net/fec: add mac field into platform data and consolidate fec_get_mac
[linux-2.6/btrfs-unstable.git] / include / linux / skbuff.h
blob20ec0a64cb9ff0f8708a66ddaeb2ffd1d3ed72c9
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 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
133 __u32 page_offset;
134 __u32 size;
135 #else
136 __u16 page_offset;
137 __u16 size;
138 #endif
141 #define HAVE_HW_TIME_STAMP
144 * struct skb_shared_hwtstamps - hardware time stamps
145 * @hwtstamp: hardware time stamp transformed into duration
146 * since arbitrary point in time
147 * @syststamp: hwtstamp transformed to system time base
149 * Software time stamps generated by ktime_get_real() are stored in
150 * skb->tstamp. The relation between the different kinds of time
151 * stamps is as follows:
153 * syststamp and tstamp can be compared against each other in
154 * arbitrary combinations. The accuracy of a
155 * syststamp/tstamp/"syststamp from other device" comparison is
156 * limited by the accuracy of the transformation into system time
157 * base. This depends on the device driver and its underlying
158 * hardware.
160 * hwtstamps can only be compared against other hwtstamps from
161 * the same device.
163 * This structure is attached to packets as part of the
164 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
166 struct skb_shared_hwtstamps {
167 ktime_t hwtstamp;
168 ktime_t syststamp;
171 /* Definitions for tx_flags in struct skb_shared_info */
172 enum {
173 /* generate hardware time stamp */
174 SKBTX_HW_TSTAMP = 1 << 0,
176 /* generate software time stamp */
177 SKBTX_SW_TSTAMP = 1 << 1,
179 /* device driver is going to provide hardware time stamp */
180 SKBTX_IN_PROGRESS = 1 << 2,
182 /* ensure the originating sk reference is available on driver level */
183 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
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 __u8 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 /* Intermediate layers must ensure that destructor_arg
206 * remains valid until skb destructor */
207 void * destructor_arg;
208 /* must be last field, see pskb_expand_head() */
209 skb_frag_t frags[MAX_SKB_FRAGS];
212 /* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
223 #define SKB_DATAREF_SHIFT 16
224 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
227 enum {
228 SKB_FCLONE_UNAVAILABLE,
229 SKB_FCLONE_ORIG,
230 SKB_FCLONE_CLONE,
233 enum {
234 SKB_GSO_TCPV4 = 1 << 0,
235 SKB_GSO_UDP = 1 << 1,
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY = 1 << 2,
240 /* This indicates the tcp segment has CWR set. */
241 SKB_GSO_TCP_ECN = 1 << 3,
243 SKB_GSO_TCPV6 = 1 << 4,
245 SKB_GSO_FCOE = 1 << 5,
248 #if BITS_PER_LONG > 32
249 #define NET_SKBUFF_DATA_USES_OFFSET 1
250 #endif
252 #ifdef NET_SKBUFF_DATA_USES_OFFSET
253 typedef unsigned int sk_buff_data_t;
254 #else
255 typedef unsigned char *sk_buff_data_t;
256 #endif
258 /**
259 * struct sk_buff - socket buffer
260 * @next: Next buffer in list
261 * @prev: Previous buffer in list
262 * @sk: Socket we are owned by
263 * @tstamp: Time we arrived
264 * @dev: Device we arrived on/are leaving by
265 * @transport_header: Transport layer header
266 * @network_header: Network layer header
267 * @mac_header: Link layer header
268 * @_skb_refdst: destination entry (with norefcount bit)
269 * @sp: the security path, used for xfrm
270 * @cb: Control buffer. Free for use by every layer. Put private vars here
271 * @len: Length of actual data
272 * @data_len: Data length
273 * @mac_len: Length of link layer header
274 * @hdr_len: writable header length of cloned skb
275 * @csum: Checksum (must include start/offset pair)
276 * @csum_start: Offset from skb->head where checksumming should start
277 * @csum_offset: Offset from csum_start where checksum should be stored
278 * @local_df: allow local fragmentation
279 * @cloned: Head may be cloned (check refcnt to be sure)
280 * @nohdr: Payload reference only, must not modify header
281 * @pkt_type: Packet class
282 * @fclone: skbuff clone status
283 * @ip_summed: Driver fed us an IP checksum
284 * @priority: Packet queueing priority
285 * @users: User count - see {datagram,tcp}.c
286 * @protocol: Packet protocol from driver
287 * @truesize: Buffer size
288 * @head: Head of buffer
289 * @data: Data head pointer
290 * @tail: Tail pointer
291 * @end: End pointer
292 * @destructor: Destruct function
293 * @mark: Generic packet mark
294 * @nfct: Associated connection, if any
295 * @ipvs_property: skbuff is owned by ipvs
296 * @peeked: this packet has been seen already, so stats have been
297 * done for it, don't do them again
298 * @nf_trace: netfilter packet trace flag
299 * @nfctinfo: Relationship of this skb to the connection
300 * @nfct_reasm: netfilter conntrack re-assembly pointer
301 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
302 * @skb_iif: ifindex of device we arrived on
303 * @rxhash: the packet hash computed on receive
304 * @queue_mapping: Queue mapping for multiqueue devices
305 * @tc_index: Traffic control index
306 * @tc_verd: traffic control verdict
307 * @ndisc_nodetype: router type (from link layer)
308 * @dma_cookie: a cookie to one of several possible DMA operations
309 * done by skb DMA functions
310 * @secmark: security marking
311 * @vlan_tci: vlan tag control information
314 struct sk_buff {
315 /* These two members must be first. */
316 struct sk_buff *next;
317 struct sk_buff *prev;
319 ktime_t tstamp;
321 struct sock *sk;
322 struct net_device *dev;
325 * This is the control buffer. It is free to use for every
326 * layer. Please put your private variables there. If you
327 * want to keep them across layers you have to do a skb_clone()
328 * first. This is owned by whoever has the skb queued ATM.
