1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * Definitions for the 'struct sk_buff' memory handlers.
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
40 #include <net/page_pool.h>
41 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
42 #include <linux/netfilter/nf_conntrack_common.h>
45 /* The interface for checksum offload between the stack and networking drivers
48 * A. IP checksum related features
50 * Drivers advertise checksum offload capabilities in the features of a device.
51 * From the stack's point of view these are capabilities offered by the driver.
52 * A driver typically only advertises features that it is capable of offloading
55 * The checksum related features are:
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
67 * is TCP or UDP. The IPv4 header may contain IP options.
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
75 * IPv6|UDP where the Next Header field in the IPv6
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
83 * This flag is only used to disable the RX checksum
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
88 * B. Checksumming of received packets by device. Indication of checksum
89 * verification is set in skb->ip_summed. Possible values are:
93 * Device did not checksum this packet e.g. due to lack of capabilities.
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
97 * CHECKSUM_UNNECESSARY:
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
113 * FCOE: indicates the CRC in FC frame has been validated.
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
119 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
120 * two. If the device were only able to verify the UDP checksum and not
121 * GRE, either because it doesn't support GRE checksum or because GRE
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
127 * This is the most generic way. The device supplied checksum of the _whole_
128 * packet as seen by netif_rx() and fills in skb->csum. This means the
129 * hardware doesn't need to parse L3/L4 headers to implement this.
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
153 * The driver is required to checksum the packet as seen by hard_start_xmit()
154 * from skb->csum_start up to the end, and to record/write the checksum at
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
157 * offset of the packet, but it should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum -- it is the
159 * purview of the stack to validate that csum_start and csum_offset are set
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
177 * The skb was already checksummed by the protocol, or a checksum is not
180 * CHECKSUM_UNNECESSARY:
182 * This has the same meaning as CHECKSUM_NONE for checksum offload on
186 * Not used in checksum output. If a driver observes a packet with this value
187 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
189 * D. Non-IP checksum (CRC) offloads
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
193 * will set csum_start and csum_offset accordingly, set ip_summed to
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
204 * accordingly. Note that there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
206 * both IP checksum offload and FCOE CRC offload must verify which offload
207 * is configured for a packet, presumably by inspecting packet headers.
209 * E. Checksumming on output with GSO.
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
216 * csum_offset are set to refer to the outermost checksum being offloaded
217 * (two offloaded checksums are possible with UDP encapsulation).
220 /* Don't change this without changing skb_csum_unnecessary! */
221 #define CHECKSUM_NONE 0
222 #define CHECKSUM_UNNECESSARY 1
223 #define CHECKSUM_COMPLETE 2
224 #define CHECKSUM_PARTIAL 3
226 /* Maximum value in skb->csum_level */
227 #define SKB_MAX_CSUM_LEVEL 3
229 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
230 #define SKB_WITH_OVERHEAD(X) \
231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
232 #define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
234 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
237 /* return minimum truesize of one skb containing X bytes of data */
238 #define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
242 struct ahash_request;
245 struct pipe_inode_info;
252 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
253 struct nf_bridge_info {
255 BRNF_PROTO_UNCHANGED,
263 struct net_device *physindev;
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
268 /* prerouting: detect dnat in orig/reply direction */
270 struct in6_addr ipv6_daddr;
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
276 char neigh_header[8];
281 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
282 /* Chain in tc_skb_ext will be used to share the tc chain with
283 * ovs recirc_id. It will be set to the current chain by tc
284 * and read by ovs to recirc_id.
293 struct sk_buff_head {
294 /* These two members must be first. */
295 struct sk_buff *next;
296 struct sk_buff *prev;
304 /* To allow 64K frame to be packed as single skb without frag_list we
305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
306 * buffers which do not start on a page boundary.
308 * Since GRO uses frags we allocate at least 16 regardless of page
311 #if (65536/PAGE_SIZE + 1) < 16
312 #define MAX_SKB_FRAGS 16UL
314 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
316 extern int sysctl_max_skb_frags;
318 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
319 * segment using its current segmentation instead.
321 #define GSO_BY_FRAGS 0xFFFF
323 typedef struct bio_vec skb_frag_t;
326 * skb_frag_size() - Returns the size of a skb fragment
327 * @frag: skb fragment
329 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
335 * skb_frag_size_set() - Sets the size of a skb fragment
336 * @frag: skb fragment
337 * @size: size of fragment
339 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
346 * @frag: skb fragment
347 * @delta: value to add
349 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
351 frag->bv_len += delta;
355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
356 * @frag: skb fragment
357 * @delta: value to subtract
359 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
361 frag->bv_len -= delta;
365 * skb_frag_must_loop - Test if %p is a high memory page
366 * @p: fragment's page
368 static inline bool skb_frag_must_loop(struct page *p)
370 #if defined(CONFIG_HIGHMEM)
371 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
378 * skb_frag_foreach_page - loop over pages in a fragment
380 * @f: skb frag to operate on
381 * @f_off: offset from start of f->bv_page
382 * @f_len: length from f_off to loop over
383 * @p: (temp var) current page
384 * @p_off: (temp var) offset from start of current page,
385 * non-zero only on first page.
386 * @p_len: (temp var) length in current page,
387 * < PAGE_SIZE only on first and last page.
388 * @copied: (temp var) length so far, excluding current p_len.
390 * A fragment can hold a compound page, in which case per-page
391 * operations, notably kmap_atomic, must be called for each
394 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
396 p_off = (f_off) & (PAGE_SIZE - 1), \
397 p_len = skb_frag_must_loop(p) ? \
398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
401 copied += p_len, p++, p_off = 0, \
402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
404 #define HAVE_HW_TIME_STAMP
407 * struct skb_shared_hwtstamps - hardware time stamps
408 * @hwtstamp: hardware time stamp transformed into duration
409 * since arbitrary point in time
411 * Software time stamps generated by ktime_get_real() are stored in
414 * hwtstamps can only be compared against other hwtstamps from
417 * This structure is attached to packets as part of the
418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
420 struct skb_shared_hwtstamps {
424 /* Definitions for tx_flags in struct skb_shared_info */
426 /* generate hardware time stamp */
427 SKBTX_HW_TSTAMP = 1 << 0,
429 /* generate software time stamp when queueing packet to NIC */
430 SKBTX_SW_TSTAMP = 1 << 1,
432 /* device driver is going to provide hardware time stamp */
433 SKBTX_IN_PROGRESS = 1 << 2,
435 /* generate wifi status information (where possible) */
436 SKBTX_WIFI_STATUS = 1 << 4,
438 /* generate software time stamp when entering packet scheduling */
439 SKBTX_SCHED_TSTAMP = 1 << 6,
442 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
444 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
446 /* Definitions for flags in struct skb_shared_info */
448 /* use zcopy routines */
449 SKBFL_ZEROCOPY_ENABLE = BIT(0),
451 /* This indicates at least one fragment might be overwritten
452 * (as in vmsplice(), sendfile() ...)
453 * If we need to compute a TX checksum, we'll need to copy
454 * all frags to avoid possible bad checksum
456 SKBFL_SHARED_FRAG = BIT(1),
459 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
462 * The callback notifies userspace to release buffers when skb DMA is done in
463 * lower device, the skb last reference should be 0 when calling this.
464 * The zerocopy_success argument is true if zero copy transmit occurred,
465 * false on data copy or out of memory error caused by data copy attempt.
466 * The ctx field is used to track device context.
467 * The desc field is used to track userspace buffer index.
470 void (*callback)(struct sk_buff *, struct ubuf_info *,
471 bool zerocopy_success);
488 struct user_struct *user;
493 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
495 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
496 void mm_unaccount_pinned_pages(struct mmpin *mmp);
498 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
499 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
500 struct ubuf_info *uarg);
502 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
504 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
507 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
508 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
509 struct msghdr *msg, int len,
510 struct ubuf_info *uarg);
512 /* This data is invariant across clones and lives at
513 * the end of the header data, ie. at skb->end.
515 struct skb_shared_info {
520 unsigned short gso_size;
521 /* Warning: this field is not always filled in (UFO)! */
522 unsigned short gso_segs;
523 struct sk_buff *frag_list;
524 struct skb_shared_hwtstamps hwtstamps;
525 unsigned int gso_type;
529 * Warning : all fields before dataref are cleared in __alloc_skb()
533 /* Intermediate layers must ensure that destructor_arg
534 * remains valid until skb destructor */
535 void * destructor_arg;
537 /* must be last field, see pskb_expand_head() */
538 skb_frag_t frags[MAX_SKB_FRAGS];
541 /* We divide dataref into two halves. The higher 16 bits hold references
542 * to the payload part of skb->data. The lower 16 bits hold references to
543 * the entire skb->data. A clone of a headerless skb holds the length of
544 * the header in skb->hdr_len.
546 * All users must obey the rule that the skb->data reference count must be
547 * greater than or equal to the payload reference count.
549 * Holding a reference to the payload part means that the user does not
550 * care about modifications to the header part of skb->data.
552 #define SKB_DATAREF_SHIFT 16
553 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
557 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
558 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
559 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
563 SKB_GSO_TCPV4 = 1 << 0,
565 /* This indicates the skb is from an untrusted source. */
566 SKB_GSO_DODGY = 1 << 1,
568 /* This indicates the tcp segment has CWR set. */
569 SKB_GSO_TCP_ECN = 1 << 2,
571 SKB_GSO_TCP_FIXEDID = 1 << 3,
573 SKB_GSO_TCPV6 = 1 << 4,
575 SKB_GSO_FCOE = 1 << 5,
577 SKB_GSO_GRE = 1 << 6,
579 SKB_GSO_GRE_CSUM = 1 << 7,
581 SKB_GSO_IPXIP4 = 1 << 8,
583 SKB_GSO_IPXIP6 = 1 << 9,
585 SKB_GSO_UDP_TUNNEL = 1 << 10,
587 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
589 SKB_GSO_PARTIAL = 1 << 12,
591 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
593 SKB_GSO_SCTP = 1 << 14,
595 SKB_GSO_ESP = 1 << 15,
597 SKB_GSO_UDP = 1 << 16,
599 SKB_GSO_UDP_L4 = 1 << 17,
601 SKB_GSO_FRAGLIST = 1 << 18,
604 #if BITS_PER_LONG > 32
605 #define NET_SKBUFF_DATA_USES_OFFSET 1
608 #ifdef NET_SKBUFF_DATA_USES_OFFSET
609 typedef unsigned int sk_buff_data_t;
611 typedef unsigned char *sk_buff_data_t;
615 * struct sk_buff - socket buffer
616 * @next: Next buffer in list
617 * @prev: Previous buffer in list
618 * @tstamp: Time we arrived/left
619 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
620 * for retransmit timer
621 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
623 * @sk: Socket we are owned by
624 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
625 * fragmentation management
626 * @dev: Device we arrived on/are leaving by
627 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
628 * @cb: Control buffer. Free for use by every layer. Put private vars here
629 * @_skb_refdst: destination entry (with norefcount bit)
630 * @sp: the security path, used for xfrm
631 * @len: Length of actual data
632 * @data_len: Data length
633 * @mac_len: Length of link layer header
634 * @hdr_len: writable header length of cloned skb
635 * @csum: Checksum (must include start/offset pair)
636 * @csum_start: Offset from skb->head where checksumming should start
637 * @csum_offset: Offset from csum_start where checksum should be stored
638 * @priority: Packet queueing priority
639 * @ignore_df: allow local fragmentation
640 * @cloned: Head may be cloned (check refcnt to be sure)
641 * @ip_summed: Driver fed us an IP checksum
642 * @nohdr: Payload reference only, must not modify header
643 * @pkt_type: Packet class
644 * @fclone: skbuff clone status
645 * @ipvs_property: skbuff is owned by ipvs
646 * @inner_protocol_type: whether the inner protocol is
647 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
648 * @remcsum_offload: remote checksum offload is enabled
649 * @offload_fwd_mark: Packet was L2-forwarded in hardware
650 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
651 * @tc_skip_classify: do not classify packet. set by IFB device
652 * @tc_at_ingress: used within tc_classify to distinguish in/egress
653 * @redirected: packet was redirected by packet classifier
654 * @from_ingress: packet was redirected from the ingress path
655 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
656 * @peeked: this packet has been seen already, so stats have been
657 * done for it, don't do them again
658 * @nf_trace: netfilter packet trace flag
659 * @protocol: Packet protocol from driver
660 * @destructor: Destruct function
661 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
662 * @_sk_redir: socket redirection information for skmsg
663 * @_nfct: Associated connection, if any (with nfctinfo bits)
664 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
665 * @skb_iif: ifindex of device we arrived on
666 * @tc_index: Traffic control index
667 * @hash: the packet hash
668 * @queue_mapping: Queue mapping for multiqueue devices
669 * @head_frag: skb was allocated from page fragments,
670 * not allocated by kmalloc() or vmalloc().
