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 {
167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
172 * @prevent_sk_orphan: make sk reference available on driver level
173 * @flags: all shared_tx flags
175 * These flags are attached to packets as part of the
176 * &skb_shared_info. Use skb_tx() to get a pointer.
178 union skb_shared_tx {
188 /* This data is invariant across clones and lives at
189 * the end of the header data, ie. at skb->end.
191 struct skb_shared_info {
192 unsigned short nr_frags;
193 unsigned short gso_size;
194 /* Warning: this field is not always filled in (UFO)! */
195 unsigned short gso_segs;
196 unsigned short gso_type;
198 union skb_shared_tx tx_flags;
199 struct sk_buff *frag_list;
200 struct skb_shared_hwtstamps hwtstamps;
203 * Warning : all fields before dataref are cleared in __alloc_skb()
207 /* Intermediate layers must ensure that destructor_arg
208 * remains valid until skb destructor */
209 void * destructor_arg;
210 /* must be last field, see pskb_expand_head() */
211 skb_frag_t frags[MAX_SKB_FRAGS];
214 /* We divide dataref into two halves. The higher 16 bits hold references
215 * to the payload part of skb->data. The lower 16 bits hold references to
216 * the entire skb->data. A clone of a headerless skb holds the length of
217 * the header in skb->hdr_len.
219 * All users must obey the rule that the skb->data reference count must be
220 * greater than or equal to the payload reference count.
222 * Holding a reference to the payload part means that the user does not
223 * care about modifications to the header part of skb->data.
225 #define SKB_DATAREF_SHIFT 16
226 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
230 SKB_FCLONE_UNAVAILABLE,
236 SKB_GSO_TCPV4 = 1 << 0,
237 SKB_GSO_UDP = 1 << 1,
239 /* This indicates the skb is from an untrusted source. */
240 SKB_GSO_DODGY = 1 << 2,
242 /* This indicates the tcp segment has CWR set. */
243 SKB_GSO_TCP_ECN = 1 << 3,
245 SKB_GSO_TCPV6 = 1 << 4,
247 SKB_GSO_FCOE = 1 << 5,
250 #if BITS_PER_LONG > 32
251 #define NET_SKBUFF_DATA_USES_OFFSET 1
254 #ifdef NET_SKBUFF_DATA_USES_OFFSET
255 typedef unsigned int sk_buff_data_t;
257 typedef unsigned char *sk_buff_data_t;
261 * struct sk_buff - socket buffer
262 * @next: Next buffer in list
263 * @prev: Previous buffer in list
264 * @sk: Socket we are owned by
265 * @tstamp: Time we arrived
266 * @dev: Device we arrived on/are leaving by
267 * @transport_header: Transport layer header
268 * @network_header: Network layer header
269 * @mac_header: Link layer header
270 * @_skb_refdst: destination entry (with norefcount bit)
271 * @sp: the security path, used for xfrm
272 * @cb: Control buffer. Free for use by every layer. Put private vars here
273 * @len: Length of actual data
274 * @data_len: Data length
275 * @mac_len: Length of link layer header
276 * @hdr_len: writable header length of cloned skb
277 * @csum: Checksum (must include start/offset pair)
278 * @csum_start: Offset from skb->head where checksumming should start
279 * @csum_offset: Offset from csum_start where checksum should be stored
280 * @local_df: allow local fragmentation
281 * @cloned: Head may be cloned (check refcnt to be sure)
282 * @nohdr: Payload reference only, must not modify header
283 * @pkt_type: Packet class
284 * @fclone: skbuff clone status
285 * @ip_summed: Driver fed us an IP checksum
286 * @priority: Packet queueing priority
287 * @users: User count - see {datagram,tcp}.c
288 * @protocol: Packet protocol from driver
289 * @truesize: Buffer size
290 * @head: Head of buffer
291 * @data: Data head pointer
292 * @tail: Tail pointer
294 * @destructor: Destruct function
295 * @mark: Generic packet mark
296 * @nfct: Associated connection, if any
297 * @ipvs_property: skbuff is owned by ipvs
298 * @peeked: this packet has been seen already, so stats have been
299 * done for it, don't do them again
300 * @nf_trace: netfilter packet trace flag
301 * @nfctinfo: Relationship of this skb to the connection
302 * @nfct_reasm: netfilter conntrack re-assembly pointer
303 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
304 * @skb_iif: ifindex of device we arrived on
305 * @rxhash: the packet hash computed on receive
306 * @queue_mapping: Queue mapping for multiqueue devices
307 * @tc_index: Traffic control index
308 * @tc_verd: traffic control verdict
309 * @ndisc_nodetype: router type (from link layer)
310 * @dma_cookie: a cookie to one of several possible DMA operations
311 * done by skb DMA functions
312 * @secmark: security marking
313 * @vlan_tci: vlan tag control information
317 /* These two members must be first. */
318 struct sk_buff *next;
319 struct sk_buff *prev;
324 struct net_device *dev;
327 * This is the control buffer. It is free to use for every
328 * layer. Please put your private variables there. If you
329 * want to keep them across layers you have to do a skb_clone()
330 * first. This is owned by whoever has the skb queued ATM.
