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 <net/checksum.h>
27 #include <linux/rcupdate.h>
28 #include <linux/hrtimer.h>
29 #include <linux/dma-mapping.h>
30 #include <linux/netdev_features.h>
31 #include <net/flow_dissector.h>
32 #include <linux/in6.h>
33 #include <linux/if_packet.h>
34 #include <linux/llist.h>
36 #include <net/page_pool.h>
37 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
38 #include <linux/netfilter/nf_conntrack_common.h>
40 #include <net/net_debug.h>
41 #include <net/dropreason.h>
46 * The interface for checksum offload between the stack and networking drivers
49 * IP checksum related features
50 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
52 * Drivers advertise checksum offload capabilities in the features of a device.
53 * From the stack's point of view these are capabilities offered by the driver.
54 * A driver typically only advertises features that it is capable of offloading
57 * .. flat-table:: Checksum related device features
60 * * - %NETIF_F_HW_CSUM
61 * - The driver (or its device) is able to compute one
62 * IP (one's complement) checksum for any combination
63 * of protocols or protocol layering. The checksum is
64 * computed and set in a packet per the CHECKSUM_PARTIAL
65 * interface (see below).
67 * * - %NETIF_F_IP_CSUM
68 * - Driver (device) is only able to checksum plain
69 * TCP or UDP packets over IPv4. These are specifically
70 * unencapsulated packets of the form IPv4|TCP or
71 * IPv4|UDP where the Protocol field in the IPv4 header
72 * is TCP or UDP. The IPv4 header may contain IP options.
73 * This feature cannot be set in features for a device
74 * with NETIF_F_HW_CSUM also set. This feature is being
75 * DEPRECATED (see below).
77 * * - %NETIF_F_IPV6_CSUM
78 * - Driver (device) is only able to checksum plain
79 * TCP or UDP packets over IPv6. These are specifically
80 * unencapsulated packets of the form IPv6|TCP or
81 * IPv6|UDP where the Next Header field in the IPv6
82 * header is either TCP or UDP. IPv6 extension headers
83 * are not supported with this feature. This feature
84 * cannot be set in features for a device with
85 * NETIF_F_HW_CSUM also set. This feature is being
86 * DEPRECATED (see below).
89 * - Driver (device) performs receive checksum offload.
90 * This flag is only used to disable the RX checksum
91 * feature for a device. The stack will accept receive
92 * checksum indication in packets received on a device
93 * regardless of whether NETIF_F_RXCSUM is set.
95 * Checksumming of received packets by device
96 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
98 * Indication of checksum verification is set in &sk_buff.ip_summed.
99 * Possible values are:
103 * Device did not checksum this packet e.g. due to lack of capabilities.
104 * The packet contains full (though not verified) checksum in packet but
105 * not in skb->csum. Thus, skb->csum is undefined in this case.
107 * - %CHECKSUM_UNNECESSARY
109 * The hardware you're dealing with doesn't calculate the full checksum
110 * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
111 * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
112 * if their checksums are okay. &sk_buff.csum is still undefined in this case
113 * though. A driver or device must never modify the checksum field in the
114 * packet even if checksum is verified.
116 * %CHECKSUM_UNNECESSARY is applicable to following protocols:
118 * - TCP: IPv6 and IPv4.
119 * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
120 * zero UDP checksum for either IPv4 or IPv6, the networking stack
121 * may perform further validation in this case.
122 * - GRE: only if the checksum is present in the header.
123 * - SCTP: indicates the CRC in SCTP header has been validated.
124 * - FCOE: indicates the CRC in FC frame has been validated.
126 * &sk_buff.csum_level indicates the number of consecutive checksums found in
127 * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
128 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
129 * and a device is able to verify the checksums for UDP (possibly zero),
130 * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
131 * two. If the device were only able to verify the UDP checksum and not
132 * GRE, either because it doesn't support GRE checksum or because GRE
133 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
134 * not considered in this case).
136 * - %CHECKSUM_COMPLETE
138 * This is the most generic way. The device supplied checksum of the _whole_
139 * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
140 * hardware doesn't need to parse L3/L4 headers to implement this.
144 * - Even if device supports only some protocols, but is able to produce
145 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
146 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
148 * - %CHECKSUM_PARTIAL
150 * A checksum is set up to be offloaded to a device as described in the
151 * output description for CHECKSUM_PARTIAL. This may occur on a packet
152 * received directly from another Linux OS, e.g., a virtualized Linux kernel
153 * on the same host, or it may be set in the input path in GRO or remote
154 * checksum offload. For the purposes of checksum verification, the checksum
155 * referred to by skb->csum_start + skb->csum_offset and any preceding
156 * checksums in the packet are considered verified. Any checksums in the
157 * packet that are after the checksum being offloaded are not considered to
160 * Checksumming on transmit for non-GSO
161 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
163 * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
166 * - %CHECKSUM_PARTIAL
168 * The driver is required to checksum the packet as seen by hard_start_xmit()
169 * from &sk_buff.csum_start up to the end, and to record/write the checksum at
170 * offset &sk_buff.csum_start + &sk_buff.csum_offset.
171 * A driver may verify that the
172 * csum_start and csum_offset values are valid values given the length and
173 * offset of the packet, but it should not attempt to validate that the
174 * checksum refers to a legitimate transport layer checksum -- it is the
175 * purview of the stack to validate that csum_start and csum_offset are set
178 * When the stack requests checksum offload for a packet, the driver MUST
179 * ensure that the checksum is set correctly. A driver can either offload the
180 * checksum calculation to the device, or call skb_checksum_help (in the case
181 * that the device does not support offload for a particular checksum).
183 * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
184 * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
185 * checksum offload capability.
186 * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
187 * on network device checksumming capabilities: if a packet does not match
188 * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
189 * &sk_buff.csum_not_inet, see :ref:`crc`)
190 * is called to resolve the checksum.
194 * The skb was already checksummed by the protocol, or a checksum is not
197 * - %CHECKSUM_UNNECESSARY
199 * This has the same meaning as CHECKSUM_NONE for checksum offload on
202 * - %CHECKSUM_COMPLETE
204 * Not used in checksum output. If a driver observes a packet with this value
205 * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
209 * Non-IP checksum (CRC) offloads
210 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
215 * * - %NETIF_F_SCTP_CRC
216 * - This feature indicates that a device is capable of
217 * offloading the SCTP CRC in a packet. To perform this offload the stack
218 * will set csum_start and csum_offset accordingly, set ip_summed to
219 * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
220 * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
221 * A driver that supports both IP checksum offload and SCTP CRC32c offload
222 * must verify which offload is configured for a packet by testing the
223 * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
224 * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
226 * * - %NETIF_F_FCOE_CRC
227 * - This feature indicates that a device is capable of offloading the FCOE
228 * CRC in a packet. To perform this offload the stack will set ip_summed
229 * to %CHECKSUM_PARTIAL and set csum_start and csum_offset
230 * accordingly. Note that there is no indication in the skbuff that the
231 * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
232 * both IP checksum offload and FCOE CRC offload must verify which offload
233 * is configured for a packet, presumably by inspecting packet headers.
235 * Checksumming on output with GSO
236 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
238 * In the case of a GSO packet (skb_is_gso() is true), checksum offload
239 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
240 * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
241 * part of the GSO operation is implied. If a checksum is being offloaded
242 * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
243 * csum_offset are set to refer to the outermost checksum being offloaded
244 * (two offloaded checksums are possible with UDP encapsulation).
247 /* Don't change this without changing skb_csum_unnecessary! */
248 #define CHECKSUM_NONE 0
249 #define CHECKSUM_UNNECESSARY 1
250 #define CHECKSUM_COMPLETE 2
251 #define CHECKSUM_PARTIAL 3
253 /* Maximum value in skb->csum_level */
254 #define SKB_MAX_CSUM_LEVEL 3
256 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
257 #define SKB_WITH_OVERHEAD(X) \
258 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
259 #define SKB_MAX_ORDER(X, ORDER) \
260 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
261 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
262 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
264 /* return minimum truesize of one skb containing X bytes of data */
265 #define SKB_TRUESIZE(X) ((X) + \
266 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
267 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
269 struct ahash_request;
272 struct pipe_inode_info;
280 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
281 struct nf_bridge_info {
283 BRNF_PROTO_UNCHANGED,
291 struct net_device *physindev;
293 /* always valid & non-NULL from FORWARD on, for physdev match */
294 struct net_device *physoutdev;
296 /* prerouting: detect dnat in orig/reply direction */
298 struct in6_addr ipv6_daddr;
300 /* after prerouting + nat detected: store original source
301 * mac since neigh resolution overwrites it, only used while
302 * skb is out in neigh layer.
304 char neigh_header[8];
309 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
310 /* Chain in tc_skb_ext will be used to share the tc chain with
311 * ovs recirc_id. It will be set to the current chain by tc
312 * and read by ovs to recirc_id.
324 struct sk_buff_head {
325 /* These two members must be first to match sk_buff. */
326 struct_group_tagged(sk_buff_list, list,
327 struct sk_buff *next;
328 struct sk_buff *prev;
337 /* To allow 64K frame to be packed as single skb without frag_list we
338 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
339 * buffers which do not start on a page boundary.
341 * Since GRO uses frags we allocate at least 16 regardless of page
344 #if (65536/PAGE_SIZE + 1) < 16
345 #define MAX_SKB_FRAGS 16UL
347 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
349 extern int sysctl_max_skb_frags;
351 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
352 * segment using its current segmentation instead.
354 #define GSO_BY_FRAGS 0xFFFF
356 typedef struct bio_vec skb_frag_t;
359 * skb_frag_size() - Returns the size of a skb fragment
360 * @frag: skb fragment
362 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
368 * skb_frag_size_set() - Sets the size of a skb fragment
369 * @frag: skb fragment
370 * @size: size of fragment
372 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
378 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
379 * @frag: skb fragment
380 * @delta: value to add
382 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
384 frag->bv_len += delta;
388 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
389 * @frag: skb fragment
390 * @delta: value to subtract
392 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
394 frag->bv_len -= delta;
398 * skb_frag_must_loop - Test if %p is a high memory page
399 * @p: fragment's page
401 static inline bool skb_frag_must_loop(struct page *p)
403 #if defined(CONFIG_HIGHMEM)
404 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
411 * skb_frag_foreach_page - loop over pages in a fragment
413 * @f: skb frag to operate on
414 * @f_off: offset from start of f->bv_page
415 * @f_len: length from f_off to loop over
416 * @p: (temp var) current page
417 * @p_off: (temp var) offset from start of current page,
418 * non-zero only on first page.
419 * @p_len: (temp var) length in current page,
420 * < PAGE_SIZE only on first and last page.
421 * @copied: (temp var) length so far, excluding current p_len.
423 * A fragment can hold a compound page, in which case per-page
424 * operations, notably kmap_atomic, must be called for each
427 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
428 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
429 p_off = (f_off) & (PAGE_SIZE - 1), \
430 p_len = skb_frag_must_loop(p) ? \
431 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
434 copied += p_len, p++, p_off = 0, \
435 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
437 #define HAVE_HW_TIME_STAMP
440 * struct skb_shared_hwtstamps - hardware time stamps
441 * @hwtstamp: hardware time stamp transformed into duration
442 * since arbitrary point in time
443 * @netdev_data: address/cookie of network device driver used as
444 * reference to actual hardware time stamp
446 * Software time stamps generated by ktime_get_real() are stored in
449 * hwtstamps can only be compared against other hwtstamps from
452 * This structure is attached to packets as part of the
453 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
455 struct skb_shared_hwtstamps {
462 /* Definitions for tx_flags in struct skb_shared_info */
464 /* generate hardware time stamp */
465 SKBTX_HW_TSTAMP = 1 << 0,
467 /* generate software time stamp when queueing packet to NIC */
468 SKBTX_SW_TSTAMP = 1 << 1,
470 /* device driver is going to provide hardware time stamp */
471 SKBTX_IN_PROGRESS = 1 << 2,
473 /* generate hardware time stamp based on cycles if supported */
474 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3,
476 /* generate wifi status information (where possible) */
477 SKBTX_WIFI_STATUS = 1 << 4,
479 /* determine hardware time stamp based on time or cycles */
480 SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
482 /* generate software time stamp when entering packet scheduling */
483 SKBTX_SCHED_TSTAMP = 1 << 6,
486 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
488 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \
489 SKBTX_HW_TSTAMP_USE_CYCLES | \
492 /* Definitions for flags in struct skb_shared_info */
494 /* use zcopy routines */
495 SKBFL_ZEROCOPY_ENABLE = BIT(0),
497 /* This indicates at least one fragment might be overwritten
498 * (as in vmsplice(), sendfile() ...)
499 * If we need to compute a TX checksum, we'll need to copy
500 * all frags to avoid possible bad checksum
502 SKBFL_SHARED_FRAG = BIT(1),
504 /* segment contains only zerocopy data and should not be
505 * charged to the kernel memory.
507 SKBFL_PURE_ZEROCOPY = BIT(2),
509 SKBFL_DONT_ORPHAN = BIT(3),
511 /* page references are managed by the ubuf_info, so it's safe to
512 * use frags only up until ubuf_info is released
514 SKBFL_MANAGED_FRAG_REFS = BIT(4),
517 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
518 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \
519 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS)
522 * The callback notifies userspace to release buffers when skb DMA is done in
523 * lower device, the skb last reference should be 0 when calling this.
524 * The zerocopy_success argument is true if zero copy transmit occurred,
525 * false on data copy or out of memory error caused by data copy attempt.
526 * The ctx field is used to track device context.
527 * The desc field is used to track userspace buffer index.
