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/prefetch.h>
32 #include <linux/hrtimer.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,
193 /* This data is invariant across clones and lives at
194 * the end of the header data, ie. at skb->end.
196 struct skb_shared_info {
197 unsigned short nr_frags;
198 unsigned short gso_size;
199 /* Warning: this field is not always filled in (UFO)! */
200 unsigned short gso_segs;
201 unsigned short gso_type;
204 struct sk_buff *frag_list;
205 struct skb_shared_hwtstamps hwtstamps;
208 * Warning : all fields before dataref are cleared in __alloc_skb()
212 /* Intermediate layers must ensure that destructor_arg
213 * remains valid until skb destructor */
214 void * destructor_arg;
215 /* must be last field, see pskb_expand_head() */
216 skb_frag_t frags[MAX_SKB_FRAGS];
219 /* We divide dataref into two halves. The higher 16 bits hold references
220 * to the payload part of skb->data. The lower 16 bits hold references to
221 * the entire skb->data. A clone of a headerless skb holds the length of
222 * the header in skb->hdr_len.
224 * All users must obey the rule that the skb->data reference count must be
225 * greater than or equal to the payload reference count.
227 * Holding a reference to the payload part means that the user does not
228 * care about modifications to the header part of skb->data.
230 #define SKB_DATAREF_SHIFT 16
231 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
235 SKB_FCLONE_UNAVAILABLE,
241 SKB_GSO_TCPV4 = 1 << 0,
242 SKB_GSO_UDP = 1 << 1,
244 /* This indicates the skb is from an untrusted source. */
245 SKB_GSO_DODGY = 1 << 2,
247 /* This indicates the tcp segment has CWR set. */
248 SKB_GSO_TCP_ECN = 1 << 3,
250 SKB_GSO_TCPV6 = 1 << 4,
252 SKB_GSO_FCOE = 1 << 5,
255 #if BITS_PER_LONG > 32
256 #define NET_SKBUFF_DATA_USES_OFFSET 1
259 #ifdef NET_SKBUFF_DATA_USES_OFFSET
260 typedef unsigned int sk_buff_data_t;
262 typedef unsigned char *sk_buff_data_t;
265 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
266 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
267 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
271 * struct sk_buff - socket buffer
272 * @next: Next buffer in list
273 * @prev: Previous buffer in list
274 * @sk: Socket we are owned by
275 * @tstamp: Time we arrived
276 * @dev: Device we arrived on/are leaving by
277 * @transport_header: Transport layer header
278 * @network_header: Network layer header
279 * @mac_header: Link layer header
280 * @_skb_refdst: destination entry (with norefcount bit)
281 * @sp: the security path, used for xfrm
282 * @cb: Control buffer. Free for use by every layer. Put private vars here
283 * @len: Length of actual data
284 * @data_len: Data length
285 * @mac_len: Length of link layer header
286 * @hdr_len: writable header length of cloned skb
287 * @csum: Checksum (must include start/offset pair)
288 * @csum_start: Offset from skb->head where checksumming should start
289 * @csum_offset: Offset from csum_start where checksum should be stored
290 * @local_df: allow local fragmentation
291 * @cloned: Head may be cloned (check refcnt to be sure)
292 * @nohdr: Payload reference only, must not modify header
293 * @pkt_type: Packet class
294 * @fclone: skbuff clone status
295 * @ip_summed: Driver fed us an IP checksum
296 * @priority: Packet queueing priority
297 * @users: User count - see {datagram,tcp}.c
298 * @protocol: Packet protocol from driver
299 * @truesize: Buffer size
300 * @head: Head of buffer
301 * @data: Data head pointer
302 * @tail: Tail pointer
304 * @destructor: Destruct function
305 * @mark: Generic packet mark
306 * @nfct: Associated connection, if any
307 * @ipvs_property: skbuff is owned by ipvs
308 * @peeked: this packet has been seen already, so stats have been
309 * done for it, don't do them again
310 * @nf_trace: netfilter packet trace flag
311 * @nfctinfo: Relationship of this skb to the connection
312 * @nfct_reasm: netfilter conntrack re-assembly pointer
313 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
314 * @skb_iif: ifindex of device we arrived on
315 * @rxhash: the packet hash computed on receive
316 * @queue_mapping: Queue mapping for multiqueue devices
317 * @tc_index: Traffic control index
318 * @tc_verd: traffic control verdict
319 * @ndisc_nodetype: router type (from link layer)
320 * @dma_cookie: a cookie to one of several possible DMA operations
321 * done by skb DMA functions
322 * @secmark: security marking
323 * @vlan_tci: vlan tag control information
327 /* These two members must be first. */
328 struct sk_buff *next;
329 struct sk_buff *prev;
334 struct net_device *dev;
337 * This is the control buffer. It is free to use for every
338 * layer. Please put your private variables there. If you
339 * want to keep them across layers you have to do a skb_clone()
340 * first. This is owned by whoever has the skb queued ATM.