671 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
672 * @pp_recycle: mark the packet for recycling instead of freeing (implies
673 * page_pool support on driver)
674 * @active_extensions: active extensions (skb_ext_id types)
675 * @ndisc_nodetype: router type (from link layer)
676 * @ooo_okay: allow the mapping of a socket to a queue to be changed
677 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
679 * @sw_hash: indicates hash was computed in software stack
680 * @wifi_acked_valid: wifi_acked was set
681 * @wifi_acked: whether frame was acked on wifi or not
682 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
683 * @encapsulation: indicates the inner headers in the skbuff are valid
684 * @encap_hdr_csum: software checksum is needed
685 * @csum_valid: checksum is already valid
686 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
687 * @csum_complete_sw: checksum was completed by software
688 * @csum_level: indicates the number of consecutive checksums found in
689 * the packet minus one that have been verified as
690 * CHECKSUM_UNNECESSARY (max 3)
691 * @dst_pending_confirm: need to confirm neighbour
692 * @decrypted: Decrypted SKB
693 * @slow_gro: state present at GRO time, slower prepare step required
694 * @napi_id: id of the NAPI struct this skb came from
695 * @sender_cpu: (aka @napi_id) source CPU in XPS
696 * @secmark: security marking
697 * @mark: Generic packet mark
698 * @reserved_tailroom: (aka @mark) number of bytes of free space available
699 * at the tail of an sk_buff
700 * @vlan_present: VLAN tag is present
701 * @vlan_proto: vlan encapsulation protocol
702 * @vlan_tci: vlan tag control information
703 * @inner_protocol: Protocol (encapsulation)
704 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
705 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
706 * @inner_transport_header: Inner transport layer header (encapsulation)
707 * @inner_network_header: Network layer header (encapsulation)
708 * @inner_mac_header: Link layer header (encapsulation)
709 * @transport_header: Transport layer header
710 * @network_header: Network layer header
711 * @mac_header: Link layer header
712 * @kcov_handle: KCOV remote handle for remote coverage collection
713 * @tail: Tail pointer
715 * @head: Head of buffer
716 * @data: Data head pointer
717 * @truesize: Buffer size
718 * @users: User count - see {datagram,tcp}.c
719 * @extensions: allocated extensions, valid if active_extensions is nonzero
725 /* These two members must be first. */
726 struct sk_buff *next;
727 struct sk_buff *prev;
730 struct net_device *dev;
731 /* Some protocols might use this space to store information,
732 * while device pointer would be NULL.
733 * UDP receive path is one user.
735 unsigned long dev_scratch;
738 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
739 struct list_head list;
744 int ip_defrag_offset;
749 u64 skb_mstamp_ns; /* earliest departure time */
752 * This is the control buffer. It is free to use for every
753 * layer. Please put your private variables there. If you
754 * want to keep them across layers you have to do a skb_clone()
755 * first. This is owned by whoever has the skb queued ATM.
757 char cb[48] __aligned(8);
761 unsigned long _skb_refdst;
762 void (*destructor)(struct sk_buff *skb);
764 struct list_head tcp_tsorted_anchor;
765 #ifdef CONFIG_NET_SOCK_MSG
766 unsigned long _sk_redir;
770 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
778 /* Following fields are _not_ copied in __copy_skb_header()
779 * Note that queue_mapping is here mostly to fill a hole.
783 /* if you move cloned around you also must adapt those constants */
784 #ifdef __BIG_ENDIAN_BITFIELD
785 #define CLONED_MASK (1 << 7)
787 #define CLONED_MASK 1
789 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
792 __u8 __cloned_offset[0];
800 pp_recycle:1; /* page_pool recycle indicator */
801 #ifdef CONFIG_SKB_EXTENSIONS
802 __u8 active_extensions;
805 /* fields enclosed in headers_start/headers_end are copied
806 * using a single memcpy() in __copy_skb_header()
809 __u32 headers_start[0];
812 /* if you move pkt_type around you also must adapt those constants */
813 #ifdef __BIG_ENDIAN_BITFIELD
814 #define PKT_TYPE_MAX (7 << 5)
816 #define PKT_TYPE_MAX 7
818 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
821 __u8 __pkt_type_offset[0];
831 __u8 wifi_acked_valid:1;
834 /* Indicates the inner headers are valid in the skbuff. */
835 __u8 encapsulation:1;
836 __u8 encap_hdr_csum:1;
839 #ifdef __BIG_ENDIAN_BITFIELD
840 #define PKT_VLAN_PRESENT_BIT 7
842 #define PKT_VLAN_PRESENT_BIT 0
844 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
846 __u8 __pkt_vlan_present_offset[0];
849 __u8 csum_complete_sw:1;
851 __u8 csum_not_inet:1;
852 __u8 dst_pending_confirm:1;
853 #ifdef CONFIG_IPV6_NDISC_NODETYPE
854 __u8 ndisc_nodetype:2;
857 __u8 ipvs_property:1;
858 __u8 inner_protocol_type:1;
859 __u8 remcsum_offload:1;
860 #ifdef CONFIG_NET_SWITCHDEV
861 __u8 offload_fwd_mark:1;
862 __u8 offload_l3_fwd_mark:1;
864 #ifdef CONFIG_NET_CLS_ACT
865 __u8 tc_skip_classify:1;
866 __u8 tc_at_ingress:1;
869 #ifdef CONFIG_NET_REDIRECT
872 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
873 __u8 nf_skip_egress:1;
875 #ifdef CONFIG_TLS_DEVICE
880 #ifdef CONFIG_NET_SCHED
881 __u16 tc_index; /* traffic control index */
896 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
898 unsigned int napi_id;
899 unsigned int sender_cpu;
902 #ifdef CONFIG_NETWORK_SECMARK
908 __u32 reserved_tailroom;
912 __be16 inner_protocol;
916 __u16 inner_transport_header;
917 __u16 inner_network_header;
918 __u16 inner_mac_header;
921 __u16 transport_header;
922 __u16 network_header;
930 __u32 headers_end[0];
933 /* These elements must be at the end, see alloc_skb() for details. */
938 unsigned int truesize;
941 #ifdef CONFIG_SKB_EXTENSIONS
942 /* only useable after checking ->active_extensions != 0 */
943 struct skb_ext *extensions;
949 * Handling routines are only of interest to the kernel
952 #define SKB_ALLOC_FCLONE 0x01
953 #define SKB_ALLOC_RX 0x02
954 #define SKB_ALLOC_NAPI 0x04
957 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
960 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
962 return unlikely(skb->pfmemalloc);
966 * skb might have a dst pointer attached, refcounted or not.
967 * _skb_refdst low order bit is set if refcount was _not_ taken
969 #define SKB_DST_NOREF 1UL
970 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
973 * skb_dst - returns skb dst_entry
976 * Returns skb dst_entry, regardless of reference taken or not.
978 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
980 /* If refdst was not refcounted, check we still are in a
981 * rcu_read_lock section
983 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
984 !rcu_read_lock_held() &&
985 !rcu_read_lock_bh_held());
986 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
990 * skb_dst_set - sets skb dst
994 * Sets skb dst, assuming a reference was taken on dst and should
995 * be released by skb_dst_drop()
997 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
999 skb->slow_gro |= !!dst;
1000 skb->_skb_refdst = (unsigned long)dst;
1004 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1008 * Sets skb dst, assuming a reference was not taken on dst.
1009 * If dst entry is cached, we do not take reference and dst_release
1010 * will be avoided by refdst_drop. If dst entry is not cached, we take
1011 * reference, so that last dst_release can destroy the dst immediately.
1013 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1015 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1016 skb->slow_gro |= !!dst;
1017 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1021 * skb_dst_is_noref - Test if skb dst isn't refcounted
1024 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1026 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1030 * skb_rtable - Returns the skb &rtable
1033 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1035 return (struct rtable *)skb_dst(skb);
1038 /* For mangling skb->pkt_type from user space side from applications
1039 * such as nft, tc, etc, we only allow a conservative subset of
1040 * possible pkt_types to be set.
1042 static inline bool skb_pkt_type_ok(u32 ptype)
1044 return ptype <= PACKET_OTHERHOST;
1048 * skb_napi_id - Returns the skb's NAPI id
1051 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1053 #ifdef CONFIG_NET_RX_BUSY_POLL
1054 return skb->napi_id;
1061 * skb_unref - decrement the skb's reference count
1064 * Returns true if we can free the skb.
1066 static inline bool skb_unref(struct sk_buff *skb)
1070 if (likely(refcount_read(&skb->users) == 1))
1072 else if (likely(!refcount_dec_and_test(&skb->users)))
1078 void skb_release_head_state(struct sk_buff *skb);
1079 void kfree_skb(struct sk_buff *skb);
1080 void kfree_skb_list(struct sk_buff *segs);
1081 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1082 void skb_tx_error(struct sk_buff *skb);
1084 #ifdef CONFIG_TRACEPOINTS
1085 void consume_skb(struct sk_buff *skb);
1087 static inline void consume_skb(struct sk_buff *skb)
1089 return kfree_skb(skb);
1093 void __consume_stateless_skb(struct sk_buff *skb);
1094 void __kfree_skb(struct sk_buff *skb);
1095 extern struct kmem_cache *skbuff_head_cache;
1097 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1098 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1099 bool *fragstolen, int *delta_truesize);
1101 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1103 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1104 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1105 struct sk_buff *build_skb_around(struct sk_buff *skb,
1106 void *data, unsigned int frag_size);
1108 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1111 * alloc_skb - allocate a network buffer
1112 * @size: size to allocate
1113 * @priority: allocation mask
1115 * This function is a convenient wrapper around __alloc_skb().
1117 static inline struct sk_buff *alloc_skb(unsigned int size,
1120 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1123 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1124 unsigned long data_len,
1128 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1130 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1131 struct sk_buff_fclones {
1132 struct sk_buff skb1;
1134 struct sk_buff skb2;
1136 refcount_t fclone_ref;
1140 * skb_fclone_busy - check if fclone is busy
1144 * Returns true if skb is a fast clone, and its clone is not freed.
1145 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1146 * so we also check that this didnt happen.
1148 static inline bool skb_fclone_busy(const struct sock *sk,
1149 const struct sk_buff *skb)
1151 const struct sk_buff_fclones *fclones;
1153 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1155 return skb->fclone == SKB_FCLONE_ORIG &&
1156 refcount_read(&fclones->fclone_ref) > 1 &&
1157 READ_ONCE(fclones->skb2.sk) == sk;
1161 * alloc_skb_fclone - allocate a network buffer from fclone cache
1162 * @size: size to allocate
1163 * @priority: allocation mask
1165 * This function is a convenient wrapper around __alloc_skb().
1167 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1170 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1173 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1174 void skb_headers_offset_update(struct sk_buff *skb, int off);
1175 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1176 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1177 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1178 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1179 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1180 gfp_t gfp_mask, bool fclone);
1181 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1184 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1187 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1188 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1189 unsigned int headroom);
1190 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1191 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1192 int newtailroom, gfp_t priority);
1193 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1194 int offset, int len);
1195 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1196 int offset, int len);
1197 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1198 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1201 * skb_pad - zero pad the tail of an skb
1202 * @skb: buffer to pad
1203 * @pad: space to pad
1205 * Ensure that a buffer is followed by a padding area that is zero
1206 * filled. Used by network drivers which may DMA or transfer data
1207 * beyond the buffer end onto the wire.
