2 * net-sysfs.c - network device class and attributes
4 * Copyright (c) 2003 Stephen Hemminger <shemminger@osdl.org>
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
12 #include <linux/capability.h>
13 #include <linux/kernel.h>
14 #include <linux/netdevice.h>
15 #include <net/switchdev.h>
16 #include <linux/if_arp.h>
17 #include <linux/slab.h>
18 #include <linux/nsproxy.h>
20 #include <net/net_namespace.h>
21 #include <linux/rtnetlink.h>
22 #include <linux/vmalloc.h>
23 #include <linux/export.h>
24 #include <linux/jiffies.h>
25 #include <linux/pm_runtime.h>
27 #include <linux/of_net.h>
29 #include "net-sysfs.h"
32 static const char fmt_hex[] = "%#x\n";
33 static const char fmt_dec[] = "%d\n";
34 static const char fmt_ulong[] = "%lu\n";
35 static const char fmt_u64[] = "%llu\n";
37 static inline int dev_isalive(const struct net_device *dev)
39 return dev->reg_state <= NETREG_REGISTERED;
42 /* use same locking rules as GIF* ioctl's */
43 static ssize_t netdev_show(const struct device *dev,
44 struct device_attribute *attr, char *buf,
45 ssize_t (*format)(const struct net_device *, char *))
47 struct net_device *ndev = to_net_dev(dev);
48 ssize_t ret = -EINVAL;
50 read_lock(&dev_base_lock);
51 if (dev_isalive(ndev))
52 ret = (*format)(ndev, buf);
53 read_unlock(&dev_base_lock);
58 /* generate a show function for simple field */
59 #define NETDEVICE_SHOW(field, format_string) \
60 static ssize_t format_##field(const struct net_device *dev, char *buf) \
62 return sprintf(buf, format_string, dev->field); \
64 static ssize_t field##_show(struct device *dev, \
65 struct device_attribute *attr, char *buf) \
67 return netdev_show(dev, attr, buf, format_##field); \
70 #define NETDEVICE_SHOW_RO(field, format_string) \
71 NETDEVICE_SHOW(field, format_string); \
72 static DEVICE_ATTR_RO(field)
74 #define NETDEVICE_SHOW_RW(field, format_string) \
75 NETDEVICE_SHOW(field, format_string); \
76 static DEVICE_ATTR_RW(field)
78 /* use same locking and permission rules as SIF* ioctl's */
79 static ssize_t netdev_store(struct device *dev, struct device_attribute *attr,
80 const char *buf, size_t len,
81 int (*set)(struct net_device *, unsigned long))
83 struct net_device *netdev = to_net_dev(dev);
84 struct net *net = dev_net(netdev);
88 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
91 ret = kstrtoul(buf, 0, &new);
96 return restart_syscall();
98 if (dev_isalive(netdev)) {
99 if ((ret = (*set)(netdev, new)) == 0)
107 NETDEVICE_SHOW_RO(dev_id, fmt_hex);
108 NETDEVICE_SHOW_RO(dev_port, fmt_dec);
109 NETDEVICE_SHOW_RO(addr_assign_type, fmt_dec);
110 NETDEVICE_SHOW_RO(addr_len, fmt_dec);
111 NETDEVICE_SHOW_RO(ifindex, fmt_dec);
112 NETDEVICE_SHOW_RO(type, fmt_dec);
113 NETDEVICE_SHOW_RO(link_mode, fmt_dec);
115 static ssize_t iflink_show(struct device *dev, struct device_attribute *attr,
118 struct net_device *ndev = to_net_dev(dev);
120 return sprintf(buf, fmt_dec, dev_get_iflink(ndev));
122 static DEVICE_ATTR_RO(iflink);
124 static ssize_t format_name_assign_type(const struct net_device *dev, char *buf)
126 return sprintf(buf, fmt_dec, dev->name_assign_type);
129 static ssize_t name_assign_type_show(struct device *dev,
130 struct device_attribute *attr,
133 struct net_device *ndev = to_net_dev(dev);
134 ssize_t ret = -EINVAL;
136 if (ndev->name_assign_type != NET_NAME_UNKNOWN)
137 ret = netdev_show(dev, attr, buf, format_name_assign_type);
141 static DEVICE_ATTR_RO(name_assign_type);
143 /* use same locking rules as GIFHWADDR ioctl's */
144 static ssize_t address_show(struct device *dev, struct device_attribute *attr,
147 struct net_device *ndev = to_net_dev(dev);
148 ssize_t ret = -EINVAL;
150 read_lock(&dev_base_lock);
151 if (dev_isalive(ndev))
152 ret = sysfs_format_mac(buf, ndev->dev_addr, ndev->addr_len);
153 read_unlock(&dev_base_lock);
156 static DEVICE_ATTR_RO(address);
158 static ssize_t broadcast_show(struct device *dev,
159 struct device_attribute *attr, char *buf)
161 struct net_device *ndev = to_net_dev(dev);
162 if (dev_isalive(ndev))
163 return sysfs_format_mac(buf, ndev->broadcast, ndev->addr_len);
166 static DEVICE_ATTR_RO(broadcast);
168 static int change_carrier(struct net_device *dev, unsigned long new_carrier)
170 if (!netif_running(dev))
172 return dev_change_carrier(dev, (bool) new_carrier);
175 static ssize_t carrier_store(struct device *dev, struct device_attribute *attr,
176 const char *buf, size_t len)
178 return netdev_store(dev, attr, buf, len, change_carrier);
181 static ssize_t carrier_show(struct device *dev,
