ARM64: dts: devkits*: Return to standard DTS without fstab mounting
[GitHub/LineageOS/G12/android_kernel_amlogic_linux-4.9.git] / include / linux / skbuff.h
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
16
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
25
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_dissector.h>
38 #include <linux/splice.h>
39 #include <linux/in6.h>
40 #include <linux/if_packet.h>
41 #include <net/flow.h>
42
43 /* The interface for checksum offload between the stack and networking drivers
44 * is as follows...
45 *
46 * A. IP checksum related features
47 *
48 * Drivers advertise checksum offload capabilities in the features of a device.
49 * From the stack's point of view these are capabilities offered by the driver,
50 * a driver typically only advertises features that it is capable of offloading
51 * to its device.
52 *
53 * The checksum related features are:
54 *
55 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
56 * IP (one's complement) checksum for any combination
57 * of protocols or protocol layering. The checksum is
58 * computed and set in a packet per the CHECKSUM_PARTIAL
59 * interface (see below).
60 *
61 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
62 * TCP or UDP packets over IPv4. These are specifically
63 * unencapsulated packets of the form IPv4|TCP or
64 * IPv4|UDP where the Protocol field in the IPv4 header
65 * is TCP or UDP. The IPv4 header may contain IP options
66 * This feature cannot be set in features for a device
67 * with NETIF_F_HW_CSUM also set. This feature is being
68 * DEPRECATED (see below).
69 *
70 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
71 * TCP or UDP packets over IPv6. These are specifically
72 * unencapsulated packets of the form IPv6|TCP or
73 * IPv4|UDP where the Next Header field in the IPv6
74 * header is either TCP or UDP. IPv6 extension headers
75 * are not supported with this feature. This feature
76 * cannot be set in features for a device with
77 * NETIF_F_HW_CSUM also set. This feature is being
78 * DEPRECATED (see below).
79 *
80 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
81 * This flag is used only used to disable the RX checksum
82 * feature for a device. The stack will accept receive
83 * checksum indication in packets received on a device
84 * regardless of whether NETIF_F_RXCSUM is set.
85 *
86 * B. Checksumming of received packets by device. Indication of checksum
87 * verification is in set skb->ip_summed. Possible values are:
88 *
89 * CHECKSUM_NONE:
90 *
91 * Device did not checksum this packet e.g. due to lack of capabilities.
92 * The packet contains full (though not verified) checksum in packet but
93 * not in skb->csum. Thus, skb->csum is undefined in this case.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * The hardware you're dealing with doesn't calculate the full checksum
98 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
99 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
100 * if their checksums are okay. skb->csum is still undefined in this case
101 * though. A driver or device must never modify the checksum field in the
102 * packet even if checksum is verified.
103 *
104 * CHECKSUM_UNNECESSARY is applicable to following protocols:
105 * TCP: IPv6 and IPv4.
106 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
107 * zero UDP checksum for either IPv4 or IPv6, the networking stack
108 * may perform further validation in this case.
109 * GRE: only if the checksum is present in the header.
110 * SCTP: indicates the CRC in SCTP header has been validated.
111 *
112 * skb->csum_level indicates the number of consecutive checksums found in
113 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
114 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
115 * and a device is able to verify the checksums for UDP (possibly zero),
116 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
117 * two. If the device were only able to verify the UDP checksum and not
118 * GRE, either because it doesn't support GRE checksum of because GRE
119 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
120 * not considered in this case).
121 *
122 * CHECKSUM_COMPLETE:
123 *
124 * This is the most generic way. The device supplied checksum of the _whole_
125 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
126 * hardware doesn't need to parse L3/L4 headers to implement this.
127 *
128 * Note: Even if device supports only some protocols, but is able to produce
129 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
130 *
131 * CHECKSUM_PARTIAL:
132 *
133 * A checksum is set up to be offloaded to a device as described in the
134 * output description for CHECKSUM_PARTIAL. This may occur on a packet
135 * received directly from another Linux OS, e.g., a virtualized Linux kernel
136 * on the same host, or it may be set in the input path in GRO or remote
137 * checksum offload. For the purposes of checksum verification, the checksum
138 * referred to by skb->csum_start + skb->csum_offset and any preceding
139 * checksums in the packet are considered verified. Any checksums in the
140 * packet that are after the checksum being offloaded are not considered to
141 * be verified.
142 *
143 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
144 * in the skb->ip_summed for a packet. Values are:
145 *
146 * CHECKSUM_PARTIAL:
147 *
148 * The driver is required to checksum the packet as seen by hard_start_xmit()
149 * from skb->csum_start up to the end, and to record/write the checksum at
150 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
151 * csum_start and csum_offset values are valid values given the length and
152 * offset of the packet, however they should not attempt to validate that the
153 * checksum refers to a legitimate transport layer checksum-- it is the
154 * purview of the stack to validate that csum_start and csum_offset are set
155 * correctly.
156 *
157 * When the stack requests checksum offload for a packet, the driver MUST
158 * ensure that the checksum is set correctly. A driver can either offload the
159 * checksum calculation to the device, or call skb_checksum_help (in the case
160 * that the device does not support offload for a particular checksum).
161 *
162 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
163 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
164 * checksum offload capability. If a device has limited checksum capabilities
165 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
166 * described above) a helper function can be called to resolve
167 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
168 * function takes a spec argument that describes the protocol layer that is
169 * supported for checksum offload and can be called for each packet. If a
170 * packet does not match the specification for offload, skb_checksum_help
171 * is called to resolve the checksum.
172 *
173 * CHECKSUM_NONE:
174 *
175 * The skb was already checksummed by the protocol, or a checksum is not
176 * required.
177 *
178 * CHECKSUM_UNNECESSARY:
179 *
180 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
181 * output.
182 *
183 * CHECKSUM_COMPLETE:
184 * Not used in checksum output. If a driver observes a packet with this value
185 * set in skbuff, if should treat as CHECKSUM_NONE being set.
186 *
187 * D. Non-IP checksum (CRC) offloads
188 *
189 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
190 * offloading the SCTP CRC in a packet. To perform this offload the stack
191 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
192 * accordingly. Note the there is no indication in the skbuff that the
193 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
194 * both IP checksum offload and SCTP CRC offload must verify which offload
195 * is configured for a packet presumably by inspecting packet headers.
196 *
197 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
198 * offloading the FCOE CRC in a packet. To perform this offload the stack
199 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
200 * accordingly. Note the there is no indication in the skbuff that the
201 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
202 * both IP checksum offload and FCOE CRC offload must verify which offload
203 * is configured for a packet presumably by inspecting packet headers.
204 *
205 * E. Checksumming on output with GSO.
206 *
207 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
208 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
209 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
210 * part of the GSO operation is implied. If a checksum is being offloaded
211 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
212 * are set to refer to the outermost checksum being offload (two offloaded
213 * checksums are possible with UDP encapsulation).
214 */
215
216 /* Don't change this without changing skb_csum_unnecessary! */
217 #define CHECKSUM_NONE 0
218 #define CHECKSUM_UNNECESSARY 1
219 #define CHECKSUM_COMPLETE 2
220 #define CHECKSUM_PARTIAL 3
221
222 /* Maximum value in skb->csum_level */
223 #define SKB_MAX_CSUM_LEVEL 3
224
225 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
226 #define SKB_WITH_OVERHEAD(X) \
227 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
228 #define SKB_MAX_ORDER(X, ORDER) \
229 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
230 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
231 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
232
233 /* return minimum truesize of one skb containing X bytes of data */
234 #define SKB_TRUESIZE(X) ((X) + \
235 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
236 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
237
238 struct net_device;
239 struct scatterlist;
240 struct pipe_inode_info;
241 struct iov_iter;
242 struct napi_struct;
243
244 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
245 struct nf_conntrack {
246 atomic_t use;
247 };
248 #endif
249
250 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
251 struct nf_bridge_info {
252 atomic_t use;
253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
257 } orig_proto:8;
258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
261 __u16 frag_max_size;
262 struct net_device *physindev;
263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
266 union {
267 /* prerouting: detect dnat in orig/reply direction */
268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
276 };
277 };
278 #endif
279
280 struct sk_buff_head {
281 /* These two members must be first. */
282 struct sk_buff *next;
283 struct sk_buff *prev;
284
285 __u32 qlen;
286 spinlock_t lock;
287 };
288
289 struct sk_buff;
290
291 /* To allow 64K frame to be packed as single skb without frag_list we
292 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
293 * buffers which do not start on a page boundary.
294 *
295 * Since GRO uses frags we allocate at least 16 regardless of page
296 * size.
297 */
298 #if (65536/PAGE_SIZE + 1) < 16
299 #define MAX_SKB_FRAGS 16UL
300 #else
301 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
302 #endif
303 extern int sysctl_max_skb_frags;
304
305 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
306 * segment using its current segmentation instead.
307 */
308 #define GSO_BY_FRAGS 0xFFFF
309
310 typedef struct skb_frag_struct skb_frag_t;
311
312 struct skb_frag_struct {
313 struct {
314 struct page *p;
315 } page;
316 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
317 __u32 page_offset;
318 __u32 size;
319 #else
320 __u16 page_offset;
321 __u16 size;
322 #endif
323 };
324
325 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326 {
327 return frag->size;
328 }
329
330 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
331 {
332 frag->size = size;
333 }
334
335 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
336 {
337 frag->size += delta;
338 }
339
340 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
341 {
342 frag->size -= delta;
343 }
344
345 #define HAVE_HW_TIME_STAMP
346
347 /**
348 * struct skb_shared_hwtstamps - hardware time stamps
349 * @hwtstamp: hardware time stamp transformed into duration
350 * since arbitrary point in time
351 *
352 * Software time stamps generated by ktime_get_real() are stored in
353 * skb->tstamp.
354 *
355 * hwtstamps can only be compared against other hwtstamps from
356 * the same device.
357 *
358 * This structure is attached to packets as part of the
359 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
360 */
361 struct skb_shared_hwtstamps {
362 ktime_t hwtstamp;
363 };
364
365 /* Definitions for tx_flags in struct skb_shared_info */
366 enum {
367 /* generate hardware time stamp */
368 SKBTX_HW_TSTAMP = 1 << 0,
369
370 /* generate software time stamp when queueing packet to NIC */
371 SKBTX_SW_TSTAMP = 1 << 1,
372
373 /* device driver is going to provide hardware time stamp */
374 SKBTX_IN_PROGRESS = 1 << 2,
375
376 /* device driver supports TX zero-copy buffers */
377 SKBTX_DEV_ZEROCOPY = 1 << 3,
378
379 /* generate wifi status information (where possible) */
380 SKBTX_WIFI_STATUS = 1 << 4,
381
382 /* This indicates at least one fragment might be overwritten
383 * (as in vmsplice(), sendfile() ...)
384 * If we need to compute a TX checksum, we'll need to copy
385 * all frags to avoid possible bad checksum
386 */
387 SKBTX_SHARED_FRAG = 1 << 5,
388
389 /* generate software time stamp when entering packet scheduling */
390 SKBTX_SCHED_TSTAMP = 1 << 6,
391 };
392
393 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
394 SKBTX_SCHED_TSTAMP)
395 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
396
397 /*
398 * The callback notifies userspace to release buffers when skb DMA is done in
399 * lower device, the skb last reference should be 0 when calling this.
400 * The zerocopy_success argument is true if zero copy transmit occurred,
401 * false on data copy or out of memory error caused by data copy attempt.
402 * The ctx field is used to track device context.
403 * The desc field is used to track userspace buffer index.
404 */
405 struct ubuf_info {
406 void (*callback)(struct ubuf_info *, bool zerocopy_success);
407 void *ctx;
408 unsigned long desc;
409 };
410
411 /* This data is invariant across clones and lives at
412 * the end of the header data, ie. at skb->end.
413 */
414 struct skb_shared_info {
415 unsigned char nr_frags;
416 __u8 tx_flags;
417 unsigned short gso_size;
418 /* Warning: this field is not always filled in (UFO)! */
419 unsigned short gso_segs;
420 unsigned short gso_type;
421 struct sk_buff *frag_list;
422 struct skb_shared_hwtstamps hwtstamps;
423 u32 tskey;
424 __be32 ip6_frag_id;
425
426 /*
427 * Warning : all fields before dataref are cleared in __alloc_skb()
428 */
429 atomic_t dataref;
430
431 /* Intermediate layers must ensure that destructor_arg
432 * remains valid until skb destructor */
433 void * destructor_arg;
434
435 /* must be last field, see pskb_expand_head() */
436 skb_frag_t frags[MAX_SKB_FRAGS];
437 };
438
439 /* We divide dataref into two halves. The higher 16 bits hold references
440 * to the payload part of skb->data. The lower 16 bits hold references to
441 * the entire skb->data. A clone of a headerless skb holds the length of
442 * the header in skb->hdr_len.
443 *
444 * All users must obey the rule that the skb->data reference count must be
445 * greater than or equal to the payload reference count.
446 *
447 * Holding a reference to the payload part means that the user does not
448 * care about modifications to the header part of skb->data.