1209 * May return error in out of memory cases. The skb is freed on error.
1211 static inline int skb_pad(struct sk_buff *skb, int pad)
1213 return __skb_pad(skb, pad, true);
1215 #define dev_kfree_skb(a) consume_skb(a)
1217 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1218 int offset, size_t size);
1220 struct skb_seq_state {
1224 __u32 stepped_offset;
1225 struct sk_buff *root_skb;
1226 struct sk_buff *cur_skb;
1231 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1232 unsigned int to, struct skb_seq_state *st);
1233 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1234 struct skb_seq_state *st);
1235 void skb_abort_seq_read(struct skb_seq_state *st);
1237 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1238 unsigned int to, struct ts_config *config);
1241 * Packet hash types specify the type of hash in skb_set_hash.
1243 * Hash types refer to the protocol layer addresses which are used to
1244 * construct a packet's hash. The hashes are used to differentiate or identify
1245 * flows of the protocol layer for the hash type. Hash types are either
1246 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1248 * Properties of hashes:
1250 * 1) Two packets in different flows have different hash values
1251 * 2) Two packets in the same flow should have the same hash value
1253 * A hash at a higher layer is considered to be more specific. A driver should
1254 * set the most specific hash possible.
1256 * A driver cannot indicate a more specific hash than the layer at which a hash
1257 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1259 * A driver may indicate a hash level which is less specific than the
1260 * actual layer the hash was computed on. For instance, a hash computed
1261 * at L4 may be considered an L3 hash. This should only be done if the
1262 * driver can't unambiguously determine that the HW computed the hash at
1263 * the higher layer. Note that the "should" in the second property above
1266 enum pkt_hash_types {
1267 PKT_HASH_TYPE_NONE, /* Undefined type */
1268 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1269 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1270 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1273 static inline void skb_clear_hash(struct sk_buff *skb)
1280 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1283 skb_clear_hash(skb);
1287 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1289 skb->l4_hash = is_l4;
1290 skb->sw_hash = is_sw;
1295 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1297 /* Used by drivers to set hash from HW */
1298 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1302 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1304 __skb_set_hash(skb, hash, true, is_l4);
1307 void __skb_get_hash(struct sk_buff *skb);
1308 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1309 u32 skb_get_poff(const struct sk_buff *skb);
1310 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1311 const struct flow_keys_basic *keys, int hlen);
1312 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1313 const void *data, int hlen_proto);
1315 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1316 int thoff, u8 ip_proto)
1318 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1321 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1322 const struct flow_dissector_key *key,
1323 unsigned int key_count);
1325 struct bpf_flow_dissector;
1326 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1327 __be16 proto, int nhoff, int hlen, unsigned int flags);
1329 bool __skb_flow_dissect(const struct net *net,
1330 const struct sk_buff *skb,
1331 struct flow_dissector *flow_dissector,
1332 void *target_container, const void *data,
1333 __be16 proto, int nhoff, int hlen, unsigned int flags);
1335 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1336 struct flow_dissector *flow_dissector,
1337 void *target_container, unsigned int flags)
1339 return __skb_flow_dissect(NULL, skb, flow_dissector,
1340 target_container, NULL, 0, 0, 0, flags);
1343 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1344 struct flow_keys *flow,
1347 memset(flow, 0, sizeof(*flow));
1348 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1349 flow, NULL, 0, 0, 0, flags);
1353 skb_flow_dissect_flow_keys_basic(const struct net *net,
1354 const struct sk_buff *skb,
1355 struct flow_keys_basic *flow,
1356 const void *data, __be16 proto,
1357 int nhoff, int hlen, unsigned int flags)
1359 memset(flow, 0, sizeof(*flow));
1360 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1361 data, proto, nhoff, hlen, flags);
1364 void skb_flow_dissect_meta(const struct sk_buff *skb,
1365 struct flow_dissector *flow_dissector,
1366 void *target_container);
1368 /* Gets a skb connection tracking info, ctinfo map should be a
1369 * map of mapsize to translate enum ip_conntrack_info states
1373 skb_flow_dissect_ct(const struct sk_buff *skb,
1374 struct flow_dissector *flow_dissector,
1375 void *target_container,
1376 u16 *ctinfo_map, size_t mapsize,
1379 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1380 struct flow_dissector *flow_dissector,
1381 void *target_container);
1383 void skb_flow_dissect_hash(const struct sk_buff *skb,
1384 struct flow_dissector *flow_dissector,
1385 void *target_container);
1387 static inline __u32 skb_get_hash(struct sk_buff *skb)
1389 if (!skb->l4_hash && !skb->sw_hash)
1390 __skb_get_hash(skb);
1395 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1397 if (!skb->l4_hash && !skb->sw_hash) {
1398 struct flow_keys keys;
1399 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1401 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1407 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1408 const siphash_key_t *perturb);
1410 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1415 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1417 to->hash = from->hash;
1418 to->sw_hash = from->sw_hash;
1419 to->l4_hash = from->l4_hash;
1422 static inline void skb_copy_decrypted(struct sk_buff *to,
1423 const struct sk_buff *from)
1425 #ifdef CONFIG_TLS_DEVICE
1426 to->decrypted = from->decrypted;
1430 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1431 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1433 return skb->head + skb->end;
1436 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1441 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1446 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1448 return skb->end - skb->head;
1453 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1455 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1457 return &skb_shinfo(skb)->hwtstamps;
1460 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1462 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1464 return is_zcopy ? skb_uarg(skb) : NULL;
1467 static inline void net_zcopy_get(struct ubuf_info *uarg)
1469 refcount_inc(&uarg->refcnt);
1472 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1474 skb_shinfo(skb)->destructor_arg = uarg;
1475 skb_shinfo(skb)->flags |= uarg->flags;
1478 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1481 if (skb && uarg && !skb_zcopy(skb)) {
1482 if (unlikely(have_ref && *have_ref))
1485 net_zcopy_get(uarg);
1486 skb_zcopy_init(skb, uarg);
1490 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1492 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1493 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1496 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1498 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1501 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1503 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1506 static inline void net_zcopy_put(struct ubuf_info *uarg)
1509 uarg->callback(NULL, uarg, true);
1512 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1515 if (uarg->callback == msg_zerocopy_callback)
1516 msg_zerocopy_put_abort(uarg, have_uref);
1518 net_zcopy_put(uarg);
1522 /* Release a reference on a zerocopy structure */
1523 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1525 struct ubuf_info *uarg = skb_zcopy(skb);
1528 if (!skb_zcopy_is_nouarg(skb))
1529 uarg->callback(skb, uarg, zerocopy_success);
1531 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
1535 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1540 /* Iterate through singly-linked GSO fragments of an skb. */
1541 #define skb_list_walk_safe(first, skb, next_skb) \
1542 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1543 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1545 static inline void skb_list_del_init(struct sk_buff *skb)
1547 __list_del_entry(&skb->list);
1548 skb_mark_not_on_list(skb);
1552 * skb_queue_empty - check if a queue is empty
1555 * Returns true if the queue is empty, false otherwise.
1557 static inline int skb_queue_empty(const struct sk_buff_head *list)
1559 return list->next == (const struct sk_buff *) list;
1563 * skb_queue_empty_lockless - check if a queue is empty
1566 * Returns true if the queue is empty, false otherwise.
1567 * This variant can be used in lockless contexts.
1569 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1571 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1576 * skb_queue_is_last - check if skb is the last entry in the queue
1580 * Returns true if @skb is the last buffer on the list.
1582 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1583 const struct sk_buff *skb)
1585 return skb->next == (const struct sk_buff *) list;
1589 * skb_queue_is_first - check if skb is the first entry in the queue
1593 * Returns true if @skb is the first buffer on the list.
1595 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1596 const struct sk_buff *skb)
1598 return skb->prev == (const struct sk_buff *) list;
1602 * skb_queue_next - return the next packet in the queue
1604 * @skb: current buffer
1606 * Return the next packet in @list after @skb. It is only valid to
1607 * call this if skb_queue_is_last() evaluates to false.
1609 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1610 const struct sk_buff *skb)
1612 /* This BUG_ON may seem severe, but if we just return then we
1613 * are going to dereference garbage.
1615 BUG_ON(skb_queue_is_last(list, skb));
1620 * skb_queue_prev - return the prev packet in the queue
1622 * @skb: current buffer
1624 * Return the prev packet in @list before @skb. It is only valid to
1625 * call this if skb_queue_is_first() evaluates to false.
1627 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1628 const struct sk_buff *skb)
1630 /* This BUG_ON may seem severe, but if we just return then we
1631 * are going to dereference garbage.
1633 BUG_ON(skb_queue_is_first(list, skb));
1638 * skb_get - reference buffer
1639 * @skb: buffer to reference
1641 * Makes another reference to a socket buffer and returns a pointer
1644 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1646 refcount_inc(&skb->users);
1651 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1655 * skb_cloned - is the buffer a clone
1656 * @skb: buffer to check
1658 * Returns true if the buffer was generated with skb_clone() and is
1659 * one of multiple shared copies of the buffer. Cloned buffers are
1660 * shared data so must not be written to under normal circumstances.
1662 static inline int skb_cloned(const struct sk_buff *skb)
1664 return skb->cloned &&
1665 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1668 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1670 might_sleep_if(gfpflags_allow_blocking(pri));
1672 if (skb_cloned(skb))
1673 return pskb_expand_head(skb, 0, 0, pri);
1679 * skb_header_cloned - is the header a clone
1680 * @skb: buffer to check
1682 * Returns true if modifying the header part of the buffer requires
1683 * the data to be copied.
1685 static inline int skb_header_cloned(const struct sk_buff *skb)
1692 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1693 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1694 return dataref != 1;
1697 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1699 might_sleep_if(gfpflags_allow_blocking(pri));
1701 if (skb_header_cloned(skb))
1702 return pskb_expand_head(skb, 0, 0, pri);
1708 * __skb_header_release - release reference to header
1709 * @skb: buffer to operate on
1711 static inline void __skb_header_release(struct sk_buff *skb)
1714 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1719 * skb_shared - is the buffer shared
1720 * @skb: buffer to check
1722 * Returns true if more than one person has a reference to this
1725 static inline int skb_shared(const struct sk_buff *skb)
1727 return refcount_read(&skb->users) != 1;
1731 * skb_share_check - check if buffer is shared and if so clone it
1732 * @skb: buffer to check
1733 * @pri: priority for memory allocation
1735 * If the buffer is shared the buffer is cloned and the old copy
1736 * drops a reference. A new clone with a single reference is returned.
1737 * If the buffer is not shared the original buffer is returned. When
1738 * being called from interrupt status or with spinlocks held pri must
1741 * NULL is returned on a memory allocation failure.
1743 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1745 might_sleep_if(gfpflags_allow_blocking(pri));
1746 if (skb_shared(skb)) {
1747 struct sk_buff *nskb = skb_clone(skb, pri);
1759 * Copy shared buffers into a new sk_buff. We effectively do COW on
1760 * packets to handle cases where we have a local reader and forward
1761 * and a couple of other messy ones. The normal one is tcpdumping
1762 * a packet thats being forwarded.
1766 * skb_unshare - make a copy of a shared buffer
1767 * @skb: buffer to check
1768 * @pri: priority for memory allocation
1770 * If the socket buffer is a clone then this function creates a new
1771 * copy of the data, drops a reference count on the old copy and returns
1772 * the new copy with the reference count at 1. If the buffer is not a clone
1773 * the original buffer is returned. When called with a spinlock held or
1774 * from interrupt state @pri must be %GFP_ATOMIC
1776 * %NULL is returned on a memory allocation failure.
