]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/skbuff.h
net: add rb_to_skb() and other rb tree helpers
[mirror_ubuntu-bionic-kernel.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 #include <linux/refcount.h>
26
27 #include <linux/atomic.h>
28 #include <asm/types.h>
29 #include <linux/spinlock.h>
30 #include <linux/net.h>
31 #include <linux/textsearch.h>
32 #include <net/checksum.h>
33 #include <linux/rcupdate.h>
34 #include <linux/hrtimer.h>
35 #include <linux/dma-mapping.h>
36 #include <linux/netdev_features.h>
37 #include <linux/sched.h>
38 #include <linux/sched/clock.h>
39 #include <net/flow_dissector.h>
40 #include <linux/splice.h>
41 #include <linux/in6.h>
42 #include <linux/if_packet.h>
43 #include <net/flow.h>
44
45 /* The interface for checksum offload between the stack and networking drivers
46 * is as follows...
47 *
48 * A. IP checksum related features
49 *
50 * Drivers advertise checksum offload capabilities in the features of a device.
51 * From the stack's point of view these are capabilities offered by the driver,
52 * a driver typically only advertises features that it is capable of offloading
53 * to its device.
54 *
55 * The checksum related features are:
56 *
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
62 *
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
67 * is TCP or UDP. The IPv4 header may contain IP options
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
71 *
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
75 * IPv4|UDP where the Next Header field in the IPv6
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
81 *
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
83 * This flag is used only used to disable the RX checksum
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
87 *
88 * B. Checksumming of received packets by device. Indication of checksum
89 * verification is in set skb->ip_summed. Possible values are:
90 *
91 * CHECKSUM_NONE:
92 *
93 * Device did not checksum this packet e.g. due to lack of capabilities.
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
96 *
97 * CHECKSUM_UNNECESSARY:
98 *
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
105 *
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
113 * FCOE: indicates the CRC in FC frame has been validated.
114 *
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
119 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
120 * two. If the device were only able to verify the UDP checksum and not
121 * GRE, either because it doesn't support GRE checksum of because GRE
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
124 *
125 * CHECKSUM_COMPLETE:
126 *
127 * This is the most generic way. The device supplied checksum of the _whole_
128 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
129 * hardware doesn't need to parse L3/L4 headers to implement this.
130 *
131 * Notes:
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
135 *
136 * CHECKSUM_PARTIAL:
137 *
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
146 * be verified.
147 *
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
150 *
151 * CHECKSUM_PARTIAL:
152 *
153 * The driver is required to checksum the packet as seen by hard_start_xmit()
154 * from skb->csum_start up to the end, and to record/write the checksum at
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
157 * offset of the packet, however they should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum-- it is the
159 * purview of the stack to validate that csum_start and csum_offset are set
160 * correctly.
161 *
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
166 *
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
174 *
175 * CHECKSUM_NONE:
176 *
177 * The skb was already checksummed by the protocol, or a checksum is not
178 * required.
179 *
180 * CHECKSUM_UNNECESSARY:
181 *
182 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
183 * output.
184 *
185 * CHECKSUM_COMPLETE:
186 * Not used in checksum output. If a driver observes a packet with this value
187 * set in skbuff, if should treat as CHECKSUM_NONE being set.
188 *
189 * D. Non-IP checksum (CRC) offloads
190 *
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
193 * will set set csum_start and csum_offset accordingly, set ip_summed to
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
200 *
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
204 * accordingly. Note the there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
206 * both IP checksum offload and FCOE CRC offload must verify which offload
207 * is configured for a packet presumably by inspecting packet headers.
208 *
209 * E. Checksumming on output with GSO.
210 *
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
216 * are set to refer to the outermost checksum being offload (two offloaded
217 * checksums are possible with UDP encapsulation).
218 */
219
220 /* Don't change this without changing skb_csum_unnecessary! */
221 #define CHECKSUM_NONE 0
222 #define CHECKSUM_UNNECESSARY 1
223 #define CHECKSUM_COMPLETE 2
224 #define CHECKSUM_PARTIAL 3
225
226 /* Maximum value in skb->csum_level */
227 #define SKB_MAX_CSUM_LEVEL 3
228
229 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
230 #define SKB_WITH_OVERHEAD(X) \
231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
232 #define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
234 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
236
237 /* return minimum truesize of one skb containing X bytes of data */
238 #define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
241
242 struct net_device;
243 struct scatterlist;
244 struct pipe_inode_info;
245 struct iov_iter;
246 struct napi_struct;
247
248 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
249 struct nf_conntrack {
250 atomic_t use;
251 };
252 #endif
253
254 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
255 struct nf_bridge_info {
256 refcount_t use;
257 enum {
258 BRNF_PROTO_UNCHANGED,
259 BRNF_PROTO_8021Q,
260 BRNF_PROTO_PPPOE
261 } orig_proto:8;
262 u8 pkt_otherhost:1;
263 u8 in_prerouting:1;
264 u8 bridged_dnat:1;
265 __u16 frag_max_size;
266 struct net_device *physindev;
267
268 /* always valid & non-NULL from FORWARD on, for physdev match */
269 struct net_device *physoutdev;
270 union {
271 /* prerouting: detect dnat in orig/reply direction */
272 __be32 ipv4_daddr;
273 struct in6_addr ipv6_daddr;
274
275 /* after prerouting + nat detected: store original source
276 * mac since neigh resolution overwrites it, only used while
277 * skb is out in neigh layer.
278 */
279 char neigh_header[8];
280 };
281 };
282 #endif
283
284 struct sk_buff_head {
285 /* These two members must be first. */
286 struct sk_buff *next;
287 struct sk_buff *prev;
288
289 __u32 qlen;
290 spinlock_t lock;
291 };
292
293 struct sk_buff;
294
295 /* To allow 64K frame to be packed as single skb without frag_list we
296 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
297 * buffers which do not start on a page boundary.
298 *
299 * Since GRO uses frags we allocate at least 16 regardless of page
300 * size.
301 */
302 #if (65536/PAGE_SIZE + 1) < 16
303 #define MAX_SKB_FRAGS 16UL
304 #else
305 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
306 #endif
307 extern int sysctl_max_skb_frags;
308
309 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
310 * segment using its current segmentation instead.
311 */
312 #define GSO_BY_FRAGS 0xFFFF
313
314 typedef struct skb_frag_struct skb_frag_t;
315
316 struct skb_frag_struct {
317 struct {
318 struct page *p;
319 } page;
320 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
321 __u32 page_offset;
322 __u32 size;
323 #else
324 __u16 page_offset;
325 __u16 size;
326 #endif
327 };
328
329 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330 {
331 return frag->size;
332 }
333
334 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
335 {
336 frag->size = size;
337 }
338
339 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
340 {
341 frag->size += delta;
342 }
343
344 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
345 {
346 frag->size -= delta;
347 }
348
349 static inline bool skb_frag_must_loop(struct page *p)
350 {
351 #if defined(CONFIG_HIGHMEM)
352 if (PageHighMem(p))
353 return true;
354 #endif
355 return false;
356 }
357
358 /**
359 * skb_frag_foreach_page - loop over pages in a fragment
360 *
361 * @f: skb frag to operate on
362 * @f_off: offset from start of f->page.p
363 * @f_len: length from f_off to loop over
364 * @p: (temp var) current page
365 * @p_off: (temp var) offset from start of current page,
366 * non-zero only on first page.
367 * @p_len: (temp var) length in current page,
368 * < PAGE_SIZE only on first and last page.
369 * @copied: (temp var) length so far, excluding current p_len.
370 *
371 * A fragment can hold a compound page, in which case per-page
372 * operations, notably kmap_atomic, must be called for each
373 * regular page.
374 */
375 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
376 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
377 p_off = (f_off) & (PAGE_SIZE - 1), \
378 p_len = skb_frag_must_loop(p) ? \
379 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
380 copied = 0; \
381 copied < f_len; \
382 copied += p_len, p++, p_off = 0, \
383 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
384
385 #define HAVE_HW_TIME_STAMP
386
387 /**
388 * struct skb_shared_hwtstamps - hardware time stamps
389 * @hwtstamp: hardware time stamp transformed into duration
390 * since arbitrary point in time
391 *
392 * Software time stamps generated by ktime_get_real() are stored in
393 * skb->tstamp.
394 *
395 * hwtstamps can only be compared against other hwtstamps from
396 * the same device.
397 *
398 * This structure is attached to packets as part of the
399 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
400 */
401 struct skb_shared_hwtstamps {
402 ktime_t hwtstamp;
403 };
404
405 /* Definitions for tx_flags in struct skb_shared_info */
406 enum {
407 /* generate hardware time stamp */
408 SKBTX_HW_TSTAMP = 1 << 0,
409
410 /* generate software time stamp when queueing packet to NIC */
411 SKBTX_SW_TSTAMP = 1 << 1,
412
413 /* device driver is going to provide hardware time stamp */
414 SKBTX_IN_PROGRESS = 1 << 2,
415
416 /* device driver supports TX zero-copy buffers */
417 SKBTX_DEV_ZEROCOPY = 1 << 3,
418
419 /* generate wifi status information (where possible) */
420 SKBTX_WIFI_STATUS = 1 << 4,
421
422 /* This indicates at least one fragment might be overwritten
423 * (as in vmsplice(), sendfile() ...)
424 * If we need to compute a TX checksum, we'll need to copy
425 * all frags to avoid possible bad checksum
426 */
427 SKBTX_SHARED_FRAG = 1 << 5,
428
429 /* generate software time stamp when entering packet scheduling */
430 SKBTX_SCHED_TSTAMP = 1 << 6,
431 };
432
433 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
434 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
435 SKBTX_SCHED_TSTAMP)
436 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
437
438 /*
439 * The callback notifies userspace to release buffers when skb DMA is done in
440 * lower device, the skb last reference should be 0 when calling this.
441 * The zerocopy_success argument is true if zero copy transmit occurred,
442 * false on data copy or out of memory error caused by data copy attempt.
443 * The ctx field is used to track device context.
444 * The desc field is used to track userspace buffer index.
445 */
446 struct ubuf_info {
447 void (*callback)(struct ubuf_info *, bool zerocopy_success);
448 union {
449 struct {
450 unsigned long desc;
451 void *ctx;
452 };
453 struct {
454 u32 id;
455 u16 len;
456 u16 zerocopy:1;
457 u32 bytelen;
458 };
459 };
460 refcount_t refcnt;
461
462 struct mmpin {
463 struct user_struct *user;
464 unsigned int num_pg;
465 } mmp;
466 };
467
468 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
469
470 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
471 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
472 struct ubuf_info *uarg);
473
474 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
475 {
476 refcount_inc(&uarg->refcnt);
477 }
478
479 void sock_zerocopy_put(struct ubuf_info *uarg);
480 void sock_zerocopy_put_abort(struct ubuf_info *uarg);
481
482 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
483
484 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
485 struct msghdr *msg, int len,
486 struct ubuf_info *uarg);
487
488 /* This data is invariant across clones and lives at
489 * the end of the header data, ie. at skb->end.
490 */
491 struct skb_shared_info {
492 __u8 __unused;
493 __u8 meta_len;
494 __u8 nr_frags;
495 __u8 tx_flags;
496 unsigned short gso_size;
497 /* Warning: this field is not always filled in (UFO)! */
498 unsigned short gso_segs;
499 struct sk_buff *frag_list;
500 struct skb_shared_hwtstamps hwtstamps;
501 unsigned int gso_type;
502 u32 tskey;
503 __be32 ip6_frag_id;
504
505 /*
506 * Warning : all fields before dataref are cleared in __alloc_skb()
507 */
508 atomic_t dataref;
509
510 /* Intermediate layers must ensure that destructor_arg
511 * remains valid until skb destructor */
512 void * destructor_arg;
513
514 /* must be last field, see pskb_expand_head() */
515 skb_frag_t frags[MAX_SKB_FRAGS];
516 };
517
518 /* We divide dataref into two halves. The higher 16 bits hold references
519 * to the payload part of skb->data. The lower 16 bits hold references to
520 * the entire skb->data. A clone of a headerless skb holds the length of
521 * the header in skb->hdr_len.
522 *
523 * All users must obey the rule that the skb->data reference count must be
524 * greater than or equal to the payload reference count.
525 *
526 * Holding a reference to the payload part means that the user does not
527 * care about modifications to the header part of skb->data.