342 char cb[48] __aligned(8);
344 unsigned long _skb_refdst;
360 kmemcheck_bitfield_begin(flags1);
371 kmemcheck_bitfield_end(flags1);
374 void (*destructor)(struct sk_buff *skb);
375 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
376 struct nf_conntrack *nfct;
378 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
379 struct sk_buff *nfct_reasm;
381 #ifdef CONFIG_BRIDGE_NETFILTER
382 struct nf_bridge_info *nf_bridge;
386 #ifdef CONFIG_NET_SCHED
387 __u16 tc_index; /* traffic control index */
388 #ifdef CONFIG_NET_CLS_ACT
389 __u16 tc_verd; /* traffic control verdict */
396 kmemcheck_bitfield_begin(flags2);
397 #ifdef CONFIG_IPV6_NDISC_NODETYPE
398 __u8 ndisc_nodetype:2;
401 kmemcheck_bitfield_end(flags2);
405 #ifdef CONFIG_NET_DMA
406 dma_cookie_t dma_cookie;
408 #ifdef CONFIG_NETWORK_SECMARK
418 sk_buff_data_t transport_header;
419 sk_buff_data_t network_header;
420 sk_buff_data_t mac_header;
421 /* These elements must be at the end, see alloc_skb() for details. */
426 unsigned int truesize;
432 * Handling routines are only of interest to the kernel
434 #include <linux/slab.h>
436 #include <asm/system.h>
439 * skb might have a dst pointer attached, refcounted or not.
440 * _skb_refdst low order bit is set if refcount was _not_ taken
442 #define SKB_DST_NOREF 1UL
443 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
446 * skb_dst - returns skb dst_entry
449 * Returns skb dst_entry, regardless of reference taken or not.
451 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
453 /* If refdst was not refcounted, check we still are in a
454 * rcu_read_lock section
456 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
457 !rcu_read_lock_held() &&
458 !rcu_read_lock_bh_held());
459 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
463 * skb_dst_set - sets skb dst
467 * Sets skb dst, assuming a reference was taken on dst and should
468 * be released by skb_dst_drop()
470 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
472 skb->_skb_refdst = (unsigned long)dst;
475 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
478 * skb_dst_is_noref - Test if skb dst isn't refcounted
481 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
483 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
486 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
488 return (struct rtable *)skb_dst(skb);
491 extern void kfree_skb(struct sk_buff *skb);
492 extern void consume_skb(struct sk_buff *skb);
493 extern void __kfree_skb(struct sk_buff *skb);
494 extern struct sk_buff *__alloc_skb(unsigned int size,
495 gfp_t priority, int fclone, int node);
496 static inline struct sk_buff *alloc_skb(unsigned int size,
499 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
502 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
505 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
508 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
510 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
511 extern struct sk_buff *skb_clone(struct sk_buff *skb,
513 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
515 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
517 extern int pskb_expand_head(struct sk_buff *skb,
518 int nhead, int ntail,
520 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
521 unsigned int headroom);
522 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
523 int newheadroom, int newtailroom,
525 extern int skb_to_sgvec(struct sk_buff *skb,
526 struct scatterlist *sg, int offset,
528 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
529 struct sk_buff **trailer);
530 extern int skb_pad(struct sk_buff *skb, int pad);
531 #define dev_kfree_skb(a) consume_skb(a)
533 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
534 int getfrag(void *from, char *to, int offset,
535 int len,int odd, struct sk_buff *skb),
536 void *from, int length);
538 struct skb_seq_state {
542 __u32 stepped_offset;
543 struct sk_buff *root_skb;
544 struct sk_buff *cur_skb;
548 extern void skb_prepare_seq_read(struct sk_buff *skb,
549 unsigned int from, unsigned int to,
550 struct skb_seq_state *st);
551 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
552 struct skb_seq_state *st);
553 extern void skb_abort_seq_read(struct skb_seq_state *st);
555 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
556 unsigned int to, struct ts_config *config,
557 struct ts_state *state);
559 extern __u32 __skb_get_rxhash(struct sk_buff *skb);
560 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
563 skb->rxhash = __skb_get_rxhash(skb);
568 #ifdef NET_SKBUFF_DATA_USES_OFFSET
569 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
571 return skb->head + skb->end;
574 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
581 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
583 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
585 return &skb_shinfo(skb)->hwtstamps;
589 * skb_queue_empty - check if a queue is empty
592 * Returns true if the queue is empty, false otherwise.
594 static inline int skb_queue_empty(const struct sk_buff_head *list)
596 return list->next == (struct sk_buff *)list;
600 * skb_queue_is_last - check if skb is the last entry in the queue
604 * Returns true if @skb is the last buffer on the list.
606 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
607 const struct sk_buff *skb)
609 return skb->next == (struct sk_buff *)list;
613 * skb_queue_is_first - check if skb is the first entry in the queue
617 * Returns true if @skb is the first buffer on the list.
619 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
620 const struct sk_buff *skb)
622 return skb->prev == (struct sk_buff *)list;
626 * skb_queue_next - return the next packet in the queue
628 * @skb: current buffer
630 * Return the next packet in @list after @skb. It is only valid to
631 * call this if skb_queue_is_last() evaluates to false.
633 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
634 const struct sk_buff *skb)
636 /* This BUG_ON may seem severe, but if we just return then we
637 * are going to dereference garbage.
639 BUG_ON(skb_queue_is_last(list, skb));
644 * skb_queue_prev - return the prev packet in the queue
646 * @skb: current buffer
648 * Return the prev packet in @list before @skb. It is only valid to
649 * call this if skb_queue_is_first() evaluates to false.
651 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
652 const struct sk_buff *skb)
654 /* This BUG_ON may seem severe, but if we just return then we
655 * are going to dereference garbage.
