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> |
Vegard Nossum | fe55f6d | 2008-08-30 12:16:35 +0200 | [diff] [blame] | 18 | #include <linux/kmemcheck.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 19 | #include <linux/compiler.h> |
| 20 | #include <linux/time.h> |
Paul Gortmaker | 187f188 | 2011-11-23 20:12:59 -0500 | [diff] [blame] | 21 | #include <linux/bug.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include <linux/cache.h> |
Eric Dumazet | 56b1742 | 2014-11-03 08:19:53 -0800 | [diff] [blame] | 23 | #include <linux/rbtree.h> |
David S. Miller | 51f3d02 | 2014-11-05 16:46:40 -0500 | [diff] [blame] | 24 | #include <linux/socket.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | |
Arun Sharma | 60063497 | 2011-07-26 16:09:06 -0700 | [diff] [blame] | 26 | #include <linux/atomic.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | #include <asm/types.h> |
| 28 | #include <linux/spinlock.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 29 | #include <linux/net.h> |
Thomas Graf | 3fc7e8a | 2005-06-23 21:00:17 -0700 | [diff] [blame] | 30 | #include <linux/textsearch.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | #include <net/checksum.h> |
Al Viro | a80958f | 2006-12-04 20:41:19 +0000 | [diff] [blame] | 32 | #include <linux/rcupdate.h> |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 33 | #include <linux/hrtimer.h> |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 34 | #include <linux/dma-mapping.h> |
Michał Mirosław | c8f44af | 2011-11-15 15:29:55 +0000 | [diff] [blame] | 35 | #include <linux/netdev_features.h> |
Eric Dumazet | 363ec3923 | 2014-02-26 14:02:11 -0800 | [diff] [blame] | 36 | #include <linux/sched.h> |
Jiri Pirko | 1bd758e | 2015-05-12 14:56:07 +0200 | [diff] [blame] | 37 | #include <net/flow_dissector.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 38 | |
Daniel Borkmann | 78ea85f | 2013-12-16 23:27:09 +0100 | [diff] [blame] | 39 | /* A. Checksumming of received packets by device. |
| 40 | * |
| 41 | * CHECKSUM_NONE: |
| 42 | * |
| 43 | * Device failed to checksum this packet e.g. due to lack of capabilities. |
| 44 | * The packet contains full (though not verified) checksum in packet but |
| 45 | * not in skb->csum. Thus, skb->csum is undefined in this case. |
| 46 | * |
| 47 | * CHECKSUM_UNNECESSARY: |
| 48 | * |
| 49 | * The hardware you're dealing with doesn't calculate the full checksum |
| 50 | * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums |
Tom Herbert | 77cffe2 | 2014-08-27 21:26:46 -0700 | [diff] [blame] | 51 | * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY |
| 52 | * if their checksums are okay. skb->csum is still undefined in this case |
| 53 | * though. It is a bad option, but, unfortunately, nowadays most vendors do |
| 54 | * this. Apparently with the secret goal to sell you new devices, when you |
| 55 | * will add new protocol to your host, f.e. IPv6 8) |
| 56 | * |
| 57 | * CHECKSUM_UNNECESSARY is applicable to following protocols: |
| 58 | * TCP: IPv6 and IPv4. |
| 59 | * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a |
| 60 | * zero UDP checksum for either IPv4 or IPv6, the networking stack |
| 61 | * may perform further validation in this case. |
| 62 | * GRE: only if the checksum is present in the header. |
| 63 | * SCTP: indicates the CRC in SCTP header has been validated. |
| 64 | * |
| 65 | * skb->csum_level indicates the number of consecutive checksums found in |
| 66 | * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. |
| 67 | * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet |
| 68 | * and a device is able to verify the checksums for UDP (possibly zero), |
| 69 | * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to |
| 70 | * two. If the device were only able to verify the UDP checksum and not |
| 71 | * GRE, either because it doesn't support GRE checksum of because GRE |
| 72 | * checksum is bad, skb->csum_level would be set to zero (TCP checksum is |
| 73 | * not considered in this case). |
Daniel Borkmann | 78ea85f | 2013-12-16 23:27:09 +0100 | [diff] [blame] | 74 | * |
| 75 | * CHECKSUM_COMPLETE: |
| 76 | * |
| 77 | * This is the most generic way. The device supplied checksum of the _whole_ |
| 78 | * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the |
| 79 | * hardware doesn't need to parse L3/L4 headers to implement this. |
| 80 | * |
| 81 | * Note: Even if device supports only some protocols, but is able to produce |
| 82 | * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. |
| 83 | * |
| 84 | * CHECKSUM_PARTIAL: |
| 85 | * |
Tom Herbert | 6edec0e | 2015-02-10 16:30:28 -0800 | [diff] [blame] | 86 | * A checksum is set up to be offloaded to a device as described in the |
| 87 | * output description for CHECKSUM_PARTIAL. This may occur on a packet |
Daniel Borkmann | 78ea85f | 2013-12-16 23:27:09 +0100 | [diff] [blame] | 88 | * received directly from another Linux OS, e.g., a virtualized Linux kernel |
Tom Herbert | 6edec0e | 2015-02-10 16:30:28 -0800 | [diff] [blame] | 89 | * on the same host, or it may be set in the input path in GRO or remote |
| 90 | * checksum offload. For the purposes of checksum verification, the checksum |
| 91 | * referred to by skb->csum_start + skb->csum_offset and any preceding |
| 92 | * checksums in the packet are considered verified. Any checksums in the |
| 93 | * packet that are after the checksum being offloaded are not considered to |
| 94 | * be verified. |
Daniel Borkmann | 78ea85f | 2013-12-16 23:27:09 +0100 | [diff] [blame] | 95 | * |
| 96 | * B. Checksumming on output. |
| 97 | * |
| 98 | * CHECKSUM_NONE: |
| 99 | * |
| 100 | * The skb was already checksummed by the protocol, or a checksum is not |
| 101 | * required. |
| 102 | * |
| 103 | * CHECKSUM_PARTIAL: |
| 104 | * |
| 105 | * The device is required to checksum the packet as seen by hard_start_xmit() |
| 106 | * from skb->csum_start up to the end, and to record/write the checksum at |
| 107 | * offset skb->csum_start + skb->csum_offset. |
| 108 | * |
| 109 | * The device must show its capabilities in dev->features, set up at device |
| 110 | * setup time, e.g. netdev_features.h: |
| 111 | * |
| 112 | * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything. |
| 113 | * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over |
| 114 | * IPv4. Sigh. Vendors like this way for an unknown reason. |
| 115 | * Though, see comment above about CHECKSUM_UNNECESSARY. 8) |
| 116 | * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead. |
| 117 | * NETIF_F_... - Well, you get the picture. |
| 118 | * |
| 119 | * CHECKSUM_UNNECESSARY: |
| 120 | * |
| 121 | * Normally, the device will do per protocol specific checksumming. Protocol |
| 122 | * implementations that do not want the NIC to perform the checksum |
| 123 | * calculation should use this flag in their outgoing skbs. |
| 124 | * |
| 125 | * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC |
| 126 | * offload. Correspondingly, the FCoE protocol driver |
| 127 | * stack should use CHECKSUM_UNNECESSARY. |
| 128 | * |
| 129 | * Any questions? No questions, good. --ANK |
| 130 | */ |
| 131 | |
Herbert Xu | 6047637 | 2007-04-09 11:59:39 -0700 | [diff] [blame] | 132 | /* Don't change this without changing skb_csum_unnecessary! */ |
Daniel Borkmann | 78ea85f | 2013-12-16 23:27:09 +0100 | [diff] [blame] | 133 | #define CHECKSUM_NONE 0 |
| 134 | #define CHECKSUM_UNNECESSARY 1 |
| 135 | #define CHECKSUM_COMPLETE 2 |
| 136 | #define CHECKSUM_PARTIAL 3 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 137 | |
Tom Herbert | 77cffe2 | 2014-08-27 21:26:46 -0700 | [diff] [blame] | 138 | /* Maximum value in skb->csum_level */ |
| 139 | #define SKB_MAX_CSUM_LEVEL 3 |
| 140 | |
Tobias Klauser | 0bec8c8 | 2014-07-22 12:06:23 +0200 | [diff] [blame] | 141 | #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) |
David S. Miller | fc910a2 | 2007-03-25 20:27:59 -0700 | [diff] [blame] | 142 | #define SKB_WITH_OVERHEAD(X) \ |
Herbert Xu | deea84b | 2007-10-21 16:27:46 -0700 | [diff] [blame] | 143 | ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) |
David S. Miller | fc910a2 | 2007-03-25 20:27:59 -0700 | [diff] [blame] | 144 | #define SKB_MAX_ORDER(X, ORDER) \ |
| 145 | SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 146 | #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) |
| 147 | #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) |
| 148 | |
Eric Dumazet | 87fb4b7 | 2011-10-13 07:28:54 +0000 | [diff] [blame] | 149 | /* return minimum truesize of one skb containing X bytes of data */ |
| 150 | #define SKB_TRUESIZE(X) ((X) + \ |
| 151 | SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ |
| 152 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) |
| 153 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 154 | struct net_device; |
David Howells | 716ea3a | 2007-04-02 20:19:53 -0700 | [diff] [blame] | 155 | struct scatterlist; |
Jens Axboe | 9c55e01 | 2007-11-06 23:30:13 -0800 | [diff] [blame] | 156 | struct pipe_inode_info; |
Herbert Xu | a8f820aa | 2014-11-07 21:22:22 +0800 | [diff] [blame] | 157 | struct iov_iter; |
Alexander Duyck | fd11a83 | 2014-12-09 19:40:49 -0800 | [diff] [blame] | 158 | struct napi_struct; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 159 | |
Yasuyuki Kozakai | 5f79e0f | 2007-03-23 11:17:07 -0700 | [diff] [blame] | 160 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 161 | struct nf_conntrack { |
| 162 | atomic_t use; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 163 | }; |
Yasuyuki Kozakai | 5f79e0f | 2007-03-23 11:17:07 -0700 | [diff] [blame] | 164 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 165 | |
Pablo Neira Ayuso | 34666d4 | 2014-09-18 11:29:03 +0200 | [diff] [blame] | 166 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 167 | struct nf_bridge_info { |
Eric Dumazet | bf1ac5c | 2012-04-18 23:19:25 +0000 | [diff] [blame] | 168 | atomic_t use; |
Florian Westphal | 3eaf402 | 2015-04-02 14:31:44 +0200 | [diff] [blame] | 169 | enum { |
| 170 | BRNF_PROTO_UNCHANGED, |
| 171 | BRNF_PROTO_8021Q, |
| 172 | BRNF_PROTO_PPPOE |
Florian Westphal | 7fb48c5 | 2015-05-03 22:05:28 +0200 | [diff] [blame] | 173 | } orig_proto:8; |
Florian Westphal | a1e6795 | 2015-04-02 14:31:45 +0200 | [diff] [blame] | 174 | bool pkt_otherhost; |
Eric Dumazet | bf1ac5c | 2012-04-18 23:19:25 +0000 | [diff] [blame] | 175 | unsigned int mask; |
| 176 | struct net_device *physindev; |
Florian Westphal | 7fb48c5 | 2015-05-03 22:05:28 +0200 | [diff] [blame] | 177 | union { |
| 178 | struct net_device *physoutdev; |
| 179 | char neigh_header[8]; |
| 180 | }; |
Florian Westphal | faecbb4 | 2015-05-20 13:42:25 +0200 | [diff] [blame] | 181 | __be32 ipv4_daddr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | }; |
| 183 | #endif |
| 184 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 185 | struct sk_buff_head { |
| 186 | /* These two members must be first. */ |
| 187 | struct sk_buff *next; |
| 188 | struct sk_buff *prev; |
| 189 | |
| 190 | __u32 qlen; |
| 191 | spinlock_t lock; |
| 192 | }; |
| 193 | |
| 194 | struct sk_buff; |
| 195 | |
Ian Campbell | 9d4dde5 | 2011-12-22 23:39:14 +0000 | [diff] [blame] | 196 | /* To allow 64K frame to be packed as single skb without frag_list we |
| 197 | * require 64K/PAGE_SIZE pages plus 1 additional page to allow for |
| 198 | * buffers which do not start on a page boundary. |
| 199 | * |
| 200 | * Since GRO uses frags we allocate at least 16 regardless of page |
| 201 | * size. |
Anton Blanchard | a715dea | 2011-03-27 14:57:26 +0000 | [diff] [blame] | 202 | */ |
Ian Campbell | 9d4dde5 | 2011-12-22 23:39:14 +0000 | [diff] [blame] | 203 | #if (65536/PAGE_SIZE + 1) < 16 |
David S. Miller | eec0095 | 2011-03-29 23:34:08 -0700 | [diff] [blame] | 204 | #define MAX_SKB_FRAGS 16UL |
Anton Blanchard | a715dea | 2011-03-27 14:57:26 +0000 | [diff] [blame] | 205 | #else |
Ian Campbell | 9d4dde5 | 2011-12-22 23:39:14 +0000 | [diff] [blame] | 206 | #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) |
Anton Blanchard | a715dea | 2011-03-27 14:57:26 +0000 | [diff] [blame] | 207 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 208 | |
| 209 | typedef struct skb_frag_struct skb_frag_t; |
| 210 | |
| 211 | struct skb_frag_struct { |
Ian Campbell | a8605c6 | 2011-10-19 23:01:49 +0000 | [diff] [blame] | 212 | struct { |
| 213 | struct page *p; |
| 214 | } page; |
Eric Dumazet | cb4dfe5 | 2010-09-23 05:06:54 +0000 | [diff] [blame] | 215 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) |
David S. Miller | a309bb0 | 2007-07-30 18:47:03 -0700 | [diff] [blame] | 216 | __u32 page_offset; |
| 217 | __u32 size; |
Eric Dumazet | cb4dfe5 | 2010-09-23 05:06:54 +0000 | [diff] [blame] | 218 | #else |
| 219 | __u16 page_offset; |
| 220 | __u16 size; |
| 221 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 222 | }; |
| 223 | |
Eric Dumazet | 9e903e0 | 2011-10-18 21:00:24 +0000 | [diff] [blame] | 224 | static inline unsigned int skb_frag_size(const skb_frag_t *frag) |
| 225 | { |
| 226 | return frag->size; |
| 227 | } |
| 228 | |
| 229 | static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) |
| 230 | { |
| 231 | frag->size = size; |
| 232 | } |
| 233 | |
| 234 | static inline void skb_frag_size_add(skb_frag_t *frag, int delta) |
| 235 | { |
| 236 | frag->size += delta; |
| 237 | } |
| 238 | |
| 239 | static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) |
| 240 | { |
| 241 | frag->size -= delta; |
| 242 | } |
| 243 | |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 244 | #define HAVE_HW_TIME_STAMP |
| 245 | |
| 246 | /** |
Randy Dunlap | d3a21be | 2009-03-02 03:15:58 -0800 | [diff] [blame] | 247 | * struct skb_shared_hwtstamps - hardware time stamps |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 248 | * @hwtstamp: hardware time stamp transformed into duration |
| 249 | * since arbitrary point in time |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 250 | * |
| 251 | * Software time stamps generated by ktime_get_real() are stored in |
Willem de Bruijn | 4d276eb | 2014-07-25 18:01:32 -0400 | [diff] [blame] | 252 | * skb->tstamp. |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 253 | * |
| 254 | * hwtstamps can only be compared against other hwtstamps from |
| 255 | * the same device. |
| 256 | * |
| 257 | * This structure is attached to packets as part of the |
| 258 | * &skb_shared_info. Use skb_hwtstamps() to get a pointer. |
| 259 | */ |
| 260 | struct skb_shared_hwtstamps { |
| 261 | ktime_t hwtstamp; |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 262 | }; |
| 263 | |
Oliver Hartkopp | 2244d07 | 2010-08-17 08:59:14 +0000 | [diff] [blame] | 264 | /* Definitions for tx_flags in struct skb_shared_info */ |
| 265 | enum { |
| 266 | /* generate hardware time stamp */ |
| 267 | SKBTX_HW_TSTAMP = 1 << 0, |
| 268 | |
Willem de Bruijn | e7fd288 | 2014-08-04 22:11:48 -0400 | [diff] [blame] | 269 | /* generate software time stamp when queueing packet to NIC */ |
Oliver Hartkopp | 2244d07 | 2010-08-17 08:59:14 +0000 | [diff] [blame] | 270 | SKBTX_SW_TSTAMP = 1 << 1, |
| 271 | |
| 272 | /* device driver is going to provide hardware time stamp */ |
| 273 | SKBTX_IN_PROGRESS = 1 << 2, |
| 274 | |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 275 | /* device driver supports TX zero-copy buffers */ |
Eric Dumazet | 62b1a8a | 2012-06-14 06:42:44 +0000 | [diff] [blame] | 276 | SKBTX_DEV_ZEROCOPY = 1 << 3, |
Johannes Berg | 6e3e939 | 2011-11-09 10:15:42 +0100 | [diff] [blame] | 277 | |
| 278 | /* generate wifi status information (where possible) */ |
Eric Dumazet | 62b1a8a | 2012-06-14 06:42:44 +0000 | [diff] [blame] | 279 | SKBTX_WIFI_STATUS = 1 << 4, |
Pravin B Shelar | c9af6db | 2013-02-11 09:27:41 +0000 | [diff] [blame] | 280 | |
| 281 | /* This indicates at least one fragment might be overwritten |
| 282 | * (as in vmsplice(), sendfile() ...) |
| 283 | * If we need to compute a TX checksum, we'll need to copy |
| 284 | * all frags to avoid possible bad checksum |
| 285 | */ |
| 286 | SKBTX_SHARED_FRAG = 1 << 5, |
Willem de Bruijn | e7fd288 | 2014-08-04 22:11:48 -0400 | [diff] [blame] | 287 | |
| 288 | /* generate software time stamp when entering packet scheduling */ |
| 289 | SKBTX_SCHED_TSTAMP = 1 << 6, |
Willem de Bruijn | e1c8a60 | 2014-08-04 22:11:50 -0400 | [diff] [blame] | 290 | |
| 291 | /* generate software timestamp on peer data acknowledgment */ |
| 292 | SKBTX_ACK_TSTAMP = 1 << 7, |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 293 | }; |
| 294 | |
Willem de Bruijn | e1c8a60 | 2014-08-04 22:11:50 -0400 | [diff] [blame] | 295 | #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ |
| 296 | SKBTX_SCHED_TSTAMP | \ |
| 297 | SKBTX_ACK_TSTAMP) |
Willem de Bruijn | f24b9be | 2014-08-04 22:11:45 -0400 | [diff] [blame] | 298 | #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) |
| 299 | |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 300 | /* |
| 301 | * The callback notifies userspace to release buffers when skb DMA is done in |
| 302 | * lower device, the skb last reference should be 0 when calling this. |
Michael S. Tsirkin | e19d676 | 2012-11-01 09:16:22 +0000 | [diff] [blame] | 303 | * The zerocopy_success argument is true if zero copy transmit occurred, |
| 304 | * false on data copy or out of memory error caused by data copy attempt. |
Michael S. Tsirkin | ca8f4fb | 2012-04-09 00:24:02 +0000 | [diff] [blame] | 305 | * The ctx field is used to track device context. |
| 306 | * The desc field is used to track userspace buffer index. |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 307 | */ |
| 308 | struct ubuf_info { |
Michael S. Tsirkin | e19d676 | 2012-11-01 09:16:22 +0000 | [diff] [blame] | 309 | void (*callback)(struct ubuf_info *, bool zerocopy_success); |
Michael S. Tsirkin | ca8f4fb | 2012-04-09 00:24:02 +0000 | [diff] [blame] | 310 | void *ctx; |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 311 | unsigned long desc; |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 312 | }; |
| 313 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 314 | /* This data is invariant across clones and lives at |
| 315 | * the end of the header data, ie. at skb->end. |
| 316 | */ |
| 317 | struct skb_shared_info { |
Ian Campbell | 9f42f12 | 2012-01-05 07:13:39 +0000 | [diff] [blame] | 318 | unsigned char nr_frags; |
| 319 | __u8 tx_flags; |
Herbert Xu | 7967168 | 2006-06-22 02:40:14 -0700 | [diff] [blame] | 320 | unsigned short gso_size; |
| 321 | /* Warning: this field is not always filled in (UFO)! */ |
| 322 | unsigned short gso_segs; |
| 323 | unsigned short gso_type; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 324 | struct sk_buff *frag_list; |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 325 | struct skb_shared_hwtstamps hwtstamps; |
Willem de Bruijn | 09c2d25 | 2014-08-04 22:11:47 -0400 | [diff] [blame] | 326 | u32 tskey; |
Ian Campbell | 9f42f12 | 2012-01-05 07:13:39 +0000 | [diff] [blame] | 327 | __be32 ip6_frag_id; |
Eric Dumazet | ec7d2f2 | 2010-05-05 01:07:37 -0700 | [diff] [blame] | 328 | |
| 329 | /* |
| 330 | * Warning : all fields before dataref are cleared in __alloc_skb() |
| 331 | */ |
| 332 | atomic_t dataref; |
| 333 | |
Johann Baudy | 69e3c75 | 2009-05-18 22:11:22 -0700 | [diff] [blame] | 334 | /* Intermediate layers must ensure that destructor_arg |
| 335 | * remains valid until skb destructor */ |
| 336 | void * destructor_arg; |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 337 | |
Eric Dumazet | fed6638 | 2010-07-22 19:09:08 +0000 | [diff] [blame] | 338 | /* must be last field, see pskb_expand_head() */ |
| 339 | skb_frag_t frags[MAX_SKB_FRAGS]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 340 | }; |
| 341 | |
| 342 | /* We divide dataref into two halves. The higher 16 bits hold references |
| 343 | * to the payload part of skb->data. The lower 16 bits hold references to |
Patrick McHardy | 334a813 | 2007-06-25 04:35:20 -0700 | [diff] [blame] | 344 | * the entire skb->data. A clone of a headerless skb holds the length of |
| 345 | * the header in skb->hdr_len. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 346 | * |
| 347 | * All users must obey the rule that the skb->data reference count must be |
| 348 | * greater than or equal to the payload reference count. |
| 349 | * |
| 350 | * Holding a reference to the payload part means that the user does not |
| 351 | * care about modifications to the header part of skb->data. |
| 352 | */ |
| 353 | #define SKB_DATAREF_SHIFT 16 |
| 354 | #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) |
| 355 | |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 356 | |
| 357 | enum { |
Vijay Subramanian | c8753d5 | 2014-10-02 10:00:43 -0700 | [diff] [blame] | 358 | SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ |
| 359 | SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ |
| 360 | SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 361 | }; |
| 362 | |
Herbert Xu | 7967168 | 2006-06-22 02:40:14 -0700 | [diff] [blame] | 363 | enum { |
| 364 | SKB_GSO_TCPV4 = 1 << 0, |
Herbert Xu | f83ef8c | 2006-06-30 13:37:03 -0700 | [diff] [blame] | 365 | SKB_GSO_UDP = 1 << 1, |
Herbert Xu | 576a30e | 2006-06-27 13:22:38 -0700 | [diff] [blame] | 366 | |
| 367 | /* This indicates the skb is from an untrusted source. */ |
| 368 | SKB_GSO_DODGY = 1 << 2, |
Michael Chan | b0da8537 | 2006-06-29 12:30:00 -0700 | [diff] [blame] | 369 | |
| 370 | /* This indicates the tcp segment has CWR set. */ |
Herbert Xu | f83ef8c | 2006-06-30 13:37:03 -0700 | [diff] [blame] | 371 | SKB_GSO_TCP_ECN = 1 << 3, |
| 372 | |
| 373 | SKB_GSO_TCPV6 = 1 << 4, |
Chris Leech | 01d5b2f | 2009-02-27 14:06:49 -0800 | [diff] [blame] | 374 | |
| 375 | SKB_GSO_FCOE = 1 << 5, |
Pravin B Shelar | 68c3316 | 2013-02-14 14:02:41 +0000 | [diff] [blame] | 376 | |
| 377 | SKB_GSO_GRE = 1 << 6, |
Pravin B Shelar | 7313626 | 2013-03-07 13:21:51 +0000 | [diff] [blame] | 378 | |
Tom Herbert | 4b28252 | 2014-06-14 23:23:52 -0700 | [diff] [blame] | 379 | SKB_GSO_GRE_CSUM = 1 << 7, |
Simon Horman | 0d89d20 | 2013-05-23 21:02:52 +0000 | [diff] [blame] | 380 | |
Tom Herbert | 4b28252 | 2014-06-14 23:23:52 -0700 | [diff] [blame] | 381 | SKB_GSO_IPIP = 1 << 8, |
Eric Dumazet | cb32f51 | 2013-10-19 11:42:57 -0700 | [diff] [blame] | 382 | |
Tom Herbert | 4b28252 | 2014-06-14 23:23:52 -0700 | [diff] [blame] | 383 | SKB_GSO_SIT = 1 << 9, |
Eric Dumazet | 61c1db7 | 2013-10-20 20:47:30 -0700 | [diff] [blame] | 384 | |
Tom Herbert | 4b28252 | 2014-06-14 23:23:52 -0700 | [diff] [blame] | 385 | SKB_GSO_UDP_TUNNEL = 1 << 10, |
Tom Herbert | 0f4f4ff | 2014-06-04 17:20:16 -0700 | [diff] [blame] | 386 | |
| 387 | SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, |
Tom Herbert | 4749c09 | 2014-06-04 17:20:23 -0700 | [diff] [blame] | 388 | |
