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.
294 struct sk_buff_head {
295 /* These two members must be first. */
296 struct sk_buff *next;
297 struct sk_buff *prev;
305 /* To allow 64K frame to be packed as single skb without frag_list we
306 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
307 * buffers which do not start on a page boundary.
309 * Since GRO uses frags we allocate at least 16 regardless of page
312 #if (65536/PAGE_SIZE + 1) < 16
313 #define MAX_SKB_FRAGS 16UL
315 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
317 extern int sysctl_max_skb_frags;
319 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
320 * segment using its current segmentation instead.
322 #define GSO_BY_FRAGS 0xFFFF
324 typedef struct bio_vec skb_frag_t;
327 * skb_frag_size() - Returns the size of a skb fragment
328 * @frag: skb fragment
330 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
336 * skb_frag_size_set() - Sets the size of a skb fragment
337 * @frag: skb fragment
338 * @size: size of fragment
340 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
346 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
347 * @frag: skb fragment
348 * @delta: value to add
350 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
352 frag->bv_len += delta;
356 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
357 * @frag: skb fragment
358 * @delta: value to subtract
360 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
362 frag->bv_len -= delta;
366 * skb_frag_must_loop - Test if %p is a high memory page
367 * @p: fragment's page
369 static inline bool skb_frag_must_loop(struct page *p)
371 #if defined(CONFIG_HIGHMEM)
372 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
379 * skb_frag_foreach_page - loop over pages in a fragment
381 * @f: skb frag to operate on
382 * @f_off: offset from start of f->bv_page
383 * @f_len: length from f_off to loop over
384 * @p: (temp var) current page
385 * @p_off: (temp var) offset from start of current page,
386 * non-zero only on first page.
387 * @p_len: (temp var) length in current page,
388 * < PAGE_SIZE only on first and last page.
389 * @copied: (temp var) length so far, excluding current p_len.
391 * A fragment can hold a compound page, in which case per-page
392 * operations, notably kmap_atomic, must be called for each
395 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
396 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
397 p_off = (f_off) & (PAGE_SIZE - 1), \
398 p_len = skb_frag_must_loop(p) ? \
399 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
402 copied += p_len, p++, p_off = 0, \
403 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
405 #define HAVE_HW_TIME_STAMP
408 * struct skb_shared_hwtstamps - hardware time stamps
409 * @hwtstamp: hardware time stamp transformed into duration
410 * since arbitrary point in time
412 * Software time stamps generated by ktime_get_real() are stored in
415 * hwtstamps can only be compared against other hwtstamps from
418 * This structure is attached to packets as part of the
419 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
421 struct skb_shared_hwtstamps {
425 /* Definitions for tx_flags in struct skb_shared_info */
427 /* generate hardware time stamp */
428 SKBTX_HW_TSTAMP = 1 << 0,
430 /* generate software time stamp when queueing packet to NIC */
431 SKBTX_SW_TSTAMP = 1 << 1,
433 /* device driver is going to provide hardware time stamp */
434 SKBTX_IN_PROGRESS = 1 << 2,
436 /* generate wifi status information (where possible) */
437 SKBTX_WIFI_STATUS = 1 << 4,
439 /* generate software time stamp when entering packet scheduling */
440 SKBTX_SCHED_TSTAMP = 1 << 6,
443 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
445 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
447 /* Definitions for flags in struct skb_shared_info */
449 /* use zcopy routines */
450 SKBFL_ZEROCOPY_ENABLE = BIT(0),
452 /* This indicates at least one fragment might be overwritten
453 * (as in vmsplice(), sendfile() ...)
454 * If we need to compute a TX checksum, we'll need to copy
455 * all frags to avoid possible bad checksum
457 SKBFL_SHARED_FRAG = BIT(1),
459 /* segment contains only zerocopy data and should not be
460 * charged to the kernel memory.
462 SKBFL_PURE_ZEROCOPY = BIT(2),
465 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
466 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY)
469 * The callback notifies userspace to release buffers when skb DMA is done in
470 * lower device, the skb last reference should be 0 when calling this.
471 * The zerocopy_success argument is true if zero copy transmit occurred,
472 * false on data copy or out of memory error caused by data copy attempt.
473 * The ctx field is used to track device context.
474 * The desc field is used to track userspace buffer index.
477 void (*callback)(struct sk_buff *, struct ubuf_info *,
478 bool zerocopy_success);
495 struct user_struct *user;
500 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
502 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
503 void mm_unaccount_pinned_pages(struct mmpin *mmp);
505 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
506 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
507 struct ubuf_info *uarg);
509 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
511 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
514 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
515 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
516 struct msghdr *msg, int len,
517 struct ubuf_info *uarg);
519 /* This data is invariant across clones and lives at
520 * the end of the header data, ie. at skb->end.
522 struct skb_shared_info {
527 unsigned short gso_size;
528 /* Warning: this field is not always filled in (UFO)! */
529 unsigned short gso_segs;
530 struct sk_buff *frag_list;
531 struct skb_shared_hwtstamps hwtstamps;
532 unsigned int gso_type;
536 * Warning : all fields before dataref are cleared in __alloc_skb()
540 /* Intermediate layers must ensure that destructor_arg
541 * remains valid until skb destructor */
542 void * destructor_arg;
544 /* must be last field, see pskb_expand_head() */
545 skb_frag_t frags[MAX_SKB_FRAGS];
548 /* We divide dataref into two halves. The higher 16 bits hold references
549 * to the payload part of skb->data. The lower 16 bits hold references to
550 * the entire skb->data. A clone of a headerless skb holds the length of
551 * the header in skb->hdr_len.
553 * All users must obey the rule that the skb->data reference count must be
554 * greater than or equal to the payload reference count.
556 * Holding a reference to the payload part means that the user does not
557 * care about modifications to the header part of skb->data.
559 #define SKB_DATAREF_SHIFT 16
560 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
564 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
565 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
566 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
570 SKB_GSO_TCPV4 = 1 << 0,
572 /* This indicates the skb is from an untrusted source. */
573 SKB_GSO_DODGY = 1 << 1,
575 /* This indicates the tcp segment has CWR set. */
576 SKB_GSO_TCP_ECN = 1 << 2,
578 SKB_GSO_TCP_FIXEDID = 1 << 3,
580 SKB_GSO_TCPV6 = 1 << 4,
582 SKB_GSO_FCOE = 1 << 5,
584 SKB_GSO_GRE = 1 << 6,
586 SKB_GSO_GRE_CSUM = 1 << 7,
588 SKB_GSO_IPXIP4 = 1 << 8,
590 SKB_GSO_IPXIP6 = 1 << 9,
592 SKB_GSO_UDP_TUNNEL = 1 << 10,
594 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
596 SKB_GSO_PARTIAL = 1 << 12,
598 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
600 SKB_GSO_SCTP = 1 << 14,
602 SKB_GSO_ESP = 1 << 15,
604 SKB_GSO_UDP = 1 << 16,
606 SKB_GSO_UDP_L4 = 1 << 17,
608 SKB_GSO_FRAGLIST = 1 << 18,
611 #if BITS_PER_LONG > 32
612 #define NET_SKBUFF_DATA_USES_OFFSET 1
615 #ifdef NET_SKBUFF_DATA_USES_OFFSET
616 typedef unsigned int sk_buff_data_t;
618 typedef unsigned char *sk_buff_data_t;
622 * struct sk_buff - socket buffer
623 * @next: Next buffer in list
624 * @prev: Previous buffer in list
625 * @tstamp: Time we arrived/left
626 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
627 * for retransmit timer
628 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
630 * @sk: Socket we are owned by
631 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
632 * fragmentation management
633 * @dev: Device we arrived on/are leaving by
634 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
635 * @cb: Control buffer. Free for use by every layer. Put private vars here
636 * @_skb_refdst: destination entry (with norefcount bit)
637 * @sp: the security path, used for xfrm
638 * @len: Length of actual data
639 * @data_len: Data length
640 * @mac_len: Length of link layer header
641 * @hdr_len: writable header length of cloned skb
642 * @csum: Checksum (must include start/offset pair)
643 * @csum_start: Offset from skb->head where checksumming should start
644 * @csum_offset: Offset from csum_start where checksum should be stored
645 * @priority: Packet queueing priority
646 * @ignore_df: allow local fragmentation
647 * @cloned: Head may be cloned (check refcnt to be sure)
648 * @ip_summed: Driver fed us an IP checksum
649 * @nohdr: Payload reference only, must not modify header
650 * @pkt_type: Packet class
651 * @fclone: skbuff clone status
652 * @ipvs_property: skbuff is owned by ipvs
653 * @inner_protocol_type: whether the inner protocol is
654 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
655 * @remcsum_offload: remote checksum offload is enabled
656 * @offload_fwd_mark: Packet was L2-forwarded in hardware
657 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
658 * @tc_skip_classify: do not classify packet. set by IFB device
659 * @tc_at_ingress: used within tc_classify to distinguish in/egress
660 * @redirected: packet was redirected by packet classifier
661 * @from_ingress: packet was redirected from the ingress path
662 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
663 * @peeked: this packet has been seen already, so stats have been
664 * done for it, don't do them again
665 * @nf_trace: netfilter packet trace flag
666 * @protocol: Packet protocol from driver
667 * @destructor: Destruct function
668 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
669 * @_sk_redir: socket redirection information for skmsg
670 * @_nfct: Associated connection, if any (with nfctinfo bits)
671 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
672 * @skb_iif: ifindex of device we arrived on
673 * @tc_index: Traffic control index
674 * @hash: the packet hash
675 * @queue_mapping: Queue mapping for multiqueue devices
676 * @head_frag: skb was allocated from page fragments,
677 * not allocated by kmalloc() or vmalloc().
678 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
679 * @pp_recycle: mark the packet for recycling instead of freeing (implies
680 * page_pool support on driver)
681 * @active_extensions: active extensions (skb_ext_id types)
682 * @ndisc_nodetype: router type (from link layer)
683 * @ooo_okay: allow the mapping of a socket to a queue to be changed
684 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
686 * @sw_hash: indicates hash was computed in software stack
687 * @wifi_acked_valid: wifi_acked was set
688 * @wifi_acked: whether frame was acked on wifi or not
689 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
690 * @encapsulation: indicates the inner headers in the skbuff are valid
691 * @encap_hdr_csum: software checksum is needed
692 * @csum_valid: checksum is already valid
693 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
694 * @csum_complete_sw: checksum was completed by software
695 * @csum_level: indicates the number of consecutive checksums found in
696 * the packet minus one that have been verified as
697 * CHECKSUM_UNNECESSARY (max 3)
698 * @dst_pending_confirm: need to confirm neighbour
699 * @decrypted: Decrypted SKB
700 * @slow_gro: state present at GRO time, slower prepare step required
701 * @napi_id: id of the NAPI struct this skb came from
702 * @sender_cpu: (aka @napi_id) source CPU in XPS
703 * @secmark: security marking
704 * @mark: Generic packet mark
705 * @reserved_tailroom: (aka @mark) number of bytes of free space available
706 * at the tail of an sk_buff
707 * @vlan_present: VLAN tag is present
708 * @vlan_proto: vlan encapsulation protocol
709 * @vlan_tci: vlan tag control information
710 * @inner_protocol: Protocol (encapsulation)
711 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
712 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
713 * @inner_transport_header: Inner transport layer header (encapsulation)
714 * @inner_network_header: Network layer header (encapsulation)
715 * @inner_mac_header: Link layer header (encapsulation)
716 * @transport_header: Transport layer header
717 * @network_header: Network layer header
718 * @mac_header: Link layer header
719 * @kcov_handle: KCOV remote handle for remote coverage collection
720 * @tail: Tail pointer
722 * @head: Head of buffer
723 * @data: Data head pointer
724 * @truesize: Buffer size
725 * @users: User count - see {datagram,tcp}.c
726 * @extensions: allocated extensions, valid if active_extensions is nonzero
732 /* These two members must be first. */
733 struct sk_buff *next;
734 struct sk_buff *prev;
737 struct net_device *dev;
738 /* Some protocols might use this space to store information,
739 * while device pointer would be NULL.
