net: add driver hook for tx time stamping.
[pohmelfs.git] / include / linux / skbuff.h
bloba1b0400c8d8648c41c47efba7e6f039b61f383f9
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/cache.h>
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/dmaengine.h>
31 #include <linux/hrtimer.h>
33 /* Don't change this without changing skb_csum_unnecessary! */
34 #define CHECKSUM_NONE 0
35 #define CHECKSUM_UNNECESSARY 1
36 #define CHECKSUM_COMPLETE 2
37 #define CHECKSUM_PARTIAL 3
39 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
41 #define SKB_WITH_OVERHEAD(X) \
42 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
43 #define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * COMPLETE: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use COMPLETE,
63 * not UNNECESSARY.
65 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
72 * B. Checksumming on output.
74 * NONE: skb is checksummed by protocol or csum is not required.
76 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
77 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
83 * everything.
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
89 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
91 * Any questions? No questions, good. --ANK
94 struct net_device;
95 struct scatterlist;
96 struct pipe_inode_info;
98 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
99 struct nf_conntrack {
100 atomic_t use;
102 #endif
104 #ifdef CONFIG_BRIDGE_NETFILTER
105 struct nf_bridge_info {
106 atomic_t use;
107 struct net_device *physindev;
108 struct net_device *physoutdev;
109 unsigned int mask;
110 unsigned long data[32 / sizeof(unsigned long)];
112 #endif
114 struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
119 __u32 qlen;
120 spinlock_t lock;
123 struct sk_buff;
125 /* To allow 64K frame to be packed as single skb without frag_list */
126 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
128 typedef struct skb_frag_struct skb_frag_t;
130 struct skb_frag_struct {
131 struct page *page;
132 __u32 page_offset;
133 __u32 size;
136 #define HAVE_HW_TIME_STAMP
139 * struct skb_shared_hwtstamps - hardware time stamps
140 * @hwtstamp: hardware time stamp transformed into duration
141 * since arbitrary point in time
142 * @syststamp: hwtstamp transformed to system time base
144 * Software time stamps generated by ktime_get_real() are stored in
145 * skb->tstamp. The relation between the different kinds of time
146 * stamps is as follows:
148 * syststamp and tstamp can be compared against each other in
149 * arbitrary combinations. The accuracy of a
150 * syststamp/tstamp/"syststamp from other device" comparison is
151 * limited by the accuracy of the transformation into system time
152 * base. This depends on the device driver and its underlying
153 * hardware.
155 * hwtstamps can only be compared against other hwtstamps from
156 * the same device.
158 * This structure is attached to packets as part of the
159 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
161 struct skb_shared_hwtstamps {
162 ktime_t hwtstamp;
163 ktime_t syststamp;
167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
172 * @flags: all shared_tx flags
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
177 union skb_shared_tx {
178 struct {
179 __u8 hardware:1,
180 software:1,
181 in_progress:1;
183 __u8 flags;
186 /* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
189 struct skb_shared_info {
190 unsigned short nr_frags;
191 unsigned short gso_size;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs;
194 unsigned short gso_type;
195 __be32 ip6_frag_id;
196 union skb_shared_tx tx_flags;
197 struct sk_buff *frag_list;
198 struct skb_shared_hwtstamps hwtstamps;
201 * Warning : all fields before dataref are cleared in __alloc_skb()
203 atomic_t dataref;
205 skb_frag_t frags[MAX_SKB_FRAGS];
206 /* Intermediate layers must ensure that destructor_arg
207 * remains valid until skb destructor */
208 void * destructor_arg;
211 /* We divide dataref into two halves. The higher 16 bits hold references
212 * to the payload part of skb->data. The lower 16 bits hold references to
213 * the entire skb->data. A clone of a headerless skb holds the length of
214 * the header in skb->hdr_len.
216 * All users must obey the rule that the skb->data reference count must be
217 * greater than or equal to the payload reference count.
219 * Holding a reference to the payload part means that the user does not
220 * care about modifications to the header part of skb->data.
222 #define SKB_DATAREF_SHIFT 16
223 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
226 enum {
227 SKB_FCLONE_UNAVAILABLE,
228 SKB_FCLONE_ORIG,
229 SKB_FCLONE_CLONE,
232 enum {
233 SKB_GSO_TCPV4 = 1 << 0,
234 SKB_GSO_UDP = 1 << 1,
236 /* This indicates the skb is from an untrusted source. */
237 SKB_GSO_DODGY = 1 << 2,
239 /* This indicates the tcp segment has CWR set. */
240 SKB_GSO_TCP_ECN = 1 << 3,
242 SKB_GSO_TCPV6 = 1 << 4,
244 SKB_GSO_FCOE = 1 << 5,
247 #if BITS_PER_LONG > 32
248 #define NET_SKBUFF_DATA_USES_OFFSET 1
249 #endif
251 #ifdef NET_SKBUFF_DATA_USES_OFFSET
252 typedef unsigned int sk_buff_data_t;
253 #else
254 typedef unsigned char *sk_buff_data_t;
255 #endif
257 /**
258 * struct sk_buff - socket buffer
259 * @next: Next buffer in list
260 * @prev: Previous buffer in list
261 * @sk: Socket we are owned by
262 * @tstamp: Time we arrived
263 * @dev: Device we arrived on/are leaving by
264 * @transport_header: Transport layer header
265 * @network_header: Network layer header
266 * @mac_header: Link layer header
267 * @_skb_refdst: destination entry (with norefcount bit)
268 * @sp: the security path, used for xfrm
269 * @cb: Control buffer. Free for use by every layer. Put private vars here
270 * @len: Length of actual data
271 * @data_len: Data length
272 * @mac_len: Length of link layer header
273 * @hdr_len: writable header length of cloned skb
274 * @csum: Checksum (must include start/offset pair)
275 * @csum_start: Offset from skb->head where checksumming should start
276 * @csum_offset: Offset from csum_start where checksum should be stored
277 * @local_df: allow local fragmentation
278 * @cloned: Head may be cloned (check refcnt to be sure)
279 * @nohdr: Payload reference only, must not modify header
280 * @pkt_type: Packet class
281 * @fclone: skbuff clone status
282 * @ip_summed: Driver fed us an IP checksum
283 * @priority: Packet queueing priority
284 * @users: User count - see {datagram,tcp}.c
285 * @protocol: Packet protocol from driver
286 * @truesize: Buffer size
287 * @head: Head of buffer
288 * @data: Data head pointer
289 * @tail: Tail pointer
290 * @end: End pointer
291 * @destructor: Destruct function
292 * @mark: Generic packet mark
293 * @nfct: Associated connection, if any
294 * @ipvs_property: skbuff is owned by ipvs
295 * @peeked: this packet has been seen already, so stats have been
296 * done for it, don't do them again
297 * @nf_trace: netfilter packet trace flag
298 * @nfctinfo: Relationship of this skb to the connection
299 * @nfct_reasm: netfilter conntrack re-assembly pointer
300 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
301 * @skb_iif: ifindex of device we arrived on
302 * @rxhash: the packet hash computed on receive
303 * @queue_mapping: Queue mapping for multiqueue devices
304 * @tc_index: Traffic control index
305 * @tc_verd: traffic control verdict
306 * @ndisc_nodetype: router type (from link layer)
307 * @dma_cookie: a cookie to one of several possible DMA operations
308 * done by skb DMA functions
309 * @secmark: security marking
310 * @vlan_tci: vlan tag control information
313 struct sk_buff {
314 /* These two members must be first. */
315 struct sk_buff *next;
316 struct sk_buff *prev;
318 ktime_t tstamp;
320 struct sock *sk;
321 struct net_device *dev;
324 * This is the control buffer. It is free to use for every
325 * layer. Please put your private variables there. If you
326 * want to keep them across layers you have to do a skb_clone()
327 * first. This is owned by whoever has the skb queued ATM.
