dect
/
linux-2.6
Archived
13
0
Fork 0
This repository has been archived on 2022-02-17. You can view files and clone it, but cannot push or open issues or pull requests.
linux-2.6/net/core/net-sysfs.c

819 lines
20 KiB
C
Raw Normal View History

/*
* net-sysfs.c - network device class and attributes
*
* Copyright (c) 2003 Stephen Hemminger <shemminger@osdl.org>
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#include <linux/capability.h>
#include <linux/kernel.h>
#include <linux/netdevice.h>
#include <linux/if_arp.h>
#include <net/sock.h>
#include <linux/rtnetlink.h>
#include <linux/wireless.h>
#include <net/wext.h>
#include "net-sysfs.h"
#ifdef CONFIG_SYSFS
static const char fmt_hex[] = "%#x\n";
static const char fmt_long_hex[] = "%#lx\n";
static const char fmt_dec[] = "%d\n";
static const char fmt_ulong[] = "%lu\n";
static inline int dev_isalive(const struct net_device *dev)
{
return dev->reg_state <= NETREG_REGISTERED;
}
/* use same locking rules as GIF* ioctl's */
static ssize_t netdev_show(const struct device *dev,
struct device_attribute *attr, char *buf,
ssize_t (*format)(const struct net_device *, char *))
{
struct net_device *net = to_net_dev(dev);
ssize_t ret = -EINVAL;
read_lock(&dev_base_lock);
if (dev_isalive(net))
ret = (*format)(net, buf);
read_unlock(&dev_base_lock);
return ret;
}
/* generate a show function for simple field */
#define NETDEVICE_SHOW(field, format_string) \
static ssize_t format_##field(const struct net_device *net, char *buf) \
{ \
return sprintf(buf, format_string, net->field); \
} \
static ssize_t show_##field(struct device *dev, \
struct device_attribute *attr, char *buf) \
{ \
return netdev_show(dev, attr, buf, format_##field); \
}
/* use same locking and permission rules as SIF* ioctl's */
static ssize_t netdev_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t len,
int (*set)(struct net_device *, unsigned long))
{
struct net_device *net = to_net_dev(dev);
char *endp;
unsigned long new;
int ret = -EINVAL;
if (!capable(CAP_NET_ADMIN))
return -EPERM;
new = simple_strtoul(buf, &endp, 0);
if (endp == buf)
goto err;
if (!rtnl_trylock())
return restart_syscall();
if (dev_isalive(net)) {
if ((ret = (*set)(net, new)) == 0)
ret = len;
}
rtnl_unlock();
err:
return ret;
}
NETDEVICE_SHOW(dev_id, fmt_hex);
NETDEVICE_SHOW(addr_len, fmt_dec);
NETDEVICE_SHOW(iflink, fmt_dec);
NETDEVICE_SHOW(ifindex, fmt_dec);
NETDEVICE_SHOW(features, fmt_long_hex);
NETDEVICE_SHOW(type, fmt_dec);
NETDEVICE_SHOW(link_mode, fmt_dec);
/* use same locking rules as GIFHWADDR ioctl's */
static ssize_t show_address(struct device *dev, struct device_attribute *attr,
char *buf)
{
struct net_device *net = to_net_dev(dev);
ssize_t ret = -EINVAL;
read_lock(&dev_base_lock);
if (dev_isalive(net))
ret = sysfs_format_mac(buf, net->dev_addr, net->addr_len);
read_unlock(&dev_base_lock);
return ret;
}
static ssize_t show_broadcast(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct net_device *net = to_net_dev(dev);
if (dev_isalive(net))
