2 * Definitions for the 'struct sk_buff' memory handlers.
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 <linux/sched.h>
36 #include <net/flow_keys.h>
38 /* A. Checksumming of received packets by device.
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 * CHECKSUM_UNNECESSARY:
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
73 * B. Checksumming on output.
77 * The skb was already checksummed by the protocol, or a checksum is not
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
96 * CHECKSUM_UNNECESSARY:
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
102 * NETIF_F_FCOE_CRC - This indicates that 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 /* Don't change this without changing skb_csum_unnecessary! */
110 #define CHECKSUM_NONE 0
111 #define CHECKSUM_UNNECESSARY 1
112 #define CHECKSUM_COMPLETE 2
113 #define CHECKSUM_PARTIAL 3
115 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
116 ~(SMP_CACHE_BYTES - 1))
117 #define SKB_WITH_OVERHEAD(X) \
118 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
119 #define SKB_MAX_ORDER(X, ORDER) \
120 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
121 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
122 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
124 /* return minimum truesize of one skb containing X bytes of data */
125 #define SKB_TRUESIZE(X) ((X) + \
126 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
127 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
131 struct pipe_inode_info;
133 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
134 struct nf_conntrack {
139 #ifdef CONFIG_BRIDGE_NETFILTER
140 struct nf_bridge_info {
143 struct net_device *physindev;
144 struct net_device *physoutdev;
145 unsigned long data[32 / sizeof(unsigned long)];
149 struct sk_buff_head {
150 /* These two members must be first. */
151 struct sk_buff *next;
152 struct sk_buff *prev;
160 /* To allow 64K frame to be packed as single skb without frag_list we
161 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
162 * buffers which do not start on a page boundary.
164 * Since GRO uses frags we allocate at least 16 regardless of page
167 #if (65536/PAGE_SIZE + 1) < 16
168 #define MAX_SKB_FRAGS 16UL
170 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
173 typedef struct skb_frag_struct skb_frag_t;
175 struct skb_frag_struct {
179 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
188 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
193 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
198 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
203 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
208 #define HAVE_HW_TIME_STAMP
211 * struct skb_shared_hwtstamps - hardware time stamps
212 * @hwtstamp: hardware time stamp transformed into duration
213 * since arbitrary point in time
214 * @syststamp: hwtstamp transformed to system time base
216 * Software time stamps generated by ktime_get_real() are stored in
217 * skb->tstamp. The relation between the different kinds of time
218 * stamps is as follows:
220 * syststamp and tstamp can be compared against each other in
221 * arbitrary combinations. The accuracy of a
222 * syststamp/tstamp/"syststamp from other device" comparison is
223 * limited by the accuracy of the transformation into system time
224 * base. This depends on the device driver and its underlying
227 * hwtstamps can only be compared against other hwtstamps from
230 * This structure is attached to packets as part of the
231 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
233 struct skb_shared_hwtstamps {
238 /* Definitions for tx_flags in struct skb_shared_info */
240 /* generate hardware time stamp */
241 SKBTX_HW_TSTAMP = 1 << 0,
243 /* generate software time stamp */
244 SKBTX_SW_TSTAMP = 1 << 1,
246 /* device driver is going to provide hardware time stamp */
247 SKBTX_IN_PROGRESS = 1 << 2,
249 /* device driver supports TX zero-copy buffers */
250 SKBTX_DEV_ZEROCOPY = 1 << 3,
252 /* generate wifi status information (where possible) */
253 SKBTX_WIFI_STATUS = 1 << 4,
255 /* This indicates at least one fragment might be overwritten
256 * (as in vmsplice(), sendfile() ...)
257 * If we need to compute a TX checksum, we'll need to copy
258 * all frags to avoid possible bad checksum
260 SKBTX_SHARED_FRAG = 1 << 5,
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
277 /* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
280 struct skb_shared_info {
281 unsigned char nr_frags;
283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
287 struct sk_buff *frag_list;
288 struct skb_shared_hwtstamps hwtstamps;
292 * Warning : all fields before dataref are cleared in __alloc_skb()
296 /* Intermediate layers must ensure that destructor_arg
297 * remains valid until skb destructor */
298 void * destructor_arg;
300 /* must be last field, see pskb_expand_head() */
301 skb_frag_t frags[MAX_SKB_FRAGS];
304 /* We divide dataref into two halves. The higher 16 bits hold references
305 * to the payload part of skb->data. The lower 16 bits hold references to
306 * the entire skb->data. A clone of a headerless skb holds the length of
307 * the header in skb->hdr_len.
309 * All users must obey the rule that the skb->data reference count must be
310 * greater than or equal to the payload reference count.
312 * Holding a reference to the payload part means that the user does not
313 * care about modifications to the header part of skb->data.
315 #define SKB_DATAREF_SHIFT 16
316 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
320 SKB_FCLONE_UNAVAILABLE,
326 SKB_GSO_TCPV4 = 1 << 0,
327 SKB_GSO_UDP = 1 << 1,
329 /* This indicates the skb is from an untrusted source. */
330 SKB_GSO_DODGY = 1 << 2,
332 /* This indicates the tcp segment has CWR set. */
333 SKB_GSO_TCP_ECN = 1 << 3,
335 SKB_GSO_TCPV6 = 1 << 4,
337 SKB_GSO_FCOE = 1 << 5,
339 SKB_GSO_GRE = 1 << 6,
341 SKB_GSO_IPIP = 1 << 7,
343 SKB_GSO_SIT = 1 << 8,
345 SKB_GSO_UDP_TUNNEL = 1 << 9,
347 SKB_GSO_MPLS = 1 << 10,
350 #if BITS_PER_LONG > 32
351 #define NET_SKBUFF_DATA_USES_OFFSET 1
354 #ifdef NET_SKBUFF_DATA_USES_OFFSET
355 typedef unsigned int sk_buff_data_t;
357 typedef unsigned char *sk_buff_data_t;
361 * struct skb_mstamp - multi resolution time stamps
362 * @stamp_us: timestamp in us resolution
363 * @stamp_jiffies: timestamp in jiffies
376 * skb_mstamp_get - get current timestamp
377 * @cl: place to store timestamps
379 static inline void skb_mstamp_get(struct skb_mstamp *cl)
381 u64 val = local_clock();
383 do_div(val, NSEC_PER_USEC);
384 cl->stamp_us = (u32)val;
385 cl->stamp_jiffies = (u32)jiffies;
389 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
390 * @t1: pointer to newest sample
391 * @t0: pointer to oldest sample
393 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
394 const struct skb_mstamp *t0)
396 s32 delta_us = t1->stamp_us - t0->stamp_us;
397 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
399 /* If delta_us is negative, this might be because interval is too big,
400 * or local_clock() drift is too big : fallback using jiffies.