332 char cb[48] __aligned(8);
334 unsigned long _skb_refdst;
350 kmemcheck_bitfield_begin(flags1);
361 kmemcheck_bitfield_end(flags1);
364 void (*destructor)(struct sk_buff *skb);
365 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
366 struct nf_conntrack *nfct;
367 struct sk_buff *nfct_reasm;
369 #ifdef CONFIG_BRIDGE_NETFILTER
370 struct nf_bridge_info *nf_bridge;
374 #ifdef CONFIG_NET_SCHED
375 __u16 tc_index; /* traffic control index */
376 #ifdef CONFIG_NET_CLS_ACT
377 __u16 tc_verd; /* traffic control verdict */
383 kmemcheck_bitfield_begin(flags2);
384 __u16 queue_mapping:16;
385 #ifdef CONFIG_IPV6_NDISC_NODETYPE
386 __u8 ndisc_nodetype:2,
389 __u8 deliver_no_wcard:1;
391 kmemcheck_bitfield_end(flags2);
395 #ifdef CONFIG_NET_DMA
396 dma_cookie_t dma_cookie;
398 #ifdef CONFIG_NETWORK_SECMARK
408 sk_buff_data_t transport_header;
409 sk_buff_data_t network_header;
410 sk_buff_data_t mac_header;
411 /* These elements must be at the end, see alloc_skb() for details. */
416 unsigned int truesize;
422 * Handling routines are only of interest to the kernel
424 #include <linux/slab.h>
426 #include <asm/system.h>
429 * skb might have a dst pointer attached, refcounted or not.
430 * _skb_refdst low order bit is set if refcount was _not_ taken
432 #define SKB_DST_NOREF 1UL
433 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
436 * skb_dst - returns skb dst_entry
439 * Returns skb dst_entry, regardless of reference taken or not.
441 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
443 /* If refdst was not refcounted, check we still are in a
444 * rcu_read_lock section
446 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
447 !rcu_read_lock_held() &&
448 !rcu_read_lock_bh_held());
449 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
453 * skb_dst_set - sets skb dst
457 * Sets skb dst, assuming a reference was taken on dst and should
458 * be released by skb_dst_drop()
460 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
462 skb->_skb_refdst = (unsigned long)dst;
466 * skb_dst_set_noref - sets skb dst, without a reference
470 * Sets skb dst, assuming a reference was not taken on dst
471 * skb_dst_drop() should not dst_release() this dst
473 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
475 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
476 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
480 * skb_dst_is_noref - Test if skb dst isnt refcounted
483 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
485 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
488 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
490 return (struct rtable *)skb_dst(skb);
493 extern void kfree_skb(struct sk_buff *skb);
494 extern void consume_skb(struct sk_buff *skb);
495 extern void __kfree_skb(struct sk_buff *skb);
496 extern struct sk_buff *__alloc_skb(unsigned int size,
497 gfp_t priority, int fclone, int node);
498 static inline struct sk_buff *alloc_skb(unsigned int size,
501 return __alloc_skb(size, priority, 0, -1);
504 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
507 return __alloc_skb(size, priority, 1, -1);
510 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
512 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
513 extern struct sk_buff *skb_clone(struct sk_buff *skb,
515 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
517 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
519 extern int pskb_expand_head(struct sk_buff *skb,
520 int nhead, int ntail,
522 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
523 unsigned int headroom);
524 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
525 int newheadroom, int newtailroom,
527 extern int skb_to_sgvec(struct sk_buff *skb,
528 struct scatterlist *sg, int offset,
530 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
531 struct sk_buff **trailer);
532 extern int skb_pad(struct sk_buff *skb, int pad);
533 #define dev_kfree_skb(a) consume_skb(a)
535 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
536 int getfrag(void *from, char *to, int offset,
537 int len,int odd, struct sk_buff *skb),
538 void *from, int length);
540 struct skb_seq_state {
544 __u32 stepped_offset;
545 struct sk_buff *root_skb;
546 struct sk_buff *cur_skb;
550 extern void skb_prepare_seq_read(struct sk_buff *skb,
551 unsigned int from, unsigned int to,
552 struct skb_seq_state *st);
553 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
554 struct skb_seq_state *st);
555 extern void skb_abort_seq_read(struct skb_seq_state *st);
557 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
558 unsigned int to, struct ts_config *config,
559 struct ts_state *state);
561 #ifdef NET_SKBUFF_DATA_USES_OFFSET
562 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
564 return skb->head + skb->end;
567 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
574 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
576 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
578 return &skb_shinfo(skb)->hwtstamps;
581 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
583 return &skb_shinfo(skb)->tx_flags;
587 * skb_queue_empty - check if a queue is empty
590 * Returns true if the queue is empty, false otherwise.
592 static inline int skb_queue_empty(const struct sk_buff_head *list)
594 return list->next == (struct sk_buff *)list;
598 * skb_queue_is_last - check if skb is the last entry in the queue
602 * Returns true if @skb is the last buffer on the list.
604 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
605 const struct sk_buff *skb)
607 return (skb->next == (struct sk_buff *) list);
611 * skb_queue_is_first - check if skb is the first entry in the queue
615 * Returns true if @skb is the first buffer on the list.
617 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
618 const struct sk_buff *skb)
620 return (skb->prev == (struct sk_buff *) list);
624 * skb_queue_next - return the next packet in the queue
626 * @skb: current buffer
628 * Return the next packet in @list after @skb. It is only valid to
629 * call this if skb_queue_is_last() evaluates to false.
631 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
632 const struct sk_buff *skb)
634 /* This BUG_ON may seem severe, but if we just return then we
635 * are going to dereference garbage.
