2 * Definitions for the 'struct sk_buff' memory handlers.
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,
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
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
96 struct pipe_inode_info;
98 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
104 #ifdef CONFIG_BRIDGE_NETFILTER
105 struct nf_bridge_info {
107 struct net_device *physindev;
108 struct net_device *physoutdev;
110 unsigned long data[32 / sizeof(unsigned long)];
114 struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
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 {
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
155 * hwtstamps can only be compared against other hwtstamps from
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 {
166 /* Definitions for tx_flags in struct skb_shared_info */
168 /* generate hardware time stamp */
169 SKBTX_HW_TSTAMP = 1 << 0,
171 /* generate software time stamp */
172 SKBTX_SW_TSTAMP = 1 << 1,
174 /* device driver is going to provide hardware time stamp */
175 SKBTX_IN_PROGRESS = 1 << 2,
177 /* ensure the originating sk reference is available on driver level */
178 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
181 /* This data is invariant across clones and lives at
182 * the end of the header data, ie. at skb->end.
184 struct skb_shared_info {
185 unsigned short nr_frags;
186 unsigned short gso_size;
187 /* Warning: this field is not always filled in (UFO)! */
188 unsigned short gso_segs;
189 unsigned short gso_type;
192 struct sk_buff *frag_list;
193 struct skb_shared_hwtstamps hwtstamps;
196 * Warning : all fields before dataref are cleared in __alloc_skb()
200 /* Intermediate layers must ensure that destructor_arg
201 * remains valid until skb destructor */
202 void * destructor_arg;
203 /* must be last field, see pskb_expand_head() */
204 skb_frag_t frags[MAX_SKB_FRAGS];
207 /* We divide dataref into two halves. The higher 16 bits hold references
208 * to the payload part of skb->data. The lower 16 bits hold references to
209 * the entire skb->data. A clone of a headerless skb holds the length of
210 * the header in skb->hdr_len.
212 * All users must obey the rule that the skb->data reference count must be
213 * greater than or equal to the payload reference count.
215 * Holding a reference to the payload part means that the user does not
216 * care about modifications to the header part of skb->data.
218 #define SKB_DATAREF_SHIFT 16
219 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
223 SKB_FCLONE_UNAVAILABLE,
229 SKB_GSO_TCPV4 = 1 << 0,
230 SKB_GSO_UDP = 1 << 1,
232 /* This indicates the skb is from an untrusted source. */
233 SKB_GSO_DODGY = 1 << 2,
235 /* This indicates the tcp segment has CWR set. */
236 SKB_GSO_TCP_ECN = 1 << 3,
238 SKB_GSO_TCPV6 = 1 << 4,
240 SKB_GSO_FCOE = 1 << 5,
243 #if BITS_PER_LONG > 32
244 #define NET_SKBUFF_DATA_USES_OFFSET 1
247 #ifdef NET_SKBUFF_DATA_USES_OFFSET
248 typedef unsigned int sk_buff_data_t;
250 typedef unsigned char *sk_buff_data_t;
254 * struct sk_buff - socket buffer
255 * @next: Next buffer in list
256 * @prev: Previous buffer in list
257 * @sk: Socket we are owned by
258 * @tstamp: Time we arrived
259 * @dev: Device we arrived on/are leaving by
260 * @transport_header: Transport layer header
261 * @network_header: Network layer header
262 * @mac_header: Link layer header
263 * @_skb_refdst: destination entry (with norefcount bit)
264 * @sp: the security path, used for xfrm
265 * @cb: Control buffer. Free for use by every layer. Put private vars here
266 * @len: Length of actual data
267 * @data_len: Data length
