Skip to contentSkip to conversation

net: vlan: make non-hw-accel rx path similar to hw-accel

PATCH In review Jiri Pirko 11 participants 92 messages Apr 8, 2011 – Jun 12, 2011

Subsystems (2)
lore
Share & tools
lore

Jiri Pirko

author
#0 lore raw

PATCH v2 1/1 net: vlan: make non-hw-accel rx path similar to hw-accel

Now there are 2 paths for rx vlan frames. When rx-vlan-hw-accel is enabled, skb is untagged by NIC, vlan_tci is set and the skb gets into vlan code in __netif_receive_skb - vlan_hwaccel_do_receive.

For non-rx-vlan-hw-accel however, tagged skb goes thru whole __netif_receive_skb, it's untagged in ptype_base hander and reinjected

This incosistency is fixed by this patch. Vlan untagging happens early in __netif_receive_skb so the rest of code (ptype_all handlers, rx_handlers) see the skb like it was untagged by hw.

v1->v2:
remove "inline" from vlan_core.c functions

---
 include/linux/if_vlan.h |   10 ++-
 net/8021q/vlan.c        |    8 --
 net/8021q/vlan.h        |    2 -
 net/8021q/vlan_core.c   |   85 +++++++++++++++++++++++-
 net/8021q/vlan_dev.c    |  173 -----------------------------------------------
 net/core/dev.c          |    8 ++-
 6 files changed, 99 insertions(+), 187 deletions(-)
include/linux/if_vlan.h +8 -2
index 635e1fa..998b299 100644
@@ -132,7 +132,8 @@ extern u16 vlan_dev_vlan_id(const struct net_device *dev);
 
 extern int __vlan_hwaccel_rx(struct sk_buff *skb, struct vlan_group *grp,
 			     u16 vlan_tci, int polling);
-extern bool vlan_hwaccel_do_receive(struct sk_buff **skb);
+extern bool vlan_do_receive(struct sk_buff **skb);
+extern struct sk_buff *vlan_untag(struct sk_buff *skb);
 extern gro_result_t
 vlan_gro_receive(struct napi_struct *napi, struct vlan_group *grp,
 		 unsigned int vlan_tci, struct sk_buff *skb);
@@ -166,13 +167,18 @@ static inline int __vlan_hwaccel_rx(struct sk_buff *skb, struct vlan_group *grp,
 	return NET_XMIT_SUCCESS;
 }
 
-static inline bool vlan_hwaccel_do_receive(struct sk_buff **skb)
+static inline bool vlan_do_receive(struct sk_buff **skb)
 {
 	if ((*skb)->vlan_tci & VLAN_VID_MASK)
 		(*skb)->pkt_type = PACKET_OTHERHOST;
 	return false;
 }
 
+inline struct sk_buff *vlan_untag(struct sk_buff *skb)
+{
+	return skb;
+}
+
 static inline gro_result_t
 vlan_gro_receive(struct napi_struct *napi, struct vlan_group *grp,
 		 unsigned int vlan_tci, struct sk_buff *skb)
net/8021q/vlan.c +0 -8
index e47600b..14ef5ef 100644
@@ -49,11 +49,6 @@ const char vlan_version[] = DRV_VERSION;
 static const char vlan_copyright[] = "Ben Greear <[email protected]>";
 static const char vlan_buggyright[] = "David S. Miller <[email protected]>";
 
-static struct packet_type vlan_packet_type __read_mostly = {
-	.type = cpu_to_be16(ETH_P_8021Q),
-	.func = vlan_skb_recv, /* VLAN receive method */
-};
-
 /* End of global variables definitions. */
 
 static void vlan_group_free(struct vlan_group *grp)
@@ -684,7 +679,6 @@ static int __init vlan_proto_init(void)
 	if (err < 0)
 		goto err4;
 
-	dev_add_pack(&vlan_packet_type);
 	vlan_ioctl_set(vlan_ioctl_handler);
 	return 0;
 
@@ -705,8 +699,6 @@ static void __exit vlan_cleanup_module(void)
 
 	unregister_netdevice_notifier(&vlan_notifier_block);
 
-	dev_remove_pack(&vlan_packet_type);
-
 	unregister_pernet_subsys(&vlan_net_ops);
 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
 
net/8021q/vlan.h +0 -2
index 5687c9b..c3408de 100644
@@ -75,8 +75,6 @@ static inline struct vlan_dev_info *vlan_dev_info(const struct net_device *dev)
 }
 
 /* found in vlan_dev.c */
-int vlan_skb_recv(struct sk_buff *skb, struct net_device *dev,
-		  struct packet_type *ptype, struct net_device *orig_dev);
 void vlan_dev_set_ingress_priority(const struct net_device *dev,
 				   u32 skb_prio, u16 vlan_prio);
 int vlan_dev_set_egress_priority(const struct net_device *dev,
net/8021q/vlan_core.c +84 -1

94 lines not on this page · raw message

net/8021q/vlan_dev.c +0 -173

179 lines not on this page · raw message

net/core/dev.c +7 -1
index 5d0b4f6..52aed0a 100644
@@ -3130,6 +3130,12 @@ another_round:
 
 	__this_cpu_inc(softnet_data.processed);
 
+	if (skb->protocol == cpu_to_be16(ETH_P_8021Q)) {
+		skb = vlan_untag(skb);
+		if (unlikely(!skb))
+			goto out;
+	}
+
 #ifdef CONFIG_NET_CLS_ACT
 	if (skb->tc_verd & TC_NCLS) {
 		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
@@ -3177,7 +3183,7 @@ ncls:
 			ret = deliver_skb(skb, pt_prev, orig_dev);
 			pt_prev = NULL;
 		}
-		if (vlan_hwaccel_do_receive(&skb)) {
+		if (vlan_do_receive(&skb)) {
 			ret = __netif_receive_skb(skb);
 			goto out;
 		} else if (unlikely(!skb))
Signature · 1 line
1.7.4
maintainer
#1 lore raw

From: Jiri Pirko <[email protected]>
Date: Fri, 8 Apr 2011 07:48:33 +0200

Quoted 7 lines

Now there are 2 paths for rx vlan frames. When rx-vlan-hw-accel is enabled, skb is untagged by NIC, vlan_tci is set and the skb gets into vlan code in __netif_receive_skb - vlan_hwaccel_do_receive.

For non-rx-vlan-hw-accel however, tagged skb goes thru whole __netif_receive_skb, it's untagged in ptype_base hander and reinjected

Quoted 8 lines

This incosistency is fixed by this patch. Vlan untagging happens early in __netif_receive_skb so the rest of code (ptype_all handlers, rx_handlers) see the skb like it was untagged by hw.

v1->v2:
remove "inline" from vlan_core.c functions

Ok, I've applied this, let's see what happens :-)

Thanks!

#2 lore raw
Quoted 14 lines · David Miller

From: Jiri Pirko <[email protected]>
Date: Fri,  8 Apr 2011 07:48:33 +0200

Quoted 11 lines

Now there are 2 paths for rx vlan frames. When rx-vlan-hw-accel is enabled, skb is untagged by NIC, vlan_tci is set and the skb gets into vlan code in __netif_receive_skb - vlan_hwaccel_do_receive.

For non-rx-vlan-hw-accel however, tagged skb goes thru whole __netif_receive_skb, it's untagged in ptype_base hander and reinjected

This incosistency is fixed by this patch. Vlan untagging happens early in __netif_receive_skb so the rest of code (ptype_all handlers, rx_handlers) see the skb like it was untagged by hw.

Quoted 7 lines
Quoted 4 lines

v1->v2:
      remove "inline" from vlan_core.c functions

Ok, I've applied this, let's see what happens :-)

I think we should revert it.

File: net/8021q/vlan_core.c:

161 skb_pull_rcsum(skb, VLAN_HLEN);

skb->data and skb->len are updated, but network_header and transport_header are left unchanged. This will break the assumption in net_sched.

for example:
file: cls_u32.c
104 unsigned int off = skb_network_offset(skb);
After this patch, skb_network_offset may be negative.

162 vlan_set_encap_proto(skb, vhdr);
163
164 skb = vlan_check_reorder_header(skb);
vlan_check_reorder_header assume skb->dev is a vlan_dev. Even though the correct dev is assigned temporarily, we should not reorder the header here as HW accelerated vlan RX does, as this may breaks the bridging comes later.

165 if (unlikely(!skb))
166 goto err_free;

The hardware accelerated vlan RX doesn't always do the "right" things as it strips the vlan header, so we should not emulate it in software all the time.

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#3 lore raw
Quoted 46 lines · Sat, May 21, 2011 at 03:11:05AM CEST, [email protected] wrote:
Quoted 14 lines · David Miller

From: Jiri Pirko <[email protected]>
Date: Fri,  8 Apr 2011 07:48:33 +0200

Quoted 11 lines

Now there are 2 paths for rx vlan frames. When rx-vlan-hw-accel is enabled, skb is untagged by NIC, vlan_tci is set and the skb gets into vlan code in __netif_receive_skb - vlan_hwaccel_do_receive.

For non-rx-vlan-hw-accel however, tagged skb goes thru whole __netif_receive_skb, it's untagged in ptype_base hander and reinjected

This incosistency is fixed by this patch. Vlan untagging happens early in __netif_receive_skb so the rest of code (ptype_all handlers, rx_handlers) see the skb like it was untagged by hw.

Quoted 7 lines
Quoted 4 lines

v1->v2:
      remove "inline" from vlan_core.c functions

Ok, I've applied this, let's see what happens :-)

I think we should revert it.

File: net/8021q/vlan_core.c:

161 skb_pull_rcsum(skb, VLAN_HLEN);

skb->data and skb->len are updated, but network_header and transport_header are left unchanged. This will break the assumption in net_sched.

for example:
file: cls_u32.c
104 unsigned int off = skb_network_offset(skb);
After this patch, skb_network_offset may be negative.

162 vlan_set_encap_proto(skb, vhdr);
163
164 skb = vlan_check_reorder_header(skb);
vlan_check_reorder_header assume skb->dev is a vlan_dev. Even though the correct dev is assigned temporarily, we should not reorder the header here as HW accelerated vlan RX does, as this may breaks the bridging comes later.

Quoted 6 lines

165 if (unlikely(!skb))
166 goto err_free;

The hardware accelerated vlan RX doesn't always do the "right" things as it strips the vlan header, so we should not emulate it in software all the time.

I do not see a reason why to not emulate that. To make paths as much similar as they can be, that is the point of this patch.

I think it would be better to fix an issue you are pointing at rather that revert this.

Jirka

Quoted 4 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])
#4 lore raw
Quoted 7 lines · Jiri Pirko

I do not see a reason why to not emulate that. To make paths as much similar as they can be, that is the point of this patch.

I think it would be better to fix an issue you are pointing at rather that revert this.

In my opinion, the hardware accelerated VLAN RX is just a special case of the non hardware accelerated VLAN RX with header reordering. For promiscuous NICs and bridges, hw-accel-vlan-rx is just disabled.

I have tried to fix all the issues, but failed in a clean way. Please try.

Thanks.

Signature · 2 lines
Regards,
Changli Gao([email protected])
#5 lore raw
Quoted 5 lines · Changli Gao
Quoted 4 lines · Jiri Pirko

I do not see a reason why to not emulate that. To make paths as much similar as they can be, that is the point of this patch.

Quoted 7 lines
Quoted 3 lines

I think it would be better to fix an issue you are pointing at rather that revert this.

In my opinion, the hardware accelerated VLAN RX is just a special case of the non hardware accelerated VLAN RX with header reordering. For promiscuous NICs and bridges, hw-accel-vlan-rx is just disabled.

I strongly agree with that.

The fact that a skb holds a VLAN tag is not a good enough reason to always remove this tag before giving the skb to protocol handlers.

If the user ask for VLAN tag removal, we should remove the tag, possibly using hw-accel untagging if available else software untagging. And if the user doesn't ask for tag removal, we should not untag.

In other words, if the user doesn't setup any vlan interface on top of another interface, there is no reason to untag the skb : both hw-accel untagging and software untagging should be disabled.

Also, the skb should be delivered untagged or tagged to protocol handlers, depending on the particular device the protocol handlers registered at. The same skb might need to be delivered tagged to a ptype_all handler registered at eth0 and untagged to a ptype_base handler registered at eth0.100.

rx_handler still sounds the right place to do software untagging, because software untagging is a per-device process. A vlan_untagging rx_handler should be installed on the devices that have vlan child device and (that lack hw-accel or where hw-accel is disabled). This would also cause __netif_receive_skb() not to hold any vlan specific code, which is cleaner.

I perfectly understand that this might require several rx_handlers per device, for advanced setup, but I think for long that several rx_handlers is a powerful feature we need.

Nicolas.

#6 lore raw
Quoted 20 lines · Nicolas de Pesloüan
Quoted 6 lines · Changli Gao
Quoted 4 lines · Jiri Pirko

I do not see a reason why to not emulate that. To make paths as much similar as they can be, that is the point of this patch.

Quoted 7 lines
Quoted 3 lines

I think it would be better to fix an issue you are pointing at rather that revert this.

In my opinion, the hardware accelerated VLAN RX is just a special case of the non hardware accelerated VLAN RX with header reordering. For promiscuous NICs and bridges, hw-accel-vlan-rx is just disabled.

I strongly agree with that.

The fact that a skb holds a VLAN tag is not a good enough reason to always remove this tag before giving the skb to protocol handlers.

Quoted 7 lines

If the user ask for VLAN tag removal, we should remove the tag, possibly using hw-accel untagging if available else software untagging. And if the user doesn't ask for tag removal, we should not untag.

In other words, if the user doesn't setup any vlan interface on top of another interface, there is no reason to untag the skb : both hw-accel untagging and software untagging should be disabled.

The problem is that for most hardware vlan stripping is actually the common case, not the exception. When you try to disable it frequently there are hidden restrictions that cause problems. A few examples:
* Some NICs can't disable stripping at all.
* Some NICs can only do tag insertion if stripping is configured on receive.
* Some NICs can only do hardware offloads (checksum, TSO) if tag insertion is used on transmit.

So if you are using vlans then acceleration is pretty much a fact of life and the best possible way we can deal with it is to make the accelerated and non-accelerated cases behave as similarly as possible.

Before we were trying to dynamically enable/disable vlan acceleration based on whether a vlan group was configured and that worked fine for vlan devices because acceleration was enabled for it. However, it caused an endless series of problems for other devices (such as bridging while trunking vlans) due to lost tags, driver bugs, and the restrictions above. Some of these can be fixed with driver changes but the fact is that dynamically changing behavior just leads to problems for the less common cases that are supposedly being fixed. It's much better to do the same thing all the time.

#7 lore raw
Quoted 42 lines · On Sat, 21 May 2011 10:54:39 -0700 Jesse Gross <[email protected]> wrote:
Quoted 20 lines · Nicolas de Pesloüan
Quoted 6 lines · Changli Gao
Quoted 4 lines · Jiri Pirko

I do not see a reason why to not emulate that. To make paths as much similar as they can be, that is the point of this patch.

Quoted 7 lines
Quoted 3 lines

I think it would be better to fix an issue you are pointing at rather that revert this.

In my opinion, the hardware accelerated VLAN RX is just a special case of the non hardware accelerated VLAN RX with header reordering. For promiscuous NICs and bridges, hw-accel-vlan-rx is just disabled.

I strongly agree with that.

The fact that a skb holds a VLAN tag is not a good enough reason to always remove this tag before giving the skb to protocol handlers.

Quoted 7 lines

If the user ask for VLAN tag removal, we should remove the tag, possibly using hw-accel untagging if available else software untagging. And if the user doesn't ask for tag removal, we should not untag.

In other words, if the user doesn't setup any vlan interface on top of another interface, there is no reason to untag the skb : both hw-accel untagging and software untagging should be disabled.

The problem is that for most hardware vlan stripping is actually the common case, not the exception. When you try to disable it frequently there are hidden restrictions that cause problems. A few examples:
* Some NICs can't disable stripping at all.
* Some NICs can only do tag insertion if stripping is configured on receive.
* Some NICs can only do hardware offloads (checksum, TSO) if tag insertion is used on transmit.

So if you are using vlans then acceleration is pretty much a fact of life and the best possible way we can deal with it is to make the accelerated and non-accelerated cases behave as similarly as possible.

Quoted 10 lines

Before we were trying to dynamically enable/disable vlan acceleration based on whether a vlan group was configured and that worked fine for vlan devices because acceleration was enabled for it. However, it caused an endless series of problems for other devices (such as bridging while trunking vlans) due to lost tags, driver bugs, and the restrictions above. Some of these can be fixed with driver changes but the fact is that dynamically changing behavior just leads to problems for the less common cases that are supposedly being fixed. It's much better to do the same thing all the time.

The old code was also fundamentally broken if doing any kind of Qos because the TC filter would have to know whether skb had extra overhead of VLAN tag at the start. This meant the TC filter setup had to be different depending on whether the hardware supported HW acceleration or not.

#8 lore raw
In reply to Jesse Gross
Quoted 42 lines · Jesse Gross
Quoted 20 lines · Nicolas de Pesloüan
Quoted 6 lines · Changli Gao
Quoted 4 lines · Jiri Pirko

I do not see a reason why to not emulate that. To make paths as much similar as they can be, that is the point of this patch.

Quoted 7 lines
Quoted 3 lines

I think it would be better to fix an issue you are pointing at rather that revert this.

In my opinion, the hardware accelerated VLAN RX is just a special case of the non hardware accelerated VLAN RX with header reordering. For promiscuous NICs and bridges, hw-accel-vlan-rx is just disabled.

I strongly agree with that.

The fact that a skb holds a VLAN tag is not a good enough reason to always remove this tag before giving the skb to protocol handlers.

Quoted 7 lines

If the user ask for VLAN tag removal, we should remove the tag, possibly using hw-accel untagging if available else software untagging. And if the user doesn't ask for tag removal, we should not untag.

In other words, if the user doesn't setup any vlan interface on top of another interface, there is no reason to untag the skb : both hw-accel untagging and software untagging should be disabled.

The problem is that for most hardware vlan stripping is actually the common case, not the exception. When you try to disable it frequently there are hidden restrictions that cause problems. A few examples:
* Some NICs can't disable stripping at all.
* Some NICs can only do tag insertion if stripping is configured on receive.
* Some NICs can only do hardware offloads (checksum, TSO) if tag insertion is used on transmit.

So if you are using vlans then acceleration is pretty much a fact of life and the best possible way we can deal with it is to make the accelerated and non-accelerated cases behave as similarly as possible.