449 */
450 #define SKB_DATAREF_SHIFT 16
451 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
452
453
454 enum {
455 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
456 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
457 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
458 };
459
460 enum {
461 SKB_GSO_TCPV4 = 1 << 0,
462 SKB_GSO_UDP = 1 << 1,
463
464 /* This indicates the skb is from an untrusted source. */
465 SKB_GSO_DODGY = 1 << 2,
466
467 /* This indicates the tcp segment has CWR set. */
468 SKB_GSO_TCP_ECN = 1 << 3,
469
470 SKB_GSO_TCP_FIXEDID = 1 << 4,
471
472 SKB_GSO_TCPV6 = 1 << 5,
473
474 SKB_GSO_FCOE = 1 << 6,
475
476 SKB_GSO_GRE = 1 << 7,
477
478 SKB_GSO_GRE_CSUM = 1 << 8,
479
480 SKB_GSO_IPXIP4 = 1 << 9,
481
482 SKB_GSO_IPXIP6 = 1 << 10,
483
484 SKB_GSO_UDP_TUNNEL = 1 << 11,
485
486 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
487
488 SKB_GSO_PARTIAL = 1 << 13,
489
490 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
491
492 SKB_GSO_SCTP = 1 << 15,
493 };
494
495 #if BITS_PER_LONG > 32
496 #define NET_SKBUFF_DATA_USES_OFFSET 1
497 #endif
498
499 #ifdef NET_SKBUFF_DATA_USES_OFFSET
500 typedef unsigned int sk_buff_data_t;
501 #else
502 typedef unsigned char *sk_buff_data_t;
503 #endif
504
505 /**
506 * struct skb_mstamp - multi resolution time stamps
507 * @stamp_us: timestamp in us resolution
508 * @stamp_jiffies: timestamp in jiffies
509 */
510 struct skb_mstamp {
511 union {
512 u64 v64;
513 struct {
514 u32 stamp_us;
515 u32 stamp_jiffies;
516 };
517 };
518 };
519
520 /**
521 * skb_mstamp_get - get current timestamp
522 * @cl: place to store timestamps
523 */
524 static inline void skb_mstamp_get(struct skb_mstamp *cl)
525 {
526 u64 val = local_clock();
527
528 do_div(val, NSEC_PER_USEC);
529 cl->stamp_us = (u32)val;
530 cl->stamp_jiffies = (u32)jiffies;
531 }
532
533 /**
534 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
535 * @t1: pointer to newest sample
536 * @t0: pointer to oldest sample
537 */
538 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
539 const struct skb_mstamp *t0)
540 {
541 s32 delta_us = t1->stamp_us - t0->stamp_us;
542 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
543
544 /* If delta_us is negative, this might be because interval is too big,
545 * or local_clock() drift is too big : fallback using jiffies.
546 */
547 if (delta_us <= 0 ||
548 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
549
550 delta_us = jiffies_to_usecs(delta_jiffies);
551
552 return delta_us;
553 }
554
555 static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
556 const struct skb_mstamp *t0)
557 {
558 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
559
560 if (!diff)
561 diff = t1->stamp_us - t0->stamp_us;
562 return diff > 0;
563 }
564
565 /**
566 * struct sk_buff - socket buffer
567 * @next: Next buffer in list
568 * @prev: Previous buffer in list
569 * @tstamp: Time we arrived/left
570 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
571 * @sk: Socket we are owned by
572 * @dev: Device we arrived on/are leaving by
573 * @cb: Control buffer. Free for use by every layer. Put private vars here
574 * @_skb_refdst: destination entry (with norefcount bit)
575 * @sp: the security path, used for xfrm
576 * @len: Length of actual data
577 * @data_len: Data length
578 * @mac_len: Length of link layer header
579 * @hdr_len: writable header length of cloned skb
580 * @csum: Checksum (must include start/offset pair)
581 * @csum_start: Offset from skb->head where checksumming should start
582 * @csum_offset: Offset from csum_start where checksum should be stored
583 * @priority: Packet queueing priority
584 * @ignore_df: allow local fragmentation
585 * @cloned: Head may be cloned (check refcnt to be sure)
586 * @ip_summed: Driver fed us an IP checksum
587 * @nohdr: Payload reference only, must not modify header
588 * @nfctinfo: Relationship of this skb to the connection
589 * @pkt_type: Packet class
590 * @fclone: skbuff clone status
591 * @ipvs_property: skbuff is owned by ipvs
592 * @peeked: this packet has been seen already, so stats have been
593 * done for it, don't do them again
594 * @nf_trace: netfilter packet trace flag
595 * @protocol: Packet protocol from driver
596 * @destructor: Destruct function
597 * @nfct: Associated connection, if any
598 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
599 * @skb_iif: ifindex of device we arrived on
600 * @tc_index: Traffic control index
601 * @tc_verd: traffic control verdict
602 * @hash: the packet hash
603 * @queue_mapping: Queue mapping for multiqueue devices
604 * @xmit_more: More SKBs are pending for this queue
605 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
606 * @ndisc_nodetype: router type (from link layer)
607 * @ooo_okay: allow the mapping of a socket to a queue to be changed
608 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
609 * ports.
610 * @sw_hash: indicates hash was computed in software stack
611 * @wifi_acked_valid: wifi_acked was set
612 * @wifi_acked: whether frame was acked on wifi or not
613 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
614 * @napi_id: id of the NAPI struct this skb came from
615 * @secmark: security marking
616 * @mark: Generic packet mark
617 * @vlan_proto: vlan encapsulation protocol
618 * @vlan_tci: vlan tag control information
619 * @inner_protocol: Protocol (encapsulation)
620 * @inner_transport_header: Inner transport layer header (encapsulation)
621 * @inner_network_header: Network layer header (encapsulation)
622 * @inner_mac_header: Link layer header (encapsulation)
623 * @transport_header: Transport layer header
624 * @network_header: Network layer header
625 * @mac_header: Link layer header
626 * @tail: Tail pointer
627 * @end: End pointer
628 * @head: Head of buffer
629 * @data: Data head pointer
630 * @truesize: Buffer size
631 * @users: User count - see {datagram,tcp}.c
632 */
633
634 struct sk_buff {
635 union {
636 struct {
637 /* These two members must be first. */
638 struct sk_buff *next;
639 struct sk_buff *prev;
640
641 union {
642 ktime_t tstamp;
643 struct skb_mstamp skb_mstamp;
644 };
645 };
646 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
647 };
648
649 union {
650 struct sock *sk;
651 int ip_defrag_offset;
652 };
653
654 struct net_device *dev;
655
656 /*
657 * This is the control buffer. It is free to use for every
658 * layer. Please put your private variables there. If you
659 * want to keep them across layers you have to do a skb_clone()
660 * first. This is owned by whoever has the skb queued ATM.
661 */
662 char cb[48] __aligned(8);
663
664 unsigned long _skb_refdst;
665 void (*destructor)(struct sk_buff *skb);
666 #ifdef CONFIG_XFRM
667 struct sec_path *sp;
668 #endif
669 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
670 struct nf_conntrack *nfct;
671 #endif
672 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
673 struct nf_bridge_info *nf_bridge;
674 #endif
675 unsigned int len,
676 data_len;
677 __u16 mac_len,
678 hdr_len;
679
680 /* Following fields are _not_ copied in __copy_skb_header()
681 * Note that queue_mapping is here mostly to fill a hole.
682 */
683 kmemcheck_bitfield_begin(flags1);
684 __u16 queue_mapping;
685
686 /* if you move cloned around you also must adapt those constants */
687 #ifdef __BIG_ENDIAN_BITFIELD
688 #define CLONED_MASK (1 << 7)
689 #else
690 #define CLONED_MASK 1
691 #endif
692 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
693
694 __u8 __cloned_offset[0];
695 __u8 cloned:1,
696 nohdr:1,
697 fclone:2,
698 peeked:1,
699 head_frag:1,
700 xmit_more:1,
701 pfmemalloc:1;
702 kmemcheck_bitfield_end(flags1);
703
704 /* fields enclosed in headers_start/headers_end are copied
705 * using a single memcpy() in __copy_skb_header()
706 */
707 /* private: */
708 __u32 headers_start[0];
709 /* public: */
710
711 /* if you move pkt_type around you also must adapt those constants */
712 #ifdef __BIG_ENDIAN_BITFIELD
713 #define PKT_TYPE_MAX (7 << 5)
714 #else
715 #define PKT_TYPE_MAX 7
716 #endif
717 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
718
719 __u8 __pkt_type_offset[0];
720 __u8 pkt_type:3;
721 __u8 ignore_df:1;
722 __u8 nfctinfo:3;
723 __u8 nf_trace:1;
724
725 __u8 ip_summed:2;
726 __u8 ooo_okay:1;
727 __u8 l4_hash:1;
728 __u8 sw_hash:1;
729 __u8 wifi_acked_valid:1;
730 __u8 wifi_acked:1;
731 __u8 no_fcs:1;
732
733 /* Indicates the inner headers are valid in the skbuff. */
734 __u8 encapsulation:1;
735 __u8 encap_hdr_csum:1;
736 __u8 csum_valid:1;
737 __u8 csum_complete_sw:1;
738 __u8 csum_level:2;
739 __u8 csum_bad:1;
740 #ifdef CONFIG_IPV6_NDISC_NODETYPE
741 __u8 ndisc_nodetype:2;
742 #endif
743 __u8 ipvs_property:1;
744
745 __u8 inner_protocol_type:1;
746 __u8 remcsum_offload:1;
747 #ifdef CONFIG_NET_SWITCHDEV
748 __u8 offload_fwd_mark:1;
749 #endif
750 /* 2, 4 or 5 bit hole */
751
752 #ifdef CONFIG_NET_SCHED
753 __u16 tc_index; /* traffic control index */
754 #ifdef CONFIG_NET_CLS_ACT
755 __u16 tc_verd; /* traffic control verdict */
756 #endif
757 #endif
758
759 union {
760 __wsum csum;
761 struct {
762 __u16 csum_start;
763 __u16 csum_offset;
764 };
765 };
766 __u32 priority;
767 int skb_iif;
768 __u32 hash;
769 __be16 vlan_proto;
770 __u16 vlan_tci;
771 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
772 union {
773 unsigned int napi_id;
774 unsigned int sender_cpu;
775 };
776 #endif
777 #ifdef CONFIG_NETWORK_SECMARK
778 __u32 secmark;
779 #endif
780
781 union {
782 __u32 mark;
783 __u32 reserved_tailroom;
784 };
785
786 union {
787 __be16 inner_protocol;
788 __u8 inner_ipproto;
789 };
790
791 __u16 inner_transport_header;
792 __u16 inner_network_header;
793 __u16 inner_mac_header;
794
795 __be16 protocol;
796 __u16 transport_header;
797 __u16 network_header;
798 __u16 mac_header;
799
800 /* private: */
801 __u32 headers_end[0];
802 /* public: */
803
804 /* These elements must be at the end, see alloc_skb() for details. */
805 sk_buff_data_t tail;
806 sk_buff_data_t end;
807 unsigned char *head,
808 *data;
809 unsigned int truesize;
810 atomic_t users;
811 };
812
813 #ifdef __KERNEL__
814 /*
815 * Handling routines are only of interest to the kernel
816 */
817 #include <linux/slab.h>
818
819
820 #define SKB_ALLOC_FCLONE 0x01
821 #define SKB_ALLOC_RX 0x02
822 #define SKB_ALLOC_NAPI 0x04
823
824 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
825 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
826 {
827 return unlikely(skb->pfmemalloc);
828 }
829
830 /*
831 * skb might have a dst pointer attached, refcounted or not.
832 * _skb_refdst low order bit is set if refcount was _not_ taken
833 */
834 #define SKB_DST_NOREF 1UL
835 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
836
837 /**
838 * skb_dst - returns skb dst_entry
839 * @skb: buffer
840 *
841 * Returns skb dst_entry, regardless of reference taken or not.
842 */
843 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
844 {
845 /* If refdst was not refcounted, check we still are in a
846 * rcu_read_lock section
847 */
848 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
849 !rcu_read_lock_held() &&
850 !rcu_read_lock_bh_held());
851 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
852 }
853
854 /**
855 * skb_dst_set - sets skb dst
856 * @skb: buffer
857 * @dst: dst entry
858 *
859 * Sets skb dst, assuming a reference was taken on dst and should
860 * be released by skb_dst_drop()
861 */
862 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
863 {
864 skb->_skb_refdst = (unsigned long)dst;
865 }
866
867 /**
868 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
869 * @skb: buffer
870 * @dst: dst entry
871 *
872 * Sets skb dst, assuming a reference was not taken on dst.
873 * If dst entry is cached, we do not take reference and dst_release
874 * will be avoided by refdst_drop. If dst entry is not cached, we take
875 * reference, so that last dst_release can destroy the dst immediately.
876 */
877 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
878 {
879 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
880 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
881 }
882
883 /**
884 * skb_dst_is_noref - Test if skb dst isn't refcounted
885 * @skb: buffer
886 */
887 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
888 {
889 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
890 }
891
892 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
893 {
894 return (struct rtable *)skb_dst(skb);
895 }
896
897 /* For mangling skb->pkt_type from user space side from applications
898 * such as nft, tc, etc, we only allow a conservative subset of
899 * possible pkt_types to be set.
900 */
901 static inline bool skb_pkt_type_ok(u32 ptype)
902 {
903 return ptype <= PACKET_OTHERHOST;
904 }
905
906 void kfree_skb(struct sk_buff *skb);
907 void kfree_skb_list(struct sk_buff *segs);
908 void skb_tx_error(struct sk_buff *skb);
909 void consume_skb(struct sk_buff *skb);
910 void __kfree_skb(struct sk_buff *skb);
911 extern struct kmem_cache *skbuff_head_cache;
912
913 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
914 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
915 bool *fragstolen, int *delta_truesize);
916
917 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
918 int node);
919 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
920 struct sk_buff *build_skb(void *data, unsigned int frag_size);
921 static inline struct sk_buff *alloc_skb(unsigned int size,
922 gfp_t priority)
923 {
924 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
925 }
926
927 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
928 unsigned long data_len,
929 int max_page_order,
930 int *errcode,
931 gfp_t gfp_mask);
932
933 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
934 struct sk_buff_fclones {
935 struct sk_buff skb1;
936
937 struct sk_buff skb2;
938
939 atomic_t fclone_ref;
940 };
941
942 /**
943 * skb_fclone_busy - check if fclone is busy
944 * @sk: socket
945 * @skb: buffer
946 *
947 * Returns true if skb is a fast clone, and its clone is not freed.
