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/sched/signal.h>
19 #include <linux/nsproxy.h>
21 #include <net/net_namespace.h>
22 #include <linux/rtnetlink.h>
23 #include <linux/vmalloc.h>
24 #include <linux/export.h>
25 #include <linux/jiffies.h>
26 #include <linux/pm_runtime.h>
28 #include <linux/of_net.h>
30 #include "net-sysfs.h"
33 static const char fmt_hex
[] = "%#x\n";
34 static const char fmt_dec
[] = "%d\n";
35 static const char fmt_ulong
[] = "%lu\n";
36 static const char fmt_u64
[] = "%llu\n";
38 static inline int dev_isalive(const struct net_device
*dev
)
40 return dev
->reg_state
<= NETREG_REGISTERED
;
43 /* use same locking rules as GIF* ioctl's */
44 static ssize_t
netdev_show(const struct device
*dev
,
45 struct device_attribute
*attr
, char *buf
,
46 ssize_t (*format
)(const struct net_device
*, char *))
48 struct net_device
*ndev
= to_net_dev(dev
);
49 ssize_t ret
= -EINVAL
;
51 read_lock(&dev_base_lock
);
52 if (dev_isalive(ndev
))
53 ret
= (*format
)(ndev
, buf
);
54 read_unlock(&dev_base_lock
);
59 /* generate a show function for simple field */
60 #define NETDEVICE_SHOW(field, format_string) \
61 static ssize_t format_##field(const struct net_device *dev, char *buf) \
63 return sprintf(buf, format_string, dev->field); \
65 static ssize_t field##_show(struct device *dev, \
66 struct device_attribute *attr, char *buf) \
68 return netdev_show(dev, attr, buf, format_##field); \
71 #define NETDEVICE_SHOW_RO(field, format_string) \
72 NETDEVICE_SHOW(field, format_string); \
73 static DEVICE_ATTR_RO(field)
75 #define NETDEVICE_SHOW_RW(field, format_string) \
76 NETDEVICE_SHOW(field, format_string); \
77 static DEVICE_ATTR_RW(field)
79 /* use same locking and permission rules as SIF* ioctl's */
80 static ssize_t
netdev_store(struct device
*dev
, struct device_attribute
*attr
,
81 const char *buf
, size_t len
,
82 int (*set
)(struct net_device
*, unsigned long))
84 struct net_device
*netdev
= to_net_dev(dev
);
85 struct net
*net
= dev_net(netdev
);
89 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
92 ret
= kstrtoul(buf
, 0, &new);
97 return restart_syscall();
99 if (dev_isalive(netdev
)) {
100 if ((ret
= (*set
)(netdev
, new)) == 0)
108 NETDEVICE_SHOW_RO(dev_id
, fmt_hex
);
109 NETDEVICE_SHOW_RO(dev_port
, fmt_dec
);
110 NETDEVICE_SHOW_RO(addr_assign_type
, fmt_dec
);
111 NETDEVICE_SHOW_RO(addr_len
, fmt_dec
);
112 NETDEVICE_SHOW_RO(ifindex
, fmt_dec
);
113 NETDEVICE_SHOW_RO(type
, fmt_dec
);
114 NETDEVICE_SHOW_RO(link_mode
, fmt_dec
);
116 static ssize_t
iflink_show(struct device
*dev
, struct device_attribute
*attr
,
119 struct net_device
*ndev
= to_net_dev(dev
);
121 return sprintf(buf
, fmt_dec
, dev_get_iflink(ndev
));
123 static DEVICE_ATTR_RO(iflink
);
125 static ssize_t
format_name_assign_type(const struct net_device
*dev
, char *buf
)
127 return sprintf(buf
, fmt_dec
, dev
->name_assign_type
);
130 static ssize_t
name_assign_type_show(struct device
*dev
,
131 struct device_attribute
*attr
,
134 struct net_device
*ndev
= to_net_dev(dev
);
135 ssize_t ret
= -EINVAL
;
137 if (ndev
->name_assign_type
!= NET_NAME_UNKNOWN
)
138 ret
= netdev_show(dev
, attr
, buf
, format_name_assign_type
);
142 static DEVICE_ATTR_RO(name_assign_type
);
144 /* use same locking rules as GIFHWADDR ioctl's */
145 static ssize_t
address_show(struct device
*dev
, struct device_attribute
*attr
,
148 struct net_device
*ndev
= to_net_dev(dev
);
149 ssize_t ret
= -EINVAL
;
151 read_lock(&dev_base_lock
);
152 if (dev_isalive(ndev
))
153 ret
= sysfs_format_mac(buf
, ndev
->dev_addr
, ndev
->addr_len
);
154 read_unlock(&dev_base_lock
);
157 static DEVICE_ATTR_RO(address
);
159 static ssize_t
broadcast_show(struct device
*dev
,
160 struct device_attribute
*attr
, char *buf
)
162 struct net_device
*ndev
= to_net_dev(dev
);
163 if (dev_isalive(ndev
))
164 return sysfs_format_mac(buf
, ndev
->broadcast
, ndev
->addr_len
);
167 static DEVICE_ATTR_RO(broadcast
);
169 static int change_carrier(struct net_device
*dev
, unsigned long new_carrier
)
171 if (!netif_running(dev
))
173 return dev_change_carrier(dev
, (bool) new_carrier
);
176 static ssize_t
carrier_store(struct device
*dev
, struct device_attribute
*attr
,
177 const char *buf
, size_t len
)
179 return netdev_store(dev
, attr
, buf
, len
, change_carrier
);
182 static ssize_t
carrier_show(struct device
*dev
,
183 struct device_attribute
*attr
, char *buf
)
185 struct net_device
*netdev
= to_net_dev(dev
);
186 if (netif_running(netdev
)) {
187 return sprintf(buf
, fmt_dec
, !!netif_carrier_ok(netdev
));
191 static DEVICE_ATTR_RW(carrier
);
193 static ssize_t
speed_show(struct device
*dev
,
194 struct device_attribute
*attr