330 char cb[48] __aligned(8);
332 unsigned long _skb_refdst;
333 #ifdef CONFIG_XFRM
334 struct sec_path *sp;
335 #endif
336 unsigned int len,
337 data_len;
338 __u16 mac_len,
339 hdr_len;
340 union {
341 __wsum csum;
342 struct {
343 __u16 csum_start;
344 __u16 csum_offset;
347 __u32 priority;
348 kmemcheck_bitfield_begin(flags1);
349 __u8 local_df:1,
350 cloned:1,
351 ip_summed:2,
352 nohdr:1,
353 nfctinfo:3;
354 __u8 pkt_type:3,
355 fclone:2,
356 ipvs_property:1,
357 peeked:1,
358 nf_trace:1;
359 kmemcheck_bitfield_end(flags1);
360 __be16 protocol;
362 void (*destructor)(struct sk_buff *skb);
363 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
364 struct nf_conntrack *nfct;
365 struct sk_buff *nfct_reasm;
366 #endif
367 #ifdef CONFIG_BRIDGE_NETFILTER
368 struct nf_bridge_info *nf_bridge;
369 #endif
371 int skb_iif;
372 #ifdef CONFIG_NET_SCHED
373 __u16 tc_index; /* traffic control index */
374 #ifdef CONFIG_NET_CLS_ACT
375 __u16 tc_verd; /* traffic control verdict */
376 #endif
377 #endif
379 __u32 rxhash;
381 kmemcheck_bitfield_begin(flags2);
382 __u16 queue_mapping:16;
383 #ifdef CONFIG_IPV6_NDISC_NODETYPE
384 __u8 ndisc_nodetype:2,
385 deliver_no_wcard:1;
386 #else
387 __u8 deliver_no_wcard:1;
388 #endif
389 __u8 ooo_okay:1;
390 kmemcheck_bitfield_end(flags2);
392 /* 0/13 bit hole */
394 #ifdef CONFIG_NET_DMA
395 dma_cookie_t dma_cookie;
396 #endif
397 #ifdef CONFIG_NETWORK_SECMARK
398 __u32 secmark;
399 #endif
400 union {
401 __u32 mark;
402 __u32 dropcount;
405 __u16 vlan_tci;
407 sk_buff_data_t transport_header;
408 sk_buff_data_t network_header;
409 sk_buff_data_t mac_header;
410 /* These elements must be at the end, see alloc_skb() for details. */
411 sk_buff_data_t tail;
412 sk_buff_data_t end;
413 unsigned char *head,
414 *data;
415 unsigned int truesize;
416 atomic_t users;
419 #ifdef __KERNEL__
421 * Handling routines are only of interest to the kernel
423 #include <linux/slab.h>
425 #include <asm/system.h>
428 * skb might have a dst pointer attached, refcounted or not.
429 * _skb_refdst low order bit is set if refcount was _not_ taken
431 #define SKB_DST_NOREF 1UL
432 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
435 * skb_dst - returns skb dst_entry
436 * @skb: buffer
438 * Returns skb dst_entry, regardless of reference taken or not.
440 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
442 /* If refdst was not refcounted, check we still are in a
443 * rcu_read_lock section
445 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
446 !rcu_read_lock_held() &&
447 !rcu_read_lock_bh_held());
448 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
452 * skb_dst_set - sets skb dst
453 * @skb: buffer
454 * @dst: dst entry
456 * Sets skb dst, assuming a reference was taken on dst and should
457 * be released by skb_dst_drop()
459 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
461 skb->_skb_refdst = (unsigned long)dst;
464 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
467 * skb_dst_is_noref - Test if skb dst isnt refcounted
468 * @skb: buffer
470 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
472 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
475 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
477 return (struct rtable *)skb_dst(skb);
480 extern void kfree_skb(struct sk_buff *skb);
481 extern void consume_skb(struct sk_buff *skb);
482 extern void __kfree_skb(struct sk_buff *skb);
483 extern struct sk_buff *__alloc_skb(unsigned int size,
484 gfp_t priority, int fclone, int node);
485 static inline struct sk_buff *alloc_skb(unsigned int size,
486 gfp_t priority)
488 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
491 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
492 gfp_t priority)
494 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
497 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
499 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
500 extern struct sk_buff *skb_clone(struct sk_buff *skb,
501 gfp_t priority);
502 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
503 gfp_t priority);
504 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
505 gfp_t gfp_mask);
506 extern int pskb_expand_head(struct sk_buff *skb,
507 int nhead, int ntail,
508 gfp_t gfp_mask);
509 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
510 unsigned int headroom);
511 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
512 int newheadroom, int newtailroom,
513 gfp_t priority);
514 extern int skb_to_sgvec(struct sk_buff *skb,
515 struct scatterlist *sg, int offset,
516 int len);
517 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
518 struct sk_buff **trailer);
519 extern int skb_pad(struct sk_buff *skb, int pad);
520 #define dev_kfree_skb(a) consume_skb(a)
522 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
523 int getfrag(void *from, char *to, int offset,
524 int len,int odd, struct sk_buff *skb),
525 void *from, int length);
527 struct skb_seq_state {
528 __u32 lower_offset;
529 __u32 upper_offset;
530 __u32 frag_idx;
531 __u32 stepped_offset;
532 struct sk_buff *root_skb;
533 struct sk_buff *cur_skb;
534 __u8 *frag_data;
537 extern void skb_prepare_seq_read(struct sk_buff *skb,
538 unsigned int from, unsigned int to,
539 struct skb_seq_state *st);
540 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
541 struct skb_seq_state *st);
542 extern void skb_abort_seq_read(struct skb_seq_state *st);
544 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
545 unsigned int to, struct ts_config *config,
546 struct ts_state *state);
548 extern __u32 __skb_get_rxhash(struct sk_buff *skb);
549 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
551 if (!skb->rxhash)
552 skb->rxhash = __skb_get_rxhash(skb);
554 return skb->rxhash;
557 #ifdef NET_SKBUFF_DATA_USES_OFFSET
558 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
560 return skb->head + skb->end;
562 #else
563 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
565 return skb->end;
567 #endif
569 /* Internal */
570 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
572 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
574 return &skb_shinfo(skb)->hwtstamps;
578 * skb_queue_empty - check if a queue is empty
579 * @list: queue head
581 * Returns true if the queue is empty, false otherwise.
583 static inline int skb_queue_empty(const struct sk_buff_head *list)
585 return list->next == (struct sk_buff *)list;
589 * skb_queue_is_last - check if skb is the last entry in the queue
590 * @list: queue head
591 * @skb: buffer
593 * Returns true if @skb is the last buffer on the list.