948 * Some drivers call skb_orphan() in their ndo_start_xmit(),
949 * so we also check that this didnt happen.
950 */
951 static inline bool skb_fclone_busy(const struct sock *sk,
952 const struct sk_buff *skb)
953 {
954 const struct sk_buff_fclones *fclones;
955
956 fclones = container_of(skb, struct sk_buff_fclones, skb1);
957
958 return skb->fclone == SKB_FCLONE_ORIG &&
959 atomic_read(&fclones->fclone_ref) > 1 &&
960 fclones->skb2.sk == sk;
961 }
962
963 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
964 gfp_t priority)
965 {
966 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
967 }
968
969 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
970 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
971 {
972 return __alloc_skb_head(priority, -1);
973 }
974
975 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
976 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
977 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
978 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
979 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
980 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
981 gfp_t gfp_mask, bool fclone);
982 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
983 gfp_t gfp_mask)
984 {
985 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
986 }
987
988 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
989 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
990 unsigned int headroom);
991 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
992 int newtailroom, gfp_t priority);
993 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
994 int offset, int len);
995 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
996 int offset, int len);
997 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
998 int skb_pad(struct sk_buff *skb, int pad);
999 #define dev_kfree_skb(a) consume_skb(a)
1000
1001 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1002 int getfrag(void *from, char *to, int offset,
1003 int len, int odd, struct sk_buff *skb),
1004 void *from, int length);
1005
1006 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1007 int offset, size_t size);
1008
1009 struct skb_seq_state {
1010 __u32 lower_offset;
1011 __u32 upper_offset;
1012 __u32 frag_idx;
1013 __u32 stepped_offset;
1014 struct sk_buff *root_skb;
1015 struct sk_buff *cur_skb;
1016 __u8 *frag_data;
1017 };
1018
1019 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1020 unsigned int to, struct skb_seq_state *st);
1021 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1022 struct skb_seq_state *st);
1023 void skb_abort_seq_read(struct skb_seq_state *st);
1024
1025 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1026 unsigned int to, struct ts_config *config);
1027
1028 /*
1029 * Packet hash types specify the type of hash in skb_set_hash.
1030 *
1031 * Hash types refer to the protocol layer addresses which are used to
1032 * construct a packet's hash. The hashes are used to differentiate or identify
1033 * flows of the protocol layer for the hash type. Hash types are either
1034 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1035 *
1036 * Properties of hashes:
1037 *
1038 * 1) Two packets in different flows have different hash values
1039 * 2) Two packets in the same flow should have the same hash value
1040 *
1041 * A hash at a higher layer is considered to be more specific. A driver should
1042 * set the most specific hash possible.
1043 *
1044 * A driver cannot indicate a more specific hash than the layer at which a hash
1045 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1046 *
1047 * A driver may indicate a hash level which is less specific than the
1048 * actual layer the hash was computed on. For instance, a hash computed
1049 * at L4 may be considered an L3 hash. This should only be done if the
1050 * driver can't unambiguously determine that the HW computed the hash at
1051 * the higher layer. Note that the "should" in the second property above
1052 * permits this.
1053 */
1054 enum pkt_hash_types {
1055 PKT_HASH_TYPE_NONE, /* Undefined type */
1056 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1057 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1058 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1059 };
1060
1061 static inline void skb_clear_hash(struct sk_buff *skb)
1062 {
1063 skb->hash = 0;
1064 skb->sw_hash = 0;
1065 skb->l4_hash = 0;
1066 }
1067
1068 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1069 {
1070 if (!skb->l4_hash)
1071 skb_clear_hash(skb);
1072 }
1073
1074 static inline void
1075 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1076 {
1077 skb->l4_hash = is_l4;
1078 skb->sw_hash = is_sw;
1079 skb->hash = hash;
1080 }
1081
1082 static inline void
1083 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1084 {
1085 /* Used by drivers to set hash from HW */
1086 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1087 }
1088
1089 static inline void
1090 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1091 {
1092 __skb_set_hash(skb, hash, true, is_l4);
1093 }
1094
1095 void __skb_get_hash(struct sk_buff *skb);
1096 u32 __skb_get_hash_symmetric(struct sk_buff *skb);
1097 u32 skb_get_poff(const struct sk_buff *skb);
1098 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1099 const struct flow_keys *keys, int hlen);
1100 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1101 void *data, int hlen_proto);
1102
1103 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1104 int thoff, u8 ip_proto)
1105 {
1106 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1107 }
1108
1109 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1110 const struct flow_dissector_key *key,
1111 unsigned int key_count);
1112
1113 bool __skb_flow_dissect(const struct sk_buff *skb,
1114 struct flow_dissector *flow_dissector,
1115 void *target_container,
1116 void *data, __be16 proto, int nhoff, int hlen,
1117 unsigned int flags);
1118
1119 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1120 struct flow_dissector *flow_dissector,
1121 void *target_container, unsigned int flags)
1122 {
1123 return __skb_flow_dissect(skb, flow_dissector, target_container,
1124 NULL, 0, 0, 0, flags);
1125 }
1126
1127 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1128 struct flow_keys *flow,
1129 unsigned int flags)
1130 {
1131 memset(flow, 0, sizeof(*flow));
1132 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
1133 NULL, 0, 0, 0, flags);
1134 }
1135
1136 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1137 void *data, __be16 proto,
1138 int nhoff, int hlen,
1139 unsigned int flags)
1140 {
1141 memset(flow, 0, sizeof(*flow));
1142 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
1143 data, proto, nhoff, hlen, flags);
1144 }
1145
1146 static inline __u32 skb_get_hash(struct sk_buff *skb)
1147 {
1148 if (!skb->l4_hash && !skb->sw_hash)
1149 __skb_get_hash(skb);
1150
1151 return skb->hash;
1152 }
1153
1154 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
1155
1156 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1157 {
1158 if (!skb->l4_hash && !skb->sw_hash) {
1159 struct flow_keys keys;
1160 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1161
1162 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1163 }
1164
1165 return skb->hash;
1166 }
1167
1168 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
1169
1170 static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
1171 {
1172 if (!skb->l4_hash && !skb->sw_hash) {
1173 struct flow_keys keys;
1174 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
1175
1176 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1177 }
1178
1179 return skb->hash;
1180 }
1181
1182 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1183 const siphash_key_t *perturb);
1184
1185 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1186 {
1187 return skb->hash;
1188 }
1189
1190 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1191 {
1192 to->hash = from->hash;
1193 to->sw_hash = from->sw_hash;
1194 to->l4_hash = from->l4_hash;
1195 };
1196
1197 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1198 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1199 {
1200 return skb->head + skb->end;
1201 }
1202
1203 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1204 {
1205 return skb->end;
1206 }
1207 #else
1208 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1209 {
1210 return skb->end;
1211 }
1212
1213 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1214 {
1215 return skb->end - skb->head;
1216 }
1217 #endif
1218
1219 /* Internal */
1220 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1221
1222 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1223 {
1224 return &skb_shinfo(skb)->hwtstamps;
1225 }
1226
1227 /**
1228 * skb_queue_empty - check if a queue is empty
1229 * @list: queue head
1230 *
1231 * Returns true if the queue is empty, false otherwise.
1232 */
1233 static inline int skb_queue_empty(const struct sk_buff_head *list)
1234 {
1235 return list->next == (const struct sk_buff *) list;
1236 }
1237
1238 /**
1239 * skb_queue_is_last - check if skb is the last entry in the queue
1240 * @list: queue head
1241 * @skb: buffer
1242 *
1243 * Returns true if @skb is the last buffer on the list.
1244 */
1245 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1246 const struct sk_buff *skb)
1247 {
1248 return skb->next == (const struct sk_buff *) list;
1249 }
1250
1251 /**
1252 * skb_queue_is_first - check if skb is the first entry in the queue
1253 * @list: queue head
1254 * @skb: buffer
1255 *
1256 * Returns true if @skb is the first buffer on the list.
1257 */
1258 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1259 const struct sk_buff *skb)
1260 {
1261 return skb->prev == (const struct sk_buff *) list;
1262 }
1263
1264 /**
1265 * skb_queue_next - return the next packet in the queue
1266 * @list: queue head
1267 * @skb: current buffer
1268 *
1269 * Return the next packet in @list after @skb. It is only valid to
1270 * call this if skb_queue_is_last() evaluates to false.
1271 */
1272 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1273 const struct sk_buff *skb)
1274 {
1275 /* This BUG_ON may seem severe, but if we just return then we
1276 * are going to dereference garbage.
1277 */
1278 BUG_ON(skb_queue_is_last(list, skb));
1279 return skb->next;
1280 }
1281
1282 /**
1283 * skb_queue_prev - return the prev packet in the queue
1284 * @list: queue head
1285 * @skb: current buffer
1286 *
1287 * Return the prev packet in @list before @skb. It is only valid to
1288 * call this if skb_queue_is_first() evaluates to false.
1289 */
1290 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1291 const struct sk_buff *skb)
1292 {
1293 /* This BUG_ON may seem severe, but if we just return then we
1294 * are going to dereference garbage.
1295 */
1296 BUG_ON(skb_queue_is_first(list, skb));
1297 return skb->prev;
1298 }
1299
1300 /**
1301 * skb_get - reference buffer
1302 * @skb: buffer to reference
1303 *
1304 * Makes another reference to a socket buffer and returns a pointer
1305 * to the buffer.
1306 */
1307 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1308 {
1309 atomic_inc(&skb->users);
1310 return skb;
1311 }
1312
1313 /*
1314 * If users == 1, we are the only owner and are can avoid redundant
1315 * atomic change.
1316 */
1317
1318 /**
1319 * skb_cloned - is the buffer a clone
1320 * @skb: buffer to check
1321 *
1322 * Returns true if the buffer was generated with skb_clone() and is
1323 * one of multiple shared copies of the buffer. Cloned buffers are
1324 * shared data so must not be written to under normal circumstances.
1325 */
1326 static inline int skb_cloned(const struct sk_buff *skb)
1327 {
1328 return skb->cloned &&
1329 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1330 }
1331
1332 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1333 {
1334 might_sleep_if(gfpflags_allow_blocking(pri));
1335
1336 if (skb_cloned(skb))
1337 return pskb_expand_head(skb, 0, 0, pri);
1338
1339 return 0;
1340 }
1341
1342 /**
1343 * skb_header_cloned - is the header a clone
1344 * @skb: buffer to check
1345 *
1346 * Returns true if modifying the header part of the buffer requires
1347 * the data to be copied.
1348 */
1349 static inline int skb_header_cloned(const struct sk_buff *skb)
1350 {
1351 int dataref;
1352
1353 if (!skb->cloned)
1354 return 0;
1355
1356 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1357 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1358 return dataref != 1;
1359 }
1360
1361 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1362 {
1363 might_sleep_if(gfpflags_allow_blocking(pri));
1364
1365 if (skb_header_cloned(skb))
1366 return pskb_expand_head(skb, 0, 0, pri);
1367
1368 return 0;
1369 }
1370
1371 /**
1372 * skb_header_release - release reference to header
1373 * @skb: buffer to operate on
1374 *
1375 * Drop a reference to the header part of the buffer. This is done
1376 * by acquiring a payload reference. You must not read from the header
1377 * part of skb->data after this.
1378 * Note : Check if you can use __skb_header_release() instead.
1379 */
1380 static inline void skb_header_release(struct sk_buff *skb)
1381 {
1382 BUG_ON(skb->nohdr);
1383 skb->nohdr = 1;
1384 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1385 }
1386
1387 /**
1388 * __skb_header_release - release reference to header
1389 * @skb: buffer to operate on
1390 *
1391 * Variant of skb_header_release() assuming skb is private to caller.
1392 * We can avoid one atomic operation.
1393 */
1394 static inline void __skb_header_release(struct sk_buff *skb)
1395 {
1396 skb->nohdr = 1;
1397 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1398 }
1399
1400
1401 /**
1402 * skb_shared - is the buffer shared
1403 * @skb: buffer to check
1404 *
1405 * Returns true if more than one person has a reference to this
1406 * buffer.
1407 */
1408 static inline int skb_shared(const struct sk_buff *skb)
1409 {
1410 return atomic_read(&skb->users) != 1;
1411 }
1412
1413 /**
1414 * skb_share_check - check if buffer is shared and if so clone it
1415 * @skb: buffer to check
1416 * @pri: priority for memory allocation
1417 *
1418 * If the buffer is shared the buffer is cloned and the old copy
1419 * drops a reference. A new clone with a single reference is returned.
1420 * If the buffer is not shared the original buffer is returned. When
1421 * being called from interrupt status or with spinlocks held pri must
1422 * be GFP_ATOMIC.
1423 *
1424 * NULL is returned on a memory allocation failure.
1425 */
1426 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1427 {
1428 might_sleep_if(gfpflags_allow_blocking(pri));
1429 if (skb_shared(skb)) {
1430 struct sk_buff *nskb = skb_clone(skb, pri);
1431
1432 if (likely(nskb))
1433 consume_skb(skb);
1434 else
1435 kfree_skb(skb);
1436 skb = nskb;
1437 }
1438 return skb;
1439 }
1440
1441 /*
1442 * Copy shared buffers into a new sk_buff. We effectively do COW on
1443 * packets to handle cases where we have a local reader and forward
1444 * and a couple of other messy ones. The normal one is tcpdumping
1445 * a packet thats being forwarded.