182 struct device_attribute *attr, char *buf)
184 struct net_device *netdev = to_net_dev(dev);
185 if (netif_running(netdev)) {
186 return sprintf(buf, fmt_dec, !!netif_carrier_ok(netdev));
190 static DEVICE_ATTR_RW(carrier);
192 static ssize_t speed_show(struct device *dev,
193 struct device_attribute *attr, char *buf)
195 struct net_device *netdev = to_net_dev(dev);
199 return restart_syscall();
201 if (netif_running(netdev)) {
202 struct ethtool_link_ksettings cmd;
204 if (!__ethtool_get_link_ksettings(netdev, &cmd))
205 ret = sprintf(buf, fmt_dec, cmd.base.speed);
210 static DEVICE_ATTR_RO(speed);
212 static ssize_t duplex_show(struct device *dev,
213 struct device_attribute *attr, char *buf)
215 struct net_device *netdev = to_net_dev(dev);
219 return restart_syscall();
221 if (netif_running(netdev)) {
222 struct ethtool_link_ksettings cmd;
224 if (!__ethtool_get_link_ksettings(netdev, &cmd)) {
227 switch (cmd.base.duplex) {
238 ret = sprintf(buf, "%s\n", duplex);
244 static DEVICE_ATTR_RO(duplex);
246 static ssize_t dormant_show(struct device *dev,
247 struct device_attribute *attr, char *buf)
249 struct net_device *netdev = to_net_dev(dev);
251 if (netif_running(netdev))
252 return sprintf(buf, fmt_dec, !!netif_dormant(netdev));
256 static DEVICE_ATTR_RO(dormant);
258 static const char *const operstates[] = {
260 "notpresent", /* currently unused */
263 "testing", /* currently unused */
268 static ssize_t operstate_show(struct device *dev,
269 struct device_attribute *attr, char *buf)
271 const struct net_device *netdev = to_net_dev(dev);
272 unsigned char operstate;
274 read_lock(&dev_base_lock);
275 operstate = netdev->operstate;
276 if (!netif_running(netdev))
277 operstate = IF_OPER_DOWN;
278 read_unlock(&dev_base_lock);
280 if (operstate >= ARRAY_SIZE(operstates))
281 return -EINVAL; /* should not happen */
283 return sprintf(buf, "%s\n", operstates[operstate]);
285 static DEVICE_ATTR_RO(operstate);
287 static ssize_t carrier_changes_show(struct device *dev,
288 struct device_attribute *attr,
291 struct net_device *netdev = to_net_dev(dev);
292 return sprintf(buf, fmt_dec,
293 atomic_read(&netdev->carrier_changes));
295 static DEVICE_ATTR_RO(carrier_changes);
297 /* read-write attributes */
299 static int change_mtu(struct net_device *dev, unsigned long new_mtu)
301 return dev_set_mtu(dev, (int) new_mtu);
304 static ssize_t mtu_store(struct device *dev, struct device_attribute *attr,
305 const char *buf, size_t len)
307 return netdev_store(dev, attr, buf, len, change_mtu);
309 NETDEVICE_SHOW_RW(mtu, fmt_dec);
311 static int change_flags(struct net_device *dev, unsigned long new_flags)
313 return dev_change_flags(dev, (unsigned int) new_flags);
316 static ssize_t flags_store(struct device *dev, struct device_attribute *attr,
317 const char *buf, size_t len)
319 return netdev_store(dev, attr, buf, len, change_flags);
321 NETDEVICE_SHOW_RW(flags, fmt_hex);
323 static int change_tx_queue_len(struct net_device *dev, unsigned long new_len)
325 dev->tx_queue_len = new_len;
329 static ssize_t tx_queue_len_store(struct device *dev,
330 struct device_attribute *attr,
331 const char *buf, size_t len)
333 if (!capable(CAP_NET_ADMIN))
336 return netdev_store(dev, attr, buf, len, change_tx_queue_len);
338 NETDEVICE_SHOW_RW(tx_queue_len, fmt_ulong);
340 static int change_gro_flush_timeout(struct net_device *dev, unsigned long val)
342 dev->gro_flush_timeout = val;
346 static ssize_t gro_flush_timeout_store(struct device *dev,
347 struct device_attribute *attr,
348 const char *buf, size_t len)
350 if (!capable(CAP_NET_ADMIN))
353 return netdev_store(dev, attr, buf, len, change_gro_flush_timeout);
355 NETDEVICE_SHOW_RW(gro_flush_timeout, fmt_ulong);
357 static ssize_t ifalias_store(struct device *dev, struct device_attribute *attr,
358 const char *buf, size_t len)
360 struct net_device *netdev = to_net_dev(dev);
361 struct net *net = dev_net(netdev);
365 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
368 /* ignore trailing newline */
369 if (len > 0 && buf[len - 1] == '\n')
373 return restart_syscall();
374 ret = dev_set_alias(netdev, buf, count);
377 return ret < 0 ? ret : len;
380 static ssize_t ifalias_show(struct device *dev,
381 struct device_attribute *attr, char *buf)
383 const struct net_device *netdev = to_net_dev(dev);
387 return restart_syscall();
389 ret = sprintf(buf, "%s\n", netdev->ifalias);
393 static DEVICE_ATTR_RW(ifalias);
395 static int change_group(struct net_device *dev, unsigned long new_group)