595 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
596 const struct sk_buff *skb)
598 return skb->next == (struct sk_buff *)list;
602 * skb_queue_is_first - check if skb is the first entry in the queue
603 * @list: queue head
604 * @skb: buffer
606 * Returns true if @skb is the first buffer on the list.
608 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
609 const struct sk_buff *skb)
611 return skb->prev == (struct sk_buff *)list;
615 * skb_queue_next - return the next packet in the queue
616 * @list: queue head
617 * @skb: current buffer
619 * Return the next packet in @list after @skb. It is only valid to
620 * call this if skb_queue_is_last() evaluates to false.
622 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
623 const struct sk_buff *skb)
625 /* This BUG_ON may seem severe, but if we just return then we
626 * are going to dereference garbage.
628 BUG_ON(skb_queue_is_last(list, skb));
629 return skb->next;
633 * skb_queue_prev - return the prev packet in the queue
634 * @list: queue head
635 * @skb: current buffer
637 * Return the prev packet in @list before @skb. It is only valid to
638 * call this if skb_queue_is_first() evaluates to false.
640 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
641 const struct sk_buff *skb)
643 /* This BUG_ON may seem severe, but if we just return then we
644 * are going to dereference garbage.
646 BUG_ON(skb_queue_is_first(list, skb));
647 return skb->prev;
651 * skb_get - reference buffer
652 * @skb: buffer to reference
654 * Makes another reference to a socket buffer and returns a pointer
655 * to the buffer.
657 static inline struct sk_buff *skb_get(struct sk_buff *skb)
659 atomic_inc(&skb->users);
660 return skb;
664 * If users == 1, we are the only owner and are can avoid redundant
665 * atomic change.
669 * skb_cloned - is the buffer a clone
670 * @skb: buffer to check
672 * Returns true if the buffer was generated with skb_clone() and is
673 * one of multiple shared copies of the buffer. Cloned buffers are
674 * shared data so must not be written to under normal circumstances.
676 static inline int skb_cloned(const struct sk_buff *skb)
678 return skb->cloned &&
679 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
683 * skb_header_cloned - is the header a clone
684 * @skb: buffer to check
686 * Returns true if modifying the header part of the buffer requires
687 * the data to be copied.
689 static inline int skb_header_cloned(const struct sk_buff *skb)
691 int dataref;
693 if (!skb->cloned)
694 return 0;
696 dataref = atomic_read(&skb_shinfo(skb)->dataref);
697 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
698 return dataref != 1;
702 * skb_header_release - release reference to header
703 * @skb: buffer to operate on
705 * Drop a reference to the header part of the buffer. This is done
706 * by acquiring a payload reference. You must not read from the header
707 * part of skb->data after this.
709 static inline void skb_header_release(struct sk_buff *skb)
711 BUG_ON(skb->nohdr);
712 skb->nohdr = 1;
713 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
717 * skb_shared - is the buffer shared
718 * @skb: buffer to check
720 * Returns true if more than one person has a reference to this
721 * buffer.
723 static inline int skb_shared(const struct sk_buff *skb)
725 return atomic_read(&skb->users) != 1;
729 * skb_share_check - check if buffer is shared and if so clone it
730 * @skb: buffer to check
731 * @pri: priority for memory allocation
733 * If the buffer is shared the buffer is cloned and the old copy
734 * drops a reference. A new clone with a single reference is returned.
735 * If the buffer is not shared the original buffer is returned. When
736 * being called from interrupt status or with spinlocks held pri must
737 * be GFP_ATOMIC.
739 * NULL is returned on a memory allocation failure.
741 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
742 gfp_t pri)
744 might_sleep_if(pri & __GFP_WAIT);
745 if (skb_shared(skb)) {
746 struct sk_buff *nskb = skb_clone(skb, pri);
747 kfree_skb(skb);
748 skb = nskb;
750 return skb;
754 * Copy shared buffers into a new sk_buff. We effectively do COW on
755 * packets to handle cases where we have a local reader and forward
756 * and a couple of other messy ones. The normal one is tcpdumping
757 * a packet thats being forwarded.
761 * skb_unshare - make a copy of a shared buffer
762 * @skb: buffer to check
763 * @pri: priority for memory allocation
765 * If the socket buffer is a clone then this function creates a new
766 * copy of the data, drops a reference count on the old copy and returns
767 * the new copy with the reference count at 1. If the buffer is not a clone
768 * the original buffer is returned. When called with a spinlock held or
769 * from interrupt state @pri must be %GFP_ATOMIC
771 * %NULL is returned on a memory allocation failure.
773 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
774 gfp_t pri)
776 might_sleep_if(pri & __GFP_WAIT);
777 if (skb_cloned(skb)) {
778 struct sk_buff *nskb = skb_copy(skb, pri);
779 kfree_skb(skb); /* Free our shared copy */
780 skb = nskb;
782 return skb;
786 * skb_peek - peek at the head of an &sk_buff_head
787 * @list_: list to peek at
789 * Peek an &sk_buff. Unlike most other operations you _MUST_
790 * be careful with this one. A peek leaves the buffer on the
791 * list and someone else may run off with it. You must hold
792 * the appropriate locks or have a private queue to do this.
794 * Returns %NULL for an empty list or a pointer to the head element.
795 * The reference count is not incremented and the reference is therefore
796 * volatile. Use with caution.
798 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
800 struct sk_buff *list = ((struct sk_buff *)list_)->next;
801 if (list == (struct sk_buff *)list_)
802 list = NULL;
803 return list;
807 * skb_peek_tail - peek at the tail of an &sk_buff_head
808 * @list_: list to peek at
810 * Peek an &sk_buff. Unlike most other operations you _MUST_
811 * be careful with this one. A peek leaves the buffer on the
812 * list and someone else may run off with it. You must hold
813 * the appropriate locks or have a private queue to do this.
815 * Returns %NULL for an empty list or a pointer to the tail element.
816 * The reference count is not incremented and the reference is therefore
817 * volatile. Use with caution.
819 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
821 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
822 if (list == (struct sk_buff *)list_)
823 list = NULL;
824 return list;
828 * skb_queue_len - get queue length
829 * @list_: list to measure
831 * Return the length of an &sk_buff queue.
