Skip to contentSkip to conversation

skb: expose and constify hash primitives

PATCH Applied Stephen Hemminger 4 participants 17 messages Mar 20, 2009 – Mar 22, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 skb: expose and constify hash primitives

Some minor changes to queue hashing:
1. Use const on accessor functions
2. Export skb_tx_hash for use in drivers (see ixgbe)

---
 include/linux/skbuff.h |    9 ++++++---
 net/core/dev.c         |    3 ++-
 2 files changed, 8 insertions(+), 4 deletions(-)
include/linux/skbuff.h +6 -3
@@ -1969,7 +1969,7 @@ static inline void skb_set_queue_mapping
 	skb->queue_mapping = queue_mapping;
 }
 
-static inline u16 skb_get_queue_mapping(struct sk_buff *skb)
+static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
 {
 	return skb->queue_mapping;
 }
@@ -1984,16 +1984,19 @@ static inline void skb_record_rx_queue(s
 	skb->queue_mapping = rx_queue + 1;
 }
 
-static inline u16 skb_get_rx_queue(struct sk_buff *skb)
+static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
 {
 	return skb->queue_mapping - 1;
 }
 
-static inline bool skb_rx_queue_recorded(struct sk_buff *skb)
+static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
 {
 	return (skb->queue_mapping != 0);
 }
 
+extern u16 skb_tx_hash(const struct net_device *dev,
+		       const struct sk_buff *skb);
+
 #ifdef CONFIG_XFRM
 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
 {
net/core/dev.c +2 -1
@@ -1725,7 +1725,7 @@ out_kfree_skb:
 
 static u32 skb_tx_hashrnd;
 
-static u16 skb_tx_hash(struct net_device *dev, struct sk_buff *skb)
+u16 skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb)
 {
 	u32 hash;
 
@@ -1740,6 +1740,7 @@ static u16 skb_tx_hash(struct net_device
 
 	return (u16) (((u64) hash * dev->real_num_tx_queues) >> 32);
 }
+EXPORT_SYMBOL(skb_tx_hash);
 
 static struct netdev_queue *dev_pick_tx(struct net_device *dev,
 					struct sk_buff *skb)
#1 lore raw

PATCH 2/2 ixgbe: fix select_queue management

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just check the flag.

---
 drivers/net/ixgbe/ixgbe_dcb_nl.c |    8 --------
 drivers/net/ixgbe/ixgbe_main.c   |   11 +++++++++++
 2 files changed, 11 insertions(+), 8 deletions(-)
drivers/net/ixgbe/ixgbe_dcb_nl.c +0 -8
@@ -102,12 +102,6 @@ static u8 ixgbe_dcbnl_get_state(struct n
 	return !!(adapter->flags & IXGBE_FLAG_DCB_ENABLED);
 }
 
-static u16 ixgbe_dcb_select_queue(struct net_device *dev, struct sk_buff *skb)
-{
-	/* All traffic should default to class 0 */
-	return 0;
-}
-
 static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 {
 	u8 err = 0;
@@ -135,7 +129,6 @@ static u8 ixgbe_dcbnl_set_state(struct n
 		kfree(adapter->rx_ring);
 		adapter->tx_ring = NULL;
 		adapter->rx_ring = NULL;
-		netdev->select_queue = &ixgbe_dcb_select_queue;
 
 		adapter->flags &= ~IXGBE_FLAG_RSS_ENABLED;
 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
@@ -154,7 +147,6 @@ static u8 ixgbe_dcbnl_set_state(struct n
 			kfree(adapter->rx_ring);
 			adapter->tx_ring = NULL;
 			adapter->rx_ring = NULL;
-			netdev->select_queue = NULL;
 
 			adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
 			adapter->flags |= IXGBE_FLAG_RSS_ENABLED;
drivers/net/ixgbe/ixgbe_main.c +11 -0
@@ -4272,6 +4272,16 @@ static int ixgbe_maybe_stop_tx(struct ne
 	return __ixgbe_maybe_stop_tx(netdev, tx_ring, size);
 }
 