return sysfs_format_mac(buf, net->broadcast, net->addr_len);
return -EINVAL;
}
static ssize_t show_carrier(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct net_device *netdev = to_net_dev(dev);
if (netif_running(netdev)) {
return sprintf(buf, fmt_dec, !!netif_carrier_ok(netdev));
}
return -EINVAL;
}
static ssize_t show_speed(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct net_device *netdev = to_net_dev(dev);
int ret = -EINVAL;
if (!rtnl_trylock())
return restart_syscall();
if (netif_running(netdev) &&
netdev->ethtool_ops &&
netdev->ethtool_ops->get_settings) {
struct ethtool_cmd cmd = { ETHTOOL_GSET };
if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
ret = sprintf(buf, fmt_dec, ethtool_cmd_speed(&cmd));
}
rtnl_unlock();
return ret;
}
static ssize_t show_duplex(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct net_device *netdev = to_net_dev(dev);
int ret = -EINVAL;
if (!rtnl_trylock())
return restart_syscall();
if (netif_running(netdev) &&
netdev->ethtool_ops &&
netdev->ethtool_ops->get_settings) {
struct ethtool_cmd cmd = { ETHTOOL_GSET };
if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
ret = sprintf(buf, "%s\n", cmd.duplex ? "full" : "half");
}
rtnl_unlock();
return ret;
}
static ssize_t show_dormant(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct net_device *netdev = to_net_dev(dev);
if (netif_running(netdev))
return sprintf(buf, fmt_dec, !!netif_dormant(netdev));
return -EINVAL;
}
static const char *const operstates[] = {
"unknown",
"notpresent", /* currently unused */
"down",
"lowerlayerdown",
"testing", /* currently unused */
"dormant",
"up"
};
static ssize_t show_operstate(struct device *dev,
struct device_attribute *attr, char *buf)
{
const struct net_device *netdev = to_net_dev(dev);
unsigned char operstate;
read_lock(&dev_base_lock);
operstate = netdev->operstate;
if (!netif_running(netdev))
operstate = IF_OPER_DOWN;
read_unlock(&dev_base_lock);
if (operstate >= ARRAY_SIZE(operstates))
return -EINVAL; /* should not happen */
return sprintf(buf, "%s\n", operstates[operstate]);
}
/* read-write attributes */
NETDEVICE_SHOW(mtu, fmt_dec);
static int change_mtu(struct net_device *net, unsigned long new_mtu)
{
return dev_set_mtu(net, (int) new_mtu);
}
static ssize_t store_mtu(struct device *dev, struct device_attribute *attr,
const char *buf, size_t len)
{
return netdev_store(dev, attr, buf, len, change_mtu);
}
NETDEVICE_SHOW(flags, fmt_hex);
static int change_flags(struct net_device *net, unsigned long new_flags)
{
return dev_change_flags(net, (unsigned) new_flags);
}
static ssize_t store_flags(struct device *dev, struct device_attribute *attr,
const char *buf, size_t len)
{
return netdev_store(dev, attr, buf, len, change_flags);
}
NETDEVICE_SHOW(tx_queue_len, fmt_ulong);
static int change_tx_queue_len(struct net_device *net, unsigned long new_len)
{
net->tx_queue_len = new_len;
return 0;
}
static ssize_t store_tx_queue_len(struct device *dev,
struct device_attribute *attr,
const char *buf, size_t len)
{