329 char cb[48] __aligned(8);
331 unsigned long _skb_refdst;
332 #ifdef CONFIG_XFRM
333 struct sec_path *sp;
334 #endif
335 unsigned int len,
336 data_len;
337 __u16 mac_len,
338 hdr_len;
339 union {
340 __wsum csum;
341 struct {
342 __u16 csum_start;
343 __u16 csum_offset;
346 __u32 priority;
347 kmemcheck_bitfield_begin(flags1);
348 __u8 local_df:1,
349 cloned:1,
350 ip_summed:2,
351 nohdr:1,
352 nfctinfo:3;
353 __u8 pkt_type:3,
354 fclone:2,
355 ipvs_property:1,
356 peeked:1,
357 nf_trace:1;
358 kmemcheck_bitfield_end(flags1);
359 __be16 protocol;
361 void (*destructor)(struct sk_buff *skb);
362 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
363 struct nf_conntrack *nfct;
364 struct sk_buff *nfct_reasm;
365 #endif
366 #ifdef CONFIG_BRIDGE_NETFILTER
367 struct nf_bridge_info *nf_bridge;
368 #endif
370 int skb_iif;
371 #ifdef CONFIG_NET_SCHED
372 __u16 tc_index; /* traffic control index */
373 #ifdef CONFIG_NET_CLS_ACT
374 __u16 tc_verd; /* traffic control verdict */
375 #endif
376 #endif
378 __u32 rxhash;
380 kmemcheck_bitfield_begin(flags2);
381 __u16 queue_mapping:16;
382 #ifdef CONFIG_IPV6_NDISC_NODETYPE
383 __u8 ndisc_nodetype:2,
384 deliver_no_wcard:1;
385 #else
386 __u8 deliver_no_wcard:1;
387 #endif
388 kmemcheck_bitfield_end(flags2);
390 /* 0/14 bit hole */
392 #ifdef CONFIG_NET_DMA
393 dma_cookie_t dma_cookie;
394 #endif
395 #ifdef CONFIG_NETWORK_SECMARK
396 __u32 secmark;
397 #endif
398 union {
399 __u32 mark;
400 __u32 dropcount;
403 __u16 vlan_tci;
405 sk_buff_data_t transport_header;
406 sk_buff_data_t network_header;
407 sk_buff_data_t mac_header;
408 /* These elements must be at the end, see alloc_skb() for details. */
409 sk_buff_data_t tail;
410 sk_buff_data_t end;
411 unsigned char *head,
412 *data;
413 unsigned int truesize;
414 atomic_t users;
417 #ifdef __KERNEL__
419 * Handling routines are only of interest to the kernel
421 #include <linux/slab.h>
423 #include <asm/system.h>
426 * skb might have a dst pointer attached, refcounted or not.
427 * _skb_refdst low order bit is set if refcount was _not_ taken
429 #define SKB_DST_NOREF 1UL
430 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
433 * skb_dst - returns skb dst_entry
434 * @skb: buffer
436 * Returns skb dst_entry, regardless of reference taken or not.
438 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
440 /* If refdst was not refcounted, check we still are in a
441 * rcu_read_lock section
443 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
444 !rcu_read_lock_held() &&
445 !rcu_read_lock_bh_held());
446 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
450 * skb_dst_set - sets skb dst
451 * @skb: buffer
452 * @dst: dst entry
454 * Sets skb dst, assuming a reference was taken on dst and should
455 * be released by skb_dst_drop()
457 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
459 skb->_skb_refdst = (unsigned long)dst;
463 * skb_dst_set_noref - sets skb dst, without a reference
464 * @skb: buffer
465 * @dst: dst entry
467 * Sets skb dst, assuming a reference was not taken on dst
468 * skb_dst_drop() should not dst_release() this dst
470 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
472 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
473 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
477 * skb_dst_is_noref - Test if skb dst isnt refcounted
478 * @skb: buffer
480 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
482 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
485 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
487 return (struct rtable *)skb_dst(skb);
490 extern void kfree_skb(struct sk_buff *skb);
491 extern void consume_skb(struct sk_buff *skb);
492 extern void __kfree_skb(struct sk_buff *skb);
493 extern struct sk_buff *__alloc_skb(unsigned int size,
494 gfp_t priority, int fclone, int node);
495 static inline struct sk_buff *alloc_skb(unsigned int size,
496 gfp_t priority)
498 return __alloc_skb(size, priority, 0, -1);
501 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
502 gfp_t priority)
504 return __alloc_skb(size, priority, 1, -1);
507 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
509 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
510 extern struct sk_buff *skb_clone(struct sk_buff *skb,
511 gfp_t priority);
512 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
513 gfp_t priority);
514 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
515 gfp_t gfp_mask);
516 extern int pskb_expand_head(struct sk_buff *skb,
517 int nhead, int ntail,
518 gfp_t gfp_mask);
519 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
520 unsigned int headroom);
521 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
522 int newheadroom, int newtailroom,
523 gfp_t priority);
524 extern int skb_to_sgvec(struct sk_buff *skb,
525 struct scatterlist *sg, int offset,
526 int len);
527 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
528 struct sk_buff **trailer);
529 extern int skb_pad(struct sk_buff *skb, int pad);
530 #define dev_kfree_skb(a) consume_skb(a)
532 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
533 int getfrag(void *from, char *to, int offset,
534 int len,int odd, struct sk_buff *skb),
535 void *from, int length);
537 struct skb_seq_state {
538 __u32 lower_offset;
539 __u32 upper_offset;
540 __u32 frag_idx;
541 __u32 stepped_offset;
542 struct sk_buff *root_skb;
543 struct sk_buff *cur_skb;
544 __u8 *frag_data;
547 extern void skb_prepare_seq_read(struct sk_buff *skb,
548 unsigned int from, unsigned int to,
549 struct skb_seq_state *st);
550 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
551 struct skb_seq_state *st);
552 extern void skb_abort_seq_read(struct skb_seq_state *st);
554 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
555 unsigned int to, struct ts_config *config,
556 struct ts_state *state);
558 #ifdef NET_SKBUFF_DATA_USES_OFFSET
559 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
561 return skb->head + skb->end;
563 #else
564 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
566 return skb->end;
568 #endif
570 /* Internal */
571 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
573 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
575 return &skb_shinfo(skb)->hwtstamps;
578 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
580 return &skb_shinfo(skb)->tx_flags;
584 * skb_queue_empty - check if a queue is empty
585 * @list: queue head
587 * Returns true if the queue is empty, false otherwise.