637 BUG_ON(skb_queue_is_last(list, skb));
642 * skb_queue_prev - return the prev packet in the queue
644 * @skb: current buffer
646 * Return the prev packet in @list before @skb. It is only valid to
647 * call this if skb_queue_is_first() evaluates to false.
649 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
650 const struct sk_buff *skb)
652 /* This BUG_ON may seem severe, but if we just return then we
653 * are going to dereference garbage.
655 BUG_ON(skb_queue_is_first(list, skb));
660 * skb_get - reference buffer
661 * @skb: buffer to reference
663 * Makes another reference to a socket buffer and returns a pointer
666 static inline struct sk_buff *skb_get(struct sk_buff *skb)
668 atomic_inc(&skb->users);
673 * If users == 1, we are the only owner and are can avoid redundant
678 * skb_cloned - is the buffer a clone
679 * @skb: buffer to check
681 * Returns true if the buffer was generated with skb_clone() and is
682 * one of multiple shared copies of the buffer. Cloned buffers are
683 * shared data so must not be written to under normal circumstances.
685 static inline int skb_cloned(const struct sk_buff *skb)
687 return skb->cloned &&
688 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
692 * skb_header_cloned - is the header a clone
693 * @skb: buffer to check
695 * Returns true if modifying the header part of the buffer requires
696 * the data to be copied.
698 static inline int skb_header_cloned(const struct sk_buff *skb)
705 dataref = atomic_read(&skb_shinfo(skb)->dataref);
706 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
711 * skb_header_release - release reference to header
712 * @skb: buffer to operate on
714 * Drop a reference to the header part of the buffer. This is done
715 * by acquiring a payload reference. You must not read from the header
716 * part of skb->data after this.
718 static inline void skb_header_release(struct sk_buff *skb)
722 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
726 * skb_shared - is the buffer shared
727 * @skb: buffer to check
729 * Returns true if more than one person has a reference to this
732 static inline int skb_shared(const struct sk_buff *skb)
734 return atomic_read(&skb->users) != 1;
738 * skb_share_check - check if buffer is shared and if so clone it
739 * @skb: buffer to check
740 * @pri: priority for memory allocation
742 * If the buffer is shared the buffer is cloned and the old copy
743 * drops a reference. A new clone with a single reference is returned.
744 * If the buffer is not shared the original buffer is returned. When
745 * being called from interrupt status or with spinlocks held pri must
748 * NULL is returned on a memory allocation failure.
750 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
753 might_sleep_if(pri & __GFP_WAIT);
754 if (skb_shared(skb)) {
755 struct sk_buff *nskb = skb_clone(skb, pri);
763 * Copy shared buffers into a new sk_buff. We effectively do COW on
764 * packets to handle cases where we have a local reader and forward
765 * and a couple of other messy ones. The normal one is tcpdumping
766 * a packet thats being forwarded.
770 * skb_unshare - make a copy of a shared buffer
771 * @skb: buffer to check
772 * @pri: priority for memory allocation
774 * If the socket buffer is a clone then this function creates a new
775 * copy of the data, drops a reference count on the old copy and returns
776 * the new copy with the reference count at 1. If the buffer is not a clone
777 * the original buffer is returned. When called with a spinlock held or
778 * from interrupt state @pri must be %GFP_ATOMIC
780 * %NULL is returned on a memory allocation failure.
782 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
785 might_sleep_if(pri & __GFP_WAIT);
786 if (skb_cloned(skb)) {
787 struct sk_buff *nskb = skb_copy(skb, pri);
788 kfree_skb(skb); /* Free our shared copy */
795 * skb_peek - peek at the head of an &sk_buff_head
796 * @list_: list to peek at
798 * Peek an &sk_buff. Unlike most other operations you _MUST_
799 * be careful with this one. A peek leaves the buffer on the
800 * list and someone else may run off with it. You must hold
801 * the appropriate locks or have a private queue to do this.
803 * Returns %NULL for an empty list or a pointer to the head element.
804 * The reference count is not incremented and the reference is therefore
805 * volatile. Use with caution.
807 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
809 struct sk_buff *list = ((struct sk_buff *)list_)->next;
810 if (list == (struct sk_buff *)list_)
816 * skb_peek_tail - peek at the tail of an &sk_buff_head
817 * @list_: list to peek at
819 * Peek an &sk_buff. Unlike most other operations you _MUST_
820 * be careful with this one. A peek leaves the buffer on the
821 * list and someone else may run off with it. You must hold
822 * the appropriate locks or have a private queue to do this.
824 * Returns %NULL for an empty list or a pointer to the tail element.
825 * The reference count is not incremented and the reference is therefore
826 * volatile. Use with caution.
828 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
830 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
831 if (list == (struct sk_buff *)list_)
837 * skb_queue_len - get queue length
838 * @list_: list to measure
840 * Return the length of an &sk_buff queue.
842 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
848 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
849 * @list: queue to initialize
851 * This initializes only the list and queue length aspects of
852 * an sk_buff_head object. This allows to initialize the list
853 * aspects of an sk_buff_head without reinitializing things like
854 * the spinlock. It can also be used for on-stack sk_buff_head
855 * objects where the spinlock is known to not be used.
857 static inline void __skb_queue_head_init(struct sk_buff_head *list)
859 list->prev = list->next = (struct sk_buff *)list;
864 * This function creates a split out lock class for each invocation;
865 * this is needed for now since a whole lot of users of the skb-queue
866 * infrastructure in drivers have different locking usage (in hardirq)
867 * than the networking core (in softirq only). In the long run either the
868 * network layer or drivers should need annotation to consolidate the
869 * main types of usage into 3 classes.