Pravin B Shelar | 59b93b4 | 2014-11-05 15:27:48 -0800 | [diff] [blame] | 389 | SKB_GSO_TUNNEL_REMCSUM = 1 << 12, |
Herbert Xu | 7967168 | 2006-06-22 02:40:14 -0700 | [diff] [blame] | 390 | }; |
| 391 | |
Arnaldo Carvalho de Melo | 2e07fa9 | 2007-04-10 21:22:35 -0700 | [diff] [blame] | 392 | #if BITS_PER_LONG > 32 |
| 393 | #define NET_SKBUFF_DATA_USES_OFFSET 1 |
| 394 | #endif |
| 395 | |
| 396 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
| 397 | typedef unsigned int sk_buff_data_t; |
| 398 | #else |
| 399 | typedef unsigned char *sk_buff_data_t; |
| 400 | #endif |
| 401 | |
Eric Dumazet | 363ec3923 | 2014-02-26 14:02:11 -0800 | [diff] [blame] | 402 | /** |
| 403 | * struct skb_mstamp - multi resolution time stamps |
| 404 | * @stamp_us: timestamp in us resolution |
| 405 | * @stamp_jiffies: timestamp in jiffies |
| 406 | */ |
| 407 | struct skb_mstamp { |
| 408 | union { |
| 409 | u64 v64; |
| 410 | struct { |
| 411 | u32 stamp_us; |
| 412 | u32 stamp_jiffies; |
| 413 | }; |
| 414 | }; |
| 415 | }; |
| 416 | |
| 417 | /** |
| 418 | * skb_mstamp_get - get current timestamp |
| 419 | * @cl: place to store timestamps |
| 420 | */ |
| 421 | static inline void skb_mstamp_get(struct skb_mstamp *cl) |
| 422 | { |
| 423 | u64 val = local_clock(); |
| 424 | |
| 425 | do_div(val, NSEC_PER_USEC); |
| 426 | cl->stamp_us = (u32)val; |
| 427 | cl->stamp_jiffies = (u32)jiffies; |
| 428 | } |
| 429 | |
| 430 | /** |
| 431 | * skb_mstamp_delta - compute the difference in usec between two skb_mstamp |
| 432 | * @t1: pointer to newest sample |
| 433 | * @t0: pointer to oldest sample |
| 434 | */ |
| 435 | static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1, |
| 436 | const struct skb_mstamp *t0) |
| 437 | { |
| 438 | s32 delta_us = t1->stamp_us - t0->stamp_us; |
| 439 | u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies; |
| 440 | |
| 441 | /* If delta_us is negative, this might be because interval is too big, |
| 442 | * or local_clock() drift is too big : fallback using jiffies. |
| 443 | */ |
| 444 | if (delta_us <= 0 || |
| 445 | delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ))) |
| 446 | |
| 447 | delta_us = jiffies_to_usecs(delta_jiffies); |
| 448 | |
| 449 | return delta_us; |
| 450 | } |
| 451 | |
| 452 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 453 | /** |
| 454 | * struct sk_buff - socket buffer |
| 455 | * @next: Next buffer in list |
| 456 | * @prev: Previous buffer in list |
Eric Dumazet | 363ec3923 | 2014-02-26 14:02:11 -0800 | [diff] [blame] | 457 | * @tstamp: Time we arrived/left |
Eric Dumazet | 56b1742 | 2014-11-03 08:19:53 -0800 | [diff] [blame] | 458 | * @rbnode: RB tree node, alternative to next/prev for netem/tcp |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 459 | * @sk: Socket we are owned by |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 460 | * @dev: Device we arrived on/are leaving by |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 461 | * @cb: Control buffer. Free for use by every layer. Put private vars here |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 462 | * @_skb_refdst: destination entry (with norefcount bit) |
Martin Waitz | 67be2dd | 2005-05-01 08:59:26 -0700 | [diff] [blame] | 463 | * @sp: the security path, used for xfrm |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 464 | * @len: Length of actual data |
| 465 | * @data_len: Data length |
| 466 | * @mac_len: Length of link layer header |
Patrick McHardy | 334a813 | 2007-06-25 04:35:20 -0700 | [diff] [blame] | 467 | * @hdr_len: writable header length of cloned skb |
Herbert Xu | 663ead3 | 2007-04-09 11:59:07 -0700 | [diff] [blame] | 468 | * @csum: Checksum (must include start/offset pair) |
| 469 | * @csum_start: Offset from skb->head where checksumming should start |
| 470 | * @csum_offset: Offset from csum_start where checksum should be stored |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 471 | * @priority: Packet queueing priority |
WANG Cong | 60ff746 | 2014-05-04 16:39:18 -0700 | [diff] [blame] | 472 | * @ignore_df: allow local fragmentation |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 473 | * @cloned: Head may be cloned (check refcnt to be sure) |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 474 | * @ip_summed: Driver fed us an IP checksum |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 475 | * @nohdr: Payload reference only, must not modify header |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 476 | * @nfctinfo: Relationship of this skb to the connection |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 477 | * @pkt_type: Packet class |
Randy Dunlap | c83c248 | 2005-10-18 22:07:41 -0700 | [diff] [blame] | 478 | * @fclone: skbuff clone status |
Randy Dunlap | c83c248 | 2005-10-18 22:07:41 -0700 | [diff] [blame] | 479 | * @ipvs_property: skbuff is owned by ipvs |
Randy Dunlap | 3172936 | 2008-02-18 20:52:13 -0800 | [diff] [blame] | 480 | * @peeked: this packet has been seen already, so stats have been |
| 481 | * done for it, don't do them again |
Jozsef Kadlecsik | ba9dda3 | 2007-07-07 22:21:23 -0700 | [diff] [blame] | 482 | * @nf_trace: netfilter packet trace flag |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 483 | * @protocol: Packet protocol from driver |
| 484 | * @destructor: Destruct function |
| 485 | * @nfct: Associated connection, if any |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 486 | * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c |
Eric Dumazet | 8964be4 | 2009-11-20 15:35:04 -0800 | [diff] [blame] | 487 | * @skb_iif: ifindex of device we arrived on |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 488 | * @tc_index: Traffic control index |
| 489 | * @tc_verd: traffic control verdict |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 490 | * @hash: the packet hash |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 491 | * @queue_mapping: Queue mapping for multiqueue devices |
David S. Miller | 0b725a2 | 2014-08-25 15:51:53 -0700 | [diff] [blame] | 492 | * @xmit_more: More SKBs are pending for this queue |
Randy Dunlap | 553a567 | 2008-04-20 10:51:01 -0700 | [diff] [blame] | 493 | * @ndisc_nodetype: router type (from link layer) |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 494 | * @ooo_okay: allow the mapping of a socket to a queue to be changed |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 495 | * @l4_hash: indicate hash is a canonical 4-tuple hash over transport |
Changli Gao | 4ca2462 | 2011-08-19 07:26:44 -0700 | [diff] [blame] | 496 | * ports. |
Tom Herbert | a3b18dd | 2014-07-01 21:33:17 -0700 | [diff] [blame] | 497 | * @sw_hash: indicates hash was computed in software stack |
Johannes Berg | 6e3e939 | 2011-11-09 10:15:42 +0100 | [diff] [blame] | 498 | * @wifi_acked_valid: wifi_acked was set |
| 499 | * @wifi_acked: whether frame was acked on wifi or not |
Ben Greear | 3bdc0eb | 2012-02-11 15:39:30 +0000 | [diff] [blame] | 500 | * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS |
Eliezer Tamir | 0602129 | 2013-06-10 11:39:50 +0300 | [diff] [blame] | 501 | * @napi_id: id of the NAPI struct this skb came from |
James Morris | 984bc16 | 2006-06-09 00:29:17 -0700 | [diff] [blame] | 502 | * @secmark: security marking |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 503 | * @mark: Generic packet mark |
Patrick McHardy | 86a9bad | 2013-04-19 02:04:30 +0000 | [diff] [blame] | 504 | * @vlan_proto: vlan encapsulation protocol |
Patrick McHardy | 6aa895b | 2008-07-14 22:49:06 -0700 | [diff] [blame] | 505 | * @vlan_tci: vlan tag control information |
Simon Horman | 0d89d20 | 2013-05-23 21:02:52 +0000 | [diff] [blame] | 506 | * @inner_protocol: Protocol (encapsulation) |
Joseph Gasparakis | 6a674e9 | 2012-12-07 14:14:14 +0000 | [diff] [blame] | 507 | * @inner_transport_header: Inner transport layer header (encapsulation) |
| 508 | * @inner_network_header: Network layer header (encapsulation) |
Pravin B Shelar | aefbd2b | 2013-03-07 13:21:46 +0000 | [diff] [blame] | 509 | * @inner_mac_header: Link layer header (encapsulation) |
Daniel Baluta | d84e0bd | 2011-07-10 07:04:04 -0700 | [diff] [blame] | 510 | * @transport_header: Transport layer header |
| 511 | * @network_header: Network layer header |
| 512 | * @mac_header: Link layer header |
| 513 | * @tail: Tail pointer |
| 514 | * @end: End pointer |
| 515 | * @head: Head of buffer |
| 516 | * @data: Data head pointer |
| 517 | * @truesize: Buffer size |
| 518 | * @users: User count - see {datagram,tcp}.c |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 519 | */ |
| 520 | |
| 521 | struct sk_buff { |
Eric Dumazet | 363ec3923 | 2014-02-26 14:02:11 -0800 | [diff] [blame] | 522 | union { |
Eric Dumazet | 56b1742 | 2014-11-03 08:19:53 -0800 | [diff] [blame] | 523 | struct { |
| 524 | /* These two members must be first. */ |
| 525 | struct sk_buff *next; |
| 526 | struct sk_buff *prev; |
Felix Fietkau | da3f5cf | 2010-02-23 11:45:51 +0000 | [diff] [blame] | 527 | |
Eric Dumazet | 56b1742 | 2014-11-03 08:19:53 -0800 | [diff] [blame] | 528 | union { |
| 529 | ktime_t tstamp; |
| 530 | struct skb_mstamp skb_mstamp; |
| 531 | }; |
| 532 | }; |
| 533 | struct rb_node rbnode; /* used in netem & tcp stack */ |
| 534 | }; |
Felix Fietkau | da3f5cf | 2010-02-23 11:45:51 +0000 | [diff] [blame] | 535 | struct sock *sk; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 536 | struct net_device *dev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 537 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 538 | /* |
| 539 | * This is the control buffer. It is free to use for every |
| 540 | * layer. Please put your private variables there. If you |
| 541 | * want to keep them across layers you have to do a skb_clone() |
| 542 | * first. This is owned by whoever has the skb queued ATM. |
| 543 | */ |
Felix Fietkau | da3f5cf | 2010-02-23 11:45:51 +0000 | [diff] [blame] | 544 | char cb[48] __aligned(8); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 545 | |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 546 | unsigned long _skb_refdst; |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 547 | void (*destructor)(struct sk_buff *skb); |
Felix Fietkau | da3f5cf | 2010-02-23 11:45:51 +0000 | [diff] [blame] | 548 | #ifdef CONFIG_XFRM |
| 549 | struct sec_path *sp; |
| 550 | #endif |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 551 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
| 552 | struct nf_conntrack *nfct; |
| 553 | #endif |
Pablo Neira Ayuso | 34666d4 | 2014-09-18 11:29:03 +0200 | [diff] [blame] | 554 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 555 | struct nf_bridge_info *nf_bridge; |
| 556 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 557 | unsigned int len, |
Patrick McHardy | 334a813 | 2007-06-25 04:35:20 -0700 | [diff] [blame] | 558 | data_len; |
| 559 | __u16 mac_len, |
| 560 | hdr_len; |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 561 | |
| 562 | /* Following fields are _not_ copied in __copy_skb_header() |
| 563 | * Note that queue_mapping is here mostly to fill a hole. |
| 564 | */ |
Vegard Nossum | fe55f6d | 2008-08-30 12:16:35 +0200 | [diff] [blame] | 565 | kmemcheck_bitfield_begin(flags1); |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 566 | __u16 queue_mapping; |
| 567 | __u8 cloned:1, |
Harald Welte | 6869c4d | 2005-08-09 19:24:19 -0700 | [diff] [blame] | 568 | nohdr:1, |
Patrick McHardy | b84f4cc | 2005-11-20 21:19:21 -0800 | [diff] [blame] | 569 | fclone:2, |
Herbert Xu | a59322b | 2007-12-05 01:53:40 -0800 | [diff] [blame] | 570 | peeked:1, |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 571 | head_frag:1, |
| 572 | xmit_more:1; |
| 573 | /* one bit hole */ |
Vegard Nossum | fe55f6d | 2008-08-30 12:16:35 +0200 | [diff] [blame] | 574 | kmemcheck_bitfield_end(flags1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 575 | |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 576 | /* fields enclosed in headers_start/headers_end are copied |
| 577 | * using a single memcpy() in __copy_skb_header() |
| 578 | */ |
Randy Dunlap | ebcf34f | 2014-10-26 19:14:06 -0700 | [diff] [blame] | 579 | /* private: */ |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 580 | __u32 headers_start[0]; |
Randy Dunlap | ebcf34f | 2014-10-26 19:14:06 -0700 | [diff] [blame] | 581 | /* public: */ |
Hannes Frederic Sowa | 233577a | 2014-09-12 14:04:43 +0200 | [diff] [blame] | 582 | |
| 583 | /* if you move pkt_type around you also must adapt those constants */ |
| 584 | #ifdef __BIG_ENDIAN_BITFIELD |
| 585 | #define PKT_TYPE_MAX (7 << 5) |
| 586 | #else |
| 587 | #define PKT_TYPE_MAX 7 |
KOVACS Krisztian | 2fc72c7 | 2011-01-12 20:25:08 +0100 | [diff] [blame] | 588 | #endif |
Hannes Frederic Sowa | 233577a | 2014-09-12 14:04:43 +0200 | [diff] [blame] | 589 | #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) |
| 590 | |
| 591 | __u8 __pkt_type_offset[0]; |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 592 | __u8 pkt_type:3; |
| 593 | __u8 pfmemalloc:1; |
| 594 | __u8 ignore_df:1; |
| 595 | __u8 nfctinfo:3; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 596 | |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 597 | __u8 nf_trace:1; |
| 598 | __u8 ip_summed:2; |
| 599 | __u8 ooo_okay:1; |
| 600 | __u8 l4_hash:1; |
| 601 | __u8 sw_hash:1; |
| 602 | __u8 wifi_acked_valid:1; |
| 603 | __u8 wifi_acked:1; |
| 604 | |
| 605 | __u8 no_fcs:1; |
| 606 | /* Indicates the inner headers are valid in the skbuff. */ |
| 607 | __u8 encapsulation:1; |
| 608 | __u8 encap_hdr_csum:1; |
| 609 | __u8 csum_valid:1; |
| 610 | __u8 csum_complete_sw:1; |
| 611 | __u8 csum_level:2; |
| 612 | __u8 csum_bad:1; |
| 613 | |
| 614 | #ifdef CONFIG_IPV6_NDISC_NODETYPE |
| 615 | __u8 ndisc_nodetype:2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 616 | #endif |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 617 | __u8 ipvs_property:1; |
Tom Herbert | 8bce6d7 | 2014-09-29 20:22:29 -0700 | [diff] [blame] | 618 | __u8 inner_protocol_type:1; |
Tom Herbert | e585f23 | 2014-11-04 09:06:54 -0800 | [diff] [blame] | 619 | __u8 remcsum_offload:1; |
| 620 | /* 3 or 5 bit hole */ |
Alexander Duyck | 4031ae6 | 2012-01-27 06:22:53 +0000 | [diff] [blame] | 621 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 622 | #ifdef CONFIG_NET_SCHED |
Patrick McHardy | b6b99eb | 2005-08-09 19:33:51 -0700 | [diff] [blame] | 623 | __u16 tc_index; /* traffic control index */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 624 | #ifdef CONFIG_NET_CLS_ACT |
Patrick McHardy | b6b99eb | 2005-08-09 19:33:51 -0700 | [diff] [blame] | 625 | __u16 tc_verd; /* traffic control verdict */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 626 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 627 | #endif |
Vegard Nossum | fe55f6d | 2008-08-30 12:16:35 +0200 | [diff] [blame] | 628 | |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 629 | union { |
| 630 | __wsum csum; |
| 631 | struct { |
| 632 | __u16 csum_start; |
| 633 | __u16 csum_offset; |
| 634 | }; |
| 635 | }; |
| 636 | __u32 priority; |
| 637 | int skb_iif; |
| 638 | __u32 hash; |
| 639 | __be16 vlan_proto; |
| 640 | __u16 vlan_tci; |
Eric Dumazet | 2bd8248 | 2015-02-03 23:48:24 -0800 | [diff] [blame] | 641 | #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) |
| 642 | union { |
| 643 | unsigned int napi_id; |
| 644 | unsigned int sender_cpu; |
| 645 | }; |
Chris Leech | 97fc2f0 | 2006-05-23 17:55:33 -0700 | [diff] [blame] | 646 | #endif |
James Morris | 984bc16 | 2006-06-09 00:29:17 -0700 | [diff] [blame] | 647 | #ifdef CONFIG_NETWORK_SECMARK |
| 648 | __u32 secmark; |
| 649 | #endif |
Neil Horman | 3b88578 | 2009-10-12 13:26:31 -0700 | [diff] [blame] | 650 | union { |
| 651 | __u32 mark; |
Eric Dumazet | 16fad69 | 2013-03-14 05:40:32 +0000 | [diff] [blame] | 652 | __u32 reserved_tailroom; |
Neil Horman | 3b88578 | 2009-10-12 13:26:31 -0700 | [diff] [blame] | 653 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 654 | |
Tom Herbert | 8bce6d7 | 2014-09-29 20:22:29 -0700 | [diff] [blame] | 655 | union { |
| 656 | __be16 inner_protocol; |
| 657 | __u8 inner_ipproto; |
| 658 | }; |
| 659 | |
Simon Horman | 1a37e41 | 2013-05-23 21:02:51 +0000 | [diff] [blame] | 660 | __u16 inner_transport_header; |
| 661 | __u16 inner_network_header; |
| 662 | __u16 inner_mac_header; |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 663 | |
| 664 | __be16 protocol; |
Simon Horman | 1a37e41 | 2013-05-23 21:02:51 +0000 | [diff] [blame] | 665 | __u16 transport_header; |
| 666 | __u16 network_header; |
| 667 | __u16 mac_header; |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 668 | |
Randy Dunlap | ebcf34f | 2014-10-26 19:14:06 -0700 | [diff] [blame] | 669 | /* private: */ |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 670 | __u32 headers_end[0]; |
Randy Dunlap | ebcf34f | 2014-10-26 19:14:06 -0700 | [diff] [blame] | 671 | /* public: */ |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 672 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 673 | /* These elements must be at the end, see alloc_skb() for details. */ |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 674 | sk_buff_data_t tail; |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 675 | sk_buff_data_t end; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 676 | unsigned char *head, |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 677 | *data; |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 678 | unsigned int truesize; |
| 679 | atomic_t users; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 680 | }; |
| 681 | |
| 682 | #ifdef __KERNEL__ |
| 683 | /* |
| 684 | * Handling routines are only of interest to the kernel |
| 685 | */ |
| 686 | #include <linux/slab.h> |
| 687 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 688 | |
Mel Gorman | c93bdd0 | 2012-07-31 16:44:19 -0700 | [diff] [blame] | 689 | #define SKB_ALLOC_FCLONE 0x01 |
| 690 | #define SKB_ALLOC_RX 0x02 |
Alexander Duyck | fd11a83 | 2014-12-09 19:40:49 -0800 | [diff] [blame] | 691 | #define SKB_ALLOC_NAPI 0x04 |
Mel Gorman | c93bdd0 | 2012-07-31 16:44:19 -0700 | [diff] [blame] | 692 | |
| 693 | /* Returns true if the skb was allocated from PFMEMALLOC reserves */ |
| 694 | static inline bool skb_pfmemalloc(const struct sk_buff *skb) |
| 695 | { |
| 696 | return unlikely(skb->pfmemalloc); |
| 697 | } |
| 698 | |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 699 | /* |
| 700 | * skb might have a dst pointer attached, refcounted or not. |
| 701 | * _skb_refdst low order bit is set if refcount was _not_ taken |
| 702 | */ |
| 703 | #define SKB_DST_NOREF 1UL |
| 704 | #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) |
| 705 | |
| 706 | /** |
| 707 | * skb_dst - returns skb dst_entry |
| 708 | * @skb: buffer |
| 709 | * |
| 710 | * Returns skb dst_entry, regardless of reference taken or not. |
| 711 | */ |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 712 | static inline struct dst_entry *skb_dst(const struct sk_buff *skb) |
| 713 | { |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 714 | /* If refdst was not refcounted, check we still are in a |
| 715 | * rcu_read_lock section |
| 716 | */ |
| 717 | WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && |
| 718 | !rcu_read_lock_held() && |
| 719 | !rcu_read_lock_bh_held()); |
| 720 | return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 721 | } |
| 722 | |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 723 | /** |
| 724 | * skb_dst_set - sets skb dst |
| 725 | * @skb: buffer |
| 726 | * @dst: dst entry |
| 727 | * |
| 728 | * Sets skb dst, assuming a reference was taken on dst and should |
| 729 | * be released by skb_dst_drop() |
| 730 | */ |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 731 | static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) |
| 732 | { |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 733 | skb->_skb_refdst = (unsigned long)dst; |
| 734 | } |
| 735 | |
Julian Anastasov | 932bc4d | 2013-03-21 11:57:58 +0200 | [diff] [blame] | 736 | /** |
| 737 | * skb_dst_set_noref - sets skb dst, hopefully, without taking reference |
| 738 | * @skb: buffer |
| 739 | * @dst: dst entry |
| 740 | * |
| 741 | * Sets skb dst, assuming a reference was not taken on dst. |
| 742 | * If dst entry is cached, we do not take reference and dst_release |
| 743 | * will be avoided by refdst_drop. If dst entry is not cached, we take |
| 744 | * reference, so that last dst_release can destroy the dst immediately. |
| 745 | */ |
| 746 | static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) |
| 747 | { |
Hannes Frederic Sowa | dbfc4fb | 2014-12-06 19:19:42 +0100 | [diff] [blame] | 748 | WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); |
| 749 | skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; |
Julian Anastasov | 932bc4d | 2013-03-21 11:57:58 +0200 | [diff] [blame] | 750 | } |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 751 | |
| 752 | /** |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 753 | * skb_dst_is_noref - Test if skb dst isn't refcounted |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 754 | * @skb: buffer |
| 755 | */ |
| 756 | static inline bool skb_dst_is_noref(const struct sk_buff *skb) |
| 757 | { |
| 758 | return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 759 | } |
| 760 | |
Eric Dumazet | 511c3f9 | 2009-06-02 05:14:27 +0000 | [diff] [blame] | 761 | static inline struct rtable *skb_rtable(const struct sk_buff *skb) |
| 762 | { |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 763 | return (struct rtable *)skb_dst(skb); |
Eric Dumazet | 511c3f9 | 2009-06-02 05:14:27 +0000 | [diff] [blame] | 764 | } |
| 765 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 766 | void kfree_skb(struct sk_buff *skb); |
| 767 | void kfree_skb_list(struct sk_buff *segs); |
| 768 | void skb_tx_error(struct sk_buff *skb); |
| 769 | void consume_skb(struct sk_buff *skb); |
| 770 | void __kfree_skb(struct sk_buff *skb); |
Eric Dumazet | d7e8883 | 2012-04-30 08:10:34 +0000 | [diff] [blame] | 771 | extern struct kmem_cache *skbuff_head_cache; |
Eric Dumazet | bad43ca | 2012-05-19 03:02:02 +0000 | [diff] [blame] | 772 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 773 | void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); |
| 774 | bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, |
| 775 | bool *fragstolen, int *delta_truesize); |
Eric Dumazet | bad43ca | 2012-05-19 03:02:02 +0000 | [diff] [blame] | 776 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 777 | struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, |
| 778 | int node); |