return netdev_store(dev, attr, buf, len, change_tx_queue_len);
}
static ssize_t store_ifalias(struct device *dev, struct device_attribute *attr,
const char *buf, size_t len)
{
struct net_device *netdev = to_net_dev(dev);
size_t count = len;
ssize_t ret;
if (!capable(CAP_NET_ADMIN))
return -EPERM;
/* ignore trailing newline */
if (len > 0 && buf[len - 1] == '\n')
--count;
if (!rtnl_trylock())
return restart_syscall();
ret = dev_set_alias(netdev, buf, count);
rtnl_unlock();
return ret < 0 ? ret : len;
}
static ssize_t show_ifalias(struct device *dev,
struct device_attribute *attr, char *buf)
{
const struct net_device *netdev = to_net_dev(dev);
ssize_t ret = 0;
if (!rtnl_trylock())
return restart_syscall();
if (netdev->ifalias)
ret = sprintf(buf, "%s\n", netdev->ifalias);
rtnl_unlock();
return ret;
}
static struct device_attribute net_class_attributes[] = {
__ATTR(addr_len, S_IRUGO, show_addr_len, NULL),
__ATTR(dev_id, S_IRUGO, show_dev_id, NULL),
__ATTR(ifalias, S_IRUGO | S_IWUSR, show_ifalias, store_ifalias),
__ATTR(iflink, S_IRUGO, show_iflink, NULL),
__ATTR(ifindex, S_IRUGO, show_ifindex, NULL),
__ATTR(features, S_IRUGO, show_features, NULL),
__ATTR(type, S_IRUGO, show_type, NULL),
__ATTR(link_mode, S_IRUGO, show_link_mode, NULL),
__ATTR(address, S_IRUGO, show_address, NULL),
__ATTR(broadcast, S_IRUGO, show_broadcast, NULL),
__ATTR(carrier, S_IRUGO, show_carrier, NULL),
__ATTR(speed, S_IRUGO, show_speed, NULL),
__ATTR(duplex, S_IRUGO, show_duplex, NULL),
__ATTR(dormant, S_IRUGO, show_dormant, NULL),
__ATTR(operstate, S_IRUGO, show_operstate, NULL),
__ATTR(mtu, S_IRUGO | S_IWUSR, show_mtu, store_mtu),
__ATTR(flags, S_IRUGO | S_IWUSR, show_flags, store_flags),
__ATTR(tx_queue_len, S_IRUGO | S_IWUSR, show_tx_queue_len,
store_tx_queue_len),
{}
};
/* Show a given an attribute in the statistics group */
static ssize_t netstat_show(const struct device *d,
struct device_attribute *attr, char *buf,
unsigned long offset)
{
struct net_device *dev = to_net_dev(d);
ssize_t ret = -EINVAL;
WARN_ON(offset > sizeof(struct net_device_stats) ||
offset % sizeof(unsigned long) != 0);
read_lock(&dev_base_lock);
if (dev_isalive(dev)) {
const struct net_device_stats *stats = dev_get_stats(dev);
ret = sprintf(buf, fmt_ulong,
*(unsigned long *)(((u8 *) stats) + offset));
}
read_unlock(&dev_base_lock);
return ret;
}
/* generate a read-only statistics attribute */
#define NETSTAT_ENTRY(name) \
static ssize_t show_##name(struct device *d, \
struct device_attribute *attr, char *buf) \
{ \
return netstat_show(d, attr, buf, \
offsetof(struct net_device_stats, name)); \
} \
static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL)
NETSTAT_ENTRY(rx_packets);
NETSTAT_ENTRY(tx_packets);
NETSTAT_ENTRY(rx_bytes);
NETSTAT_ENTRY(tx_bytes);
NETSTAT_ENTRY(rx_errors);
NETSTAT_ENTRY(tx_errors);
NETSTAT_ENTRY(rx_dropped);
NETSTAT_ENTRY(tx_dropped);
NETSTAT_ENTRY(multicast);
NETSTAT_ENTRY(collisions);
NETSTAT_ENTRY(rx_length_errors);
NETSTAT_ENTRY(rx_over_errors);
NETSTAT_ENTRY(rx_crc_errors);
