initial commit with v3.6.7
[linux-3.6.7-moxart.git] / include / linux / skbuff.h
blobf3165d26c1636cfeda2837e48da5ead1e20314e3
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
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.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
24 #include <linux/atomic.h>
25 #include <asm/types.h>
26 #include <linux/spinlock.h>
27 #include <linux/net.h>
28 #include <linux/textsearch.h>
29 #include <net/checksum.h>
30 #include <linux/rcupdate.h>
31 #include <linux/dmaengine.h>
32 #include <linux/hrtimer.h>
33 #include <linux/dma-mapping.h>
34 #include <linux/netdev_features.h>
36 /* Don't change this without changing skb_csum_unnecessary! */
37 #define CHECKSUM_NONE 0
38 #define CHECKSUM_UNNECESSARY 1
39 #define CHECKSUM_COMPLETE 2
40 #define CHECKSUM_PARTIAL 3
42 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
44 #define SKB_WITH_OVERHEAD(X) \
45 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
46 #define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
48 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
51 /* return minimum truesize of one skb containing X bytes of data */
52 #define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
56 /* A. Checksumming of received packets by device.
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
67 * COMPLETE: the most generic way. Device supplied checksum of _all_
68 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
70 * is able to produce some skb->csum, it MUST use COMPLETE,
71 * not UNNECESSARY.
73 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
80 * B. Checksumming on output.
82 * NONE: skb is checksummed by protocol or csum is not required.
84 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
85 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
92 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
96 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
98 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
105 * Any questions? No questions, good. --ANK
108 struct net_device;
109 struct scatterlist;
110 struct pipe_inode_info;
112 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
113 struct nf_conntrack {
114 atomic_t use;
116 #endif
118 #ifdef CONFIG_BRIDGE_NETFILTER
119 struct nf_bridge_info {
120 atomic_t use;
121 unsigned int mask;
122 struct net_device *physindev;
123 struct net_device *physoutdev;
124 unsigned long data[32 / sizeof(unsigned long)];
126 #endif
128 struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
133 __u32 qlen;
134 spinlock_t lock;
137 struct sk_buff;
139 /* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
146 #if (65536/PAGE_SIZE + 1) < 16
147 #define MAX_SKB_FRAGS 16UL
148 #else
149 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
150 #endif
152 typedef struct skb_frag_struct skb_frag_t;
154 struct skb_frag_struct {
155 struct {
156 struct page *p;
157 } page;
158 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
159 __u32 page_offset;
160 __u32 size;
161 #else
162 __u16 page_offset;
163 __u16 size;
164 #endif
167 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
169 return frag->size;
172 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
174 frag->size = size;
177 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
179 frag->size += delta;
182 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
184 frag->size -= delta;
187 #define HAVE_HW_TIME_STAMP
190 * struct skb_shared_hwtstamps - hardware time stamps
191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
212 struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
217 /* Definitions for tx_flags in struct skb_shared_info */
218 enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
228 /* device driver supports TX zero-copy buffers */
229 SKBTX_DEV_ZEROCOPY = 1 << 3,
231 /* generate wifi status information (where possible) */
232 SKBTX_WIFI_STATUS = 1 << 4,
236 * The callback notifies userspace to release buffers when skb DMA is done in
237 * lower device, the skb last reference should be 0 when calling this.
238 * The ctx field is used to track device context.
239 * The desc field is used to track userspace buffer index.
241 struct ubuf_info {
242 void (*callback)(struct ubuf_info *);
243 void *ctx;
244 unsigned long desc;
247 /* This data is invariant across clones and lives at
248 * the end of the header data, ie. at skb->end.
250 struct skb_shared_info {
251 unsigned char nr_frags;
252 __u8 tx_flags;
253 unsigned short gso_size;
254 /* Warning: this field is not always filled in (UFO)! */
255 unsigned short gso_segs;
256 unsigned short gso_type;
257 struct sk_buff *frag_list;
258 struct skb_shared_hwtstamps hwtstamps;
259 __be32 ip6_frag_id;
262 * Warning : all fields before dataref are cleared in __alloc_skb()
264 atomic_t dataref;
266 /* Intermediate layers must ensure that destructor_arg
267 * remains valid until skb destructor */
268 void * destructor_arg;
270 /* must be last field, see pskb_expand_head() */
271 skb_frag_t frags[MAX_SKB_FRAGS];
274 /* We divide dataref into two halves. The higher 16 bits hold references
275 * to the payload part of skb->data. The lower 16 bits hold references to
276 * the entire skb->data. A clone of a headerless skb holds the length of
277 * the header in skb->hdr_len.
279 * All users must obey the rule that the skb->data reference count must be
280 * greater than or equal to the payload reference count.
282 * Holding a reference to the payload part means that the user does not
283 * care about modifications to the header part of skb->data.
285 #define SKB_DATAREF_SHIFT 16
286 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
289 enum {
290 SKB_FCLONE_UNAVAILABLE,
291 SKB_FCLONE_ORIG,
292 SKB_FCLONE_CLONE,
295 enum {
296 SKB_GSO_TCPV4 = 1 << 0,
297 SKB_GSO_UDP = 1 << 1,
299 /* This indicates the skb is from an untrusted source. */
300 SKB_GSO_DODGY = 1 << 2,
302 /* This indicates the tcp segment has CWR set. */
303 SKB_GSO_TCP_ECN = 1 << 3,
305 SKB_GSO_TCPV6 = 1 << 4,
307 SKB_GSO_FCOE = 1 << 5,
310 #if BITS_PER_LONG > 32
311 #define NET_SKBUFF_DATA_USES_OFFSET 1
312 #endif
314 #ifdef NET_SKBUFF_DATA_USES_OFFSET
315 typedef unsigned int sk_buff_data_t;
316 #else
317 typedef unsigned char *sk_buff_data_t;
318 #endif
320 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
321 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
322 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
323 #endif
325 /**
326 * struct sk_buff - socket buffer
327 * @next: Next buffer in list
328 * @prev: Previous buffer in list
329 * @tstamp: Time we arrived
330 * @sk: Socket we are owned by
331 * @dev: Device we arrived on/are leaving by
332 * @cb: Control buffer. Free for use by every layer. Put private vars here
333 * @_skb_refdst: destination entry (with norefcount bit)
334 * @sp: the security path, used for xfrm
335 * @len: Length of actual data
336 * @data_len: Data length
337 * @mac_len: Length of link layer header
338 * @hdr_len: writable header length of cloned skb
339 * @csum: Checksum (must include start/offset pair)
340 * @csum_start: Offset from skb->head where checksumming should start
341 * @csum_offset: Offset from csum_start where checksum should be stored
342 * @priority: Packet queueing priority
343 * @local_df: allow local fragmentation
344 * @cloned: Head may be cloned (check refcnt to be sure)
345 * @ip_summed: Driver fed us an IP checksum
346 * @nohdr: Payload reference only, must not modify header
347 * @nfctinfo: Relationship of this skb to the connection
348 * @pkt_type: Packet class
349 * @fclone: skbuff clone status
350 * @ipvs_property: skbuff is owned by ipvs
351 * @peeked: this packet has been seen already, so stats have been
352 * done for it, don't do them again
353 * @nf_trace: netfilter packet trace flag
354 * @protocol: Packet protocol from driver
355 * @destructor: Destruct function
356 * @nfct: Associated connection, if any
357 * @nfct_reasm: netfilter conntrack re-assembly pointer
358 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
359 * @skb_iif: ifindex of device we arrived on
360 * @tc_index: Traffic control index
361 * @tc_verd: traffic control verdict
362 * @rxhash: the packet hash computed on receive
363 * @queue_mapping: Queue mapping for multiqueue devices
364 * @ndisc_nodetype: router type (from link layer)
365 * @ooo_okay: allow the mapping of a socket to a queue to be changed
366 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
367 * ports.
