2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/dmaengine.h>
31 #include <linux/hrtimer.h>
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. Since
126 * GRO uses frags we allocate at least 16 regardless of page size.
128 #if (65536/PAGE_SIZE + 2) < 16
129 #define MAX_SKB_FRAGS 16UL
131 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
134 typedef struct skb_frag_struct skb_frag_t
;
136 struct skb_frag_struct
{
138 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
147 #define HAVE_HW_TIME_STAMP
150 * struct skb_shared_hwtstamps - hardware time stamps
151 * @hwtstamp: hardware time stamp transformed into duration
152 * since arbitrary point in time
153 * @syststamp: hwtstamp transformed to system time base
155 * Software time stamps generated by ktime_get_real() are stored in
156 * skb->tstamp. The relation between the different kinds of time
157 * stamps is as follows:
159 * syststamp and tstamp can be compared against each other in
160 * arbitrary combinations. The accuracy of a
161 * syststamp/tstamp/"syststamp from other device" comparison is
162 * limited by the accuracy of the transformation into system time
163 * base. This depends on the device driver and its underlying
166 * hwtstamps can only be compared against other hwtstamps from
169 * This structure is attached to packets as part of the
170 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
172 struct skb_shared_hwtstamps
{
177 /* Definitions for tx_flags in struct skb_shared_info */
179 /* generate hardware time stamp */
180 SKBTX_HW_TSTAMP
= 1 << 0,
182 /* generate software time stamp */
183 SKBTX_SW_TSTAMP
= 1 << 1,
185 /* device driver is going to provide hardware time stamp */
186 SKBTX_IN_PROGRESS
= 1 << 2,
188 /* ensure the originating sk reference is available on driver level */
189 SKBTX_DRV_NEEDS_SK_REF
= 1 << 3,
191 /* device driver supports TX zero-copy buffers */
192 SKBTX_DEV_ZEROCOPY
= 1 << 4,
196 * The callback notifies userspace to release buffers when skb DMA is done in
197 * lower device, the skb last reference should be 0 when calling this.
198 * The desc is used to track userspace buffer index.
201 void (*callback
)(void *);
206 /* This data is invariant across clones and lives at
207 * the end of the header data, ie. at skb->end.
209 struct skb_shared_info
{
210 unsigned short nr_frags
;
211 unsigned short gso_size
;
212 /* Warning: this field is not always filled in (UFO)! */
213 unsigned short gso_segs
;
214 unsigned short gso_type
;
217 struct sk_buff
*frag_list
;
218 struct skb_shared_hwtstamps hwtstamps
;
221 * Warning : all fields before dataref are cleared in __alloc_skb()
225 /* Intermediate layers must ensure that destructor_arg
226 * remains valid until skb destructor */
227 void * destructor_arg
;
229 /* must be last field, see pskb_expand_head() */
230 skb_frag_t frags
[MAX_SKB_FRAGS
];
233 /* We divide dataref into two halves. The higher 16 bits hold references
234 * to the payload part of skb->data. The lower 16 bits hold references to
235 * the entire skb->data. A clone of a headerless skb holds the length of
236 * the header in skb->hdr_len.
238 * All users must obey the rule that the skb->data reference count must be
239 * greater than or equal to the payload reference count.
241 * Holding a reference to the payload part means that the user does not
242 * care about modifications to the header part of skb->data.
244 #define SKB_DATAREF_SHIFT 16
245 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
249 SKB_FCLONE_UNAVAILABLE
,
255 SKB_GSO_TCPV4
= 1 << 0,
256 SKB_GSO_UDP
= 1 << 1,
258 /* This indicates the skb is from an untrusted source. */
259 SKB_GSO_DODGY
= 1 << 2,
261 /* This indicates the tcp segment has CWR set. */
262 SKB_GSO_TCP_ECN
= 1 << 3,
264 SKB_GSO_TCPV6
= 1 << 4,
266 SKB_GSO_FCOE
= 1 << 5,
269 #if BITS_PER_LONG > 32
270 #define NET_SKBUFF_DATA_USES_OFFSET 1
273 #ifdef NET_SKBUFF_DATA_USES_OFFSET
274 typedef unsigned int sk_buff_data_t
;
276 typedef unsigned char *sk_buff_data_t
;
279 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
280 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
281 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
285 * struct sk_buff - socket buffer
286 * @next: Next buffer in list
287 * @prev: Previous buffer in list
288 * @tstamp: Time we arrived
289 * @sk: Socket we are owned by
290 * @dev: Device we arrived on/are leaving by
291 * @cb: Control buffer. Free for use by every layer. Put private vars here
292 * @_skb_refdst: destination entry (with norefcount bit)
293 * @sp: the security path, used for xfrm
294 * @len: Length of actual data
295 * @data_len: Data length
296 * @mac_len: Length of link layer header
297 * @hdr_len: writable header length of cloned skb
298 * @csum: Checksum (must include start/offset pair)
299 * @csum_start: Offset from skb->head where checksumming should start
300 * @csum_offset: Offset from csum_start where checksum should be stored
301 * @priority: Packet queueing priority
302 * @local_df: allow local fragmentation
303 * @cloned: Head may be cloned (check refcnt to be sure)
304 * @ip_summed: Driver fed us an IP checksum
305 * @nohdr: Payload reference only, must not modify header
306 * @nfctinfo: Relationship of this skb to the connection
307 * @pkt_type: Packet class
308 * @fclone: skbuff clone status
309 * @ipvs_property: skbuff is owned by ipvs
310 * @peeked: this packet has been seen already, so stats have been
311 * done for it, don't do them again
312 * @nf_trace: netfilter packet trace flag
313 * @protocol: Packet protocol from driver
314 * @destructor: Destruct function
315 * @nfct: Associated connection, if any
316 * @nfct_reasm: netfilter conntrack re-assembly pointer
317 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
318 * @skb_iif: ifindex of device we arrived on
319 * @tc_index: Traffic control index
320 * @tc_verd: traffic control verdict
321 * @rxhash: the packet hash computed on receive
322 * @queue_mapping: Queue mapping for multiqueue devices
323 * @ndisc_nodetype: router type (from link layer)
324 * @ooo_okay: allow the mapping of a socket to a queue to be changed
325 * @dma_cookie: a cookie to one of several possible DMA operations
326 * done by skb DMA functions
327 * @secmark: security marking
328 * @mark: Generic packet mark
329 * @dropcount: total number of sk_receive_queue overflows
330 * @vlan_tci: vlan tag control information
331 * @transport_header: Transport layer header
332 * @network_header: Network layer header
333 * @mac_header: Link layer header
334 * @tail: Tail pointer
336 * @head: Head of buffer
337 * @data: Data head pointer
338 * @truesize: Buffer size
339 * @users: User count - see {datagram,tcp}.c
343 /* These two members must be first. */
344 struct sk_buff
*next
;
345 struct sk_buff
*prev
;
350 struct net_device
*dev
;
353 * This is the control buffer. It is free to use for every
354 * layer. Please put your private variables there. If you
355 * want to keep them across layers you have to do a skb_clone()
356 * first. This is owned by whoever has the skb queued ATM.