397 dev_set_group(dev, (int) new_group);
401 static ssize_t group_store(struct device *dev, struct device_attribute *attr,
402 const char *buf, size_t len)
404 return netdev_store(dev, attr, buf, len, change_group);
406 NETDEVICE_SHOW(group, fmt_dec);
407 static DEVICE_ATTR(netdev_group, S_IRUGO | S_IWUSR, group_show, group_store);
409 static int change_proto_down(struct net_device *dev, unsigned long proto_down)
411 return dev_change_proto_down(dev, (bool) proto_down);
414 static ssize_t proto_down_store(struct device *dev,
415 struct device_attribute *attr,
416 const char *buf, size_t len)
418 return netdev_store(dev, attr, buf, len, change_proto_down);
420 NETDEVICE_SHOW_RW(proto_down, fmt_dec);
422 static ssize_t phys_port_id_show(struct device *dev,
423 struct device_attribute *attr, char *buf)
425 struct net_device *netdev = to_net_dev(dev);
426 ssize_t ret = -EINVAL;
429 return restart_syscall();
431 if (dev_isalive(netdev)) {
432 struct netdev_phys_item_id ppid;
434 ret = dev_get_phys_port_id(netdev, &ppid);
436 ret = sprintf(buf, "%*phN\n", ppid.id_len, ppid.id);
442 static DEVICE_ATTR_RO(phys_port_id);
444 static ssize_t phys_port_name_show(struct device *dev,
445 struct device_attribute *attr, char *buf)
447 struct net_device *netdev = to_net_dev(dev);
448 ssize_t ret = -EINVAL;
451 return restart_syscall();
453 if (dev_isalive(netdev)) {
456 ret = dev_get_phys_port_name(netdev, name, sizeof(name));
458 ret = sprintf(buf, "%s\n", name);
464 static DEVICE_ATTR_RO(phys_port_name);
466 static ssize_t phys_switch_id_show(struct device *dev,
467 struct device_attribute *attr, char *buf)
469 struct net_device *netdev = to_net_dev(dev);
470 ssize_t ret = -EINVAL;
473 return restart_syscall();
475 if (dev_isalive(netdev)) {
476 struct switchdev_attr attr = {
478 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
479 .flags = SWITCHDEV_F_NO_RECURSE,
482 ret = switchdev_port_attr_get(netdev, &attr);
484 ret = sprintf(buf, "%*phN\n", attr.u.ppid.id_len,
491 static DEVICE_ATTR_RO(phys_switch_id);
493 static struct attribute *net_class_attrs[] = {
494 &dev_attr_netdev_group.attr,
496 &dev_attr_dev_id.attr,
497 &dev_attr_dev_port.attr,
498 &dev_attr_iflink.attr,
499 &dev_attr_ifindex.attr,
500 &dev_attr_name_assign_type.attr,
501 &dev_attr_addr_assign_type.attr,
502 &dev_attr_addr_len.attr,
503 &dev_attr_link_mode.attr,
504 &dev_attr_address.attr,
505 &dev_attr_broadcast.attr,
506 &dev_attr_speed.attr,
507 &dev_attr_duplex.attr,
508 &dev_attr_dormant.attr,
509 &dev_attr_operstate.attr,
510 &dev_attr_carrier_changes.attr,
511 &dev_attr_ifalias.attr,
512 &dev_attr_carrier.attr,
514 &dev_attr_flags.attr,
515 &dev_attr_tx_queue_len.attr,
516 &dev_attr_gro_flush_timeout.attr,
517 &dev_attr_phys_port_id.attr,
518 &dev_attr_phys_port_name.attr,
519 &dev_attr_phys_switch_id.attr,
520 &dev_attr_proto_down.attr,
523 ATTRIBUTE_GROUPS(net_class);
525 /* Show a given an attribute in the statistics group */
526 static ssize_t netstat_show(const struct device *d,
527 struct device_attribute *attr, char *buf,
528 unsigned long offset)
530 struct net_device *dev = to_net_dev(d);
531 ssize_t ret = -EINVAL;
533 WARN_ON(offset > sizeof(struct rtnl_link_stats64) ||
534 offset % sizeof(u64) != 0);
536 read_lock(&dev_base_lock);
537 if (dev_isalive(dev)) {
538 struct rtnl_link_stats64 temp;
539 const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp);
541 ret = sprintf(buf, fmt_u64, *(u64 *)(((u8 *) stats) + offset));
543 read_unlock(&dev_base_lock);
547 /* generate a read-only statistics attribute */
548 #define NETSTAT_ENTRY(name) \
549 static ssize_t name##_show(struct device *d, \
550 struct device_attribute *attr, char *buf) \
552 return netstat_show(d, attr, buf, \
553 offsetof(struct rtnl_link_stats64, name)); \
555 static DEVICE_ATTR_RO(name)
557 NETSTAT_ENTRY(rx_packets);
558 NETSTAT_ENTRY(tx_packets);
559 NETSTAT_ENTRY(rx_bytes);
560 NETSTAT_ENTRY(tx_bytes);
561 NETSTAT_ENTRY(rx_errors);
562 NETSTAT_ENTRY(tx_errors);
563 NETSTAT_ENTRY(rx_dropped);
564 NETSTAT_ENTRY(tx_dropped);
565 NETSTAT_ENTRY(multicast);
566 NETSTAT_ENTRY(collisions);
567 NETSTAT_ENTRY(rx_length_errors);
568 NETSTAT_ENTRY(rx_over_errors);
569 NETSTAT_ENTRY(rx_crc_errors);
570 NETSTAT_ENTRY(rx_frame_errors);
571 NETSTAT_ENTRY(rx_fifo_errors);
572 NETSTAT_ENTRY(rx_missed_errors);
573 NETSTAT_ENTRY(tx_aborted_errors);
574 NETSTAT_ENTRY(tx_carrier_errors);
575 NETSTAT_ENTRY(tx_fifo_errors);
576 NETSTAT_ENTRY(tx_heartbeat_errors);
577 NETSTAT_ENTRY(tx_window_errors);
