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 <linux/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>
32 #include <linux/dma-mapping.h>
34 /* Don't change this without changing skb_csum_unnecessary! */
35 #define CHECKSUM_NONE 0
36 #define CHECKSUM_UNNECESSARY 1
37 #define CHECKSUM_COMPLETE 2
38 #define CHECKSUM_PARTIAL 3
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_WITH_OVERHEAD(X) \
43 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
44 #define SKB_MAX_ORDER(X, ORDER) \
45 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
46 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
47 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
49 /* A. Checksumming of received packets by device.
51 * NONE: device failed to checksum this packet.
52 * skb->csum is undefined.
54 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
55 * skb->csum is undefined.
56 * It is bad option, but, unfortunately, many of vendors do this.
57 * Apparently with secret goal to sell you new device, when you
58 * will add new protocol to your host. F.e. IPv6. 8)
60 * COMPLETE: the most generic way. Device supplied checksum of _all_
61 * the packet as seen by netif_rx in skb->csum.
62 * NOTE: Even if device supports only some protocols, but
63 * is able to produce some skb->csum, it MUST use COMPLETE,
66 * PARTIAL: identical to the case for output below. This may occur
67 * on a packet received directly from another Linux OS, e.g.,
68 * a virtualised Linux kernel on the same host. The packet can
69 * be treated in the same way as UNNECESSARY except that on
70 * output (i.e., forwarding) the checksum must be filled in
71 * by the OS or the hardware.
73 * B. Checksumming on output.
75 * NONE: skb is checksummed by protocol or csum is not required.
77 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
78 * from skb->csum_start to the end and to record the checksum
79 * at skb->csum_start + skb->csum_offset.
81 * Device must show its capabilities in dev->features, set
82 * at device setup time.
83 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
85 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
86 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
87 * TCP/UDP over IPv4. Sigh. Vendors like this
88 * way by an unknown reason. Though, see comment above
89 * about CHECKSUM_UNNECESSARY. 8)
90 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
92 * Any questions? No questions, good. --ANK
97 struct pipe_inode_info;
99 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
100 struct nf_conntrack {
105 #ifdef CONFIG_BRIDGE_NETFILTER
106 struct nf_bridge_info {
108 struct net_device *physindev;
109 struct net_device *physoutdev;
111 unsigned long data[32 / sizeof(unsigned long)];
115 struct sk_buff_head {
116 /* These two members must be first. */
117 struct sk_buff *next;
118 struct sk_buff *prev;
126 /* To allow 64K frame to be packed as single skb without frag_list. Since
127 * GRO uses frags we allocate at least 16 regardless of page size.
129 #if (65536/PAGE_SIZE + 2) < 16
130 #define MAX_SKB_FRAGS 16UL
132 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
135 typedef struct skb_frag_struct skb_frag_t;
137 struct skb_frag_struct {
139 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
148 #define HAVE_HW_TIME_STAMP
151 * struct skb_shared_hwtstamps - hardware time stamps
152 * @hwtstamp: hardware time stamp transformed into duration
153 * since arbitrary point in time
154 * @syststamp: hwtstamp transformed to system time base
156 * Software time stamps generated by ktime_get_real() are stored in
157 * skb->tstamp. The relation between the different kinds of time
158 * stamps is as follows:
160 * syststamp and tstamp can be compared against each other in
161 * arbitrary combinations. The accuracy of a
162 * syststamp/tstamp/"syststamp from other device" comparison is
163 * limited by the accuracy of the transformation into system time
164 * base. This depends on the device driver and its underlying
167 * hwtstamps can only be compared against other hwtstamps from
170 * This structure is attached to packets as part of the
171 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
173 struct skb_shared_hwtstamps {
178 /* Definitions for tx_flags in struct skb_shared_info */
180 /* generate hardware time stamp */
181 SKBTX_HW_TSTAMP = 1 << 0,
183 /* generate software time stamp */
184 SKBTX_SW_TSTAMP = 1 << 1,
186 /* device driver is going to provide hardware time stamp */
187 SKBTX_IN_PROGRESS = 1 << 2,
189 /* ensure the originating sk reference is available on driver level */
190 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
192 /* device driver supports TX zero-copy buffers */
193 SKBTX_DEV_ZEROCOPY = 1 << 4,
197 * The callback notifies userspace to release buffers when skb DMA is done in
198 * lower device, the skb last reference should be 0 when calling this.
199 * The desc is used to track userspace buffer index.
202 void (*callback)(void *);
207 /* This data is invariant across clones and lives at
208 * the end of the header data, ie. at skb->end.
210 struct skb_shared_info {
211 unsigned short nr_frags;
212 unsigned short gso_size;
213 /* Warning: this field is not always filled in (UFO)! */
214 unsigned short gso_segs;
215 unsigned short gso_type;
218 struct sk_buff *frag_list;
219 struct skb_shared_hwtstamps hwtstamps;
222 * Warning : all fields before dataref are cleared in __alloc_skb()
226 /* Intermediate layers must ensure that destructor_arg
227 * remains valid until skb destructor */
228 void * destructor_arg;
230 /* must be last field, see pskb_expand_head() */
231 skb_frag_t frags[MAX_SKB_FRAGS];
234 /* We divide dataref into two halves. The higher 16 bits hold references
235 * to the payload part of skb->data. The lower 16 bits hold references to
236 * the entire skb->data. A clone of a headerless skb holds the length of
237 * the header in skb->hdr_len.
239 * All users must obey the rule that the skb->data reference count must be
240 * greater than or equal to the payload reference count.
242 * Holding a reference to the payload part means that the user does not
243 * care about modifications to the header part of skb->data.