358 char cb
[48] __aligned(8);
360 unsigned long _skb_refdst
;
376 kmemcheck_bitfield_begin(flags1
);
387 kmemcheck_bitfield_end(flags1
);
390 void (*destructor
)(struct sk_buff
*skb
);
391 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
392 struct nf_conntrack
*nfct
;
394 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
395 struct sk_buff
*nfct_reasm
;
397 #ifdef CONFIG_BRIDGE_NETFILTER
398 struct nf_bridge_info
*nf_bridge
;
402 #ifdef CONFIG_NET_SCHED
403 __u16 tc_index
; /* traffic control index */
404 #ifdef CONFIG_NET_CLS_ACT
405 __u16 tc_verd
; /* traffic control verdict */
412 kmemcheck_bitfield_begin(flags2
);
413 #ifdef CONFIG_IPV6_NDISC_NODETYPE
414 __u8 ndisc_nodetype
:2;
417 kmemcheck_bitfield_end(flags2
);
421 #ifdef CONFIG_NET_DMA
422 dma_cookie_t dma_cookie
;
424 #ifdef CONFIG_NETWORK_SECMARK
434 sk_buff_data_t transport_header
;
435 sk_buff_data_t network_header
;
436 sk_buff_data_t mac_header
;
437 /* These elements must be at the end, see alloc_skb() for details. */
442 unsigned int truesize
;
448 * Handling routines are only of interest to the kernel
450 #include <linux/slab.h>
452 #include <asm/system.h>
455 * skb might have a dst pointer attached, refcounted or not.
456 * _skb_refdst low order bit is set if refcount was _not_ taken
458 #define SKB_DST_NOREF 1UL
459 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
462 * skb_dst - returns skb dst_entry
465 * Returns skb dst_entry, regardless of reference taken or not.
467 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
469 /* If refdst was not refcounted, check we still are in a
470 * rcu_read_lock section
472 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
473 !rcu_read_lock_held() &&
474 !rcu_read_lock_bh_held());
475 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
479 * skb_dst_set - sets skb dst
483 * Sets skb dst, assuming a reference was taken on dst and should
484 * be released by skb_dst_drop()
486 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
488 skb
->_skb_refdst
= (unsigned long)dst
;
491 extern void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
);
494 * skb_dst_is_noref - Test if skb dst isn't refcounted
497 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
499 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
502 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
504 return (struct rtable
*)skb_dst(skb
);
507 extern void kfree_skb(struct sk_buff
*skb
);
508 extern void consume_skb(struct sk_buff
*skb
);
509 extern void __kfree_skb(struct sk_buff
*skb
);
510 extern struct sk_buff
*__alloc_skb(unsigned int size
,
511 gfp_t priority
, int fclone
, int node
);
512 static inline struct sk_buff
*alloc_skb(unsigned int size
,
515 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
518 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
521 return __alloc_skb(size
, priority
, 1, NUMA_NO_NODE
);
524 extern bool skb_recycle_check(struct sk_buff
*skb
, int skb_size
);
526 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
527 extern int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
528 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
530 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
532 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
534 extern int pskb_expand_head(struct sk_buff
*skb
,
535 int nhead
, int ntail
,
537 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
538 unsigned int headroom
);
539 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
540 int newheadroom
, int newtailroom
,
542 extern int skb_to_sgvec(struct sk_buff
*skb
,
543 struct scatterlist
*sg
, int offset
,
545 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
546 struct sk_buff
**trailer
);
547 extern int skb_pad(struct sk_buff
*skb
, int pad
);
548 #define dev_kfree_skb(a) consume_skb(a)
550 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
551 int getfrag(void *from
, char *to
, int offset
,
552 int len
,int odd
, struct sk_buff
*skb
),
553 void *from
, int length
);
555 struct skb_seq_state
{
559 __u32 stepped_offset
;
560 struct sk_buff
*root_skb
;
561 struct sk_buff
*cur_skb
;
565 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
566 unsigned int from
, unsigned int to
,
567 struct skb_seq_state
*st
);
568 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
569 struct skb_seq_state
*st
);
570 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
572 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
573 unsigned int to
, struct ts_config
*config
,
574 struct ts_state
*state
);
576 extern __u32
__skb_get_rxhash(struct sk_buff
*skb
);
577 static inline __u32
skb_get_rxhash(struct sk_buff
*skb
)
580 skb
->rxhash
= __skb_get_rxhash(skb
);
585 #ifdef NET_SKBUFF_DATA_USES_OFFSET
586 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
588 return skb
->head
+ skb
->end
;
591 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
598 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
600 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
602 return &skb_shinfo(skb
)->hwtstamps
;
606 * skb_queue_empty - check if a queue is empty
609 * Returns true if the queue is empty, false otherwise.
