Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[linux-2.6.git] / include / linux / skbuff.h
blob5afefa01a13c892e1fe9eb7ce6fc3f047e5880ba
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
4 * Authors:
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/bug.h>
22 #include <linux/cache.h>
24 #include <linux/atomic.h>
25 #include <asm/types.h>
26 #include <linux/spinlock.h>
27 #include <linux/net.h>
28 #include <linux/textsearch.h>
29 #include <net/checksum.h>
30 #include <linux/rcupdate.h>
31 #include <linux/dmaengine.h>
32 #include <linux/hrtimer.h>
33 #include <linux/dma-mapping.h>
34 #include <linux/netdev_features.h>
35 #include <net/flow_keys.h>
37 /* Don't change this without changing skb_csum_unnecessary! */
38 #define CHECKSUM_NONE 0
39 #define CHECKSUM_UNNECESSARY 1
40 #define CHECKSUM_COMPLETE 2
41 #define CHECKSUM_PARTIAL 3
43 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_WITH_OVERHEAD(X) \
46 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
47 #define SKB_MAX_ORDER(X, ORDER) \
48 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
49 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
50 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
52 /* return minimum truesize of one skb containing X bytes of data */
53 #define SKB_TRUESIZE(X) ((X) + \
54 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
55 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
57 /* A. Checksumming of received packets by device.
59 * NONE: device failed to checksum this packet.
60 * skb->csum is undefined.
62 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
63 * skb->csum is undefined.
64 * It is bad option, but, unfortunately, many of vendors do this.
65 * Apparently with secret goal to sell you new device, when you
66 * will add new protocol to your host. F.e. IPv6. 8)
68 * COMPLETE: the most generic way. Device supplied checksum of _all_
69 * the packet as seen by netif_rx in skb->csum.
70 * NOTE: Even if device supports only some protocols, but
71 * is able to produce some skb->csum, it MUST use COMPLETE,
72 * not UNNECESSARY.
74 * PARTIAL: identical to the case for output below. This may occur
75 * on a packet received directly from another Linux OS, e.g.,
76 * a virtualised Linux kernel on the same host. The packet can
77 * be treated in the same way as UNNECESSARY except that on
78 * output (i.e., forwarding) the checksum must be filled in
79 * by the OS or the hardware.
81 * B. Checksumming on output.
83 * NONE: skb is checksummed by protocol or csum is not required.
85 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
86 * from skb->csum_start to the end and to record the checksum
87 * at skb->csum_start + skb->csum_offset.
89 * Device must show its capabilities in dev->features, set
90 * at device setup time.
91 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
92 * everything.
93 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
94 * TCP/UDP over IPv4. Sigh. Vendors like this
95 * way by an unknown reason. Though, see comment above
96 * about CHECKSUM_UNNECESSARY. 8)
97 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
99 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
100 * that do not want net to perform the checksum calculation should use
101 * this flag in their outgoing skbs.
102 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
106 * Any questions? No questions, good. --ANK
109 struct net_device;
110 struct scatterlist;
111 struct pipe_inode_info;
113 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
114 struct nf_conntrack {
115 atomic_t use;
117 #endif
119 #ifdef CONFIG_BRIDGE_NETFILTER
120 struct nf_bridge_info {
121 atomic_t use;
122 unsigned int mask;
123 struct net_device *physindev;
124 struct net_device *physoutdev;
125 unsigned long data[32 / sizeof(unsigned long)];
127 #endif
129 struct sk_buff_head {
130 /* These two members must be first. */
131 struct sk_buff *next;
132 struct sk_buff *prev;
134 __u32 qlen;
135 spinlock_t lock;
138 struct sk_buff;
140 /* To allow 64K frame to be packed as single skb without frag_list we
141 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
142 * buffers which do not start on a page boundary.
144 * Since GRO uses frags we allocate at least 16 regardless of page
145 * size.
147 #if (65536/PAGE_SIZE + 1) < 16
148 #define MAX_SKB_FRAGS 16UL
149 #else
150 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
151 #endif
153 typedef struct skb_frag_struct skb_frag_t;
155 struct skb_frag_struct {
156 struct {
157 struct page *p;
158 } page;
159 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
160 __u32 page_offset;
161 __u32 size;
162 #else
163 __u16 page_offset;
164 __u16 size;
165 #endif
168 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
170 return frag->size;
173 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
175 frag->size = size;
178 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
180 frag->size += delta;
183 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
185 frag->size -= delta;
188 #define HAVE_HW_TIME_STAMP
191 * struct skb_shared_hwtstamps - hardware time stamps
192 * @hwtstamp: hardware time stamp transformed into duration
193 * since arbitrary point in time
194 * @syststamp: hwtstamp transformed to system time base
196 * Software time stamps generated by ktime_get_real() are stored in
197 * skb->tstamp. The relation between the different kinds of time
198 * stamps is as follows:
200 * syststamp and tstamp can be compared against each other in
201 * arbitrary combinations. The accuracy of a
202 * syststamp/tstamp/"syststamp from other device" comparison is
203 * limited by the accuracy of the transformation into system time
204 * base. This depends on the device driver and its underlying
205 * hardware.
207 * hwtstamps can only be compared against other hwtstamps from
208 * the same device.
210 * This structure is attached to packets as part of the
211 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
213 struct skb_shared_hwtstamps {
214 ktime_t hwtstamp;
215 ktime_t syststamp;
218 /* Definitions for tx_flags in struct skb_shared_info */
219 enum {
220 /* generate hardware time stamp */
221 SKBTX_HW_TSTAMP = 1 << 0,
223 /* generate software time stamp */
224 SKBTX_SW_TSTAMP = 1 << 1,
226 /* device driver is going to provide hardware time stamp */
227 SKBTX_IN_PROGRESS = 1 << 2,
229 /* device driver supports TX zero-copy buffers */
230 SKBTX_DEV_ZEROCOPY = 1 << 3,
232 /* generate wifi status information (where possible) */
233 SKBTX_WIFI_STATUS = 1 << 4,
235 /* This indicates at least one fragment might be overwritten
236 * (as in vmsplice(), sendfile() ...)
237 * If we need to compute a TX checksum, we'll need to copy
238 * all frags to avoid possible bad checksum
240 SKBTX_SHARED_FRAG = 1 << 5,
244 * The callback notifies userspace to release buffers when skb DMA is done in
245 * lower device, the skb last reference should be 0 when calling this.
246 * The zerocopy_success argument is true if zero copy transmit occurred,
247 * false on data copy or out of memory error caused by data copy attempt.
248 * The ctx field is used to track device context.
249 * The desc field is used to track userspace buffer index.
251 struct ubuf_info {
252 void (*callback)(struct ubuf_info *, bool zerocopy_success);
253 void *ctx;
254 unsigned long desc;
257 /* This data is invariant across clones and lives at
258 * the end of the header data, ie. at skb->end.
260 struct skb_shared_info {
261 unsigned char nr_frags;
262 __u8 tx_flags;
263 unsigned short gso_size;
264 /* Warning: this field is not always filled in (UFO)! */
265 unsigned short gso_segs;
266 unsigned short gso_type;
267 struct sk_buff *frag_list;
268 struct skb_shared_hwtstamps hwtstamps;
269 __be32 ip6_frag_id;
272 * Warning : all fields before dataref are cleared in __alloc_skb()
274 atomic_t dataref;
276 /* Intermediate layers must ensure that destructor_arg
277 * remains valid until skb destructor */
278 void * destructor_arg;
280 /* must be last field, see pskb_expand_head() */
281 skb_frag_t frags[MAX_SKB_FRAGS];
284 /* We divide dataref into two halves. The higher 16 bits hold references
285 * to the payload part of skb->data. The lower 16 bits hold references to
286 * the entire skb->data. A clone of a headerless skb holds the length of
287 * the header in skb->hdr_len.
289 * All users must obey the rule that the skb->data reference count must be
290 * greater than or equal to the payload reference count.
292 * Holding a reference to the payload part means that the user does not
293 * care about modifications to the header part of skb->data.
