* dsa: dsa_slave_port_obj_del calls multiple times with SWITCHDEV_OBJ_ID_HOST_MDB obj id @ 2017-12-06 2:33 Tristram.Ha 2017-12-06 13:56 ` Andrew Lunn 0 siblings, 1 reply; 3+ messages in thread From: Tristram.Ha @ 2017-12-06 2:33 UTC (permalink / raw) To: andrew, f.fainelli; +Cc: netdev, UNGLinuxDriver I found the latest net-next kernel calls dsa_slave_port_obj_del() multiple times, one for each port, with host port as the parameter. As the base driver cannot find an entry with that host port, it returns an error and so users will see a lot of failures from the DSA switch. Is this a new behavior and the driver needs to handle that? In previous versions I do not think I saw that. Typical operation is a PC connected to a port in a switch wants to send multicast packets. It broadcasts an IGMP membership join message. Function dsa_slave_port_obj_add is called to setup an entry in the lookup table. When IGMP membership leave message is received dsa_slave_port_obj_del will be called after a delay. But then it is called for each port with host port as the parameter. Another issue is the host port can setup an entry for IPv6 neighbor discovery like 33:33:FF:??:??:??. When it leaves a failure message will be displayed for lan2, lan3, and so on. It seems the first deletion is coming from lan1. ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: dsa: dsa_slave_port_obj_del calls multiple times with SWITCHDEV_OBJ_ID_HOST_MDB obj id 2017-12-06 2:33 dsa: dsa_slave_port_obj_del calls multiple times with SWITCHDEV_OBJ_ID_HOST_MDB obj id Tristram.Ha @ 2017-12-06 13:56 ` Andrew Lunn 2017-12-06 19:31 ` Tristram.Ha 0 siblings, 1 reply; 3+ messages in thread From: Andrew Lunn @ 2017-12-06 13:56 UTC (permalink / raw) To: Tristram.Ha; +Cc: f.fainelli, netdev, UNGLinuxDriver On Wed, Dec 06, 2017 at 02:33:07AM +0000, Tristram.Ha@microchip.com wrote: > I found the latest net-next kernel calls dsa_slave_port_obj_del() multiple times, > one for each port, with host port as the parameter. Hi Tristram SWITCHDEV_OBJ_ID_HOST_MDB is used, when there is a join/leave on the bridge interface. It happens for each interface in the bridge, and it means, packets which match the group that ingress on that interface should be forwarded to the CPU. > As the base driver cannot find an entry with that host port, it returns an error > and so users will see a lot of failures from the DSA switch. You have a few options: 1) Just forward all multicast traffic to the cpu, and ignore SWITCHDEV_OBJ_ID_HOST_MDB. 2) Implement SWITCHDEV_OBJ_ID_HOST_MDB so you setup your tables to just forward the requested multicast to the cpu. 3) You can also forward a bit too much, e.g. if you cannot set filters per ingress port, just send all the traffic for the group from any port. The bridge will discard whatever it does not need. > Is this a new behavior and the driver needs to handle that? In previous versions > I do not think I saw that. SWITCHDEV_OBJ_ID_HOST_MDB is new. However, dsa_slave_port_obj_del() can be called for all sorts of objects, and you should only be reacting on those your support. So adding a new object should not of changed anything. > Typical operation is a PC connected to a port in a switch wants to send multicast > packets. It broadcasts an IGMP membership join message. Function > dsa_slave_port_obj_add is called to setup an entry in the lookup table. When > IGMP membership leave message is received dsa_slave_port_obj_del will be > called after a delay. But then it is called for each port with host port as the > parameter. Correct. switchdev is a generic API. It also needs to work for Top of Rack switches, which generally have a match/action architecture. I can imagine that this match/action happens per port, so we need to call switchdev per port. However, switches supported by DSA tend to have central management of all ports, so one call would be sufficient. With a DSA driver, you just need to expect redundant calls, and do the right thing. Andrew ^ permalink raw reply [flat|nested] 3+ messages in thread
