1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * Definitions for the 'struct sk_buff' memory handlers.
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
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 * @peeked: this packet has been seen already, so stats have been
656 * done for it, don't do them again
657 * @nf_trace: netfilter packet trace flag
658 * @protocol: Packet protocol from driver
659 * @destructor: Destruct function
660 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
661 * @_sk_redir: socket redirection information for skmsg
662 * @_nfct: Associated connection, if any (with nfctinfo bits)
663 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
664 * @skb_iif: ifindex of device we arrived on
665 * @tc_index: Traffic control index
666 * @hash: the packet hash
667 * @queue_mapping: Queue mapping for multiqueue devices
668 * @head_frag: skb was allocated from page fragments,
669 * not allocated by kmalloc() or vmalloc().
670 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
671 * @pp_recycle: mark the packet for recycling instead of freeing (implies
672 * page_pool support on driver)
673 * @active_extensions: active extensions (skb_ext_id types)
674 * @ndisc_nodetype: router type (from link layer)
675 * @ooo_okay: allow the mapping of a socket to a queue to be changed
676 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
678 * @sw_hash: indicates hash was computed in software stack
679 * @wifi_acked_valid: wifi_acked was set
680 * @wifi_acked: whether frame was acked on wifi or not
681 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
682 * @encapsulation: indicates the inner headers in the skbuff are valid
683 * @encap_hdr_csum: software checksum is needed
684 * @csum_valid: checksum is already valid
685 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
686 * @csum_complete_sw: checksum was completed by software
687 * @csum_level: indicates the number of consecutive checksums found in
688 * the packet minus one that have been verified as
689 * CHECKSUM_UNNECESSARY (max 3)
690 * @dst_pending_confirm: need to confirm neighbour
691 * @decrypted: Decrypted SKB
692 * @slow_gro: state present at GRO time, slower prepare step required
693 * @napi_id: id of the NAPI struct this skb came from
694 * @sender_cpu: (aka @napi_id) source CPU in XPS
695 * @secmark: security marking
696 * @mark: Generic packet mark
697 * @reserved_tailroom: (aka @mark) number of bytes of free space available
698 * at the tail of an sk_buff
699 * @vlan_present: VLAN tag is present
700 * @vlan_proto: vlan encapsulation protocol
701 * @vlan_tci: vlan tag control information
702 * @inner_protocol: Protocol (encapsulation)
703 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
704 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
705 * @inner_transport_header: Inner transport layer header (encapsulation)
706 * @inner_network_header: Network layer header (encapsulation)
707 * @inner_mac_header: Link layer header (encapsulation)
708 * @transport_header: Transport layer header
709 * @network_header: Network layer header
710 * @mac_header: Link layer header
711 * @kcov_handle: KCOV remote handle for remote coverage collection
712 * @tail: Tail pointer
714 * @head: Head of buffer
715 * @data: Data head pointer
716 * @truesize: Buffer size
717 * @users: User count - see {datagram,tcp}.c
718 * @extensions: allocated extensions, valid if active_extensions is nonzero
724 /* These two members must be first. */
725 struct sk_buff *next;
726 struct sk_buff *prev;
729 struct net_device *dev;
730 /* Some protocols might use this space to store information,
731 * while device pointer would be NULL.
732 * UDP receive path is one user.
734 unsigned long dev_scratch;
737 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
738 struct list_head list;
743 int ip_defrag_offset;
748 u64 skb_mstamp_ns; /* earliest departure time */
751 * This is the control buffer. It is free to use for every
752 * layer. Please put your private variables there. If you
753 * want to keep them across layers you have to do a skb_clone()
754 * first. This is owned by whoever has the skb queued ATM.
756 char cb[48] __aligned(8);
760 unsigned long _skb_refdst;
761 void (*destructor)(struct sk_buff *skb);
763 struct list_head tcp_tsorted_anchor;
764 #ifdef CONFIG_NET_SOCK_MSG
765 unsigned long _sk_redir;
769 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
777 /* Following fields are _not_ copied in __copy_skb_header()
778 * Note that queue_mapping is here mostly to fill a hole.
782 /* if you move cloned around you also must adapt those constants */
783 #ifdef __BIG_ENDIAN_BITFIELD
784 #define CLONED_MASK (1 << 7)
786 #define CLONED_MASK 1
788 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
791 __u8 __cloned_offset[0];
799 pp_recycle:1; /* page_pool recycle indicator */
800 #ifdef CONFIG_SKB_EXTENSIONS
801 __u8 active_extensions;
804 /* fields enclosed in headers_start/headers_end are copied
805 * using a single memcpy() in __copy_skb_header()
808 __u32 headers_start[0];
811 /* if you move pkt_type around you also must adapt those constants */
812 #ifdef __BIG_ENDIAN_BITFIELD
813 #define PKT_TYPE_MAX (7 << 5)
815 #define PKT_TYPE_MAX 7
817 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
820 __u8 __pkt_type_offset[0];
830 __u8 wifi_acked_valid:1;
833 /* Indicates the inner headers are valid in the skbuff. */
834 __u8 encapsulation:1;
835 __u8 encap_hdr_csum:1;
838 #ifdef __BIG_ENDIAN_BITFIELD
839 #define PKT_VLAN_PRESENT_BIT 7
841 #define PKT_VLAN_PRESENT_BIT 0
843 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
845 __u8 __pkt_vlan_present_offset[0];
848 __u8 csum_complete_sw:1;
850 __u8 csum_not_inet:1;
851 __u8 dst_pending_confirm:1;
852 #ifdef CONFIG_IPV6_NDISC_NODETYPE
853 __u8 ndisc_nodetype:2;
856 __u8 ipvs_property:1;
857 __u8 inner_protocol_type:1;
858 __u8 remcsum_offload:1;
859 #ifdef CONFIG_NET_SWITCHDEV
860 __u8 offload_fwd_mark:1;
861 __u8 offload_l3_fwd_mark:1;
863 #ifdef CONFIG_NET_CLS_ACT
864 __u8 tc_skip_classify:1;
865 __u8 tc_at_ingress:1;
868 #ifdef CONFIG_NET_REDIRECT
871 #ifdef CONFIG_TLS_DEVICE
876 #ifdef CONFIG_NET_SCHED
877 __u16 tc_index; /* traffic control index */
892 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
894 unsigned int napi_id;
895 unsigned int sender_cpu;
898 #ifdef CONFIG_NETWORK_SECMARK
904 __u32 reserved_tailroom;
908 __be16 inner_protocol;
912 __u16 inner_transport_header;
913 __u16 inner_network_header;
914 __u16 inner_mac_header;
917 __u16 transport_header;
918 __u16 network_header;
926 __u32 headers_end[0];
929 /* These elements must be at the end, see alloc_skb() for details. */
934 unsigned int truesize;
937 #ifdef CONFIG_SKB_EXTENSIONS
938 /* only useable after checking ->active_extensions != 0 */
939 struct skb_ext *extensions;
945 * Handling routines are only of interest to the kernel
948 #define SKB_ALLOC_FCLONE 0x01
949 #define SKB_ALLOC_RX 0x02
950 #define SKB_ALLOC_NAPI 0x04
953 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
956 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
958 return unlikely(skb->pfmemalloc);
962 * skb might have a dst pointer attached, refcounted or not.
963 * _skb_refdst low order bit is set if refcount was _not_ taken
965 #define SKB_DST_NOREF 1UL
966 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
969 * skb_dst - returns skb dst_entry
972 * Returns skb dst_entry, regardless of reference taken or not.
974 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
976 /* If refdst was not refcounted, check we still are in a
977 * rcu_read_lock section
979 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
980 !rcu_read_lock_held() &&
981 !rcu_read_lock_bh_held());
982 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
986 * skb_dst_set - sets skb dst
990 * Sets skb dst, assuming a reference was taken on dst and should
991 * be released by skb_dst_drop()
993 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
995 skb->slow_gro |= !!dst;
996 skb->_skb_refdst = (unsigned long)dst;
1000 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1004 * Sets skb dst, assuming a reference was not taken on dst.
1005 * If dst entry is cached, we do not take reference and dst_release
1006 * will be avoided by refdst_drop. If dst entry is not cached, we take
1007 * reference, so that last dst_release can destroy the dst immediately.
1009 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1011 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1012 skb->slow_gro |= !!dst;
1013 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1017 * skb_dst_is_noref - Test if skb dst isn't refcounted
1020 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1022 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1026 * skb_rtable - Returns the skb &rtable
1029 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1031 return (struct rtable *)skb_dst(skb);
1034 /* For mangling skb->pkt_type from user space side from applications
1035 * such as nft, tc, etc, we only allow a conservative subset of
1036 * possible pkt_types to be set.
1038 static inline bool skb_pkt_type_ok(u32 ptype)
1040 return ptype <= PACKET_OTHERHOST;
1044 * skb_napi_id - Returns the skb's NAPI id
1047 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1049 #ifdef CONFIG_NET_RX_BUSY_POLL
1050 return skb->napi_id;
1057 * skb_unref - decrement the skb's reference count
1060 * Returns true if we can free the skb.
1062 static inline bool skb_unref(struct sk_buff *skb)
1066 if (likely(refcount_read(&skb->users) == 1))
1068 else if (likely(!refcount_dec_and_test(&skb->users)))
1074 void skb_release_head_state(struct sk_buff *skb);
1075 void kfree_skb(struct sk_buff *skb);
1076 void kfree_skb_list(struct sk_buff *segs);
1077 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1078 void skb_tx_error(struct sk_buff *skb);
1080 #ifdef CONFIG_TRACEPOINTS
1081 void consume_skb(struct sk_buff *skb);
1083 static inline void consume_skb(struct sk_buff *skb)
1085 return kfree_skb(skb);
1089 void __consume_stateless_skb(struct sk_buff *skb);
1090 void __kfree_skb(struct sk_buff *skb);
1091 extern struct kmem_cache *skbuff_head_cache;
1093 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1094 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1095 bool *fragstolen, int *delta_truesize);
1097 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1099 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1100 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1101 struct sk_buff *build_skb_around(struct sk_buff *skb,
1102 void *data, unsigned int frag_size);
1104 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1107 * alloc_skb - allocate a network buffer
1108 * @size: size to allocate
1109 * @priority: allocation mask
1111 * This function is a convenient wrapper around __alloc_skb().