Eric Dumazet | 2ea2f62 | 2015-04-24 16:05:01 -0700 | [diff] [blame] | 779 | struct sk_buff *__build_skb(void *data, unsigned int frag_size); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 780 | struct sk_buff *build_skb(void *data, unsigned int frag_size); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 781 | static inline struct sk_buff *alloc_skb(unsigned int size, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 782 | gfp_t priority) |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 783 | { |
Eric Dumazet | 564824b | 2010-10-11 19:05:25 +0000 | [diff] [blame] | 784 | return __alloc_skb(size, priority, 0, NUMA_NO_NODE); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 785 | } |
| 786 | |
Eric Dumazet | 2e4e441 | 2014-09-17 04:49:49 -0700 | [diff] [blame] | 787 | struct sk_buff *alloc_skb_with_frags(unsigned long header_len, |
| 788 | unsigned long data_len, |
| 789 | int max_page_order, |
| 790 | int *errcode, |
| 791 | gfp_t gfp_mask); |
| 792 | |
Eric Dumazet | d0bf4a9 | 2014-09-29 13:29:15 -0700 | [diff] [blame] | 793 | /* Layout of fast clones : [skb1][skb2][fclone_ref] */ |
| 794 | struct sk_buff_fclones { |
| 795 | struct sk_buff skb1; |
| 796 | |
| 797 | struct sk_buff skb2; |
| 798 | |
| 799 | atomic_t fclone_ref; |
| 800 | }; |
| 801 | |
| 802 | /** |
| 803 | * skb_fclone_busy - check if fclone is busy |
| 804 | * @skb: buffer |
| 805 | * |
| 806 | * Returns true is skb is a fast clone, and its clone is not freed. |
Eric Dumazet | 39bb5e6 | 2014-10-30 10:32:34 -0700 | [diff] [blame] | 807 | * Some drivers call skb_orphan() in their ndo_start_xmit(), |
| 808 | * so we also check that this didnt happen. |
Eric Dumazet | d0bf4a9 | 2014-09-29 13:29:15 -0700 | [diff] [blame] | 809 | */ |
Eric Dumazet | 39bb5e6 | 2014-10-30 10:32:34 -0700 | [diff] [blame] | 810 | static inline bool skb_fclone_busy(const struct sock *sk, |
| 811 | const struct sk_buff *skb) |
Eric Dumazet | d0bf4a9 | 2014-09-29 13:29:15 -0700 | [diff] [blame] | 812 | { |
| 813 | const struct sk_buff_fclones *fclones; |
| 814 | |
| 815 | fclones = container_of(skb, struct sk_buff_fclones, skb1); |
| 816 | |
| 817 | return skb->fclone == SKB_FCLONE_ORIG && |
Eric Dumazet | 6ffe75eb | 2014-12-03 17:04:39 -0800 | [diff] [blame] | 818 | atomic_read(&fclones->fclone_ref) > 1 && |
Eric Dumazet | 39bb5e6 | 2014-10-30 10:32:34 -0700 | [diff] [blame] | 819 | fclones->skb2.sk == sk; |
Eric Dumazet | d0bf4a9 | 2014-09-29 13:29:15 -0700 | [diff] [blame] | 820 | } |
| 821 | |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 822 | static inline struct sk_buff *alloc_skb_fclone(unsigned int size, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 823 | gfp_t priority) |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 824 | { |
Mel Gorman | c93bdd0 | 2012-07-31 16:44:19 -0700 | [diff] [blame] | 825 | return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); |
David S. Miller | d179cd1 | 2005-08-17 14:57:30 -0700 | [diff] [blame] | 826 | } |
| 827 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 828 | struct sk_buff *__alloc_skb_head(gfp_t priority, int node); |
Patrick McHardy | 0ebd0ac | 2013-04-17 06:46:58 +0000 | [diff] [blame] | 829 | static inline struct sk_buff *alloc_skb_head(gfp_t priority) |
| 830 | { |
| 831 | return __alloc_skb_head(priority, -1); |
| 832 | } |
| 833 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 834 | struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); |
| 835 | int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); |
| 836 | struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); |
| 837 | struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); |
Octavian Purdila | bad93e9 | 2014-06-12 01:36:26 +0300 | [diff] [blame] | 838 | struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, |
| 839 | gfp_t gfp_mask, bool fclone); |
| 840 | static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, |
| 841 | gfp_t gfp_mask) |
| 842 | { |
| 843 | return __pskb_copy_fclone(skb, headroom, gfp_mask, false); |
| 844 | } |
Eric Dumazet | 117632e | 2011-12-03 21:39:53 +0000 | [diff] [blame] | 845 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 846 | int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); |
| 847 | struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, |
| 848 | unsigned int headroom); |
| 849 | struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, |
| 850 | int newtailroom, gfp_t priority); |
Fan Du | 25a91d8 | 2014-01-18 09:54:23 +0800 | [diff] [blame] | 851 | int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, |
| 852 | int offset, int len); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 853 | int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset, |
| 854 | int len); |
| 855 | int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); |
| 856 | int skb_pad(struct sk_buff *skb, int pad); |
Neil Horman | ead2ceb | 2009-03-11 09:49:55 +0000 | [diff] [blame] | 857 | #define dev_kfree_skb(a) consume_skb(a) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 858 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 859 | int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, |
| 860 | int getfrag(void *from, char *to, int offset, |
| 861 | int len, int odd, struct sk_buff *skb), |
| 862 | void *from, int length); |
Ananda Raju | e89e9cf | 2005-10-18 15:46:41 -0700 | [diff] [blame] | 863 | |
Hannes Frederic Sowa | be12a1f | 2015-05-21 16:59:58 +0200 | [diff] [blame^] | 864 | int skb_append_pagefrags(struct sk_buff *skb, struct page *page, |
| 865 | int offset, size_t size); |
| 866 | |
Eric Dumazet | d94d9fe | 2009-11-04 09:50:58 -0800 | [diff] [blame] | 867 | struct skb_seq_state { |
Thomas Graf | 677e90e | 2005-06-23 20:59:51 -0700 | [diff] [blame] | 868 | __u32 lower_offset; |
| 869 | __u32 upper_offset; |
| 870 | __u32 frag_idx; |
| 871 | __u32 stepped_offset; |
| 872 | struct sk_buff *root_skb; |
| 873 | struct sk_buff *cur_skb; |
| 874 | __u8 *frag_data; |
| 875 | }; |
| 876 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 877 | void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, |
| 878 | unsigned int to, struct skb_seq_state *st); |
| 879 | unsigned int skb_seq_read(unsigned int consumed, const u8 **data, |
| 880 | struct skb_seq_state *st); |
| 881 | void skb_abort_seq_read(struct skb_seq_state *st); |
Thomas Graf | 677e90e | 2005-06-23 20:59:51 -0700 | [diff] [blame] | 882 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 883 | unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, |
Bojan Prtvar | 059a244 | 2015-02-22 11:46:35 +0100 | [diff] [blame] | 884 | unsigned int to, struct ts_config *config); |
Thomas Graf | 3fc7e8a | 2005-06-23 21:00:17 -0700 | [diff] [blame] | 885 | |
Tom Herbert | 09323cc | 2013-12-15 22:16:19 -0800 | [diff] [blame] | 886 | /* |
| 887 | * Packet hash types specify the type of hash in skb_set_hash. |
| 888 | * |
| 889 | * Hash types refer to the protocol layer addresses which are used to |
| 890 | * construct a packet's hash. The hashes are used to differentiate or identify |
| 891 | * flows of the protocol layer for the hash type. Hash types are either |
| 892 | * layer-2 (L2), layer-3 (L3), or layer-4 (L4). |
| 893 | * |
| 894 | * Properties of hashes: |
| 895 | * |
| 896 | * 1) Two packets in different flows have different hash values |
| 897 | * 2) Two packets in the same flow should have the same hash value |
| 898 | * |
| 899 | * A hash at a higher layer is considered to be more specific. A driver should |
| 900 | * set the most specific hash possible. |
| 901 | * |
| 902 | * A driver cannot indicate a more specific hash than the layer at which a hash |
| 903 | * was computed. For instance an L3 hash cannot be set as an L4 hash. |
| 904 | * |
| 905 | * A driver may indicate a hash level which is less specific than the |
| 906 | * actual layer the hash was computed on. For instance, a hash computed |
| 907 | * at L4 may be considered an L3 hash. This should only be done if the |
| 908 | * driver can't unambiguously determine that the HW computed the hash at |
| 909 | * the higher layer. Note that the "should" in the second property above |
| 910 | * permits this. |
| 911 | */ |
| 912 | enum pkt_hash_types { |
| 913 | PKT_HASH_TYPE_NONE, /* Undefined type */ |
| 914 | PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ |
| 915 | PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ |
| 916 | PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ |
| 917 | }; |
| 918 | |
| 919 | static inline void |
| 920 | skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) |
| 921 | { |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 922 | skb->l4_hash = (type == PKT_HASH_TYPE_L4); |
Tom Herbert | a3b18dd | 2014-07-01 21:33:17 -0700 | [diff] [blame] | 923 | skb->sw_hash = 0; |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 924 | skb->hash = hash; |
Tom Herbert | 09323cc | 2013-12-15 22:16:19 -0800 | [diff] [blame] | 925 | } |
| 926 | |
Tom Herbert | 3958afa1b | 2013-12-15 22:12:06 -0800 | [diff] [blame] | 927 | static inline __u32 skb_get_hash(struct sk_buff *skb) |
Krishna Kumar | bfb564e | 2010-08-04 06:15:52 +0000 | [diff] [blame] | 928 | { |
Tom Herbert | a3b18dd | 2014-07-01 21:33:17 -0700 | [diff] [blame] | 929 | if (!skb->l4_hash && !skb->sw_hash) |
Tom Herbert | 3958afa1b | 2013-12-15 22:12:06 -0800 | [diff] [blame] | 930 | __skb_get_hash(skb); |
Krishna Kumar | bfb564e | 2010-08-04 06:15:52 +0000 | [diff] [blame] | 931 | |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 932 | return skb->hash; |
Krishna Kumar | bfb564e | 2010-08-04 06:15:52 +0000 | [diff] [blame] | 933 | } |
| 934 | |
Tom Herbert | 50fb799 | 2015-05-01 11:30:12 -0700 | [diff] [blame] | 935 | __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb); |
| 936 | |
Tom Herbert | 57bdf7f4 | 2014-01-15 08:57:54 -0800 | [diff] [blame] | 937 | static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) |
| 938 | { |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 939 | return skb->hash; |
Tom Herbert | 57bdf7f4 | 2014-01-15 08:57:54 -0800 | [diff] [blame] | 940 | } |
| 941 | |
Tom Herbert | 7539fad | 2013-12-15 22:12:18 -0800 | [diff] [blame] | 942 | static inline void skb_clear_hash(struct sk_buff *skb) |
| 943 | { |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 944 | skb->hash = 0; |
Tom Herbert | a3b18dd | 2014-07-01 21:33:17 -0700 | [diff] [blame] | 945 | skb->sw_hash = 0; |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 946 | skb->l4_hash = 0; |
Tom Herbert | 7539fad | 2013-12-15 22:12:18 -0800 | [diff] [blame] | 947 | } |
| 948 | |
| 949 | static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) |
| 950 | { |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 951 | if (!skb->l4_hash) |
Tom Herbert | 7539fad | 2013-12-15 22:12:18 -0800 | [diff] [blame] | 952 | skb_clear_hash(skb); |
| 953 | } |
| 954 | |
Tom Herbert | 3df7a74 | 2013-12-15 22:16:29 -0800 | [diff] [blame] | 955 | static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) |
| 956 | { |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 957 | to->hash = from->hash; |
Tom Herbert | a3b18dd | 2014-07-01 21:33:17 -0700 | [diff] [blame] | 958 | to->sw_hash = from->sw_hash; |
Tom Herbert | 61b905d | 2014-03-24 15:34:47 -0700 | [diff] [blame] | 959 | to->l4_hash = from->l4_hash; |
Tom Herbert | 3df7a74 | 2013-12-15 22:16:29 -0800 | [diff] [blame] | 960 | }; |
| 961 | |
Eric Dumazet | c29390c | 2015-03-11 18:42:02 -0700 | [diff] [blame] | 962 | static inline void skb_sender_cpu_clear(struct sk_buff *skb) |
| 963 | { |
| 964 | #ifdef CONFIG_XPS |
| 965 | skb->sender_cpu = 0; |
| 966 | #endif |
| 967 | } |
| 968 | |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 969 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
| 970 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) |
| 971 | { |
| 972 | return skb->head + skb->end; |
| 973 | } |
Alexander Duyck | ec47ea8 | 2012-05-04 14:26:56 +0000 | [diff] [blame] | 974 | |
| 975 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) |
| 976 | { |
| 977 | return skb->end; |
| 978 | } |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 979 | #else |
| 980 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) |
| 981 | { |
| 982 | return skb->end; |
| 983 | } |
Alexander Duyck | ec47ea8 | 2012-05-04 14:26:56 +0000 | [diff] [blame] | 984 | |
| 985 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) |
| 986 | { |
| 987 | return skb->end - skb->head; |
| 988 | } |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 989 | #endif |
| 990 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 991 | /* Internal */ |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 992 | #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 993 | |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 994 | static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) |
| 995 | { |
| 996 | return &skb_shinfo(skb)->hwtstamps; |
| 997 | } |
| 998 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 999 | /** |
| 1000 | * skb_queue_empty - check if a queue is empty |
| 1001 | * @list: queue head |
| 1002 | * |
| 1003 | * Returns true if the queue is empty, false otherwise. |
| 1004 | */ |
| 1005 | static inline int skb_queue_empty(const struct sk_buff_head *list) |
| 1006 | { |
Daniel Borkmann | fd44b93 | 2014-01-07 23:23:44 +0100 | [diff] [blame] | 1007 | return list->next == (const struct sk_buff *) list; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1008 | } |
| 1009 | |
| 1010 | /** |
David S. Miller | fc7ebb2 | 2008-09-23 00:34:07 -0700 | [diff] [blame] | 1011 | * skb_queue_is_last - check if skb is the last entry in the queue |
| 1012 | * @list: queue head |
| 1013 | * @skb: buffer |
| 1014 | * |
| 1015 | * Returns true if @skb is the last buffer on the list. |
| 1016 | */ |
| 1017 | static inline bool skb_queue_is_last(const struct sk_buff_head *list, |
| 1018 | const struct sk_buff *skb) |
| 1019 | { |
Daniel Borkmann | fd44b93 | 2014-01-07 23:23:44 +0100 | [diff] [blame] | 1020 | return skb->next == (const struct sk_buff *) list; |
David S. Miller | fc7ebb2 | 2008-09-23 00:34:07 -0700 | [diff] [blame] | 1021 | } |
| 1022 | |
| 1023 | /** |
Ilpo Järvinen | 832d11c | 2008-11-24 21:20:15 -0800 | [diff] [blame] | 1024 | * skb_queue_is_first - check if skb is the first entry in the queue |
| 1025 | * @list: queue head |
| 1026 | * @skb: buffer |
| 1027 | * |
| 1028 | * Returns true if @skb is the first buffer on the list. |
| 1029 | */ |
| 1030 | static inline bool skb_queue_is_first(const struct sk_buff_head *list, |
| 1031 | const struct sk_buff *skb) |
| 1032 | { |
Daniel Borkmann | fd44b93 | 2014-01-07 23:23:44 +0100 | [diff] [blame] | 1033 | return skb->prev == (const struct sk_buff *) list; |
Ilpo Järvinen | 832d11c | 2008-11-24 21:20:15 -0800 | [diff] [blame] | 1034 | } |
| 1035 | |
| 1036 | /** |
David S. Miller | 249c8b4 | 2008-09-23 00:44:42 -0700 | [diff] [blame] | 1037 | * skb_queue_next - return the next packet in the queue |
| 1038 | * @list: queue head |
| 1039 | * @skb: current buffer |
| 1040 | * |
| 1041 | * Return the next packet in @list after @skb. It is only valid to |
| 1042 | * call this if skb_queue_is_last() evaluates to false. |
| 1043 | */ |
| 1044 | static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, |
| 1045 | const struct sk_buff *skb) |
| 1046 | { |
| 1047 | /* This BUG_ON may seem severe, but if we just return then we |
| 1048 | * are going to dereference garbage. |
| 1049 | */ |
| 1050 | BUG_ON(skb_queue_is_last(list, skb)); |
| 1051 | return skb->next; |
| 1052 | } |
| 1053 | |
| 1054 | /** |
Ilpo Järvinen | 832d11c | 2008-11-24 21:20:15 -0800 | [diff] [blame] | 1055 | * skb_queue_prev - return the prev packet in the queue |
| 1056 | * @list: queue head |
| 1057 | * @skb: current buffer |
| 1058 | * |
| 1059 | * Return the prev packet in @list before @skb. It is only valid to |
| 1060 | * call this if skb_queue_is_first() evaluates to false. |
| 1061 | */ |
| 1062 | static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, |
| 1063 | const struct sk_buff *skb) |
| 1064 | { |
| 1065 | /* This BUG_ON may seem severe, but if we just return then we |
| 1066 | * are going to dereference garbage. |
| 1067 | */ |
| 1068 | BUG_ON(skb_queue_is_first(list, skb)); |
| 1069 | return skb->prev; |
| 1070 | } |
| 1071 | |
| 1072 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1073 | * skb_get - reference buffer |
| 1074 | * @skb: buffer to reference |
| 1075 | * |
| 1076 | * Makes another reference to a socket buffer and returns a pointer |
| 1077 | * to the buffer. |
| 1078 | */ |
| 1079 | static inline struct sk_buff *skb_get(struct sk_buff *skb) |
| 1080 | { |
| 1081 | atomic_inc(&skb->users); |
| 1082 | return skb; |
| 1083 | } |
| 1084 | |
| 1085 | /* |
| 1086 | * If users == 1, we are the only owner and are can avoid redundant |
| 1087 | * atomic change. |
| 1088 | */ |
| 1089 | |
| 1090 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1091 | * skb_cloned - is the buffer a clone |
| 1092 | * @skb: buffer to check |
| 1093 | * |
| 1094 | * Returns true if the buffer was generated with skb_clone() and is |
| 1095 | * one of multiple shared copies of the buffer. Cloned buffers are |
| 1096 | * shared data so must not be written to under normal circumstances. |
| 1097 | */ |
| 1098 | static inline int skb_cloned(const struct sk_buff *skb) |
| 1099 | { |
| 1100 | return skb->cloned && |
| 1101 | (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; |
| 1102 | } |
| 1103 | |
Pravin B Shelar | 14bbd6a | 2013-02-14 09:44:49 +0000 | [diff] [blame] | 1104 | static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) |
| 1105 | { |
| 1106 | might_sleep_if(pri & __GFP_WAIT); |
| 1107 | |
| 1108 | if (skb_cloned(skb)) |
| 1109 | return pskb_expand_head(skb, 0, 0, pri); |
| 1110 | |
| 1111 | return 0; |
| 1112 | } |
| 1113 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1114 | /** |
| 1115 | * skb_header_cloned - is the header a clone |
| 1116 | * @skb: buffer to check |
| 1117 | * |
| 1118 | * Returns true if modifying the header part of the buffer requires |
| 1119 | * the data to be copied. |
| 1120 | */ |
| 1121 | static inline int skb_header_cloned(const struct sk_buff *skb) |
| 1122 | { |
| 1123 | int dataref; |
| 1124 | |
| 1125 | if (!skb->cloned) |
| 1126 | return 0; |
| 1127 | |
| 1128 | dataref = atomic_read(&skb_shinfo(skb)->dataref); |
| 1129 | dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); |
| 1130 | return dataref != 1; |
| 1131 | } |
| 1132 | |
| 1133 | /** |
| 1134 | * skb_header_release - release reference to header |
| 1135 | * @skb: buffer to operate on |
| 1136 | * |
| 1137 | * Drop a reference to the header part of the buffer. This is done |
| 1138 | * by acquiring a payload reference. You must not read from the header |
| 1139 | * part of skb->data after this. |
Eric Dumazet | f4a775d | 2014-09-22 16:29:32 -0700 | [diff] [blame] | 1140 | * Note : Check if you can use __skb_header_release() instead. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1141 | */ |
| 1142 | static inline void skb_header_release(struct sk_buff *skb) |
| 1143 | { |
| 1144 | BUG_ON(skb->nohdr); |
| 1145 | skb->nohdr = 1; |
| 1146 | atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); |
| 1147 | } |
| 1148 | |
| 1149 | /** |
Eric Dumazet | f4a775d | 2014-09-22 16:29:32 -0700 | [diff] [blame] | 1150 | * __skb_header_release - release reference to header |
| 1151 | * @skb: buffer to operate on |
| 1152 | * |
| 1153 | * Variant of skb_header_release() assuming skb is private to caller. |
| 1154 | * We can avoid one atomic operation. |
| 1155 | */ |
| 1156 | static inline void __skb_header_release(struct sk_buff *skb) |
| 1157 | { |
| 1158 | skb->nohdr = 1; |
| 1159 | atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); |
| 1160 | } |
| 1161 | |
| 1162 | |
| 1163 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1164 | * skb_shared - is the buffer shared |
| 1165 | * @skb: buffer to check |
| 1166 | * |
| 1167 | * Returns true if more than one person has a reference to this |
| 1168 | * buffer. |
| 1169 | */ |
| 1170 | static inline int skb_shared(const struct sk_buff *skb) |
| 1171 | { |
| 1172 | return atomic_read(&skb->users) != 1; |
| 1173 | } |
| 1174 | |
| 1175 | /** |
| 1176 | * skb_share_check - check if buffer is shared and if so clone it |
| 1177 | * @skb: buffer to check |
| 1178 | * @pri: priority for memory allocation |
| 1179 | * |
| 1180 | * If the buffer is shared the buffer is cloned and the old copy |
| 1181 | * drops a reference. A new clone with a single reference is returned. |
| 1182 | * If the buffer is not shared the original buffer is returned. When |
| 1183 | * being called from interrupt status or with spinlocks held pri must |
| 1184 | * be GFP_ATOMIC. |
| 1185 | * |
| 1186 | * NULL is returned on a memory allocation failure. |
| 1187 | */ |
Eric Dumazet | 47061bc | 2012-08-03 20:54:15 +0000 | [diff] [blame] | 1188 | static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1189 | { |
| 1190 | might_sleep_if(pri & __GFP_WAIT); |
| 1191 | if (skb_shared(skb)) { |
| 1192 | struct sk_buff *nskb = skb_clone(skb, pri); |
Eric Dumazet | 47061bc | 2012-08-03 20:54:15 +0000 | [diff] [blame] | 1193 | |
| 1194 | if (likely(nskb)) |
| 1195 | consume_skb(skb); |
| 1196 | else |
| 1197 | kfree_skb(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1198 | skb = nskb; |
| 1199 | } |
| 1200 | return skb; |
| 1201 | } |
| 1202 | |
| 1203 | /* |
| 1204 | * Copy shared buffers into a new sk_buff. We effectively do COW on |
| 1205 | * packets to handle cases where we have a local reader and forward |
| 1206 | * and a couple of other messy ones. The normal one is tcpdumping |
| 1207 | * a packet thats being forwarded. |
| 1208 | */ |
| 1209 | |
| 1210 | /** |
| 1211 | * skb_unshare - make a copy of a shared buffer |
| 1212 | * @skb: buffer to check |
| 1213 | * @pri: priority for memory allocation |
| 1214 | * |
| 1215 | * If the socket buffer is a clone then this function creates a new |
| 1216 | * copy of the data, drops a reference count on the old copy and returns |
| 1217 | * the new copy with the reference count at 1. If the buffer is not a clone |
| 1218 | * the original buffer is returned. When called with a spinlock held or |
| 1219 | * from interrupt state @pri must be %GFP_ATOMIC |
| 1220 | * |
| 1221 | * %NULL is returned on a memory allocation failure. |
| 1222 | */ |
Victor Fusco | e2bf521 | 2005-07-18 13:36:38 -0700 | [diff] [blame] | 1223 | static inline struct sk_buff *skb_unshare(struct sk_buff *skb, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1224 | gfp_t pri) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1225 | { |
| 1226 | might_sleep_if(pri & __GFP_WAIT); |
| 1227 | if (skb_cloned(skb)) { |