611 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
613 return list
->next
== (struct sk_buff
*)list
;
617 * skb_queue_is_last - check if skb is the last entry in the queue
621 * Returns true if @skb is the last buffer on the list.
623 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
624 const struct sk_buff
*skb
)
626 return skb
->next
== (struct sk_buff
*)list
;
630 * skb_queue_is_first - check if skb is the first entry in the queue
634 * Returns true if @skb is the first buffer on the list.
636 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
637 const struct sk_buff
*skb
)
639 return skb
->prev
== (struct sk_buff
*)list
;
643 * skb_queue_next - return the next packet in the queue
645 * @skb: current buffer
647 * Return the next packet in @list after @skb. It is only valid to
648 * call this if skb_queue_is_last() evaluates to false.
650 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
651 const struct sk_buff
*skb
)
653 /* This BUG_ON may seem severe, but if we just return then we
654 * are going to dereference garbage.
656 BUG_ON(skb_queue_is_last(list
, skb
));
661 * skb_queue_prev - return the prev packet in the queue
663 * @skb: current buffer
665 * Return the prev packet in @list before @skb. It is only valid to
666 * call this if skb_queue_is_first() evaluates to false.
668 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
669 const struct sk_buff
*skb
)
671 /* This BUG_ON may seem severe, but if we just return then we
672 * are going to dereference garbage.
674 BUG_ON(skb_queue_is_first(list
, skb
));
679 * skb_get - reference buffer
680 * @skb: buffer to reference
682 * Makes another reference to a socket buffer and returns a pointer
685 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
687 atomic_inc(&skb
->users
);
692 * If users == 1, we are the only owner and are can avoid redundant
697 * skb_cloned - is the buffer a clone
698 * @skb: buffer to check
700 * Returns true if the buffer was generated with skb_clone() and is
701 * one of multiple shared copies of the buffer. Cloned buffers are
702 * shared data so must not be written to under normal circumstances.
704 static inline int skb_cloned(const struct sk_buff
*skb
)
706 return skb
->cloned
&&
707 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
711 * skb_header_cloned - is the header a clone
712 * @skb: buffer to check
714 * Returns true if modifying the header part of the buffer requires
715 * the data to be copied.
717 static inline int skb_header_cloned(const struct sk_buff
*skb
)
724 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
725 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
730 * skb_header_release - release reference to header
731 * @skb: buffer to operate on
733 * Drop a reference to the header part of the buffer. This is done
734 * by acquiring a payload reference. You must not read from the header
735 * part of skb->data after this.
737 static inline void skb_header_release(struct sk_buff
*skb
)
741 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
745 * skb_shared - is the buffer shared
746 * @skb: buffer to check
748 * Returns true if more than one person has a reference to this
751 static inline int skb_shared(const struct sk_buff
*skb
)
753 return atomic_read(&skb
->users
) != 1;
757 * skb_share_check - check if buffer is shared and if so clone it
758 * @skb: buffer to check
759 * @pri: priority for memory allocation
761 * If the buffer is shared the buffer is cloned and the old copy
762 * drops a reference. A new clone with a single reference is returned.
763 * If the buffer is not shared the original buffer is returned. When
764 * being called from interrupt status or with spinlocks held pri must
767 * NULL is returned on a memory allocation failure.
769 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
772 might_sleep_if(pri
& __GFP_WAIT
);
773 if (skb_shared(skb
)) {
774 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
782 * Copy shared buffers into a new sk_buff. We effectively do COW on
783 * packets to handle cases where we have a local reader and forward
784 * and a couple of other messy ones. The normal one is tcpdumping
785 * a packet thats being forwarded.
789 * skb_unshare - make a copy of a shared buffer
790 * @skb: buffer to check
791 * @pri: priority for memory allocation
793 * If the socket buffer is a clone then this function creates a new
794 * copy of the data, drops a reference count on the old copy and returns
795 * the new copy with the reference count at 1. If the buffer is not a clone
796 * the original buffer is returned. When called with a spinlock held or
797 * from interrupt state @pri must be %GFP_ATOMIC
799 * %NULL is returned on a memory allocation failure.
801 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
804 might_sleep_if(pri
& __GFP_WAIT
);
805 if (skb_cloned(skb
)) {
806 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
807 kfree_skb(skb
); /* Free our shared copy */
814 * skb_peek - peek at the head of an &sk_buff_head
815 * @list_: list to peek at
817 * Peek an &sk_buff. Unlike most other operations you _MUST_
818 * be careful with this one. A peek leaves the buffer on the
819 * list and someone else may run off with it. You must hold
820 * the appropriate locks or have a private queue to do this.
822 * Returns %NULL for an empty list or a pointer to the head element.
823 * The reference count is not incremented and the reference is therefore
824 * volatile. Use with caution.
826 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
828 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
829 if (list
== (struct sk_buff
*)list_
)
835 * skb_peek_tail - peek at the tail of an &sk_buff_head
836 * @list_: list to peek at
838 * Peek an &sk_buff. Unlike most other operations you _MUST_
839 * be careful with this one. A peek leaves the buffer on the
840 * list and someone else may run off with it. You must hold
841 * the appropriate locks or have a private queue to do this.
843 * Returns %NULL for an empty list or a pointer to the tail element.