657 BUG_ON(skb_queue_is_first(list, skb));
662 * skb_get - reference buffer
663 * @skb: buffer to reference
665 * Makes another reference to a socket buffer and returns a pointer
668 static inline struct sk_buff *skb_get(struct sk_buff *skb)
670 atomic_inc(&skb->users);
675 * If users == 1, we are the only owner and are can avoid redundant
680 * skb_cloned - is the buffer a clone
681 * @skb: buffer to check
683 * Returns true if the buffer was generated with skb_clone() and is
684 * one of multiple shared copies of the buffer. Cloned buffers are
685 * shared data so must not be written to under normal circumstances.
687 static inline int skb_cloned(const struct sk_buff *skb)
689 return skb->cloned &&
690 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
694 * skb_header_cloned - is the header a clone
695 * @skb: buffer to check
697 * Returns true if modifying the header part of the buffer requires
698 * the data to be copied.
700 static inline int skb_header_cloned(const struct sk_buff *skb)
707 dataref = atomic_read(&skb_shinfo(skb)->dataref);
708 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
713 * skb_header_release - release reference to header
714 * @skb: buffer to operate on
716 * Drop a reference to the header part of the buffer. This is done
717 * by acquiring a payload reference. You must not read from the header
718 * part of skb->data after this.
720 static inline void skb_header_release(struct sk_buff *skb)
724 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
728 * skb_shared - is the buffer shared
729 * @skb: buffer to check
731 * Returns true if more than one person has a reference to this
734 static inline int skb_shared(const struct sk_buff *skb)
736 return atomic_read(&skb->users) != 1;
740 * skb_share_check - check if buffer is shared and if so clone it
741 * @skb: buffer to check
742 * @pri: priority for memory allocation
744 * If the buffer is shared the buffer is cloned and the old copy
745 * drops a reference. A new clone with a single reference is returned.
746 * If the buffer is not shared the original buffer is returned. When
747 * being called from interrupt status or with spinlocks held pri must
750 * NULL is returned on a memory allocation failure.
752 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
755 might_sleep_if(pri & __GFP_WAIT);
756 if (skb_shared(skb)) {
757 struct sk_buff *nskb = skb_clone(skb, pri);
765 * Copy shared buffers into a new sk_buff. We effectively do COW on
766 * packets to handle cases where we have a local reader and forward
767 * and a couple of other messy ones. The normal one is tcpdumping
768 * a packet thats being forwarded.
772 * skb_unshare - make a copy of a shared buffer
773 * @skb: buffer to check
774 * @pri: priority for memory allocation
776 * If the socket buffer is a clone then this function creates a new
777 * copy of the data, drops a reference count on the old copy and returns
778 * the new copy with the reference count at 1. If the buffer is not a clone
779 * the original buffer is returned. When called with a spinlock held or
780 * from interrupt state @pri must be %GFP_ATOMIC
782 * %NULL is returned on a memory allocation failure.
784 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
787 might_sleep_if(pri & __GFP_WAIT);
788 if (skb_cloned(skb)) {
789 struct sk_buff *nskb = skb_copy(skb, pri);
790 kfree_skb(skb); /* Free our shared copy */
797 * skb_peek - peek at the head of an &sk_buff_head
798 * @list_: list to peek at
800 * Peek an &sk_buff. Unlike most other operations you _MUST_
801 * be careful with this one. A peek leaves the buffer on the
802 * list and someone else may run off with it. You must hold
803 * the appropriate locks or have a private queue to do this.
805 * Returns %NULL for an empty list or a pointer to the head element.
806 * The reference count is not incremented and the reference is therefore
807 * volatile. Use with caution.
809 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
811 struct sk_buff *list = ((struct sk_buff *)list_)->next;
812 if (list == (struct sk_buff *)list_)
818 * skb_peek_tail - peek at the tail 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 tail 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_tail(struct sk_buff_head *list_)
832 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
833 if (list == (struct sk_buff *)list_)
839 * skb_queue_len - get queue length
840 * @list_: list to measure
842 * Return the length of an &sk_buff queue.
844 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
850 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
851 * @list: queue to initialize
853 * This initializes only the list and queue length aspects of
854 * an sk_buff_head object. This allows to initialize the list
855 * aspects of an sk_buff_head without reinitializing things like
856 * the spinlock. It can also be used for on-stack sk_buff_head
857 * objects where the spinlock is known to not be used.
859 static inline void __skb_queue_head_init(struct sk_buff_head *list)
861 list->prev = list->next = (struct sk_buff *)list;
866 * This function creates a split out lock class for each invocation;
867 * this is needed for now since a whole lot of users of the skb-queue
868 * infrastructure in drivers have different locking usage (in hardirq)
869 * than the networking core (in softirq only). In the long run either the
870 * network layer or drivers should need annotation to consolidate the
871 * main types of usage into 3 classes.
873 static inline void skb_queue_head_init(struct sk_buff_head *list)
875 spin_lock_init(&list->lock);
876 __skb_queue_head_init(list);
879 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
880 struct lock_class_key *class)
882 skb_queue_head_init(list);
883 lockdep_set_class(&list->lock, class);
887 * Insert an sk_buff on a list.
889 * The "__skb_xxxx()" functions are the non-atomic ones that
890 * can only be called with interrupts disabled.