, char *buf
)
196 struct net_device
*netdev
= to_net_dev(dev
);
200 return restart_syscall();
202 if (netif_running(netdev
)) {
203 struct ethtool_link_ksettings cmd
;
205 if (!__ethtool_get_link_ksettings(netdev
, &cmd
))
206 ret
= sprintf(buf
, fmt_dec
, cmd
.base
.speed
);
211 static DEVICE_ATTR_RO(speed
);
213 static ssize_t
duplex_show(struct device
*dev
,
214 struct device_attribute
*attr
, char *buf
)
216 struct net_device
*netdev
= to_net_dev(dev
);
220 return restart_syscall();
222 if (netif_running(netdev
)) {
223 struct ethtool_link_ksettings cmd
;
225 if (!__ethtool_get_link_ksettings(netdev
, &cmd
)) {
228 switch (cmd
.base
.duplex
) {
239 ret
= sprintf(buf
, "%s\n", duplex
);
245 static DEVICE_ATTR_RO(duplex
);
247 static ssize_t
dormant_show(struct device
*dev
,
248 struct device_attribute
*attr
, char *buf
)
250 struct net_device
*netdev
= to_net_dev(dev
);
252 if (netif_running(netdev
))
253 return sprintf(buf
, fmt_dec
, !!netif_dormant(netdev
));
257 static DEVICE_ATTR_RO(dormant
);
259 static const char *const operstates
[] = {
261 "notpresent", /* currently unused */
264 "testing", /* currently unused */
269 static ssize_t
operstate_show(struct device
*dev
,
270 struct device_attribute
*attr
, char *buf
)
272 const struct net_device
*netdev
= to_net_dev(dev
);
273 unsigned char operstate
;
275 read_lock(&dev_base_lock
);
276 operstate
= netdev
->operstate
;
277 if (!netif_running(netdev
))
278 operstate
= IF_OPER_DOWN
;
279 read_unlock(&dev_base_lock
);
281 if (operstate
>= ARRAY_SIZE(operstates
))
282 return -EINVAL
; /* should not happen */
284 return sprintf(buf
, "%s\n", operstates
[operstate
]);
286 static DEVICE_ATTR_RO(operstate
);
288 static ssize_t
carrier_changes_show(struct device
*dev
,
289 struct device_attribute
*attr
,
292 struct net_device
*netdev
= to_net_dev(dev
);
293 return sprintf(buf
, fmt_dec
,
294 atomic_read(&netdev
->carrier_changes
));
296 static DEVICE_ATTR_RO(carrier_changes
);
298 /* read-write attributes */
300 static int change_mtu(struct net_device
*dev
, unsigned long new_mtu
)
302 return dev_set_mtu(dev
, (int) new_mtu
);
305 static ssize_t
mtu_store(struct device
*dev
, struct device_attribute
*attr
,
306 const char *buf
, size_t len
)
308 return netdev_store(dev
, attr
, buf
, len
, change_mtu
);
310 NETDEVICE_SHOW_RW(mtu
, fmt_dec
);
312 static int change_flags(struct net_device
*dev
, unsigned long new_flags
)
314 return dev_change_flags(dev
, (unsigned int) new_flags
);
317 static ssize_t
flags_store(struct device
*dev
, struct device_attribute
*attr
,
318 const char *buf
, size_t len
)
320 return netdev_store(dev
, attr
, buf
, len
, change_flags
);
322 NETDEVICE_SHOW_RW(flags
, fmt_hex
);
324 static int change_tx_queue_len(struct net_device
*dev
, unsigned long new_len
)
326 unsigned int orig_len
= dev
->tx_queue_len
;
329 if (new_len
!= (unsigned int)new_len
)
332 if (new_len
!= orig_len
) {
333 dev
->tx_queue_len
= new_len
;
334 res
= call_netdevice_notifiers(NETDEV_CHANGE_TX_QUEUE_LEN
, dev
);
335 res
= notifier_to_errno(res
);
338 "refused to change device tx_queue_len\n");
339 dev
->tx_queue_len
= orig_len
;
347 static ssize_t
tx_queue_len_store(struct device
*dev
,
348 struct device_attribute
*attr
,
349 const char *buf
, size_t len
)
351 if (!capable(CAP_NET_ADMIN
))
354 return netdev_store(dev
, attr
, buf
, len
, change_tx_queue_len
);
356 NETDEVICE_SHOW_RW(tx_queue_len
, fmt_dec
);
358 static int change_gro_flush_timeout(struct net_device
*dev
, unsigned long val
)
360 dev
->gro_flush_timeout
= val
;
364 static ssize_t
gro_flush_timeout_store(struct device
*dev
,
365 struct device_attribute
*attr
,
366 const char *buf
, size_t len
)
368 if (!capable(CAP_NET_ADMIN
))
371 return netdev_store(dev
, attr
, buf
, len
, change_gro_flush_timeout
);
373 NETDEVICE_SHOW_RW(gro_flush_timeout
, fmt_ulong
);
375 static ssize_t
ifalias_store(struct device
*dev
, struct device_attribute
*attr
,
376 const char *buf
, size_t len
)
378 struct net_device
*netdev
= to_net_dev(dev
);
379 struct net
*net
= dev_net(netdev
);
383 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
386 /* ignore trailing newline */
387 if (len
> 0 && buf
[len
- 1] == '\n')
391 return restart_syscall();
392 ret
= dev_set_alias(netdev
, buf
, count
);
395 return ret
< 0 ? ret
: len
;
398 static ssize_t
ifalias_show(struct device
*dev
,
399 struct device_attribute
*attr
, char *buf
)
401 const struct net_device
*netdev
= to_net_dev(dev
);
405 return restart_syscall();
407 ret
= sprintf(buf
, "%s\n", netdev
->ifalias
);
411 static DEVICE_ATTR_RW(ifalias
);
413 static int change_group(struct net_device
*dev
, unsigned long new_group
)
415 dev_set_group(dev