740 * UDP receive path is one user.
742 unsigned long dev_scratch;
745 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
746 struct list_head list;
747 struct llist_node ll_node;
752 int ip_defrag_offset;
757 u64 skb_mstamp_ns; /* earliest departure time */
760 * This is the control buffer. It is free to use for every
761 * layer. Please put your private variables there. If you
762 * want to keep them across layers you have to do a skb_clone()
763 * first. This is owned by whoever has the skb queued ATM.
765 char cb[48] __aligned(8);
769 unsigned long _skb_refdst;
770 void (*destructor)(struct sk_buff *skb);
772 struct list_head tcp_tsorted_anchor;
773 #ifdef CONFIG_NET_SOCK_MSG
774 unsigned long _sk_redir;
778 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
786 /* Following fields are _not_ copied in __copy_skb_header()
787 * Note that queue_mapping is here mostly to fill a hole.
791 /* if you move cloned around you also must adapt those constants */
792 #ifdef __BIG_ENDIAN_BITFIELD
793 #define CLONED_MASK (1 << 7)
795 #define CLONED_MASK 1
797 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
800 __u8 __cloned_offset[0];
808 pp_recycle:1; /* page_pool recycle indicator */
809 #ifdef CONFIG_SKB_EXTENSIONS
810 __u8 active_extensions;
813 /* fields enclosed in headers_start/headers_end are copied
814 * using a single memcpy() in __copy_skb_header()
817 __u32 headers_start[0];
820 /* if you move pkt_type around you also must adapt those constants */
821 #ifdef __BIG_ENDIAN_BITFIELD
822 #define PKT_TYPE_MAX (7 << 5)
824 #define PKT_TYPE_MAX 7
826 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
829 __u8 __pkt_type_offset[0];
839 __u8 wifi_acked_valid:1;
842 /* Indicates the inner headers are valid in the skbuff. */
843 __u8 encapsulation:1;
844 __u8 encap_hdr_csum:1;
847 #ifdef __BIG_ENDIAN_BITFIELD
848 #define PKT_VLAN_PRESENT_BIT 7
850 #define PKT_VLAN_PRESENT_BIT 0
852 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
854 __u8 __pkt_vlan_present_offset[0];
857 __u8 csum_complete_sw:1;
859 __u8 csum_not_inet:1;
860 __u8 dst_pending_confirm:1;
861 #ifdef CONFIG_IPV6_NDISC_NODETYPE
862 __u8 ndisc_nodetype:2;
865 __u8 ipvs_property:1;
866 __u8 inner_protocol_type:1;
867 __u8 remcsum_offload:1;
868 #ifdef CONFIG_NET_SWITCHDEV
869 __u8 offload_fwd_mark:1;
870 __u8 offload_l3_fwd_mark:1;
872 #ifdef CONFIG_NET_CLS_ACT
873 __u8 tc_skip_classify:1;
874 __u8 tc_at_ingress:1;
877 #ifdef CONFIG_NET_REDIRECT
880 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
881 __u8 nf_skip_egress:1;
883 #ifdef CONFIG_TLS_DEVICE
888 #ifdef CONFIG_NET_SCHED
889 __u16 tc_index; /* traffic control index */
904 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
906 unsigned int napi_id;
907 unsigned int sender_cpu;
910 #ifdef CONFIG_NETWORK_SECMARK
916 __u32 reserved_tailroom;
920 __be16 inner_protocol;
924 __u16 inner_transport_header;
925 __u16 inner_network_header;
926 __u16 inner_mac_header;
929 __u16 transport_header;
930 __u16 network_header;
938 __u32 headers_end[0];
941 /* These elements must be at the end, see alloc_skb() for details. */
946 unsigned int truesize;
949 #ifdef CONFIG_SKB_EXTENSIONS
950 /* only useable after checking ->active_extensions != 0 */
951 struct skb_ext *extensions;
957 * Handling routines are only of interest to the kernel
960 #define SKB_ALLOC_FCLONE 0x01
961 #define SKB_ALLOC_RX 0x02
962 #define SKB_ALLOC_NAPI 0x04
965 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
968 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
970 return unlikely(skb->pfmemalloc);
974 * skb might have a dst pointer attached, refcounted or not.
975 * _skb_refdst low order bit is set if refcount was _not_ taken
977 #define SKB_DST_NOREF 1UL
978 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
981 * skb_dst - returns skb dst_entry
984 * Returns skb dst_entry, regardless of reference taken or not.
986 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
988 /* If refdst was not refcounted, check we still are in a
989 * rcu_read_lock section
991 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
992 !rcu_read_lock_held() &&
993 !rcu_read_lock_bh_held());
994 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
998 * skb_dst_set - sets skb dst
1002 * Sets skb dst, assuming a reference was taken on dst and should
1003 * be released by skb_dst_drop()
1005 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1007 skb->slow_gro |= !!dst;
1008 skb->_skb_refdst = (unsigned long)dst;
1012 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1016 * Sets skb dst, assuming a reference was not taken on dst.
1017 * If dst entry is cached, we do not take reference and dst_release
1018 * will be avoided by refdst_drop. If dst entry is not cached, we take
1019 * reference, so that last dst_release can destroy the dst immediately.
1021 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1023 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1024 skb->slow_gro |= !!dst;
1025 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1029 * skb_dst_is_noref - Test if skb dst isn't refcounted
1032 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1034 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1038 * skb_rtable - Returns the skb &rtable
1041 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1043 return (struct rtable *)skb_dst(skb);
1046 /* For mangling skb->pkt_type from user space side from applications
1047 * such as nft, tc, etc, we only allow a conservative subset of
1048 * possible pkt_types to be set.
1050 static inline bool skb_pkt_type_ok(u32 ptype)
1052 return ptype <= PACKET_OTHERHOST;
1056 * skb_napi_id - Returns the skb's NAPI id
1059 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1061 #ifdef CONFIG_NET_RX_BUSY_POLL
1062 return skb->napi_id;
1069 * skb_unref - decrement the skb's reference count
1072 * Returns true if we can free the skb.
1074 static inline bool skb_unref(struct sk_buff *skb)
1078 if (likely(refcount_read(&skb->users) == 1))
1080 else if (likely(!refcount_dec_and_test(&skb->users)))
1086 void skb_release_head_state(struct sk_buff *skb);
1087 void kfree_skb(struct sk_buff *skb);
1088 void kfree_skb_list(struct sk_buff *segs);
1089 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1090 void skb_tx_error(struct sk_buff *skb);
1092 #ifdef CONFIG_TRACEPOINTS
1093 void consume_skb(struct sk_buff *skb);
1095 static inline void consume_skb(struct sk_buff *skb)
1097 return kfree_skb(skb);
1101 void __consume_stateless_skb(struct sk_buff *skb);
1102 void __kfree_skb(struct sk_buff *skb);
1103 extern struct kmem_cache *skbuff_head_cache;
1105 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1106 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1107 bool *fragstolen, int *delta_truesize);
1109 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1111 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1112 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1113 struct sk_buff *build_skb_around(struct sk_buff *skb,
1114 void *data, unsigned int frag_size);
1116 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1119 * alloc_skb - allocate a network buffer
1120 * @size: size to allocate
1121 * @priority: allocation mask
1123 * This function is a convenient wrapper around __alloc_skb().
1125 static inline struct sk_buff *alloc_skb(unsigned int size,
1128 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1131 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1132 unsigned long data_len,
1136 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1138 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1139 struct sk_buff_fclones {
1140 struct sk_buff skb1;
1142 struct sk_buff skb2;
1144 refcount_t fclone_ref;
1148 * skb_fclone_busy - check if fclone is busy
1152 * Returns true if skb is a fast clone, and its clone is not freed.
1153 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1154 * so we also check that this didnt happen.
1156 static inline bool skb_fclone_busy(const struct sock *sk,
1157 const struct sk_buff *skb)
1159 const struct sk_buff_fclones *fclones;
1161 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1163 return skb->fclone == SKB_FCLONE_ORIG &&
1164 refcount_read(&fclones->fclone_ref) > 1 &&
1165 READ_ONCE(fclones->skb2.sk) == sk;
1169 * alloc_skb_fclone - allocate a network buffer from fclone cache
1170 * @size: size to allocate
1171 * @priority: allocation mask
1173 * This function is a convenient wrapper around __alloc_skb().
1175 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1178 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1181 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1182 void skb_headers_offset_update(struct sk_buff *skb, int off);
1183 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1184 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1185 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1186 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1187 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1188 gfp_t gfp_mask, bool fclone);
1189 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1192 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1195 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1196 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1197 unsigned int headroom);
1198 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1199 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1200 int newtailroom, gfp_t priority);
1201 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1202 int offset, int len);
1203 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1204 int offset, int len);
1205 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1206 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1209 * skb_pad - zero pad the tail of an skb
1210 * @skb: buffer to pad
1211 * @pad: space to pad
1213 * Ensure that a buffer is followed by a padding area that is zero
1214 * filled. Used by network drivers which may DMA or transfer data
1215 * beyond the buffer end onto the wire.
1217 * May return error in out of memory cases. The skb is freed on error.
1219 static inline int skb_pad(struct sk_buff *skb, int pad)
1221 return __skb_pad(skb, pad, true);
1223 #define dev_kfree_skb(a) consume_skb(a)
1225 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1226 int offset, size_t size);
1228 struct skb_seq_state {
1232 __u32 stepped_offset;
1233 struct sk_buff *root_skb;
1234 struct sk_buff *cur_skb;
1239 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1240 unsigned int to, struct skb_seq_state *st);
1241 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1242 struct skb_seq_state *st);
1243 void skb_abort_seq_read(struct skb_seq_state *st);
1245 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1246 unsigned int to, struct ts_config *config);
1249 * Packet hash types specify the type of hash in skb_set_hash.
1251 * Hash types refer to the protocol layer addresses which are used to
1252 * construct a packet's hash. The hashes are used to differentiate or identify
1253 * flows of the protocol layer for the hash type. Hash types are either
1254 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1256 * Properties of hashes:
1258 * 1) Two packets in different flows have different hash values
1259 * 2) Two packets in the same flow should have the same hash value
1261 * A hash at a higher layer is considered to be more specific. A driver should
1262 * set the most specific hash possible.
