1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * Definitions for the 'struct sk_buff' memory handlers.
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
39 #include <linux/llist.h>
41 #include <net/page_pool.h>
42 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
43 #include <linux/netfilter/nf_conntrack_common.h>
45 #include <net/net_debug.h>
46 #include <net/dropreason.h>
51 * The interface for checksum offload between the stack and networking drivers
54 * IP checksum related features
55 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
57 * Drivers advertise checksum offload capabilities in the features of a device.
58 * From the stack's point of view these are capabilities offered by the driver.
59 * A driver typically only advertises features that it is capable of offloading
62 * .. flat-table:: Checksum related device features
65 * * - %NETIF_F_HW_CSUM
66 * - The driver (or its device) is able to compute one
67 * IP (one's complement) checksum for any combination
68 * of protocols or protocol layering. The checksum is
69 * computed and set in a packet per the CHECKSUM_PARTIAL
70 * interface (see below).
72 * * - %NETIF_F_IP_CSUM
73 * - Driver (device) is only able to checksum plain
74 * TCP or UDP packets over IPv4. These are specifically
75 * unencapsulated packets of the form IPv4|TCP or
76 * IPv4|UDP where the Protocol field in the IPv4 header
77 * is TCP or UDP. The IPv4 header may contain IP options.
78 * This feature cannot be set in features for a device
79 * with NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
82 * * - %NETIF_F_IPV6_CSUM
83 * - Driver (device) is only able to checksum plain
84 * TCP or UDP packets over IPv6. These are specifically
85 * unencapsulated packets of the form IPv6|TCP or
86 * IPv6|UDP where the Next Header field in the IPv6
87 * header is either TCP or UDP. IPv6 extension headers
88 * are not supported with this feature. This feature
89 * cannot be set in features for a device with
90 * NETIF_F_HW_CSUM also set. This feature is being
91 * DEPRECATED (see below).
94 * - Driver (device) performs receive checksum offload.
95 * This flag is only used to disable the RX checksum
96 * feature for a device. The stack will accept receive
97 * checksum indication in packets received on a device
98 * regardless of whether NETIF_F_RXCSUM is set.
100 * Checksumming of received packets by device
101 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
103 * Indication of checksum verification is set in &sk_buff.ip_summed.
104 * Possible values are:
108 * Device did not checksum this packet e.g. due to lack of capabilities.
109 * The packet contains full (though not verified) checksum in packet but
110 * not in skb->csum. Thus, skb->csum is undefined in this case.
112 * - %CHECKSUM_UNNECESSARY
114 * The hardware you're dealing with doesn't calculate the full checksum
115 * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
116 * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
117 * if their checksums are okay. &sk_buff.csum is still undefined in this case
118 * though. A driver or device must never modify the checksum field in the
119 * packet even if checksum is verified.
121 * %CHECKSUM_UNNECESSARY is applicable to following protocols:
123 * - TCP: IPv6 and IPv4.
124 * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
125 * zero UDP checksum for either IPv4 or IPv6, the networking stack
126 * may perform further validation in this case.
127 * - GRE: only if the checksum is present in the header.
128 * - SCTP: indicates the CRC in SCTP header has been validated.
129 * - FCOE: indicates the CRC in FC frame has been validated.
131 * &sk_buff.csum_level indicates the number of consecutive checksums found in
132 * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
133 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
134 * and a device is able to verify the checksums for UDP (possibly zero),
135 * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
136 * two. If the device were only able to verify the UDP checksum and not
137 * GRE, either because it doesn't support GRE checksum or because GRE
138 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
139 * not considered in this case).
141 * - %CHECKSUM_COMPLETE
143 * This is the most generic way. The device supplied checksum of the _whole_
144 * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
145 * hardware doesn't need to parse L3/L4 headers to implement this.
149 * - Even if device supports only some protocols, but is able to produce
150 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
151 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
153 * - %CHECKSUM_PARTIAL
155 * A checksum is set up to be offloaded to a device as described in the
156 * output description for CHECKSUM_PARTIAL. This may occur on a packet
157 * received directly from another Linux OS, e.g., a virtualized Linux kernel
158 * on the same host, or it may be set in the input path in GRO or remote
159 * checksum offload. For the purposes of checksum verification, the checksum
160 * referred to by skb->csum_start + skb->csum_offset and any preceding
161 * checksums in the packet are considered verified. Any checksums in the
162 * packet that are after the checksum being offloaded are not considered to
165 * Checksumming on transmit for non-GSO
166 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
168 * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
171 * - %CHECKSUM_PARTIAL
173 * The driver is required to checksum the packet as seen by hard_start_xmit()
174 * from &sk_buff.csum_start up to the end, and to record/write the checksum at
175 * offset &sk_buff.csum_start + &sk_buff.csum_offset.
176 * A driver may verify that the
177 * csum_start and csum_offset values are valid values given the length and
178 * offset of the packet, but it should not attempt to validate that the
179 * checksum refers to a legitimate transport layer checksum -- it is the
180 * purview of the stack to validate that csum_start and csum_offset are set
183 * When the stack requests checksum offload for a packet, the driver MUST
184 * ensure that the checksum is set correctly. A driver can either offload the
185 * checksum calculation to the device, or call skb_checksum_help (in the case
186 * that the device does not support offload for a particular checksum).
188 * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
189 * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
190 * checksum offload capability.
191 * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
192 * on network device checksumming capabilities: if a packet does not match
193 * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
194 * &sk_buff.csum_not_inet, see :ref:`crc`)
195 * is called to resolve the checksum.
199 * The skb was already checksummed by the protocol, or a checksum is not
202 * - %CHECKSUM_UNNECESSARY
204 * This has the same meaning as CHECKSUM_NONE for checksum offload on
207 * - %CHECKSUM_COMPLETE
209 * Not used in checksum output. If a driver observes a packet with this value
210 * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
214 * Non-IP checksum (CRC) offloads
215 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
220 * * - %NETIF_F_SCTP_CRC
221 * - This feature indicates that a device is capable of
222 * offloading the SCTP CRC in a packet. To perform this offload the stack
223 * will set csum_start and csum_offset accordingly, set ip_summed to
224 * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
225 * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
226 * A driver that supports both IP checksum offload and SCTP CRC32c offload
227 * must verify which offload is configured for a packet by testing the
228 * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
229 * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
231 * * - %NETIF_F_FCOE_CRC
232 * - This feature indicates that a device is capable of offloading the FCOE
233 * CRC in a packet. To perform this offload the stack will set ip_summed
234 * to %CHECKSUM_PARTIAL and set csum_start and csum_offset
235 * accordingly. Note that there is no indication in the skbuff that the
236 * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
237 * both IP checksum offload and FCOE CRC offload must verify which offload
238 * is configured for a packet, presumably by inspecting packet headers.
240 * Checksumming on output with GSO
241 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
243 * In the case of a GSO packet (skb_is_gso() is true), checksum offload
244 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
245 * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
246 * part of the GSO operation is implied. If a checksum is being offloaded
247 * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
248 * csum_offset are set to refer to the outermost checksum being offloaded
249 * (two offloaded checksums are possible with UDP encapsulation).
252 /* Don't change this without changing skb_csum_unnecessary! */
253 #define CHECKSUM_NONE 0
254 #define CHECKSUM_UNNECESSARY 1
255 #define CHECKSUM_COMPLETE 2
256 #define CHECKSUM_PARTIAL 3
258 /* Maximum value in skb->csum_level */
259 #define SKB_MAX_CSUM_LEVEL 3
261 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
262 #define SKB_WITH_OVERHEAD(X) \
263 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
264 #define SKB_MAX_ORDER(X, ORDER) \
265 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
266 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
267 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
269 /* return minimum truesize of one skb containing X bytes of data */
270 #define SKB_TRUESIZE(X) ((X) + \
271 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
272 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
274 struct ahash_request;
277 struct pipe_inode_info;
284 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
285 struct nf_bridge_info {
287 BRNF_PROTO_UNCHANGED,
294 u8 sabotage_in_done:1;
296 struct net_device *physindev;
298 /* always valid & non-NULL from FORWARD on, for physdev match */
299 struct net_device *physoutdev;
301 /* prerouting: detect dnat in orig/reply direction */
303 struct in6_addr ipv6_daddr;
305 /* after prerouting + nat detected: store original source
306 * mac since neigh resolution overwrites it, only used while
307 * skb is out in neigh layer.
309 char neigh_header[8];
314 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
315 /* Chain in tc_skb_ext will be used to share the tc chain with
316 * ovs recirc_id. It will be set to the current chain by tc
317 * and read by ovs to recirc_id.
329 struct sk_buff_head {
330 /* These two members must be first to match sk_buff. */
331 struct_group_tagged(sk_buff_list, list,
332 struct sk_buff *next;
333 struct sk_buff *prev;
342 /* To allow 64K frame to be packed as single skb without frag_list we
343 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
344 * buffers which do not start on a page boundary.
346 * Since GRO uses frags we allocate at least 16 regardless of page
349 #if (65536/PAGE_SIZE + 1) < 16
350 #define MAX_SKB_FRAGS 16UL
352 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
354 extern int sysctl_max_skb_frags;
356 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
357 * segment using its current segmentation instead.
359 #define GSO_BY_FRAGS 0xFFFF
361 typedef struct bio_vec skb_frag_t;
364 * skb_frag_size() - Returns the size of a skb fragment
365 * @frag: skb fragment
367 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
373 * skb_frag_size_set() - Sets the size of a skb fragment
374 * @frag: skb fragment
375 * @size: size of fragment
377 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
383 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
384 * @frag: skb fragment
385 * @delta: value to add
387 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
389 frag->bv_len += delta;
393 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
394 * @frag: skb fragment
395 * @delta: value to subtract
397 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
399 frag->bv_len -= delta;
403 * skb_frag_must_loop - Test if %p is a high memory page
404 * @p: fragment's page
406 static inline bool skb_frag_must_loop(struct page *p)
408 #if defined(CONFIG_HIGHMEM)
409 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
416 * skb_frag_foreach_page - loop over pages in a fragment
418 * @f: skb frag to operate on
419 * @f_off: offset from start of f->bv_page
420 * @f_len: length from f_off to loop over
421 * @p: (temp var) current page
422 * @p_off: (temp var) offset from start of current page,
423 * non-zero only on first page.
424 * @p_len: (temp var) length in current page,
425 * < PAGE_SIZE only on first and last page.
426 * @copied: (temp var) length so far, excluding current p_len.
428 * A fragment can hold a compound page, in which case per-page
429 * operations, notably kmap_atomic, must be called for each
432 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
433 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
434 p_off = (f_off) & (PAGE_SIZE - 1), \
435 p_len = skb_frag_must_loop(p) ? \
436 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
439 copied += p_len, p++, p_off = 0, \
440 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
442 #define HAVE_HW_TIME_STAMP
445 * struct skb_shared_hwtstamps - hardware time stamps
446 * @hwtstamp: hardware time stamp transformed into duration
447 * since arbitrary point in time
448 * @netdev_data: address/cookie of network device driver used as
449 * reference to actual hardware time stamp
451 * Software time stamps generated by ktime_get_real() are stored in
454 * hwtstamps can only be compared against other hwtstamps from
457 * This structure is attached to packets as part of the
458 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
460 struct skb_shared_hwtstamps {
467 /* Definitions for tx_flags in struct skb_shared_info */
469 /* generate hardware time stamp */
470 SKBTX_HW_TSTAMP = 1 << 0,
472 /* generate software time stamp when queueing packet to NIC */
473 SKBTX_SW_TSTAMP = 1 << 1,
475 /* device driver is going to provide hardware time stamp */
476 SKBTX_IN_PROGRESS = 1 << 2,
478 /* generate hardware time stamp based on cycles if supported */
479 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3,
481 /* generate wifi status information (where possible) */
482 SKBTX_WIFI_STATUS = 1 << 4,
484 /* determine hardware time stamp based on time or cycles */
485 SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
487 /* generate software time stamp when entering packet scheduling */
488 SKBTX_SCHED_TSTAMP = 1 << 6,
491 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
493 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \
494 SKBTX_HW_TSTAMP_USE_CYCLES | \
497 /* Definitions for flags in struct skb_shared_info */
499 /* use zcopy routines */
500 SKBFL_ZEROCOPY_ENABLE = BIT(0),
502 /* This indicates at least one fragment might be overwritten
503 * (as in vmsplice(), sendfile() ...)
504 * If we need to compute a TX checksum, we'll need to copy
505 * all frags to avoid possible bad checksum
507 SKBFL_SHARED_FRAG = BIT(1),
509 /* segment contains only zerocopy data and should not be
510 * charged to the kernel memory.
512 SKBFL_PURE_ZEROCOPY = BIT(2),
514 SKBFL_DONT_ORPHAN = BIT(3),
516 /* page references are managed by the ubuf_info, so it's safe to
517 * use frags only up until ubuf_info is released
519 SKBFL_MANAGED_FRAG_REFS = BIT(4),
522 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
523 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \
524 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS)
527 * The callback notifies userspace to release buffers when skb DMA is done in
528 * lower device, the skb last reference should be 0 when calling this.
529 * The zerocopy_success argument is true if zero copy transmit occurred,
530 * false on data copy or out of memory error caused by data copy attempt.
531 * The ctx field is used to track device context.
532 * The desc field is used to track userspace buffer index.
535 void (*callback)(struct sk_buff *, struct ubuf_info *,
536 bool zerocopy_success);
541 struct ubuf_info_msgzc {
542 struct ubuf_info ubuf;
558 struct user_struct *user;
563 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
564 #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \
567 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
568 void mm_unaccount_pinned_pages(struct mmpin *mmp);
570 /* This data is invariant across clones and lives at
571 * the end of the header data, ie. at skb->end.