833 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
835 return list_->qlen;
839 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
840 * @list: queue to initialize
842 * This initializes only the list and queue length aspects of
843 * an sk_buff_head object. This allows to initialize the list
844 * aspects of an sk_buff_head without reinitializing things like
845 * the spinlock. It can also be used for on-stack sk_buff_head
846 * objects where the spinlock is known to not be used.
848 static inline void __skb_queue_head_init(struct sk_buff_head *list)
850 list->prev = list->next = (struct sk_buff *)list;
851 list->qlen = 0;
855 * This function creates a split out lock class for each invocation;
856 * this is needed for now since a whole lot of users of the skb-queue
857 * infrastructure in drivers have different locking usage (in hardirq)
858 * than the networking core (in softirq only). In the long run either the
859 * network layer or drivers should need annotation to consolidate the
860 * main types of usage into 3 classes.
862 static inline void skb_queue_head_init(struct sk_buff_head *list)
864 spin_lock_init(&list->lock);
865 __skb_queue_head_init(list);
868 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
869 struct lock_class_key *class)
871 skb_queue_head_init(list);
872 lockdep_set_class(&list->lock, class);
876 * Insert an sk_buff on a list.
878 * The "__skb_xxxx()" functions are the non-atomic ones that
879 * can only be called with interrupts disabled.
881 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
882 static inline void __skb_insert(struct sk_buff *newsk,
883 struct sk_buff *prev, struct sk_buff *next,
884 struct sk_buff_head *list)
886 newsk->next = next;
887 newsk->prev = prev;
888 next->prev = prev->next = newsk;
889 list->qlen++;
892 static inline void __skb_queue_splice(const struct sk_buff_head *list,
893 struct sk_buff *prev,
894 struct sk_buff *next)
896 struct sk_buff *first = list->next;
897 struct sk_buff *last = list->prev;
899 first->prev = prev;
900 prev->next = first;
902 last->next = next;
903 next->prev = last;
907 * skb_queue_splice - join two skb lists, this is designed for stacks
908 * @list: the new list to add
909 * @head: the place to add it in the first list
911 static inline void skb_queue_splice(const struct sk_buff_head *list,
912 struct sk_buff_head *head)
914 if (!skb_queue_empty(list)) {
915 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
916 head->qlen += list->qlen;
921 * skb_queue_splice - join two skb lists and reinitialise the emptied list
922 * @list: the new list to add
923 * @head: the place to add it in the first list
925 * The list at @list is reinitialised
927 static inline void skb_queue_splice_init(struct sk_buff_head *list,
928 struct sk_buff_head *head)
930 if (!skb_queue_empty(list)) {
931 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
932 head->qlen += list->qlen;
933 __skb_queue_head_init(list);
938 * skb_queue_splice_tail - join two skb lists, each list being a queue
939 * @list: the new list to add
940 * @head: the place to add it in the first list
942 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
943 struct sk_buff_head *head)
945 if (!skb_queue_empty(list)) {
946 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
947 head->qlen += list->qlen;
952 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
953 * @list: the new list to add
954 * @head: the place to add it in the first list
956 * Each of the lists is a queue.
957 * The list at @list is reinitialised
959 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
960 struct sk_buff_head *head)
962 if (!skb_queue_empty(list)) {
963 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
964 head->qlen += list->qlen;
965 __skb_queue_head_init(list);
970 * __skb_queue_after - queue a buffer at the list head
971 * @list: list to use
972 * @prev: place after this buffer
973 * @newsk: buffer to queue
975 * Queue a buffer int the middle of a list. This function takes no locks
976 * and you must therefore hold required locks before calling it.
978 * A buffer cannot be placed on two lists at the same time.
980 static inline void __skb_queue_after(struct sk_buff_head *list,
981 struct sk_buff *prev,
982 struct sk_buff *newsk)
984 __skb_insert(newsk, prev, prev->next, list);
987 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
988 struct sk_buff_head *list);
990 static inline void __skb_queue_before(struct sk_buff_head *list,
991 struct sk_buff *next,
992 struct sk_buff *newsk)
994 __skb_insert(newsk, next->prev, next, list);
998 * __skb_queue_head - queue a buffer at the list head
999 * @list: list to use
1000 * @newsk: buffer to queue
1002 * Queue a buffer at the start of a list. This function takes no locks
1003 * and you must therefore hold required locks before calling it.
1005 * A buffer cannot be placed on two lists at the same time.
1007 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1008 static inline void __skb_queue_head(struct sk_buff_head *list,
1009 struct sk_buff *newsk)
1011 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1015 * __skb_queue_tail - queue a buffer at the list tail
1016 * @list: list to use
1017 * @newsk: buffer to queue
1019 * Queue a buffer at the end of a list. This function takes no locks
1020 * and you must therefore hold required locks before calling it.
1022 * A buffer cannot be placed on two lists at the same time.
1024 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1025 static inline void __skb_queue_tail(struct sk_buff_head *list,
1026 struct sk_buff *newsk)
1028 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1032 * remove sk_buff from list. _Must_ be called atomically, and with
1033 * the list known..
1035 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1036 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1038 struct sk_buff *next, *prev;
1040 list->qlen--;
1041 next = skb->next;
1042 prev = skb->prev;
1043 skb->next = skb->prev = NULL;
1044 next->prev = prev;
1045 prev->next = next;
1049 * __skb_dequeue - remove from the head of the queue
1050 * @list: list to dequeue from
1052 * Remove the head of the list. This function does not take any locks
1053 * so must be used with appropriate locks held only. The head item is
1054 * returned or %NULL if the list is empty.
1056 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1057 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1059 struct sk_buff *skb = skb_peek(list);
1060 if (skb)
1061 __skb_unlink(skb, list);
1062 return skb;
1066 * __skb_dequeue_tail - remove from the tail of the queue
1067 * @list: list to dequeue from
1069 * Remove the tail of the list. This function does not take any locks
1070 * so must be used with appropriate locks held only. The tail item is
1071 * returned or %NULL if the list is empty.
