]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/skbuff.h
net: Fix usage of pskb_trim_rcsum
[mirror_ubuntu-bionic-kernel.git] / include / linux / skbuff.h
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
16
17 #include <linux/kernel.h>
18 #include <linux/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 __u8 __unused;
492 __u8 meta_len;
493 __u8 nr_frags;
494 __u8 tx_flags;
495 unsigned short gso_size;
496 /* Warning: this field is not always filled in (UFO)! */
497 unsigned short gso_segs;
498 struct sk_buff *frag_list;
499 struct skb_shared_hwtstamps hwtstamps;
500 unsigned int gso_type;
501 u32 tskey;
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 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
621 * @_nfct: Associated connection, if any (with nfctinfo bits)
622 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
623 * @skb_iif: ifindex of device we arrived on
624 * @tc_index: Traffic control index
625 * @hash: the packet hash
626 * @queue_mapping: Queue mapping for multiqueue devices
627 * @xmit_more: More SKBs are pending for this queue
628 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
629 * @ndisc_nodetype: router type (from link layer)
630 * @ooo_okay: allow the mapping of a socket to a queue to be changed
631 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
632 * ports.
633 * @sw_hash: indicates hash was computed in software stack
634 * @wifi_acked_valid: wifi_acked was set
635 * @wifi_acked: whether frame was acked on wifi or not
636 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
637 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
638 * @dst_pending_confirm: need to confirm neighbour
639 * @napi_id: id of the NAPI struct this skb came from
640 * @secmark: security marking
641 * @mark: Generic packet mark
642 * @vlan_proto: vlan encapsulation protocol
643 * @vlan_tci: vlan tag control information
644 * @inner_protocol: Protocol (encapsulation)
645 * @inner_transport_header: Inner transport layer header (encapsulation)
646 * @inner_network_header: Network layer header (encapsulation)
647 * @inner_mac_header: Link layer header (encapsulation)
648 * @transport_header: Transport layer header
649 * @network_header: Network layer header
650 * @mac_header: Link layer header
651 * @tail: Tail pointer
652 * @end: End pointer
653 * @head: Head of buffer
654 * @data: Data head pointer
655 * @truesize: Buffer size
656 * @users: User count - see {datagram,tcp}.c
657 */
658
659 struct sk_buff {
660 union {
661 struct {
662 /* These two members must be first. */
663 struct sk_buff *next;
664 struct sk_buff *prev;
665
666 union {
667 struct net_device *dev;
668 /* Some protocols might use this space to store information,
669 * while device pointer would be NULL.
670 * UDP receive path is one user.
671 */
672 unsigned long dev_scratch;
673 int ip_defrag_offset;
674 };
675 };
676 struct rb_node rbnode; /* used in netem & tcp stack */
677 };
678 struct sock *sk;
679
680 union {
681 ktime_t tstamp;
682 u64 skb_mstamp;
683 };
684 /*
685 * This is the control buffer. It is free to use for every
686 * layer. Please put your private variables there. If you
687 * want to keep them across layers you have to do a skb_clone()
688 * first. This is owned by whoever has the skb queued ATM.
689 */
690 char cb[48] __aligned(8);
691
692 union {
693 struct {
694 unsigned long _skb_refdst;
695 void (*destructor)(struct sk_buff *skb);
696 };
697 struct list_head tcp_tsorted_anchor;
698 };
699
700 #ifdef CONFIG_XFRM
701 struct sec_path *sp;
702 #endif
703 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
704 unsigned long _nfct;
705 #endif
706 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
707 struct nf_bridge_info *nf_bridge;
708 #endif
709 unsigned int len,
710 data_len;
711 __u16 mac_len,
712 hdr_len;
713
714 /* Following fields are _not_ copied in __copy_skb_header()
715 * Note that queue_mapping is here mostly to fill a hole.