245 #define SKB_DATAREF_SHIFT 16
246 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
250 SKB_FCLONE_UNAVAILABLE,
256 SKB_GSO_TCPV4 = 1 << 0,
257 SKB_GSO_UDP = 1 << 1,
259 /* This indicates the skb is from an untrusted source. */
260 SKB_GSO_DODGY = 1 << 2,
262 /* This indicates the tcp segment has CWR set. */
263 SKB_GSO_TCP_ECN = 1 << 3,
265 SKB_GSO_TCPV6 = 1 << 4,
267 SKB_GSO_FCOE = 1 << 5,
270 #if BITS_PER_LONG > 32
271 #define NET_SKBUFF_DATA_USES_OFFSET 1
274 #ifdef NET_SKBUFF_DATA_USES_OFFSET
275 typedef unsigned int sk_buff_data_t;
277 typedef unsigned char *sk_buff_data_t;
280 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
281 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
282 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
286 * struct sk_buff - socket buffer
287 * @next: Next buffer in list
288 * @prev: Previous buffer in list
289 * @tstamp: Time we arrived
290 * @sk: Socket we are owned by
291 * @dev: Device we arrived on/are leaving by
292 * @cb: Control buffer. Free for use by every layer. Put private vars here
293 * @_skb_refdst: destination entry (with norefcount bit)
294 * @sp: the security path, used for xfrm
295 * @len: Length of actual data
296 * @data_len: Data length
297 * @mac_len: Length of link layer header
298 * @hdr_len: writable header length of cloned skb
299 * @csum: Checksum (must include start/offset pair)
300 * @csum_start: Offset from skb->head where checksumming should start
301 * @csum_offset: Offset from csum_start where checksum should be stored
302 * @priority: Packet queueing priority
303 * @local_df: allow local fragmentation
304 * @cloned: Head may be cloned (check refcnt to be sure)
305 * @ip_summed: Driver fed us an IP checksum
306 * @nohdr: Payload reference only, must not modify header
307 * @nfctinfo: Relationship of this skb to the connection
308 * @pkt_type: Packet class
309 * @fclone: skbuff clone status
310 * @ipvs_property: skbuff is owned by ipvs
311 * @peeked: this packet has been seen already, so stats have been
312 * done for it, don't do them again
313 * @nf_trace: netfilter packet trace flag
314 * @protocol: Packet protocol from driver
315 * @destructor: Destruct function
316 * @nfct: Associated connection, if any
317 * @nfct_reasm: netfilter conntrack re-assembly pointer
318 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
319 * @skb_iif: ifindex of device we arrived on
320 * @tc_index: Traffic control index
321 * @tc_verd: traffic control verdict
322 * @rxhash: the packet hash computed on receive
323 * @queue_mapping: Queue mapping for multiqueue devices
324 * @ndisc_nodetype: router type (from link layer)
325 * @ooo_okay: allow the mapping of a socket to a queue to be changed
326 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
328 * @dma_cookie: a cookie to one of several possible DMA operations
329 * done by skb DMA functions
330 * @secmark: security marking
331 * @mark: Generic packet mark
332 * @dropcount: total number of sk_receive_queue overflows
333 * @vlan_tci: vlan tag control information
334 * @transport_header: Transport layer header
335 * @network_header: Network layer header
336 * @mac_header: Link layer header
337 * @tail: Tail pointer
339 * @head: Head of buffer
340 * @data: Data head pointer
341 * @truesize: Buffer size
342 * @users: User count - see {datagram,tcp}.c
346 /* These two members must be first. */
347 struct sk_buff *next;
348 struct sk_buff *prev;
353 struct net_device *dev;
356 * This is the control buffer. It is free to use for every
357 * layer. Please put your private variables there. If you
358 * want to keep them across layers you have to do a skb_clone()
359 * first. This is owned by whoever has the skb queued ATM.
361 char cb[48] __aligned(8);
363 unsigned long _skb_refdst;
379 kmemcheck_bitfield_begin(flags1);
390 kmemcheck_bitfield_end(flags1);
393 void (*destructor)(struct sk_buff *skb);
394 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
395 struct nf_conntrack *nfct;
397 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
398 struct sk_buff *nfct_reasm;
400 #ifdef CONFIG_BRIDGE_NETFILTER
401 struct nf_bridge_info *nf_bridge;
405 #ifdef CONFIG_NET_SCHED
406 __u16 tc_index; /* traffic control index */
407 #ifdef CONFIG_NET_CLS_ACT
408 __u16 tc_verd; /* traffic control verdict */
415 kmemcheck_bitfield_begin(flags2);
416 #ifdef CONFIG_IPV6_NDISC_NODETYPE
417 __u8 ndisc_nodetype:2;
421 kmemcheck_bitfield_end(flags2);
425 #ifdef CONFIG_NET_DMA
426 dma_cookie_t dma_cookie;
428 #ifdef CONFIG_NETWORK_SECMARK
438 sk_buff_data_t transport_header;
439 sk_buff_data_t network_header;
440 sk_buff_data_t mac_header;
441 /* These elements must be at the end, see alloc_skb() for details. */
446 unsigned int truesize;
452 * Handling routines are only of interest to the kernel
454 #include <linux/slab.h>
456 #include <asm/system.h>
459 * skb might have a dst pointer attached, refcounted or not.
460 * _skb_refdst low order bit is set if refcount was _not_ taken
462 #define SKB_DST_NOREF 1UL
463 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
466 * skb_dst - returns skb dst_entry
469 * Returns skb dst_entry, regardless of reference taken or not.
471 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
473 /* If refdst was not refcounted, check we still are in a
474 * rcu_read_lock section
476 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
477 !rcu_read_lock_held() &&
478 !rcu_read_lock_bh_held());
479 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
483 * skb_dst_set - sets skb dst
487 * Sets skb dst, assuming a reference was taken on dst and should
488 * be released by skb_dst_drop()
490 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
492 skb->_skb_refdst = (unsigned long)dst;
495 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
498 * skb_dst_is_noref - Test if skb dst isn't refcounted
501 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
503 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
506 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
508 return (struct rtable *)skb_dst(skb);
511 extern void kfree_skb(struct sk_buff *skb);
512 extern void consume_skb(struct sk_buff *skb);
513 extern void __kfree_skb(struct sk_buff *skb);
514 extern struct sk_buff *__alloc_skb(unsigned int size,
515 gfp_t priority, int fclone, int node);
516 static inline struct sk_buff *alloc_skb(unsigned int size,
519 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
522 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
525 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
528 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
530 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
531 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
532 extern struct sk_buff *skb_clone(struct sk_buff *skb,
534 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
536 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
538 extern int pskb_expand_head(struct sk_buff *skb,
539 int nhead, int ntail,
541 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
542 unsigned int headroom);
543 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
544 int newheadroom, int newtailroom,
546 extern int skb_to_sgvec(struct sk_buff *skb,
547 struct scatterlist *sg, int offset,
549 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
550 struct sk_buff **trailer);
551 extern int skb_pad(struct sk_buff *skb, int pad);
552 #define dev_kfree_skb(a) consume_skb(a)
554 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
555 int getfrag(void *from, char *to, int offset,
556 int len,int odd, struct sk_buff *skb),
557 void *from, int length);
559 struct skb_seq_state {
563 __u32 stepped_offset;
564 struct sk_buff *root_skb;
565 struct sk_buff *cur_skb;
569 extern void skb_prepare_seq_read(struct sk_buff *skb,
570 unsigned int from, unsigned int to,
571 struct skb_seq_state *st);
572 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
573 struct skb_seq_state *st);
574 extern void skb_abort_seq_read(struct skb_seq_state *st);
576 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
577 unsigned int to, struct ts_config *config,
578 struct ts_state *state);
580 extern void __skb_get_rxhash(struct sk_buff *skb);
581 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
584 __skb_get_rxhash(skb);
589 #ifdef NET_SKBUFF_DATA_USES_OFFSET
590 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
592 return skb->head + skb->end;
595 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
602 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
604 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
606 return &skb_shinfo(skb)->hwtstamps;
610 * skb_queue_empty - check if a queue is empty
613 * Returns true if the queue is empty, false otherwise.