1446 */
1447
1448 /**
1449 * skb_unshare - make a copy of a shared buffer
1450 * @skb: buffer to check
1451 * @pri: priority for memory allocation
1452 *
1453 * If the socket buffer is a clone then this function creates a new
1454 * copy of the data, drops a reference count on the old copy and returns
1455 * the new copy with the reference count at 1. If the buffer is not a clone
1456 * the original buffer is returned. When called with a spinlock held or
1457 * from interrupt state @pri must be %GFP_ATOMIC
1458 *
1459 * %NULL is returned on a memory allocation failure.
1460 */
1461 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1462 gfp_t pri)
1463 {
1464 might_sleep_if(gfpflags_allow_blocking(pri));
1465 if (skb_cloned(skb)) {
1466 struct sk_buff *nskb = skb_copy(skb, pri);
1467
1468 /* Free our shared copy */
1469 if (likely(nskb))
1470 consume_skb(skb);
1471 else
1472 kfree_skb(skb);
1473 skb = nskb;
1474 }
1475 return skb;
1476 }
1477
1478 /**
1479 * skb_peek - peek at the head of an &sk_buff_head
1480 * @list_: list to peek at
1481 *
1482 * Peek an &sk_buff. Unlike most other operations you _MUST_
1483 * be careful with this one. A peek leaves the buffer on the
1484 * list and someone else may run off with it. You must hold
1485 * the appropriate locks or have a private queue to do this.
1486 *
1487 * Returns %NULL for an empty list or a pointer to the head element.
1488 * The reference count is not incremented and the reference is therefore
1489 * volatile. Use with caution.
1490 */
1491 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1492 {
1493 struct sk_buff *skb = list_->next;
1494
1495 if (skb == (struct sk_buff *)list_)
1496 skb = NULL;
1497 return skb;
1498 }
1499
1500 /**
1501 * skb_peek_next - peek skb following the given one from a queue
1502 * @skb: skb to start from
1503 * @list_: list to peek at
1504 *
1505 * Returns %NULL when the end of the list is met or a pointer to the
1506 * next element. The reference count is not incremented and the
1507 * reference is therefore volatile. Use with caution.
1508 */
1509 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1510 const struct sk_buff_head *list_)
1511 {
1512 struct sk_buff *next = skb->next;
1513
1514 if (next == (struct sk_buff *)list_)
1515 next = NULL;
1516 return next;
1517 }
1518
1519 /**
1520 * skb_peek_tail - peek at the tail of an &sk_buff_head
1521 * @list_: list to peek at
1522 *
1523 * Peek an &sk_buff. Unlike most other operations you _MUST_
1524 * be careful with this one. A peek leaves the buffer on the
1525 * list and someone else may run off with it. You must hold
1526 * the appropriate locks or have a private queue to do this.
1527 *
1528 * Returns %NULL for an empty list or a pointer to the tail element.
1529 * The reference count is not incremented and the reference is therefore
1530 * volatile. Use with caution.
1531 */
1532 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1533 {
1534 struct sk_buff *skb = list_->prev;
1535
1536 if (skb == (struct sk_buff *)list_)
1537 skb = NULL;
1538 return skb;
1539
1540 }
1541
1542 /**
1543 * skb_queue_len - get queue length
1544 * @list_: list to measure
1545 *
1546 * Return the length of an &sk_buff queue.
1547 */
1548 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1549 {
1550 return list_->qlen;
1551 }
1552
1553 /**
1554 * skb_queue_len_lockless - get queue length
1555 * @list_: list to measure
1556 *
1557 * Return the length of an &sk_buff queue.
1558 * This variant can be used in lockless contexts.
1559 */
1560 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1561 {
1562 return READ_ONCE(list_->qlen);
1563 }
1564
1565 /**
1566 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1567 * @list: queue to initialize
1568 *
1569 * This initializes only the list and queue length aspects of
1570 * an sk_buff_head object. This allows to initialize the list
1571 * aspects of an sk_buff_head without reinitializing things like
1572 * the spinlock. It can also be used for on-stack sk_buff_head
1573 * objects where the spinlock is known to not be used.
1574 */
1575 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1576 {
1577 list->prev = list->next = (struct sk_buff *)list;
1578 list->qlen = 0;
1579 }
1580
1581 /*
1582 * This function creates a split out lock class for each invocation;
1583 * this is needed for now since a whole lot of users of the skb-queue
1584 * infrastructure in drivers have different locking usage (in hardirq)
1585 * than the networking core (in softirq only). In the long run either the
1586 * network layer or drivers should need annotation to consolidate the
1587 * main types of usage into 3 classes.
1588 */
1589 static inline void skb_queue_head_init(struct sk_buff_head *list)
1590 {
1591 spin_lock_init(&list->lock);
1592 __skb_queue_head_init(list);
1593 }
1594
1595 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1596 struct lock_class_key *class)
1597 {
1598 skb_queue_head_init(list);
1599 lockdep_set_class(&list->lock, class);
1600 }
1601
1602 /*
1603 * Insert an sk_buff on a list.
1604 *
1605 * The "__skb_xxxx()" functions are the non-atomic ones that
1606 * can only be called with interrupts disabled.
1607 */
1608 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1609 struct sk_buff_head *list);
1610 static inline void __skb_insert(struct sk_buff *newsk,
1611 struct sk_buff *prev, struct sk_buff *next,
1612 struct sk_buff_head *list)
1613 {
1614 newsk->next = next;
1615 newsk->prev = prev;
1616 next->prev = prev->next = newsk;
1617 WRITE_ONCE(list->qlen, list->qlen + 1);
1618 }
1619
1620 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1621 struct sk_buff *prev,
1622 struct sk_buff *next)
1623 {
1624 struct sk_buff *first = list->next;
1625 struct sk_buff *last = list->prev;
1626
1627 first->prev = prev;
1628 prev->next = first;
1629
1630 last->next = next;
1631 next->prev = last;
1632 }
1633
1634 /**
1635 * skb_queue_splice - join two skb lists, this is designed for stacks
1636 * @list: the new list to add
1637 * @head: the place to add it in the first list
1638 */
1639 static inline void skb_queue_splice(const struct sk_buff_head *list,
1640 struct sk_buff_head *head)
1641 {
1642 if (!skb_queue_empty(list)) {
1643 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1644 head->qlen += list->qlen;
1645 }
1646 }
1647
1648 /**
1649 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1650 * @list: the new list to add
1651 * @head: the place to add it in the first list
1652 *
1653 * The list at @list is reinitialised
1654 */
1655 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1656 struct sk_buff_head *head)
1657 {
1658 if (!skb_queue_empty(list)) {
1659 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1660 head->qlen += list->qlen;
1661 __skb_queue_head_init(list);
1662 }
1663 }
1664
1665 /**
1666 * skb_queue_splice_tail - join two skb lists, each list being a queue
1667 * @list: the new list to add
1668 * @head: the place to add it in the first list
1669 */
1670 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1671 struct sk_buff_head *head)
1672 {
1673 if (!skb_queue_empty(list)) {
1674 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1675 head->qlen += list->qlen;
1676 }
1677 }
1678
1679 /**
1680 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1681 * @list: the new list to add
1682 * @head: the place to add it in the first list
1683 *
1684 * Each of the lists is a queue.
1685 * The list at @list is reinitialised
1686 */
1687 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1688 struct sk_buff_head *head)
1689 {
1690 if (!skb_queue_empty(list)) {
1691 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1692 head->qlen += list->qlen;
1693 __skb_queue_head_init(list);
1694 }
1695 }
1696
1697 /**
1698 * __skb_queue_after - queue a buffer at the list head
1699 * @list: list to use
1700 * @prev: place after this buffer
1701 * @newsk: buffer to queue
1702 *
1703 * Queue a buffer int the middle of a list. This function takes no locks
1704 * and you must therefore hold required locks before calling it.
1705 *
1706 * A buffer cannot be placed on two lists at the same time.
1707 */
1708 static inline void __skb_queue_after(struct sk_buff_head *list,
1709 struct sk_buff *prev,
1710 struct sk_buff *newsk)
1711 {
1712 __skb_insert(newsk, prev, prev->next, list);
1713 }
1714
1715 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1716 struct sk_buff_head *list);
1717
1718 static inline void __skb_queue_before(struct sk_buff_head *list,
1719 struct sk_buff *next,
1720 struct sk_buff *newsk)
1721 {
1722 __skb_insert(newsk, next->prev, next, list);
1723 }
1724
1725 /**
1726 * __skb_queue_head - queue a buffer at the list head
1727 * @list: list to use
1728 * @newsk: buffer to queue
1729 *
1730 * Queue a buffer at the start of a list. This function takes no locks
1731 * and you must therefore hold required locks before calling it.
1732 *
1733 * A buffer cannot be placed on two lists at the same time.
1734 */
1735 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1736 static inline void __skb_queue_head(struct sk_buff_head *list,
1737 struct sk_buff *newsk)
1738 {
1739 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1740 }
1741
1742 /**
1743 * __skb_queue_tail - queue a buffer at the list tail
1744 * @list: list to use
1745 * @newsk: buffer to queue
1746 *
1747 * Queue a buffer at the end of a list. This function takes no locks
1748 * and you must therefore hold required locks before calling it.
1749 *
1750 * A buffer cannot be placed on two lists at the same time.
1751 */
1752 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1753 static inline void __skb_queue_tail(struct sk_buff_head *list,
1754 struct sk_buff *newsk)
1755 {
1756 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1757 }
1758
1759 /*
1760 * remove sk_buff from list. _Must_ be called atomically, and with
1761 * the list known..
1762 */
1763 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1764 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1765 {
1766 struct sk_buff *next, *prev;
1767
1768 WRITE_ONCE(list->qlen, list->qlen - 1);
1769 next = skb->next;
1770 prev = skb->prev;
1771 skb->next = skb->prev = NULL;
1772 next->prev = prev;
1773 prev->next = next;
1774 }
1775
1776 /**
1777 * __skb_dequeue - remove from the head of the queue
1778 * @list: list to dequeue from
1779 *
1780 * Remove the head of the list. This function does not take any locks
1781 * so must be used with appropriate locks held only. The head item is
1782 * returned or %NULL if the list is empty.
1783 */
1784 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1785 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1786 {
1787 struct sk_buff *skb = skb_peek(list);
1788 if (skb)
1789 __skb_unlink(skb, list);
1790 return skb;
1791 }
1792
1793 /**
1794 * __skb_dequeue_tail - remove from the tail of the queue
1795 * @list: list to dequeue from
1796 *
1797 * Remove the tail of the list. This function does not take any locks
1798 * so must be used with appropriate locks held only. The tail item is
1799 * returned or %NULL if the list is empty.
1800 */
1801 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1802 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1803 {
1804 struct sk_buff *skb = skb_peek_tail(list);
1805 if (skb)
1806 __skb_unlink(skb, list);
1807 return skb;
1808 }
1809
1810
1811 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1812 {
1813 return skb->data_len;
1814 }
1815
1816 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1817 {
1818 return skb->len - skb->data_len;
1819 }
1820
1821 static inline int skb_pagelen(const struct sk_buff *skb)
1822 {
1823 int i, len = 0;
1824
1825 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1826 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1827 return len + skb_headlen(skb);
1828 }
1829
1830 /**
1831 * __skb_fill_page_desc - initialise a paged fragment in an skb
1832 * @skb: buffer containing fragment to be initialised
1833 * @i: paged fragment index to initialise
1834 * @page: the page to use for this fragment
1835 * @off: the offset to the data with @page
1836 * @size: the length of the data
1837 *
1838 * Initialises the @i'th fragment of @skb to point to &size bytes at
1839 * offset @off within @page.
1840 *
1841 * Does not take any additional reference on the fragment.
1842 */
1843 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1844 struct page *page, int off, int size)
1845 {
1846 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1847
1848 /*
1849 * Propagate page pfmemalloc to the skb if we can. The problem is
1850 * that not all callers have unique ownership of the page but rely
1851 * on page_is_pfmemalloc doing the right thing(tm).
1852 */
1853 frag->page.p = page;
1854 frag->page_offset = off;
1855 skb_frag_size_set(frag, size);
1856
1857 page = compound_head(page);
1858 if (page_is_pfmemalloc(page))
1859 skb->pfmemalloc = true;
1860 }
1861
1862 /**
1863 * skb_fill_page_desc - initialise a paged fragment in an skb
1864 * @skb: buffer containing fragment to be initialised
1865 * @i: paged fragment index to initialise
1866 * @page: the page to use for this fragment
1867 * @off: the offset to the data with @page
1868 * @size: the length of the data
1869 *
1870 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1871 * @skb to point to @size bytes at offset @off within @page. In
1872 * addition updates @skb such that @i is the last fragment.
1873 *
1874 * Does not take any additional reference on the fragment.