530 void (*callback)(struct sk_buff *, struct ubuf_info *,
531 bool zerocopy_success);
536 struct ubuf_info_msgzc {
537 struct ubuf_info ubuf;
553 struct user_struct *user;
558 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
559 #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \
562 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
563 void mm_unaccount_pinned_pages(struct mmpin *mmp);
565 /* This data is invariant across clones and lives at
566 * the end of the header data, ie. at skb->end.
568 struct skb_shared_info {
573 unsigned short gso_size;
574 /* Warning: this field is not always filled in (UFO)! */
575 unsigned short gso_segs;
576 struct sk_buff *frag_list;
577 struct skb_shared_hwtstamps hwtstamps;
578 unsigned int gso_type;
582 * Warning : all fields before dataref are cleared in __alloc_skb()
585 unsigned int xdp_frags_size;
587 /* Intermediate layers must ensure that destructor_arg
588 * remains valid until skb destructor */
589 void * destructor_arg;
591 /* must be last field, see pskb_expand_head() */
592 skb_frag_t frags[MAX_SKB_FRAGS];
596 * DOC: dataref and headerless skbs
598 * Transport layers send out clones of payload skbs they hold for
599 * retransmissions. To allow lower layers of the stack to prepend their headers
600 * we split &skb_shared_info.dataref into two halves.
601 * The lower 16 bits count the overall number of references.
602 * The higher 16 bits indicate how many of the references are payload-only.
603 * skb_header_cloned() checks if skb is allowed to add / write the headers.
605 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
606 * (via __skb_header_release()). Any clone created from marked skb will get
607 * &sk_buff.hdr_len populated with the available headroom.
608 * If there's the only clone in existence it's able to modify the headroom
609 * at will. The sequence of calls inside the transport layer is::
613 * __skb_header_release()
615 * // send the clone down the stack
617 * This is not a very generic construct and it depends on the transport layers
618 * doing the right thing. In practice there's usually only one payload-only skb.
619 * Having multiple payload-only skbs with different lengths of hdr_len is not
620 * possible. The payload-only skbs should never leave their owner.
622 #define SKB_DATAREF_SHIFT 16
623 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
627 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
628 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
629 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
633 SKB_GSO_TCPV4 = 1 << 0,
635 /* This indicates the skb is from an untrusted source. */
636 SKB_GSO_DODGY = 1 << 1,
638 /* This indicates the tcp segment has CWR set. */
639 SKB_GSO_TCP_ECN = 1 << 2,
641 SKB_GSO_TCP_FIXEDID = 1 << 3,
643 SKB_GSO_TCPV6 = 1 << 4,
645 SKB_GSO_FCOE = 1 << 5,
647 SKB_GSO_GRE = 1 << 6,
649 SKB_GSO_GRE_CSUM = 1 << 7,
651 SKB_GSO_IPXIP4 = 1 << 8,
653 SKB_GSO_IPXIP6 = 1 << 9,
655 SKB_GSO_UDP_TUNNEL = 1 << 10,
657 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
659 SKB_GSO_PARTIAL = 1 << 12,
661 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
663 SKB_GSO_SCTP = 1 << 14,
665 SKB_GSO_ESP = 1 << 15,
667 SKB_GSO_UDP = 1 << 16,
669 SKB_GSO_UDP_L4 = 1 << 17,
671 SKB_GSO_FRAGLIST = 1 << 18,
674 #if BITS_PER_LONG > 32
675 #define NET_SKBUFF_DATA_USES_OFFSET 1
678 #ifdef NET_SKBUFF_DATA_USES_OFFSET
679 typedef unsigned int sk_buff_data_t;
681 typedef unsigned char *sk_buff_data_t;
685 * DOC: Basic sk_buff geometry
687 * struct sk_buff itself is a metadata structure and does not hold any packet
688 * data. All the data is held in associated buffers.
690 * &sk_buff.head points to the main "head" buffer. The head buffer is divided
693 * - data buffer, containing headers and sometimes payload;
694 * this is the part of the skb operated on by the common helpers
695 * such as skb_put() or skb_pull();
696 * - shared info (struct skb_shared_info) which holds an array of pointers
697 * to read-only data in the (page, offset, length) format.
699 * Optionally &skb_shared_info.frag_list may point to another skb.
701 * Basic diagram may look like this::
706 * ,--------------------------- + head
707 * / ,----------------- + data
708 * / / ,----------- + tail
712 * -----------------------------------------------
713 * | headroom | data | tailroom | skb_shared_info |
714 * -----------------------------------------------
718 * + [page frag] ---------
719 * + frag_list --> | sk_buff |
725 * struct sk_buff - socket buffer
726 * @next: Next buffer in list
727 * @prev: Previous buffer in list
728 * @tstamp: Time we arrived/left
729 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
730 * for retransmit timer
731 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
733 * @ll_node: anchor in an llist (eg socket defer_list)
734 * @sk: Socket we are owned by
735 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
736 * fragmentation management
737 * @dev: Device we arrived on/are leaving by
738 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
739 * @cb: Control buffer. Free for use by every layer. Put private vars here
740 * @_skb_refdst: destination entry (with norefcount bit)
741 * @sp: the security path, used for xfrm
742 * @len: Length of actual data
743 * @data_len: Data length
744 * @mac_len: Length of link layer header
745 * @hdr_len: writable header length of cloned skb
746 * @csum: Checksum (must include start/offset pair)
747 * @csum_start: Offset from skb->head where checksumming should start
748 * @csum_offset: Offset from csum_start where checksum should be stored
749 * @priority: Packet queueing priority
750 * @ignore_df: allow local fragmentation
751 * @cloned: Head may be cloned (check refcnt to be sure)
752 * @ip_summed: Driver fed us an IP checksum
753 * @nohdr: Payload reference only, must not modify header
754 * @pkt_type: Packet class
755 * @fclone: skbuff clone status
756 * @ipvs_property: skbuff is owned by ipvs
757 * @inner_protocol_type: whether the inner protocol is
758 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
759 * @remcsum_offload: remote checksum offload is enabled
760 * @offload_fwd_mark: Packet was L2-forwarded in hardware
761 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
762 * @tc_skip_classify: do not classify packet. set by IFB device
763 * @tc_at_ingress: used within tc_classify to distinguish in/egress
764 * @redirected: packet was redirected by packet classifier
765 * @from_ingress: packet was redirected from the ingress path
766 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
767 * @peeked: this packet has been seen already, so stats have been
768 * done for it, don't do them again
769 * @nf_trace: netfilter packet trace flag
770 * @protocol: Packet protocol from driver
771 * @destructor: Destruct function
772 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
773 * @_sk_redir: socket redirection information for skmsg
774 * @_nfct: Associated connection, if any (with nfctinfo bits)
775 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
776 * @skb_iif: ifindex of device we arrived on
777 * @tc_index: Traffic control index
778 * @hash: the packet hash
779 * @queue_mapping: Queue mapping for multiqueue devices
780 * @head_frag: skb was allocated from page fragments,
781 * not allocated by kmalloc() or vmalloc().
782 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
783 * @pp_recycle: mark the packet for recycling instead of freeing (implies
784 * page_pool support on driver)
785 * @active_extensions: active extensions (skb_ext_id types)
786 * @ndisc_nodetype: router type (from link layer)
787 * @ooo_okay: allow the mapping of a socket to a queue to be changed
788 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
790 * @sw_hash: indicates hash was computed in software stack
791 * @wifi_acked_valid: wifi_acked was set
792 * @wifi_acked: whether frame was acked on wifi or not
793 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
794 * @encapsulation: indicates the inner headers in the skbuff are valid
795 * @encap_hdr_csum: software checksum is needed
796 * @csum_valid: checksum is already valid
797 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
798 * @csum_complete_sw: checksum was completed by software
799 * @csum_level: indicates the number of consecutive checksums found in
800 * the packet minus one that have been verified as
801 * CHECKSUM_UNNECESSARY (max 3)
802 * @scm_io_uring: SKB holds io_uring registered files
803 * @dst_pending_confirm: need to confirm neighbour
804 * @decrypted: Decrypted SKB
805 * @slow_gro: state present at GRO time, slower prepare step required
806 * @mono_delivery_time: When set, skb->tstamp has the
807 * delivery_time in mono clock base (i.e. EDT). Otherwise, the
808 * skb->tstamp has the (rcv) timestamp at ingress and
809 * delivery_time at egress.
810 * @napi_id: id of the NAPI struct this skb came from
811 * @sender_cpu: (aka @napi_id) source CPU in XPS
812 * @alloc_cpu: CPU which did the skb allocation.
813 * @secmark: security marking
814 * @mark: Generic packet mark
815 * @reserved_tailroom: (aka @mark) number of bytes of free space available
816 * at the tail of an sk_buff
817 * @vlan_all: vlan fields (proto & tci)
818 * @vlan_proto: vlan encapsulation protocol
819 * @vlan_tci: vlan tag control information
820 * @inner_protocol: Protocol (encapsulation)
821 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
822 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
823 * @inner_transport_header: Inner transport layer header (encapsulation)
824 * @inner_network_header: Network layer header (encapsulation)
825 * @inner_mac_header: Link layer header (encapsulation)
826 * @transport_header: Transport layer header
827 * @network_header: Network layer header
828 * @mac_header: Link layer header
829 * @kcov_handle: KCOV remote handle for remote coverage collection
830 * @tail: Tail pointer
832 * @head: Head of buffer
833 * @data: Data head pointer
834 * @truesize: Buffer size
835 * @users: User count - see {datagram,tcp}.c
836 * @extensions: allocated extensions, valid if active_extensions is nonzero
842 /* These two members must be first to match sk_buff_head. */
843 struct sk_buff *next;
844 struct sk_buff *prev;
847 struct net_device *dev;
848 /* Some protocols might use this space to store information,
849 * while device pointer would be NULL.
850 * UDP receive path is one user.
852 unsigned long dev_scratch;
855 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
856 struct list_head list;
857 struct llist_node ll_node;
862 int ip_defrag_offset;
867 u64 skb_mstamp_ns; /* earliest departure time */
870 * This is the control buffer. It is free to use for every
871 * layer. Please put your private variables there. If you
872 * want to keep them across layers you have to do a skb_clone()
873 * first. This is owned by whoever has the skb queued ATM.
875 char cb[48] __aligned(8);
879 unsigned long _skb_refdst;
880 void (*destructor)(struct sk_buff *skb);
882 struct list_head tcp_tsorted_anchor;
883 #ifdef CONFIG_NET_SOCK_MSG
884 unsigned long _sk_redir;
888 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
896 /* Following fields are _not_ copied in __copy_skb_header()
897 * Note that queue_mapping is here mostly to fill a hole.
901 /* if you move cloned around you also must adapt those constants */
902 #ifdef __BIG_ENDIAN_BITFIELD
903 #define CLONED_MASK (1 << 7)
905 #define CLONED_MASK 1
907 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
910 __u8 __cloned_offset[0];
918 pp_recycle:1; /* page_pool recycle indicator */
919 #ifdef CONFIG_SKB_EXTENSIONS
920 __u8 active_extensions;
923 /* Fields enclosed in headers group are copied
924 * using a single memcpy() in __copy_skb_header()
926 struct_group(headers,
929 __u8 __pkt_type_offset[0];
931 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
939 __u8 wifi_acked_valid:1;
942 /* Indicates the inner headers are valid in the skbuff. */
943 __u8 encapsulation:1;
944 __u8 encap_hdr_csum:1;
948 __u8 __pkt_vlan_present_offset[0];
950 __u8 remcsum_offload:1;
951 __u8 csum_complete_sw:1;
953 __u8 dst_pending_confirm:1;
954 __u8 mono_delivery_time:1; /* See SKB_MONO_DELIVERY_TIME_MASK */
955 #ifdef CONFIG_NET_CLS_ACT
956 __u8 tc_skip_classify:1;
957 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */
959 #ifdef CONFIG_IPV6_NDISC_NODETYPE
960 __u8 ndisc_nodetype:2;
963 __u8 ipvs_property:1;
964 __u8 inner_protocol_type:1;
965 #ifdef CONFIG_NET_SWITCHDEV
966 __u8 offload_fwd_mark:1;
967 __u8 offload_l3_fwd_mark:1;
970 #ifdef CONFIG_NET_REDIRECT
973 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
974 __u8 nf_skip_egress:1;
976 #ifdef CONFIG_TLS_DEVICE
980 __u8 csum_not_inet:1;
983 #ifdef CONFIG_NET_SCHED
984 __u16 tc_index; /* traffic control index */
1004 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1006 unsigned int napi_id;
1007 unsigned int sender_cpu;
1011 #ifdef CONFIG_NETWORK_SECMARK
1017 __u32 reserved_tailroom;
1021 __be16 inner_protocol;
1025 __u16 inner_transport_header;
1026 __u16 inner_network_header;
1027 __u16 inner_mac_header;
1030 __u16 transport_header;
1031 __u16 network_header;
1038 ); /* end headers group */
1040 /* These elements must be at the end, see alloc_skb() for details. */
1041 sk_buff_data_t tail;
1043 unsigned char *head,
1045 unsigned int truesize;
1048 #ifdef CONFIG_SKB_EXTENSIONS
1049 /* only useable after checking ->active_extensions != 0 */
1050 struct skb_ext *extensions;
1054 /* if you move pkt_type around you also must adapt those constants */
1055 #ifdef __BIG_ENDIAN_BITFIELD
1056 #define PKT_TYPE_MAX (7 << 5)
1058 #define PKT_TYPE_MAX 7
1060 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset)
1062 /* if you move tc_at_ingress or mono_delivery_time
1063 * around, you also must adapt these constants.
1065 #ifdef __BIG_ENDIAN_BITFIELD
1066 #define TC_AT_INGRESS_MASK (1 << 0)
1067 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 2)
1069 #define TC_AT_INGRESS_MASK (1 << 7)
1070 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 5)
1072 #define PKT_VLAN_PRESENT_OFFSET offsetof(struct sk_buff, __pkt_vlan_present_offset)
1076 * Handling routines are only of interest to the kernel
1079 #define SKB_ALLOC_FCLONE 0x01
1080 #define SKB_ALLOC_RX 0x02
1081 #define SKB_ALLOC_NAPI 0x04
1084 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1087 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1089 return unlikely(skb->pfmemalloc);
1093 * skb might have a dst pointer attached, refcounted or not.
1094 * _skb_refdst low order bit is set if refcount was _not_ taken
1096 #define SKB_DST_NOREF 1UL
1097 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1100 * skb_dst - returns skb dst_entry
1103 * Returns skb dst_entry, regardless of reference taken or not.
1105 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1107 /* If refdst was not refcounted, check we still are in a
1108 * rcu_read_lock section
1110 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1111 !rcu_read_lock_held() &&
1112 !rcu_read_lock_bh_held());
1113 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1117 * skb_dst_set - sets skb dst
1121 * Sets skb dst, assuming a reference was taken on dst and should
1122 * be released by skb_dst_drop()
1124 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1126 skb->slow_gro |= !!dst;
1127 skb->_skb_refdst = (unsigned long)dst;
1131 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1135 * Sets skb dst, assuming a reference was not taken on dst.
1136 * If dst entry is cached, we do not take reference and dst_release
1137 * will be avoided by refdst_drop. If dst entry is not cached, we take
1138 * reference, so that last dst_release can destroy the dst immediately.
1140 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1142 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1143 skb->slow_gro |= !!dst;
1144 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1148 * skb_dst_is_noref - Test if skb dst isn't refcounted
1151 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1153 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1157 * skb_rtable - Returns the skb &rtable
1160 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1162 return (struct rtable *)skb_dst(skb);
1165 /* For mangling skb->pkt_type from user space side from applications
1166 * such as nft, tc, etc, we only allow a conservative subset of
1167 * possible pkt_types to be set.
1169 static inline bool skb_pkt_type_ok(u32 ptype)
1171 return ptype <= PACKET_OTHERHOST;
1175 * skb_napi_id - Returns the skb's NAPI id
1178 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1180 #ifdef CONFIG_NET_RX_BUSY_POLL
1181 return skb->napi_id;
1188 * skb_unref - decrement the skb's reference count
1191 * Returns true if we can free the skb.
1193 static inline bool skb_unref(struct sk_buff *skb)
1197 if (likely(refcount_read(&skb->users) == 1))
1199 else if (likely(!refcount_dec_and_test(&skb->users)))
1206 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1209 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1210 * @skb: buffer to free
1212 static inline void kfree_skb(struct sk_buff *skb)
1214 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1217 void skb_release_head_state(struct sk_buff *skb);
1218 void kfree_skb_list_reason(struct sk_buff *segs,
1219 enum skb_drop_reason reason);
1220 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1221 void skb_tx_error(struct sk_buff *skb);
1223 static inline void kfree_skb_list(struct sk_buff *segs)
1225 kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED);
1228 #ifdef CONFIG_TRACEPOINTS
1229 void consume_skb(struct sk_buff *skb);
1231 static inline void consume_skb(struct sk_buff *skb)
1233 return kfree_skb(skb);
1237 void __consume_stateless_skb(struct sk_buff *skb);
1238 void __kfree_skb(struct sk_buff *skb);
1239 extern struct kmem_cache *skbuff_head_cache;
1241 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1242 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1243 bool *fragstolen, int *delta_truesize);
1245 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1247 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1248 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1249 struct sk_buff *build_skb_around(struct sk_buff *skb,
1250 void *data, unsigned int frag_size);
1251 void skb_attempt_defer_free(struct sk_buff *skb);
1253 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1254 struct sk_buff *slab_build_skb(void *data);
1257 * alloc_skb - allocate a network buffer
1258 * @size: size to allocate
1259 * @priority: allocation mask
1261 * This function is a convenient wrapper around __alloc_skb().
1263 static inline struct sk_buff *alloc_skb(unsigned int size,
1266 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1269 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1270 unsigned long data_len,
1274 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1276 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1277 struct sk_buff_fclones {
1278 struct sk_buff skb1;
1280 struct sk_buff skb2;
1282 refcount_t fclone_ref;
1286 * skb_fclone_busy - check if fclone is busy
1290 * Returns true if skb is a fast clone, and its clone is not freed.
1291 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1292 * so we also check that this didnt happen.
1294 static inline bool skb_fclone_busy(const struct sock *sk,
1295 const struct sk_buff *skb)
1297 const struct sk_buff_fclones *fclones;
1299 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1301 return skb->fclone == SKB_FCLONE_ORIG &&
1302 refcount_read(&fclones->fclone_ref) > 1 &&
1303 READ_ONCE(fclones->skb2.sk) == sk;
1307 * alloc_skb_fclone - allocate a network buffer from fclone cache
1308 * @size: size to allocate
1309 * @priority: allocation mask
1311 * This function is a convenient wrapper around __alloc_skb().
1313 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1316 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1319 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1320 void skb_headers_offset_update(struct sk_buff *skb, int off);
1321 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1322 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1323 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1324 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1325 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1326 gfp_t gfp_mask, bool fclone);
1327 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1330 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1333 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1334 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1335 unsigned int headroom);
1336 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1337 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1338 int newtailroom, gfp_t priority);
1339 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1340 int offset, int len);
1341 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1342 int offset, int len);
1343 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1344 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1347 * skb_pad - zero pad the tail of an skb
1348 * @skb: buffer to pad
1349 * @pad: space to pad
1351 * Ensure that a buffer is followed by a padding area that is zero
1352 * filled. Used by network drivers which may DMA or transfer data
1353 * beyond the buffer end onto the wire.
1355 * May return error in out of memory cases. The skb is freed on error.
1357 static inline int skb_pad(struct sk_buff *skb, int pad)
1359 return __skb_pad(skb, pad, true);
1361 #define dev_kfree_skb(a) consume_skb(a)
1363 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1364 int offset, size_t size);
1366 struct skb_seq_state {
1370 __u32 stepped_offset;
1371 struct sk_buff *root_skb;
1372 struct sk_buff *cur_skb;
1377 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1378 unsigned int to, struct skb_seq_state *st);
1379 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1380 struct skb_seq_state *st);
1381 void skb_abort_seq_read(struct skb_seq_state *st);
1383 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1384 unsigned int to, struct ts_config *config);
1387 * Packet hash types specify the type of hash in skb_set_hash.
1389 * Hash types refer to the protocol layer addresses which are used to
1390 * construct a packet's hash. The hashes are used to differentiate or identify
1391 * flows of the protocol layer for the hash type. Hash types are either
1392 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1394 * Properties of hashes:
1396 * 1) Two packets in different flows have different hash values
1397 * 2) Two packets in the same flow should have the same hash value
1399 * A hash at a higher layer is considered to be more specific. A driver should
1400 * set the most specific hash possible.
1402 * A driver cannot indicate a more specific hash than the layer at which a hash
1403 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1405 * A driver may indicate a hash level which is less specific than the
1406 * actual layer the hash was computed on. For instance, a hash computed
1407 * at L4 may be considered an L3 hash. This should only be done if the
1408 * driver can't unambiguously determine that the HW computed the hash at
1409 * the higher layer. Note that the "should" in the second property above
1412 enum pkt_hash_types {
1413 PKT_HASH_TYPE_NONE, /* Undefined type */
1414 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1415 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1416 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1419 static inline void skb_clear_hash(struct sk_buff *skb)
1426 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1429 skb_clear_hash(skb);
1433 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1435 skb->l4_hash = is_l4;
1436 skb->sw_hash = is_sw;
1441 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1443 /* Used by drivers to set hash from HW */
1444 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1448 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1450 __skb_set_hash(skb, hash, true, is_l4);
1453 void __skb_get_hash(struct sk_buff *skb);
1454 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1455 u32 skb_get_poff(const struct sk_buff *skb);
1456 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1457 const struct flow_keys_basic *keys, int hlen);
1458 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1459 const void *data, int hlen_proto);
1461 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1462 int thoff, u8 ip_proto)
1464 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1467 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1468 const struct flow_dissector_key *key,
1469 unsigned int key_count);
1471 struct bpf_flow_dissector;
1472 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1473 __be16 proto, int nhoff, int hlen, unsigned int flags);
1475 bool __skb_flow_dissect(const struct net *net,
1476 const struct sk_buff *skb,
1477 struct flow_dissector *flow_dissector,
1478 void *target_container, const void *data,
1479 __be16 proto, int nhoff, int hlen, unsigned int flags);
1481 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1482 struct flow_dissector *flow_dissector,
1483 void *target_container, unsigned int flags)
1485 return __skb_flow_dissect(NULL, skb, flow_dissector,
1486 target_container, NULL, 0, 0, 0, flags);
1489 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1490 struct flow_keys *flow,
1493 memset(flow, 0, sizeof(*flow));
1494 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1495 flow, NULL, 0, 0, 0, flags);
1499 skb_flow_dissect_flow_keys_basic(const struct net *net,
1500 const struct sk_buff *skb,
1501 struct flow_keys_basic *flow,
1502 const void *data, __be16 proto,
1503 int nhoff, int hlen, unsigned int flags)
1505 memset(flow, 0, sizeof(*flow));
1506 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1507 data, proto, nhoff, hlen, flags);
1510 void skb_flow_dissect_meta(const struct sk_buff *skb,
1511 struct flow_dissector *flow_dissector,
1512 void *target_container);
1514 /* Gets a skb connection tracking info, ctinfo map should be a
1515 * map of mapsize to translate enum ip_conntrack_info states
1519 skb_flow_dissect_ct(const struct sk_buff *skb,
1520 struct flow_dissector *flow_dissector,
1521 void *target_container,
1522 u16 *ctinfo_map, size_t mapsize,
1523 bool post_ct, u16 zone);
1525 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1526 struct flow_dissector *flow_dissector,
1527 void *target_container);
1529 void skb_flow_dissect_hash(const struct sk_buff *skb,
1530 struct flow_dissector *flow_dissector,
1531 void *target_container);
1533 static inline __u32 skb_get_hash(struct sk_buff *skb)
1535 if (!skb->l4_hash && !skb->sw_hash)
1536 __skb_get_hash(skb);
1541 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1543 if (!skb->l4_hash && !skb->sw_hash) {
1544 struct flow_keys keys;
1545 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1547 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1553 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1554 const siphash_key_t *perturb);
1556 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1561 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1563 to->hash = from->hash;
1564 to->sw_hash = from->sw_hash;
1565 to->l4_hash = from->l4_hash;
1568 static inline void skb_copy_decrypted(struct sk_buff *to,
1569 const struct sk_buff *from)
1571 #ifdef CONFIG_TLS_DEVICE
1572 to->decrypted = from->decrypted;
1576 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1577 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1579 return skb->head + skb->end;
1582 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1587 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1592 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1597 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1599 return skb->end - skb->head;
1602 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1604 skb->end = skb->head + offset;
1608 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
1609 struct ubuf_info *uarg);
1611 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
1613 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
1616 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
1617 struct sk_buff *skb, struct iov_iter *from,
1620 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb,
1621 struct msghdr *msg, int len)
1623 return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len);
1626 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
1627 struct msghdr *msg, int len,
1628 struct ubuf_info *uarg);
1631 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1633 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1635 return &skb_shinfo(skb)->hwtstamps;
1638 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1640 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1642 return is_zcopy ? skb_uarg(skb) : NULL;
1645 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1647 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1650 static inline bool skb_zcopy_managed(const struct sk_buff *skb)
1652 return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS;
1655 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1656 const struct sk_buff *skb2)
1658 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1661 static inline void net_zcopy_get(struct ubuf_info *uarg)
1663 refcount_inc(&uarg->refcnt);
1666 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1668 skb_shinfo(skb)->destructor_arg = uarg;
1669 skb_shinfo(skb)->flags |= uarg->flags;
1672 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1675 if (skb && uarg && !skb_zcopy(skb)) {
1676 if (unlikely(have_ref && *have_ref))
1679 net_zcopy_get(uarg);
1680 skb_zcopy_init(skb, uarg);
1684 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1686 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1687 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1690 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1692 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1695 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1697 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1700 static inline void net_zcopy_put(struct ubuf_info *uarg)
1703 uarg->callback(NULL, uarg, true);
1706 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1709 if (uarg->callback == msg_zerocopy_callback)
1710 msg_zerocopy_put_abort(uarg, have_uref);
1712 net_zcopy_put(uarg);
1716 /* Release a reference on a zerocopy structure */
1717 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1719 struct ubuf_info *uarg = skb_zcopy(skb);
1722 if (!skb_zcopy_is_nouarg(skb))
1723 uarg->callback(skb, uarg, zerocopy_success);
1725 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1729 void __skb_zcopy_downgrade_managed(struct sk_buff *skb);
1731 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb)
1733 if (unlikely(skb_zcopy_managed(skb)))
1734 __skb_zcopy_downgrade_managed(skb);
1737 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1742 /* Iterate through singly-linked GSO fragments of an skb. */
1743 #define skb_list_walk_safe(first, skb, next_skb) \
1744 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1745 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1747 static inline void skb_list_del_init(struct sk_buff *skb)
1749 __list_del_entry(&skb->list);
1750 skb_mark_not_on_list(skb);
1754 * skb_queue_empty - check if a queue is empty
1757 * Returns true if the queue is empty, false otherwise.
