tcp: remove Appropriate Byte Count support
[linux-2.6/btrfs-unstable.git] / include / linux / skbuff.h
blob0259b719bebf504d6ab7001a36a68cdebe41dacd
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 zerocopy_success argument is true if zero copy transmit occurred,
239 * false on data copy or out of memory error caused by data copy attempt.
240 * The ctx field is used to track device context.
241 * The desc field is used to track userspace buffer index.
243 struct ubuf_info {
244 void (*callback)(struct ubuf_info *, bool zerocopy_success);
245 void *ctx;
246 unsigned long desc;
249 /* This data is invariant across clones and lives at
250 * the end of the header data, ie. at skb->end.
252 struct skb_shared_info {
253 unsigned char nr_frags;
254 __u8 tx_flags;
255 unsigned short gso_size;
256 /* Warning: this field is not always filled in (UFO)! */
257 unsigned short gso_segs;
258 unsigned short gso_type;
259 struct sk_buff *frag_list;
260 struct skb_shared_hwtstamps hwtstamps;
261 __be32 ip6_frag_id;
264 * Warning : all fields before dataref are cleared in __alloc_skb()
266 atomic_t dataref;
268 /* Intermediate layers must ensure that destructor_arg
269 * remains valid until skb destructor */
270 void * destructor_arg;
272 /* must be last field, see pskb_expand_head() */
273 skb_frag_t frags[MAX_SKB_FRAGS];
276 /* We divide dataref into two halves. The higher 16 bits hold references
277 * to the payload part of skb->data. The lower 16 bits hold references to
278 * the entire skb->data. A clone of a headerless skb holds the length of
279 * the header in skb->hdr_len.
281 * All users must obey the rule that the skb->data reference count must be
282 * greater than or equal to the payload reference count.
284 * Holding a reference to the payload part means that the user does not
285 * care about modifications to the header part of skb->data.
287 #define SKB_DATAREF_SHIFT 16
288 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
291 enum {
292 SKB_FCLONE_UNAVAILABLE,
293 SKB_FCLONE_ORIG,
294 SKB_FCLONE_CLONE,
297 enum {
298 SKB_GSO_TCPV4 = 1 << 0,
299 SKB_GSO_UDP = 1 << 1,
301 /* This indicates the skb is from an untrusted source. */
302 SKB_GSO_DODGY = 1 << 2,
304 /* This indicates the tcp segment has CWR set. */
305 SKB_GSO_TCP_ECN = 1 << 3,
307 SKB_GSO_TCPV6 = 1 << 4,
309 SKB_GSO_FCOE = 1 << 5,
311 /* This indicates at least one fragment might be overwritten
312 * (as in vmsplice(), sendfile() ...)
313 * If we need to compute a TX checksum, we'll need to copy
314 * all frags to avoid possible bad checksum
316 SKB_GSO_SHARED_FRAG = 1 << 6,
319 #if BITS_PER_LONG > 32
320 #define NET_SKBUFF_DATA_USES_OFFSET 1
321 #endif
323 #ifdef NET_SKBUFF_DATA_USES_OFFSET
324 typedef unsigned int sk_buff_data_t;
325 #else
326 typedef unsigned char *sk_buff_data_t;
327 #endif
329 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
330 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
331 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
332 #endif
334 /**
335 * struct sk_buff - socket buffer
336 * @next: Next buffer in list
337 * @prev: Previous buffer in list
338 * @tstamp: Time we arrived
339 * @sk: Socket we are owned by
340 * @dev: Device we arrived on/are leaving by
341 * @cb: Control buffer. Free for use by every layer. Put private vars here
342 * @_skb_refdst: destination entry (with norefcount bit)
343 * @sp: the security path, used for xfrm
344 * @len: Length of actual data
345 * @data_len: Data length
346 * @mac_len: Length of link layer header
347 * @hdr_len: writable header length of cloned skb
348 * @csum: Checksum (must include start/offset pair)
349 * @csum_start: Offset from skb->head where checksumming should start
350 * @csum_offset: Offset from csum_start where checksum should be stored
351 * @priority: Packet queueing priority
352 * @local_df: allow local fragmentation
353 * @cloned: Head may be cloned (check refcnt to be sure)
354 * @ip_summed: Driver fed us an IP checksum
355 * @nohdr: Payload reference only, must not modify header
356 * @nfctinfo: Relationship of this skb to the connection
357 * @pkt_type: Packet class
358 * @fclone: skbuff clone status
359 * @ipvs_property: skbuff is owned by ipvs
360 * @peeked: this packet has been seen already, so stats have been
361 * done for it, don't do them again
362 * @nf_trace: netfilter packet trace flag
363 * @protocol: Packet protocol from driver
364 * @destructor: Destruct function
365 * @nfct: Associated connection, if any
366 * @nfct_reasm: netfilter conntrack re-assembly pointer
367 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
368 * @skb_iif: ifindex of device we arrived on
369 * @tc_index: Traffic control index
370 * @tc_verd: traffic control verdict
371 * @rxhash: the packet hash computed on receive
372 * @queue_mapping: Queue mapping for multiqueue devices
373 * @ndisc_nodetype: router type (from link layer)
374 * @ooo_okay: allow the mapping of a socket to a queue to be changed
375 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
376 * ports.
377 * @wifi_acked_valid: wifi_acked was set
378 * @wifi_acked: whether frame was acked on wifi or not
379 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
380 * @dma_cookie: a cookie to one of several possible DMA operations
381 * done by skb DMA functions
382 * @secmark: security marking
383 * @mark: Generic packet mark
384 * @dropcount: total number of sk_receive_queue overflows
385 * @vlan_tci: vlan tag control information
386 * @inner_transport_header: Inner transport layer header (encapsulation)
387 * @inner_network_header: Network layer header (encapsulation)
388 * @transport_header: Transport layer header
389 * @network_header: Network layer header
390 * @mac_header: Link layer header
391 * @tail: Tail pointer
392 * @end: End pointer
393 * @head: Head of buffer
394 * @data: Data head pointer
395 * @truesize: Buffer size
396 * @users: User count - see {datagram,tcp}.c
399 struct sk_buff {
400 /* These two members must be first. */
401 struct sk_buff *next;
402 struct sk_buff *prev;
404 ktime_t tstamp;
406 struct sock *sk;
407 struct net_device *dev;
410 * This is the control buffer. It is free to use for every
411 * layer. Please put your private variables there. If you
412 * want to keep them across layers you have to do a skb_clone()
413 * first. This is owned by whoever has the skb queued ATM.
