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,
262 u8 sabotage_in_done:1;
264 struct net_device *physindev;
266 /* always valid & non-NULL from FORWARD on, for physdev match */
267 struct net_device *physoutdev;
269 /* prerouting: detect dnat in orig/reply direction */
271 struct in6_addr ipv6_daddr;
273 /* after prerouting + nat detected: store original source
274 * mac since neigh resolution overwrites it, only used while
275 * skb is out in neigh layer.
277 char neigh_header[8];
282 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
283 /* Chain in tc_skb_ext will be used to share the tc chain with
284 * ovs recirc_id. It will be set to the current chain by tc
285 * and read by ovs to recirc_id.
297 struct sk_buff_head {
298 /* These two members must be first. */
299 struct sk_buff *next;
300 struct sk_buff *prev;
308 /* The reason of skb drop, which is used in kfree_skb_reason().
309 * en...maybe they should be splited by group?
311 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
312 * used to translate the reason to string.
314 enum skb_drop_reason {
315 SKB_DROP_REASON_NOT_SPECIFIED, /* drop reason is not specified */
316 SKB_DROP_REASON_NO_SOCKET, /* socket not found */
317 SKB_DROP_REASON_PKT_TOO_SMALL, /* packet size is too small */
318 SKB_DROP_REASON_TCP_CSUM, /* TCP checksum error */
319 SKB_DROP_REASON_SOCKET_FILTER, /* dropped by socket filter */
320 SKB_DROP_REASON_UDP_CSUM, /* UDP checksum error */
321 SKB_DROP_REASON_NETFILTER_DROP, /* dropped by netfilter */
322 SKB_DROP_REASON_OTHERHOST, /* packet don't belong to current
323 * host (interface is in promisc
326 SKB_DROP_REASON_IP_CSUM, /* IP checksum error */
327 SKB_DROP_REASON_IP_INHDR, /* there is something wrong with
329 * IPSTATS_MIB_INHDRERRORS)
331 SKB_DROP_REASON_IP_RPFILTER, /* IP rpfilter validate failed.
332 * see the document for rp_filter
333 * in ip-sysctl.rst for more
336 SKB_DROP_REASON_UNICAST_IN_L2_MULTICAST, /* destination address of L2
337 * is multicast, but L3 is
343 /* To allow 64K frame to be packed as single skb without frag_list we
344 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
345 * buffers which do not start on a page boundary.
347 * Since GRO uses frags we allocate at least 16 regardless of page
350 #if (65536/PAGE_SIZE + 1) < 16
351 #define MAX_SKB_FRAGS 16UL
353 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
355 extern int sysctl_max_skb_frags;
357 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
358 * segment using its current segmentation instead.
360 #define GSO_BY_FRAGS 0xFFFF
362 typedef struct bio_vec skb_frag_t;
365 * skb_frag_size() - Returns the size of a skb fragment
366 * @frag: skb fragment
368 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
374 * skb_frag_size_set() - Sets the size of a skb fragment
375 * @frag: skb fragment
376 * @size: size of fragment
378 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
384 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
385 * @frag: skb fragment
386 * @delta: value to add
388 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
390 frag->bv_len += delta;
394 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
395 * @frag: skb fragment
396 * @delta: value to subtract
398 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
400 frag->bv_len -= delta;
404 * skb_frag_must_loop - Test if %p is a high memory page
405 * @p: fragment's page
407 static inline bool skb_frag_must_loop(struct page *p)
409 #if defined(CONFIG_HIGHMEM)
410 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
417 * skb_frag_foreach_page - loop over pages in a fragment
419 * @f: skb frag to operate on
420 * @f_off: offset from start of f->bv_page
421 * @f_len: length from f_off to loop over
422 * @p: (temp var) current page
423 * @p_off: (temp var) offset from start of current page,
424 * non-zero only on first page.
425 * @p_len: (temp var) length in current page,
426 * < PAGE_SIZE only on first and last page.
427 * @copied: (temp var) length so far, excluding current p_len.
429 * A fragment can hold a compound page, in which case per-page
430 * operations, notably kmap_atomic, must be called for each
433 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
434 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
435 p_off = (f_off) & (PAGE_SIZE - 1), \
436 p_len = skb_frag_must_loop(p) ? \
437 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
440 copied += p_len, p++, p_off = 0, \
441 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
443 #define HAVE_HW_TIME_STAMP
446 * struct skb_shared_hwtstamps - hardware time stamps
447 * @hwtstamp: hardware time stamp transformed into duration
448 * since arbitrary point in time
450 * Software time stamps generated by ktime_get_real() are stored in
453 * hwtstamps can only be compared against other hwtstamps from
456 * This structure is attached to packets as part of the
457 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
459 struct skb_shared_hwtstamps {
463 /* Definitions for tx_flags in struct skb_shared_info */
465 /* generate hardware time stamp */
466 SKBTX_HW_TSTAMP = 1 << 0,
468 /* generate software time stamp when queueing packet to NIC */
469 SKBTX_SW_TSTAMP = 1 << 1,
471 /* device driver is going to provide hardware time stamp */
472 SKBTX_IN_PROGRESS = 1 << 2,
474 /* generate wifi status information (where possible) */
475 SKBTX_WIFI_STATUS = 1 << 4,
477 /* generate software time stamp when entering packet scheduling */
478 SKBTX_SCHED_TSTAMP = 1 << 6,
481 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
483 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
485 /* Definitions for flags in struct skb_shared_info */
487 /* use zcopy routines */
488 SKBFL_ZEROCOPY_ENABLE = BIT(0),
490 /* This indicates at least one fragment might be overwritten
491 * (as in vmsplice(), sendfile() ...)
492 * If we need to compute a TX checksum, we'll need to copy
493 * all frags to avoid possible bad checksum
495 SKBFL_SHARED_FRAG = BIT(1),
498 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
501 * The callback notifies userspace to release buffers when skb DMA is done in
502 * lower device, the skb last reference should be 0 when calling this.
503 * The zerocopy_success argument is true if zero copy transmit occurred,
504 * false on data copy or out of memory error caused by data copy attempt.
505 * The ctx field is used to track device context.
506 * The desc field is used to track userspace buffer index.
509 void (*callback)(struct sk_buff *, struct ubuf_info *,
510 bool zerocopy_success);
527 struct user_struct *user;
532 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
534 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
535 void mm_unaccount_pinned_pages(struct mmpin *mmp);
537 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
538 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
539 struct ubuf_info *uarg);
541 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
543 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
546 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
547 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
548 struct msghdr *msg, int len,
549 struct ubuf_info *uarg);
551 /* This data is invariant across clones and lives at
552 * the end of the header data, ie. at skb->end.
554 struct skb_shared_info {
559 unsigned short gso_size;
560 /* Warning: this field is not always filled in (UFO)! */
561 unsigned short gso_segs;
562 struct sk_buff *frag_list;
563 struct skb_shared_hwtstamps hwtstamps;
564 unsigned int gso_type;
568 * Warning : all fields before dataref are cleared in __alloc_skb()
572 /* Intermediate layers must ensure that destructor_arg
573 * remains valid until skb destructor */
574 void * destructor_arg;
576 /* must be last field, see pskb_expand_head() */
577 skb_frag_t frags[MAX_SKB_FRAGS];
580 /* We divide dataref into two halves. The higher 16 bits hold references
581 * to the payload part of skb->data. The lower 16 bits hold references to
582 * the entire skb->data. A clone of a headerless skb holds the length of
583 * the header in skb->hdr_len.
585 * All users must obey the rule that the skb->data reference count must be
586 * greater than or equal to the payload reference count.
588 * Holding a reference to the payload part means that the user does not
589 * care about modifications to the header part of skb->data.
591 #define SKB_DATAREF_SHIFT 16
592 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
596 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
597 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
598 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
602 SKB_GSO_TCPV4 = 1 << 0,
604 /* This indicates the skb is from an untrusted source. */
605 SKB_GSO_DODGY = 1 << 1,
607 /* This indicates the tcp segment has CWR set. */
608 SKB_GSO_TCP_ECN = 1 << 2,
610 SKB_GSO_TCP_FIXEDID = 1 << 3,
612 SKB_GSO_TCPV6 = 1 << 4,
614 SKB_GSO_FCOE = 1 << 5,
616 SKB_GSO_GRE = 1 << 6,
618 SKB_GSO_GRE_CSUM = 1 << 7,
620 SKB_GSO_IPXIP4 = 1 << 8,
622 SKB_GSO_IPXIP6 = 1 << 9,
624 SKB_GSO_UDP_TUNNEL = 1 << 10,
626 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
628 SKB_GSO_PARTIAL = 1 << 12,
630 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
632 SKB_GSO_SCTP = 1 << 14,
634 SKB_GSO_ESP = 1 << 15,
636 SKB_GSO_UDP = 1 << 16,
638 SKB_GSO_UDP_L4 = 1 << 17,
640 SKB_GSO_FRAGLIST = 1 << 18,
643 #if BITS_PER_LONG > 32
644 #define NET_SKBUFF_DATA_USES_OFFSET 1
647 #ifdef NET_SKBUFF_DATA_USES_OFFSET
648 typedef unsigned int sk_buff_data_t;
650 typedef unsigned char *sk_buff_data_t;
654 * struct sk_buff - socket buffer
655 * @next: Next buffer in list
656 * @prev: Previous buffer in list
657 * @tstamp: Time we arrived/left
658 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
659 * for retransmit timer
660 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
662 * @sk: Socket we are owned by
663 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
664 * fragmentation management
665 * @dev: Device we arrived on/are leaving by
666 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
667 * @cb: Control buffer. Free for use by every layer. Put private vars here
668 * @_skb_refdst: destination entry (with norefcount bit)
669 * @sp: the security path, used for xfrm
670 * @len: Length of actual data
671 * @data_len: Data length
672 * @mac_len: Length of link layer header
673 * @hdr_len: writable header length of cloned skb
674 * @csum: Checksum (must include start/offset pair)
675 * @csum_start: Offset from skb->head where checksumming should start
676 * @csum_offset: Offset from csum_start where checksum should be stored
677 * @priority: Packet queueing priority
678 * @ignore_df: allow local fragmentation
679 * @cloned: Head may be cloned (check refcnt to be sure)
680 * @ip_summed: Driver fed us an IP checksum
681 * @nohdr: Payload reference only, must not modify header
682 * @pkt_type: Packet class
683 * @fclone: skbuff clone status
684 * @ipvs_property: skbuff is owned by ipvs
685 * @inner_protocol_type: whether the inner protocol is
686 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
687 * @remcsum_offload: remote checksum offload is enabled
688 * @offload_fwd_mark: Packet was L2-forwarded in hardware
689 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
690 * @tc_skip_classify: do not classify packet. set by IFB device
691 * @tc_at_ingress: used within tc_classify to distinguish in/egress
692 * @redirected: packet was redirected by packet classifier
693 * @from_ingress: packet was redirected from the ingress path
694 * @peeked: this packet has been seen already, so stats have been
695 * done for it, don't do them again
696 * @nf_trace: netfilter packet trace flag
697 * @protocol: Packet protocol from driver
698 * @destructor: Destruct function
699 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
700 * @_sk_redir: socket redirection information for skmsg
701 * @_nfct: Associated connection, if any (with nfctinfo bits)
702 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
703 * @skb_iif: ifindex of device we arrived on
704 * @tc_index: Traffic control index
705 * @hash: the packet hash
706 * @queue_mapping: Queue mapping for multiqueue devices
707 * @head_frag: skb was allocated from page fragments,
708 * not allocated by kmalloc() or vmalloc().
