[PATCH 12/19] NET: set PF_FSTRANS while holding rtnl_lock

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



As rtnl_mutex can be taken while holding sk_lock, and sk_lock can be
taken while performing memory reclaim (at least when loop-back NFS is
active), any memory allocation under rtnl_mutex must avoid __GFP_FS,
which is most easily done by setting PF_MEMALLOC.


        CPU0                    CPU1
        ----                    ----
   lock(rtnl_mutex);
                                lock(sk_lock-AF_INET);
                                lock(rtnl_mutex);
   <Memory allocation/reclaim>
     lock(sk_lock-AF_INET);

  *** DEADLOCK ***

1/ rtnl_mutex is taken while holding sk_lock:

    [<ffffffff81abb442>] rtnl_lock+0x12/0x20
    [<ffffffff81b28c3a>] ip_mc_leave_group+0x2a/0x160
    [<ffffffff81aec70b>] do_ip_setsockopt.isra.18+0x96b/0xed0
    [<ffffffff81aecc97>] ip_setsockopt+0x27/0x90
    [<ffffffff81b151c6>] udp_setsockopt+0x16/0x30
    [<ffffffff81a9144f>] sock_common_setsockopt+0xf/0x20
    [<ffffffff81a907de>] SyS_setsockopt+0x5e/0xc0

2/ memory is allocated under rtnl_mutex:
    [<ffffffff8166eb41>] kobject_set_name_vargs+0x21/0x70
    [<ffffffff81840d92>] dev_set_name+0x42/0x50
    [<ffffffff81ac5e97>] netdev_register_kobject+0x57/0x130
    [<ffffffff81aaf574>] register_netdevice+0x354/0x550
    [<ffffffff81aaf785>] register_netdev+0x15/0x30


Signed-off-by: NeilBrown <neilb@xxxxxxx>
---
 net/core/rtnetlink.c |    9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 120eecc0f5a4..6870211e93a6 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -61,15 +61,18 @@ struct rtnl_link {
 };
 
 static DEFINE_MUTEX(rtnl_mutex);
+static int rtnl_pflags;
 
 void rtnl_lock(void)
 {
 	mutex_lock(&rtnl_mutex);
+	current_set_flags_nested(&rtnl_pflags, PF_FSTRANS);
 }
 EXPORT_SYMBOL(rtnl_lock);
 
 void __rtnl_unlock(void)
 {
+	current_restore_flags_nested(&rtnl_pflags, PF_FSTRANS);
 	mutex_unlock(&rtnl_mutex);
 }
 
@@ -82,7 +85,11 @@ EXPORT_SYMBOL(rtnl_unlock);
 
 int rtnl_trylock(void)
 {
-	return mutex_trylock(&rtnl_mutex);
+	if (mutex_trylock(&rtnl_mutex)) {
+		current_set_flags_nested(&rtnl_pflags, PF_FSTRANS);
+		return 1;
+	}
+	return 0;
 }
 EXPORT_SYMBOL(rtnl_trylock);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]