, (int) new_group
);
419 static ssize_t
group_store(struct device
*dev
, struct device_attribute
*attr
,
420 const char *buf
, size_t len
)
422 return netdev_store(dev
, attr
, buf
, len
, change_group
);
424 NETDEVICE_SHOW(group
, fmt_dec
);
425 static DEVICE_ATTR(netdev_group
, S_IRUGO
| S_IWUSR
, group_show
, group_store
);
427 static int change_proto_down(struct net_device
*dev
, unsigned long proto_down
)
429 return dev_change_proto_down(dev
, (bool) proto_down
);
432 static ssize_t
proto_down_store(struct device
*dev
,
433 struct device_attribute
*attr
,
434 const char *buf
, size_t len
)
436 return netdev_store(dev
, attr
, buf
, len
, change_proto_down
);
438 NETDEVICE_SHOW_RW(proto_down
, fmt_dec
);
440 static ssize_t
phys_port_id_show(struct device
*dev
,
441 struct device_attribute
*attr
, char *buf
)
443 struct net_device
*netdev
= to_net_dev(dev
);
444 ssize_t ret
= -EINVAL
;
447 return restart_syscall();
449 if (dev_isalive(netdev
)) {
450 struct netdev_phys_item_id ppid
;
452 ret
= dev_get_phys_port_id(netdev
, &ppid
);
454 ret
= sprintf(buf
, "%*phN\n", ppid
.id_len
, ppid
.id
);
460 static DEVICE_ATTR_RO(phys_port_id
);
462 static ssize_t
phys_port_name_show(struct device
*dev
,
463 struct device_attribute
*attr
, char *buf
)
465 struct net_device
*netdev
= to_net_dev(dev
);
466 ssize_t ret
= -EINVAL
;
469 return restart_syscall();
471 if (dev_isalive(netdev
)) {
474 ret
= dev_get_phys_port_name(netdev
, name
, sizeof(name
));
476 ret
= sprintf(buf
, "%s\n", name
);
482 static DEVICE_ATTR_RO(phys_port_name
);
484 static ssize_t
phys_switch_id_show(struct device
*dev
,
485 struct device_attribute
*attr
, char *buf
)
487 struct net_device
*netdev
= to_net_dev(dev
);
488 ssize_t ret
= -EINVAL
;
491 return restart_syscall();
493 if (dev_isalive(netdev
)) {
494 struct switchdev_attr attr
= {
496 .id
= SWITCHDEV_ATTR_ID_PORT_PARENT_ID
,
497 .flags
= SWITCHDEV_F_NO_RECURSE
,
500 ret
= switchdev_port_attr_get(netdev
, &attr
);
502 ret
= sprintf(buf
, "%*phN\n", attr
.u
.ppid
.id_len
,
509 static DEVICE_ATTR_RO(phys_switch_id
);
511 static struct attribute
*net_class_attrs
[] = {
512 &dev_attr_netdev_group
.attr
,
514 &dev_attr_dev_id
.attr
,
515 &dev_attr_dev_port
.attr
,
516 &dev_attr_iflink
.attr
,
517 &dev_attr_ifindex
.attr
,
518 &dev_attr_name_assign_type
.attr
,
519 &dev_attr_addr_assign_type
.attr
,
520 &dev_attr_addr_len
.attr
,
521 &dev_attr_link_mode
.attr
,
522 &dev_attr_address
.attr
,
523 &dev_attr_broadcast
.attr
,
524 &dev_attr_speed
.attr
,
525 &dev_attr_duplex
.attr
,
526 &dev_attr_dormant
.attr
,
527 &dev_attr_operstate
.attr
,
528 &dev_attr_carrier_changes
.attr
,
529 &dev_attr_ifalias
.attr
,
530 &dev_attr_carrier
.attr
,
532 &dev_attr_flags
.attr
,
533 &dev_attr_tx_queue_len
.attr
,
534 &dev_attr_gro_flush_timeout
.attr
,
535 &dev_attr_phys_port_id
.attr
,
536 &dev_attr_phys_port_name
.attr
,
537 &dev_attr_phys_switch_id
.attr
,
538 &dev_attr_proto_down
.attr
,
541 ATTRIBUTE_GROUPS(net_class
);
543 /* Show a given an attribute in the statistics group */
544 static ssize_t
netstat_show(const struct device
*d
,
545 struct device_attribute
*attr
, char *buf
,
546 unsigned long offset
)
548 struct net_device
*dev
= to_net_dev(d
);
549 ssize_t ret
= -EINVAL
;
551 WARN_ON(offset
> sizeof(struct rtnl_link_stats64
) ||
552 offset
% sizeof(u64
) != 0);
554 read_lock(&dev_base_lock
);
555 if (dev_isalive(dev
)) {
556 struct rtnl_link_stats64 temp
;
557 const struct rtnl_link_stats64
*stats
= dev_get_stats(dev
, &temp
);
559 ret
= sprintf(buf
, fmt_u64
, *(u64
*)(((u8
*) stats
) + offset
));
561 read_unlock(&dev_base_lock
);
565 /* generate a read-only statistics attribute */
566 #define NETSTAT_ENTRY(name) \
567 static ssize_t name##_show(struct device *d, \
568 struct device_attribute *attr, char *buf) \
570 return netstat_show(d, attr, buf, \
571 offsetof(struct rtnl_link_stats64, name)); \
573 static DEVICE_ATTR_RO(name)
575 NETSTAT_ENTRY(rx_packets
);
576 NETSTAT_ENTRY(tx_packets
);
577 NETSTAT_ENTRY(rx_bytes
);
578 NETSTAT_ENTRY(tx_bytes
);
579 NETSTAT_ENTRY(rx_errors
);
580 NETSTAT_ENTRY(tx_errors
);
581 NETSTAT_ENTRY(rx_dropped
);
582 NETSTAT_ENTRY(tx_dropped
);
583 NETSTAT_ENTRY(multicast
);
584 NETSTAT_ENTRY(collisions
);
585 NETSTAT_ENTRY(rx_length_errors
);
586 NETSTAT_ENTRY(rx_over_errors
);
587 NETSTAT_ENTRY(rx_crc_errors
);
588 NETSTAT_ENTRY(rx_frame_errors
);
589 NETSTAT_ENTRY(rx_fifo_errors
);
590 NETSTAT_ENTRY(rx_missed_errors
);
591 NETSTAT_ENTRY(tx_aborted_errors
);
592 NETSTAT_ENTRY(tx_carrier_errors
);
593 NETSTAT_ENTRY(tx_fifo_errors
);
594 NETSTAT_ENTRY(tx_heartbeat_errors
);
595 NETSTAT_ENTRY(tx_window_errors
);
