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>
39 #include <linux/llist.h>
41 #include <net/page_pool.h>
42 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
43 #include <linux/netfilter/nf_conntrack_common.h>
46 /* The interface for checksum offload between the stack and networking drivers
49 * A. IP checksum related features
51 * Drivers advertise checksum offload capabilities in the features of a device.
52 * From the stack's point of view these are capabilities offered by the driver.
53 * A driver typically only advertises features that it is capable of offloading
56 * The checksum related features are:
58 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
59 * IP (one's complement) checksum for any combination
60 * of protocols or protocol layering. The checksum is
61 * computed and set in a packet per the CHECKSUM_PARTIAL
62 * interface (see below).
64 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
65 * TCP or UDP packets over IPv4. These are specifically
66 * unencapsulated packets of the form IPv4|TCP or
67 * IPv4|UDP where the Protocol field in the IPv4 header
68 * is TCP or UDP. The IPv4 header may contain IP options.
69 * This feature cannot be set in features for a device
70 * with NETIF_F_HW_CSUM also set. This feature is being
71 * DEPRECATED (see below).
73 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
74 * TCP or UDP packets over IPv6. These are specifically
75 * unencapsulated packets of the form IPv6|TCP or
76 * IPv6|UDP where the Next Header field in the IPv6
77 * header is either TCP or UDP. IPv6 extension headers
78 * are not supported with this feature. This feature
79 * cannot be set in features for a device with
80 * NETIF_F_HW_CSUM also set. This feature is being
81 * DEPRECATED (see below).
83 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
84 * This flag is only used to disable the RX checksum
85 * feature for a device. The stack will accept receive
86 * checksum indication in packets received on a device
87 * regardless of whether NETIF_F_RXCSUM is set.
89 * B. Checksumming of received packets by device. Indication of checksum
90 * verification is set in skb->ip_summed. Possible values are:
94 * Device did not checksum this packet e.g. due to lack of capabilities.
95 * The packet contains full (though not verified) checksum in packet but
96 * not in skb->csum. Thus, skb->csum is undefined in this case.
98 * CHECKSUM_UNNECESSARY:
100 * The hardware you're dealing with doesn't calculate the full checksum
101 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
102 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
103 * if their checksums are okay. skb->csum is still undefined in this case
104 * though. A driver or device must never modify the checksum field in the
105 * packet even if checksum is verified.
107 * CHECKSUM_UNNECESSARY is applicable to following protocols:
108 * TCP: IPv6 and IPv4.
109 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
110 * zero UDP checksum for either IPv4 or IPv6, the networking stack
111 * may perform further validation in this case.
112 * GRE: only if the checksum is present in the header.
113 * SCTP: indicates the CRC in SCTP header has been validated.
114 * FCOE: indicates the CRC in FC frame has been validated.
116 * skb->csum_level indicates the number of consecutive checksums found in
117 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
118 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
119 * and a device is able to verify the checksums for UDP (possibly zero),
120 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
121 * two. If the device were only able to verify the UDP checksum and not
122 * GRE, either because it doesn't support GRE checksum or because GRE
123 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
124 * not considered in this case).
128 * This is the most generic way. The device supplied checksum of the _whole_
129 * packet as seen by netif_rx() and fills in skb->csum. This means the
130 * hardware doesn't need to parse L3/L4 headers to implement this.
133 * - Even if device supports only some protocols, but is able to produce
134 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
135 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
139 * A checksum is set up to be offloaded to a device as described in the
140 * output description for CHECKSUM_PARTIAL. This may occur on a packet
141 * received directly from another Linux OS, e.g., a virtualized Linux kernel
142 * on the same host, or it may be set in the input path in GRO or remote
143 * checksum offload. For the purposes of checksum verification, the checksum
144 * referred to by skb->csum_start + skb->csum_offset and any preceding
145 * checksums in the packet are considered verified. Any checksums in the
146 * packet that are after the checksum being offloaded are not considered to
149 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
150 * in the skb->ip_summed for a packet. Values are:
154 * The driver is required to checksum the packet as seen by hard_start_xmit()
155 * from skb->csum_start up to the end, and to record/write the checksum at
156 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
157 * csum_start and csum_offset values are valid values given the length and
158 * offset of the packet, but it should not attempt to validate that the
159 * checksum refers to a legitimate transport layer checksum -- it is the
160 * purview of the stack to validate that csum_start and csum_offset are set
163 * When the stack requests checksum offload for a packet, the driver MUST
164 * ensure that the checksum is set correctly. A driver can either offload the
165 * checksum calculation to the device, or call skb_checksum_help (in the case
166 * that the device does not support offload for a particular checksum).
168 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
169 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
170 * checksum offload capability.
171 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
172 * on network device checksumming capabilities: if a packet does not match
173 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
174 * csum_not_inet, see item D.) is called to resolve the checksum.
178 * The skb was already checksummed by the protocol, or a checksum is not
181 * CHECKSUM_UNNECESSARY:
183 * This has the same meaning as CHECKSUM_NONE for checksum offload on
187 * Not used in checksum output. If a driver observes a packet with this value
188 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
190 * D. Non-IP checksum (CRC) offloads
192 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
193 * offloading the SCTP CRC in a packet. To perform this offload the stack
194 * will set csum_start and csum_offset accordingly, set ip_summed to
195 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
196 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
197 * A driver that supports both IP checksum offload and SCTP CRC32c offload
198 * must verify which offload is configured for a packet by testing the
199 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
200 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
202 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
203 * offloading the FCOE CRC in a packet. To perform this offload the stack
204 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
205 * accordingly. Note that there is no indication in the skbuff that the
206 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
207 * both IP checksum offload and FCOE CRC offload must verify which offload
208 * is configured for a packet, presumably by inspecting packet headers.
210 * E. Checksumming on output with GSO.
212 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
213 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
214 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
215 * part of the GSO operation is implied. If a checksum is being offloaded
216 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
217 * csum_offset are set to refer to the outermost checksum being offloaded
218 * (two offloaded checksums are possible with UDP encapsulation).
221 /* Don't change this without changing skb_csum_unnecessary! */
222 #define CHECKSUM_NONE 0
223 #define CHECKSUM_UNNECESSARY 1
224 #define CHECKSUM_COMPLETE 2
225 #define CHECKSUM_PARTIAL 3
227 /* Maximum value in skb->csum_level */
228 #define SKB_MAX_CSUM_LEVEL 3
230 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
231 #define SKB_WITH_OVERHEAD(X) \
232 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
233 #define SKB_MAX_ORDER(X, ORDER) \
234 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
235 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
236 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
238 /* return minimum truesize of one skb containing X bytes of data */
239 #define SKB_TRUESIZE(X) ((X) + \
240 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
241 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
243 struct ahash_request;
246 struct pipe_inode_info;
253 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
254 struct nf_bridge_info {
256 BRNF_PROTO_UNCHANGED,
264 struct net_device *physindev;
266 /* always valid & non-NULL from FORWARD on, for physdev match */
267 struct net_device *physoutdev;
269 /* prerouting: detect dnat in orig/reply direction */
271 struct in6_addr ipv6_daddr;
273 /* after prerouting + nat detected: store original source
274 * mac since neigh resolution overwrites it, only used while
275 * skb is out in neigh layer.
277 char neigh_header[8];
282 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
283 /* Chain in tc_skb_ext will be used to share the tc chain with
284 * ovs recirc_id. It will be set to the current chain by tc
285 * and read by ovs to recirc_id.
297 struct sk_buff_head {
298 /* These two members must be first to match sk_buff. */
299 struct_group_tagged(sk_buff_list, list,
300 struct sk_buff *next;
301 struct sk_buff *prev;
310 /* The reason of skb drop, which is used in kfree_skb_reason().
311 * en...maybe they should be splited by group?
313 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
314 * used to translate the reason to string.
316 enum skb_drop_reason {
317 SKB_DROP_REASON_NOT_SPECIFIED,
318 SKB_DROP_REASON_NO_SOCKET,
319 SKB_DROP_REASON_PKT_TOO_SMALL,
320 SKB_DROP_REASON_TCP_CSUM,
321 SKB_DROP_REASON_TCP_FILTER,
322 SKB_DROP_REASON_UDP_CSUM,
326 /* To allow 64K frame to be packed as single skb without frag_list we
327 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
328 * buffers which do not start on a page boundary.
330 * Since GRO uses frags we allocate at least 16 regardless of page
333 #if (65536/PAGE_SIZE + 1) < 16
334 #define MAX_SKB_FRAGS 16UL
336 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
338 extern int sysctl_max_skb_frags;
340 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
341 * segment using its current segmentation instead.
343 #define GSO_BY_FRAGS 0xFFFF
345 typedef struct bio_vec skb_frag_t;
348 * skb_frag_size() - Returns the size of a skb fragment
349 * @frag: skb fragment
351 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
357 * skb_frag_size_set() - Sets the size of a skb fragment
358 * @frag: skb fragment
359 * @size: size of fragment
361 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
367 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
368 * @frag: skb fragment
369 * @delta: value to add
371 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
373 frag->bv_len += delta;
377 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
378 * @frag: skb fragment
379 * @delta: value to subtract
381 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
383 frag->bv_len -= delta;
387 * skb_frag_must_loop - Test if %p is a high memory page
388 * @p: fragment's page
390 static inline bool skb_frag_must_loop(struct page *p)
392 #if defined(CONFIG_HIGHMEM)
393 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
400 * skb_frag_foreach_page - loop over pages in a fragment
402 * @f: skb frag to operate on
403 * @f_off: offset from start of f->bv_page
404 * @f_len: length from f_off to loop over
405 * @p: (temp var) current page
406 * @p_off: (temp var) offset from start of current page,
407 * non-zero only on first page.
408 * @p_len: (temp var) length in current page,
409 * < PAGE_SIZE only on first and last page.
410 * @copied: (temp var) length so far, excluding current p_len.
412 * A fragment can hold a compound page, in which case per-page
413 * operations, notably kmap_atomic, must be called for each
416 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
417 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
418 p_off = (f_off) & (PAGE_SIZE - 1), \
419 p_len = skb_frag_must_loop(p) ? \
420 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
423 copied += p_len, p++, p_off = 0, \
424 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
426 #define HAVE_HW_TIME_STAMP
429 * struct skb_shared_hwtstamps - hardware time stamps
430 * @hwtstamp: hardware time stamp transformed into duration
431 * since arbitrary point in time
433 * Software time stamps generated by ktime_get_real() are stored in
436 * hwtstamps can only be compared against other hwtstamps from
439 * This structure is attached to packets as part of the
440 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
442 struct skb_shared_hwtstamps {
446 /* Definitions for tx_flags in struct skb_shared_info */
448 /* generate hardware time stamp */
449 SKBTX_HW_TSTAMP = 1 << 0,
451 /* generate software time stamp when queueing packet to NIC */
452 SKBTX_SW_TSTAMP = 1 << 1,
454 /* device driver is going to provide hardware time stamp */
455 SKBTX_IN_PROGRESS = 1 << 2,
457 /* generate wifi status information (where possible) */
458 SKBTX_WIFI_STATUS = 1 << 4,
460 /* generate software time stamp when entering packet scheduling */
461 SKBTX_SCHED_TSTAMP = 1 << 6,
464 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
466 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
468 /* Definitions for flags in struct skb_shared_info */
470 /* use zcopy routines */
471 SKBFL_ZEROCOPY_ENABLE = BIT(0),
473 /* This indicates at least one fragment might be overwritten
474 * (as in vmsplice(), sendfile() ...)
475 * If we need to compute a TX checksum, we'll need to copy
476 * all frags to avoid possible bad checksum
478 SKBFL_SHARED_FRAG = BIT(1),
480 /* segment contains only zerocopy data and should not be
481 * charged to the kernel memory.
483 SKBFL_PURE_ZEROCOPY = BIT(2),
486 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
487 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY)
490 * The callback notifies userspace to release buffers when skb DMA is done in
491 * lower device, the skb last reference should be 0 when calling this.
492 * The zerocopy_success argument is true if zero copy transmit occurred,
493 * false on data copy or out of memory error caused by data copy attempt.
494 * The ctx field is used to track device context.
495 * The desc field is used to track userspace buffer index.
498 void (*callback)(struct sk_buff *, struct ubuf_info *,
499 bool zerocopy_success);
516 struct user_struct *user;
521 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
523 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
524 void mm_unaccount_pinned_pages(struct mmpin *mmp);
526 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
527 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
528 struct ubuf_info *uarg);
530 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
532 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
535 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
536 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
537 struct msghdr *msg, int len,
538 struct ubuf_info *uarg);
540 /* This data is invariant across clones and lives at
541 * the end of the header data, ie. at skb->end.