+static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb)
+{
+	struct ixgbe_adapter *adapter = netdev_priv(dev);
+
+	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
+		return 0;  /* All traffic should default to class 0 */
+
+	return skb_tx_hash(dev, skb);
+}
+
 static int ixgbe_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
 {
 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
@@ -4401,6 +4411,7 @@ static const struct net_device_ops ixgbe
 	.ndo_open 		= ixgbe_open,
 	.ndo_stop		= ixgbe_close,
 	.ndo_start_xmit		= ixgbe_xmit_frame,
+	.ndo_select_queue	= ixgbe_select_queue,
 	.ndo_get_stats		= ixgbe_get_stats,
 	.ndo_set_rx_mode        = ixgbe_set_rx_mode,
 	.ndo_set_multicast_list	= ixgbe_set_rx_mode,

------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com

#2 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management

Quoted 5 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just check the flag.

Thanks Stephen. I was looking at reassigning a DCB netdev_ops struct when DCB is enabled, and then having a default netdev_ops struct when it's not enabled. I agree the check is cleaner the way you have it below, but it's another conditional check in the Tx hotpath, which we have too many of in the first place.

On a related side note, why is the netdev_ops member of net_device declared const?

Cheers,
-PJ Waskiewicz

#3 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management

Quoted 6 lines · Peter P
Quoted 4 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just check the flag.

Quoted 7 lines
Quoted 1 line

Thanks Stephen. I was looking at reassigning a DCB netdev_ops struct when DCB is enabled, and then having a default netdev_ops struct when it's not enabled. I agree the check is cleaner the way you have it below, but it's another conditional check in the Tx hotpath, which we have too many of in the first place.

Changing number of tx queues is actually the fastest, since then indirection is not needed

Quoted 2 lines

On a related side note, why is the netdev_ops member of net_device declared const?

The purpose of having an ops structure is two fold. First, the ops can be in read-only section (if driver wants) to avoid cache issues. More importantly only one instance is necessary when there are multiple boards, or 1000's of vlans.

Quoted 2 lines

Cheers,
-PJ Waskiewicz

#4 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management

Quoted 20 lines · Stephen Hemminger
Quoted 6 lines · Peter P
Quoted 4 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just check the flag.

Quoted 7 lines
Quoted 1 line

Thanks Stephen. I was looking at reassigning a DCB netdev_ops struct when DCB is enabled, and then having a default netdev_ops struct when it's not enabled. I agree the check is cleaner the way you have it below, but it's another conditional check in the Tx hotpath, which we have too many of in the first place.

Changing number of tx queues is actually the fastest, since then indirection is not needed

Quoted 7 lines
Quoted 2 lines

On a related side note, why is the netdev_ops member of net_device declared const?

The purpose of having an ops structure is two fold. First, the ops can be in read-only section (if driver wants) to avoid cache issues. More importantly only one instance is necessary when there are multiple boards, or 1000's of vlans.

Ah, gotcha. Clears up my questions.

Acked-by
#5 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management

Quoted 30 lines · Peter P
Quoted 20 lines · Stephen Hemminger
Quoted 6 lines · Peter P
Quoted 4 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just check the flag.

Quoted 7 lines
Quoted 1 line

Thanks Stephen.  I was looking at reassigning a DCB netdev_ops struct when DCB is enabled, and then having a default netdev_ops struct when it's not enabled.  I agree the check is cleaner the way you have it below, but it's another conditional check in the Tx hotpath, which we have too many of in the first place.

Changing number of tx queues is actually the fastest, since then indirection is not needed

Quoted 7 lines
Quoted 2 lines

On a related side note, why is the netdev_ops member of net_device declared const?