615 static inline int skb_queue_empty(const struct sk_buff_head *list)
617 return list->next == (struct sk_buff *)list;
621 * skb_queue_is_last - check if skb is the last entry in the queue
625 * Returns true if @skb is the last buffer on the list.
627 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
628 const struct sk_buff *skb)
630 return skb->next == (struct sk_buff *)list;
634 * skb_queue_is_first - check if skb is the first entry in the queue
638 * Returns true if @skb is the first buffer on the list.
640 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
641 const struct sk_buff *skb)
643 return skb->prev == (struct sk_buff *)list;
647 * skb_queue_next - return the next packet in the queue
649 * @skb: current buffer
651 * Return the next packet in @list after @skb. It is only valid to
652 * call this if skb_queue_is_last() evaluates to false.
654 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
655 const struct sk_buff *skb)
657 /* This BUG_ON may seem severe, but if we just return then we
658 * are going to dereference garbage.
660 BUG_ON(skb_queue_is_last(list, skb));
665 * skb_queue_prev - return the prev packet in the queue
667 * @skb: current buffer
669 * Return the prev packet in @list before @skb. It is only valid to
670 * call this if skb_queue_is_first() evaluates to false.
672 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
673 const struct sk_buff *skb)
675 /* This BUG_ON may seem severe, but if we just return then we
676 * are going to dereference garbage.
678 BUG_ON(skb_queue_is_first(list, skb));
683 * skb_get - reference buffer
684 * @skb: buffer to reference
686 * Makes another reference to a socket buffer and returns a pointer
689 static inline struct sk_buff *skb_get(struct sk_buff *skb)
691 atomic_inc(&skb->users);
696 * If users == 1, we are the only owner and are can avoid redundant
701 * skb_cloned - is the buffer a clone
702 * @skb: buffer to check
704 * Returns true if the buffer was generated with skb_clone() and is
705 * one of multiple shared copies of the buffer. Cloned buffers are
706 * shared data so must not be written to under normal circumstances.
708 static inline int skb_cloned(const struct sk_buff *skb)
710 return skb->cloned &&
711 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
715 * skb_header_cloned - is the header a clone
716 * @skb: buffer to check
718 * Returns true if modifying the header part of the buffer requires
719 * the data to be copied.
721 static inline int skb_header_cloned(const struct sk_buff *skb)
728 dataref = atomic_read(&skb_shinfo(skb)->dataref);
729 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
734 * skb_header_release - release reference to header
735 * @skb: buffer to operate on
737 * Drop a reference to the header part of the buffer. This is done
738 * by acquiring a payload reference. You must not read from the header
739 * part of skb->data after this.
741 static inline void skb_header_release(struct sk_buff *skb)
745 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
749 * skb_shared - is the buffer shared
750 * @skb: buffer to check
752 * Returns true if more than one person has a reference to this
755 static inline int skb_shared(const struct sk_buff *skb)
757 return atomic_read(&skb->users) != 1;
761 * skb_share_check - check if buffer is shared and if so clone it
762 * @skb: buffer to check
763 * @pri: priority for memory allocation
765 * If the buffer is shared the buffer is cloned and the old copy
766 * drops a reference. A new clone with a single reference is returned.
767 * If the buffer is not shared the original buffer is returned. When
768 * being called from interrupt status or with spinlocks held pri must
771 * NULL is returned on a memory allocation failure.
773 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
776 might_sleep_if(pri & __GFP_WAIT);
777 if (skb_shared(skb)) {
778 struct sk_buff *nskb = skb_clone(skb, pri);
786 * Copy shared buffers into a new sk_buff. We effectively do COW on
787 * packets to handle cases where we have a local reader and forward
788 * and a couple of other messy ones. The normal one is tcpdumping
789 * a packet thats being forwarded.
793 * skb_unshare - make a copy of a shared buffer
794 * @skb: buffer to check
795 * @pri: priority for memory allocation
797 * If the socket buffer is a clone then this function creates a new
798 * copy of the data, drops a reference count on the old copy and returns
799 * the new copy with the reference count at 1. If the buffer is not a clone
800 * the original buffer is returned. When called with a spinlock held or
801 * from interrupt state @pri must be %GFP_ATOMIC
803 * %NULL is returned on a memory allocation failure.
805 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
808 might_sleep_if(pri & __GFP_WAIT);
809 if (skb_cloned(skb)) {
810 struct sk_buff *nskb = skb_copy(skb, pri);
811 kfree_skb(skb); /* Free our shared copy */
818 * skb_peek - peek at the head of an &sk_buff_head
819 * @list_: list to peek at
821 * Peek an &sk_buff. Unlike most other operations you _MUST_
822 * be careful with this one. A peek leaves the buffer on the
823 * list and someone else may run off with it. You must hold
824 * the appropriate locks or have a private queue to do this.
826 * Returns %NULL for an empty list or a pointer to the head element.
827 * The reference count is not incremented and the reference is therefore
828 * volatile. Use with caution.
830 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
832 struct sk_buff *list = ((struct sk_buff *)list_)->next;
833 if (list == (struct sk_buff *)list_)
839 * skb_peek_tail - peek at the tail of an &sk_buff_head
840 * @list_: list to peek at
842 * Peek an &sk_buff. Unlike most other operations you _MUST_
843 * be careful with this one. A peek leaves the buffer on the
844 * list and someone else may run off with it. You must hold
845 * the appropriate locks or have a private queue to do this.
847 * Returns %NULL for an empty list or a pointer to the tail element.
848 * The reference count is not incremented and the reference is therefore
849 * volatile. Use with caution.
851 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
853 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
854 if (list == (struct sk_buff *)list_)
860 * skb_queue_len - get queue length
861 * @list_: list to measure
863 * Return the length of an &sk_buff queue.