573 struct skb_shared_info {
578 unsigned short gso_size;
579 /* Warning: this field is not always filled in (UFO)! */
580 unsigned short gso_segs;
581 struct sk_buff *frag_list;
582 struct skb_shared_hwtstamps hwtstamps;
583 unsigned int gso_type;
587 * Warning : all fields before dataref are cleared in __alloc_skb()
590 unsigned int xdp_frags_size;
592 /* Intermediate layers must ensure that destructor_arg
593 * remains valid until skb destructor */
594 void * destructor_arg;
596 /* must be last field, see pskb_expand_head() */
597 skb_frag_t frags[MAX_SKB_FRAGS];
601 * DOC: dataref and headerless skbs
603 * Transport layers send out clones of payload skbs they hold for
604 * retransmissions. To allow lower layers of the stack to prepend their headers
605 * we split &skb_shared_info.dataref into two halves.
606 * The lower 16 bits count the overall number of references.
607 * The higher 16 bits indicate how many of the references are payload-only.
608 * skb_header_cloned() checks if skb is allowed to add / write the headers.
610 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
611 * (via __skb_header_release()). Any clone created from marked skb will get
612 * &sk_buff.hdr_len populated with the available headroom.
613 * If there's the only clone in existence it's able to modify the headroom
614 * at will. The sequence of calls inside the transport layer is::
618 * __skb_header_release()
620 * // send the clone down the stack
622 * This is not a very generic construct and it depends on the transport layers
623 * doing the right thing. In practice there's usually only one payload-only skb.
624 * Having multiple payload-only skbs with different lengths of hdr_len is not
625 * possible. The payload-only skbs should never leave their owner.
627 #define SKB_DATAREF_SHIFT 16
628 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
632 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
633 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
634 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
638 SKB_GSO_TCPV4 = 1 << 0,
640 /* This indicates the skb is from an untrusted source. */
641 SKB_GSO_DODGY = 1 << 1,
643 /* This indicates the tcp segment has CWR set. */
644 SKB_GSO_TCP_ECN = 1 << 2,
646 SKB_GSO_TCP_FIXEDID = 1 << 3,
648 SKB_GSO_TCPV6 = 1 << 4,
650 SKB_GSO_FCOE = 1 << 5,
652 SKB_GSO_GRE = 1 << 6,
654 SKB_GSO_GRE_CSUM = 1 << 7,
656 SKB_GSO_IPXIP4 = 1 << 8,
658 SKB_GSO_IPXIP6 = 1 << 9,
660 SKB_GSO_UDP_TUNNEL = 1 << 10,
662 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
664 SKB_GSO_PARTIAL = 1 << 12,
666 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
668 SKB_GSO_SCTP = 1 << 14,
670 SKB_GSO_ESP = 1 << 15,
672 SKB_GSO_UDP = 1 << 16,
674 SKB_GSO_UDP_L4 = 1 << 17,
676 SKB_GSO_FRAGLIST = 1 << 18,
679 #if BITS_PER_LONG > 32
680 #define NET_SKBUFF_DATA_USES_OFFSET 1
683 #ifdef NET_SKBUFF_DATA_USES_OFFSET
684 typedef unsigned int sk_buff_data_t;
686 typedef unsigned char *sk_buff_data_t;
690 * DOC: Basic sk_buff geometry
692 * struct sk_buff itself is a metadata structure and does not hold any packet
693 * data. All the data is held in associated buffers.
695 * &sk_buff.head points to the main "head" buffer. The head buffer is divided
698 * - data buffer, containing headers and sometimes payload;
699 * this is the part of the skb operated on by the common helpers
700 * such as skb_put() or skb_pull();
701 * - shared info (struct skb_shared_info) which holds an array of pointers
702 * to read-only data in the (page, offset, length) format.
704 * Optionally &skb_shared_info.frag_list may point to another skb.
706 * Basic diagram may look like this::
711 * ,--------------------------- + head
712 * / ,----------------- + data
713 * / / ,----------- + tail
717 * -----------------------------------------------
718 * | headroom | data | tailroom | skb_shared_info |
719 * -----------------------------------------------
723 * + [page frag] ---------
724 * + frag_list --> | sk_buff |
730 * struct sk_buff - socket buffer
731 * @next: Next buffer in list
732 * @prev: Previous buffer in list
733 * @tstamp: Time we arrived/left
734 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
735 * for retransmit timer
736 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
738 * @ll_node: anchor in an llist (eg socket defer_list)
739 * @sk: Socket we are owned by
740 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
741 * fragmentation management
742 * @dev: Device we arrived on/are leaving by
743 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
744 * @cb: Control buffer. Free for use by every layer. Put private vars here
745 * @_skb_refdst: destination entry (with norefcount bit)
746 * @sp: the security path, used for xfrm
747 * @len: Length of actual data
748 * @data_len: Data length
749 * @mac_len: Length of link layer header
750 * @hdr_len: writable header length of cloned skb
751 * @csum: Checksum (must include start/offset pair)
752 * @csum_start: Offset from skb->head where checksumming should start
753 * @csum_offset: Offset from csum_start where checksum should be stored
754 * @priority: Packet queueing priority
755 * @ignore_df: allow local fragmentation
756 * @cloned: Head may be cloned (check refcnt to be sure)
757 * @ip_summed: Driver fed us an IP checksum
758 * @nohdr: Payload reference only, must not modify header
759 * @pkt_type: Packet class
760 * @fclone: skbuff clone status
761 * @ipvs_property: skbuff is owned by ipvs
762 * @inner_protocol_type: whether the inner protocol is
763 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
764 * @remcsum_offload: remote checksum offload is enabled
765 * @offload_fwd_mark: Packet was L2-forwarded in hardware
766 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
767 * @tc_skip_classify: do not classify packet. set by IFB device
768 * @tc_at_ingress: used within tc_classify to distinguish in/egress
769 * @redirected: packet was redirected by packet classifier
770 * @from_ingress: packet was redirected from the ingress path
771 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
772 * @peeked: this packet has been seen already, so stats have been
773 * done for it, don't do them again
774 * @nf_trace: netfilter packet trace flag
775 * @protocol: Packet protocol from driver
776 * @destructor: Destruct function
777 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
778 * @_sk_redir: socket redirection information for skmsg
779 * @_nfct: Associated connection, if any (with nfctinfo bits)
780 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
781 * @skb_iif: ifindex of device we arrived on
782 * @tc_index: Traffic control index
783 * @hash: the packet hash
784 * @queue_mapping: Queue mapping for multiqueue devices
785 * @head_frag: skb was allocated from page fragments,
786 * not allocated by kmalloc() or vmalloc().
787 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
788 * @pp_recycle: mark the packet for recycling instead of freeing (implies
789 * page_pool support on driver)
790 * @active_extensions: active extensions (skb_ext_id types)
791 * @ndisc_nodetype: router type (from link layer)
792 * @ooo_okay: allow the mapping of a socket to a queue to be changed
793 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
795 * @sw_hash: indicates hash was computed in software stack
796 * @wifi_acked_valid: wifi_acked was set
797 * @wifi_acked: whether frame was acked on wifi or not
798 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
799 * @encapsulation: indicates the inner headers in the skbuff are valid
800 * @encap_hdr_csum: software checksum is needed
801 * @csum_valid: checksum is already valid
802 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
803 * @csum_complete_sw: checksum was completed by software
804 * @csum_level: indicates the number of consecutive checksums found in
805 * the packet minus one that have been verified as
806 * CHECKSUM_UNNECESSARY (max 3)
807 * @scm_io_uring: SKB holds io_uring registered files
808 * @dst_pending_confirm: need to confirm neighbour
809 * @decrypted: Decrypted SKB
810 * @slow_gro: state present at GRO time, slower prepare step required
811 * @mono_delivery_time: When set, skb->tstamp has the
812 * delivery_time in mono clock base (i.e. EDT). Otherwise, the
813 * skb->tstamp has the (rcv) timestamp at ingress and
814 * delivery_time at egress.
815 * @napi_id: id of the NAPI struct this skb came from
816 * @sender_cpu: (aka @napi_id) source CPU in XPS
817 * @alloc_cpu: CPU which did the skb allocation.
818 * @secmark: security marking
819 * @mark: Generic packet mark
820 * @reserved_tailroom: (aka @mark) number of bytes of free space available
821 * at the tail of an sk_buff
822 * @vlan_present: VLAN tag is present
823 * @vlan_proto: vlan encapsulation protocol
824 * @vlan_tci: vlan tag control information
825 * @inner_protocol: Protocol (encapsulation)
826 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
827 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
828 * @inner_transport_header: Inner transport layer header (encapsulation)
829 * @inner_network_header: Network layer header (encapsulation)
830 * @inner_mac_header: Link layer header (encapsulation)
831 * @transport_header: Transport layer header
832 * @network_header: Network layer header
833 * @mac_header: Link layer header
834 * @kcov_handle: KCOV remote handle for remote coverage collection
835 * @tail: Tail pointer
837 * @head: Head of buffer
838 * @data: Data head pointer
839 * @truesize: Buffer size
840 * @users: User count - see {datagram,tcp}.c
841 * @extensions: allocated extensions, valid if active_extensions is nonzero
847 /* These two members must be first to match sk_buff_head. */
848 struct sk_buff *next;
849 struct sk_buff *prev;
852 struct net_device *dev;
853 /* Some protocols might use this space to store information,
854 * while device pointer would be NULL.
855 * UDP receive path is one user.
857 unsigned long dev_scratch;
860 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
861 struct list_head list;
862 struct llist_node ll_node;
867 int ip_defrag_offset;
872 u64 skb_mstamp_ns; /* earliest departure time */
875 * This is the control buffer. It is free to use for every
876 * layer. Please put your private variables there. If you
877 * want to keep them across layers you have to do a skb_clone()
878 * first. This is owned by whoever has the skb queued ATM.
880 char cb[48] __aligned(8);
884 unsigned long _skb_refdst;
885 void (*destructor)(struct sk_buff *skb);
887 struct list_head tcp_tsorted_anchor;
888 #ifdef CONFIG_NET_SOCK_MSG
889 unsigned long _sk_redir;
893 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
901 /* Following fields are _not_ copied in __copy_skb_header()
902 * Note that queue_mapping is here mostly to fill a hole.
906 /* if you move cloned around you also must adapt those constants */
907 #ifdef __BIG_ENDIAN_BITFIELD
908 #define CLONED_MASK (1 << 7)
910 #define CLONED_MASK 1
912 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
915 __u8 __cloned_offset[0];
923 pp_recycle:1; /* page_pool recycle indicator */
924 #ifdef CONFIG_SKB_EXTENSIONS
925 __u8 active_extensions;
928 /* Fields enclosed in headers group are copied
929 * using a single memcpy() in __copy_skb_header()
931 struct_group(headers,
934 __u8 __pkt_type_offset[0];
936 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
944 __u8 wifi_acked_valid:1;
947 /* Indicates the inner headers are valid in the skbuff. */
948 __u8 encapsulation:1;
949 __u8 encap_hdr_csum:1;
953 __u8 __pkt_vlan_present_offset[0];
955 __u8 vlan_present:1; /* See PKT_VLAN_PRESENT_BIT */
956 __u8 csum_complete_sw:1;
958 __u8 dst_pending_confirm:1;
959 __u8 mono_delivery_time:1; /* See SKB_MONO_DELIVERY_TIME_MASK */
960 #ifdef CONFIG_NET_CLS_ACT
961 __u8 tc_skip_classify:1;
962 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */
964 #ifdef CONFIG_IPV6_NDISC_NODETYPE
965 __u8 ndisc_nodetype:2;
968 __u8 ipvs_property:1;
969 __u8 inner_protocol_type:1;
970 __u8 remcsum_offload:1;
971 #ifdef CONFIG_NET_SWITCHDEV
972 __u8 offload_fwd_mark:1;
973 __u8 offload_l3_fwd_mark:1;
976 #ifdef CONFIG_NET_REDIRECT
979 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
980 __u8 nf_skip_egress:1;
982 #ifdef CONFIG_TLS_DEVICE
986 __u8 csum_not_inet:1;
989 #ifdef CONFIG_NET_SCHED
990 __u16 tc_index; /* traffic control index */
1005 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1007 unsigned int napi_id;
1008 unsigned int sender_cpu;
1012 #ifdef CONFIG_NETWORK_SECMARK
1018 __u32 reserved_tailroom;
1022 __be16 inner_protocol;
1026 __u16 inner_transport_header;
1027 __u16 inner_network_header;
1028 __u16 inner_mac_header;
1031 __u16 transport_header;
1032 __u16 network_header;
1039 ); /* end headers group */
1041 /* These elements must be at the end, see alloc_skb() for details. */
1042 sk_buff_data_t tail;
1044 unsigned char *head,
1046 unsigned int truesize;
1049 #ifdef CONFIG_SKB_EXTENSIONS
1050 /* only useable after checking ->active_extensions != 0 */
1051 struct skb_ext *extensions;
1055 /* if you move pkt_type around you also must adapt those constants */
1056 #ifdef __BIG_ENDIAN_BITFIELD
1057 #define PKT_TYPE_MAX (7 << 5)
1059 #define PKT_TYPE_MAX 7
1061 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset)
1063 /* if you move pkt_vlan_present, tc_at_ingress, or mono_delivery_time
1064 * around, you also must adapt these constants.
