Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 2a2a459eeeff48640dc557548ce576d666ab06ed

Authored by Eric Dumazet
Committed by David S. Miller
1 parent 3556485f15
Exists in smarc-l5.0.0_1.0.0-ga and in 5 other branches imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga

net: fix napi_reuse_skb() skb reserve 

napi->skb is allocated in napi_get_frags() using
netdev_alloc_skb_ip_align(), with a reserve of NET_SKB_PAD +
NET_IP_ALIGN bytes.

However, when such skb is recycled in napi_reuse_skb(), it ends with a
reserve of NET_IP_ALIGN which is suboptimal.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Showing 1 changed file with 2 additions and 1 deletions Inline Diff

1 /* 1 /*
2 * NET3 Protocol independent device support routines. 2 * NET3 Protocol independent device support routines.
3 * 3 *
4 * This program is free software; you can redistribute it and/or 4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License 5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version 6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version. 7 * 2 of the License, or (at your option) any later version.
8 * 8 *
9 * Derived from the non IP parts of dev.c 1.0.19 9 * Derived from the non IP parts of dev.c 1.0.19
10 * Authors: Ross Biro 10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Mark Evans, <evansmp@uhura.aston.ac.uk> 12 * Mark Evans, <evansmp@uhura.aston.ac.uk>
13 * 13 *
14 * Additional Authors: 14 * Additional Authors:
15 * Florian la Roche <rzsfl@rz.uni-sb.de> 15 * Florian la Roche <rzsfl@rz.uni-sb.de>
16 * Alan Cox <gw4pts@gw4pts.ampr.org> 16 * Alan Cox <gw4pts@gw4pts.ampr.org>
17 * David Hinds <dahinds@users.sourceforge.net> 17 * David Hinds <dahinds@users.sourceforge.net>
18 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 18 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
19 * Adam Sulmicki <adam@cfar.umd.edu> 19 * Adam Sulmicki <adam@cfar.umd.edu>
20 * Pekka Riikonen <priikone@poesidon.pspt.fi> 20 * Pekka Riikonen <priikone@poesidon.pspt.fi>
21 * 21 *
22 * Changes: 22 * Changes:
23 * D.J. Barrow : Fixed bug where dev->refcnt gets set 23 * D.J. Barrow : Fixed bug where dev->refcnt gets set
24 * to 2 if register_netdev gets called 24 * to 2 if register_netdev gets called
25 * before net_dev_init & also removed a 25 * before net_dev_init & also removed a
26 * few lines of code in the process. 26 * few lines of code in the process.
27 * Alan Cox : device private ioctl copies fields back. 27 * Alan Cox : device private ioctl copies fields back.
28 * Alan Cox : Transmit queue code does relevant 28 * Alan Cox : Transmit queue code does relevant
29 * stunts to keep the queue safe. 29 * stunts to keep the queue safe.
30 * Alan Cox : Fixed double lock. 30 * Alan Cox : Fixed double lock.
31 * Alan Cox : Fixed promisc NULL pointer trap 31 * Alan Cox : Fixed promisc NULL pointer trap
32 * ???????? : Support the full private ioctl range 32 * ???????? : Support the full private ioctl range
33 * Alan Cox : Moved ioctl permission check into 33 * Alan Cox : Moved ioctl permission check into
34 * drivers 34 * drivers
35 * Tim Kordas : SIOCADDMULTI/SIOCDELMULTI 35 * Tim Kordas : SIOCADDMULTI/SIOCDELMULTI
36 * Alan Cox : 100 backlog just doesn't cut it when 36 * Alan Cox : 100 backlog just doesn't cut it when
37 * you start doing multicast video 8) 37 * you start doing multicast video 8)
38 * Alan Cox : Rewrote net_bh and list manager. 38 * Alan Cox : Rewrote net_bh and list manager.
39 * Alan Cox : Fix ETH_P_ALL echoback lengths. 39 * Alan Cox : Fix ETH_P_ALL echoback lengths.
40 * Alan Cox : Took out transmit every packet pass 40 * Alan Cox : Took out transmit every packet pass
41 * Saved a few bytes in the ioctl handler 41 * Saved a few bytes in the ioctl handler
42 * Alan Cox : Network driver sets packet type before 42 * Alan Cox : Network driver sets packet type before
43 * calling netif_rx. Saves a function 43 * calling netif_rx. Saves a function
44 * call a packet. 44 * call a packet.
45 * Alan Cox : Hashed net_bh() 45 * Alan Cox : Hashed net_bh()
46 * Richard Kooijman: Timestamp fixes. 46 * Richard Kooijman: Timestamp fixes.
47 * Alan Cox : Wrong field in SIOCGIFDSTADDR 47 * Alan Cox : Wrong field in SIOCGIFDSTADDR
48 * Alan Cox : Device lock protection. 48 * Alan Cox : Device lock protection.
49 * Alan Cox : Fixed nasty side effect of device close 49 * Alan Cox : Fixed nasty side effect of device close
50 * changes. 50 * changes.
51 * Rudi Cilibrasi : Pass the right thing to 51 * Rudi Cilibrasi : Pass the right thing to
52 * set_mac_address() 52 * set_mac_address()
53 * Dave Miller : 32bit quantity for the device lock to 53 * Dave Miller : 32bit quantity for the device lock to
54 * make it work out on a Sparc. 54 * make it work out on a Sparc.
55 * Bjorn Ekwall : Added KERNELD hack. 55 * Bjorn Ekwall : Added KERNELD hack.
56 * Alan Cox : Cleaned up the backlog initialise. 56 * Alan Cox : Cleaned up the backlog initialise.
57 * Craig Metz : SIOCGIFCONF fix if space for under 57 * Craig Metz : SIOCGIFCONF fix if space for under
58 * 1 device. 58 * 1 device.
59 * Thomas Bogendoerfer : Return ENODEV for dev_open, if there 59 * Thomas Bogendoerfer : Return ENODEV for dev_open, if there
60 * is no device open function. 60 * is no device open function.
61 * Andi Kleen : Fix error reporting for SIOCGIFCONF 61 * Andi Kleen : Fix error reporting for SIOCGIFCONF
62 * Michael Chastain : Fix signed/unsigned for SIOCGIFCONF 62 * Michael Chastain : Fix signed/unsigned for SIOCGIFCONF
63 * Cyrus Durgin : Cleaned for KMOD 63 * Cyrus Durgin : Cleaned for KMOD
64 * Adam Sulmicki : Bug Fix : Network Device Unload 64 * Adam Sulmicki : Bug Fix : Network Device Unload
65 * A network device unload needs to purge 65 * A network device unload needs to purge
66 * the backlog queue. 66 * the backlog queue.
67 * Paul Rusty Russell : SIOCSIFNAME 67 * Paul Rusty Russell : SIOCSIFNAME
68 * Pekka Riikonen : Netdev boot-time settings code 68 * Pekka Riikonen : Netdev boot-time settings code
69 * Andrew Morton : Make unregister_netdevice wait 69 * Andrew Morton : Make unregister_netdevice wait
70 * indefinitely on dev->refcnt 70 * indefinitely on dev->refcnt
71 * J Hadi Salim : - Backlog queue sampling 71 * J Hadi Salim : - Backlog queue sampling
72 * - netif_rx() feedback 72 * - netif_rx() feedback
73 */ 73 */
74 74
75 #include <asm/uaccess.h> 75 #include <asm/uaccess.h>
76 #include <asm/system.h> 76 #include <asm/system.h>
77 #include <linux/bitops.h> 77 #include <linux/bitops.h>
78 #include <linux/capability.h> 78 #include <linux/capability.h>
79 #include <linux/cpu.h> 79 #include <linux/cpu.h>
80 #include <linux/types.h> 80 #include <linux/types.h>
81 #include <linux/kernel.h> 81 #include <linux/kernel.h>
82 #include <linux/hash.h> 82 #include <linux/hash.h>
83 #include <linux/slab.h> 83 #include <linux/slab.h>
84 #include <linux/sched.h> 84 #include <linux/sched.h>
85 #include <linux/mutex.h> 85 #include <linux/mutex.h>
86 #include <linux/string.h> 86 #include <linux/string.h>
87 #include <linux/mm.h> 87 #include <linux/mm.h>
88 #include <linux/socket.h> 88 #include <linux/socket.h>
89 #include <linux/sockios.h> 89 #include <linux/sockios.h>
90 #include <linux/errno.h> 90 #include <linux/errno.h>
91 #include <linux/interrupt.h> 91 #include <linux/interrupt.h>
92 #include <linux/if_ether.h> 92 #include <linux/if_ether.h>
93 #include <linux/netdevice.h> 93 #include <linux/netdevice.h>
94 #include <linux/etherdevice.h> 94 #include <linux/etherdevice.h>
95 #include <linux/ethtool.h> 95 #include <linux/ethtool.h>
96 #include <linux/notifier.h> 96 #include <linux/notifier.h>
97 #include <linux/skbuff.h> 97 #include <linux/skbuff.h>
98 #include <net/net_namespace.h> 98 #include <net/net_namespace.h>
99 #include <net/sock.h> 99 #include <net/sock.h>
100 #include <linux/rtnetlink.h> 100 #include <linux/rtnetlink.h>
101 #include <linux/proc_fs.h> 101 #include <linux/proc_fs.h>
102 #include <linux/seq_file.h> 102 #include <linux/seq_file.h>
103 #include <linux/stat.h> 103 #include <linux/stat.h>
104 #include <net/dst.h> 104 #include <net/dst.h>
105 #include <net/pkt_sched.h> 105 #include <net/pkt_sched.h>
106 #include <net/checksum.h> 106 #include <net/checksum.h>
107 #include <net/xfrm.h> 107 #include <net/xfrm.h>
108 #include <linux/highmem.h> 108 #include <linux/highmem.h>
109 #include <linux/init.h> 109 #include <linux/init.h>
110 #include <linux/kmod.h> 110 #include <linux/kmod.h>
111 #include <linux/module.h> 111 #include <linux/module.h>
112 #include <linux/netpoll.h> 112 #include <linux/netpoll.h>
113 #include <linux/rcupdate.h> 113 #include <linux/rcupdate.h>
114 #include <linux/delay.h> 114 #include <linux/delay.h>
115 #include <net/wext.h> 115 #include <net/wext.h>
116 #include <net/iw_handler.h> 116 #include <net/iw_handler.h>
117 #include <asm/current.h> 117 #include <asm/current.h>
118 #include <linux/audit.h> 118 #include <linux/audit.h>
119 #include <linux/dmaengine.h> 119 #include <linux/dmaengine.h>
120 #include <linux/err.h> 120 #include <linux/err.h>
121 #include <linux/ctype.h> 121 #include <linux/ctype.h>
122 #include <linux/if_arp.h> 122 #include <linux/if_arp.h>
123 #include <linux/if_vlan.h> 123 #include <linux/if_vlan.h>
124 #include <linux/ip.h> 124 #include <linux/ip.h>
125 #include <net/ip.h> 125 #include <net/ip.h>
126 #include <linux/ipv6.h> 126 #include <linux/ipv6.h>
127 #include <linux/in.h> 127 #include <linux/in.h>
128 #include <linux/jhash.h> 128 #include <linux/jhash.h>
129 #include <linux/random.h> 129 #include <linux/random.h>
130 #include <trace/events/napi.h> 130 #include <trace/events/napi.h>
131 #include <trace/events/net.h> 131 #include <trace/events/net.h>
132 #include <trace/events/skb.h> 132 #include <trace/events/skb.h>
133 #include <linux/pci.h> 133 #include <linux/pci.h>
134 #include <linux/inetdevice.h> 134 #include <linux/inetdevice.h>
135 #include <linux/cpu_rmap.h> 135 #include <linux/cpu_rmap.h>
136 #include <linux/net_tstamp.h> 136 #include <linux/net_tstamp.h>
137 #include <linux/static_key.h> 137 #include <linux/static_key.h>
138 #include <net/flow_keys.h> 138 #include <net/flow_keys.h>
139 139
140 #include "net-sysfs.h" 140 #include "net-sysfs.h"
141 141
142 /* Instead of increasing this, you should create a hash table. */ 142 /* Instead of increasing this, you should create a hash table. */
143 #define MAX_GRO_SKBS 8 143 #define MAX_GRO_SKBS 8
144 144
145 /* This should be increased if a protocol with a bigger head is added. */ 145 /* This should be increased if a protocol with a bigger head is added. */
146 #define GRO_MAX_HEAD (MAX_HEADER + 128) 146 #define GRO_MAX_HEAD (MAX_HEADER + 128)
147 147
148 /* 148 /*
149 * The list of packet types we will receive (as opposed to discard) 149 * The list of packet types we will receive (as opposed to discard)
150 * and the routines to invoke. 150 * and the routines to invoke.
151 * 151 *
152 * Why 16. Because with 16 the only overlap we get on a hash of the 152 * Why 16. Because with 16 the only overlap we get on a hash of the
153 * low nibble of the protocol value is RARP/SNAP/X.25. 153 * low nibble of the protocol value is RARP/SNAP/X.25.
154 * 154 *
155 * NOTE: That is no longer true with the addition of VLAN tags. Not 155 * NOTE: That is no longer true with the addition of VLAN tags. Not
156 * sure which should go first, but I bet it won't make much 156 * sure which should go first, but I bet it won't make much
157 * difference if we are running VLANs. The good news is that 157 * difference if we are running VLANs. The good news is that
158 * this protocol won't be in the list unless compiled in, so 158 * this protocol won't be in the list unless compiled in, so
159 * the average user (w/out VLANs) will not be adversely affected. 159 * the average user (w/out VLANs) will not be adversely affected.
160 * --BLG 160 * --BLG
161 * 161 *
162 * 0800 IP 162 * 0800 IP
163 * 8100 802.1Q VLAN 163 * 8100 802.1Q VLAN
164 * 0001 802.3 164 * 0001 802.3
165 * 0002 AX.25 165 * 0002 AX.25
166 * 0004 802.2 166 * 0004 802.2
167 * 8035 RARP 167 * 8035 RARP
168 * 0005 SNAP 168 * 0005 SNAP
169 * 0805 X.25 169 * 0805 X.25
170 * 0806 ARP 170 * 0806 ARP
171 * 8137 IPX 171 * 8137 IPX
172 * 0009 Localtalk 172 * 0009 Localtalk
173 * 86DD IPv6 173 * 86DD IPv6
174 */ 174 */
175 175
176 #define PTYPE_HASH_SIZE (16) 176 #define PTYPE_HASH_SIZE (16)
177 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 177 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
178 178
179 static DEFINE_SPINLOCK(ptype_lock); 179 static DEFINE_SPINLOCK(ptype_lock);
180 static struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 180 static struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
181 static struct list_head ptype_all __read_mostly; /* Taps */ 181 static struct list_head ptype_all __read_mostly; /* Taps */
182 182
183 /* 183 /*
184 * The @dev_base_head list is protected by @dev_base_lock and the rtnl 184 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
185 * semaphore. 185 * semaphore.
186 * 186 *
187 * Pure readers hold dev_base_lock for reading, or rcu_read_lock() 187 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
188 * 188 *
189 * Writers must hold the rtnl semaphore while they loop through the 189 * Writers must hold the rtnl semaphore while they loop through the
190 * dev_base_head list, and hold dev_base_lock for writing when they do the 190 * dev_base_head list, and hold dev_base_lock for writing when they do the
191 * actual updates. This allows pure readers to access the list even 191 * actual updates. This allows pure readers to access the list even
192 * while a writer is preparing to update it. 192 * while a writer is preparing to update it.
193 * 193 *
194 * To put it another way, dev_base_lock is held for writing only to 194 * To put it another way, dev_base_lock is held for writing only to
195 * protect against pure readers; the rtnl semaphore provides the 195 * protect against pure readers; the rtnl semaphore provides the
196 * protection against other writers. 196 * protection against other writers.
197 * 197 *
198 * See, for example usages, register_netdevice() and 198 * See, for example usages, register_netdevice() and
199 * unregister_netdevice(), which must be called with the rtnl 199 * unregister_netdevice(), which must be called with the rtnl
200 * semaphore held. 200 * semaphore held.
201 */ 201 */
202 DEFINE_RWLOCK(dev_base_lock); 202 DEFINE_RWLOCK(dev_base_lock);
203 EXPORT_SYMBOL(dev_base_lock); 203 EXPORT_SYMBOL(dev_base_lock);
204 204
205 static inline void dev_base_seq_inc(struct net *net) 205 static inline void dev_base_seq_inc(struct net *net)
206 { 206 {
207 while (++net->dev_base_seq == 0); 207 while (++net->dev_base_seq == 0);
208 } 208 }
209 209
210 static inline struct hlist_head *dev_name_hash(struct net *net, const char *name) 210 static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
211 { 211 {
212 unsigned hash = full_name_hash(name, strnlen(name, IFNAMSIZ)); 212 unsigned hash = full_name_hash(name, strnlen(name, IFNAMSIZ));
213 return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)]; 213 return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
214 } 214 }
215 215
216 static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex) 216 static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
217 { 217 {
218 return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)]; 218 return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
219 } 219 }
220 220
221 static inline void rps_lock(struct softnet_data *sd) 221 static inline void rps_lock(struct softnet_data *sd)
222 { 222 {
223 #ifdef CONFIG_RPS 223 #ifdef CONFIG_RPS
224 spin_lock(&sd->input_pkt_queue.lock); 224 spin_lock(&sd->input_pkt_queue.lock);
225 #endif 225 #endif
226 } 226 }
227 227
228 static inline void rps_unlock(struct softnet_data *sd) 228 static inline void rps_unlock(struct softnet_data *sd)
229 { 229 {
230 #ifdef CONFIG_RPS 230 #ifdef CONFIG_RPS
231 spin_unlock(&sd->input_pkt_queue.lock); 231 spin_unlock(&sd->input_pkt_queue.lock);
232 #endif 232 #endif
233 } 233 }
234 234
235 /* Device list insertion */ 235 /* Device list insertion */
236 static int list_netdevice(struct net_device *dev) 236 static int list_netdevice(struct net_device *dev)
237 { 237 {
238 struct net *net = dev_net(dev); 238 struct net *net = dev_net(dev);
239 239
240 ASSERT_RTNL(); 240 ASSERT_RTNL();
241 241
242 write_lock_bh(&dev_base_lock); 242 write_lock_bh(&dev_base_lock);
243 list_add_tail_rcu(&dev->dev_list, &net->dev_base_head); 243 list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
244 hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name)); 244 hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
245 hlist_add_head_rcu(&dev->index_hlist, 245 hlist_add_head_rcu(&dev->index_hlist,
246 dev_index_hash(net, dev->ifindex)); 246 dev_index_hash(net, dev->ifindex));
247 write_unlock_bh(&dev_base_lock); 247 write_unlock_bh(&dev_base_lock);
248 248
249 dev_base_seq_inc(net); 249 dev_base_seq_inc(net);
250 250
251 return 0; 251 return 0;
252 } 252 }
253 253
254 /* Device list removal 254 /* Device list removal
255 * caller must respect a RCU grace period before freeing/reusing dev 255 * caller must respect a RCU grace period before freeing/reusing dev
256 */ 256 */
257 static void unlist_netdevice(struct net_device *dev) 257 static void unlist_netdevice(struct net_device *dev)
258 { 258 {
259 ASSERT_RTNL(); 259 ASSERT_RTNL();
260 260
261 /* Unlink dev from the device chain */ 261 /* Unlink dev from the device chain */
262 write_lock_bh(&dev_base_lock); 262 write_lock_bh(&dev_base_lock);
263 list_del_rcu(&dev->dev_list); 263 list_del_rcu(&dev->dev_list);
264 hlist_del_rcu(&dev->name_hlist); 264 hlist_del_rcu(&dev->name_hlist);
265 hlist_del_rcu(&dev->index_hlist); 265 hlist_del_rcu(&dev->index_hlist);
266 write_unlock_bh(&dev_base_lock); 266 write_unlock_bh(&dev_base_lock);
267 267
268 dev_base_seq_inc(dev_net(dev)); 268 dev_base_seq_inc(dev_net(dev));
269 } 269 }
270 270
271 /* 271 /*
272 * Our notifier list 272 * Our notifier list
273 */ 273 */
274 274
275 static RAW_NOTIFIER_HEAD(netdev_chain); 275 static RAW_NOTIFIER_HEAD(netdev_chain);
276 276
277 /* 277 /*
278 * Device drivers call our routines to queue packets here. We empty the 278 * Device drivers call our routines to queue packets here. We empty the
279 * queue in the local softnet handler. 279 * queue in the local softnet handler.
280 */ 280 */
281 281
282 DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 282 DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
283 EXPORT_PER_CPU_SYMBOL(softnet_data); 283 EXPORT_PER_CPU_SYMBOL(softnet_data);
284 284
285 #ifdef CONFIG_LOCKDEP 285 #ifdef CONFIG_LOCKDEP
286 /* 286 /*
287 * register_netdevice() inits txq->_xmit_lock and sets lockdep class 287 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
288 * according to dev->type 288 * according to dev->type
289 */ 289 */
290 static const unsigned short netdev_lock_type[] = 290 static const unsigned short netdev_lock_type[] =
291 {ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25, 291 {ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25,
292 ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET, 292 ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET,
293 ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM, 293 ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM,
294 ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP, 294 ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP,
295 ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD, 295 ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD,
296 ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25, 296 ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25,
297 ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP, 297 ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP,
298 ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD, 298 ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD,
299 ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI, 299 ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI,
300 ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE, 300 ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE,
301 ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET, 301 ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET,
302 ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL, 302 ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL,
303 ARPHRD_FCFABRIC, ARPHRD_IEEE802_TR, ARPHRD_IEEE80211, 303 ARPHRD_FCFABRIC, ARPHRD_IEEE802_TR, ARPHRD_IEEE80211,
304 ARPHRD_IEEE80211_PRISM, ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, 304 ARPHRD_IEEE80211_PRISM, ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET,
305 ARPHRD_PHONET_PIPE, ARPHRD_IEEE802154, 305 ARPHRD_PHONET_PIPE, ARPHRD_IEEE802154,
306 ARPHRD_VOID, ARPHRD_NONE}; 306 ARPHRD_VOID, ARPHRD_NONE};
307 307
308 static const char *const netdev_lock_name[] = 308 static const char *const netdev_lock_name[] =
309 {"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25", 309 {"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25",
310 "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET", 310 "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET",
311 "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM", 311 "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM",
312 "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP", 312 "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP",
313 "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD", 313 "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD",
314 "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25", 314 "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25",
315 "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP", 315 "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP",
316 "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD", 316 "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD",
317 "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI", 317 "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI",
318 "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE", 318 "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE",
319 "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET", 319 "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET",
320 "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL", 320 "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL",
321 "_xmit_FCFABRIC", "_xmit_IEEE802_TR", "_xmit_IEEE80211", 321 "_xmit_FCFABRIC", "_xmit_IEEE802_TR", "_xmit_IEEE80211",
322 "_xmit_IEEE80211_PRISM", "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", 322 "_xmit_IEEE80211_PRISM", "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET",
323 "_xmit_PHONET_PIPE", "_xmit_IEEE802154", 323 "_xmit_PHONET_PIPE", "_xmit_IEEE802154",
324 "_xmit_VOID", "_xmit_NONE"}; 324 "_xmit_VOID", "_xmit_NONE"};
325 325
326 static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)]; 326 static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
327 static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)]; 327 static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
328 328
329 static inline unsigned short netdev_lock_pos(unsigned short dev_type) 329 static inline unsigned short netdev_lock_pos(unsigned short dev_type)
330 { 330 {
331 int i; 331 int i;
332 332
333 for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++) 333 for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++)
334 if (netdev_lock_type[i] == dev_type) 334 if (netdev_lock_type[i] == dev_type)
335 return i; 335 return i;
336 /* the last key is used by default */ 336 /* the last key is used by default */
337 return ARRAY_SIZE(netdev_lock_type) - 1; 337 return ARRAY_SIZE(netdev_lock_type) - 1;
338 } 338 }
339 339
340 static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, 340 static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
341 unsigned short dev_type) 341 unsigned short dev_type)
342 { 342 {
343 int i; 343 int i;
344 344
345 i = netdev_lock_pos(dev_type); 345 i = netdev_lock_pos(dev_type);
346 lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i], 346 lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
347 netdev_lock_name[i]); 347 netdev_lock_name[i]);
348 } 348 }
349 349
350 static inline void netdev_set_addr_lockdep_class(struct net_device *dev) 350 static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
351 { 351 {
352 int i; 352 int i;
353 353
354 i = netdev_lock_pos(dev->type); 354 i = netdev_lock_pos(dev->type);
355 lockdep_set_class_and_name(&dev->addr_list_lock, 355 lockdep_set_class_and_name(&dev->addr_list_lock,
356 &netdev_addr_lock_key[i], 356 &netdev_addr_lock_key[i],
357 netdev_lock_name[i]); 357 netdev_lock_name[i]);
358 } 358 }
359 #else 359 #else
360 static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock, 360 static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
361 unsigned short dev_type) 361 unsigned short dev_type)
362 { 362 {
363 } 363 }
364 static inline void netdev_set_addr_lockdep_class(struct net_device *dev) 364 static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
365 { 365 {
366 } 366 }
367 #endif 367 #endif
368 368
369 /******************************************************************************* 369 /*******************************************************************************
370 370
371 Protocol management and registration routines 371 Protocol management and registration routines
372 372
373 *******************************************************************************/ 373 *******************************************************************************/
374 374
375 /* 375 /*
376 * Add a protocol ID to the list. Now that the input handler is 376 * Add a protocol ID to the list. Now that the input handler is
377 * smarter we can dispense with all the messy stuff that used to be 377 * smarter we can dispense with all the messy stuff that used to be
378 * here. 378 * here.
379 * 379 *
380 * BEWARE!!! Protocol handlers, mangling input packets, 380 * BEWARE!!! Protocol handlers, mangling input packets,
381 * MUST BE last in hash buckets and checking protocol handlers 381 * MUST BE last in hash buckets and checking protocol handlers
382 * MUST start from promiscuous ptype_all chain in net_bh. 382 * MUST start from promiscuous ptype_all chain in net_bh.
383 * It is true now, do not change it. 383 * It is true now, do not change it.
384 * Explanation follows: if protocol handler, mangling packet, will 384 * Explanation follows: if protocol handler, mangling packet, will
385 * be the first on list, it is not able to sense, that packet 385 * be the first on list, it is not able to sense, that packet
386 * is cloned and should be copied-on-write, so that it will 386 * is cloned and should be copied-on-write, so that it will
387 * change it and subsequent readers will get broken packet. 387 * change it and subsequent readers will get broken packet.
388 * --ANK (980803) 388 * --ANK (980803)
389 */ 389 */
390 390
391 static inline struct list_head *ptype_head(const struct packet_type *pt) 391 static inline struct list_head *ptype_head(const struct packet_type *pt)
392 { 392 {
393 if (pt->type == htons(ETH_P_ALL)) 393 if (pt->type == htons(ETH_P_ALL))
394 return &ptype_all; 394 return &ptype_all;
395 else 395 else
396 return &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK]; 396 return &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
397 } 397 }
398 398
399 /** 399 /**
400 * dev_add_pack - add packet handler 400 * dev_add_pack - add packet handler
401 * @pt: packet type declaration 401 * @pt: packet type declaration
402 * 402 *
403 * Add a protocol handler to the networking stack. The passed &packet_type 403 * Add a protocol handler to the networking stack. The passed &packet_type
404 * is linked into kernel lists and may not be freed until it has been 404 * is linked into kernel lists and may not be freed until it has been
405 * removed from the kernel lists. 405 * removed from the kernel lists.
406 * 406 *
407 * This call does not sleep therefore it can not 407 * This call does not sleep therefore it can not
408 * guarantee all CPU's that are in middle of receiving packets 408 * guarantee all CPU's that are in middle of receiving packets
409 * will see the new packet type (until the next received packet). 409 * will see the new packet type (until the next received packet).
410 */ 410 */
411 411
412 void dev_add_pack(struct packet_type *pt) 412 void dev_add_pack(struct packet_type *pt)
413 { 413 {
414 struct list_head *head = ptype_head(pt); 414 struct list_head *head = ptype_head(pt);
415 415
416 spin_lock(&ptype_lock); 416 spin_lock(&ptype_lock);
417 list_add_rcu(&pt->list, head); 417 list_add_rcu(&pt->list, head);
418 spin_unlock(&ptype_lock); 418 spin_unlock(&ptype_lock);
419 } 419 }
420 EXPORT_SYMBOL(dev_add_pack); 420 EXPORT_SYMBOL(dev_add_pack);
421 421
422 /** 422 /**
423 * __dev_remove_pack - remove packet handler 423 * __dev_remove_pack - remove packet handler
424 * @pt: packet type declaration 424 * @pt: packet type declaration
425 * 425 *
426 * Remove a protocol handler that was previously added to the kernel 426 * Remove a protocol handler that was previously added to the kernel
427 * protocol handlers by dev_add_pack(). The passed &packet_type is removed 427 * protocol handlers by dev_add_pack(). The passed &packet_type is removed
428 * from the kernel lists and can be freed or reused once this function 428 * from the kernel lists and can be freed or reused once this function
429 * returns. 429 * returns.
430 * 430 *
431 * The packet type might still be in use by receivers 431 * The packet type might still be in use by receivers
432 * and must not be freed until after all the CPU's have gone 432 * and must not be freed until after all the CPU's have gone
433 * through a quiescent state. 433 * through a quiescent state.
434 */ 434 */
435 void __dev_remove_pack(struct packet_type *pt) 435 void __dev_remove_pack(struct packet_type *pt)
436 { 436 {
437 struct list_head *head = ptype_head(pt); 437 struct list_head *head = ptype_head(pt);
438 struct packet_type *pt1; 438 struct packet_type *pt1;
439 439
440 spin_lock(&ptype_lock); 440 spin_lock(&ptype_lock);
441 441
442 list_for_each_entry(pt1, head, list) { 442 list_for_each_entry(pt1, head, list) {
443 if (pt == pt1) { 443 if (pt == pt1) {
444 list_del_rcu(&pt->list); 444 list_del_rcu(&pt->list);
445 goto out; 445 goto out;
446 } 446 }
447 } 447 }
448 448
449 pr_warn("dev_remove_pack: %p not found\n", pt); 449 pr_warn("dev_remove_pack: %p not found\n", pt);
450 out: 450 out:
451 spin_unlock(&ptype_lock); 451 spin_unlock(&ptype_lock);
452 } 452 }
453 EXPORT_SYMBOL(__dev_remove_pack); 453 EXPORT_SYMBOL(__dev_remove_pack);
454 454
455 /** 455 /**
456 * dev_remove_pack - remove packet handler 456 * dev_remove_pack - remove packet handler
457 * @pt: packet type declaration 457 * @pt: packet type declaration
458 * 458 *
459 * Remove a protocol handler that was previously added to the kernel 459 * Remove a protocol handler that was previously added to the kernel
460 * protocol handlers by dev_add_pack(). The passed &packet_type is removed 460 * protocol handlers by dev_add_pack(). The passed &packet_type is removed
461 * from the kernel lists and can be freed or reused once this function 461 * from the kernel lists and can be freed or reused once this function
462 * returns. 462 * returns.
463 * 463 *
464 * This call sleeps to guarantee that no CPU is looking at the packet 464 * This call sleeps to guarantee that no CPU is looking at the packet
465 * type after return. 465 * type after return.
466 */ 466 */
467 void dev_remove_pack(struct packet_type *pt) 467 void dev_remove_pack(struct packet_type *pt)
468 { 468 {
469 __dev_remove_pack(pt); 469 __dev_remove_pack(pt);
470 470
471 synchronize_net(); 471 synchronize_net();
472 } 472 }
473 EXPORT_SYMBOL(dev_remove_pack); 473 EXPORT_SYMBOL(dev_remove_pack);
474 474
475 /****************************************************************************** 475 /******************************************************************************
476 476
477 Device Boot-time Settings Routines 477 Device Boot-time Settings Routines
478 478
479 *******************************************************************************/ 479 *******************************************************************************/
480 480
481 /* Boot time configuration table */ 481 /* Boot time configuration table */
482 static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX]; 482 static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX];
483 483
484 /** 484 /**
485 * netdev_boot_setup_add - add new setup entry 485 * netdev_boot_setup_add - add new setup entry
486 * @name: name of the device 486 * @name: name of the device
487 * @map: configured settings for the device 487 * @map: configured settings for the device
488 * 488 *
489 * Adds new setup entry to the dev_boot_setup list. The function 489 * Adds new setup entry to the dev_boot_setup list. The function
490 * returns 0 on error and 1 on success. This is a generic routine to 490 * returns 0 on error and 1 on success. This is a generic routine to
491 * all netdevices. 491 * all netdevices.
492 */ 492 */
493 static int netdev_boot_setup_add(char *name, struct ifmap *map) 493 static int netdev_boot_setup_add(char *name, struct ifmap *map)
494 { 494 {
495 struct netdev_boot_setup *s; 495 struct netdev_boot_setup *s;
496 int i; 496 int i;
497 497
498 s = dev_boot_setup; 498 s = dev_boot_setup;
499 for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { 499 for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) {
500 if (s[i].name[0] == '\0' || s[i].name[0] == ' ') { 500 if (s[i].name[0] == '\0' || s[i].name[0] == ' ') {
501 memset(s[i].name, 0, sizeof(s[i].name)); 501 memset(s[i].name, 0, sizeof(s[i].name));
502 strlcpy(s[i].name, name, IFNAMSIZ); 502 strlcpy(s[i].name, name, IFNAMSIZ);
503 memcpy(&s[i].map, map, sizeof(s[i].map)); 503 memcpy(&s[i].map, map, sizeof(s[i].map));
504 break; 504 break;
505 } 505 }
506 } 506 }
507 507
508 return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1; 508 return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1;
509 } 509 }
510 510
511 /** 511 /**
512 * netdev_boot_setup_check - check boot time settings 512 * netdev_boot_setup_check - check boot time settings
513 * @dev: the netdevice 513 * @dev: the netdevice
514 * 514 *
515 * Check boot time settings for the device. 515 * Check boot time settings for the device.
516 * The found settings are set for the device to be used 516 * The found settings are set for the device to be used
517 * later in the device probing. 517 * later in the device probing.
518 * Returns 0 if no settings found, 1 if they are. 518 * Returns 0 if no settings found, 1 if they are.
519 */ 519 */
520 int netdev_boot_setup_check(struct net_device *dev) 520 int netdev_boot_setup_check(struct net_device *dev)
521 { 521 {
522 struct netdev_boot_setup *s = dev_boot_setup; 522 struct netdev_boot_setup *s = dev_boot_setup;
523 int i; 523 int i;
524 524
525 for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) { 525 for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) {
526 if (s[i].name[0] != '\0' && s[i].name[0] != ' ' && 526 if (s[i].name[0] != '\0' && s[i].name[0] != ' ' &&
527 !strcmp(dev->name, s[i].name)) { 527 !strcmp(dev->name, s[i].name)) {
528 dev->irq = s[i].map.irq; 528 dev->irq = s[i].map.irq;
529 dev->base_addr = s[i].map.base_addr; 529 dev->base_addr = s[i].map.base_addr;
530 dev->mem_start = s[i].map.mem_start; 530 dev->mem_start = s[i].map.mem_start;
531 dev->mem_end = s[i].map.mem_end; 531 dev->mem_end = s[i].map.mem_end;
532 return 1; 532 return 1;
533 } 533 }
534 } 534 }
535 return 0; 535 return 0;
536 } 536 }
537 EXPORT_SYMBOL(netdev_boot_setup_check); 537 EXPORT_SYMBOL(netdev_boot_setup_check);
538 538
539 539
540 /** 540 /**
541 * netdev_boot_base - get address from boot time settings 541 * netdev_boot_base - get address from boot time settings
542 * @prefix: prefix for network device 542 * @prefix: prefix for network device
543 * @unit: id for network device 543 * @unit: id for network device
544 * 544 *
545 * Check boot time settings for the base address of device. 545 * Check boot time settings for the base address of device.
546 * The found settings are set for the device to be used 546 * The found settings are set for the device to be used
547 * later in the device probing. 547 * later in the device probing.
548 * Returns 0 if no settings found. 548 * Returns 0 if no settings found.
549 */ 549 */
550 unsigned long netdev_boot_base(const char *prefix, int unit) 550 unsigned long netdev_boot_base(const char *prefix, int unit)
551 { 551 {
552 const struct netdev_boot_setup *s = dev_boot_setup; 552 const struct netdev_boot_setup *s = dev_boot_setup;
553 char name[IFNAMSIZ]; 553 char name[IFNAMSIZ];
554 int i; 554 int i;
555 555
556 sprintf(name, "%s%d", prefix, unit); 556 sprintf(name, "%s%d", prefix, unit);
557 557
558 /* 558 /*
559 * If device already registered then return base of 1 559 * If device already registered then return base of 1
560 * to indicate not to probe for this interface 560 * to indicate not to probe for this interface
561 */ 561 */
562 if (__dev_get_by_name(&init_net, name)) 562 if (__dev_get_by_name(&init_net, name))
563 return 1; 563 return 1;
564 564
565 for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) 565 for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++)
566 if (!strcmp(name, s[i].name)) 566 if (!strcmp(name, s[i].name))
567 return s[i].map.base_addr; 567 return s[i].map.base_addr;
568 return 0; 568 return 0;
569 } 569 }
570 570
571 /* 571 /*
572 * Saves at boot time configured settings for any netdevice. 572 * Saves at boot time configured settings for any netdevice.
573 */ 573 */
574 int __init netdev_boot_setup(char *str) 574 int __init netdev_boot_setup(char *str)
575 { 575 {
576 int ints[5]; 576 int ints[5];
577 struct ifmap map; 577 struct ifmap map;
578 578
579 str = get_options(str, ARRAY_SIZE(ints), ints); 579 str = get_options(str, ARRAY_SIZE(ints), ints);
580 if (!str || !*str) 580 if (!str || !*str)
581 return 0; 581 return 0;
582 582
583 /* Save settings */ 583 /* Save settings */
584 memset(&map, 0, sizeof(map)); 584 memset(&map, 0, sizeof(map));
585 if (ints[0] > 0) 585 if (ints[0] > 0)
586 map.irq = ints[1]; 586 map.irq = ints[1];
587 if (ints[0] > 1) 587 if (ints[0] > 1)
588 map.base_addr = ints[2]; 588 map.base_addr = ints[2];
589 if (ints[0] > 2) 589 if (ints[0] > 2)
590 map.mem_start = ints[3]; 590 map.mem_start = ints[3];
591 if (ints[0] > 3) 591 if (ints[0] > 3)
592 map.mem_end = ints[4]; 592 map.mem_end = ints[4];
593 593
594 /* Add new entry to the list */ 594 /* Add new entry to the list */
595 return netdev_boot_setup_add(str, &map); 595 return netdev_boot_setup_add(str, &map);
596 } 596 }
597 597
598 __setup("netdev=", netdev_boot_setup); 598 __setup("netdev=", netdev_boot_setup);
599 599
600 /******************************************************************************* 600 /*******************************************************************************
601 601
602 Device Interface Subroutines 602 Device Interface Subroutines
603 603
604 *******************************************************************************/ 604 *******************************************************************************/
605 605
606 /** 606 /**
607 * __dev_get_by_name - find a device by its name 607 * __dev_get_by_name - find a device by its name
608 * @net: the applicable net namespace 608 * @net: the applicable net namespace
609 * @name: name to find 609 * @name: name to find
610 * 610 *
611 * Find an interface by name. Must be called under RTNL semaphore 611 * Find an interface by name. Must be called under RTNL semaphore
612 * or @dev_base_lock. If the name is found a pointer to the device 612 * or @dev_base_lock. If the name is found a pointer to the device
613 * is returned. If the name is not found then %NULL is returned. The 613 * is returned. If the name is not found then %NULL is returned. The
614 * reference counters are not incremented so the caller must be 614 * reference counters are not incremented so the caller must be
615 * careful with locks. 615 * careful with locks.
616 */ 616 */
617 617
618 struct net_device *__dev_get_by_name(struct net *net, const char *name) 618 struct net_device *__dev_get_by_name(struct net *net, const char *name)
619 { 619 {
620 struct hlist_node *p; 620 struct hlist_node *p;
621 struct net_device *dev; 621 struct net_device *dev;
622 struct hlist_head *head = dev_name_hash(net, name); 622 struct hlist_head *head = dev_name_hash(net, name);
623 623
624 hlist_for_each_entry(dev, p, head, name_hlist) 624 hlist_for_each_entry(dev, p, head, name_hlist)
625 if (!strncmp(dev->name, name, IFNAMSIZ)) 625 if (!strncmp(dev->name, name, IFNAMSIZ))
626 return dev; 626 return dev;
627 627
628 return NULL; 628 return NULL;
629 } 629 }
630 EXPORT_SYMBOL(__dev_get_by_name); 630 EXPORT_SYMBOL(__dev_get_by_name);
631 631
632 /** 632 /**
633 * dev_get_by_name_rcu - find a device by its name 633 * dev_get_by_name_rcu - find a device by its name
634 * @net: the applicable net namespace 634 * @net: the applicable net namespace
635 * @name: name to find 635 * @name: name to find
636 * 636 *
637 * Find an interface by name. 637 * Find an interface by name.
638 * If the name is found a pointer to the device is returned. 638 * If the name is found a pointer to the device is returned.
639 * If the name is not found then %NULL is returned. 639 * If the name is not found then %NULL is returned.
640 * The reference counters are not incremented so the caller must be 640 * The reference counters are not incremented so the caller must be
641 * careful with locks. The caller must hold RCU lock. 641 * careful with locks. The caller must hold RCU lock.
642 */ 642 */
643 643
644 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name) 644 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name)
645 { 645 {
646 struct hlist_node *p; 646 struct hlist_node *p;
647 struct net_device *dev; 647 struct net_device *dev;
648 struct hlist_head *head = dev_name_hash(net, name); 648 struct hlist_head *head = dev_name_hash(net, name);
649 649
650 hlist_for_each_entry_rcu(dev, p, head, name_hlist) 650 hlist_for_each_entry_rcu(dev, p, head, name_hlist)
651 if (!strncmp(dev->name, name, IFNAMSIZ)) 651 if (!strncmp(dev->name, name, IFNAMSIZ))
652 return dev; 652 return dev;
653 653
654 return NULL; 654 return NULL;
655 } 655 }
656 EXPORT_SYMBOL(dev_get_by_name_rcu); 656 EXPORT_SYMBOL(dev_get_by_name_rcu);
657 657
658 /** 658 /**
659 * dev_get_by_name - find a device by its name 659 * dev_get_by_name - find a device by its name
660 * @net: the applicable net namespace 660 * @net: the applicable net namespace
661 * @name: name to find 661 * @name: name to find
662 * 662 *
663 * Find an interface by name. This can be called from any 663 * Find an interface by name. This can be called from any
664 * context and does its own locking. The returned handle has 664 * context and does its own locking. The returned handle has
665 * the usage count incremented and the caller must use dev_put() to 665 * the usage count incremented and the caller must use dev_put() to
666 * release it when it is no longer needed. %NULL is returned if no 666 * release it when it is no longer needed. %NULL is returned if no
667 * matching device is found. 667 * matching device is found.
668 */ 668 */
669 669
670 struct net_device *dev_get_by_name(struct net *net, const char *name) 670 struct net_device *dev_get_by_name(struct net *net, const char *name)
671 { 671 {
672 struct net_device *dev; 672 struct net_device *dev;
673 673
674 rcu_read_lock(); 674 rcu_read_lock();
675 dev = dev_get_by_name_rcu(net, name); 675 dev = dev_get_by_name_rcu(net, name);
676 if (dev) 676 if (dev)
677 dev_hold(dev); 677 dev_hold(dev);
678 rcu_read_unlock(); 678 rcu_read_unlock();
679 return dev; 679 return dev;
680 } 680 }
681 EXPORT_SYMBOL(dev_get_by_name); 681 EXPORT_SYMBOL(dev_get_by_name);
682 682
683 /** 683 /**
684 * __dev_get_by_index - find a device by its ifindex 684 * __dev_get_by_index - find a device by its ifindex
685 * @net: the applicable net namespace 685 * @net: the applicable net namespace
686 * @ifindex: index of device 686 * @ifindex: index of device
687 * 687 *
688 * Search for an interface by index. Returns %NULL if the device 688 * Search for an interface by index. Returns %NULL if the device
689 * is not found or a pointer to the device. The device has not 689 * is not found or a pointer to the device. The device has not
690 * had its reference counter increased so the caller must be careful 690 * had its reference counter increased so the caller must be careful
691 * about locking. The caller must hold either the RTNL semaphore 691 * about locking. The caller must hold either the RTNL semaphore
692 * or @dev_base_lock. 692 * or @dev_base_lock.
693 */ 693 */
694 694
695 struct net_device *__dev_get_by_index(struct net *net, int ifindex) 695 struct net_device *__dev_get_by_index(struct net *net, int ifindex)
696 { 696 {
697 struct hlist_node *p; 697 struct hlist_node *p;
698 struct net_device *dev; 698 struct net_device *dev;
699 struct hlist_head *head = dev_index_hash(net, ifindex); 699 struct hlist_head *head = dev_index_hash(net, ifindex);
700 700
701 hlist_for_each_entry(dev, p, head, index_hlist) 701 hlist_for_each_entry(dev, p, head, index_hlist)
702 if (dev->ifindex == ifindex) 702 if (dev->ifindex == ifindex)
703 return dev; 703 return dev;
704 704
705 return NULL; 705 return NULL;
706 } 706 }
707 EXPORT_SYMBOL(__dev_get_by_index); 707 EXPORT_SYMBOL(__dev_get_by_index);
708 708
709 /** 709 /**
710 * dev_get_by_index_rcu - find a device by its ifindex 710 * dev_get_by_index_rcu - find a device by its ifindex
711 * @net: the applicable net namespace 711 * @net: the applicable net namespace
712 * @ifindex: index of device 712 * @ifindex: index of device
713 * 713 *
714 * Search for an interface by index. Returns %NULL if the device 714 * Search for an interface by index. Returns %NULL if the device
715 * is not found or a pointer to the device. The device has not 715 * is not found or a pointer to the device. The device has not
716 * had its reference counter increased so the caller must be careful 716 * had its reference counter increased so the caller must be careful
717 * about locking. The caller must hold RCU lock. 717 * about locking. The caller must hold RCU lock.
718 */ 718 */
719 719
720 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex) 720 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex)
721 { 721 {
722 struct hlist_node *p; 722 struct hlist_node *p;
723 struct net_device *dev; 723 struct net_device *dev;
724 struct hlist_head *head = dev_index_hash(net, ifindex); 724 struct hlist_head *head = dev_index_hash(net, ifindex);
725 725
726 hlist_for_each_entry_rcu(dev, p, head, index_hlist) 726 hlist_for_each_entry_rcu(dev, p, head, index_hlist)
727 if (dev->ifindex == ifindex) 727 if (dev->ifindex == ifindex)
728 return dev; 728 return dev;
729 729
730 return NULL; 730 return NULL;
731 } 731 }
732 EXPORT_SYMBOL(dev_get_by_index_rcu); 732 EXPORT_SYMBOL(dev_get_by_index_rcu);
733 733
734 734
735 /** 735 /**
736 * dev_get_by_index - find a device by its ifindex 736 * dev_get_by_index - find a device by its ifindex
737 * @net: the applicable net namespace 737 * @net: the applicable net namespace
738 * @ifindex: index of device 738 * @ifindex: index of device
739 * 739 *
740 * Search for an interface by index. Returns NULL if the device 740 * Search for an interface by index. Returns NULL if the device
741 * is not found or a pointer to the device. The device returned has 741 * is not found or a pointer to the device. The device returned has
742 * had a reference added and the pointer is safe until the user calls 742 * had a reference added and the pointer is safe until the user calls
743 * dev_put to indicate they have finished with it. 743 * dev_put to indicate they have finished with it.
744 */ 744 */
745 745
746 struct net_device *dev_get_by_index(struct net *net, int ifindex) 746 struct net_device *dev_get_by_index(struct net *net, int ifindex)
747 { 747 {
748 struct net_device *dev; 748 struct net_device *dev;
749 749
750 rcu_read_lock(); 750 rcu_read_lock();
751 dev = dev_get_by_index_rcu(net, ifindex); 751 dev = dev_get_by_index_rcu(net, ifindex);
752 if (dev) 752 if (dev)
753 dev_hold(dev); 753 dev_hold(dev);
754 rcu_read_unlock(); 754 rcu_read_unlock();
755 return dev; 755 return dev;
756 } 756 }
757 EXPORT_SYMBOL(dev_get_by_index); 757 EXPORT_SYMBOL(dev_get_by_index);
758 758
759 /** 759 /**
760 * dev_getbyhwaddr_rcu - find a device by its hardware address 760 * dev_getbyhwaddr_rcu - find a device by its hardware address
761 * @net: the applicable net namespace 761 * @net: the applicable net namespace
762 * @type: media type of device 762 * @type: media type of device
763 * @ha: hardware address 763 * @ha: hardware address
764 * 764 *
765 * Search for an interface by MAC address. Returns NULL if the device 765 * Search for an interface by MAC address. Returns NULL if the device
766 * is not found or a pointer to the device. 766 * is not found or a pointer to the device.
767 * The caller must hold RCU or RTNL. 767 * The caller must hold RCU or RTNL.
768 * The returned device has not had its ref count increased 768 * The returned device has not had its ref count increased
769 * and the caller must therefore be careful about locking 769 * and the caller must therefore be careful about locking
770 * 770 *
771 */ 771 */
772 772
773 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 773 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
774 const char *ha) 774 const char *ha)
775 { 775 {
776 struct net_device *dev; 776 struct net_device *dev;
777 777
778 for_each_netdev_rcu(net, dev) 778 for_each_netdev_rcu(net, dev)
779 if (dev->type == type && 779 if (dev->type == type &&
780 !memcmp(dev->dev_addr, ha, dev->addr_len)) 780 !memcmp(dev->dev_addr, ha, dev->addr_len))
781 return dev; 781 return dev;
782 782
783 return NULL; 783 return NULL;
784 } 784 }
785 EXPORT_SYMBOL(dev_getbyhwaddr_rcu); 785 EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
786 786
787 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type) 787 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
788 { 788 {
789 struct net_device *dev; 789 struct net_device *dev;
790 790
791 ASSERT_RTNL(); 791 ASSERT_RTNL();
792 for_each_netdev(net, dev) 792 for_each_netdev(net, dev)
793 if (dev->type == type) 793 if (dev->type == type)
794 return dev; 794 return dev;
795 795
796 return NULL; 796 return NULL;
797 } 797 }
798 EXPORT_SYMBOL(__dev_getfirstbyhwtype); 798 EXPORT_SYMBOL(__dev_getfirstbyhwtype);
799 799
800 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type) 800 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
801 { 801 {
802 struct net_device *dev, *ret = NULL; 802 struct net_device *dev, *ret = NULL;
803 803
804 rcu_read_lock(); 804 rcu_read_lock();
805 for_each_netdev_rcu(net, dev) 805 for_each_netdev_rcu(net, dev)
806 if (dev->type == type) { 806 if (dev->type == type) {
807 dev_hold(dev); 807 dev_hold(dev);
808 ret = dev; 808 ret = dev;
809 break; 809 break;
810 } 810 }
811 rcu_read_unlock(); 811 rcu_read_unlock();
812 return ret; 812 return ret;
813 } 813 }
814 EXPORT_SYMBOL(dev_getfirstbyhwtype); 814 EXPORT_SYMBOL(dev_getfirstbyhwtype);
815 815
816 /** 816 /**
817 * dev_get_by_flags_rcu - find any device with given flags 817 * dev_get_by_flags_rcu - find any device with given flags
818 * @net: the applicable net namespace 818 * @net: the applicable net namespace
819 * @if_flags: IFF_* values 819 * @if_flags: IFF_* values
820 * @mask: bitmask of bits in if_flags to check 820 * @mask: bitmask of bits in if_flags to check
821 * 821 *
822 * Search for any interface with the given flags. Returns NULL if a device 822 * Search for any interface with the given flags. Returns NULL if a device
823 * is not found or a pointer to the device. Must be called inside 823 * is not found or a pointer to the device. Must be called inside
824 * rcu_read_lock(), and result refcount is unchanged. 824 * rcu_read_lock(), and result refcount is unchanged.
825 */ 825 */
826 826
827 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short if_flags, 827 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short if_flags,
828 unsigned short mask) 828 unsigned short mask)
829 { 829 {
830 struct net_device *dev, *ret; 830 struct net_device *dev, *ret;
831 831
832 ret = NULL; 832 ret = NULL;
833 for_each_netdev_rcu(net, dev) { 833 for_each_netdev_rcu(net, dev) {
834 if (((dev->flags ^ if_flags) & mask) == 0) { 834 if (((dev->flags ^ if_flags) & mask) == 0) {
835 ret = dev; 835 ret = dev;
836 break; 836 break;
837 } 837 }
838 } 838 }
839 return ret; 839 return ret;
840 } 840 }
841 EXPORT_SYMBOL(dev_get_by_flags_rcu); 841 EXPORT_SYMBOL(dev_get_by_flags_rcu);
842 842
843 /** 843 /**
844 * dev_valid_name - check if name is okay for network device 844 * dev_valid_name - check if name is okay for network device
845 * @name: name string 845 * @name: name string
846 * 846 *
847 * Network device names need to be valid file names to 847 * Network device names need to be valid file names to
848 * to allow sysfs to work. We also disallow any kind of 848 * to allow sysfs to work. We also disallow any kind of
849 * whitespace. 849 * whitespace.
850 */ 850 */
851 bool dev_valid_name(const char *name) 851 bool dev_valid_name(const char *name)
852 { 852 {
853 if (*name == '\0') 853 if (*name == '\0')
854 return false; 854 return false;
855 if (strlen(name) >= IFNAMSIZ) 855 if (strlen(name) >= IFNAMSIZ)
856 return false; 856 return false;
857 if (!strcmp(name, ".") || !strcmp(name, "..")) 857 if (!strcmp(name, ".") || !strcmp(name, ".."))