1264 * A driver cannot indicate a more specific hash than the layer at which a hash
1265 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1267 * A driver may indicate a hash level which is less specific than the
1268 * actual layer the hash was computed on. For instance, a hash computed
1269 * at L4 may be considered an L3 hash. This should only be done if the
1270 * driver can't unambiguously determine that the HW computed the hash at
1271 * the higher layer. Note that the "should" in the second property above
1274 enum pkt_hash_types {
1275 PKT_HASH_TYPE_NONE, /* Undefined type */
1276 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1277 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1278 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1281 static inline void skb_clear_hash(struct sk_buff *skb)
1288 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1291 skb_clear_hash(skb);
1295 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1297 skb->l4_hash = is_l4;
1298 skb->sw_hash = is_sw;
1303 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1305 /* Used by drivers to set hash from HW */
1306 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1310 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1312 __skb_set_hash(skb, hash, true, is_l4);
1315 void __skb_get_hash(struct sk_buff *skb);
1316 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1317 u32 skb_get_poff(const struct sk_buff *skb);
1318 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1319 const struct flow_keys_basic *keys, int hlen);
1320 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1321 const void *data, int hlen_proto);
1323 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1324 int thoff, u8 ip_proto)
1326 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1329 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1330 const struct flow_dissector_key *key,
1331 unsigned int key_count);
1333 struct bpf_flow_dissector;
1334 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1335 __be16 proto, int nhoff, int hlen, unsigned int flags);
1337 bool __skb_flow_dissect(const struct net *net,
1338 const struct sk_buff *skb,
1339 struct flow_dissector *flow_dissector,
1340 void *target_container, const void *data,
1341 __be16 proto, int nhoff, int hlen, unsigned int flags);
1343 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1344 struct flow_dissector *flow_dissector,
1345 void *target_container, unsigned int flags)
1347 return __skb_flow_dissect(NULL, skb, flow_dissector,
1348 target_container, NULL, 0, 0, 0, flags);
1351 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1352 struct flow_keys *flow,
1355 memset(flow, 0, sizeof(*flow));
1356 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1357 flow, NULL, 0, 0, 0, flags);
1361 skb_flow_dissect_flow_keys_basic(const struct net *net,
1362 const struct sk_buff *skb,
1363 struct flow_keys_basic *flow,
1364 const void *data, __be16 proto,
1365 int nhoff, int hlen, unsigned int flags)
1367 memset(flow, 0, sizeof(*flow));
1368 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1369 data, proto, nhoff, hlen, flags);
1372 void skb_flow_dissect_meta(const struct sk_buff *skb,
1373 struct flow_dissector *flow_dissector,
1374 void *target_container);
1376 /* Gets a skb connection tracking info, ctinfo map should be a
1377 * map of mapsize to translate enum ip_conntrack_info states
1381 skb_flow_dissect_ct(const struct sk_buff *skb,
1382 struct flow_dissector *flow_dissector,
1383 void *target_container,
1384 u16 *ctinfo_map, size_t mapsize,
1387 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1388 struct flow_dissector *flow_dissector,
1389 void *target_container);
1391 void skb_flow_dissect_hash(const struct sk_buff *skb,
1392 struct flow_dissector *flow_dissector,
1393 void *target_container);
1395 static inline __u32 skb_get_hash(struct sk_buff *skb)
1397 if (!skb->l4_hash && !skb->sw_hash)
1398 __skb_get_hash(skb);
1403 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1405 if (!skb->l4_hash && !skb->sw_hash) {
1406 struct flow_keys keys;
1407 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1409 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1415 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1416 const siphash_key_t *perturb);
1418 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1423 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1425 to->hash = from->hash;
1426 to->sw_hash = from->sw_hash;
1427 to->l4_hash = from->l4_hash;
1430 static inline void skb_copy_decrypted(struct sk_buff *to,
1431 const struct sk_buff *from)
1433 #ifdef CONFIG_TLS_DEVICE
1434 to->decrypted = from->decrypted;
1438 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1439 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1441 return skb->head + skb->end;
1444 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1449 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1454 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1456 return skb->end - skb->head;
1461 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1463 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1465 return &skb_shinfo(skb)->hwtstamps;
1468 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1470 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1472 return is_zcopy ? skb_uarg(skb) : NULL;
1475 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1477 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1480 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1481 const struct sk_buff *skb2)
1483 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1486 static inline void net_zcopy_get(struct ubuf_info *uarg)
1488 refcount_inc(&uarg->refcnt);
1491 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1493 skb_shinfo(skb)->destructor_arg = uarg;
1494 skb_shinfo(skb)->flags |= uarg->flags;
1497 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1500 if (skb && uarg && !skb_zcopy(skb)) {
1501 if (unlikely(have_ref && *have_ref))
1504 net_zcopy_get(uarg);
1505 skb_zcopy_init(skb, uarg);
1509 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1511 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1512 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1515 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1517 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1520 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1522 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1525 static inline void net_zcopy_put(struct ubuf_info *uarg)
1528 uarg->callback(NULL, uarg, true);
1531 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1534 if (uarg->callback == msg_zerocopy_callback)
1535 msg_zerocopy_put_abort(uarg, have_uref);
1537 net_zcopy_put(uarg);
1541 /* Release a reference on a zerocopy structure */
1542 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1544 struct ubuf_info *uarg = skb_zcopy(skb);
1547 if (!skb_zcopy_is_nouarg(skb))
1548 uarg->callback(skb, uarg, zerocopy_success);
1550 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1554 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1559 /* Iterate through singly-linked GSO fragments of an skb. */
1560 #define skb_list_walk_safe(first, skb, next_skb) \
1561 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1562 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1564 static inline void skb_list_del_init(struct sk_buff *skb)
1566 __list_del_entry(&skb->list);
1567 skb_mark_not_on_list(skb);
1571 * skb_queue_empty - check if a queue is empty
1574 * Returns true if the queue is empty, false otherwise.
1576 static inline int skb_queue_empty(const struct sk_buff_head *list)
1578 return list->next == (const struct sk_buff *) list;
1582 * skb_queue_empty_lockless - check if a queue is empty
1585 * Returns true if the queue is empty, false otherwise.
1586 * This variant can be used in lockless contexts.
1588 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1590 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1595 * skb_queue_is_last - check if skb is the last entry in the queue
1599 * Returns true if @skb is the last buffer on the list.
1601 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1602 const struct sk_buff *skb)
1604 return skb->next == (const struct sk_buff *) list;
1608 * skb_queue_is_first - check if skb is the first entry in the queue
1612 * Returns true if @skb is the first buffer on the list.
1614 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1615 const struct sk_buff *skb)
1617 return skb->prev == (const struct sk_buff *) list;
1621 * skb_queue_next - return the next packet in the queue
1623 * @skb: current buffer
1625 * Return the next packet in @list after @skb. It is only valid to
1626 * call this if skb_queue_is_last() evaluates to false.
1628 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1629 const struct sk_buff *skb)
1631 /* This BUG_ON may seem severe, but if we just return then we
1632 * are going to dereference garbage.
1634 BUG_ON(skb_queue_is_last(list, skb));
1639 * skb_queue_prev - return the prev packet in the queue
1641 * @skb: current buffer
1643 * Return the prev packet in @list before @skb. It is only valid to
1644 * call this if skb_queue_is_first() evaluates to false.
1646 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1647 const struct sk_buff *skb)
1649 /* This BUG_ON may seem severe, but if we just return then we
1650 * are going to dereference garbage.
1652 BUG_ON(skb_queue_is_first(list, skb));
1657 * skb_get - reference buffer
1658 * @skb: buffer to reference
1660 * Makes another reference to a socket buffer and returns a pointer
1663 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1665 refcount_inc(&skb->users);
1670 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1674 * skb_cloned - is the buffer a clone
1675 * @skb: buffer to check
1677 * Returns true if the buffer was generated with skb_clone() and is
1678 * one of multiple shared copies of the buffer. Cloned buffers are
1679 * shared data so must not be written to under normal circumstances.
1681 static inline int skb_cloned(const struct sk_buff *skb)
1683 return skb->cloned &&
1684 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1687 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1689 might_sleep_if(gfpflags_allow_blocking(pri));
1691 if (skb_cloned(skb))
1692 return pskb_expand_head(skb, 0, 0, pri);
1697 /* This variant of skb_unclone() makes sure skb->truesize is not changed */
1698 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1700 might_sleep_if(gfpflags_allow_blocking(pri));
1702 if (skb_cloned(skb)) {
1703 unsigned int save = skb->truesize;
1706 res = pskb_expand_head(skb, 0, 0, pri);
1707 skb->truesize = save;
1714 * skb_header_cloned - is the header a clone
1715 * @skb: buffer to check
1717 * Returns true if modifying the header part of the buffer requires
1718 * the data to be copied.
1720 static inline int skb_header_cloned(const struct sk_buff *skb)
1727 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1728 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1729 return dataref != 1;
1732 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1734 might_sleep_if(gfpflags_allow_blocking(pri));
1736 if (skb_header_cloned(skb))
1737 return pskb_expand_head(skb, 0, 0, pri);
1743 * __skb_header_release - release reference to header
1744 * @skb: buffer to operate on
1746 static inline void __skb_header_release(struct sk_buff *skb)
1749 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1754 * skb_shared - is the buffer shared
1755 * @skb: buffer to check
1757 * Returns true if more than one person has a reference to this
1760 static inline int skb_shared(const struct sk_buff *skb)
1762 return refcount_read(&skb->users) != 1;
1766 * skb_share_check - check if buffer is shared and if so clone it
1767 * @skb: buffer to check
1768 * @pri: priority for memory allocation
1770 * If the buffer is shared the buffer is cloned and the old copy
1771 * drops a reference. A new clone with a single reference is returned.
1772 * If the buffer is not shared the original buffer is returned. When
1773 * being called from interrupt status or with spinlocks held pri must
1776 * NULL is returned on a memory allocation failure.
1778 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1780 might_sleep_if(gfpflags_allow_blocking(pri));
1781 if (skb_shared(skb)) {
1782 struct sk_buff *nskb = skb_clone(skb, pri);
1794 * Copy shared buffers into a new sk_buff. We effectively do COW on
1795 * packets to handle cases where we have a local reader and forward
1796 * and a couple of other messy ones. The normal one is tcpdumping
1797 * a packet thats being forwarded.
1801 * skb_unshare - make a copy of a shared buffer
1802 * @skb: buffer to check
1803 * @pri: priority for memory allocation
1805 * If the socket buffer is a clone then this function creates a new
1806 * copy of the data, drops a reference count on the old copy and returns
1807 * the new copy with the reference count at 1. If the buffer is not a clone
1808 * the original buffer is returned. When called with a spinlock held or
1809 * from interrupt state @pri must be %GFP_ATOMIC
1811 * %NULL is returned on a memory allocation failure.
1813 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1816 might_sleep_if(gfpflags_allow_blocking(pri));
1817 if (skb_cloned(skb)) {
1818 struct sk_buff *nskb = skb_copy(skb, pri);
1820 /* Free our shared copy */
1831 * skb_peek - peek at the head of an &sk_buff_head
1832 * @list_: list to peek at
1834 * Peek an &sk_buff. Unlike most other operations you _MUST_
1835 * be careful with this one. A peek leaves the buffer on the
1836 * list and someone else may run off with it. You must hold
1837 * the appropriate locks or have a private queue to do this.
1839 * Returns %NULL for an empty list or a pointer to the head element.
1840 * The reference count is not incremented and the reference is therefore
1841 * volatile. Use with caution.
1843 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1845 struct sk_buff *skb = list_->next;
1847 if (skb == (struct sk_buff *)list_)
1853 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1854 * @list_: list to peek at
1856 * Like skb_peek(), but the caller knows that the list is not empty.
1858 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1864 * skb_peek_next - peek skb following the given one from a queue
1865 * @skb: skb to start from
1866 * @list_: list to peek at
1868 * Returns %NULL when the end of the list is met or a pointer to the
1869 * next element. The reference count is not incremented and the
1870 * reference is therefore volatile. Use with caution.
1872 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1873 const struct sk_buff_head *list_)
1875 struct sk_buff *next = skb->next;
1877 if (next == (struct sk_buff *)list_)
1883 * skb_peek_tail - peek at the tail of an &sk_buff_head
1884 * @list_: list to peek at
1886 * Peek an &sk_buff. Unlike most other operations you _MUST_
1887 * be careful with this one. A peek leaves the buffer on the
1888 * list and someone else may run off with it. You must hold
1889 * the appropriate locks or have a private queue to do this.
1891 * Returns %NULL for an empty list or a pointer to the tail element.
