[RAMEN9610-21564]net/flow_dissector: switch to siphash
[GitHub/MotorolaMobilityLLC/kernel-slsi.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/compiler.h>
19 #include <linux/time.h>
20 #include <linux/bug.h>
21 #include <linux/cache.h>
22 #include <linux/rbtree.h>
23 #include <linux/socket.h>
24 #include <linux/refcount.h>
25
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <linux/sched/clock.h>
38 #include <net/flow_dissector.h>
39 #include <linux/splice.h>
40 #include <linux/in6.h>
41 #include <linux/if_packet.h>
42 #include <net/flow.h>
43
44 /* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
89 *
90 * CHECKSUM_NONE:
91 *
92 * Device did not checksum this packet e.g. due to lack of capabilities.
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
112 * FCOE: indicates the CRC in FC frame has been validated.
113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
134 *
135 * CHECKSUM_PARTIAL:
136 *
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
146 *
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
183 *
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
217 */
218
219 /* Don't change this without changing skb_csum_unnecessary! */
220 #define CHECKSUM_NONE 0
221 #define CHECKSUM_UNNECESSARY 1
222 #define CHECKSUM_COMPLETE 2
223 #define CHECKSUM_PARTIAL 3
224
225 /* Maximum value in skb->csum_level */
226 #define SKB_MAX_CSUM_LEVEL 3
227
228 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
229 #define SKB_WITH_OVERHEAD(X) \
230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
231 #define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
233 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
236 /* return minimum truesize of one skb containing X bytes of data */
237 #define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
241 struct net_device;
242 struct scatterlist;
243 struct pipe_inode_info;
244 struct iov_iter;
245 struct napi_struct;
246
247 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
248 struct nf_conntrack {
249 atomic_t use;
250 };
251 #endif
252
253 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
254 struct nf_bridge_info {
255 refcount_t use;
256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
260 } orig_proto:8;
261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
264 __u16 frag_max_size;
265 struct net_device *physindev;
266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
269 union {
270 /* prerouting: detect dnat in orig/reply direction */
271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
279 };
280 };
281 #endif
282
283 struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290 };
291
292 struct sk_buff;
293
294 /* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
300 */
301 #if (65536/PAGE_SIZE + 1) < 16
302 #define MAX_SKB_FRAGS 16UL
303 #else
304 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
305 #endif
306 extern int sysctl_max_skb_frags;
307
308 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311 #define GSO_BY_FRAGS 0xFFFF
312
313 typedef struct skb_frag_struct skb_frag_t;
314
315 struct skb_frag_struct {
316 struct {
317 struct page *p;
318 } page;
319 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
320 __u32 page_offset;
321 __u32 size;
322 #else
323 __u16 page_offset;
324 __u16 size;
325 #endif
326 };
327
328 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329 {
330 return frag->size;
331 }
332
333 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334 {
335 frag->size = size;
336 }
337
338 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339 {
340 frag->size += delta;
341 }
342
343 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344 {
345 frag->size -= delta;
346 }
347
348 static inline bool skb_frag_must_loop(struct page *p)
349 {
350 #if defined(CONFIG_HIGHMEM)
351 if (PageHighMem(p))
352 return true;
353 #endif
354 return false;
355 }
356
357 /**
358 * skb_frag_foreach_page - loop over pages in a fragment
359 *
360 * @f: skb frag to operate on
361 * @f_off: offset from start of f->page.p
362 * @f_len: length from f_off to loop over
363 * @p: (temp var) current page
364 * @p_off: (temp var) offset from start of current page,
365 * non-zero only on first page.
366 * @p_len: (temp var) length in current page,
367 * < PAGE_SIZE only on first and last page.
368 * @copied: (temp var) length so far, excluding current p_len.
369 *
370 * A fragment can hold a compound page, in which case per-page
371 * operations, notably kmap_atomic, must be called for each
372 * regular page.
373 */
374 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
375 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
376 p_off = (f_off) & (PAGE_SIZE - 1), \
377 p_len = skb_frag_must_loop(p) ? \
378 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
379 copied = 0; \
380 copied < f_len; \
381 copied += p_len, p++, p_off = 0, \
382 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
383
384 #define HAVE_HW_TIME_STAMP
385
386 /**
387 * struct skb_shared_hwtstamps - hardware time stamps
388 * @hwtstamp: hardware time stamp transformed into duration
389 * since arbitrary point in time
390 *
391 * Software time stamps generated by ktime_get_real() are stored in
392 * skb->tstamp.
393 *
394 * hwtstamps can only be compared against other hwtstamps from
395 * the same device.
396 *
397 * This structure is attached to packets as part of the
398 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
399 */
400 struct skb_shared_hwtstamps {
401 ktime_t hwtstamp;
402 };
403
404 /* Definitions for tx_flags in struct skb_shared_info */
405 enum {
406 /* generate hardware time stamp */
407 SKBTX_HW_TSTAMP = 1 << 0,
408
409 /* generate software time stamp when queueing packet to NIC */
410 SKBTX_SW_TSTAMP = 1 << 1,
411
412 /* device driver is going to provide hardware time stamp */
413 SKBTX_IN_PROGRESS = 1 << 2,
414
415 /* device driver supports TX zero-copy buffers */
416 SKBTX_DEV_ZEROCOPY = 1 << 3,
417
418 /* generate wifi status information (where possible) */
419 SKBTX_WIFI_STATUS = 1 << 4,
420
421 /* This indicates at least one fragment might be overwritten
422 * (as in vmsplice(), sendfile() ...)
423 * If we need to compute a TX checksum, we'll need to copy
424 * all frags to avoid possible bad checksum
425 */
426 SKBTX_SHARED_FRAG = 1 << 5,
427
428 /* generate software time stamp when entering packet scheduling */
429 SKBTX_SCHED_TSTAMP = 1 << 6,
430 };
431
432 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
433 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
434 SKBTX_SCHED_TSTAMP)
435 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
436
437 /*
438 * The callback notifies userspace to release buffers when skb DMA is done in
439 * lower device, the skb last reference should be 0 when calling this.
440 * The zerocopy_success argument is true if zero copy transmit occurred,
441 * false on data copy or out of memory error caused by data copy attempt.
442 * The ctx field is used to track device context.
443 * The desc field is used to track userspace buffer index.
444 */
445 struct ubuf_info {
446 void (*callback)(struct ubuf_info *, bool zerocopy_success);
447 union {
448 struct {
449 unsigned long desc;
450 void *ctx;
451 };
452 struct {
453 u32 id;
454 u16 len;
455 u16 zerocopy:1;
456 u32 bytelen;
457 };
458 };
459 refcount_t refcnt;
460
461 struct mmpin {
462 struct user_struct *user;
463 unsigned int num_pg;
464 } mmp;
465 };
466
467 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
468
469 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
470 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
471 struct ubuf_info *uarg);
472
473 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
474 {
475 refcount_inc(&uarg->refcnt);
476 }
477
478 void sock_zerocopy_put(struct ubuf_info *uarg);
479 void sock_zerocopy_put_abort(struct ubuf_info *uarg);
480
481 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
482
483 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
484 struct msghdr *msg, int len,
485 struct ubuf_info *uarg);
486
487 /* This data is invariant across clones and lives at
488 * the end of the header data, ie. at skb->end.
489 */
490 struct skb_shared_info {
491 unsigned short _unused;
492 unsigned char nr_frags;
493 __u8 tx_flags;
494 unsigned short gso_size;
495 /* Warning: this field is not always filled in (UFO)! */
496 unsigned short gso_segs;
497 struct sk_buff *frag_list;
498 struct skb_shared_hwtstamps hwtstamps;
499 unsigned int gso_type;
500 u32 tskey;
501 __be32 ip6_frag_id;
502
503 /*
504 * Warning : all fields before dataref are cleared in __alloc_skb()
505 */
506 atomic_t dataref;
507
508 /* Intermediate layers must ensure that destructor_arg
509 * remains valid until skb destructor */
510 void * destructor_arg;
511
512 /* must be last field, see pskb_expand_head() */
513 skb_frag_t frags[MAX_SKB_FRAGS];
514 };
515
516 /* We divide dataref into two halves. The higher 16 bits hold references
517 * to the payload part of skb->data. The lower 16 bits hold references to
518 * the entire skb->data. A clone of a headerless skb holds the length of
519 * the header in skb->hdr_len.
520 *
521 * All users must obey the rule that the skb->data reference count must be
522 * greater than or equal to the payload reference count.
523 *
524 * Holding a reference to the payload part means that the user does not
525 * care about modifications to the header part of skb->data.
526 */
527 #define SKB_DATAREF_SHIFT 16
528 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
529
530
531 enum {
532 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
533 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
534 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
535 };
536
537 enum {
538 SKB_GSO_TCPV4 = 1 << 0,
539
540 /* This indicates the skb is from an untrusted source. */
541 SKB_GSO_DODGY = 1 << 1,
542
543 /* This indicates the tcp segment has CWR set. */
544 SKB_GSO_TCP_ECN = 1 << 2,
545
546 SKB_GSO_TCP_FIXEDID = 1 << 3,
547
548 SKB_GSO_TCPV6 = 1 << 4,
549
550 SKB_GSO_FCOE = 1 << 5,
551
552 SKB_GSO_GRE = 1 << 6,
553
554 SKB_GSO_GRE_CSUM = 1 << 7,
555
556 SKB_GSO_IPXIP4 = 1 << 8,
557
558 SKB_GSO_IPXIP6 = 1 << 9,
559
560 SKB_GSO_UDP_TUNNEL = 1 << 10,
561
562 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
563
564 SKB_GSO_PARTIAL = 1 << 12,
565
566 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
567
568 SKB_GSO_SCTP = 1 << 14,
569
570 SKB_GSO_ESP = 1 << 15,
571
572 SKB_GSO_UDP = 1 << 16,
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 * @_nfct: Associated connection, if any (with nfctinfo bits)
621 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
622 * @skb_iif: ifindex of device we arrived on
623 * @tc_index: Traffic control index
624 * @hash: the packet hash
625 * @queue_mapping: Queue mapping for multiqueue devices
626 * @xmit_more: More SKBs are pending for this queue
627 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
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, ip4 defrag, and tcp stack */
675 struct list_head list;
676 };
677
678 union {
679 struct sock *sk;
680 int ip_defrag_offset;
681 };
682
683 union {
684 ktime_t tstamp;
685 u64 skb_mstamp;
686 };
687 /*
688 * This is the control buffer. It is free to use for every
689 * layer. Please put your private variables there. If you
690 * want to keep them across layers you have to do a skb_clone()
691 * first. This is owned by whoever has the skb queued ATM.
692 */
693 char cb[48] __aligned(8);
694
695 unsigned long _skb_refdst;
696 void (*destructor)(struct sk_buff *skb);
697 #ifdef CONFIG_XFRM
698 struct sec_path *sp;
699 #endif
700 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
701 unsigned long _nfct;
702 #endif
703 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
704 struct nf_bridge_info *nf_bridge;
705 #endif
706 unsigned int len,
707 data_len;
708 __u16 mac_len,
709 hdr_len;
710
711 /* Following fields are _not_ copied in __copy_skb_header()
712 * Note that queue_mapping is here mostly to fill a hole.