892 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
893 static inline void __skb_insert(struct sk_buff *newsk,
894 struct sk_buff *prev, struct sk_buff *next,
895 struct sk_buff_head *list)
899 next->prev = prev->next = newsk;
903 static inline void __skb_queue_splice(const struct sk_buff_head *list,
904 struct sk_buff *prev,
905 struct sk_buff *next)
907 struct sk_buff *first = list->next;
908 struct sk_buff *last = list->prev;
918 * skb_queue_splice - join two skb lists, this is designed for stacks
919 * @list: the new list to add
920 * @head: the place to add it in the first list
922 static inline void skb_queue_splice(const struct sk_buff_head *list,
923 struct sk_buff_head *head)
925 if (!skb_queue_empty(list)) {
926 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
927 head->qlen += list->qlen;
932 * skb_queue_splice - join two skb lists and reinitialise the emptied list
933 * @list: the new list to add
934 * @head: the place to add it in the first list
936 * The list at @list is reinitialised
938 static inline void skb_queue_splice_init(struct sk_buff_head *list,
939 struct sk_buff_head *head)
941 if (!skb_queue_empty(list)) {
942 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
943 head->qlen += list->qlen;
944 __skb_queue_head_init(list);
949 * skb_queue_splice_tail - join two skb lists, each list being a queue
950 * @list: the new list to add
951 * @head: the place to add it in the first list
953 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
954 struct sk_buff_head *head)
956 if (!skb_queue_empty(list)) {
957 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
958 head->qlen += list->qlen;
963 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
964 * @list: the new list to add
965 * @head: the place to add it in the first list
967 * Each of the lists is a queue.
968 * The list at @list is reinitialised
970 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
971 struct sk_buff_head *head)
973 if (!skb_queue_empty(list)) {
974 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
975 head->qlen += list->qlen;
976 __skb_queue_head_init(list);
981 * __skb_queue_after - queue a buffer at the list head
983 * @prev: place after this buffer
984 * @newsk: buffer to queue
986 * Queue a buffer int the middle of a list. This function takes no locks
987 * and you must therefore hold required locks before calling it.
989 * A buffer cannot be placed on two lists at the same time.
991 static inline void __skb_queue_after(struct sk_buff_head *list,
992 struct sk_buff *prev,
993 struct sk_buff *newsk)
995 __skb_insert(newsk, prev, prev->next, list);
998 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
999 struct sk_buff_head *list);
1001 static inline void __skb_queue_before(struct sk_buff_head *list,
1002 struct sk_buff *next,
1003 struct sk_buff *newsk)
1005 __skb_insert(newsk, next->prev, next, list);
1009 * __skb_queue_head - queue a buffer at the list head
1010 * @list: list to use
1011 * @newsk: buffer to queue
1013 * Queue a buffer at the start of a list. This function takes no locks
1014 * and you must therefore hold required locks before calling it.
1016 * A buffer cannot be placed on two lists at the same time.
1018 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1019 static inline void __skb_queue_head(struct sk_buff_head *list,
1020 struct sk_buff *newsk)
1022 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1026 * __skb_queue_tail - queue a buffer at the list tail
1027 * @list: list to use
1028 * @newsk: buffer to queue
1030 * Queue a buffer at the end of a list. This function takes no locks
1031 * and you must therefore hold required locks before calling it.
1033 * A buffer cannot be placed on two lists at the same time.
1035 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1036 static inline void __skb_queue_tail(struct sk_buff_head *list,
1037 struct sk_buff *newsk)
1039 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1043 * remove sk_buff from list. _Must_ be called atomically, and with
1046 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1047 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1049 struct sk_buff *next, *prev;
1054 skb->next = skb->prev = NULL;
1060 * __skb_dequeue - remove from the head of the queue
1061 * @list: list to dequeue from
1063 * Remove the head of the list. This function does not take any locks
1064 * so must be used with appropriate locks held only. The head item is
1065 * returned or %NULL if the list is empty.
1067 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1068 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1070 struct sk_buff *skb = skb_peek(list);
1072 __skb_unlink(skb, list);
1077 * __skb_dequeue_tail - remove from the tail of the queue
1078 * @list: list to dequeue from
1080 * Remove the tail of the list. This function does not take any locks
1081 * so must be used with appropriate locks held only. The tail item is
1082 * returned or %NULL if the list is empty.
1084 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1085 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1087 struct sk_buff *skb = skb_peek_tail(list);
1089 __skb_unlink(skb, list);
1094 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1096 return skb->data_len;
1099 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1101 return skb->len - skb->data_len;
1104 static inline int skb_pagelen(const struct sk_buff *skb)
1108 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1109 len += skb_shinfo(skb)->frags[i].size;
1110 return len + skb_headlen(skb);
1113 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1114 struct page *page, int off, int size)
1116 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1119 frag->page_offset = off;
1121 skb_shinfo(skb)->nr_frags = i + 1;
1124 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1127 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1128 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1129 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1131 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1132 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1134 return skb->head + skb->tail;
1137 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1139 skb->tail = skb->data - skb->head;
1142 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1144 skb_reset_tail_pointer(skb);
1145 skb->tail += offset;
1147 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1148 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1153 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1155 skb->tail = skb->data;
1158 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1160 skb->tail = skb->data + offset;
1163 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1166 * Add data to an sk_buff
1168 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1169 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1171 unsigned char *tmp = skb_tail_pointer(skb);
1172 SKB_LINEAR_ASSERT(skb);
1178 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1179 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1186 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1187 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1190 BUG_ON(skb->len < skb->data_len);
1191 return skb->data += len;
1194 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1196 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1199 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1201 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1203 if (len > skb_headlen(skb) &&
1204 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1207 return skb->data += len;
1210 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1212 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1215 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1217 if (likely(len <= skb_headlen(skb)))
1219 if (unlikely(len > skb->len))
1221 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1225 * skb_headroom - bytes at buffer head
1226 * @skb: buffer to check
1228 * Return the number of bytes of free space at the head of an &sk_buff.