596 NETSTAT_ENTRY(rx_compressed
);
597 NETSTAT_ENTRY(tx_compressed
);
598 NETSTAT_ENTRY(rx_nohandler
);
600 static struct attribute
*netstat_attrs
[] = {
601 &dev_attr_rx_packets
.attr
,
602 &dev_attr_tx_packets
.attr
,
603 &dev_attr_rx_bytes
.attr
,
604 &dev_attr_tx_bytes
.attr
,
605 &dev_attr_rx_errors
.attr
,
606 &dev_attr_tx_errors
.attr
,
607 &dev_attr_rx_dropped
.attr
,
608 &dev_attr_tx_dropped
.attr
,
609 &dev_attr_multicast
.attr
,
610 &dev_attr_collisions
.attr
,
611 &dev_attr_rx_length_errors
.attr
,
612 &dev_attr_rx_over_errors
.attr
,
613 &dev_attr_rx_crc_errors
.attr
,
614 &dev_attr_rx_frame_errors
.attr
,
615 &dev_attr_rx_fifo_errors
.attr
,
616 &dev_attr_rx_missed_errors
.attr
,
617 &dev_attr_tx_aborted_errors
.attr
,
618 &dev_attr_tx_carrier_errors
.attr
,
619 &dev_attr_tx_fifo_errors
.attr
,
620 &dev_attr_tx_heartbeat_errors
.attr
,
621 &dev_attr_tx_window_errors
.attr
,
622 &dev_attr_rx_compressed
.attr
,
623 &dev_attr_tx_compressed
.attr
,
624 &dev_attr_rx_nohandler
.attr
,
629 static const struct attribute_group netstat_group
= {
630 .name
= "statistics",
631 .attrs
= netstat_attrs
,
634 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
635 static struct attribute
*wireless_attrs
[] = {
639 static const struct attribute_group wireless_group
= {
641 .attrs
= wireless_attrs
,
645 #else /* CONFIG_SYSFS */
646 #define net_class_groups NULL
647 #endif /* CONFIG_SYSFS */
650 #define to_rx_queue_attr(_attr) container_of(_attr, \
651 struct rx_queue_attribute, attr)
653 #define to_rx_queue(obj) container_of(obj, struct netdev_rx_queue, kobj)
655 static ssize_t
rx_queue_attr_show(struct kobject
*kobj
, struct attribute
*attr
,
658 struct rx_queue_attribute
*attribute
= to_rx_queue_attr(attr
);
659 struct netdev_rx_queue
*queue
= to_rx_queue(kobj
);
661 if (!attribute
->show
)
664 return attribute
->show(queue
, attribute
, buf
);
667 static ssize_t
rx_queue_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
668 const char *buf
, size_t count
)
670 struct rx_queue_attribute
*attribute
= to_rx_queue_attr(attr
);
671 struct netdev_rx_queue
*queue
= to_rx_queue(kobj
);
673 if (!attribute
->store
)
676 return attribute
->store(queue
, attribute
, buf
, count
);
679 static const struct sysfs_ops rx_queue_sysfs_ops
= {
680 .show
= rx_queue_attr_show
,
681 .store
= rx_queue_attr_store
,
685 static ssize_t
show_rps_map(struct netdev_rx_queue
*queue
,
686 struct rx_queue_attribute
*attribute
, char *buf
)
692 if (!zalloc_cpumask_var(&mask
, GFP_KERNEL
))
696 map
= rcu_dereference(queue
->rps_map
);
698 for (i
= 0; i
< map
->len
; i
++)
699 cpumask_set_cpu(map
->cpus
[i
], mask
);
701 len
= snprintf(buf
, PAGE_SIZE
, "%*pb\n", cpumask_pr_args(mask
));
703 free_cpumask_var(mask
);
705 return len
< PAGE_SIZE
? len
: -EINVAL
;
708 static ssize_t
store_rps_map(struct netdev_rx_queue
*queue
,
709 struct rx_queue_attribute
*attribute
,
710 const char *buf
, size_t len
)
712 struct rps_map
*old_map
, *map
;
715 static DEFINE_MUTEX(rps_map_mutex
);
717 if (!capable(CAP_NET_ADMIN
))
720 if (!alloc_cpumask_var(&mask
, GFP_KERNEL
))
723 err
= bitmap_parse(buf
, len
, cpumask_bits(mask
), nr_cpumask_bits
);
725 free_cpumask_var(mask
);
729 map
= kzalloc(max_t(unsigned int,
730 RPS_MAP_SIZE(cpumask_weight(mask
)), L1_CACHE_BYTES
),
733 free_cpumask_var(mask
);
738 for_each_cpu_and(cpu
, mask
, cpu_online_mask
)
739 map
->cpus
[i
++] = cpu
;
748 mutex_lock(&rps_map_mutex
);
749 old_map
= rcu_dereference_protected(queue
->rps_map
,
750 mutex_is_locked(&rps_map_mutex
));
751 rcu_assign_pointer(queue
->rps_map
, map
);
754 static_key_slow_inc(&rps_needed
);
756 static_key_slow_dec(&rps_needed
);
758 mutex_unlock(&rps_map_mutex
);
761 kfree_rcu(old_map
, rcu
);
763 free_cpumask_var(mask
);
767 static ssize_t
show_rps_dev_flow_table_cnt(struct netdev_rx_queue
*queue
,
768 struct rx_queue_attribute
*attr
,
771 struct rps_dev_flow_table
*flow_table
;
772 unsigned long val
= 0;
775 flow_table
= rcu_dereference(queue
->rps_flow_table
);
777 val
= (unsigned long)flow_table
->mask
+ 1;
780 return sprintf(buf
, "%lu\n", val
);
783 static void rps_dev_flow_table_release(struct rcu_head
*rcu
)
785 struct rps_dev_flow_table
*table
= container_of(rcu
,
786 struct rps_dev_flow_table
, rcu
);
790 static ssize_t
store_rps_dev_flow_table_cnt(struct netdev_rx_queue
*queue
,
791 struct rx_queue_attribute
*attr
,
792 const char *buf
, size_t len
)
794 unsigned long mask
, count
;
795 struct rps_dev_flow_table
*table
, *old_table
;
796 static DEFINE_SPINLOCK(rps_dev_flow_lock
);
799 if (!capable(CAP_NET_ADMIN
))
802 rc
= kstrtoul(buf
, 0, &count
);
808 /* mask = roundup_pow_of_two(count) - 1;
809 * without overflows...