713 */
714 __u16 queue_mapping;
715
716 /* if you move cloned around you also must adapt those constants */
717 #ifdef __BIG_ENDIAN_BITFIELD
718 #define CLONED_MASK (1 << 7)
719 #else
720 #define CLONED_MASK 1
721 #endif
722 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
723
724 __u8 __cloned_offset[0];
725 __u8 cloned:1,
726 nohdr:1,
727 fclone:2,
728 peeked:1,
729 head_frag:1,
730 xmit_more:1,
731 pfmemalloc:1;
732
733 /* fields enclosed in headers_start/headers_end are copied
734 * using a single memcpy() in __copy_skb_header()
735 */
736 /* private: */
737 __u32 headers_start[0];
738 /* public: */
739
740 /* if you move pkt_type around you also must adapt those constants */
741 #ifdef __BIG_ENDIAN_BITFIELD
742 #define PKT_TYPE_MAX (7 << 5)
743 #else
744 #define PKT_TYPE_MAX 7
745 #endif
746 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
747
748 __u8 __pkt_type_offset[0];
749 __u8 pkt_type:3;
750 __u8 ignore_df:1;
751 __u8 nf_trace:1;
752 __u8 ip_summed:2;
753 __u8 ooo_okay:1;
754
755 __u8 l4_hash:1;
756 __u8 sw_hash:1;
757 __u8 wifi_acked_valid:1;
758 __u8 wifi_acked:1;
759 __u8 no_fcs:1;
760 /* Indicates the inner headers are valid in the skbuff. */
761 __u8 encapsulation:1;
762 __u8 encap_hdr_csum:1;
763 __u8 csum_valid:1;
764
765 __u8 csum_complete_sw:1;
766 __u8 csum_level:2;
767 __u8 csum_not_inet:1;
768 __u8 dst_pending_confirm:1;
769 #ifdef CONFIG_IPV6_NDISC_NODETYPE
770 __u8 ndisc_nodetype:2;
771 #endif
772 __u8 ipvs_property:1;
773
774 __u8 inner_protocol_type:1;
775 __u8 remcsum_offload:1;
776 #ifdef CONFIG_NET_SWITCHDEV
777 __u8 offload_fwd_mark:1;
778 #endif
779 #ifdef CONFIG_NET_CLS_ACT
780 __u8 tc_skip_classify:1;
781 __u8 tc_at_ingress:1;
782 __u8 tc_redirected:1;
783 __u8 tc_from_ingress:1;
784 #endif
785
786 #ifdef CONFIG_NET_SCHED
787 __u16 tc_index; /* traffic control index */
788 #endif
789
790 union {
791 __wsum csum;
792 struct {
793 __u16 csum_start;
794 __u16 csum_offset;
795 };
796 };
797 __u32 priority;
798 int skb_iif;
799 __u32 hash;
800 __be16 vlan_proto;
801 __u16 vlan_tci;
802 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
803 union {
804 unsigned int napi_id;
805 unsigned int sender_cpu;
806 };
807 #endif
808 #ifdef CONFIG_NETWORK_SECMARK
809 __u32 secmark;
810 #endif
811
812 union {
813 __u32 mark;
814 __u32 reserved_tailroom;
815 };
816
817 union {
818 __be16 inner_protocol;
819 __u8 inner_ipproto;
820 };
821
822 __u16 inner_transport_header;
823 __u16 inner_network_header;
824 __u16 inner_mac_header;
825
826 __be16 protocol;
827 __u16 transport_header;
828 __u16 network_header;
829 __u16 mac_header;
830
831 /* private: */
832 __u32 headers_end[0];
833 /* public: */
834
835 /* These elements must be at the end, see alloc_skb() for details. */
836 sk_buff_data_t tail;
837 sk_buff_data_t end;
838 unsigned char *head,
839 *data;
840 unsigned int truesize;
841 refcount_t users;
842 };
843
844 #ifdef __KERNEL__
845 /*
846 * Handling routines are only of interest to the kernel
847 */
848 #include <linux/slab.h>
849
850
851 #define SKB_ALLOC_FCLONE 0x01
852 #define SKB_ALLOC_RX 0x02
853 #define SKB_ALLOC_NAPI 0x04
854
855 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
856 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
857 {
858 return unlikely(skb->pfmemalloc);
859 }
860
861 /*
862 * skb might have a dst pointer attached, refcounted or not.
863 * _skb_refdst low order bit is set if refcount was _not_ taken
864 */
865 #define SKB_DST_NOREF 1UL
866 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
867
868 #define SKB_NFCT_PTRMASK ~(7UL)
869 /**
870 * skb_dst - returns skb dst_entry
871 * @skb: buffer
872 *
873 * Returns skb dst_entry, regardless of reference taken or not.
874 */
875 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
876 {
877 /* If refdst was not refcounted, check we still are in a
878 * rcu_read_lock section
879 */
880 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
881 !rcu_read_lock_held() &&
882 !rcu_read_lock_bh_held());
883 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
884 }
885
886 /**
887 * skb_dst_set - sets skb dst
888 * @skb: buffer
889 * @dst: dst entry
890 *
891 * Sets skb dst, assuming a reference was taken on dst and should
892 * be released by skb_dst_drop()
893 */
894 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
895 {
896 skb->_skb_refdst = (unsigned long)dst;
897 }
898
899 /**
900 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
901 * @skb: buffer
902 * @dst: dst entry
903 *
904 * Sets skb dst, assuming a reference was not taken on dst.
905 * If dst entry is cached, we do not take reference and dst_release
906 * will be avoided by refdst_drop. If dst entry is not cached, we take
907 * reference, so that last dst_release can destroy the dst immediately.
908 */
909 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
910 {
911 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
912 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
913 }
914
915 /**
916 * skb_dst_is_noref - Test if skb dst isn't refcounted
917 * @skb: buffer
918 */
919 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
920 {
921 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
922 }
923
924 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
925 {
926 return (struct rtable *)skb_dst(skb);
927 }
928
929 /* For mangling skb->pkt_type from user space side from applications
930 * such as nft, tc, etc, we only allow a conservative subset of
931 * possible pkt_types to be set.
932 */
933 static inline bool skb_pkt_type_ok(u32 ptype)
934 {
935 return ptype <= PACKET_OTHERHOST;
936 }
937
938 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
939 {
940 #ifdef CONFIG_NET_RX_BUSY_POLL
941 return skb->napi_id;
942 #else
943 return 0;
944 #endif
945 }
946
947 /* decrement the reference count and return true if we can free the skb */
948 static inline bool skb_unref(struct sk_buff *skb)
949 {
950 if (unlikely(!skb))
951 return false;
952 if (likely(refcount_read(&skb->users) == 1))
953 smp_rmb();
954 else if (likely(!refcount_dec_and_test(&skb->users)))
955 return false;
956
957 return true;
958 }
959
960 void skb_release_head_state(struct sk_buff *skb);
961 void kfree_skb(struct sk_buff *skb);
962 void kfree_skb_list(struct sk_buff *segs);
963 void skb_tx_error(struct sk_buff *skb);
964 void consume_skb(struct sk_buff *skb);
965 void __consume_stateless_skb(struct sk_buff *skb);
966 void __kfree_skb(struct sk_buff *skb);
967 extern struct kmem_cache *skbuff_head_cache;
968
969 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
970 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
971 bool *fragstolen, int *delta_truesize);
972
973 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
974 int node);
975 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
976 struct sk_buff *build_skb(void *data, unsigned int frag_size);
977 static inline struct sk_buff *alloc_skb(unsigned int size,
978 gfp_t priority)
979 {
980 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
981 }
982
983 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
984 unsigned long data_len,
985 int max_page_order,
986 int *errcode,
987 gfp_t gfp_mask);
988
989 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
990 struct sk_buff_fclones {
991 struct sk_buff skb1;
992
993 struct sk_buff skb2;
994
995 refcount_t fclone_ref;
996 };
997
998 /**
999 * skb_fclone_busy - check if fclone is busy
1000 * @sk: socket
1001 * @skb: buffer
1002 *
1003 * Returns true if skb is a fast clone, and its clone is not freed.
1004 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1005 * so we also check that this didnt happen.
1006 */
1007 static inline bool skb_fclone_busy(const struct sock *sk,
1008 const struct sk_buff *skb)
1009 {
1010 const struct sk_buff_fclones *fclones;
1011
1012 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1013
1014 return skb->fclone == SKB_FCLONE_ORIG &&
1015 refcount_read(&fclones->fclone_ref) > 1 &&
1016 fclones->skb2.sk == sk;
1017 }
1018
1019 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1020 gfp_t priority)
1021 {
1022 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1023 }
1024
1025 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1026 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1027 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1028 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1029 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1030 gfp_t gfp_mask, bool fclone);
1031 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1032 gfp_t gfp_mask)
1033 {
1034 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1035 }
1036
1037 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1038 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1039 unsigned int headroom);
1040 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1041 int newtailroom, gfp_t priority);
1042 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1043 int offset, int len);
1044 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1045 int offset, int len);
1046 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1047 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1048
1049 /**
1050 * skb_pad - zero pad the tail of an skb
1051 * @skb: buffer to pad
1052 * @pad: space to pad
1053 *
1054 * Ensure that a buffer is followed by a padding area that is zero
1055 * filled. Used by network drivers which may DMA or transfer data
1056 * beyond the buffer end onto the wire.
1057 *
1058 * May return error in out of memory cases. The skb is freed on error.
1059 */
1060 static inline int skb_pad(struct sk_buff *skb, int pad)
1061 {
1062 return __skb_pad(skb, pad, true);
1063 }
1064 #define dev_kfree_skb(a) consume_skb(a)
1065
1066 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1067 int getfrag(void *from, char *to, int offset,
1068 int len, int odd, struct sk_buff *skb),
1069 void *from, int length);
1070
1071 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1072 int offset, size_t size);
1073
1074 struct skb_seq_state {
1075 __u32 lower_offset;
1076 __u32 upper_offset;
1077 __u32 frag_idx;
1078 __u32 stepped_offset;
1079 struct sk_buff *root_skb;
1080 struct sk_buff *cur_skb;
1081 __u8 *frag_data;
1082 };
1083
1084 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1085 unsigned int to, struct skb_seq_state *st);
1086 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1087 struct skb_seq_state *st);
1088 void skb_abort_seq_read(struct skb_seq_state *st);
1089
1090 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1091 unsigned int to, struct ts_config *config);
1092
1093 /*
1094 * Packet hash types specify the type of hash in skb_set_hash.
1095 *
1096 * Hash types refer to the protocol layer addresses which are used to
1097 * construct a packet's hash. The hashes are used to differentiate or identify
1098 * flows of the protocol layer for the hash type. Hash types are either
1099 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1100 *
1101 * Properties of hashes:
1102 *
1103 * 1) Two packets in different flows have different hash values
1104 * 2) Two packets in the same flow should have the same hash value
1105 *
1106 * A hash at a higher layer is considered to be more specific. A driver should
1107 * set the most specific hash possible.
1108 *
1109 * A driver cannot indicate a more specific hash than the layer at which a hash
1110 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1111 *
1112 * A driver may indicate a hash level which is less specific than the
1113 * actual layer the hash was computed on. For instance, a hash computed
1114 * at L4 may be considered an L3 hash. This should only be done if the
1115 * driver can't unambiguously determine that the HW computed the hash at
1116 * the higher layer. Note that the "should" in the second property above
1117 * permits this.
1118 */
1119 enum pkt_hash_types {
1120 PKT_HASH_TYPE_NONE, /* Undefined type */
1121 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1122 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1123 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1124 };
1125
1126 static inline void skb_clear_hash(struct sk_buff *skb)
1127 {
1128 skb->hash = 0;
1129 skb->sw_hash = 0;
1130 skb->l4_hash = 0;
1131 }
1132
1133 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1134 {
1135 if (!skb->l4_hash)
1136 skb_clear_hash(skb);
1137 }
1138
1139 static inline void
1140 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1141 {
1142 skb->l4_hash = is_l4;
1143 skb->sw_hash = is_sw;
1144 skb->hash = hash;
1145 }
1146
1147 static inline void
1148 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1149 {
1150 /* Used by drivers to set hash from HW */
1151 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1152 }
1153
1154 static inline void
1155 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1156 {
1157 __skb_set_hash(skb, hash, true, is_l4);
1158 }
1159
1160 void __skb_get_hash(struct sk_buff *skb);
1161 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1162 u32 skb_get_poff(const struct sk_buff *skb);
1163 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1164 const struct flow_keys *keys, int hlen);
1165 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1166 void *data, int hlen_proto);
1167
1168 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1169 int thoff, u8 ip_proto)
1170 {
1171 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1172 }
1173
1174 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1175 const struct flow_dissector_key *key,
1176 unsigned int key_count);
1177
1178 bool __skb_flow_dissect(const struct sk_buff *skb,
1179 struct flow_dissector *flow_dissector,
1180 void *target_container,
1181 void *data, __be16 proto, int nhoff, int hlen,
1182 unsigned int flags);
1183
1184 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1185 struct flow_dissector *flow_dissector,
1186 void *target_container, unsigned int flags)
1187 {
1188 return __skb_flow_dissect(skb, flow_dissector, target_container,
1189 NULL, 0, 0, 0, flags);
1190 }
1191
1192 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1193 struct flow_keys *flow,
1194 unsigned int flags)
1195 {
1196 memset(flow, 0, sizeof(*flow));
1197 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
1198 NULL, 0, 0, 0, flags);
1199 }
1200
1201 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1202 void *data, __be16 proto,
1203 int nhoff, int hlen,
1204 unsigned int flags)
1205 {
1206 memset(flow, 0, sizeof(*flow));
1207 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
1208 data, proto, nhoff, hlen, flags);
1209 }
1210
1211 static inline __u32 skb_get_hash(struct sk_buff *skb)
1212 {
1213 if (!skb->l4_hash && !skb->sw_hash)
1214 __skb_get_hash(skb);
1215
1216 return skb->hash;
1217 }
1218
1219 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1220 {
1221 if (!skb->l4_hash && !skb->sw_hash) {
1222 struct flow_keys keys;
1223 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1224
1225 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1226 }
1227
1228 return skb->hash;
1229 }
1230
1231 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1232 const siphash_key_t *perturb);
1233
1234 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1235 {
1236 return skb->hash;
1237 }
1238
1239 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1240 {
1241 to->hash = from->hash;
1242 to->sw_hash = from->sw_hash;
1243 to->l4_hash = from->l4_hash;
1244 };
1245
1246 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1247 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1248 {
1249 return skb->head + skb->end;
1250 }
1251
1252 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1253 {
1254 return skb->end;
1255 }
1256 #else
1257 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1258 {
1259 return skb->end;
1260 }
1261
1262 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1263 {
1264 return skb->end - skb->head;
1265 }
1266 #endif
1267
1268 /* Internal */
1269 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1270
1271 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1272 {
1273 return &skb_shinfo(skb)->hwtstamps;
1274 }
1275
1276 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1277 {
1278 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1279
1280 return is_zcopy ? skb_uarg(skb) : NULL;
1281 }
1282
1283 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1284 {
1285 if (skb && uarg && !skb_zcopy(skb)) {
1286 sock_zerocopy_get(uarg);
1287 skb_shinfo(skb)->destructor_arg = uarg;
1288 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1289 }
1290 }
1291
1292 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1293 {
1294 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1295 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1296 }
1297
1298 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1299 {
1300 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1301 }
1302
1303 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1304 {
1305 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1306 }
1307
1308 /* Release a reference on a zerocopy structure */
1309 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1310 {
1311 struct ubuf_info *uarg = skb_zcopy(skb);
1312
1313 if (uarg) {
1314 if (uarg->callback == sock_zerocopy_callback) {
1315 uarg->zerocopy = uarg->zerocopy && zerocopy;
1316 sock_zerocopy_put(uarg);
1317 } else if (!skb_zcopy_is_nouarg(skb)) {
1318 uarg->callback(uarg, zerocopy);
1319 }
1320
1321 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1322 }
1323 }
1324
1325 /* Abort a zerocopy operation and revert zckey on error in send syscall */
1326 static inline void skb_zcopy_abort(struct sk_buff *skb)
1327 {
1328 struct ubuf_info *uarg = skb_zcopy(skb);
1329
1330 if (uarg) {
1331 sock_zerocopy_put_abort(uarg);
1332 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1333 }
1334 }
1335
1336 /**
1337 * skb_queue_empty - check if a queue is empty
1338 * @list: queue head
1339 *
1340 * Returns true if the queue is empty, false otherwise.
