Kernel 2.6 (2.6.11)
When ethernet-bridge forward a packet and such ethernet-frame has
VLAN-tag, bridge should update skb->prioriry for properly QoS
handling. This small patch does this.
Based upon discussion during last week I added pskb_may_pull()
checking and simple mapping from 802.1p/user_priority to skb->priority.
Patch-by: Leo Yuriev <[email protected]>
--
Best regards,
Leo mailto:[email protected]
-----BEGIN PGP PUBLIC KEY BLOCK-----
Version: 2.6
mQCPA0HMImkBbQEEALnWXpnchl1dHaCfwnXe2RO8ft9e7K5IGRW1lE9RERMy6R3L
JnMXMPiuWm/4tx6H/yTRUML8LbuACzf2Np9oHTxbDWxP40wGxQKrPDlzv/9gLEp6
ZwGF8mSfvZrPHux1LwIbryQxjlmwfNCkE+qiVOBWMsD7yAFCWrZztbWTWJ6pABEB
AAG0GkxlbyBZdXJpZXYgPGxlb0B5dXJpZXYucnU+iQCVAwUQQcwiabZztbWTWJ6p
AQG6cAP9H0O+MMa0WDlGE2JGG+SWuu9Iuqg76Hp6tjtrz2pLWEzbq8oqCkE0THff
/YUUaKqnrLELwEaptE+MrWSv9Zt1K/PauMpKUWXhlYqGcGB2NqJL69AONQte0M4B
rPSSts4CU7gK2Zuds1DOLiON7e9Sbpjc5T+4D7Jw5XGKMz66nhY=
=qGIk
-----END PGP PUBLIC KEY BLOCK-----
--- net/bridge/br_input.c.orig 2005-03-02 10:37:50.000000000 +0300
+++ net/bridge/br_input.c 2005-03-11 16:28:02.000000000 +0300
@@ -5,6 +5,10 @@
* Authors:
* Lennert Buytenhek <[email protected]>
*
+ * Changes:
+ * 11/Mar/2005 - LY (Leo Yuriev <[email protected]>)
+ * Update skb->priority for packets with VLAN-tag.
+ *
* $Id: br_input.c,v 1.10 2001/12/24 04:50:20 davem Exp $
*
* This program is free software; you can redistribute it and/or
@@ -17,6 +21,10 @@
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/netfilter_bridge.h>
+#ifdef CONFIG_NET_SCHED
+# include <linux/pkt_sched.h>
+# include <linux/if_vlan.h>
+#endif /* CONFIG_NET_SCHED*/
#include "br_private.h"
const unsigned char bridge_ula[6] = { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
@@ -45,6 +53,75 @@ static void br_pass_frame_up(struct net_
br_pass_frame_up_finish);
}
+
+#ifdef CONFIG_NET_SCHED
+/*
+ * LY (Leo Yuriev):
+ * Just update skb->priority for properly QoS handling in case
+ * frame in the skb is contain VLAN-header.
+ *
+ * SANITY NOTE:
+ * We are referencing to the VLAN_HDR frields, which MAY be
+ * stored UNALIGNED in the memory.
+ * According to Dave Miller & Alexey, it will always be aligned,
+ * so there doesn't need to be any of the unaligned stuff.