871 static inline void skb_queue_head_init(struct sk_buff_head *list)
873 spin_lock_init(&list->lock);
874 __skb_queue_head_init(list);
877 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
878 struct lock_class_key *class)
880 skb_queue_head_init(list);
881 lockdep_set_class(&list->lock, class);
885 * Insert an sk_buff on a list.
887 * The "__skb_xxxx()" functions are the non-atomic ones that
888 * can only be called with interrupts disabled.
890 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
891 static inline void __skb_insert(struct sk_buff *newsk,
892 struct sk_buff *prev, struct sk_buff *next,
893 struct sk_buff_head *list)
897 next->prev = prev->next = newsk;
901 static inline void __skb_queue_splice(const struct sk_buff_head *list,
902 struct sk_buff *prev,
903 struct sk_buff *next)
905 struct sk_buff *first = list->next;
906 struct sk_buff *last = list->prev;
916 * skb_queue_splice - join two skb lists, this is designed for stacks
917 * @list: the new list to add
918 * @head: the place to add it in the first list
920 static inline void skb_queue_splice(const struct sk_buff_head *list,
921 struct sk_buff_head *head)
923 if (!skb_queue_empty(list)) {
924 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
925 head->qlen += list->qlen;
930 * skb_queue_splice - join two skb lists and reinitialise the emptied list
931 * @list: the new list to add
932 * @head: the place to add it in the first list
934 * The list at @list is reinitialised
936 static inline void skb_queue_splice_init(struct sk_buff_head *list,
937 struct sk_buff_head *head)
939 if (!skb_queue_empty(list)) {
940 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
941 head->qlen += list->qlen;
942 __skb_queue_head_init(list);
947 * skb_queue_splice_tail - join two skb lists, each list being a queue
948 * @list: the new list to add
949 * @head: the place to add it in the first list
951 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
952 struct sk_buff_head *head)
954 if (!skb_queue_empty(list)) {
955 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
956 head->qlen += list->qlen;
961 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
962 * @list: the new list to add
963 * @head: the place to add it in the first list
965 * Each of the lists is a queue.
966 * The list at @list is reinitialised
968 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
969 struct sk_buff_head *head)
971 if (!skb_queue_empty(list)) {
972 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
973 head->qlen += list->qlen;
974 __skb_queue_head_init(list);
979 * __skb_queue_after - queue a buffer at the list head
981 * @prev: place after this buffer
982 * @newsk: buffer to queue
984 * Queue a buffer int the middle of a list. This function takes no locks
985 * and you must therefore hold required locks before calling it.
987 * A buffer cannot be placed on two lists at the same time.
989 static inline void __skb_queue_after(struct sk_buff_head *list,
990 struct sk_buff *prev,
991 struct sk_buff *newsk)
993 __skb_insert(newsk, prev, prev->next, list);
996 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
997 struct sk_buff_head *list);
999 static inline void __skb_queue_before(struct sk_buff_head *list,
1000 struct sk_buff *next,
1001 struct sk_buff *newsk)
1003 __skb_insert(newsk, next->prev, next, list);
1007 * __skb_queue_head - queue a buffer at the list head
1008 * @list: list to use
1009 * @newsk: buffer to queue
1011 * Queue a buffer at the start of a list. This function takes no locks
1012 * and you must therefore hold required locks before calling it.
1014 * A buffer cannot be placed on two lists at the same time.
1016 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1017 static inline void __skb_queue_head(struct sk_buff_head *list,
1018 struct sk_buff *newsk)
1020 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1024 * __skb_queue_tail - queue a buffer at the list tail
1025 * @list: list to use
1026 * @newsk: buffer to queue
1028 * Queue a buffer at the end of a list. This function takes no locks
1029 * and you must therefore hold required locks before calling it.
1031 * A buffer cannot be placed on two lists at the same time.
1033 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1034 static inline void __skb_queue_tail(struct sk_buff_head *list,
1035 struct sk_buff *newsk)
1037 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1041 * remove sk_buff from list. _Must_ be called atomically, and with
1044 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1045 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1047 struct sk_buff *next, *prev;
1052 skb->next = skb->prev = NULL;
1058 * __skb_dequeue - remove from the head of the queue
1059 * @list: list to dequeue from
1061 * Remove the head of the list. This function does not take any locks
1062 * so must be used with appropriate locks held only. The head item is
1063 * returned or %NULL if the list is empty.
1065 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1066 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1068 struct sk_buff *skb = skb_peek(list);
1070 __skb_unlink(skb, list);
1075 * __skb_dequeue_tail - remove from the tail of the queue
1076 * @list: list to dequeue from
1078 * Remove the tail of the list. This function does not take any locks
1079 * so must be used with appropriate locks held only. The tail item is
1080 * returned or %NULL if the list is empty.