1073 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1074 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1076 struct sk_buff *skb = skb_peek_tail(list);
1077 if (skb)
1078 __skb_unlink(skb, list);
1079 return skb;
1083 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1085 return skb->data_len;
1088 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1090 return skb->len - skb->data_len;
1093 static inline int skb_pagelen(const struct sk_buff *skb)
1095 int i, len = 0;
1097 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1098 len += skb_shinfo(skb)->frags[i].size;
1099 return len + skb_headlen(skb);
1102 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1103 struct page *page, int off, int size)
1105 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1107 frag->page = page;
1108 frag->page_offset = off;
1109 frag->size = size;
1110 skb_shinfo(skb)->nr_frags = i + 1;
1113 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1114 int off, int size);
1116 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1117 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1118 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1120 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1121 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1123 return skb->head + skb->tail;
1126 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1128 skb->tail = skb->data - skb->head;
1131 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1133 skb_reset_tail_pointer(skb);
1134 skb->tail += offset;
1136 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1137 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1139 return skb->tail;
1142 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1144 skb->tail = skb->data;
1147 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1149 skb->tail = skb->data + offset;
1152 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1155 * Add data to an sk_buff
1157 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1158 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1160 unsigned char *tmp = skb_tail_pointer(skb);
1161 SKB_LINEAR_ASSERT(skb);
1162 skb->tail += len;
1163 skb->len += len;
1164 return tmp;
1167 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1168 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1170 skb->data -= len;
1171 skb->len += len;
1172 return skb->data;
1175 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1176 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1178 skb->len -= len;
1179 BUG_ON(skb->len < skb->data_len);
1180 return skb->data += len;
1183 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1185 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1188 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1190 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1192 if (len > skb_headlen(skb) &&
1193 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1194 return NULL;
1195 skb->len -= len;
1196 return skb->data += len;
1199 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1201 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1204 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1206 if (likely(len <= skb_headlen(skb)))
1207 return 1;
1208 if (unlikely(len > skb->len))
1209 return 0;
1210 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1214 * skb_headroom - bytes at buffer head
1215 * @skb: buffer to check
1217 * Return the number of bytes of free space at the head of an &sk_buff.
1219 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1221 return skb->data - skb->head;
1225 * skb_tailroom - bytes at buffer end
1226 * @skb: buffer to check
1228 * Return the number of bytes of free space at the tail of an sk_buff
1230 static inline int skb_tailroom(const struct sk_buff *skb)
1232 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1236 * skb_reserve - adjust headroom
1237 * @skb: buffer to alter
1238 * @len: bytes to move
1240 * Increase the headroom of an empty &sk_buff by reducing the tail
1241 * room. This is only allowed for an empty buffer.
1243 static inline void skb_reserve(struct sk_buff *skb, int len)
1245 skb->data += len;
1246 skb->tail += len;
1249 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1250 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1252 return skb->head + skb->transport_header;
1255 static inline void skb_reset_transport_header(struct sk_buff *skb)
1257 skb->transport_header = skb->data - skb->head;
1260 static inline void skb_set_transport_header(struct sk_buff *skb,
1261 const int offset)
1263 skb_reset_transport_header(skb);
1264 skb->transport_header += offset;
1267 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1269 return skb->head + skb->network_header;
1272 static inline void skb_reset_network_header(struct sk_buff *skb)
1274 skb->network_header = skb->data - skb->head;
1277 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1279 skb_reset_network_header(skb);
1280 skb->network_header += offset;
1283 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1285 return skb->head + skb->mac_header;
1288 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1290 return skb->mac_header != ~0U;
1293 static inline void skb_reset_mac_header(struct sk_buff *skb)
1295 skb->mac_header = skb->data - skb->head;
1298 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1300 skb_reset_mac_header(skb);
1301 skb->mac_header += offset;
1304 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1306 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1308 return skb->transport_header;
1311 static inline void skb_reset_transport_header(struct sk_buff *skb)
1313 skb->transport_header = skb->data;
1316 static inline void skb_set_transport_header(struct sk_buff *skb,
1317 const int offset)
1319 skb->transport_header = skb->data + offset;
1322 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1324 return skb->network_header;
1327 static inline void skb_reset_network_header(struct sk_buff *skb)
1329 skb->network_header = skb->data;
1332 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1334 skb->network_header = skb->data + offset;
1337 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1339 return skb->mac_header;
1342 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1344 return skb->mac_header != NULL;
1347 static inline void skb_reset_mac_header(struct sk_buff *skb)
1349 skb->mac_header = skb->data;
1352 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1354 skb->mac_header = skb->data + offset;
1356 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1358 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1360 return skb->csum_start - skb_headroom(skb);
1363 static inline int skb_transport_offset(const struct sk_buff *skb)
1365 return skb_transport_header(skb) - skb->data;
1368 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1370 return skb->transport_header - skb->network_header;
1373 static inline int skb_network_offset(const struct sk_buff *skb)
1375 return skb_network_header(skb) - skb->data;
1378 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1380 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1384 * CPUs often take a performance hit when accessing unaligned memory
1385 * locations. The actual performance hit varies, it can be small if the
1386 * hardware handles it or large if we have to take an exception and fix it
1387 * in software.
1389 * Since an ethernet header is 14 bytes network drivers often end up with
1390 * the IP header at an unaligned offset. The IP header can be aligned by
1391 * shifting the start of the packet by 2 bytes. Drivers should do this
1392 * with:
1394 * skb_reserve(skb, NET_IP_ALIGN);
1396 * The downside to this alignment of the IP header is that the DMA is now
1397 * unaligned. On some architectures the cost of an unaligned DMA is high
1398 * and this cost outweighs the gains made by aligning the IP header.
1400 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1401 * to be overridden.
1403 #ifndef NET_IP_ALIGN
1404 #define NET_IP_ALIGN 2
1405 #endif
1408 * The networking layer reserves some headroom in skb data (via
1409 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1410 * the header has to grow. In the default case, if the header has to grow
1411 * 32 bytes or less we avoid the reallocation.
1413 * Unfortunately this headroom changes the DMA alignment of the resulting
1414 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1415 * on some architectures. An architecture can override this value,
1416 * perhaps setting it to a cacheline in size (since that will maintain
1417 * cacheline alignment of the DMA). It must be a power of 2.
1419 * Various parts of the networking layer expect at least 32 bytes of
1420 * headroom, you should not reduce this.
1422 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1423 * to reduce average number of cache lines per packet.