+ *
+ */
+static __inline__ void br_update_skb_priority_if_vlan(struct sk_buff *skb)
+{
+ /*
+ * MAC-Layer Prioritization (802.1p as part of 802.1D-1998)
+ * - priority 0:
+ * Noncritical traffic such as backups, noncritical
+ * replications, some electronic mail, and so on;
+ * - priority 1:
+ * The default mode if none is specified;
+ * - priority 2:
+ * Best-effort traffic;
+ * - priority 3:
+ * Better than best effort, which would include important
+ * business traffic that can tolerate some delay;
+ * - priority 4:
+ * Controlled load, latency-sensitive traffic
+ * such as SNA transactions;
+ * - priority 5:
+ * Video, which is high bandwidth and sensitive to jitter;
+ * - priority 6:
+ * Voice traffic, such as NetMeeting, that is especially
+ * sensitive to jitter;
+ * - priority 7:
+ * Network control traffic
+ * such as router configuration messages;
+ */
+ static const __u8 mac_userprio2prio[8] = {
+ TC_PRIO_FILLER, /* prio 0 => band 2 */
+ TC_PRIO_BULK, /* prio 1 => band 2 */
+ TC_PRIO_BESTEFFORT, /* prio 2 => band 1 */
+ TC_PRIO_BESTEFFORT, /* prio 3 => band 1 */
+ TC_PRIO_INTERACTIVE_BULK, /* prio 4 => band 1 */
+ TC_PRIO_INTERACTIVE_BULK, /* prio 5 => band 1 */
+ TC_PRIO_INTERACTIVE, /* prio 6 => band 0 */
+ TC_PRIO_CONTROL, /* prio 7 => band 0 */
+ };
+ unsigned short vlan_TCI;
+ struct vlan_hdr *vhdr;
+
+ if (__constant_htons(ETH_P_8021Q) == skb->protocol && pskb_may_pull(skb, VLAN_HLEN)) {
+ vhdr = (struct vlan_hdr *)(skb->data);
+ /* vlan_TCI = ntohs(get_unaligned(&vhdr->h_vlan_TCI)); */
+ vlan_TCI = ntohs(vhdr->h_vlan_TCI);
+#ifdef VLAN_DEBUG
+ printk(VLAN_DBG "%s: skb: %p vlan_id: %hx\n",
+ __FUNCTION__, skb, (vlan_TCI & VLAN_VID_MASK));
+#endif
+ /* LY: whether we can assume here vlan_TCI <= 65535 always ?
+ * I think yes, then "& 7" is needless after shifting.
+ */
+ skb->priority = mac_userprio2prio[vlan_TCI >> 13];
+ }
+}
+#endif /* CONFIG_NET_SCHED */
+
/* note: already called with rcu_read_lock (preempt_disabled) */
int br_handle_frame_finish(struct sk_buff *skb)
{
@@ -54,6 +131,10 @@ int br_handle_frame_finish(struct sk_buf
struct net_bridge_fdb_entry *dst;
int passedup = 0;
+#ifdef CONFIG_NET_SCHED
+ br_update_skb_priority_if_vlan(skb);
+#endif /* CONFIG_NET_SCHED*/
+
if (br->dev->flags & IFF_PROMISC) {
struct sk_buff *skb2;
On Fri, 11 Mar 2005 17:20:22 +0300
Leo Yuriev <[email protected]> wrote:
> Kernel 2.6 (2.6.11)
>
> When ethernet-bridge forward a packet and such ethernet-frame has
> VLAN-tag, bridge should update skb->prioriry for properly QoS
> handling. This small patch does this.
>
> Based upon discussion during last week I added pskb_may_pull()
> checking and simple mapping from 802.1p/user_priority to skb->priority.
>
> Patch-by: Leo Yuriev <[email protected]>
Do this as an ebtables module please.
Actually there is a case to be made for this to be part of the
bridge code. A VLAN is a single collision domain which is mappable to a
collission domain that is a bridge.
infact the old VLAN code written by Lennert (and somebody else) had
those two intermingled.
cheers,
jamal
On Fri, 2005-03-11 at 12:37, Stephen Hemminger wrote:
> On Fri, 11 Mar 2005 17:20:22 +0300
> Leo Yuriev <[email protected]> wrote:
>
> > Kernel 2.6 (2.6.11)
> >
> > When ethernet-bridge forward a packet and such ethernet-frame has
> > VLAN-tag, bridge should update skb->prioriry for properly QoS
> > handling. This small patch does this.
> >
> > Based upon discussion during last week I added pskb_may_pull()
> > checking and simple mapping from 802.1p/user_priority to skb->priority.
> >
> > Patch-by: Leo Yuriev <[email protected]>
>
> Do this as an ebtables module please.
>