716 */
717 __u16 queue_mapping;
718
719 /* if you move cloned around you also must adapt those constants */
720 #ifdef __BIG_ENDIAN_BITFIELD
721 #define CLONED_MASK (1 << 7)
722 #else
723 #define CLONED_MASK 1
724 #endif
725 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
726
727 __u8 __cloned_offset[0];
728 __u8 cloned:1,
729 nohdr:1,
730 fclone:2,
731 peeked:1,
732 head_frag:1,
733 xmit_more:1,
734 pfmemalloc:1;
735
736 /* fields enclosed in headers_start/headers_end are copied
737 * using a single memcpy() in __copy_skb_header()
738 */
739 /* private: */
740 __u32 headers_start[0];
741 /* public: */
742
743 /* if you move pkt_type around you also must adapt those constants */
744 #ifdef __BIG_ENDIAN_BITFIELD
745 #define PKT_TYPE_MAX (7 << 5)
746 #else
747 #define PKT_TYPE_MAX 7
748 #endif
749 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
750
751 __u8 __pkt_type_offset[0];
752 __u8 pkt_type:3;
753 __u8 ignore_df:1;
754 __u8 nf_trace:1;
755 __u8 ip_summed:2;
756 __u8 ooo_okay:1;
757
758 __u8 l4_hash:1;
759 __u8 sw_hash:1;
760 __u8 wifi_acked_valid:1;
761 __u8 wifi_acked:1;
762 __u8 no_fcs:1;
763 /* Indicates the inner headers are valid in the skbuff. */
764 __u8 encapsulation:1;
765 __u8 encap_hdr_csum:1;
766 __u8 csum_valid:1;
767
768 __u8 csum_complete_sw:1;
769 __u8 csum_level:2;
770 __u8 csum_not_inet:1;
771 __u8 dst_pending_confirm:1;
772 #ifdef CONFIG_IPV6_NDISC_NODETYPE
773 __u8 ndisc_nodetype:2;
774 #endif
775 __u8 ipvs_property:1;
776
777 __u8 inner_protocol_type:1;
778 __u8 remcsum_offload:1;
779 #ifdef CONFIG_NET_SWITCHDEV
780 __u8 offload_fwd_mark:1;
781 __u8 offload_mr_fwd_mark:1;
782 #endif
783 #ifdef CONFIG_NET_CLS_ACT
784 __u8 tc_skip_classify:1;
785 __u8 tc_at_ingress:1;
786 __u8 tc_redirected:1;
787 __u8 tc_from_ingress:1;
788 #endif
789
790 #ifdef CONFIG_NET_SCHED
791 __u16 tc_index; /* traffic control index */
792 #endif
793
794 union {
795 __wsum csum;
796 struct {
797 __u16 csum_start;
798 __u16 csum_offset;
799 };
800 };
801 __u32 priority;
802 int skb_iif;
803 __u32 hash;
804 __be16 vlan_proto;
805 __u16 vlan_tci;
806 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
807 union {
808 unsigned int napi_id;
809 unsigned int sender_cpu;
810 };
811 #endif
812 #ifdef CONFIG_NETWORK_SECMARK
813 __u32 secmark;
814 #endif
815
816 union {
817 __u32 mark;
818 __u32 reserved_tailroom;
819 };
820
821 union {
822 __be16 inner_protocol;
823 __u8 inner_ipproto;
824 };
825
826 __u16 inner_transport_header;
827 __u16 inner_network_header;
828 __u16 inner_mac_header;
829
830 __be16 protocol;
831 __u16 transport_header;
832 __u16 network_header;
833 __u16 mac_header;
834
835 /* private: */
836 __u32 headers_end[0];
837 /* public: */
838
839 /* These elements must be at the end, see alloc_skb() for details. */
840 sk_buff_data_t tail;
841 sk_buff_data_t end;
842 unsigned char *head,
843 *data;
844 unsigned int truesize;
845 refcount_t users;
846 };
847
848 #ifdef __KERNEL__
849 /*
850 * Handling routines are only of interest to the kernel
851 */
852 #include <linux/slab.h>
853
854
855 #define SKB_ALLOC_FCLONE 0x01
856 #define SKB_ALLOC_RX 0x02
857 #define SKB_ALLOC_NAPI 0x04
858
859 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
860 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
861 {
862 return unlikely(skb->pfmemalloc);
863 }
864
865 /*
866 * skb might have a dst pointer attached, refcounted or not.