1113 static inline struct sk_buff *alloc_skb(unsigned int size,
1116 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1119 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1120 unsigned long data_len,
1124 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1126 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1127 struct sk_buff_fclones {
1128 struct sk_buff skb1;
1130 struct sk_buff skb2;
1132 refcount_t fclone_ref;
1136 * skb_fclone_busy - check if fclone is busy
1140 * Returns true if skb is a fast clone, and its clone is not freed.
1141 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1142 * so we also check that this didnt happen.
1144 static inline bool skb_fclone_busy(const struct sock *sk,
1145 const struct sk_buff *skb)
1147 const struct sk_buff_fclones *fclones;
1149 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1151 return skb->fclone == SKB_FCLONE_ORIG &&
1152 refcount_read(&fclones->fclone_ref) > 1 &&
1153 READ_ONCE(fclones->skb2.sk) == sk;
1157 * alloc_skb_fclone - allocate a network buffer from fclone cache
1158 * @size: size to allocate
1159 * @priority: allocation mask
1161 * This function is a convenient wrapper around __alloc_skb().
1163 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1166 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1169 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1170 void skb_headers_offset_update(struct sk_buff *skb, int off);
1171 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1172 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1173 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1174 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1175 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1176 gfp_t gfp_mask, bool fclone);
1177 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1180 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1183 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1184 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1185 unsigned int headroom);
1186 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1187 int newtailroom, gfp_t priority);
1188 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1189 int offset, int len);
1190 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1191 int offset, int len);
1192 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1193 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1196 * skb_pad - zero pad the tail of an skb
1197 * @skb: buffer to pad
1198 * @pad: space to pad
1200 * Ensure that a buffer is followed by a padding area that is zero
1201 * filled. Used by network drivers which may DMA or transfer data
1202 * beyond the buffer end onto the wire.
1204 * May return error in out of memory cases. The skb is freed on error.
1206 static inline int skb_pad(struct sk_buff *skb, int pad)
1208 return __skb_pad(skb, pad, true);
1210 #define dev_kfree_skb(a) consume_skb(a)
1212 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1213 int offset, size_t size);
1215 struct skb_seq_state {
1219 __u32 stepped_offset;
1220 struct sk_buff *root_skb;
1221 struct sk_buff *cur_skb;
1226 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1227 unsigned int to, struct skb_seq_state *st);
1228 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1229 struct skb_seq_state *st);
1230 void skb_abort_seq_read(struct skb_seq_state *st);
1232 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1233 unsigned int to, struct ts_config *config);
1236 * Packet hash types specify the type of hash in skb_set_hash.
1238 * Hash types refer to the protocol layer addresses which are used to
1239 * construct a packet's hash. The hashes are used to differentiate or identify
1240 * flows of the protocol layer for the hash type. Hash types are either
1241 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1243 * Properties of hashes:
1245 * 1) Two packets in different flows have different hash values
1246 * 2) Two packets in the same flow should have the same hash value
1248 * A hash at a higher layer is considered to be more specific. A driver should
1249 * set the most specific hash possible.
1251 * A driver cannot indicate a more specific hash than the layer at which a hash
1252 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1254 * A driver may indicate a hash level which is less specific than the
1255 * actual layer the hash was computed on. For instance, a hash computed
1256 * at L4 may be considered an L3 hash. This should only be done if the
1257 * driver can't unambiguously determine that the HW computed the hash at
1258 * the higher layer. Note that the "should" in the second property above
1261 enum pkt_hash_types {
1262 PKT_HASH_TYPE_NONE, /* Undefined type */
1263 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1264 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1265 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1268 static inline void skb_clear_hash(struct sk_buff *skb)
1275 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1278 skb_clear_hash(skb);
1282 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1284 skb->l4_hash = is_l4;
1285 skb->sw_hash = is_sw;
1290 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1292 /* Used by drivers to set hash from HW */
1293 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1297 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1299 __skb_set_hash(skb, hash, true, is_l4);
1302 void __skb_get_hash(struct sk_buff *skb);
1303 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1304 u32 skb_get_poff(const struct sk_buff *skb);
1305 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1306 const struct flow_keys_basic *keys, int hlen);
1307 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1308 const void *data, int hlen_proto);
1310 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1311 int thoff, u8 ip_proto)
1313 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1316 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1317 const struct flow_dissector_key *key,
1318 unsigned int key_count);
1320 struct bpf_flow_dissector;
1321 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1322 __be16 proto, int nhoff, int hlen, unsigned int flags);
1324 bool __skb_flow_dissect(const struct net *net,
1325 const struct sk_buff *skb,
1326 struct flow_dissector *flow_dissector,
1327 void *target_container, const void *data,
1328 __be16 proto, int nhoff, int hlen, unsigned int flags);
1330 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1331 struct flow_dissector *flow_dissector,
1332 void *target_container, unsigned int flags)
1334 return __skb_flow_dissect(NULL, skb, flow_dissector,
1335 target_container, NULL, 0, 0, 0, flags);
1338 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1339 struct flow_keys *flow,
1342 memset(flow, 0, sizeof(*flow));
1343 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1344 flow, NULL, 0, 0, 0, flags);
1348 skb_flow_dissect_flow_keys_basic(const struct net *net,
1349 const struct sk_buff *skb,
1350 struct flow_keys_basic *flow,
1351 const void *data, __be16 proto,
1352 int nhoff, int hlen, unsigned int flags)
1354 memset(flow, 0, sizeof(*flow));
1355 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1356 data, proto, nhoff, hlen, flags);
1359 void skb_flow_dissect_meta(const struct sk_buff *skb,
1360 struct flow_dissector *flow_dissector,
1361 void *target_container);
1363 /* Gets a skb connection tracking info, ctinfo map should be a
1364 * map of mapsize to translate enum ip_conntrack_info states
1368 skb_flow_dissect_ct(const struct sk_buff *skb,
1369 struct flow_dissector *flow_dissector,
1370 void *target_container,
1371 u16 *ctinfo_map, size_t mapsize,
1374 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1375 struct flow_dissector *flow_dissector,
1376 void *target_container);
1378 void skb_flow_dissect_hash(const struct sk_buff *skb,
1379 struct flow_dissector *flow_dissector,
1380 void *target_container);
1382 static inline __u32 skb_get_hash(struct sk_buff *skb)
1384 if (!skb->l4_hash && !skb->sw_hash)
1385 __skb_get_hash(skb);
1390 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1392 if (!skb->l4_hash && !skb->sw_hash) {
1393 struct flow_keys keys;
1394 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1396 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1402 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1403 const siphash_key_t *perturb);
1405 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1410 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1412 to->hash = from->hash;
1413 to->sw_hash = from->sw_hash;
1414 to->l4_hash = from->l4_hash;
1417 static inline void skb_copy_decrypted(struct sk_buff *to,
1418 const struct sk_buff *from)
1420 #ifdef CONFIG_TLS_DEVICE
1421 to->decrypted = from->decrypted;
1425 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1426 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1428 return skb->head + skb->end;
1431 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1436 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1441 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1443 return skb->end - skb->head;
1448 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1450 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1452 return &skb_shinfo(skb)->hwtstamps;
1455 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1457 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1459 return is_zcopy ? skb_uarg(skb) : NULL;
1462 static inline void net_zcopy_get(struct ubuf_info *uarg)
1464 refcount_inc(&uarg->refcnt);
1467 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1469 skb_shinfo(skb)->destructor_arg = uarg;
1470 skb_shinfo(skb)->flags |= uarg->flags;
1473 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1476 if (skb && uarg && !skb_zcopy(skb)) {
1477 if (unlikely(have_ref && *have_ref))
1480 net_zcopy_get(uarg);
1481 skb_zcopy_init(skb, uarg);
1485 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1487 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1488 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1491 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1493 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1496 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1498 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1501 static inline void net_zcopy_put(struct ubuf_info *uarg)
1504 uarg->callback(NULL, uarg, true);
1507 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1510 if (uarg->callback == msg_zerocopy_callback)
1511 msg_zerocopy_put_abort(uarg, have_uref);
1513 net_zcopy_put(uarg);
1517 /* Release a reference on a zerocopy structure */
1518 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1520 struct ubuf_info *uarg = skb_zcopy(skb);
1523 if (!skb_zcopy_is_nouarg(skb))
1524 uarg->callback(skb, uarg, zerocopy_success);
1526 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
1530 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1535 /* Iterate through singly-linked GSO fragments of an skb. */
1536 #define skb_list_walk_safe(first, skb, next_skb) \
1537 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1538 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1540 static inline void skb_list_del_init(struct sk_buff *skb)
1542 __list_del_entry(&skb->list);
1543 skb_mark_not_on_list(skb);
1547 * skb_queue_empty - check if a queue is empty
1550 * Returns true if the queue is empty, false otherwise.
1552 static inline int skb_queue_empty(const struct sk_buff_head *list)
1554 return list->next == (const struct sk_buff *) list;
1558 * skb_queue_empty_lockless - check if a queue is empty
1561 * Returns true if the queue is empty, false otherwise.
1562 * This variant can be used in lockless contexts.
1564 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1566 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1571 * skb_queue_is_last - check if skb is the last entry in the queue
1575 * Returns true if @skb is the last buffer on the list.
1577 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1578 const struct sk_buff *skb)
1580 return skb->next == (const struct sk_buff *) list;
1584 * skb_queue_is_first - check if skb is the first entry in the queue
1588 * Returns true if @skb is the first buffer on the list.
1590 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1591 const struct sk_buff *skb)
1593 return skb->prev == (const struct sk_buff *) list;
1597 * skb_queue_next - return the next packet in the queue
1599 * @skb: current buffer
1601 * Return the next packet in @list after @skb. It is only valid to
1602 * call this if skb_queue_is_last() evaluates to false.
1604 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1605 const struct sk_buff *skb)
1607 /* This BUG_ON may seem severe, but if we just return then we
1608 * are going to dereference garbage.
1610 BUG_ON(skb_queue_is_last(list, skb));
1615 * skb_queue_prev - return the prev packet in the queue
1617 * @skb: current buffer
1619 * Return the prev packet in @list before @skb. It is only valid to
1620 * call this if skb_queue_is_first() evaluates to false.
1622 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1623 const struct sk_buff *skb)
1625 /* This BUG_ON may seem severe, but if we just return then we
1626 * are going to dereference garbage.
1628 BUG_ON(skb_queue_is_first(list, skb));
1633 * skb_get - reference buffer
1634 * @skb: buffer to reference
1636 * Makes another reference to a socket buffer and returns a pointer
1639 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1641 refcount_inc(&skb->users);
1646 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1650 * skb_cloned - is the buffer a clone
1651 * @skb: buffer to check
1653 * Returns true if the buffer was generated with skb_clone() and is
1654 * one of multiple shared copies of the buffer. Cloned buffers are
1655 * shared data so must not be written to under normal circumstances.