| 1228 | struct sk_buff *nskb = skb_copy(skb, pri); |
Alexander Aring | 31eff81 | 2014-10-10 23:10:47 +0200 | [diff] [blame] | 1229 | |
| 1230 | /* Free our shared copy */ |
| 1231 | if (likely(nskb)) |
| 1232 | consume_skb(skb); |
| 1233 | else |
| 1234 | kfree_skb(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1235 | skb = nskb; |
| 1236 | } |
| 1237 | return skb; |
| 1238 | } |
| 1239 | |
| 1240 | /** |
Ben Hutchings | 1a5778a | 2010-02-14 22:35:47 -0800 | [diff] [blame] | 1241 | * skb_peek - peek at the head of an &sk_buff_head |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1242 | * @list_: list to peek at |
| 1243 | * |
| 1244 | * Peek an &sk_buff. Unlike most other operations you _MUST_ |
| 1245 | * be careful with this one. A peek leaves the buffer on the |
| 1246 | * list and someone else may run off with it. You must hold |
| 1247 | * the appropriate locks or have a private queue to do this. |
| 1248 | * |
| 1249 | * Returns %NULL for an empty list or a pointer to the head element. |
| 1250 | * The reference count is not incremented and the reference is therefore |
| 1251 | * volatile. Use with caution. |
| 1252 | */ |
Eric Dumazet | 05bdd2f | 2011-10-20 17:45:43 -0400 | [diff] [blame] | 1253 | static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1254 | { |
Eric Dumazet | 18d0700 | 2012-04-30 16:31:46 +0000 | [diff] [blame] | 1255 | struct sk_buff *skb = list_->next; |
| 1256 | |
| 1257 | if (skb == (struct sk_buff *)list_) |
| 1258 | skb = NULL; |
| 1259 | return skb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1260 | } |
| 1261 | |
| 1262 | /** |
Pavel Emelyanov | da5ef6e | 2012-02-21 07:31:18 +0000 | [diff] [blame] | 1263 | * skb_peek_next - peek skb following the given one from a queue |
| 1264 | * @skb: skb to start from |
| 1265 | * @list_: list to peek at |
| 1266 | * |
| 1267 | * Returns %NULL when the end of the list is met or a pointer to the |
| 1268 | * next element. The reference count is not incremented and the |
| 1269 | * reference is therefore volatile. Use with caution. |
| 1270 | */ |
| 1271 | static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, |
| 1272 | const struct sk_buff_head *list_) |
| 1273 | { |
| 1274 | struct sk_buff *next = skb->next; |
Eric Dumazet | 18d0700 | 2012-04-30 16:31:46 +0000 | [diff] [blame] | 1275 | |
Pavel Emelyanov | da5ef6e | 2012-02-21 07:31:18 +0000 | [diff] [blame] | 1276 | if (next == (struct sk_buff *)list_) |
| 1277 | next = NULL; |
| 1278 | return next; |
| 1279 | } |
| 1280 | |
| 1281 | /** |
Ben Hutchings | 1a5778a | 2010-02-14 22:35:47 -0800 | [diff] [blame] | 1282 | * skb_peek_tail - peek at the tail of an &sk_buff_head |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1283 | * @list_: list to peek at |
| 1284 | * |
| 1285 | * Peek an &sk_buff. Unlike most other operations you _MUST_ |
| 1286 | * be careful with this one. A peek leaves the buffer on the |
| 1287 | * list and someone else may run off with it. You must hold |
| 1288 | * the appropriate locks or have a private queue to do this. |
| 1289 | * |
| 1290 | * Returns %NULL for an empty list or a pointer to the tail element. |
| 1291 | * The reference count is not incremented and the reference is therefore |
| 1292 | * volatile. Use with caution. |
| 1293 | */ |
Eric Dumazet | 05bdd2f | 2011-10-20 17:45:43 -0400 | [diff] [blame] | 1294 | static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1295 | { |
Eric Dumazet | 18d0700 | 2012-04-30 16:31:46 +0000 | [diff] [blame] | 1296 | struct sk_buff *skb = list_->prev; |
| 1297 | |
| 1298 | if (skb == (struct sk_buff *)list_) |
| 1299 | skb = NULL; |
| 1300 | return skb; |
| 1301 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1302 | } |
| 1303 | |
| 1304 | /** |
| 1305 | * skb_queue_len - get queue length |
| 1306 | * @list_: list to measure |
| 1307 | * |
| 1308 | * Return the length of an &sk_buff queue. |
| 1309 | */ |
| 1310 | static inline __u32 skb_queue_len(const struct sk_buff_head *list_) |
| 1311 | { |
| 1312 | return list_->qlen; |
| 1313 | } |
| 1314 | |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1315 | /** |
| 1316 | * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head |
| 1317 | * @list: queue to initialize |
| 1318 | * |
| 1319 | * This initializes only the list and queue length aspects of |
| 1320 | * an sk_buff_head object. This allows to initialize the list |
| 1321 | * aspects of an sk_buff_head without reinitializing things like |
| 1322 | * the spinlock. It can also be used for on-stack sk_buff_head |
| 1323 | * objects where the spinlock is known to not be used. |
| 1324 | */ |
| 1325 | static inline void __skb_queue_head_init(struct sk_buff_head *list) |
| 1326 | { |
| 1327 | list->prev = list->next = (struct sk_buff *)list; |
| 1328 | list->qlen = 0; |
| 1329 | } |
| 1330 | |
Arjan van de Ven | 76f10ad | 2006-08-02 14:06:55 -0700 | [diff] [blame] | 1331 | /* |
| 1332 | * This function creates a split out lock class for each invocation; |
| 1333 | * this is needed for now since a whole lot of users of the skb-queue |
| 1334 | * infrastructure in drivers have different locking usage (in hardirq) |
| 1335 | * than the networking core (in softirq only). In the long run either the |
| 1336 | * network layer or drivers should need annotation to consolidate the |
| 1337 | * main types of usage into 3 classes. |
| 1338 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1339 | static inline void skb_queue_head_init(struct sk_buff_head *list) |
| 1340 | { |
| 1341 | spin_lock_init(&list->lock); |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1342 | __skb_queue_head_init(list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1343 | } |
| 1344 | |
Pavel Emelianov | c2ecba7 | 2007-04-17 12:45:31 -0700 | [diff] [blame] | 1345 | static inline void skb_queue_head_init_class(struct sk_buff_head *list, |
| 1346 | struct lock_class_key *class) |
| 1347 | { |
| 1348 | skb_queue_head_init(list); |
| 1349 | lockdep_set_class(&list->lock, class); |
| 1350 | } |
| 1351 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1352 | /* |
Gerrit Renker | bf29927 | 2008-04-14 00:04:51 -0700 | [diff] [blame] | 1353 | * Insert an sk_buff on a list. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1354 | * |
| 1355 | * The "__skb_xxxx()" functions are the non-atomic ones that |
| 1356 | * can only be called with interrupts disabled. |
| 1357 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1358 | void skb_insert(struct sk_buff *old, struct sk_buff *newsk, |
| 1359 | struct sk_buff_head *list); |
Gerrit Renker | bf29927 | 2008-04-14 00:04:51 -0700 | [diff] [blame] | 1360 | static inline void __skb_insert(struct sk_buff *newsk, |
| 1361 | struct sk_buff *prev, struct sk_buff *next, |
| 1362 | struct sk_buff_head *list) |
| 1363 | { |
| 1364 | newsk->next = next; |
| 1365 | newsk->prev = prev; |
| 1366 | next->prev = prev->next = newsk; |
| 1367 | list->qlen++; |
| 1368 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1369 | |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1370 | static inline void __skb_queue_splice(const struct sk_buff_head *list, |
| 1371 | struct sk_buff *prev, |
| 1372 | struct sk_buff *next) |
| 1373 | { |
| 1374 | struct sk_buff *first = list->next; |
| 1375 | struct sk_buff *last = list->prev; |
| 1376 | |
| 1377 | first->prev = prev; |
| 1378 | prev->next = first; |
| 1379 | |
| 1380 | last->next = next; |
| 1381 | next->prev = last; |
| 1382 | } |
| 1383 | |
| 1384 | /** |
| 1385 | * skb_queue_splice - join two skb lists, this is designed for stacks |
| 1386 | * @list: the new list to add |
| 1387 | * @head: the place to add it in the first list |
| 1388 | */ |
| 1389 | static inline void skb_queue_splice(const struct sk_buff_head *list, |
| 1390 | struct sk_buff_head *head) |
| 1391 | { |
| 1392 | if (!skb_queue_empty(list)) { |
| 1393 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); |
David S. Miller | 1d4a31d | 2008-09-22 21:57:21 -0700 | [diff] [blame] | 1394 | head->qlen += list->qlen; |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1395 | } |
| 1396 | } |
| 1397 | |
| 1398 | /** |
Eric Dumazet | d96194966 | 2012-04-30 21:29:16 +0000 | [diff] [blame] | 1399 | * skb_queue_splice_init - join two skb lists and reinitialise the emptied list |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1400 | * @list: the new list to add |
| 1401 | * @head: the place to add it in the first list |
| 1402 | * |
| 1403 | * The list at @list is reinitialised |
| 1404 | */ |
| 1405 | static inline void skb_queue_splice_init(struct sk_buff_head *list, |
| 1406 | struct sk_buff_head *head) |
| 1407 | { |
| 1408 | if (!skb_queue_empty(list)) { |
| 1409 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); |
David S. Miller | 1d4a31d | 2008-09-22 21:57:21 -0700 | [diff] [blame] | 1410 | head->qlen += list->qlen; |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1411 | __skb_queue_head_init(list); |
| 1412 | } |
| 1413 | } |
| 1414 | |
| 1415 | /** |
| 1416 | * skb_queue_splice_tail - join two skb lists, each list being a queue |
| 1417 | * @list: the new list to add |
| 1418 | * @head: the place to add it in the first list |
| 1419 | */ |
| 1420 | static inline void skb_queue_splice_tail(const struct sk_buff_head *list, |
| 1421 | struct sk_buff_head *head) |
| 1422 | { |
| 1423 | if (!skb_queue_empty(list)) { |
| 1424 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); |
David S. Miller | 1d4a31d | 2008-09-22 21:57:21 -0700 | [diff] [blame] | 1425 | head->qlen += list->qlen; |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1426 | } |
| 1427 | } |
| 1428 | |
| 1429 | /** |
Eric Dumazet | d96194966 | 2012-04-30 21:29:16 +0000 | [diff] [blame] | 1430 | * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1431 | * @list: the new list to add |
| 1432 | * @head: the place to add it in the first list |
| 1433 | * |
| 1434 | * Each of the lists is a queue. |
| 1435 | * The list at @list is reinitialised |
| 1436 | */ |
| 1437 | static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, |
| 1438 | struct sk_buff_head *head) |
| 1439 | { |
| 1440 | if (!skb_queue_empty(list)) { |
| 1441 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); |
David S. Miller | 1d4a31d | 2008-09-22 21:57:21 -0700 | [diff] [blame] | 1442 | head->qlen += list->qlen; |
David S. Miller | 67fed45 | 2008-09-21 22:36:24 -0700 | [diff] [blame] | 1443 | __skb_queue_head_init(list); |
| 1444 | } |
| 1445 | } |
| 1446 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1447 | /** |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 1448 | * __skb_queue_after - queue a buffer at the list head |
| 1449 | * @list: list to use |
| 1450 | * @prev: place after this buffer |
| 1451 | * @newsk: buffer to queue |
| 1452 | * |
| 1453 | * Queue a buffer int the middle of a list. This function takes no locks |
| 1454 | * and you must therefore hold required locks before calling it. |
| 1455 | * |
| 1456 | * A buffer cannot be placed on two lists at the same time. |
| 1457 | */ |
| 1458 | static inline void __skb_queue_after(struct sk_buff_head *list, |
| 1459 | struct sk_buff *prev, |
| 1460 | struct sk_buff *newsk) |
| 1461 | { |
Gerrit Renker | bf29927 | 2008-04-14 00:04:51 -0700 | [diff] [blame] | 1462 | __skb_insert(newsk, prev, prev->next, list); |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 1463 | } |
| 1464 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1465 | void skb_append(struct sk_buff *old, struct sk_buff *newsk, |
| 1466 | struct sk_buff_head *list); |
Gerrit Renker | 7de6c03 | 2008-04-14 00:05:09 -0700 | [diff] [blame] | 1467 | |
Gerrit Renker | f557285 | 2008-04-14 00:05:28 -0700 | [diff] [blame] | 1468 | static inline void __skb_queue_before(struct sk_buff_head *list, |
| 1469 | struct sk_buff *next, |
| 1470 | struct sk_buff *newsk) |
| 1471 | { |
| 1472 | __skb_insert(newsk, next->prev, next, list); |
| 1473 | } |
| 1474 | |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 1475 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1476 | * __skb_queue_head - queue a buffer at the list head |
| 1477 | * @list: list to use |
| 1478 | * @newsk: buffer to queue |
| 1479 | * |
| 1480 | * Queue a buffer at the start of a list. This function takes no locks |
| 1481 | * and you must therefore hold required locks before calling it. |
| 1482 | * |
| 1483 | * A buffer cannot be placed on two lists at the same time. |
| 1484 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1485 | void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1486 | static inline void __skb_queue_head(struct sk_buff_head *list, |
| 1487 | struct sk_buff *newsk) |
| 1488 | { |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 1489 | __skb_queue_after(list, (struct sk_buff *)list, newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1490 | } |
| 1491 | |
| 1492 | /** |
| 1493 | * __skb_queue_tail - queue a buffer at the list tail |
| 1494 | * @list: list to use |
| 1495 | * @newsk: buffer to queue |
| 1496 | * |
| 1497 | * Queue a buffer at the end of a list. This function takes no locks |
| 1498 | * and you must therefore hold required locks before calling it. |
| 1499 | * |
| 1500 | * A buffer cannot be placed on two lists at the same time. |
| 1501 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1502 | void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1503 | static inline void __skb_queue_tail(struct sk_buff_head *list, |
| 1504 | struct sk_buff *newsk) |
| 1505 | { |
Gerrit Renker | f557285 | 2008-04-14 00:05:28 -0700 | [diff] [blame] | 1506 | __skb_queue_before(list, (struct sk_buff *)list, newsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1507 | } |
| 1508 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1509 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1510 | * remove sk_buff from list. _Must_ be called atomically, and with |
| 1511 | * the list known.. |
| 1512 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1513 | void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1514 | static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) |
| 1515 | { |
| 1516 | struct sk_buff *next, *prev; |
| 1517 | |
| 1518 | list->qlen--; |
| 1519 | next = skb->next; |
| 1520 | prev = skb->prev; |
| 1521 | skb->next = skb->prev = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1522 | next->prev = prev; |
| 1523 | prev->next = next; |
| 1524 | } |
| 1525 | |
Gerrit Renker | f525c06 | 2008-04-14 00:04:12 -0700 | [diff] [blame] | 1526 | /** |
| 1527 | * __skb_dequeue - remove from the head of the queue |
| 1528 | * @list: list to dequeue from |
| 1529 | * |
| 1530 | * Remove the head of the list. This function does not take any locks |
| 1531 | * so must be used with appropriate locks held only. The head item is |
| 1532 | * returned or %NULL if the list is empty. |
| 1533 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1534 | struct sk_buff *skb_dequeue(struct sk_buff_head *list); |
Gerrit Renker | f525c06 | 2008-04-14 00:04:12 -0700 | [diff] [blame] | 1535 | static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) |
| 1536 | { |
| 1537 | struct sk_buff *skb = skb_peek(list); |
| 1538 | if (skb) |
| 1539 | __skb_unlink(skb, list); |
| 1540 | return skb; |
| 1541 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1542 | |
| 1543 | /** |
| 1544 | * __skb_dequeue_tail - remove from the tail of the queue |
| 1545 | * @list: list to dequeue from |
| 1546 | * |
| 1547 | * Remove the tail of the list. This function does not take any locks |
| 1548 | * so must be used with appropriate locks held only. The tail item is |
| 1549 | * returned or %NULL if the list is empty. |
| 1550 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1551 | struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1552 | static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) |
| 1553 | { |
| 1554 | struct sk_buff *skb = skb_peek_tail(list); |
| 1555 | if (skb) |
| 1556 | __skb_unlink(skb, list); |
| 1557 | return skb; |
| 1558 | } |
| 1559 | |
| 1560 | |
David S. Miller | bdcc092 | 2012-03-07 20:53:36 -0500 | [diff] [blame] | 1561 | static inline bool skb_is_nonlinear(const struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1562 | { |
| 1563 | return skb->data_len; |
| 1564 | } |
| 1565 | |
| 1566 | static inline unsigned int skb_headlen(const struct sk_buff *skb) |
| 1567 | { |
| 1568 | return skb->len - skb->data_len; |
| 1569 | } |
| 1570 | |
| 1571 | static inline int skb_pagelen(const struct sk_buff *skb) |
| 1572 | { |
| 1573 | int i, len = 0; |
| 1574 | |
| 1575 | for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) |
Eric Dumazet | 9e903e0 | 2011-10-18 21:00:24 +0000 | [diff] [blame] | 1576 | len += skb_frag_size(&skb_shinfo(skb)->frags[i]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1577 | return len + skb_headlen(skb); |
| 1578 | } |
| 1579 | |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 1580 | /** |
| 1581 | * __skb_fill_page_desc - initialise a paged fragment in an skb |
| 1582 | * @skb: buffer containing fragment to be initialised |
| 1583 | * @i: paged fragment index to initialise |
| 1584 | * @page: the page to use for this fragment |
| 1585 | * @off: the offset to the data with @page |
| 1586 | * @size: the length of the data |
| 1587 | * |
| 1588 | * Initialises the @i'th fragment of @skb to point to &size bytes at |
| 1589 | * offset @off within @page. |
| 1590 | * |
| 1591 | * Does not take any additional reference on the fragment. |
| 1592 | */ |
| 1593 | static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, |
| 1594 | struct page *page, int off, int size) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1595 | { |
| 1596 | skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; |
| 1597 | |
Mel Gorman | c48a11c | 2012-07-31 16:44:23 -0700 | [diff] [blame] | 1598 | /* |
| 1599 | * Propagate page->pfmemalloc to the skb if we can. The problem is |
| 1600 | * that not all callers have unique ownership of the page. If |
| 1601 | * pfmemalloc is set, we check the mapping as a mapping implies |
| 1602 | * page->index is set (index and pfmemalloc share space). |
| 1603 | * If it's a valid mapping, we cannot use page->pfmemalloc but we |
| 1604 | * do not lose pfmemalloc information as the pages would not be |
| 1605 | * allocated using __GFP_MEMALLOC. |
| 1606 | */ |
Ian Campbell | a8605c6 | 2011-10-19 23:01:49 +0000 | [diff] [blame] | 1607 | frag->page.p = page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1608 | frag->page_offset = off; |
Eric Dumazet | 9e903e0 | 2011-10-18 21:00:24 +0000 | [diff] [blame] | 1609 | skb_frag_size_set(frag, size); |
Pavel Emelyanov | cca7af3 | 2013-03-14 03:29:40 +0000 | [diff] [blame] | 1610 | |
| 1611 | page = compound_head(page); |
| 1612 | if (page->pfmemalloc && !page->mapping) |
| 1613 | skb->pfmemalloc = true; |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 1614 | } |
| 1615 | |
| 1616 | /** |
| 1617 | * skb_fill_page_desc - initialise a paged fragment in an skb |
| 1618 | * @skb: buffer containing fragment to be initialised |
| 1619 | * @i: paged fragment index to initialise |
| 1620 | * @page: the page to use for this fragment |
| 1621 | * @off: the offset to the data with @page |
| 1622 | * @size: the length of the data |
| 1623 | * |
| 1624 | * As per __skb_fill_page_desc() -- initialises the @i'th fragment of |
Mathias Krause | bc32383 | 2013-11-07 14:18:26 +0100 | [diff] [blame] | 1625 | * @skb to point to @size bytes at offset @off within @page. In |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 1626 | * addition updates @skb such that @i is the last fragment. |
| 1627 | * |
| 1628 | * Does not take any additional reference on the fragment. |
| 1629 | */ |
| 1630 | static inline void skb_fill_page_desc(struct sk_buff *skb, int i, |
| 1631 | struct page *page, int off, int size) |
| 1632 | { |
| 1633 | __skb_fill_page_desc(skb, i, page, off, size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1634 | skb_shinfo(skb)->nr_frags = i + 1; |
| 1635 | } |
| 1636 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1637 | void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, |
| 1638 | int size, unsigned int truesize); |
Peter Zijlstra | 654bed1 | 2008-10-07 14:22:33 -0700 | [diff] [blame] | 1639 | |
Jason Wang | f8e617e | 2013-11-01 14:07:47 +0800 | [diff] [blame] | 1640 | void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, |
| 1641 | unsigned int truesize); |
| 1642 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1643 | #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) |
David S. Miller | 21dc330 | 2010-08-23 00:13:46 -0700 | [diff] [blame] | 1644 | #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1645 | #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) |
| 1646 | |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 1647 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
| 1648 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) |
| 1649 | { |
| 1650 | return skb->head + skb->tail; |
| 1651 | } |
| 1652 | |
| 1653 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) |
| 1654 | { |
| 1655 | skb->tail = skb->data - skb->head; |
| 1656 | } |
| 1657 | |
| 1658 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) |
| 1659 | { |
| 1660 | skb_reset_tail_pointer(skb); |
| 1661 | skb->tail += offset; |
| 1662 | } |
Simon Horman | 7cc4619 | 2013-05-28 20:34:29 +0000 | [diff] [blame] | 1663 | |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 1664 | #else /* NET_SKBUFF_DATA_USES_OFFSET */ |
| 1665 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) |
| 1666 | { |
| 1667 | return skb->tail; |
| 1668 | } |
| 1669 | |
| 1670 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) |
| 1671 | { |
| 1672 | skb->tail = skb->data; |
| 1673 | } |
| 1674 | |
| 1675 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) |
| 1676 | { |
| 1677 | skb->tail = skb->data + offset; |
| 1678 | } |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 1679 | |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 1680 | #endif /* NET_SKBUFF_DATA_USES_OFFSET */ |
| 1681 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1682 | /* |
| 1683 | * Add data to an sk_buff |
| 1684 | */ |
Mathias Krause | 0c7ddf3 | 2013-11-07 14:18:24 +0100 | [diff] [blame] | 1685 | unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1686 | unsigned char *skb_put(struct sk_buff *skb, unsigned int len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1687 | static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) |
| 1688 | { |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 1689 | unsigned char *tmp = skb_tail_pointer(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1690 | SKB_LINEAR_ASSERT(skb); |
| 1691 | skb->tail += len; |