Quoted 9 lines

Before we were trying to dynamically enable/disable vlan acceleration based on whether a vlan group was configured and that worked fine for vlan devices because acceleration was enabled for it. However, it caused an endless series of problems for other devices (such as bridging while trunking vlans) due to lost tags, driver bugs, and the restrictions above. Some of these can be fixed with driver changes but the fact is that dynamically changing behavior just leads to problems for the less common cases that are supposedly being fixed. It's much better to do the same thing all the time.

Thanks for clarifying.

So, because many limited/buggy hardware exist, we must mimic the behavior in software. 'Sounds good to me.

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Of course, I don't say we should always re-tag: if no protocol handler nor rx_handler were registered on the parent interface, we don't need the extra work of re-tagging.

What I say is that it shouldn't be the job of protocol handlers or rx_handlers that expect the skb to be tagged to fix the improper untagging. A generic feature should do it when necessary.

And all this being said, it doesn't mean that we should pollute __netif_receive_skb with special code for vlan handling.

May be, as suggested by Eric W. Biederman in the V1 thread for this patch, software untagging for the first level of header should happen before __netif_receive_skb if we only try to mimic hardware behavior.

And possible later untagging (due to vlan nesting) should be done generically inside __netif_receive_skb, using rx_handler when appropriate. This would cleanup the general case where no vlan is involved at all.

Nicolas.

author
#9 lore raw

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

#10 lore raw
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

Eric

#11 lore raw
Quoted 14 lines · Eric W. Biederman
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

bridge is another. More precisely, there is a difference between the following two setups:

1/ eth0 - eth0.100 - br0 - eth1.200 - eth1

2/ eth0 - br0 - eth1

In case 1, it is normal and desirable for the bridge to see untagged skb.

In case 2, it is desirable for the bridge to see untouched (possibly tagged) skb. If current bridge implementation is able to handle skb from which we removed a tag, in this situation, it means that bridge currently "fix improper untagging" by itself, by forcing re-tagging on output. I think is should not be the job of protocol handlers to fix this. Again, a generic feature should to it when necessary.

Think of the following setups:

3/ eth0 - br0 - eth1.200 - eth1.
4/ eth0 - eth0.100 - br0 - eth1

What if one expect this setup to add (3) or remove (4) one level of vlan nesting? This is precisely what this setup suggest. How can we instruct the bridge to do so? It is not the bridge responsibility to do any vlan processing. bridge is expected to... bridge !

Nicolas.

#12 lore raw

2011/5/22 Nicolas de Pesloüan <[email protected]>:

Quoted 38 lines · Eric W. Biederman
Quoted 15 lines
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

bridge is another. More precisely, there is a difference between the following two setups:

1/ eth0 - eth0.100 - br0 - eth1.200 - eth1

2/ eth0 - br0 - eth1

In case 1, it is normal and desirable for the bridge to see untagged skb.

In case 2, it is desirable for the bridge to see untouched (possibly tagged) skb. If current bridge implementation is able to handle skb from which we removed a tag, in this situation, it means that bridge currently "fix improper untagging" by itself, by forcing re-tagging on output. I think is should not be the job of protocol handlers to fix this. Again, a generic feature should to it when necessary.

Think of the following setups:

Quoted 7 lines

3/ eth0 - br0 - eth1.200 - eth1.
4/ eth0 - eth0.100 - br0 - eth1

What if one expect this setup to add (3) or remove (4) one level of vlan nesting? This is precisely what this setup suggest. How can we instruct the bridge to do so? It is not the bridge responsibility to do any vlan processing. bridge is expected to... bridge !

I assumed that this untaging Jiri is implementing does not remove the tag. It moves the information from skb->data to skb->vlan_tci, but the information contained is not otherwise changing. All your examples should work regardless of where the tag is stored.

Best Regards,
Michał Mirosław

#13 lore raw
Quoted 47 lines · Michał Mirosław

2011/5/22 Nicolas de Pesloüan<[email protected]>:

Quoted 38 lines · Eric W. Biederman
Quoted 15 lines
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

bridge is another. More precisely, there is a difference between the following two setups:

1/ eth0 - eth0.100 - br0 - eth1.200 - eth1

2/ eth0 - br0 - eth1

In case 1, it is normal and desirable for the bridge to see untagged skb.

In case 2, it is desirable for the bridge to see untouched (possibly tagged) skb. If current bridge implementation is able to handle skb from which we removed a tag, in this situation, it means that bridge currently "fix improper untagging" by itself, by forcing re-tagging on output. I think is should not be the job of protocol handlers to fix this. Again, a generic feature should to it when necessary.

Think of the following setups:

Quoted 7 lines

3/ eth0 - br0 - eth1.200 - eth1.
4/ eth0 - eth0.100 - br0 - eth1

What if one expect this setup to add (3) or remove (4) one level of vlan nesting? This is precisely what this setup suggest. How can we instruct the bridge to do so? It is not the bridge responsibility to do any vlan processing. bridge is expected to... bridge !

Quoted 4 lines

I assumed that this untaging Jiri is implementing does not remove the tag. It moves the information from skb->data to skb->vlan_tci, but the information contained is not otherwise changing. All your examples should work regardless of where the tag is stored.

I assumed (but didn't tested) that this untagging also change the starting point of the payload of the packet. So protocol handlers expecting to have the raw packet won't see the vlan header.

Nicolas.

#14 lore raw

W dniu 22 maja 2011 11:10 użytkownik Nicolas de Pesloüan

Quoted 55 lines · <[email protected]> napisał:
Quoted 54 lines · Michał Mirosław

2011/5/22 Nicolas de Pesloüan<[email protected]>:

Quoted 51 lines · Eric W. Biederman
Quoted 16 lines
Quoted 13 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

bridge is another. More precisely, there is a difference between the following two setups:

1/ eth0 - eth0.100 - br0 - eth1.200 - eth1

2/ eth0 - br0 - eth1

In case 1, it is normal and desirable for the bridge to see untagged skb.

In case 2, it is desirable for the bridge to see untouched (possibly tagged)
skb. If current bridge implementation is able to handle skb from which we removed a tag, in this situation, it means that bridge currently "fix improper untagging" by itself, by forcing re-tagging on output. I think is
should not be the job of protocol handlers to fix this. Again, a generic feature should to it when necessary.

Think of the following setups:

3/ eth0 - br0 - eth1.200 - eth1.
4/ eth0 - eth0.100 - br0 - eth1

What if one expect this setup to add (3) or remove (4) one level of vlan nesting? This is precisely what this setup suggest. How can we instruct the
bridge to do so? It is not the bridge responsibility to do any vlan processing. bridge is expected to... bridge !

Quoted 8 lines
Quoted 4 lines

I assumed that this untaging Jiri is implementing does not remove the tag. It moves the information from skb->data to skb->vlan_tci, but the information contained is not otherwise changing. All your examples should work regardless of where the tag is stored.

I assumed (but didn't tested) that this untagging also change the starting point of the payload of the packet. So protocol handlers expecting to have the raw packet won't see the vlan header.

That would also be the case with hardware stripped tags - they need to look into skb->vlan_tci anyway.

Best Regards,
Michał Mirosław

author
#15 lore raw
Quoted 62 lines · Sun, May 22, 2011 at 11:20:09AM CEST, [email protected] wrote:

W dniu 22 maja 2011 11:10 użytkownik Nicolas de Pesloüan

Quoted 60 lines · <[email protected]> napisał:
Quoted 58 lines · Michał Mirosław

2011/5/22 Nicolas de Pesloüan<[email protected]>:

Quoted 43 lines · Eric W. Biederman
Quoted 16 lines
Quoted 13 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

bridge is another. More precisely, there is a difference between the following two setups:

1/ eth0 - eth0.100 - br0 - eth1.200 - eth1

2/ eth0 - br0 - eth1

In case 1, it is normal and desirable for the bridge to see untagged skb.

In case 2, it is desirable for the bridge to see untouched (possibly tagged)
skb. If current bridge implementation is able to handle skb from which we removed a tag, in this situation, it means that bridge currently "fix improper untagging" by itself, by forcing re-tagging on output. I think is
should not be the job of protocol handlers to fix this. Again, a generic feature should to it when necessary.

Think of the following setups:

Quoted 8 lines

3/ eth0 - br0 - eth1.200 - eth1.
4/ eth0 - eth0.100 - br0 - eth1

What if one expect this setup to add (3) or remove (4) one level of vlan nesting? This is precisely what this setup suggest. How can we instruct the
bridge to do so? It is not the bridge responsibility to do any vlan processing. bridge is expected to... bridge !

I assumed that this untaging Jiri is implementing does not remove the tag. It moves the information from skb->data to skb->vlan_tci, but the information contained is not otherwise changing. All your examples should work regardless of where the tag is stored.

Quoted 6 lines
Quoted 3 lines

I assumed (but didn't tested) that this untagging also change the starting point of the payload of the packet. So protocol handlers expecting to have the raw packet won't see the vlan header.

That would also be the case with hardware stripped tags - they need to look into skb->vlan_tci anyway.

Exactly. Nicolas, I do not see anything wrong on always untagging in all your setups. As Michal said, vlan_tci keeps the info.

Jirka

Quoted 3 lines

Best Regards,
Michał Mirosław

#16 lore raw
Quoted 67 lines · Jiri Pirko
Quoted 66 lines · Sun, May 22, 2011 at 11:20:09AM CEST, [email protected] wrote:

W dniu 22 maja 2011 11:10 użytkownik Nicolas de Pesloüan

Quoted 63 lines · <[email protected]> napisał:
Quoted 54 lines · Michał Mirosław

2011/5/22 Nicolas de Pesloüan<[email protected]>:

Quoted 43 lines · Eric W. Biederman
Quoted 16 lines
Quoted 13 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

bridge is another. More precisely, there is a difference between the following two setups:

1/ eth0 - eth0.100 - br0 - eth1.200 - eth1

2/ eth0 - br0 - eth1

In case 1, it is normal and desirable for the bridge to see untagged skb.

In case 2, it is desirable for the bridge to see untouched (possibly tagged)
skb. If current bridge implementation is able to handle skb from which we removed a tag, in this situation, it means that bridge currently "fix improper untagging" by itself, by forcing re-tagging on output. I think is
should not be the job of protocol handlers to fix this. Again, a generic feature should to it when necessary.

Think of the following setups:

Quoted 8 lines

3/ eth0 - br0 - eth1.200 - eth1.
4/ eth0 - eth0.100 - br0 - eth1

What if one expect this setup to add (3) or remove (4) one level of vlan nesting? This is precisely what this setup suggest. How can we instruct the
bridge to do so? It is not the bridge responsibility to do any vlan processing. bridge is expected to... bridge !

Quoted 4 lines

I assumed that this untaging Jiri is implementing does not remove the tag. It moves the information from skb->data to skb->vlan_tci, but the information contained is not otherwise changing. All your examples should work regardless of where the tag is stored.

I assumed (but didn't tested) that this untagging also change the starting point of the payload of the packet. So protocol handlers expecting to have the raw packet won't see the vlan header.

Quoted 5 lines
Quoted 2 lines

That would also be the case with hardware stripped tags - they need to look into skb->vlan_tci anyway.

Exactly. Nicolas, I do not see anything wrong on always untagging in all your setups. As Michal said, vlan_tci keeps the info.

I understand this.

But I don't understand how the bridge code is expected to know whether it should re-tag the packet or not before forwarding and which value to use as the egress vlan tag.

1/ eth0 - br0 - eth1 : the bridge is expected to retag using skb->vlan_tci value.

2/ eth0 - eth0.100 - br0 - eth1.200 - eth1 : the bridge is expected to retag using a different value than skb->vlan_tci.

3/ eth0 - eth0.100 - br0 - eth1 : the bridge is expected not to re-tag, because the expected behavior of this setup is to untag while crossing the bridge.

4/ eth0 - eth0.100 - eth0.100.300 - br0 - eth1.400 - eth1.200 - eth1 : the bridge is expected to retag using a different value than skb->vlan_tci. What value would skb->vlan_tci hold when the skb will be delivered to the bridge? 100 or 300?

From my point of view, in both setup, the bridge will receive a single value in skb->vlan_tci and will lack any other indication to help it decide how to retag when forwarding.

I'm not against your idea to mimic hw-accel in software. But I'm concern about troubles for those who expect to have access to untouched packet. Your patch didn't cause those troubles to start happening, but cause them to always happen. Before your patch, someone had the option to use non-hw-accel NIC or to disable hw-accel feature if possible. Now, it's no more possible.

Nicolas.

#17 lore raw

2011/5/22 Nicolas de Pesloüan <[email protected]>:

Quoted 35 lines · Jiri Pirko
Quoted 13 lines
Quoted 12 lines · Sun, May 22, 2011 at 11:20:09AM CEST, [email protected] wrote:

W dniu 22 maja 2011 11:10 użytkownik Nicolas de Pesloüan

Quoted 10 lines · <[email protected]>  napisał:
Quoted 4 lines · Michał Mirosław

I assumed that this untaging Jiri is implementing does not remove the tag. It moves the information from skb->data to skb->vlan_tci, but the information contained is not otherwise changing. All your examples should work regardless of where the tag is stored.

I assumed (but didn't tested) that this untagging also change the starting
point of the payload of the packet. So protocol handlers expecting to have
the raw packet won't see the vlan header.

Quoted 4 lines
Quoted 2 lines

That would also be the case with hardware stripped tags - they need to look into skb->vlan_tci anyway.

Exactly. Nicolas, I do not see anything wrong on always untagging in all your setups. As Michal said, vlan_tci keeps the info.

I understand this.

But I don't understand how the bridge code is expected to know whether it should re-tag the packet or not before forwarding and which value to use as the egress vlan tag.

1/ eth0 - br0 - eth1 : the bridge is expected to retag using skb->vlan_tci value.

2/ eth0 - eth0.100 - br0 - eth1.200 - eth1 : the bridge is expected to retag using a different value than skb->vlan_tci.

3/ eth0 - eth0.100 - br0 - eth1 : the bridge is expected not to re-tag, because the expected behavior of this setup is to untag while crossing the bridge.

Quoted 8 lines

4/ eth0 - eth0.100 - eth0.100.300 - br0 - eth1.400 - eth1.200 - eth1 : the bridge is expected to retag using a different value than skb->vlan_tci. What value would skb->vlan_tci hold when the skb will be delivered to the bridge? 100 or 300?

From my point of view, in both setup, the bridge will receive a single value in skb->vlan_tci and will lack any other indication to help it decide how to retag when forwarding.

Packets looking like they came from eth0.100 will have skb->vlan_tci cleared (like taking packet out of a tunnel) and then possibly filled again with inner tag. It's really convenient to thing of VLANs as tunnels.

Best Regards,
Michał Mirosław

#18 lore raw
In reply to Eric W. Biederman
Quoted 14 lines · Eric W. Biederman
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

There are some drivers still using the old vlan model that will drop tags or packets when no vlan group is configured but that's a driver problem, not one with networking core or tcpdump.

#19 lore raw
Quoted 16 lines · Jesse Gross
Quoted 14 lines · Eric W. Biederman
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

There are some drivers still using the old vlan model that will drop tags or packets when no vlan group is configured but that's a driver problem, not one with networking core or tcpdump.

On receive if we have stripped the vlan header and then we go to deliver the interrupt to a pf_packet socket (on a non-vlan interface) before or as part of the deliver of the packet to user space we need to re-add the vlan header. Additionally the socket filter on a pf_packet socket needs to behave as though we have a vlan header.

So no I am not talking about anything that is driver specific. I am talking about reasonable userspace expectations. Because otherwise we simply loose the information that a packet was vlan tagged, and in doing so we break existing userspace applications because of our bugs.

Eric

#20 lore raw
Quoted 27 lines · [email protected] (Eric W. Biederman) writes:
Quoted 16 lines · Jesse Gross
Quoted 14 lines · Eric W. Biederman
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

tcpdump on the non-vlan interface for one.

There are some drivers still using the old vlan model that will drop tags or packets when no vlan group is configured but that's a driver problem, not one with networking core or tcpdump.

Quoted 5 lines

On receive if we have stripped the vlan header and then we go to deliver the interrupt to a pf_packet socket (on a non-vlan interface) before or as part of the deliver of the packet to user space we need to re-add the vlan header. Additionally the socket filter on a pf_packet socket needs to behave as though we have a vlan header.

Hmm. Taking a second look the pf_packet code and with hardware vlan header removal isn't completely broken. It is possible to receive packet auxdata and get the information from the vlan header.

It stills seems like a pretty messed up interface to me. Especially since the socket filter can't get at the information in the stripped vlan header, but that is another matter entirely.

Eric

#21 lore raw

PATCH 1/3 vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Jesse Gross

Simplify the vlan handling code by not supporing clearing of VLAN_FLAG_REORDER_HDR. Which means we always make the vlan handling code strip the vlan header from the packets, and always insert the vlan header when transmitting packets.

Not stripping the vlan header has alwasy been broken in combination with vlan hardware accelleration. Now that we are making everything look like accelerated vlan handling not stripping the vlan header is always broken.

I don't think anyone actually cares so simply stop supporting the broken case.

---
 net/8021q/vlan_dev.c     |    3 +--
 net/8021q/vlan_netlink.c |    4 +---
 2 files changed, 2 insertions(+), 5 deletions(-)
net/8021q/vlan_dev.c +1 -2
index f247f5b..20629fe 100644
@@ -242,8 +242,7 @@ int vlan_dev_change_flags(const struct net_device *dev, u32 flags, u32 mask)
 	struct vlan_dev_info *vlan = vlan_dev_info(dev);
 	u32 old_flags = vlan->flags;
 
-	if (mask & ~(VLAN_FLAG_REORDER_HDR | VLAN_FLAG_GVRP |
-		     VLAN_FLAG_LOOSE_BINDING))
+	if (mask & ~(VLAN_FLAG_GVRP | VLAN_FLAG_LOOSE_BINDING))
 		return -EINVAL;
 
 	vlan->flags = (old_flags & ~mask) | (flags & mask);
net/8021q/vlan_netlink.c +1 -3
index be9a5c1..a638a4c 100644
@@ -59,9 +59,7 @@ static int vlan_validate(struct nlattr *tb[], struct nlattr *data[])
 	}
 	if (data[IFLA_VLAN_FLAGS]) {
 		flags = nla_data(data[IFLA_VLAN_FLAGS]);
-		if ((flags->flags & flags->mask) &
-		    ~(VLAN_FLAG_REORDER_HDR | VLAN_FLAG_GVRP |
-		      VLAN_FLAG_LOOSE_BINDING))
+		if (flags->mask & ~(VLAN_FLAG_GVRP | VLAN_FLAG_LOOSE_BINDING))
 			return -EINVAL;
 	}
 
Signature · 1 line
1.7.5.1.217.g4e3aa
#22 lore raw

PATCH 2/3 vlan: Always strip the vlan header in vlan_untag.