1657 static inline int skb_cloned(const struct sk_buff *skb)
1659 return skb->cloned &&
1660 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1663 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1665 might_sleep_if(gfpflags_allow_blocking(pri));
1667 if (skb_cloned(skb))
1668 return pskb_expand_head(skb, 0, 0, pri);
1674 * skb_header_cloned - is the header a clone
1675 * @skb: buffer to check
1677 * Returns true if modifying the header part of the buffer requires
1678 * the data to be copied.
1680 static inline int skb_header_cloned(const struct sk_buff *skb)
1687 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1688 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1689 return dataref != 1;
1692 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1694 might_sleep_if(gfpflags_allow_blocking(pri));
1696 if (skb_header_cloned(skb))
1697 return pskb_expand_head(skb, 0, 0, pri);
1703 * __skb_header_release - release reference to header
1704 * @skb: buffer to operate on
1706 static inline void __skb_header_release(struct sk_buff *skb)
1709 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1714 * skb_shared - is the buffer shared
1715 * @skb: buffer to check
1717 * Returns true if more than one person has a reference to this
1720 static inline int skb_shared(const struct sk_buff *skb)
1722 return refcount_read(&skb->users) != 1;
1726 * skb_share_check - check if buffer is shared and if so clone it
1727 * @skb: buffer to check
1728 * @pri: priority for memory allocation
1730 * If the buffer is shared the buffer is cloned and the old copy
1731 * drops a reference. A new clone with a single reference is returned.
1732 * If the buffer is not shared the original buffer is returned. When
1733 * being called from interrupt status or with spinlocks held pri must
1736 * NULL is returned on a memory allocation failure.
1738 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1740 might_sleep_if(gfpflags_allow_blocking(pri));
1741 if (skb_shared(skb)) {
1742 struct sk_buff *nskb = skb_clone(skb, pri);
1754 * Copy shared buffers into a new sk_buff. We effectively do COW on
1755 * packets to handle cases where we have a local reader and forward
1756 * and a couple of other messy ones. The normal one is tcpdumping
1757 * a packet thats being forwarded.
1761 * skb_unshare - make a copy of a shared buffer
1762 * @skb: buffer to check
1763 * @pri: priority for memory allocation
1765 * If the socket buffer is a clone then this function creates a new
1766 * copy of the data, drops a reference count on the old copy and returns
1767 * the new copy with the reference count at 1. If the buffer is not a clone
1768 * the original buffer is returned. When called with a spinlock held or
1769 * from interrupt state @pri must be %GFP_ATOMIC
1771 * %NULL is returned on a memory allocation failure.
1773 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1776 might_sleep_if(gfpflags_allow_blocking(pri));
1777 if (skb_cloned(skb)) {
1778 struct sk_buff *nskb = skb_copy(skb, pri);
1780 /* Free our shared copy */
1791 * skb_peek - peek at the head of an &sk_buff_head
1792 * @list_: list to peek at
1794 * Peek an &sk_buff. Unlike most other operations you _MUST_
1795 * be careful with this one. A peek leaves the buffer on the
1796 * list and someone else may run off with it. You must hold
1797 * the appropriate locks or have a private queue to do this.
1799 * Returns %NULL for an empty list or a pointer to the head element.
1800 * The reference count is not incremented and the reference is therefore
1801 * volatile. Use with caution.
1803 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1805 struct sk_buff *skb = list_->next;
1807 if (skb == (struct sk_buff *)list_)
1813 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1814 * @list_: list to peek at
1816 * Like skb_peek(), but the caller knows that the list is not empty.
1818 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1824 * skb_peek_next - peek skb following the given one from a queue
1825 * @skb: skb to start from
1826 * @list_: list to peek at
1828 * Returns %NULL when the end of the list is met or a pointer to the
1829 * next element. The reference count is not incremented and the
1830 * reference is therefore volatile. Use with caution.
1832 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1833 const struct sk_buff_head *list_)
1835 struct sk_buff *next = skb->next;
1837 if (next == (struct sk_buff *)list_)
1843 * skb_peek_tail - peek at the tail of an &sk_buff_head
1844 * @list_: list to peek at
1846 * Peek an &sk_buff. Unlike most other operations you _MUST_
1847 * be careful with this one. A peek leaves the buffer on the
1848 * list and someone else may run off with it. You must hold
1849 * the appropriate locks or have a private queue to do this.
1851 * Returns %NULL for an empty list or a pointer to the tail element.
1852 * The reference count is not incremented and the reference is therefore
1853 * volatile. Use with caution.
1855 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1857 struct sk_buff *skb = READ_ONCE(list_->prev);
1859 if (skb == (struct sk_buff *)list_)
1866 * skb_queue_len - get queue length
1867 * @list_: list to measure
1869 * Return the length of an &sk_buff queue.
1871 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1877 * skb_queue_len_lockless - get queue length
1878 * @list_: list to measure
1880 * Return the length of an &sk_buff queue.
1881 * This variant can be used in lockless contexts.
1883 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1885 return READ_ONCE(list_->qlen);
1889 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1890 * @list: queue to initialize
1892 * This initializes only the list and queue length aspects of
1893 * an sk_buff_head object. This allows to initialize the list
1894 * aspects of an sk_buff_head without reinitializing things like
1895 * the spinlock. It can also be used for on-stack sk_buff_head
1896 * objects where the spinlock is known to not be used.
1898 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1900 list->prev = list->next = (struct sk_buff *)list;
1905 * This function creates a split out lock class for each invocation;
1906 * this is needed for now since a whole lot of users of the skb-queue
1907 * infrastructure in drivers have different locking usage (in hardirq)
1908 * than the networking core (in softirq only). In the long run either the
1909 * network layer or drivers should need annotation to consolidate the
1910 * main types of usage into 3 classes.
1912 static inline void skb_queue_head_init(struct sk_buff_head *list)
1914 spin_lock_init(&list->lock);
1915 __skb_queue_head_init(list);
1918 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1919 struct lock_class_key *class)
1921 skb_queue_head_init(list);
1922 lockdep_set_class(&list->lock, class);
1926 * Insert an sk_buff on a list.
1928 * The "__skb_xxxx()" functions are the non-atomic ones that
1929 * can only be called with interrupts disabled.
1931 static inline void __skb_insert(struct sk_buff *newsk,
1932 struct sk_buff *prev, struct sk_buff *next,
1933 struct sk_buff_head *list)
1935 /* See skb_queue_empty_lockless() and skb_peek_tail()
1936 * for the opposite READ_ONCE()
1938 WRITE_ONCE(newsk->next, next);
1939 WRITE_ONCE(newsk->prev, prev);
1940 WRITE_ONCE(next->prev, newsk);
1941 WRITE_ONCE(prev->next, newsk);
1945 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1946 struct sk_buff *prev,
1947 struct sk_buff *next)
1949 struct sk_buff *first = list->next;
1950 struct sk_buff *last = list->prev;
1952 WRITE_ONCE(first->prev, prev);
1953 WRITE_ONCE(prev->next, first);
1955 WRITE_ONCE(last->next, next);
1956 WRITE_ONCE(next->prev, last);
1960 * skb_queue_splice - join two skb lists, this is designed for stacks
1961 * @list: the new list to add
1962 * @head: the place to add it in the first list
1964 static inline void skb_queue_splice(const struct sk_buff_head *list,
1965 struct sk_buff_head *head)
1967 if (!skb_queue_empty(list)) {
1968 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1969 head->qlen += list->qlen;
1974 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1975 * @list: the new list to add
1976 * @head: the place to add it in the first list
1978 * The list at @list is reinitialised
1980 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1981 struct sk_buff_head *head)
1983 if (!skb_queue_empty(list)) {
1984 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1985 head->qlen += list->qlen;
1986 __skb_queue_head_init(list);
1991 * skb_queue_splice_tail - join two skb lists, each list being a queue
1992 * @list: the new list to add
1993 * @head: the place to add it in the first list
1995 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1996 struct sk_buff_head *head)
1998 if (!skb_queue_empty(list)) {
1999 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2000 head->qlen += list->qlen;
2005 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2006 * @list: the new list to add
2007 * @head: the place to add it in the first list
2009 * Each of the lists is a queue.
2010 * The list at @list is reinitialised
2012 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2013 struct sk_buff_head *head)
2015 if (!skb_queue_empty(list)) {
2016 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2017 head->qlen += list->qlen;
2018 __skb_queue_head_init(list);
2023 * __skb_queue_after - queue a buffer at the list head
2024 * @list: list to use
2025 * @prev: place after this buffer
2026 * @newsk: buffer to queue
2028 * Queue a buffer int the middle of a list. This function takes no locks
2029 * and you must therefore hold required locks before calling it.
2031 * A buffer cannot be placed on two lists at the same time.
2033 static inline void __skb_queue_after(struct sk_buff_head *list,
2034 struct sk_buff *prev,
2035 struct sk_buff *newsk)
2037 __skb_insert(newsk, prev, prev->next, list);
2040 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2041 struct sk_buff_head *list);
2043 static inline void __skb_queue_before(struct sk_buff_head *list,
2044 struct sk_buff *next,
2045 struct sk_buff *newsk)
2047 __skb_insert(newsk, next->prev, next, list);
2051 * __skb_queue_head - queue a buffer at the list head
2052 * @list: list to use
2053 * @newsk: buffer to queue
2055 * Queue a buffer at the start of a list. This function takes no locks
2056 * and you must therefore hold required locks before calling it.
2058 * A buffer cannot be placed on two lists at the same time.
2060 static inline void __skb_queue_head(struct sk_buff_head *list,
2061 struct sk_buff *newsk)
2063 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2065 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2068 * __skb_queue_tail - queue a buffer at the list tail
2069 * @list: list to use
2070 * @newsk: buffer to queue
2072 * Queue a buffer at the end of a list. This function takes no locks
2073 * and you must therefore hold required locks before calling it.
2075 * A buffer cannot be placed on two lists at the same time.
2077 static inline void __skb_queue_tail(struct sk_buff_head *list,
2078 struct sk_buff *newsk)
2080 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2082 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2085 * remove sk_buff from list. _Must_ be called atomically, and with
2088 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2089 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2091 struct sk_buff *next, *prev;
2093 WRITE_ONCE(list->qlen, list->qlen - 1);
2096 skb->next = skb->prev = NULL;
2097 WRITE_ONCE(next->prev, prev);
2098 WRITE_ONCE(prev->next, next);
2102 * __skb_dequeue - remove from the head of the queue
2103 * @list: list to dequeue from
2105 * Remove the head of the list. This function does not take any locks
2106 * so must be used with appropriate locks held only. The head item is
2107 * returned or %NULL if the list is empty.