1230 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1232 return skb->data - skb->head;
1236 * skb_tailroom - bytes at buffer end
1237 * @skb: buffer to check
1239 * Return the number of bytes of free space at the tail of an sk_buff
1241 static inline int skb_tailroom(const struct sk_buff *skb)
1243 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1247 * skb_reserve - adjust headroom
1248 * @skb: buffer to alter
1249 * @len: bytes to move
1251 * Increase the headroom of an empty &sk_buff by reducing the tail
1252 * room. This is only allowed for an empty buffer.
1254 static inline void skb_reserve(struct sk_buff *skb, int len)
1260 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1261 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1263 return skb->head + skb->transport_header;
1266 static inline void skb_reset_transport_header(struct sk_buff *skb)
1268 skb->transport_header = skb->data - skb->head;
1271 static inline void skb_set_transport_header(struct sk_buff *skb,
1274 skb_reset_transport_header(skb);
1275 skb->transport_header += offset;
1278 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1280 return skb->head + skb->network_header;
1283 static inline void skb_reset_network_header(struct sk_buff *skb)
1285 skb->network_header = skb->data - skb->head;
1288 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1290 skb_reset_network_header(skb);
1291 skb->network_header += offset;
1294 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1296 return skb->head + skb->mac_header;
1299 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1301 return skb->mac_header != ~0U;
1304 static inline void skb_reset_mac_header(struct sk_buff *skb)
1306 skb->mac_header = skb->data - skb->head;
1309 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1311 skb_reset_mac_header(skb);
1312 skb->mac_header += offset;
1315 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1317 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1319 return skb->transport_header;
1322 static inline void skb_reset_transport_header(struct sk_buff *skb)
1324 skb->transport_header = skb->data;
1327 static inline void skb_set_transport_header(struct sk_buff *skb,
1330 skb->transport_header = skb->data + offset;
1333 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1335 return skb->network_header;
1338 static inline void skb_reset_network_header(struct sk_buff *skb)
1340 skb->network_header = skb->data;
1343 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1345 skb->network_header = skb->data + offset;
1348 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1350 return skb->mac_header;
1353 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1355 return skb->mac_header != NULL;
1358 static inline void skb_reset_mac_header(struct sk_buff *skb)
1360 skb->mac_header = skb->data;
1363 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1365 skb->mac_header = skb->data + offset;
1367 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1369 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1371 return skb->csum_start - skb_headroom(skb);
1374 static inline int skb_transport_offset(const struct sk_buff *skb)
1376 return skb_transport_header(skb) - skb->data;
1379 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1381 return skb->transport_header - skb->network_header;
1384 static inline int skb_network_offset(const struct sk_buff *skb)
1386 return skb_network_header(skb) - skb->data;
1389 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1391 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1395 * CPUs often take a performance hit when accessing unaligned memory
1396 * locations. The actual performance hit varies, it can be small if the
1397 * hardware handles it or large if we have to take an exception and fix it
1400 * Since an ethernet header is 14 bytes network drivers often end up with
1401 * the IP header at an unaligned offset. The IP header can be aligned by
1402 * shifting the start of the packet by 2 bytes. Drivers should do this
1405 * skb_reserve(skb, NET_IP_ALIGN);
1407 * The downside to this alignment of the IP header is that the DMA is now
1408 * unaligned. On some architectures the cost of an unaligned DMA is high
1409 * and this cost outweighs the gains made by aligning the IP header.
1411 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1414 #ifndef NET_IP_ALIGN
1415 #define NET_IP_ALIGN 2
1419 * The networking layer reserves some headroom in skb data (via
1420 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1421 * the header has to grow. In the default case, if the header has to grow
1422 * 32 bytes or less we avoid the reallocation.
1424 * Unfortunately this headroom changes the DMA alignment of the resulting
1425 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1426 * on some architectures. An architecture can override this value,
1427 * perhaps setting it to a cacheline in size (since that will maintain
1428 * cacheline alignment of the DMA). It must be a power of 2.
1430 * Various parts of the networking layer expect at least 32 bytes of
1431 * headroom, you should not reduce this.
1433 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1434 * to reduce average number of cache lines per packet.
1435 * get_rps_cpus() for example only access one 64 bytes aligned block :
1436 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1439 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1442 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1444 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1446 if (unlikely(skb->data_len)) {
1451 skb_set_tail_pointer(skb, len);
1454 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1456 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1459 return ___pskb_trim(skb, len);
1460 __skb_trim(skb, len);
1464 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1466 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1470 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1471 * @skb: buffer to alter
1474 * This is identical to pskb_trim except that the caller knows that
1475 * the skb is not cloned so we should never get an error due to out-
1478 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1480 int err = pskb_trim(skb, len);
1485 * skb_orphan - orphan a buffer
1486 * @skb: buffer to orphan
1488 * If a buffer currently has an owner then we call the owner's
1489 * destructor function and make the @skb unowned. The buffer continues
1490 * to exist but is no longer charged to its former owner.
1492 static inline void skb_orphan(struct sk_buff *skb)
1494 if (skb->destructor)
1495 skb->destructor(skb);
1496 skb->destructor = NULL;
1501 * __skb_queue_purge - empty a list
1502 * @list: list to empty
1504 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1505 * the list and one reference dropped. This function does not take the
1506 * list lock and the caller must hold the relevant locks to use it.