415 char cb[48] __aligned(8);
417 unsigned long _skb_refdst;
418 #ifdef CONFIG_XFRM
419 struct sec_path *sp;
420 #endif
421 unsigned int len,
422 data_len;
423 __u16 mac_len,
424 hdr_len;
425 union {
426 __wsum csum;
427 struct {
428 __u16 csum_start;
429 __u16 csum_offset;
432 __u32 priority;
433 kmemcheck_bitfield_begin(flags1);
434 __u8 local_df:1,
435 cloned:1,
436 ip_summed:2,
437 nohdr:1,
438 nfctinfo:3;
439 __u8 pkt_type:3,
440 fclone:2,
441 ipvs_property:1,
442 peeked:1,
443 nf_trace:1;
444 kmemcheck_bitfield_end(flags1);
445 __be16 protocol;
447 void (*destructor)(struct sk_buff *skb);
448 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
449 struct nf_conntrack *nfct;
450 #endif
451 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
452 struct sk_buff *nfct_reasm;
453 #endif
454 #ifdef CONFIG_BRIDGE_NETFILTER
455 struct nf_bridge_info *nf_bridge;
456 #endif
458 int skb_iif;
460 __u32 rxhash;
462 __u16 vlan_tci;
464 #ifdef CONFIG_NET_SCHED
465 __u16 tc_index; /* traffic control index */
466 #ifdef CONFIG_NET_CLS_ACT
467 __u16 tc_verd; /* traffic control verdict */
468 #endif
469 #endif
471 __u16 queue_mapping;
472 kmemcheck_bitfield_begin(flags2);
473 #ifdef CONFIG_IPV6_NDISC_NODETYPE
474 __u8 ndisc_nodetype:2;
475 #endif
476 __u8 pfmemalloc:1;
477 __u8 ooo_okay:1;
478 __u8 l4_rxhash:1;
479 __u8 wifi_acked_valid:1;
480 __u8 wifi_acked:1;
481 __u8 no_fcs:1;
482 __u8 head_frag:1;
483 /* Encapsulation protocol and NIC drivers should use
484 * this flag to indicate to each other if the skb contains
485 * encapsulated packet or not and maybe use the inner packet
486 * headers if needed
488 __u8 encapsulation:1;
489 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
490 kmemcheck_bitfield_end(flags2);
492 #ifdef CONFIG_NET_DMA
493 dma_cookie_t dma_cookie;
494 #endif
495 #ifdef CONFIG_NETWORK_SECMARK
496 __u32 secmark;
497 #endif
498 union {
499 __u32 mark;
500 __u32 dropcount;
501 __u32 avail_size;
504 sk_buff_data_t inner_transport_header;
505 sk_buff_data_t inner_network_header;
506 sk_buff_data_t transport_header;
507 sk_buff_data_t network_header;
508 sk_buff_data_t mac_header;
509 /* These elements must be at the end, see alloc_skb() for details. */
510 sk_buff_data_t tail;
511 sk_buff_data_t end;
512 unsigned char *head,
513 *data;
514 unsigned int truesize;
515 atomic_t users;
518 #ifdef __KERNEL__
520 * Handling routines are only of interest to the kernel
522 #include <linux/slab.h>
525 #define SKB_ALLOC_FCLONE 0x01
526 #define SKB_ALLOC_RX 0x02
528 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
529 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
531 return unlikely(skb->pfmemalloc);
535 * skb might have a dst pointer attached, refcounted or not.
536 * _skb_refdst low order bit is set if refcount was _not_ taken
538 #define SKB_DST_NOREF 1UL
539 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
542 * skb_dst - returns skb dst_entry
543 * @skb: buffer
545 * Returns skb dst_entry, regardless of reference taken or not.
547 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
549 /* If refdst was not refcounted, check we still are in a
550 * rcu_read_lock section
552 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
553 !rcu_read_lock_held() &&
554 !rcu_read_lock_bh_held());
555 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
559 * skb_dst_set - sets skb dst
560 * @skb: buffer
561 * @dst: dst entry
563 * Sets skb dst, assuming a reference was taken on dst and should
564 * be released by skb_dst_drop()
566 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
568 skb->_skb_refdst = (unsigned long)dst;
571 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
574 * skb_dst_is_noref - Test if skb dst isn't refcounted
575 * @skb: buffer
577 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
579 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
582 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
584 return (struct rtable *)skb_dst(skb);
587 extern void kfree_skb(struct sk_buff *skb);
588 extern void skb_tx_error(struct sk_buff *skb);
589 extern void consume_skb(struct sk_buff *skb);
590 extern void __kfree_skb(struct sk_buff *skb);
591 extern struct kmem_cache *skbuff_head_cache;
593 extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
594 extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
595 bool *fragstolen, int *delta_truesize);
597 extern struct sk_buff *__alloc_skb(unsigned int size,
598 gfp_t priority, int flags, int node);
599 extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
600 static inline struct sk_buff *alloc_skb(unsigned int size,
601 gfp_t priority)
603 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
606 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
607 gfp_t priority)
609 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
612 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
613 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
614 extern struct sk_buff *skb_clone(struct sk_buff *skb,
615 gfp_t priority);
616 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
617 gfp_t priority);
618 extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
619 int headroom, gfp_t gfp_mask);
621 extern int pskb_expand_head(struct sk_buff *skb,
622 int nhead, int ntail,
623 gfp_t gfp_mask);
624 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
625 unsigned int headroom);
626 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
627 int newheadroom, int newtailroom,
628 gfp_t priority);
629 extern int skb_to_sgvec(struct sk_buff *skb,
630 struct scatterlist *sg, int offset,
631 int len);
632 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
633 struct sk_buff **trailer);
634 extern int skb_pad(struct sk_buff *skb, int pad);
635 #define dev_kfree_skb(a) consume_skb(a)
637 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
638 int getfrag(void *from, char *to, int offset,
639 int len,int odd, struct sk_buff *skb),
640 void *from, int length);
642 struct skb_seq_state {
643 __u32 lower_offset;
644 __u32 upper_offset;
645 __u32 frag_idx;
646 __u32 stepped_offset;
647 struct sk_buff *root_skb;
648 struct sk_buff *cur_skb;
649 __u8 *frag_data;
652 extern void skb_prepare_seq_read(struct sk_buff *skb,
653 unsigned int from, unsigned int to,
654 struct skb_seq_state *st);
655 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
656 struct skb_seq_state *st);
657 extern void skb_abort_seq_read(struct skb_seq_state *st);
659 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
660 unsigned int to, struct ts_config *config,
661 struct ts_state *state);
663 extern void __skb_get_rxhash(struct sk_buff *skb);
664 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
666 if (!skb->l4_rxhash)
667 __skb_get_rxhash(skb);
669 return skb->rxhash;
672 #ifdef NET_SKBUFF_DATA_USES_OFFSET
673 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
675 return skb->head + skb->end;
678 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
680 return skb->end;
682 #else
683 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
685 return skb->end;
688 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
690 return skb->end - skb->head;
692 #endif
694 /* Internal */
695 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
697 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
699 return &skb_shinfo(skb)->hwtstamps;
703 * skb_queue_empty - check if a queue is empty
704 * @list: queue head
706 * Returns true if the queue is empty, false otherwise.
708 static inline int skb_queue_empty(const struct sk_buff_head *list)
710 return list->next == (struct sk_buff *)list;
714 * skb_queue_is_last - check if skb is the last entry in the queue
715 * @list: queue head
716 * @skb: buffer
718 * Returns true if @skb is the last buffer on the list.
720 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
721 const struct sk_buff *skb)
723 return skb->next == (struct sk_buff *)list;
727 * skb_queue_is_first - check if skb is the first entry in the queue
728 * @list: queue head
729 * @skb: buffer
731 * Returns true if @skb is the first buffer on the list.
733 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
734 const struct sk_buff *skb)
736 return skb->prev == (struct sk_buff *)list;
740 * skb_queue_next - return the next packet in the queue
741 * @list: queue head
742 * @skb: current buffer
744 * Return the next packet in @list after @skb. It is only valid to
745 * call this if skb_queue_is_last() evaluates to false.
747 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
748 const struct sk_buff *skb)
750 /* This BUG_ON may seem severe, but if we just return then we
751 * are going to dereference garbage.
753 BUG_ON(skb_queue_is_last(list, skb));
754 return skb->next;
758 * skb_queue_prev - return the prev packet in the queue
759 * @list: queue head
760 * @skb: current buffer
762 * Return the prev packet in @list before @skb. It is only valid to
763 * call this if skb_queue_is_first() evaluates to false.
765 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
766 const struct sk_buff *skb)
768 /* This BUG_ON may seem severe, but if we just return then we
769 * are going to dereference garbage.