865 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
871 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
872 * @list: queue to initialize
874 * This initializes only the list and queue length aspects of
875 * an sk_buff_head object. This allows to initialize the list
876 * aspects of an sk_buff_head without reinitializing things like
877 * the spinlock. It can also be used for on-stack sk_buff_head
878 * objects where the spinlock is known to not be used.
880 static inline void __skb_queue_head_init(struct sk_buff_head *list)
882 list->prev = list->next = (struct sk_buff *)list;
887 * This function creates a split out lock class for each invocation;
888 * this is needed for now since a whole lot of users of the skb-queue
889 * infrastructure in drivers have different locking usage (in hardirq)
890 * than the networking core (in softirq only). In the long run either the
891 * network layer or drivers should need annotation to consolidate the
892 * main types of usage into 3 classes.
894 static inline void skb_queue_head_init(struct sk_buff_head *list)
896 spin_lock_init(&list->lock);
897 __skb_queue_head_init(list);
900 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
901 struct lock_class_key *class)
903 skb_queue_head_init(list);
904 lockdep_set_class(&list->lock, class);
908 * Insert an sk_buff on a list.
910 * The "__skb_xxxx()" functions are the non-atomic ones that
911 * can only be called with interrupts disabled.
913 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
914 static inline void __skb_insert(struct sk_buff *newsk,
915 struct sk_buff *prev, struct sk_buff *next,
916 struct sk_buff_head *list)
920 next->prev = prev->next = newsk;
924 static inline void __skb_queue_splice(const struct sk_buff_head *list,
925 struct sk_buff *prev,
926 struct sk_buff *next)
928 struct sk_buff *first = list->next;
929 struct sk_buff *last = list->prev;
939 * skb_queue_splice - join two skb lists, this is designed for stacks
940 * @list: the new list to add
941 * @head: the place to add it in the first list
943 static inline void skb_queue_splice(const struct sk_buff_head *list,
944 struct sk_buff_head *head)
946 if (!skb_queue_empty(list)) {
947 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
948 head->qlen += list->qlen;
953 * skb_queue_splice - join two skb lists and reinitialise the emptied list
954 * @list: the new list to add
955 * @head: the place to add it in the first list
957 * The list at @list is reinitialised
959 static inline void skb_queue_splice_init(struct sk_buff_head *list,
960 struct sk_buff_head *head)
962 if (!skb_queue_empty(list)) {
963 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
964 head->qlen += list->qlen;
965 __skb_queue_head_init(list);
970 * skb_queue_splice_tail - join two skb lists, each list being a queue
971 * @list: the new list to add
972 * @head: the place to add it in the first list
974 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
975 struct sk_buff_head *head)
977 if (!skb_queue_empty(list)) {
978 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
979 head->qlen += list->qlen;
984 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
985 * @list: the new list to add
986 * @head: the place to add it in the first list
988 * Each of the lists is a queue.
989 * The list at @list is reinitialised
991 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
992 struct sk_buff_head *head)
994 if (!skb_queue_empty(list)) {
995 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
996 head->qlen += list->qlen;
997 __skb_queue_head_init(list);
1002 * __skb_queue_after - queue a buffer at the list head
1003 * @list: list to use
1004 * @prev: place after this buffer
1005 * @newsk: buffer to queue
1007 * Queue a buffer int the middle of a list. This function takes no locks
1008 * and you must therefore hold required locks before calling it.
1010 * A buffer cannot be placed on two lists at the same time.
1012 static inline void __skb_queue_after(struct sk_buff_head *list,
1013 struct sk_buff *prev,
1014 struct sk_buff *newsk)
1016 __skb_insert(newsk, prev, prev->next, list);
1019 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1020 struct sk_buff_head *list);
1022 static inline void __skb_queue_before(struct sk_buff_head *list,
1023 struct sk_buff *next,
1024 struct sk_buff *newsk)
1026 __skb_insert(newsk, next->prev, next, list);
1030 * __skb_queue_head - queue a buffer at the list head
1031 * @list: list to use
1032 * @newsk: buffer to queue
1034 * Queue a buffer at the start of a list. This function takes no locks
1035 * and you must therefore hold required locks before calling it.
1037 * A buffer cannot be placed on two lists at the same time.
1039 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1040 static inline void __skb_queue_head(struct sk_buff_head *list,
1041 struct sk_buff *newsk)
1043 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1047 * __skb_queue_tail - queue a buffer at the list tail
1048 * @list: list to use
1049 * @newsk: buffer to queue
1051 * Queue a buffer at the end of a list. This function takes no locks
1052 * and you must therefore hold required locks before calling it.
1054 * A buffer cannot be placed on two lists at the same time.
1056 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1057 static inline void __skb_queue_tail(struct sk_buff_head *list,
1058 struct sk_buff *newsk)
1060 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1064 * remove sk_buff from list. _Must_ be called atomically, and with
1067 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1068 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1070 struct sk_buff *next, *prev;
1075 skb->next = skb->prev = NULL;
1081 * __skb_dequeue - remove from the head of the queue
1082 * @list: list to dequeue from
1084 * Remove the head of the list. This function does not take any locks
1085 * so must be used with appropriate locks held only. The head item is
1086 * returned or %NULL if the list is empty.
1088 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1089 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1091 struct sk_buff *skb = skb_peek(list);
1093 __skb_unlink(skb, list);
1098 * __skb_dequeue_tail - remove from the tail of the queue
1099 * @list: list to dequeue from
1101 * Remove the tail of the list. This function does not take any locks
1102 * so must be used with appropriate locks held only. The tail item is
1103 * returned or %NULL if the list is empty.
1105 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1106 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1108 struct sk_buff *skb = skb_peek_tail(list);
1110 __skb_unlink(skb, list);
1115 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1117 return skb->data_len;
1120 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1122 return skb->len - skb->data_len;
1125 static inline int skb_pagelen(const struct sk_buff *skb)
1129 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1130 len += skb_shinfo(skb)->frags[i].size;
1131 return len + skb_headlen(skb);
1135 * __skb_fill_page_desc - initialise a paged fragment in an skb
1136 * @skb: buffer containing fragment to be initialised
1137 * @i: paged fragment index to initialise
1138 * @page: the page to use for this fragment
1139 * @off: the offset to the data with @page
1140 * @size: the length of the data
1142 * Initialises the @i'th fragment of @skb to point to &size bytes at
1143 * offset @off within @page.
1145 * Does not take any additional reference on the fragment.