543 struct skb_shared_info {
548 unsigned short gso_size;
549 /* Warning: this field is not always filled in (UFO)! */
550 unsigned short gso_segs;
551 struct sk_buff *frag_list;
552 struct skb_shared_hwtstamps hwtstamps;
553 unsigned int gso_type;
557 * Warning : all fields before dataref are cleared in __alloc_skb()
561 /* Intermediate layers must ensure that destructor_arg
562 * remains valid until skb destructor */
563 void * destructor_arg;
565 /* must be last field, see pskb_expand_head() */
566 skb_frag_t frags[MAX_SKB_FRAGS];
569 /* We divide dataref into two halves. The higher 16 bits hold references
570 * to the payload part of skb->data. The lower 16 bits hold references to
571 * the entire skb->data. A clone of a headerless skb holds the length of
572 * the header in skb->hdr_len.
574 * All users must obey the rule that the skb->data reference count must be
575 * greater than or equal to the payload reference count.
577 * Holding a reference to the payload part means that the user does not
578 * care about modifications to the header part of skb->data.
580 #define SKB_DATAREF_SHIFT 16
581 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
585 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
586 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
587 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
591 SKB_GSO_TCPV4 = 1 << 0,
593 /* This indicates the skb is from an untrusted source. */
594 SKB_GSO_DODGY = 1 << 1,
596 /* This indicates the tcp segment has CWR set. */
597 SKB_GSO_TCP_ECN = 1 << 2,
599 SKB_GSO_TCP_FIXEDID = 1 << 3,
601 SKB_GSO_TCPV6 = 1 << 4,
603 SKB_GSO_FCOE = 1 << 5,
605 SKB_GSO_GRE = 1 << 6,
607 SKB_GSO_GRE_CSUM = 1 << 7,
609 SKB_GSO_IPXIP4 = 1 << 8,
611 SKB_GSO_IPXIP6 = 1 << 9,
613 SKB_GSO_UDP_TUNNEL = 1 << 10,
615 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
617 SKB_GSO_PARTIAL = 1 << 12,
619 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
621 SKB_GSO_SCTP = 1 << 14,
623 SKB_GSO_ESP = 1 << 15,
625 SKB_GSO_UDP = 1 << 16,
627 SKB_GSO_UDP_L4 = 1 << 17,
629 SKB_GSO_FRAGLIST = 1 << 18,
632 #if BITS_PER_LONG > 32
633 #define NET_SKBUFF_DATA_USES_OFFSET 1
636 #ifdef NET_SKBUFF_DATA_USES_OFFSET
637 typedef unsigned int sk_buff_data_t;
639 typedef unsigned char *sk_buff_data_t;
643 * struct sk_buff - socket buffer
644 * @next: Next buffer in list
645 * @prev: Previous buffer in list
646 * @tstamp: Time we arrived/left
647 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
648 * for retransmit timer
649 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
651 * @ll_node: anchor in an llist (eg socket defer_list)
652 * @sk: Socket we are owned by
653 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
654 * fragmentation management
655 * @dev: Device we arrived on/are leaving by
656 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
657 * @cb: Control buffer. Free for use by every layer. Put private vars here
658 * @_skb_refdst: destination entry (with norefcount bit)
659 * @sp: the security path, used for xfrm
660 * @len: Length of actual data
661 * @data_len: Data length
662 * @mac_len: Length of link layer header
663 * @hdr_len: writable header length of cloned skb
664 * @csum: Checksum (must include start/offset pair)
665 * @csum_start: Offset from skb->head where checksumming should start
666 * @csum_offset: Offset from csum_start where checksum should be stored
667 * @priority: Packet queueing priority
668 * @ignore_df: allow local fragmentation
669 * @cloned: Head may be cloned (check refcnt to be sure)
670 * @ip_summed: Driver fed us an IP checksum
671 * @nohdr: Payload reference only, must not modify header
672 * @pkt_type: Packet class
673 * @fclone: skbuff clone status
674 * @ipvs_property: skbuff is owned by ipvs
675 * @inner_protocol_type: whether the inner protocol is
676 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
677 * @remcsum_offload: remote checksum offload is enabled
678 * @offload_fwd_mark: Packet was L2-forwarded in hardware
679 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
680 * @tc_skip_classify: do not classify packet. set by IFB device
681 * @tc_at_ingress: used within tc_classify to distinguish in/egress
682 * @redirected: packet was redirected by packet classifier
683 * @from_ingress: packet was redirected from the ingress path
684 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
685 * @peeked: this packet has been seen already, so stats have been
686 * done for it, don't do them again
687 * @nf_trace: netfilter packet trace flag
688 * @protocol: Packet protocol from driver
689 * @destructor: Destruct function
690 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
691 * @_sk_redir: socket redirection information for skmsg
692 * @_nfct: Associated connection, if any (with nfctinfo bits)
693 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
694 * @skb_iif: ifindex of device we arrived on
695 * @tc_index: Traffic control index
696 * @hash: the packet hash
697 * @queue_mapping: Queue mapping for multiqueue devices
698 * @head_frag: skb was allocated from page fragments,
699 * not allocated by kmalloc() or vmalloc().
700 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
701 * @pp_recycle: mark the packet for recycling instead of freeing (implies
702 * page_pool support on driver)
703 * @active_extensions: active extensions (skb_ext_id types)
704 * @ndisc_nodetype: router type (from link layer)
705 * @ooo_okay: allow the mapping of a socket to a queue to be changed
706 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
708 * @sw_hash: indicates hash was computed in software stack
709 * @wifi_acked_valid: wifi_acked was set
710 * @wifi_acked: whether frame was acked on wifi or not
711 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
712 * @encapsulation: indicates the inner headers in the skbuff are valid
713 * @encap_hdr_csum: software checksum is needed
714 * @csum_valid: checksum is already valid
715 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
716 * @csum_complete_sw: checksum was completed by software
717 * @csum_level: indicates the number of consecutive checksums found in
718 * the packet minus one that have been verified as
719 * CHECKSUM_UNNECESSARY (max 3)
720 * @dst_pending_confirm: need to confirm neighbour
721 * @decrypted: Decrypted SKB
722 * @slow_gro: state present at GRO time, slower prepare step required
723 * @napi_id: id of the NAPI struct this skb came from
724 * @sender_cpu: (aka @napi_id) source CPU in XPS
725 * @secmark: security marking
726 * @mark: Generic packet mark
727 * @reserved_tailroom: (aka @mark) number of bytes of free space available
728 * at the tail of an sk_buff
729 * @vlan_present: VLAN tag is present
730 * @vlan_proto: vlan encapsulation protocol
731 * @vlan_tci: vlan tag control information
732 * @inner_protocol: Protocol (encapsulation)
733 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
734 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
735 * @inner_transport_header: Inner transport layer header (encapsulation)
736 * @inner_network_header: Network layer header (encapsulation)
737 * @inner_mac_header: Link layer header (encapsulation)
738 * @transport_header: Transport layer header
739 * @network_header: Network layer header
740 * @mac_header: Link layer header
741 * @kcov_handle: KCOV remote handle for remote coverage collection
742 * @tail: Tail pointer
744 * @head: Head of buffer
745 * @data: Data head pointer
746 * @truesize: Buffer size
747 * @users: User count - see {datagram,tcp}.c
748 * @extensions: allocated extensions, valid if active_extensions is nonzero
754 /* These two members must be first to match sk_buff_head. */
755 struct sk_buff *next;
756 struct sk_buff *prev;
759 struct net_device *dev;
760 /* Some protocols might use this space to store information,
761 * while device pointer would be NULL.
762 * UDP receive path is one user.
764 unsigned long dev_scratch;
767 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
768 struct list_head list;
769 struct llist_node ll_node;
774 int ip_defrag_offset;
779 u64 skb_mstamp_ns; /* earliest departure time */
782 * This is the control buffer. It is free to use for every
783 * layer. Please put your private variables there. If you
784 * want to keep them across layers you have to do a skb_clone()
785 * first. This is owned by whoever has the skb queued ATM.
787 char cb[48] __aligned(8);
791 unsigned long _skb_refdst;
792 void (*destructor)(struct sk_buff *skb);
794 struct list_head tcp_tsorted_anchor;
795 #ifdef CONFIG_NET_SOCK_MSG
796 unsigned long _sk_redir;
800 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
808 /* Following fields are _not_ copied in __copy_skb_header()
809 * Note that queue_mapping is here mostly to fill a hole.
813 /* if you move cloned around you also must adapt those constants */
814 #ifdef __BIG_ENDIAN_BITFIELD
815 #define CLONED_MASK (1 << 7)
817 #define CLONED_MASK 1
819 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
822 __u8 __cloned_offset[0];
830 pp_recycle:1; /* page_pool recycle indicator */
831 #ifdef CONFIG_SKB_EXTENSIONS
832 __u8 active_extensions;
835 /* Fields enclosed in headers group are copied
836 * using a single memcpy() in __copy_skb_header()
838 struct_group(headers,
841 __u8 __pkt_type_offset[0];
843 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
851 __u8 wifi_acked_valid:1;
854 /* Indicates the inner headers are valid in the skbuff. */
855 __u8 encapsulation:1;
856 __u8 encap_hdr_csum:1;
860 __u8 __pkt_vlan_present_offset[0];
862 __u8 vlan_present:1; /* See PKT_VLAN_PRESENT_BIT */
863 __u8 csum_complete_sw:1;
865 __u8 csum_not_inet:1;
866 __u8 dst_pending_confirm:1;
867 #ifdef CONFIG_IPV6_NDISC_NODETYPE
868 __u8 ndisc_nodetype:2;
871 __u8 ipvs_property:1;
872 __u8 inner_protocol_type:1;
873 __u8 remcsum_offload:1;
874 #ifdef CONFIG_NET_SWITCHDEV
875 __u8 offload_fwd_mark:1;
876 __u8 offload_l3_fwd_mark:1;
878 #ifdef CONFIG_NET_CLS_ACT
879 __u8 tc_skip_classify:1;
880 __u8 tc_at_ingress:1;
883 #ifdef CONFIG_NET_REDIRECT
886 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
887 __u8 nf_skip_egress:1;
889 #ifdef CONFIG_TLS_DEVICE
894 #ifdef CONFIG_NET_SCHED
895 __u16 tc_index; /* traffic control index */
910 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
912 unsigned int napi_id;
913 unsigned int sender_cpu;
916 #ifdef CONFIG_NETWORK_SECMARK
922 __u32 reserved_tailroom;
926 __be16 inner_protocol;
930 __u16 inner_transport_header;
931 __u16 inner_network_header;
932 __u16 inner_mac_header;
935 __u16 transport_header;
936 __u16 network_header;
943 ); /* end headers group */
945 /* These elements must be at the end, see alloc_skb() for details. */
950 unsigned int truesize;
953 #ifdef CONFIG_SKB_EXTENSIONS
954 /* only useable after checking ->active_extensions != 0 */
955 struct skb_ext *extensions;
959 /* if you move pkt_type around you also must adapt those constants */
960 #ifdef __BIG_ENDIAN_BITFIELD
961 #define PKT_TYPE_MAX (7 << 5)
963 #define PKT_TYPE_MAX 7
965 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset)
967 /* if you move pkt_vlan_present around you also must adapt these constants */
968 #ifdef __BIG_ENDIAN_BITFIELD
969 #define PKT_VLAN_PRESENT_BIT 7
971 #define PKT_VLAN_PRESENT_BIT 0
973 #define PKT_VLAN_PRESENT_OFFSET offsetof(struct sk_buff, __pkt_vlan_present_offset)
977 * Handling routines are only of interest to the kernel
980 #define SKB_ALLOC_FCLONE 0x01
981 #define SKB_ALLOC_RX 0x02
982 #define SKB_ALLOC_NAPI 0x04
985 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
988 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
990 return unlikely(skb->pfmemalloc);
994 * skb might have a dst pointer attached, refcounted or not.
995 * _skb_refdst low order bit is set if refcount was _not_ taken
997 #define SKB_DST_NOREF 1UL
998 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1001 * skb_dst - returns skb dst_entry
1004 * Returns skb dst_entry, regardless of reference taken or not.
1006 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1008 /* If refdst was not refcounted, check we still are in a
1009 * rcu_read_lock section
1011 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1012 !rcu_read_lock_held() &&
1013 !rcu_read_lock_bh_held());
1014 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1018 * skb_dst_set - sets skb dst
1022 * Sets skb dst, assuming a reference was taken on dst and should
1023 * be released by skb_dst_drop()
1025 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1027 skb->slow_gro |= !!dst;
1028 skb->_skb_refdst = (unsigned long)dst;
1032 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1036 * Sets skb dst, assuming a reference was not taken on dst.
