > -	list_for_each_entry_rcu(sdata, &local->interfaces, list) {
> > +	list_for_each_entry_rcu(sdata, &local->interfaces, list,
> > +				lockdep_is_held(&rx->local->rx_path_lock)) {
> >  		if (!ieee80211_sdata_running(sdata))
> >  			continue;
> 
> This is not related at all.

I analysed the following traces:
ieee80211_rx_handlers() -> ieee80211_rx_handlers_result() -> ieee80211_rx_cooked_monitor()

here ieee80211_rx_handlers() is holding the rx->local->rx_path_lock and
therefore I used this for the cond argument.

 If this is not right, can you help me in figuring out that which other
 lock is held?

and 
__ieee80211_rx_handle_packet() -> ieee80211_prepare_and_rx_handle() -> ieee80211_invoke_rx_handlers() -> 
ieee80211_rx_handlers_result() -> ieee80211_rx_cooked_monitor()

Here __ieee80211_rx_handle_packet() should be called under
rcu_read_lock protection.
So this trace seems okay and no need to pass any cond.

I may have missed something, please correct me in that case.

> > @@ -4114,7 +4115,8 @@ void __ieee80211_check_fast_rx_iface(struct ieee80211_sub_if_data *sdata)
> >  
> >  	lockdep_assert_held(&local->sta_mtx);
> >  
> > -	list_for_each_entry_rcu(sta, &local->sta_list, list) {
> > +	list_for_each_entry_rcu(sta, &local->sta_list, list,
> > +				lockdep_is_held(&local->sta_mtx)) {
> 
> And this isn't even a real RCU iteration, since we _must_ hold the mutex
> here.
>
Yeah exactly, dropping _rcu (use list_for_each_entry()) would be a good option in this case.
Let me know if that is alright and I will send a new patch with all the
changes required.

Thank you,
Madhuparna

> johannes
>