1147 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1148 struct page *page, int off, int size)
1150 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1153 frag->page_offset = off;
1158 * skb_fill_page_desc - initialise a paged fragment in an skb
1159 * @skb: buffer containing fragment to be initialised
1160 * @i: paged fragment index to initialise
1161 * @page: the page to use for this fragment
1162 * @off: the offset to the data with @page
1163 * @size: the length of the data
1165 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1166 * @skb to point to &size bytes at offset @off within @page. In
1167 * addition updates @skb such that @i is the last fragment.
1169 * Does not take any additional reference on the fragment.
1171 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1172 struct page *page, int off, int size)
1174 __skb_fill_page_desc(skb, i, page, off, size);
1175 skb_shinfo(skb)->nr_frags = i + 1;
1178 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1181 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1182 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1183 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1185 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1186 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1188 return skb->head + skb->tail;
1191 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1193 skb->tail = skb->data - skb->head;
1196 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1198 skb_reset_tail_pointer(skb);
1199 skb->tail += offset;
1201 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1202 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1207 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1209 skb->tail = skb->data;
1212 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1214 skb->tail = skb->data + offset;
1217 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1220 * Add data to an sk_buff
1222 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1223 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1225 unsigned char *tmp = skb_tail_pointer(skb);
1226 SKB_LINEAR_ASSERT(skb);
1232 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1233 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1240 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1241 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1244 BUG_ON(skb->len < skb->data_len);
1245 return skb->data += len;
1248 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1250 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1253 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1255 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1257 if (len > skb_headlen(skb) &&
1258 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1261 return skb->data += len;
1264 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1266 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1269 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1271 if (likely(len <= skb_headlen(skb)))
1273 if (unlikely(len > skb->len))
1275 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1279 * skb_headroom - bytes at buffer head
1280 * @skb: buffer to check
1282 * Return the number of bytes of free space at the head of an &sk_buff.
1284 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1286 return skb->data - skb->head;
1290 * skb_tailroom - bytes at buffer end
1291 * @skb: buffer to check
1293 * Return the number of bytes of free space at the tail of an sk_buff
1295 static inline int skb_tailroom(const struct sk_buff *skb)
1297 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1301 * skb_reserve - adjust headroom
1302 * @skb: buffer to alter
1303 * @len: bytes to move
1305 * Increase the headroom of an empty &sk_buff by reducing the tail
1306 * room. This is only allowed for an empty buffer.
1308 static inline void skb_reserve(struct sk_buff *skb, int len)
1314 static inline void skb_reset_mac_len(struct sk_buff *skb)
1316 skb->mac_len = skb->network_header - skb->mac_header;
1319 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1320 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1322 return skb->head + skb->transport_header;
1325 static inline void skb_reset_transport_header(struct sk_buff *skb)
1327 skb->transport_header = skb->data - skb->head;
1330 static inline void skb_set_transport_header(struct sk_buff *skb,
1333 skb_reset_transport_header(skb);
1334 skb->transport_header += offset;
1337 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1339 return skb->head + skb->network_header;
1342 static inline void skb_reset_network_header(struct sk_buff *skb)
1344 skb->network_header = skb->data - skb->head;
1347 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1349 skb_reset_network_header(skb);
1350 skb->network_header += offset;
1353 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1355 return skb->head + skb->mac_header;
1358 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1360 return skb->mac_header != ~0U;
1363 static inline void skb_reset_mac_header(struct sk_buff *skb)
1365 skb->mac_header = skb->data - skb->head;
1368 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1370 skb_reset_mac_header(skb);
1371 skb->mac_header += offset;
1374 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1376 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1378 return skb->transport_header;
1381 static inline void skb_reset_transport_header(struct sk_buff *skb)
1383 skb->transport_header = skb->data;
1386 static inline void skb_set_transport_header(struct sk_buff *skb,
1389 skb->transport_header = skb->data + offset;
1392 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1394 return skb->network_header;
1397 static inline void skb_reset_network_header(struct sk_buff *skb)
1399 skb->network_header = skb->data;
1402 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1404 skb->network_header = skb->data + offset;
1407 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1409 return skb->mac_header;
1412 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1414 return skb->mac_header != NULL;
1417 static inline void skb_reset_mac_header(struct sk_buff *skb)
1419 skb->mac_header = skb->data;
1422 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1424 skb->mac_header = skb->data + offset;
1426 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1428 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1430 return skb->csum_start - skb_headroom(skb);
1433 static inline int skb_transport_offset(const struct sk_buff *skb)
1435 return skb_transport_header(skb) - skb->data;
1438 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1440 return skb->transport_header - skb->network_header;
1443 static inline int skb_network_offset(const struct sk_buff *skb)
1445 return skb_network_header(skb) - skb->data;
1448 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1450 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1454 * CPUs often take a performance hit when accessing unaligned memory
1455 * locations. The actual performance hit varies, it can be small if the
1456 * hardware handles it or large if we have to take an exception and fix it
1459 * Since an ethernet header is 14 bytes network drivers often end up with
1460 * the IP header at an unaligned offset. The IP header can be aligned by
1461 * shifting the start of the packet by 2 bytes. Drivers should do this
1464 * skb_reserve(skb, NET_IP_ALIGN);
1466 * The downside to this alignment of the IP header is that the DMA is now
1467 * unaligned. On some architectures the cost of an unaligned DMA is high
1468 * and this cost outweighs the gains made by aligning the IP header.
1470 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1473 #ifndef NET_IP_ALIGN
1474 #define NET_IP_ALIGN 2
1478 * The networking layer reserves some headroom in skb data (via
1479 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1480 * the header has to grow. In the default case, if the header has to grow
1481 * 32 bytes or less we avoid the reallocation.
1483 * Unfortunately this headroom changes the DMA alignment of the resulting
1484 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1485 * on some architectures. An architecture can override this value,
1486 * perhaps setting it to a cacheline in size (since that will maintain
1487 * cacheline alignment of the DMA). It must be a power of 2.
1489 * Various parts of the networking layer expect at least 32 bytes of
1490 * headroom, you should not reduce this.
1492 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1493 * to reduce average number of cache lines per packet.
1494 * get_rps_cpus() for example only access one 64 bytes aligned block :
1495 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1498 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1501 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1503 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1505 if (unlikely(skb_is_nonlinear(skb))) {
1510 skb_set_tail_pointer(skb, len);
1513 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1515 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1518 return ___pskb_trim(skb, len);
1519 __skb_trim(skb, len);
1523 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1525 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1529 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1530 * @skb: buffer to alter
1533 * This is identical to pskb_trim except that the caller knows that
1534 * the skb is not cloned so we should never get an error due to out-
1537 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1539 int err = pskb_trim(skb, len);
1544 * skb_orphan - orphan a buffer
1545 * @skb: buffer to orphan
1547 * If a buffer currently has an owner then we call the owner's
1548 * destructor function and make the @skb unowned. The buffer continues
1549 * to exist but is no longer charged to its former owner.