867 * _skb_refdst low order bit is set if refcount was _not_ taken
868 */
869 #define SKB_DST_NOREF 1UL
870 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
871
872 #define SKB_NFCT_PTRMASK ~(7UL)
873 /**
874 * skb_dst - returns skb dst_entry
875 * @skb: buffer
876 *
877 * Returns skb dst_entry, regardless of reference taken or not.
878 */
879 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
880 {
881 /* If refdst was not refcounted, check we still are in a
882 * rcu_read_lock section
883 */
884 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
885 !rcu_read_lock_held() &&
886 !rcu_read_lock_bh_held());
887 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
888 }
889
890 /**
891 * skb_dst_set - sets skb dst
892 * @skb: buffer
893 * @dst: dst entry
894 *
895 * Sets skb dst, assuming a reference was taken on dst and should
896 * be released by skb_dst_drop()
897 */
898 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
899 {
900 skb->_skb_refdst = (unsigned long)dst;
901 }
902
903 /**
904 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
905 * @skb: buffer
906 * @dst: dst entry
907 *
908 * Sets skb dst, assuming a reference was not taken on dst.
909 * If dst entry is cached, we do not take reference and dst_release
910 * will be avoided by refdst_drop. If dst entry is not cached, we take
911 * reference, so that last dst_release can destroy the dst immediately.
912 */
913 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
914 {
915 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
916 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
917 }
918
919 /**
920 * skb_dst_is_noref - Test if skb dst isn't refcounted
921 * @skb: buffer
922 */
923 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
924 {
925 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
926 }
927
928 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
929 {
930 return (struct rtable *)skb_dst(skb);
931 }
932
933 /* For mangling skb->pkt_type from user space side from applications
934 * such as nft, tc, etc, we only allow a conservative subset of
935 * possible pkt_types to be set.
936 */
937 static inline bool skb_pkt_type_ok(u32 ptype)
938 {
939 return ptype <= PACKET_OTHERHOST;
940 }
941
942 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
943 {
944 #ifdef CONFIG_NET_RX_BUSY_POLL
945 return skb->napi_id;
946 #else
947 return 0;
948 #endif
949 }
950
951 /* decrement the reference count and return true if we can free the skb */
952 static inline bool skb_unref(struct sk_buff *skb)
953 {
954 if (unlikely(!skb))
955 return false;
956 if (likely(refcount_read(&skb->users) == 1))
957 smp_rmb();
958 else if (likely(!refcount_dec_and_test(&skb->users)))
959 return false;
960
961 return true;
962 }
963
964 void skb_release_head_state(struct sk_buff *skb);
965 void kfree_skb(struct sk_buff *skb);
966 void kfree_skb_list(struct sk_buff *segs);
967 void skb_tx_error(struct sk_buff *skb);
968 void consume_skb(struct sk_buff *skb);
969 void __consume_stateless_skb(struct sk_buff *skb);
970 void __kfree_skb(struct sk_buff *skb);
971 extern struct kmem_cache *skbuff_head_cache;
972
973 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
974 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
975 bool *fragstolen, int *delta_truesize);
976
977 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
978 int node);
979 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
980 struct sk_buff *build_skb(void *data, unsigned int frag_size);
981 static inline struct sk_buff *alloc_skb(unsigned int size,
982 gfp_t priority)
983 {
984 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
985 }
986
987 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
988 unsigned long data_len,
989 int max_page_order,
990 int *errcode,
991 gfp_t gfp_mask);
992
993 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
994 struct sk_buff_fclones {
995 struct sk_buff skb1;
996
997 struct sk_buff skb2;
998
999 refcount_t fclone_ref;
1000 };
1001
1002 /**
1003 * skb_fclone_busy - check if fclone is busy
1004 * @sk: socket
1005 * @skb: buffer
1006 *
1007 * Returns true if skb is a fast clone, and its clone is not freed.