811 while ((mask
| (mask
>> 1)) != mask
)
813 /* On 64 bit arches, must check mask fits in table->mask (u32),
814 * and on 32bit arches, must check
815 * RPS_DEV_FLOW_TABLE_SIZE(mask + 1) doesn't overflow.
817 #if BITS_PER_LONG > 32
818 if (mask
> (unsigned long)(u32
)mask
)
821 if (mask
> (ULONG_MAX
- RPS_DEV_FLOW_TABLE_SIZE(1))
822 / sizeof(struct rps_dev_flow
)) {
823 /* Enforce a limit to prevent overflow */
827 table
= vmalloc(RPS_DEV_FLOW_TABLE_SIZE(mask
+ 1));
832 for (count
= 0; count
<= mask
; count
++)
833 table
->flows
[count
].cpu
= RPS_NO_CPU
;
837 spin_lock(&rps_dev_flow_lock
);
838 old_table
= rcu_dereference_protected(queue
->rps_flow_table
,
839 lockdep_is_held(&rps_dev_flow_lock
));
840 rcu_assign_pointer(queue
->rps_flow_table
, table
);
841 spin_unlock(&rps_dev_flow_lock
);
844 call_rcu(&old_table
->rcu
, rps_dev_flow_table_release
);
849 static struct rx_queue_attribute rps_cpus_attribute
=
850 __ATTR(rps_cpus
, S_IRUGO
| S_IWUSR
, show_rps_map
, store_rps_map
);
853 static struct rx_queue_attribute rps_dev_flow_table_cnt_attribute
=
854 __ATTR(rps_flow_cnt
, S_IRUGO
| S_IWUSR
,
855 show_rps_dev_flow_table_cnt
, store_rps_dev_flow_table_cnt
);
856 #endif /* CONFIG_RPS */
858 static struct attribute
*rx_queue_default_attrs
[] = {
860 &rps_cpus_attribute
.attr
,
861 &rps_dev_flow_table_cnt_attribute
.attr
,
866 static void rx_queue_release(struct kobject
*kobj
)
868 struct netdev_rx_queue
*queue
= to_rx_queue(kobj
);
871 struct rps_dev_flow_table
*flow_table
;
874 map
= rcu_dereference_protected(queue
->rps_map
, 1);
876 RCU_INIT_POINTER(queue
->rps_map
, NULL
);
880 flow_table
= rcu_dereference_protected(queue
->rps_flow_table
, 1);
882 RCU_INIT_POINTER(queue
->rps_flow_table
, NULL
);
883 call_rcu(&flow_table
->rcu
, rps_dev_flow_table_release
);
887 memset(kobj
, 0, sizeof(*kobj
));
891 static const void *rx_queue_namespace(struct kobject
*kobj
)
893 struct netdev_rx_queue
*queue
= to_rx_queue(kobj
);
894 struct device
*dev
= &queue
->dev
->dev
;
895 const void *ns
= NULL
;
897 if (dev
->class && dev
->class->ns_type
)
898 ns
= dev
->class->namespace(dev
);
903 static struct kobj_type rx_queue_ktype
= {
904 .sysfs_ops
= &rx_queue_sysfs_ops
,
905 .release
= rx_queue_release
,
906 .default_attrs
= rx_queue_default_attrs
,
907 .namespace = rx_queue_namespace
910 static int rx_queue_add_kobject(struct net_device
*dev
, int index
)
912 struct netdev_rx_queue
*queue
= dev
->_rx
+ index
;
913 struct kobject
*kobj
= &queue
->kobj
;
916 kobj
->kset
= dev
->queues_kset
;
917 error
= kobject_init_and_add(kobj
, &rx_queue_ktype
, NULL
,
922 if (dev
->sysfs_rx_queue_group
) {
923 error
= sysfs_create_group(kobj
, dev
->sysfs_rx_queue_group
);
928 kobject_uevent(kobj
, KOBJ_ADD
);
929 dev_hold(queue
->dev
);
936 #endif /* CONFIG_SYSFS */
939 net_rx_queue_update_kobjects(struct net_device
*dev
, int old_num
, int new_num
)
946 if (!dev
->sysfs_rx_queue_group
)
949 for (i
= old_num
; i
< new_num
; i
++) {
950 error
= rx_queue_add_kobject(dev
, i
);
957 while (--i
>= new_num
) {
958 struct kobject
*kobj
= &dev
->_rx
[i
].kobj
;
960 if (!atomic_read(&dev_net(dev
)->count
))
961 kobj
->uevent_suppress
= 1;
962 if (dev
->sysfs_rx_queue_group
)
963 sysfs_remove_group(kobj
, dev
->sysfs_rx_queue_group
);
975 * netdev_queue sysfs structures and functions.