771 BUG_ON(skb_queue_is_first(list, skb));
772 return skb->prev;
776 * skb_get - reference buffer
777 * @skb: buffer to reference
779 * Makes another reference to a socket buffer and returns a pointer
780 * to the buffer.
782 static inline struct sk_buff *skb_get(struct sk_buff *skb)
784 atomic_inc(&skb->users);
785 return skb;
789 * If users == 1, we are the only owner and are can avoid redundant
790 * atomic change.
794 * skb_cloned - is the buffer a clone
795 * @skb: buffer to check
797 * Returns true if the buffer was generated with skb_clone() and is
798 * one of multiple shared copies of the buffer. Cloned buffers are
799 * shared data so must not be written to under normal circumstances.
801 static inline int skb_cloned(const struct sk_buff *skb)
803 return skb->cloned &&
804 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
808 * skb_header_cloned - is the header a clone
809 * @skb: buffer to check
811 * Returns true if modifying the header part of the buffer requires
812 * the data to be copied.
814 static inline int skb_header_cloned(const struct sk_buff *skb)
816 int dataref;
818 if (!skb->cloned)
819 return 0;
821 dataref = atomic_read(&skb_shinfo(skb)->dataref);
822 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
823 return dataref != 1;
827 * skb_header_release - release reference to header
828 * @skb: buffer to operate on
830 * Drop a reference to the header part of the buffer. This is done
831 * by acquiring a payload reference. You must not read from the header
832 * part of skb->data after this.
834 static inline void skb_header_release(struct sk_buff *skb)
836 BUG_ON(skb->nohdr);
837 skb->nohdr = 1;
838 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
842 * skb_shared - is the buffer shared
843 * @skb: buffer to check
845 * Returns true if more than one person has a reference to this
846 * buffer.
848 static inline int skb_shared(const struct sk_buff *skb)
850 return atomic_read(&skb->users) != 1;
854 * skb_share_check - check if buffer is shared and if so clone it
855 * @skb: buffer to check
856 * @pri: priority for memory allocation
858 * If the buffer is shared the buffer is cloned and the old copy
859 * drops a reference. A new clone with a single reference is returned.
860 * If the buffer is not shared the original buffer is returned. When
861 * being called from interrupt status or with spinlocks held pri must
862 * be GFP_ATOMIC.
864 * NULL is returned on a memory allocation failure.
866 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
868 might_sleep_if(pri & __GFP_WAIT);
869 if (skb_shared(skb)) {
870 struct sk_buff *nskb = skb_clone(skb, pri);
872 if (likely(nskb))
873 consume_skb(skb);
874 else
875 kfree_skb(skb);
876 skb = nskb;
878 return skb;
882 * Copy shared buffers into a new sk_buff. We effectively do COW on
883 * packets to handle cases where we have a local reader and forward
884 * and a couple of other messy ones. The normal one is tcpdumping
885 * a packet thats being forwarded.
889 * skb_unshare - make a copy of a shared buffer
890 * @skb: buffer to check
891 * @pri: priority for memory allocation
893 * If the socket buffer is a clone then this function creates a new
894 * copy of the data, drops a reference count on the old copy and returns
895 * the new copy with the reference count at 1. If the buffer is not a clone
896 * the original buffer is returned. When called with a spinlock held or
897 * from interrupt state @pri must be %GFP_ATOMIC
899 * %NULL is returned on a memory allocation failure.
901 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
902 gfp_t pri)
904 might_sleep_if(pri & __GFP_WAIT);
905 if (skb_cloned(skb)) {
906 struct sk_buff *nskb = skb_copy(skb, pri);
907 kfree_skb(skb); /* Free our shared copy */
908 skb = nskb;
910 return skb;
914 * skb_peek - peek at the head of an &sk_buff_head
915 * @list_: list to peek at
917 * Peek an &sk_buff. Unlike most other operations you _MUST_
918 * be careful with this one. A peek leaves the buffer on the
919 * list and someone else may run off with it. You must hold
920 * the appropriate locks or have a private queue to do this.
922 * Returns %NULL for an empty list or a pointer to the head element.
923 * The reference count is not incremented and the reference is therefore
924 * volatile. Use with caution.
926 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
928 struct sk_buff *skb = list_->next;
930 if (skb == (struct sk_buff *)list_)
931 skb = NULL;
932 return skb;
936 * skb_peek_next - peek skb following the given one from a queue
937 * @skb: skb to start from
938 * @list_: list to peek at
940 * Returns %NULL when the end of the list is met or a pointer to the
941 * next element. The reference count is not incremented and the
942 * reference is therefore volatile. Use with caution.
944 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
945 const struct sk_buff_head *list_)
947 struct sk_buff *next = skb->next;
949 if (next == (struct sk_buff *)list_)
950 next = NULL;
951 return next;
955 * skb_peek_tail - peek at the tail of an &sk_buff_head
956 * @list_: list to peek at
958 * Peek an &sk_buff. Unlike most other operations you _MUST_
959 * be careful with this one. A peek leaves the buffer on the
960 * list and someone else may run off with it. You must hold
961 * the appropriate locks or have a private queue to do this.
963 * Returns %NULL for an empty list or a pointer to the tail element.
964 * The reference count is not incremented and the reference is therefore
965 * volatile. Use with caution.
967 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
969 struct sk_buff *skb = list_->prev;
971 if (skb == (struct sk_buff *)list_)
972 skb = NULL;
973 return skb;
978 * skb_queue_len - get queue length
979 * @list_: list to measure
981 * Return the length of an &sk_buff queue.
983 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
985 return list_->qlen;
989 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
990 * @list: queue to initialize
992 * This initializes only the list and queue length aspects of
993 * an sk_buff_head object. This allows to initialize the list
994 * aspects of an sk_buff_head without reinitializing things like
995 * the spinlock. It can also be used for on-stack sk_buff_head
996 * objects where the spinlock is known to not be used.
998 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1000 list->prev = list->next = (struct sk_buff *)list;
1001 list->qlen = 0;
1005 * This function creates a split out lock class for each invocation;
1006 * this is needed for now since a whole lot of users of the skb-queue
1007 * infrastructure in drivers have different locking usage (in hardirq)
1008 * than the networking core (in softirq only). In the long run either the
1009 * network layer or drivers should need annotation to consolidate the
1010 * main types of usage into 3 classes.
1012 static inline void skb_queue_head_init(struct sk_buff_head *list)
1014 spin_lock_init(&list->lock);
1015 __skb_queue_head_init(list);
1018 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1019 struct lock_class_key *class)
1021 skb_queue_head_init(list);
1022 lockdep_set_class(&list->lock, class);
1026 * Insert an sk_buff on a list.
1028 * The "__skb_xxxx()" functions are the non-atomic ones that
1029 * can only be called with interrupts disabled.