1551 static inline void skb_orphan(struct sk_buff *skb)
1553 if (skb->destructor)
1554 skb->destructor(skb);
1555 skb->destructor = NULL;
1560 * __skb_queue_purge - empty a list
1561 * @list: list to empty
1563 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1564 * the list and one reference dropped. This function does not take the
1565 * list lock and the caller must hold the relevant locks to use it.
1567 extern void skb_queue_purge(struct sk_buff_head *list);
1568 static inline void __skb_queue_purge(struct sk_buff_head *list)
1570 struct sk_buff *skb;
1571 while ((skb = __skb_dequeue(list)) != NULL)
1576 * __dev_alloc_skb - allocate an skbuff for receiving
1577 * @length: length to allocate
1578 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1580 * Allocate a new &sk_buff and assign it a usage count of one. The
1581 * buffer has unspecified headroom built in. Users should allocate
1582 * the headroom they think they need without accounting for the
1583 * built in space. The built in space is used for optimisations.
1585 * %NULL is returned if there is no free memory.
1587 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1590 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1592 skb_reserve(skb, NET_SKB_PAD);
1596 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1598 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1599 unsigned int length, gfp_t gfp_mask);
1602 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1603 * @dev: network device to receive on
1604 * @length: length to allocate
1606 * Allocate a new &sk_buff and assign it a usage count of one. The
1607 * buffer has unspecified headroom built in. Users should allocate
1608 * the headroom they think they need without accounting for the
1609 * built in space. The built in space is used for optimisations.
1611 * %NULL is returned if there is no free memory. Although this function
1612 * allocates memory it can be called from an interrupt.
1614 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1615 unsigned int length)
1617 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1620 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1621 unsigned int length, gfp_t gfp)
1623 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1625 if (NET_IP_ALIGN && skb)
1626 skb_reserve(skb, NET_IP_ALIGN);
1630 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1631 unsigned int length)
1633 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1637 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1638 * @dev: network device to receive on
1639 * @gfp_mask: alloc_pages_node mask
1641 * Allocate a new page. dev currently unused.
1643 * %NULL is returned if there is no free memory.
1645 static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1647 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1651 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1652 * @dev: network device to receive on
1654 * Allocate a new page. dev currently unused.
1656 * %NULL is returned if there is no free memory.
1658 static inline struct page *netdev_alloc_page(struct net_device *dev)
1660 return __netdev_alloc_page(dev, GFP_ATOMIC);
1663 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1669 * skb_frag_page - retrieve the page refered to by a paged fragment
1670 * @frag: the paged fragment
1672 * Returns the &struct page associated with @frag.
1674 static inline struct page *skb_frag_page(const skb_frag_t *frag)
1680 * __skb_frag_ref - take an addition reference on a paged fragment.
1681 * @frag: the paged fragment
1683 * Takes an additional reference on the paged fragment @frag.
1685 static inline void __skb_frag_ref(skb_frag_t *frag)
1687 get_page(skb_frag_page(frag));
1691 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1693 * @f: the fragment offset.
1695 * Takes an additional reference on the @f'th paged fragment of @skb.
1697 static inline void skb_frag_ref(struct sk_buff *skb, int f)
1699 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1703 * __skb_frag_unref - release a reference on a paged fragment.
1704 * @frag: the paged fragment
1706 * Releases a reference on the paged fragment @frag.
1708 static inline void __skb_frag_unref(skb_frag_t *frag)
1710 put_page(skb_frag_page(frag));
1714 * skb_frag_unref - release a reference on a paged fragment of an skb.
1716 * @f: the fragment offset
1718 * Releases a reference on the @f'th paged fragment of @skb.
1720 static inline void skb_frag_unref(struct sk_buff *skb, int f)
1722 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1726 * skb_frag_address - gets the address of the data contained in a paged fragment
1727 * @frag: the paged fragment buffer
1729 * Returns the address of the data within @frag. The page must already
1732 static inline void *skb_frag_address(const skb_frag_t *frag)
1734 return page_address(skb_frag_page(frag)) + frag->page_offset;
1738 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1739 * @frag: the paged fragment buffer
1741 * Returns the address of the data within @frag. Checks that the page
1742 * is mapped and returns %NULL otherwise.
1744 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1746 void *ptr = page_address(skb_frag_page(frag));
1750 return ptr + frag->page_offset;
1754 * __skb_frag_set_page - sets the page contained in a paged fragment
1755 * @frag: the paged fragment
1756 * @page: the page to set
1758 * Sets the fragment @frag to contain @page.
1760 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1763 __skb_frag_ref(frag);
1767 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1769 * @f: the fragment offset
1770 * @page: the page to set
1772 * Sets the @f'th fragment of @skb to contain @page.
1774 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1777 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1781 * skb_frag_dma_map - maps a paged fragment via the DMA API
1782 * @device: the device to map the fragment to
1783 * @frag: the paged fragment to map
1784 * @offset: the offset within the fragment (starting at the
1785 * fragment's own offset)
1786 * @size: the number of bytes to map
1787 * @direction: the direction of the mapping (%PCI_DMA_*)
1789 * Maps the page associated with @frag to @device.
1791 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1792 const skb_frag_t *frag,
1793 size_t offset, size_t size,
1794 enum dma_data_direction dir)
1796 return dma_map_page(dev, skb_frag_page(frag),
1797 frag->page_offset + offset, size, dir);
1801 * skb_clone_writable - is the header of a clone writable
1802 * @skb: buffer to check
1803 * @len: length up to which to write
1805 * Returns true if modifying the header part of the cloned buffer
1806 * does not requires the data to be copied.
1808 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1810 return !skb_header_cloned(skb) &&
1811 skb_headroom(skb) + len <= skb->hdr_len;
1814 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1819 if (headroom < NET_SKB_PAD)
1820 headroom = NET_SKB_PAD;
1821 if (headroom > skb_headroom(skb))
1822 delta = headroom - skb_headroom(skb);
1824 if (delta || cloned)
1825 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1831 * skb_cow - copy header of skb when it is required
1832 * @skb: buffer to cow
1833 * @headroom: needed headroom
1835 * If the skb passed lacks sufficient headroom or its data part
1836 * is shared, data is reallocated. If reallocation fails, an error
1837 * is returned and original skb is not changed.
1839 * The result is skb with writable area skb->head...skb->tail
1840 * and at least @headroom of space at head.