Calling vlan_dev_info() on a non-vlan device gives completely undefined results, and is no longer needed now that we no longer support keeping the vlan tag.

Simplify and correct the code by remove the check for VLAN_FLAG_REORDER_HDR.

Additionally rename vlan_check_reorder_header to just vlan_reorder_header as there is no check involved now.

---
 net/8021q/vlan_core.c |   20 +++++++++-----------
 1 files changed, 9 insertions(+), 11 deletions(-)
net/8021q/vlan_core.c +9 -11
index 41495dc..302d8e3 100644
@@ -89,17 +89,15 @@ gro_result_t vlan_gro_frags(struct napi_struct *napi, struct vlan_group *grp,
 }
 EXPORT_SYMBOL(vlan_gro_frags);
 
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)
 {
-	if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-		if (skb_cow(skb, skb_headroom(skb)) < 0)
-			skb = NULL;
-		if (skb) {
-			/* Lifted from Gleb's VLAN code... */
-			memmove(skb->data - ETH_HLEN,
-				skb->data - VLAN_ETH_HLEN, 12);
-			skb->mac_header += VLAN_HLEN;
-		}
+	if (skb_cow(skb, skb_headroom(skb)) < 0)
+		skb = NULL;
+	if (skb) {
+		/* Lifted from Gleb's VLAN code... */
+		memmove(skb->data - ETH_HLEN,
+			skb->data - VLAN_ETH_HLEN, 12);
+		skb->mac_header += VLAN_HLEN;
 	}
 	return skb;
 }
@@ -161,7 +159,7 @@ struct sk_buff *vlan_untag(struct sk_buff *skb)
 	skb_pull_rcsum(skb, VLAN_HLEN);
 	vlan_set_encap_proto(skb, vhdr);
 
-	skb = vlan_check_reorder_header(skb);
+	skb = vlan_reorder_header(skb);
 	if (unlikely(!skb))
 		goto err_free;
 
Signature · 1 line
1.7.5.1.217.g4e3aa
#23 lore raw

PATCH 3/3 vlan: Simplify the code now that VLAN_FLAG_REORDER_HDR is always set

Now that we no longer support clearing VLAN_FLAG_REORDER_HDR remove the code that was needed to cope with the case when it was cleared.

---
 net/8021q/vlan_dev.c |   45 +++++----------------------------------------
 1 files changed, 5 insertions(+), 40 deletions(-)
net/8021q/vlan_dev.c +5 -40
index 20629fe..2b3ca1e 100644
@@ -96,63 +96,28 @@ static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
 				const void *daddr, const void *saddr,
 				unsigned int len)
 {
-	struct vlan_hdr *vhdr;
-	unsigned int vhdrlen = 0;
-	u16 vlan_tci = 0;
 	int rc;
 
-	if (!(vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR)) {
-		vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
-
-		vlan_tci = vlan_dev_info(dev)->vlan_id;
-		vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
-		vhdr->h_vlan_TCI = htons(vlan_tci);
-
-		/*
-		 *  Set the protocol type. For a packet of type ETH_P_802_3/2 we
-		 *  put the length in here instead.
-		 */
-		if (type != ETH_P_802_3 && type != ETH_P_802_2)
-			vhdr->h_vlan_encapsulated_proto = htons(type);
-		else
-			vhdr->h_vlan_encapsulated_proto = htons(len);
-
-		skb->protocol = htons(ETH_P_8021Q);
-		type = ETH_P_8021Q;
-		vhdrlen = VLAN_HLEN;
-	}
-
 	/* Before delegating work to the lower layer, enter our MAC-address */
 	if (saddr == NULL)
 		saddr = dev->dev_addr;
 
 	/* Now make the underlying real hard header */
 	dev = vlan_dev_info(dev)->real_dev;
-	rc = dev_hard_header(skb, dev, type, daddr, saddr, len + vhdrlen);
-	if (rc > 0)
-		rc += vhdrlen;
+	rc = dev_hard_header(skb, dev, type, daddr, saddr, len);
 	return rc;
 }
 
 static netdev_tx_t vlan_dev_hard_start_xmit(struct sk_buff *skb,
 					    struct net_device *dev)
 {
-	struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
 	unsigned int len;
+	u16 vlan_tci;
 	int ret;
 
-	/* Handle non-VLAN frames if they are sent to us, for example by DHCP.
-	 *
-	 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
-	 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
-	 */
-	if (veth->h_vlan_proto != htons(ETH_P_8021Q) ||
-	    vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR) {
-		u16 vlan_tci;
-		vlan_tci = vlan_dev_info(dev)->vlan_id;
-		vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
-		skb = __vlan_hwaccel_put_tag(skb, vlan_tci);
-	}
+	vlan_tci = vlan_dev_info(dev)->vlan_id;
+	vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
+	skb = __vlan_hwaccel_put_tag(skb, vlan_tci);
 
 	skb_set_dev(skb, vlan_dev_info(dev)->real_dev);
 	len = skb->len;
Signature · 1 line
1.7.5.1.217.g4e3aa
author
#24 lore raw

Re: [PATCH 3/3] vlan: Simplify the code now that VLAN_FLAG_REORDER_HDR is always set

Quoted 60 lines · Sun, May 22, 2011 at 09:42:20PM CEST, [email protected] wrote:

Now that we no longer support clearing VLAN_FLAG_REORDER_HDR remove the code that was needed to cope with the case when it was cleared.

---
net/8021q/vlan_dev.c | 45 +++++----------------------------------------

1 files changed, 5 insertions(+), 40 deletions(-)
net/8021q/vlan_dev.c +1 -28
index 20629fe..2b3ca1e 100644
@@ -96,63 +96,28 @@ static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
 				const void *daddr, const void *saddr,
 				unsigned int len)
 {
-	struct vlan_hdr *vhdr;
-	unsigned int vhdrlen = 0;
-	u16 vlan_tci = 0;
 	int rc;
 
-	if (!(vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR)) {
-		vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
-
-		vlan_tci = vlan_dev_info(dev)->vlan_id;
-		vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
-		vhdr->h_vlan_TCI = htons(vlan_tci);
-
-		/*
-		 *  Set the protocol type. For a packet of type ETH_P_802_3/2 we
-		 *  put the length in here instead.
-		 */
-		if (type != ETH_P_802_3 && type != ETH_P_802_2)
-			vhdr->h_vlan_encapsulated_proto = htons(type);
-		else
-			vhdr->h_vlan_encapsulated_proto = htons(len);
-
-		skb->protocol = htons(ETH_P_8021Q);
-		type = ETH_P_8021Q;
-		vhdrlen = VLAN_HLEN;
-	}
-
 	/* Before delegating work to the lower layer, enter our MAC-address */
 	if (saddr == NULL)
 		saddr = dev->dev_addr;
 
 	/* Now make the underlying real hard header */
 	dev = vlan_dev_info(dev)->real_dev;
-	rc = dev_hard_header(skb, dev, type, daddr, saddr, len + vhdrlen);
-	if (rc > 0)
-		rc += vhdrlen;
+	rc = dev_hard_header(skb, dev, type, daddr, saddr, len);
 	return rc;
 }
 
 static netdev_tx_t vlan_dev_hard_start_xmit(struct sk_buff *skb,
 					    struct net_device *dev)
 {
-	struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);

net/8021q/vlan_dev.c static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev, · line 149

Quoted 10 lines
net/8021q/vlan_dev.c +1 -6
 	unsigned int len;
+	u16 vlan_tci;
 	int ret;
 
-	/* Handle non-VLAN frames if they are sent to us, for example by DHCP.
-	 *
-	 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
-	 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
-	 */
-	if (veth->h_vlan_proto != htons(ETH_P_8021Q) ||
	    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	    		this should stay here.
Quoted 15 lines
net/8021q/vlan_dev.c +3 -6
-	    vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR) {
-		u16 vlan_tci;
-		vlan_tci = vlan_dev_info(dev)->vlan_id;
-		vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
-		skb = __vlan_hwaccel_put_tag(skb, vlan_tci);
-	}
+	vlan_tci = vlan_dev_info(dev)->vlan_id;
+	vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
+	skb = __vlan_hwaccel_put_tag(skb, vlan_tci);
 
 	skb_set_dev(skb, vlan_dev_info(dev)->real_dev);
 	len = skb->len;
Signature · 1 line
1.7.5.1.217.g4e3aa
#25 lore raw

Re: [PATCH 3/3] vlan: Simplify the code now that VLAN_FLAG_REORDER_HDR is always set

Quoted 65 lines · Jiri Pirko
Quoted 64 lines · Sun, May 22, 2011 at 09:42:20PM CEST, [email protected] wrote:

Now that we no longer support clearing VLAN_FLAG_REORDER_HDR remove the code that was needed to cope with the case when it was cleared.

---
net/8021q/vlan_dev.c | 45 +++++----------------------------------------

1 files changed, 5 insertions(+), 40 deletions(-)
net/8021q/vlan_dev.c +2 -28
index 20629fe..2b3ca1e 100644
@@ -96,63 +96,28 @@ static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev,
 				const void *daddr, const void *saddr,
 				unsigned int len)
 {
-	struct vlan_hdr *vhdr;
-	unsigned int vhdrlen = 0;
-	u16 vlan_tci = 0;
 	int rc;
 
-	if (!(vlan_dev_info(dev)->flags & VLAN_FLAG_REORDER_HDR)) {
-		vhdr = (struct vlan_hdr *) skb_push(skb, VLAN_HLEN);
-
-		vlan_tci = vlan_dev_info(dev)->vlan_id;
-		vlan_tci |= vlan_dev_get_egress_qos_mask(dev, skb);
-		vhdr->h_vlan_TCI = htons(vlan_tci);
-
-		/*
-		 *  Set the protocol type. For a packet of type ETH_P_802_3/2 we
-		 *  put the length in here instead.
-		 */
-		if (type != ETH_P_802_3 && type != ETH_P_802_2)
-			vhdr->h_vlan_encapsulated_proto = htons(type);
-		else
-			vhdr->h_vlan_encapsulated_proto = htons(len);
-
-		skb->protocol = htons(ETH_P_8021Q);
-		type = ETH_P_8021Q;
-		vhdrlen = VLAN_HLEN;
-	}
-
 	/* Before delegating work to the lower layer, enter our MAC-address */
 	if (saddr == NULL)
 		saddr = dev->dev_addr;
 
 	/* Now make the underlying real hard header */
 	dev = vlan_dev_info(dev)->real_dev;
-	rc = dev_hard_header(skb, dev, type, daddr, saddr, len + vhdrlen);
-	if (rc > 0)
-		rc += vhdrlen;
+	rc = dev_hard_header(skb, dev, type, daddr, saddr, len);
 	return rc;
 }
 
 static netdev_tx_t vlan_dev_hard_start_xmit(struct sk_buff *skb,
 					    struct net_device *dev)
 {
-	struct vlan_ethhdr *veth = (struct vlan_ethhdr *)(skb->data);
 	unsigned int len;
+	u16 vlan_tci;
 	int ret;
 
Quoted 9 lines

net/8021q/vlan_dev.c static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev, · line 149

Quoted 6 lines
net/8021q/vlan_dev.c +0 -6
-	/* Handle non-VLAN frames if they are sent to us, for example by DHCP.
-	 *
-	 * NOTE: THIS ASSUMES DIX ETHERNET, SPECIFICALLY NOT SUPPORTING
-	 * OTHER THINGS LIKE FDDI/TokenRing/802.3 SNAPs...
-	 */
-	if (veth->h_vlan_proto != htons(ETH_P_8021Q) ||
	    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	    		this should stay here.

I admit this is a change in behavior from today so we need to be careful here.

At least the comment needs to change.

When this code was written the assumption was that everything would come in tagged and this code was a special exception for pf_packet sockets.

Now everything comes in untagged this code becomes a special exception for pf_packet sockets of not putting on a vlan header.

To me this special exception pf_packet sockets looks like a bug, that should have been conditionality on VLAN_FLAG_REORDER_HDR but was over looked.

Now maybe we want to be bug compatible, or do this as a separate patch.

I would expect sending tagged packets out a vlan device would result in double tagged packets but apparently not always.

Eric

#26 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Eric W. Biederman
Quoted 6 lines · Eric W. Biederman

Simplify the vlan handling code by not supporing clearing of VLAN_FLAG_REORDER_HDR. Which means we always make the vlan handling code strip the vlan header from the packets, and always insert the vlan header when transmitting packets.

Quoted 7 lines

Not stripping the vlan header has alwasy been broken in combination with vlan hardware accelleration. Now that we are making everything look like accelerated vlan handling not stripping the vlan header is always broken.

I don't think anyone actually cares so simply stop supporting the broken case.

I've lost track of the VLAN code a bit. Is there any documentation somewhere about what happens in these various cases:

* Open a raw packet socket on eth0.
* Do we get tagged VLAN packets? (I'd expect yes.)
* If we sent a tagged VLAN packet, it's sent without modification? (I'd expect yes.) ** Without "yes" to the two above, one cannot do user-space bridging properly.

* Open a raw packet socket on VLAN eth0.5

  *  Do we get tagged VLAN packets?  (I'd expect no.)
  *  If we send an un-tagged packet, I expect it would be tagged?
  *  What if we sent a tagged packet with same VID?
  *  With different VID?

Many years ago we supported the REORDER, but we suggested disabling it for most users because it was a performance drag. Funny that now it seems to be the opposite!

Thanks,
Ben

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#27 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 12 lines · Ben Greear
Quoted 11 lines · Eric W. Biederman

Simplify the vlan handling code by not supporing clearing of VLAN_FLAG_REORDER_HDR. Which means we always make the vlan handling code strip the vlan header from the packets, and always insert the vlan header when transmitting packets.

Not stripping the vlan header has alwasy been broken in combination with vlan hardware accelleration. Now that we are making everything look like accelerated vlan handling not stripping the vlan header is always broken.

Quoted 5 lines
Quoted 2 lines

I don't think anyone actually cares so simply stop supporting the broken case.

I've lost track of the VLAN code a bit. Is there any documentation somewhere about what happens in these various cases:

Other than the code I don't know about documentation.

Quoted 1 line

* Open a raw packet socket on eth0.

I assume you mean a pf_packet socket.

Quoted 1 line

* Do we get tagged VLAN packets? (I'd expect yes.)

yes.

Quoted 2 lines

* If we sent a tagged VLAN packet, it's sent without modification? (I'd expect yes.) ** Without "yes" to the two above, one cannot do user-space bridging properly.

This is sort of. If you set the PACKET_AUXDATA option and use recv_cmsg you get the priority and the vlan identifier in the auxdata.

I think that is a pretty horrible answer myself but it has been that way since sometime mid 2008. So I'm not immediately prepared to call it a regression, or a bug.

Quoted 2 lines

* Open a raw packet socket on VLAN eth0.5
* Do we get tagged VLAN packets? (I'd expect no.)

No.

Quoted 1 line

* If we send an un-tagged packet, I expect it would be tagged?

Yes.

Quoted 2 lines

* What if we sent a tagged packet with same VID?
* With different VID?

The vlan tag is ignored and another vlan tag is added.

Quoted 3 lines

Many years ago we supported the REORDER, but we suggested disabling it for most users because it was a performance drag. Funny that now it seems to be the opposite!

Yes it is funny. I looked in history a while back and what I saw was that REORDER was always enabled by default and it took some serious effort to figure out how to get vconfig to disable REORDER. ip doesn't admit that REORDER can be disabled at all.

Eric

#28 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 10 lines · Eric W. Biederman
Quoted 4 lines

Many years ago we supported the REORDER, but we suggested disabling it for most users because it was a performance drag.  Funny that now it seems to be the opposite!

Yes it is funny.  I looked in history a while back and what I saw was that REORDER was always enabled by default and it took some serious effort to figure out how to get vconfig to disable REORDER. ip doesn't admit that REORDER can be disabled at all.

Really?

Quoted from the manual page of vconfig

       set_flag [vlan-device] 0 | 1

When 1, ethernet header reorders are turned on. Dumping the device will appear as a common ethernet device without vlans. When 0(default) however, ethernet headers are not reordered, which results in vlan tagged packets when dumping the device. Usually the default gives no problems, but some packet filtering programs might have problems with it.

reordered is disabled by default. I also concern the performance. Untag and then tag are expensive for the NICs which don't support hw-accel-vlan-rx/tx.

Signature · 2 lines
Regards,
Changli Gao([email protected])
#29 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 24 lines · Changli Gao
Quoted 10 lines · Eric W. Biederman
Quoted 4 lines

Many years ago we supported the REORDER, but we suggested disabling it for most users because it was a performance drag.  Funny that now it seems to be the opposite!

Yes it is funny.  I looked in history a while back and what I saw was that REORDER was always enabled by default and it took some serious effort to figure out how to get vconfig to disable REORDER. ip doesn't admit that REORDER can be disabled at all.

Really?

Quoted from the manual page of vconfig
      set_flag [vlan-device] 0 | 1
             When  1,  ethernet  header  reorders  are turned on. Dumping the device will appear as a common ethernet  device  without  vlans. When  0(default)  however,  ethernet  headers are not reordered, which results in vlan tagged packets when  dumping  the  device. Usually the default gives no problems, but some packet filtering programs might have problems with it.

Quoted 4 lines

reordered is disabled by default. I also concern the performance. Untag and then tag are expensive for the NICs which don't support hw-accel-vlan-rx/tx.

For ip:
localhost ~ # ip link add link eth0 vlan1 type vlan help
Usage: ... vlan id VLANID [ FLAG-LIST ]
[ ingress-qos-map QOS-MAP ] [ egress-qos-map QOS-MAP ]

VLANID := 0-4095
FLAG-LIST := [ FLAG-LIST ] FLAG
FLAG := [ reorder_hdr { on | off } ] [ gvrp { on | off } ]
[ loose_binding { on | off } ]
QOS-MAP := [ QOS-MAP ] QOS-MAPPING
QOS-MAPPING := FROM:TO

After checking the code, I found reorder_hdr is off by default.

In another side, is there a specification which defines the hw-accel-vlan-rx?

Signature · 2 lines
Regards,
Changli Gao([email protected])
#30 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 30 lines · Changli Gao
Quoted 24 lines · Changli Gao
Quoted 10 lines · Eric W. Biederman
Quoted 4 lines

Many years ago we supported the REORDER, but we suggested disabling it for most users because it was a performance drag.  Funny that now it seems to be the opposite!

Yes it is funny.  I looked in history a while back and what I saw was that REORDER was always enabled by default and it took some serious effort to figure out how to get vconfig to disable REORDER. ip doesn't admit that REORDER can be disabled at all.