2109 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2111 struct sk_buff *skb = skb_peek(list);
2113 __skb_unlink(skb, list);
2116 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2119 * __skb_dequeue_tail - remove from the tail of the queue
2120 * @list: list to dequeue from
2122 * Remove the tail of the list. This function does not take any locks
2123 * so must be used with appropriate locks held only. The tail item is
2124 * returned or %NULL if the list is empty.
2126 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2128 struct sk_buff *skb = skb_peek_tail(list);
2130 __skb_unlink(skb, list);
2133 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2136 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2138 return skb->data_len;
2141 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2143 return skb->len - skb->data_len;
2146 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2148 unsigned int i, len = 0;
2150 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2151 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2155 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2157 return skb_headlen(skb) + __skb_pagelen(skb);
2161 * __skb_fill_page_desc - initialise a paged fragment in an skb
2162 * @skb: buffer containing fragment to be initialised
2163 * @i: paged fragment index to initialise
2164 * @page: the page to use for this fragment
2165 * @off: the offset to the data with @page
2166 * @size: the length of the data
2168 * Initialises the @i'th fragment of @skb to point to &size bytes at
2169 * offset @off within @page.
2171 * Does not take any additional reference on the fragment.
2173 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2174 struct page *page, int off, int size)
2176 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2179 * Propagate page pfmemalloc to the skb if we can. The problem is
2180 * that not all callers have unique ownership of the page but rely
2181 * on page_is_pfmemalloc doing the right thing(tm).
2183 frag->bv_page = page;
2184 frag->bv_offset = off;
2185 skb_frag_size_set(frag, size);
2187 page = compound_head(page);
2188 if (page_is_pfmemalloc(page))
2189 skb->pfmemalloc = true;
2193 * skb_fill_page_desc - initialise a paged fragment in an skb
2194 * @skb: buffer containing fragment to be initialised
2195 * @i: paged fragment index to initialise
2196 * @page: the page to use for this fragment
2197 * @off: the offset to the data with @page
2198 * @size: the length of the data
2200 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2201 * @skb to point to @size bytes at offset @off within @page. In
2202 * addition updates @skb such that @i is the last fragment.
2204 * Does not take any additional reference on the fragment.
2206 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2207 struct page *page, int off, int size)
2209 __skb_fill_page_desc(skb, i, page, off, size);
2210 skb_shinfo(skb)->nr_frags = i + 1;
2213 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2214 int size, unsigned int truesize);
2216 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2217 unsigned int truesize);
2219 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2221 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2222 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2224 return skb->head + skb->tail;
2227 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2229 skb->tail = skb->data - skb->head;
2232 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2234 skb_reset_tail_pointer(skb);
2235 skb->tail += offset;
2238 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2239 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2244 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2246 skb->tail = skb->data;
2249 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2251 skb->tail = skb->data + offset;
2254 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2257 * Add data to an sk_buff
2259 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2260 void *skb_put(struct sk_buff *skb, unsigned int len);
2261 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2263 void *tmp = skb_tail_pointer(skb);
2264 SKB_LINEAR_ASSERT(skb);
2270 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2272 void *tmp = __skb_put(skb, len);
2274 memset(tmp, 0, len);
2278 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2281 void *tmp = __skb_put(skb, len);
2283 memcpy(tmp, data, len);
2287 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2289 *(u8 *)__skb_put(skb, 1) = val;
2292 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2294 void *tmp = skb_put(skb, len);
2296 memset(tmp, 0, len);
2301 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2304 void *tmp = skb_put(skb, len);
2306 memcpy(tmp, data, len);
2311 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2313 *(u8 *)skb_put(skb, 1) = val;
2316 void *skb_push(struct sk_buff *skb, unsigned int len);
2317 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2324 void *skb_pull(struct sk_buff *skb, unsigned int len);
2325 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2328 BUG_ON(skb->len < skb->data_len);
2329 return skb->data += len;
2332 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2334 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2337 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2339 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2341 if (len > skb_headlen(skb) &&
2342 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2345 return skb->data += len;
2348 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2350 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2353 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2355 if (likely(len <= skb_headlen(skb)))
2357 if (unlikely(len > skb->len))
2359 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2362 void skb_condense(struct sk_buff *skb);
2365 * skb_headroom - bytes at buffer head
2366 * @skb: buffer to check
2368 * Return the number of bytes of free space at the head of an &sk_buff.
2370 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2372 return skb->data - skb->head;
2376 * skb_tailroom - bytes at buffer end
2377 * @skb: buffer to check
2379 * Return the number of bytes of free space at the tail of an sk_buff
2381 static inline int skb_tailroom(const struct sk_buff *skb)
2383 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2387 * skb_availroom - bytes at buffer end
2388 * @skb: buffer to check
2390 * Return the number of bytes of free space at the tail of an sk_buff
2391 * allocated by sk_stream_alloc()
2393 static inline int skb_availroom(const struct sk_buff *skb)
2395 if (skb_is_nonlinear(skb))
2398 return skb->end - skb->tail - skb->reserved_tailroom;
2402 * skb_reserve - adjust headroom
2403 * @skb: buffer to alter
2404 * @len: bytes to move
2406 * Increase the headroom of an empty &sk_buff by reducing the tail
2407 * room. This is only allowed for an empty buffer.
2409 static inline void skb_reserve(struct sk_buff *skb, int len)
2416 * skb_tailroom_reserve - adjust reserved_tailroom
2417 * @skb: buffer to alter
2418 * @mtu: maximum amount of headlen permitted
2419 * @needed_tailroom: minimum amount of reserved_tailroom
2421 * Set reserved_tailroom so that headlen can be as large as possible but
2422 * not larger than mtu and tailroom cannot be smaller than
2424 * The required headroom should already have been reserved before using
2427 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2428 unsigned int needed_tailroom)
2430 SKB_LINEAR_ASSERT(skb);
2431 if (mtu < skb_tailroom(skb) - needed_tailroom)
2432 /* use at most mtu */
2433 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2435 /* use up to all available space */
2436 skb->reserved_tailroom = needed_tailroom;
2439 #define ENCAP_TYPE_ETHER 0
2440 #define ENCAP_TYPE_IPPROTO 1
2442 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2445 skb->inner_protocol = protocol;
2446 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2449 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2452 skb->inner_ipproto = ipproto;
2453 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2456 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2458 skb->inner_mac_header = skb->mac_header;
2459 skb->inner_network_header = skb->network_header;
2460 skb->inner_transport_header = skb->transport_header;
2463 static inline void skb_reset_mac_len(struct sk_buff *skb)
2465 skb->mac_len = skb->network_header - skb->mac_header;
2468 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2471 return skb->head + skb->inner_transport_header;
2474 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2476 return skb_inner_transport_header(skb) - skb->data;
2479 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2481 skb->inner_transport_header = skb->data - skb->head;
2484 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2487 skb_reset_inner_transport_header(skb);
2488 skb->inner_transport_header += offset;
2491 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2493 return skb->head + skb->inner_network_header;
2496 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2498 skb->inner_network_header = skb->data - skb->head;
2501 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2504 skb_reset_inner_network_header(skb);
2505 skb->inner_network_header += offset;
2508 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2510 return skb->head + skb->inner_mac_header;
2513 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2515 skb->inner_mac_header = skb->data - skb->head;
2518 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2521 skb_reset_inner_mac_header(skb);
2522 skb->inner_mac_header += offset;
2524 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2526 return skb->transport_header != (typeof(skb->transport_header))~0U;
2529 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2531 return skb->head + skb->transport_header;
2534 static inline void skb_reset_transport_header(struct sk_buff *skb)
2536 skb->transport_header = skb->data - skb->head;
2539 static inline void skb_set_transport_header(struct sk_buff *skb,
2542 skb_reset_transport_header(skb);
2543 skb->transport_header += offset;
2546 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2548 return skb->head + skb->network_header;
2551 static inline void skb_reset_network_header(struct sk_buff *skb)
2553 skb->network_header = skb->data - skb->head;
2556 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2558 skb_reset_network_header(skb);
2559 skb->network_header += offset;
2562 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2564 return skb->head + skb->mac_header;
2567 static inline int skb_mac_offset(const struct sk_buff *skb)
2569 return skb_mac_header(skb) - skb->data;
2572 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2574 return skb->network_header - skb->mac_header;
2577 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2579 return skb->mac_header != (typeof(skb->mac_header))~0U;
2582 static inline void skb_unset_mac_header(struct sk_buff *skb)
2584 skb->mac_header = (typeof(skb->mac_header))~0U;
2587 static inline void skb_reset_mac_header(struct sk_buff *skb)
2589 skb->mac_header = skb->data - skb->head;
2592 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2594 skb_reset_mac_header(skb);
2595 skb->mac_header += offset;
2598 static inline void skb_pop_mac_header(struct sk_buff *skb)
2600 skb->mac_header = skb->network_header;
2603 static inline void skb_probe_transport_header(struct sk_buff *skb)
2605 struct flow_keys_basic keys;
2607 if (skb_transport_header_was_set(skb))
2610 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2612 skb_set_transport_header(skb, keys.control.thoff);
2615 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2617 if (skb_mac_header_was_set(skb)) {
2618 const unsigned char *old_mac = skb_mac_header(skb);
2620 skb_set_mac_header(skb, -skb->mac_len);
2621 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2625 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2627 return skb->csum_start - skb_headroom(skb);
2630 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2632 return skb->head + skb->csum_start;
2635 static inline int skb_transport_offset(const struct sk_buff *skb)
2637 return skb_transport_header(skb) - skb->data;
2640 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2642 return skb->transport_header - skb->network_header;
2645 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2647 return skb->inner_transport_header - skb->inner_network_header;
2650 static inline int skb_network_offset(const struct sk_buff *skb)
2652 return skb_network_header(skb) - skb->data;
2655 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2657 return skb_inner_network_header(skb) - skb->data;
2660 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2662 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2666 * CPUs often take a performance hit when accessing unaligned memory
2667 * locations. The actual performance hit varies, it can be small if the
2668 * hardware handles it or large if we have to take an exception and fix it
2671 * Since an ethernet header is 14 bytes network drivers often end up with
2672 * the IP header at an unaligned offset. The IP header can be aligned by
2673 * shifting the start of the packet by 2 bytes. Drivers should do this
2676 * skb_reserve(skb, NET_IP_ALIGN);
2678 * The downside to this alignment of the IP header is that the DMA is now
2679 * unaligned. On some architectures the cost of an unaligned DMA is high
2680 * and this cost outweighs the gains made by aligning the IP header.