268 * @mac_len: Length of link layer header
269 * @hdr_len: writable header length of cloned skb
270 * @csum: Checksum (must include start/offset pair)
271 * @csum_start: Offset from skb->head where checksumming should start
272 * @csum_offset: Offset from csum_start where checksum should be stored
273 * @local_df: allow local fragmentation
274 * @cloned: Head may be cloned (check refcnt to be sure)
275 * @nohdr: Payload reference only, must not modify header
276 * @pkt_type: Packet class
277 * @fclone: skbuff clone status
278 * @ip_summed: Driver fed us an IP checksum
279 * @priority: Packet queueing priority
280 * @users: User count - see {datagram,tcp}.c
281 * @protocol: Packet protocol from driver
282 * @truesize: Buffer size
283 * @head: Head of buffer
284 * @data: Data head pointer
285 * @tail: Tail pointer
287 * @destructor: Destruct function
288 * @mark: Generic packet mark
289 * @nfct: Associated connection, if any
290 * @ipvs_property: skbuff is owned by ipvs
291 * @peeked: this packet has been seen already, so stats have been
292 * done for it, don't do them again
293 * @nf_trace: netfilter packet trace flag
294 * @nfctinfo: Relationship of this skb to the connection
295 * @nfct_reasm: netfilter conntrack re-assembly pointer
296 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
297 * @skb_iif: ifindex of device we arrived on
298 * @rxhash: the packet hash computed on receive
299 * @queue_mapping: Queue mapping for multiqueue devices
300 * @tc_index: Traffic control index
301 * @tc_verd: traffic control verdict
302 * @ndisc_nodetype: router type (from link layer)
303 * @dma_cookie: a cookie to one of several possible DMA operations
304 * done by skb DMA functions
305 * @secmark: security marking
306 * @vlan_tci: vlan tag control information
310 /* These two members must be first. */
311 struct sk_buff *next;
312 struct sk_buff *prev;
317 struct net_device *dev;
320 * This is the control buffer. It is free to use for every
321 * layer. Please put your private variables there. If you
322 * want to keep them across layers you have to do a skb_clone()
323 * first. This is owned by whoever has the skb queued ATM.
325 char cb[48] __aligned(8);
327 unsigned long _skb_refdst;
343 kmemcheck_bitfield_begin(flags1);
354 kmemcheck_bitfield_end(flags1);
357 void (*destructor)(struct sk_buff *skb);
358 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
359 struct nf_conntrack *nfct;
360 struct sk_buff *nfct_reasm;
362 #ifdef CONFIG_BRIDGE_NETFILTER
363 struct nf_bridge_info *nf_bridge;
367 #ifdef CONFIG_NET_SCHED
368 __u16 tc_index; /* traffic control index */
369 #ifdef CONFIG_NET_CLS_ACT
370 __u16 tc_verd; /* traffic control verdict */
376 kmemcheck_bitfield_begin(flags2);
377 __u16 queue_mapping:16;
378 #ifdef CONFIG_IPV6_NDISC_NODETYPE
379 __u8 ndisc_nodetype:2,
382 __u8 deliver_no_wcard:1;
384 kmemcheck_bitfield_end(flags2);
388 #ifdef CONFIG_NET_DMA
389 dma_cookie_t dma_cookie;
391 #ifdef CONFIG_NETWORK_SECMARK
401 sk_buff_data_t transport_header;
402 sk_buff_data_t network_header;
403 sk_buff_data_t mac_header;
404 /* These elements must be at the end, see alloc_skb() for details. */
409 unsigned int truesize;
415 * Handling routines are only of interest to the kernel
417 #include <linux/slab.h>
419 #include <asm/system.h>
422 * skb might have a dst pointer attached, refcounted or not.
423 * _skb_refdst low order bit is set if refcount was _not_ taken
425 #define SKB_DST_NOREF 1UL
426 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
429 * skb_dst - returns skb dst_entry
432 * Returns skb dst_entry, regardless of reference taken or not.