578 NETSTAT_ENTRY(rx_compressed);
579 NETSTAT_ENTRY(tx_compressed);
580 NETSTAT_ENTRY(rx_nohandler);
582 static struct attribute *netstat_attrs[] = {
583 &dev_attr_rx_packets.attr,
584 &dev_attr_tx_packets.attr,
585 &dev_attr_rx_bytes.attr,
586 &dev_attr_tx_bytes.attr,
587 &dev_attr_rx_errors.attr,
588 &dev_attr_tx_errors.attr,
589 &dev_attr_rx_dropped.attr,
590 &dev_attr_tx_dropped.attr,
591 &dev_attr_multicast.attr,
592 &dev_attr_collisions.attr,
593 &dev_attr_rx_length_errors.attr,
594 &dev_attr_rx_over_errors.attr,
595 &dev_attr_rx_crc_errors.attr,
596 &dev_attr_rx_frame_errors.attr,
597 &dev_attr_rx_fifo_errors.attr,
598 &dev_attr_rx_missed_errors.attr,
599 &dev_attr_tx_aborted_errors.attr,
600 &dev_attr_tx_carrier_errors.attr,
601 &dev_attr_tx_fifo_errors.attr,
602 &dev_attr_tx_heartbeat_errors.attr,
603 &dev_attr_tx_window_errors.attr,
604 &dev_attr_rx_compressed.attr,
605 &dev_attr_tx_compressed.attr,
606 &dev_attr_rx_nohandler.attr,
611 static struct attribute_group netstat_group = {
612 .name = "statistics",
613 .attrs = netstat_attrs,
616 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
617 static struct attribute *wireless_attrs[] = {
621 static struct attribute_group wireless_group = {
623 .attrs = wireless_attrs,
627 #else /* CONFIG_SYSFS */
628 #define net_class_groups NULL
629 #endif /* CONFIG_SYSFS */
632 #define to_rx_queue_attr(_attr) container_of(_attr, \
633 struct rx_queue_attribute, attr)
635 #define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj)
637 static ssize_t rx_queue_attr_show(struct kobject *kobj, struct attribute *attr,
640 struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
641 struct netdev_rx_queue *queue = to_rx_queue(kobj);
643 if (!attribute->show)
646 return attribute->show(queue, attribute, buf);
649 static ssize_t rx_queue_attr_store(struct kobject *kobj, struct attribute *attr,
650 const char *buf, size_t count)
652 struct rx_queue_attribute *attribute = to_rx_queue_attr(attr);
653 struct netdev_rx_queue *queue = to_rx_queue(kobj);
655 if (!attribute->store)
658 return attribute->store(queue, attribute, buf, count);
661 static const struct sysfs_ops rx_queue_sysfs_ops = {
662 .show = rx_queue_attr_show,
663 .store = rx_queue_attr_store,
667 static ssize_t show_rps_map(struct netdev_rx_queue *queue,
668 struct rx_queue_attribute *attribute, char *buf)
674 if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
678 map = rcu_dereference(queue->rps_map);
680 for (i = 0; i < map->len; i++)
681 cpumask_set_cpu(map->cpus[i], mask);
683 len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
685 free_cpumask_var(mask);
687 return len < PAGE_SIZE ? len : -EINVAL;
690 static ssize_t store_rps_map(struct netdev_rx_queue *queue,
691 struct rx_queue_attribute *attribute,
692 const char *buf, size_t len)
694 struct rps_map *old_map, *map;
697 static DEFINE_MUTEX(rps_map_mutex);
699 if (!capable(CAP_NET_ADMIN))
702 if (!alloc_cpumask_var(&mask, GFP_KERNEL))
705 err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
707 free_cpumask_var(mask);
711 map = kzalloc(max_t(unsigned int,
712 RPS_MAP_SIZE(cpumask_weight(mask)), L1_CACHE_BYTES),
715 free_cpumask_var(mask);
720 for_each_cpu_and(cpu, mask, cpu_online_mask)
721 map->cpus[i++] = cpu;
730 mutex_lock(&rps_map_mutex);
731 old_map = rcu_dereference_protected(queue->rps_map,
732 mutex_is_locked(&rps_map_mutex));
733 rcu_assign_pointer(queue->rps_map, map);
736 static_key_slow_inc(&rps_needed);
738 static_key_slow_dec(&rps_needed);
740 mutex_unlock(&rps_map_mutex);
743 kfree_rcu(old_map, rcu);
745 free_cpumask_var(mask);
749 static ssize_t show_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
750 struct rx_queue_attribute *attr,
753 struct rps_dev_flow_table *flow_table;
754 unsigned long val = 0;
757 flow_table = rcu_dereference(queue->rps_flow_table);
759 val = (unsigned long)flow_table->mask + 1;
762 return sprintf(buf, "%lu\n", val);
765 static void rps_dev_flow_table_release(struct rcu_head *rcu)
767 struct rps_dev_flow_table *table = container_of(rcu,
768 struct rps_dev_flow_table, rcu);
772 static ssize_t store_rps_dev_flow_table_cnt(struct netdev_rx_queue *queue,
773 struct rx_queue_attribute *attr,
774 const char *buf, size_t len)
776 unsigned long mask, count;
777 struct rps_dev_flow_table *table, *old_table;
778 static DEFINE_SPINLOCK(rps_dev_flow_lock);
781 if (!capable(CAP_NET_ADMIN))
784 rc = kstrtoul(buf, 0, &count);
790 /* mask = roundup_pow_of_two(count) - 1;
791 * without overflows...