403 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
405 delta_us = jiffies_to_usecs(delta_jiffies);
412 * struct sk_buff - socket buffer
413 * @next: Next buffer in list
414 * @prev: Previous buffer in list
415 * @tstamp: Time we arrived/left
416 * @sk: Socket we are owned by
417 * @dev: Device we arrived on/are leaving by
418 * @cb: Control buffer. Free for use by every layer. Put private vars here
419 * @_skb_refdst: destination entry (with norefcount bit)
420 * @sp: the security path, used for xfrm
421 * @len: Length of actual data
422 * @data_len: Data length
423 * @mac_len: Length of link layer header
424 * @hdr_len: writable header length of cloned skb
425 * @csum: Checksum (must include start/offset pair)
426 * @csum_start: Offset from skb->head where checksumming should start
427 * @csum_offset: Offset from csum_start where checksum should be stored
428 * @priority: Packet queueing priority
429 * @local_df: allow local fragmentation
430 * @cloned: Head may be cloned (check refcnt to be sure)
431 * @ip_summed: Driver fed us an IP checksum
432 * @nohdr: Payload reference only, must not modify header
433 * @nfctinfo: Relationship of this skb to the connection
434 * @pkt_type: Packet class
435 * @fclone: skbuff clone status
436 * @ipvs_property: skbuff is owned by ipvs
437 * @peeked: this packet has been seen already, so stats have been
438 * done for it, don't do them again
439 * @nf_trace: netfilter packet trace flag
440 * @protocol: Packet protocol from driver
441 * @destructor: Destruct function
442 * @nfct: Associated connection, if any
443 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
444 * @skb_iif: ifindex of device we arrived on
445 * @tc_index: Traffic control index
446 * @tc_verd: traffic control verdict
447 * @hash: the packet hash
448 * @queue_mapping: Queue mapping for multiqueue devices
449 * @ndisc_nodetype: router type (from link layer)
450 * @ooo_okay: allow the mapping of a socket to a queue to be changed
451 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
453 * @wifi_acked_valid: wifi_acked was set
454 * @wifi_acked: whether frame was acked on wifi or not
455 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
456 * @dma_cookie: a cookie to one of several possible DMA operations
457 * done by skb DMA functions
458 * @napi_id: id of the NAPI struct this skb came from
459 * @secmark: security marking
460 * @mark: Generic packet mark
461 * @dropcount: total number of sk_receive_queue overflows
462 * @vlan_proto: vlan encapsulation protocol
463 * @vlan_tci: vlan tag control information
464 * @inner_protocol: Protocol (encapsulation)
465 * @inner_transport_header: Inner transport layer header (encapsulation)
466 * @inner_network_header: Network layer header (encapsulation)
467 * @inner_mac_header: Link layer header (encapsulation)
468 * @transport_header: Transport layer header
469 * @network_header: Network layer header
470 * @mac_header: Link layer header
471 * @tail: Tail pointer
473 * @head: Head of buffer
474 * @data: Data head pointer
475 * @truesize: Buffer size
476 * @users: User count - see {datagram,tcp}.c
480 /* These two members must be first. */
481 struct sk_buff *next;
482 struct sk_buff *prev;
486 struct skb_mstamp skb_mstamp;
490 struct net_device *dev;
493 * This is the control buffer. It is free to use for every
494 * layer. Please put your private variables there. If you
495 * want to keep them across layers you have to do a skb_clone()
496 * first. This is owned by whoever has the skb queued ATM.
498 char cb[48] __aligned(8);
500 unsigned long _skb_refdst;
516 kmemcheck_bitfield_begin(flags1);
527 kmemcheck_bitfield_end(flags1);
530 void (*destructor)(struct sk_buff *skb);
531 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
532 struct nf_conntrack *nfct;
534 #ifdef CONFIG_BRIDGE_NETFILTER
535 struct nf_bridge_info *nf_bridge;
545 #ifdef CONFIG_NET_SCHED
546 __u16 tc_index; /* traffic control index */
547 #ifdef CONFIG_NET_CLS_ACT
548 __u16 tc_verd; /* traffic control verdict */
553 kmemcheck_bitfield_begin(flags2);
554 #ifdef CONFIG_IPV6_NDISC_NODETYPE
555 __u8 ndisc_nodetype:2;
560 __u8 wifi_acked_valid:1;
564 /* Encapsulation protocol and NIC drivers should use
565 * this flag to indicate to each other if the skb contains
566 * encapsulated packet or not and maybe use the inner packet
569 __u8 encapsulation:1;
570 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
571 kmemcheck_bitfield_end(flags2);
573 #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
575 unsigned int napi_id;
576 dma_cookie_t dma_cookie;
579 #ifdef CONFIG_NETWORK_SECMARK
585 __u32 reserved_tailroom;
588 __be16 inner_protocol;
589 __u16 inner_transport_header;
590 __u16 inner_network_header;
591 __u16 inner_mac_header;
592 __u16 transport_header;
593 __u16 network_header;
595 /* These elements must be at the end, see alloc_skb() for details. */
600 unsigned int truesize;
606 * Handling routines are only of interest to the kernel
608 #include <linux/slab.h>
611 #define SKB_ALLOC_FCLONE 0x01
612 #define SKB_ALLOC_RX 0x02
614 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
615 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
617 return unlikely(skb->pfmemalloc);
621 * skb might have a dst pointer attached, refcounted or not.
622 * _skb_refdst low order bit is set if refcount was _not_ taken
624 #define SKB_DST_NOREF 1UL
625 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
628 * skb_dst - returns skb dst_entry
631 * Returns skb dst_entry, regardless of reference taken or not.
633 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
635 /* If refdst was not refcounted, check we still are in a
636 * rcu_read_lock section
638 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
639 !rcu_read_lock_held() &&
640 !rcu_read_lock_bh_held());
641 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
645 * skb_dst_set - sets skb dst
649 * Sets skb dst, assuming a reference was taken on dst and should
650 * be released by skb_dst_drop()
652 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
654 skb->_skb_refdst = (unsigned long)dst;
657 void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
661 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
665 * Sets skb dst, assuming a reference was not taken on dst.
666 * If dst entry is cached, we do not take reference and dst_release
667 * will be avoided by refdst_drop. If dst entry is not cached, we take
668 * reference, so that last dst_release can destroy the dst immediately.
670 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
672 __skb_dst_set_noref(skb, dst, false);
676 * skb_dst_set_noref_force - sets skb dst, without taking reference
680 * Sets skb dst, assuming a reference was not taken on dst.
681 * No reference is taken and no dst_release will be called. While for
682 * cached dsts deferred reclaim is a basic feature, for entries that are
683 * not cached it is caller's job to guarantee that last dst_release for
684 * provided dst happens when nobody uses it, eg. after a RCU grace period.
686 static inline void skb_dst_set_noref_force(struct sk_buff *skb,
687 struct dst_entry *dst)
689 __skb_dst_set_noref(skb, dst, true);
693 * skb_dst_is_noref - Test if skb dst isn't refcounted
696 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
698 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
701 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
703 return (struct rtable *)skb_dst(skb);
706 void kfree_skb(struct sk_buff *skb);
707 void kfree_skb_list(struct sk_buff *segs);
708 void skb_tx_error(struct sk_buff *skb);
709 void consume_skb(struct sk_buff *skb);
710 void __kfree_skb(struct sk_buff *skb);
711 extern struct kmem_cache *skbuff_head_cache;
713 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
714 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
715 bool *fragstolen, int *delta_truesize);
717 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
719 struct sk_buff *build_skb(void *data, unsigned int frag_size);
720 static inline struct sk_buff *alloc_skb(unsigned int size,
723 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
726 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
729 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
732 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
733 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
735 return __alloc_skb_head(priority, -1);
738 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
739 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
740 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
741 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
742 struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
744 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
745 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
746 unsigned int headroom);
747 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
748 int newtailroom, gfp_t priority);
749 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
750 int offset, int len);
751 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
753 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
754 int skb_pad(struct sk_buff *skb, int pad);
755 #define dev_kfree_skb(a) consume_skb(a)
757 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
758 int getfrag(void *from, char *to, int offset,
759 int len, int odd, struct sk_buff *skb),
760 void *from, int length);
762 struct skb_seq_state {
766 __u32 stepped_offset;
767 struct sk_buff *root_skb;
768 struct sk_buff *cur_skb;
772 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
773 unsigned int to, struct skb_seq_state *st);
774 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
775 struct skb_seq_state *st);
776 void skb_abort_seq_read(struct skb_seq_state *st);
778 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
779 unsigned int to, struct ts_config *config,
780 struct ts_state *state);
783 * Packet hash types specify the type of hash in skb_set_hash.
785 * Hash types refer to the protocol layer addresses which are used to
786 * construct a packet's hash. The hashes are used to differentiate or identify
787 * flows of the protocol layer for the hash type. Hash types are either
788 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
790 * Properties of hashes:
792 * 1) Two packets in different flows have different hash values
793 * 2) Two packets in the same flow should have the same hash value
795 * A hash at a higher layer is considered to be more specific. A driver should
796 * set the most specific hash possible.
798 * A driver cannot indicate a more specific hash than the layer at which a hash
799 * was computed. For instance an L3 hash cannot be set as an L4 hash.
801 * A driver may indicate a hash level which is less specific than the
802 * actual layer the hash was computed on. For instance, a hash computed
803 * at L4 may be considered an L3 hash. This should only be done if the
804 * driver can't unambiguously determine that the HW computed the hash at
805 * the higher layer. Note that the "should" in the second property above
808 enum pkt_hash_types {
809 PKT_HASH_TYPE_NONE, /* Undefined type */
810 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
811 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
812 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
816 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
818 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
822 void __skb_get_hash(struct sk_buff *skb);
823 static inline __u32 skb_get_hash(struct sk_buff *skb)
831 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
836 static inline void skb_clear_hash(struct sk_buff *skb)
842 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
848 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
850 to->hash = from->hash;
851 to->l4_hash = from->l4_hash;
854 #ifdef NET_SKBUFF_DATA_USES_OFFSET
855 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
857 return skb->head + skb->end;
860 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
865 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
870 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
872 return skb->end - skb->head;
877 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
879 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
881 return &skb_shinfo(skb)->hwtstamps;
885 * skb_queue_empty - check if a queue is empty
888 * Returns true if the queue is empty, false otherwise.
890 static inline int skb_queue_empty(const struct sk_buff_head *list)
892 return list->next == (const struct sk_buff *) list;
896 * skb_queue_is_last - check if skb is the last entry in the queue
900 * Returns true if @skb is the last buffer on the list.
902 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
903 const struct sk_buff *skb)
905 return skb->next == (const struct sk_buff *) list;
909 * skb_queue_is_first - check if skb is the first entry in the queue
913 * Returns true if @skb is the first buffer on the list.
915 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
916 const struct sk_buff *skb)
918 return skb->prev == (const struct sk_buff *) list;
922 * skb_queue_next - return the next packet in the queue
924 * @skb: current buffer
926 * Return the next packet in @list after @skb. It is only valid to
927 * call this if skb_queue_is_last() evaluates to false.
929 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
930 const struct sk_buff *skb)
932 /* This BUG_ON may seem severe, but if we just return then we
933 * are going to dereference garbage.
935 BUG_ON(skb_queue_is_last(list, skb));
940 * skb_queue_prev - return the prev packet in the queue
942 * @skb: current buffer
944 * Return the prev packet in @list before @skb. It is only valid to
945 * call this if skb_queue_is_first() evaluates to false.
947 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
948 const struct sk_buff *skb)
950 /* This BUG_ON may seem severe, but if we just return then we
951 * are going to dereference garbage.
953 BUG_ON(skb_queue_is_first(list, skb));
958 * skb_get - reference buffer
959 * @skb: buffer to reference
961 * Makes another reference to a socket buffer and returns a pointer
964 static inline struct sk_buff *skb_get(struct sk_buff *skb)
966 atomic_inc(&skb->users);
971 * If users == 1, we are the only owner and are can avoid redundant
976 * skb_cloned - is the buffer a clone
977 * @skb: buffer to check
979 * Returns true if the buffer was generated with skb_clone() and is
980 * one of multiple shared copies of the buffer. Cloned buffers are
981 * shared data so must not be written to under normal circumstances.
983 static inline int skb_cloned(const struct sk_buff *skb)
985 return skb->cloned &&
986 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
989 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
991 might_sleep_if(pri & __GFP_WAIT);
994 return pskb_expand_head(skb, 0, 0, pri);
1000 * skb_header_cloned - is the header a clone
1001 * @skb: buffer to check
1003 * Returns true if modifying the header part of the buffer requires
1004 * the data to be copied.
1006 static inline int skb_header_cloned(const struct sk_buff *skb)
1013 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1014 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1015 return dataref != 1;
1019 * skb_header_release - release reference to header
1020 * @skb: buffer to operate on
1022 * Drop a reference to the header part of the buffer. This is done
1023 * by acquiring a payload reference. You must not read from the header
1024 * part of skb->data after this.
1026 static inline void skb_header_release(struct sk_buff *skb)
1030 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1034 * skb_shared - is the buffer shared
1035 * @skb: buffer to check
1037 * Returns true if more than one person has a reference to this
1040 static inline int skb_shared(const struct sk_buff *skb)
1042 return atomic_read(&skb->users) != 1;
1046 * skb_share_check - check if buffer is shared and if so clone it
1047 * @skb: buffer to check
1048 * @pri: priority for memory allocation
1050 * If the buffer is shared the buffer is cloned and the old copy
1051 * drops a reference. A new clone with a single reference is returned.
1052 * If the buffer is not shared the original buffer is returned. When
1053 * being called from interrupt status or with spinlocks held pri must
1056 * NULL is returned on a memory allocation failure.
1058 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1060 might_sleep_if(pri & __GFP_WAIT);
1061 if (skb_shared(skb)) {
1062 struct sk_buff *nskb = skb_clone(skb, pri);
1074 * Copy shared buffers into a new sk_buff. We effectively do COW on
1075 * packets to handle cases where we have a local reader and forward
1076 * and a couple of other messy ones. The normal one is tcpdumping
1077 * a packet thats being forwarded.
1081 * skb_unshare - make a copy of a shared buffer
1082 * @skb: buffer to check
1083 * @pri: priority for memory allocation
1085 * If the socket buffer is a clone then this function creates a new
1086 * copy of the data, drops a reference count on the old copy and returns
1087 * the new copy with the reference count at 1. If the buffer is not a clone
1088 * the original buffer is returned. When called with a spinlock held or
1089 * from interrupt state @pri must be %GFP_ATOMIC
1091 * %NULL is returned on a memory allocation failure.
1093 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1096 might_sleep_if(pri & __GFP_WAIT);
1097 if (skb_cloned(skb)) {
1098 struct sk_buff *nskb = skb_copy(skb, pri);
1099 kfree_skb(skb); /* Free our shared copy */
1106 * skb_peek - peek at the head of an &sk_buff_head
1107 * @list_: list to peek at
1109 * Peek an &sk_buff. Unlike most other operations you _MUST_
1110 * be careful with this one. A peek leaves the buffer on the
1111 * list and someone else may run off with it. You must hold
1112 * the appropriate locks or have a private queue to do this.
1114 * Returns %NULL for an empty list or a pointer to the head element.
1115 * The reference count is not incremented and the reference is therefore
1116 * volatile. Use with caution.
1118 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1120 struct sk_buff *skb = list_->next;
1122 if (skb == (struct sk_buff *)list_)
1128 * skb_peek_next - peek skb following the given one from a queue
1129 * @skb: skb to start from
1130 * @list_: list to peek at
1132 * Returns %NULL when the end of the list is met or a pointer to the
1133 * next element. The reference count is not incremented and the
1134 * reference is therefore volatile. Use with caution.
1136 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1137 const struct sk_buff_head *list_)
1139 struct sk_buff *next = skb->next;
1141 if (next == (struct sk_buff *)list_)
1147 * skb_peek_tail - peek at the tail of an &sk_buff_head
1148 * @list_: list to peek at
1150 * Peek an &sk_buff. Unlike most other operations you _MUST_
1151 * be careful with this one. A peek leaves the buffer on the
1152 * list and someone else may run off with it. You must hold
1153 * the appropriate locks or have a private queue to do this.
1155 * Returns %NULL for an empty list or a pointer to the tail element.
1156 * The reference count is not incremented and the reference is therefore
1157 * volatile. Use with caution.
1159 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1161 struct sk_buff *skb = list_->prev;
1163 if (skb == (struct sk_buff *)list_)
1170 * skb_queue_len - get queue length
1171 * @list_: list to measure
1173 * Return the length of an &sk_buff queue.
1175 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1181 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1182 * @list: queue to initialize
1184 * This initializes only the list and queue length aspects of
1185 * an sk_buff_head object. This allows to initialize the list
1186 * aspects of an sk_buff_head without reinitializing things like
1187 * the spinlock. It can also be used for on-stack sk_buff_head
1188 * objects where the spinlock is known to not be used.
1190 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1192 list->prev = list->next = (struct sk_buff *)list;
1197 * This function creates a split out lock class for each invocation;
1198 * this is needed for now since a whole lot of users of the skb-queue
1199 * infrastructure in drivers have different locking usage (in hardirq)
1200 * than the networking core (in softirq only). In the long run either the
1201 * network layer or drivers should need annotation to consolidate the
1202 * main types of usage into 3 classes.
1204 static inline void skb_queue_head_init(struct sk_buff_head *list)
1206 spin_lock_init(&list->lock);
1207 __skb_queue_head_init(list);
1210 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1211 struct lock_class_key *class)
1213 skb_queue_head_init(list);
1214 lockdep_set_class(&list->lock, class);
1218 * Insert an sk_buff on a list.
1220 * The "__skb_xxxx()" functions are the non-atomic ones that
1221 * can only be called with interrupts disabled.
1223 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1224 struct sk_buff_head *list);
1225 static inline void __skb_insert(struct sk_buff *newsk,
1226 struct sk_buff *prev, struct sk_buff *next,
1227 struct sk_buff_head *list)
1231 next->prev = prev->next = newsk;
1235 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1236 struct sk_buff *prev,
1237 struct sk_buff *next)
1239 struct sk_buff *first = list->next;
1240 struct sk_buff *last = list->prev;
1250 * skb_queue_splice - join two skb lists, this is designed for stacks
1251 * @list: the new list to add
1252 * @head: the place to add it in the first list
1254 static inline void skb_queue_splice(const struct sk_buff_head *list,
1255 struct sk_buff_head *head)
1257 if (!skb_queue_empty(list)) {
1258 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1259 head->qlen += list->qlen;
1264 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1265 * @list: the new list to add
1266 * @head: the place to add it in the first list
1268 * The list at @list is reinitialised
1270 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1271 struct sk_buff_head *head)
1273 if (!skb_queue_empty(list)) {
1274 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1275 head->qlen += list->qlen;
1276 __skb_queue_head_init(list);
1281 * skb_queue_splice_tail - join two skb lists, each list being a queue
1282 * @list: the new list to add
1283 * @head: the place to add it in the first list
1285 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1286 struct sk_buff_head *head)
1288 if (!skb_queue_empty(list)) {
1289 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1290 head->qlen += list->qlen;
1295 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1296 * @list: the new list to add
1297 * @head: the place to add it in the first list
1299 * Each of the lists is a queue.
1300 * The list at @list is reinitialised
1302 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1303 struct sk_buff_head *head)
1305 if (!skb_queue_empty(list)) {
1306 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1307 head->qlen += list->qlen;
1308 __skb_queue_head_init(list);
1313 * __skb_queue_after - queue a buffer at the list head
1314 * @list: list to use
1315 * @prev: place after this buffer
1316 * @newsk: buffer to queue
1318 * Queue a buffer int the middle of a list. This function takes no locks
1319 * and you must therefore hold required locks before calling it.
1321 * A buffer cannot be placed on two lists at the same time.
1323 static inline void __skb_queue_after(struct sk_buff_head *list,
1324 struct sk_buff *prev,
1325 struct sk_buff *newsk)
1327 __skb_insert(newsk, prev, prev->next, list);
1330 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1331 struct sk_buff_head *list);
1333 static inline void __skb_queue_before(struct sk_buff_head *list,
1334 struct sk_buff *next,
1335 struct sk_buff *newsk)
1337 __skb_insert(newsk, next->prev, next, list);
1341 * __skb_queue_head - queue a buffer at the list head
1342 * @list: list to use
1343 * @newsk: buffer to queue
1345 * Queue a buffer at the start of a list. This function takes no locks
1346 * and you must therefore hold required locks before calling it.
1348 * A buffer cannot be placed on two lists at the same time.
1350 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1351 static inline void __skb_queue_head(struct sk_buff_head *list,
1352 struct sk_buff *newsk)
1354 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1358 * __skb_queue_tail - queue a buffer at the list tail
1359 * @list: list to use
1360 * @newsk: buffer to queue
1362 * Queue a buffer at the end of a list. This function takes no locks
1363 * and you must therefore hold required locks before calling it.
1365 * A buffer cannot be placed on two lists at the same time.
1367 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1368 static inline void __skb_queue_tail(struct sk_buff_head *list,
1369 struct sk_buff *newsk)
1371 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1375 * remove sk_buff from list. _Must_ be called atomically, and with
1378 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1379 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1381 struct sk_buff *next, *prev;
1386 skb->next = skb->prev = NULL;
1392 * __skb_dequeue - remove from the head of the queue
1393 * @list: list to dequeue from
1395 * Remove the head of the list. This function does not take any locks
1396 * so must be used with appropriate locks held only. The head item is
1397 * returned or %NULL if the list is empty.
1399 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1400 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1402 struct sk_buff *skb = skb_peek(list);
1404 __skb_unlink(skb, list);
1409 * __skb_dequeue_tail - remove from the tail of the queue
1410 * @list: list to dequeue from
1412 * Remove the tail of the list. This function does not take any locks
1413 * so must be used with appropriate locks held only. The tail item is
1414 * returned or %NULL if the list is empty.
1416 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1417 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1419 struct sk_buff *skb = skb_peek_tail(list);
1421 __skb_unlink(skb, list);
1426 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1428 return skb->data_len;
1431 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1433 return skb->len - skb->data_len;
1436 static inline int skb_pagelen(const struct sk_buff *skb)
1440 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1441 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1442 return len + skb_headlen(skb);
1446 * __skb_fill_page_desc - initialise a paged fragment in an skb
1447 * @skb: buffer containing fragment to be initialised
1448 * @i: paged fragment index to initialise
1449 * @page: the page to use for this fragment
1450 * @off: the offset to the data with @page
1451 * @size: the length of the data
1453 * Initialises the @i'th fragment of @skb to point to &size bytes at
1454 * offset @off within @page.
1456 * Does not take any additional reference on the fragment.
1458 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1459 struct page *page, int off, int size)
1461 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1464 * Propagate page->pfmemalloc to the skb if we can. The problem is
1465 * that not all callers have unique ownership of the page. If
1466 * pfmemalloc is set, we check the mapping as a mapping implies
1467 * page->index is set (index and pfmemalloc share space).
1468 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1469 * do not lose pfmemalloc information as the pages would not be
1470 * allocated using __GFP_MEMALLOC.
1472 frag->page.p = page;
1473 frag->page_offset = off;
1474 skb_frag_size_set(frag, size);
1476 page = compound_head(page);
1477 if (page->pfmemalloc && !page->mapping)
1478 skb->pfmemalloc = true;
1482 * skb_fill_page_desc - initialise a paged fragment in an skb
1483 * @skb: buffer containing fragment to be initialised
1484 * @i: paged fragment index to initialise
1485 * @page: the page to use for this fragment
1486 * @off: the offset to the data with @page
1487 * @size: the length of the data
1489 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1490 * @skb to point to @size bytes at offset @off within @page. In
1491 * addition updates @skb such that @i is the last fragment.
1493 * Does not take any additional reference on the fragment.
1495 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1496 struct page *page, int off, int size)
1498 __skb_fill_page_desc(skb, i, page, off, size);
1499 skb_shinfo(skb)->nr_frags = i + 1;
1502 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1503 int size, unsigned int truesize);
1505 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1506 unsigned int truesize);
1508 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1509 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1510 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1512 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1513 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1515 return skb->head + skb->tail;
1518 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1520 skb->tail = skb->data - skb->head;
1523 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1525 skb_reset_tail_pointer(skb);
1526 skb->tail += offset;
1529 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1530 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1535 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1537 skb->tail = skb->data;
1540 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1542 skb->tail = skb->data + offset;
1545 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1548 * Add data to an sk_buff
1550 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1551 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1552 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1554 unsigned char *tmp = skb_tail_pointer(skb);
1555 SKB_LINEAR_ASSERT(skb);
1561 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1562 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1569 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1570 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1573 BUG_ON(skb->len < skb->data_len);
1574 return skb->data += len;
1577 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1579 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1582 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1584 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1586 if (len > skb_headlen(skb) &&
1587 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1590 return skb->data += len;
1593 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1595 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1598 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1600 if (likely(len <= skb_headlen(skb)))
1602 if (unlikely(len > skb->len))
1604 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1608 * skb_headroom - bytes at buffer head
1609 * @skb: buffer to check
1611 * Return the number of bytes of free space at the head of an &sk_buff.
1613 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1615 return skb->data - skb->head;
1619 * skb_tailroom - bytes at buffer end
1620 * @skb: buffer to check
1622 * Return the number of bytes of free space at the tail of an sk_buff
1624 static inline int skb_tailroom(const struct sk_buff *skb)
1626 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1630 * skb_availroom - bytes at buffer end
1631 * @skb: buffer to check
1633 * Return the number of bytes of free space at the tail of an sk_buff
1634 * allocated by sk_stream_alloc()
1636 static inline int skb_availroom(const struct sk_buff *skb)
1638 if (skb_is_nonlinear(skb))
1641 return skb->end - skb->tail - skb->reserved_tailroom;
1645 * skb_reserve - adjust headroom
1646 * @skb: buffer to alter
1647 * @len: bytes to move
1649 * Increase the headroom of an empty &sk_buff by reducing the tail
1650 * room. This is only allowed for an empty buffer.
1652 static inline void skb_reserve(struct sk_buff *skb, int len)
1658 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1660 skb->inner_mac_header = skb->mac_header;
1661 skb->inner_network_header = skb->network_header;
1662 skb->inner_transport_header = skb->transport_header;
1665 static inline void skb_reset_mac_len(struct sk_buff *skb)
1667 skb->mac_len = skb->network_header - skb->mac_header;
1670 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1673 return skb->head + skb->inner_transport_header;
1676 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1678 skb->inner_transport_header = skb->data - skb->head;
1681 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1684 skb_reset_inner_transport_header(skb);
1685 skb->inner_transport_header += offset;
1688 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1690 return skb->head + skb->inner_network_header;
1693 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1695 skb->inner_network_header = skb->data - skb->head;
1698 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1701 skb_reset_inner_network_header(skb);
1702 skb->inner_network_header += offset;
1705 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1707 return skb->head + skb->inner_mac_header;
1710 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1712 skb->inner_mac_header = skb->data - skb->head;
1715 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1718 skb_reset_inner_mac_header(skb);
1719 skb->inner_mac_header += offset;
1721 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1723 return skb->transport_header != (typeof(skb->transport_header))~0U;
1726 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1728 return skb->head + skb->transport_header;
1731 static inline void skb_reset_transport_header(struct sk_buff *skb)
1733 skb->transport_header = skb->data - skb->head;
1736 static inline void skb_set_transport_header(struct sk_buff *skb,
1739 skb_reset_transport_header(skb);
1740 skb->transport_header += offset;
1743 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1745 return skb->head + skb->network_header;
1748 static inline void skb_reset_network_header(struct sk_buff *skb)
1750 skb->network_header = skb->data - skb->head;
1753 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1755 skb_reset_network_header(skb);
1756 skb->network_header += offset;
1759 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1761 return skb->head + skb->mac_header;
1764 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1766 return skb->mac_header != (typeof(skb->mac_header))~0U;
1769 static inline void skb_reset_mac_header(struct sk_buff *skb)
1771 skb->mac_header = skb->data - skb->head;
1774 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1776 skb_reset_mac_header(skb);
1777 skb->mac_header += offset;
1780 static inline void skb_pop_mac_header(struct sk_buff *skb)
1782 skb->mac_header = skb->network_header;
1785 static inline void skb_probe_transport_header(struct sk_buff *skb,
1786 const int offset_hint)
1788 struct flow_keys keys;
1790 if (skb_transport_header_was_set(skb))
1792 else if (skb_flow_dissect(skb, &keys))
1793 skb_set_transport_header(skb, keys.thoff);
1795 skb_set_transport_header(skb, offset_hint);
1798 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1800 if (skb_mac_header_was_set(skb)) {
1801 const unsigned char *old_mac = skb_mac_header(skb);
1803 skb_set_mac_header(skb, -skb->mac_len);
1804 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1808 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1810 return skb->csum_start - skb_headroom(skb);
1813 static inline int skb_transport_offset(const struct sk_buff *skb)
1815 return skb_transport_header(skb) - skb->data;
1818 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1820 return skb->transport_header - skb->network_header;
1823 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1825 return skb->inner_transport_header - skb->inner_network_header;
1828 static inline int skb_network_offset(const struct sk_buff *skb)
1830 return skb_network_header(skb) - skb->data;
1833 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1835 return skb_inner_network_header(skb) - skb->data;
1838 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1840 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1844 * CPUs often take a performance hit when accessing unaligned memory
1845 * locations. The actual performance hit varies, it can be small if the
1846 * hardware handles it or large if we have to take an exception and fix it
1849 * Since an ethernet header is 14 bytes network drivers often end up with
1850 * the IP header at an unaligned offset. The IP header can be aligned by
1851 * shifting the start of the packet by 2 bytes. Drivers should do this
1854 * skb_reserve(skb, NET_IP_ALIGN);
1856 * The downside to this alignment of the IP header is that the DMA is now
1857 * unaligned. On some architectures the cost of an unaligned DMA is high
1858 * and this cost outweighs the gains made by aligning the IP header.
1860 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1863 #ifndef NET_IP_ALIGN
1864 #define NET_IP_ALIGN 2
1868 * The networking layer reserves some headroom in skb data (via
1869 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1870 * the header has to grow. In the default case, if the header has to grow
1871 * 32 bytes or less we avoid the reallocation.
1873 * Unfortunately this headroom changes the DMA alignment of the resulting
1874 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1875 * on some architectures. An architecture can override this value,
1876 * perhaps setting it to a cacheline in size (since that will maintain
1877 * cacheline alignment of the DMA). It must be a power of 2.
1879 * Various parts of the networking layer expect at least 32 bytes of
1880 * headroom, you should not reduce this.
1882 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1883 * to reduce average number of cache lines per packet.
1884 * get_rps_cpus() for example only access one 64 bytes aligned block :
1885 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1888 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1891 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1893 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1895 if (unlikely(skb_is_nonlinear(skb))) {
1900 skb_set_tail_pointer(skb, len);
1903 void skb_trim(struct sk_buff *skb, unsigned int len);
1905 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1908 return ___pskb_trim(skb, len);
1909 __skb_trim(skb, len);
1913 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1915 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1919 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1920 * @skb: buffer to alter
1923 * This is identical to pskb_trim except that the caller knows that
1924 * the skb is not cloned so we should never get an error due to out-
1927 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1929 int err = pskb_trim(skb, len);
1934 * skb_orphan - orphan a buffer
1935 * @skb: buffer to orphan
1937 * If a buffer currently has an owner then we call the owner's
1938 * destructor function and make the @skb unowned. The buffer continues
1939 * to exist but is no longer charged to its former owner.
1941 static inline void skb_orphan(struct sk_buff *skb)
1943 if (skb->destructor) {
1944 skb->destructor(skb);
1945 skb->destructor = NULL;
1953 * skb_orphan_frags - orphan the frags contained in a buffer
1954 * @skb: buffer to orphan frags from
1955 * @gfp_mask: allocation mask for replacement pages
1957 * For each frag in the SKB which needs a destructor (i.e. has an
1958 * owner) create a copy of that frag and release the original
1959 * page by calling the destructor.
1961 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1963 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1965 return skb_copy_ubufs(skb, gfp_mask);
1969 * __skb_queue_purge - empty a list
1970 * @list: list to empty
1972 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1973 * the list and one reference dropped. This function does not take the
1974 * list lock and the caller must hold the relevant locks to use it.
1976 void skb_queue_purge(struct sk_buff_head *list);
1977 static inline void __skb_queue_purge(struct sk_buff_head *list)
1979 struct sk_buff *skb;
1980 while ((skb = __skb_dequeue(list)) != NULL)
1984 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1985 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1986 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1988 void *netdev_alloc_frag(unsigned int fragsz);
1990 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1994 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1995 * @dev: network device to receive on
1996 * @length: length to allocate
1998 * Allocate a new &sk_buff and assign it a usage count of one. The
1999 * buffer has unspecified headroom built in. Users should allocate
2000 * the headroom they think they need without accounting for the
2001 * built in space. The built in space is used for optimisations.
2003 * %NULL is returned if there is no free memory. Although this function
2004 * allocates memory it can be called from an interrupt.
2006 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2007 unsigned int length)
2009 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2012 /* legacy helper around __netdev_alloc_skb() */
2013 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2016 return __netdev_alloc_skb(NULL, length, gfp_mask);
2019 /* legacy helper around netdev_alloc_skb() */
2020 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2022 return netdev_alloc_skb(NULL, length);
2026 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2027 unsigned int length, gfp_t gfp)
2029 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2031 if (NET_IP_ALIGN && skb)
2032 skb_reserve(skb, NET_IP_ALIGN);
2036 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2037 unsigned int length)
2039 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2043 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
2044 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2045 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2046 * @order: size of the allocation
2048 * Allocate a new page.
2050 * %NULL is returned if there is no free memory.
2052 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2053 struct sk_buff *skb,
2058 gfp_mask |= __GFP_COLD;
2060 if (!(gfp_mask & __GFP_NOMEMALLOC))
2061 gfp_mask |= __GFP_MEMALLOC;
2063 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2064 if (skb && page && page->pfmemalloc)
2065 skb->pfmemalloc = true;
2071 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2072 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2073 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2075 * Allocate a new page.
2077 * %NULL is returned if there is no free memory.
2079 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2080 struct sk_buff *skb)
2082 return __skb_alloc_pages(gfp_mask, skb, 0);
2086 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2087 * @page: The page that was allocated from skb_alloc_page
2088 * @skb: The skb that may need pfmemalloc set
2090 static inline void skb_propagate_pfmemalloc(struct page *page,
2091 struct sk_buff *skb)
2093 if (page && page->pfmemalloc)
2094 skb->pfmemalloc = true;
2098 * skb_frag_page - retrieve the page referred to by a paged fragment
2099 * @frag: the paged fragment
2101 * Returns the &struct page associated with @frag.
2103 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2105 return frag->page.p;
2109 * __skb_frag_ref - take an addition reference on a paged fragment.
2110 * @frag: the paged fragment
2112 * Takes an additional reference on the paged fragment @frag.
2114 static inline void __skb_frag_ref(skb_frag_t *frag)
2116 get_page(skb_frag_page(frag));
2120 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2122 * @f: the fragment offset.
2124 * Takes an additional reference on the @f'th paged fragment of @skb.
2126 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2128 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2132 * __skb_frag_unref - release a reference on a paged fragment.
2133 * @frag: the paged fragment
2135 * Releases a reference on the paged fragment @frag.
2137 static inline void __skb_frag_unref(skb_frag_t *frag)
2139 put_page(skb_frag_page(frag));
2143 * skb_frag_unref - release a reference on a paged fragment of an skb.
2145 * @f: the fragment offset
2147 * Releases a reference on the @f'th paged fragment of @skb.
2149 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2151 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2155 * skb_frag_address - gets the address of the data contained in a paged fragment
2156 * @frag: the paged fragment buffer
2158 * Returns the address of the data within @frag. The page must already
2161 static inline void *skb_frag_address(const skb_frag_t *frag)
2163 return page_address(skb_frag_page(frag)) + frag->page_offset;
2167 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2168 * @frag: the paged fragment buffer
2170 * Returns the address of the data within @frag. Checks that the page
2171 * is mapped and returns %NULL otherwise.
2173 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2175 void *ptr = page_address(skb_frag_page(frag));
2179 return ptr + frag->page_offset;
2183 * __skb_frag_set_page - sets the page contained in a paged fragment
2184 * @frag: the paged fragment
2185 * @page: the page to set
2187 * Sets the fragment @frag to contain @page.
2189 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2191 frag->page.p = page;
2195 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2197 * @f: the fragment offset
2198 * @page: the page to set
2200 * Sets the @f'th fragment of @skb to contain @page.
2202 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2205 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2208 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2211 * skb_frag_dma_map - maps a paged fragment via the DMA API
2212 * @dev: the device to map the fragment to
2213 * @frag: the paged fragment to map
2214 * @offset: the offset within the fragment (starting at the
2215 * fragment's own offset)
2216 * @size: the number of bytes to map
2217 * @dir: the direction of the mapping (%PCI_DMA_*)
2219 * Maps the page associated with @frag to @device.
2221 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2222 const skb_frag_t *frag,
2223 size_t offset, size_t size,
2224 enum dma_data_direction dir)
2226 return dma_map_page(dev, skb_frag_page(frag),
2227 frag->page_offset + offset, size, dir);
2230 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2233 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2237 * skb_clone_writable - is the header of a clone writable
2238 * @skb: buffer to check
2239 * @len: length up to which to write
2241 * Returns true if modifying the header part of the cloned buffer
2242 * does not requires the data to be copied.
2244 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2246 return !skb_header_cloned(skb) &&
2247 skb_headroom(skb) + len <= skb->hdr_len;
2250 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2255 if (headroom > skb_headroom(skb))
2256 delta = headroom - skb_headroom(skb);
2258 if (delta || cloned)
2259 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2265 * skb_cow - copy header of skb when it is required
2266 * @skb: buffer to cow
2267 * @headroom: needed headroom
2269 * If the skb passed lacks sufficient headroom or its data part
2270 * is shared, data is reallocated. If reallocation fails, an error
2271 * is returned and original skb is not changed.
2273 * The result is skb with writable area skb->head...skb->tail
2274 * and at least @headroom of space at head.