Really?

Quoted from the manual page of vconfig
      set_flag [vlan-device] 0 | 1
             When  1,  ethernet  header  reorders  are turned on. Dumping the device will appear as a common ethernet  device  without  vlans. When  0(default)  however,  ethernet  headers are not reordered, which results in vlan tagged packets when  dumping  the  device. Usually the default gives no problems, but some packet filtering programs might have problems with it.

Quoted 4 lines

reordered is disabled by default. I also concern the performance. Untag and then tag are expensive for the NICs which don't support hw-accel-vlan-rx/tx.

Quoted 4 lines

For ip:
localhost ~ # ip link add link eth0 vlan1 type vlan help
Usage: ... vlan id VLANID [ FLAG-LIST ]
[ ingress-qos-map QOS-MAP ] [ egress-qos-map QOS-MAP ]

Apparently I was blind when I looked at iproute. I am certain I didn't see it there but iproute clearly has the option to set or clear reorder_hdr.

Quoted 8 lines

VLANID := 0-4095
FLAG-LIST := [ FLAG-LIST ] FLAG
FLAG := [ reorder_hdr { on | off } ] [ gvrp { on | off } ]
[ loose_binding { on | off } ]
QOS-MAP := [ QOS-MAP ] QOS-MAPPING
QOS-MAPPING := FROM:TO

After checking the code, I found reorder_hdr is off by default.

In iproute reorder_hdr is not modified by default which is subtly different.

Quoted 2 lines

In another side, is there a specification which defines the hw-accel-vlan-rx?

I don't know.

I have just been trying to clean up the mess since some of the hw-accel-vlan code broke my use case, by delivering packets with priority but no vlan (aka vlan 0 packets) twice to my pf_packet sockets.

Eric

#31 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 9 lines · Eric W. Biederman
Quoted 2 lines

In another side, is there a specification which defines the hw-accel-vlan-rx?

I don't know.

I have just been trying to clean up the mess since some of the hw-accel-vlan code broke my use case, by delivering packets with priority but no vlan (aka vlan 0 packets) twice to my pf_packet sockets.

OK. But if we have decided to simulate the hw-accel-vlan-rx, I think we'd better adjust the place where we put the emulation code. The very beginnings of netif_rx() and neif_receive_skb() are better. Then rps can support vlan packets without any change.

Signature · 2 lines
Regards,
Changli Gao([email protected])
#32 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 12 lines · Changli Gao
Quoted 9 lines · Eric W. Biederman
Quoted 2 lines

In another side, is there a specification which defines the hw-accel-vlan-rx?

I don't know.

I have just been trying to clean up the mess since some of the hw-accel-vlan code broke my use case, by delivering packets with priority but no vlan (aka vlan 0 packets) twice to my pf_packet sockets.

Quoted 4 lines

OK. But if we have decided to simulate the hw-accel-vlan-rx, I think we'd better adjust the place where we put the emulation code. The very beginnings of netif_rx() and neif_receive_skb() are better. Then rps can support vlan packets without any change.

That sounds nice. Patches are welcome.

In principle it should be doable with some code motion. I don't think moving vlan_untag earlier constitutes a bug fix.

In my investigation earlier I found a non-trivial number of paths into __netif_receive_skb. So it was not clear to me in the slightest how to move the check earlier without modifying every networking driver and a few other pieces of code.

Why should receive packet steering be affected by vlan tags at all?

Eric

author
#33 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 19 lines · Mon, May 23, 2011 at 11:41:22AM CEST, [email protected] wrote:
Quoted 12 lines · Changli Gao
Quoted 9 lines · Eric W. Biederman
Quoted 2 lines

In another side, is there a specification which defines the hw-accel-vlan-rx?

I don't know.

I have just been trying to clean up the mess since some of the hw-accel-vlan code broke my use case, by delivering packets with priority but no vlan (aka vlan 0 packets) twice to my pf_packet sockets.

Quoted 4 lines

OK. But if we have decided to simulate the hw-accel-vlan-rx, I think we'd better adjust the place where we put the emulation code. The very beginnings of netif_rx() and neif_receive_skb() are better. Then rps can support vlan packets without any change.

Quoted 4 lines

That sounds nice. Patches are welcome.

In principle it should be doable with some code motion. I don't think moving vlan_untag earlier constitutes a bug fix.

I do not think that is doable. Consider multi tagged packets. The place just after "another_round" takes care about that.

Btw what's the rationale to move untag to earlier position?

Quoted 10 lines

In my investigation earlier I found a non-trivial number of paths into __netif_receive_skb. So it was not clear to me in the slightest how to move the check earlier without modifying every networking driver and a few other pieces of code.

Why should receive packet steering be affected by vlan tags at all?

Eric

#34 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 22 lines · Jiri Pirko
Quoted 21 lines · Mon, May 23, 2011 at 11:41:22AM CEST, [email protected] wrote:
Quoted 12 lines · Changli Gao
Quoted 9 lines · Eric W. Biederman
Quoted 2 lines

In another side, is there a specification which defines the hw-accel-vlan-rx?

I don't know.

I have just been trying to clean up the mess since some of the hw-accel-vlan code broke my use case, by delivering packets with priority but no vlan (aka vlan 0 packets) twice to my pf_packet sockets.

Quoted 4 lines

OK. But if we have decided to simulate the hw-accel-vlan-rx, I think we'd better adjust the place where we put the emulation code. The very beginnings of netif_rx() and neif_receive_skb() are better. Then rps can support vlan packets without any change.

That sounds nice. Patches are welcome.

Quoted 7 lines
Quoted 2 lines

In principle it should be doable with some code motion. I don't think moving vlan_untag earlier constitutes a bug fix.

I do not think that is doable. Consider multi tagged packets. The place just after "another_round" takes care about that.

Btw what's the rationale to move untag to earlier position?

Maybe simply because we try to mimic hw-accel, and hw-accel untagging definitely happens before we enter __netif_receive_skb and only happens once.

So having software untagging inside the __netif_receive_skb loop looks different.

Nicolas.

author
#35 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 29 lines · Mon, May 23, 2011 at 09:48:12PM CEST, [email protected] wrote:
Quoted 28 lines · Jiri Pirko
Quoted 19 lines · Mon, May 23, 2011 at 11:41:22AM CEST, [email protected] wrote:
Quoted 12 lines · Changli Gao
Quoted 9 lines · Eric W. Biederman
Quoted 2 lines

In another side, is there a specification which defines the hw-accel-vlan-rx?

I don't know.

I have just been trying to clean up the mess since some of the hw-accel-vlan code broke my use case, by delivering packets with priority but no vlan (aka vlan 0 packets) twice to my pf_packet sockets.

Quoted 4 lines

OK. But if we have decided to simulate the hw-accel-vlan-rx, I think we'd better adjust the place where we put the emulation code. The very beginnings of netif_rx() and neif_receive_skb() are better. Then rps can support vlan packets without any change.

Quoted 4 lines

That sounds nice. Patches are welcome.

In principle it should be doable with some code motion. I don't think moving vlan_untag earlier constitutes a bug fix.

I do not think that is doable. Consider multi tagged packets. The place just after "another_round" takes care about that.

Quoted 7 lines
Quoted 1 line

Btw what's the rationale to move untag to earlier position?

Maybe simply because we try to mimic hw-accel, and hw-accel untagging definitely happens before we enter __netif_receive_skb and only happens once.

So having software untagging inside the __netif_receive_skb loop looks different.

I understand. But what code prior to current vlan_untag position needs to see the skb untagged?

Quoted 2 lines

Nicolas.

#36 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Le 24/05/2011 07:58, Jiri Pirko a écrit :
<snip>

Quoted 10 lines
Quoted 7 lines
Quoted 1 line

Btw what's the rationale to move untag to earlier position?

Maybe simply because we try to mimic hw-accel, and hw-accel untagging definitely happens before we enter __netif_receive_skb and only happens once.

So having software untagging inside the __netif_receive_skb loop looks different.

I understand. But what code prior to current vlan_untag position needs to see the skb untagged?

Any protocol handlers (ptype_all or ptype_base) registered on the parent interface may need to see the skb untagged, for all the reasons given in this thread. Arguably, doing software untagging earlier wouldn't help. :-) We need a strong logic to decide whether and when to untag and/or possibly re-tag.

Nicolas.

#37 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Changli Gao
Quoted 24 lines · Changli Gao
Quoted 10 lines · Eric W. Biederman
Quoted 4 lines

Many years ago we supported the REORDER, but we suggested disabling it for most users because it was a performance drag.  Funny that now it seems to be the opposite!

Yes it is funny.  I looked in history a while back and what I saw was that REORDER was always enabled by default and it took some serious effort to figure out how to get vconfig to disable REORDER. ip doesn't admit that REORDER can be disabled at all.

Really?

Quoted from the manual page of vconfig

       set_flag [vlan-device] 0 | 1

When 1, ethernet header reorders are turned on. Dumping the device will appear as a common ethernet device without vlans. When 0(default) however, ethernet headers are not reordered, which results in vlan tagged packets when dumping the device. Usually the default gives no problems, but some packet filtering programs might have problems with it.

Quoted 3 lines

reordered is disabled by default. I also concern the performance. Untag and then tag are expensive for the NICs which don't support hw-accel-vlan-rx/tx.

Yes really the vconfig man page is wrong. Reorder has been enabled by default since the code was merged. See below.

Dhcp among other things does not work if you disable reorder.

As for performace we are moving 12 bytes which should be cache hot 4 bytes later in what should be the same cache line. I expect my 16Mhz 386 will slow down a little for an operations like that, I don't expect much else will. I can't see anything short of benchmark numbers being persuasive.

The other reality is that the code is honestly and truly broken in the corner cases. If we don't consolidate the code paths it will never operate correctly.

Not that I agree 100% with all of the decisions but the code has been broken for a while for most users so we really need to make it consistent and fix the bugs.

Furthermore to the best of my knowledge no one actually clears the reorder bit. Changli the fact that you don't know that the reorder bit is set by default suggest that you don't clear it in the cases you are concerned about.

If no one is clearing the reorder bit in practice all that we really have for the vlan changes are code motion and simplification which should only have a positive impact on performance.

Eric

Aka the vlan code came in at:

commit 15a435fe2c0f649e2879e51e05fa466c2bb12731

Author: torvalds <torvalds>
Date:   Tue Feb 5 20:30:01 2002 +0000

v2.4.13.5 -> v2.4.13.6

- me: remember to bump the version number ;)
- Hugh Dickins: export "free_lru_page()" for modules
- Jeff Garzik: don't change nopage arguments, just make the last a dummy one
- David Miller: sparc and net updates (netfilter, VLAN etc)
- Nikita Danilov: reiserfs cleanups
- Jan Kara: quota initialization race
- Tigran Aivazian: make the x86 microcode update driver happy about hyperthreaded P4's
- me: shrink dcache/icache more aggressively
- me: fix up oom-killer so that it actually works

BKrev: 3c6040c9uewuD0S9AwFCfH3_YMyRHQ

[snip]

net/8021q/vlan.c new file +0 -0
new file mode 100644
index 0000000..e549e33

[snip]
+/* DO reorder the header by default */
+unsigned short vlan_default_dev_flags = 1;

[snip]

+/*  Attach a VLAN device to a mac address (ie Ethernet Card).
+ *  Returns the device that was created, or NULL if there was
+ *  an error of some kind.

+ */
+struct net_device *register_802_1Q_vlan_device(const char* eth_IF_name,
+ unsigned short VLAN_ID)
+{
[snip]
+ VLAN_DEV_INFO(new_dev)->vlan_id = VLAN_ID; /* 1 through 0xFFF */
+ VLAN_DEV_INFO(new_dev)->real_dev = real_dev;
+ VLAN_DEV_INFO(new_dev)->dent = NULL;
+ VLAN_DEV_INFO(new_dev)->flags = vlan_default_dev_flags;
[snip]

#38 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Eric W. Biederman
Quoted 14 lines · Eric W. Biederman
Quoted 12 lines · Ben Greear
Quoted 11 lines · Eric W. Biederman

Simplify the vlan handling code by not supporing clearing of VLAN_FLAG_REORDER_HDR. Which means we always make the vlan handling code strip the vlan header from the packets, and always insert the vlan header when transmitting packets.

Not stripping the vlan header has alwasy been broken in combination with vlan hardware accelleration. Now that we are making everything look like accelerated vlan handling not stripping the vlan header is always broken.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I don't think anyone actually cares so simply stop supporting the broken case.

I've lost track of the VLAN code a bit. Is there any documentation somewhere about what happens in these various cases:

Other than the code I don't know about documentation.

These cases are tricky and probably have changed over the years. It would be nice to have it written down somewhere, even if just in comments somewhere in the VLAN code.

Quoted 3 lines
Quoted 2 lines

* Open a raw packet socket on eth0.

I assume you mean a pf_packet socket.

Yes.

Quoted 6 lines
Quoted 1 line

* Do we get tagged VLAN packets? (I'd expect yes.)

yes.

Quoted 2 lines

* If we sent a tagged VLAN packet, it's sent without modification? (I'd expect yes.) ** Without "yes" to the two above, one cannot do user-space bridging properly.

Quoted 6 lines

This is sort of. If you set the PACKET_AUXDATA option and use recv_cmsg you get the priority and the vlan identifier in the auxdata.

I think that is a pretty horrible answer myself but it has been that way since sometime mid 2008. So I'm not immediately prepared to call it a regression, or a bug.

I believe we have been getting tagged VLAN packets properly in our test cases. We would not be creating any VLAN devices in this case, so perhaps the NIC isn't doing any stripping.

To me, it seems like we should get the fully tagged packet without having to go muck with aux-data, though it would be fine if it were *also* in aux-data.

I'll try to test this again this coming week to make sure it's working like I think it is.

Thanks,
Ben

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#39 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 47 lines · Ben Greear
Quoted 14 lines · Eric W. Biederman
Quoted 12 lines · Ben Greear
Quoted 11 lines · Eric W. Biederman

Simplify the vlan handling code by not supporing clearing of VLAN_FLAG_REORDER_HDR. Which means we always make the vlan handling code strip the vlan header from the packets, and always insert the vlan header when transmitting packets.

Not stripping the vlan header has alwasy been broken in combination with vlan hardware accelleration. Now that we are making everything look like accelerated vlan handling not stripping the vlan header is always broken.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I don't think anyone actually cares so simply stop supporting the broken case.

I've lost track of the VLAN code a bit. Is there any documentation somewhere about what happens in these various cases:

Other than the code I don't know about documentation.

These cases are tricky and probably have changed over the years. It would be nice to have it written down somewhere, even if just in comments somewhere in the VLAN code.

Quoted 3 lines
Quoted 2 lines

* Open a raw packet socket on eth0.

I assume you mean a pf_packet socket.

Yes.

Quoted 6 lines
Quoted 1 line

* Do we get tagged VLAN packets? (I'd expect yes.)

yes.

Quoted 2 lines

* If we sent a tagged VLAN packet, it's sent without modification? (I'd expect yes.) ** Without "yes" to the two above, one cannot do user-space bridging properly.

Quoted 6 lines

This is sort of. If you set the PACKET_AUXDATA option and use recv_cmsg you get the priority and the vlan identifier in the auxdata.

I think that is a pretty horrible answer myself but it has been that way since sometime mid 2008. So I'm not immediately prepared to call it a regression, or a bug.

Quoted 7 lines

I believe we have been getting tagged VLAN packets properly in our test cases. We would not be creating any VLAN devices in this case, so perhaps the NIC isn't doing any stripping.

To me, it seems like we should get the fully tagged packet without having to go muck with aux-data, though it would be fine if it were *also* in aux-data.

Given that pf_packet is a portable interface that works on multiple OS's I tend to agree. Certainly my users would be happier if they don't have to change their code and not having to change tcpdump would also be nice.

I'm not certain exactly where in the code it makes sense to put the vlan header back on for pf_packet sockets. The simplest thing would be just before we run the socket filter. If we don't do the simplest thing this raises the question how do we avoid breaking socket filters that look at the packet data and know there is going to be a vlan header there.

Still the current situation is better than seeing vlan 0 tagged packets twice.

My gut feel says if we can cheaply get the socket filters to act like it sees the vlan tag (where the vlan tag belongs) we should not actually put the vlan tag back on until we copy the packet to userspace.

Having the perspective of someone who cares whose hardware supports the vlan tagging optimizations would be nice to have.

Quoted 2 lines

I'll try to test this again this coming week to make sure it's working like I think it is.

Thanks.

Eric

#40 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 15 lines · Eric W. Biederman
Quoted 7 lines · Ben Greear

I believe we have been getting tagged VLAN packets properly in our test cases. We would not be creating any VLAN devices in this case, so perhaps the NIC isn't doing any stripping.

To me, it seems like we should get the fully tagged packet without having to go muck with aux-data, though it would be fine if it were *also* in aux-data.

Given that pf_packet is a portable interface that works on multiple OS's I tend to agree. Certainly my users would be happier if they don't have to change their code and not having to change tcpdump would also be nice.

Quoted 6 lines

I'm not certain exactly where in the code it makes sense to put the vlan header back on for pf_packet sockets. The simplest thing would be just before we run the socket filter. If we don't do the simplest thing this raises the question how do we avoid breaking socket filters that look at the packet data and know there is going to be a vlan header there.

That is going to be tricky, since the VLAN header would adjust offsets and users could be using some filter that uses offsets with no actual mention of VLANs (but expecting it to take the VLAN header into account).

Quoted 6 lines

Still the current situation is better than seeing vlan 0 tagged packets twice.

My gut feel says if we can cheaply get the socket filters to act like it sees the vlan tag (where the vlan tag belongs) we should not actually put the vlan tag back on until we copy the packet to userspace.

Maybe keep a count of how many sockets with filters and/or pf_packet sockets are open, and how many things are registered in
the 'ptype_all' chain, and only re-add (or never remove) the header if that is > 0?

(And, let the bridging and other kernel logic deal with vlans via auxillary methods as well as checking in-line headers.)

Thanks,
Ben

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#41 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 35 lines · Ben Greear
Quoted 15 lines · Eric W. Biederman
Quoted 7 lines · Ben Greear

I believe we have been getting tagged VLAN packets properly in our test cases. We would not be creating any VLAN devices in this case, so perhaps the NIC isn't doing any stripping.

To me, it seems like we should get the fully tagged packet without having to go muck with aux-data, though it would be fine if it were *also* in aux-data.

Given that pf_packet is a portable interface that works on multiple OS's I tend to agree. Certainly my users would be happier if they don't have to change their code and not having to change tcpdump would also be nice.