528 */
529 #define SKB_DATAREF_SHIFT 16
530 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
531
532
533 enum {
534 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
535 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
536 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
537 };
538
539 enum {
540 SKB_GSO_TCPV4 = 1 << 0,
541
542 /* This indicates the skb is from an untrusted source. */
543 SKB_GSO_DODGY = 1 << 1,
544
545 /* This indicates the tcp segment has CWR set. */
546 SKB_GSO_TCP_ECN = 1 << 2,
547
548 SKB_GSO_TCP_FIXEDID = 1 << 3,
549
550 SKB_GSO_TCPV6 = 1 << 4,
551
552 SKB_GSO_FCOE = 1 << 5,
553
554 SKB_GSO_GRE = 1 << 6,
555
556 SKB_GSO_GRE_CSUM = 1 << 7,
557
558 SKB_GSO_IPXIP4 = 1 << 8,
559
560 SKB_GSO_IPXIP6 = 1 << 9,
561
562 SKB_GSO_UDP_TUNNEL = 1 << 10,
563
564 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
565
566 SKB_GSO_PARTIAL = 1 << 12,
567
568 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
569
570 SKB_GSO_SCTP = 1 << 14,
571
572 SKB_GSO_ESP = 1 << 15,
573 };
574
575 #if BITS_PER_LONG > 32
576 #define NET_SKBUFF_DATA_USES_OFFSET 1
577 #endif
578
579 #ifdef NET_SKBUFF_DATA_USES_OFFSET
580 typedef unsigned int sk_buff_data_t;
581 #else
582 typedef unsigned char *sk_buff_data_t;
583 #endif
584
585 /**
586 * struct sk_buff - socket buffer
587 * @next: Next buffer in list
588 * @prev: Previous buffer in list
589 * @tstamp: Time we arrived/left
590 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
591 * @sk: Socket we are owned by
592 * @dev: Device we arrived on/are leaving by
593 * @cb: Control buffer. Free for use by every layer. Put private vars here
594 * @_skb_refdst: destination entry (with norefcount bit)
595 * @sp: the security path, used for xfrm
596 * @len: Length of actual data
597 * @data_len: Data length
598 * @mac_len: Length of link layer header
599 * @hdr_len: writable header length of cloned skb
600 * @csum: Checksum (must include start/offset pair)
601 * @csum_start: Offset from skb->head where checksumming should start
602 * @csum_offset: Offset from csum_start where checksum should be stored
603 * @priority: Packet queueing priority
604 * @ignore_df: allow local fragmentation
605 * @cloned: Head may be cloned (check refcnt to be sure)
606 * @ip_summed: Driver fed us an IP checksum
607 * @nohdr: Payload reference only, must not modify header
608 * @pkt_type: Packet class
609 * @fclone: skbuff clone status
610 * @ipvs_property: skbuff is owned by ipvs
611 * @tc_skip_classify: do not classify packet. set by IFB device
612 * @tc_at_ingress: used within tc_classify to distinguish in/egress
613 * @tc_redirected: packet was redirected by a tc action
614 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
615 * @peeked: this packet has been seen already, so stats have been
616 * done for it, don't do them again
617 * @nf_trace: netfilter packet trace flag
618 * @protocol: Packet protocol from driver
619 * @destructor: Destruct function
620 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
621 * @_nfct: Associated connection, if any (with nfctinfo bits)
622 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
623 * @skb_iif: ifindex of device we arrived on
624 * @tc_index: Traffic control index
625 * @hash: the packet hash
626 * @queue_mapping: Queue mapping for multiqueue devices
627 * @xmit_more: More SKBs are pending for this queue
628 * @ndisc_nodetype: router type (from link layer)
629 * @ooo_okay: allow the mapping of a socket to a queue to be changed
630 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
631 * ports.
632 * @sw_hash: indicates hash was computed in software stack
633 * @wifi_acked_valid: wifi_acked was set
634 * @wifi_acked: whether frame was acked on wifi or not
635 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
636 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
637 * @dst_pending_confirm: need to confirm neighbour
638 * @napi_id: id of the NAPI struct this skb came from
639 * @secmark: security marking
640 * @mark: Generic packet mark
641 * @vlan_proto: vlan encapsulation protocol
642 * @vlan_tci: vlan tag control information
643 * @inner_protocol: Protocol (encapsulation)
644 * @inner_transport_header: Inner transport layer header (encapsulation)
645 * @inner_network_header: Network layer header (encapsulation)
646 * @inner_mac_header: Link layer header (encapsulation)
647 * @transport_header: Transport layer header
648 * @network_header: Network layer header
649 * @mac_header: Link layer header
650 * @tail: Tail pointer
651 * @end: End pointer
652 * @head: Head of buffer
653 * @data: Data head pointer
654 * @truesize: Buffer size
655 * @users: User count - see {datagram,tcp}.c
656 */
657
658 struct sk_buff {
659 union {
660 struct {
661 /* These two members must be first. */
662 struct sk_buff *next;
663 struct sk_buff *prev;
664
665 union {
666 struct net_device *dev;
667 /* Some protocols might use this space to store information,
668 * while device pointer would be NULL.
669 * UDP receive path is one user.
670 */
671 unsigned long dev_scratch;
672 };
673 };
674 struct rb_node rbnode; /* used in netem & tcp stack */
675 };
676 struct sock *sk;
677
678 union {
679 ktime_t tstamp;
680 u64 skb_mstamp;
681 };
682 /*
683 * This is the control buffer. It is free to use for every
684 * layer. Please put your private variables there. If you
685 * want to keep them across layers you have to do a skb_clone()
686 * first. This is owned by whoever has the skb queued ATM.
687 */
688 char cb[48] __aligned(8);
689
690 union {
691 struct {
692 unsigned long _skb_refdst;
693 void (*destructor)(struct sk_buff *skb);
694 };
695 struct list_head tcp_tsorted_anchor;
696 };
697
698 #ifdef CONFIG_XFRM
699 struct sec_path *sp;
700 #endif
701 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
702 unsigned long _nfct;
703 #endif
704 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
705 struct nf_bridge_info *nf_bridge;
706 #endif
707 unsigned int len,
708 data_len;
709 __u16 mac_len,
710 hdr_len;
711
712 /* Following fields are _not_ copied in __copy_skb_header()
713 * Note that queue_mapping is here mostly to fill a hole.
714 */
715 kmemcheck_bitfield_begin(flags1);
716 __u16 queue_mapping;
717
718 /* if you move cloned around you also must adapt those constants */
719 #ifdef __BIG_ENDIAN_BITFIELD
720 #define CLONED_MASK (1 << 7)
721 #else
722 #define CLONED_MASK 1
723 #endif
724 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
725
726 __u8 __cloned_offset[0];
727 __u8 cloned:1,
728 nohdr:1,
729 fclone:2,
730 peeked:1,
731 head_frag:1,
732 xmit_more:1,
733 __unused:1; /* one bit hole */
734 kmemcheck_bitfield_end(flags1);
735
736 /* fields enclosed in headers_start/headers_end are copied
737 * using a single memcpy() in __copy_skb_header()
738 */
739 /* private: */
740 __u32 headers_start[0];
741 /* public: */
742
743 /* if you move pkt_type around you also must adapt those constants */
744 #ifdef __BIG_ENDIAN_BITFIELD
745 #define PKT_TYPE_MAX (7 << 5)
746 #else
747 #define PKT_TYPE_MAX 7
748 #endif
749 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
750
751 __u8 __pkt_type_offset[0];
752 __u8 pkt_type:3;
753 __u8 pfmemalloc:1;
754 __u8 ignore_df:1;
755
756 __u8 nf_trace:1;
757 __u8 ip_summed:2;
758 __u8 ooo_okay:1;
759 __u8 l4_hash:1;
760 __u8 sw_hash:1;
761 __u8 wifi_acked_valid:1;
762 __u8 wifi_acked:1;
763
764 __u8 no_fcs:1;
765 /* Indicates the inner headers are valid in the skbuff. */
766 __u8 encapsulation:1;
767 __u8 encap_hdr_csum:1;
768 __u8 csum_valid:1;
769 __u8 csum_complete_sw:1;
770 __u8 csum_level:2;
771 __u8 csum_not_inet:1;
772
773 __u8 dst_pending_confirm:1;
774 #ifdef CONFIG_IPV6_NDISC_NODETYPE
775 __u8 ndisc_nodetype:2;
776 #endif
777 __u8 ipvs_property:1;
778 __u8 inner_protocol_type:1;
779 __u8 remcsum_offload:1;
780 #ifdef CONFIG_NET_SWITCHDEV
781 __u8 offload_fwd_mark:1;
782 __u8 offload_mr_fwd_mark:1;
783 #endif
784 #ifdef CONFIG_NET_CLS_ACT
785 __u8 tc_skip_classify:1;
786 __u8 tc_at_ingress:1;
787 __u8 tc_redirected:1;
788 __u8 tc_from_ingress:1;
789 #endif
790
791 #ifdef CONFIG_NET_SCHED
792 __u16 tc_index; /* traffic control index */
793 #endif
794
795 union {
796 __wsum csum;
797 struct {
798 __u16 csum_start;
799 __u16 csum_offset;
800 };
801 };
802 __u32 priority;
803 int skb_iif;
804 __u32 hash;
805 __be16 vlan_proto;
806 __u16 vlan_tci;
807 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
808 union {
809 unsigned int napi_id;
810 unsigned int sender_cpu;
811 };
812 #endif
813 #ifdef CONFIG_NETWORK_SECMARK
814 __u32 secmark;
815 #endif
816
817 union {
818 __u32 mark;
819 __u32 reserved_tailroom;
820 };
821
822 union {
823 __be16 inner_protocol;
824 __u8 inner_ipproto;
825 };
826
827 __u16 inner_transport_header;
828 __u16 inner_network_header;
829 __u16 inner_mac_header;
830
831 __be16 protocol;
832 __u16 transport_header;
833 __u16 network_header;
834 __u16 mac_header;
835
836 /* private: */
837 __u32 headers_end[0];
838 /* public: */
839
840 /* These elements must be at the end, see alloc_skb() for details. */
841 sk_buff_data_t tail;
842 sk_buff_data_t end;
843 unsigned char *head,
844 *data;
845 unsigned int truesize;
846 refcount_t users;
847 };
848
849 #ifdef __KERNEL__
850 /*
851 * Handling routines are only of interest to the kernel
852 */
853 #include <linux/slab.h>
854
855
856 #define SKB_ALLOC_FCLONE 0x01
857 #define SKB_ALLOC_RX 0x02
858 #define SKB_ALLOC_NAPI 0x04
859
860 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
861 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
862 {
863 return unlikely(skb->pfmemalloc);
864 }
865
866 /*
867 * skb might have a dst pointer attached, refcounted or not.
868 * _skb_refdst low order bit is set if refcount was _not_ taken
869 */
870 #define SKB_DST_NOREF 1UL
871 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
872
873 #define SKB_NFCT_PTRMASK ~(7UL)
874 /**
875 * skb_dst - returns skb dst_entry
876 * @skb: buffer
877 *
878 * Returns skb dst_entry, regardless of reference taken or not.
879 */
880 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
881 {
882 /* If refdst was not refcounted, check we still are in a
883 * rcu_read_lock section
884 */
885 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
886 !rcu_read_lock_held() &&
887 !rcu_read_lock_bh_held());
888 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
889 }
890
891 /**
892 * skb_dst_set - sets skb dst
893 * @skb: buffer
894 * @dst: dst entry
895 *
896 * Sets skb dst, assuming a reference was taken on dst and should
897 * be released by skb_dst_drop()
898 */
899 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
900 {
901 skb->_skb_refdst = (unsigned long)dst;
902 }
903
904 /**
905 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
906 * @skb: buffer
907 * @dst: dst entry
908 *
909 * Sets skb dst, assuming a reference was not taken on dst.
910 * If dst entry is cached, we do not take reference and dst_release
911 * will be avoided by refdst_drop. If dst entry is not cached, we take
912 * reference, so that last dst_release can destroy the dst immediately.
913 */
914 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
915 {
916 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
917 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
918 }
919
920 /**
921 * skb_dst_is_noref - Test if skb dst isn't refcounted
922 * @skb: buffer
923 */
924 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
925 {
926 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
927 }
928
929 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
930 {
931 return (struct rtable *)skb_dst(skb);
932 }
933
934 /* For mangling skb->pkt_type from user space side from applications
935 * such as nft, tc, etc, we only allow a conservative subset of
936 * possible pkt_types to be set.
937 */
938 static inline bool skb_pkt_type_ok(u32 ptype)
939 {
940 return ptype <= PACKET_OTHERHOST;
941 }
942
943 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
944 {
945 #ifdef CONFIG_NET_RX_BUSY_POLL
946 return skb->napi_id;
947 #else
948 return 0;
949 #endif
950 }
951
952 /* decrement the reference count and return true if we can free the skb */
953 static inline bool skb_unref(struct sk_buff *skb)
954 {
955 if (unlikely(!skb))
956 return false;
957 if (likely(refcount_read(&skb->users) == 1))
958 smp_rmb();
959 else if (likely(!refcount_dec_and_test(&skb->users)))
960 return false;
961
962 return true;
963 }
964
965 void skb_release_head_state(struct sk_buff *skb);
966 void kfree_skb(struct sk_buff *skb);
967 void kfree_skb_list(struct sk_buff *segs);
968 void skb_tx_error(struct sk_buff *skb);
969 void consume_skb(struct sk_buff *skb);
970 void __consume_stateless_skb(struct sk_buff *skb);
971 void __kfree_skb(struct sk_buff *skb);
972 extern struct kmem_cache *skbuff_head_cache;
973
974 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
975 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
976 bool *fragstolen, int *delta_truesize);
977
978 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
979 int node);
980 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
981 struct sk_buff *build_skb(void *data, unsigned int frag_size);
982 static inline struct sk_buff *alloc_skb(unsigned int size,
983 gfp_t priority)
984 {
985 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
986 }
987
988 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
989 unsigned long data_len,
990 int max_page_order,
991 int *errcode,
992 gfp_t gfp_mask);
993
994 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
995 struct sk_buff_fclones {
996 struct sk_buff skb1;
997
998 struct sk_buff skb2;
999
1000 refcount_t fclone_ref;
1001 };
1002
1003 /**
1004 * skb_fclone_busy - check if fclone is busy
1005 * @sk: socket
1006 * @skb: buffer
1007 *
1008 * Returns true if skb is a fast clone, and its clone is not freed.
1009 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1010 * so we also check that this didnt happen.
1011 */
1012 static inline bool skb_fclone_busy(const struct sock *sk,
1013 const struct sk_buff *skb)
1014 {
1015 const struct sk_buff_fclones *fclones;
1016
1017 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1018
1019 return skb->fclone == SKB_FCLONE_ORIG &&
1020 refcount_read(&fclones->fclone_ref) > 1 &&
1021 fclones->skb2.sk == sk;
1022 }
1023
1024 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1025 gfp_t priority)
1026 {
1027 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1028 }
1029
1030 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1031 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1032 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1033 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1034 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1035 gfp_t gfp_mask, bool fclone);
1036 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1037 gfp_t gfp_mask)
1038 {
1039 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1040 }
1041
1042 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1043 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1044 unsigned int headroom);
1045 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1046 int newtailroom, gfp_t priority);
1047 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1048 int offset, int len);
1049 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1050 int offset, int len);
1051 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1052 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1053
1054 /**
1055 * skb_pad - zero pad the tail of an skb
1056 * @skb: buffer to pad
1057 * @pad: space to pad
1058 *
1059 * Ensure that a buffer is followed by a padding area that is zero
1060 * filled. Used by network drivers which may DMA or transfer data
1061 * beyond the buffer end onto the wire.