1037 * If dst entry is cached, we do not take reference and dst_release
1038 * will be avoided by refdst_drop. If dst entry is not cached, we take
1039 * reference, so that last dst_release can destroy the dst immediately.
1041 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1043 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1044 skb->slow_gro |= !!dst;
1045 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1049 * skb_dst_is_noref - Test if skb dst isn't refcounted
1052 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1054 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1058 * skb_rtable - Returns the skb &rtable
1061 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1063 return (struct rtable *)skb_dst(skb);
1066 /* For mangling skb->pkt_type from user space side from applications
1067 * such as nft, tc, etc, we only allow a conservative subset of
1068 * possible pkt_types to be set.
1070 static inline bool skb_pkt_type_ok(u32 ptype)
1072 return ptype <= PACKET_OTHERHOST;
1076 * skb_napi_id - Returns the skb's NAPI id
1079 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1081 #ifdef CONFIG_NET_RX_BUSY_POLL
1082 return skb->napi_id;
1089 * skb_unref - decrement the skb's reference count
1092 * Returns true if we can free the skb.
1094 static inline bool skb_unref(struct sk_buff *skb)
1098 if (likely(refcount_read(&skb->users) == 1))
1100 else if (likely(!refcount_dec_and_test(&skb->users)))
1106 void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1109 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1110 * @skb: buffer to free
1112 static inline void kfree_skb(struct sk_buff *skb)
1114 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1117 void skb_release_head_state(struct sk_buff *skb);
1118 void kfree_skb_list(struct sk_buff *segs);
1119 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1120 void skb_tx_error(struct sk_buff *skb);
1122 #ifdef CONFIG_TRACEPOINTS
1123 void consume_skb(struct sk_buff *skb);
1125 static inline void consume_skb(struct sk_buff *skb)
1127 return kfree_skb(skb);
1131 void __consume_stateless_skb(struct sk_buff *skb);
1132 void __kfree_skb(struct sk_buff *skb);
1133 extern struct kmem_cache *skbuff_head_cache;
1135 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1136 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1137 bool *fragstolen, int *delta_truesize);
1139 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1141 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1142 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1143 struct sk_buff *build_skb_around(struct sk_buff *skb,
1144 void *data, unsigned int frag_size);
1146 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1149 * alloc_skb - allocate a network buffer
1150 * @size: size to allocate
1151 * @priority: allocation mask
1153 * This function is a convenient wrapper around __alloc_skb().
1155 static inline struct sk_buff *alloc_skb(unsigned int size,
1158 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1161 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1162 unsigned long data_len,
1166 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1168 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1169 struct sk_buff_fclones {
1170 struct sk_buff skb1;
1172 struct sk_buff skb2;
1174 refcount_t fclone_ref;
1178 * skb_fclone_busy - check if fclone is busy
1182 * Returns true if skb is a fast clone, and its clone is not freed.
1183 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1184 * so we also check that this didnt happen.
1186 static inline bool skb_fclone_busy(const struct sock *sk,
1187 const struct sk_buff *skb)
1189 const struct sk_buff_fclones *fclones;
1191 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1193 return skb->fclone == SKB_FCLONE_ORIG &&
1194 refcount_read(&fclones->fclone_ref) > 1 &&
1195 READ_ONCE(fclones->skb2.sk) == sk;
1199 * alloc_skb_fclone - allocate a network buffer from fclone cache
1200 * @size: size to allocate
1201 * @priority: allocation mask
1203 * This function is a convenient wrapper around __alloc_skb().
1205 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1208 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1211 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1212 void skb_headers_offset_update(struct sk_buff *skb, int off);
1213 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1214 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1215 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1216 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1217 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1218 gfp_t gfp_mask, bool fclone);
1219 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1222 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1225 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1226 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1227 unsigned int headroom);
1228 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1229 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1230 int newtailroom, gfp_t priority);
1231 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1232 int offset, int len);
1233 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1234 int offset, int len);
1235 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1236 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1239 * skb_pad - zero pad the tail of an skb
1240 * @skb: buffer to pad
1241 * @pad: space to pad
1243 * Ensure that a buffer is followed by a padding area that is zero
1244 * filled. Used by network drivers which may DMA or transfer data
1245 * beyond the buffer end onto the wire.
1247 * May return error in out of memory cases. The skb is freed on error.
1249 static inline int skb_pad(struct sk_buff *skb, int pad)
1251 return __skb_pad(skb, pad, true);
1253 #define dev_kfree_skb(a) consume_skb(a)
1255 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1256 int offset, size_t size);
1258 struct skb_seq_state {
1262 __u32 stepped_offset;
1263 struct sk_buff *root_skb;
1264 struct sk_buff *cur_skb;
1269 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1270 unsigned int to, struct skb_seq_state *st);
1271 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1272 struct skb_seq_state *st);
1273 void skb_abort_seq_read(struct skb_seq_state *st);
1275 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1276 unsigned int to, struct ts_config *config);
1279 * Packet hash types specify the type of hash in skb_set_hash.
1281 * Hash types refer to the protocol layer addresses which are used to
1282 * construct a packet's hash. The hashes are used to differentiate or identify
1283 * flows of the protocol layer for the hash type. Hash types are either
1284 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1286 * Properties of hashes:
1288 * 1) Two packets in different flows have different hash values
1289 * 2) Two packets in the same flow should have the same hash value
1291 * A hash at a higher layer is considered to be more specific. A driver should
1292 * set the most specific hash possible.
1294 * A driver cannot indicate a more specific hash than the layer at which a hash
1295 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1297 * A driver may indicate a hash level which is less specific than the
1298 * actual layer the hash was computed on. For instance, a hash computed
1299 * at L4 may be considered an L3 hash. This should only be done if the
1300 * driver can't unambiguously determine that the HW computed the hash at
1301 * the higher layer. Note that the "should" in the second property above
1304 enum pkt_hash_types {
1305 PKT_HASH_TYPE_NONE, /* Undefined type */
1306 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1307 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1308 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1311 static inline void skb_clear_hash(struct sk_buff *skb)
1318 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1321 skb_clear_hash(skb);
1325 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1327 skb->l4_hash = is_l4;
1328 skb->sw_hash = is_sw;
1333 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1335 /* Used by drivers to set hash from HW */
1336 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1340 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1342 __skb_set_hash(skb, hash, true, is_l4);
1345 void __skb_get_hash(struct sk_buff *skb);
1346 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1347 u32 skb_get_poff(const struct sk_buff *skb);
1348 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1349 const struct flow_keys_basic *keys, int hlen);
1350 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1351 const void *data, int hlen_proto);
1353 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1354 int thoff, u8 ip_proto)
1356 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1359 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1360 const struct flow_dissector_key *key,
1361 unsigned int key_count);
1363 struct bpf_flow_dissector;
1364 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1365 __be16 proto, int nhoff, int hlen, unsigned int flags);
1367 bool __skb_flow_dissect(const struct net *net,
1368 const struct sk_buff *skb,
1369 struct flow_dissector *flow_dissector,
1370 void *target_container, const void *data,
1371 __be16 proto, int nhoff, int hlen, unsigned int flags);
1373 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1374 struct flow_dissector *flow_dissector,
1375 void *target_container, unsigned int flags)
1377 return __skb_flow_dissect(NULL, skb, flow_dissector,
1378 target_container, NULL, 0, 0, 0, flags);
1381 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1382 struct flow_keys *flow,
1385 memset(flow, 0, sizeof(*flow));
1386 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1387 flow, NULL, 0, 0, 0, flags);
1391 skb_flow_dissect_flow_keys_basic(const struct net *net,
1392 const struct sk_buff *skb,
1393 struct flow_keys_basic *flow,
1394 const void *data, __be16 proto,
1395 int nhoff, int hlen, unsigned int flags)
1397 memset(flow, 0, sizeof(*flow));
1398 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1399 data, proto, nhoff, hlen, flags);
1402 void skb_flow_dissect_meta(const struct sk_buff *skb,
1403 struct flow_dissector *flow_dissector,
1404 void *target_container);
1406 /* Gets a skb connection tracking info, ctinfo map should be a
1407 * map of mapsize to translate enum ip_conntrack_info states
1411 skb_flow_dissect_ct(const struct sk_buff *skb,
1412 struct flow_dissector *flow_dissector,
1413 void *target_container,
1414 u16 *ctinfo_map, size_t mapsize,
1415 bool post_ct, u16 zone);
1417 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1418 struct flow_dissector *flow_dissector,
1419 void *target_container);
1421 void skb_flow_dissect_hash(const struct sk_buff *skb,
1422 struct flow_dissector *flow_dissector,
1423 void *target_container);
1425 static inline __u32 skb_get_hash(struct sk_buff *skb)
1427 if (!skb->l4_hash && !skb->sw_hash)
1428 __skb_get_hash(skb);
1433 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1435 if (!skb->l4_hash && !skb->sw_hash) {
1436 struct flow_keys keys;
1437 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1439 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1445 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1446 const siphash_key_t *perturb);
1448 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1453 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1455 to->hash = from->hash;
1456 to->sw_hash = from->sw_hash;
1457 to->l4_hash = from->l4_hash;
1460 static inline void skb_copy_decrypted(struct sk_buff *to,
1461 const struct sk_buff *from)
1463 #ifdef CONFIG_TLS_DEVICE
1464 to->decrypted = from->decrypted;
1468 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1469 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1471 return skb->head + skb->end;
1474 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1479 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1484 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1486 return skb->end - skb->head;
1491 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1493 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1495 return &skb_shinfo(skb)->hwtstamps;
1498 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1500 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1502 return is_zcopy ? skb_uarg(skb) : NULL;
1505 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1507 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1510 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1511 const struct sk_buff *skb2)
1513 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1516 static inline void net_zcopy_get(struct ubuf_info *uarg)
1518 refcount_inc(&uarg->refcnt);
1521 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1523 skb_shinfo(skb)->destructor_arg = uarg;
1524 skb_shinfo(skb)->flags |= uarg->flags;
1527 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1530 if (skb && uarg && !skb_zcopy(skb)) {
1531 if (unlikely(have_ref && *have_ref))
1534 net_zcopy_get(uarg);
1535 skb_zcopy_init(skb, uarg);
1539 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1541 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1542 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1545 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1547 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1550 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1552 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1555 static inline void net_zcopy_put(struct ubuf_info *uarg)
1558 uarg->callback(NULL, uarg, true);
1561 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1564 if (uarg->callback == msg_zerocopy_callback)
1565 msg_zerocopy_put_abort(uarg, have_uref);
1567 net_zcopy_put(uarg);
1571 /* Release a reference on a zerocopy structure */
1572 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1574 struct ubuf_info *uarg = skb_zcopy(skb);
1577 if (!skb_zcopy_is_nouarg(skb))
1578 uarg->callback(skb, uarg, zerocopy_success);
1580 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1584 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1589 /* Iterate through singly-linked GSO fragments of an skb. */
1590 #define skb_list_walk_safe(first, skb, next_skb) \
1591 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1592 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1594 static inline void skb_list_del_init(struct sk_buff *skb)
1596 __list_del_entry(&skb->list);
1597 skb_mark_not_on_list(skb);
1601 * skb_queue_empty - check if a queue is empty
1604 * Returns true if the queue is empty, false otherwise.
1606 static inline int skb_queue_empty(const struct sk_buff_head *list)
1608 return list->next == (const struct sk_buff *) list;
1612 * skb_queue_empty_lockless - check if a queue is empty
1615 * Returns true if the queue is empty, false otherwise.
1616 * This variant can be used in lockless contexts.
1618 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1620 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1625 * skb_queue_is_last - check if skb is the last entry in the queue
1629 * Returns true if @skb is the last buffer on the list.
1631 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1632 const struct sk_buff *skb)
1634 return skb->next == (const struct sk_buff *) list;
1638 * skb_queue_is_first - check if skb is the first entry in the queue
1642 * Returns true if @skb is the first buffer on the list.
1644 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1645 const struct sk_buff *skb)
1647 return skb->prev == (const struct sk_buff *) list;
1651 * skb_queue_next - return the next packet in the queue
1653 * @skb: current buffer
1655 * Return the next packet in @list after @skb. It is only valid to
1656 * call this if skb_queue_is_last() evaluates to false.
1658 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1659 const struct sk_buff *skb)
1661 /* This BUG_ON may seem severe, but if we just return then we
1662 * are going to dereference garbage.
1664 BUG_ON(skb_queue_is_last(list, skb));
1669 * skb_queue_prev - return the prev packet in the queue
1671 * @skb: current buffer
1673 * Return the prev packet in @list before @skb. It is only valid to
1674 * call this if skb_queue_is_first() evaluates to false.
1676 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1677 const struct sk_buff *skb)
1679 /* This BUG_ON may seem severe, but if we just return then we
1680 * are going to dereference garbage.