858 return false; 858 return false;
859 859
860 while (*name) { 860 while (*name) {
861 if (*name == '/' || isspace(*name)) 861 if (*name == '/' || isspace(*name))
862 return false; 862 return false;
863 name++; 863 name++;
864 } 864 }
865 return true; 865 return true;
866 } 866 }
867 EXPORT_SYMBOL(dev_valid_name); 867 EXPORT_SYMBOL(dev_valid_name);
868 868
869 /** 869 /**
870 * __dev_alloc_name - allocate a name for a device 870 * __dev_alloc_name - allocate a name for a device
871 * @net: network namespace to allocate the device name in 871 * @net: network namespace to allocate the device name in
872 * @name: name format string 872 * @name: name format string
873 * @buf: scratch buffer and result name string 873 * @buf: scratch buffer and result name string
874 * 874 *
875 * Passed a format string - eg "lt%d" it will try and find a suitable 875 * Passed a format string - eg "lt%d" it will try and find a suitable
876 * id. It scans list of devices to build up a free map, then chooses 876 * id. It scans list of devices to build up a free map, then chooses
877 * the first empty slot. The caller must hold the dev_base or rtnl lock 877 * the first empty slot. The caller must hold the dev_base or rtnl lock
878 * while allocating the name and adding the device in order to avoid 878 * while allocating the name and adding the device in order to avoid
879 * duplicates. 879 * duplicates.
880 * Limited to bits_per_byte * page size devices (ie 32K on most platforms). 880 * Limited to bits_per_byte * page size devices (ie 32K on most platforms).
881 * Returns the number of the unit assigned or a negative errno code. 881 * Returns the number of the unit assigned or a negative errno code.
882 */ 882 */
883 883
884 static int __dev_alloc_name(struct net *net, const char *name, char *buf) 884 static int __dev_alloc_name(struct net *net, const char *name, char *buf)
885 { 885 {
886 int i = 0; 886 int i = 0;
887 const char *p; 887 const char *p;
888 const int max_netdevices = 8*PAGE_SIZE; 888 const int max_netdevices = 8*PAGE_SIZE;
889 unsigned long *inuse; 889 unsigned long *inuse;
890 struct net_device *d; 890 struct net_device *d;
891 891
892 p = strnchr(name, IFNAMSIZ-1, '%'); 892 p = strnchr(name, IFNAMSIZ-1, '%');
893 if (p) { 893 if (p) {
894 /* 894 /*
895 * Verify the string as this thing may have come from 895 * Verify the string as this thing may have come from
896 * the user. There must be either one "%d" and no other "%" 896 * the user. There must be either one "%d" and no other "%"
897 * characters. 897 * characters.
898 */ 898 */
899 if (p[1] != 'd' || strchr(p + 2, '%')) 899 if (p[1] != 'd' || strchr(p + 2, '%'))
900 return -EINVAL; 900 return -EINVAL;
901 901
902 /* Use one page as a bit array of possible slots */ 902 /* Use one page as a bit array of possible slots */
903 inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC); 903 inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
904 if (!inuse) 904 if (!inuse)
905 return -ENOMEM; 905 return -ENOMEM;
906 906
907 for_each_netdev(net, d) { 907 for_each_netdev(net, d) {
908 if (!sscanf(d->name, name, &i)) 908 if (!sscanf(d->name, name, &i))
909 continue; 909 continue;
910 if (i < 0 || i >= max_netdevices) 910 if (i < 0 || i >= max_netdevices)
911 continue; 911 continue;
912 912
913 /* avoid cases where sscanf is not exact inverse of printf */ 913 /* avoid cases where sscanf is not exact inverse of printf */
914 snprintf(buf, IFNAMSIZ, name, i); 914 snprintf(buf, IFNAMSIZ, name, i);
915 if (!strncmp(buf, d->name, IFNAMSIZ)) 915 if (!strncmp(buf, d->name, IFNAMSIZ))
916 set_bit(i, inuse); 916 set_bit(i, inuse);
917 } 917 }
918 918
919 i = find_first_zero_bit(inuse, max_netdevices); 919 i = find_first_zero_bit(inuse, max_netdevices);
920 free_page((unsigned long) inuse); 920 free_page((unsigned long) inuse);
921 } 921 }
922 922
923 if (buf != name) 923 if (buf != name)
924 snprintf(buf, IFNAMSIZ, name, i); 924 snprintf(buf, IFNAMSIZ, name, i);
925 if (!__dev_get_by_name(net, buf)) 925 if (!__dev_get_by_name(net, buf))
926 return i; 926 return i;
927 927
928 /* It is possible to run out of possible slots 928 /* It is possible to run out of possible slots
929 * when the name is long and there isn't enough space left 929 * when the name is long and there isn't enough space left
930 * for the digits, or if all bits are used. 930 * for the digits, or if all bits are used.
931 */ 931 */
932 return -ENFILE; 932 return -ENFILE;
933 } 933 }
934 934
935 /** 935 /**
936 * dev_alloc_name - allocate a name for a device 936 * dev_alloc_name - allocate a name for a device
937 * @dev: device 937 * @dev: device
938 * @name: name format string 938 * @name: name format string
939 * 939 *
940 * Passed a format string - eg "lt%d" it will try and find a suitable 940 * Passed a format string - eg "lt%d" it will try and find a suitable
941 * id. It scans list of devices to build up a free map, then chooses 941 * id. It scans list of devices to build up a free map, then chooses
942 * the first empty slot. The caller must hold the dev_base or rtnl lock 942 * the first empty slot. The caller must hold the dev_base or rtnl lock
943 * while allocating the name and adding the device in order to avoid 943 * while allocating the name and adding the device in order to avoid
944 * duplicates. 944 * duplicates.
945 * Limited to bits_per_byte * page size devices (ie 32K on most platforms). 945 * Limited to bits_per_byte * page size devices (ie 32K on most platforms).
946 * Returns the number of the unit assigned or a negative errno code. 946 * Returns the number of the unit assigned or a negative errno code.
947 */ 947 */
948 948
949 int dev_alloc_name(struct net_device *dev, const char *name) 949 int dev_alloc_name(struct net_device *dev, const char *name)
950 { 950 {
951 char buf[IFNAMSIZ]; 951 char buf[IFNAMSIZ];
952 struct net *net; 952 struct net *net;
953 int ret; 953 int ret;
954 954
955 BUG_ON(!dev_net(dev)); 955 BUG_ON(!dev_net(dev));
956 net = dev_net(dev); 956 net = dev_net(dev);
957 ret = __dev_alloc_name(net, name, buf); 957 ret = __dev_alloc_name(net, name, buf);
958 if (ret >= 0) 958 if (ret >= 0)
959 strlcpy(dev->name, buf, IFNAMSIZ); 959 strlcpy(dev->name, buf, IFNAMSIZ);
960 return ret; 960 return ret;
961 } 961 }
962 EXPORT_SYMBOL(dev_alloc_name); 962 EXPORT_SYMBOL(dev_alloc_name);
963 963
964 static int dev_get_valid_name(struct net_device *dev, const char *name) 964 static int dev_get_valid_name(struct net_device *dev, const char *name)
965 { 965 {
966 struct net *net; 966 struct net *net;
967 967
968 BUG_ON(!dev_net(dev)); 968 BUG_ON(!dev_net(dev));
969 net = dev_net(dev); 969 net = dev_net(dev);
970 970
971 if (!dev_valid_name(name)) 971 if (!dev_valid_name(name))
972 return -EINVAL; 972 return -EINVAL;
973 973
974 if (strchr(name, '%')) 974 if (strchr(name, '%'))
975 return dev_alloc_name(dev, name); 975 return dev_alloc_name(dev, name);
976 else if (__dev_get_by_name(net, name)) 976 else if (__dev_get_by_name(net, name))
977 return -EEXIST; 977 return -EEXIST;
978 else if (dev->name != name) 978 else if (dev->name != name)
979 strlcpy(dev->name, name, IFNAMSIZ); 979 strlcpy(dev->name, name, IFNAMSIZ);
980 980
981 return 0; 981 return 0;
982 } 982 }
983 983
984 /** 984 /**
985 * dev_change_name - change name of a device 985 * dev_change_name - change name of a device
986 * @dev: device 986 * @dev: device
987 * @newname: name (or format string) must be at least IFNAMSIZ 987 * @newname: name (or format string) must be at least IFNAMSIZ
988 * 988 *
989 * Change name of a device, can pass format strings "eth%d". 989 * Change name of a device, can pass format strings "eth%d".
990 * for wildcarding. 990 * for wildcarding.
991 */ 991 */
992 int dev_change_name(struct net_device *dev, const char *newname) 992 int dev_change_name(struct net_device *dev, const char *newname)
993 { 993 {
994 char oldname[IFNAMSIZ]; 994 char oldname[IFNAMSIZ];
995 int err = 0; 995 int err = 0;
996 int ret; 996 int ret;
997 struct net *net; 997 struct net *net;
998 998
999 ASSERT_RTNL(); 999 ASSERT_RTNL();
1000 BUG_ON(!dev_net(dev)); 1000 BUG_ON(!dev_net(dev));
1001 1001
1002 net = dev_net(dev); 1002 net = dev_net(dev);
1003 if (dev->flags & IFF_UP) 1003 if (dev->flags & IFF_UP)
1004 return -EBUSY; 1004 return -EBUSY;
1005 1005
1006 if (strncmp(newname, dev->name, IFNAMSIZ) == 0) 1006 if (strncmp(newname, dev->name, IFNAMSIZ) == 0)
1007 return 0; 1007 return 0;
1008 1008
1009 memcpy(oldname, dev->name, IFNAMSIZ); 1009 memcpy(oldname, dev->name, IFNAMSIZ);
1010 1010
1011 err = dev_get_valid_name(dev, newname); 1011 err = dev_get_valid_name(dev, newname);
1012 if (err < 0) 1012 if (err < 0)
1013 return err; 1013 return err;
1014 1014
1015 rollback: 1015 rollback:
1016 ret = device_rename(&dev->dev, dev->name); 1016 ret = device_rename(&dev->dev, dev->name);
1017 if (ret) { 1017 if (ret) {
1018 memcpy(dev->name, oldname, IFNAMSIZ); 1018 memcpy(dev->name, oldname, IFNAMSIZ);
1019 return ret; 1019 return ret;
1020 } 1020 }
1021 1021
1022 write_lock_bh(&dev_base_lock); 1022 write_lock_bh(&dev_base_lock);
1023 hlist_del_rcu(&dev->name_hlist); 1023 hlist_del_rcu(&dev->name_hlist);
1024 write_unlock_bh(&dev_base_lock); 1024 write_unlock_bh(&dev_base_lock);
1025 1025
1026 synchronize_rcu(); 1026 synchronize_rcu();
1027 1027
1028 write_lock_bh(&dev_base_lock); 1028 write_lock_bh(&dev_base_lock);
1029 hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name)); 1029 hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
1030 write_unlock_bh(&dev_base_lock); 1030 write_unlock_bh(&dev_base_lock);
1031 1031
1032 ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); 1032 ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1033 ret = notifier_to_errno(ret); 1033 ret = notifier_to_errno(ret);
1034 1034
1035 if (ret) { 1035 if (ret) {
1036 /* err >= 0 after dev_alloc_name() or stores the first errno */ 1036 /* err >= 0 after dev_alloc_name() or stores the first errno */
1037 if (err >= 0) { 1037 if (err >= 0) {
1038 err = ret; 1038 err = ret;
1039 memcpy(dev->name, oldname, IFNAMSIZ); 1039 memcpy(dev->name, oldname, IFNAMSIZ);
1040 goto rollback; 1040 goto rollback;
1041 } else { 1041 } else {
1042 pr_err("%s: name change rollback failed: %d\n", 1042 pr_err("%s: name change rollback failed: %d\n",
1043 dev->name, ret); 1043 dev->name, ret);
1044 } 1044 }
1045 } 1045 }
1046 1046
1047 return err; 1047 return err;
1048 } 1048 }
1049 1049
1050 /** 1050 /**
1051 * dev_set_alias - change ifalias of a device 1051 * dev_set_alias - change ifalias of a device
1052 * @dev: device 1052 * @dev: device
1053 * @alias: name up to IFALIASZ 1053 * @alias: name up to IFALIASZ
1054 * @len: limit of bytes to copy from info 1054 * @len: limit of bytes to copy from info
1055 * 1055 *
1056 * Set ifalias for a device, 1056 * Set ifalias for a device,
1057 */ 1057 */
1058 int dev_set_alias(struct net_device *dev, const char *alias, size_t len) 1058 int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
1059 { 1059 {
1060 ASSERT_RTNL(); 1060 ASSERT_RTNL();
1061 1061
1062 if (len >= IFALIASZ) 1062 if (len >= IFALIASZ)
1063 return -EINVAL; 1063 return -EINVAL;
1064 1064
1065 if (!len) { 1065 if (!len) {
1066 if (dev->ifalias) { 1066 if (dev->ifalias) {
1067 kfree(dev->ifalias); 1067 kfree(dev->ifalias);
1068 dev->ifalias = NULL; 1068 dev->ifalias = NULL;
1069 } 1069 }
1070 return 0; 1070 return 0;
1071 } 1071 }
1072 1072
1073 dev->ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL); 1073 dev->ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
1074 if (!dev->ifalias) 1074 if (!dev->ifalias)
1075 return -ENOMEM; 1075 return -ENOMEM;
1076 1076
1077 strlcpy(dev->ifalias, alias, len+1); 1077 strlcpy(dev->ifalias, alias, len+1);
1078 return len; 1078 return len;
1079 } 1079 }
1080 1080
1081 1081
1082 /** 1082 /**
1083 * netdev_features_change - device changes features 1083 * netdev_features_change - device changes features
1084 * @dev: device to cause notification 1084 * @dev: device to cause notification
1085 * 1085 *
1086 * Called to indicate a device has changed features. 1086 * Called to indicate a device has changed features.
1087 */ 1087 */
1088 void netdev_features_change(struct net_device *dev) 1088 void netdev_features_change(struct net_device *dev)
1089 { 1089 {
1090 call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev); 1090 call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1091 } 1091 }
1092 EXPORT_SYMBOL(netdev_features_change); 1092 EXPORT_SYMBOL(netdev_features_change);
1093 1093
1094 /** 1094 /**
1095 * netdev_state_change - device changes state 1095 * netdev_state_change - device changes state
1096 * @dev: device to cause notification 1096 * @dev: device to cause notification
1097 * 1097 *
1098 * Called to indicate a device has changed state. This function calls 1098 * Called to indicate a device has changed state. This function calls
1099 * the notifier chains for netdev_chain and sends a NEWLINK message 1099 * the notifier chains for netdev_chain and sends a NEWLINK message
1100 * to the routing socket. 1100 * to the routing socket.
1101 */ 1101 */
1102 void netdev_state_change(struct net_device *dev) 1102 void netdev_state_change(struct net_device *dev)
1103 { 1103 {
1104 if (dev->flags & IFF_UP) { 1104 if (dev->flags & IFF_UP) {
1105 call_netdevice_notifiers(NETDEV_CHANGE, dev); 1105 call_netdevice_notifiers(NETDEV_CHANGE, dev);
1106 rtmsg_ifinfo(RTM_NEWLINK, dev, 0); 1106 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1107 } 1107 }
1108 } 1108 }
1109 EXPORT_SYMBOL(netdev_state_change); 1109 EXPORT_SYMBOL(netdev_state_change);
1110 1110
1111 int netdev_bonding_change(struct net_device *dev, unsigned long event) 1111 int netdev_bonding_change(struct net_device *dev, unsigned long event)
1112 { 1112 {
1113 return call_netdevice_notifiers(event, dev); 1113 return call_netdevice_notifiers(event, dev);
1114 } 1114 }
1115 EXPORT_SYMBOL(netdev_bonding_change); 1115 EXPORT_SYMBOL(netdev_bonding_change);
1116 1116
1117 /** 1117 /**
1118 * dev_load - load a network module 1118 * dev_load - load a network module
1119 * @net: the applicable net namespace 1119 * @net: the applicable net namespace
1120 * @name: name of interface 1120 * @name: name of interface
1121 * 1121 *
1122 * If a network interface is not present and the process has suitable 1122 * If a network interface is not present and the process has suitable
1123 * privileges this function loads the module. If module loading is not 1123 * privileges this function loads the module. If module loading is not
1124 * available in this kernel then it becomes a nop. 1124 * available in this kernel then it becomes a nop.
1125 */ 1125 */
1126 1126
1127 void dev_load(struct net *net, const char *name) 1127 void dev_load(struct net *net, const char *name)
1128 { 1128 {
1129 struct net_device *dev; 1129 struct net_device *dev;
1130 int no_module; 1130 int no_module;
1131 1131
1132 rcu_read_lock(); 1132 rcu_read_lock();
1133 dev = dev_get_by_name_rcu(net, name); 1133 dev = dev_get_by_name_rcu(net, name);
1134 rcu_read_unlock(); 1134 rcu_read_unlock();
1135 1135
1136 no_module = !dev; 1136 no_module = !dev;
1137 if (no_module && capable(CAP_NET_ADMIN)) 1137 if (no_module && capable(CAP_NET_ADMIN))
1138 no_module = request_module("netdev-%s", name); 1138 no_module = request_module("netdev-%s", name);
1139 if (no_module && capable(CAP_SYS_MODULE)) { 1139 if (no_module && capable(CAP_SYS_MODULE)) {
1140 if (!request_module("%s", name)) 1140 if (!request_module("%s", name))
1141 pr_err("Loading kernel module for a network device with CAP_SYS_MODULE (deprecated). Use CAP_NET_ADMIN and alias netdev-%s instead.\n", 1141 pr_err("Loading kernel module for a network device with CAP_SYS_MODULE (deprecated). Use CAP_NET_ADMIN and alias netdev-%s instead.\n",
1142 name); 1142 name);
1143 } 1143 }
1144 } 1144 }
1145 EXPORT_SYMBOL(dev_load); 1145 EXPORT_SYMBOL(dev_load);
1146 1146
1147 static int __dev_open(struct net_device *dev) 1147 static int __dev_open(struct net_device *dev)
1148 { 1148 {
1149 const struct net_device_ops *ops = dev->netdev_ops; 1149 const struct net_device_ops *ops = dev->netdev_ops;
1150 int ret; 1150 int ret;
1151 1151
1152 ASSERT_RTNL(); 1152 ASSERT_RTNL();
1153 1153
1154 if (!netif_device_present(dev)) 1154 if (!netif_device_present(dev))
1155 return -ENODEV; 1155 return -ENODEV;
1156 1156
1157 ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev); 1157 ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
1158 ret = notifier_to_errno(ret); 1158 ret = notifier_to_errno(ret);
1159 if (ret) 1159 if (ret)
1160 return ret; 1160 return ret;
1161 1161
1162 set_bit(__LINK_STATE_START, &dev->state); 1162 set_bit(__LINK_STATE_START, &dev->state);
1163 1163
1164 if (ops->ndo_validate_addr) 1164 if (ops->ndo_validate_addr)
1165 ret = ops->ndo_validate_addr(dev); 1165 ret = ops->ndo_validate_addr(dev);
1166 1166
1167 if (!ret && ops->ndo_open) 1167 if (!ret && ops->ndo_open)
1168 ret = ops->ndo_open(dev); 1168 ret = ops->ndo_open(dev);
1169 1169
1170 if (ret) 1170 if (ret)
1171 clear_bit(__LINK_STATE_START, &dev->state); 1171 clear_bit(__LINK_STATE_START, &dev->state);
1172 else { 1172 else {
1173 dev->flags |= IFF_UP; 1173 dev->flags |= IFF_UP;
1174 net_dmaengine_get(); 1174 net_dmaengine_get();
1175 dev_set_rx_mode(dev); 1175 dev_set_rx_mode(dev);
1176 dev_activate(dev); 1176 dev_activate(dev);
1177 } 1177 }
1178 1178
1179 return ret; 1179 return ret;
1180 } 1180 }
1181 1181
1182 /** 1182 /**
1183 * dev_open - prepare an interface for use. 1183 * dev_open - prepare an interface for use.
1184 * @dev: device to open 1184 * @dev: device to open
1185 * 1185 *
1186 * Takes a device from down to up state. The device's private open 1186 * Takes a device from down to up state. The device's private open
1187 * function is invoked and then the multicast lists are loaded. Finally 1187 * function is invoked and then the multicast lists are loaded. Finally
1188 * the device is moved into the up state and a %NETDEV_UP message is 1188 * the device is moved into the up state and a %NETDEV_UP message is
1189 * sent to the netdev notifier chain. 1189 * sent to the netdev notifier chain.
1190 * 1190 *
1191 * Calling this function on an active interface is a nop. On a failure 1191 * Calling this function on an active interface is a nop. On a failure
1192 * a negative errno code is returned. 1192 * a negative errno code is returned.
1193 */ 1193 */
1194 int dev_open(struct net_device *dev) 1194 int dev_open(struct net_device *dev)
1195 { 1195 {
1196 int ret; 1196 int ret;
1197 1197
1198 if (dev->flags & IFF_UP) 1198 if (dev->flags & IFF_UP)
1199 return 0; 1199 return 0;
1200 1200
1201 ret = __dev_open(dev); 1201 ret = __dev_open(dev);
1202 if (ret < 0) 1202 if (ret < 0)
1203 return ret; 1203 return ret;
1204 1204
1205 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING); 1205 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1206 call_netdevice_notifiers(NETDEV_UP, dev); 1206 call_netdevice_notifiers(NETDEV_UP, dev);
1207 1207
1208 return ret; 1208 return ret;
1209 } 1209 }
1210 EXPORT_SYMBOL(dev_open); 1210 EXPORT_SYMBOL(dev_open);
1211 1211
1212 static int __dev_close_many(struct list_head *head) 1212 static int __dev_close_many(struct list_head *head)
1213 { 1213 {
1214 struct net_device *dev; 1214 struct net_device *dev;
1215 1215
1216 ASSERT_RTNL(); 1216 ASSERT_RTNL();
1217 might_sleep(); 1217 might_sleep();
1218 1218
1219 list_for_each_entry(dev, head, unreg_list) { 1219 list_for_each_entry(dev, head, unreg_list) {
1220 call_netdevice_notifiers(NETDEV_GOING_DOWN, dev); 1220 call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
1221 1221
1222 clear_bit(__LINK_STATE_START, &dev->state); 1222 clear_bit(__LINK_STATE_START, &dev->state);
1223 1223
1224 /* Synchronize to scheduled poll. We cannot touch poll list, it 1224 /* Synchronize to scheduled poll. We cannot touch poll list, it
1225 * can be even on different cpu. So just clear netif_running(). 1225 * can be even on different cpu. So just clear netif_running().
1226 * 1226 *
1227 * dev->stop() will invoke napi_disable() on all of it's 1227 * dev->stop() will invoke napi_disable() on all of it's
1228 * napi_struct instances on this device. 1228 * napi_struct instances on this device.
1229 */ 1229 */
1230 smp_mb__after_clear_bit(); /* Commit netif_running(). */ 1230 smp_mb__after_clear_bit(); /* Commit netif_running(). */
1231 } 1231 }
1232 1232
1233 dev_deactivate_many(head); 1233 dev_deactivate_many(head);
1234 1234
1235 list_for_each_entry(dev, head, unreg_list) { 1235 list_for_each_entry(dev, head, unreg_list) {
1236 const struct net_device_ops *ops = dev->netdev_ops; 1236 const struct net_device_ops *ops = dev->netdev_ops;
1237 1237
1238 /* 1238 /*
1239 * Call the device specific close. This cannot fail. 1239 * Call the device specific close. This cannot fail.
1240 * Only if device is UP 1240 * Only if device is UP
1241 * 1241 *
1242 * We allow it to be called even after a DETACH hot-plug 1242 * We allow it to be called even after a DETACH hot-plug
1243 * event. 1243 * event.
1244 */ 1244 */
1245 if (ops->ndo_stop) 1245 if (ops->ndo_stop)
1246 ops->ndo_stop(dev); 1246 ops->ndo_stop(dev);
1247 1247
1248 dev->flags &= ~IFF_UP; 1248 dev->flags &= ~IFF_UP;
1249 net_dmaengine_put(); 1249 net_dmaengine_put();
1250 } 1250 }
1251 1251
1252 return 0; 1252 return 0;
1253 } 1253 }
1254 1254
1255 static int __dev_close(struct net_device *dev) 1255 static int __dev_close(struct net_device *dev)
1256 { 1256 {
1257 int retval; 1257 int retval;
1258 LIST_HEAD(single); 1258 LIST_HEAD(single);
1259 1259
1260 list_add(&dev->unreg_list, &single); 1260 list_add(&dev->unreg_list, &single);
1261 retval = __dev_close_many(&single); 1261 retval = __dev_close_many(&single);
1262 list_del(&single); 1262 list_del(&single);
1263 return retval; 1263 return retval;
1264 } 1264 }
1265 1265
1266 static int dev_close_many(struct list_head *head) 1266 static int dev_close_many(struct list_head *head)
1267 { 1267 {
1268 struct net_device *dev, *tmp; 1268 struct net_device *dev, *tmp;
1269 LIST_HEAD(tmp_list); 1269 LIST_HEAD(tmp_list);
1270 1270
1271 list_for_each_entry_safe(dev, tmp, head, unreg_list) 1271 list_for_each_entry_safe(dev, tmp, head, unreg_list)
1272 if (!(dev->flags & IFF_UP)) 1272 if (!(dev->flags & IFF_UP))
1273 list_move(&dev->unreg_list, &tmp_list); 1273 list_move(&dev->unreg_list, &tmp_list);
1274 1274
1275 __dev_close_many(head); 1275 __dev_close_many(head);
1276 1276
1277 list_for_each_entry(dev, head, unreg_list) { 1277 list_for_each_entry(dev, head, unreg_list) {
1278 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING); 1278 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1279 call_netdevice_notifiers(NETDEV_DOWN, dev); 1279 call_netdevice_notifiers(NETDEV_DOWN, dev);
1280 } 1280 }
1281 1281
1282 /* rollback_registered_many needs the complete original list */ 1282 /* rollback_registered_many needs the complete original list */
1283 list_splice(&tmp_list, head); 1283 list_splice(&tmp_list, head);
1284 return 0; 1284 return 0;
1285 } 1285 }
1286 1286
1287 /** 1287 /**
1288 * dev_close - shutdown an interface. 1288 * dev_close - shutdown an interface.
1289 * @dev: device to shutdown 1289 * @dev: device to shutdown
1290 * 1290 *
1291 * This function moves an active device into down state. A 1291 * This function moves an active device into down state. A
1292 * %NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device 1292 * %NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device
1293 * is then deactivated and finally a %NETDEV_DOWN is sent to the notifier 1293 * is then deactivated and finally a %NETDEV_DOWN is sent to the notifier
1294 * chain. 1294 * chain.
1295 */ 1295 */
1296 int dev_close(struct net_device *dev) 1296 int dev_close(struct net_device *dev)
1297 { 1297 {
1298 if (dev->flags & IFF_UP) { 1298 if (dev->flags & IFF_UP) {
1299 LIST_HEAD(single); 1299 LIST_HEAD(single);
1300 1300
1301 list_add(&dev->unreg_list, &single); 1301 list_add(&dev->unreg_list, &single);
1302 dev_close_many(&single); 1302 dev_close_many(&single);
1303 list_del(&single); 1303 list_del(&single);
1304 } 1304 }
1305 return 0; 1305 return 0;
1306 } 1306 }
1307 EXPORT_SYMBOL(dev_close); 1307 EXPORT_SYMBOL(dev_close);
1308 1308
1309 1309
1310 /** 1310 /**
1311 * dev_disable_lro - disable Large Receive Offload on a device 1311 * dev_disable_lro - disable Large Receive Offload on a device
1312 * @dev: device 1312 * @dev: device
1313 * 1313 *
1314 * Disable Large Receive Offload (LRO) on a net device. Must be 1314 * Disable Large Receive Offload (LRO) on a net device. Must be
1315 * called under RTNL. This is needed if received packets may be 1315 * called under RTNL. This is needed if received packets may be
1316 * forwarded to another interface. 1316 * forwarded to another interface.
1317 */ 1317 */
1318 void dev_disable_lro(struct net_device *dev) 1318 void dev_disable_lro(struct net_device *dev)
1319 { 1319 {
1320 /* 1320 /*
1321 * If we're trying to disable lro on a vlan device 1321 * If we're trying to disable lro on a vlan device
1322 * use the underlying physical device instead 1322 * use the underlying physical device instead
1323 */ 1323 */
1324 if (is_vlan_dev(dev)) 1324 if (is_vlan_dev(dev))
1325 dev = vlan_dev_real_dev(dev); 1325 dev = vlan_dev_real_dev(dev);
1326 1326
1327 dev->wanted_features &= ~NETIF_F_LRO; 1327 dev->wanted_features &= ~NETIF_F_LRO;
1328 netdev_update_features(dev); 1328 netdev_update_features(dev);
1329 1329
1330 if (unlikely(dev->features & NETIF_F_LRO)) 1330 if (unlikely(dev->features & NETIF_F_LRO))
1331 netdev_WARN(dev, "failed to disable LRO!\n"); 1331 netdev_WARN(dev, "failed to disable LRO!\n");
1332 } 1332 }
1333 EXPORT_SYMBOL(dev_disable_lro); 1333 EXPORT_SYMBOL(dev_disable_lro);
1334 1334
1335 1335
1336 static int dev_boot_phase = 1; 1336 static int dev_boot_phase = 1;
1337 1337
1338 /** 1338 /**
1339 * register_netdevice_notifier - register a network notifier block 1339 * register_netdevice_notifier - register a network notifier block
1340 * @nb: notifier 1340 * @nb: notifier
1341 * 1341 *
1342 * Register a notifier to be called when network device events occur. 1342 * Register a notifier to be called when network device events occur.
1343 * The notifier passed is linked into the kernel structures and must 1343 * The notifier passed is linked into the kernel structures and must
1344 * not be reused until it has been unregistered. A negative errno code 1344 * not be reused until it has been unregistered. A negative errno code
1345 * is returned on a failure. 1345 * is returned on a failure.
1346 * 1346 *
1347 * When registered all registration and up events are replayed 1347 * When registered all registration and up events are replayed
1348 * to the new notifier to allow device to have a race free 1348 * to the new notifier to allow device to have a race free
1349 * view of the network device list. 1349 * view of the network device list.
1350 */ 1350 */
1351 1351
1352 int register_netdevice_notifier(struct notifier_block *nb) 1352 int register_netdevice_notifier(struct notifier_block *nb)
1353 { 1353 {
1354 struct net_device *dev; 1354 struct net_device *dev;
1355 struct net_device *last; 1355 struct net_device *last;
1356 struct net *net; 1356 struct net *net;
1357 int err; 1357 int err;
1358 1358
1359 rtnl_lock(); 1359 rtnl_lock();
1360 err = raw_notifier_chain_register(&netdev_chain, nb); 1360 err = raw_notifier_chain_register(&netdev_chain, nb);
1361 if (err) 1361 if (err)
1362 goto unlock; 1362 goto unlock;
1363 if (dev_boot_phase) 1363 if (dev_boot_phase)
1364 goto unlock; 1364 goto unlock;
1365 for_each_net(net) { 1365 for_each_net(net) {
1366 for_each_netdev(net, dev) { 1366 for_each_netdev(net, dev) {
1367 err = nb->notifier_call(nb, NETDEV_REGISTER, dev); 1367 err = nb->notifier_call(nb, NETDEV_REGISTER, dev);
1368 err = notifier_to_errno(err); 1368 err = notifier_to_errno(err);
1369 if (err) 1369 if (err)
1370 goto rollback; 1370 goto rollback;
1371 1371
1372 if (!(dev->flags & IFF_UP)) 1372 if (!(dev->flags & IFF_UP))
1373 continue; 1373 continue;
1374 1374
1375 nb->notifier_call(nb, NETDEV_UP, dev); 1375 nb->notifier_call(nb, NETDEV_UP, dev);
1376 } 1376 }
1377 } 1377 }
1378 1378
1379 unlock: 1379 unlock:
1380 rtnl_unlock(); 1380 rtnl_unlock();
1381 return err; 1381 return err;
1382 1382
1383 rollback: 1383 rollback:
1384 last = dev; 1384 last = dev;
1385 for_each_net(net) { 1385 for_each_net(net) {
1386 for_each_netdev(net, dev) { 1386 for_each_netdev(net, dev) {
1387 if (dev == last) 1387 if (dev == last)
1388 goto outroll; 1388 goto outroll;
1389 1389
1390 if (dev->flags & IFF_UP) { 1390 if (dev->flags & IFF_UP) {
1391 nb->notifier_call(nb, NETDEV_GOING_DOWN, dev); 1391 nb->notifier_call(nb, NETDEV_GOING_DOWN, dev);
1392 nb->notifier_call(nb, NETDEV_DOWN, dev); 1392 nb->notifier_call(nb, NETDEV_DOWN, dev);
1393 } 1393 }
1394 nb->notifier_call(nb, NETDEV_UNREGISTER, dev); 1394 nb->notifier_call(nb, NETDEV_UNREGISTER, dev);
1395 nb->notifier_call(nb, NETDEV_UNREGISTER_BATCH, dev); 1395 nb->notifier_call(nb, NETDEV_UNREGISTER_BATCH, dev);
1396 } 1396 }
1397 } 1397 }
1398 1398
1399 outroll: 1399 outroll:
1400 raw_notifier_chain_unregister(&netdev_chain, nb); 1400 raw_notifier_chain_unregister(&netdev_chain, nb);
1401 goto unlock; 1401 goto unlock;
1402 } 1402 }
1403 EXPORT_SYMBOL(register_netdevice_notifier); 1403 EXPORT_SYMBOL(register_netdevice_notifier);
1404 1404
1405 /** 1405 /**
1406 * unregister_netdevice_notifier - unregister a network notifier block 1406 * unregister_netdevice_notifier - unregister a network notifier block
1407 * @nb: notifier 1407 * @nb: notifier
1408 * 1408 *
1409 * Unregister a notifier previously registered by 1409 * Unregister a notifier previously registered by
1410 * register_netdevice_notifier(). The notifier is unlinked into the 1410 * register_netdevice_notifier(). The notifier is unlinked into the
1411 * kernel structures and may then be reused. A negative errno code 1411 * kernel structures and may then be reused. A negative errno code
1412 * is returned on a failure. 1412 * is returned on a failure.
1413 */ 1413 */
1414 1414
1415 int unregister_netdevice_notifier(struct notifier_block *nb) 1415 int unregister_netdevice_notifier(struct notifier_block *nb)
1416 { 1416 {
1417 int err; 1417 int err;
1418 1418
1419 rtnl_lock(); 1419 rtnl_lock();
1420 err = raw_notifier_chain_unregister(&netdev_chain, nb); 1420 err = raw_notifier_chain_unregister(&netdev_chain, nb);
1421 rtnl_unlock(); 1421 rtnl_unlock();
1422 return err; 1422 return err;
1423 } 1423 }
1424 EXPORT_SYMBOL(unregister_netdevice_notifier); 1424 EXPORT_SYMBOL(unregister_netdevice_notifier);
1425 1425
1426 /** 1426 /**
1427 * call_netdevice_notifiers - call all network notifier blocks 1427 * call_netdevice_notifiers - call all network notifier blocks
1428 * @val: value passed unmodified to notifier function 1428 * @val: value passed unmodified to notifier function
1429 * @dev: net_device pointer passed unmodified to notifier function 1429 * @dev: net_device pointer passed unmodified to notifier function
1430 * 1430 *
1431 * Call all network notifier blocks. Parameters and return value 1431 * Call all network notifier blocks. Parameters and return value
1432 * are as for raw_notifier_call_chain(). 1432 * are as for raw_notifier_call_chain().
1433 */ 1433 */
1434 1434
1435 int call_netdevice_notifiers(unsigned long val, struct net_device *dev) 1435 int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
1436 { 1436 {
1437 ASSERT_RTNL(); 1437 ASSERT_RTNL();
1438 return raw_notifier_call_chain(&netdev_chain, val, dev); 1438 return raw_notifier_call_chain(&netdev_chain, val, dev);
1439 } 1439 }
1440 EXPORT_SYMBOL(call_netdevice_notifiers); 1440 EXPORT_SYMBOL(call_netdevice_notifiers);
1441 1441
1442 static struct static_key netstamp_needed __read_mostly; 1442 static struct static_key netstamp_needed __read_mostly;
1443 #ifdef HAVE_JUMP_LABEL 1443 #ifdef HAVE_JUMP_LABEL
1444 /* We are not allowed to call static_key_slow_dec() from irq context 1444 /* We are not allowed to call static_key_slow_dec() from irq context
1445 * If net_disable_timestamp() is called from irq context, defer the 1445 * If net_disable_timestamp() is called from irq context, defer the
1446 * static_key_slow_dec() calls. 1446 * static_key_slow_dec() calls.
1447 */ 1447 */
1448 static atomic_t netstamp_needed_deferred; 1448 static atomic_t netstamp_needed_deferred;
1449 #endif 1449 #endif
1450 1450
1451 void net_enable_timestamp(void) 1451 void net_enable_timestamp(void)
1452 { 1452 {
1453 #ifdef HAVE_JUMP_LABEL 1453 #ifdef HAVE_JUMP_LABEL
1454 int deferred = atomic_xchg(&netstamp_needed_deferred, 0); 1454 int deferred = atomic_xchg(&netstamp_needed_deferred, 0);
1455 1455
1456 if (deferred) { 1456 if (deferred) {
1457 while (--deferred) 1457 while (--deferred)
1458 static_key_slow_dec(&netstamp_needed); 1458 static_key_slow_dec(&netstamp_needed);
1459 return; 1459 return;
1460 } 1460 }
1461 #endif 1461 #endif
1462 WARN_ON(in_interrupt()); 1462 WARN_ON(in_interrupt());
1463 static_key_slow_inc(&netstamp_needed); 1463 static_key_slow_inc(&netstamp_needed);
1464 } 1464 }
1465 EXPORT_SYMBOL(net_enable_timestamp); 1465 EXPORT_SYMBOL(net_enable_timestamp);
1466 1466
1467 void net_disable_timestamp(void) 1467 void net_disable_timestamp(void)
1468 { 1468 {
1469 #ifdef HAVE_JUMP_LABEL 1469 #ifdef HAVE_JUMP_LABEL
1470 if (in_interrupt()) { 1470 if (in_interrupt()) {
1471 atomic_inc(&netstamp_needed_deferred); 1471 atomic_inc(&netstamp_needed_deferred);
1472 return; 1472 return;
1473 } 1473 }
1474 #endif 1474 #endif
1475 static_key_slow_dec(&netstamp_needed); 1475 static_key_slow_dec(&netstamp_needed);
1476 } 1476 }
1477 EXPORT_SYMBOL(net_disable_timestamp); 1477 EXPORT_SYMBOL(net_disable_timestamp);
1478 1478
1479 static inline void net_timestamp_set(struct sk_buff *skb) 1479 static inline void net_timestamp_set(struct sk_buff *skb)
1480 { 1480 {
1481 skb->tstamp.tv64 = 0; 1481 skb->tstamp.tv64 = 0;
1482 if (static_key_false(&netstamp_needed)) 1482 if (static_key_false(&netstamp_needed))
1483 __net_timestamp(skb); 1483 __net_timestamp(skb);
1484 } 1484 }
1485 1485
1486 #define net_timestamp_check(COND, SKB) \ 1486 #define net_timestamp_check(COND, SKB) \
1487 if (static_key_false(&netstamp_needed)) { \ 1487 if (static_key_false(&netstamp_needed)) { \
1488 if ((COND) && !(SKB)->tstamp.tv64) \ 1488 if ((COND) && !(SKB)->tstamp.tv64) \
1489 __net_timestamp(SKB); \ 1489 __net_timestamp(SKB); \
1490 } \ 1490 } \
1491 1491
1492 static int net_hwtstamp_validate(struct ifreq *ifr) 1492 static int net_hwtstamp_validate(struct ifreq *ifr)
1493 { 1493 {
1494 struct hwtstamp_config cfg; 1494 struct hwtstamp_config cfg;
1495 enum hwtstamp_tx_types tx_type; 1495 enum hwtstamp_tx_types tx_type;
1496 enum hwtstamp_rx_filters rx_filter; 1496 enum hwtstamp_rx_filters rx_filter;
1497 int tx_type_valid = 0; 1497 int tx_type_valid = 0;
1498 int rx_filter_valid = 0; 1498 int rx_filter_valid = 0;
1499 1499
1500 if (copy_from_user(&cfg, ifr->ifr_data, sizeof(cfg))) 1500 if (copy_from_user(&cfg, ifr->ifr_data, sizeof(cfg)))
1501 return -EFAULT; 1501 return -EFAULT;
1502 1502
1503 if (cfg.flags) /* reserved for future extensions */ 1503 if (cfg.flags) /* reserved for future extensions */
1504 return -EINVAL; 1504 return -EINVAL;
1505 1505
1506 tx_type = cfg.tx_type; 1506 tx_type = cfg.tx_type;
1507 rx_filter = cfg.rx_filter; 1507 rx_filter = cfg.rx_filter;
1508 1508
1509 switch (tx_type) { 1509 switch (tx_type) {
1510 case HWTSTAMP_TX_OFF: 1510 case HWTSTAMP_TX_OFF:
1511 case HWTSTAMP_TX_ON: 1511 case HWTSTAMP_TX_ON:
1512 case HWTSTAMP_TX_ONESTEP_SYNC: 1512 case HWTSTAMP_TX_ONESTEP_SYNC:
1513 tx_type_valid = 1; 1513 tx_type_valid = 1;
1514 break; 1514 break;
1515 } 1515 }
1516 1516
1517 switch (rx_filter) { 1517 switch (rx_filter) {
1518 case HWTSTAMP_FILTER_NONE: 1518 case HWTSTAMP_FILTER_NONE:
1519 case HWTSTAMP_FILTER_ALL: 1519 case HWTSTAMP_FILTER_ALL:
1520 case HWTSTAMP_FILTER_SOME: 1520 case HWTSTAMP_FILTER_SOME:
1521 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT: 1521 case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
1522 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC: 1522 case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
1523 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ: 1523 case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
1524 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT: 1524 case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
1525 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC: 1525 case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
1526 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ: 1526 case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
1527 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT: 1527 case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
1528 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC: 1528 case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
1529 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ: 1529 case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
1530 case HWTSTAMP_FILTER_PTP_V2_EVENT: 1530 case HWTSTAMP_FILTER_PTP_V2_EVENT:
1531 case HWTSTAMP_FILTER_PTP_V2_SYNC: 1531 case HWTSTAMP_FILTER_PTP_V2_SYNC:
1532 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ: 1532 case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
1533 rx_filter_valid = 1; 1533 rx_filter_valid = 1;
1534 break; 1534 break;
1535 } 1535 }
1536 1536
1537 if (!tx_type_valid || !rx_filter_valid) 1537 if (!tx_type_valid || !rx_filter_valid)
1538 return -ERANGE; 1538 return -ERANGE;
1539 1539
1540 return 0; 1540 return 0;
1541 } 1541 }
1542 1542
1543 static inline bool is_skb_forwardable(struct net_device *dev, 1543 static inline bool is_skb_forwardable(struct net_device *dev,
1544 struct sk_buff *skb) 1544 struct sk_buff *skb)
1545 { 1545 {
1546 unsigned int len; 1546 unsigned int len;
1547 1547
1548 if (!(dev->flags & IFF_UP)) 1548 if (!(dev->flags & IFF_UP))
1549 return false; 1549 return false;
1550 1550
1551 len = dev->mtu + dev->hard_header_len + VLAN_HLEN; 1551 len = dev->mtu + dev->hard_header_len + VLAN_HLEN;
1552 if (skb->len <= len) 1552 if (skb->len <= len)
1553 return true; 1553 return true;
1554 1554
1555 /* if TSO is enabled, we don't care about the length as the packet 1555 /* if TSO is enabled, we don't care about the length as the packet
1556 * could be forwarded without being segmented before 1556 * could be forwarded without being segmented before
1557 */ 1557 */
1558 if (skb_is_gso(skb)) 1558 if (skb_is_gso(skb))
1559 return true; 1559 return true;
1560 1560
1561 return false; 1561 return false;
1562 } 1562 }
1563 1563
1564 /** 1564 /**
1565 * dev_forward_skb - loopback an skb to another netif 1565 * dev_forward_skb - loopback an skb to another netif
1566 * 1566 *
1567 * @dev: destination network device 1567 * @dev: destination network device
1568 * @skb: buffer to forward 1568 * @skb: buffer to forward
1569 * 1569 *
1570 * return values: 1570 * return values:
1571 * NET_RX_SUCCESS (no congestion) 1571 * NET_RX_SUCCESS (no congestion)
1572 * NET_RX_DROP (packet was dropped, but freed) 1572 * NET_RX_DROP (packet was dropped, but freed)
1573 * 1573 *
1574 * dev_forward_skb can be used for injecting an skb from the 1574 * dev_forward_skb can be used for injecting an skb from the
1575 * start_xmit function of one device into the receive queue 1575 * start_xmit function of one device into the receive queue
1576 * of another device. 1576 * of another device.
1577 * 1577 *
1578 * The receiving device may be in another namespace, so 1578 * The receiving device may be in another namespace, so
1579 * we have to clear all information in the skb that could 1579 * we have to clear all information in the skb that could
1580 * impact namespace isolation. 1580 * impact namespace isolation.
1581 */ 1581 */
1582 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb) 1582 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
1583 { 1583 {
1584 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) { 1584 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
1585 if (skb_copy_ubufs(skb, GFP_ATOMIC)) { 1585 if (skb_copy_ubufs(skb, GFP_ATOMIC)) {
1586 atomic_long_inc(&dev->rx_dropped); 1586 atomic_long_inc(&dev->rx_dropped);
1587 kfree_skb(skb); 1587 kfree_skb(skb);
1588 return NET_RX_DROP; 1588 return NET_RX_DROP;
1589 } 1589 }
1590 } 1590 }
1591 1591
1592 skb_orphan(skb); 1592 skb_orphan(skb);
1593 nf_reset(skb); 1593 nf_reset(skb);
1594 1594
1595 if (unlikely(!is_skb_forwardable(dev, skb))) { 1595 if (unlikely(!is_skb_forwardable(dev, skb))) {
1596 atomic_long_inc(&dev->rx_dropped); 1596 atomic_long_inc(&dev->rx_dropped);
1597 kfree_skb(skb); 1597 kfree_skb(skb);
1598 return NET_RX_DROP; 1598 return NET_RX_DROP;
1599 } 1599 }
1600 skb_set_dev(skb, dev); 1600 skb_set_dev(skb, dev);
1601 skb->tstamp.tv64 = 0; 1601 skb->tstamp.tv64 = 0;
1602 skb->pkt_type = PACKET_HOST; 1602 skb->pkt_type = PACKET_HOST;
1603 skb->protocol = eth_type_trans(skb, dev); 1603 skb->protocol = eth_type_trans(skb, dev);
1604 return netif_rx(skb); 1604 return netif_rx(skb);
1605 } 1605 }
1606 EXPORT_SYMBOL_GPL(dev_forward_skb); 1606 EXPORT_SYMBOL_GPL(dev_forward_skb);
1607 1607
1608 static inline int deliver_skb(struct sk_buff *skb, 1608 static inline int deliver_skb(struct sk_buff *skb,
1609 struct packet_type *pt_prev, 1609 struct packet_type *pt_prev,
1610 struct net_device *orig_dev) 1610 struct net_device *orig_dev)
1611 { 1611 {
1612 atomic_inc(&skb->users); 1612 atomic_inc(&skb->users);
1613 return pt_prev->func(skb, skb->dev, pt_prev, orig_dev); 1613 return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
1614 } 1614 }
1615 1615
1616 /* 1616 /*
1617 * Support routine. Sends outgoing frames to any network 1617 * Support routine. Sends outgoing frames to any network
1618 * taps currently in use. 1618 * taps currently in use.
1619 */ 1619 */
1620 1620
1621 static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev) 1621 static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
1622 { 1622 {
1623 struct packet_type *ptype; 1623 struct packet_type *ptype;
1624 struct sk_buff *skb2 = NULL; 1624 struct sk_buff *skb2 = NULL;
1625 struct packet_type *pt_prev = NULL; 1625 struct packet_type *pt_prev = NULL;
1626 1626
1627 rcu_read_lock(); 1627 rcu_read_lock();
1628 list_for_each_entry_rcu(ptype, &ptype_all, list) { 1628 list_for_each_entry_rcu(ptype, &ptype_all, list) {
1629 /* Never send packets back to the socket 1629 /* Never send packets back to the socket
1630 * they originated from - MvS (miquels@drinkel.ow.org) 1630 * they originated from - MvS (miquels@drinkel.ow.org)
1631 */ 1631 */
1632 if ((ptype->dev == dev || !ptype->dev) && 1632 if ((ptype->dev == dev || !ptype->dev) &&
1633 (ptype->af_packet_priv == NULL || 1633 (ptype->af_packet_priv == NULL ||
1634 (struct sock *)ptype->af_packet_priv != skb->sk)) { 1634 (struct sock *)ptype->af_packet_priv != skb->sk)) {
1635 if (pt_prev) { 1635 if (pt_prev) {
1636 deliver_skb(skb2, pt_prev, skb->dev); 1636 deliver_skb(skb2, pt_prev, skb->dev);
1637 pt_prev = ptype; 1637 pt_prev = ptype;
1638 continue; 1638 continue;
1639 } 1639 }
1640 1640
1641 skb2 = skb_clone(skb, GFP_ATOMIC); 1641 skb2 = skb_clone(skb, GFP_ATOMIC);
1642 if (!skb2) 1642 if (!skb2)
1643 break; 1643 break;
1644 1644
1645 net_timestamp_set(skb2); 1645 net_timestamp_set(skb2);
1646 1646
1647 /* skb->nh should be correctly 1647 /* skb->nh should be correctly
1648 set by sender, so that the second statement is 1648 set by sender, so that the second statement is
1649 just protection against buggy protocols. 1649 just protection against buggy protocols.
1650 */ 1650 */
1651 skb_reset_mac_header(skb2); 1651 skb_reset_mac_header(skb2);
1652 1652
1653 if (skb_network_header(skb2) < skb2->data || 1653 if (skb_network_header(skb2) < skb2->data ||
1654 skb2->network_header > skb2->tail) { 1654 skb2->network_header > skb2->tail) {
1655 if (net_ratelimit()) 1655 if (net_ratelimit())
1656 pr_crit("protocol %04x is buggy, dev %s\n", 1656 pr_crit("protocol %04x is buggy, dev %s\n",
1657 ntohs(skb2->protocol), 1657 ntohs(skb2->protocol),
1658 dev->name); 1658 dev->name);
1659 skb_reset_network_header(skb2); 1659 skb_reset_network_header(skb2);
1660 } 1660 }
1661 1661
1662 skb2->transport_header = skb2->network_header; 1662 skb2->transport_header = skb2->network_header;
1663 skb2->pkt_type = PACKET_OUTGOING; 1663 skb2->pkt_type = PACKET_OUTGOING;
1664 pt_prev = ptype; 1664 pt_prev = ptype;
1665 } 1665 }
1666 } 1666 }
1667 if (pt_prev) 1667 if (pt_prev)
1668 pt_prev->func(skb2, skb->dev, pt_prev, skb->dev); 1668 pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
1669 rcu_read_unlock(); 1669 rcu_read_unlock();
1670 } 1670 }
1671 1671
1672 /* netif_setup_tc - Handle tc mappings on real_num_tx_queues change 1672 /* netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1673 * @dev: Network device 1673 * @dev: Network device
1674 * @txq: number of queues available 1674 * @txq: number of queues available
1675 * 1675 *
1676 * If real_num_tx_queues is changed the tc mappings may no longer be 1676 * If real_num_tx_queues is changed the tc mappings may no longer be
1677 * valid. To resolve this verify the tc mapping remains valid and if 1677 * valid. To resolve this verify the tc mapping remains valid and if
1678 * not NULL the mapping. With no priorities mapping to this 1678 * not NULL the mapping. With no priorities mapping to this
1679 * offset/count pair it will no longer be used. In the worst case TC0 1679 * offset/count pair it will no longer be used. In the worst case TC0
1680 * is invalid nothing can be done so disable priority mappings. If is 1680 * is invalid nothing can be done so disable priority mappings. If is
1681 * expected that drivers will fix this mapping if they can before 1681 * expected that drivers will fix this mapping if they can before
1682 * calling netif_set_real_num_tx_queues. 1682 * calling netif_set_real_num_tx_queues.
1683 */ 1683 */
1684 static void netif_setup_tc(struct net_device *dev, unsigned int txq) 1684 static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1685 { 1685 {
1686 int i; 1686 int i;
1687 struct netdev_tc_txq *tc = &dev->tc_to_txq[0]; 1687 struct netdev_tc_txq *tc = &dev->tc_to_txq[0];
1688 1688
1689 /* If TC0 is invalidated disable TC mapping */ 1689 /* If TC0 is invalidated disable TC mapping */
1690 if (tc->offset + tc->count > txq) { 1690 if (tc->offset + tc->count > txq) {
1691 pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n"); 1691 pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1692 dev->num_tc = 0; 1692 dev->num_tc = 0;
1693 return; 1693 return;
1694 } 1694 }
1695 1695
1696 /* Invalidated prio to tc mappings set to TC0 */ 1696 /* Invalidated prio to tc mappings set to TC0 */
1697 for (i = 1; i < TC_BITMASK + 1; i++) { 1697 for (i = 1; i < TC_BITMASK + 1; i++) {
1698 int q = netdev_get_prio_tc_map(dev, i); 1698 int q = netdev_get_prio_tc_map(dev, i);
1699 1699
1700 tc = &dev->tc_to_txq[q]; 1700 tc = &dev->tc_to_txq[q];
1701 if (tc->offset + tc->count > txq) { 1701 if (tc->offset + tc->count > txq) {
1702 pr_warn("Number of in use tx queues changed. Priority %i to tc mapping %i is no longer valid. Setting map to 0\n", 1702 pr_warn("Number of in use tx queues changed. Priority %i to tc mapping %i is no longer valid. Setting map to 0\n",
1703 i, q); 1703 i, q);
1704 netdev_set_prio_tc_map(dev, i, 0); 1704 netdev_set_prio_tc_map(dev, i, 0);
1705 } 1705 }
1706 } 1706 }
1707 } 1707 }
1708 1708
1709 /* 1709 /*
1710 * Routine to help set real_num_tx_queues. To avoid skbs mapped to queues 1710 * Routine to help set real_num_tx_queues. To avoid skbs mapped to queues
1711 * greater then real_num_tx_queues stale skbs on the qdisc must be flushed. 1711 * greater then real_num_tx_queues stale skbs on the qdisc must be flushed.