1062 *
1063 * May return error in out of memory cases. The skb is freed on error.
1064 */
1065 static inline int skb_pad(struct sk_buff *skb, int pad)
1066 {
1067 return __skb_pad(skb, pad, true);
1068 }
1069 #define dev_kfree_skb(a) consume_skb(a)
1070
1071 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1072 int getfrag(void *from, char *to, int offset,
1073 int len, int odd, struct sk_buff *skb),
1074 void *from, int length);
1075
1076 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1077 int offset, size_t size);
1078
1079 struct skb_seq_state {
1080 __u32 lower_offset;
1081 __u32 upper_offset;
1082 __u32 frag_idx;
1083 __u32 stepped_offset;
1084 struct sk_buff *root_skb;
1085 struct sk_buff *cur_skb;
1086 __u8 *frag_data;
1087 };
1088
1089 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1090 unsigned int to, struct skb_seq_state *st);
1091 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1092 struct skb_seq_state *st);
1093 void skb_abort_seq_read(struct skb_seq_state *st);
1094
1095 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1096 unsigned int to, struct ts_config *config);
1097
1098 /*
1099 * Packet hash types specify the type of hash in skb_set_hash.
1100 *
1101 * Hash types refer to the protocol layer addresses which are used to
1102 * construct a packet's hash. The hashes are used to differentiate or identify
1103 * flows of the protocol layer for the hash type. Hash types are either
1104 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1105 *
1106 * Properties of hashes:
1107 *
1108 * 1) Two packets in different flows have different hash values
1109 * 2) Two packets in the same flow should have the same hash value
1110 *
1111 * A hash at a higher layer is considered to be more specific. A driver should
1112 * set the most specific hash possible.
1113 *
1114 * A driver cannot indicate a more specific hash than the layer at which a hash
1115 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1116 *
1117 * A driver may indicate a hash level which is less specific than the
1118 * actual layer the hash was computed on. For instance, a hash computed
1119 * at L4 may be considered an L3 hash. This should only be done if the
1120 * driver can't unambiguously determine that the HW computed the hash at
1121 * the higher layer. Note that the "should" in the second property above
1122 * permits this.
1123 */
1124 enum pkt_hash_types {
1125 PKT_HASH_TYPE_NONE, /* Undefined type */
1126 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1127 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1128 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1129 };
1130
1131 static inline void skb_clear_hash(struct sk_buff *skb)
1132 {
1133 skb->hash = 0;
1134 skb->sw_hash = 0;
1135 skb->l4_hash = 0;
1136 }
1137
1138 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1139 {
1140 if (!skb->l4_hash)
1141 skb_clear_hash(skb);
1142 }
1143
1144 static inline void
1145 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1146 {
1147 skb->l4_hash = is_l4;
1148 skb->sw_hash = is_sw;
1149 skb->hash = hash;
1150 }
1151
1152 static inline void
1153 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1154 {
1155 /* Used by drivers to set hash from HW */
1156 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1157 }
1158
1159 static inline void
1160 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1161 {
1162 __skb_set_hash(skb, hash, true, is_l4);
1163 }
1164
1165 void __skb_get_hash(struct sk_buff *skb);
1166 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1167 u32 skb_get_poff(const struct sk_buff *skb);
1168 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1169 const struct flow_keys *keys, int hlen);
1170 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1171 void *data, int hlen_proto);
1172
1173 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1174 int thoff, u8 ip_proto)
1175 {
1176 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1177 }
1178
1179 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1180 const struct flow_dissector_key *key,
1181 unsigned int key_count);
1182
1183 bool __skb_flow_dissect(const struct sk_buff *skb,
1184 struct flow_dissector *flow_dissector,
1185 void *target_container,
1186 void *data, __be16 proto, int nhoff, int hlen,
1187 unsigned int flags);
1188
1189 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1190 struct flow_dissector *flow_dissector,
1191 void *target_container, unsigned int flags)
1192 {
1193 return __skb_flow_dissect(skb, flow_dissector, target_container,
1194 NULL, 0, 0, 0, flags);
1195 }
1196
1197 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1198 struct flow_keys *flow,
1199 unsigned int flags)
1200 {
1201 memset(flow, 0, sizeof(*flow));
1202 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
1203 NULL, 0, 0, 0, flags);
1204 }
1205
1206 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1207 void *data, __be16 proto,
1208 int nhoff, int hlen,
1209 unsigned int flags)
1210 {
1211 memset(flow, 0, sizeof(*flow));
1212 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
1213 data, proto, nhoff, hlen, flags);
1214 }
1215
1216 static inline __u32 skb_get_hash(struct sk_buff *skb)
1217 {
1218 if (!skb->l4_hash && !skb->sw_hash)
1219 __skb_get_hash(skb);
1220
1221 return skb->hash;
1222 }
1223
1224 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1225 {
1226 if (!skb->l4_hash && !skb->sw_hash) {
1227 struct flow_keys keys;
1228 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1229
1230 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1231 }
1232
1233 return skb->hash;
1234 }
1235
1236 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1237
1238 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1239 {
1240 return skb->hash;
1241 }
1242
1243 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1244 {
1245 to->hash = from->hash;
1246 to->sw_hash = from->sw_hash;
1247 to->l4_hash = from->l4_hash;
1248 };
1249
1250 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1251 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1252 {
1253 return skb->head + skb->end;
1254 }
1255
1256 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1257 {
1258 return skb->end;
1259 }
1260 #else
1261 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1262 {
1263 return skb->end;
1264 }
1265
1266 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1267 {
1268 return skb->end - skb->head;
1269 }
1270 #endif
1271
1272 /* Internal */
1273 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1274
1275 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1276 {
1277 return &skb_shinfo(skb)->hwtstamps;
1278 }
1279
1280 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1281 {
1282 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1283
1284 return is_zcopy ? skb_uarg(skb) : NULL;
1285 }
1286
1287 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1288 {
1289 if (skb && uarg && !skb_zcopy(skb)) {
1290 sock_zerocopy_get(uarg);
1291 skb_shinfo(skb)->destructor_arg = uarg;
1292 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1293 }
1294 }
1295
1296 /* Release a reference on a zerocopy structure */
1297 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1298 {
1299 struct ubuf_info *uarg = skb_zcopy(skb);
1300
1301 if (uarg) {
1302 if (uarg->callback == sock_zerocopy_callback) {
1303 uarg->zerocopy = uarg->zerocopy && zerocopy;
1304 sock_zerocopy_put(uarg);
1305 } else {
1306 uarg->callback(uarg, zerocopy);
1307 }
1308
1309 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1310 }
1311 }
1312
1313 /* Abort a zerocopy operation and revert zckey on error in send syscall */
1314 static inline void skb_zcopy_abort(struct sk_buff *skb)
1315 {
1316 struct ubuf_info *uarg = skb_zcopy(skb);
1317
1318 if (uarg) {
1319 sock_zerocopy_put_abort(uarg);
1320 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1321 }
1322 }
1323
1324 /**
1325 * skb_queue_empty - check if a queue is empty
1326 * @list: queue head
1327 *
1328 * Returns true if the queue is empty, false otherwise.
1329 */
1330 static inline int skb_queue_empty(const struct sk_buff_head *list)
1331 {
1332 return list->next == (const struct sk_buff *) list;
1333 }
1334
1335 /**
1336 * skb_queue_is_last - check if skb is the last entry in the queue
1337 * @list: queue head
1338 * @skb: buffer
1339 *
1340 * Returns true if @skb is the last buffer on the list.
1341 */
1342 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1343 const struct sk_buff *skb)
1344 {
1345 return skb->next == (const struct sk_buff *) list;
1346 }
1347
1348 /**
1349 * skb_queue_is_first - check if skb is the first entry in the queue
1350 * @list: queue head
1351 * @skb: buffer
1352 *
1353 * Returns true if @skb is the first buffer on the list.
1354 */
1355 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1356 const struct sk_buff *skb)
1357 {
1358 return skb->prev == (const struct sk_buff *) list;
1359 }
1360
1361 /**
1362 * skb_queue_next - return the next packet in the queue
1363 * @list: queue head
1364 * @skb: current buffer
1365 *
1366 * Return the next packet in @list after @skb. It is only valid to
1367 * call this if skb_queue_is_last() evaluates to false.
1368 */
1369 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1370 const struct sk_buff *skb)
1371 {
1372 /* This BUG_ON may seem severe, but if we just return then we
1373 * are going to dereference garbage.
1374 */
1375 BUG_ON(skb_queue_is_last(list, skb));
1376 return skb->next;
1377 }
1378
1379 /**
1380 * skb_queue_prev - return the prev packet in the queue
1381 * @list: queue head
1382 * @skb: current buffer
1383 *
1384 * Return the prev packet in @list before @skb. It is only valid to
1385 * call this if skb_queue_is_first() evaluates to false.
1386 */
1387 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1388 const struct sk_buff *skb)
1389 {
1390 /* This BUG_ON may seem severe, but if we just return then we
1391 * are going to dereference garbage.
1392 */
1393 BUG_ON(skb_queue_is_first(list, skb));
1394 return skb->prev;
1395 }
1396
1397 /**
1398 * skb_get - reference buffer
1399 * @skb: buffer to reference
1400 *
1401 * Makes another reference to a socket buffer and returns a pointer
1402 * to the buffer.
1403 */
1404 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1405 {
1406 refcount_inc(&skb->users);
1407 return skb;
1408 }
1409
1410 /*
1411 * If users == 1, we are the only owner and are can avoid redundant
1412 * atomic change.
1413 */
1414
1415 /**
1416 * skb_cloned - is the buffer a clone
1417 * @skb: buffer to check
1418 *
1419 * Returns true if the buffer was generated with skb_clone() and is
1420 * one of multiple shared copies of the buffer. Cloned buffers are
1421 * shared data so must not be written to under normal circumstances.
1422 */
1423 static inline int skb_cloned(const struct sk_buff *skb)
1424 {
1425 return skb->cloned &&
1426 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1427 }
1428
1429 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1430 {
1431 might_sleep_if(gfpflags_allow_blocking(pri));
1432
1433 if (skb_cloned(skb))
1434 return pskb_expand_head(skb, 0, 0, pri);
1435
1436 return 0;
1437 }
1438
1439 /**
1440 * skb_header_cloned - is the header a clone
1441 * @skb: buffer to check
1442 *
1443 * Returns true if modifying the header part of the buffer requires
1444 * the data to be copied.
1445 */
1446 static inline int skb_header_cloned(const struct sk_buff *skb)
1447 {
1448 int dataref;
1449
1450 if (!skb->cloned)
1451 return 0;
1452
1453 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1454 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1455 return dataref != 1;
1456 }
1457
1458 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1459 {
1460 might_sleep_if(gfpflags_allow_blocking(pri));
1461
1462 if (skb_header_cloned(skb))
1463 return pskb_expand_head(skb, 0, 0, pri);
1464
1465 return 0;
1466 }
1467
1468 /**
1469 * __skb_header_release - release reference to header
1470 * @skb: buffer to operate on
1471 */
1472 static inline void __skb_header_release(struct sk_buff *skb)
1473 {
1474 skb->nohdr = 1;
1475 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1476 }
1477
1478
1479 /**
1480 * skb_shared - is the buffer shared
1481 * @skb: buffer to check
1482 *
1483 * Returns true if more than one person has a reference to this
1484 * buffer.
1485 */
1486 static inline int skb_shared(const struct sk_buff *skb)
1487 {
1488 return refcount_read(&skb->users) != 1;
1489 }
1490
1491 /**
1492 * skb_share_check - check if buffer is shared and if so clone it
1493 * @skb: buffer to check
1494 * @pri: priority for memory allocation
1495 *
1496 * If the buffer is shared the buffer is cloned and the old copy
1497 * drops a reference. A new clone with a single reference is returned.
1498 * If the buffer is not shared the original buffer is returned. When
1499 * being called from interrupt status or with spinlocks held pri must
1500 * be GFP_ATOMIC.
1501 *
1502 * NULL is returned on a memory allocation failure.
1503 */
1504 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1505 {
1506 might_sleep_if(gfpflags_allow_blocking(pri));
1507 if (skb_shared(skb)) {
1508 struct sk_buff *nskb = skb_clone(skb, pri);
1509
1510 if (likely(nskb))
1511 consume_skb(skb);
1512 else
1513 kfree_skb(skb);
1514 skb = nskb;
1515 }
1516 return skb;
1517 }
1518
1519 /*
1520 * Copy shared buffers into a new sk_buff. We effectively do COW on
1521 * packets to handle cases where we have a local reader and forward
1522 * and a couple of other messy ones. The normal one is tcpdumping
1523 * a packet thats being forwarded.
1524 */
1525
1526 /**
1527 * skb_unshare - make a copy of a shared buffer
1528 * @skb: buffer to check
1529 * @pri: priority for memory allocation
1530 *
1531 * If the socket buffer is a clone then this function creates a new
1532 * copy of the data, drops a reference count on the old copy and returns
1533 * the new copy with the reference count at 1. If the buffer is not a clone
1534 * the original buffer is returned. When called with a spinlock held or
1535 * from interrupt state @pri must be %GFP_ATOMIC
1536 *
1537 * %NULL is returned on a memory allocation failure.
1538 */
1539 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1540 gfp_t pri)
1541 {
1542 might_sleep_if(gfpflags_allow_blocking(pri));
1543 if (skb_cloned(skb)) {
1544 struct sk_buff *nskb = skb_copy(skb, pri);
1545
1546 /* Free our shared copy */
1547 if (likely(nskb))
1548 consume_skb(skb);
1549 else
1550 kfree_skb(skb);
1551 skb = nskb;
1552 }
1553 return skb;
1554 }
1555
1556 /**
1557 * skb_peek - peek at the head of an &sk_buff_head
1558 * @list_: list to peek at
1559 *
1560 * Peek an &sk_buff. Unlike most other operations you _MUST_
1561 * be careful with this one. A peek leaves the buffer on the
1562 * list and someone else may run off with it. You must hold
1563 * the appropriate locks or have a private queue to do this.