368 * @wifi_acked_valid: wifi_acked was set
369 * @wifi_acked: whether frame was acked on wifi or not
370 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
371 * @dma_cookie: a cookie to one of several possible DMA operations
372 * done by skb DMA functions
373 * @secmark: security marking
374 * @mark: Generic packet mark
375 * @dropcount: total number of sk_receive_queue overflows
376 * @vlan_tci: vlan tag control information
377 * @transport_header: Transport layer header
378 * @network_header: Network layer header
379 * @mac_header: Link layer header
380 * @tail: Tail pointer
381 * @end: End pointer
382 * @head: Head of buffer
383 * @data: Data head pointer
384 * @truesize: Buffer size
385 * @users: User count - see {datagram,tcp}.c
388 struct sk_buff {
389 /* These two members must be first. */
390 struct sk_buff *next;
391 struct sk_buff *prev;
393 ktime_t tstamp;
395 struct sock *sk;
396 struct net_device *dev;
399 * This is the control buffer. It is free to use for every
400 * layer. Please put your private variables there. If you
401 * want to keep them across layers you have to do a skb_clone()
402 * first. This is owned by whoever has the skb queued ATM.
404 char cb[48] __aligned(8);
406 unsigned long _skb_refdst;
407 #ifdef CONFIG_XFRM
408 struct sec_path *sp;
409 #endif
410 unsigned int len,
411 data_len;
412 __u16 mac_len,
413 hdr_len;
414 union {
415 __wsum csum;
416 struct {
417 __u16 csum_start;
418 __u16 csum_offset;
421 __u32 priority;
422 kmemcheck_bitfield_begin(flags1);
423 __u8 local_df:1,
424 cloned:1,
425 ip_summed:2,
426 nohdr:1,
427 nfctinfo:3;
428 __u8 pkt_type:3,
429 fclone:2,
430 ipvs_property:1,
431 peeked:1,
432 nf_trace:1;
433 kmemcheck_bitfield_end(flags1);
434 __be16 protocol;
436 void (*destructor)(struct sk_buff *skb);
437 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
438 struct nf_conntrack *nfct;
439 #endif
440 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
441 struct sk_buff *nfct_reasm;
442 #endif
443 #ifdef CONFIG_BRIDGE_NETFILTER
444 struct nf_bridge_info *nf_bridge;
445 #endif
447 int skb_iif;
449 __u32 rxhash;
451 __u16 vlan_tci;
453 #ifdef CONFIG_NET_SCHED
454 __u16 tc_index; /* traffic control index */
455 #ifdef CONFIG_NET_CLS_ACT
456 __u16 tc_verd; /* traffic control verdict */
457 #endif
458 #endif
460 __u16 queue_mapping;
461 kmemcheck_bitfield_begin(flags2);
462 #ifdef CONFIG_IPV6_NDISC_NODETYPE
463 __u8 ndisc_nodetype:2;
464 #endif
465 __u8 pfmemalloc:1;
466 __u8 ooo_okay:1;
467 __u8 l4_rxhash:1;
468 __u8 wifi_acked_valid:1;
469 __u8 wifi_acked:1;
470 __u8 no_fcs:1;
471 __u8 head_frag:1;
472 /* 8/10 bit hole (depending on ndisc_nodetype presence) */
473 kmemcheck_bitfield_end(flags2);
475 #ifdef CONFIG_NET_DMA
476 dma_cookie_t dma_cookie;
477 #endif
478 #ifdef CONFIG_NETWORK_SECMARK
479 __u32 secmark;
480 #endif
481 union {
482 __u32 mark;
483 __u32 dropcount;
484 __u32 avail_size;
487 sk_buff_data_t transport_header;
488 sk_buff_data_t network_header;
489 sk_buff_data_t mac_header;
490 /* These elements must be at the end, see alloc_skb() for details. */
491 sk_buff_data_t tail;
492 sk_buff_data_t end;
493 unsigned char *head,
494 *data;
495 unsigned int truesize;
496 atomic_t users;
499 #ifdef __KERNEL__
501 * Handling routines are only of interest to the kernel
503 #include <linux/slab.h>
506 #define SKB_ALLOC_FCLONE 0x01
507 #define SKB_ALLOC_RX 0x02
509 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
510 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
512 return unlikely(skb->pfmemalloc);
516 * skb might have a dst pointer attached, refcounted or not.
517 * _skb_refdst low order bit is set if refcount was _not_ taken
519 #define SKB_DST_NOREF 1UL
520 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
523 * skb_dst - returns skb dst_entry
524 * @skb: buffer
526 * Returns skb dst_entry, regardless of reference taken or not.
528 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
530 /* If refdst was not refcounted, check we still are in a
531 * rcu_read_lock section
533 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
534 !rcu_read_lock_held() &&
535 !rcu_read_lock_bh_held());
536 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
540 * skb_dst_set - sets skb dst
541 * @skb: buffer
542 * @dst: dst entry
544 * Sets skb dst, assuming a reference was taken on dst and should
545 * be released by skb_dst_drop()
547 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
549 skb->_skb_refdst = (unsigned long)dst;
552 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
555 * skb_dst_is_noref - Test if skb dst isn't refcounted
556 * @skb: buffer
558 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
560 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
563 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
565 return (struct rtable *)skb_dst(skb);
568 extern void kfree_skb(struct sk_buff *skb);
569 extern void consume_skb(struct sk_buff *skb);
570 extern void __kfree_skb(struct sk_buff *skb);
571 extern struct kmem_cache *skbuff_head_cache;
573 extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
574 extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
575 bool *fragstolen, int *delta_truesize);
577 extern struct sk_buff *__alloc_skb(unsigned int size,
578 gfp_t priority, int flags, int node);
579 extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
580 static inline struct sk_buff *alloc_skb(unsigned int size,
581 gfp_t priority)
583 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
586 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
587 gfp_t priority)
589 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
592 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
593 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
594 extern struct sk_buff *skb_clone(struct sk_buff *skb,
595 gfp_t priority);
596 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
597 gfp_t priority);
598 extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
599 int headroom, gfp_t gfp_mask);
601 extern int pskb_expand_head(struct sk_buff *skb,
602 int nhead, int ntail,
603 gfp_t gfp_mask);
604 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
605 unsigned int headroom);
606 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
607 int newheadroom, int newtailroom,
608 gfp_t priority);
609 extern int skb_to_sgvec(struct sk_buff *skb,
610 struct scatterlist *sg, int offset,
611 int len);
612 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
613 struct sk_buff **trailer);
614 extern int skb_pad(struct sk_buff *skb, int pad);
615 #define dev_kfree_skb(a) consume_skb(a)
617 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
618 int getfrag(void *from, char *to, int offset,
619 int len,int odd, struct sk_buff *skb),
620 void *from, int length);
622 struct skb_seq_state {
623 __u32 lower_offset;
624 __u32 upper_offset;
625 __u32 frag_idx;
626 __u32 stepped_offset;
627 struct sk_buff *root_skb;
628 struct sk_buff *cur_skb;
629 __u8 *frag_data;
632 extern void skb_prepare_seq_read(struct sk_buff *skb,
633 unsigned int from, unsigned int to,
634 struct skb_seq_state *st);
635 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
636 struct skb_seq_state *st);
637 extern void skb_abort_seq_read(struct skb_seq_state *st);
639 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
640 unsigned int to, struct ts_config *config,
641 struct ts_state *state);
643 extern void __skb_get_rxhash(struct sk_buff *skb);
644 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
646 if (!skb->rxhash)
647 __skb_get_rxhash(skb);
649 return skb->rxhash;
652 #ifdef NET_SKBUFF_DATA_USES_OFFSET
653 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
655 return skb->head + skb->end;
658 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
660 return skb->end;
662 #else
663 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
665 return skb->end;
668 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
670 return skb->end - skb->head;
672 #endif
674 /* Internal */
675 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
677 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
679 return &skb_shinfo(skb)->hwtstamps;
683 * skb_queue_empty - check if a queue is empty
684 * @list: queue head
686 * Returns true if the queue is empty, false otherwise.