1712 */ 1712 */
1713 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq) 1713 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
1714 { 1714 {
1715 int rc; 1715 int rc;
1716 1716
1717 if (txq < 1 || txq > dev->num_tx_queues) 1717 if (txq < 1 || txq > dev->num_tx_queues)
1718 return -EINVAL; 1718 return -EINVAL;
1719 1719
1720 if (dev->reg_state == NETREG_REGISTERED || 1720 if (dev->reg_state == NETREG_REGISTERED ||
1721 dev->reg_state == NETREG_UNREGISTERING) { 1721 dev->reg_state == NETREG_UNREGISTERING) {
1722 ASSERT_RTNL(); 1722 ASSERT_RTNL();
1723 1723
1724 rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues, 1724 rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
1725 txq); 1725 txq);
1726 if (rc) 1726 if (rc)
1727 return rc; 1727 return rc;
1728 1728
1729 if (dev->num_tc) 1729 if (dev->num_tc)
1730 netif_setup_tc(dev, txq); 1730 netif_setup_tc(dev, txq);
1731 1731
1732 if (txq < dev->real_num_tx_queues) 1732 if (txq < dev->real_num_tx_queues)
1733 qdisc_reset_all_tx_gt(dev, txq); 1733 qdisc_reset_all_tx_gt(dev, txq);
1734 } 1734 }
1735 1735
1736 dev->real_num_tx_queues = txq; 1736 dev->real_num_tx_queues = txq;
1737 return 0; 1737 return 0;
1738 } 1738 }
1739 EXPORT_SYMBOL(netif_set_real_num_tx_queues); 1739 EXPORT_SYMBOL(netif_set_real_num_tx_queues);
1740 1740
1741 #ifdef CONFIG_RPS 1741 #ifdef CONFIG_RPS
1742 /** 1742 /**
1743 * netif_set_real_num_rx_queues - set actual number of RX queues used 1743 * netif_set_real_num_rx_queues - set actual number of RX queues used
1744 * @dev: Network device 1744 * @dev: Network device
1745 * @rxq: Actual number of RX queues 1745 * @rxq: Actual number of RX queues
1746 * 1746 *
1747 * This must be called either with the rtnl_lock held or before 1747 * This must be called either with the rtnl_lock held or before
1748 * registration of the net device. Returns 0 on success, or a 1748 * registration of the net device. Returns 0 on success, or a
1749 * negative error code. If called before registration, it always 1749 * negative error code. If called before registration, it always
1750 * succeeds. 1750 * succeeds.
1751 */ 1751 */
1752 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq) 1752 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
1753 { 1753 {
1754 int rc; 1754 int rc;
1755 1755
1756 if (rxq < 1 || rxq > dev->num_rx_queues) 1756 if (rxq < 1 || rxq > dev->num_rx_queues)
1757 return -EINVAL; 1757 return -EINVAL;
1758 1758
1759 if (dev->reg_state == NETREG_REGISTERED) { 1759 if (dev->reg_state == NETREG_REGISTERED) {
1760 ASSERT_RTNL(); 1760 ASSERT_RTNL();
1761 1761
1762 rc = net_rx_queue_update_kobjects(dev, dev->real_num_rx_queues, 1762 rc = net_rx_queue_update_kobjects(dev, dev->real_num_rx_queues,
1763 rxq); 1763 rxq);
1764 if (rc) 1764 if (rc)
1765 return rc; 1765 return rc;
1766 } 1766 }
1767 1767
1768 dev->real_num_rx_queues = rxq; 1768 dev->real_num_rx_queues = rxq;
1769 return 0; 1769 return 0;
1770 } 1770 }
1771 EXPORT_SYMBOL(netif_set_real_num_rx_queues); 1771 EXPORT_SYMBOL(netif_set_real_num_rx_queues);
1772 #endif 1772 #endif
1773 1773
1774 static inline void __netif_reschedule(struct Qdisc *q) 1774 static inline void __netif_reschedule(struct Qdisc *q)
1775 { 1775 {
1776 struct softnet_data *sd; 1776 struct softnet_data *sd;
1777 unsigned long flags; 1777 unsigned long flags;
1778 1778
1779 local_irq_save(flags); 1779 local_irq_save(flags);
1780 sd = &__get_cpu_var(softnet_data); 1780 sd = &__get_cpu_var(softnet_data);
1781 q->next_sched = NULL; 1781 q->next_sched = NULL;
1782 *sd->output_queue_tailp = q; 1782 *sd->output_queue_tailp = q;
1783 sd->output_queue_tailp = &q->next_sched; 1783 sd->output_queue_tailp = &q->next_sched;
1784 raise_softirq_irqoff(NET_TX_SOFTIRQ); 1784 raise_softirq_irqoff(NET_TX_SOFTIRQ);
1785 local_irq_restore(flags); 1785 local_irq_restore(flags);
1786 } 1786 }
1787 1787
1788 void __netif_schedule(struct Qdisc *q) 1788 void __netif_schedule(struct Qdisc *q)
1789 { 1789 {
1790 if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state)) 1790 if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state))
1791 __netif_reschedule(q); 1791 __netif_reschedule(q);
1792 } 1792 }
1793 EXPORT_SYMBOL(__netif_schedule); 1793 EXPORT_SYMBOL(__netif_schedule);
1794 1794
1795 void dev_kfree_skb_irq(struct sk_buff *skb) 1795 void dev_kfree_skb_irq(struct sk_buff *skb)
1796 { 1796 {
1797 if (atomic_dec_and_test(&skb->users)) { 1797 if (atomic_dec_and_test(&skb->users)) {
1798 struct softnet_data *sd; 1798 struct softnet_data *sd;
1799 unsigned long flags; 1799 unsigned long flags;
1800 1800
1801 local_irq_save(flags); 1801 local_irq_save(flags);
1802 sd = &__get_cpu_var(softnet_data); 1802 sd = &__get_cpu_var(softnet_data);
1803 skb->next = sd->completion_queue; 1803 skb->next = sd->completion_queue;
1804 sd->completion_queue = skb; 1804 sd->completion_queue = skb;
1805 raise_softirq_irqoff(NET_TX_SOFTIRQ); 1805 raise_softirq_irqoff(NET_TX_SOFTIRQ);
1806 local_irq_restore(flags); 1806 local_irq_restore(flags);
1807 } 1807 }
1808 } 1808 }
1809 EXPORT_SYMBOL(dev_kfree_skb_irq); 1809 EXPORT_SYMBOL(dev_kfree_skb_irq);
1810 1810
1811 void dev_kfree_skb_any(struct sk_buff *skb) 1811 void dev_kfree_skb_any(struct sk_buff *skb)
1812 { 1812 {
1813 if (in_irq() || irqs_disabled()) 1813 if (in_irq() || irqs_disabled())
1814 dev_kfree_skb_irq(skb); 1814 dev_kfree_skb_irq(skb);
1815 else 1815 else
1816 dev_kfree_skb(skb); 1816 dev_kfree_skb(skb);
1817 } 1817 }
1818 EXPORT_SYMBOL(dev_kfree_skb_any); 1818 EXPORT_SYMBOL(dev_kfree_skb_any);
1819 1819
1820 1820
1821 /** 1821 /**
1822 * netif_device_detach - mark device as removed 1822 * netif_device_detach - mark device as removed
1823 * @dev: network device 1823 * @dev: network device
1824 * 1824 *
1825 * Mark device as removed from system and therefore no longer available. 1825 * Mark device as removed from system and therefore no longer available.
1826 */ 1826 */
1827 void netif_device_detach(struct net_device *dev) 1827 void netif_device_detach(struct net_device *dev)
1828 { 1828 {
1829 if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) && 1829 if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
1830 netif_running(dev)) { 1830 netif_running(dev)) {
1831 netif_tx_stop_all_queues(dev); 1831 netif_tx_stop_all_queues(dev);
1832 } 1832 }
1833 } 1833 }
1834 EXPORT_SYMBOL(netif_device_detach); 1834 EXPORT_SYMBOL(netif_device_detach);
1835 1835
1836 /** 1836 /**
1837 * netif_device_attach - mark device as attached 1837 * netif_device_attach - mark device as attached
1838 * @dev: network device 1838 * @dev: network device
1839 * 1839 *
1840 * Mark device as attached from system and restart if needed. 1840 * Mark device as attached from system and restart if needed.
1841 */ 1841 */
1842 void netif_device_attach(struct net_device *dev) 1842 void netif_device_attach(struct net_device *dev)
1843 { 1843 {
1844 if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) && 1844 if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
1845 netif_running(dev)) { 1845 netif_running(dev)) {
1846 netif_tx_wake_all_queues(dev); 1846 netif_tx_wake_all_queues(dev);
1847 __netdev_watchdog_up(dev); 1847 __netdev_watchdog_up(dev);
1848 } 1848 }
1849 } 1849 }
1850 EXPORT_SYMBOL(netif_device_attach); 1850 EXPORT_SYMBOL(netif_device_attach);
1851 1851
1852 /** 1852 /**
1853 * skb_dev_set -- assign a new device to a buffer 1853 * skb_dev_set -- assign a new device to a buffer
1854 * @skb: buffer for the new device 1854 * @skb: buffer for the new device
1855 * @dev: network device 1855 * @dev: network device
1856 * 1856 *
1857 * If an skb is owned by a device already, we have to reset 1857 * If an skb is owned by a device already, we have to reset
1858 * all data private to the namespace a device belongs to 1858 * all data private to the namespace a device belongs to
1859 * before assigning it a new device. 1859 * before assigning it a new device.
1860 */ 1860 */
1861 #ifdef CONFIG_NET_NS 1861 #ifdef CONFIG_NET_NS
1862 void skb_set_dev(struct sk_buff *skb, struct net_device *dev) 1862 void skb_set_dev(struct sk_buff *skb, struct net_device *dev)
1863 { 1863 {
1864 skb_dst_drop(skb); 1864 skb_dst_drop(skb);
1865 if (skb->dev && !net_eq(dev_net(skb->dev), dev_net(dev))) { 1865 if (skb->dev && !net_eq(dev_net(skb->dev), dev_net(dev))) {
1866 secpath_reset(skb); 1866 secpath_reset(skb);
1867 nf_reset(skb); 1867 nf_reset(skb);
1868 skb_init_secmark(skb); 1868 skb_init_secmark(skb);
1869 skb->mark = 0; 1869 skb->mark = 0;
1870 skb->priority = 0; 1870 skb->priority = 0;
1871 skb->nf_trace = 0; 1871 skb->nf_trace = 0;
1872 skb->ipvs_property = 0; 1872 skb->ipvs_property = 0;
1873 #ifdef CONFIG_NET_SCHED 1873 #ifdef CONFIG_NET_SCHED
1874 skb->tc_index = 0; 1874 skb->tc_index = 0;
1875 #endif 1875 #endif
1876 } 1876 }
1877 skb->dev = dev; 1877 skb->dev = dev;
1878 } 1878 }
1879 EXPORT_SYMBOL(skb_set_dev); 1879 EXPORT_SYMBOL(skb_set_dev);
1880 #endif /* CONFIG_NET_NS */ 1880 #endif /* CONFIG_NET_NS */
1881 1881
1882 static void skb_warn_bad_offload(const struct sk_buff *skb) 1882 static void skb_warn_bad_offload(const struct sk_buff *skb)
1883 { 1883 {
1884 static const netdev_features_t null_features = 0; 1884 static const netdev_features_t null_features = 0;
1885 struct net_device *dev = skb->dev; 1885 struct net_device *dev = skb->dev;
1886 const char *driver = ""; 1886 const char *driver = "";
1887 1887
1888 if (dev && dev->dev.parent) 1888 if (dev && dev->dev.parent)
1889 driver = dev_driver_string(dev->dev.parent); 1889 driver = dev_driver_string(dev->dev.parent);
1890 1890
1891 WARN(1, "%s: caps=(%pNF, %pNF) len=%d data_len=%d gso_size=%d " 1891 WARN(1, "%s: caps=(%pNF, %pNF) len=%d data_len=%d gso_size=%d "
1892 "gso_type=%d ip_summed=%d\n", 1892 "gso_type=%d ip_summed=%d\n",
1893 driver, dev ? &dev->features : &null_features, 1893 driver, dev ? &dev->features : &null_features,
1894 skb->sk ? &skb->sk->sk_route_caps : &null_features, 1894 skb->sk ? &skb->sk->sk_route_caps : &null_features,
1895 skb->len, skb->data_len, skb_shinfo(skb)->gso_size, 1895 skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
1896 skb_shinfo(skb)->gso_type, skb->ip_summed); 1896 skb_shinfo(skb)->gso_type, skb->ip_summed);
1897 } 1897 }
1898 1898
1899 /* 1899 /*
1900 * Invalidate hardware checksum when packet is to be mangled, and 1900 * Invalidate hardware checksum when packet is to be mangled, and
1901 * complete checksum manually on outgoing path. 1901 * complete checksum manually on outgoing path.
1902 */ 1902 */
1903 int skb_checksum_help(struct sk_buff *skb) 1903 int skb_checksum_help(struct sk_buff *skb)
1904 { 1904 {
1905 __wsum csum; 1905 __wsum csum;
1906 int ret = 0, offset; 1906 int ret = 0, offset;
1907 1907
1908 if (skb->ip_summed == CHECKSUM_COMPLETE) 1908 if (skb->ip_summed == CHECKSUM_COMPLETE)
1909 goto out_set_summed; 1909 goto out_set_summed;
1910 1910
1911 if (unlikely(skb_shinfo(skb)->gso_size)) { 1911 if (unlikely(skb_shinfo(skb)->gso_size)) {
1912 skb_warn_bad_offload(skb); 1912 skb_warn_bad_offload(skb);
1913 return -EINVAL; 1913 return -EINVAL;
1914 } 1914 }
1915 1915
1916 offset = skb_checksum_start_offset(skb); 1916 offset = skb_checksum_start_offset(skb);
1917 BUG_ON(offset >= skb_headlen(skb)); 1917 BUG_ON(offset >= skb_headlen(skb));
1918 csum = skb_checksum(skb, offset, skb->len - offset, 0); 1918 csum = skb_checksum(skb, offset, skb->len - offset, 0);
1919 1919
1920 offset += skb->csum_offset; 1920 offset += skb->csum_offset;
1921 BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb)); 1921 BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb));
1922 1922
1923 if (skb_cloned(skb) && 1923 if (skb_cloned(skb) &&
1924 !skb_clone_writable(skb, offset + sizeof(__sum16))) { 1924 !skb_clone_writable(skb, offset + sizeof(__sum16))) {
1925 ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 1925 ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
1926 if (ret) 1926 if (ret)
1927 goto out; 1927 goto out;
1928 } 1928 }
1929 1929
1930 *(__sum16 *)(skb->data + offset) = csum_fold(csum); 1930 *(__sum16 *)(skb->data + offset) = csum_fold(csum);
1931 out_set_summed: 1931 out_set_summed:
1932 skb->ip_summed = CHECKSUM_NONE; 1932 skb->ip_summed = CHECKSUM_NONE;
1933 out: 1933 out:
1934 return ret; 1934 return ret;
1935 } 1935 }
1936 EXPORT_SYMBOL(skb_checksum_help); 1936 EXPORT_SYMBOL(skb_checksum_help);
1937 1937
1938 /** 1938 /**
1939 * skb_gso_segment - Perform segmentation on skb. 1939 * skb_gso_segment - Perform segmentation on skb.
1940 * @skb: buffer to segment 1940 * @skb: buffer to segment
1941 * @features: features for the output path (see dev->features) 1941 * @features: features for the output path (see dev->features)
1942 * 1942 *
1943 * This function segments the given skb and returns a list of segments. 1943 * This function segments the given skb and returns a list of segments.
1944 * 1944 *
1945 * It may return NULL if the skb requires no segmentation. This is 1945 * It may return NULL if the skb requires no segmentation. This is
1946 * only possible when GSO is used for verifying header integrity. 1946 * only possible when GSO is used for verifying header integrity.
1947 */ 1947 */
1948 struct sk_buff *skb_gso_segment(struct sk_buff *skb, 1948 struct sk_buff *skb_gso_segment(struct sk_buff *skb,
1949 netdev_features_t features) 1949 netdev_features_t features)
1950 { 1950 {
1951 struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT); 1951 struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT);
1952 struct packet_type *ptype; 1952 struct packet_type *ptype;
1953 __be16 type = skb->protocol; 1953 __be16 type = skb->protocol;
1954 int vlan_depth = ETH_HLEN; 1954 int vlan_depth = ETH_HLEN;
1955 int err; 1955 int err;
1956 1956
1957 while (type == htons(ETH_P_8021Q)) { 1957 while (type == htons(ETH_P_8021Q)) {
1958 struct vlan_hdr *vh; 1958 struct vlan_hdr *vh;
1959 1959
1960 if (unlikely(!pskb_may_pull(skb, vlan_depth + VLAN_HLEN))) 1960 if (unlikely(!pskb_may_pull(skb, vlan_depth + VLAN_HLEN)))
1961 return ERR_PTR(-EINVAL); 1961 return ERR_PTR(-EINVAL);
1962 1962
1963 vh = (struct vlan_hdr *)(skb->data + vlan_depth); 1963 vh = (struct vlan_hdr *)(skb->data + vlan_depth);
1964 type = vh->h_vlan_encapsulated_proto; 1964 type = vh->h_vlan_encapsulated_proto;
1965 vlan_depth += VLAN_HLEN; 1965 vlan_depth += VLAN_HLEN;
1966 } 1966 }
1967 1967
1968 skb_reset_mac_header(skb); 1968 skb_reset_mac_header(skb);
1969 skb->mac_len = skb->network_header - skb->mac_header; 1969 skb->mac_len = skb->network_header - skb->mac_header;
1970 __skb_pull(skb, skb->mac_len); 1970 __skb_pull(skb, skb->mac_len);
1971 1971
1972 if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { 1972 if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
1973 skb_warn_bad_offload(skb); 1973 skb_warn_bad_offload(skb);
1974 1974
1975 if (skb_header_cloned(skb) && 1975 if (skb_header_cloned(skb) &&
1976 (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) 1976 (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
1977 return ERR_PTR(err); 1977 return ERR_PTR(err);
1978 } 1978 }
1979 1979
1980 rcu_read_lock(); 1980 rcu_read_lock();
1981 list_for_each_entry_rcu(ptype, 1981 list_for_each_entry_rcu(ptype,
1982 &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { 1982 &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
1983 if (ptype->type == type && !ptype->dev && ptype->gso_segment) { 1983 if (ptype->type == type && !ptype->dev && ptype->gso_segment) {
1984 if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) { 1984 if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
1985 err = ptype->gso_send_check(skb); 1985 err = ptype->gso_send_check(skb);
1986 segs = ERR_PTR(err); 1986 segs = ERR_PTR(err);
1987 if (err || skb_gso_ok(skb, features)) 1987 if (err || skb_gso_ok(skb, features))
1988 break; 1988 break;
1989 __skb_push(skb, (skb->data - 1989 __skb_push(skb, (skb->data -
1990 skb_network_header(skb))); 1990 skb_network_header(skb)));
1991 } 1991 }
1992 segs = ptype->gso_segment(skb, features); 1992 segs = ptype->gso_segment(skb, features);
1993 break; 1993 break;
1994 } 1994 }
1995 } 1995 }
1996 rcu_read_unlock(); 1996 rcu_read_unlock();
1997 1997
1998 __skb_push(skb, skb->data - skb_mac_header(skb)); 1998 __skb_push(skb, skb->data - skb_mac_header(skb));
1999 1999
2000 return segs; 2000 return segs;
2001 } 2001 }
2002 EXPORT_SYMBOL(skb_gso_segment); 2002 EXPORT_SYMBOL(skb_gso_segment);
2003 2003
2004 /* Take action when hardware reception checksum errors are detected. */ 2004 /* Take action when hardware reception checksum errors are detected. */
2005 #ifdef CONFIG_BUG 2005 #ifdef CONFIG_BUG
2006 void netdev_rx_csum_fault(struct net_device *dev) 2006 void netdev_rx_csum_fault(struct net_device *dev)
2007 { 2007 {
2008 if (net_ratelimit()) { 2008 if (net_ratelimit()) {
2009 pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>"); 2009 pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2010 dump_stack(); 2010 dump_stack();
2011 } 2011 }
2012 } 2012 }
2013 EXPORT_SYMBOL(netdev_rx_csum_fault); 2013 EXPORT_SYMBOL(netdev_rx_csum_fault);
2014 #endif 2014 #endif
2015 2015
2016 /* Actually, we should eliminate this check as soon as we know, that: 2016 /* Actually, we should eliminate this check as soon as we know, that:
2017 * 1. IOMMU is present and allows to map all the memory. 2017 * 1. IOMMU is present and allows to map all the memory.
2018 * 2. No high memory really exists on this machine. 2018 * 2. No high memory really exists on this machine.
2019 */ 2019 */
2020 2020
2021 static int illegal_highdma(struct net_device *dev, struct sk_buff *skb) 2021 static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
2022 { 2022 {
2023 #ifdef CONFIG_HIGHMEM 2023 #ifdef CONFIG_HIGHMEM
2024 int i; 2024 int i;
2025 if (!(dev->features & NETIF_F_HIGHDMA)) { 2025 if (!(dev->features & NETIF_F_HIGHDMA)) {
2026 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) { 2026 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2027 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2027 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2028 if (PageHighMem(skb_frag_page(frag))) 2028 if (PageHighMem(skb_frag_page(frag)))
2029 return 1; 2029 return 1;
2030 } 2030 }
2031 } 2031 }
2032 2032
2033 if (PCI_DMA_BUS_IS_PHYS) { 2033 if (PCI_DMA_BUS_IS_PHYS) {
2034 struct device *pdev = dev->dev.parent; 2034 struct device *pdev = dev->dev.parent;
2035 2035
2036 if (!pdev) 2036 if (!pdev)
2037 return 0; 2037 return 0;
2038 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) { 2038 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2039 skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; 2039 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2040 dma_addr_t addr = page_to_phys(skb_frag_page(frag)); 2040 dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2041 if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask) 2041 if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
2042 return 1; 2042 return 1;
2043 } 2043 }
2044 } 2044 }
2045 #endif 2045 #endif
2046 return 0; 2046 return 0;
2047 } 2047 }
2048 2048
2049 struct dev_gso_cb { 2049 struct dev_gso_cb {
2050 void (*destructor)(struct sk_buff *skb); 2050 void (*destructor)(struct sk_buff *skb);
2051 }; 2051 };
2052 2052
2053 #define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb) 2053 #define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb)
2054 2054
2055 static void dev_gso_skb_destructor(struct sk_buff *skb) 2055 static void dev_gso_skb_destructor(struct sk_buff *skb)
2056 { 2056 {
2057 struct dev_gso_cb *cb; 2057 struct dev_gso_cb *cb;
2058 2058
2059 do { 2059 do {
2060 struct sk_buff *nskb = skb->next; 2060 struct sk_buff *nskb = skb->next;
2061 2061
2062 skb->next = nskb->next; 2062 skb->next = nskb->next;
2063 nskb->next = NULL; 2063 nskb->next = NULL;
2064 kfree_skb(nskb); 2064 kfree_skb(nskb);
2065 } while (skb->next); 2065 } while (skb->next);
2066 2066
2067 cb = DEV_GSO_CB(skb); 2067 cb = DEV_GSO_CB(skb);
2068 if (cb->destructor) 2068 if (cb->destructor)
2069 cb->destructor(skb); 2069 cb->destructor(skb);
2070 } 2070 }
2071 2071
2072 /** 2072 /**
2073 * dev_gso_segment - Perform emulated hardware segmentation on skb. 2073 * dev_gso_segment - Perform emulated hardware segmentation on skb.
2074 * @skb: buffer to segment 2074 * @skb: buffer to segment
2075 * @features: device features as applicable to this skb 2075 * @features: device features as applicable to this skb
2076 * 2076 *
2077 * This function segments the given skb and stores the list of segments 2077 * This function segments the given skb and stores the list of segments
2078 * in skb->next. 2078 * in skb->next.
2079 */ 2079 */
2080 static int dev_gso_segment(struct sk_buff *skb, netdev_features_t features) 2080 static int dev_gso_segment(struct sk_buff *skb, netdev_features_t features)
2081 { 2081 {
2082 struct sk_buff *segs; 2082 struct sk_buff *segs;
2083 2083
2084 segs = skb_gso_segment(skb, features); 2084 segs = skb_gso_segment(skb, features);
2085 2085
2086 /* Verifying header integrity only. */ 2086 /* Verifying header integrity only. */
2087 if (!segs) 2087 if (!segs)
2088 return 0; 2088 return 0;
2089 2089
2090 if (IS_ERR(segs)) 2090 if (IS_ERR(segs))
2091 return PTR_ERR(segs); 2091 return PTR_ERR(segs);
2092 2092
2093 skb->next = segs; 2093 skb->next = segs;
2094 DEV_GSO_CB(skb)->destructor = skb->destructor; 2094 DEV_GSO_CB(skb)->destructor = skb->destructor;
2095 skb->destructor = dev_gso_skb_destructor; 2095 skb->destructor = dev_gso_skb_destructor;
2096 2096
2097 return 0; 2097 return 0;
2098 } 2098 }
2099 2099
2100 /* 2100 /*
2101 * Try to orphan skb early, right before transmission by the device. 2101 * Try to orphan skb early, right before transmission by the device.
2102 * We cannot orphan skb if tx timestamp is requested or the sk-reference 2102 * We cannot orphan skb if tx timestamp is requested or the sk-reference
2103 * is needed on driver level for other reasons, e.g. see net/can/raw.c 2103 * is needed on driver level for other reasons, e.g. see net/can/raw.c
2104 */ 2104 */
2105 static inline void skb_orphan_try(struct sk_buff *skb) 2105 static inline void skb_orphan_try(struct sk_buff *skb)
2106 { 2106 {
2107 struct sock *sk = skb->sk; 2107 struct sock *sk = skb->sk;
2108 2108
2109 if (sk && !skb_shinfo(skb)->tx_flags) { 2109 if (sk && !skb_shinfo(skb)->tx_flags) {
2110 /* skb_tx_hash() wont be able to get sk. 2110 /* skb_tx_hash() wont be able to get sk.
2111 * We copy sk_hash into skb->rxhash 2111 * We copy sk_hash into skb->rxhash
2112 */ 2112 */
2113 if (!skb->rxhash) 2113 if (!skb->rxhash)
2114 skb->rxhash = sk->sk_hash; 2114 skb->rxhash = sk->sk_hash;
2115 skb_orphan(skb); 2115 skb_orphan(skb);
2116 } 2116 }
2117 } 2117 }
2118 2118
2119 static bool can_checksum_protocol(netdev_features_t features, __be16 protocol) 2119 static bool can_checksum_protocol(netdev_features_t features, __be16 protocol)
2120 { 2120 {
2121 return ((features & NETIF_F_GEN_CSUM) || 2121 return ((features & NETIF_F_GEN_CSUM) ||
2122 ((features & NETIF_F_V4_CSUM) && 2122 ((features & NETIF_F_V4_CSUM) &&
2123 protocol == htons(ETH_P_IP)) || 2123 protocol == htons(ETH_P_IP)) ||
2124 ((features & NETIF_F_V6_CSUM) && 2124 ((features & NETIF_F_V6_CSUM) &&
2125 protocol == htons(ETH_P_IPV6)) || 2125 protocol == htons(ETH_P_IPV6)) ||
2126 ((features & NETIF_F_FCOE_CRC) && 2126 ((features & NETIF_F_FCOE_CRC) &&
2127 protocol == htons(ETH_P_FCOE))); 2127 protocol == htons(ETH_P_FCOE)));
2128 } 2128 }
2129 2129
2130 static netdev_features_t harmonize_features(struct sk_buff *skb, 2130 static netdev_features_t harmonize_features(struct sk_buff *skb,
2131 __be16 protocol, netdev_features_t features) 2131 __be16 protocol, netdev_features_t features)
2132 { 2132 {
2133 if (!can_checksum_protocol(features, protocol)) { 2133 if (!can_checksum_protocol(features, protocol)) {
2134 features &= ~NETIF_F_ALL_CSUM; 2134 features &= ~NETIF_F_ALL_CSUM;
2135 features &= ~NETIF_F_SG; 2135 features &= ~NETIF_F_SG;
2136 } else if (illegal_highdma(skb->dev, skb)) { 2136 } else if (illegal_highdma(skb->dev, skb)) {
2137 features &= ~NETIF_F_SG; 2137 features &= ~NETIF_F_SG;
2138 } 2138 }
2139 2139
2140 return features; 2140 return features;
2141 } 2141 }
2142 2142
2143 netdev_features_t netif_skb_features(struct sk_buff *skb) 2143 netdev_features_t netif_skb_features(struct sk_buff *skb)
2144 { 2144 {
2145 __be16 protocol = skb->protocol; 2145 __be16 protocol = skb->protocol;
2146 netdev_features_t features = skb->dev->features; 2146 netdev_features_t features = skb->dev->features;
2147 2147
2148 if (protocol == htons(ETH_P_8021Q)) { 2148 if (protocol == htons(ETH_P_8021Q)) {
2149 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data; 2149 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
2150 protocol = veh->h_vlan_encapsulated_proto; 2150 protocol = veh->h_vlan_encapsulated_proto;
2151 } else if (!vlan_tx_tag_present(skb)) { 2151 } else if (!vlan_tx_tag_present(skb)) {
2152 return harmonize_features(skb, protocol, features); 2152 return harmonize_features(skb, protocol, features);
2153 } 2153 }
2154 2154
2155 features &= (skb->dev->vlan_features | NETIF_F_HW_VLAN_TX); 2155 features &= (skb->dev->vlan_features | NETIF_F_HW_VLAN_TX);
2156 2156
2157 if (protocol != htons(ETH_P_8021Q)) { 2157 if (protocol != htons(ETH_P_8021Q)) {
2158 return harmonize_features(skb, protocol, features); 2158 return harmonize_features(skb, protocol, features);
2159 } else { 2159 } else {
2160 features &= NETIF_F_SG | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | 2160 features &= NETIF_F_SG | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST |
2161 NETIF_F_GEN_CSUM | NETIF_F_HW_VLAN_TX; 2161 NETIF_F_GEN_CSUM | NETIF_F_HW_VLAN_TX;
2162 return harmonize_features(skb, protocol, features); 2162 return harmonize_features(skb, protocol, features);
2163 } 2163 }
2164 } 2164 }
2165 EXPORT_SYMBOL(netif_skb_features); 2165 EXPORT_SYMBOL(netif_skb_features);
2166 2166
2167 /* 2167 /*
2168 * Returns true if either: 2168 * Returns true if either:
2169 * 1. skb has frag_list and the device doesn't support FRAGLIST, or 2169 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2170 * 2. skb is fragmented and the device does not support SG, or if 2170 * 2. skb is fragmented and the device does not support SG, or if
2171 * at least one of fragments is in highmem and device does not 2171 * at least one of fragments is in highmem and device does not
2172 * support DMA from it. 2172 * support DMA from it.
2173 */ 2173 */
2174 static inline int skb_needs_linearize(struct sk_buff *skb, 2174 static inline int skb_needs_linearize(struct sk_buff *skb,
2175 int features) 2175 int features)
2176 { 2176 {
2177 return skb_is_nonlinear(skb) && 2177 return skb_is_nonlinear(skb) &&
2178 ((skb_has_frag_list(skb) && 2178 ((skb_has_frag_list(skb) &&
2179 !(features & NETIF_F_FRAGLIST)) || 2179 !(features & NETIF_F_FRAGLIST)) ||
2180 (skb_shinfo(skb)->nr_frags && 2180 (skb_shinfo(skb)->nr_frags &&
2181 !(features & NETIF_F_SG))); 2181 !(features & NETIF_F_SG)));
2182 } 2182 }
2183 2183
2184 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 2184 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2185 struct netdev_queue *txq) 2185 struct netdev_queue *txq)
2186 { 2186 {
2187 const struct net_device_ops *ops = dev->netdev_ops; 2187 const struct net_device_ops *ops = dev->netdev_ops;
2188 int rc = NETDEV_TX_OK; 2188 int rc = NETDEV_TX_OK;
2189 unsigned int skb_len; 2189 unsigned int skb_len;
2190 2190
2191 if (likely(!skb->next)) { 2191 if (likely(!skb->next)) {
2192 netdev_features_t features; 2192 netdev_features_t features;
2193 2193
2194 /* 2194 /*
2195 * If device doesn't need skb->dst, release it right now while 2195 * If device doesn't need skb->dst, release it right now while
2196 * its hot in this cpu cache 2196 * its hot in this cpu cache
2197 */ 2197 */
2198 if (dev->priv_flags & IFF_XMIT_DST_RELEASE) 2198 if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
2199 skb_dst_drop(skb); 2199 skb_dst_drop(skb);
2200 2200
2201 if (!list_empty(&ptype_all)) 2201 if (!list_empty(&ptype_all))
2202 dev_queue_xmit_nit(skb, dev); 2202 dev_queue_xmit_nit(skb, dev);
2203 2203
2204 skb_orphan_try(skb); 2204 skb_orphan_try(skb);
2205 2205
2206 features = netif_skb_features(skb); 2206 features = netif_skb_features(skb);
2207 2207
2208 if (vlan_tx_tag_present(skb) && 2208 if (vlan_tx_tag_present(skb) &&
2209 !(features & NETIF_F_HW_VLAN_TX)) { 2209 !(features & NETIF_F_HW_VLAN_TX)) {
2210 skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb)); 2210 skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb));
2211 if (unlikely(!skb)) 2211 if (unlikely(!skb))
2212 goto out; 2212 goto out;
2213 2213
2214 skb->vlan_tci = 0; 2214 skb->vlan_tci = 0;
2215 } 2215 }
2216 2216
2217 if (netif_needs_gso(skb, features)) { 2217 if (netif_needs_gso(skb, features)) {
2218 if (unlikely(dev_gso_segment(skb, features))) 2218 if (unlikely(dev_gso_segment(skb, features)))
2219 goto out_kfree_skb; 2219 goto out_kfree_skb;
2220 if (skb->next) 2220 if (skb->next)
2221 goto gso; 2221 goto gso;
2222 } else { 2222 } else {
2223 if (skb_needs_linearize(skb, features) && 2223 if (skb_needs_linearize(skb, features) &&
2224 __skb_linearize(skb)) 2224 __skb_linearize(skb))
2225 goto out_kfree_skb; 2225 goto out_kfree_skb;
2226 2226
2227 /* If packet is not checksummed and device does not 2227 /* If packet is not checksummed and device does not
2228 * support checksumming for this protocol, complete 2228 * support checksumming for this protocol, complete
2229 * checksumming here. 2229 * checksumming here.
2230 */ 2230 */
2231 if (skb->ip_summed == CHECKSUM_PARTIAL) { 2231 if (skb->ip_summed == CHECKSUM_PARTIAL) {
2232 skb_set_transport_header(skb, 2232 skb_set_transport_header(skb,
2233 skb_checksum_start_offset(skb)); 2233 skb_checksum_start_offset(skb));
2234 if (!(features & NETIF_F_ALL_CSUM) && 2234 if (!(features & NETIF_F_ALL_CSUM) &&
2235 skb_checksum_help(skb)) 2235 skb_checksum_help(skb))
2236 goto out_kfree_skb; 2236 goto out_kfree_skb;
2237 } 2237 }
2238 } 2238 }
2239 2239
2240 skb_len = skb->len; 2240 skb_len = skb->len;
2241 rc = ops->ndo_start_xmit(skb, dev); 2241 rc = ops->ndo_start_xmit(skb, dev);
2242 trace_net_dev_xmit(skb, rc, dev, skb_len); 2242 trace_net_dev_xmit(skb, rc, dev, skb_len);
2243 if (rc == NETDEV_TX_OK) 2243 if (rc == NETDEV_TX_OK)
2244 txq_trans_update(txq); 2244 txq_trans_update(txq);
2245 return rc; 2245 return rc;
2246 } 2246 }
2247 2247
2248 gso: 2248 gso:
2249 do { 2249 do {
2250 struct sk_buff *nskb = skb->next; 2250 struct sk_buff *nskb = skb->next;
2251 2251
2252 skb->next = nskb->next; 2252 skb->next = nskb->next;
2253 nskb->next = NULL; 2253 nskb->next = NULL;
2254 2254
2255 /* 2255 /*
2256 * If device doesn't need nskb->dst, release it right now while 2256 * If device doesn't need nskb->dst, release it right now while
2257 * its hot in this cpu cache 2257 * its hot in this cpu cache
2258 */ 2258 */
2259 if (dev->priv_flags & IFF_XMIT_DST_RELEASE) 2259 if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
2260 skb_dst_drop(nskb); 2260 skb_dst_drop(nskb);
2261 2261
2262 skb_len = nskb->len; 2262 skb_len = nskb->len;
2263 rc = ops->ndo_start_xmit(nskb, dev); 2263 rc = ops->ndo_start_xmit(nskb, dev);
2264 trace_net_dev_xmit(nskb, rc, dev, skb_len); 2264 trace_net_dev_xmit(nskb, rc, dev, skb_len);
2265 if (unlikely(rc != NETDEV_TX_OK)) { 2265 if (unlikely(rc != NETDEV_TX_OK)) {
2266 if (rc & ~NETDEV_TX_MASK) 2266 if (rc & ~NETDEV_TX_MASK)
2267 goto out_kfree_gso_skb; 2267 goto out_kfree_gso_skb;
2268 nskb->next = skb->next; 2268 nskb->next = skb->next;
2269 skb->next = nskb; 2269 skb->next = nskb;
2270 return rc; 2270 return rc;
2271 } 2271 }
2272 txq_trans_update(txq); 2272 txq_trans_update(txq);
2273 if (unlikely(netif_xmit_stopped(txq) && skb->next)) 2273 if (unlikely(netif_xmit_stopped(txq) && skb->next))
2274 return NETDEV_TX_BUSY; 2274 return NETDEV_TX_BUSY;
2275 } while (skb->next); 2275 } while (skb->next);
2276 2276
2277 out_kfree_gso_skb: 2277 out_kfree_gso_skb:
2278 if (likely(skb->next == NULL)) 2278 if (likely(skb->next == NULL))
2279 skb->destructor = DEV_GSO_CB(skb)->destructor; 2279 skb->destructor = DEV_GSO_CB(skb)->destructor;
2280 out_kfree_skb: 2280 out_kfree_skb:
2281 kfree_skb(skb); 2281 kfree_skb(skb);
2282 out: 2282 out:
2283 return rc; 2283 return rc;
2284 } 2284 }
2285 2285
2286 static u32 hashrnd __read_mostly; 2286 static u32 hashrnd __read_mostly;
2287 2287
2288 /* 2288 /*
2289 * Returns a Tx hash based on the given packet descriptor a Tx queues' number 2289 * Returns a Tx hash based on the given packet descriptor a Tx queues' number
2290 * to be used as a distribution range. 2290 * to be used as a distribution range.
2291 */ 2291 */
2292 u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb, 2292 u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2293 unsigned int num_tx_queues) 2293 unsigned int num_tx_queues)
2294 { 2294 {
2295 u32 hash; 2295 u32 hash;
2296 u16 qoffset = 0; 2296 u16 qoffset = 0;
2297 u16 qcount = num_tx_queues; 2297 u16 qcount = num_tx_queues;
2298 2298
2299 if (skb_rx_queue_recorded(skb)) { 2299 if (skb_rx_queue_recorded(skb)) {
2300 hash = skb_get_rx_queue(skb); 2300 hash = skb_get_rx_queue(skb);
2301 while (unlikely(hash >= num_tx_queues)) 2301 while (unlikely(hash >= num_tx_queues))
2302 hash -= num_tx_queues; 2302 hash -= num_tx_queues;
2303 return hash; 2303 return hash;
2304 } 2304 }
2305 2305
2306 if (dev->num_tc) { 2306 if (dev->num_tc) {
2307 u8 tc = netdev_get_prio_tc_map(dev, skb->priority); 2307 u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
2308 qoffset = dev->tc_to_txq[tc].offset; 2308 qoffset = dev->tc_to_txq[tc].offset;
2309 qcount = dev->tc_to_txq[tc].count; 2309 qcount = dev->tc_to_txq[tc].count;
2310 } 2310 }
2311 2311
2312 if (skb->sk && skb->sk->sk_hash) 2312 if (skb->sk && skb->sk->sk_hash)
2313 hash = skb->sk->sk_hash; 2313 hash = skb->sk->sk_hash;
2314 else 2314 else
2315 hash = (__force u16) skb->protocol ^ skb->rxhash; 2315 hash = (__force u16) skb->protocol ^ skb->rxhash;
2316 hash = jhash_1word(hash, hashrnd); 2316 hash = jhash_1word(hash, hashrnd);
2317 2317
2318 return (u16) (((u64) hash * qcount) >> 32) + qoffset; 2318 return (u16) (((u64) hash * qcount) >> 32) + qoffset;
2319 } 2319 }
2320 EXPORT_SYMBOL(__skb_tx_hash); 2320 EXPORT_SYMBOL(__skb_tx_hash);
2321 2321
2322 static inline u16 dev_cap_txqueue(struct net_device *dev, u16 queue_index) 2322 static inline u16 dev_cap_txqueue(struct net_device *dev, u16 queue_index)
2323 { 2323 {
2324 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 2324 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2325 if (net_ratelimit()) { 2325 if (net_ratelimit()) {
2326 pr_warn("%s selects TX queue %d, but real number of TX queues is %d\n", 2326 pr_warn("%s selects TX queue %d, but real number of TX queues is %d\n",
2327 dev->name, queue_index, 2327 dev->name, queue_index,
2328 dev->real_num_tx_queues); 2328 dev->real_num_tx_queues);
2329 } 2329 }
2330 return 0; 2330 return 0;
2331 } 2331 }
2332 return queue_index; 2332 return queue_index;
2333 } 2333 }
2334 2334
2335 static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb) 2335 static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
2336 { 2336 {
2337 #ifdef CONFIG_XPS 2337 #ifdef CONFIG_XPS
2338 struct xps_dev_maps *dev_maps; 2338 struct xps_dev_maps *dev_maps;
2339 struct xps_map *map; 2339 struct xps_map *map;
2340 int queue_index = -1; 2340 int queue_index = -1;
2341 2341
2342 rcu_read_lock(); 2342 rcu_read_lock();
2343 dev_maps = rcu_dereference(dev->xps_maps); 2343 dev_maps = rcu_dereference(dev->xps_maps);
2344 if (dev_maps) { 2344 if (dev_maps) {
2345 map = rcu_dereference( 2345 map = rcu_dereference(
2346 dev_maps->cpu_map[raw_smp_processor_id()]); 2346 dev_maps->cpu_map[raw_smp_processor_id()]);
2347 if (map) { 2347 if (map) {
2348 if (map->len == 1) 2348 if (map->len == 1)
2349 queue_index = map->queues[0]; 2349 queue_index = map->queues[0];
2350 else { 2350 else {
2351 u32 hash; 2351 u32 hash;
2352 if (skb->sk && skb->sk->sk_hash) 2352 if (skb->sk && skb->sk->sk_hash)
2353 hash = skb->sk->sk_hash; 2353 hash = skb->sk->sk_hash;
2354 else 2354 else
2355 hash = (__force u16) skb->protocol ^ 2355 hash = (__force u16) skb->protocol ^
2356 skb->rxhash; 2356 skb->rxhash;
2357 hash = jhash_1word(hash, hashrnd); 2357 hash = jhash_1word(hash, hashrnd);
2358 queue_index = map->queues[ 2358 queue_index = map->queues[
2359 ((u64)hash * map->len) >> 32]; 2359 ((u64)hash * map->len) >> 32];
2360 } 2360 }
2361 if (unlikely(queue_index >= dev->real_num_tx_queues)) 2361 if (unlikely(queue_index >= dev->real_num_tx_queues))
2362 queue_index = -1; 2362 queue_index = -1;
2363 } 2363 }
2364 } 2364 }
2365 rcu_read_unlock(); 2365 rcu_read_unlock();
2366 2366
2367 return queue_index; 2367 return queue_index;
2368 #else 2368 #else
2369 return -1; 2369 return -1;
2370 #endif 2370 #endif
2371 } 2371 }
2372 2372
2373 static struct netdev_queue *dev_pick_tx(struct net_device *dev, 2373 static struct netdev_queue *dev_pick_tx(struct net_device *dev,
2374 struct sk_buff *skb) 2374 struct sk_buff *skb)
2375 { 2375 {
2376 int queue_index; 2376 int queue_index;
2377 const struct net_device_ops *ops = dev->netdev_ops; 2377 const struct net_device_ops *ops = dev->netdev_ops;
2378 2378
2379 if (dev->real_num_tx_queues == 1) 2379 if (dev->real_num_tx_queues == 1)
2380 queue_index = 0; 2380 queue_index = 0;
2381 else if (ops->ndo_select_queue) { 2381 else if (ops->ndo_select_queue) {
2382 queue_index = ops->ndo_select_queue(dev, skb); 2382 queue_index = ops->ndo_select_queue(dev, skb);
2383 queue_index = dev_cap_txqueue(dev, queue_index); 2383 queue_index = dev_cap_txqueue(dev, queue_index);
2384 } else { 2384 } else {
2385 struct sock *sk = skb->sk; 2385 struct sock *sk = skb->sk;
2386 queue_index = sk_tx_queue_get(sk); 2386 queue_index = sk_tx_queue_get(sk);
2387 2387
2388 if (queue_index < 0 || skb->ooo_okay || 2388 if (queue_index < 0 || skb->ooo_okay ||
2389 queue_index >= dev->real_num_tx_queues) { 2389 queue_index >= dev->real_num_tx_queues) {
2390 int old_index = queue_index; 2390 int old_index = queue_index;
2391 2391
2392 queue_index = get_xps_queue(dev, skb); 2392 queue_index = get_xps_queue(dev, skb);
2393 if (queue_index < 0) 2393 if (queue_index < 0)
2394 queue_index = skb_tx_hash(dev, skb); 2394 queue_index = skb_tx_hash(dev, skb);
2395 2395
2396 if (queue_index != old_index && sk) { 2396 if (queue_index != old_index && sk) {
2397 struct dst_entry *dst = 2397 struct dst_entry *dst =
2398 rcu_dereference_check(sk->sk_dst_cache, 1); 2398 rcu_dereference_check(sk->sk_dst_cache, 1);
2399 2399
2400 if (dst && skb_dst(skb) == dst) 2400 if (dst && skb_dst(skb) == dst)
2401 sk_tx_queue_set(sk, queue_index); 2401 sk_tx_queue_set(sk, queue_index);
2402 } 2402 }
2403 } 2403 }
2404 } 2404 }
2405 2405
2406 skb_set_queue_mapping(skb, queue_index); 2406 skb_set_queue_mapping(skb, queue_index);
2407 return netdev_get_tx_queue(dev, queue_index); 2407 return netdev_get_tx_queue(dev, queue_index);
2408 } 2408 }
2409 2409
2410 static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q, 2410 static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q,
2411 struct net_device *dev, 2411 struct net_device *dev,
2412 struct netdev_queue *txq) 2412 struct netdev_queue *txq)
2413 { 2413 {
2414 spinlock_t *root_lock = qdisc_lock(q); 2414 spinlock_t *root_lock = qdisc_lock(q);
2415 bool contended; 2415 bool contended;
2416 int rc; 2416 int rc;
2417 2417
2418 qdisc_skb_cb(skb)->pkt_len = skb->len; 2418 qdisc_skb_cb(skb)->pkt_len = skb->len;
2419 qdisc_calculate_pkt_len(skb, q); 2419 qdisc_calculate_pkt_len(skb, q);
2420 /* 2420 /*
2421 * Heuristic to force contended enqueues to serialize on a 2421 * Heuristic to force contended enqueues to serialize on a
2422 * separate lock before trying to get qdisc main lock. 2422 * separate lock before trying to get qdisc main lock.
2423 * This permits __QDISC_STATE_RUNNING owner to get the lock more often 2423 * This permits __QDISC_STATE_RUNNING owner to get the lock more often
2424 * and dequeue packets faster. 2424 * and dequeue packets faster.
2425 */ 2425 */
2426 contended = qdisc_is_running(q); 2426 contended = qdisc_is_running(q);
2427 if (unlikely(contended)) 2427 if (unlikely(contended))
2428 spin_lock(&q->busylock); 2428 spin_lock(&q->busylock);
2429 2429
2430 spin_lock(root_lock); 2430 spin_lock(root_lock);
2431 if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) { 2431 if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
2432 kfree_skb(skb); 2432 kfree_skb(skb);
2433 rc = NET_XMIT_DROP; 2433 rc = NET_XMIT_DROP;
2434 } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) && 2434 } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
2435 qdisc_run_begin(q)) { 2435 qdisc_run_begin(q)) {
2436 /* 2436 /*
2437 * This is a work-conserving queue; there are no old skbs 2437 * This is a work-conserving queue; there are no old skbs
2438 * waiting to be sent out; and the qdisc is not running - 2438 * waiting to be sent out; and the qdisc is not running -
2439 * xmit the skb directly. 2439 * xmit the skb directly.
2440 */ 2440 */
2441 if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE)) 2441 if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE))
2442 skb_dst_force(skb); 2442 skb_dst_force(skb);
2443 2443
2444 qdisc_bstats_update(q, skb); 2444 qdisc_bstats_update(q, skb);
2445 2445
2446 if (sch_direct_xmit(skb, q, dev, txq, root_lock)) { 2446 if (sch_direct_xmit(skb, q, dev, txq, root_lock)) {
2447 if (unlikely(contended)) { 2447 if (unlikely(contended)) {
2448 spin_unlock(&q->busylock); 2448 spin_unlock(&q->busylock);
2449 contended = false; 2449 contended = false;
2450 } 2450 }
2451 __qdisc_run(q); 2451 __qdisc_run(q);
2452 } else 2452 } else
2453 qdisc_run_end(q); 2453 qdisc_run_end(q);
2454 2454
2455 rc = NET_XMIT_SUCCESS; 2455 rc = NET_XMIT_SUCCESS;
2456 } else { 2456 } else {
2457 skb_dst_force(skb); 2457 skb_dst_force(skb);
2458 rc = q->enqueue(skb, q) & NET_XMIT_MASK; 2458 rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2459 if (qdisc_run_begin(q)) { 2459 if (qdisc_run_begin(q)) {
2460 if (unlikely(contended)) { 2460 if (unlikely(contended)) {
2461 spin_unlock(&q->busylock); 2461 spin_unlock(&q->busylock);
2462 contended = false; 2462 contended = false;
2463 } 2463 }
2464 __qdisc_run(q); 2464 __qdisc_run(q);
2465 } 2465 }
2466 } 2466 }
2467 spin_unlock(root_lock); 2467 spin_unlock(root_lock);
2468 if (unlikely(contended)) 2468 if (unlikely(contended))
2469 spin_unlock(&q->busylock); 2469 spin_unlock(&q->busylock);
2470 return rc; 2470 return rc;
2471 } 2471 }
2472 2472
2473 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP) 2473 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
2474 static void skb_update_prio(struct sk_buff *skb) 2474 static void skb_update_prio(struct sk_buff *skb)
2475 { 2475 {
2476 struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap); 2476 struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2477 2477
2478 if ((!skb->priority) && (skb->sk) && map) 2478 if ((!skb->priority) && (skb->sk) && map)
2479 skb->priority = map->priomap[skb->sk->sk_cgrp_prioidx]; 2479 skb->priority = map->priomap[skb->sk->sk_cgrp_prioidx];
2480 } 2480 }
2481 #else 2481 #else
2482 #define skb_update_prio(skb) 2482 #define skb_update_prio(skb)
2483 #endif 2483 #endif
2484 2484
2485 static DEFINE_PER_CPU(int, xmit_recursion); 2485 static DEFINE_PER_CPU(int, xmit_recursion);
2486 #define RECURSION_LIMIT 10 2486 #define RECURSION_LIMIT 10
2487 2487
2488 /** 2488 /**
2489 * dev_queue_xmit - transmit a buffer 2489 * dev_queue_xmit - transmit a buffer
2490 * @skb: buffer to transmit 2490 * @skb: buffer to transmit
2491 * 2491 *
2492 * Queue a buffer for transmission to a network device. The caller must 2492 * Queue a buffer for transmission to a network device. The caller must
2493 * have set the device and priority and built the buffer before calling 2493 * have set the device and priority and built the buffer before calling
2494 * this function. The function can be called from an interrupt. 2494 * this function. The function can be called from an interrupt.
2495 * 2495 *
2496 * A negative errno code is returned on a failure. A success does not 2496 * A negative errno code is returned on a failure. A success does not
2497 * guarantee the frame will be transmitted as it may be dropped due 2497 * guarantee the frame will be transmitted as it may be dropped due
2498 * to congestion or traffic shaping. 2498 * to congestion or traffic shaping.
2499 * 2499 *
2500 * ----------------------------------------------------------------------------------- 2500 * -----------------------------------------------------------------------------------
2501 * I notice this method can also return errors from the queue disciplines, 2501 * I notice this method can also return errors from the queue disciplines,
2502 * including NET_XMIT_DROP, which is a positive value. So, errors can also 2502 * including NET_XMIT_DROP, which is a positive value. So, errors can also
2503 * be positive. 2503 * be positive.
2504 * 2504 *
2505 * Regardless of the return value, the skb is consumed, so it is currently 2505 * Regardless of the return value, the skb is consumed, so it is currently
2506 * difficult to retry a send to this method. (You can bump the ref count 2506 * difficult to retry a send to this method. (You can bump the ref count
2507 * before sending to hold a reference for retry if you are careful.) 2507 * before sending to hold a reference for retry if you are careful.)
2508 * 2508 *
2509 * When calling this method, interrupts MUST be enabled. This is because 2509 * When calling this method, interrupts MUST be enabled. This is because
2510 * the BH enable code must have IRQs enabled so that it will not deadlock. 2510 * the BH enable code must have IRQs enabled so that it will not deadlock.