2682 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2685 #ifndef NET_IP_ALIGN
2686 #define NET_IP_ALIGN 2
2690 * The networking layer reserves some headroom in skb data (via
2691 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2692 * the header has to grow. In the default case, if the header has to grow
2693 * 32 bytes or less we avoid the reallocation.
2695 * Unfortunately this headroom changes the DMA alignment of the resulting
2696 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2697 * on some architectures. An architecture can override this value,
2698 * perhaps setting it to a cacheline in size (since that will maintain
2699 * cacheline alignment of the DMA). It must be a power of 2.
2701 * Various parts of the networking layer expect at least 32 bytes of
2702 * headroom, you should not reduce this.
2704 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2705 * to reduce average number of cache lines per packet.
2706 * get_rps_cpu() for example only access one 64 bytes aligned block :
2707 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2710 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2713 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2715 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2717 if (WARN_ON(skb_is_nonlinear(skb)))
2720 skb_set_tail_pointer(skb, len);
2723 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2725 __skb_set_length(skb, len);
2728 void skb_trim(struct sk_buff *skb, unsigned int len);
2730 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2733 return ___pskb_trim(skb, len);
2734 __skb_trim(skb, len);
2738 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2740 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2744 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2745 * @skb: buffer to alter
2748 * This is identical to pskb_trim except that the caller knows that
2749 * the skb is not cloned so we should never get an error due to out-
2752 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2754 int err = pskb_trim(skb, len);
2758 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2760 unsigned int diff = len - skb->len;
2762 if (skb_tailroom(skb) < diff) {
2763 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2768 __skb_set_length(skb, len);
2773 * skb_orphan - orphan a buffer
2774 * @skb: buffer to orphan
2776 * If a buffer currently has an owner then we call the owner's
2777 * destructor function and make the @skb unowned. The buffer continues
2778 * to exist but is no longer charged to its former owner.
2780 static inline void skb_orphan(struct sk_buff *skb)
2782 if (skb->destructor) {
2783 skb->destructor(skb);
2784 skb->destructor = NULL;
2792 * skb_orphan_frags - orphan the frags contained in a buffer
2793 * @skb: buffer to orphan frags from
2794 * @gfp_mask: allocation mask for replacement pages
2796 * For each frag in the SKB which needs a destructor (i.e. has an
2797 * owner) create a copy of that frag and release the original
2798 * page by calling the destructor.
2800 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2802 if (likely(!skb_zcopy(skb)))
2804 if (!skb_zcopy_is_nouarg(skb) &&
2805 skb_uarg(skb)->callback == msg_zerocopy_callback)
2807 return skb_copy_ubufs(skb, gfp_mask);
2810 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2811 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2813 if (likely(!skb_zcopy(skb)))
2815 return skb_copy_ubufs(skb, gfp_mask);
2819 * __skb_queue_purge - empty a list
2820 * @list: list to empty
2822 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2823 * the list and one reference dropped. This function does not take the
2824 * list lock and the caller must hold the relevant locks to use it.
2826 static inline void __skb_queue_purge(struct sk_buff_head *list)
2828 struct sk_buff *skb;
2829 while ((skb = __skb_dequeue(list)) != NULL)
2832 void skb_queue_purge(struct sk_buff_head *list);
2834 unsigned int skb_rbtree_purge(struct rb_root *root);
2836 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2839 * netdev_alloc_frag - allocate a page fragment
2840 * @fragsz: fragment size
2842 * Allocates a frag from a page for receive buffer.
2843 * Uses GFP_ATOMIC allocations.
2845 static inline void *netdev_alloc_frag(unsigned int fragsz)
2847 return __netdev_alloc_frag_align(fragsz, ~0u);
2850 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2853 WARN_ON_ONCE(!is_power_of_2(align));
2854 return __netdev_alloc_frag_align(fragsz, -align);
2857 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2861 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2862 * @dev: network device to receive on
2863 * @length: length to allocate
2865 * Allocate a new &sk_buff and assign it a usage count of one. The
2866 * buffer has unspecified headroom built in. Users should allocate
2867 * the headroom they think they need without accounting for the
2868 * built in space. The built in space is used for optimisations.
2870 * %NULL is returned if there is no free memory. Although this function
2871 * allocates memory it can be called from an interrupt.
2873 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2874 unsigned int length)
2876 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2879 /* legacy helper around __netdev_alloc_skb() */
2880 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2883 return __netdev_alloc_skb(NULL, length, gfp_mask);
2886 /* legacy helper around netdev_alloc_skb() */
2887 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2889 return netdev_alloc_skb(NULL, length);
2893 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2894 unsigned int length, gfp_t gfp)
2896 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2898 if (NET_IP_ALIGN && skb)
2899 skb_reserve(skb, NET_IP_ALIGN);
2903 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2904 unsigned int length)
2906 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2909 static inline void skb_free_frag(void *addr)
2911 page_frag_free(addr);
2914 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2916 static inline void *napi_alloc_frag(unsigned int fragsz)
2918 return __napi_alloc_frag_align(fragsz, ~0u);
2921 static inline void *napi_alloc_frag_align(unsigned int fragsz,
2924 WARN_ON_ONCE(!is_power_of_2(align));
2925 return __napi_alloc_frag_align(fragsz, -align);
2928 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2929 unsigned int length, gfp_t gfp_mask);
2930 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2931 unsigned int length)
2933 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2935 void napi_consume_skb(struct sk_buff *skb, int budget);
2937 void napi_skb_free_stolen_head(struct sk_buff *skb);
2938 void __kfree_skb_defer(struct sk_buff *skb);
2941 * __dev_alloc_pages - allocate page for network Rx
2942 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2943 * @order: size of the allocation
2945 * Allocate a new page.
2947 * %NULL is returned if there is no free memory.
2949 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2952 /* This piece of code contains several assumptions.
2953 * 1. This is for device Rx, therefor a cold page is preferred.
2954 * 2. The expectation is the user wants a compound page.
2955 * 3. If requesting a order 0 page it will not be compound
2956 * due to the check to see if order has a value in prep_new_page
2957 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2958 * code in gfp_to_alloc_flags that should be enforcing this.
2960 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2962 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2965 static inline struct page *dev_alloc_pages(unsigned int order)
2967 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2971 * __dev_alloc_page - allocate a page for network Rx
2972 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2974 * Allocate a new page.
2976 * %NULL is returned if there is no free memory.
2978 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2980 return __dev_alloc_pages(gfp_mask, 0);
2983 static inline struct page *dev_alloc_page(void)
2985 return dev_alloc_pages(0);
2989 * dev_page_is_reusable - check whether a page can be reused for network Rx
2990 * @page: the page to test
2992 * A page shouldn't be considered for reusing/recycling if it was allocated
2993 * under memory pressure or at a distant memory node.
2995 * Returns false if this page should be returned to page allocator, true
2998 static inline bool dev_page_is_reusable(const struct page *page)
3000 return likely(page_to_nid(page) == numa_mem_id() &&
3001 !page_is_pfmemalloc(page));
3005 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3006 * @page: The page that was allocated from skb_alloc_page
3007 * @skb: The skb that may need pfmemalloc set
3009 static inline void skb_propagate_pfmemalloc(const struct page *page,
3010 struct sk_buff *skb)
3012 if (page_is_pfmemalloc(page))
3013 skb->pfmemalloc = true;
3017 * skb_frag_off() - Returns the offset of a skb fragment
3018 * @frag: the paged fragment
3020 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3022 return frag->bv_offset;
3026 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3027 * @frag: skb fragment
3028 * @delta: value to add
3030 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3032 frag->bv_offset += delta;
3036 * skb_frag_off_set() - Sets the offset of a skb fragment
3037 * @frag: skb fragment
3038 * @offset: offset of fragment
3040 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3042 frag->bv_offset = offset;
3046 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3047 * @fragto: skb fragment where offset is set
3048 * @fragfrom: skb fragment offset is copied from
3050 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3051 const skb_frag_t *fragfrom)
3053 fragto->bv_offset = fragfrom->bv_offset;
3057 * skb_frag_page - retrieve the page referred to by a paged fragment
3058 * @frag: the paged fragment
3060 * Returns the &struct page associated with @frag.
3062 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3064 return frag->bv_page;
3068 * __skb_frag_ref - take an addition reference on a paged fragment.
3069 * @frag: the paged fragment
3071 * Takes an additional reference on the paged fragment @frag.
3073 static inline void __skb_frag_ref(skb_frag_t *frag)
3075 get_page(skb_frag_page(frag));
3079 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3081 * @f: the fragment offset.
3083 * Takes an additional reference on the @f'th paged fragment of @skb.
3085 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3087 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3091 * __skb_frag_unref - release a reference on a paged fragment.
3092 * @frag: the paged fragment
3093 * @recycle: recycle the page if allocated via page_pool
3095 * Releases a reference on the paged fragment @frag
3096 * or recycles the page via the page_pool API.
3098 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3100 struct page *page = skb_frag_page(frag);
3102 #ifdef CONFIG_PAGE_POOL
3103 if (recycle && page_pool_return_skb_page(page))
3110 * skb_frag_unref - release a reference on a paged fragment of an skb.
3112 * @f: the fragment offset
3114 * Releases a reference on the @f'th paged fragment of @skb.
3116 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3118 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3122 * skb_frag_address - gets the address of the data contained in a paged fragment
3123 * @frag: the paged fragment buffer
3125 * Returns the address of the data within @frag. The page must already
3128 static inline void *skb_frag_address(const skb_frag_t *frag)
3130 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3134 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3135 * @frag: the paged fragment buffer
3137 * Returns the address of the data within @frag. Checks that the page
3138 * is mapped and returns %NULL otherwise.
3140 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3142 void *ptr = page_address(skb_frag_page(frag));
3146 return ptr + skb_frag_off(frag);
3150 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3151 * @fragto: skb fragment where page is set
3152 * @fragfrom: skb fragment page is copied from
3154 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3155 const skb_frag_t *fragfrom)
3157 fragto->bv_page = fragfrom->bv_page;
3161 * __skb_frag_set_page - sets the page contained in a paged fragment
3162 * @frag: the paged fragment
3163 * @page: the page to set
3165 * Sets the fragment @frag to contain @page.
