BATMAN currently does not allow QoS policies to be used with fragmented packets. The first patch copies the skb priority from the unfragmented packet into the fragments as they are built, so the lower layers can then prioritise as appropriated.
The second patch places the priority into the fragment header, and upon reception of a fragment, copies it into the skb priority, so that when the fragment is retransmitted, it is correctly prioritised.
Andrew Lunn (2): batman-adv: Set skb priority in fragments batman-adv: Include frame priority in fragment header
net/batman-adv/fragmentation.c | 4 +++- net/batman-adv/packet.h | 7 +++++-- net/batman-adv/routing.c | 2 ++ 3 files changed, 10 insertions(+), 3 deletions(-)
BATMAN will set the skb->priority based on the IP precedence or 802.1q tag. However, if it needs to fragment the frame, it currently leaves the fragment skb with the default priority and actually overwrites the priority in the unfragmented frame. Fix this.
Signed-off-by: Andrew Lunn andrew@lunn.ch --- net/batman-adv/fragmentation.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/batman-adv/fragmentation.c b/net/batman-adv/fragmentation.c index 65536db..a3fefbd 100644 --- a/net/batman-adv/fragmentation.c +++ b/net/batman-adv/fragmentation.c @@ -414,7 +414,7 @@ static struct sk_buff *batadv_frag_create(struct sk_buff *skb, if (!skb_fragment) goto err;
- skb->priority = TC_PRIO_CONTROL; + skb_fragment->priority = skb->priority;
/* Eat the last mtu-bytes of the skb */ skb_reserve(skb_fragment, header_size + ETH_HLEN);
On Monday 09 May 2016 20:03:35 Andrew Lunn wrote:
BATMAN will set the skb->priority based on the IP precedence or 802.1q tag. However, if it needs to fragment the frame, it currently leaves the fragment skb with the default priority and actually overwrites the priority in the unfragmented frame. Fix this.
Signed-off-by: Andrew Lunn andrew@lunn.ch
net/batman-adv/fragmentation.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/batman-adv/fragmentation.c b/net/batman-adv/fragmentation.c index 65536db..a3fefbd 100644 --- a/net/batman-adv/fragmentation.c +++ b/net/batman-adv/fragmentation.c @@ -414,7 +414,7 @@ static struct sk_buff *batadv_frag_create(struct sk_buff *skb, if (!skb_fragment) goto err;
- skb->priority = TC_PRIO_CONTROL;
skb_fragment->priority = skb->priority;
/* Eat the last mtu-bytes of the skb */ skb_reserve(skb_fragment, header_size + ETH_HLEN);
Please remove linux/pkt_sched.h in fragmentation.c because it is only used for TC_PRIO_CONTROL in master.
Kind regards, Sven
On Monday, May 09, 2016 20:03:35 Andrew Lunn wrote:
BATMAN will set the skb->priority based on the IP precedence or 802.1q tag. However, if it needs to fragment the frame, it currently leaves the fragment skb with the default priority and actually overwrites the priority in the unfragmented frame. Fix this.
Signed-off-by: Andrew Lunn andrew@lunn.ch
net/batman-adv/fragmentation.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Applied in revision e374d75.
Thanks, Marek
Unfragmented frames which traverse a node have their skb->priority set by looking at the IP ToS byte, or the 802.1p header. However for fragments this is not possible, only one of the fragments will contain the headers. Instead, place the priority into the fragment header and on receiving a fragment, use this information to set the skb->priority for when the fragment is forwarded.
Signed-off-by: Andrew Lunn andrew@lunn.ch --- net/batman-adv/fragmentation.c | 2 ++ net/batman-adv/packet.h | 7 +++++-- net/batman-adv/routing.c | 2 ++ 3 files changed, 9 insertions(+), 2 deletions(-)
diff --git a/net/batman-adv/fragmentation.c b/net/batman-adv/fragmentation.c index a3fefbd..c703767 100644 --- a/net/batman-adv/fragmentation.c +++ b/net/batman-adv/fragmentation.c @@ -473,6 +473,8 @@ bool batadv_frag_send_packet(struct sk_buff *skb, frag_header.reserved = 0; frag_header.no = 0; frag_header.total_size = htons(skb->len); + if (skb->priority >= 256 && skb->priority <= 263) + frag_header.priority = skb->priority - 256; ether_addr_copy(frag_header.orig, primary_if->net_dev->dev_addr); ether_addr_copy(frag_header.dest, orig_node->orig);
diff --git a/net/batman-adv/packet.h b/net/batman-adv/packet.h index b45460d..7acc6a2 100644 --- a/net/batman-adv/packet.h +++ b/net/batman-adv/packet.h @@ -384,6 +384,7 @@ struct batadv_unicast_4addr_packet { * @dest: final destination used when routing fragments * @orig: originator of the fragment used when merging the packet * @no: fragment number within this sequence + * @priority: priority of frame, from ToS IP precedence or 802.1p * @reserved: reserved byte for alignment * @seqno: sequence identification * @total_size: size of the merged packet @@ -394,9 +395,11 @@ struct batadv_frag_packet { u8 ttl; #if defined(__BIG_ENDIAN_BITFIELD) u8 no:4; - u8 reserved:4; + u8 priority:3; + u8 reserved:1; #elif defined(__LITTLE_ENDIAN_BITFIELD) - u8 reserved:4; + u8 reserved:1; + u8 priority:3; u8 no:4; #else #error "unknown bitfield endianness" diff --git a/net/batman-adv/routing.c b/net/batman-adv/routing.c index ae850f2..01d36e3 100644 --- a/net/batman-adv/routing.c +++ b/net/batman-adv/routing.c @@ -1004,6 +1004,8 @@ int batadv_recv_frag_packet(struct sk_buff *skb, if (!orig_node_src) goto out;
+ skb->priority = frag_packet->priority + 256; + /* Route the fragment if it is not for us and too big to be merged. */ if (!batadv_is_my_mac(bat_priv, frag_packet->dest) && batadv_frag_skb_fwd(skb, recv_if, orig_node_src)) {
On Monday, May 09, 2016 20:03:36 Andrew Lunn wrote:
Unfragmented frames which traverse a node have their skb->priority set by looking at the IP ToS byte, or the 802.1p header. However for fragments this is not possible, only one of the fragments will contain the headers. Instead, place the priority into the fragment header and on receiving a fragment, use this information to set the skb->priority for when the fragment is forwarded.
Signed-off-by: Andrew Lunn andrew@lunn.ch
net/batman-adv/fragmentation.c | 2 ++ net/batman-adv/packet.h | 7 +++++-- net/batman-adv/routing.c | 2 ++ 3 files changed, 9 insertions(+), 2 deletions(-)
Applied in revision 4f241fc.
Thanks, Marek
On Monday 09 May 2016 20:03:34 Andrew Lunn wrote:
BATMAN currently does not allow QoS policies to be used with fragmented packets. The first patch copies the skb priority from the unfragmented packet into the fragments as they are built, so the lower layers can then prioritise as appropriated.
The second patch places the priority into the fragment header, and upon reception of a fragment, copies it into the skb priority, so that when the fragment is retransmitted, it is correctly prioritised.
Andrew Lunn (2): batman-adv: Set skb priority in fragments batman-adv: Include frame priority in fragment header
Series looks good to me.
Reviewed-by: Simon Wunderlich sw@simonwunderlich.de
Thanks, Simon
b.a.t.m.a.n@lists.open-mesh.org