2511 * --BLG 2511 * --BLG
2512 */ 2512 */
2513 int dev_queue_xmit(struct sk_buff *skb) 2513 int dev_queue_xmit(struct sk_buff *skb)
2514 { 2514 {
2515 struct net_device *dev = skb->dev; 2515 struct net_device *dev = skb->dev;
2516 struct netdev_queue *txq; 2516 struct netdev_queue *txq;
2517 struct Qdisc *q; 2517 struct Qdisc *q;
2518 int rc = -ENOMEM; 2518 int rc = -ENOMEM;
2519 2519
2520 /* Disable soft irqs for various locks below. Also 2520 /* Disable soft irqs for various locks below. Also
2521 * stops preemption for RCU. 2521 * stops preemption for RCU.
2522 */ 2522 */
2523 rcu_read_lock_bh(); 2523 rcu_read_lock_bh();
2524 2524
2525 skb_update_prio(skb); 2525 skb_update_prio(skb);
2526 2526
2527 txq = dev_pick_tx(dev, skb); 2527 txq = dev_pick_tx(dev, skb);
2528 q = rcu_dereference_bh(txq->qdisc); 2528 q = rcu_dereference_bh(txq->qdisc);
2529 2529
2530 #ifdef CONFIG_NET_CLS_ACT 2530 #ifdef CONFIG_NET_CLS_ACT
2531 skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS); 2531 skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
2532 #endif 2532 #endif
2533 trace_net_dev_queue(skb); 2533 trace_net_dev_queue(skb);
2534 if (q->enqueue) { 2534 if (q->enqueue) {
2535 rc = __dev_xmit_skb(skb, q, dev, txq); 2535 rc = __dev_xmit_skb(skb, q, dev, txq);
2536 goto out; 2536 goto out;
2537 } 2537 }
2538 2538
2539 /* The device has no queue. Common case for software devices: 2539 /* The device has no queue. Common case for software devices:
2540 loopback, all the sorts of tunnels... 2540 loopback, all the sorts of tunnels...
2541 2541
2542 Really, it is unlikely that netif_tx_lock protection is necessary 2542 Really, it is unlikely that netif_tx_lock protection is necessary
2543 here. (f.e. loopback and IP tunnels are clean ignoring statistics 2543 here. (f.e. loopback and IP tunnels are clean ignoring statistics
2544 counters.) 2544 counters.)
2545 However, it is possible, that they rely on protection 2545 However, it is possible, that they rely on protection
2546 made by us here. 2546 made by us here.
2547 2547
2548 Check this and shot the lock. It is not prone from deadlocks. 2548 Check this and shot the lock. It is not prone from deadlocks.
2549 Either shot noqueue qdisc, it is even simpler 8) 2549 Either shot noqueue qdisc, it is even simpler 8)
2550 */ 2550 */
2551 if (dev->flags & IFF_UP) { 2551 if (dev->flags & IFF_UP) {
2552 int cpu = smp_processor_id(); /* ok because BHs are off */ 2552 int cpu = smp_processor_id(); /* ok because BHs are off */
2553 2553
2554 if (txq->xmit_lock_owner != cpu) { 2554 if (txq->xmit_lock_owner != cpu) {
2555 2555
2556 if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT) 2556 if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
2557 goto recursion_alert; 2557 goto recursion_alert;
2558 2558
2559 HARD_TX_LOCK(dev, txq, cpu); 2559 HARD_TX_LOCK(dev, txq, cpu);
2560 2560
2561 if (!netif_xmit_stopped(txq)) { 2561 if (!netif_xmit_stopped(txq)) {
2562 __this_cpu_inc(xmit_recursion); 2562 __this_cpu_inc(xmit_recursion);
2563 rc = dev_hard_start_xmit(skb, dev, txq); 2563 rc = dev_hard_start_xmit(skb, dev, txq);
2564 __this_cpu_dec(xmit_recursion); 2564 __this_cpu_dec(xmit_recursion);
2565 if (dev_xmit_complete(rc)) { 2565 if (dev_xmit_complete(rc)) {
2566 HARD_TX_UNLOCK(dev, txq); 2566 HARD_TX_UNLOCK(dev, txq);
2567 goto out; 2567 goto out;
2568 } 2568 }
2569 } 2569 }
2570 HARD_TX_UNLOCK(dev, txq); 2570 HARD_TX_UNLOCK(dev, txq);
2571 if (net_ratelimit()) 2571 if (net_ratelimit())
2572 pr_crit("Virtual device %s asks to queue packet!\n", 2572 pr_crit("Virtual device %s asks to queue packet!\n",
2573 dev->name); 2573 dev->name);
2574 } else { 2574 } else {
2575 /* Recursion is detected! It is possible, 2575 /* Recursion is detected! It is possible,
2576 * unfortunately 2576 * unfortunately
2577 */ 2577 */
2578 recursion_alert: 2578 recursion_alert:
2579 if (net_ratelimit()) 2579 if (net_ratelimit())
2580 pr_crit("Dead loop on virtual device %s, fix it urgently!\n", 2580 pr_crit("Dead loop on virtual device %s, fix it urgently!\n",
2581 dev->name); 2581 dev->name);
2582 } 2582 }
2583 } 2583 }
2584 2584
2585 rc = -ENETDOWN; 2585 rc = -ENETDOWN;
2586 rcu_read_unlock_bh(); 2586 rcu_read_unlock_bh();
2587 2587
2588 kfree_skb(skb); 2588 kfree_skb(skb);
2589 return rc; 2589 return rc;
2590 out: 2590 out:
2591 rcu_read_unlock_bh(); 2591 rcu_read_unlock_bh();
2592 return rc; 2592 return rc;
2593 } 2593 }
2594 EXPORT_SYMBOL(dev_queue_xmit); 2594 EXPORT_SYMBOL(dev_queue_xmit);
2595 2595
2596 2596
2597 /*======================================================================= 2597 /*=======================================================================
2598 Receiver routines 2598 Receiver routines
2599 =======================================================================*/ 2599 =======================================================================*/
2600 2600
2601 int netdev_max_backlog __read_mostly = 1000; 2601 int netdev_max_backlog __read_mostly = 1000;
2602 int netdev_tstamp_prequeue __read_mostly = 1; 2602 int netdev_tstamp_prequeue __read_mostly = 1;
2603 int netdev_budget __read_mostly = 300; 2603 int netdev_budget __read_mostly = 300;
2604 int weight_p __read_mostly = 64; /* old backlog weight */ 2604 int weight_p __read_mostly = 64; /* old backlog weight */
2605 2605
2606 /* Called with irq disabled */ 2606 /* Called with irq disabled */
2607 static inline void ____napi_schedule(struct softnet_data *sd, 2607 static inline void ____napi_schedule(struct softnet_data *sd,
2608 struct napi_struct *napi) 2608 struct napi_struct *napi)
2609 { 2609 {
2610 list_add_tail(&napi->poll_list, &sd->poll_list); 2610 list_add_tail(&napi->poll_list, &sd->poll_list);
2611 __raise_softirq_irqoff(NET_RX_SOFTIRQ); 2611 __raise_softirq_irqoff(NET_RX_SOFTIRQ);
2612 } 2612 }
2613 2613
2614 /* 2614 /*
2615 * __skb_get_rxhash: calculate a flow hash based on src/dst addresses 2615 * __skb_get_rxhash: calculate a flow hash based on src/dst addresses
2616 * and src/dst port numbers. Sets rxhash in skb to non-zero hash value 2616 * and src/dst port numbers. Sets rxhash in skb to non-zero hash value
2617 * on success, zero indicates no valid hash. Also, sets l4_rxhash in skb 2617 * on success, zero indicates no valid hash. Also, sets l4_rxhash in skb
2618 * if hash is a canonical 4-tuple hash over transport ports. 2618 * if hash is a canonical 4-tuple hash over transport ports.
2619 */ 2619 */
2620 void __skb_get_rxhash(struct sk_buff *skb) 2620 void __skb_get_rxhash(struct sk_buff *skb)
2621 { 2621 {
2622 struct flow_keys keys; 2622 struct flow_keys keys;
2623 u32 hash; 2623 u32 hash;
2624 2624
2625 if (!skb_flow_dissect(skb, &keys)) 2625 if (!skb_flow_dissect(skb, &keys))
2626 return; 2626 return;
2627 2627
2628 if (keys.ports) { 2628 if (keys.ports) {
2629 if ((__force u16)keys.port16[1] < (__force u16)keys.port16[0]) 2629 if ((__force u16)keys.port16[1] < (__force u16)keys.port16[0])
2630 swap(keys.port16[0], keys.port16[1]); 2630 swap(keys.port16[0], keys.port16[1]);
2631 skb->l4_rxhash = 1; 2631 skb->l4_rxhash = 1;
2632 } 2632 }
2633 2633
2634 /* get a consistent hash (same value on both flow directions) */ 2634 /* get a consistent hash (same value on both flow directions) */
2635 if ((__force u32)keys.dst < (__force u32)keys.src) 2635 if ((__force u32)keys.dst < (__force u32)keys.src)
2636 swap(keys.dst, keys.src); 2636 swap(keys.dst, keys.src);
2637 2637
2638 hash = jhash_3words((__force u32)keys.dst, 2638 hash = jhash_3words((__force u32)keys.dst,
2639 (__force u32)keys.src, 2639 (__force u32)keys.src,
2640 (__force u32)keys.ports, hashrnd); 2640 (__force u32)keys.ports, hashrnd);
2641 if (!hash) 2641 if (!hash)
2642 hash = 1; 2642 hash = 1;
2643 2643
2644 skb->rxhash = hash; 2644 skb->rxhash = hash;
2645 } 2645 }
2646 EXPORT_SYMBOL(__skb_get_rxhash); 2646 EXPORT_SYMBOL(__skb_get_rxhash);
2647 2647
2648 #ifdef CONFIG_RPS 2648 #ifdef CONFIG_RPS
2649 2649
2650 /* One global table that all flow-based protocols share. */ 2650 /* One global table that all flow-based protocols share. */
2651 struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly; 2651 struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
2652 EXPORT_SYMBOL(rps_sock_flow_table); 2652 EXPORT_SYMBOL(rps_sock_flow_table);
2653 2653
2654 struct static_key rps_needed __read_mostly; 2654 struct static_key rps_needed __read_mostly;
2655 2655
2656 static struct rps_dev_flow * 2656 static struct rps_dev_flow *
2657 set_rps_cpu(struct net_device *dev, struct sk_buff *skb, 2657 set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
2658 struct rps_dev_flow *rflow, u16 next_cpu) 2658 struct rps_dev_flow *rflow, u16 next_cpu)
2659 { 2659 {
2660 if (next_cpu != RPS_NO_CPU) { 2660 if (next_cpu != RPS_NO_CPU) {
2661 #ifdef CONFIG_RFS_ACCEL 2661 #ifdef CONFIG_RFS_ACCEL
2662 struct netdev_rx_queue *rxqueue; 2662 struct netdev_rx_queue *rxqueue;
2663 struct rps_dev_flow_table *flow_table; 2663 struct rps_dev_flow_table *flow_table;
2664 struct rps_dev_flow *old_rflow; 2664 struct rps_dev_flow *old_rflow;
2665 u32 flow_id; 2665 u32 flow_id;
2666 u16 rxq_index; 2666 u16 rxq_index;
2667 int rc; 2667 int rc;
2668 2668
2669 /* Should we steer this flow to a different hardware queue? */ 2669 /* Should we steer this flow to a different hardware queue? */
2670 if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap || 2670 if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
2671 !(dev->features & NETIF_F_NTUPLE)) 2671 !(dev->features & NETIF_F_NTUPLE))
2672 goto out; 2672 goto out;
2673 rxq_index = cpu_rmap_lookup_index(dev->rx_cpu_rmap, next_cpu); 2673 rxq_index = cpu_rmap_lookup_index(dev->rx_cpu_rmap, next_cpu);
2674 if (rxq_index == skb_get_rx_queue(skb)) 2674 if (rxq_index == skb_get_rx_queue(skb))
2675 goto out; 2675 goto out;
2676 2676
2677 rxqueue = dev->_rx + rxq_index; 2677 rxqueue = dev->_rx + rxq_index;
2678 flow_table = rcu_dereference(rxqueue->rps_flow_table); 2678 flow_table = rcu_dereference(rxqueue->rps_flow_table);
2679 if (!flow_table) 2679 if (!flow_table)
2680 goto out; 2680 goto out;
2681 flow_id = skb->rxhash & flow_table->mask; 2681 flow_id = skb->rxhash & flow_table->mask;
2682 rc = dev->netdev_ops->ndo_rx_flow_steer(dev, skb, 2682 rc = dev->netdev_ops->ndo_rx_flow_steer(dev, skb,
2683 rxq_index, flow_id); 2683 rxq_index, flow_id);
2684 if (rc < 0) 2684 if (rc < 0)
2685 goto out; 2685 goto out;
2686 old_rflow = rflow; 2686 old_rflow = rflow;
2687 rflow = &flow_table->flows[flow_id]; 2687 rflow = &flow_table->flows[flow_id];
2688 rflow->filter = rc; 2688 rflow->filter = rc;
2689 if (old_rflow->filter == rflow->filter) 2689 if (old_rflow->filter == rflow->filter)
2690 old_rflow->filter = RPS_NO_FILTER; 2690 old_rflow->filter = RPS_NO_FILTER;
2691 out: 2691 out:
2692 #endif 2692 #endif
2693 rflow->last_qtail = 2693 rflow->last_qtail =
2694 per_cpu(softnet_data, next_cpu).input_queue_head; 2694 per_cpu(softnet_data, next_cpu).input_queue_head;
2695 } 2695 }
2696 2696
2697 rflow->cpu = next_cpu; 2697 rflow->cpu = next_cpu;
2698 return rflow; 2698 return rflow;
2699 } 2699 }
2700 2700
2701 /* 2701 /*
2702 * get_rps_cpu is called from netif_receive_skb and returns the target 2702 * get_rps_cpu is called from netif_receive_skb and returns the target
2703 * CPU from the RPS map of the receiving queue for a given skb. 2703 * CPU from the RPS map of the receiving queue for a given skb.
2704 * rcu_read_lock must be held on entry. 2704 * rcu_read_lock must be held on entry.
2705 */ 2705 */
2706 static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb, 2706 static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb,
2707 struct rps_dev_flow **rflowp) 2707 struct rps_dev_flow **rflowp)
2708 { 2708 {
2709 struct netdev_rx_queue *rxqueue; 2709 struct netdev_rx_queue *rxqueue;
2710 struct rps_map *map; 2710 struct rps_map *map;
2711 struct rps_dev_flow_table *flow_table; 2711 struct rps_dev_flow_table *flow_table;
2712 struct rps_sock_flow_table *sock_flow_table; 2712 struct rps_sock_flow_table *sock_flow_table;
2713 int cpu = -1; 2713 int cpu = -1;
2714 u16 tcpu; 2714 u16 tcpu;
2715 2715
2716 if (skb_rx_queue_recorded(skb)) { 2716 if (skb_rx_queue_recorded(skb)) {
2717 u16 index = skb_get_rx_queue(skb); 2717 u16 index = skb_get_rx_queue(skb);
2718 if (unlikely(index >= dev->real_num_rx_queues)) { 2718 if (unlikely(index >= dev->real_num_rx_queues)) {
2719 WARN_ONCE(dev->real_num_rx_queues > 1, 2719 WARN_ONCE(dev->real_num_rx_queues > 1,
2720 "%s received packet on queue %u, but number " 2720 "%s received packet on queue %u, but number "
2721 "of RX queues is %u\n", 2721 "of RX queues is %u\n",
2722 dev->name, index, dev->real_num_rx_queues); 2722 dev->name, index, dev->real_num_rx_queues);
2723 goto done; 2723 goto done;
2724 } 2724 }
2725 rxqueue = dev->_rx + index; 2725 rxqueue = dev->_rx + index;
2726 } else 2726 } else
2727 rxqueue = dev->_rx; 2727 rxqueue = dev->_rx;
2728 2728
2729 map = rcu_dereference(rxqueue->rps_map); 2729 map = rcu_dereference(rxqueue->rps_map);
2730 if (map) { 2730 if (map) {
2731 if (map->len == 1 && 2731 if (map->len == 1 &&
2732 !rcu_access_pointer(rxqueue->rps_flow_table)) { 2732 !rcu_access_pointer(rxqueue->rps_flow_table)) {
2733 tcpu = map->cpus[0]; 2733 tcpu = map->cpus[0];
2734 if (cpu_online(tcpu)) 2734 if (cpu_online(tcpu))
2735 cpu = tcpu; 2735 cpu = tcpu;
2736 goto done; 2736 goto done;
2737 } 2737 }
2738 } else if (!rcu_access_pointer(rxqueue->rps_flow_table)) { 2738 } else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
2739 goto done; 2739 goto done;
2740 } 2740 }
2741 2741
2742 skb_reset_network_header(skb); 2742 skb_reset_network_header(skb);
2743 if (!skb_get_rxhash(skb)) 2743 if (!skb_get_rxhash(skb))
2744 goto done; 2744 goto done;
2745 2745
2746 flow_table = rcu_dereference(rxqueue->rps_flow_table); 2746 flow_table = rcu_dereference(rxqueue->rps_flow_table);
2747 sock_flow_table = rcu_dereference(rps_sock_flow_table); 2747 sock_flow_table = rcu_dereference(rps_sock_flow_table);
2748 if (flow_table && sock_flow_table) { 2748 if (flow_table && sock_flow_table) {
2749 u16 next_cpu; 2749 u16 next_cpu;
2750 struct rps_dev_flow *rflow; 2750 struct rps_dev_flow *rflow;
2751 2751
2752 rflow = &flow_table->flows[skb->rxhash & flow_table->mask]; 2752 rflow = &flow_table->flows[skb->rxhash & flow_table->mask];
2753 tcpu = rflow->cpu; 2753 tcpu = rflow->cpu;
2754 2754
2755 next_cpu = sock_flow_table->ents[skb->rxhash & 2755 next_cpu = sock_flow_table->ents[skb->rxhash &
2756 sock_flow_table->mask]; 2756 sock_flow_table->mask];
2757 2757
2758 /* 2758 /*
2759 * If the desired CPU (where last recvmsg was done) is 2759 * If the desired CPU (where last recvmsg was done) is
2760 * different from current CPU (one in the rx-queue flow 2760 * different from current CPU (one in the rx-queue flow
2761 * table entry), switch if one of the following holds: 2761 * table entry), switch if one of the following holds:
2762 * - Current CPU is unset (equal to RPS_NO_CPU). 2762 * - Current CPU is unset (equal to RPS_NO_CPU).
2763 * - Current CPU is offline. 2763 * - Current CPU is offline.
2764 * - The current CPU's queue tail has advanced beyond the 2764 * - The current CPU's queue tail has advanced beyond the
2765 * last packet that was enqueued using this table entry. 2765 * last packet that was enqueued using this table entry.
2766 * This guarantees that all previous packets for the flow 2766 * This guarantees that all previous packets for the flow
2767 * have been dequeued, thus preserving in order delivery. 2767 * have been dequeued, thus preserving in order delivery.
2768 */ 2768 */
2769 if (unlikely(tcpu != next_cpu) && 2769 if (unlikely(tcpu != next_cpu) &&
2770 (tcpu == RPS_NO_CPU || !cpu_online(tcpu) || 2770 (tcpu == RPS_NO_CPU || !cpu_online(tcpu) ||
2771 ((int)(per_cpu(softnet_data, tcpu).input_queue_head - 2771 ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
2772 rflow->last_qtail)) >= 0)) 2772 rflow->last_qtail)) >= 0))
2773 rflow = set_rps_cpu(dev, skb, rflow, next_cpu); 2773 rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
2774 2774
2775 if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) { 2775 if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
2776 *rflowp = rflow; 2776 *rflowp = rflow;
2777 cpu = tcpu; 2777 cpu = tcpu;
2778 goto done; 2778 goto done;
2779 } 2779 }
2780 } 2780 }
2781 2781
2782 if (map) { 2782 if (map) {
2783 tcpu = map->cpus[((u64) skb->rxhash * map->len) >> 32]; 2783 tcpu = map->cpus[((u64) skb->rxhash * map->len) >> 32];
2784 2784
2785 if (cpu_online(tcpu)) { 2785 if (cpu_online(tcpu)) {
2786 cpu = tcpu; 2786 cpu = tcpu;
2787 goto done; 2787 goto done;
2788 } 2788 }
2789 } 2789 }
2790 2790
2791 done: 2791 done:
2792 return cpu; 2792 return cpu;
2793 } 2793 }
2794 2794
2795 #ifdef CONFIG_RFS_ACCEL 2795 #ifdef CONFIG_RFS_ACCEL
2796 2796
2797 /** 2797 /**
2798 * rps_may_expire_flow - check whether an RFS hardware filter may be removed 2798 * rps_may_expire_flow - check whether an RFS hardware filter may be removed
2799 * @dev: Device on which the filter was set 2799 * @dev: Device on which the filter was set
2800 * @rxq_index: RX queue index 2800 * @rxq_index: RX queue index
2801 * @flow_id: Flow ID passed to ndo_rx_flow_steer() 2801 * @flow_id: Flow ID passed to ndo_rx_flow_steer()
2802 * @filter_id: Filter ID returned by ndo_rx_flow_steer() 2802 * @filter_id: Filter ID returned by ndo_rx_flow_steer()
2803 * 2803 *
2804 * Drivers that implement ndo_rx_flow_steer() should periodically call 2804 * Drivers that implement ndo_rx_flow_steer() should periodically call
2805 * this function for each installed filter and remove the filters for 2805 * this function for each installed filter and remove the filters for
2806 * which it returns %true. 2806 * which it returns %true.
2807 */ 2807 */
2808 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, 2808 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
2809 u32 flow_id, u16 filter_id) 2809 u32 flow_id, u16 filter_id)
2810 { 2810 {
2811 struct netdev_rx_queue *rxqueue = dev->_rx + rxq_index; 2811 struct netdev_rx_queue *rxqueue = dev->_rx + rxq_index;
2812 struct rps_dev_flow_table *flow_table; 2812 struct rps_dev_flow_table *flow_table;
2813 struct rps_dev_flow *rflow; 2813 struct rps_dev_flow *rflow;
2814 bool expire = true; 2814 bool expire = true;
2815 int cpu; 2815 int cpu;
2816 2816
2817 rcu_read_lock(); 2817 rcu_read_lock();
2818 flow_table = rcu_dereference(rxqueue->rps_flow_table); 2818 flow_table = rcu_dereference(rxqueue->rps_flow_table);
2819 if (flow_table && flow_id <= flow_table->mask) { 2819 if (flow_table && flow_id <= flow_table->mask) {
2820 rflow = &flow_table->flows[flow_id]; 2820 rflow = &flow_table->flows[flow_id];
2821 cpu = ACCESS_ONCE(rflow->cpu); 2821 cpu = ACCESS_ONCE(rflow->cpu);
2822 if (rflow->filter == filter_id && cpu != RPS_NO_CPU && 2822 if (rflow->filter == filter_id && cpu != RPS_NO_CPU &&
2823 ((int)(per_cpu(softnet_data, cpu).input_queue_head - 2823 ((int)(per_cpu(softnet_data, cpu).input_queue_head -
2824 rflow->last_qtail) < 2824 rflow->last_qtail) <
2825 (int)(10 * flow_table->mask))) 2825 (int)(10 * flow_table->mask)))
2826 expire = false; 2826 expire = false;
2827 } 2827 }
2828 rcu_read_unlock(); 2828 rcu_read_unlock();
2829 return expire; 2829 return expire;
2830 } 2830 }
2831 EXPORT_SYMBOL(rps_may_expire_flow); 2831 EXPORT_SYMBOL(rps_may_expire_flow);
2832 2832
2833 #endif /* CONFIG_RFS_ACCEL */ 2833 #endif /* CONFIG_RFS_ACCEL */
2834 2834
2835 /* Called from hardirq (IPI) context */ 2835 /* Called from hardirq (IPI) context */
2836 static void rps_trigger_softirq(void *data) 2836 static void rps_trigger_softirq(void *data)
2837 { 2837 {
2838 struct softnet_data *sd = data; 2838 struct softnet_data *sd = data;
2839 2839
2840 ____napi_schedule(sd, &sd->backlog); 2840 ____napi_schedule(sd, &sd->backlog);
2841 sd->received_rps++; 2841 sd->received_rps++;
2842 } 2842 }
2843 2843
2844 #endif /* CONFIG_RPS */ 2844 #endif /* CONFIG_RPS */
2845 2845
2846 /* 2846 /*
2847 * Check if this softnet_data structure is another cpu one 2847 * Check if this softnet_data structure is another cpu one
2848 * If yes, queue it to our IPI list and return 1 2848 * If yes, queue it to our IPI list and return 1
2849 * If no, return 0 2849 * If no, return 0
2850 */ 2850 */
2851 static int rps_ipi_queued(struct softnet_data *sd) 2851 static int rps_ipi_queued(struct softnet_data *sd)
2852 { 2852 {
2853 #ifdef CONFIG_RPS 2853 #ifdef CONFIG_RPS
2854 struct softnet_data *mysd = &__get_cpu_var(softnet_data); 2854 struct softnet_data *mysd = &__get_cpu_var(softnet_data);
2855 2855
2856 if (sd != mysd) { 2856 if (sd != mysd) {
2857 sd->rps_ipi_next = mysd->rps_ipi_list; 2857 sd->rps_ipi_next = mysd->rps_ipi_list;
2858 mysd->rps_ipi_list = sd; 2858 mysd->rps_ipi_list = sd;
2859 2859
2860 __raise_softirq_irqoff(NET_RX_SOFTIRQ); 2860 __raise_softirq_irqoff(NET_RX_SOFTIRQ);
2861 return 1; 2861 return 1;
2862 } 2862 }
2863 #endif /* CONFIG_RPS */ 2863 #endif /* CONFIG_RPS */
2864 return 0; 2864 return 0;
2865 } 2865 }
2866 2866
2867 /* 2867 /*
2868 * enqueue_to_backlog is called to queue an skb to a per CPU backlog 2868 * enqueue_to_backlog is called to queue an skb to a per CPU backlog
2869 * queue (may be a remote CPU queue). 2869 * queue (may be a remote CPU queue).
2870 */ 2870 */
2871 static int enqueue_to_backlog(struct sk_buff *skb, int cpu, 2871 static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
2872 unsigned int *qtail) 2872 unsigned int *qtail)
2873 { 2873 {
2874 struct softnet_data *sd; 2874 struct softnet_data *sd;
2875 unsigned long flags; 2875 unsigned long flags;
2876 2876
2877 sd = &per_cpu(softnet_data, cpu); 2877 sd = &per_cpu(softnet_data, cpu);
2878 2878
2879 local_irq_save(flags); 2879 local_irq_save(flags);
2880 2880
2881 rps_lock(sd); 2881 rps_lock(sd);
2882 if (skb_queue_len(&sd->input_pkt_queue) <= netdev_max_backlog) { 2882 if (skb_queue_len(&sd->input_pkt_queue) <= netdev_max_backlog) {
2883 if (skb_queue_len(&sd->input_pkt_queue)) { 2883 if (skb_queue_len(&sd->input_pkt_queue)) {
2884 enqueue: 2884 enqueue:
2885 __skb_queue_tail(&sd->input_pkt_queue, skb); 2885 __skb_queue_tail(&sd->input_pkt_queue, skb);
2886 input_queue_tail_incr_save(sd, qtail); 2886 input_queue_tail_incr_save(sd, qtail);
2887 rps_unlock(sd); 2887 rps_unlock(sd);
2888 local_irq_restore(flags); 2888 local_irq_restore(flags);
2889 return NET_RX_SUCCESS; 2889 return NET_RX_SUCCESS;
2890 } 2890 }
2891 2891
2892 /* Schedule NAPI for backlog device 2892 /* Schedule NAPI for backlog device
2893 * We can use non atomic operation since we own the queue lock 2893 * We can use non atomic operation since we own the queue lock
2894 */ 2894 */
2895 if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) { 2895 if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) {
2896 if (!rps_ipi_queued(sd)) 2896 if (!rps_ipi_queued(sd))
2897 ____napi_schedule(sd, &sd->backlog); 2897 ____napi_schedule(sd, &sd->backlog);
2898 } 2898 }
2899 goto enqueue; 2899 goto enqueue;
2900 } 2900 }
2901 2901
2902 sd->dropped++; 2902 sd->dropped++;
2903 rps_unlock(sd); 2903 rps_unlock(sd);
2904 2904
2905 local_irq_restore(flags); 2905 local_irq_restore(flags);
2906 2906
2907 atomic_long_inc(&skb->dev->rx_dropped); 2907 atomic_long_inc(&skb->dev->rx_dropped);
2908 kfree_skb(skb); 2908 kfree_skb(skb);
2909 return NET_RX_DROP; 2909 return NET_RX_DROP;
2910 } 2910 }
2911 2911
2912 /** 2912 /**
2913 * netif_rx - post buffer to the network code 2913 * netif_rx - post buffer to the network code
2914 * @skb: buffer to post 2914 * @skb: buffer to post
2915 * 2915 *
2916 * This function receives a packet from a device driver and queues it for 2916 * This function receives a packet from a device driver and queues it for
2917 * the upper (protocol) levels to process. It always succeeds. The buffer 2917 * the upper (protocol) levels to process. It always succeeds. The buffer
2918 * may be dropped during processing for congestion control or by the 2918 * may be dropped during processing for congestion control or by the
2919 * protocol layers. 2919 * protocol layers.
2920 * 2920 *
2921 * return values: 2921 * return values:
2922 * NET_RX_SUCCESS (no congestion) 2922 * NET_RX_SUCCESS (no congestion)
2923 * NET_RX_DROP (packet was dropped) 2923 * NET_RX_DROP (packet was dropped)
2924 * 2924 *
2925 */ 2925 */
2926 2926
2927 int netif_rx(struct sk_buff *skb) 2927 int netif_rx(struct sk_buff *skb)
2928 { 2928 {
2929 int ret; 2929 int ret;
2930 2930
2931 /* if netpoll wants it, pretend we never saw it */ 2931 /* if netpoll wants it, pretend we never saw it */
2932 if (netpoll_rx(skb)) 2932 if (netpoll_rx(skb))
2933 return NET_RX_DROP; 2933 return NET_RX_DROP;
2934 2934
2935 net_timestamp_check(netdev_tstamp_prequeue, skb); 2935 net_timestamp_check(netdev_tstamp_prequeue, skb);
2936 2936
2937 trace_netif_rx(skb); 2937 trace_netif_rx(skb);
2938 #ifdef CONFIG_RPS 2938 #ifdef CONFIG_RPS
2939 if (static_key_false(&rps_needed)) { 2939 if (static_key_false(&rps_needed)) {
2940 struct rps_dev_flow voidflow, *rflow = &voidflow; 2940 struct rps_dev_flow voidflow, *rflow = &voidflow;
2941 int cpu; 2941 int cpu;
2942 2942
2943 preempt_disable(); 2943 preempt_disable();
2944 rcu_read_lock(); 2944 rcu_read_lock();
2945 2945
2946 cpu = get_rps_cpu(skb->dev, skb, &rflow); 2946 cpu = get_rps_cpu(skb->dev, skb, &rflow);
2947 if (cpu < 0) 2947 if (cpu < 0)
2948 cpu = smp_processor_id(); 2948 cpu = smp_processor_id();
2949 2949
2950 ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail); 2950 ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
2951 2951
2952 rcu_read_unlock(); 2952 rcu_read_unlock();
2953 preempt_enable(); 2953 preempt_enable();
2954 } else 2954 } else
2955 #endif 2955 #endif
2956 { 2956 {
2957 unsigned int qtail; 2957 unsigned int qtail;
2958 ret = enqueue_to_backlog(skb, get_cpu(), &qtail); 2958 ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
2959 put_cpu(); 2959 put_cpu();
2960 } 2960 }
2961 return ret; 2961 return ret;
2962 } 2962 }
2963 EXPORT_SYMBOL(netif_rx); 2963 EXPORT_SYMBOL(netif_rx);
2964 2964
2965 int netif_rx_ni(struct sk_buff *skb) 2965 int netif_rx_ni(struct sk_buff *skb)
2966 { 2966 {
2967 int err; 2967 int err;
2968 2968
2969 preempt_disable(); 2969 preempt_disable();
2970 err = netif_rx(skb); 2970 err = netif_rx(skb);
2971 if (local_softirq_pending()) 2971 if (local_softirq_pending())
2972 do_softirq(); 2972 do_softirq();
2973 preempt_enable(); 2973 preempt_enable();
2974 2974
2975 return err; 2975 return err;
2976 } 2976 }
2977 EXPORT_SYMBOL(netif_rx_ni); 2977 EXPORT_SYMBOL(netif_rx_ni);
2978 2978
2979 static void net_tx_action(struct softirq_action *h) 2979 static void net_tx_action(struct softirq_action *h)
2980 { 2980 {
2981 struct softnet_data *sd = &__get_cpu_var(softnet_data); 2981 struct softnet_data *sd = &__get_cpu_var(softnet_data);
2982 2982
2983 if (sd->completion_queue) { 2983 if (sd->completion_queue) {
2984 struct sk_buff *clist; 2984 struct sk_buff *clist;
2985 2985
2986 local_irq_disable(); 2986 local_irq_disable();
2987 clist = sd->completion_queue; 2987 clist = sd->completion_queue;
2988 sd->completion_queue = NULL; 2988 sd->completion_queue = NULL;
2989 local_irq_enable(); 2989 local_irq_enable();
2990 2990
2991 while (clist) { 2991 while (clist) {
2992 struct sk_buff *skb = clist; 2992 struct sk_buff *skb = clist;
2993 clist = clist->next; 2993 clist = clist->next;
2994 2994
2995 WARN_ON(atomic_read(&skb->users)); 2995 WARN_ON(atomic_read(&skb->users));
2996 trace_kfree_skb(skb, net_tx_action); 2996 trace_kfree_skb(skb, net_tx_action);
2997 __kfree_skb(skb); 2997 __kfree_skb(skb);
2998 } 2998 }
2999 } 2999 }
3000 3000
3001 if (sd->output_queue) { 3001 if (sd->output_queue) {
3002 struct Qdisc *head; 3002 struct Qdisc *head;
3003 3003
3004 local_irq_disable(); 3004 local_irq_disable();
3005 head = sd->output_queue; 3005 head = sd->output_queue;
3006 sd->output_queue = NULL; 3006 sd->output_queue = NULL;
3007 sd->output_queue_tailp = &sd->output_queue; 3007 sd->output_queue_tailp = &sd->output_queue;
3008 local_irq_enable(); 3008 local_irq_enable();
3009 3009
3010 while (head) { 3010 while (head) {
3011 struct Qdisc *q = head; 3011 struct Qdisc *q = head;
3012 spinlock_t *root_lock; 3012 spinlock_t *root_lock;
3013 3013
3014 head = head->next_sched; 3014 head = head->next_sched;
3015 3015
3016 root_lock = qdisc_lock(q); 3016 root_lock = qdisc_lock(q);
3017 if (spin_trylock(root_lock)) { 3017 if (spin_trylock(root_lock)) {
3018 smp_mb__before_clear_bit(); 3018 smp_mb__before_clear_bit();
3019 clear_bit(__QDISC_STATE_SCHED, 3019 clear_bit(__QDISC_STATE_SCHED,
3020 &q->state); 3020 &q->state);
3021 qdisc_run(q); 3021 qdisc_run(q);
3022 spin_unlock(root_lock); 3022 spin_unlock(root_lock);
3023 } else { 3023 } else {
3024 if (!test_bit(__QDISC_STATE_DEACTIVATED, 3024 if (!test_bit(__QDISC_STATE_DEACTIVATED,
3025 &q->state)) { 3025 &q->state)) {
3026 __netif_reschedule(q); 3026 __netif_reschedule(q);
3027 } else { 3027 } else {
3028 smp_mb__before_clear_bit(); 3028 smp_mb__before_clear_bit();
3029 clear_bit(__QDISC_STATE_SCHED, 3029 clear_bit(__QDISC_STATE_SCHED,
3030 &q->state); 3030 &q->state);
3031 } 3031 }
3032 } 3032 }
3033 } 3033 }
3034 } 3034 }
3035 } 3035 }
3036 3036
3037 #if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \ 3037 #if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
3038 (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE)) 3038 (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3039 /* This hook is defined here for ATM LANE */ 3039 /* This hook is defined here for ATM LANE */
3040 int (*br_fdb_test_addr_hook)(struct net_device *dev, 3040 int (*br_fdb_test_addr_hook)(struct net_device *dev,
3041 unsigned char *addr) __read_mostly; 3041 unsigned char *addr) __read_mostly;
3042 EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook); 3042 EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3043 #endif 3043 #endif
3044 3044
3045 #ifdef CONFIG_NET_CLS_ACT 3045 #ifdef CONFIG_NET_CLS_ACT
3046 /* TODO: Maybe we should just force sch_ingress to be compiled in 3046 /* TODO: Maybe we should just force sch_ingress to be compiled in
3047 * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions 3047 * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions
3048 * a compare and 2 stores extra right now if we dont have it on 3048 * a compare and 2 stores extra right now if we dont have it on
3049 * but have CONFIG_NET_CLS_ACT 3049 * but have CONFIG_NET_CLS_ACT
3050 * NOTE: This doesn't stop any functionality; if you dont have 3050 * NOTE: This doesn't stop any functionality; if you dont have
3051 * the ingress scheduler, you just can't add policies on ingress. 3051 * the ingress scheduler, you just can't add policies on ingress.
3052 * 3052 *
3053 */ 3053 */
3054 static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq) 3054 static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
3055 { 3055 {
3056 struct net_device *dev = skb->dev; 3056 struct net_device *dev = skb->dev;
3057 u32 ttl = G_TC_RTTL(skb->tc_verd); 3057 u32 ttl = G_TC_RTTL(skb->tc_verd);
3058 int result = TC_ACT_OK; 3058 int result = TC_ACT_OK;
3059 struct Qdisc *q; 3059 struct Qdisc *q;
3060 3060
3061 if (unlikely(MAX_RED_LOOP < ttl++)) { 3061 if (unlikely(MAX_RED_LOOP < ttl++)) {
3062 if (net_ratelimit()) 3062 if (net_ratelimit())
3063 pr_warn("Redir loop detected Dropping packet (%d->%d)\n", 3063 pr_warn("Redir loop detected Dropping packet (%d->%d)\n",
3064 skb->skb_iif, dev->ifindex); 3064 skb->skb_iif, dev->ifindex);
3065 return TC_ACT_SHOT; 3065 return TC_ACT_SHOT;
3066 } 3066 }
3067 3067
3068 skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl); 3068 skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl);
3069 skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS); 3069 skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
3070 3070
3071 q = rxq->qdisc; 3071 q = rxq->qdisc;
3072 if (q != &noop_qdisc) { 3072 if (q != &noop_qdisc) {
3073 spin_lock(qdisc_lock(q)); 3073 spin_lock(qdisc_lock(q));
3074 if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) 3074 if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
3075 result = qdisc_enqueue_root(skb, q); 3075 result = qdisc_enqueue_root(skb, q);
3076 spin_unlock(qdisc_lock(q)); 3076 spin_unlock(qdisc_lock(q));
3077 } 3077 }
3078 3078
3079 return result; 3079 return result;
3080 } 3080 }
3081 3081
3082 static inline struct sk_buff *handle_ing(struct sk_buff *skb, 3082 static inline struct sk_buff *handle_ing(struct sk_buff *skb,
3083 struct packet_type **pt_prev, 3083 struct packet_type **pt_prev,
3084 int *ret, struct net_device *orig_dev) 3084 int *ret, struct net_device *orig_dev)
3085 { 3085 {
3086 struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue); 3086 struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);
3087 3087
3088 if (!rxq || rxq->qdisc == &noop_qdisc) 3088 if (!rxq || rxq->qdisc == &noop_qdisc)
3089 goto out; 3089 goto out;
3090 3090
3091 if (*pt_prev) { 3091 if (*pt_prev) {
3092 *ret = deliver_skb(skb, *pt_prev, orig_dev); 3092 *ret = deliver_skb(skb, *pt_prev, orig_dev);
3093 *pt_prev = NULL; 3093 *pt_prev = NULL;
3094 } 3094 }
3095 3095
3096 switch (ing_filter(skb, rxq)) { 3096 switch (ing_filter(skb, rxq)) {
3097 case TC_ACT_SHOT: 3097 case TC_ACT_SHOT:
3098 case TC_ACT_STOLEN: 3098 case TC_ACT_STOLEN:
3099 kfree_skb(skb); 3099 kfree_skb(skb);
3100 return NULL; 3100 return NULL;
3101 } 3101 }
3102 3102
3103 out: 3103 out:
3104 skb->tc_verd = 0; 3104 skb->tc_verd = 0;
3105 return skb; 3105 return skb;
3106 } 3106 }
3107 #endif 3107 #endif
3108 3108
3109 /** 3109 /**
3110 * netdev_rx_handler_register - register receive handler 3110 * netdev_rx_handler_register - register receive handler
3111 * @dev: device to register a handler for 3111 * @dev: device to register a handler for
3112 * @rx_handler: receive handler to register 3112 * @rx_handler: receive handler to register
3113 * @rx_handler_data: data pointer that is used by rx handler 3113 * @rx_handler_data: data pointer that is used by rx handler
3114 * 3114 *
3115 * Register a receive hander for a device. This handler will then be 3115 * Register a receive hander for a device. This handler will then be
3116 * called from __netif_receive_skb. A negative errno code is returned 3116 * called from __netif_receive_skb. A negative errno code is returned
3117 * on a failure. 3117 * on a failure.
3118 * 3118 *
3119 * The caller must hold the rtnl_mutex. 3119 * The caller must hold the rtnl_mutex.
3120 * 3120 *
3121 * For a general description of rx_handler, see enum rx_handler_result. 3121 * For a general description of rx_handler, see enum rx_handler_result.
3122 */ 3122 */
3123 int netdev_rx_handler_register(struct net_device *dev, 3123 int netdev_rx_handler_register(struct net_device *dev,
3124 rx_handler_func_t *rx_handler, 3124 rx_handler_func_t *rx_handler,
3125 void *rx_handler_data) 3125 void *rx_handler_data)
3126 { 3126 {
3127 ASSERT_RTNL(); 3127 ASSERT_RTNL();
3128 3128
3129 if (dev->rx_handler) 3129 if (dev->rx_handler)
3130 return -EBUSY; 3130 return -EBUSY;
3131 3131
3132 rcu_assign_pointer(dev->rx_handler_data, rx_handler_data); 3132 rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3133 rcu_assign_pointer(dev->rx_handler, rx_handler); 3133 rcu_assign_pointer(dev->rx_handler, rx_handler);
3134 3134
3135 return 0; 3135 return 0;
3136 } 3136 }
3137 EXPORT_SYMBOL_GPL(netdev_rx_handler_register); 3137 EXPORT_SYMBOL_GPL(netdev_rx_handler_register);
3138 3138
3139 /** 3139 /**
3140 * netdev_rx_handler_unregister - unregister receive handler 3140 * netdev_rx_handler_unregister - unregister receive handler
3141 * @dev: device to unregister a handler from 3141 * @dev: device to unregister a handler from
3142 * 3142 *
3143 * Unregister a receive hander from a device. 3143 * Unregister a receive hander from a device.
3144 * 3144 *
3145 * The caller must hold the rtnl_mutex. 3145 * The caller must hold the rtnl_mutex.
3146 */ 3146 */
3147 void netdev_rx_handler_unregister(struct net_device *dev) 3147 void netdev_rx_handler_unregister(struct net_device *dev)
3148 { 3148 {
3149 3149
3150 ASSERT_RTNL(); 3150 ASSERT_RTNL();
3151 RCU_INIT_POINTER(dev->rx_handler, NULL); 3151 RCU_INIT_POINTER(dev->rx_handler, NULL);
3152 RCU_INIT_POINTER(dev->rx_handler_data, NULL); 3152 RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3153 } 3153 }
3154 EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister); 3154 EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);
3155 3155
3156 static int __netif_receive_skb(struct sk_buff *skb) 3156 static int __netif_receive_skb(struct sk_buff *skb)
3157 { 3157 {
3158 struct packet_type *ptype, *pt_prev; 3158 struct packet_type *ptype, *pt_prev;
3159 rx_handler_func_t *rx_handler; 3159 rx_handler_func_t *rx_handler;
3160 struct net_device *orig_dev; 3160 struct net_device *orig_dev;
3161 struct net_device *null_or_dev; 3161 struct net_device *null_or_dev;
3162 bool deliver_exact = false; 3162 bool deliver_exact = false;
3163 int ret = NET_RX_DROP; 3163 int ret = NET_RX_DROP;
3164 __be16 type; 3164 __be16 type;
3165 3165
3166 net_timestamp_check(!netdev_tstamp_prequeue, skb); 3166 net_timestamp_check(!netdev_tstamp_prequeue, skb);
3167 3167
3168 trace_netif_receive_skb(skb); 3168 trace_netif_receive_skb(skb);
3169 3169
3170 /* if we've gotten here through NAPI, check netpoll */ 3170 /* if we've gotten here through NAPI, check netpoll */
3171 if (netpoll_receive_skb(skb)) 3171 if (netpoll_receive_skb(skb))
3172 return NET_RX_DROP; 3172 return NET_RX_DROP;
3173 3173
3174 if (!skb->skb_iif) 3174 if (!skb->skb_iif)
3175 skb->skb_iif = skb->dev->ifindex; 3175 skb->skb_iif = skb->dev->ifindex;
3176 orig_dev = skb->dev; 3176 orig_dev = skb->dev;
3177 3177
3178 skb_reset_network_header(skb); 3178 skb_reset_network_header(skb);
3179 skb_reset_transport_header(skb); 3179 skb_reset_transport_header(skb);
3180 skb_reset_mac_len(skb); 3180 skb_reset_mac_len(skb);
3181 3181
3182 pt_prev = NULL; 3182 pt_prev = NULL;
3183 3183
3184 rcu_read_lock(); 3184 rcu_read_lock();
3185 3185
3186 another_round: 3186 another_round:
3187 3187
3188 __this_cpu_inc(softnet_data.processed); 3188 __this_cpu_inc(softnet_data.processed);
3189 3189
3190 if (skb->protocol == cpu_to_be16(ETH_P_8021Q)) { 3190 if (skb->protocol == cpu_to_be16(ETH_P_8021Q)) {
3191 skb = vlan_untag(skb); 3191 skb = vlan_untag(skb);
3192 if (unlikely(!skb)) 3192 if (unlikely(!skb))
3193 goto out; 3193 goto out;
3194 } 3194 }
3195 3195
3196 #ifdef CONFIG_NET_CLS_ACT 3196 #ifdef CONFIG_NET_CLS_ACT
3197 if (skb->tc_verd & TC_NCLS) { 3197 if (skb->tc_verd & TC_NCLS) {
3198 skb->tc_verd = CLR_TC_NCLS(skb->tc_verd); 3198 skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
3199 goto ncls; 3199 goto ncls;
3200 } 3200 }
3201 #endif 3201 #endif
3202 3202
3203 list_for_each_entry_rcu(ptype, &ptype_all, list) { 3203 list_for_each_entry_rcu(ptype, &ptype_all, list) {
3204 if (!ptype->dev || ptype->dev == skb->dev) { 3204 if (!ptype->dev || ptype->dev == skb->dev) {
3205 if (pt_prev) 3205 if (pt_prev)
3206 ret = deliver_skb(skb, pt_prev, orig_dev); 3206 ret = deliver_skb(skb, pt_prev, orig_dev);
3207 pt_prev = ptype; 3207 pt_prev = ptype;
3208 } 3208 }
3209 } 3209 }
3210 3210
3211 #ifdef CONFIG_NET_CLS_ACT 3211 #ifdef CONFIG_NET_CLS_ACT
3212 skb = handle_ing(skb, &pt_prev, &ret, orig_dev); 3212 skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
3213 if (!skb) 3213 if (!skb)
3214 goto out; 3214 goto out;
3215 ncls: 3215 ncls:
3216 #endif 3216 #endif
3217 3217
3218 rx_handler = rcu_dereference(skb->dev->rx_handler); 3218 rx_handler = rcu_dereference(skb->dev->rx_handler);
3219 if (vlan_tx_tag_present(skb)) { 3219 if (vlan_tx_tag_present(skb)) {
3220 if (pt_prev) { 3220 if (pt_prev) {
3221 ret = deliver_skb(skb, pt_prev, orig_dev); 3221 ret = deliver_skb(skb, pt_prev, orig_dev);
3222 pt_prev = NULL; 3222 pt_prev = NULL;
3223 } 3223 }
3224 if (vlan_do_receive(&skb, !rx_handler)) 3224 if (vlan_do_receive(&skb, !rx_handler))
3225 goto another_round; 3225 goto another_round;
3226 else if (unlikely(!skb)) 3226 else if (unlikely(!skb))
3227 goto out; 3227 goto out;
3228 } 3228 }
3229 3229
3230 if (rx_handler) { 3230 if (rx_handler) {
3231 if (pt_prev) { 3231 if (pt_prev) {
3232 ret = deliver_skb(skb, pt_prev, orig_dev); 3232 ret = deliver_skb(skb, pt_prev, orig_dev);
3233 pt_prev = NULL; 3233 pt_prev = NULL;
3234 } 3234 }
3235 switch (rx_handler(&skb)) { 3235 switch (rx_handler(&skb)) {
3236 case RX_HANDLER_CONSUMED: 3236 case RX_HANDLER_CONSUMED:
3237 goto out; 3237 goto out;
3238 case RX_HANDLER_ANOTHER: 3238 case RX_HANDLER_ANOTHER:
3239 goto another_round; 3239 goto another_round;
3240 case RX_HANDLER_EXACT: 3240 case RX_HANDLER_EXACT:
3241 deliver_exact = true; 3241 deliver_exact = true;
3242 case RX_HANDLER_PASS: 3242 case RX_HANDLER_PASS:
3243 break; 3243 break;
3244 default: 3244 default:
3245 BUG(); 3245 BUG();
3246 } 3246 }
3247 } 3247 }
3248 3248
3249 /* deliver only exact match when indicated */ 3249 /* deliver only exact match when indicated */
3250 null_or_dev = deliver_exact ? skb->dev : NULL; 3250 null_or_dev = deliver_exact ? skb->dev : NULL;
3251 3251
3252 type = skb->protocol; 3252 type = skb->protocol;
3253 list_for_each_entry_rcu(ptype, 3253 list_for_each_entry_rcu(ptype,
3254 &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) { 3254 &ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3255 if (ptype->type == type && 3255 if (ptype->type == type &&
3256 (ptype->dev == null_or_dev || ptype->dev == skb->dev || 3256 (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
3257 ptype->dev == orig_dev)) { 3257 ptype->dev == orig_dev)) {
3258 if (pt_prev) 3258 if (pt_prev)
3259 ret = deliver_skb(skb, pt_prev, orig_dev); 3259 ret = deliver_skb(skb, pt_prev, orig_dev);
3260 pt_prev = ptype; 3260 pt_prev = ptype;
3261 } 3261 }
3262 } 3262 }
3263 3263
3264 if (pt_prev) { 3264 if (pt_prev) {
3265 ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev); 3265 ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
3266 } else { 3266 } else {
3267 atomic_long_inc(&skb->dev->rx_dropped); 3267 atomic_long_inc(&skb->dev->rx_dropped);
3268 kfree_skb(skb); 3268 kfree_skb(skb);
3269 /* Jamal, now you will not able to escape explaining 3269 /* Jamal, now you will not able to escape explaining
3270 * me how you were going to use this. :-) 3270 * me how you were going to use this. :-)
3271 */ 3271 */
3272 ret = NET_RX_DROP; 3272 ret = NET_RX_DROP;
3273 } 3273 }
3274 3274
3275 out: 3275 out:
3276 rcu_read_unlock(); 3276 rcu_read_unlock();
3277 return ret; 3277 return ret;
3278 } 3278 }
3279 3279
3280 /** 3280 /**
3281 * netif_receive_skb - process receive buffer from network 3281 * netif_receive_skb - process receive buffer from network
3282 * @skb: buffer to process 3282 * @skb: buffer to process
3283 * 3283 *
3284 * netif_receive_skb() is the main receive data processing function. 3284 * netif_receive_skb() is the main receive data processing function.
3285 * It always succeeds. The buffer may be dropped during processing 3285 * It always succeeds. The buffer may be dropped during processing
3286 * for congestion control or by the protocol layers. 3286 * for congestion control or by the protocol layers.
3287 * 3287 *
3288 * This function may only be called from softirq context and interrupts 3288 * This function may only be called from softirq context and interrupts
3289 * should be enabled. 3289 * should be enabled.
3290 * 3290 *
3291 * Return values (usually ignored): 3291 * Return values (usually ignored):
3292 * NET_RX_SUCCESS: no congestion 3292 * NET_RX_SUCCESS: no congestion
3293 * NET_RX_DROP: packet was dropped 3293 * NET_RX_DROP: packet was dropped
3294 */ 3294 */
3295 int netif_receive_skb(struct sk_buff *skb) 3295 int netif_receive_skb(struct sk_buff *skb)
3296 { 3296 {
3297 net_timestamp_check(netdev_tstamp_prequeue, skb); 3297 net_timestamp_check(netdev_tstamp_prequeue, skb);
3298 3298
3299 if (skb_defer_rx_timestamp(skb)) 3299 if (skb_defer_rx_timestamp(skb))
3300 return NET_RX_SUCCESS; 3300 return NET_RX_SUCCESS;
3301 3301
3302 #ifdef CONFIG_RPS 3302 #ifdef CONFIG_RPS
3303 if (static_key_false(&rps_needed)) { 3303 if (static_key_false(&rps_needed)) {
3304 struct rps_dev_flow voidflow, *rflow = &voidflow; 3304 struct rps_dev_flow voidflow, *rflow = &voidflow;
3305 int cpu, ret; 3305 int cpu, ret;
3306 3306
3307 rcu_read_lock(); 3307 rcu_read_lock();
3308 3308
3309 cpu = get_rps_cpu(skb->dev, skb, &rflow); 3309 cpu = get_rps_cpu(skb->dev, skb, &rflow);
3310 3310
3311 if (cpu >= 0) { 3311 if (cpu >= 0) {
3312 ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail); 3312 ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
3313 rcu_read_unlock(); 3313 rcu_read_unlock();
3314 return ret; 3314 return ret;
3315 } 3315 }
3316 rcu_read_unlock(); 3316 rcu_read_unlock();
3317 } 3317 }
3318 #endif 3318 #endif
3319 return __netif_receive_skb(skb); 3319 return __netif_receive_skb(skb);
3320 } 3320 }
3321 EXPORT_SYMBOL(netif_receive_skb); 3321 EXPORT_SYMBOL(netif_receive_skb);
3322 3322
3323 /* Network device is going away, flush any packets still pending 3323 /* Network device is going away, flush any packets still pending
3324 * Called with irqs disabled. 3324 * Called with irqs disabled.