1759 static inline int skb_queue_empty(const struct sk_buff_head *list)
1761 return list->next == (const struct sk_buff *) list;
1765 * skb_queue_empty_lockless - check if a queue is empty
1768 * Returns true if the queue is empty, false otherwise.
1769 * This variant can be used in lockless contexts.
1771 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1773 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1778 * skb_queue_is_last - check if skb is the last entry in the queue
1782 * Returns true if @skb is the last buffer on the list.
1784 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1785 const struct sk_buff *skb)
1787 return skb->next == (const struct sk_buff *) list;
1791 * skb_queue_is_first - check if skb is the first entry in the queue
1795 * Returns true if @skb is the first buffer on the list.
1797 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1798 const struct sk_buff *skb)
1800 return skb->prev == (const struct sk_buff *) list;
1804 * skb_queue_next - return the next packet in the queue
1806 * @skb: current buffer
1808 * Return the next packet in @list after @skb. It is only valid to
1809 * call this if skb_queue_is_last() evaluates to false.
1811 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1812 const struct sk_buff *skb)
1814 /* This BUG_ON may seem severe, but if we just return then we
1815 * are going to dereference garbage.
1817 BUG_ON(skb_queue_is_last(list, skb));
1822 * skb_queue_prev - return the prev packet in the queue
1824 * @skb: current buffer
1826 * Return the prev packet in @list before @skb. It is only valid to
1827 * call this if skb_queue_is_first() evaluates to false.
1829 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1830 const struct sk_buff *skb)
1832 /* This BUG_ON may seem severe, but if we just return then we
1833 * are going to dereference garbage.
1835 BUG_ON(skb_queue_is_first(list, skb));
1840 * skb_get - reference buffer
1841 * @skb: buffer to reference
1843 * Makes another reference to a socket buffer and returns a pointer
1846 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1848 refcount_inc(&skb->users);
1853 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1857 * skb_cloned - is the buffer a clone
1858 * @skb: buffer to check
1860 * Returns true if the buffer was generated with skb_clone() and is
1861 * one of multiple shared copies of the buffer. Cloned buffers are
1862 * shared data so must not be written to under normal circumstances.
1864 static inline int skb_cloned(const struct sk_buff *skb)
1866 return skb->cloned &&
1867 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1870 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1872 might_sleep_if(gfpflags_allow_blocking(pri));
1874 if (skb_cloned(skb))
1875 return pskb_expand_head(skb, 0, 0, pri);
1880 /* This variant of skb_unclone() makes sure skb->truesize
1881 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1883 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1884 * when various debugging features are in place.
1886 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
1887 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1889 might_sleep_if(gfpflags_allow_blocking(pri));
1891 if (skb_cloned(skb))
1892 return __skb_unclone_keeptruesize(skb, pri);
1897 * skb_header_cloned - is the header a clone
1898 * @skb: buffer to check
1900 * Returns true if modifying the header part of the buffer requires
1901 * the data to be copied.
1903 static inline int skb_header_cloned(const struct sk_buff *skb)
1910 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1911 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1912 return dataref != 1;
1915 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1917 might_sleep_if(gfpflags_allow_blocking(pri));
1919 if (skb_header_cloned(skb))
1920 return pskb_expand_head(skb, 0, 0, pri);
1926 * __skb_header_release() - allow clones to use the headroom
1927 * @skb: buffer to operate on
1929 * See "DOC: dataref and headerless skbs".
1931 static inline void __skb_header_release(struct sk_buff *skb)
1934 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1939 * skb_shared - is the buffer shared
1940 * @skb: buffer to check
1942 * Returns true if more than one person has a reference to this
1945 static inline int skb_shared(const struct sk_buff *skb)
1947 return refcount_read(&skb->users) != 1;
1951 * skb_share_check - check if buffer is shared and if so clone it
1952 * @skb: buffer to check
1953 * @pri: priority for memory allocation
1955 * If the buffer is shared the buffer is cloned and the old copy
1956 * drops a reference. A new clone with a single reference is returned.
1957 * If the buffer is not shared the original buffer is returned. When
1958 * being called from interrupt status or with spinlocks held pri must
1961 * NULL is returned on a memory allocation failure.
1963 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1965 might_sleep_if(gfpflags_allow_blocking(pri));
1966 if (skb_shared(skb)) {
1967 struct sk_buff *nskb = skb_clone(skb, pri);
1979 * Copy shared buffers into a new sk_buff. We effectively do COW on
1980 * packets to handle cases where we have a local reader and forward
1981 * and a couple of other messy ones. The normal one is tcpdumping
1982 * a packet thats being forwarded.
1986 * skb_unshare - make a copy of a shared buffer
1987 * @skb: buffer to check
1988 * @pri: priority for memory allocation
1990 * If the socket buffer is a clone then this function creates a new
1991 * copy of the data, drops a reference count on the old copy and returns
1992 * the new copy with the reference count at 1. If the buffer is not a clone
1993 * the original buffer is returned. When called with a spinlock held or
1994 * from interrupt state @pri must be %GFP_ATOMIC
1996 * %NULL is returned on a memory allocation failure.
1998 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
2001 might_sleep_if(gfpflags_allow_blocking(pri));
2002 if (skb_cloned(skb)) {
2003 struct sk_buff *nskb = skb_copy(skb, pri);
2005 /* Free our shared copy */
2016 * skb_peek - peek at the head of an &sk_buff_head
2017 * @list_: list to peek at
2019 * Peek an &sk_buff. Unlike most other operations you _MUST_
2020 * be careful with this one. A peek leaves the buffer on the
2021 * list and someone else may run off with it. You must hold
2022 * the appropriate locks or have a private queue to do this.
2024 * Returns %NULL for an empty list or a pointer to the head element.
2025 * The reference count is not incremented and the reference is therefore
2026 * volatile. Use with caution.
2028 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
2030 struct sk_buff *skb = list_->next;
2032 if (skb == (struct sk_buff *)list_)
2038 * __skb_peek - peek at the head of a non-empty &sk_buff_head
2039 * @list_: list to peek at
2041 * Like skb_peek(), but the caller knows that the list is not empty.
2043 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
2049 * skb_peek_next - peek skb following the given one from a queue
2050 * @skb: skb to start from
2051 * @list_: list to peek at
2053 * Returns %NULL when the end of the list is met or a pointer to the
2054 * next element. The reference count is not incremented and the
2055 * reference is therefore volatile. Use with caution.
2057 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
2058 const struct sk_buff_head *list_)
2060 struct sk_buff *next = skb->next;
2062 if (next == (struct sk_buff *)list_)
2068 * skb_peek_tail - peek at the tail of an &sk_buff_head
2069 * @list_: list to peek at
2071 * Peek an &sk_buff. Unlike most other operations you _MUST_
2072 * be careful with this one. A peek leaves the buffer on the
2073 * list and someone else may run off with it. You must hold
2074 * the appropriate locks or have a private queue to do this.
2076 * Returns %NULL for an empty list or a pointer to the tail element.
2077 * The reference count is not incremented and the reference is therefore
2078 * volatile. Use with caution.
2080 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
2082 struct sk_buff *skb = READ_ONCE(list_->prev);
2084 if (skb == (struct sk_buff *)list_)
2091 * skb_queue_len - get queue length
2092 * @list_: list to measure
2094 * Return the length of an &sk_buff queue.
2096 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
2102 * skb_queue_len_lockless - get queue length
2103 * @list_: list to measure
2105 * Return the length of an &sk_buff queue.
2106 * This variant can be used in lockless contexts.
2108 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
2110 return READ_ONCE(list_->qlen);
2114 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
2115 * @list: queue to initialize
2117 * This initializes only the list and queue length aspects of
2118 * an sk_buff_head object. This allows to initialize the list
2119 * aspects of an sk_buff_head without reinitializing things like
2120 * the spinlock. It can also be used for on-stack sk_buff_head
2121 * objects where the spinlock is known to not be used.
2123 static inline void __skb_queue_head_init(struct sk_buff_head *list)
2125 list->prev = list->next = (struct sk_buff *)list;
2130 * This function creates a split out lock class for each invocation;
2131 * this is needed for now since a whole lot of users of the skb-queue
2132 * infrastructure in drivers have different locking usage (in hardirq)
2133 * than the networking core (in softirq only). In the long run either the
2134 * network layer or drivers should need annotation to consolidate the
2135 * main types of usage into 3 classes.
2137 static inline void skb_queue_head_init(struct sk_buff_head *list)
2139 spin_lock_init(&list->lock);
2140 __skb_queue_head_init(list);
2143 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2144 struct lock_class_key *class)
2146 skb_queue_head_init(list);
2147 lockdep_set_class(&list->lock, class);
2151 * Insert an sk_buff on a list.
2153 * The "__skb_xxxx()" functions are the non-atomic ones that
2154 * can only be called with interrupts disabled.
2156 static inline void __skb_insert(struct sk_buff *newsk,
2157 struct sk_buff *prev, struct sk_buff *next,
2158 struct sk_buff_head *list)
2160 /* See skb_queue_empty_lockless() and skb_peek_tail()
2161 * for the opposite READ_ONCE()
2163 WRITE_ONCE(newsk->next, next);
2164 WRITE_ONCE(newsk->prev, prev);
2165 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2166 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
2167 WRITE_ONCE(list->qlen, list->qlen + 1);
2170 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2171 struct sk_buff *prev,
2172 struct sk_buff *next)
2174 struct sk_buff *first = list->next;
2175 struct sk_buff *last = list->prev;
2177 WRITE_ONCE(first->prev, prev);
2178 WRITE_ONCE(prev->next, first);
2180 WRITE_ONCE(last->next, next);
2181 WRITE_ONCE(next->prev, last);
2185 * skb_queue_splice - join two skb lists, this is designed for stacks
2186 * @list: the new list to add
2187 * @head: the place to add it in the first list
2189 static inline void skb_queue_splice(const struct sk_buff_head *list,
2190 struct sk_buff_head *head)
2192 if (!skb_queue_empty(list)) {
2193 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2194 head->qlen += list->qlen;
2199 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2200 * @list: the new list to add
2201 * @head: the place to add it in the first list
2203 * The list at @list is reinitialised
2205 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2206 struct sk_buff_head *head)
2208 if (!skb_queue_empty(list)) {
2209 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2210 head->qlen += list->qlen;
2211 __skb_queue_head_init(list);
2216 * skb_queue_splice_tail - join two skb lists, each list being a queue
2217 * @list: the new list to add
2218 * @head: the place to add it in the first list
2220 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2221 struct sk_buff_head *head)
2223 if (!skb_queue_empty(list)) {
2224 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2225 head->qlen += list->qlen;
2230 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2231 * @list: the new list to add
2232 * @head: the place to add it in the first list
2234 * Each of the lists is a queue.
2235 * The list at @list is reinitialised
2237 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2238 struct sk_buff_head *head)
2240 if (!skb_queue_empty(list)) {
2241 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2242 head->qlen += list->qlen;
2243 __skb_queue_head_init(list);
2248 * __skb_queue_after - queue a buffer at the list head
2249 * @list: list to use
2250 * @prev: place after this buffer
2251 * @newsk: buffer to queue
2253 * Queue a buffer int the middle of a list. This function takes no locks
2254 * and you must therefore hold required locks before calling it.
2256 * A buffer cannot be placed on two lists at the same time.
2258 static inline void __skb_queue_after(struct sk_buff_head *list,
2259 struct sk_buff *prev,
2260 struct sk_buff *newsk)
2262 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
2265 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2266 struct sk_buff_head *list);
2268 static inline void __skb_queue_before(struct sk_buff_head *list,
2269 struct sk_buff *next,
2270 struct sk_buff *newsk)
2272 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
2276 * __skb_queue_head - queue a buffer at the list head
2277 * @list: list to use
2278 * @newsk: buffer to queue
2280 * Queue a buffer at the start of a list. This function takes no locks
2281 * and you must therefore hold required locks before calling it.
2283 * A buffer cannot be placed on two lists at the same time.
2285 static inline void __skb_queue_head(struct sk_buff_head *list,
2286 struct sk_buff *newsk)
2288 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2290 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2293 * __skb_queue_tail - queue a buffer at the list tail
2294 * @list: list to use
2295 * @newsk: buffer to queue
2297 * Queue a buffer at the end of a list. This function takes no locks
2298 * and you must therefore hold required locks before calling it.
2300 * A buffer cannot be placed on two lists at the same time.
2302 static inline void __skb_queue_tail(struct sk_buff_head *list,
2303 struct sk_buff *newsk)
2305 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2307 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2310 * remove sk_buff from list. _Must_ be called atomically, and with
2313 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2314 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2316 struct sk_buff *next, *prev;
2318 WRITE_ONCE(list->qlen, list->qlen - 1);
2321 skb->next = skb->prev = NULL;
2322 WRITE_ONCE(next->prev, prev);
2323 WRITE_ONCE(prev->next, next);
2327 * __skb_dequeue - remove from the head of the queue
2328 * @list: list to dequeue from
2330 * Remove the head of the list. This function does not take any locks
2331 * so must be used with appropriate locks held only. The head item is
2332 * returned or %NULL if the list is empty.
2334 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2336 struct sk_buff *skb = skb_peek(list);
2338 __skb_unlink(skb, list);
2341 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2344 * __skb_dequeue_tail - remove from the tail of the queue
2345 * @list: list to dequeue from
2347 * Remove the tail of the list. This function does not take any locks
2348 * so must be used with appropriate locks held only. The tail item is
2349 * returned or %NULL if the list is empty.