1778 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1781 might_sleep_if(gfpflags_allow_blocking(pri));
1782 if (skb_cloned(skb)) {
1783 struct sk_buff *nskb = skb_copy(skb, pri);
1785 /* Free our shared copy */
1796 * skb_peek - peek at the head of an &sk_buff_head
1797 * @list_: list to peek at
1799 * Peek an &sk_buff. Unlike most other operations you _MUST_
1800 * be careful with this one. A peek leaves the buffer on the
1801 * list and someone else may run off with it. You must hold
1802 * the appropriate locks or have a private queue to do this.
1804 * Returns %NULL for an empty list or a pointer to the head element.
1805 * The reference count is not incremented and the reference is therefore
1806 * volatile. Use with caution.
1808 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1810 struct sk_buff *skb = list_->next;
1812 if (skb == (struct sk_buff *)list_)
1818 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1819 * @list_: list to peek at
1821 * Like skb_peek(), but the caller knows that the list is not empty.
1823 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1829 * skb_peek_next - peek skb following the given one from a queue
1830 * @skb: skb to start from
1831 * @list_: list to peek at
1833 * Returns %NULL when the end of the list is met or a pointer to the
1834 * next element. The reference count is not incremented and the
1835 * reference is therefore volatile. Use with caution.
1837 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1838 const struct sk_buff_head *list_)
1840 struct sk_buff *next = skb->next;
1842 if (next == (struct sk_buff *)list_)
1848 * skb_peek_tail - peek at the tail of an &sk_buff_head
1849 * @list_: list to peek at
1851 * Peek an &sk_buff. Unlike most other operations you _MUST_
1852 * be careful with this one. A peek leaves the buffer on the
1853 * list and someone else may run off with it. You must hold
1854 * the appropriate locks or have a private queue to do this.
1856 * Returns %NULL for an empty list or a pointer to the tail element.
1857 * The reference count is not incremented and the reference is therefore
1858 * volatile. Use with caution.
1860 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1862 struct sk_buff *skb = READ_ONCE(list_->prev);
1864 if (skb == (struct sk_buff *)list_)
1871 * skb_queue_len - get queue length
1872 * @list_: list to measure
1874 * Return the length of an &sk_buff queue.
1876 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1882 * skb_queue_len_lockless - get queue length
1883 * @list_: list to measure
1885 * Return the length of an &sk_buff queue.
1886 * This variant can be used in lockless contexts.
1888 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1890 return READ_ONCE(list_->qlen);
1894 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1895 * @list: queue to initialize
1897 * This initializes only the list and queue length aspects of
1898 * an sk_buff_head object. This allows to initialize the list
1899 * aspects of an sk_buff_head without reinitializing things like
1900 * the spinlock. It can also be used for on-stack sk_buff_head
1901 * objects where the spinlock is known to not be used.
1903 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1905 list->prev = list->next = (struct sk_buff *)list;
1910 * This function creates a split out lock class for each invocation;
1911 * this is needed for now since a whole lot of users of the skb-queue
1912 * infrastructure in drivers have different locking usage (in hardirq)
1913 * than the networking core (in softirq only). In the long run either the
1914 * network layer or drivers should need annotation to consolidate the
1915 * main types of usage into 3 classes.
1917 static inline void skb_queue_head_init(struct sk_buff_head *list)
1919 spin_lock_init(&list->lock);
1920 __skb_queue_head_init(list);
1923 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1924 struct lock_class_key *class)
1926 skb_queue_head_init(list);
1927 lockdep_set_class(&list->lock, class);
1931 * Insert an sk_buff on a list.
1933 * The "__skb_xxxx()" functions are the non-atomic ones that
1934 * can only be called with interrupts disabled.
1936 static inline void __skb_insert(struct sk_buff *newsk,
1937 struct sk_buff *prev, struct sk_buff *next,
1938 struct sk_buff_head *list)
1940 /* See skb_queue_empty_lockless() and skb_peek_tail()
1941 * for the opposite READ_ONCE()
1943 WRITE_ONCE(newsk->next, next);
1944 WRITE_ONCE(newsk->prev, prev);
1945 WRITE_ONCE(next->prev, newsk);
1946 WRITE_ONCE(prev->next, newsk);
1947 WRITE_ONCE(list->qlen, list->qlen + 1);
1950 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1951 struct sk_buff *prev,
1952 struct sk_buff *next)
1954 struct sk_buff *first = list->next;
1955 struct sk_buff *last = list->prev;
1957 WRITE_ONCE(first->prev, prev);
1958 WRITE_ONCE(prev->next, first);
1960 WRITE_ONCE(last->next, next);
1961 WRITE_ONCE(next->prev, last);
1965 * skb_queue_splice - join two skb lists, this is designed for stacks
1966 * @list: the new list to add
1967 * @head: the place to add it in the first list
1969 static inline void skb_queue_splice(const struct sk_buff_head *list,
1970 struct sk_buff_head *head)
1972 if (!skb_queue_empty(list)) {
1973 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1974 head->qlen += list->qlen;
1979 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1980 * @list: the new list to add
1981 * @head: the place to add it in the first list
1983 * The list at @list is reinitialised
1985 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1986 struct sk_buff_head *head)
1988 if (!skb_queue_empty(list)) {
1989 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1990 head->qlen += list->qlen;
1991 __skb_queue_head_init(list);
1996 * skb_queue_splice_tail - join two skb lists, each list being a queue
1997 * @list: the new list to add
1998 * @head: the place to add it in the first list
2000 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2001 struct sk_buff_head *head)
2003 if (!skb_queue_empty(list)) {
2004 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2005 head->qlen += list->qlen;
2010 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2011 * @list: the new list to add
2012 * @head: the place to add it in the first list
2014 * Each of the lists is a queue.
2015 * The list at @list is reinitialised
2017 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2018 struct sk_buff_head *head)
2020 if (!skb_queue_empty(list)) {
2021 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2022 head->qlen += list->qlen;
2023 __skb_queue_head_init(list);
2028 * __skb_queue_after - queue a buffer at the list head
2029 * @list: list to use
2030 * @prev: place after this buffer
2031 * @newsk: buffer to queue
2033 * Queue a buffer int the middle of a list. This function takes no locks
2034 * and you must therefore hold required locks before calling it.
2036 * A buffer cannot be placed on two lists at the same time.
2038 static inline void __skb_queue_after(struct sk_buff_head *list,
2039 struct sk_buff *prev,
2040 struct sk_buff *newsk)
2042 __skb_insert(newsk, prev, prev->next, list);
2045 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2046 struct sk_buff_head *list);
2048 static inline void __skb_queue_before(struct sk_buff_head *list,
2049 struct sk_buff *next,
2050 struct sk_buff *newsk)
2052 __skb_insert(newsk, next->prev, next, list);
2056 * __skb_queue_head - queue a buffer at the list head
2057 * @list: list to use
2058 * @newsk: buffer to queue
2060 * Queue a buffer at the start of a list. This function takes no locks
2061 * and you must therefore hold required locks before calling it.
2063 * A buffer cannot be placed on two lists at the same time.
2065 static inline void __skb_queue_head(struct sk_buff_head *list,
2066 struct sk_buff *newsk)
2068 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2070 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2073 * __skb_queue_tail - queue a buffer at the list tail
2074 * @list: list to use
2075 * @newsk: buffer to queue
2077 * Queue a buffer at the end of a list. This function takes no locks
2078 * and you must therefore hold required locks before calling it.
2080 * A buffer cannot be placed on two lists at the same time.
2082 static inline void __skb_queue_tail(struct sk_buff_head *list,
2083 struct sk_buff *newsk)
2085 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2087 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2090 * remove sk_buff from list. _Must_ be called atomically, and with
2093 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2094 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2096 struct sk_buff *next, *prev;
2098 WRITE_ONCE(list->qlen, list->qlen - 1);
2101 skb->next = skb->prev = NULL;
2102 WRITE_ONCE(next->prev, prev);
2103 WRITE_ONCE(prev->next, next);
2107 * __skb_dequeue - remove from the head of the queue
2108 * @list: list to dequeue from
2110 * Remove the head of the list. This function does not take any locks
2111 * so must be used with appropriate locks held only. The head item is
2112 * returned or %NULL if the list is empty.
2114 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2116 struct sk_buff *skb = skb_peek(list);
2118 __skb_unlink(skb, list);
2121 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2124 * __skb_dequeue_tail - remove from the tail of the queue
2125 * @list: list to dequeue from
2127 * Remove the tail of the list. This function does not take any locks
2128 * so must be used with appropriate locks held only. The tail item is
2129 * returned or %NULL if the list is empty.
2131 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2133 struct sk_buff *skb = skb_peek_tail(list);
2135 __skb_unlink(skb, list);
2138 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2141 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2143 return skb->data_len;
2146 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2148 return skb->len - skb->data_len;
2151 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2153 unsigned int i, len = 0;
2155 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2156 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2160 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2162 return skb_headlen(skb) + __skb_pagelen(skb);
2166 * __skb_fill_page_desc - initialise a paged fragment in an skb
2167 * @skb: buffer containing fragment to be initialised
2168 * @i: paged fragment index to initialise
2169 * @page: the page to use for this fragment
2170 * @off: the offset to the data with @page
2171 * @size: the length of the data
2173 * Initialises the @i'th fragment of @skb to point to &size bytes at
2174 * offset @off within @page.
2176 * Does not take any additional reference on the fragment.
2178 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2179 struct page *page, int off, int size)
2181 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2184 * Propagate page pfmemalloc to the skb if we can. The problem is
2185 * that not all callers have unique ownership of the page but rely
2186 * on page_is_pfmemalloc doing the right thing(tm).
2188 frag->bv_page = page;
2189 frag->bv_offset = off;
2190 skb_frag_size_set(frag, size);
2192 page = compound_head(page);
2193 if (page_is_pfmemalloc(page))
2194 skb->pfmemalloc = true;
2198 * skb_fill_page_desc - initialise a paged fragment in an skb
2199 * @skb: buffer containing fragment to be initialised
2200 * @i: paged fragment index to initialise
2201 * @page: the page to use for this fragment
2202 * @off: the offset to the data with @page
2203 * @size: the length of the data
2205 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2206 * @skb to point to @size bytes at offset @off within @page. In
2207 * addition updates @skb such that @i is the last fragment.
2209 * Does not take any additional reference on the fragment.
2211 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2212 struct page *page, int off, int size)
2214 __skb_fill_page_desc(skb, i, page, off, size);
2215 skb_shinfo(skb)->nr_frags = i + 1;
2218 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2219 int size, unsigned int truesize);
2221 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2222 unsigned int truesize);
2224 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2226 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2227 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2229 return skb->head + skb->tail;
2232 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2234 skb->tail = skb->data - skb->head;
2237 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2239 skb_reset_tail_pointer(skb);
2240 skb->tail += offset;
2243 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2244 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2249 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2251 skb->tail = skb->data;
2254 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2256 skb->tail = skb->data + offset;
2259 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2262 * Add data to an sk_buff
2264 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2265 void *skb_put(struct sk_buff *skb, unsigned int len);
2266 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2268 void *tmp = skb_tail_pointer(skb);
2269 SKB_LINEAR_ASSERT(skb);
2275 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2277 void *tmp = __skb_put(skb, len);
2279 memset(tmp, 0, len);
2283 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2286 void *tmp = __skb_put(skb, len);
2288 memcpy(tmp, data, len);
2292 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2294 *(u8 *)__skb_put(skb, 1) = val;
2297 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2299 void *tmp = skb_put(skb, len);
2301 memset(tmp, 0, len);
2306 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2309 void *tmp = skb_put(skb, len);
2311 memcpy(tmp, data, len);
2316 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2318 *(u8 *)skb_put(skb, 1) = val;
2321 void *skb_push(struct sk_buff *skb, unsigned int len);
2322 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2329 void *skb_pull(struct sk_buff *skb, unsigned int len);
2330 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2333 BUG_ON(skb->len < skb->data_len);
2334 return skb->data += len;
2337 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2339 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2342 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2344 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2346 if (len > skb_headlen(skb) &&
2347 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2350 return skb->data += len;
2353 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2355 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2358 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2360 if (likely(len <= skb_headlen(skb)))
2362 if (unlikely(len > skb->len))
2364 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2367 void skb_condense(struct sk_buff *skb);
2370 * skb_headroom - bytes at buffer head
2371 * @skb: buffer to check
2373 * Return the number of bytes of free space at the head of an &sk_buff.