The purpose of having an ops structure is two fold. First, the ops can be in read-only section (if driver wants) to avoid cache issues. More importantly only one instance is necessary when there are multiple boards, or 1000's of vlans.

Quoted 4 lines

Ah, gotcha.  Clears up my questions.

--

Dave - feel free to apply this patch when you take the first patch in this series, I am not adding this to me tree.

Signature · 13 lines
Cheers,
Jeff

------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are
powering Web 2.0 with engaging, cross-platform capabilities. Quickly and
easily build your RIAs with Flex Builder, the Eclipse(TM)based development
software that enables intelligent coding and step-through debugging.
Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
maintainer Applied
#6 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management

In reply to Waskiewicz Jr, Peter P

From: "Waskiewicz Jr, Peter P" <[email protected]>
Date: Fri, 20 Mar 2009 10:14:38 -0700 (Pacific Daylight Time)

Quoted 12 lines · Stephen Hemminger
Quoted 11 lines
Quoted 7 lines · Peter P
Quoted 5 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just check the flag.

...

Quoted 1 line

Applied.

#7 lore raw

PATCH 2/2 ixgbe: fix select_queue management (v2)

In reply to Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just change number of available transmit queues.

---
 drivers/net/ixgbe/ixgbe_dcb_nl.c |   48 +++++++++++++++++----------------------
 1 file changed, 21 insertions(+), 27 deletions(-)
drivers/net/ixgbe/ixgbe_dcb_nl.c +3 -8
@@ -102,12 +102,6 @@ static u8 ixgbe_dcbnl_get_state(struct n
 	return !!(adapter->flags & IXGBE_FLAG_DCB_ENABLED);
 }
 
-static u16 ixgbe_dcb_select_queue(struct net_device *dev, struct sk_buff *skb)
-{
-	/* All traffic should default to class 0 */
-	return 0;
-}
-
 static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 {
 	u8 err = 0;
@@ -135,7 +129,7 @@ static u8 ixgbe_dcbnl_set_state(struct n
 		kfree(adapter->rx_ring);
 		adapter->tx_ring = NULL;
 		adapter->rx_ring = NULL;
-		netdev->select_queue = &ixgbe_dcb_select_queue;
+		netdev->real_num_tx_queues = 1;
 
 		adapter->flags &= ~IXGBE_FLAG_RSS_ENABLED;
 		adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
@@ -147,6 +141,7 @@ static u8 ixgbe_dcbnl_set_state(struct n
 		if (adapter->flags & IXGBE_FLAG_DCB_ENABLED) {
 			if (netif_running(netdev))
 				netdev->netdev_ops->ndo_stop(netdev);
+
 			ixgbe_reset_interrupt_capability(adapter);
 			ixgbe_napi_del_all(adapter);
 			INIT_LIST_HEAD(&netdev->napi_list);
@@ -154,7 +149,7 @@ static u8 ixgbe_dcbnl_set_state(struct n
 			kfree(adapter->rx_ring);
 			adapter->tx_ring = NULL;
 			adapter->rx_ring = NULL;
-			netdev->select_queue = NULL;
+			netdev->real_num_tx_queues = MAX_TX_QUEUES;
 
 			adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
 			adapter->flags |= IXGBE_FLAG_RSS_ENABLED;
#8 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

Quoted 26 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just change number of available transmit queues.

---
 drivers/net/ixgbe/ixgbe_dcb_nl.c |   48 +++++++++++++++++----------------------
 1 file changed, 21 insertions(+), 27 deletions(-)
drivers/net/ixgbe/ixgbe_dcb_nl.c +0 -6
@@ -102,12 +102,6 @@ static u8 ixgbe_dcbnl_get_state(struct n
 	return !!(adapter->flags & IXGBE_FLAG_DCB_ENABLED);
 }
 