977 struct netdev_queue_attribute
{
978 struct attribute attr
;
979 ssize_t (*show
)(struct netdev_queue
*queue
,
980 struct netdev_queue_attribute
*attr
, char *buf
);
981 ssize_t (*store
)(struct netdev_queue
*queue
,
982 struct netdev_queue_attribute
*attr
, const char *buf
, size_t len
);
984 #define to_netdev_queue_attr(_attr) container_of(_attr, \
985 struct netdev_queue_attribute, attr)
987 #define to_netdev_queue(obj) container_of(obj, struct netdev_queue, kobj)
989 static ssize_t
netdev_queue_attr_show(struct kobject
*kobj
,
990 struct attribute
*attr
, char *buf
)
992 struct netdev_queue_attribute
*attribute
= to_netdev_queue_attr(attr
);
993 struct netdev_queue
*queue
= to_netdev_queue(kobj
);
995 if (!attribute
->show
)
998 return attribute
->show(queue
, attribute
, buf
);
1001 static ssize_t
netdev_queue_attr_store(struct kobject
*kobj
,
1002 struct attribute
*attr
,
1003 const char *buf
, size_t count
)
1005 struct netdev_queue_attribute
*attribute
= to_netdev_queue_attr(attr
);
1006 struct netdev_queue
*queue
= to_netdev_queue(kobj
);
1008 if (!attribute
->store
)
1011 return attribute
->store(queue
, attribute
, buf
, count
);
1014 static const struct sysfs_ops netdev_queue_sysfs_ops
= {
1015 .show
= netdev_queue_attr_show
,
1016 .store
= netdev_queue_attr_store
,
1019 static ssize_t
show_trans_timeout(struct netdev_queue
*queue
,
1020 struct netdev_queue_attribute
*attribute
,
1023 unsigned long trans_timeout
;
1025 spin_lock_irq(&queue
->_xmit_lock
);
1026 trans_timeout
= queue
->trans_timeout
;
1027 spin_unlock_irq(&queue
->_xmit_lock
);
1029 return sprintf(buf
, "%lu", trans_timeout
);
1032 static unsigned int get_netdev_queue_index(struct netdev_queue
*queue
)
1034 struct net_device
*dev
= queue
->dev
;
1037 i
= queue
- dev
->_tx
;
1038 BUG_ON(i
>= dev
->num_tx_queues
);
1043 static ssize_t
show_traffic_class(struct netdev_queue
*queue
,
1044 struct netdev_queue_attribute
*attribute
,
1047 struct net_device
*dev
= queue
->dev
;
1048 int index
= get_netdev_queue_index(queue
);
1049 int tc
= netdev_txq_to_tc(dev
, index
);
1054 return sprintf(buf
, "%u\n", tc
);
1058 static ssize_t
show_tx_maxrate(struct netdev_queue
*queue
,
1059 struct netdev_queue_attribute
*attribute
,
1062 return sprintf(buf
, "%lu\n", queue
->tx_maxrate
);
1065 static ssize_t
set_tx_maxrate(struct netdev_queue
*queue
,
1066 struct netdev_queue_attribute
*attribute
,
1067 const char *buf
, size_t len
)
1069 struct net_device
*dev
= queue
->dev
;
1070 int err
, index
= get_netdev_queue_index(queue
);
1073 err
= kstrtou32(buf
, 10, &rate
);
1077 if (!rtnl_trylock())
1078 return restart_syscall();
1081 if (dev
->netdev_ops
->ndo_set_tx_maxrate
)
1082 err
= dev
->netdev_ops
->ndo_set_tx_maxrate(dev
, index
, rate
);
1086 queue
->tx_maxrate
= rate
;
1092 static struct netdev_queue_attribute queue_tx_maxrate
=
1093 __ATTR(tx_maxrate
, S_IRUGO
| S_IWUSR
,
1094 show_tx_maxrate
, set_tx_maxrate
);
1097 static struct netdev_queue_attribute queue_trans_timeout
=
1098 __ATTR(tx_timeout
, S_IRUGO
, show_trans_timeout
, NULL
);
1100 static struct netdev_queue_attribute queue_traffic_class
=
1101 __ATTR(traffic_class
, S_IRUGO
, show_traffic_class
, NULL
);
1105 * Byte queue limits sysfs structures and functions.
1107 static ssize_t
bql_show(char *buf
, unsigned int value
)
1109 return sprintf(buf
, "%u\n", value
);
1112 static ssize_t
bql_set(const char *buf
, const size_t count
,
1113 unsigned int *pvalue
)
1118 if (!strcmp(buf
, "max") || !strcmp(buf
, "max\n"))
1119 value
= DQL_MAX_LIMIT
;
1121 err
= kstrtouint(buf
, 10, &value
);
1124 if (value
> DQL_MAX_LIMIT
)
1133 static ssize_t
bql_show_hold_time(struct netdev_queue
*queue
,
1134 struct netdev_queue_attribute
*attr
,
1137 struct dql
*dql
= &queue
->dql
;
1139 return sprintf(buf
, "%u\n", jiffies_to_msecs(dql
->slack_hold_time
));
1142 static ssize_t
bql_set_hold_time(struct netdev_queue
*queue
,
1143 struct netdev_queue_attribute
*attribute
,
1144 const char *buf
, size_t len
)
1146 struct dql
*dql
= &queue
->dql
;
1150 err
= kstrtouint(buf
, 10, &value
);
1154 dql
->slack_hold_time
= msecs_to_jiffies(value
);
1159 static struct netdev_queue_attribute bql_hold_time_attribute
=
1160 __ATTR(hold_time
, S_IRUGO
| S_IWUSR
, bql_show_hold_time
,
1163 static ssize_t
bql_show_inflight(struct netdev_queue
*queue
,
1164 struct netdev_queue_attribute
*attr
,
1167 struct dql
*dql
= &queue
->dql
;
1169 return sprintf(buf
, "%u\n", dql
->num_queued
- dql
->num_completed
);
1172 static struct netdev_queue_attribute bql_inflight_attribute
=
1173 __ATTR(inflight
, S_IRUGO
, bql_show_inflight
, NULL
);
1175 #define BQL_ATTR(NAME, FIELD) \
1176 static ssize_t bql_show_ ## NAME(struct netdev_queue *queue, \
1177 struct netdev_queue_attribute *attr, \
1180 return bql_show(buf, queue->dql.FIELD); \