1424 * get_rps_cpus() for example only access one 64 bytes aligned block :
1425 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1427 #ifndef NET_SKB_PAD
1428 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1429 #endif
1431 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1433 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1435 if (unlikely(skb->data_len)) {
1436 WARN_ON(1);
1437 return;
1439 skb->len = len;
1440 skb_set_tail_pointer(skb, len);
1443 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1445 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1447 if (skb->data_len)
1448 return ___pskb_trim(skb, len);
1449 __skb_trim(skb, len);
1450 return 0;
1453 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1455 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1459 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1460 * @skb: buffer to alter
1461 * @len: new length
1463 * This is identical to pskb_trim except that the caller knows that
1464 * the skb is not cloned so we should never get an error due to out-
1465 * of-memory.
1467 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1469 int err = pskb_trim(skb, len);
1470 BUG_ON(err);
1474 * skb_orphan - orphan a buffer
1475 * @skb: buffer to orphan
1477 * If a buffer currently has an owner then we call the owner's
1478 * destructor function and make the @skb unowned. The buffer continues
1479 * to exist but is no longer charged to its former owner.
1481 static inline void skb_orphan(struct sk_buff *skb)
1483 if (skb->destructor)
1484 skb->destructor(skb);
1485 skb->destructor = NULL;
1486 skb->sk = NULL;
1490 * __skb_queue_purge - empty a list
1491 * @list: list to empty
1493 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1494 * the list and one reference dropped. This function does not take the
1495 * list lock and the caller must hold the relevant locks to use it.
1497 extern void skb_queue_purge(struct sk_buff_head *list);
1498 static inline void __skb_queue_purge(struct sk_buff_head *list)
1500 struct sk_buff *skb;
1501 while ((skb = __skb_dequeue(list)) != NULL)
1502 kfree_skb(skb);
1506 * __dev_alloc_skb - allocate an skbuff for receiving
1507 * @length: length to allocate
1508 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1510 * Allocate a new &sk_buff and assign it a usage count of one. The
1511 * buffer has unspecified headroom built in. Users should allocate
1512 * the headroom they think they need without accounting for the
1513 * built in space. The built in space is used for optimisations.
1515 * %NULL is returned if there is no free memory.
1517 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1518 gfp_t gfp_mask)
1520 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1521 if (likely(skb))
1522 skb_reserve(skb, NET_SKB_PAD);
1523 return skb;
1526 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1528 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1529 unsigned int length, gfp_t gfp_mask);
1532 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1533 * @dev: network device to receive on
1534 * @length: length to allocate
1536 * Allocate a new &sk_buff and assign it a usage count of one. The
1537 * buffer has unspecified headroom built in. Users should allocate
1538 * the headroom they think they need without accounting for the
1539 * built in space. The built in space is used for optimisations.
1541 * %NULL is returned if there is no free memory. Although this function
1542 * allocates memory it can be called from an interrupt.
1544 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1545 unsigned int length)
1547 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1550 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1551 unsigned int length)
1553 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1555 if (NET_IP_ALIGN && skb)
1556 skb_reserve(skb, NET_IP_ALIGN);
1557 return skb;
1561 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1562 * @dev: network device to receive on
1563 * @gfp_mask: alloc_pages_node mask
1565 * Allocate a new page. dev currently unused.
1567 * %NULL is returned if there is no free memory.
1569 static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1571 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1575 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1576 * @dev: network device to receive on
1578 * Allocate a new page. dev currently unused.
1580 * %NULL is returned if there is no free memory.
1582 static inline struct page *netdev_alloc_page(struct net_device *dev)
1584 return __netdev_alloc_page(dev, GFP_ATOMIC);
1587 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1589 __free_page(page);
1593 * skb_clone_writable - is the header of a clone writable
1594 * @skb: buffer to check
1595 * @len: length up to which to write
1597 * Returns true if modifying the header part of the cloned buffer
1598 * does not requires the data to be copied.
1600 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1602 return !skb_header_cloned(skb) &&
1603 skb_headroom(skb) + len <= skb->hdr_len;
1606 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1607 int cloned)
1609 int delta = 0;
1611 if (headroom < NET_SKB_PAD)
1612 headroom = NET_SKB_PAD;
1613 if (headroom > skb_headroom(skb))
1614 delta = headroom - skb_headroom(skb);
1616 if (delta || cloned)
1617 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1618 GFP_ATOMIC);
1619 return 0;
1623 * skb_cow - copy header of skb when it is required
1624 * @skb: buffer to cow
1625 * @headroom: needed headroom
1627 * If the skb passed lacks sufficient headroom or its data part
1628 * is shared, data is reallocated. If reallocation fails, an error
1629 * is returned and original skb is not changed.
1631 * The result is skb with writable area skb->head...skb->tail
1632 * and at least @headroom of space at head.
1634 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1636 return __skb_cow(skb, headroom, skb_cloned(skb));
1640 * skb_cow_head - skb_cow but only making the head writable
1641 * @skb: buffer to cow
1642 * @headroom: needed headroom
1644 * This function is identical to skb_cow except that we replace the
1645 * skb_cloned check by skb_header_cloned. It should be used when
1646 * you only need to push on some header and do not need to modify
1647 * the data.
1649 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1651 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1655 * skb_padto - pad an skbuff up to a minimal size
1656 * @skb: buffer to pad
1657 * @len: minimal length
1659 * Pads up a buffer to ensure the trailing bytes exist and are
1660 * blanked. If the buffer already contains sufficient data it
1661 * is untouched. Otherwise it is extended. Returns zero on
1662 * success. The skb is freed on error.
1665 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1667 unsigned int size = skb->len;
1668 if (likely(size >= len))
1669 return 0;
1670 return skb_pad(skb, len - size);
1673 static inline int skb_add_data(struct sk_buff *skb,
1674 char __user *from, int copy)
1676 const int off = skb->len;
1678 if (skb->ip_summed == CHECKSUM_NONE) {
1679 int err = 0;
1680 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1681 copy, 0, &err);
1682 if (!err) {
1683 skb->csum = csum_block_add(skb->csum, csum, off);
1684 return 0;
1686 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1687 return 0;
1689 __skb_trim(skb, off);
1690 return -EFAULT;
1693 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1694 struct page *page, int off)
1696 if (i) {
1697 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1699 return page == frag->page &&
1700 off == frag->page_offset + frag->size;
1702 return 0;
1705 static inline int __skb_linearize(struct sk_buff *skb)
1707 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1711 * skb_linearize - convert paged skb to linear one
1712 * @skb: buffer to linarize
1714 * If there is no free memory -ENOMEM is returned, otherwise zero
1715 * is returned and the old skb data released.