2351 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2353 struct sk_buff *skb = skb_peek_tail(list);
2355 __skb_unlink(skb, list);
2358 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2361 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2363 return skb->data_len;
2366 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2368 return skb->len - skb->data_len;
2371 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2373 unsigned int i, len = 0;
2375 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2376 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2380 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2382 return skb_headlen(skb) + __skb_pagelen(skb);
2385 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2386 int i, struct page *page,
2389 skb_frag_t *frag = &shinfo->frags[i];
2392 * Propagate page pfmemalloc to the skb if we can. The problem is
2393 * that not all callers have unique ownership of the page but rely
2394 * on page_is_pfmemalloc doing the right thing(tm).
2396 frag->bv_page = page;
2397 frag->bv_offset = off;
2398 skb_frag_size_set(frag, size);
2402 * skb_len_add - adds a number to len fields of skb
2403 * @skb: buffer to add len to
2404 * @delta: number of bytes to add
2406 static inline void skb_len_add(struct sk_buff *skb, int delta)
2409 skb->data_len += delta;
2410 skb->truesize += delta;
2414 * __skb_fill_page_desc - initialise a paged fragment in an skb
2415 * @skb: buffer containing fragment to be initialised
2416 * @i: paged fragment index to initialise
2417 * @page: the page to use for this fragment
2418 * @off: the offset to the data with @page
2419 * @size: the length of the data
2421 * Initialises the @i'th fragment of @skb to point to &size bytes at
2422 * offset @off within @page.
2424 * Does not take any additional reference on the fragment.
2426 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2427 struct page *page, int off, int size)
2429 __skb_fill_page_desc_noacc(skb_shinfo(skb), i, page, off, size);
2430 page = compound_head(page);
2431 if (page_is_pfmemalloc(page))
2432 skb->pfmemalloc = true;
2436 * skb_fill_page_desc - initialise a paged fragment in an skb
2437 * @skb: buffer containing fragment to be initialised
2438 * @i: paged fragment index to initialise
2439 * @page: the page to use for this fragment
2440 * @off: the offset to the data with @page
2441 * @size: the length of the data
2443 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2444 * @skb to point to @size bytes at offset @off within @page. In
2445 * addition updates @skb such that @i is the last fragment.
2447 * Does not take any additional reference on the fragment.
2449 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2450 struct page *page, int off, int size)
2452 __skb_fill_page_desc(skb, i, page, off, size);
2453 skb_shinfo(skb)->nr_frags = i + 1;
2457 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2458 * @skb: buffer containing fragment to be initialised
2459 * @i: paged fragment index to initialise
2460 * @page: the page to use for this fragment
2461 * @off: the offset to the data with @page
2462 * @size: the length of the data
2464 * Variant of skb_fill_page_desc() which does not deal with
2465 * pfmemalloc, if page is not owned by us.
2467 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2468 struct page *page, int off,
2471 struct skb_shared_info *shinfo = skb_shinfo(skb);
2473 __skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2474 shinfo->nr_frags = i + 1;
2477 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2478 int size, unsigned int truesize);
2480 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2481 unsigned int truesize);
2483 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2485 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2486 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2488 return skb->head + skb->tail;
2491 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2493 skb->tail = skb->data - skb->head;
2496 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2498 skb_reset_tail_pointer(skb);
2499 skb->tail += offset;
2502 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2503 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2508 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2510 skb->tail = skb->data;
2513 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2515 skb->tail = skb->data + offset;
2518 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2520 static inline void skb_assert_len(struct sk_buff *skb)
2522 #ifdef CONFIG_DEBUG_NET
2523 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2524 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2525 #endif /* CONFIG_DEBUG_NET */
2529 * Add data to an sk_buff
2531 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2532 void *skb_put(struct sk_buff *skb, unsigned int len);
2533 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2535 void *tmp = skb_tail_pointer(skb);
2536 SKB_LINEAR_ASSERT(skb);
2542 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2544 void *tmp = __skb_put(skb, len);
2546 memset(tmp, 0, len);
2550 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2553 void *tmp = __skb_put(skb, len);
2555 memcpy(tmp, data, len);
2559 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2561 *(u8 *)__skb_put(skb, 1) = val;
2564 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2566 void *tmp = skb_put(skb, len);
2568 memset(tmp, 0, len);
2573 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2576 void *tmp = skb_put(skb, len);
2578 memcpy(tmp, data, len);
2583 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2585 *(u8 *)skb_put(skb, 1) = val;
2588 void *skb_push(struct sk_buff *skb, unsigned int len);
2589 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2596 void *skb_pull(struct sk_buff *skb, unsigned int len);
2597 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2600 if (unlikely(skb->len < skb->data_len)) {
2601 #if defined(CONFIG_DEBUG_NET)
2603 pr_err("__skb_pull(len=%u)\n", len);
2604 skb_dump(KERN_ERR, skb, false);
2608 return skb->data += len;
2611 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2613 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2616 void *skb_pull_data(struct sk_buff *skb, size_t len);
2618 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2620 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2622 if (likely(len <= skb_headlen(skb)))
2624 if (unlikely(len > skb->len))
2626 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2629 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2631 if (!pskb_may_pull(skb, len))
2635 return skb->data += len;
2638 void skb_condense(struct sk_buff *skb);
2641 * skb_headroom - bytes at buffer head
2642 * @skb: buffer to check
2644 * Return the number of bytes of free space at the head of an &sk_buff.
2646 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2648 return skb->data - skb->head;
2652 * skb_tailroom - bytes at buffer end
2653 * @skb: buffer to check
2655 * Return the number of bytes of free space at the tail of an sk_buff
2657 static inline int skb_tailroom(const struct sk_buff *skb)
2659 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2663 * skb_availroom - bytes at buffer end
2664 * @skb: buffer to check
2666 * Return the number of bytes of free space at the tail of an sk_buff
2667 * allocated by sk_stream_alloc()
2669 static inline int skb_availroom(const struct sk_buff *skb)
2671 if (skb_is_nonlinear(skb))
2674 return skb->end - skb->tail - skb->reserved_tailroom;
2678 * skb_reserve - adjust headroom
2679 * @skb: buffer to alter
2680 * @len: bytes to move
2682 * Increase the headroom of an empty &sk_buff by reducing the tail
2683 * room. This is only allowed for an empty buffer.
2685 static inline void skb_reserve(struct sk_buff *skb, int len)
2692 * skb_tailroom_reserve - adjust reserved_tailroom
2693 * @skb: buffer to alter
2694 * @mtu: maximum amount of headlen permitted
2695 * @needed_tailroom: minimum amount of reserved_tailroom
2697 * Set reserved_tailroom so that headlen can be as large as possible but
2698 * not larger than mtu and tailroom cannot be smaller than
2700 * The required headroom should already have been reserved before using
2703 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2704 unsigned int needed_tailroom)
2706 SKB_LINEAR_ASSERT(skb);
2707 if (mtu < skb_tailroom(skb) - needed_tailroom)
2708 /* use at most mtu */
2709 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2711 /* use up to all available space */
2712 skb->reserved_tailroom = needed_tailroom;
2715 #define ENCAP_TYPE_ETHER 0
2716 #define ENCAP_TYPE_IPPROTO 1
2718 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2721 skb->inner_protocol = protocol;
2722 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2725 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2728 skb->inner_ipproto = ipproto;
2729 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2732 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2734 skb->inner_mac_header = skb->mac_header;
2735 skb->inner_network_header = skb->network_header;
2736 skb->inner_transport_header = skb->transport_header;
2739 static inline void skb_reset_mac_len(struct sk_buff *skb)
2741 skb->mac_len = skb->network_header - skb->mac_header;
2744 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2747 return skb->head + skb->inner_transport_header;
2750 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2752 return skb_inner_transport_header(skb) - skb->data;
2755 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2757 skb->inner_transport_header = skb->data - skb->head;
2760 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2763 skb_reset_inner_transport_header(skb);
2764 skb->inner_transport_header += offset;
2767 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2769 return skb->head + skb->inner_network_header;
2772 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2774 skb->inner_network_header = skb->data - skb->head;
2777 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2780 skb_reset_inner_network_header(skb);
2781 skb->inner_network_header += offset;
2784 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2786 return skb->head + skb->inner_mac_header;
2789 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2791 skb->inner_mac_header = skb->data - skb->head;
2794 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2797 skb_reset_inner_mac_header(skb);
2798 skb->inner_mac_header += offset;
2800 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2802 return skb->transport_header != (typeof(skb->transport_header))~0U;
2805 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2807 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
2808 return skb->head + skb->transport_header;
2811 static inline void skb_reset_transport_header(struct sk_buff *skb)
2813 skb->transport_header = skb->data - skb->head;
2816 static inline void skb_set_transport_header(struct sk_buff *skb,
2819 skb_reset_transport_header(skb);
2820 skb->transport_header += offset;
2823 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2825 return skb->head + skb->network_header;
2828 static inline void skb_reset_network_header(struct sk_buff *skb)
2830 skb->network_header = skb->data - skb->head;
2833 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2835 skb_reset_network_header(skb);
2836 skb->network_header += offset;
2839 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2841 return skb->mac_header != (typeof(skb->mac_header))~0U;
2844 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2846 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
2847 return skb->head + skb->mac_header;
2850 static inline int skb_mac_offset(const struct sk_buff *skb)
2852 return skb_mac_header(skb) - skb->data;
2855 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2857 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
2858 return skb->network_header - skb->mac_header;
2861 static inline void skb_unset_mac_header(struct sk_buff *skb)
2863 skb->mac_header = (typeof(skb->mac_header))~0U;
2866 static inline void skb_reset_mac_header(struct sk_buff *skb)
2868 skb->mac_header = skb->data - skb->head;
2871 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2873 skb_reset_mac_header(skb);
2874 skb->mac_header += offset;
2877 static inline void skb_pop_mac_header(struct sk_buff *skb)
2879 skb->mac_header = skb->network_header;
2882 static inline void skb_probe_transport_header(struct sk_buff *skb)
2884 struct flow_keys_basic keys;
2886 if (skb_transport_header_was_set(skb))
2889 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2891 skb_set_transport_header(skb, keys.control.thoff);
2894 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2896 if (skb_mac_header_was_set(skb)) {
2897 const unsigned char *old_mac = skb_mac_header(skb);
2899 skb_set_mac_header(skb, -skb->mac_len);
2900 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2904 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2906 return skb->csum_start - skb_headroom(skb);
2909 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2911 return skb->head + skb->csum_start;
2914 static inline int skb_transport_offset(const struct sk_buff *skb)
2916 return skb_transport_header(skb) - skb->data;
2919 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2921 return skb->transport_header - skb->network_header;
2924 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2926 return skb->inner_transport_header - skb->inner_network_header;
2929 static inline int skb_network_offset(const struct sk_buff *skb)
2931 return skb_network_header(skb) - skb->data;
2934 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2936 return skb_inner_network_header(skb) - skb->data;
2939 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2941 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2945 * CPUs often take a performance hit when accessing unaligned memory
2946 * locations. The actual performance hit varies, it can be small if the
2947 * hardware handles it or large if we have to take an exception and fix it
2950 * Since an ethernet header is 14 bytes network drivers often end up with
2951 * the IP header at an unaligned offset. The IP header can be aligned by
2952 * shifting the start of the packet by 2 bytes. Drivers should do this
2955 * skb_reserve(skb, NET_IP_ALIGN);
2957 * The downside to this alignment of the IP header is that the DMA is now
2958 * unaligned. On some architectures the cost of an unaligned DMA is high
2959 * and this cost outweighs the gains made by aligning the IP header.
2961 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2964 #ifndef NET_IP_ALIGN
2965 #define NET_IP_ALIGN 2
2969 * The networking layer reserves some headroom in skb data (via
2970 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2971 * the header has to grow. In the default case, if the header has to grow
2972 * 32 bytes or less we avoid the reallocation.
2974 * Unfortunately this headroom changes the DMA alignment of the resulting
2975 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2976 * on some architectures. An architecture can override this value,
2977 * perhaps setting it to a cacheline in size (since that will maintain
2978 * cacheline alignment of the DMA). It must be a power of 2.
2980 * Various parts of the networking layer expect at least 32 bytes of
2981 * headroom, you should not reduce this.
2983 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2984 * to reduce average number of cache lines per packet.
2985 * get_rps_cpu() for example only access one 64 bytes aligned block :
2986 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2989 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2992 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2994 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2996 if (WARN_ON(skb_is_nonlinear(skb)))
2999 skb_set_tail_pointer(skb, len);
3002 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3004 __skb_set_length(skb, len);
3007 void skb_trim(struct sk_buff *skb, unsigned int len);
3009 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3012 return ___pskb_trim(skb, len);
3013 __skb_trim(skb, len);
3017 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3019 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3023 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3024 * @skb: buffer to alter
3027 * This is identical to pskb_trim except that the caller knows that
3028 * the skb is not cloned so we should never get an error due to out-
3031 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3033 int err = pskb_trim(skb, len);
3037 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3039 unsigned int diff = len - skb->len;
3041 if (skb_tailroom(skb) < diff) {
3042 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3047 __skb_set_length(skb, len);
3052 * skb_orphan - orphan a buffer
3053 * @skb: buffer to orphan
3055 * If a buffer currently has an owner then we call the owner's
3056 * destructor function and make the @skb unowned. The buffer continues
3057 * to exist but is no longer charged to its former owner.
3059 static inline void skb_orphan(struct sk_buff *skb)
3061 if (skb->destructor) {
3062 skb->destructor(skb);
3063 skb->destructor = NULL;
3071 * skb_orphan_frags - orphan the frags contained in a buffer
3072 * @skb: buffer to orphan frags from
3073 * @gfp_mask: allocation mask for replacement pages
3075 * For each frag in the SKB which needs a destructor (i.e. has an
3076 * owner) create a copy of that frag and release the original
3077 * page by calling the destructor.
3079 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3081 if (likely(!skb_zcopy(skb)))
3083 if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN)
3085 return skb_copy_ubufs(skb, gfp_mask);
3088 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
3089 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3091 if (likely(!skb_zcopy(skb)))
3093 return skb_copy_ubufs(skb, gfp_mask);
3097 * __skb_queue_purge - empty a list
3098 * @list: list to empty
3100 * Delete all buffers on an &sk_buff list. Each buffer is removed from
3101 * the list and one reference dropped. This function does not take the
3102 * list lock and the caller must hold the relevant locks to use it.
3104 static inline void __skb_queue_purge(struct sk_buff_head *list)
3106 struct sk_buff *skb;
3107 while ((skb = __skb_dequeue(list)) != NULL)
3110 void skb_queue_purge(struct sk_buff_head *list);
3112 unsigned int skb_rbtree_purge(struct rb_root *root);
3114 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3117 * netdev_alloc_frag - allocate a page fragment
3118 * @fragsz: fragment size
3120 * Allocates a frag from a page for receive buffer.
3121 * Uses GFP_ATOMIC allocations.
3123 static inline void *netdev_alloc_frag(unsigned int fragsz)
3125 return __netdev_alloc_frag_align(fragsz, ~0u);
3128 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3131 WARN_ON_ONCE(!is_power_of_2(align));
3132 return __netdev_alloc_frag_align(fragsz, -align);
3135 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3139 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
3140 * @dev: network device to receive on
3141 * @length: length to allocate
3143 * Allocate a new &sk_buff and assign it a usage count of one. The
3144 * buffer has unspecified headroom built in. Users should allocate
3145 * the headroom they think they need without accounting for the
3146 * built in space. The built in space is used for optimisations.
3148 * %NULL is returned if there is no free memory. Although this function
3149 * allocates memory it can be called from an interrupt.
3151 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3152 unsigned int length)
3154 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3157 /* legacy helper around __netdev_alloc_skb() */
3158 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3161 return __netdev_alloc_skb(NULL, length, gfp_mask);
3164 /* legacy helper around netdev_alloc_skb() */
3165 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3167 return netdev_alloc_skb(NULL, length);
3171 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3172 unsigned int length, gfp_t gfp)
3174 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3176 if (NET_IP_ALIGN && skb)
3177 skb_reserve(skb, NET_IP_ALIGN);
3181 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3182 unsigned int length)
3184 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3187 static inline void skb_free_frag(void *addr)
3189 page_frag_free(addr);
3192 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3194 static inline void *napi_alloc_frag(unsigned int fragsz)
3196 return __napi_alloc_frag_align(fragsz, ~0u);
3199 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3202 WARN_ON_ONCE(!is_power_of_2(align));
3203 return __napi_alloc_frag_align(fragsz, -align);
3206 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3207 unsigned int length, gfp_t gfp_mask);
3208 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3209 unsigned int length)
3211 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3213 void napi_consume_skb(struct sk_buff *skb, int budget);
3215 void napi_skb_free_stolen_head(struct sk_buff *skb);
3216 void __kfree_skb_defer(struct sk_buff *skb);
3219 * __dev_alloc_pages - allocate page for network Rx
3220 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3221 * @order: size of the allocation
3223 * Allocate a new page.
3225 * %NULL is returned if there is no free memory.
3227 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3230 /* This piece of code contains several assumptions.
3231 * 1. This is for device Rx, therefor a cold page is preferred.
3232 * 2. The expectation is the user wants a compound page.
3233 * 3. If requesting a order 0 page it will not be compound
3234 * due to the check to see if order has a value in prep_new_page
3235 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3236 * code in gfp_to_alloc_flags that should be enforcing this.
3238 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3240 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3243 static inline struct page *dev_alloc_pages(unsigned int order)
3245 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3249 * __dev_alloc_page - allocate a page for network Rx
3250 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3252 * Allocate a new page.
3254 * %NULL is returned if there is no free memory.
3256 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3258 return __dev_alloc_pages(gfp_mask, 0);
3261 static inline struct page *dev_alloc_page(void)
3263 return dev_alloc_pages(0);
3267 * dev_page_is_reusable - check whether a page can be reused for network Rx
3268 * @page: the page to test
3270 * A page shouldn't be considered for reusing/recycling if it was allocated
3271 * under memory pressure or at a distant memory node.
3273 * Returns false if this page should be returned to page allocator, true
3276 static inline bool dev_page_is_reusable(const struct page *page)
3278 return likely(page_to_nid(page) == numa_mem_id() &&
3279 !page_is_pfmemalloc(page));
3283 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3284 * @page: The page that was allocated from skb_alloc_page
3285 * @skb: The skb that may need pfmemalloc set
3287 static inline void skb_propagate_pfmemalloc(const struct page *page,
3288 struct sk_buff *skb)
3290 if (page_is_pfmemalloc(page))
3291 skb->pfmemalloc = true;
3295 * skb_frag_off() - Returns the offset of a skb fragment
3296 * @frag: the paged fragment
3298 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3300 return frag->bv_offset;
3304 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3305 * @frag: skb fragment
3306 * @delta: value to add
3308 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3310 frag->bv_offset += delta;
3314 * skb_frag_off_set() - Sets the offset of a skb fragment
3315 * @frag: skb fragment
3316 * @offset: offset of fragment
3318 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3320 frag->bv_offset = offset;
3324 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3325 * @fragto: skb fragment where offset is set
3326 * @fragfrom: skb fragment offset is copied from
3328 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3329 const skb_frag_t *fragfrom)
3331 fragto->bv_offset = fragfrom->bv_offset;
3335 * skb_frag_page - retrieve the page referred to by a paged fragment
3336 * @frag: the paged fragment
3338 * Returns the &struct page associated with @frag.
3340 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3342 return frag->bv_page;
3346 * __skb_frag_ref - take an addition reference on a paged fragment.
3347 * @frag: the paged fragment
3349 * Takes an additional reference on the paged fragment @frag.
3351 static inline void __skb_frag_ref(skb_frag_t *frag)
3353 get_page(skb_frag_page(frag));
3357 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3359 * @f: the fragment offset.
3361 * Takes an additional reference on the @f'th paged fragment of @skb.
3363 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3365 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3369 * __skb_frag_unref - release a reference on a paged fragment.
3370 * @frag: the paged fragment
3371 * @recycle: recycle the page if allocated via page_pool
3373 * Releases a reference on the paged fragment @frag
3374 * or recycles the page via the page_pool API.
3376 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3378 struct page *page = skb_frag_page(frag);
3380 #ifdef CONFIG_PAGE_POOL
3381 if (recycle && page_pool_return_skb_page(page))
3388 * skb_frag_unref - release a reference on a paged fragment of an skb.
3390 * @f: the fragment offset
3392 * Releases a reference on the @f'th paged fragment of @skb.
3394 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3396 struct skb_shared_info *shinfo = skb_shinfo(skb);
3398 if (!skb_zcopy_managed(skb))
3399 __skb_frag_unref(&shinfo->frags[f], skb->pp_recycle);
3403 * skb_frag_address - gets the address of the data contained in a paged fragment
3404 * @frag: the paged fragment buffer
3406 * Returns the address of the data within @frag. The page must already
3409 static inline void *skb_frag_address(const skb_frag_t *frag)
3411 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3415 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3416 * @frag: the paged fragment buffer
3418 * Returns the address of the data within @frag. Checks that the page
3419 * is mapped and returns %NULL otherwise.
3421 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3423 void *ptr = page_address(skb_frag_page(frag));
3427 return ptr + skb_frag_off(frag);
3431 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3432 * @fragto: skb fragment where page is set
3433 * @fragfrom: skb fragment page is copied from
3435 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3436 const skb_frag_t *fragfrom)
3438 fragto->bv_page = fragfrom->bv_page;
3442 * __skb_frag_set_page - sets the page contained in a paged fragment
3443 * @frag: the paged fragment
3444 * @page: the page to set
3446 * Sets the fragment @frag to contain @page.
3448 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3450 frag->bv_page = page;
3454 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3456 * @f: the fragment offset
3457 * @page: the page to set
3459 * Sets the @f'th fragment of @skb to contain @page.
3461 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3464 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3467 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3470 * skb_frag_dma_map - maps a paged fragment via the DMA API
3471 * @dev: the device to map the fragment to
3472 * @frag: the paged fragment to map
3473 * @offset: the offset within the fragment (starting at the
3474 * fragment's own offset)
3475 * @size: the number of bytes to map
3476 * @dir: the direction of the mapping (``PCI_DMA_*``)
3478 * Maps the page associated with @frag to @device.
3480 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3481 const skb_frag_t *frag,
3482 size_t offset, size_t size,
3483 enum dma_data_direction dir)
3485 return dma_map_page(dev, skb_frag_page(frag),
3486 skb_frag_off(frag) + offset, size, dir);
3489 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3492 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3496 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3499 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3504 * skb_clone_writable - is the header of a clone writable
3505 * @skb: buffer to check
3506 * @len: length up to which to write
3508 * Returns true if modifying the header part of the cloned buffer
3509 * does not requires the data to be copied.
3511 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3513 return !skb_header_cloned(skb) &&
3514 skb_headroom(skb) + len <= skb->hdr_len;
3517 static inline int skb_try_make_writable(struct sk_buff *skb,
3518 unsigned int write_len)
3520 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3521 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3524 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3529 if (headroom > skb_headroom(skb))
3530 delta = headroom - skb_headroom(skb);
3532 if (delta || cloned)
3533 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3539 * skb_cow - copy header of skb when it is required
3540 * @skb: buffer to cow
3541 * @headroom: needed headroom
3543 * If the skb passed lacks sufficient headroom or its data part
3544 * is shared, data is reallocated. If reallocation fails, an error
3545 * is returned and original skb is not changed.
3547 * The result is skb with writable area skb->head...skb->tail
3548 * and at least @headroom of space at head.
3550 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3552 return __skb_cow(skb, headroom, skb_cloned(skb));
3556 * skb_cow_head - skb_cow but only making the head writable
3557 * @skb: buffer to cow
3558 * @headroom: needed headroom
3560 * This function is identical to skb_cow except that we replace the
3561 * skb_cloned check by skb_header_cloned. It should be used when
3562 * you only need to push on some header and do not need to modify
3565 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3567 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3571 * skb_padto - pad an skbuff up to a minimal size
3572 * @skb: buffer to pad
3573 * @len: minimal length
3575 * Pads up a buffer to ensure the trailing bytes exist and are
3576 * blanked. If the buffer already contains sufficient data it
3577 * is untouched. Otherwise it is extended. Returns zero on
3578 * success. The skb is freed on error.
3580 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3582 unsigned int size = skb->len;
3583 if (likely(size >= len))
3585 return skb_pad(skb, len - size);
3589 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3590 * @skb: buffer to pad
3591 * @len: minimal length
3592 * @free_on_error: free buffer on error
3594 * Pads up a buffer to ensure the trailing bytes exist and are
3595 * blanked. If the buffer already contains sufficient data it
3596 * is untouched. Otherwise it is extended. Returns zero on
3597 * success. The skb is freed on error if @free_on_error is true.
3599 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3603 unsigned int size = skb->len;
3605 if (unlikely(size < len)) {
3607 if (__skb_pad(skb, len, free_on_error))
3609 __skb_put(skb, len);
3615 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3616 * @skb: buffer to pad
3617 * @len: minimal length
3619 * Pads up a buffer to ensure the trailing bytes exist and are
3620 * blanked. If the buffer already contains sufficient data it
3621 * is untouched. Otherwise it is extended. Returns zero on
3622 * success. The skb is freed on error.
3624 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3626 return __skb_put_padto(skb, len, true);
3629 static inline int skb_add_data(struct sk_buff *skb,
3630 struct iov_iter *from, int copy)
3632 const int off = skb->len;
3634 if (skb->ip_summed == CHECKSUM_NONE) {
3636 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3638 skb->csum = csum_block_add(skb->csum, csum, off);
3641 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3644 __skb_trim(skb, off);
3648 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3649 const struct page *page, int off)
3654 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3656 return page == skb_frag_page(frag) &&
3657 off == skb_frag_off(frag) + skb_frag_size(frag);
3662 static inline int __skb_linearize(struct sk_buff *skb)
3664 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3668 * skb_linearize - convert paged skb to linear one
3669 * @skb: buffer to linarize
3671 * If there is no free memory -ENOMEM is returned, otherwise zero
3672 * is returned and the old skb data released.
3674 static inline int skb_linearize(struct sk_buff *skb)
3676 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3680 * skb_has_shared_frag - can any frag be overwritten
3681 * @skb: buffer to test
3683 * Return true if the skb has at least one frag that might be modified
3684 * by an external entity (as in vmsplice()/sendfile())
3686 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3688 return skb_is_nonlinear(skb) &&
3689 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3693 * skb_linearize_cow - make sure skb is linear and writable
3694 * @skb: buffer to process
3696 * If there is no free memory -ENOMEM is returned, otherwise zero
3697 * is returned and the old skb data released.