1082 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1083 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1085 struct sk_buff *skb = skb_peek_tail(list);
1087 __skb_unlink(skb, list);
1092 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1094 return skb->data_len;
1097 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1099 return skb->len - skb->data_len;
1102 static inline int skb_pagelen(const struct sk_buff *skb)
1106 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1107 len += skb_shinfo(skb)->frags[i].size;
1108 return len + skb_headlen(skb);
1111 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1112 struct page *page, int off, int size)
1114 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1117 frag->page_offset = off;
1119 skb_shinfo(skb)->nr_frags = i + 1;
1122 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1125 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1126 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1127 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1129 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1130 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1132 return skb->head + skb->tail;
1135 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1137 skb->tail = skb->data - skb->head;
1140 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1142 skb_reset_tail_pointer(skb);
1143 skb->tail += offset;
1145 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1146 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1151 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1153 skb->tail = skb->data;
1156 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1158 skb->tail = skb->data + offset;
1161 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1164 * Add data to an sk_buff
1166 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1167 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1169 unsigned char *tmp = skb_tail_pointer(skb);
1170 SKB_LINEAR_ASSERT(skb);
1176 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1177 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1184 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1185 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1188 BUG_ON(skb->len < skb->data_len);
1189 return skb->data += len;
1192 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1194 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1197 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1199 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1201 if (len > skb_headlen(skb) &&
1202 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1205 return skb->data += len;
1208 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1210 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1213 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1215 if (likely(len <= skb_headlen(skb)))
1217 if (unlikely(len > skb->len))
1219 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1223 * skb_headroom - bytes at buffer head
1224 * @skb: buffer to check
1226 * Return the number of bytes of free space at the head of an &sk_buff.
1228 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1230 return skb->data - skb->head;
1234 * skb_tailroom - bytes at buffer end
1235 * @skb: buffer to check
1237 * Return the number of bytes of free space at the tail of an sk_buff
1239 static inline int skb_tailroom(const struct sk_buff *skb)
1241 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1245 * skb_reserve - adjust headroom
1246 * @skb: buffer to alter
1247 * @len: bytes to move
1249 * Increase the headroom of an empty &sk_buff by reducing the tail
1250 * room. This is only allowed for an empty buffer.
1252 static inline void skb_reserve(struct sk_buff *skb, int len)
1258 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1259 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1261 return skb->head + skb->transport_header;
1264 static inline void skb_reset_transport_header(struct sk_buff *skb)
1266 skb->transport_header = skb->data - skb->head;
1269 static inline void skb_set_transport_header(struct sk_buff *skb,
1272 skb_reset_transport_header(skb);
1273 skb->transport_header += offset;
1276 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1278 return skb->head + skb->network_header;
1281 static inline void skb_reset_network_header(struct sk_buff *skb)
1283 skb->network_header = skb->data - skb->head;
1286 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1288 skb_reset_network_header(skb);
1289 skb->network_header += offset;
1292 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1294 return skb->head + skb->mac_header;
1297 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1299 return skb->mac_header != ~0U;
1302 static inline void skb_reset_mac_header(struct sk_buff *skb)
1304 skb->mac_header = skb->data - skb->head;
1307 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1309 skb_reset_mac_header(skb);
1310 skb->mac_header += offset;
1313 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1315 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1317 return skb->transport_header;
1320 static inline void skb_reset_transport_header(struct sk_buff *skb)
1322 skb->transport_header = skb->data;
1325 static inline void skb_set_transport_header(struct sk_buff *skb,
1328 skb->transport_header = skb->data + offset;
1331 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1333 return skb->network_header;
1336 static inline void skb_reset_network_header(struct sk_buff *skb)
1338 skb->network_header = skb->data;
1341 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1343 skb->network_header = skb->data + offset;
1346 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1348 return skb->mac_header;
1351 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1353 return skb->mac_header != NULL;
1356 static inline void skb_reset_mac_header(struct sk_buff *skb)
1358 skb->mac_header = skb->data;
1361 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1363 skb->mac_header = skb->data + offset;
1365 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1367 static inline int skb_transport_offset(const struct sk_buff *skb)
1369 return skb_transport_header(skb) - skb->data;
1372 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1374 return skb->transport_header - skb->network_header;
1377 static inline int skb_network_offset(const struct sk_buff *skb)
1379 return skb_network_header(skb) - skb->data;
1382 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1384 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1388 * CPUs often take a performance hit when accessing unaligned memory
1389 * locations. The actual performance hit varies, it can be small if the
1390 * hardware handles it or large if we have to take an exception and fix it
1393 * Since an ethernet header is 14 bytes network drivers often end up with
1394 * the IP header at an unaligned offset. The IP header can be aligned by
1395 * shifting the start of the packet by 2 bytes. Drivers should do this
1398 * skb_reserve(skb, NET_IP_ALIGN);
1400 * The downside to this alignment of the IP header is that the DMA is now
1401 * unaligned. On some architectures the cost of an unaligned DMA is high
1402 * and this cost outweighs the gains made by aligning the IP header.
1404 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1407 #ifndef NET_IP_ALIGN
1408 #define NET_IP_ALIGN 2
1412 * The networking layer reserves some headroom in skb data (via
1413 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1414 * the header has to grow. In the default case, if the header has to grow
1415 * 32 bytes or less we avoid the reallocation.
1417 * Unfortunately this headroom changes the DMA alignment of the resulting
1418 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1419 * on some architectures. An architecture can override this value,
1420 * perhaps setting it to a cacheline in size (since that will maintain
1421 * cacheline alignment of the DMA). It must be a power of 2.
1423 * Various parts of the networking layer expect at least 32 bytes of
1424 * headroom, you should not reduce this.
1426 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1427 * to reduce average number of cache lines per packet.