589 static inline int skb_queue_empty(const struct sk_buff_head *list)
591 return list->next == (struct sk_buff *)list;
595 * skb_queue_is_last - check if skb is the last entry in the queue
596 * @list: queue head
597 * @skb: buffer
599 * Returns true if @skb is the last buffer on the list.
601 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
602 const struct sk_buff *skb)
604 return (skb->next == (struct sk_buff *) list);
608 * skb_queue_is_first - check if skb is the first entry in the queue
609 * @list: queue head
610 * @skb: buffer
612 * Returns true if @skb is the first buffer on the list.
614 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
615 const struct sk_buff *skb)
617 return (skb->prev == (struct sk_buff *) list);
621 * skb_queue_next - return the next packet in the queue
622 * @list: queue head
623 * @skb: current buffer
625 * Return the next packet in @list after @skb. It is only valid to
626 * call this if skb_queue_is_last() evaluates to false.
628 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
629 const struct sk_buff *skb)
631 /* This BUG_ON may seem severe, but if we just return then we
632 * are going to dereference garbage.
634 BUG_ON(skb_queue_is_last(list, skb));
635 return skb->next;
639 * skb_queue_prev - return the prev packet in the queue
640 * @list: queue head
641 * @skb: current buffer
643 * Return the prev packet in @list before @skb. It is only valid to
644 * call this if skb_queue_is_first() evaluates to false.
646 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
647 const struct sk_buff *skb)
649 /* This BUG_ON may seem severe, but if we just return then we
650 * are going to dereference garbage.
652 BUG_ON(skb_queue_is_first(list, skb));
653 return skb->prev;
657 * skb_get - reference buffer
658 * @skb: buffer to reference
660 * Makes another reference to a socket buffer and returns a pointer
661 * to the buffer.
663 static inline struct sk_buff *skb_get(struct sk_buff *skb)
665 atomic_inc(&skb->users);
666 return skb;
670 * If users == 1, we are the only owner and are can avoid redundant
671 * atomic change.
675 * skb_cloned - is the buffer a clone
676 * @skb: buffer to check
678 * Returns true if the buffer was generated with skb_clone() and is
679 * one of multiple shared copies of the buffer. Cloned buffers are
680 * shared data so must not be written to under normal circumstances.
682 static inline int skb_cloned(const struct sk_buff *skb)
684 return skb->cloned &&
685 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
689 * skb_header_cloned - is the header a clone
690 * @skb: buffer to check
692 * Returns true if modifying the header part of the buffer requires
693 * the data to be copied.
695 static inline int skb_header_cloned(const struct sk_buff *skb)
697 int dataref;
699 if (!skb->cloned)
700 return 0;
702 dataref = atomic_read(&skb_shinfo(skb)->dataref);
703 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
704 return dataref != 1;
708 * skb_header_release - release reference to header
709 * @skb: buffer to operate on
711 * Drop a reference to the header part of the buffer. This is done
712 * by acquiring a payload reference. You must not read from the header
713 * part of skb->data after this.
715 static inline void skb_header_release(struct sk_buff *skb)
717 BUG_ON(skb->nohdr);
718 skb->nohdr = 1;
719 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
723 * skb_shared - is the buffer shared
724 * @skb: buffer to check
726 * Returns true if more than one person has a reference to this
727 * buffer.
729 static inline int skb_shared(const struct sk_buff *skb)
731 return atomic_read(&skb->users) != 1;
735 * skb_share_check - check if buffer is shared and if so clone it
736 * @skb: buffer to check
737 * @pri: priority for memory allocation
739 * If the buffer is shared the buffer is cloned and the old copy
740 * drops a reference. A new clone with a single reference is returned.
741 * If the buffer is not shared the original buffer is returned. When
742 * being called from interrupt status or with spinlocks held pri must
743 * be GFP_ATOMIC.
745 * NULL is returned on a memory allocation failure.
747 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
748 gfp_t pri)
750 might_sleep_if(pri & __GFP_WAIT);
751 if (skb_shared(skb)) {
752 struct sk_buff *nskb = skb_clone(skb, pri);
753 kfree_skb(skb);
754 skb = nskb;
756 return skb;
760 * Copy shared buffers into a new sk_buff. We effectively do COW on
761 * packets to handle cases where we have a local reader and forward
762 * and a couple of other messy ones. The normal one is tcpdumping
763 * a packet thats being forwarded.
767 * skb_unshare - make a copy of a shared buffer
768 * @skb: buffer to check
769 * @pri: priority for memory allocation
771 * If the socket buffer is a clone then this function creates a new
772 * copy of the data, drops a reference count on the old copy and returns
773 * the new copy with the reference count at 1. If the buffer is not a clone
774 * the original buffer is returned. When called with a spinlock held or
775 * from interrupt state @pri must be %GFP_ATOMIC
777 * %NULL is returned on a memory allocation failure.