434 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
436 /* If refdst was not refcounted, check we still are in a
437 * rcu_read_lock section
439 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
440 !rcu_read_lock_held() &&
441 !rcu_read_lock_bh_held());
442 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
446 * skb_dst_set - sets skb dst
450 * Sets skb dst, assuming a reference was taken on dst and should
451 * be released by skb_dst_drop()
453 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
455 skb->_skb_refdst = (unsigned long)dst;
459 * skb_dst_set_noref - sets skb dst, without a reference
463 * Sets skb dst, assuming a reference was not taken on dst
464 * skb_dst_drop() should not dst_release() this dst
466 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
468 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
469 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
473 * skb_dst_is_noref - Test if skb dst isnt refcounted
476 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
478 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
481 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
483 return (struct rtable *)skb_dst(skb);
486 extern void kfree_skb(struct sk_buff *skb);
487 extern void consume_skb(struct sk_buff *skb);
488 extern void __kfree_skb(struct sk_buff *skb);
489 extern struct sk_buff *__alloc_skb(unsigned int size,
490 gfp_t priority, int fclone, int node);
491 static inline struct sk_buff *alloc_skb(unsigned int size,
494 return __alloc_skb(size, priority, 0, -1);
497 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
500 return __alloc_skb(size, priority, 1, -1);
503 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
505 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
506 extern struct sk_buff *skb_clone(struct sk_buff *skb,
508 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
510 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
512 extern int pskb_expand_head(struct sk_buff *skb,
513 int nhead, int ntail,
515 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
516 unsigned int headroom);
517 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
518 int newheadroom, int newtailroom,
520 extern int skb_to_sgvec(struct sk_buff *skb,
521 struct scatterlist *sg, int offset,
523 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
524 struct sk_buff **trailer);
525 extern int skb_pad(struct sk_buff *skb, int pad);
526 #define dev_kfree_skb(a) consume_skb(a)
528 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
529 int getfrag(void *from, char *to, int offset,
530 int len,int odd, struct sk_buff *skb),
531 void *from, int length);
533 struct skb_seq_state {
537 __u32 stepped_offset;
538 struct sk_buff *root_skb;
539 struct sk_buff *cur_skb;
543 extern void skb_prepare_seq_read(struct sk_buff *skb,
544 unsigned int from, unsigned int to,
545 struct skb_seq_state *st);
546 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
547 struct skb_seq_state *st);
548 extern void skb_abort_seq_read(struct skb_seq_state *st);
550 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
551 unsigned int to, struct ts_config *config,
552 struct ts_state *state);
554 extern __u32 __skb_get_rxhash(struct sk_buff *skb);
555 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
558 skb->rxhash = __skb_get_rxhash(skb);
563 #ifdef NET_SKBUFF_DATA_USES_OFFSET
564 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
566 return skb->head + skb->end;
569 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
576 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
578 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
580 return &skb_shinfo(skb)->hwtstamps;
584 * skb_queue_empty - check if a queue is empty
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
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
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
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));
639 * skb_queue_prev - return the prev packet in the queue
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));
657 * skb_get - reference buffer
658 * @skb: buffer to reference
660 * Makes another reference to a socket buffer and returns a pointer
663 static inline struct sk_buff *skb_get(struct sk_buff *skb)
665 atomic_inc(&skb->users);
670 * If users == 1, we are the only owner and are can avoid redundant
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)
702 dataref = atomic_read(&skb_shinfo(skb)->dataref);
703 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
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)
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
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
745 * NULL is returned on a memory allocation failure.
747 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
750 might_sleep_if(pri & __GFP_WAIT);
751 if (skb_shared(skb)) {
752 struct sk_buff *nskb = skb_clone(skb, pri);
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,
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 */
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_)
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_)
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_)
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;
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)
894 next->prev = prev->next = newsk;
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;
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
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
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;
1049 skb->next = skb->prev = NULL;
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);
1067 __skb_unlink(skb, list);
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);
1084 __skb_unlink(skb, list);
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)
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];
1114 frag->page_offset = off;
1116 skb_shinfo(skb)->nr_frags = i + 1;
1119 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1122 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1123 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(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)
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);
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)
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)
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)))
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)))
1214 if (unlikely(len > skb->len))
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)
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,
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,
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;
1379 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1381 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1385 * CPUs often take a performance hit when accessing unaligned memory
1386 * locations. The actual performance hit varies, it can be small if the
1387 * hardware handles it or large if we have to take an exception and fix it
1390 * Since an ethernet header is 14 bytes network drivers often end up with
1391 * the IP header at an unaligned offset. The IP header can be aligned by
1392 * shifting the start of the packet by 2 bytes. Drivers should do this
1395 * skb_reserve(skb, NET_IP_ALIGN);
1397 * The downside to this alignment of the IP header is that the DMA is now
1398 * unaligned. On some architectures the cost of an unaligned DMA is high
1399 * and this cost outweighs the gains made by aligning the IP header.
1401 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1404 #ifndef NET_IP_ALIGN
1405 #define NET_IP_ALIGN 2
1409 * The networking layer reserves some headroom in skb data (via
1410 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1411 * the header has to grow. In the default case, if the header has to grow
1412 * 32 bytes or less we avoid the reallocation.