NETSTAT_ENTRY(rx_frame_errors);
NETSTAT_ENTRY(rx_fifo_errors);
NETSTAT_ENTRY(rx_missed_errors);
NETSTAT_ENTRY(tx_aborted_errors);
NETSTAT_ENTRY(tx_carrier_errors);
NETSTAT_ENTRY(tx_fifo_errors);
NETSTAT_ENTRY(tx_heartbeat_errors);
NETSTAT_ENTRY(tx_window_errors);
NETSTAT_ENTRY(rx_compressed);
NETSTAT_ENTRY(tx_compressed);
static struct attribute *netstat_attrs[] = {
&dev_attr_rx_packets.attr,
&dev_attr_tx_packets.attr,
&dev_attr_rx_bytes.attr,
&dev_attr_tx_bytes.attr,
&dev_attr_rx_errors.attr,
&dev_attr_tx_errors.attr,
&dev_attr_rx_dropped.attr,
&dev_attr_tx_dropped.attr,
&dev_attr_multicast.attr,
&dev_attr_collisions.attr,
&dev_attr_rx_length_errors.attr,
&dev_attr_rx_over_errors.attr,
&dev_attr_rx_crc_errors.attr,
&dev_attr_rx_frame_errors.attr,
&dev_attr_rx_fifo_errors.attr,
&dev_attr_rx_missed_errors.attr,
&dev_attr_tx_aborted_errors.attr,
&dev_attr_tx_carrier_errors.attr,
&dev_attr_tx_fifo_errors.attr,
&dev_attr_tx_heartbeat_errors.attr,
&dev_attr_tx_window_errors.attr,
&dev_attr_rx_compressed.attr,
&dev_attr_tx_compressed.attr,
NULL
};
static struct attribute_group netstat_group = {
.name = "statistics",
.attrs = netstat_attrs,
};
#ifdef CONFIG_WIRELESS_EXT_SYSFS
/* helper function that does all the locking etc for wireless stats */
static ssize_t wireless_show(struct device *d, char *buf,
ssize_t (*format)(const struct iw_statistics *,
char *))
{
struct net_device *dev = to_net_dev(d);
const struct iw_statistics *iw;
ssize_t ret = -EINVAL;
if (!rtnl_trylock())
return restart_syscall();
if (dev_isalive(dev)) {
iw = get_wireless_stats(dev);
if (iw)
ret = (*format)(iw, buf);
}
rtnl_unlock();
return ret;
}
/* show function template for wireless fields */
#define WIRELESS_SHOW(name, field, format_string) \
static ssize_t format_iw_##name(const struct iw_statistics *iw, char *buf) \
{ \
return sprintf(buf, format_string, iw->field); \
} \
static ssize_t show_iw_##name(struct device *d, \
struct device_attribute *attr, char *buf) \
{ \
return wireless_show(d, buf, format_iw_##name); \
} \
static DEVICE_ATTR(name, S_IRUGO, show_iw_##name, NULL)
WIRELESS_SHOW(status, status, fmt_hex);
WIRELESS_SHOW(link, qual.qual, fmt_dec);
WIRELESS_SHOW(level, qual.level, fmt_dec);
WIRELESS_SHOW(noise, qual.noise, fmt_dec);
WIRELESS_SHOW(nwid, discard.nwid, fmt_dec);
WIRELESS_SHOW(crypt, discard.code, fmt_dec);
WIRELESS_SHOW(fragment, discard.fragment, fmt_dec);
WIRELESS_SHOW(misc, discard.misc, fmt_dec);
WIRELESS_SHOW(retries, discard.retries, fmt_dec);
WIRELESS_SHOW(beacon, miss.beacon, fmt_dec);
static struct attribute *wireless_attrs[] = {
&dev_attr_status.attr,
&dev_attr_link.attr,
&dev_attr_level.attr,
&dev_attr_noise.attr,
&dev_attr_nwid.attr,
&dev_attr_crypt.attr,
&dev_attr_fragment.attr,
&dev_attr_retries.attr,
&dev_attr_misc.attr,
&dev_attr_beacon.attr,
NULL
};
static struct attribute_group wireless_group = {
.name = "wireless",
.attrs = wireless_attrs,
};
#endif
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
/*
* RX queue sysfs structures and functions.