1031 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1032 static inline void __skb_insert(struct sk_buff *newsk,
1033 struct sk_buff *prev, struct sk_buff *next,
1034 struct sk_buff_head *list)
1036 newsk->next = next;
1037 newsk->prev = prev;
1038 next->prev = prev->next = newsk;
1039 list->qlen++;
1042 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1043 struct sk_buff *prev,
1044 struct sk_buff *next)
1046 struct sk_buff *first = list->next;
1047 struct sk_buff *last = list->prev;
1049 first->prev = prev;
1050 prev->next = first;
1052 last->next = next;
1053 next->prev = last;
1057 * skb_queue_splice - join two skb lists, this is designed for stacks
1058 * @list: the new list to add
1059 * @head: the place to add it in the first list
1061 static inline void skb_queue_splice(const struct sk_buff_head *list,
1062 struct sk_buff_head *head)
1064 if (!skb_queue_empty(list)) {
1065 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1066 head->qlen += list->qlen;
1071 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1072 * @list: the new list to add
1073 * @head: the place to add it in the first list
1075 * The list at @list is reinitialised
1077 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1078 struct sk_buff_head *head)
1080 if (!skb_queue_empty(list)) {
1081 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1082 head->qlen += list->qlen;
1083 __skb_queue_head_init(list);
1088 * skb_queue_splice_tail - join two skb lists, each list being a queue
1089 * @list: the new list to add
1090 * @head: the place to add it in the first list
1092 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1093 struct sk_buff_head *head)
1095 if (!skb_queue_empty(list)) {
1096 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1097 head->qlen += list->qlen;
1102 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1103 * @list: the new list to add
1104 * @head: the place to add it in the first list
1106 * Each of the lists is a queue.
1107 * The list at @list is reinitialised
1109 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1110 struct sk_buff_head *head)
1112 if (!skb_queue_empty(list)) {
1113 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1114 head->qlen += list->qlen;
1115 __skb_queue_head_init(list);
1120 * __skb_queue_after - queue a buffer at the list head
1121 * @list: list to use
1122 * @prev: place after this buffer
1123 * @newsk: buffer to queue
1125 * Queue a buffer int the middle of a list. This function takes no locks
1126 * and you must therefore hold required locks before calling it.
1128 * A buffer cannot be placed on two lists at the same time.
1130 static inline void __skb_queue_after(struct sk_buff_head *list,
1131 struct sk_buff *prev,
1132 struct sk_buff *newsk)
1134 __skb_insert(newsk, prev, prev->next, list);
1137 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1138 struct sk_buff_head *list);
1140 static inline void __skb_queue_before(struct sk_buff_head *list,
1141 struct sk_buff *next,
1142 struct sk_buff *newsk)
1144 __skb_insert(newsk, next->prev, next, list);
1148 * __skb_queue_head - queue a buffer at the list head
1149 * @list: list to use
1150 * @newsk: buffer to queue
1152 * Queue a buffer at the start of a list. This function takes no locks
1153 * and you must therefore hold required locks before calling it.
1155 * A buffer cannot be placed on two lists at the same time.
1157 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1158 static inline void __skb_queue_head(struct sk_buff_head *list,
1159 struct sk_buff *newsk)
1161 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1165 * __skb_queue_tail - queue a buffer at the list tail
1166 * @list: list to use
1167 * @newsk: buffer to queue
1169 * Queue a buffer at the end of a list. This function takes no locks
1170 * and you must therefore hold required locks before calling it.
1172 * A buffer cannot be placed on two lists at the same time.
1174 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1175 static inline void __skb_queue_tail(struct sk_buff_head *list,
1176 struct sk_buff *newsk)
1178 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1182 * remove sk_buff from list. _Must_ be called atomically, and with
1183 * the list known..
1185 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1186 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1188 struct sk_buff *next, *prev;
1190 list->qlen--;
1191 next = skb->next;
1192 prev = skb->prev;
1193 skb->next = skb->prev = NULL;
1194 next->prev = prev;
1195 prev->next = next;
1199 * __skb_dequeue - remove from the head of the queue
1200 * @list: list to dequeue from
1202 * Remove the head of the list. This function does not take any locks
1203 * so must be used with appropriate locks held only. The head item is
1204 * returned or %NULL if the list is empty.
1206 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1207 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1209 struct sk_buff *skb = skb_peek(list);
1210 if (skb)
1211 __skb_unlink(skb, list);
1212 return skb;
1216 * __skb_dequeue_tail - remove from the tail of the queue
1217 * @list: list to dequeue from
1219 * Remove the tail of the list. This function does not take any locks
1220 * so must be used with appropriate locks held only. The tail item is
1221 * returned or %NULL if the list is empty.
1223 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1224 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1226 struct sk_buff *skb = skb_peek_tail(list);
1227 if (skb)
1228 __skb_unlink(skb, list);
1229 return skb;
1233 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1235 return skb->data_len;
1238 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1240 return skb->len - skb->data_len;
1243 static inline int skb_pagelen(const struct sk_buff *skb)
1245 int i, len = 0;
1247 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1248 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1249 return len + skb_headlen(skb);
1253 * __skb_fill_page_desc - initialise a paged fragment in an skb
1254 * @skb: buffer containing fragment to be initialised
1255 * @i: paged fragment index to initialise
1256 * @page: the page to use for this fragment
1257 * @off: the offset to the data with @page
1258 * @size: the length of the data
1260 * Initialises the @i'th fragment of @skb to point to &size bytes at
1261 * offset @off within @page.
1263 * Does not take any additional reference on the fragment.
1265 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1266 struct page *page, int off, int size)
1268 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1271 * Propagate page->pfmemalloc to the skb if we can. The problem is
1272 * that not all callers have unique ownership of the page. If
1273 * pfmemalloc is set, we check the mapping as a mapping implies
1274 * page->index is set (index and pfmemalloc share space).
1275 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1276 * do not lose pfmemalloc information as the pages would not be
1277 * allocated using __GFP_MEMALLOC.
1279 if (page->pfmemalloc && !page->mapping)
1280 skb->pfmemalloc = true;
1281 frag->page.p = page;
1282 frag->page_offset = off;
1283 skb_frag_size_set(frag, size);
1287 * skb_fill_page_desc - initialise a paged fragment in an skb
1288 * @skb: buffer containing fragment to be initialised
1289 * @i: paged fragment index to initialise
1290 * @page: the page to use for this fragment
1291 * @off: the offset to the data with @page
1292 * @size: the length of the data
1294 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1295 * @skb to point to &size bytes at offset @off within @page. In
1296 * addition updates @skb such that @i is the last fragment.
1298 * Does not take any additional reference on the fragment.
1300 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1301 struct page *page, int off, int size)
1303 __skb_fill_page_desc(skb, i, page, off, size);
1304 skb_shinfo(skb)->nr_frags = i + 1;
1307 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1308 int off, int size, unsigned int truesize);
1310 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1311 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1312 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1314 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1315 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1317 return skb->head + skb->tail;
1320 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1322 skb->tail = skb->data - skb->head;
1325 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1327 skb_reset_tail_pointer(skb);
1328 skb->tail += offset;
1330 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1331 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1333 return skb->tail;
1336 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1338 skb->tail = skb->data;
1341 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1343 skb->tail = skb->data + offset;
1346 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1349 * Add data to an sk_buff
1351 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1352 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1354 unsigned char *tmp = skb_tail_pointer(skb);
1355 SKB_LINEAR_ASSERT(skb);
1356 skb->tail += len;
1357 skb->len += len;
1358 return tmp;
1361 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1362 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1364 skb->data -= len;
1365 skb->len += len;
1366 return skb->data;
1369 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1370 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1372 skb->len -= len;
1373 BUG_ON(skb->len < skb->data_len);
1374 return skb->data += len;
1377 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1379 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1382 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1384 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1386 if (len > skb_headlen(skb) &&
1387 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1388 return NULL;
1389 skb->len -= len;
1390 return skb->data += len;
1393 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1395 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1398 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1400 if (likely(len <= skb_headlen(skb)))
1401 return 1;
1402 if (unlikely(len > skb->len))
1403 return 0;
1404 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1408 * skb_headroom - bytes at buffer head
1409 * @skb: buffer to check
1411 * Return the number of bytes of free space at the head of an &sk_buff.