1414 * Unfortunately this headroom changes the DMA alignment of the resulting
1415 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1416 * on some architectures. An architecture can override this value,
1417 * perhaps setting it to a cacheline in size (since that will maintain
1418 * cacheline alignment of the DMA). It must be a power of 2.
1420 * Various parts of the networking layer expect at least 32 bytes of
1421 * headroom, you should not reduce this.
1423 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1424 * to reduce average number of cache lines per packet.
1425 * get_rps_cpus() for example only access one 64 bytes aligned block :
1426 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1429 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1432 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1434 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1436 if (unlikely(skb->data_len)) {
1441 skb_set_tail_pointer(skb, len);
1444 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1446 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1449 return ___pskb_trim(skb, len);
1450 __skb_trim(skb, len);
1454 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1456 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1460 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1461 * @skb: buffer to alter
1464 * This is identical to pskb_trim except that the caller knows that
1465 * the skb is not cloned so we should never get an error due to out-
1468 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1470 int err = pskb_trim(skb, len);
1475 * skb_orphan - orphan a buffer
1476 * @skb: buffer to orphan
1478 * If a buffer currently has an owner then we call the owner's
1479 * destructor function and make the @skb unowned. The buffer continues
1480 * to exist but is no longer charged to its former owner.
1482 static inline void skb_orphan(struct sk_buff *skb)
1484 if (skb->destructor)
1485 skb->destructor(skb);
1486 skb->destructor = NULL;
1491 * __skb_queue_purge - empty a list
1492 * @list: list to empty
1494 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1495 * the list and one reference dropped. This function does not take the
1496 * list lock and the caller must hold the relevant locks to use it.
1498 extern void skb_queue_purge(struct sk_buff_head *list);
1499 static inline void __skb_queue_purge(struct sk_buff_head *list)
1501 struct sk_buff *skb;
1502 while ((skb = __skb_dequeue(list)) != NULL)
1507 * __dev_alloc_skb - allocate an skbuff for receiving
1508 * @length: length to allocate
1509 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1511 * Allocate a new &sk_buff and assign it a usage count of one. The
1512 * buffer has unspecified headroom built in. Users should allocate
1513 * the headroom they think they need without accounting for the
1514 * built in space. The built in space is used for optimisations.
1516 * %NULL is returned if there is no free memory.
1518 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1521 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1523 skb_reserve(skb, NET_SKB_PAD);
1527 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1529 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1530 unsigned int length, gfp_t gfp_mask);
1533 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1534 * @dev: network device to receive on
1535 * @length: length to allocate
1537 * Allocate a new &sk_buff and assign it a usage count of one. The
1538 * buffer has unspecified headroom built in. Users should allocate
1539 * the headroom they think they need without accounting for the
1540 * built in space. The built in space is used for optimisations.
1542 * %NULL is returned if there is no free memory. Although this function
1543 * allocates memory it can be called from an interrupt.
1545 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1546 unsigned int length)
1548 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1551 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1552 unsigned int length)
1554 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1556 if (NET_IP_ALIGN && skb)
1557 skb_reserve(skb, NET_IP_ALIGN);
1561 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1564 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1565 * @dev: network device to receive on
1567 * Allocate a new page node local to the specified device.
1569 * %NULL is returned if there is no free memory.
1571 static inline struct page *netdev_alloc_page(struct net_device *dev)
1573 return __netdev_alloc_page(dev, GFP_ATOMIC);
1576 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1582 * skb_clone_writable - is the header of a clone writable
1583 * @skb: buffer to check
1584 * @len: length up to which to write
1586 * Returns true if modifying the header part of the cloned buffer
1587 * does not requires the data to be copied.
1589 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1591 return !skb_header_cloned(skb) &&
1592 skb_headroom(skb) + len <= skb->hdr_len;
1595 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1600 if (headroom < NET_SKB_PAD)
1601 headroom = NET_SKB_PAD;
1602 if (headroom > skb_headroom(skb))
1603 delta = headroom - skb_headroom(skb);
1605 if (delta || cloned)
1606 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1612 * skb_cow - copy header of skb when it is required
1613 * @skb: buffer to cow
1614 * @headroom: needed headroom
1616 * If the skb passed lacks sufficient headroom or its data part
1617 * is shared, data is reallocated. If reallocation fails, an error
1618 * is returned and original skb is not changed.
