From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933178AbXBWTF6 (ORCPT ); Fri, 23 Feb 2007 14:05:58 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S933179AbXBWTF6 (ORCPT ); Fri, 23 Feb 2007 14:05:58 -0500 Received: from mga03.intel.com ([143.182.124.21]:10288 "EHLO mga03.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933178AbXBWTFz convert rfc822-to-8bit (ORCPT ); Fri, 23 Feb 2007 14:05:55 -0500 X-ExtLoop1: 1 X-IronPort-AV: i="4.14,212,1170662400"; d="scan'208"; a="186269680:sNHT1291657598" X-MimeOLE: Produced By Microsoft Exchange V6.5 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 8BIT Subject: RE: [PATCH 1/2] NET: Multiple queue network device support Date: Fri, 23 Feb 2007 11:05:50 -0800 Message-ID: In-Reply-To: <78C9135A3D2ECE4B8162EBDCE82CAD77013F383F@nekter> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PATCH 1/2] NET: Multiple queue network device support thread-index: AcdL3LXJN6SzN3nAQOClpFmdrE0iwwLS8fmwAAACOYAAAABb4AAVLVsQ From: "Waskiewicz Jr, Peter P" To: "Sreenivasa Honnur" , "David Miller" , "Garzik, Jeff" , , Cc: "Kok, Auke" X-OriginalArrivalTime: 23 Feb 2007 19:05:51.0802 (UTC) FILETIME=[A2D931A0:01C7577D] Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org map_queue will always return 0 in the pfifo_fast Qdisc. This is because pfifo_fast is the default scheduler when a device is brought up. I didn't want to make any assumptions about a device's functionality, so using queue 0 seems the least dangerous. However, in the prio qdisc, there is a small problem with the queue to band assignment logic. If you have the same number of bands as queues (e.g. 4 prio bands and 4 Tx queues), then you will have all bands assigned to queue 0. I have a fix for that, which I plan to send when I finish up a documentation piece for this patch's repost. Thanks for the feedback, PJ Waskiewicz > -----Original Message----- > From: Sreenivasa Honnur [mailto:Sreenivasa.Honnur@neterion.com] > Sent: Friday, February 23, 2007 1:01 AM > To: Kok, Auke-jan H; David Miller; Garzik, Jeff; > netdev@vger.kernel.org; linux-kernel@vger.kernel.org > Cc: Waskiewicz Jr, Peter P; Brandeburg, Jesse; Kok, Auke; > Ronciak, John > Subject: RE: [PATCH 1/2] NET: Multiple queue network device support > > Fucntion "map_queue" returns queue index as '0'. There is no > support to return different queue indexes. > > -----Original Message----- > From: netdev-owner@vger.kernel.org > [mailto:netdev-owner@vger.kernel.org] > On Behalf Of Kok, Auke > Sent: Friday, February 09, 2007 5:40 AM > To: David Miller; Garzik, Jeff; netdev@vger.kernel.org; > linux-kernel@vger.kernel.org > Cc: Kok, Auke; Peter Waskiewicz Jr; Brandeburg, Jesse; Kok, > Auke; Ronciak, John > Subject: [PATCH 1/2] NET: Multiple queue network device support > > > From: Peter Waskiewicz Jr > > Added an API and associated supporting routines for > multiqueue network devices. This allows network devices > supporting multiple TX queues to configure each queue within > the netdevice and manage each queue independantly. Changes > to the PRIO Qdisc also allow a user to map multiple flows to > individual TX queues, taking advantage of each queue on the device. > > Signed-off-by: Peter Waskiewicz Jr > Signed-off-by: Auke Kok > --- > > include/linux/netdevice.h | 73 ++++++++++++++++++++++++++++ > include/net/sch_generic.h | 3 + > net/Kconfig | 20 ++++++++ > net/core/dev.c | 51 ++++++++++++++++++++ > net/sched/sch_generic.c | 117 > +++++++++++++++++++++++++++++++++++++++++++++ > net/sched/sch_prio.c | 106 > ++++++++++++++++++++++++++++++++++++++--- > 6 files changed, 364 insertions(+), 6 deletions(-) > > diff --git a/include/linux/netdevice.h > b/include/linux/netdevice.h index > 2e37f50..c7f94a8 100644 > --- a/include/linux/netdevice.h > +++ b/include/linux/netdevice.h > @@ -106,6 +106,16 @@ struct netpoll_info; #define MAX_HEADER > (LL_MAX_HEADER + 48) #endif > > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + > +struct net_device_subqueue > +{ > + /* Give a lock and a flow control state for each queue */ > + unsigned long state; > + spinlock_t queue_lock ____cacheline_aligned_in_smp; > +}; > +#endif > + > /* > * Network device statistics. Akin to the 2.0 ether stats but > * with byte counters. > @@ -339,6 +349,10 @@ struct net_device > #define NETIF_F_GEN_CSUM (NETIF_F_NO_CSUM | NETIF_F_HW_CSUM) > #define NETIF_F_ALL_CSUM (NETIF_F_IP_CSUM | NETIF_F_GEN_CSUM) > > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > +#define NETIF_F_MULTI_QUEUE 8192 /* Has multiple TX/RX queues */ > +#endif > + > struct net_device *next_sched; > > /* Interface index. Unique device identifier */ > @@ -532,6 +546,19 @@ struct net_device > struct device dev; > /* space for optional statistics and wireless sysfs groups */ > struct attribute_group *sysfs_groups[3]; > + > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + /* To retrieve statistics per subqueue - FOR FUTURE USE */ > + struct net_device_stats* (*get_subqueue_stats)(struct net_device > *dev, > + int > queue_index); > + > + /* The TX queue control structures */ > + struct net_device_subqueue *egress_subqueue; > + int egress_subqueue_count; > + int (*hard_start_subqueue_xmit)(struct > sk_buff *skb, > + struct > net_device *dev, > + int > queue_index); > +#endif /* CONFIG_NET_MULTI_QUEUE_DEVICE */ > }; > #define to_net_dev(d) container_of(d, struct net_device, dev) > > @@ -673,6 +700,52 @@ static inline int netif_running(const > struct net_device *dev) > return test_bit(__LINK_STATE_START, &dev->state); } > > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + > +/* > + * Routines to manage the subqueues on a device. We only need start > + * stop, and a check if it's stopped. All other device management is > + * done at the overall netdevice level. > + * Also test the netif state if we're multi-queue at all. > + */ > +static inline void netif_start_subqueue(struct net_device *dev, int > +queue_index) { > + clear_bit(__LINK_STATE_XOFF, > +&dev->egress_subqueue[queue_index].state); > +} > + > +static inline void netif_stop_subqueue(struct net_device *dev, int > +queue_index) { #ifdef CONFIG_NETPOLL_TRAP > + if (netpoll_trap()) > + return; > +#endif > + set_bit(__LINK_STATE_XOFF, > &dev->egress_subqueue[queue_index].state); > +} > + > +static inline int netif_subqueue_stopped(const struct > net_device *dev, > + int queue_index) { > + return test_bit(__LINK_STATE_XOFF, > + &dev->egress_subqueue[queue_index].state); > +} > + > +static inline void netif_wake_subqueue(struct net_device *dev, int > +queue_index) { #ifdef CONFIG_NETPOLL_TRAP > + if (netpoll_trap()) > + return; > +#endif > + if (test_and_clear_bit(__LINK_STATE_XOFF, > + > &dev->egress_subqueue[queue_index].state)) > + __netif_schedule(dev); > +} > + > +static inline int netif_is_multiqueue(const struct net_device *dev) { > + return (!!(NETIF_F_MULTI_QUEUE & dev->features)); } #endif /* > +CONFIG_NET_MULTI_QUEUE_DEVICE */ > + > > /* Use this variant when it is known for sure that it > * is executing from interrupt context. > diff --git a/include/net/sch_generic.h > b/include/net/sch_generic.h index > 8208639..8751ba6 100644 > --- a/include/net/sch_generic.h > +++ b/include/net/sch_generic.h > @@ -102,6 +102,9 @@ struct Qdisc_ops > > int (*dump)(struct Qdisc *, struct sk_buff > *); > int (*dump_stats)(struct Qdisc *, struct > gnet_dump *); > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int (*map_queue)(struct sk_buff *, struct > Qdisc *); > +#endif > > struct module *owner; > }; > diff --git a/net/Kconfig b/net/Kconfig > index 7dfc949..796edb3 100644 > --- a/net/Kconfig > +++ b/net/Kconfig > @@ -38,6 +38,26 @@ source "net/packet/Kconfig" > source "net/unix/Kconfig" > source "net/xfrm/Kconfig" > > +config NET_MULTI_QUEUE_DEVICE > + bool "Multiple queue network device support (EXPERIMENTAL)" > + depends on NET_SCHED && EXPERIMENTAL > + help > + Saying Y here will add support for network devices that have > more than > + one physical TX queue and/or RX queue. > + > + Multiple queue devices will require qdiscs that understand how > to > + queue to multiple targets. The default packet scheduler will > default > + to the first queue in a device. In other words, if you need > the > + ability to spread traffic across queues, your queueing > discipline > + needs to know how to do that. > + > + Note that saying Y here will give preferential treatment to > multiple > + queue devices in the network stack. A slight drop in > single-queue > + device performance may be seen. > + > + Say N here unless you know your network device supports > multiple > + TX and/or RX queues. > + > config INET > bool "TCP/IP networking" > ---help--- > diff --git a/net/core/dev.c b/net/core/dev.c index 455d589..42b635c > 100644 > --- a/net/core/dev.c > +++ b/net/core/dev.c > @@ -1477,6 +1477,49 @@ gso: > skb->tc_verd = SET_TC_AT(skb->tc_verd,AT_EGRESS); > #endif > if (q->enqueue) { > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int queue_index; > + /* If we're a multi-queue device, get a queue index to > lock */ > + if (netif_is_multiqueue(dev)) > + { > + /* Get the queue index and lock it. */ > + if (likely(q->ops->map_queue)) { > + queue_index = q->ops->map_queue(skb, q); > + > spin_lock(&dev->egress_subqueue[queue_index].queue_lock); > + rc = q->enqueue(skb, q); > + /* > + * Unlock because the underlying qdisc > + * may queue and send a packet from a > + * different queue. > + */ > + > spin_unlock(&dev->egress_subqueue[queue_index].queue_lock); > + qdisc_run(dev); > + rc = rc == NET_XMIT_BYPASS > + ? NET_XMIT_SUCCESS : rc; > + goto out; > + } else { > + printk(KERN_CRIT "Device %s is " > + "multiqueue, but map_queue is " > + "undefined in the qdisc!!\n", > + dev->name); > + kfree_skb(skb); > + } > + } else { > + /* We're not a multi-queue device. */ > + spin_lock(&dev->queue_lock); > + q = dev->qdisc; > + if (q->enqueue) { > + rc = q->enqueue(skb, q); > + qdisc_run(dev); > + spin_unlock(&dev->queue_lock); > + rc = rc == NET_XMIT_BYPASS > + ? NET_XMIT_SUCCESS : rc; > + goto out; > + } > + spin_unlock(&dev->queue_lock); > + } > +#else /* No multi-queue support compiled in */ > + > /* Grab device queue */ > spin_lock(&dev->queue_lock); > q = dev->qdisc; > @@ -1489,6 +1532,7 @@ gso: > goto out; > } > spin_unlock(&dev->queue_lock); > +#endif /* CONFIG_NET_MULTI_QUEUE_DEVICE */ > } > > /* The device has no queue. Common case for software devices: > @@ -2879,6 +2923,9 @@ int register_netdevice(struct net_device *dev) > struct hlist_head *head; > struct hlist_node *p; > int ret; > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int i; > +#endif > > BUG_ON(dev_boot_phase); > ASSERT_RTNL(); > @@ -2894,6 +2941,10 @@ int register_netdevice(struct > net_device *dev) #ifdef CONFIG_NET_CLS_ACT > spin_lock_init(&dev->ingress_lock); > #endif > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + for (i = 0; i < dev->egress_subqueue_count; i++) > + spin_lock_init(&dev->egress_subqueue[i].queue_lock); > +#endif > > dev->iflink = -1; > > diff --git a/net/sched/sch_generic.c > b/net/sched/sch_generic.c index bc116bd..62ca23e 100644 > --- a/net/sched/sch_generic.c > +++ b/net/sched/sch_generic.c > @@ -42,6 +42,7 @@ > to data, participating in scheduling must be additionally > protected with dev->queue_lock spinlock. > > + For single-queue devices: > The idea is the following: > - enqueue, dequeue are serialized via