1008 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1009 * so we also check that this didnt happen.
1010 */
1011 static inline bool skb_fclone_busy(const struct sock *sk,
1012 const struct sk_buff *skb)
1013 {
1014 const struct sk_buff_fclones *fclones;
1015
1016 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1017
1018 return skb->fclone == SKB_FCLONE_ORIG &&
1019 refcount_read(&fclones->fclone_ref) > 1 &&
1020 fclones->skb2.sk == sk;
1021 }
1022
1023 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1024 gfp_t priority)
1025 {
1026 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1027 }
1028
1029 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1030 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1031 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1032 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1033 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1034 gfp_t gfp_mask, bool fclone);
1035 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1036 gfp_t gfp_mask)
1037 {
1038 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1039 }
1040
1041 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1042 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1043 unsigned int headroom);
1044 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1045 int newtailroom, gfp_t priority);
1046 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1047 int offset, int len);
1048 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1049 int offset, int len);
1050 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1051 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1052
1053 /**
1054 * skb_pad - zero pad the tail of an skb
1055 * @skb: buffer to pad
1056 * @pad: space to pad
1057 *
1058 * Ensure that a buffer is followed by a padding area that is zero
1059 * filled. Used by network drivers which may DMA or transfer data
1060 * beyond the buffer end onto the wire.
1061 *
1062 * May return error in out of memory cases. The skb is freed on error.
1063 */
1064 static inline int skb_pad(struct sk_buff *skb, int pad)
1065 {
1066 return __skb_pad(skb, pad, true);
1067 }
1068 #define dev_kfree_skb(a) consume_skb(a)
1069
1070 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1071 int getfrag(void *from, char *to, int offset,
1072 int len, int odd, struct sk_buff *skb),
1073 void *from, int length);
1074
1075 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1076 int offset, size_t size);
1077
1078 struct skb_seq_state {
1079 __u32 lower_offset;
1080 __u32 upper_offset;
1081 __u32 frag_idx;
1082 __u32 stepped_offset;
1083 struct sk_buff *root_skb;
1084 struct sk_buff *cur_skb;
1085 __u8 *frag_data;
1086 };
1087
1088 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1089 unsigned int to, struct skb_seq_state *st);
1090 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1091 struct skb_seq_state *st);
1092 void skb_abort_seq_read(struct skb_seq_state *st);
1093
1094 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1095 unsigned int to, struct ts_config *config);
1096
1097 /*
1098 * Packet hash types specify the type of hash in skb_set_hash.
1099 *
1100 * Hash types refer to the protocol layer addresses which are used to
1101 * construct a packet's hash. The hashes are used to differentiate or identify
1102 * flows of the protocol layer for the hash type. Hash types are either
1103 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1104 *
1105 * Properties of hashes:
1106 *
1107 * 1) Two packets in different flows have different hash values
1108 * 2) Two packets in the same flow should have the same hash value
1109 *
1110 * A hash at a higher layer is considered to be more specific. A driver should
1111 * set the most specific hash possible.
1112 *
1113 * A driver cannot indicate a more specific hash than the layer at which a hash
1114 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1115 *
1116 * A driver may indicate a hash level which is less specific than the
1117 * actual layer the hash was computed on. For instance, a hash computed
1118 * at L4 may be considered an L3 hash. This should only be done if the
1119 * driver can't unambiguously determine that the HW computed the hash at
1120 * the higher layer. Note that the "should" in the second property above
1121 * permits this.
1122 */
1123 enum pkt_hash_types {
1124 PKT_HASH_TYPE_NONE, /* Undefined type */
1125 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1126 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1127 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1128 };
1129
1130 static inline void skb_clear_hash(struct sk_buff *skb)
1131 {
1132 skb->hash = 0;
1133 skb->sw_hash = 0;
1134 skb->l4_hash = 0;
1135 }
1136
1137 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1138 {
1139 if (!skb->l4_hash)
1140 skb_clear_hash(skb);
1141 }
1142
1143 static inline void
1144 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1145 {
1146 skb->l4_hash = is_l4;
1147 skb->sw_hash = is_sw;
1148 skb->hash = hash;
1149 }
1150
1151 static inline void
1152 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1153 {
1154 /* Used by drivers to set hash from HW */
1155 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1156 }
1157
1158 static inline void
1159 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1160 {
1161 __skb_set_hash(skb, hash, true, is_l4);
1162 }
1163
1164 void __skb_get_hash(struct sk_buff *skb);
1165 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1166 u32 skb_get_poff(const struct sk_buff *skb);
1167 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1168 const struct flow_keys *keys, int hlen);
1169 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1170 void *data, int hlen_proto);
1171
1172 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1173 int thoff, u8 ip_proto)
1174 {
1175 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1176 }
1177
1178 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1179 const struct flow_dissector_key *key,
1180 unsigned int key_count);
1181
1182 bool __skb_flow_dissect(const struct sk_buff *skb,
1183 struct flow_dissector *flow_dissector,
1184 void *target_container,
1185 void *data, __be16 proto, int nhoff, int hlen,
1186 unsigned int flags);
1187
1188 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1189 struct flow_dissector *flow_dissector,
1190 void *target_container, unsigned int flags)
1191 {
1192 return __skb_flow_dissect(skb, flow_dissector, target_container,
1193 NULL, 0, 0, 0, flags);
1194 }
1195
1196 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1197 struct flow_keys *flow,
1198 unsigned int flags)
1199 {
1200 memset(flow, 0, sizeof(*flow));
1201 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
1202 NULL, 0, 0, 0, flags);
1203 }
1204
1205 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1206 void *data, __be16 proto,
1207 int nhoff, int hlen,
1208 unsigned int flags)
1209 {
1210 memset(flow, 0, sizeof(*flow));
1211 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
1212 data, proto, nhoff, hlen, flags);
1213 }
1214
1215 static inline __u32 skb_get_hash(struct sk_buff *skb)
1216 {
1217 if (!skb->l4_hash && !skb->sw_hash)
1218 __skb_get_hash(skb);
1219
1220 return skb->hash;
1221 }
1222
1223 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1224 {
1225 if (!skb->l4_hash && !skb->sw_hash) {
1226 struct flow_keys keys;
1227 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1228
1229 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1230 }
1231
1232 return skb->hash;
1233 }
1234
1235 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1236
1237 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1238 {
1239 return skb->hash;
1240 }
1241
1242 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1243 {
1244 to->hash = from->hash;
1245 to->sw_hash = from->sw_hash;
1246 to->l4_hash = from->l4_hash;
1247 };
1248
1249 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1250 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1251 {
1252 return skb->head + skb->end;
1253 }
1254
1255 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1256 {
1257 return skb->end;
1258 }
1259 #else
1260 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1261 {
1262 return skb->end;
1263 }
1264
1265 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1266 {
1267 return skb->end - skb->head;
1268 }
1269 #endif
1270
1271 /* Internal */
1272 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1273
1274 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1275 {
1276 return &skb_shinfo(skb)->hwtstamps;
1277 }
1278
1279 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1280 {
1281 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1282
1283 return is_zcopy ? skb_uarg(skb) : NULL;