*/
struct rx_queue_attribute {
struct attribute attr;
ssize_t (*show)(struct netdev_rx_queue *queue,
struct rx_queue_attribute *attr, char *buf);
ssize_t (*store)(struct netdev_rx_queue *queue,
struct rx_queue_attribute *attr, const char *buf, size_t len);
};
#define to_rx_queue_attr(_attr) container_of(_attr, \
struct rx_queue_attribute, attr)
#define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj)
static ssize_t rx_queue_attr_show(struct kobject *kobj, struct attribute *attr,
char *buf)
{
struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
struct netdev_rx_queue *queue = to_rx_queue(kobj);
if (!attribute->show)
return -EIO;
return attribute->show(queue, attribute, buf);
}
static ssize_t rx_queue_attr_store(struct kobject *kobj, struct attribute *attr,
const char *buf, size_t count)
{
struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
struct netdev_rx_queue *queue = to_rx_queue(kobj);
if (!attribute->store)
return -EIO;
return attribute->store(queue, attribute, buf, count);
}
static struct sysfs_ops rx_queue_sysfs_ops = {
.show = rx_queue_attr_show,
.store = rx_queue_attr_store,
};
static ssize_t show_rps_map(struct netdev_rx_queue *queue,
struct rx_queue_attribute *attribute, char *buf)
{
struct rps_map *map;
cpumask_var_t mask;
size_t len = 0;
int i;
if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
return -ENOMEM;
rcu_read_lock();
map = rcu_dereference(queue->rps_map);
if (map)
for (i = 0; i < map->len; i++)
cpumask_set_cpu(map->cpus[i], mask);
len += cpumask_scnprintf(buf + len, PAGE_SIZE, mask);
if (PAGE_SIZE - len < 3) {
rcu_read_unlock();
free_cpumask_var(mask);
return -EINVAL;
}
rcu_read_unlock();
free_cpumask_var(mask);
len += sprintf(buf + len, "\n");
return len;
}
static void rps_map_release(struct rcu_head *rcu)
{
struct rps_map *map = container_of(rcu, struct rps_map, rcu);
kfree(map);
}
ssize_t store_rps_map(struct netdev_rx_queue *queue,
struct rx_queue_attribute *attribute,
const char *buf, size_t len)
{
struct rps_map *old_map, *map;
cpumask_var_t mask;
int err, cpu, i;
static DEFINE_SPINLOCK(rps_map_lock);
if (!capable(CAP_NET_ADMIN))
return -EPERM;
if (!alloc_cpumask_var(&mask, GFP_KERNEL))
return -ENOMEM;
err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
if (err) {
free_cpumask_var(mask);
return err;
}
map = kzalloc(max_t(unsigned,
RPS_MAP_SIZE(cpumask_weight(mask)), L1_CACHE_BYTES),
GFP_KERNEL);
if (!map) {
free_cpumask_var(mask);
return -ENOMEM;
}
i = 0;
for_each_cpu_and(cpu, mask, cpu_online_mask)
map->cpus[i++] = cpu;
if (i)
map->len = i;
else {
kfree(map);
map = NULL;
}
spin_lock(&rps_map_lock);
old_map = queue->rps_map;
rcu_assign_pointer(queue->rps_map, map);
spin_unlock(&rps_map_lock);
if (old_map)
call_rcu(&old_map->rcu, rps_map_release);
free_cpumask_var(mask);
return len;
}
static struct rx_queue_attribute rps_cpus_attribute =
__ATTR(rps_cpus, S_IRUGO | S_IWUSR, show_rps_map, store_rps_map);
static struct attribute *rx_queue_default_attrs[] = {
&rps_cpus_attribute.attr,
NULL
};
static void rx_queue_release(struct kobject *kobj)
{
struct netdev_rx_queue *queue = to_rx_queue(kobj);
struct rps_map *map = queue->rps_map;
struct netdev_rx_queue *first = queue->first;
if (map)
call_rcu(&map->rcu, rps_map_release);
if (atomic_dec_and_test(&first->count))
kfree(first);
}
static struct kobj_type rx_queue_ktype = {
.sysfs_ops = &rx_queue_sysfs_ops,
.release = rx_queue_release,
.default_attrs = rx_queue_default_attrs,
};
static int rx_queue_add_kobject(struct net_device *net, int index)
{
struct netdev_rx_queue *queue = net->_rx + index;
struct kobject *kobj = &queue->kobj;
int error = 0;
kobj->kset = net->queues_kset;