1564 *
1565 * Returns %NULL for an empty list or a pointer to the head element.
1566 * The reference count is not incremented and the reference is therefore
1567 * volatile. Use with caution.
1568 */
1569 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1570 {
1571 struct sk_buff *skb = list_->next;
1572
1573 if (skb == (struct sk_buff *)list_)
1574 skb = NULL;
1575 return skb;
1576 }
1577
1578 /**
1579 * skb_peek_next - peek skb following the given one from a queue
1580 * @skb: skb to start from
1581 * @list_: list to peek at
1582 *
1583 * Returns %NULL when the end of the list is met or a pointer to the
1584 * next element. The reference count is not incremented and the
1585 * reference is therefore volatile. Use with caution.
1586 */
1587 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1588 const struct sk_buff_head *list_)
1589 {
1590 struct sk_buff *next = skb->next;
1591
1592 if (next == (struct sk_buff *)list_)
1593 next = NULL;
1594 return next;
1595 }
1596
1597 /**
1598 * skb_peek_tail - peek at the tail of an &sk_buff_head
1599 * @list_: list to peek at
1600 *
1601 * Peek an &sk_buff. Unlike most other operations you _MUST_
1602 * be careful with this one. A peek leaves the buffer on the
1603 * list and someone else may run off with it. You must hold
1604 * the appropriate locks or have a private queue to do this.
1605 *
1606 * Returns %NULL for an empty list or a pointer to the tail element.
1607 * The reference count is not incremented and the reference is therefore
1608 * volatile. Use with caution.
1609 */
1610 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1611 {
1612 struct sk_buff *skb = list_->prev;
1613
1614 if (skb == (struct sk_buff *)list_)
1615 skb = NULL;
1616 return skb;
1617
1618 }
1619
1620 /**
1621 * skb_queue_len - get queue length
1622 * @list_: list to measure
1623 *
1624 * Return the length of an &sk_buff queue.
1625 */
1626 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1627 {
1628 return list_->qlen;
1629 }
1630
1631 /**
1632 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1633 * @list: queue to initialize
1634 *
1635 * This initializes only the list and queue length aspects of
1636 * an sk_buff_head object. This allows to initialize the list
1637 * aspects of an sk_buff_head without reinitializing things like
1638 * the spinlock. It can also be used for on-stack sk_buff_head
1639 * objects where the spinlock is known to not be used.
1640 */
1641 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1642 {
1643 list->prev = list->next = (struct sk_buff *)list;
1644 list->qlen = 0;
1645 }
1646
1647 /*
1648 * This function creates a split out lock class for each invocation;
1649 * this is needed for now since a whole lot of users of the skb-queue
1650 * infrastructure in drivers have different locking usage (in hardirq)
1651 * than the networking core (in softirq only). In the long run either the
1652 * network layer or drivers should need annotation to consolidate the
1653 * main types of usage into 3 classes.
1654 */
1655 static inline void skb_queue_head_init(struct sk_buff_head *list)
1656 {
1657 spin_lock_init(&list->lock);
1658 __skb_queue_head_init(list);
1659 }
1660
1661 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1662 struct lock_class_key *class)
1663 {
1664 skb_queue_head_init(list);
1665 lockdep_set_class(&list->lock, class);
1666 }
1667
1668 /*
1669 * Insert an sk_buff on a list.
1670 *
1671 * The "__skb_xxxx()" functions are the non-atomic ones that
1672 * can only be called with interrupts disabled.
1673 */
1674 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1675 struct sk_buff_head *list);
1676 static inline void __skb_insert(struct sk_buff *newsk,
1677 struct sk_buff *prev, struct sk_buff *next,
1678 struct sk_buff_head *list)
1679 {
1680 newsk->next = next;
1681 newsk->prev = prev;
1682 next->prev = prev->next = newsk;
1683 list->qlen++;
1684 }
1685
1686 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1687 struct sk_buff *prev,
1688 struct sk_buff *next)
1689 {
1690 struct sk_buff *first = list->next;
1691 struct sk_buff *last = list->prev;
1692
1693 first->prev = prev;
1694 prev->next = first;
1695
1696 last->next = next;
1697 next->prev = last;
1698 }
1699
1700 /**
1701 * skb_queue_splice - join two skb lists, this is designed for stacks
1702 * @list: the new list to add
1703 * @head: the place to add it in the first list
1704 */
1705 static inline void skb_queue_splice(const struct sk_buff_head *list,
1706 struct sk_buff_head *head)
1707 {
1708 if (!skb_queue_empty(list)) {
1709 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1710 head->qlen += list->qlen;
1711 }
1712 }
1713
1714 /**
1715 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1716 * @list: the new list to add
1717 * @head: the place to add it in the first list
1718 *
1719 * The list at @list is reinitialised
1720 */
1721 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1722 struct sk_buff_head *head)
1723 {
1724 if (!skb_queue_empty(list)) {
1725 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1726 head->qlen += list->qlen;
1727 __skb_queue_head_init(list);
1728 }
1729 }
1730
1731 /**
1732 * skb_queue_splice_tail - join two skb lists, each list being a queue
1733 * @list: the new list to add
1734 * @head: the place to add it in the first list
1735 */
1736 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1737 struct sk_buff_head *head)
1738 {
1739 if (!skb_queue_empty(list)) {
1740 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1741 head->qlen += list->qlen;
1742 }
1743 }
1744
1745 /**
1746 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1747 * @list: the new list to add
1748 * @head: the place to add it in the first list
1749 *
1750 * Each of the lists is a queue.
1751 * The list at @list is reinitialised
1752 */
1753 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1754 struct sk_buff_head *head)
1755 {
1756 if (!skb_queue_empty(list)) {
1757 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1758 head->qlen += list->qlen;
1759 __skb_queue_head_init(list);
1760 }
1761 }
1762
1763 /**
1764 * __skb_queue_after - queue a buffer at the list head
1765 * @list: list to use
1766 * @prev: place after this buffer
1767 * @newsk: buffer to queue
1768 *
1769 * Queue a buffer int the middle of a list. This function takes no locks
1770 * and you must therefore hold required locks before calling it.
1771 *
1772 * A buffer cannot be placed on two lists at the same time.
1773 */
1774 static inline void __skb_queue_after(struct sk_buff_head *list,
1775 struct sk_buff *prev,
1776 struct sk_buff *newsk)
1777 {
1778 __skb_insert(newsk, prev, prev->next, list);
1779 }
1780
1781 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1782 struct sk_buff_head *list);
1783
1784 static inline void __skb_queue_before(struct sk_buff_head *list,
1785 struct sk_buff *next,
1786 struct sk_buff *newsk)
1787 {
1788 __skb_insert(newsk, next->prev, next, list);
1789 }
1790
1791 /**
1792 * __skb_queue_head - queue a buffer at the list head
1793 * @list: list to use
1794 * @newsk: buffer to queue
1795 *
1796 * Queue a buffer at the start of a list. This function takes no locks
1797 * and you must therefore hold required locks before calling it.
1798 *
1799 * A buffer cannot be placed on two lists at the same time.
1800 */
1801 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1802 static inline void __skb_queue_head(struct sk_buff_head *list,
1803 struct sk_buff *newsk)
1804 {
1805 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1806 }
1807
1808 /**
1809 * __skb_queue_tail - queue a buffer at the list tail
1810 * @list: list to use
1811 * @newsk: buffer to queue
1812 *
1813 * Queue a buffer at the end of a list. This function takes no locks
1814 * and you must therefore hold required locks before calling it.
1815 *
1816 * A buffer cannot be placed on two lists at the same time.
1817 */
1818 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1819 static inline void __skb_queue_tail(struct sk_buff_head *list,
1820 struct sk_buff *newsk)
1821 {
1822 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1823 }
1824
1825 /*
1826 * remove sk_buff from list. _Must_ be called atomically, and with
1827 * the list known..
1828 */
1829 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1830 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1831 {
1832 struct sk_buff *next, *prev;
1833
1834 list->qlen--;
1835 next = skb->next;
1836 prev = skb->prev;
1837 skb->next = skb->prev = NULL;
1838 next->prev = prev;
1839 prev->next = next;
1840 }
1841
1842 /**
1843 * __skb_dequeue - remove from the head of the queue
1844 * @list: list to dequeue from
1845 *
1846 * Remove the head of the list. This function does not take any locks
1847 * so must be used with appropriate locks held only. The head item is
1848 * returned or %NULL if the list is empty.
1849 */
1850 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1851 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1852 {
1853 struct sk_buff *skb = skb_peek(list);
1854 if (skb)
1855 __skb_unlink(skb, list);
1856 return skb;
1857 }
1858
1859 /**
1860 * __skb_dequeue_tail - remove from the tail of the queue
1861 * @list: list to dequeue from
1862 *
1863 * Remove the tail of the list. This function does not take any locks
1864 * so must be used with appropriate locks held only. The tail item is
1865 * returned or %NULL if the list is empty.
1866 */
1867 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1868 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1869 {
1870 struct sk_buff *skb = skb_peek_tail(list);
1871 if (skb)
1872 __skb_unlink(skb, list);
1873 return skb;
1874 }
1875
1876
1877 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1878 {
1879 return skb->data_len;
1880 }
1881
1882 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1883 {
1884 return skb->len - skb->data_len;
1885 }
1886
1887 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1888 {
1889 unsigned int i, len = 0;
1890
1891 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
1892 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1893 return len;
1894 }
1895
1896 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1897 {
1898 return skb_headlen(skb) + __skb_pagelen(skb);
1899 }
1900
1901 /**
1902 * __skb_fill_page_desc - initialise a paged fragment in an skb
1903 * @skb: buffer containing fragment to be initialised
1904 * @i: paged fragment index to initialise
1905 * @page: the page to use for this fragment
1906 * @off: the offset to the data with @page
1907 * @size: the length of the data
1908 *
1909 * Initialises the @i'th fragment of @skb to point to &size bytes at
1910 * offset @off within @page.
1911 *
1912 * Does not take any additional reference on the fragment.
1913 */
1914 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1915 struct page *page, int off, int size)
1916 {
1917 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1918
1919 /*
1920 * Propagate page pfmemalloc to the skb if we can. The problem is
1921 * that not all callers have unique ownership of the page but rely
1922 * on page_is_pfmemalloc doing the right thing(tm).
1923 */
1924 frag->page.p = page;
1925 frag->page_offset = off;
1926 skb_frag_size_set(frag, size);
1927
1928 page = compound_head(page);
1929 if (page_is_pfmemalloc(page))
1930 skb->pfmemalloc = true;
1931 }
1932
1933 /**
1934 * skb_fill_page_desc - initialise a paged fragment in an skb
1935 * @skb: buffer containing fragment to be initialised
1936 * @i: paged fragment index to initialise
1937 * @page: the page to use for this fragment
1938 * @off: the offset to the data with @page
1939 * @size: the length of the data
1940 *
1941 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1942 * @skb to point to @size bytes at offset @off within @page. In
1943 * addition updates @skb such that @i is the last fragment.
1944 *
1945 * Does not take any additional reference on the fragment.
1946 */
1947 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1948 struct page *page, int off, int size)
1949 {
1950 __skb_fill_page_desc(skb, i, page, off, size);
1951 skb_shinfo(skb)->nr_frags = i + 1;
1952 }
1953
1954 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1955 int size, unsigned int truesize);
1956
1957 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1958 unsigned int truesize);
1959
1960 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1961 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1962 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1963
1964 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1965 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1966 {
1967 return skb->head + skb->tail;
1968 }
1969
1970 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1971 {
1972 skb->tail = skb->data - skb->head;
1973 }
1974
1975 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1976 {
1977 skb_reset_tail_pointer(skb);
1978 skb->tail += offset;
1979 }
1980
1981 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1982 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1983 {
1984 return skb->tail;
1985 }
1986
1987 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1988 {
1989 skb->tail = skb->data;
1990 }
1991
1992 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1993 {
1994 skb->tail = skb->data + offset;
1995 }
1996
1997 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1998
1999 /*
2000 * Add data to an sk_buff
2001 */
2002 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2003 void *skb_put(struct sk_buff *skb, unsigned int len);
2004 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2005 {
2006 void *tmp = skb_tail_pointer(skb);
2007 SKB_LINEAR_ASSERT(skb);
2008 skb->tail += len;
2009 skb->len += len;
2010 return tmp;
2011 }
2012
2013 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2014 {
2015 void *tmp = __skb_put(skb, len);
2016
2017 memset(tmp, 0, len);
2018 return tmp;
2019 }
2020
2021 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2022 unsigned int len)
2023 {
2024 void *tmp = __skb_put(skb, len);
2025
2026 memcpy(tmp, data, len);
2027 return tmp;
2028 }
2029
2030 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2031 {
2032 *(u8 *)__skb_put(skb, 1) = val;
2033 }
2034
2035 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2036 {
2037 void *tmp = skb_put(skb, len);
2038
2039 memset(tmp, 0, len);
2040
2041 return tmp;
2042 }
2043
2044 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2045 unsigned int len)
2046 {
2047 void *tmp = skb_put(skb, len);
2048
2049 memcpy(tmp, data, len);
2050
2051 return tmp;
2052 }
2053
2054 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2055 {
2056 *(u8 *)skb_put(skb, 1) = val;
2057 }
2058
2059 void *skb_push(struct sk_buff *skb, unsigned int len);
2060 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2061 {
2062 skb->data -= len;
2063 skb->len += len;
2064 return skb->data;
2065 }
2066
2067 void *skb_pull(struct sk_buff *skb, unsigned int len);
2068 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2069 {
2070 skb->len -= len;
2071 BUG_ON(skb->len < skb->data_len);
2072 return skb->data += len;
2073 }
2074
2075 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2076 {
2077 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2078 }
2079
2080 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2081
2082 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2083 {
2084 if (len > skb_headlen(skb) &&
2085 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2086 return NULL;
2087 skb->len -= len;
2088 return skb->data += len;
2089 }
2090
2091 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2092 {
2093 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2094 }
2095
2096 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2097 {
2098 if (likely(len <= skb_headlen(skb)))
2099 return 1;
2100 if (unlikely(len > skb->len))
2101 return 0;
2102 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2103 }
2104
2105 void skb_condense(struct sk_buff *skb);
2106
2107 /**
2108 * skb_headroom - bytes at buffer head
2109 * @skb: buffer to check
2110 *
2111 * Return the number of bytes of free space at the head of an &sk_buff.