3167 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3169 frag->bv_page = page;
3173 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3175 * @f: the fragment offset
3176 * @page: the page to set
3178 * Sets the @f'th fragment of @skb to contain @page.
3180 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3183 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3186 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3189 * skb_frag_dma_map - maps a paged fragment via the DMA API
3190 * @dev: the device to map the fragment to
3191 * @frag: the paged fragment to map
3192 * @offset: the offset within the fragment (starting at the
3193 * fragment's own offset)
3194 * @size: the number of bytes to map
3195 * @dir: the direction of the mapping (``PCI_DMA_*``)
3197 * Maps the page associated with @frag to @device.
3199 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3200 const skb_frag_t *frag,
3201 size_t offset, size_t size,
3202 enum dma_data_direction dir)
3204 return dma_map_page(dev, skb_frag_page(frag),
3205 skb_frag_off(frag) + offset, size, dir);
3208 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3211 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3215 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3218 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3223 * skb_clone_writable - is the header of a clone writable
3224 * @skb: buffer to check
3225 * @len: length up to which to write
3227 * Returns true if modifying the header part of the cloned buffer
3228 * does not requires the data to be copied.
3230 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3232 return !skb_header_cloned(skb) &&
3233 skb_headroom(skb) + len <= skb->hdr_len;
3236 static inline int skb_try_make_writable(struct sk_buff *skb,
3237 unsigned int write_len)
3239 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3240 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3243 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3248 if (headroom > skb_headroom(skb))
3249 delta = headroom - skb_headroom(skb);
3251 if (delta || cloned)
3252 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3258 * skb_cow - copy header of skb when it is required
3259 * @skb: buffer to cow
3260 * @headroom: needed headroom
3262 * If the skb passed lacks sufficient headroom or its data part
3263 * is shared, data is reallocated. If reallocation fails, an error
3264 * is returned and original skb is not changed.
3266 * The result is skb with writable area skb->head...skb->tail
3267 * and at least @headroom of space at head.
3269 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3271 return __skb_cow(skb, headroom, skb_cloned(skb));
3275 * skb_cow_head - skb_cow but only making the head writable
3276 * @skb: buffer to cow
3277 * @headroom: needed headroom
3279 * This function is identical to skb_cow except that we replace the
3280 * skb_cloned check by skb_header_cloned. It should be used when
3281 * you only need to push on some header and do not need to modify
3284 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3286 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3290 * skb_padto - pad an skbuff up to a minimal size
3291 * @skb: buffer to pad
3292 * @len: minimal length
3294 * Pads up a buffer to ensure the trailing bytes exist and are
3295 * blanked. If the buffer already contains sufficient data it
3296 * is untouched. Otherwise it is extended. Returns zero on
3297 * success. The skb is freed on error.
3299 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3301 unsigned int size = skb->len;
3302 if (likely(size >= len))
3304 return skb_pad(skb, len - size);
3308 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3309 * @skb: buffer to pad
3310 * @len: minimal length
3311 * @free_on_error: free buffer on error
3313 * Pads up a buffer to ensure the trailing bytes exist and are
3314 * blanked. If the buffer already contains sufficient data it
3315 * is untouched. Otherwise it is extended. Returns zero on
3316 * success. The skb is freed on error if @free_on_error is true.
3318 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3322 unsigned int size = skb->len;
3324 if (unlikely(size < len)) {
3326 if (__skb_pad(skb, len, free_on_error))
3328 __skb_put(skb, len);
3334 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3335 * @skb: buffer to pad
3336 * @len: minimal length
3338 * Pads up a buffer to ensure the trailing bytes exist and are
3339 * blanked. If the buffer already contains sufficient data it
3340 * is untouched. Otherwise it is extended. Returns zero on
3341 * success. The skb is freed on error.
3343 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3345 return __skb_put_padto(skb, len, true);
3348 static inline int skb_add_data(struct sk_buff *skb,
3349 struct iov_iter *from, int copy)
3351 const int off = skb->len;
3353 if (skb->ip_summed == CHECKSUM_NONE) {
3355 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3357 skb->csum = csum_block_add(skb->csum, csum, off);
3360 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3363 __skb_trim(skb, off);
3367 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3368 const struct page *page, int off)
3373 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3375 return page == skb_frag_page(frag) &&
3376 off == skb_frag_off(frag) + skb_frag_size(frag);
3381 static inline int __skb_linearize(struct sk_buff *skb)
3383 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3387 * skb_linearize - convert paged skb to linear one
3388 * @skb: buffer to linarize
3390 * If there is no free memory -ENOMEM is returned, otherwise zero
3391 * is returned and the old skb data released.
3393 static inline int skb_linearize(struct sk_buff *skb)
3395 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3399 * skb_has_shared_frag - can any frag be overwritten
3400 * @skb: buffer to test
3402 * Return true if the skb has at least one frag that might be modified
3403 * by an external entity (as in vmsplice()/sendfile())
3405 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3407 return skb_is_nonlinear(skb) &&
3408 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3412 * skb_linearize_cow - make sure skb is linear and writable
3413 * @skb: buffer to process
3415 * If there is no free memory -ENOMEM is returned, otherwise zero
3416 * is returned and the old skb data released.
3418 static inline int skb_linearize_cow(struct sk_buff *skb)
3420 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3421 __skb_linearize(skb) : 0;
3424 static __always_inline void
3425 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3428 if (skb->ip_summed == CHECKSUM_COMPLETE)
3429 skb->csum = csum_block_sub(skb->csum,
3430 csum_partial(start, len, 0), off);
3431 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3432 skb_checksum_start_offset(skb) < 0)
3433 skb->ip_summed = CHECKSUM_NONE;
3437 * skb_postpull_rcsum - update checksum for received skb after pull
3438 * @skb: buffer to update
3439 * @start: start of data before pull
3440 * @len: length of data pulled
3442 * After doing a pull on a received packet, you need to call this to
3443 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3444 * CHECKSUM_NONE so that it can be recomputed from scratch.
3446 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3447 const void *start, unsigned int len)
3449 __skb_postpull_rcsum(skb, start, len, 0);
3452 static __always_inline void
3453 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3456 if (skb->ip_summed == CHECKSUM_COMPLETE)
3457 skb->csum = csum_block_add(skb->csum,
3458 csum_partial(start, len, 0), off);
3462 * skb_postpush_rcsum - update checksum for received skb after push
3463 * @skb: buffer to update
3464 * @start: start of data after push
3465 * @len: length of data pushed
3467 * After doing a push on a received packet, you need to call this to
3468 * update the CHECKSUM_COMPLETE checksum.
3470 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3471 const void *start, unsigned int len)
3473 __skb_postpush_rcsum(skb, start, len, 0);
3476 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3479 * skb_push_rcsum - push skb and update receive checksum
3480 * @skb: buffer to update
3481 * @len: length of data pulled
3483 * This function performs an skb_push on the packet and updates
3484 * the CHECKSUM_COMPLETE checksum. It should be used on
3485 * receive path processing instead of skb_push unless you know
3486 * that the checksum difference is zero (e.g., a valid IP header)
3487 * or you are setting ip_summed to CHECKSUM_NONE.
3489 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3492 skb_postpush_rcsum(skb, skb->data, len);
3496 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3498 * pskb_trim_rcsum - trim received skb and update checksum
3499 * @skb: buffer to trim
3502 * This is exactly the same as pskb_trim except that it ensures the
3503 * checksum of received packets are still valid after the operation.
3504 * It can change skb pointers.
3507 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3509 if (likely(len >= skb->len))
3511 return pskb_trim_rcsum_slow(skb, len);
3514 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3516 if (skb->ip_summed == CHECKSUM_COMPLETE)
3517 skb->ip_summed = CHECKSUM_NONE;
3518 __skb_trim(skb, len);
3522 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3524 if (skb->ip_summed == CHECKSUM_COMPLETE)
3525 skb->ip_summed = CHECKSUM_NONE;
3526 return __skb_grow(skb, len);
3529 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3530 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3531 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3532 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3533 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3535 #define skb_queue_walk(queue, skb) \
3536 for (skb = (queue)->next; \
3537 skb != (struct sk_buff *)(queue); \
3540 #define skb_queue_walk_safe(queue, skb, tmp) \
3541 for (skb = (queue)->next, tmp = skb->next; \
3542 skb != (struct sk_buff *)(queue); \
3543 skb = tmp, tmp = skb->next)
3545 #define skb_queue_walk_from(queue, skb) \
3546 for (; skb != (struct sk_buff *)(queue); \
3549 #define skb_rbtree_walk(skb, root) \
3550 for (skb = skb_rb_first(root); skb != NULL; \
3551 skb = skb_rb_next(skb))
3553 #define skb_rbtree_walk_from(skb) \
3554 for (; skb != NULL; \
3555 skb = skb_rb_next(skb))
3557 #define skb_rbtree_walk_from_safe(skb, tmp) \
3558 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3561 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3562 for (tmp = skb->next; \
3563 skb != (struct sk_buff *)(queue); \
3564 skb = tmp, tmp = skb->next)
3566 #define skb_queue_reverse_walk(queue, skb) \
3567 for (skb = (queue)->prev; \
3568 skb != (struct sk_buff *)(queue); \
3571 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3572 for (skb = (queue)->prev, tmp = skb->prev; \
3573 skb != (struct sk_buff *)(queue); \
3574 skb = tmp, tmp = skb->prev)
3576 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3577 for (tmp = skb->prev; \
3578 skb != (struct sk_buff *)(queue); \
3579 skb = tmp, tmp = skb->prev)
3581 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3583 return skb_shinfo(skb)->frag_list != NULL;
3586 static inline void skb_frag_list_init(struct sk_buff *skb)
3588 skb_shinfo(skb)->frag_list = NULL;
3591 #define skb_walk_frags(skb, iter) \
3592 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3595 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3596 int *err, long *timeo_p,
3597 const struct sk_buff *skb);
3598 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3599 struct sk_buff_head *queue,
3602 struct sk_buff **last);
3603 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3604 struct sk_buff_head *queue,
3605 unsigned int flags, int *off, int *err,
3606 struct sk_buff **last);
3607 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3608 struct sk_buff_head *sk_queue,
3609 unsigned int flags, int *off, int *err);
3610 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3612 __poll_t datagram_poll(struct file *file, struct socket *sock,
3613 struct poll_table_struct *wait);
3614 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3615 struct iov_iter *to, int size);
3616 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3617 struct msghdr *msg, int size)
3619 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3621 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3622 struct msghdr *msg);
3623 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3624 struct iov_iter *to, int len,
3625 struct ahash_request *hash);
3626 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3627 struct iov_iter *from, int len);
3628 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3629 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3630 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3631 static inline void skb_free_datagram_locked(struct sock *sk,
3632 struct sk_buff *skb)
3634 __skb_free_datagram_locked(sk, skb, 0);
3636 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3637 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3638 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3639 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3641 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3642 struct pipe_inode_info *pipe, unsigned int len,
3643 unsigned int flags);
3644 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3646 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3647 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3648 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3649 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3651 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3652 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3653 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3654 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3655 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3656 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3657 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3658 unsigned int offset);
3659 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3660 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3661 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3662 int skb_vlan_pop(struct sk_buff *skb);
3663 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3664 int skb_eth_pop(struct sk_buff *skb);
3665 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3666 const unsigned char *src);
3667 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3668 int mac_len, bool ethernet);
3669 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3671 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3672 int skb_mpls_dec_ttl(struct sk_buff *skb);
3673 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3676 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3678 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3681 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3683 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3686 struct skb_checksum_ops {
3687 __wsum (*update)(const void *mem, int len, __wsum wsum);
3688 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3691 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3693 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3694 __wsum csum, const struct skb_checksum_ops *ops);
3695 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3698 static inline void * __must_check
3699 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3700 const void *data, int hlen, void *buffer)
3702 if (likely(hlen - offset >= len))
3703 return (void *)data + offset;
3705 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3711 static inline void * __must_check
3712 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3714 return __skb_header_pointer(skb, offset, len, skb->data,
3715 skb_headlen(skb), buffer);
3719 * skb_needs_linearize - check if we need to linearize a given skb
3720 * depending on the given device features.