2375 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2377 return skb->data - skb->head;
2381 * skb_tailroom - bytes at buffer end
2382 * @skb: buffer to check
2384 * Return the number of bytes of free space at the tail of an sk_buff
2386 static inline int skb_tailroom(const struct sk_buff *skb)
2388 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2392 * skb_availroom - bytes at buffer end
2393 * @skb: buffer to check
2395 * Return the number of bytes of free space at the tail of an sk_buff
2396 * allocated by sk_stream_alloc()
2398 static inline int skb_availroom(const struct sk_buff *skb)
2400 if (skb_is_nonlinear(skb))
2403 return skb->end - skb->tail - skb->reserved_tailroom;
2407 * skb_reserve - adjust headroom
2408 * @skb: buffer to alter
2409 * @len: bytes to move
2411 * Increase the headroom of an empty &sk_buff by reducing the tail
2412 * room. This is only allowed for an empty buffer.
2414 static inline void skb_reserve(struct sk_buff *skb, int len)
2421 * skb_tailroom_reserve - adjust reserved_tailroom
2422 * @skb: buffer to alter
2423 * @mtu: maximum amount of headlen permitted
2424 * @needed_tailroom: minimum amount of reserved_tailroom
2426 * Set reserved_tailroom so that headlen can be as large as possible but
2427 * not larger than mtu and tailroom cannot be smaller than
2429 * The required headroom should already have been reserved before using
2432 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2433 unsigned int needed_tailroom)
2435 SKB_LINEAR_ASSERT(skb);
2436 if (mtu < skb_tailroom(skb) - needed_tailroom)
2437 /* use at most mtu */
2438 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2440 /* use up to all available space */
2441 skb->reserved_tailroom = needed_tailroom;
2444 #define ENCAP_TYPE_ETHER 0
2445 #define ENCAP_TYPE_IPPROTO 1
2447 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2450 skb->inner_protocol = protocol;
2451 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2454 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2457 skb->inner_ipproto = ipproto;
2458 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2461 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2463 skb->inner_mac_header = skb->mac_header;
2464 skb->inner_network_header = skb->network_header;
2465 skb->inner_transport_header = skb->transport_header;
2468 static inline void skb_reset_mac_len(struct sk_buff *skb)
2470 skb->mac_len = skb->network_header - skb->mac_header;
2473 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2476 return skb->head + skb->inner_transport_header;
2479 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2481 return skb_inner_transport_header(skb) - skb->data;
2484 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2486 skb->inner_transport_header = skb->data - skb->head;
2489 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2492 skb_reset_inner_transport_header(skb);
2493 skb->inner_transport_header += offset;
2496 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2498 return skb->head + skb->inner_network_header;
2501 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2503 skb->inner_network_header = skb->data - skb->head;
2506 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2509 skb_reset_inner_network_header(skb);
2510 skb->inner_network_header += offset;
2513 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2515 return skb->head + skb->inner_mac_header;
2518 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2520 skb->inner_mac_header = skb->data - skb->head;
2523 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2526 skb_reset_inner_mac_header(skb);
2527 skb->inner_mac_header += offset;
2529 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2531 return skb->transport_header != (typeof(skb->transport_header))~0U;
2534 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2536 return skb->head + skb->transport_header;
2539 static inline void skb_reset_transport_header(struct sk_buff *skb)
2541 skb->transport_header = skb->data - skb->head;
2544 static inline void skb_set_transport_header(struct sk_buff *skb,
2547 skb_reset_transport_header(skb);
2548 skb->transport_header += offset;
2551 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2553 return skb->head + skb->network_header;
2556 static inline void skb_reset_network_header(struct sk_buff *skb)
2558 skb->network_header = skb->data - skb->head;
2561 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2563 skb_reset_network_header(skb);
2564 skb->network_header += offset;
2567 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2569 return skb->head + skb->mac_header;
2572 static inline int skb_mac_offset(const struct sk_buff *skb)
2574 return skb_mac_header(skb) - skb->data;
2577 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2579 return skb->network_header - skb->mac_header;
2582 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2584 return skb->mac_header != (typeof(skb->mac_header))~0U;
2587 static inline void skb_unset_mac_header(struct sk_buff *skb)
2589 skb->mac_header = (typeof(skb->mac_header))~0U;
2592 static inline void skb_reset_mac_header(struct sk_buff *skb)
2594 skb->mac_header = skb->data - skb->head;
2597 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2599 skb_reset_mac_header(skb);
2600 skb->mac_header += offset;
2603 static inline void skb_pop_mac_header(struct sk_buff *skb)
2605 skb->mac_header = skb->network_header;
2608 static inline void skb_probe_transport_header(struct sk_buff *skb)
2610 struct flow_keys_basic keys;
2612 if (skb_transport_header_was_set(skb))
2615 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2617 skb_set_transport_header(skb, keys.control.thoff);
2620 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2622 if (skb_mac_header_was_set(skb)) {
2623 const unsigned char *old_mac = skb_mac_header(skb);
2625 skb_set_mac_header(skb, -skb->mac_len);
2626 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2630 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2632 return skb->csum_start - skb_headroom(skb);
2635 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2637 return skb->head + skb->csum_start;
2640 static inline int skb_transport_offset(const struct sk_buff *skb)
2642 return skb_transport_header(skb) - skb->data;
2645 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2647 return skb->transport_header - skb->network_header;
2650 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2652 return skb->inner_transport_header - skb->inner_network_header;
2655 static inline int skb_network_offset(const struct sk_buff *skb)
2657 return skb_network_header(skb) - skb->data;
2660 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2662 return skb_inner_network_header(skb) - skb->data;
2665 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2667 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2671 * CPUs often take a performance hit when accessing unaligned memory
2672 * locations. The actual performance hit varies, it can be small if the
2673 * hardware handles it or large if we have to take an exception and fix it
2676 * Since an ethernet header is 14 bytes network drivers often end up with
2677 * the IP header at an unaligned offset. The IP header can be aligned by
2678 * shifting the start of the packet by 2 bytes. Drivers should do this
2681 * skb_reserve(skb, NET_IP_ALIGN);
2683 * The downside to this alignment of the IP header is that the DMA is now
2684 * unaligned. On some architectures the cost of an unaligned DMA is high
2685 * and this cost outweighs the gains made by aligning the IP header.
2687 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2690 #ifndef NET_IP_ALIGN
2691 #define NET_IP_ALIGN 2
2695 * The networking layer reserves some headroom in skb data (via
2696 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2697 * the header has to grow. In the default case, if the header has to grow
2698 * 32 bytes or less we avoid the reallocation.
2700 * Unfortunately this headroom changes the DMA alignment of the resulting
2701 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2702 * on some architectures. An architecture can override this value,
2703 * perhaps setting it to a cacheline in size (since that will maintain
2704 * cacheline alignment of the DMA). It must be a power of 2.
2706 * Various parts of the networking layer expect at least 32 bytes of
2707 * headroom, you should not reduce this.
2709 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2710 * to reduce average number of cache lines per packet.
2711 * get_rps_cpu() for example only access one 64 bytes aligned block :
2712 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2715 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2718 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2720 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2722 if (WARN_ON(skb_is_nonlinear(skb)))
2725 skb_set_tail_pointer(skb, len);
2728 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2730 __skb_set_length(skb, len);
2733 void skb_trim(struct sk_buff *skb, unsigned int len);
2735 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2738 return ___pskb_trim(skb, len);
2739 __skb_trim(skb, len);
2743 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2745 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2749 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2750 * @skb: buffer to alter
2753 * This is identical to pskb_trim except that the caller knows that
2754 * the skb is not cloned so we should never get an error due to out-
2757 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2759 int err = pskb_trim(skb, len);
2763 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2765 unsigned int diff = len - skb->len;
2767 if (skb_tailroom(skb) < diff) {
2768 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2773 __skb_set_length(skb, len);
2778 * skb_orphan - orphan a buffer
2779 * @skb: buffer to orphan
2781 * If a buffer currently has an owner then we call the owner's
2782 * destructor function and make the @skb unowned. The buffer continues
2783 * to exist but is no longer charged to its former owner.
2785 static inline void skb_orphan(struct sk_buff *skb)
2787 if (skb->destructor) {
2788 skb->destructor(skb);
2789 skb->destructor = NULL;
2797 * skb_orphan_frags - orphan the frags contained in a buffer
2798 * @skb: buffer to orphan frags from
2799 * @gfp_mask: allocation mask for replacement pages
2801 * For each frag in the SKB which needs a destructor (i.e. has an
2802 * owner) create a copy of that frag and release the original
2803 * page by calling the destructor.
2805 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2807 if (likely(!skb_zcopy(skb)))
2809 if (!skb_zcopy_is_nouarg(skb) &&
2810 skb_uarg(skb)->callback == msg_zerocopy_callback)
2812 return skb_copy_ubufs(skb, gfp_mask);
2815 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2816 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2818 if (likely(!skb_zcopy(skb)))
2820 return skb_copy_ubufs(skb, gfp_mask);
2824 * __skb_queue_purge - empty a list
2825 * @list: list to empty
2827 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2828 * the list and one reference dropped. This function does not take the
2829 * list lock and the caller must hold the relevant locks to use it.
2831 static inline void __skb_queue_purge(struct sk_buff_head *list)
2833 struct sk_buff *skb;
2834 while ((skb = __skb_dequeue(list)) != NULL)
2837 void skb_queue_purge(struct sk_buff_head *list);
2839 unsigned int skb_rbtree_purge(struct rb_root *root);
2841 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2844 * netdev_alloc_frag - allocate a page fragment
2845 * @fragsz: fragment size
2847 * Allocates a frag from a page for receive buffer.
2848 * Uses GFP_ATOMIC allocations.
2850 static inline void *netdev_alloc_frag(unsigned int fragsz)
2852 return __netdev_alloc_frag_align(fragsz, ~0u);
2855 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2858 WARN_ON_ONCE(!is_power_of_2(align));
2859 return __netdev_alloc_frag_align(fragsz, -align);
2862 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2866 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2867 * @dev: network device to receive on
2868 * @length: length to allocate
2870 * Allocate a new &sk_buff and assign it a usage count of one. The
2871 * buffer has unspecified headroom built in. Users should allocate
2872 * the headroom they think they need without accounting for the
2873 * built in space. The built in space is used for optimisations.
2875 * %NULL is returned if there is no free memory. Although this function
2876 * allocates memory it can be called from an interrupt.
2878 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2879 unsigned int length)
2881 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2884 /* legacy helper around __netdev_alloc_skb() */
2885 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2888 return __netdev_alloc_skb(NULL, length, gfp_mask);
2891 /* legacy helper around netdev_alloc_skb() */
2892 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2894 return netdev_alloc_skb(NULL, length);
2898 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2899 unsigned int length, gfp_t gfp)
2901 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2903 if (NET_IP_ALIGN && skb)
2904 skb_reserve(skb, NET_IP_ALIGN);
2908 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2909 unsigned int length)
2911 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2914 static inline void skb_free_frag(void *addr)
2916 page_frag_free(addr);
2919 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2921 static inline void *napi_alloc_frag(unsigned int fragsz)
2923 return __napi_alloc_frag_align(fragsz, ~0u);
2926 static inline void *napi_alloc_frag_align(unsigned int fragsz,
2929 WARN_ON_ONCE(!is_power_of_2(align));
2930 return __napi_alloc_frag_align(fragsz, -align);
2933 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2934 unsigned int length, gfp_t gfp_mask);
2935 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2936 unsigned int length)
2938 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2940 void napi_consume_skb(struct sk_buff *skb, int budget);
2942 void napi_skb_free_stolen_head(struct sk_buff *skb);
2943 void __kfree_skb_defer(struct sk_buff *skb);
2946 * __dev_alloc_pages - allocate page for network Rx
2947 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2948 * @order: size of the allocation
2950 * Allocate a new page.