793 while ((mask | (mask >> 1)) != mask)
795 /* On 64 bit arches, must check mask fits in table->mask (u32),
796 * and on 32bit arches, must check
797 * RPS_DEV_FLOW_TABLE_SIZE(mask + 1) doesn't overflow.
799 #if BITS_PER_LONG > 32
800 if (mask > (unsigned long)(u32)mask)
803 if (mask > (ULONG_MAX - RPS_DEV_FLOW_TABLE_SIZE(1))
804 / sizeof(struct rps_dev_flow)) {
805 /* Enforce a limit to prevent overflow */
809 table = vmalloc(RPS_DEV_FLOW_TABLE_SIZE(mask + 1));
814 for (count = 0; count <= mask; count++)
815 table->flows[count].cpu = RPS_NO_CPU;
819 spin_lock(&rps_dev_flow_lock);
820 old_table = rcu_dereference_protected(queue->rps_flow_table,
821 lockdep_is_held(&rps_dev_flow_lock));
822 rcu_assign_pointer(queue->rps_flow_table, table);
823 spin_unlock(&rps_dev_flow_lock);
826 call_rcu(&old_table->rcu, rps_dev_flow_table_release);
831 static struct rx_queue_attribute rps_cpus_attribute =
832 __ATTR(rps_cpus, S_IRUGO | S_IWUSR, show_rps_map, store_rps_map);
835 static struct rx_queue_attribute rps_dev_flow_table_cnt_attribute =
836 __ATTR(rps_flow_cnt, S_IRUGO | S_IWUSR,
837 show_rps_dev_flow_table_cnt, store_rps_dev_flow_table_cnt);
838 #endif /* CONFIG_RPS */
840 static struct attribute *rx_queue_default_attrs[] = {
842 &rps_cpus_attribute.attr,
843 &rps_dev_flow_table_cnt_attribute.attr,
848 static void rx_queue_release(struct kobject *kobj)
850 struct netdev_rx_queue *queue = to_rx_queue(kobj);
853 struct rps_dev_flow_table *flow_table;
856 map = rcu_dereference_protected(queue->rps_map, 1);
858 RCU_INIT_POINTER(queue->rps_map, NULL);
862 flow_table = rcu_dereference_protected(queue->rps_flow_table, 1);
864 RCU_INIT_POINTER(queue->rps_flow_table, NULL);
865 call_rcu(&flow_table->rcu, rps_dev_flow_table_release);
869 memset(kobj, 0, sizeof(*kobj));
873 static const void *rx_queue_namespace(struct kobject *kobj)
875 struct netdev_rx_queue *queue = to_rx_queue(kobj);
876 struct device *dev = &queue->dev->dev;
877 const void *ns = NULL;
879 if (dev->class && dev->class->ns_type)
880 ns = dev->class->namespace(dev);
885 static struct kobj_type rx_queue_ktype = {
886 .sysfs_ops = &rx_queue_sysfs_ops,
887 .release = rx_queue_release,
888 .default_attrs = rx_queue_default_attrs,
889 .namespace = rx_queue_namespace
892 static int rx_queue_add_kobject(struct net_device *dev, int index)
894 struct netdev_rx_queue *queue = dev->_rx + index;
895 struct kobject *kobj = &queue->kobj;
898 kobj->kset = dev->queues_kset;
899 error = kobject_init_and_add(kobj, &rx_queue_ktype, NULL,
904 if (dev->sysfs_rx_queue_group) {
905 error = sysfs_create_group(kobj, dev->sysfs_rx_queue_group);
910 kobject_uevent(kobj, KOBJ_ADD);
911 dev_hold(queue->dev);
918 #endif /* CONFIG_SYSFS */
921 net_rx_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
928 if (!dev->sysfs_rx_queue_group)
931 for (i = old_num; i < new_num; i++) {
932 error = rx_queue_add_kobject(dev, i);
939 while (--i >= new_num) {
940 if (dev->sysfs_rx_queue_group)
941 sysfs_remove_group(&dev->_rx[i].kobj,
942 dev->sysfs_rx_queue_group);
943 kobject_put(&dev->_rx[i].kobj);
954 * netdev_queue sysfs structures and functions.