295 #define SKB_DATAREF_SHIFT 16
296 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
299 enum {
300 SKB_FCLONE_UNAVAILABLE,
301 SKB_FCLONE_ORIG,
302 SKB_FCLONE_CLONE,
305 enum {
306 SKB_GSO_TCPV4 = 1 << 0,
307 SKB_GSO_UDP = 1 << 1,
309 /* This indicates the skb is from an untrusted source. */
310 SKB_GSO_DODGY = 1 << 2,
312 /* This indicates the tcp segment has CWR set. */
313 SKB_GSO_TCP_ECN = 1 << 3,
315 SKB_GSO_TCPV6 = 1 << 4,
317 SKB_GSO_FCOE = 1 << 5,
319 SKB_GSO_GRE = 1 << 6,
321 SKB_GSO_UDP_TUNNEL = 1 << 7,
323 SKB_GSO_MPLS = 1 << 8,
326 #if BITS_PER_LONG > 32
327 #define NET_SKBUFF_DATA_USES_OFFSET 1
328 #endif
330 #ifdef NET_SKBUFF_DATA_USES_OFFSET
331 typedef unsigned int sk_buff_data_t;
332 #else
333 typedef unsigned char *sk_buff_data_t;
334 #endif
336 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
337 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
338 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
339 #endif
341 /**
342 * struct sk_buff - socket buffer
343 * @next: Next buffer in list
344 * @prev: Previous buffer in list
345 * @tstamp: Time we arrived
346 * @sk: Socket we are owned by
347 * @dev: Device we arrived on/are leaving by
348 * @cb: Control buffer. Free for use by every layer. Put private vars here
349 * @_skb_refdst: destination entry (with norefcount bit)
350 * @sp: the security path, used for xfrm
351 * @len: Length of actual data
352 * @data_len: Data length
353 * @mac_len: Length of link layer header
354 * @hdr_len: writable header length of cloned skb
355 * @csum: Checksum (must include start/offset pair)
356 * @csum_start: Offset from skb->head where checksumming should start
357 * @csum_offset: Offset from csum_start where checksum should be stored
358 * @priority: Packet queueing priority
359 * @local_df: allow local fragmentation
360 * @cloned: Head may be cloned (check refcnt to be sure)
361 * @ip_summed: Driver fed us an IP checksum
362 * @nohdr: Payload reference only, must not modify header
363 * @nfctinfo: Relationship of this skb to the connection
364 * @pkt_type: Packet class
365 * @fclone: skbuff clone status
366 * @ipvs_property: skbuff is owned by ipvs
367 * @peeked: this packet has been seen already, so stats have been
368 * done for it, don't do them again
369 * @nf_trace: netfilter packet trace flag
370 * @protocol: Packet protocol from driver
371 * @destructor: Destruct function
372 * @nfct: Associated connection, if any
373 * @nfct_reasm: netfilter conntrack re-assembly pointer
374 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
375 * @skb_iif: ifindex of device we arrived on
376 * @tc_index: Traffic control index
377 * @tc_verd: traffic control verdict
378 * @rxhash: the packet hash computed on receive
379 * @queue_mapping: Queue mapping for multiqueue devices
380 * @ndisc_nodetype: router type (from link layer)
381 * @ooo_okay: allow the mapping of a socket to a queue to be changed
382 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
383 * ports.
384 * @wifi_acked_valid: wifi_acked was set
385 * @wifi_acked: whether frame was acked on wifi or not
386 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
387 * @dma_cookie: a cookie to one of several possible DMA operations
388 * done by skb DMA functions
389 * @napi_id: id of the NAPI struct this skb came from
390 * @secmark: security marking
391 * @mark: Generic packet mark
392 * @dropcount: total number of sk_receive_queue overflows
393 * @vlan_proto: vlan encapsulation protocol
394 * @vlan_tci: vlan tag control information
395 * @inner_protocol: Protocol (encapsulation)
396 * @inner_transport_header: Inner transport layer header (encapsulation)
397 * @inner_network_header: Network layer header (encapsulation)
398 * @inner_mac_header: Link layer header (encapsulation)
399 * @transport_header: Transport layer header
400 * @network_header: Network layer header
401 * @mac_header: Link layer header
402 * @tail: Tail pointer
403 * @end: End pointer
404 * @head: Head of buffer
405 * @data: Data head pointer
406 * @truesize: Buffer size
407 * @users: User count - see {datagram,tcp}.c
410 struct sk_buff {
411 /* These two members must be first. */
412 struct sk_buff *next;
413 struct sk_buff *prev;
415 ktime_t tstamp;
417 struct sock *sk;
418 struct net_device *dev;
421 * This is the control buffer. It is free to use for every
422 * layer. Please put your private variables there. If you
423 * want to keep them across layers you have to do a skb_clone()
424 * first. This is owned by whoever has the skb queued ATM.
426 char cb[48] __aligned(8);
428 unsigned long _skb_refdst;
429 #ifdef CONFIG_XFRM
430 struct sec_path *sp;
431 #endif
432 unsigned int len,
433 data_len;
434 __u16 mac_len,
435 hdr_len;
436 union {
437 __wsum csum;
438 struct {
439 __u16 csum_start;
440 __u16 csum_offset;
443 __u32 priority;
444 kmemcheck_bitfield_begin(flags1);
445 __u8 local_df:1,
446 cloned:1,
447 ip_summed:2,
448 nohdr:1,
449 nfctinfo:3;
450 __u8 pkt_type:3,
451 fclone:2,
452 ipvs_property:1,
453 peeked:1,
454 nf_trace:1;
455 kmemcheck_bitfield_end(flags1);
456 __be16 protocol;
458 void (*destructor)(struct sk_buff *skb);
459 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
460 struct nf_conntrack *nfct;
461 #endif
462 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
463 struct sk_buff *nfct_reasm;
464 #endif
465 #ifdef CONFIG_BRIDGE_NETFILTER
466 struct nf_bridge_info *nf_bridge;
467 #endif
469 int skb_iif;
471 __u32 rxhash;
473 __be16 vlan_proto;
474 __u16 vlan_tci;
476 #ifdef CONFIG_NET_SCHED
477 __u16 tc_index; /* traffic control index */
478 #ifdef CONFIG_NET_CLS_ACT
479 __u16 tc_verd; /* traffic control verdict */
480 #endif
481 #endif
483 __u16 queue_mapping;
484 kmemcheck_bitfield_begin(flags2);
485 #ifdef CONFIG_IPV6_NDISC_NODETYPE
486 __u8 ndisc_nodetype:2;
487 #endif
488 __u8 pfmemalloc:1;
489 __u8 ooo_okay:1;
490 __u8 l4_rxhash:1;
491 __u8 wifi_acked_valid:1;
492 __u8 wifi_acked:1;
493 __u8 no_fcs:1;
494 __u8 head_frag:1;
495 /* Encapsulation protocol and NIC drivers should use
496 * this flag to indicate to each other if the skb contains
497 * encapsulated packet or not and maybe use the inner packet
498 * headers if needed
500 __u8 encapsulation:1;
501 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
502 kmemcheck_bitfield_end(flags2);
504 #if defined CONFIG_NET_DMA || defined CONFIG_NET_LL_RX_POLL
505 union {
506 unsigned int napi_id;
507 dma_cookie_t dma_cookie;
509 #endif
510 #ifdef CONFIG_NETWORK_SECMARK
511 __u32 secmark;
512 #endif
513 union {
514 __u32 mark;
515 __u32 dropcount;
516 __u32 reserved_tailroom;
519 __be16 inner_protocol;
520 __u16 inner_transport_header;
521 __u16 inner_network_header;
522 __u16 inner_mac_header;
523 __u16 transport_header;
524 __u16 network_header;
525 __u16 mac_header;
526 /* These elements must be at the end, see alloc_skb() for details. */
527 sk_buff_data_t tail;
528 sk_buff_data_t end;
529 unsigned char *head,
530 *data;
531 unsigned int truesize;
532 atomic_t users;
535 #ifdef __KERNEL__
537 * Handling routines are only of interest to the kernel
539 #include <linux/slab.h>
542 #define SKB_ALLOC_FCLONE 0x01
543 #define SKB_ALLOC_RX 0x02
545 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
546 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
548 return unlikely(skb->pfmemalloc);
552 * skb might have a dst pointer attached, refcounted or not.
553 * _skb_refdst low order bit is set if refcount was _not_ taken
555 #define SKB_DST_NOREF 1UL
556 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
559 * skb_dst - returns skb dst_entry
560 * @skb: buffer
562 * Returns skb dst_entry, regardless of reference taken or not.
564 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
566 /* If refdst was not refcounted, check we still are in a
567 * rcu_read_lock section
569 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
570 !rcu_read_lock_held() &&
571 !rcu_read_lock_bh_held());
572 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
576 * skb_dst_set - sets skb dst
577 * @skb: buffer
578 * @dst: dst entry
580 * Sets skb dst, assuming a reference was taken on dst and should
581 * be released by skb_dst_drop()
583 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
585 skb->_skb_refdst = (unsigned long)dst;
588 extern void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
589 bool force);
592 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
593 * @skb: buffer
594 * @dst: dst entry
596 * Sets skb dst, assuming a reference was not taken on dst.
597 * If dst entry is cached, we do not take reference and dst_release
598 * will be avoided by refdst_drop. If dst entry is not cached, we take
599 * reference, so that last dst_release can destroy the dst immediately.
601 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
603 __skb_dst_set_noref(skb, dst, false);
607 * skb_dst_set_noref_force - sets skb dst, without taking reference
608 * @skb: buffer
609 * @dst: dst entry
611 * Sets skb dst, assuming a reference was not taken on dst.
612 * No reference is taken and no dst_release will be called. While for
613 * cached dsts deferred reclaim is a basic feature, for entries that are
614 * not cached it is caller's job to guarantee that last dst_release for
615 * provided dst happens when nobody uses it, eg. after a RCU grace period.
617 static inline void skb_dst_set_noref_force(struct sk_buff *skb,
618 struct dst_entry *dst)
620 __skb_dst_set_noref(skb, dst, true);
624 * skb_dst_is_noref - Test if skb dst isn't refcounted
625 * @skb: buffer
627 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
629 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
632 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
634 return (struct rtable *)skb_dst(skb);
637 extern void kfree_skb(struct sk_buff *skb);
638 extern void kfree_skb_list(struct sk_buff *segs);
639 extern void skb_tx_error(struct sk_buff *skb);
640 extern void consume_skb(struct sk_buff *skb);
641 extern void __kfree_skb(struct sk_buff *skb);
642 extern struct kmem_cache *skbuff_head_cache;
644 extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
645 extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
646 bool *fragstolen, int *delta_truesize);
648 extern struct sk_buff *__alloc_skb(unsigned int size,
649 gfp_t priority, int flags, int node);
650 extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
651 static inline struct sk_buff *alloc_skb(unsigned int size,
652 gfp_t priority)
654 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
657 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
658 gfp_t priority)
660 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
663 extern struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
664 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
666 return __alloc_skb_head(priority, -1);
669 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
670 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
671 extern struct sk_buff *skb_clone(struct sk_buff *skb,
672 gfp_t priority);
673 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
674 gfp_t priority);
675 extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
676 int headroom, gfp_t gfp_mask);
678 extern int pskb_expand_head(struct sk_buff *skb,
679 int nhead, int ntail,
680 gfp_t gfp_mask);
681 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
682 unsigned int headroom);
683 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
684 int newheadroom, int newtailroom,
685 gfp_t priority);
686 extern int skb_to_sgvec(struct sk_buff *skb,
687 struct scatterlist *sg, int offset,
688 int len);
689 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
690 struct sk_buff **trailer);
691 extern int skb_pad(struct sk_buff *skb, int pad);
692 #define dev_kfree_skb(a) consume_skb(a)
694 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
695 int getfrag(void *from, char *to, int offset,
696 int len,int odd, struct sk_buff *skb),
697 void *from, int length);
699 struct skb_seq_state {
700 __u32 lower_offset;
701 __u32 upper_offset;
702 __u32 frag_idx;
703 __u32 stepped_offset;
704 struct sk_buff *root_skb;
705 struct sk_buff *cur_skb;
706 __u8 *frag_data;
709 extern void skb_prepare_seq_read(struct sk_buff *skb,
710 unsigned int from, unsigned int to,
711 struct skb_seq_state *st);
712 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
713 struct skb_seq_state *st);
714 extern void skb_abort_seq_read(struct skb_seq_state *st);
716 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
717 unsigned int to, struct ts_config *config,
718 struct ts_state *state);
720 extern void __skb_get_rxhash(struct sk_buff *skb);
721 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
723 if (!skb->l4_rxhash)
724 __skb_get_rxhash(skb);
726 return skb->rxhash;
729 #ifdef NET_SKBUFF_DATA_USES_OFFSET
730 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
732 return skb->head + skb->end;
735 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
737 return skb->end;
739 #else
740 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
742 return skb->end;
745 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
747 return skb->end - skb->head;
749 #endif
751 /* Internal */
752 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
754 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
756 return &skb_shinfo(skb)->hwtstamps;
760 * skb_queue_empty - check if a queue is empty
761 * @list: queue head
763 * Returns true if the queue is empty, false otherwise.
765 static inline int skb_queue_empty(const struct sk_buff_head *list)
767 return list->next == (struct sk_buff *)list;
771 * skb_queue_is_last - check if skb is the last entry in the queue
772 * @list: queue head
773 * @skb: buffer
775 * Returns true if @skb is the last buffer on the list.
777 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
778 const struct sk_buff *skb)
780 return skb->next == (struct sk_buff *)list;
784 * skb_queue_is_first - check if skb is the first entry in the queue
785 * @list: queue head
786 * @skb: buffer
788 * Returns true if @skb is the first buffer on the list.
790 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
791 const struct sk_buff *skb)
793 return skb->prev == (struct sk_buff *)list;
797 * skb_queue_next - return the next packet in the queue
798 * @list: queue head
799 * @skb: current buffer
801 * Return the next packet in @list after @skb. It is only valid to
802 * call this if skb_queue_is_last() evaluates to false.
804 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
805 const struct sk_buff *skb)
807 /* This BUG_ON may seem severe, but if we just return then we
808 * are going to dereference garbage.
810 BUG_ON(skb_queue_is_last(list, skb));
811 return skb->next;
815 * skb_queue_prev - return the prev packet in the queue
816 * @list: queue head
817 * @skb: current buffer
819 * Return the prev packet in @list before @skb. It is only valid to
820 * call this if skb_queue_is_first() evaluates to false.
822 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
823 const struct sk_buff *skb)
825 /* This BUG_ON may seem severe, but if we just return then we
826 * are going to dereference garbage.
828 BUG_ON(skb_queue_is_first(list, skb));
829 return skb->prev;
833 * skb_get - reference buffer
834 * @skb: buffer to reference
836 * Makes another reference to a socket buffer and returns a pointer
837 * to the buffer.
839 static inline struct sk_buff *skb_get(struct sk_buff *skb)
841 atomic_inc(&skb->users);
842 return skb;
846 * If users == 1, we are the only owner and are can avoid redundant
847 * atomic change.
851 * skb_cloned - is the buffer a clone
852 * @skb: buffer to check
854 * Returns true if the buffer was generated with skb_clone() and is
855 * one of multiple shared copies of the buffer. Cloned buffers are
856 * shared data so must not be written to under normal circumstances.
858 static inline int skb_cloned(const struct sk_buff *skb)
860 return skb->cloned &&
861 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
864 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
866 might_sleep_if(pri & __GFP_WAIT);
868 if (skb_cloned(skb))
869 return pskb_expand_head(skb, 0, 0, pri);
871 return 0;
875 * skb_header_cloned - is the header a clone
876 * @skb: buffer to check
878 * Returns true if modifying the header part of the buffer requires
879 * the data to be copied.
881 static inline int skb_header_cloned(const struct sk_buff *skb)
883 int dataref;
885 if (!skb->cloned)
886 return 0;
888 dataref = atomic_read(&skb_shinfo(skb)->dataref);
889 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
890 return dataref != 1;
894 * skb_header_release - release reference to header
895 * @skb: buffer to operate on
897 * Drop a reference to the header part of the buffer. This is done
898 * by acquiring a payload reference. You must not read from the header
899 * part of skb->data after this.
901 static inline void skb_header_release(struct sk_buff *skb)
903 BUG_ON(skb->nohdr);
904 skb->nohdr = 1;
905 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
909 * skb_shared - is the buffer shared
910 * @skb: buffer to check
912 * Returns true if more than one person has a reference to this
913 * buffer.
915 static inline int skb_shared(const struct sk_buff *skb)
917 return atomic_read(&skb->users) != 1;
921 * skb_share_check - check if buffer is shared and if so clone it
922 * @skb: buffer to check
923 * @pri: priority for memory allocation
925 * If the buffer is shared the buffer is cloned and the old copy
926 * drops a reference. A new clone with a single reference is returned.
927 * If the buffer is not shared the original buffer is returned. When
928 * being called from interrupt status or with spinlocks held pri must
929 * be GFP_ATOMIC.
931 * NULL is returned on a memory allocation failure.
933 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
935 might_sleep_if(pri & __GFP_WAIT);
936 if (skb_shared(skb)) {
937 struct sk_buff *nskb = skb_clone(skb, pri);
939 if (likely(nskb))
940 consume_skb(skb);
941 else
942 kfree_skb(skb);
943 skb = nskb;
945 return skb;
949 * Copy shared buffers into a new sk_buff. We effectively do COW on
950 * packets to handle cases where we have a local reader and forward
951 * and a couple of other messy ones. The normal one is tcpdumping
952 * a packet thats being forwarded.