1717 static inline int skb_linearize(struct sk_buff *skb)
1719 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1723 * skb_linearize_cow - make sure skb is linear and writable
1724 * @skb: buffer to process
1726 * If there is no free memory -ENOMEM is returned, otherwise zero
1727 * is returned and the old skb data released.
1729 static inline int skb_linearize_cow(struct sk_buff *skb)
1731 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1732 __skb_linearize(skb) : 0;
1736 * skb_postpull_rcsum - update checksum for received skb after pull
1737 * @skb: buffer to update
1738 * @start: start of data before pull
1739 * @len: length of data pulled
1741 * After doing a pull on a received packet, you need to call this to
1742 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1743 * CHECKSUM_NONE so that it can be recomputed from scratch.
1746 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1747 const void *start, unsigned int len)
1749 if (skb->ip_summed == CHECKSUM_COMPLETE)
1750 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1753 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1756 * pskb_trim_rcsum - trim received skb and update checksum
1757 * @skb: buffer to trim
1758 * @len: new length
1760 * This is exactly the same as pskb_trim except that it ensures the
1761 * checksum of received packets are still valid after the operation.
1764 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1766 if (likely(len >= skb->len))
1767 return 0;
1768 if (skb->ip_summed == CHECKSUM_COMPLETE)
1769 skb->ip_summed = CHECKSUM_NONE;
1770 return __pskb_trim(skb, len);
1773 #define skb_queue_walk(queue, skb) \
1774 for (skb = (queue)->next; \
1775 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1776 skb = skb->next)
1778 #define skb_queue_walk_safe(queue, skb, tmp) \
1779 for (skb = (queue)->next, tmp = skb->next; \
1780 skb != (struct sk_buff *)(queue); \
1781 skb = tmp, tmp = skb->next)
1783 #define skb_queue_walk_from(queue, skb) \
1784 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1785 skb = skb->next)
1787 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1788 for (tmp = skb->next; \
1789 skb != (struct sk_buff *)(queue); \
1790 skb = tmp, tmp = skb->next)
1792 #define skb_queue_reverse_walk(queue, skb) \
1793 for (skb = (queue)->prev; \
1794 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1795 skb = skb->prev)
1798 static inline bool skb_has_frag_list(const struct sk_buff *skb)
1800 return skb_shinfo(skb)->frag_list != NULL;
1803 static inline void skb_frag_list_init(struct sk_buff *skb)
1805 skb_shinfo(skb)->frag_list = NULL;
1808 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1810 frag->next = skb_shinfo(skb)->frag_list;
1811 skb_shinfo(skb)->frag_list = frag;
1814 #define skb_walk_frags(skb, iter) \
1815 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1817 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1818 int *peeked, int *err);
1819 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1820 int noblock, int *err);
1821 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1822 struct poll_table_struct *wait);
1823 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1824 int offset, struct iovec *to,
1825 int size);
1826 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1827 int hlen,
1828 struct iovec *iov);
1829 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1830 int offset,
1831 const struct iovec *from,
1832 int from_offset,
1833 int len);
1834 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1835 int offset,
1836 const struct iovec *to,
1837 int to_offset,
1838 int size);
1839 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1840 extern void skb_free_datagram_locked(struct sock *sk,
1841 struct sk_buff *skb);
1842 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1843 unsigned int flags);
1844 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1845 int len, __wsum csum);
1846 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1847 void *to, int len);
1848 extern int skb_store_bits(struct sk_buff *skb, int offset,
1849 const void *from, int len);
1850 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1851 int offset, u8 *to, int len,
1852 __wsum csum);
1853 extern int skb_splice_bits(struct sk_buff *skb,
1854 unsigned int offset,
1855 struct pipe_inode_info *pipe,
1856 unsigned int len,
1857 unsigned int flags);
1858 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1859 extern void skb_split(struct sk_buff *skb,
1860 struct sk_buff *skb1, const u32 len);
1861 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1862 int shiftlen);
1864 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1866 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1867 int len, void *buffer)
1869 int hlen = skb_headlen(skb);
1871 if (hlen - offset >= len)
1872 return skb->data + offset;
1874 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1875 return NULL;
1877 return buffer;
1880 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1881 void *to,
1882 const unsigned int len)
1884 memcpy(to, skb->data, len);
1887 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1888 const int offset, void *to,
1889 const unsigned int len)
1891 memcpy(to, skb->data + offset, len);
1894 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1895 const void *from,
1896 const unsigned int len)
1898 memcpy(skb->data, from, len);
1901 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1902 const int offset,
1903 const void *from,
1904 const unsigned int len)
1906 memcpy(skb->data + offset, from, len);
1909 extern void skb_init(void);
1911 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1913 return skb->tstamp;
1917 * skb_get_timestamp - get timestamp from a skb
1918 * @skb: skb to get stamp from
1919 * @stamp: pointer to struct timeval to store stamp in
1921 * Timestamps are stored in the skb as offsets to a base timestamp.
1922 * This function converts the offset back to a struct timeval and stores
1923 * it in stamp.
1925 static inline void skb_get_timestamp(const struct sk_buff *skb,
1926 struct timeval *stamp)
1928 *stamp = ktime_to_timeval(skb->tstamp);
1931 static inline void skb_get_timestampns(const struct sk_buff *skb,
1932 struct timespec *stamp)
1934 *stamp = ktime_to_timespec(skb->tstamp);
1937 static inline void __net_timestamp(struct sk_buff *skb)
1939 skb->tstamp = ktime_get_real();
1942 static inline ktime_t net_timedelta(ktime_t t)
1944 return ktime_sub(ktime_get_real(), t);
1947 static inline ktime_t net_invalid_timestamp(void)
1949 return ktime_set(0, 0);
1952 extern void skb_timestamping_init(void);
1954 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1956 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1957 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1959 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1961 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1965 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1967 return false;
1970 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1973 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1975 * @skb: clone of the the original outgoing packet
1976 * @hwtstamps: hardware time stamps
1979 void skb_complete_tx_timestamp(struct sk_buff *skb,
1980 struct skb_shared_hwtstamps *hwtstamps);
1983 * skb_tstamp_tx - queue clone of skb with send time stamps
1984 * @orig_skb: the original outgoing packet
1985 * @hwtstamps: hardware time stamps, may be NULL if not available
1987 * If the skb has a socket associated, then this function clones the
1988 * skb (thus sharing the actual data and optional structures), stores
1989 * the optional hardware time stamping information (if non NULL) or
1990 * generates a software time stamp (otherwise), then queues the clone
1991 * to the error queue of the socket. Errors are silently ignored.