956 struct netdev_queue_attribute {
957 struct attribute attr;
958 ssize_t (*show)(struct netdev_queue *queue,
959 struct netdev_queue_attribute *attr, char *buf);
960 ssize_t (*store)(struct netdev_queue *queue,
961 struct netdev_queue_attribute *attr, const char *buf, size_t len);
963 #define to_netdev_queue_attr(_attr) container_of(_attr, \
964 struct netdev_queue_attribute, attr)
966 #define to_netdev_queue(obj) container_of(obj, struct netdev_queue, kobj)
968 static ssize_t netdev_queue_attr_show(struct kobject *kobj,
969 struct attribute *attr, char *buf)
971 struct netdev_queue_attribute *attribute = to_netdev_queue_attr(attr);
972 struct netdev_queue *queue = to_netdev_queue(kobj);
974 if (!attribute->show)
977 return attribute->show(queue, attribute, buf);
980 static ssize_t netdev_queue_attr_store(struct kobject *kobj,
981 struct attribute *attr,
982 const char *buf, size_t count)
984 struct netdev_queue_attribute *attribute = to_netdev_queue_attr(attr);
985 struct netdev_queue *queue = to_netdev_queue(kobj);
987 if (!attribute->store)
990 return attribute->store(queue, attribute, buf, count);
993 static const struct sysfs_ops netdev_queue_sysfs_ops = {
994 .show = netdev_queue_attr_show,
995 .store = netdev_queue_attr_store,
998 static ssize_t show_trans_timeout(struct netdev_queue *queue,
999 struct netdev_queue_attribute *attribute,
1002 unsigned long trans_timeout;
1004 spin_lock_irq(&queue->_xmit_lock);
1005 trans_timeout = queue->trans_timeout;
1006 spin_unlock_irq(&queue->_xmit_lock);
1008 return sprintf(buf, "%lu", trans_timeout);
1012 static unsigned int get_netdev_queue_index(struct netdev_queue *queue)
1014 struct net_device *dev = queue->dev;
1017 i = queue - dev->_tx;
1018 BUG_ON(i >= dev->num_tx_queues);
1023 static ssize_t show_tx_maxrate(struct netdev_queue *queue,
1024 struct netdev_queue_attribute *attribute,
1027 return sprintf(buf, "%lu\n", queue->tx_maxrate);
1030 static ssize_t set_tx_maxrate(struct netdev_queue *queue,
1031 struct netdev_queue_attribute *attribute,
1032 const char *buf, size_t len)
1034 struct net_device *dev = queue->dev;
1035 int err, index = get_netdev_queue_index(queue);
1038 err = kstrtou32(buf, 10, &rate);
1042 if (!rtnl_trylock())
1043 return restart_syscall();
1046 if (dev->netdev_ops->ndo_set_tx_maxrate)
1047 err = dev->netdev_ops->ndo_set_tx_maxrate(dev, index, rate);
1051 queue->tx_maxrate = rate;
1057 static struct netdev_queue_attribute queue_tx_maxrate =
1058 __ATTR(tx_maxrate, S_IRUGO | S_IWUSR,
1059 show_tx_maxrate, set_tx_maxrate);
1062 static struct netdev_queue_attribute queue_trans_timeout =
1063 __ATTR(tx_timeout, S_IRUGO, show_trans_timeout, NULL);
1067 * Byte queue limits sysfs structures and functions.
1069 static ssize_t bql_show(char *buf, unsigned int value)
1071 return sprintf(buf, "%u\n", value);
1074 static ssize_t bql_set(const char *buf, const size_t count,
1075 unsigned int *pvalue)
1080 if (!strcmp(buf, "max") || !strcmp(buf, "max\n"))
1081 value = DQL_MAX_LIMIT;
1083 err = kstrtouint(buf, 10, &value);
1086 if (value > DQL_MAX_LIMIT)
1095 static ssize_t bql_show_hold_time(struct netdev_queue *queue,
1096 struct netdev_queue_attribute *attr,
1099 struct dql *dql = &queue->dql;
1101 return sprintf(buf, "%u\n", jiffies_to_msecs(dql->slack_hold_time));
1104 static ssize_t bql_set_hold_time(struct netdev_queue *queue,
1105 struct netdev_queue_attribute *attribute,
1106 const char *buf, size_t len)
1108 struct dql *dql = &queue->dql;
1112 err = kstrtouint(buf, 10, &value);
1116 dql->slack_hold_time = msecs_to_jiffies(value);
1121 static struct netdev_queue_attribute bql_hold_time_attribute =
1122 __ATTR(hold_time, S_IRUGO | S_IWUSR, bql_show_hold_time,
1125 static ssize_t bql_show_inflight(struct netdev_queue *queue,
1126 struct netdev_queue_attribute *attr,
1129 struct dql *dql = &queue->dql;
1131 return sprintf(buf, "%u\n", dql->num_queued - dql->num_completed);
1134 static struct netdev_queue_attribute bql_inflight_attribute =
1135 __ATTR(inflight, S_IRUGO, bql_show_inflight, NULL);
1137 #define BQL_ATTR(NAME, FIELD) \
1138 static ssize_t bql_show_ ## NAME(struct netdev_queue *queue, \
1139 struct netdev_queue_attribute *attr, \
1142 return bql_show(buf, queue->dql.FIELD); \
1145 static ssize_t bql_set_ ## NAME(struct netdev_queue *queue, \
1146 struct netdev_queue_attribute *attr, \
1147 const char *buf, size_t len) \
