commit 381c759d9916 ("ipv4: Avoid crashing in ip_error")
fixes a problem where processed packet comes from device
with destroyed inetdev (dev->ip_ptr). This is not expected
because inetdev_destroy is called in NETDEV_UNREGISTER
phase and packets should not be processed after
dev_close_many() and synchronize_net(). But backlog
processing is deactivated later after NETDEV_UNREGISTER_FINAL
phase which allows CPUs to initiate processing for new
packets during and after the NETDEV_UNREGISTER phase.

Fix it by moving flush_backlog after the device driver
is stopped and before the synchronize_net() call. This
allows dev->ip_ptr to be always valid during packet
processing.

Reported-by: Vittorio Gambaletta <linuxb...@vittgam.net>
Fixes: 6e583ce5242f ("net: eliminate refcounting in backlog queue")
Signed-off-by: Julian Anastasov <j...@ssi.bg>
---
 net/core/dev.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

Tested by reverting commit 381c759d9916 and using the
test commands provided by Eric Biederman:

http://marc.info/?l=linux-netdev&m=143239231905533&w=2

Problem does not occur every time. SMP may be needed
to reproduce.

diff --git a/net/core/dev.c b/net/core/dev.c
index aa82f9a..67132b3 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6027,6 +6027,7 @@ static void rollback_registered_many(struct list_head 
*head)
                unlist_netdevice(dev);
 
                dev->reg_state = NETREG_UNREGISTERING;
+               on_each_cpu(flush_backlog, dev, 1);
        }
 
        synchronize_net();
@@ -6650,8 +6651,6 @@ void netdev_run_todo(void)
 
                dev->reg_state = NETREG_UNREGISTERED;
 
-               on_each_cpu(flush_backlog, dev, 1);
-
                netdev_wait_allrefs(dev);
 
                /* paranoia */
-- 
1.9.3

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

Reply via email to