956 * skb_unshare - make a copy of a shared buffer
957 * @skb: buffer to check
958 * @pri: priority for memory allocation
960 * If the socket buffer is a clone then this function creates a new
961 * copy of the data, drops a reference count on the old copy and returns
962 * the new copy with the reference count at 1. If the buffer is not a clone
963 * the original buffer is returned. When called with a spinlock held or
964 * from interrupt state @pri must be %GFP_ATOMIC
966 * %NULL is returned on a memory allocation failure.
968 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
969 gfp_t pri)
971 might_sleep_if(pri & __GFP_WAIT);
972 if (skb_cloned(skb)) {
973 struct sk_buff *nskb = skb_copy(skb, pri);
974 kfree_skb(skb); /* Free our shared copy */
975 skb = nskb;
977 return skb;
981 * skb_peek - peek at the head of an &sk_buff_head
982 * @list_: list to peek at
984 * Peek an &sk_buff. Unlike most other operations you _MUST_
985 * be careful with this one. A peek leaves the buffer on the
986 * list and someone else may run off with it. You must hold
987 * the appropriate locks or have a private queue to do this.
989 * Returns %NULL for an empty list or a pointer to the head element.
990 * The reference count is not incremented and the reference is therefore
991 * volatile. Use with caution.
993 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
995 struct sk_buff *skb = list_->next;
997 if (skb == (struct sk_buff *)list_)
998 skb = NULL;
999 return skb;
1003 * skb_peek_next - peek skb following the given one from a queue
1004 * @skb: skb to start from
1005 * @list_: list to peek at
1007 * Returns %NULL when the end of the list is met or a pointer to the
1008 * next element. The reference count is not incremented and the
1009 * reference is therefore volatile. Use with caution.
1011 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1012 const struct sk_buff_head *list_)
1014 struct sk_buff *next = skb->next;
1016 if (next == (struct sk_buff *)list_)
1017 next = NULL;
1018 return next;
1022 * skb_peek_tail - peek at the tail of an &sk_buff_head
1023 * @list_: list to peek at
1025 * Peek an &sk_buff. Unlike most other operations you _MUST_
1026 * be careful with this one. A peek leaves the buffer on the
1027 * list and someone else may run off with it. You must hold
1028 * the appropriate locks or have a private queue to do this.
1030 * Returns %NULL for an empty list or a pointer to the tail element.
1031 * The reference count is not incremented and the reference is therefore
1032 * volatile. Use with caution.
1034 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1036 struct sk_buff *skb = list_->prev;
1038 if (skb == (struct sk_buff *)list_)
1039 skb = NULL;
1040 return skb;
1045 * skb_queue_len - get queue length
1046 * @list_: list to measure
1048 * Return the length of an &sk_buff queue.
1050 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1052 return list_->qlen;
1056 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1057 * @list: queue to initialize
1059 * This initializes only the list and queue length aspects of
1060 * an sk_buff_head object. This allows to initialize the list
1061 * aspects of an sk_buff_head without reinitializing things like
1062 * the spinlock. It can also be used for on-stack sk_buff_head
1063 * objects where the spinlock is known to not be used.
1065 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1067 list->prev = list->next = (struct sk_buff *)list;
1068 list->qlen = 0;
1072 * This function creates a split out lock class for each invocation;
1073 * this is needed for now since a whole lot of users of the skb-queue
1074 * infrastructure in drivers have different locking usage (in hardirq)
1075 * than the networking core (in softirq only). In the long run either the
1076 * network layer or drivers should need annotation to consolidate the
1077 * main types of usage into 3 classes.
1079 static inline void skb_queue_head_init(struct sk_buff_head *list)
1081 spin_lock_init(&list->lock);
1082 __skb_queue_head_init(list);
1085 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1086 struct lock_class_key *class)
1088 skb_queue_head_init(list);
1089 lockdep_set_class(&list->lock, class);
1093 * Insert an sk_buff on a list.
1095 * The "__skb_xxxx()" functions are the non-atomic ones that
1096 * can only be called with interrupts disabled.
1098 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1099 static inline void __skb_insert(struct sk_buff *newsk,
1100 struct sk_buff *prev, struct sk_buff *next,
1101 struct sk_buff_head *list)
1103 newsk->next = next;
1104 newsk->prev = prev;
1105 next->prev = prev->next = newsk;
1106 list->qlen++;
1109 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1110 struct sk_buff *prev,
1111 struct sk_buff *next)
1113 struct sk_buff *first = list->next;
1114 struct sk_buff *last = list->prev;
1116 first->prev = prev;
1117 prev->next = first;
1119 last->next = next;
1120 next->prev = last;
1124 * skb_queue_splice - join two skb lists, this is designed for stacks
1125 * @list: the new list to add
1126 * @head: the place to add it in the first list
1128 static inline void skb_queue_splice(const struct sk_buff_head *list,
1129 struct sk_buff_head *head)
1131 if (!skb_queue_empty(list)) {
1132 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1133 head->qlen += list->qlen;
1138 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1139 * @list: the new list to add
1140 * @head: the place to add it in the first list
1142 * The list at @list is reinitialised
1144 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1145 struct sk_buff_head *head)
1147 if (!skb_queue_empty(list)) {
1148 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1149 head->qlen += list->qlen;
1150 __skb_queue_head_init(list);
1155 * skb_queue_splice_tail - join two skb lists, each list being a queue
1156 * @list: the new list to add
1157 * @head: the place to add it in the first list
1159 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1160 struct sk_buff_head *head)
1162 if (!skb_queue_empty(list)) {
1163 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1164 head->qlen += list->qlen;
1169 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1170 * @list: the new list to add
1171 * @head: the place to add it in the first list
1173 * Each of the lists is a queue.
1174 * The list at @list is reinitialised
1176 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1177 struct sk_buff_head *head)
1179 if (!skb_queue_empty(list)) {
1180 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1181 head->qlen += list->qlen;
1182 __skb_queue_head_init(list);
1187 * __skb_queue_after - queue a buffer at the list head
1188 * @list: list to use
1189 * @prev: place after this buffer
1190 * @newsk: buffer to queue
1192 * Queue a buffer int the middle of a list. This function takes no locks
1193 * and you must therefore hold required locks before calling it.
1195 * A buffer cannot be placed on two lists at the same time.
1197 static inline void __skb_queue_after(struct sk_buff_head *list,
1198 struct sk_buff *prev,
1199 struct sk_buff *newsk)
1201 __skb_insert(newsk, prev, prev->next, list);
1204 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1205 struct sk_buff_head *list);
1207 static inline void __skb_queue_before(struct sk_buff_head *list,
1208 struct sk_buff *next,
1209 struct sk_buff *newsk)
1211 __skb_insert(newsk, next->prev, next, list);
1215 * __skb_queue_head - queue a buffer at the list head
1216 * @list: list to use
1217 * @newsk: buffer to queue
1219 * Queue a buffer at the start of a list. This function takes no locks
1220 * and you must therefore hold required locks before calling it.
1222 * A buffer cannot be placed on two lists at the same time.
1224 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1225 static inline void __skb_queue_head(struct sk_buff_head *list,
1226 struct sk_buff *newsk)
1228 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1232 * __skb_queue_tail - queue a buffer at the list tail
1233 * @list: list to use
1234 * @newsk: buffer to queue
1236 * Queue a buffer at the end of a list. This function takes no locks
1237 * and you must therefore hold required locks before calling it.
1239 * A buffer cannot be placed on two lists at the same time.
1241 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1242 static inline void __skb_queue_tail(struct sk_buff_head *list,
1243 struct sk_buff *newsk)
1245 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1249 * remove sk_buff from list. _Must_ be called atomically, and with
1250 * the list known..
1252 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1253 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1255 struct sk_buff *next, *prev;
1257 list->qlen--;
1258 next = skb->next;
1259 prev = skb->prev;
1260 skb->next = skb->prev = NULL;
1261 next->prev = prev;
1262 prev->next = next;
1266 * __skb_dequeue - remove from the head of the queue
1267 * @list: list to dequeue from
1269 * Remove the head of the list. This function does not take any locks
1270 * so must be used with appropriate locks held only. The head item is
1271 * returned or %NULL if the list is empty.
1273 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1274 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1276 struct sk_buff *skb = skb_peek(list);
1277 if (skb)
1278 __skb_unlink(skb, list);
1279 return skb;
1283 * __skb_dequeue_tail - remove from the tail of the queue
1284 * @list: list to dequeue from
1286 * Remove the tail of the list. This function does not take any locks
1287 * so must be used with appropriate locks held only. The tail item is
1288 * returned or %NULL if the list is empty.
1290 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1291 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1293 struct sk_buff *skb = skb_peek_tail(list);
1294 if (skb)
1295 __skb_unlink(skb, list);
1296 return skb;
1300 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1302 return skb->data_len;
1305 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1307 return skb->len - skb->data_len;
1310 static inline int skb_pagelen(const struct sk_buff *skb)
1312 int i, len = 0;
1314 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1315 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1316 return len + skb_headlen(skb);
1320 * __skb_fill_page_desc - initialise a paged fragment in an skb
1321 * @skb: buffer containing fragment to be initialised
1322 * @i: paged fragment index to initialise
1323 * @page: the page to use for this fragment
1324 * @off: the offset to the data with @page
1325 * @size: the length of the data
1327 * Initialises the @i'th fragment of @skb to point to &size bytes at
1328 * offset @off within @page.
1330 * Does not take any additional reference on the fragment.
1332 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1333 struct page *page, int off, int size)
1335 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1338 * Propagate page->pfmemalloc to the skb if we can. The problem is
1339 * that not all callers have unique ownership of the page. If
1340 * pfmemalloc is set, we check the mapping as a mapping implies
1341 * page->index is set (index and pfmemalloc share space).
1342 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1343 * do not lose pfmemalloc information as the pages would not be
1344 * allocated using __GFP_MEMALLOC.
1346 frag->page.p = page;
1347 frag->page_offset = off;
1348 skb_frag_size_set(frag, size);
1350 page = compound_head(page);
1351 if (page->pfmemalloc && !page->mapping)
1352 skb->pfmemalloc = true;
1356 * skb_fill_page_desc - initialise a paged fragment in an skb
1357 * @skb: buffer containing fragment to be initialised
1358 * @i: paged fragment index to initialise
1359 * @page: the page to use for this fragment
1360 * @off: the offset to the data with @page
1361 * @size: the length of the data
1363 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1364 * @skb to point to &size bytes at offset @off within @page. In
1365 * addition updates @skb such that @i is the last fragment.
1367 * Does not take any additional reference on the fragment.
1369 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1370 struct page *page, int off, int size)
1372 __skb_fill_page_desc(skb, i, page, off, size);
1373 skb_shinfo(skb)->nr_frags = i + 1;
1376 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1377 int off, int size, unsigned int truesize);
1379 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1380 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1381 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1383 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1384 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1386 return skb->head + skb->tail;
1389 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1391 skb->tail = skb->data - skb->head;
1394 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1396 skb_reset_tail_pointer(skb);
1397 skb->tail += offset;
1400 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1401 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1403 return skb->tail;
1406 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1408 skb->tail = skb->data;
1411 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1413 skb->tail = skb->data + offset;
1416 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1419 * Add data to an sk_buff
1421 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1422 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1424 unsigned char *tmp = skb_tail_pointer(skb);
1425 SKB_LINEAR_ASSERT(skb);
1426 skb->tail += len;
1427 skb->len += len;
1428 return tmp;
1431 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1432 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1434 skb->data -= len;
1435 skb->len += len;
1436 return skb->data;
1439 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1440 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1442 skb->len -= len;
1443 BUG_ON(skb->len < skb->data_len);
1444 return skb->data += len;
1447 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1449 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1452 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1454 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1456 if (len > skb_headlen(skb) &&
1457 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1458 return NULL;
1459 skb->len -= len;
1460 return skb->data += len;
1463 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1465 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1468 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1470 if (likely(len <= skb_headlen(skb)))
1471 return 1;
1472 if (unlikely(len > skb->len))
1473 return 0;
1474 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1478 * skb_headroom - bytes at buffer head
1479 * @skb: buffer to check
1481 * Return the number of bytes of free space at the head of an &sk_buff.
1483 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1485 return skb->data - skb->head;
1489 * skb_tailroom - bytes at buffer end
1490 * @skb: buffer to check
1492 * Return the number of bytes of free space at the tail of an sk_buff
1494 static inline int skb_tailroom(const struct sk_buff *skb)
1496 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1500 * skb_availroom - bytes at buffer end
1501 * @skb: buffer to check
1503 * Return the number of bytes of free space at the tail of an sk_buff
1504 * allocated by sk_stream_alloc()
1506 static inline int skb_availroom(const struct sk_buff *skb)
1508 if (skb_is_nonlinear(skb))
1509 return 0;
1511 return skb->end - skb->tail - skb->reserved_tailroom;
1515 * skb_reserve - adjust headroom
1516 * @skb: buffer to alter
1517 * @len: bytes to move
1519 * Increase the headroom of an empty &sk_buff by reducing the tail
1520 * room. This is only allowed for an empty buffer.
1522 static inline void skb_reserve(struct sk_buff *skb, int len)
1524 skb->data += len;
1525 skb->tail += len;
1528 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1530 skb->inner_mac_header = skb->mac_header;
1531 skb->inner_network_header = skb->network_header;
1532 skb->inner_transport_header = skb->transport_header;
1535 static inline void skb_reset_mac_len(struct sk_buff *skb)
1537 skb->mac_len = skb->network_header - skb->mac_header;
1540 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1541 *skb)
1543 return skb->head + skb->inner_transport_header;
1546 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1548 skb->inner_transport_header = skb->data - skb->head;
1551 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1552 const int offset)
1554 skb_reset_inner_transport_header(skb);
1555 skb->inner_transport_header += offset;
1558 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1560 return skb->head + skb->inner_network_header;
1563 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1565 skb->inner_network_header = skb->data - skb->head;
1568 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1569 const int offset)
1571 skb_reset_inner_network_header(skb);
1572 skb->inner_network_header += offset;
1575 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1577 return skb->head + skb->inner_mac_header;
1580 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1582 skb->inner_mac_header = skb->data - skb->head;
1585 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1586 const int offset)
1588 skb_reset_inner_mac_header(skb);
1589 skb->inner_mac_header += offset;
1591 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1593 return skb->transport_header != (typeof(skb->transport_header))~0U;
1596 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1598 return skb->head + skb->transport_header;
1601 static inline void skb_reset_transport_header(struct sk_buff *skb)
1603 skb->transport_header = skb->data - skb->head;
1606 static inline void skb_set_transport_header(struct sk_buff *skb,
1607 const int offset)
1609 skb_reset_transport_header(skb);
1610 skb->transport_header += offset;
1613 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1615 return skb->head + skb->network_header;
1618 static inline void skb_reset_network_header(struct sk_buff *skb)
1620 skb->network_header = skb->data - skb->head;
1623 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1625 skb_reset_network_header(skb);
1626 skb->network_header += offset;
1629 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1631 return skb->head + skb->mac_header;
1634 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1636 return skb->mac_header != (typeof(skb->mac_header))~0U;
1639 static inline void skb_reset_mac_header(struct sk_buff *skb)
1641 skb->mac_header = skb->data - skb->head;
1644 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1646 skb_reset_mac_header(skb);
1647 skb->mac_header += offset;
1650 static inline void skb_probe_transport_header(struct sk_buff *skb,
1651 const int offset_hint)
1653 struct flow_keys keys;
1655 if (skb_transport_header_was_set(skb))
1656 return;
1657 else if (skb_flow_dissect(skb, &keys))
1658 skb_set_transport_header(skb, keys.thoff);
1659 else
1660 skb_set_transport_header(skb, offset_hint);
1663 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1665 if (skb_mac_header_was_set(skb)) {
1666 const unsigned char *old_mac = skb_mac_header(skb);
1668 skb_set_mac_header(skb, -skb->mac_len);
1669 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1673 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1675 return skb->csum_start - skb_headroom(skb);
1678 static inline int skb_transport_offset(const struct sk_buff *skb)
1680 return skb_transport_header(skb) - skb->data;
1683 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1685 return skb->transport_header - skb->network_header;
1688 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1690 return skb->inner_transport_header - skb->inner_network_header;
1693 static inline int skb_network_offset(const struct sk_buff *skb)
1695 return skb_network_header(skb) - skb->data;
1698 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1700 return skb_inner_network_header(skb) - skb->data;
1703 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1705 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1709 * CPUs often take a performance hit when accessing unaligned memory
1710 * locations. The actual performance hit varies, it can be small if the
1711 * hardware handles it or large if we have to take an exception and fix it
1712 * in software.
1714 * Since an ethernet header is 14 bytes network drivers often end up with
1715 * the IP header at an unaligned offset. The IP header can be aligned by
1716 * shifting the start of the packet by 2 bytes. Drivers should do this
1717 * with:
1719 * skb_reserve(skb, NET_IP_ALIGN);
1721 * The downside to this alignment of the IP header is that the DMA is now
1722 * unaligned. On some architectures the cost of an unaligned DMA is high
1723 * and this cost outweighs the gains made by aligning the IP header.
1725 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1726 * to be overridden.
1728 #ifndef NET_IP_ALIGN
1729 #define NET_IP_ALIGN 2
1730 #endif
1733 * The networking layer reserves some headroom in skb data (via
1734 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1735 * the header has to grow. In the default case, if the header has to grow
1736 * 32 bytes or less we avoid the reallocation.
1738 * Unfortunately this headroom changes the DMA alignment of the resulting
1739 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1740 * on some architectures. An architecture can override this value,
1741 * perhaps setting it to a cacheline in size (since that will maintain
1742 * cacheline alignment of the DMA). It must be a power of 2.
1744 * Various parts of the networking layer expect at least 32 bytes of
1745 * headroom, you should not reduce this.
1747 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1748 * to reduce average number of cache lines per packet.
1749 * get_rps_cpus() for example only access one 64 bytes aligned block :
1750 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1752 #ifndef NET_SKB_PAD
1753 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1754 #endif
1756 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1758 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1760 if (unlikely(skb_is_nonlinear(skb))) {
1761 WARN_ON(1);
1762 return;
1764 skb->len = len;
1765 skb_set_tail_pointer(skb, len);
1768 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1770 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1772 if (skb->data_len)
1773 return ___pskb_trim(skb, len);
1774 __skb_trim(skb, len);
1775 return 0;
1778 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1780 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1784 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1785 * @skb: buffer to alter
1786 * @len: new length
1788 * This is identical to pskb_trim except that the caller knows that
1789 * the skb is not cloned so we should never get an error due to out-
1790 * of-memory.
1792 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1794 int err = pskb_trim(skb, len);
1795 BUG_ON(err);
1799 * skb_orphan - orphan a buffer
1800 * @skb: buffer to orphan
1802 * If a buffer currently has an owner then we call the owner's
1803 * destructor function and make the @skb unowned. The buffer continues
1804 * to exist but is no longer charged to its former owner.
1806 static inline void skb_orphan(struct sk_buff *skb)
1808 if (skb->destructor)
1809 skb->destructor(skb);
1810 skb->destructor = NULL;
1811 skb->sk = NULL;
1815 * skb_orphan_frags - orphan the frags contained in a buffer
1816 * @skb: buffer to orphan frags from
1817 * @gfp_mask: allocation mask for replacement pages
1819 * For each frag in the SKB which needs a destructor (i.e. has an
1820 * owner) create a copy of that frag and release the original
1821 * page by calling the destructor.
1823 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1825 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1826 return 0;
1827 return skb_copy_ubufs(skb, gfp_mask);
1831 * __skb_queue_purge - empty a list
1832 * @list: list to empty
1834 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1835 * the list and one reference dropped. This function does not take the
1836 * list lock and the caller must hold the relevant locks to use it.
1838 extern void skb_queue_purge(struct sk_buff_head *list);
1839 static inline void __skb_queue_purge(struct sk_buff_head *list)
1841 struct sk_buff *skb;
1842 while ((skb = __skb_dequeue(list)) != NULL)
1843 kfree_skb(skb);
1846 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1847 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1848 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1850 extern void *netdev_alloc_frag(unsigned int fragsz);
1852 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1853 unsigned int length,
1854 gfp_t gfp_mask);
1857 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1858 * @dev: network device to receive on
1859 * @length: length to allocate
1861 * Allocate a new &sk_buff and assign it a usage count of one. The
1862 * buffer has unspecified headroom built in. Users should allocate
1863 * the headroom they think they need without accounting for the
1864 * built in space. The built in space is used for optimisations.
1866 * %NULL is returned if there is no free memory. Although this function
1867 * allocates memory it can be called from an interrupt.
1869 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1870 unsigned int length)
1872 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1875 /* legacy helper around __netdev_alloc_skb() */
1876 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1877 gfp_t gfp_mask)
1879 return __netdev_alloc_skb(NULL, length, gfp_mask);
1882 /* legacy helper around netdev_alloc_skb() */
1883 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1885 return netdev_alloc_skb(NULL, length);
1889 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1890 unsigned int length, gfp_t gfp)
1892 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1894 if (NET_IP_ALIGN && skb)
1895 skb_reserve(skb, NET_IP_ALIGN);
1896 return skb;
1899 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1900 unsigned int length)
1902 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1906 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1907 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1908 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1909 * @order: size of the allocation
1911 * Allocate a new page.
1913 * %NULL is returned if there is no free memory.
1915 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1916 struct sk_buff *skb,
1917 unsigned int order)
1919 struct page *page;
1921 gfp_mask |= __GFP_COLD;
1923 if (!(gfp_mask & __GFP_NOMEMALLOC))
1924 gfp_mask |= __GFP_MEMALLOC;
1926 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1927 if (skb && page && page->pfmemalloc)
1928 skb->pfmemalloc = true;
1930 return page;
1934 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1935 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1936 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1938 * Allocate a new page.
1940 * %NULL is returned if there is no free memory.
1942 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1943 struct sk_buff *skb)
1945 return __skb_alloc_pages(gfp_mask, skb, 0);
1949 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1950 * @page: The page that was allocated from skb_alloc_page
1951 * @skb: The skb that may need pfmemalloc set
1953 static inline void skb_propagate_pfmemalloc(struct page *page,
1954 struct sk_buff *skb)
1956 if (page && page->pfmemalloc)
1957 skb->pfmemalloc = true;
1961 * skb_frag_page - retrieve the page refered to by a paged fragment
1962 * @frag: the paged fragment
1964 * Returns the &struct page associated with @frag.
1966 static inline struct page *skb_frag_page(const skb_frag_t *frag)
1968 return frag->page.p;
1972 * __skb_frag_ref - take an addition reference on a paged fragment.
1973 * @frag: the paged fragment
1975 * Takes an additional reference on the paged fragment @frag.
1977 static inline void __skb_frag_ref(skb_frag_t *frag)
1979 get_page(skb_frag_page(frag));
1983 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1984 * @skb: the buffer
1985 * @f: the fragment offset.
1987 * Takes an additional reference on the @f'th paged fragment of @skb.
1989 static inline void skb_frag_ref(struct sk_buff *skb, int f)
1991 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1995 * __skb_frag_unref - release a reference on a paged fragment.
1996 * @frag: the paged fragment
1998 * Releases a reference on the paged fragment @frag.
2000 static inline void __skb_frag_unref(skb_frag_t *frag)
2002 put_page(skb_frag_page(frag));
2006 * skb_frag_unref - release a reference on a paged fragment of an skb.
2007 * @skb: the buffer
2008 * @f: the fragment offset
2010 * Releases a reference on the @f'th paged fragment of @skb.
2012 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2014 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2018 * skb_frag_address - gets the address of the data contained in a paged fragment
2019 * @frag: the paged fragment buffer
2021 * Returns the address of the data within @frag. The page must already
2022 * be mapped.
2024 static inline void *skb_frag_address(const skb_frag_t *frag)
2026 return page_address(skb_frag_page(frag)) + frag->page_offset;
2030 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2031 * @frag: the paged fragment buffer
2033 * Returns the address of the data within @frag. Checks that the page
2034 * is mapped and returns %NULL otherwise.
2036 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2038 void *ptr = page_address(skb_frag_page(frag));
2039 if (unlikely(!ptr))
2040 return NULL;
2042 return ptr + frag->page_offset;
2046 * __skb_frag_set_page - sets the page contained in a paged fragment
2047 * @frag: the paged fragment
2048 * @page: the page to set
2050 * Sets the fragment @frag to contain @page.
2052 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2054 frag->page.p = page;
2058 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2059 * @skb: the buffer
2060 * @f: the fragment offset
2061 * @page: the page to set
2063 * Sets the @f'th fragment of @skb to contain @page.
2065 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2066 struct page *page)
2068 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2072 * skb_frag_dma_map - maps a paged fragment via the DMA API
2073 * @dev: the device to map the fragment to
2074 * @frag: the paged fragment to map
2075 * @offset: the offset within the fragment (starting at the
2076 * fragment's own offset)
2077 * @size: the number of bytes to map
2078 * @dir: the direction of the mapping (%PCI_DMA_*)
2080 * Maps the page associated with @frag to @device.
2082 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2083 const skb_frag_t *frag,
2084 size_t offset, size_t size,
2085 enum dma_data_direction dir)
2087 return dma_map_page(dev, skb_frag_page(frag),
2088 frag->page_offset + offset, size, dir);
2091 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2092 gfp_t gfp_mask)
2094 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2098 * skb_clone_writable - is the header of a clone writable
2099 * @skb: buffer to check
2100 * @len: length up to which to write
2102 * Returns true if modifying the header part of the cloned buffer
2103 * does not requires the data to be copied.
2105 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2107 return !skb_header_cloned(skb) &&
2108 skb_headroom(skb) + len <= skb->hdr_len;
2111 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2112 int cloned)
2114 int delta = 0;
2116 if (headroom > skb_headroom(skb))
2117 delta = headroom - skb_headroom(skb);
2119 if (delta || cloned)
2120 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2121 GFP_ATOMIC);
2122 return 0;
2126 * skb_cow - copy header of skb when it is required
2127 * @skb: buffer to cow
2128 * @headroom: needed headroom
2130 * If the skb passed lacks sufficient headroom or its data part
2131 * is shared, data is reallocated. If reallocation fails, an error
2132 * is returned and original skb is not changed.
2134 * The result is skb with writable area skb->head...skb->tail
2135 * and at least @headroom of space at head.
2137 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2139 return __skb_cow(skb, headroom, skb_cloned(skb));
2143 * skb_cow_head - skb_cow but only making the head writable
2144 * @skb: buffer to cow
2145 * @headroom: needed headroom
2147 * This function is identical to skb_cow except that we replace the
2148 * skb_cloned check by skb_header_cloned. It should be used when
2149 * you only need to push on some header and do not need to modify
2150 * the data.
2152 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2154 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2158 * skb_padto - pad an skbuff up to a minimal size
2159 * @skb: buffer to pad
2160 * @len: minimal length
2162 * Pads up a buffer to ensure the trailing bytes exist and are
2163 * blanked. If the buffer already contains sufficient data it
2164 * is untouched. Otherwise it is extended. Returns zero on
2165 * success. The skb is freed on error.
2168 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2170 unsigned int size = skb->len;
2171 if (likely(size >= len))
2172 return 0;
2173 return skb_pad(skb, len - size);
2176 static inline int skb_add_data(struct sk_buff *skb,
2177 char __user *from, int copy)
2179 const int off = skb->len;
2181 if (skb->ip_summed == CHECKSUM_NONE) {
2182 int err = 0;
2183 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2184 copy, 0, &err);
2185 if (!err) {
2186 skb->csum = csum_block_add(skb->csum, csum, off);
2187 return 0;
2189 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2190 return 0;
2192 __skb_trim(skb, off);
2193 return -EFAULT;
2196 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2197 const struct page *page, int off)
2199 if (i) {
2200 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2202 return page == skb_frag_page(frag) &&
2203 off == frag->page_offset + skb_frag_size(frag);
2205 return false;
2208 static inline int __skb_linearize(struct sk_buff *skb)
2210 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2214 * skb_linearize - convert paged skb to linear one
2215 * @skb: buffer to linarize
2217 * If there is no free memory -ENOMEM is returned, otherwise zero
2218 * is returned and the old skb data released.
2220 static inline int skb_linearize(struct sk_buff *skb)
2222 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2226 * skb_has_shared_frag - can any frag be overwritten
2227 * @skb: buffer to test
2229 * Return true if the skb has at least one frag that might be modified
2230 * by an external entity (as in vmsplice()/sendfile())
2232 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2234 return skb_is_nonlinear(skb) &&
2235 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2239 * skb_linearize_cow - make sure skb is linear and writable
2240 * @skb: buffer to process
2242 * If there is no free memory -ENOMEM is returned, otherwise zero
2243 * is returned and the old skb data released.
2245 static inline int skb_linearize_cow(struct sk_buff *skb)
2247 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2248 __skb_linearize(skb) : 0;
2252 * skb_postpull_rcsum - update checksum for received skb after pull
2253 * @skb: buffer to update
2254 * @start: start of data before pull
2255 * @len: length of data pulled
2257 * After doing a pull on a received packet, you need to call this to
2258 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2259 * CHECKSUM_NONE so that it can be recomputed from scratch.
2262 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2263 const void *start, unsigned int len)
2265 if (skb->ip_summed == CHECKSUM_COMPLETE)
2266 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2269 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2272 * pskb_trim_rcsum - trim received skb and update checksum
2273 * @skb: buffer to trim
2274 * @len: new length
2276 * This is exactly the same as pskb_trim except that it ensures the
2277 * checksum of received packets are still valid after the operation.
2280 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2282 if (likely(len >= skb->len))
2283 return 0;
2284 if (skb->ip_summed == CHECKSUM_COMPLETE)
2285 skb->ip_summed = CHECKSUM_NONE;
2286 return __pskb_trim(skb, len);
2289 #define skb_queue_walk(queue, skb) \
2290 for (skb = (queue)->next; \
2291 skb != (struct sk_buff *)(queue); \
2292 skb = skb->next)
2294 #define skb_queue_walk_safe(queue, skb, tmp) \
2295 for (skb = (queue)->next, tmp = skb->next; \
2296 skb != (struct sk_buff *)(queue); \
2297 skb = tmp, tmp = skb->next)
2299 #define skb_queue_walk_from(queue, skb) \
2300 for (; skb != (struct sk_buff *)(queue); \
2301 skb = skb->next)
2303 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2304 for (tmp = skb->next; \
2305 skb != (struct sk_buff *)(queue); \
2306 skb = tmp, tmp = skb->next)
2308 #define skb_queue_reverse_walk(queue, skb) \
2309 for (skb = (queue)->prev; \
2310 skb != (struct sk_buff *)(queue); \
2311 skb = skb->prev)
2313 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2314 for (skb = (queue)->prev, tmp = skb->prev; \
2315 skb != (struct sk_buff *)(queue); \
2316 skb = tmp, tmp = skb->prev)
2318 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2319 for (tmp = skb->prev; \
2320 skb != (struct sk_buff *)(queue); \
2321 skb = tmp, tmp = skb->prev)
2323 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2325 return skb_shinfo(skb)->frag_list != NULL;
2328 static inline void skb_frag_list_init(struct sk_buff *skb)
2330 skb_shinfo(skb)->frag_list = NULL;
2333 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2335 frag->next = skb_shinfo(skb)->frag_list;
2336 skb_shinfo(skb)->frag_list = frag;
2339 #define skb_walk_frags(skb, iter) \
2340 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2342 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2343 int *peeked, int *off, int *err);
2344 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2345 int noblock, int *err);
2346 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2347 struct poll_table_struct *wait);
2348 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2349 int offset, struct iovec *to,
2350 int size);
2351 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
2352 int hlen,
2353 struct iovec *iov);
2354 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2355 int offset,
2356 const struct iovec *from,
2357 int from_offset,
2358 int len);
2359 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2360 int offset,
2361 const struct iovec *to,
2362 int to_offset,
2363 int size);
2364 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2365 extern void skb_free_datagram_locked(struct sock *sk,
2366 struct sk_buff *skb);
2367 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
2368 unsigned int flags);
2369 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2370 int len, __wsum csum);
2371 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2372 void *to, int len);
2373 extern int skb_store_bits(struct sk_buff *skb, int offset,
2374 const void *from, int len);
2375 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
2376 int offset, u8 *to, int len,
2377 __wsum csum);
2378 extern int skb_splice_bits(struct sk_buff *skb,
2379 unsigned int offset,
2380 struct pipe_inode_info *pipe,
2381 unsigned int len,
2382 unsigned int flags);
2383 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2384 extern void skb_split(struct sk_buff *skb,
2385 struct sk_buff *skb1, const u32 len);
2386 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2387 int shiftlen);
2388 extern void skb_scrub_packet(struct sk_buff *skb);
2390 extern struct sk_buff *skb_segment(struct sk_buff *skb,
2391 netdev_features_t features);
2393 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2394 int len, void *buffer)
2396 int hlen = skb_headlen(skb);
2398 if (hlen - offset >= len)
2399 return skb->data + offset;
2401 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2402 return NULL;
2404 return buffer;
2407 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2408 void *to,
2409 const unsigned int len)
2411 memcpy(to, skb->data, len);
2414 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2415 const int offset, void *to,
2416 const unsigned int len)
2418 memcpy(to, skb->data + offset, len);
2421 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2422 const void *from,
2423 const unsigned int len)
2425 memcpy(skb->data, from, len);
2428 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2429 const int offset,
2430 const void *from,
2431 const unsigned int len)
2433 memcpy(skb->data + offset, from, len);
2436 extern void skb_init(void);
2438 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2440 return skb->tstamp;
2444 * skb_get_timestamp - get timestamp from a skb
2445 * @skb: skb to get stamp from
2446 * @stamp: pointer to struct timeval to store stamp in
2448 * Timestamps are stored in the skb as offsets to a base timestamp.
2449 * This function converts the offset back to a struct timeval and stores
2450 * it in stamp.