1892 * The reference count is not incremented and the reference is therefore
1893 * volatile. Use with caution.
1895 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1897 struct sk_buff *skb = READ_ONCE(list_->prev);
1899 if (skb == (struct sk_buff *)list_)
1906 * skb_queue_len - get queue length
1907 * @list_: list to measure
1909 * Return the length of an &sk_buff queue.
1911 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1917 * skb_queue_len_lockless - get queue length
1918 * @list_: list to measure
1920 * Return the length of an &sk_buff queue.
1921 * This variant can be used in lockless contexts.
1923 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1925 return READ_ONCE(list_->qlen);
1929 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1930 * @list: queue to initialize
1932 * This initializes only the list and queue length aspects of
1933 * an sk_buff_head object. This allows to initialize the list
1934 * aspects of an sk_buff_head without reinitializing things like
1935 * the spinlock. It can also be used for on-stack sk_buff_head
1936 * objects where the spinlock is known to not be used.
1938 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1940 list->prev = list->next = (struct sk_buff *)list;
1945 * This function creates a split out lock class for each invocation;
1946 * this is needed for now since a whole lot of users of the skb-queue
1947 * infrastructure in drivers have different locking usage (in hardirq)
1948 * than the networking core (in softirq only). In the long run either the
1949 * network layer or drivers should need annotation to consolidate the
1950 * main types of usage into 3 classes.
1952 static inline void skb_queue_head_init(struct sk_buff_head *list)
1954 spin_lock_init(&list->lock);
1955 __skb_queue_head_init(list);
1958 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1959 struct lock_class_key *class)
1961 skb_queue_head_init(list);
1962 lockdep_set_class(&list->lock, class);
1966 * Insert an sk_buff on a list.
1968 * The "__skb_xxxx()" functions are the non-atomic ones that
1969 * can only be called with interrupts disabled.
1971 static inline void __skb_insert(struct sk_buff *newsk,
1972 struct sk_buff *prev, struct sk_buff *next,
1973 struct sk_buff_head *list)
1975 /* See skb_queue_empty_lockless() and skb_peek_tail()
1976 * for the opposite READ_ONCE()
1978 WRITE_ONCE(newsk->next, next);
1979 WRITE_ONCE(newsk->prev, prev);
1980 WRITE_ONCE(next->prev, newsk);
1981 WRITE_ONCE(prev->next, newsk);
1982 WRITE_ONCE(list->qlen, list->qlen + 1);
1985 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1986 struct sk_buff *prev,
1987 struct sk_buff *next)
1989 struct sk_buff *first = list->next;
1990 struct sk_buff *last = list->prev;
1992 WRITE_ONCE(first->prev, prev);
1993 WRITE_ONCE(prev->next, first);
1995 WRITE_ONCE(last->next, next);
1996 WRITE_ONCE(next->prev, last);
2000 * skb_queue_splice - join two skb lists, this is designed for stacks
2001 * @list: the new list to add
2002 * @head: the place to add it in the first list
2004 static inline void skb_queue_splice(const struct sk_buff_head *list,
2005 struct sk_buff_head *head)
2007 if (!skb_queue_empty(list)) {
2008 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2009 head->qlen += list->qlen;
2014 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2015 * @list: the new list to add
2016 * @head: the place to add it in the first list
2018 * The list at @list is reinitialised
2020 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2021 struct sk_buff_head *head)
2023 if (!skb_queue_empty(list)) {
2024 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2025 head->qlen += list->qlen;
2026 __skb_queue_head_init(list);
2031 * skb_queue_splice_tail - join two skb lists, each list being a queue
2032 * @list: the new list to add
2033 * @head: the place to add it in the first list
2035 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2036 struct sk_buff_head *head)
2038 if (!skb_queue_empty(list)) {
2039 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2040 head->qlen += list->qlen;
2045 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2046 * @list: the new list to add
2047 * @head: the place to add it in the first list
2049 * Each of the lists is a queue.
2050 * The list at @list is reinitialised
2052 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2053 struct sk_buff_head *head)
2055 if (!skb_queue_empty(list)) {
2056 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2057 head->qlen += list->qlen;
2058 __skb_queue_head_init(list);
2063 * __skb_queue_after - queue a buffer at the list head
2064 * @list: list to use
2065 * @prev: place after this buffer
2066 * @newsk: buffer to queue
2068 * Queue a buffer int the middle of a list. This function takes no locks
2069 * and you must therefore hold required locks before calling it.
2071 * A buffer cannot be placed on two lists at the same time.
2073 static inline void __skb_queue_after(struct sk_buff_head *list,
2074 struct sk_buff *prev,
2075 struct sk_buff *newsk)
2077 __skb_insert(newsk, prev, prev->next, list);
2080 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2081 struct sk_buff_head *list);
2083 static inline void __skb_queue_before(struct sk_buff_head *list,
2084 struct sk_buff *next,
2085 struct sk_buff *newsk)
2087 __skb_insert(newsk, next->prev, next, list);
2091 * __skb_queue_head - queue a buffer at the list head
2092 * @list: list to use
2093 * @newsk: buffer to queue
2095 * Queue a buffer at the start of a list. This function takes no locks
2096 * and you must therefore hold required locks before calling it.
2098 * A buffer cannot be placed on two lists at the same time.
2100 static inline void __skb_queue_head(struct sk_buff_head *list,
2101 struct sk_buff *newsk)
2103 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2105 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2108 * __skb_queue_tail - queue a buffer at the list tail
2109 * @list: list to use
2110 * @newsk: buffer to queue
2112 * Queue a buffer at the end of a list. This function takes no locks
2113 * and you must therefore hold required locks before calling it.
2115 * A buffer cannot be placed on two lists at the same time.
2117 static inline void __skb_queue_tail(struct sk_buff_head *list,
2118 struct sk_buff *newsk)
2120 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2122 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2125 * remove sk_buff from list. _Must_ be called atomically, and with
2128 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2129 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2131 struct sk_buff *next, *prev;
2133 WRITE_ONCE(list->qlen, list->qlen - 1);
2136 skb->next = skb->prev = NULL;
2137 WRITE_ONCE(next->prev, prev);
2138 WRITE_ONCE(prev->next, next);
2142 * __skb_dequeue - remove from the head of the queue
2143 * @list: list to dequeue from
2145 * Remove the head of the list. This function does not take any locks
2146 * so must be used with appropriate locks held only. The head item is
2147 * returned or %NULL if the list is empty.
2149 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2151 struct sk_buff *skb = skb_peek(list);
2153 __skb_unlink(skb, list);
2156 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2159 * __skb_dequeue_tail - remove from the tail of the queue
2160 * @list: list to dequeue from
2162 * Remove the tail of the list. This function does not take any locks
2163 * so must be used with appropriate locks held only. The tail item is
2164 * returned or %NULL if the list is empty.
2166 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2168 struct sk_buff *skb = skb_peek_tail(list);
2170 __skb_unlink(skb, list);
2173 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2176 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2178 return skb->data_len;
2181 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2183 return skb->len - skb->data_len;
2186 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2188 unsigned int i, len = 0;
2190 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2191 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2195 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2197 return skb_headlen(skb) + __skb_pagelen(skb);
2201 * __skb_fill_page_desc - initialise a paged fragment in an skb
2202 * @skb: buffer containing fragment to be initialised
2203 * @i: paged fragment index to initialise
2204 * @page: the page to use for this fragment
2205 * @off: the offset to the data with @page
2206 * @size: the length of the data
2208 * Initialises the @i'th fragment of @skb to point to &size bytes at
2209 * offset @off within @page.
2211 * Does not take any additional reference on the fragment.
2213 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2214 struct page *page, int off, int size)
2216 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2219 * Propagate page pfmemalloc to the skb if we can. The problem is
2220 * that not all callers have unique ownership of the page but rely
2221 * on page_is_pfmemalloc doing the right thing(tm).
2223 frag->bv_page = page;
2224 frag->bv_offset = off;
2225 skb_frag_size_set(frag, size);
2227 page = compound_head(page);
2228 if (page_is_pfmemalloc(page))
2229 skb->pfmemalloc = true;
2233 * skb_fill_page_desc - initialise a paged fragment in an skb
2234 * @skb: buffer containing fragment to be initialised
2235 * @i: paged fragment index to initialise
2236 * @page: the page to use for this fragment
2237 * @off: the offset to the data with @page
2238 * @size: the length of the data
2240 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2241 * @skb to point to @size bytes at offset @off within @page. In
2242 * addition updates @skb such that @i is the last fragment.
2244 * Does not take any additional reference on the fragment.
2246 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2247 struct page *page, int off, int size)
2249 __skb_fill_page_desc(skb, i, page, off, size);
2250 skb_shinfo(skb)->nr_frags = i + 1;
2253 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2254 int size, unsigned int truesize);
2256 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2257 unsigned int truesize);
2259 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2261 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2262 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2264 return skb->head + skb->tail;
2267 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2269 skb->tail = skb->data - skb->head;
2272 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2274 skb_reset_tail_pointer(skb);
2275 skb->tail += offset;
2278 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2279 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2284 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2286 skb->tail = skb->data;
2289 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2291 skb->tail = skb->data + offset;
2294 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2297 * Add data to an sk_buff
2299 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2300 void *skb_put(struct sk_buff *skb, unsigned int len);
2301 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2303 void *tmp = skb_tail_pointer(skb);
2304 SKB_LINEAR_ASSERT(skb);
2310 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2312 void *tmp = __skb_put(skb, len);
2314 memset(tmp, 0, len);
2318 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2321 void *tmp = __skb_put(skb, len);
2323 memcpy(tmp, data, len);
2327 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2329 *(u8 *)__skb_put(skb, 1) = val;
2332 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2334 void *tmp = skb_put(skb, len);
2336 memset(tmp, 0, len);
2341 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2344 void *tmp = skb_put(skb, len);
2346 memcpy(tmp, data, len);
2351 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2353 *(u8 *)skb_put(skb, 1) = val;
2356 void *skb_push(struct sk_buff *skb, unsigned int len);
2357 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2364 void *skb_pull(struct sk_buff *skb, unsigned int len);
2365 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2368 BUG_ON(skb->len < skb->data_len);
2369 return skb->data += len;
2372 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2374 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2377 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2379 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2381 if (len > skb_headlen(skb) &&
2382 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2385 return skb->data += len;
2388 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2390 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2393 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2395 if (likely(len <= skb_headlen(skb)))
2397 if (unlikely(len > skb->len))
2399 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2402 void skb_condense(struct sk_buff *skb);
2405 * skb_headroom - bytes at buffer head
2406 * @skb: buffer to check
2408 * Return the number of bytes of free space at the head of an &sk_buff.
2410 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2412 return skb->data - skb->head;
2416 * skb_tailroom - bytes at buffer end
2417 * @skb: buffer to check
2419 * Return the number of bytes of free space at the tail of an sk_buff
2421 static inline int skb_tailroom(const struct sk_buff *skb)
2423 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2427 * skb_availroom - bytes at buffer end
2428 * @skb: buffer to check
2430 * Return the number of bytes of free space at the tail of an sk_buff
2431 * allocated by sk_stream_alloc()
2433 static inline int skb_availroom(const struct sk_buff *skb)
2435 if (skb_is_nonlinear(skb))
2438 return skb->end - skb->tail - skb->reserved_tailroom;
2442 * skb_reserve - adjust headroom
2443 * @skb: buffer to alter
2444 * @len: bytes to move
2446 * Increase the headroom of an empty &sk_buff by reducing the tail
2447 * room. This is only allowed for an empty buffer.
2449 static inline void skb_reserve(struct sk_buff *skb, int len)
2456 * skb_tailroom_reserve - adjust reserved_tailroom
2457 * @skb: buffer to alter
2458 * @mtu: maximum amount of headlen permitted
2459 * @needed_tailroom: minimum amount of reserved_tailroom
2461 * Set reserved_tailroom so that headlen can be as large as possible but
2462 * not larger than mtu and tailroom cannot be smaller than
2464 * The required headroom should already have been reserved before using
2467 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2468 unsigned int needed_tailroom)
2470 SKB_LINEAR_ASSERT(skb);
2471 if (mtu < skb_tailroom(skb) - needed_tailroom)
2472 /* use at most mtu */
2473 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2475 /* use up to all available space */
2476 skb->reserved_tailroom = needed_tailroom;
2479 #define ENCAP_TYPE_ETHER 0
2480 #define ENCAP_TYPE_IPPROTO 1
2482 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2485 skb->inner_protocol = protocol;
2486 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2489 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2492 skb->inner_ipproto = ipproto;
2493 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2496 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2498 skb->inner_mac_header = skb->mac_header;
2499 skb->inner_network_header = skb->network_header;
2500 skb->inner_transport_header = skb->transport_header;
2503 static inline void skb_reset_mac_len(struct sk_buff *skb)
2505 skb->mac_len = skb->network_header - skb->mac_header;
2508 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2511 return skb->head + skb->inner_transport_header;
2514 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2516 return skb_inner_transport_header(skb) - skb->data;
2519 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2521 skb->inner_transport_header = skb->data - skb->head;
2524 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2527 skb_reset_inner_transport_header(skb);
2528 skb->inner_transport_header += offset;
2531 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2533 return skb->head + skb->inner_network_header;
2536 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2538 skb->inner_network_header = skb->data - skb->head;
2541 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2544 skb_reset_inner_network_header(skb);
2545 skb->inner_network_header += offset;
2548 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2550 return skb->head + skb->inner_mac_header;
2553 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2555 skb->inner_mac_header = skb->data - skb->head;
2558 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2561 skb_reset_inner_mac_header(skb);
2562 skb->inner_mac_header += offset;
2564 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2566 return skb->transport_header != (typeof(skb->transport_header))~0U;
2569 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2571 return skb->head + skb->transport_header;
2574 static inline void skb_reset_transport_header(struct sk_buff *skb)
2576 skb->transport_header = skb->data - skb->head;
2579 static inline void skb_set_transport_header(struct sk_buff *skb,
2582 skb_reset_transport_header(skb);
2583 skb->transport_header += offset;
2586 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2588 return skb->head + skb->network_header;
2591 static inline void skb_reset_network_header(struct sk_buff *skb)
2593 skb->network_header = skb->data - skb->head;
2596 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2598 skb_reset_network_header(skb);
2599 skb->network_header += offset;
2602 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2604 return skb->head + skb->mac_header;
2607 static inline int skb_mac_offset(const struct sk_buff *skb)
2609 return skb_mac_header(skb) - skb->data;
2612 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2614 return skb->network_header - skb->mac_header;
2617 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2619 return skb->mac_header != (typeof(skb->mac_header))~0U;
2622 static inline void skb_unset_mac_header(struct sk_buff *skb)
2624 skb->mac_header = (typeof(skb->mac_header))~0U;
2627 static inline void skb_reset_mac_header(struct sk_buff *skb)
2629 skb->mac_header = skb->data - skb->head;
2632 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2634 skb_reset_mac_header(skb);
2635 skb->mac_header += offset;
2638 static inline void skb_pop_mac_header(struct sk_buff *skb)
2640 skb->mac_header = skb->network_header;
2643 static inline void skb_probe_transport_header(struct sk_buff *skb)
2645 struct flow_keys_basic keys;
2647 if (skb_transport_header_was_set(skb))
2650 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2652 skb_set_transport_header(skb, keys.control.thoff);
2655 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2657 if (skb_mac_header_was_set(skb)) {
2658 const unsigned char *old_mac = skb_mac_header(skb);
2660 skb_set_mac_header(skb, -skb->mac_len);
2661 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2665 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2667 return skb->csum_start - skb_headroom(skb);
2670 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2672 return skb->head + skb->csum_start;
2675 static inline int skb_transport_offset(const struct sk_buff *skb)
2677 return skb_transport_header(skb) - skb->data;
2680 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2682 return skb->transport_header - skb->network_header;
2685 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2687 return skb->inner_transport_header - skb->inner_network_header;
2690 static inline int skb_network_offset(const struct sk_buff *skb)
2692 return skb_network_header(skb) - skb->data;
2695 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2697 return skb_inner_network_header(skb) - skb->data;
2700 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2702 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2706 * CPUs often take a performance hit when accessing unaligned memory
2707 * locations. The actual performance hit varies, it can be small if the
2708 * hardware handles it or large if we have to take an exception and fix it
2711 * Since an ethernet header is 14 bytes network drivers often end up with
2712 * the IP header at an unaligned offset. The IP header can be aligned by
2713 * shifting the start of the packet by 2 bytes. Drivers should do this
2716 * skb_reserve(skb, NET_IP_ALIGN);
2718 * The downside to this alignment of the IP header is that the DMA is now
2719 * unaligned. On some architectures the cost of an unaligned DMA is high
2720 * and this cost outweighs the gains made by aligning the IP header.
2722 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2725 #ifndef NET_IP_ALIGN
2726 #define NET_IP_ALIGN 2
2730 * The networking layer reserves some headroom in skb data (via
2731 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2732 * the header has to grow. In the default case, if the header has to grow
2733 * 32 bytes or less we avoid the reallocation.
2735 * Unfortunately this headroom changes the DMA alignment of the resulting
2736 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2737 * on some architectures. An architecture can override this value,
2738 * perhaps setting it to a cacheline in size (since that will maintain
2739 * cacheline alignment of the DMA). It must be a power of 2.
2741 * Various parts of the networking layer expect at least 32 bytes of
2742 * headroom, you should not reduce this.
2744 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2745 * to reduce average number of cache lines per packet.
2746 * get_rps_cpu() for example only access one 64 bytes aligned block :
2747 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2750 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2753 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2755 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2757 if (WARN_ON(skb_is_nonlinear(skb)))
2760 skb_set_tail_pointer(skb, len);
2763 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2765 __skb_set_length(skb, len);
2768 void skb_trim(struct sk_buff *skb, unsigned int len);
2770 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2773 return ___pskb_trim(skb, len);
2774 __skb_trim(skb, len);
2778 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2780 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2784 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2785 * @skb: buffer to alter
2788 * This is identical to pskb_trim except that the caller knows that
2789 * the skb is not cloned so we should never get an error due to out-
2792 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2794 int err = pskb_trim(skb, len);
2798 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2800 unsigned int diff = len - skb->len;
2802 if (skb_tailroom(skb) < diff) {
2803 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2808 __skb_set_length(skb, len);
2813 * skb_orphan - orphan a buffer
2814 * @skb: buffer to orphan
2816 * If a buffer currently has an owner then we call the owner's
2817 * destructor function and make the @skb unowned. The buffer continues
2818 * to exist but is no longer charged to its former owner.
2820 static inline void skb_orphan(struct sk_buff *skb)
2822 if (skb->destructor) {
2823 skb->destructor(skb);
2824 skb->destructor = NULL;
2832 * skb_orphan_frags - orphan the frags contained in a buffer
2833 * @skb: buffer to orphan frags from
2834 * @gfp_mask: allocation mask for replacement pages
2836 * For each frag in the SKB which needs a destructor (i.e. has an
2837 * owner) create a copy of that frag and release the original
2838 * page by calling the destructor.
2840 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2842 if (likely(!skb_zcopy(skb)))
2844 if (!skb_zcopy_is_nouarg(skb) &&
2845 skb_uarg(skb)->callback == msg_zerocopy_callback)
2847 return skb_copy_ubufs(skb, gfp_mask);
2850 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2851 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2853 if (likely(!skb_zcopy(skb)))
2855 return skb_copy_ubufs(skb, gfp_mask);
2859 * __skb_queue_purge - empty a list
2860 * @list: list to empty
2862 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2863 * the list and one reference dropped. This function does not take the
2864 * list lock and the caller must hold the relevant locks to use it.
2866 static inline void __skb_queue_purge(struct sk_buff_head *list)
2868 struct sk_buff *skb;
2869 while ((skb = __skb_dequeue(list)) != NULL)
2872 void skb_queue_purge(struct sk_buff_head *list);
2874 unsigned int skb_rbtree_purge(struct rb_root *root);
2876 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2879 * netdev_alloc_frag - allocate a page fragment
2880 * @fragsz: fragment size
2882 * Allocates a frag from a page for receive buffer.
2883 * Uses GFP_ATOMIC allocations.
2885 static inline void *netdev_alloc_frag(unsigned int fragsz)
2887 return __netdev_alloc_frag_align(fragsz, ~0u);
2890 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2893 WARN_ON_ONCE(!is_power_of_2(align));
2894 return __netdev_alloc_frag_align(fragsz, -align);
2897 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2901 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2902 * @dev: network device to receive on
2903 * @length: length to allocate
2905 * Allocate a new &sk_buff and assign it a usage count of one. The
2906 * buffer has unspecified headroom built in. Users should allocate
2907 * the headroom they think they need without accounting for the
2908 * built in space. The built in space is used for optimisations.
2910 * %NULL is returned if there is no free memory. Although this function
2911 * allocates memory it can be called from an interrupt.
2913 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2914 unsigned int length)
2916 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2919 /* legacy helper around __netdev_alloc_skb() */
2920 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2923 return __netdev_alloc_skb(NULL, length, gfp_mask);
2926 /* legacy helper around netdev_alloc_skb() */
2927 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2929 return netdev_alloc_skb(NULL, length);
2933 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2934 unsigned int length, gfp_t gfp)
2936 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2938 if (NET_IP_ALIGN && skb)
2939 skb_reserve(skb, NET_IP_ALIGN);
2943 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2944 unsigned int length)
2946 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2949 static inline void skb_free_frag(void *addr)
2951 page_frag_free(addr);
2954 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2956 static inline void *napi_alloc_frag(unsigned int fragsz)
2958 return __napi_alloc_frag_align(fragsz, ~0u);
2961 static inline void *napi_alloc_frag_align(unsigned int fragsz,
2964 WARN_ON_ONCE(!is_power_of_2(align));
2965 return __napi_alloc_frag_align(fragsz, -align);
2968 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2969 unsigned int length, gfp_t gfp_mask);
2970 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2971 unsigned int length)
2973 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2975 void napi_consume_skb(struct sk_buff *skb, int budget);
2977 void napi_skb_free_stolen_head(struct sk_buff *skb);
2978 void __kfree_skb_defer(struct sk_buff *skb);
2981 * __dev_alloc_pages - allocate page for network Rx
2982 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2983 * @order: size of the allocation
2985 * Allocate a new page.
2987 * %NULL is returned if there is no free memory.
2989 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2992 /* This piece of code contains several assumptions.
2993 * 1. This is for device Rx, therefor a cold page is preferred.
2994 * 2. The expectation is the user wants a compound page.
2995 * 3. If requesting a order 0 page it will not be compound
2996 * due to the check to see if order has a value in prep_new_page
2997 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2998 * code in gfp_to_alloc_flags that should be enforcing this.
3000 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3002 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3005 static inline struct page *dev_alloc_pages(unsigned int order)
3007 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3011 * __dev_alloc_page - allocate a page for network Rx
3012 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3014 * Allocate a new page.
3016 * %NULL is returned if there is no free memory.