1284 }
1285
1286 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
1287 {
1288 if (skb && uarg && !skb_zcopy(skb)) {
1289 sock_zerocopy_get(uarg);
1290 skb_shinfo(skb)->destructor_arg = uarg;
1291 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1292 }
1293 }
1294
1295 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1296 {
1297 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1298 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1299 }
1300
1301 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1302 {
1303 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1304 }
1305
1306 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1307 {
1308 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1309 }
1310
1311 /* Release a reference on a zerocopy structure */
1312 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1313 {
1314 struct ubuf_info *uarg = skb_zcopy(skb);
1315
1316 if (uarg) {
1317 if (uarg->callback == sock_zerocopy_callback) {
1318 uarg->zerocopy = uarg->zerocopy && zerocopy;
1319 sock_zerocopy_put(uarg);
1320 } else if (!skb_zcopy_is_nouarg(skb)) {
1321 uarg->callback(uarg, zerocopy);
1322 }
1323
1324 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1325 }
1326 }
1327
1328 /* Abort a zerocopy operation and revert zckey on error in send syscall */
1329 static inline void skb_zcopy_abort(struct sk_buff *skb)
1330 {
1331 struct ubuf_info *uarg = skb_zcopy(skb);
1332
1333 if (uarg) {
1334 sock_zerocopy_put_abort(uarg);
1335 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1336 }
1337 }
1338
1339 /**
1340 * skb_queue_empty - check if a queue is empty
1341 * @list: queue head
1342 *
1343 * Returns true if the queue is empty, false otherwise.
1344 */
1345 static inline int skb_queue_empty(const struct sk_buff_head *list)
1346 {
1347 return list->next == (const struct sk_buff *) list;
1348 }
1349
1350 /**
1351 * skb_queue_is_last - check if skb is the last entry in the queue
1352 * @list: queue head
1353 * @skb: buffer
1354 *
1355 * Returns true if @skb is the last buffer on the list.
1356 */
1357 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1358 const struct sk_buff *skb)
1359 {
1360 return skb->next == (const struct sk_buff *) list;
1361 }
1362
1363 /**
1364 * skb_queue_is_first - check if skb is the first entry in the queue
1365 * @list: queue head
1366 * @skb: buffer
1367 *
1368 * Returns true if @skb is the first buffer on the list.
1369 */
1370 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1371 const struct sk_buff *skb)
1372 {
1373 return skb->prev == (const struct sk_buff *) list;
1374 }
1375
1376 /**
1377 * skb_queue_next - return the next packet in the queue
1378 * @list: queue head
1379 * @skb: current buffer
1380 *
1381 * Return the next packet in @list after @skb. It is only valid to
1382 * call this if skb_queue_is_last() evaluates to false.
1383 */
1384 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1385 const struct sk_buff *skb)
1386 {
1387 /* This BUG_ON may seem severe, but if we just return then we
1388 * are going to dereference garbage.
1389 */
1390 BUG_ON(skb_queue_is_last(list, skb));
1391 return skb->next;
1392 }
1393
1394 /**
1395 * skb_queue_prev - return the prev packet in the queue
1396 * @list: queue head
1397 * @skb: current buffer
1398 *
1399 * Return the prev packet in @list before @skb. It is only valid to
1400 * call this if skb_queue_is_first() evaluates to false.
1401 */
1402 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1403 const struct sk_buff *skb)
1404 {
1405 /* This BUG_ON may seem severe, but if we just return then we
1406 * are going to dereference garbage.
1407 */
1408 BUG_ON(skb_queue_is_first(list, skb));
1409 return skb->prev;
1410 }
1411
1412 /**
1413 * skb_get - reference buffer
1414 * @skb: buffer to reference
1415 *
1416 * Makes another reference to a socket buffer and returns a pointer
1417 * to the buffer.
1418 */
1419 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1420 {
1421 refcount_inc(&skb->users);
1422 return skb;
1423 }
1424
1425 /*
1426 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1427 */
1428
1429 /**
1430 * skb_cloned - is the buffer a clone
1431 * @skb: buffer to check
1432 *
1433 * Returns true if the buffer was generated with skb_clone() and is
1434 * one of multiple shared copies of the buffer. Cloned buffers are
1435 * shared data so must not be written to under normal circumstances.
