search for: switchdev

To be able to add dynamic FDB entries to drivers from userspace, the dynamic flag must be added when sending RTM_NEWNEIGH events down. Signed-off-by: Hans J. Schultz <netdev at kapio-technology.com> --- include/net/switchdev.h | 1 + net/bridge/br_switchdev.c | 1 + 2 files changed, 2 insertions(+) diff --git a/include/net/switchdev.h b/include/net/switchdev.h index ca0312b78294..aaf918d4ba67 100644 --- a/include/net/switchdev.h +++ b/include/net/switchdev.h @@ -249,6 +249,7 @@ struct switchdev_notifier_fdb_info {...

...23 at 07:57:10PM +0100, Hans J. Schultz wrote: > To be able to add dynamic FDB entries to drivers from userspace, the > dynamic flag must be added when sending RTM_NEWNEIGH events down. > > Signed-off-by: Hans J. Schultz <netdev at kapio-technology.com> > --- > include/net/switchdev.h | 1 + > net/bridge/br_switchdev.c | 1 + > 2 files changed, 2 insertions(+) > > diff --git a/include/net/switchdev.h b/include/net/switchdev.h > index ca0312b78294..aaf918d4ba67 100644 > --- a/include/net/switchdev.h > +++ b/include/net/switchdev.h > @@ -249,6 +249,7 @...

...Hans J. Schultz wrote: >> To be able to add dynamic FDB entries to drivers from userspace, the >> dynamic flag must be added when sending RTM_NEWNEIGH events down. >> >> Signed-off-by: Hans J. Schultz <netdev at kapio-technology.com> >> --- >> include/net/switchdev.h | 1 + >> net/bridge/br_switchdev.c | 1 + >> 2 files changed, 2 insertions(+) >> >> diff --git a/include/net/switchdev.h b/include/net/switchdev.h >> index ca0312b78294..aaf918d4ba67 100644 >> --- a/include/net/switchdev.h >> +++ b/include/net/switchd...

There is a structural problem in switchdev, where the flag bits in struct switchdev_notifier_fdb_info (added_by_user, is_local etc) only represent a simplified / denatured view of what's in struct net_bridge_fdb_entry :: flags (BR_FDB_ADDED_BY_USER, BR_FDB_LOCAL etc). Each time we want to pass more information about struct net_bridge_fd...

On Mon, Apr 10, 2023 at 11:49:51PM +0300, Vladimir Oltean wrote: > There is a structural problem in switchdev, where the flag bits in > struct switchdev_notifier_fdb_info (added_by_user, is_local etc) only > represent a simplified / denatured view of what's in struct > net_bridge_fdb_entry :: flags (BR_FDB_ADDED_BY_USER, BR_FDB_LOCAL etc). > Each time we want to pass more information about...

On Wed, Apr 12, 2023 at 07:24:07PM +0300, Vladimir Oltean wrote: > I'll send v2 with BR_FDB_ADDED_BY_EXT_LEARN not prevented from being > notified from switchdev. > > Unless you have any objection, I won't send v2 like this: > > if (test_bit(BR_FDB_ADDED_BY_USER, &fdb->flags) && > !test_bit(BR_FDB_STATIC, &fdb->flags) && > !test_bit(BR_FDB_ADDED_BY_EXT_LEARN, &fdb->flags)) > return...

On Wed, Apr 12, 2023 at 07:00:02PM +0300, Ido Schimmel wrote: > On Wed, Apr 12, 2023 at 05:27:33PM +0300, Vladimir Oltean wrote: > > How are extern_learn FDB entries processed by spectrum's > > SWITCHDEV_FDB_ADD_TO_DEVICE handler? > > No different than "BR_FDB_STATIC", which is a bug I'm aware of and > intend to fix in net-next when I get the time (together with all the > other combinations enabled by the bridge). Entry has ageing disabled, > but can roam in which case...

...: > > which idea is that, again? So I cannot us the offloaded flag as it is added by DSA in the common case when using 'bridge fdb replace ... dynamic'. The idea is then to use the ext_learn flag instead, which is not aged by the bridge. To do this the driver (mv88e6xxx) will send a SWITCHDEV_FDB_ADD_TO_BRIDGE switchdev event when the new dynamic flag is true. The function sending this event will then be named mv88e6xxx_add_fdb_synth_learned() in drivers/net/dsa/mv88e6xxx/switchdev.c, replacing the mv88e6xxx_set_fdb_offloaded() function but in most part the same content, just another ev...

...y not? I find it reasonable that the software bridge does not age out a dynamic FDB entry that is offloaded to hardware... the hardware should do that ("dynamic" being the key). At least, I find it more reasonable than the current behavior, where the bridge notifies dynamic FDB entries to switchdev, but doesn't say they're dynamic, and switchdev treats them as static, so they don't roam from one bridge port to another until software sees a packet with that MAC DA, and they have the potential of blocking traffic because of that. If for some reason you do think that behavior is use...