| 1692 | skb->len += len; |
| 1693 | return tmp; |
| 1694 | } |
| 1695 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1696 | unsigned char *skb_push(struct sk_buff *skb, unsigned int len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1697 | static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) |
| 1698 | { |
| 1699 | skb->data -= len; |
| 1700 | skb->len += len; |
| 1701 | return skb->data; |
| 1702 | } |
| 1703 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1704 | unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1705 | static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) |
| 1706 | { |
| 1707 | skb->len -= len; |
| 1708 | BUG_ON(skb->len < skb->data_len); |
| 1709 | return skb->data += len; |
| 1710 | } |
| 1711 | |
David S. Miller | 47d2964 | 2010-05-02 02:21:44 -0700 | [diff] [blame] | 1712 | static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len) |
| 1713 | { |
| 1714 | return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); |
| 1715 | } |
| 1716 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 1717 | unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1718 | |
| 1719 | static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) |
| 1720 | { |
| 1721 | if (len > skb_headlen(skb) && |
Gerrit Renker | 987c402 | 2008-08-11 18:17:17 -0700 | [diff] [blame] | 1722 | !__pskb_pull_tail(skb, len - skb_headlen(skb))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1723 | return NULL; |
| 1724 | skb->len -= len; |
| 1725 | return skb->data += len; |
| 1726 | } |
| 1727 | |
| 1728 | static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) |
| 1729 | { |
| 1730 | return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); |
| 1731 | } |
| 1732 | |
| 1733 | static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) |
| 1734 | { |
| 1735 | if (likely(len <= skb_headlen(skb))) |
| 1736 | return 1; |
| 1737 | if (unlikely(len > skb->len)) |
| 1738 | return 0; |
Gerrit Renker | 987c402 | 2008-08-11 18:17:17 -0700 | [diff] [blame] | 1739 | return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1740 | } |
| 1741 | |
| 1742 | /** |
| 1743 | * skb_headroom - bytes at buffer head |
| 1744 | * @skb: buffer to check |
| 1745 | * |
| 1746 | * Return the number of bytes of free space at the head of an &sk_buff. |
| 1747 | */ |
Chuck Lever | c2636b4 | 2007-10-23 21:07:32 -0700 | [diff] [blame] | 1748 | static inline unsigned int skb_headroom(const struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1749 | { |
| 1750 | return skb->data - skb->head; |
| 1751 | } |
| 1752 | |
| 1753 | /** |
| 1754 | * skb_tailroom - bytes at buffer end |
| 1755 | * @skb: buffer to check |
| 1756 | * |
| 1757 | * Return the number of bytes of free space at the tail of an sk_buff |
| 1758 | */ |
| 1759 | static inline int skb_tailroom(const struct sk_buff *skb) |
| 1760 | { |
Arnaldo Carvalho de Melo | 4305b54 | 2007-04-19 20:43:29 -0700 | [diff] [blame] | 1761 | return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1762 | } |
| 1763 | |
| 1764 | /** |
Eric Dumazet | a21d457 | 2012-04-10 20:30:48 +0000 | [diff] [blame] | 1765 | * skb_availroom - bytes at buffer end |
| 1766 | * @skb: buffer to check |
| 1767 | * |
| 1768 | * Return the number of bytes of free space at the tail of an sk_buff |
| 1769 | * allocated by sk_stream_alloc() |
| 1770 | */ |
| 1771 | static inline int skb_availroom(const struct sk_buff *skb) |
| 1772 | { |
Eric Dumazet | 16fad69 | 2013-03-14 05:40:32 +0000 | [diff] [blame] | 1773 | if (skb_is_nonlinear(skb)) |
| 1774 | return 0; |
| 1775 | |
| 1776 | return skb->end - skb->tail - skb->reserved_tailroom; |
Eric Dumazet | a21d457 | 2012-04-10 20:30:48 +0000 | [diff] [blame] | 1777 | } |
| 1778 | |
| 1779 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1780 | * skb_reserve - adjust headroom |
| 1781 | * @skb: buffer to alter |
| 1782 | * @len: bytes to move |
| 1783 | * |
| 1784 | * Increase the headroom of an empty &sk_buff by reducing the tail |
| 1785 | * room. This is only allowed for an empty buffer. |
| 1786 | */ |
David S. Miller | 8243126c | 2006-01-17 02:54:21 -0800 | [diff] [blame] | 1787 | static inline void skb_reserve(struct sk_buff *skb, int len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1788 | { |
| 1789 | skb->data += len; |
| 1790 | skb->tail += len; |
| 1791 | } |
| 1792 | |
Tom Herbert | 8bce6d7 | 2014-09-29 20:22:29 -0700 | [diff] [blame] | 1793 | #define ENCAP_TYPE_ETHER 0 |
| 1794 | #define ENCAP_TYPE_IPPROTO 1 |
| 1795 | |
| 1796 | static inline void skb_set_inner_protocol(struct sk_buff *skb, |
| 1797 | __be16 protocol) |
| 1798 | { |
| 1799 | skb->inner_protocol = protocol; |
| 1800 | skb->inner_protocol_type = ENCAP_TYPE_ETHER; |
| 1801 | } |
| 1802 | |
| 1803 | static inline void skb_set_inner_ipproto(struct sk_buff *skb, |
| 1804 | __u8 ipproto) |
| 1805 | { |
| 1806 | skb->inner_ipproto = ipproto; |
| 1807 | skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; |
| 1808 | } |
| 1809 | |
Joseph Gasparakis | 6a674e9 | 2012-12-07 14:14:14 +0000 | [diff] [blame] | 1810 | static inline void skb_reset_inner_headers(struct sk_buff *skb) |
| 1811 | { |
Pravin B Shelar | aefbd2b | 2013-03-07 13:21:46 +0000 | [diff] [blame] | 1812 | skb->inner_mac_header = skb->mac_header; |
Joseph Gasparakis | 6a674e9 | 2012-12-07 14:14:14 +0000 | [diff] [blame] | 1813 | skb->inner_network_header = skb->network_header; |
| 1814 | skb->inner_transport_header = skb->transport_header; |
| 1815 | } |
| 1816 | |
Jiri Pirko | 0b5c9db | 2011-06-10 06:56:58 +0000 | [diff] [blame] | 1817 | static inline void skb_reset_mac_len(struct sk_buff *skb) |
| 1818 | { |
| 1819 | skb->mac_len = skb->network_header - skb->mac_header; |
| 1820 | } |
| 1821 | |
Joseph Gasparakis | 6a674e9 | 2012-12-07 14:14:14 +0000 | [diff] [blame] | 1822 | static inline unsigned char *skb_inner_transport_header(const struct sk_buff |
| 1823 | *skb) |
| 1824 | { |
| 1825 | return skb->head + skb->inner_transport_header; |
| 1826 | } |
| 1827 | |
| 1828 | static inline void skb_reset_inner_transport_header(struct sk_buff *skb) |
| 1829 | { |
| 1830 | skb->inner_transport_header = skb->data - skb->head; |
| 1831 | } |
| 1832 | |
| 1833 | static inline void skb_set_inner_transport_header(struct sk_buff *skb, |
| 1834 | const int offset) |
| 1835 | { |
| 1836 | skb_reset_inner_transport_header(skb); |
| 1837 | skb->inner_transport_header += offset; |
| 1838 | } |
| 1839 | |
| 1840 | static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) |
| 1841 | { |
| 1842 | return skb->head + skb->inner_network_header; |
| 1843 | } |
| 1844 | |
| 1845 | static inline void skb_reset_inner_network_header(struct sk_buff *skb) |
| 1846 | { |
| 1847 | skb->inner_network_header = skb->data - skb->head; |
| 1848 | } |
| 1849 | |
| 1850 | static inline void skb_set_inner_network_header(struct sk_buff *skb, |
| 1851 | const int offset) |
| 1852 | { |
| 1853 | skb_reset_inner_network_header(skb); |
| 1854 | skb->inner_network_header += offset; |
| 1855 | } |
| 1856 | |
Pravin B Shelar | aefbd2b | 2013-03-07 13:21:46 +0000 | [diff] [blame] | 1857 | static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) |
| 1858 | { |
| 1859 | return skb->head + skb->inner_mac_header; |
| 1860 | } |
| 1861 | |
| 1862 | static inline void skb_reset_inner_mac_header(struct sk_buff *skb) |
| 1863 | { |
| 1864 | skb->inner_mac_header = skb->data - skb->head; |
| 1865 | } |
| 1866 | |
| 1867 | static inline void skb_set_inner_mac_header(struct sk_buff *skb, |
| 1868 | const int offset) |
| 1869 | { |
| 1870 | skb_reset_inner_mac_header(skb); |
| 1871 | skb->inner_mac_header += offset; |
| 1872 | } |
Eric Dumazet | fda55ec | 2013-01-07 09:28:21 +0000 | [diff] [blame] | 1873 | static inline bool skb_transport_header_was_set(const struct sk_buff *skb) |
| 1874 | { |
Cong Wang | 35d0461 | 2013-05-29 15:16:05 +0800 | [diff] [blame] | 1875 | return skb->transport_header != (typeof(skb->transport_header))~0U; |
Eric Dumazet | fda55ec | 2013-01-07 09:28:21 +0000 | [diff] [blame] | 1876 | } |
| 1877 | |
Arnaldo Carvalho de Melo | 2e07fa9 | 2007-04-10 21:22:35 -0700 | [diff] [blame] | 1878 | static inline unsigned char *skb_transport_header(const struct sk_buff *skb) |
| 1879 | { |
| 1880 | return skb->head + skb->transport_header; |
| 1881 | } |
| 1882 | |
| 1883 | static inline void skb_reset_transport_header(struct sk_buff *skb) |
| 1884 | { |
| 1885 | skb->transport_header = skb->data - skb->head; |
| 1886 | } |
| 1887 | |
| 1888 | static inline void skb_set_transport_header(struct sk_buff *skb, |
| 1889 | const int offset) |
| 1890 | { |
| 1891 | skb_reset_transport_header(skb); |
| 1892 | skb->transport_header += offset; |
| 1893 | } |
| 1894 | |
| 1895 | static inline unsigned char *skb_network_header(const struct sk_buff *skb) |
| 1896 | { |
| 1897 | return skb->head + skb->network_header; |
| 1898 | } |
| 1899 | |
| 1900 | static inline void skb_reset_network_header(struct sk_buff *skb) |
| 1901 | { |
| 1902 | skb->network_header = skb->data - skb->head; |
| 1903 | } |
| 1904 | |
| 1905 | static inline void skb_set_network_header(struct sk_buff *skb, const int offset) |
| 1906 | { |
| 1907 | skb_reset_network_header(skb); |
| 1908 | skb->network_header += offset; |
| 1909 | } |
| 1910 | |
| 1911 | static inline unsigned char *skb_mac_header(const struct sk_buff *skb) |
| 1912 | { |
| 1913 | return skb->head + skb->mac_header; |
| 1914 | } |
| 1915 | |
| 1916 | static inline int skb_mac_header_was_set(const struct sk_buff *skb) |
| 1917 | { |
Cong Wang | 35d0461 | 2013-05-29 15:16:05 +0800 | [diff] [blame] | 1918 | return skb->mac_header != (typeof(skb->mac_header))~0U; |
Arnaldo Carvalho de Melo | 2e07fa9 | 2007-04-10 21:22:35 -0700 | [diff] [blame] | 1919 | } |
| 1920 | |
| 1921 | static inline void skb_reset_mac_header(struct sk_buff *skb) |
| 1922 | { |
| 1923 | skb->mac_header = skb->data - skb->head; |
| 1924 | } |
| 1925 | |
| 1926 | static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) |
| 1927 | { |
| 1928 | skb_reset_mac_header(skb); |
| 1929 | skb->mac_header += offset; |
| 1930 | } |
| 1931 | |
Timo Teräs | 0e3da5b | 2013-12-16 11:02:09 +0200 | [diff] [blame] | 1932 | static inline void skb_pop_mac_header(struct sk_buff *skb) |
| 1933 | { |
| 1934 | skb->mac_header = skb->network_header; |
| 1935 | } |
| 1936 | |
Ying Xue | fbbdb8f | 2013-03-27 16:46:06 +0000 | [diff] [blame] | 1937 | static inline void skb_probe_transport_header(struct sk_buff *skb, |
| 1938 | const int offset_hint) |
| 1939 | { |
| 1940 | struct flow_keys keys; |
| 1941 | |
| 1942 | if (skb_transport_header_was_set(skb)) |
| 1943 | return; |
Jiri Pirko | 06635a3 | 2015-05-12 14:56:16 +0200 | [diff] [blame] | 1944 | else if (skb_flow_dissect_flow_keys(skb, &keys)) |
| 1945 | skb_set_transport_header(skb, keys.basic.thoff); |
Ying Xue | fbbdb8f | 2013-03-27 16:46:06 +0000 | [diff] [blame] | 1946 | else |
| 1947 | skb_set_transport_header(skb, offset_hint); |
| 1948 | } |
| 1949 | |
Eric Dumazet | 0360689 | 2012-02-23 10:55:02 +0000 | [diff] [blame] | 1950 | static inline void skb_mac_header_rebuild(struct sk_buff *skb) |
| 1951 | { |
| 1952 | if (skb_mac_header_was_set(skb)) { |
| 1953 | const unsigned char *old_mac = skb_mac_header(skb); |
| 1954 | |
| 1955 | skb_set_mac_header(skb, -skb->mac_len); |
| 1956 | memmove(skb_mac_header(skb), old_mac, skb->mac_len); |
| 1957 | } |
| 1958 | } |
| 1959 | |
Michał Mirosław | 04fb451 | 2010-12-14 15:24:08 +0000 | [diff] [blame] | 1960 | static inline int skb_checksum_start_offset(const struct sk_buff *skb) |
| 1961 | { |
| 1962 | return skb->csum_start - skb_headroom(skb); |
| 1963 | } |
| 1964 | |
Arnaldo Carvalho de Melo | 2e07fa9 | 2007-04-10 21:22:35 -0700 | [diff] [blame] | 1965 | static inline int skb_transport_offset(const struct sk_buff *skb) |
| 1966 | { |
| 1967 | return skb_transport_header(skb) - skb->data; |
| 1968 | } |
| 1969 | |
| 1970 | static inline u32 skb_network_header_len(const struct sk_buff *skb) |
| 1971 | { |
| 1972 | return skb->transport_header - skb->network_header; |
| 1973 | } |
| 1974 | |
Joseph Gasparakis | 6a674e9 | 2012-12-07 14:14:14 +0000 | [diff] [blame] | 1975 | static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) |
| 1976 | { |
| 1977 | return skb->inner_transport_header - skb->inner_network_header; |
| 1978 | } |
| 1979 | |
Arnaldo Carvalho de Melo | 2e07fa9 | 2007-04-10 21:22:35 -0700 | [diff] [blame] | 1980 | static inline int skb_network_offset(const struct sk_buff *skb) |
| 1981 | { |
| 1982 | return skb_network_header(skb) - skb->data; |
| 1983 | } |
Arnaldo Carvalho de Melo | 48d49d0c | 2007-03-10 12:30:58 -0300 | [diff] [blame] | 1984 | |
Joseph Gasparakis | 6a674e9 | 2012-12-07 14:14:14 +0000 | [diff] [blame] | 1985 | static inline int skb_inner_network_offset(const struct sk_buff *skb) |
| 1986 | { |
| 1987 | return skb_inner_network_header(skb) - skb->data; |
| 1988 | } |
| 1989 | |
Changli Gao | f9599ce | 2010-08-04 04:43:44 +0000 | [diff] [blame] | 1990 | static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) |
| 1991 | { |
| 1992 | return pskb_may_pull(skb, skb_network_offset(skb) + len); |
| 1993 | } |
| 1994 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1995 | /* |
| 1996 | * CPUs often take a performance hit when accessing unaligned memory |
| 1997 | * locations. The actual performance hit varies, it can be small if the |
| 1998 | * hardware handles it or large if we have to take an exception and fix it |
| 1999 | * in software. |
| 2000 | * |
| 2001 | * Since an ethernet header is 14 bytes network drivers often end up with |
| 2002 | * the IP header at an unaligned offset. The IP header can be aligned by |
| 2003 | * shifting the start of the packet by 2 bytes. Drivers should do this |
| 2004 | * with: |
| 2005 | * |
Tobias Klauser | 8660c12 | 2009-07-13 22:48:16 +0000 | [diff] [blame] | 2006 | * skb_reserve(skb, NET_IP_ALIGN); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2007 | * |
| 2008 | * The downside to this alignment of the IP header is that the DMA is now |
| 2009 | * unaligned. On some architectures the cost of an unaligned DMA is high |
| 2010 | * and this cost outweighs the gains made by aligning the IP header. |
Tobias Klauser | 8660c12 | 2009-07-13 22:48:16 +0000 | [diff] [blame] | 2011 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2012 | * Since this trade off varies between architectures, we allow NET_IP_ALIGN |
| 2013 | * to be overridden. |
| 2014 | */ |
| 2015 | #ifndef NET_IP_ALIGN |
| 2016 | #define NET_IP_ALIGN 2 |
| 2017 | #endif |
| 2018 | |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 2019 | /* |
| 2020 | * The networking layer reserves some headroom in skb data (via |
| 2021 | * dev_alloc_skb). This is used to avoid having to reallocate skb data when |
| 2022 | * the header has to grow. In the default case, if the header has to grow |
David S. Miller | d6301d3 | 2009-02-08 19:24:13 -0800 | [diff] [blame] | 2023 | * 32 bytes or less we avoid the reallocation. |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 2024 | * |
| 2025 | * Unfortunately this headroom changes the DMA alignment of the resulting |
| 2026 | * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive |
| 2027 | * on some architectures. An architecture can override this value, |
| 2028 | * perhaps setting it to a cacheline in size (since that will maintain |
| 2029 | * cacheline alignment of the DMA). It must be a power of 2. |
| 2030 | * |
David S. Miller | d6301d3 | 2009-02-08 19:24:13 -0800 | [diff] [blame] | 2031 | * Various parts of the networking layer expect at least 32 bytes of |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 2032 | * headroom, you should not reduce this. |
Eric Dumazet | 5933dd2 | 2010-06-15 18:16:43 -0700 | [diff] [blame] | 2033 | * |
| 2034 | * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) |
| 2035 | * to reduce average number of cache lines per packet. |
| 2036 | * get_rps_cpus() for example only access one 64 bytes aligned block : |
Eric Dumazet | 18e8c13 | 2010-05-06 21:58:51 -0700 | [diff] [blame] | 2037 | * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 2038 | */ |
| 2039 | #ifndef NET_SKB_PAD |
Eric Dumazet | 5933dd2 | 2010-06-15 18:16:43 -0700 | [diff] [blame] | 2040 | #define NET_SKB_PAD max(32, L1_CACHE_BYTES) |
Anton Blanchard | 025be81 | 2006-03-31 02:27:06 -0800 | [diff] [blame] | 2041 | #endif |
| 2042 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2043 | int ___pskb_trim(struct sk_buff *skb, unsigned int len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2044 | |
| 2045 | static inline void __skb_trim(struct sk_buff *skb, unsigned int len) |
| 2046 | { |
Emmanuel Grumbach | c4264f2 | 2011-05-21 19:46:09 +0000 | [diff] [blame] | 2047 | if (unlikely(skb_is_nonlinear(skb))) { |
Herbert Xu | 3cc0e87 | 2006-06-09 16:13:38 -0700 | [diff] [blame] | 2048 | WARN_ON(1); |
| 2049 | return; |
| 2050 | } |
Arnaldo Carvalho de Melo | 27a884d | 2007-04-19 20:29:13 -0700 | [diff] [blame] | 2051 | skb->len = len; |
| 2052 | skb_set_tail_pointer(skb, len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2053 | } |
| 2054 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2055 | void skb_trim(struct sk_buff *skb, unsigned int len); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2056 | |
| 2057 | static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) |
| 2058 | { |
Herbert Xu | 3cc0e87 | 2006-06-09 16:13:38 -0700 | [diff] [blame] | 2059 | if (skb->data_len) |
| 2060 | return ___pskb_trim(skb, len); |
| 2061 | __skb_trim(skb, len); |
| 2062 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2063 | } |
| 2064 | |
| 2065 | static inline int pskb_trim(struct sk_buff *skb, unsigned int len) |
| 2066 | { |
| 2067 | return (len < skb->len) ? __pskb_trim(skb, len) : 0; |
| 2068 | } |
| 2069 | |
| 2070 | /** |
Herbert Xu | e9fa4f7 | 2006-08-13 20:12:58 -0700 | [diff] [blame] | 2071 | * pskb_trim_unique - remove end from a paged unique (not cloned) buffer |
| 2072 | * @skb: buffer to alter |
| 2073 | * @len: new length |
| 2074 | * |
| 2075 | * This is identical to pskb_trim except that the caller knows that |
| 2076 | * the skb is not cloned so we should never get an error due to out- |
| 2077 | * of-memory. |
| 2078 | */ |
| 2079 | static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) |
| 2080 | { |
| 2081 | int err = pskb_trim(skb, len); |
| 2082 | BUG_ON(err); |
| 2083 | } |
| 2084 | |
| 2085 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2086 | * skb_orphan - orphan a buffer |
| 2087 | * @skb: buffer to orphan |
| 2088 | * |
| 2089 | * If a buffer currently has an owner then we call the owner's |
| 2090 | * destructor function and make the @skb unowned. The buffer continues |
| 2091 | * to exist but is no longer charged to its former owner. |
| 2092 | */ |
| 2093 | static inline void skb_orphan(struct sk_buff *skb) |
| 2094 | { |
Eric Dumazet | c34a761 | 2013-07-30 16:11:15 -0700 | [diff] [blame] | 2095 | if (skb->destructor) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2096 | skb->destructor(skb); |
Eric Dumazet | c34a761 | 2013-07-30 16:11:15 -0700 | [diff] [blame] | 2097 | skb->destructor = NULL; |
| 2098 | skb->sk = NULL; |
Eric Dumazet | 376c731 | 2013-08-01 11:43:08 -0700 | [diff] [blame] | 2099 | } else { |
| 2100 | BUG_ON(skb->sk); |
Eric Dumazet | c34a761 | 2013-07-30 16:11:15 -0700 | [diff] [blame] | 2101 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2102 | } |
| 2103 | |
| 2104 | /** |
Michael S. Tsirkin | a353e0c | 2012-07-20 09:23:07 +0000 | [diff] [blame] | 2105 | * skb_orphan_frags - orphan the frags contained in a buffer |
| 2106 | * @skb: buffer to orphan frags from |
| 2107 | * @gfp_mask: allocation mask for replacement pages |
| 2108 | * |
| 2109 | * For each frag in the SKB which needs a destructor (i.e. has an |
| 2110 | * owner) create a copy of that frag and release the original |
| 2111 | * page by calling the destructor. |
| 2112 | */ |
| 2113 | static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) |
| 2114 | { |
| 2115 | if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY))) |
| 2116 | return 0; |
| 2117 | return skb_copy_ubufs(skb, gfp_mask); |
| 2118 | } |
| 2119 | |
| 2120 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2121 | * __skb_queue_purge - empty a list |
| 2122 | * @list: list to empty |
| 2123 | * |
| 2124 | * Delete all buffers on an &sk_buff list. Each buffer is removed from |
| 2125 | * the list and one reference dropped. This function does not take the |
| 2126 | * list lock and the caller must hold the relevant locks to use it. |
| 2127 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2128 | void skb_queue_purge(struct sk_buff_head *list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2129 | static inline void __skb_queue_purge(struct sk_buff_head *list) |
| 2130 | { |
| 2131 | struct sk_buff *skb; |
| 2132 | while ((skb = __skb_dequeue(list)) != NULL) |
| 2133 | kfree_skb(skb); |
| 2134 | } |
| 2135 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2136 | void *netdev_alloc_frag(unsigned int fragsz); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2137 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2138 | struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, |
| 2139 | gfp_t gfp_mask); |
Christoph Hellwig | 8af2745 | 2006-07-31 22:35:23 -0700 | [diff] [blame] | 2140 | |
| 2141 | /** |
| 2142 | * netdev_alloc_skb - allocate an skbuff for rx on a specific device |
| 2143 | * @dev: network device to receive on |
| 2144 | * @length: length to allocate |
| 2145 | * |
| 2146 | * Allocate a new &sk_buff and assign it a usage count of one. The |
| 2147 | * buffer has unspecified headroom built in. Users should allocate |
| 2148 | * the headroom they think they need without accounting for the |
| 2149 | * built in space. The built in space is used for optimisations. |
| 2150 | * |
| 2151 | * %NULL is returned if there is no free memory. Although this function |
| 2152 | * allocates memory it can be called from an interrupt. |
| 2153 | */ |
| 2154 | static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, |
Eric Dumazet | 6f53261 | 2012-05-18 05:12:12 +0000 | [diff] [blame] | 2155 | unsigned int length) |
Christoph Hellwig | 8af2745 | 2006-07-31 22:35:23 -0700 | [diff] [blame] | 2156 | { |
| 2157 | return __netdev_alloc_skb(dev, length, GFP_ATOMIC); |
| 2158 | } |
| 2159 | |
Eric Dumazet | 6f53261 | 2012-05-18 05:12:12 +0000 | [diff] [blame] | 2160 | /* legacy helper around __netdev_alloc_skb() */ |
| 2161 | static inline struct sk_buff *__dev_alloc_skb(unsigned int length, |
| 2162 | gfp_t gfp_mask) |
| 2163 | { |
| 2164 | return __netdev_alloc_skb(NULL, length, gfp_mask); |
| 2165 | } |
| 2166 | |