779 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
780 gfp_t pri)
782 might_sleep_if(pri & __GFP_WAIT);
783 if (skb_cloned(skb)) {
784 struct sk_buff *nskb = skb_copy(skb, pri);
785 kfree_skb(skb); /* Free our shared copy */
786 skb = nskb;
788 return skb;
792 * skb_peek - peek at the head of an &sk_buff_head
793 * @list_: list to peek at
795 * Peek an &sk_buff. Unlike most other operations you _MUST_
796 * be careful with this one. A peek leaves the buffer on the
797 * list and someone else may run off with it. You must hold
798 * the appropriate locks or have a private queue to do this.
800 * Returns %NULL for an empty list or a pointer to the head element.
801 * The reference count is not incremented and the reference is therefore
802 * volatile. Use with caution.
804 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
806 struct sk_buff *list = ((struct sk_buff *)list_)->next;
807 if (list == (struct sk_buff *)list_)
808 list = NULL;
809 return list;
813 * skb_peek_tail - peek at the tail of an &sk_buff_head
814 * @list_: list to peek at
816 * Peek an &sk_buff. Unlike most other operations you _MUST_
817 * be careful with this one. A peek leaves the buffer on the
818 * list and someone else may run off with it. You must hold
819 * the appropriate locks or have a private queue to do this.
821 * Returns %NULL for an empty list or a pointer to the tail element.
822 * The reference count is not incremented and the reference is therefore
823 * volatile. Use with caution.
825 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
827 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
828 if (list == (struct sk_buff *)list_)
829 list = NULL;
830 return list;
834 * skb_queue_len - get queue length
835 * @list_: list to measure
837 * Return the length of an &sk_buff queue.
839 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
841 return list_->qlen;
845 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
846 * @list: queue to initialize
848 * This initializes only the list and queue length aspects of
849 * an sk_buff_head object. This allows to initialize the list
850 * aspects of an sk_buff_head without reinitializing things like
851 * the spinlock. It can also be used for on-stack sk_buff_head
852 * objects where the spinlock is known to not be used.
854 static inline void __skb_queue_head_init(struct sk_buff_head *list)
856 list->prev = list->next = (struct sk_buff *)list;
857 list->qlen = 0;
861 * This function creates a split out lock class for each invocation;
862 * this is needed for now since a whole lot of users of the skb-queue
863 * infrastructure in drivers have different locking usage (in hardirq)
864 * than the networking core (in softirq only). In the long run either the
865 * network layer or drivers should need annotation to consolidate the
866 * main types of usage into 3 classes.
868 static inline void skb_queue_head_init(struct sk_buff_head *list)
870 spin_lock_init(&list->lock);
871 __skb_queue_head_init(list);
874 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
875 struct lock_class_key *class)
877 skb_queue_head_init(list);
878 lockdep_set_class(&list->lock, class);
882 * Insert an sk_buff on a list.
884 * The "__skb_xxxx()" functions are the non-atomic ones that
885 * can only be called with interrupts disabled.
887 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
888 static inline void __skb_insert(struct sk_buff *newsk,
889 struct sk_buff *prev, struct sk_buff *next,
890 struct sk_buff_head *list)
892 newsk->next = next;
893 newsk->prev = prev;
894 next->prev = prev->next = newsk;
895 list->qlen++;
898 static inline void __skb_queue_splice(const struct sk_buff_head *list,
899 struct sk_buff *prev,
900 struct sk_buff *next)
902 struct sk_buff *first = list->next;
903 struct sk_buff *last = list->prev;
905 first->prev = prev;
906 prev->next = first;
908 last->next = next;
909 next->prev = last;
913 * skb_queue_splice - join two skb lists, this is designed for stacks
914 * @list: the new list to add
915 * @head: the place to add it in the first list
917 static inline void skb_queue_splice(const struct sk_buff_head *list,
918 struct sk_buff_head *head)
920 if (!skb_queue_empty(list)) {
921 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
922 head->qlen += list->qlen;
927 * skb_queue_splice - join two skb lists and reinitialise the emptied list
928 * @list: the new list to add
929 * @head: the place to add it in the first list
931 * The list at @list is reinitialised
933 static inline void skb_queue_splice_init(struct sk_buff_head *list,
934 struct sk_buff_head *head)
936 if (!skb_queue_empty(list)) {
937 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
938 head->qlen += list->qlen;
939 __skb_queue_head_init(list);
944 * skb_queue_splice_tail - join two skb lists, each list being a queue
945 * @list: the new list to add
946 * @head: the place to add it in the first list
948 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
949 struct sk_buff_head *head)
951 if (!skb_queue_empty(list)) {
952 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
953 head->qlen += list->qlen;
958 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
959 * @list: the new list to add
960 * @head: the place to add it in the first list
962 * Each of the lists is a queue.
963 * The list at @list is reinitialised
965 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
966 struct sk_buff_head *head)
968 if (!skb_queue_empty(list)) {
969 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
970 head->qlen += list->qlen;
971 __skb_queue_head_init(list);
976 * __skb_queue_after - queue a buffer at the list head
977 * @list: list to use
978 * @prev: place after this buffer
979 * @newsk: buffer to queue
981 * Queue a buffer int the middle of a list. This function takes no locks
982 * and you must therefore hold required locks before calling it.
984 * A buffer cannot be placed on two lists at the same time.
986 static inline void __skb_queue_after(struct sk_buff_head *list,
987 struct sk_buff *prev,
988 struct sk_buff *newsk)
990 __skb_insert(newsk, prev, prev->next, list);
993 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
994 struct sk_buff_head *list);
996 static inline void __skb_queue_before(struct sk_buff_head *list,
997 struct sk_buff *next,
998 struct sk_buff *newsk)
1000 __skb_insert(newsk, next->prev, next, list);
1004 * __skb_queue_head - queue a buffer at the list head
1005 * @list: list to use
1006 * @newsk: buffer to queue
1008 * Queue a buffer at the start of a list. This function takes no locks
1009 * and you must therefore hold required locks before calling it.
1011 * A buffer cannot be placed on two lists at the same time.
1013 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1014 static inline void __skb_queue_head(struct sk_buff_head *list,
1015 struct sk_buff *newsk)
1017 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1021 * __skb_queue_tail - queue a buffer at the list tail
1022 * @list: list to use
1023 * @newsk: buffer to queue
1025 * Queue a buffer at the end of a list. This function takes no locks
1026 * and you must therefore hold required locks before calling it.