2276 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2278 return __skb_cow(skb, headroom, skb_cloned(skb));
2282 * skb_cow_head - skb_cow but only making the head writable
2283 * @skb: buffer to cow
2284 * @headroom: needed headroom
2286 * This function is identical to skb_cow except that we replace the
2287 * skb_cloned check by skb_header_cloned. It should be used when
2288 * you only need to push on some header and do not need to modify
2291 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2293 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2297 * skb_padto - pad an skbuff up to a minimal size
2298 * @skb: buffer to pad
2299 * @len: minimal length
2301 * Pads up a buffer to ensure the trailing bytes exist and are
2302 * blanked. If the buffer already contains sufficient data it
2303 * is untouched. Otherwise it is extended. Returns zero on
2304 * success. The skb is freed on error.
2307 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2309 unsigned int size = skb->len;
2310 if (likely(size >= len))
2312 return skb_pad(skb, len - size);
2315 static inline int skb_add_data(struct sk_buff *skb,
2316 char __user *from, int copy)
2318 const int off = skb->len;
2320 if (skb->ip_summed == CHECKSUM_NONE) {
2322 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2325 skb->csum = csum_block_add(skb->csum, csum, off);
2328 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2331 __skb_trim(skb, off);
2335 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2336 const struct page *page, int off)
2339 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2341 return page == skb_frag_page(frag) &&
2342 off == frag->page_offset + skb_frag_size(frag);
2347 static inline int __skb_linearize(struct sk_buff *skb)
2349 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2353 * skb_linearize - convert paged skb to linear one
2354 * @skb: buffer to linarize
2356 * If there is no free memory -ENOMEM is returned, otherwise zero
2357 * is returned and the old skb data released.
2359 static inline int skb_linearize(struct sk_buff *skb)
2361 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2365 * skb_has_shared_frag - can any frag be overwritten
2366 * @skb: buffer to test
2368 * Return true if the skb has at least one frag that might be modified
2369 * by an external entity (as in vmsplice()/sendfile())
2371 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2373 return skb_is_nonlinear(skb) &&
2374 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2378 * skb_linearize_cow - make sure skb is linear and writable
2379 * @skb: buffer to process
2381 * If there is no free memory -ENOMEM is returned, otherwise zero
2382 * is returned and the old skb data released.
2384 static inline int skb_linearize_cow(struct sk_buff *skb)
2386 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2387 __skb_linearize(skb) : 0;
2391 * skb_postpull_rcsum - update checksum for received skb after pull
2392 * @skb: buffer to update
2393 * @start: start of data before pull
2394 * @len: length of data pulled
2396 * After doing a pull on a received packet, you need to call this to
2397 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2398 * CHECKSUM_NONE so that it can be recomputed from scratch.
2401 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2402 const void *start, unsigned int len)
2404 if (skb->ip_summed == CHECKSUM_COMPLETE)
2405 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2408 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2411 * pskb_trim_rcsum - trim received skb and update checksum
2412 * @skb: buffer to trim
2415 * This is exactly the same as pskb_trim except that it ensures the
2416 * checksum of received packets are still valid after the operation.
2419 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2421 if (likely(len >= skb->len))
2423 if (skb->ip_summed == CHECKSUM_COMPLETE)
2424 skb->ip_summed = CHECKSUM_NONE;
2425 return __pskb_trim(skb, len);
2428 #define skb_queue_walk(queue, skb) \
2429 for (skb = (queue)->next; \
2430 skb != (struct sk_buff *)(queue); \
2433 #define skb_queue_walk_safe(queue, skb, tmp) \
2434 for (skb = (queue)->next, tmp = skb->next; \
2435 skb != (struct sk_buff *)(queue); \
2436 skb = tmp, tmp = skb->next)
2438 #define skb_queue_walk_from(queue, skb) \
2439 for (; skb != (struct sk_buff *)(queue); \
2442 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2443 for (tmp = skb->next; \
2444 skb != (struct sk_buff *)(queue); \
2445 skb = tmp, tmp = skb->next)
2447 #define skb_queue_reverse_walk(queue, skb) \
2448 for (skb = (queue)->prev; \
2449 skb != (struct sk_buff *)(queue); \
2452 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2453 for (skb = (queue)->prev, tmp = skb->prev; \
2454 skb != (struct sk_buff *)(queue); \
2455 skb = tmp, tmp = skb->prev)
2457 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2458 for (tmp = skb->prev; \
2459 skb != (struct sk_buff *)(queue); \
2460 skb = tmp, tmp = skb->prev)
2462 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2464 return skb_shinfo(skb)->frag_list != NULL;
2467 static inline void skb_frag_list_init(struct sk_buff *skb)
2469 skb_shinfo(skb)->frag_list = NULL;
2472 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2474 frag->next = skb_shinfo(skb)->frag_list;
2475 skb_shinfo(skb)->frag_list = frag;
2478 #define skb_walk_frags(skb, iter) \
2479 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2481 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2482 int *peeked, int *off, int *err);
2483 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2485 unsigned int datagram_poll(struct file *file, struct socket *sock,
2486 struct poll_table_struct *wait);
2487 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2488 struct iovec *to, int size);
2489 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2491 int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2492 const struct iovec *from, int from_offset,
2494 int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2495 int offset, size_t count);
2496 int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2497 const struct iovec *to, int to_offset,
2499 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2500 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2501 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2502 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2503 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2504 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2505 int len, __wsum csum);
2506 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2507 struct pipe_inode_info *pipe, unsigned int len,
2508 unsigned int flags);
2509 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2510 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2511 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2513 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2514 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2515 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2516 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2517 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2519 struct skb_checksum_ops {
2520 __wsum (*update)(const void *mem, int len, __wsum wsum);
2521 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2524 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2525 __wsum csum, const struct skb_checksum_ops *ops);
2526 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2529 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2530 int len, void *buffer)
2532 int hlen = skb_headlen(skb);
2534 if (hlen - offset >= len)
2535 return skb->data + offset;
2537 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2544 * skb_needs_linearize - check if we need to linearize a given skb
2545 * depending on the given device features.
2546 * @skb: socket buffer to check
2547 * @features: net device features
2549 * Returns true if either:
2550 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2551 * 2. skb is fragmented and the device does not support SG.
2553 static inline bool skb_needs_linearize(struct sk_buff *skb,
2554 netdev_features_t features)
2556 return skb_is_nonlinear(skb) &&
2557 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2558 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2561 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2563 const unsigned int len)
2565 memcpy(to, skb->data, len);
2568 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2569 const int offset, void *to,
2570 const unsigned int len)
2572 memcpy(to, skb->data + offset, len);
2575 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2577 const unsigned int len)
2579 memcpy(skb->data, from, len);
2582 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2585 const unsigned int len)
2587 memcpy(skb->data + offset, from, len);
2590 void skb_init(void);
2592 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2598 * skb_get_timestamp - get timestamp from a skb
2599 * @skb: skb to get stamp from
2600 * @stamp: pointer to struct timeval to store stamp in
2602 * Timestamps are stored in the skb as offsets to a base timestamp.
2603 * This function converts the offset back to a struct timeval and stores
2606 static inline void skb_get_timestamp(const struct sk_buff *skb,
2607 struct timeval *stamp)
2609 *stamp = ktime_to_timeval(skb->tstamp);
2612 static inline void skb_get_timestampns(const struct sk_buff *skb,
2613 struct timespec *stamp)
2615 *stamp = ktime_to_timespec(skb->tstamp);
2618 static inline void __net_timestamp(struct sk_buff *skb)
2620 skb->tstamp = ktime_get_real();
2623 static inline ktime_t net_timedelta(ktime_t t)
2625 return ktime_sub(ktime_get_real(), t);
2628 static inline ktime_t net_invalid_timestamp(void)
2630 return ktime_set(0, 0);
2633 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2635 void skb_clone_tx_timestamp(struct sk_buff *skb);
2636 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2638 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2640 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2644 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2649 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2652 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2654 * PHY drivers may accept clones of transmitted packets for
2655 * timestamping via their phy_driver.txtstamp method. These drivers
2656 * must call this function to return the skb back to the stack, with
2657 * or without a timestamp.