2112 */
2113 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2114 {
2115 return skb->data - skb->head;
2116 }
2117
2118 /**
2119 * skb_tailroom - bytes at buffer end
2120 * @skb: buffer to check
2121 *
2122 * Return the number of bytes of free space at the tail of an sk_buff
2123 */
2124 static inline int skb_tailroom(const struct sk_buff *skb)
2125 {
2126 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2127 }
2128
2129 /**
2130 * skb_availroom - bytes at buffer end
2131 * @skb: buffer to check
2132 *
2133 * Return the number of bytes of free space at the tail of an sk_buff
2134 * allocated by sk_stream_alloc()
2135 */
2136 static inline int skb_availroom(const struct sk_buff *skb)
2137 {
2138 if (skb_is_nonlinear(skb))
2139 return 0;
2140
2141 return skb->end - skb->tail - skb->reserved_tailroom;
2142 }
2143
2144 /**
2145 * skb_reserve - adjust headroom
2146 * @skb: buffer to alter
2147 * @len: bytes to move
2148 *
2149 * Increase the headroom of an empty &sk_buff by reducing the tail
2150 * room. This is only allowed for an empty buffer.
2151 */
2152 static inline void skb_reserve(struct sk_buff *skb, int len)
2153 {
2154 skb->data += len;
2155 skb->tail += len;
2156 }
2157
2158 /**
2159 * skb_tailroom_reserve - adjust reserved_tailroom
2160 * @skb: buffer to alter
2161 * @mtu: maximum amount of headlen permitted
2162 * @needed_tailroom: minimum amount of reserved_tailroom
2163 *
2164 * Set reserved_tailroom so that headlen can be as large as possible but
2165 * not larger than mtu and tailroom cannot be smaller than
2166 * needed_tailroom.
2167 * The required headroom should already have been reserved before using
2168 * this function.
2169 */
2170 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2171 unsigned int needed_tailroom)
2172 {
2173 SKB_LINEAR_ASSERT(skb);
2174 if (mtu < skb_tailroom(skb) - needed_tailroom)
2175 /* use at most mtu */
2176 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2177 else
2178 /* use up to all available space */
2179 skb->reserved_tailroom = needed_tailroom;
2180 }
2181
2182 #define ENCAP_TYPE_ETHER 0
2183 #define ENCAP_TYPE_IPPROTO 1
2184
2185 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2186 __be16 protocol)
2187 {
2188 skb->inner_protocol = protocol;
2189 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2190 }
2191
2192 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2193 __u8 ipproto)
2194 {
2195 skb->inner_ipproto = ipproto;
2196 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2197 }
2198
2199 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2200 {
2201 skb->inner_mac_header = skb->mac_header;
2202 skb->inner_network_header = skb->network_header;
2203 skb->inner_transport_header = skb->transport_header;
2204 }
2205
2206 static inline void skb_reset_mac_len(struct sk_buff *skb)
2207 {
2208 skb->mac_len = skb->network_header - skb->mac_header;
2209 }
2210
2211 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2212 *skb)
2213 {
2214 return skb->head + skb->inner_transport_header;
2215 }
2216
2217 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2218 {
2219 return skb_inner_transport_header(skb) - skb->data;
2220 }
2221
2222 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2223 {
2224 skb->inner_transport_header = skb->data - skb->head;
2225 }
2226
2227 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2228 const int offset)
2229 {
2230 skb_reset_inner_transport_header(skb);
2231 skb->inner_transport_header += offset;
2232 }
2233
2234 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2235 {
2236 return skb->head + skb->inner_network_header;
2237 }
2238
2239 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2240 {
2241 skb->inner_network_header = skb->data - skb->head;
2242 }
2243
2244 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2245 const int offset)
2246 {
2247 skb_reset_inner_network_header(skb);
2248 skb->inner_network_header += offset;
2249 }
2250
2251 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2252 {
2253 return skb->head + skb->inner_mac_header;
2254 }
2255
2256 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2257 {
2258 skb->inner_mac_header = skb->data - skb->head;
2259 }
2260
2261 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2262 const int offset)
2263 {
2264 skb_reset_inner_mac_header(skb);
2265 skb->inner_mac_header += offset;
2266 }
2267 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2268 {
2269 return skb->transport_header != (typeof(skb->transport_header))~0U;
2270 }
2271
2272 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2273 {
2274 return skb->head + skb->transport_header;
2275 }
2276
2277 static inline void skb_reset_transport_header(struct sk_buff *skb)
2278 {
2279 skb->transport_header = skb->data - skb->head;
2280 }
2281
2282 static inline void skb_set_transport_header(struct sk_buff *skb,
2283 const int offset)
2284 {
2285 skb_reset_transport_header(skb);
2286 skb->transport_header += offset;
2287 }
2288
2289 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2290 {
2291 return skb->head + skb->network_header;
2292 }
2293
2294 static inline void skb_reset_network_header(struct sk_buff *skb)
2295 {
2296 skb->network_header = skb->data - skb->head;
2297 }
2298
2299 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2300 {
2301 skb_reset_network_header(skb);
2302 skb->network_header += offset;
2303 }
2304
2305 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2306 {
2307 return skb->head + skb->mac_header;
2308 }
2309
2310 static inline int skb_mac_offset(const struct sk_buff *skb)
2311 {
2312 return skb_mac_header(skb) - skb->data;
2313 }
2314
2315 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2316 {
2317 return skb->network_header - skb->mac_header;
2318 }
2319
2320 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2321 {
2322 return skb->mac_header != (typeof(skb->mac_header))~0U;
2323 }
2324
2325 static inline void skb_reset_mac_header(struct sk_buff *skb)
2326 {
2327 skb->mac_header = skb->data - skb->head;
2328 }
2329
2330 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2331 {
2332 skb_reset_mac_header(skb);
2333 skb->mac_header += offset;
2334 }
2335
2336 static inline void skb_pop_mac_header(struct sk_buff *skb)
2337 {
2338 skb->mac_header = skb->network_header;
2339 }
2340
2341 static inline void skb_probe_transport_header(struct sk_buff *skb,
2342 const int offset_hint)
2343 {
2344 struct flow_keys keys;
2345
2346 if (skb_transport_header_was_set(skb))
2347 return;
2348 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
2349 skb_set_transport_header(skb, keys.control.thoff);
2350 else
2351 skb_set_transport_header(skb, offset_hint);
2352 }
2353
2354 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2355 {
2356 if (skb_mac_header_was_set(skb)) {
2357 const unsigned char *old_mac = skb_mac_header(skb);
2358
2359 skb_set_mac_header(skb, -skb->mac_len);
2360 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2361 }
2362 }
2363
2364 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2365 {
2366 return skb->csum_start - skb_headroom(skb);
2367 }
2368
2369 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2370 {
2371 return skb->head + skb->csum_start;
2372 }
2373
2374 static inline int skb_transport_offset(const struct sk_buff *skb)
2375 {
2376 return skb_transport_header(skb) - skb->data;
2377 }
2378
2379 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2380 {
2381 return skb->transport_header - skb->network_header;
2382 }
2383
2384 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2385 {
2386 return skb->inner_transport_header - skb->inner_network_header;
2387 }
2388
2389 static inline int skb_network_offset(const struct sk_buff *skb)
2390 {
2391 return skb_network_header(skb) - skb->data;
2392 }
2393
2394 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2395 {
2396 return skb_inner_network_header(skb) - skb->data;
2397 }
2398
2399 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2400 {
2401 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2402 }
2403
2404 /*
2405 * CPUs often take a performance hit when accessing unaligned memory
2406 * locations. The actual performance hit varies, it can be small if the
2407 * hardware handles it or large if we have to take an exception and fix it
2408 * in software.
2409 *
2410 * Since an ethernet header is 14 bytes network drivers often end up with
2411 * the IP header at an unaligned offset. The IP header can be aligned by
2412 * shifting the start of the packet by 2 bytes. Drivers should do this
2413 * with:
2414 *
2415 * skb_reserve(skb, NET_IP_ALIGN);
2416 *
2417 * The downside to this alignment of the IP header is that the DMA is now
2418 * unaligned. On some architectures the cost of an unaligned DMA is high
2419 * and this cost outweighs the gains made by aligning the IP header.
2420 *
2421 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2422 * to be overridden.
2423 */
2424 #ifndef NET_IP_ALIGN
2425 #define NET_IP_ALIGN 2
2426 #endif
2427
2428 /*
2429 * The networking layer reserves some headroom in skb data (via
2430 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2431 * the header has to grow. In the default case, if the header has to grow
2432 * 32 bytes or less we avoid the reallocation.
2433 *
2434 * Unfortunately this headroom changes the DMA alignment of the resulting
2435 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2436 * on some architectures. An architecture can override this value,
2437 * perhaps setting it to a cacheline in size (since that will maintain
2438 * cacheline alignment of the DMA). It must be a power of 2.
2439 *
2440 * Various parts of the networking layer expect at least 32 bytes of
2441 * headroom, you should not reduce this.
2442 *
2443 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2444 * to reduce average number of cache lines per packet.
2445 * get_rps_cpus() for example only access one 64 bytes aligned block :
2446 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2447 */
2448 #ifndef NET_SKB_PAD
2449 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2450 #endif
2451
2452 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2453
2454 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2455 {
2456 if (unlikely(skb_is_nonlinear(skb))) {
2457 WARN_ON(1);
2458 return;
2459 }
2460 skb->len = len;
2461 skb_set_tail_pointer(skb, len);
2462 }
2463
2464 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2465 {
2466 __skb_set_length(skb, len);
2467 }
2468
2469 void skb_trim(struct sk_buff *skb, unsigned int len);
2470
2471 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2472 {
2473 if (skb->data_len)
2474 return ___pskb_trim(skb, len);
2475 __skb_trim(skb, len);
2476 return 0;
2477 }
2478
2479 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2480 {
2481 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2482 }
2483
2484 /**
2485 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2486 * @skb: buffer to alter
2487 * @len: new length
2488 *
2489 * This is identical to pskb_trim except that the caller knows that
2490 * the skb is not cloned so we should never get an error due to out-
2491 * of-memory.
2492 */
2493 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2494 {
2495 int err = pskb_trim(skb, len);
2496 BUG_ON(err);
2497 }
2498
2499 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2500 {
2501 unsigned int diff = len - skb->len;
2502
2503 if (skb_tailroom(skb) < diff) {
2504 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2505 GFP_ATOMIC);
2506 if (ret)
2507 return ret;
2508 }
2509 __skb_set_length(skb, len);
2510 return 0;
2511 }
2512
2513 /**
2514 * skb_orphan - orphan a buffer
2515 * @skb: buffer to orphan
2516 *
2517 * If a buffer currently has an owner then we call the owner's
2518 * destructor function and make the @skb unowned. The buffer continues
2519 * to exist but is no longer charged to its former owner.
2520 */
2521 static inline void skb_orphan(struct sk_buff *skb)
2522 {
2523 if (skb->destructor) {
2524 skb->destructor(skb);
2525 skb->destructor = NULL;
2526 skb->sk = NULL;
2527 } else {
2528 BUG_ON(skb->sk);
2529 }
2530 }
2531
2532 /**
2533 * skb_orphan_frags - orphan the frags contained in a buffer
2534 * @skb: buffer to orphan frags from
2535 * @gfp_mask: allocation mask for replacement pages
2536 *
2537 * For each frag in the SKB which needs a destructor (i.e. has an
2538 * owner) create a copy of that frag and release the original
2539 * page by calling the destructor.
2540 */
2541 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2542 {
2543 if (likely(!skb_zcopy(skb)))
2544 return 0;
2545 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2546 return 0;
2547 return skb_copy_ubufs(skb, gfp_mask);
2548 }
2549
2550 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2551 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2552 {
2553 if (likely(!skb_zcopy(skb)))
2554 return 0;
2555 return skb_copy_ubufs(skb, gfp_mask);
2556 }
2557
2558 /**
2559 * __skb_queue_purge - empty a list
2560 * @list: list to empty
2561 *
2562 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2563 * the list and one reference dropped. This function does not take the
2564 * list lock and the caller must hold the relevant locks to use it.
2565 */
2566 void skb_queue_purge(struct sk_buff_head *list);
2567 static inline void __skb_queue_purge(struct sk_buff_head *list)
2568 {
2569 struct sk_buff *skb;
2570 while ((skb = __skb_dequeue(list)) != NULL)
2571 kfree_skb(skb);
2572 }
2573
2574 void skb_rbtree_purge(struct rb_root *root);
2575
2576 void *netdev_alloc_frag(unsigned int fragsz);
2577
2578 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2579 gfp_t gfp_mask);
2580
2581 /**
2582 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2583 * @dev: network device to receive on
2584 * @length: length to allocate
2585 *
2586 * Allocate a new &sk_buff and assign it a usage count of one. The
2587 * buffer has unspecified headroom built in. Users should allocate
2588 * the headroom they think they need without accounting for the
2589 * built in space. The built in space is used for optimisations.
2590 *
2591 * %NULL is returned if there is no free memory. Although this function
2592 * allocates memory it can be called from an interrupt.
2593 */
2594 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2595 unsigned int length)
2596 {
2597 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2598 }
2599
2600 /* legacy helper around __netdev_alloc_skb() */
2601 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2602 gfp_t gfp_mask)
2603 {
2604 return __netdev_alloc_skb(NULL, length, gfp_mask);
2605 }
2606
2607 /* legacy helper around netdev_alloc_skb() */
2608 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2609 {
2610 return netdev_alloc_skb(NULL, length);
2611 }
2612
2613
2614 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2615 unsigned int length, gfp_t gfp)
2616 {
2617 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2618
2619 if (NET_IP_ALIGN && skb)
2620 skb_reserve(skb, NET_IP_ALIGN);
2621 return skb;
2622 }
2623
2624 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2625 unsigned int length)
2626 {
2627 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2628 }
2629
2630 static inline void skb_free_frag(void *addr)
2631 {
2632 page_frag_free(addr);
2633 }
2634
2635 void *napi_alloc_frag(unsigned int fragsz);
2636 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2637 unsigned int length, gfp_t gfp_mask);
2638 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2639 unsigned int length)
2640 {
2641 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2642 }
2643 void napi_consume_skb(struct sk_buff *skb, int budget);
2644
2645 void __kfree_skb_flush(void);
2646 void __kfree_skb_defer(struct sk_buff *skb);
2647
2648 /**
2649 * __dev_alloc_pages - allocate page for network Rx
2650 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2651 * @order: size of the allocation
2652 *
2653 * Allocate a new page.
2654 *
2655 * %NULL is returned if there is no free memory.
2656 */
2657 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2658 unsigned int order)
2659 {
2660 /* This piece of code contains several assumptions.
2661 * 1. This is for device Rx, therefor a cold page is preferred.
2662 * 2. The expectation is the user wants a compound page.
2663 * 3. If requesting a order 0 page it will not be compound
2664 * due to the check to see if order has a value in prep_new_page
2665 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2666 * code in gfp_to_alloc_flags that should be enforcing this.
2667 */
2668 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2669
2670 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2671 }
2672
2673 static inline struct page *dev_alloc_pages(unsigned int order)
2674 {
2675 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2676 }
2677
2678 /**
2679 * __dev_alloc_page - allocate a page for network Rx
2680 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2681 *
2682 * Allocate a new page.
2683 *
2684 * %NULL is returned if there is no free memory.
2685 */
2686 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2687 {
2688 return __dev_alloc_pages(gfp_mask, 0);
2689 }
2690
2691 static inline struct page *dev_alloc_page(void)
2692 {
2693 return dev_alloc_pages(0);
2694 }
2695
2696 /**
2697 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2698 * @page: The page that was allocated from skb_alloc_page
2699 * @skb: The skb that may need pfmemalloc set
2700 */
2701 static inline void skb_propagate_pfmemalloc(struct page *page,
2702 struct sk_buff *skb)
2703 {
2704 if (page_is_pfmemalloc(page))
2705 skb->pfmemalloc = true;
2706 }
2707
2708 /**
2709 * skb_frag_page - retrieve the page referred to by a paged fragment
2710 * @frag: the paged fragment
2711 *
2712 * Returns the &struct page associated with @frag.
2713 */
2714 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2715 {
2716 return frag->page.p;
2717 }
2718
2719 /**
2720 * __skb_frag_ref - take an addition reference on a paged fragment.
2721 * @frag: the paged fragment
2722 *
2723 * Takes an additional reference on the paged fragment @frag.
2724 */
2725 static inline void __skb_frag_ref(skb_frag_t *frag)
2726 {
2727 get_page(skb_frag_page(frag));
2728 }
2729
2730 /**
2731 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2732 * @skb: the buffer
2733 * @f: the fragment offset.
2734 *
2735 * Takes an additional reference on the @f'th paged fragment of @skb.
2736 */
2737 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2738 {
2739 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2740 }
2741
2742 /**
2743 * __skb_frag_unref - release a reference on a paged fragment.
2744 * @frag: the paged fragment
2745 *
2746 * Releases a reference on the paged fragment @frag.
2747 */
2748 static inline void __skb_frag_unref(skb_frag_t *frag)
2749 {
2750 put_page(skb_frag_page(frag));
2751 }
2752
2753 /**
2754 * skb_frag_unref - release a reference on a paged fragment of an skb.
2755 * @skb: the buffer
2756 * @f: the fragment offset
2757 *
2758 * Releases a reference on the @f'th paged fragment of @skb.
2759 */
2760 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2761 {
2762 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2763 }
2764
2765 /**
2766 * skb_frag_address - gets the address of the data contained in a paged fragment
2767 * @frag: the paged fragment buffer
2768 *
2769 * Returns the address of the data within @frag. The page must already
2770 * be mapped.
2771 */
2772 static inline void *skb_frag_address(const skb_frag_t *frag)
2773 {
2774 return page_address(skb_frag_page(frag)) + frag->page_offset;
2775 }
2776
2777 /**
2778 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2779 * @frag: the paged fragment buffer
2780 *
2781 * Returns the address of the data within @frag. Checks that the page
2782 * is mapped and returns %NULL otherwise.
2783 */
2784 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2785 {
2786 void *ptr = page_address(skb_frag_page(frag));
2787 if (unlikely(!ptr))
2788 return NULL;
2789
2790 return ptr + frag->page_offset;
2791 }
2792
2793 /**
2794 * __skb_frag_set_page - sets the page contained in a paged fragment
2795 * @frag: the paged fragment
2796 * @page: the page to set
2797 *
2798 * Sets the fragment @frag to contain @page.
2799 */
2800 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2801 {
2802 frag->page.p = page;
2803 }
2804
2805 /**
2806 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2807 * @skb: the buffer
2808 * @f: the fragment offset
2809 * @page: the page to set
2810 *
2811 * Sets the @f'th fragment of @skb to contain @page.
2812 */
2813 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2814 struct page *page)
2815 {
2816 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2817 }
2818
2819 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2820
2821 /**
2822 * skb_frag_dma_map - maps a paged fragment via the DMA API
2823 * @dev: the device to map the fragment to
2824 * @frag: the paged fragment to map
2825 * @offset: the offset within the fragment (starting at the
2826 * fragment's own offset)
2827 * @size: the number of bytes to map
2828 * @dir: the direction of the mapping (``PCI_DMA_*``)
2829 *
2830 * Maps the page associated with @frag to @device.
2831 */
2832 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2833 const skb_frag_t *frag,
2834 size_t offset, size_t size,
2835 enum dma_data_direction dir)
2836 {
2837 return dma_map_page(dev, skb_frag_page(frag),
2838 frag->page_offset + offset, size, dir);
2839 }
2840
2841 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2842 gfp_t gfp_mask)
2843 {
2844 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2845 }
2846
2847
2848 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2849 gfp_t gfp_mask)
2850 {
2851 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2852 }
2853
2854
2855 /**
2856 * skb_clone_writable - is the header of a clone writable
2857 * @skb: buffer to check
2858 * @len: length up to which to write
2859 *
2860 * Returns true if modifying the header part of the cloned buffer
2861 * does not requires the data to be copied.
2862 */
2863 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2864 {
2865 return !skb_header_cloned(skb) &&
2866 skb_headroom(skb) + len <= skb->hdr_len;
2867 }
2868
2869 static inline int skb_try_make_writable(struct sk_buff *skb,
2870 unsigned int write_len)
2871 {
2872 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2873 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2874 }
2875
2876 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2877 int cloned)
2878 {
2879 int delta = 0;
2880
2881 if (headroom > skb_headroom(skb))
2882 delta = headroom - skb_headroom(skb);
2883
2884 if (delta || cloned)
2885 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2886 GFP_ATOMIC);
2887 return 0;
2888 }
2889
2890 /**
2891 * skb_cow - copy header of skb when it is required
2892 * @skb: buffer to cow
2893 * @headroom: needed headroom
2894 *
2895 * If the skb passed lacks sufficient headroom or its data part
2896 * is shared, data is reallocated. If reallocation fails, an error
2897 * is returned and original skb is not changed.
2898 *
2899 * The result is skb with writable area skb->head...skb->tail
2900 * and at least @headroom of space at head.
2901 */
2902 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2903 {
2904 return __skb_cow(skb, headroom, skb_cloned(skb));
2905 }
2906
2907 /**
2908 * skb_cow_head - skb_cow but only making the head writable
2909 * @skb: buffer to cow
2910 * @headroom: needed headroom
2911 *
2912 * This function is identical to skb_cow except that we replace the
2913 * skb_cloned check by skb_header_cloned. It should be used when
2914 * you only need to push on some header and do not need to modify
2915 * the data.
2916 */
2917 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2918 {
2919 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2920 }
2921
2922 /**
2923 * skb_padto - pad an skbuff up to a minimal size
2924 * @skb: buffer to pad
2925 * @len: minimal length
2926 *
2927 * Pads up a buffer to ensure the trailing bytes exist and are
2928 * blanked. If the buffer already contains sufficient data it
2929 * is untouched. Otherwise it is extended. Returns zero on
2930 * success. The skb is freed on error.
2931 */
2932 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2933 {
2934 unsigned int size = skb->len;
2935 if (likely(size >= len))
2936 return 0;
2937 return skb_pad(skb, len - size);
2938 }
2939
2940 /**
2941 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2942 * @skb: buffer to pad
2943 * @len: minimal length
2944 * @free_on_error: free buffer on error
2945 *
2946 * Pads up a buffer to ensure the trailing bytes exist and are
2947 * blanked. If the buffer already contains sufficient data it
2948 * is untouched. Otherwise it is extended. Returns zero on
2949 * success. The skb is freed on error if @free_on_error is true.
2950 */
2951 static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2952 bool free_on_error)
2953 {
2954 unsigned int size = skb->len;
2955
2956 if (unlikely(size < len)) {
2957 len -= size;
2958 if (__skb_pad(skb, len, free_on_error))
2959 return -ENOMEM;
2960 __skb_put(skb, len);
2961 }
2962 return 0;
2963 }
2964
2965 /**
2966 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2967 * @skb: buffer to pad
2968 * @len: minimal length
2969 *
2970 * Pads up a buffer to ensure the trailing bytes exist and are
2971 * blanked. If the buffer already contains sufficient data it
2972 * is untouched. Otherwise it is extended. Returns zero on
2973 * success. The skb is freed on error.
2974 */
2975 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2976 {
2977 return __skb_put_padto(skb, len, true);
2978 }
2979
2980 static inline int skb_add_data(struct sk_buff *skb,
2981 struct iov_iter *from, int copy)
2982 {
2983 const int off = skb->len;
2984
2985 if (skb->ip_summed == CHECKSUM_NONE) {
2986 __wsum csum = 0;
2987 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2988 &csum, from)) {
2989 skb->csum = csum_block_add(skb->csum, csum, off);
2990 return 0;
2991 }
2992 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
2993 return 0;
2994
2995 __skb_trim(skb, off);
2996 return -EFAULT;
2997 }
2998
2999 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3000 const struct page *page, int off)
3001 {
3002 if (skb_zcopy(skb))
3003 return false;
3004 if (i) {
3005 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
3006
3007 return page == skb_frag_page(frag) &&
3008 off == frag->page_offset + skb_frag_size(frag);
3009 }
3010 return false;
3011 }
3012
3013 static inline int __skb_linearize(struct sk_buff *skb)
3014 {
3015 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3016 }
3017
3018 /**
3019 * skb_linearize - convert paged skb to linear one
3020 * @skb: buffer to linarize
3021 *
3022 * If there is no free memory -ENOMEM is returned, otherwise zero
3023 * is returned and the old skb data released.
3024 */
3025 static inline int skb_linearize(struct sk_buff *skb)
3026 {
3027 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3028 }
3029
3030 /**
3031 * skb_has_shared_frag - can any frag be overwritten
3032 * @skb: buffer to test
3033 *
3034 * Return true if the skb has at least one frag that might be modified
3035 * by an external entity (as in vmsplice()/sendfile())
3036 */
3037 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3038 {
3039 return skb_is_nonlinear(skb) &&
3040 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3041 }
3042
3043 /**
3044 * skb_linearize_cow - make sure skb is linear and writable
3045 * @skb: buffer to process
3046 *
3047 * If there is no free memory -ENOMEM is returned, otherwise zero
3048 * is returned and the old skb data released.
3049 */
3050 static inline int skb_linearize_cow(struct sk_buff *skb)
3051 {
3052 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3053 __skb_linearize(skb) : 0;
3054 }
3055
3056 static __always_inline void
3057 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3058 unsigned int off)
3059 {
3060 if (skb->ip_summed == CHECKSUM_COMPLETE)
3061 skb->csum = csum_block_sub(skb->csum,
3062 csum_partial(start, len, 0), off);
3063 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3064 skb_checksum_start_offset(skb) < 0)
3065 skb->ip_summed = CHECKSUM_NONE;
3066 }
3067
3068 /**
3069 * skb_postpull_rcsum - update checksum for received skb after pull
3070 * @skb: buffer to update
3071 * @start: start of data before pull
3072 * @len: length of data pulled
3073 *
3074 * After doing a pull on a received packet, you need to call this to
3075 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3076 * CHECKSUM_NONE so that it can be recomputed from scratch.
3077 */
3078 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3079 const void *start, unsigned int len)
3080 {
3081 __skb_postpull_rcsum(skb, start, len, 0);
3082 }
3083
3084 static __always_inline void
3085 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3086 unsigned int off)
3087 {
3088 if (skb->ip_summed == CHECKSUM_COMPLETE)
3089 skb->csum = csum_block_add(skb->csum,
3090 csum_partial(start, len, 0), off);
3091 }
3092
3093 /**
3094 * skb_postpush_rcsum - update checksum for received skb after push
3095 * @skb: buffer to update
3096 * @start: start of data after push
3097 * @len: length of data pushed
3098 *
3099 * After doing a push on a received packet, you need to call this to
3100 * update the CHECKSUM_COMPLETE checksum.
3101 */
3102 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3103 const void *start, unsigned int len)
3104 {
3105 __skb_postpush_rcsum(skb, start, len, 0);
3106 }
3107
3108 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3109
3110 /**
3111 * skb_push_rcsum - push skb and update receive checksum
3112 * @skb: buffer to update
3113 * @len: length of data pulled
3114 *
3115 * This function performs an skb_push on the packet and updates
3116 * the CHECKSUM_COMPLETE checksum. It should be used on
3117 * receive path processing instead of skb_push unless you know
3118 * that the checksum difference is zero (e.g., a valid IP header)
3119 * or you are setting ip_summed to CHECKSUM_NONE.