1066 #ifdef __BIG_ENDIAN_BITFIELD
1067 #define PKT_VLAN_PRESENT_BIT 7
1068 #define TC_AT_INGRESS_MASK (1 << 0)
1069 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 2)
1071 #define PKT_VLAN_PRESENT_BIT 0
1072 #define TC_AT_INGRESS_MASK (1 << 7)
1073 #define SKB_MONO_DELIVERY_TIME_MASK (1 << 5)
1075 #define PKT_VLAN_PRESENT_OFFSET offsetof(struct sk_buff, __pkt_vlan_present_offset)
1079 * Handling routines are only of interest to the kernel
1082 #define SKB_ALLOC_FCLONE 0x01
1083 #define SKB_ALLOC_RX 0x02
1084 #define SKB_ALLOC_NAPI 0x04
1087 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1090 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1092 return unlikely(skb->pfmemalloc);
1096 * skb might have a dst pointer attached, refcounted or not.
1097 * _skb_refdst low order bit is set if refcount was _not_ taken
1099 #define SKB_DST_NOREF 1UL
1100 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1103 * skb_dst - returns skb dst_entry
1106 * Returns skb dst_entry, regardless of reference taken or not.
1108 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1110 /* If refdst was not refcounted, check we still are in a
1111 * rcu_read_lock section
1113 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1114 !rcu_read_lock_held() &&
1115 !rcu_read_lock_bh_held());
1116 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1120 * skb_dst_set - sets skb dst
1124 * Sets skb dst, assuming a reference was taken on dst and should
1125 * be released by skb_dst_drop()
1127 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1129 skb->slow_gro |= !!dst;
1130 skb->_skb_refdst = (unsigned long)dst;
1134 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1138 * Sets skb dst, assuming a reference was not taken on dst.
1139 * If dst entry is cached, we do not take reference and dst_release
1140 * will be avoided by refdst_drop. If dst entry is not cached, we take
1141 * reference, so that last dst_release can destroy the dst immediately.
1143 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1145 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1146 skb->slow_gro |= !!dst;
1147 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1151 * skb_dst_is_noref - Test if skb dst isn't refcounted
1154 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1156 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1160 * skb_rtable - Returns the skb &rtable
1163 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1165 return (struct rtable *)skb_dst(skb);
1168 /* For mangling skb->pkt_type from user space side from applications
1169 * such as nft, tc, etc, we only allow a conservative subset of
1170 * possible pkt_types to be set.
1172 static inline bool skb_pkt_type_ok(u32 ptype)
1174 return ptype <= PACKET_OTHERHOST;
1178 * skb_napi_id - Returns the skb's NAPI id
1181 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1183 #ifdef CONFIG_NET_RX_BUSY_POLL
1184 return skb->napi_id;
1191 * skb_unref - decrement the skb's reference count
1194 * Returns true if we can free the skb.
1196 static inline bool skb_unref(struct sk_buff *skb)
1200 if (likely(refcount_read(&skb->users) == 1))
1202 else if (likely(!refcount_dec_and_test(&skb->users)))
1209 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1212 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1213 * @skb: buffer to free
1215 static inline void kfree_skb(struct sk_buff *skb)
1217 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1220 void skb_release_head_state(struct sk_buff *skb);
1221 void kfree_skb_list_reason(struct sk_buff *segs,
1222 enum skb_drop_reason reason);
1223 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1224 void skb_tx_error(struct sk_buff *skb);
1226 static inline void kfree_skb_list(struct sk_buff *segs)
1228 kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED);
1231 #ifdef CONFIG_TRACEPOINTS
1232 void consume_skb(struct sk_buff *skb);
1234 static inline void consume_skb(struct sk_buff *skb)
1236 return kfree_skb(skb);
1240 void __consume_stateless_skb(struct sk_buff *skb);
1241 void __kfree_skb(struct sk_buff *skb);
1242 extern struct kmem_cache *skbuff_head_cache;
1244 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1245 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1246 bool *fragstolen, int *delta_truesize);
1248 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1250 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1251 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1252 struct sk_buff *build_skb_around(struct sk_buff *skb,
1253 void *data, unsigned int frag_size);
1254 void skb_attempt_defer_free(struct sk_buff *skb);
1256 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1259 * alloc_skb - allocate a network buffer
1260 * @size: size to allocate
1261 * @priority: allocation mask
1263 * This function is a convenient wrapper around __alloc_skb().
1265 static inline struct sk_buff *alloc_skb(unsigned int size,
1268 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1271 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1272 unsigned long data_len,
1276 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1278 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1279 struct sk_buff_fclones {
1280 struct sk_buff skb1;
1282 struct sk_buff skb2;
1284 refcount_t fclone_ref;
1288 * skb_fclone_busy - check if fclone is busy
1292 * Returns true if skb is a fast clone, and its clone is not freed.
1293 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1294 * so we also check that this didnt happen.
1296 static inline bool skb_fclone_busy(const struct sock *sk,
1297 const struct sk_buff *skb)
1299 const struct sk_buff_fclones *fclones;
1301 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1303 return skb->fclone == SKB_FCLONE_ORIG &&
1304 refcount_read(&fclones->fclone_ref) > 1 &&
1305 READ_ONCE(fclones->skb2.sk) == sk;
1309 * alloc_skb_fclone - allocate a network buffer from fclone cache
1310 * @size: size to allocate
1311 * @priority: allocation mask
1313 * This function is a convenient wrapper around __alloc_skb().
1315 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1318 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1321 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1322 void skb_headers_offset_update(struct sk_buff *skb, int off);
1323 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1324 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1325 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1326 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1327 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1328 gfp_t gfp_mask, bool fclone);
1329 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1332 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1335 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1336 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1337 unsigned int headroom);
1338 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1339 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1340 int newtailroom, gfp_t priority);
1341 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1342 int offset, int len);
1343 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1344 int offset, int len);
1345 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1346 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1349 * skb_pad - zero pad the tail of an skb
1350 * @skb: buffer to pad
1351 * @pad: space to pad
1353 * Ensure that a buffer is followed by a padding area that is zero
1354 * filled. Used by network drivers which may DMA or transfer data
1355 * beyond the buffer end onto the wire.
1357 * May return error in out of memory cases. The skb is freed on error.
1359 static inline int skb_pad(struct sk_buff *skb, int pad)
1361 return __skb_pad(skb, pad, true);
1363 #define dev_kfree_skb(a) consume_skb(a)
1365 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1366 int offset, size_t size);
1368 struct skb_seq_state {
1372 __u32 stepped_offset;
1373 struct sk_buff *root_skb;
1374 struct sk_buff *cur_skb;
1379 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1380 unsigned int to, struct skb_seq_state *st);
1381 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1382 struct skb_seq_state *st);
1383 void skb_abort_seq_read(struct skb_seq_state *st);
1385 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1386 unsigned int to, struct ts_config *config);
1389 * Packet hash types specify the type of hash in skb_set_hash.
1391 * Hash types refer to the protocol layer addresses which are used to
1392 * construct a packet's hash. The hashes are used to differentiate or identify
1393 * flows of the protocol layer for the hash type. Hash types are either
1394 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1396 * Properties of hashes:
1398 * 1) Two packets in different flows have different hash values
1399 * 2) Two packets in the same flow should have the same hash value
1401 * A hash at a higher layer is considered to be more specific. A driver should
1402 * set the most specific hash possible.
1404 * A driver cannot indicate a more specific hash than the layer at which a hash
1405 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1407 * A driver may indicate a hash level which is less specific than the
1408 * actual layer the hash was computed on. For instance, a hash computed
1409 * at L4 may be considered an L3 hash. This should only be done if the
1410 * driver can't unambiguously determine that the HW computed the hash at
1411 * the higher layer. Note that the "should" in the second property above
1414 enum pkt_hash_types {
1415 PKT_HASH_TYPE_NONE, /* Undefined type */
1416 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1417 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1418 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1421 static inline void skb_clear_hash(struct sk_buff *skb)
1428 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1431 skb_clear_hash(skb);
1435 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1437 skb->l4_hash = is_l4;
1438 skb->sw_hash = is_sw;
1443 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1445 /* Used by drivers to set hash from HW */
1446 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1450 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1452 __skb_set_hash(skb, hash, true, is_l4);
1455 void __skb_get_hash(struct sk_buff *skb);
1456 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1457 u32 skb_get_poff(const struct sk_buff *skb);
1458 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1459 const struct flow_keys_basic *keys, int hlen);
1460 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1461 const void *data, int hlen_proto);
1463 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1464 int thoff, u8 ip_proto)
1466 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1469 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1470 const struct flow_dissector_key *key,
1471 unsigned int key_count);
1473 struct bpf_flow_dissector;
1474 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1475 __be16 proto, int nhoff, int hlen, unsigned int flags);
1477 bool __skb_flow_dissect(const struct net *net,
1478 const struct sk_buff *skb,
1479 struct flow_dissector *flow_dissector,
1480 void *target_container, const void *data,
1481 __be16 proto, int nhoff, int hlen, unsigned int flags);
1483 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1484 struct flow_dissector *flow_dissector,
1485 void *target_container, unsigned int flags)
1487 return __skb_flow_dissect(NULL, skb, flow_dissector,
1488 target_container, NULL, 0, 0, 0, flags);
1491 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1492 struct flow_keys *flow,
1495 memset(flow, 0, sizeof(*flow));
1496 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1497 flow, NULL, 0, 0, 0, flags);
1501 skb_flow_dissect_flow_keys_basic(const struct net *net,
1502 const struct sk_buff *skb,
1503 struct flow_keys_basic *flow,
1504 const void *data, __be16 proto,
1505 int nhoff, int hlen, unsigned int flags)
1507 memset(flow, 0, sizeof(*flow));
1508 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1509 data, proto, nhoff, hlen, flags);
1512 void skb_flow_dissect_meta(const struct sk_buff *skb,
1513 struct flow_dissector *flow_dissector,
1514 void *target_container);
1516 /* Gets a skb connection tracking info, ctinfo map should be a
1517 * map of mapsize to translate enum ip_conntrack_info states
1521 skb_flow_dissect_ct(const struct sk_buff *skb,
1522 struct flow_dissector *flow_dissector,
1523 void *target_container,
1524 u16 *ctinfo_map, size_t mapsize,
1525 bool post_ct, u16 zone);
1527 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1528 struct flow_dissector *flow_dissector,
1529 void *target_container);
1531 void skb_flow_dissect_hash(const struct sk_buff *skb,
1532 struct flow_dissector *flow_dissector,
1533 void *target_container);
1535 static inline __u32 skb_get_hash(struct sk_buff *skb)
1537 if (!skb->l4_hash && !skb->sw_hash)
1538 __skb_get_hash(skb);
1543 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1545 if (!skb->l4_hash && !skb->sw_hash) {
1546 struct flow_keys keys;
1547 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1549 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1555 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1556 const siphash_key_t *perturb);
1558 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1563 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1565 to->hash = from->hash;
1566 to->sw_hash = from->sw_hash;
1567 to->l4_hash = from->l4_hash;
1570 static inline void skb_copy_decrypted(struct sk_buff *to,
1571 const struct sk_buff *from)
1573 #ifdef CONFIG_TLS_DEVICE
1574 to->decrypted = from->decrypted;
1578 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1579 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1581 return skb->head + skb->end;
1584 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1589 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1594 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1599 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1601 return skb->end - skb->head;
1604 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1606 skb->end = skb->head + offset;
1610 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
1611 struct ubuf_info *uarg);
1613 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
1615 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
1618 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
1619 struct sk_buff *skb, struct iov_iter *from,
1622 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb,
1623 struct msghdr *msg, int len)
1625 return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len);
1628 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
1629 struct msghdr *msg, int len,
1630 struct ubuf_info *uarg);
1633 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1635 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1637 return &skb_shinfo(skb)->hwtstamps;
1640 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1642 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1644 return is_zcopy ? skb_uarg(skb) : NULL;
1647 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1649 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1652 static inline bool skb_zcopy_managed(const struct sk_buff *skb)
1654 return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS;
1657 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1658 const struct sk_buff *skb2)
1660 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1663 static inline void net_zcopy_get(struct ubuf_info *uarg)
1665 refcount_inc(&uarg->refcnt);
1668 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1670 skb_shinfo(skb)->destructor_arg = uarg;
1671 skb_shinfo(skb)->flags |= uarg->flags;
1674 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1677 if (skb && uarg && !skb_zcopy(skb)) {
1678 if (unlikely(have_ref && *have_ref))
1681 net_zcopy_get(uarg);
1682 skb_zcopy_init(skb, uarg);
1686 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1688 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1689 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1692 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1694 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1697 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1699 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1702 static inline void net_zcopy_put(struct ubuf_info *uarg)
1705 uarg->callback(NULL, uarg, true);
1708 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1711 if (uarg->callback == msg_zerocopy_callback)
1712 msg_zerocopy_put_abort(uarg, have_uref);
1714 net_zcopy_put(uarg);
1718 /* Release a reference on a zerocopy structure */
1719 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1721 struct ubuf_info *uarg = skb_zcopy(skb);
1724 if (!skb_zcopy_is_nouarg(skb))
1725 uarg->callback(skb, uarg, zerocopy_success);
1727 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1731 void __skb_zcopy_downgrade_managed(struct sk_buff *skb);
1733 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb)
1735 if (unlikely(skb_zcopy_managed(skb)))
1736 __skb_zcopy_downgrade_managed(skb);
1739 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1744 /* Iterate through singly-linked GSO fragments of an skb. */
1745 #define skb_list_walk_safe(first, skb, next_skb) \
1746 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1747 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1749 static inline void skb_list_del_init(struct sk_buff *skb)
1751 __list_del_entry(&skb->list);
1752 skb_mark_not_on_list(skb);
1756 * skb_queue_empty - check if a queue is empty
1759 * Returns true if the queue is empty, false otherwise.