844 * The reference count is not incremented and the reference is therefore
845 * volatile. Use with caution.
847 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
849 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
850 if (list
== (struct sk_buff
*)list_
)
856 * skb_queue_len - get queue length
857 * @list_: list to measure
859 * Return the length of an &sk_buff queue.
861 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
867 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
868 * @list: queue to initialize
870 * This initializes only the list and queue length aspects of
871 * an sk_buff_head object. This allows to initialize the list
872 * aspects of an sk_buff_head without reinitializing things like
873 * the spinlock. It can also be used for on-stack sk_buff_head
874 * objects where the spinlock is known to not be used.
876 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
878 list
->prev
= list
->next
= (struct sk_buff
*)list
;
883 * This function creates a split out lock class for each invocation;
884 * this is needed for now since a whole lot of users of the skb-queue
885 * infrastructure in drivers have different locking usage (in hardirq)
886 * than the networking core (in softirq only). In the long run either the
887 * network layer or drivers should need annotation to consolidate the
888 * main types of usage into 3 classes.
890 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
892 spin_lock_init(&list
->lock
);
893 __skb_queue_head_init(list
);
896 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
897 struct lock_class_key
*class)
899 skb_queue_head_init(list
);
900 lockdep_set_class(&list
->lock
, class);
904 * Insert an sk_buff on a list.
906 * The "__skb_xxxx()" functions are the non-atomic ones that
907 * can only be called with interrupts disabled.
909 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
910 static inline void __skb_insert(struct sk_buff
*newsk
,
911 struct sk_buff
*prev
, struct sk_buff
*next
,
912 struct sk_buff_head
*list
)
916 next
->prev
= prev
->next
= newsk
;
920 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
921 struct sk_buff
*prev
,
922 struct sk_buff
*next
)
924 struct sk_buff
*first
= list
->next
;
925 struct sk_buff
*last
= list
->prev
;
935 * skb_queue_splice - join two skb lists, this is designed for stacks
936 * @list: the new list to add
937 * @head: the place to add it in the first list
939 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
940 struct sk_buff_head
*head
)
942 if (!skb_queue_empty(list
)) {
943 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
944 head
->qlen
+= list
->qlen
;
949 * skb_queue_splice - join two skb lists and reinitialise the emptied list
950 * @list: the new list to add
951 * @head: the place to add it in the first list
953 * The list at @list is reinitialised
955 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
956 struct sk_buff_head
*head
)
958 if (!skb_queue_empty(list
)) {
959 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
960 head
->qlen
+= list
->qlen
;
961 __skb_queue_head_init(list
);
966 * skb_queue_splice_tail - join two skb lists, each list being a queue
967 * @list: the new list to add
968 * @head: the place to add it in the first list
970 static inline void skb_queue_splice_tail(const 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
;
980 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
981 * @list: the new list to add
982 * @head: the place to add it in the first list
984 * Each of the lists is a queue.
985 * The list at @list is reinitialised
987 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
988 struct sk_buff_head
*head
)
990 if (!skb_queue_empty(list
)) {
991 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
992 head
->qlen
+= list
->qlen
;
993 __skb_queue_head_init(list
);
998 * __skb_queue_after - queue a buffer at the list head
1000 * @prev: place after this buffer
1001 * @newsk: buffer to queue
1003 * Queue a buffer int the middle of a list. This function takes no locks
1004 * and you must therefore hold required locks before calling it.
1006 * A buffer cannot be placed on two lists at the same time.
1008 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1009 struct sk_buff
*prev
,
1010 struct sk_buff
*newsk
)
1012 __skb_insert(newsk
, prev
, prev
->next
, list
);
1015 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1016 struct sk_buff_head
*list
);
1018 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1019 struct sk_buff
*next
,
1020 struct sk_buff
*newsk
)
1022 __skb_insert(newsk
, next
->prev
, next
, list
);
1026 * __skb_queue_head - queue a buffer at the list head
1027 * @list: list to use
1028 * @newsk: buffer to queue
1030 * Queue a buffer at the start 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_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1036 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1037 struct sk_buff
*newsk
)
1039 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1043 * __skb_queue_tail - queue a buffer at the list tail
1044 * @list: list to use
1045 * @newsk: buffer to queue
1047 * Queue a buffer at the end of a list. This function takes no locks
1048 * and you must therefore hold required locks before calling it.
1050 * A buffer cannot be placed on two lists at the same time.
1052 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1053 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1054 struct sk_buff
*newsk
)
1056 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1060 * remove sk_buff from list. _Must_ be called atomically, and with
1063 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1064 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1066 struct sk_buff
*next
, *prev
;
1071 skb
->next
= skb
->prev
= NULL
;
1077 * __skb_dequeue - remove from the head of the queue
1078 * @list: list to dequeue from
1080 * Remove the head of the list. This function does not take any locks
1081 * so must be used with appropriate locks held only. The head item is
1082 * returned or %NULL if the list is empty.
1084 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1085 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1087 struct sk_buff
*skb
= skb_peek(list
);
1089 __skb_unlink(skb
, list
);
1094 * __skb_dequeue_tail - remove from the tail of the queue
1095 * @list: list to dequeue from
1097 * Remove the tail of the list. This function does not take any locks
1098 * so must be used with appropriate locks held only. The tail item is
1099 * returned or %NULL if the list is empty.