709 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
710 * @pp_recycle: mark the packet for recycling instead of freeing (implies
711 * page_pool support on driver)
712 * @active_extensions: active extensions (skb_ext_id types)
713 * @ndisc_nodetype: router type (from link layer)
714 * @ooo_okay: allow the mapping of a socket to a queue to be changed
715 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
717 * @sw_hash: indicates hash was computed in software stack
718 * @wifi_acked_valid: wifi_acked was set
719 * @wifi_acked: whether frame was acked on wifi or not
720 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
721 * @encapsulation: indicates the inner headers in the skbuff are valid
722 * @encap_hdr_csum: software checksum is needed
723 * @csum_valid: checksum is already valid
724 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
725 * @csum_complete_sw: checksum was completed by software
726 * @csum_level: indicates the number of consecutive checksums found in
727 * the packet minus one that have been verified as
728 * CHECKSUM_UNNECESSARY (max 3)
729 * @scm_io_uring: SKB holds io_uring registered files
730 * @dst_pending_confirm: need to confirm neighbour
731 * @decrypted: Decrypted SKB
732 * @slow_gro: state present at GRO time, slower prepare step required
733 * @napi_id: id of the NAPI struct this skb came from
734 * @sender_cpu: (aka @napi_id) source CPU in XPS
735 * @secmark: security marking
736 * @mark: Generic packet mark
737 * @reserved_tailroom: (aka @mark) number of bytes of free space available
738 * at the tail of an sk_buff
739 * @vlan_present: VLAN tag is present
740 * @vlan_proto: vlan encapsulation protocol
741 * @vlan_tci: vlan tag control information
742 * @inner_protocol: Protocol (encapsulation)
743 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
744 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
745 * @inner_transport_header: Inner transport layer header (encapsulation)
746 * @inner_network_header: Network layer header (encapsulation)
747 * @inner_mac_header: Link layer header (encapsulation)
748 * @transport_header: Transport layer header
749 * @network_header: Network layer header
750 * @mac_header: Link layer header
751 * @kcov_handle: KCOV remote handle for remote coverage collection
752 * @tail: Tail pointer
754 * @head: Head of buffer
755 * @data: Data head pointer
756 * @truesize: Buffer size
757 * @users: User count - see {datagram,tcp}.c
758 * @extensions: allocated extensions, valid if active_extensions is nonzero
764 /* These two members must be first. */
765 struct sk_buff *next;
766 struct sk_buff *prev;
769 struct net_device *dev;
770 /* Some protocols might use this space to store information,
771 * while device pointer would be NULL.
772 * UDP receive path is one user.
774 unsigned long dev_scratch;
777 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
778 struct list_head list;
783 int ip_defrag_offset;
788 u64 skb_mstamp_ns; /* earliest departure time */
791 * This is the control buffer. It is free to use for every
792 * layer. Please put your private variables there. If you
793 * want to keep them across layers you have to do a skb_clone()
794 * first. This is owned by whoever has the skb queued ATM.
796 char cb[48] __aligned(8);
800 unsigned long _skb_refdst;
801 void (*destructor)(struct sk_buff *skb);
803 struct list_head tcp_tsorted_anchor;
804 #ifdef CONFIG_NET_SOCK_MSG
805 unsigned long _sk_redir;
809 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
817 /* Following fields are _not_ copied in __copy_skb_header()
818 * Note that queue_mapping is here mostly to fill a hole.
822 /* if you move cloned around you also must adapt those constants */
823 #ifdef __BIG_ENDIAN_BITFIELD
824 #define CLONED_MASK (1 << 7)
826 #define CLONED_MASK 1
828 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
831 __u8 __cloned_offset[0];
839 pp_recycle:1; /* page_pool recycle indicator */
840 #ifdef CONFIG_SKB_EXTENSIONS
841 __u8 active_extensions;
844 /* fields enclosed in headers_start/headers_end are copied
845 * using a single memcpy() in __copy_skb_header()
848 __u32 headers_start[0];
851 /* if you move pkt_type around you also must adapt those constants */
852 #ifdef __BIG_ENDIAN_BITFIELD
853 #define PKT_TYPE_MAX (7 << 5)
855 #define PKT_TYPE_MAX 7
857 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
860 __u8 __pkt_type_offset[0];
870 __u8 wifi_acked_valid:1;
873 /* Indicates the inner headers are valid in the skbuff. */
874 __u8 encapsulation:1;
875 __u8 encap_hdr_csum:1;
878 #ifdef __BIG_ENDIAN_BITFIELD
879 #define PKT_VLAN_PRESENT_BIT 7
881 #define PKT_VLAN_PRESENT_BIT 0
883 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
885 __u8 __pkt_vlan_present_offset[0];
888 __u8 csum_complete_sw:1;
890 __u8 csum_not_inet:1;
891 __u8 dst_pending_confirm:1;
892 #ifdef CONFIG_IPV6_NDISC_NODETYPE
893 __u8 ndisc_nodetype:2;
896 __u8 ipvs_property:1;
897 __u8 inner_protocol_type:1;
898 __u8 remcsum_offload:1;
899 #ifdef CONFIG_NET_SWITCHDEV
900 __u8 offload_fwd_mark:1;
901 __u8 offload_l3_fwd_mark:1;
903 #ifdef CONFIG_NET_CLS_ACT
904 __u8 tc_skip_classify:1;
905 __u8 tc_at_ingress:1;
908 #ifdef CONFIG_NET_REDIRECT
911 #ifdef CONFIG_TLS_DEVICE
917 #ifdef CONFIG_NET_SCHED
918 __u16 tc_index; /* traffic control index */
933 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
935 unsigned int napi_id;
936 unsigned int sender_cpu;
939 #ifdef CONFIG_NETWORK_SECMARK
945 __u32 reserved_tailroom;
949 __be16 inner_protocol;
953 __u16 inner_transport_header;
954 __u16 inner_network_header;
955 __u16 inner_mac_header;
958 __u16 transport_header;
959 __u16 network_header;
967 __u32 headers_end[0];
970 /* These elements must be at the end, see alloc_skb() for details. */
975 unsigned int truesize;
978 #ifdef CONFIG_SKB_EXTENSIONS
979 /* only useable after checking ->active_extensions != 0 */
980 struct skb_ext *extensions;
986 * Handling routines are only of interest to the kernel
989 #define SKB_ALLOC_FCLONE 0x01
990 #define SKB_ALLOC_RX 0x02
991 #define SKB_ALLOC_NAPI 0x04
994 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
997 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
999 return unlikely(skb->pfmemalloc);
1003 * skb might have a dst pointer attached, refcounted or not.
1004 * _skb_refdst low order bit is set if refcount was _not_ taken
1006 #define SKB_DST_NOREF 1UL
1007 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1010 * skb_dst - returns skb dst_entry
1013 * Returns skb dst_entry, regardless of reference taken or not.
1015 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1017 /* If refdst was not refcounted, check we still are in a
1018 * rcu_read_lock section
1020 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1021 !rcu_read_lock_held() &&
1022 !rcu_read_lock_bh_held());
1023 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1027 * skb_dst_set - sets skb dst
1031 * Sets skb dst, assuming a reference was taken on dst and should
1032 * be released by skb_dst_drop()
1034 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1036 skb->slow_gro |= !!dst;
1037 skb->_skb_refdst = (unsigned long)dst;
1041 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1045 * Sets skb dst, assuming a reference was not taken on dst.
1046 * If dst entry is cached, we do not take reference and dst_release
1047 * will be avoided by refdst_drop. If dst entry is not cached, we take
1048 * reference, so that last dst_release can destroy the dst immediately.
1050 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1052 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1053 skb->slow_gro |= !!dst;
1054 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1058 * skb_dst_is_noref - Test if skb dst isn't refcounted
1061 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1063 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1067 * skb_rtable - Returns the skb &rtable
1070 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1072 return (struct rtable *)skb_dst(skb);
1075 /* For mangling skb->pkt_type from user space side from applications
1076 * such as nft, tc, etc, we only allow a conservative subset of
1077 * possible pkt_types to be set.
1079 static inline bool skb_pkt_type_ok(u32 ptype)
1081 return ptype <= PACKET_OTHERHOST;
1085 * skb_napi_id - Returns the skb's NAPI id
1088 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1090 #ifdef CONFIG_NET_RX_BUSY_POLL
1091 return skb->napi_id;
1098 * skb_unref - decrement the skb's reference count
1101 * Returns true if we can free the skb.
1103 static inline bool skb_unref(struct sk_buff *skb)
1107 if (likely(refcount_read(&skb->users) == 1))
1109 else if (likely(!refcount_dec_and_test(&skb->users)))
1115 void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1118 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1119 * @skb: buffer to free
1121 static inline void kfree_skb(struct sk_buff *skb)
1123 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1126 void skb_release_head_state(struct sk_buff *skb);
1127 void kfree_skb_list(struct sk_buff *segs);
1128 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1129 void skb_tx_error(struct sk_buff *skb);
1131 #ifdef CONFIG_TRACEPOINTS
1132 void consume_skb(struct sk_buff *skb);
1134 static inline void consume_skb(struct sk_buff *skb)
1136 return kfree_skb(skb);
1140 void __consume_stateless_skb(struct sk_buff *skb);
1141 void __kfree_skb(struct sk_buff *skb);
1142 extern struct kmem_cache *skbuff_head_cache;
1144 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1145 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1146 bool *fragstolen, int *delta_truesize);
1148 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1150 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1151 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1152 struct sk_buff *build_skb_around(struct sk_buff *skb,
1153 void *data, unsigned int frag_size);
1155 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1158 * alloc_skb - allocate a network buffer
1159 * @size: size to allocate
1160 * @priority: allocation mask
1162 * This function is a convenient wrapper around __alloc_skb().
1164 static inline struct sk_buff *alloc_skb(unsigned int size,
1167 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1170 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1171 unsigned long data_len,
1175 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1177 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1178 struct sk_buff_fclones {
1179 struct sk_buff skb1;
1181 struct sk_buff skb2;
1183 refcount_t fclone_ref;
1187 * skb_fclone_busy - check if fclone is busy
1191 * Returns true if skb is a fast clone, and its clone is not freed.
1192 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1193 * so we also check that this didnt happen.
1195 static inline bool skb_fclone_busy(const struct sock *sk,
1196 const struct sk_buff *skb)
1198 const struct sk_buff_fclones *fclones;
1200 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1202 return skb->fclone == SKB_FCLONE_ORIG &&
1203 refcount_read(&fclones->fclone_ref) > 1 &&
1204 READ_ONCE(fclones->skb2.sk) == sk;
1208 * alloc_skb_fclone - allocate a network buffer from fclone cache
1209 * @size: size to allocate
1210 * @priority: allocation mask
1212 * This function is a convenient wrapper around __alloc_skb().
1214 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1217 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1220 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1221 void skb_headers_offset_update(struct sk_buff *skb, int off);
1222 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1223 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1224 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1225 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1226 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1227 gfp_t gfp_mask, bool fclone);
1228 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1231 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1234 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1235 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1236 unsigned int headroom);
1237 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1238 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1239 int newtailroom, gfp_t priority);
1240 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1241 int offset, int len);
1242 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1243 int offset, int len);
1244 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1245 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1248 * skb_pad - zero pad the tail of an skb
1249 * @skb: buffer to pad
1250 * @pad: space to pad
1252 * Ensure that a buffer is followed by a padding area that is zero
1253 * filled. Used by network drivers which may DMA or transfer data
1254 * beyond the buffer end onto the wire.
1256 * May return error in out of memory cases. The skb is freed on error.
1258 static inline int skb_pad(struct sk_buff *skb, int pad)
1260 return __skb_pad(skb, pad, true);
1262 #define dev_kfree_skb(a) consume_skb(a)
1264 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1265 int offset, size_t size);
1267 struct skb_seq_state {
1271 __u32 stepped_offset;
1272 struct sk_buff *root_skb;
1273 struct sk_buff *cur_skb;
1278 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1279 unsigned int to, struct skb_seq_state *st);
1280 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1281 struct skb_seq_state *st);
1282 void skb_abort_seq_read(struct skb_seq_state *st);
1284 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1285 unsigned int to, struct ts_config *config);
1288 * Packet hash types specify the type of hash in skb_set_hash.