top level device > spinlock dev->queue_lock. > @@ -51,6 +52,12 @@ > hence this lock may be made without local bh disabling. > > qdisc_tree_lock must be grabbed BEFORE dev->queue_lock! > + > + For multi-queue devices: > + - enqueue, dequeue are serialized with individual queue locks, > + dev->egress_subqueue[queue_index].queue_lock. > + - Everything else with tree protection with single queue devices > apply > + to multiqueue devices. > */ > DEFINE_RWLOCK(qdisc_tree_lock); > > @@ -92,6 +99,9 @@ static inline int qdisc_restart(struct net_device > *dev) { > struct Qdisc *q = dev->qdisc; > struct sk_buff *skb; > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int queue_index = 0; > +#endif > > /* Dequeue packet */ > if (((skb = dev->gso_skb)) || ((skb = q->dequeue(q)))) { @@ > -130,6 +140,72 @@ static inline int qdisc_restart(struct net_device > *dev) > } > > { > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + if (netif_is_multiqueue(dev)) { > + if (likely(q->ops->map_queue)) { > + queue_index = > q->ops->map_queue(skb, q); > + } else { > + printk(KERN_CRIT "Device %s is " > + "multiqueue, but > map_queue is " > + "undefined in the > qdisc!!\n", > + dev->name); > + goto requeue; > + } > + > spin_unlock(&dev->egress_subqueue[queue_index].queue_lock); > + /* Check top level device, and any > sub-device */ > + if ((!netif_queue_stopped(dev)) && > + (!netif_subqueue_stopped(dev, > queue_index))) { > + int ret; > + ret = > dev->hard_start_subqueue_xmit(skb, dev, queue_index); > + if (ret == NETDEV_TX_OK) { > + if (!nolock) { > + > netif_tx_unlock(dev); > + } > + return -1; > + } > + if (ret == NETDEV_TX_LOCKED && > nolock) { > + > spin_lock(&dev->egress_subqueue[queue_index].queue_lock); > + goto collision; > + } > + } > + /* NETDEV_TX_BUSY - we need to requeue > */ > + /* Release the driver */ > + if (!nolock) { > + netif_tx_unlock(dev); > + } > + > spin_lock(&dev->egress_subqueue[queue_index].queue_lock); > + q = dev->qdisc; > + } > + else > + { > + /* We're a single-queue device */ > + /* And release queue */ > + spin_unlock(&dev->queue_lock); > + if (!netif_queue_stopped(dev)) { > + int ret; > + > + ret = dev->hard_start_xmit(skb, > dev); > + if (ret == NETDEV_TX_OK) { > + if (!nolock) { > + > netif_tx_unlock(dev); > + } > + > spin_lock(&dev->queue_lock); > + return -1; > + } > + if (ret == NETDEV_TX_LOCKED && > nolock) { > + > spin_lock(&dev->queue_lock); > + goto collision; > + } > + } > + /* NETDEV_TX_BUSY - we need to requeue > */ > + /* Release the driver */ > + if (!nolock) { > + netif_tx_unlock(dev); > + } > + spin_lock(&dev->queue_lock); > + q = dev->qdisc; > + } > +#else /* CONFIG_NET_MULTI_QUEUE_DEVICE */ > /* And release queue */ > spin_unlock(&dev->queue_lock); > > @@ -157,6 +233,7 @@ static inline int qdisc_restart(struct net_device > *dev) > } > spin_lock(&dev->queue_lock); > q = dev->qdisc; > +#endif /* CONFIG_NET_MULTI_QUEUE_DEVICE */ > } > > /* Device kicked us out :( > @@ -175,9 +252,17 @@ requeue: > else > q->ops->requeue(skb, q); > netif_schedule(dev); > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + if (netif_is_multiqueue(dev)) > + > spin_unlock(&dev->egress_subqueue[queue_index].queue_lock); > +#endif > return 1; > } > BUG_ON((int) q->q.qlen < 0); > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + if (netif_is_multiqueue(dev)) > + > spin_unlock(&dev->egress_subqueue[queue_index].queue_lock); > +#endif > return q->q.qlen; > } > > @@ -287,12 +372,22 @@ static int noop_requeue(struct sk_buff > *skb, struct Qdisc* qdisc) > return NET_XMIT_CN; > } > > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > +static int noop_map_queue(struct sk_buff *skb, struct Qdisc *qdisc) { > + return 0; > +} > +#endif > + > struct Qdisc_ops noop_qdisc_ops = { > .id = "noop", > .priv_size = 