1682 BUG_ON(skb_queue_is_first(list, skb));
1687 * skb_get - reference buffer
1688 * @skb: buffer to reference
1690 * Makes another reference to a socket buffer and returns a pointer
1693 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1695 refcount_inc(&skb->users);
1700 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1704 * skb_cloned - is the buffer a clone
1705 * @skb: buffer to check
1707 * Returns true if the buffer was generated with skb_clone() and is
1708 * one of multiple shared copies of the buffer. Cloned buffers are
1709 * shared data so must not be written to under normal circumstances.
1711 static inline int skb_cloned(const struct sk_buff *skb)
1713 return skb->cloned &&
1714 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1717 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1719 might_sleep_if(gfpflags_allow_blocking(pri));
1721 if (skb_cloned(skb))
1722 return pskb_expand_head(skb, 0, 0, pri);
1727 /* This variant of skb_unclone() makes sure skb->truesize is not changed */
1728 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1730 might_sleep_if(gfpflags_allow_blocking(pri));
1732 if (skb_cloned(skb)) {
1733 unsigned int save = skb->truesize;
1736 res = pskb_expand_head(skb, 0, 0, pri);
1737 skb->truesize = save;
1744 * skb_header_cloned - is the header a clone
1745 * @skb: buffer to check
1747 * Returns true if modifying the header part of the buffer requires
1748 * the data to be copied.
1750 static inline int skb_header_cloned(const struct sk_buff *skb)
1757 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1758 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1759 return dataref != 1;
1762 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1764 might_sleep_if(gfpflags_allow_blocking(pri));
1766 if (skb_header_cloned(skb))
1767 return pskb_expand_head(skb, 0, 0, pri);
1773 * __skb_header_release - release reference to header
1774 * @skb: buffer to operate on
1776 static inline void __skb_header_release(struct sk_buff *skb)
1779 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1784 * skb_shared - is the buffer shared
1785 * @skb: buffer to check
1787 * Returns true if more than one person has a reference to this
1790 static inline int skb_shared(const struct sk_buff *skb)
1792 return refcount_read(&skb->users) != 1;
1796 * skb_share_check - check if buffer is shared and if so clone it
1797 * @skb: buffer to check
1798 * @pri: priority for memory allocation
1800 * If the buffer is shared the buffer is cloned and the old copy
1801 * drops a reference. A new clone with a single reference is returned.
1802 * If the buffer is not shared the original buffer is returned. When
1803 * being called from interrupt status or with spinlocks held pri must
1806 * NULL is returned on a memory allocation failure.
1808 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1810 might_sleep_if(gfpflags_allow_blocking(pri));
1811 if (skb_shared(skb)) {
1812 struct sk_buff *nskb = skb_clone(skb, pri);
1824 * Copy shared buffers into a new sk_buff. We effectively do COW on
1825 * packets to handle cases where we have a local reader and forward
1826 * and a couple of other messy ones. The normal one is tcpdumping
1827 * a packet thats being forwarded.
1831 * skb_unshare - make a copy of a shared buffer
1832 * @skb: buffer to check
1833 * @pri: priority for memory allocation
1835 * If the socket buffer is a clone then this function creates a new
1836 * copy of the data, drops a reference count on the old copy and returns
1837 * the new copy with the reference count at 1. If the buffer is not a clone
1838 * the original buffer is returned. When called with a spinlock held or
1839 * from interrupt state @pri must be %GFP_ATOMIC
1841 * %NULL is returned on a memory allocation failure.
1843 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1846 might_sleep_if(gfpflags_allow_blocking(pri));
1847 if (skb_cloned(skb)) {
1848 struct sk_buff *nskb = skb_copy(skb, pri);
1850 /* Free our shared copy */
1861 * skb_peek - peek at the head of an &sk_buff_head
1862 * @list_: list to peek at
1864 * Peek an &sk_buff. Unlike most other operations you _MUST_
1865 * be careful with this one. A peek leaves the buffer on the
1866 * list and someone else may run off with it. You must hold
1867 * the appropriate locks or have a private queue to do this.
1869 * Returns %NULL for an empty list or a pointer to the head element.
1870 * The reference count is not incremented and the reference is therefore
1871 * volatile. Use with caution.
1873 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1875 struct sk_buff *skb = list_->next;
1877 if (skb == (struct sk_buff *)list_)
1883 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1884 * @list_: list to peek at
1886 * Like skb_peek(), but the caller knows that the list is not empty.
1888 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1894 * skb_peek_next - peek skb following the given one from a queue
1895 * @skb: skb to start from
1896 * @list_: list to peek at
1898 * Returns %NULL when the end of the list is met or a pointer to the
1899 * next element. The reference count is not incremented and the
1900 * reference is therefore volatile. Use with caution.
1902 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1903 const struct sk_buff_head *list_)
1905 struct sk_buff *next = skb->next;
1907 if (next == (struct sk_buff *)list_)
1913 * skb_peek_tail - peek at the tail of an &sk_buff_head
1914 * @list_: list to peek at
1916 * Peek an &sk_buff. Unlike most other operations you _MUST_
1917 * be careful with this one. A peek leaves the buffer on the
1918 * list and someone else may run off with it. You must hold
1919 * the appropriate locks or have a private queue to do this.
1921 * Returns %NULL for an empty list or a pointer to the tail element.
1922 * The reference count is not incremented and the reference is therefore
1923 * volatile. Use with caution.
1925 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1927 struct sk_buff *skb = READ_ONCE(list_->prev);
1929 if (skb == (struct sk_buff *)list_)
1936 * skb_queue_len - get queue length
1937 * @list_: list to measure
1939 * Return the length of an &sk_buff queue.
1941 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1947 * skb_queue_len_lockless - get queue length
1948 * @list_: list to measure
1950 * Return the length of an &sk_buff queue.
1951 * This variant can be used in lockless contexts.
1953 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1955 return READ_ONCE(list_->qlen);
1959 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1960 * @list: queue to initialize
1962 * This initializes only the list and queue length aspects of
1963 * an sk_buff_head object. This allows to initialize the list
1964 * aspects of an sk_buff_head without reinitializing things like
1965 * the spinlock. It can also be used for on-stack sk_buff_head
1966 * objects where the spinlock is known to not be used.
1968 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1970 list->prev = list->next = (struct sk_buff *)list;
1975 * This function creates a split out lock class for each invocation;
1976 * this is needed for now since a whole lot of users of the skb-queue
1977 * infrastructure in drivers have different locking usage (in hardirq)
1978 * than the networking core (in softirq only). In the long run either the
1979 * network layer or drivers should need annotation to consolidate the
1980 * main types of usage into 3 classes.
1982 static inline void skb_queue_head_init(struct sk_buff_head *list)
1984 spin_lock_init(&list->lock);
1985 __skb_queue_head_init(list);
1988 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1989 struct lock_class_key *class)
1991 skb_queue_head_init(list);
1992 lockdep_set_class(&list->lock, class);
1996 * Insert an sk_buff on a list.
1998 * The "__skb_xxxx()" functions are the non-atomic ones that
1999 * can only be called with interrupts disabled.
2001 static inline void __skb_insert(struct sk_buff *newsk,
2002 struct sk_buff *prev, struct sk_buff *next,
2003 struct sk_buff_head *list)
2005 /* See skb_queue_empty_lockless() and skb_peek_tail()
2006 * for the opposite READ_ONCE()
2008 WRITE_ONCE(newsk->next, next);
2009 WRITE_ONCE(newsk->prev, prev);
2010 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2011 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
2012 WRITE_ONCE(list->qlen, list->qlen + 1);
2015 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2016 struct sk_buff *prev,
2017 struct sk_buff *next)
2019 struct sk_buff *first = list->next;
2020 struct sk_buff *last = list->prev;
2022 WRITE_ONCE(first->prev, prev);
2023 WRITE_ONCE(prev->next, first);
2025 WRITE_ONCE(last->next, next);
2026 WRITE_ONCE(next->prev, last);
2030 * skb_queue_splice - join two skb lists, this is designed for stacks
2031 * @list: the new list to add
2032 * @head: the place to add it in the first list
2034 static inline void skb_queue_splice(const struct sk_buff_head *list,
2035 struct sk_buff_head *head)
2037 if (!skb_queue_empty(list)) {
2038 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2039 head->qlen += list->qlen;
2044 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2045 * @list: the new list to add
2046 * @head: the place to add it in the first list
2048 * The list at @list is reinitialised
2050 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2051 struct sk_buff_head *head)
2053 if (!skb_queue_empty(list)) {
2054 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2055 head->qlen += list->qlen;
2056 __skb_queue_head_init(list);
2061 * skb_queue_splice_tail - join two skb lists, each list being a queue
2062 * @list: the new list to add
2063 * @head: the place to add it in the first list
2065 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2066 struct sk_buff_head *head)
2068 if (!skb_queue_empty(list)) {
2069 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2070 head->qlen += list->qlen;
2075 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2076 * @list: the new list to add
2077 * @head: the place to add it in the first list
2079 * Each of the lists is a queue.
2080 * The list at @list is reinitialised
2082 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2083 struct sk_buff_head *head)
2085 if (!skb_queue_empty(list)) {
2086 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2087 head->qlen += list->qlen;
2088 __skb_queue_head_init(list);
2093 * __skb_queue_after - queue a buffer at the list head
2094 * @list: list to use
2095 * @prev: place after this buffer
2096 * @newsk: buffer to queue
2098 * Queue a buffer int the middle of a list. This function takes no locks
2099 * and you must therefore hold required locks before calling it.
2101 * A buffer cannot be placed on two lists at the same time.
2103 static inline void __skb_queue_after(struct sk_buff_head *list,
2104 struct sk_buff *prev,
2105 struct sk_buff *newsk)
2107 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
2110 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2111 struct sk_buff_head *list);
2113 static inline void __skb_queue_before(struct sk_buff_head *list,
2114 struct sk_buff *next,
2115 struct sk_buff *newsk)
2117 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
2121 * __skb_queue_head - queue a buffer at the list head
2122 * @list: list to use
2123 * @newsk: buffer to queue
2125 * Queue a buffer at the start of a list. This function takes no locks
2126 * and you must therefore hold required locks before calling it.
2128 * A buffer cannot be placed on two lists at the same time.
2130 static inline void __skb_queue_head(struct sk_buff_head *list,
2131 struct sk_buff *newsk)
2133 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2135 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2138 * __skb_queue_tail - queue a buffer at the list tail
2139 * @list: list to use
2140 * @newsk: buffer to queue
2142 * Queue a buffer at the end of a list. This function takes no locks
2143 * and you must therefore hold required locks before calling it.
2145 * A buffer cannot be placed on two lists at the same time.
2147 static inline void __skb_queue_tail(struct sk_buff_head *list,
2148 struct sk_buff *newsk)
2150 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2152 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2155 * remove sk_buff from list. _Must_ be called atomically, and with
2158 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2159 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2161 struct sk_buff *next, *prev;
2163 WRITE_ONCE(list->qlen, list->qlen - 1);
2166 skb->next = skb->prev = NULL;
2167 WRITE_ONCE(next->prev, prev);
2168 WRITE_ONCE(prev->next, next);
2172 * __skb_dequeue - remove from the head of the queue
2173 * @list: list to dequeue from
2175 * Remove the head of the list. This function does not take any locks
2176 * so must be used with appropriate locks held only. The head item is
2177 * returned or %NULL if the list is empty.
2179 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2181 struct sk_buff *skb = skb_peek(list);
2183 __skb_unlink(skb, list);
2186 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2189 * __skb_dequeue_tail - remove from the tail of the queue
2190 * @list: list to dequeue from
2192 * Remove the tail of the list. This function does not take any locks
2193 * so must be used with appropriate locks held only. The tail item is
2194 * returned or %NULL if the list is empty.
2196 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2198 struct sk_buff *skb = skb_peek_tail(list);
2200 __skb_unlink(skb, list);
2203 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2206 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2208 return skb->data_len;
2211 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2213 return skb->len - skb->data_len;
2216 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2218 unsigned int i, len = 0;
2220 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2221 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2225 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2227 return skb_headlen(skb) + __skb_pagelen(skb);
2231 * __skb_fill_page_desc - initialise a paged fragment in an skb
2232 * @skb: buffer containing fragment to be initialised
2233 * @i: paged fragment index to initialise
2234 * @page: the page to use for this fragment
2235 * @off: the offset to the data with @page
2236 * @size: the length of the data
2238 * Initialises the @i'th fragment of @skb to point to &size bytes at
2239 * offset @off within @page.
2241 * Does not take any additional reference on the fragment.
2243 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2244 struct page *page, int off, int size)
2246 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2249 * Propagate page pfmemalloc to the skb if we can. The problem is
2250 * that not all callers have unique ownership of the page but rely
2251 * on page_is_pfmemalloc doing the right thing(tm).