| 2167 | /* legacy helper around netdev_alloc_skb() */ |
| 2168 | static inline struct sk_buff *dev_alloc_skb(unsigned int length) |
| 2169 | { |
| 2170 | return netdev_alloc_skb(NULL, length); |
| 2171 | } |
| 2172 | |
| 2173 | |
Eric Dumazet | 4915a0d | 2011-07-11 20:08:34 -0700 | [diff] [blame] | 2174 | static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, |
| 2175 | unsigned int length, gfp_t gfp) |
Eric Dumazet | 61321bb | 2009-10-07 17:11:23 +0000 | [diff] [blame] | 2176 | { |
Eric Dumazet | 4915a0d | 2011-07-11 20:08:34 -0700 | [diff] [blame] | 2177 | struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); |
Eric Dumazet | 61321bb | 2009-10-07 17:11:23 +0000 | [diff] [blame] | 2178 | |
| 2179 | if (NET_IP_ALIGN && skb) |
| 2180 | skb_reserve(skb, NET_IP_ALIGN); |
| 2181 | return skb; |
| 2182 | } |
| 2183 | |
Eric Dumazet | 4915a0d | 2011-07-11 20:08:34 -0700 | [diff] [blame] | 2184 | static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, |
| 2185 | unsigned int length) |
| 2186 | { |
| 2187 | return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); |
| 2188 | } |
| 2189 | |
Alexander Duyck | 181edb2 | 2015-05-06 21:12:03 -0700 | [diff] [blame] | 2190 | static inline void skb_free_frag(void *addr) |
| 2191 | { |
| 2192 | __free_page_frag(addr); |
| 2193 | } |
| 2194 | |
Alexander Duyck | ffde732 | 2014-12-09 19:40:42 -0800 | [diff] [blame] | 2195 | void *napi_alloc_frag(unsigned int fragsz); |
Alexander Duyck | fd11a83 | 2014-12-09 19:40:49 -0800 | [diff] [blame] | 2196 | struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, |
| 2197 | unsigned int length, gfp_t gfp_mask); |
| 2198 | static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, |
| 2199 | unsigned int length) |
| 2200 | { |
| 2201 | return __napi_alloc_skb(napi, length, GFP_ATOMIC); |
| 2202 | } |
Alexander Duyck | ffde732 | 2014-12-09 19:40:42 -0800 | [diff] [blame] | 2203 | |
Florian Fainelli | bc6fc9f | 2013-08-30 15:36:14 +0100 | [diff] [blame] | 2204 | /** |
Alexander Duyck | 71dfda5 | 2014-11-11 09:26:34 -0800 | [diff] [blame] | 2205 | * __dev_alloc_pages - allocate page for network Rx |
| 2206 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx |
| 2207 | * @order: size of the allocation |
| 2208 | * |
| 2209 | * Allocate a new page. |
| 2210 | * |
| 2211 | * %NULL is returned if there is no free memory. |
| 2212 | */ |
| 2213 | static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, |
| 2214 | unsigned int order) |
| 2215 | { |
| 2216 | /* This piece of code contains several assumptions. |
| 2217 | * 1. This is for device Rx, therefor a cold page is preferred. |
| 2218 | * 2. The expectation is the user wants a compound page. |
| 2219 | * 3. If requesting a order 0 page it will not be compound |
| 2220 | * due to the check to see if order has a value in prep_new_page |
| 2221 | * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to |
| 2222 | * code in gfp_to_alloc_flags that should be enforcing this. |
| 2223 | */ |
| 2224 | gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC; |
| 2225 | |
| 2226 | return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); |
| 2227 | } |
| 2228 | |
| 2229 | static inline struct page *dev_alloc_pages(unsigned int order) |
| 2230 | { |
| 2231 | return __dev_alloc_pages(GFP_ATOMIC, order); |
| 2232 | } |
| 2233 | |
| 2234 | /** |
| 2235 | * __dev_alloc_page - allocate a page for network Rx |
| 2236 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx |
| 2237 | * |
| 2238 | * Allocate a new page. |
| 2239 | * |
| 2240 | * %NULL is returned if there is no free memory. |
| 2241 | */ |
| 2242 | static inline struct page *__dev_alloc_page(gfp_t gfp_mask) |
| 2243 | { |
| 2244 | return __dev_alloc_pages(gfp_mask, 0); |
| 2245 | } |
| 2246 | |
| 2247 | static inline struct page *dev_alloc_page(void) |
| 2248 | { |
| 2249 | return __dev_alloc_page(GFP_ATOMIC); |
| 2250 | } |
| 2251 | |
| 2252 | /** |
Mel Gorman | 0614002 | 2012-07-31 16:44:24 -0700 | [diff] [blame] | 2253 | * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page |
| 2254 | * @page: The page that was allocated from skb_alloc_page |
| 2255 | * @skb: The skb that may need pfmemalloc set |
| 2256 | */ |
| 2257 | static inline void skb_propagate_pfmemalloc(struct page *page, |
| 2258 | struct sk_buff *skb) |
| 2259 | { |
| 2260 | if (page && page->pfmemalloc) |
| 2261 | skb->pfmemalloc = true; |
| 2262 | } |
| 2263 | |
Eric Dumazet | 564824b | 2010-10-11 19:05:25 +0000 | [diff] [blame] | 2264 | /** |
Masanari Iida | e227867 | 2014-02-18 22:54:36 +0900 | [diff] [blame] | 2265 | * skb_frag_page - retrieve the page referred to by a paged fragment |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2266 | * @frag: the paged fragment |
| 2267 | * |
| 2268 | * Returns the &struct page associated with @frag. |
| 2269 | */ |
| 2270 | static inline struct page *skb_frag_page(const skb_frag_t *frag) |
| 2271 | { |
Ian Campbell | a8605c6 | 2011-10-19 23:01:49 +0000 | [diff] [blame] | 2272 | return frag->page.p; |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2273 | } |
| 2274 | |
| 2275 | /** |
| 2276 | * __skb_frag_ref - take an addition reference on a paged fragment. |
| 2277 | * @frag: the paged fragment |
| 2278 | * |
| 2279 | * Takes an additional reference on the paged fragment @frag. |
| 2280 | */ |
| 2281 | static inline void __skb_frag_ref(skb_frag_t *frag) |
| 2282 | { |
| 2283 | get_page(skb_frag_page(frag)); |
| 2284 | } |
| 2285 | |
| 2286 | /** |
| 2287 | * skb_frag_ref - take an addition reference on a paged fragment of an skb. |
| 2288 | * @skb: the buffer |
| 2289 | * @f: the fragment offset. |
| 2290 | * |
| 2291 | * Takes an additional reference on the @f'th paged fragment of @skb. |
| 2292 | */ |
| 2293 | static inline void skb_frag_ref(struct sk_buff *skb, int f) |
| 2294 | { |
| 2295 | __skb_frag_ref(&skb_shinfo(skb)->frags[f]); |
| 2296 | } |
| 2297 | |
| 2298 | /** |
| 2299 | * __skb_frag_unref - release a reference on a paged fragment. |
| 2300 | * @frag: the paged fragment |
| 2301 | * |
| 2302 | * Releases a reference on the paged fragment @frag. |
| 2303 | */ |
| 2304 | static inline void __skb_frag_unref(skb_frag_t *frag) |
| 2305 | { |
| 2306 | put_page(skb_frag_page(frag)); |
| 2307 | } |
| 2308 | |
| 2309 | /** |
| 2310 | * skb_frag_unref - release a reference on a paged fragment of an skb. |
| 2311 | * @skb: the buffer |
| 2312 | * @f: the fragment offset |
| 2313 | * |
| 2314 | * Releases a reference on the @f'th paged fragment of @skb. |
| 2315 | */ |
| 2316 | static inline void skb_frag_unref(struct sk_buff *skb, int f) |
| 2317 | { |
| 2318 | __skb_frag_unref(&skb_shinfo(skb)->frags[f]); |
| 2319 | } |
| 2320 | |
| 2321 | /** |
| 2322 | * skb_frag_address - gets the address of the data contained in a paged fragment |
| 2323 | * @frag: the paged fragment buffer |
| 2324 | * |
| 2325 | * Returns the address of the data within @frag. The page must already |
| 2326 | * be mapped. |
| 2327 | */ |
| 2328 | static inline void *skb_frag_address(const skb_frag_t *frag) |
| 2329 | { |
| 2330 | return page_address(skb_frag_page(frag)) + frag->page_offset; |
| 2331 | } |
| 2332 | |
| 2333 | /** |
| 2334 | * skb_frag_address_safe - gets the address of the data contained in a paged fragment |
| 2335 | * @frag: the paged fragment buffer |
| 2336 | * |
| 2337 | * Returns the address of the data within @frag. Checks that the page |
| 2338 | * is mapped and returns %NULL otherwise. |
| 2339 | */ |
| 2340 | static inline void *skb_frag_address_safe(const skb_frag_t *frag) |
| 2341 | { |
| 2342 | void *ptr = page_address(skb_frag_page(frag)); |
| 2343 | if (unlikely(!ptr)) |
| 2344 | return NULL; |
| 2345 | |
| 2346 | return ptr + frag->page_offset; |
| 2347 | } |
| 2348 | |
| 2349 | /** |
| 2350 | * __skb_frag_set_page - sets the page contained in a paged fragment |
| 2351 | * @frag: the paged fragment |
| 2352 | * @page: the page to set |
| 2353 | * |
| 2354 | * Sets the fragment @frag to contain @page. |
| 2355 | */ |
| 2356 | static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) |
| 2357 | { |
Ian Campbell | a8605c6 | 2011-10-19 23:01:49 +0000 | [diff] [blame] | 2358 | frag->page.p = page; |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2359 | } |
| 2360 | |
| 2361 | /** |
| 2362 | * skb_frag_set_page - sets the page contained in a paged fragment of an skb |
| 2363 | * @skb: the buffer |
| 2364 | * @f: the fragment offset |
| 2365 | * @page: the page to set |
| 2366 | * |
| 2367 | * Sets the @f'th fragment of @skb to contain @page. |
| 2368 | */ |
| 2369 | static inline void skb_frag_set_page(struct sk_buff *skb, int f, |
| 2370 | struct page *page) |
| 2371 | { |
| 2372 | __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); |
| 2373 | } |
| 2374 | |
Eric Dumazet | 400dfd3 | 2013-10-17 16:27:07 -0700 | [diff] [blame] | 2375 | bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); |
| 2376 | |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2377 | /** |
| 2378 | * skb_frag_dma_map - maps a paged fragment via the DMA API |
Marcos Paulo de Souza | f83347d | 2011-10-31 15:11:45 +0000 | [diff] [blame] | 2379 | * @dev: the device to map the fragment to |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2380 | * @frag: the paged fragment to map |
| 2381 | * @offset: the offset within the fragment (starting at the |
| 2382 | * fragment's own offset) |
| 2383 | * @size: the number of bytes to map |
Marcos Paulo de Souza | f83347d | 2011-10-31 15:11:45 +0000 | [diff] [blame] | 2384 | * @dir: the direction of the mapping (%PCI_DMA_*) |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2385 | * |
| 2386 | * Maps the page associated with @frag to @device. |
| 2387 | */ |
| 2388 | static inline dma_addr_t skb_frag_dma_map(struct device *dev, |
| 2389 | const skb_frag_t *frag, |
| 2390 | size_t offset, size_t size, |
| 2391 | enum dma_data_direction dir) |
| 2392 | { |
| 2393 | return dma_map_page(dev, skb_frag_page(frag), |
| 2394 | frag->page_offset + offset, size, dir); |
| 2395 | } |
| 2396 | |
Eric Dumazet | 117632e | 2011-12-03 21:39:53 +0000 | [diff] [blame] | 2397 | static inline struct sk_buff *pskb_copy(struct sk_buff *skb, |
| 2398 | gfp_t gfp_mask) |
| 2399 | { |
| 2400 | return __pskb_copy(skb, skb_headroom(skb), gfp_mask); |
| 2401 | } |
| 2402 | |
Octavian Purdila | bad93e9 | 2014-06-12 01:36:26 +0300 | [diff] [blame] | 2403 | |
| 2404 | static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, |
| 2405 | gfp_t gfp_mask) |
| 2406 | { |
| 2407 | return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); |
| 2408 | } |
| 2409 | |
| 2410 | |
Ian Campbell | 131ea66 | 2011-08-19 06:25:00 +0000 | [diff] [blame] | 2411 | /** |
Patrick McHardy | 334a813 | 2007-06-25 04:35:20 -0700 | [diff] [blame] | 2412 | * skb_clone_writable - is the header of a clone writable |
| 2413 | * @skb: buffer to check |
| 2414 | * @len: length up to which to write |
| 2415 | * |
| 2416 | * Returns true if modifying the header part of the cloned buffer |
| 2417 | * does not requires the data to be copied. |
| 2418 | */ |
Eric Dumazet | 05bdd2f | 2011-10-20 17:45:43 -0400 | [diff] [blame] | 2419 | static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) |
Patrick McHardy | 334a813 | 2007-06-25 04:35:20 -0700 | [diff] [blame] | 2420 | { |
| 2421 | return !skb_header_cloned(skb) && |
| 2422 | skb_headroom(skb) + len <= skb->hdr_len; |
| 2423 | } |
| 2424 | |
Herbert Xu | d9cc204 | 2007-09-16 16:21:16 -0700 | [diff] [blame] | 2425 | static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, |
| 2426 | int cloned) |
| 2427 | { |
| 2428 | int delta = 0; |
| 2429 | |
Herbert Xu | d9cc204 | 2007-09-16 16:21:16 -0700 | [diff] [blame] | 2430 | if (headroom > skb_headroom(skb)) |
| 2431 | delta = headroom - skb_headroom(skb); |
| 2432 | |
| 2433 | if (delta || cloned) |
| 2434 | return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, |
| 2435 | GFP_ATOMIC); |
| 2436 | return 0; |
| 2437 | } |
| 2438 | |
Patrick McHardy | 334a813 | 2007-06-25 04:35:20 -0700 | [diff] [blame] | 2439 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2440 | * skb_cow - copy header of skb when it is required |
| 2441 | * @skb: buffer to cow |
| 2442 | * @headroom: needed headroom |
| 2443 | * |
| 2444 | * If the skb passed lacks sufficient headroom or its data part |
| 2445 | * is shared, data is reallocated. If reallocation fails, an error |
| 2446 | * is returned and original skb is not changed. |
| 2447 | * |
| 2448 | * The result is skb with writable area skb->head...skb->tail |
| 2449 | * and at least @headroom of space at head. |
| 2450 | */ |
| 2451 | static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) |
| 2452 | { |
Herbert Xu | d9cc204 | 2007-09-16 16:21:16 -0700 | [diff] [blame] | 2453 | return __skb_cow(skb, headroom, skb_cloned(skb)); |
| 2454 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2455 | |
Herbert Xu | d9cc204 | 2007-09-16 16:21:16 -0700 | [diff] [blame] | 2456 | /** |
| 2457 | * skb_cow_head - skb_cow but only making the head writable |
| 2458 | * @skb: buffer to cow |
| 2459 | * @headroom: needed headroom |
| 2460 | * |
| 2461 | * This function is identical to skb_cow except that we replace the |
| 2462 | * skb_cloned check by skb_header_cloned. It should be used when |
| 2463 | * you only need to push on some header and do not need to modify |
| 2464 | * the data. |
| 2465 | */ |
| 2466 | static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) |
| 2467 | { |
| 2468 | return __skb_cow(skb, headroom, skb_header_cloned(skb)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2469 | } |
| 2470 | |
| 2471 | /** |
| 2472 | * skb_padto - pad an skbuff up to a minimal size |
| 2473 | * @skb: buffer to pad |
| 2474 | * @len: minimal length |
| 2475 | * |
| 2476 | * Pads up a buffer to ensure the trailing bytes exist and are |
| 2477 | * blanked. If the buffer already contains sufficient data it |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 2478 | * is untouched. Otherwise it is extended. Returns zero on |
| 2479 | * success. The skb is freed on error. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2480 | */ |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 2481 | static inline int skb_padto(struct sk_buff *skb, unsigned int len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2482 | { |
| 2483 | unsigned int size = skb->len; |
| 2484 | if (likely(size >= len)) |
Herbert Xu | 5b057c6 | 2006-06-23 02:06:41 -0700 | [diff] [blame] | 2485 | return 0; |
Gerrit Renker | 987c402 | 2008-08-11 18:17:17 -0700 | [diff] [blame] | 2486 | return skb_pad(skb, len - size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2487 | } |
| 2488 | |
Alexander Duyck | 9c0c112 | 2014-12-03 08:17:33 -0800 | [diff] [blame] | 2489 | /** |
| 2490 | * skb_put_padto - increase size and pad an skbuff up to a minimal size |
| 2491 | * @skb: buffer to pad |
| 2492 | * @len: minimal length |
| 2493 | * |
| 2494 | * Pads up a buffer to ensure the trailing bytes exist and are |
| 2495 | * blanked. If the buffer already contains sufficient data it |
| 2496 | * is untouched. Otherwise it is extended. Returns zero on |
| 2497 | * success. The skb is freed on error. |
| 2498 | */ |
| 2499 | static inline int skb_put_padto(struct sk_buff *skb, unsigned int len) |
| 2500 | { |
| 2501 | unsigned int size = skb->len; |
| 2502 | |
| 2503 | if (unlikely(size < len)) { |
| 2504 | len -= size; |
| 2505 | if (skb_pad(skb, len)) |
| 2506 | return -ENOMEM; |
| 2507 | __skb_put(skb, len); |
| 2508 | } |
| 2509 | return 0; |
| 2510 | } |
| 2511 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2512 | static inline int skb_add_data(struct sk_buff *skb, |
Al Viro | af2b040 | 2014-11-27 21:44:24 -0500 | [diff] [blame] | 2513 | struct iov_iter *from, int copy) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2514 | { |
| 2515 | const int off = skb->len; |
| 2516 | |
| 2517 | if (skb->ip_summed == CHECKSUM_NONE) { |
Al Viro | af2b040 | 2014-11-27 21:44:24 -0500 | [diff] [blame] | 2518 | __wsum csum = 0; |
| 2519 | if (csum_and_copy_from_iter(skb_put(skb, copy), copy, |
| 2520 | &csum, from) == copy) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2521 | skb->csum = csum_block_add(skb->csum, csum, off); |
| 2522 | return 0; |
| 2523 | } |
Al Viro | af2b040 | 2014-11-27 21:44:24 -0500 | [diff] [blame] | 2524 | } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2525 | return 0; |
| 2526 | |
| 2527 | __skb_trim(skb, off); |
| 2528 | return -EFAULT; |
| 2529 | } |
| 2530 | |
Eric Dumazet | 38ba0a6 | 2012-04-23 17:48:27 +0000 | [diff] [blame] | 2531 | static inline bool skb_can_coalesce(struct sk_buff *skb, int i, |
| 2532 | const struct page *page, int off) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2533 | { |
| 2534 | if (i) { |
Eric Dumazet | 9e903e0 | 2011-10-18 21:00:24 +0000 | [diff] [blame] | 2535 | const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2536 | |
Ian Campbell | ea2ab69 | 2011-08-22 23:44:58 +0000 | [diff] [blame] | 2537 | return page == skb_frag_page(frag) && |
Eric Dumazet | 9e903e0 | 2011-10-18 21:00:24 +0000 | [diff] [blame] | 2538 | off == frag->page_offset + skb_frag_size(frag); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2539 | } |
Eric Dumazet | 38ba0a6 | 2012-04-23 17:48:27 +0000 | [diff] [blame] | 2540 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2541 | } |
| 2542 | |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 2543 | static inline int __skb_linearize(struct sk_buff *skb) |
| 2544 | { |
| 2545 | return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; |
| 2546 | } |
| 2547 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2548 | /** |
| 2549 | * skb_linearize - convert paged skb to linear one |
| 2550 | * @skb: buffer to linarize |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2551 | * |
| 2552 | * If there is no free memory -ENOMEM is returned, otherwise zero |
| 2553 | * is returned and the old skb data released. |
| 2554 | */ |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 2555 | static inline int skb_linearize(struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2556 | { |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 2557 | return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; |
| 2558 | } |
| 2559 | |
| 2560 | /** |
Eric Dumazet | cef401d | 2013-01-25 20:34:37 +0000 | [diff] [blame] | 2561 | * skb_has_shared_frag - can any frag be overwritten |
| 2562 | * @skb: buffer to test |
| 2563 | * |
| 2564 | * Return true if the skb has at least one frag that might be modified |
| 2565 | * by an external entity (as in vmsplice()/sendfile()) |
| 2566 | */ |
| 2567 | static inline bool skb_has_shared_frag(const struct sk_buff *skb) |
| 2568 | { |
Pravin B Shelar | c9af6db | 2013-02-11 09:27:41 +0000 | [diff] [blame] | 2569 | return skb_is_nonlinear(skb) && |
| 2570 | skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG; |
Eric Dumazet | cef401d | 2013-01-25 20:34:37 +0000 | [diff] [blame] | 2571 | } |
| 2572 | |
| 2573 | /** |
Herbert Xu | 364c6ba | 2006-06-09 16:10:40 -0700 | [diff] [blame] | 2574 | * skb_linearize_cow - make sure skb is linear and writable |
| 2575 | * @skb: buffer to process |
| 2576 | * |
| 2577 | * If there is no free memory -ENOMEM is returned, otherwise zero |
| 2578 | * is returned and the old skb data released. |
| 2579 | */ |
| 2580 | static inline int skb_linearize_cow(struct sk_buff *skb) |
| 2581 | { |
| 2582 | return skb_is_nonlinear(skb) || skb_cloned(skb) ? |
| 2583 | __skb_linearize(skb) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2584 | } |
| 2585 | |
| 2586 | /** |
| 2587 | * skb_postpull_rcsum - update checksum for received skb after pull |
| 2588 | * @skb: buffer to update |
| 2589 | * @start: start of data before pull |
| 2590 | * @len: length of data pulled |
| 2591 | * |
| 2592 | * 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] | 2593 | * update the CHECKSUM_COMPLETE checksum, or set ip_summed to |
| 2594 | * CHECKSUM_NONE so that it can be recomputed from scratch. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2595 | */ |
| 2596 | |
| 2597 | static inline void skb_postpull_rcsum(struct sk_buff *skb, |
Herbert Xu | cbb042f | 2006-03-20 22:43:56 -0800 | [diff] [blame] | 2598 | const void *start, unsigned int len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2599 | { |
Patrick McHardy | 84fa793 | 2006-08-29 16:44:56 -0700 | [diff] [blame] | 2600 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2601 | skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); |
| 2602 | } |
| 2603 | |
Herbert Xu | cbb042f | 2006-03-20 22:43:56 -0800 | [diff] [blame] | 2604 | unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); |
| 2605 | |
David S. Miller | 7ce5a27 | 2013-12-02 17:26:05 -0500 | [diff] [blame] | 2606 | /** |
| 2607 | * pskb_trim_rcsum - trim received skb and update checksum |
| 2608 | * @skb: buffer to trim |
| 2609 | * @len: new length |
| 2610 | * |
| 2611 | * This is exactly the same as pskb_trim except that it ensures the |
| 2612 | * checksum of received packets are still valid after the operation. |
| 2613 | */ |
| 2614 | |
| 2615 | static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) |
| 2616 | { |
| 2617 | if (likely(len >= skb->len)) |
| 2618 | return 0; |
| 2619 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
| 2620 | skb->ip_summed = CHECKSUM_NONE; |
| 2621 | return __pskb_trim(skb, len); |
| 2622 | } |
| 2623 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2624 | #define skb_queue_walk(queue, skb) \ |
| 2625 | for (skb = (queue)->next; \ |
Linus Torvalds | a1e4891 | 2011-05-22 16:51:43 -0700 | [diff] [blame] | 2626 | skb != (struct sk_buff *)(queue); \ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2627 | skb = skb->next) |
| 2628 | |
James Chapman | 46f8914 | 2007-04-30 00:07:31 -0700 | [diff] [blame] | 2629 | #define skb_queue_walk_safe(queue, skb, tmp) \ |
| 2630 | for (skb = (queue)->next, tmp = skb->next; \ |
| 2631 | skb != (struct sk_buff *)(queue); \ |
| 2632 | skb = tmp, tmp = skb->next) |
| 2633 | |
David S. Miller | 1164f52 | 2008-09-23 00:49:44 -0700 | [diff] [blame] | 2634 | #define skb_queue_walk_from(queue, skb) \ |
Linus Torvalds | a1e4891 | 2011-05-22 16:51:43 -0700 | [diff] [blame] | 2635 | for (; skb != (struct sk_buff *)(queue); \ |
David S. Miller | 1164f52 | 2008-09-23 00:49:44 -0700 | [diff] [blame] | 2636 | skb = skb->next) |
| 2637 | |
| 2638 | #define skb_queue_walk_from_safe(queue, skb, tmp) \ |
| 2639 | for (tmp = skb->next; \ |
| 2640 | skb != (struct sk_buff *)(queue); \ |
| 2641 | skb = tmp, tmp = skb->next) |
| 2642 | |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 2643 | #define skb_queue_reverse_walk(queue, skb) \ |