-static u16 ixgbe_dcb_select_queue(struct net_device *dev, struct sk_buff *skb)
-{
-	/* All traffic should default to class 0 */
-	return 0;
-}
-
 static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 {
 	u8 err = 0;

drivers/net/ixgbe/ixgbe_dcb_nl.c static u8 ixgbe_dcbnl_set_state(struct n · line 132

Quoted 5 lines
drivers/net/ixgbe/ixgbe_dcb_nl.c +1 -1
@@ -135,7 +129,7 @@ static u8 ixgbe_dcbnl_set_state(struct n
 		kfree(adapter->rx_ring);
 		adapter->tx_ring = NULL;
 		adapter->rx_ring = NULL;
-		netdev->select_queue = &ixgbe_dcb_select_queue;
+		netdev->real_num_tx_queues = 1;

NAK. The point of dcb_select_queue() isn't because DCB mode only uses 1 Tx queue. DCB has 8 priorities, and allocates 8 Tx queues, one for each priority. The DCB spec says that any traffic not being filtered by some kind of mechanism needs to go through priority 0, or queue 0. So select_queue is meant to tag all traffic to queue 0, then have the attached qdisc and tc filters get the majority of the traffic into the different priority queues.

If we did not push the unfiltered traffic into queue 0, then skb_tx_hash() would put traffic randomly into queues with higher priority, which is not what we want.

I'd prefer your original patch to fix this up, where you check if DCB is enabled, and return 0.

-PJ Waskiewicz

#9 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

Quoted 42 lines · Peter P
Quoted 26 lines · Stephen Hemminger

Convert ixgbe to use net_device_ops properly. Rather than changing the select_queue function pointer just change number of available transmit queues.

---
 drivers/net/ixgbe/ixgbe_dcb_nl.c |   48 +++++++++++++++++----------------------
 1 file changed, 21 insertions(+), 27 deletions(-)
drivers/net/ixgbe/ixgbe_dcb_nl.c +0 -6
@@ -102,12 +102,6 @@ static u8 ixgbe_dcbnl_get_state(struct n
 	return !!(adapter->flags & IXGBE_FLAG_DCB_ENABLED);
 }
 
-static u16 ixgbe_dcb_select_queue(struct net_device *dev, struct sk_buff *skb)
-{
-	/* All traffic should default to class 0 */
-	return 0;
-}
-
 static u8 ixgbe_dcbnl_set_state(struct net_device *netdev, u8 state)
 {
 	u8 err = 0;

drivers/net/ixgbe/ixgbe_dcb_nl.c static u8 ixgbe_dcbnl_set_state(struct n · line 132

Quoted 5 lines
drivers/net/ixgbe/ixgbe_dcb_nl.c +1 -1
@@ -135,7 +129,7 @@ static u8 ixgbe_dcbnl_set_state(struct n
 		kfree(adapter->rx_ring);
 		adapter->tx_ring = NULL;
 		adapter->rx_ring = NULL;
-		netdev->select_queue = &ixgbe_dcb_select_queue;
+		netdev->real_num_tx_queues = 1;

NAK. The point of dcb_select_queue() isn't because DCB mode only uses 1 Tx queue. DCB has 8 priorities, and allocates 8 Tx queues, one for each priority. The DCB spec says that any traffic not being filtered by some kind of mechanism needs to go through priority 0, or queue 0. So select_queue is meant to tag all traffic to queue 0, then have the attached qdisc and tc filters get the majority of the traffic into the different priority queues.

Quoted 8 lines

If we did not push the unfiltered traffic into queue 0, then skb_tx_hash() would put traffic randomly into queues with higher priority, which is not what we want.

I'd prefer your original patch to fix this up, where you check if DCB is enabled, and return 0.