1341 */
1342 static inline int skb_queue_empty(const struct sk_buff_head *list)
1343 {
1344 return list->next == (const struct sk_buff *) list;
1345 }
1346
1347 /**
1348 * skb_queue_is_last - check if skb is the last entry in the queue
1349 * @list: queue head
1350 * @skb: buffer
1351 *
1352 * Returns true if @skb is the last buffer on the list.
1353 */
1354 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1355 const struct sk_buff *skb)
1356 {
1357 return skb->next == (const struct sk_buff *) list;
1358 }
1359
1360 /**
1361 * skb_queue_is_first - check if skb is the first entry in the queue
1362 * @list: queue head
1363 * @skb: buffer
1364 *
1365 * Returns true if @skb is the first buffer on the list.
1366 */
1367 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1368 const struct sk_buff *skb)
1369 {
1370 return skb->prev == (const struct sk_buff *) list;
1371 }
1372
1373 /**
1374 * skb_queue_next - return the next packet in the queue
1375 * @list: queue head
1376 * @skb: current buffer
1377 *
1378 * Return the next packet in @list after @skb. It is only valid to
1379 * call this if skb_queue_is_last() evaluates to false.
1380 */
1381 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1382 const struct sk_buff *skb)
1383 {
1384 /* This BUG_ON may seem severe, but if we just return then we
1385 * are going to dereference garbage.
1386 */
1387 BUG_ON(skb_queue_is_last(list, skb));
1388 return skb->next;
1389 }
1390
1391 /**
1392 * skb_queue_prev - return the prev packet in the queue
1393 * @list: queue head
1394 * @skb: current buffer
1395 *
1396 * Return the prev packet in @list before @skb. It is only valid to
1397 * call this if skb_queue_is_first() evaluates to false.
1398 */
1399 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1400 const struct sk_buff *skb)
1401 {
1402 /* This BUG_ON may seem severe, but if we just return then we
1403 * are going to dereference garbage.
1404 */
1405 BUG_ON(skb_queue_is_first(list, skb));
1406 return skb->prev;
1407 }
1408
1409 /**
1410 * skb_get - reference buffer
1411 * @skb: buffer to reference
1412 *
1413 * Makes another reference to a socket buffer and returns a pointer
1414 * to the buffer.
1415 */
1416 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1417 {
1418 refcount_inc(&skb->users);
1419 return skb;
1420 }
1421
1422 /*
1423 * If users == 1, we are the only owner and are can avoid redundant
1424 * atomic change.
1425 */
1426
1427 /**
1428 * skb_cloned - is the buffer a clone
1429 * @skb: buffer to check
1430 *
1431 * Returns true if the buffer was generated with skb_clone() and is
1432 * one of multiple shared copies of the buffer. Cloned buffers are
1433 * shared data so must not be written to under normal circumstances.
1434 */
1435 static inline int skb_cloned(const struct sk_buff *skb)
1436 {
1437 return skb->cloned &&
1438 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1439 }
1440
1441 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1442 {
1443 might_sleep_if(gfpflags_allow_blocking(pri));
1444
1445 if (skb_cloned(skb))
1446 return pskb_expand_head(skb, 0, 0, pri);
1447
1448 return 0;
1449 }
1450
1451 /**
1452 * skb_header_cloned - is the header a clone
1453 * @skb: buffer to check
1454 *
1455 * Returns true if modifying the header part of the buffer requires
1456 * the data to be copied.
1457 */
1458 static inline int skb_header_cloned(const struct sk_buff *skb)
1459 {
1460 int dataref;
1461
1462 if (!skb->cloned)
1463 return 0;
1464
1465 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1466 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1467 return dataref != 1;
1468 }
1469
1470 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1471 {
1472 might_sleep_if(gfpflags_allow_blocking(pri));
1473
1474 if (skb_header_cloned(skb))
1475 return pskb_expand_head(skb, 0, 0, pri);
1476
1477 return 0;
1478 }
1479
1480 /**
1481 * skb_header_release - release reference to header
1482 * @skb: buffer to operate on
1483 *
1484 * Drop a reference to the header part of the buffer. This is done
1485 * by acquiring a payload reference. You must not read from the header
1486 * part of skb->data after this.
1487 * Note : Check if you can use __skb_header_release() instead.
1488 */
1489 static inline void skb_header_release(struct sk_buff *skb)
1490 {
1491 BUG_ON(skb->nohdr);
1492 skb->nohdr = 1;
1493 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1494 }
1495
1496 /**
1497 * __skb_header_release - release reference to header
1498 * @skb: buffer to operate on
1499 *
1500 * Variant of skb_header_release() assuming skb is private to caller.
1501 * We can avoid one atomic operation.
1502 */
1503 static inline void __skb_header_release(struct sk_buff *skb)
1504 {
1505 skb->nohdr = 1;
1506 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1507 }
1508
1509
1510 /**
1511 * skb_shared - is the buffer shared
1512 * @skb: buffer to check
1513 *
1514 * Returns true if more than one person has a reference to this
1515 * buffer.
1516 */
1517 static inline int skb_shared(const struct sk_buff *skb)
1518 {
1519 return refcount_read(&skb->users) != 1;
1520 }
1521
1522 /**
1523 * skb_share_check - check if buffer is shared and if so clone it
1524 * @skb: buffer to check
1525 * @pri: priority for memory allocation
1526 *
1527 * If the buffer is shared the buffer is cloned and the old copy
1528 * drops a reference. A new clone with a single reference is returned.
1529 * If the buffer is not shared the original buffer is returned. When
1530 * being called from interrupt status or with spinlocks held pri must
1531 * be GFP_ATOMIC.
1532 *
1533 * NULL is returned on a memory allocation failure.
1534 */
1535 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1536 {
1537 might_sleep_if(gfpflags_allow_blocking(pri));
1538 if (skb_shared(skb)) {
1539 struct sk_buff *nskb = skb_clone(skb, pri);
1540
1541 if (likely(nskb))
1542 consume_skb(skb);
1543 else
1544 kfree_skb(skb);
1545 skb = nskb;
1546 }
1547 return skb;
1548 }
1549
1550 /*
1551 * Copy shared buffers into a new sk_buff. We effectively do COW on
1552 * packets to handle cases where we have a local reader and forward
1553 * and a couple of other messy ones. The normal one is tcpdumping
1554 * a packet thats being forwarded.
1555 */
1556
1557 /**
1558 * skb_unshare - make a copy of a shared buffer
1559 * @skb: buffer to check
1560 * @pri: priority for memory allocation
1561 *
1562 * If the socket buffer is a clone then this function creates a new
1563 * copy of the data, drops a reference count on the old copy and returns
1564 * the new copy with the reference count at 1. If the buffer is not a clone
1565 * the original buffer is returned. When called with a spinlock held or
1566 * from interrupt state @pri must be %GFP_ATOMIC
1567 *
1568 * %NULL is returned on a memory allocation failure.
1569 */
1570 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1571 gfp_t pri)
1572 {
1573 might_sleep_if(gfpflags_allow_blocking(pri));
1574 if (skb_cloned(skb)) {
1575 struct sk_buff *nskb = skb_copy(skb, pri);
1576
1577 /* Free our shared copy */
1578 if (likely(nskb))
1579 consume_skb(skb);
1580 else
1581 kfree_skb(skb);
1582 skb = nskb;
1583 }
1584 return skb;
1585 }
1586
1587 /**
1588 * skb_peek - peek at the head of an &sk_buff_head
1589 * @list_: list to peek at
1590 *
1591 * Peek an &sk_buff. Unlike most other operations you _MUST_
1592 * be careful with this one. A peek leaves the buffer on the
1593 * list and someone else may run off with it. You must hold
1594 * the appropriate locks or have a private queue to do this.
1595 *
1596 * Returns %NULL for an empty list or a pointer to the head element.
1597 * The reference count is not incremented and the reference is therefore
1598 * volatile. Use with caution.
1599 */
1600 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1601 {
1602 struct sk_buff *skb = list_->next;
1603
1604 if (skb == (struct sk_buff *)list_)
1605 skb = NULL;
1606 return skb;
1607 }
1608
1609 /**
1610 * skb_peek_next - peek skb following the given one from a queue
1611 * @skb: skb to start from
1612 * @list_: list to peek at
1613 *
1614 * Returns %NULL when the end of the list is met or a pointer to the
1615 * next element. The reference count is not incremented and the
1616 * reference is therefore volatile. Use with caution.
1617 */
1618 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1619 const struct sk_buff_head *list_)
1620 {
1621 struct sk_buff *next = skb->next;
1622
1623 if (next == (struct sk_buff *)list_)
1624 next = NULL;
1625 return next;
1626 }
1627
1628 /**
1629 * skb_peek_tail - peek at the tail of an &sk_buff_head
1630 * @list_: list to peek at
1631 *
1632 * Peek an &sk_buff. Unlike most other operations you _MUST_
1633 * be careful with this one. A peek leaves the buffer on the
1634 * list and someone else may run off with it. You must hold
1635 * the appropriate locks or have a private queue to do this.
1636 *
1637 * Returns %NULL for an empty list or a pointer to the tail element.
1638 * The reference count is not incremented and the reference is therefore
1639 * volatile. Use with caution.
1640 */
1641 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1642 {
1643 struct sk_buff *skb = list_->prev;
1644
1645 if (skb == (struct sk_buff *)list_)
1646 skb = NULL;
1647 return skb;
1648
1649 }
1650
1651 /**
1652 * skb_queue_len - get queue length
1653 * @list_: list to measure
1654 *
1655 * Return the length of an &sk_buff queue.
1656 */
1657 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1658 {
1659 return list_->qlen;
1660 }
1661
1662 /**
1663 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1664 * @list: queue to initialize
1665 *
1666 * This initializes only the list and queue length aspects of
1667 * an sk_buff_head object. This allows to initialize the list
1668 * aspects of an sk_buff_head without reinitializing things like
1669 * the spinlock. It can also be used for on-stack sk_buff_head
1670 * objects where the spinlock is known to not be used.
1671 */
1672 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1673 {
1674 list->prev = list->next = (struct sk_buff *)list;
1675 list->qlen = 0;
1676 }
1677
1678 /*
1679 * This function creates a split out lock class for each invocation;
1680 * this is needed for now since a whole lot of users of the skb-queue
1681 * infrastructure in drivers have different locking usage (in hardirq)
1682 * than the networking core (in softirq only). In the long run either the
1683 * network layer or drivers should need annotation to consolidate the
1684 * main types of usage into 3 classes.
1685 */
1686 static inline void skb_queue_head_init(struct sk_buff_head *list)
1687 {
1688 spin_lock_init(&list->lock);
1689 __skb_queue_head_init(list);
1690 }
1691
1692 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1693 struct lock_class_key *class)
1694 {
1695 skb_queue_head_init(list);
1696 lockdep_set_class(&list->lock, class);
1697 }
1698
1699 /*
1700 * Insert an sk_buff on a list.
1701 *
1702 * The "__skb_xxxx()" functions are the non-atomic ones that
1703 * can only be called with interrupts disabled.
1704 */
1705 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1706 struct sk_buff_head *list);
1707 static inline void __skb_insert(struct sk_buff *newsk,
1708 struct sk_buff *prev, struct sk_buff *next,
1709 struct sk_buff_head *list)
1710 {
1711 newsk->next = next;
1712 newsk->prev = prev;
1713 next->prev = prev->next = newsk;
1714 list->qlen++;
1715 }
1716
1717 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1718 struct sk_buff *prev,
1719 struct sk_buff *next)
1720 {
1721 struct sk_buff *first = list->next;
1722 struct sk_buff *last = list->prev;
1723
1724 first->prev = prev;
1725 prev->next = first;
1726
1727 last->next = next;
1728 next->prev = last;
1729 }
1730
1731 /**
1732 * skb_queue_splice - join two skb lists, this is designed for stacks
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(const struct sk_buff_head *list,
1737 struct sk_buff_head *head)
1738 {
1739 if (!skb_queue_empty(list)) {
1740 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1741 head->qlen += list->qlen;
1742 }
1743 }
1744
1745 /**
1746 * skb_queue_splice_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 * The list at @list is reinitialised
1751 */
1752 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1753 struct sk_buff_head *head)
1754 {
1755 if (!skb_queue_empty(list)) {
1756 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1757 head->qlen += list->qlen;
1758 __skb_queue_head_init(list);
1759 }
1760 }
1761
1762 /**
1763 * skb_queue_splice_tail - join two skb lists, each list being a queue
1764 * @list: the new list to add
1765 * @head: the place to add it in the first list
1766 */
1767 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1768 struct sk_buff_head *head)
1769 {
1770 if (!skb_queue_empty(list)) {
1771 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1772 head->qlen += list->qlen;
1773 }
1774 }
1775
1776 /**
1777 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1778 * @list: the new list to add
1779 * @head: the place to add it in the first list
1780 *
1781 * Each of the lists is a queue.
1782 * The list at @list is reinitialised
1783 */
1784 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1785 struct sk_buff_head *head)
1786 {
1787 if (!skb_queue_empty(list)) {
1788 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1789 head->qlen += list->qlen;
1790 __skb_queue_head_init(list);
1791 }
1792 }
1793
1794 /**
1795 * __skb_queue_after - queue a buffer at the list head
1796 * @list: list to use
1797 * @prev: place after this buffer
1798 * @newsk: buffer to queue
1799 *
1800 * Queue a buffer int the middle of a list. This function takes no locks
1801 * and you must therefore hold required locks before calling it.
1802 *
1803 * A buffer cannot be placed on two lists at the same time.
1804 */
1805 static inline void __skb_queue_after(struct sk_buff_head *list,
1806 struct sk_buff *prev,
1807 struct sk_buff *newsk)
1808 {
1809 __skb_insert(newsk, prev, prev->next, list);
1810 }
1811
1812 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1813 struct sk_buff_head *list);
1814
1815 static inline void __skb_queue_before(struct sk_buff_head *list,
1816 struct sk_buff *next,
1817 struct sk_buff *newsk)
1818 {
1819 __skb_insert(newsk, next->prev, next, list);
1820 }
1821
1822 /**
1823 * __skb_queue_head - queue a buffer at the list head
1824 * @list: list to use
1825 * @newsk: buffer to queue
1826 *
1827 * Queue a buffer at the start of a list. This function takes no locks
1828 * and you must therefore hold required locks before calling it.
1829 *
1830 * A buffer cannot be placed on two lists at the same time.
1831 */
1832 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1833 static inline void __skb_queue_head(struct sk_buff_head *list,
1834 struct sk_buff *newsk)
1835 {
1836 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1837 }
1838
1839 /**
1840 * __skb_queue_tail - queue a buffer at the list tail
1841 * @list: list to use
1842 * @newsk: buffer to queue
1843 *
1844 * Queue a buffer at the end of a list. This function takes no locks
1845 * and you must therefore hold required locks before calling it.
1846 *
1847 * A buffer cannot be placed on two lists at the same time.
1848 */
1849 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1850 static inline void __skb_queue_tail(struct sk_buff_head *list,
1851 struct sk_buff *newsk)
1852 {
1853 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1854 }
1855
1856 /*
1857 * remove sk_buff from list. _Must_ be called atomically, and with
1858 * the list known..
1859 */
1860 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1861 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1862 {
1863 struct sk_buff *next, *prev;
1864
1865 list->qlen--;
1866 next = skb->next;
1867 prev = skb->prev;
1868 skb->next = skb->prev = NULL;
1869 next->prev = prev;
1870 prev->next = next;
1871 }
1872
1873 /**
1874 * __skb_dequeue - remove from the head of the queue
1875 * @list: list to dequeue from
1876 *
1877 * Remove the head of the list. This function does not take any locks
1878 * so must be used with appropriate locks held only. The head item is
1879 * returned or %NULL if the list is empty.
1880 */
1881 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1882 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1883 {
1884 struct sk_buff *skb = skb_peek(list);
1885 if (skb)
1886 __skb_unlink(skb, list);
1887 return skb;
1888 }
1889
1890 /**
1891 * __skb_dequeue_tail - remove from the tail of the queue
1892 * @list: list to dequeue from
1893 *
1894 * Remove the tail of the list. This function does not take any locks
1895 * so must be used with appropriate locks held only. The tail item is
1896 * returned or %NULL if the list is empty.
1897 */
1898 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1899 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1900 {
1901 struct sk_buff *skb = skb_peek_tail(list);
1902 if (skb)
1903 __skb_unlink(skb, list);
1904 return skb;
1905 }
1906
1907
1908 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1909 {
1910 return skb->data_len;
1911 }
1912
1913 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1914 {
1915 return skb->len - skb->data_len;
1916 }
1917
1918 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1919 {
1920 unsigned int i, len = 0;
1921
1922 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
1923 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1924 return len;
1925 }
1926
1927 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1928 {
1929 return skb_headlen(skb) + __skb_pagelen(skb);
1930 }
1931
1932 /**
1933 * __skb_fill_page_desc - initialise a paged fragment in an skb
1934 * @skb: buffer containing fragment to be initialised
1935 * @i: paged fragment index to initialise
1936 * @page: the page to use for this fragment
1937 * @off: the offset to the data with @page
1938 * @size: the length of the data
1939 *
1940 * Initialises the @i'th fragment of @skb to point to &size bytes at
1941 * offset @off within @page.
1942 *
1943 * Does not take any additional reference on the fragment.
1944 */
1945 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1946 struct page *page, int off, int size)
1947 {
1948 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1949
1950 /*
1951 * Propagate page pfmemalloc to the skb if we can. The problem is
1952 * that not all callers have unique ownership of the page but rely
1953 * on page_is_pfmemalloc doing the right thing(tm).
1954 */
1955 frag->page.p = page;
1956 frag->page_offset = off;
1957 skb_frag_size_set(frag, size);
1958
1959 page = compound_head(page);
1960 if (page_is_pfmemalloc(page))
1961 skb->pfmemalloc = true;
1962 }
1963
1964 /**
1965 * skb_fill_page_desc - initialise a paged fragment in an skb
1966 * @skb: buffer containing fragment to be initialised
1967 * @i: paged fragment index to initialise
1968 * @page: the page to use for this fragment
1969 * @off: the offset to the data with @page
1970 * @size: the length of the data
1971 *
1972 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1973 * @skb to point to @size bytes at offset @off within @page. In
1974 * addition updates @skb such that @i is the last fragment.
1975 *
1976 * Does not take any additional reference on the fragment.
1977 */
1978 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1979 struct page *page, int off, int size)
1980 {
1981 __skb_fill_page_desc(skb, i, page, off, size);
1982 skb_shinfo(skb)->nr_frags = i + 1;
1983 }
1984
1985 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1986 int size, unsigned int truesize);
1987
1988 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1989 unsigned int truesize);
1990
1991 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1992 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1993 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1994
1995 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1996 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1997 {
1998 return skb->head + skb->tail;
1999 }
2000
2001 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2002 {
2003 skb->tail = skb->data - skb->head;
2004 }
2005
2006 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2007 {
2008 skb_reset_tail_pointer(skb);
2009 skb->tail += offset;
2010 }
2011
2012 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2013 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2014 {
2015 return skb->tail;
2016 }
2017
2018 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2019 {
2020 skb->tail = skb->data;
2021 }
2022
2023 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2024 {
2025 skb->tail = skb->data + offset;
2026 }
2027
2028 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2029
2030 /*
2031 * Add data to an sk_buff
2032 */
2033 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2034 void *skb_put(struct sk_buff *skb, unsigned int len);
2035 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2036 {
2037 void *tmp = skb_tail_pointer(skb);
2038 SKB_LINEAR_ASSERT(skb);
2039 skb->tail += len;
2040 skb->len += len;
2041 return tmp;
2042 }
2043
2044 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2045 {
2046 void *tmp = __skb_put(skb, len);
2047
2048 memset(tmp, 0, len);
2049 return tmp;
2050 }
2051
2052 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2053 unsigned int len)
2054 {
2055 void *tmp = __skb_put(skb, len);
2056
2057 memcpy(tmp, data, len);
2058 return tmp;
2059 }
2060
2061 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2062 {
2063 *(u8 *)__skb_put(skb, 1) = val;
2064 }
2065
2066 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2067 {
2068 void *tmp = skb_put(skb, len);
2069
2070 memset(tmp, 0, len);
2071
2072 return tmp;
2073 }
2074
2075 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2076 unsigned int len)
2077 {
2078 void *tmp = skb_put(skb, len);
2079
2080 memcpy(tmp, data, len);
2081
2082 return tmp;
2083 }
2084
2085 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2086 {
2087 *(u8 *)skb_put(skb, 1) = val;
2088 }
2089
2090 void *skb_push(struct sk_buff *skb, unsigned int len);
2091 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2092 {
2093 skb->data -= len;
2094 skb->len += len;
2095 return skb->data;
2096 }
2097
2098 void *skb_pull(struct sk_buff *skb, unsigned int len);
2099 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2100 {
2101 skb->len -= len;
2102 BUG_ON(skb->len < skb->data_len);
2103 return skb->data += len;
2104 }
2105
2106 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2107 {
2108 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2109 }
2110
2111 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2112
2113 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2114 {
2115 if (len > skb_headlen(skb) &&
2116 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2117 return NULL;
2118 skb->len -= len;
2119 return skb->data += len;
2120 }
2121
2122 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2123 {
2124 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2125 }
2126
2127 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2128 {
2129 if (likely(len <= skb_headlen(skb)))
2130 return 1;
2131 if (unlikely(len > skb->len))
2132 return 0;
2133 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2134 }
2135
2136 void skb_condense(struct sk_buff *skb);
2137
2138 /**
2139 * skb_headroom - bytes at buffer head
2140 * @skb: buffer to check
2141 *
2142 * Return the number of bytes of free space at the head of an &sk_buff.
2143 */
2144 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2145 {
2146 return skb->data - skb->head;
2147 }
2148
2149 /**
2150 * skb_tailroom - bytes at buffer end
2151 * @skb: buffer to check
2152 *
2153 * Return the number of bytes of free space at the tail of an sk_buff
2154 */
2155 static inline int skb_tailroom(const struct sk_buff *skb)
2156 {
2157 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2158 }
2159
2160 /**
2161 * skb_availroom - bytes at buffer end
2162 * @skb: buffer to check
2163 *
2164 * Return the number of bytes of free space at the tail of an sk_buff
2165 * allocated by sk_stream_alloc()
2166 */
2167 static inline int skb_availroom(const struct sk_buff *skb)
2168 {
2169 if (skb_is_nonlinear(skb))
2170 return 0;
2171
2172 return skb->end - skb->tail - skb->reserved_tailroom;
2173 }
2174
2175 /**
2176 * skb_reserve - adjust headroom
2177 * @skb: buffer to alter
2178 * @len: bytes to move
2179 *
2180 * Increase the headroom of an empty &sk_buff by reducing the tail
2181 * room. This is only allowed for an empty buffer.
2182 */
2183 static inline void skb_reserve(struct sk_buff *skb, int len)
2184 {
2185 skb->data += len;
2186 skb->tail += len;
2187 }
2188
2189 /**
2190 * skb_tailroom_reserve - adjust reserved_tailroom
2191 * @skb: buffer to alter
2192 * @mtu: maximum amount of headlen permitted
2193 * @needed_tailroom: minimum amount of reserved_tailroom
2194 *
2195 * Set reserved_tailroom so that headlen can be as large as possible but
2196 * not larger than mtu and tailroom cannot be smaller than
2197 * needed_tailroom.
2198 * The required headroom should already have been reserved before using
2199 * this function.
2200 */
2201 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2202 unsigned int needed_tailroom)
2203 {
2204 SKB_LINEAR_ASSERT(skb);
2205 if (mtu < skb_tailroom(skb) - needed_tailroom)
2206 /* use at most mtu */
2207 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2208 else
2209 /* use up to all available space */
2210 skb->reserved_tailroom = needed_tailroom;
2211 }
2212
2213 #define ENCAP_TYPE_ETHER 0
2214 #define ENCAP_TYPE_IPPROTO 1
2215
2216 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2217 __be16 protocol)
2218 {
2219 skb->inner_protocol = protocol;
2220 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2221 }
2222
2223 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2224 __u8 ipproto)
2225 {
2226 skb->inner_ipproto = ipproto;
2227 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2228 }
2229
2230 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2231 {
2232 skb->inner_mac_header = skb->mac_header;
2233 skb->inner_network_header = skb->network_header;
2234 skb->inner_transport_header = skb->transport_header;
2235 }
2236
2237 static inline void skb_reset_mac_len(struct sk_buff *skb)
2238 {
2239 skb->mac_len = skb->network_header - skb->mac_header;
2240 }
2241
2242 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2243 *skb)
2244 {
2245 return skb->head + skb->inner_transport_header;
2246 }
2247
2248 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2249 {
2250 return skb_inner_transport_header(skb) - skb->data;
2251 }
2252
2253 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2254 {
2255 skb->inner_transport_header = skb->data - skb->head;
2256 }
2257
2258 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2259 const int offset)
2260 {
2261 skb_reset_inner_transport_header(skb);
2262 skb->inner_transport_header += offset;
2263 }
2264
2265 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2266 {
2267 return skb->head + skb->inner_network_header;
2268 }
2269
2270 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2271 {
2272 skb->inner_network_header = skb->data - skb->head;
2273 }
2274
2275 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2276 const int offset)
2277 {
2278 skb_reset_inner_network_header(skb);
2279 skb->inner_network_header += offset;
2280 }
2281
2282 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2283 {
2284 return skb->head + skb->inner_mac_header;
2285 }
2286
2287 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2288 {
2289 skb->inner_mac_header = skb->data - skb->head;
2290 }
2291
2292 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2293 const int offset)
2294 {
2295 skb_reset_inner_mac_header(skb);
2296 skb->inner_mac_header += offset;
2297 }
2298 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2299 {
2300 return skb->transport_header != (typeof(skb->transport_header))~0U;
2301 }
2302
2303 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2304 {
2305 return skb->head + skb->transport_header;
2306 }
2307
2308 static inline void skb_reset_transport_header(struct sk_buff *skb)
2309 {
2310 skb->transport_header = skb->data - skb->head;
2311 }
2312
2313 static inline void skb_set_transport_header(struct sk_buff *skb,
2314 const int offset)
2315 {
2316 skb_reset_transport_header(skb);
2317 skb->transport_header += offset;
2318 }
2319
2320 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2321 {
2322 return skb->head + skb->network_header;
2323 }
2324
2325 static inline void skb_reset_network_header(struct sk_buff *skb)
2326 {
2327 skb->network_header = skb->data - skb->head;
2328 }
2329
2330 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2331 {
2332 skb_reset_network_header(skb);
2333 skb->network_header += offset;
2334 }
2335
2336 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2337 {
2338 return skb->head + skb->mac_header;
2339 }
2340
2341 static inline int skb_mac_offset(const struct sk_buff *skb)
2342 {
2343 return skb_mac_header(skb) - skb->data;
2344 }
2345
2346 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2347 {
2348 return skb->network_header - skb->mac_header;
2349 }
2350
2351 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2352 {
2353 return skb->mac_header != (typeof(skb->mac_header))~0U;
2354 }
2355
2356 static inline void skb_reset_mac_header(struct sk_buff *skb)
2357 {
2358 skb->mac_header = skb->data - skb->head;
2359 }
2360
2361 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2362 {
2363 skb_reset_mac_header(skb);
2364 skb->mac_header += offset;
2365 }
2366
2367 static inline void skb_pop_mac_header(struct sk_buff *skb)
2368 {
2369 skb->mac_header = skb->network_header;
2370 }
2371
2372 static inline void skb_probe_transport_header(struct sk_buff *skb,
2373 const int offset_hint)
2374 {
2375 struct flow_keys keys;
2376
2377 if (skb_transport_header_was_set(skb))
2378 return;
2379 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
2380 skb_set_transport_header(skb, keys.control.thoff);
2381 else if (offset_hint >= 0)
2382 skb_set_transport_header(skb, offset_hint);
2383 }
2384
2385 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2386 {
2387 if (skb_mac_header_was_set(skb)) {
2388 const unsigned char *old_mac = skb_mac_header(skb);
2389
2390 skb_set_mac_header(skb, -skb->mac_len);
2391 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2392 }
2393 }
2394
2395 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2396 {
2397 return skb->csum_start - skb_headroom(skb);
2398 }
2399
2400 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2401 {
2402 return skb->head + skb->csum_start;
2403 }
2404
2405 static inline int skb_transport_offset(const struct sk_buff *skb)
2406 {
2407 return skb_transport_header(skb) - skb->data;
2408 }
2409
2410 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2411 {
2412 return skb->transport_header - skb->network_header;
2413 }
2414
2415 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2416 {
2417 return skb->inner_transport_header - skb->inner_network_header;
2418 }
2419
2420 static inline int skb_network_offset(const struct sk_buff *skb)
2421 {
2422 return skb_network_header(skb) - skb->data;
2423 }
2424
2425 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2426 {
2427 return skb_inner_network_header(skb) - skb->data;
2428 }
2429
2430 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2431 {
2432 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2433 }
2434
2435 /*
2436 * CPUs often take a performance hit when accessing unaligned memory
2437 * locations. The actual performance hit varies, it can be small if the
2438 * hardware handles it or large if we have to take an exception and fix it
2439 * in software.
2440 *
2441 * Since an ethernet header is 14 bytes network drivers often end up with
2442 * the IP header at an unaligned offset. The IP header can be aligned by
2443 * shifting the start of the packet by 2 bytes. Drivers should do this
2444 * with:
2445 *
2446 * skb_reserve(skb, NET_IP_ALIGN);
2447 *
2448 * The downside to this alignment of the IP header is that the DMA is now
2449 * unaligned. On some architectures the cost of an unaligned DMA is high
2450 * and this cost outweighs the gains made by aligning the IP header.
2451 *
2452 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2453 * to be overridden.
2454 */
2455 #ifndef NET_IP_ALIGN
2456 #define NET_IP_ALIGN 2
2457 #endif
2458
2459 /*
2460 * The networking layer reserves some headroom in skb data (via
2461 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2462 * the header has to grow. In the default case, if the header has to grow
2463 * 32 bytes or less we avoid the reallocation.
2464 *
2465 * Unfortunately this headroom changes the DMA alignment of the resulting
2466 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2467 * on some architectures. An architecture can override this value,
2468 * perhaps setting it to a cacheline in size (since that will maintain
2469 * cacheline alignment of the DMA). It must be a power of 2.
2470 *
2471 * Various parts of the networking layer expect at least 32 bytes of
2472 * headroom, you should not reduce this.
2473 *
2474 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2475 * to reduce average number of cache lines per packet.
2476 * get_rps_cpus() for example only access one 64 bytes aligned block :
2477 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2478 */
2479 #ifndef NET_SKB_PAD
2480 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2481 #endif
2482
2483 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2484
2485 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2486 {
2487 if (unlikely(skb_is_nonlinear(skb))) {
2488 WARN_ON(1);
2489 return;
2490 }
2491 skb->len = len;
2492 skb_set_tail_pointer(skb, len);
2493 }
2494
2495 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2496 {
2497 __skb_set_length(skb, len);
2498 }
2499
2500 void skb_trim(struct sk_buff *skb, unsigned int len);
2501
2502 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2503 {
2504 if (skb->data_len)
2505 return ___pskb_trim(skb, len);
2506 __skb_trim(skb, len);
2507 return 0;
2508 }
2509
2510 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2511 {
2512 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2513 }
2514
2515 /**
2516 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2517 * @skb: buffer to alter
2518 * @len: new length
2519 *
2520 * This is identical to pskb_trim except that the caller knows that
2521 * the skb is not cloned so we should never get an error due to out-
2522 * of-memory.
2523 */
2524 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2525 {
2526 int err = pskb_trim(skb, len);
2527 BUG_ON(err);
2528 }
2529
2530 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2531 {
2532 unsigned int diff = len - skb->len;
2533
2534 if (skb_tailroom(skb) < diff) {
2535 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2536 GFP_ATOMIC);
2537 if (ret)
2538 return ret;
2539 }
2540 __skb_set_length(skb, len);
2541 return 0;
2542 }
2543
2544 /**
2545 * skb_orphan - orphan a buffer
2546 * @skb: buffer to orphan
2547 *
2548 * If a buffer currently has an owner then we call the owner's
2549 * destructor function and make the @skb unowned. The buffer continues
2550 * to exist but is no longer charged to its former owner.
2551 */
2552 static inline void skb_orphan(struct sk_buff *skb)
2553 {
2554 if (skb->destructor) {
2555 skb->destructor(skb);
2556 skb->destructor = NULL;
2557 skb->sk = NULL;
2558 } else {
2559 BUG_ON(skb->sk);
2560 }
2561 }
2562
2563 /**
2564 * skb_orphan_frags - orphan the frags contained in a buffer
2565 * @skb: buffer to orphan frags from
2566 * @gfp_mask: allocation mask for replacement pages
2567 *
2568 * For each frag in the SKB which needs a destructor (i.e. has an
2569 * owner) create a copy of that frag and release the original
2570 * page by calling the destructor.
2571 */
2572 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2573 {
2574 if (likely(!skb_zcopy(skb)))
2575 return 0;
2576 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2577 return 0;
2578 return skb_copy_ubufs(skb, gfp_mask);
2579 }
2580
2581 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2582 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2583 {
2584 if (likely(!skb_zcopy(skb)))
2585 return 0;
2586 return skb_copy_ubufs(skb, gfp_mask);
2587 }
2588
2589 /**
2590 * __skb_queue_purge - empty a list
2591 * @list: list to empty
2592 *
2593 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2594 * the list and one reference dropped. This function does not take the
2595 * list lock and the caller must hold the relevant locks to use it.
2596 */
2597 void skb_queue_purge(struct sk_buff_head *list);
2598 static inline void __skb_queue_purge(struct sk_buff_head *list)
2599 {
2600 struct sk_buff *skb;
2601 while ((skb = __skb_dequeue(list)) != NULL)
2602 kfree_skb(skb);
2603 }
2604
2605 unsigned int skb_rbtree_purge(struct rb_root *root);
2606
2607 void *netdev_alloc_frag(unsigned int fragsz);
2608
2609 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2610 gfp_t gfp_mask);
2611
2612 /**
2613 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2614 * @dev: network device to receive on
2615 * @length: length to allocate
2616 *
2617 * Allocate a new &sk_buff and assign it a usage count of one. The
2618 * buffer has unspecified headroom built in. Users should allocate
2619 * the headroom they think they need without accounting for the
2620 * built in space. The built in space is used for optimisations.
2621 *
2622 * %NULL is returned if there is no free memory. Although this function
2623 * allocates memory it can be called from an interrupt.
2624 */
2625 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2626 unsigned int length)
2627 {
2628 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2629 }
2630
2631 /* legacy helper around __netdev_alloc_skb() */
2632 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2633 gfp_t gfp_mask)
2634 {
2635 return __netdev_alloc_skb(NULL, length, gfp_mask);
2636 }
2637
2638 /* legacy helper around netdev_alloc_skb() */
2639 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2640 {
2641 return netdev_alloc_skb(NULL, length);
2642 }
2643
2644
2645 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2646 unsigned int length, gfp_t gfp)
2647 {
2648 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2649
2650 if (NET_IP_ALIGN && skb)
2651 skb_reserve(skb, NET_IP_ALIGN);
2652 return skb;
2653 }
2654
2655 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2656 unsigned int length)
2657 {
2658 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2659 }
2660
2661 static inline void skb_free_frag(void *addr)
2662 {
2663 page_frag_free(addr);
2664 }
2665
2666 void *napi_alloc_frag(unsigned int fragsz);
2667 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2668 unsigned int length, gfp_t gfp_mask);
2669 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2670 unsigned int length)
2671 {
2672 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2673 }
2674 void napi_consume_skb(struct sk_buff *skb, int budget);
2675
2676 void __kfree_skb_flush(void);
2677 void __kfree_skb_defer(struct sk_buff *skb);
2678
2679 /**
2680 * __dev_alloc_pages - allocate page for network Rx
2681 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2682 * @order: size of the allocation
2683 *
2684 * Allocate a new page.
2685 *
2686 * %NULL is returned if there is no free memory.
2687 */
2688 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2689 unsigned int order)
2690 {
2691 /* This piece of code contains several assumptions.
2692 * 1. This is for device Rx, therefor a cold page is preferred.
2693 * 2. The expectation is the user wants a compound page.
2694 * 3. If requesting a order 0 page it will not be compound
2695 * due to the check to see if order has a value in prep_new_page
2696 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2697 * code in gfp_to_alloc_flags that should be enforcing this.
2698 */
2699 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2700
2701 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2702 }
2703
2704 static inline struct page *dev_alloc_pages(unsigned int order)
2705 {
2706 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2707 }
2708
2709 /**
2710 * __dev_alloc_page - allocate a page for network Rx
2711 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2712 *
2713 * Allocate a new page.
2714 *
2715 * %NULL is returned if there is no free memory.
2716 */
2717 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2718 {
2719 return __dev_alloc_pages(gfp_mask, 0);
2720 }
2721
2722 static inline struct page *dev_alloc_page(void)
2723 {
2724 return dev_alloc_pages(0);
2725 }
2726
2727 /**
2728 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2729 * @page: The page that was allocated from skb_alloc_page
2730 * @skb: The skb that may need pfmemalloc set
2731 */
2732 static inline void skb_propagate_pfmemalloc(struct page *page,
2733 struct sk_buff *skb)
2734 {
2735 if (page_is_pfmemalloc(page))
2736 skb->pfmemalloc = true;
2737 }
2738
2739 /**
2740 * skb_frag_page - retrieve the page referred to by a paged fragment
2741 * @frag: the paged fragment
2742 *
2743 * Returns the &struct page associated with @frag.
2744 */
2745 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2746 {
2747 return frag->page.p;
2748 }
2749
2750 /**
2751 * __skb_frag_ref - take an addition reference on a paged fragment.
2752 * @frag: the paged fragment
2753 *
2754 * Takes an additional reference on the paged fragment @frag.
2755 */
2756 static inline void __skb_frag_ref(skb_frag_t *frag)
2757 {
2758 get_page(skb_frag_page(frag));
2759 }
2760
2761 /**
2762 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2763 * @skb: the buffer
2764 * @f: the fragment offset.
2765 *
2766 * Takes an additional reference on the @f'th paged fragment of @skb.
2767 */
2768 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2769 {
2770 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2771 }
2772
2773 /**
2774 * __skb_frag_unref - release a reference on a paged fragment.
2775 * @frag: the paged fragment
2776 *
2777 * Releases a reference on the paged fragment @frag.
2778 */
2779 static inline void __skb_frag_unref(skb_frag_t *frag)
2780 {
2781 put_page(skb_frag_page(frag));
2782 }
2783
2784 /**
2785 * skb_frag_unref - release a reference on a paged fragment of an skb.
2786 * @skb: the buffer
2787 * @f: the fragment offset
2788 *
2789 * Releases a reference on the @f'th paged fragment of @skb.
2790 */
2791 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2792 {
2793 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2794 }
2795
2796 /**
2797 * skb_frag_address - gets the address of the data contained in a paged fragment
2798 * @frag: the paged fragment buffer
2799 *
2800 * Returns the address of the data within @frag. The page must already
2801 * be mapped.
2802 */
2803 static inline void *skb_frag_address(const skb_frag_t *frag)
2804 {
2805 return page_address(skb_frag_page(frag)) + frag->page_offset;
2806 }
2807
2808 /**
2809 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2810 * @frag: the paged fragment buffer
2811 *
2812 * Returns the address of the data within @frag. Checks that the page
2813 * is mapped and returns %NULL otherwise.
2814 */
2815 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2816 {
2817 void *ptr = page_address(skb_frag_page(frag));
2818 if (unlikely(!ptr))
2819 return NULL;
2820
2821 return ptr + frag->page_offset;
2822 }
2823
2824 /**
2825 * __skb_frag_set_page - sets the page contained in a paged fragment
2826 * @frag: the paged fragment
2827 * @page: the page to set
2828 *
2829 * Sets the fragment @frag to contain @page.
2830 */
2831 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2832 {
2833 frag->page.p = page;
2834 }
2835
2836 /**
2837 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2838 * @skb: the buffer
2839 * @f: the fragment offset
2840 * @page: the page to set
2841 *
2842 * Sets the @f'th fragment of @skb to contain @page.
2843 */
2844 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2845 struct page *page)
2846 {
2847 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2848 }
2849
2850 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2851
2852 /**
2853 * skb_frag_dma_map - maps a paged fragment via the DMA API
2854 * @dev: the device to map the fragment to
2855 * @frag: the paged fragment to map
2856 * @offset: the offset within the fragment (starting at the
2857 * fragment's own offset)
2858 * @size: the number of bytes to map
2859 * @dir: the direction of the mapping (``PCI_DMA_*``)
2860 *
2861 * Maps the page associated with @frag to @device.
2862 */
2863 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2864 const skb_frag_t *frag,
2865 size_t offset, size_t size,
2866 enum dma_data_direction dir)
2867 {
2868 return dma_map_page(dev, skb_frag_page(frag),
2869 frag->page_offset + offset, size, dir);
2870 }
2871
2872 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2873 gfp_t gfp_mask)
2874 {
2875 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2876 }
2877
2878
2879 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2880 gfp_t gfp_mask)
2881 {
2882 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2883 }
2884
2885
2886 /**
2887 * skb_clone_writable - is the header of a clone writable
2888 * @skb: buffer to check
2889 * @len: length up to which to write
2890 *
2891 * Returns true if modifying the header part of the cloned buffer
2892 * does not requires the data to be copied.
2893 */
2894 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2895 {
2896 return !skb_header_cloned(skb) &&
2897 skb_headroom(skb) + len <= skb->hdr_len;
2898 }
2899
2900 static inline int skb_try_make_writable(struct sk_buff *skb,
2901 unsigned int write_len)
2902 {
2903 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2904 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2905 }
2906
2907 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2908 int cloned)
2909 {
2910 int delta = 0;
2911
2912 if (headroom > skb_headroom(skb))
2913 delta = headroom - skb_headroom(skb);
2914
2915 if (delta || cloned)
2916 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2917 GFP_ATOMIC);
2918 return 0;
2919 }
2920
2921 /**
2922 * skb_cow - copy header of skb when it is required
2923 * @skb: buffer to cow
2924 * @headroom: needed headroom
2925 *
2926 * If the skb passed lacks sufficient headroom or its data part
2927 * is shared, data is reallocated. If reallocation fails, an error
2928 * is returned and original skb is not changed.
2929 *
2930 * The result is skb with writable area skb->head...skb->tail
2931 * and at least @headroom of space at head.
2932 */
2933 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2934 {
2935 return __skb_cow(skb, headroom, skb_cloned(skb));
2936 }
2937
2938 /**
2939 * skb_cow_head - skb_cow but only making the head writable
2940 * @skb: buffer to cow
2941 * @headroom: needed headroom
2942 *
2943 * This function is identical to skb_cow except that we replace the
2944 * skb_cloned check by skb_header_cloned. It should be used when
2945 * you only need to push on some header and do not need to modify
2946 * the data.
2947 */
2948 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2949 {
2950 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2951 }
2952
2953 /**
2954 * skb_padto - pad an skbuff up to a minimal size
2955 * @skb: buffer to pad
2956 * @len: minimal length
2957 *
2958 * Pads up a buffer to ensure the trailing bytes exist and are
2959 * blanked. If the buffer already contains sufficient data it
2960 * is untouched. Otherwise it is extended. Returns zero on
2961 * success. The skb is freed on error.
2962 */
2963 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2964 {
2965 unsigned int size = skb->len;
2966 if (likely(size >= len))
2967 return 0;
2968 return skb_pad(skb, len - size);
2969 }
2970
2971 /**
2972 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2973 * @skb: buffer to pad
2974 * @len: minimal length
2975 * @free_on_error: free buffer on error
2976 *
2977 * Pads up a buffer to ensure the trailing bytes exist and are
2978 * blanked. If the buffer already contains sufficient data it
2979 * is untouched. Otherwise it is extended. Returns zero on
2980 * success. The skb is freed on error if @free_on_error is true.
2981 */
2982 static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2983 bool free_on_error)
2984 {
2985 unsigned int size = skb->len;
2986
2987 if (unlikely(size < len)) {
2988 len -= size;
2989 if (__skb_pad(skb, len, free_on_error))
2990 return -ENOMEM;
2991 __skb_put(skb, len);
2992 }
2993 return 0;
2994 }
2995
2996 /**
2997 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2998 * @skb: buffer to pad
2999 * @len: minimal length
3000 *
3001 * Pads up a buffer to ensure the trailing bytes exist and are
3002 * blanked. If the buffer already contains sufficient data it
3003 * is untouched. Otherwise it is extended. Returns zero on
3004 * success. The skb is freed on error.
3005 */
3006 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3007 {
3008 return __skb_put_padto(skb, len, true);
3009 }
3010
3011 static inline int skb_add_data(struct sk_buff *skb,
3012 struct iov_iter *from, int copy)
3013 {
3014 const int off = skb->len;
3015
3016 if (skb->ip_summed == CHECKSUM_NONE) {
3017 __wsum csum = 0;
3018 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3019 &csum, from)) {
3020 skb->csum = csum_block_add(skb->csum, csum, off);
3021 return 0;
3022 }
3023 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3024 return 0;
3025
3026 __skb_trim(skb, off);
3027 return -EFAULT;
3028 }
3029
3030 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3031 const struct page *page, int off)
3032 {
3033 if (skb_zcopy(skb))
3034 return false;
3035 if (i) {
3036 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
3037
3038 return page == skb_frag_page(frag) &&
3039 off == frag->page_offset + skb_frag_size(frag);
3040 }
3041 return false;
3042 }
3043
3044 static inline int __skb_linearize(struct sk_buff *skb)
3045 {
3046 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3047 }
3048
3049 /**
3050 * skb_linearize - convert paged skb to linear one
3051 * @skb: buffer to linarize
3052 *
3053 * If there is no free memory -ENOMEM is returned, otherwise zero
3054 * is returned and the old skb data released.
3055 */
3056 static inline int skb_linearize(struct sk_buff *skb)
3057 {
3058 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3059 }
3060
3061 /**
3062 * skb_has_shared_frag - can any frag be overwritten
3063 * @skb: buffer to test
3064 *
3065 * Return true if the skb has at least one frag that might be modified
3066 * by an external entity (as in vmsplice()/sendfile())
3067 */
3068 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3069 {
3070 return skb_is_nonlinear(skb) &&
3071 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3072 }
3073
3074 /**
3075 * skb_linearize_cow - make sure skb is linear and writable
3076 * @skb: buffer to process
3077 *
3078 * If there is no free memory -ENOMEM is returned, otherwise zero
3079 * is returned and the old skb data released.
3080 */
3081 static inline int skb_linearize_cow(struct sk_buff *skb)
3082 {
3083 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3084 __skb_linearize(skb) : 0;
3085 }
3086
3087 static __always_inline void
3088 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3089 unsigned int off)
3090 {
3091 if (skb->ip_summed == CHECKSUM_COMPLETE)
3092 skb->csum = csum_block_sub(skb->csum,
3093 csum_partial(start, len, 0), off);
3094 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3095 skb_checksum_start_offset(skb) < 0)
3096 skb->ip_summed = CHECKSUM_NONE;
3097 }
3098
3099 /**
3100 * skb_postpull_rcsum - update checksum for received skb after pull
3101 * @skb: buffer to update
3102 * @start: start of data before pull
3103 * @len: length of data pulled
3104 *
3105 * After doing a pull on a received packet, you need to call this to
3106 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3107 * CHECKSUM_NONE so that it can be recomputed from scratch.
3108 */
3109 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3110 const void *start, unsigned int len)
3111 {
3112 __skb_postpull_rcsum(skb, start, len, 0);
3113 }
3114
3115 static __always_inline void
3116 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3117 unsigned int off)
3118 {
3119 if (skb->ip_summed == CHECKSUM_COMPLETE)
3120 skb->csum = csum_block_add(skb->csum,
3121 csum_partial(start, len, 0), off);
3122 }
3123
3124 /**
3125 * skb_postpush_rcsum - update checksum for received skb after push
3126 * @skb: buffer to update
3127 * @start: start of data after push
3128 * @len: length of data pushed
3129 *
3130 * After doing a push on a received packet, you need to call this to
3131 * update the CHECKSUM_COMPLETE checksum.
3132 */
3133 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3134 const void *start, unsigned int len)
3135 {
3136 __skb_postpush_rcsum(skb, start, len, 0);
3137 }
3138
3139 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3140
3141 /**
3142 * skb_push_rcsum - push skb and update receive checksum
3143 * @skb: buffer to update
3144 * @len: length of data pulled
3145 *
3146 * This function performs an skb_push on the packet and updates
3147 * the CHECKSUM_COMPLETE checksum. It should be used on
3148 * receive path processing instead of skb_push unless you know
3149 * that the checksum difference is zero (e.g., a valid IP header)
3150 * or you are setting ip_summed to CHECKSUM_NONE.
3151 */
3152 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3153 {
3154 skb_push(skb, len);
3155 skb_postpush_rcsum(skb, skb->data, len);
3156 return skb->data;
3157 }
3158
3159 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3160 /**
3161 * pskb_trim_rcsum - trim received skb and update checksum
3162 * @skb: buffer to trim
3163 * @len: new length
3164 *
3165 * This is exactly the same as pskb_trim except that it ensures the
3166 * checksum of received packets are still valid after the operation.
3167 * It can change skb pointers.
3168 */
3169
3170 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3171 {
3172 if (likely(len >= skb->len))
3173 return 0;
3174 return pskb_trim_rcsum_slow(skb, len);
3175 }
3176
3177 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3178 {
3179 if (skb->ip_summed == CHECKSUM_COMPLETE)
3180 skb->ip_summed = CHECKSUM_NONE;
3181 __skb_trim(skb, len);
3182 return 0;
3183 }
3184
3185 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3186 {
3187 if (skb->ip_summed == CHECKSUM_COMPLETE)
3188 skb->ip_summed = CHECKSUM_NONE;
3189 return __skb_grow(skb, len);
3190 }
3191
3192 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3193
3194 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3195 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3196 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3197 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3198 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3199
3200 #define skb_queue_walk(queue, skb) \
3201 for (skb = (queue)->next; \
3202 skb != (struct sk_buff *)(queue); \
3203 skb = skb->next)
3204
3205 #define skb_queue_walk_safe(queue, skb, tmp) \
3206 for (skb = (queue)->next, tmp = skb->next; \
3207 skb != (struct sk_buff *)(queue); \
3208 skb = tmp, tmp = skb->next)
3209
3210 #define skb_queue_walk_from(queue, skb) \
3211 for (; skb != (struct sk_buff *)(queue); \
3212 skb = skb->next)
3213
3214 #define skb_rbtree_walk(skb, root) \
3215 for (skb = skb_rb_first(root); skb != NULL; \
3216 skb = skb_rb_next(skb))
3217
3218 #define skb_rbtree_walk_from(skb) \
3219 for (; skb != NULL; \
3220 skb = skb_rb_next(skb))
3221
3222 #define skb_rbtree_walk_from_safe(skb, tmp) \
3223 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3224 skb = tmp)
3225
3226 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3227 for (tmp = skb->next; \
3228 skb != (struct sk_buff *)(queue); \
3229 skb = tmp, tmp = skb->next)
3230
3231 #define skb_queue_reverse_walk(queue, skb) \
3232 for (skb = (queue)->prev; \
3233 skb != (struct sk_buff *)(queue); \
3234 skb = skb->prev)
3235
3236 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3237 for (skb = (queue)->prev, tmp = skb->prev; \
3238 skb != (struct sk_buff *)(queue); \
3239 skb = tmp, tmp = skb->prev)
3240
3241 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3242 for (tmp = skb->prev; \
3243 skb != (struct sk_buff *)(queue); \
3244 skb = tmp, tmp = skb->prev)
3245
3246 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3247 {
3248 return skb_shinfo(skb)->frag_list != NULL;
3249 }
3250
3251 static inline void skb_frag_list_init(struct sk_buff *skb)
3252 {
3253 skb_shinfo(skb)->frag_list = NULL;
3254 }
3255
3256 #define skb_walk_frags(skb, iter) \
3257 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3258
3259
3260 int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3261 const struct sk_buff *skb);
3262 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3263 struct sk_buff_head *queue,
3264 unsigned int flags,
3265 void (*destructor)(struct sock *sk,
3266 struct sk_buff *skb),
3267 int *peeked, int *off, int *err,
3268 struct sk_buff **last);
3269 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3270 void (*destructor)(struct sock *sk,
3271 struct sk_buff *skb),
3272 int *peeked, int *off, int *err,
3273 struct sk_buff **last);
3274 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3275 void (*destructor)(struct sock *sk,
3276 struct sk_buff *skb),
3277 int *peeked, int *off, int *err);
3278 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3279 int *err);
3280 unsigned int datagram_poll(struct file *file, struct socket *sock,
3281 struct poll_table_struct *wait);
3282 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3283 struct iov_iter *to, int size);
3284 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3285 struct msghdr *msg, int size)
3286 {
3287 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3288 }
3289 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3290 struct msghdr *msg);
3291 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3292 struct iov_iter *from, int len);
3293 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3294 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3295 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3296 static inline void skb_free_datagram_locked(struct sock *sk,
3297 struct sk_buff *skb)
3298 {
3299 __skb_free_datagram_locked(sk, skb, 0);
3300 }
3301 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3302 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3303 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3304 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3305 int len, __wsum csum);
3306 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3307 struct pipe_inode_info *pipe, unsigned int len,
3308 unsigned int flags);
3309 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3310 int len);
3311 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3312 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3313 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3314 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3315 int len, int hlen);
3316 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3317 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3318 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3319 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
3320 bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
3321 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3322 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3323 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3324 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3325 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3326 int skb_vlan_pop(struct sk_buff *skb);
3327 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3328 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3329 gfp_t gfp);
3330
3331 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3332 {
3333 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3334 }
3335
3336 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3337 {
3338 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3339 }
3340
3341 struct skb_checksum_ops {
3342 __wsum (*update)(const void *mem, int len, __wsum wsum);
3343 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3344 };
3345
3346 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3347
3348 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3349 __wsum csum, const struct skb_checksum_ops *ops);
3350 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3351 __wsum csum);
3352
3353 static inline void * __must_check
3354 __skb_header_pointer(const struct sk_buff *skb, int offset,
3355 int len, void *data, int hlen, void *buffer)
3356 {
3357 if (hlen - offset >= len)
3358 return data + offset;
3359
3360 if (!skb ||
3361 skb_copy_bits(skb, offset, buffer, len) < 0)
3362 return NULL;
3363
3364 return buffer;
3365 }
3366
3367 static inline void * __must_check
3368 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3369 {
3370 return __skb_header_pointer(skb, offset, len, skb->data,
3371 skb_headlen(skb), buffer);
3372 }
3373
3374 /**
3375 * skb_needs_linearize - check if we need to linearize a given skb
3376 * depending on the given device features.
3377 * @skb: socket buffer to check
3378 * @features: net device features
3379 *
3380 * Returns true if either:
3381 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3382 * 2. skb is fragmented and the device does not support SG.
3383 */
3384 static inline bool skb_needs_linearize(struct sk_buff *skb,
3385 netdev_features_t features)
3386 {
3387 return skb_is_nonlinear(skb) &&
3388 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3389 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3390 }
3391
3392 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3393 void *to,
3394 const unsigned int len)
3395 {
3396 memcpy(to, skb->data, len);
3397 }
3398
3399 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3400 const int offset, void *to,
3401 const unsigned int len)
3402 {
3403 memcpy(to, skb->data + offset, len);
3404 }
3405
3406 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3407 const void *from,
3408 const unsigned int len)
3409 {
3410 memcpy(skb->data, from, len);
3411 }
3412
3413 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3414 const int offset,
3415 const void *from,
3416 const unsigned int len)
3417 {
3418 memcpy(skb->data + offset, from, len);
3419 }
3420
3421 void skb_init(void);
3422
3423 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3424 {
3425 return skb->tstamp;
3426 }
3427
3428 /**
3429 * skb_get_timestamp - get timestamp from a skb
3430 * @skb: skb to get stamp from
3431 * @stamp: pointer to struct timeval to store stamp in
3432 *
3433 * Timestamps are stored in the skb as offsets to a base timestamp.
3434 * This function converts the offset back to a struct timeval and stores
3435 * it in stamp.
3436 */
3437 static inline void skb_get_timestamp(const struct sk_buff *skb,
3438 struct timeval *stamp)
3439 {
3440 *stamp = ktime_to_timeval(skb->tstamp);
3441 }
3442
3443 static inline void skb_get_timestampns(const struct sk_buff *skb,
3444 struct timespec *stamp)
3445 {
3446 *stamp = ktime_to_timespec(skb->tstamp);
3447 }
3448
3449 static inline void __net_timestamp(struct sk_buff *skb)
3450 {
3451 skb->tstamp = ktime_get_real();
3452 }
3453
3454 static inline ktime_t net_timedelta(ktime_t t)
3455 {
3456 return ktime_sub(ktime_get_real(), t);
3457 }
3458
3459 static inline ktime_t net_invalid_timestamp(void)
3460 {
3461 return 0;
3462 }
3463
3464 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3465
3466 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3467
3468 void skb_clone_tx_timestamp(struct sk_buff *skb);
3469 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3470
3471 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3472
3473 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3474 {
3475 }
3476
3477 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3478 {
3479 return false;
3480 }
3481
3482 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3483
3484 /**
3485 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3486 *
3487 * PHY drivers may accept clones of transmitted packets for
3488 * timestamping via their phy_driver.txtstamp method. These drivers
3489 * must call this function to return the skb back to the stack with a
3490 * timestamp.
3491 *
3492 * @skb: clone of the the original outgoing packet
3493 * @hwtstamps: hardware time stamps
3494 *
3495 */
3496 void skb_complete_tx_timestamp(struct sk_buff *skb,
3497 struct skb_shared_hwtstamps *hwtstamps);
3498
3499 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3500 struct skb_shared_hwtstamps *hwtstamps,
3501 struct sock *sk, int tstype);
3502
3503 /**
3504 * skb_tstamp_tx - queue clone of skb with send time stamps
3505 * @orig_skb: the original outgoing packet
3506 * @hwtstamps: hardware time stamps, may be NULL if not available
3507 *
3508 * If the skb has a socket associated, then this function clones the
3509 * skb (thus sharing the actual data and optional structures), stores
3510 * the optional hardware time stamping information (if non NULL) or
3511 * generates a software time stamp (otherwise), then queues the clone
3512 * to the error queue of the socket. Errors are silently ignored.
3513 */
3514 void skb_tstamp_tx(struct sk_buff *orig_skb,
3515 struct skb_shared_hwtstamps *hwtstamps);
3516
3517 /**
3518 * skb_tx_timestamp() - Driver hook for transmit timestamping
3519 *
3520 * Ethernet MAC Drivers should call this function in their hard_xmit()
3521 * function immediately before giving the sk_buff to the MAC hardware.
3522 *
3523 * Specifically, one should make absolutely sure that this function is
3524 * called before TX completion of this packet can trigger. Otherwise
3525 * the packet could potentially already be freed.
3526 *
3527 * @skb: A socket buffer.
3528 */
3529 static inline void skb_tx_timestamp(struct sk_buff *skb)
3530 {
3531 skb_clone_tx_timestamp(skb);
3532 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3533 skb_tstamp_tx(skb, NULL);
3534 }
3535
3536 /**
3537 * skb_complete_wifi_ack - deliver skb with wifi status
3538 *
3539 * @skb: the original outgoing packet
3540 * @acked: ack status
3541 *
3542 */
3543 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3544
3545 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3546 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3547
3548 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3549 {
3550 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3551 skb->csum_valid ||
3552 (skb->ip_summed == CHECKSUM_PARTIAL &&
3553 skb_checksum_start_offset(skb) >= 0));
3554 }
3555
3556 /**
3557 * skb_checksum_complete - Calculate checksum of an entire packet
3558 * @skb: packet to process
3559 *
3560 * This function calculates the checksum over the entire packet plus
3561 * the value of skb->csum. The latter can be used to supply the
3562 * checksum of a pseudo header as used by TCP/UDP. It returns the
3563 * checksum.
3564 *
3565 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3566 * this function can be used to verify that checksum on received
3567 * packets. In that case the function should return zero if the
3568 * checksum is correct. In particular, this function will return zero
3569 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3570 * hardware has already verified the correctness of the checksum.
3571 */
3572 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3573 {
3574 return skb_csum_unnecessary(skb) ?
3575 0 : __skb_checksum_complete(skb);
3576 }
3577
3578 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3579 {
3580 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3581 if (skb->csum_level == 0)
3582 skb->ip_summed = CHECKSUM_NONE;
3583 else
3584 skb->csum_level--;
3585 }
3586 }
3587
3588 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3589 {
3590 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3591 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3592 skb->csum_level++;
3593 } else if (skb->ip_summed == CHECKSUM_NONE) {
3594 skb->ip_summed = CHECKSUM_UNNECESSARY;
3595 skb->csum_level = 0;
3596 }
3597 }
3598
3599 /* Check if we need to perform checksum complete validation.
3600 *
3601 * Returns true if checksum complete is needed, false otherwise
3602 * (either checksum is unnecessary or zero checksum is allowed).
3603 */
3604 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3605 bool zero_okay,
3606 __sum16 check)
3607 {
3608 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3609 skb->csum_valid = 1;
3610 __skb_decr_checksum_unnecessary(skb);
3611 return false;
3612 }
3613
3614 return true;
3615 }
3616
3617 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3618 * in checksum_init.
3619 */
3620 #define CHECKSUM_BREAK 76
3621
3622 /* Unset checksum-complete
3623 *
3624 * Unset checksum complete can be done when packet is being modified
3625 * (uncompressed for instance) and checksum-complete value is
3626 * invalidated.
3627 */
3628 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3629 {
3630 if (skb->ip_summed == CHECKSUM_COMPLETE)
3631 skb->ip_summed = CHECKSUM_NONE;
3632 }
3633
3634 /* Validate (init) checksum based on checksum complete.