1875 */
1876 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1877 struct page *page, int off, int size)
1878 {
1879 __skb_fill_page_desc(skb, i, page, off, size);
1880 skb_shinfo(skb)->nr_frags = i + 1;
1881 }
1882
1883 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1884 int size, unsigned int truesize);
1885
1886 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1887 unsigned int truesize);
1888
1889 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1890 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1891 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1892
1893 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1894 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1895 {
1896 return skb->head + skb->tail;
1897 }
1898
1899 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1900 {
1901 skb->tail = skb->data - skb->head;
1902 }
1903
1904 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1905 {
1906 skb_reset_tail_pointer(skb);
1907 skb->tail += offset;
1908 }
1909
1910 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1911 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1912 {
1913 return skb->tail;
1914 }
1915
1916 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1917 {
1918 skb->tail = skb->data;
1919 }
1920
1921 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1922 {
1923 skb->tail = skb->data + offset;
1924 }
1925
1926 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1927
1928 /*
1929 * Add data to an sk_buff
1930 */
1931 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1932 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1933 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1934 {
1935 unsigned char *tmp = skb_tail_pointer(skb);
1936 SKB_LINEAR_ASSERT(skb);
1937 skb->tail += len;
1938 skb->len += len;
1939 return tmp;
1940 }
1941
1942 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1943 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1944 {
1945 skb->data -= len;
1946 skb->len += len;
1947 return skb->data;
1948 }
1949
1950 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1951 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1952 {
1953 skb->len -= len;
1954 BUG_ON(skb->len < skb->data_len);
1955 return skb->data += len;
1956 }
1957
1958 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1959 {
1960 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1961 }
1962
1963 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1964
1965 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1966 {
1967 if (len > skb_headlen(skb) &&
1968 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1969 return NULL;
1970 skb->len -= len;
1971 return skb->data += len;
1972 }
1973
1974 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1975 {
1976 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1977 }
1978
1979 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1980 {
1981 if (likely(len <= skb_headlen(skb)))
1982 return 1;
1983 if (unlikely(len > skb->len))
1984 return 0;
1985 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1986 }
1987
1988 /**
1989 * skb_headroom - bytes at buffer head
1990 * @skb: buffer to check
1991 *
1992 * Return the number of bytes of free space at the head of an &sk_buff.
1993 */
1994 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1995 {
1996 return skb->data - skb->head;
1997 }
1998
1999 /**
2000 * skb_tailroom - bytes at buffer end
2001 * @skb: buffer to check
2002 *
2003 * Return the number of bytes of free space at the tail of an sk_buff
2004 */
2005 static inline int skb_tailroom(const struct sk_buff *skb)
2006 {
2007 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2008 }
2009
2010 /**
2011 * skb_availroom - bytes at buffer end
2012 * @skb: buffer to check
2013 *
2014 * Return the number of bytes of free space at the tail of an sk_buff
2015 * allocated by sk_stream_alloc()
2016 */
2017 static inline int skb_availroom(const struct sk_buff *skb)
2018 {
2019 if (skb_is_nonlinear(skb))
2020 return 0;
2021
2022 return skb->end - skb->tail - skb->reserved_tailroom;
2023 }
2024
2025 /**
2026 * skb_reserve - adjust headroom
2027 * @skb: buffer to alter
2028 * @len: bytes to move
2029 *
2030 * Increase the headroom of an empty &sk_buff by reducing the tail
2031 * room. This is only allowed for an empty buffer.
2032 */
2033 static inline void skb_reserve(struct sk_buff *skb, int len)
2034 {
2035 skb->data += len;
2036 skb->tail += len;
2037 }
2038
2039 /**
2040 * skb_tailroom_reserve - adjust reserved_tailroom
2041 * @skb: buffer to alter
2042 * @mtu: maximum amount of headlen permitted
2043 * @needed_tailroom: minimum amount of reserved_tailroom
2044 *
2045 * Set reserved_tailroom so that headlen can be as large as possible but
2046 * not larger than mtu and tailroom cannot be smaller than
2047 * needed_tailroom.
2048 * The required headroom should already have been reserved before using
2049 * this function.
2050 */
2051 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2052 unsigned int needed_tailroom)
2053 {
2054 SKB_LINEAR_ASSERT(skb);
2055 if (mtu < skb_tailroom(skb) - needed_tailroom)
2056 /* use at most mtu */
2057 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2058 else
2059 /* use up to all available space */
2060 skb->reserved_tailroom = needed_tailroom;
2061 }
2062
2063 #define ENCAP_TYPE_ETHER 0
2064 #define ENCAP_TYPE_IPPROTO 1
2065
2066 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2067 __be16 protocol)
2068 {
2069 skb->inner_protocol = protocol;
2070 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2071 }
2072
2073 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2074 __u8 ipproto)
2075 {
2076 skb->inner_ipproto = ipproto;
2077 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2078 }
2079
2080 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2081 {
2082 skb->inner_mac_header = skb->mac_header;
2083 skb->inner_network_header = skb->network_header;
2084 skb->inner_transport_header = skb->transport_header;
2085 }
2086
2087 static inline void skb_reset_mac_len(struct sk_buff *skb)
2088 {
2089 skb->mac_len = skb->network_header - skb->mac_header;
2090 }
2091
2092 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2093 *skb)
2094 {
2095 return skb->head + skb->inner_transport_header;
2096 }
2097
2098 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2099 {
2100 return skb_inner_transport_header(skb) - skb->data;
2101 }
2102
2103 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2104 {
2105 skb->inner_transport_header = skb->data - skb->head;
2106 }
2107
2108 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2109 const int offset)
2110 {
2111 skb_reset_inner_transport_header(skb);
2112 skb->inner_transport_header += offset;
2113 }
2114
2115 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2116 {
2117 return skb->head + skb->inner_network_header;
2118 }
2119
2120 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2121 {
2122 skb->inner_network_header = skb->data - skb->head;
2123 }
2124
2125 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2126 const int offset)
2127 {
2128 skb_reset_inner_network_header(skb);
2129 skb->inner_network_header += offset;
2130 }
2131
2132 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2133 {
2134 return skb->head + skb->inner_mac_header;
2135 }
2136
2137 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2138 {
2139 skb->inner_mac_header = skb->data - skb->head;
2140 }
2141
2142 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2143 const int offset)
2144 {
2145 skb_reset_inner_mac_header(skb);
2146 skb->inner_mac_header += offset;
2147 }
2148 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2149 {
2150 return skb->transport_header != (typeof(skb->transport_header))~0U;
2151 }
2152
2153 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2154 {
2155 return skb->head + skb->transport_header;
2156 }
2157
2158 static inline void skb_reset_transport_header(struct sk_buff *skb)
2159 {
2160 skb->transport_header = skb->data - skb->head;
2161 }
2162
2163 static inline void skb_set_transport_header(struct sk_buff *skb,
2164 const int offset)
2165 {
2166 skb_reset_transport_header(skb);
2167 skb->transport_header += offset;
2168 }
2169
2170 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2171 {
2172 return skb->head + skb->network_header;
2173 }
2174
2175 static inline void skb_reset_network_header(struct sk_buff *skb)
2176 {
2177 skb->network_header = skb->data - skb->head;
2178 }
2179
2180 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2181 {
2182 skb_reset_network_header(skb);
2183 skb->network_header += offset;
2184 }
2185
2186 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2187 {
2188 return skb->head + skb->mac_header;
2189 }
2190
2191 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2192 {
2193 return skb->mac_header != (typeof(skb->mac_header))~0U;
2194 }
2195
2196 static inline void skb_reset_mac_header(struct sk_buff *skb)
2197 {
2198 skb->mac_header = skb->data - skb->head;
2199 }
2200
2201 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2202 {
2203 skb_reset_mac_header(skb);
2204 skb->mac_header += offset;
2205 }
2206
2207 static inline void skb_pop_mac_header(struct sk_buff *skb)
2208 {
2209 skb->mac_header = skb->network_header;
2210 }
2211
2212 static inline void skb_probe_transport_header(struct sk_buff *skb,
2213 const int offset_hint)
2214 {
2215 struct flow_keys keys;
2216
2217 if (skb_transport_header_was_set(skb))
2218 return;
2219 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
2220 skb_set_transport_header(skb, keys.control.thoff);
2221 else
2222 skb_set_transport_header(skb, offset_hint);
2223 }
2224
2225 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2226 {
2227 if (skb_mac_header_was_set(skb)) {
2228 const unsigned char *old_mac = skb_mac_header(skb);
2229
2230 skb_set_mac_header(skb, -skb->mac_len);
2231 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2232 }
2233 }
2234
2235 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2236 {
2237 return skb->csum_start - skb_headroom(skb);
2238 }
2239
2240 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2241 {
2242 return skb->head + skb->csum_start;
2243 }
2244
2245 static inline int skb_transport_offset(const struct sk_buff *skb)
2246 {
2247 return skb_transport_header(skb) - skb->data;
2248 }
2249
2250 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2251 {
2252 return skb->transport_header - skb->network_header;
2253 }
2254
2255 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2256 {
2257 return skb->inner_transport_header - skb->inner_network_header;
2258 }
2259
2260 static inline int skb_network_offset(const struct sk_buff *skb)
2261 {
2262 return skb_network_header(skb) - skb->data;
2263 }
2264
2265 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2266 {
2267 return skb_inner_network_header(skb) - skb->data;
2268 }
2269
2270 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2271 {
2272 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2273 }
2274
2275 /*
2276 * CPUs often take a performance hit when accessing unaligned memory
2277 * locations. The actual performance hit varies, it can be small if the
2278 * hardware handles it or large if we have to take an exception and fix it
2279 * in software.
2280 *
2281 * Since an ethernet header is 14 bytes network drivers often end up with
2282 * the IP header at an unaligned offset. The IP header can be aligned by
2283 * shifting the start of the packet by 2 bytes. Drivers should do this
2284 * with:
2285 *
2286 * skb_reserve(skb, NET_IP_ALIGN);
2287 *
2288 * The downside to this alignment of the IP header is that the DMA is now
2289 * unaligned. On some architectures the cost of an unaligned DMA is high
2290 * and this cost outweighs the gains made by aligning the IP header.
2291 *
2292 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2293 * to be overridden.
2294 */
2295 #ifndef NET_IP_ALIGN
2296 #define NET_IP_ALIGN 2
2297 #endif
2298
2299 /*
2300 * The networking layer reserves some headroom in skb data (via
2301 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2302 * the header has to grow. In the default case, if the header has to grow
2303 * 32 bytes or less we avoid the reallocation.
2304 *
2305 * Unfortunately this headroom changes the DMA alignment of the resulting
2306 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2307 * on some architectures. An architecture can override this value,
2308 * perhaps setting it to a cacheline in size (since that will maintain
2309 * cacheline alignment of the DMA). It must be a power of 2.
2310 *
2311 * Various parts of the networking layer expect at least 32 bytes of
2312 * headroom, you should not reduce this.
2313 *
2314 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2315 * to reduce average number of cache lines per packet.
2316 * get_rps_cpus() for example only access one 64 bytes aligned block :
2317 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2318 */
2319 #ifndef NET_SKB_PAD
2320 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2321 #endif
2322
2323 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2324
2325 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2326 {
2327 if (unlikely(skb_is_nonlinear(skb))) {
2328 WARN_ON(1);
2329 return;
2330 }
2331 skb->len = len;
2332 skb_set_tail_pointer(skb, len);
2333 }
2334
2335 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2336 {
2337 __skb_set_length(skb, len);
2338 }
2339
2340 void skb_trim(struct sk_buff *skb, unsigned int len);
2341
2342 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2343 {
2344 if (skb->data_len)
2345 return ___pskb_trim(skb, len);
2346 __skb_trim(skb, len);
2347 return 0;
2348 }
2349
2350 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2351 {
2352 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2353 }
2354
2355 /**
2356 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2357 * @skb: buffer to alter
2358 * @len: new length
2359 *
2360 * This is identical to pskb_trim except that the caller knows that
2361 * the skb is not cloned so we should never get an error due to out-
2362 * of-memory.
2363 */
2364 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2365 {
2366 int err = pskb_trim(skb, len);
2367 BUG_ON(err);
2368 }
2369
2370 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2371 {
2372 unsigned int diff = len - skb->len;
2373
2374 if (skb_tailroom(skb) < diff) {
2375 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2376 GFP_ATOMIC);
2377 if (ret)
2378 return ret;
2379 }
2380 __skb_set_length(skb, len);
2381 return 0;
2382 }
2383
2384 /**
2385 * skb_orphan - orphan a buffer
2386 * @skb: buffer to orphan
2387 *
2388 * If a buffer currently has an owner then we call the owner's
2389 * destructor function and make the @skb unowned. The buffer continues
2390 * to exist but is no longer charged to its former owner.
2391 */
2392 static inline void skb_orphan(struct sk_buff *skb)
2393 {
2394 if (skb->destructor) {
2395 skb->destructor(skb);
2396 skb->destructor = NULL;
2397 skb->sk = NULL;
2398 } else {
2399 BUG_ON(skb->sk);
2400 }
2401 }
2402
2403 /**
2404 * skb_orphan_frags - orphan the frags contained in a buffer
2405 * @skb: buffer to orphan frags from
2406 * @gfp_mask: allocation mask for replacement pages
2407 *
2408 * For each frag in the SKB which needs a destructor (i.e. has an
2409 * owner) create a copy of that frag and release the original
2410 * page by calling the destructor.
2411 */
2412 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2413 {
2414 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2415 return 0;
2416 return skb_copy_ubufs(skb, gfp_mask);
2417 }
2418
2419 /**
2420 * __skb_queue_purge - empty a list
2421 * @list: list to empty
2422 *
2423 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2424 * the list and one reference dropped. This function does not take the
2425 * list lock and the caller must hold the relevant locks to use it.
2426 */
2427 void skb_queue_purge(struct sk_buff_head *list);
2428 static inline void __skb_queue_purge(struct sk_buff_head *list)
2429 {
2430 struct sk_buff *skb;
2431 while ((skb = __skb_dequeue(list)) != NULL)
2432 kfree_skb(skb);
2433 }
2434
2435 unsigned int skb_rbtree_purge(struct rb_root *root);
2436
2437 void *netdev_alloc_frag(unsigned int fragsz);
2438
2439 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2440 gfp_t gfp_mask);
2441
2442 /**
2443 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2444 * @dev: network device to receive on
2445 * @length: length to allocate
2446 *
2447 * Allocate a new &sk_buff and assign it a usage count of one. The
2448 * buffer has unspecified headroom built in. Users should allocate
2449 * the headroom they think they need without accounting for the
2450 * built in space. The built in space is used for optimisations.
2451 *
2452 * %NULL is returned if there is no free memory. Although this function
2453 * allocates memory it can be called from an interrupt.
2454 */
2455 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2456 unsigned int length)
2457 {
2458 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2459 }
2460
2461 /* legacy helper around __netdev_alloc_skb() */
2462 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2463 gfp_t gfp_mask)
2464 {
2465 return __netdev_alloc_skb(NULL, length, gfp_mask);
2466 }
2467
2468 /* legacy helper around netdev_alloc_skb() */
2469 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2470 {
2471 return netdev_alloc_skb(NULL, length);
2472 }
2473
2474
2475 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2476 unsigned int length, gfp_t gfp)
2477 {
2478 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2479
2480 if (NET_IP_ALIGN && skb)
2481 skb_reserve(skb, NET_IP_ALIGN);
2482 return skb;
2483 }
2484
2485 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2486 unsigned int length)
2487 {
2488 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2489 }
2490
2491 static inline void skb_free_frag(void *addr)
2492 {
2493 __free_page_frag(addr);
2494 }
2495
2496 void *napi_alloc_frag(unsigned int fragsz);
2497 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2498 unsigned int length, gfp_t gfp_mask);
2499 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2500 unsigned int length)
2501 {
2502 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2503 }
2504 void napi_consume_skb(struct sk_buff *skb, int budget);
2505
2506 void __kfree_skb_flush(void);
2507 void __kfree_skb_defer(struct sk_buff *skb);
2508
2509 /**
2510 * __dev_alloc_pages - allocate page for network Rx
2511 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2512 * @order: size of the allocation
2513 *
2514 * Allocate a new page.
2515 *
2516 * %NULL is returned if there is no free memory.
2517 */
2518 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2519 unsigned int order)
2520 {
2521 /* This piece of code contains several assumptions.
2522 * 1. This is for device Rx, therefor a cold page is preferred.
2523 * 2. The expectation is the user wants a compound page.
2524 * 3. If requesting a order 0 page it will not be compound
2525 * due to the check to see if order has a value in prep_new_page
2526 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2527 * code in gfp_to_alloc_flags that should be enforcing this.
2528 */
2529 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2530
2531 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2532 }
2533
2534 static inline struct page *dev_alloc_pages(unsigned int order)
2535 {
2536 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2537 }
2538
2539 /**
2540 * __dev_alloc_page - allocate a page for network Rx
2541 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2542 *
2543 * Allocate a new page.
2544 *
2545 * %NULL is returned if there is no free memory.
2546 */
2547 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2548 {
2549 return __dev_alloc_pages(gfp_mask, 0);
2550 }
2551
2552 static inline struct page *dev_alloc_page(void)
2553 {
2554 return dev_alloc_pages(0);
2555 }
2556
2557 /**
2558 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2559 * @page: The page that was allocated from skb_alloc_page
2560 * @skb: The skb that may need pfmemalloc set
2561 */
2562 static inline void skb_propagate_pfmemalloc(struct page *page,
2563 struct sk_buff *skb)
2564 {
2565 if (page_is_pfmemalloc(page))
2566 skb->pfmemalloc = true;
2567 }
2568
2569 /**
2570 * skb_frag_page - retrieve the page referred to by a paged fragment
2571 * @frag: the paged fragment
2572 *
2573 * Returns the &struct page associated with @frag.
2574 */
2575 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2576 {
2577 return frag->page.p;
2578 }
2579
2580 /**
2581 * __skb_frag_ref - take an addition reference on a paged fragment.
2582 * @frag: the paged fragment
2583 *
2584 * Takes an additional reference on the paged fragment @frag.
2585 */
2586 static inline void __skb_frag_ref(skb_frag_t *frag)
2587 {
2588 get_page(skb_frag_page(frag));
2589 }
2590
2591 /**
2592 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2593 * @skb: the buffer
2594 * @f: the fragment offset.
2595 *
2596 * Takes an additional reference on the @f'th paged fragment of @skb.
2597 */
2598 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2599 {
2600 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2601 }
2602
2603 /**
2604 * __skb_frag_unref - release a reference on a paged fragment.
2605 * @frag: the paged fragment
2606 *
2607 * Releases a reference on the paged fragment @frag.
2608 */
2609 static inline void __skb_frag_unref(skb_frag_t *frag)
2610 {
2611 put_page(skb_frag_page(frag));
2612 }
2613
2614 /**
2615 * skb_frag_unref - release a reference on a paged fragment of an skb.
2616 * @skb: the buffer
2617 * @f: the fragment offset
2618 *
2619 * Releases a reference on the @f'th paged fragment of @skb.
2620 */
2621 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2622 {
2623 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2624 }
2625
2626 /**
2627 * skb_frag_address - gets the address of the data contained in a paged fragment
2628 * @frag: the paged fragment buffer
2629 *
2630 * Returns the address of the data within @frag. The page must already
2631 * be mapped.
2632 */
2633 static inline void *skb_frag_address(const skb_frag_t *frag)
2634 {
2635 return page_address(skb_frag_page(frag)) + frag->page_offset;
2636 }
2637
2638 /**
2639 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2640 * @frag: the paged fragment buffer
2641 *
2642 * Returns the address of the data within @frag. Checks that the page
2643 * is mapped and returns %NULL otherwise.
2644 */
2645 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2646 {
2647 void *ptr = page_address(skb_frag_page(frag));
2648 if (unlikely(!ptr))
2649 return NULL;
2650
2651 return ptr + frag->page_offset;
2652 }
2653
2654 /**
2655 * __skb_frag_set_page - sets the page contained in a paged fragment
2656 * @frag: the paged fragment
2657 * @page: the page to set
2658 *
2659 * Sets the fragment @frag to contain @page.
2660 */
2661 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2662 {
2663 frag->page.p = page;
2664 }
2665
2666 /**
2667 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2668 * @skb: the buffer
2669 * @f: the fragment offset
2670 * @page: the page to set
2671 *
2672 * Sets the @f'th fragment of @skb to contain @page.
2673 */
2674 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2675 struct page *page)
2676 {
2677 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2678 }
2679
2680 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2681
2682 /**
2683 * skb_frag_dma_map - maps a paged fragment via the DMA API
2684 * @dev: the device to map the fragment to
2685 * @frag: the paged fragment to map
2686 * @offset: the offset within the fragment (starting at the
2687 * fragment's own offset)
2688 * @size: the number of bytes to map
2689 * @dir: the direction of the mapping (%PCI_DMA_*)
2690 *
2691 * Maps the page associated with @frag to @device.
2692 */
2693 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2694 const skb_frag_t *frag,
2695 size_t offset, size_t size,
2696 enum dma_data_direction dir)
2697 {
2698 return dma_map_page(dev, skb_frag_page(frag),
2699 frag->page_offset + offset, size, dir);
2700 }
2701
2702 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2703 gfp_t gfp_mask)
2704 {
2705 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2706 }
2707
2708
2709 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2710 gfp_t gfp_mask)
2711 {
2712 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2713 }
2714
2715
2716 /**
2717 * skb_clone_writable - is the header of a clone writable
2718 * @skb: buffer to check
2719 * @len: length up to which to write
2720 *
2721 * Returns true if modifying the header part of the cloned buffer
2722 * does not requires the data to be copied.
2723 */
2724 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2725 {
2726 return !skb_header_cloned(skb) &&
2727 skb_headroom(skb) + len <= skb->hdr_len;
2728 }
2729
2730 static inline int skb_try_make_writable(struct sk_buff *skb,
2731 unsigned int write_len)
2732 {
2733 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2734 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2735 }
2736
2737 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2738 int cloned)
2739 {
2740 int delta = 0;
2741
2742 if (headroom > skb_headroom(skb))
2743 delta = headroom - skb_headroom(skb);
2744
2745 if (delta || cloned)
2746 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2747 GFP_ATOMIC);
2748 return 0;
2749 }
2750
2751 /**
2752 * skb_cow - copy header of skb when it is required
2753 * @skb: buffer to cow
2754 * @headroom: needed headroom
2755 *
2756 * If the skb passed lacks sufficient headroom or its data part
2757 * is shared, data is reallocated. If reallocation fails, an error
2758 * is returned and original skb is not changed.
2759 *
2760 * The result is skb with writable area skb->head...skb->tail
2761 * and at least @headroom of space at head.
2762 */
2763 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2764 {
2765 return __skb_cow(skb, headroom, skb_cloned(skb));
2766 }
2767
2768 /**
2769 * skb_cow_head - skb_cow but only making the head writable
2770 * @skb: buffer to cow
2771 * @headroom: needed headroom
2772 *
2773 * This function is identical to skb_cow except that we replace the
2774 * skb_cloned check by skb_header_cloned. It should be used when
2775 * you only need to push on some header and do not need to modify
2776 * the data.
2777 */
2778 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2779 {
2780 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2781 }
2782
2783 /**
2784 * skb_padto - pad an skbuff up to a minimal size
2785 * @skb: buffer to pad
2786 * @len: minimal length
2787 *
2788 * Pads up a buffer to ensure the trailing bytes exist and are
2789 * blanked. If the buffer already contains sufficient data it
2790 * is untouched. Otherwise it is extended. Returns zero on
2791 * success. The skb is freed on error.
2792 */
2793 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2794 {
2795 unsigned int size = skb->len;
2796 if (likely(size >= len))
2797 return 0;
2798 return skb_pad(skb, len - size);
2799 }
2800
2801 /**
2802 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2803 * @skb: buffer to pad
2804 * @len: minimal length
2805 *
2806 * Pads up a buffer to ensure the trailing bytes exist and are
2807 * blanked. If the buffer already contains sufficient data it
2808 * is untouched. Otherwise it is extended. Returns zero on
2809 * success. The skb is freed on error.
2810 */
2811 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
2812 {
2813 unsigned int size = skb->len;
2814
2815 if (unlikely(size < len)) {
2816 len -= size;
2817 if (skb_pad(skb, len))
2818 return -ENOMEM;
2819 __skb_put(skb, len);
2820 }
2821 return 0;
2822 }
2823
2824 static inline int skb_add_data(struct sk_buff *skb,
2825 struct iov_iter *from, int copy)
2826 {
2827 const int off = skb->len;
2828
2829 if (skb->ip_summed == CHECKSUM_NONE) {
2830 __wsum csum = 0;
2831 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2832 &csum, from) == copy) {
2833 skb->csum = csum_block_add(skb->csum, csum, off);
2834 return 0;
2835 }
2836 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
2837 return 0;
2838
2839 __skb_trim(skb, off);
2840 return -EFAULT;
2841 }
2842
2843 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2844 const struct page *page, int off)
2845 {
2846 if (i) {
2847 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2848
2849 return page == skb_frag_page(frag) &&
2850 off == frag->page_offset + skb_frag_size(frag);
2851 }
2852 return false;
2853 }
2854
2855 static inline int __skb_linearize(struct sk_buff *skb)
2856 {
2857 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2858 }
2859
2860 /**
2861 * skb_linearize - convert paged skb to linear one
2862 * @skb: buffer to linarize
2863 *
2864 * If there is no free memory -ENOMEM is returned, otherwise zero
2865 * is returned and the old skb data released.
2866 */
2867 static inline int skb_linearize(struct sk_buff *skb)
2868 {
2869 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2870 }
2871
2872 /**
2873 * skb_has_shared_frag - can any frag be overwritten
2874 * @skb: buffer to test
2875 *
2876 * Return true if the skb has at least one frag that might be modified
2877 * by an external entity (as in vmsplice()/sendfile())
2878 */
2879 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2880 {
2881 return skb_is_nonlinear(skb) &&
2882 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2883 }
2884
2885 /**
2886 * skb_linearize_cow - make sure skb is linear and writable
2887 * @skb: buffer to process
2888 *
2889 * If there is no free memory -ENOMEM is returned, otherwise zero
2890 * is returned and the old skb data released.
2891 */
2892 static inline int skb_linearize_cow(struct sk_buff *skb)
2893 {
2894 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2895 __skb_linearize(skb) : 0;
2896 }
2897
2898 static __always_inline void
2899 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2900 unsigned int off)
2901 {
2902 if (skb->ip_summed == CHECKSUM_COMPLETE)
2903 skb->csum = csum_block_sub(skb->csum,
2904 csum_partial(start, len, 0), off);
2905 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2906 skb_checksum_start_offset(skb) < 0)
2907 skb->ip_summed = CHECKSUM_NONE;
2908 }
2909
2910 /**
2911 * skb_postpull_rcsum - update checksum for received skb after pull
2912 * @skb: buffer to update
2913 * @start: start of data before pull
2914 * @len: length of data pulled
2915 *
2916 * After doing a pull on a received packet, you need to call this to
2917 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2918 * CHECKSUM_NONE so that it can be recomputed from scratch.
2919 */
2920 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2921 const void *start, unsigned int len)
2922 {
2923 __skb_postpull_rcsum(skb, start, len, 0);
2924 }
2925
2926 static __always_inline void
2927 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2928 unsigned int off)
2929 {
2930 if (skb->ip_summed == CHECKSUM_COMPLETE)
2931 skb->csum = csum_block_add(skb->csum,
2932 csum_partial(start, len, 0), off);
2933 }
2934
2935 /**
2936 * skb_postpush_rcsum - update checksum for received skb after push
2937 * @skb: buffer to update
2938 * @start: start of data after push
2939 * @len: length of data pushed
2940 *
2941 * After doing a push on a received packet, you need to call this to
2942 * update the CHECKSUM_COMPLETE checksum.
2943 */
2944 static inline void skb_postpush_rcsum(struct sk_buff *skb,
2945 const void *start, unsigned int len)
2946 {
2947 __skb_postpush_rcsum(skb, start, len, 0);
2948 }
2949
2950 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2951
2952 /**
2953 * skb_push_rcsum - push skb and update receive checksum
2954 * @skb: buffer to update
2955 * @len: length of data pulled
2956 *
2957 * This function performs an skb_push on the packet and updates
2958 * the CHECKSUM_COMPLETE checksum. It should be used on
2959 * receive path processing instead of skb_push unless you know
2960 * that the checksum difference is zero (e.g., a valid IP header)
2961 * or you are setting ip_summed to CHECKSUM_NONE.
2962 */
2963 static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2964 unsigned int len)
2965 {
2966 skb_push(skb, len);
2967 skb_postpush_rcsum(skb, skb->data, len);
2968 return skb->data;
2969 }
2970
2971 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
2972 /**
2973 * pskb_trim_rcsum - trim received skb and update checksum
2974 * @skb: buffer to trim
2975 * @len: new length
2976 *
2977 * This is exactly the same as pskb_trim except that it ensures the
2978 * checksum of received packets are still valid after the operation.
2979 * It can change skb pointers.
2980 */
2981
2982 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2983 {
2984 if (likely(len >= skb->len))
2985 return 0;
2986 return pskb_trim_rcsum_slow(skb, len);
2987 }
2988
2989 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2990 {
2991 if (skb->ip_summed == CHECKSUM_COMPLETE)
2992 skb->ip_summed = CHECKSUM_NONE;
2993 __skb_trim(skb, len);
2994 return 0;
2995 }
2996
2997 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2998 {
2999 if (skb->ip_summed == CHECKSUM_COMPLETE)
3000 skb->ip_summed = CHECKSUM_NONE;
3001 return __skb_grow(skb, len);
3002 }
3003
3004 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3005
3006 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3007 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3008 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3009 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3010 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3011
3012 #define skb_queue_walk(queue, skb) \
3013 for (skb = (queue)->next; \
3014 skb != (struct sk_buff *)(queue); \
3015 skb = skb->next)
3016
3017 #define skb_queue_walk_safe(queue, skb, tmp) \
3018 for (skb = (queue)->next, tmp = skb->next; \
3019 skb != (struct sk_buff *)(queue); \
3020 skb = tmp, tmp = skb->next)
3021
3022 #define skb_queue_walk_from(queue, skb) \
3023 for (; skb != (struct sk_buff *)(queue); \
3024 skb = skb->next)
3025
3026 #define skb_rbtree_walk(skb, root) \
3027 for (skb = skb_rb_first(root); skb != NULL; \
3028 skb = skb_rb_next(skb))
3029
3030 #define skb_rbtree_walk_from(skb) \
3031 for (; skb != NULL; \
3032 skb = skb_rb_next(skb))
3033
3034 #define skb_rbtree_walk_from_safe(skb, tmp) \
3035 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3036 skb = tmp)
3037
3038 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3039 for (tmp = skb->next; \
3040 skb != (struct sk_buff *)(queue); \
3041 skb = tmp, tmp = skb->next)
3042
3043 #define skb_queue_reverse_walk(queue, skb) \
3044 for (skb = (queue)->prev; \
3045 skb != (struct sk_buff *)(queue); \
3046 skb = skb->prev)
3047
3048 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3049 for (skb = (queue)->prev, tmp = skb->prev; \
3050 skb != (struct sk_buff *)(queue); \
3051 skb = tmp, tmp = skb->prev)
3052
3053 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3054 for (tmp = skb->prev; \
3055 skb != (struct sk_buff *)(queue); \
3056 skb = tmp, tmp = skb->prev)
3057
3058 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3059 {
3060 return skb_shinfo(skb)->frag_list != NULL;
3061 }
3062
3063 static inline void skb_frag_list_init(struct sk_buff *skb)
3064 {
3065 skb_shinfo(skb)->frag_list = NULL;
3066 }
3067
3068 #define skb_walk_frags(skb, iter) \
3069 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3070
3071
3072 int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3073 const struct sk_buff *skb);
3074 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3075 int *peeked, int *off, int *err,
3076 struct sk_buff **last);
3077 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3078 int *peeked, int *off, int *err);
3079 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3080 int *err);
3081 unsigned int datagram_poll(struct file *file, struct socket *sock,
3082 struct poll_table_struct *wait);
3083 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3084 struct iov_iter *to, int size);
3085 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3086 struct msghdr *msg, int size)
3087 {
3088 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3089 }
3090 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3091 struct msghdr *msg);
3092 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3093 struct iov_iter *from, int len);
3094 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3095 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3096 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3097 static inline void skb_free_datagram_locked(struct sock *sk,
3098 struct sk_buff *skb)
3099 {
3100 __skb_free_datagram_locked(sk, skb, 0);
3101 }
3102 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3103 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3104 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3105 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3106 int len, __wsum csum);
3107 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3108 struct pipe_inode_info *pipe, unsigned int len,
3109 unsigned int flags);
3110 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3111 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3112 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3113 int len, int hlen);
3114 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3115 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3116 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3117 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
3118 bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
3119 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3120 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3121 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3122 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3123 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3124 int skb_vlan_pop(struct sk_buff *skb);
3125 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3126 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3127 gfp_t gfp);
3128
3129 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3130 {
3131 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3132 }
3133
3134 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3135 {
3136 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3137 }
3138
3139 struct skb_checksum_ops {
3140 __wsum (*update)(const void *mem, int len, __wsum wsum);
3141 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3142 };
3143
3144 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3145 __wsum csum, const struct skb_checksum_ops *ops);
3146 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3147 __wsum csum);
3148
3149 static inline void * __must_check
3150 __skb_header_pointer(const struct sk_buff *skb, int offset,
3151 int len, void *data, int hlen, void *buffer)
3152 {
3153 if (hlen - offset >= len)
3154 return data + offset;
3155
3156 if (!skb ||
3157 skb_copy_bits(skb, offset, buffer, len) < 0)
3158 return NULL;
3159
3160 return buffer;
3161 }
3162
3163 static inline void * __must_check
3164 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3165 {
3166 return __skb_header_pointer(skb, offset, len, skb->data,
3167 skb_headlen(skb), buffer);
3168 }
3169
3170 /**
3171 * skb_needs_linearize - check if we need to linearize a given skb
3172 * depending on the given device features.
3173 * @skb: socket buffer to check
3174 * @features: net device features
3175 *
3176 * Returns true if either:
3177 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3178 * 2. skb is fragmented and the device does not support SG.
3179 */
3180 static inline bool skb_needs_linearize(struct sk_buff *skb,
3181 netdev_features_t features)
3182 {
3183 return skb_is_nonlinear(skb) &&
3184 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3185 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3186 }
3187
3188 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3189 void *to,
3190 const unsigned int len)
3191 {
3192 memcpy(to, skb->data, len);
3193 }
3194
3195 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3196 const int offset, void *to,
3197 const unsigned int len)
3198 {
3199 memcpy(to, skb->data + offset, len);
3200 }
3201
3202 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3203 const void *from,
3204 const unsigned int len)
3205 {
3206 memcpy(skb->data, from, len);
3207 }
3208
3209 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3210 const int offset,
3211 const void *from,
3212 const unsigned int len)
3213 {
3214 memcpy(skb->data + offset, from, len);
3215 }
3216
3217 void skb_init(void);
3218
3219 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3220 {
3221 return skb->tstamp;
3222 }
3223
3224 /**
3225 * skb_get_timestamp - get timestamp from a skb
3226 * @skb: skb to get stamp from
3227 * @stamp: pointer to struct timeval to store stamp in
3228 *
3229 * Timestamps are stored in the skb as offsets to a base timestamp.
3230 * This function converts the offset back to a struct timeval and stores
3231 * it in stamp.
3232 */
3233 static inline void skb_get_timestamp(const struct sk_buff *skb,
3234 struct timeval *stamp)
3235 {
3236 *stamp = ktime_to_timeval(skb->tstamp);
3237 }
3238
3239 static inline void skb_get_timestampns(const struct sk_buff *skb,
3240 struct timespec *stamp)
3241 {
3242 *stamp = ktime_to_timespec(skb->tstamp);
3243 }
3244
3245 static inline void __net_timestamp(struct sk_buff *skb)
3246 {
3247 skb->tstamp = ktime_get_real();
3248 }
3249
3250 static inline ktime_t net_timedelta(ktime_t t)
3251 {
3252 return ktime_sub(ktime_get_real(), t);
3253 }
3254
3255 static inline ktime_t net_invalid_timestamp(void)
3256 {
3257 return ktime_set(0, 0);
3258 }
3259
3260 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3261
3262 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3263
3264 void skb_clone_tx_timestamp(struct sk_buff *skb);
3265 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3266
3267 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3268
3269 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3270 {
3271 }
3272
3273 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3274 {
3275 return false;
3276 }
3277
3278 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3279
3280 /**
3281 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3282 *
3283 * PHY drivers may accept clones of transmitted packets for
3284 * timestamping via their phy_driver.txtstamp method. These drivers
3285 * must call this function to return the skb back to the stack with a
3286 * timestamp.
3287 *
3288 * @skb: clone of the the original outgoing packet
3289 * @hwtstamps: hardware time stamps
3290 *
3291 */
3292 void skb_complete_tx_timestamp(struct sk_buff *skb,
3293 struct skb_shared_hwtstamps *hwtstamps);
3294
3295 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3296 struct skb_shared_hwtstamps *hwtstamps,
3297 struct sock *sk, int tstype);
3298
3299 /**
3300 * skb_tstamp_tx - queue clone of skb with send time stamps
3301 * @orig_skb: the original outgoing packet
3302 * @hwtstamps: hardware time stamps, may be NULL if not available
3303 *
3304 * If the skb has a socket associated, then this function clones the
3305 * skb (thus sharing the actual data and optional structures), stores
3306 * the optional hardware time stamping information (if non NULL) or
3307 * generates a software time stamp (otherwise), then queues the clone
3308 * to the error queue of the socket. Errors are silently ignored.
3309 */
3310 void skb_tstamp_tx(struct sk_buff *orig_skb,
3311 struct skb_shared_hwtstamps *hwtstamps);
3312
3313 static inline void sw_tx_timestamp(struct sk_buff *skb)
3314 {
3315 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3316 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
3317 skb_tstamp_tx(skb, NULL);
3318 }
3319
3320 /**
3321 * skb_tx_timestamp() - Driver hook for transmit timestamping
3322 *
3323 * Ethernet MAC Drivers should call this function in their hard_xmit()
3324 * function immediately before giving the sk_buff to the MAC hardware.
3325 *
3326 * Specifically, one should make absolutely sure that this function is
3327 * called before TX completion of this packet can trigger. Otherwise
3328 * the packet could potentially already be freed.
3329 *
3330 * @skb: A socket buffer.
3331 */
3332 static inline void skb_tx_timestamp(struct sk_buff *skb)
3333 {
3334 skb_clone_tx_timestamp(skb);
3335 sw_tx_timestamp(skb);
3336 }
3337
3338 /**
3339 * skb_complete_wifi_ack - deliver skb with wifi status
3340 *
3341 * @skb: the original outgoing packet
3342 * @acked: ack status
3343 *
3344 */
3345 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3346
3347 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3348 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3349
3350 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3351 {
3352 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3353 skb->csum_valid ||
3354 (skb->ip_summed == CHECKSUM_PARTIAL &&
3355 skb_checksum_start_offset(skb) >= 0));
3356 }
3357
3358 /**
3359 * skb_checksum_complete - Calculate checksum of an entire packet
3360 * @skb: packet to process
3361 *
3362 * This function calculates the checksum over the entire packet plus
3363 * the value of skb->csum. The latter can be used to supply the
3364 * checksum of a pseudo header as used by TCP/UDP. It returns the
3365 * checksum.
3366 *
3367 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3368 * this function can be used to verify that checksum on received
3369 * packets. In that case the function should return zero if the
3370 * checksum is correct. In particular, this function will return zero
3371 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3372 * hardware has already verified the correctness of the checksum.
3373 */
3374 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3375 {
3376 return skb_csum_unnecessary(skb) ?
3377 0 : __skb_checksum_complete(skb);
3378 }
3379
3380 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3381 {
3382 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3383 if (skb->csum_level == 0)
3384 skb->ip_summed = CHECKSUM_NONE;
3385 else
3386 skb->csum_level--;
3387 }
3388 }
3389
3390 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3391 {
3392 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3393 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3394 skb->csum_level++;
3395 } else if (skb->ip_summed == CHECKSUM_NONE) {
3396 skb->ip_summed = CHECKSUM_UNNECESSARY;
3397 skb->csum_level = 0;
3398 }
3399 }
3400
3401 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3402 {
3403 /* Mark current checksum as bad (typically called from GRO
3404 * path). In the case that ip_summed is CHECKSUM_NONE
3405 * this must be the first checksum encountered in the packet.
3406 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3407 * checksum after the last one validated. For UDP, a zero
3408 * checksum can not be marked as bad.
3409 */
3410
3411 if (skb->ip_summed == CHECKSUM_NONE ||
3412 skb->ip_summed == CHECKSUM_UNNECESSARY)
3413 skb->csum_bad = 1;
3414 }
3415
3416 /* Check if we need to perform checksum complete validation.
3417 *
3418 * Returns true if checksum complete is needed, false otherwise
3419 * (either checksum is unnecessary or zero checksum is allowed).
3420 */
3421 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3422 bool zero_okay,
3423 __sum16 check)
3424 {
3425 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3426 skb->csum_valid = 1;
3427 __skb_decr_checksum_unnecessary(skb);
3428 return false;
3429 }
3430
3431 return true;
3432 }
3433
3434 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3435 * in checksum_init.