1290 * Hash types refer to the protocol layer addresses which are used to
1291 * construct a packet's hash. The hashes are used to differentiate or identify
1292 * flows of the protocol layer for the hash type. Hash types are either
1293 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1295 * Properties of hashes:
1297 * 1) Two packets in different flows have different hash values
1298 * 2) Two packets in the same flow should have the same hash value
1300 * A hash at a higher layer is considered to be more specific. A driver should
1301 * set the most specific hash possible.
1303 * A driver cannot indicate a more specific hash than the layer at which a hash
1304 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1306 * A driver may indicate a hash level which is less specific than the
1307 * actual layer the hash was computed on. For instance, a hash computed
1308 * at L4 may be considered an L3 hash. This should only be done if the
1309 * driver can't unambiguously determine that the HW computed the hash at
1310 * the higher layer. Note that the "should" in the second property above
1313 enum pkt_hash_types {
1314 PKT_HASH_TYPE_NONE, /* Undefined type */
1315 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1316 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1317 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1320 static inline void skb_clear_hash(struct sk_buff *skb)
1327 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1330 skb_clear_hash(skb);
1334 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1336 skb->l4_hash = is_l4;
1337 skb->sw_hash = is_sw;
1342 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1344 /* Used by drivers to set hash from HW */
1345 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1349 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1351 __skb_set_hash(skb, hash, true, is_l4);
1354 void __skb_get_hash(struct sk_buff *skb);
1355 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1356 u32 skb_get_poff(const struct sk_buff *skb);
1357 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1358 const struct flow_keys_basic *keys, int hlen);
1359 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1360 const void *data, int hlen_proto);
1362 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1363 int thoff, u8 ip_proto)
1365 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1368 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1369 const struct flow_dissector_key *key,
1370 unsigned int key_count);
1372 struct bpf_flow_dissector;
1373 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1374 __be16 proto, int nhoff, int hlen, unsigned int flags);
1376 bool __skb_flow_dissect(const struct net *net,
1377 const struct sk_buff *skb,
1378 struct flow_dissector *flow_dissector,
1379 void *target_container, const void *data,
1380 __be16 proto, int nhoff, int hlen, unsigned int flags);
1382 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1383 struct flow_dissector *flow_dissector,
1384 void *target_container, unsigned int flags)
1386 return __skb_flow_dissect(NULL, skb, flow_dissector,
1387 target_container, NULL, 0, 0, 0, flags);
1390 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1391 struct flow_keys *flow,
1394 memset(flow, 0, sizeof(*flow));
1395 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1396 flow, NULL, 0, 0, 0, flags);
1400 skb_flow_dissect_flow_keys_basic(const struct net *net,
1401 const struct sk_buff *skb,
1402 struct flow_keys_basic *flow,
1403 const void *data, __be16 proto,
1404 int nhoff, int hlen, unsigned int flags)
1406 memset(flow, 0, sizeof(*flow));
1407 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1408 data, proto, nhoff, hlen, flags);
1411 void skb_flow_dissect_meta(const struct sk_buff *skb,
1412 struct flow_dissector *flow_dissector,
1413 void *target_container);
1415 /* Gets a skb connection tracking info, ctinfo map should be a
1416 * map of mapsize to translate enum ip_conntrack_info states
1420 skb_flow_dissect_ct(const struct sk_buff *skb,
1421 struct flow_dissector *flow_dissector,
1422 void *target_container,
1423 u16 *ctinfo_map, size_t mapsize,
1424 bool post_ct, u16 zone);
1426 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1427 struct flow_dissector *flow_dissector,
1428 void *target_container);
1430 void skb_flow_dissect_hash(const struct sk_buff *skb,
1431 struct flow_dissector *flow_dissector,
1432 void *target_container);
1434 static inline __u32 skb_get_hash(struct sk_buff *skb)
1436 if (!skb->l4_hash && !skb->sw_hash)
1437 __skb_get_hash(skb);
1442 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1444 if (!skb->l4_hash && !skb->sw_hash) {
1445 struct flow_keys keys;
1446 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1448 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1454 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1455 const siphash_key_t *perturb);
1457 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1462 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1464 to->hash = from->hash;
1465 to->sw_hash = from->sw_hash;
1466 to->l4_hash = from->l4_hash;
1469 static inline void skb_copy_decrypted(struct sk_buff *to,
1470 const struct sk_buff *from)
1472 #ifdef CONFIG_TLS_DEVICE
1473 to->decrypted = from->decrypted;
1477 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1478 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1480 return skb->head + skb->end;
1483 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1488 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1493 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1498 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1500 return skb->end - skb->head;
1503 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1505 skb->end = skb->head + offset;
1510 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1512 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1514 return &skb_shinfo(skb)->hwtstamps;
1517 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1519 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1521 return is_zcopy ? skb_uarg(skb) : NULL;
1524 static inline void net_zcopy_get(struct ubuf_info *uarg)
1526 refcount_inc(&uarg->refcnt);
1529 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1531 skb_shinfo(skb)->destructor_arg = uarg;
1532 skb_shinfo(skb)->flags |= uarg->flags;
1535 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1538 if (skb && uarg && !skb_zcopy(skb)) {
1539 if (unlikely(have_ref && *have_ref))
1542 net_zcopy_get(uarg);
1543 skb_zcopy_init(skb, uarg);
1547 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1549 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1550 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1553 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1555 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1558 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1560 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1563 static inline void net_zcopy_put(struct ubuf_info *uarg)
1566 uarg->callback(NULL, uarg, true);
1569 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1572 if (uarg->callback == msg_zerocopy_callback)
1573 msg_zerocopy_put_abort(uarg, have_uref);
1575 net_zcopy_put(uarg);
1579 /* Release a reference on a zerocopy structure */
1580 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1582 struct ubuf_info *uarg = skb_zcopy(skb);
1585 if (!skb_zcopy_is_nouarg(skb))
1586 uarg->callback(skb, uarg, zerocopy_success);
1588 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
1592 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1597 /* Iterate through singly-linked GSO fragments of an skb. */
1598 #define skb_list_walk_safe(first, skb, next_skb) \
1599 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1600 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1602 static inline void skb_list_del_init(struct sk_buff *skb)
1604 __list_del_entry(&skb->list);
1605 skb_mark_not_on_list(skb);
1609 * skb_queue_empty - check if a queue is empty
1612 * Returns true if the queue is empty, false otherwise.
1614 static inline int skb_queue_empty(const struct sk_buff_head *list)
1616 return list->next == (const struct sk_buff *) list;
1620 * skb_queue_empty_lockless - check if a queue is empty
1623 * Returns true if the queue is empty, false otherwise.
1624 * This variant can be used in lockless contexts.
1626 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1628 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1633 * skb_queue_is_last - check if skb is the last entry in the queue
1637 * Returns true if @skb is the last buffer on the list.
1639 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1640 const struct sk_buff *skb)
1642 return skb->next == (const struct sk_buff *) list;
1646 * skb_queue_is_first - check if skb is the first entry in the queue
1650 * Returns true if @skb is the first buffer on the list.
1652 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1653 const struct sk_buff *skb)
1655 return skb->prev == (const struct sk_buff *) list;
1659 * skb_queue_next - return the next packet in the queue
1661 * @skb: current buffer
1663 * Return the next packet in @list after @skb. It is only valid to
1664 * call this if skb_queue_is_last() evaluates to false.
1666 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1667 const struct sk_buff *skb)
1669 /* This BUG_ON may seem severe, but if we just return then we
1670 * are going to dereference garbage.
1672 BUG_ON(skb_queue_is_last(list, skb));
1677 * skb_queue_prev - return the prev packet in the queue
1679 * @skb: current buffer
1681 * Return the prev packet in @list before @skb. It is only valid to
1682 * call this if skb_queue_is_first() evaluates to false.
1684 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1685 const struct sk_buff *skb)
1687 /* This BUG_ON may seem severe, but if we just return then we
1688 * are going to dereference garbage.
1690 BUG_ON(skb_queue_is_first(list, skb));
1695 * skb_get - reference buffer
1696 * @skb: buffer to reference
1698 * Makes another reference to a socket buffer and returns a pointer
1701 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1703 refcount_inc(&skb->users);
1708 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1712 * skb_cloned - is the buffer a clone
1713 * @skb: buffer to check
1715 * Returns true if the buffer was generated with skb_clone() and is
1716 * one of multiple shared copies of the buffer. Cloned buffers are
1717 * shared data so must not be written to under normal circumstances.
1719 static inline int skb_cloned(const struct sk_buff *skb)
1721 return skb->cloned &&
1722 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1725 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1727 might_sleep_if(gfpflags_allow_blocking(pri));
1729 if (skb_cloned(skb))
1730 return pskb_expand_head(skb, 0, 0, pri);
1735 /* This variant of skb_unclone() makes sure skb->truesize
1736 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1738 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1739 * when various debugging features are in place.
1741 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
1742 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1744 might_sleep_if(gfpflags_allow_blocking(pri));
1746 if (skb_cloned(skb))
1747 return __skb_unclone_keeptruesize(skb, pri);
1752 * skb_header_cloned - is the header a clone
1753 * @skb: buffer to check
1755 * Returns true if modifying the header part of the buffer requires
1756 * the data to be copied.
1758 static inline int skb_header_cloned(const struct sk_buff *skb)
1765 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1766 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1767 return dataref != 1;
1770 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1772 might_sleep_if(gfpflags_allow_blocking(pri));
1774 if (skb_header_cloned(skb))
1775 return pskb_expand_head(skb, 0, 0, pri);
1781 * __skb_header_release - release reference to header
1782 * @skb: buffer to operate on
1784 static inline void __skb_header_release(struct sk_buff *skb)
1787 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1792 * skb_shared - is the buffer shared
1793 * @skb: buffer to check
1795 * Returns true if more than one person has a reference to this
1798 static inline int skb_shared(const struct sk_buff *skb)
1800 return refcount_read(&skb->users) != 1;
1804 * skb_share_check - check if buffer is shared and if so clone it
1805 * @skb: buffer to check
1806 * @pri: priority for memory allocation
1808 * If the buffer is shared the buffer is cloned and the old copy
1809 * drops a reference. A new clone with a single reference is returned.
1810 * If the buffer is not shared the original buffer is returned. When
1811 * being called from interrupt status or with spinlocks held pri must
1814 * NULL is returned on a memory allocation failure.
1816 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1818 might_sleep_if(gfpflags_allow_blocking(pri));
1819 if (skb_shared(skb)) {
1820 struct sk_buff *nskb = skb_clone(skb, pri);
1832 * Copy shared buffers into a new sk_buff. We effectively do COW on
1833 * packets to handle cases where we have a local reader and forward
1834 * and a couple of other messy ones. The normal one is tcpdumping
1835 * a packet thats being forwarded.
1839 * skb_unshare - make a copy of a shared buffer
1840 * @skb: buffer to check
1841 * @pri: priority for memory allocation
1843 * If the socket buffer is a clone then this function creates a new
1844 * copy of the data, drops a reference count on the old copy and returns
1845 * the new copy with the reference count at 1. If the buffer is not a clone
1846 * the original buffer is returned. When called with a spinlock held or
1847 * from interrupt state @pri must be %GFP_ATOMIC
1849 * %NULL is returned on a memory allocation failure.
1851 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1854 might_sleep_if(gfpflags_allow_blocking(pri));
1855 if (skb_cloned(skb)) {
1856 struct sk_buff *nskb = skb_copy(skb, pri);
1858 /* Free our shared copy */
1869 * skb_peek - peek at the head of an &sk_buff_head
1870 * @list_: list to peek at
1872 * Peek an &sk_buff. Unlike most other operations you _MUST_
1873 * be careful with this one. A peek leaves the buffer on the
1874 * list and someone else may run off with it. You must hold
1875 * the appropriate locks or have a private queue to do this.
1877 * Returns %NULL for an empty list or a pointer to the head element.
1878 * The reference count is not incremented and the reference is therefore
1879 * volatile. Use with caution.
1881 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1883 struct sk_buff *skb = list_->next;
1885 if (skb == (struct sk_buff *)list_)
1891 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1892 * @list_: list to peek at
1894 * Like skb_peek(), but the caller knows that the list is not empty.
1896 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1902 * skb_peek_next - peek skb following the given one from a queue
1903 * @skb: skb to start from
1904 * @list_: list to peek at
1906 * Returns %NULL when the end of the list is met or a pointer to the
1907 * next element. The reference count is not incremented and the
1908 * reference is therefore volatile. Use with caution.
1910 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1911 const struct sk_buff_head *list_)
1913 struct sk_buff *next = skb->next;
1915 if (next == (struct sk_buff *)list_)
1921 * skb_peek_tail - peek at the tail of an &sk_buff_head
1922 * @list_: list to peek at
1924 * Peek an &sk_buff. Unlike most other operations you _MUST_
1925 * be careful with this one. A peek leaves the buffer on the
1926 * list and someone else may run off with it. You must hold
1927 * the appropriate locks or have a private queue to do this.
1929 * Returns %NULL for an empty list or a pointer to the tail element.
1930 * The reference count is not incremented and the reference is therefore
1931 * volatile. Use with caution.
1933 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1935 struct sk_buff *skb = READ_ONCE(list_->prev);
1937 if (skb == (struct sk_buff *)list_)
1944 * skb_queue_len - get queue length
1945 * @list_: list to measure
1947 * Return the length of an &sk_buff queue.
1949 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1955 * skb_queue_len_lockless - get queue length
1956 * @list_: list to measure
1958 * Return the length of an &sk_buff queue.
1959 * This variant can be used in lockless contexts.
1961 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1963 return READ_ONCE(list_->qlen);
1967 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1968 * @list: queue to initialize
1970 * This initializes only the list and queue length aspects of
1971 * an sk_buff_head object. This allows to initialize the list
1972 * aspects of an sk_buff_head without reinitializing things like
1973 * the spinlock. It can also be used for on-stack sk_buff_head
1974 * objects where the spinlock is known to not be used.
1976 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1978 list->prev = list->next = (struct sk_buff *)list;
1983 * This function creates a split out lock class for each invocation;
1984 * this is needed for now since a whole lot of users of the skb-queue
1985 * infrastructure in drivers have different locking usage (in hardirq)
1986 * than the networking core (in softirq only). In the long run either the
1987 * network layer or drivers should need annotation to consolidate the
1988 * main types of usage into 3 classes.
1990 static inline void skb_queue_head_init(struct sk_buff_head *list)
1992 spin_lock_init(&list->lock);
1993 __skb_queue_head_init(list);
1996 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1997 struct lock_class_key *class)
1999 skb_queue_head_init(list);
2000 lockdep_set_class(&list->lock, class);
2004 * Insert an sk_buff on a list.
2006 * The "__skb_xxxx()" functions are the non-atomic ones that
2007 * can only be called with interrupts disabled.
2009 static inline void __skb_insert(struct sk_buff *newsk,
2010 struct sk_buff *prev, struct sk_buff *next,
2011 struct sk_buff_head *list)
2013 /* See skb_queue_empty_lockless() and skb_peek_tail()
2014 * for the opposite READ_ONCE()
2016 WRITE_ONCE(newsk->next, next);
2017 WRITE_ONCE(newsk->prev, prev);
2018 WRITE_ONCE(next->prev, newsk);
2019 WRITE_ONCE(prev->next, newsk);
2020 WRITE_ONCE(list->qlen, list->qlen + 1);
2023 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2024 struct sk_buff *prev,
2025 struct sk_buff *next)
2027 struct sk_buff *first = list->next;
2028 struct sk_buff *last = list->prev;
2030 WRITE_ONCE(first->prev, prev);
2031 WRITE_ONCE(prev->next, first);
2033 WRITE_ONCE(last->next, next);
2034 WRITE_ONCE(next->prev, last);
2038 * skb_queue_splice - join two skb lists, this is designed for stacks
2039 * @list: the new list to add
2040 * @head: the place to add it in the first list
2042 static inline void skb_queue_splice(const struct sk_buff_head *list,
2043 struct sk_buff_head *head)
2045 if (!skb_queue_empty(list)) {
2046 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2047 head->qlen += list->qlen;
2052 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2053 * @list: the new list to add
2054 * @head: the place to add it in the first list
2056 * The list at @list is reinitialised
2058 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2059 struct sk_buff_head *head)
2061 if (!skb_queue_empty(list)) {
2062 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2063 head->qlen += list->qlen;
2064 __skb_queue_head_init(list);
2069 * skb_queue_splice_tail - join two skb lists, each list being a queue
2070 * @list: the new list to add
2071 * @head: the place to add it in the first list
2073 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2074 struct sk_buff_head *head)
2076 if (!skb_queue_empty(list)) {
2077 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2078 head->qlen += list->qlen;
2083 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2084 * @list: the new list to add
2085 * @head: the place to add it in the first list
2087 * Each of the lists is a queue.
2088 * The list at @list is reinitialised
2090 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2091 struct sk_buff_head *head)
2093 if (!skb_queue_empty(list)) {
2094 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2095 head->qlen += list->qlen;
2096 __skb_queue_head_init(list);
2101 * __skb_queue_after - queue a buffer at the list head
2102 * @list: list to use
2103 * @prev: place after this buffer
2104 * @newsk: buffer to queue
2106 * Queue a buffer int the middle of a list. This function takes no locks
2107 * and you must therefore hold required locks before calling it.
2109 * A buffer cannot be placed on two lists at the same time.
2111 static inline void __skb_queue_after(struct sk_buff_head *list,
2112 struct sk_buff *prev,
2113 struct sk_buff *newsk)
2115 __skb_insert(newsk, prev, prev->next, list);
2118 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2119 struct sk_buff_head *list);
2121 static inline void __skb_queue_before(struct sk_buff_head *list,
2122 struct sk_buff *next,
2123 struct sk_buff *newsk)
2125 __skb_insert(newsk, next->prev, next, list);
2129 * __skb_queue_head - queue a buffer at the list head
2130 * @list: list to use
2131 * @newsk: buffer to queue
2133 * Queue a buffer at the start of a list. This function takes no locks
2134 * and you must therefore hold required locks before calling it.
2136 * A buffer cannot be placed on two lists at the same time.
2138 static inline void __skb_queue_head(struct sk_buff_head *list,
2139 struct sk_buff *newsk)
2141 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2143 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2146 * __skb_queue_tail - queue a buffer at the list tail
2147 * @list: list to use
2148 * @newsk: buffer to queue
2150 * Queue a buffer at the end of a list. This function takes no locks
2151 * and you must therefore hold required locks before calling it.
2153 * A buffer cannot be placed on two lists at the same time.
2155 static inline void __skb_queue_tail(struct sk_buff_head *list,
2156 struct sk_buff *newsk)
2158 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2160 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2163 * remove sk_buff from list. _Must_ be called atomically, and with
2166 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2167 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2169 struct sk_buff *next, *prev;
2171 WRITE_ONCE(list->qlen, list->qlen - 1);
2174 skb->next = skb->prev = NULL;
2175 WRITE_ONCE(next->prev, prev);
2176 WRITE_ONCE(prev->next, next);
2180 * __skb_dequeue - remove from the head of the queue
2181 * @list: list to dequeue from
2183 * Remove the head of the list. This function does not take any locks
2184 * so must be used with appropriate locks held only. The head item is
2185 * returned or %NULL if the list is empty.
2187 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2189 struct sk_buff *skb = skb_peek(list);
2191 __skb_unlink(skb, list);
2194 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2197 * __skb_dequeue_tail - remove from the tail of the queue
2198 * @list: list to dequeue from
2200 * Remove the tail of the list. This function does not take any locks
2201 * so must be used with appropriate locks held only. The tail item is
2202 * returned or %NULL if the list is empty.
2204 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2206 struct sk_buff *skb = skb_peek_tail(list);
2208 __skb_unlink(skb, list);
2211 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2214 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2216 return skb->data_len;
2219 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2221 return skb->len - skb->data_len;
2224 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2226 unsigned int i, len = 0;
2228 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2229 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2233 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2235 return skb_headlen(skb) + __skb_pagelen(skb);
2238 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2239 int i, struct page *page,
2242 skb_frag_t *frag = &shinfo->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);
2255 * __skb_fill_page_desc - initialise a paged fragment in an skb
2256 * @skb: buffer containing fragment to be initialised
2257 * @i: paged fragment index to initialise
2258 * @page: the page to use for this fragment
2259 * @off: the offset to the data with @page
2260 * @size: the length of the data
2262 * Initialises the @i'th fragment of @skb to point to &size bytes at
2263 * offset @off within @page.
2265 * Does not take any additional reference on the fragment.
2267 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2268 struct page *page, int off, int size)
2270 __skb_fill_page_desc_noacc(skb_shinfo(skb), i, page, off, size);
2271 page = compound_head(page);
2272 if (page_is_pfmemalloc(page))
2273 skb->pfmemalloc = true;
2277 * skb_fill_page_desc - initialise a paged fragment in an skb
2278 * @skb: buffer containing fragment to be initialised
2279 * @i: paged fragment index to initialise
2280 * @page: the page to use for this fragment
2281 * @off: the offset to the data with @page
2282 * @size: the length of the data
2284 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2285 * @skb to point to @size bytes at offset @off within @page. In
2286 * addition updates @skb such that @i is the last fragment.
2288 * Does not take any additional reference on the fragment.
2290 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2291 struct page *page, int off, int size)
2293 __skb_fill_page_desc(skb, i, page, off, size);
2294 skb_shinfo(skb)->nr_frags = i + 1;
2298 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2299 * @skb: buffer containing fragment to be initialised
2300 * @i: paged fragment index to initialise
2301 * @page: the page to use for this fragment
2302 * @off: the offset to the data with @page
2303 * @size: the length of the data
2305 * Variant of skb_fill_page_desc() which does not deal with
2306 * pfmemalloc, if page is not owned by us.
2308 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2309 struct page *page, int off,
2312 struct skb_shared_info *shinfo = skb_shinfo(skb);
2314 __skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2315 shinfo->nr_frags = i + 1;
2318 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2319 int size, unsigned int truesize);
2321 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2322 unsigned int truesize);
2324 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2326 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2327 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2329 return skb->head + skb->tail;
2332 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2334 skb->tail = skb->data - skb->head;
2337 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2339 skb_reset_tail_pointer(skb);
2340 skb->tail += offset;
2343 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2344 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2349 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2351 skb->tail = skb->data;
2354 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2356 skb->tail = skb->data + offset;
2359 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2361 static inline void skb_assert_len(struct sk_buff *skb)
2363 #ifdef CONFIG_DEBUG_NET
2364 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2365 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2366 #endif /* CONFIG_DEBUG_NET */
2370 * Add data to an sk_buff
2372 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2373 void *skb_put(struct sk_buff *skb, unsigned int len);
2374 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2376 void *tmp = skb_tail_pointer(skb);
2377 SKB_LINEAR_ASSERT(skb);
2383 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2385 void *tmp = __skb_put(skb, len);
2387 memset(tmp, 0, len);
2391 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2394 void *tmp = __skb_put(skb, len);
2396 memcpy(tmp, data, len);
2400 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2402 *(u8 *)__skb_put(skb, 1) = val;
2405 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2407 void *tmp = skb_put(skb, len);
2409 memset(tmp, 0, len);
2414 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2417 void *tmp = skb_put(skb, len);
2419 memcpy(tmp, data, len);
2424 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2426 *(u8 *)skb_put(skb, 1) = val;
2429 void *skb_push(struct sk_buff *skb, unsigned int len);
2430 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2437 void *skb_pull(struct sk_buff *skb, unsigned int len);
2438 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2441 BUG_ON(skb->len < skb->data_len);
2442 return skb->data += len;
2445 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2447 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2450 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2452 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2454 if (len > skb_headlen(skb) &&
2455 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2458 return skb->data += len;
2461 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2463 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2466 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2468 if (likely(len <= skb_headlen(skb)))
2470 if (unlikely(len > skb->len))
2472 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2475 void skb_condense(struct sk_buff *skb);
2478 * skb_headroom - bytes at buffer head
2479 * @skb: buffer to check
2481 * Return the number of bytes of free space at the head of an &sk_buff.