3120 */
3121 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3122 {
3123 skb_push(skb, len);
3124 skb_postpush_rcsum(skb, skb->data, len);
3125 return skb->data;
3126 }
3127
3128 /**
3129 * pskb_trim_rcsum - trim received skb and update checksum
3130 * @skb: buffer to trim
3131 * @len: new length
3132 *
3133 * This is exactly the same as pskb_trim except that it ensures the
3134 * checksum of received packets are still valid after the operation.
3135 */
3136
3137 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3138 {
3139 if (likely(len >= skb->len))
3140 return 0;
3141 if (skb->ip_summed == CHECKSUM_COMPLETE)
3142 skb->ip_summed = CHECKSUM_NONE;
3143 return __pskb_trim(skb, len);
3144 }
3145
3146 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3147 {
3148 if (skb->ip_summed == CHECKSUM_COMPLETE)
3149 skb->ip_summed = CHECKSUM_NONE;
3150 __skb_trim(skb, len);
3151 return 0;
3152 }
3153
3154 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3155 {
3156 if (skb->ip_summed == CHECKSUM_COMPLETE)
3157 skb->ip_summed = CHECKSUM_NONE;
3158 return __skb_grow(skb, len);
3159 }
3160
3161 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3162 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3163 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3164 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3165 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3166
3167 #define skb_queue_walk(queue, skb) \
3168 for (skb = (queue)->next; \
3169 skb != (struct sk_buff *)(queue); \
3170 skb = skb->next)
3171
3172 #define skb_queue_walk_safe(queue, skb, tmp) \
3173 for (skb = (queue)->next, tmp = skb->next; \
3174 skb != (struct sk_buff *)(queue); \
3175 skb = tmp, tmp = skb->next)
3176
3177 #define skb_queue_walk_from(queue, skb) \
3178 for (; skb != (struct sk_buff *)(queue); \
3179 skb = skb->next)
3180
3181 #define skb_rbtree_walk(skb, root) \
3182 for (skb = skb_rb_first(root); skb != NULL; \
3183 skb = skb_rb_next(skb))
3184
3185 #define skb_rbtree_walk_from(skb) \
3186 for (; skb != NULL; \
3187 skb = skb_rb_next(skb))
3188
3189 #define skb_rbtree_walk_from_safe(skb, tmp) \
3190 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3191 skb = tmp)
3192
3193 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3194 for (tmp = skb->next; \
3195 skb != (struct sk_buff *)(queue); \
3196 skb = tmp, tmp = skb->next)
3197
3198 #define skb_queue_reverse_walk(queue, skb) \
3199 for (skb = (queue)->prev; \
3200 skb != (struct sk_buff *)(queue); \
3201 skb = skb->prev)
3202
3203 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3204 for (skb = (queue)->prev, tmp = skb->prev; \
3205 skb != (struct sk_buff *)(queue); \
3206 skb = tmp, tmp = skb->prev)
3207
3208 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3209 for (tmp = skb->prev; \
3210 skb != (struct sk_buff *)(queue); \
3211 skb = tmp, tmp = skb->prev)
3212
3213 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3214 {
3215 return skb_shinfo(skb)->frag_list != NULL;
3216 }
3217
3218 static inline void skb_frag_list_init(struct sk_buff *skb)
3219 {
3220 skb_shinfo(skb)->frag_list = NULL;
3221 }
3222
3223 #define skb_walk_frags(skb, iter) \
3224 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3225
3226
3227 int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3228 const struct sk_buff *skb);
3229 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3230 struct sk_buff_head *queue,
3231 unsigned int flags,
3232 void (*destructor)(struct sock *sk,
3233 struct sk_buff *skb),
3234 int *peeked, int *off, int *err,
3235 struct sk_buff **last);
3236 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3237 void (*destructor)(struct sock *sk,
3238 struct sk_buff *skb),
3239 int *peeked, int *off, int *err,
3240 struct sk_buff **last);
3241 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3242 void (*destructor)(struct sock *sk,
3243 struct sk_buff *skb),
3244 int *peeked, int *off, int *err);
3245 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3246 int *err);
3247 unsigned int datagram_poll(struct file *file, struct socket *sock,
3248 struct poll_table_struct *wait);
3249 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3250 struct iov_iter *to, int size);
3251 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3252 struct msghdr *msg, int size)
3253 {
3254 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3255 }
3256 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3257 struct msghdr *msg);
3258 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3259 struct iov_iter *from, int len);
3260 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3261 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3262 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3263 static inline void skb_free_datagram_locked(struct sock *sk,
3264 struct sk_buff *skb)
3265 {
3266 __skb_free_datagram_locked(sk, skb, 0);
3267 }
3268 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3269 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3270 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3271 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3272 int len, __wsum csum);
3273 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3274 struct pipe_inode_info *pipe, unsigned int len,
3275 unsigned int flags);
3276 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3277 int len);
3278 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3279 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3280 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3281 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3282 int len, int hlen);
3283 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3284 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3285 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3286 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
3287 bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
3288 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3289 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3290 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3291 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3292 int skb_vlan_pop(struct sk_buff *skb);
3293 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3294 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3295 gfp_t gfp);
3296
3297 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3298 {
3299 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3300 }
3301
3302 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3303 {
3304 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3305 }
3306
3307 struct skb_checksum_ops {
3308 __wsum (*update)(const void *mem, int len, __wsum wsum);
3309 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3310 };
3311
3312 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3313
3314 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3315 __wsum csum, const struct skb_checksum_ops *ops);
3316 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3317 __wsum csum);
3318
3319 static inline void * __must_check
3320 __skb_header_pointer(const struct sk_buff *skb, int offset,
3321 int len, void *data, int hlen, void *buffer)
3322 {
3323 if (hlen - offset >= len)
3324 return data + offset;
3325
3326 if (!skb ||
3327 skb_copy_bits(skb, offset, buffer, len) < 0)
3328 return NULL;
3329
3330 return buffer;
3331 }
3332
3333 static inline void * __must_check
3334 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3335 {
3336 return __skb_header_pointer(skb, offset, len, skb->data,
3337 skb_headlen(skb), buffer);
3338 }
3339
3340 /**
3341 * skb_needs_linearize - check if we need to linearize a given skb
3342 * depending on the given device features.
3343 * @skb: socket buffer to check
3344 * @features: net device features
3345 *
3346 * Returns true if either:
3347 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3348 * 2. skb is fragmented and the device does not support SG.
3349 */
3350 static inline bool skb_needs_linearize(struct sk_buff *skb,
3351 netdev_features_t features)
3352 {
3353 return skb_is_nonlinear(skb) &&
3354 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3355 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3356 }
3357
3358 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3359 void *to,
3360 const unsigned int len)
3361 {
3362 memcpy(to, skb->data, len);
3363 }
3364
3365 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3366 const int offset, void *to,
3367 const unsigned int len)
3368 {
3369 memcpy(to, skb->data + offset, len);
3370 }
3371
3372 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3373 const void *from,
3374 const unsigned int len)
3375 {
3376 memcpy(skb->data, from, len);
3377 }
3378
3379 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3380 const int offset,
3381 const void *from,
3382 const unsigned int len)
3383 {
3384 memcpy(skb->data + offset, from, len);
3385 }
3386
3387 void skb_init(void);
3388
3389 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3390 {
3391 return skb->tstamp;
3392 }
3393
3394 /**
3395 * skb_get_timestamp - get timestamp from a skb
3396 * @skb: skb to get stamp from
3397 * @stamp: pointer to struct timeval to store stamp in
3398 *
3399 * Timestamps are stored in the skb as offsets to a base timestamp.
3400 * This function converts the offset back to a struct timeval and stores
3401 * it in stamp.
3402 */
3403 static inline void skb_get_timestamp(const struct sk_buff *skb,
3404 struct timeval *stamp)
3405 {
3406 *stamp = ktime_to_timeval(skb->tstamp);
3407 }
3408
3409 static inline void skb_get_timestampns(const struct sk_buff *skb,
3410 struct timespec *stamp)
3411 {
3412 *stamp = ktime_to_timespec(skb->tstamp);
3413 }
3414
3415 static inline void __net_timestamp(struct sk_buff *skb)
3416 {
3417 skb->tstamp = ktime_get_real();
3418 }
3419
3420 static inline ktime_t net_timedelta(ktime_t t)
3421 {
3422 return ktime_sub(ktime_get_real(), t);
3423 }
3424
3425 static inline ktime_t net_invalid_timestamp(void)
3426 {
3427 return 0;
3428 }
3429
3430 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3431 {
3432 return skb_shinfo(skb)->meta_len;
3433 }
3434
3435 static inline void *skb_metadata_end(const struct sk_buff *skb)
3436 {
3437 return skb_mac_header(skb);
3438 }
3439
3440 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3441 const struct sk_buff *skb_b,
3442 u8 meta_len)
3443 {
3444 const void *a = skb_metadata_end(skb_a);
3445 const void *b = skb_metadata_end(skb_b);
3446 /* Using more efficient varaiant than plain call to memcmp(). */
3447 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3448 u64 diffs = 0;
3449
3450 switch (meta_len) {
3451 #define __it(x, op) (x -= sizeof(u##op))
3452 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3453 case 32: diffs |= __it_diff(a, b, 64);
3454 case 24: diffs |= __it_diff(a, b, 64);
3455 case 16: diffs |= __it_diff(a, b, 64);
3456 case 8: diffs |= __it_diff(a, b, 64);
3457 break;
3458 case 28: diffs |= __it_diff(a, b, 64);
3459 case 20: diffs |= __it_diff(a, b, 64);
3460 case 12: diffs |= __it_diff(a, b, 64);
3461 case 4: diffs |= __it_diff(a, b, 32);
3462 break;
3463 }
3464 return diffs;
3465 #else
3466 return memcmp(a - meta_len, b - meta_len, meta_len);
3467 #endif
3468 }
3469
3470 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3471 const struct sk_buff *skb_b)
3472 {
3473 u8 len_a = skb_metadata_len(skb_a);
3474 u8 len_b = skb_metadata_len(skb_b);
3475
3476 if (!(len_a | len_b))
3477 return false;
3478
3479 return len_a != len_b ?
3480 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3481 }
3482
3483 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3484 {
3485 skb_shinfo(skb)->meta_len = meta_len;
3486 }
3487
3488 static inline void skb_metadata_clear(struct sk_buff *skb)
3489 {
3490 skb_metadata_set(skb, 0);
3491 }
3492
3493 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3494
3495 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3496
3497 void skb_clone_tx_timestamp(struct sk_buff *skb);
3498 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3499
3500 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3501
3502 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3503 {
3504 }
3505
3506 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3507 {
3508 return false;
3509 }
3510
3511 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3512
3513 /**
3514 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3515 *
3516 * PHY drivers may accept clones of transmitted packets for
3517 * timestamping via their phy_driver.txtstamp method. These drivers
3518 * must call this function to return the skb back to the stack with a
3519 * timestamp.
3520 *
3521 * @skb: clone of the the original outgoing packet
3522 * @hwtstamps: hardware time stamps
3523 *
3524 */
3525 void skb_complete_tx_timestamp(struct sk_buff *skb,
3526 struct skb_shared_hwtstamps *hwtstamps);
3527
3528 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3529 struct skb_shared_hwtstamps *hwtstamps,
3530 struct sock *sk, int tstype);
3531
3532 /**
3533 * skb_tstamp_tx - queue clone of skb with send time stamps
3534 * @orig_skb: the original outgoing packet
3535 * @hwtstamps: hardware time stamps, may be NULL if not available
3536 *
3537 * If the skb has a socket associated, then this function clones the
3538 * skb (thus sharing the actual data and optional structures), stores
3539 * the optional hardware time stamping information (if non NULL) or
3540 * generates a software time stamp (otherwise), then queues the clone
3541 * to the error queue of the socket. Errors are silently ignored.
3542 */
3543 void skb_tstamp_tx(struct sk_buff *orig_skb,
3544 struct skb_shared_hwtstamps *hwtstamps);
3545
3546 /**
3547 * skb_tx_timestamp() - Driver hook for transmit timestamping
3548 *
3549 * Ethernet MAC Drivers should call this function in their hard_xmit()
3550 * function immediately before giving the sk_buff to the MAC hardware.
3551 *
3552 * Specifically, one should make absolutely sure that this function is
3553 * called before TX completion of this packet can trigger. Otherwise
3554 * the packet could potentially already be freed.
3555 *
3556 * @skb: A socket buffer.
3557 */
3558 static inline void skb_tx_timestamp(struct sk_buff *skb)
3559 {
3560 skb_clone_tx_timestamp(skb);
3561 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3562 skb_tstamp_tx(skb, NULL);
3563 }
3564
3565 /**
3566 * skb_complete_wifi_ack - deliver skb with wifi status
3567 *
3568 * @skb: the original outgoing packet
3569 * @acked: ack status
3570 *
3571 */
3572 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3573
3574 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3575 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3576
3577 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3578 {
3579 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3580 skb->csum_valid ||
3581 (skb->ip_summed == CHECKSUM_PARTIAL &&
3582 skb_checksum_start_offset(skb) >= 0));
3583 }
3584
3585 /**
3586 * skb_checksum_complete - Calculate checksum of an entire packet
3587 * @skb: packet to process
3588 *
3589 * This function calculates the checksum over the entire packet plus
3590 * the value of skb->csum. The latter can be used to supply the
3591 * checksum of a pseudo header as used by TCP/UDP. It returns the
3592 * checksum.
3593 *
3594 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3595 * this function can be used to verify that checksum on received
3596 * packets. In that case the function should return zero if the
3597 * checksum is correct. In particular, this function will return zero
3598 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3599 * hardware has already verified the correctness of the checksum.
