Re: [Linux-kernel-mentees] [PATCH net 2/2 RESEND] ipmr: Add lockdep expression to ipmr_for_each_table macro

From: Jakub Kicinski <kuba@kernel.org>
To: Amol Grover <frextrite@gmail.com>
Cc: "Paul E . McKenney" <paulmck@kernel.org>,
	Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
	Madhuparna Bhowmik <madhuparnabhowmik10@gmail.com>,
	Joel Fernandes <joel@joelfernandes.org>,
	Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>,
	linux-kernel-mentees@lists.linuxfoundation.org,
	"David S . Miller" <davem@davemloft.net>
Subject: Re: [Linux-kernel-mentees] [PATCH net 2/2 RESEND] ipmr: Add lockdep expression to ipmr_for_each_table macro
Date: Sat, 9 May 2020 14:19:38 -0700	[thread overview]
Message-ID: <20200509141938.028fa959@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com> (raw)
In-Reply-To: <20200509072243.3141-2-frextrite@gmail.com>

On Sat,  9 May 2020 12:52:44 +0530 Amol Grover wrote:
> ipmr_for_each_table() uses list_for_each_entry_rcu() for
> traversing outside of an RCU read-side critical section but
> under the protection of pernet_ops_rwsem. Hence add the
> corresponding lockdep expression to silence the following
> false-positive warning at boot:

Thanks for the fix, the warning has been annoying me as well!

> [    0.645292] =============================
> [    0.645294] WARNING: suspicious RCU usage
> [    0.645296] 5.5.4-stable #17 Not tainted
> [    0.645297] -----------------------------
> [    0.645299] net/ipv4/ipmr.c:136 RCU-list traversed in non-reader section!!

please provide a fuller stack trace, it would have helped the review

> Signed-off-by: Amol Grover <frextrite@gmail.com>
> ---
>  net/ipv4/ipmr.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
> index 99c864eb6e34..950ffe9943da 100644
> --- a/net/ipv4/ipmr.c
> +++ b/net/ipv4/ipmr.c
> @@ -109,9 +109,10 @@ static void mroute_clean_tables(struct mr_table *mrt, int flags);
>  static void ipmr_expire_process(struct timer_list *t);
>  
>  #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
> -#define ipmr_for_each_table(mrt, net) \
> -	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list, \
> -				lockdep_rtnl_is_held())
> +#define ipmr_for_each_table(mrt, net)					\
> +	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,	\
> +				lockdep_rtnl_is_held() ||		\
> +				lockdep_is_held(&pernet_ops_rwsem))

This is a strange condition, IMHO. How can we be fine with either
lock.. This is supposed to be the writer side lock, one can't have 
two writer side locks..

I think what is happening is this:

ipmr_net_init() -> ipmr_rules_init() -> ipmr_new_table()

ipmr_new_table() returns an existing table if there is one, but
obviously none can exist at init.  So a better fix would be:

#define ipmr_for_each_table(mrt, net)					\
	list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list,	\
				lockdep_rtnl_is_held() ||		\
				list_empty(&net->ipv4.mr_tables))

Thoughts?
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees