1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * Definitions for the 'struct sk_buff' memory handlers.
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
40 #include <net/page_pool.h>
41 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
42 #include <linux/netfilter/nf_conntrack_common.h>
45 /* The interface for checksum offload between the stack and networking drivers
48 * A. IP checksum related features
50 * Drivers advertise checksum offload capabilities in the features of a device.
51 * From the stack's point of view these are capabilities offered by the driver.
52 * A driver typically only advertises features that it is capable of offloading
55 * The checksum related features are:
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
67 * is TCP or UDP. The IPv4 header may contain IP options.
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
75 * IPv6|UDP where the Next Header field in the IPv6
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
83 * This flag is only used to disable the RX checksum
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
88 * B. Checksumming of received packets by device. Indication of checksum
89 * verification is set in skb->ip_summed. Possible values are:
93 * Device did not checksum this packet e.g. due to lack of capabilities.
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
97 * CHECKSUM_UNNECESSARY:
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
113 * FCOE: indicates the CRC in FC frame has been validated.
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
119 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
120 * two. If the device were only able to verify the UDP checksum and not
121 * GRE, either because it doesn't support GRE checksum or because GRE
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
127 * This is the most generic way. The device supplied checksum of the _whole_
128 * packet as seen by netif_rx() and fills in skb->csum. This means the
129 * hardware doesn't need to parse L3/L4 headers to implement this.
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
153 * The driver is required to checksum the packet as seen by hard_start_xmit()
154 * from skb->csum_start up to the end, and to record/write the checksum at
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
157 * offset of the packet, but it should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum -- it is the
159 * purview of the stack to validate that csum_start and csum_offset are set
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
177 * The skb was already checksummed by the protocol, or a checksum is not
180 * CHECKSUM_UNNECESSARY:
182 * This has the same meaning as CHECKSUM_NONE for checksum offload on
186 * Not used in checksum output. If a driver observes a packet with this value
187 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
189 * D. Non-IP checksum (CRC) offloads
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
193 * will set csum_start and csum_offset accordingly, set ip_summed to
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
204 * accordingly. Note that there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
206 * both IP checksum offload and FCOE CRC offload must verify which offload
207 * is configured for a packet, presumably by inspecting packet headers.
209 * E. Checksumming on output with GSO.
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
216 * csum_offset are set to refer to the outermost checksum being offloaded
217 * (two offloaded checksums are possible with UDP encapsulation).
220 /* Don't change this without changing skb_csum_unnecessary! */
221 #define CHECKSUM_NONE 0
222 #define CHECKSUM_UNNECESSARY 1
223 #define CHECKSUM_COMPLETE 2
224 #define CHECKSUM_PARTIAL 3
226 /* Maximum value in skb->csum_level */
227 #define SKB_MAX_CSUM_LEVEL 3
229 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
230 #define SKB_WITH_OVERHEAD(X) \
231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
232 #define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
234 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
237 /* return minimum truesize of one skb containing X bytes of data */
238 #define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
242 struct ahash_request;
245 struct pipe_inode_info;
252 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
253 struct nf_bridge_info {
255 BRNF_PROTO_UNCHANGED,
263 struct net_device *physindev;
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
268 /* prerouting: detect dnat in orig/reply direction */
270 struct in6_addr ipv6_daddr;
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
276 char neigh_header[8];
281 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
282 /* Chain in tc_skb_ext will be used to share the tc chain with
283 * ovs recirc_id. It will be set to the current chain by tc
284 * and read by ovs to recirc_id.
296 struct sk_buff_head {
297 /* These two members must be first. */
298 struct sk_buff *next;
299 struct sk_buff *prev;
307 /* The reason of skb drop, which is used in kfree_skb_reason().
308 * en...maybe they should be splited by group?
310 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
311 * used to translate the reason to string.
313 enum skb_drop_reason {
314 SKB_DROP_REASON_NOT_SPECIFIED, /* drop reason is not specified */
315 SKB_DROP_REASON_NO_SOCKET, /* socket not found */
316 SKB_DROP_REASON_PKT_TOO_SMALL, /* packet size is too small */
317 SKB_DROP_REASON_TCP_CSUM, /* TCP checksum error */
318 SKB_DROP_REASON_SOCKET_FILTER, /* dropped by socket filter */
319 SKB_DROP_REASON_UDP_CSUM, /* UDP checksum error */
320 SKB_DROP_REASON_NETFILTER_DROP, /* dropped by netfilter */
321 SKB_DROP_REASON_OTHERHOST, /* packet don't belong to current
322 * host (interface is in promisc
325 SKB_DROP_REASON_IP_CSUM, /* IP checksum error */
326 SKB_DROP_REASON_IP_INHDR, /* there is something wrong with
328 * IPSTATS_MIB_INHDRERRORS)
333 /* To allow 64K frame to be packed as single skb without frag_list we
334 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
335 * buffers which do not start on a page boundary.
337 * Since GRO uses frags we allocate at least 16 regardless of page
340 #if (65536/PAGE_SIZE + 1) < 16
341 #define MAX_SKB_FRAGS 16UL
343 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
345 extern int sysctl_max_skb_frags;
347 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
348 * segment using its current segmentation instead.
350 #define GSO_BY_FRAGS 0xFFFF
352 typedef struct bio_vec skb_frag_t;
355 * skb_frag_size() - Returns the size of a skb fragment
356 * @frag: skb fragment
358 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
364 * skb_frag_size_set() - Sets the size of a skb fragment
365 * @frag: skb fragment
366 * @size: size of fragment
368 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
374 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
375 * @frag: skb fragment
376 * @delta: value to add
378 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
380 frag->bv_len += delta;
384 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
385 * @frag: skb fragment
386 * @delta: value to subtract
388 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
390 frag->bv_len -= delta;
394 * skb_frag_must_loop - Test if %p is a high memory page
395 * @p: fragment's page
397 static inline bool skb_frag_must_loop(struct page *p)
399 #if defined(CONFIG_HIGHMEM)
400 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
407 * skb_frag_foreach_page - loop over pages in a fragment
409 * @f: skb frag to operate on
410 * @f_off: offset from start of f->bv_page
411 * @f_len: length from f_off to loop over
412 * @p: (temp var) current page
413 * @p_off: (temp var) offset from start of current page,
414 * non-zero only on first page.
415 * @p_len: (temp var) length in current page,
416 * < PAGE_SIZE only on first and last page.
417 * @copied: (temp var) length so far, excluding current p_len.
419 * A fragment can hold a compound page, in which case per-page
420 * operations, notably kmap_atomic, must be called for each
423 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
424 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
425 p_off = (f_off) & (PAGE_SIZE - 1), \
426 p_len = skb_frag_must_loop(p) ? \
427 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
430 copied += p_len, p++, p_off = 0, \
431 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
433 #define HAVE_HW_TIME_STAMP
436 * struct skb_shared_hwtstamps - hardware time stamps
437 * @hwtstamp: hardware time stamp transformed into duration
438 * since arbitrary point in time
440 * Software time stamps generated by ktime_get_real() are stored in
443 * hwtstamps can only be compared against other hwtstamps from
446 * This structure is attached to packets as part of the
447 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
449 struct skb_shared_hwtstamps {
453 /* Definitions for tx_flags in struct skb_shared_info */
455 /* generate hardware time stamp */
456 SKBTX_HW_TSTAMP = 1 << 0,
458 /* generate software time stamp when queueing packet to NIC */
459 SKBTX_SW_TSTAMP = 1 << 1,
461 /* device driver is going to provide hardware time stamp */
462 SKBTX_IN_PROGRESS = 1 << 2,
464 /* generate wifi status information (where possible) */
465 SKBTX_WIFI_STATUS = 1 << 4,
467 /* generate software time stamp when entering packet scheduling */
468 SKBTX_SCHED_TSTAMP = 1 << 6,
471 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
473 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
475 /* Definitions for flags in struct skb_shared_info */
477 /* use zcopy routines */
478 SKBFL_ZEROCOPY_ENABLE = BIT(0),
480 /* This indicates at least one fragment might be overwritten
481 * (as in vmsplice(), sendfile() ...)
482 * If we need to compute a TX checksum, we'll need to copy
483 * all frags to avoid possible bad checksum
485 SKBFL_SHARED_FRAG = BIT(1),
488 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
491 * The callback notifies userspace to release buffers when skb DMA is done in
492 * lower device, the skb last reference should be 0 when calling this.
493 * The zerocopy_success argument is true if zero copy transmit occurred,
494 * false on data copy or out of memory error caused by data copy attempt.
495 * The ctx field is used to track device context.
496 * The desc field is used to track userspace buffer index.
499 void (*callback)(struct sk_buff *, struct ubuf_info *,
500 bool zerocopy_success);
517 struct user_struct *user;
522 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
524 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
525 void mm_unaccount_pinned_pages(struct mmpin *mmp);
527 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
528 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
529 struct ubuf_info *uarg);
531 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
533 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
536 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
537 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
538 struct msghdr *msg, int len,
539 struct ubuf_info *uarg);
541 /* This data is invariant across clones and lives at
542 * the end of the header data, ie. at skb->end.
544 struct skb_shared_info {
549 unsigned short gso_size;
550 /* Warning: this field is not always filled in (UFO)! */
551 unsigned short gso_segs;
552 struct sk_buff *frag_list;
553 struct skb_shared_hwtstamps hwtstamps;
554 unsigned int gso_type;
558 * Warning : all fields before dataref are cleared in __alloc_skb()
562 /* Intermediate layers must ensure that destructor_arg
563 * remains valid until skb destructor */
564 void * destructor_arg;
566 /* must be last field, see pskb_expand_head() */
567 skb_frag_t frags[MAX_SKB_FRAGS];
570 /* We divide dataref into two halves. The higher 16 bits hold references
571 * to the payload part of skb->data. The lower 16 bits hold references to
572 * the entire skb->data. A clone of a headerless skb holds the length of
573 * the header in skb->hdr_len.
575 * All users must obey the rule that the skb->data reference count must be
576 * greater than or equal to the payload reference count.
578 * Holding a reference to the payload part means that the user does not
579 * care about modifications to the header part of skb->data.
581 #define SKB_DATAREF_SHIFT 16
582 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
586 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
587 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
588 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
592 SKB_GSO_TCPV4 = 1 << 0,
594 /* This indicates the skb is from an untrusted source. */
595 SKB_GSO_DODGY = 1 << 1,
597 /* This indicates the tcp segment has CWR set. */
598 SKB_GSO_TCP_ECN = 1 << 2,
600 SKB_GSO_TCP_FIXEDID = 1 << 3,
602 SKB_GSO_TCPV6 = 1 << 4,
604 SKB_GSO_FCOE = 1 << 5,
606 SKB_GSO_GRE = 1 << 6,
608 SKB_GSO_GRE_CSUM = 1 << 7,
610 SKB_GSO_IPXIP4 = 1 << 8,
612 SKB_GSO_IPXIP6 = 1 << 9,
614 SKB_GSO_UDP_TUNNEL = 1 << 10,
616 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
618 SKB_GSO_PARTIAL = 1 << 12,
620 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
622 SKB_GSO_SCTP = 1 << 14,
624 SKB_GSO_ESP = 1 << 15,
626 SKB_GSO_UDP = 1 << 16,
628 SKB_GSO_UDP_L4 = 1 << 17,
630 SKB_GSO_FRAGLIST = 1 << 18,
633 #if BITS_PER_LONG > 32
634 #define NET_SKBUFF_DATA_USES_OFFSET 1
637 #ifdef NET_SKBUFF_DATA_USES_OFFSET
638 typedef unsigned int sk_buff_data_t;
640 typedef unsigned char *sk_buff_data_t;
644 * struct sk_buff - socket buffer
645 * @next: Next buffer in list
646 * @prev: Previous buffer in list
647 * @tstamp: Time we arrived/left
648 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
649 * for retransmit timer
650 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
652 * @sk: Socket we are owned by
653 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
654 * fragmentation management
655 * @dev: Device we arrived on/are leaving by
656 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
657 * @cb: Control buffer. Free for use by every layer. Put private vars here
658 * @_skb_refdst: destination entry (with norefcount bit)
659 * @sp: the security path, used for xfrm
660 * @len: Length of actual data
661 * @data_len: Data length
662 * @mac_len: Length of link layer header
663 * @hdr_len: writable header length of cloned skb
664 * @csum: Checksum (must include start/offset pair)
665 * @csum_start: Offset from skb->head where checksumming should start
666 * @csum_offset: Offset from csum_start where checksum should be stored
667 * @priority: Packet queueing priority
668 * @ignore_df: allow local fragmentation
669 * @cloned: Head may be cloned (check refcnt to be sure)
670 * @ip_summed: Driver fed us an IP checksum
671 * @nohdr: Payload reference only, must not modify header
672 * @pkt_type: Packet class
673 * @fclone: skbuff clone status
674 * @ipvs_property: skbuff is owned by ipvs
675 * @inner_protocol_type: whether the inner protocol is
676 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
677 * @remcsum_offload: remote checksum offload is enabled
678 * @offload_fwd_mark: Packet was L2-forwarded in hardware
679 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
680 * @tc_skip_classify: do not classify packet. set by IFB device
681 * @tc_at_ingress: used within tc_classify to distinguish in/egress
682 * @redirected: packet was redirected by packet classifier
683 * @from_ingress: packet was redirected from the ingress path
684 * @peeked: this packet has been seen already, so stats have been
685 * done for it, don't do them again
686 * @nf_trace: netfilter packet trace flag
687 * @protocol: Packet protocol from driver
688 * @destructor: Destruct function
689 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
690 * @_sk_redir: socket redirection information for skmsg
691 * @_nfct: Associated connection, if any (with nfctinfo bits)
692 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
693 * @skb_iif: ifindex of device we arrived on
694 * @tc_index: Traffic control index
695 * @hash: the packet hash
696 * @queue_mapping: Queue mapping for multiqueue devices
697 * @head_frag: skb was allocated from page fragments,
698 * not allocated by kmalloc() or vmalloc().
699 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
700 * @pp_recycle: mark the packet for recycling instead of freeing (implies
701 * page_pool support on driver)
702 * @active_extensions: active extensions (skb_ext_id types)
703 * @ndisc_nodetype: router type (from link layer)
704 * @ooo_okay: allow the mapping of a socket to a queue to be changed
705 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
707 * @sw_hash: indicates hash was computed in software stack
708 * @wifi_acked_valid: wifi_acked was set
709 * @wifi_acked: whether frame was acked on wifi or not
710 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
711 * @encapsulation: indicates the inner headers in the skbuff are valid
712 * @encap_hdr_csum: software checksum is needed
713 * @csum_valid: checksum is already valid
714 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
715 * @csum_complete_sw: checksum was completed by software
716 * @csum_level: indicates the number of consecutive checksums found in
717 * the packet minus one that have been verified as
718 * CHECKSUM_UNNECESSARY (max 3)
719 * @dst_pending_confirm: need to confirm neighbour
720 * @decrypted: Decrypted SKB
721 * @slow_gro: state present at GRO time, slower prepare step required
722 * @napi_id: id of the NAPI struct this skb came from
723 * @sender_cpu: (aka @napi_id) source CPU in XPS
724 * @secmark: security marking
725 * @mark: Generic packet mark
726 * @reserved_tailroom: (aka @mark) number of bytes of free space available
727 * at the tail of an sk_buff
728 * @vlan_present: VLAN tag is present
729 * @vlan_proto: vlan encapsulation protocol
730 * @vlan_tci: vlan tag control information
731 * @inner_protocol: Protocol (encapsulation)
732 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
733 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
734 * @inner_transport_header: Inner transport layer header (encapsulation)
735 * @inner_network_header: Network layer header (encapsulation)
736 * @inner_mac_header: Link layer header (encapsulation)
737 * @transport_header: Transport layer header
738 * @network_header: Network layer header
739 * @mac_header: Link layer header
740 * @kcov_handle: KCOV remote handle for remote coverage collection
741 * @tail: Tail pointer
743 * @head: Head of buffer
744 * @data: Data head pointer
745 * @truesize: Buffer size
746 * @users: User count - see {datagram,tcp}.c
747 * @extensions: allocated extensions, valid if active_extensions is nonzero
753 /* These two members must be first. */
754 struct sk_buff *next;
755 struct sk_buff *prev;
758 struct net_device *dev;
759 /* Some protocols might use this space to store information,
760 * while device pointer would be NULL.
761 * UDP receive path is one user.
763 unsigned long dev_scratch;
766 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
767 struct list_head list;
772 int ip_defrag_offset;
777 u64 skb_mstamp_ns; /* earliest departure time */
780 * This is the control buffer. It is free to use for every
781 * layer. Please put your private variables there. If you
782 * want to keep them across layers you have to do a skb_clone()
783 * first. This is owned by whoever has the skb queued ATM.
785 char cb[48] __aligned(8);
789 unsigned long _skb_refdst;
790 void (*destructor)(struct sk_buff *skb);
792 struct list_head tcp_tsorted_anchor;
793 #ifdef CONFIG_NET_SOCK_MSG
794 unsigned long _sk_redir;
798 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
806 /* Following fields are _not_ copied in __copy_skb_header()
807 * Note that queue_mapping is here mostly to fill a hole.
811 /* if you move cloned around you also must adapt those constants */
812 #ifdef __BIG_ENDIAN_BITFIELD
813 #define CLONED_MASK (1 << 7)
815 #define CLONED_MASK 1
817 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
820 __u8 __cloned_offset[0];
828 pp_recycle:1; /* page_pool recycle indicator */
829 #ifdef CONFIG_SKB_EXTENSIONS
830 __u8 active_extensions;
833 /* fields enclosed in headers_start/headers_end are copied
834 * using a single memcpy() in __copy_skb_header()
837 __u32 headers_start[0];
840 /* if you move pkt_type around you also must adapt those constants */
841 #ifdef __BIG_ENDIAN_BITFIELD
842 #define PKT_TYPE_MAX (7 << 5)
844 #define PKT_TYPE_MAX 7
846 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
849 __u8 __pkt_type_offset[0];
859 __u8 wifi_acked_valid:1;
862 /* Indicates the inner headers are valid in the skbuff. */
863 __u8 encapsulation:1;
864 __u8 encap_hdr_csum:1;
867 #ifdef __BIG_ENDIAN_BITFIELD
868 #define PKT_VLAN_PRESENT_BIT 7
870 #define PKT_VLAN_PRESENT_BIT 0
872 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
874 __u8 __pkt_vlan_present_offset[0];
877 __u8 csum_complete_sw:1;
879 __u8 csum_not_inet:1;
880 __u8 dst_pending_confirm:1;
881 #ifdef CONFIG_IPV6_NDISC_NODETYPE
882 __u8 ndisc_nodetype:2;
885 __u8 ipvs_property:1;
886 __u8 inner_protocol_type:1;
887 __u8 remcsum_offload:1;
888 #ifdef CONFIG_NET_SWITCHDEV
889 __u8 offload_fwd_mark:1;
890 __u8 offload_l3_fwd_mark:1;
892 #ifdef CONFIG_NET_CLS_ACT
893 __u8 tc_skip_classify:1;
894 __u8 tc_at_ingress:1;
897 #ifdef CONFIG_NET_REDIRECT
900 #ifdef CONFIG_TLS_DEVICE
905 #ifdef CONFIG_NET_SCHED
906 __u16 tc_index; /* traffic control index */
921 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
923 unsigned int napi_id;
924 unsigned int sender_cpu;
927 #ifdef CONFIG_NETWORK_SECMARK
933 __u32 reserved_tailroom;
937 __be16 inner_protocol;
941 __u16 inner_transport_header;
942 __u16 inner_network_header;
943 __u16 inner_mac_header;
946 __u16 transport_header;
947 __u16 network_header;
955 __u32 headers_end[0];
958 /* These elements must be at the end, see alloc_skb() for details. */
963 unsigned int truesize;
966 #ifdef CONFIG_SKB_EXTENSIONS
967 /* only useable after checking ->active_extensions != 0 */
968 struct skb_ext *extensions;
974 * Handling routines are only of interest to the kernel
977 #define SKB_ALLOC_FCLONE 0x01
978 #define SKB_ALLOC_RX 0x02
979 #define SKB_ALLOC_NAPI 0x04
982 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
985 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
987 return unlikely(skb->pfmemalloc);
991 * skb might have a dst pointer attached, refcounted or not.
992 * _skb_refdst low order bit is set if refcount was _not_ taken
994 #define SKB_DST_NOREF 1UL
995 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
998 * skb_dst - returns skb dst_entry
1001 * Returns skb dst_entry, regardless of reference taken or not.
1003 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1005 /* If refdst was not refcounted, check we still are in a
1006 * rcu_read_lock section
1008 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1009 !rcu_read_lock_held() &&
1010 !rcu_read_lock_bh_held());
1011 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1015 * skb_dst_set - sets skb dst
1019 * Sets skb dst, assuming a reference was taken on dst and should
1020 * be released by skb_dst_drop()
1022 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1024 skb->slow_gro |= !!dst;
1025 skb->_skb_refdst = (unsigned long)dst;
1029 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1033 * Sets skb dst, assuming a reference was not taken on dst.
1034 * If dst entry is cached, we do not take reference and dst_release
1035 * will be avoided by refdst_drop. If dst entry is not cached, we take
1036 * reference, so that last dst_release can destroy the dst immediately.
1038 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1040 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1041 skb->slow_gro |= !!dst;
1042 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1046 * skb_dst_is_noref - Test if skb dst isn't refcounted
1049 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1051 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1055 * skb_rtable - Returns the skb &rtable
1058 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1060 return (struct rtable *)skb_dst(skb);
1063 /* For mangling skb->pkt_type from user space side from applications
1064 * such as nft, tc, etc, we only allow a conservative subset of
1065 * possible pkt_types to be set.
1067 static inline bool skb_pkt_type_ok(u32 ptype)
1069 return ptype <= PACKET_OTHERHOST;
1073 * skb_napi_id - Returns the skb's NAPI id
1076 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1078 #ifdef CONFIG_NET_RX_BUSY_POLL
1079 return skb->napi_id;
1086 * skb_unref - decrement the skb's reference count
1089 * Returns true if we can free the skb.
1091 static inline bool skb_unref(struct sk_buff *skb)
1095 if (likely(refcount_read(&skb->users) == 1))
1097 else if (likely(!refcount_dec_and_test(&skb->users)))
1103 void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1106 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1107 * @skb: buffer to free
1109 static inline void kfree_skb(struct sk_buff *skb)
1111 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1114 void skb_release_head_state(struct sk_buff *skb);
1115 void kfree_skb_list(struct sk_buff *segs);
1116 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1117 void skb_tx_error(struct sk_buff *skb);
1119 #ifdef CONFIG_TRACEPOINTS
1120 void consume_skb(struct sk_buff *skb);
1122 static inline void consume_skb(struct sk_buff *skb)
1124 return kfree_skb(skb);
1128 void __consume_stateless_skb(struct sk_buff *skb);
1129 void __kfree_skb(struct sk_buff *skb);
1130 extern struct kmem_cache *skbuff_head_cache;
1132 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1133 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1134 bool *fragstolen, int *delta_truesize);
1136 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1138 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1139 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1140 struct sk_buff *build_skb_around(struct sk_buff *skb,
1141 void *data, unsigned int frag_size);
1143 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1146 * alloc_skb - allocate a network buffer
1147 * @size: size to allocate
1148 * @priority: allocation mask
1150 * This function is a convenient wrapper around __alloc_skb().
1152 static inline struct sk_buff *alloc_skb(unsigned int size,
1155 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1158 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1159 unsigned long data_len,
1163 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1165 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1166 struct sk_buff_fclones {
1167 struct sk_buff skb1;
1169 struct sk_buff skb2;
1171 refcount_t fclone_ref;
1175 * skb_fclone_busy - check if fclone is busy
1179 * Returns true if skb is a fast clone, and its clone is not freed.
1180 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1181 * so we also check that this didnt happen.
1183 static inline bool skb_fclone_busy(const struct sock *sk,
1184 const struct sk_buff *skb)
1186 const struct sk_buff_fclones *fclones;
1188 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1190 return skb->fclone == SKB_FCLONE_ORIG &&
1191 refcount_read(&fclones->fclone_ref) > 1 &&
1192 READ_ONCE(fclones->skb2.sk) == sk;
1196 * alloc_skb_fclone - allocate a network buffer from fclone cache
1197 * @size: size to allocate
1198 * @priority: allocation mask
1200 * This function is a convenient wrapper around __alloc_skb().
1202 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1205 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1208 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1209 void skb_headers_offset_update(struct sk_buff *skb, int off);
1210 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1211 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1212 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1213 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1214 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1215 gfp_t gfp_mask, bool fclone);
1216 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1219 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1222 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1223 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1224 unsigned int headroom);
1225 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1226 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1227 int newtailroom, gfp_t priority);
1228 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1229 int offset, int len);
1230 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1231 int offset, int len);
1232 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1233 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1236 * skb_pad - zero pad the tail of an skb
1237 * @skb: buffer to pad
1238 * @pad: space to pad
1240 * Ensure that a buffer is followed by a padding area that is zero
1241 * filled. Used by network drivers which may DMA or transfer data
1242 * beyond the buffer end onto the wire.
1244 * May return error in out of memory cases. The skb is freed on error.
1246 static inline int skb_pad(struct sk_buff *skb, int pad)
1248 return __skb_pad(skb, pad, true);
1250 #define dev_kfree_skb(a) consume_skb(a)
1252 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1253 int offset, size_t size);
1255 struct skb_seq_state {
1259 __u32 stepped_offset;
1260 struct sk_buff *root_skb;
1261 struct sk_buff *cur_skb;
1266 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1267 unsigned int to, struct skb_seq_state *st);
1268 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1269 struct skb_seq_state *st);
1270 void skb_abort_seq_read(struct skb_seq_state *st);
1272 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1273 unsigned int to, struct ts_config *config);
1276 * Packet hash types specify the type of hash in skb_set_hash.
1278 * Hash types refer to the protocol layer addresses which are used to
1279 * construct a packet's hash. The hashes are used to differentiate or identify
1280 * flows of the protocol layer for the hash type. Hash types are either
1281 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1283 * Properties of hashes:
1285 * 1) Two packets in different flows have different hash values
1286 * 2) Two packets in the same flow should have the same hash value
1288 * A hash at a higher layer is considered to be more specific. A driver should
1289 * set the most specific hash possible.
1291 * A driver cannot indicate a more specific hash than the layer at which a hash
1292 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1294 * A driver may indicate a hash level which is less specific than the
1295 * actual layer the hash was computed on. For instance, a hash computed
1296 * at L4 may be considered an L3 hash. This should only be done if the
1297 * driver can't unambiguously determine that the HW computed the hash at
1298 * the higher layer. Note that the "should" in the second property above
1301 enum pkt_hash_types {
1302 PKT_HASH_TYPE_NONE, /* Undefined type */
1303 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1304 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1305 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1308 static inline void skb_clear_hash(struct sk_buff *skb)
1315 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1318 skb_clear_hash(skb);
1322 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1324 skb->l4_hash = is_l4;
1325 skb->sw_hash = is_sw;
1330 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1332 /* Used by drivers to set hash from HW */
1333 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1337 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1339 __skb_set_hash(skb, hash, true, is_l4);
1342 void __skb_get_hash(struct sk_buff *skb);
1343 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1344 u32 skb_get_poff(const struct sk_buff *skb);
1345 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1346 const struct flow_keys_basic *keys, int hlen);
1347 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1348 const void *data, int hlen_proto);
1350 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1351 int thoff, u8 ip_proto)
1353 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1356 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1357 const struct flow_dissector_key *key,
1358 unsigned int key_count);
1360 struct bpf_flow_dissector;
1361 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1362 __be16 proto, int nhoff, int hlen, unsigned int flags);
1364 bool __skb_flow_dissect(const struct net *net,
1365 const struct sk_buff *skb,
1366 struct flow_dissector *flow_dissector,
1367 void *target_container, const void *data,
1368 __be16 proto, int nhoff, int hlen, unsigned int flags);
1370 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1371 struct flow_dissector *flow_dissector,
1372 void *target_container, unsigned int flags)
1374 return __skb_flow_dissect(NULL, skb, flow_dissector,
1375 target_container, NULL, 0, 0, 0, flags);
1378 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1379 struct flow_keys *flow,
1382 memset(flow, 0, sizeof(*flow));
1383 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1384 flow, NULL, 0, 0, 0, flags);
1388 skb_flow_dissect_flow_keys_basic(const struct net *net,
1389 const struct sk_buff *skb,
1390 struct flow_keys_basic *flow,
1391 const void *data, __be16 proto,
1392 int nhoff, int hlen, unsigned int flags)
1394 memset(flow, 0, sizeof(*flow));
1395 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1396 data, proto, nhoff, hlen, flags);
1399 void skb_flow_dissect_meta(const struct sk_buff *skb,
1400 struct flow_dissector *flow_dissector,
1401 void *target_container);
1403 /* Gets a skb connection tracking info, ctinfo map should be a
1404 * map of mapsize to translate enum ip_conntrack_info states
1408 skb_flow_dissect_ct(const struct sk_buff *skb,
1409 struct flow_dissector *flow_dissector,
1410 void *target_container,
1411 u16 *ctinfo_map, size_t mapsize,
1412 bool post_ct, u16 zone);
1414 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1415 struct flow_dissector *flow_dissector,
1416 void *target_container);
1418 void skb_flow_dissect_hash(const struct sk_buff *skb,
1419 struct flow_dissector *flow_dissector,
1420 void *target_container);
1422 static inline __u32 skb_get_hash(struct sk_buff *skb)
1424 if (!skb->l4_hash && !skb->sw_hash)
1425 __skb_get_hash(skb);
1430 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1432 if (!skb->l4_hash && !skb->sw_hash) {
1433 struct flow_keys keys;
1434 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1436 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1442 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1443 const siphash_key_t *perturb);
1445 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1450 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1452 to->hash = from->hash;
1453 to->sw_hash = from->sw_hash;
1454 to->l4_hash = from->l4_hash;
1457 static inline void skb_copy_decrypted(struct sk_buff *to,
1458 const struct sk_buff *from)
1460 #ifdef CONFIG_TLS_DEVICE
1461 to->decrypted = from->decrypted;
1465 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1466 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1468 return skb->head + skb->end;
1471 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1476 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1481 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1486 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1488 return skb->end - skb->head;
1491 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1493 skb->end = skb->head + offset;
1498 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1500 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1502 return &skb_shinfo(skb)->hwtstamps;
1505 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1507 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1509 return is_zcopy ? skb_uarg(skb) : NULL;
1512 static inline void net_zcopy_get(struct ubuf_info *uarg)
1514 refcount_inc(&uarg->refcnt);
1517 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1519 skb_shinfo(skb)->destructor_arg = uarg;
1520 skb_shinfo(skb)->flags |= uarg->flags;
1523 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1526 if (skb && uarg && !skb_zcopy(skb)) {
1527 if (unlikely(have_ref && *have_ref))
1530 net_zcopy_get(uarg);
1531 skb_zcopy_init(skb, uarg);
1535 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1537 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1538 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1541 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1543 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1546 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1548 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1551 static inline void net_zcopy_put(struct ubuf_info *uarg)
1554 uarg->callback(NULL, uarg, true);
1557 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1560 if (uarg->callback == msg_zerocopy_callback)
1561 msg_zerocopy_put_abort(uarg, have_uref);
1563 net_zcopy_put(uarg);
1567 /* Release a reference on a zerocopy structure */
1568 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1570 struct ubuf_info *uarg = skb_zcopy(skb);
1573 if (!skb_zcopy_is_nouarg(skb))
1574 uarg->callback(skb, uarg, zerocopy_success);
1576 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
1580 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1585 /* Iterate through singly-linked GSO fragments of an skb. */
1586 #define skb_list_walk_safe(first, skb, next_skb) \
1587 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1588 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1590 static inline void skb_list_del_init(struct sk_buff *skb)
1592 __list_del_entry(&skb->list);
1593 skb_mark_not_on_list(skb);
1597 * skb_queue_empty - check if a queue is empty
1600 * Returns true if the queue is empty, false otherwise.
1602 static inline int skb_queue_empty(const struct sk_buff_head *list)
1604 return list->next == (const struct sk_buff *) list;
1608 * skb_queue_empty_lockless - check if a queue is empty
1611 * Returns true if the queue is empty, false otherwise.
1612 * This variant can be used in lockless contexts.
1614 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1616 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1621 * skb_queue_is_last - check if skb is the last entry in the queue
1625 * Returns true if @skb is the last buffer on the list.
1627 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1628 const struct sk_buff *skb)
1630 return skb->next == (const struct sk_buff *) list;
1634 * skb_queue_is_first - check if skb is the first entry in the queue
1638 * Returns true if @skb is the first buffer on the list.
1640 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1641 const struct sk_buff *skb)
1643 return skb->prev == (const struct sk_buff *) list;
1647 * skb_queue_next - return the next packet in the queue
1649 * @skb: current buffer
1651 * Return the next packet in @list after @skb. It is only valid to
1652 * call this if skb_queue_is_last() evaluates to false.
1654 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1655 const struct sk_buff *skb)
1657 /* This BUG_ON may seem severe, but if we just return then we
1658 * are going to dereference garbage.
1660 BUG_ON(skb_queue_is_last(list, skb));
1665 * skb_queue_prev - return the prev packet in the queue
1667 * @skb: current buffer
1669 * Return the prev packet in @list before @skb. It is only valid to
1670 * call this if skb_queue_is_first() evaluates to false.
1672 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1673 const struct sk_buff *skb)
1675 /* This BUG_ON may seem severe, but if we just return then we
1676 * are going to dereference garbage.
1678 BUG_ON(skb_queue_is_first(list, skb));
1683 * skb_get - reference buffer
1684 * @skb: buffer to reference
1686 * Makes another reference to a socket buffer and returns a pointer
1689 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1691 refcount_inc(&skb->users);
1696 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1700 * skb_cloned - is the buffer a clone
1701 * @skb: buffer to check
1703 * Returns true if the buffer was generated with skb_clone() and is
1704 * one of multiple shared copies of the buffer. Cloned buffers are
1705 * shared data so must not be written to under normal circumstances.
1707 static inline int skb_cloned(const struct sk_buff *skb)
1709 return skb->cloned &&
1710 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1713 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1715 might_sleep_if(gfpflags_allow_blocking(pri));
1717 if (skb_cloned(skb))
1718 return pskb_expand_head(skb, 0, 0, pri);
1723 /* This variant of skb_unclone() makes sure skb->truesize
1724 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1726 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1727 * when various debugging features are in place.
1729 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
1730 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1732 might_sleep_if(gfpflags_allow_blocking(pri));
1734 if (skb_cloned(skb))
1735 return __skb_unclone_keeptruesize(skb, pri);
1740 * skb_header_cloned - is the header a clone
1741 * @skb: buffer to check
1743 * Returns true if modifying the header part of the buffer requires
1744 * the data to be copied.
1746 static inline int skb_header_cloned(const struct sk_buff *skb)
1753 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1754 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1755 return dataref != 1;
1758 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1760 might_sleep_if(gfpflags_allow_blocking(pri));
1762 if (skb_header_cloned(skb))
1763 return pskb_expand_head(skb, 0, 0, pri);
1769 * __skb_header_release - release reference to header
1770 * @skb: buffer to operate on
1772 static inline void __skb_header_release(struct sk_buff *skb)
1775 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1780 * skb_shared - is the buffer shared
1781 * @skb: buffer to check
1783 * Returns true if more than one person has a reference to this
1786 static inline int skb_shared(const struct sk_buff *skb)
1788 return refcount_read(&skb->users) != 1;
1792 * skb_share_check - check if buffer is shared and if so clone it
1793 * @skb: buffer to check
1794 * @pri: priority for memory allocation
1796 * If the buffer is shared the buffer is cloned and the old copy
1797 * drops a reference. A new clone with a single reference is returned.
1798 * If the buffer is not shared the original buffer is returned. When
1799 * being called from interrupt status or with spinlocks held pri must
1802 * NULL is returned on a memory allocation failure.
1804 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1806 might_sleep_if(gfpflags_allow_blocking(pri));
1807 if (skb_shared(skb)) {
1808 struct sk_buff *nskb = skb_clone(skb, pri);
1820 * Copy shared buffers into a new sk_buff. We effectively do COW on
1821 * packets to handle cases where we have a local reader and forward
1822 * and a couple of other messy ones. The normal one is tcpdumping
1823 * a packet thats being forwarded.
1827 * skb_unshare - make a copy of a shared buffer
1828 * @skb: buffer to check
1829 * @pri: priority for memory allocation
1831 * If the socket buffer is a clone then this function creates a new
1832 * copy of the data, drops a reference count on the old copy and returns
1833 * the new copy with the reference count at 1. If the buffer is not a clone
1834 * the original buffer is returned. When called with a spinlock held or
1835 * from interrupt state @pri must be %GFP_ATOMIC
1837 * %NULL is returned on a memory allocation failure.
1839 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1842 might_sleep_if(gfpflags_allow_blocking(pri));
1843 if (skb_cloned(skb)) {
1844 struct sk_buff *nskb = skb_copy(skb, pri);
1846 /* Free our shared copy */
1857 * skb_peek - peek at the head of an &sk_buff_head
1858 * @list_: list to peek at
1860 * Peek an &sk_buff. Unlike most other operations you _MUST_
1861 * be careful with this one. A peek leaves the buffer on the
1862 * list and someone else may run off with it. You must hold
1863 * the appropriate locks or have a private queue to do this.
1865 * Returns %NULL for an empty list or a pointer to the head element.
1866 * The reference count is not incremented and the reference is therefore
1867 * volatile. Use with caution.
1869 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1871 struct sk_buff *skb = list_->next;
1873 if (skb == (struct sk_buff *)list_)
1879 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1880 * @list_: list to peek at
1882 * Like skb_peek(), but the caller knows that the list is not empty.
1884 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1890 * skb_peek_next - peek skb following the given one from a queue
1891 * @skb: skb to start from
1892 * @list_: list to peek at
1894 * Returns %NULL when the end of the list is met or a pointer to the
1895 * next element. The reference count is not incremented and the
1896 * reference is therefore volatile. Use with caution.
1898 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1899 const struct sk_buff_head *list_)
1901 struct sk_buff *next = skb->next;
1903 if (next == (struct sk_buff *)list_)
1909 * skb_peek_tail - peek at the tail of an &sk_buff_head
1910 * @list_: list to peek at
1912 * Peek an &sk_buff. Unlike most other operations you _MUST_
1913 * be careful with this one. A peek leaves the buffer on the
1914 * list and someone else may run off with it. You must hold
1915 * the appropriate locks or have a private queue to do this.
1917 * Returns %NULL for an empty list or a pointer to the tail element.
1918 * The reference count is not incremented and the reference is therefore
1919 * volatile. Use with caution.
1921 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1923 struct sk_buff *skb = READ_ONCE(list_->prev);
1925 if (skb == (struct sk_buff *)list_)
1932 * skb_queue_len - get queue length
1933 * @list_: list to measure
1935 * Return the length of an &sk_buff queue.
1937 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1943 * skb_queue_len_lockless - get queue length
1944 * @list_: list to measure
1946 * Return the length of an &sk_buff queue.
1947 * This variant can be used in lockless contexts.
1949 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1951 return READ_ONCE(list_->qlen);
1955 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1956 * @list: queue to initialize
1958 * This initializes only the list and queue length aspects of
1959 * an sk_buff_head object. This allows to initialize the list
1960 * aspects of an sk_buff_head without reinitializing things like
1961 * the spinlock. It can also be used for on-stack sk_buff_head
1962 * objects where the spinlock is known to not be used.
1964 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1966 list->prev = list->next = (struct sk_buff *)list;
1971 * This function creates a split out lock class for each invocation;
1972 * this is needed for now since a whole lot of users of the skb-queue
1973 * infrastructure in drivers have different locking usage (in hardirq)
1974 * than the networking core (in softirq only). In the long run either the
1975 * network layer or drivers should need annotation to consolidate the
1976 * main types of usage into 3 classes.
1978 static inline void skb_queue_head_init(struct sk_buff_head *list)
1980 spin_lock_init(&list->lock);
1981 __skb_queue_head_init(list);
1984 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1985 struct lock_class_key *class)
1987 skb_queue_head_init(list);
1988 lockdep_set_class(&list->lock, class);
1992 * Insert an sk_buff on a list.
1994 * The "__skb_xxxx()" functions are the non-atomic ones that
1995 * can only be called with interrupts disabled.
1997 static inline void __skb_insert(struct sk_buff *newsk,
1998 struct sk_buff *prev, struct sk_buff *next,
1999 struct sk_buff_head *list)
2001 /* See skb_queue_empty_lockless() and skb_peek_tail()
2002 * for the opposite READ_ONCE()
2004 WRITE_ONCE(newsk->next, next);
2005 WRITE_ONCE(newsk->prev, prev);
2006 WRITE_ONCE(next->prev, newsk);
2007 WRITE_ONCE(prev->next, newsk);
2008 WRITE_ONCE(list->qlen, list->qlen + 1);
2011 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2012 struct sk_buff *prev,
2013 struct sk_buff *next)
2015 struct sk_buff *first = list->next;
2016 struct sk_buff *last = list->prev;
2018 WRITE_ONCE(first->prev, prev);
2019 WRITE_ONCE(prev->next, first);
2021 WRITE_ONCE(last->next, next);
2022 WRITE_ONCE(next->prev, last);
2026 * skb_queue_splice - join two skb lists, this is designed for stacks
2027 * @list: the new list to add
2028 * @head: the place to add it in the first list
2030 static inline void skb_queue_splice(const struct sk_buff_head *list,
2031 struct sk_buff_head *head)
2033 if (!skb_queue_empty(list)) {
2034 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2035 head->qlen += list->qlen;
2040 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2041 * @list: the new list to add
2042 * @head: the place to add it in the first list
2044 * The list at @list is reinitialised
2046 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2047 struct sk_buff_head *head)
2049 if (!skb_queue_empty(list)) {
2050 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2051 head->qlen += list->qlen;
2052 __skb_queue_head_init(list);
2057 * skb_queue_splice_tail - join two skb lists, each list being a queue
2058 * @list: the new list to add
2059 * @head: the place to add it in the first list
2061 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2062 struct sk_buff_head *head)
2064 if (!skb_queue_empty(list)) {
2065 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2066 head->qlen += list->qlen;
2071 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2072 * @list: the new list to add
2073 * @head: the place to add it in the first list
2075 * Each of the lists is a queue.
2076 * The list at @list is reinitialised
2078 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2079 struct sk_buff_head *head)
2081 if (!skb_queue_empty(list)) {
2082 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2083 head->qlen += list->qlen;
2084 __skb_queue_head_init(list);
2089 * __skb_queue_after - queue a buffer at the list head
2090 * @list: list to use
2091 * @prev: place after this buffer
2092 * @newsk: buffer to queue
2094 * Queue a buffer int the middle of a list. This function takes no locks
2095 * and you must therefore hold required locks before calling it.
2097 * A buffer cannot be placed on two lists at the same time.
2099 static inline void __skb_queue_after(struct sk_buff_head *list,
2100 struct sk_buff *prev,
2101 struct sk_buff *newsk)
2103 __skb_insert(newsk, prev, prev->next, list);
2106 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2107 struct sk_buff_head *list);
2109 static inline void __skb_queue_before(struct sk_buff_head *list,
2110 struct sk_buff *next,
2111 struct sk_buff *newsk)
2113 __skb_insert(newsk, next->prev, next, list);
2117 * __skb_queue_head - queue a buffer at the list head
2118 * @list: list to use
2119 * @newsk: buffer to queue
2121 * Queue a buffer at the start of a list. This function takes no locks
2122 * and you must therefore hold required locks before calling it.
2124 * A buffer cannot be placed on two lists at the same time.
2126 static inline void __skb_queue_head(struct sk_buff_head *list,
2127 struct sk_buff *newsk)
2129 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2131 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2134 * __skb_queue_tail - queue a buffer at the list tail
2135 * @list: list to use
2136 * @newsk: buffer to queue
2138 * Queue a buffer at the end of a list. This function takes no locks
2139 * and you must therefore hold required locks before calling it.
2141 * A buffer cannot be placed on two lists at the same time.
2143 static inline void __skb_queue_tail(struct sk_buff_head *list,
2144 struct sk_buff *newsk)
2146 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2148 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2151 * remove sk_buff from list. _Must_ be called atomically, and with
2154 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2155 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2157 struct sk_buff *next, *prev;
2159 WRITE_ONCE(list->qlen, list->qlen - 1);
2162 skb->next = skb->prev = NULL;
2163 WRITE_ONCE(next->prev, prev);
2164 WRITE_ONCE(prev->next, next);
2168 * __skb_dequeue - remove from the head of the queue
2169 * @list: list to dequeue from
2171 * Remove the head of the list. This function does not take any locks
2172 * so must be used with appropriate locks held only. The head item is
2173 * returned or %NULL if the list is empty.
2175 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2177 struct sk_buff *skb = skb_peek(list);
2179 __skb_unlink(skb, list);
2182 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2185 * __skb_dequeue_tail - remove from the tail of the queue
2186 * @list: list to dequeue from
2188 * Remove the tail of the list. This function does not take any locks
2189 * so must be used with appropriate locks held only. The tail item is
2190 * returned or %NULL if the list is empty.
2192 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2194 struct sk_buff *skb = skb_peek_tail(list);
2196 __skb_unlink(skb, list);
2199 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2202 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2204 return skb->data_len;
2207 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2209 return skb->len - skb->data_len;
2212 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2214 unsigned int i, len = 0;
2216 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2217 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2221 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2223 return skb_headlen(skb) + __skb_pagelen(skb);
2227 * __skb_fill_page_desc - initialise a paged fragment in an skb
2228 * @skb: buffer containing fragment to be initialised
2229 * @i: paged fragment index to initialise
2230 * @page: the page to use for this fragment
2231 * @off: the offset to the data with @page
2232 * @size: the length of the data
2234 * Initialises the @i'th fragment of @skb to point to &size bytes at
2235 * offset @off within @page.
2237 * Does not take any additional reference on the fragment.
2239 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2240 struct page *page, int off, int size)
2242 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2245 * Propagate page pfmemalloc to the skb if we can. The problem is
2246 * that not all callers have unique ownership of the page but rely
2247 * on page_is_pfmemalloc doing the right thing(tm).
2249 frag->bv_page = page;
2250 frag->bv_offset = off;
2251 skb_frag_size_set(frag, size);
2253 page = compound_head(page);
2254 if (page_is_pfmemalloc(page))
2255 skb->pfmemalloc = true;
2259 * skb_fill_page_desc - initialise a paged fragment in an skb
2260 * @skb: buffer containing fragment to be initialised
2261 * @i: paged fragment index to initialise
2262 * @page: the page to use for this fragment
2263 * @off: the offset to the data with @page
2264 * @size: the length of the data
2266 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2267 * @skb to point to @size bytes at offset @off within @page. In
2268 * addition updates @skb such that @i is the last fragment.
2270 * Does not take any additional reference on the fragment.
2272 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2273 struct page *page, int off, int size)
2275 __skb_fill_page_desc(skb, i, page, off, size);
2276 skb_shinfo(skb)->nr_frags = i + 1;
2279 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2280 int size, unsigned int truesize);
2282 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2283 unsigned int truesize);
2285 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2287 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2288 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2290 return skb->head + skb->tail;
2293 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2295 skb->tail = skb->data - skb->head;
2298 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2300 skb_reset_tail_pointer(skb);
2301 skb->tail += offset;
2304 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2305 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2310 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2312 skb->tail = skb->data;
2315 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2317 skb->tail = skb->data + offset;
2320 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2323 * Add data to an sk_buff
2325 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2326 void *skb_put(struct sk_buff *skb, unsigned int len);
2327 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2329 void *tmp = skb_tail_pointer(skb);
2330 SKB_LINEAR_ASSERT(skb);
2336 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2338 void *tmp = __skb_put(skb, len);
2340 memset(tmp, 0, len);
2344 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2347 void *tmp = __skb_put(skb, len);
2349 memcpy(tmp, data, len);
2353 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2355 *(u8 *)__skb_put(skb, 1) = val;
2358 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2360 void *tmp = skb_put(skb, len);
2362 memset(tmp, 0, len);
2367 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2370 void *tmp = skb_put(skb, len);
2372 memcpy(tmp, data, len);
2377 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2379 *(u8 *)skb_put(skb, 1) = val;
2382 void *skb_push(struct sk_buff *skb, unsigned int len);
2383 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2390 void *skb_pull(struct sk_buff *skb, unsigned int len);
2391 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2394 BUG_ON(skb->len < skb->data_len);
2395 return skb->data += len;
2398 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2400 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2403 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2405 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2407 if (len > skb_headlen(skb) &&
2408 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2411 return skb->data += len;
2414 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2416 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2419 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2421 if (likely(len <= skb_headlen(skb)))
2423 if (unlikely(len > skb->len))
2425 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2428 void skb_condense(struct sk_buff *skb);
2431 * skb_headroom - bytes at buffer head
2432 * @skb: buffer to check
2434 * Return the number of bytes of free space at the head of an &sk_buff.
2436 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2438 return skb->data - skb->head;
2442 * skb_tailroom - bytes at buffer end
2443 * @skb: buffer to check
2445 * Return the number of bytes of free space at the tail of an sk_buff
2447 static inline int skb_tailroom(const struct sk_buff *skb)
2449 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2453 * skb_availroom - bytes at buffer end
2454 * @skb: buffer to check
2456 * Return the number of bytes of free space at the tail of an sk_buff
2457 * allocated by sk_stream_alloc()
2459 static inline int skb_availroom(const struct sk_buff *skb)
2461 if (skb_is_nonlinear(skb))
2464 return skb->end - skb->tail - skb->reserved_tailroom;
2468 * skb_reserve - adjust headroom
2469 * @skb: buffer to alter
2470 * @len: bytes to move
2472 * Increase the headroom of an empty &sk_buff by reducing the tail
2473 * room. This is only allowed for an empty buffer.
2475 static inline void skb_reserve(struct sk_buff *skb, int len)
2482 * skb_tailroom_reserve - adjust reserved_tailroom
2483 * @skb: buffer to alter
2484 * @mtu: maximum amount of headlen permitted
2485 * @needed_tailroom: minimum amount of reserved_tailroom
2487 * Set reserved_tailroom so that headlen can be as large as possible but
2488 * not larger than mtu and tailroom cannot be smaller than
2490 * The required headroom should already have been reserved before using
2493 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2494 unsigned int needed_tailroom)
2496 SKB_LINEAR_ASSERT(skb);
2497 if (mtu < skb_tailroom(skb) - needed_tailroom)
2498 /* use at most mtu */
2499 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2501 /* use up to all available space */
2502 skb->reserved_tailroom = needed_tailroom;
2505 #define ENCAP_TYPE_ETHER 0
2506 #define ENCAP_TYPE_IPPROTO 1
2508 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2511 skb->inner_protocol = protocol;
2512 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2515 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2518 skb->inner_ipproto = ipproto;
2519 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2522 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2524 skb->inner_mac_header = skb->mac_header;
2525 skb->inner_network_header = skb->network_header;
2526 skb->inner_transport_header = skb->transport_header;
2529 static inline void skb_reset_mac_len(struct sk_buff *skb)
2531 skb->mac_len = skb->network_header - skb->mac_header;
2534 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2537 return skb->head + skb->inner_transport_header;
2540 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2542 return skb_inner_transport_header(skb) - skb->data;
2545 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2547 skb->inner_transport_header = skb->data - skb->head;
2550 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2553 skb_reset_inner_transport_header(skb);
2554 skb->inner_transport_header += offset;
2557 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2559 return skb->head + skb->inner_network_header;
2562 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2564 skb->inner_network_header = skb->data - skb->head;
2567 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2570 skb_reset_inner_network_header(skb);
2571 skb->inner_network_header += offset;
2574 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2576 return skb->head + skb->inner_mac_header;
2579 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2581 skb->inner_mac_header = skb->data - skb->head;
2584 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2587 skb_reset_inner_mac_header(skb);
2588 skb->inner_mac_header += offset;
2590 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2592 return skb->transport_header != (typeof(skb->transport_header))~0U;
2595 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2597 return skb->head + skb->transport_header;
2600 static inline void skb_reset_transport_header(struct sk_buff *skb)
2602 skb->transport_header = skb->data - skb->head;
2605 static inline void skb_set_transport_header(struct sk_buff *skb,
2608 skb_reset_transport_header(skb);
2609 skb->transport_header += offset;
2612 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2614 return skb->head + skb->network_header;
2617 static inline void skb_reset_network_header(struct sk_buff *skb)
2619 skb->network_header = skb->data - skb->head;
2622 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2624 skb_reset_network_header(skb);
2625 skb->network_header += offset;
2628 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2630 return skb->head + skb->mac_header;
2633 static inline int skb_mac_offset(const struct sk_buff *skb)
2635 return skb_mac_header(skb) - skb->data;
2638 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2640 return skb->network_header - skb->mac_header;
2643 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2645 return skb->mac_header != (typeof(skb->mac_header))~0U;
2648 static inline void skb_unset_mac_header(struct sk_buff *skb)
2650 skb->mac_header = (typeof(skb->mac_header))~0U;
2653 static inline void skb_reset_mac_header(struct sk_buff *skb)
2655 skb->mac_header = skb->data - skb->head;
2658 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2660 skb_reset_mac_header(skb);
2661 skb->mac_header += offset;
2664 static inline void skb_pop_mac_header(struct sk_buff *skb)
2666 skb->mac_header = skb->network_header;
2669 static inline void skb_probe_transport_header(struct sk_buff *skb)
2671 struct flow_keys_basic keys;
2673 if (skb_transport_header_was_set(skb))
2676 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2678 skb_set_transport_header(skb, keys.control.thoff);
2681 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2683 if (skb_mac_header_was_set(skb)) {
2684 const unsigned char *old_mac = skb_mac_header(skb);
2686 skb_set_mac_header(skb, -skb->mac_len);
2687 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2691 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2693 return skb->csum_start - skb_headroom(skb);
2696 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2698 return skb->head + skb->csum_start;
2701 static inline int skb_transport_offset(const struct sk_buff *skb)
2703 return skb_transport_header(skb) - skb->data;
2706 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2708 return skb->transport_header - skb->network_header;
2711 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2713 return skb->inner_transport_header - skb->inner_network_header;
2716 static inline int skb_network_offset(const struct sk_buff *skb)
2718 return skb_network_header(skb) - skb->data;
2721 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2723 return skb_inner_network_header(skb) - skb->data;
2726 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2728 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2732 * CPUs often take a performance hit when accessing unaligned memory
2733 * locations. The actual performance hit varies, it can be small if the
2734 * hardware handles it or large if we have to take an exception and fix it
2737 * Since an ethernet header is 14 bytes network drivers often end up with
2738 * the IP header at an unaligned offset. The IP header can be aligned by
2739 * shifting the start of the packet by 2 bytes. Drivers should do this
2742 * skb_reserve(skb, NET_IP_ALIGN);
2744 * The downside to this alignment of the IP header is that the DMA is now
2745 * unaligned. On some architectures the cost of an unaligned DMA is high
2746 * and this cost outweighs the gains made by aligning the IP header.
2748 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2751 #ifndef NET_IP_ALIGN
2752 #define NET_IP_ALIGN 2
2756 * The networking layer reserves some headroom in skb data (via
2757 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2758 * the header has to grow. In the default case, if the header has to grow
2759 * 32 bytes or less we avoid the reallocation.
2761 * Unfortunately this headroom changes the DMA alignment of the resulting
2762 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2763 * on some architectures. An architecture can override this value,
2764 * perhaps setting it to a cacheline in size (since that will maintain
2765 * cacheline alignment of the DMA). It must be a power of 2.
2767 * Various parts of the networking layer expect at least 32 bytes of
2768 * headroom, you should not reduce this.
2770 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2771 * to reduce average number of cache lines per packet.
2772 * get_rps_cpu() for example only access one 64 bytes aligned block :
2773 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2776 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2779 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2781 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2783 if (WARN_ON(skb_is_nonlinear(skb)))
2786 skb_set_tail_pointer(skb, len);
2789 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2791 __skb_set_length(skb, len);
2794 void skb_trim(struct sk_buff *skb, unsigned int len);
2796 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2799 return ___pskb_trim(skb, len);
2800 __skb_trim(skb, len);
2804 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2806 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2810 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2811 * @skb: buffer to alter
2814 * This is identical to pskb_trim except that the caller knows that
2815 * the skb is not cloned so we should never get an error due to out-
2818 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2820 int err = pskb_trim(skb, len);
2824 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2826 unsigned int diff = len - skb->len;
2828 if (skb_tailroom(skb) < diff) {
2829 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2834 __skb_set_length(skb, len);
2839 * skb_orphan - orphan a buffer
2840 * @skb: buffer to orphan
2842 * If a buffer currently has an owner then we call the owner's
2843 * destructor function and make the @skb unowned. The buffer continues
2844 * to exist but is no longer charged to its former owner.
2846 static inline void skb_orphan(struct sk_buff *skb)
2848 if (skb->destructor) {
2849 skb->destructor(skb);
2850 skb->destructor = NULL;
2858 * skb_orphan_frags - orphan the frags contained in a buffer
2859 * @skb: buffer to orphan frags from
2860 * @gfp_mask: allocation mask for replacement pages
2862 * For each frag in the SKB which needs a destructor (i.e. has an
2863 * owner) create a copy of that frag and release the original
2864 * page by calling the destructor.
2866 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2868 if (likely(!skb_zcopy(skb)))
2870 if (!skb_zcopy_is_nouarg(skb) &&
2871 skb_uarg(skb)->callback == msg_zerocopy_callback)
2873 return skb_copy_ubufs(skb, gfp_mask);
2876 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2877 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2879 if (likely(!skb_zcopy(skb)))
2881 return skb_copy_ubufs(skb, gfp_mask);
2885 * __skb_queue_purge - empty a list
2886 * @list: list to empty
2888 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2889 * the list and one reference dropped. This function does not take the
2890 * list lock and the caller must hold the relevant locks to use it.
2892 static inline void __skb_queue_purge(struct sk_buff_head *list)
2894 struct sk_buff *skb;
2895 while ((skb = __skb_dequeue(list)) != NULL)
2898 void skb_queue_purge(struct sk_buff_head *list);
2900 unsigned int skb_rbtree_purge(struct rb_root *root);
2902 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2905 * netdev_alloc_frag - allocate a page fragment
2906 * @fragsz: fragment size
2908 * Allocates a frag from a page for receive buffer.
2909 * Uses GFP_ATOMIC allocations.
2911 static inline void *netdev_alloc_frag(unsigned int fragsz)
2913 return __netdev_alloc_frag_align(fragsz, ~0u);
2916 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2919 WARN_ON_ONCE(!is_power_of_2(align));
2920 return __netdev_alloc_frag_align(fragsz, -align);
2923 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2927 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2928 * @dev: network device to receive on
2929 * @length: length to allocate
2931 * Allocate a new &sk_buff and assign it a usage count of one. The
2932 * buffer has unspecified headroom built in. Users should allocate
2933 * the headroom they think they need without accounting for the
2934 * built in space. The built in space is used for optimisations.
2936 * %NULL is returned if there is no free memory. Although this function
2937 * allocates memory it can be called from an interrupt.
2939 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2940 unsigned int length)
2942 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2945 /* legacy helper around __netdev_alloc_skb() */
2946 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2949 return __netdev_alloc_skb(NULL, length, gfp_mask);
2952 /* legacy helper around netdev_alloc_skb() */
2953 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2955 return netdev_alloc_skb(NULL, length);
2959 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2960 unsigned int length, gfp_t gfp)
2962 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2964 if (NET_IP_ALIGN && skb)
2965 skb_reserve(skb, NET_IP_ALIGN);
2969 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2970 unsigned int length)
2972 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2975 static inline void skb_free_frag(void *addr)
2977 page_frag_free(addr);
2980 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2982 static inline void *napi_alloc_frag(unsigned int fragsz)
2984 return __napi_alloc_frag_align(fragsz, ~0u);
2987 static inline void *napi_alloc_frag_align(unsigned int fragsz,
2990 WARN_ON_ONCE(!is_power_of_2(align));
2991 return __napi_alloc_frag_align(fragsz, -align);
2994 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2995 unsigned int length, gfp_t gfp_mask);
2996 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2997 unsigned int length)
2999 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3001 void napi_consume_skb(struct sk_buff *skb, int budget);
3003 void napi_skb_free_stolen_head(struct sk_buff *skb);
3004 void __kfree_skb_defer(struct sk_buff *skb);
3007 * __dev_alloc_pages - allocate page for network Rx
3008 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3009 * @order: size of the allocation
3011 * Allocate a new page.
3013 * %NULL is returned if there is no free memory.
3015 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3018 /* This piece of code contains several assumptions.
3019 * 1. This is for device Rx, therefor a cold page is preferred.
3020 * 2. The expectation is the user wants a compound page.
3021 * 3. If requesting a order 0 page it will not be compound
3022 * due to the check to see if order has a value in prep_new_page
3023 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3024 * code in gfp_to_alloc_flags that should be enforcing this.
3026 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3028 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3031 static inline struct page *dev_alloc_pages(unsigned int order)
3033 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3037 * __dev_alloc_page - allocate a page for network Rx
3038 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3040 * Allocate a new page.
3042 * %NULL is returned if there is no free memory.
3044 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3046 return __dev_alloc_pages(gfp_mask, 0);
3049 static inline struct page *dev_alloc_page(void)
3051 return dev_alloc_pages(0);
3055 * dev_page_is_reusable - check whether a page can be reused for network Rx
3056 * @page: the page to test
3058 * A page shouldn't be considered for reusing/recycling if it was allocated
3059 * under memory pressure or at a distant memory node.
3061 * Returns false if this page should be returned to page allocator, true
3064 static inline bool dev_page_is_reusable(const struct page *page)
3066 return likely(page_to_nid(page) == numa_mem_id() &&
3067 !page_is_pfmemalloc(page));
3071 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3072 * @page: The page that was allocated from skb_alloc_page
3073 * @skb: The skb that may need pfmemalloc set
3075 static inline void skb_propagate_pfmemalloc(const struct page *page,
3076 struct sk_buff *skb)
3078 if (page_is_pfmemalloc(page))
3079 skb->pfmemalloc = true;
3083 * skb_frag_off() - Returns the offset of a skb fragment
3084 * @frag: the paged fragment
3086 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3088 return frag->bv_offset;
3092 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3093 * @frag: skb fragment
3094 * @delta: value to add
3096 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3098 frag->bv_offset += delta;
3102 * skb_frag_off_set() - Sets the offset of a skb fragment
3103 * @frag: skb fragment
3104 * @offset: offset of fragment
3106 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3108 frag->bv_offset = offset;
3112 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3113 * @fragto: skb fragment where offset is set
3114 * @fragfrom: skb fragment offset is copied from
3116 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3117 const skb_frag_t *fragfrom)
3119 fragto->bv_offset = fragfrom->bv_offset;
3123 * skb_frag_page - retrieve the page referred to by a paged fragment
3124 * @frag: the paged fragment
3126 * Returns the &struct page associated with @frag.
3128 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3130 return frag->bv_page;
3134 * __skb_frag_ref - take an addition reference on a paged fragment.
3135 * @frag: the paged fragment
3137 * Takes an additional reference on the paged fragment @frag.
3139 static inline void __skb_frag_ref(skb_frag_t *frag)
3141 get_page(skb_frag_page(frag));
3145 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3147 * @f: the fragment offset.
3149 * Takes an additional reference on the @f'th paged fragment of @skb.
3151 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3153 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3157 * __skb_frag_unref - release a reference on a paged fragment.
3158 * @frag: the paged fragment
3159 * @recycle: recycle the page if allocated via page_pool
3161 * Releases a reference on the paged fragment @frag
3162 * or recycles the page via the page_pool API.
3164 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3166 struct page *page = skb_frag_page(frag);
3168 #ifdef CONFIG_PAGE_POOL
3169 if (recycle && page_pool_return_skb_page(page))
3176 * skb_frag_unref - release a reference on a paged fragment of an skb.
3178 * @f: the fragment offset
3180 * Releases a reference on the @f'th paged fragment of @skb.
3182 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3184 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3188 * skb_frag_address - gets the address of the data contained in a paged fragment
3189 * @frag: the paged fragment buffer
3191 * Returns the address of the data within @frag. The page must already
3194 static inline void *skb_frag_address(const skb_frag_t *frag)
3196 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3200 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3201 * @frag: the paged fragment buffer
3203 * Returns the address of the data within @frag. Checks that the page
3204 * is mapped and returns %NULL otherwise.
3206 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3208 void *ptr = page_address(skb_frag_page(frag));
3212 return ptr + skb_frag_off(frag);
3216 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3217 * @fragto: skb fragment where page is set
3218 * @fragfrom: skb fragment page is copied from
3220 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3221 const skb_frag_t *fragfrom)
3223 fragto->bv_page = fragfrom->bv_page;
3227 * __skb_frag_set_page - sets the page contained in a paged fragment
3228 * @frag: the paged fragment
3229 * @page: the page to set
3231 * Sets the fragment @frag to contain @page.
3233 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3235 frag->bv_page = page;
3239 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3241 * @f: the fragment offset
3242 * @page: the page to set
3244 * Sets the @f'th fragment of @skb to contain @page.
3246 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3249 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3252 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3255 * skb_frag_dma_map - maps a paged fragment via the DMA API
3256 * @dev: the device to map the fragment to
3257 * @frag: the paged fragment to map
3258 * @offset: the offset within the fragment (starting at the
3259 * fragment's own offset)
3260 * @size: the number of bytes to map
3261 * @dir: the direction of the mapping (``PCI_DMA_*``)
3263 * Maps the page associated with @frag to @device.
3265 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3266 const skb_frag_t *frag,
3267 size_t offset, size_t size,
3268 enum dma_data_direction dir)
3270 return dma_map_page(dev, skb_frag_page(frag),
3271 skb_frag_off(frag) + offset, size, dir);
3274 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3277 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3281 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3284 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3289 * skb_clone_writable - is the header of a clone writable
3290 * @skb: buffer to check
3291 * @len: length up to which to write
3293 * Returns true if modifying the header part of the cloned buffer
3294 * does not requires the data to be copied.
3296 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3298 return !skb_header_cloned(skb) &&
3299 skb_headroom(skb) + len <= skb->hdr_len;
3302 static inline int skb_try_make_writable(struct sk_buff *skb,
3303 unsigned int write_len)
3305 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3306 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3309 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3314 if (headroom > skb_headroom(skb))
3315 delta = headroom - skb_headroom(skb);
3317 if (delta || cloned)
3318 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3324 * skb_cow - copy header of skb when it is required
3325 * @skb: buffer to cow
3326 * @headroom: needed headroom
3328 * If the skb passed lacks sufficient headroom or its data part
3329 * is shared, data is reallocated. If reallocation fails, an error
3330 * is returned and original skb is not changed.
3332 * The result is skb with writable area skb->head...skb->tail
3333 * and at least @headroom of space at head.
3335 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3337 return __skb_cow(skb, headroom, skb_cloned(skb));
3341 * skb_cow_head - skb_cow but only making the head writable
3342 * @skb: buffer to cow
3343 * @headroom: needed headroom
3345 * This function is identical to skb_cow except that we replace the
3346 * skb_cloned check by skb_header_cloned. It should be used when
3347 * you only need to push on some header and do not need to modify
3350 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3352 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3356 * skb_padto - pad an skbuff up to a minimal size
3357 * @skb: buffer to pad
3358 * @len: minimal length
3360 * Pads up a buffer to ensure the trailing bytes exist and are
3361 * blanked. If the buffer already contains sufficient data it
3362 * is untouched. Otherwise it is extended. Returns zero on
3363 * success. The skb is freed on error.
3365 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3367 unsigned int size = skb->len;
3368 if (likely(size >= len))
3370 return skb_pad(skb, len - size);
3374 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3375 * @skb: buffer to pad
3376 * @len: minimal length
3377 * @free_on_error: free buffer on error
3379 * Pads up a buffer to ensure the trailing bytes exist and are
3380 * blanked. If the buffer already contains sufficient data it
3381 * is untouched. Otherwise it is extended. Returns zero on
3382 * success. The skb is freed on error if @free_on_error is true.
3384 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3388 unsigned int size = skb->len;
3390 if (unlikely(size < len)) {
3392 if (__skb_pad(skb, len, free_on_error))
3394 __skb_put(skb, len);
3400 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3401 * @skb: buffer to pad
3402 * @len: minimal length
3404 * Pads up a buffer to ensure the trailing bytes exist and are
3405 * blanked. If the buffer already contains sufficient data it
3406 * is untouched. Otherwise it is extended. Returns zero on
3407 * success. The skb is freed on error.
3409 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3411 return __skb_put_padto(skb, len, true);
3414 static inline int skb_add_data(struct sk_buff *skb,
3415 struct iov_iter *from, int copy)
3417 const int off = skb->len;
3419 if (skb->ip_summed == CHECKSUM_NONE) {
3421 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3423 skb->csum = csum_block_add(skb->csum, csum, off);
3426 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3429 __skb_trim(skb, off);
3433 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3434 const struct page *page, int off)
3439 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3441 return page == skb_frag_page(frag) &&
3442 off == skb_frag_off(frag) + skb_frag_size(frag);
3447 static inline int __skb_linearize(struct sk_buff *skb)
3449 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3453 * skb_linearize - convert paged skb to linear one
3454 * @skb: buffer to linarize
3456 * If there is no free memory -ENOMEM is returned, otherwise zero
3457 * is returned and the old skb data released.
3459 static inline int skb_linearize(struct sk_buff *skb)
3461 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3465 * skb_has_shared_frag - can any frag be overwritten
3466 * @skb: buffer to test
3468 * Return true if the skb has at least one frag that might be modified
3469 * by an external entity (as in vmsplice()/sendfile())
3471 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3473 return skb_is_nonlinear(skb) &&
3474 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3478 * skb_linearize_cow - make sure skb is linear and writable
3479 * @skb: buffer to process
3481 * If there is no free memory -ENOMEM is returned, otherwise zero
3482 * is returned and the old skb data released.
3484 static inline int skb_linearize_cow(struct sk_buff *skb)
3486 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3487 __skb_linearize(skb) : 0;
3490 static __always_inline void
3491 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3494 if (skb->ip_summed == CHECKSUM_COMPLETE)
3495 skb->csum = csum_block_sub(skb->csum,
3496 csum_partial(start, len, 0), off);
3497 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3498 skb_checksum_start_offset(skb) < 0)
3499 skb->ip_summed = CHECKSUM_NONE;
3503 * skb_postpull_rcsum - update checksum for received skb after pull
3504 * @skb: buffer to update
3505 * @start: start of data before pull
3506 * @len: length of data pulled
3508 * After doing a pull on a received packet, you need to call this to
3509 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3510 * CHECKSUM_NONE so that it can be recomputed from scratch.
3512 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3513 const void *start, unsigned int len)
3515 __skb_postpull_rcsum(skb, start, len, 0);
3518 static __always_inline void
3519 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3522 if (skb->ip_summed == CHECKSUM_COMPLETE)
3523 skb->csum = csum_block_add(skb->csum,
3524 csum_partial(start, len, 0), off);
3528 * skb_postpush_rcsum - update checksum for received skb after push
3529 * @skb: buffer to update
3530 * @start: start of data after push
3531 * @len: length of data pushed
3533 * After doing a push on a received packet, you need to call this to
3534 * update the CHECKSUM_COMPLETE checksum.
3536 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3537 const void *start, unsigned int len)
3539 __skb_postpush_rcsum(skb, start, len, 0);
3542 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3545 * skb_push_rcsum - push skb and update receive checksum
3546 * @skb: buffer to update
3547 * @len: length of data pulled
3549 * This function performs an skb_push on the packet and updates
3550 * the CHECKSUM_COMPLETE checksum. It should be used on
3551 * receive path processing instead of skb_push unless you know
3552 * that the checksum difference is zero (e.g., a valid IP header)
3553 * or you are setting ip_summed to CHECKSUM_NONE.
3555 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3558 skb_postpush_rcsum(skb, skb->data, len);
3562 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3564 * pskb_trim_rcsum - trim received skb and update checksum
3565 * @skb: buffer to trim
3568 * This is exactly the same as pskb_trim except that it ensures the
3569 * checksum of received packets are still valid after the operation.
3570 * It can change skb pointers.
3573 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3575 if (likely(len >= skb->len))
3577 return pskb_trim_rcsum_slow(skb, len);
3580 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3582 if (skb->ip_summed == CHECKSUM_COMPLETE)
3583 skb->ip_summed = CHECKSUM_NONE;
3584 __skb_trim(skb, len);
3588 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3590 if (skb->ip_summed == CHECKSUM_COMPLETE)
3591 skb->ip_summed = CHECKSUM_NONE;
3592 return __skb_grow(skb, len);
3595 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3596 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3597 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3598 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3599 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3601 #define skb_queue_walk(queue, skb) \
3602 for (skb = (queue)->next; \
3603 skb != (struct sk_buff *)(queue); \
3606 #define skb_queue_walk_safe(queue, skb, tmp) \
3607 for (skb = (queue)->next, tmp = skb->next; \
3608 skb != (struct sk_buff *)(queue); \
3609 skb = tmp, tmp = skb->next)
3611 #define skb_queue_walk_from(queue, skb) \
3612 for (; skb != (struct sk_buff *)(queue); \
3615 #define skb_rbtree_walk(skb, root) \
3616 for (skb = skb_rb_first(root); skb != NULL; \
3617 skb = skb_rb_next(skb))
3619 #define skb_rbtree_walk_from(skb) \
3620 for (; skb != NULL; \
3621 skb = skb_rb_next(skb))
3623 #define skb_rbtree_walk_from_safe(skb, tmp) \
3624 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3627 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3628 for (tmp = skb->next; \
3629 skb != (struct sk_buff *)(queue); \
3630 skb = tmp, tmp = skb->next)
3632 #define skb_queue_reverse_walk(queue, skb) \
3633 for (skb = (queue)->prev; \
3634 skb != (struct sk_buff *)(queue); \
3637 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3638 for (skb = (queue)->prev, tmp = skb->prev; \
3639 skb != (struct sk_buff *)(queue); \
3640 skb = tmp, tmp = skb->prev)
3642 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3643 for (tmp = skb->prev; \
3644 skb != (struct sk_buff *)(queue); \
3645 skb = tmp, tmp = skb->prev)
3647 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3649 return skb_shinfo(skb)->frag_list != NULL;
3652 static inline void skb_frag_list_init(struct sk_buff *skb)
3654 skb_shinfo(skb)->frag_list = NULL;
3657 #define skb_walk_frags(skb, iter) \
3658 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3661 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3662 int *err, long *timeo_p,
3663 const struct sk_buff *skb);
3664 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3665 struct sk_buff_head *queue,
3668 struct sk_buff **last);
3669 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3670 struct sk_buff_head *queue,
3671 unsigned int flags, int *off, int *err,
3672 struct sk_buff **last);
3673 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3674 struct sk_buff_head *sk_queue,
3675 unsigned int flags, int *off, int *err);
3676 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3678 __poll_t datagram_poll(struct file *file, struct socket *sock,
3679 struct poll_table_struct *wait);
3680 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3681 struct iov_iter *to, int size);
3682 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3683 struct msghdr *msg, int size)
3685 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3687 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3688 struct msghdr *msg);
3689 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3690 struct iov_iter *to, int len,
3691 struct ahash_request *hash);
3692 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3693 struct iov_iter *from, int len);
3694 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3695 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3696 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3697 static inline void skb_free_datagram_locked(struct sock *sk,
3698 struct sk_buff *skb)
3700 __skb_free_datagram_locked(sk, skb, 0);
3702 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3703 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3704 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3705 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3707 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3708 struct pipe_inode_info *pipe, unsigned int len,
3709 unsigned int flags);
3710 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3712 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3713 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3714 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3715 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3717 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3718 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3719 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3720 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3721 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3722 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3723 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3724 unsigned int offset);
3725 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3726 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3727 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3728 int skb_vlan_pop(struct sk_buff *skb);
3729 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3730 int skb_eth_pop(struct sk_buff *skb);
3731 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3732 const unsigned char *src);
3733 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3734 int mac_len, bool ethernet);
3735 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3737 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3738 int skb_mpls_dec_ttl(struct sk_buff *skb);
3739 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3742 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3744 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3747 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3749 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3752 struct skb_checksum_ops {
3753 __wsum (*update)(const void *mem, int len, __wsum wsum);
3754 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3757 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3759 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3760 __wsum csum, const struct skb_checksum_ops *ops);
3761 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3764 static inline void * __must_check
3765 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3766 const void *data, int hlen, void *buffer)
3768 if (likely(hlen - offset >= len))
3769 return (void *)data + offset;
3771 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3777 static inline void * __must_check
3778 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3780 return __skb_header_pointer(skb, offset, len, skb->data,
3781 skb_headlen(skb), buffer);
3785 * skb_needs_linearize - check if we need to linearize a given skb
3786 * depending on the given device features.
3787 * @skb: socket buffer to check
3788 * @features: net device features
3790 * Returns true if either:
3791 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3792 * 2. skb is fragmented and the device does not support SG.
3794 static inline bool skb_needs_linearize(struct sk_buff *skb,
3795 netdev_features_t features)
3797 return skb_is_nonlinear(skb) &&
3798 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3799 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3802 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3804 const unsigned int len)
3806 memcpy(to, skb->data, len);
3809 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3810 const int offset, void *to,
3811 const unsigned int len)
3813 memcpy(to, skb->data + offset, len);
3816 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3818 const unsigned int len)
3820 memcpy(skb->data, from, len);
3823 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3826 const unsigned int len)
3828 memcpy(skb->data + offset, from, len);
3831 void skb_init(void);
3833 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3839 * skb_get_timestamp - get timestamp from a skb
3840 * @skb: skb to get stamp from
3841 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3843 * Timestamps are stored in the skb as offsets to a base timestamp.
3844 * This function converts the offset back to a struct timeval and stores
3847 static inline void skb_get_timestamp(const struct sk_buff *skb,
3848 struct __kernel_old_timeval *stamp)
3850 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3853 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3854 struct __kernel_sock_timeval *stamp)
3856 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3858 stamp->tv_sec = ts.tv_sec;
3859 stamp->tv_usec = ts.tv_nsec / 1000;
3862 static inline void skb_get_timestampns(const struct sk_buff *skb,
3863 struct __kernel_old_timespec *stamp)
3865 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3867 stamp->tv_sec = ts.tv_sec;
3868 stamp->tv_nsec = ts.tv_nsec;
3871 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3872 struct __kernel_timespec *stamp)
3874 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3876 stamp->tv_sec = ts.tv_sec;
3877 stamp->tv_nsec = ts.tv_nsec;
3880 static inline void __net_timestamp(struct sk_buff *skb)
3882 skb->tstamp = ktime_get_real();
3885 static inline ktime_t net_timedelta(ktime_t t)
3887 return ktime_sub(ktime_get_real(), t);
3890 static inline ktime_t net_invalid_timestamp(void)
3895 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3897 return skb_shinfo(skb)->meta_len;
3900 static inline void *skb_metadata_end(const struct sk_buff *skb)
3902 return skb_mac_header(skb);
3905 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3906 const struct sk_buff *skb_b,
3909 const void *a = skb_metadata_end(skb_a);
3910 const void *b = skb_metadata_end(skb_b);
3911 /* Using more efficient varaiant than plain call to memcmp(). */
3912 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3916 #define __it(x, op) (x -= sizeof(u##op))
3917 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3918 case 32: diffs |= __it_diff(a, b, 64);
3920 case 24: diffs |= __it_diff(a, b, 64);
3922 case 16: diffs |= __it_diff(a, b, 64);
3924 case 8: diffs |= __it_diff(a, b, 64);
3926 case 28: diffs |= __it_diff(a, b, 64);
3928 case 20: diffs |= __it_diff(a, b, 64);
3930 case 12: diffs |= __it_diff(a, b, 64);
3932 case 4: diffs |= __it_diff(a, b, 32);
3937 return memcmp(a - meta_len, b - meta_len, meta_len);
3941 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3942 const struct sk_buff *skb_b)
3944 u8 len_a = skb_metadata_len(skb_a);
3945 u8 len_b = skb_metadata_len(skb_b);
3947 if (!(len_a | len_b))
3950 return len_a != len_b ?
3951 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3954 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3956 skb_shinfo(skb)->meta_len = meta_len;
3959 static inline void skb_metadata_clear(struct sk_buff *skb)
3961 skb_metadata_set(skb, 0);
3964 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3966 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3968 void skb_clone_tx_timestamp(struct sk_buff *skb);
3969 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3971 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3973 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3977 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3982 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3985 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3987 * PHY drivers may accept clones of transmitted packets for
3988 * timestamping via their phy_driver.txtstamp method. These drivers
3989 * must call this function to return the skb back to the stack with a
3992 * @skb: clone of the original outgoing packet
3993 * @hwtstamps: hardware time stamps
3996 void skb_complete_tx_timestamp(struct sk_buff *skb,
3997 struct skb_shared_hwtstamps *hwtstamps);
3999 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4000 struct skb_shared_hwtstamps *hwtstamps,
4001 struct sock *sk, int tstype);
4004 * skb_tstamp_tx - queue clone of skb with send time stamps
4005 * @orig_skb: the original outgoing packet
4006 * @hwtstamps: hardware time stamps, may be NULL if not available
4008 * If the skb has a socket associated, then this function clones the
4009 * skb (thus sharing the actual data and optional structures), stores
4010 * the optional hardware time stamping information (if non NULL) or
4011 * generates a software time stamp (otherwise), then queues the clone
4012 * to the error queue of the socket. Errors are silently ignored.
4014 void skb_tstamp_tx(struct sk_buff *orig_skb,
4015 struct skb_shared_hwtstamps *hwtstamps);
4018 * skb_tx_timestamp() - Driver hook for transmit timestamping
4020 * Ethernet MAC Drivers should call this function in their hard_xmit()
4021 * function immediately before giving the sk_buff to the MAC hardware.
4023 * Specifically, one should make absolutely sure that this function is
4024 * called before TX completion of this packet can trigger. Otherwise
4025 * the packet could potentially already be freed.
4027 * @skb: A socket buffer.
4029 static inline void skb_tx_timestamp(struct sk_buff *skb)
4031 skb_clone_tx_timestamp(skb);
4032 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4033 skb_tstamp_tx(skb, NULL);
4037 * skb_complete_wifi_ack - deliver skb with wifi status
4039 * @skb: the original outgoing packet
4040 * @acked: ack status
4043 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4045 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4046 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4048 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4050 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4052 (skb->ip_summed == CHECKSUM_PARTIAL &&
4053 skb_checksum_start_offset(skb) >= 0));
4057 * skb_checksum_complete - Calculate checksum of an entire packet
4058 * @skb: packet to process
4060 * This function calculates the checksum over the entire packet plus
4061 * the value of skb->csum. The latter can be used to supply the
4062 * checksum of a pseudo header as used by TCP/UDP. It returns the
4065 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4066 * this function can be used to verify that checksum on received
4067 * packets. In that case the function should return zero if the
4068 * checksum is correct. In particular, this function will return zero
4069 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4070 * hardware has already verified the correctness of the checksum.
4072 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4074 return skb_csum_unnecessary(skb) ?
4075 0 : __skb_checksum_complete(skb);
4078 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4080 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4081 if (skb->csum_level == 0)
4082 skb->ip_summed = CHECKSUM_NONE;
4088 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4090 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4091 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4093 } else if (skb->ip_summed == CHECKSUM_NONE) {
4094 skb->ip_summed = CHECKSUM_UNNECESSARY;
4095 skb->csum_level = 0;
4099 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4101 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4102 skb->ip_summed = CHECKSUM_NONE;
4103 skb->csum_level = 0;
4107 /* Check if we need to perform checksum complete validation.
4109 * Returns true if checksum complete is needed, false otherwise
4110 * (either checksum is unnecessary or zero checksum is allowed).
4112 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4116 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4117 skb->csum_valid = 1;
4118 __skb_decr_checksum_unnecessary(skb);
4125 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4128 #define CHECKSUM_BREAK 76
4130 /* Unset checksum-complete
4132 * Unset checksum complete can be done when packet is being modified
4133 * (uncompressed for instance) and checksum-complete value is
4136 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4138 if (skb->ip_summed == CHECKSUM_COMPLETE)
4139 skb->ip_summed = CHECKSUM_NONE;
4142 /* Validate (init) checksum based on checksum complete.
4145 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4146 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4147 * checksum is stored in skb->csum for use in __skb_checksum_complete
4148 * non-zero: value of invalid checksum
4151 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4155 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4156 if (!csum_fold(csum_add(psum, skb->csum))) {
4157 skb->csum_valid = 1;
4164 if (complete || skb->len <= CHECKSUM_BREAK) {
4167 csum = __skb_checksum_complete(skb);
4168 skb->csum_valid = !csum;
4175 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4180 /* Perform checksum validate (init). Note that this is a macro since we only
4181 * want to calculate the pseudo header which is an input function if necessary.
4182 * First we try to validate without any computation (checksum unnecessary) and
4183 * then calculate based on checksum complete calling the function to compute
4187 * 0: checksum is validated or try to in skb_checksum_complete
4188 * non-zero: value of invalid checksum
4190 #define __skb_checksum_validate(skb, proto, complete, \
4191 zero_okay, check, compute_pseudo) \
4193 __sum16 __ret = 0; \
4194 skb->csum_valid = 0; \
4195 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4196 __ret = __skb_checksum_validate_complete(skb, \
4197 complete, compute_pseudo(skb, proto)); \
4201 #define skb_checksum_init(skb, proto, compute_pseudo) \
4202 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4204 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4205 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4207 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4208 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4210 #define skb_checksum_validate_zero_check(skb, proto, check, \
4212 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4214 #define skb_checksum_simple_validate(skb) \
4215 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4217 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4219 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4222 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4224 skb->csum = ~pseudo;
4225 skb->ip_summed = CHECKSUM_COMPLETE;
4228 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4230 if (__skb_checksum_convert_check(skb)) \
4231 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4234 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4235 u16 start, u16 offset)
4237 skb->ip_summed = CHECKSUM_PARTIAL;
4238 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4239 skb->csum_offset = offset - start;
4242 /* Update skbuf and packet to reflect the remote checksum offload operation.
4243 * When called, ptr indicates the starting point for skb->csum when
4244 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4245 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4247 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4248 int start, int offset, bool nopartial)
4253 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4257 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4258 __skb_checksum_complete(skb);
4259 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4262 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4264 /* Adjust skb->csum since we changed the packet */
4265 skb->csum = csum_add(skb->csum, delta);
4268 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4270 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4271 return (void *)(skb->_nfct & NFCT_PTRMASK);
4277 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4279 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4286 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4288 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4289 skb->slow_gro |= !!nfct;
4294 #ifdef CONFIG_SKB_EXTENSIONS
4296 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4302 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4305 #if IS_ENABLED(CONFIG_MPTCP)
4308 SKB_EXT_NUM, /* must be last */
4312 * struct skb_ext - sk_buff extensions
4313 * @refcnt: 1 on allocation, deallocated on 0
4314 * @offset: offset to add to @data to obtain extension address
4315 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4316 * @data: start of extension data, variable sized
4318 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4319 * to use 'u8' types while allowing up to 2kb worth of extension data.
4323 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4324 u8 chunks; /* same */
4325 char data[] __aligned(8);
4328 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4329 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4330 struct skb_ext *ext);
4331 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4332 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4333 void __skb_ext_put(struct skb_ext *ext);
4335 static inline void skb_ext_put(struct sk_buff *skb)
4337 if (skb->active_extensions)
4338 __skb_ext_put(skb->extensions);
4341 static inline void __skb_ext_copy(struct sk_buff *dst,
4342 const struct sk_buff *src)
4344 dst->active_extensions = src->active_extensions;
4346 if (src->active_extensions) {
4347 struct skb_ext *ext = src->extensions;
4349 refcount_inc(&ext->refcnt);
4350 dst->extensions = ext;
4354 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4357 __skb_ext_copy(dst, src);
4360 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4362 return !!ext->offset[i];
4365 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4367 return skb->active_extensions & (1 << id);
4370 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4372 if (skb_ext_exist(skb, id))
4373 __skb_ext_del(skb, id);
4376 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4378 if (skb_ext_exist(skb, id)) {
4379 struct skb_ext *ext = skb->extensions;
4381 return (void *)ext + (ext->offset[id] << 3);
4387 static inline void skb_ext_reset(struct sk_buff *skb)
4389 if (unlikely(skb->active_extensions)) {
4390 __skb_ext_put(skb->extensions);
4391 skb->active_extensions = 0;
4395 static inline bool skb_has_extensions(struct sk_buff *skb)
4397 return unlikely(skb->active_extensions);
4400 static inline void skb_ext_put(struct sk_buff *skb) {}
4401 static inline void skb_ext_reset(struct sk_buff *skb) {}
4402 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4403 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4404 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4405 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4406 #endif /* CONFIG_SKB_EXTENSIONS */
4408 static inline void nf_reset_ct(struct sk_buff *skb)
4410 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4411 nf_conntrack_put(skb_nfct(skb));
4416 static inline void nf_reset_trace(struct sk_buff *skb)
4418 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4423 static inline void ipvs_reset(struct sk_buff *skb)
4425 #if IS_ENABLED(CONFIG_IP_VS)
4426 skb->ipvs_property = 0;
4430 /* Note: This doesn't put any conntrack info in dst. */
4431 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4434 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4435 dst->_nfct = src->_nfct;
4436 nf_conntrack_get(skb_nfct(src));
4438 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4440 dst->nf_trace = src->nf_trace;
4444 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4446 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4447 nf_conntrack_put(skb_nfct(dst));
4449 dst->slow_gro = src->slow_gro;
4450 __nf_copy(dst, src, true);
4453 #ifdef CONFIG_NETWORK_SECMARK
4454 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4456 to->secmark = from->secmark;
4459 static inline void skb_init_secmark(struct sk_buff *skb)
4464 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4467 static inline void skb_init_secmark(struct sk_buff *skb)
4471 static inline int secpath_exists(const struct sk_buff *skb)
4474 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4480 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4482 return !skb->destructor &&
4483 !secpath_exists(skb) &&
4485 !skb->_skb_refdst &&
4486 !skb_has_frag_list(skb);
4489 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4491 skb->queue_mapping = queue_mapping;
4494 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4496 return skb->queue_mapping;
4499 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4501 to->queue_mapping = from->queue_mapping;
4504 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4506 skb->queue_mapping = rx_queue + 1;
4509 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4511 return skb->queue_mapping - 1;
4514 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4516 return skb->queue_mapping != 0;
4519 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4521 skb->dst_pending_confirm = val;
4524 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4526 return skb->dst_pending_confirm != 0;
4529 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4532 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4538 /* Keeps track of mac header offset relative to skb->head.
4539 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4540 * For non-tunnel skb it points to skb_mac_header() and for
4541 * tunnel skb it points to outer mac header.
4542 * Keeps track of level of encapsulation of network headers.
4553 #define SKB_GSO_CB_OFFSET 32
4554 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4556 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4558 return (skb_mac_header(inner_skb) - inner_skb->head) -
4559 SKB_GSO_CB(inner_skb)->mac_offset;
4562 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4564 int new_headroom, headroom;
4567 headroom = skb_headroom(skb);
4568 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4572 new_headroom = skb_headroom(skb);
4573 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4577 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4579 /* Do not update partial checksums if remote checksum is enabled. */
4580 if (skb->remcsum_offload)
4583 SKB_GSO_CB(skb)->csum = res;
4584 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4587 /* Compute the checksum for a gso segment. First compute the checksum value
4588 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4589 * then add in skb->csum (checksum from csum_start to end of packet).
4590 * skb->csum and csum_start are then updated to reflect the checksum of the
4591 * resultant packet starting from the transport header-- the resultant checksum
4592 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4595 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4597 unsigned char *csum_start = skb_transport_header(skb);
4598 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4599 __wsum partial = SKB_GSO_CB(skb)->csum;
4601 SKB_GSO_CB(skb)->csum = res;
4602 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4604 return csum_fold(csum_partial(csum_start, plen, partial));
4607 static inline bool skb_is_gso(const struct sk_buff *skb)
4609 return skb_shinfo(skb)->gso_size;
4612 /* Note: Should be called only if skb_is_gso(skb) is true */
4613 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4615 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4618 /* Note: Should be called only if skb_is_gso(skb) is true */
4619 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4621 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4624 /* Note: Should be called only if skb_is_gso(skb) is true */
4625 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4627 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4630 static inline void skb_gso_reset(struct sk_buff *skb)
4632 skb_shinfo(skb)->gso_size = 0;
4633 skb_shinfo(skb)->gso_segs = 0;
4634 skb_shinfo(skb)->gso_type = 0;
4637 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4640 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4642 shinfo->gso_size += increment;
4645 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4648 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4650 shinfo->gso_size -= decrement;
4653 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4655 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4657 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4658 * wanted then gso_type will be set. */
4659 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4661 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4662 unlikely(shinfo->gso_type == 0)) {
4663 __skb_warn_lro_forwarding(skb);
4669 static inline void skb_forward_csum(struct sk_buff *skb)
4671 /* Unfortunately we don't support this one. Any brave souls? */
4672 if (skb->ip_summed == CHECKSUM_COMPLETE)
4673 skb->ip_summed = CHECKSUM_NONE;
4677 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4678 * @skb: skb to check
4680 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4681 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4682 * use this helper, to document places where we make this assertion.
4684 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4687 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4691 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4693 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4694 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4695 unsigned int transport_len,
4696 __sum16(*skb_chkf)(struct sk_buff *skb));
4699 * skb_head_is_locked - Determine if the skb->head is locked down
4700 * @skb: skb to check
4702 * The head on skbs build around a head frag can be removed if they are
4703 * not cloned. This function returns true if the skb head is locked down
4704 * due to either being allocated via kmalloc, or by being a clone with
4705 * multiple references to the head.
4707 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4709 return !skb->head_frag || skb_cloned(skb);
4712 /* Local Checksum Offload.
4713 * Compute outer checksum based on the assumption that the
4714 * inner checksum will be offloaded later.
4715 * See Documentation/networking/checksum-offloads.rst for
4716 * explanation of how this works.
4717 * Fill in outer checksum adjustment (e.g. with sum of outer
4718 * pseudo-header) before calling.
4719 * Also ensure that inner checksum is in linear data area.
4721 static inline __wsum lco_csum(struct sk_buff *skb)
4723 unsigned char *csum_start = skb_checksum_start(skb);
4724 unsigned char *l4_hdr = skb_transport_header(skb);
4727 /* Start with complement of inner checksum adjustment */
4728 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4731 /* Add in checksum of our headers (incl. outer checksum
4732 * adjustment filled in by caller) and return result.
4734 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4737 static inline bool skb_is_redirected(const struct sk_buff *skb)
4739 return skb->redirected;
4742 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4744 skb->redirected = 1;
4745 #ifdef CONFIG_NET_REDIRECT
4746 skb->from_ingress = from_ingress;
4747 if (skb->from_ingress)
4752 static inline void skb_reset_redirect(struct sk_buff *skb)
4754 skb->redirected = 0;
4757 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4759 return skb->csum_not_inet;
4762 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4763 const u64 kcov_handle)
4766 skb->kcov_handle = kcov_handle;
4770 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4773 return skb->kcov_handle;
4779 #ifdef CONFIG_PAGE_POOL
4780 static inline void skb_mark_for_recycle(struct sk_buff *skb)
4782 skb->pp_recycle = 1;
4786 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4788 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4790 return page_pool_return_skb_page(virt_to_page(data));
4793 #endif /* __KERNEL__ */
4794 #endif /* _LINUX_SKBUFF_H */