1428 * get_rps_cpus() for example only access one 64 bytes aligned block :
1429 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1432 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1435 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1437 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1439 if (unlikely(skb->data_len)) {
1444 skb_set_tail_pointer(skb, len);
1447 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1449 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1452 return ___pskb_trim(skb, len);
1453 __skb_trim(skb, len);
1457 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1459 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1463 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1464 * @skb: buffer to alter
1467 * This is identical to pskb_trim except that the caller knows that
1468 * the skb is not cloned so we should never get an error due to out-
1471 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1473 int err = pskb_trim(skb, len);
1478 * skb_orphan - orphan a buffer
1479 * @skb: buffer to orphan
1481 * If a buffer currently has an owner then we call the owner's
1482 * destructor function and make the @skb unowned. The buffer continues
1483 * to exist but is no longer charged to its former owner.
1485 static inline void skb_orphan(struct sk_buff *skb)
1487 if (skb->destructor)
1488 skb->destructor(skb);
1489 skb->destructor = NULL;
1494 * __skb_queue_purge - empty a list
1495 * @list: list to empty
1497 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1498 * the list and one reference dropped. This function does not take the
1499 * list lock and the caller must hold the relevant locks to use it.
1501 extern void skb_queue_purge(struct sk_buff_head *list);
1502 static inline void __skb_queue_purge(struct sk_buff_head *list)
1504 struct sk_buff *skb;
1505 while ((skb = __skb_dequeue(list)) != NULL)
1510 * __dev_alloc_skb - allocate an skbuff for receiving
1511 * @length: length to allocate
1512 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1514 * Allocate a new &sk_buff and assign it a usage count of one. The
1515 * buffer has unspecified headroom built in. Users should allocate
1516 * the headroom they think they need without accounting for the
1517 * built in space. The built in space is used for optimisations.
1519 * %NULL is returned if there is no free memory.
1521 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1524 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1526 skb_reserve(skb, NET_SKB_PAD);
1530 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1532 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1533 unsigned int length, gfp_t gfp_mask);
1536 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1537 * @dev: network device to receive on
1538 * @length: length to allocate
1540 * Allocate a new &sk_buff and assign it a usage count of one. The
1541 * buffer has unspecified headroom built in. Users should allocate
1542 * the headroom they think they need without accounting for the
1543 * built in space. The built in space is used for optimisations.
1545 * %NULL is returned if there is no free memory. Although this function
1546 * allocates memory it can be called from an interrupt.
1548 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1549 unsigned int length)
1551 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1554 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1555 unsigned int length)
1557 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1559 if (NET_IP_ALIGN && skb)
1560 skb_reserve(skb, NET_IP_ALIGN);
1564 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1567 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1568 * @dev: network device to receive on
1570 * Allocate a new page node local to the specified device.
1572 * %NULL is returned if there is no free memory.
1574 static inline struct page *netdev_alloc_page(struct net_device *dev)
1576 return __netdev_alloc_page(dev, GFP_ATOMIC);
1579 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1585 * skb_clone_writable - is the header of a clone writable
1586 * @skb: buffer to check
1587 * @len: length up to which to write
1589 * Returns true if modifying the header part of the cloned buffer
1590 * does not requires the data to be copied.
1592 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1594 return !skb_header_cloned(skb) &&
1595 skb_headroom(skb) + len <= skb->hdr_len;
1598 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1603 if (headroom < NET_SKB_PAD)
1604 headroom = NET_SKB_PAD;
1605 if (headroom > skb_headroom(skb))
1606 delta = headroom - skb_headroom(skb);
1608 if (delta || cloned)
1609 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1615 * skb_cow - copy header of skb when it is required
1616 * @skb: buffer to cow
1617 * @headroom: needed headroom
1619 * If the skb passed lacks sufficient headroom or its data part
1620 * is shared, data is reallocated. If reallocation fails, an error
1621 * is returned and original skb is not changed.
1623 * The result is skb with writable area skb->head...skb->tail
1624 * and at least @headroom of space at head.
1626 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1628 return __skb_cow(skb, headroom, skb_cloned(skb));
1632 * skb_cow_head - skb_cow but only making the head writable
1633 * @skb: buffer to cow
1634 * @headroom: needed headroom
1636 * This function is identical to skb_cow except that we replace the
1637 * skb_cloned check by skb_header_cloned. It should be used when
1638 * you only need to push on some header and do not need to modify
1641 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1643 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1647 * skb_padto - pad an skbuff up to a minimal size
1648 * @skb: buffer to pad
1649 * @len: minimal length
1651 * Pads up a buffer to ensure the trailing bytes exist and are
1652 * blanked. If the buffer already contains sufficient data it
1653 * is untouched. Otherwise it is extended. Returns zero on
1654 * success. The skb is freed on error.
1657 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1659 unsigned int size = skb->len;
1660 if (likely(size >= len))
1662 return skb_pad(skb, len - size);
1665 static inline int skb_add_data(struct sk_buff *skb,
1666 char __user *from, int copy)
1668 const int off = skb->len;
1670 if (skb->ip_summed == CHECKSUM_NONE) {
1672 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1675 skb->csum = csum_block_add(skb->csum, csum, off);
1678 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1681 __skb_trim(skb, off);
1685 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1686 struct page *page, int off)
1689 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1691 return page == frag->page &&
1692 off == frag->page_offset + frag->size;
1697 static inline int __skb_linearize(struct sk_buff *skb)
1699 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1703 * skb_linearize - convert paged skb to linear one
1704 * @skb: buffer to linarize
1706 * If there is no free memory -ENOMEM is returned, otherwise zero
1707 * is returned and the old skb data released.
1709 static inline int skb_linearize(struct sk_buff *skb)
1711 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1715 * skb_linearize_cow - make sure skb is linear and writable
1716 * @skb: buffer to process
1718 * If there is no free memory -ENOMEM is returned, otherwise zero
1719 * is returned and the old skb data released.
1721 static inline int skb_linearize_cow(struct sk_buff *skb)
1723 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1724 __skb_linearize(skb) : 0;
1728 * skb_postpull_rcsum - update checksum for received skb after pull
1729 * @skb: buffer to update
1730 * @start: start of data before pull
1731 * @len: length of data pulled
1733 * After doing a pull on a received packet, you need to call this to
1734 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1735 * CHECKSUM_NONE so that it can be recomputed from scratch.
1738 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1739 const void *start, unsigned int len)
1741 if (skb->ip_summed == CHECKSUM_COMPLETE)
1742 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1745 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1748 * pskb_trim_rcsum - trim received skb and update checksum
1749 * @skb: buffer to trim
1752 * This is exactly the same as pskb_trim except that it ensures the
1753 * checksum of received packets are still valid after the operation.
1756 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1758 if (likely(len >= skb->len))
1760 if (skb->ip_summed == CHECKSUM_COMPLETE)
1761 skb->ip_summed = CHECKSUM_NONE;
1762 return __pskb_trim(skb, len);
1765 #define skb_queue_walk(queue, skb) \
1766 for (skb = (queue)->next; \
1767 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1770 #define skb_queue_walk_safe(queue, skb, tmp) \
1771 for (skb = (queue)->next, tmp = skb->next; \
1772 skb != (struct sk_buff *)(queue); \
1773 skb = tmp, tmp = skb->next)
1775 #define skb_queue_walk_from(queue, skb) \
1776 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1779 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1780 for (tmp = skb->next; \
1781 skb != (struct sk_buff *)(queue); \
1782 skb = tmp, tmp = skb->next)
1784 #define skb_queue_reverse_walk(queue, skb) \
1785 for (skb = (queue)->prev; \
1786 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1790 static inline bool skb_has_frags(const struct sk_buff *skb)
1792 return skb_shinfo(skb)->frag_list != NULL;
1795 static inline void skb_frag_list_init(struct sk_buff *skb)
1797 skb_shinfo(skb)->frag_list = NULL;
1800 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1802 frag->next = skb_shinfo(skb)->frag_list;
1803 skb_shinfo(skb)->frag_list = frag;
1806 #define skb_walk_frags(skb, iter) \
1807 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1809 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1810 int *peeked, int *err);
1811 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1812 int noblock, int *err);
1813 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1814 struct poll_table_struct *wait);
1815 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1816 int offset, struct iovec *to,
1818 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1821 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1823 const struct iovec *from,
1826 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1828 const struct iovec *to,
1831 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1832 extern void skb_free_datagram_locked(struct sock *sk,
1833 struct sk_buff *skb);
1834 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1835 unsigned int flags);
1836 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1837 int len, __wsum csum);
1838 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1840 extern int skb_store_bits(struct sk_buff *skb, int offset,
1841 const void *from, int len);
1842 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1843 int offset, u8 *to, int len,
1845 extern int skb_splice_bits(struct sk_buff *skb,
1846 unsigned int offset,
1847 struct pipe_inode_info *pipe,
1849 unsigned int flags);
1850 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1851 extern void skb_split(struct sk_buff *skb,
1852 struct sk_buff *skb1, const u32 len);
1853 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1856 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1858 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1859 int len, void *buffer)
1861 int hlen = skb_headlen(skb);
1863 if (hlen - offset >= len)
1864 return skb->data + offset;
1866 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1872 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1874 const unsigned int len)
1876 memcpy(to, skb->data, len);
1879 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1880 const int offset, void *to,
1881 const unsigned int len)
1883 memcpy(to, skb->data + offset, len);
1886 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1888 const unsigned int len)
1890 memcpy(skb->data, from, len);
1893 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1896 const unsigned int len)
1898 memcpy(skb->data + offset, from, len);
1901 extern void skb_init(void);
1903 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1909 * skb_get_timestamp - get timestamp from a skb
1910 * @skb: skb to get stamp from
1911 * @stamp: pointer to struct timeval to store stamp in
1913 * Timestamps are stored in the skb as offsets to a base timestamp.
1914 * This function converts the offset back to a struct timeval and stores
1917 static inline void skb_get_timestamp(const struct sk_buff *skb,
1918 struct timeval *stamp)
1920 *stamp = ktime_to_timeval(skb->tstamp);
1923 static inline void skb_get_timestampns(const struct sk_buff *skb,
1924 struct timespec *stamp)
1926 *stamp = ktime_to_timespec(skb->tstamp);
1929 static inline void __net_timestamp(struct sk_buff *skb)
1931 skb->tstamp = ktime_get_real();
1934 static inline ktime_t net_timedelta(ktime_t t)
1936 return ktime_sub(ktime_get_real(), t);
1939 static inline ktime_t net_invalid_timestamp(void)
1941 return ktime_set(0, 0);
1944 extern void skb_timestamping_init(void);
1946 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1948 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1949 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1951 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1953 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1957 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1962 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1965 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1967 * @skb: clone of the the original outgoing packet
1968 * @hwtstamps: hardware time stamps
1971 void skb_complete_tx_timestamp(struct sk_buff *skb,
1972 struct skb_shared_hwtstamps *hwtstamps);
1975 * skb_tstamp_tx - queue clone of skb with send time stamps
1976 * @orig_skb: the original outgoing packet
1977 * @hwtstamps: hardware time stamps, may be NULL if not available
1979 * If the skb has a socket associated, then this function clones the
1980 * skb (thus sharing the actual data and optional structures), stores
1981 * the optional hardware time stamping information (if non NULL) or
1982 * generates a software time stamp (otherwise), then queues the clone
1983 * to the error queue of the socket. Errors are silently ignored.