2952 * %NULL is returned if there is no free memory.
2954 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2957 /* This piece of code contains several assumptions.
2958 * 1. This is for device Rx, therefor a cold page is preferred.
2959 * 2. The expectation is the user wants a compound page.
2960 * 3. If requesting a order 0 page it will not be compound
2961 * due to the check to see if order has a value in prep_new_page
2962 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2963 * code in gfp_to_alloc_flags that should be enforcing this.
2965 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2967 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2970 static inline struct page *dev_alloc_pages(unsigned int order)
2972 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2976 * __dev_alloc_page - allocate a page for network Rx
2977 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2979 * Allocate a new page.
2981 * %NULL is returned if there is no free memory.
2983 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2985 return __dev_alloc_pages(gfp_mask, 0);
2988 static inline struct page *dev_alloc_page(void)
2990 return dev_alloc_pages(0);
2994 * dev_page_is_reusable - check whether a page can be reused for network Rx
2995 * @page: the page to test
2997 * A page shouldn't be considered for reusing/recycling if it was allocated
2998 * under memory pressure or at a distant memory node.
3000 * Returns false if this page should be returned to page allocator, true
3003 static inline bool dev_page_is_reusable(const struct page *page)
3005 return likely(page_to_nid(page) == numa_mem_id() &&
3006 !page_is_pfmemalloc(page));
3010 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3011 * @page: The page that was allocated from skb_alloc_page
3012 * @skb: The skb that may need pfmemalloc set
3014 static inline void skb_propagate_pfmemalloc(const struct page *page,
3015 struct sk_buff *skb)
3017 if (page_is_pfmemalloc(page))
3018 skb->pfmemalloc = true;
3022 * skb_frag_off() - Returns the offset of a skb fragment
3023 * @frag: the paged fragment
3025 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3027 return frag->bv_offset;
3031 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3032 * @frag: skb fragment
3033 * @delta: value to add
3035 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3037 frag->bv_offset += delta;
3041 * skb_frag_off_set() - Sets the offset of a skb fragment
3042 * @frag: skb fragment
3043 * @offset: offset of fragment
3045 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3047 frag->bv_offset = offset;
3051 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3052 * @fragto: skb fragment where offset is set
3053 * @fragfrom: skb fragment offset is copied from
3055 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3056 const skb_frag_t *fragfrom)
3058 fragto->bv_offset = fragfrom->bv_offset;
3062 * skb_frag_page - retrieve the page referred to by a paged fragment
3063 * @frag: the paged fragment
3065 * Returns the &struct page associated with @frag.
3067 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3069 return frag->bv_page;
3073 * __skb_frag_ref - take an addition reference on a paged fragment.
3074 * @frag: the paged fragment
3076 * Takes an additional reference on the paged fragment @frag.
3078 static inline void __skb_frag_ref(skb_frag_t *frag)
3080 get_page(skb_frag_page(frag));
3084 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3086 * @f: the fragment offset.
3088 * Takes an additional reference on the @f'th paged fragment of @skb.
3090 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3092 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3096 * __skb_frag_unref - release a reference on a paged fragment.
3097 * @frag: the paged fragment
3098 * @recycle: recycle the page if allocated via page_pool
3100 * Releases a reference on the paged fragment @frag
3101 * or recycles the page via the page_pool API.
3103 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3105 struct page *page = skb_frag_page(frag);
3107 #ifdef CONFIG_PAGE_POOL
3108 if (recycle && page_pool_return_skb_page(page))
3115 * skb_frag_unref - release a reference on a paged fragment of an skb.
3117 * @f: the fragment offset
3119 * Releases a reference on the @f'th paged fragment of @skb.
3121 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3123 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3127 * skb_frag_address - gets the address of the data contained in a paged fragment
3128 * @frag: the paged fragment buffer
3130 * Returns the address of the data within @frag. The page must already
3133 static inline void *skb_frag_address(const skb_frag_t *frag)
3135 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3139 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3140 * @frag: the paged fragment buffer
3142 * Returns the address of the data within @frag. Checks that the page
3143 * is mapped and returns %NULL otherwise.
3145 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3147 void *ptr = page_address(skb_frag_page(frag));
3151 return ptr + skb_frag_off(frag);
3155 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3156 * @fragto: skb fragment where page is set
3157 * @fragfrom: skb fragment page is copied from
3159 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3160 const skb_frag_t *fragfrom)
3162 fragto->bv_page = fragfrom->bv_page;
3166 * __skb_frag_set_page - sets the page contained in a paged fragment
3167 * @frag: the paged fragment
3168 * @page: the page to set
3170 * Sets the fragment @frag to contain @page.
3172 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3174 frag->bv_page = page;
3178 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3180 * @f: the fragment offset
3181 * @page: the page to set
3183 * Sets the @f'th fragment of @skb to contain @page.
3185 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3188 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3191 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3194 * skb_frag_dma_map - maps a paged fragment via the DMA API
3195 * @dev: the device to map the fragment to
3196 * @frag: the paged fragment to map
3197 * @offset: the offset within the fragment (starting at the
3198 * fragment's own offset)
3199 * @size: the number of bytes to map
3200 * @dir: the direction of the mapping (``PCI_DMA_*``)
3202 * Maps the page associated with @frag to @device.
3204 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3205 const skb_frag_t *frag,
3206 size_t offset, size_t size,
3207 enum dma_data_direction dir)
3209 return dma_map_page(dev, skb_frag_page(frag),
3210 skb_frag_off(frag) + offset, size, dir);
3213 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3216 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3220 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3223 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3228 * skb_clone_writable - is the header of a clone writable
3229 * @skb: buffer to check
3230 * @len: length up to which to write
3232 * Returns true if modifying the header part of the cloned buffer
3233 * does not requires the data to be copied.
3235 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3237 return !skb_header_cloned(skb) &&
3238 skb_headroom(skb) + len <= skb->hdr_len;
3241 static inline int skb_try_make_writable(struct sk_buff *skb,
3242 unsigned int write_len)
3244 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3245 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3248 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3253 if (headroom > skb_headroom(skb))
3254 delta = headroom - skb_headroom(skb);
3256 if (delta || cloned)
3257 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3263 * skb_cow - copy header of skb when it is required
3264 * @skb: buffer to cow
3265 * @headroom: needed headroom
3267 * If the skb passed lacks sufficient headroom or its data part
3268 * is shared, data is reallocated. If reallocation fails, an error
3269 * is returned and original skb is not changed.
3271 * The result is skb with writable area skb->head...skb->tail
3272 * and at least @headroom of space at head.
3274 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3276 return __skb_cow(skb, headroom, skb_cloned(skb));
3280 * skb_cow_head - skb_cow but only making the head writable
3281 * @skb: buffer to cow
3282 * @headroom: needed headroom
3284 * This function is identical to skb_cow except that we replace the
3285 * skb_cloned check by skb_header_cloned. It should be used when
3286 * you only need to push on some header and do not need to modify
3289 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3291 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3295 * skb_padto - pad an skbuff up to a minimal size
3296 * @skb: buffer to pad
3297 * @len: minimal length
3299 * Pads up a buffer to ensure the trailing bytes exist and are
3300 * blanked. If the buffer already contains sufficient data it
3301 * is untouched. Otherwise it is extended. Returns zero on
3302 * success. The skb is freed on error.
3304 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3306 unsigned int size = skb->len;
3307 if (likely(size >= len))
3309 return skb_pad(skb, len - size);
3313 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3314 * @skb: buffer to pad
3315 * @len: minimal length
3316 * @free_on_error: free buffer on error
3318 * Pads up a buffer to ensure the trailing bytes exist and are
3319 * blanked. If the buffer already contains sufficient data it
3320 * is untouched. Otherwise it is extended. Returns zero on
3321 * success. The skb is freed on error if @free_on_error is true.
3323 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3327 unsigned int size = skb->len;
3329 if (unlikely(size < len)) {
3331 if (__skb_pad(skb, len, free_on_error))
3333 __skb_put(skb, len);
3339 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3340 * @skb: buffer to pad
3341 * @len: minimal length
3343 * Pads up a buffer to ensure the trailing bytes exist and are
3344 * blanked. If the buffer already contains sufficient data it
3345 * is untouched. Otherwise it is extended. Returns zero on
3346 * success. The skb is freed on error.
3348 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3350 return __skb_put_padto(skb, len, true);
3353 static inline int skb_add_data(struct sk_buff *skb,
3354 struct iov_iter *from, int copy)
3356 const int off = skb->len;
3358 if (skb->ip_summed == CHECKSUM_NONE) {
3360 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3362 skb->csum = csum_block_add(skb->csum, csum, off);
3365 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3368 __skb_trim(skb, off);
3372 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3373 const struct page *page, int off)
3378 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3380 return page == skb_frag_page(frag) &&
3381 off == skb_frag_off(frag) + skb_frag_size(frag);
3386 static inline int __skb_linearize(struct sk_buff *skb)
3388 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3392 * skb_linearize - convert paged skb to linear one
3393 * @skb: buffer to linarize
3395 * If there is no free memory -ENOMEM is returned, otherwise zero
3396 * is returned and the old skb data released.
3398 static inline int skb_linearize(struct sk_buff *skb)
3400 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3404 * skb_has_shared_frag - can any frag be overwritten
3405 * @skb: buffer to test
3407 * Return true if the skb has at least one frag that might be modified
3408 * by an external entity (as in vmsplice()/sendfile())
3410 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3412 return skb_is_nonlinear(skb) &&
3413 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3417 * skb_linearize_cow - make sure skb is linear and writable
3418 * @skb: buffer to process
3420 * If there is no free memory -ENOMEM is returned, otherwise zero
3421 * is returned and the old skb data released.
3423 static inline int skb_linearize_cow(struct sk_buff *skb)
3425 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3426 __skb_linearize(skb) : 0;
3429 static __always_inline void
3430 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3433 if (skb->ip_summed == CHECKSUM_COMPLETE)
3434 skb->csum = csum_block_sub(skb->csum,
3435 csum_partial(start, len, 0), off);
3436 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3437 skb_checksum_start_offset(skb) < 0)
3438 skb->ip_summed = CHECKSUM_NONE;
3442 * skb_postpull_rcsum - update checksum for received skb after pull
3443 * @skb: buffer to update
3444 * @start: start of data before pull
3445 * @len: length of data pulled
3447 * After doing a pull on a received packet, you need to call this to
3448 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3449 * CHECKSUM_NONE so that it can be recomputed from scratch.
3451 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3452 const void *start, unsigned int len)
3454 __skb_postpull_rcsum(skb, start, len, 0);
3457 static __always_inline void
3458 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3461 if (skb->ip_summed == CHECKSUM_COMPLETE)
3462 skb->csum = csum_block_add(skb->csum,
3463 csum_partial(start, len, 0), off);
3467 * skb_postpush_rcsum - update checksum for received skb after push
3468 * @skb: buffer to update
3469 * @start: start of data after push
3470 * @len: length of data pushed
3472 * After doing a push on a received packet, you need to call this to
3473 * update the CHECKSUM_COMPLETE checksum.
3475 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3476 const void *start, unsigned int len)
3478 __skb_postpush_rcsum(skb, start, len, 0);
3481 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3484 * skb_push_rcsum - push skb and update receive checksum
3485 * @skb: buffer to update
3486 * @len: length of data pulled
3488 * This function performs an skb_push on the packet and updates
3489 * the CHECKSUM_COMPLETE checksum. It should be used on
3490 * receive path processing instead of skb_push unless you know
3491 * that the checksum difference is zero (e.g., a valid IP header)
3492 * or you are setting ip_summed to CHECKSUM_NONE.
3494 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3497 skb_postpush_rcsum(skb, skb->data, len);
3501 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3503 * pskb_trim_rcsum - trim received skb and update checksum
3504 * @skb: buffer to trim
3507 * This is exactly the same as pskb_trim except that it ensures the
3508 * checksum of received packets are still valid after the operation.