1436 */
1437 static inline int skb_cloned(const struct sk_buff *skb)
1438 {
1439 return skb->cloned &&
1440 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1441 }
1442
1443 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1444 {
1445 might_sleep_if(gfpflags_allow_blocking(pri));
1446
1447 if (skb_cloned(skb))
1448 return pskb_expand_head(skb, 0, 0, pri);
1449
1450 return 0;
1451 }
1452
1453 /**
1454 * skb_header_cloned - is the header a clone
1455 * @skb: buffer to check
1456 *
1457 * Returns true if modifying the header part of the buffer requires
1458 * the data to be copied.
1459 */
1460 static inline int skb_header_cloned(const struct sk_buff *skb)
1461 {
1462 int dataref;
1463
1464 if (!skb->cloned)
1465 return 0;
1466
1467 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1468 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1469 return dataref != 1;
1470 }
1471
1472 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1473 {
1474 might_sleep_if(gfpflags_allow_blocking(pri));
1475
1476 if (skb_header_cloned(skb))
1477 return pskb_expand_head(skb, 0, 0, pri);
1478
1479 return 0;
1480 }
1481
1482 /**
1483 * __skb_header_release - release reference to header
1484 * @skb: buffer to operate on
1485 */
1486 static inline void __skb_header_release(struct sk_buff *skb)
1487 {
1488 skb->nohdr = 1;
1489 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1490 }
1491
1492
1493 /**
1494 * skb_shared - is the buffer shared
1495 * @skb: buffer to check
1496 *
1497 * Returns true if more than one person has a reference to this
1498 * buffer.
1499 */
1500 static inline int skb_shared(const struct sk_buff *skb)
1501 {
1502 return refcount_read(&skb->users) != 1;
1503 }
1504
1505 /**
1506 * skb_share_check - check if buffer is shared and if so clone it
1507 * @skb: buffer to check
1508 * @pri: priority for memory allocation
1509 *
1510 * If the buffer is shared the buffer is cloned and the old copy
1511 * drops a reference. A new clone with a single reference is returned.
1512 * If the buffer is not shared the original buffer is returned. When
1513 * being called from interrupt status or with spinlocks held pri must
1514 * be GFP_ATOMIC.
1515 *
1516 * NULL is returned on a memory allocation failure.
1517 */
1518 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1519 {
1520 might_sleep_if(gfpflags_allow_blocking(pri));
1521 if (skb_shared(skb)) {
1522 struct sk_buff *nskb = skb_clone(skb, pri);
1523
1524 if (likely(nskb))
1525 consume_skb(skb);
1526 else
1527 kfree_skb(skb);
1528 skb = nskb;
1529 }
1530 return skb;
1531 }
1532
1533 /*
1534 * Copy shared buffers into a new sk_buff. We effectively do COW on
1535 * packets to handle cases where we have a local reader and forward
1536 * and a couple of other messy ones. The normal one is tcpdumping
1537 * a packet thats being forwarded.
1538 */
1539
1540 /**
1541 * skb_unshare - make a copy of a shared buffer
1542 * @skb: buffer to check
1543 * @pri: priority for memory allocation
1544 *
1545 * If the socket buffer is a clone then this function creates a new
1546 * copy of the data, drops a reference count on the old copy and returns
1547 * the new copy with the reference count at 1. If the buffer is not a clone
1548 * the original buffer is returned. When called with a spinlock held or
1549 * from interrupt state @pri must be %GFP_ATOMIC
1550 *
1551 * %NULL is returned on a memory allocation failure.