688 static inline int skb_queue_empty(const struct sk_buff_head *list)
690 return list->next == (struct sk_buff *)list;
694 * skb_queue_is_last - check if skb is the last entry in the queue
695 * @list: queue head
696 * @skb: buffer
698 * Returns true if @skb is the last buffer on the list.
700 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
701 const struct sk_buff *skb)
703 return skb->next == (struct sk_buff *)list;
707 * skb_queue_is_first - check if skb is the first entry in the queue
708 * @list: queue head
709 * @skb: buffer
711 * Returns true if @skb is the first buffer on the list.
713 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
714 const struct sk_buff *skb)
716 return skb->prev == (struct sk_buff *)list;
720 * skb_queue_next - return the next packet in the queue
721 * @list: queue head
722 * @skb: current buffer
724 * Return the next packet in @list after @skb. It is only valid to
725 * call this if skb_queue_is_last() evaluates to false.
727 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
728 const struct sk_buff *skb)
730 /* This BUG_ON may seem severe, but if we just return then we
731 * are going to dereference garbage.
733 BUG_ON(skb_queue_is_last(list, skb));
734 return skb->next;
738 * skb_queue_prev - return the prev packet in the queue
739 * @list: queue head
740 * @skb: current buffer
742 * Return the prev packet in @list before @skb. It is only valid to
743 * call this if skb_queue_is_first() evaluates to false.
745 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
746 const struct sk_buff *skb)
748 /* This BUG_ON may seem severe, but if we just return then we
749 * are going to dereference garbage.
751 BUG_ON(skb_queue_is_first(list, skb));
752 return skb->prev;
756 * skb_get - reference buffer
757 * @skb: buffer to reference
759 * Makes another reference to a socket buffer and returns a pointer
760 * to the buffer.
762 static inline struct sk_buff *skb_get(struct sk_buff *skb)
764 atomic_inc(&skb->users);
765 return skb;
769 * If users == 1, we are the only owner and are can avoid redundant
770 * atomic change.
774 * skb_cloned - is the buffer a clone
775 * @skb: buffer to check
777 * Returns true if the buffer was generated with skb_clone() and is
778 * one of multiple shared copies of the buffer. Cloned buffers are
779 * shared data so must not be written to under normal circumstances.
781 static inline int skb_cloned(const struct sk_buff *skb)
783 return skb->cloned &&
784 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
788 * skb_header_cloned - is the header a clone
789 * @skb: buffer to check
791 * Returns true if modifying the header part of the buffer requires
792 * the data to be copied.
794 static inline int skb_header_cloned(const struct sk_buff *skb)
796 int dataref;
798 if (!skb->cloned)
799 return 0;
801 dataref = atomic_read(&skb_shinfo(skb)->dataref);
802 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
803 return dataref != 1;
807 * skb_header_release - release reference to header
808 * @skb: buffer to operate on
810 * Drop a reference to the header part of the buffer. This is done
811 * by acquiring a payload reference. You must not read from the header
812 * part of skb->data after this.
814 static inline void skb_header_release(struct sk_buff *skb)
816 BUG_ON(skb->nohdr);
817 skb->nohdr = 1;
818 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
822 * skb_shared - is the buffer shared
823 * @skb: buffer to check
825 * Returns true if more than one person has a reference to this
826 * buffer.
828 static inline int skb_shared(const struct sk_buff *skb)
830 return atomic_read(&skb->users) != 1;
834 * skb_share_check - check if buffer is shared and if so clone it
835 * @skb: buffer to check
836 * @pri: priority for memory allocation
838 * If the buffer is shared the buffer is cloned and the old copy
839 * drops a reference. A new clone with a single reference is returned.
840 * If the buffer is not shared the original buffer is returned. When
841 * being called from interrupt status or with spinlocks held pri must
842 * be GFP_ATOMIC.
844 * NULL is returned on a memory allocation failure.
846 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
847 gfp_t pri)
849 might_sleep_if(pri & __GFP_WAIT);
850 if (skb_shared(skb)) {
851 struct sk_buff *nskb = skb_clone(skb, pri);
852 kfree_skb(skb);
853 skb = nskb;
855 return skb;
859 * Copy shared buffers into a new sk_buff. We effectively do COW on
860 * packets to handle cases where we have a local reader and forward
861 * and a couple of other messy ones. The normal one is tcpdumping
862 * a packet thats being forwarded.
866 * skb_unshare - make a copy of a shared buffer
867 * @skb: buffer to check
868 * @pri: priority for memory allocation
870 * If the socket buffer is a clone then this function creates a new
871 * copy of the data, drops a reference count on the old copy and returns
872 * the new copy with the reference count at 1. If the buffer is not a clone
873 * the original buffer is returned. When called with a spinlock held or
874 * from interrupt state @pri must be %GFP_ATOMIC
876 * %NULL is returned on a memory allocation failure.
878 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
879 gfp_t pri)
881 might_sleep_if(pri & __GFP_WAIT);
882 if (skb_cloned(skb)) {
883 struct sk_buff *nskb = skb_copy(skb, pri);
884 kfree_skb(skb); /* Free our shared copy */
885 skb = nskb;
887 return skb;
891 * skb_peek - peek at the head of an &sk_buff_head
892 * @list_: list to peek at
894 * Peek an &sk_buff. Unlike most other operations you _MUST_
895 * be careful with this one. A peek leaves the buffer on the
896 * list and someone else may run off with it. You must hold
897 * the appropriate locks or have a private queue to do this.
899 * Returns %NULL for an empty list or a pointer to the head element.
900 * The reference count is not incremented and the reference is therefore
901 * volatile. Use with caution.
903 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
905 struct sk_buff *skb = list_->next;
907 if (skb == (struct sk_buff *)list_)
908 skb = NULL;
909 return skb;
913 * skb_peek_next - peek skb following the given one from a queue
914 * @skb: skb to start from
915 * @list_: list to peek at
917 * Returns %NULL when the end of the list is met or a pointer to the
918 * next element. The reference count is not incremented and the
919 * reference is therefore volatile. Use with caution.
921 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
922 const struct sk_buff_head *list_)
924 struct sk_buff *next = skb->next;
926 if (next == (struct sk_buff *)list_)
927 next = NULL;
928 return next;
932 * skb_peek_tail - peek at the tail of an &sk_buff_head
933 * @list_: list to peek at
935 * Peek an &sk_buff. Unlike most other operations you _MUST_
936 * be careful with this one. A peek leaves the buffer on the
937 * list and someone else may run off with it. You must hold
938 * the appropriate locks or have a private queue to do this.
940 * Returns %NULL for an empty list or a pointer to the tail element.
941 * The reference count is not incremented and the reference is therefore
942 * volatile. Use with caution.
944 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
946 struct sk_buff *skb = list_->prev;
948 if (skb == (struct sk_buff *)list_)
949 skb = NULL;
950 return skb;
955 * skb_queue_len - get queue length
956 * @list_: list to measure
958 * Return the length of an &sk_buff queue.