2253 frag->bv_page = page;
2254 frag->bv_offset = off;
2255 skb_frag_size_set(frag, size);
2257 page = compound_head(page);
2258 if (page_is_pfmemalloc(page))
2259 skb->pfmemalloc = true;
2263 * skb_fill_page_desc - initialise a paged fragment in an skb
2264 * @skb: buffer containing fragment to be initialised
2265 * @i: paged fragment index to initialise
2266 * @page: the page to use for this fragment
2267 * @off: the offset to the data with @page
2268 * @size: the length of the data
2270 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2271 * @skb to point to @size bytes at offset @off within @page. In
2272 * addition updates @skb such that @i is the last fragment.
2274 * Does not take any additional reference on the fragment.
2276 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2277 struct page *page, int off, int size)
2279 __skb_fill_page_desc(skb, i, page, off, size);
2280 skb_shinfo(skb)->nr_frags = i + 1;
2283 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2284 int size, unsigned int truesize);
2286 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2287 unsigned int truesize);
2289 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2291 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2292 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2294 return skb->head + skb->tail;
2297 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2299 skb->tail = skb->data - skb->head;
2302 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2304 skb_reset_tail_pointer(skb);
2305 skb->tail += offset;
2308 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2309 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2314 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2316 skb->tail = skb->data;
2319 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2321 skb->tail = skb->data + offset;
2324 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2327 * Add data to an sk_buff
2329 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2330 void *skb_put(struct sk_buff *skb, unsigned int len);
2331 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2333 void *tmp = skb_tail_pointer(skb);
2334 SKB_LINEAR_ASSERT(skb);
2340 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2342 void *tmp = __skb_put(skb, len);
2344 memset(tmp, 0, len);
2348 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2351 void *tmp = __skb_put(skb, len);
2353 memcpy(tmp, data, len);
2357 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2359 *(u8 *)__skb_put(skb, 1) = val;
2362 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2364 void *tmp = skb_put(skb, len);
2366 memset(tmp, 0, len);
2371 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2374 void *tmp = skb_put(skb, len);
2376 memcpy(tmp, data, len);
2381 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2383 *(u8 *)skb_put(skb, 1) = val;
2386 void *skb_push(struct sk_buff *skb, unsigned int len);
2387 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2394 void *skb_pull(struct sk_buff *skb, unsigned int len);
2395 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2398 BUG_ON(skb->len < skb->data_len);
2399 return skb->data += len;
2402 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2404 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2407 void *skb_pull_data(struct sk_buff *skb, size_t len);
2409 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2411 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2413 if (len > skb_headlen(skb) &&
2414 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2417 return skb->data += len;
2420 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2422 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2425 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2427 if (likely(len <= skb_headlen(skb)))
2429 if (unlikely(len > skb->len))
2431 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2434 void skb_condense(struct sk_buff *skb);
2437 * skb_headroom - bytes at buffer head
2438 * @skb: buffer to check
2440 * Return the number of bytes of free space at the head of an &sk_buff.
2442 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2444 return skb->data - skb->head;
2448 * skb_tailroom - bytes at buffer end
2449 * @skb: buffer to check
2451 * Return the number of bytes of free space at the tail of an sk_buff
2453 static inline int skb_tailroom(const struct sk_buff *skb)
2455 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2459 * skb_availroom - bytes at buffer end
2460 * @skb: buffer to check
2462 * Return the number of bytes of free space at the tail of an sk_buff
2463 * allocated by sk_stream_alloc()
2465 static inline int skb_availroom(const struct sk_buff *skb)
2467 if (skb_is_nonlinear(skb))
2470 return skb->end - skb->tail - skb->reserved_tailroom;
2474 * skb_reserve - adjust headroom
2475 * @skb: buffer to alter
2476 * @len: bytes to move
2478 * Increase the headroom of an empty &sk_buff by reducing the tail
2479 * room. This is only allowed for an empty buffer.
2481 static inline void skb_reserve(struct sk_buff *skb, int len)
2488 * skb_tailroom_reserve - adjust reserved_tailroom
2489 * @skb: buffer to alter
2490 * @mtu: maximum amount of headlen permitted
2491 * @needed_tailroom: minimum amount of reserved_tailroom
2493 * Set reserved_tailroom so that headlen can be as large as possible but
2494 * not larger than mtu and tailroom cannot be smaller than
2496 * The required headroom should already have been reserved before using
2499 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2500 unsigned int needed_tailroom)
2502 SKB_LINEAR_ASSERT(skb);
2503 if (mtu < skb_tailroom(skb) - needed_tailroom)
2504 /* use at most mtu */
2505 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2507 /* use up to all available space */
2508 skb->reserved_tailroom = needed_tailroom;
2511 #define ENCAP_TYPE_ETHER 0
2512 #define ENCAP_TYPE_IPPROTO 1
2514 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2517 skb->inner_protocol = protocol;
2518 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2521 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2524 skb->inner_ipproto = ipproto;
2525 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2528 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2530 skb->inner_mac_header = skb->mac_header;
2531 skb->inner_network_header = skb->network_header;
2532 skb->inner_transport_header = skb->transport_header;
2535 static inline void skb_reset_mac_len(struct sk_buff *skb)
2537 skb->mac_len = skb->network_header - skb->mac_header;
2540 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2543 return skb->head + skb->inner_transport_header;
2546 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2548 return skb_inner_transport_header(skb) - skb->data;
2551 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2553 skb->inner_transport_header = skb->data - skb->head;
2556 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2559 skb_reset_inner_transport_header(skb);
2560 skb->inner_transport_header += offset;
2563 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2565 return skb->head + skb->inner_network_header;
2568 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2570 skb->inner_network_header = skb->data - skb->head;
2573 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2576 skb_reset_inner_network_header(skb);
2577 skb->inner_network_header += offset;
2580 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2582 return skb->head + skb->inner_mac_header;
2585 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2587 skb->inner_mac_header = skb->data - skb->head;
2590 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2593 skb_reset_inner_mac_header(skb);
2594 skb->inner_mac_header += offset;
2596 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2598 return skb->transport_header != (typeof(skb->transport_header))~0U;
2601 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2603 return skb->head + skb->transport_header;
2606 static inline void skb_reset_transport_header(struct sk_buff *skb)
2608 skb->transport_header = skb->data - skb->head;
2611 static inline void skb_set_transport_header(struct sk_buff *skb,
2614 skb_reset_transport_header(skb);
2615 skb->transport_header += offset;
2618 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2620 return skb->head + skb->network_header;
2623 static inline void skb_reset_network_header(struct sk_buff *skb)
2625 skb->network_header = skb->data - skb->head;
2628 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2630 skb_reset_network_header(skb);
2631 skb->network_header += offset;
2634 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2636 return skb->head + skb->mac_header;
2639 static inline int skb_mac_offset(const struct sk_buff *skb)
2641 return skb_mac_header(skb) - skb->data;
2644 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2646 return skb->network_header - skb->mac_header;
2649 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2651 return skb->mac_header != (typeof(skb->mac_header))~0U;
2654 static inline void skb_unset_mac_header(struct sk_buff *skb)
2656 skb->mac_header = (typeof(skb->mac_header))~0U;
2659 static inline void skb_reset_mac_header(struct sk_buff *skb)
2661 skb->mac_header = skb->data - skb->head;
2664 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2666 skb_reset_mac_header(skb);
2667 skb->mac_header += offset;
2670 static inline void skb_pop_mac_header(struct sk_buff *skb)
2672 skb->mac_header = skb->network_header;
2675 static inline void skb_probe_transport_header(struct sk_buff *skb)
2677 struct flow_keys_basic keys;
2679 if (skb_transport_header_was_set(skb))
2682 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2684 skb_set_transport_header(skb, keys.control.thoff);
2687 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2689 if (skb_mac_header_was_set(skb)) {
2690 const unsigned char *old_mac = skb_mac_header(skb);
2692 skb_set_mac_header(skb, -skb->mac_len);
2693 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2697 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2699 return skb->csum_start - skb_headroom(skb);
2702 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2704 return skb->head + skb->csum_start;
2707 static inline int skb_transport_offset(const struct sk_buff *skb)
2709 return skb_transport_header(skb) - skb->data;
2712 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2714 return skb->transport_header - skb->network_header;
2717 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2719 return skb->inner_transport_header - skb->inner_network_header;
2722 static inline int skb_network_offset(const struct sk_buff *skb)
2724 return skb_network_header(skb) - skb->data;
2727 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2729 return skb_inner_network_header(skb) - skb->data;
2732 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2734 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2738 * CPUs often take a performance hit when accessing unaligned memory
2739 * locations. The actual performance hit varies, it can be small if the
2740 * hardware handles it or large if we have to take an exception and fix it
2743 * Since an ethernet header is 14 bytes network drivers often end up with
2744 * the IP header at an unaligned offset. The IP header can be aligned by
2745 * shifting the start of the packet by 2 bytes. Drivers should do this
2748 * skb_reserve(skb, NET_IP_ALIGN);
2750 * The downside to this alignment of the IP header is that the DMA is now
2751 * unaligned. On some architectures the cost of an unaligned DMA is high
2752 * and this cost outweighs the gains made by aligning the IP header.
2754 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2757 #ifndef NET_IP_ALIGN
2758 #define NET_IP_ALIGN 2
2762 * The networking layer reserves some headroom in skb data (via
2763 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2764 * the header has to grow. In the default case, if the header has to grow
2765 * 32 bytes or less we avoid the reallocation.
2767 * Unfortunately this headroom changes the DMA alignment of the resulting
2768 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2769 * on some architectures. An architecture can override this value,
2770 * perhaps setting it to a cacheline in size (since that will maintain
2771 * cacheline alignment of the DMA). It must be a power of 2.
2773 * Various parts of the networking layer expect at least 32 bytes of
2774 * headroom, you should not reduce this.
2776 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2777 * to reduce average number of cache lines per packet.
2778 * get_rps_cpu() for example only access one 64 bytes aligned block :
2779 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2782 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2785 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2787 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2789 if (WARN_ON(skb_is_nonlinear(skb)))
2792 skb_set_tail_pointer(skb, len);
2795 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2797 __skb_set_length(skb, len);
2800 void skb_trim(struct sk_buff *skb, unsigned int len);
2802 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2805 return ___pskb_trim(skb, len);
2806 __skb_trim(skb, len);
2810 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2812 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2816 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2817 * @skb: buffer to alter
2820 * This is identical to pskb_trim except that the caller knows that
2821 * the skb is not cloned so we should never get an error due to out-
2824 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2826 int err = pskb_trim(skb, len);
2830 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2832 unsigned int diff = len - skb->len;
2834 if (skb_tailroom(skb) < diff) {
2835 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2840 __skb_set_length(skb, len);
2845 * skb_orphan - orphan a buffer
2846 * @skb: buffer to orphan
2848 * If a buffer currently has an owner then we call the owner's
2849 * destructor function and make the @skb unowned. The buffer continues
2850 * to exist but is no longer charged to its former owner.
2852 static inline void skb_orphan(struct sk_buff *skb)
2854 if (skb->destructor) {
2855 skb->destructor(skb);
2856 skb->destructor = NULL;
2864 * skb_orphan_frags - orphan the frags contained in a buffer
2865 * @skb: buffer to orphan frags from
2866 * @gfp_mask: allocation mask for replacement pages
2868 * For each frag in the SKB which needs a destructor (i.e. has an
2869 * owner) create a copy of that frag and release the original
2870 * page by calling the destructor.
2872 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2874 if (likely(!skb_zcopy(skb)))
2876 if (!skb_zcopy_is_nouarg(skb) &&
2877 skb_uarg(skb)->callback == msg_zerocopy_callback)
2879 return skb_copy_ubufs(skb, gfp_mask);
2882 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2883 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2885 if (likely(!skb_zcopy(skb)))
2887 return skb_copy_ubufs(skb, gfp_mask);
2891 * __skb_queue_purge - empty a list
2892 * @list: list to empty
2894 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2895 * the list and one reference dropped. This function does not take the
2896 * list lock and the caller must hold the relevant locks to use it.
2898 static inline void __skb_queue_purge(struct sk_buff_head *list)
2900 struct sk_buff *skb;
2901 while ((skb = __skb_dequeue(list)) != NULL)
2904 void skb_queue_purge(struct sk_buff_head *list);
2906 unsigned int skb_rbtree_purge(struct rb_root *root);
2908 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2911 * netdev_alloc_frag - allocate a page fragment
2912 * @fragsz: fragment size
2914 * Allocates a frag from a page for receive buffer.
2915 * Uses GFP_ATOMIC allocations.