3325 */ 3325 */
3326 static void flush_backlog(void *arg) 3326 static void flush_backlog(void *arg)
3327 { 3327 {
3328 struct net_device *dev = arg; 3328 struct net_device *dev = arg;
3329 struct softnet_data *sd = &__get_cpu_var(softnet_data); 3329 struct softnet_data *sd = &__get_cpu_var(softnet_data);
3330 struct sk_buff *skb, *tmp; 3330 struct sk_buff *skb, *tmp;
3331 3331
3332 rps_lock(sd); 3332 rps_lock(sd);
3333 skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) { 3333 skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3334 if (skb->dev == dev) { 3334 if (skb->dev == dev) {
3335 __skb_unlink(skb, &sd->input_pkt_queue); 3335 __skb_unlink(skb, &sd->input_pkt_queue);
3336 kfree_skb(skb); 3336 kfree_skb(skb);
3337 input_queue_head_incr(sd); 3337 input_queue_head_incr(sd);
3338 } 3338 }
3339 } 3339 }
3340 rps_unlock(sd); 3340 rps_unlock(sd);
3341 3341
3342 skb_queue_walk_safe(&sd->process_queue, skb, tmp) { 3342 skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
3343 if (skb->dev == dev) { 3343 if (skb->dev == dev) {
3344 __skb_unlink(skb, &sd->process_queue); 3344 __skb_unlink(skb, &sd->process_queue);
3345 kfree_skb(skb); 3345 kfree_skb(skb);
3346 input_queue_head_incr(sd); 3346 input_queue_head_incr(sd);
3347 } 3347 }
3348 } 3348 }
3349 } 3349 }
3350 3350
3351 static int napi_gro_complete(struct sk_buff *skb) 3351 static int napi_gro_complete(struct sk_buff *skb)
3352 { 3352 {
3353 struct packet_type *ptype; 3353 struct packet_type *ptype;
3354 __be16 type = skb->protocol; 3354 __be16 type = skb->protocol;
3355 struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; 3355 struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK];
3356 int err = -ENOENT; 3356 int err = -ENOENT;
3357 3357
3358 if (NAPI_GRO_CB(skb)->count == 1) { 3358 if (NAPI_GRO_CB(skb)->count == 1) {
3359 skb_shinfo(skb)->gso_size = 0; 3359 skb_shinfo(skb)->gso_size = 0;
3360 goto out; 3360 goto out;
3361 } 3361 }
3362 3362
3363 rcu_read_lock(); 3363 rcu_read_lock();
3364 list_for_each_entry_rcu(ptype, head, list) { 3364 list_for_each_entry_rcu(ptype, head, list) {
3365 if (ptype->type != type || ptype->dev || !ptype->gro_complete) 3365 if (ptype->type != type || ptype->dev || !ptype->gro_complete)
3366 continue; 3366 continue;
3367 3367
3368 err = ptype->gro_complete(skb); 3368 err = ptype->gro_complete(skb);
3369 break; 3369 break;
3370 } 3370 }
3371 rcu_read_unlock(); 3371 rcu_read_unlock();
3372 3372
3373 if (err) { 3373 if (err) {
3374 WARN_ON(&ptype->list == head); 3374 WARN_ON(&ptype->list == head);
3375 kfree_skb(skb); 3375 kfree_skb(skb);
3376 return NET_RX_SUCCESS; 3376 return NET_RX_SUCCESS;
3377 } 3377 }
3378 3378
3379 out: 3379 out:
3380 return netif_receive_skb(skb); 3380 return netif_receive_skb(skb);
3381 } 3381 }
3382 3382
3383 inline void napi_gro_flush(struct napi_struct *napi) 3383 inline void napi_gro_flush(struct napi_struct *napi)
3384 { 3384 {
3385 struct sk_buff *skb, *next; 3385 struct sk_buff *skb, *next;
3386 3386
3387 for (skb = napi->gro_list; skb; skb = next) { 3387 for (skb = napi->gro_list; skb; skb = next) {
3388 next = skb->next; 3388 next = skb->next;
3389 skb->next = NULL; 3389 skb->next = NULL;
3390 napi_gro_complete(skb); 3390 napi_gro_complete(skb);
3391 } 3391 }
3392 3392
3393 napi->gro_count = 0; 3393 napi->gro_count = 0;
3394 napi->gro_list = NULL; 3394 napi->gro_list = NULL;
3395 } 3395 }
3396 EXPORT_SYMBOL(napi_gro_flush); 3396 EXPORT_SYMBOL(napi_gro_flush);
3397 3397
3398 enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb) 3398 enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3399 { 3399 {
3400 struct sk_buff **pp = NULL; 3400 struct sk_buff **pp = NULL;
3401 struct packet_type *ptype; 3401 struct packet_type *ptype;
3402 __be16 type = skb->protocol; 3402 __be16 type = skb->protocol;
3403 struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK]; 3403 struct list_head *head = &ptype_base[ntohs(type) & PTYPE_HASH_MASK];
3404 int same_flow; 3404 int same_flow;
3405 int mac_len; 3405 int mac_len;
3406 enum gro_result ret; 3406 enum gro_result ret;
3407 3407
3408 if (!(skb->dev->features & NETIF_F_GRO) || netpoll_rx_on(skb)) 3408 if (!(skb->dev->features & NETIF_F_GRO) || netpoll_rx_on(skb))
3409 goto normal; 3409 goto normal;
3410 3410
3411 if (skb_is_gso(skb) || skb_has_frag_list(skb)) 3411 if (skb_is_gso(skb) || skb_has_frag_list(skb))
3412 goto normal; 3412 goto normal;
3413 3413
3414 rcu_read_lock(); 3414 rcu_read_lock();
3415 list_for_each_entry_rcu(ptype, head, list) { 3415 list_for_each_entry_rcu(ptype, head, list) {
3416 if (ptype->type != type || ptype->dev || !ptype->gro_receive) 3416 if (ptype->type != type || ptype->dev || !ptype->gro_receive)
3417 continue; 3417 continue;
3418 3418
3419 skb_set_network_header(skb, skb_gro_offset(skb)); 3419 skb_set_network_header(skb, skb_gro_offset(skb));
3420 mac_len = skb->network_header - skb->mac_header; 3420 mac_len = skb->network_header - skb->mac_header;
3421 skb->mac_len = mac_len; 3421 skb->mac_len = mac_len;
3422 NAPI_GRO_CB(skb)->same_flow = 0; 3422 NAPI_GRO_CB(skb)->same_flow = 0;
3423 NAPI_GRO_CB(skb)->flush = 0; 3423 NAPI_GRO_CB(skb)->flush = 0;
3424 NAPI_GRO_CB(skb)->free = 0; 3424 NAPI_GRO_CB(skb)->free = 0;
3425 3425
3426 pp = ptype->gro_receive(&napi->gro_list, skb); 3426 pp = ptype->gro_receive(&napi->gro_list, skb);
3427 break; 3427 break;
3428 } 3428 }
3429 rcu_read_unlock(); 3429 rcu_read_unlock();
3430 3430
3431 if (&ptype->list == head) 3431 if (&ptype->list == head)
3432 goto normal; 3432 goto normal;
3433 3433
3434 same_flow = NAPI_GRO_CB(skb)->same_flow; 3434 same_flow = NAPI_GRO_CB(skb)->same_flow;
3435 ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED; 3435 ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
3436 3436
3437 if (pp) { 3437 if (pp) {
3438 struct sk_buff *nskb = *pp; 3438 struct sk_buff *nskb = *pp;
3439 3439
3440 *pp = nskb->next; 3440 *pp = nskb->next;
3441 nskb->next = NULL; 3441 nskb->next = NULL;
3442 napi_gro_complete(nskb); 3442 napi_gro_complete(nskb);
3443 napi->gro_count--; 3443 napi->gro_count--;
3444 } 3444 }
3445 3445
3446 if (same_flow) 3446 if (same_flow)
3447 goto ok; 3447 goto ok;
3448 3448
3449 if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS) 3449 if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS)
3450 goto normal; 3450 goto normal;
3451 3451
3452 napi->gro_count++; 3452 napi->gro_count++;
3453 NAPI_GRO_CB(skb)->count = 1; 3453 NAPI_GRO_CB(skb)->count = 1;
3454 skb_shinfo(skb)->gso_size = skb_gro_len(skb); 3454 skb_shinfo(skb)->gso_size = skb_gro_len(skb);
3455 skb->next = napi->gro_list; 3455 skb->next = napi->gro_list;
3456 napi->gro_list = skb; 3456 napi->gro_list = skb;
3457 ret = GRO_HELD; 3457 ret = GRO_HELD;
3458 3458
3459 pull: 3459 pull:
3460 if (skb_headlen(skb) < skb_gro_offset(skb)) { 3460 if (skb_headlen(skb) < skb_gro_offset(skb)) {
3461 int grow = skb_gro_offset(skb) - skb_headlen(skb); 3461 int grow = skb_gro_offset(skb) - skb_headlen(skb);
3462 3462
3463 BUG_ON(skb->end - skb->tail < grow); 3463 BUG_ON(skb->end - skb->tail < grow);
3464 3464
3465 memcpy(skb_tail_pointer(skb), NAPI_GRO_CB(skb)->frag0, grow); 3465 memcpy(skb_tail_pointer(skb), NAPI_GRO_CB(skb)->frag0, grow);
3466 3466
3467 skb->tail += grow; 3467 skb->tail += grow;
3468 skb->data_len -= grow; 3468 skb->data_len -= grow;
3469 3469
3470 skb_shinfo(skb)->frags[0].page_offset += grow; 3470 skb_shinfo(skb)->frags[0].page_offset += grow;
3471 skb_frag_size_sub(&skb_shinfo(skb)->frags[0], grow); 3471 skb_frag_size_sub(&skb_shinfo(skb)->frags[0], grow);
3472 3472
3473 if (unlikely(!skb_frag_size(&skb_shinfo(skb)->frags[0]))) { 3473 if (unlikely(!skb_frag_size(&skb_shinfo(skb)->frags[0]))) {
3474 skb_frag_unref(skb, 0); 3474 skb_frag_unref(skb, 0);
3475 memmove(skb_shinfo(skb)->frags, 3475 memmove(skb_shinfo(skb)->frags,
3476 skb_shinfo(skb)->frags + 1, 3476 skb_shinfo(skb)->frags + 1,
3477 --skb_shinfo(skb)->nr_frags * sizeof(skb_frag_t)); 3477 --skb_shinfo(skb)->nr_frags * sizeof(skb_frag_t));
3478 } 3478 }
3479 } 3479 }
3480 3480
3481 ok: 3481 ok:
3482 return ret; 3482 return ret;
3483 3483
3484 normal: 3484 normal:
3485 ret = GRO_NORMAL; 3485 ret = GRO_NORMAL;
3486 goto pull; 3486 goto pull;
3487 } 3487 }
3488 EXPORT_SYMBOL(dev_gro_receive); 3488 EXPORT_SYMBOL(dev_gro_receive);
3489 3489
3490 static inline gro_result_t 3490 static inline gro_result_t
3491 __napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) 3491 __napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3492 { 3492 {
3493 struct sk_buff *p; 3493 struct sk_buff *p;
3494 unsigned int maclen = skb->dev->hard_header_len; 3494 unsigned int maclen = skb->dev->hard_header_len;
3495 3495
3496 for (p = napi->gro_list; p; p = p->next) { 3496 for (p = napi->gro_list; p; p = p->next) {
3497 unsigned long diffs; 3497 unsigned long diffs;
3498 3498
3499 diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev; 3499 diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
3500 diffs |= p->vlan_tci ^ skb->vlan_tci; 3500 diffs |= p->vlan_tci ^ skb->vlan_tci;
3501 if (maclen == ETH_HLEN) 3501 if (maclen == ETH_HLEN)
3502 diffs |= compare_ether_header(skb_mac_header(p), 3502 diffs |= compare_ether_header(skb_mac_header(p),
3503 skb_gro_mac_header(skb)); 3503 skb_gro_mac_header(skb));
3504 else if (!diffs) 3504 else if (!diffs)
3505 diffs = memcmp(skb_mac_header(p), 3505 diffs = memcmp(skb_mac_header(p),
3506 skb_gro_mac_header(skb), 3506 skb_gro_mac_header(skb),
3507 maclen); 3507 maclen);
3508 NAPI_GRO_CB(p)->same_flow = !diffs; 3508 NAPI_GRO_CB(p)->same_flow = !diffs;
3509 NAPI_GRO_CB(p)->flush = 0; 3509 NAPI_GRO_CB(p)->flush = 0;
3510 } 3510 }
3511 3511
3512 return dev_gro_receive(napi, skb); 3512 return dev_gro_receive(napi, skb);
3513 } 3513 }
3514 3514
3515 gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb) 3515 gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
3516 { 3516 {
3517 switch (ret) { 3517 switch (ret) {
3518 case GRO_NORMAL: 3518 case GRO_NORMAL:
3519 if (netif_receive_skb(skb)) 3519 if (netif_receive_skb(skb))
3520 ret = GRO_DROP; 3520 ret = GRO_DROP;
3521 break; 3521 break;
3522 3522
3523 case GRO_DROP: 3523 case GRO_DROP:
3524 case GRO_MERGED_FREE: 3524 case GRO_MERGED_FREE:
3525 kfree_skb(skb); 3525 kfree_skb(skb);
3526 break; 3526 break;
3527 3527
3528 case GRO_HELD: 3528 case GRO_HELD:
3529 case GRO_MERGED: 3529 case GRO_MERGED:
3530 break; 3530 break;
3531 } 3531 }
3532 3532
3533 return ret; 3533 return ret;
3534 } 3534 }
3535 EXPORT_SYMBOL(napi_skb_finish); 3535 EXPORT_SYMBOL(napi_skb_finish);
3536 3536
3537 void skb_gro_reset_offset(struct sk_buff *skb) 3537 void skb_gro_reset_offset(struct sk_buff *skb)
3538 { 3538 {
3539 NAPI_GRO_CB(skb)->data_offset = 0; 3539 NAPI_GRO_CB(skb)->data_offset = 0;
3540 NAPI_GRO_CB(skb)->frag0 = NULL; 3540 NAPI_GRO_CB(skb)->frag0 = NULL;
3541 NAPI_GRO_CB(skb)->frag0_len = 0; 3541 NAPI_GRO_CB(skb)->frag0_len = 0;
3542 3542
3543 if (skb->mac_header == skb->tail && 3543 if (skb->mac_header == skb->tail &&
3544 !PageHighMem(skb_frag_page(&skb_shinfo(skb)->frags[0]))) { 3544 !PageHighMem(skb_frag_page(&skb_shinfo(skb)->frags[0]))) {
3545 NAPI_GRO_CB(skb)->frag0 = 3545 NAPI_GRO_CB(skb)->frag0 =
3546 skb_frag_address(&skb_shinfo(skb)->frags[0]); 3546 skb_frag_address(&skb_shinfo(skb)->frags[0]);
3547 NAPI_GRO_CB(skb)->frag0_len = skb_frag_size(&skb_shinfo(skb)->frags[0]); 3547 NAPI_GRO_CB(skb)->frag0_len = skb_frag_size(&skb_shinfo(skb)->frags[0]);
3548 } 3548 }
3549 } 3549 }
3550 EXPORT_SYMBOL(skb_gro_reset_offset); 3550 EXPORT_SYMBOL(skb_gro_reset_offset);
3551 3551
3552 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb) 3552 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3553 { 3553 {
3554 skb_gro_reset_offset(skb); 3554 skb_gro_reset_offset(skb);
3555 3555
3556 return napi_skb_finish(__napi_gro_receive(napi, skb), skb); 3556 return napi_skb_finish(__napi_gro_receive(napi, skb), skb);
3557 } 3557 }
3558 EXPORT_SYMBOL(napi_gro_receive); 3558 EXPORT_SYMBOL(napi_gro_receive);
3559 3559
3560 static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb) 3560 static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
3561 { 3561 {
3562 __skb_pull(skb, skb_headlen(skb)); 3562 __skb_pull(skb, skb_headlen(skb));
3563 skb_reserve(skb, NET_IP_ALIGN - skb_headroom(skb)); 3563 /* restore the reserve we had after netdev_alloc_skb_ip_align() */
3564 skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
3564 skb->vlan_tci = 0; 3565 skb->vlan_tci = 0;
3565 skb->dev = napi->dev; 3566 skb->dev = napi->dev;
3566 skb->skb_iif = 0; 3567 skb->skb_iif = 0;
3567 3568
3568 napi->skb = skb; 3569 napi->skb = skb;
3569 } 3570 }
3570 3571
3571 struct sk_buff *napi_get_frags(struct napi_struct *napi) 3572 struct sk_buff *napi_get_frags(struct napi_struct *napi)
3572 { 3573 {
3573 struct sk_buff *skb = napi->skb; 3574 struct sk_buff *skb = napi->skb;
3574 3575
3575 if (!skb) { 3576 if (!skb) {
3576 skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD); 3577 skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD);
3577 if (skb) 3578 if (skb)
3578 napi->skb = skb; 3579 napi->skb = skb;
3579 } 3580 }
3580 return skb; 3581 return skb;
3581 } 3582 }
3582 EXPORT_SYMBOL(napi_get_frags); 3583 EXPORT_SYMBOL(napi_get_frags);
3583 3584
3584 gro_result_t napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb, 3585 gro_result_t napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb,
3585 gro_result_t ret) 3586 gro_result_t ret)
3586 { 3587 {
3587 switch (ret) { 3588 switch (ret) {
3588 case GRO_NORMAL: 3589 case GRO_NORMAL:
3589 case GRO_HELD: 3590 case GRO_HELD:
3590 skb->protocol = eth_type_trans(skb, skb->dev); 3591 skb->protocol = eth_type_trans(skb, skb->dev);
3591 3592
3592 if (ret == GRO_HELD) 3593 if (ret == GRO_HELD)
3593 skb_gro_pull(skb, -ETH_HLEN); 3594 skb_gro_pull(skb, -ETH_HLEN);
3594 else if (netif_receive_skb(skb)) 3595 else if (netif_receive_skb(skb))
3595 ret = GRO_DROP; 3596 ret = GRO_DROP;
3596 break; 3597 break;
3597 3598
3598 case GRO_DROP: 3599 case GRO_DROP:
3599 case GRO_MERGED_FREE: 3600 case GRO_MERGED_FREE:
3600 napi_reuse_skb(napi, skb); 3601 napi_reuse_skb(napi, skb);
3601 break; 3602 break;
3602 3603
3603 case GRO_MERGED: 3604 case GRO_MERGED:
3604 break; 3605 break;
3605 } 3606 }
3606 3607
3607 return ret; 3608 return ret;
3608 } 3609 }
3609 EXPORT_SYMBOL(napi_frags_finish); 3610 EXPORT_SYMBOL(napi_frags_finish);
3610 3611
3611 struct sk_buff *napi_frags_skb(struct napi_struct *napi) 3612 struct sk_buff *napi_frags_skb(struct napi_struct *napi)
3612 { 3613 {
3613 struct sk_buff *skb = napi->skb; 3614 struct sk_buff *skb = napi->skb;
3614 struct ethhdr *eth; 3615 struct ethhdr *eth;
3615 unsigned int hlen; 3616 unsigned int hlen;
3616 unsigned int off; 3617 unsigned int off;
3617 3618
3618 napi->skb = NULL; 3619 napi->skb = NULL;
3619 3620
3620 skb_reset_mac_header(skb); 3621 skb_reset_mac_header(skb);
3621 skb_gro_reset_offset(skb); 3622 skb_gro_reset_offset(skb);
3622 3623
3623 off = skb_gro_offset(skb); 3624 off = skb_gro_offset(skb);
3624 hlen = off + sizeof(*eth); 3625 hlen = off + sizeof(*eth);
3625 eth = skb_gro_header_fast(skb, off); 3626 eth = skb_gro_header_fast(skb, off);
3626 if (skb_gro_header_hard(skb, hlen)) { 3627 if (skb_gro_header_hard(skb, hlen)) {
3627 eth = skb_gro_header_slow(skb, hlen, off); 3628 eth = skb_gro_header_slow(skb, hlen, off);
3628 if (unlikely(!eth)) { 3629 if (unlikely(!eth)) {
3629 napi_reuse_skb(napi, skb); 3630 napi_reuse_skb(napi, skb);
3630 skb = NULL; 3631 skb = NULL;
3631 goto out; 3632 goto out;
3632 } 3633 }
3633 } 3634 }
3634 3635
3635 skb_gro_pull(skb, sizeof(*eth)); 3636 skb_gro_pull(skb, sizeof(*eth));
3636 3637
3637 /* 3638 /*
3638 * This works because the only protocols we care about don't require 3639 * This works because the only protocols we care about don't require
3639 * special handling. We'll fix it up properly at the end. 3640 * special handling. We'll fix it up properly at the end.
3640 */ 3641 */
3641 skb->protocol = eth->h_proto; 3642 skb->protocol = eth->h_proto;
3642 3643
3643 out: 3644 out:
3644 return skb; 3645 return skb;
3645 } 3646 }
3646 EXPORT_SYMBOL(napi_frags_skb); 3647 EXPORT_SYMBOL(napi_frags_skb);
3647 3648
3648 gro_result_t napi_gro_frags(struct napi_struct *napi) 3649 gro_result_t napi_gro_frags(struct napi_struct *napi)
3649 { 3650 {
3650 struct sk_buff *skb = napi_frags_skb(napi); 3651 struct sk_buff *skb = napi_frags_skb(napi);
3651 3652
3652 if (!skb) 3653 if (!skb)
3653 return GRO_DROP; 3654 return GRO_DROP;
3654 3655
3655 return napi_frags_finish(napi, skb, __napi_gro_receive(napi, skb)); 3656 return napi_frags_finish(napi, skb, __napi_gro_receive(napi, skb));
3656 } 3657 }
3657 EXPORT_SYMBOL(napi_gro_frags); 3658 EXPORT_SYMBOL(napi_gro_frags);
3658 3659
3659 /* 3660 /*
3660 * net_rps_action sends any pending IPI's for rps. 3661 * net_rps_action sends any pending IPI's for rps.
3661 * Note: called with local irq disabled, but exits with local irq enabled. 3662 * Note: called with local irq disabled, but exits with local irq enabled.
3662 */ 3663 */
3663 static void net_rps_action_and_irq_enable(struct softnet_data *sd) 3664 static void net_rps_action_and_irq_enable(struct softnet_data *sd)
3664 { 3665 {
3665 #ifdef CONFIG_RPS 3666 #ifdef CONFIG_RPS
3666 struct softnet_data *remsd = sd->rps_ipi_list; 3667 struct softnet_data *remsd = sd->rps_ipi_list;
3667 3668
3668 if (remsd) { 3669 if (remsd) {
3669 sd->rps_ipi_list = NULL; 3670 sd->rps_ipi_list = NULL;
3670 3671
3671 local_irq_enable(); 3672 local_irq_enable();
3672 3673
3673 /* Send pending IPI's to kick RPS processing on remote cpus. */ 3674 /* Send pending IPI's to kick RPS processing on remote cpus. */
3674 while (remsd) { 3675 while (remsd) {
3675 struct softnet_data *next = remsd->rps_ipi_next; 3676 struct softnet_data *next = remsd->rps_ipi_next;
3676 3677
3677 if (cpu_online(remsd->cpu)) 3678 if (cpu_online(remsd->cpu))
3678 __smp_call_function_single(remsd->cpu, 3679 __smp_call_function_single(remsd->cpu,
3679 &remsd->csd, 0); 3680 &remsd->csd, 0);
3680 remsd = next; 3681 remsd = next;
3681 } 3682 }
3682 } else 3683 } else
3683 #endif 3684 #endif
3684 local_irq_enable(); 3685 local_irq_enable();
3685 } 3686 }
3686 3687
3687 static int process_backlog(struct napi_struct *napi, int quota) 3688 static int process_backlog(struct napi_struct *napi, int quota)
3688 { 3689 {
3689 int work = 0; 3690 int work = 0;
3690 struct softnet_data *sd = container_of(napi, struct softnet_data, backlog); 3691 struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
3691 3692
3692 #ifdef CONFIG_RPS 3693 #ifdef CONFIG_RPS
3693 /* Check if we have pending ipi, its better to send them now, 3694 /* Check if we have pending ipi, its better to send them now,
3694 * not waiting net_rx_action() end. 3695 * not waiting net_rx_action() end.
3695 */ 3696 */
3696 if (sd->rps_ipi_list) { 3697 if (sd->rps_ipi_list) {
3697 local_irq_disable(); 3698 local_irq_disable();
3698 net_rps_action_and_irq_enable(sd); 3699 net_rps_action_and_irq_enable(sd);
3699 } 3700 }
3700 #endif 3701 #endif
3701 napi->weight = weight_p; 3702 napi->weight = weight_p;
3702 local_irq_disable(); 3703 local_irq_disable();
3703 while (work < quota) { 3704 while (work < quota) {
3704 struct sk_buff *skb; 3705 struct sk_buff *skb;
3705 unsigned int qlen; 3706 unsigned int qlen;
3706 3707
3707 while ((skb = __skb_dequeue(&sd->process_queue))) { 3708 while ((skb = __skb_dequeue(&sd->process_queue))) {
3708 local_irq_enable(); 3709 local_irq_enable();
3709 __netif_receive_skb(skb); 3710 __netif_receive_skb(skb);
3710 local_irq_disable(); 3711 local_irq_disable();
3711 input_queue_head_incr(sd); 3712 input_queue_head_incr(sd);
3712 if (++work >= quota) { 3713 if (++work >= quota) {
3713 local_irq_enable(); 3714 local_irq_enable();
3714 return work; 3715 return work;
3715 } 3716 }
3716 } 3717 }
3717 3718
3718 rps_lock(sd); 3719 rps_lock(sd);
3719 qlen = skb_queue_len(&sd->input_pkt_queue); 3720 qlen = skb_queue_len(&sd->input_pkt_queue);
3720 if (qlen) 3721 if (qlen)
3721 skb_queue_splice_tail_init(&sd->input_pkt_queue, 3722 skb_queue_splice_tail_init(&sd->input_pkt_queue,
3722 &sd->process_queue); 3723 &sd->process_queue);
3723 3724
3724 if (qlen < quota - work) { 3725 if (qlen < quota - work) {
3725 /* 3726 /*
3726 * Inline a custom version of __napi_complete(). 3727 * Inline a custom version of __napi_complete().
3727 * only current cpu owns and manipulates this napi, 3728 * only current cpu owns and manipulates this napi,
3728 * and NAPI_STATE_SCHED is the only possible flag set on backlog. 3729 * and NAPI_STATE_SCHED is the only possible flag set on backlog.
3729 * we can use a plain write instead of clear_bit(), 3730 * we can use a plain write instead of clear_bit(),
3730 * and we dont need an smp_mb() memory barrier. 3731 * and we dont need an smp_mb() memory barrier.
3731 */ 3732 */
3732 list_del(&napi->poll_list); 3733 list_del(&napi->poll_list);
3733 napi->state = 0; 3734 napi->state = 0;
3734 3735
3735 quota = work + qlen; 3736 quota = work + qlen;
3736 } 3737 }
3737 rps_unlock(sd); 3738 rps_unlock(sd);
3738 } 3739 }
3739 local_irq_enable(); 3740 local_irq_enable();
3740 3741
3741 return work; 3742 return work;
3742 } 3743 }
3743 3744
3744 /** 3745 /**
3745 * __napi_schedule - schedule for receive 3746 * __napi_schedule - schedule for receive
3746 * @n: entry to schedule 3747 * @n: entry to schedule
3747 * 3748 *
3748 * The entry's receive function will be scheduled to run 3749 * The entry's receive function will be scheduled to run
3749 */ 3750 */
3750 void __napi_schedule(struct napi_struct *n) 3751 void __napi_schedule(struct napi_struct *n)
3751 { 3752 {
3752 unsigned long flags; 3753 unsigned long flags;
3753 3754
3754 local_irq_save(flags); 3755 local_irq_save(flags);
3755 ____napi_schedule(&__get_cpu_var(softnet_data), n); 3756 ____napi_schedule(&__get_cpu_var(softnet_data), n);
3756 local_irq_restore(flags); 3757 local_irq_restore(flags);
3757 } 3758 }
3758 EXPORT_SYMBOL(__napi_schedule); 3759 EXPORT_SYMBOL(__napi_schedule);
3759 3760
3760 void __napi_complete(struct napi_struct *n) 3761 void __napi_complete(struct napi_struct *n)
3761 { 3762 {
3762 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 3763 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
3763 BUG_ON(n->gro_list); 3764 BUG_ON(n->gro_list);
3764 3765
3765 list_del(&n->poll_list); 3766 list_del(&n->poll_list);
3766 smp_mb__before_clear_bit(); 3767 smp_mb__before_clear_bit();
3767 clear_bit(NAPI_STATE_SCHED, &n->state); 3768 clear_bit(NAPI_STATE_SCHED, &n->state);
3768 } 3769 }
3769 EXPORT_SYMBOL(__napi_complete); 3770 EXPORT_SYMBOL(__napi_complete);
3770 3771
3771 void napi_complete(struct napi_struct *n) 3772 void napi_complete(struct napi_struct *n)
3772 { 3773 {
3773 unsigned long flags; 3774 unsigned long flags;
3774 3775
3775 /* 3776 /*
3776 * don't let napi dequeue from the cpu poll list 3777 * don't let napi dequeue from the cpu poll list
3777 * just in case its running on a different cpu 3778 * just in case its running on a different cpu
3778 */ 3779 */
3779 if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state))) 3780 if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state)))
3780 return; 3781 return;
3781 3782
3782 napi_gro_flush(n); 3783 napi_gro_flush(n);
3783 local_irq_save(flags); 3784 local_irq_save(flags);
3784 __napi_complete(n); 3785 __napi_complete(n);
3785 local_irq_restore(flags); 3786 local_irq_restore(flags);
3786 } 3787 }
3787 EXPORT_SYMBOL(napi_complete); 3788 EXPORT_SYMBOL(napi_complete);
3788 3789
3789 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 3790 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
3790 int (*poll)(struct napi_struct *, int), int weight) 3791 int (*poll)(struct napi_struct *, int), int weight)
3791 { 3792 {
3792 INIT_LIST_HEAD(&napi->poll_list); 3793 INIT_LIST_HEAD(&napi->poll_list);
3793 napi->gro_count = 0; 3794 napi->gro_count = 0;
3794 napi->gro_list = NULL; 3795 napi->gro_list = NULL;
3795 napi->skb = NULL; 3796 napi->skb = NULL;
3796 napi->poll = poll; 3797 napi->poll = poll;
3797 napi->weight = weight; 3798 napi->weight = weight;
3798 list_add(&napi->dev_list, &dev->napi_list); 3799 list_add(&napi->dev_list, &dev->napi_list);
3799 napi->dev = dev; 3800 napi->dev = dev;
3800 #ifdef CONFIG_NETPOLL 3801 #ifdef CONFIG_NETPOLL
3801 spin_lock_init(&napi->poll_lock); 3802 spin_lock_init(&napi->poll_lock);
3802 napi->poll_owner = -1; 3803 napi->poll_owner = -1;
3803 #endif 3804 #endif
3804 set_bit(NAPI_STATE_SCHED, &napi->state); 3805 set_bit(NAPI_STATE_SCHED, &napi->state);
3805 } 3806 }
3806 EXPORT_SYMBOL(netif_napi_add); 3807 EXPORT_SYMBOL(netif_napi_add);
3807 3808
3808 void netif_napi_del(struct napi_struct *napi) 3809 void netif_napi_del(struct napi_struct *napi)
3809 { 3810 {
3810 struct sk_buff *skb, *next; 3811 struct sk_buff *skb, *next;
3811 3812
3812 list_del_init(&napi->dev_list); 3813 list_del_init(&napi->dev_list);
3813 napi_free_frags(napi); 3814 napi_free_frags(napi);
3814 3815
3815 for (skb = napi->gro_list; skb; skb = next) { 3816 for (skb = napi->gro_list; skb; skb = next) {
3816 next = skb->next; 3817 next = skb->next;
3817 skb->next = NULL; 3818 skb->next = NULL;
3818 kfree_skb(skb); 3819 kfree_skb(skb);
3819 } 3820 }
3820 3821
3821 napi->gro_list = NULL; 3822 napi->gro_list = NULL;
3822 napi->gro_count = 0; 3823 napi->gro_count = 0;
3823 } 3824 }
3824 EXPORT_SYMBOL(netif_napi_del); 3825 EXPORT_SYMBOL(netif_napi_del);
3825 3826
3826 static void net_rx_action(struct softirq_action *h) 3827 static void net_rx_action(struct softirq_action *h)
3827 { 3828 {
3828 struct softnet_data *sd = &__get_cpu_var(softnet_data); 3829 struct softnet_data *sd = &__get_cpu_var(softnet_data);
3829 unsigned long time_limit = jiffies + 2; 3830 unsigned long time_limit = jiffies + 2;
3830 int budget = netdev_budget; 3831 int budget = netdev_budget;
3831 void *have; 3832 void *have;
3832 3833
3833 local_irq_disable(); 3834 local_irq_disable();
3834 3835
3835 while (!list_empty(&sd->poll_list)) { 3836 while (!list_empty(&sd->poll_list)) {
3836 struct napi_struct *n; 3837 struct napi_struct *n;
3837 int work, weight; 3838 int work, weight;
3838 3839
3839 /* If softirq window is exhuasted then punt. 3840 /* If softirq window is exhuasted then punt.
3840 * Allow this to run for 2 jiffies since which will allow 3841 * Allow this to run for 2 jiffies since which will allow
3841 * an average latency of 1.5/HZ. 3842 * an average latency of 1.5/HZ.
3842 */ 3843 */
3843 if (unlikely(budget <= 0 || time_after(jiffies, time_limit))) 3844 if (unlikely(budget <= 0 || time_after(jiffies, time_limit)))
3844 goto softnet_break; 3845 goto softnet_break;
3845 3846
3846 local_irq_enable(); 3847 local_irq_enable();
3847 3848
3848 /* Even though interrupts have been re-enabled, this 3849 /* Even though interrupts have been re-enabled, this
3849 * access is safe because interrupts can only add new 3850 * access is safe because interrupts can only add new
3850 * entries to the tail of this list, and only ->poll() 3851 * entries to the tail of this list, and only ->poll()
3851 * calls can remove this head entry from the list. 3852 * calls can remove this head entry from the list.
3852 */ 3853 */
3853 n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list); 3854 n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list);
3854 3855
3855 have = netpoll_poll_lock(n); 3856 have = netpoll_poll_lock(n);
3856 3857
3857 weight = n->weight; 3858 weight = n->weight;
3858 3859
3859 /* This NAPI_STATE_SCHED test is for avoiding a race 3860 /* This NAPI_STATE_SCHED test is for avoiding a race
3860 * with netpoll's poll_napi(). Only the entity which 3861 * with netpoll's poll_napi(). Only the entity which
3861 * obtains the lock and sees NAPI_STATE_SCHED set will 3862 * obtains the lock and sees NAPI_STATE_SCHED set will
3862 * actually make the ->poll() call. Therefore we avoid 3863 * actually make the ->poll() call. Therefore we avoid
3863 * accidentally calling ->poll() when NAPI is not scheduled. 3864 * accidentally calling ->poll() when NAPI is not scheduled.
3864 */ 3865 */
3865 work = 0; 3866 work = 0;
3866 if (test_bit(NAPI_STATE_SCHED, &n->state)) { 3867 if (test_bit(NAPI_STATE_SCHED, &n->state)) {
3867 work = n->poll(n, weight); 3868 work = n->poll(n, weight);
3868 trace_napi_poll(n); 3869 trace_napi_poll(n);
3869 } 3870 }
3870 3871
3871 WARN_ON_ONCE(work > weight); 3872 WARN_ON_ONCE(work > weight);
3872 3873
3873 budget -= work; 3874 budget -= work;
3874 3875
3875 local_irq_disable(); 3876 local_irq_disable();
3876 3877
3877 /* Drivers must not modify the NAPI state if they 3878 /* Drivers must not modify the NAPI state if they
3878 * consume the entire weight. In such cases this code 3879 * consume the entire weight. In such cases this code
3879 * still "owns" the NAPI instance and therefore can 3880 * still "owns" the NAPI instance and therefore can
3880 * move the instance around on the list at-will. 3881 * move the instance around on the list at-will.
3881 */ 3882 */
3882 if (unlikely(work == weight)) { 3883 if (unlikely(work == weight)) {
3883 if (unlikely(napi_disable_pending(n))) { 3884 if (unlikely(napi_disable_pending(n))) {
3884 local_irq_enable(); 3885 local_irq_enable();
3885 napi_complete(n); 3886 napi_complete(n);
3886 local_irq_disable(); 3887 local_irq_disable();
3887 } else 3888 } else
3888 list_move_tail(&n->poll_list, &sd->poll_list); 3889 list_move_tail(&n->poll_list, &sd->poll_list);
3889 } 3890 }
3890 3891
3891 netpoll_poll_unlock(have); 3892 netpoll_poll_unlock(have);
3892 } 3893 }
3893 out: 3894 out:
3894 net_rps_action_and_irq_enable(sd); 3895 net_rps_action_and_irq_enable(sd);
3895 3896
3896 #ifdef CONFIG_NET_DMA 3897 #ifdef CONFIG_NET_DMA
3897 /* 3898 /*
3898 * There may not be any more sk_buffs coming right now, so push 3899 * There may not be any more sk_buffs coming right now, so push
3899 * any pending DMA copies to hardware 3900 * any pending DMA copies to hardware
3900 */ 3901 */
3901 dma_issue_pending_all(); 3902 dma_issue_pending_all();
3902 #endif 3903 #endif
3903 3904
3904 return; 3905 return;
3905 3906
3906 softnet_break: 3907 softnet_break:
3907 sd->time_squeeze++; 3908 sd->time_squeeze++;
3908 __raise_softirq_irqoff(NET_RX_SOFTIRQ); 3909 __raise_softirq_irqoff(NET_RX_SOFTIRQ);
3909 goto out; 3910 goto out;
3910 } 3911 }
3911 3912
3912 static gifconf_func_t *gifconf_list[NPROTO]; 3913 static gifconf_func_t *gifconf_list[NPROTO];
3913 3914
3914 /** 3915 /**
3915 * register_gifconf - register a SIOCGIF handler 3916 * register_gifconf - register a SIOCGIF handler
3916 * @family: Address family 3917 * @family: Address family
3917 * @gifconf: Function handler 3918 * @gifconf: Function handler
3918 * 3919 *
3919 * Register protocol dependent address dumping routines. The handler 3920 * Register protocol dependent address dumping routines. The handler
3920 * that is passed must not be freed or reused until it has been replaced 3921 * that is passed must not be freed or reused until it has been replaced
3921 * by another handler. 3922 * by another handler.
3922 */ 3923 */
3923 int register_gifconf(unsigned int family, gifconf_func_t *gifconf) 3924 int register_gifconf(unsigned int family, gifconf_func_t *gifconf)
3924 { 3925 {
3925 if (family >= NPROTO) 3926 if (family >= NPROTO)
3926 return -EINVAL; 3927 return -EINVAL;
3927 gifconf_list[family] = gifconf; 3928 gifconf_list[family] = gifconf;
3928 return 0; 3929 return 0;
3929 } 3930 }
3930 EXPORT_SYMBOL(register_gifconf); 3931 EXPORT_SYMBOL(register_gifconf);
3931 3932
3932 3933
3933 /* 3934 /*
3934 * Map an interface index to its name (SIOCGIFNAME) 3935 * Map an interface index to its name (SIOCGIFNAME)
3935 */ 3936 */
3936 3937
3937 /* 3938 /*
3938 * We need this ioctl for efficient implementation of the 3939 * We need this ioctl for efficient implementation of the
3939 * if_indextoname() function required by the IPv6 API. Without 3940 * if_indextoname() function required by the IPv6 API. Without
3940 * it, we would have to search all the interfaces to find a 3941 * it, we would have to search all the interfaces to find a
3941 * match. --pb 3942 * match. --pb
3942 */ 3943 */
3943 3944
3944 static int dev_ifname(struct net *net, struct ifreq __user *arg) 3945 static int dev_ifname(struct net *net, struct ifreq __user *arg)
3945 { 3946 {
3946 struct net_device *dev; 3947 struct net_device *dev;
3947 struct ifreq ifr; 3948 struct ifreq ifr;
3948 3949
3949 /* 3950 /*
3950 * Fetch the caller's info block. 3951 * Fetch the caller's info block.
3951 */ 3952 */
3952 3953
3953 if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) 3954 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
3954 return -EFAULT; 3955 return -EFAULT;
3955 3956
3956 rcu_read_lock(); 3957 rcu_read_lock();
3957 dev = dev_get_by_index_rcu(net, ifr.ifr_ifindex); 3958 dev = dev_get_by_index_rcu(net, ifr.ifr_ifindex);
3958 if (!dev) { 3959 if (!dev) {
3959 rcu_read_unlock(); 3960 rcu_read_unlock();
3960 return -ENODEV; 3961 return -ENODEV;
3961 } 3962 }
3962 3963
3963 strcpy(ifr.ifr_name, dev->name); 3964 strcpy(ifr.ifr_name, dev->name);
3964 rcu_read_unlock(); 3965 rcu_read_unlock();
3965 3966
3966 if (copy_to_user(arg, &ifr, sizeof(struct ifreq))) 3967 if (copy_to_user(arg, &ifr, sizeof(struct ifreq)))
3967 return -EFAULT; 3968 return -EFAULT;
3968 return 0; 3969 return 0;
3969 } 3970 }
3970 3971
3971 /* 3972 /*
3972 * Perform a SIOCGIFCONF call. This structure will change 3973 * Perform a SIOCGIFCONF call. This structure will change
3973 * size eventually, and there is nothing I can do about it. 3974 * size eventually, and there is nothing I can do about it.
3974 * Thus we will need a 'compatibility mode'. 3975 * Thus we will need a 'compatibility mode'.
3975 */ 3976 */
3976 3977
3977 static int dev_ifconf(struct net *net, char __user *arg) 3978 static int dev_ifconf(struct net *net, char __user *arg)
3978 { 3979 {
3979 struct ifconf ifc; 3980 struct ifconf ifc;
3980 struct net_device *dev; 3981 struct net_device *dev;
3981 char __user *pos; 3982 char __user *pos;
3982 int len; 3983 int len;
3983 int total; 3984 int total;
3984 int i; 3985 int i;
3985 3986
3986 /* 3987 /*
3987 * Fetch the caller's info block. 3988 * Fetch the caller's info block.
3988 */ 3989 */
3989 3990
3990 if (copy_from_user(&ifc, arg, sizeof(struct ifconf))) 3991 if (copy_from_user(&ifc, arg, sizeof(struct ifconf)))
3991 return -EFAULT; 3992 return -EFAULT;
3992 3993
3993 pos = ifc.ifc_buf; 3994 pos = ifc.ifc_buf;
3994 len = ifc.ifc_len; 3995 len = ifc.ifc_len;
3995 3996
3996 /* 3997 /*
3997 * Loop over the interfaces, and write an info block for each. 3998 * Loop over the interfaces, and write an info block for each.
3998 */ 3999 */
3999 4000
4000 total = 0; 4001 total = 0;
4001 for_each_netdev(net, dev) { 4002 for_each_netdev(net, dev) {
4002 for (i = 0; i < NPROTO; i++) { 4003 for (i = 0; i < NPROTO; i++) {
4003 if (gifconf_list[i]) { 4004 if (gifconf_list[i]) {
4004 int done; 4005 int done;
4005 if (!pos) 4006 if (!pos)
4006 done = gifconf_list[i](dev, NULL, 0); 4007 done = gifconf_list[i](dev, NULL, 0);
4007 else 4008 else
4008 done = gifconf_list[i](dev, pos + total, 4009 done = gifconf_list[i](dev, pos + total,
4009 len - total); 4010 len - total);
4010 if (done < 0) 4011 if (done < 0)
4011 return -EFAULT; 4012 return -EFAULT;
4012 total += done; 4013 total += done;
4013 } 4014 }
4014 } 4015 }
4015 } 4016 }
4016 4017
4017 /* 4018 /*
4018 * All done. Write the updated control block back to the caller. 4019 * All done. Write the updated control block back to the caller.
4019 */ 4020 */
4020 ifc.ifc_len = total; 4021 ifc.ifc_len = total;
4021 4022
4022 /* 4023 /*
4023 * Both BSD and Solaris return 0 here, so we do too. 4024 * Both BSD and Solaris return 0 here, so we do too.
4024 */ 4025 */
4025 return copy_to_user(arg, &ifc, sizeof(struct ifconf)) ? -EFAULT : 0; 4026 return copy_to_user(arg, &ifc, sizeof(struct ifconf)) ? -EFAULT : 0;
4026 } 4027 }
4027 4028
4028 #ifdef CONFIG_PROC_FS 4029 #ifdef CONFIG_PROC_FS
4029 4030
4030 #define BUCKET_SPACE (32 - NETDEV_HASHBITS) 4031 #define BUCKET_SPACE (32 - NETDEV_HASHBITS)
4031 4032
4032 struct dev_iter_state { 4033 struct dev_iter_state {
4033 struct seq_net_private p; 4034 struct seq_net_private p;
4034 unsigned int pos; /* bucket << BUCKET_SPACE + offset */ 4035 unsigned int pos; /* bucket << BUCKET_SPACE + offset */
4035 }; 4036 };
4036 4037
4037 #define get_bucket(x) ((x) >> BUCKET_SPACE) 4038 #define get_bucket(x) ((x) >> BUCKET_SPACE)
4038 #define get_offset(x) ((x) & ((1 << BUCKET_SPACE) - 1)) 4039 #define get_offset(x) ((x) & ((1 << BUCKET_SPACE) - 1))
4039 #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) 4040 #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o))
4040 4041
4041 static inline struct net_device *dev_from_same_bucket(struct seq_file *seq) 4042 static inline struct net_device *dev_from_same_bucket(struct seq_file *seq)
4042 { 4043 {
4043 struct dev_iter_state *state = seq->private; 4044 struct dev_iter_state *state = seq->private;
4044 struct net *net = seq_file_net(seq); 4045 struct net *net = seq_file_net(seq);
4045 struct net_device *dev; 4046 struct net_device *dev;
4046 struct hlist_node *p; 4047 struct hlist_node *p;
4047 struct hlist_head *h; 4048 struct hlist_head *h;
4048 unsigned int count, bucket, offset; 4049 unsigned int count, bucket, offset;
4049 4050
4050 bucket = get_bucket(state->pos); 4051 bucket = get_bucket(state->pos);
4051 offset = get_offset(state->pos); 4052 offset = get_offset(state->pos);
4052 h = &net->dev_name_head[bucket]; 4053 h = &net->dev_name_head[bucket];
4053 count = 0; 4054 count = 0;
4054 hlist_for_each_entry_rcu(dev, p, h, name_hlist) { 4055 hlist_for_each_entry_rcu(dev, p, h, name_hlist) {
4055 if (count++ == offset) { 4056 if (count++ == offset) {
4056 state->pos = set_bucket_offset(bucket, count); 4057 state->pos = set_bucket_offset(bucket, count);
4057 return dev; 4058 return dev;
4058 } 4059 }
4059 } 4060 }
4060 4061
4061 return NULL; 4062 return NULL;
4062 } 4063 }
4063 4064
4064 static inline struct net_device *dev_from_new_bucket(struct seq_file *seq) 4065 static inline struct net_device *dev_from_new_bucket(struct seq_file *seq)
4065 { 4066 {
4066 struct dev_iter_state *state = seq->private; 4067 struct dev_iter_state *state = seq->private;
4067 struct net_device *dev; 4068 struct net_device *dev;
4068 unsigned int bucket; 4069 unsigned int bucket;
4069 4070
4070 bucket = get_bucket(state->pos); 4071 bucket = get_bucket(state->pos);
4071 do { 4072 do {
4072 dev = dev_from_same_bucket(seq); 4073 dev = dev_from_same_bucket(seq);
4073 if (dev) 4074 if (dev)
4074 return dev; 4075 return dev;
4075 4076
4076 bucket++; 4077 bucket++;
4077 state->pos = set_bucket_offset(bucket, 0); 4078 state->pos = set_bucket_offset(bucket, 0);
4078 } while (bucket < NETDEV_HASHENTRIES); 4079 } while (bucket < NETDEV_HASHENTRIES);
4079 4080
4080 return NULL; 4081 return NULL;
4081 } 4082 }
4082 4083
4083 /* 4084 /*
4084 * This is invoked by the /proc filesystem handler to display a device 4085 * This is invoked by the /proc filesystem handler to display a device
4085 * in detail. 4086 * in detail.
4086 */ 4087 */
4087 void *dev_seq_start(struct seq_file *seq, loff_t *pos) 4088 void *dev_seq_start(struct seq_file *seq, loff_t *pos)
4088 __acquires(RCU) 4089 __acquires(RCU)
4089 { 4090 {
4090 struct dev_iter_state *state = seq->private; 4091 struct dev_iter_state *state = seq->private;
4091 4092
4092 rcu_read_lock(); 4093 rcu_read_lock();
4093 if (!*pos) 4094 if (!*pos)
4094 return SEQ_START_TOKEN; 4095 return SEQ_START_TOKEN;
4095 4096
4096 /* check for end of the hash */ 4097 /* check for end of the hash */
4097 if (state->pos == 0 && *pos > 1) 4098 if (state->pos == 0 && *pos > 1)
4098 return NULL; 4099 return NULL;
4099 4100
4100 return dev_from_new_bucket(seq); 4101 return dev_from_new_bucket(seq);
4101 } 4102 }
4102 4103
4103 void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4104 void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4104 { 4105 {
4105 struct net_device *dev; 4106 struct net_device *dev;
4106 4107
4107 ++*pos; 4108 ++*pos;
4108 4109
4109 if (v == SEQ_START_TOKEN) 4110 if (v == SEQ_START_TOKEN)
4110 return dev_from_new_bucket(seq); 4111 return dev_from_new_bucket(seq);
4111 4112
4112 dev = dev_from_same_bucket(seq); 4113 dev = dev_from_same_bucket(seq);
4113 if (dev) 4114 if (dev)
4114 return dev; 4115 return dev;
4115 4116
4116 return dev_from_new_bucket(seq); 4117 return dev_from_new_bucket(seq);
4117 } 4118 }
4118 4119
4119 void dev_seq_stop(struct seq_file *seq, void *v) 4120 void dev_seq_stop(struct seq_file *seq, void *v)
4120 __releases(RCU) 4121 __releases(RCU)
4121 { 4122 {
4122 rcu_read_unlock(); 4123 rcu_read_unlock();
4123 } 4124 }
4124 4125
4125 static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev) 4126 static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev)
4126 { 4127 {
4127 struct rtnl_link_stats64 temp; 4128 struct rtnl_link_stats64 temp;
4128 const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp); 4129 const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp);
4129 4130
4130 seq_printf(seq, "%6s: %7llu %7llu %4llu %4llu %4llu %5llu %10llu %9llu " 4131 seq_printf(seq, "%6s: %7llu %7llu %4llu %4llu %4llu %5llu %10llu %9llu "
4131 "%8llu %7llu %4llu %4llu %4llu %5llu %7llu %10llu\n", 4132 "%8llu %7llu %4llu %4llu %4llu %5llu %7llu %10llu\n",
4132 dev->name, stats->rx_bytes, stats->rx_packets, 4133 dev->name, stats->rx_bytes, stats->rx_packets,
4133 stats->rx_errors, 4134 stats->rx_errors,
4134 stats->rx_dropped + stats->rx_missed_errors, 4135 stats->rx_dropped + stats->rx_missed_errors,
4135 stats->rx_fifo_errors, 4136 stats->rx_fifo_errors,
4136 stats->rx_length_errors + stats->rx_over_errors + 4137 stats->rx_length_errors + stats->rx_over_errors +
4137 stats->rx_crc_errors + stats->rx_frame_errors, 4138 stats->rx_crc_errors + stats->rx_frame_errors,
4138 stats->rx_compressed, stats->multicast, 4139 stats->rx_compressed, stats->multicast,
4139 stats->tx_bytes, stats->tx_packets, 4140 stats->tx_bytes, stats->tx_packets,
4140 stats->tx_errors, stats->tx_dropped, 4141 stats->tx_errors, stats->tx_dropped,
4141 stats->tx_fifo_errors, stats->collisions, 4142 stats->tx_fifo_errors, stats->collisions,
4142 stats->tx_carrier_errors + 4143 stats->tx_carrier_errors +
4143 stats->tx_aborted_errors + 4144 stats->tx_aborted_errors +
4144 stats->tx_window_errors + 4145 stats->tx_window_errors +
4145 stats->tx_heartbeat_errors, 4146 stats->tx_heartbeat_errors,
4146 stats->tx_compressed); 4147 stats->tx_compressed);
4147 } 4148 }
4148 4149
4149 /* 4150 /*
4150 * Called from the PROCfs module. This now uses the new arbitrary sized 4151 * Called from the PROCfs module. This now uses the new arbitrary sized
4151 * /proc/net interface to create /proc/net/dev 4152 * /proc/net interface to create /proc/net/dev
4152 */ 4153 */
4153 static int dev_seq_show(struct seq_file *seq, void *v) 4154 static int dev_seq_show(struct seq_file *seq, void *v)
4154 { 4155 {
4155 if (v == SEQ_START_TOKEN) 4156 if (v == SEQ_START_TOKEN)
4156 seq_puts(seq, "Inter-| Receive " 4157 seq_puts(seq, "Inter-| Receive "
4157 " | Transmit\n" 4158 " | Transmit\n"
4158 " face |bytes packets errs drop fifo frame " 4159 " face |bytes packets errs drop fifo frame "
4159 "compressed multicast|bytes packets errs " 4160 "compressed multicast|bytes packets errs "
4160 "drop fifo colls carrier compressed\n"); 4161 "drop fifo colls carrier compressed\n");
4161 else 4162 else
4162 dev_seq_printf_stats(seq, v); 4163 dev_seq_printf_stats(seq, v);
4163 return 0; 4164 return 0;
4164 } 4165 }
4165 4166
4166 static struct softnet_data *softnet_get_online(loff_t *pos) 4167 static struct softnet_data *softnet_get_online(loff_t *pos)
4167 { 4168 {
4168 struct softnet_data *sd = NULL; 4169 struct softnet_data *sd = NULL;
4169 4170
4170 while (*pos < nr_cpu_ids) 4171 while (*pos < nr_cpu_ids)
4171 if (cpu_online(*pos)) { 4172 if (cpu_online(*pos)) {
4172 sd = &per_cpu(softnet_data, *pos); 4173 sd = &per_cpu(softnet_data, *pos);
4173 break; 4174 break;
4174 } else 4175 } else
4175 ++*pos; 4176 ++*pos;
4176 return sd; 4177 return sd;
4177 } 4178 }
4178 4179
4179 static void *softnet_seq_start(struct seq_file *seq, loff_t *pos) 4180 static void *softnet_seq_start(struct seq_file *seq, loff_t *pos)
4180 { 4181 {
4181 return softnet_get_online(pos); 4182 return softnet_get_online(pos);
4182 } 4183 }
4183 4184
4184 static void *softnet_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4185 static void *softnet_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4185 { 4186 {
4186 ++*pos; 4187 ++*pos;
4187 return softnet_get_online(pos); 4188 return softnet_get_online(pos);
4188 } 4189 }
4189 4190
4190 static void softnet_seq_stop(struct seq_file *seq, void *v) 4191 static void softnet_seq_stop(struct seq_file *seq, void *v)
4191 { 4192 {
4192 } 4193 }
4193 4194
4194 static int softnet_seq_show(struct seq_file *seq, void *v) 4195 static int softnet_seq_show(struct seq_file *seq, void *v)
4195 { 4196 {
4196 struct softnet_data *sd = v; 4197 struct softnet_data *sd = v;
4197 4198
4198 seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x\n", 4199 seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x\n",
4199 sd->processed, sd->dropped, sd->time_squeeze, 0, 4200 sd->processed, sd->dropped, sd->time_squeeze, 0,
4200 0, 0, 0, 0, /* was fastroute */ 4201 0, 0, 0, 0, /* was fastroute */
4201 sd->cpu_collision, sd->received_rps); 4202 sd->cpu_collision, sd->received_rps);
4202 return 0; 4203 return 0;
4203 } 4204 }
4204 4205
4205 static const struct seq_operations dev_seq_ops = { 4206 static const struct seq_operations dev_seq_ops = {
4206 .start = dev_seq_start, 4207 .start = dev_seq_start,
4207 .next = dev_seq_next, 4208 .next = dev_seq_next,
4208 .stop = dev_seq_stop, 4209 .stop = dev_seq_stop,
4209 .show = dev_seq_show, 4210 .show = dev_seq_show,
4210 }; 4211 };
4211 4212
4212 static int dev_seq_open(struct inode *inode, struct file *file) 4213 static int dev_seq_open(struct inode *inode, struct file *file)
4213 { 4214 {
4214 return seq_open_net(inode, file, &dev_seq_ops, 4215 return seq_open_net(inode, file, &dev_seq_ops,
4215 sizeof(struct dev_iter_state)); 4216 sizeof(struct dev_iter_state));
4216 } 4217 }
4217 4218
4218 int dev_seq_open_ops(struct inode *inode, struct file *file, 4219 int dev_seq_open_ops(struct inode *inode, struct file *file,
4219 const struct seq_operations *ops) 4220 const struct seq_operations *ops)
4220 { 4221 {
4221 return seq_open_net(inode, file, ops, sizeof(struct dev_iter_state)); 4222 return seq_open_net(inode, file, ops, sizeof(struct dev_iter_state));
4222 } 4223 }
4223 4224
4224 static const struct file_operations dev_seq_fops = { 4225 static const struct file_operations dev_seq_fops = {
4225 .owner = THIS_MODULE, 4226 .owner = THIS_MODULE,
4226 .open = dev_seq_open, 4227 .open = dev_seq_open,
4227 .read = seq_read, 4228 .read = seq_read,
4228 .llseek = seq_lseek, 4229 .llseek = seq_lseek,
4229 .release = seq_release_net, 4230 .release = seq_release_net,
4230 }; 4231 };
4231 4232
4232 static const struct seq_operations softnet_seq_ops = { 4233 static const struct seq_operations softnet_seq_ops = {
4233 .start = softnet_seq_start, 4234 .start = softnet_seq_start,
4234 .next = softnet_seq_next, 4235 .next = softnet_seq_next,
4235 .stop = softnet_seq_stop, 4236 .stop = softnet_seq_stop,
4236 .show = softnet_seq_show, 4237 .show = softnet_seq_show,
4237 }; 4238 };
4238 4239
4239 static int softnet_seq_open(struct inode *inode, struct file *file) 4240 static int softnet_seq_open(struct inode *inode, struct file *file)
4240 { 4241 {
4241 return seq_open(file, &softnet_seq_ops); 4242 return seq_open(file, &softnet_seq_ops);
4242 } 4243 }
4243 4244
4244 static const struct file_operations softnet_seq_fops = { 4245 static const struct file_operations softnet_seq_fops = {
4245 .owner = THIS_MODULE, 4246 .owner = THIS_MODULE,
4246 .open = softnet_seq_open, 4247 .open = softnet_seq_open,
4247 .read = seq_read, 4248 .read = seq_read,
4248 .llseek = seq_lseek, 4249 .llseek = seq_lseek,
4249 .release = seq_release, 4250 .release = seq_release,
4250 }; 4251 };
4251 4252
4252 static void *ptype_get_idx(loff_t pos) 4253 static void *ptype_get_idx(loff_t pos)
4253 { 4254 {
4254 struct packet_type *pt = NULL; 4255 struct packet_type *pt = NULL;
4255 loff_t i = 0; 4256 loff_t i = 0;
4256 int t; 4257 int t;
4257 4258
4258 list_for_each_entry_rcu(pt, &ptype_all, list) { 4259 list_for_each_entry_rcu(pt, &ptype_all, list) {
4259 if (i == pos) 4260 if (i == pos)
4260 return pt; 4261 return pt;
4261 ++i; 4262 ++i;
4262 } 4263 }
4263 4264
4264 for (t = 0; t < PTYPE_HASH_SIZE; t++) { 4265 for (t = 0; t < PTYPE_HASH_SIZE; t++) {
4265 list_for_each_entry_rcu(pt, &ptype_base[t], list) { 4266 list_for_each_entry_rcu(pt, &ptype_base[t], list) {
4266 if (i == pos) 4267 if (i == pos)
4267 return pt; 4268 return pt;
4268 ++i; 4269 ++i;
4269 } 4270 }
4270 } 4271 }
4271 return NULL; 4272 return NULL;
4272 } 4273 }
4273 4274
4274 static void *ptype_seq_start(struct seq_file *seq, loff_t *pos) 4275 static void *ptype_seq_start(struct seq_file *seq, loff_t *pos)
4275 __acquires(RCU) 4276 __acquires(RCU)
4276 { 4277 {
4277 rcu_read_lock(); 4278 rcu_read_lock();
4278 return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN; 4279 return *pos ? ptype_get_idx(*pos - 1) : SEQ_START_TOKEN;
4279 } 4280 }
4280 4281
4281 static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4282 static void *ptype_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4282 { 4283 {
4283 struct packet_type *pt; 4284 struct packet_type *pt;
4284 struct list_head *nxt; 4285 struct list_head *nxt;
4285 int hash; 4286 int hash;
4286 4287
4287 ++*pos; 4288 ++*pos;
4288 if (v == SEQ_START_TOKEN) 4289 if (v == SEQ_START_TOKEN)
4289 return ptype_get_idx(0); 4290 return ptype_get_idx(0);
4290 4291
4291 pt = v; 4292 pt = v;
4292 nxt = pt->list.next; 4293 nxt = pt->list.next;
4293 if (pt->type == htons(ETH_P_ALL)) { 4294 if (pt->type == htons(ETH_P_ALL)) {
4294 if (nxt != &ptype_all) 4295 if (nxt != &ptype_all)
4295 goto found; 4296 goto found;
4296 hash = 0; 4297 hash = 0;
4297 nxt = ptype_base[0].next; 4298 nxt = ptype_base[0].next;
4298 } else 4299 } else
4299 hash = ntohs(pt->type) & PTYPE_HASH_MASK; 4300 hash = ntohs(pt->type) & PTYPE_HASH_MASK;
4300 4301
4301 while (nxt == &ptype_base[hash]) { 4302 while (nxt == &ptype_base[hash]) {
4302 if (++hash >= PTYPE_HASH_SIZE) 4303 if (++hash >= PTYPE_HASH_SIZE)
4303 return NULL; 4304 return NULL;
4304 nxt = ptype_base[hash].next; 4305 nxt = ptype_base[hash].next;
4305 } 4306 }
4306 found: 4307 found:
4307 return list_entry(nxt, struct packet_type, list); 4308 return list_entry(nxt, struct packet_type, list);
4308 } 4309 }
4309 4310
4310 static void ptype_seq_stop(struct seq_file *seq, void *v) 4311 static void ptype_seq_stop(struct seq_file *seq, void *v)
4311 __releases(RCU) 4312 __releases(RCU)
4312 { 4313 {
4313 rcu_read_unlock(); 4314 rcu_read_unlock();
4314 } 4315 }
4315 4316
4316 static int ptype_seq_show(struct seq_file *seq, void *v) 4317 static int ptype_seq_show(struct seq_file *seq, void *v)
4317 { 4318 {
4318 struct packet_type *pt = v; 4319 struct packet_type *pt = v;
4319 4320
4320 if (v == SEQ_START_TOKEN) 4321 if (v == SEQ_START_TOKEN)
4321 seq_puts(seq, "Type Device Function\n"); 4322 seq_puts(seq, "Type Device Function\n");
4322 else if (pt->dev == NULL || dev_net(pt->dev) == seq_file_net(seq)) { 4323 else if (pt->dev == NULL || dev_net(pt->dev) == seq_file_net(seq)) {
4323 if (pt->type == htons(ETH_P_ALL)) 4324 if (pt->type == htons(ETH_P_ALL))
4324 seq_puts(seq, "ALL "); 4325 seq_puts(seq, "ALL ");
4325 else 4326 else
4326 seq_printf(seq, "%04x", ntohs(pt->type)); 4327 seq_printf(seq, "%04x", ntohs(pt->type));
4327 4328
4328 seq_printf(seq, " %-8s %pF\n", 4329 seq_printf(seq, " %-8s %pF\n",
4329 pt->dev ? pt->dev->name : "", pt->func); 4330 pt->dev ? pt->dev->name : "", pt->func);
4330 } 4331 }
4331 4332
4332 return 0; 4333 return 0;
4333 } 4334 }
4334 4335
4335 static const struct seq_operations ptype_seq_ops = { 4336 static const struct seq_operations ptype_seq_ops = {
4336 .start = ptype_seq_start, 4337 .start = ptype_seq_start,
4337 .next = ptype_seq_next, 4338 .next = ptype_seq_next,
4338 .stop = ptype_seq_stop, 4339 .stop = ptype_seq_stop,
4339 .show = ptype_seq_show, 4340 .show = ptype_seq_show,
4340 }; 4341 };
4341 4342
4342 static int ptype_seq_open(struct inode *inode, struct file *file) 4343 static int ptype_seq_open(struct inode *inode, struct file *file)
4343 { 4344 {
4344 return seq_open_net(inode, file, &ptype_seq_ops, 4345 return seq_open_net(inode, file, &ptype_seq_ops,
4345 sizeof(struct seq_net_private)); 4346 sizeof(struct seq_net_private));
4346 } 4347 }
4347 4348
4348 static const struct file_operations ptype_seq_fops = { 4349 static const struct file_operations ptype_seq_fops = {
4349 .owner = THIS_MODULE, 4350 .owner = THIS_MODULE,
4350 .open = ptype_seq_open, 4351 .open = ptype_seq_open,
4351 .read = seq_read, 4352 .read = seq_read,
4352 .llseek = seq_lseek, 4353 .llseek = seq_lseek,
4353 .release = seq_release_net, 4354 .release = seq_release_net,
4354 }; 4355 };
4355 4356
4356 4357
4357 static int __net_init dev_proc_net_init(struct net *net) 4358 static int __net_init dev_proc_net_init(struct net *net)
4358 { 4359 {
4359 int rc = -ENOMEM; 4360 int rc = -ENOMEM;
4360 4361
4361 if (!proc_net_fops_create(net, "dev", S_IRUGO, &dev_seq_fops)) 4362 if (!proc_net_fops_create(net, "dev", S_IRUGO, &dev_seq_fops))
4362 goto out; 4363 goto out;
4363 if (!proc_net_fops_create(net, "softnet_stat", S_IRUGO, &softnet_seq_fops)) 4364 if (!proc_net_fops_create(net, "softnet_stat", S_IRUGO, &softnet_seq_fops))
4364 goto out_dev; 4365 goto out_dev;
4365 if (!proc_net_fops_create(net, "ptype", S_IRUGO, &ptype_seq_fops)) 4366 if (!proc_net_fops_create(net, "ptype", S_IRUGO, &ptype_seq_fops))
4366 goto out_softnet; 4367 goto out_softnet;
4367 4368
4368 if (wext_proc_init(net)) 4369 if (wext_proc_init(net))
4369 goto out_ptype; 4370 goto out_ptype;
4370 rc = 0; 4371 rc = 0;
4371 out: 4372 out:
4372 return rc; 4373 return rc;
4373 out_ptype: 4374 out_ptype:
4374 proc_net_remove(net, "ptype"); 4375 proc_net_remove(net, "ptype");
4375 out_softnet: 4376 out_softnet:
4376 proc_net_remove(net, "softnet_stat"); 4377 proc_net_remove(net, "softnet_stat");
4377 out_dev: 4378 out_dev:
4378 proc_net_remove(net, "dev"); 4379 proc_net_remove(net, "dev");
4379 goto out; 4380 goto out;
4380 } 4381 }
4381 4382
4382 static void __net_exit dev_proc_net_exit(struct net *net) 4383 static void __net_exit dev_proc_net_exit(struct net *net)
4383 { 4384 {
4384 wext_proc_exit(net); 4385 wext_proc_exit(net);
4385 4386
4386 proc_net_remove(net, "ptype"); 4387 proc_net_remove(net, "ptype");
4387 proc_net_remove(net, "softnet_stat"); 4388 proc_net_remove(net, "softnet_stat");
4388 proc_net_remove(net, "dev"); 4389 proc_net_remove(net, "dev");
4389 } 4390 }
4390 4391
4391 static struct pernet_operations __net_initdata dev_proc_ops = { 4392 static struct pernet_operations __net_initdata dev_proc_ops = {
4392 .init = dev_proc_net_init, 4393 .init = dev_proc_net_init,
4393 .exit = dev_proc_net_exit, 4394 .exit = dev_proc_net_exit,
4394 }; 4395 };
4395 4396
4396 static int __init dev_proc_init(void) 4397 static int __init dev_proc_init(void)
4397 { 4398 {
4398 return register_pernet_subsys(&dev_proc_ops); 4399 return register_pernet_subsys(&dev_proc_ops);
4399 } 4400 }
4400 #else 4401 #else
4401 #define dev_proc_init() 0 4402 #define dev_proc_init() 0
4402 #endif /* CONFIG_PROC_FS */ 4403 #endif /* CONFIG_PROC_FS */
4403 4404
4404 4405
4405 /** 4406 /**
4406 * netdev_set_master - set up master pointer 4407 * netdev_set_master - set up master pointer
4407 * @slave: slave device 4408 * @slave: slave device
4408 * @master: new master device 4409 * @master: new master device
4409 * 4410 *
4410 * Changes the master device of the slave. Pass %NULL to break the 4411 * Changes the master device of the slave. Pass %NULL to break the
4411 * bonding. The caller must hold the RTNL semaphore. On a failure 4412 * bonding. The caller must hold the RTNL semaphore. On a failure
4412 * a negative errno code is returned. On success the reference counts 4413 * a negative errno code is returned. On success the reference counts
4413 * are adjusted and the function returns zero. 4414 * are adjusted and the function returns zero.
4414 */ 4415 */
4415 int netdev_set_master(struct net_device *slave, struct net_device *master) 4416 int netdev_set_master(struct net_device *slave, struct net_device *master)
4416 { 4417 {
4417 struct net_device *old = slave->master; 4418 struct net_device *old = slave->master;
4418 4419
4419 ASSERT_RTNL(); 4420 ASSERT_RTNL();
4420 4421
4421 if (master) { 4422 if (master) {
4422 if (old) 4423 if (old)
4423 return -EBUSY; 4424 return -EBUSY;
4424 dev_hold(master); 4425 dev_hold(master);
4425 } 4426 }
4426 4427
4427 slave->master = master; 4428 slave->master = master;
4428 4429
4429 if (old) 4430 if (old)
4430 dev_put(old); 4431 dev_put(old);
4431 return 0; 4432 return 0;
4432 } 4433 }
4433 EXPORT_SYMBOL(netdev_set_master); 4434 EXPORT_SYMBOL(netdev_set_master);
4434 4435
4435 /** 4436 /**
4436 * netdev_set_bond_master - set up bonding master/slave pair 4437 * netdev_set_bond_master - set up bonding master/slave pair
4437 * @slave: slave device 4438 * @slave: slave device
4438 * @master: new master device 4439 * @master: new master device
4439 * 4440 *
4440 * Changes the master device of the slave. Pass %NULL to break the 4441 * Changes the master device of the slave. Pass %NULL to break the
4441 * bonding. The caller must hold the RTNL semaphore. On a failure 4442 * bonding. The caller must hold the RTNL semaphore. On a failure
4442 * a negative errno code is returned. On success %RTM_NEWLINK is sent 4443 * a negative errno code is returned. On success %RTM_NEWLINK is sent
4443 * to the routing socket and the function returns zero. 4444 * to the routing socket and the function returns zero.
4444 */ 4445 */
4445 int netdev_set_bond_master(struct net_device *slave, struct net_device *master) 4446 int netdev_set_bond_master(struct net_device *slave, struct net_device *master)
4446 { 4447 {
4447 int err; 4448 int err;
4448 4449
4449 ASSERT_RTNL(); 4450 ASSERT_RTNL();
4450 4451
4451 err = netdev_set_master(slave, master); 4452 err = netdev_set_master(slave, master);
4452 if (err) 4453 if (err)
4453 return err; 4454 return err;
4454 if (master) 4455 if (master)
4455 slave->flags |= IFF_SLAVE; 4456 slave->flags |= IFF_SLAVE;
4456 else 4457 else
4457 slave->flags &= ~IFF_SLAVE; 4458 slave->flags &= ~IFF_SLAVE;
4458 4459
4459 rtmsg_ifinfo(RTM_NEWLINK, slave, IFF_SLAVE); 4460 rtmsg_ifinfo(RTM_NEWLINK, slave, IFF_SLAVE);
4460 return 0; 4461 return 0;
4461 } 4462 }
4462 EXPORT_SYMBOL(netdev_set_bond_master); 4463 EXPORT_SYMBOL(netdev_set_bond_master);
4463 4464
4464 static void dev_change_rx_flags(struct net_device *dev, int flags) 4465 static void dev_change_rx_flags(struct net_device *dev, int flags)
4465 { 4466 {
4466 const struct net_device_ops *ops = dev->netdev_ops; 4467 const struct net_device_ops *ops = dev->netdev_ops;
4467 4468
4468 if ((dev->flags & IFF_UP) && ops->ndo_change_rx_flags) 4469 if ((dev->flags & IFF_UP) && ops->ndo_change_rx_flags)
4469 ops->ndo_change_rx_flags(dev, flags); 4470 ops->ndo_change_rx_flags(dev, flags);
4470 } 4471 }
4471 4472
4472 static int __dev_set_promiscuity(struct net_device *dev, int inc) 4473 static int __dev_set_promiscuity(struct net_device *dev, int inc)
4473 { 4474 {
4474 unsigned int old_flags = dev->flags; 4475 unsigned int old_flags = dev->flags;
4475 uid_t uid; 4476 uid_t uid;
4476 gid_t gid; 4477 gid_t gid;
4477 4478
4478 ASSERT_RTNL(); 4479 ASSERT_RTNL();
4479 4480
4480 dev->flags |= IFF_PROMISC; 4481 dev->flags |= IFF_PROMISC;
4481 dev->promiscuity += inc; 4482 dev->promiscuity += inc;
4482 if (dev->promiscuity == 0) { 4483 if (dev->promiscuity == 0) {
4483 /* 4484 /*
4484 * Avoid overflow. 4485 * Avoid overflow.
4485 * If inc causes overflow, untouch promisc and return error. 4486 * If inc causes overflow, untouch promisc and return error.
4486 */ 4487 */
4487 if (inc < 0) 4488 if (inc < 0)
4488 dev->flags &= ~IFF_PROMISC; 4489 dev->flags &= ~IFF_PROMISC;
4489 else { 4490 else {
4490 dev->promiscuity -= inc; 4491 dev->promiscuity -= inc;
4491 pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n", 4492 pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
4492 dev->name); 4493 dev->name);
4493 return -EOVERFLOW; 4494 return -EOVERFLOW;
4494 } 4495 }
4495 } 4496 }
4496 if (dev->flags != old_flags) { 4497 if (dev->flags != old_flags) {
4497 pr_info("device %s %s promiscuous mode\n", 4498 pr_info("device %s %s promiscuous mode\n",
4498 dev->name, 4499 dev->name,
4499 dev->flags & IFF_PROMISC ? "entered" : "left"); 4500 dev->flags & IFF_PROMISC ? "entered" : "left");
4500 if (audit_enabled) { 4501 if (audit_enabled) {
4501 current_uid_gid(&uid, &gid); 4502 current_uid_gid(&uid, &gid);
4502 audit_log(current->audit_context, GFP_ATOMIC, 4503 audit_log(current->audit_context, GFP_ATOMIC,
4503 AUDIT_ANOM_PROMISCUOUS, 4504 AUDIT_ANOM_PROMISCUOUS,
4504 "dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u", 4505 "dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u",
4505 dev->name, (dev->flags & IFF_PROMISC), 4506 dev->name, (dev->flags & IFF_PROMISC),
4506 (old_flags & IFF_PROMISC), 4507 (old_flags & IFF_PROMISC),
4507 audit_get_loginuid(current), 4508 audit_get_loginuid(current),
4508 uid, gid, 4509 uid, gid,
4509 audit_get_sessionid(current)); 4510 audit_get_sessionid(current));
4510 } 4511 }
4511 4512
4512 dev_change_rx_flags(dev, IFF_PROMISC); 4513 dev_change_rx_flags(dev, IFF_PROMISC);
4513 } 4514 }
4514 return 0; 4515 return 0;
4515 } 4516 }
4516 4517
4517 /** 4518 /**
4518 * dev_set_promiscuity - update promiscuity count on a device 4519 * dev_set_promiscuity - update promiscuity count on a device
4519 * @dev: device 4520 * @dev: device
4520 * @inc: modifier 4521 * @inc: modifier
4521 * 4522 *
4522 * Add or remove promiscuity from a device. While the count in the device 4523 * Add or remove promiscuity from a device. While the count in the device
4523 * remains above zero the interface remains promiscuous. Once it hits zero 4524 * remains above zero the interface remains promiscuous. Once it hits zero
4524 * the device reverts back to normal filtering operation. A negative inc 4525 * the device reverts back to normal filtering operation. A negative inc
4525 * value is used to drop promiscuity on the device. 4526 * value is used to drop promiscuity on the device.
4526 * Return 0 if successful or a negative errno code on error. 4527 * Return 0 if successful or a negative errno code on error.
4527 */ 4528 */
4528 int dev_set_promiscuity(struct net_device *dev, int inc) 4529 int dev_set_promiscuity(struct net_device *dev, int inc)
4529 { 4530 {
4530 unsigned int old_flags = dev->flags; 4531 unsigned int old_flags = dev->flags;
4531 int err; 4532 int err;
4532 4533
4533 err = __dev_set_promiscuity(dev, inc); 4534 err = __dev_set_promiscuity(dev, inc);
4534 if (err < 0) 4535 if (err < 0)
4535 return err; 4536 return err;
4536 if (dev->flags != old_flags) 4537 if (dev->flags != old_flags)
4537 dev_set_rx_mode(dev); 4538 dev_set_rx_mode(dev);
4538 return err; 4539 return err;
4539 } 4540 }
4540 EXPORT_SYMBOL(dev_set_promiscuity); 4541 EXPORT_SYMBOL(dev_set_promiscuity);
4541 4542
4542 /** 4543 /**
4543 * dev_set_allmulti - update allmulti count on a device 4544 * dev_set_allmulti - update allmulti count on a device
4544 * @dev: device 4545 * @dev: device
4545 * @inc: modifier 4546 * @inc: modifier
4546 * 4547 *
4547 * Add or remove reception of all multicast frames to a device. While the 4548 * Add or remove reception of all multicast frames to a device. While the
4548 * count in the device remains above zero the interface remains listening 4549 * count in the device remains above zero the interface remains listening
4549 * to all interfaces. Once it hits zero the device reverts back to normal 4550 * to all interfaces. Once it hits zero the device reverts back to normal
4550 * filtering operation. A negative @inc value is used to drop the counter 4551 * filtering operation. A negative @inc value is used to drop the counter
4551 * when releasing a resource needing all multicasts. 4552 * when releasing a resource needing all multicasts.
4552 * Return 0 if successful or a negative errno code on error. 4553 * Return 0 if successful or a negative errno code on error.
4553 */ 4554 */
4554 4555
4555 int dev_set_allmulti(struct net_device *dev, int inc) 4556 int dev_set_allmulti(struct net_device *dev, int inc)
4556 { 4557 {
4557 unsigned int old_flags = dev->flags; 4558 unsigned int old_flags = dev->flags;
4558 4559
4559 ASSERT_RTNL(); 4560 ASSERT_RTNL();
4560 4561
4561 dev->flags |= IFF_ALLMULTI; 4562 dev->flags |= IFF_ALLMULTI;
4562 dev->allmulti += inc; 4563 dev->allmulti += inc;
4563 if (dev->allmulti == 0) { 4564 if (dev->allmulti == 0) {
4564 /* 4565 /*
4565 * Avoid overflow. 4566 * Avoid overflow.
4566 * If inc causes overflow, untouch allmulti and return error. 4567 * If inc causes overflow, untouch allmulti and return error.
4567 */ 4568 */
4568 if (inc < 0) 4569 if (inc < 0)
4569 dev->flags &= ~IFF_ALLMULTI; 4570 dev->flags &= ~IFF_ALLMULTI;
4570 else { 4571 else {
4571 dev->allmulti -= inc; 4572 dev->allmulti -= inc;
4572 pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n", 4573 pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
4573 dev->name); 4574 dev->name);
4574 return -EOVERFLOW; 4575 return -EOVERFLOW;
4575 } 4576 }
4576 } 4577 }
4577 if (dev->flags ^ old_flags) { 4578 if (dev->flags ^ old_flags) {
4578 dev_change_rx_flags(dev, IFF_ALLMULTI); 4579 dev_change_rx_flags(dev, IFF_ALLMULTI);
4579 dev_set_rx_mode(dev); 4580 dev_set_rx_mode(dev);
4580 } 4581 }
4581 return 0; 4582 return 0;
4582 } 4583 }
4583 EXPORT_SYMBOL(dev_set_allmulti); 4584 EXPORT_SYMBOL(dev_set_allmulti);
4584 4585
4585 /* 4586 /*
4586 * Upload unicast and multicast address lists to device and 4587 * Upload unicast and multicast address lists to device and
4587 * configure RX filtering. When the device doesn't support unicast 4588 * configure RX filtering. When the device doesn't support unicast
4588 * filtering it is put in promiscuous mode while unicast addresses 4589 * filtering it is put in promiscuous mode while unicast addresses
4589 * are present. 4590 * are present.
4590 */ 4591 */
4591 void __dev_set_rx_mode(struct net_device *dev) 4592 void __dev_set_rx_mode(struct net_device *dev)
4592 { 4593 {
4593 const struct net_device_ops *ops = dev->netdev_ops; 4594 const struct net_device_ops *ops = dev->netdev_ops;
4594 4595
4595 /* dev_open will call this function so the list will stay sane. */ 4596 /* dev_open will call this function so the list will stay sane. */
4596 if (!(dev->flags&IFF_UP)) 4597 if (!(dev->flags&IFF_UP))
4597 return; 4598 return;
4598 4599
4599 if (!netif_device_present(dev)) 4600 if (!netif_device_present(dev))
4600 return; 4601 return;
4601 4602
4602 if (!(dev->priv_flags & IFF_UNICAST_FLT)) { 4603 if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
4603 /* Unicast addresses changes may only happen under the rtnl, 4604 /* Unicast addresses changes may only happen under the rtnl,
4604 * therefore calling __dev_set_promiscuity here is safe. 4605 * therefore calling __dev_set_promiscuity here is safe.
4605 */ 4606 */
4606 if (!netdev_uc_empty(dev) && !dev->uc_promisc) { 4607 if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
4607 __dev_set_promiscuity(dev, 1); 4608 __dev_set_promiscuity(dev, 1);
4608 dev->uc_promisc = true; 4609 dev->uc_promisc = true;
4609 } else if (netdev_uc_empty(dev) && dev->uc_promisc) { 4610 } else if (netdev_uc_empty(dev) && dev->uc_promisc) {
4610 __dev_set_promiscuity(dev, -1); 4611 __dev_set_promiscuity(dev, -1);
4611 dev->uc_promisc = false; 4612 dev->uc_promisc = false;
4612 } 4613 }
4613 } 4614 }
4614 4615
4615 if (ops->ndo_set_rx_mode) 4616 if (ops->ndo_set_rx_mode)
4616 ops->ndo_set_rx_mode(dev); 4617 ops->ndo_set_rx_mode(dev);
4617 } 4618 }
4618 4619
4619 void dev_set_rx_mode(struct net_device *dev) 4620 void dev_set_rx_mode(struct net_device *dev)
4620 { 4621 {
4621 netif_addr_lock_bh(dev); 4622 netif_addr_lock_bh(dev);
4622 __dev_set_rx_mode(dev); 4623 __dev_set_rx_mode(dev);
4623 netif_addr_unlock_bh(dev); 4624 netif_addr_unlock_bh(dev);
4624 } 4625 }
4625 4626
4626 /** 4627 /**
4627 * dev_get_flags - get flags reported to userspace 4628 * dev_get_flags - get flags reported to userspace
4628 * @dev: device 4629 * @dev: device
4629 * 4630 *
4630 * Get the combination of flag bits exported through APIs to userspace. 4631 * Get the combination of flag bits exported through APIs to userspace.
4631 */ 4632 */
4632 unsigned dev_get_flags(const struct net_device *dev) 4633 unsigned dev_get_flags(const struct net_device *dev)
4633 { 4634 {
4634 unsigned flags; 4635 unsigned flags;
4635 4636
4636 flags = (dev->flags & ~(IFF_PROMISC | 4637 flags = (dev->flags & ~(IFF_PROMISC |
4637 IFF_ALLMULTI | 4638 IFF_ALLMULTI |
4638 IFF_RUNNING | 4639 IFF_RUNNING |
4639 IFF_LOWER_UP | 4640 IFF_LOWER_UP |
4640 IFF_DORMANT)) | 4641 IFF_DORMANT)) |
4641 (dev->gflags & (IFF_PROMISC | 4642 (dev->gflags & (IFF_PROMISC |
4642 IFF_ALLMULTI)); 4643 IFF_ALLMULTI));
4643 4644
4644 if (netif_running(dev)) { 4645 if (netif_running(dev)) {
4645 if (netif_oper_up(dev)) 4646 if (netif_oper_up(dev))
4646 flags |= IFF_RUNNING; 4647 flags |= IFF_RUNNING;
4647 if (netif_carrier_ok(dev)) 4648 if (netif_carrier_ok(dev))
4648 flags |= IFF_LOWER_UP; 4649 flags |= IFF_LOWER_UP;
4649 if (netif_dormant(dev)) 4650 if (netif_dormant(dev))
4650 flags |= IFF_DORMANT; 4651 flags |= IFF_DORMANT;
4651 } 4652 }
4652 4653
4653 return flags; 4654 return flags;
4654 } 4655 }
4655 EXPORT_SYMBOL(dev_get_flags); 4656 EXPORT_SYMBOL(dev_get_flags);
4656 4657
4657 int __dev_change_flags(struct net_device *dev, unsigned int flags) 4658 int __dev_change_flags(struct net_device *dev, unsigned int flags)
4658 { 4659 {
4659 unsigned int old_flags = dev->flags; 4660 unsigned int old_flags = dev->flags;
4660 int ret; 4661 int ret;
4661 4662
4662 ASSERT_RTNL(); 4663 ASSERT_RTNL();
4663 4664
4664 /* 4665 /*
4665 * Set the flags on our device. 4666 * Set the flags on our device.
4666 */ 4667 */
4667 4668
4668 dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP | 4669 dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP |
4669 IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL | 4670 IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL |
4670 IFF_AUTOMEDIA)) | 4671 IFF_AUTOMEDIA)) |
4671 (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC | 4672 (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC |
4672 IFF_ALLMULTI)); 4673 IFF_ALLMULTI));
4673 4674
4674 /* 4675 /*
4675 * Load in the correct multicast list now the flags have changed. 4676 * Load in the correct multicast list now the flags have changed.
4676 */ 4677 */
4677 4678
4678 if ((old_flags ^ flags) & IFF_MULTICAST) 4679 if ((old_flags ^ flags) & IFF_MULTICAST)
4679 dev_change_rx_flags(dev, IFF_MULTICAST); 4680 dev_change_rx_flags(dev, IFF_MULTICAST);
4680 4681
4681 dev_set_rx_mode(dev); 4682 dev_set_rx_mode(dev);
4682 4683
4683 /* 4684 /*
4684 * Have we downed the interface. We handle IFF_UP ourselves 4685 * Have we downed the interface. We handle IFF_UP ourselves
4685 * according to user attempts to set it, rather than blindly 4686 * according to user attempts to set it, rather than blindly
4686 * setting it. 4687 * setting it.
4687 */ 4688 */
4688 4689
4689 ret = 0; 4690 ret = 0;
4690 if ((old_flags ^ flags) & IFF_UP) { /* Bit is different ? */ 4691 if ((old_flags ^ flags) & IFF_UP) { /* Bit is different ? */
4691 ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev); 4692 ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
4692 4693
4693 if (!ret) 4694 if (!ret)
4694 dev_set_rx_mode(dev); 4695 dev_set_rx_mode(dev);
4695 } 4696 }
4696 4697
4697 if ((flags ^ dev->gflags) & IFF_PROMISC) { 4698 if ((flags ^ dev->gflags) & IFF_PROMISC) {
4698 int inc = (flags & IFF_PROMISC) ? 1 : -1; 4699 int inc = (flags & IFF_PROMISC) ? 1 : -1;
4699 4700
4700 dev->gflags ^= IFF_PROMISC; 4701 dev->gflags ^= IFF_PROMISC;
4701 dev_set_promiscuity(dev, inc); 4702 dev_set_promiscuity(dev, inc);
4702 } 4703 }
4703 4704
4704 /* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI 4705 /* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI
4705 is important. Some (broken) drivers set IFF_PROMISC, when 4706 is important. Some (broken) drivers set IFF_PROMISC, when
4706 IFF_ALLMULTI is requested not asking us and not reporting. 4707 IFF_ALLMULTI is requested not asking us and not reporting.
4707 */ 4708 */
4708 if ((flags ^ dev->gflags) & IFF_ALLMULTI) { 4709 if ((flags ^ dev->gflags) & IFF_ALLMULTI) {
4709 int inc = (flags & IFF_ALLMULTI) ? 1 : -1; 4710 int inc = (flags & IFF_ALLMULTI) ? 1 : -1;
4710 4711
4711 dev->gflags ^= IFF_ALLMULTI; 4712 dev->gflags ^= IFF_ALLMULTI;
4712 dev_set_allmulti(dev, inc); 4713 dev_set_allmulti(dev, inc);
4713 } 4714 }
4714 4715
4715 return ret; 4716 return ret;
4716 } 4717 }
4717 4718
4718 void __dev_notify_flags(struct net_device *dev, unsigned int old_flags) 4719 void __dev_notify_flags(struct net_device *dev, unsigned int old_flags)
4719 { 4720 {
4720 unsigned int changes = dev->flags ^ old_flags; 4721 unsigned int changes = dev->flags ^ old_flags;
4721 4722
4722 if (changes & IFF_UP) { 4723 if (changes & IFF_UP) {
4723 if (dev->flags & IFF_UP) 4724 if (dev->flags & IFF_UP)
4724 call_netdevice_notifiers(NETDEV_UP, dev); 4725 call_netdevice_notifiers(NETDEV_UP, dev);
4725 else 4726 else
4726 call_netdevice_notifiers(NETDEV_DOWN, dev); 4727 call_netdevice_notifiers(NETDEV_DOWN, dev);
4727 } 4728 }
4728 4729
4729 if (dev->flags & IFF_UP && 4730 if (dev->flags & IFF_UP &&
4730 (changes & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE))) 4731 (changes & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE)))
4731 call_netdevice_notifiers(NETDEV_CHANGE, dev); 4732 call_netdevice_notifiers(NETDEV_CHANGE, dev);
4732 } 4733 }
4733 4734
4734 /** 4735 /**
4735 * dev_change_flags - change device settings 4736 * dev_change_flags - change device settings
4736 * @dev: device 4737 * @dev: device
4737 * @flags: device state flags 4738 * @flags: device state flags
4738 * 4739 *
4739 * Change settings on device based state flags. The flags are 4740 * Change settings on device based state flags. The flags are
4740 * in the userspace exported format. 4741 * in the userspace exported format.
4741 */ 4742 */
4742 int dev_change_flags(struct net_device *dev, unsigned int flags) 4743 int dev_change_flags(struct net_device *dev, unsigned int flags)
4743 { 4744 {
4744 int ret; 4745 int ret;
4745 unsigned int changes, old_flags = dev->flags; 4746 unsigned int changes, old_flags = dev->flags;
4746 4747
4747 ret = __dev_change_flags(dev, flags); 4748 ret = __dev_change_flags(dev, flags);
4748 if (ret < 0) 4749 if (ret < 0)
4749 return ret; 4750 return ret;
4750 4751
4751 changes = old_flags ^ dev->flags; 4752 changes = old_flags ^ dev->flags;
4752 if (changes) 4753 if (changes)
4753 rtmsg_ifinfo(RTM_NEWLINK, dev, changes); 4754 rtmsg_ifinfo(RTM_NEWLINK, dev, changes);
4754 4755
4755 __dev_notify_flags(dev, old_flags); 4756 __dev_notify_flags(dev, old_flags);
4756 return ret; 4757 return ret;
4757 } 4758 }
4758 EXPORT_SYMBOL(dev_change_flags); 4759 EXPORT_SYMBOL(dev_change_flags);
4759 4760
4760 /** 4761 /**
4761 * dev_set_mtu - Change maximum transfer unit 4762 * dev_set_mtu - Change maximum transfer unit
4762 * @dev: device 4763 * @dev: device
4763 * @new_mtu: new transfer unit 4764 * @new_mtu: new transfer unit
4764 * 4765 *
4765 * Change the maximum transfer size of the network device. 4766 * Change the maximum transfer size of the network device.
4766 */ 4767 */
4767 int dev_set_mtu(struct net_device *dev, int new_mtu) 4768 int dev_set_mtu(struct net_device *dev, int new_mtu)
4768 { 4769 {
4769 const struct net_device_ops *ops = dev->netdev_ops; 4770 const struct net_device_ops *ops = dev->netdev_ops;
4770 int err; 4771 int err;
4771 4772
4772 if (new_mtu == dev->mtu) 4773 if (new_mtu == dev->mtu)
4773 return 0; 4774 return 0;
4774 4775
4775 /* MTU must be positive. */ 4776 /* MTU must be positive. */
4776 if (new_mtu < 0) 4777 if (new_mtu < 0)
4777 return -EINVAL; 4778 return -EINVAL;
4778 4779
4779 if (!netif_device_present(dev)) 4780 if (!netif_device_present(dev))
4780 return -ENODEV; 4781 return -ENODEV;
4781 4782
4782 err = 0; 4783 err = 0;
4783 if (ops->ndo_change_mtu) 4784 if (ops->ndo_change_mtu)
4784 err = ops->ndo_change_mtu(dev, new_mtu); 4785 err = ops->ndo_change_mtu(dev, new_mtu);
4785 else 4786 else
4786 dev->mtu = new_mtu; 4787 dev->mtu = new_mtu;
4787 4788
4788 if (!err && dev->flags & IFF_UP) 4789 if (!err && dev->flags & IFF_UP)
4789 call_netdevice_notifiers(NETDEV_CHANGEMTU, dev); 4790 call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
4790 return err; 4791 return err;
4791 } 4792 }
4792 EXPORT_SYMBOL(dev_set_mtu); 4793 EXPORT_SYMBOL(dev_set_mtu);
4793 4794
4794 /** 4795 /**
4795 * dev_set_group - Change group this device belongs to 4796 * dev_set_group - Change group this device belongs to
4796 * @dev: device 4797 * @dev: device
4797 * @new_group: group this device should belong to 4798 * @new_group: group this device should belong to
4798 */ 4799 */
4799 void dev_set_group(struct net_device *dev, int new_group) 4800 void dev_set_group(struct net_device *dev, int new_group)
4800 { 4801 {
4801 dev->group = new_group; 4802 dev->group = new_group;
4802 } 4803 }
4803 EXPORT_SYMBOL(dev_set_group); 4804 EXPORT_SYMBOL(dev_set_group);
4804 4805
4805 /** 4806 /**
4806 * dev_set_mac_address - Change Media Access Control Address 4807 * dev_set_mac_address - Change Media Access Control Address
4807 * @dev: device 4808 * @dev: device
4808 * @sa: new address 4809 * @sa: new address
4809 * 4810 *
4810 * Change the hardware (MAC) address of the device 4811 * Change the hardware (MAC) address of the device
4811 */ 4812 */
4812 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa) 4813 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
4813 { 4814 {
4814 const struct net_device_ops *ops = dev->netdev_ops; 4815 const struct net_device_ops *ops = dev->netdev_ops;
4815 int err; 4816 int err;
4816 4817
4817 if (!ops->ndo_set_mac_address) 4818 if (!ops->ndo_set_mac_address)
4818 return -EOPNOTSUPP; 4819 return -EOPNOTSUPP;
4819 if (sa->sa_family != dev->type) 4820 if (sa->sa_family != dev->type)
4820 return -EINVAL; 4821 return -EINVAL;
4821 if (!netif_device_present(dev)) 4822 if (!netif_device_present(dev))
4822 return -ENODEV; 4823 return -ENODEV;
4823 err = ops->ndo_set_mac_address(dev, sa); 4824 err = ops->ndo_set_mac_address(dev, sa);
4824 if (!err) 4825 if (!err)
4825 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 4826 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
4826 return err; 4827 return err;
4827 } 4828 }
4828 EXPORT_SYMBOL(dev_set_mac_address); 4829 EXPORT_SYMBOL(dev_set_mac_address);
4829 4830
4830 /* 4831 /*
4831 * Perform the SIOCxIFxxx calls, inside rcu_read_lock() 4832 * Perform the SIOCxIFxxx calls, inside rcu_read_lock()
4832 */ 4833 */
4833 static int dev_ifsioc_locked(struct net *net, struct ifreq *ifr, unsigned int cmd) 4834 static int dev_ifsioc_locked(struct net *net, struct ifreq *ifr, unsigned int cmd)
4834 { 4835 {
4835 int err; 4836 int err;
4836 struct net_device *dev = dev_get_by_name_rcu(net, ifr->ifr_name); 4837 struct net_device *dev = dev_get_by_name_rcu(net, ifr->ifr_name);
4837 4838
4838 if (!dev) 4839 if (!dev)
4839 return -ENODEV; 4840 return -ENODEV;
4840 4841
4841 switch (cmd) { 4842 switch (cmd) {
4842 case SIOCGIFFLAGS: /* Get interface flags */ 4843 case SIOCGIFFLAGS: /* Get interface flags */
4843 ifr->ifr_flags = (short) dev_get_flags(dev); 4844 ifr->ifr_flags = (short) dev_get_flags(dev);
4844 return 0; 4845 return 0;
4845 4846
4846 case SIOCGIFMETRIC: /* Get the metric on the interface 4847 case SIOCGIFMETRIC: /* Get the metric on the interface
4847 (currently unused) */ 4848 (currently unused) */
4848 ifr->ifr_metric = 0; 4849 ifr->ifr_metric = 0;
4849 return 0; 4850 return 0;
4850 4851
4851 case SIOCGIFMTU: /* Get the MTU of a device */ 4852 case SIOCGIFMTU: /* Get the MTU of a device */
4852 ifr->ifr_mtu = dev->mtu; 4853 ifr->ifr_mtu = dev->mtu;
4853 return 0; 4854 return 0;
4854 4855
4855 case SIOCGIFHWADDR: 4856 case SIOCGIFHWADDR:
4856 if (!dev->addr_len) 4857 if (!dev->addr_len)
4857 memset(ifr->ifr_hwaddr.sa_data, 0, sizeof ifr->ifr_hwaddr.sa_data); 4858 memset(ifr->ifr_hwaddr.sa_data, 0, sizeof ifr->ifr_hwaddr.sa_data);
4858 else 4859 else
4859 memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr, 4860 memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr,
4860 min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); 4861 min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len));
4861 ifr->ifr_hwaddr.sa_family = dev->type; 4862 ifr->ifr_hwaddr.sa_family = dev->type;
4862 return 0; 4863 return 0;
4863 4864
4864 case SIOCGIFSLAVE: 4865 case SIOCGIFSLAVE:
4865 err = -EINVAL; 4866 err = -EINVAL;
4866 break; 4867 break;
4867 4868
4868 case SIOCGIFMAP: 4869 case SIOCGIFMAP:
4869 ifr->ifr_map.mem_start = dev->mem_start; 4870 ifr->ifr_map.mem_start = dev->mem_start;
4870 ifr->ifr_map.mem_end = dev->mem_end; 4871 ifr->ifr_map.mem_end = dev->mem_end;
4871 ifr->ifr_map.base_addr = dev->base_addr; 4872 ifr->ifr_map.base_addr = dev->base_addr;
4872 ifr->ifr_map.irq = dev->irq; 4873 ifr->ifr_map.irq = dev->irq;
4873 ifr->ifr_map.dma = dev->dma; 4874 ifr->ifr_map.dma = dev->dma;
4874 ifr->ifr_map.port = dev->if_port; 4875 ifr->ifr_map.port = dev->if_port;
4875 return 0; 4876 return 0;
4876 4877
4877 case SIOCGIFINDEX: 4878 case SIOCGIFINDEX:
4878 ifr->ifr_ifindex = dev->ifindex; 4879 ifr->ifr_ifindex = dev->ifindex;
4879 return 0; 4880 return 0;
4880 4881
4881 case SIOCGIFTXQLEN: 4882 case SIOCGIFTXQLEN:
4882 ifr->ifr_qlen = dev->tx_queue_len; 4883 ifr->ifr_qlen = dev->tx_queue_len;
4883 return 0; 4884 return 0;
4884 4885
4885 default: 4886 default:
4886 /* dev_ioctl() should ensure this case 4887 /* dev_ioctl() should ensure this case
4887 * is never reached 4888 * is never reached
4888 */ 4889 */
4889 WARN_ON(1); 4890 WARN_ON(1);
4890 err = -ENOTTY; 4891 err = -ENOTTY;
4891 break; 4892 break;
4892 4893
4893 } 4894 }
4894 return err; 4895 return err;
4895 } 4896 }
4896 4897
4897 /* 4898 /*
4898 * Perform the SIOCxIFxxx calls, inside rtnl_lock() 4899 * Perform the SIOCxIFxxx calls, inside rtnl_lock()
4899 */ 4900 */
4900 static int dev_ifsioc(struct net *net, struct ifreq *ifr, unsigned int cmd) 4901 static int dev_ifsioc(struct net *net, struct ifreq *ifr, unsigned int cmd)
4901 { 4902 {
4902 int err; 4903 int err;
4903 struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name); 4904 struct net_device *dev = __dev_get_by_name(net, ifr->ifr_name);
4904 const struct net_device_ops *ops; 4905 const struct net_device_ops *ops;
4905 4906
4906 if (!dev) 4907 if (!dev)
4907 return -ENODEV; 4908 return -ENODEV;
4908 4909
4909 ops = dev->netdev_ops; 4910 ops = dev->netdev_ops;
4910 4911
4911 switch (cmd) { 4912 switch (cmd) {
4912 case SIOCSIFFLAGS: /* Set interface flags */ 4913 case SIOCSIFFLAGS: /* Set interface flags */
4913 return dev_change_flags(dev, ifr->ifr_flags); 4914 return dev_change_flags(dev, ifr->ifr_flags);
4914 4915
4915 case SIOCSIFMETRIC: /* Set the metric on the interface 4916 case SIOCSIFMETRIC: /* Set the metric on the interface
4916 (currently unused) */ 4917 (currently unused) */
4917 return -EOPNOTSUPP; 4918 return -EOPNOTSUPP;
4918 4919
4919 case SIOCSIFMTU: /* Set the MTU of a device */ 4920 case SIOCSIFMTU: /* Set the MTU of a device */
4920 return dev_set_mtu(dev, ifr->ifr_mtu); 4921 return dev_set_mtu(dev, ifr->ifr_mtu);
4921 4922
4922 case SIOCSIFHWADDR: 4923 case SIOCSIFHWADDR:
4923 return dev_set_mac_address(dev, &ifr->ifr_hwaddr); 4924 return dev_set_mac_address(dev, &ifr->ifr_hwaddr);
4924 4925
4925 case SIOCSIFHWBROADCAST: 4926 case SIOCSIFHWBROADCAST:
4926 if (ifr->ifr_hwaddr.sa_family != dev->type) 4927 if (ifr->ifr_hwaddr.sa_family != dev->type)
4927 return -EINVAL; 4928 return -EINVAL;
4928 memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data, 4929 memcpy(dev->broadcast, ifr->ifr_hwaddr.sa_data,
4929 min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len)); 4930 min(sizeof ifr->ifr_hwaddr.sa_data, (size_t) dev->addr_len));
4930 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 4931 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
4931 return 0; 4932 return 0;
4932 4933
4933 case SIOCSIFMAP: 4934 case SIOCSIFMAP:
4934 if (ops->ndo_set_config) { 4935 if (ops->ndo_set_config) {
4935 if (!netif_device_present(dev)) 4936 if (!netif_device_present(dev))
4936 return -ENODEV; 4937 return -ENODEV;
4937 return ops->ndo_set_config(dev, &ifr->ifr_map); 4938 return ops->ndo_set_config(dev, &ifr->ifr_map);
4938 } 4939 }
4939 return -EOPNOTSUPP; 4940 return -EOPNOTSUPP;
4940 4941
4941 case SIOCADDMULTI: 4942 case SIOCADDMULTI:
4942 if (!ops->ndo_set_rx_mode || 4943 if (!ops->ndo_set_rx_mode ||
4943 ifr->ifr_hwaddr.sa_family != AF_UNSPEC) 4944 ifr->ifr_hwaddr.sa_family != AF_UNSPEC)
4944 return -EINVAL; 4945 return -EINVAL;
4945 if (!netif_device_present(dev)) 4946 if (!netif_device_present(dev))
4946 return -ENODEV; 4947 return -ENODEV;
4947 return dev_mc_add_global(dev, ifr->ifr_hwaddr.sa_data); 4948 return dev_mc_add_global(dev, ifr->ifr_hwaddr.sa_data);
4948 4949
4949 case SIOCDELMULTI: 4950 case SIOCDELMULTI:
4950 if (!ops->ndo_set_rx_mode || 4951 if (!ops->ndo_set_rx_mode ||
4951 ifr->ifr_hwaddr.sa_family != AF_UNSPEC) 4952 ifr->ifr_hwaddr.sa_family != AF_UNSPEC)
4952 return -EINVAL; 4953 return -EINVAL;
4953 if (!netif_device_present(dev)) 4954 if (!netif_device_present(dev))
4954 return -ENODEV; 4955 return -ENODEV;
4955 return dev_mc_del_global(dev, ifr->ifr_hwaddr.sa_data); 4956 return dev_mc_del_global(dev, ifr->ifr_hwaddr.sa_data);
4956 4957
4957 case SIOCSIFTXQLEN: 4958 case SIOCSIFTXQLEN:
4958 if (ifr->ifr_qlen < 0) 4959 if (ifr->ifr_qlen < 0)
4959 return -EINVAL; 4960 return -EINVAL;
4960 dev->tx_queue_len = ifr->ifr_qlen; 4961 dev->tx_queue_len = ifr->ifr_qlen;
4961 return 0; 4962 return 0;
4962 4963
4963 case SIOCSIFNAME: 4964 case SIOCSIFNAME:
4964 ifr->ifr_newname[IFNAMSIZ-1] = '\0'; 4965 ifr->ifr_newname[IFNAMSIZ-1] = '\0';
4965 return dev_change_name(dev, ifr->ifr_newname); 4966 return dev_change_name(dev, ifr->ifr_newname);
4966 4967
4967 case SIOCSHWTSTAMP: 4968 case SIOCSHWTSTAMP:
4968 err = net_hwtstamp_validate(ifr); 4969 err = net_hwtstamp_validate(ifr);
4969 if (err) 4970 if (err)
4970 return err; 4971 return err;
4971 /* fall through */ 4972 /* fall through */
4972 4973
4973 /* 4974 /*
4974 * Unknown or private ioctl 4975 * Unknown or private ioctl
4975 */ 4976 */
4976 default: 4977 default:
4977 if ((cmd >= SIOCDEVPRIVATE && 4978 if ((cmd >= SIOCDEVPRIVATE &&
4978 cmd <= SIOCDEVPRIVATE + 15) || 4979 cmd <= SIOCDEVPRIVATE + 15) ||
4979 cmd == SIOCBONDENSLAVE || 4980 cmd == SIOCBONDENSLAVE ||
4980 cmd == SIOCBONDRELEASE || 4981 cmd == SIOCBONDRELEASE ||
4981 cmd == SIOCBONDSETHWADDR || 4982 cmd == SIOCBONDSETHWADDR ||
4982 cmd == SIOCBONDSLAVEINFOQUERY || 4983 cmd == SIOCBONDSLAVEINFOQUERY ||
4983 cmd == SIOCBONDINFOQUERY || 4984 cmd == SIOCBONDINFOQUERY ||
4984 cmd == SIOCBONDCHANGEACTIVE || 4985 cmd == SIOCBONDCHANGEACTIVE ||
4985 cmd == SIOCGMIIPHY || 4986 cmd == SIOCGMIIPHY ||
4986 cmd == SIOCGMIIREG || 4987 cmd == SIOCGMIIREG ||
4987 cmd == SIOCSMIIREG || 4988 cmd == SIOCSMIIREG ||
4988 cmd == SIOCBRADDIF || 4989 cmd == SIOCBRADDIF ||
4989 cmd == SIOCBRDELIF || 4990 cmd == SIOCBRDELIF ||
4990 cmd == SIOCSHWTSTAMP || 4991 cmd == SIOCSHWTSTAMP ||
4991 cmd == SIOCWANDEV) { 4992 cmd == SIOCWANDEV) {
4992 err = -EOPNOTSUPP; 4993 err = -EOPNOTSUPP;
4993 if (ops->ndo_do_ioctl) { 4994 if (ops->ndo_do_ioctl) {
4994 if (netif_device_present(dev)) 4995 if (netif_device_present(dev))
4995 err = ops->ndo_do_ioctl(dev, ifr, cmd); 4996 err = ops->ndo_do_ioctl(dev, ifr, cmd);
4996 else 4997 else
4997 err = -ENODEV; 4998 err = -ENODEV;
4998 } 4999 }
4999 } else 5000 } else
5000 err = -EINVAL; 5001 err = -EINVAL;
5001 5002
5002 } 5003 }
5003 return err; 5004 return err;
5004 } 5005 }
5005 5006
5006 /* 5007 /*
5007 * This function handles all "interface"-type I/O control requests. The actual 5008 * This function handles all "interface"-type I/O control requests. The actual
5008 * 'doing' part of this is dev_ifsioc above. 5009 * 'doing' part of this is dev_ifsioc above.
5009 */ 5010 */
5010 5011
5011 /** 5012 /**
5012 * dev_ioctl - network device ioctl 5013 * dev_ioctl - network device ioctl
5013 * @net: the applicable net namespace 5014 * @net: the applicable net namespace
5014 * @cmd: command to issue 5015 * @cmd: command to issue
5015 * @arg: pointer to a struct ifreq in user space 5016 * @arg: pointer to a struct ifreq in user space
5016 * 5017 *
5017 * Issue ioctl functions to devices. This is normally called by the 5018 * Issue ioctl functions to devices. This is normally called by the
5018 * user space syscall interfaces but can sometimes be useful for 5019 * user space syscall interfaces but can sometimes be useful for
5019 * other purposes. The return value is the return from the syscall if 5020 * other purposes. The return value is the return from the syscall if
5020 * positive or a negative errno code on error. 5021 * positive or a negative errno code on error.
5021 */ 5022 */
5022 5023
5023 int dev_ioctl(struct net *net, unsigned int cmd, void __user *arg) 5024 int dev_ioctl(struct net *net, unsigned int cmd, void __user *arg)
5024 { 5025 {
5025 struct ifreq ifr; 5026 struct ifreq ifr;
5026 int ret; 5027 int ret;
5027 char *colon; 5028 char *colon;
5028 5029
5029 /* One special case: SIOCGIFCONF takes ifconf argument 5030 /* One special case: SIOCGIFCONF takes ifconf argument
5030 and requires shared lock, because it sleeps writing 5031 and requires shared lock, because it sleeps writing
5031 to user space. 5032 to user space.
5032 */ 5033 */
5033 5034
5034 if (cmd == SIOCGIFCONF) { 5035 if (cmd == SIOCGIFCONF) {
5035 rtnl_lock(); 5036 rtnl_lock();
5036 ret = dev_ifconf(net, (char __user *) arg); 5037 ret = dev_ifconf(net, (char __user *) arg);
5037 rtnl_unlock(); 5038 rtnl_unlock();
5038 return ret; 5039 return ret;
5039 } 5040 }
5040 if (cmd == SIOCGIFNAME) 5041 if (cmd == SIOCGIFNAME)
5041 return dev_ifname(net, (struct ifreq __user *)arg); 5042 return dev_ifname(net, (struct ifreq __user *)arg);
5042 5043
5043 if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) 5044 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
5044 return -EFAULT; 5045 return -EFAULT;
5045 5046
5046 ifr.ifr_name[IFNAMSIZ-1] = 0; 5047 ifr.ifr_name[IFNAMSIZ-1] = 0;
5047 5048
5048 colon = strchr(ifr.ifr_name, ':'); 5049 colon = strchr(ifr.ifr_name, ':');
5049 if (colon) 5050 if (colon)
5050 *colon = 0; 5051 *colon = 0;
5051 5052
5052 /* 5053 /*
5053 * See which interface the caller is talking about. 5054 * See which interface the caller is talking about.
5054 */ 5055 */
5055 5056
5056 switch (cmd) { 5057 switch (cmd) {
5057 /* 5058 /*
5058 * These ioctl calls: 5059 * These ioctl calls:
5059 * - can be done by all. 5060 * - can be done by all.
5060 * - atomic and do not require locking. 5061 * - atomic and do not require locking.
5061 * - return a value 5062 * - return a value
5062 */ 5063 */
5063 case SIOCGIFFLAGS: 5064 case SIOCGIFFLAGS:
5064 case SIOCGIFMETRIC: 5065 case SIOCGIFMETRIC:
5065 case SIOCGIFMTU: 5066 case SIOCGIFMTU:
5066 case SIOCGIFHWADDR: 5067 case SIOCGIFHWADDR:
5067 case SIOCGIFSLAVE: 5068 case SIOCGIFSLAVE:
5068 case SIOCGIFMAP: 5069 case SIOCGIFMAP:
5069 case SIOCGIFINDEX: 5070 case SIOCGIFINDEX:
5070 case SIOCGIFTXQLEN: 5071 case SIOCGIFTXQLEN:
5071 dev_load(net, ifr.ifr_name); 5072 dev_load(net, ifr.ifr_name);
5072 rcu_read_lock(); 5073 rcu_read_lock();
5073 ret = dev_ifsioc_locked(net, &ifr, cmd); 5074 ret = dev_ifsioc_locked(net, &ifr, cmd);
5074 rcu_read_unlock(); 5075 rcu_read_unlock();
5075 if (!ret) { 5076 if (!ret) {
5076 if (colon) 5077 if (colon)
5077 *colon = ':'; 5078 *colon = ':';
5078 if (copy_to_user(arg, &ifr, 5079 if (copy_to_user(arg, &ifr,
5079 sizeof(struct ifreq))) 5080 sizeof(struct ifreq)))
5080 ret = -EFAULT; 5081 ret = -EFAULT;
5081 } 5082 }
5082 return ret; 5083 return ret;
5083 5084
5084 case SIOCETHTOOL: 5085 case SIOCETHTOOL:
5085 dev_load(net, ifr.ifr_name); 5086 dev_load(net, ifr.ifr_name);
5086 rtnl_lock(); 5087 rtnl_lock();
5087 ret = dev_ethtool(net, &ifr); 5088 ret = dev_ethtool(net, &ifr);
5088 rtnl_unlock(); 5089 rtnl_unlock();
5089 if (!ret) { 5090 if (!ret) {
5090 if (colon) 5091 if (colon)
5091 *colon = ':'; 5092 *colon = ':';
5092 if (copy_to_user(arg, &ifr, 5093 if (copy_to_user(arg, &ifr,
5093 sizeof(struct ifreq))) 5094 sizeof(struct ifreq)))
5094 ret = -EFAULT; 5095 ret = -EFAULT;
5095 } 5096 }
5096 return ret; 5097 return ret;
5097 5098
5098 /* 5099 /*
5099 * These ioctl calls: 5100 * These ioctl calls:
5100 * - require superuser power. 5101 * - require superuser power.
5101 * - require strict serialization. 5102 * - require strict serialization.
5102 * - return a value 5103 * - return a value
5103 */ 5104 */
5104 case SIOCGMIIPHY: 5105 case SIOCGMIIPHY:
5105 case SIOCGMIIREG: 5106 case SIOCGMIIREG:
5106 case SIOCSIFNAME: 5107 case SIOCSIFNAME:
5107 if (!capable(CAP_NET_ADMIN)) 5108 if (!capable(CAP_NET_ADMIN))
5108 return -EPERM; 5109 return -EPERM;
5109 dev_load(net, ifr.ifr_name); 5110 dev_load(net, ifr.ifr_name);
5110 rtnl_lock(); 5111 rtnl_lock();
5111 ret = dev_ifsioc(net, &ifr, cmd); 5112 ret = dev_ifsioc(net, &ifr, cmd);
5112 rtnl_unlock(); 5113 rtnl_unlock();
5113 if (!ret) { 5114 if (!ret) {
5114 if (colon) 5115 if (colon)
5115 *colon = ':'; 5116 *colon = ':';
5116 if (copy_to_user(arg, &ifr, 5117 if (copy_to_user(arg, &ifr,
5117 sizeof(struct ifreq))) 5118 sizeof(struct ifreq)))
5118 ret = -EFAULT; 5119 ret = -EFAULT;
5119 } 5120 }
5120 return ret; 5121 return ret;
5121 5122
5122 /* 5123 /*
5123 * These ioctl calls: 5124 * These ioctl calls:
5124 * - require superuser power. 5125 * - require superuser power.
5125 * - require strict serialization. 5126 * - require strict serialization.
5126 * - do not return a value 5127 * - do not return a value
5127 */ 5128 */
5128 case SIOCSIFFLAGS: 5129 case SIOCSIFFLAGS:
5129 case SIOCSIFMETRIC: 5130 case SIOCSIFMETRIC:
5130 case SIOCSIFMTU: 5131 case SIOCSIFMTU:
5131 case SIOCSIFMAP: 5132 case SIOCSIFMAP:
5132 case SIOCSIFHWADDR: 5133 case SIOCSIFHWADDR:
5133 case SIOCSIFSLAVE: 5134 case SIOCSIFSLAVE:
5134 case SIOCADDMULTI: 5135 case SIOCADDMULTI:
5135 case SIOCDELMULTI: 5136 case SIOCDELMULTI:
5136 case SIOCSIFHWBROADCAST: 5137 case SIOCSIFHWBROADCAST:
5137 case SIOCSIFTXQLEN: 5138 case SIOCSIFTXQLEN:
5138 case SIOCSMIIREG: 5139 case SIOCSMIIREG:
5139 case SIOCBONDENSLAVE: 5140 case SIOCBONDENSLAVE:
5140 case SIOCBONDRELEASE: 5141 case SIOCBONDRELEASE:
5141 case SIOCBONDSETHWADDR: 5142 case SIOCBONDSETHWADDR:
5142 case SIOCBONDCHANGEACTIVE: 5143 case SIOCBONDCHANGEACTIVE:
5143 case SIOCBRADDIF: 5144 case SIOCBRADDIF:
5144 case SIOCBRDELIF: 5145 case SIOCBRDELIF:
5145 case SIOCSHWTSTAMP: 5146 case SIOCSHWTSTAMP:
5146 if (!capable(CAP_NET_ADMIN)) 5147 if (!capable(CAP_NET_ADMIN))
5147 return -EPERM; 5148 return -EPERM;
5148 /* fall through */ 5149 /* fall through */
5149 case SIOCBONDSLAVEINFOQUERY: 5150 case SIOCBONDSLAVEINFOQUERY:
5150 case SIOCBONDINFOQUERY: 5151 case SIOCBONDINFOQUERY:
5151 dev_load(net, ifr.ifr_name); 5152 dev_load(net, ifr.ifr_name);
5152 rtnl_lock(); 5153 rtnl_lock();
5153 ret = dev_ifsioc(net, &ifr, cmd); 5154 ret = dev_ifsioc(net, &ifr, cmd);
5154 rtnl_unlock(); 5155 rtnl_unlock();
5155 return ret; 5156 return ret;
5156 5157
5157 case SIOCGIFMEM: 5158 case SIOCGIFMEM:
5158 /* Get the per device memory space. We can add this but 5159 /* Get the per device memory space. We can add this but
5159 * currently do not support it */ 5160 * currently do not support it */
5160 case SIOCSIFMEM: 5161 case SIOCSIFMEM:
5161 /* Set the per device memory buffer space. 5162 /* Set the per device memory buffer space.
5162 * Not applicable in our case */ 5163 * Not applicable in our case */
5163 case SIOCSIFLINK: 5164 case SIOCSIFLINK:
5164 return -ENOTTY; 5165 return -ENOTTY;
5165 5166
5166 /* 5167 /*
5167 * Unknown or private ioctl. 5168 * Unknown or private ioctl.
5168 */ 5169 */
5169 default: 5170 default:
5170 if (cmd == SIOCWANDEV || 5171 if (cmd == SIOCWANDEV ||
5171 (cmd >= SIOCDEVPRIVATE && 5172 (cmd >= SIOCDEVPRIVATE &&
5172 cmd <= SIOCDEVPRIVATE + 15)) { 5173 cmd <= SIOCDEVPRIVATE + 15)) {
5173 dev_load(net, ifr.ifr_name); 5174 dev_load(net, ifr.ifr_name);
5174 rtnl_lock(); 5175 rtnl_lock();
5175 ret = dev_ifsioc(net, &ifr, cmd); 5176 ret = dev_ifsioc(net, &ifr, cmd);
5176 rtnl_unlock(); 5177 rtnl_unlock();
5177 if (!ret && copy_to_user(arg, &ifr, 5178 if (!ret && copy_to_user(arg, &ifr,
5178 sizeof(struct ifreq))) 5179 sizeof(struct ifreq)))
5179 ret = -EFAULT; 5180 ret = -EFAULT;
5180 return ret; 5181 return ret;
5181 } 5182 }
5182 /* Take care of Wireless Extensions */ 5183 /* Take care of Wireless Extensions */
5183 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) 5184 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)
5184 return wext_handle_ioctl(net, &ifr, cmd, arg); 5185 return wext_handle_ioctl(net, &ifr, cmd, arg);
5185 return -ENOTTY; 5186 return -ENOTTY;
5186 } 5187 }
5187 } 5188 }
5188 5189
5189 5190
5190 /** 5191 /**
5191 * dev_new_index - allocate an ifindex 5192 * dev_new_index - allocate an ifindex
5192 * @net: the applicable net namespace 5193 * @net: the applicable net namespace
5193 * 5194 *
5194 * Returns a suitable unique value for a new device interface 5195 * Returns a suitable unique value for a new device interface
5195 * number. The caller must hold the rtnl semaphore or the 5196 * number. The caller must hold the rtnl semaphore or the
5196 * dev_base_lock to be sure it remains unique. 5197 * dev_base_lock to be sure it remains unique.
5197 */ 5198 */
5198 static int dev_new_index(struct net *net) 5199 static int dev_new_index(struct net *net)
5199 { 5200 {
5200 static int ifindex; 5201 static int ifindex;
5201 for (;;) { 5202 for (;;) {
5202 if (++ifindex <= 0) 5203 if (++ifindex <= 0)
5203 ifindex = 1; 5204 ifindex = 1;
5204 if (!__dev_get_by_index(net, ifindex)) 5205 if (!__dev_get_by_index(net, ifindex))
5205 return ifindex; 5206 return ifindex;
5206 } 5207 }
5207 } 5208 }
5208 5209
5209 /* Delayed registration/unregisteration */ 5210 /* Delayed registration/unregisteration */
5210 static LIST_HEAD(net_todo_list); 5211 static LIST_HEAD(net_todo_list);
5211 5212
5212 static void net_set_todo(struct net_device *dev) 5213 static void net_set_todo(struct net_device *dev)
5213 { 5214 {
5214 list_add_tail(&dev->todo_list, &net_todo_list); 5215 list_add_tail(&dev->todo_list, &net_todo_list);
5215 } 5216 }
5216 5217
5217 static void rollback_registered_many(struct list_head *head) 5218 static void rollback_registered_many(struct list_head *head)
5218 { 5219 {
5219 struct net_device *dev, *tmp; 5220 struct net_device *dev, *tmp;
5220 5221
5221 BUG_ON(dev_boot_phase); 5222 BUG_ON(dev_boot_phase);
5222 ASSERT_RTNL(); 5223 ASSERT_RTNL();
5223 5224
5224 list_for_each_entry_safe(dev, tmp, head, unreg_list) { 5225 list_for_each_entry_safe(dev, tmp, head, unreg_list) {
5225 /* Some devices call without registering 5226 /* Some devices call without registering
5226 * for initialization unwind. Remove those 5227 * for initialization unwind. Remove those
5227 * devices and proceed with the remaining. 5228 * devices and proceed with the remaining.
5228 */ 5229 */
5229 if (dev->reg_state == NETREG_UNINITIALIZED) { 5230 if (dev->reg_state == NETREG_UNINITIALIZED) {
5230 pr_debug("unregister_netdevice: device %s/%p never was registered\n", 5231 pr_debug("unregister_netdevice: device %s/%p never was registered\n",
5231 dev->name, dev); 5232 dev->name, dev);
5232 5233
5233 WARN_ON(1); 5234 WARN_ON(1);
5234 list_del(&dev->unreg_list); 5235 list_del(&dev->unreg_list);
5235 continue; 5236 continue;
5236 } 5237 }
5237 dev->dismantle = true; 5238 dev->dismantle = true;
5238 BUG_ON(dev->reg_state != NETREG_REGISTERED); 5239 BUG_ON(dev->reg_state != NETREG_REGISTERED);
5239 } 5240 }
5240 5241
5241 /* If device is running, close it first. */ 5242 /* If device is running, close it first. */
5242 dev_close_many(head); 5243 dev_close_many(head);
5243 5244
5244 list_for_each_entry(dev, head, unreg_list) { 5245 list_for_each_entry(dev, head, unreg_list) {
5245 /* And unlink it from device chain. */ 5246 /* And unlink it from device chain. */
5246 unlist_netdevice(dev); 5247 unlist_netdevice(dev);
5247 5248
5248 dev->reg_state = NETREG_UNREGISTERING; 5249 dev->reg_state = NETREG_UNREGISTERING;
5249 } 5250 }
5250 5251
5251 synchronize_net(); 5252 synchronize_net();
5252 5253
5253 list_for_each_entry(dev, head, unreg_list) { 5254 list_for_each_entry(dev, head, unreg_list) {
5254 /* Shutdown queueing discipline. */ 5255 /* Shutdown queueing discipline. */
5255 dev_shutdown(dev); 5256 dev_shutdown(dev);
5256 5257
5257 5258
5258 /* Notify protocols, that we are about to destroy 5259 /* Notify protocols, that we are about to destroy
5259 this device. They should clean all the things. 5260 this device. They should clean all the things.
5260 */ 5261 */
5261 call_netdevice_notifiers(NETDEV_UNREGISTER, dev); 5262 call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5262 5263
5263 if (!dev->rtnl_link_ops || 5264 if (!dev->rtnl_link_ops ||
5264 dev->rtnl_link_state == RTNL_LINK_INITIALIZED) 5265 dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5265 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U); 5266 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
5266 5267
5267 /* 5268 /*
5268 * Flush the unicast and multicast chains 5269 * Flush the unicast and multicast chains
5269 */ 5270 */
5270 dev_uc_flush(dev); 5271 dev_uc_flush(dev);
5271 dev_mc_flush(dev); 5272 dev_mc_flush(dev);
5272 5273
5273 if (dev->netdev_ops->ndo_uninit) 5274 if (dev->netdev_ops->ndo_uninit)
5274 dev->netdev_ops->ndo_uninit(dev); 5275 dev->netdev_ops->ndo_uninit(dev);
5275 5276
5276 /* Notifier chain MUST detach us from master device. */ 5277 /* Notifier chain MUST detach us from master device. */
5277 WARN_ON(dev->master); 5278 WARN_ON(dev->master);
5278 5279
5279 /* Remove entries from kobject tree */ 5280 /* Remove entries from kobject tree */
5280 netdev_unregister_kobject(dev); 5281 netdev_unregister_kobject(dev);
5281 } 5282 }
5282 5283
5283 /* Process any work delayed until the end of the batch */ 5284 /* Process any work delayed until the end of the batch */
5284 dev = list_first_entry(head, struct net_device, unreg_list); 5285 dev = list_first_entry(head, struct net_device, unreg_list);
5285 call_netdevice_notifiers(NETDEV_UNREGISTER_BATCH, dev); 5286 call_netdevice_notifiers(NETDEV_UNREGISTER_BATCH, dev);
5286 5287
5287 synchronize_net(); 5288 synchronize_net();
5288 5289
5289 list_for_each_entry(dev, head, unreg_list) 5290 list_for_each_entry(dev, head, unreg_list)
5290 dev_put(dev); 5291 dev_put(dev);
5291 } 5292 }
5292 5293
5293 static void rollback_registered(struct net_device *dev) 5294 static void rollback_registered(struct net_device *dev)
5294 { 5295 {
5295 LIST_HEAD(single); 5296 LIST_HEAD(single);
5296 5297
5297 list_add(&dev->unreg_list, &single); 5298 list_add(&dev->unreg_list, &single);
5298 rollback_registered_many(&single); 5299 rollback_registered_many(&single);
5299 list_del(&single); 5300 list_del(&single);
5300 } 5301 }
5301 5302
5302 static netdev_features_t netdev_fix_features(struct net_device *dev, 5303 static netdev_features_t netdev_fix_features(struct net_device *dev,
5303 netdev_features_t features) 5304 netdev_features_t features)
5304 { 5305 {
5305 /* Fix illegal checksum combinations */ 5306 /* Fix illegal checksum combinations */
5306 if ((features & NETIF_F_HW_CSUM) && 5307 if ((features & NETIF_F_HW_CSUM) &&
5307 (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { 5308 (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
5308 netdev_warn(dev, "mixed HW and IP checksum settings.\n"); 5309 netdev_warn(dev, "mixed HW and IP checksum settings.\n");
5309 features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5310 features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5310 } 5311 }
5311 5312
5312 /* Fix illegal SG+CSUM combinations. */ 5313 /* Fix illegal SG+CSUM combinations. */
5313 if ((features & NETIF_F_SG) && 5314 if ((features & NETIF_F_SG) &&
5314 !(features & NETIF_F_ALL_CSUM)) { 5315 !(features & NETIF_F_ALL_CSUM)) {
5315 netdev_dbg(dev, 5316 netdev_dbg(dev,
5316 "Dropping NETIF_F_SG since no checksum feature.\n"); 5317 "Dropping NETIF_F_SG since no checksum feature.\n");
5317 features &= ~NETIF_F_SG; 5318 features &= ~NETIF_F_SG;
5318 } 5319 }
5319 5320
5320 /* TSO requires that SG is present as well. */ 5321 /* TSO requires that SG is present as well. */
5321 if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) { 5322 if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
5322 netdev_dbg(dev, "Dropping TSO features since no SG feature.\n"); 5323 netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
5323 features &= ~NETIF_F_ALL_TSO; 5324 features &= ~NETIF_F_ALL_TSO;
5324 } 5325 }
5325 5326
5326 /* TSO ECN requires that TSO is present as well. */ 5327 /* TSO ECN requires that TSO is present as well. */
5327 if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN) 5328 if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
5328 features &= ~NETIF_F_TSO_ECN; 5329 features &= ~NETIF_F_TSO_ECN;
5329 5330
5330 /* Software GSO depends on SG. */ 5331 /* Software GSO depends on SG. */
5331 if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) { 5332 if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
5332 netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n"); 5333 netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
5333 features &= ~NETIF_F_GSO; 5334 features &= ~NETIF_F_GSO;
5334 } 5335 }
5335 5336
5336 /* UFO needs SG and checksumming */ 5337 /* UFO needs SG and checksumming */
5337 if (features & NETIF_F_UFO) { 5338 if (features & NETIF_F_UFO) {
5338 /* maybe split UFO into V4 and V6? */ 5339 /* maybe split UFO into V4 and V6? */
5339 if (!((features & NETIF_F_GEN_CSUM) || 5340 if (!((features & NETIF_F_GEN_CSUM) ||
5340 (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM)) 5341 (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))
5341 == (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) { 5342 == (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
5342 netdev_dbg(dev, 5343 netdev_dbg(dev,
5343 "Dropping NETIF_F_UFO since no checksum offload features.\n"); 5344 "Dropping NETIF_F_UFO since no checksum offload features.\n");
5344 features &= ~NETIF_F_UFO; 5345 features &= ~NETIF_F_UFO;
5345 } 5346 }
5346 5347
5347 if (!(features & NETIF_F_SG)) { 5348 if (!(features & NETIF_F_SG)) {
5348 netdev_dbg(dev, 5349 netdev_dbg(dev,
5349 "Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n"); 5350 "Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
5350 features &= ~NETIF_F_UFO; 5351 features &= ~NETIF_F_UFO;
5351 } 5352 }
5352 } 5353 }
5353 5354
5354 return features; 5355 return features;
5355 } 5356 }
5356 5357
5357 int __netdev_update_features(struct net_device *dev) 5358 int __netdev_update_features(struct net_device *dev)
5358 { 5359 {
5359 netdev_features_t features; 5360 netdev_features_t features;
5360 int err = 0; 5361 int err = 0;
5361 5362
5362 ASSERT_RTNL(); 5363 ASSERT_RTNL();
5363 5364
5364 features = netdev_get_wanted_features(dev); 5365 features = netdev_get_wanted_features(dev);
5365 5366
5366 if (dev->netdev_ops->ndo_fix_features) 5367 if (dev->netdev_ops->ndo_fix_features)
5367 features = dev->netdev_ops->ndo_fix_features(dev, features); 5368 features = dev->netdev_ops->ndo_fix_features(dev, features);
5368 5369
5369 /* driver might be less strict about feature dependencies */ 5370 /* driver might be less strict about feature dependencies */
5370 features = netdev_fix_features(dev, features); 5371 features = netdev_fix_features(dev, features);
5371 5372
5372 if (dev->features == features) 5373 if (dev->features == features)
5373 return 0; 5374 return 0;
5374 5375
5375 netdev_dbg(dev, "Features changed: %pNF -> %pNF\n", 5376 netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
5376 &dev->features, &features); 5377 &dev->features, &features);
5377 5378
5378 if (dev->netdev_ops->ndo_set_features) 5379 if (dev->netdev_ops->ndo_set_features)
5379 err = dev->netdev_ops->ndo_set_features(dev, features); 5380 err = dev->netdev_ops->ndo_set_features(dev, features);
5380 5381
5381 if (unlikely(err < 0)) { 5382 if (unlikely(err < 0)) {
5382 netdev_err(dev, 5383 netdev_err(dev,
5383 "set_features() failed (%d); wanted %pNF, left %pNF\n", 5384 "set_features() failed (%d); wanted %pNF, left %pNF\n",
5384 err, &features, &dev->features); 5385 err, &features, &dev->features);
5385 return -1; 5386 return -1;
5386 } 5387 }
5387 5388
5388 if (!err) 5389 if (!err)
5389 dev->features = features; 5390 dev->features = features;
5390 5391
5391 return 1; 5392 return 1;
5392 } 5393 }
5393 5394
5394 /** 5395 /**
5395 * netdev_update_features - recalculate device features 5396 * netdev_update_features - recalculate device features
5396 * @dev: the device to check 5397 * @dev: the device to check
5397 * 5398 *
5398 * Recalculate dev->features set and send notifications if it 5399 * Recalculate dev->features set and send notifications if it
5399 * has changed. Should be called after driver or hardware dependent 5400 * has changed. Should be called after driver or hardware dependent
5400 * conditions might have changed that influence the features. 5401 * conditions might have changed that influence the features.
5401 */ 5402 */
5402 void netdev_update_features(struct net_device *dev) 5403 void netdev_update_features(struct net_device *dev)
5403 { 5404 {
5404 if (__netdev_update_features(dev)) 5405 if (__netdev_update_features(dev))
5405 netdev_features_change(dev); 5406 netdev_features_change(dev);
5406 } 5407 }
5407 EXPORT_SYMBOL(netdev_update_features); 5408 EXPORT_SYMBOL(netdev_update_features);
5408 5409
5409 /** 5410 /**
5410 * netdev_change_features - recalculate device features 5411 * netdev_change_features - recalculate device features
5411 * @dev: the device to check 5412 * @dev: the device to check
5412 * 5413 *
5413 * Recalculate dev->features set and send notifications even 5414 * Recalculate dev->features set and send notifications even
5414 * if they have not changed. Should be called instead of 5415 * if they have not changed. Should be called instead of
5415 * netdev_update_features() if also dev->vlan_features might 5416 * netdev_update_features() if also dev->vlan_features might
5416 * have changed to allow the changes to be propagated to stacked 5417 * have changed to allow the changes to be propagated to stacked
5417 * VLAN devices. 5418 * VLAN devices.
5418 */ 5419 */
5419 void netdev_change_features(struct net_device *dev) 5420 void netdev_change_features(struct net_device *dev)
5420 { 5421 {
5421 __netdev_update_features(dev); 5422 __netdev_update_features(dev);
5422 netdev_features_change(dev); 5423 netdev_features_change(dev);
5423 } 5424 }
5424 EXPORT_SYMBOL(netdev_change_features); 5425 EXPORT_SYMBOL(netdev_change_features);
5425 5426
5426 /** 5427 /**
5427 * netif_stacked_transfer_operstate - transfer operstate 5428 * netif_stacked_transfer_operstate - transfer operstate
5428 * @rootdev: the root or lower level device to transfer state from 5429 * @rootdev: the root or lower level device to transfer state from
5429 * @dev: the device to transfer operstate to 5430 * @dev: the device to transfer operstate to
5430 * 5431 *
5431 * Transfer operational state from root to device. This is normally 5432 * Transfer operational state from root to device. This is normally
5432 * called when a stacking relationship exists between the root 5433 * called when a stacking relationship exists between the root
5433 * device and the device(a leaf device). 5434 * device and the device(a leaf device).
5434 */ 5435 */
5435 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5436 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5436 struct net_device *dev) 5437 struct net_device *dev)
5437 { 5438 {
5438 if (rootdev->operstate == IF_OPER_DORMANT) 5439 if (rootdev->operstate == IF_OPER_DORMANT)
5439 netif_dormant_on(dev); 5440 netif_dormant_on(dev);
5440 else 5441 else
5441 netif_dormant_off(dev); 5442 netif_dormant_off(dev);
5442 5443
5443 if (netif_carrier_ok(rootdev)) { 5444 if (netif_carrier_ok(rootdev)) {
5444 if (!netif_carrier_ok(dev)) 5445 if (!netif_carrier_ok(dev))
5445 netif_carrier_on(dev); 5446 netif_carrier_on(dev);
5446 } else { 5447 } else {
5447 if (netif_carrier_ok(dev)) 5448 if (netif_carrier_ok(dev))
5448 netif_carrier_off(dev); 5449 netif_carrier_off(dev);
5449 } 5450 }
5450 } 5451 }
5451 EXPORT_SYMBOL(netif_stacked_transfer_operstate); 5452 EXPORT_SYMBOL(netif_stacked_transfer_operstate);
5452 5453
5453 #ifdef CONFIG_RPS 5454 #ifdef CONFIG_RPS
5454 static int netif_alloc_rx_queues(struct net_device *dev) 5455 static int netif_alloc_rx_queues(struct net_device *dev)
5455 { 5456 {
5456 unsigned int i, count = dev->num_rx_queues; 5457 unsigned int i, count = dev->num_rx_queues;
5457 struct netdev_rx_queue *rx; 5458 struct netdev_rx_queue *rx;
5458 5459
5459 BUG_ON(count < 1); 5460 BUG_ON(count < 1);
5460 5461
5461 rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL); 5462 rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
5462 if (!rx) { 5463 if (!rx) {
5463 pr_err("netdev: Unable to allocate %u rx queues\n", count); 5464 pr_err("netdev: Unable to allocate %u rx queues\n", count);
5464 return -ENOMEM; 5465 return -ENOMEM;
5465 } 5466 }
5466 dev->_rx = rx; 5467 dev->_rx = rx;
5467 5468
5468 for (i = 0; i < count; i++) 5469 for (i = 0; i < count; i++)
5469 rx[i].dev = dev; 5470 rx[i].dev = dev;
5470 return 0; 5471 return 0;
5471 } 5472 }
5472 #endif 5473 #endif
5473 5474
5474 static void netdev_init_one_queue(struct net_device *dev, 5475 static void netdev_init_one_queue(struct net_device *dev,
5475 struct netdev_queue *queue, void *_unused) 5476 struct netdev_queue *queue, void *_unused)
5476 { 5477 {
5477 /* Initialize queue lock */ 5478 /* Initialize queue lock */
5478 spin_lock_init(&queue->_xmit_lock); 5479 spin_lock_init(&queue->_xmit_lock);
5479 netdev_set_xmit_lockdep_class(&queue->_xmit_lock, dev->type); 5480 netdev_set_xmit_lockdep_class(&queue->_xmit_lock, dev->type);
5480 queue->xmit_lock_owner = -1; 5481 queue->xmit_lock_owner = -1;
5481 netdev_queue_numa_node_write(queue, NUMA_NO_NODE); 5482 netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
5482 queue->dev = dev; 5483 queue->dev = dev;
5483 #ifdef CONFIG_BQL 5484 #ifdef CONFIG_BQL
5484 dql_init(&queue->dql, HZ); 5485 dql_init(&queue->dql, HZ);
5485 #endif 5486 #endif
5486 } 5487 }
5487 5488
5488 static int netif_alloc_netdev_queues(struct net_device *dev) 5489 static int netif_alloc_netdev_queues(struct net_device *dev)
5489 { 5490 {
5490 unsigned int count = dev->num_tx_queues; 5491 unsigned int count = dev->num_tx_queues;
5491 struct netdev_queue *tx; 5492 struct netdev_queue *tx;
5492 5493
5493 BUG_ON(count < 1); 5494 BUG_ON(count < 1);
5494 5495
5495 tx = kcalloc(count, sizeof(struct netdev_queue), GFP_KERNEL); 5496 tx = kcalloc(count, sizeof(struct netdev_queue), GFP_KERNEL);
5496 if (!tx) { 5497 if (!tx) {
5497 pr_err("netdev: Unable to allocate %u tx queues\n", count); 5498 pr_err("netdev: Unable to allocate %u tx queues\n", count);
5498 return -ENOMEM; 5499 return -ENOMEM;
5499 } 5500 }
5500 dev->_tx = tx; 5501 dev->_tx = tx;
5501 5502
5502 netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL); 5503 netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
5503 spin_lock_init(&dev->tx_global_lock); 5504 spin_lock_init(&dev->tx_global_lock);
5504 5505
5505 return 0; 5506 return 0;
5506 } 5507 }
5507 5508
5508 /** 5509 /**
5509 * register_netdevice - register a network device 5510 * register_netdevice - register a network device
5510 * @dev: device to register 5511 * @dev: device to register
5511 * 5512 *
5512 * Take a completed network device structure and add it to the kernel 5513 * Take a completed network device structure and add it to the kernel
5513 * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier 5514 * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
5514 * chain. 0 is returned on success. A negative errno code is returned 5515 * chain. 0 is returned on success. A negative errno code is returned
5515 * on a failure to set up the device, or if the name is a duplicate. 5516 * on a failure to set up the device, or if the name is a duplicate.
5516 * 5517 *
5517 * Callers must hold the rtnl semaphore. You may want 5518 * Callers must hold the rtnl semaphore. You may want
5518 * register_netdev() instead of this. 5519 * register_netdev() instead of this.
5519 * 5520 *
5520 * BUGS: 5521 * BUGS:
5521 * The locking appears insufficient to guarantee two parallel registers 5522 * The locking appears insufficient to guarantee two parallel registers
5522 * will not get the same name. 5523 * will not get the same name.
5523 */ 5524 */
5524 5525
5525 int register_netdevice(struct net_device *dev) 5526 int register_netdevice(struct net_device *dev)
5526 { 5527 {
5527 int ret; 5528 int ret;
5528 struct net *net = dev_net(dev); 5529 struct net *net = dev_net(dev);
5529 5530
5530 BUG_ON(dev_boot_phase); 5531 BUG_ON(dev_boot_phase);
5531 ASSERT_RTNL(); 5532 ASSERT_RTNL();
5532 5533
5533 might_sleep(); 5534 might_sleep();
5534 5535
5535 /* When net_device's are persistent, this will be fatal. */ 5536 /* When net_device's are persistent, this will be fatal. */
5536 BUG_ON(dev->reg_state != NETREG_UNINITIALIZED); 5537 BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
5537 BUG_ON(!net); 5538 BUG_ON(!net);
5538 5539
5539 spin_lock_init(&dev->addr_list_lock); 5540 spin_lock_init(&dev->addr_list_lock);
5540 netdev_set_addr_lockdep_class(dev); 5541 netdev_set_addr_lockdep_class(dev);
5541 5542
5542 dev->iflink = -1; 5543 dev->iflink = -1;
5543 5544
5544 ret = dev_get_valid_name(dev, dev->name); 5545 ret = dev_get_valid_name(dev, dev->name);
5545 if (ret < 0) 5546 if (ret < 0)
5546 goto out; 5547 goto out;
5547 5548
5548 /* Init, if this function is available */ 5549 /* Init, if this function is available */
5549 if (dev->netdev_ops->ndo_init) { 5550 if (dev->netdev_ops->ndo_init) {
5550 ret = dev->netdev_ops->ndo_init(dev); 5551 ret = dev->netdev_ops->ndo_init(dev);
5551 if (ret) { 5552 if (ret) {
5552 if (ret > 0) 5553 if (ret > 0)
5553 ret = -EIO; 5554 ret = -EIO;
5554 goto out; 5555 goto out;
5555 } 5556 }
5556 } 5557 }
5557 5558
5558 dev->ifindex = dev_new_index(net); 5559 dev->ifindex = dev_new_index(net);
5559 if (dev->iflink == -1) 5560 if (dev->iflink == -1)
5560 dev->iflink = dev->ifindex; 5561 dev->iflink = dev->ifindex;
5561 5562
5562 /* Transfer changeable features to wanted_features and enable 5563 /* Transfer changeable features to wanted_features and enable
5563 * software offloads (GSO and GRO). 5564 * software offloads (GSO and GRO).
5564 */ 5565 */
5565 dev->hw_features |= NETIF_F_SOFT_FEATURES; 5566 dev->hw_features |= NETIF_F_SOFT_FEATURES;
5566 dev->features |= NETIF_F_SOFT_FEATURES; 5567 dev->features |= NETIF_F_SOFT_FEATURES;
5567 dev->wanted_features = dev->features & dev->hw_features; 5568 dev->wanted_features = dev->features & dev->hw_features;
5568 5569
5569 /* Turn on no cache copy if HW is doing checksum */ 5570 /* Turn on no cache copy if HW is doing checksum */
5570 if (!(dev->flags & IFF_LOOPBACK)) { 5571 if (!(dev->flags & IFF_LOOPBACK)) {
5571 dev->hw_features |= NETIF_F_NOCACHE_COPY; 5572 dev->hw_features |= NETIF_F_NOCACHE_COPY;
5572 if (dev->features & NETIF_F_ALL_CSUM) { 5573 if (dev->features & NETIF_F_ALL_CSUM) {
5573 dev->wanted_features |= NETIF_F_NOCACHE_COPY; 5574 dev->wanted_features |= NETIF_F_NOCACHE_COPY;
5574 dev->features |= NETIF_F_NOCACHE_COPY; 5575 dev->features |= NETIF_F_NOCACHE_COPY;
5575 } 5576 }
5576 } 5577 }
5577 5578
5578 /* Make NETIF_F_HIGHDMA inheritable to VLAN devices. 5579 /* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
5579 */ 5580 */
5580 dev->vlan_features |= NETIF_F_HIGHDMA; 5581 dev->vlan_features |= NETIF_F_HIGHDMA;
5581 5582
5582 ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev); 5583 ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
5583 ret = notifier_to_errno(ret); 5584 ret = notifier_to_errno(ret);
5584 if (ret) 5585 if (ret)
5585 goto err_uninit; 5586 goto err_uninit;
5586 5587
5587 ret = netdev_register_kobject(dev); 5588 ret = netdev_register_kobject(dev);
5588 if (ret) 5589 if (ret)
5589 goto err_uninit; 5590 goto err_uninit;
5590 dev->reg_state = NETREG_REGISTERED; 5591 dev->reg_state = NETREG_REGISTERED;
5591 5592
5592 __netdev_update_features(dev); 5593 __netdev_update_features(dev);
5593 5594
5594 /* 5595 /*
5595 * Default initial state at registry is that the 5596 * Default initial state at registry is that the
5596 * device is present. 5597 * device is present.
5597 */ 5598 */
5598 5599
5599 set_bit(__LINK_STATE_PRESENT, &dev->state); 5600 set_bit(__LINK_STATE_PRESENT, &dev->state);
5600 5601
5601 dev_init_scheduler(dev); 5602 dev_init_scheduler(dev);
5602 dev_hold(dev); 5603 dev_hold(dev);
5603 list_netdevice(dev); 5604 list_netdevice(dev);
5604 5605
5605 /* Notify protocols, that a new device appeared. */ 5606 /* Notify protocols, that a new device appeared. */
5606 ret = call_netdevice_notifiers(NETDEV_REGISTER, dev); 5607 ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
5607 ret = notifier_to_errno(ret); 5608 ret = notifier_to_errno(ret);
5608 if (ret) { 5609 if (ret) {
5609 rollback_registered(dev); 5610 rollback_registered(dev);
5610 dev->reg_state = NETREG_UNREGISTERED; 5611 dev->reg_state = NETREG_UNREGISTERED;
5611 } 5612 }
5612 /* 5613 /*
5613 * Prevent userspace races by waiting until the network 5614 * Prevent userspace races by waiting until the network
5614 * device is fully setup before sending notifications. 5615 * device is fully setup before sending notifications.
5615 */ 5616 */
5616 if (!dev->rtnl_link_ops || 5617 if (!dev->rtnl_link_ops ||
5617 dev->rtnl_link_state == RTNL_LINK_INITIALIZED) 5618 dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
5618 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U); 5619 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
5619 5620
5620 out: 5621 out:
5621 return ret; 5622 return ret;
5622 5623
5623 err_uninit: 5624 err_uninit:
5624 if (dev->netdev_ops->ndo_uninit) 5625 if (dev->netdev_ops->ndo_uninit)
5625 dev->netdev_ops->ndo_uninit(dev); 5626 dev->netdev_ops->ndo_uninit(dev);
5626 goto out; 5627 goto out;
5627 } 5628 }
5628 EXPORT_SYMBOL(register_netdevice); 5629 EXPORT_SYMBOL(register_netdevice);
5629 5630
5630 /** 5631 /**
5631 * init_dummy_netdev - init a dummy network device for NAPI 5632 * init_dummy_netdev - init a dummy network device for NAPI
5632 * @dev: device to init 5633 * @dev: device to init
5633 * 5634 *
5634 * This takes a network device structure and initialize the minimum 5635 * This takes a network device structure and initialize the minimum
5635 * amount of fields so it can be used to schedule NAPI polls without 5636 * amount of fields so it can be used to schedule NAPI polls without
5636 * registering a full blown interface. This is to be used by drivers 5637 * registering a full blown interface. This is to be used by drivers
5637 * that need to tie several hardware interfaces to a single NAPI 5638 * that need to tie several hardware interfaces to a single NAPI
5638 * poll scheduler due to HW limitations. 5639 * poll scheduler due to HW limitations.
5639 */ 5640 */
5640 int init_dummy_netdev(struct net_device *dev) 5641 int init_dummy_netdev(struct net_device *dev)
5641 { 5642 {
5642 /* Clear everything. Note we don't initialize spinlocks 5643 /* Clear everything. Note we don't initialize spinlocks
5643 * are they aren't supposed to be taken by any of the 5644 * are they aren't supposed to be taken by any of the
5644 * NAPI code and this dummy netdev is supposed to be 5645 * NAPI code and this dummy netdev is supposed to be
5645 * only ever used for NAPI polls 5646 * only ever used for NAPI polls
5646 */ 5647 */
5647 memset(dev, 0, sizeof(struct net_device)); 5648 memset(dev, 0, sizeof(struct net_device));
5648 5649
5649 /* make sure we BUG if trying to hit standard 5650 /* make sure we BUG if trying to hit standard
5650 * register/unregister code path 5651 * register/unregister code path
5651 */ 5652 */
5652 dev->reg_state = NETREG_DUMMY; 5653 dev->reg_state = NETREG_DUMMY;
5653 5654
5654 /* NAPI wants this */ 5655 /* NAPI wants this */
5655 INIT_LIST_HEAD(&dev->napi_list); 5656 INIT_LIST_HEAD(&dev->napi_list);
5656 5657
5657 /* a dummy interface is started by default */ 5658 /* a dummy interface is started by default */
5658 set_bit(__LINK_STATE_PRESENT, &dev->state); 5659 set_bit(__LINK_STATE_PRESENT, &dev->state);
5659 set_bit(__LINK_STATE_START, &dev->state); 5660 set_bit(__LINK_STATE_START, &dev->state);
5660 5661
5661 /* Note : We dont allocate pcpu_refcnt for dummy devices, 5662 /* Note : We dont allocate pcpu_refcnt for dummy devices,
5662 * because users of this 'device' dont need to change 5663 * because users of this 'device' dont need to change
5663 * its refcount. 5664 * its refcount.
5664 */ 5665 */
5665 5666
5666 return 0; 5667 return 0;
5667 } 5668 }
5668 EXPORT_SYMBOL_GPL(init_dummy_netdev); 5669 EXPORT_SYMBOL_GPL(init_dummy_netdev);
5669 5670
5670 5671
5671 /** 5672 /**
5672 * register_netdev - register a network device 5673 * register_netdev - register a network device
5673 * @dev: device to register 5674 * @dev: device to register
5674 * 5675 *
5675 * Take a completed network device structure and add it to the kernel 5676 * Take a completed network device structure and add it to the kernel
5676 * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier 5677 * interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
5677 * chain. 0 is returned on success. A negative errno code is returned 5678 * chain. 0 is returned on success. A negative errno code is returned
5678 * on a failure to set up the device, or if the name is a duplicate. 5679 * on a failure to set up the device, or if the name is a duplicate.
5679 * 5680 *
5680 * This is a wrapper around register_netdevice that takes the rtnl semaphore 5681 * This is a wrapper around register_netdevice that takes the rtnl semaphore
5681 * and expands the device name if you passed a format string to 5682 * and expands the device name if you passed a format string to
5682 * alloc_netdev. 5683 * alloc_netdev.
5683 */ 5684 */
5684 int register_netdev(struct net_device *dev) 5685 int register_netdev(struct net_device *dev)
5685 { 5686 {
5686 int err; 5687 int err;
5687 5688
5688 rtnl_lock(); 5689 rtnl_lock();
5689 err = register_netdevice(dev); 5690 err = register_netdevice(dev);
5690 rtnl_unlock(); 5691 rtnl_unlock();
5691 return err; 5692 return err;
5692 } 5693 }
5693 EXPORT_SYMBOL(register_netdev); 5694 EXPORT_SYMBOL(register_netdev);
5694 5695
5695 int netdev_refcnt_read(const struct net_device *dev) 5696 int netdev_refcnt_read(const struct net_device *dev)
5696 { 5697 {
5697 int i, refcnt = 0; 5698 int i, refcnt = 0;
5698 5699
5699 for_each_possible_cpu(i) 5700 for_each_possible_cpu(i)
5700 refcnt += *per_cpu_ptr(dev->pcpu_refcnt, i); 5701 refcnt += *per_cpu_ptr(dev->pcpu_refcnt, i);
5701 return refcnt; 5702 return refcnt;
5702 } 5703 }
5703 EXPORT_SYMBOL(netdev_refcnt_read); 5704 EXPORT_SYMBOL(netdev_refcnt_read);
5704 5705
5705 /* 5706 /*
5706 * netdev_wait_allrefs - wait until all references are gone. 5707 * netdev_wait_allrefs - wait until all references are gone.
5707 * 5708 *
5708 * This is called when unregistering network devices. 5709 * This is called when unregistering network devices.
5709 * 5710 *
5710 * Any protocol or device that holds a reference should register 5711 * Any protocol or device that holds a reference should register
5711 * for netdevice notification, and cleanup and put back the 5712 * for netdevice notification, and cleanup and put back the
5712 * reference if they receive an UNREGISTER event. 5713 * reference if they receive an UNREGISTER event.
5713 * We can get stuck here if buggy protocols don't correctly 5714 * We can get stuck here if buggy protocols don't correctly
5714 * call dev_put. 5715 * call dev_put.
5715 */ 5716 */
5716 static void netdev_wait_allrefs(struct net_device *dev) 5717 static void netdev_wait_allrefs(struct net_device *dev)
5717 { 5718 {
5718 unsigned long rebroadcast_time, warning_time; 5719 unsigned long rebroadcast_time, warning_time;
5719 int refcnt; 5720 int refcnt;
5720 5721
5721 linkwatch_forget_dev(dev); 5722 linkwatch_forget_dev(dev);
5722 5723
5723 rebroadcast_time = warning_time = jiffies; 5724 rebroadcast_time = warning_time = jiffies;
5724 refcnt = netdev_refcnt_read(dev); 5725 refcnt = netdev_refcnt_read(dev);
5725 5726
5726 while (refcnt != 0) { 5727 while (refcnt != 0) {
5727 if (time_after(jiffies, rebroadcast_time + 1 * HZ)) { 5728 if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
5728 rtnl_lock(); 5729 rtnl_lock();
5729 5730
5730 /* Rebroadcast unregister notification */ 5731 /* Rebroadcast unregister notification */
5731 call_netdevice_notifiers(NETDEV_UNREGISTER, dev); 5732 call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5732 /* don't resend NETDEV_UNREGISTER_BATCH, _BATCH users 5733 /* don't resend NETDEV_UNREGISTER_BATCH, _BATCH users
5733 * should have already handle it the first time */ 5734 * should have already handle it the first time */
5734 5735
5735 if (test_bit(__LINK_STATE_LINKWATCH_PENDING, 5736 if (test_bit(__LINK_STATE_LINKWATCH_PENDING,
5736 &dev->state)) { 5737 &dev->state)) {
5737 /* We must not have linkwatch events 5738 /* We must not have linkwatch events
5738 * pending on unregister. If this 5739 * pending on unregister. If this
5739 * happens, we simply run the queue 5740 * happens, we simply run the queue
5740 * unscheduled, resulting in a noop 5741 * unscheduled, resulting in a noop
5741 * for this device. 5742 * for this device.
5742 */ 5743 */
5743 linkwatch_run_queue(); 5744 linkwatch_run_queue();
5744 } 5745 }
5745 5746
5746 __rtnl_unlock(); 5747 __rtnl_unlock();
5747 5748
5748 rebroadcast_time = jiffies; 5749 rebroadcast_time = jiffies;
5749 } 5750 }
5750 5751
5751 msleep(250); 5752 msleep(250);
5752 5753
5753 refcnt = netdev_refcnt_read(dev); 5754 refcnt = netdev_refcnt_read(dev);
5754 5755
5755 if (time_after(jiffies, warning_time + 10 * HZ)) { 5756 if (time_after(jiffies, warning_time + 10 * HZ)) {
5756 pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n", 5757 pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
5757 dev->name, refcnt); 5758 dev->name, refcnt);
5758 warning_time = jiffies; 5759 warning_time = jiffies;
5759 } 5760 }
5760 } 5761 }
5761 } 5762 }
5762 5763
5763 /* The sequence is: 5764 /* The sequence is:
5764 * 5765 *
5765 * rtnl_lock(); 5766 * rtnl_lock();
5766 * ... 5767 * ...
5767 * register_netdevice(x1); 5768 * register_netdevice(x1);
5768 * register_netdevice(x2); 5769 * register_netdevice(x2);
5769 * ... 5770 * ...
5770 * unregister_netdevice(y1); 5771 * unregister_netdevice(y1);
5771 * unregister_netdevice(y2); 5772 * unregister_netdevice(y2);
5772 * ... 5773 * ...
5773 * rtnl_unlock(); 5774 * rtnl_unlock();
5774 * free_netdev(y1); 5775 * free_netdev(y1);
5775 * free_netdev(y2); 5776 * free_netdev(y2);
5776 * 5777 *
5777 * We are invoked by rtnl_unlock(). 5778 * We are invoked by rtnl_unlock().
5778 * This allows us to deal with problems: 5779 * This allows us to deal with problems:
5779 * 1) We can delete sysfs objects which invoke hotplug 5780 * 1) We can delete sysfs objects which invoke hotplug
5780 * without deadlocking with linkwatch via keventd. 5781 * without deadlocking with linkwatch via keventd.
5781 * 2) Since we run with the RTNL semaphore not held, we can sleep 5782 * 2) Since we run with the RTNL semaphore not held, we can sleep
5782 * safely in order to wait for the netdev refcnt to drop to zero. 5783 * safely in order to wait for the netdev refcnt to drop to zero.
5783 * 5784 *
5784 * We must not return until all unregister events added during 5785 * We must not return until all unregister events added during
5785 * the interval the lock was held have been completed. 5786 * the interval the lock was held have been completed.
5786 */ 5787 */
5787 void netdev_run_todo(void) 5788 void netdev_run_todo(void)
5788 { 5789 {
5789 struct list_head list; 5790 struct list_head list;
5790 5791
5791 /* Snapshot list, allow later requests */ 5792 /* Snapshot list, allow later requests */
5792 list_replace_init(&net_todo_list, &list); 5793 list_replace_init(&net_todo_list, &list);
5793 5794
5794 __rtnl_unlock(); 5795 __rtnl_unlock();
5795 5796
5796 /* Wait for rcu callbacks to finish before attempting to drain 5797 /* Wait for rcu callbacks to finish before attempting to drain
5797 * the device list. This usually avoids a 250ms wait. 5798 * the device list. This usually avoids a 250ms wait.
5798 */ 5799 */
5799 if (!list_empty(&list)) 5800 if (!list_empty(&list))
5800 rcu_barrier(); 5801 rcu_barrier();
5801 5802
5802 while (!list_empty(&list)) { 5803 while (!list_empty(&list)) {
5803 struct net_device *dev 5804 struct net_device *dev
5804 = list_first_entry(&list, struct net_device, todo_list); 5805 = list_first_entry(&list, struct net_device, todo_list);
5805 list_del(&dev->todo_list); 5806 list_del(&dev->todo_list);
5806 5807
5807 if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) { 5808 if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
5808 pr_err("network todo '%s' but state %d\n", 5809 pr_err("network todo '%s' but state %d\n",
5809 dev->name, dev->reg_state); 5810 dev->name, dev->reg_state);
5810 dump_stack(); 5811 dump_stack();
5811 continue; 5812 continue;
5812 } 5813 }
5813 5814
5814 dev->reg_state = NETREG_UNREGISTERED; 5815 dev->reg_state = NETREG_UNREGISTERED;
5815 5816
5816 on_each_cpu(flush_backlog, dev, 1); 5817 on_each_cpu(flush_backlog, dev, 1);
5817 5818
5818 netdev_wait_allrefs(dev); 5819 netdev_wait_allrefs(dev);
5819 5820
5820 /* paranoia */ 5821 /* paranoia */
5821 BUG_ON(netdev_refcnt_read(dev)); 5822 BUG_ON(netdev_refcnt_read(dev));
5822 WARN_ON(rcu_access_pointer(dev->ip_ptr)); 5823 WARN_ON(rcu_access_pointer(dev->ip_ptr));
5823 WARN_ON(rcu_access_pointer(dev->ip6_ptr)); 5824 WARN_ON(rcu_access_pointer(dev->ip6_ptr));
5824 WARN_ON(dev->dn_ptr); 5825 WARN_ON(dev->dn_ptr);
5825 5826
5826 if (dev->destructor) 5827 if (dev->destructor)
5827 dev->destructor(dev); 5828 dev->destructor(dev);
5828 5829
5829 /* Free network device */ 5830 /* Free network device */
5830 kobject_put(&dev->dev.kobj); 5831 kobject_put(&dev->dev.kobj);
5831 } 5832 }
5832 } 5833 }
5833 5834
5834 /* Convert net_device_stats to rtnl_link_stats64. They have the same 5835 /* Convert net_device_stats to rtnl_link_stats64. They have the same
5835 * fields in the same order, with only the type differing. 5836 * fields in the same order, with only the type differing.
5836 */ 5837 */
5837 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5838 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
5838 const struct net_device_stats *netdev_stats) 5839 const struct net_device_stats *netdev_stats)
5839 { 5840 {
5840 #if BITS_PER_LONG == 64 5841 #if BITS_PER_LONG == 64
5841 BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats)); 5842 BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
5842 memcpy(stats64, netdev_stats, sizeof(*stats64)); 5843 memcpy(stats64, netdev_stats, sizeof(*stats64));
5843 #else 5844 #else
5844 size_t i, n = sizeof(*stats64) / sizeof(u64); 5845 size_t i, n = sizeof(*stats64) / sizeof(u64);
5845 const unsigned long *src = (const unsigned long *)netdev_stats; 5846 const unsigned long *src = (const unsigned long *)netdev_stats;
5846 u64 *dst = (u64 *)stats64; 5847 u64 *dst = (u64 *)stats64;
5847 5848
5848 BUILD_BUG_ON(sizeof(*netdev_stats) / sizeof(unsigned long) != 5849 BUILD_BUG_ON(sizeof(*netdev_stats) / sizeof(unsigned long) !=
5849 sizeof(*stats64) / sizeof(u64)); 5850 sizeof(*stats64) / sizeof(u64));
5850 for (i = 0; i < n; i++) 5851 for (i = 0; i < n; i++)
5851 dst[i] = src[i]; 5852 dst[i] = src[i];
5852 #endif 5853 #endif
5853 } 5854 }
5854 EXPORT_SYMBOL(netdev_stats_to_stats64); 5855 EXPORT_SYMBOL(netdev_stats_to_stats64);
5855 5856
5856 /** 5857 /**
5857 * dev_get_stats - get network device statistics 5858 * dev_get_stats - get network device statistics
5858 * @dev: device to get statistics from 5859 * @dev: device to get statistics from
5859 * @storage: place to store stats 5860 * @storage: place to store stats
5860 * 5861 *
5861 * Get network statistics from device. Return @storage. 5862 * Get network statistics from device. Return @storage.
5862 * The device driver may provide its own method by setting 5863 * The device driver may provide its own method by setting
5863 * dev->netdev_ops->get_stats64 or dev->netdev_ops->get_stats; 5864 * dev->netdev_ops->get_stats64 or dev->netdev_ops->get_stats;
5864 * otherwise the internal statistics structure is used. 5865 * otherwise the internal statistics structure is used.
5865 */ 5866 */
5866 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 5867 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
5867 struct rtnl_link_stats64 *storage) 5868 struct rtnl_link_stats64 *storage)
5868 { 5869 {
5869 const struct net_device_ops *ops = dev->netdev_ops; 5870 const struct net_device_ops *ops = dev->netdev_ops;
5870 5871
5871 if (ops->ndo_get_stats64) { 5872 if (ops->ndo_get_stats64) {
5872 memset(storage, 0, sizeof(*storage)); 5873 memset(storage, 0, sizeof(*storage));
5873 ops->ndo_get_stats64(dev, storage); 5874 ops->ndo_get_stats64(dev, storage);
5874 } else if (ops->ndo_get_stats) { 5875 } else if (ops->ndo_get_stats) {
5875 netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev)); 5876 netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
5876 } else { 5877 } else {
5877 netdev_stats_to_stats64(storage, &dev->stats); 5878 netdev_stats_to_stats64(storage, &dev->stats);
5878 } 5879 }
5879 storage->rx_dropped += atomic_long_read(&dev->rx_dropped); 5880 storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
5880 return storage; 5881 return storage;
5881 } 5882 }
5882 EXPORT_SYMBOL(dev_get_stats); 5883 EXPORT_SYMBOL(dev_get_stats);
5883 5884
5884 struct netdev_queue *dev_ingress_queue_create(struct net_device *dev) 5885 struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
5885 { 5886 {
5886 struct netdev_queue *queue = dev_ingress_queue(dev); 5887 struct netdev_queue *queue = dev_ingress_queue(dev);
5887 5888
5888 #ifdef CONFIG_NET_CLS_ACT 5889 #ifdef CONFIG_NET_CLS_ACT
5889 if (queue) 5890 if (queue)
5890 return queue; 5891 return queue;
5891 queue = kzalloc(sizeof(*queue), GFP_KERNEL); 5892 queue = kzalloc(sizeof(*queue), GFP_KERNEL);
5892 if (!queue) 5893 if (!queue)
5893 return NULL; 5894 return NULL;
5894 netdev_init_one_queue(dev, queue, NULL); 5895 netdev_init_one_queue(dev, queue, NULL);
5895 queue->qdisc = &noop_qdisc; 5896 queue->qdisc = &noop_qdisc;
5896 queue->qdisc_sleeping = &noop_qdisc; 5897 queue->qdisc_sleeping = &noop_qdisc;
5897 rcu_assign_pointer(dev->ingress_queue, queue); 5898 rcu_assign_pointer(dev->ingress_queue, queue);
5898 #endif 5899 #endif
5899 return queue; 5900 return queue;
5900 } 5901 }
5901 5902
5902 /** 5903 /**
5903 * alloc_netdev_mqs - allocate network device 5904 * alloc_netdev_mqs - allocate network device
5904 * @sizeof_priv: size of private data to allocate space for 5905 * @sizeof_priv: size of private data to allocate space for
5905 * @name: device name format string 5906 * @name: device name format string
5906 * @setup: callback to initialize device 5907 * @setup: callback to initialize device
5907 * @txqs: the number of TX subqueues to allocate 5908 * @txqs: the number of TX subqueues to allocate
5908 * @rxqs: the number of RX subqueues to allocate 5909 * @rxqs: the number of RX subqueues to allocate
5909 * 5910 *
5910 * Allocates a struct net_device with private data area for driver use 5911 * Allocates a struct net_device with private data area for driver use
5911 * and performs basic initialization. Also allocates subquue structs 5912 * and performs basic initialization. Also allocates subquue structs
5912 * for each queue on the device. 5913 * for each queue on the device.
5913 */ 5914 */
5914 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 5915 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
5915 void (*setup)(struct net_device *), 5916 void (*setup)(struct net_device *),
5916 unsigned int txqs, unsigned int rxqs) 5917 unsigned int txqs, unsigned int rxqs)
5917 { 5918 {
5918 struct net_device *dev; 5919 struct net_device *dev;
5919 size_t alloc_size; 5920 size_t alloc_size;
5920 struct net_device *p; 5921 struct net_device *p;
5921 5922
5922 BUG_ON(strlen(name) >= sizeof(dev->name)); 5923 BUG_ON(strlen(name) >= sizeof(dev->name));
5923 5924
5924 if (txqs < 1) { 5925 if (txqs < 1) {
5925 pr_err("alloc_netdev: Unable to allocate device with zero queues\n"); 5926 pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
5926 return NULL; 5927 return NULL;
5927 } 5928 }
5928 5929
5929 #ifdef CONFIG_RPS 5930 #ifdef CONFIG_RPS
5930 if (rxqs < 1) { 5931 if (rxqs < 1) {
5931 pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n"); 5932 pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
5932 return NULL; 5933 return NULL;
5933 } 5934 }
5934 #endif 5935 #endif
5935 5936
5936 alloc_size = sizeof(struct net_device); 5937 alloc_size = sizeof(struct net_device);
5937 if (sizeof_priv) { 5938 if (sizeof_priv) {
5938 /* ensure 32-byte alignment of private area */ 5939 /* ensure 32-byte alignment of private area */
5939 alloc_size = ALIGN(alloc_size, NETDEV_ALIGN); 5940 alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
5940 alloc_size += sizeof_priv; 5941 alloc_size += sizeof_priv;
5941 } 5942 }
5942 /* ensure 32-byte alignment of whole construct */ 5943 /* ensure 32-byte alignment of whole construct */
5943 alloc_size += NETDEV_ALIGN - 1; 5944 alloc_size += NETDEV_ALIGN - 1;
5944 5945
5945 p = kzalloc(alloc_size, GFP_KERNEL); 5946 p = kzalloc(alloc_size, GFP_KERNEL);
5946 if (!p) { 5947 if (!p) {
5947 pr_err("alloc_netdev: Unable to allocate device\n"); 5948 pr_err("alloc_netdev: Unable to allocate device\n");
5948 return NULL; 5949 return NULL;
5949 } 5950 }
5950 5951
5951 dev = PTR_ALIGN(p, NETDEV_ALIGN); 5952 dev = PTR_ALIGN(p, NETDEV_ALIGN);
5952 dev->padded = (char *)dev - (char *)p; 5953 dev->padded = (char *)dev - (char *)p;
5953 5954
5954 dev->pcpu_refcnt = alloc_percpu(int); 5955 dev->pcpu_refcnt = alloc_percpu(int);
5955 if (!dev->pcpu_refcnt) 5956 if (!dev->pcpu_refcnt)
5956 goto free_p; 5957 goto free_p;
5957 5958
5958 if (dev_addr_init(dev)) 5959 if (dev_addr_init(dev))
5959 goto free_pcpu; 5960 goto free_pcpu;
5960 5961
5961 dev_mc_init(dev); 5962 dev_mc_init(dev);
5962 dev_uc_init(dev); 5963 dev_uc_init(dev);
5963 5964
5964 dev_net_set(dev, &init_net); 5965 dev_net_set(dev, &init_net);
5965 5966
5966 dev->gso_max_size = GSO_MAX_SIZE; 5967 dev->gso_max_size = GSO_MAX_SIZE;
5967 5968
5968 INIT_LIST_HEAD(&dev->napi_list); 5969 INIT_LIST_HEAD(&dev->napi_list);
5969 INIT_LIST_HEAD(&dev->unreg_list); 5970 INIT_LIST_HEAD(&dev->unreg_list);
5970 INIT_LIST_HEAD(&dev->link_watch_list); 5971 INIT_LIST_HEAD(&dev->link_watch_list);
5971 dev->priv_flags = IFF_XMIT_DST_RELEASE; 5972 dev->priv_flags = IFF_XMIT_DST_RELEASE;
5972 setup(dev); 5973 setup(dev);
5973 5974
5974 dev->num_tx_queues = txqs; 5975 dev->num_tx_queues = txqs;
5975 dev->real_num_tx_queues = txqs; 5976 dev->real_num_tx_queues = txqs;
5976 if (netif_alloc_netdev_queues(dev)) 5977 if (netif_alloc_netdev_queues(dev))
5977 goto free_all; 5978 goto free_all;
5978 5979
5979 #ifdef CONFIG_RPS 5980 #ifdef CONFIG_RPS
5980 dev->num_rx_queues = rxqs; 5981 dev->num_rx_queues = rxqs;
5981 dev->real_num_rx_queues = rxqs; 5982 dev->real_num_rx_queues = rxqs;
5982 if (netif_alloc_rx_queues(dev)) 5983 if (netif_alloc_rx_queues(dev))
5983 goto free_all; 5984 goto free_all;
5984 #endif 5985 #endif
5985 5986
5986 strcpy(dev->name, name); 5987 strcpy(dev->name, name);
5987 dev->group = INIT_NETDEV_GROUP; 5988 dev->group = INIT_NETDEV_GROUP;
5988 return dev; 5989 return dev;
5989 5990
5990 free_all: 5991 free_all:
5991 free_netdev(dev); 5992 free_netdev(dev);
5992 return NULL; 5993 return NULL;
5993 5994
5994 free_pcpu: 5995 free_pcpu:
5995 free_percpu(dev->pcpu_refcnt); 5996 free_percpu(dev->pcpu_refcnt);
5996 kfree(dev->_tx); 5997 kfree(dev->_tx);
5997 #ifdef CONFIG_RPS 5998 #ifdef CONFIG_RPS
5998 kfree(dev->_rx); 5999 kfree(dev->_rx);
5999 #endif 6000 #endif
6000 6001
6001 free_p: 6002 free_p:
6002 kfree(p); 6003 kfree(p);
6003 return NULL; 6004 return NULL;
6004 } 6005 }
6005 EXPORT_SYMBOL(alloc_netdev_mqs); 6006 EXPORT_SYMBOL(alloc_netdev_mqs);
6006 6007
6007 /** 6008 /**
6008 * free_netdev - free network device 6009 * free_netdev - free network device
6009 * @dev: device 6010 * @dev: device
6010 * 6011 *
6011 * This function does the last stage of destroying an allocated device 6012 * This function does the last stage of destroying an allocated device
6012 * interface. The reference to the device object is released. 6013 * interface. The reference to the device object is released.
6013 * If this is the last reference then it will be freed. 6014 * If this is the last reference then it will be freed.
6014 */ 6015 */
6015 void free_netdev(struct net_device *dev) 6016 void free_netdev(struct net_device *dev)
6016 { 6017 {
6017 struct napi_struct *p, *n; 6018 struct napi_struct *p, *n;
6018 6019
6019 release_net(dev_net(dev)); 6020 release_net(dev_net(dev));
6020 6021
6021 kfree(dev->_tx); 6022 kfree(dev->_tx);
6022 #ifdef CONFIG_RPS 6023 #ifdef CONFIG_RPS
6023 kfree(dev->_rx); 6024 kfree(dev->_rx);
6024 #endif 6025 #endif
6025 6026
6026 kfree(rcu_dereference_protected(dev->ingress_queue, 1)); 6027 kfree(rcu_dereference_protected(dev->ingress_queue, 1));
6027 6028
6028 /* Flush device addresses */ 6029 /* Flush device addresses */
6029 dev_addr_flush(dev); 6030 dev_addr_flush(dev);
6030 6031
6031 list_for_each_entry_safe(p, n, &dev->napi_list, dev_list) 6032 list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
6032 netif_napi_del(p); 6033 netif_napi_del(p);
6033 6034
6034 free_percpu(dev->pcpu_refcnt); 6035 free_percpu(dev->pcpu_refcnt);
6035 dev->pcpu_refcnt = NULL; 6036 dev->pcpu_refcnt = NULL;
6036 6037
6037 /* Compatibility with error handling in drivers */ 6038 /* Compatibility with error handling in drivers */
6038 if (dev->reg_state == NETREG_UNINITIALIZED) { 6039 if (dev->reg_state == NETREG_UNINITIALIZED) {
6039 kfree((char *)dev - dev->padded); 6040 kfree((char *)dev - dev->padded);
6040 return; 6041 return;
6041 } 6042 }
6042 6043
6043 BUG_ON(dev->reg_state != NETREG_UNREGISTERED); 6044 BUG_ON(dev->reg_state != NETREG_UNREGISTERED);
6044 dev->reg_state = NETREG_RELEASED; 6045 dev->reg_state = NETREG_RELEASED;
6045 6046
6046 /* will free via device release */ 6047 /* will free via device release */
6047 put_device(&dev->dev); 6048 put_device(&dev->dev);
6048 } 6049 }
6049 EXPORT_SYMBOL(free_netdev); 6050 EXPORT_SYMBOL(free_netdev);
6050 6051
6051 /** 6052 /**
6052 * synchronize_net - Synchronize with packet receive processing 6053 * synchronize_net - Synchronize with packet receive processing
6053 * 6054 *
6054 * Wait for packets currently being received to be done. 6055 * Wait for packets currently being received to be done.
6055 * Does not block later packets from starting. 6056 * Does not block later packets from starting.
6056 */ 6057 */
6057 void synchronize_net(void) 6058 void synchronize_net(void)
6058 { 6059 {
6059 might_sleep(); 6060 might_sleep();
6060 if (rtnl_is_locked()) 6061 if (rtnl_is_locked())
6061 synchronize_rcu_expedited(); 6062 synchronize_rcu_expedited();
6062 else 6063 else
6063 synchronize_rcu(); 6064 synchronize_rcu();
6064 } 6065 }
6065 EXPORT_SYMBOL(synchronize_net); 6066 EXPORT_SYMBOL(synchronize_net);
6066 6067
6067 /** 6068 /**
6068 * unregister_netdevice_queue - remove device from the kernel 6069 * unregister_netdevice_queue - remove device from the kernel
6069 * @dev: device 6070 * @dev: device
6070 * @head: list 6071 * @head: list
6071 * 6072 *
6072 * This function shuts down a device interface and removes it 6073 * This function shuts down a device interface and removes it
6073 * from the kernel tables. 6074 * from the kernel tables.
6074 * If head not NULL, device is queued to be unregistered later. 6075 * If head not NULL, device is queued to be unregistered later.
6075 * 6076 *
6076 * Callers must hold the rtnl semaphore. You may want 6077 * Callers must hold the rtnl semaphore. You may want
6077 * unregister_netdev() instead of this. 6078 * unregister_netdev() instead of this.
6078 */ 6079 */
6079 6080
6080 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head) 6081 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
6081 { 6082 {
6082 ASSERT_RTNL(); 6083 ASSERT_RTNL();
6083 6084
6084 if (head) { 6085 if (head) {
6085 list_move_tail(&dev->unreg_list, head); 6086 list_move_tail(&dev->unreg_list, head);
6086 } else { 6087 } else {
6087 rollback_registered(dev); 6088 rollback_registered(dev);
6088 /* Finish processing unregister after unlock */ 6089 /* Finish processing unregister after unlock */
6089 net_set_todo(dev); 6090 net_set_todo(dev);
6090 } 6091 }
6091 } 6092 }
6092 EXPORT_SYMBOL(unregister_netdevice_queue); 6093 EXPORT_SYMBOL(unregister_netdevice_queue);
6093 6094
6094 /** 6095 /**
6095 * unregister_netdevice_many - unregister many devices 6096 * unregister_netdevice_many - unregister many devices
6096 * @head: list of devices 6097 * @head: list of devices
6097 */ 6098 */
6098 void unregister_netdevice_many(struct list_head *head) 6099 void unregister_netdevice_many(struct list_head *head)
6099 { 6100 {
6100 struct net_device *dev; 6101 struct net_device *dev;
6101 6102
6102 if (!list_empty(head)) { 6103 if (!list_empty(head)) {
6103 rollback_registered_many(head); 6104 rollback_registered_many(head);
6104 list_for_each_entry(dev, head, unreg_list) 6105 list_for_each_entry(dev, head, unreg_list)
6105 net_set_todo(dev); 6106 net_set_todo(dev);
6106 } 6107 }
6107 } 6108 }
6108 EXPORT_SYMBOL(unregister_netdevice_many); 6109 EXPORT_SYMBOL(unregister_netdevice_many);
6109 6110
6110 /** 6111 /**
6111 * unregister_netdev - remove device from the kernel 6112 * unregister_netdev - remove device from the kernel
6112 * @dev: device 6113 * @dev: device
6113 * 6114 *
6114 * This function shuts down a device interface and removes it 6115 * This function shuts down a device interface and removes it
6115 * from the kernel tables. 6116 * from the kernel tables.
6116 * 6117 *
6117 * This is just a wrapper for unregister_netdevice that takes 6118 * This is just a wrapper for unregister_netdevice that takes
6118 * the rtnl semaphore. In general you want to use this and not 6119 * the rtnl semaphore. In general you want to use this and not
6119 * unregister_netdevice. 6120 * unregister_netdevice.
6120 */ 6121 */
6121 void unregister_netdev(struct net_device *dev) 6122 void unregister_netdev(struct net_device *dev)
6122 { 6123 {
6123 rtnl_lock(); 6124 rtnl_lock();
6124 unregister_netdevice(dev); 6125 unregister_netdevice(dev);
6125 rtnl_unlock(); 6126 rtnl_unlock();
6126 } 6127 }
6127 EXPORT_SYMBOL(unregister_netdev); 6128 EXPORT_SYMBOL(unregister_netdev);
6128 6129
6129 /** 6130 /**
6130 * dev_change_net_namespace - move device to different nethost namespace 6131 * dev_change_net_namespace - move device to different nethost namespace
6131 * @dev: device 6132 * @dev: device
6132 * @net: network namespace 6133 * @net: network namespace
6133 * @pat: If not NULL name pattern to try if the current device name 6134 * @pat: If not NULL name pattern to try if the current device name
6134 * is already taken in the destination network namespace. 6135 * is already taken in the destination network namespace.
6135 * 6136 *
6136 * This function shuts down a device interface and moves it 6137 * This function shuts down a device interface and moves it
6137 * to a new network namespace. On success 0 is returned, on 6138 * to a new network namespace. On success 0 is returned, on
6138 * a failure a netagive errno code is returned. 6139 * a failure a netagive errno code is returned.
6139 * 6140 *
6140 * Callers must hold the rtnl semaphore. 6141 * Callers must hold the rtnl semaphore.
6141 */ 6142 */
6142 6143
6143 int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat) 6144 int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat)
6144 { 6145 {
6145 int err; 6146 int err;
6146 6147
6147 ASSERT_RTNL(); 6148 ASSERT_RTNL();
6148 6149
6149 /* Don't allow namespace local devices to be moved. */ 6150 /* Don't allow namespace local devices to be moved. */
6150 err = -EINVAL; 6151 err = -EINVAL;
6151 if (dev->features & NETIF_F_NETNS_LOCAL) 6152 if (dev->features & NETIF_F_NETNS_LOCAL)
6152 goto out; 6153 goto out;
6153 6154
6154 /* Ensure the device has been registrered */ 6155 /* Ensure the device has been registrered */
6155 err = -EINVAL; 6156 err = -EINVAL;
6156 if (dev->reg_state != NETREG_REGISTERED) 6157 if (dev->reg_state != NETREG_REGISTERED)
6157 goto out; 6158 goto out;
6158 6159
6159 /* Get out if there is nothing todo */ 6160 /* Get out if there is nothing todo */
6160 err = 0; 6161 err = 0;
6161 if (net_eq(dev_net(dev), net)) 6162 if (net_eq(dev_net(dev), net))
6162 goto out; 6163 goto out;
6163 6164
6164 /* Pick the destination device name, and ensure 6165 /* Pick the destination device name, and ensure
6165 * we can use it in the destination network namespace. 6166 * we can use it in the destination network namespace.
6166 */ 6167 */
6167 err = -EEXIST; 6168 err = -EEXIST;
6168 if (__dev_get_by_name(net, dev->name)) { 6169 if (__dev_get_by_name(net, dev->name)) {
6169 /* We get here if we can't use the current device name */ 6170 /* We get here if we can't use the current device name */
6170 if (!pat) 6171 if (!pat)
6171 goto out; 6172 goto out;
6172 if (dev_get_valid_name(dev, pat) < 0) 6173 if (dev_get_valid_name(dev, pat) < 0)
6173 goto out; 6174 goto out;
6174 } 6175 }
6175 6176
6176 /* 6177 /*
6177 * And now a mini version of register_netdevice unregister_netdevice. 6178 * And now a mini version of register_netdevice unregister_netdevice.
6178 */ 6179 */
6179 6180
6180 /* If device is running close it first. */ 6181 /* If device is running close it first. */
6181 dev_close(dev); 6182 dev_close(dev);
6182 6183
6183 /* And unlink it from device chain */ 6184 /* And unlink it from device chain */
6184 err = -ENODEV; 6185 err = -ENODEV;
6185 unlist_netdevice(dev); 6186 unlist_netdevice(dev);
6186 6187
6187 synchronize_net(); 6188 synchronize_net();
6188 6189
6189 /* Shutdown queueing discipline. */ 6190 /* Shutdown queueing discipline. */
6190 dev_shutdown(dev); 6191 dev_shutdown(dev);
6191 6192
6192 /* Notify protocols, that we are about to destroy 6193 /* Notify protocols, that we are about to destroy
6193 this device. They should clean all the things. 6194 this device. They should clean all the things.
6194 6195
6195 Note that dev->reg_state stays at NETREG_REGISTERED. 6196 Note that dev->reg_state stays at NETREG_REGISTERED.
6196 This is wanted because this way 8021q and macvlan know 6197 This is wanted because this way 8021q and macvlan know
6197 the device is just moving and can keep their slaves up. 6198 the device is just moving and can keep their slaves up.
6198 */ 6199 */
6199 call_netdevice_notifiers(NETDEV_UNREGISTER, dev); 6200 call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
6200 call_netdevice_notifiers(NETDEV_UNREGISTER_BATCH, dev); 6201 call_netdevice_notifiers(NETDEV_UNREGISTER_BATCH, dev);
6201 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U); 6202 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
6202 6203
6203 /* 6204 /*
6204 * Flush the unicast and multicast chains 6205 * Flush the unicast and multicast chains
6205 */ 6206 */
6206 dev_uc_flush(dev); 6207 dev_uc_flush(dev);
6207 dev_mc_flush(dev); 6208 dev_mc_flush(dev);
6208 6209
6209 /* Actually switch the network namespace */ 6210 /* Actually switch the network namespace */
6210 dev_net_set(dev, net); 6211 dev_net_set(dev, net);
6211 6212
6212 /* If there is an ifindex conflict assign a new one */ 6213 /* If there is an ifindex conflict assign a new one */
6213 if (__dev_get_by_index(net, dev->ifindex)) { 6214 if (__dev_get_by_index(net, dev->ifindex)) {
6214 int iflink = (dev->iflink == dev->ifindex); 6215 int iflink = (dev->iflink == dev->ifindex);
6215 dev->ifindex = dev_new_index(net); 6216 dev->ifindex = dev_new_index(net);
6216 if (iflink) 6217 if (iflink)
6217 dev->iflink = dev->ifindex; 6218 dev->iflink = dev->ifindex;
6218 } 6219 }
6219 6220
6220 /* Fixup kobjects */ 6221 /* Fixup kobjects */
6221 err = device_rename(&dev->dev, dev->name); 6222 err = device_rename(&dev->dev, dev->name);
6222 WARN_ON(err); 6223 WARN_ON(err);
6223 6224
6224 /* Add the device back in the hashes */ 6225 /* Add the device back in the hashes */
6225 list_netdevice(dev); 6226 list_netdevice(dev);
6226 6227
6227 /* Notify protocols, that a new device appeared. */ 6228 /* Notify protocols, that a new device appeared. */
6228 call_netdevice_notifiers(NETDEV_REGISTER, dev); 6229 call_netdevice_notifiers(NETDEV_REGISTER, dev);
6229 6230
6230 /* 6231 /*
6231 * Prevent userspace races by waiting until the network 6232 * Prevent userspace races by waiting until the network
6232 * device is fully setup before sending notifications. 6233 * device is fully setup before sending notifications.
6233 */ 6234 */
6234 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U); 6235 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
6235 6236
6236 synchronize_net(); 6237 synchronize_net();
6237 err = 0; 6238 err = 0;
6238 out: 6239 out:
6239 return err; 6240 return err;
6240 } 6241 }
6241 EXPORT_SYMBOL_GPL(dev_change_net_namespace); 6242 EXPORT_SYMBOL_GPL(dev_change_net_namespace);
6242 6243
6243 static int dev_cpu_callback(struct notifier_block *nfb, 6244 static int dev_cpu_callback(struct notifier_block *nfb,
6244 unsigned long action, 6245 unsigned long action,
6245 void *ocpu) 6246 void *ocpu)
6246 { 6247 {
6247 struct sk_buff **list_skb; 6248 struct sk_buff **list_skb;
6248 struct sk_buff *skb; 6249 struct sk_buff *skb;
6249 unsigned int cpu, oldcpu = (unsigned long)ocpu; 6250 unsigned int cpu, oldcpu = (unsigned long)ocpu;
6250 struct softnet_data *sd, *oldsd; 6251 struct softnet_data *sd, *oldsd;
6251 6252
6252 if (action != CPU_DEAD && action != CPU_DEAD_FROZEN) 6253 if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
6253 return NOTIFY_OK; 6254 return NOTIFY_OK;
6254 6255
6255 local_irq_disable(); 6256 local_irq_disable();
6256 cpu = smp_processor_id(); 6257 cpu = smp_processor_id();
6257 sd = &per_cpu(softnet_data, cpu); 6258 sd = &per_cpu(softnet_data, cpu);
6258 oldsd = &per_cpu(softnet_data, oldcpu); 6259 oldsd = &per_cpu(softnet_data, oldcpu);
6259 6260
6260 /* Find end of our completion_queue. */ 6261 /* Find end of our completion_queue. */
6261 list_skb = &sd->completion_queue; 6262 list_skb = &sd->completion_queue;
6262 while (*list_skb) 6263 while (*list_skb)
6263 list_skb = &(*list_skb)->next; 6264 list_skb = &(*list_skb)->next;
6264 /* Append completion queue from offline CPU. */ 6265 /* Append completion queue from offline CPU. */
6265 *list_skb = oldsd->completion_queue; 6266 *list_skb = oldsd->completion_queue;
6266 oldsd->completion_queue = NULL; 6267 oldsd->completion_queue = NULL;
6267 6268
6268 /* Append output queue from offline CPU. */ 6269 /* Append output queue from offline CPU. */
6269 if (oldsd->output_queue) { 6270 if (oldsd->output_queue) {
6270 *sd->output_queue_tailp = oldsd->output_queue; 6271 *sd->output_queue_tailp = oldsd->output_queue;
6271 sd->output_queue_tailp = oldsd->output_queue_tailp; 6272 sd->output_queue_tailp = oldsd->output_queue_tailp;
6272 oldsd->output_queue = NULL; 6273 oldsd->output_queue = NULL;
6273 oldsd->output_queue_tailp = &oldsd->output_queue; 6274 oldsd->output_queue_tailp = &oldsd->output_queue;
6274 } 6275 }
6275 /* Append NAPI poll list from offline CPU. */ 6276 /* Append NAPI poll list from offline CPU. */
6276 if (!list_empty(&oldsd->poll_list)) { 6277 if (!list_empty(&oldsd->poll_list)) {
6277 list_splice_init(&oldsd->poll_list, &sd->poll_list); 6278 list_splice_init(&oldsd->poll_list, &sd->poll_list);
6278 raise_softirq_irqoff(NET_RX_SOFTIRQ); 6279 raise_softirq_irqoff(NET_RX_SOFTIRQ);
6279 } 6280 }
6280 6281
6281 raise_softirq_irqoff(NET_TX_SOFTIRQ); 6282 raise_softirq_irqoff(NET_TX_SOFTIRQ);
6282 local_irq_enable(); 6283 local_irq_enable();
6283 6284
6284 /* Process offline CPU's input_pkt_queue */ 6285 /* Process offline CPU's input_pkt_queue */
6285 while ((skb = __skb_dequeue(&oldsd->process_queue))) { 6286 while ((skb = __skb_dequeue(&oldsd->process_queue))) {
6286 netif_rx(skb); 6287 netif_rx(skb);
6287 input_queue_head_incr(oldsd); 6288 input_queue_head_incr(oldsd);
6288 } 6289 }
6289 while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) { 6290 while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
6290 netif_rx(skb); 6291 netif_rx(skb);
6291 input_queue_head_incr(oldsd); 6292 input_queue_head_incr(oldsd);
6292 } 6293 }
6293 6294
6294 return NOTIFY_OK; 6295 return NOTIFY_OK;
6295 } 6296 }
6296 6297
6297 6298
6298 /** 6299 /**
6299 * netdev_increment_features - increment feature set by one 6300 * netdev_increment_features - increment feature set by one
6300 * @all: current feature set 6301 * @all: current feature set
6301 * @one: new feature set 6302 * @one: new feature set
6302 * @mask: mask feature set 6303 * @mask: mask feature set
6303 * 6304 *
6304 * Computes a new feature set after adding a device with feature set 6305 * Computes a new feature set after adding a device with feature set
6305 * @one to the master device with current feature set @all. Will not 6306 * @one to the master device with current feature set @all. Will not
6306 * enable anything that is off in @mask. Returns the new feature set. 6307 * enable anything that is off in @mask. Returns the new feature set.
6307 */ 6308 */
6308 netdev_features_t netdev_increment_features(netdev_features_t all, 6309 netdev_features_t netdev_increment_features(netdev_features_t all,
6309 netdev_features_t one, netdev_features_t mask) 6310 netdev_features_t one, netdev_features_t mask)
6310 { 6311 {
6311 if (mask & NETIF_F_GEN_CSUM) 6312 if (mask & NETIF_F_GEN_CSUM)
6312 mask |= NETIF_F_ALL_CSUM; 6313 mask |= NETIF_F_ALL_CSUM;
6313 mask |= NETIF_F_VLAN_CHALLENGED; 6314 mask |= NETIF_F_VLAN_CHALLENGED;
6314 6315
6315 all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask; 6316 all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
6316 all &= one | ~NETIF_F_ALL_FOR_ALL; 6317 all &= one | ~NETIF_F_ALL_FOR_ALL;
6317 6318
6318 /* If one device supports hw checksumming, set for all. */ 6319 /* If one device supports hw checksumming, set for all. */
6319 if (all & NETIF_F_GEN_CSUM) 6320 if (all & NETIF_F_GEN_CSUM)
6320 all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM); 6321 all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
6321 6322
6322 return all; 6323 return all;
6323 } 6324 }
6324 EXPORT_SYMBOL(netdev_increment_features); 6325 EXPORT_SYMBOL(netdev_increment_features);
6325 6326
6326 static struct hlist_head *netdev_create_hash(void) 6327 static struct hlist_head *netdev_create_hash(void)
6327 { 6328 {
6328 int i; 6329 int i;
6329 struct hlist_head *hash; 6330 struct hlist_head *hash;
6330 6331
6331 hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL); 6332 hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL);
6332 if (hash != NULL) 6333 if (hash != NULL)
6333 for (i = 0; i < NETDEV_HASHENTRIES; i++) 6334 for (i = 0; i < NETDEV_HASHENTRIES; i++)
6334 INIT_HLIST_HEAD(&hash[i]); 6335 INIT_HLIST_HEAD(&hash[i]);
6335 6336
6336 return hash; 6337 return hash;
6337 } 6338 }
6338 6339
6339 /* Initialize per network namespace state */ 6340 /* Initialize per network namespace state */
6340 static int __net_init netdev_init(struct net *net) 6341 static int __net_init netdev_init(struct net *net)
6341 { 6342 {
6342 INIT_LIST_HEAD(&net->dev_base_head); 6343 INIT_LIST_HEAD(&net->dev_base_head);
6343 6344
6344 net->dev_name_head = netdev_create_hash(); 6345 net->dev_name_head = netdev_create_hash();
6345 if (net->dev_name_head == NULL) 6346 if (net->dev_name_head == NULL)
6346 goto err_name; 6347 goto err_name;
6347 6348
6348 net->dev_index_head = netdev_create_hash(); 6349 net->dev_index_head = netdev_create_hash();
6349 if (net->dev_index_head == NULL) 6350 if (net->dev_index_head == NULL)
6350 goto err_idx; 6351 goto err_idx;
6351 6352
6352 return 0; 6353 return 0;
6353 6354
6354 err_idx: 6355 err_idx:
6355 kfree(net->dev_name_head); 6356 kfree(net->dev_name_head);
6356 err_name: 6357 err_name:
6357 return -ENOMEM; 6358 return -ENOMEM;
6358 } 6359 }
6359 6360
6360 /** 6361 /**
6361 * netdev_drivername - network driver for the device 6362 * netdev_drivername - network driver for the device
6362 * @dev: network device 6363 * @dev: network device
6363 * 6364 *
6364 * Determine network driver for device. 6365 * Determine network driver for device.
6365 */ 6366 */
6366 const char *netdev_drivername(const struct net_device *dev) 6367 const char *netdev_drivername(const struct net_device *dev)
6367 { 6368 {
6368 const struct device_driver *driver; 6369 const struct device_driver *driver;
6369 const struct device *parent; 6370 const struct device *parent;
6370 const char *empty = ""; 6371 const char *empty = "";
6371 6372
6372 parent = dev->dev.parent; 6373 parent = dev->dev.parent;
6373 if (!parent) 6374 if (!parent)
6374 return empty; 6375 return empty;
6375 6376
6376 driver = parent->driver; 6377 driver = parent->driver;
6377 if (driver && driver->name) 6378 if (driver && driver->name)
6378 return driver->name; 6379 return driver->name;
6379 return empty; 6380 return empty;
6380 } 6381 }
6381 6382
6382 int __netdev_printk(const char *level, const struct net_device *dev, 6383 int __netdev_printk(const char *level, const struct net_device *dev,
6383 struct va_format *vaf) 6384 struct va_format *vaf)
6384 { 6385 {
6385 int r; 6386 int r;
6386 6387
6387 if (dev && dev->dev.parent) 6388 if (dev && dev->dev.parent)
6388 r = dev_printk(level, dev->dev.parent, "%s: %pV", 6389 r = dev_printk(level, dev->dev.parent, "%s: %pV",
6389 netdev_name(dev), vaf); 6390 netdev_name(dev), vaf);
6390 else if (dev) 6391 else if (dev)
6391 r = printk("%s%s: %pV", level, netdev_name(dev), vaf); 6392 r = printk("%s%s: %pV", level, netdev_name(dev), vaf);
6392 else 6393 else
6393 r = printk("%s(NULL net_device): %pV", level, vaf); 6394 r = printk("%s(NULL net_device): %pV", level, vaf);
6394 6395
6395 return r; 6396 return r;
6396 } 6397 }
6397 EXPORT_SYMBOL(__netdev_printk); 6398 EXPORT_SYMBOL(__netdev_printk);
6398 6399
6399 int netdev_printk(const char *level, const struct net_device *dev, 6400 int netdev_printk(const char *level, const struct net_device *dev,
6400 const char *format, ...) 6401 const char *format, ...)
6401 { 6402 {
6402 struct va_format vaf; 6403 struct va_format vaf;
6403 va_list args; 6404 va_list args;
6404 int r; 6405 int r;
6405 6406
6406 va_start(args, format); 6407 va_start(args, format);
6407 6408
6408 vaf.fmt = format; 6409 vaf.fmt = format;
6409 vaf.va = &args; 6410 vaf.va = &args;
6410 6411
6411 r = __netdev_printk(level, dev, &vaf); 6412 r = __netdev_printk(level, dev, &vaf);
6412 va_end(args); 6413 va_end(args);
6413 6414
6414 return r; 6415 return r;
6415 } 6416 }
6416 EXPORT_SYMBOL(netdev_printk); 6417 EXPORT_SYMBOL(netdev_printk);
6417 6418
6418 #define define_netdev_printk_level(func, level) \ 6419 #define define_netdev_printk_level(func, level) \
6419 int func(const struct net_device *dev, const char *fmt, ...) \ 6420 int func(const struct net_device *dev, const char *fmt, ...) \
6420 { \ 6421 { \
6421 int r; \ 6422 int r; \
6422 struct va_format vaf; \ 6423 struct va_format vaf; \
6423 va_list args; \ 6424 va_list args; \
6424 \ 6425 \
6425 va_start(args, fmt); \ 6426 va_start(args, fmt); \
6426 \ 6427 \
6427 vaf.fmt = fmt; \ 6428 vaf.fmt = fmt; \
6428 vaf.va = &args; \ 6429 vaf.va = &args; \
6429 \ 6430 \
6430 r = __netdev_printk(level, dev, &vaf); \ 6431 r = __netdev_printk(level, dev, &vaf); \
6431 va_end(args); \ 6432 va_end(args); \
6432 \ 6433 \
6433 return r; \ 6434 return r; \
6434 } \ 6435 } \
6435 EXPORT_SYMBOL(func); 6436 EXPORT_SYMBOL(func);
6436 6437
6437 define_netdev_printk_level(netdev_emerg, KERN_EMERG); 6438 define_netdev_printk_level(netdev_emerg, KERN_EMERG);
6438 define_netdev_printk_level(netdev_alert, KERN_ALERT); 6439 define_netdev_printk_level(netdev_alert, KERN_ALERT);
6439 define_netdev_printk_level(netdev_crit, KERN_CRIT); 6440 define_netdev_printk_level(netdev_crit, KERN_CRIT);
6440 define_netdev_printk_level(netdev_err, KERN_ERR); 6441 define_netdev_printk_level(netdev_err, KERN_ERR);
6441 define_netdev_printk_level(netdev_warn, KERN_WARNING); 6442 define_netdev_printk_level(netdev_warn, KERN_WARNING);
6442 define_netdev_printk_level(netdev_notice, KERN_NOTICE); 6443 define_netdev_printk_level(netdev_notice, KERN_NOTICE);
6443 define_netdev_printk_level(netdev_info, KERN_INFO); 6444 define_netdev_printk_level(netdev_info, KERN_INFO);
6444 6445
6445 static void __net_exit netdev_exit(struct net *net) 6446 static void __net_exit netdev_exit(struct net *net)
6446 { 6447 {
6447 kfree(net->dev_name_head); 6448 kfree(net->dev_name_head);
6448 kfree(net->dev_index_head); 6449 kfree(net->dev_index_head);
6449 } 6450 }
6450 6451
6451 static struct pernet_operations __net_initdata netdev_net_ops = { 6452 static struct pernet_operations __net_initdata netdev_net_ops = {
6452 .init = netdev_init, 6453 .init = netdev_init,
6453 .exit = netdev_exit, 6454 .exit = netdev_exit,
6454 }; 6455 };
6455 6456
6456 static void __net_exit default_device_exit(struct net *net) 6457 static void __net_exit default_device_exit(struct net *net)
6457 { 6458 {
6458 struct net_device *dev, *aux; 6459 struct net_device *dev, *aux;
6459 /* 6460 /*
6460 * Push all migratable network devices back to the 6461 * Push all migratable network devices back to the
6461 * initial network namespace 6462 * initial network namespace
6462 */ 6463 */
6463 rtnl_lock(); 6464 rtnl_lock();
6464 for_each_netdev_safe(net, dev, aux) { 6465 for_each_netdev_safe(net, dev, aux) {
6465 int err; 6466 int err;
6466 char fb_name[IFNAMSIZ]; 6467 char fb_name[IFNAMSIZ];
6467 6468
6468 /* Ignore unmoveable devices (i.e. loopback) */ 6469 /* Ignore unmoveable devices (i.e. loopback) */
6469 if (dev->features & NETIF_F_NETNS_LOCAL) 6470 if (dev->features & NETIF_F_NETNS_LOCAL)
6470 continue; 6471 continue;
6471 6472
6472 /* Leave virtual devices for the generic cleanup */ 6473 /* Leave virtual devices for the generic cleanup */
6473 if (dev->rtnl_link_ops) 6474 if (dev->rtnl_link_ops)
6474 continue; 6475 continue;
6475 6476
6476 /* Push remaining network devices to init_net */ 6477 /* Push remaining network devices to init_net */
6477 snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex); 6478 snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
6478 err = dev_change_net_namespace(dev, &init_net, fb_name); 6479 err = dev_change_net_namespace(dev, &init_net, fb_name);
6479 if (err) { 6480 if (err) {
6480 pr_emerg("%s: failed to move %s to init_net: %d\n", 6481 pr_emerg("%s: failed to move %s to init_net: %d\n",
6481 __func__, dev->name, err); 6482 __func__, dev->name, err);
6482 BUG(); 6483 BUG();
6483 } 6484 }
6484 } 6485 }
6485 rtnl_unlock(); 6486 rtnl_unlock();
6486 } 6487 }
6487 6488
6488 static void __net_exit default_device_exit_batch(struct list_head *net_list) 6489 static void __net_exit default_device_exit_batch(struct list_head *net_list)
6489 { 6490 {
6490 /* At exit all network devices most be removed from a network 6491 /* At exit all network devices most be removed from a network
6491 * namespace. Do this in the reverse order of registration. 6492 * namespace. Do this in the reverse order of registration.
6492 * Do this across as many network namespaces as possible to 6493 * Do this across as many network namespaces as possible to
6493 * improve batching efficiency. 6494 * improve batching efficiency.
6494 */ 6495 */
6495 struct net_device *dev; 6496 struct net_device *dev;
6496 struct net *net; 6497 struct net *net;
6497 LIST_HEAD(dev_kill_list); 6498 LIST_HEAD(dev_kill_list);
6498 6499
6499 rtnl_lock(); 6500 rtnl_lock();
6500 list_for_each_entry(net, net_list, exit_list) { 6501 list_for_each_entry(net, net_list, exit_list) {
6501 for_each_netdev_reverse(net, dev) { 6502 for_each_netdev_reverse(net, dev) {
6502 if (dev->rtnl_link_ops) 6503 if (dev->rtnl_link_ops)
6503 dev->rtnl_link_ops->dellink(dev, &dev_kill_list); 6504 dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
6504 else 6505 else
6505 unregister_netdevice_queue(dev, &dev_kill_list); 6506 unregister_netdevice_queue(dev, &dev_kill_list);
6506 } 6507 }
6507 } 6508 }
6508 unregister_netdevice_many(&dev_kill_list); 6509 unregister_netdevice_many(&dev_kill_list);
6509 list_del(&dev_kill_list); 6510 list_del(&dev_kill_list);
6510 rtnl_unlock(); 6511 rtnl_unlock();
6511 } 6512 }
6512 6513
6513 static struct pernet_operations __net_initdata default_device_ops = { 6514 static struct pernet_operations __net_initdata default_device_ops = {
6514 .exit = default_device_exit, 6515 .exit = default_device_exit,
6515 .exit_batch = default_device_exit_batch, 6516 .exit_batch = default_device_exit_batch,
6516 }; 6517 };
6517 6518
6518 /* 6519 /*
6519 * Initialize the DEV module. At boot time this walks the device list and 6520 * Initialize the DEV module. At boot time this walks the device list and
6520 * unhooks any devices that fail to initialise (normally hardware not 6521 * unhooks any devices that fail to initialise (normally hardware not
6521 * present) and leaves us with a valid list of present and active devices. 6522 * present) and leaves us with a valid list of present and active devices.
6522 * 6523 *
6523 */ 6524 */
6524 6525
6525 /* 6526 /*
6526 * This is called single threaded during boot, so no need 6527 * This is called single threaded during boot, so no need
6527 * to take the rtnl semaphore. 6528 * to take the rtnl semaphore.
6528 */ 6529 */
6529 static int __init net_dev_init(void) 6530 static int __init net_dev_init(void)
6530 { 6531 {
6531 int i, rc = -ENOMEM; 6532 int i, rc = -ENOMEM;
6532 6533
6533 BUG_ON(!dev_boot_phase); 6534 BUG_ON(!dev_boot_phase);
6534 6535
6535 if (dev_proc_init()) 6536 if (dev_proc_init())
6536 goto out; 6537 goto out;
6537 6538
6538 if (netdev_kobject_init()) 6539 if (netdev_kobject_init())
6539 goto out; 6540 goto out;
6540 6541
6541 INIT_LIST_HEAD(&ptype_all); 6542 INIT_LIST_HEAD(&ptype_all);
6542 for (i = 0; i < PTYPE_HASH_SIZE; i++) 6543 for (i = 0; i < PTYPE_HASH_SIZE; i++)
6543 INIT_LIST_HEAD(&ptype_base[i]); 6544 INIT_LIST_HEAD(&ptype_base[i]);
6544 6545
6545 if (register_pernet_subsys(&netdev_net_ops)) 6546 if (register_pernet_subsys(&netdev_net_ops))
6546 goto out; 6547 goto out;
6547 6548
6548 /* 6549 /*
6549 * Initialise the packet receive queues. 6550 * Initialise the packet receive queues.
6550 */ 6551 */
6551 6552
6552 for_each_possible_cpu(i) { 6553 for_each_possible_cpu(i) {
6553 struct softnet_data *sd = &per_cpu(softnet_data, i); 6554 struct softnet_data *sd = &per_cpu(softnet_data, i);
6554 6555
6555 memset(sd, 0, sizeof(*sd)); 6556 memset(sd, 0, sizeof(*sd));
6556 skb_queue_head_init(&sd->input_pkt_queue); 6557 skb_queue_head_init(&sd->input_pkt_queue);
6557 skb_queue_head_init(&sd->process_queue); 6558 skb_queue_head_init(&sd->process_queue);
6558 sd->completion_queue = NULL; 6559 sd->completion_queue = NULL;
6559 INIT_LIST_HEAD(&sd->poll_list); 6560 INIT_LIST_HEAD(&sd->poll_list);
6560 sd->output_queue = NULL; 6561 sd->output_queue = NULL;
6561 sd->output_queue_tailp = &sd->output_queue; 6562 sd->output_queue_tailp = &sd->output_queue;
6562 #ifdef CONFIG_RPS 6563 #ifdef CONFIG_RPS
6563 sd->csd.func = rps_trigger_softirq; 6564 sd->csd.func = rps_trigger_softirq;
6564 sd->csd.info = sd; 6565 sd->csd.info = sd;
6565 sd->csd.flags = 0; 6566 sd->csd.flags = 0;
6566 sd->cpu = i; 6567 sd->cpu = i;
6567 #endif 6568 #endif
6568 6569
6569 sd->backlog.poll = process_backlog; 6570 sd->backlog.poll = process_backlog;
6570 sd->backlog.weight = weight_p; 6571 sd->backlog.weight = weight_p;
6571 sd->backlog.gro_list = NULL; 6572 sd->backlog.gro_list = NULL;
6572 sd->backlog.gro_count = 0; 6573 sd->backlog.gro_count = 0;
6573 } 6574 }
6574 6575
6575 dev_boot_phase = 0; 6576 dev_boot_phase = 0;
6576 6577
6577 /* The loopback device is special if any other network devices 6578 /* The loopback device is special if any other network devices
6578 * is present in a network namespace the loopback device must 6579 * is present in a network namespace the loopback device must
6579 * be present. Since we now dynamically allocate and free the 6580 * be present. Since we now dynamically allocate and free the
6580 * loopback device ensure this invariant is maintained by 6581 * loopback device ensure this invariant is maintained by
6581 * keeping the loopback device as the first device on the 6582 * keeping the loopback device as the first device on the
6582 * list of network devices. Ensuring the loopback devices 6583 * list of network devices. Ensuring the loopback devices
6583 * is the first device that appears and the last network device 6584 * is the first device that appears and the last network device
6584 * that disappears. 6585 * that disappears.
6585 */ 6586 */
6586 if (register_pernet_device(&loopback_net_ops)) 6587 if (register_pernet_device(&loopback_net_ops))
6587 goto out; 6588 goto out;
6588 6589
6589 if (register_pernet_device(&default_device_ops)) 6590 if (register_pernet_device(&default_device_ops))
6590 goto out; 6591 goto out;
6591 6592
6592 open_softirq(NET_TX_SOFTIRQ, net_tx_action); 6593 open_softirq(NET_TX_SOFTIRQ, net_tx_action);
6593 open_softirq(NET_RX_SOFTIRQ, net_rx_action); 6594 open_softirq(NET_RX_SOFTIRQ, net_rx_action);
6594 6595
6595 hotcpu_notifier(dev_cpu_callback, 0); 6596 hotcpu_notifier(dev_cpu_callback, 0);
6596 dst_init(); 6597 dst_init();
6597 dev_mcast_init(); 6598 dev_mcast_init();
6598 rc = 0; 6599 rc = 0;
6599 out: 6600 out:
6600 return rc; 6601 return rc;
6601 } 6602 }
6602 6603
6603 subsys_initcall(net_dev_init); 6604 subsys_initcall(net_dev_init);
6604 6605
6605 static int __init initialize_hashrnd(void) 6606 static int __init initialize_hashrnd(void)
6606 { 6607 {
6607 get_random_bytes(&hashrnd, sizeof(hashrnd)); 6608 get_random_bytes(&hashrnd, sizeof(hashrnd));
6608 return 0; 6609 return 0;
6609 } 6610 }
6610 6611
6611 late_initcall_sync(initialize_hashrnd); 6612 late_initcall_sync(initialize_hashrnd);
6612 6613
6613 6614