3436 */
3437 #define CHECKSUM_BREAK 76
3438
3439 /* Unset checksum-complete
3440 *
3441 * Unset checksum complete can be done when packet is being modified
3442 * (uncompressed for instance) and checksum-complete value is
3443 * invalidated.
3444 */
3445 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3446 {
3447 if (skb->ip_summed == CHECKSUM_COMPLETE)
3448 skb->ip_summed = CHECKSUM_NONE;
3449 }
3450
3451 /* Validate (init) checksum based on checksum complete.
3452 *
3453 * Return values:
3454 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3455 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3456 * checksum is stored in skb->csum for use in __skb_checksum_complete
3457 * non-zero: value of invalid checksum
3458 *
3459 */
3460 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3461 bool complete,
3462 __wsum psum)
3463 {
3464 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3465 if (!csum_fold(csum_add(psum, skb->csum))) {
3466 skb->csum_valid = 1;
3467 return 0;
3468 }
3469 } else if (skb->csum_bad) {
3470 /* ip_summed == CHECKSUM_NONE in this case */
3471 return (__force __sum16)1;
3472 }
3473
3474 skb->csum = psum;
3475
3476 if (complete || skb->len <= CHECKSUM_BREAK) {
3477 __sum16 csum;
3478
3479 csum = __skb_checksum_complete(skb);
3480 skb->csum_valid = !csum;
3481 return csum;
3482 }
3483
3484 return 0;
3485 }
3486
3487 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3488 {
3489 return 0;
3490 }
3491
3492 /* Perform checksum validate (init). Note that this is a macro since we only
3493 * want to calculate the pseudo header which is an input function if necessary.
3494 * First we try to validate without any computation (checksum unnecessary) and
3495 * then calculate based on checksum complete calling the function to compute
3496 * pseudo header.
3497 *
3498 * Return values:
3499 * 0: checksum is validated or try to in skb_checksum_complete
3500 * non-zero: value of invalid checksum
3501 */
3502 #define __skb_checksum_validate(skb, proto, complete, \
3503 zero_okay, check, compute_pseudo) \
3504 ({ \
3505 __sum16 __ret = 0; \
3506 skb->csum_valid = 0; \
3507 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3508 __ret = __skb_checksum_validate_complete(skb, \
3509 complete, compute_pseudo(skb, proto)); \
3510 __ret; \
3511 })
3512
3513 #define skb_checksum_init(skb, proto, compute_pseudo) \
3514 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3515
3516 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3517 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3518
3519 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3520 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3521
3522 #define skb_checksum_validate_zero_check(skb, proto, check, \
3523 compute_pseudo) \
3524 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3525
3526 #define skb_checksum_simple_validate(skb) \
3527 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3528
3529 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3530 {
3531 return (skb->ip_summed == CHECKSUM_NONE &&
3532 skb->csum_valid && !skb->csum_bad);
3533 }
3534
3535 static inline void __skb_checksum_convert(struct sk_buff *skb,
3536 __sum16 check, __wsum pseudo)
3537 {
3538 skb->csum = ~pseudo;
3539 skb->ip_summed = CHECKSUM_COMPLETE;
3540 }
3541
3542 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3543 do { \
3544 if (__skb_checksum_convert_check(skb)) \
3545 __skb_checksum_convert(skb, check, \
3546 compute_pseudo(skb, proto)); \
3547 } while (0)
3548
3549 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3550 u16 start, u16 offset)
3551 {
3552 skb->ip_summed = CHECKSUM_PARTIAL;
3553 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3554 skb->csum_offset = offset - start;
3555 }
3556
3557 /* Update skbuf and packet to reflect the remote checksum offload operation.
3558 * When called, ptr indicates the starting point for skb->csum when
3559 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3560 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3561 */
3562 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3563 int start, int offset, bool nopartial)
3564 {
3565 __wsum delta;
3566
3567 if (!nopartial) {
3568 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3569 return;
3570 }
3571
3572 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3573 __skb_checksum_complete(skb);
3574 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3575 }
3576
3577 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3578
3579 /* Adjust skb->csum since we changed the packet */
3580 skb->csum = csum_add(skb->csum, delta);
3581 }
3582
3583 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3584 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3585 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3586 {
3587 if (nfct && atomic_dec_and_test(&nfct->use))
3588 nf_conntrack_destroy(nfct);
3589 }
3590 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3591 {
3592 if (nfct)
3593 atomic_inc(&nfct->use);
3594 }
3595 #endif
3596 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3597 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3598 {
3599 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3600 kfree(nf_bridge);
3601 }
3602 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3603 {
3604 if (nf_bridge)
3605 atomic_inc(&nf_bridge->use);
3606 }
3607 #endif /* CONFIG_BRIDGE_NETFILTER */
3608 static inline void nf_reset(struct sk_buff *skb)
3609 {
3610 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3611 nf_conntrack_put(skb->nfct);
3612 skb->nfct = NULL;
3613 #endif
3614 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3615 nf_bridge_put(skb->nf_bridge);
3616 skb->nf_bridge = NULL;
3617 #endif
3618 }
3619
3620 static inline void nf_reset_trace(struct sk_buff *skb)
3621 {
3622 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3623 skb->nf_trace = 0;
3624 #endif
3625 }
3626
3627 static inline void ipvs_reset(struct sk_buff *skb)
3628 {
3629 #if IS_ENABLED(CONFIG_IP_VS)
3630 skb->ipvs_property = 0;
3631 #endif
3632 }
3633
3634 /* Note: This doesn't put any conntrack and bridge info in dst. */
3635 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3636 bool copy)
3637 {
3638 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3639 dst->nfct = src->nfct;
3640 nf_conntrack_get(src->nfct);
3641 if (copy)
3642 dst->nfctinfo = src->nfctinfo;
3643 #endif
3644 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3645 dst->nf_bridge = src->nf_bridge;
3646 nf_bridge_get(src->nf_bridge);
3647 #endif
3648 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3649 if (copy)
3650 dst->nf_trace = src->nf_trace;
3651 #endif
3652 }
3653
3654 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3655 {
3656 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3657 nf_conntrack_put(dst->nfct);
3658 #endif
3659 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3660 nf_bridge_put(dst->nf_bridge);
3661 #endif
3662 __nf_copy(dst, src, true);
3663 }
3664
3665 #ifdef CONFIG_NETWORK_SECMARK
3666 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3667 {
3668 to->secmark = from->secmark;
3669 }
3670
3671 static inline void skb_init_secmark(struct sk_buff *skb)
3672 {
3673 skb->secmark = 0;
3674 }
3675 #else
3676 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3677 { }
3678
3679 static inline void skb_init_secmark(struct sk_buff *skb)
3680 { }
3681 #endif
3682
3683 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3684 {
3685 return !skb->destructor &&
3686 #if IS_ENABLED(CONFIG_XFRM)
3687 !skb->sp &&
3688 #endif
3689 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3690 !skb->nfct &&
3691 #endif
3692 !skb->_skb_refdst &&
3693 !skb_has_frag_list(skb);
3694 }
3695
3696 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3697 {
3698 skb->queue_mapping = queue_mapping;
3699 }
3700
3701 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3702 {
3703 return skb->queue_mapping;
3704 }
3705
3706 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3707 {
3708 to->queue_mapping = from->queue_mapping;
3709 }
3710
3711 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3712 {
3713 skb->queue_mapping = rx_queue + 1;
3714 }
3715
3716 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3717 {
3718 return skb->queue_mapping - 1;
3719 }
3720
3721 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3722 {
3723 return skb->queue_mapping != 0;
3724 }
3725
3726 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3727 {
3728 #ifdef CONFIG_XFRM
3729 return skb->sp;
3730 #else
3731 return NULL;
3732 #endif
3733 }
3734
3735 /* Keeps track of mac header offset relative to skb->head.
3736 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3737 * For non-tunnel skb it points to skb_mac_header() and for
3738 * tunnel skb it points to outer mac header.
3739 * Keeps track of level of encapsulation of network headers.
3740 */
3741 struct skb_gso_cb {
3742 union {
3743 int mac_offset;
3744 int data_offset;
3745 };
3746 int encap_level;
3747 __wsum csum;
3748 __u16 csum_start;
3749 };
3750 #define SKB_SGO_CB_OFFSET 32
3751 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
3752
3753 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3754 {
3755 return (skb_mac_header(inner_skb) - inner_skb->head) -
3756 SKB_GSO_CB(inner_skb)->mac_offset;
3757 }
3758
3759 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3760 {
3761 int new_headroom, headroom;
3762 int ret;
3763
3764 headroom = skb_headroom(skb);
3765 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3766 if (ret)
3767 return ret;
3768
3769 new_headroom = skb_headroom(skb);
3770 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3771 return 0;
3772 }
3773
3774 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3775 {
3776 /* Do not update partial checksums if remote checksum is enabled. */
3777 if (skb->remcsum_offload)
3778 return;
3779
3780 SKB_GSO_CB(skb)->csum = res;
3781 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3782 }
3783
3784 /* Compute the checksum for a gso segment. First compute the checksum value
3785 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3786 * then add in skb->csum (checksum from csum_start to end of packet).
3787 * skb->csum and csum_start are then updated to reflect the checksum of the
3788 * resultant packet starting from the transport header-- the resultant checksum
3789 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3790 * header.
3791 */
3792 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3793 {
3794 unsigned char *csum_start = skb_transport_header(skb);
3795 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3796 __wsum partial = SKB_GSO_CB(skb)->csum;
3797
3798 SKB_GSO_CB(skb)->csum = res;
3799 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
3800
3801 return csum_fold(csum_partial(csum_start, plen, partial));
3802 }
3803
3804 static inline bool skb_is_gso(const struct sk_buff *skb)
3805 {
3806 return skb_shinfo(skb)->gso_size;
3807 }
3808
3809 /* Note: Should be called only if skb_is_gso(skb) is true */
3810 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3811 {
3812 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3813 }
3814
3815 static inline void skb_gso_reset(struct sk_buff *skb)
3816 {
3817 skb_shinfo(skb)->gso_size = 0;
3818 skb_shinfo(skb)->gso_segs = 0;
3819 skb_shinfo(skb)->gso_type = 0;
3820 }
3821
3822 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3823
3824 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3825 {
3826 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3827 * wanted then gso_type will be set. */
3828 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3829
3830 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3831 unlikely(shinfo->gso_type == 0)) {
3832 __skb_warn_lro_forwarding(skb);
3833 return true;
3834 }
3835 return false;
3836 }
3837
3838 static inline void skb_forward_csum(struct sk_buff *skb)
3839 {
3840 /* Unfortunately we don't support this one. Any brave souls? */
3841 if (skb->ip_summed == CHECKSUM_COMPLETE)
3842 skb->ip_summed = CHECKSUM_NONE;
3843 }
3844
3845 /**
3846 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3847 * @skb: skb to check
3848 *
3849 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3850 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3851 * use this helper, to document places where we make this assertion.
3852 */
3853 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3854 {
3855 #ifdef DEBUG
3856 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3857 #endif
3858 }
3859
3860 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3861
3862 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3863 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3864 unsigned int transport_len,
3865 __sum16(*skb_chkf)(struct sk_buff *skb));
3866
3867 /**
3868 * skb_head_is_locked - Determine if the skb->head is locked down
3869 * @skb: skb to check
3870 *
3871 * The head on skbs build around a head frag can be removed if they are
3872 * not cloned. This function returns true if the skb head is locked down
3873 * due to either being allocated via kmalloc, or by being a clone with
3874 * multiple references to the head.
3875 */
3876 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3877 {
3878 return !skb->head_frag || skb_cloned(skb);
3879 }
3880
3881 /**
3882 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3883 *
3884 * @skb: GSO skb
3885 *
3886 * skb_gso_network_seglen is used to determine the real size of the
3887 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3888 *
3889 * The MAC/L2 header is not accounted for.
3890 */
3891 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3892 {
3893 unsigned int hdr_len = skb_transport_header(skb) -
3894 skb_network_header(skb);
3895 return hdr_len + skb_gso_transport_seglen(skb);
3896 }
3897
3898 /**
3899 * skb_gso_mac_seglen - Return length of individual segments of a gso packet
3900 *
3901 * @skb: GSO skb
3902 *
3903 * skb_gso_mac_seglen is used to determine the real size of the
3904 * individual segments, including MAC/L2, Layer3 (IP, IPv6) and L4
3905 * headers (TCP/UDP).
3906 */
3907 static inline unsigned int skb_gso_mac_seglen(const struct sk_buff *skb)
3908 {
3909 unsigned int hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
3910 return hdr_len + skb_gso_transport_seglen(skb);
3911 }
3912
3913 /* Local Checksum Offload.
3914 * Compute outer checksum based on the assumption that the
3915 * inner checksum will be offloaded later.
3916 * See Documentation/networking/checksum-offloads.txt for
3917 * explanation of how this works.
3918 * Fill in outer checksum adjustment (e.g. with sum of outer
3919 * pseudo-header) before calling.
3920 * Also ensure that inner checksum is in linear data area.
3921 */
3922 static inline __wsum lco_csum(struct sk_buff *skb)
3923 {
3924 unsigned char *csum_start = skb_checksum_start(skb);
3925 unsigned char *l4_hdr = skb_transport_header(skb);
3926 __wsum partial;
3927
3928 /* Start with complement of inner checksum adjustment */
3929 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3930 skb->csum_offset));
3931
3932 /* Add in checksum of our headers (incl. outer checksum
3933 * adjustment filled in by caller) and return result.
3934 */
3935 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
3936 }
3937
3938 #endif /* __KERNEL__ */
3939 #endif /* _LINUX_SKBUFF_H */