| 2644 | for (skb = (queue)->prev; \ |
Linus Torvalds | a1e4891 | 2011-05-22 16:51:43 -0700 | [diff] [blame] | 2645 | skb != (struct sk_buff *)(queue); \ |
Stephen Hemminger | 300ce17 | 2005-10-30 13:47:34 -0800 | [diff] [blame] | 2646 | skb = skb->prev) |
| 2647 | |
David S. Miller | 686a295 | 2011-01-20 22:47:32 -0800 | [diff] [blame] | 2648 | #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ |
| 2649 | for (skb = (queue)->prev, tmp = skb->prev; \ |
| 2650 | skb != (struct sk_buff *)(queue); \ |
| 2651 | skb = tmp, tmp = skb->prev) |
| 2652 | |
| 2653 | #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ |
| 2654 | for (tmp = skb->prev; \ |
| 2655 | skb != (struct sk_buff *)(queue); \ |
| 2656 | skb = tmp, tmp = skb->prev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2657 | |
David S. Miller | 21dc330 | 2010-08-23 00:13:46 -0700 | [diff] [blame] | 2658 | static inline bool skb_has_frag_list(const struct sk_buff *skb) |
David S. Miller | ee03987 | 2009-06-09 00:17:13 -0700 | [diff] [blame] | 2659 | { |
| 2660 | return skb_shinfo(skb)->frag_list != NULL; |
| 2661 | } |
| 2662 | |
| 2663 | static inline void skb_frag_list_init(struct sk_buff *skb) |
| 2664 | { |
| 2665 | skb_shinfo(skb)->frag_list = NULL; |
| 2666 | } |
| 2667 | |
| 2668 | static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag) |
| 2669 | { |
| 2670 | frag->next = skb_shinfo(skb)->frag_list; |
| 2671 | skb_shinfo(skb)->frag_list = frag; |
| 2672 | } |
| 2673 | |
| 2674 | #define skb_walk_frags(skb, iter) \ |
| 2675 | for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) |
| 2676 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2677 | struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, |
| 2678 | int *peeked, int *off, int *err); |
| 2679 | struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, |
| 2680 | int *err); |
| 2681 | unsigned int datagram_poll(struct file *file, struct socket *sock, |
| 2682 | struct poll_table_struct *wait); |
Al Viro | c0371da | 2014-11-24 10:42:55 -0500 | [diff] [blame] | 2683 | int skb_copy_datagram_iter(const struct sk_buff *from, int offset, |
| 2684 | struct iov_iter *to, int size); |
David S. Miller | 51f3d02 | 2014-11-05 16:46:40 -0500 | [diff] [blame] | 2685 | static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, |
| 2686 | struct msghdr *msg, int size) |
| 2687 | { |
Al Viro | e5a4b0b | 2014-11-24 18:17:55 -0500 | [diff] [blame] | 2688 | return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); |
David S. Miller | 51f3d02 | 2014-11-05 16:46:40 -0500 | [diff] [blame] | 2689 | } |
Al Viro | e5a4b0b | 2014-11-24 18:17:55 -0500 | [diff] [blame] | 2690 | int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, |
| 2691 | struct msghdr *msg); |
Al Viro | 3a654f9 | 2014-06-19 14:15:22 -0400 | [diff] [blame] | 2692 | int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, |
| 2693 | struct iov_iter *from, int len); |
Al Viro | 3a654f9 | 2014-06-19 14:15:22 -0400 | [diff] [blame] | 2694 | int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2695 | void skb_free_datagram(struct sock *sk, struct sk_buff *skb); |
| 2696 | void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb); |
| 2697 | int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2698 | int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); |
| 2699 | int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); |
| 2700 | __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, |
| 2701 | int len, __wsum csum); |
| 2702 | int skb_splice_bits(struct sk_buff *skb, unsigned int offset, |
| 2703 | struct pipe_inode_info *pipe, unsigned int len, |
| 2704 | unsigned int flags); |
| 2705 | void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); |
Thomas Graf | af2806f | 2013-12-13 15:22:17 +0100 | [diff] [blame] | 2706 | unsigned int skb_zerocopy_headlen(const struct sk_buff *from); |
Zoltan Kiss | 36d5fe6 | 2014-03-26 22:37:45 +0000 | [diff] [blame] | 2707 | int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, |
| 2708 | int len, int hlen); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2709 | void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); |
| 2710 | int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); |
| 2711 | void skb_scrub_packet(struct sk_buff *skb, bool xnet); |
Florian Westphal | de960aa | 2014-01-26 10:58:16 +0100 | [diff] [blame] | 2712 | unsigned int skb_gso_transport_seglen(const struct sk_buff *skb); |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2713 | struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); |
Vlad Yasevich | 0d5501c | 2014-08-08 14:42:13 -0400 | [diff] [blame] | 2714 | struct sk_buff *skb_vlan_untag(struct sk_buff *skb); |
Jiri Pirko | e219512 | 2014-11-19 14:05:01 +0100 | [diff] [blame] | 2715 | int skb_ensure_writable(struct sk_buff *skb, int write_len); |
Jiri Pirko | 93515d5 | 2014-11-19 14:05:02 +0100 | [diff] [blame] | 2716 | int skb_vlan_pop(struct sk_buff *skb); |
| 2717 | int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 2718 | |
Al Viro | 6ce8e9c | 2014-04-06 21:25:44 -0400 | [diff] [blame] | 2719 | static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) |
| 2720 | { |
Al Viro | 21226ab | 2014-11-28 15:48:29 -0500 | [diff] [blame] | 2721 | return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
Al Viro | 6ce8e9c | 2014-04-06 21:25:44 -0400 | [diff] [blame] | 2722 | } |
| 2723 | |
Al Viro | 7eab8d9 | 2014-04-06 21:51:23 -0400 | [diff] [blame] | 2724 | static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) |
| 2725 | { |
Al Viro | e5a4b0b | 2014-11-24 18:17:55 -0500 | [diff] [blame] | 2726 | return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
Al Viro | 7eab8d9 | 2014-04-06 21:51:23 -0400 | [diff] [blame] | 2727 | } |
| 2728 | |
Daniel Borkmann | 2817a33 | 2013-10-30 11:50:51 +0100 | [diff] [blame] | 2729 | struct skb_checksum_ops { |
| 2730 | __wsum (*update)(const void *mem, int len, __wsum wsum); |
| 2731 | __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); |
| 2732 | }; |
| 2733 | |
| 2734 | __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, |
| 2735 | __wsum csum, const struct skb_checksum_ops *ops); |
| 2736 | __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, |
| 2737 | __wsum csum); |
| 2738 | |
David S. Miller | 690e36e | 2014-08-23 12:13:41 -0700 | [diff] [blame] | 2739 | static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset, |
| 2740 | int len, void *data, int hlen, void *buffer) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2741 | { |
Patrick McHardy | 55820ee | 2005-07-05 14:08:10 -0700 | [diff] [blame] | 2742 | if (hlen - offset >= len) |
David S. Miller | 690e36e | 2014-08-23 12:13:41 -0700 | [diff] [blame] | 2743 | return data + offset; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2744 | |
David S. Miller | 690e36e | 2014-08-23 12:13:41 -0700 | [diff] [blame] | 2745 | if (!skb || |
| 2746 | skb_copy_bits(skb, offset, buffer, len) < 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2747 | return NULL; |
| 2748 | |
| 2749 | return buffer; |
| 2750 | } |
| 2751 | |
David S. Miller | 690e36e | 2014-08-23 12:13:41 -0700 | [diff] [blame] | 2752 | static inline void *skb_header_pointer(const struct sk_buff *skb, int offset, |
| 2753 | int len, void *buffer) |
| 2754 | { |
| 2755 | return __skb_header_pointer(skb, offset, len, skb->data, |
| 2756 | skb_headlen(skb), buffer); |
| 2757 | } |
| 2758 | |
Daniel Borkmann | 4262e5c | 2013-12-06 11:36:16 +0100 | [diff] [blame] | 2759 | /** |
| 2760 | * skb_needs_linearize - check if we need to linearize a given skb |
| 2761 | * depending on the given device features. |
| 2762 | * @skb: socket buffer to check |
| 2763 | * @features: net device features |
| 2764 | * |
| 2765 | * Returns true if either: |
| 2766 | * 1. skb has frag_list and the device doesn't support FRAGLIST, or |
| 2767 | * 2. skb is fragmented and the device does not support SG. |
| 2768 | */ |
| 2769 | static inline bool skb_needs_linearize(struct sk_buff *skb, |
| 2770 | netdev_features_t features) |
| 2771 | { |
| 2772 | return skb_is_nonlinear(skb) && |
| 2773 | ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || |
| 2774 | (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); |
| 2775 | } |
| 2776 | |
Arnaldo Carvalho de Melo | d626f62 | 2007-03-27 18:55:52 -0300 | [diff] [blame] | 2777 | static inline void skb_copy_from_linear_data(const struct sk_buff *skb, |
| 2778 | void *to, |
| 2779 | const unsigned int len) |
| 2780 | { |
| 2781 | memcpy(to, skb->data, len); |
| 2782 | } |
| 2783 | |
| 2784 | static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, |
| 2785 | const int offset, void *to, |
| 2786 | const unsigned int len) |
| 2787 | { |
| 2788 | memcpy(to, skb->data + offset, len); |
| 2789 | } |
| 2790 | |
Arnaldo Carvalho de Melo | 27d7ff4 | 2007-03-31 11:55:19 -0300 | [diff] [blame] | 2791 | static inline void skb_copy_to_linear_data(struct sk_buff *skb, |
| 2792 | const void *from, |
| 2793 | const unsigned int len) |
| 2794 | { |
| 2795 | memcpy(skb->data, from, len); |
| 2796 | } |
| 2797 | |
| 2798 | static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, |
| 2799 | const int offset, |
| 2800 | const void *from, |
| 2801 | const unsigned int len) |
| 2802 | { |
| 2803 | memcpy(skb->data + offset, from, len); |
| 2804 | } |
| 2805 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2806 | void skb_init(void); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2807 | |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 2808 | static inline ktime_t skb_get_ktime(const struct sk_buff *skb) |
| 2809 | { |
| 2810 | return skb->tstamp; |
| 2811 | } |
| 2812 | |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2813 | /** |
| 2814 | * skb_get_timestamp - get timestamp from a skb |
| 2815 | * @skb: skb to get stamp from |
| 2816 | * @stamp: pointer to struct timeval to store stamp in |
| 2817 | * |
| 2818 | * Timestamps are stored in the skb as offsets to a base timestamp. |
| 2819 | * This function converts the offset back to a struct timeval and stores |
| 2820 | * it in stamp. |
| 2821 | */ |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 2822 | static inline void skb_get_timestamp(const struct sk_buff *skb, |
| 2823 | struct timeval *stamp) |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2824 | { |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 2825 | *stamp = ktime_to_timeval(skb->tstamp); |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2826 | } |
| 2827 | |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 2828 | static inline void skb_get_timestampns(const struct sk_buff *skb, |
| 2829 | struct timespec *stamp) |
| 2830 | { |
| 2831 | *stamp = ktime_to_timespec(skb->tstamp); |
| 2832 | } |
| 2833 | |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 2834 | static inline void __net_timestamp(struct sk_buff *skb) |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2835 | { |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 2836 | skb->tstamp = ktime_get_real(); |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2837 | } |
| 2838 | |
Stephen Hemminger | 164891a | 2007-04-23 22:26:16 -0700 | [diff] [blame] | 2839 | static inline ktime_t net_timedelta(ktime_t t) |
| 2840 | { |
| 2841 | return ktime_sub(ktime_get_real(), t); |
| 2842 | } |
| 2843 | |
Ilpo Järvinen | b9ce204 | 2007-06-15 15:08:43 -0700 | [diff] [blame] | 2844 | static inline ktime_t net_invalid_timestamp(void) |
| 2845 | { |
| 2846 | return ktime_set(0, 0); |
| 2847 | } |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2848 | |
Alexander Duyck | 62bccb8 | 2014-09-04 13:31:35 -0400 | [diff] [blame] | 2849 | struct sk_buff *skb_clone_sk(struct sk_buff *skb); |
| 2850 | |
Richard Cochran | c1f19b5 | 2010-07-17 08:49:36 +0000 | [diff] [blame] | 2851 | #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING |
| 2852 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2853 | void skb_clone_tx_timestamp(struct sk_buff *skb); |
| 2854 | bool skb_defer_rx_timestamp(struct sk_buff *skb); |
Richard Cochran | c1f19b5 | 2010-07-17 08:49:36 +0000 | [diff] [blame] | 2855 | |
| 2856 | #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ |
| 2857 | |
| 2858 | static inline void skb_clone_tx_timestamp(struct sk_buff *skb) |
| 2859 | { |
| 2860 | } |
| 2861 | |
| 2862 | static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) |
| 2863 | { |
| 2864 | return false; |
| 2865 | } |
| 2866 | |
| 2867 | #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ |
| 2868 | |
| 2869 | /** |
| 2870 | * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps |
| 2871 | * |
Richard Cochran | da92b19 | 2011-10-21 00:49:15 +0000 | [diff] [blame] | 2872 | * PHY drivers may accept clones of transmitted packets for |
| 2873 | * timestamping via their phy_driver.txtstamp method. These drivers |
| 2874 | * must call this function to return the skb back to the stack, with |
| 2875 | * or without a timestamp. |
| 2876 | * |
Richard Cochran | c1f19b5 | 2010-07-17 08:49:36 +0000 | [diff] [blame] | 2877 | * @skb: clone of the the original outgoing packet |
Richard Cochran | da92b19 | 2011-10-21 00:49:15 +0000 | [diff] [blame] | 2878 | * @hwtstamps: hardware time stamps, may be NULL if not available |
Richard Cochran | c1f19b5 | 2010-07-17 08:49:36 +0000 | [diff] [blame] | 2879 | * |
| 2880 | */ |
| 2881 | void skb_complete_tx_timestamp(struct sk_buff *skb, |
| 2882 | struct skb_shared_hwtstamps *hwtstamps); |
| 2883 | |
Willem de Bruijn | e7fd288 | 2014-08-04 22:11:48 -0400 | [diff] [blame] | 2884 | void __skb_tstamp_tx(struct sk_buff *orig_skb, |
| 2885 | struct skb_shared_hwtstamps *hwtstamps, |
| 2886 | struct sock *sk, int tstype); |
| 2887 | |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 2888 | /** |
| 2889 | * skb_tstamp_tx - queue clone of skb with send time stamps |
| 2890 | * @orig_skb: the original outgoing packet |
| 2891 | * @hwtstamps: hardware time stamps, may be NULL if not available |
| 2892 | * |
| 2893 | * If the skb has a socket associated, then this function clones the |
| 2894 | * skb (thus sharing the actual data and optional structures), stores |
| 2895 | * the optional hardware time stamping information (if non NULL) or |
| 2896 | * generates a software time stamp (otherwise), then queues the clone |
| 2897 | * to the error queue of the socket. Errors are silently ignored. |
| 2898 | */ |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2899 | void skb_tstamp_tx(struct sk_buff *orig_skb, |
| 2900 | struct skb_shared_hwtstamps *hwtstamps); |
Patrick Ohly | ac45f60 | 2009-02-12 05:03:37 +0000 | [diff] [blame] | 2901 | |
Richard Cochran | 4507a71 | 2010-07-17 08:48:28 +0000 | [diff] [blame] | 2902 | static inline void sw_tx_timestamp(struct sk_buff *skb) |
| 2903 | { |
Oliver Hartkopp | 2244d07 | 2010-08-17 08:59:14 +0000 | [diff] [blame] | 2904 | if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP && |
| 2905 | !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) |
Richard Cochran | 4507a71 | 2010-07-17 08:48:28 +0000 | [diff] [blame] | 2906 | skb_tstamp_tx(skb, NULL); |
| 2907 | } |
| 2908 | |
| 2909 | /** |
| 2910 | * skb_tx_timestamp() - Driver hook for transmit timestamping |
| 2911 | * |
| 2912 | * Ethernet MAC Drivers should call this function in their hard_xmit() |
Richard Cochran | 4ff75b7 | 2011-06-19 03:31:39 +0000 | [diff] [blame] | 2913 | * function immediately before giving the sk_buff to the MAC hardware. |
Richard Cochran | 4507a71 | 2010-07-17 08:48:28 +0000 | [diff] [blame] | 2914 | * |
David S. Miller | 73409f3 | 2013-12-27 13:04:33 -0500 | [diff] [blame] | 2915 | * Specifically, one should make absolutely sure that this function is |
| 2916 | * called before TX completion of this packet can trigger. Otherwise |
| 2917 | * the packet could potentially already be freed. |
| 2918 | * |
Richard Cochran | 4507a71 | 2010-07-17 08:48:28 +0000 | [diff] [blame] | 2919 | * @skb: A socket buffer. |
| 2920 | */ |
| 2921 | static inline void skb_tx_timestamp(struct sk_buff *skb) |
| 2922 | { |
Richard Cochran | c1f19b5 | 2010-07-17 08:49:36 +0000 | [diff] [blame] | 2923 | skb_clone_tx_timestamp(skb); |
Richard Cochran | 4507a71 | 2010-07-17 08:48:28 +0000 | [diff] [blame] | 2924 | sw_tx_timestamp(skb); |
| 2925 | } |
| 2926 | |
Johannes Berg | 6e3e939 | 2011-11-09 10:15:42 +0100 | [diff] [blame] | 2927 | /** |
| 2928 | * skb_complete_wifi_ack - deliver skb with wifi status |
| 2929 | * |
| 2930 | * @skb: the original outgoing packet |
| 2931 | * @acked: ack status |
| 2932 | * |
| 2933 | */ |
| 2934 | void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); |
| 2935 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 2936 | __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); |
| 2937 | __sum16 __skb_checksum_complete(struct sk_buff *skb); |
Herbert Xu | fb286bb | 2005-11-10 13:01:24 -0800 | [diff] [blame] | 2938 | |
Herbert Xu | 6047637 | 2007-04-09 11:59:39 -0700 | [diff] [blame] | 2939 | static inline int skb_csum_unnecessary(const struct sk_buff *skb) |
| 2940 | { |
Tom Herbert | 6edec0e | 2015-02-10 16:30:28 -0800 | [diff] [blame] | 2941 | return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || |
| 2942 | skb->csum_valid || |
| 2943 | (skb->ip_summed == CHECKSUM_PARTIAL && |
| 2944 | skb_checksum_start_offset(skb) >= 0)); |
Herbert Xu | 6047637 | 2007-04-09 11:59:39 -0700 | [diff] [blame] | 2945 | } |
| 2946 | |
Herbert Xu | fb286bb | 2005-11-10 13:01:24 -0800 | [diff] [blame] | 2947 | /** |
| 2948 | * skb_checksum_complete - Calculate checksum of an entire packet |
| 2949 | * @skb: packet to process |
| 2950 | * |
| 2951 | * This function calculates the checksum over the entire packet plus |
| 2952 | * the value of skb->csum. The latter can be used to supply the |
| 2953 | * checksum of a pseudo header as used by TCP/UDP. It returns the |
| 2954 | * checksum. |
| 2955 | * |
| 2956 | * For protocols that contain complete checksums such as ICMP/TCP/UDP, |
| 2957 | * this function can be used to verify that checksum on received |
| 2958 | * packets. In that case the function should return zero if the |
| 2959 | * checksum is correct. In particular, this function will return zero |
| 2960 | * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the |
| 2961 | * hardware has already verified the correctness of the checksum. |
| 2962 | */ |
Al Viro | 4381ca3 | 2007-07-15 21:00:11 +0100 | [diff] [blame] | 2963 | static inline __sum16 skb_checksum_complete(struct sk_buff *skb) |
Herbert Xu | fb286bb | 2005-11-10 13:01:24 -0800 | [diff] [blame] | 2964 | { |
Herbert Xu | 6047637 | 2007-04-09 11:59:39 -0700 | [diff] [blame] | 2965 | return skb_csum_unnecessary(skb) ? |
| 2966 | 0 : __skb_checksum_complete(skb); |
Herbert Xu | fb286bb | 2005-11-10 13:01:24 -0800 | [diff] [blame] | 2967 | } |
| 2968 | |
Tom Herbert | 77cffe2 | 2014-08-27 21:26:46 -0700 | [diff] [blame] | 2969 | static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) |
| 2970 | { |
| 2971 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { |
| 2972 | if (skb->csum_level == 0) |
| 2973 | skb->ip_summed = CHECKSUM_NONE; |
| 2974 | else |
| 2975 | skb->csum_level--; |
| 2976 | } |
| 2977 | } |
| 2978 | |
| 2979 | static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) |
| 2980 | { |
| 2981 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { |
| 2982 | if (skb->csum_level < SKB_MAX_CSUM_LEVEL) |
| 2983 | skb->csum_level++; |
| 2984 | } else if (skb->ip_summed == CHECKSUM_NONE) { |
| 2985 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
| 2986 | skb->csum_level = 0; |
| 2987 | } |
| 2988 | } |
| 2989 | |
Tom Herbert | 5a21232 | 2014-08-31 15:12:41 -0700 | [diff] [blame] | 2990 | static inline void __skb_mark_checksum_bad(struct sk_buff *skb) |
| 2991 | { |
| 2992 | /* Mark current checksum as bad (typically called from GRO |
| 2993 | * path). In the case that ip_summed is CHECKSUM_NONE |
| 2994 | * this must be the first checksum encountered in the packet. |
| 2995 | * When ip_summed is CHECKSUM_UNNECESSARY, this is the first |
| 2996 | * checksum after the last one validated. For UDP, a zero |
| 2997 | * checksum can not be marked as bad. |
| 2998 | */ |
| 2999 | |
| 3000 | if (skb->ip_summed == CHECKSUM_NONE || |
| 3001 | skb->ip_summed == CHECKSUM_UNNECESSARY) |
| 3002 | skb->csum_bad = 1; |
| 3003 | } |
| 3004 | |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3005 | /* Check if we need to perform checksum complete validation. |
| 3006 | * |
| 3007 | * Returns true if checksum complete is needed, false otherwise |
| 3008 | * (either checksum is unnecessary or zero checksum is allowed). |
| 3009 | */ |
| 3010 | static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, |
| 3011 | bool zero_okay, |
| 3012 | __sum16 check) |
| 3013 | { |
Tom Herbert | 5d0c2b9 | 2014-06-10 18:54:13 -0700 | [diff] [blame] | 3014 | if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { |
| 3015 | skb->csum_valid = 1; |
Tom Herbert | 77cffe2 | 2014-08-27 21:26:46 -0700 | [diff] [blame] | 3016 | __skb_decr_checksum_unnecessary(skb); |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3017 | return false; |
| 3018 | } |
| 3019 | |
| 3020 | return true; |
| 3021 | } |
| 3022 | |
| 3023 | /* For small packets <= CHECKSUM_BREAK peform checksum complete directly |
| 3024 | * in checksum_init. |
| 3025 | */ |
| 3026 | #define CHECKSUM_BREAK 76 |
| 3027 | |
Tom Herbert | 4e18b9a | 2015-04-20 14:10:04 -0700 | [diff] [blame] | 3028 | /* Unset checksum-complete |
| 3029 | * |
| 3030 | * Unset checksum complete can be done when packet is being modified |
| 3031 | * (uncompressed for instance) and checksum-complete value is |
| 3032 | * invalidated. |
| 3033 | */ |
| 3034 | static inline void skb_checksum_complete_unset(struct sk_buff *skb) |
| 3035 | { |
| 3036 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
| 3037 | skb->ip_summed = CHECKSUM_NONE; |
| 3038 | } |
| 3039 | |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3040 | /* Validate (init) checksum based on checksum complete. |
| 3041 | * |
| 3042 | * Return values: |
| 3043 | * 0: checksum is validated or try to in skb_checksum_complete. In the latter |
| 3044 | * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo |
| 3045 | * checksum is stored in skb->csum for use in __skb_checksum_complete |
| 3046 | * non-zero: value of invalid checksum |
| 3047 | * |
| 3048 | */ |
| 3049 | static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, |
| 3050 | bool complete, |
| 3051 | __wsum psum) |
| 3052 | { |
| 3053 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
| 3054 | if (!csum_fold(csum_add(psum, skb->csum))) { |
Tom Herbert | 5d0c2b9 | 2014-06-10 18:54:13 -0700 | [diff] [blame] | 3055 | skb->csum_valid = 1; |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3056 | return 0; |