1101 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1102 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1104 struct sk_buff
*skb
= skb_peek_tail(list
);
1106 __skb_unlink(skb
, list
);
1111 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
1113 return skb
->data_len
;
1116 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1118 return skb
->len
- skb
->data_len
;
1121 static inline int skb_pagelen(const struct sk_buff
*skb
)
1125 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1126 len
+= skb_shinfo(skb
)->frags
[i
].size
;
1127 return len
+ skb_headlen(skb
);
1130 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1131 struct page
*page
, int off
, int size
)
1133 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1136 frag
->page_offset
= off
;
1138 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1141 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1144 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1145 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1146 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1148 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1149 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1151 return skb
->head
+ skb
->tail
;
1154 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1156 skb
->tail
= skb
->data
- skb
->head
;
1159 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1161 skb_reset_tail_pointer(skb
);
1162 skb
->tail
+= offset
;
1164 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1165 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1170 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1172 skb
->tail
= skb
->data
;
1175 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1177 skb
->tail
= skb
->data
+ offset
;
1180 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1183 * Add data to an sk_buff
1185 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1186 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1188 unsigned char *tmp
= skb_tail_pointer(skb
);
1189 SKB_LINEAR_ASSERT(skb
);
1195 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1196 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1203 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1204 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1207 BUG_ON(skb
->len
< skb
->data_len
);
1208 return skb
->data
+= len
;
1211 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1213 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1216 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1218 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1220 if (len
> skb_headlen(skb
) &&
1221 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1224 return skb
->data
+= len
;
1227 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1229 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1232 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1234 if (likely(len
<= skb_headlen(skb
)))
1236 if (unlikely(len
> skb
->len
))
1238 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1242 * skb_headroom - bytes at buffer head
1243 * @skb: buffer to check
1245 * Return the number of bytes of free space at the head of an &sk_buff.
1247 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1249 return skb
->data
- skb
->head
;
1253 * skb_tailroom - bytes at buffer end
1254 * @skb: buffer to check
1256 * Return the number of bytes of free space at the tail of an sk_buff
1258 static inline int skb_tailroom(const struct sk_buff
*skb
)
1260 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1264 * skb_reserve - adjust headroom
1265 * @skb: buffer to alter
1266 * @len: bytes to move
1268 * Increase the headroom of an empty &sk_buff by reducing the tail
1269 * room. This is only allowed for an empty buffer.
1271 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1277 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1279 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1282 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1283 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1285 return skb
->head
+ skb
->transport_header
;
1288 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1290 skb
->transport_header
= skb
->data
- skb
->head
;
1293 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1296 skb_reset_transport_header(skb
);
1297 skb
->transport_header
+= offset
;
1300 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1302 return skb
->head
+ skb
->network_header
;
1305 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1307 skb
->network_header
= skb
->data
- skb
->head
;
1310 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1312 skb_reset_network_header(skb
);
1313 skb
->network_header
+= offset
;
1316 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1318 return skb
->head
+ skb
->mac_header
;
1321 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1323 return skb
->mac_header
!= ~0U;
1326 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1328 skb
->mac_header
= skb
->data
- skb
->head
;
1331 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1333 skb_reset_mac_header(skb
);
1334 skb
->mac_header
+= offset
;
1337 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1339 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1341 return skb
->transport_header
;
1344 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1346 skb
->transport_header
= skb
->data
;
1349 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1352 skb
->transport_header
= skb
->data
+ offset
;
1355 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1357 return skb
->network_header
;
1360 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1362 skb
->network_header
= skb
->data
;
1365 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1367 skb
->network_header
= skb
->data
+ offset
;
1370 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1372 return skb
->mac_header
;
1375 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1377 return skb
->mac_header
!= NULL
;
1380 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1382 skb
->mac_header
= skb
->data
;
1385 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1387 skb
->mac_header
= skb
->data
+ offset
;
1389 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1391 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1393 return skb
->csum_start
- skb_headroom(skb
);
1396 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1398 return skb_transport_header(skb
) - skb
->data
;
1401 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1403 return skb
->transport_header
- skb
->network_header
;
1406 static inline int skb_network_offset(const struct sk_buff
*skb
)
1408 return skb_network_header(skb
) - skb
->data
;
1411 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1413 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1417 * CPUs often take a performance hit when accessing unaligned memory
1418 * locations. The actual performance hit varies, it can be small if the
1419 * hardware handles it or large if we have to take an exception and fix it
1422 * Since an ethernet header is 14 bytes network drivers often end up with
1423 * the IP header at an unaligned offset. The IP header can be aligned by
1424 * shifting the start of the packet by 2 bytes. Drivers should do this
1427 * skb_reserve(skb, NET_IP_ALIGN);
1429 * The downside to this alignment of the IP header is that the DMA is now
1430 * unaligned. On some architectures the cost of an unaligned DMA is high
1431 * and this cost outweighs the gains made by aligning the IP header.
1433 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1436 #ifndef NET_IP_ALIGN
1437 #define NET_IP_ALIGN 2
1441 * The networking layer reserves some headroom in skb data (via
1442 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1443 * the header has to grow. In the default case, if the header has to grow
1444 * 32 bytes or less we avoid the reallocation.
1446 * Unfortunately this headroom changes the DMA alignment of the resulting
1447 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1448 * on some architectures. An architecture can override this value,
1449 * perhaps setting it to a cacheline in size (since that will maintain
1450 * cacheline alignment of the DMA). It must be a power of 2.
1452 * Various parts of the networking layer expect at least 32 bytes of
1453 * headroom, you should not reduce this.
1455 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1456 * to reduce average number of cache lines per packet.
1457 * get_rps_cpus() for example only access one 64 bytes aligned block :
1458 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1461 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1464 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1466 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1468 if (unlikely(skb_is_nonlinear(skb
))) {
1473 skb_set_tail_pointer(skb
, len
);
1476 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1478 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1481 return ___pskb_trim(skb
, len
);
1482 __skb_trim(skb
, len
);
1486 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1488 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1492 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1493 * @skb: buffer to alter
1496 * This is identical to pskb_trim except that the caller knows that
1497 * the skb is not cloned so we should never get an error due to out-
1500 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1502 int err
= pskb_trim(skb
, len
);
1507 * skb_orphan - orphan a buffer
1508 * @skb: buffer to orphan
1510 * If a buffer currently has an owner then we call the owner's
1511 * destructor function and make the @skb unowned. The buffer continues
1512 * to exist but is no longer charged to its former owner.