2659 * @skb: clone of the the original outgoing packet
2660 * @hwtstamps: hardware time stamps, may be NULL if not available
2663 void skb_complete_tx_timestamp(struct sk_buff *skb,
2664 struct skb_shared_hwtstamps *hwtstamps);
2667 * skb_tstamp_tx - queue clone of skb with send time stamps
2668 * @orig_skb: the original outgoing packet
2669 * @hwtstamps: hardware time stamps, may be NULL if not available
2671 * If the skb has a socket associated, then this function clones the
2672 * skb (thus sharing the actual data and optional structures), stores
2673 * the optional hardware time stamping information (if non NULL) or
2674 * generates a software time stamp (otherwise), then queues the clone
2675 * to the error queue of the socket. Errors are silently ignored.
2677 void skb_tstamp_tx(struct sk_buff *orig_skb,
2678 struct skb_shared_hwtstamps *hwtstamps);
2680 static inline void sw_tx_timestamp(struct sk_buff *skb)
2682 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2683 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2684 skb_tstamp_tx(skb, NULL);
2688 * skb_tx_timestamp() - Driver hook for transmit timestamping
2690 * Ethernet MAC Drivers should call this function in their hard_xmit()
2691 * function immediately before giving the sk_buff to the MAC hardware.
2693 * Specifically, one should make absolutely sure that this function is
2694 * called before TX completion of this packet can trigger. Otherwise
2695 * the packet could potentially already be freed.
2697 * @skb: A socket buffer.
2699 static inline void skb_tx_timestamp(struct sk_buff *skb)
2701 skb_clone_tx_timestamp(skb);
2702 sw_tx_timestamp(skb);
2706 * skb_complete_wifi_ack - deliver skb with wifi status
2708 * @skb: the original outgoing packet
2709 * @acked: ack status
2712 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2714 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2715 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2717 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2719 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2723 * skb_checksum_complete - Calculate checksum of an entire packet
2724 * @skb: packet to process
2726 * This function calculates the checksum over the entire packet plus
2727 * the value of skb->csum. The latter can be used to supply the
2728 * checksum of a pseudo header as used by TCP/UDP. It returns the
2731 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2732 * this function can be used to verify that checksum on received
2733 * packets. In that case the function should return zero if the
2734 * checksum is correct. In particular, this function will return zero
2735 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2736 * hardware has already verified the correctness of the checksum.
2738 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2740 return skb_csum_unnecessary(skb) ?
2741 0 : __skb_checksum_complete(skb);
2744 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2745 void nf_conntrack_destroy(struct nf_conntrack *nfct);
2746 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2748 if (nfct && atomic_dec_and_test(&nfct->use))
2749 nf_conntrack_destroy(nfct);
2751 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2754 atomic_inc(&nfct->use);
2757 #ifdef CONFIG_BRIDGE_NETFILTER
2758 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2760 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2763 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2766 atomic_inc(&nf_bridge->use);
2768 #endif /* CONFIG_BRIDGE_NETFILTER */
2769 static inline void nf_reset(struct sk_buff *skb)
2771 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2772 nf_conntrack_put(skb->nfct);
2775 #ifdef CONFIG_BRIDGE_NETFILTER
2776 nf_bridge_put(skb->nf_bridge);
2777 skb->nf_bridge = NULL;
2781 static inline void nf_reset_trace(struct sk_buff *skb)
2783 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2788 /* Note: This doesn't put any conntrack and bridge info in dst. */
2789 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2791 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2792 dst->nfct = src->nfct;
2793 nf_conntrack_get(src->nfct);
2794 dst->nfctinfo = src->nfctinfo;
2796 #ifdef CONFIG_BRIDGE_NETFILTER
2797 dst->nf_bridge = src->nf_bridge;
2798 nf_bridge_get(src->nf_bridge);
2800 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2801 dst->nf_trace = src->nf_trace;
2805 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2807 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2808 nf_conntrack_put(dst->nfct);
2810 #ifdef CONFIG_BRIDGE_NETFILTER
2811 nf_bridge_put(dst->nf_bridge);
2813 __nf_copy(dst, src);
2816 #ifdef CONFIG_NETWORK_SECMARK
2817 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2819 to->secmark = from->secmark;
2822 static inline void skb_init_secmark(struct sk_buff *skb)
2827 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2830 static inline void skb_init_secmark(struct sk_buff *skb)
2834 static inline bool skb_irq_freeable(const struct sk_buff *skb)
2836 return !skb->destructor &&
2837 #if IS_ENABLED(CONFIG_XFRM)
2840 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
2843 !skb->_skb_refdst &&
2844 !skb_has_frag_list(skb);
2847 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2849 skb->queue_mapping = queue_mapping;
2852 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2854 return skb->queue_mapping;
2857 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2859 to->queue_mapping = from->queue_mapping;
2862 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2864 skb->queue_mapping = rx_queue + 1;
2867 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2869 return skb->queue_mapping - 1;
2872 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2874 return skb->queue_mapping != 0;
2877 u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2878 unsigned int num_tx_queues);
2880 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2889 /* Keeps track of mac header offset relative to skb->head.
2890 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2891 * For non-tunnel skb it points to skb_mac_header() and for
2892 * tunnel skb it points to outer mac header.
2893 * Keeps track of level of encapsulation of network headers.
2899 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2901 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2903 return (skb_mac_header(inner_skb) - inner_skb->head) -
2904 SKB_GSO_CB(inner_skb)->mac_offset;
2907 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2909 int new_headroom, headroom;
2912 headroom = skb_headroom(skb);
2913 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2917 new_headroom = skb_headroom(skb);
2918 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2922 static inline bool skb_is_gso(const struct sk_buff *skb)
2924 return skb_shinfo(skb)->gso_size;
2927 /* Note: Should be called only if skb_is_gso(skb) is true */
2928 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2930 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2933 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2935 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2937 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2938 * wanted then gso_type will be set. */
2939 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2941 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2942 unlikely(shinfo->gso_type == 0)) {
2943 __skb_warn_lro_forwarding(skb);
2949 static inline void skb_forward_csum(struct sk_buff *skb)
2951 /* Unfortunately we don't support this one. Any brave souls? */
2952 if (skb->ip_summed == CHECKSUM_COMPLETE)
2953 skb->ip_summed = CHECKSUM_NONE;
2957 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2958 * @skb: skb to check
2960 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2961 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2962 * use this helper, to document places where we make this assertion.
2964 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2967 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2971 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2973 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
2975 u32 __skb_get_poff(const struct sk_buff *skb);
2978 * skb_head_is_locked - Determine if the skb->head is locked down
2979 * @skb: skb to check
2981 * The head on skbs build around a head frag can be removed if they are
2982 * not cloned. This function returns true if the skb head is locked down
2983 * due to either being allocated via kmalloc, or by being a clone with
2984 * multiple references to the head.
2986 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2988 return !skb->head_frag || skb_cloned(skb);
2992 * skb_gso_network_seglen - Return length of individual segments of a gso packet
2996 * skb_gso_network_seglen is used to determine the real size of the
2997 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
2999 * The MAC/L2 header is not accounted for.
3001 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3003 unsigned int hdr_len = skb_transport_header(skb) -
3004 skb_network_header(skb);
3005 return hdr_len + skb_gso_transport_seglen(skb);
3007 #endif /* __KERNEL__ */
3008 #endif /* _LINUX_SKBUFF_H */