1842 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1844 return __skb_cow(skb, headroom, skb_cloned(skb));
1848 * skb_cow_head - skb_cow but only making the head writable
1849 * @skb: buffer to cow
1850 * @headroom: needed headroom
1852 * This function is identical to skb_cow except that we replace the
1853 * skb_cloned check by skb_header_cloned. It should be used when
1854 * you only need to push on some header and do not need to modify
1857 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1859 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1863 * skb_padto - pad an skbuff up to a minimal size
1864 * @skb: buffer to pad
1865 * @len: minimal length
1867 * Pads up a buffer to ensure the trailing bytes exist and are
1868 * blanked. If the buffer already contains sufficient data it
1869 * is untouched. Otherwise it is extended. Returns zero on
1870 * success. The skb is freed on error.
1873 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1875 unsigned int size = skb->len;
1876 if (likely(size >= len))
1878 return skb_pad(skb, len - size);
1881 static inline int skb_add_data(struct sk_buff *skb,
1882 char __user *from, int copy)
1884 const int off = skb->len;
1886 if (skb->ip_summed == CHECKSUM_NONE) {
1888 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1891 skb->csum = csum_block_add(skb->csum, csum, off);
1894 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1897 __skb_trim(skb, off);
1901 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1902 const struct page *page, int off)
1905 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1907 return page == skb_frag_page(frag) &&
1908 off == frag->page_offset + frag->size;
1913 static inline int __skb_linearize(struct sk_buff *skb)
1915 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1919 * skb_linearize - convert paged skb to linear one
1920 * @skb: buffer to linarize
1922 * If there is no free memory -ENOMEM is returned, otherwise zero
1923 * is returned and the old skb data released.
1925 static inline int skb_linearize(struct sk_buff *skb)
1927 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1931 * skb_linearize_cow - make sure skb is linear and writable
1932 * @skb: buffer to process
1934 * If there is no free memory -ENOMEM is returned, otherwise zero
1935 * is returned and the old skb data released.
1937 static inline int skb_linearize_cow(struct sk_buff *skb)
1939 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1940 __skb_linearize(skb) : 0;
1944 * skb_postpull_rcsum - update checksum for received skb after pull
1945 * @skb: buffer to update
1946 * @start: start of data before pull
1947 * @len: length of data pulled
1949 * After doing a pull on a received packet, you need to call this to
1950 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1951 * CHECKSUM_NONE so that it can be recomputed from scratch.
1954 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1955 const void *start, unsigned int len)
1957 if (skb->ip_summed == CHECKSUM_COMPLETE)
1958 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1961 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1964 * pskb_trim_rcsum - trim received skb and update checksum
1965 * @skb: buffer to trim
1968 * This is exactly the same as pskb_trim except that it ensures the
1969 * checksum of received packets are still valid after the operation.
1972 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1974 if (likely(len >= skb->len))
1976 if (skb->ip_summed == CHECKSUM_COMPLETE)
1977 skb->ip_summed = CHECKSUM_NONE;
1978 return __pskb_trim(skb, len);
1981 #define skb_queue_walk(queue, skb) \
1982 for (skb = (queue)->next; \
1983 skb != (struct sk_buff *)(queue); \
1986 #define skb_queue_walk_safe(queue, skb, tmp) \
1987 for (skb = (queue)->next, tmp = skb->next; \
1988 skb != (struct sk_buff *)(queue); \
1989 skb = tmp, tmp = skb->next)
1991 #define skb_queue_walk_from(queue, skb) \
1992 for (; skb != (struct sk_buff *)(queue); \
1995 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1996 for (tmp = skb->next; \
1997 skb != (struct sk_buff *)(queue); \
1998 skb = tmp, tmp = skb->next)
2000 #define skb_queue_reverse_walk(queue, skb) \
2001 for (skb = (queue)->prev; \
2002 skb != (struct sk_buff *)(queue); \
2005 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2006 for (skb = (queue)->prev, tmp = skb->prev; \
2007 skb != (struct sk_buff *)(queue); \
2008 skb = tmp, tmp = skb->prev)
2010 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2011 for (tmp = skb->prev; \
2012 skb != (struct sk_buff *)(queue); \
2013 skb = tmp, tmp = skb->prev)
2015 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2017 return skb_shinfo(skb)->frag_list != NULL;
2020 static inline void skb_frag_list_init(struct sk_buff *skb)
2022 skb_shinfo(skb)->frag_list = NULL;
2025 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2027 frag->next = skb_shinfo(skb)->frag_list;
2028 skb_shinfo(skb)->frag_list = frag;
2031 #define skb_walk_frags(skb, iter) \
2032 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2034 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2035 int *peeked, int *err);
2036 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2037 int noblock, int *err);
2038 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2039 struct poll_table_struct *wait);
2040 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2041 int offset, struct iovec *to,
2043 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
2046 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2048 const struct iovec *from,
2051 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2053 const struct iovec *to,
2056 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2057 extern void skb_free_datagram_locked(struct sock *sk,
2058 struct sk_buff *skb);
2059 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
2060 unsigned int flags);
2061 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2062 int len, __wsum csum);
2063 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2065 extern int skb_store_bits(struct sk_buff *skb, int offset,
2066 const void *from, int len);
2067 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
2068 int offset, u8 *to, int len,
2070 extern int skb_splice_bits(struct sk_buff *skb,
2071 unsigned int offset,
2072 struct pipe_inode_info *pipe,
2074 unsigned int flags);
2075 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2076 extern void skb_split(struct sk_buff *skb,
2077 struct sk_buff *skb1, const u32 len);
2078 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2081 extern struct sk_buff *skb_segment(struct sk_buff *skb, u32 features);
2083 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2084 int len, void *buffer)
2086 int hlen = skb_headlen(skb);
2088 if (hlen - offset >= len)
2089 return skb->data + offset;
2091 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2097 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2099 const unsigned int len)
2101 memcpy(to, skb->data, len);
2104 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2105 const int offset, void *to,
2106 const unsigned int len)
2108 memcpy(to, skb->data + offset, len);
2111 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2113 const unsigned int len)
2115 memcpy(skb->data, from, len);
2118 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2121 const unsigned int len)
2123 memcpy(skb->data + offset, from, len);
2126 extern void skb_init(void);
2128 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2134 * skb_get_timestamp - get timestamp from a skb
2135 * @skb: skb to get stamp from
2136 * @stamp: pointer to struct timeval to store stamp in
2138 * Timestamps are stored in the skb as offsets to a base timestamp.