0, > .enqueue = noop_enqueue, > .dequeue = noop_dequeue, > .requeue = noop_requeue, > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + .map_queue = noop_map_queue, > +#endif > .owner = THIS_MODULE, > }; > > @@ -310,6 +405,9 @@ static struct Qdisc_ops noqueue_qdisc_ops = { > .enqueue = noop_enqueue, > .dequeue = noop_dequeue, > .requeue = noop_requeue, > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + .map_queue = noop_map_queue, > +#endif > .owner = THIS_MODULE, > }; > > @@ -356,10 +454,18 @@ static struct sk_buff *pfifo_fast_dequeue(struct > Qdisc* qdisc) > struct sk_buff_head *list = qdisc_priv(qdisc); > > for (prio = 0; prio < PFIFO_FAST_BANDS; prio++) { > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + if (netif_is_multiqueue(qdisc->dev)) > + > spin_lock(&qdisc->dev->egress_subqueue[0].queue_lock); > +#endif > if (!skb_queue_empty(list + prio)) { > qdisc->q.qlen--; > return __qdisc_dequeue_head(qdisc, list + prio); > } > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + if (netif_is_multiqueue(qdisc->dev)) > + > spin_unlock(&qdisc->dev->egress_subqueue[0].queue_lock); > +#endif > } > > return NULL; > @@ -406,6 +512,14 @@ static int pfifo_fast_init(struct Qdisc > *qdisc, struct rtattr *opt) > return 0; > } > > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > +static int pfifo_fast_map_queue(struct sk_buff *skb, struct Qdisc > +*qdisc) { > + /* Generic qdisc, so always return queue index 0. */ > + return 0; > +} > +#endif > + > static struct Qdisc_ops pfifo_fast_ops = { > .id = "pfifo_fast", > .priv_size = PFIFO_FAST_BANDS * sizeof(struct > sk_buff_head), > @@ -415,6 +529,9 @@ static struct Qdisc_ops pfifo_fast_ops = { > .init = pfifo_fast_init, > .reset = pfifo_fast_reset, > .dump = pfifo_fast_dump, > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + .map_queue = pfifo_fast_map_queue, > +#endif > .owner = THIS_MODULE, > }; > > diff --git a/net/sched/sch_prio.c b/net/sched/sch_prio.c index > 2567b4c..0e24071 100644 > --- a/net/sched/sch_prio.c > +++ b/net/sched/sch_prio.c > @@ -43,16 +43,19 @@ struct prio_sched_data > struct tcf_proto *filter_list; > u8 prio2band[TC_PRIO_MAX+1]; > struct Qdisc *queues[TCQ_PRIO_BANDS]; > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int band2queue[TC_PRIO_MAX + 1]; > +#endif > }; > > - > static struct Qdisc * > -prio_classify(struct sk_buff *skb, struct Qdisc *sch, int *qerr) > +prio_classify(struct sk_buff *skb, struct Qdisc *sch, int *qerr, int > +*retband) > { > struct prio_sched_data *q = qdisc_priv(sch); > u32 band = skb->priority; > struct tcf_result res; > > + *retband = band; > *qerr = NET_XMIT_BYPASS; > if (TC_H_MAJ(skb->priority) != sch->handle) { #ifdef > CONFIG_NET_CLS_ACT @@ -68,13 +71,16 @@ prio_classify(struct > sk_buff *skb, struct Qdisc *sch, int *qerr) #else > if (!q->filter_list || tc_classify(skb, q->filter_list, > &res)) { #endif > - if (TC_H_MAJ(band)) > + if (TC_H_MAJ(band)) { > band = 0; > + *retband = 0; > + } > return > q->queues[q->prio2band[band&TC_PRIO_MAX]]; > } > band = res.classid; > } > band = TC_H_MIN(band) - 1; > + *retband = band; > if (band > q->bands) > return q->queues[q->prio2band[0]]; > > @@ -86,8 +92,15 @@ prio_enqueue(struct sk_buff *skb, struct > Qdisc *sch) { > struct Qdisc *qdisc; > int ret; > + int band; /* Unused in this function, just here for multi-queue > */ > + > + /* > + * NOTE: if a device is multiqueue, then it is imperative the > + * underlying qdisc NOT be another PRIO qdisc. Otherwise we're > going > + * to deadlock. Fixme please. > + */ > + qdisc = prio_classify(skb, sch, &ret, &band); > > - qdisc = prio_classify(skb, sch, &ret); > #ifdef CONFIG_NET_CLS_ACT > if (qdisc == NULL) { > > @@ -108,14 +121,34 @@ prio_enqueue(struct sk_buff *skb, struct Qdisc > *sch) > return