1508 extern void skb_queue_purge(struct sk_buff_head *list);
1509 static inline void __skb_queue_purge(struct sk_buff_head *list)
1511 struct sk_buff *skb;
1512 while ((skb = __skb_dequeue(list)) != NULL)
1517 * __dev_alloc_skb - allocate an skbuff for receiving
1518 * @length: length to allocate
1519 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1521 * Allocate a new &sk_buff and assign it a usage count of one. The
1522 * buffer has unspecified headroom built in. Users should allocate
1523 * the headroom they think they need without accounting for the
1524 * built in space. The built in space is used for optimisations.
1526 * %NULL is returned if there is no free memory.
1528 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1531 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1533 skb_reserve(skb, NET_SKB_PAD);
1537 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1539 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1540 unsigned int length, gfp_t gfp_mask);
1543 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1544 * @dev: network device to receive on
1545 * @length: length to allocate
1547 * Allocate a new &sk_buff and assign it a usage count of one. The
1548 * buffer has unspecified headroom built in. Users should allocate
1549 * the headroom they think they need without accounting for the
1550 * built in space. The built in space is used for optimisations.
1552 * %NULL is returned if there is no free memory. Although this function
1553 * allocates memory it can be called from an interrupt.
1555 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1556 unsigned int length)
1558 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1561 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1562 unsigned int length)
1564 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1566 if (NET_IP_ALIGN && skb)
1567 skb_reserve(skb, NET_IP_ALIGN);
1572 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1573 * @dev: network device to receive on
1574 * @gfp_mask: alloc_pages_node mask
1576 * Allocate a new page. dev currently unused.
1578 * %NULL is returned if there is no free memory.
1580 static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1582 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1586 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1587 * @dev: network device to receive on
1589 * Allocate a new page. dev currently unused.
1591 * %NULL is returned if there is no free memory.
1593 static inline struct page *netdev_alloc_page(struct net_device *dev)
1595 return __netdev_alloc_page(dev, GFP_ATOMIC);
1598 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1604 * skb_clone_writable - is the header of a clone writable
1605 * @skb: buffer to check
1606 * @len: length up to which to write
1608 * Returns true if modifying the header part of the cloned buffer
1609 * does not requires the data to be copied.
1611 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1613 return !skb_header_cloned(skb) &&
1614 skb_headroom(skb) + len <= skb->hdr_len;
1617 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1622 if (headroom < NET_SKB_PAD)
1623 headroom = NET_SKB_PAD;
1624 if (headroom > skb_headroom(skb))
1625 delta = headroom - skb_headroom(skb);
1627 if (delta || cloned)
1628 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1634 * skb_cow - copy header of skb when it is required
1635 * @skb: buffer to cow
1636 * @headroom: needed headroom
1638 * If the skb passed lacks sufficient headroom or its data part
1639 * is shared, data is reallocated. If reallocation fails, an error
1640 * is returned and original skb is not changed.
1642 * The result is skb with writable area skb->head...skb->tail
1643 * and at least @headroom of space at head.
1645 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1647 return __skb_cow(skb, headroom, skb_cloned(skb));
1651 * skb_cow_head - skb_cow but only making the head writable
1652 * @skb: buffer to cow
1653 * @headroom: needed headroom
1655 * This function is identical to skb_cow except that we replace the
1656 * skb_cloned check by skb_header_cloned. It should be used when
1657 * you only need to push on some header and do not need to modify
1660 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1662 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1666 * skb_padto - pad an skbuff up to a minimal size
1667 * @skb: buffer to pad
1668 * @len: minimal length
1670 * Pads up a buffer to ensure the trailing bytes exist and are
1671 * blanked. If the buffer already contains sufficient data it
1672 * is untouched. Otherwise it is extended. Returns zero on
1673 * success. The skb is freed on error.
1676 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1678 unsigned int size = skb->len;
1679 if (likely(size >= len))
1681 return skb_pad(skb, len - size);
1684 static inline int skb_add_data(struct sk_buff *skb,
1685 char __user *from, int copy)
1687 const int off = skb->len;
1689 if (skb->ip_summed == CHECKSUM_NONE) {
1691 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1694 skb->csum = csum_block_add(skb->csum, csum, off);
1697 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1700 __skb_trim(skb, off);
1704 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1705 struct page *page, int off)
1708 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1710 return page == frag->page &&
1711 off == frag->page_offset + frag->size;
1716 static inline int __skb_linearize(struct sk_buff *skb)
1718 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1722 * skb_linearize - convert paged skb to linear one
1723 * @skb: buffer to linarize
1725 * If there is no free memory -ENOMEM is returned, otherwise zero
1726 * is returned and the old skb data released.
1728 static inline int skb_linearize(struct sk_buff *skb)
1730 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1734 * skb_linearize_cow - make sure skb is linear and writable
1735 * @skb: buffer to process
1737 * If there is no free memory -ENOMEM is returned, otherwise zero
1738 * is returned and the old skb data released.
1740 static inline int skb_linearize_cow(struct sk_buff *skb)
1742 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1743 __skb_linearize(skb) : 0;
1747 * skb_postpull_rcsum - update checksum for received skb after pull
1748 * @skb: buffer to update
1749 * @start: start of data before pull
1750 * @len: length of data pulled
1752 * After doing a pull on a received packet, you need to call this to
1753 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1754 * CHECKSUM_NONE so that it can be recomputed from scratch.