2917 static inline void *netdev_alloc_frag(unsigned int fragsz)
2919 return __netdev_alloc_frag_align(fragsz, ~0u);
2922 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2925 WARN_ON_ONCE(!is_power_of_2(align));
2926 return __netdev_alloc_frag_align(fragsz, -align);
2929 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2933 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2934 * @dev: network device to receive on
2935 * @length: length to allocate
2937 * Allocate a new &sk_buff and assign it a usage count of one. The
2938 * buffer has unspecified headroom built in. Users should allocate
2939 * the headroom they think they need without accounting for the
2940 * built in space. The built in space is used for optimisations.
2942 * %NULL is returned if there is no free memory. Although this function
2943 * allocates memory it can be called from an interrupt.
2945 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2946 unsigned int length)
2948 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2951 /* legacy helper around __netdev_alloc_skb() */
2952 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2955 return __netdev_alloc_skb(NULL, length, gfp_mask);
2958 /* legacy helper around netdev_alloc_skb() */
2959 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2961 return netdev_alloc_skb(NULL, length);
2965 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2966 unsigned int length, gfp_t gfp)
2968 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2970 if (NET_IP_ALIGN && skb)
2971 skb_reserve(skb, NET_IP_ALIGN);
2975 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2976 unsigned int length)
2978 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2981 static inline void skb_free_frag(void *addr)
2983 page_frag_free(addr);
2986 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2988 static inline void *napi_alloc_frag(unsigned int fragsz)
2990 return __napi_alloc_frag_align(fragsz, ~0u);
2993 static inline void *napi_alloc_frag_align(unsigned int fragsz,
2996 WARN_ON_ONCE(!is_power_of_2(align));
2997 return __napi_alloc_frag_align(fragsz, -align);
3000 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3001 unsigned int length, gfp_t gfp_mask);
3002 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3003 unsigned int length)
3005 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3007 void napi_consume_skb(struct sk_buff *skb, int budget);
3009 void napi_skb_free_stolen_head(struct sk_buff *skb);
3010 void __kfree_skb_defer(struct sk_buff *skb);
3013 * __dev_alloc_pages - allocate page for network Rx
3014 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3015 * @order: size of the allocation
3017 * Allocate a new page.
3019 * %NULL is returned if there is no free memory.
3021 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3024 /* This piece of code contains several assumptions.
3025 * 1. This is for device Rx, therefor a cold page is preferred.
3026 * 2. The expectation is the user wants a compound page.
3027 * 3. If requesting a order 0 page it will not be compound
3028 * due to the check to see if order has a value in prep_new_page
3029 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3030 * code in gfp_to_alloc_flags that should be enforcing this.
3032 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3034 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3037 static inline struct page *dev_alloc_pages(unsigned int order)
3039 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3043 * __dev_alloc_page - allocate a page for network Rx
3044 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3046 * Allocate a new page.
3048 * %NULL is returned if there is no free memory.
3050 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3052 return __dev_alloc_pages(gfp_mask, 0);
3055 static inline struct page *dev_alloc_page(void)
3057 return dev_alloc_pages(0);
3061 * dev_page_is_reusable - check whether a page can be reused for network Rx
3062 * @page: the page to test
3064 * A page shouldn't be considered for reusing/recycling if it was allocated
3065 * under memory pressure or at a distant memory node.
3067 * Returns false if this page should be returned to page allocator, true
3070 static inline bool dev_page_is_reusable(const struct page *page)
3072 return likely(page_to_nid(page) == numa_mem_id() &&
3073 !page_is_pfmemalloc(page));
3077 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3078 * @page: The page that was allocated from skb_alloc_page
3079 * @skb: The skb that may need pfmemalloc set
3081 static inline void skb_propagate_pfmemalloc(const struct page *page,
3082 struct sk_buff *skb)
3084 if (page_is_pfmemalloc(page))
3085 skb->pfmemalloc = true;
3089 * skb_frag_off() - Returns the offset of a skb fragment
3090 * @frag: the paged fragment
3092 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3094 return frag->bv_offset;
3098 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3099 * @frag: skb fragment
3100 * @delta: value to add
3102 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3104 frag->bv_offset += delta;
3108 * skb_frag_off_set() - Sets the offset of a skb fragment
3109 * @frag: skb fragment
3110 * @offset: offset of fragment
3112 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3114 frag->bv_offset = offset;
3118 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3119 * @fragto: skb fragment where offset is set
3120 * @fragfrom: skb fragment offset is copied from
3122 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3123 const skb_frag_t *fragfrom)
3125 fragto->bv_offset = fragfrom->bv_offset;
3129 * skb_frag_page - retrieve the page referred to by a paged fragment
3130 * @frag: the paged fragment
3132 * Returns the &struct page associated with @frag.
3134 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3136 return frag->bv_page;
3140 * __skb_frag_ref - take an addition reference on a paged fragment.
3141 * @frag: the paged fragment
3143 * Takes an additional reference on the paged fragment @frag.
3145 static inline void __skb_frag_ref(skb_frag_t *frag)
3147 get_page(skb_frag_page(frag));
3151 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3153 * @f: the fragment offset.
3155 * Takes an additional reference on the @f'th paged fragment of @skb.
3157 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3159 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3163 * __skb_frag_unref - release a reference on a paged fragment.
3164 * @frag: the paged fragment
3165 * @recycle: recycle the page if allocated via page_pool
3167 * Releases a reference on the paged fragment @frag
3168 * or recycles the page via the page_pool API.
3170 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3172 struct page *page = skb_frag_page(frag);
3174 #ifdef CONFIG_PAGE_POOL
3175 if (recycle && page_pool_return_skb_page(page))
3182 * skb_frag_unref - release a reference on a paged fragment of an skb.
3184 * @f: the fragment offset
3186 * Releases a reference on the @f'th paged fragment of @skb.
3188 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3190 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3194 * skb_frag_address - gets the address of the data contained in a paged fragment
3195 * @frag: the paged fragment buffer
3197 * Returns the address of the data within @frag. The page must already
3200 static inline void *skb_frag_address(const skb_frag_t *frag)
3202 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3206 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3207 * @frag: the paged fragment buffer
3209 * Returns the address of the data within @frag. Checks that the page
3210 * is mapped and returns %NULL otherwise.
3212 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3214 void *ptr = page_address(skb_frag_page(frag));
3218 return ptr + skb_frag_off(frag);
3222 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3223 * @fragto: skb fragment where page is set
3224 * @fragfrom: skb fragment page is copied from
3226 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3227 const skb_frag_t *fragfrom)
3229 fragto->bv_page = fragfrom->bv_page;
3233 * __skb_frag_set_page - sets the page contained in a paged fragment
3234 * @frag: the paged fragment
3235 * @page: the page to set
3237 * Sets the fragment @frag to contain @page.
3239 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3241 frag->bv_page = page;
3245 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3247 * @f: the fragment offset
3248 * @page: the page to set
3250 * Sets the @f'th fragment of @skb to contain @page.
3252 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3255 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3258 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3261 * skb_frag_dma_map - maps a paged fragment via the DMA API
3262 * @dev: the device to map the fragment to
3263 * @frag: the paged fragment to map
3264 * @offset: the offset within the fragment (starting at the
3265 * fragment's own offset)
3266 * @size: the number of bytes to map
3267 * @dir: the direction of the mapping (``PCI_DMA_*``)
3269 * Maps the page associated with @frag to @device.
3271 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3272 const skb_frag_t *frag,
3273 size_t offset, size_t size,
3274 enum dma_data_direction dir)
3276 return dma_map_page(dev, skb_frag_page(frag),
3277 skb_frag_off(frag) + offset, size, dir);
3280 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3283 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3287 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3290 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3295 * skb_clone_writable - is the header of a clone writable
3296 * @skb: buffer to check
3297 * @len: length up to which to write
3299 * Returns true if modifying the header part of the cloned buffer
3300 * does not requires the data to be copied.
3302 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3304 return !skb_header_cloned(skb) &&
3305 skb_headroom(skb) + len <= skb->hdr_len;
3308 static inline int skb_try_make_writable(struct sk_buff *skb,
3309 unsigned int write_len)
3311 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3312 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3315 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3320 if (headroom > skb_headroom(skb))
3321 delta = headroom - skb_headroom(skb);
3323 if (delta || cloned)
3324 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3330 * skb_cow - copy header of skb when it is required
3331 * @skb: buffer to cow
3332 * @headroom: needed headroom
3334 * If the skb passed lacks sufficient headroom or its data part
3335 * is shared, data is reallocated. If reallocation fails, an error
3336 * is returned and original skb is not changed.
3338 * The result is skb with writable area skb->head...skb->tail
3339 * and at least @headroom of space at head.
3341 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3343 return __skb_cow(skb, headroom, skb_cloned(skb));
3347 * skb_cow_head - skb_cow but only making the head writable
3348 * @skb: buffer to cow
3349 * @headroom: needed headroom
3351 * This function is identical to skb_cow except that we replace the
3352 * skb_cloned check by skb_header_cloned. It should be used when
3353 * you only need to push on some header and do not need to modify
3356 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3358 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3362 * skb_padto - pad an skbuff up to a minimal size
3363 * @skb: buffer to pad
3364 * @len: minimal length
3366 * Pads up a buffer to ensure the trailing bytes exist and are
3367 * blanked. If the buffer already contains sufficient data it
3368 * is untouched. Otherwise it is extended. Returns zero on
3369 * success. The skb is freed on error.
3371 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3373 unsigned int size = skb->len;
3374 if (likely(size >= len))
3376 return skb_pad(skb, len - size);
3380 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3381 * @skb: buffer to pad
3382 * @len: minimal length
3383 * @free_on_error: free buffer on error
3385 * Pads up a buffer to ensure the trailing bytes exist and are
3386 * blanked. If the buffer already contains sufficient data it
3387 * is untouched. Otherwise it is extended. Returns zero on
3388 * success. The skb is freed on error if @free_on_error is true.
3390 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3394 unsigned int size = skb->len;
3396 if (unlikely(size < len)) {
3398 if (__skb_pad(skb, len, free_on_error))
3400 __skb_put(skb, len);
3406 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3407 * @skb: buffer to pad
3408 * @len: minimal length
3410 * Pads up a buffer to ensure the trailing bytes exist and are
3411 * blanked. If the buffer already contains sufficient data it
3412 * is untouched. Otherwise it is extended. Returns zero on
3413 * success. The skb is freed on error.
3415 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3417 return __skb_put_padto(skb, len, true);
3420 static inline int skb_add_data(struct sk_buff *skb,
3421 struct iov_iter *from, int copy)
3423 const int off = skb->len;
3425 if (skb->ip_summed == CHECKSUM_NONE) {
3427 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3429 skb->csum = csum_block_add(skb->csum, csum, off);
3432 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3435 __skb_trim(skb, off);
3439 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3440 const struct page *page, int off)
3445 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3447 return page == skb_frag_page(frag) &&
3448 off == skb_frag_off(frag) + skb_frag_size(frag);
3453 static inline int __skb_linearize(struct sk_buff *skb)
3455 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3459 * skb_linearize - convert paged skb to linear one
3460 * @skb: buffer to linarize
3462 * If there is no free memory -ENOMEM is returned, otherwise zero
3463 * is returned and the old skb data released.
3465 static inline int skb_linearize(struct sk_buff *skb)
3467 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3471 * skb_has_shared_frag - can any frag be overwritten
3472 * @skb: buffer to test
3474 * Return true if the skb has at least one frag that might be modified
3475 * by an external entity (as in vmsplice()/sendfile())
3477 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3479 return skb_is_nonlinear(skb) &&
3480 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3484 * skb_linearize_cow - make sure skb is linear and writable
3485 * @skb: buffer to process
3487 * If there is no free memory -ENOMEM is returned, otherwise zero
3488 * is returned and the old skb data released.
3490 static inline int skb_linearize_cow(struct sk_buff *skb)
3492 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3493 __skb_linearize(skb) : 0;
3496 static __always_inline void
3497 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3500 if (skb->ip_summed == CHECKSUM_COMPLETE)
3501 skb->csum = csum_block_sub(skb->csum,
3502 csum_partial(start, len, 0), off);
3503 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3504 skb_checksum_start_offset(skb) < 0)
3505 skb->ip_summed = CHECKSUM_NONE;
3509 * skb_postpull_rcsum - update checksum for received skb after pull
3510 * @skb: buffer to update
3511 * @start: start of data before pull
3512 * @len: length of data pulled
3514 * After doing a pull on a received packet, you need to call this to
3515 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3516 * CHECKSUM_NONE so that it can be recomputed from scratch.