1620 * The result is skb with writable area skb->head...skb->tail
1621 * and at least @headroom of space at head.
1623 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1625 return __skb_cow(skb, headroom, skb_cloned(skb));
1629 * skb_cow_head - skb_cow but only making the head writable
1630 * @skb: buffer to cow
1631 * @headroom: needed headroom
1633 * This function is identical to skb_cow except that we replace the
1634 * skb_cloned check by skb_header_cloned. It should be used when
1635 * you only need to push on some header and do not need to modify
1638 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1640 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1644 * skb_padto - pad an skbuff up to a minimal size
1645 * @skb: buffer to pad
1646 * @len: minimal length
1648 * Pads up a buffer to ensure the trailing bytes exist and are
1649 * blanked. If the buffer already contains sufficient data it
1650 * is untouched. Otherwise it is extended. Returns zero on
1651 * success. The skb is freed on error.
1654 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1656 unsigned int size = skb->len;
1657 if (likely(size >= len))
1659 return skb_pad(skb, len - size);
1662 static inline int skb_add_data(struct sk_buff *skb,
1663 char __user *from, int copy)
1665 const int off = skb->len;
1667 if (skb->ip_summed == CHECKSUM_NONE) {
1669 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1672 skb->csum = csum_block_add(skb->csum, csum, off);
1675 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1678 __skb_trim(skb, off);
1682 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1683 struct page *page, int off)
1686 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1688 return page == frag->page &&
1689 off == frag->page_offset + frag->size;
1694 static inline int __skb_linearize(struct sk_buff *skb)
1696 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1700 * skb_linearize - convert paged skb to linear one
1701 * @skb: buffer to linarize
1703 * If there is no free memory -ENOMEM is returned, otherwise zero
1704 * is returned and the old skb data released.
1706 static inline int skb_linearize(struct sk_buff *skb)
1708 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1712 * skb_linearize_cow - make sure skb is linear and writable
1713 * @skb: buffer to process
1715 * If there is no free memory -ENOMEM is returned, otherwise zero
1716 * is returned and the old skb data released.
1718 static inline int skb_linearize_cow(struct sk_buff *skb)
1720 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1721 __skb_linearize(skb) : 0;
1725 * skb_postpull_rcsum - update checksum for received skb after pull
1726 * @skb: buffer to update
1727 * @start: start of data before pull
1728 * @len: length of data pulled
1730 * After doing a pull on a received packet, you need to call this to
1731 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1732 * CHECKSUM_NONE so that it can be recomputed from scratch.
1735 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1736 const void *start, unsigned int len)
1738 if (skb->ip_summed == CHECKSUM_COMPLETE)
1739 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1742 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1745 * pskb_trim_rcsum - trim received skb and update checksum
1746 * @skb: buffer to trim
1749 * This is exactly the same as pskb_trim except that it ensures the
1750 * checksum of received packets are still valid after the operation.
1753 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1755 if (likely(len >= skb->len))
1757 if (skb->ip_summed == CHECKSUM_COMPLETE)
1758 skb->ip_summed = CHECKSUM_NONE;
1759 return __pskb_trim(skb, len);
1762 #define skb_queue_walk(queue, skb) \
1763 for (skb = (queue)->next; \
1764 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1767 #define skb_queue_walk_safe(queue, skb, tmp) \
1768 for (skb = (queue)->next, tmp = skb->next; \
1769 skb != (struct sk_buff *)(queue); \
1770 skb = tmp, tmp = skb->next)
1772 #define skb_queue_walk_from(queue, skb) \
1773 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1776 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1777 for (tmp = skb->next; \
1778 skb != (struct sk_buff *)(queue); \
1779 skb = tmp, tmp = skb->next)
1781 #define skb_queue_reverse_walk(queue, skb) \
1782 for (skb = (queue)->prev; \
1783 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1787 static inline bool skb_has_frag_list(const struct sk_buff *skb)
1789 return skb_shinfo(skb)->frag_list != NULL;
1792 static inline void skb_frag_list_init(struct sk_buff *skb)
1794 skb_shinfo(skb)->frag_list = NULL;
1797 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1799 frag->next = skb_shinfo(skb)->frag_list;
1800 skb_shinfo(skb)->frag_list = frag;
1803 #define skb_walk_frags(skb, iter) \
1804 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1806 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1807 int *peeked, int *err);
1808 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1809 int noblock, int *err);