1757 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1758 const void *start, unsigned int len)
1760 if (skb->ip_summed == CHECKSUM_COMPLETE)
1761 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1764 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1767 * pskb_trim_rcsum - trim received skb and update checksum
1768 * @skb: buffer to trim
1771 * This is exactly the same as pskb_trim except that it ensures the
1772 * checksum of received packets are still valid after the operation.
1775 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1777 if (likely(len >= skb->len))
1779 if (skb->ip_summed == CHECKSUM_COMPLETE)
1780 skb->ip_summed = CHECKSUM_NONE;
1781 return __pskb_trim(skb, len);
1784 #define skb_queue_walk(queue, skb) \
1785 for (skb = (queue)->next; \
1786 (skb != (struct sk_buff *)(queue)); \
1789 #define skb_queue_walk_safe(queue, skb, tmp) \
1790 for (skb = (queue)->next, tmp = skb->next; \
1791 skb != (struct sk_buff *)(queue); \
1792 skb = tmp, tmp = skb->next)
1794 #define skb_queue_walk_from(queue, skb) \
1795 for (; (skb != (struct sk_buff *)(queue)); \
1798 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1799 for (tmp = skb->next; \
1800 skb != (struct sk_buff *)(queue); \
1801 skb = tmp, tmp = skb->next)
1803 #define skb_queue_reverse_walk(queue, skb) \
1804 for (skb = (queue)->prev; \
1805 (skb != (struct sk_buff *)(queue)); \
1808 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1809 for (skb = (queue)->prev, tmp = skb->prev; \
1810 skb != (struct sk_buff *)(queue); \
1811 skb = tmp, tmp = skb->prev)
1813 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1814 for (tmp = skb->prev; \
1815 skb != (struct sk_buff *)(queue); \
1816 skb = tmp, tmp = skb->prev)
1818 static inline bool skb_has_frag_list(const struct sk_buff *skb)
1820 return skb_shinfo(skb)->frag_list != NULL;
1823 static inline void skb_frag_list_init(struct sk_buff *skb)
1825 skb_shinfo(skb)->frag_list = NULL;
1828 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1830 frag->next = skb_shinfo(skb)->frag_list;
1831 skb_shinfo(skb)->frag_list = frag;
1834 #define skb_walk_frags(skb, iter) \
1835 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1837 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1838 int *peeked, int *err);
1839 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1840 int noblock, int *err);
1841 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1842 struct poll_table_struct *wait);
1843 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1844 int offset, struct iovec *to,
1846 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1849 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1851 const struct iovec *from,
1854 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1856 const struct iovec *to,
1859 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1860 extern void skb_free_datagram_locked(struct sock *sk,
1861 struct sk_buff *skb);
1862 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1863 unsigned int flags);
1864 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1865 int len, __wsum csum);
1866 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1868 extern int skb_store_bits(struct sk_buff *skb, int offset,
1869 const void *from, int len);
1870 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1871 int offset, u8 *to, int len,
1873 extern int skb_splice_bits(struct sk_buff *skb,
1874 unsigned int offset,
1875 struct pipe_inode_info *pipe,
1877 unsigned int flags);
1878 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1879 extern void skb_split(struct sk_buff *skb,
1880 struct sk_buff *skb1, const u32 len);
1881 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1884 extern struct sk_buff *skb_segment(struct sk_buff *skb, u32 features);
1886 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1887 int len, void *buffer)
1889 int hlen = skb_headlen(skb);
1891 if (hlen - offset >= len)
1892 return skb->data + offset;
1894 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1900 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1902 const unsigned int len)
1904 memcpy(to, skb->data, len);
1907 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1908 const int offset, void *to,
1909 const unsigned int len)
1911 memcpy(to, skb->data + offset, len);
1914 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1916 const unsigned int len)
1918 memcpy(skb->data, from, len);
1921 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1924 const unsigned int len)
1926 memcpy(skb->data + offset, from, len);
1929 extern void skb_init(void);
1931 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1937 * skb_get_timestamp - get timestamp from a skb
1938 * @skb: skb to get stamp from
1939 * @stamp: pointer to struct timeval to store stamp in
1941 * Timestamps are stored in the skb as offsets to a base timestamp.
1942 * This function converts the offset back to a struct timeval and stores
1945 static inline void skb_get_timestamp(const struct sk_buff *skb,
1946 struct timeval *stamp)
1948 *stamp = ktime_to_timeval(skb->tstamp);
1951 static inline void skb_get_timestampns(const struct sk_buff *skb,
1952 struct timespec *stamp)
1954 *stamp = ktime_to_timespec(skb->tstamp);
1957 static inline void __net_timestamp(struct sk_buff *skb)
1959 skb->tstamp = ktime_get_real();
1962 static inline ktime_t net_timedelta(ktime_t t)
1964 return ktime_sub(ktime_get_real(), t);
1967 static inline ktime_t net_invalid_timestamp(void)
1969 return ktime_set(0, 0);
1972 extern void skb_timestamping_init(void);
1974 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1976 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1977 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1979 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1981 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1985 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1990 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1993 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1995 * @skb: clone of the the original outgoing packet
1996 * @hwtstamps: hardware time stamps
1999 void skb_complete_tx_timestamp(struct sk_buff *skb,
2000 struct skb_shared_hwtstamps *hwtstamps);
2003 * skb_tstamp_tx - queue clone of skb with send time stamps
2004 * @orig_skb: the original outgoing packet
2005 * @hwtstamps: hardware time stamps, may be NULL if not available
2007 * If the skb has a socket associated, then this function clones the
2008 * skb (thus sharing the actual data and optional structures), stores
2009 * the optional hardware time stamping information (if non NULL) or
2010 * generates a software time stamp (otherwise), then queues the clone
2011 * to the error queue of the socket. Errors are silently ignored.