| 3057 | } |
Tom Herbert | 5a21232 | 2014-08-31 15:12:41 -0700 | [diff] [blame] | 3058 | } else if (skb->csum_bad) { |
| 3059 | /* ip_summed == CHECKSUM_NONE in this case */ |
Eric Dumazet | c91d460 | 2015-05-15 05:48:07 -0700 | [diff] [blame] | 3060 | return (__force __sum16)1; |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3061 | } |
| 3062 | |
| 3063 | skb->csum = psum; |
| 3064 | |
Tom Herbert | 5d0c2b9 | 2014-06-10 18:54:13 -0700 | [diff] [blame] | 3065 | if (complete || skb->len <= CHECKSUM_BREAK) { |
| 3066 | __sum16 csum; |
| 3067 | |
| 3068 | csum = __skb_checksum_complete(skb); |
| 3069 | skb->csum_valid = !csum; |
| 3070 | return csum; |
| 3071 | } |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3072 | |
| 3073 | return 0; |
| 3074 | } |
| 3075 | |
| 3076 | static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) |
| 3077 | { |
| 3078 | return 0; |
| 3079 | } |
| 3080 | |
| 3081 | /* Perform checksum validate (init). Note that this is a macro since we only |
| 3082 | * want to calculate the pseudo header which is an input function if necessary. |
| 3083 | * First we try to validate without any computation (checksum unnecessary) and |
| 3084 | * then calculate based on checksum complete calling the function to compute |
| 3085 | * pseudo header. |
| 3086 | * |
| 3087 | * Return values: |
| 3088 | * 0: checksum is validated or try to in skb_checksum_complete |
| 3089 | * non-zero: value of invalid checksum |
| 3090 | */ |
| 3091 | #define __skb_checksum_validate(skb, proto, complete, \ |
| 3092 | zero_okay, check, compute_pseudo) \ |
| 3093 | ({ \ |
| 3094 | __sum16 __ret = 0; \ |
Tom Herbert | 5d0c2b9 | 2014-06-10 18:54:13 -0700 | [diff] [blame] | 3095 | skb->csum_valid = 0; \ |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3096 | if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ |
| 3097 | __ret = __skb_checksum_validate_complete(skb, \ |
| 3098 | complete, compute_pseudo(skb, proto)); \ |
| 3099 | __ret; \ |
| 3100 | }) |
| 3101 | |
| 3102 | #define skb_checksum_init(skb, proto, compute_pseudo) \ |
| 3103 | __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) |
| 3104 | |
| 3105 | #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ |
| 3106 | __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) |
| 3107 | |
| 3108 | #define skb_checksum_validate(skb, proto, compute_pseudo) \ |
| 3109 | __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) |
| 3110 | |
| 3111 | #define skb_checksum_validate_zero_check(skb, proto, check, \ |
| 3112 | compute_pseudo) \ |
Sabrina Dubroca | 096a4cf | 2015-02-06 18:54:19 +0100 | [diff] [blame] | 3113 | __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) |
Tom Herbert | 76ba0aa | 2014-05-02 16:29:18 -0700 | [diff] [blame] | 3114 | |
| 3115 | #define skb_checksum_simple_validate(skb) \ |
| 3116 | __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) |
| 3117 | |
Tom Herbert | d96535a | 2014-08-31 15:12:42 -0700 | [diff] [blame] | 3118 | static inline bool __skb_checksum_convert_check(struct sk_buff *skb) |
| 3119 | { |
| 3120 | return (skb->ip_summed == CHECKSUM_NONE && |
| 3121 | skb->csum_valid && !skb->csum_bad); |
| 3122 | } |
| 3123 | |
| 3124 | static inline void __skb_checksum_convert(struct sk_buff *skb, |
| 3125 | __sum16 check, __wsum pseudo) |
| 3126 | { |
| 3127 | skb->csum = ~pseudo; |
| 3128 | skb->ip_summed = CHECKSUM_COMPLETE; |
| 3129 | } |
| 3130 | |
| 3131 | #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \ |
| 3132 | do { \ |
| 3133 | if (__skb_checksum_convert_check(skb)) \ |
| 3134 | __skb_checksum_convert(skb, check, \ |
| 3135 | compute_pseudo(skb, proto)); \ |
| 3136 | } while (0) |
| 3137 | |
Tom Herbert | 15e2396 | 2015-02-10 16:30:31 -0800 | [diff] [blame] | 3138 | static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, |
| 3139 | u16 start, u16 offset) |
| 3140 | { |
| 3141 | skb->ip_summed = CHECKSUM_PARTIAL; |
| 3142 | skb->csum_start = ((unsigned char *)ptr + start) - skb->head; |
| 3143 | skb->csum_offset = offset - start; |
| 3144 | } |
| 3145 | |
Tom Herbert | dcdc899 | 2015-02-02 16:07:34 -0800 | [diff] [blame] | 3146 | /* Update skbuf and packet to reflect the remote checksum offload operation. |
| 3147 | * When called, ptr indicates the starting point for skb->csum when |
| 3148 | * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete |
| 3149 | * here, skb_postpull_rcsum is done so skb->csum start is ptr. |
| 3150 | */ |
| 3151 | static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, |
Tom Herbert | 15e2396 | 2015-02-10 16:30:31 -0800 | [diff] [blame] | 3152 | int start, int offset, bool nopartial) |
Tom Herbert | dcdc899 | 2015-02-02 16:07:34 -0800 | [diff] [blame] | 3153 | { |
| 3154 | __wsum delta; |
| 3155 | |
Tom Herbert | 15e2396 | 2015-02-10 16:30:31 -0800 | [diff] [blame] | 3156 | if (!nopartial) { |
| 3157 | skb_remcsum_adjust_partial(skb, ptr, start, offset); |
| 3158 | return; |
| 3159 | } |
| 3160 | |
Tom Herbert | dcdc899 | 2015-02-02 16:07:34 -0800 | [diff] [blame] | 3161 | if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { |
| 3162 | __skb_checksum_complete(skb); |
| 3163 | skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); |
| 3164 | } |
| 3165 | |
| 3166 | delta = remcsum_adjust(ptr, skb->csum, start, offset); |
| 3167 | |
| 3168 | /* Adjust skb->csum since we changed the packet */ |
| 3169 | skb->csum = csum_add(skb->csum, delta); |
| 3170 | } |
| 3171 | |
Yasuyuki Kozakai | 5f79e0f | 2007-03-23 11:17:07 -0700 | [diff] [blame] | 3172 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 3173 | void nf_conntrack_destroy(struct nf_conntrack *nfct); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3174 | static inline void nf_conntrack_put(struct nf_conntrack *nfct) |
| 3175 | { |
| 3176 | if (nfct && atomic_dec_and_test(&nfct->use)) |
Yasuyuki Kozakai | de6e05c | 2007-03-23 11:17:27 -0700 | [diff] [blame] | 3177 | nf_conntrack_destroy(nfct); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3178 | } |
| 3179 | static inline void nf_conntrack_get(struct nf_conntrack *nfct) |
| 3180 | { |
| 3181 | if (nfct) |
| 3182 | atomic_inc(&nfct->use); |
| 3183 | } |
KOVACS Krisztian | 2fc72c7 | 2011-01-12 20:25:08 +0100 | [diff] [blame] | 3184 | #endif |
Pablo Neira Ayuso | 34666d4 | 2014-09-18 11:29:03 +0200 | [diff] [blame] | 3185 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3186 | static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) |
| 3187 | { |
| 3188 | if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) |
| 3189 | kfree(nf_bridge); |
| 3190 | } |
| 3191 | static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) |
| 3192 | { |
| 3193 | if (nf_bridge) |
| 3194 | atomic_inc(&nf_bridge->use); |
| 3195 | } |
| 3196 | #endif /* CONFIG_BRIDGE_NETFILTER */ |
Patrick McHardy | a193a4a | 2006-03-20 19:23:05 -0800 | [diff] [blame] | 3197 | static inline void nf_reset(struct sk_buff *skb) |
| 3198 | { |
Yasuyuki Kozakai | 5f79e0f | 2007-03-23 11:17:07 -0700 | [diff] [blame] | 3199 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
Patrick McHardy | a193a4a | 2006-03-20 19:23:05 -0800 | [diff] [blame] | 3200 | nf_conntrack_put(skb->nfct); |
| 3201 | skb->nfct = NULL; |
KOVACS Krisztian | 2fc72c7 | 2011-01-12 20:25:08 +0100 | [diff] [blame] | 3202 | #endif |
Pablo Neira Ayuso | 34666d4 | 2014-09-18 11:29:03 +0200 | [diff] [blame] | 3203 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
Patrick McHardy | a193a4a | 2006-03-20 19:23:05 -0800 | [diff] [blame] | 3204 | nf_bridge_put(skb->nf_bridge); |
| 3205 | skb->nf_bridge = NULL; |
| 3206 | #endif |
| 3207 | } |
| 3208 | |
Patrick McHardy | 124dff0 | 2013-04-05 20:42:05 +0200 | [diff] [blame] | 3209 | static inline void nf_reset_trace(struct sk_buff *skb) |
| 3210 | { |
Florian Westphal | 478b360 | 2014-02-15 23:48:45 +0100 | [diff] [blame] | 3211 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) |
Gao feng | 130549fe | 2013-03-21 19:48:41 +0000 | [diff] [blame] | 3212 | skb->nf_trace = 0; |
| 3213 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3214 | } |
| 3215 | |
Yasuyuki Kozakai | edda553 | 2007-03-14 16:43:37 -0700 | [diff] [blame] | 3216 | /* Note: This doesn't put any conntrack and bridge info in dst. */ |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 3217 | static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, |
| 3218 | bool copy) |
Yasuyuki Kozakai | edda553 | 2007-03-14 16:43:37 -0700 | [diff] [blame] | 3219 | { |
Yasuyuki Kozakai | 5f79e0f | 2007-03-23 11:17:07 -0700 | [diff] [blame] | 3220 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
Yasuyuki Kozakai | edda553 | 2007-03-14 16:43:37 -0700 | [diff] [blame] | 3221 | dst->nfct = src->nfct; |
| 3222 | nf_conntrack_get(src->nfct); |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 3223 | if (copy) |
| 3224 | dst->nfctinfo = src->nfctinfo; |
KOVACS Krisztian | 2fc72c7 | 2011-01-12 20:25:08 +0100 | [diff] [blame] | 3225 | #endif |
Pablo Neira Ayuso | 34666d4 | 2014-09-18 11:29:03 +0200 | [diff] [blame] | 3226 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
Yasuyuki Kozakai | edda553 | 2007-03-14 16:43:37 -0700 | [diff] [blame] | 3227 | dst->nf_bridge = src->nf_bridge; |
| 3228 | nf_bridge_get(src->nf_bridge); |
| 3229 | #endif |
Florian Westphal | 478b360 | 2014-02-15 23:48:45 +0100 | [diff] [blame] | 3230 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 3231 | if (copy) |
| 3232 | dst->nf_trace = src->nf_trace; |
Florian Westphal | 478b360 | 2014-02-15 23:48:45 +0100 | [diff] [blame] | 3233 | #endif |
Yasuyuki Kozakai | edda553 | 2007-03-14 16:43:37 -0700 | [diff] [blame] | 3234 | } |
| 3235 | |
Yasuyuki Kozakai | e7ac05f | 2007-03-14 16:44:01 -0700 | [diff] [blame] | 3236 | static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) |
| 3237 | { |
Yasuyuki Kozakai | e7ac05f | 2007-03-14 16:44:01 -0700 | [diff] [blame] | 3238 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
Yasuyuki Kozakai | 5f79e0f | 2007-03-23 11:17:07 -0700 | [diff] [blame] | 3239 | nf_conntrack_put(dst->nfct); |
KOVACS Krisztian | 2fc72c7 | 2011-01-12 20:25:08 +0100 | [diff] [blame] | 3240 | #endif |
Pablo Neira Ayuso | 34666d4 | 2014-09-18 11:29:03 +0200 | [diff] [blame] | 3241 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
Yasuyuki Kozakai | e7ac05f | 2007-03-14 16:44:01 -0700 | [diff] [blame] | 3242 | nf_bridge_put(dst->nf_bridge); |
| 3243 | #endif |
Eric Dumazet | b193722 | 2014-09-28 22:18:47 -0700 | [diff] [blame] | 3244 | __nf_copy(dst, src, true); |
Yasuyuki Kozakai | e7ac05f | 2007-03-14 16:44:01 -0700 | [diff] [blame] | 3245 | } |
| 3246 | |
James Morris | 984bc16 | 2006-06-09 00:29:17 -0700 | [diff] [blame] | 3247 | #ifdef CONFIG_NETWORK_SECMARK |
| 3248 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) |
| 3249 | { |
| 3250 | to->secmark = from->secmark; |
| 3251 | } |
| 3252 | |
| 3253 | static inline void skb_init_secmark(struct sk_buff *skb) |
| 3254 | { |
| 3255 | skb->secmark = 0; |
| 3256 | } |
| 3257 | #else |
| 3258 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) |
| 3259 | { } |
| 3260 | |
| 3261 | static inline void skb_init_secmark(struct sk_buff *skb) |
| 3262 | { } |
| 3263 | #endif |
| 3264 | |
Eric W. Biederman | 574f719 | 2014-04-01 12:20:24 -0700 | [diff] [blame] | 3265 | static inline bool skb_irq_freeable(const struct sk_buff *skb) |
| 3266 | { |
| 3267 | return !skb->destructor && |
| 3268 | #if IS_ENABLED(CONFIG_XFRM) |
| 3269 | !skb->sp && |
| 3270 | #endif |
| 3271 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
| 3272 | !skb->nfct && |
| 3273 | #endif |
| 3274 | !skb->_skb_refdst && |
| 3275 | !skb_has_frag_list(skb); |
| 3276 | } |
| 3277 | |
Peter P Waskiewicz Jr | f25f4e4 | 2007-07-06 13:36:20 -0700 | [diff] [blame] | 3278 | static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) |
| 3279 | { |
Peter P Waskiewicz Jr | f25f4e4 | 2007-07-06 13:36:20 -0700 | [diff] [blame] | 3280 | skb->queue_mapping = queue_mapping; |
Peter P Waskiewicz Jr | f25f4e4 | 2007-07-06 13:36:20 -0700 | [diff] [blame] | 3281 | } |
| 3282 | |
Stephen Hemminger | 9247744 | 2009-03-21 13:39:26 -0700 | [diff] [blame] | 3283 | static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) |
Pavel Emelyanov | 4e3ab47 | 2007-10-21 17:01:29 -0700 | [diff] [blame] | 3284 | { |
Pavel Emelyanov | 4e3ab47 | 2007-10-21 17:01:29 -0700 | [diff] [blame] | 3285 | return skb->queue_mapping; |
Pavel Emelyanov | 4e3ab47 | 2007-10-21 17:01:29 -0700 | [diff] [blame] | 3286 | } |
| 3287 | |
Peter P Waskiewicz Jr | f25f4e4 | 2007-07-06 13:36:20 -0700 | [diff] [blame] | 3288 | static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) |
| 3289 | { |
Peter P Waskiewicz Jr | f25f4e4 | 2007-07-06 13:36:20 -0700 | [diff] [blame] | 3290 | to->queue_mapping = from->queue_mapping; |
Peter P Waskiewicz Jr | f25f4e4 | 2007-07-06 13:36:20 -0700 | [diff] [blame] | 3291 | } |
| 3292 | |
David S. Miller | d5a9e24 | 2009-01-27 16:22:11 -0800 | [diff] [blame] | 3293 | static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) |
| 3294 | { |
| 3295 | skb->queue_mapping = rx_queue + 1; |
| 3296 | } |
| 3297 | |
Stephen Hemminger | 9247744 | 2009-03-21 13:39:26 -0700 | [diff] [blame] | 3298 | static inline u16 skb_get_rx_queue(const struct sk_buff *skb) |
David S. Miller | d5a9e24 | 2009-01-27 16:22:11 -0800 | [diff] [blame] | 3299 | { |
| 3300 | return skb->queue_mapping - 1; |
| 3301 | } |
| 3302 | |
Stephen Hemminger | 9247744 | 2009-03-21 13:39:26 -0700 | [diff] [blame] | 3303 | static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) |
David S. Miller | d5a9e24 | 2009-01-27 16:22:11 -0800 | [diff] [blame] | 3304 | { |
Eric Dumazet | a02cec2 | 2010-09-22 20:43:57 +0000 | [diff] [blame] | 3305 | return skb->queue_mapping != 0; |
David S. Miller | d5a9e24 | 2009-01-27 16:22:11 -0800 | [diff] [blame] | 3306 | } |
| 3307 | |
Denis Kirjanov | 0b3d8e0 | 2013-10-02 05:58:32 +0400 | [diff] [blame] | 3308 | static inline struct sec_path *skb_sec_path(struct sk_buff *skb) |
| 3309 | { |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 3310 | #ifdef CONFIG_XFRM |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 3311 | return skb->sp; |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 3312 | #else |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 3313 | return NULL; |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 3314 | #endif |
Denis Kirjanov | 0b3d8e0 | 2013-10-02 05:58:32 +0400 | [diff] [blame] | 3315 | } |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 3316 | |
Pravin B Shelar | 68c3316 | 2013-02-14 14:02:41 +0000 | [diff] [blame] | 3317 | /* Keeps track of mac header offset relative to skb->head. |
| 3318 | * It is useful for TSO of Tunneling protocol. e.g. GRE. |
| 3319 | * For non-tunnel skb it points to skb_mac_header() and for |
Eric Dumazet | 3347c96 | 2013-10-19 11:42:56 -0700 | [diff] [blame] | 3320 | * tunnel skb it points to outer mac header. |
| 3321 | * Keeps track of level of encapsulation of network headers. |
| 3322 | */ |
Pravin B Shelar | 68c3316 | 2013-02-14 14:02:41 +0000 | [diff] [blame] | 3323 | struct skb_gso_cb { |
Eric Dumazet | 3347c96 | 2013-10-19 11:42:56 -0700 | [diff] [blame] | 3324 | int mac_offset; |
| 3325 | int encap_level; |
Tom Herbert | 7e2b10c | 2014-06-04 17:20:02 -0700 | [diff] [blame] | 3326 | __u16 csum_start; |
Pravin B Shelar | 68c3316 | 2013-02-14 14:02:41 +0000 | [diff] [blame] | 3327 | }; |
| 3328 | #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb) |
| 3329 | |
| 3330 | static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) |
| 3331 | { |
| 3332 | return (skb_mac_header(inner_skb) - inner_skb->head) - |
| 3333 | SKB_GSO_CB(inner_skb)->mac_offset; |
| 3334 | } |
| 3335 | |
Pravin B Shelar | 1e2bd51 | 2013-05-30 06:45:27 +0000 | [diff] [blame] | 3336 | static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) |
| 3337 | { |
| 3338 | int new_headroom, headroom; |
| 3339 | int ret; |
| 3340 | |
| 3341 | headroom = skb_headroom(skb); |
| 3342 | ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); |
| 3343 | if (ret) |
| 3344 | return ret; |
| 3345 | |
| 3346 | new_headroom = skb_headroom(skb); |
| 3347 | SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); |
| 3348 | return 0; |
| 3349 | } |
| 3350 | |
Tom Herbert | 7e2b10c | 2014-06-04 17:20:02 -0700 | [diff] [blame] | 3351 | /* Compute the checksum for a gso segment. First compute the checksum value |
| 3352 | * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and |
| 3353 | * then add in skb->csum (checksum from csum_start to end of packet). |
| 3354 | * skb->csum and csum_start are then updated to reflect the checksum of the |
| 3355 | * resultant packet starting from the transport header-- the resultant checksum |
| 3356 | * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo |
| 3357 | * header. |
| 3358 | */ |
| 3359 | static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) |
| 3360 | { |
| 3361 | int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) - |
Eric Dumazet | c91d460 | 2015-05-15 05:48:07 -0700 | [diff] [blame] | 3362 | skb_transport_offset(skb); |
| 3363 | __wsum partial; |
Tom Herbert | 7e2b10c | 2014-06-04 17:20:02 -0700 | [diff] [blame] | 3364 | |
Eric Dumazet | c91d460 | 2015-05-15 05:48:07 -0700 | [diff] [blame] | 3365 | partial = csum_partial(skb_transport_header(skb), plen, skb->csum); |
Tom Herbert | 7e2b10c | 2014-06-04 17:20:02 -0700 | [diff] [blame] | 3366 | skb->csum = res; |
| 3367 | SKB_GSO_CB(skb)->csum_start -= plen; |
| 3368 | |
Eric Dumazet | c91d460 | 2015-05-15 05:48:07 -0700 | [diff] [blame] | 3369 | return csum_fold(partial); |
Tom Herbert | 7e2b10c | 2014-06-04 17:20:02 -0700 | [diff] [blame] | 3370 | } |
| 3371 | |
David S. Miller | bdcc092 | 2012-03-07 20:53:36 -0500 | [diff] [blame] | 3372 | static inline bool skb_is_gso(const struct sk_buff *skb) |
Herbert Xu | 89114af | 2006-07-08 13:34:32 -0700 | [diff] [blame] | 3373 | { |
| 3374 | return skb_shinfo(skb)->gso_size; |
| 3375 | } |
| 3376 | |
Eric Dumazet | 36a8f39 | 2013-09-29 01:21:32 -0700 | [diff] [blame] | 3377 | /* Note: Should be called only if skb_is_gso(skb) is true */ |
David S. Miller | bdcc092 | 2012-03-07 20:53:36 -0500 | [diff] [blame] | 3378 | static inline bool skb_is_gso_v6(const struct sk_buff *skb) |
Brice Goglin | eabd7e3 | 2007-10-13 12:33:32 +0200 | [diff] [blame] | 3379 | { |
| 3380 | return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; |
| 3381 | } |
| 3382 | |
Joe Perches | 7965bd4 | 2013-09-26 14:48:15 -0700 | [diff] [blame] | 3383 | void __skb_warn_lro_forwarding(const struct sk_buff *skb); |
Ben Hutchings | 4497b07 | 2008-06-19 16:22:28 -0700 | [diff] [blame] | 3384 | |
| 3385 | static inline bool skb_warn_if_lro(const struct sk_buff *skb) |
| 3386 | { |
| 3387 | /* LRO sets gso_size but not gso_type, whereas if GSO is really |
| 3388 | * wanted then gso_type will be set. */ |
Eric Dumazet | 05bdd2f | 2011-10-20 17:45:43 -0400 | [diff] [blame] | 3389 | const struct skb_shared_info *shinfo = skb_shinfo(skb); |
| 3390 | |
Alexander Duyck | b78462e | 2010-06-02 12:24:37 +0000 | [diff] [blame] | 3391 | if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && |
| 3392 | unlikely(shinfo->gso_type == 0)) { |
Ben Hutchings | 4497b07 | 2008-06-19 16:22:28 -0700 | [diff] [blame] | 3393 | __skb_warn_lro_forwarding(skb); |
| 3394 | return true; |
| 3395 | } |
| 3396 | return false; |
| 3397 | } |
| 3398 | |
Herbert Xu | 35fc92a | 2007-03-26 23:22:20 -0700 | [diff] [blame] | 3399 | static inline void skb_forward_csum(struct sk_buff *skb) |
| 3400 | { |
| 3401 | /* Unfortunately we don't support this one. Any brave souls? */ |
| 3402 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
| 3403 | skb->ip_summed = CHECKSUM_NONE; |
| 3404 | } |
| 3405 | |
Eric Dumazet | bc8acf2 | 2010-09-02 13:07:41 -0700 | [diff] [blame] | 3406 | /** |
| 3407 | * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE |
| 3408 | * @skb: skb to check |
| 3409 | * |
| 3410 | * fresh skbs have their ip_summed set to CHECKSUM_NONE. |
| 3411 | * Instead of forcing ip_summed to CHECKSUM_NONE, we can |
| 3412 | * use this helper, to document places where we make this assertion. |
| 3413 | */ |
Eric Dumazet | 05bdd2f | 2011-10-20 17:45:43 -0400 | [diff] [blame] | 3414 | static inline void skb_checksum_none_assert(const struct sk_buff *skb) |
Eric Dumazet | bc8acf2 | 2010-09-02 13:07:41 -0700 | [diff] [blame] | 3415 | { |
| 3416 | #ifdef DEBUG |
| 3417 | BUG_ON(skb->ip_summed != CHECKSUM_NONE); |
| 3418 | #endif |
| 3419 | } |
| 3420 | |
Rusty Russell | f35d9d8 | 2008-02-04 23:49:54 -0500 | [diff] [blame] | 3421 | bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); |
Shirley Ma | a6686f2 | 2011-07-06 12:22:12 +0000 | [diff] [blame] | 3422 | |
Paul Durrant | ed1f50c | 2014-01-09 10:02:46 +0000 | [diff] [blame] | 3423 | int skb_checksum_setup(struct sk_buff *skb, bool recalculate); |
Linus Lüssing | 9afd85c | 2015-05-02 14:01:07 +0200 | [diff] [blame] | 3424 | struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, |
| 3425 | unsigned int transport_len, |
| 3426 | __sum16(*skb_chkf)(struct sk_buff *skb)); |
Paul Durrant | ed1f50c | 2014-01-09 10:02:46 +0000 | [diff] [blame] | 3427 | |
Alexander Duyck | 3a7c1ee4 | 2012-05-03 01:09:42 +0000 | [diff] [blame] | 3428 | /** |
| 3429 | * skb_head_is_locked - Determine if the skb->head is locked down |
| 3430 | * @skb: skb to check |
| 3431 | * |
| 3432 | * The head on skbs build around a head frag can be removed if they are |
| 3433 | * not cloned. This function returns true if the skb head is locked down |
| 3434 | * due to either being allocated via kmalloc, or by being a clone with |
| 3435 | * multiple references to the head. |
| 3436 | */ |
| 3437 | static inline bool skb_head_is_locked(const struct sk_buff *skb) |
| 3438 | { |
| 3439 | return !skb->head_frag || skb_cloned(skb); |
| 3440 | } |
Florian Westphal | fe6cc55 | 2014-02-13 23:09:12 +0100 | [diff] [blame] | 3441 | |
| 3442 | /** |
| 3443 | * skb_gso_network_seglen - Return length of individual segments of a gso packet |
| 3444 | * |
| 3445 | * @skb: GSO skb |
| 3446 | * |
| 3447 | * skb_gso_network_seglen is used to determine the real size of the |
| 3448 | * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP). |
| 3449 | * |
| 3450 | * The MAC/L2 header is not accounted for. |
| 3451 | */ |
| 3452 | static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb) |
| 3453 | { |
| 3454 | unsigned int hdr_len = skb_transport_header(skb) - |
| 3455 | skb_network_header(skb); |
| 3456 | return hdr_len + skb_gso_transport_seglen(skb); |
| 3457 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3458 | #endif /* __KERNEL__ */ |
| 3459 | #endif /* _LINUX_SKBUFF_H */ |