3018 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3020 return __dev_alloc_pages(gfp_mask, 0);
3023 static inline struct page *dev_alloc_page(void)
3025 return dev_alloc_pages(0);
3029 * dev_page_is_reusable - check whether a page can be reused for network Rx
3030 * @page: the page to test
3032 * A page shouldn't be considered for reusing/recycling if it was allocated
3033 * under memory pressure or at a distant memory node.
3035 * Returns false if this page should be returned to page allocator, true
3038 static inline bool dev_page_is_reusable(const struct page *page)
3040 return likely(page_to_nid(page) == numa_mem_id() &&
3041 !page_is_pfmemalloc(page));
3045 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3046 * @page: The page that was allocated from skb_alloc_page
3047 * @skb: The skb that may need pfmemalloc set
3049 static inline void skb_propagate_pfmemalloc(const struct page *page,
3050 struct sk_buff *skb)
3052 if (page_is_pfmemalloc(page))
3053 skb->pfmemalloc = true;
3057 * skb_frag_off() - Returns the offset of a skb fragment
3058 * @frag: the paged fragment
3060 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3062 return frag->bv_offset;
3066 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3067 * @frag: skb fragment
3068 * @delta: value to add
3070 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3072 frag->bv_offset += delta;
3076 * skb_frag_off_set() - Sets the offset of a skb fragment
3077 * @frag: skb fragment
3078 * @offset: offset of fragment
3080 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3082 frag->bv_offset = offset;
3086 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3087 * @fragto: skb fragment where offset is set
3088 * @fragfrom: skb fragment offset is copied from
3090 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3091 const skb_frag_t *fragfrom)
3093 fragto->bv_offset = fragfrom->bv_offset;
3097 * skb_frag_page - retrieve the page referred to by a paged fragment
3098 * @frag: the paged fragment
3100 * Returns the &struct page associated with @frag.
3102 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3104 return frag->bv_page;
3108 * __skb_frag_ref - take an addition reference on a paged fragment.
3109 * @frag: the paged fragment
3111 * Takes an additional reference on the paged fragment @frag.
3113 static inline void __skb_frag_ref(skb_frag_t *frag)
3115 get_page(skb_frag_page(frag));
3119 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3121 * @f: the fragment offset.
3123 * Takes an additional reference on the @f'th paged fragment of @skb.
3125 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3127 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3131 * __skb_frag_unref - release a reference on a paged fragment.
3132 * @frag: the paged fragment
3133 * @recycle: recycle the page if allocated via page_pool
3135 * Releases a reference on the paged fragment @frag
3136 * or recycles the page via the page_pool API.
3138 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3140 struct page *page = skb_frag_page(frag);
3142 #ifdef CONFIG_PAGE_POOL
3143 if (recycle && page_pool_return_skb_page(page))
3150 * skb_frag_unref - release a reference on a paged fragment of an skb.
3152 * @f: the fragment offset
3154 * Releases a reference on the @f'th paged fragment of @skb.
3156 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3158 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3162 * skb_frag_address - gets the address of the data contained in a paged fragment
3163 * @frag: the paged fragment buffer
3165 * Returns the address of the data within @frag. The page must already
3168 static inline void *skb_frag_address(const skb_frag_t *frag)
3170 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3174 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3175 * @frag: the paged fragment buffer
3177 * Returns the address of the data within @frag. Checks that the page
3178 * is mapped and returns %NULL otherwise.
3180 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3182 void *ptr = page_address(skb_frag_page(frag));
3186 return ptr + skb_frag_off(frag);
3190 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3191 * @fragto: skb fragment where page is set
3192 * @fragfrom: skb fragment page is copied from
3194 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3195 const skb_frag_t *fragfrom)
3197 fragto->bv_page = fragfrom->bv_page;
3201 * __skb_frag_set_page - sets the page contained in a paged fragment
3202 * @frag: the paged fragment
3203 * @page: the page to set
3205 * Sets the fragment @frag to contain @page.
3207 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3209 frag->bv_page = page;
3213 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3215 * @f: the fragment offset
3216 * @page: the page to set
3218 * Sets the @f'th fragment of @skb to contain @page.
3220 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3223 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3226 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3229 * skb_frag_dma_map - maps a paged fragment via the DMA API
3230 * @dev: the device to map the fragment to
3231 * @frag: the paged fragment to map
3232 * @offset: the offset within the fragment (starting at the
3233 * fragment's own offset)
3234 * @size: the number of bytes to map
3235 * @dir: the direction of the mapping (``PCI_DMA_*``)
3237 * Maps the page associated with @frag to @device.
3239 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3240 const skb_frag_t *frag,
3241 size_t offset, size_t size,
3242 enum dma_data_direction dir)
3244 return dma_map_page(dev, skb_frag_page(frag),
3245 skb_frag_off(frag) + offset, size, dir);
3248 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3251 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3255 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3258 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3263 * skb_clone_writable - is the header of a clone writable
3264 * @skb: buffer to check
3265 * @len: length up to which to write
3267 * Returns true if modifying the header part of the cloned buffer
3268 * does not requires the data to be copied.
3270 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3272 return !skb_header_cloned(skb) &&
3273 skb_headroom(skb) + len <= skb->hdr_len;
3276 static inline int skb_try_make_writable(struct sk_buff *skb,
3277 unsigned int write_len)
3279 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3280 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3283 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3288 if (headroom > skb_headroom(skb))
3289 delta = headroom - skb_headroom(skb);
3291 if (delta || cloned)
3292 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3298 * skb_cow - copy header of skb when it is required
3299 * @skb: buffer to cow
3300 * @headroom: needed headroom
3302 * If the skb passed lacks sufficient headroom or its data part
3303 * is shared, data is reallocated. If reallocation fails, an error
3304 * is returned and original skb is not changed.
3306 * The result is skb with writable area skb->head...skb->tail
3307 * and at least @headroom of space at head.
3309 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3311 return __skb_cow(skb, headroom, skb_cloned(skb));
3315 * skb_cow_head - skb_cow but only making the head writable
3316 * @skb: buffer to cow
3317 * @headroom: needed headroom
3319 * This function is identical to skb_cow except that we replace the
3320 * skb_cloned check by skb_header_cloned. It should be used when
3321 * you only need to push on some header and do not need to modify
3324 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3326 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3330 * skb_padto - pad an skbuff up to a minimal size
3331 * @skb: buffer to pad
3332 * @len: minimal length
3334 * Pads up a buffer to ensure the trailing bytes exist and are
3335 * blanked. If the buffer already contains sufficient data it
3336 * is untouched. Otherwise it is extended. Returns zero on
3337 * success. The skb is freed on error.
3339 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3341 unsigned int size = skb->len;
3342 if (likely(size >= len))
3344 return skb_pad(skb, len - size);
3348 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3349 * @skb: buffer to pad
3350 * @len: minimal length
3351 * @free_on_error: free buffer on error
3353 * Pads up a buffer to ensure the trailing bytes exist and are
3354 * blanked. If the buffer already contains sufficient data it
3355 * is untouched. Otherwise it is extended. Returns zero on
3356 * success. The skb is freed on error if @free_on_error is true.
3358 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3362 unsigned int size = skb->len;
3364 if (unlikely(size < len)) {
3366 if (__skb_pad(skb, len, free_on_error))
3368 __skb_put(skb, len);
3374 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3375 * @skb: buffer to pad
3376 * @len: minimal length
3378 * Pads up a buffer to ensure the trailing bytes exist and are
3379 * blanked. If the buffer already contains sufficient data it
3380 * is untouched. Otherwise it is extended. Returns zero on
3381 * success. The skb is freed on error.
3383 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3385 return __skb_put_padto(skb, len, true);
3388 static inline int skb_add_data(struct sk_buff *skb,
3389 struct iov_iter *from, int copy)
3391 const int off = skb->len;
3393 if (skb->ip_summed == CHECKSUM_NONE) {
3395 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3397 skb->csum = csum_block_add(skb->csum, csum, off);
3400 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3403 __skb_trim(skb, off);
3407 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3408 const struct page *page, int off)
3413 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3415 return page == skb_frag_page(frag) &&
3416 off == skb_frag_off(frag) + skb_frag_size(frag);
3421 static inline int __skb_linearize(struct sk_buff *skb)
3423 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3427 * skb_linearize - convert paged skb to linear one
3428 * @skb: buffer to linarize
3430 * If there is no free memory -ENOMEM is returned, otherwise zero
3431 * is returned and the old skb data released.
3433 static inline int skb_linearize(struct sk_buff *skb)
3435 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3439 * skb_has_shared_frag - can any frag be overwritten
3440 * @skb: buffer to test
3442 * Return true if the skb has at least one frag that might be modified
3443 * by an external entity (as in vmsplice()/sendfile())
3445 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3447 return skb_is_nonlinear(skb) &&
3448 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3452 * skb_linearize_cow - make sure skb is linear and writable
3453 * @skb: buffer to process
3455 * If there is no free memory -ENOMEM is returned, otherwise zero
3456 * is returned and the old skb data released.
3458 static inline int skb_linearize_cow(struct sk_buff *skb)
3460 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3461 __skb_linearize(skb) : 0;
3464 static __always_inline void
3465 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3468 if (skb->ip_summed == CHECKSUM_COMPLETE)
3469 skb->csum = csum_block_sub(skb->csum,
3470 csum_partial(start, len, 0), off);
3471 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3472 skb_checksum_start_offset(skb) < 0)
3473 skb->ip_summed = CHECKSUM_NONE;
3477 * skb_postpull_rcsum - update checksum for received skb after pull
3478 * @skb: buffer to update
3479 * @start: start of data before pull
3480 * @len: length of data pulled
3482 * After doing a pull on a received packet, you need to call this to
3483 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3484 * CHECKSUM_NONE so that it can be recomputed from scratch.
3486 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3487 const void *start, unsigned int len)
3489 __skb_postpull_rcsum(skb, start, len, 0);
3492 static __always_inline void
3493 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3496 if (skb->ip_summed == CHECKSUM_COMPLETE)
3497 skb->csum = csum_block_add(skb->csum,
3498 csum_partial(start, len, 0), off);
3502 * skb_postpush_rcsum - update checksum for received skb after push
3503 * @skb: buffer to update
3504 * @start: start of data after push
3505 * @len: length of data pushed
3507 * After doing a push on a received packet, you need to call this to
3508 * update the CHECKSUM_COMPLETE checksum.
3510 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3511 const void *start, unsigned int len)
3513 __skb_postpush_rcsum(skb, start, len, 0);
3516 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3519 * skb_push_rcsum - push skb and update receive checksum
3520 * @skb: buffer to update
3521 * @len: length of data pulled
3523 * This function performs an skb_push on the packet and updates
3524 * the CHECKSUM_COMPLETE checksum. It should be used on
3525 * receive path processing instead of skb_push unless you know
3526 * that the checksum difference is zero (e.g., a valid IP header)
3527 * or you are setting ip_summed to CHECKSUM_NONE.
3529 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3532 skb_postpush_rcsum(skb, skb->data, len);
3536 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3538 * pskb_trim_rcsum - trim received skb and update checksum
3539 * @skb: buffer to trim
3542 * This is exactly the same as pskb_trim except that it ensures the
3543 * checksum of received packets are still valid after the operation.
3544 * It can change skb pointers.