1810 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1811 struct poll_table_struct *wait);
1812 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1813 int offset, struct iovec *to,
1815 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1818 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1820 const struct iovec *from,
1823 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1825 const struct iovec *to,
1828 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1829 extern void skb_free_datagram_locked(struct sock *sk,
1830 struct sk_buff *skb);
1831 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1832 unsigned int flags);
1833 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1834 int len, __wsum csum);
1835 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1837 extern int skb_store_bits(struct sk_buff *skb, int offset,
1838 const void *from, int len);
1839 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1840 int offset, u8 *to, int len,
1842 extern int skb_splice_bits(struct sk_buff *skb,
1843 unsigned int offset,
1844 struct pipe_inode_info *pipe,
1846 unsigned int flags);
1847 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1848 extern void skb_split(struct sk_buff *skb,
1849 struct sk_buff *skb1, const u32 len);
1850 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1853 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1855 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1856 int len, void *buffer)
1858 int hlen = skb_headlen(skb);
1860 if (hlen - offset >= len)
1861 return skb->data + offset;
1863 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1869 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1871 const unsigned int len)
1873 memcpy(to, skb->data, len);
1876 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1877 const int offset, void *to,
1878 const unsigned int len)
1880 memcpy(to, skb->data + offset, len);
1883 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1885 const unsigned int len)
1887 memcpy(skb->data, from, len);
1890 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1893 const unsigned int len)
1895 memcpy(skb->data + offset, from, len);
1898 extern void skb_init(void);
1900 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1906 * skb_get_timestamp - get timestamp from a skb
1907 * @skb: skb to get stamp from
1908 * @stamp: pointer to struct timeval to store stamp in
1910 * Timestamps are stored in the skb as offsets to a base timestamp.
1911 * This function converts the offset back to a struct timeval and stores
1914 static inline void skb_get_timestamp(const struct sk_buff *skb,
1915 struct timeval *stamp)
1917 *stamp = ktime_to_timeval(skb->tstamp);
1920 static inline void skb_get_timestampns(const struct sk_buff *skb,
1921 struct timespec *stamp)
1923 *stamp = ktime_to_timespec(skb->tstamp);
1926 static inline void __net_timestamp(struct sk_buff *skb)
1928 skb->tstamp = ktime_get_real();
1931 static inline ktime_t net_timedelta(ktime_t t)
1933 return ktime_sub(ktime_get_real(), t);
1936 static inline ktime_t net_invalid_timestamp(void)
1938 return ktime_set(0, 0);
1941 extern void skb_timestamping_init(void);
1943 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1945 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1946 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1948 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1950 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1954 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1959 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1962 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1964 * @skb: clone of the the original outgoing packet
1965 * @hwtstamps: hardware time stamps
1968 void skb_complete_tx_timestamp(struct sk_buff *skb,
1969 struct skb_shared_hwtstamps *hwtstamps);
1972 * skb_tstamp_tx - queue clone of skb with send time stamps
1973 * @orig_skb: the original outgoing packet
1974 * @hwtstamps: hardware time stamps, may be NULL if not available
1976 * If the skb has a socket associated, then this function clones the
1977 * skb (thus sharing the actual data and optional structures), stores
1978 * the optional hardware time stamping information (if non NULL) or
1979 * generates a software time stamp (otherwise), then queues the clone
1980 * to the error queue of the socket. Errors are silently ignored.
1982 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1983 struct skb_shared_hwtstamps *hwtstamps);
1985 static inline void sw_tx_timestamp(struct sk_buff *skb)
1987 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
1988 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
1989 skb_tstamp_tx(skb, NULL);
1993 * skb_tx_timestamp() - Driver hook for transmit timestamping
1995 * Ethernet MAC Drivers should call this function in their hard_xmit()
1996 * function as soon as possible after giving the sk_buff to the MAC
1997 * hardware, but before freeing the sk_buff.
1999 * @skb: A socket buffer.