3518 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3519 const void *start, unsigned int len)
3521 if (skb->ip_summed == CHECKSUM_COMPLETE)
3522 skb->csum = wsum_negate(csum_partial(start, len,
3523 wsum_negate(skb->csum)));
3524 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3525 skb_checksum_start_offset(skb) < 0)
3526 skb->ip_summed = CHECKSUM_NONE;
3529 static __always_inline void
3530 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3533 if (skb->ip_summed == CHECKSUM_COMPLETE)
3534 skb->csum = csum_block_add(skb->csum,
3535 csum_partial(start, len, 0), off);
3539 * skb_postpush_rcsum - update checksum for received skb after push
3540 * @skb: buffer to update
3541 * @start: start of data after push
3542 * @len: length of data pushed
3544 * After doing a push on a received packet, you need to call this to
3545 * update the CHECKSUM_COMPLETE checksum.
3547 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3548 const void *start, unsigned int len)
3550 __skb_postpush_rcsum(skb, start, len, 0);
3553 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3556 * skb_push_rcsum - push skb and update receive checksum
3557 * @skb: buffer to update
3558 * @len: length of data pulled
3560 * This function performs an skb_push on the packet and updates
3561 * the CHECKSUM_COMPLETE checksum. It should be used on
3562 * receive path processing instead of skb_push unless you know
3563 * that the checksum difference is zero (e.g., a valid IP header)
3564 * or you are setting ip_summed to CHECKSUM_NONE.
3566 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3569 skb_postpush_rcsum(skb, skb->data, len);
3573 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3575 * pskb_trim_rcsum - trim received skb and update checksum
3576 * @skb: buffer to trim
3579 * This is exactly the same as pskb_trim except that it ensures the
3580 * checksum of received packets are still valid after the operation.
3581 * It can change skb pointers.
3584 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3586 if (likely(len >= skb->len))
3588 return pskb_trim_rcsum_slow(skb, len);
3591 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3593 if (skb->ip_summed == CHECKSUM_COMPLETE)
3594 skb->ip_summed = CHECKSUM_NONE;
3595 __skb_trim(skb, len);
3599 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3601 if (skb->ip_summed == CHECKSUM_COMPLETE)
3602 skb->ip_summed = CHECKSUM_NONE;
3603 return __skb_grow(skb, len);
3606 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3607 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3608 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3609 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3610 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3612 #define skb_queue_walk(queue, skb) \
3613 for (skb = (queue)->next; \
3614 skb != (struct sk_buff *)(queue); \
3617 #define skb_queue_walk_safe(queue, skb, tmp) \
3618 for (skb = (queue)->next, tmp = skb->next; \
3619 skb != (struct sk_buff *)(queue); \
3620 skb = tmp, tmp = skb->next)
3622 #define skb_queue_walk_from(queue, skb) \
3623 for (; skb != (struct sk_buff *)(queue); \
3626 #define skb_rbtree_walk(skb, root) \
3627 for (skb = skb_rb_first(root); skb != NULL; \
3628 skb = skb_rb_next(skb))
3630 #define skb_rbtree_walk_from(skb) \
3631 for (; skb != NULL; \
3632 skb = skb_rb_next(skb))
3634 #define skb_rbtree_walk_from_safe(skb, tmp) \
3635 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3638 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3639 for (tmp = skb->next; \
3640 skb != (struct sk_buff *)(queue); \
3641 skb = tmp, tmp = skb->next)
3643 #define skb_queue_reverse_walk(queue, skb) \
3644 for (skb = (queue)->prev; \
3645 skb != (struct sk_buff *)(queue); \
3648 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3649 for (skb = (queue)->prev, tmp = skb->prev; \
3650 skb != (struct sk_buff *)(queue); \
3651 skb = tmp, tmp = skb->prev)
3653 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3654 for (tmp = skb->prev; \
3655 skb != (struct sk_buff *)(queue); \
3656 skb = tmp, tmp = skb->prev)
3658 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3660 return skb_shinfo(skb)->frag_list != NULL;
3663 static inline void skb_frag_list_init(struct sk_buff *skb)
3665 skb_shinfo(skb)->frag_list = NULL;
3668 #define skb_walk_frags(skb, iter) \
3669 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3672 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3673 int *err, long *timeo_p,
3674 const struct sk_buff *skb);
3675 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3676 struct sk_buff_head *queue,
3679 struct sk_buff **last);
3680 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3681 struct sk_buff_head *queue,
3682 unsigned int flags, int *off, int *err,
3683 struct sk_buff **last);
3684 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3685 struct sk_buff_head *sk_queue,
3686 unsigned int flags, int *off, int *err);
3687 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3689 __poll_t datagram_poll(struct file *file, struct socket *sock,
3690 struct poll_table_struct *wait);
3691 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3692 struct iov_iter *to, int size);
3693 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3694 struct msghdr *msg, int size)
3696 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3698 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3699 struct msghdr *msg);
3700 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3701 struct iov_iter *to, int len,
3702 struct ahash_request *hash);
3703 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3704 struct iov_iter *from, int len);
3705 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3706 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3707 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3708 static inline void skb_free_datagram_locked(struct sock *sk,
3709 struct sk_buff *skb)
3711 __skb_free_datagram_locked(sk, skb, 0);
3713 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3714 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3715 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3716 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3718 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3719 struct pipe_inode_info *pipe, unsigned int len,
3720 unsigned int flags);
3721 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3723 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3724 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3725 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3726 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3728 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3729 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3730 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3731 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3732 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3733 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3734 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3735 unsigned int offset);
3736 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3737 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3738 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3739 int skb_vlan_pop(struct sk_buff *skb);
3740 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3741 int skb_eth_pop(struct sk_buff *skb);
3742 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3743 const unsigned char *src);
3744 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3745 int mac_len, bool ethernet);
3746 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3748 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3749 int skb_mpls_dec_ttl(struct sk_buff *skb);
3750 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3753 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3755 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3758 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3760 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3763 struct skb_checksum_ops {
3764 __wsum (*update)(const void *mem, int len, __wsum wsum);
3765 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3768 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3770 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3771 __wsum csum, const struct skb_checksum_ops *ops);
3772 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3775 static inline void * __must_check
3776 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3777 const void *data, int hlen, void *buffer)
3779 if (likely(hlen - offset >= len))
3780 return (void *)data + offset;
3782 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3788 static inline void * __must_check
3789 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3791 return __skb_header_pointer(skb, offset, len, skb->data,
3792 skb_headlen(skb), buffer);
3796 * skb_needs_linearize - check if we need to linearize a given skb
3797 * depending on the given device features.
3798 * @skb: socket buffer to check
3799 * @features: net device features
3801 * Returns true if either:
3802 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3803 * 2. skb is fragmented and the device does not support SG.
3805 static inline bool skb_needs_linearize(struct sk_buff *skb,
3806 netdev_features_t features)
3808 return skb_is_nonlinear(skb) &&
3809 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3810 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3813 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3815 const unsigned int len)
3817 memcpy(to, skb->data, len);
3820 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3821 const int offset, void *to,
3822 const unsigned int len)
3824 memcpy(to, skb->data + offset, len);
3827 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3829 const unsigned int len)
3831 memcpy(skb->data, from, len);
3834 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3837 const unsigned int len)
3839 memcpy(skb->data + offset, from, len);
3842 void skb_init(void);
3844 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3850 * skb_get_timestamp - get timestamp from a skb
3851 * @skb: skb to get stamp from
3852 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3854 * Timestamps are stored in the skb as offsets to a base timestamp.
3855 * This function converts the offset back to a struct timeval and stores
3858 static inline void skb_get_timestamp(const struct sk_buff *skb,
3859 struct __kernel_old_timeval *stamp)
3861 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3864 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3865 struct __kernel_sock_timeval *stamp)
3867 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3869 stamp->tv_sec = ts.tv_sec;
3870 stamp->tv_usec = ts.tv_nsec / 1000;
3873 static inline void skb_get_timestampns(const struct sk_buff *skb,
3874 struct __kernel_old_timespec *stamp)
3876 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3878 stamp->tv_sec = ts.tv_sec;
3879 stamp->tv_nsec = ts.tv_nsec;
3882 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3883 struct __kernel_timespec *stamp)
3885 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3887 stamp->tv_sec = ts.tv_sec;
3888 stamp->tv_nsec = ts.tv_nsec;
3891 static inline void __net_timestamp(struct sk_buff *skb)
3893 skb->tstamp = ktime_get_real();
3896 static inline ktime_t net_timedelta(ktime_t t)
3898 return ktime_sub(ktime_get_real(), t);
3901 static inline ktime_t net_invalid_timestamp(void)
3906 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3908 return skb_shinfo(skb)->meta_len;
3911 static inline void *skb_metadata_end(const struct sk_buff *skb)
3913 return skb_mac_header(skb);
3916 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3917 const struct sk_buff *skb_b,
3920 const void *a = skb_metadata_end(skb_a);
3921 const void *b = skb_metadata_end(skb_b);
3922 /* Using more efficient varaiant than plain call to memcmp(). */
3923 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3927 #define __it(x, op) (x -= sizeof(u##op))
3928 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3929 case 32: diffs |= __it_diff(a, b, 64);
3931 case 24: diffs |= __it_diff(a, b, 64);
3933 case 16: diffs |= __it_diff(a, b, 64);
3935 case 8: diffs |= __it_diff(a, b, 64);
3937 case 28: diffs |= __it_diff(a, b, 64);
3939 case 20: diffs |= __it_diff(a, b, 64);
3941 case 12: diffs |= __it_diff(a, b, 64);
3943 case 4: diffs |= __it_diff(a, b, 32);
3948 return memcmp(a - meta_len, b - meta_len, meta_len);
3952 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3953 const struct sk_buff *skb_b)
3955 u8 len_a = skb_metadata_len(skb_a);
3956 u8 len_b = skb_metadata_len(skb_b);
3958 if (!(len_a | len_b))
3961 return len_a != len_b ?
3962 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3965 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3967 skb_shinfo(skb)->meta_len = meta_len;
3970 static inline void skb_metadata_clear(struct sk_buff *skb)
3972 skb_metadata_set(skb, 0);
3975 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3977 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3979 void skb_clone_tx_timestamp(struct sk_buff *skb);
3980 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3982 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3984 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3988 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3993 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3996 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3998 * PHY drivers may accept clones of transmitted packets for
3999 * timestamping via their phy_driver.txtstamp method. These drivers
4000 * must call this function to return the skb back to the stack with a
4003 * @skb: clone of the original outgoing packet
4004 * @hwtstamps: hardware time stamps
4007 void skb_complete_tx_timestamp(struct sk_buff *skb,
4008 struct skb_shared_hwtstamps *hwtstamps);
4010 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4011 struct skb_shared_hwtstamps *hwtstamps,
4012 struct sock *sk, int tstype);
4015 * skb_tstamp_tx - queue clone of skb with send time stamps
4016 * @orig_skb: the original outgoing packet
4017 * @hwtstamps: hardware time stamps, may be NULL if not available
4019 * If the skb has a socket associated, then this function clones the
4020 * skb (thus sharing the actual data and optional structures), stores
4021 * the optional hardware time stamping information (if non NULL) or
4022 * generates a software time stamp (otherwise), then queues the clone
4023 * to the error queue of the socket. Errors are silently ignored.
4025 void skb_tstamp_tx(struct sk_buff *orig_skb,
4026 struct skb_shared_hwtstamps *hwtstamps);
4029 * skb_tx_timestamp() - Driver hook for transmit timestamping
4031 * Ethernet MAC Drivers should call this function in their hard_xmit()
4032 * function immediately before giving the sk_buff to the MAC hardware.
4034 * Specifically, one should make absolutely sure that this function is
4035 * called before TX completion of this packet can trigger. Otherwise
4036 * the packet could potentially already be freed.
4038 * @skb: A socket buffer.
4040 static inline void skb_tx_timestamp(struct sk_buff *skb)
4042 skb_clone_tx_timestamp(skb);
4043 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4044 skb_tstamp_tx(skb, NULL);
4048 * skb_complete_wifi_ack - deliver skb with wifi status
4050 * @skb: the original outgoing packet
4051 * @acked: ack status
4054 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4056 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4057 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4059 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4061 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4063 (skb->ip_summed == CHECKSUM_PARTIAL &&
4064 skb_checksum_start_offset(skb) >= 0));
4068 * skb_checksum_complete - Calculate checksum of an entire packet
4069 * @skb: packet to process
4071 * This function calculates the checksum over the entire packet plus
4072 * the value of skb->csum. The latter can be used to supply the
4073 * checksum of a pseudo header as used by TCP/UDP. It returns the
4076 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4077 * this function can be used to verify that checksum on received
4078 * packets. In that case the function should return zero if the
4079 * checksum is correct. In particular, this function will return zero
4080 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4081 * hardware has already verified the correctness of the checksum.
4083 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4085 return skb_csum_unnecessary(skb) ?