error = kobject_init_and_add(kobj, &rx_queue_ktype, NULL,
"rx-%u", index);
if (error) {
kobject_put(kobj);
return error;
}
kobject_uevent(kobj, KOBJ_ADD);
return error;
}
static int rx_queue_register_kobjects(struct net_device *net)
{
int i;
int error = 0;
net->queues_kset = kset_create_and_add("queues",
NULL, &net->dev.kobj);
if (!net->queues_kset)
return -ENOMEM;
for (i = 0; i < net->num_rx_queues; i++) {
error = rx_queue_add_kobject(net, i);
if (error)
break;
}
if (error)
while (--i >= 0)
kobject_put(&net->_rx[i].kobj);
return error;
}
static void rx_queue_remove_kobjects(struct net_device *net)
{
int i;
for (i = 0; i < net->num_rx_queues; i++)
kobject_put(&net->_rx[i].kobj);
kset_unregister(net->queues_kset);
}
#endif /* CONFIG_SYSFS */
#ifdef CONFIG_HOTPLUG
static int netdev_uevent(struct device *d, struct kobj_uevent_env *env)
{
struct net_device *dev = to_net_dev(d);
int retval;
if (!net_eq(dev_net(dev), &init_net))
return 0;
/* pass interface to uevent. */
retval = add_uevent_var(env, "INTERFACE=%s", dev->name);
if (retval)
goto exit;
/* pass ifindex to uevent.
* ifindex is useful as it won't change (interface name may change)
* and is what RtNetlink uses natively. */
retval = add_uevent_var(env, "IFINDEX=%d", dev->ifindex);
exit:
return retval;
}
#endif
/*
* netdev_release -- destroy and free a dead device.
* Called when last reference to device kobject is gone.
*/
static void netdev_release(struct device *d)
{
struct net_device *dev = to_net_dev(d);
BUG_ON(dev->reg_state != NETREG_RELEASED);
kfree(dev->ifalias);
kfree((char *)dev - dev->padded);
}
static struct class net_class = {
.name = "net",
.dev_release = netdev_release,
#ifdef CONFIG_SYSFS
.dev_attrs = net_class_attributes,
#endif /* CONFIG_SYSFS */
#ifdef CONFIG_HOTPLUG
.dev_uevent = netdev_uevent,
#endif
};
/* Delete sysfs entries but hold kobject reference until after all
* netdev references are gone.
*/
void netdev_unregister_kobject(struct net_device * net)
{
struct device *dev = &(net->dev);
kobject_get(&dev->kobj);
if (!net_eq(dev_net(net), &init_net))
return;
#ifdef CONFIG_SYSFS
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
rx_queue_remove_kobjects(net);
#endif
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
device_del(dev);
}
/* Create sysfs entries for network device. */
int netdev_register_kobject(struct net_device *net)
{
struct device *dev = &(net->dev);
const struct attribute_group **groups = net->sysfs_groups;
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
int error = 0;
dev->class = &net_class;
dev->platform_data = net;
dev->groups = groups;
dev_set_name(dev, "%s", net->name);
#ifdef CONFIG_SYSFS
/* Allow for a device specific group */
if (*groups)
groups++;
*groups++ = &netstat_group;
#ifdef CONFIG_WIRELESS_EXT_SYSFS
if (net->ieee80211_ptr)
*groups++ = &wireless_group;
#ifdef CONFIG_WIRELESS_EXT
else if (net->wireless_handlers)
*groups++ = &wireless_group;
#endif
#endif
#endif /* CONFIG_SYSFS */
if (!net_eq(dev_net(net), &init_net))
return 0;
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
error = device_add(dev);
if (error)
return error;
#ifdef CONFIG_SYSFS
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
error = rx_queue_register_kobjects(net);
if (error) {
device_del(dev);
return error;
}
#endif