1761 static inline int skb_queue_empty(const struct sk_buff_head *list)
1763 return list->next == (const struct sk_buff *) list;
1767 * skb_queue_empty_lockless - check if a queue is empty
1770 * Returns true if the queue is empty, false otherwise.
1771 * This variant can be used in lockless contexts.
1773 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1775 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1780 * skb_queue_is_last - check if skb is the last entry in the queue
1784 * Returns true if @skb is the last buffer on the list.
1786 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1787 const struct sk_buff *skb)
1789 return skb->next == (const struct sk_buff *) list;
1793 * skb_queue_is_first - check if skb is the first entry in the queue
1797 * Returns true if @skb is the first buffer on the list.
1799 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1800 const struct sk_buff *skb)
1802 return skb->prev == (const struct sk_buff *) list;
1806 * skb_queue_next - return the next packet in the queue
1808 * @skb: current buffer
1810 * Return the next packet in @list after @skb. It is only valid to
1811 * call this if skb_queue_is_last() evaluates to false.
1813 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1814 const struct sk_buff *skb)
1816 /* This BUG_ON may seem severe, but if we just return then we
1817 * are going to dereference garbage.
1819 BUG_ON(skb_queue_is_last(list, skb));
1824 * skb_queue_prev - return the prev packet in the queue
1826 * @skb: current buffer
1828 * Return the prev packet in @list before @skb. It is only valid to
1829 * call this if skb_queue_is_first() evaluates to false.
1831 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1832 const struct sk_buff *skb)
1834 /* This BUG_ON may seem severe, but if we just return then we
1835 * are going to dereference garbage.
1837 BUG_ON(skb_queue_is_first(list, skb));
1842 * skb_get - reference buffer
1843 * @skb: buffer to reference
1845 * Makes another reference to a socket buffer and returns a pointer
1848 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1850 refcount_inc(&skb->users);
1855 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1859 * skb_cloned - is the buffer a clone
1860 * @skb: buffer to check
1862 * Returns true if the buffer was generated with skb_clone() and is
1863 * one of multiple shared copies of the buffer. Cloned buffers are
1864 * shared data so must not be written to under normal circumstances.
1866 static inline int skb_cloned(const struct sk_buff *skb)
1868 return skb->cloned &&
1869 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1872 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1874 might_sleep_if(gfpflags_allow_blocking(pri));
1876 if (skb_cloned(skb))
1877 return pskb_expand_head(skb, 0, 0, pri);
1882 /* This variant of skb_unclone() makes sure skb->truesize
1883 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1885 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1886 * when various debugging features are in place.
1888 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
1889 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1891 might_sleep_if(gfpflags_allow_blocking(pri));
1893 if (skb_cloned(skb))
1894 return __skb_unclone_keeptruesize(skb, pri);
1899 * skb_header_cloned - is the header a clone
1900 * @skb: buffer to check
1902 * Returns true if modifying the header part of the buffer requires
1903 * the data to be copied.
1905 static inline int skb_header_cloned(const struct sk_buff *skb)
1912 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1913 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1914 return dataref != 1;
1917 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1919 might_sleep_if(gfpflags_allow_blocking(pri));
1921 if (skb_header_cloned(skb))
1922 return pskb_expand_head(skb, 0, 0, pri);
1928 * __skb_header_release() - allow clones to use the headroom
1929 * @skb: buffer to operate on
1931 * See "DOC: dataref and headerless skbs".
1933 static inline void __skb_header_release(struct sk_buff *skb)
1936 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1941 * skb_shared - is the buffer shared
1942 * @skb: buffer to check
1944 * Returns true if more than one person has a reference to this
1947 static inline int skb_shared(const struct sk_buff *skb)
1949 return refcount_read(&skb->users) != 1;
1953 * skb_share_check - check if buffer is shared and if so clone it
1954 * @skb: buffer to check
1955 * @pri: priority for memory allocation
1957 * If the buffer is shared the buffer is cloned and the old copy
1958 * drops a reference. A new clone with a single reference is returned.
1959 * If the buffer is not shared the original buffer is returned. When
1960 * being called from interrupt status or with spinlocks held pri must
1963 * NULL is returned on a memory allocation failure.
1965 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1967 might_sleep_if(gfpflags_allow_blocking(pri));
1968 if (skb_shared(skb)) {
1969 struct sk_buff *nskb = skb_clone(skb, pri);
1981 * Copy shared buffers into a new sk_buff. We effectively do COW on
1982 * packets to handle cases where we have a local reader and forward
1983 * and a couple of other messy ones. The normal one is tcpdumping
1984 * a packet thats being forwarded.
1988 * skb_unshare - make a copy of a shared buffer
1989 * @skb: buffer to check
1990 * @pri: priority for memory allocation
1992 * If the socket buffer is a clone then this function creates a new
1993 * copy of the data, drops a reference count on the old copy and returns
1994 * the new copy with the reference count at 1. If the buffer is not a clone
1995 * the original buffer is returned. When called with a spinlock held or
1996 * from interrupt state @pri must be %GFP_ATOMIC
1998 * %NULL is returned on a memory allocation failure.
2000 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
2003 might_sleep_if(gfpflags_allow_blocking(pri));
2004 if (skb_cloned(skb)) {
2005 struct sk_buff *nskb = skb_copy(skb, pri);
2007 /* Free our shared copy */
2018 * skb_peek - peek at the head of an &sk_buff_head
2019 * @list_: list to peek at
2021 * Peek an &sk_buff. Unlike most other operations you _MUST_
2022 * be careful with this one. A peek leaves the buffer on the
2023 * list and someone else may run off with it. You must hold
2024 * the appropriate locks or have a private queue to do this.
2026 * Returns %NULL for an empty list or a pointer to the head element.
2027 * The reference count is not incremented and the reference is therefore
2028 * volatile. Use with caution.
2030 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
2032 struct sk_buff *skb = list_->next;
2034 if (skb == (struct sk_buff *)list_)
2040 * __skb_peek - peek at the head of a non-empty &sk_buff_head
2041 * @list_: list to peek at
2043 * Like skb_peek(), but the caller knows that the list is not empty.
2045 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
2051 * skb_peek_next - peek skb following the given one from a queue
2052 * @skb: skb to start from
2053 * @list_: list to peek at
2055 * Returns %NULL when the end of the list is met or a pointer to the
2056 * next element. The reference count is not incremented and the
2057 * reference is therefore volatile. Use with caution.
2059 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
2060 const struct sk_buff_head *list_)
2062 struct sk_buff *next = skb->next;
2064 if (next == (struct sk_buff *)list_)
2070 * skb_peek_tail - peek at the tail of an &sk_buff_head
2071 * @list_: list to peek at
2073 * Peek an &sk_buff. Unlike most other operations you _MUST_
2074 * be careful with this one. A peek leaves the buffer on the
2075 * list and someone else may run off with it. You must hold
2076 * the appropriate locks or have a private queue to do this.
2078 * Returns %NULL for an empty list or a pointer to the tail element.
2079 * The reference count is not incremented and the reference is therefore
2080 * volatile. Use with caution.
2082 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
2084 struct sk_buff *skb = READ_ONCE(list_->prev);
2086 if (skb == (struct sk_buff *)list_)
2093 * skb_queue_len - get queue length
2094 * @list_: list to measure
2096 * Return the length of an &sk_buff queue.
2098 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
2104 * skb_queue_len_lockless - get queue length
2105 * @list_: list to measure
2107 * Return the length of an &sk_buff queue.
2108 * This variant can be used in lockless contexts.
2110 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
2112 return READ_ONCE(list_->qlen);
2116 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
2117 * @list: queue to initialize
2119 * This initializes only the list and queue length aspects of
2120 * an sk_buff_head object. This allows to initialize the list
2121 * aspects of an sk_buff_head without reinitializing things like
2122 * the spinlock. It can also be used for on-stack sk_buff_head
2123 * objects where the spinlock is known to not be used.
2125 static inline void __skb_queue_head_init(struct sk_buff_head *list)
2127 list->prev = list->next = (struct sk_buff *)list;
2132 * This function creates a split out lock class for each invocation;
2133 * this is needed for now since a whole lot of users of the skb-queue
2134 * infrastructure in drivers have different locking usage (in hardirq)
2135 * than the networking core (in softirq only). In the long run either the
2136 * network layer or drivers should need annotation to consolidate the
2137 * main types of usage into 3 classes.
2139 static inline void skb_queue_head_init(struct sk_buff_head *list)
2141 spin_lock_init(&list->lock);
2142 __skb_queue_head_init(list);
2145 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2146 struct lock_class_key *class)
2148 skb_queue_head_init(list);
2149 lockdep_set_class(&list->lock, class);
2153 * Insert an sk_buff on a list.
2155 * The "__skb_xxxx()" functions are the non-atomic ones that
2156 * can only be called with interrupts disabled.
2158 static inline void __skb_insert(struct sk_buff *newsk,
2159 struct sk_buff *prev, struct sk_buff *next,
2160 struct sk_buff_head *list)
2162 /* See skb_queue_empty_lockless() and skb_peek_tail()
2163 * for the opposite READ_ONCE()
2165 WRITE_ONCE(newsk->next, next);
2166 WRITE_ONCE(newsk->prev, prev);
2167 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2168 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
2169 WRITE_ONCE(list->qlen, list->qlen + 1);
2172 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2173 struct sk_buff *prev,
2174 struct sk_buff *next)
2176 struct sk_buff *first = list->next;
2177 struct sk_buff *last = list->prev;
2179 WRITE_ONCE(first->prev, prev);
2180 WRITE_ONCE(prev->next, first);
2182 WRITE_ONCE(last->next, next);
2183 WRITE_ONCE(next->prev, last);
2187 * skb_queue_splice - join two skb lists, this is designed for stacks
2188 * @list: the new list to add
2189 * @head: the place to add it in the first list
2191 static inline void skb_queue_splice(const struct sk_buff_head *list,
2192 struct sk_buff_head *head)
2194 if (!skb_queue_empty(list)) {
2195 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2196 head->qlen += list->qlen;
2201 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2202 * @list: the new list to add
2203 * @head: the place to add it in the first list
2205 * The list at @list is reinitialised
2207 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2208 struct sk_buff_head *head)
2210 if (!skb_queue_empty(list)) {
2211 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2212 head->qlen += list->qlen;
2213 __skb_queue_head_init(list);
2218 * skb_queue_splice_tail - join two skb lists, each list being a queue
2219 * @list: the new list to add
2220 * @head: the place to add it in the first list
2222 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2223 struct sk_buff_head *head)
2225 if (!skb_queue_empty(list)) {
2226 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2227 head->qlen += list->qlen;
2232 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2233 * @list: the new list to add
2234 * @head: the place to add it in the first list
2236 * Each of the lists is a queue.
2237 * The list at @list is reinitialised
2239 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2240 struct sk_buff_head *head)
2242 if (!skb_queue_empty(list)) {
2243 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2244 head->qlen += list->qlen;
2245 __skb_queue_head_init(list);
2250 * __skb_queue_after - queue a buffer at the list head
2251 * @list: list to use
2252 * @prev: place after this buffer
2253 * @newsk: buffer to queue
2255 * Queue a buffer int the middle of a list. This function takes no locks
2256 * and you must therefore hold required locks before calling it.
2258 * A buffer cannot be placed on two lists at the same time.
2260 static inline void __skb_queue_after(struct sk_buff_head *list,
2261 struct sk_buff *prev,
2262 struct sk_buff *newsk)
2264 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
2267 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2268 struct sk_buff_head *list);
2270 static inline void __skb_queue_before(struct sk_buff_head *list,
2271 struct sk_buff *next,
2272 struct sk_buff *newsk)
2274 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
2278 * __skb_queue_head - queue a buffer at the list head
2279 * @list: list to use
2280 * @newsk: buffer to queue
2282 * Queue a buffer at the start of a list. This function takes no locks
2283 * and you must therefore hold required locks before calling it.
2285 * A buffer cannot be placed on two lists at the same time.
2287 static inline void __skb_queue_head(struct sk_buff_head *list,
2288 struct sk_buff *newsk)
2290 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2292 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2295 * __skb_queue_tail - queue a buffer at the list tail
2296 * @list: list to use
2297 * @newsk: buffer to queue
2299 * Queue a buffer at the end of a list. This function takes no locks
2300 * and you must therefore hold required locks before calling it.
2302 * A buffer cannot be placed on two lists at the same time.
2304 static inline void __skb_queue_tail(struct sk_buff_head *list,
2305 struct sk_buff *newsk)
2307 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2309 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2312 * remove sk_buff from list. _Must_ be called atomically, and with
2315 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2316 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2318 struct sk_buff *next, *prev;
2320 WRITE_ONCE(list->qlen, list->qlen - 1);
2323 skb->next = skb->prev = NULL;
2324 WRITE_ONCE(next->prev, prev);
2325 WRITE_ONCE(prev->next, next);
2329 * __skb_dequeue - remove from the head of the queue
2330 * @list: list to dequeue from
2332 * Remove the head of the list. This function does not take any locks
2333 * so must be used with appropriate locks held only. The head item is
2334 * returned or %NULL if the list is empty.
2336 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2338 struct sk_buff *skb = skb_peek(list);
2340 __skb_unlink(skb, list);
2343 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2346 * __skb_dequeue_tail - remove from the tail of the queue
2347 * @list: list to dequeue from
2349 * Remove the tail of the list. This function does not take any locks
2350 * so must be used with appropriate locks held only. The tail item is
2351 * returned or %NULL if the list is empty.
2353 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2355 struct sk_buff *skb = skb_peek_tail(list);
2357 __skb_unlink(skb, list);
2360 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2363 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2365 return skb->data_len;
2368 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2370 return skb->len - skb->data_len;
2373 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2375 unsigned int i, len = 0;
2377 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2378 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2382 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2384 return skb_headlen(skb) + __skb_pagelen(skb);
2387 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2388 int i, struct page *page,
2391 skb_frag_t *frag = &shinfo->frags[i];
2394 * Propagate page pfmemalloc to the skb if we can. The problem is
2395 * that not all callers have unique ownership of the page but rely
2396 * on page_is_pfmemalloc doing the right thing(tm).
2398 frag->bv_page = page;
2399 frag->bv_offset = off;
2400 skb_frag_size_set(frag, size);
2404 * skb_len_add - adds a number to len fields of skb
2405 * @skb: buffer to add len to
2406 * @delta: number of bytes to add
2408 static inline void skb_len_add(struct sk_buff *skb, int delta)
2411 skb->data_len += delta;
2412 skb->truesize += delta;
2416 * __skb_fill_page_desc - initialise a paged fragment in an skb
2417 * @skb: buffer containing fragment to be initialised
2418 * @i: paged fragment index to initialise
2419 * @page: the page to use for this fragment
2420 * @off: the offset to the data with @page
2421 * @size: the length of the data
2423 * Initialises the @i'th fragment of @skb to point to &size bytes at
2424 * offset @off within @page.
2426 * Does not take any additional reference on the fragment.
2428 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2429 struct page *page, int off, int size)
2431 __skb_fill_page_desc_noacc(skb_shinfo(skb), i, page, off, size);
2432 page = compound_head(page);
2433 if (page_is_pfmemalloc(page))
2434 skb->pfmemalloc = true;
2438 * skb_fill_page_desc - initialise a paged fragment in an skb
2439 * @skb: buffer containing fragment to be initialised
2440 * @i: paged fragment index to initialise
2441 * @page: the page to use for this fragment
2442 * @off: the offset to the data with @page
2443 * @size: the length of the data
2445 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2446 * @skb to point to @size bytes at offset @off within @page. In
2447 * addition updates @skb such that @i is the last fragment.
2449 * Does not take any additional reference on the fragment.
2451 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2452 struct page *page, int off, int size)
2454 __skb_fill_page_desc(skb, i, page, off, size);
2455 skb_shinfo(skb)->nr_frags = i + 1;
2459 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2460 * @skb: buffer containing fragment to be initialised
2461 * @i: paged fragment index to initialise
2462 * @page: the page to use for this fragment
2463 * @off: the offset to the data with @page
2464 * @size: the length of the data
2466 * Variant of skb_fill_page_desc() which does not deal with
2467 * pfmemalloc, if page is not owned by us.
2469 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2470 struct page *page, int off,
2473 struct skb_shared_info *shinfo = skb_shinfo(skb);
2475 __skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2476 shinfo->nr_frags = i + 1;
2479 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2480 int size, unsigned int truesize);
2482 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2483 unsigned int truesize);
2485 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2487 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2488 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2490 return skb->head + skb->tail;
2493 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2495 skb->tail = skb->data - skb->head;
2498 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2500 skb_reset_tail_pointer(skb);
2501 skb->tail += offset;
2504 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2505 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2510 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2512 skb->tail = skb->data;
2515 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2517 skb->tail = skb->data + offset;
2520 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2522 static inline void skb_assert_len(struct sk_buff *skb)
2524 #ifdef CONFIG_DEBUG_NET
2525 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2526 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2527 #endif /* CONFIG_DEBUG_NET */
2531 * Add data to an sk_buff
2533 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2534 void *skb_put(struct sk_buff *skb, unsigned int len);
2535 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2537 void *tmp = skb_tail_pointer(skb);
2538 SKB_LINEAR_ASSERT(skb);
2544 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2546 void *tmp = __skb_put(skb, len);
2548 memset(tmp, 0, len);
2552 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2555 void *tmp = __skb_put(skb, len);
2557 memcpy(tmp, data, len);
2561 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2563 *(u8 *)__skb_put(skb, 1) = val;
2566 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2568 void *tmp = skb_put(skb, len);
2570 memset(tmp, 0, len);
2575 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2578 void *tmp = skb_put(skb, len);
2580 memcpy(tmp, data, len);
2585 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2587 *(u8 *)skb_put(skb, 1) = val;
2590 void *skb_push(struct sk_buff *skb, unsigned int len);
2591 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2598 void *skb_pull(struct sk_buff *skb, unsigned int len);
2599 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2602 if (unlikely(skb->len < skb->data_len)) {
2603 #if defined(CONFIG_DEBUG_NET)
2605 pr_err("__skb_pull(len=%u)\n", len);
2606 skb_dump(KERN_ERR, skb, false);
2610 return skb->data += len;
2613 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2615 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2618 void *skb_pull_data(struct sk_buff *skb, size_t len);
2620 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2622 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2624 if (likely(len <= skb_headlen(skb)))
2626 if (unlikely(len > skb->len))
2628 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2631 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2633 if (!pskb_may_pull(skb, len))
2637 return skb->data += len;
2640 void skb_condense(struct sk_buff *skb);
2643 * skb_headroom - bytes at buffer head
2644 * @skb: buffer to check
2646 * Return the number of bytes of free space at the head of an &sk_buff.
2648 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2650 return skb->data - skb->head;
2654 * skb_tailroom - bytes at buffer end
2655 * @skb: buffer to check
2657 * Return the number of bytes of free space at the tail of an sk_buff
2659 static inline int skb_tailroom(const struct sk_buff *skb)
2661 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2665 * skb_availroom - bytes at buffer end
2666 * @skb: buffer to check
2668 * Return the number of bytes of free space at the tail of an sk_buff
2669 * allocated by sk_stream_alloc()
2671 static inline int skb_availroom(const struct sk_buff *skb)
2673 if (skb_is_nonlinear(skb))
2676 return skb->end - skb->tail - skb->reserved_tailroom;
2680 * skb_reserve - adjust headroom
2681 * @skb: buffer to alter
2682 * @len: bytes to move
2684 * Increase the headroom of an empty &sk_buff by reducing the tail
2685 * room. This is only allowed for an empty buffer.
2687 static inline void skb_reserve(struct sk_buff *skb, int len)
2694 * skb_tailroom_reserve - adjust reserved_tailroom
2695 * @skb: buffer to alter
2696 * @mtu: maximum amount of headlen permitted
2697 * @needed_tailroom: minimum amount of reserved_tailroom
2699 * Set reserved_tailroom so that headlen can be as large as possible but
2700 * not larger than mtu and tailroom cannot be smaller than
2702 * The required headroom should already have been reserved before using
2705 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2706 unsigned int needed_tailroom)
2708 SKB_LINEAR_ASSERT(skb);
2709 if (mtu < skb_tailroom(skb) - needed_tailroom)
2710 /* use at most mtu */
2711 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2713 /* use up to all available space */
2714 skb->reserved_tailroom = needed_tailroom;
2717 #define ENCAP_TYPE_ETHER 0
2718 #define ENCAP_TYPE_IPPROTO 1
2720 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2723 skb->inner_protocol = protocol;
2724 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2727 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2730 skb->inner_ipproto = ipproto;
2731 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2734 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2736 skb->inner_mac_header = skb->mac_header;
2737 skb->inner_network_header = skb->network_header;
2738 skb->inner_transport_header = skb->transport_header;
2741 static inline void skb_reset_mac_len(struct sk_buff *skb)
2743 skb->mac_len = skb->network_header - skb->mac_header;
2746 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2749 return skb->head + skb->inner_transport_header;
2752 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2754 return skb_inner_transport_header(skb) - skb->data;
2757 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2759 skb->inner_transport_header = skb->data - skb->head;
2762 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2765 skb_reset_inner_transport_header(skb);
2766 skb->inner_transport_header += offset;
2769 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2771 return skb->head + skb->inner_network_header;
2774 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2776 skb->inner_network_header = skb->data - skb->head;
2779 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2782 skb_reset_inner_network_header(skb);
2783 skb->inner_network_header += offset;
2786 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2788 return skb->head + skb->inner_mac_header;
2791 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2793 skb->inner_mac_header = skb->data - skb->head;
2796 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2799 skb_reset_inner_mac_header(skb);
2800 skb->inner_mac_header += offset;
2802 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2804 return skb->transport_header != (typeof(skb->transport_header))~0U;
2807 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2809 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
2810 return skb->head + skb->transport_header;
2813 static inline void skb_reset_transport_header(struct sk_buff *skb)
2815 skb->transport_header = skb->data - skb->head;
2818 static inline void skb_set_transport_header(struct sk_buff *skb,
2821 skb_reset_transport_header(skb);
2822 skb->transport_header += offset;
2825 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2827 return skb->head + skb->network_header;
2830 static inline void skb_reset_network_header(struct sk_buff *skb)
2832 skb->network_header = skb->data - skb->head;
2835 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2837 skb_reset_network_header(skb);
2838 skb->network_header += offset;
2841 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2843 return skb->mac_header != (typeof(skb->mac_header))~0U;
2846 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2848 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
2849 return skb->head + skb->mac_header;
2852 static inline int skb_mac_offset(const struct sk_buff *skb)
2854 return skb_mac_header(skb) - skb->data;
2857 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2859 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
2860 return skb->network_header - skb->mac_header;
2863 static inline void skb_unset_mac_header(struct sk_buff *skb)
2865 skb->mac_header = (typeof(skb->mac_header))~0U;
2868 static inline void skb_reset_mac_header(struct sk_buff *skb)
2870 skb->mac_header = skb->data - skb->head;
2873 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2875 skb_reset_mac_header(skb);
2876 skb->mac_header += offset;
2879 static inline void skb_pop_mac_header(struct sk_buff *skb)
2881 skb->mac_header = skb->network_header;
2884 static inline void skb_probe_transport_header(struct sk_buff *skb)
2886 struct flow_keys_basic keys;
2888 if (skb_transport_header_was_set(skb))
2891 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2893 skb_set_transport_header(skb, keys.control.thoff);
2896 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2898 if (skb_mac_header_was_set(skb)) {
2899 const unsigned char *old_mac = skb_mac_header(skb);
2901 skb_set_mac_header(skb, -skb->mac_len);
2902 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2906 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2908 return skb->csum_start - skb_headroom(skb);
2911 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2913 return skb->head + skb->csum_start;
2916 static inline int skb_transport_offset(const struct sk_buff *skb)
2918 return skb_transport_header(skb) - skb->data;
2921 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2923 return skb->transport_header - skb->network_header;
2926 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2928 return skb->inner_transport_header - skb->inner_network_header;
2931 static inline int skb_network_offset(const struct sk_buff *skb)
2933 return skb_network_header(skb) - skb->data;
2936 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2938 return skb_inner_network_header(skb) - skb->data;
2941 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2943 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2947 * CPUs often take a performance hit when accessing unaligned memory
2948 * locations. The actual performance hit varies, it can be small if the
2949 * hardware handles it or large if we have to take an exception and fix it
2952 * Since an ethernet header is 14 bytes network drivers often end up with
2953 * the IP header at an unaligned offset. The IP header can be aligned by
2954 * shifting the start of the packet by 2 bytes. Drivers should do this
2957 * skb_reserve(skb, NET_IP_ALIGN);
2959 * The downside to this alignment of the IP header is that the DMA is now
2960 * unaligned. On some architectures the cost of an unaligned DMA is high
2961 * and this cost outweighs the gains made by aligning the IP header.
2963 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2966 #ifndef NET_IP_ALIGN
2967 #define NET_IP_ALIGN 2
2971 * The networking layer reserves some headroom in skb data (via
2972 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2973 * the header has to grow. In the default case, if the header has to grow
2974 * 32 bytes or less we avoid the reallocation.
2976 * Unfortunately this headroom changes the DMA alignment of the resulting
2977 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2978 * on some architectures. An architecture can override this value,
2979 * perhaps setting it to a cacheline in size (since that will maintain
2980 * cacheline alignment of the DMA). It must be a power of 2.
2982 * Various parts of the networking layer expect at least 32 bytes of
2983 * headroom, you should not reduce this.
2985 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2986 * to reduce average number of cache lines per packet.
2987 * get_rps_cpu() for example only access one 64 bytes aligned block :
2988 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2991 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2994 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2996 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2998 if (WARN_ON(skb_is_nonlinear(skb)))
3001 skb_set_tail_pointer(skb, len);
3004 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3006 __skb_set_length(skb, len);
3009 void skb_trim(struct sk_buff *skb, unsigned int len);
3011 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3014 return ___pskb_trim(skb, len);
3015 __skb_trim(skb, len);
3019 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3021 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3025 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3026 * @skb: buffer to alter
3029 * This is identical to pskb_trim except that the caller knows that
3030 * the skb is not cloned so we should never get an error due to out-
3033 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3035 int err = pskb_trim(skb, len);
3039 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3041 unsigned int diff = len - skb->len;
3043 if (skb_tailroom(skb) < diff) {
3044 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3049 __skb_set_length(skb, len);
3054 * skb_orphan - orphan a buffer
3055 * @skb: buffer to orphan
3057 * If a buffer currently has an owner then we call the owner's
3058 * destructor function and make the @skb unowned. The buffer continues
3059 * to exist but is no longer charged to its former owner.