1985 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1986 struct skb_shared_hwtstamps *hwtstamps);
1988 static inline void sw_tx_timestamp(struct sk_buff *skb)
1990 union skb_shared_tx *shtx = skb_tx(skb);
1991 if (shtx->software && !shtx->in_progress)
1992 skb_tstamp_tx(skb, NULL);
1996 * skb_tx_timestamp() - Driver hook for transmit timestamping
1998 * Ethernet MAC Drivers should call this function in their hard_xmit()
1999 * function as soon as possible after giving the sk_buff to the MAC
2000 * hardware, but before freeing the sk_buff.
2002 * @skb: A socket buffer.
2004 static inline void skb_tx_timestamp(struct sk_buff *skb)
2006 skb_clone_tx_timestamp(skb);
2007 sw_tx_timestamp(skb);
2010 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2011 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2013 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2015 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2019 * skb_checksum_complete - Calculate checksum of an entire packet
2020 * @skb: packet to process
2022 * This function calculates the checksum over the entire packet plus
2023 * the value of skb->csum. The latter can be used to supply the
2024 * checksum of a pseudo header as used by TCP/UDP. It returns the
2027 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2028 * this function can be used to verify that checksum on received
2029 * packets. In that case the function should return zero if the
2030 * checksum is correct. In particular, this function will return zero
2031 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2032 * hardware has already verified the correctness of the checksum.
2034 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2036 return skb_csum_unnecessary(skb) ?
2037 0 : __skb_checksum_complete(skb);
2040 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2041 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2042 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2044 if (nfct && atomic_dec_and_test(&nfct->use))
2045 nf_conntrack_destroy(nfct);
2047 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2050 atomic_inc(&nfct->use);
2052 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2055 atomic_inc(&skb->users);
2057 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2063 #ifdef CONFIG_BRIDGE_NETFILTER
2064 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2066 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2069 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2072 atomic_inc(&nf_bridge->use);
2074 #endif /* CONFIG_BRIDGE_NETFILTER */
2075 static inline void nf_reset(struct sk_buff *skb)
2077 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2078 nf_conntrack_put(skb->nfct);
2080 nf_conntrack_put_reasm(skb->nfct_reasm);
2081 skb->nfct_reasm = NULL;
2083 #ifdef CONFIG_BRIDGE_NETFILTER
2084 nf_bridge_put(skb->nf_bridge);
2085 skb->nf_bridge = NULL;
2089 /* Note: This doesn't put any conntrack and bridge info in dst. */
2090 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2092 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2093 dst->nfct = src->nfct;
2094 nf_conntrack_get(src->nfct);
2095 dst->nfctinfo = src->nfctinfo;
2096 dst->nfct_reasm = src->nfct_reasm;
2097 nf_conntrack_get_reasm(src->nfct_reasm);
2099 #ifdef CONFIG_BRIDGE_NETFILTER
2100 dst->nf_bridge = src->nf_bridge;
2101 nf_bridge_get(src->nf_bridge);
2105 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2107 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2108 nf_conntrack_put(dst->nfct);
2109 nf_conntrack_put_reasm(dst->nfct_reasm);
2111 #ifdef CONFIG_BRIDGE_NETFILTER
2112 nf_bridge_put(dst->nf_bridge);
2114 __nf_copy(dst, src);
2117 #ifdef CONFIG_NETWORK_SECMARK
2118 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2120 to->secmark = from->secmark;
2123 static inline void skb_init_secmark(struct sk_buff *skb)
2128 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2131 static inline void skb_init_secmark(struct sk_buff *skb)
2135 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2137 skb->queue_mapping = queue_mapping;
2140 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2142 return skb->queue_mapping;
2145 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2147 to->queue_mapping = from->queue_mapping;
2150 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2152 skb->queue_mapping = rx_queue + 1;
2155 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2157 return skb->queue_mapping - 1;
2160 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2162 return (skb->queue_mapping != 0);
2165 extern u16 skb_tx_hash(const struct net_device *dev,
2166 const struct sk_buff *skb);
2169 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2174 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2180 static inline int skb_is_gso(const struct sk_buff *skb)
2182 return skb_shinfo(skb)->gso_size;
2185 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2187 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2190 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2192 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2194 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2195 * wanted then gso_type will be set. */
2196 struct skb_shared_info *shinfo = skb_shinfo(skb);
2197 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2198 unlikely(shinfo->gso_type == 0)) {
2199 __skb_warn_lro_forwarding(skb);
2205 static inline void skb_forward_csum(struct sk_buff *skb)
2207 /* Unfortunately we don't support this one. Any brave souls? */
2208 if (skb->ip_summed == CHECKSUM_COMPLETE)
2209 skb->ip_summed = CHECKSUM_NONE;
2212 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2213 #endif /* __KERNEL__ */
2214 #endif /* _LINUX_SKBUFF_H */