1149 return bql_set(buf, len, &queue->dql.FIELD); \
1152 static struct netdev_queue_attribute bql_ ## NAME ## _attribute = \
1153 __ATTR(NAME, S_IRUGO | S_IWUSR, bql_show_ ## NAME, \
1156 BQL_ATTR(limit, limit)
1157 BQL_ATTR(limit_max, max_limit)
1158 BQL_ATTR(limit_min, min_limit)
1160 static struct attribute *dql_attrs[] = {
1161 &bql_limit_attribute.attr,
1162 &bql_limit_max_attribute.attr,
1163 &bql_limit_min_attribute.attr,
1164 &bql_hold_time_attribute.attr,
1165 &bql_inflight_attribute.attr,
1169 static struct attribute_group dql_group = {
1170 .name = "byte_queue_limits",
1173 #endif /* CONFIG_BQL */
1176 static ssize_t show_xps_map(struct netdev_queue *queue,
1177 struct netdev_queue_attribute *attribute, char *buf)
1179 struct net_device *dev = queue->dev;
1180 struct xps_dev_maps *dev_maps;
1182 unsigned long index;
1185 if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
1188 index = get_netdev_queue_index(queue);
1191 dev_maps = rcu_dereference(dev->xps_maps);
1193 for_each_possible_cpu(i) {
1194 struct xps_map *map =
1195 rcu_dereference(dev_maps->cpu_map[i]);
1198 for (j = 0; j < map->len; j++) {
1199 if (map->queues[j] == index) {
1200 cpumask_set_cpu(i, mask);
1209 len = snprintf(buf, PAGE_SIZE, "%*pb\n", cpumask_pr_args(mask));
1210 free_cpumask_var(mask);
1211 return len < PAGE_SIZE ? len : -EINVAL;
1214 static ssize_t store_xps_map(struct netdev_queue *queue,
1215 struct netdev_queue_attribute *attribute,
1216 const char *buf, size_t len)
1218 struct net_device *dev = queue->dev;
1219 unsigned long index;
1223 if (!capable(CAP_NET_ADMIN))
1226 if (!alloc_cpumask_var(&mask, GFP_KERNEL))
1229 index = get_netdev_queue_index(queue);
1231 err = bitmap_parse(buf, len, cpumask_bits(mask), nr_cpumask_bits);
1233 free_cpumask_var(mask);
1237 err = netif_set_xps_queue(dev, mask, index);
1239 free_cpumask_var(mask);
1244 static struct netdev_queue_attribute xps_cpus_attribute =
1245 __ATTR(xps_cpus, S_IRUGO | S_IWUSR, show_xps_map, store_xps_map);
1246 #endif /* CONFIG_XPS */
1248 static struct attribute *netdev_queue_default_attrs[] = {
1249 &queue_trans_timeout.attr,
1251 &xps_cpus_attribute.attr,
1252 &queue_tx_maxrate.attr,
1257 static void netdev_queue_release(struct kobject *kobj)
1259 struct netdev_queue *queue = to_netdev_queue(kobj);
1261 memset(kobj, 0, sizeof(*kobj));
1262 dev_put(queue->dev);
1265 static const void *netdev_queue_namespace(struct kobject *kobj)
1267 struct netdev_queue *queue = to_netdev_queue(kobj);
1268 struct device *dev = &queue->dev->dev;
1269 const void *ns = NULL;
1271 if (dev->class && dev->class->ns_type)
1272 ns = dev->class->namespace(dev);
1277 static struct kobj_type netdev_queue_ktype = {
1278 .sysfs_ops = &netdev_queue_sysfs_ops,
1279 .release = netdev_queue_release,
1280 .default_attrs = netdev_queue_default_attrs,
1281 .namespace = netdev_queue_namespace,
1284 static int netdev_queue_add_kobject(struct net_device *dev, int index)
1286 struct netdev_queue *queue = dev->_tx + index;
1287 struct kobject *kobj = &queue->kobj;
1290 kobj->kset = dev->queues_kset;
1291 error = kobject_init_and_add(kobj, &netdev_queue_ktype, NULL,
1297 error = sysfs_create_group(kobj, &dql_group);
1302 kobject_uevent(kobj, KOBJ_ADD);
1303 dev_hold(queue->dev);
1310 #endif /* CONFIG_SYSFS */
1313 netdev_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
1319 for (i = old_num; i < new_num; i++) {
1320 error = netdev_queue_add_kobject(dev, i);
1327 while (--i >= new_num) {
1328 struct netdev_queue *queue = dev->_tx + i;
1331 sysfs_remove_group(&queue->kobj, &dql_group);
1333 kobject_put(&queue->kobj);
1339 #endif /* CONFIG_SYSFS */
1342 static int register_queue_kobjects(struct net_device *dev)
1344 int error = 0, txq = 0, rxq = 0, real_rx = 0, real_tx = 0;
1347 dev->queues_kset = kset_create_and_add("queues",
1348 NULL, &dev->dev.kobj);
1349 if (!dev->queues_kset)
1351 real_rx = dev->real_num_rx_queues;
1353 real_tx = dev->real_num_tx_queues;
1355 error = net_rx_queue_update_kobjects(dev, 0, real_rx);
1360 error = netdev_queue_update_kobjects(dev, 0, real_tx);
1368 netdev_queue_update_kobjects(dev, txq, 0);
1369 net_rx_queue_update_kobjects(dev, rxq, 0);
1373 static void remove_queue_kobjects(struct net_device *dev)
1375 int real_rx = 0, real_tx = 0;
1378 real_rx = dev->real_num_rx_queues;
1380 real_tx = dev->real_num_tx_queues;
1382 net_rx_queue_update_kobjects(dev, real_rx, 0);