-PJ Waskiewicz

The default select queue function in the kernel is:

static struct netdev_queue *dev_pick_tx(struct net_device *dev,
struct sk_buff *skb)

{
	const struct net_device_ops *ops = dev->netdev_ops;
	u16 queue_index = 0;

	if (ops->ndo_select_queue)
		queue_index = ops->ndo_select_queue(dev, skb);
	else if (dev->real_num_tx_queues > 1)
		queue_index = skb_tx_hash(dev, skb);

	skb_set_queue_mapping(skb, queue_index);
	return netdev_get_tx_queue(dev, queue_index);
}

So if driver (re)sets real_num_tx_queues to 1 then queue_index will always 0 and all traffic will go to one queue. This is the same as having your own select_queue function.

#10 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

Quoted 17 lines · Stephen Hemminger

The default select queue function in the kernel is:

static struct netdev_queue *dev_pick_tx(struct net_device *dev,
struct sk_buff *skb)

{
	const struct net_device_ops *ops = dev->netdev_ops;
	u16 queue_index = 0;

	if (ops->ndo_select_queue)
		queue_index = ops->ndo_select_queue(dev, skb);
	else if (dev->real_num_tx_queues > 1)
		queue_index = skb_tx_hash(dev, skb);

	skb_set_queue_mapping(skb, queue_index);
	return netdev_get_tx_queue(dev, queue_index);
}
Quoted 3 lines

So if driver (re)sets real_num_tx_queues to 1 then queue_index will always 0 and all traffic will go to one queue. This is the same as having your own select_queue function.

I see your point, but it is a hack in my opinion. The device will have 8 real Tx queues, not 1. I'd much rather go with the original proposal, since if the code in dev_pick_tx() changed, it could silently break ixgbe.

-PJ Waskiewicz

maintainer
#11 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

From: "Waskiewicz Jr, Peter P" <[email protected]>
Date: Fri, 20 Mar 2009 23:21:38 -0700 (Pacific Daylight Time)

Quoted 3 lines

I see your point, but it is a hack in my opinion. The device will have 8 real Tx queues, not 1. I'd much rather go with the original proposal, since if the code in dev_pick_tx() changed, it could silently break ixgbe.

It can't, if you only advertise one transmit queue the kernel can never ever choose anything other than queue zero. It's impossible.

Stephen's right, you guys don't need your select queue override.

And if you recall I suspected this from the very beginning.

You guys never ever think out of the box, ever... if it's not straightforward, you guys won't got for it. That makes it very frustrating to get anything done.

------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com

#12 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

Quoted 11 lines · David Miller

From: "Waskiewicz Jr, Peter P" <[email protected]>
Date: Fri, 20 Mar 2009 23:21:38 -0700 (Pacific Daylight Time)

Quoted 3 lines

I see your point, but it is a hack in my opinion. The device will have 8 real Tx queues, not 1. I'd much rather go with the original proposal, since if the code in dev_pick_tx() changed, it could silently break ixgbe.

It can't, if you only advertise one transmit queue the kernel can never ever choose anything other than queue zero. It's impossible.

Quoted 7 lines

Stephen's right, you guys don't need your select queue override.

And if you recall I suspected this from the very beginning.

You guys never ever think out of the box, ever... if it's not straightforward, you guys won't got for it. That makes it very frustrating to get anything done.

This patch will break DCB in ixgbe. We need all 8 queues, because the user will be assigning tc filters to the sch_multiq qdisc to get traffic into priority queues. If we take Stephen's patch and tell the stack we have 1 real_num_tx_queues, then we get 1 band in sch_multiq, which makes it impossible to assign traffic to priorities 1 through 8:

static int multiq_tune(struct Qdisc *sch, struct nlattr *opt)

{
        struct multiq_sched_data *q = qdisc_priv(sch);
        struct tc_multiq_qopt *qopt;
        int i;

        if (!netif_is_multiqueue(qdisc_dev(sch)))
                return -EOPNOTSUPP;
        if (nla_len(opt) < sizeof(*qopt))
                return -EINVAL;

        qopt = nla_data(opt);

        qopt->bands = qdisc_dev(sch)->real_num_tx_queues;

This is not what we want, rather, we want all 8 Tx queues that we expose. The only reason we override the select_queue is to catch the unfiltered traffic and send it to queue 0.