2452 static inline void skb_get_timestamp(const struct sk_buff *skb,
2453 struct timeval *stamp)
2455 *stamp = ktime_to_timeval(skb->tstamp);
2458 static inline void skb_get_timestampns(const struct sk_buff *skb,
2459 struct timespec *stamp)
2461 *stamp = ktime_to_timespec(skb->tstamp);
2464 static inline void __net_timestamp(struct sk_buff *skb)
2466 skb->tstamp = ktime_get_real();
2469 static inline ktime_t net_timedelta(ktime_t t)
2471 return ktime_sub(ktime_get_real(), t);
2474 static inline ktime_t net_invalid_timestamp(void)
2476 return ktime_set(0, 0);
2479 extern void skb_timestamping_init(void);
2481 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2483 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2484 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2486 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2488 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2492 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2494 return false;
2497 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2500 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2502 * PHY drivers may accept clones of transmitted packets for
2503 * timestamping via their phy_driver.txtstamp method. These drivers
2504 * must call this function to return the skb back to the stack, with
2505 * or without a timestamp.
2507 * @skb: clone of the the original outgoing packet
2508 * @hwtstamps: hardware time stamps, may be NULL if not available
2511 void skb_complete_tx_timestamp(struct sk_buff *skb,
2512 struct skb_shared_hwtstamps *hwtstamps);
2515 * skb_tstamp_tx - queue clone of skb with send time stamps
2516 * @orig_skb: the original outgoing packet
2517 * @hwtstamps: hardware time stamps, may be NULL if not available
2519 * If the skb has a socket associated, then this function clones the
2520 * skb (thus sharing the actual data and optional structures), stores
2521 * the optional hardware time stamping information (if non NULL) or
2522 * generates a software time stamp (otherwise), then queues the clone
2523 * to the error queue of the socket. Errors are silently ignored.
2525 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2526 struct skb_shared_hwtstamps *hwtstamps);
2528 static inline void sw_tx_timestamp(struct sk_buff *skb)
2530 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2531 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2532 skb_tstamp_tx(skb, NULL);
2536 * skb_tx_timestamp() - Driver hook for transmit timestamping
2538 * Ethernet MAC Drivers should call this function in their hard_xmit()
2539 * function immediately before giving the sk_buff to the MAC hardware.
2541 * @skb: A socket buffer.
2543 static inline void skb_tx_timestamp(struct sk_buff *skb)
2545 skb_clone_tx_timestamp(skb);
2546 sw_tx_timestamp(skb);
2550 * skb_complete_wifi_ack - deliver skb with wifi status
2552 * @skb: the original outgoing packet
2553 * @acked: ack status
2556 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2558 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2559 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2561 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2563 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2567 * skb_checksum_complete - Calculate checksum of an entire packet
2568 * @skb: packet to process
2570 * This function calculates the checksum over the entire packet plus
2571 * the value of skb->csum. The latter can be used to supply the
2572 * checksum of a pseudo header as used by TCP/UDP. It returns the
2573 * checksum.
2575 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2576 * this function can be used to verify that checksum on received
2577 * packets. In that case the function should return zero if the
2578 * checksum is correct. In particular, this function will return zero
2579 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2580 * hardware has already verified the correctness of the checksum.
2582 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2584 return skb_csum_unnecessary(skb) ?
2585 0 : __skb_checksum_complete(skb);
2588 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2589 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2590 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2592 if (nfct && atomic_dec_and_test(&nfct->use))
2593 nf_conntrack_destroy(nfct);
2595 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2597 if (nfct)
2598 atomic_inc(&nfct->use);
2600 #endif
2601 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2602 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2604 if (skb)
2605 atomic_inc(&skb->users);
2607 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2609 if (skb)
2610 kfree_skb(skb);
2612 #endif
2613 #ifdef CONFIG_BRIDGE_NETFILTER
2614 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2616 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2617 kfree(nf_bridge);
2619 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2621 if (nf_bridge)
2622 atomic_inc(&nf_bridge->use);
2624 #endif /* CONFIG_BRIDGE_NETFILTER */
2625 static inline void nf_reset(struct sk_buff *skb)
2627 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2628 nf_conntrack_put(skb->nfct);
2629 skb->nfct = NULL;
2630 #endif
2631 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2632 nf_conntrack_put_reasm(skb->nfct_reasm);
2633 skb->nfct_reasm = NULL;
2634 #endif
2635 #ifdef CONFIG_BRIDGE_NETFILTER
2636 nf_bridge_put(skb->nf_bridge);
2637 skb->nf_bridge = NULL;
2638 #endif
2641 static inline void nf_reset_trace(struct sk_buff *skb)
2643 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2644 skb->nf_trace = 0;
2645 #endif
2648 /* Note: This doesn't put any conntrack and bridge info in dst. */
2649 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2651 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2652 dst->nfct = src->nfct;
2653 nf_conntrack_get(src->nfct);
2654 dst->nfctinfo = src->nfctinfo;
2655 #endif
2656 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2657 dst->nfct_reasm = src->nfct_reasm;
2658 nf_conntrack_get_reasm(src->nfct_reasm);
2659 #endif
2660 #ifdef CONFIG_BRIDGE_NETFILTER
2661 dst->nf_bridge = src->nf_bridge;
2662 nf_bridge_get(src->nf_bridge);
2663 #endif
2666 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2668 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2669 nf_conntrack_put(dst->nfct);
2670 #endif
2671 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2672 nf_conntrack_put_reasm(dst->nfct_reasm);
2673 #endif
2674 #ifdef CONFIG_BRIDGE_NETFILTER
2675 nf_bridge_put(dst->nf_bridge);
2676 #endif
2677 __nf_copy(dst, src);
2680 #ifdef CONFIG_NETWORK_SECMARK
2681 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2683 to->secmark = from->secmark;
2686 static inline void skb_init_secmark(struct sk_buff *skb)
2688 skb->secmark = 0;
2690 #else
2691 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2694 static inline void skb_init_secmark(struct sk_buff *skb)
2696 #endif
2698 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2700 skb->queue_mapping = queue_mapping;
2703 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2705 return skb->queue_mapping;
2708 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2710 to->queue_mapping = from->queue_mapping;
2713 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2715 skb->queue_mapping = rx_queue + 1;
2718 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2720 return skb->queue_mapping - 1;
2723 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2725 return skb->queue_mapping != 0;
2728 extern u16 __skb_tx_hash(const struct net_device *dev,
2729 const struct sk_buff *skb,
2730 unsigned int num_tx_queues);
2732 #ifdef CONFIG_XFRM
2733 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2735 return skb->sp;
2737 #else
2738 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2740 return NULL;
2742 #endif
2744 /* Keeps track of mac header offset relative to skb->head.
2745 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2746 * For non-tunnel skb it points to skb_mac_header() and for
2747 * tunnel skb it points to outer mac header. */
2748 struct skb_gso_cb {
2749 int mac_offset;
2751 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2753 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2755 return (skb_mac_header(inner_skb) - inner_skb->head) -
2756 SKB_GSO_CB(inner_skb)->mac_offset;
2759 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2761 int new_headroom, headroom;
2762 int ret;
2764 headroom = skb_headroom(skb);
2765 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2766 if (ret)
2767 return ret;
2769 new_headroom = skb_headroom(skb);
2770 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2771 return 0;
2774 static inline bool skb_is_gso(const struct sk_buff *skb)
2776 return skb_shinfo(skb)->gso_size;
2779 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2781 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2784 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2786 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2788 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2789 * wanted then gso_type will be set. */
2790 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2792 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2793 unlikely(shinfo->gso_type == 0)) {
2794 __skb_warn_lro_forwarding(skb);
2795 return true;
2797 return false;
2800 static inline void skb_forward_csum(struct sk_buff *skb)
2802 /* Unfortunately we don't support this one. Any brave souls? */
2803 if (skb->ip_summed == CHECKSUM_COMPLETE)
2804 skb->ip_summed = CHECKSUM_NONE;
2808 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2809 * @skb: skb to check
2811 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2812 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2813 * use this helper, to document places where we make this assertion.
2815 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2817 #ifdef DEBUG
2818 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2819 #endif
2822 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2824 u32 __skb_get_poff(const struct sk_buff *skb);
2827 * skb_head_is_locked - Determine if the skb->head is locked down
2828 * @skb: skb to check
2830 * The head on skbs build around a head frag can be removed if they are
2831 * not cloned. This function returns true if the skb head is locked down
2832 * due to either being allocated via kmalloc, or by being a clone with
2833 * multiple references to the head.
2835 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2837 return !skb->head_frag || skb_cloned(skb);
2839 #endif /* __KERNEL__ */
2840 #endif /* _LINUX_SKBUFF_H */