1383 netdev_queue_update_kobjects(dev, real_tx, 0);
1385 kset_unregister(dev->queues_kset);
1389 static bool net_current_may_mount(void)
1391 struct net *net = current->nsproxy->net_ns;
1393 return ns_capable(net->user_ns, CAP_SYS_ADMIN);
1396 static void *net_grab_current_ns(void)
1398 struct net *ns = current->nsproxy->net_ns;
1399 #ifdef CONFIG_NET_NS
1401 atomic_inc(&ns->passive);
1406 static const void *net_initial_ns(void)
1411 static const void *net_netlink_ns(struct sock *sk)
1413 return sock_net(sk);
1416 struct kobj_ns_type_operations net_ns_type_operations = {
1417 .type = KOBJ_NS_TYPE_NET,
1418 .current_may_mount = net_current_may_mount,
1419 .grab_current_ns = net_grab_current_ns,
1420 .netlink_ns = net_netlink_ns,
1421 .initial_ns = net_initial_ns,
1422 .drop_ns = net_drop_ns,
1424 EXPORT_SYMBOL_GPL(net_ns_type_operations);
1426 static int netdev_uevent(struct device *d, struct kobj_uevent_env *env)
1428 struct net_device *dev = to_net_dev(d);
1431 /* pass interface to uevent. */
1432 retval = add_uevent_var(env, "INTERFACE=%s", dev->name);
1436 /* pass ifindex to uevent.
1437 * ifindex is useful as it won't change (interface name may change)
1438 * and is what RtNetlink uses natively. */
1439 retval = add_uevent_var(env, "IFINDEX=%d", dev->ifindex);
1446 * netdev_release -- destroy and free a dead device.
1447 * Called when last reference to device kobject is gone.
1449 static void netdev_release(struct device *d)
1451 struct net_device *dev = to_net_dev(d);
1453 BUG_ON(dev->reg_state != NETREG_RELEASED);
1455 kfree(dev->ifalias);
1456 netdev_freemem(dev);
1459 static const void *net_namespace(struct device *d)
1461 struct net_device *dev = to_net_dev(d);
1463 return dev_net(dev);
1466 static struct class net_class = {
1468 .dev_release = netdev_release,
1469 .dev_groups = net_class_groups,
1470 .dev_uevent = netdev_uevent,
1471 .ns_type = &net_ns_type_operations,
1472 .namespace = net_namespace,
1475 #ifdef CONFIG_OF_NET
1476 static int of_dev_node_match(struct device *dev, const void *data)
1481 ret = dev->parent->of_node == data;
1483 return ret == 0 ? dev->of_node == data : ret;
1487 * of_find_net_device_by_node - lookup the net device for the device node
1488 * @np: OF device node
1490 * Looks up the net_device structure corresponding with the device node.
1491 * If successful, returns a pointer to the net_device with the embedded
1492 * struct device refcount incremented by one, or NULL on failure. The
1493 * refcount must be dropped when done with the net_device.
1495 struct net_device *of_find_net_device_by_node(struct device_node *np)
1499 dev = class_find_device(&net_class, NULL, np, of_dev_node_match);
1503 return to_net_dev(dev);
1505 EXPORT_SYMBOL(of_find_net_device_by_node);
1508 /* Delete sysfs entries but hold kobject reference until after all
1509 * netdev references are gone.
1511 void netdev_unregister_kobject(struct net_device *ndev)
1513 struct device *dev = &(ndev->dev);
1515 kobject_get(&dev->kobj);
1517 remove_queue_kobjects(ndev);
1519 pm_runtime_set_memalloc_noio(dev, false);
1524 /* Create sysfs entries for network device. */
1525 int netdev_register_kobject(struct net_device *ndev)
1527 struct device *dev = &(ndev->dev);
1528 const struct attribute_group **groups = ndev->sysfs_groups;
1531 device_initialize(dev);
1532 dev->class = &net_class;
1533 dev->platform_data = ndev;
1534 dev->groups = groups;
1536 dev_set_name(dev, "%s", ndev->name);
1539 /* Allow for a device specific group */
1543 *groups++ = &netstat_group;
1545 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
1546 if (ndev->ieee80211_ptr)
1547 *groups++ = &wireless_group;
1548 #if IS_ENABLED(CONFIG_WIRELESS_EXT)
1549 else if (ndev->wireless_handlers)
1550 *groups++ = &wireless_group;
1553 #endif /* CONFIG_SYSFS */
1555 error = device_add(dev);
1559 error = register_queue_kobjects(ndev);
1565 pm_runtime_set_memalloc_noio(dev, true);
1570 int netdev_class_create_file_ns(struct class_attribute *class_attr,
1573 return class_create_file_ns(&net_class, class_attr, ns);
1575 EXPORT_SYMBOL(netdev_class_create_file_ns);
1577 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
1580 class_remove_file_ns(&net_class, class_attr, ns);
1582 EXPORT_SYMBOL(netdev_class_remove_file_ns);
1584 int __init netdev_kobject_init(void)
1586 kobj_ns_type_register(&net_ns_type_operations);
1587 return class_register(&net_class);