Quoted 6 lines

I'm not certain exactly where in the code it makes sense to put the vlan header back on for pf_packet sockets. The simplest thing would be just before we run the socket filter. If we don't do the simplest thing this raises the question how do we avoid breaking socket filters that look at the packet data and know there is going to be a vlan header there.

That is going to be tricky, since the VLAN header would adjust offsets and users could be using some filter that uses offsets with no actual mention of VLANs (but expecting it to take the VLAN header into account).

Quoted 6 lines

Still the current situation is better than seeing vlan 0 tagged packets twice.

My gut feel says if we can cheaply get the socket filters to act like it sees the vlan tag (where the vlan tag belongs) we should not actually put the vlan tag back on until we copy the packet to userspace.

Quoted 6 lines

Maybe keep a count of how many sockets with filters and/or pf_packet sockets are open, and how many things are registered in
the 'ptype_all' chain, and only re-add (or never remove) the header if that is > 0?

(And, let the bridging and other kernel logic deal with vlans via auxillary methods as well as checking in-line headers.)

Well, this doesn't sound very different from my previous proposal: if a protocol handler is registered at parent interface level, can't we simply assume this protocol handler expect the raw packet?

Also, I think the main problem is that ptype_all and ptype_base deliveries are different in __netif_receive_skb, for no good reason (at least, to my knowledge).

In http://patchwork.ozlabs.org/patch/85578/, I proposed to process protocol handlers that are registered on a particular interface inside the loop, then process protocol handlers registered to NULL upon exit from the loop.

That way, we would have the opportunity to deliver different skb (tagged/untagged) to different level of interface, while we cross the interface stack.

I know there is a possible performance penalty in this implementation, but this can be fixed by properly ordering the ptype_all/ptype_base list.

Nicolas.

#42 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 41 lines · Nicolas de Pesloüan
Quoted 40 lines · Ben Greear
Quoted 15 lines · Eric W. Biederman
Quoted 7 lines · Ben Greear

I believe we have been getting tagged VLAN packets properly in our test cases. We would not be creating any VLAN devices in this case, so perhaps the NIC isn't doing any stripping.

To me, it seems like we should get the fully tagged packet without having to go muck with aux-data, though it would be fine if it were *also* in aux-data.

Given that pf_packet is a portable interface that works on multiple OS's I tend to agree. Certainly my users would be happier if they don't have to change their code and not having to change tcpdump would also be nice.

Quoted 6 lines

I'm not certain exactly where in the code it makes sense to put the vlan header back on for pf_packet sockets. The simplest thing would be just before we run the socket filter. If we don't do the simplest thing this raises the question how do we avoid breaking socket filters that look at the packet data and know there is going to be a vlan header there.

That is going to be tricky, since the VLAN header would adjust offsets and users could be using some filter that uses offsets with no actual mention of VLANs (but expecting it to take the VLAN header into account).

Quoted 6 lines

Still the current situation is better than seeing vlan 0 tagged packets twice.

My gut feel says if we can cheaply get the socket filters to act like it sees the vlan tag (where the vlan tag belongs) we should not actually put the vlan tag back on until we copy the packet to userspace.

Maybe keep a count of how many sockets with filters and/or pf_packet sockets are open, and how many things are registered in
the 'ptype_all' chain, and only re-add (or never remove) the header if that is > 0?

Quoted 6 lines
Quoted 2 lines

(And, let the bridging and other kernel logic deal with vlans via auxillary methods as well as checking in-line headers.)

Well, this doesn't sound very different from my previous proposal: if a protocol handler is registered at parent interface level, can't we simply assume this protocol handler expect the raw packet?

Well, yes..unless perhaps when the REORDER_HDR flag is enabled.

As for when the tag is added/removed, as long as we don't end up doing a remove and then an add (in software), and as long as the pkt is correct going to user-space, it doesn't matter to me.

It would be lame to remove it in software and then re-add it, unless absolutely required for some reason.

Thanks,
Ben

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#43 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 9 lines

Well, yes..unless perhaps when the REORDER_HDR flag is enabled.

As for when the tag is added/removed, as long as we don't end up doing a remove and then an add (in software), and as long as the pkt is correct going to user-space, it doesn't matter to me.

It would be lame to remove it in software and then re-add it, unless absolutely required for some reason.

IMHO the REORDER_HDR flag was a design mistake. It means supporting two different API's for the application. For a packet capture application it means the format of the packet will be different based on how the VLAN was created. And the vlan was created outside the application.

If there was a requirement to support both formats, then it should be a property of the AF_PACKET socket. The application can then do an setsockopt() to control the format of the data. The issue is for things like mmap/AF_PACKET the re-inserted form will be slower since data has to be copied.

maintainer
#44 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

From: Stephen Hemminger <[email protected]>
Date: Mon, 23 May 2011 14:00:48 -0700

Quoted 10 lines

IMHO the REORDER_HDR flag was a design mistake. It means supporting two different API's for the application. For a packet capture application it means the format of the packet will be different based on how the VLAN was created. And the vlan was created outside the application.

If there was a requirement to support both formats, then it should be a property of the AF_PACKET socket. The application can then do an setsockopt() to control the format of the data. The issue is for things like mmap/AF_PACKET the re-inserted form will be slower since data has to be copied.

Indeed, it was a foolish thing to support in the first place.

I guess we couldn't see the hw offloads in the future at that point.

I'm all for finding a way to deprecate this over time.

#45 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 14 lines · David Miller

From: Stephen Hemminger <[email protected]>
Date: Mon, 23 May 2011 14:00:48 -0700

Quoted 10 lines

IMHO the REORDER_HDR flag was a design mistake. It means supporting two different API's for the application. For a packet capture application it means the format of the packet will be different based on how the VLAN was created. And the vlan was created outside the application.

If there was a requirement to support both formats, then it should be a property of the AF_PACKET socket. The application can then do an setsockopt() to control the format of the data. The issue is for things like mmap/AF_PACKET the re-inserted form will be slower since data has to be copied.

Quoted 6 lines

Indeed, it was a foolish thing to support in the first place.

I guess we couldn't see the hw offloads in the future at that point.

I'm all for finding a way to deprecate this over time.

There are two very similar issues here, that affect two slightly different cases.

Let's assume the configuration is:
eth0 - Talks to the outside world.
vlan2000 - Is the vlan interface of interest.

With REORDER_HDR set when I tcpdump on vlan2000 I don't see the vlan header, however I continue to see the vlan header when I tcpdump on eth0.

With vlan hardware acceleration. When I tcpdump on eth0 I don't see the vlan header. Nor do I see the vlan header when I tcpdump on vlan2000.

Right now both cases are problematic in Linus's tree.

For inbound packets We are testing the wrong interface for the to see if we should play with REORDER_HDR.

Hardware acceleration vlan tagging we are hiding the vlan header from portable applications that simply dump eth0. Which is non-intuitive and apparent to everyone now that this happens on both software and hardware interfaces.

So we have a couple of questions.
1) Do we revert the software emulation of vlan header tagging to fix it's implementation issues in 2.6.40?

The big issues.
- vlan_untag is currently reading undefined data.
- Clearing REORDER_HDR no longer works.

I expect the issues are small enough that we can address them now.

2) Do we instead move the software implementation of vlan header acceleration back into the net-next.

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

Eric

#46 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 64 lines · On Mon, 23 May 2011 15:05:54 -0700 [email protected] (Eric W. Biederman)…
Quoted 14 lines · David Miller

From: Stephen Hemminger <[email protected]>
Date: Mon, 23 May 2011 14:00:48 -0700

Quoted 10 lines

IMHO the REORDER_HDR flag was a design mistake. It means supporting two different API's for the application. For a packet capture application it means the format of the packet will be different based on how the VLAN was created. And the vlan was created outside the application.

If there was a requirement to support both formats, then it should be a property of the AF_PACKET socket. The application can then do an setsockopt() to control the format of the data. The issue is for things like mmap/AF_PACKET the re-inserted form will be slower since data has to be copied.

Quoted 6 lines

Indeed, it was a foolish thing to support in the first place.

I guess we couldn't see the hw offloads in the future at that point.

I'm all for finding a way to deprecate this over time.

There are two very similar issues here, that affect two slightly different cases.

Let's assume the configuration is:
eth0 - Talks to the outside world.
vlan2000 - Is the vlan interface of interest.

With REORDER_HDR set when I tcpdump on vlan2000 I don't see the vlan header, however I continue to see the vlan header when I tcpdump on eth0.

With vlan hardware acceleration. When I tcpdump on eth0 I don't see the vlan header. Nor do I see the vlan header when I tcpdump on vlan2000.

Right now both cases are problematic in Linus's tree.

For inbound packets We are testing the wrong interface for the to see if we should play with REORDER_HDR.

Hardware acceleration vlan tagging we are hiding the vlan header from portable applications that simply dump eth0. Which is non-intuitive and apparent to everyone now that this happens on both software and hardware interfaces.

So we have a couple of questions.
1) Do we revert the software emulation of vlan header tagging to fix it's implementation issues in 2.6.40?

The big issues.
- vlan_untag is currently reading undefined data.
- Clearing REORDER_HDR no longer works.

I expect the issues are small enough that we can address them now.

2) Do we instead move the software implementation of vlan header acceleration back into the net-next.

Quoted 6 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

Why not just make libpcap smarter? it already has Linux specific stuff for checksum offload. The portability to BSD for AF_PACKET is a non-issue.

#47 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 71 lines · Stephen Hemminger
Quoted 68 lines · On Mon, 23 May 2011 15:05:54 -0700 [email protected] (Eric W. Biederman)…
Quoted 14 lines · David Miller

From: Stephen Hemminger <[email protected]>
Date: Mon, 23 May 2011 14:00:48 -0700

Quoted 10 lines

IMHO the REORDER_HDR flag was a design mistake. It means supporting two different API's for the application. For a packet capture application it means the format of the packet will be different based on how the VLAN was created. And the vlan was created outside the application.

If there was a requirement to support both formats, then it should be a property of the AF_PACKET socket. The application can then do an setsockopt() to control the format of the data. The issue is for things like mmap/AF_PACKET the re-inserted form will be slower since data has to be copied.

Quoted 6 lines

Indeed, it was a foolish thing to support in the first place.

I guess we couldn't see the hw offloads in the future at that point.

I'm all for finding a way to deprecate this over time.

There are two very similar issues here, that affect two slightly different cases.

Let's assume the configuration is:
eth0 - Talks to the outside world.
vlan2000 - Is the vlan interface of interest.

With REORDER_HDR set when I tcpdump on vlan2000 I don't see the vlan header, however I continue to see the vlan header when I tcpdump on eth0.

With vlan hardware acceleration. When I tcpdump on eth0 I don't see the vlan header. Nor do I see the vlan header when I tcpdump on vlan2000.

Right now both cases are problematic in Linus's tree.

For inbound packets We are testing the wrong interface for the to see if we should play with REORDER_HDR.

Hardware acceleration vlan tagging we are hiding the vlan header from portable applications that simply dump eth0. Which is non-intuitive and apparent to everyone now that this happens on both software and hardware interfaces.

So we have a couple of questions.
1) Do we revert the software emulation of vlan header tagging to fix it's implementation issues in 2.6.40?

The big issues.
- vlan_untag is currently reading undefined data.
- Clearing REORDER_HDR no longer works.

I expect the issues are small enough that we can address them now.

2) Do we instead move the software implementation of vlan header acceleration back into the net-next.

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Quoted 6 lines
Quoted 2 lines

Do we provide an option to keep the vlan header? Should that option be on by default?

Why not just make libpcap smarter? it already has Linux specific stuff for checksum offload. The portability to BSD for AF_PACKET is a non-issue.

That is not an option I had considered, and it sounds intriguing.

Even if libpcap can cope we are still not backwards compatible with older linux applications. Some of which work up to 2.6.39 because the interface the vlan tagged packets come in on does not implement vlan hardware acceleration.

There is also the issue that socket filters can't currently see skb->tci so we are really messed up in the case we want vlan hardware acceleration.

Eric

#48 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Eric W. Biederman
Quoted 27 lines · Eric W. Biederman
Quoted 14 lines · David Miller

From: Stephen Hemminger<[email protected]>
Date: Mon, 23 May 2011 14:00:48 -0700

Quoted 10 lines

IMHO the REORDER_HDR flag was a design mistake. It means supporting two different API's for the application. For a packet capture application it means the format of the packet will be different based on how the VLAN was created. And the vlan was created outside the application.

If there was a requirement to support both formats, then it should be a property of the AF_PACKET socket. The application can then do an setsockopt() to control the format of the data. The issue is for things like mmap/AF_PACKET the re-inserted form will be slower since data has to be copied.

Quoted 6 lines

Indeed, it was a foolish thing to support in the first place.

I guess we couldn't see the hw offloads in the future at that point.

I'm all for finding a way to deprecate this over time.

There are two very similar issues here, that affect two slightly different cases.

Quoted 7 lines

Let's assume the configuration is:
eth0 - Talks to the outside world.
vlan2000 - Is the vlan interface of interest.

With REORDER_HDR set when I tcpdump on vlan2000 I don't see the vlan header, however I continue to see the vlan header when I tcpdump on eth0.

That seems good. This is what originally let dhcp function. Not sure if it's still required for dhcp, but there could easily be other programs that have similar issues.

With REORDER_HDR off, we should see vlan tagged packets on both eth0 and vlan2000.

If REORDER_HDR dissappears entirely, I think you have to default to stripping the header on vlan2000.

Quoted 4 lines

With vlan hardware acceleration. When I tcpdump on eth0 I don't see the vlan header. Nor do I see the vlan header when I tcpdump on vlan2000.

I think you should see the header on eth0 regardless of hw acceleration or not. Users should not have to know if their NIC/driver supports vlan tag stripping in one mode or another.

Quoted 24 lines

Right now both cases are problematic in Linus's tree.

For inbound packets We are testing the wrong interface for the to see if we should play with REORDER_HDR.

Hardware acceleration vlan tagging we are hiding the vlan header from portable applications that simply dump eth0. Which is non-intuitive and apparent to everyone now that this happens on both software and hardware interfaces.

So we have a couple of questions.
1) Do we revert the software emulation of vlan header tagging to fix it's implementation issues in 2.6.40?

The big issues.
- vlan_untag is currently reading undefined data.
- Clearing REORDER_HDR no longer works.

I expect the issues are small enough that we can address them now.

2) Do we instead move the software implementation of vlan header acceleration back into the net-next.

Quoted 6 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

At the least we need to have the header kept when using pf_packet on eth0.

I think it's best to have it available on vlan2000, but perhaps have it stripped by default for backwards compatibility.

Thanks,
Ben

Quoted 2 lines

Eric

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#49 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 4 lines · Ben Greear

On 05/23/2011 03:05 PM, Eric W. Biederman wrote:

If REORDER_HDR dissappears entirely, I think you have to default to stripping the header on vlan2000.

Which is what patches that started this thread are doing.

Quoted 7 lines
Quoted 3 lines

With vlan hardware acceleration. When I tcpdump on eth0 I don't see the vlan header. Nor do I see the vlan header when I tcpdump on vlan2000.

I think you should see the header on eth0 regardless of hw acceleration or not. Users should not have to know if their NIC/driver supports vlan tag stripping in one mode or another.

But is it acceptable to fix this in libpcap?

Quoted 9 lines
Quoted 6 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

At the least we need to have the header kept when using pf_packet on eth0.

Start with the assumption that vlan hardware acceleration is in place and the hardware has stripped the vlan tag and put it in skb->tci. Sure there are dumb divers out there that don't do this today but either we need to throw out vlan hardware acceleration completely or emulate it in software because otherwise the test matrix is just too big.

Quoted 2 lines

I think it's best to have it available on vlan2000, but perhaps have it stripped by default for backwards compatibility.

Anything that deals with raw packets pretty much breaks if you don't strip the vlan header from visibility on vlan2000. Plus you loose any advantage there is from vlan hardware acceleration, which is available on must modern NICs. So I don't think we can seriously consider having the vlan header for present on the vlan2000 device.

All that is interesting is what to do with eth0, and pf_packet sockets. And the only question that seems really interesting there is do we put the vlan header back on with libpcap magic or with pf_packet logic.

We have to start with a the assumption that we come in with a pf_packet with the vlan tag only in skb->tci.

Eric

#50 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 13 lines · Eric W. Biederman
Quoted 4 lines · Ben Greear

On 05/23/2011 03:05 PM, Eric W. Biederman wrote:

If REORDER_HDR dissappears entirely, I think you have to default to stripping the header on vlan2000.

Which is what patches that started this thread are doing.

Quoted 4 lines
Quoted 3 lines

With vlan hardware acceleration. When I tcpdump on eth0 I don't see the vlan header. Nor do I see the vlan header when I tcpdump on vlan2000.

Quoted 5 lines
Quoted 3 lines

I think you should see the header on eth0 regardless of hw acceleration or not. Users should not have to know if their NIC/driver supports vlan tag stripping in one mode or another.

But is it acceptable to fix this in libpcap?

No, because libpcap is not the only thing that opens packet sockets. Our user-space network emulator bridges two ethernet interfaces using packet-sockets. We have to have the VLAN header available to properly bridge a VLAN packet out the other side. It would be possible to use some aux data, but I think that is a very nasty hack.

Surely we and libpcap are not the only users...

Also, anything using a packet filter might need the header in place.

Quoted 10 lines
Quoted 9 lines
Quoted 6 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

At the least we need to have the header kept when using pf_packet on eth0.

Quoted 6 lines

Start with the assumption that vlan hardware acceleration is in place and the hardware has stripped the vlan tag and put it in skb->tci. Sure there are dumb divers out there that don't do this today but either we need to throw out vlan hardware acceleration completely or emulate it in software because otherwise the test matrix is just too big.

It's a binary case..either tag exists in skb or it doesn't. It might double the test cases, but that is still a reasonable number. We should be able to write a test app to mostly automate testing for that matter.

I still haven't had time to do detailed testing, but I *believe* that even smart drivers like e1000e and igb don't strip tags if you do not have any VLANs created on the NIC. So, sometimes you get tags on eth0 even when using fancy NICs.

Quoted 9 lines
Quoted 3 lines

I think it's best to have it available on vlan2000, but perhaps have it stripped by default for backwards compatibility.

Anything that deals with raw packets pretty much breaks if you don't strip the vlan header from visibility on vlan2000. Plus you loose any advantage there is from vlan hardware acceleration, which is available on must modern NICs. So I don't think we can seriously consider having the vlan header for present on the vlan2000 device.

I'm fine with stripping them from the vlan2000 frame, though hopefully one day we can optimize to only do this if there are actual consumers of the raw packet in user-space.