1514 static inline void skb_orphan(struct sk_buff
*skb
)
1516 if (skb
->destructor
)
1517 skb
->destructor(skb
);
1518 skb
->destructor
= NULL
;
1523 * __skb_queue_purge - empty a list
1524 * @list: list to empty
1526 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1527 * the list and one reference dropped. This function does not take the
1528 * list lock and the caller must hold the relevant locks to use it.
1530 extern void skb_queue_purge(struct sk_buff_head
*list
);
1531 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1533 struct sk_buff
*skb
;
1534 while ((skb
= __skb_dequeue(list
)) != NULL
)
1539 * __dev_alloc_skb - allocate an skbuff for receiving
1540 * @length: length to allocate
1541 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1543 * Allocate a new &sk_buff and assign it a usage count of one. The
1544 * buffer has unspecified headroom built in. Users should allocate
1545 * the headroom they think they need without accounting for the
1546 * built in space. The built in space is used for optimisations.
1548 * %NULL is returned if there is no free memory.
1550 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1553 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1555 skb_reserve(skb
, NET_SKB_PAD
);
1559 extern struct sk_buff
*dev_alloc_skb(unsigned int length
);
1561 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1562 unsigned int length
, gfp_t gfp_mask
);
1565 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1566 * @dev: network device to receive on
1567 * @length: length to allocate
1569 * Allocate a new &sk_buff and assign it a usage count of one. The
1570 * buffer has unspecified headroom built in. Users should allocate
1571 * the headroom they think they need without accounting for the
1572 * built in space. The built in space is used for optimisations.
1574 * %NULL is returned if there is no free memory. Although this function
1575 * allocates memory it can be called from an interrupt.
1577 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1578 unsigned int length
)
1580 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1583 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
1584 unsigned int length
, gfp_t gfp
)
1586 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
1588 if (NET_IP_ALIGN
&& skb
)
1589 skb_reserve(skb
, NET_IP_ALIGN
);
1593 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1594 unsigned int length
)
1596 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
1600 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1601 * @dev: network device to receive on
1602 * @gfp_mask: alloc_pages_node mask
1604 * Allocate a new page. dev currently unused.
1606 * %NULL is returned if there is no free memory.
1608 static inline struct page
*__netdev_alloc_page(struct net_device
*dev
, gfp_t gfp_mask
)
1610 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, 0);
1614 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1615 * @dev: network device to receive on
1617 * Allocate a new page. dev currently unused.
1619 * %NULL is returned if there is no free memory.
1621 static inline struct page
*netdev_alloc_page(struct net_device
*dev
)
1623 return __netdev_alloc_page(dev
, GFP_ATOMIC
);
1626 static inline void netdev_free_page(struct net_device
*dev
, struct page
*page
)
1632 * skb_clone_writable - is the header of a clone writable
1633 * @skb: buffer to check
1634 * @len: length up to which to write
1636 * Returns true if modifying the header part of the cloned buffer
1637 * does not requires the data to be copied.
1639 static inline int skb_clone_writable(struct sk_buff
*skb
, unsigned int len
)
1641 return !skb_header_cloned(skb
) &&
1642 skb_headroom(skb
) + len
<= skb
->hdr_len
;
1645 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
1650 if (headroom
< NET_SKB_PAD
)
1651 headroom
= NET_SKB_PAD
;
1652 if (headroom
> skb_headroom(skb
))
1653 delta
= headroom
- skb_headroom(skb
);
1655 if (delta
|| cloned
)
1656 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
1662 * skb_cow - copy header of skb when it is required
1663 * @skb: buffer to cow
1664 * @headroom: needed headroom
1666 * If the skb passed lacks sufficient headroom or its data part
1667 * is shared, data is reallocated. If reallocation fails, an error
1668 * is returned and original skb is not changed.
1670 * The result is skb with writable area skb->head...skb->tail
1671 * and at least @headroom of space at head.
1673 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1675 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
1679 * skb_cow_head - skb_cow but only making the head writable
1680 * @skb: buffer to cow
1681 * @headroom: needed headroom
1683 * This function is identical to skb_cow except that we replace the
1684 * skb_cloned check by skb_header_cloned. It should be used when
1685 * you only need to push on some header and do not need to modify
1688 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
1690 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
1694 * skb_padto - pad an skbuff up to a minimal size
1695 * @skb: buffer to pad
1696 * @len: minimal length
1698 * Pads up a buffer to ensure the trailing bytes exist and are
1699 * blanked. If the buffer already contains sufficient data it
1700 * is untouched. Otherwise it is extended. Returns zero on
1701 * success. The skb is freed on error.
1704 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1706 unsigned int size
= skb
->len
;
1707 if (likely(size
>= len
))
1709 return skb_pad(skb
, len
- size
);
1712 static inline int skb_add_data(struct sk_buff
*skb
,
1713 char __user
*from
, int copy
)
1715 const int off
= skb
->len
;
1717 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1719 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1722 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1725 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1728 __skb_trim(skb
, off
);
1732 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1733 struct page
*page
, int off
)
1736 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1738 return page
== frag
->page
&&
1739 off
== frag
->page_offset
+ frag
->size
;
1744 static inline int __skb_linearize(struct sk_buff
*skb
)
1746 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1750 * skb_linearize - convert paged skb to linear one
1751 * @skb: buffer to linarize
1753 * If there is no free memory -ENOMEM is returned, otherwise zero
1754 * is returned and the old skb data released.