3635 *
3636 * Return values:
3637 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3638 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3639 * checksum is stored in skb->csum for use in __skb_checksum_complete
3640 * non-zero: value of invalid checksum
3641 *
3642 */
3643 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3644 bool complete,
3645 __wsum psum)
3646 {
3647 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3648 if (!csum_fold(csum_add(psum, skb->csum))) {
3649 skb->csum_valid = 1;
3650 return 0;
3651 }
3652 }
3653
3654 skb->csum = psum;
3655
3656 if (complete || skb->len <= CHECKSUM_BREAK) {
3657 __sum16 csum;
3658
3659 csum = __skb_checksum_complete(skb);
3660 skb->csum_valid = !csum;
3661 return csum;
3662 }
3663
3664 return 0;
3665 }
3666
3667 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3668 {
3669 return 0;
3670 }
3671
3672 /* Perform checksum validate (init). Note that this is a macro since we only
3673 * want to calculate the pseudo header which is an input function if necessary.
3674 * First we try to validate without any computation (checksum unnecessary) and
3675 * then calculate based on checksum complete calling the function to compute
3676 * pseudo header.
3677 *
3678 * Return values:
3679 * 0: checksum is validated or try to in skb_checksum_complete
3680 * non-zero: value of invalid checksum
3681 */
3682 #define __skb_checksum_validate(skb, proto, complete, \
3683 zero_okay, check, compute_pseudo) \
3684 ({ \
3685 __sum16 __ret = 0; \
3686 skb->csum_valid = 0; \
3687 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3688 __ret = __skb_checksum_validate_complete(skb, \
3689 complete, compute_pseudo(skb, proto)); \
3690 __ret; \
3691 })
3692
3693 #define skb_checksum_init(skb, proto, compute_pseudo) \
3694 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3695
3696 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3697 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3698
3699 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3700 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3701
3702 #define skb_checksum_validate_zero_check(skb, proto, check, \
3703 compute_pseudo) \
3704 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3705
3706 #define skb_checksum_simple_validate(skb) \
3707 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3708
3709 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3710 {
3711 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
3712 }
3713
3714 static inline void __skb_checksum_convert(struct sk_buff *skb,
3715 __sum16 check, __wsum pseudo)
3716 {
3717 skb->csum = ~pseudo;
3718 skb->ip_summed = CHECKSUM_COMPLETE;
3719 }
3720
3721 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3722 do { \
3723 if (__skb_checksum_convert_check(skb)) \
3724 __skb_checksum_convert(skb, check, \
3725 compute_pseudo(skb, proto)); \
3726 } while (0)
3727
3728 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3729 u16 start, u16 offset)
3730 {
3731 skb->ip_summed = CHECKSUM_PARTIAL;
3732 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3733 skb->csum_offset = offset - start;
3734 }
3735
3736 /* Update skbuf and packet to reflect the remote checksum offload operation.
3737 * When called, ptr indicates the starting point for skb->csum when
3738 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3739 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3740 */
3741 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3742 int start, int offset, bool nopartial)
3743 {
3744 __wsum delta;
3745
3746 if (!nopartial) {
3747 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3748 return;
3749 }
3750
3751 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3752 __skb_checksum_complete(skb);
3753 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3754 }
3755
3756 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3757
3758 /* Adjust skb->csum since we changed the packet */
3759 skb->csum = csum_add(skb->csum, delta);
3760 }
3761
3762 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3763 {
3764 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3765 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
3766 #else
3767 return NULL;
3768 #endif
3769 }
3770
3771 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3772 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3773 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3774 {
3775 if (nfct && atomic_dec_and_test(&nfct->use))
3776 nf_conntrack_destroy(nfct);
3777 }
3778 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3779 {
3780 if (nfct)
3781 atomic_inc(&nfct->use);
3782 }
3783 #endif
3784 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3785 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3786 {
3787 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
3788 kfree(nf_bridge);
3789 }
3790 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3791 {
3792 if (nf_bridge)
3793 refcount_inc(&nf_bridge->use);
3794 }
3795 #endif /* CONFIG_BRIDGE_NETFILTER */
3796 static inline void nf_reset(struct sk_buff *skb)
3797 {
3798 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3799 nf_conntrack_put(skb_nfct(skb));
3800 skb->_nfct = 0;
3801 #endif
3802 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3803 nf_bridge_put(skb->nf_bridge);
3804 skb->nf_bridge = NULL;
3805 #endif
3806 }
3807
3808 static inline void nf_reset_trace(struct sk_buff *skb)
3809 {
3810 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3811 skb->nf_trace = 0;
3812 #endif
3813 }
3814
3815 static inline void ipvs_reset(struct sk_buff *skb)
3816 {
3817 #if IS_ENABLED(CONFIG_IP_VS)
3818 skb->ipvs_property = 0;
3819 #endif
3820 }
3821
3822 /* Note: This doesn't put any conntrack and bridge info in dst. */
3823 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3824 bool copy)
3825 {
3826 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3827 dst->_nfct = src->_nfct;
3828 nf_conntrack_get(skb_nfct(src));
3829 #endif
3830 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3831 dst->nf_bridge = src->nf_bridge;
3832 nf_bridge_get(src->nf_bridge);
3833 #endif
3834 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3835 if (copy)
3836 dst->nf_trace = src->nf_trace;
3837 #endif
3838 }
3839
3840 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3841 {
3842 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3843 nf_conntrack_put(skb_nfct(dst));
3844 #endif
3845 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3846 nf_bridge_put(dst->nf_bridge);
3847 #endif
3848 __nf_copy(dst, src, true);
3849 }
3850
3851 #ifdef CONFIG_NETWORK_SECMARK
3852 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3853 {
3854 to->secmark = from->secmark;
3855 }
3856
3857 static inline void skb_init_secmark(struct sk_buff *skb)
3858 {
3859 skb->secmark = 0;
3860 }
3861 #else
3862 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3863 { }
3864
3865 static inline void skb_init_secmark(struct sk_buff *skb)
3866 { }
3867 #endif
3868
3869 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3870 {
3871 return !skb->destructor &&
3872 #if IS_ENABLED(CONFIG_XFRM)
3873 !skb->sp &&
3874 #endif
3875 !skb_nfct(skb) &&
3876 !skb->_skb_refdst &&
3877 !skb_has_frag_list(skb);
3878 }
3879
3880 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3881 {
3882 skb->queue_mapping = queue_mapping;
3883 }
3884
3885 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3886 {
3887 return skb->queue_mapping;
3888 }
3889
3890 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3891 {
3892 to->queue_mapping = from->queue_mapping;
3893 }
3894
3895 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3896 {
3897 skb->queue_mapping = rx_queue + 1;
3898 }
3899
3900 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3901 {
3902 return skb->queue_mapping - 1;
3903 }
3904
3905 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3906 {
3907 return skb->queue_mapping != 0;
3908 }
3909
3910 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3911 {
3912 skb->dst_pending_confirm = val;
3913 }
3914
3915 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3916 {
3917 return skb->dst_pending_confirm != 0;
3918 }
3919
3920 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3921 {
3922 #ifdef CONFIG_XFRM
3923 return skb->sp;
3924 #else
3925 return NULL;
3926 #endif
3927 }
3928
3929 /* Keeps track of mac header offset relative to skb->head.
3930 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3931 * For non-tunnel skb it points to skb_mac_header() and for
3932 * tunnel skb it points to outer mac header.
3933 * Keeps track of level of encapsulation of network headers.
3934 */
3935 struct skb_gso_cb {
3936 union {
3937 int mac_offset;
3938 int data_offset;
3939 };
3940 int encap_level;
3941 __wsum csum;
3942 __u16 csum_start;
3943 };
3944 #define SKB_SGO_CB_OFFSET 32
3945 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
3946
3947 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3948 {
3949 return (skb_mac_header(inner_skb) - inner_skb->head) -
3950 SKB_GSO_CB(inner_skb)->mac_offset;
3951 }
3952
3953 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3954 {
3955 int new_headroom, headroom;
3956 int ret;
3957
3958 headroom = skb_headroom(skb);
3959 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3960 if (ret)
3961 return ret;
3962
3963 new_headroom = skb_headroom(skb);
3964 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3965 return 0;
3966 }
3967
3968 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3969 {
3970 /* Do not update partial checksums if remote checksum is enabled. */
3971 if (skb->remcsum_offload)
3972 return;
3973
3974 SKB_GSO_CB(skb)->csum = res;
3975 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3976 }
3977
3978 /* Compute the checksum for a gso segment. First compute the checksum value
3979 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3980 * then add in skb->csum (checksum from csum_start to end of packet).
3981 * skb->csum and csum_start are then updated to reflect the checksum of the
3982 * resultant packet starting from the transport header-- the resultant checksum
3983 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3984 * header.
3985 */
3986 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3987 {
3988 unsigned char *csum_start = skb_transport_header(skb);
3989 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3990 __wsum partial = SKB_GSO_CB(skb)->csum;
3991
3992 SKB_GSO_CB(skb)->csum = res;
3993 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
3994
3995 return csum_fold(csum_partial(csum_start, plen, partial));
3996 }
3997
3998 static inline bool skb_is_gso(const struct sk_buff *skb)
3999 {
4000 return skb_shinfo(skb)->gso_size;
4001 }
4002
4003 /* Note: Should be called only if skb_is_gso(skb) is true */
4004 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4005 {
4006 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4007 }
4008
4009 static inline void skb_gso_reset(struct sk_buff *skb)
4010 {
4011 skb_shinfo(skb)->gso_size = 0;
4012 skb_shinfo(skb)->gso_segs = 0;
4013 skb_shinfo(skb)->gso_type = 0;
4014 }
4015
4016 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4017
4018 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4019 {
4020 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4021 * wanted then gso_type will be set. */
4022 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4023
4024 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4025 unlikely(shinfo->gso_type == 0)) {
4026 __skb_warn_lro_forwarding(skb);
4027 return true;
4028 }
4029 return false;
4030 }
4031
4032 static inline void skb_forward_csum(struct sk_buff *skb)
4033 {
4034 /* Unfortunately we don't support this one. Any brave souls? */
4035 if (skb->ip_summed == CHECKSUM_COMPLETE)
4036 skb->ip_summed = CHECKSUM_NONE;
4037 }
4038
4039 /**
4040 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4041 * @skb: skb to check
4042 *
4043 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4044 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4045 * use this helper, to document places where we make this assertion.
4046 */
4047 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4048 {
4049 #ifdef DEBUG
4050 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4051 #endif
4052 }
4053
4054 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4055
4056 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4057 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4058 unsigned int transport_len,
4059 __sum16(*skb_chkf)(struct sk_buff *skb));
4060
4061 /**
4062 * skb_head_is_locked - Determine if the skb->head is locked down
4063 * @skb: skb to check
4064 *
4065 * The head on skbs build around a head frag can be removed if they are
4066 * not cloned. This function returns true if the skb head is locked down
4067 * due to either being allocated via kmalloc, or by being a clone with
4068 * multiple references to the head.
4069 */
4070 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4071 {
4072 return !skb->head_frag || skb_cloned(skb);
4073 }
4074
4075 /**
4076 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4077 *
4078 * @skb: GSO skb
4079 *
4080 * skb_gso_network_seglen is used to determine the real size of the
4081 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4082 *
4083 * The MAC/L2 header is not accounted for.
4084 */
4085 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4086 {
4087 unsigned int hdr_len = skb_transport_header(skb) -
4088 skb_network_header(skb);
4089 return hdr_len + skb_gso_transport_seglen(skb);
4090 }
4091
4092 /**
4093 * skb_gso_mac_seglen - Return length of individual segments of a gso packet
4094 *
4095 * @skb: GSO skb
4096 *
4097 * skb_gso_mac_seglen is used to determine the real size of the
4098 * individual segments, including MAC/L2, Layer3 (IP, IPv6) and L4
4099 * headers (TCP/UDP).
4100 */
4101 static inline unsigned int skb_gso_mac_seglen(const struct sk_buff *skb)
4102 {
4103 unsigned int hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
4104 return hdr_len + skb_gso_transport_seglen(skb);
4105 }
4106
4107 /* Local Checksum Offload.
4108 * Compute outer checksum based on the assumption that the
4109 * inner checksum will be offloaded later.
4110 * See Documentation/networking/checksum-offloads.txt for
4111 * explanation of how this works.
4112 * Fill in outer checksum adjustment (e.g. with sum of outer
4113 * pseudo-header) before calling.
4114 * Also ensure that inner checksum is in linear data area.
4115 */
4116 static inline __wsum lco_csum(struct sk_buff *skb)
4117 {
4118 unsigned char *csum_start = skb_checksum_start(skb);
4119 unsigned char *l4_hdr = skb_transport_header(skb);
4120 __wsum partial;
4121
4122 /* Start with complement of inner checksum adjustment */
4123 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4124 skb->csum_offset));
4125
4126 /* Add in checksum of our headers (incl. outer checksum
4127 * adjustment filled in by caller) and return result.
4128 */
4129 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4130 }
4131
4132 #endif /* __KERNEL__ */
4133 #endif /* _LINUX_SKBUFF_H */