3547 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3549 if (likely(len >= skb->len))
3551 return pskb_trim_rcsum_slow(skb, len);
3554 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3556 if (skb->ip_summed == CHECKSUM_COMPLETE)
3557 skb->ip_summed = CHECKSUM_NONE;
3558 __skb_trim(skb, len);
3562 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3564 if (skb->ip_summed == CHECKSUM_COMPLETE)
3565 skb->ip_summed = CHECKSUM_NONE;
3566 return __skb_grow(skb, len);
3569 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3570 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3571 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3572 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3573 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3575 #define skb_queue_walk(queue, skb) \
3576 for (skb = (queue)->next; \
3577 skb != (struct sk_buff *)(queue); \
3580 #define skb_queue_walk_safe(queue, skb, tmp) \
3581 for (skb = (queue)->next, tmp = skb->next; \
3582 skb != (struct sk_buff *)(queue); \
3583 skb = tmp, tmp = skb->next)
3585 #define skb_queue_walk_from(queue, skb) \
3586 for (; skb != (struct sk_buff *)(queue); \
3589 #define skb_rbtree_walk(skb, root) \
3590 for (skb = skb_rb_first(root); skb != NULL; \
3591 skb = skb_rb_next(skb))
3593 #define skb_rbtree_walk_from(skb) \
3594 for (; skb != NULL; \
3595 skb = skb_rb_next(skb))
3597 #define skb_rbtree_walk_from_safe(skb, tmp) \
3598 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3601 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3602 for (tmp = skb->next; \
3603 skb != (struct sk_buff *)(queue); \
3604 skb = tmp, tmp = skb->next)
3606 #define skb_queue_reverse_walk(queue, skb) \
3607 for (skb = (queue)->prev; \
3608 skb != (struct sk_buff *)(queue); \
3611 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3612 for (skb = (queue)->prev, tmp = skb->prev; \
3613 skb != (struct sk_buff *)(queue); \
3614 skb = tmp, tmp = skb->prev)
3616 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3617 for (tmp = skb->prev; \
3618 skb != (struct sk_buff *)(queue); \
3619 skb = tmp, tmp = skb->prev)
3621 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3623 return skb_shinfo(skb)->frag_list != NULL;
3626 static inline void skb_frag_list_init(struct sk_buff *skb)
3628 skb_shinfo(skb)->frag_list = NULL;
3631 #define skb_walk_frags(skb, iter) \
3632 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3635 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3636 int *err, long *timeo_p,
3637 const struct sk_buff *skb);
3638 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3639 struct sk_buff_head *queue,
3642 struct sk_buff **last);
3643 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3644 struct sk_buff_head *queue,
3645 unsigned int flags, int *off, int *err,
3646 struct sk_buff **last);
3647 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3648 struct sk_buff_head *sk_queue,
3649 unsigned int flags, int *off, int *err);
3650 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3652 __poll_t datagram_poll(struct file *file, struct socket *sock,
3653 struct poll_table_struct *wait);
3654 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3655 struct iov_iter *to, int size);
3656 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3657 struct msghdr *msg, int size)
3659 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3661 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3662 struct msghdr *msg);
3663 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3664 struct iov_iter *to, int len,
3665 struct ahash_request *hash);
3666 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3667 struct iov_iter *from, int len);
3668 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3669 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3670 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3671 static inline void skb_free_datagram_locked(struct sock *sk,
3672 struct sk_buff *skb)
3674 __skb_free_datagram_locked(sk, skb, 0);
3676 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3677 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3678 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3679 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3681 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3682 struct pipe_inode_info *pipe, unsigned int len,
3683 unsigned int flags);
3684 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3686 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3687 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3688 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3689 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3691 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3692 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3693 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3694 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3695 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3696 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3697 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3698 unsigned int offset);
3699 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3700 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3701 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3702 int skb_vlan_pop(struct sk_buff *skb);
3703 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3704 int skb_eth_pop(struct sk_buff *skb);
3705 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3706 const unsigned char *src);
3707 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3708 int mac_len, bool ethernet);
3709 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3711 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3712 int skb_mpls_dec_ttl(struct sk_buff *skb);
3713 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3716 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3718 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3721 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3723 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3726 struct skb_checksum_ops {
3727 __wsum (*update)(const void *mem, int len, __wsum wsum);
3728 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3731 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3733 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3734 __wsum csum, const struct skb_checksum_ops *ops);
3735 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3738 static inline void * __must_check
3739 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3740 const void *data, int hlen, void *buffer)
3742 if (likely(hlen - offset >= len))
3743 return (void *)data + offset;
3745 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3751 static inline void * __must_check
3752 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3754 return __skb_header_pointer(skb, offset, len, skb->data,
3755 skb_headlen(skb), buffer);
3759 * skb_needs_linearize - check if we need to linearize a given skb
3760 * depending on the given device features.
3761 * @skb: socket buffer to check
3762 * @features: net device features
3764 * Returns true if either:
3765 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3766 * 2. skb is fragmented and the device does not support SG.
3768 static inline bool skb_needs_linearize(struct sk_buff *skb,
3769 netdev_features_t features)
3771 return skb_is_nonlinear(skb) &&
3772 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3773 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3776 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3778 const unsigned int len)
3780 memcpy(to, skb->data, len);
3783 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3784 const int offset, void *to,
3785 const unsigned int len)
3787 memcpy(to, skb->data + offset, len);
3790 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3792 const unsigned int len)
3794 memcpy(skb->data, from, len);
3797 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3800 const unsigned int len)
3802 memcpy(skb->data + offset, from, len);
3805 void skb_init(void);
3807 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3813 * skb_get_timestamp - get timestamp from a skb
3814 * @skb: skb to get stamp from
3815 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3817 * Timestamps are stored in the skb as offsets to a base timestamp.
3818 * This function converts the offset back to a struct timeval and stores
3821 static inline void skb_get_timestamp(const struct sk_buff *skb,
3822 struct __kernel_old_timeval *stamp)
3824 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3827 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3828 struct __kernel_sock_timeval *stamp)
3830 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3832 stamp->tv_sec = ts.tv_sec;
3833 stamp->tv_usec = ts.tv_nsec / 1000;
3836 static inline void skb_get_timestampns(const struct sk_buff *skb,
3837 struct __kernel_old_timespec *stamp)
3839 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3841 stamp->tv_sec = ts.tv_sec;
3842 stamp->tv_nsec = ts.tv_nsec;
3845 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3846 struct __kernel_timespec *stamp)
3848 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3850 stamp->tv_sec = ts.tv_sec;
3851 stamp->tv_nsec = ts.tv_nsec;
3854 static inline void __net_timestamp(struct sk_buff *skb)
3856 skb->tstamp = ktime_get_real();
3859 static inline ktime_t net_timedelta(ktime_t t)
3861 return ktime_sub(ktime_get_real(), t);
3864 static inline ktime_t net_invalid_timestamp(void)
3869 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3871 return skb_shinfo(skb)->meta_len;
3874 static inline void *skb_metadata_end(const struct sk_buff *skb)
3876 return skb_mac_header(skb);
3879 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3880 const struct sk_buff *skb_b,
3883 const void *a = skb_metadata_end(skb_a);
3884 const void *b = skb_metadata_end(skb_b);
3885 /* Using more efficient varaiant than plain call to memcmp(). */
3886 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3890 #define __it(x, op) (x -= sizeof(u##op))
3891 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3892 case 32: diffs |= __it_diff(a, b, 64);
3894 case 24: diffs |= __it_diff(a, b, 64);
3896 case 16: diffs |= __it_diff(a, b, 64);
3898 case 8: diffs |= __it_diff(a, b, 64);
3900 case 28: diffs |= __it_diff(a, b, 64);
3902 case 20: diffs |= __it_diff(a, b, 64);
3904 case 12: diffs |= __it_diff(a, b, 64);
3906 case 4: diffs |= __it_diff(a, b, 32);
3911 return memcmp(a - meta_len, b - meta_len, meta_len);
3915 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3916 const struct sk_buff *skb_b)
3918 u8 len_a = skb_metadata_len(skb_a);
3919 u8 len_b = skb_metadata_len(skb_b);
3921 if (!(len_a | len_b))
3924 return len_a != len_b ?
3925 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3928 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3930 skb_shinfo(skb)->meta_len = meta_len;
3933 static inline void skb_metadata_clear(struct sk_buff *skb)
3935 skb_metadata_set(skb, 0);
3938 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3940 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3942 void skb_clone_tx_timestamp(struct sk_buff *skb);
3943 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3945 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3947 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3951 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3956 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3959 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3961 * PHY drivers may accept clones of transmitted packets for
3962 * timestamping via their phy_driver.txtstamp method. These drivers
3963 * must call this function to return the skb back to the stack with a
3966 * @skb: clone of the original outgoing packet
3967 * @hwtstamps: hardware time stamps
3970 void skb_complete_tx_timestamp(struct sk_buff *skb,
3971 struct skb_shared_hwtstamps *hwtstamps);
3973 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
3974 struct skb_shared_hwtstamps *hwtstamps,
3975 struct sock *sk, int tstype);
3978 * skb_tstamp_tx - queue clone of skb with send time stamps
3979 * @orig_skb: the original outgoing packet
3980 * @hwtstamps: hardware time stamps, may be NULL if not available
3982 * If the skb has a socket associated, then this function clones the
3983 * skb (thus sharing the actual data and optional structures), stores
3984 * the optional hardware time stamping information (if non NULL) or
3985 * generates a software time stamp (otherwise), then queues the clone
3986 * to the error queue of the socket. Errors are silently ignored.
3988 void skb_tstamp_tx(struct sk_buff *orig_skb,
3989 struct skb_shared_hwtstamps *hwtstamps);
3992 * skb_tx_timestamp() - Driver hook for transmit timestamping
3994 * Ethernet MAC Drivers should call this function in their hard_xmit()
3995 * function immediately before giving the sk_buff to the MAC hardware.
3997 * Specifically, one should make absolutely sure that this function is
3998 * called before TX completion of this packet can trigger. Otherwise
3999 * the packet could potentially already be freed.
4001 * @skb: A socket buffer.
4003 static inline void skb_tx_timestamp(struct sk_buff *skb)
4005 skb_clone_tx_timestamp(skb);
4006 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4007 skb_tstamp_tx(skb, NULL);
4011 * skb_complete_wifi_ack - deliver skb with wifi status
4013 * @skb: the original outgoing packet
4014 * @acked: ack status
4017 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4019 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4020 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4022 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4024 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4026 (skb->ip_summed == CHECKSUM_PARTIAL &&
4027 skb_checksum_start_offset(skb) >= 0));
4031 * skb_checksum_complete - Calculate checksum of an entire packet
4032 * @skb: packet to process
4034 * This function calculates the checksum over the entire packet plus
4035 * the value of skb->csum. The latter can be used to supply the
4036 * checksum of a pseudo header as used by TCP/UDP. It returns the
4039 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4040 * this function can be used to verify that checksum on received
4041 * packets. In that case the function should return zero if the
4042 * checksum is correct. In particular, this function will return zero
4043 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4044 * hardware has already verified the correctness of the checksum.
4046 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4048 return skb_csum_unnecessary(skb) ?
4049 0 : __skb_checksum_complete(skb);
4052 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4054 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4055 if (skb->csum_level == 0)
4056 skb->ip_summed = CHECKSUM_NONE;
4062 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4064 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4065 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4067 } else if (skb->ip_summed == CHECKSUM_NONE) {
4068 skb->ip_summed = CHECKSUM_UNNECESSARY;
4069 skb->csum_level = 0;
4073 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4075 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4076 skb->ip_summed = CHECKSUM_NONE;
4077 skb->csum_level = 0;
4081 /* Check if we need to perform checksum complete validation.
4083 * Returns true if checksum complete is needed, false otherwise
4084 * (either checksum is unnecessary or zero checksum is allowed).
4086 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4090 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4091 skb->csum_valid = 1;
4092 __skb_decr_checksum_unnecessary(skb);
4099 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4102 #define CHECKSUM_BREAK 76
4104 /* Unset checksum-complete
4106 * Unset checksum complete can be done when packet is being modified
4107 * (uncompressed for instance) and checksum-complete value is
4110 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4112 if (skb->ip_summed == CHECKSUM_COMPLETE)
4113 skb->ip_summed = CHECKSUM_NONE;
4116 /* Validate (init) checksum based on checksum complete.