Quoted 6 lines

All that is interesting is what to do with eth0, and pf_packet sockets. And the only question that seems really interesting there is do we put the vlan header back on with libpcap magic or with pf_packet logic.

We have to start with a the assumption that we come in with a pf_packet with the vlan tag only in skb->tci.

Then I think we should put it back with pf_packet logic. Possibly with a per-socket option to disable this and send it as only aux data if that is more efficient.

If it turns out the NIC is not stripping VLAN tags for whatever reason, we might be able to optimize things so that it never does the HW emulation so that it never has to un-do it later.

If we can agree on the desired behaviour, I'll put some effort into a test system that can test a 4-port system:

eth0 { loopback cable } eth1 { pf-socket based software bridge } eth2 { looped-back-cable } eth3

Could iterate through a matrix of vlans in various places and test each time to make sure eth0 and eth3 receive expected data. I could split the sender/consumer logic into one process and the bridge into another so that it could be done on two systems with 2 ports each.

Thanks,
Ben

Quoted 2 lines

Eric

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#51 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Le 24/05/2011 06:20, Ben Greear a écrit :
<snip>

Quoted 3 lines

Then I think we should put it back with pf_packet logic. Possibly with a per-socket option to disable this and send it as only aux data if that is more efficient.

Why would we need a per-socket option for that?

If the socket listen on eth0, it probably expect to receive the raw packet. If it happens to expect the untagged packet, it should listen on vlan2000.

The only reason I can imagine to listen on eth0 while expecting the packet to be untagged is to listen to several VLAN at the same time on a single socket, while still expecting the kernel or the NIC to extract the vlan ID for us. But I don't have a real life use case for this in mind.

And maybe, for that particular situation, we should have a special vlan interface, with wildcard vlan ID:

       +--eth0.100
       |
eth0 --+--eth0.200
       |
       +--eth0.any

- Someone listening on eth0 would receive raw packet.
- Someone listening on eth0.100/eth0.200 would receive untagged packets originally having 100/200 as the VLAN ID.
- Someone listening on eth0.any would receive untagged packets originally having any VLAN ID and would not receive non-originally-tagged packets.

Quoted 3 lines

If it turns out the NIC is not stripping VLAN tags for whatever reason, we might be able to optimize things so that it never does the HW emulation so that it never has to un-do it later.

It might be very tricky to avoid any do-undo-redo situation. I will latter try and describe all the possible situations: with/without hw-accel, with/without a protocol handler registered on the parent interface, with/without a child interface build on top of a particular parent interface and with the corresponding VLAN ID...

Nicolas.

#52 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

2011/5/24 Nicolas de Pesloüan <[email protected]>:

Quoted 8 lines
Quoted 3 lines

If it turns out the NIC is not stripping VLAN tags for whatever reason, we might be able to optimize things so that it never does the HW emulation so that it never has to un-do it later.

It might be very tricky to avoid any do-undo-redo situation. I will latter try and describe all the possible situations: with/without hw-accel, with/without a protocol handler registered on the parent interface, with/without a child interface build on top of a particular parent interface and with the corresponding VLAN ID...

Hardware tag stripping could be disabled whenever AF_PACKET listener is present on the base interface. It should be easy now with the new features handling.

BTW, what's the performance gain from hw tag stripping? Besides saving 12-byte memmove() in one cacheline?

Best Regards,
Michał Mirosław

#53 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Nicolas de Pesloüan
Quoted 8 lines · Nicolas de Pesloüan

Le 24/05/2011 06:20, Ben Greear a écrit :
<snip>

Quoted 3 lines

Then I think we should put it back with pf_packet logic. Possibly with a per-socket option to disable this and send it as only aux data if that is more efficient.

Why would we need a per-socket option for that?

I'd want the tag available at all times, but if it is a performance win to put the tag in 'aux data', then a socket could opt for that method. Otherwise, the tag should always be inline in the packet.

Quoted 7 lines
Quoted 3 lines

If it turns out the NIC is not stripping VLAN tags for whatever reason, we might be able to optimize things so that it never does the HW emulation so that it never has to un-do it later.

It might be very tricky to avoid any do-undo-redo situation. I will latter try and describe all the possible situations: with/without hw-accel, with/without a protocol handler registered on the parent interface, with/without a child interface build on top of a particular parent interface and with the corresponding VLAN ID...

You could optimize for some common cases, but leave fix-up code in place to add/remove the tag properly before handing the packet to user-space, filters, etc.

Thanks,
Ben

Quoted 2 lines

Nicolas.

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
maintainer
#54 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Eric W. Biederman

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 15:05:54 -0700

Quoted 7 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

The vlan_tci in the V2 pf_packet auxdata was intended for this purpose.

So no matter what variant of behavior is occurring, apps can properly reconstitute the VLAN header if they inspect the vlan_tci in the auxdata.

The only thing that seems to be missing is an indication that a VLAN tag was present at all, ie. vlan_tx_tag_present(), in this manner an application could then differentiate between no VLAN header and a VLAN tag of zero.

#55 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

Quoted 11 lines · David Miller

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 15:05:54 -0700

Quoted 7 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

Quoted 6 lines

The vlan_tci in the V2 pf_packet auxdata was intended for this purpose.

So no matter what variant of behavior is occurring, apps can properly reconstitute the VLAN header if they inspect the vlan_tci in the auxdata.

It sucks a little bit to deal with that but that is fair.

Quoted 4 lines

The only thing that seems to be missing is an indication that a VLAN tag was present at all, ie. vlan_tx_tag_present(), in this manner an application could then differentiate between no VLAN header and a VLAN tag of zero.

Good point.

I had seen that we were putting the data there but I missed the fact that we deleted the indicator. That makes packets not destined for a vlan but that just have priority bits set hard to detect. Especially if the priority bits are 0.

Would it cause many problems if we added used tp_status to hold a flag indicating the presence of a vlan header?

We also have an issue that the socket filter doesn't have access to any of the vlan information at present.

Eric

#56 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to David Miller
Quoted 11 lines · David Miller

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 15:05:54 -0700

Quoted 7 lines

3) What do we do with pf_packet and vlan hardware acceleration when dumping not the vlan interface but the interface below the vlan interface?

Do we provide an option to keep the vlan header? Should that option be on by default?

Quoted 6 lines

The vlan_tci in the V2 pf_packet auxdata was intended for this purpose.

So no matter what variant of behavior is occurring, apps can properly reconstitute the VLAN header if they inspect the vlan_tci in the auxdata.

When using pf-packet on eth0, with no VLAN devices existing, would you still be putting the VLAN tags in auxdata, or would the tags be inline in the skb?

Quoted 4 lines

The only thing that seems to be missing is an indication that a VLAN tag was present at all, ie. vlan_tx_tag_present(), in this manner an application could then differentiate between no VLAN header and a VLAN tag of zero.

For nested VLANs, the outside VLAN data is in the auxdata, and the rest is inline in the packet?

Thanks,
Ben

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#57 lore raw

PATCH 1/1 vlan: Fix the b0rked ingress VLAN_FLAG_REORDER_HDR check.

In reply to David Miller

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive. In attempting to test VLAN_FLAG_REORDER_HDR we are attempting to deference vlan private network device members a non vlan device, which is just wrong. Move the test into vlan_do_receive so that the vlan header will not be properly put on the packet in the case of vlan header acceleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code. Until then this keeps it working correctly.

---
 include/linux/if_vlan.h |   25 ++++++++++++++++++++++---
 net/8021q/vlan_core.c   |   25 ++++++++++++++-----------
 2 files changed, 36 insertions(+), 14 deletions(-)
include/linux/if_vlan.h +22 -3
index 290bd8a..26f3c92 100644
@@ -220,7 +220,7 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
 }
 
 /**
- * __vlan_put_tag - regular VLAN tag inserting
+ * vlan_insert_tag - regular VLAN tag inserting
  * @skb: skbuff to tag
  * @vlan_tci: VLAN TCI to insert
  *
@@ -229,8 +229,10 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
  *
  * Following the skb_unshare() example, in case of error, the calling function
  * doesn't have to worry about freeing the original skb.
+ *
+ * Does not change skb->protocol so this function can be used for receive.
  */
-static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+static inline struct sk_buff *vlan_insert_tag(struct sk_buff *skb, u16 vlan_tci)
 {
 	struct vlan_ethhdr *veth;
 
@@ -250,8 +252,25 @@ static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
 	/* now, the TCI */
 	veth->h_vlan_TCI = htons(vlan_tci);
 
-	skb->protocol = htons(ETH_P_8021Q);
+	return skb;
+}
 
+/**
+ * __vlan_put_tag - regular VLAN tag inserting
+ * @skb: skbuff to tag
+ * @vlan_tci: VLAN TCI to insert
+ *
+ * Inserts the VLAN tag into @skb as part of the payload
+ * Returns a VLAN tagged skb. If a new skb is created, @skb is freed.
+ *
+ * Following the skb_unshare() example, in case of error, the calling function
+ * doesn't have to worry about freeing the original skb.
+ */
+static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+{
+	skb = vlan_insert_tag(skb, vlan_tci);
+	if (skb)
+		skb->protocol = htons(ETH_P_8021Q);
 	return skb;
 }
 
net/8021q/vlan_core.c +14 -11
index 41495dc..c08dae7 100644
@@ -23,6 +23,11 @@ bool vlan_do_receive(struct sk_buff **skbp)
 		return false;
 
 	skb->dev = vlan_dev;
+	if (unlikely(!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR))) {
+		skb = *skbp = vlan_insert_tag(skb, skb->vlan_tci);
+		if (!skb)
+			return false;
+	}
 	skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
 	skb->vlan_tci = 0;
 
@@ -89,17 +94,15 @@ gro_result_t vlan_gro_frags(struct napi_struct *napi, struct vlan_group *grp,
 }
 EXPORT_SYMBOL(vlan_gro_frags);
 
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)
 {
-	if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-		if (skb_cow(skb, skb_headroom(skb)) < 0)
-			skb = NULL;
-		if (skb) {
-			/* Lifted from Gleb's VLAN code... */
-			memmove(skb->data - ETH_HLEN,
-				skb->data - VLAN_ETH_HLEN, 12);
-			skb->mac_header += VLAN_HLEN;
-		}
+	if (skb_cow(skb, skb_headroom(skb)) < 0)
+		skb = NULL;
+	if (skb) {
+		/* Lifted from Gleb's VLAN code... */
+		memmove(skb->data - ETH_HLEN,
+			skb->data - VLAN_ETH_HLEN, 12);
+		skb->mac_header += VLAN_HLEN;
 	}
 	return skb;
 }
@@ -161,7 +164,7 @@ struct sk_buff *vlan_untag(struct sk_buff *skb)
 	skb_pull_rcsum(skb, VLAN_HLEN);
 	vlan_set_encap_proto(skb, vhdr);
 
-	skb = vlan_check_reorder_header(skb);
+	skb = vlan_reorder_header(skb);
 	if (unlikely(!skb))
 		goto err_free;
 
Signature · 1 line
1.7.5.1.217.g4e3aa
maintainer
#58 lore raw

Re: [PATCH] vlan: Fix the b0rked ingress VLAN_FLAG_REORDER_HDR check.

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 17:11:51 -0700

Quoted 6 lines
net/8021q/vlan_core.c +0 -0
index 41495dc..c08dae7 100644
@@ -23,6 +23,11 @@ bool vlan_do_receive(struct sk_buff **skbp)
 		return false;

net/8021q/vlan_core.c bool vlan_do_receive(struct sk_buff **skbp) · line 33

Quoted 10 lines
net/8021q/vlan_core.c +5 -0
 
 	skb->dev = vlan_dev;
+	if (unlikely(!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR))) {
+		skb = *skbp = vlan_insert_tag(skb, skb->vlan_tci);
+		if (!skb)
+			return false;
+	}
 	skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
 	skb->vlan_tci = 0;
 

Below we do a eth_hdr(skb) based check on the ethernet address in the PACKET_OTHERHOST case.

Won't this VLAN tag insertion change skb->mac_header and thus screw up that test?

Touching this code requires surgical precision and long audits, because there are so many subtble dependencies all over the place like this.

#59 lore raw

Re: [PATCH] vlan: Fix the b0rked ingress VLAN_FLAG_REORDER_HDR check.

Quoted 23 lines · David Miller

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 17:11:51 -0700

Quoted 6 lines
net/8021q/vlan_core.c +0 -0
index 41495dc..c08dae7 100644
@@ -23,6 +23,11 @@ bool vlan_do_receive(struct sk_buff **skbp)
 		return false;

net/8021q/vlan_core.c bool vlan_do_receive(struct sk_buff **skbp) · line 33

Quoted 10 lines
net/8021q/vlan_core.c +5 -0
 
 	skb->dev = vlan_dev;
+	if (unlikely(!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR))) {
+		skb = *skbp = vlan_insert_tag(skb, skb->vlan_tci);
+		if (!skb)
+			return false;
+	}
 	skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
 	skb->vlan_tci = 0;
 

Below we do a eth_hdr(skb) based check on the ethernet address in the PACKET_OTHERHOST case.

Quoted 2 lines

Won't this VLAN tag insertion change skb->mac_header and thus screw up that test?

Yes the vlan tag insertion does in fact change the skb->mac_header index. However we also move the location of the data as well, so I fail to see any reason to be concerned about the PACKET_OTHERHOST case. Things were moved around and we updated the appropriate references.

Feel free to read through the code, to convince yourself it is correct. In addition the code is untouched from the vlan header insertion for emulation of vlan header acceleration in dev_hard_start_xmit() which presumably has been working for quite awhile.

Quoted 2 lines

Touching this code requires surgical precision and long audits, because there are so many subtble dependencies all over the place like this.

Certainly.

I think you will find that I have applied great precision and restraint in this case.

Eric

maintainer
#60 lore raw

Re: [PATCH] vlan: Fix the b0rked ingress VLAN_FLAG_REORDER_HDR check.

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 23:18:02 -0700

Quoted 4 lines

Feel free to read through the code, to convince yourself it is correct. In addition the code is untouched from the vlan header insertion for emulation of vlan header acceleration in dev_hard_start_xmit() which presumably has been working for quite awhile.

I'm not keeping code there that does eth_hdr(skb)->foo when there can be either a vlan_hdr(skb) or a eth_hdr(skb) there.

That's just asking for trouble.

#61 lore raw

Re: [PATCH] vlan: Fix the b0rked ingress VLAN_FLAG_REORDER_HDR check.

Quoted 8 lines · David Miller

From: [email protected] (Eric W. Biederman)
Date: Mon, 23 May 2011 23:18:02 -0700

Quoted 4 lines

Feel free to read through the code, to convince yourself it is correct. In addition the code is untouched from the vlan header insertion for emulation of vlan header acceleration in dev_hard_start_xmit() which presumably has been working for quite awhile.

Quoted 4 lines

I'm not keeping code there that does eth_hdr(skb)->foo when there can be either a vlan_hdr(skb) or a eth_hdr(skb) there.

That's just asking for trouble.

How so? eth_hdr(skb) is a proper subset of vlan_hdr(skb)?

vlan_insert_tag() moves the ethernet header to make space for the vlan header after it, but before the rest of the data. With the appropriate fixups applied to the skb->mac_pointer.

I can see not leaving a vlan_hdr(skb) reference, but beyond that I'm not seeing the problem.

Certainly we need to do the insert before we update the statics or we will get rx_bytes wrong.

Would you be happier if the pkt_type check was moved earlier in the function like say:

net/8021q/vlan_core.c +15 -15
index c08dae7..7571637 100644
@@ -22,21 +22,6 @@ bool vlan_do_receive(struct sk_buff **skbp)
 	if (unlikely(!skb))
 		return false;
 
-	skb->dev = vlan_dev;
-	if (unlikely(!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR))) {
-		skb = *skbp = vlan_insert_tag(skb, skb->vlan_tci);
-		if (!skb)
-			return false;
-	}
-	skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
-	skb->vlan_tci = 0;
-
-	rx_stats = this_cpu_ptr(vlan_dev_info(vlan_dev)->vlan_pcpu_stats);
-
-	u64_stats_update_begin(&rx_stats->syncp);
-	rx_stats->rx_packets++;
-	rx_stats->rx_bytes += skb->len;
-
 	switch (skb->pkt_type) {
 	case PACKET_BROADCAST:
 		break;
@@ -52,6 +37,21 @@ bool vlan_do_receive(struct sk_buff **skbp)
 			skb->pkt_type = PACKET_HOST;
 		break;
 	}
+
+	skb->dev = vlan_dev;
+	if (unlikely(!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR))) {
+		skb = *skbp = vlan_insert_tag(skb, skb->vlan_tci);
+		if (!skb)
+			return false;
+	}
+	skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
+	skb->vlan_tci = 0;
+
+	rx_stats = this_cpu_ptr(vlan_dev_info(vlan_dev)->vlan_pcpu_stats);
+
+	u64_stats_update_begin(&rx_stats->syncp);
+	rx_stats->rx_packets++;
+	rx_stats->rx_bytes += skb->len;
 	u64_stats_update_end(&rx_stats->syncp);
 
 	return true;

Eric

maintainer
#62 lore raw

Re: [PATCH] vlan: Fix the b0rked ingress VLAN_FLAG_REORDER_HDR check.

From: [email protected] (Eric W. Biederman)
Date: Tue, 24 May 2011 00:38:34 -0700

Quoted 2 lines

Would you be happier if the pkt_type check was moved earlier in the function like say:

Yes, I like this best because it's fully consistent and it fixes the bug.

#63 lore raw

PATCH 1/1 vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive. Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code. Until then this keeps it working correctly.

---
 include/linux/if_vlan.h |   25 ++++++++++++++++++++--
 net/8021q/vlan_core.c   |   50 ++++++++++++++++++++++------------------------
 2 files changed, 46 insertions(+), 29 deletions(-)
include/linux/if_vlan.h +22 -3
index 290bd8a..821f0e3 100644
@@ -220,7 +220,7 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
 }
 
 /**
- * __vlan_put_tag - regular VLAN tag inserting
+ * vlan_insert_tag - regular VLAN tag inserting
  * @skb: skbuff to tag
  * @vlan_tci: VLAN TCI to insert
  *
@@ -229,8 +229,10 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
  *
  * Following the skb_unshare() example, in case of error, the calling function
  * doesn't have to worry about freeing the original skb.
+ *
+ * Does not change skb->protocol so this function can be used during receive.
  */
-static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+static inline struct sk_buff *vlan_insert_tag(struct sk_buff *skb, u16 vlan_tci)
 {
 	struct vlan_ethhdr *veth;
 
@@ -250,8 +252,25 @@ static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
 	/* now, the TCI */
 	veth->h_vlan_TCI = htons(vlan_tci);
 
-	skb->protocol = htons(ETH_P_8021Q);
+	return skb;
+}
 
+/**
+ * __vlan_put_tag - regular VLAN tag inserting
+ * @skb: skbuff to tag
+ * @vlan_tci: VLAN TCI to insert
+ *
+ * Inserts the VLAN tag into @skb as part of the payload
+ * Returns a VLAN tagged skb. If a new skb is created, @skb is freed.
+ *
+ * Following the skb_unshare() example, in case of error, the calling function
+ * doesn't have to worry about freeing the original skb.
+ */
+static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+{
+	skb = vlan_insert_tag(skb, vlan_tci);
+	if (skb)
+		skb->protocol = htons(ETH_P_8021Q);
 	return skb;
 }
 
net/8021q/vlan_core.c +24 -26

76 lines not on this page · raw message

Signature · 1 line
1.7.5.1.217.g4e3aa
author Reviewed-by on v1 1/1
#64 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Quoted 17 lines · Thu, Jun 02, 2011 at 03:03:53PM CEST, [email protected] wrote:

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive. Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Quoted 5 lines

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code. Until then this keeps it working correctly.