1028 * A buffer cannot be placed on two lists at the same time.
1030 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1031 static inline void __skb_queue_tail(struct sk_buff_head *list,
1032 struct sk_buff *newsk)
1034 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1038 * remove sk_buff from list. _Must_ be called atomically, and with
1039 * the list known..
1041 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1042 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1044 struct sk_buff *next, *prev;
1046 list->qlen--;
1047 next = skb->next;
1048 prev = skb->prev;
1049 skb->next = skb->prev = NULL;
1050 next->prev = prev;
1051 prev->next = next;
1055 * __skb_dequeue - remove from the head of the queue
1056 * @list: list to dequeue from
1058 * Remove the head of the list. This function does not take any locks
1059 * so must be used with appropriate locks held only. The head item is
1060 * returned or %NULL if the list is empty.
1062 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1063 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1065 struct sk_buff *skb = skb_peek(list);
1066 if (skb)
1067 __skb_unlink(skb, list);
1068 return skb;
1072 * __skb_dequeue_tail - remove from the tail of the queue
1073 * @list: list to dequeue from
1075 * Remove the tail of the list. This function does not take any locks
1076 * so must be used with appropriate locks held only. The tail item is
1077 * returned or %NULL if the list is empty.
1079 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1080 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1082 struct sk_buff *skb = skb_peek_tail(list);
1083 if (skb)
1084 __skb_unlink(skb, list);
1085 return skb;
1089 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1091 return skb->data_len;
1094 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1096 return skb->len - skb->data_len;
1099 static inline int skb_pagelen(const struct sk_buff *skb)
1101 int i, len = 0;
1103 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1104 len += skb_shinfo(skb)->frags[i].size;
1105 return len + skb_headlen(skb);
1108 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1109 struct page *page, int off, int size)
1111 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1113 frag->page = page;
1114 frag->page_offset = off;
1115 frag->size = size;
1116 skb_shinfo(skb)->nr_frags = i + 1;
1119 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1120 int off, int size);
1122 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1123 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1124 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1126 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1127 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1129 return skb->head + skb->tail;
1132 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1134 skb->tail = skb->data - skb->head;
1137 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1139 skb_reset_tail_pointer(skb);
1140 skb->tail += offset;
1142 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1143 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1145 return skb->tail;
1148 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1150 skb->tail = skb->data;
1153 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1155 skb->tail = skb->data + offset;
1158 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1161 * Add data to an sk_buff
1163 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1164 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1166 unsigned char *tmp = skb_tail_pointer(skb);
1167 SKB_LINEAR_ASSERT(skb);
1168 skb->tail += len;
1169 skb->len += len;
1170 return tmp;
1173 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1174 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1176 skb->data -= len;
1177 skb->len += len;
1178 return skb->data;
1181 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1182 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1184 skb->len -= len;
1185 BUG_ON(skb->len < skb->data_len);
1186 return skb->data += len;
1189 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1191 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1194 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1196 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1198 if (len > skb_headlen(skb) &&
1199 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1200 return NULL;
1201 skb->len -= len;
1202 return skb->data += len;
1205 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1207 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1210 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1212 if (likely(len <= skb_headlen(skb)))
1213 return 1;
1214 if (unlikely(len > skb->len))
1215 return 0;
1216 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1220 * skb_headroom - bytes at buffer head
1221 * @skb: buffer to check
1223 * Return the number of bytes of free space at the head of an &sk_buff.
1225 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1227 return skb->data - skb->head;
1231 * skb_tailroom - bytes at buffer end
1232 * @skb: buffer to check
1234 * Return the number of bytes of free space at the tail of an sk_buff
1236 static inline int skb_tailroom(const struct sk_buff *skb)
1238 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1242 * skb_reserve - adjust headroom
1243 * @skb: buffer to alter
1244 * @len: bytes to move
1246 * Increase the headroom of an empty &sk_buff by reducing the tail
1247 * room. This is only allowed for an empty buffer.
1249 static inline void skb_reserve(struct sk_buff *skb, int len)
1251 skb->data += len;
1252 skb->tail += len;
1255 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1256 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1258 return skb->head + skb->transport_header;
1261 static inline void skb_reset_transport_header(struct sk_buff *skb)
1263 skb->transport_header = skb->data - skb->head;
1266 static inline void skb_set_transport_header(struct sk_buff *skb,
1267 const int offset)
1269 skb_reset_transport_header(skb);
1270 skb->transport_header += offset;
1273 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1275 return skb->head + skb->network_header;
1278 static inline void skb_reset_network_header(struct sk_buff *skb)
1280 skb->network_header = skb->data - skb->head;
1283 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1285 skb_reset_network_header(skb);
1286 skb->network_header += offset;
1289 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1291 return skb->head + skb->mac_header;
1294 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1296 return skb->mac_header != ~0U;
1299 static inline void skb_reset_mac_header(struct sk_buff *skb)
1301 skb->mac_header = skb->data - skb->head;
1304 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1306 skb_reset_mac_header(skb);
1307 skb->mac_header += offset;
1310 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1312 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1314 return skb->transport_header;
1317 static inline void skb_reset_transport_header(struct sk_buff *skb)
1319 skb->transport_header = skb->data;
1322 static inline void skb_set_transport_header(struct sk_buff *skb,
1323 const int offset)
1325 skb->transport_header = skb->data + offset;
1328 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1330 return skb->network_header;
1333 static inline void skb_reset_network_header(struct sk_buff *skb)
1335 skb->network_header = skb->data;
1338 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1340 skb->network_header = skb->data + offset;
1343 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1345 return skb->mac_header;
1348 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1350 return skb->mac_header != NULL;
1353 static inline void skb_reset_mac_header(struct sk_buff *skb)
1355 skb->mac_header = skb->data;
1358 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1360 skb->mac_header = skb->data + offset;
1362 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1364 static inline int skb_transport_offset(const struct sk_buff *skb)
1366 return skb_transport_header(skb) - skb->data;
1369 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1371 return skb->transport_header - skb->network_header;
1374 static inline int skb_network_offset(const struct sk_buff *skb)
1376 return skb_network_header(skb) - skb->data;
1380 * CPUs often take a performance hit when accessing unaligned memory
1381 * locations. The actual performance hit varies, it can be small if the
1382 * hardware handles it or large if we have to take an exception and fix it
1383 * in software.