2013 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2014 struct skb_shared_hwtstamps *hwtstamps);
2016 static inline void sw_tx_timestamp(struct sk_buff *skb)
2018 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2019 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2020 skb_tstamp_tx(skb, NULL);
2024 * skb_tx_timestamp() - Driver hook for transmit timestamping
2026 * Ethernet MAC Drivers should call this function in their hard_xmit()
2027 * function as soon as possible after giving the sk_buff to the MAC
2028 * hardware, but before freeing the sk_buff.
2030 * @skb: A socket buffer.
2032 static inline void skb_tx_timestamp(struct sk_buff *skb)
2034 skb_clone_tx_timestamp(skb);
2035 sw_tx_timestamp(skb);
2038 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2039 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2041 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2043 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2047 * skb_checksum_complete - Calculate checksum of an entire packet
2048 * @skb: packet to process
2050 * This function calculates the checksum over the entire packet plus
2051 * the value of skb->csum. The latter can be used to supply the
2052 * checksum of a pseudo header as used by TCP/UDP. It returns the
2055 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2056 * this function can be used to verify that checksum on received
2057 * packets. In that case the function should return zero if the
2058 * checksum is correct. In particular, this function will return zero
2059 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2060 * hardware has already verified the correctness of the checksum.
2062 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2064 return skb_csum_unnecessary(skb) ?
2065 0 : __skb_checksum_complete(skb);
2068 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2069 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2070 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2072 if (nfct && atomic_dec_and_test(&nfct->use))
2073 nf_conntrack_destroy(nfct);
2075 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2078 atomic_inc(&nfct->use);
2081 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2082 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2085 atomic_inc(&skb->users);
2087 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2093 #ifdef CONFIG_BRIDGE_NETFILTER
2094 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2096 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2099 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2102 atomic_inc(&nf_bridge->use);
2104 #endif /* CONFIG_BRIDGE_NETFILTER */
2105 static inline void nf_reset(struct sk_buff *skb)
2107 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2108 nf_conntrack_put(skb->nfct);
2111 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2112 nf_conntrack_put_reasm(skb->nfct_reasm);
2113 skb->nfct_reasm = NULL;
2115 #ifdef CONFIG_BRIDGE_NETFILTER
2116 nf_bridge_put(skb->nf_bridge);
2117 skb->nf_bridge = NULL;
2121 /* Note: This doesn't put any conntrack and bridge info in dst. */
2122 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2124 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2125 dst->nfct = src->nfct;
2126 nf_conntrack_get(src->nfct);
2127 dst->nfctinfo = src->nfctinfo;
2129 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2130 dst->nfct_reasm = src->nfct_reasm;
2131 nf_conntrack_get_reasm(src->nfct_reasm);
2133 #ifdef CONFIG_BRIDGE_NETFILTER
2134 dst->nf_bridge = src->nf_bridge;
2135 nf_bridge_get(src->nf_bridge);
2139 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2141 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2142 nf_conntrack_put(dst->nfct);
2144 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2145 nf_conntrack_put_reasm(dst->nfct_reasm);
2147 #ifdef CONFIG_BRIDGE_NETFILTER
2148 nf_bridge_put(dst->nf_bridge);
2150 __nf_copy(dst, src);
2153 #ifdef CONFIG_NETWORK_SECMARK
2154 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2156 to->secmark = from->secmark;
2159 static inline void skb_init_secmark(struct sk_buff *skb)
2164 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2167 static inline void skb_init_secmark(struct sk_buff *skb)
2171 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2173 skb->queue_mapping = queue_mapping;
2176 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2178 return skb->queue_mapping;
2181 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2183 to->queue_mapping = from->queue_mapping;
2186 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2188 skb->queue_mapping = rx_queue + 1;
2191 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2193 return skb->queue_mapping - 1;
2196 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2198 return skb->queue_mapping != 0;
2201 extern u16 __skb_tx_hash(const struct net_device *dev,
2202 const struct sk_buff *skb,
2203 unsigned int num_tx_queues);
2206 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2211 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2217 static inline int skb_is_gso(const struct sk_buff *skb)
2219 return skb_shinfo(skb)->gso_size;
2222 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2224 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2227 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2229 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2231 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2232 * wanted then gso_type will be set. */
2233 struct skb_shared_info *shinfo = skb_shinfo(skb);
2234 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2235 unlikely(shinfo->gso_type == 0)) {
2236 __skb_warn_lro_forwarding(skb);
2242 static inline void skb_forward_csum(struct sk_buff *skb)
2244 /* Unfortunately we don't support this one. Any brave souls? */
2245 if (skb->ip_summed == CHECKSUM_COMPLETE)
2246 skb->ip_summed = CHECKSUM_NONE;
2250 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2251 * @skb: skb to check
2253 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2254 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2255 * use this helper, to document places where we make this assertion.
2257 static inline void skb_checksum_none_assert(struct sk_buff *skb)
2260 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2264 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2265 #endif /* __KERNEL__ */
2266 #endif /* _LINUX_SKBUFF_H */