2483 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2485 return skb->data - skb->head;
2489 * skb_tailroom - bytes at buffer end
2490 * @skb: buffer to check
2492 * Return the number of bytes of free space at the tail of an sk_buff
2494 static inline int skb_tailroom(const struct sk_buff *skb)
2496 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2500 * skb_availroom - bytes at buffer end
2501 * @skb: buffer to check
2503 * Return the number of bytes of free space at the tail of an sk_buff
2504 * allocated by sk_stream_alloc()
2506 static inline int skb_availroom(const struct sk_buff *skb)
2508 if (skb_is_nonlinear(skb))
2511 return skb->end - skb->tail - skb->reserved_tailroom;
2515 * skb_reserve - adjust headroom
2516 * @skb: buffer to alter
2517 * @len: bytes to move
2519 * Increase the headroom of an empty &sk_buff by reducing the tail
2520 * room. This is only allowed for an empty buffer.
2522 static inline void skb_reserve(struct sk_buff *skb, int len)
2529 * skb_tailroom_reserve - adjust reserved_tailroom
2530 * @skb: buffer to alter
2531 * @mtu: maximum amount of headlen permitted
2532 * @needed_tailroom: minimum amount of reserved_tailroom
2534 * Set reserved_tailroom so that headlen can be as large as possible but
2535 * not larger than mtu and tailroom cannot be smaller than
2537 * The required headroom should already have been reserved before using
2540 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2541 unsigned int needed_tailroom)
2543 SKB_LINEAR_ASSERT(skb);
2544 if (mtu < skb_tailroom(skb) - needed_tailroom)
2545 /* use at most mtu */
2546 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2548 /* use up to all available space */
2549 skb->reserved_tailroom = needed_tailroom;
2552 #define ENCAP_TYPE_ETHER 0
2553 #define ENCAP_TYPE_IPPROTO 1
2555 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2558 skb->inner_protocol = protocol;
2559 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2562 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2565 skb->inner_ipproto = ipproto;
2566 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2569 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2571 skb->inner_mac_header = skb->mac_header;
2572 skb->inner_network_header = skb->network_header;
2573 skb->inner_transport_header = skb->transport_header;
2576 static inline void skb_reset_mac_len(struct sk_buff *skb)
2578 skb->mac_len = skb->network_header - skb->mac_header;
2581 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2584 return skb->head + skb->inner_transport_header;
2587 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2589 return skb_inner_transport_header(skb) - skb->data;
2592 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2594 skb->inner_transport_header = skb->data - skb->head;
2597 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2600 skb_reset_inner_transport_header(skb);
2601 skb->inner_transport_header += offset;
2604 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2606 return skb->head + skb->inner_network_header;
2609 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2611 skb->inner_network_header = skb->data - skb->head;
2614 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2617 skb_reset_inner_network_header(skb);
2618 skb->inner_network_header += offset;
2621 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2623 return skb->head + skb->inner_mac_header;
2626 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2628 skb->inner_mac_header = skb->data - skb->head;
2631 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2634 skb_reset_inner_mac_header(skb);
2635 skb->inner_mac_header += offset;
2637 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2639 return skb->transport_header != (typeof(skb->transport_header))~0U;
2642 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2644 return skb->head + skb->transport_header;
2647 static inline void skb_reset_transport_header(struct sk_buff *skb)
2649 skb->transport_header = skb->data - skb->head;
2652 static inline void skb_set_transport_header(struct sk_buff *skb,
2655 skb_reset_transport_header(skb);
2656 skb->transport_header += offset;
2659 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2661 return skb->head + skb->network_header;
2664 static inline void skb_reset_network_header(struct sk_buff *skb)
2666 skb->network_header = skb->data - skb->head;
2669 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2671 skb_reset_network_header(skb);
2672 skb->network_header += offset;
2675 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2677 return skb->head + skb->mac_header;
2680 static inline int skb_mac_offset(const struct sk_buff *skb)
2682 return skb_mac_header(skb) - skb->data;
2685 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2687 return skb->network_header - skb->mac_header;
2690 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2692 return skb->mac_header != (typeof(skb->mac_header))~0U;
2695 static inline void skb_unset_mac_header(struct sk_buff *skb)
2697 skb->mac_header = (typeof(skb->mac_header))~0U;
2700 static inline void skb_reset_mac_header(struct sk_buff *skb)
2702 skb->mac_header = skb->data - skb->head;
2705 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2707 skb_reset_mac_header(skb);
2708 skb->mac_header += offset;
2711 static inline void skb_pop_mac_header(struct sk_buff *skb)
2713 skb->mac_header = skb->network_header;
2716 static inline void skb_probe_transport_header(struct sk_buff *skb)
2718 struct flow_keys_basic keys;
2720 if (skb_transport_header_was_set(skb))
2723 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2725 skb_set_transport_header(skb, keys.control.thoff);
2728 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2730 if (skb_mac_header_was_set(skb)) {
2731 const unsigned char *old_mac = skb_mac_header(skb);
2733 skb_set_mac_header(skb, -skb->mac_len);
2734 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2738 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2740 return skb->csum_start - skb_headroom(skb);
2743 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2745 return skb->head + skb->csum_start;
2748 static inline int skb_transport_offset(const struct sk_buff *skb)
2750 return skb_transport_header(skb) - skb->data;
2753 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2755 return skb->transport_header - skb->network_header;
2758 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2760 return skb->inner_transport_header - skb->inner_network_header;
2763 static inline int skb_network_offset(const struct sk_buff *skb)
2765 return skb_network_header(skb) - skb->data;
2768 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2770 return skb_inner_network_header(skb) - skb->data;
2773 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2775 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2779 * CPUs often take a performance hit when accessing unaligned memory
2780 * locations. The actual performance hit varies, it can be small if the
2781 * hardware handles it or large if we have to take an exception and fix it
2784 * Since an ethernet header is 14 bytes network drivers often end up with
2785 * the IP header at an unaligned offset. The IP header can be aligned by
2786 * shifting the start of the packet by 2 bytes. Drivers should do this
2789 * skb_reserve(skb, NET_IP_ALIGN);
2791 * The downside to this alignment of the IP header is that the DMA is now
2792 * unaligned. On some architectures the cost of an unaligned DMA is high
2793 * and this cost outweighs the gains made by aligning the IP header.
2795 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2798 #ifndef NET_IP_ALIGN
2799 #define NET_IP_ALIGN 2
2803 * The networking layer reserves some headroom in skb data (via
2804 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2805 * the header has to grow. In the default case, if the header has to grow
2806 * 32 bytes or less we avoid the reallocation.
2808 * Unfortunately this headroom changes the DMA alignment of the resulting
2809 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2810 * on some architectures. An architecture can override this value,
2811 * perhaps setting it to a cacheline in size (since that will maintain
2812 * cacheline alignment of the DMA). It must be a power of 2.
2814 * Various parts of the networking layer expect at least 32 bytes of
2815 * headroom, you should not reduce this.
2817 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2818 * to reduce average number of cache lines per packet.
2819 * get_rps_cpu() for example only access one 64 bytes aligned block :
2820 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2823 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2826 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2828 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2830 if (WARN_ON(skb_is_nonlinear(skb)))
2833 skb_set_tail_pointer(skb, len);
2836 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2838 __skb_set_length(skb, len);
2841 void skb_trim(struct sk_buff *skb, unsigned int len);
2843 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2846 return ___pskb_trim(skb, len);
2847 __skb_trim(skb, len);
2851 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2853 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2857 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2858 * @skb: buffer to alter
2861 * This is identical to pskb_trim except that the caller knows that
2862 * the skb is not cloned so we should never get an error due to out-
2865 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2867 int err = pskb_trim(skb, len);
2871 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2873 unsigned int diff = len - skb->len;
2875 if (skb_tailroom(skb) < diff) {
2876 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2881 __skb_set_length(skb, len);
2886 * skb_orphan - orphan a buffer
2887 * @skb: buffer to orphan
2889 * If a buffer currently has an owner then we call the owner's
2890 * destructor function and make the @skb unowned. The buffer continues
2891 * to exist but is no longer charged to its former owner.
2893 static inline void skb_orphan(struct sk_buff *skb)
2895 if (skb->destructor) {
2896 skb->destructor(skb);
2897 skb->destructor = NULL;
2905 * skb_orphan_frags - orphan the frags contained in a buffer
2906 * @skb: buffer to orphan frags from
2907 * @gfp_mask: allocation mask for replacement pages
2909 * For each frag in the SKB which needs a destructor (i.e. has an
2910 * owner) create a copy of that frag and release the original
2911 * page by calling the destructor.
2913 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2915 if (likely(!skb_zcopy(skb)))
2917 if (!skb_zcopy_is_nouarg(skb) &&
2918 skb_uarg(skb)->callback == msg_zerocopy_callback)
2920 return skb_copy_ubufs(skb, gfp_mask);
2923 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2924 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2926 if (likely(!skb_zcopy(skb)))
2928 return skb_copy_ubufs(skb, gfp_mask);
2932 * __skb_queue_purge - empty a list
2933 * @list: list to empty
2935 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2936 * the list and one reference dropped. This function does not take the
2937 * list lock and the caller must hold the relevant locks to use it.
2939 static inline void __skb_queue_purge(struct sk_buff_head *list)
2941 struct sk_buff *skb;
2942 while ((skb = __skb_dequeue(list)) != NULL)
2945 void skb_queue_purge(struct sk_buff_head *list);
2947 unsigned int skb_rbtree_purge(struct rb_root *root);
2949 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2952 * netdev_alloc_frag - allocate a page fragment
2953 * @fragsz: fragment size
2955 * Allocates a frag from a page for receive buffer.
2956 * Uses GFP_ATOMIC allocations.
2958 static inline void *netdev_alloc_frag(unsigned int fragsz)
2960 return __netdev_alloc_frag_align(fragsz, ~0u);
2963 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2966 WARN_ON_ONCE(!is_power_of_2(align));
2967 return __netdev_alloc_frag_align(fragsz, -align);
2970 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2974 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2975 * @dev: network device to receive on
2976 * @length: length to allocate
2978 * Allocate a new &sk_buff and assign it a usage count of one. The
2979 * buffer has unspecified headroom built in. Users should allocate
2980 * the headroom they think they need without accounting for the
2981 * built in space. The built in space is used for optimisations.
2983 * %NULL is returned if there is no free memory. Although this function
2984 * allocates memory it can be called from an interrupt.
2986 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2987 unsigned int length)
2989 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2992 /* legacy helper around __netdev_alloc_skb() */
2993 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2996 return __netdev_alloc_skb(NULL, length, gfp_mask);
2999 /* legacy helper around netdev_alloc_skb() */
3000 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3002 return netdev_alloc_skb(NULL, length);
3006 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3007 unsigned int length, gfp_t gfp)
3009 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3011 if (NET_IP_ALIGN && skb)
3012 skb_reserve(skb, NET_IP_ALIGN);
3016 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3017 unsigned int length)
3019 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3022 static inline void skb_free_frag(void *addr)
3024 page_frag_free(addr);
3027 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3029 static inline void *napi_alloc_frag(unsigned int fragsz)
3031 return __napi_alloc_frag_align(fragsz, ~0u);
3034 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3037 WARN_ON_ONCE(!is_power_of_2(align));
3038 return __napi_alloc_frag_align(fragsz, -align);
3041 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3042 unsigned int length, gfp_t gfp_mask);
3043 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3044 unsigned int length)
3046 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3048 void napi_consume_skb(struct sk_buff *skb, int budget);
3050 void napi_skb_free_stolen_head(struct sk_buff *skb);
3051 void __kfree_skb_defer(struct sk_buff *skb);
3054 * __dev_alloc_pages - allocate page for network Rx
3055 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3056 * @order: size of the allocation
3058 * Allocate a new page.
3060 * %NULL is returned if there is no free memory.
3062 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3065 /* This piece of code contains several assumptions.
3066 * 1. This is for device Rx, therefor a cold page is preferred.
3067 * 2. The expectation is the user wants a compound page.
3068 * 3. If requesting a order 0 page it will not be compound
3069 * due to the check to see if order has a value in prep_new_page
3070 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3071 * code in gfp_to_alloc_flags that should be enforcing this.
3073 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3075 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3078 static inline struct page *dev_alloc_pages(unsigned int order)
3080 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3084 * __dev_alloc_page - allocate a page for network Rx
3085 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3087 * Allocate a new page.
3089 * %NULL is returned if there is no free memory.
3091 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3093 return __dev_alloc_pages(gfp_mask, 0);
3096 static inline struct page *dev_alloc_page(void)
3098 return dev_alloc_pages(0);
3102 * dev_page_is_reusable - check whether a page can be reused for network Rx
3103 * @page: the page to test
3105 * A page shouldn't be considered for reusing/recycling if it was allocated
3106 * under memory pressure or at a distant memory node.
3108 * Returns false if this page should be returned to page allocator, true
3111 static inline bool dev_page_is_reusable(const struct page *page)
3113 return likely(page_to_nid(page) == numa_mem_id() &&
3114 !page_is_pfmemalloc(page));
3118 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3119 * @page: The page that was allocated from skb_alloc_page
3120 * @skb: The skb that may need pfmemalloc set
3122 static inline void skb_propagate_pfmemalloc(const struct page *page,
3123 struct sk_buff *skb)
3125 if (page_is_pfmemalloc(page))
3126 skb->pfmemalloc = true;
3130 * skb_frag_off() - Returns the offset of a skb fragment
3131 * @frag: the paged fragment
3133 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3135 return frag->bv_offset;
3139 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3140 * @frag: skb fragment
3141 * @delta: value to add
3143 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3145 frag->bv_offset += delta;
3149 * skb_frag_off_set() - Sets the offset of a skb fragment
3150 * @frag: skb fragment
3151 * @offset: offset of fragment
3153 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3155 frag->bv_offset = offset;
3159 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3160 * @fragto: skb fragment where offset is set
3161 * @fragfrom: skb fragment offset is copied from
3163 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3164 const skb_frag_t *fragfrom)
3166 fragto->bv_offset = fragfrom->bv_offset;
3170 * skb_frag_page - retrieve the page referred to by a paged fragment
3171 * @frag: the paged fragment
3173 * Returns the &struct page associated with @frag.
3175 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3177 return frag->bv_page;
3181 * __skb_frag_ref - take an addition reference on a paged fragment.
3182 * @frag: the paged fragment
3184 * Takes an additional reference on the paged fragment @frag.
3186 static inline void __skb_frag_ref(skb_frag_t *frag)
3188 get_page(skb_frag_page(frag));
3192 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3194 * @f: the fragment offset.
3196 * Takes an additional reference on the @f'th paged fragment of @skb.
3198 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3200 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3204 * __skb_frag_unref - release a reference on a paged fragment.
3205 * @frag: the paged fragment
3206 * @recycle: recycle the page if allocated via page_pool
3208 * Releases a reference on the paged fragment @frag
3209 * or recycles the page via the page_pool API.
3211 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3213 struct page *page = skb_frag_page(frag);
3215 #ifdef CONFIG_PAGE_POOL
3216 if (recycle && page_pool_return_skb_page(page))
3223 * skb_frag_unref - release a reference on a paged fragment of an skb.
3225 * @f: the fragment offset
3227 * Releases a reference on the @f'th paged fragment of @skb.
3229 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3231 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3235 * skb_frag_address - gets the address of the data contained in a paged fragment
3236 * @frag: the paged fragment buffer
3238 * Returns the address of the data within @frag. The page must already
3241 static inline void *skb_frag_address(const skb_frag_t *frag)
3243 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3247 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3248 * @frag: the paged fragment buffer
3250 * Returns the address of the data within @frag. Checks that the page
3251 * is mapped and returns %NULL otherwise.
3253 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3255 void *ptr = page_address(skb_frag_page(frag));
3259 return ptr + skb_frag_off(frag);
3263 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3264 * @fragto: skb fragment where page is set
3265 * @fragfrom: skb fragment page is copied from
3267 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3268 const skb_frag_t *fragfrom)
3270 fragto->bv_page = fragfrom->bv_page;
3274 * __skb_frag_set_page - sets the page contained in a paged fragment
3275 * @frag: the paged fragment
3276 * @page: the page to set
3278 * Sets the fragment @frag to contain @page.
3280 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3282 frag->bv_page = page;
3286 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3288 * @f: the fragment offset
3289 * @page: the page to set
3291 * Sets the @f'th fragment of @skb to contain @page.
3293 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3296 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3299 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3302 * skb_frag_dma_map - maps a paged fragment via the DMA API
3303 * @dev: the device to map the fragment to
3304 * @frag: the paged fragment to map
3305 * @offset: the offset within the fragment (starting at the
3306 * fragment's own offset)
3307 * @size: the number of bytes to map
3308 * @dir: the direction of the mapping (``PCI_DMA_*``)
3310 * Maps the page associated with @frag to @device.
3312 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3313 const skb_frag_t *frag,
3314 size_t offset, size_t size,
3315 enum dma_data_direction dir)
3317 return dma_map_page(dev, skb_frag_page(frag),
3318 skb_frag_off(frag) + offset, size, dir);
3321 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3324 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3328 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3331 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3336 * skb_clone_writable - is the header of a clone writable
3337 * @skb: buffer to check
3338 * @len: length up to which to write
3340 * Returns true if modifying the header part of the cloned buffer
3341 * does not requires the data to be copied.
3343 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3345 return !skb_header_cloned(skb) &&
3346 skb_headroom(skb) + len <= skb->hdr_len;
3349 static inline int skb_try_make_writable(struct sk_buff *skb,
3350 unsigned int write_len)
3352 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3353 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3356 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3361 if (headroom > skb_headroom(skb))
3362 delta = headroom - skb_headroom(skb);
3364 if (delta || cloned)
3365 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3371 * skb_cow - copy header of skb when it is required
3372 * @skb: buffer to cow
3373 * @headroom: needed headroom
3375 * If the skb passed lacks sufficient headroom or its data part
3376 * is shared, data is reallocated. If reallocation fails, an error
3377 * is returned and original skb is not changed.
3379 * The result is skb with writable area skb->head...skb->tail
3380 * and at least @headroom of space at head.
3382 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3384 return __skb_cow(skb, headroom, skb_cloned(skb));
3388 * skb_cow_head - skb_cow but only making the head writable
3389 * @skb: buffer to cow
3390 * @headroom: needed headroom
3392 * This function is identical to skb_cow except that we replace the
3393 * skb_cloned check by skb_header_cloned. It should be used when
3394 * you only need to push on some header and do not need to modify
3397 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3399 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3403 * skb_padto - pad an skbuff up to a minimal size
3404 * @skb: buffer to pad
3405 * @len: minimal length
3407 * Pads up a buffer to ensure the trailing bytes exist and are
3408 * blanked. If the buffer already contains sufficient data it
3409 * is untouched. Otherwise it is extended. Returns zero on
3410 * success. The skb is freed on error.
3412 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3414 unsigned int size = skb->len;
3415 if (likely(size >= len))
3417 return skb_pad(skb, len - size);
3421 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3422 * @skb: buffer to pad
3423 * @len: minimal length
3424 * @free_on_error: free buffer on error
3426 * Pads up a buffer to ensure the trailing bytes exist and are
3427 * blanked. If the buffer already contains sufficient data it
3428 * is untouched. Otherwise it is extended. Returns zero on
3429 * success. The skb is freed on error if @free_on_error is true.
3431 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3435 unsigned int size = skb->len;
3437 if (unlikely(size < len)) {
3439 if (__skb_pad(skb, len, free_on_error))
3441 __skb_put(skb, len);
3447 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3448 * @skb: buffer to pad
3449 * @len: minimal length
3451 * Pads up a buffer to ensure the trailing bytes exist and are
3452 * blanked. If the buffer already contains sufficient data it
3453 * is untouched. Otherwise it is extended. Returns zero on
3454 * success. The skb is freed on error.
3456 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3458 return __skb_put_padto(skb, len, true);
3461 static inline int skb_add_data(struct sk_buff *skb,
3462 struct iov_iter *from, int copy)
3464 const int off = skb->len;
3466 if (skb->ip_summed == CHECKSUM_NONE) {
3468 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3470 skb->csum = csum_block_add(skb->csum, csum, off);
3473 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3476 __skb_trim(skb, off);
3480 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3481 const struct page *page, int off)
3486 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3488 return page == skb_frag_page(frag) &&
3489 off == skb_frag_off(frag) + skb_frag_size(frag);
3494 static inline int __skb_linearize(struct sk_buff *skb)
3496 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3500 * skb_linearize - convert paged skb to linear one
3501 * @skb: buffer to linarize
3503 * If there is no free memory -ENOMEM is returned, otherwise zero
3504 * is returned and the old skb data released.
3506 static inline int skb_linearize(struct sk_buff *skb)
3508 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3512 * skb_has_shared_frag - can any frag be overwritten
3513 * @skb: buffer to test
3515 * Return true if the skb has at least one frag that might be modified
3516 * by an external entity (as in vmsplice()/sendfile())
3518 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3520 return skb_is_nonlinear(skb) &&
3521 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3525 * skb_linearize_cow - make sure skb is linear and writable
3526 * @skb: buffer to process
3528 * If there is no free memory -ENOMEM is returned, otherwise zero
3529 * is returned and the old skb data released.
3531 static inline int skb_linearize_cow(struct sk_buff *skb)
3533 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3534 __skb_linearize(skb) : 0;
3537 static __always_inline void
3538 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3541 if (skb->ip_summed == CHECKSUM_COMPLETE)
3542 skb->csum = csum_block_sub(skb->csum,
3543 csum_partial(start, len, 0), off);
3544 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3545 skb_checksum_start_offset(skb) < 0)
3546 skb->ip_summed = CHECKSUM_NONE;
3550 * skb_postpull_rcsum - update checksum for received skb after pull
3551 * @skb: buffer to update
3552 * @start: start of data before pull
3553 * @len: length of data pulled
3555 * After doing a pull on a received packet, you need to call this to
3556 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3557 * CHECKSUM_NONE so that it can be recomputed from scratch.
3559 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3560 const void *start, unsigned int len)
3562 __skb_postpull_rcsum(skb, start, len, 0);
3565 static __always_inline void
3566 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3569 if (skb->ip_summed == CHECKSUM_COMPLETE)
3570 skb->csum = csum_block_add(skb->csum,
3571 csum_partial(start, len, 0), off);
3575 * skb_postpush_rcsum - update checksum for received skb after push
3576 * @skb: buffer to update
3577 * @start: start of data after push
3578 * @len: length of data pushed
3580 * After doing a push on a received packet, you need to call this to
3581 * update the CHECKSUM_COMPLETE checksum.
3583 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3584 const void *start, unsigned int len)
3586 __skb_postpush_rcsum(skb, start, len, 0);
3589 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3592 * skb_push_rcsum - push skb and update receive checksum
3593 * @skb: buffer to update
3594 * @len: length of data pulled
3596 * This function performs an skb_push on the packet and updates
3597 * the CHECKSUM_COMPLETE checksum. It should be used on
3598 * receive path processing instead of skb_push unless you know
3599 * that the checksum difference is zero (e.g., a valid IP header)
3600 * or you are setting ip_summed to CHECKSUM_NONE.
3602 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3605 skb_postpush_rcsum(skb, skb->data, len);
3609 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3611 * pskb_trim_rcsum - trim received skb and update checksum
3612 * @skb: buffer to trim
3615 * This is exactly the same as pskb_trim except that it ensures the
3616 * checksum of received packets are still valid after the operation.
3617 * It can change skb pointers.
3620 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3622 if (likely(len >= skb->len))
3624 return pskb_trim_rcsum_slow(skb, len);
3627 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3629 if (skb->ip_summed == CHECKSUM_COMPLETE)
3630 skb->ip_summed = CHECKSUM_NONE;
3631 __skb_trim(skb, len);
3635 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3637 if (skb->ip_summed == CHECKSUM_COMPLETE)
3638 skb->ip_summed = CHECKSUM_NONE;
3639 return __skb_grow(skb, len);
3642 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3643 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3644 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3645 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3646 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3648 #define skb_queue_walk(queue, skb) \
3649 for (skb = (queue)->next; \
3650 skb != (struct sk_buff *)(queue); \
3653 #define skb_queue_walk_safe(queue, skb, tmp) \
3654 for (skb = (queue)->next, tmp = skb->next; \
3655 skb != (struct sk_buff *)(queue); \
3656 skb = tmp, tmp = skb->next)
3658 #define skb_queue_walk_from(queue, skb) \
3659 for (; skb != (struct sk_buff *)(queue); \
3662 #define skb_rbtree_walk(skb, root) \
3663 for (skb = skb_rb_first(root); skb != NULL; \
3664 skb = skb_rb_next(skb))
3666 #define skb_rbtree_walk_from(skb) \
3667 for (; skb != NULL; \
3668 skb = skb_rb_next(skb))
3670 #define skb_rbtree_walk_from_safe(skb, tmp) \
3671 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3674 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3675 for (tmp = skb->next; \
3676 skb != (struct sk_buff *)(queue); \
3677 skb = tmp, tmp = skb->next)
3679 #define skb_queue_reverse_walk(queue, skb) \
3680 for (skb = (queue)->prev; \
3681 skb != (struct sk_buff *)(queue); \
3684 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3685 for (skb = (queue)->prev, tmp = skb->prev; \
3686 skb != (struct sk_buff *)(queue); \
3687 skb = tmp, tmp = skb->prev)
3689 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3690 for (tmp = skb->prev; \
3691 skb != (struct sk_buff *)(queue); \
3692 skb = tmp, tmp = skb->prev)
3694 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3696 return skb_shinfo(skb)->frag_list != NULL;
3699 static inline void skb_frag_list_init(struct sk_buff *skb)
3701 skb_shinfo(skb)->frag_list = NULL;
3704 #define skb_walk_frags(skb, iter) \
3705 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3708 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3709 int *err, long *timeo_p,
3710 const struct sk_buff *skb);
3711 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3712 struct sk_buff_head *queue,
3715 struct sk_buff **last);
3716 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3717 struct sk_buff_head *queue,
3718 unsigned int flags, int *off, int *err,
3719 struct sk_buff **last);
3720 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3721 struct sk_buff_head *sk_queue,
3722 unsigned int flags, int *off, int *err);
3723 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3725 __poll_t datagram_poll(struct file *file, struct socket *sock,
3726 struct poll_table_struct *wait);
3727 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3728 struct iov_iter *to, int size);
3729 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3730 struct msghdr *msg, int size)
3732 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3734 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3735 struct msghdr *msg);
3736 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3737 struct iov_iter *to, int len,
3738 struct ahash_request *hash);
3739 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3740 struct iov_iter *from, int len);
3741 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3742 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3743 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3744 static inline void skb_free_datagram_locked(struct sock *sk,
3745 struct sk_buff *skb)
3747 __skb_free_datagram_locked(sk, skb, 0);
3749 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3750 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3751 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3752 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3754 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3755 struct pipe_inode_info *pipe, unsigned int len,
3756 unsigned int flags);
3757 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3759 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3760 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3761 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3762 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3764 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3765 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3766 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3767 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3768 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3769 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3770 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3771 unsigned int offset);
3772 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3773 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3774 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3775 int skb_vlan_pop(struct sk_buff *skb);
3776 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3777 int skb_eth_pop(struct sk_buff *skb);
3778 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3779 const unsigned char *src);
3780 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3781 int mac_len, bool ethernet);
3782 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3784 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3785 int skb_mpls_dec_ttl(struct sk_buff *skb);
3786 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3789 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3791 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3794 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3796 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3799 struct skb_checksum_ops {
3800 __wsum (*update)(const void *mem, int len, __wsum wsum);
3801 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3804 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3806 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3807 __wsum csum, const struct skb_checksum_ops *ops);
3808 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3811 static inline void * __must_check
3812 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3813 const void *data, int hlen, void *buffer)
3815 if (likely(hlen - offset >= len))
3816 return (void *)data + offset;
3818 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3824 static inline void * __must_check
3825 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3827 return __skb_header_pointer(skb, offset, len, skb->data,
3828 skb_headlen(skb), buffer);
3832 * skb_needs_linearize - check if we need to linearize a given skb
3833 * depending on the given device features.
3834 * @skb: socket buffer to check
3835 * @features: net device features
3837 * Returns true if either:
3838 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3839 * 2. skb is fragmented and the device does not support SG.
3841 static inline bool skb_needs_linearize(struct sk_buff *skb,
3842 netdev_features_t features)
3844 return skb_is_nonlinear(skb) &&
3845 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3846 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3849 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3851 const unsigned int len)
3853 memcpy(to, skb->data, len);
3856 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3857 const int offset, void *to,
3858 const unsigned int len)
3860 memcpy(to, skb->data + offset, len);
3863 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3865 const unsigned int len)
3867 memcpy(skb->data, from, len);
3870 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3873 const unsigned int len)
3875 memcpy(skb->data + offset, from, len);
3878 void skb_init(void);
3880 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3886 * skb_get_timestamp - get timestamp from a skb
3887 * @skb: skb to get stamp from
3888 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3890 * Timestamps are stored in the skb as offsets to a base timestamp.
3891 * This function converts the offset back to a struct timeval and stores
3894 static inline void skb_get_timestamp(const struct sk_buff *skb,
3895 struct __kernel_old_timeval *stamp)
3897 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3900 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3901 struct __kernel_sock_timeval *stamp)
3903 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3905 stamp->tv_sec = ts.tv_sec;
3906 stamp->tv_usec = ts.tv_nsec / 1000;
3909 static inline void skb_get_timestampns(const struct sk_buff *skb,
3910 struct __kernel_old_timespec *stamp)
3912 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3914 stamp->tv_sec = ts.tv_sec;
3915 stamp->tv_nsec = ts.tv_nsec;
3918 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3919 struct __kernel_timespec *stamp)
3921 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3923 stamp->tv_sec = ts.tv_sec;
3924 stamp->tv_nsec = ts.tv_nsec;
3927 static inline void __net_timestamp(struct sk_buff *skb)
3929 skb->tstamp = ktime_get_real();
3932 static inline ktime_t net_timedelta(ktime_t t)
3934 return ktime_sub(ktime_get_real(), t);
3937 static inline ktime_t net_invalid_timestamp(void)
3942 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3944 return skb_shinfo(skb)->meta_len;
3947 static inline void *skb_metadata_end(const struct sk_buff *skb)
3949 return skb_mac_header(skb);
3952 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3953 const struct sk_buff *skb_b,
3956 const void *a = skb_metadata_end(skb_a);
3957 const void *b = skb_metadata_end(skb_b);
3958 /* Using more efficient varaiant than plain call to memcmp(). */
3959 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3963 #define __it(x, op) (x -= sizeof(u##op))
3964 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3965 case 32: diffs |= __it_diff(a, b, 64);
3967 case 24: diffs |= __it_diff(a, b, 64);
3969 case 16: diffs |= __it_diff(a, b, 64);
3971 case 8: diffs |= __it_diff(a, b, 64);
3973 case 28: diffs |= __it_diff(a, b, 64);
3975 case 20: diffs |= __it_diff(a, b, 64);
3977 case 12: diffs |= __it_diff(a, b, 64);
3979 case 4: diffs |= __it_diff(a, b, 32);
3984 return memcmp(a - meta_len, b - meta_len, meta_len);
3988 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3989 const struct sk_buff *skb_b)
3991 u8 len_a = skb_metadata_len(skb_a);
3992 u8 len_b = skb_metadata_len(skb_b);
3994 if (!(len_a | len_b))
3997 return len_a != len_b ?
3998 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4001 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4003 skb_shinfo(skb)->meta_len = meta_len;
4006 static inline void skb_metadata_clear(struct sk_buff *skb)
4008 skb_metadata_set(skb, 0);
4011 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4013 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4015 void skb_clone_tx_timestamp(struct sk_buff *skb);
4016 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4018 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4020 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4024 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4029 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4032 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4034 * PHY drivers may accept clones of transmitted packets for
4035 * timestamping via their phy_driver.txtstamp method. These drivers
4036 * must call this function to return the skb back to the stack with a
4039 * @skb: clone of the original outgoing packet
4040 * @hwtstamps: hardware time stamps
4043 void skb_complete_tx_timestamp(struct sk_buff *skb,
4044 struct skb_shared_hwtstamps *hwtstamps);
4046 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4047 struct skb_shared_hwtstamps *hwtstamps,
4048 struct sock *sk, int tstype);
4051 * skb_tstamp_tx - queue clone of skb with send time stamps
4052 * @orig_skb: the original outgoing packet
4053 * @hwtstamps: hardware time stamps, may be NULL if not available
4055 * If the skb has a socket associated, then this function clones the
4056 * skb (thus sharing the actual data and optional structures), stores
4057 * the optional hardware time stamping information (if non NULL) or
4058 * generates a software time stamp (otherwise), then queues the clone
4059 * to the error queue of the socket. Errors are silently ignored.
4061 void skb_tstamp_tx(struct sk_buff *orig_skb,
4062 struct skb_shared_hwtstamps *hwtstamps);
4065 * skb_tx_timestamp() - Driver hook for transmit timestamping
4067 * Ethernet MAC Drivers should call this function in their hard_xmit()
4068 * function immediately before giving the sk_buff to the MAC hardware.
4070 * Specifically, one should make absolutely sure that this function is
4071 * called before TX completion of this packet can trigger. Otherwise
4072 * the packet could potentially already be freed.
4074 * @skb: A socket buffer.
4076 static inline void skb_tx_timestamp(struct sk_buff *skb)
4078 skb_clone_tx_timestamp(skb);
4079 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4080 skb_tstamp_tx(skb, NULL);
4084 * skb_complete_wifi_ack - deliver skb with wifi status
4086 * @skb: the original outgoing packet
4087 * @acked: ack status
4090 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4092 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4093 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4095 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4097 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4099 (skb->ip_summed == CHECKSUM_PARTIAL &&
4100 skb_checksum_start_offset(skb) >= 0));
4104 * skb_checksum_complete - Calculate checksum of an entire packet
4105 * @skb: packet to process
4107 * This function calculates the checksum over the entire packet plus
4108 * the value of skb->csum. The latter can be used to supply the
4109 * checksum of a pseudo header as used by TCP/UDP. It returns the
4112 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4113 * this function can be used to verify that checksum on received
4114 * packets. In that case the function should return zero if the
4115 * checksum is correct. In particular, this function will return zero
4116 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4117 * hardware has already verified the correctness of the checksum.
4119 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4121 return skb_csum_unnecessary(skb) ?
4122 0 : __skb_checksum_complete(skb);
4125 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4127 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4128 if (skb->csum_level == 0)
4129 skb->ip_summed = CHECKSUM_NONE;
4135 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4137 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4138 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4140 } else if (skb->ip_summed == CHECKSUM_NONE) {
4141 skb->ip_summed = CHECKSUM_UNNECESSARY;
4142 skb->csum_level = 0;
4146 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4148 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4149 skb->ip_summed = CHECKSUM_NONE;
4150 skb->csum_level = 0;
4154 /* Check if we need to perform checksum complete validation.
4156 * Returns true if checksum complete is needed, false otherwise
4157 * (either checksum is unnecessary or zero checksum is allowed).