2139 * This function converts the offset back to a struct timeval and stores
2142 static inline void skb_get_timestamp(const struct sk_buff *skb,
2143 struct timeval *stamp)
2145 *stamp = ktime_to_timeval(skb->tstamp);
2148 static inline void skb_get_timestampns(const struct sk_buff *skb,
2149 struct timespec *stamp)
2151 *stamp = ktime_to_timespec(skb->tstamp);
2154 static inline void __net_timestamp(struct sk_buff *skb)
2156 skb->tstamp = ktime_get_real();
2159 static inline ktime_t net_timedelta(ktime_t t)
2161 return ktime_sub(ktime_get_real(), t);
2164 static inline ktime_t net_invalid_timestamp(void)
2166 return ktime_set(0, 0);
2169 extern void skb_timestamping_init(void);
2171 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2173 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2174 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2176 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2178 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2182 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2187 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2190 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2192 * @skb: clone of the the original outgoing packet
2193 * @hwtstamps: hardware time stamps
2196 void skb_complete_tx_timestamp(struct sk_buff *skb,
2197 struct skb_shared_hwtstamps *hwtstamps);
2200 * skb_tstamp_tx - queue clone of skb with send time stamps
2201 * @orig_skb: the original outgoing packet
2202 * @hwtstamps: hardware time stamps, may be NULL if not available
2204 * If the skb has a socket associated, then this function clones the
2205 * skb (thus sharing the actual data and optional structures), stores
2206 * the optional hardware time stamping information (if non NULL) or
2207 * generates a software time stamp (otherwise), then queues the clone
2208 * to the error queue of the socket. Errors are silently ignored.
2210 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2211 struct skb_shared_hwtstamps *hwtstamps);
2213 static inline void sw_tx_timestamp(struct sk_buff *skb)
2215 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2216 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2217 skb_tstamp_tx(skb, NULL);
2221 * skb_tx_timestamp() - Driver hook for transmit timestamping
2223 * Ethernet MAC Drivers should call this function in their hard_xmit()
2224 * function immediately before giving the sk_buff to the MAC hardware.
2226 * @skb: A socket buffer.
2228 static inline void skb_tx_timestamp(struct sk_buff *skb)
2230 skb_clone_tx_timestamp(skb);
2231 sw_tx_timestamp(skb);
2234 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2235 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2237 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2239 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2243 * skb_checksum_complete - Calculate checksum of an entire packet
2244 * @skb: packet to process
2246 * This function calculates the checksum over the entire packet plus
2247 * the value of skb->csum. The latter can be used to supply the
2248 * checksum of a pseudo header as used by TCP/UDP. It returns the
2251 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2252 * this function can be used to verify that checksum on received
2253 * packets. In that case the function should return zero if the
2254 * checksum is correct. In particular, this function will return zero
2255 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2256 * hardware has already verified the correctness of the checksum.
2258 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2260 return skb_csum_unnecessary(skb) ?
2261 0 : __skb_checksum_complete(skb);
2264 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2265 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2266 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2268 if (nfct && atomic_dec_and_test(&nfct->use))
2269 nf_conntrack_destroy(nfct);
2271 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2274 atomic_inc(&nfct->use);
2277 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2278 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2281 atomic_inc(&skb->users);
2283 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2289 #ifdef CONFIG_BRIDGE_NETFILTER
2290 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2292 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2295 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2298 atomic_inc(&nf_bridge->use);
2300 #endif /* CONFIG_BRIDGE_NETFILTER */
2301 static inline void nf_reset(struct sk_buff *skb)
2303 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2304 nf_conntrack_put(skb->nfct);
2307 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2308 nf_conntrack_put_reasm(skb->nfct_reasm);
2309 skb->nfct_reasm = NULL;
2311 #ifdef CONFIG_BRIDGE_NETFILTER
2312 nf_bridge_put(skb->nf_bridge);
2313 skb->nf_bridge = NULL;
2317 /* Note: This doesn't put any conntrack and bridge info in dst. */
2318 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2320 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2321 dst->nfct = src->nfct;
2322 nf_conntrack_get(src->nfct);
2323 dst->nfctinfo = src->nfctinfo;
2325 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2326 dst->nfct_reasm = src->nfct_reasm;
2327 nf_conntrack_get_reasm(src->nfct_reasm);
2329 #ifdef CONFIG_BRIDGE_NETFILTER
2330 dst->nf_bridge = src->nf_bridge;
2331 nf_bridge_get(src->nf_bridge);
2335 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2337 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2338 nf_conntrack_put(dst->nfct);
2340 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2341 nf_conntrack_put_reasm(dst->nfct_reasm);
2343 #ifdef CONFIG_BRIDGE_NETFILTER
2344 nf_bridge_put(dst->nf_bridge);
2346 __nf_copy(dst, src);
2349 #ifdef CONFIG_NETWORK_SECMARK
2350 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2352 to->secmark = from->secmark;
2355 static inline void skb_init_secmark(struct sk_buff *skb)
2360 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2363 static inline void skb_init_secmark(struct sk_buff *skb)
2367 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2369 skb->queue_mapping = queue_mapping;
2372 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2374 return skb->queue_mapping;
2377 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2379 to->queue_mapping = from->queue_mapping;
2382 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2384 skb->queue_mapping = rx_queue + 1;
2387 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2389 return skb->queue_mapping - 1;
2392 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2394 return skb->queue_mapping != 0;
2397 extern u16 __skb_tx_hash(const struct net_device *dev,
2398 const struct sk_buff *skb,
2399 unsigned int num_tx_queues);
2402 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2407 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2413 static inline int skb_is_gso(const struct sk_buff *skb)
2415 return skb_shinfo(skb)->gso_size;
2418 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2420 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2423 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2425 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2427 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2428 * wanted then gso_type will be set. */
2429 struct skb_shared_info *shinfo = skb_shinfo(skb);
2430 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2431 unlikely(shinfo->gso_type == 0)) {
2432 __skb_warn_lro_forwarding(skb);
2438 static inline void skb_forward_csum(struct sk_buff *skb)
2440 /* Unfortunately we don't support this one. Any brave souls? */
2441 if (skb->ip_summed == CHECKSUM_COMPLETE)
2442 skb->ip_summed = CHECKSUM_NONE;
2446 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2447 * @skb: skb to check
2449 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2450 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2451 * use this helper, to document places where we make this assertion.
2453 static inline void skb_checksum_none_assert(struct sk_buff *skb)
2456 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2460 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2462 #endif /* __KERNEL__ */
2463 #endif /* _LINUX_SKBUFF_H */