1756 static inline int skb_linearize(struct sk_buff
*skb
)
1758 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1762 * skb_linearize_cow - make sure skb is linear and writable
1763 * @skb: buffer to process
1765 * If there is no free memory -ENOMEM is returned, otherwise zero
1766 * is returned and the old skb data released.
1768 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1770 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1771 __skb_linearize(skb
) : 0;
1775 * skb_postpull_rcsum - update checksum for received skb after pull
1776 * @skb: buffer to update
1777 * @start: start of data before pull
1778 * @len: length of data pulled
1780 * After doing a pull on a received packet, you need to call this to
1781 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1782 * CHECKSUM_NONE so that it can be recomputed from scratch.
1785 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1786 const void *start
, unsigned int len
)
1788 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1789 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1792 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1795 * pskb_trim_rcsum - trim received skb and update checksum
1796 * @skb: buffer to trim
1799 * This is exactly the same as pskb_trim except that it ensures the
1800 * checksum of received packets are still valid after the operation.
1803 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1805 if (likely(len
>= skb
->len
))
1807 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1808 skb
->ip_summed
= CHECKSUM_NONE
;
1809 return __pskb_trim(skb
, len
);
1812 #define skb_queue_walk(queue, skb) \
1813 for (skb = (queue)->next; \
1814 skb != (struct sk_buff *)(queue); \
1817 #define skb_queue_walk_safe(queue, skb, tmp) \
1818 for (skb = (queue)->next, tmp = skb->next; \
1819 skb != (struct sk_buff *)(queue); \
1820 skb = tmp, tmp = skb->next)
1822 #define skb_queue_walk_from(queue, skb) \
1823 for (; skb != (struct sk_buff *)(queue); \
1826 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1827 for (tmp = skb->next; \
1828 skb != (struct sk_buff *)(queue); \
1829 skb = tmp, tmp = skb->next)
1831 #define skb_queue_reverse_walk(queue, skb) \
1832 for (skb = (queue)->prev; \
1833 skb != (struct sk_buff *)(queue); \
1836 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1837 for (skb = (queue)->prev, tmp = skb->prev; \
1838 skb != (struct sk_buff *)(queue); \
1839 skb = tmp, tmp = skb->prev)
1841 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1842 for (tmp = skb->prev; \
1843 skb != (struct sk_buff *)(queue); \
1844 skb = tmp, tmp = skb->prev)
1846 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
1848 return skb_shinfo(skb
)->frag_list
!= NULL
;
1851 static inline void skb_frag_list_init(struct sk_buff
*skb
)
1853 skb_shinfo(skb
)->frag_list
= NULL
;
1856 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
1858 frag
->next
= skb_shinfo(skb
)->frag_list
;
1859 skb_shinfo(skb
)->frag_list
= frag
;
1862 #define skb_walk_frags(skb, iter) \
1863 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1865 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1866 int *peeked
, int *err
);
1867 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1868 int noblock
, int *err
);
1869 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1870 struct poll_table_struct
*wait
);
1871 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1872 int offset
, struct iovec
*to
,
1874 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1877 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
1879 const struct iovec
*from
,
1882 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
1884 const struct iovec
*to
,
1887 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1888 extern void skb_free_datagram_locked(struct sock
*sk
,
1889 struct sk_buff
*skb
);
1890 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1891 unsigned int flags
);
1892 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1893 int len
, __wsum csum
);
1894 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1896 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
1897 const void *from
, int len
);
1898 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1899 int offset
, u8
*to
, int len
,
1901 extern int skb_splice_bits(struct sk_buff
*skb
,
1902 unsigned int offset
,
1903 struct pipe_inode_info
*pipe
,
1905 unsigned int flags
);
1906 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1907 extern void skb_split(struct sk_buff
*skb
,
1908 struct sk_buff
*skb1
, const u32 len
);
1909 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
1912 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, u32 features
);
1914 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1915 int len
, void *buffer
)
1917 int hlen
= skb_headlen(skb
);
1919 if (hlen
- offset
>= len
)
1920 return skb
->data
+ offset
;
1922 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1928 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
1930 const unsigned int len
)
1932 memcpy(to
, skb
->data
, len
);
1935 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
1936 const int offset
, void *to
,
1937 const unsigned int len
)
1939 memcpy(to
, skb
->data
+ offset
, len
);
1942 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
1944 const unsigned int len
)
1946 memcpy(skb
->data
, from
, len
);
1949 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
1952 const unsigned int len
)
1954 memcpy(skb
->data
+ offset
, from
, len
);
1957 extern void skb_init(void);
1959 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
1965 * skb_get_timestamp - get timestamp from a skb
1966 * @skb: skb to get stamp from
1967 * @stamp: pointer to struct timeval to store stamp in
1969 * Timestamps are stored in the skb as offsets to a base timestamp.
1970 * This function converts the offset back to a struct timeval and stores
1973 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
1974 struct timeval
*stamp
)
1976 *stamp
= ktime_to_timeval(skb
->tstamp
);
1979 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
1980 struct timespec
*stamp
)
1982 *stamp
= ktime_to_timespec(skb
->tstamp
);
1985 static inline void __net_timestamp(struct sk_buff
*skb
)
1987 skb
->tstamp
= ktime_get_real();
1990 static inline ktime_t
net_timedelta(ktime_t t
)
1992 return ktime_sub(ktime_get_real(), t
);
1995 static inline ktime_t
net_invalid_timestamp(void)
1997 return ktime_set(0, 0);
2000 extern void skb_timestamping_init(void);
2002 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2004 extern void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2005 extern bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2007 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2009 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2013 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2018 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2021 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2023 * PHY drivers may accept clones of transmitted packets for
2024 * timestamping via their phy_driver.txtstamp method. These drivers
2025 * must call this function to return the skb back to the stack, with
2026 * or without a timestamp.