2001 static inline void skb_tx_timestamp(struct sk_buff *skb)
2003 skb_clone_tx_timestamp(skb);
2004 sw_tx_timestamp(skb);
2007 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2008 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2010 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2012 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2016 * skb_checksum_complete - Calculate checksum of an entire packet
2017 * @skb: packet to process
2019 * This function calculates the checksum over the entire packet plus
2020 * the value of skb->csum. The latter can be used to supply the
2021 * checksum of a pseudo header as used by TCP/UDP. It returns the
2024 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2025 * this function can be used to verify that checksum on received
2026 * packets. In that case the function should return zero if the
2027 * checksum is correct. In particular, this function will return zero
2028 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2029 * hardware has already verified the correctness of the checksum.
2031 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2033 return skb_csum_unnecessary(skb) ?
2034 0 : __skb_checksum_complete(skb);
2037 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2038 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2039 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2041 if (nfct && atomic_dec_and_test(&nfct->use))
2042 nf_conntrack_destroy(nfct);
2044 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2047 atomic_inc(&nfct->use);
2049 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2052 atomic_inc(&skb->users);
2054 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2060 #ifdef CONFIG_BRIDGE_NETFILTER
2061 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2063 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2066 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2069 atomic_inc(&nf_bridge->use);
2071 #endif /* CONFIG_BRIDGE_NETFILTER */
2072 static inline void nf_reset(struct sk_buff *skb)
2074 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2075 nf_conntrack_put(skb->nfct);
2077 nf_conntrack_put_reasm(skb->nfct_reasm);
2078 skb->nfct_reasm = NULL;
2080 #ifdef CONFIG_BRIDGE_NETFILTER
2081 nf_bridge_put(skb->nf_bridge);
2082 skb->nf_bridge = NULL;
2086 /* Note: This doesn't put any conntrack and bridge info in dst. */
2087 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2089 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2090 dst->nfct = src->nfct;
2091 nf_conntrack_get(src->nfct);
2092 dst->nfctinfo = src->nfctinfo;
2093 dst->nfct_reasm = src->nfct_reasm;
2094 nf_conntrack_get_reasm(src->nfct_reasm);
2096 #ifdef CONFIG_BRIDGE_NETFILTER
2097 dst->nf_bridge = src->nf_bridge;
2098 nf_bridge_get(src->nf_bridge);
2102 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2104 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2105 nf_conntrack_put(dst->nfct);
2106 nf_conntrack_put_reasm(dst->nfct_reasm);
2108 #ifdef CONFIG_BRIDGE_NETFILTER
2109 nf_bridge_put(dst->nf_bridge);
2111 __nf_copy(dst, src);
2114 #ifdef CONFIG_NETWORK_SECMARK
2115 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2117 to->secmark = from->secmark;
2120 static inline void skb_init_secmark(struct sk_buff *skb)
2125 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2128 static inline void skb_init_secmark(struct sk_buff *skb)
2132 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2134 skb->queue_mapping = queue_mapping;
2137 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2139 return skb->queue_mapping;
2142 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2144 to->queue_mapping = from->queue_mapping;
2147 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2149 skb->queue_mapping = rx_queue + 1;
2152 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2154 return skb->queue_mapping - 1;
2157 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2159 return (skb->queue_mapping != 0);
2162 extern u16 skb_tx_hash(const struct net_device *dev,
2163 const struct sk_buff *skb);
2166 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2171 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2177 static inline int skb_is_gso(const struct sk_buff *skb)
2179 return skb_shinfo(skb)->gso_size;
2182 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2184 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2187 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2189 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2191 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2192 * wanted then gso_type will be set. */
2193 struct skb_shared_info *shinfo = skb_shinfo(skb);
2194 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2195 unlikely(shinfo->gso_type == 0)) {
2196 __skb_warn_lro_forwarding(skb);
2202 static inline void skb_forward_csum(struct sk_buff *skb)
2204 /* Unfortunately we don't support this one. Any brave souls? */
2205 if (skb->ip_summed == CHECKSUM_COMPLETE)
2206 skb->ip_summed = CHECKSUM_NONE;
2210 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2211 * @skb: skb to check
2213 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2214 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2215 * use this helper, to document places where we make this assertion.
2217 static inline void skb_checksum_none_assert(struct sk_buff *skb)
2220 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2224 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2225 #endif /* __KERNEL__ */
2226 #endif /* _LINUX_SKBUFF_H */