3509 * It can change skb pointers.
3512 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3514 if (likely(len >= skb->len))
3516 return pskb_trim_rcsum_slow(skb, len);
3519 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3521 if (skb->ip_summed == CHECKSUM_COMPLETE)
3522 skb->ip_summed = CHECKSUM_NONE;
3523 __skb_trim(skb, len);
3527 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3529 if (skb->ip_summed == CHECKSUM_COMPLETE)
3530 skb->ip_summed = CHECKSUM_NONE;
3531 return __skb_grow(skb, len);
3534 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3535 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3536 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3537 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3538 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3540 #define skb_queue_walk(queue, skb) \
3541 for (skb = (queue)->next; \
3542 skb != (struct sk_buff *)(queue); \
3545 #define skb_queue_walk_safe(queue, skb, tmp) \
3546 for (skb = (queue)->next, tmp = skb->next; \
3547 skb != (struct sk_buff *)(queue); \
3548 skb = tmp, tmp = skb->next)
3550 #define skb_queue_walk_from(queue, skb) \
3551 for (; skb != (struct sk_buff *)(queue); \
3554 #define skb_rbtree_walk(skb, root) \
3555 for (skb = skb_rb_first(root); skb != NULL; \
3556 skb = skb_rb_next(skb))
3558 #define skb_rbtree_walk_from(skb) \
3559 for (; skb != NULL; \
3560 skb = skb_rb_next(skb))
3562 #define skb_rbtree_walk_from_safe(skb, tmp) \
3563 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3566 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3567 for (tmp = skb->next; \
3568 skb != (struct sk_buff *)(queue); \
3569 skb = tmp, tmp = skb->next)
3571 #define skb_queue_reverse_walk(queue, skb) \
3572 for (skb = (queue)->prev; \
3573 skb != (struct sk_buff *)(queue); \
3576 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3577 for (skb = (queue)->prev, tmp = skb->prev; \
3578 skb != (struct sk_buff *)(queue); \
3579 skb = tmp, tmp = skb->prev)
3581 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3582 for (tmp = skb->prev; \
3583 skb != (struct sk_buff *)(queue); \
3584 skb = tmp, tmp = skb->prev)
3586 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3588 return skb_shinfo(skb)->frag_list != NULL;
3591 static inline void skb_frag_list_init(struct sk_buff *skb)
3593 skb_shinfo(skb)->frag_list = NULL;
3596 #define skb_walk_frags(skb, iter) \
3597 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3600 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3601 int *err, long *timeo_p,
3602 const struct sk_buff *skb);
3603 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3604 struct sk_buff_head *queue,
3607 struct sk_buff **last);
3608 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3609 struct sk_buff_head *queue,
3610 unsigned int flags, int *off, int *err,
3611 struct sk_buff **last);
3612 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3613 struct sk_buff_head *sk_queue,
3614 unsigned int flags, int *off, int *err);
3615 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3617 __poll_t datagram_poll(struct file *file, struct socket *sock,
3618 struct poll_table_struct *wait);
3619 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3620 struct iov_iter *to, int size);
3621 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3622 struct msghdr *msg, int size)
3624 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3626 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3627 struct msghdr *msg);
3628 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3629 struct iov_iter *to, int len,
3630 struct ahash_request *hash);
3631 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3632 struct iov_iter *from, int len);
3633 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3634 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3635 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3636 static inline void skb_free_datagram_locked(struct sock *sk,
3637 struct sk_buff *skb)
3639 __skb_free_datagram_locked(sk, skb, 0);
3641 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3642 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3643 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3644 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3646 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3647 struct pipe_inode_info *pipe, unsigned int len,
3648 unsigned int flags);
3649 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3651 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3652 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3653 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3654 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3656 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3657 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3658 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3659 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3660 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3661 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3662 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3663 unsigned int offset);
3664 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3665 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3666 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3667 int skb_vlan_pop(struct sk_buff *skb);
3668 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3669 int skb_eth_pop(struct sk_buff *skb);
3670 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3671 const unsigned char *src);
3672 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3673 int mac_len, bool ethernet);
3674 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3676 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3677 int skb_mpls_dec_ttl(struct sk_buff *skb);
3678 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3681 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3683 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3686 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3688 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3691 struct skb_checksum_ops {
3692 __wsum (*update)(const void *mem, int len, __wsum wsum);
3693 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3696 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3698 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3699 __wsum csum, const struct skb_checksum_ops *ops);
3700 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3703 static inline void * __must_check
3704 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3705 const void *data, int hlen, void *buffer)
3707 if (likely(hlen - offset >= len))
3708 return (void *)data + offset;
3710 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3716 static inline void * __must_check
3717 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3719 return __skb_header_pointer(skb, offset, len, skb->data,
3720 skb_headlen(skb), buffer);
3724 * skb_needs_linearize - check if we need to linearize a given skb
3725 * depending on the given device features.
3726 * @skb: socket buffer to check
3727 * @features: net device features
3729 * Returns true if either:
3730 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3731 * 2. skb is fragmented and the device does not support SG.
3733 static inline bool skb_needs_linearize(struct sk_buff *skb,
3734 netdev_features_t features)
3736 return skb_is_nonlinear(skb) &&
3737 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3738 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3741 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3743 const unsigned int len)
3745 memcpy(to, skb->data, len);
3748 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3749 const int offset, void *to,
3750 const unsigned int len)
3752 memcpy(to, skb->data + offset, len);
3755 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3757 const unsigned int len)
3759 memcpy(skb->data, from, len);
3762 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3765 const unsigned int len)
3767 memcpy(skb->data + offset, from, len);
3770 void skb_init(void);
3772 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3778 * skb_get_timestamp - get timestamp from a skb
3779 * @skb: skb to get stamp from
3780 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3782 * Timestamps are stored in the skb as offsets to a base timestamp.
3783 * This function converts the offset back to a struct timeval and stores
3786 static inline void skb_get_timestamp(const struct sk_buff *skb,
3787 struct __kernel_old_timeval *stamp)
3789 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3792 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3793 struct __kernel_sock_timeval *stamp)
3795 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3797 stamp->tv_sec = ts.tv_sec;
3798 stamp->tv_usec = ts.tv_nsec / 1000;
3801 static inline void skb_get_timestampns(const struct sk_buff *skb,
3802 struct __kernel_old_timespec *stamp)
3804 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3806 stamp->tv_sec = ts.tv_sec;
3807 stamp->tv_nsec = ts.tv_nsec;
3810 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3811 struct __kernel_timespec *stamp)
3813 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3815 stamp->tv_sec = ts.tv_sec;
3816 stamp->tv_nsec = ts.tv_nsec;
3819 static inline void __net_timestamp(struct sk_buff *skb)
3821 skb->tstamp = ktime_get_real();
3824 static inline ktime_t net_timedelta(ktime_t t)
3826 return ktime_sub(ktime_get_real(), t);
3829 static inline ktime_t net_invalid_timestamp(void)
3834 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3836 return skb_shinfo(skb)->meta_len;
3839 static inline void *skb_metadata_end(const struct sk_buff *skb)
3841 return skb_mac_header(skb);
3844 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3845 const struct sk_buff *skb_b,
3848 const void *a = skb_metadata_end(skb_a);
3849 const void *b = skb_metadata_end(skb_b);
3850 /* Using more efficient varaiant than plain call to memcmp(). */
3851 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3855 #define __it(x, op) (x -= sizeof(u##op))
3856 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3857 case 32: diffs |= __it_diff(a, b, 64);
3859 case 24: diffs |= __it_diff(a, b, 64);
3861 case 16: diffs |= __it_diff(a, b, 64);
3863 case 8: diffs |= __it_diff(a, b, 64);
3865 case 28: diffs |= __it_diff(a, b, 64);
3867 case 20: diffs |= __it_diff(a, b, 64);
3869 case 12: diffs |= __it_diff(a, b, 64);
3871 case 4: diffs |= __it_diff(a, b, 32);
3876 return memcmp(a - meta_len, b - meta_len, meta_len);
3880 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3881 const struct sk_buff *skb_b)
3883 u8 len_a = skb_metadata_len(skb_a);
3884 u8 len_b = skb_metadata_len(skb_b);
3886 if (!(len_a | len_b))
3889 return len_a != len_b ?
3890 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3893 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3895 skb_shinfo(skb)->meta_len = meta_len;
3898 static inline void skb_metadata_clear(struct sk_buff *skb)
3900 skb_metadata_set(skb, 0);
3903 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3905 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3907 void skb_clone_tx_timestamp(struct sk_buff *skb);
3908 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3910 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3912 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3916 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3921 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3924 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3926 * PHY drivers may accept clones of transmitted packets for
3927 * timestamping via their phy_driver.txtstamp method. These drivers
3928 * must call this function to return the skb back to the stack with a
3931 * @skb: clone of the original outgoing packet
3932 * @hwtstamps: hardware time stamps
3935 void skb_complete_tx_timestamp(struct sk_buff *skb,
3936 struct skb_shared_hwtstamps *hwtstamps);
3938 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
3939 struct skb_shared_hwtstamps *hwtstamps,
3940 struct sock *sk, int tstype);
3943 * skb_tstamp_tx - queue clone of skb with send time stamps
3944 * @orig_skb: the original outgoing packet
3945 * @hwtstamps: hardware time stamps, may be NULL if not available
3947 * If the skb has a socket associated, then this function clones the
3948 * skb (thus sharing the actual data and optional structures), stores
3949 * the optional hardware time stamping information (if non NULL) or
3950 * generates a software time stamp (otherwise), then queues the clone
3951 * to the error queue of the socket. Errors are silently ignored.
3953 void skb_tstamp_tx(struct sk_buff *orig_skb,
3954 struct skb_shared_hwtstamps *hwtstamps);
3957 * skb_tx_timestamp() - Driver hook for transmit timestamping
3959 * Ethernet MAC Drivers should call this function in their hard_xmit()
3960 * function immediately before giving the sk_buff to the MAC hardware.
3962 * Specifically, one should make absolutely sure that this function is
3963 * called before TX completion of this packet can trigger. Otherwise
3964 * the packet could potentially already be freed.
3966 * @skb: A socket buffer.
3968 static inline void skb_tx_timestamp(struct sk_buff *skb)
3970 skb_clone_tx_timestamp(skb);
3971 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3972 skb_tstamp_tx(skb, NULL);
3976 * skb_complete_wifi_ack - deliver skb with wifi status
3978 * @skb: the original outgoing packet
3979 * @acked: ack status
3982 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3984 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3985 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3987 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3989 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3991 (skb->ip_summed == CHECKSUM_PARTIAL &&
3992 skb_checksum_start_offset(skb) >= 0));
3996 * skb_checksum_complete - Calculate checksum of an entire packet
3997 * @skb: packet to process
3999 * This function calculates the checksum over the entire packet plus
4000 * the value of skb->csum. The latter can be used to supply the
4001 * checksum of a pseudo header as used by TCP/UDP. It returns the
4004 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4005 * this function can be used to verify that checksum on received
4006 * packets. In that case the function should return zero if the
4007 * checksum is correct. In particular, this function will return zero
4008 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4009 * hardware has already verified the correctness of the checksum.
4011 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4013 return skb_csum_unnecessary(skb) ?
4014 0 : __skb_checksum_complete(skb);
4017 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4019 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4020 if (skb->csum_level == 0)
4021 skb->ip_summed = CHECKSUM_NONE;
4027 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4029 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4030 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4032 } else if (skb->ip_summed == CHECKSUM_NONE) {
4033 skb->ip_summed = CHECKSUM_UNNECESSARY;
4034 skb->csum_level = 0;
4038 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4040 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4041 skb->ip_summed = CHECKSUM_NONE;
4042 skb->csum_level = 0;
4046 /* Check if we need to perform checksum complete validation.
4048 * Returns true if checksum complete is needed, false otherwise
4049 * (either checksum is unnecessary or zero checksum is allowed).