Looking good to me. Thanks Eric.

#65 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

In reply to Eric W. Biederman
Quoted 8 lines · Eric W. Biederman
net/8021q/vlan_core.c +1 -1
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)

 {
-       if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-               if (skb_cow(skb, skb_headroom(skb)) < 0)
-                       skb = NULL;
-               if (skb) {

Quoted 8 lines

-                       /* Lifted from Gleb's VLAN code... */
-                       memmove(skb->data - ETH_HLEN,
-                               skb->data - VLAN_ETH_HLEN, 12);
-                       skb->mac_header += VLAN_HLEN;
-               }
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               skb = NULL;
+       if (skb) {

I think an else branch maybe more readable here.

Quoted 4 lines

+               /* Lifted from Gleb's VLAN code... */
+               memmove(skb->data - ETH_HLEN,
+                       skb->data - VLAN_ETH_HLEN, 12);
+               skb->mac_header += VLAN_HLEN;

skb->mac_len should be adjusted too.

Quoted 3 lines

       }
       return skb;
 }

Signature · 2 lines
Regards,
Changli Gao([email protected])
#66 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Quoted 11 lines · Changli Gao
Quoted 10 lines · Eric W. Biederman
(quoted diff) +1 -1
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)

 {
-       if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-               if (skb_cow(skb, skb_headroom(skb)) < 0)
-                       skb = NULL;
-               if (skb) {
-                       /* Lifted from Gleb's VLAN code... */
-                       memmove(skb->data - ETH_HLEN,

Quoted 8 lines
Quoted 6 lines

-                               skb->data - VLAN_ETH_HLEN, 12);
-                       skb->mac_header += VLAN_HLEN;
-               }
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               skb = NULL;
+       if (skb) {

I think an else branch maybe more readable here.

Probably most readable would be simply returning NULL immediately on error. In this patch I just removed the if statement, and I would like to avoid mixing different bug fixes in the same patch if possible.

Quoted 6 lines
Quoted 4 lines

+               /* Lifted from Gleb's VLAN code... */
+               memmove(skb->data - ETH_HLEN,
+                       skb->data - VLAN_ETH_HLEN, 12);
+               skb->mac_header += VLAN_HLEN;

skb->mac_len should be adjusted too.

Given how vlan_untag is called at the moment it does appear that the skb->mac_len = skb->network_header - skb->mac_header in __netif_receive_skb that used to handle this for is no longer doing this for us.

My feel is that either we need to do all of the header resets and the vlan_untagging together. So we either need this all together before or after the another_round label:

So the proper fix is probably something like this.

net/core/dev.c +3 -3
index bcb05cb..8fe50d4 100644
@@ -3102,9 +3102,6 @@ static int __netif_receive_skb(struct sk_buff *skb)
 		skb->skb_iif = skb->dev->ifindex;
 	orig_dev = skb->dev;
 
-	skb_reset_network_header(skb);
-	skb_reset_transport_header(skb);
-	skb->mac_len = skb->network_header - skb->mac_header;
 
 	pt_prev = NULL;
 
@@ -3119,6 +3116,9 @@ another_round:
 		if (unlikely(!skb))
 			goto out;
 	}
+	skb_reset_network_header(skb);
+	skb_reset_transport_header(skb);
+	skb->mac_len = skb->network_header - skb->mac_header;
 
 #ifdef CONFIG_NET_CLS_ACT
 	if (skb->tc_verd & TC_NCLS) {
#67 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Quoted 22 lines · Eric W. Biederman
Quoted 11 lines · Changli Gao
Quoted 10 lines · Eric W. Biederman
(quoted diff) +1 -1
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)

 {
-       if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-               if (skb_cow(skb, skb_headroom(skb)) < 0)
-                       skb = NULL;
-               if (skb) {
-                       /* Lifted from Gleb's VLAN code... */
-                       memmove(skb->data - ETH_HLEN,

Quoted 8 lines
Quoted 6 lines

-                               skb->data - VLAN_ETH_HLEN, 12);
-                       skb->mac_header += VLAN_HLEN;
-               }
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               skb = NULL;
+       if (skb) {

I think an else branch maybe more readable here.

Quoted 4 lines

Probably most readable would be simply returning NULL immediately on error.  In this patch I just removed the if statement, and I would like to avoid mixing different bug fixes in the same patch if possible.

OK. It is minor.

Quoted 12 lines
Quoted 6 lines
Quoted 4 lines

+               /* Lifted from Gleb's VLAN code... */
+               memmove(skb->data - ETH_HLEN,
+                       skb->data - VLAN_ETH_HLEN, 12);
+               skb->mac_header += VLAN_HLEN;

skb->mac_len should be adjusted too.

Given how vlan_untag is called at the moment it does appear that the skb->mac_len = skb->network_header - skb->mac_header in __netif_receive_skb that used to handle this for is no longer doing this for us.

Quoted 6 lines

My feel is that either we need to do all of the header resets and the vlan_untagging together.  So we either need this all together before or after the another_round label:

So the proper fix is probably something like this.

OK, it is right. Thanks.

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#68 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

In reply to Eric W. Biederman
Quoted 59 lines · Thu, Jun 02, 2011 at 05:26:51PM CEST, [email protected] wrote:
Quoted 11 lines · Changli Gao
Quoted 10 lines · Eric W. Biederman
(quoted diff) +1 -1
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)

 {
-       if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-               if (skb_cow(skb, skb_headroom(skb)) < 0)
-                       skb = NULL;
-               if (skb) {
-                       /* Lifted from Gleb's VLAN code... */
-                       memmove(skb->data - ETH_HLEN,

Quoted 8 lines
Quoted 6 lines

-                               skb->data - VLAN_ETH_HLEN, 12);
-                       skb->mac_header += VLAN_HLEN;
-               }
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               skb = NULL;
+       if (skb) {

I think an else branch maybe more readable here.

Probably most readable would be simply returning NULL immediately on error. In this patch I just removed the if statement, and I would like to avoid mixing different bug fixes in the same patch if possible.

Quoted 6 lines
Quoted 4 lines

+               /* Lifted from Gleb's VLAN code... */
+               memmove(skb->data - ETH_HLEN,
+                       skb->data - VLAN_ETH_HLEN, 12);
+               skb->mac_header += VLAN_HLEN;

skb->mac_len should be adjusted too.

Given how vlan_untag is called at the moment it does appear that the skb->mac_len = skb->network_header - skb->mac_header in __netif_receive_skb that used to handle this for is no longer doing this for us.

My feel is that either we need to do all of the header resets and the vlan_untagging together. So we either need this all together before or after the another_round label:

So the proper fix is probably something like this.

net/core/dev.c +0 -3
index bcb05cb..8fe50d4 100644
@@ -3102,9 +3102,6 @@ static int __netif_receive_skb(struct sk_buff *skb)
 		skb->skb_iif = skb->dev->ifindex;
 	orig_dev = skb->dev;
 
-	skb_reset_network_header(skb);
-	skb_reset_transport_header(skb);
-	skb->mac_len = skb->network_header - skb->mac_header;
 
 	pt_prev = NULL;
 

net/core/dev.c another_round: · line 3124

Quoted 9 lines
net/core/dev.c +3 -0
@@ -3119,6 +3116,9 @@ another_round:
 		if (unlikely(!skb))
 			goto out;
 	}
+	skb_reset_network_header(skb);
+	skb_reset_transport_header(skb);
+	skb->mac_len = skb->network_header - skb->mac_header;
 
 #ifdef CONFIG_NET_CLS_ACT
 	if (skb->tc_verd & TC_NCLS) {

This looks good to me. This does not need to be done before vlan_untag.

#69 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

In reply to Eric W. Biederman

Doesnt __vlan_put_tag()/vlan_insert_tag() depend on skb->data pointing to ethernet header.
Is'nt the skb->data pointing past ethernet header in netif_receive_skb(). Am I missing anything?
Regards,
Padmanabh

Quoted 171 lines · Eric W. Biederman

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive.  Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code.  Until then this keeps it working correctly.

---
 include/linux/if_vlan.h |   25 ++++++++++++++++++++-- net/8021q/vlan_core.c   |   50 ++++++++++++++++++++++------------------------

 2 files changed, 46 insertions(+), 29 deletions(-)
include/linux/if_vlan.h +22 -3
index 290bd8a..821f0e3 100644
@@ -220,7 +220,7 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
  }
 
  /**
- * __vlan_put_tag - regular VLAN tag inserting
+ * vlan_insert_tag - regular VLAN tag inserting
  * @skb: skbuff to tag
  * @vlan_tci: VLAN TCI to insert
  *
@@ -229,8 +229,10 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
  *
  * Following the skb_unshare() example, in case of error, the calling function
  * doesn't have to worry about freeing the original skb.
+ *
+ * Does not change skb->protocol so this function can be used during receive.
  */
-static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+static inline struct sk_buff *vlan_insert_tag(struct sk_buff *skb, u16 vlan_tci)
  {
        struct vlan_ethhdr *veth;
 
@@ -250,8 +252,25 @@ static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
        /* now, the TCI */
        veth->h_vlan_TCI = htons(vlan_tci);
 
-       skb->protocol = htons(ETH_P_8021Q);
+       return skb;
+}
 
+/**
+ * __vlan_put_tag - regular VLAN tag inserting
+ * @skb: skbuff to tag
+ * @vlan_tci: VLAN TCI to insert
+ *
+ * Inserts the VLAN tag into @skb as part of the payload
+ * Returns a VLAN tagged skb. If a new skb is created, @skb is freed.
+ *
+ * Following the skb_unshare() example, in case of error, the calling function
+ * doesn't have to worry about freeing the original skb.
+ */
+static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+{
+       skb = vlan_insert_tag(skb, vlan_tci);
+       if (skb)
+               skb->protocol = htons(ETH_P_8021Q);
        return skb;
  }
 
net/8021q/vlan_core.c +24 -27
index 41495dc..735c818 100644
@@ -23,6 +23,20 @@ bool vlan_do_receive(struct sk_buff **skbp)
                return false;
 
        skb->dev = vlan_dev;
+       if (skb->pkt_type == PACKET_OTHERHOST) {
+               /* Our lower layer thinks this is not local, let's make sure.
+                * This allows the VLAN to have a different MAC than the
+                * underlying device, and still route correctly. */
+               if (!compare_ether_addr(eth_hdr(skb)->h_dest,
+                                       vlan_dev->dev_addr))
+                       skb->pkt_type = PACKET_HOST;
+       }
+
+       if (unlikely(!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR))) {
+               skb = *skbp = vlan_insert_tag(skb, skb->vlan_tci);
+               if (!skb)
+                       return false;
+       }
        skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
        skb->vlan_tci = 0;
 
@@ -31,22 +45,8 @@ bool vlan_do_receive(struct sk_buff **skbp)
        u64_stats_update_begin(&rx_stats->syncp);
        rx_stats->rx_packets++;
        rx_stats->rx_bytes += skb->len;
-
-       switch (skb->pkt_type) {
-       case PACKET_BROADCAST:
-               break;
-       case PACKET_MULTICAST:
+       if (skb->pkt_type == PACKET_MULTICAST)
                rx_stats->rx_multicast++;
-               break;
-       case PACKET_OTHERHOST:
-               /* Our lower layer thinks this is not local, let's make sure.
-                * This allows the VLAN to have a different MAC than the
-                * underlying device, and still route correctly. */
-               if (!compare_ether_addr(eth_hdr(skb)->h_dest,
-                                       vlan_dev->dev_addr))
-                       skb->pkt_type = PACKET_HOST;
-               break;
-       }
        u64_stats_update_end(&rx_stats->syncp);
 
        return true;
@@ -89,17 +89,15 @@ gro_result_t vlan_gro_frags(struct napi_struct *napi, struct vlan_group *grp,
  }
  EXPORT_SYMBOL(vlan_gro_frags);
 
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)
  {
-       if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-               if (skb_cow(skb, skb_headroom(skb)) < 0)
-                       skb = NULL;
-               if (skb) {
-                       /* Lifted from Gleb's VLAN code... */
-                       memmove(skb->data - ETH_HLEN,
-                               skb->data - VLAN_ETH_HLEN, 12);
-                       skb->mac_header += VLAN_HLEN;
-               }
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               skb = NULL;
+       if (skb) {
+               /* Lifted from Gleb's VLAN code... */
+               memmove(skb->data - ETH_HLEN,
+                       skb->data - VLAN_ETH_HLEN, 12);
+               skb->mac_header += VLAN_HLEN;
        }
        return skb;
  }
@@ -161,7 +159,7 @@ struct sk_buff *vlan_untag(struct sk_buff *skb)
        skb_pull_rcsum(skb, VLAN_HLEN);
        vlan_set_encap_proto(skb, vhdr);
 
-       skb = vlan_check_reorder_header(skb);
+       skb = vlan_reorder_header(skb);
        if (unlikely(!skb))
                goto err_free;
 
--

1.7.5.1.217.g4e3aa

--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

#70 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Quoted 4 lines · padmanabh ratnakar

Doesnt __vlan_put_tag()/vlan_insert_tag() depend on skb->data pointing to ethernet header.
Is'nt the skb->data pointing past ethernet header in netif_receive_skb(). Am I missing anything?

Yes, you are right. skb->data should be adjusted before feeding to vlan_insert_tag(), or we make vlan_insert_tag() rely on skb->mac_header instead.

Signature · 2 lines
Regards,
Changli Gao([email protected])
maintainer
#71 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

From: Changli Gao <[email protected]>
Date: Fri, 3 Jun 2011 11:59:26 +0800

Quoted 10 lines
Quoted 4 lines · padmanabh ratnakar

Doesnt __vlan_put_tag()/vlan_insert_tag() depend on skb->data pointing to ethernet header.
Is'nt the skb->data pointing past ethernet header in netif_receive_skb(). Am I missing anything?

Yes, you are right. skb->data should be adjusted before feeding to vlan_insert_tag(), or we make vlan_insert_tag() rely on skb->mac_header instead.

I told you guys this is not simple stuff :-)

Even here we have 3 or 4 people who study this code all the time, and are still having trouble sorting out all of these details.

author
#72 lore raw

PATCH v3 1/1 vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Please review. I'm not sure if there wouldn't be necessary to control for rx accelerated skbs. Smoke tested.

Subject: [patch net-2.6 v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive. Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code. Until then this keeps it working correctly.

Use vlan_unreorder_header to revert actions previously done by vlan_untag->vlan_reorder_header in case VLAN_FLAG_REORDER_HDR is not set

---
 net/8021q/vlan_core.c |   75 ++++++++++++++++++++++++++++--------------------
 1 files changed, 44 insertions(+), 31 deletions(-)
net/8021q/vlan_core.c +44 -31
index 41495dc..d71cc81 100644
@@ -4,6 +4,33 @@
 #include <linux/netpoll.h>
 #include "vlan.h"
 
+static struct sk_buff *vlan_reorder_header(struct sk_buff *skb)
+{
+	if (skb_cow(skb, skb_headroom(skb)) < 0)
+		return NULL;
+	/* Lifted from Gleb's VLAN code... */
+	memmove(skb->data - ETH_HLEN, skb->data - VLAN_ETH_HLEN, 12);
+	skb->mac_header += VLAN_HLEN;
+	return skb;
+}
+
+/* Should be used only to revert vlan_reorder_header action */
+static struct sk_buff *vlan_unreorder_header(struct sk_buff *skb)
+{
+	unsigned char *mac_header;
+	struct vlan_ethhdr *veth;
+
+	if (skb_cow(skb, skb_headroom(skb)) < 0)
+		return NULL;
+	skb->mac_header -= VLAN_HLEN;
+	mac_header = skb_mac_header(skb);
+	memmove(mac_header, mac_header + VLAN_HLEN, 12);
+	veth = (struct vlan_ethhdr *) mac_header;
+	veth->h_vlan_proto = htons(ETH_P_8021Q);
+	veth->h_vlan_TCI = htons(skb->vlan_tci);
+	return skb;
+}
+
 bool vlan_do_receive(struct sk_buff **skbp)
 {
 	struct sk_buff *skb = *skbp;
@@ -23,6 +50,21 @@ bool vlan_do_receive(struct sk_buff **skbp)
 		return false;
 
 	skb->dev = vlan_dev;
+	if (skb->pkt_type == PACKET_OTHERHOST) {
+		/* Our lower layer thinks this is not local, let's make sure.
+		 * This allows the VLAN to have a different MAC than the
+		 * underlying device, and still route correctly. */
+		if (!compare_ether_addr(eth_hdr(skb)->h_dest,
+					vlan_dev->dev_addr))
+			skb->pkt_type = PACKET_HOST;
+	}
+
+	if (!(vlan_dev_info(vlan_dev)->flags & VLAN_FLAG_REORDER_HDR)) {
+		skb = *skbp = vlan_unreorder_header(skb);
+		if (!skb)
+			return false;
+	}
+
 	skb->priority = vlan_get_ingress_priority(vlan_dev, skb->vlan_tci);
 	skb->vlan_tci = 0;
 
@@ -31,22 +73,8 @@ bool vlan_do_receive(struct sk_buff **skbp)
 	u64_stats_update_begin(&rx_stats->syncp);
 	rx_stats->rx_packets++;
 	rx_stats->rx_bytes += skb->len;
-
-	switch (skb->pkt_type) {
-	case PACKET_BROADCAST:
-		break;
-	case PACKET_MULTICAST:
+	if (skb->pkt_type == PACKET_MULTICAST)
 		rx_stats->rx_multicast++;
-		break;
-	case PACKET_OTHERHOST:
-		/* Our lower layer thinks this is not local, let's make sure.
-		 * This allows the VLAN to have a different MAC than the
-		 * underlying device, and still route correctly. */
-		if (!compare_ether_addr(eth_hdr(skb)->h_dest,
-					vlan_dev->dev_addr))
-			skb->pkt_type = PACKET_HOST;
-		break;
-	}
 	u64_stats_update_end(&rx_stats->syncp);
 
 	return true;
@@ -89,21 +117,6 @@ gro_result_t vlan_gro_frags(struct napi_struct *napi, struct vlan_group *grp,
 }
 EXPORT_SYMBOL(vlan_gro_frags);
 
-static struct sk_buff *vlan_check_reorder_header(struct sk_buff *skb)
-{
-	if (vlan_dev_info(skb->dev)->flags & VLAN_FLAG_REORDER_HDR) {
-		if (skb_cow(skb, skb_headroom(skb)) < 0)
-			skb = NULL;
-		if (skb) {
-			/* Lifted from Gleb's VLAN code... */
-			memmove(skb->data - ETH_HLEN,
-				skb->data - VLAN_ETH_HLEN, 12);
-			skb->mac_header += VLAN_HLEN;
-		}
-	}
-	return skb;
-}
-
 static void vlan_set_encap_proto(struct sk_buff *skb, struct vlan_hdr *vhdr)
 {
 	__be16 proto;
@@ -161,7 +174,7 @@ struct sk_buff *vlan_untag(struct sk_buff *skb)
 	skb_pull_rcsum(skb, VLAN_HLEN);
 	vlan_set_encap_proto(skb, vhdr);
 
-	skb = vlan_check_reorder_header(skb);
+	skb = vlan_reorder_header(skb);
 	if (unlikely(!skb))
 		goto err_free;
 
Signature · 1 line
1.7.5.2
#73 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

net/8021q/vlan_core.c line 24

Quoted 9 lines · Jiri Pirko
net/8021q/vlan_core.c +9 -0
+
+/* Should be used only to revert vlan_reorder_header action */
+static struct sk_buff *vlan_unreorder_header(struct sk_buff *skb)
+{
+       unsigned char *mac_header;
+       struct vlan_ethhdr *veth;
+
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               return NULL;

I think we need to make sure if there is enough headroom for this header expansion.