3600 */
3601 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3602 {
3603 return skb_csum_unnecessary(skb) ?
3604 0 : __skb_checksum_complete(skb);
3605 }
3606
3607 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3608 {
3609 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3610 if (skb->csum_level == 0)
3611 skb->ip_summed = CHECKSUM_NONE;
3612 else
3613 skb->csum_level--;
3614 }
3615 }
3616
3617 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3618 {
3619 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3620 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3621 skb->csum_level++;
3622 } else if (skb->ip_summed == CHECKSUM_NONE) {
3623 skb->ip_summed = CHECKSUM_UNNECESSARY;
3624 skb->csum_level = 0;
3625 }
3626 }
3627
3628 /* Check if we need to perform checksum complete validation.
3629 *
3630 * Returns true if checksum complete is needed, false otherwise
3631 * (either checksum is unnecessary or zero checksum is allowed).
3632 */
3633 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3634 bool zero_okay,
3635 __sum16 check)
3636 {
3637 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3638 skb->csum_valid = 1;
3639 __skb_decr_checksum_unnecessary(skb);
3640 return false;
3641 }
3642
3643 return true;
3644 }
3645
3646 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3647 * in checksum_init.
3648 */
3649 #define CHECKSUM_BREAK 76
3650
3651 /* Unset checksum-complete
3652 *
3653 * Unset checksum complete can be done when packet is being modified
3654 * (uncompressed for instance) and checksum-complete value is
3655 * invalidated.
3656 */
3657 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3658 {
3659 if (skb->ip_summed == CHECKSUM_COMPLETE)
3660 skb->ip_summed = CHECKSUM_NONE;
3661 }
3662
3663 /* Validate (init) checksum based on checksum complete.
3664 *
3665 * Return values:
3666 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3667 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3668 * checksum is stored in skb->csum for use in __skb_checksum_complete
3669 * non-zero: value of invalid checksum
3670 *
3671 */
3672 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3673 bool complete,
3674 __wsum psum)
3675 {
3676 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3677 if (!csum_fold(csum_add(psum, skb->csum))) {
3678 skb->csum_valid = 1;
3679 return 0;
3680 }
3681 }
3682
3683 skb->csum = psum;
3684
3685 if (complete || skb->len <= CHECKSUM_BREAK) {
3686 __sum16 csum;
3687
3688 csum = __skb_checksum_complete(skb);
3689 skb->csum_valid = !csum;
3690 return csum;
3691 }
3692
3693 return 0;
3694 }
3695
3696 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3697 {
3698 return 0;
3699 }
3700
3701 /* Perform checksum validate (init). Note that this is a macro since we only
3702 * want to calculate the pseudo header which is an input function if necessary.
3703 * First we try to validate without any computation (checksum unnecessary) and
3704 * then calculate based on checksum complete calling the function to compute
3705 * pseudo header.
3706 *
3707 * Return values:
3708 * 0: checksum is validated or try to in skb_checksum_complete
3709 * non-zero: value of invalid checksum
3710 */
3711 #define __skb_checksum_validate(skb, proto, complete, \
3712 zero_okay, check, compute_pseudo) \
3713 ({ \
3714 __sum16 __ret = 0; \
3715 skb->csum_valid = 0; \
3716 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3717 __ret = __skb_checksum_validate_complete(skb, \
3718 complete, compute_pseudo(skb, proto)); \
3719 __ret; \
3720 })
3721
3722 #define skb_checksum_init(skb, proto, compute_pseudo) \
3723 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3724
3725 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3726 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3727
3728 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3729 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3730
3731 #define skb_checksum_validate_zero_check(skb, proto, check, \
3732 compute_pseudo) \
3733 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3734
3735 #define skb_checksum_simple_validate(skb) \
3736 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3737
3738 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3739 {
3740 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
3741 }
3742
3743 static inline void __skb_checksum_convert(struct sk_buff *skb,
3744 __sum16 check, __wsum pseudo)
3745 {
3746 skb->csum = ~pseudo;
3747 skb->ip_summed = CHECKSUM_COMPLETE;
3748 }
3749
3750 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3751 do { \
3752 if (__skb_checksum_convert_check(skb)) \
3753 __skb_checksum_convert(skb, check, \
3754 compute_pseudo(skb, proto)); \
3755 } while (0)
3756
3757 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3758 u16 start, u16 offset)
3759 {
3760 skb->ip_summed = CHECKSUM_PARTIAL;
3761 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3762 skb->csum_offset = offset - start;
3763 }
3764
3765 /* Update skbuf and packet to reflect the remote checksum offload operation.
3766 * When called, ptr indicates the starting point for skb->csum when
3767 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3768 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3769 */
3770 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3771 int start, int offset, bool nopartial)
3772 {
3773 __wsum delta;
3774
3775 if (!nopartial) {
3776 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3777 return;
3778 }
3779
3780 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3781 __skb_checksum_complete(skb);
3782 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3783 }
3784
3785 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3786
3787 /* Adjust skb->csum since we changed the packet */
3788 skb->csum = csum_add(skb->csum, delta);
3789 }
3790
3791 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3792 {
3793 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3794 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
3795 #else
3796 return NULL;
3797 #endif
3798 }
3799
3800 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3801 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3802 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3803 {
3804 if (nfct && atomic_dec_and_test(&nfct->use))
3805 nf_conntrack_destroy(nfct);
3806 }
3807 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3808 {
3809 if (nfct)
3810 atomic_inc(&nfct->use);
3811 }
3812 #endif
3813 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3814 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3815 {
3816 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
3817 kfree(nf_bridge);
3818 }
3819 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3820 {
3821 if (nf_bridge)
3822 refcount_inc(&nf_bridge->use);
3823 }
3824 #endif /* CONFIG_BRIDGE_NETFILTER */
3825 static inline void nf_reset(struct sk_buff *skb)
3826 {
3827 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3828 nf_conntrack_put(skb_nfct(skb));
3829 skb->_nfct = 0;
3830 #endif
3831 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3832 nf_bridge_put(skb->nf_bridge);
3833 skb->nf_bridge = NULL;
3834 #endif
3835 }
3836
3837 static inline void nf_reset_trace(struct sk_buff *skb)
3838 {
3839 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3840 skb->nf_trace = 0;
3841 #endif
3842 }
3843
3844 /* Note: This doesn't put any conntrack and bridge info in dst. */
3845 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3846 bool copy)
3847 {
3848 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3849 dst->_nfct = src->_nfct;
3850 nf_conntrack_get(skb_nfct(src));
3851 #endif
3852 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3853 dst->nf_bridge = src->nf_bridge;
3854 nf_bridge_get(src->nf_bridge);
3855 #endif
3856 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3857 if (copy)
3858 dst->nf_trace = src->nf_trace;
3859 #endif
3860 }
3861
3862 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3863 {
3864 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3865 nf_conntrack_put(skb_nfct(dst));
3866 #endif
3867 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3868 nf_bridge_put(dst->nf_bridge);
3869 #endif
3870 __nf_copy(dst, src, true);
3871 }
3872
3873 #ifdef CONFIG_NETWORK_SECMARK
3874 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3875 {
3876 to->secmark = from->secmark;
3877 }
3878
3879 static inline void skb_init_secmark(struct sk_buff *skb)
3880 {
3881 skb->secmark = 0;
3882 }
3883 #else
3884 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3885 { }
3886
3887 static inline void skb_init_secmark(struct sk_buff *skb)
3888 { }
3889 #endif
3890
3891 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3892 {
3893 return !skb->destructor &&
3894 #if IS_ENABLED(CONFIG_XFRM)
3895 !skb->sp &&
3896 #endif
3897 !skb_nfct(skb) &&
3898 !skb->_skb_refdst &&
3899 !skb_has_frag_list(skb);
3900 }
3901
3902 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3903 {
3904 skb->queue_mapping = queue_mapping;
3905 }
3906
3907 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3908 {
3909 return skb->queue_mapping;
3910 }
3911
3912 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3913 {
3914 to->queue_mapping = from->queue_mapping;
3915 }
3916
3917 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3918 {
3919 skb->queue_mapping = rx_queue + 1;
3920 }
3921
3922 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3923 {
3924 return skb->queue_mapping - 1;
3925 }
3926
3927 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3928 {
3929 return skb->queue_mapping != 0;
3930 }
3931
3932 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3933 {
3934 skb->dst_pending_confirm = val;
3935 }
3936
3937 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3938 {
3939 return skb->dst_pending_confirm != 0;
3940 }
3941
3942 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3943 {
3944 #ifdef CONFIG_XFRM
3945 return skb->sp;
3946 #else
3947 return NULL;
3948 #endif
3949 }
3950
3951 /* Keeps track of mac header offset relative to skb->head.
3952 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3953 * For non-tunnel skb it points to skb_mac_header() and for
3954 * tunnel skb it points to outer mac header.
3955 * Keeps track of level of encapsulation of network headers.
3956 */
3957 struct skb_gso_cb {
3958 union {
3959 int mac_offset;
3960 int data_offset;
3961 };
3962 int encap_level;
3963 __wsum csum;
3964 __u16 csum_start;
3965 };
3966 #define SKB_SGO_CB_OFFSET 32
3967 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
3968
3969 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3970 {
3971 return (skb_mac_header(inner_skb) - inner_skb->head) -
3972 SKB_GSO_CB(inner_skb)->mac_offset;
3973 }
3974
3975 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3976 {
3977 int new_headroom, headroom;
3978 int ret;
3979
3980 headroom = skb_headroom(skb);
3981 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3982 if (ret)
3983 return ret;
3984
3985 new_headroom = skb_headroom(skb);
3986 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3987 return 0;
3988 }
3989
3990 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3991 {
3992 /* Do not update partial checksums if remote checksum is enabled. */
3993 if (skb->remcsum_offload)
3994 return;
3995
3996 SKB_GSO_CB(skb)->csum = res;
3997 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3998 }
3999
4000 /* Compute the checksum for a gso segment. First compute the checksum value
4001 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4002 * then add in skb->csum (checksum from csum_start to end of packet).
4003 * skb->csum and csum_start are then updated to reflect the checksum of the
4004 * resultant packet starting from the transport header-- the resultant checksum
4005 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4006 * header.
4007 */
4008 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4009 {
4010 unsigned char *csum_start = skb_transport_header(skb);
4011 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4012 __wsum partial = SKB_GSO_CB(skb)->csum;
4013
4014 SKB_GSO_CB(skb)->csum = res;
4015 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4016
4017 return csum_fold(csum_partial(csum_start, plen, partial));
4018 }
4019
4020 static inline bool skb_is_gso(const struct sk_buff *skb)
4021 {
4022 return skb_shinfo(skb)->gso_size;
4023 }
4024
4025 /* Note: Should be called only if skb_is_gso(skb) is true */
4026 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4027 {
4028 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4029 }
4030
4031 static inline void skb_gso_reset(struct sk_buff *skb)
4032 {
4033 skb_shinfo(skb)->gso_size = 0;
4034 skb_shinfo(skb)->gso_segs = 0;
4035 skb_shinfo(skb)->gso_type = 0;
4036 }
4037
4038 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4039
4040 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4041 {
4042 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4043 * wanted then gso_type will be set. */
4044 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4045
4046 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4047 unlikely(shinfo->gso_type == 0)) {
4048 __skb_warn_lro_forwarding(skb);
4049 return true;
4050 }
4051 return false;
4052 }
4053
4054 static inline void skb_forward_csum(struct sk_buff *skb)
4055 {
4056 /* Unfortunately we don't support this one. Any brave souls? */
4057 if (skb->ip_summed == CHECKSUM_COMPLETE)
4058 skb->ip_summed = CHECKSUM_NONE;
4059 }
4060
4061 /**
4062 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4063 * @skb: skb to check
4064 *
4065 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4066 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4067 * use this helper, to document places where we make this assertion.
4068 */
4069 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4070 {
4071 #ifdef DEBUG
4072 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4073 #endif
4074 }
4075
4076 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4077
4078 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4079 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4080 unsigned int transport_len,
4081 __sum16(*skb_chkf)(struct sk_buff *skb));
4082
4083 /**
4084 * skb_head_is_locked - Determine if the skb->head is locked down
4085 * @skb: skb to check
4086 *
4087 * The head on skbs build around a head frag can be removed if they are
4088 * not cloned. This function returns true if the skb head is locked down
4089 * due to either being allocated via kmalloc, or by being a clone with
4090 * multiple references to the head.
4091 */
4092 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4093 {
4094 return !skb->head_frag || skb_cloned(skb);
4095 }
4096
4097 /**
4098 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4099 *
4100 * @skb: GSO skb
4101 *
4102 * skb_gso_network_seglen is used to determine the real size of the
4103 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4104 *
4105 * The MAC/L2 header is not accounted for.
4106 */
4107 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4108 {
4109 unsigned int hdr_len = skb_transport_header(skb) -
4110 skb_network_header(skb);
4111 return hdr_len + skb_gso_transport_seglen(skb);
4112 }
4113
4114 /* Local Checksum Offload.
4115 * Compute outer checksum based on the assumption that the
4116 * inner checksum will be offloaded later.
4117 * See Documentation/networking/checksum-offloads.txt for
4118 * explanation of how this works.
4119 * Fill in outer checksum adjustment (e.g. with sum of outer
4120 * pseudo-header) before calling.
4121 * Also ensure that inner checksum is in linear data area.
4122 */
4123 static inline __wsum lco_csum(struct sk_buff *skb)
4124 {
4125 unsigned char *csum_start = skb_checksum_start(skb);
4126 unsigned char *l4_hdr = skb_transport_header(skb);
4127 __wsum partial;
4128
4129 /* Start with complement of inner checksum adjustment */
4130 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4131 skb->csum_offset));
4132
4133 /* Add in checksum of our headers (incl. outer checksum
4134 * adjustment filled in by caller) and return result.
4135 */
4136 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4137 }
4138
4139 #endif /* __KERNEL__ */
4140 #endif /* _LINUX_SKBUFF_H */