Tue, Oct 13, 2015 at 05:28:20AM CEST, sfel...@gmail.com wrote:
>On Mon, Oct 12, 2015 at 11:03 AM, Jiri Pirko <j...@resnulli.us> wrote:
>> From: Jiri Pirko <j...@mellanox.com>
>>
>> Since spinlock is held here, defer the switchdev operation.
>>
>> Signed-off-by: Jiri Pirko <j...@mellanox.com>
>> ---
>> net/bridge/br_fdb.c | 5 ++++-
>> net/bridge/br_if.c | 3 +++
>> 2 files changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
>> index f5e7da0..c88bd8e 100644
>> --- a/net/bridge/br_fdb.c
>> +++ b/net/bridge/br_fdb.c
>> @@ -134,7 +134,10 @@ static void fdb_del_hw_addr(struct net_bridge *br,
>> const unsigned char *addr)
>> static void fdb_del_external_learn(struct net_bridge_fdb_entry *f)
>> {
>> struct switchdev_obj_port_fdb fdb = {
>> - .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
>> + .obj = {
>> + .id = SWITCHDEV_OBJ_ID_PORT_FDB,
>> + .flags = SWITCHDEV_F_DEFER,
>> + },
>> .vid = f->vlan_id,
>> };
>>
>> diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
>> index 934cae9..09147cb 100644
>> --- a/net/bridge/br_if.c
>> +++ b/net/bridge/br_if.c
>> @@ -24,6 +24,7 @@
>> #include <linux/slab.h>
>> #include <net/sock.h>
>> #include <linux/if_vlan.h>
>> +#include <net/switchdev.h>
>>
>> #include "br_private.h"
>>
>> @@ -249,6 +250,8 @@ static void del_nbp(struct net_bridge_port *p)
>> list_del_rcu(&p->list);
>>
>> br_fdb_delete_by_port(br, p, 0, 1);
>> + switchdev_flush_deferred();
>> +
>
>This potentially flushes other (valid) work on the deferred queue not
>related to FDB del.
flush_deferred is implemented by flush_workqueue. This enforces the
workqueue to be processed heve and sleeps until it is done. Nothing is
lost.
>
>I wonder if this flush step is necessary at all? The work we deferred
>to delete the FDB entry can still happen after the port has been
>removed (del_nbp). If the port driver/device find the FDB entry, then
>delete it, otherwise ignore it.
This step is necessary. Consider following scenario:
spin_lock
switchdev_schedule_deferred (workx)
spin_unlock
netdev_upper_dev_unlink
workx is executed-> driver op is executed.
But during netdev_upper_dev_unlink, resources may have been removed, so
the workx after that would fail (this happens in case of rocker).
So I added switchdev_flush_deferred so the caller may force to safely
process the pending work in right time, for example after he released
the spinlock (after br_fdb_delete_by_port in this case).
--
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