* RE: dsa: dsa_slave_port_obj_del calls multiple times with SWITCHDEV_OBJ_ID_HOST_MDB obj id 2017-12-06 13:56 ` Andrew Lunn @ 2017-12-06 19:31 ` Tristram.Ha 0 siblings, 0 replies; 3+ messages in thread From: Tristram.Ha @ 2017-12-06 19:31 UTC (permalink / raw) To: andrew; +Cc: f.fainelli, netdev, UNGLinuxDriver > SWITCHDEV_OBJ_ID_HOST_MDB is used, when there is a join/leave on the > bridge interface. It happens for each interface in the bridge, and it > means, packets which match the group that ingress on that interface > should be forwarded to the CPU. > > > As the base driver cannot find an entry with that host port, it returns an > error > > and so users will see a lot of failures from the DSA switch. > > You have a few options: > > 1) Just forward all multicast traffic to the cpu, and ignore > SWITCHDEV_OBJ_ID_HOST_MDB. > > 2) Implement SWITCHDEV_OBJ_ID_HOST_MDB so you setup your tables to > just forward the requested multicast to the cpu. > > 3) You can also forward a bit too much, e.g. if you cannot set filters > per ingress port, just send all the traffic for the group from any > port. > > > The bridge will discard whatever it does not need. > > > Is this a new behavior and the driver needs to handle that? In previous > versions > > I do not think I saw that. > > SWITCHDEV_OBJ_ID_HOST_MDB is new. However, > dsa_slave_port_obj_del() > can be called for all sorts of objects, and you should only be > reacting on those your support. So adding a new object should not of > changed anything. > > > Typical operation is a PC connected to a port in a switch wants to send > multicast > > packets. It broadcasts an IGMP membership join message. Function > > dsa_slave_port_obj_add is called to setup an entry in the lookup table. > When > > IGMP membership leave message is received dsa_slave_port_obj_del will > be > > called after a delay. But then it is called for each port with host port as the > > parameter. > > Correct. switchdev is a generic API. It also needs to work for Top of > Rack switches, which generally have a match/action architecture. I can > imagine that this match/action happens per port, so we need to call > switchdev per port. However, switches supported by DSA tend to have > central management of all ports, so one call would be sufficient. With > a DSA driver, you just need to expect redundant calls, and do the > right thing. The base DSA driver only knows about port_mdb_add and port_mdb_del. It does not really know about SWITCHDEV_OBJ_ID_HOST_MDB. It is fine to use port_mdb_add and port_mdb_del for SWITCHDEV_OBJ_ID_HOST_MDB as hardware can program an entry for the host port, but receiving many port_mdb_del calls with the same parameter makes it difficult to handle. Example: port_mdb_add 2 xx:xx:xx:xx:xx:xx port_mdb_del 2 xx:xx:xx:xx:xx:xx port_mdb_del 4 xx:xx:xx:xx:xx:xx; lan1 port_mdb_del 4 xx:xx:xx:xx:xx:xx; lan2 port_mdb_del 4 xx:xx:xx:xx:xx:xx; lan3 There is no indication in the port_mdb_del call to show this call is for port 1, 2, and so on. port_mdb_add can be used to add an entry for the host. Again it is called as many times as number of ports. I think the port_mdb_* call needs an extra parameter to make it useful in this situation. An easy fix is not to return any error in port_mdb_del, or ignore host port. It seems port_mdb_prepare is used to indicate an entry can be added before the actual port_mdb_add call, which returns void and so cannot indicate failure to add. Currently there is no implementation for that port_mdb_prepare call and the Marvell driver just displays a warning inside its port_mdb_add function. ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2017-12-06 19:31 UTC | newest] Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2017-12-06 2:33 dsa: dsa_slave_port_obj_del calls multiple times with SWITCHDEV_OBJ_ID_HOST_MDB obj id Tristram.Ha 2017-12-06 13:56 ` Andrew Lunn 2017-12-06 19:31 ` Tristram.Ha
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.