Quoted 1 line

+       skb->mac_header -= VLAN_HLEN;

skb->mac_len isn't adjusted. You forgot to move the headers resetting after vlan_untag().

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#74 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Quoted 13 lines · Fri, Jun 10, 2011 at 11:26:06AM CEST, [email protected] wrote:

net/8021q/vlan_core.c line 24

Quoted 9 lines · Jiri Pirko
net/8021q/vlan_core.c +9 -0
+
+/* Should be used only to revert vlan_reorder_header action */
+static struct sk_buff *vlan_unreorder_header(struct sk_buff *skb)
+{
+       unsigned char *mac_header;
+       struct vlan_ethhdr *veth;
+
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               return NULL;

I think we need to make sure if there is enough headroom for this header expansion.

Well the header expansion was previously there so there should be place there in all cases, or am I wrong?

Quoted 5 lines
Quoted 2 lines

+       skb->mac_header -= VLAN_HLEN;

skb->mac_len isn't adjusted. You forgot to move the headers resetting after vlan_untag().

that is correct, I'll move that. I'll also add reset after unreorder.

Thanks Changli.

Jirka

Quoted 6 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])
#75 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Quoted 15 lines · Jiri Pirko
Quoted 13 lines · Fri, Jun 10, 2011 at 11:26:06AM CEST, [email protected] wrote:

net/8021q/vlan_core.c line 24

Quoted 9 lines · Jiri Pirko
net/8021q/vlan_core.c +9 -0
+
+/* Should be used only to revert vlan_reorder_header action */
+static struct sk_buff *vlan_unreorder_header(struct sk_buff *skb)
+{
+       unsigned char *mac_header;
+       struct vlan_ethhdr *veth;
+
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               return NULL;

I think we need to make sure if there is enough headroom for this header expansion.

Quoted 2 lines

Well the header expansion was previously there so there should be place there in all cases, or am I wrong?

For hw-accel-vlan-rx, is the headroom for this header expansion reserved? I don't think so. Thanks.

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#76 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Quoted 16 lines · Fri, Jun 10, 2011 at 11:49:27AM CEST, [email protected] wrote:
Quoted 15 lines · Jiri Pirko
Quoted 13 lines · Fri, Jun 10, 2011 at 11:26:06AM CEST, [email protected] wrote:

net/8021q/vlan_core.c line 24

Quoted 9 lines · Jiri Pirko
net/8021q/vlan_core.c +9 -0
+
+/* Should be used only to revert vlan_reorder_header action */
+static struct sk_buff *vlan_unreorder_header(struct sk_buff *skb)
+{
+       unsigned char *mac_header;
+       struct vlan_ethhdr *veth;
+
+       if (skb_cow(skb, skb_headroom(skb)) < 0)
+               return NULL;

I think we need to make sure if there is enough headroom for this header expansion.

Quoted 5 lines
Quoted 2 lines

Well the header expansion was previously there so there should be place there in all cases, or am I wrong?

For hw-accel-vlan-rx, is the headroom for this header expansion reserved? I don't think so. Thanks.

But this wasn't done for hw-accel-vlan-rx previously right? So why don't check for hw-accel-vlan-rx and don do unreorder in that case?

Quoted 4 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])
#77 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Quoted 8 lines · Jiri Pirko
Quoted 3 lines · Fri, Jun 10, 2011 at 11:49:27AM CEST, [email protected] wrote:

For hw-accel-vlan-rx, is the headroom for this header expansion reserved? I don't think so. Thanks.

But this wasn't done for hw-accel-vlan-rx previously right? So why don't check for hw-accel-vlan-rx and don do unreorder in that case?

Yes, it wasn't done before. We want to make the behaviors of hw-accel-vlan-rx and non-hw-accel-vlan-rx are the same. Is it your original goal of your vlan-untag patch?

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#78 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Quoted 6 lines · Fri, Jun 10, 2011 at 01:20:29PM CEST, [email protected] wrote:
Quoted 5 lines · Jiri Pirko
Quoted 3 lines · Fri, Jun 10, 2011 at 11:49:27AM CEST, [email protected] wrote:

For hw-accel-vlan-rx, is the headroom for this header expansion reserved? I don't think so. Thanks.

Quoted 7 lines
Quoted 3 lines

But this wasn't done for hw-accel-vlan-rx previously right? So why don't check for hw-accel-vlan-rx and don do unreorder in that case?

Yes, it wasn't done before. We want to make the behaviors of hw-accel-vlan-rx and non-hw-accel-vlan-rx are the same. Is it your original goal of your vlan-untag patch?

You are correct. I just thought that since disabling reordering will be most likely removed in near future, non-hw-accel-vlan-rx reorder disable would not be needed.

Quoted 4 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])
author
#79 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

In reply to Changli Gao

This time heavily based on Eric's V2. mac_len is reset at appropriate places. Also skb->data is adjusted to point to beginning of mac header before calling vlan_insert_tag.

Please review (fingers crossed).

Subject: [patch net-2.6 v2.1] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive. Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code. Until then this keeps it working correctly.

---
 include/linux/if_vlan.h |   25 +++++++++++++++++--
 include/linux/skbuff.h  |    5 ++++
 net/8021q/vlan_core.c   |   60 +++++++++++++++++++++++++---------------------
 net/core/dev.c          |    2 +-
 4 files changed, 61 insertions(+), 31 deletions(-)
include/linux/if_vlan.h +22 -3
index dc01681..affa273 100644
@@ -225,7 +225,7 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
 }
 
 /**
- * __vlan_put_tag - regular VLAN tag inserting
+ * vlan_insert_tag - regular VLAN tag inserting
  * @skb: skbuff to tag
  * @vlan_tci: VLAN TCI to insert
  *
@@ -234,8 +234,10 @@ static inline int vlan_hwaccel_receive_skb(struct sk_buff *skb,
  *
  * Following the skb_unshare() example, in case of error, the calling function
  * doesn't have to worry about freeing the original skb.
+ *
+ * Does not change skb->protocol so this function can be used during receive.
  */
-static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+static inline struct sk_buff *vlan_insert_tag(struct sk_buff *skb, u16 vlan_tci)
 {
 	struct vlan_ethhdr *veth;
 
@@ -255,8 +257,25 @@ static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
 	/* now, the TCI */
 	veth->h_vlan_TCI = htons(vlan_tci);
 
-	skb->protocol = htons(ETH_P_8021Q);
+	return skb;
+}
 
+/**
+ * __vlan_put_tag - regular VLAN tag inserting
+ * @skb: skbuff to tag
+ * @vlan_tci: VLAN TCI to insert
+ *
+ * Inserts the VLAN tag into @skb as part of the payload
+ * Returns a VLAN tagged skb. If a new skb is created, @skb is freed.
+ *
+ * Following the skb_unshare() example, in case of error, the calling function
+ * doesn't have to worry about freeing the original skb.
+ */
+static inline struct sk_buff *__vlan_put_tag(struct sk_buff *skb, u16 vlan_tci)
+{
+	skb = vlan_insert_tag(skb, vlan_tci);
+	if (skb)
+		skb->protocol = htons(ETH_P_8021Q);
 	return skb;
 }
 
include/linux/skbuff.h +5 -0
index e8b78ce..c0a4f3a 100644
@@ -1256,6 +1256,11 @@ static inline void skb_reserve(struct sk_buff *skb, int len)
 	skb->tail += len;
 }
 
+static inline void skb_reset_mac_len(struct sk_buff *skb)
+{
+	skb->mac_len = skb->network_header - skb->mac_header;
+}
+
 #ifdef NET_SKBUFF_DATA_USES_OFFSET
 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
 {
net/8021q/vlan_core.c +33 -27

86 lines not on this page · raw message

net/core/dev.c +1 -1
index a54c9f8..9c58c1e 100644
@@ -3114,7 +3114,7 @@ static int __netif_receive_skb(struct sk_buff *skb)
 
 	skb_reset_network_header(skb);
 	skb_reset_transport_header(skb);
-	skb->mac_len = skb->network_header - skb->mac_header;
+	skb_reset_mac_len(skb);
 
 	pt_prev = NULL;
 
Signature · 1 line
1.7.5.2
Acked-by
#80 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Quoted 24 lines · Jiri Pirko

This time heavily based on Eric's V2. mac_len is reset at appropriate places. Also skb->data is adjusted to point to beginning of mac header before calling vlan_insert_tag.

Please review (fingers crossed).

Subject: [patch net-2.6 v2.1] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive.  Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Quoted 6 lines

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code.  Until then this keeps it working correctly.

Signature · 2 lines
Regards,
Changli Gao([email protected])
maintainer Applied
#81 lore raw

Re: [PATCH v3] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

From: Changli Gao <[email protected]>
Date: Sat, 11 Jun 2011 08:05:37 +0800

Quoted 25 lines · Jiri Pirko
Quoted 24 lines

This time heavily based on Eric's V2. mac_len is reset at appropriate places. Also skb->data is adjusted to point to beginning of mac header before calling vlan_insert_tag.

Please review (fingers crossed).

Subject: [patch net-2.6 v2.1] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive.  Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Quoted 7 lines
Quoted 6 lines

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code.  Until then this keeps it working correctly.

Applied, great work everyone.

author
#82 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

In reply to Eric W. Biederman
Quoted 13 lines · Thu, Jun 02, 2011 at 03:03:53PM CEST, [email protected] wrote:

Testing of VLAN_FLAG_REORDER_HDR does not belong in vlan_untag but rather in vlan_do_receive. Otherwise the vlan header will not be properly put on the packet in the case of vlan header accelleration.

As we remove the check from vlan_check_reorder_header rename it vlan_reorder_header to keep the naming clean.

Fix up the skb->pkt_type early so we don't look at the packet after adding the vlan tag, which guarantees we don't goof and look at the wrong field.

Quoted 7 lines

Use a simple if statement instead of a complicated switch statement to decided that we need to increment rx_stats for a multicast packet.

Hopefully at somepoint we will just declare the case where VLAN_FLAG_REORDER_HDR is cleared as unsupported and remove the code. Until then this keeps it working correctly.

Why we can't remove this right away?

#83 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Quoted 3 lines · Jiri Pirko

Why we can't remove this right away?

I think the reason is this patch is a bug fix for the stable kernel, so we should not introduce new features or remove old features.

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#84 lore raw

Re: [PATCH] vlan: Fix the ingress VLAN_FLAG_REORDER_HDR check v2

Quoted 7 lines · Thu, Jun 09, 2011 at 01:08:13AM CEST, [email protected] wrote:
Quoted 3 lines · Jiri Pirko

Why we can't remove this right away?

I think the reason is this patch is a bug fix for the stable kernel, so we should not introduce new features or remove old features.

But for net-next if should be ok, right?

What's the reason for possibility of removing the reorder flag? Why would user want do it?

Thanks

Jirka

Quoted 4 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])
author
#85 lore raw

Re: [PATCH 1/3] vlan: Do not support clearing VLAN_FLAG_REORDER_HDR

In reply to Eric W. Biederman

How about to drop this patch and silently ignore the flag? That would feel the same for user who is setting it.

Quoted 52 lines · Sun, May 22, 2011 at 09:39:18PM CEST, [email protected] wrote:

Simplify the vlan handling code by not supporing clearing of VLAN_FLAG_REORDER_HDR. Which means we always make the vlan handling code strip the vlan header from the packets, and always insert the vlan header when transmitting packets.

Not stripping the vlan header has alwasy been broken in combination with vlan hardware accelleration. Now that we are making everything look like accelerated vlan handling not stripping the vlan header is always broken.

I don't think anyone actually cares so simply stop supporting the broken case.

---

net/8021q/vlan_dev.c     |    3 +--
net/8021q/vlan_netlink.c |    4 +---
2 files changed, 2 insertions(+), 5 deletions(-)
net/8021q/vlan_dev.c +1 -2
index f247f5b..20629fe 100644
@@ -242,8 +242,7 @@ int vlan_dev_change_flags(const struct net_device *dev, u32 flags, u32 mask)
 	struct vlan_dev_info *vlan = vlan_dev_info(dev);
 	u32 old_flags = vlan->flags;
 
-	if (mask & ~(VLAN_FLAG_REORDER_HDR | VLAN_FLAG_GVRP |
-		     VLAN_FLAG_LOOSE_BINDING))
+	if (mask & ~(VLAN_FLAG_GVRP | VLAN_FLAG_LOOSE_BINDING))
 		return -EINVAL;
 
 	vlan->flags = (old_flags & ~mask) | (flags & mask);
net/8021q/vlan_netlink.c +1 -3
index be9a5c1..a638a4c 100644
@@ -59,9 +59,7 @@ static int vlan_validate(struct nlattr *tb[], struct nlattr *data[])
 	}
 	if (data[IFLA_VLAN_FLAGS]) {
 		flags = nla_data(data[IFLA_VLAN_FLAGS]);
-		if ((flags->flags & flags->mask) &
-		    ~(VLAN_FLAG_REORDER_HDR | VLAN_FLAG_GVRP |
-		      VLAN_FLAG_LOOSE_BINDING))
+		if (flags->mask & ~(VLAN_FLAG_GVRP | VLAN_FLAG_LOOSE_BINDING))
 			return -EINVAL;
 	}
 
Signature · 1 line
1.7.5.1.217.g4e3aa
#86 lore raw
In reply to Jiri Pirko
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 3 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

For a transparent bridge with ports: eth0 and eth1, the vlan tags need to be preserved.

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#87 lore raw
Quoted 13 lines · Sun, May 22, 2011 at 10:38:45AM CEST, [email protected] wrote:
Quoted 12 lines · Jiri Pirko

Sun, May 22, 2011 at 04:59:49AM CEST, [email protected] wrote:

<snip>

Quoted 8 lines

And because some setups may still require the skb not to be untagged, may be we need the ability to re-tag the skb in some situations... When a protocol handler or rx_handler is explicitly registered on a net_device which expect to receive tagged skb, we should deliver tagged skb to it... Arguably, this may sound incredible for the general case, but may be required for not-so-special cases like bridge or protocol analyzer.

Quoted 6 lines
Quoted 3 lines

Wait, what setups/code require the skb not to be untagged? If there's such, it should be fixed.

For a transparent bridge with ports: eth0 and eth1, the vlan tags need to be preserved.

Yet they are - in skb->vlan_tci. I see no problem here. It's the same as if the NIC does hw-untagging itself.

Jirka

Quoted 4 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])
#88 lore raw
Quoted 7 lines · Jiri Pirko
Quoted 3 lines

For a transparent bridge with ports: eth0 and eth1, the vlan tags need to be preserved.

Yet they are - in skb->vlan_tci. I see no problem here. It's the same as if the NIC does hw-untagging itself.

Although they are in skb->vlan_tci, but the xmit nic, which doesn't support hw-accel-vlan-tx, doesn't use them to tag the outgoing packets. Correct me if I am wrong.

I know the current software emulation doesn't strip the vlan headers, but I want to know in what way you fix the current misuse of vlan_check_reorder_header(). If you emulation what the HW does exactly, I am afraid more work are needed. IMHO, this way is more dangerous than just a revert for this bugfix cycle.

Signature · 2 lines
Regards,
Changli Gao([email protected])
#89 lore raw
Quoted 5 lines · Changli Gao
Quoted 4 lines · Jiri Pirko
Quoted 3 lines

For a transparent bridge with ports: eth0 and eth1, the vlan tags need to be preserved.

Quoted 7 lines
Quoted 3 lines

Yet they are - in skb->vlan_tci. I see no problem here. It's the same as if the NIC does hw-untagging itself.

Although they are in skb->vlan_tci, but the xmit nic, which doesn't support hw-accel-vlan-tx, doesn't use them to tag the outgoing packets. Correct me if I am wrong.

There is software emulation in dev_hard_start_xmit for the nics that don't support the feature. So we will properly add tags.

Eric

#90 lore raw
Quoted 4 lines · Eric W. Biederman

There is software emulation in dev_hard_start_xmit for the nics that don't support the feature.  So we will properly add tags.

Thanks. I got it. :)

Signature · 2 lines
Regards,
Changli Gao([email protected])
author
#91 lore raw
In reply to Changli Gao

<snip>

Quoted 6 lines

I know the current software emulation doesn't strip the vlan headers, but I want to know in what way you fix the current misuse of vlan_check_reorder_header(). If you emulation what the HW does exactly, I am afraid more work are needed. IMHO, this way is more dangerous than just a revert for this bugfix cycle.

Looking at this. Will follow-up soon.

Quoted 4 lines
Signature · 3 lines
-- 
Regards,
Changli Gao([email protected])