1993 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1994 struct skb_shared_hwtstamps *hwtstamps);
1996 static inline void sw_tx_timestamp(struct sk_buff *skb)
1998 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
1999 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2000 skb_tstamp_tx(skb, NULL);
2004 * skb_tx_timestamp() - Driver hook for transmit timestamping
2006 * Ethernet MAC Drivers should call this function in their hard_xmit()
2007 * function as soon as possible after giving the sk_buff to the MAC
2008 * hardware, but before freeing the sk_buff.
2010 * @skb: A socket buffer.
2012 static inline void skb_tx_timestamp(struct sk_buff *skb)
2014 skb_clone_tx_timestamp(skb);
2015 sw_tx_timestamp(skb);
2018 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2019 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2021 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2023 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2027 * skb_checksum_complete - Calculate checksum of an entire packet
2028 * @skb: packet to process
2030 * This function calculates the checksum over the entire packet plus
2031 * the value of skb->csum. The latter can be used to supply the
2032 * checksum of a pseudo header as used by TCP/UDP. It returns the
2033 * checksum.
2035 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2036 * this function can be used to verify that checksum on received
2037 * packets. In that case the function should return zero if the
2038 * checksum is correct. In particular, this function will return zero
2039 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2040 * hardware has already verified the correctness of the checksum.
2042 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2044 return skb_csum_unnecessary(skb) ?
2045 0 : __skb_checksum_complete(skb);
2048 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2049 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2050 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2052 if (nfct && atomic_dec_and_test(&nfct->use))
2053 nf_conntrack_destroy(nfct);
2055 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2057 if (nfct)
2058 atomic_inc(&nfct->use);
2060 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2062 if (skb)
2063 atomic_inc(&skb->users);
2065 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2067 if (skb)
2068 kfree_skb(skb);
2070 #endif
2071 #ifdef CONFIG_BRIDGE_NETFILTER
2072 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2074 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2075 kfree(nf_bridge);
2077 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2079 if (nf_bridge)
2080 atomic_inc(&nf_bridge->use);
2082 #endif /* CONFIG_BRIDGE_NETFILTER */
2083 static inline void nf_reset(struct sk_buff *skb)
2085 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2086 nf_conntrack_put(skb->nfct);
2087 skb->nfct = NULL;
2088 nf_conntrack_put_reasm(skb->nfct_reasm);
2089 skb->nfct_reasm = NULL;
2090 #endif
2091 #ifdef CONFIG_BRIDGE_NETFILTER
2092 nf_bridge_put(skb->nf_bridge);
2093 skb->nf_bridge = NULL;
2094 #endif
2097 /* Note: This doesn't put any conntrack and bridge info in dst. */
2098 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2100 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2101 dst->nfct = src->nfct;
2102 nf_conntrack_get(src->nfct);
2103 dst->nfctinfo = src->nfctinfo;
2104 dst->nfct_reasm = src->nfct_reasm;
2105 nf_conntrack_get_reasm(src->nfct_reasm);
2106 #endif
2107 #ifdef CONFIG_BRIDGE_NETFILTER
2108 dst->nf_bridge = src->nf_bridge;
2109 nf_bridge_get(src->nf_bridge);
2110 #endif
2113 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2115 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2116 nf_conntrack_put(dst->nfct);
2117 nf_conntrack_put_reasm(dst->nfct_reasm);
2118 #endif
2119 #ifdef CONFIG_BRIDGE_NETFILTER
2120 nf_bridge_put(dst->nf_bridge);
2121 #endif
2122 __nf_copy(dst, src);
2125 #ifdef CONFIG_NETWORK_SECMARK
2126 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2128 to->secmark = from->secmark;
2131 static inline void skb_init_secmark(struct sk_buff *skb)
2133 skb->secmark = 0;
2135 #else
2136 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2139 static inline void skb_init_secmark(struct sk_buff *skb)
2141 #endif
2143 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2145 skb->queue_mapping = queue_mapping;
2148 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2150 return skb->queue_mapping;
2153 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2155 to->queue_mapping = from->queue_mapping;
2158 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2160 skb->queue_mapping = rx_queue + 1;
2163 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2165 return skb->queue_mapping - 1;
2168 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2170 return skb->queue_mapping != 0;
2173 extern u16 __skb_tx_hash(const struct net_device *dev,
2174 const struct sk_buff *skb,
2175 unsigned int num_tx_queues);
2177 #ifdef CONFIG_XFRM
2178 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2180 return skb->sp;
2182 #else
2183 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2185 return NULL;
2187 #endif
2189 static inline int skb_is_gso(const struct sk_buff *skb)
2191 return skb_shinfo(skb)->gso_size;
2194 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2196 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2199 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2201 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2203 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2204 * wanted then gso_type will be set. */
2205 struct skb_shared_info *shinfo = skb_shinfo(skb);
2206 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2207 unlikely(shinfo->gso_type == 0)) {
2208 __skb_warn_lro_forwarding(skb);
2209 return true;
2211 return false;
2214 static inline void skb_forward_csum(struct sk_buff *skb)
2216 /* Unfortunately we don't support this one. Any brave souls? */
2217 if (skb->ip_summed == CHECKSUM_COMPLETE)
2218 skb->ip_summed = CHECKSUM_NONE;
2222 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2223 * @skb: skb to check
2225 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2226 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2227 * use this helper, to document places where we make this assertion.
2229 static inline void skb_checksum_none_assert(struct sk_buff *skb)
2231 #ifdef DEBUG
2232 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2233 #endif
2236 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2237 #endif /* __KERNEL__ */
2238 #endif /* _LINUX_SKBUFF_H */