2028 * @skb: clone of the the original outgoing packet
2029 * @hwtstamps: hardware time stamps, may be NULL if not available
2032 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2033 struct skb_shared_hwtstamps
*hwtstamps
);
2036 * skb_tstamp_tx - queue clone of skb with send time stamps
2037 * @orig_skb: the original outgoing packet
2038 * @hwtstamps: hardware time stamps, may be NULL if not available
2040 * If the skb has a socket associated, then this function clones the
2041 * skb (thus sharing the actual data and optional structures), stores
2042 * the optional hardware time stamping information (if non NULL) or
2043 * generates a software time stamp (otherwise), then queues the clone
2044 * to the error queue of the socket. Errors are silently ignored.
2046 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2047 struct skb_shared_hwtstamps
*hwtstamps
);
2049 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2051 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2052 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2053 skb_tstamp_tx(skb
, NULL
);
2057 * skb_tx_timestamp() - Driver hook for transmit timestamping
2059 * Ethernet MAC Drivers should call this function in their hard_xmit()
2060 * function immediately before giving the sk_buff to the MAC hardware.
2062 * @skb: A socket buffer.
2064 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2066 skb_clone_tx_timestamp(skb
);
2067 sw_tx_timestamp(skb
);
2070 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2071 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2073 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2075 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
2079 * skb_checksum_complete - Calculate checksum of an entire packet
2080 * @skb: packet to process
2082 * This function calculates the checksum over the entire packet plus
2083 * the value of skb->csum. The latter can be used to supply the
2084 * checksum of a pseudo header as used by TCP/UDP. It returns the
2087 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2088 * this function can be used to verify that checksum on received
2089 * packets. In that case the function should return zero if the
2090 * checksum is correct. In particular, this function will return zero
2091 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2092 * hardware has already verified the correctness of the checksum.
2094 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2096 return skb_csum_unnecessary(skb
) ?
2097 0 : __skb_checksum_complete(skb
);
2100 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2101 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
2102 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
2104 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
2105 nf_conntrack_destroy(nfct
);
2107 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
2110 atomic_inc(&nfct
->use
);
2113 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2114 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
2117 atomic_inc(&skb
->users
);
2119 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
2125 #ifdef CONFIG_BRIDGE_NETFILTER
2126 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2128 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2131 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2134 atomic_inc(&nf_bridge
->use
);
2136 #endif /* CONFIG_BRIDGE_NETFILTER */
2137 static inline void nf_reset(struct sk_buff
*skb
)
2139 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2140 nf_conntrack_put(skb
->nfct
);
2143 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2144 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2145 skb
->nfct_reasm
= NULL
;
2147 #ifdef CONFIG_BRIDGE_NETFILTER
2148 nf_bridge_put(skb
->nf_bridge
);
2149 skb
->nf_bridge
= NULL
;
2153 /* Note: This doesn't put any conntrack and bridge info in dst. */
2154 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2156 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2157 dst
->nfct
= src
->nfct
;
2158 nf_conntrack_get(src
->nfct
);
2159 dst
->nfctinfo
= src
->nfctinfo
;
2161 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2162 dst
->nfct_reasm
= src
->nfct_reasm
;
2163 nf_conntrack_get_reasm(src
->nfct_reasm
);
2165 #ifdef CONFIG_BRIDGE_NETFILTER
2166 dst
->nf_bridge
= src
->nf_bridge
;
2167 nf_bridge_get(src
->nf_bridge
);
2171 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2173 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2174 nf_conntrack_put(dst
->nfct
);
2176 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2177 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2179 #ifdef CONFIG_BRIDGE_NETFILTER
2180 nf_bridge_put(dst
->nf_bridge
);
2182 __nf_copy(dst
, src
);
2185 #ifdef CONFIG_NETWORK_SECMARK
2186 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2188 to
->secmark
= from
->secmark
;
2191 static inline void skb_init_secmark(struct sk_buff
*skb
)
2196 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2199 static inline void skb_init_secmark(struct sk_buff
*skb
)
2203 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2205 skb
->queue_mapping
= queue_mapping
;
2208 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2210 return skb
->queue_mapping
;
2213 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2215 to
->queue_mapping
= from
->queue_mapping
;
2218 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2220 skb
->queue_mapping
= rx_queue
+ 1;
2223 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2225 return skb
->queue_mapping
- 1;
2228 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2230 return skb
->queue_mapping
!= 0;
2233 extern u16
__skb_tx_hash(const struct net_device
*dev
,
2234 const struct sk_buff
*skb
,
2235 unsigned int num_tx_queues
);
2238 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2243 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2249 static inline int skb_is_gso(const struct sk_buff
*skb
)
2251 return skb_shinfo(skb
)->gso_size
;
2254 static inline int skb_is_gso_v6(const struct sk_buff
*skb
)
2256 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2259 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2261 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2263 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2264 * wanted then gso_type will be set. */
2265 struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2266 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
2267 unlikely(shinfo
->gso_type
== 0)) {
2268 __skb_warn_lro_forwarding(skb
);
2274 static inline void skb_forward_csum(struct sk_buff
*skb
)
2276 /* Unfortunately we don't support this one. Any brave souls? */
2277 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2278 skb
->ip_summed
= CHECKSUM_NONE
;
2282 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2283 * @skb: skb to check
2285 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2286 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2287 * use this helper, to document places where we make this assertion.
2289 static inline void skb_checksum_none_assert(struct sk_buff
*skb
)
2292 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
2296 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2298 #endif /* __KERNEL__ */
2299 #endif /* _LINUX_SKBUFF_H */