1183 static ssize_t bql_set_ ## NAME(struct netdev_queue *queue, \
1184 struct netdev_queue_attribute *attr, \
1185 const char *buf, size_t len) \
1187 return bql_set(buf, len, &queue->dql.FIELD); \
1190 static struct netdev_queue_attribute bql_ ## NAME ## _attribute = \
1191 __ATTR(NAME, S_IRUGO | S_IWUSR, bql_show_ ## NAME, \
1194 BQL_ATTR(limit
, limit
)
1195 BQL_ATTR(limit_max
, max_limit
)
1196 BQL_ATTR(limit_min
, min_limit
)
1198 static struct attribute
*dql_attrs
[] = {
1199 &bql_limit_attribute
.attr
,
1200 &bql_limit_max_attribute
.attr
,
1201 &bql_limit_min_attribute
.attr
,
1202 &bql_hold_time_attribute
.attr
,
1203 &bql_inflight_attribute
.attr
,
1207 static const struct attribute_group dql_group
= {
1208 .name
= "byte_queue_limits",
1211 #endif /* CONFIG_BQL */
1214 static ssize_t
show_xps_map(struct netdev_queue
*queue
,
1215 struct netdev_queue_attribute
*attribute
, char *buf
)
1217 struct net_device
*dev
= queue
->dev
;
1218 int cpu
, len
, num_tc
= 1, tc
= 0;
1219 struct xps_dev_maps
*dev_maps
;
1221 unsigned long index
;
1223 if (!zalloc_cpumask_var(&mask
, GFP_KERNEL
))
1226 index
= get_netdev_queue_index(queue
);
1229 num_tc
= dev
->num_tc
;
1230 tc
= netdev_txq_to_tc(dev
, index
);
1236 dev_maps
= rcu_dereference(dev
->xps_maps
);
1238 for_each_possible_cpu(cpu
) {
1239 int i
, tci
= cpu
* num_tc
+ tc
;
1240 struct xps_map
*map
;
1242 map
= rcu_dereference(dev_maps
->cpu_map
[tci
]);
1246 for (i
= map
->len
; i
--;) {
1247 if (map
->queues
[i
] == index
) {
1248 cpumask_set_cpu(cpu
, mask
);
1256 len
= snprintf(buf
, PAGE_SIZE
, "%*pb\n", cpumask_pr_args(mask
));
1257 free_cpumask_var(mask
);
1258 return len
< PAGE_SIZE
? len
: -EINVAL
;
1261 static ssize_t
store_xps_map(struct netdev_queue
*queue
,
1262 struct netdev_queue_attribute
*attribute
,
1263 const char *buf
, size_t len
)
1265 struct net_device
*dev
= queue
->dev
;
1266 unsigned long index
;
1270 if (!capable(CAP_NET_ADMIN
))
1273 if (!alloc_cpumask_var(&mask
, GFP_KERNEL
))
1276 index
= get_netdev_queue_index(queue
);
1278 err
= bitmap_parse(buf
, len
, cpumask_bits(mask
), nr_cpumask_bits
);
1280 free_cpumask_var(mask
);
1284 err
= netif_set_xps_queue(dev
, mask
, index
);
1286 free_cpumask_var(mask
);
1291 static struct netdev_queue_attribute xps_cpus_attribute
=
1292 __ATTR(xps_cpus
, S_IRUGO
| S_IWUSR
, show_xps_map
, store_xps_map
);
1293 #endif /* CONFIG_XPS */
1295 static struct attribute
*netdev_queue_default_attrs
[] = {
1296 &queue_trans_timeout
.attr
,
1297 &queue_traffic_class
.attr
,
1299 &xps_cpus_attribute
.attr
,
1300 &queue_tx_maxrate
.attr
,
1305 static void netdev_queue_release(struct kobject
*kobj
)
1307 struct netdev_queue
*queue
= to_netdev_queue(kobj
);
1309 memset(kobj
, 0, sizeof(*kobj
));
1310 dev_put(queue
->dev
);
1313 static const void *netdev_queue_namespace(struct kobject
*kobj
)
1315 struct netdev_queue
*queue
= to_netdev_queue(kobj
);
1316 struct device
*dev
= &queue
->dev
->dev
;
1317 const void *ns
= NULL
;
1319 if (dev
->class && dev
->class->ns_type
)
1320 ns
= dev
->class->namespace(dev
);
1325 static struct kobj_type netdev_queue_ktype
= {
1326 .sysfs_ops
= &netdev_queue_sysfs_ops
,
1327 .release
= netdev_queue_release
,
1328 .default_attrs
= netdev_queue_default_attrs
,
1329 .namespace = netdev_queue_namespace
,
1332 static int netdev_queue_add_kobject(struct net_device
*dev
, int index
)
1334 struct netdev_queue
*queue
= dev
->_tx
+ index
;
1335 struct kobject
*kobj
= &queue
->kobj
;
1338 kobj
->kset
= dev
->queues_kset
;
1339 error
= kobject_init_and_add(kobj
, &netdev_queue_ktype
, NULL
,
1345 error
= sysfs_create_group(kobj
, &dql_group
);
1350 kobject_uevent(kobj
, KOBJ_ADD
);
1351 dev_hold(queue
->dev
);
1358 #endif /* CONFIG_SYSFS */
1361 netdev_queue_update_kobjects(struct net_device
*dev
, int old_num
, int new_num
)
1367 for (i
= old_num
; i
< new_num
; i
++) {
1368 error
= netdev_queue_add_kobject(dev
, i
);
1375 while (--i
>= new_num
) {
1376 struct netdev_queue
*queue
= dev
->_tx
+ i
;
1378 if (!atomic_read(&dev_net(dev
)->count
))
1379 queue
->kobj
.uevent_suppress
= 1;
1381 sysfs_remove_group(&queue
->kobj
, &dql_group
);
1383 kobject_put(&queue
->kobj
);
1389 #endif /* CONFIG_SYSFS */
1392 static int register_queue_kobjects(struct net_device
*dev
)
1394 int error
= 0, txq
= 0, rxq
= 0, real_rx
= 0, real_tx
= 0;
1397 dev
->queues_kset
= kset_create_and_add("queues",
1398 NULL
, &dev
->dev
.kobj
);
1399 if (!dev
->queues_kset
)
1401 real_rx
= dev
->real_num_rx_queues
;
1403 real_tx
= dev
->real_num_tx_queues
;
1405 error
= net_rx_queue_update_kobjects(dev
, 0, real_rx
);
1410 error
= netdev_queue_update_kobjects(dev
, 0, real_tx
);
1418 netdev_queue_update_kobjects(dev
, txq
, 0);
1419 net_rx_queue_update_kobjects(dev
, rxq
, 0);
1423 static void remove_queue_kobjects(struct net_device