...eem to suggest otherwise. > > If you have a suggestion, feel free. Didn't I explain what I would do from the first reply on this thread? https://patchwork.kernel.org/project/netdevbpf/patch/20230318141010.513424-3-netdev at kapio-technology.com/#25270613 As a bug fix, stop reporting to switchdev those FDB entries with BR_FDB_ADDED_BY_USER && !BR_FDB_STATIC. Then, after "net" is merged into "net-next" next Thursday (the ship has sailed for today), add "bool static" to the switchdev notifier info, and make all switchdev drivers (everywhere where a SWITCH...

On Wed, Apr 12, 2023 at 05:27:33PM +0300, Vladimir Oltean wrote: > How are extern_learn FDB entries processed by spectrum's > SWITCHDEV_FDB_ADD_TO_DEVICE handler? No different than "BR_FDB_STATIC", which is a bug I'm aware of and intend to fix in net-next when I get the time (together with all the other combinations enabled by the bridge). Entry has ageing disabled, but can roam in which case it becomes age-able. TB...

...>flags); > > > > Why reverse logic? Why not just name this "is_static" and leave any > > further interpretations up to the consumer? > > My reasoning for this is that the common case is to have static entries, > thus is_dyn=false, so whenever someone uses a switchdev_notifier_fdb_info > struct the common case does not need to be entered. > Otherwise it might also break something when someone uses this struct and if > it was 'is_static' and they forget to code is_static=true they will get > dynamic entries without wanting it and it can be har...

...f you post a v2 that > takes "BR_FDB_ADDED_BY_EXT_LEARN" into account, then I can ask Petr to > run it through our regression and report back (not sure we will make it > to this week's PR though). > > Thanks How are extern_learn FDB entries processed by spectrum's SWITCHDEV_FDB_ADD_TO_DEVICE handler?

...t; > > Why reverse logic? Why not just name this "is_static" and leave any > > > further interpretations up to the consumer? > > > > My reasoning for this is that the common case is to have static entries, > > thus is_dyn=false, so whenever someone uses a switchdev_notifier_fdb_info > > struct the common case does not need to be entered. > > Otherwise it might also break something when someone uses this struct and if > > it was 'is_static' and they forget to code is_static=true they will get > > dynamic entries without wanting...

...2:59:04PM +0200, Hans Schultz wrote: >> On Thu, Mar 30, 2023 at 15:43, Vladimir Oltean <olteanv at gmail.com> wrote: >> > On Tue, Mar 28, 2023 at 09:45:26PM +0200, Hans Schultz wrote: >> >> So the solution would be to not let the DSA layer send the >> >> SWITCHDEV_FDB_OFFLOADED event in the case when the new dynamic flag is >> >> set? >> > >> > I have never said that. >> >> No, I was just thinking of a solution based on your previous comment >> that dynamic fdb entries with the offloaded flag set should not be...

...state is not forwarding. 4) SKB_DROP_REASON_BRIDGE_NOT_ALLOWED_EGRESS: the packet is not allowed to go out through the port due to vlan filtering. 5) SKB_DROP_REASON_BRIDGE_SWDEV_NOT_ALLOWED_EGRESS: the packet is not allowed to go out through the port which is offloaded by a hardware switchdev, checked by nbp_switchdev_allowed_egress(). 6) SKB_DROP_REASON_BRIDGE_BOTH_PORT_ISOLATED: both source port and dest port are in BR_ISOLATED state when bridge forwarding. Signed-off-by: xu xin <xu.xin16 at zte.com.cn> Reviewed-by: Zhang Yunkai <zhang.yunkai at zte.com.cn> Reviewed-b...

...r tha > > propagate, -EOPNOTSUPP. > > I looked at that, but changing the caller is also a bit ugly. Answering on behalf of Simon, and hoping he will agree. You are missing a big opportunity to make the kernel avoid doing useless work. The dsa_slave_fdb_event() function runs in atomic switchdev notifier context, and schedules a deferred workqueue item - dsa_schedule_work() - to get sleepable context to program hardware. Only that scheduling a deferred work item is not exactly cheap, so we try to avoid doing that unless we know that we'll end up doing something with that FDB entry onc...

A firmware error is triggered when configuring a 9k MTU on the PF after switching to switchdev mode and then using a vdpa device with larger (1k) rings: mlx5_cmd_out_err: CREATE_GENERAL_OBJECT(0xa00) op_mod(0xd) failed, status bad resource(0x5), syndrome (0xf6db90), err(-22) This is due to the fact that the hw VQ size parameters are computed based on the umem_1/2/3_buffer_param_a/b capabili...

..._not_filtering = true; > > + /* Since dynamic FDB entries are legacy, all switch drivers should > + * support the flag at least by just installing a static entry and > + * letting the bridge age it. > + */ > + ds->supported_fdb_flags = DSA_FDB_FLAG_DYNAMIC; I believe that switchdev has a structural problem in the fact that FDB entries with flags that aren't interpreted by drivers (so they don't know if those flags are set or unset) are still passed to the switchdev notifier chains by default. I don't believe that anybody used 'bridge fdb add <mac> <d...

...and pretty orthogonal to host/guest interface. > > Jason said virtio net is sometimes preferable. > > If that's the case don't make it a standby. > > > > More advanced use-cases do exist and e.g. Alexander Duyck > > suggested using a switch-dev. > > The switchdev case would need a new feature bit, right? > > -Siwei I think it would need to be a completely new device. > > failover isn't it though. > > > > -- > > MST