Cheers,
-PJ Waskiewicz

#13 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

Quoted 21 lines · Peter P
Quoted 11 lines · David Miller

From: "Waskiewicz Jr, Peter P" <[email protected]>
Date: Fri, 20 Mar 2009 23:21:38 -0700 (Pacific Daylight Time)

Quoted 3 lines

I see your point, but it is a hack in my opinion. The device will have 8 real Tx queues, not 1. I'd much rather go with the original proposal, since if the code in dev_pick_tx() changed, it could silently break ixgbe.

It can't, if you only advertise one transmit queue the kernel can never ever choose anything other than queue zero. It's impossible.

Quoted 7 lines

Stephen's right, you guys don't need your select queue override.

And if you recall I suspected this from the very beginning.

You guys never ever think out of the box, ever... if it's not straightforward, you guys won't got for it. That makes it very frustrating to get anything done.

Quoted 6 lines

This patch will break DCB in ixgbe. We need all 8 queues, because the user will be assigning tc filters to the sch_multiq qdisc to get traffic into priority queues. If we take Stephen's patch and tell the stack we have 1 real_num_tx_queues, then we get 1 band in sch_multiq, which makes it impossible to assign traffic to priorities 1 through 8:

How does it make sense to say you have 8 bands, but only use one.

------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com

#14 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

Quoted 24 lines · Stephen Hemminger
Quoted 21 lines · Peter P
Quoted 18 lines · David Miller

From: "Waskiewicz Jr, Peter P" <[email protected]>
Date: Fri, 20 Mar 2009 23:21:38 -0700 (Pacific Daylight Time)

Quoted 3 lines

I see your point, but it is a hack in my opinion. The device will have 8 real Tx queues, not 1. I'd much rather go with the original proposal, since if the code in dev_pick_tx() changed, it could silently break ixgbe.

It can't, if you only advertise one transmit queue the kernel can never ever choose anything other than queue zero. It's impossible.

Stephen's right, you guys don't need your select queue override.

And if you recall I suspected this from the very beginning.

You guys never ever think out of the box, ever... if it's not straightforward, you guys won't got for it. That makes it very frustrating to get anything done.

Quoted 8 lines
Quoted 6 lines

This patch will break DCB in ixgbe. We need all 8 queues, because the user will be assigning tc filters to the sch_multiq qdisc to get traffic into priority queues. If we take Stephen's patch and tell the stack we have 1 real_num_tx_queues, then we get 1 band in sch_multiq, which makes it impossible to assign traffic to priorities 1 through 8:

How does it make sense to say you have 8 bands, but only use one.

That's not how DCB works. We use sch_multiq to identify the traffic we want to put into the 8 bands. So in other words, the user will add tc filters to move the traffic around. We override select queue to filter the rest of the traffic into a single queue, so we don't randomly put traffic into the other hardware priority queues.

Either way, thanks for cleaning this up Stephen. It was something I needed to do, and haven't done yet. So I very much appreciate it.

Cheers,
-PJ Waskiewicz

------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com

maintainer
#15 lore raw

Re: [PATCH 2/2] ixgbe: fix select_queue management (v2)

From: "Waskiewicz Jr, Peter P" <[email protected]>
Date: Sat, 21 Mar 2009 18:48:18 -0700 (Pacific Daylight Time)

Quoted 5 lines

That's not how DCB works. We use sch_multiq to identify the traffic we want to put into the 8 bands. So in other words, the user will add tc filters to move the traffic around. We override select queue to filter the rest of the traffic into a single queue, so we don't randomly put traffic into the other hardware priority queues.

It's escaping me why the multiq rules can't handle this?

In the end, it's a decision of where the logic lives. Currently the default handling logic is in the ->select_queue() override, and I'm still not at all convinced it has to be there.

------------------------------------------------------------------------------
Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com