960 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
962 return list_->qlen;
966 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
967 * @list: queue to initialize
969 * This initializes only the list and queue length aspects of
970 * an sk_buff_head object. This allows to initialize the list
971 * aspects of an sk_buff_head without reinitializing things like
972 * the spinlock. It can also be used for on-stack sk_buff_head
973 * objects where the spinlock is known to not be used.
975 static inline void __skb_queue_head_init(struct sk_buff_head *list)
977 list->prev = list->next = (struct sk_buff *)list;
978 list->qlen = 0;
982 * This function creates a split out lock class for each invocation;
983 * this is needed for now since a whole lot of users of the skb-queue
984 * infrastructure in drivers have different locking usage (in hardirq)
985 * than the networking core (in softirq only). In the long run either the
986 * network layer or drivers should need annotation to consolidate the
987 * main types of usage into 3 classes.
989 static inline void skb_queue_head_init(struct sk_buff_head *list)
991 spin_lock_init(&list->lock);
992 __skb_queue_head_init(list);
995 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
996 struct lock_class_key *class)
998 skb_queue_head_init(list);
999 lockdep_set_class(&list->lock, class);
1003 * Insert an sk_buff on a list.
1005 * The "__skb_xxxx()" functions are the non-atomic ones that
1006 * can only be called with interrupts disabled.
1008 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1009 static inline void __skb_insert(struct sk_buff *newsk,
1010 struct sk_buff *prev, struct sk_buff *next,
1011 struct sk_buff_head *list)
1013 newsk->next = next;
1014 newsk->prev = prev;
1015 next->prev = prev->next = newsk;
1016 list->qlen++;
1019 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1020 struct sk_buff *prev,
1021 struct sk_buff *next)
1023 struct sk_buff *first = list->next;
1024 struct sk_buff *last = list->prev;
1026 first->prev = prev;
1027 prev->next = first;
1029 last->next = next;
1030 next->prev = last;
1034 * skb_queue_splice - join two skb lists, this is designed for stacks
1035 * @list: the new list to add
1036 * @head: the place to add it in the first list
1038 static inline void skb_queue_splice(const struct sk_buff_head *list,
1039 struct sk_buff_head *head)
1041 if (!skb_queue_empty(list)) {
1042 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1043 head->qlen += list->qlen;
1048 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1049 * @list: the new list to add
1050 * @head: the place to add it in the first list
1052 * The list at @list is reinitialised
1054 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1055 struct sk_buff_head *head)
1057 if (!skb_queue_empty(list)) {
1058 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1059 head->qlen += list->qlen;
1060 __skb_queue_head_init(list);
1065 * skb_queue_splice_tail - join two skb lists, each list being a queue
1066 * @list: the new list to add
1067 * @head: the place to add it in the first list
1069 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1070 struct sk_buff_head *head)
1072 if (!skb_queue_empty(list)) {
1073 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1074 head->qlen += list->qlen;
1079 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1080 * @list: the new list to add
1081 * @head: the place to add it in the first list
1083 * Each of the lists is a queue.
1084 * The list at @list is reinitialised
1086 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1087 struct sk_buff_head *head)
1089 if (!skb_queue_empty(list)) {
1090 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1091 head->qlen += list->qlen;
1092 __skb_queue_head_init(list);
1097 * __skb_queue_after - queue a buffer at the list head
1098 * @list: list to use
1099 * @prev: place after this buffer
1100 * @newsk: buffer to queue
1102 * Queue a buffer int the middle of a list. This function takes no locks
1103 * and you must therefore hold required locks before calling it.
1105 * A buffer cannot be placed on two lists at the same time.
1107 static inline void __skb_queue_after(struct sk_buff_head *list,
1108 struct sk_buff *prev,
1109 struct sk_buff *newsk)
1111 __skb_insert(newsk, prev, prev->next, list);
1114 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1115 struct sk_buff_head *list);
1117 static inline void __skb_queue_before(struct sk_buff_head *list,
1118 struct sk_buff *next,
1119 struct sk_buff *newsk)
1121 __skb_insert(newsk, next->prev, next, list);
1125 * __skb_queue_head - queue a buffer at the list head
1126 * @list: list to use
1127 * @newsk: buffer to queue
1129 * Queue a buffer at the start of a list. This function takes no locks
1130 * and you must therefore hold required locks before calling it.
1132 * A buffer cannot be placed on two lists at the same time.
1134 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1135 static inline void __skb_queue_head(struct sk_buff_head *list,
1136 struct sk_buff *newsk)
1138 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1142 * __skb_queue_tail - queue a buffer at the list tail
1143 * @list: list to use
1144 * @newsk: buffer to queue
1146 * Queue a buffer at the end of a list. This function takes no locks
1147 * and you must therefore hold required locks before calling it.
1149 * A buffer cannot be placed on two lists at the same time.
1151 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1152 static inline void __skb_queue_tail(struct sk_buff_head *list,
1153 struct sk_buff *newsk)
1155 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1159 * remove sk_buff from list. _Must_ be called atomically, and with
1160 * the list known..
1162 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1163 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1165 struct sk_buff *next, *prev;
1167 list->qlen--;
1168 next = skb->next;
1169 prev = skb->prev;
1170 skb->next = skb->prev = NULL;
1171 next->prev = prev;
1172 prev->next = next;
1176 * __skb_dequeue - remove from the head of the queue
1177 * @list: list to dequeue from
1179 * Remove the head of the list. This function does not take any locks
1180 * so must be used with appropriate locks held only. The head item is
1181 * returned or %NULL if the list is empty.
1183 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1184 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1186 struct sk_buff *skb = skb_peek(list);
1187 if (skb)
1188 __skb_unlink(skb, list);
1189 return skb;
1193 * __skb_dequeue_tail - remove from the tail of the queue
1194 * @list: list to dequeue from
1196 * Remove the tail of the list. This function does not take any locks
1197 * so must be used with appropriate locks held only. The tail item is
1198 * returned or %NULL if the list is empty.
1200 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1201 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1203 struct sk_buff *skb = skb_peek_tail(list);
1204 if (skb)
1205 __skb_unlink(skb, list);
1206 return skb;
1210 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1212 return skb->data_len;
1215 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1217 return skb->len - skb->data_len;
1220 static inline int skb_pagelen(const struct sk_buff *skb)
1222 int i, len = 0;
1224 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1225 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1226 return len + skb_headlen(skb);
1230 * __skb_fill_page_desc - initialise a paged fragment in an skb
1231 * @skb: buffer containing fragment to be initialised
1232 * @i: paged fragment index to initialise
1233 * @page: the page to use for this fragment
1234 * @off: the offset to the data with @page
1235 * @size: the length of the data
1237 * Initialises the @i'th fragment of @skb to point to &size bytes at
1238 * offset @off within @page.
1240 * Does not take any additional reference on the fragment.
1242 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1243 struct page *page, int off, int size)
1245 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1248 * Propagate page->pfmemalloc to the skb if we can. The problem is
1249 * that not all callers have unique ownership of the page. If
1250 * pfmemalloc is set, we check the mapping as a mapping implies
1251 * page->index is set (index and pfmemalloc share space).
1252 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1253 * do not lose pfmemalloc information as the pages would not be
1254 * allocated using __GFP_MEMALLOC.
1256 if (page->pfmemalloc && !page->mapping)
1257 skb->pfmemalloc = true;
1258 frag->page.p = page;
1259 frag->page_offset = off;
1260 skb_frag_size_set(frag, size);
1264 * skb_fill_page_desc - initialise a paged fragment in an skb
1265 * @skb: buffer containing fragment to be initialised
1266 * @i: paged fragment index to initialise
1267 * @page: the page to use for this fragment
1268 * @off: the offset to the data with @page
1269 * @size: the length of the data
1271 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1272 * @skb to point to &size bytes at offset @off within @page. In
1273 * addition updates @skb such that @i is the last fragment.
