[Bridge] [RFC PATCH] batman-adv: Reserve needed_headroom for fragments
annika.wickert at exaring.de
Thu Nov 26 08:58:45 UTC 2020
>I find your output slightly confusing. Maybe you can change your printk stuff
> to something more like:
>printk("%s %s:%u max_headroom %u\n", __FILE__, __func__, __LINE__, max_headroom);
Will add this thx.
>On Thursday, 26 November 2020 00:14:35 CET Annika Wickert wrote:
>> This is what I get from the bridge when bat0 is enslaved with the vxlan interface as member of batman ( https://elixir.bootlin.com/linux/latest/source/net/bridge/br_if.c#L311 )
>> [ 36.959547] Bridge firewalling registered
>> [ 522.221767] SKB Bridge br_if.c: max_headroom 0
>> [ 522.221781] SKB Bridge br_if.c: new_hr 0
>> [ 627.186129] SKB Bridge br_if.c: max_headroom 0
>> [ 627.186139] SKB Bridge br_if.c: new_hr 0
>> [ 627.616650] SKB Bridge br_if.c: new_hr 102
>When is this shown? Does the batadv interface already have its hardif (slave)
>interfaces attached at that point? And did the vxlan report the correct
>needed_headroom to batadv before you've tried to attach the batadv interface
>to the bridge?
BATMAN already has the vxlan interface as hardif here is the script I use to generate the config:
ip link add mesh-vpn type vxlan id 4831583 local fe80::2e0:2fff:fe18:dc2f remote fe80::281:8eff:fef0:73aa dstport 8472 dev wg-uplink
ip link del bat-welt
batctl ra BATMAN_V
batctl meshif bat-welt interface add mesh-vpn
ip link set up bat-welt
ip link set dev bat-welt master br-welt
> Because the bridge can also only change its needed_headroom on interface add
>> Also BATMAN reports itself when initialized and seems not to propagate stuff up the stack on change?: (https://github.com/open-mesh-mirror/batman-adv/blob/master/net/batman-adv/hard-interface.c#L555 )
>> [ 3350.212116] SKB hard-interface.h: lower_headroom 70
>> [ 3350.212126] SKB hard-interface.h: needed_headroom 102
>Afaik, it is "propagating" its stuff by adjusting its own needed_headroom/
>tailroom at this point. But there is no way to notify that the headroom/
>tailroom was changed and the upper layers should recalculate it.
Which should already include the headroom needed by vxlan as it's already present as hardif.
>If you need something like this then we might to have a new
>NETDEV_RESERVED_SPACE_CHANGE (or a better name OR maybe use a netdev_cmd with
>a similar meaning). And then call this whenever the needed_headroom/
>tailroom/... of an interface changes during its lifetime. And bridge/batman-
>adv/ovs/... have to check the headroom in their notifier_call again when they
>receive this event.
>Could it be that the vxlan didn't had the correct needed_headroom when you've
>added it to you batadv interface? Or that the vxlan interface didn't set the
>correct needed_headroom for its lower_dev (see vxlan_config_apply)?
The vxlan interface was added first. So it should propagate it?
>If you have the "slow" setup, can you please do following steps:
>* keep vxlan as is (I hope you specify a fixed lowerdev)
>- but try to print the needed headroom in vxlan_config_apply and compare it
> to the ones from vxlan_build_skb
>* remove the vxlan from your batadv interface
>* add your vxlan again from the batadv interface
> - check if the headroom numbers are now looking better in
> * remove batadv interface from the bridge
> * add your batadv interface again to the bridge
> - is update_headroom() now using the correct headroom information?
As the setup is always doing the pskb_expand_head() it should be possible to test this.
More information about the Bridge