3721 * @skb: socket buffer to check
3722 * @features: net device features
3724 * Returns true if either:
3725 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3726 * 2. skb is fragmented and the device does not support SG.
3728 static inline bool skb_needs_linearize(struct sk_buff *skb,
3729 netdev_features_t features)
3731 return skb_is_nonlinear(skb) &&
3732 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3733 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3736 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3738 const unsigned int len)
3740 memcpy(to, skb->data, len);
3743 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3744 const int offset, void *to,
3745 const unsigned int len)
3747 memcpy(to, skb->data + offset, len);
3750 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3752 const unsigned int len)
3754 memcpy(skb->data, from, len);
3757 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3760 const unsigned int len)
3762 memcpy(skb->data + offset, from, len);
3765 void skb_init(void);
3767 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3773 * skb_get_timestamp - get timestamp from a skb
3774 * @skb: skb to get stamp from
3775 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3777 * Timestamps are stored in the skb as offsets to a base timestamp.
3778 * This function converts the offset back to a struct timeval and stores
3781 static inline void skb_get_timestamp(const struct sk_buff *skb,
3782 struct __kernel_old_timeval *stamp)
3784 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3787 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3788 struct __kernel_sock_timeval *stamp)
3790 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3792 stamp->tv_sec = ts.tv_sec;
3793 stamp->tv_usec = ts.tv_nsec / 1000;
3796 static inline void skb_get_timestampns(const struct sk_buff *skb,
3797 struct __kernel_old_timespec *stamp)
3799 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3801 stamp->tv_sec = ts.tv_sec;
3802 stamp->tv_nsec = ts.tv_nsec;
3805 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3806 struct __kernel_timespec *stamp)
3808 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3810 stamp->tv_sec = ts.tv_sec;
3811 stamp->tv_nsec = ts.tv_nsec;
3814 static inline void __net_timestamp(struct sk_buff *skb)
3816 skb->tstamp = ktime_get_real();
3819 static inline ktime_t net_timedelta(ktime_t t)
3821 return ktime_sub(ktime_get_real(), t);
3824 static inline ktime_t net_invalid_timestamp(void)
3829 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3831 return skb_shinfo(skb)->meta_len;
3834 static inline void *skb_metadata_end(const struct sk_buff *skb)
3836 return skb_mac_header(skb);
3839 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3840 const struct sk_buff *skb_b,
3843 const void *a = skb_metadata_end(skb_a);
3844 const void *b = skb_metadata_end(skb_b);
3845 /* Using more efficient varaiant than plain call to memcmp(). */
3846 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3850 #define __it(x, op) (x -= sizeof(u##op))
3851 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3852 case 32: diffs |= __it_diff(a, b, 64);
3854 case 24: diffs |= __it_diff(a, b, 64);
3856 case 16: diffs |= __it_diff(a, b, 64);
3858 case 8: diffs |= __it_diff(a, b, 64);
3860 case 28: diffs |= __it_diff(a, b, 64);
3862 case 20: diffs |= __it_diff(a, b, 64);
3864 case 12: diffs |= __it_diff(a, b, 64);
3866 case 4: diffs |= __it_diff(a, b, 32);
3871 return memcmp(a - meta_len, b - meta_len, meta_len);
3875 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3876 const struct sk_buff *skb_b)
3878 u8 len_a = skb_metadata_len(skb_a);
3879 u8 len_b = skb_metadata_len(skb_b);
3881 if (!(len_a | len_b))
3884 return len_a != len_b ?
3885 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3888 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3890 skb_shinfo(skb)->meta_len = meta_len;
3893 static inline void skb_metadata_clear(struct sk_buff *skb)
3895 skb_metadata_set(skb, 0);
3898 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3900 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3902 void skb_clone_tx_timestamp(struct sk_buff *skb);
3903 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3905 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3907 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3911 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3916 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3919 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3921 * PHY drivers may accept clones of transmitted packets for
3922 * timestamping via their phy_driver.txtstamp method. These drivers
3923 * must call this function to return the skb back to the stack with a
3926 * @skb: clone of the original outgoing packet
3927 * @hwtstamps: hardware time stamps
3930 void skb_complete_tx_timestamp(struct sk_buff *skb,
3931 struct skb_shared_hwtstamps *hwtstamps);
3933 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
3934 struct skb_shared_hwtstamps *hwtstamps,
3935 struct sock *sk, int tstype);
3938 * skb_tstamp_tx - queue clone of skb with send time stamps
3939 * @orig_skb: the original outgoing packet
3940 * @hwtstamps: hardware time stamps, may be NULL if not available
3942 * If the skb has a socket associated, then this function clones the
3943 * skb (thus sharing the actual data and optional structures), stores
3944 * the optional hardware time stamping information (if non NULL) or
3945 * generates a software time stamp (otherwise), then queues the clone
3946 * to the error queue of the socket. Errors are silently ignored.
3948 void skb_tstamp_tx(struct sk_buff *orig_skb,
3949 struct skb_shared_hwtstamps *hwtstamps);
3952 * skb_tx_timestamp() - Driver hook for transmit timestamping
3954 * Ethernet MAC Drivers should call this function in their hard_xmit()
3955 * function immediately before giving the sk_buff to the MAC hardware.
3957 * Specifically, one should make absolutely sure that this function is
3958 * called before TX completion of this packet can trigger. Otherwise
3959 * the packet could potentially already be freed.
3961 * @skb: A socket buffer.
3963 static inline void skb_tx_timestamp(struct sk_buff *skb)
3965 skb_clone_tx_timestamp(skb);
3966 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3967 skb_tstamp_tx(skb, NULL);
3971 * skb_complete_wifi_ack - deliver skb with wifi status
3973 * @skb: the original outgoing packet
3974 * @acked: ack status
3977 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3979 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3980 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3982 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3984 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3986 (skb->ip_summed == CHECKSUM_PARTIAL &&
3987 skb_checksum_start_offset(skb) >= 0));
3991 * skb_checksum_complete - Calculate checksum of an entire packet
3992 * @skb: packet to process
3994 * This function calculates the checksum over the entire packet plus
3995 * the value of skb->csum. The latter can be used to supply the
3996 * checksum of a pseudo header as used by TCP/UDP. It returns the
3999 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4000 * this function can be used to verify that checksum on received
4001 * packets. In that case the function should return zero if the
4002 * checksum is correct. In particular, this function will return zero
4003 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4004 * hardware has already verified the correctness of the checksum.
4006 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4008 return skb_csum_unnecessary(skb) ?
4009 0 : __skb_checksum_complete(skb);
4012 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4014 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4015 if (skb->csum_level == 0)
4016 skb->ip_summed = CHECKSUM_NONE;
4022 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4024 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4025 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4027 } else if (skb->ip_summed == CHECKSUM_NONE) {
4028 skb->ip_summed = CHECKSUM_UNNECESSARY;
4029 skb->csum_level = 0;
4033 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4035 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4036 skb->ip_summed = CHECKSUM_NONE;
4037 skb->csum_level = 0;
4041 /* Check if we need to perform checksum complete validation.
4043 * Returns true if checksum complete is needed, false otherwise
4044 * (either checksum is unnecessary or zero checksum is allowed).
4046 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4050 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4051 skb->csum_valid = 1;
4052 __skb_decr_checksum_unnecessary(skb);
4059 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4062 #define CHECKSUM_BREAK 76
4064 /* Unset checksum-complete
4066 * Unset checksum complete can be done when packet is being modified
4067 * (uncompressed for instance) and checksum-complete value is
4070 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4072 if (skb->ip_summed == CHECKSUM_COMPLETE)
4073 skb->ip_summed = CHECKSUM_NONE;
4076 /* Validate (init) checksum based on checksum complete.
4079 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4080 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4081 * checksum is stored in skb->csum for use in __skb_checksum_complete
4082 * non-zero: value of invalid checksum
4085 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4089 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4090 if (!csum_fold(csum_add(psum, skb->csum))) {
4091 skb->csum_valid = 1;
4098 if (complete || skb->len <= CHECKSUM_BREAK) {
4101 csum = __skb_checksum_complete(skb);
4102 skb->csum_valid = !csum;
4109 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4114 /* Perform checksum validate (init). Note that this is a macro since we only
4115 * want to calculate the pseudo header which is an input function if necessary.