1275 * Does not take any additional reference on the fragment.
1277 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1278 struct page *page, int off, int size)
1280 __skb_fill_page_desc(skb, i, page, off, size);
1281 skb_shinfo(skb)->nr_frags = i + 1;
1284 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1285 int off, int size, unsigned int truesize);
1287 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1288 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1289 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1291 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1292 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1294 return skb->head + skb->tail;
1297 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1299 skb->tail = skb->data - skb->head;
1302 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1304 skb_reset_tail_pointer(skb);
1305 skb->tail += offset;
1307 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1308 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1310 return skb->tail;
1313 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1315 skb->tail = skb->data;
1318 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1320 skb->tail = skb->data + offset;
1323 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1326 * Add data to an sk_buff
1328 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1329 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1331 unsigned char *tmp = skb_tail_pointer(skb);
1332 SKB_LINEAR_ASSERT(skb);
1333 skb->tail += len;
1334 skb->len += len;
1335 return tmp;
1338 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1339 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1341 skb->data -= len;
1342 skb->len += len;
1343 return skb->data;
1346 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1347 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1349 skb->len -= len;
1350 BUG_ON(skb->len < skb->data_len);
1351 return skb->data += len;
1354 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1356 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1359 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1361 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1363 if (len > skb_headlen(skb) &&
1364 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1365 return NULL;
1366 skb->len -= len;
1367 return skb->data += len;
1370 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1372 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1375 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1377 if (likely(len <= skb_headlen(skb)))
1378 return 1;
1379 if (unlikely(len > skb->len))
1380 return 0;
1381 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1385 * skb_headroom - bytes at buffer head
1386 * @skb: buffer to check
1388 * Return the number of bytes of free space at the head of an &sk_buff.
1390 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1392 return skb->data - skb->head;
1396 * skb_tailroom - bytes at buffer end
1397 * @skb: buffer to check
1399 * Return the number of bytes of free space at the tail of an sk_buff
1401 static inline int skb_tailroom(const struct sk_buff *skb)
1403 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1407 * skb_availroom - bytes at buffer end
1408 * @skb: buffer to check
1410 * Return the number of bytes of free space at the tail of an sk_buff
1411 * allocated by sk_stream_alloc()
1413 static inline int skb_availroom(const struct sk_buff *skb)
1415 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1419 * skb_reserve - adjust headroom
1420 * @skb: buffer to alter
1421 * @len: bytes to move
1423 * Increase the headroom of an empty &sk_buff by reducing the tail
1424 * room. This is only allowed for an empty buffer.
1426 static inline void skb_reserve(struct sk_buff *skb, int len)
1428 skb->data += len;
1429 skb->tail += len;
1432 static inline void skb_reset_mac_len(struct sk_buff *skb)
1434 skb->mac_len = skb->network_header - skb->mac_header;
1437 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1438 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1440 return skb->head + skb->transport_header;
1443 static inline void skb_reset_transport_header(struct sk_buff *skb)
1445 skb->transport_header = skb->data - skb->head;
1448 static inline void skb_set_transport_header(struct sk_buff *skb,
1449 const int offset)
1451 skb_reset_transport_header(skb);
1452 skb->transport_header += offset;
1455 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1457 return skb->head + skb->network_header;
1460 static inline void skb_reset_network_header(struct sk_buff *skb)
1462 skb->network_header = skb->data - skb->head;
1465 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1467 skb_reset_network_header(skb);
1468 skb->network_header += offset;
1471 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1473 return skb->head + skb->mac_header;
1476 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1478 return skb->mac_header != ~0U;
1481 static inline void skb_reset_mac_header(struct sk_buff *skb)
1483 skb->mac_header = skb->data - skb->head;
1486 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1488 skb_reset_mac_header(skb);
1489 skb->mac_header += offset;
1492 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1494 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1496 return skb->transport_header;
1499 static inline void skb_reset_transport_header(struct sk_buff *skb)
1501 skb->transport_header = skb->data;
1504 static inline void skb_set_transport_header(struct sk_buff *skb,
1505 const int offset)
1507 skb->transport_header = skb->data + offset;
1510 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1512 return skb->network_header;
1515 static inline void skb_reset_network_header(struct sk_buff *skb)
1517 skb->network_header = skb->data;
1520 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1522 skb->network_header = skb->data + offset;
1525 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1527 return skb->mac_header;
1530 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1532 return skb->mac_header != NULL;
1535 static inline void skb_reset_mac_header(struct sk_buff *skb)
1537 skb->mac_header = skb->data;
1540 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1542 skb->mac_header = skb->data + offset;
1544 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1546 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1548 if (skb_mac_header_was_set(skb)) {
1549 const unsigned char *old_mac = skb_mac_header(skb);
1551 skb_set_mac_header(skb, -skb->mac_len);
1552 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1556 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1558 return skb->csum_start - skb_headroom(skb);
1561 static inline int skb_transport_offset(const struct sk_buff *skb)
1563 return skb_transport_header(skb) - skb->data;
1566 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1568 return skb->transport_header - skb->network_header;
1571 static inline int skb_network_offset(const struct sk_buff *skb)
1573 return skb_network_header(skb) - skb->data;
1576 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1578 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1582 * CPUs often take a performance hit when accessing unaligned memory
1583 * locations. The actual performance hit varies, it can be small if the
1584 * hardware handles it or large if we have to take an exception and fix it
1585 * in software.
1587 * Since an ethernet header is 14 bytes network drivers often end up with
1588 * the IP header at an unaligned offset. The IP header can be aligned by
1589 * shifting the start of the packet by 2 bytes. Drivers should do this
1590 * with:
1592 * skb_reserve(skb, NET_IP_ALIGN);
1594 * The downside to this alignment of the IP header is that the DMA is now
1595 * unaligned. On some architectures the cost of an unaligned DMA is high
1596 * and this cost outweighs the gains made by aligning the IP header.
1598 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1599 * to be overridden.
1601 #ifndef NET_IP_ALIGN
1602 #define NET_IP_ALIGN 2
1603 #endif
1606 * The networking layer reserves some headroom in skb data (via
1607 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1608 * the header has to grow. In the default case, if the header has to grow
1609 * 32 bytes or less we avoid the reallocation.
1611 * Unfortunately this headroom changes the DMA alignment of the resulting
1612 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1613 * on some architectures. An architecture can override this value,
1614 * perhaps setting it to a cacheline in size (since that will maintain
1615 * cacheline alignment of the DMA). It must be a power of 2.
1617 * Various parts of the networking layer expect at least 32 bytes of
1618 * headroom, you should not reduce this.
1620 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1621 * to reduce average number of cache lines per packet.
1622 * get_rps_cpus() for example only access one 64 bytes aligned block :
1623 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1625 #ifndef NET_SKB_PAD
1626 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1627 #endif
1629 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1631 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1633 if (unlikely(skb_is_nonlinear(skb))) {
1634 WARN_ON(1);
1635 return;
1637 skb->len = len;
1638 skb_set_tail_pointer(skb, len);
1641 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1643 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1645 if (skb->data_len)
1646 return ___pskb_trim(skb, len);
1647 __skb_trim(skb, len);
1648 return 0;
1651 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1653 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1657 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1658 * @skb: buffer to alter
1659 * @len: new length
1661 * This is identical to pskb_trim except that the caller knows that
1662 * the skb is not cloned so we should never get an error due to out-
1663 * of-memory.
1665 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1667 int err = pskb_trim(skb, len);
1668 BUG_ON(err);
1672 * skb_orphan - orphan a buffer
1673 * @skb: buffer to orphan
1675 * If a buffer currently has an owner then we call the owner's
1676 * destructor function and make the @skb unowned. The buffer continues
1677 * to exist but is no longer charged to its former owner.
