Message ID | 1426079690-28271-2-git-send-email-ogerlitz@mellanox.com |
---|---|
State | Changes Requested, archived |
Delegated to: | David Miller |
Headers | show |
On 03/11/2015 06:14 AM, Or Gerlitz wrote: > From: John Fastabend <john.r.fastabend@intel.com> > > This adds a tx_maxrate attribute to the tx queue sysfs entry allowing > for max-rate limiting. Along with DCB-ETS and BQL this provides another > knob to tune queue performance. The limit units are Mbps. > > By default it is disabled. To disable the rate limitation after it > has been set for a queue, it should be set to zero. > > Signed-off-by: John Fastabend <john.r.fastabend@intel.com> > Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> > --- > include/linux/netdevice.h | 9 ++++++ > net/core/net-sysfs.c | 70 +++++++++++++++++++++++++++++++++++++-------- > 2 files changed, 67 insertions(+), 12 deletions(-) hmm I'm not sure sysfs is entirely the correct interface although it is convenient. Did you consider porting to netlink? To be honest I can't remember why I didn't do it netlink to start with. I think at the time I was more interested in the controller, feedback loop aspect. Basically I was dynamically managing the rate limiters across NICs from an agent to try and manage some types of congestion. .John -- 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
On Wed, Mar 11, 2015, John Fastabend <john.r.fastabend@intel.com> wrote: > On 03/11/2015 06:14 AM, Or Gerlitz wrote: >> This adds a tx_maxrate attribute to the tx queue sysfs entry allowing >> for max-rate limiting. Along with DCB-ETS and BQL this provides another >> knob to tune queue performance. The limit units are Mbps. >> >> By default it is disabled. To disable the rate limitation after it >> has been set for a queue, it should be set to zero. >> include/linux/netdevice.h | 9 ++++++ >> net/core/net-sysfs.c | 70 +++++++++++++++++++++++++++++++++++++-------- > hmm I'm not sure sysfs is entirely the correct interface although it is convenient. Very convenient indeed, and we use it for bunch of other tunables (xps, affinities and such). So I would vote for keeping it. > Did you consider porting to netlink? To be honest I can't remember why > I didn't do it netlink to start with. I think at the time > I was more interested in the controller, feedback loop aspect. Basically > I was dynamically managing the rate limiters across NICs from an agent to > try and manage some types of congestion. Sure, but we can add later netlink interface on top of the basic patch that defines the ndo. I assume you refer to rtnetlink and the changelink callback? can a driver instantiate it w.o having newlink/dellink bindings? -- 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
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index 1354ae8..e727677 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -587,6 +587,7 @@ struct netdev_queue { #ifdef CONFIG_BQL struct dql dql; #endif + unsigned long tx_maxrate; } ____cacheline_aligned_in_smp; static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) @@ -1025,6 +1026,11 @@ struct fib_info; * the set of features that the stack has calculated and it returns * those the driver believes to be appropriate. * + * int (*ndo_set_tx_maxrate)(struct net_device *dev, + int queue_index, + u32 maxrate); + * Called when a user wants to set a max-rate limitation of specific + * TX queue. * int (*ndo_switch_parent_id_get)(struct net_device *dev, * struct netdev_phys_item_id *psid); * Called to get an ID of the switch chip this port is part of. @@ -1197,6 +1203,9 @@ struct net_device_ops { netdev_features_t (*ndo_features_check) (struct sk_buff *skb, struct net_device *dev, netdev_features_t features); + int (*ndo_set_tx_maxrate)(struct net_device *dev, + int queue_index, + u32 maxrate); #ifdef CONFIG_NET_SWITCHDEV int (*ndo_switch_parent_id_get)(struct net_device *dev, struct netdev_phys_item_id *psid); diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c index cf30620..7822dda 100644 --- a/net/core/net-sysfs.c +++ b/net/core/net-sysfs.c @@ -951,6 +951,63 @@ static ssize_t show_trans_timeout(struct netdev_queue *queue, return sprintf(buf, "%lu", trans_timeout); } +#ifdef CONFIG_XPS +static inline unsigned int get_netdev_queue_index(struct netdev_queue *queue) +{ + struct net_device *dev = queue->dev; + int i; + + for (i = 0; i < dev->num_tx_queues; i++) + if (queue == &dev->_tx[i]) + break; + + BUG_ON(i >= dev->num_tx_queues); + + return i; +} + +static ssize_t show_tx_maxrate(struct netdev_queue *queue, + struct netdev_queue_attribute *attribute, + char *buf) +{ + return sprintf(buf, "%lu\n", queue->tx_maxrate); +} + +static ssize_t set_tx_maxrate(struct netdev_queue *queue, + struct netdev_queue_attribute *attribute, + const char *buf, size_t len) +{ + struct net_device *dev = queue->dev; + int err, index = get_netdev_queue_index(queue); + u32 rate = 0; + + err = kstrtou32(buf, 10, &rate); + if (err < 0) + return err; + + if (!rtnl_trylock()) + return restart_syscall(); + + if (dev->netdev_ops->ndo_set_tx_maxrate) { + err = dev->netdev_ops->ndo_set_tx_maxrate(dev, index, rate); + } else { + rtnl_unlock(); + return -EOPNOTSUPP; + } + + rtnl_unlock(); + if (!err) { + queue->tx_maxrate = rate; + return len; + } + return err; +} + +static struct netdev_queue_attribute queue_tx_maxrate = + __ATTR(tx_maxrate, S_IRUGO | S_IWUSR, + show_tx_maxrate, set_tx_maxrate); +#endif + static struct netdev_queue_attribute queue_trans_timeout = __ATTR(tx_timeout, S_IRUGO, show_trans_timeout, NULL); @@ -1065,18 +1122,6 @@ static struct attribute_group dql_group = { #endif /* CONFIG_BQL */ #ifdef CONFIG_XPS -static unsigned int get_netdev_queue_index(struct netdev_queue *queue) -{ - struct net_device *dev = queue->dev; - unsigned int i; - - i = queue - dev->_tx; - BUG_ON(i >= dev->num_tx_queues); - - return i; -} - - static ssize_t show_xps_map(struct netdev_queue *queue, struct netdev_queue_attribute *attribute, char *buf) { @@ -1153,6 +1198,7 @@ static struct attribute *netdev_queue_default_attrs[] = { &queue_trans_timeout.attr, #ifdef CONFIG_XPS &xps_cpus_attribute.attr, + &queue_tx_maxrate.attr, #endif NULL };