*dev
)
1425 int real_rx
= 0, real_tx
= 0;
1428 real_rx
= dev
->real_num_rx_queues
;
1430 real_tx
= dev
->real_num_tx_queues
;
1432 net_rx_queue_update_kobjects(dev
, real_rx
, 0);
1433 netdev_queue_update_kobjects(dev
, real_tx
, 0);
1435 kset_unregister(dev
->queues_kset
);
1439 static bool net_current_may_mount(void)
1441 struct net
*net
= current
->nsproxy
->net_ns
;
1443 return ns_capable(net
->user_ns
, CAP_SYS_ADMIN
);
1446 static void *net_grab_current_ns(void)
1448 struct net
*ns
= current
->nsproxy
->net_ns
;
1449 #ifdef CONFIG_NET_NS
1451 refcount_inc(&ns
->passive
);
1456 static const void *net_initial_ns(void)
1461 static const void *net_netlink_ns(struct sock
*sk
)
1463 return sock_net(sk
);
1466 struct kobj_ns_type_operations net_ns_type_operations
= {
1467 .type
= KOBJ_NS_TYPE_NET
,
1468 .current_may_mount
= net_current_may_mount
,
1469 .grab_current_ns
= net_grab_current_ns
,
1470 .netlink_ns
= net_netlink_ns
,
1471 .initial_ns
= net_initial_ns
,
1472 .drop_ns
= net_drop_ns
,
1474 EXPORT_SYMBOL_GPL(net_ns_type_operations
);
1476 static int netdev_uevent(struct device
*d
, struct kobj_uevent_env
*env
)
1478 struct net_device
*dev
= to_net_dev(d
);
1481 /* pass interface to uevent. */
1482 retval
= add_uevent_var(env
, "INTERFACE=%s", dev
->name
);
1486 /* pass ifindex to uevent.
1487 * ifindex is useful as it won't change (interface name may change)
1488 * and is what RtNetlink uses natively. */
1489 retval
= add_uevent_var(env
, "IFINDEX=%d", dev
->ifindex
);
1496 * netdev_release -- destroy and free a dead device.
1497 * Called when last reference to device kobject is gone.
1499 static void netdev_release(struct device
*d
)
1501 struct net_device
*dev
= to_net_dev(d
);
1503 BUG_ON(dev
->reg_state
!= NETREG_RELEASED
);
1505 kfree(dev
->ifalias
);
1506 netdev_freemem(dev
);
1509 static const void *net_namespace(struct device
*d
)
1511 struct net_device
*dev
= to_net_dev(d
);
1513 return dev_net(dev
);
1516 static struct class net_class
= {
1518 .dev_release
= netdev_release
,
1519 .dev_groups
= net_class_groups
,
1520 .dev_uevent
= netdev_uevent
,
1521 .ns_type
= &net_ns_type_operations
,
1522 .namespace = net_namespace
,
1525 #ifdef CONFIG_OF_NET
1526 static int of_dev_node_match(struct device
*dev
, const void *data
)
1531 ret
= dev
->parent
->of_node
== data
;
1533 return ret
== 0 ? dev
->of_node
== data
: ret
;
1537 * of_find_net_device_by_node - lookup the net device for the device node
1538 * @np: OF device node
1540 * Looks up the net_device structure corresponding with the device node.
1541 * If successful, returns a pointer to the net_device with the embedded
1542 * struct device refcount incremented by one, or NULL on failure. The
1543 * refcount must be dropped when done with the net_device.
1545 struct net_device
*of_find_net_device_by_node(struct device_node
*np
)
1549 dev
= class_find_device(&net_class
, NULL
, np
, of_dev_node_match
);
1553 return to_net_dev(dev
);
1555 EXPORT_SYMBOL(of_find_net_device_by_node
);
1558 /* Delete sysfs entries but hold kobject reference until after all
1559 * netdev references are gone.
1561 void netdev_unregister_kobject(struct net_device
*ndev
)
1563 struct device
*dev
= &(ndev
->dev
);
1565 if (!atomic_read(&dev_net(ndev
)->count
))
1566 dev_set_uevent_suppress(dev
, 1);
1568 kobject_get(&dev
->kobj
);
1570 remove_queue_kobjects(ndev
);
1572 pm_runtime_set_memalloc_noio(dev
, false);
1577 /* Create sysfs entries for network device. */
1578 int netdev_register_kobject(struct net_device
*ndev
)
1580 struct device
*dev
= &(ndev
->dev
);
1581 const struct attribute_group
**groups
= ndev
->sysfs_groups
;
1584 device_initialize(dev
);
1585 dev
->class = &net_class
;
1586 dev
->platform_data
= ndev
;
1587 dev
->groups
= groups
;
1589 dev_set_name(dev
, "%s", ndev
->name
);
1592 /* Allow for a device specific group */
1596 *groups
++ = &netstat_group
;
1598 #if IS_ENABLED(CONFIG_WIRELESS_EXT) || IS_ENABLED(CONFIG_CFG80211)
1599 if (ndev
->ieee80211_ptr
)
1600 *groups
++ = &wireless_group
;
1601 #if IS_ENABLED(CONFIG_WIRELESS_EXT)
1602 else if (ndev
->wireless_handlers
)
1603 *groups
++ = &wireless_group
;
1606 #endif /* CONFIG_SYSFS */
1608 error
= device_add(dev
);
1612 error
= register_queue_kobjects(ndev
);
1618 pm_runtime_set_memalloc_noio(dev
, true);
1623 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
1626 return class_create_file_ns(&net_class
, class_attr
, ns
);
1628 EXPORT_SYMBOL(netdev_class_create_file_ns
);
1630 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
1633 class_remove_file_ns(&net_class
, class_attr
, ns
);
1635 EXPORT_SYMBOL(netdev_class_remove_file_ns
);
1637 int __init
netdev_kobject_init(void)
1639 kobj_ns_type_register(&net_ns_type_operations
);
1640 return class_register(&net_class
);