3061 static inline void skb_orphan(struct sk_buff *skb)
3063 if (skb->destructor) {
3064 skb->destructor(skb);
3065 skb->destructor = NULL;
3073 * skb_orphan_frags - orphan the frags contained in a buffer
3074 * @skb: buffer to orphan frags from
3075 * @gfp_mask: allocation mask for replacement pages
3077 * For each frag in the SKB which needs a destructor (i.e. has an
3078 * owner) create a copy of that frag and release the original
3079 * page by calling the destructor.
3081 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3083 if (likely(!skb_zcopy(skb)))
3085 if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN)
3087 return skb_copy_ubufs(skb, gfp_mask);
3090 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
3091 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3093 if (likely(!skb_zcopy(skb)))
3095 return skb_copy_ubufs(skb, gfp_mask);
3099 * __skb_queue_purge - empty a list
3100 * @list: list to empty
3102 * Delete all buffers on an &sk_buff list. Each buffer is removed from
3103 * the list and one reference dropped. This function does not take the
3104 * list lock and the caller must hold the relevant locks to use it.
3106 static inline void __skb_queue_purge(struct sk_buff_head *list)
3108 struct sk_buff *skb;
3109 while ((skb = __skb_dequeue(list)) != NULL)
3112 void skb_queue_purge(struct sk_buff_head *list);
3114 unsigned int skb_rbtree_purge(struct rb_root *root);
3116 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3119 * netdev_alloc_frag - allocate a page fragment
3120 * @fragsz: fragment size
3122 * Allocates a frag from a page for receive buffer.
3123 * Uses GFP_ATOMIC allocations.
3125 static inline void *netdev_alloc_frag(unsigned int fragsz)
3127 return __netdev_alloc_frag_align(fragsz, ~0u);
3130 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3133 WARN_ON_ONCE(!is_power_of_2(align));
3134 return __netdev_alloc_frag_align(fragsz, -align);
3137 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3141 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
3142 * @dev: network device to receive on
3143 * @length: length to allocate
3145 * Allocate a new &sk_buff and assign it a usage count of one. The
3146 * buffer has unspecified headroom built in. Users should allocate
3147 * the headroom they think they need without accounting for the
3148 * built in space. The built in space is used for optimisations.
3150 * %NULL is returned if there is no free memory. Although this function
3151 * allocates memory it can be called from an interrupt.
3153 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3154 unsigned int length)
3156 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3159 /* legacy helper around __netdev_alloc_skb() */
3160 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3163 return __netdev_alloc_skb(NULL, length, gfp_mask);
3166 /* legacy helper around netdev_alloc_skb() */
3167 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3169 return netdev_alloc_skb(NULL, length);
3173 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3174 unsigned int length, gfp_t gfp)
3176 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3178 if (NET_IP_ALIGN && skb)
3179 skb_reserve(skb, NET_IP_ALIGN);
3183 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3184 unsigned int length)
3186 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3189 static inline void skb_free_frag(void *addr)
3191 page_frag_free(addr);
3194 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3196 static inline void *napi_alloc_frag(unsigned int fragsz)
3198 return __napi_alloc_frag_align(fragsz, ~0u);
3201 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3204 WARN_ON_ONCE(!is_power_of_2(align));
3205 return __napi_alloc_frag_align(fragsz, -align);
3208 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3209 unsigned int length, gfp_t gfp_mask);
3210 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3211 unsigned int length)
3213 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3215 void napi_consume_skb(struct sk_buff *skb, int budget);
3217 void napi_skb_free_stolen_head(struct sk_buff *skb);
3218 void __kfree_skb_defer(struct sk_buff *skb);
3221 * __dev_alloc_pages - allocate page for network Rx
3222 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3223 * @order: size of the allocation
3225 * Allocate a new page.
3227 * %NULL is returned if there is no free memory.
3229 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3232 /* This piece of code contains several assumptions.
3233 * 1. This is for device Rx, therefor a cold page is preferred.
3234 * 2. The expectation is the user wants a compound page.
3235 * 3. If requesting a order 0 page it will not be compound
3236 * due to the check to see if order has a value in prep_new_page
3237 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3238 * code in gfp_to_alloc_flags that should be enforcing this.
3240 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3242 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3245 static inline struct page *dev_alloc_pages(unsigned int order)
3247 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3251 * __dev_alloc_page - allocate a page for network Rx
3252 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3254 * Allocate a new page.
3256 * %NULL is returned if there is no free memory.
3258 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3260 return __dev_alloc_pages(gfp_mask, 0);
3263 static inline struct page *dev_alloc_page(void)
3265 return dev_alloc_pages(0);
3269 * dev_page_is_reusable - check whether a page can be reused for network Rx
3270 * @page: the page to test
3272 * A page shouldn't be considered for reusing/recycling if it was allocated
3273 * under memory pressure or at a distant memory node.
3275 * Returns false if this page should be returned to page allocator, true
3278 static inline bool dev_page_is_reusable(const struct page *page)
3280 return likely(page_to_nid(page) == numa_mem_id() &&
3281 !page_is_pfmemalloc(page));
3285 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3286 * @page: The page that was allocated from skb_alloc_page
3287 * @skb: The skb that may need pfmemalloc set
3289 static inline void skb_propagate_pfmemalloc(const struct page *page,
3290 struct sk_buff *skb)
3292 if (page_is_pfmemalloc(page))
3293 skb->pfmemalloc = true;
3297 * skb_frag_off() - Returns the offset of a skb fragment
3298 * @frag: the paged fragment
3300 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3302 return frag->bv_offset;
3306 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3307 * @frag: skb fragment
3308 * @delta: value to add
3310 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3312 frag->bv_offset += delta;
3316 * skb_frag_off_set() - Sets the offset of a skb fragment
3317 * @frag: skb fragment
3318 * @offset: offset of fragment
3320 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3322 frag->bv_offset = offset;
3326 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3327 * @fragto: skb fragment where offset is set
3328 * @fragfrom: skb fragment offset is copied from
3330 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3331 const skb_frag_t *fragfrom)
3333 fragto->bv_offset = fragfrom->bv_offset;
3337 * skb_frag_page - retrieve the page referred to by a paged fragment
3338 * @frag: the paged fragment
3340 * Returns the &struct page associated with @frag.
3342 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3344 return frag->bv_page;
3348 * __skb_frag_ref - take an addition reference on a paged fragment.
3349 * @frag: the paged fragment
3351 * Takes an additional reference on the paged fragment @frag.
3353 static inline void __skb_frag_ref(skb_frag_t *frag)
3355 get_page(skb_frag_page(frag));
3359 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3361 * @f: the fragment offset.
3363 * Takes an additional reference on the @f'th paged fragment of @skb.
3365 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3367 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3371 * __skb_frag_unref - release a reference on a paged fragment.
3372 * @frag: the paged fragment
3373 * @recycle: recycle the page if allocated via page_pool
3375 * Releases a reference on the paged fragment @frag
3376 * or recycles the page via the page_pool API.
3378 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3380 struct page *page = skb_frag_page(frag);
3382 #ifdef CONFIG_PAGE_POOL
3383 if (recycle && page_pool_return_skb_page(page))
3390 * skb_frag_unref - release a reference on a paged fragment of an skb.
3392 * @f: the fragment offset
3394 * Releases a reference on the @f'th paged fragment of @skb.
3396 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3398 struct skb_shared_info *shinfo = skb_shinfo(skb);
3400 if (!skb_zcopy_managed(skb))
3401 __skb_frag_unref(&shinfo->frags[f], skb->pp_recycle);
3405 * skb_frag_address - gets the address of the data contained in a paged fragment
3406 * @frag: the paged fragment buffer
3408 * Returns the address of the data within @frag. The page must already
3411 static inline void *skb_frag_address(const skb_frag_t *frag)
3413 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3417 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3418 * @frag: the paged fragment buffer
3420 * Returns the address of the data within @frag. Checks that the page
3421 * is mapped and returns %NULL otherwise.
3423 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3425 void *ptr = page_address(skb_frag_page(frag));
3429 return ptr + skb_frag_off(frag);
3433 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3434 * @fragto: skb fragment where page is set
3435 * @fragfrom: skb fragment page is copied from
3437 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3438 const skb_frag_t *fragfrom)
3440 fragto->bv_page = fragfrom->bv_page;
3444 * __skb_frag_set_page - sets the page contained in a paged fragment
3445 * @frag: the paged fragment
3446 * @page: the page to set
3448 * Sets the fragment @frag to contain @page.
3450 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3452 frag->bv_page = page;
3456 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3458 * @f: the fragment offset
3459 * @page: the page to set
3461 * Sets the @f'th fragment of @skb to contain @page.
3463 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3466 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3469 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3472 * skb_frag_dma_map - maps a paged fragment via the DMA API
3473 * @dev: the device to map the fragment to
3474 * @frag: the paged fragment to map
3475 * @offset: the offset within the fragment (starting at the
3476 * fragment's own offset)
3477 * @size: the number of bytes to map
3478 * @dir: the direction of the mapping (``PCI_DMA_*``)
3480 * Maps the page associated with @frag to @device.
3482 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3483 const skb_frag_t *frag,
3484 size_t offset, size_t size,
3485 enum dma_data_direction dir)
3487 return dma_map_page(dev, skb_frag_page(frag),
3488 skb_frag_off(frag) + offset, size, dir);
3491 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3494 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3498 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3501 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3506 * skb_clone_writable - is the header of a clone writable
3507 * @skb: buffer to check
3508 * @len: length up to which to write
3510 * Returns true if modifying the header part of the cloned buffer
3511 * does not requires the data to be copied.
3513 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3515 return !skb_header_cloned(skb) &&
3516 skb_headroom(skb) + len <= skb->hdr_len;
3519 static inline int skb_try_make_writable(struct sk_buff *skb,
3520 unsigned int write_len)
3522 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3523 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3526 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3531 if (headroom > skb_headroom(skb))
3532 delta = headroom - skb_headroom(skb);
3534 if (delta || cloned)
3535 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3541 * skb_cow - copy header of skb when it is required
3542 * @skb: buffer to cow
3543 * @headroom: needed headroom
3545 * If the skb passed lacks sufficient headroom or its data part
3546 * is shared, data is reallocated. If reallocation fails, an error
3547 * is returned and original skb is not changed.
3549 * The result is skb with writable area skb->head...skb->tail
3550 * and at least @headroom of space at head.
3552 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3554 return __skb_cow(skb, headroom, skb_cloned(skb));
3558 * skb_cow_head - skb_cow but only making the head writable
3559 * @skb: buffer to cow
3560 * @headroom: needed headroom
3562 * This function is identical to skb_cow except that we replace the
3563 * skb_cloned check by skb_header_cloned. It should be used when
3564 * you only need to push on some header and do not need to modify
3567 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3569 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3573 * skb_padto - pad an skbuff up to a minimal size
3574 * @skb: buffer to pad
3575 * @len: minimal length
3577 * Pads up a buffer to ensure the trailing bytes exist and are
3578 * blanked. If the buffer already contains sufficient data it
3579 * is untouched. Otherwise it is extended. Returns zero on
3580 * success. The skb is freed on error.
3582 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3584 unsigned int size = skb->len;
3585 if (likely(size >= len))
3587 return skb_pad(skb, len - size);
3591 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3592 * @skb: buffer to pad
3593 * @len: minimal length
3594 * @free_on_error: free buffer on error
3596 * Pads up a buffer to ensure the trailing bytes exist and are
3597 * blanked. If the buffer already contains sufficient data it
3598 * is untouched. Otherwise it is extended. Returns zero on
3599 * success. The skb is freed on error if @free_on_error is true.
3601 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3605 unsigned int size = skb->len;
3607 if (unlikely(size < len)) {
3609 if (__skb_pad(skb, len, free_on_error))
3611 __skb_put(skb, len);
3617 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3618 * @skb: buffer to pad
3619 * @len: minimal length
3621 * Pads up a buffer to ensure the trailing bytes exist and are
3622 * blanked. If the buffer already contains sufficient data it
3623 * is untouched. Otherwise it is extended. Returns zero on
3624 * success. The skb is freed on error.
3626 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3628 return __skb_put_padto(skb, len, true);
3631 static inline int skb_add_data(struct sk_buff *skb,
3632 struct iov_iter *from, int copy)
3634 const int off = skb->len;
3636 if (skb->ip_summed == CHECKSUM_NONE) {
3638 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3640 skb->csum = csum_block_add(skb->csum, csum, off);
3643 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3646 __skb_trim(skb, off);
3650 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3651 const struct page *page, int off)
3656 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3658 return page == skb_frag_page(frag) &&
3659 off == skb_frag_off(frag) + skb_frag_size(frag);
3664 static inline int __skb_linearize(struct sk_buff *skb)
3666 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3670 * skb_linearize - convert paged skb to linear one
3671 * @skb: buffer to linarize
3673 * If there is no free memory -ENOMEM is returned, otherwise zero
3674 * is returned and the old skb data released.
3676 static inline int skb_linearize(struct sk_buff *skb)
3678 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3682 * skb_has_shared_frag - can any frag be overwritten
3683 * @skb: buffer to test
3685 * Return true if the skb has at least one frag that might be modified
3686 * by an external entity (as in vmsplice()/sendfile())
3688 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3690 return skb_is_nonlinear(skb) &&
3691 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3695 * skb_linearize_cow - make sure skb is linear and writable
3696 * @skb: buffer to process
3698 * If there is no free memory -ENOMEM is returned, otherwise zero
3699 * is returned and the old skb data released.