4116 * First we try to validate without any computation (checksum unnecessary) and
4117 * then calculate based on checksum complete calling the function to compute
4121 * 0: checksum is validated or try to in skb_checksum_complete
4122 * non-zero: value of invalid checksum
4124 #define __skb_checksum_validate(skb, proto, complete, \
4125 zero_okay, check, compute_pseudo) \
4127 __sum16 __ret = 0; \
4128 skb->csum_valid = 0; \
4129 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4130 __ret = __skb_checksum_validate_complete(skb, \
4131 complete, compute_pseudo(skb, proto)); \
4135 #define skb_checksum_init(skb, proto, compute_pseudo) \
4136 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4138 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4139 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4141 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4142 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4144 #define skb_checksum_validate_zero_check(skb, proto, check, \
4146 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4148 #define skb_checksum_simple_validate(skb) \
4149 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4151 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4153 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4156 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4158 skb->csum = ~pseudo;
4159 skb->ip_summed = CHECKSUM_COMPLETE;
4162 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4164 if (__skb_checksum_convert_check(skb)) \
4165 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4168 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4169 u16 start, u16 offset)
4171 skb->ip_summed = CHECKSUM_PARTIAL;
4172 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4173 skb->csum_offset = offset - start;
4176 /* Update skbuf and packet to reflect the remote checksum offload operation.
4177 * When called, ptr indicates the starting point for skb->csum when
4178 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4179 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4181 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4182 int start, int offset, bool nopartial)
4187 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4191 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4192 __skb_checksum_complete(skb);
4193 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4196 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4198 /* Adjust skb->csum since we changed the packet */
4199 skb->csum = csum_add(skb->csum, delta);
4202 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4204 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4205 return (void *)(skb->_nfct & NFCT_PTRMASK);
4211 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4213 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4220 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4222 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4223 skb->slow_gro |= !!nfct;
4228 #ifdef CONFIG_SKB_EXTENSIONS
4230 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4236 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4239 #if IS_ENABLED(CONFIG_MPTCP)
4242 SKB_EXT_NUM, /* must be last */
4246 * struct skb_ext - sk_buff extensions
4247 * @refcnt: 1 on allocation, deallocated on 0
4248 * @offset: offset to add to @data to obtain extension address
4249 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4250 * @data: start of extension data, variable sized
4252 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4253 * to use 'u8' types while allowing up to 2kb worth of extension data.
4257 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4258 u8 chunks; /* same */
4259 char data[] __aligned(8);
4262 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4263 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4264 struct skb_ext *ext);
4265 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4266 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4267 void __skb_ext_put(struct skb_ext *ext);
4269 static inline void skb_ext_put(struct sk_buff *skb)
4271 if (skb->active_extensions)
4272 __skb_ext_put(skb->extensions);
4275 static inline void __skb_ext_copy(struct sk_buff *dst,
4276 const struct sk_buff *src)
4278 dst->active_extensions = src->active_extensions;
4280 if (src->active_extensions) {
4281 struct skb_ext *ext = src->extensions;
4283 refcount_inc(&ext->refcnt);
4284 dst->extensions = ext;
4288 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4291 __skb_ext_copy(dst, src);
4294 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4296 return !!ext->offset[i];
4299 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4301 return skb->active_extensions & (1 << id);
4304 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4306 if (skb_ext_exist(skb, id))
4307 __skb_ext_del(skb, id);
4310 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4312 if (skb_ext_exist(skb, id)) {
4313 struct skb_ext *ext = skb->extensions;
4315 return (void *)ext + (ext->offset[id] << 3);
4321 static inline void skb_ext_reset(struct sk_buff *skb)
4323 if (unlikely(skb->active_extensions)) {
4324 __skb_ext_put(skb->extensions);
4325 skb->active_extensions = 0;
4329 static inline bool skb_has_extensions(struct sk_buff *skb)
4331 return unlikely(skb->active_extensions);
4334 static inline void skb_ext_put(struct sk_buff *skb) {}
4335 static inline void skb_ext_reset(struct sk_buff *skb) {}
4336 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4337 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4338 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4339 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4340 #endif /* CONFIG_SKB_EXTENSIONS */
4342 static inline void nf_reset_ct(struct sk_buff *skb)
4344 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4345 nf_conntrack_put(skb_nfct(skb));
4350 static inline void nf_reset_trace(struct sk_buff *skb)
4352 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4357 static inline void ipvs_reset(struct sk_buff *skb)
4359 #if IS_ENABLED(CONFIG_IP_VS)
4360 skb->ipvs_property = 0;
4364 /* Note: This doesn't put any conntrack info in dst. */
4365 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4368 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4369 dst->_nfct = src->_nfct;
4370 nf_conntrack_get(skb_nfct(src));
4372 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4374 dst->nf_trace = src->nf_trace;
4378 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4380 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4381 nf_conntrack_put(skb_nfct(dst));
4383 dst->slow_gro = src->slow_gro;
4384 __nf_copy(dst, src, true);
4387 #ifdef CONFIG_NETWORK_SECMARK
4388 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4390 to->secmark = from->secmark;
4393 static inline void skb_init_secmark(struct sk_buff *skb)
4398 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4401 static inline void skb_init_secmark(struct sk_buff *skb)
4405 static inline int secpath_exists(const struct sk_buff *skb)
4408 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4414 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4416 return !skb->destructor &&
4417 !secpath_exists(skb) &&
4419 !skb->_skb_refdst &&
4420 !skb_has_frag_list(skb);
4423 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4425 skb->queue_mapping = queue_mapping;
4428 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4430 return skb->queue_mapping;
4433 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4435 to->queue_mapping = from->queue_mapping;
4438 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4440 skb->queue_mapping = rx_queue + 1;
4443 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4445 return skb->queue_mapping - 1;
4448 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4450 return skb->queue_mapping != 0;
4453 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4455 skb->dst_pending_confirm = val;
4458 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4460 return skb->dst_pending_confirm != 0;
4463 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4466 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4472 /* Keeps track of mac header offset relative to skb->head.
4473 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4474 * For non-tunnel skb it points to skb_mac_header() and for
4475 * tunnel skb it points to outer mac header.
4476 * Keeps track of level of encapsulation of network headers.
4487 #define SKB_GSO_CB_OFFSET 32
4488 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4490 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4492 return (skb_mac_header(inner_skb) - inner_skb->head) -
4493 SKB_GSO_CB(inner_skb)->mac_offset;
4496 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4498 int new_headroom, headroom;
4501 headroom = skb_headroom(skb);
4502 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4506 new_headroom = skb_headroom(skb);
4507 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4511 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4513 /* Do not update partial checksums if remote checksum is enabled. */
4514 if (skb->remcsum_offload)
4517 SKB_GSO_CB(skb)->csum = res;
4518 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4521 /* Compute the checksum for a gso segment. First compute the checksum value
4522 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4523 * then add in skb->csum (checksum from csum_start to end of packet).
4524 * skb->csum and csum_start are then updated to reflect the checksum of the
4525 * resultant packet starting from the transport header-- the resultant checksum
4526 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4529 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4531 unsigned char *csum_start = skb_transport_header(skb);
4532 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4533 __wsum partial = SKB_GSO_CB(skb)->csum;
4535 SKB_GSO_CB(skb)->csum = res;
4536 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4538 return csum_fold(csum_partial(csum_start, plen, partial));
4541 static inline bool skb_is_gso(const struct sk_buff *skb)
4543 return skb_shinfo(skb)->gso_size;
4546 /* Note: Should be called only if skb_is_gso(skb) is true */
4547 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4549 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4552 /* Note: Should be called only if skb_is_gso(skb) is true */
4553 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4555 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4558 /* Note: Should be called only if skb_is_gso(skb) is true */
4559 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4561 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4564 static inline void skb_gso_reset(struct sk_buff *skb)
4566 skb_shinfo(skb)->gso_size = 0;
4567 skb_shinfo(skb)->gso_segs = 0;
4568 skb_shinfo(skb)->gso_type = 0;
4571 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4574 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4576 shinfo->gso_size += increment;
4579 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4582 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4584 shinfo->gso_size -= decrement;
4587 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4589 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4591 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4592 * wanted then gso_type will be set. */
4593 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4595 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4596 unlikely(shinfo->gso_type == 0)) {
4597 __skb_warn_lro_forwarding(skb);
4603 static inline void skb_forward_csum(struct sk_buff *skb)
4605 /* Unfortunately we don't support this one. Any brave souls? */
4606 if (skb->ip_summed == CHECKSUM_COMPLETE)
4607 skb->ip_summed = CHECKSUM_NONE;
4611 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4612 * @skb: skb to check
4614 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4615 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4616 * use this helper, to document places where we make this assertion.
4618 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4621 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4625 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4627 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4628 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4629 unsigned int transport_len,
4630 __sum16(*skb_chkf)(struct sk_buff *skb));
4633 * skb_head_is_locked - Determine if the skb->head is locked down
4634 * @skb: skb to check
4636 * The head on skbs build around a head frag can be removed if they are
4637 * not cloned. This function returns true if the skb head is locked down
4638 * due to either being allocated via kmalloc, or by being a clone with
4639 * multiple references to the head.
4641 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4643 return !skb->head_frag || skb_cloned(skb);
4646 /* Local Checksum Offload.
4647 * Compute outer checksum based on the assumption that the
4648 * inner checksum will be offloaded later.
4649 * See Documentation/networking/checksum-offloads.rst for
4650 * explanation of how this works.
4651 * Fill in outer checksum adjustment (e.g. with sum of outer
4652 * pseudo-header) before calling.
4653 * Also ensure that inner checksum is in linear data area.
4655 static inline __wsum lco_csum(struct sk_buff *skb)
4657 unsigned char *csum_start = skb_checksum_start(skb);
4658 unsigned char *l4_hdr = skb_transport_header(skb);
4661 /* Start with complement of inner checksum adjustment */
4662 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4665 /* Add in checksum of our headers (incl. outer checksum
4666 * adjustment filled in by caller) and return result.
4668 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4671 static inline bool skb_is_redirected(const struct sk_buff *skb)
4673 return skb->redirected;
4676 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4678 skb->redirected = 1;
4679 #ifdef CONFIG_NET_REDIRECT
4680 skb->from_ingress = from_ingress;
4681 if (skb->from_ingress)
4686 static inline void skb_reset_redirect(struct sk_buff *skb)
4688 skb->redirected = 0;
4691 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4693 return skb->csum_not_inet;
4696 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4697 const u64 kcov_handle)
4700 skb->kcov_handle = kcov_handle;
4704 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4707 return skb->kcov_handle;
4713 #ifdef CONFIG_PAGE_POOL
4714 static inline void skb_mark_for_recycle(struct sk_buff *skb, struct page *page,
4715 struct page_pool *pp)
4717 skb->pp_recycle = 1;
4718 page_pool_store_mem_info(page, pp);
4722 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4724 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4726 return page_pool_return_skb_page(virt_to_page(data));
4729 #endif /* __KERNEL__ */
4730 #endif /* _LINUX_SKBUFF_H */