1413 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1415 return skb->data - skb->head;
1419 * skb_tailroom - bytes at buffer end
1420 * @skb: buffer to check
1422 * Return the number of bytes of free space at the tail of an sk_buff
1424 static inline int skb_tailroom(const struct sk_buff *skb)
1426 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1430 * skb_availroom - bytes at buffer end
1431 * @skb: buffer to check
1433 * Return the number of bytes of free space at the tail of an sk_buff
1434 * allocated by sk_stream_alloc()
1436 static inline int skb_availroom(const struct sk_buff *skb)
1438 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1442 * skb_reserve - adjust headroom
1443 * @skb: buffer to alter
1444 * @len: bytes to move
1446 * Increase the headroom of an empty &sk_buff by reducing the tail
1447 * room. This is only allowed for an empty buffer.
1449 static inline void skb_reserve(struct sk_buff *skb, int len)
1451 skb->data += len;
1452 skb->tail += len;
1455 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1457 skb->inner_network_header = skb->network_header;
1458 skb->inner_transport_header = skb->transport_header;
1461 static inline void skb_reset_mac_len(struct sk_buff *skb)
1463 skb->mac_len = skb->network_header - skb->mac_header;
1466 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1467 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1468 *skb)
1470 return skb->head + skb->inner_transport_header;
1473 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1475 skb->inner_transport_header = skb->data - skb->head;
1478 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1479 const int offset)
1481 skb_reset_inner_transport_header(skb);
1482 skb->inner_transport_header += offset;
1485 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1487 return skb->head + skb->inner_network_header;
1490 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1492 skb->inner_network_header = skb->data - skb->head;
1495 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1496 const int offset)
1498 skb_reset_inner_network_header(skb);
1499 skb->inner_network_header += offset;
1502 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1504 return skb->transport_header != ~0U;
1507 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1509 return skb->head + skb->transport_header;
1512 static inline void skb_reset_transport_header(struct sk_buff *skb)
1514 skb->transport_header = skb->data - skb->head;
1517 static inline void skb_set_transport_header(struct sk_buff *skb,
1518 const int offset)
1520 skb_reset_transport_header(skb);
1521 skb->transport_header += offset;
1524 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1526 return skb->head + skb->network_header;
1529 static inline void skb_reset_network_header(struct sk_buff *skb)
1531 skb->network_header = skb->data - skb->head;
1534 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1536 skb_reset_network_header(skb);
1537 skb->network_header += offset;
1540 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1542 return skb->head + skb->mac_header;
1545 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1547 return skb->mac_header != ~0U;
1550 static inline void skb_reset_mac_header(struct sk_buff *skb)
1552 skb->mac_header = skb->data - skb->head;
1555 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1557 skb_reset_mac_header(skb);
1558 skb->mac_header += offset;
1561 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1562 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1563 *skb)
1565 return skb->inner_transport_header;
1568 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1570 skb->inner_transport_header = skb->data;
1573 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1574 const int offset)
1576 skb->inner_transport_header = skb->data + offset;
1579 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1581 return skb->inner_network_header;
1584 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1586 skb->inner_network_header = skb->data;
1589 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1590 const int offset)
1592 skb->inner_network_header = skb->data + offset;
1595 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1597 return skb->transport_header != NULL;
1600 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1602 return skb->transport_header;
1605 static inline void skb_reset_transport_header(struct sk_buff *skb)
1607 skb->transport_header = skb->data;
1610 static inline void skb_set_transport_header(struct sk_buff *skb,
1611 const int offset)
1613 skb->transport_header = skb->data + offset;
1616 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1618 return skb->network_header;
1621 static inline void skb_reset_network_header(struct sk_buff *skb)
1623 skb->network_header = skb->data;
1626 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1628 skb->network_header = skb->data + offset;
1631 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1633 return skb->mac_header;
1636 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1638 return skb->mac_header != NULL;
1641 static inline void skb_reset_mac_header(struct sk_buff *skb)
1643 skb->mac_header = skb->data;
1646 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1648 skb->mac_header = skb->data + offset;
1650 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1652 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1654 if (skb_mac_header_was_set(skb)) {
1655 const unsigned char *old_mac = skb_mac_header(skb);
1657 skb_set_mac_header(skb, -skb->mac_len);
1658 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1662 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1664 return skb->csum_start - skb_headroom(skb);
1667 static inline int skb_transport_offset(const struct sk_buff *skb)
1669 return skb_transport_header(skb) - skb->data;
1672 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1674 return skb->transport_header - skb->network_header;
1677 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1679 return skb->inner_transport_header - skb->inner_network_header;
1682 static inline int skb_network_offset(const struct sk_buff *skb)
1684 return skb_network_header(skb) - skb->data;
1687 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1689 return skb_inner_network_header(skb) - skb->data;
1692 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1694 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1698 * CPUs often take a performance hit when accessing unaligned memory
1699 * locations. The actual performance hit varies, it can be small if the
1700 * hardware handles it or large if we have to take an exception and fix it
1701 * in software.
1703 * Since an ethernet header is 14 bytes network drivers often end up with
1704 * the IP header at an unaligned offset. The IP header can be aligned by
1705 * shifting the start of the packet by 2 bytes. Drivers should do this
1706 * with:
1708 * skb_reserve(skb, NET_IP_ALIGN);
1710 * The downside to this alignment of the IP header is that the DMA is now
1711 * unaligned. On some architectures the cost of an unaligned DMA is high
1712 * and this cost outweighs the gains made by aligning the IP header.
1714 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1715 * to be overridden.
1717 #ifndef NET_IP_ALIGN
1718 #define NET_IP_ALIGN 2
1719 #endif
1722 * The networking layer reserves some headroom in skb data (via
1723 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1724 * the header has to grow. In the default case, if the header has to grow
1725 * 32 bytes or less we avoid the reallocation.
1727 * Unfortunately this headroom changes the DMA alignment of the resulting
1728 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1729 * on some architectures. An architecture can override this value,
1730 * perhaps setting it to a cacheline in size (since that will maintain
1731 * cacheline alignment of the DMA). It must be a power of 2.
1733 * Various parts of the networking layer expect at least 32 bytes of
1734 * headroom, you should not reduce this.
1736 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1737 * to reduce average number of cache lines per packet.
1738 * get_rps_cpus() for example only access one 64 bytes aligned block :
1739 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1741 #ifndef NET_SKB_PAD
1742 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1743 #endif
1745 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1747 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1749 if (unlikely(skb_is_nonlinear(skb))) {
1750 WARN_ON(1);
1751 return;
1753 skb->len = len;
1754 skb_set_tail_pointer(skb, len);
1757 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1759 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1761 if (skb->data_len)
1762 return ___pskb_trim(skb, len);
1763 __skb_trim(skb, len);
1764 return 0;
1767 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1769 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1773 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1774 * @skb: buffer to alter
1775 * @len: new length
1777 * This is identical to pskb_trim except that the caller knows that
1778 * the skb is not cloned so we should never get an error due to out-
1779 * of-memory.
1781 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1783 int err = pskb_trim(skb, len);
1784 BUG_ON(err);
1788 * skb_orphan - orphan a buffer
1789 * @skb: buffer to orphan
1791 * If a buffer currently has an owner then we call the owner's
1792 * destructor function and make the @skb unowned. The buffer continues
1793 * to exist but is no longer charged to its former owner.
1795 static inline void skb_orphan(struct sk_buff *skb)
1797 if (skb->destructor)
1798 skb->destructor(skb);
1799 skb->destructor = NULL;
1800 skb->sk = NULL;
1804 * skb_orphan_frags - orphan the frags contained in a buffer
1805 * @skb: buffer to orphan frags from
1806 * @gfp_mask: allocation mask for replacement pages
1808 * For each frag in the SKB which needs a destructor (i.e. has an
1809 * owner) create a copy of that frag and release the original
1810 * page by calling the destructor.