3701 static inline int skb_linearize_cow(struct sk_buff *skb)
3703 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3704 __skb_linearize(skb) : 0;
3707 static __always_inline void
3708 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3711 if (skb->ip_summed == CHECKSUM_COMPLETE)
3712 skb->csum = csum_block_sub(skb->csum,
3713 csum_partial(start, len, 0), off);
3714 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3715 skb_checksum_start_offset(skb) < 0)
3716 skb->ip_summed = CHECKSUM_NONE;
3720 * skb_postpull_rcsum - update checksum for received skb after pull
3721 * @skb: buffer to update
3722 * @start: start of data before pull
3723 * @len: length of data pulled
3725 * After doing a pull on a received packet, you need to call this to
3726 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3727 * CHECKSUM_NONE so that it can be recomputed from scratch.
3729 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3730 const void *start, unsigned int len)
3732 if (skb->ip_summed == CHECKSUM_COMPLETE)
3733 skb->csum = wsum_negate(csum_partial(start, len,
3734 wsum_negate(skb->csum)));
3735 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3736 skb_checksum_start_offset(skb) < 0)
3737 skb->ip_summed = CHECKSUM_NONE;
3740 static __always_inline void
3741 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3744 if (skb->ip_summed == CHECKSUM_COMPLETE)
3745 skb->csum = csum_block_add(skb->csum,
3746 csum_partial(start, len, 0), off);
3750 * skb_postpush_rcsum - update checksum for received skb after push
3751 * @skb: buffer to update
3752 * @start: start of data after push
3753 * @len: length of data pushed
3755 * After doing a push on a received packet, you need to call this to
3756 * update the CHECKSUM_COMPLETE checksum.
3758 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3759 const void *start, unsigned int len)
3761 __skb_postpush_rcsum(skb, start, len, 0);
3764 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3767 * skb_push_rcsum - push skb and update receive checksum
3768 * @skb: buffer to update
3769 * @len: length of data pulled
3771 * This function performs an skb_push on the packet and updates
3772 * the CHECKSUM_COMPLETE checksum. It should be used on
3773 * receive path processing instead of skb_push unless you know
3774 * that the checksum difference is zero (e.g., a valid IP header)
3775 * or you are setting ip_summed to CHECKSUM_NONE.
3777 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3780 skb_postpush_rcsum(skb, skb->data, len);
3784 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3786 * pskb_trim_rcsum - trim received skb and update checksum
3787 * @skb: buffer to trim
3790 * This is exactly the same as pskb_trim except that it ensures the
3791 * checksum of received packets are still valid after the operation.
3792 * It can change skb pointers.
3795 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3797 if (likely(len >= skb->len))
3799 return pskb_trim_rcsum_slow(skb, len);
3802 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3804 if (skb->ip_summed == CHECKSUM_COMPLETE)
3805 skb->ip_summed = CHECKSUM_NONE;
3806 __skb_trim(skb, len);
3810 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3812 if (skb->ip_summed == CHECKSUM_COMPLETE)
3813 skb->ip_summed = CHECKSUM_NONE;
3814 return __skb_grow(skb, len);
3817 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3818 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3819 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3820 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3821 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3823 #define skb_queue_walk(queue, skb) \
3824 for (skb = (queue)->next; \
3825 skb != (struct sk_buff *)(queue); \
3828 #define skb_queue_walk_safe(queue, skb, tmp) \
3829 for (skb = (queue)->next, tmp = skb->next; \
3830 skb != (struct sk_buff *)(queue); \
3831 skb = tmp, tmp = skb->next)
3833 #define skb_queue_walk_from(queue, skb) \
3834 for (; skb != (struct sk_buff *)(queue); \
3837 #define skb_rbtree_walk(skb, root) \
3838 for (skb = skb_rb_first(root); skb != NULL; \
3839 skb = skb_rb_next(skb))
3841 #define skb_rbtree_walk_from(skb) \
3842 for (; skb != NULL; \
3843 skb = skb_rb_next(skb))
3845 #define skb_rbtree_walk_from_safe(skb, tmp) \
3846 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3849 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3850 for (tmp = skb->next; \
3851 skb != (struct sk_buff *)(queue); \
3852 skb = tmp, tmp = skb->next)
3854 #define skb_queue_reverse_walk(queue, skb) \
3855 for (skb = (queue)->prev; \
3856 skb != (struct sk_buff *)(queue); \
3859 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3860 for (skb = (queue)->prev, tmp = skb->prev; \
3861 skb != (struct sk_buff *)(queue); \
3862 skb = tmp, tmp = skb->prev)
3864 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3865 for (tmp = skb->prev; \
3866 skb != (struct sk_buff *)(queue); \
3867 skb = tmp, tmp = skb->prev)
3869 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3871 return skb_shinfo(skb)->frag_list != NULL;
3874 static inline void skb_frag_list_init(struct sk_buff *skb)
3876 skb_shinfo(skb)->frag_list = NULL;
3879 #define skb_walk_frags(skb, iter) \
3880 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3883 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3884 int *err, long *timeo_p,
3885 const struct sk_buff *skb);
3886 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3887 struct sk_buff_head *queue,
3890 struct sk_buff **last);
3891 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3892 struct sk_buff_head *queue,
3893 unsigned int flags, int *off, int *err,
3894 struct sk_buff **last);
3895 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3896 struct sk_buff_head *sk_queue,
3897 unsigned int flags, int *off, int *err);
3898 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
3899 __poll_t datagram_poll(struct file *file, struct socket *sock,
3900 struct poll_table_struct *wait);
3901 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3902 struct iov_iter *to, int size);
3903 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3904 struct msghdr *msg, int size)
3906 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3908 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3909 struct msghdr *msg);
3910 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3911 struct iov_iter *to, int len,
3912 struct ahash_request *hash);
3913 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3914 struct iov_iter *from, int len);
3915 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3916 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3917 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3918 static inline void skb_free_datagram_locked(struct sock *sk,
3919 struct sk_buff *skb)
3921 __skb_free_datagram_locked(sk, skb, 0);
3923 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3924 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3925 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3926 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3928 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3929 struct pipe_inode_info *pipe, unsigned int len,
3930 unsigned int flags);
3931 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3933 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3934 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3935 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3936 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3938 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3939 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3940 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3941 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3942 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3943 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3944 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3945 unsigned int offset);
3946 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3947 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
3948 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3949 int skb_vlan_pop(struct sk_buff *skb);
3950 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3951 int skb_eth_pop(struct sk_buff *skb);
3952 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3953 const unsigned char *src);
3954 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3955 int mac_len, bool ethernet);
3956 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3958 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3959 int skb_mpls_dec_ttl(struct sk_buff *skb);
3960 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3963 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3965 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3968 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3970 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3973 struct skb_checksum_ops {
3974 __wsum (*update)(const void *mem, int len, __wsum wsum);
3975 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3978 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3980 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3981 __wsum csum, const struct skb_checksum_ops *ops);
3982 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3985 static inline void * __must_check
3986 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3987 const void *data, int hlen, void *buffer)
3989 if (likely(hlen - offset >= len))
3990 return (void *)data + offset;
3992 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3998 static inline void * __must_check
3999 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
4001 return __skb_header_pointer(skb, offset, len, skb->data,
4002 skb_headlen(skb), buffer);
4006 * skb_needs_linearize - check if we need to linearize a given skb
4007 * depending on the given device features.
4008 * @skb: socket buffer to check
4009 * @features: net device features
4011 * Returns true if either:
4012 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
4013 * 2. skb is fragmented and the device does not support SG.
4015 static inline bool skb_needs_linearize(struct sk_buff *skb,
4016 netdev_features_t features)
4018 return skb_is_nonlinear(skb) &&
4019 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4020 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4023 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4025 const unsigned int len)
4027 memcpy(to, skb->data, len);
4030 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4031 const int offset, void *to,
4032 const unsigned int len)
4034 memcpy(to, skb->data + offset, len);
4037 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4039 const unsigned int len)
4041 memcpy(skb->data, from, len);
4044 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4047 const unsigned int len)
4049 memcpy(skb->data + offset, from, len);
4052 void skb_init(void);
4054 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4060 * skb_get_timestamp - get timestamp from a skb
4061 * @skb: skb to get stamp from
4062 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
4064 * Timestamps are stored in the skb as offsets to a base timestamp.
4065 * This function converts the offset back to a struct timeval and stores
4068 static inline void skb_get_timestamp(const struct sk_buff *skb,
4069 struct __kernel_old_timeval *stamp)
4071 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
4074 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4075 struct __kernel_sock_timeval *stamp)
4077 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4079 stamp->tv_sec = ts.tv_sec;
4080 stamp->tv_usec = ts.tv_nsec / 1000;
4083 static inline void skb_get_timestampns(const struct sk_buff *skb,
4084 struct __kernel_old_timespec *stamp)
4086 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4088 stamp->tv_sec = ts.tv_sec;
4089 stamp->tv_nsec = ts.tv_nsec;
4092 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4093 struct __kernel_timespec *stamp)
4095 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4097 stamp->tv_sec = ts.tv_sec;
4098 stamp->tv_nsec = ts.tv_nsec;
4101 static inline void __net_timestamp(struct sk_buff *skb)
4103 skb->tstamp = ktime_get_real();
4104 skb->mono_delivery_time = 0;
4107 static inline ktime_t net_timedelta(ktime_t t)
4109 return ktime_sub(ktime_get_real(), t);
4112 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4116 skb->mono_delivery_time = kt && mono;
4119 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4121 /* It is used in the ingress path to clear the delivery_time.
4122 * If needed, set the skb->tstamp to the (rcv) timestamp.
4124 static inline void skb_clear_delivery_time(struct sk_buff *skb)
4126 if (skb->mono_delivery_time) {
4127 skb->mono_delivery_time = 0;
4128 if (static_branch_unlikely(&netstamp_needed_key))
4129 skb->tstamp = ktime_get_real();
4135 static inline void skb_clear_tstamp(struct sk_buff *skb)
4137 if (skb->mono_delivery_time)
4143 static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4145 if (skb->mono_delivery_time)
4151 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4153 if (!skb->mono_delivery_time && skb->tstamp)
4156 if (static_branch_unlikely(&netstamp_needed_key) || cond)
4157 return ktime_get_real();
4162 static inline u8 skb_metadata_len(const struct sk_buff *skb)
4164 return skb_shinfo(skb)->meta_len;
4167 static inline void *skb_metadata_end(const struct sk_buff *skb)
4169 return skb_mac_header(skb);
4172 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4173 const struct sk_buff *skb_b,
4176 const void *a = skb_metadata_end(skb_a);
4177 const void *b = skb_metadata_end(skb_b);
4178 /* Using more efficient varaiant than plain call to memcmp(). */
4179 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
4183 #define __it(x, op) (x -= sizeof(u##op))
4184 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4185 case 32: diffs |= __it_diff(a, b, 64);
4187 case 24: diffs |= __it_diff(a, b, 64);
4189 case 16: diffs |= __it_diff(a, b, 64);
4191 case 8: diffs |= __it_diff(a, b, 64);
4193 case 28: diffs |= __it_diff(a, b, 64);
4195 case 20: diffs |= __it_diff(a, b, 64);
4197 case 12: diffs |= __it_diff(a, b, 64);
4199 case 4: diffs |= __it_diff(a, b, 32);
4204 return memcmp(a - meta_len, b - meta_len, meta_len);
4208 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4209 const struct sk_buff *skb_b)
4211 u8 len_a = skb_metadata_len(skb_a);
4212 u8 len_b = skb_metadata_len(skb_b);
4214 if (!(len_a | len_b))
4217 return len_a != len_b ?
4218 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4221 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4223 skb_shinfo(skb)->meta_len = meta_len;
4226 static inline void skb_metadata_clear(struct sk_buff *skb)
4228 skb_metadata_set(skb, 0);
4231 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4233 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4235 void skb_clone_tx_timestamp(struct sk_buff *skb);
4236 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4238 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4240 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4244 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4249 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4252 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4254 * PHY drivers may accept clones of transmitted packets for
4255 * timestamping via their phy_driver.txtstamp method. These drivers
4256 * must call this function to return the skb back to the stack with a
4259 * @skb: clone of the original outgoing packet
4260 * @hwtstamps: hardware time stamps
4263 void skb_complete_tx_timestamp(struct sk_buff *skb,
4264 struct skb_shared_hwtstamps *hwtstamps);
4266 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4267 struct skb_shared_hwtstamps *hwtstamps,
4268 struct sock *sk, int tstype);
4271 * skb_tstamp_tx - queue clone of skb with send time stamps
4272 * @orig_skb: the original outgoing packet
4273 * @hwtstamps: hardware time stamps, may be NULL if not available
4275 * If the skb has a socket associated, then this function clones the
4276 * skb (thus sharing the actual data and optional structures), stores
4277 * the optional hardware time stamping information (if non NULL) or
4278 * generates a software time stamp (otherwise), then queues the clone
4279 * to the error queue of the socket. Errors are silently ignored.
4281 void skb_tstamp_tx(struct sk_buff *orig_skb,
4282 struct skb_shared_hwtstamps *hwtstamps);
4285 * skb_tx_timestamp() - Driver hook for transmit timestamping
4287 * Ethernet MAC Drivers should call this function in their hard_xmit()
4288 * function immediately before giving the sk_buff to the MAC hardware.
4290 * Specifically, one should make absolutely sure that this function is
4291 * called before TX completion of this packet can trigger. Otherwise
4292 * the packet could potentially already be freed.
4294 * @skb: A socket buffer.