rps: Receive Packet Steering This patch implements software receive side packet steering (RPS). RPS distributes the load of received packet processing across multiple CPUs. Problem statement: Protocol processing done in the NAPI context for received packets is serialized per device queue and becomes a bottleneck under high packet load. This substantially limits pps that can be achieved on a single queue NIC and provides no scaling with multiple cores. This solution queues packets early on in the receive path on the backlog queues of other CPUs. This allows protocol processing (e.g. IP and TCP) to be performed on packets in parallel. For each device (or each receive queue in a multi-queue device) a mask of CPUs is set to indicate the CPUs that can process packets. A CPU is selected on a per packet basis by hashing contents of the packet header (e.g. the TCP or UDP 4-tuple) and using the result to index into the CPU mask. The IPI mechanism is used to raise networking receive softirqs between CPUs. This effectively emulates in software what a multi-queue NIC can provide, but is generic requiring no device support. Many devices now provide a hash over the 4-tuple on a per packet basis (e.g. the Toeplitz hash). This patch allow drivers to set the HW reported hash in an skb field, and that value in turn is used to index into the RPS maps. Using the HW generated hash can avoid cache misses on the packet when steering it to a remote CPU. The CPU mask is set on a per device and per queue basis in the sysfs variable /sys/class/net/<device>/queues/rx-<n>/rps_cpus. This is a set of canonical bit maps for receive queues in the device (numbered by <n>). If a device does not support multi-queue, a single variable is used for the device (rx-0). Generally, we have found this technique increases pps capabilities of a single queue device with good CPU utilization. Optimal settings for the CPU mask seem to depend on architectures and cache hierarcy. Below are some results running 500 instances of netperf TCP_RR test with 1 byte req. and resp. Results show cumulative transaction rate and system CPU utilization. e1000e on 8 core Intel Without RPS: 108K tps at 33% CPU With RPS: 311K tps at 64% CPU forcedeth on 16 core AMD Without RPS: 156K tps at 15% CPU With RPS: 404K tps at 49% CPU bnx2x on 16 core AMD Without RPS 567K tps at 61% CPU (4 HW RX queues) Without RPS 738K tps at 96% CPU (8 HW RX queues) With RPS: 854K tps at 76% CPU (4 HW RX queues) Caveats: - The benefits of this patch are dependent on architecture and cache hierarchy. Tuning the masks to get best performance is probably necessary. - This patch adds overhead in the path for processing a single packet. In a lightly loaded server this overhead may eliminate the advantages of increased parallelism, and possibly cause some relative performance degradation. We have found that masks that are cache aware (share same caches with the interrupting CPU) mitigate much of this. - The RPS masks can be changed dynamically, however whenever the mask is changed this introduces the possibility of generating out of order packets. It's probably best not change the masks too frequently. Signed-off-by: Tom Herbert <therbert@google.com> include/linux/netdevice.h | 32 ++++- include/linux/skbuff.h | 3 + net/core/dev.c | 335 +++++++++++++++++++++++++++++++++++++-------- net/core/net-sysfs.c | 225 ++++++++++++++++++++++++++++++- net/core/skbuff.c | 2 + 5 files changed, 538 insertions(+), 59 deletions(-) Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-03-16 08:03:29 +00:00
return error;
}
int netdev_class_create_file(struct class_attribute *class_attr)
{
return class_create_file(&net_class, class_attr);
}
void netdev_class_remove_file(struct class_attribute *class_attr)
{
class_remove_file(&net_class, class_attr);
}
EXPORT_SYMBOL(netdev_class_create_file);
EXPORT_SYMBOL(netdev_class_remove_file);
void netdev_initialize_kobject(struct net_device *net)
{
struct device *device = &(net->dev);
device_initialize(device);
}
int netdev_kobject_init(void)
{
return class_register(&net_class);
}