1812 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1814 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1815 return 0;
1816 return skb_copy_ubufs(skb, gfp_mask);
1820 * __skb_queue_purge - empty a list
1821 * @list: list to empty
1823 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1824 * the list and one reference dropped. This function does not take the
1825 * list lock and the caller must hold the relevant locks to use it.
1827 extern void skb_queue_purge(struct sk_buff_head *list);
1828 static inline void __skb_queue_purge(struct sk_buff_head *list)
1830 struct sk_buff *skb;
1831 while ((skb = __skb_dequeue(list)) != NULL)
1832 kfree_skb(skb);
1835 extern void *netdev_alloc_frag(unsigned int fragsz);
1837 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1838 unsigned int length,
1839 gfp_t gfp_mask);
1842 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1843 * @dev: network device to receive on
1844 * @length: length to allocate
1846 * Allocate a new &sk_buff and assign it a usage count of one. The
1847 * buffer has unspecified headroom built in. Users should allocate
1848 * the headroom they think they need without accounting for the
1849 * built in space. The built in space is used for optimisations.
1851 * %NULL is returned if there is no free memory. Although this function
1852 * allocates memory it can be called from an interrupt.
1854 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1855 unsigned int length)
1857 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1860 /* legacy helper around __netdev_alloc_skb() */
1861 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1862 gfp_t gfp_mask)
1864 return __netdev_alloc_skb(NULL, length, gfp_mask);
1867 /* legacy helper around netdev_alloc_skb() */
1868 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1870 return netdev_alloc_skb(NULL, length);
1874 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1875 unsigned int length, gfp_t gfp)
1877 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1879 if (NET_IP_ALIGN && skb)
1880 skb_reserve(skb, NET_IP_ALIGN);
1881 return skb;
1884 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1885 unsigned int length)
1887 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1891 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1892 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1893 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1894 * @order: size of the allocation
1896 * Allocate a new page.
1898 * %NULL is returned if there is no free memory.
1900 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1901 struct sk_buff *skb,
1902 unsigned int order)
1904 struct page *page;
1906 gfp_mask |= __GFP_COLD;
1908 if (!(gfp_mask & __GFP_NOMEMALLOC))
1909 gfp_mask |= __GFP_MEMALLOC;
1911 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1912 if (skb && page && page->pfmemalloc)
1913 skb->pfmemalloc = true;
1915 return page;
1919 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1920 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1921 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1923 * Allocate a new page.
1925 * %NULL is returned if there is no free memory.
1927 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1928 struct sk_buff *skb)
1930 return __skb_alloc_pages(gfp_mask, skb, 0);
1934 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1935 * @page: The page that was allocated from skb_alloc_page
1936 * @skb: The skb that may need pfmemalloc set
1938 static inline void skb_propagate_pfmemalloc(struct page *page,
1939 struct sk_buff *skb)
1941 if (page && page->pfmemalloc)
1942 skb->pfmemalloc = true;
1946 * skb_frag_page - retrieve the page refered to by a paged fragment
1947 * @frag: the paged fragment
1949 * Returns the &struct page associated with @frag.
1951 static inline struct page *skb_frag_page(const skb_frag_t *frag)
1953 return frag->page.p;
1957 * __skb_frag_ref - take an addition reference on a paged fragment.
1958 * @frag: the paged fragment
1960 * Takes an additional reference on the paged fragment @frag.
1962 static inline void __skb_frag_ref(skb_frag_t *frag)
1964 get_page(skb_frag_page(frag));
1968 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1969 * @skb: the buffer
1970 * @f: the fragment offset.
1972 * Takes an additional reference on the @f'th paged fragment of @skb.
1974 static inline void skb_frag_ref(struct sk_buff *skb, int f)
1976 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1980 * __skb_frag_unref - release a reference on a paged fragment.
1981 * @frag: the paged fragment
1983 * Releases a reference on the paged fragment @frag.
1985 static inline void __skb_frag_unref(skb_frag_t *frag)
1987 put_page(skb_frag_page(frag));
1991 * skb_frag_unref - release a reference on a paged fragment of an skb.
1992 * @skb: the buffer
1993 * @f: the fragment offset
1995 * Releases a reference on the @f'th paged fragment of @skb.
1997 static inline void skb_frag_unref(struct sk_buff *skb, int f)
1999 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2003 * skb_frag_address - gets the address of the data contained in a paged fragment
2004 * @frag: the paged fragment buffer
2006 * Returns the address of the data within @frag. The page must already
2007 * be mapped.
2009 static inline void *skb_frag_address(const skb_frag_t *frag)
2011 return page_address(skb_frag_page(frag)) + frag->page_offset;
2015 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2016 * @frag: the paged fragment buffer
2018 * Returns the address of the data within @frag. Checks that the page
2019 * is mapped and returns %NULL otherwise.
2021 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2023 void *ptr = page_address(skb_frag_page(frag));
2024 if (unlikely(!ptr))
2025 return NULL;
2027 return ptr + frag->page_offset;
2031 * __skb_frag_set_page - sets the page contained in a paged fragment
2032 * @frag: the paged fragment
2033 * @page: the page to set
2035 * Sets the fragment @frag to contain @page.
2037 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2039 frag->page.p = page;
2043 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2044 * @skb: the buffer
2045 * @f: the fragment offset
2046 * @page: the page to set
2048 * Sets the @f'th fragment of @skb to contain @page.
2050 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2051 struct page *page)
2053 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2057 * skb_frag_dma_map - maps a paged fragment via the DMA API
2058 * @dev: the device to map the fragment to
2059 * @frag: the paged fragment to map
2060 * @offset: the offset within the fragment (starting at the
2061 * fragment's own offset)
2062 * @size: the number of bytes to map
2063 * @dir: the direction of the mapping (%PCI_DMA_*)
2065 * Maps the page associated with @frag to @device.
2067 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2068 const skb_frag_t *frag,
2069 size_t offset, size_t size,
2070 enum dma_data_direction dir)
2072 return dma_map_page(dev, skb_frag_page(frag),
2073 frag->page_offset + offset, size, dir);
2076 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2077 gfp_t gfp_mask)
2079 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2083 * skb_clone_writable - is the header of a clone writable
2084 * @skb: buffer to check
2085 * @len: length up to which to write
2087 * Returns true if modifying the header part of the cloned buffer
2088 * does not requires the data to be copied.
2090 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2092 return !skb_header_cloned(skb) &&
2093 skb_headroom(skb) + len <= skb->hdr_len;
2096 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2097 int cloned)
2099 int delta = 0;
2101 if (headroom > skb_headroom(skb))
2102 delta = headroom - skb_headroom(skb);
2104 if (delta || cloned)
2105 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2106 GFP_ATOMIC);
2107 return 0;
2111 * skb_cow - copy header of skb when it is required
2112 * @skb: buffer to cow
2113 * @headroom: needed headroom
2115 * If the skb passed lacks sufficient headroom or its data part
2116 * is shared, data is reallocated. If reallocation fails, an error
2117 * is returned and original skb is not changed.
2119 * The result is skb with writable area skb->head...skb->tail
2120 * and at least @headroom of space at head.
2122 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2124 return __skb_cow(skb, headroom, skb_cloned(skb));
2128 * skb_cow_head - skb_cow but only making the head writable
2129 * @skb: buffer to cow
2130 * @headroom: needed headroom
2132 * This function is identical to skb_cow except that we replace the
2133 * skb_cloned check by skb_header_cloned. It should be used when
2134 * you only need to push on some header and do not need to modify
2135 * the data.