1679 static inline void skb_orphan(struct sk_buff *skb)
1681 if (skb->destructor)
1682 skb->destructor(skb);
1683 skb->destructor = NULL;
1684 skb->sk = NULL;
1688 * skb_orphan_frags - orphan the frags contained in a buffer
1689 * @skb: buffer to orphan frags from
1690 * @gfp_mask: allocation mask for replacement pages
1692 * For each frag in the SKB which needs a destructor (i.e. has an
1693 * owner) create a copy of that frag and release the original
1694 * page by calling the destructor.
1696 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1698 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1699 return 0;
1700 return skb_copy_ubufs(skb, gfp_mask);
1704 * __skb_queue_purge - empty a list
1705 * @list: list to empty
1707 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1708 * the list and one reference dropped. This function does not take the
1709 * list lock and the caller must hold the relevant locks to use it.
1711 extern void skb_queue_purge(struct sk_buff_head *list);
1712 static inline void __skb_queue_purge(struct sk_buff_head *list)
1714 struct sk_buff *skb;
1715 while ((skb = __skb_dequeue(list)) != NULL)
1716 kfree_skb(skb);
1719 extern void *netdev_alloc_frag(unsigned int fragsz);
1721 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1722 unsigned int length,
1723 gfp_t gfp_mask);
1726 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1727 * @dev: network device to receive on
1728 * @length: length to allocate
1730 * Allocate a new &sk_buff and assign it a usage count of one. The
1731 * buffer has unspecified headroom built in. Users should allocate
1732 * the headroom they think they need without accounting for the
1733 * built in space. The built in space is used for optimisations.
1735 * %NULL is returned if there is no free memory. Although this function
1736 * allocates memory it can be called from an interrupt.
1738 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1739 unsigned int length)
1741 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1744 /* legacy helper around __netdev_alloc_skb() */
1745 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1746 gfp_t gfp_mask)
1748 return __netdev_alloc_skb(NULL, length, gfp_mask);
1751 /* legacy helper around netdev_alloc_skb() */
1752 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1754 return netdev_alloc_skb(NULL, length);
1758 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1759 unsigned int length, gfp_t gfp)
1761 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1763 if (NET_IP_ALIGN && skb)
1764 skb_reserve(skb, NET_IP_ALIGN);
1765 return skb;
1768 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1769 unsigned int length)
1771 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1775 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1776 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1777 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1778 * @order: size of the allocation
1780 * Allocate a new page.
1782 * %NULL is returned if there is no free memory.
1784 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1785 struct sk_buff *skb,
1786 unsigned int order)
1788 struct page *page;
1790 gfp_mask |= __GFP_COLD;
1792 if (!(gfp_mask & __GFP_NOMEMALLOC))
1793 gfp_mask |= __GFP_MEMALLOC;
1795 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1796 if (skb && page && page->pfmemalloc)
1797 skb->pfmemalloc = true;
1799 return page;
1803 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1804 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1805 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1807 * Allocate a new page.
1809 * %NULL is returned if there is no free memory.
1811 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1812 struct sk_buff *skb)
1814 return __skb_alloc_pages(gfp_mask, skb, 0);
1818 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1819 * @page: The page that was allocated from skb_alloc_page
1820 * @skb: The skb that may need pfmemalloc set
1822 static inline void skb_propagate_pfmemalloc(struct page *page,
1823 struct sk_buff *skb)
1825 if (page && page->pfmemalloc)
1826 skb->pfmemalloc = true;
1830 * skb_frag_page - retrieve the page refered to by a paged fragment
1831 * @frag: the paged fragment
1833 * Returns the &struct page associated with @frag.
1835 static inline struct page *skb_frag_page(const skb_frag_t *frag)
1837 return frag->page.p;
1841 * __skb_frag_ref - take an addition reference on a paged fragment.
1842 * @frag: the paged fragment
1844 * Takes an additional reference on the paged fragment @frag.
1846 static inline void __skb_frag_ref(skb_frag_t *frag)
1848 get_page(skb_frag_page(frag));
1852 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1853 * @skb: the buffer
1854 * @f: the fragment offset.
1856 * Takes an additional reference on the @f'th paged fragment of @skb.
1858 static inline void skb_frag_ref(struct sk_buff *skb, int f)
1860 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1864 * __skb_frag_unref - release a reference on a paged fragment.
1865 * @frag: the paged fragment
1867 * Releases a reference on the paged fragment @frag.
1869 static inline void __skb_frag_unref(skb_frag_t *frag)
1871 put_page(skb_frag_page(frag));
1875 * skb_frag_unref - release a reference on a paged fragment of an skb.
1876 * @skb: the buffer
1877 * @f: the fragment offset
1879 * Releases a reference on the @f'th paged fragment of @skb.
1881 static inline void skb_frag_unref(struct sk_buff *skb, int f)
1883 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1887 * skb_frag_address - gets the address of the data contained in a paged fragment
1888 * @frag: the paged fragment buffer
1890 * Returns the address of the data within @frag. The page must already
1891 * be mapped.
1893 static inline void *skb_frag_address(const skb_frag_t *frag)
1895 return page_address(skb_frag_page(frag)) + frag->page_offset;
1899 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1900 * @frag: the paged fragment buffer
1902 * Returns the address of the data within @frag. Checks that the page
1903 * is mapped and returns %NULL otherwise.
1905 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1907 void *ptr = page_address(skb_frag_page(frag));
1908 if (unlikely(!ptr))
1909 return NULL;
1911 return ptr + frag->page_offset;
1915 * __skb_frag_set_page - sets the page contained in a paged fragment
1916 * @frag: the paged fragment
1917 * @page: the page to set
1919 * Sets the fragment @frag to contain @page.
1921 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1923 frag->page.p = page;
1927 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1928 * @skb: the buffer
1929 * @f: the fragment offset
1930 * @page: the page to set
1932 * Sets the @f'th fragment of @skb to contain @page.
1934 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1935 struct page *page)
1937 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1941 * skb_frag_dma_map - maps a paged fragment via the DMA API
1942 * @dev: the device to map the fragment to
1943 * @frag: the paged fragment to map
1944 * @offset: the offset within the fragment (starting at the
1945 * fragment's own offset)
1946 * @size: the number of bytes to map
1947 * @dir: the direction of the mapping (%PCI_DMA_*)
1949 * Maps the page associated with @frag to @device.
1951 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1952 const skb_frag_t *frag,
1953 size_t offset, size_t size,
1954 enum dma_data_direction dir)
1956 return dma_map_page(dev, skb_frag_page(frag),
1957 frag->page_offset + offset, size, dir);
1960 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1961 gfp_t gfp_mask)
1963 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1967 * skb_clone_writable - is the header of a clone writable
1968 * @skb: buffer to check
1969 * @len: length up to which to write
1971 * Returns true if modifying the header part of the cloned buffer
1972 * does not requires the data to be copied.
1974 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
1976 return !skb_header_cloned(skb) &&
1977 skb_headroom(skb) + len <= skb->hdr_len;
1980 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1981 int cloned)
1983 int delta = 0;
1985 if (headroom > skb_headroom(skb))
1986 delta = headroom - skb_headroom(skb);
1988 if (delta || cloned)
1989 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1990 GFP_ATOMIC);
1991 return 0;
1995 * skb_cow - copy header of skb when it is required
1996 * @skb: buffer to cow
1997 * @headroom: needed headroom
1999 * If the skb passed lacks sufficient headroom or its data part
2000 * is shared, data is reallocated. If reallocation fails, an error
2001 * is returned and original skb is not changed.