ret; > } > > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > +static int > +prio_map_queue(struct sk_buff *skb, struct Qdisc *sch) { > + /* > + * We'll classify the skb here, so return the Qdisc back through > + * the function call. The return value is the queue we need to > + * lock, which will be the value returned through band. This > will help > + * speed up enqueue, since it won't have to do the classify > again. > + */ > + int ret; > + int band; > + struct prio_sched_data *q = qdisc_priv(sch); > + > + sch = prio_classify(skb, sch, &ret, &band); > + return q->band2queue[band]; > +} > +#endif > > static int > prio_requeue(struct sk_buff *skb, struct Qdisc* sch) { > struct Qdisc *qdisc; > int ret; > + int band; > + > + qdisc = prio_classify(skb, sch, &ret, &band); > > - qdisc = prio_classify(skb, sch, &ret); > #ifdef CONFIG_NET_CLS_ACT > if (qdisc == NULL) { > if (ret == NET_XMIT_BYPASS) > @@ -141,18 +174,53 @@ prio_dequeue(struct Qdisc* sch) > struct sk_buff *skb; > struct prio_sched_data *q = qdisc_priv(sch); > int prio; > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int queue; > +#endif > struct Qdisc *qdisc; > > + /* > + * If we're multiqueue, the basic approach is try the lock on > each > + * queue. If it's locked, either we're already dequeuing, or > enqueue > + * is doing something. Go to the next band if we're locked. > Once we > + * have a packet, unlock the queue. NOTE: the underlying qdisc > CANNOT > + * be a PRIO qdisc, otherwise we will deadlock. FIXME > + */ > for (prio = 0; prio < q->bands; prio++) { > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + if (netif_is_multiqueue(sch->dev)) { > + queue = q->band2queue[prio]; > + if > (spin_trylock(&sch->dev->egress_subqueue[queue].queue_lock)) { > + qdisc = q->queues[prio]; > + skb = qdisc->dequeue(qdisc); > + if (skb) { > + sch->q.qlen--; > + skb->priority = prio; > + > spin_unlock(&sch->dev->egress_subqueue[queue].queue_lock); > + return skb; > + } > + > spin_unlock(&sch->dev->egress_subqueue[queue].queue_lock); > + } > + } else { > + qdisc = q->queues[prio]; > + skb = qdisc->dequeue(qdisc); > + if (skb) { > + sch->q.qlen--; > + skb->priority = prio; > + return skb; > + } > + } > +#else > qdisc = q->queues[prio]; > skb = qdisc->dequeue(qdisc); > if (skb) { > sch->q.qlen--; > + skb->priority = prio; > return skb; > } > +#endif > } > return NULL; > - > } > > static unsigned int prio_drop(struct Qdisc* sch) @@ -205,6 > +273,10 @@ static int prio_tune(struct Qdisc *sch, struct rtattr *opt) > struct prio_sched_data *q = qdisc_priv(sch); > struct tc_prio_qopt *qopt = RTA_DATA(opt); > int i; > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + int queue; > + int qmapoffset; > +#endif > > if (opt->rta_len < RTA_LENGTH(sizeof(*qopt))) > return -EINVAL; > @@ -247,6 +319,25 @@ static int prio_tune(struct Qdisc *sch, > struct rtattr *opt) > } > } > } > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + /* setup queue to band mapping */ > + if (netif_is_multiqueue(sch->dev)) { > + if (q->bands == sch->dev->egress_subqueue_count) > + qmapoffset = 0; > + else > + qmapoffset = q->bands / > sch->dev->egress_subqueue_count; > + > + queue = 0; > + for (i = 0; i < q->bands; i++) { > + q->band2queue[i] = queue; > + if ( (((i + 1) - queue) == qmapoffset) && > + (queue < (sch->dev->egress_subqueue_count - > 1))) { > + queue++; > + } > + } > + } > + > +#endif > return 0; > } > > @@ -430,6 +521,9 @@ static struct Qdisc_ops prio_qdisc_ops = { > .destroy = prio_destroy, > .change = prio_tune, > .dump = prio_dump, > +#ifdef CONFIG_NET_MULTI_QUEUE_DEVICE > + .map_queue = prio_map_queue, > +#endif > .owner = THIS_MODULE, > }; > > > > > --- > Auke Kok > - > To unsubscribe from this list: send the line "unsubscribe > netdev" in the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >