Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Definitions for the 'struct sk_buff' memory handlers. |
| 3 | * |
| 4 | * Authors: |
| 5 | * Alan Cox, <gw4pts@gw4pts.ampr.org> |
| 6 | * Florian La Roche, <rzsfl@rz.uni-sb.de> |
| 7 | * |
| 8 | * This program is free software; you can redistribute it and/or |
| 9 | * modify it under the terms of the GNU General Public License |
| 10 | * as published by the Free Software Foundation; either version |
| 11 | * 2 of the License, or (at your option) any later version. |
| 12 | */ |
| 13 | |
| 14 | #ifndef _LINUX_SKBUFF_H |
| 15 | #define _LINUX_SKBUFF_H |
| 16 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | #include <linux/kernel.h> |
| 18 | #include <linux/compiler.h> |
| 19 | #include <linux/time.h> |
| 20 | #include <linux/cache.h> |
| 21 | |
| 22 | #include <asm/atomic.h> |
| 23 | #include <asm/types.h> |
| 24 | #include <linux/spinlock.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | #include <linux/net.h> |
Thomas Graf | 3fc7e8a | 2005-06-23 21:00:17 -0700 | [diff] [blame] | 26 | #include <linux/textsearch.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | #include <net/checksum.h> |
Al Viro | a80958f | 2006-12-04 20:41:19 +0000 | [diff] [blame] | 28 | #include <linux/rcupdate.h> |
Chris Leech | 97fc2f0 | 2006-05-23 17:55:33 -0700 | [diff] [blame] | 29 | #include <linux/dmaengine.h> |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 30 | #include <linux/hrtimer.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | |
| 32 | #define HAVE_ALLOC_SKB /* For the drivers to know */ |
| 33 | #define HAVE_ALIGNABLE_SKB /* Ditto 8) */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 34 | |
| 35 | #define CHECKSUM_NONE 0 |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 36 | #define CHECKSUM_PARTIAL 1 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 37 | #define CHECKSUM_UNNECESSARY 2 |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 38 | #define CHECKSUM_COMPLETE 3 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 39 | |
| 40 | #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \ |
| 41 | ~(SMP_CACHE_BYTES - 1)) |
David S. Miller | fc910a2 | 2007-03-25 20:27:59 -0700 | [diff] [blame] | 42 | #define SKB_WITH_OVERHEAD(X) \ |
| 43 | (((X) - sizeof(struct skb_shared_info)) & \ |
| 44 | ~(SMP_CACHE_BYTES - 1)) |
| 45 | #define SKB_MAX_ORDER(X, ORDER) \ |
| 46 | SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 47 | #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) |
| 48 | #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) |
| 49 | |
| 50 | /* A. Checksumming of received packets by device. |
| 51 | * |
| 52 | * NONE: device failed to checksum this packet. |
| 53 | * skb->csum is undefined. |
| 54 | * |
| 55 | * UNNECESSARY: device parsed packet and wouldbe verified checksum. |
| 56 | * skb->csum is undefined. |
| 57 | * It is bad option, but, unfortunately, many of vendors do this. |
| 58 | * Apparently with secret goal to sell you new device, when you |
| 59 | * will add new protocol to your host. F.e. IPv6. 8) |
| 60 | * |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 61 | * COMPLETE: the most generic way. Device supplied checksum of _all_ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 62 | * the packet as seen by netif_rx in skb->csum. |
| 63 | * NOTE: Even if device supports only some protocols, but |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 64 | * is able to produce some skb->csum, it MUST use COMPLETE, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | * not UNNECESSARY. |
| 66 | * |
| 67 | * B. Checksumming on output. |
| 68 | * |
| 69 | * NONE: skb is checksummed by protocol or csum is not required. |
| 70 | * |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 71 | * PARTIAL: device is required to csum packet as seen by hard_start_xmit |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | * from skb->h.raw to the end and to record the checksum |
| 73 | * at skb->h.raw+skb->csum. |
| 74 | * |
| 75 | * Device must show its capabilities in dev->features, set |
| 76 | * at device setup time. |
| 77 | * NETIF_F_HW_CSUM - it is clever device, it is able to checksum |
| 78 | * everything. |
| 79 | * NETIF_F_NO_CSUM - loopback or reliable single hop media. |
| 80 | * NETIF_F_IP_CSUM - device is dumb. It is able to csum only |
| 81 | * TCP/UDP over IPv4. Sigh. Vendors like this |
| 82 | * way by an unknown reason. Though, see comment above |
| 83 | * about CHECKSUM_UNNECESSARY. 8) |
| 84 | * |
| 85 | * Any questions? No questions, good. --ANK |
| 86 | */ |
| 87 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 88 | struct net_device; |
| 89 | |
| 90 | #ifdef CONFIG_NETFILTER |
| 91 | struct nf_conntrack { |
| 92 | atomic_t use; |
| 93 | void (*destroy)(struct nf_conntrack *); |
| 94 | }; |
| 95 | |
| 96 | #ifdef CONFIG_BRIDGE_NETFILTER |
| 97 | struct nf_bridge_info { |
| 98 | atomic_t use; |
| 99 | struct net_device *physindev; |
| 100 | struct net_device *physoutdev; |
| 101 | #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE) |
| 102 | struct net_device *netoutdev; |
| 103 | #endif |
| 104 | unsigned int mask; |
| 105 | unsigned long data[32 / sizeof(unsigned long)]; |
| 106 | }; |
| 107 | #endif |
| 108 | |
| 109 | #endif |
| 110 | |
| 111 | struct sk_buff_head { |
| 112 | /* These two members must be first. */ |
| 113 | struct sk_buff *next; |
| 114 | struct sk_buff *prev; |
| 115 | |
| 116 | __u32 qlen; |
| 117 | spinlock_t lock; |
| 118 | }; |
| 119 | |
| 120 | struct sk_buff; |
| 121 | |
| 122 | /* To allow 64K frame to be packed as single skb without frag_list */ |
| 123 | #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2) |
| 124 | |
| 125 | typedef struct skb_frag_struct skb_frag_t; |
| 126 | |
| 127 | struct skb_frag_struct { |
| 128 | struct page *page; |
| 129 | __u16 page_offset; |
| 130 | __u16 size; |
| 131 | }; |
| 132 | |
| 133 | /* This data is invariant across clones and lives at |
| 134 | * the end of the header data, ie. at skb->end. |
| 135 | */ |
| 136 | struct skb_shared_info { |
| 137 | atomic_t dataref; |
Benjamin LaHaise | 4947d3e | 2006-01-03 14:06:50 -0800 | [diff] [blame] | 138 | unsigned short nr_frags; |
Herbert Xu | 7967168 | 2006-06-22 02:40:14 -0700 | [diff] [blame] | 139 | unsigned short gso_size; |
| 140 | /* Warning: this field is not always filled in (UFO)! */ |
| 141 | unsigned short gso_segs; |
| 142 | unsigned short gso_type; |
Al Viro | ae08e1f | 2006-11-08 00:27:11 -0800 | [diff] [blame] | 143 | __be32 ip6_frag_id; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 144 | struct sk_buff *frag_list; |
| 145 | skb_frag_t frags[MAX_SKB_FRAGS]; |
| 146 | }; |
| 147 | |
| 148 | /* We divide dataref into two halves. The higher 16 bits hold references |
| 149 | * to the payload part of skb->data. The lower 16 bits hold references to |
| 150 | * the entire skb->data. It is up to the users of the skb to agree on |
| 151 | * where the payload starts. |
| 152 | * |
| 153 | * All users must obey the rule that the skb->data reference count must be |
| 154 | * greater than or equal to the payload reference count. |
| 155 | * |
| 156 | * Holding a reference to the payload part means that the user does not |
| 157 | * care about modifications to the header part of skb->data. |
| 158 | */ |
| 159 | #define SKB_DATAREF_SHIFT 16 |
| 160 | #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) |
| 161 | |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 162 | |
| 163 | enum { |
| 164 | SKB_FCLONE_UNAVAILABLE, |
| 165 | SKB_FCLONE_ORIG, |
| 166 | SKB_FCLONE_CLONE, |
| 167 | }; |
| 168 | |
Herbert Xu | 7967168 | 2006-06-22 02:40:14 -0700 | [diff] [blame] | 169 | enum { |
| 170 | SKB_GSO_TCPV4 = 1 << 0, |
Herbert Xu | f83ef8c | 2006-06-30 13:37:03 -0700 | [diff] [blame] | 171 | SKB_GSO_UDP = 1 << 1, |
Herbert Xu | 576a30e | 2006-06-27 13:22:38 -0700 | [diff] [blame] | 172 | |
| 173 | /* This indicates the skb is from an untrusted source. */ |
| 174 | SKB_GSO_DODGY = 1 << 2, |
Michael Chan | b0da8537 | 2006-06-29 12:30:00 -0700 | [diff] [blame] | 175 | |
| 176 | /* This indicates the tcp segment has CWR set. */ |
Herbert Xu | f83ef8c | 2006-06-30 13:37:03 -0700 | [diff] [blame] | 177 | SKB_GSO_TCP_ECN = 1 << 3, |
| 178 | |
| 179 | SKB_GSO_TCPV6 = 1 << 4, |
Herbert Xu | 7967168 | 2006-06-22 02:40:14 -0700 | [diff] [blame] | 180 | }; |
| 181 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | /** |
| 183 | * struct sk_buff - socket buffer |
| 184 | * @next: Next buffer in list |
| 185 | * @prev: Previous buffer in list |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 186 | * @sk: Socket we are owned by |
Herbert Xu | 325ed82 | 2005-10-03 13:57:23 -0700 | [diff] [blame] | 187 | * @tstamp: Time we arrived |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 188 | * @dev: Device we arrived on/are leaving by |
Patrick McHardy | c01003c | 2007-03-29 11:46:52 -0700 | [diff] [blame] | 189 | * @iif: ifindex of device we arrived on |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 190 | * @h: Transport layer header |
| 191 | * @nh: Network layer header |
| 192 | * @mac: Link layer header |
Martin Waitz | 67be2dd | 2005-05-01 08:59:26 -0700 | [diff] [blame] | 193 | * @dst: destination entry |
| 194 | * @sp: the security path, used for xfrm |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 195 | * @cb: Control buffer. Free for use by every layer. Put private vars here |
| 196 | * @len: Length of actual data |
| 197 | * @data_len: Data length |
| 198 | * @mac_len: Length of link layer header |
| 199 | * @csum: Checksum |
Martin Waitz | 67be2dd | 2005-05-01 08:59:26 -0700 | [diff] [blame] | 200 | * @local_df: allow local fragmentation |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 201 | * @cloned: Head may be cloned (check refcnt to be sure) |
| 202 | * @nohdr: Payload reference only, must not modify header |
| 203 | * @pkt_type: Packet class |
Randy Dunlap | c83c248 | 2005-10-18 22:07:41 -0700 | [diff] [blame] | 204 | * @fclone: skbuff clone status |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 205 | * @ip_summed: Driver fed us an IP checksum |
| 206 | * @priority: Packet queueing priority |
| 207 | * @users: User count - see {datagram,tcp}.c |
| 208 | * @protocol: Packet protocol from driver |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 209 | * @truesize: Buffer size |
| 210 | * @head: Head of buffer |
| 211 | * @data: Data head pointer |
| 212 | * @tail: Tail pointer |
| 213 | * @end: End pointer |
| 214 | * @destructor: Destruct function |
Thomas Graf | 82e91ff | 2006-11-09 15:19:14 -0800 | [diff] [blame] | 215 | * @mark: Generic packet mark |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 216 | * @nfct: Associated connection, if any |
Randy Dunlap | c83c248 | 2005-10-18 22:07:41 -0700 | [diff] [blame] | 217 | * @ipvs_property: skbuff is owned by ipvs |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 218 | * @nfctinfo: Relationship of this skb to the connection |
Randy Dunlap | 461ddf3 | 2005-11-20 21:25:15 -0800 | [diff] [blame] | 219 | * @nfct_reasm: netfilter conntrack re-assembly pointer |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 220 | * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 221 | * @tc_index: Traffic control index |
| 222 | * @tc_verd: traffic control verdict |
Randy Dunlap | f4b8ea7 | 2006-06-22 16:00:11 -0700 | [diff] [blame] | 223 | * @dma_cookie: a cookie to one of several possible DMA operations |
| 224 | * done by skb DMA functions |
James Morris | 984bc16 | 2006-06-09 00:29:17 -0700 | [diff] [blame] | 225 | * @secmark: security marking |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 226 | */ |
| 227 | |
| 228 | struct sk_buff { |
| 229 | /* These two members must be first. */ |
| 230 | struct sk_buff *next; |
| 231 | struct sk_buff *prev; |
| 232 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 233 | struct sock *sk; |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 234 | ktime_t tstamp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 235 | struct net_device *dev; |
Patrick McHardy | c01003c | 2007-03-29 11:46:52 -0700 | [diff] [blame] | 236 | int iif; |
| 237 | /* 4 byte hole on 64 bit*/ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 238 | |
| 239 | union { |
| 240 | struct tcphdr *th; |
| 241 | struct udphdr *uh; |
| 242 | struct icmphdr *icmph; |
| 243 | struct igmphdr *igmph; |
| 244 | struct iphdr *ipiph; |
| 245 | struct ipv6hdr *ipv6h; |
| 246 | unsigned char *raw; |
| 247 | } h; |
| 248 | |
| 249 | union { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 250 | unsigned char *raw; |
| 251 | } nh; |
| 252 | |
| 253 | union { |
| 254 | unsigned char *raw; |
| 255 | } mac; |
| 256 | |
| 257 | struct dst_entry *dst; |
| 258 | struct sec_path *sp; |
| 259 | |
| 260 | /* |
| 261 | * This is the control buffer. It is free to use for every |
| 262 | * layer. Please put your private variables there. If you |
| 263 | * want to keep them across layers you have to do a skb_clone() |
| 264 | * first. This is owned by whoever has the skb queued ATM. |
| 265 | */ |
Patrick McHardy | 3e3850e | 2006-01-06 23:04:54 -0800 | [diff] [blame] | 266 | char cb[48]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 267 | |
| 268 | unsigned int len, |
| 269 | data_len, |
Al Viro | 1f61ab5 | 2006-11-14 21:44:08 -0800 | [diff] [blame] | 270 | mac_len; |
Al Viro | ff1dcad | 2006-11-20 18:07:29 -0800 | [diff] [blame] | 271 | union { |
| 272 | __wsum csum; |
| 273 | __u32 csum_offset; |
| 274 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | __u32 priority; |
Thomas Graf | 1cbb338 | 2005-07-05 14:13:41 -0700 | [diff] [blame] | 276 | __u8 local_df:1, |
| 277 | cloned:1, |
| 278 | ip_summed:2, |
Harald Welte | 6869c4d | 2005-08-09 19:24:19 -0700 | [diff] [blame] | 279 | nohdr:1, |
| 280 | nfctinfo:3; |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 281 | __u8 pkt_type:3, |
Patrick McHardy | b84f4cc | 2005-11-20 21:19:21 -0800 | [diff] [blame] | 282 | fclone:2, |
| 283 | ipvs_property:1; |
Alexey Dobriyan | a0d3bea | 2005-08-11 16:05:50 -0700 | [diff] [blame] | 284 | __be16 protocol; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 285 | |
| 286 | void (*destructor)(struct sk_buff *skb); |
| 287 | #ifdef CONFIG_NETFILTER |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 288 | struct nf_conntrack *nfct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 289 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
| 290 | struct sk_buff *nfct_reasm; |
| 291 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 292 | #ifdef CONFIG_BRIDGE_NETFILTER |
| 293 | struct nf_bridge_info *nf_bridge; |
| 294 | #endif |
| 295 | #endif /* CONFIG_NETFILTER */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 296 | #ifdef CONFIG_NET_SCHED |
Patrick McHardy | b6b99eb | 2005-08-09 19:33:51 -0700 | [diff] [blame] | 297 | __u16 tc_index; /* traffic control index */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 298 | #ifdef CONFIG_NET_CLS_ACT |
Patrick McHardy | b6b99eb | 2005-08-09 19:33:51 -0700 | [diff] [blame] | 299 | __u16 tc_verd; /* traffic control verdict */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 300 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 301 | #endif |
Chris Leech | 97fc2f0 | 2006-05-23 17:55:33 -0700 | [diff] [blame] | 302 | #ifdef CONFIG_NET_DMA |
| 303 | dma_cookie_t dma_cookie; |
| 304 | #endif |
James Morris | 984bc16 | 2006-06-09 00:29:17 -0700 | [diff] [blame] | 305 | #ifdef CONFIG_NETWORK_SECMARK |
| 306 | __u32 secmark; |
| 307 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 308 | |
Thomas Graf | 82e91ff | 2006-11-09 15:19:14 -0800 | [diff] [blame] | 309 | __u32 mark; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 310 | |
| 311 | /* These elements must be at the end, see alloc_skb() for details. */ |
| 312 | unsigned int truesize; |
| 313 | atomic_t users; |
| 314 | unsigned char *head, |
| 315 | *data, |
| 316 | *tail, |
| 317 | *end; |
| 318 | }; |
| 319 | |
| 320 | #ifdef __KERNEL__ |
| 321 | /* |
| 322 | * Handling routines are only of interest to the kernel |
| 323 | */ |
| 324 | #include <linux/slab.h> |
| 325 | |
| 326 | #include <asm/system.h> |
| 327 | |
Jörn Engel | 231d06a | 2006-03-20 21:28:35 -0800 | [diff] [blame] | 328 | extern void kfree_skb(struct sk_buff *skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 329 | extern void __kfree_skb(struct sk_buff *skb); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 330 | extern struct sk_buff *__alloc_skb(unsigned int size, |
Christoph Hellwig | b30973f | 2006-12-06 20:32:36 -0800 | [diff] [blame] | 331 | gfp_t priority, int fclone, int node); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 332 | static inline struct sk_buff *alloc_skb(unsigned int size, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 333 | gfp_t priority) |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 334 | { |
Christoph Hellwig | b30973f | 2006-12-06 20:32:36 -0800 | [diff] [blame] | 335 | return __alloc_skb(size, priority, 0, -1); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 336 | } |
| 337 | |
| 338 | static inline struct sk_buff *alloc_skb_fclone(unsigned int size, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 339 | gfp_t priority) |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 340 | { |
Christoph Hellwig | b30973f | 2006-12-06 20:32:36 -0800 | [diff] [blame] | 341 | return __alloc_skb(size, priority, 1, -1); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 342 | } |
| 343 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 344 | extern void kfree_skbmem(struct sk_buff *skb); |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 345 | extern struct sk_buff *skb_clone(struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 346 | gfp_t priority); |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 347 | extern struct sk_buff *skb_copy(const struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 348 | gfp_t priority); |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 349 | extern struct sk_buff *pskb_copy(struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 350 | gfp_t gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 351 | extern int pskb_expand_head(struct sk_buff *skb, |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 352 | int nhead, int ntail, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 353 | gfp_t gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 354 | extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, |
| 355 | unsigned int headroom); |
| 356 | extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb, |
| 357 | int newheadroom, int newtailroom, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 358 | gfp_t priority); |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 359 | extern int skb_pad(struct sk_buff *skb, int pad); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 360 | #define dev_kfree_skb(a) kfree_skb(a) |
| 361 | extern void skb_over_panic(struct sk_buff *skb, int len, |
| 362 | void *here); |
| 363 | extern void skb_under_panic(struct sk_buff *skb, int len, |
| 364 | void *here); |
David S. Miller | dc6de33 | 2006-04-20 00:10:50 -0700 | [diff] [blame] | 365 | extern void skb_truesize_bug(struct sk_buff *skb); |
| 366 | |
| 367 | static inline void skb_truesize_check(struct sk_buff *skb) |
| 368 | { |
| 369 | if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len)) |
| 370 | skb_truesize_bug(skb); |
| 371 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 372 | |
Ananda Raju | e89e9cf | 2005-10-18 15:46:41 -0700 | [diff] [blame] | 373 | extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, |
| 374 | int getfrag(void *from, char *to, int offset, |
| 375 | int len,int odd, struct sk_buff *skb), |
| 376 | void *from, int length); |
| 377 | |
Thomas Graf | 677e90e | 2005-06-23 20:59:51 -0700 | [diff] [blame] | 378 | struct skb_seq_state |
| 379 | { |
| 380 | __u32 lower_offset; |
| 381 | __u32 upper_offset; |
| 382 | __u32 frag_idx; |
| 383 | __u32 stepped_offset; |
| 384 | struct sk_buff *root_skb; |
| 385 | struct sk_buff *cur_skb; |
| 386 | __u8 *frag_data; |
| 387 | }; |
| 388 | |
| 389 | extern void skb_prepare_seq_read(struct sk_buff *skb, |
| 390 | unsigned int from, unsigned int to, |
| 391 | struct skb_seq_state *st); |
| 392 | extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data, |
| 393 | struct skb_seq_state *st); |
| 394 | extern void skb_abort_seq_read(struct skb_seq_state *st); |
| 395 | |
Thomas Graf | 3fc7e8a | 2005-06-23 21:00:17 -0700 | [diff] [blame] | 396 | extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, |
| 397 | unsigned int to, struct ts_config *config, |
| 398 | struct ts_state *state); |
| 399 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 400 | /* Internal */ |
| 401 | #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end)) |
| 402 | |
| 403 | /** |
| 404 | * skb_queue_empty - check if a queue is empty |
| 405 | * @list: queue head |
| 406 | * |
| 407 | * Returns true if the queue is empty, false otherwise. |
| 408 | */ |
| 409 | static inline int skb_queue_empty(const struct sk_buff_head *list) |
| 410 | { |
| 411 | return list->next == (struct sk_buff *)list; |
| 412 | } |
| 413 | |
| 414 | /** |
| 415 | * skb_get - reference buffer |
| 416 | * @skb: buffer to reference |
| 417 | * |
| 418 | * Makes another reference to a socket buffer and returns a pointer |
| 419 | * to the buffer. |
| 420 | */ |
| 421 | static inline struct sk_buff *skb_get(struct sk_buff *skb) |
| 422 | { |
| 423 | atomic_inc(&skb->users); |
| 424 | return skb; |
| 425 | } |
| 426 | |
| 427 | /* |
| 428 | * If users == 1, we are the only owner and are can avoid redundant |
| 429 | * atomic change. |
| 430 | */ |
| 431 | |
| 432 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 433 | * skb_cloned - is the buffer a clone |
| 434 | * @skb: buffer to check |
| 435 | * |
| 436 | * Returns true if the buffer was generated with skb_clone() and is |
| 437 | * one of multiple shared copies of the buffer. Cloned buffers are |
| 438 | * shared data so must not be written to under normal circumstances. |
| 439 | */ |
| 440 | static inline int skb_cloned(const struct sk_buff *skb) |
| 441 | { |
| 442 | return skb->cloned && |
| 443 | (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; |
| 444 | } |
| 445 | |
| 446 | /** |
| 447 | * skb_header_cloned - is the header a clone |
| 448 | * @skb: buffer to check |
| 449 | * |
| 450 | * Returns true if modifying the header part of the buffer requires |
| 451 | * the data to be copied. |
| 452 | */ |
| 453 | static inline int skb_header_cloned(const struct sk_buff *skb) |
| 454 | { |
| 455 | int dataref; |
| 456 | |
| 457 | if (!skb->cloned) |
| 458 | return 0; |
| 459 | |
| 460 | dataref = atomic_read(&skb_shinfo(skb)->dataref); |
| 461 | dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); |
| 462 | return dataref != 1; |
| 463 | } |
| 464 | |
| 465 | /** |
| 466 | * skb_header_release - release reference to header |
| 467 | * @skb: buffer to operate on |
| 468 | * |
| 469 | * Drop a reference to the header part of the buffer. This is done |
| 470 | * by acquiring a payload reference. You must not read from the header |
| 471 | * part of skb->data after this. |
| 472 | */ |
| 473 | static inline void skb_header_release(struct sk_buff *skb) |
| 474 | { |
| 475 | BUG_ON(skb->nohdr); |
| 476 | skb->nohdr = 1; |
| 477 | atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); |
| 478 | } |
| 479 | |
| 480 | /** |
| 481 | * skb_shared - is the buffer shared |
| 482 | * @skb: buffer to check |
| 483 | * |
| 484 | * Returns true if more than one person has a reference to this |
| 485 | * buffer. |
| 486 | */ |
| 487 | static inline int skb_shared(const struct sk_buff *skb) |
| 488 | { |
| 489 | return atomic_read(&skb->users) != 1; |
| 490 | } |
| 491 | |
| 492 | /** |
| 493 | * skb_share_check - check if buffer is shared and if so clone it |
| 494 | * @skb: buffer to check |
| 495 | * @pri: priority for memory allocation |
| 496 | * |
| 497 | * If the buffer is shared the buffer is cloned and the old copy |
| 498 | * drops a reference. A new clone with a single reference is returned. |
| 499 | * If the buffer is not shared the original buffer is returned. When |
| 500 | * being called from interrupt status or with spinlocks held pri must |
| 501 | * be GFP_ATOMIC. |
| 502 | * |
| 503 | * NULL is returned on a memory allocation failure. |
| 504 | */ |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 505 | static inline struct sk_buff *skb_share_check(struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 506 | gfp_t pri) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 507 | { |
| 508 | might_sleep_if(pri & __GFP_WAIT); |
| 509 | if (skb_shared(skb)) { |
| 510 | struct sk_buff *nskb = skb_clone(skb, pri); |
| 511 | kfree_skb(skb); |
| 512 | skb = nskb; |
| 513 | } |
| 514 | return skb; |
| 515 | } |
| 516 | |
| 517 | /* |
| 518 | * Copy shared buffers into a new sk_buff. We effectively do COW on |
| 519 | * packets to handle cases where we have a local reader and forward |
| 520 | * and a couple of other messy ones. The normal one is tcpdumping |
| 521 | * a packet thats being forwarded. |
| 522 | */ |
| 523 | |
| 524 | /** |
| 525 | * skb_unshare - make a copy of a shared buffer |
| 526 | * @skb: buffer to check |
| 527 | * @pri: priority for memory allocation |
| 528 | * |
| 529 | * If the socket buffer is a clone then this function creates a new |
| 530 | * copy of the data, drops a reference count on the old copy and returns |
| 531 | * the new copy with the reference count at 1. If the buffer is not a clone |
| 532 | * the original buffer is returned. When called with a spinlock held or |
| 533 | * from interrupt state @pri must be %GFP_ATOMIC |
| 534 | * |
| 535 | * %NULL is returned on a memory allocation failure. |
| 536 | */ |
Victor Fusco | e2bf521 | 2005-07-18 13:36:38 -0700 | [diff] [blame] | 537 | static inline struct sk_buff *skb_unshare(struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 538 | gfp_t pri) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 539 | { |
| 540 | might_sleep_if(pri & __GFP_WAIT); |
| 541 | if (skb_cloned(skb)) { |
| 542 | struct sk_buff *nskb = skb_copy(skb, pri); |
| 543 | kfree_skb(skb); /* Free our shared copy */ |
| 544 | skb = nskb; |
| 545 | } |
| 546 | return skb; |
| 547 | } |
| 548 | |
| 549 | /** |
| 550 | * skb_peek |
| 551 | * @list_: list to peek at |
| 552 | * |
| 553 | * Peek an &sk_buff. Unlike most other operations you _MUST_ |
| 554 | * be careful with this one. A peek leaves the buffer on the |
| 555 | * list and someone else may run off with it. You must hold |
| 556 | * the appropriate locks or have a private queue to do this. |
| 557 | * |
| 558 | * Returns %NULL for an empty list or a pointer to the head element. |
| 559 | * The reference count is not incremented and the reference is therefore |
| 560 | * volatile. Use with caution. |
| 561 | */ |
| 562 | static inline struct sk_buff *skb_peek(struct sk_buff_head *list_) |
| 563 | { |
| 564 | struct sk_buff *list = ((struct sk_buff *)list_)->next; |
| 565 | if (list == (struct sk_buff *)list_) |
| 566 | list = NULL; |
| 567 | return list; |
| 568 | } |
| 569 | |
| 570 | /** |
| 571 | * skb_peek_tail |
| 572 | * @list_: list to peek at |
| 573 | * |
| 574 | * Peek an &sk_buff. Unlike most other operations you _MUST_ |
| 575 | * be careful with this one. A peek leaves the buffer on the |
| 576 | * list and someone else may run off with it. You must hold |
| 577 | * the appropriate locks or have a private queue to do this. |
| 578 | * |
| 579 | * Returns %NULL for an empty list or a pointer to the tail element. |
| 580 | * The reference count is not incremented and the reference is therefore |
| 581 | * volatile. Use with caution. |
| 582 | */ |
| 583 | static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_) |
| 584 | { |
| 585 | struct sk_buff *list = ((struct sk_buff *)list_)->prev; |
| 586 | if (list == (struct sk_buff *)list_) |
| 587 | list = NULL; |
| 588 | return list; |
| 589 | } |
| 590 | |
| 591 | /** |
| 592 | * skb_queue_len - get queue length |
| 593 | * @list_: list to measure |
| 594 | * |
| 595 | * Return the length of an &sk_buff queue. |
| 596 | */ |
| 597 | static inline __u32 skb_queue_len(const struct sk_buff_head *list_) |
| 598 | { |
| 599 | return list_->qlen; |
| 600 | } |
| 601 | |
Arjan van de Ven | 76f10ad | 2006-08-02 14:06:55 -0700 | [diff] [blame] | 602 | /* |
| 603 | * This function creates a split out lock class for each invocation; |
| 604 | * this is needed for now since a whole lot of users of the skb-queue |
| 605 | * infrastructure in drivers have different locking usage (in hardirq) |
| 606 | * than the networking core (in softirq only). In the long run either the |
| 607 | * network layer or drivers should need annotation to consolidate the |
| 608 | * main types of usage into 3 classes. |
| 609 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 610 | static inline void skb_queue_head_init(struct sk_buff_head *list) |
| 611 | { |
| 612 | spin_lock_init(&list->lock); |
| 613 | list->prev = list->next = (struct sk_buff *)list; |
| 614 | list->qlen = 0; |
| 615 | } |
| 616 | |
Pavel Emelianov | c2ecba7 | 2007-04-17 12:45:31 -0700 | [diff] [blame] | 617 | static inline void skb_queue_head_init_class(struct sk_buff_head *list, |
| 618 | struct lock_class_key *class) |
| 619 | { |
| 620 | skb_queue_head_init(list); |
| 621 | lockdep_set_class(&list->lock, class); |
| 622 | } |
| 623 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 624 | /* |
| 625 | * Insert an sk_buff at the start of a list. |
| 626 | * |
| 627 | * The "__skb_xxxx()" functions are the non-atomic ones that |
| 628 | * can only be called with interrupts disabled. |
| 629 | */ |
| 630 | |
| 631 | /** |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 632 | * __skb_queue_after - queue a buffer at the list head |
| 633 | * @list: list to use |
| 634 | * @prev: place after this buffer |
| 635 | * @newsk: buffer to queue |
| 636 | * |
| 637 | * Queue a buffer int the middle of a list. This function takes no locks |
| 638 | * and you must therefore hold required locks before calling it. |
| 639 | * |
| 640 | * A buffer cannot be placed on two lists at the same time. |
| 641 | */ |
| 642 | static inline void __skb_queue_after(struct sk_buff_head *list, |
| 643 | struct sk_buff *prev, |
| 644 | struct sk_buff *newsk) |
| 645 | { |
| 646 | struct sk_buff *next; |
| 647 | list->qlen++; |
| 648 | |
| 649 | next = prev->next; |
| 650 | newsk->next = next; |
| 651 | newsk->prev = prev; |
| 652 | next->prev = prev->next = newsk; |
| 653 | } |
| 654 | |
| 655 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 656 | * __skb_queue_head - queue a buffer at the list head |
| 657 | * @list: list to use |
| 658 | * @newsk: buffer to queue |
| 659 | * |
| 660 | * Queue a buffer at the start of a list. This function takes no locks |
| 661 | * and you must therefore hold required locks before calling it. |
| 662 | * |
| 663 | * A buffer cannot be placed on two lists at the same time. |
| 664 | */ |
| 665 | extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); |
| 666 | static inline void __skb_queue_head(struct sk_buff_head *list, |
| 667 | struct sk_buff *newsk) |
| 668 | { |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 669 | __skb_queue_after(list, (struct sk_buff *)list, newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 670 | } |
| 671 | |
| 672 | /** |
| 673 | * __skb_queue_tail - queue a buffer at the list tail |
| 674 | * @list: list to use |
| 675 | * @newsk: buffer to queue |
| 676 | * |
| 677 | * Queue a buffer at the end of a list. This function takes no locks |
| 678 | * and you must therefore hold required locks before calling it. |
| 679 | * |
| 680 | * A buffer cannot be placed on two lists at the same time. |
| 681 | */ |
| 682 | extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); |
| 683 | static inline void __skb_queue_tail(struct sk_buff_head *list, |
| 684 | struct sk_buff *newsk) |
| 685 | { |
| 686 | struct sk_buff *prev, *next; |
| 687 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 688 | list->qlen++; |
| 689 | next = (struct sk_buff *)list; |
| 690 | prev = next->prev; |
| 691 | newsk->next = next; |
| 692 | newsk->prev = prev; |
| 693 | next->prev = prev->next = newsk; |
| 694 | } |
| 695 | |
| 696 | |
| 697 | /** |
| 698 | * __skb_dequeue - remove from the head of the queue |
| 699 | * @list: list to dequeue from |
| 700 | * |
| 701 | * Remove the head of the list. This function does not take any locks |
| 702 | * so must be used with appropriate locks held only. The head item is |
| 703 | * returned or %NULL if the list is empty. |
| 704 | */ |
| 705 | extern struct sk_buff *skb_dequeue(struct sk_buff_head *list); |
| 706 | static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) |
| 707 | { |
| 708 | struct sk_buff *next, *prev, *result; |
| 709 | |
| 710 | prev = (struct sk_buff *) list; |
| 711 | next = prev->next; |
| 712 | result = NULL; |
| 713 | if (next != prev) { |
| 714 | result = next; |
| 715 | next = next->next; |
| 716 | list->qlen--; |
| 717 | next->prev = prev; |
| 718 | prev->next = next; |
| 719 | result->next = result->prev = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 720 | } |
| 721 | return result; |
| 722 | } |
| 723 | |
| 724 | |
| 725 | /* |
| 726 | * Insert a packet on a list. |
| 727 | */ |
David S. Miller | 8728b83 | 2005-08-09 19:25:21 -0700 | [diff] [blame] | 728 | extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 729 | static inline void __skb_insert(struct sk_buff *newsk, |
| 730 | struct sk_buff *prev, struct sk_buff *next, |
| 731 | struct sk_buff_head *list) |
| 732 | { |
| 733 | newsk->next = next; |
| 734 | newsk->prev = prev; |
| 735 | next->prev = prev->next = newsk; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 736 | list->qlen++; |
| 737 | } |
| 738 | |
| 739 | /* |
| 740 | * Place a packet after a given packet in a list. |
| 741 | */ |
David S. Miller | 8728b83 | 2005-08-09 19:25:21 -0700 | [diff] [blame] | 742 | extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list); |
| 743 | static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 744 | { |
David S. Miller | 8728b83 | 2005-08-09 19:25:21 -0700 | [diff] [blame] | 745 | __skb_insert(newsk, old, old->next, list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 746 | } |
| 747 | |
| 748 | /* |
| 749 | * remove sk_buff from list. _Must_ be called atomically, and with |
| 750 | * the list known.. |
| 751 | */ |
David S. Miller | 8728b83 | 2005-08-09 19:25:21 -0700 | [diff] [blame] | 752 | extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 753 | static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) |
| 754 | { |
| 755 | struct sk_buff *next, *prev; |
| 756 | |
| 757 | list->qlen--; |
| 758 | next = skb->next; |
| 759 | prev = skb->prev; |
| 760 | skb->next = skb->prev = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 761 | next->prev = prev; |
| 762 | prev->next = next; |
| 763 | } |
| 764 | |
| 765 | |
| 766 | /* XXX: more streamlined implementation */ |
| 767 | |
| 768 | /** |
| 769 | * __skb_dequeue_tail - remove from the tail of the queue |
| 770 | * @list: list to dequeue from |
| 771 | * |
| 772 | * Remove the tail of the list. This function does not take any locks |
| 773 | * so must be used with appropriate locks held only. The tail item is |
| 774 | * returned or %NULL if the list is empty. |
| 775 | */ |
| 776 | extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); |
| 777 | static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) |
| 778 | { |
| 779 | struct sk_buff *skb = skb_peek_tail(list); |
| 780 | if (skb) |
| 781 | __skb_unlink(skb, list); |
| 782 | return skb; |
| 783 | } |
| 784 | |
| 785 | |
| 786 | static inline int skb_is_nonlinear(const struct sk_buff *skb) |
| 787 | { |
| 788 | return skb->data_len; |
| 789 | } |
| 790 | |
| 791 | static inline unsigned int skb_headlen(const struct sk_buff *skb) |
| 792 | { |
| 793 | return skb->len - skb->data_len; |
| 794 | } |
| 795 | |
| 796 | static inline int skb_pagelen(const struct sk_buff *skb) |
| 797 | { |
| 798 | int i, len = 0; |
| 799 | |
| 800 | for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) |
| 801 | len += skb_shinfo(skb)->frags[i].size; |
| 802 | return len + skb_headlen(skb); |
| 803 | } |
| 804 | |
| 805 | static inline void skb_fill_page_desc(struct sk_buff *skb, int i, |
| 806 | struct page *page, int off, int size) |
| 807 | { |
| 808 | skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; |
| 809 | |
| 810 | frag->page = page; |
| 811 | frag->page_offset = off; |
| 812 | frag->size = size; |
| 813 | skb_shinfo(skb)->nr_frags = i + 1; |
| 814 | } |
| 815 | |
| 816 | #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) |
| 817 | #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list) |
| 818 | #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) |
| 819 | |
| 820 | /* |
| 821 | * Add data to an sk_buff |
| 822 | */ |
| 823 | static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) |
| 824 | { |
| 825 | unsigned char *tmp = skb->tail; |
| 826 | SKB_LINEAR_ASSERT(skb); |
| 827 | skb->tail += len; |
| 828 | skb->len += len; |
| 829 | return tmp; |
| 830 | } |
| 831 | |
| 832 | /** |
| 833 | * skb_put - add data to a buffer |
| 834 | * @skb: buffer to use |
| 835 | * @len: amount of data to add |
| 836 | * |
| 837 | * This function extends the used data area of the buffer. If this would |
| 838 | * exceed the total buffer size the kernel will panic. A pointer to the |
| 839 | * first byte of the extra data is returned. |
| 840 | */ |
| 841 | static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len) |
| 842 | { |
| 843 | unsigned char *tmp = skb->tail; |
| 844 | SKB_LINEAR_ASSERT(skb); |
| 845 | skb->tail += len; |
| 846 | skb->len += len; |
| 847 | if (unlikely(skb->tail>skb->end)) |
| 848 | skb_over_panic(skb, len, current_text_addr()); |
| 849 | return tmp; |
| 850 | } |
| 851 | |
| 852 | static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) |
| 853 | { |
| 854 | skb->data -= len; |
| 855 | skb->len += len; |
| 856 | return skb->data; |
| 857 | } |
| 858 | |
| 859 | /** |
| 860 | * skb_push - add data to the start of a buffer |
| 861 | * @skb: buffer to use |
| 862 | * @len: amount of data to add |
| 863 | * |
| 864 | * This function extends the used data area of the buffer at the buffer |
| 865 | * start. If this would exceed the total buffer headroom the kernel will |
| 866 | * panic. A pointer to the first byte of the extra data is returned. |
| 867 | */ |
| 868 | static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len) |
| 869 | { |
| 870 | skb->data -= len; |
| 871 | skb->len += len; |
| 872 | if (unlikely(skb->data<skb->head)) |
| 873 | skb_under_panic(skb, len, current_text_addr()); |
| 874 | return skb->data; |
| 875 | } |
| 876 | |
| 877 | static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) |
| 878 | { |
| 879 | skb->len -= len; |
| 880 | BUG_ON(skb->len < skb->data_len); |
| 881 | return skb->data += len; |
| 882 | } |
| 883 | |
| 884 | /** |
| 885 | * skb_pull - remove data from the start of a buffer |
| 886 | * @skb: buffer to use |
| 887 | * @len: amount of data to remove |
| 888 | * |
| 889 | * This function removes data from the start of a buffer, returning |
| 890 | * the memory to the headroom. A pointer to the next data in the buffer |
| 891 | * is returned. Once the data has been pulled future pushes will overwrite |
| 892 | * the old data. |
| 893 | */ |
| 894 | static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len) |
| 895 | { |
| 896 | return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); |
| 897 | } |
| 898 | |
| 899 | extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); |
| 900 | |
| 901 | static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) |
| 902 | { |
| 903 | if (len > skb_headlen(skb) && |
| 904 | !__pskb_pull_tail(skb, len-skb_headlen(skb))) |
| 905 | return NULL; |
| 906 | skb->len -= len; |
| 907 | return skb->data += len; |
| 908 | } |
| 909 | |
| 910 | static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) |
| 911 | { |
| 912 | return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); |
| 913 | } |
| 914 | |
| 915 | static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) |
| 916 | { |
| 917 | if (likely(len <= skb_headlen(skb))) |
| 918 | return 1; |
| 919 | if (unlikely(len > skb->len)) |
| 920 | return 0; |
| 921 | return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL; |
| 922 | } |
| 923 | |
| 924 | /** |
| 925 | * skb_headroom - bytes at buffer head |
| 926 | * @skb: buffer to check |
| 927 | * |
| 928 | * Return the number of bytes of free space at the head of an &sk_buff. |
| 929 | */ |
| 930 | static inline int skb_headroom(const struct sk_buff *skb) |
| 931 | { |
| 932 | return skb->data - skb->head; |
| 933 | } |
| 934 | |
| 935 | /** |
| 936 | * skb_tailroom - bytes at buffer end |
| 937 | * @skb: buffer to check |
| 938 | * |
| 939 | * Return the number of bytes of free space at the tail of an sk_buff |
| 940 | */ |
| 941 | static inline int skb_tailroom(const struct sk_buff *skb) |
| 942 | { |
| 943 | return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; |
| 944 | } |
| 945 | |
| 946 | /** |
| 947 | * skb_reserve - adjust headroom |
| 948 | * @skb: buffer to alter |
| 949 | * @len: bytes to move |
| 950 | * |
| 951 | * Increase the headroom of an empty &sk_buff by reducing the tail |
| 952 | * room. This is only allowed for an empty buffer. |
| 953 | */ |
David S. Miller | 8243126c | 2006-01-17 02:54:21 -0800 | [diff] [blame] | 954 | static inline void skb_reserve(struct sk_buff *skb, int len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 955 | { |
| 956 | skb->data += len; |
| 957 | skb->tail += len; |
| 958 | } |
| 959 | |
Arnaldo Carvalho de Melo | badff6d | 2007-03-13 13:06:52 -0300 | [diff] [blame] | 960 | static inline void skb_reset_transport_header(struct sk_buff *skb) |
| 961 | { |
| 962 | skb->h.raw = skb->data; |
| 963 | } |
| 964 | |
Arnaldo Carvalho de Melo | ea2ae17 | 2007-04-25 17:55:53 -0700 | [diff] [blame^] | 965 | static inline int skb_transport_offset(const struct sk_buff *skb) |
| 966 | { |
| 967 | return skb->h.raw - skb->data; |
| 968 | } |
| 969 | |
Arnaldo Carvalho de Melo | d56f90a | 2007-04-10 20:50:43 -0700 | [diff] [blame] | 970 | static inline unsigned char *skb_network_header(const struct sk_buff *skb) |
| 971 | { |
| 972 | return skb->nh.raw; |
| 973 | } |
| 974 | |
Arnaldo Carvalho de Melo | c1d2bbe | 2007-04-10 20:45:18 -0700 | [diff] [blame] | 975 | static inline void skb_reset_network_header(struct sk_buff *skb) |
| 976 | { |
| 977 | skb->nh.raw = skb->data; |
| 978 | } |
| 979 | |
Arnaldo Carvalho de Melo | c14d245 | 2007-03-11 22:39:41 -0300 | [diff] [blame] | 980 | static inline void skb_set_network_header(struct sk_buff *skb, const int offset) |
| 981 | { |
| 982 | skb->nh.raw = skb->data + offset; |
| 983 | } |
| 984 | |
Arnaldo Carvalho de Melo | bbe735e | 2007-03-10 22:16:10 -0300 | [diff] [blame] | 985 | static inline int skb_network_offset(const struct sk_buff *skb) |
| 986 | { |
| 987 | return skb->nh.raw - skb->data; |
| 988 | } |
| 989 | |
Arnaldo Carvalho de Melo | 98e399f | 2007-03-19 15:33:04 -0700 | [diff] [blame] | 990 | static inline unsigned char *skb_mac_header(const struct sk_buff *skb) |
| 991 | { |
| 992 | return skb->mac.raw; |
| 993 | } |
| 994 | |
| 995 | static inline int skb_mac_header_was_set(const struct sk_buff *skb) |
| 996 | { |
| 997 | return skb->mac.raw != NULL; |
| 998 | } |
| 999 | |
Arnaldo Carvalho de Melo | 459a98e | 2007-03-19 15:30:44 -0700 | [diff] [blame] | 1000 | static inline void skb_reset_mac_header(struct sk_buff *skb) |
| 1001 | { |
| 1002 | skb->mac.raw = skb->data; |
| 1003 | } |
| 1004 | |
Arnaldo Carvalho de Melo | 48d49d0c | 2007-03-10 12:30:58 -0300 | [diff] [blame] | 1005 | static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) |
| 1006 | { |
| 1007 | skb->mac.raw = skb->data + offset; |
| 1008 | } |
| 1009 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1010 | /* |
| 1011 | * CPUs often take a performance hit when accessing unaligned memory |
| 1012 | * locations. The actual performance hit varies, it can be small if the |
| 1013 | * hardware handles it or large if we have to take an exception and fix it |
| 1014 | * in software. |
| 1015 | * |
| 1016 | * Since an ethernet header is 14 bytes network drivers often end up with |
| 1017 | * the IP header at an unaligned offset. The IP header can be aligned by |
| 1018 | * shifting the start of the packet by 2 bytes. Drivers should do this |
| 1019 | * with: |
| 1020 | * |
| 1021 | * skb_reserve(NET_IP_ALIGN); |
| 1022 | * |
| 1023 | * The downside to this alignment of the IP header is that the DMA is now |
| 1024 | * unaligned. On some architectures the cost of an unaligned DMA is high |
| 1025 | * and this cost outweighs the gains made by aligning the IP header. |
| 1026 | * |
| 1027 | * Since this trade off varies between architectures, we allow NET_IP_ALIGN |
| 1028 | * to be overridden. |
| 1029 | */ |
| 1030 | #ifndef NET_IP_ALIGN |
| 1031 | #define NET_IP_ALIGN 2 |
| 1032 | #endif |
| 1033 | |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 1034 | /* |
| 1035 | * The networking layer reserves some headroom in skb data (via |
| 1036 | * dev_alloc_skb). This is used to avoid having to reallocate skb data when |
| 1037 | * the header has to grow. In the default case, if the header has to grow |
| 1038 | * 16 bytes or less we avoid the reallocation. |
| 1039 | * |
| 1040 | * Unfortunately this headroom changes the DMA alignment of the resulting |
| 1041 | * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive |
| 1042 | * on some architectures. An architecture can override this value, |
| 1043 | * perhaps setting it to a cacheline in size (since that will maintain |
| 1044 | * cacheline alignment of the DMA). It must be a power of 2. |
| 1045 | * |
| 1046 | * Various parts of the networking layer expect at least 16 bytes of |
| 1047 | * headroom, you should not reduce this. |
| 1048 | */ |
| 1049 | #ifndef NET_SKB_PAD |
| 1050 | #define NET_SKB_PAD 16 |
| 1051 | #endif |
| 1052 | |
Herbert Xu | 3cc0e87 | 2006-06-09 16:13:38 -0700 | [diff] [blame] | 1053 | extern int ___pskb_trim(struct sk_buff *skb, unsigned int len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1054 | |
| 1055 | static inline void __skb_trim(struct sk_buff *skb, unsigned int len) |
| 1056 | { |
Herbert Xu | 3cc0e87 | 2006-06-09 16:13:38 -0700 | [diff] [blame] | 1057 | if (unlikely(skb->data_len)) { |
| 1058 | WARN_ON(1); |
| 1059 | return; |
| 1060 | } |
| 1061 | skb->len = len; |
| 1062 | skb->tail = skb->data + len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1063 | } |
| 1064 | |
| 1065 | /** |
| 1066 | * skb_trim - remove end from a buffer |
| 1067 | * @skb: buffer to alter |
| 1068 | * @len: new length |
| 1069 | * |
| 1070 | * Cut the length of a buffer down by removing data from the tail. If |
| 1071 | * the buffer is already under the length specified it is not modified. |
Herbert Xu | 3cc0e87 | 2006-06-09 16:13:38 -0700 | [diff] [blame] | 1072 | * The skb must be linear. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1073 | */ |
| 1074 | static inline void skb_trim(struct sk_buff *skb, unsigned int len) |
| 1075 | { |
| 1076 | if (skb->len > len) |
| 1077 | __skb_trim(skb, len); |
| 1078 | } |
| 1079 | |
| 1080 | |
| 1081 | static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) |
| 1082 | { |
Herbert Xu | 3cc0e87 | 2006-06-09 16:13:38 -0700 | [diff] [blame] | 1083 | if (skb->data_len) |
| 1084 | return ___pskb_trim(skb, len); |
| 1085 | __skb_trim(skb, len); |
| 1086 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1087 | } |
| 1088 | |
| 1089 | static inline int pskb_trim(struct sk_buff *skb, unsigned int len) |
| 1090 | { |
| 1091 | return (len < skb->len) ? __pskb_trim(skb, len) : 0; |
| 1092 | } |
| 1093 | |
| 1094 | /** |
Herbert Xu | e9fa4f7 | 2006-08-13 20:12:58 -0700 | [diff] [blame] | 1095 | * pskb_trim_unique - remove end from a paged unique (not cloned) buffer |
| 1096 | * @skb: buffer to alter |
| 1097 | * @len: new length |
| 1098 | * |
| 1099 | * This is identical to pskb_trim except that the caller knows that |
| 1100 | * the skb is not cloned so we should never get an error due to out- |
| 1101 | * of-memory. |
| 1102 | */ |
| 1103 | static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) |
| 1104 | { |
| 1105 | int err = pskb_trim(skb, len); |
| 1106 | BUG_ON(err); |
| 1107 | } |
| 1108 | |
| 1109 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1110 | * skb_orphan - orphan a buffer |
| 1111 | * @skb: buffer to orphan |
| 1112 | * |
| 1113 | * If a buffer currently has an owner then we call the owner's |
| 1114 | * destructor function and make the @skb unowned. The buffer continues |
| 1115 | * to exist but is no longer charged to its former owner. |
| 1116 | */ |
| 1117 | static inline void skb_orphan(struct sk_buff *skb) |
| 1118 | { |
| 1119 | if (skb->destructor) |
| 1120 | skb->destructor(skb); |
| 1121 | skb->destructor = NULL; |
| 1122 | skb->sk = NULL; |
| 1123 | } |
| 1124 | |
| 1125 | /** |
| 1126 | * __skb_queue_purge - empty a list |
| 1127 | * @list: list to empty |
| 1128 | * |
| 1129 | * Delete all buffers on an &sk_buff list. Each buffer is removed from |
| 1130 | * the list and one reference dropped. This function does not take the |
| 1131 | * list lock and the caller must hold the relevant locks to use it. |
| 1132 | */ |
| 1133 | extern void skb_queue_purge(struct sk_buff_head *list); |
| 1134 | static inline void __skb_queue_purge(struct sk_buff_head *list) |
| 1135 | { |
| 1136 | struct sk_buff *skb; |
| 1137 | while ((skb = __skb_dequeue(list)) != NULL) |
| 1138 | kfree_skb(skb); |
| 1139 | } |
| 1140 | |
| 1141 | /** |
Christoph Hellwig | b4e54de | 2006-07-24 15:31:14 -0700 | [diff] [blame] | 1142 | * __dev_alloc_skb - allocate an skbuff for receiving |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1143 | * @length: length to allocate |
| 1144 | * @gfp_mask: get_free_pages mask, passed to alloc_skb |
| 1145 | * |
| 1146 | * Allocate a new &sk_buff and assign it a usage count of one. The |
| 1147 | * buffer has unspecified headroom built in. Users should allocate |
| 1148 | * the headroom they think they need without accounting for the |
| 1149 | * built in space. The built in space is used for optimisations. |
| 1150 | * |
Christoph Hellwig | 766ea8c | 2006-08-07 15:49:53 -0700 | [diff] [blame] | 1151 | * %NULL is returned if there is no free memory. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1152 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1153 | static inline struct sk_buff *__dev_alloc_skb(unsigned int length, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1154 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1155 | { |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 1156 | struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1157 | if (likely(skb)) |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 1158 | skb_reserve(skb, NET_SKB_PAD); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1159 | return skb; |
| 1160 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1161 | |
| 1162 | /** |
Christoph Hellwig | b4e54de | 2006-07-24 15:31:14 -0700 | [diff] [blame] | 1163 | * dev_alloc_skb - allocate an skbuff for receiving |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1164 | * @length: length to allocate |
| 1165 | * |
| 1166 | * Allocate a new &sk_buff and assign it a usage count of one. The |
| 1167 | * buffer has unspecified headroom built in. Users should allocate |
| 1168 | * the headroom they think they need without accounting for the |
| 1169 | * built in space. The built in space is used for optimisations. |
| 1170 | * |
Christoph Hellwig | 766ea8c | 2006-08-07 15:49:53 -0700 | [diff] [blame] | 1171 | * %NULL is returned if there is no free memory. Although this function |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1172 | * allocates memory it can be called from an interrupt. |
| 1173 | */ |
| 1174 | static inline struct sk_buff *dev_alloc_skb(unsigned int length) |
| 1175 | { |
| 1176 | return __dev_alloc_skb(length, GFP_ATOMIC); |
| 1177 | } |
| 1178 | |
Christoph Hellwig | 8af2745 | 2006-07-31 22:35:23 -0700 | [diff] [blame] | 1179 | extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev, |
| 1180 | unsigned int length, gfp_t gfp_mask); |
| 1181 | |
| 1182 | /** |
| 1183 | * netdev_alloc_skb - allocate an skbuff for rx on a specific device |
| 1184 | * @dev: network device to receive on |
| 1185 | * @length: length to allocate |
| 1186 | * |
| 1187 | * Allocate a new &sk_buff and assign it a usage count of one. The |
| 1188 | * buffer has unspecified headroom built in. Users should allocate |
| 1189 | * the headroom they think they need without accounting for the |
| 1190 | * built in space. The built in space is used for optimisations. |
| 1191 | * |
| 1192 | * %NULL is returned if there is no free memory. Although this function |
| 1193 | * allocates memory it can be called from an interrupt. |
| 1194 | */ |
| 1195 | static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, |
| 1196 | unsigned int length) |
| 1197 | { |
| 1198 | return __netdev_alloc_skb(dev, length, GFP_ATOMIC); |
| 1199 | } |
| 1200 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1201 | /** |
| 1202 | * skb_cow - copy header of skb when it is required |
| 1203 | * @skb: buffer to cow |
| 1204 | * @headroom: needed headroom |
| 1205 | * |
| 1206 | * If the skb passed lacks sufficient headroom or its data part |
| 1207 | * is shared, data is reallocated. If reallocation fails, an error |
| 1208 | * is returned and original skb is not changed. |
| 1209 | * |
| 1210 | * The result is skb with writable area skb->head...skb->tail |
| 1211 | * and at least @headroom of space at head. |
| 1212 | */ |
| 1213 | static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) |
| 1214 | { |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 1215 | int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) - |
| 1216 | skb_headroom(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1217 | |
| 1218 | if (delta < 0) |
| 1219 | delta = 0; |
| 1220 | |
| 1221 | if (delta || skb_cloned(skb)) |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 1222 | return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) & |
| 1223 | ~(NET_SKB_PAD-1), 0, GFP_ATOMIC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1224 | return 0; |
| 1225 | } |
| 1226 | |
| 1227 | /** |
| 1228 | * skb_padto - pad an skbuff up to a minimal size |
| 1229 | * @skb: buffer to pad |
| 1230 | * @len: minimal length |
| 1231 | * |
| 1232 | * Pads up a buffer to ensure the trailing bytes exist and are |
| 1233 | * blanked. If the buffer already contains sufficient data it |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 1234 | * is untouched. Otherwise it is extended. Returns zero on |
| 1235 | * success. The skb is freed on error. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1236 | */ |
| 1237 | |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 1238 | static inline int skb_padto(struct sk_buff *skb, unsigned int len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1239 | { |
| 1240 | unsigned int size = skb->len; |
| 1241 | if (likely(size >= len)) |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 1242 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1243 | return skb_pad(skb, len-size); |
| 1244 | } |
| 1245 | |
| 1246 | static inline int skb_add_data(struct sk_buff *skb, |
| 1247 | char __user *from, int copy) |
| 1248 | { |
| 1249 | const int off = skb->len; |
| 1250 | |
| 1251 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1252 | int err = 0; |
Al Viro | 5084205 | 2006-11-14 21:36:34 -0800 | [diff] [blame] | 1253 | __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy), |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1254 | copy, 0, &err); |
| 1255 | if (!err) { |
| 1256 | skb->csum = csum_block_add(skb->csum, csum, off); |
| 1257 | return 0; |
| 1258 | } |
| 1259 | } else if (!copy_from_user(skb_put(skb, copy), from, copy)) |
| 1260 | return 0; |
| 1261 | |
| 1262 | __skb_trim(skb, off); |
| 1263 | return -EFAULT; |
| 1264 | } |
| 1265 | |
| 1266 | static inline int skb_can_coalesce(struct sk_buff *skb, int i, |
| 1267 | struct page *page, int off) |
| 1268 | { |
| 1269 | if (i) { |
| 1270 | struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; |
| 1271 | |
| 1272 | return page == frag->page && |
| 1273 | off == frag->page_offset + frag->size; |
| 1274 | } |
| 1275 | return 0; |
| 1276 | } |
| 1277 | |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 1278 | static inline int __skb_linearize(struct sk_buff *skb) |
| 1279 | { |
| 1280 | return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; |
| 1281 | } |
| 1282 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1283 | /** |
| 1284 | * skb_linearize - convert paged skb to linear one |
| 1285 | * @skb: buffer to linarize |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1286 | * |
| 1287 | * If there is no free memory -ENOMEM is returned, otherwise zero |
| 1288 | * is returned and the old skb data released. |
| 1289 | */ |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 1290 | static inline int skb_linearize(struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1291 | { |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 1292 | return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; |
| 1293 | } |
| 1294 | |
| 1295 | /** |
| 1296 | * skb_linearize_cow - make sure skb is linear and writable |
| 1297 | * @skb: buffer to process |
| 1298 | * |
| 1299 | * If there is no free memory -ENOMEM is returned, otherwise zero |
| 1300 | * is returned and the old skb data released. |
| 1301 | */ |
| 1302 | static inline int skb_linearize_cow(struct sk_buff *skb) |
| 1303 | { |
| 1304 | return skb_is_nonlinear(skb) || skb_cloned(skb) ? |
| 1305 | __skb_linearize(skb) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1306 | } |
| 1307 | |
| 1308 | /** |
| 1309 | * skb_postpull_rcsum - update checksum for received skb after pull |
| 1310 | * @skb: buffer to update |
| 1311 | * @start: start of data before pull |
| 1312 | * @len: length of data pulled |
| 1313 | * |
| 1314 | * After doing a pull on a received packet, you need to call this to |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 1315 | * update the CHECKSUM_COMPLETE checksum, or set ip_summed to |
| 1316 | * CHECKSUM_NONE so that it can be recomputed from scratch. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1317 | */ |
| 1318 | |
| 1319 | static inline void skb_postpull_rcsum(struct sk_buff *skb, |
Herbert Xu | cbb042f | 2006-03-20 22:43:56 -0800 | [diff] [blame] | 1320 | const void *start, unsigned int len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1321 | { |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 1322 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1323 | skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); |
| 1324 | } |
| 1325 | |
Herbert Xu | cbb042f | 2006-03-20 22:43:56 -0800 | [diff] [blame] | 1326 | unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); |
| 1327 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1328 | /** |
| 1329 | * pskb_trim_rcsum - trim received skb and update checksum |
| 1330 | * @skb: buffer to trim |
| 1331 | * @len: new length |
| 1332 | * |
| 1333 | * This is exactly the same as pskb_trim except that it ensures the |
| 1334 | * checksum of received packets are still valid after the operation. |
| 1335 | */ |
| 1336 | |
| 1337 | static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) |
| 1338 | { |
Stephen Hemminger | 0e4e422 | 2005-09-08 12:32:03 -0700 | [diff] [blame] | 1339 | if (likely(len >= skb->len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1340 | return 0; |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 1341 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1342 | skb->ip_summed = CHECKSUM_NONE; |
| 1343 | return __pskb_trim(skb, len); |
| 1344 | } |
| 1345 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1346 | #define skb_queue_walk(queue, skb) \ |
| 1347 | for (skb = (queue)->next; \ |
| 1348 | prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \ |
| 1349 | skb = skb->next) |
| 1350 | |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 1351 | #define skb_queue_reverse_walk(queue, skb) \ |
| 1352 | for (skb = (queue)->prev; \ |
| 1353 | prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \ |
| 1354 | skb = skb->prev) |
| 1355 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1356 | |
| 1357 | extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, |
| 1358 | int noblock, int *err); |
| 1359 | extern unsigned int datagram_poll(struct file *file, struct socket *sock, |
| 1360 | struct poll_table_struct *wait); |
| 1361 | extern int skb_copy_datagram_iovec(const struct sk_buff *from, |
| 1362 | int offset, struct iovec *to, |
| 1363 | int size); |
Herbert Xu | fb286bb | 2005-11-10 13:01:24 -0800 | [diff] [blame] | 1364 | extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1365 | int hlen, |
| 1366 | struct iovec *iov); |
| 1367 | extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb); |
Herbert Xu | 3305b80 | 2005-12-13 23:16:37 -0800 | [diff] [blame] | 1368 | extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb, |
| 1369 | unsigned int flags); |
Al Viro | 2bbbc86 | 2006-11-14 21:37:14 -0800 | [diff] [blame] | 1370 | extern __wsum skb_checksum(const struct sk_buff *skb, int offset, |
| 1371 | int len, __wsum csum); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1372 | extern int skb_copy_bits(const struct sk_buff *skb, int offset, |
| 1373 | void *to, int len); |
Herbert Xu | 357b40a | 2005-04-19 22:30:14 -0700 | [diff] [blame] | 1374 | extern int skb_store_bits(const struct sk_buff *skb, int offset, |
| 1375 | void *from, int len); |
Al Viro | 81d7766 | 2006-11-14 21:37:33 -0800 | [diff] [blame] | 1376 | extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1377 | int offset, u8 *to, int len, |
Al Viro | 81d7766 | 2006-11-14 21:37:33 -0800 | [diff] [blame] | 1378 | __wsum csum); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1379 | extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); |
| 1380 | extern void skb_split(struct sk_buff *skb, |
| 1381 | struct sk_buff *skb1, const u32 len); |
| 1382 | |
Herbert Xu | 576a30e | 2006-06-27 13:22:38 -0700 | [diff] [blame] | 1383 | extern struct sk_buff *skb_segment(struct sk_buff *skb, int features); |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 1384 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1385 | static inline void *skb_header_pointer(const struct sk_buff *skb, int offset, |
| 1386 | int len, void *buffer) |
| 1387 | { |
| 1388 | int hlen = skb_headlen(skb); |
| 1389 | |
Patrick McHardy | 55820ee | 2005-07-05 14:08:10 -0700 | [diff] [blame] | 1390 | if (hlen - offset >= len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1391 | return skb->data + offset; |
| 1392 | |
| 1393 | if (skb_copy_bits(skb, offset, buffer, len) < 0) |
| 1394 | return NULL; |
| 1395 | |
| 1396 | return buffer; |
| 1397 | } |
| 1398 | |
| 1399 | extern void skb_init(void); |
| 1400 | extern void skb_add_mtu(int mtu); |
| 1401 | |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1402 | /** |
| 1403 | * skb_get_timestamp - get timestamp from a skb |
| 1404 | * @skb: skb to get stamp from |
| 1405 | * @stamp: pointer to struct timeval to store stamp in |
| 1406 | * |
| 1407 | * Timestamps are stored in the skb as offsets to a base timestamp. |
| 1408 | * This function converts the offset back to a struct timeval and stores |
| 1409 | * it in stamp. |
| 1410 | */ |
Stephen Hemminger | f2c3839 | 2005-09-06 15:48:03 -0700 | [diff] [blame] | 1411 | static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp) |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1412 | { |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 1413 | *stamp = ktime_to_timeval(skb->tstamp); |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1414 | } |
| 1415 | |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 1416 | static inline void __net_timestamp(struct sk_buff *skb) |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1417 | { |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 1418 | skb->tstamp = ktime_get_real(); |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1419 | } |
| 1420 | |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1421 | |
Herbert Xu | 759e5d0 | 2007-03-25 20:10:56 -0700 | [diff] [blame] | 1422 | extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); |
Al Viro | b51655b | 2006-11-14 21:40:42 -0800 | [diff] [blame] | 1423 | extern __sum16 __skb_checksum_complete(struct sk_buff *skb); |
Herbert Xu | fb286bb | 2005-11-10 13:01:24 -0800 | [diff] [blame] | 1424 | |
| 1425 | /** |
| 1426 | * skb_checksum_complete - Calculate checksum of an entire packet |
| 1427 | * @skb: packet to process |
| 1428 | * |
| 1429 | * This function calculates the checksum over the entire packet plus |
| 1430 | * the value of skb->csum. The latter can be used to supply the |
| 1431 | * checksum of a pseudo header as used by TCP/UDP. It returns the |
| 1432 | * checksum. |
| 1433 | * |
| 1434 | * For protocols that contain complete checksums such as ICMP/TCP/UDP, |
| 1435 | * this function can be used to verify that checksum on received |
| 1436 | * packets. In that case the function should return zero if the |
| 1437 | * checksum is correct. In particular, this function will return zero |
| 1438 | * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the |
| 1439 | * hardware has already verified the correctness of the checksum. |
| 1440 | */ |
| 1441 | static inline unsigned int skb_checksum_complete(struct sk_buff *skb) |
| 1442 | { |
| 1443 | return skb->ip_summed != CHECKSUM_UNNECESSARY && |
| 1444 | __skb_checksum_complete(skb); |
| 1445 | } |
| 1446 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1447 | #ifdef CONFIG_NETFILTER |
| 1448 | static inline void nf_conntrack_put(struct nf_conntrack *nfct) |
| 1449 | { |
| 1450 | if (nfct && atomic_dec_and_test(&nfct->use)) |
| 1451 | nfct->destroy(nfct); |
| 1452 | } |
| 1453 | static inline void nf_conntrack_get(struct nf_conntrack *nfct) |
| 1454 | { |
| 1455 | if (nfct) |
| 1456 | atomic_inc(&nfct->use); |
| 1457 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1458 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
| 1459 | static inline void nf_conntrack_get_reasm(struct sk_buff *skb) |
| 1460 | { |
| 1461 | if (skb) |
| 1462 | atomic_inc(&skb->users); |
| 1463 | } |
| 1464 | static inline void nf_conntrack_put_reasm(struct sk_buff *skb) |
| 1465 | { |
| 1466 | if (skb) |
| 1467 | kfree_skb(skb); |
| 1468 | } |
| 1469 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1470 | #ifdef CONFIG_BRIDGE_NETFILTER |
| 1471 | static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) |
| 1472 | { |
| 1473 | if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) |
| 1474 | kfree(nf_bridge); |
| 1475 | } |
| 1476 | static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) |
| 1477 | { |
| 1478 | if (nf_bridge) |
| 1479 | atomic_inc(&nf_bridge->use); |
| 1480 | } |
| 1481 | #endif /* CONFIG_BRIDGE_NETFILTER */ |
Patrick McHardy | a193a4a | 2006-03-20 19:23:05 -0800 | [diff] [blame] | 1482 | static inline void nf_reset(struct sk_buff *skb) |
| 1483 | { |
| 1484 | nf_conntrack_put(skb->nfct); |
| 1485 | skb->nfct = NULL; |
| 1486 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
| 1487 | nf_conntrack_put_reasm(skb->nfct_reasm); |
| 1488 | skb->nfct_reasm = NULL; |
| 1489 | #endif |
| 1490 | #ifdef CONFIG_BRIDGE_NETFILTER |
| 1491 | nf_bridge_put(skb->nf_bridge); |
| 1492 | skb->nf_bridge = NULL; |
| 1493 | #endif |
| 1494 | } |
| 1495 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1496 | #else /* CONFIG_NETFILTER */ |
| 1497 | static inline void nf_reset(struct sk_buff *skb) {} |
| 1498 | #endif /* CONFIG_NETFILTER */ |
| 1499 | |
James Morris | 984bc16 | 2006-06-09 00:29:17 -0700 | [diff] [blame] | 1500 | #ifdef CONFIG_NETWORK_SECMARK |
| 1501 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) |
| 1502 | { |
| 1503 | to->secmark = from->secmark; |
| 1504 | } |
| 1505 | |
| 1506 | static inline void skb_init_secmark(struct sk_buff *skb) |
| 1507 | { |
| 1508 | skb->secmark = 0; |
| 1509 | } |
| 1510 | #else |
| 1511 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) |
| 1512 | { } |
| 1513 | |
| 1514 | static inline void skb_init_secmark(struct sk_buff *skb) |
| 1515 | { } |
| 1516 | #endif |
| 1517 | |
Herbert Xu | 89114af | 2006-07-08 13:34:32 -0700 | [diff] [blame] | 1518 | static inline int skb_is_gso(const struct sk_buff *skb) |
| 1519 | { |
| 1520 | return skb_shinfo(skb)->gso_size; |
| 1521 | } |
| 1522 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1523 | #endif /* __KERNEL__ */ |
| 1524 | #endif /* _LINUX_SKBUFF_H */ |