2137 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2139 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2143 * skb_padto - pad an skbuff up to a minimal size
2144 * @skb: buffer to pad
2145 * @len: minimal length
2147 * Pads up a buffer to ensure the trailing bytes exist and are
2148 * blanked. If the buffer already contains sufficient data it
2149 * is untouched. Otherwise it is extended. Returns zero on
2150 * success. The skb is freed on error.
2153 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2155 unsigned int size = skb->len;
2156 if (likely(size >= len))
2157 return 0;
2158 return skb_pad(skb, len - size);
2161 static inline int skb_add_data(struct sk_buff *skb,
2162 char __user *from, int copy)
2164 const int off = skb->len;
2166 if (skb->ip_summed == CHECKSUM_NONE) {
2167 int err = 0;
2168 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2169 copy, 0, &err);
2170 if (!err) {
2171 skb->csum = csum_block_add(skb->csum, csum, off);
2172 return 0;
2174 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2175 return 0;
2177 __skb_trim(skb, off);
2178 return -EFAULT;
2181 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2182 const struct page *page, int off)
2184 if (i) {
2185 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2187 return page == skb_frag_page(frag) &&
2188 off == frag->page_offset + skb_frag_size(frag);
2190 return false;
2193 static inline int __skb_linearize(struct sk_buff *skb)
2195 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2199 * skb_linearize - convert paged skb to linear one
2200 * @skb: buffer to linarize
2202 * If there is no free memory -ENOMEM is returned, otherwise zero
2203 * is returned and the old skb data released.
2205 static inline int skb_linearize(struct sk_buff *skb)
2207 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2211 * skb_has_shared_frag - can any frag be overwritten
2212 * @skb: buffer to test
2214 * Return true if the skb has at least one frag that might be modified
2215 * by an external entity (as in vmsplice()/sendfile())
2217 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2219 return skb_shinfo(skb)->gso_type & SKB_GSO_SHARED_FRAG;
2223 * skb_linearize_cow - make sure skb is linear and writable
2224 * @skb: buffer to process
2226 * If there is no free memory -ENOMEM is returned, otherwise zero
2227 * is returned and the old skb data released.
2229 static inline int skb_linearize_cow(struct sk_buff *skb)
2231 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2232 __skb_linearize(skb) : 0;
2236 * skb_postpull_rcsum - update checksum for received skb after pull
2237 * @skb: buffer to update
2238 * @start: start of data before pull
2239 * @len: length of data pulled
2241 * After doing a pull on a received packet, you need to call this to
2242 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2243 * CHECKSUM_NONE so that it can be recomputed from scratch.
2246 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2247 const void *start, unsigned int len)
2249 if (skb->ip_summed == CHECKSUM_COMPLETE)
2250 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2253 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2256 * pskb_trim_rcsum - trim received skb and update checksum
2257 * @skb: buffer to trim
2258 * @len: new length
2260 * This is exactly the same as pskb_trim except that it ensures the
2261 * checksum of received packets are still valid after the operation.
2264 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2266 if (likely(len >= skb->len))
2267 return 0;
2268 if (skb->ip_summed == CHECKSUM_COMPLETE)
2269 skb->ip_summed = CHECKSUM_NONE;
2270 return __pskb_trim(skb, len);
2273 #define skb_queue_walk(queue, skb) \
2274 for (skb = (queue)->next; \
2275 skb != (struct sk_buff *)(queue); \
2276 skb = skb->next)
2278 #define skb_queue_walk_safe(queue, skb, tmp) \
2279 for (skb = (queue)->next, tmp = skb->next; \
2280 skb != (struct sk_buff *)(queue); \
2281 skb = tmp, tmp = skb->next)
2283 #define skb_queue_walk_from(queue, skb) \
2284 for (; skb != (struct sk_buff *)(queue); \
2285 skb = skb->next)
2287 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2288 for (tmp = skb->next; \
2289 skb != (struct sk_buff *)(queue); \
2290 skb = tmp, tmp = skb->next)
2292 #define skb_queue_reverse_walk(queue, skb) \
2293 for (skb = (queue)->prev; \
2294 skb != (struct sk_buff *)(queue); \
2295 skb = skb->prev)
2297 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2298 for (skb = (queue)->prev, tmp = skb->prev; \
2299 skb != (struct sk_buff *)(queue); \
2300 skb = tmp, tmp = skb->prev)
2302 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2303 for (tmp = skb->prev; \
2304 skb != (struct sk_buff *)(queue); \
2305 skb = tmp, tmp = skb->prev)
2307 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2309 return skb_shinfo(skb)->frag_list != NULL;
2312 static inline void skb_frag_list_init(struct sk_buff *skb)
2314 skb_shinfo(skb)->frag_list = NULL;
2317 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2319 frag->next = skb_shinfo(skb)->frag_list;
2320 skb_shinfo(skb)->frag_list = frag;
2323 #define skb_walk_frags(skb, iter) \
2324 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2326 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2327 int *peeked, int *off, int *err);
2328 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2329 int noblock, int *err);
2330 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2331 struct poll_table_struct *wait);
2332 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2333 int offset, struct iovec *to,
2334 int size);
2335 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
2336 int hlen,
2337 struct iovec *iov);
2338 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2339 int offset,
2340 const struct iovec *from,
2341 int from_offset,
2342 int len);
2343 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2344 int offset,
2345 const struct iovec *to,
2346 int to_offset,
2347 int size);
2348 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2349 extern void skb_free_datagram_locked(struct sock *sk,
2350 struct sk_buff *skb);
2351 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
2352 unsigned int flags);
2353 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2354 int len, __wsum csum);
2355 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2356 void *to, int len);
2357 extern int skb_store_bits(struct sk_buff *skb, int offset,
2358 const void *from, int len);
2359 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
2360 int offset, u8 *to, int len,
2361 __wsum csum);
2362 extern int skb_splice_bits(struct sk_buff *skb,
2363 unsigned int offset,
2364 struct pipe_inode_info *pipe,
2365 unsigned int len,
2366 unsigned int flags);
2367 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2368 extern void skb_split(struct sk_buff *skb,
2369 struct sk_buff *skb1, const u32 len);
2370 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2371 int shiftlen);
2373 extern struct sk_buff *skb_segment(struct sk_buff *skb,
2374 netdev_features_t features);
2376 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2377 int len, void *buffer)
2379 int hlen = skb_headlen(skb);
2381 if (hlen - offset >= len)
2382 return skb->data + offset;
2384 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2385 return NULL;
2387 return buffer;
2390 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2391 void *to,
2392 const unsigned int len)
2394 memcpy(to, skb->data, len);
2397 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2398 const int offset, void *to,
2399 const unsigned int len)
2401 memcpy(to, skb->data + offset, len);
2404 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2405 const void *from,
2406 const unsigned int len)
2408 memcpy(skb->data, from, len);
2411 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2412 const int offset,
2413 const void *from,
2414 const unsigned int len)
2416 memcpy(skb->data + offset, from, len);
2419 extern void skb_init(void);
2421 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2423 return skb->tstamp;
2427 * skb_get_timestamp - get timestamp from a skb
2428 * @skb: skb to get stamp from
2429 * @stamp: pointer to struct timeval to store stamp in
2431 * Timestamps are stored in the skb as offsets to a base timestamp.
2432 * This function converts the offset back to a struct timeval and stores
2433 * it in stamp.