4119 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4120 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4121 * checksum is stored in skb->csum for use in __skb_checksum_complete
4122 * non-zero: value of invalid checksum
4125 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4129 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4130 if (!csum_fold(csum_add(psum, skb->csum))) {
4131 skb->csum_valid = 1;
4138 if (complete || skb->len <= CHECKSUM_BREAK) {
4141 csum = __skb_checksum_complete(skb);
4142 skb->csum_valid = !csum;
4149 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4154 /* Perform checksum validate (init). Note that this is a macro since we only
4155 * want to calculate the pseudo header which is an input function if necessary.
4156 * First we try to validate without any computation (checksum unnecessary) and
4157 * then calculate based on checksum complete calling the function to compute
4161 * 0: checksum is validated or try to in skb_checksum_complete
4162 * non-zero: value of invalid checksum
4164 #define __skb_checksum_validate(skb, proto, complete, \
4165 zero_okay, check, compute_pseudo) \
4167 __sum16 __ret = 0; \
4168 skb->csum_valid = 0; \
4169 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4170 __ret = __skb_checksum_validate_complete(skb, \
4171 complete, compute_pseudo(skb, proto)); \
4175 #define skb_checksum_init(skb, proto, compute_pseudo) \
4176 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4178 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4179 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4181 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4182 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4184 #define skb_checksum_validate_zero_check(skb, proto, check, \
4186 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4188 #define skb_checksum_simple_validate(skb) \
4189 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4191 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4193 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4196 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4198 skb->csum = ~pseudo;
4199 skb->ip_summed = CHECKSUM_COMPLETE;
4202 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4204 if (__skb_checksum_convert_check(skb)) \
4205 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4208 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4209 u16 start, u16 offset)
4211 skb->ip_summed = CHECKSUM_PARTIAL;
4212 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4213 skb->csum_offset = offset - start;
4216 /* Update skbuf and packet to reflect the remote checksum offload operation.
4217 * When called, ptr indicates the starting point for skb->csum when
4218 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4219 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4221 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4222 int start, int offset, bool nopartial)
4227 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4231 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4232 __skb_checksum_complete(skb);
4233 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4236 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4238 /* Adjust skb->csum since we changed the packet */
4239 skb->csum = csum_add(skb->csum, delta);
4242 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4244 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4245 return (void *)(skb->_nfct & NFCT_PTRMASK);
4251 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4253 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4260 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4262 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4263 skb->slow_gro |= !!nfct;
4268 #ifdef CONFIG_SKB_EXTENSIONS
4270 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4276 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4279 #if IS_ENABLED(CONFIG_MPTCP)
4282 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4285 SKB_EXT_NUM, /* must be last */
4289 * struct skb_ext - sk_buff extensions
4290 * @refcnt: 1 on allocation, deallocated on 0
4291 * @offset: offset to add to @data to obtain extension address
4292 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4293 * @data: start of extension data, variable sized
4295 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4296 * to use 'u8' types while allowing up to 2kb worth of extension data.
4300 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4301 u8 chunks; /* same */
4302 char data[] __aligned(8);
4305 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4306 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4307 struct skb_ext *ext);
4308 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4309 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4310 void __skb_ext_put(struct skb_ext *ext);
4312 static inline void skb_ext_put(struct sk_buff *skb)
4314 if (skb->active_extensions)
4315 __skb_ext_put(skb->extensions);
4318 static inline void __skb_ext_copy(struct sk_buff *dst,
4319 const struct sk_buff *src)
4321 dst->active_extensions = src->active_extensions;
4323 if (src->active_extensions) {
4324 struct skb_ext *ext = src->extensions;
4326 refcount_inc(&ext->refcnt);
4327 dst->extensions = ext;
4331 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4334 __skb_ext_copy(dst, src);
4337 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4339 return !!ext->offset[i];
4342 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4344 return skb->active_extensions & (1 << id);
4347 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4349 if (skb_ext_exist(skb, id))
4350 __skb_ext_del(skb, id);
4353 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4355 if (skb_ext_exist(skb, id)) {
4356 struct skb_ext *ext = skb->extensions;
4358 return (void *)ext + (ext->offset[id] << 3);
4364 static inline void skb_ext_reset(struct sk_buff *skb)
4366 if (unlikely(skb->active_extensions)) {
4367 __skb_ext_put(skb->extensions);
4368 skb->active_extensions = 0;
4372 static inline bool skb_has_extensions(struct sk_buff *skb)
4374 return unlikely(skb->active_extensions);
4377 static inline void skb_ext_put(struct sk_buff *skb) {}
4378 static inline void skb_ext_reset(struct sk_buff *skb) {}
4379 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4380 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4381 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4382 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4383 #endif /* CONFIG_SKB_EXTENSIONS */
4385 static inline void nf_reset_ct(struct sk_buff *skb)
4387 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4388 nf_conntrack_put(skb_nfct(skb));
4393 static inline void nf_reset_trace(struct sk_buff *skb)
4395 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4400 static inline void ipvs_reset(struct sk_buff *skb)
4402 #if IS_ENABLED(CONFIG_IP_VS)
4403 skb->ipvs_property = 0;
4407 /* Note: This doesn't put any conntrack info in dst. */
4408 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4411 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4412 dst->_nfct = src->_nfct;
4413 nf_conntrack_get(skb_nfct(src));
4415 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4417 dst->nf_trace = src->nf_trace;
4421 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4423 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4424 nf_conntrack_put(skb_nfct(dst));
4426 dst->slow_gro = src->slow_gro;
4427 __nf_copy(dst, src, true);
4430 #ifdef CONFIG_NETWORK_SECMARK
4431 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4433 to->secmark = from->secmark;
4436 static inline void skb_init_secmark(struct sk_buff *skb)
4441 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4444 static inline void skb_init_secmark(struct sk_buff *skb)
4448 static inline int secpath_exists(const struct sk_buff *skb)
4451 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4457 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4459 return !skb->destructor &&
4460 !secpath_exists(skb) &&
4462 !skb->_skb_refdst &&
4463 !skb_has_frag_list(skb);
4466 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4468 skb->queue_mapping = queue_mapping;
4471 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4473 return skb->queue_mapping;
4476 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4478 to->queue_mapping = from->queue_mapping;
4481 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4483 skb->queue_mapping = rx_queue + 1;
4486 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4488 return skb->queue_mapping - 1;
4491 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4493 return skb->queue_mapping != 0;
4496 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4498 skb->dst_pending_confirm = val;
4501 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4503 return skb->dst_pending_confirm != 0;
4506 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4509 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4515 /* Keeps track of mac header offset relative to skb->head.
4516 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4517 * For non-tunnel skb it points to skb_mac_header() and for
4518 * tunnel skb it points to outer mac header.
4519 * Keeps track of level of encapsulation of network headers.
4530 #define SKB_GSO_CB_OFFSET 32
4531 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4533 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4535 return (skb_mac_header(inner_skb) - inner_skb->head) -
4536 SKB_GSO_CB(inner_skb)->mac_offset;
4539 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4541 int new_headroom, headroom;
4544 headroom = skb_headroom(skb);
4545 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4549 new_headroom = skb_headroom(skb);
4550 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4554 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4556 /* Do not update partial checksums if remote checksum is enabled. */
4557 if (skb->remcsum_offload)
4560 SKB_GSO_CB(skb)->csum = res;
4561 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4564 /* Compute the checksum for a gso segment. First compute the checksum value
4565 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4566 * then add in skb->csum (checksum from csum_start to end of packet).
4567 * skb->csum and csum_start are then updated to reflect the checksum of the
4568 * resultant packet starting from the transport header-- the resultant checksum
4569 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4572 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4574 unsigned char *csum_start = skb_transport_header(skb);
4575 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4576 __wsum partial = SKB_GSO_CB(skb)->csum;
4578 SKB_GSO_CB(skb)->csum = res;
4579 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4581 return csum_fold(csum_partial(csum_start, plen, partial));
4584 static inline bool skb_is_gso(const struct sk_buff *skb)
4586 return skb_shinfo(skb)->gso_size;
4589 /* Note: Should be called only if skb_is_gso(skb) is true */
4590 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4592 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4595 /* Note: Should be called only if skb_is_gso(skb) is true */
4596 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4598 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4601 /* Note: Should be called only if skb_is_gso(skb) is true */
4602 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4604 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4607 static inline void skb_gso_reset(struct sk_buff *skb)
4609 skb_shinfo(skb)->gso_size = 0;
4610 skb_shinfo(skb)->gso_segs = 0;
4611 skb_shinfo(skb)->gso_type = 0;
4614 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4617 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4619 shinfo->gso_size += increment;
4622 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4625 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4627 shinfo->gso_size -= decrement;
4630 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4632 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4634 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4635 * wanted then gso_type will be set. */
4636 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4638 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4639 unlikely(shinfo->gso_type == 0)) {
4640 __skb_warn_lro_forwarding(skb);
4646 static inline void skb_forward_csum(struct sk_buff *skb)
4648 /* Unfortunately we don't support this one. Any brave souls? */
4649 if (skb->ip_summed == CHECKSUM_COMPLETE)
4650 skb->ip_summed = CHECKSUM_NONE;
4654 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4655 * @skb: skb to check
4657 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4658 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4659 * use this helper, to document places where we make this assertion.
4661 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4664 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4668 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4670 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4671 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4672 unsigned int transport_len,
4673 __sum16(*skb_chkf)(struct sk_buff *skb));
4676 * skb_head_is_locked - Determine if the skb->head is locked down
4677 * @skb: skb to check
4679 * The head on skbs build around a head frag can be removed if they are
4680 * not cloned. This function returns true if the skb head is locked down
4681 * due to either being allocated via kmalloc, or by being a clone with
4682 * multiple references to the head.
4684 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4686 return !skb->head_frag || skb_cloned(skb);
4689 /* Local Checksum Offload.
4690 * Compute outer checksum based on the assumption that the
4691 * inner checksum will be offloaded later.
4692 * See Documentation/networking/checksum-offloads.rst for
4693 * explanation of how this works.
4694 * Fill in outer checksum adjustment (e.g. with sum of outer
4695 * pseudo-header) before calling.
4696 * Also ensure that inner checksum is in linear data area.
4698 static inline __wsum lco_csum(struct sk_buff *skb)
4700 unsigned char *csum_start = skb_checksum_start(skb);
4701 unsigned char *l4_hdr = skb_transport_header(skb);
4704 /* Start with complement of inner checksum adjustment */
4705 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4708 /* Add in checksum of our headers (incl. outer checksum
4709 * adjustment filled in by caller) and return result.
4711 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4714 static inline bool skb_is_redirected(const struct sk_buff *skb)
4716 return skb->redirected;
4719 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4721 skb->redirected = 1;
4722 #ifdef CONFIG_NET_REDIRECT
4723 skb->from_ingress = from_ingress;
4724 if (skb->from_ingress)
4729 static inline void skb_reset_redirect(struct sk_buff *skb)
4731 skb->redirected = 0;
4734 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4736 return skb->csum_not_inet;
4739 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4740 const u64 kcov_handle)
4743 skb->kcov_handle = kcov_handle;
4747 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4750 return skb->kcov_handle;
4756 #ifdef CONFIG_PAGE_POOL
4757 static inline void skb_mark_for_recycle(struct sk_buff *skb)
4759 skb->pp_recycle = 1;
4763 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4765 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4767 return page_pool_return_skb_page(virt_to_page(data));
4770 #endif /* __KERNEL__ */
4771 #endif /* _LINUX_SKBUFF_H */