1385 * Since an ethernet header is 14 bytes network drivers often end up with
1386 * the IP header at an unaligned offset. The IP header can be aligned by
1387 * shifting the start of the packet by 2 bytes. Drivers should do this
1388 * with:
1390 * skb_reserve(skb, NET_IP_ALIGN);
1392 * The downside to this alignment of the IP header is that the DMA is now
1393 * unaligned. On some architectures the cost of an unaligned DMA is high
1394 * and this cost outweighs the gains made by aligning the IP header.
1396 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1397 * to be overridden.
1399 #ifndef NET_IP_ALIGN
1400 #define NET_IP_ALIGN 2
1401 #endif
1404 * The networking layer reserves some headroom in skb data (via
1405 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1406 * the header has to grow. In the default case, if the header has to grow
1407 * 32 bytes or less we avoid the reallocation.
1409 * Unfortunately this headroom changes the DMA alignment of the resulting
1410 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1411 * on some architectures. An architecture can override this value,
1412 * perhaps setting it to a cacheline in size (since that will maintain
1413 * cacheline alignment of the DMA). It must be a power of 2.
1415 * Various parts of the networking layer expect at least 32 bytes of
1416 * headroom, you should not reduce this.
1418 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1419 * to reduce average number of cache lines per packet.
1420 * get_rps_cpus() for example only access one 64 bytes aligned block :
1421 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1423 #ifndef NET_SKB_PAD
1424 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1425 #endif
1427 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1429 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1431 if (unlikely(skb->data_len)) {
1432 WARN_ON(1);
1433 return;
1435 skb->len = len;
1436 skb_set_tail_pointer(skb, len);
1439 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1441 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1443 if (skb->data_len)
1444 return ___pskb_trim(skb, len);
1445 __skb_trim(skb, len);
1446 return 0;
1449 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1451 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1455 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1456 * @skb: buffer to alter
1457 * @len: new length
1459 * This is identical to pskb_trim except that the caller knows that
1460 * the skb is not cloned so we should never get an error due to out-
1461 * of-memory.
1463 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1465 int err = pskb_trim(skb, len);
1466 BUG_ON(err);
1470 * skb_orphan - orphan a buffer
1471 * @skb: buffer to orphan
1473 * If a buffer currently has an owner then we call the owner's
1474 * destructor function and make the @skb unowned. The buffer continues
1475 * to exist but is no longer charged to its former owner.
1477 static inline void skb_orphan(struct sk_buff *skb)
1479 if (skb->destructor)
1480 skb->destructor(skb);
1481 skb->destructor = NULL;
1482 skb->sk = NULL;
1486 * __skb_queue_purge - empty a list
1487 * @list: list to empty
1489 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1490 * the list and one reference dropped. This function does not take the
1491 * list lock and the caller must hold the relevant locks to use it.
1493 extern void skb_queue_purge(struct sk_buff_head *list);
1494 static inline void __skb_queue_purge(struct sk_buff_head *list)
1496 struct sk_buff *skb;
1497 while ((skb = __skb_dequeue(list)) != NULL)
1498 kfree_skb(skb);
1502 * __dev_alloc_skb - allocate an skbuff for receiving
1503 * @length: length to allocate
1504 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1506 * Allocate a new &sk_buff and assign it a usage count of one. The
1507 * buffer has unspecified headroom built in. Users should allocate
1508 * the headroom they think they need without accounting for the
1509 * built in space. The built in space is used for optimisations.
1511 * %NULL is returned if there is no free memory.
1513 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1514 gfp_t gfp_mask)
1516 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1517 if (likely(skb))
1518 skb_reserve(skb, NET_SKB_PAD);
1519 return skb;
1522 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1524 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1525 unsigned int length, gfp_t gfp_mask);
1528 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1529 * @dev: network device to receive on
1530 * @length: length to allocate
1532 * Allocate a new &sk_buff and assign it a usage count of one. The
1533 * buffer has unspecified headroom built in. Users should allocate
1534 * the headroom they think they need without accounting for the
1535 * built in space. The built in space is used for optimisations.
1537 * %NULL is returned if there is no free memory. Although this function
1538 * allocates memory it can be called from an interrupt.
1540 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1541 unsigned int length)
1543 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1546 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1547 unsigned int length)
1549 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1551 if (NET_IP_ALIGN && skb)
1552 skb_reserve(skb, NET_IP_ALIGN);
1553 return skb;
1556 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1559 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1560 * @dev: network device to receive on
1562 * Allocate a new page node local to the specified device.
1564 * %NULL is returned if there is no free memory.
1566 static inline struct page *netdev_alloc_page(struct net_device *dev)
1568 return __netdev_alloc_page(dev, GFP_ATOMIC);
1571 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1573 __free_page(page);
1577 * skb_clone_writable - is the header of a clone writable
1578 * @skb: buffer to check
1579 * @len: length up to which to write
1581 * Returns true if modifying the header part of the cloned buffer
1582 * does not requires the data to be copied.
1584 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1586 return !skb_header_cloned(skb) &&
1587 skb_headroom(skb) + len <= skb->hdr_len;
1590 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1591 int cloned)
1593 int delta = 0;
1595 if (headroom < NET_SKB_PAD)
1596 headroom = NET_SKB_PAD;
1597 if (headroom > skb_headroom(skb))
1598 delta = headroom - skb_headroom(skb);
1600 if (delta || cloned)
1601 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1602 GFP_ATOMIC);
1603 return 0;
1607 * skb_cow - copy header of skb when it is required
1608 * @skb: buffer to cow
1609 * @headroom: needed headroom
1611 * If the skb passed lacks sufficient headroom or its data part
1612 * is shared, data is reallocated. If reallocation fails, an error
1613 * is returned and original skb is not changed.
1615 * The result is skb with writable area skb->head...skb->tail
1616 * and at least @headroom of space at head.
1618 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1620 return __skb_cow(skb, headroom, skb_cloned(skb));
1624 * skb_cow_head - skb_cow but only making the head writable
1625 * @skb: buffer to cow
1626 * @headroom: needed headroom
1628 * This function is identical to skb_cow except that we replace the
1629 * skb_cloned check by skb_header_cloned. It should be used when
1630 * you only need to push on some header and do not need to modify
1631 * the data.
1633 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1635 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1639 * skb_padto - pad an skbuff up to a minimal size
1640 * @skb: buffer to pad
1641 * @len: minimal length
1643 * Pads up a buffer to ensure the trailing bytes exist and are
1644 * blanked. If the buffer already contains sufficient data it
1645 * is untouched. Otherwise it is extended. Returns zero on
1646 * success. The skb is freed on error.