3699 static inline int skb_linearize_cow(struct sk_buff *skb)
3701 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3702 __skb_linearize(skb) : 0;
3705 static __always_inline void
3706 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3709 if (skb->ip_summed == CHECKSUM_COMPLETE)
3710 skb->csum = csum_block_sub(skb->csum,
3711 csum_partial(start, len, 0), off);
3712 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3713 skb_checksum_start_offset(skb) < 0)
3714 skb->ip_summed = CHECKSUM_NONE;
3718 * skb_postpull_rcsum - update checksum for received skb after pull
3719 * @skb: buffer to update
3720 * @start: start of data before pull
3721 * @len: length of data pulled
3723 * After doing a pull on a received packet, you need to call this to
3724 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3725 * CHECKSUM_NONE so that it can be recomputed from scratch.
3727 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3728 const void *start, unsigned int len)
3730 if (skb->ip_summed == CHECKSUM_COMPLETE)
3731 skb->csum = wsum_negate(csum_partial(start, len,
3732 wsum_negate(skb->csum)));
3733 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3734 skb_checksum_start_offset(skb) < 0)
3735 skb->ip_summed = CHECKSUM_NONE;
3738 static __always_inline void
3739 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3742 if (skb->ip_summed == CHECKSUM_COMPLETE)
3743 skb->csum = csum_block_add(skb->csum,
3744 csum_partial(start, len, 0), off);
3748 * skb_postpush_rcsum - update checksum for received skb after push
3749 * @skb: buffer to update
3750 * @start: start of data after push
3751 * @len: length of data pushed
3753 * After doing a push on a received packet, you need to call this to
3754 * update the CHECKSUM_COMPLETE checksum.
3756 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3757 const void *start, unsigned int len)
3759 __skb_postpush_rcsum(skb, start, len, 0);
3762 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3765 * skb_push_rcsum - push skb and update receive checksum
3766 * @skb: buffer to update
3767 * @len: length of data pulled
3769 * This function performs an skb_push on the packet and updates
3770 * the CHECKSUM_COMPLETE checksum. It should be used on
3771 * receive path processing instead of skb_push unless you know
3772 * that the checksum difference is zero (e.g., a valid IP header)
3773 * or you are setting ip_summed to CHECKSUM_NONE.
3775 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3778 skb_postpush_rcsum(skb, skb->data, len);
3782 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3784 * pskb_trim_rcsum - trim received skb and update checksum
3785 * @skb: buffer to trim
3788 * This is exactly the same as pskb_trim except that it ensures the
3789 * checksum of received packets are still valid after the operation.
3790 * It can change skb pointers.
3793 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3795 if (likely(len >= skb->len))
3797 return pskb_trim_rcsum_slow(skb, len);
3800 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3802 if (skb->ip_summed == CHECKSUM_COMPLETE)
3803 skb->ip_summed = CHECKSUM_NONE;
3804 __skb_trim(skb, len);
3808 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3810 if (skb->ip_summed == CHECKSUM_COMPLETE)
3811 skb->ip_summed = CHECKSUM_NONE;
3812 return __skb_grow(skb, len);
3815 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3816 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3817 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3818 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3819 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3821 #define skb_queue_walk(queue, skb) \
3822 for (skb = (queue)->next; \
3823 skb != (struct sk_buff *)(queue); \
3826 #define skb_queue_walk_safe(queue, skb, tmp) \
3827 for (skb = (queue)->next, tmp = skb->next; \
3828 skb != (struct sk_buff *)(queue); \
3829 skb = tmp, tmp = skb->next)
3831 #define skb_queue_walk_from(queue, skb) \
3832 for (; skb != (struct sk_buff *)(queue); \
3835 #define skb_rbtree_walk(skb, root) \
3836 for (skb = skb_rb_first(root); skb != NULL; \
3837 skb = skb_rb_next(skb))
3839 #define skb_rbtree_walk_from(skb) \
3840 for (; skb != NULL; \
3841 skb = skb_rb_next(skb))
3843 #define skb_rbtree_walk_from_safe(skb, tmp) \
3844 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3847 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3848 for (tmp = skb->next; \
3849 skb != (struct sk_buff *)(queue); \
3850 skb = tmp, tmp = skb->next)
3852 #define skb_queue_reverse_walk(queue, skb) \
3853 for (skb = (queue)->prev; \
3854 skb != (struct sk_buff *)(queue); \
3857 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3858 for (skb = (queue)->prev, tmp = skb->prev; \
3859 skb != (struct sk_buff *)(queue); \
3860 skb = tmp, tmp = skb->prev)
3862 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3863 for (tmp = skb->prev; \
3864 skb != (struct sk_buff *)(queue); \
3865 skb = tmp, tmp = skb->prev)
3867 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3869 return skb_shinfo(skb)->frag_list != NULL;
3872 static inline void skb_frag_list_init(struct sk_buff *skb)
3874 skb_shinfo(skb)->frag_list = NULL;
3877 #define skb_walk_frags(skb, iter) \
3878 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3881 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3882 int *err, long *timeo_p,
3883 const struct sk_buff *skb);
3884 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3885 struct sk_buff_head *queue,
3888 struct sk_buff **last);
3889 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3890 struct sk_buff_head *queue,
3891 unsigned int flags, int *off, int *err,
3892 struct sk_buff **last);
3893 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3894 struct sk_buff_head *sk_queue,
3895 unsigned int flags, int *off, int *err);
3896 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
3897 __poll_t datagram_poll(struct file *file, struct socket *sock,
3898 struct poll_table_struct *wait);
3899 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3900 struct iov_iter *to, int size);
3901 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3902 struct msghdr *msg, int size)
3904 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3906 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3907 struct msghdr *msg);
3908 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3909 struct iov_iter *to, int len,
3910 struct ahash_request *hash);
3911 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3912 struct iov_iter *from, int len);
3913 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3914 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3915 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3916 static inline void skb_free_datagram_locked(struct sock *sk,
3917 struct sk_buff *skb)
3919 __skb_free_datagram_locked(sk, skb, 0);
3921 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3922 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3923 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3924 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3926 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3927 struct pipe_inode_info *pipe, unsigned int len,
3928 unsigned int flags);
3929 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3931 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3932 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3933 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3934 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3936 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3937 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3938 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3939 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3940 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3941 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3942 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3943 unsigned int offset);
3944 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3945 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
3946 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3947 int skb_vlan_pop(struct sk_buff *skb);
3948 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3949 int skb_eth_pop(struct sk_buff *skb);
3950 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3951 const unsigned char *src);
3952 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3953 int mac_len, bool ethernet);
3954 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3956 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3957 int skb_mpls_dec_ttl(struct sk_buff *skb);
3958 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3961 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3963 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3966 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3968 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3971 struct skb_checksum_ops {
3972 __wsum (*update)(const void *mem, int len, __wsum wsum);
3973 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3976 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3978 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3979 __wsum csum, const struct skb_checksum_ops *ops);
3980 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3983 static inline void * __must_check
3984 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3985 const void *data, int hlen, void *buffer)
3987 if (likely(hlen - offset >= len))
3988 return (void *)data + offset;
3990 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3996 static inline void * __must_check
3997 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3999 return __skb_header_pointer(skb, offset, len, skb->data,
4000 skb_headlen(skb), buffer);
4004 * skb_needs_linearize - check if we need to linearize a given skb
4005 * depending on the given device features.
4006 * @skb: socket buffer to check
4007 * @features: net device features
4009 * Returns true if either:
4010 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
4011 * 2. skb is fragmented and the device does not support SG.
4013 static inline bool skb_needs_linearize(struct sk_buff *skb,
4014 netdev_features_t features)
4016 return skb_is_nonlinear(skb) &&
4017 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4018 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4021 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4023 const unsigned int len)
4025 memcpy(to, skb->data, len);
4028 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4029 const int offset, void *to,
4030 const unsigned int len)
4032 memcpy(to, skb->data + offset, len);
4035 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4037 const unsigned int len)
4039 memcpy(skb->data, from, len);
4042 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4045 const unsigned int len)
4047 memcpy(skb->data + offset, from, len);
4050 void skb_init(void);
4052 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4058 * skb_get_timestamp - get timestamp from a skb
4059 * @skb: skb to get stamp from
4060 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
4062 * Timestamps are stored in the skb as offsets to a base timestamp.
4063 * This function converts the offset back to a struct timeval and stores
4066 static inline void skb_get_timestamp(const struct sk_buff *skb,
4067 struct __kernel_old_timeval *stamp)
4069 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
4072 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4073 struct __kernel_sock_timeval *stamp)
4075 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4077 stamp->tv_sec = ts.tv_sec;
4078 stamp->tv_usec = ts.tv_nsec / 1000;
4081 static inline void skb_get_timestampns(const struct sk_buff *skb,
4082 struct __kernel_old_timespec *stamp)
4084 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4086 stamp->tv_sec = ts.tv_sec;
4087 stamp->tv_nsec = ts.tv_nsec;
4090 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4091 struct __kernel_timespec *stamp)
4093 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4095 stamp->tv_sec = ts.tv_sec;
4096 stamp->tv_nsec = ts.tv_nsec;
4099 static inline void __net_timestamp(struct sk_buff *skb)
4101 skb->tstamp = ktime_get_real();
4102 skb->mono_delivery_time = 0;
4105 static inline ktime_t net_timedelta(ktime_t t)
4107 return ktime_sub(ktime_get_real(), t);
4110 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4114 skb->mono_delivery_time = kt && mono;
4117 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4119 /* It is used in the ingress path to clear the delivery_time.
4120 * If needed, set the skb->tstamp to the (rcv) timestamp.
4122 static inline void skb_clear_delivery_time(struct sk_buff *skb)
4124 if (skb->mono_delivery_time) {
4125 skb->mono_delivery_time = 0;
4126 if (static_branch_unlikely(&netstamp_needed_key))
4127 skb->tstamp = ktime_get_real();
4133 static inline void skb_clear_tstamp(struct sk_buff *skb)
4135 if (skb->mono_delivery_time)
4141 static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4143 if (skb->mono_delivery_time)
4149 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4151 if (!skb->mono_delivery_time && skb->tstamp)
4154 if (static_branch_unlikely(&netstamp_needed_key) || cond)
4155 return ktime_get_real();
4160 static inline u8 skb_metadata_len(const struct sk_buff *skb)
4162 return skb_shinfo(skb)->meta_len;
4165 static inline void *skb_metadata_end(const struct sk_buff *skb)
4167 return skb_mac_header(skb);
4170 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4171 const struct sk_buff *skb_b,
4174 const void *a = skb_metadata_end(skb_a);
4175 const void *b = skb_metadata_end(skb_b);
4176 /* Using more efficient varaiant than plain call to memcmp(). */
4177 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
4181 #define __it(x, op) (x -= sizeof(u##op))
4182 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4183 case 32: diffs |= __it_diff(a, b, 64);
4185 case 24: diffs |= __it_diff(a, b, 64);
4187 case 16: diffs |= __it_diff(a, b, 64);
4189 case 8: diffs |= __it_diff(a, b, 64);
4191 case 28: diffs |= __it_diff(a, b, 64);
4193 case 20: diffs |= __it_diff(a, b, 64);
4195 case 12: diffs |= __it_diff(a, b, 64);
4197 case 4: diffs |= __it_diff(a, b, 32);
4202 return memcmp(a - meta_len, b - meta_len, meta_len);
4206 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4207 const struct sk_buff *skb_b)
4209 u8 len_a = skb_metadata_len(skb_a);
4210 u8 len_b = skb_metadata_len(skb_b);
4212 if (!(len_a | len_b))
4215 return len_a != len_b ?
4216 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4219 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4221 skb_shinfo(skb)->meta_len = meta_len;
4224 static inline void skb_metadata_clear(struct sk_buff *skb)
4226 skb_metadata_set(skb, 0);
4229 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4231 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4233 void skb_clone_tx_timestamp(struct sk_buff *skb);
4234 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4236 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4238 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4242 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4247 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4250 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4252 * PHY drivers may accept clones of transmitted packets for
4253 * timestamping via their phy_driver.txtstamp method. These drivers
4254 * must call this function to return the skb back to the stack with a
4257 * @skb: clone of the original outgoing packet
4258 * @hwtstamps: hardware time stamps
4261 void skb_complete_tx_timestamp(struct sk_buff *skb,
4262 struct skb_shared_hwtstamps *hwtstamps);
4264 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4265 struct skb_shared_hwtstamps *hwtstamps,
4266 struct sock *sk, int tstype);
4269 * skb_tstamp_tx - queue clone of skb with send time stamps
4270 * @orig_skb: the original outgoing packet
4271 * @hwtstamps: hardware time stamps, may be NULL if not available
4273 * If the skb has a socket associated, then this function clones the
4274 * skb (thus sharing the actual data and optional structures), stores
4275 * the optional hardware time stamping information (if non NULL) or
4276 * generates a software time stamp (otherwise), then queues the clone
4277 * to the error queue of the socket. Errors are silently ignored.
4279 void skb_tstamp_tx(struct sk_buff *orig_skb,
4280 struct skb_shared_hwtstamps *hwtstamps);
4283 * skb_tx_timestamp() - Driver hook for transmit timestamping
4285 * Ethernet MAC Drivers should call this function in their hard_xmit()
4286 * function immediately before giving the sk_buff to the MAC hardware.
4288 * Specifically, one should make absolutely sure that this function is
4289 * called before TX completion of this packet can trigger. Otherwise
4290 * the packet could potentially already be freed.
4292 * @skb: A socket buffer.
4294 static inline void skb_tx_timestamp(struct sk_buff *skb)
4296 skb_clone_tx_timestamp(skb);
4297 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4298 skb_tstamp_tx(skb, NULL);
4302 * skb_complete_wifi_ack - deliver skb with wifi status
4304 * @skb: the original outgoing packet
4305 * @acked: ack status
4308 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4310 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4311 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4313 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4315 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4317 (skb->ip_summed == CHECKSUM_PARTIAL &&
4318 skb_checksum_start_offset(skb) >= 0));
4322 * skb_checksum_complete - Calculate checksum of an entire packet
4323 * @skb: packet to process
4325 * This function calculates the checksum over the entire packet plus
4326 * the value of skb->csum. The latter can be used to supply the
4327 * checksum of a pseudo header as used by TCP/UDP. It returns the
4330 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4331 * this function can be used to verify that checksum on received
4332 * packets. In that case the function should return zero if the
4333 * checksum is correct. In particular, this function will return zero
4334 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4335 * hardware has already verified the correctness of the checksum.
4337 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4339 return skb_csum_unnecessary(skb) ?
4340 0 : __skb_checksum_complete(skb);
4343 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4345 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4346 if (skb->csum_level == 0)
4347 skb->ip_summed = CHECKSUM_NONE;
4353 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4355 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4356 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4358 } else if (skb->ip_summed == CHECKSUM_NONE) {
4359 skb->ip_summed = CHECKSUM_UNNECESSARY;
4360 skb->csum_level = 0;
4364 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4366 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4367 skb->ip_summed = CHECKSUM_NONE;
4368 skb->csum_level = 0;
4372 /* Check if we need to perform checksum complete validation.
4374 * Returns true if checksum complete is needed, false otherwise
4375 * (either checksum is unnecessary or zero checksum is allowed).
4377 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4381 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4382 skb->csum_valid = 1;
4383 __skb_decr_checksum_unnecessary(skb);
4390 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4393 #define CHECKSUM_BREAK 76
4395 /* Unset checksum-complete
4397 * Unset checksum complete can be done when packet is being modified
4398 * (uncompressed for instance) and checksum-complete value is
4401 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4403 if (skb->ip_summed == CHECKSUM_COMPLETE)
4404 skb->ip_summed = CHECKSUM_NONE;
4407 /* Validate (init) checksum based on checksum complete.
4410 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4411 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4412 * checksum is stored in skb->csum for use in __skb_checksum_complete
4413 * non-zero: value of invalid checksum
4416 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4420 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4421 if (!csum_fold(csum_add(psum, skb->csum))) {
4422 skb->csum_valid = 1;
4429 if (complete || skb->len <= CHECKSUM_BREAK) {
4432 csum = __skb_checksum_complete(skb);
4433 skb->csum_valid = !csum;
4440 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4445 /* Perform checksum validate (init). Note that this is a macro since we only
4446 * want to calculate the pseudo header which is an input function if necessary.
4447 * First we try to validate without any computation (checksum unnecessary) and
4448 * then calculate based on checksum complete calling the function to compute
4452 * 0: checksum is validated or try to in skb_checksum_complete
4453 * non-zero: value of invalid checksum
4455 #define __skb_checksum_validate(skb, proto, complete, \
4456 zero_okay, check, compute_pseudo) \
4458 __sum16 __ret = 0; \
4459 skb->csum_valid = 0; \
4460 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4461 __ret = __skb_checksum_validate_complete(skb, \
4462 complete, compute_pseudo(skb, proto)); \
4466 #define skb_checksum_init(skb, proto, compute_pseudo) \
4467 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4469 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4470 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4472 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4473 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4475 #define skb_checksum_validate_zero_check(skb, proto, check, \
4477 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4479 #define skb_checksum_simple_validate(skb) \
4480 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4482 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4484 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4487 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4489 skb->csum = ~pseudo;
4490 skb->ip_summed = CHECKSUM_COMPLETE;
4493 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4495 if (__skb_checksum_convert_check(skb)) \
4496 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4499 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4500 u16 start, u16 offset)
4502 skb->ip_summed = CHECKSUM_PARTIAL;
4503 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4504 skb->csum_offset = offset - start;
4507 /* Update skbuf and packet to reflect the remote checksum offload operation.
4508 * When called, ptr indicates the starting point for skb->csum when
4509 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4510 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4512 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4513 int start, int offset, bool nopartial)
4518 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4522 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4523 __skb_checksum_complete(skb);
4524 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4527 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4529 /* Adjust skb->csum since we changed the packet */
4530 skb->csum = csum_add(skb->csum, delta);
4533 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4535 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4536 return (void *)(skb->_nfct & NFCT_PTRMASK);
4542 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4544 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4551 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4553 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4554 skb->slow_gro |= !!nfct;
4559 #ifdef CONFIG_SKB_EXTENSIONS
4561 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4567 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4570 #if IS_ENABLED(CONFIG_MPTCP)
4573 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4576 SKB_EXT_NUM, /* must be last */
4580 * struct skb_ext - sk_buff extensions
4581 * @refcnt: 1 on allocation, deallocated on 0
4582 * @offset: offset to add to @data to obtain extension address
4583 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4584 * @data: start of extension data, variable sized
4586 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4587 * to use 'u8' types while allowing up to 2kb worth of extension data.
4591 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4592 u8 chunks; /* same */
4593 char data[] __aligned(8);
4596 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4597 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4598 struct skb_ext *ext);
4599 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4600 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4601 void __skb_ext_put(struct skb_ext *ext);
4603 static inline void skb_ext_put(struct sk_buff *skb)
4605 if (skb->active_extensions)
4606 __skb_ext_put(skb->extensions);
4609 static inline void __skb_ext_copy(struct sk_buff *dst,
4610 const struct sk_buff *src)
4612 dst->active_extensions = src->active_extensions;
4614 if (src->active_extensions) {
4615 struct skb_ext *ext = src->extensions;
4617 refcount_inc(&ext->refcnt);
4618 dst->extensions = ext;
4622 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4625 __skb_ext_copy(dst, src);
4628 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4630 return !!ext->offset[i];
4633 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4635 return skb->active_extensions & (1 << id);
4638 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4640 if (skb_ext_exist(skb, id))
4641 __skb_ext_del(skb, id);
4644 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4646 if (skb_ext_exist(skb, id)) {
4647 struct skb_ext *ext = skb->extensions;
4649 return (void *)ext + (ext->offset[id] << 3);
4655 static inline void skb_ext_reset(struct sk_buff *skb)
4657 if (unlikely(skb->active_extensions)) {
4658 __skb_ext_put(skb->extensions);
4659 skb->active_extensions = 0;
4663 static inline bool skb_has_extensions(struct sk_buff *skb)
4665 return unlikely(skb->active_extensions);
4668 static inline void skb_ext_put(struct sk_buff *skb) {}
4669 static inline void skb_ext_reset(struct sk_buff *skb) {}
4670 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4671 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4672 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4673 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4674 #endif /* CONFIG_SKB_EXTENSIONS */
4676 static inline void nf_reset_ct(struct sk_buff *skb)
4678 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4679 nf_conntrack_put(skb_nfct(skb));
4684 static inline void nf_reset_trace(struct sk_buff *skb)
4686 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4691 static inline void ipvs_reset(struct sk_buff *skb)
4693 #if IS_ENABLED(CONFIG_IP_VS)
4694 skb->ipvs_property = 0;
4698 /* Note: This doesn't put any conntrack info in dst. */
4699 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4702 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4703 dst->_nfct = src->_nfct;
4704 nf_conntrack_get(skb_nfct(src));
4706 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4708 dst->nf_trace = src->nf_trace;
4712 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4714 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4715 nf_conntrack_put(skb_nfct(dst));
4717 dst->slow_gro = src->slow_gro;
4718 __nf_copy(dst, src, true);
4721 #ifdef CONFIG_NETWORK_SECMARK
4722 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4724 to->secmark = from->secmark;
4727 static inline void skb_init_secmark(struct sk_buff *skb)
4732 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4735 static inline void skb_init_secmark(struct sk_buff *skb)
4739 static inline int secpath_exists(const struct sk_buff *skb)
4742 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4748 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4750 return !skb->destructor &&
4751 !secpath_exists(skb) &&
4753 !skb->_skb_refdst &&
4754 !skb_has_frag_list(skb);
4757 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4759 skb->queue_mapping = queue_mapping;
4762 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4764 return skb->queue_mapping;
4767 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4769 to->queue_mapping = from->queue_mapping;
4772 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4774 skb->queue_mapping = rx_queue + 1;
4777 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4779 return skb->queue_mapping - 1;
4782 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4784 return skb->queue_mapping != 0;
4787 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4789 skb->dst_pending_confirm = val;
4792 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4794 return skb->dst_pending_confirm != 0;
4797 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4800 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4806 /* Keeps track of mac header offset relative to skb->head.
4807 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4808 * For non-tunnel skb it points to skb_mac_header() and for
4809 * tunnel skb it points to outer mac header.
4810 * Keeps track of level of encapsulation of network headers.
4821 #define SKB_GSO_CB_OFFSET 32
4822 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4824 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4826 return (skb_mac_header(inner_skb) - inner_skb->head) -
4827 SKB_GSO_CB(inner_skb)->mac_offset;
4830 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4832 int new_headroom, headroom;
4835 headroom = skb_headroom(skb);
4836 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4840 new_headroom = skb_headroom(skb);
4841 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4845 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4847 /* Do not update partial checksums if remote checksum is enabled. */
4848 if (skb->remcsum_offload)
4851 SKB_GSO_CB(skb)->csum = res;
4852 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4855 /* Compute the checksum for a gso segment. First compute the checksum value
4856 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4857 * then add in skb->csum (checksum from csum_start to end of packet).
4858 * skb->csum and csum_start are then updated to reflect the checksum of the
4859 * resultant packet starting from the transport header-- the resultant checksum
4860 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4863 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4865 unsigned char *csum_start = skb_transport_header(skb);
4866 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4867 __wsum partial = SKB_GSO_CB(skb)->csum;
4869 SKB_GSO_CB(skb)->csum = res;
4870 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4872 return csum_fold(csum_partial(csum_start, plen, partial));
4875 static inline bool skb_is_gso(const struct sk_buff *skb)
4877 return skb_shinfo(skb)->gso_size;
4880 /* Note: Should be called only if skb_is_gso(skb) is true */
4881 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4883 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4886 /* Note: Should be called only if skb_is_gso(skb) is true */
4887 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4889 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4892 /* Note: Should be called only if skb_is_gso(skb) is true */
4893 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4895 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4898 static inline void skb_gso_reset(struct sk_buff *skb)
4900 skb_shinfo(skb)->gso_size = 0;
4901 skb_shinfo(skb)->gso_segs = 0;
4902 skb_shinfo(skb)->gso_type = 0;
4905 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4908 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4910 shinfo->gso_size += increment;
4913 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4916 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4918 shinfo->gso_size -= decrement;
4921 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4923 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4925 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4926 * wanted then gso_type will be set. */
4927 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4929 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4930 unlikely(shinfo->gso_type == 0)) {
4931 __skb_warn_lro_forwarding(skb);
4937 static inline void skb_forward_csum(struct sk_buff *skb)
4939 /* Unfortunately we don't support this one. Any brave souls? */
4940 if (skb->ip_summed == CHECKSUM_COMPLETE)
4941 skb->ip_summed = CHECKSUM_NONE;
4945 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4946 * @skb: skb to check
4948 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4949 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4950 * use this helper, to document places where we make this assertion.
4952 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4954 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
4957 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4959 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4960 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4961 unsigned int transport_len,
4962 __sum16(*skb_chkf)(struct sk_buff *skb));
4965 * skb_head_is_locked - Determine if the skb->head is locked down
4966 * @skb: skb to check
4968 * The head on skbs build around a head frag can be removed if they are
4969 * not cloned. This function returns true if the skb head is locked down
4970 * due to either being allocated via kmalloc, or by being a clone with
4971 * multiple references to the head.
4973 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4975 return !skb->head_frag || skb_cloned(skb);
4978 /* Local Checksum Offload.
4979 * Compute outer checksum based on the assumption that the
4980 * inner checksum will be offloaded later.
4981 * See Documentation/networking/checksum-offloads.rst for
4982 * explanation of how this works.
4983 * Fill in outer checksum adjustment (e.g. with sum of outer
4984 * pseudo-header) before calling.
4985 * Also ensure that inner checksum is in linear data area.
4987 static inline __wsum lco_csum(struct sk_buff *skb)
4989 unsigned char *csum_start = skb_checksum_start(skb);
4990 unsigned char *l4_hdr = skb_transport_header(skb);
4993 /* Start with complement of inner checksum adjustment */
4994 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4997 /* Add in checksum of our headers (incl. outer checksum
4998 * adjustment filled in by caller) and return result.
5000 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
5003 static inline bool skb_is_redirected(const struct sk_buff *skb)
5005 return skb->redirected;
5008 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5010 skb->redirected = 1;
5011 #ifdef CONFIG_NET_REDIRECT
5012 skb->from_ingress = from_ingress;
5013 if (skb->from_ingress)
5014 skb_clear_tstamp(skb);
5018 static inline void skb_reset_redirect(struct sk_buff *skb)
5020 skb->redirected = 0;
5023 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5025 return skb->csum_not_inet;
5028 static inline void skb_set_kcov_handle(struct sk_buff *skb,
5029 const u64 kcov_handle)
5032 skb->kcov_handle = kcov_handle;
5036 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5039 return skb->kcov_handle;
5045 #ifdef CONFIG_PAGE_POOL
5046 static inline void skb_mark_for_recycle(struct sk_buff *skb)
5048 skb->pp_recycle = 1;
5052 #endif /* __KERNEL__ */
5053 #endif /* _LINUX_SKBUFF_H */