2435 static inline void skb_get_timestamp(const struct sk_buff *skb,
2436 struct timeval *stamp)
2438 *stamp = ktime_to_timeval(skb->tstamp);
2441 static inline void skb_get_timestampns(const struct sk_buff *skb,
2442 struct timespec *stamp)
2444 *stamp = ktime_to_timespec(skb->tstamp);
2447 static inline void __net_timestamp(struct sk_buff *skb)
2449 skb->tstamp = ktime_get_real();
2452 static inline ktime_t net_timedelta(ktime_t t)
2454 return ktime_sub(ktime_get_real(), t);
2457 static inline ktime_t net_invalid_timestamp(void)
2459 return ktime_set(0, 0);
2462 extern void skb_timestamping_init(void);
2464 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2466 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2467 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2469 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2471 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2475 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2477 return false;
2480 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2483 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2485 * PHY drivers may accept clones of transmitted packets for
2486 * timestamping via their phy_driver.txtstamp method. These drivers
2487 * must call this function to return the skb back to the stack, with
2488 * or without a timestamp.
2490 * @skb: clone of the the original outgoing packet
2491 * @hwtstamps: hardware time stamps, may be NULL if not available
2494 void skb_complete_tx_timestamp(struct sk_buff *skb,
2495 struct skb_shared_hwtstamps *hwtstamps);
2498 * skb_tstamp_tx - queue clone of skb with send time stamps
2499 * @orig_skb: the original outgoing packet
2500 * @hwtstamps: hardware time stamps, may be NULL if not available
2502 * If the skb has a socket associated, then this function clones the
2503 * skb (thus sharing the actual data and optional structures), stores
2504 * the optional hardware time stamping information (if non NULL) or
2505 * generates a software time stamp (otherwise), then queues the clone
2506 * to the error queue of the socket. Errors are silently ignored.
2508 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2509 struct skb_shared_hwtstamps *hwtstamps);
2511 static inline void sw_tx_timestamp(struct sk_buff *skb)
2513 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2514 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2515 skb_tstamp_tx(skb, NULL);
2519 * skb_tx_timestamp() - Driver hook for transmit timestamping
2521 * Ethernet MAC Drivers should call this function in their hard_xmit()
2522 * function immediately before giving the sk_buff to the MAC hardware.
2524 * @skb: A socket buffer.
2526 static inline void skb_tx_timestamp(struct sk_buff *skb)
2528 skb_clone_tx_timestamp(skb);
2529 sw_tx_timestamp(skb);
2533 * skb_complete_wifi_ack - deliver skb with wifi status
2535 * @skb: the original outgoing packet
2536 * @acked: ack status
2539 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2541 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2542 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2544 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2546 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2550 * skb_checksum_complete - Calculate checksum of an entire packet
2551 * @skb: packet to process
2553 * This function calculates the checksum over the entire packet plus
2554 * the value of skb->csum. The latter can be used to supply the
2555 * checksum of a pseudo header as used by TCP/UDP. It returns the
2556 * checksum.
2558 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2559 * this function can be used to verify that checksum on received
2560 * packets. In that case the function should return zero if the
2561 * checksum is correct. In particular, this function will return zero
2562 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2563 * hardware has already verified the correctness of the checksum.
2565 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2567 return skb_csum_unnecessary(skb) ?
2568 0 : __skb_checksum_complete(skb);
2571 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2572 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2573 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2575 if (nfct && atomic_dec_and_test(&nfct->use))
2576 nf_conntrack_destroy(nfct);
2578 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2580 if (nfct)
2581 atomic_inc(&nfct->use);
2583 #endif
2584 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2585 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2587 if (skb)
2588 atomic_inc(&skb->users);
2590 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2592 if (skb)
2593 kfree_skb(skb);
2595 #endif
2596 #ifdef CONFIG_BRIDGE_NETFILTER
2597 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2599 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2600 kfree(nf_bridge);
2602 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2604 if (nf_bridge)
2605 atomic_inc(&nf_bridge->use);
2607 #endif /* CONFIG_BRIDGE_NETFILTER */
2608 static inline void nf_reset(struct sk_buff *skb)
2610 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2611 nf_conntrack_put(skb->nfct);
2612 skb->nfct = NULL;
2613 #endif
2614 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2615 nf_conntrack_put_reasm(skb->nfct_reasm);
2616 skb->nfct_reasm = NULL;
2617 #endif
2618 #ifdef CONFIG_BRIDGE_NETFILTER
2619 nf_bridge_put(skb->nf_bridge);
2620 skb->nf_bridge = NULL;
2621 #endif
2624 /* Note: This doesn't put any conntrack and bridge info in dst. */
2625 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2627 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2628 dst->nfct = src->nfct;
2629 nf_conntrack_get(src->nfct);
2630 dst->nfctinfo = src->nfctinfo;
2631 #endif
2632 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2633 dst->nfct_reasm = src->nfct_reasm;
2634 nf_conntrack_get_reasm(src->nfct_reasm);
2635 #endif
2636 #ifdef CONFIG_BRIDGE_NETFILTER
2637 dst->nf_bridge = src->nf_bridge;
2638 nf_bridge_get(src->nf_bridge);
2639 #endif
2642 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2644 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2645 nf_conntrack_put(dst->nfct);
2646 #endif
2647 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2648 nf_conntrack_put_reasm(dst->nfct_reasm);
2649 #endif
2650 #ifdef CONFIG_BRIDGE_NETFILTER
2651 nf_bridge_put(dst->nf_bridge);
2652 #endif
2653 __nf_copy(dst, src);
2656 #ifdef CONFIG_NETWORK_SECMARK
2657 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2659 to->secmark = from->secmark;
2662 static inline void skb_init_secmark(struct sk_buff *skb)
2664 skb->secmark = 0;
2666 #else
2667 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2670 static inline void skb_init_secmark(struct sk_buff *skb)
2672 #endif
2674 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2676 skb->queue_mapping = queue_mapping;
2679 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2681 return skb->queue_mapping;
2684 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2686 to->queue_mapping = from->queue_mapping;
2689 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2691 skb->queue_mapping = rx_queue + 1;
2694 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2696 return skb->queue_mapping - 1;
2699 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2701 return skb->queue_mapping != 0;
2704 extern u16 __skb_tx_hash(const struct net_device *dev,
2705 const struct sk_buff *skb,
2706 unsigned int num_tx_queues);
2708 #ifdef CONFIG_XFRM
2709 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2711 return skb->sp;
2713 #else
2714 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2716 return NULL;
2718 #endif
2720 static inline bool skb_is_gso(const struct sk_buff *skb)
2722 return skb_shinfo(skb)->gso_size;
2725 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2727 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2730 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2732 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2734 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2735 * wanted then gso_type will be set. */
2736 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2738 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2739 unlikely(shinfo->gso_type == 0)) {
2740 __skb_warn_lro_forwarding(skb);
2741 return true;
2743 return false;
2746 static inline void skb_forward_csum(struct sk_buff *skb)
2748 /* Unfortunately we don't support this one. Any brave souls? */
2749 if (skb->ip_summed == CHECKSUM_COMPLETE)
2750 skb->ip_summed = CHECKSUM_NONE;
2754 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2755 * @skb: skb to check
2757 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2758 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2759 * use this helper, to document places where we make this assertion.
2761 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2763 #ifdef DEBUG
2764 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2765 #endif
2768 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2771 * skb_head_is_locked - Determine if the skb->head is locked down
2772 * @skb: skb to check
2774 * The head on skbs build around a head frag can be removed if they are
2775 * not cloned. This function returns true if the skb head is locked down
2776 * due to either being allocated via kmalloc, or by being a clone with
2777 * multiple references to the head.
2779 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2781 return !skb->head_frag || skb_cloned(skb);
2783 #endif /* __KERNEL__ */
2784 #endif /* _LINUX_SKBUFF_H */