2003 * The result is skb with writable area skb->head...skb->tail
2004 * and at least @headroom of space at head.
2006 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2008 return __skb_cow(skb, headroom, skb_cloned(skb));
2012 * skb_cow_head - skb_cow but only making the head writable
2013 * @skb: buffer to cow
2014 * @headroom: needed headroom
2016 * This function is identical to skb_cow except that we replace the
2017 * skb_cloned check by skb_header_cloned. It should be used when
2018 * you only need to push on some header and do not need to modify
2019 * the data.
2021 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2023 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2027 * skb_padto - pad an skbuff up to a minimal size
2028 * @skb: buffer to pad
2029 * @len: minimal length
2031 * Pads up a buffer to ensure the trailing bytes exist and are
2032 * blanked. If the buffer already contains sufficient data it
2033 * is untouched. Otherwise it is extended. Returns zero on
2034 * success. The skb is freed on error.
2037 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2039 unsigned int size = skb->len;
2040 if (likely(size >= len))
2041 return 0;
2042 return skb_pad(skb, len - size);
2045 static inline int skb_add_data(struct sk_buff *skb,
2046 char __user *from, int copy)
2048 const int off = skb->len;
2050 if (skb->ip_summed == CHECKSUM_NONE) {
2051 int err = 0;
2052 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2053 copy, 0, &err);
2054 if (!err) {
2055 skb->csum = csum_block_add(skb->csum, csum, off);
2056 return 0;
2058 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2059 return 0;
2061 __skb_trim(skb, off);
2062 return -EFAULT;
2065 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2066 const struct page *page, int off)
2068 if (i) {
2069 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2071 return page == skb_frag_page(frag) &&
2072 off == frag->page_offset + skb_frag_size(frag);
2074 return false;
2077 static inline int __skb_linearize(struct sk_buff *skb)
2079 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2083 * skb_linearize - convert paged skb to linear one
2084 * @skb: buffer to linarize
2086 * If there is no free memory -ENOMEM is returned, otherwise zero
2087 * is returned and the old skb data released.
2089 static inline int skb_linearize(struct sk_buff *skb)
2091 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2095 * skb_linearize_cow - make sure skb is linear and writable
2096 * @skb: buffer to process
2098 * If there is no free memory -ENOMEM is returned, otherwise zero
2099 * is returned and the old skb data released.
2101 static inline int skb_linearize_cow(struct sk_buff *skb)
2103 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2104 __skb_linearize(skb) : 0;
2108 * skb_postpull_rcsum - update checksum for received skb after pull
2109 * @skb: buffer to update
2110 * @start: start of data before pull
2111 * @len: length of data pulled
2113 * After doing a pull on a received packet, you need to call this to
2114 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2115 * CHECKSUM_NONE so that it can be recomputed from scratch.
2118 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2119 const void *start, unsigned int len)
2121 if (skb->ip_summed == CHECKSUM_COMPLETE)
2122 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2125 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2128 * pskb_trim_rcsum - trim received skb and update checksum
2129 * @skb: buffer to trim
2130 * @len: new length
2132 * This is exactly the same as pskb_trim except that it ensures the
2133 * checksum of received packets are still valid after the operation.
2136 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2138 if (likely(len >= skb->len))
2139 return 0;
2140 if (skb->ip_summed == CHECKSUM_COMPLETE)
2141 skb->ip_summed = CHECKSUM_NONE;
2142 return __pskb_trim(skb, len);
2145 #define skb_queue_walk(queue, skb) \
2146 for (skb = (queue)->next; \
2147 skb != (struct sk_buff *)(queue); \
2148 skb = skb->next)
2150 #define skb_queue_walk_safe(queue, skb, tmp) \
2151 for (skb = (queue)->next, tmp = skb->next; \
2152 skb != (struct sk_buff *)(queue); \
2153 skb = tmp, tmp = skb->next)
2155 #define skb_queue_walk_from(queue, skb) \
2156 for (; skb != (struct sk_buff *)(queue); \
2157 skb = skb->next)
2159 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2160 for (tmp = skb->next; \
2161 skb != (struct sk_buff *)(queue); \
2162 skb = tmp, tmp = skb->next)
2164 #define skb_queue_reverse_walk(queue, skb) \
2165 for (skb = (queue)->prev; \
2166 skb != (struct sk_buff *)(queue); \
2167 skb = skb->prev)
2169 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2170 for (skb = (queue)->prev, tmp = skb->prev; \
2171 skb != (struct sk_buff *)(queue); \
2172 skb = tmp, tmp = skb->prev)
2174 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2175 for (tmp = skb->prev; \
2176 skb != (struct sk_buff *)(queue); \
2177 skb = tmp, tmp = skb->prev)
2179 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2181 return skb_shinfo(skb)->frag_list != NULL;
2184 static inline void skb_frag_list_init(struct sk_buff *skb)
2186 skb_shinfo(skb)->frag_list = NULL;
2189 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2191 frag->next = skb_shinfo(skb)->frag_list;
2192 skb_shinfo(skb)->frag_list = frag;
2195 #define skb_walk_frags(skb, iter) \
2196 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2198 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2199 int *peeked, int *off, int *err);
2200 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2201 int noblock, int *err);
2202 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2203 struct poll_table_struct *wait);
2204 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2205 int offset, struct iovec *to,
2206 int size);
2207 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
2208 int hlen,
2209 struct iovec *iov);
2210 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2211 int offset,
2212 const struct iovec *from,
2213 int from_offset,
2214 int len);
2215 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2216 int offset,
2217 const struct iovec *to,
2218 int to_offset,
2219 int size);
2220 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2221 extern void skb_free_datagram_locked(struct sock *sk,
2222 struct sk_buff *skb);
2223 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
2224 unsigned int flags);
2225 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2226 int len, __wsum csum);
2227 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2228 void *to, int len);
2229 extern int skb_store_bits(struct sk_buff *skb, int offset,
2230 const void *from, int len);
2231 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
2232 int offset, u8 *to, int len,
2233 __wsum csum);
2234 extern int skb_splice_bits(struct sk_buff *skb,
2235 unsigned int offset,
2236 struct pipe_inode_info *pipe,
2237 unsigned int len,
2238 unsigned int flags);
2239 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2240 extern void skb_split(struct sk_buff *skb,
2241 struct sk_buff *skb1, const u32 len);
2242 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2243 int shiftlen);
2245 extern struct sk_buff *skb_segment(struct sk_buff *skb,
2246 netdev_features_t features);
2248 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2249 int len, void *buffer)
2251 int hlen = skb_headlen(skb);
2253 if (hlen - offset >= len)
2254 return skb->data + offset;
2256 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2257 return NULL;
2259 return buffer;
2262 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2263 void *to,
2264 const unsigned int len)
2266 memcpy(to, skb->data, len);
2269 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2270 const int offset, void *to,
2271 const unsigned int len)
2273 memcpy(to, skb->data + offset, len);
2276 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2277 const void *from,
2278 const unsigned int len)
2280 memcpy(skb->data, from, len);
2283 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2284 const int offset,
2285 const void *from,
2286 const unsigned int len)
2288 memcpy(skb->data + offset, from, len);
2291 extern void skb_init(void);
2293 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2295 return skb->tstamp;
2299 * skb_get_timestamp - get timestamp from a skb
2300 * @skb: skb to get stamp from
2301 * @stamp: pointer to struct timeval to store stamp in
2303 * Timestamps are stored in the skb as offsets to a base timestamp.
2304 * This function converts the offset back to a struct timeval and stores
2305 * it in stamp.
2307 static inline void skb_get_timestamp(const struct sk_buff *skb,
2308 struct timeval *stamp)
2310 *stamp = ktime_to_timeval(skb->tstamp);
2313 static inline void skb_get_timestampns(const struct sk_buff *skb,
2314 struct timespec *stamp)
2316 *stamp = ktime_to_timespec(skb->tstamp);
2319 static inline void __net_timestamp(struct sk_buff *skb)
2321 skb->tstamp = ktime_get_real();
2324 static inline ktime_t net_timedelta(ktime_t t)
2326 return ktime_sub(ktime_get_real(), t);
2329 static inline ktime_t net_invalid_timestamp(void)
2331 return ktime_set(0, 0);
2334 extern void skb_timestamping_init(void);
2336 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2338 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2339 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2341 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2343 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2347 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2349 return false;
2352 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2355 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2357 * PHY drivers may accept clones of transmitted packets for
2358 * timestamping via their phy_driver.txtstamp method. These drivers
2359 * must call this function to return the skb back to the stack, with
2360 * or without a timestamp.
2362 * @skb: clone of the the original outgoing packet
2363 * @hwtstamps: hardware time stamps, may be NULL if not available
2366 void skb_complete_tx_timestamp(struct sk_buff *skb,
2367 struct skb_shared_hwtstamps *hwtstamps);
2370 * skb_tstamp_tx - queue clone of skb with send time stamps
2371 * @orig_skb: the original outgoing packet
2372 * @hwtstamps: hardware time stamps, may be NULL if not available
2374 * If the skb has a socket associated, then this function clones the
2375 * skb (thus sharing the actual data and optional structures), stores
2376 * the optional hardware time stamping information (if non NULL) or
2377 * generates a software time stamp (otherwise), then queues the clone
2378 * to the error queue of the socket. Errors are silently ignored.
2380 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2381 struct skb_shared_hwtstamps *hwtstamps);
2383 static inline void sw_tx_timestamp(struct sk_buff *skb)
2385 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2386 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2387 skb_tstamp_tx(skb, NULL);
2391 * skb_tx_timestamp() - Driver hook for transmit timestamping
2393 * Ethernet MAC Drivers should call this function in their hard_xmit()
2394 * function immediately before giving the sk_buff to the MAC hardware.
2396 * @skb: A socket buffer.
2398 static inline void skb_tx_timestamp(struct sk_buff *skb)
2400 skb_clone_tx_timestamp(skb);
2401 sw_tx_timestamp(skb);
2405 * skb_complete_wifi_ack - deliver skb with wifi status
2407 * @skb: the original outgoing packet
2408 * @acked: ack status
2411 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2413 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2414 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2416 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2418 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2422 * skb_checksum_complete - Calculate checksum of an entire packet
2423 * @skb: packet to process
2425 * This function calculates the checksum over the entire packet plus
2426 * the value of skb->csum. The latter can be used to supply the
2427 * checksum of a pseudo header as used by TCP/UDP. It returns the
2428 * checksum.
2430 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2431 * this function can be used to verify that checksum on received
2432 * packets. In that case the function should return zero if the
2433 * checksum is correct. In particular, this function will return zero
2434 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2435 * hardware has already verified the correctness of the checksum.
2437 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2439 return skb_csum_unnecessary(skb) ?
2440 0 : __skb_checksum_complete(skb);
2443 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2444 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2445 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2447 if (nfct && atomic_dec_and_test(&nfct->use))
2448 nf_conntrack_destroy(nfct);
2450 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2452 if (nfct)
2453 atomic_inc(&nfct->use);
2455 #endif
2456 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2457 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2459 if (skb)
2460 atomic_inc(&skb->users);
2462 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2464 if (skb)
2465 kfree_skb(skb);
2467 #endif
2468 #ifdef CONFIG_BRIDGE_NETFILTER
2469 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2471 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2472 kfree(nf_bridge);
2474 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2476 if (nf_bridge)
2477 atomic_inc(&nf_bridge->use);
2479 #endif /* CONFIG_BRIDGE_NETFILTER */
2480 static inline void nf_reset(struct sk_buff *skb)
2482 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2483 nf_conntrack_put(skb->nfct);
2484 skb->nfct = NULL;
2485 #endif
2486 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2487 nf_conntrack_put_reasm(skb->nfct_reasm);
2488 skb->nfct_reasm = NULL;
2489 #endif
2490 #ifdef CONFIG_BRIDGE_NETFILTER
2491 nf_bridge_put(skb->nf_bridge);
2492 skb->nf_bridge = NULL;
2493 #endif
2496 /* Note: This doesn't put any conntrack and bridge info in dst. */
2497 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2499 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2500 dst->nfct = src->nfct;
2501 nf_conntrack_get(src->nfct);
2502 dst->nfctinfo = src->nfctinfo;
2503 #endif
2504 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2505 dst->nfct_reasm = src->nfct_reasm;
2506 nf_conntrack_get_reasm(src->nfct_reasm);
2507 #endif
2508 #ifdef CONFIG_BRIDGE_NETFILTER
2509 dst->nf_bridge = src->nf_bridge;
2510 nf_bridge_get(src->nf_bridge);
2511 #endif
2514 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2516 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2517 nf_conntrack_put(dst->nfct);
2518 #endif
2519 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2520 nf_conntrack_put_reasm(dst->nfct_reasm);
2521 #endif
2522 #ifdef CONFIG_BRIDGE_NETFILTER
2523 nf_bridge_put(dst->nf_bridge);
2524 #endif
2525 __nf_copy(dst, src);
2528 #ifdef CONFIG_NETWORK_SECMARK
2529 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2531 to->secmark = from->secmark;
2534 static inline void skb_init_secmark(struct sk_buff *skb)
2536 skb->secmark = 0;
2538 #else
2539 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2542 static inline void skb_init_secmark(struct sk_buff *skb)
2544 #endif
2546 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2548 skb->queue_mapping = queue_mapping;
2551 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2553 return skb->queue_mapping;
2556 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2558 to->queue_mapping = from->queue_mapping;
2561 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2563 skb->queue_mapping = rx_queue + 1;
2566 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2568 return skb->queue_mapping - 1;
2571 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2573 return skb->queue_mapping != 0;
2576 extern u16 __skb_tx_hash(const struct net_device *dev,
2577 const struct sk_buff *skb,
2578 unsigned int num_tx_queues);
2580 #ifdef CONFIG_XFRM
2581 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2583 return skb->sp;
2585 #else
2586 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2588 return NULL;
2590 #endif
2592 static inline bool skb_is_gso(const struct sk_buff *skb)
2594 return skb_shinfo(skb)->gso_size;
2597 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2599 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2602 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2604 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2606 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2607 * wanted then gso_type will be set. */
2608 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2610 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2611 unlikely(shinfo->gso_type == 0)) {
2612 __skb_warn_lro_forwarding(skb);
2613 return true;
2615 return false;
2618 static inline void skb_forward_csum(struct sk_buff *skb)
2620 /* Unfortunately we don't support this one. Any brave souls? */
2621 if (skb->ip_summed == CHECKSUM_COMPLETE)
2622 skb->ip_summed = CHECKSUM_NONE;
2626 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2627 * @skb: skb to check
2629 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2630 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2631 * use this helper, to document places where we make this assertion.
2633 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2635 #ifdef DEBUG
2636 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2637 #endif
2640 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2643 * skb_head_is_locked - Determine if the skb->head is locked down
2644 * @skb: skb to check
2646 * The head on skbs build around a head frag can be removed if they are
2647 * not cloned. This function returns true if the skb head is locked down
2648 * due to either being allocated via kmalloc, or by being a clone with
2649 * multiple references to the head.
2651 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2653 return !skb->head_frag || skb_cloned(skb);
2655 #endif /* __KERNEL__ */
2656 #endif /* _LINUX_SKBUFF_H */