4296 static inline void skb_tx_timestamp(struct sk_buff *skb)
4298 skb_clone_tx_timestamp(skb);
4299 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4300 skb_tstamp_tx(skb, NULL);
4304 * skb_complete_wifi_ack - deliver skb with wifi status
4306 * @skb: the original outgoing packet
4307 * @acked: ack status
4310 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4312 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4313 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4315 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4317 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4319 (skb->ip_summed == CHECKSUM_PARTIAL &&
4320 skb_checksum_start_offset(skb) >= 0));
4324 * skb_checksum_complete - Calculate checksum of an entire packet
4325 * @skb: packet to process
4327 * This function calculates the checksum over the entire packet plus
4328 * the value of skb->csum. The latter can be used to supply the
4329 * checksum of a pseudo header as used by TCP/UDP. It returns the
4332 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4333 * this function can be used to verify that checksum on received
4334 * packets. In that case the function should return zero if the
4335 * checksum is correct. In particular, this function will return zero
4336 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4337 * hardware has already verified the correctness of the checksum.
4339 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4341 return skb_csum_unnecessary(skb) ?
4342 0 : __skb_checksum_complete(skb);
4345 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4347 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4348 if (skb->csum_level == 0)
4349 skb->ip_summed = CHECKSUM_NONE;
4355 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4357 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4358 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4360 } else if (skb->ip_summed == CHECKSUM_NONE) {
4361 skb->ip_summed = CHECKSUM_UNNECESSARY;
4362 skb->csum_level = 0;
4366 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4368 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4369 skb->ip_summed = CHECKSUM_NONE;
4370 skb->csum_level = 0;
4374 /* Check if we need to perform checksum complete validation.
4376 * Returns true if checksum complete is needed, false otherwise
4377 * (either checksum is unnecessary or zero checksum is allowed).
4379 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4383 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4384 skb->csum_valid = 1;
4385 __skb_decr_checksum_unnecessary(skb);
4392 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4395 #define CHECKSUM_BREAK 76
4397 /* Unset checksum-complete
4399 * Unset checksum complete can be done when packet is being modified
4400 * (uncompressed for instance) and checksum-complete value is
4403 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4405 if (skb->ip_summed == CHECKSUM_COMPLETE)
4406 skb->ip_summed = CHECKSUM_NONE;
4409 /* Validate (init) checksum based on checksum complete.
4412 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4413 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4414 * checksum is stored in skb->csum for use in __skb_checksum_complete
4415 * non-zero: value of invalid checksum
4418 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4422 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4423 if (!csum_fold(csum_add(psum, skb->csum))) {
4424 skb->csum_valid = 1;
4431 if (complete || skb->len <= CHECKSUM_BREAK) {
4434 csum = __skb_checksum_complete(skb);
4435 skb->csum_valid = !csum;
4442 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4447 /* Perform checksum validate (init). Note that this is a macro since we only
4448 * want to calculate the pseudo header which is an input function if necessary.
4449 * First we try to validate without any computation (checksum unnecessary) and
4450 * then calculate based on checksum complete calling the function to compute
4454 * 0: checksum is validated or try to in skb_checksum_complete
4455 * non-zero: value of invalid checksum
4457 #define __skb_checksum_validate(skb, proto, complete, \
4458 zero_okay, check, compute_pseudo) \
4460 __sum16 __ret = 0; \
4461 skb->csum_valid = 0; \
4462 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4463 __ret = __skb_checksum_validate_complete(skb, \
4464 complete, compute_pseudo(skb, proto)); \
4468 #define skb_checksum_init(skb, proto, compute_pseudo) \
4469 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4471 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4472 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4474 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4475 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4477 #define skb_checksum_validate_zero_check(skb, proto, check, \
4479 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4481 #define skb_checksum_simple_validate(skb) \
4482 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4484 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4486 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4489 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4491 skb->csum = ~pseudo;
4492 skb->ip_summed = CHECKSUM_COMPLETE;
4495 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4497 if (__skb_checksum_convert_check(skb)) \
4498 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4501 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4502 u16 start, u16 offset)
4504 skb->ip_summed = CHECKSUM_PARTIAL;
4505 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4506 skb->csum_offset = offset - start;
4509 /* Update skbuf and packet to reflect the remote checksum offload operation.
4510 * When called, ptr indicates the starting point for skb->csum when
4511 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4512 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4514 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4515 int start, int offset, bool nopartial)
4520 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4524 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4525 __skb_checksum_complete(skb);
4526 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4529 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4531 /* Adjust skb->csum since we changed the packet */
4532 skb->csum = csum_add(skb->csum, delta);
4535 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4537 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4538 return (void *)(skb->_nfct & NFCT_PTRMASK);
4544 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4546 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4553 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4555 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4556 skb->slow_gro |= !!nfct;
4561 #ifdef CONFIG_SKB_EXTENSIONS
4563 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4569 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4572 #if IS_ENABLED(CONFIG_MPTCP)
4575 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4578 SKB_EXT_NUM, /* must be last */
4582 * struct skb_ext - sk_buff extensions
4583 * @refcnt: 1 on allocation, deallocated on 0
4584 * @offset: offset to add to @data to obtain extension address
4585 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4586 * @data: start of extension data, variable sized
4588 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4589 * to use 'u8' types while allowing up to 2kb worth of extension data.
4593 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4594 u8 chunks; /* same */
4595 char data[] __aligned(8);
4598 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4599 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4600 struct skb_ext *ext);
4601 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4602 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4603 void __skb_ext_put(struct skb_ext *ext);
4605 static inline void skb_ext_put(struct sk_buff *skb)
4607 if (skb->active_extensions)
4608 __skb_ext_put(skb->extensions);
4611 static inline void __skb_ext_copy(struct sk_buff *dst,
4612 const struct sk_buff *src)
4614 dst->active_extensions = src->active_extensions;
4616 if (src->active_extensions) {
4617 struct skb_ext *ext = src->extensions;
4619 refcount_inc(&ext->refcnt);
4620 dst->extensions = ext;
4624 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4627 __skb_ext_copy(dst, src);
4630 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4632 return !!ext->offset[i];
4635 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4637 return skb->active_extensions & (1 << id);
4640 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4642 if (skb_ext_exist(skb, id))
4643 __skb_ext_del(skb, id);
4646 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4648 if (skb_ext_exist(skb, id)) {
4649 struct skb_ext *ext = skb->extensions;
4651 return (void *)ext + (ext->offset[id] << 3);
4657 static inline void skb_ext_reset(struct sk_buff *skb)
4659 if (unlikely(skb->active_extensions)) {
4660 __skb_ext_put(skb->extensions);
4661 skb->active_extensions = 0;
4665 static inline bool skb_has_extensions(struct sk_buff *skb)
4667 return unlikely(skb->active_extensions);
4670 static inline void skb_ext_put(struct sk_buff *skb) {}
4671 static inline void skb_ext_reset(struct sk_buff *skb) {}
4672 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4673 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4674 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4675 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4676 #endif /* CONFIG_SKB_EXTENSIONS */
4678 static inline void nf_reset_ct(struct sk_buff *skb)
4680 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4681 nf_conntrack_put(skb_nfct(skb));
4686 static inline void nf_reset_trace(struct sk_buff *skb)
4688 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4693 static inline void ipvs_reset(struct sk_buff *skb)
4695 #if IS_ENABLED(CONFIG_IP_VS)
4696 skb->ipvs_property = 0;
4700 /* Note: This doesn't put any conntrack info in dst. */
4701 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4704 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4705 dst->_nfct = src->_nfct;
4706 nf_conntrack_get(skb_nfct(src));
4708 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4710 dst->nf_trace = src->nf_trace;
4714 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4716 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4717 nf_conntrack_put(skb_nfct(dst));
4719 dst->slow_gro = src->slow_gro;
4720 __nf_copy(dst, src, true);
4723 #ifdef CONFIG_NETWORK_SECMARK
4724 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4726 to->secmark = from->secmark;
4729 static inline void skb_init_secmark(struct sk_buff *skb)
4734 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4737 static inline void skb_init_secmark(struct sk_buff *skb)
4741 static inline int secpath_exists(const struct sk_buff *skb)
4744 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4750 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4752 return !skb->destructor &&
4753 !secpath_exists(skb) &&
4755 !skb->_skb_refdst &&
4756 !skb_has_frag_list(skb);
4759 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4761 skb->queue_mapping = queue_mapping;
4764 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4766 return skb->queue_mapping;
4769 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4771 to->queue_mapping = from->queue_mapping;
4774 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4776 skb->queue_mapping = rx_queue + 1;
4779 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4781 return skb->queue_mapping - 1;
4784 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4786 return skb->queue_mapping != 0;
4789 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4791 skb->dst_pending_confirm = val;
4794 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4796 return skb->dst_pending_confirm != 0;
4799 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4802 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4808 /* Keeps track of mac header offset relative to skb->head.
4809 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4810 * For non-tunnel skb it points to skb_mac_header() and for
4811 * tunnel skb it points to outer mac header.
4812 * Keeps track of level of encapsulation of network headers.
4823 #define SKB_GSO_CB_OFFSET 32
4824 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4826 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4828 return (skb_mac_header(inner_skb) - inner_skb->head) -
4829 SKB_GSO_CB(inner_skb)->mac_offset;
4832 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4834 int new_headroom, headroom;
4837 headroom = skb_headroom(skb);
4838 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4842 new_headroom = skb_headroom(skb);
4843 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4847 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4849 /* Do not update partial checksums if remote checksum is enabled. */
4850 if (skb->remcsum_offload)
4853 SKB_GSO_CB(skb)->csum = res;
4854 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4857 /* Compute the checksum for a gso segment. First compute the checksum value
4858 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4859 * then add in skb->csum (checksum from csum_start to end of packet).
4860 * skb->csum and csum_start are then updated to reflect the checksum of the
4861 * resultant packet starting from the transport header-- the resultant checksum
4862 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4865 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4867 unsigned char *csum_start = skb_transport_header(skb);
4868 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4869 __wsum partial = SKB_GSO_CB(skb)->csum;
4871 SKB_GSO_CB(skb)->csum = res;
4872 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4874 return csum_fold(csum_partial(csum_start, plen, partial));
4877 static inline bool skb_is_gso(const struct sk_buff *skb)
4879 return skb_shinfo(skb)->gso_size;
4882 /* Note: Should be called only if skb_is_gso(skb) is true */
4883 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4885 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4888 /* Note: Should be called only if skb_is_gso(skb) is true */
4889 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4891 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4894 /* Note: Should be called only if skb_is_gso(skb) is true */
4895 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4897 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4900 static inline void skb_gso_reset(struct sk_buff *skb)
4902 skb_shinfo(skb)->gso_size = 0;
4903 skb_shinfo(skb)->gso_segs = 0;
4904 skb_shinfo(skb)->gso_type = 0;
4907 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4910 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4912 shinfo->gso_size += increment;
4915 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4918 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4920 shinfo->gso_size -= decrement;
4923 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4925 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4927 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4928 * wanted then gso_type will be set. */
4929 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4931 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4932 unlikely(shinfo->gso_type == 0)) {
4933 __skb_warn_lro_forwarding(skb);
4939 static inline void skb_forward_csum(struct sk_buff *skb)
4941 /* Unfortunately we don't support this one. Any brave souls? */
4942 if (skb->ip_summed == CHECKSUM_COMPLETE)
4943 skb->ip_summed = CHECKSUM_NONE;
4947 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4948 * @skb: skb to check
4950 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4951 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4952 * use this helper, to document places where we make this assertion.
4954 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4956 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
4959 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4961 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4962 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4963 unsigned int transport_len,
4964 __sum16(*skb_chkf)(struct sk_buff *skb));
4967 * skb_head_is_locked - Determine if the skb->head is locked down
4968 * @skb: skb to check
4970 * The head on skbs build around a head frag can be removed if they are
4971 * not cloned. This function returns true if the skb head is locked down
4972 * due to either being allocated via kmalloc, or by being a clone with
4973 * multiple references to the head.
4975 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4977 return !skb->head_frag || skb_cloned(skb);
4980 /* Local Checksum Offload.
4981 * Compute outer checksum based on the assumption that the
4982 * inner checksum will be offloaded later.
4983 * See Documentation/networking/checksum-offloads.rst for
4984 * explanation of how this works.
4985 * Fill in outer checksum adjustment (e.g. with sum of outer
4986 * pseudo-header) before calling.
4987 * Also ensure that inner checksum is in linear data area.
4989 static inline __wsum lco_csum(struct sk_buff *skb)
4991 unsigned char *csum_start = skb_checksum_start(skb);
4992 unsigned char *l4_hdr = skb_transport_header(skb);
4995 /* Start with complement of inner checksum adjustment */
4996 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4999 /* Add in checksum of our headers (incl. outer checksum
5000 * adjustment filled in by caller) and return result.
5002 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
5005 static inline bool skb_is_redirected(const struct sk_buff *skb)
5007 return skb->redirected;
5010 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5012 skb->redirected = 1;
5013 #ifdef CONFIG_NET_REDIRECT
5014 skb->from_ingress = from_ingress;
5015 if (skb->from_ingress)
5016 skb_clear_tstamp(skb);
5020 static inline void skb_reset_redirect(struct sk_buff *skb)
5022 skb->redirected = 0;
5025 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5027 return skb->csum_not_inet;
5030 static inline void skb_set_kcov_handle(struct sk_buff *skb,
5031 const u64 kcov_handle)
5034 skb->kcov_handle = kcov_handle;
5038 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5041 return skb->kcov_handle;
5047 #ifdef CONFIG_PAGE_POOL
5048 static inline void skb_mark_for_recycle(struct sk_buff *skb)
5050 skb->pp_recycle = 1;
5054 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
5056 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
5058 return page_pool_return_skb_page(virt_to_page(data));
5061 #endif /* __KERNEL__ */
5062 #endif /* _LINUX_SKBUFF_H */