1649 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1651 unsigned int size = skb->len;
1652 if (likely(size >= len))
1653 return 0;
1654 return skb_pad(skb, len - size);
1657 static inline int skb_add_data(struct sk_buff *skb,
1658 char __user *from, int copy)
1660 const int off = skb->len;
1662 if (skb->ip_summed == CHECKSUM_NONE) {
1663 int err = 0;
1664 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1665 copy, 0, &err);
1666 if (!err) {
1667 skb->csum = csum_block_add(skb->csum, csum, off);
1668 return 0;
1670 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1671 return 0;
1673 __skb_trim(skb, off);
1674 return -EFAULT;
1677 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1678 struct page *page, int off)
1680 if (i) {
1681 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1683 return page == frag->page &&
1684 off == frag->page_offset + frag->size;
1686 return 0;
1689 static inline int __skb_linearize(struct sk_buff *skb)
1691 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1695 * skb_linearize - convert paged skb to linear one
1696 * @skb: buffer to linarize
1698 * If there is no free memory -ENOMEM is returned, otherwise zero
1699 * is returned and the old skb data released.
1701 static inline int skb_linearize(struct sk_buff *skb)
1703 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1707 * skb_linearize_cow - make sure skb is linear and writable
1708 * @skb: buffer to process
1710 * If there is no free memory -ENOMEM is returned, otherwise zero
1711 * is returned and the old skb data released.
1713 static inline int skb_linearize_cow(struct sk_buff *skb)
1715 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1716 __skb_linearize(skb) : 0;
1720 * skb_postpull_rcsum - update checksum for received skb after pull
1721 * @skb: buffer to update
1722 * @start: start of data before pull
1723 * @len: length of data pulled
1725 * After doing a pull on a received packet, you need to call this to
1726 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1727 * CHECKSUM_NONE so that it can be recomputed from scratch.
1730 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1731 const void *start, unsigned int len)
1733 if (skb->ip_summed == CHECKSUM_COMPLETE)
1734 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1737 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1740 * pskb_trim_rcsum - trim received skb and update checksum
1741 * @skb: buffer to trim
1742 * @len: new length
1744 * This is exactly the same as pskb_trim except that it ensures the
1745 * checksum of received packets are still valid after the operation.
1748 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1750 if (likely(len >= skb->len))
1751 return 0;
1752 if (skb->ip_summed == CHECKSUM_COMPLETE)
1753 skb->ip_summed = CHECKSUM_NONE;
1754 return __pskb_trim(skb, len);
1757 #define skb_queue_walk(queue, skb) \
1758 for (skb = (queue)->next; \
1759 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1760 skb = skb->next)
1762 #define skb_queue_walk_safe(queue, skb, tmp) \
1763 for (skb = (queue)->next, tmp = skb->next; \
1764 skb != (struct sk_buff *)(queue); \
1765 skb = tmp, tmp = skb->next)
1767 #define skb_queue_walk_from(queue, skb) \
1768 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1769 skb = skb->next)
1771 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1772 for (tmp = skb->next; \
1773 skb != (struct sk_buff *)(queue); \
1774 skb = tmp, tmp = skb->next)
1776 #define skb_queue_reverse_walk(queue, skb) \
1777 for (skb = (queue)->prev; \
1778 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1779 skb = skb->prev)
1782 static inline bool skb_has_frags(const struct sk_buff *skb)
1784 return skb_shinfo(skb)->frag_list != NULL;
1787 static inline void skb_frag_list_init(struct sk_buff *skb)
1789 skb_shinfo(skb)->frag_list = NULL;
1792 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1794 frag->next = skb_shinfo(skb)->frag_list;
1795 skb_shinfo(skb)->frag_list = frag;
1798 #define skb_walk_frags(skb, iter) \
1799 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1801 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1802 int *peeked, int *err);
1803 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1804 int noblock, int *err);
1805 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1806 struct poll_table_struct *wait);
1807 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1808 int offset, struct iovec *to,
1809 int size);
1810 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1811 int hlen,
1812 struct iovec *iov);
1813 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1814 int offset,
1815 const struct iovec *from,
1816 int from_offset,
1817 int len);
1818 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1819 int offset,
1820 const struct iovec *to,
1821 int to_offset,
1822 int size);
1823 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1824 extern void skb_free_datagram_locked(struct sock *sk,
1825 struct sk_buff *skb);
1826 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1827 unsigned int flags);
1828 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1829 int len, __wsum csum);
1830 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1831 void *to, int len);
1832 extern int skb_store_bits(struct sk_buff *skb, int offset,
1833 const void *from, int len);
1834 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1835 int offset, u8 *to, int len,
1836 __wsum csum);
1837 extern int skb_splice_bits(struct sk_buff *skb,
1838 unsigned int offset,
1839 struct pipe_inode_info *pipe,
1840 unsigned int len,
1841 unsigned int flags);
1842 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1843 extern void skb_split(struct sk_buff *skb,
1844 struct sk_buff *skb1, const u32 len);
1845 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1846 int shiftlen);
1848 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1850 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1851 int len, void *buffer)
1853 int hlen = skb_headlen(skb);
1855 if (hlen - offset >= len)
1856 return skb->data + offset;
1858 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1859 return NULL;
1861 return buffer;
1864 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1865 void *to,
1866 const unsigned int len)
1868 memcpy(to, skb->data, len);
1871 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1872 const int offset, void *to,
1873 const unsigned int len)
1875 memcpy(to, skb->data + offset, len);
1878 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1879 const void *from,
1880 const unsigned int len)
1882 memcpy(skb->data, from, len);
1885 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1886 const int offset,
1887 const void *from,
1888 const unsigned int len)
1890 memcpy(skb->data + offset, from, len);
1893 extern void skb_init(void);
1895 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1897 return skb->tstamp;
1901 * skb_get_timestamp - get timestamp from a skb
1902 * @skb: skb to get stamp from
1903 * @stamp: pointer to struct timeval to store stamp in
1905 * Timestamps are stored in the skb as offsets to a base timestamp.
1906 * This function converts the offset back to a struct timeval and stores
1907 * it in stamp.
