[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130521125856.GH3578@linux.vnet.ibm.com>
Date: Tue, 21 May 2013 05:58:56 -0700
From: "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To: Roman Gushchin <klamm@...dex-team.ru>
Cc: Dipankar Sarma <dipankar@...ibm.com>, zhmurov@...dex-team.ru,
linux-kernel@...r.kernel.org, netdev@...r.kernel.org,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <eric.dumazet@...il.com>,
Alexey Kuznetsov <kuznet@....inr.ac.ru>,
James Morris <jmorris@...ei.org>,
Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>,
Patrick McHardy <kaber@...sh.net>
Subject: Re: [PATCH] rcu: fix a race in hlist_nulls_for_each_entry_rcu macro
On Tue, May 21, 2013 at 04:46:54PM +0400, Roman Gushchin wrote:
> On 21.05.2013 16:09, Paul E. McKenney wrote:
> >On Tue, May 21, 2013 at 01:05:48PM +0400, Roman Gushchin wrote:
> >>Hi, all!
> >>
> >>This is a fix for a problem described here:
> >>https://lkml.org/lkml/2013/4/16/371 .
> >>---
> >>
> >>Some network functions (udp4_lib_lookup2(), for instance) use the
> >>hlist_nulls_for_each_entry_rcu macro in a way that assumes restarting
> >>of a loop. In this case, it is strictly necessary to reread the head->first
> >>value from the memory before each scan.
> >>Without additional hints, gcc caches this value in a register. In this case,
> >>if a cached node is moved to another chain during the scan, we can loop
> >>forever getting wrong nulls values and restarting the loop uninterruptedly.
> >>
> >>Signed-off-by: Roman Gushchin <klamm@...dex-team.ru>
> >>Reported-by: Boris Zhmurov <zhmurov@...dex-team.ru>
> >>---
> >> include/linux/rculist_nulls.h | 5 +++--
> >> 1 file changed, 3 insertions(+), 2 deletions(-)
> >>
> >>diff --git a/include/linux/rculist_nulls.h b/include/linux/rculist_nulls.h
> >>index 2ae1371..efd51bf 100644
> >>--- a/include/linux/rculist_nulls.h
> >>+++ b/include/linux/rculist_nulls.h
> >>@@ -37,8 +37,9 @@ static inline void hlist_nulls_del_init_rcu(struct
> >>hlist_nulls_node *n)
> >> }
> >> }
> >>
> >>-#define hlist_nulls_first_rcu(head) \
> >>- (*((struct hlist_nulls_node __rcu __force **)&(head)->first))
> >>+#define hlist_nulls_first_rcu(head) \
> >>+ (*((struct hlist_nulls_node __rcu __force **) \
> >>+ &((volatile typeof(*head) *)head)->first))
> >
> >Why not use ACCESS_ONCE() or (better) rcu_dereference_raw() here?
>
> It will be nice, but will require to keep the old variant too (for
> using in hlist_nulls_add_head_rcu() as in rcu_assign_pointer()
> argument). Do you think, it's better?
Both ACCESS_ONCE() and rcu_dereference_raw() can be used by updaters
as well as readers, so yes, I do think that it is better. Better to
keep the encapsulation rather than having to search for lots of volatile
casts should this idiom ever need to change.
Thanx, Paul
--
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
Powered by blists - more mailing lists