4051 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4055 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4056 skb->csum_valid = 1;
4057 __skb_decr_checksum_unnecessary(skb);
4064 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4067 #define CHECKSUM_BREAK 76
4069 /* Unset checksum-complete
4071 * Unset checksum complete can be done when packet is being modified
4072 * (uncompressed for instance) and checksum-complete value is
4075 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4077 if (skb->ip_summed == CHECKSUM_COMPLETE)
4078 skb->ip_summed = CHECKSUM_NONE;
4081 /* Validate (init) checksum based on checksum complete.
4084 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4085 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4086 * checksum is stored in skb->csum for use in __skb_checksum_complete
4087 * non-zero: value of invalid checksum
4090 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4094 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4095 if (!csum_fold(csum_add(psum, skb->csum))) {
4096 skb->csum_valid = 1;
4103 if (complete || skb->len <= CHECKSUM_BREAK) {
4106 csum = __skb_checksum_complete(skb);
4107 skb->csum_valid = !csum;
4114 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4119 /* Perform checksum validate (init). Note that this is a macro since we only
4120 * want to calculate the pseudo header which is an input function if necessary.
4121 * First we try to validate without any computation (checksum unnecessary) and
4122 * then calculate based on checksum complete calling the function to compute
4126 * 0: checksum is validated or try to in skb_checksum_complete
4127 * non-zero: value of invalid checksum
4129 #define __skb_checksum_validate(skb, proto, complete, \
4130 zero_okay, check, compute_pseudo) \
4132 __sum16 __ret = 0; \
4133 skb->csum_valid = 0; \
4134 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4135 __ret = __skb_checksum_validate_complete(skb, \
4136 complete, compute_pseudo(skb, proto)); \
4140 #define skb_checksum_init(skb, proto, compute_pseudo) \
4141 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4143 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4144 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4146 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4147 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4149 #define skb_checksum_validate_zero_check(skb, proto, check, \
4151 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4153 #define skb_checksum_simple_validate(skb) \
4154 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4156 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4158 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4161 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4163 skb->csum = ~pseudo;
4164 skb->ip_summed = CHECKSUM_COMPLETE;
4167 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4169 if (__skb_checksum_convert_check(skb)) \
4170 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4173 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4174 u16 start, u16 offset)
4176 skb->ip_summed = CHECKSUM_PARTIAL;
4177 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4178 skb->csum_offset = offset - start;
4181 /* Update skbuf and packet to reflect the remote checksum offload operation.
4182 * When called, ptr indicates the starting point for skb->csum when
4183 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4184 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4186 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4187 int start, int offset, bool nopartial)
4192 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4196 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4197 __skb_checksum_complete(skb);
4198 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4201 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4203 /* Adjust skb->csum since we changed the packet */
4204 skb->csum = csum_add(skb->csum, delta);
4207 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4209 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4210 return (void *)(skb->_nfct & NFCT_PTRMASK);
4216 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4218 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4225 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4227 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4228 skb->slow_gro |= !!nfct;
4233 #ifdef CONFIG_SKB_EXTENSIONS
4235 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4241 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4244 #if IS_ENABLED(CONFIG_MPTCP)
4247 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4250 SKB_EXT_NUM, /* must be last */
4254 * struct skb_ext - sk_buff extensions
4255 * @refcnt: 1 on allocation, deallocated on 0
4256 * @offset: offset to add to @data to obtain extension address
4257 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4258 * @data: start of extension data, variable sized
4260 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4261 * to use 'u8' types while allowing up to 2kb worth of extension data.
4265 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4266 u8 chunks; /* same */
4267 char data[] __aligned(8);
4270 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4271 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4272 struct skb_ext *ext);
4273 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4274 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4275 void __skb_ext_put(struct skb_ext *ext);
4277 static inline void skb_ext_put(struct sk_buff *skb)
4279 if (skb->active_extensions)
4280 __skb_ext_put(skb->extensions);
4283 static inline void __skb_ext_copy(struct sk_buff *dst,
4284 const struct sk_buff *src)
4286 dst->active_extensions = src->active_extensions;
4288 if (src->active_extensions) {
4289 struct skb_ext *ext = src->extensions;
4291 refcount_inc(&ext->refcnt);
4292 dst->extensions = ext;
4296 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4299 __skb_ext_copy(dst, src);
4302 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4304 return !!ext->offset[i];
4307 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4309 return skb->active_extensions & (1 << id);
4312 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4314 if (skb_ext_exist(skb, id))
4315 __skb_ext_del(skb, id);
4318 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4320 if (skb_ext_exist(skb, id)) {
4321 struct skb_ext *ext = skb->extensions;
4323 return (void *)ext + (ext->offset[id] << 3);
4329 static inline void skb_ext_reset(struct sk_buff *skb)
4331 if (unlikely(skb->active_extensions)) {
4332 __skb_ext_put(skb->extensions);
4333 skb->active_extensions = 0;
4337 static inline bool skb_has_extensions(struct sk_buff *skb)
4339 return unlikely(skb->active_extensions);
4342 static inline void skb_ext_put(struct sk_buff *skb) {}
4343 static inline void skb_ext_reset(struct sk_buff *skb) {}
4344 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4345 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4346 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4347 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4348 #endif /* CONFIG_SKB_EXTENSIONS */
4350 static inline void nf_reset_ct(struct sk_buff *skb)
4352 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4353 nf_conntrack_put(skb_nfct(skb));
4358 static inline void nf_reset_trace(struct sk_buff *skb)
4360 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4365 static inline void ipvs_reset(struct sk_buff *skb)
4367 #if IS_ENABLED(CONFIG_IP_VS)
4368 skb->ipvs_property = 0;
4372 /* Note: This doesn't put any conntrack info in dst. */
4373 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4376 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4377 dst->_nfct = src->_nfct;
4378 nf_conntrack_get(skb_nfct(src));
4380 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4382 dst->nf_trace = src->nf_trace;
4386 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4388 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4389 nf_conntrack_put(skb_nfct(dst));
4391 dst->slow_gro = src->slow_gro;
4392 __nf_copy(dst, src, true);
4395 #ifdef CONFIG_NETWORK_SECMARK
4396 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4398 to->secmark = from->secmark;
4401 static inline void skb_init_secmark(struct sk_buff *skb)
4406 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4409 static inline void skb_init_secmark(struct sk_buff *skb)
4413 static inline int secpath_exists(const struct sk_buff *skb)
4416 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4422 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4424 return !skb->destructor &&
4425 !secpath_exists(skb) &&
4427 !skb->_skb_refdst &&
4428 !skb_has_frag_list(skb);
4431 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4433 skb->queue_mapping = queue_mapping;
4436 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4438 return skb->queue_mapping;
4441 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4443 to->queue_mapping = from->queue_mapping;
4446 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4448 skb->queue_mapping = rx_queue + 1;
4451 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4453 return skb->queue_mapping - 1;
4456 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4458 return skb->queue_mapping != 0;
4461 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4463 skb->dst_pending_confirm = val;
4466 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4468 return skb->dst_pending_confirm != 0;
4471 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4474 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4480 /* Keeps track of mac header offset relative to skb->head.
4481 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4482 * For non-tunnel skb it points to skb_mac_header() and for
4483 * tunnel skb it points to outer mac header.
4484 * Keeps track of level of encapsulation of network headers.
4495 #define SKB_GSO_CB_OFFSET 32
4496 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4498 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4500 return (skb_mac_header(inner_skb) - inner_skb->head) -
4501 SKB_GSO_CB(inner_skb)->mac_offset;
4504 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4506 int new_headroom, headroom;
4509 headroom = skb_headroom(skb);
4510 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4514 new_headroom = skb_headroom(skb);
4515 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4519 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4521 /* Do not update partial checksums if remote checksum is enabled. */
4522 if (skb->remcsum_offload)
4525 SKB_GSO_CB(skb)->csum = res;
4526 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4529 /* Compute the checksum for a gso segment. First compute the checksum value
4530 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4531 * then add in skb->csum (checksum from csum_start to end of packet).
4532 * skb->csum and csum_start are then updated to reflect the checksum of the
4533 * resultant packet starting from the transport header-- the resultant checksum
4534 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4537 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4539 unsigned char *csum_start = skb_transport_header(skb);
4540 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4541 __wsum partial = SKB_GSO_CB(skb)->csum;
4543 SKB_GSO_CB(skb)->csum = res;
4544 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4546 return csum_fold(csum_partial(csum_start, plen, partial));
4549 static inline bool skb_is_gso(const struct sk_buff *skb)
4551 return skb_shinfo(skb)->gso_size;
4554 /* Note: Should be called only if skb_is_gso(skb) is true */
4555 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4557 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4560 /* Note: Should be called only if skb_is_gso(skb) is true */
4561 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4563 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4566 /* Note: Should be called only if skb_is_gso(skb) is true */
4567 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4569 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4572 static inline void skb_gso_reset(struct sk_buff *skb)
4574 skb_shinfo(skb)->gso_size = 0;
4575 skb_shinfo(skb)->gso_segs = 0;
4576 skb_shinfo(skb)->gso_type = 0;
4579 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4582 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4584 shinfo->gso_size += increment;
4587 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4590 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4592 shinfo->gso_size -= decrement;
4595 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4597 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4599 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4600 * wanted then gso_type will be set. */
4601 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4603 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4604 unlikely(shinfo->gso_type == 0)) {
4605 __skb_warn_lro_forwarding(skb);
4611 static inline void skb_forward_csum(struct sk_buff *skb)
4613 /* Unfortunately we don't support this one. Any brave souls? */
4614 if (skb->ip_summed == CHECKSUM_COMPLETE)
4615 skb->ip_summed = CHECKSUM_NONE;
4619 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4620 * @skb: skb to check
4622 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4623 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4624 * use this helper, to document places where we make this assertion.
4626 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4629 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4633 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4635 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4636 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4637 unsigned int transport_len,
4638 __sum16(*skb_chkf)(struct sk_buff *skb));
4641 * skb_head_is_locked - Determine if the skb->head is locked down
4642 * @skb: skb to check
4644 * The head on skbs build around a head frag can be removed if they are
4645 * not cloned. This function returns true if the skb head is locked down
4646 * due to either being allocated via kmalloc, or by being a clone with
4647 * multiple references to the head.
4649 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4651 return !skb->head_frag || skb_cloned(skb);
4654 /* Local Checksum Offload.
4655 * Compute outer checksum based on the assumption that the
4656 * inner checksum will be offloaded later.
4657 * See Documentation/networking/checksum-offloads.rst for
4658 * explanation of how this works.
4659 * Fill in outer checksum adjustment (e.g. with sum of outer
4660 * pseudo-header) before calling.
4661 * Also ensure that inner checksum is in linear data area.
4663 static inline __wsum lco_csum(struct sk_buff *skb)
4665 unsigned char *csum_start = skb_checksum_start(skb);
4666 unsigned char *l4_hdr = skb_transport_header(skb);
4669 /* Start with complement of inner checksum adjustment */
4670 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4673 /* Add in checksum of our headers (incl. outer checksum
4674 * adjustment filled in by caller) and return result.
4676 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4679 static inline bool skb_is_redirected(const struct sk_buff *skb)
4681 return skb->redirected;
4684 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4686 skb->redirected = 1;
4687 #ifdef CONFIG_NET_REDIRECT
4688 skb->from_ingress = from_ingress;
4689 if (skb->from_ingress)
4694 static inline void skb_reset_redirect(struct sk_buff *skb)
4696 skb->redirected = 0;
4699 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4701 return skb->csum_not_inet;
4704 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4705 const u64 kcov_handle)
4708 skb->kcov_handle = kcov_handle;
4712 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4715 return skb->kcov_handle;
4721 #ifdef CONFIG_PAGE_POOL
4722 static inline void skb_mark_for_recycle(struct sk_buff *skb)
4724 skb->pp_recycle = 1;
4728 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4730 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4732 return page_pool_return_skb_page(virt_to_page(data));
4735 #endif /* __KERNEL__ */
4736 #endif /* _LINUX_SKBUFF_H */