1909 static inline void skb_get_timestamp(const struct sk_buff *skb,
1910 struct timeval *stamp)
1912 *stamp = ktime_to_timeval(skb->tstamp);
1915 static inline void skb_get_timestampns(const struct sk_buff *skb,
1916 struct timespec *stamp)
1918 *stamp = ktime_to_timespec(skb->tstamp);
1921 static inline void __net_timestamp(struct sk_buff *skb)
1923 skb->tstamp = ktime_get_real();
1926 static inline ktime_t net_timedelta(ktime_t t)
1928 return ktime_sub(ktime_get_real(), t);
1931 static inline ktime_t net_invalid_timestamp(void)
1933 return ktime_set(0, 0);
1937 * skb_tstamp_tx - queue clone of skb with send time stamps
1938 * @orig_skb: the original outgoing packet
1939 * @hwtstamps: hardware time stamps, may be NULL if not available
1941 * If the skb has a socket associated, then this function clones the
1942 * skb (thus sharing the actual data and optional structures), stores
1943 * the optional hardware time stamping information (if non NULL) or
1944 * generates a software time stamp (otherwise), then queues the clone
1945 * to the error queue of the socket. Errors are silently ignored.
1947 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1948 struct skb_shared_hwtstamps *hwtstamps);
1950 static inline void sw_tx_timestamp(struct sk_buff *skb)
1952 union skb_shared_tx *shtx = skb_tx(skb);
1953 if (shtx->software && !shtx->in_progress)
1954 skb_tstamp_tx(skb, NULL);
1958 * skb_tx_timestamp() - Driver hook for transmit timestamping
1960 * Ethernet MAC Drivers should call this function in their hard_xmit()
1961 * function as soon as possible after giving the sk_buff to the MAC
1962 * hardware, but before freeing the sk_buff.
1964 * @skb: A socket buffer.
1966 static inline void skb_tx_timestamp(struct sk_buff *skb)
1968 sw_tx_timestamp(skb);
1971 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1972 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1974 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1976 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1980 * skb_checksum_complete - Calculate checksum of an entire packet
1981 * @skb: packet to process
1983 * This function calculates the checksum over the entire packet plus
1984 * the value of skb->csum. The latter can be used to supply the
1985 * checksum of a pseudo header as used by TCP/UDP. It returns the
1986 * checksum.
1988 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1989 * this function can be used to verify that checksum on received
1990 * packets. In that case the function should return zero if the
1991 * checksum is correct. In particular, this function will return zero
1992 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1993 * hardware has already verified the correctness of the checksum.
1995 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1997 return skb_csum_unnecessary(skb) ?
1998 0 : __skb_checksum_complete(skb);
2001 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2002 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2003 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2005 if (nfct && atomic_dec_and_test(&nfct->use))
2006 nf_conntrack_destroy(nfct);
2008 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2010 if (nfct)
2011 atomic_inc(&nfct->use);
2013 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2015 if (skb)
2016 atomic_inc(&skb->users);
2018 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2020 if (skb)
2021 kfree_skb(skb);
2023 #endif
2024 #ifdef CONFIG_BRIDGE_NETFILTER
2025 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2027 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2028 kfree(nf_bridge);
2030 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2032 if (nf_bridge)
2033 atomic_inc(&nf_bridge->use);
2035 #endif /* CONFIG_BRIDGE_NETFILTER */
2036 static inline void nf_reset(struct sk_buff *skb)
2038 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2039 nf_conntrack_put(skb->nfct);
2040 skb->nfct = NULL;
2041 nf_conntrack_put_reasm(skb->nfct_reasm);
2042 skb->nfct_reasm = NULL;
2043 #endif
2044 #ifdef CONFIG_BRIDGE_NETFILTER
2045 nf_bridge_put(skb->nf_bridge);
2046 skb->nf_bridge = NULL;
2047 #endif
2050 /* Note: This doesn't put any conntrack and bridge info in dst. */
2051 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2053 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2054 dst->nfct = src->nfct;
2055 nf_conntrack_get(src->nfct);
2056 dst->nfctinfo = src->nfctinfo;
2057 dst->nfct_reasm = src->nfct_reasm;
2058 nf_conntrack_get_reasm(src->nfct_reasm);
2059 #endif
2060 #ifdef CONFIG_BRIDGE_NETFILTER
2061 dst->nf_bridge = src->nf_bridge;
2062 nf_bridge_get(src->nf_bridge);
2063 #endif
2066 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2068 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2069 nf_conntrack_put(dst->nfct);
2070 nf_conntrack_put_reasm(dst->nfct_reasm);
2071 #endif
2072 #ifdef CONFIG_BRIDGE_NETFILTER
2073 nf_bridge_put(dst->nf_bridge);
2074 #endif
2075 __nf_copy(dst, src);
2078 #ifdef CONFIG_NETWORK_SECMARK
2079 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2081 to->secmark = from->secmark;
2084 static inline void skb_init_secmark(struct sk_buff *skb)
2086 skb->secmark = 0;
2088 #else
2089 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2092 static inline void skb_init_secmark(struct sk_buff *skb)
2094 #endif
2096 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2098 skb->queue_mapping = queue_mapping;
2101 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2103 return skb->queue_mapping;
2106 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2108 to->queue_mapping = from->queue_mapping;
2111 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2113 skb->queue_mapping = rx_queue + 1;
2116 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2118 return skb->queue_mapping - 1;
2121 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2123 return (skb->queue_mapping != 0);
2126 extern u16 skb_tx_hash(const struct net_device *dev,
2127 const struct sk_buff *skb);
2129 #ifdef CONFIG_XFRM
2130 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2132 return skb->sp;
2134 #else
2135 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2137 return NULL;
2139 #endif
2141 static inline int skb_is_gso(const struct sk_buff *skb)
2143 return skb_shinfo(skb)->gso_size;
2146 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2148 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2151 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2153 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2155 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2156 * wanted then gso_type will be set. */
2157 struct skb_shared_info *shinfo = skb_shinfo(skb);
2158 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2159 unlikely(shinfo->gso_type == 0)) {
2160 __skb_warn_lro_forwarding(skb);
2161 return true;
2163 return false;
2166 static inline void skb_forward_csum(struct sk_buff *skb)
2168 /* Unfortunately we don't support this one. Any brave souls? */
2169 if (skb->ip_summed == CHECKSUM_COMPLETE)
2170 skb->ip_summed = CHECKSUM_NONE;
2173 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2174 #endif /* __KERNEL__ */
2175 #endif /* _LINUX_SKBUFF_H */