4159 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4163 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4164 skb->csum_valid = 1;
4165 __skb_decr_checksum_unnecessary(skb);
4172 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4175 #define CHECKSUM_BREAK 76
4177 /* Unset checksum-complete
4179 * Unset checksum complete can be done when packet is being modified
4180 * (uncompressed for instance) and checksum-complete value is
4183 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4185 if (skb->ip_summed == CHECKSUM_COMPLETE)
4186 skb->ip_summed = CHECKSUM_NONE;
4189 /* Validate (init) checksum based on checksum complete.
4192 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4193 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4194 * checksum is stored in skb->csum for use in __skb_checksum_complete
4195 * non-zero: value of invalid checksum
4198 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4202 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4203 if (!csum_fold(csum_add(psum, skb->csum))) {
4204 skb->csum_valid = 1;
4211 if (complete || skb->len <= CHECKSUM_BREAK) {
4214 csum = __skb_checksum_complete(skb);
4215 skb->csum_valid = !csum;
4222 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4227 /* Perform checksum validate (init). Note that this is a macro since we only
4228 * want to calculate the pseudo header which is an input function if necessary.
4229 * First we try to validate without any computation (checksum unnecessary) and
4230 * then calculate based on checksum complete calling the function to compute
4234 * 0: checksum is validated or try to in skb_checksum_complete
4235 * non-zero: value of invalid checksum
4237 #define __skb_checksum_validate(skb, proto, complete, \
4238 zero_okay, check, compute_pseudo) \
4240 __sum16 __ret = 0; \
4241 skb->csum_valid = 0; \
4242 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4243 __ret = __skb_checksum_validate_complete(skb, \
4244 complete, compute_pseudo(skb, proto)); \
4248 #define skb_checksum_init(skb, proto, compute_pseudo) \
4249 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4251 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4252 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4254 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4255 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4257 #define skb_checksum_validate_zero_check(skb, proto, check, \
4259 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4261 #define skb_checksum_simple_validate(skb) \
4262 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4264 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4266 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4269 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4271 skb->csum = ~pseudo;
4272 skb->ip_summed = CHECKSUM_COMPLETE;
4275 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4277 if (__skb_checksum_convert_check(skb)) \
4278 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4281 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4282 u16 start, u16 offset)
4284 skb->ip_summed = CHECKSUM_PARTIAL;
4285 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4286 skb->csum_offset = offset - start;
4289 /* Update skbuf and packet to reflect the remote checksum offload operation.
4290 * When called, ptr indicates the starting point for skb->csum when
4291 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4292 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4294 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4295 int start, int offset, bool nopartial)
4300 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4304 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4305 __skb_checksum_complete(skb);
4306 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4309 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4311 /* Adjust skb->csum since we changed the packet */
4312 skb->csum = csum_add(skb->csum, delta);
4315 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4317 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4318 return (void *)(skb->_nfct & NFCT_PTRMASK);
4324 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4326 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4333 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4335 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4336 skb->slow_gro |= !!nfct;
4341 #ifdef CONFIG_SKB_EXTENSIONS
4343 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4349 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4352 #if IS_ENABLED(CONFIG_MPTCP)
4355 SKB_EXT_NUM, /* must be last */
4359 * struct skb_ext - sk_buff extensions
4360 * @refcnt: 1 on allocation, deallocated on 0
4361 * @offset: offset to add to @data to obtain extension address
4362 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4363 * @data: start of extension data, variable sized
4365 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4366 * to use 'u8' types while allowing up to 2kb worth of extension data.
4370 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4371 u8 chunks; /* same */
4372 char data[] __aligned(8);
4375 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4376 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4377 struct skb_ext *ext);
4378 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4379 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4380 void __skb_ext_put(struct skb_ext *ext);
4382 static inline void skb_ext_put(struct sk_buff *skb)
4384 if (skb->active_extensions)
4385 __skb_ext_put(skb->extensions);
4388 static inline void __skb_ext_copy(struct sk_buff *dst,
4389 const struct sk_buff *src)
4391 dst->active_extensions = src->active_extensions;
4393 if (src->active_extensions) {
4394 struct skb_ext *ext = src->extensions;
4396 refcount_inc(&ext->refcnt);
4397 dst->extensions = ext;
4401 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4404 __skb_ext_copy(dst, src);
4407 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4409 return !!ext->offset[i];
4412 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4414 return skb->active_extensions & (1 << id);
4417 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4419 if (skb_ext_exist(skb, id))
4420 __skb_ext_del(skb, id);
4423 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4425 if (skb_ext_exist(skb, id)) {
4426 struct skb_ext *ext = skb->extensions;
4428 return (void *)ext + (ext->offset[id] << 3);
4434 static inline void skb_ext_reset(struct sk_buff *skb)
4436 if (unlikely(skb->active_extensions)) {
4437 __skb_ext_put(skb->extensions);
4438 skb->active_extensions = 0;
4442 static inline bool skb_has_extensions(struct sk_buff *skb)
4444 return unlikely(skb->active_extensions);
4447 static inline void skb_ext_put(struct sk_buff *skb) {}
4448 static inline void skb_ext_reset(struct sk_buff *skb) {}
4449 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4450 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4451 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4452 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4453 #endif /* CONFIG_SKB_EXTENSIONS */
4455 static inline void nf_reset_ct(struct sk_buff *skb)
4457 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4458 nf_conntrack_put(skb_nfct(skb));
4463 static inline void nf_reset_trace(struct sk_buff *skb)
4465 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4470 static inline void ipvs_reset(struct sk_buff *skb)
4472 #if IS_ENABLED(CONFIG_IP_VS)
4473 skb->ipvs_property = 0;
4477 /* Note: This doesn't put any conntrack info in dst. */
4478 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4481 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4482 dst->_nfct = src->_nfct;
4483 nf_conntrack_get(skb_nfct(src));
4485 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4487 dst->nf_trace = src->nf_trace;
4491 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4493 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4494 nf_conntrack_put(skb_nfct(dst));
4496 dst->slow_gro = src->slow_gro;
4497 __nf_copy(dst, src, true);
4500 #ifdef CONFIG_NETWORK_SECMARK
4501 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4503 to->secmark = from->secmark;
4506 static inline void skb_init_secmark(struct sk_buff *skb)
4511 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4514 static inline void skb_init_secmark(struct sk_buff *skb)
4518 static inline int secpath_exists(const struct sk_buff *skb)
4521 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4527 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4529 return !skb->destructor &&
4530 !secpath_exists(skb) &&
4532 !skb->_skb_refdst &&
4533 !skb_has_frag_list(skb);
4536 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4538 skb->queue_mapping = queue_mapping;
4541 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4543 return skb->queue_mapping;
4546 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4548 to->queue_mapping = from->queue_mapping;
4551 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4553 skb->queue_mapping = rx_queue + 1;
4556 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4558 return skb->queue_mapping - 1;
4561 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4563 return skb->queue_mapping != 0;
4566 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4568 skb->dst_pending_confirm = val;
4571 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4573 return skb->dst_pending_confirm != 0;
4576 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4579 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4585 /* Keeps track of mac header offset relative to skb->head.
4586 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4587 * For non-tunnel skb it points to skb_mac_header() and for
4588 * tunnel skb it points to outer mac header.
4589 * Keeps track of level of encapsulation of network headers.
4600 #define SKB_GSO_CB_OFFSET 32
4601 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4603 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4605 return (skb_mac_header(inner_skb) - inner_skb->head) -
4606 SKB_GSO_CB(inner_skb)->mac_offset;
4609 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4611 int new_headroom, headroom;
4614 headroom = skb_headroom(skb);
4615 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4619 new_headroom = skb_headroom(skb);
4620 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4624 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4626 /* Do not update partial checksums if remote checksum is enabled. */
4627 if (skb->remcsum_offload)
4630 SKB_GSO_CB(skb)->csum = res;
4631 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4634 /* Compute the checksum for a gso segment. First compute the checksum value
4635 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4636 * then add in skb->csum (checksum from csum_start to end of packet).
4637 * skb->csum and csum_start are then updated to reflect the checksum of the
4638 * resultant packet starting from the transport header-- the resultant checksum
4639 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4642 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4644 unsigned char *csum_start = skb_transport_header(skb);
4645 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4646 __wsum partial = SKB_GSO_CB(skb)->csum;
4648 SKB_GSO_CB(skb)->csum = res;
4649 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4651 return csum_fold(csum_partial(csum_start, plen, partial));
4654 static inline bool skb_is_gso(const struct sk_buff *skb)
4656 return skb_shinfo(skb)->gso_size;
4659 /* Note: Should be called only if skb_is_gso(skb) is true */
4660 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4662 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4665 /* Note: Should be called only if skb_is_gso(skb) is true */
4666 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4668 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4671 /* Note: Should be called only if skb_is_gso(skb) is true */
4672 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4674 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4677 static inline void skb_gso_reset(struct sk_buff *skb)
4679 skb_shinfo(skb)->gso_size = 0;
4680 skb_shinfo(skb)->gso_segs = 0;
4681 skb_shinfo(skb)->gso_type = 0;
4684 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4687 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4689 shinfo->gso_size += increment;
4692 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4695 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4697 shinfo->gso_size -= decrement;
4700 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4702 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4704 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4705 * wanted then gso_type will be set. */
4706 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4708 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4709 unlikely(shinfo->gso_type == 0)) {
4710 __skb_warn_lro_forwarding(skb);
4716 static inline void skb_forward_csum(struct sk_buff *skb)
4718 /* Unfortunately we don't support this one. Any brave souls? */
4719 if (skb->ip_summed == CHECKSUM_COMPLETE)
4720 skb->ip_summed = CHECKSUM_NONE;
4724 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4725 * @skb: skb to check
4727 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4728 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4729 * use this helper, to document places where we make this assertion.
4731 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4734 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4738 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4740 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4741 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4742 unsigned int transport_len,
4743 __sum16(*skb_chkf)(struct sk_buff *skb));
4746 * skb_head_is_locked - Determine if the skb->head is locked down
4747 * @skb: skb to check
4749 * The head on skbs build around a head frag can be removed if they are
4750 * not cloned. This function returns true if the skb head is locked down
4751 * due to either being allocated via kmalloc, or by being a clone with
4752 * multiple references to the head.
4754 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4756 return !skb->head_frag || skb_cloned(skb);
4759 /* Local Checksum Offload.
4760 * Compute outer checksum based on the assumption that the
4761 * inner checksum will be offloaded later.
4762 * See Documentation/networking/checksum-offloads.rst for
4763 * explanation of how this works.
4764 * Fill in outer checksum adjustment (e.g. with sum of outer
4765 * pseudo-header) before calling.
4766 * Also ensure that inner checksum is in linear data area.
4768 static inline __wsum lco_csum(struct sk_buff *skb)
4770 unsigned char *csum_start = skb_checksum_start(skb);
4771 unsigned char *l4_hdr = skb_transport_header(skb);
4774 /* Start with complement of inner checksum adjustment */
4775 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4778 /* Add in checksum of our headers (incl. outer checksum
4779 * adjustment filled in by caller) and return result.
4781 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4784 static inline bool skb_is_redirected(const struct sk_buff *skb)
4786 return skb->redirected;
4789 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4791 skb->redirected = 1;
4792 #ifdef CONFIG_NET_REDIRECT
4793 skb->from_ingress = from_ingress;
4794 if (skb->from_ingress)
4799 static inline void skb_reset_redirect(struct sk_buff *skb)
4801 skb->redirected = 0;
4804 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4806 return skb->csum_not_inet;
4809 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4810 const u64 kcov_handle)
4813 skb->kcov_handle = kcov_handle;
4817 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4820 return skb->kcov_handle;
4826 #ifdef CONFIG_PAGE_POOL
4827 static inline void skb_mark_for_recycle(struct sk_buff *skb)
4829 skb->pp_recycle = 1;
4833 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4835 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4837 return page_pool_return_skb_page(virt_to_page(data));
4840 #endif /* __KERNEL__ */
4841 #endif /* _LINUX_SKBUFF_H */