4086 0 : __skb_checksum_complete(skb);
4089 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4091 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4092 if (skb->csum_level == 0)
4093 skb->ip_summed = CHECKSUM_NONE;
4099 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4101 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4102 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4104 } else if (skb->ip_summed == CHECKSUM_NONE) {
4105 skb->ip_summed = CHECKSUM_UNNECESSARY;
4106 skb->csum_level = 0;
4110 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4112 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4113 skb->ip_summed = CHECKSUM_NONE;
4114 skb->csum_level = 0;
4118 /* Check if we need to perform checksum complete validation.
4120 * Returns true if checksum complete is needed, false otherwise
4121 * (either checksum is unnecessary or zero checksum is allowed).
4123 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4127 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4128 skb->csum_valid = 1;
4129 __skb_decr_checksum_unnecessary(skb);
4136 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4139 #define CHECKSUM_BREAK 76
4141 /* Unset checksum-complete
4143 * Unset checksum complete can be done when packet is being modified
4144 * (uncompressed for instance) and checksum-complete value is
4147 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4149 if (skb->ip_summed == CHECKSUM_COMPLETE)
4150 skb->ip_summed = CHECKSUM_NONE;
4153 /* Validate (init) checksum based on checksum complete.
4156 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4157 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4158 * checksum is stored in skb->csum for use in __skb_checksum_complete
4159 * non-zero: value of invalid checksum
4162 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4166 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4167 if (!csum_fold(csum_add(psum, skb->csum))) {
4168 skb->csum_valid = 1;
4175 if (complete || skb->len <= CHECKSUM_BREAK) {
4178 csum = __skb_checksum_complete(skb);
4179 skb->csum_valid = !csum;
4186 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4191 /* Perform checksum validate (init). Note that this is a macro since we only
4192 * want to calculate the pseudo header which is an input function if necessary.
4193 * First we try to validate without any computation (checksum unnecessary) and
4194 * then calculate based on checksum complete calling the function to compute
4198 * 0: checksum is validated or try to in skb_checksum_complete
4199 * non-zero: value of invalid checksum
4201 #define __skb_checksum_validate(skb, proto, complete, \
4202 zero_okay, check, compute_pseudo) \
4204 __sum16 __ret = 0; \
4205 skb->csum_valid = 0; \
4206 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4207 __ret = __skb_checksum_validate_complete(skb, \
4208 complete, compute_pseudo(skb, proto)); \
4212 #define skb_checksum_init(skb, proto, compute_pseudo) \
4213 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4215 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4216 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4218 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4219 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4221 #define skb_checksum_validate_zero_check(skb, proto, check, \
4223 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4225 #define skb_checksum_simple_validate(skb) \
4226 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4228 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4230 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4233 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4235 skb->csum = ~pseudo;
4236 skb->ip_summed = CHECKSUM_COMPLETE;
4239 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4241 if (__skb_checksum_convert_check(skb)) \
4242 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4245 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4246 u16 start, u16 offset)
4248 skb->ip_summed = CHECKSUM_PARTIAL;
4249 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4250 skb->csum_offset = offset - start;
4253 /* Update skbuf and packet to reflect the remote checksum offload operation.
4254 * When called, ptr indicates the starting point for skb->csum when
4255 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4256 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4258 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4259 int start, int offset, bool nopartial)
4264 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4268 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4269 __skb_checksum_complete(skb);
4270 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4273 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4275 /* Adjust skb->csum since we changed the packet */
4276 skb->csum = csum_add(skb->csum, delta);
4279 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4281 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4282 return (void *)(skb->_nfct & NFCT_PTRMASK);
4288 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4290 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4297 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4299 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4300 skb->slow_gro |= !!nfct;
4305 #ifdef CONFIG_SKB_EXTENSIONS
4307 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4313 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4316 #if IS_ENABLED(CONFIG_MPTCP)
4319 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4322 SKB_EXT_NUM, /* must be last */
4326 * struct skb_ext - sk_buff extensions
4327 * @refcnt: 1 on allocation, deallocated on 0
4328 * @offset: offset to add to @data to obtain extension address
4329 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4330 * @data: start of extension data, variable sized
4332 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4333 * to use 'u8' types while allowing up to 2kb worth of extension data.
4337 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4338 u8 chunks; /* same */
4339 char data[] __aligned(8);
4342 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4343 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4344 struct skb_ext *ext);
4345 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4346 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4347 void __skb_ext_put(struct skb_ext *ext);
4349 static inline void skb_ext_put(struct sk_buff *skb)
4351 if (skb->active_extensions)
4352 __skb_ext_put(skb->extensions);
4355 static inline void __skb_ext_copy(struct sk_buff *dst,
4356 const struct sk_buff *src)
4358 dst->active_extensions = src->active_extensions;
4360 if (src->active_extensions) {
4361 struct skb_ext *ext = src->extensions;
4363 refcount_inc(&ext->refcnt);
4364 dst->extensions = ext;
4368 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4371 __skb_ext_copy(dst, src);
4374 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4376 return !!ext->offset[i];
4379 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4381 return skb->active_extensions & (1 << id);
4384 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4386 if (skb_ext_exist(skb, id))
4387 __skb_ext_del(skb, id);
4390 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4392 if (skb_ext_exist(skb, id)) {
4393 struct skb_ext *ext = skb->extensions;
4395 return (void *)ext + (ext->offset[id] << 3);
4401 static inline void skb_ext_reset(struct sk_buff *skb)
4403 if (unlikely(skb->active_extensions)) {
4404 __skb_ext_put(skb->extensions);
4405 skb->active_extensions = 0;
4409 static inline bool skb_has_extensions(struct sk_buff *skb)
4411 return unlikely(skb->active_extensions);
4414 static inline void skb_ext_put(struct sk_buff *skb) {}
4415 static inline void skb_ext_reset(struct sk_buff *skb) {}
4416 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4417 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4418 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4419 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4420 #endif /* CONFIG_SKB_EXTENSIONS */
4422 static inline void nf_reset_ct(struct sk_buff *skb)
4424 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4425 nf_conntrack_put(skb_nfct(skb));
4430 static inline void nf_reset_trace(struct sk_buff *skb)
4432 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4437 static inline void ipvs_reset(struct sk_buff *skb)
4439 #if IS_ENABLED(CONFIG_IP_VS)
4440 skb->ipvs_property = 0;
4444 /* Note: This doesn't put any conntrack info in dst. */
4445 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4448 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4449 dst->_nfct = src->_nfct;
4450 nf_conntrack_get(skb_nfct(src));
4452 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4454 dst->nf_trace = src->nf_trace;
4458 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4460 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4461 nf_conntrack_put(skb_nfct(dst));
4463 dst->slow_gro = src->slow_gro;
4464 __nf_copy(dst, src, true);
4467 #ifdef CONFIG_NETWORK_SECMARK
4468 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4470 to->secmark = from->secmark;
4473 static inline void skb_init_secmark(struct sk_buff *skb)
4478 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4481 static inline void skb_init_secmark(struct sk_buff *skb)
4485 static inline int secpath_exists(const struct sk_buff *skb)
4488 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4494 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4496 return !skb->destructor &&
4497 !secpath_exists(skb) &&
4499 !skb->_skb_refdst &&
4500 !skb_has_frag_list(skb);
4503 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4505 skb->queue_mapping = queue_mapping;
4508 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4510 return skb->queue_mapping;
4513 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4515 to->queue_mapping = from->queue_mapping;
4518 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4520 skb->queue_mapping = rx_queue + 1;
4523 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4525 return skb->queue_mapping - 1;
4528 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4530 return skb->queue_mapping != 0;
4533 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4535 skb->dst_pending_confirm = val;
4538 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4540 return skb->dst_pending_confirm != 0;
4543 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4546 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4552 /* Keeps track of mac header offset relative to skb->head.
4553 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4554 * For non-tunnel skb it points to skb_mac_header() and for
4555 * tunnel skb it points to outer mac header.
4556 * Keeps track of level of encapsulation of network headers.
4567 #define SKB_GSO_CB_OFFSET 32
4568 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4570 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4572 return (skb_mac_header(inner_skb) - inner_skb->head) -
4573 SKB_GSO_CB(inner_skb)->mac_offset;
4576 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4578 int new_headroom, headroom;
4581 headroom = skb_headroom(skb);
4582 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4586 new_headroom = skb_headroom(skb);
4587 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4591 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4593 /* Do not update partial checksums if remote checksum is enabled. */
4594 if (skb->remcsum_offload)
4597 SKB_GSO_CB(skb)->csum = res;
4598 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4601 /* Compute the checksum for a gso segment. First compute the checksum value
4602 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4603 * then add in skb->csum (checksum from csum_start to end of packet).
4604 * skb->csum and csum_start are then updated to reflect the checksum of the
4605 * resultant packet starting from the transport header-- the resultant checksum
4606 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4609 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4611 unsigned char *csum_start = skb_transport_header(skb);
4612 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4613 __wsum partial = SKB_GSO_CB(skb)->csum;
4615 SKB_GSO_CB(skb)->csum = res;
4616 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4618 return csum_fold(csum_partial(csum_start, plen, partial));
4621 static inline bool skb_is_gso(const struct sk_buff *skb)
4623 return skb_shinfo(skb)->gso_size;
4626 /* Note: Should be called only if skb_is_gso(skb) is true */
4627 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4629 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4632 /* Note: Should be called only if skb_is_gso(skb) is true */
4633 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4635 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4638 /* Note: Should be called only if skb_is_gso(skb) is true */
4639 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4641 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4644 static inline void skb_gso_reset(struct sk_buff *skb)
4646 skb_shinfo(skb)->gso_size = 0;
4647 skb_shinfo(skb)->gso_segs = 0;
4648 skb_shinfo(skb)->gso_type = 0;
4651 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4654 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4656 shinfo->gso_size += increment;
4659 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4662 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4664 shinfo->gso_size -= decrement;
4667 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4669 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4671 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4672 * wanted then gso_type will be set. */
4673 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4675 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4676 unlikely(shinfo->gso_type == 0)) {
4677 __skb_warn_lro_forwarding(skb);
4683 static inline void skb_forward_csum(struct sk_buff *skb)
4685 /* Unfortunately we don't support this one. Any brave souls? */
4686 if (skb->ip_summed == CHECKSUM_COMPLETE)
4687 skb->ip_summed = CHECKSUM_NONE;
4691 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4692 * @skb: skb to check
4694 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4695 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4696 * use this helper, to document places where we make this assertion.
4698 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4701 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4705 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4707 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4708 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4709 unsigned int transport_len,
4710 __sum16(*skb_chkf)(struct sk_buff *skb));
4713 * skb_head_is_locked - Determine if the skb->head is locked down
4714 * @skb: skb to check
4716 * The head on skbs build around a head frag can be removed if they are
4717 * not cloned. This function returns true if the skb head is locked down
4718 * due to either being allocated via kmalloc, or by being a clone with
4719 * multiple references to the head.
4721 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4723 return !skb->head_frag || skb_cloned(skb);
4726 /* Local Checksum Offload.
4727 * Compute outer checksum based on the assumption that the
4728 * inner checksum will be offloaded later.
4729 * See Documentation/networking/checksum-offloads.rst for
4730 * explanation of how this works.
4731 * Fill in outer checksum adjustment (e.g. with sum of outer
4732 * pseudo-header) before calling.
4733 * Also ensure that inner checksum is in linear data area.
4735 static inline __wsum lco_csum(struct sk_buff *skb)
4737 unsigned char *csum_start = skb_checksum_start(skb);
4738 unsigned char *l4_hdr = skb_transport_header(skb);
4741 /* Start with complement of inner checksum adjustment */
4742 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4745 /* Add in checksum of our headers (incl. outer checksum
4746 * adjustment filled in by caller) and return result.
4748 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4751 static inline bool skb_is_redirected(const struct sk_buff *skb)
4753 return skb->redirected;
4756 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4758 skb->redirected = 1;
4759 #ifdef CONFIG_NET_REDIRECT
4760 skb->from_ingress = from_ingress;
4761 if (skb->from_ingress)
4766 static inline void skb_reset_redirect(struct sk_buff *skb)
4768 skb->redirected = 0;
4771 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4773 return skb->csum_not_inet;
4776 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4777 const u64 kcov_handle)
4780 skb->kcov_handle = kcov_handle;
4784 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4787 return skb->kcov_handle;
4793 #ifdef CONFIG_PAGE_POOL
4794 static inline void skb_mark_for_recycle(struct sk_buff *skb)
4796 skb->pp_recycle = 1;
4800 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4802 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4804 return page_pool_return_skb_page(virt_to_page(data));
4807 #endif /* __KERNEL__ */
4808 #endif /* _LINUX_SKBUFF_H */