1552 */
1553 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1554 gfp_t pri)
1555 {
1556 might_sleep_if(gfpflags_allow_blocking(pri));
1557 if (skb_cloned(skb)) {
1558 struct sk_buff *nskb = skb_copy(skb, pri);
1559
1560 /* Free our shared copy */
1561 if (likely(nskb))
1562 consume_skb(skb);
1563 else
1564 kfree_skb(skb);
1565 skb = nskb;
1566 }
1567 return skb;
1568 }
1569
1570 /**
1571 * skb_peek - peek at the head of an &sk_buff_head
1572 * @list_: list to peek at
1573 *
1574 * Peek an &sk_buff. Unlike most other operations you _MUST_
1575 * be careful with this one. A peek leaves the buffer on the
1576 * list and someone else may run off with it. You must hold
1577 * the appropriate locks or have a private queue to do this.
1578 *
1579 * Returns %NULL for an empty list or a pointer to the head element.
1580 * The reference count is not incremented and the reference is therefore
1581 * volatile. Use with caution.
1582 */
1583 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1584 {
1585 struct sk_buff *skb = list_->next;
1586
1587 if (skb == (struct sk_buff *)list_)
1588 skb = NULL;
1589 return skb;
1590 }
1591
1592 /**
1593 * skb_peek_next - peek skb following the given one from a queue
1594 * @skb: skb to start from
1595 * @list_: list to peek at
1596 *
1597 * Returns %NULL when the end of the list is met or a pointer to the
1598 * next element. The reference count is not incremented and the
1599 * reference is therefore volatile. Use with caution.
1600 */
1601 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1602 const struct sk_buff_head *list_)
1603 {
1604 struct sk_buff *next = skb->next;
1605
1606 if (next == (struct sk_buff *)list_)
1607 next = NULL;
1608 return next;
1609 }
1610
1611 /**
1612 * skb_peek_tail - peek at the tail of an &sk_buff_head
1613 * @list_: list to peek at
1614 *
1615 * Peek an &sk_buff. Unlike most other operations you _MUST_
1616 * be careful with this one. A peek leaves the buffer on the
1617 * list and someone else may run off with it. You must hold
1618 * the appropriate locks or have a private queue to do this.
1619 *
1620 * Returns %NULL for an empty list or a pointer to the tail element.
1621 * The reference count is not incremented and the reference is therefore
1622 * volatile. Use with caution.
1623 */
1624 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1625 {
1626 struct sk_buff *skb = list_->prev;
1627
1628 if (skb == (struct sk_buff *)list_)
1629 skb = NULL;
1630 return skb;
1631
1632 }
1633
1634 /**
1635 * skb_queue_len - get queue length
1636 * @list_: list to measure
1637 *
1638 * Return the length of an &sk_buff queue.
1639 */
1640 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1641 {
1642 return list_->qlen;
1643 }
1644
1645 /**
1646 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1647 * @list: queue to initialize
1648 *
1649 * This initializes only the list and queue length aspects of
1650 * an sk_buff_head object. This allows to initialize the list
1651 * aspects of an sk_buff_head without reinitializing things like
1652 * the spinlock. It can also be used for on-stack sk_buff_head
1653 * objects where the spinlock is known to not be used.
1654 */
1655 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1656 {
1657 list->prev = list->next = (struct sk_buff *)list;
1658 list->qlen = 0;
1659 }
1660
1661 /*
1662 * This function creates a split out lock class for each invocation;
1663 * this is needed for now since a whole lot of users of the skb-queue
1664 * infrastructure in drivers have different locking usage (in hardirq)
1665 * than the networking core (in softirq only). In the long run either the
1666 * network layer or drivers should need annotation to consolidate the
1667 * main types of usage into 3 classes.
1668 */
1669 static inline void skb_queue_head_init(struct sk_buff_head *list)
1670 {
1671 spin_lock_init(&list->lock);
1672 __skb_queue_head_init(list);
1673 }
1674
1675 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1676 struct lock_class_key *class)
1677 {
1678 skb_queue_head_init(list);
1679 lockdep_set_class(&list->lock, class);
1680 }
1681
1682 /*
1683 * Insert an sk_buff on a list.
1684 *
1685 * The "__skb_xxxx()" functions are the non-atomic ones that
1686 * can only be called with interrupts disabled.
1687 */
1688 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1689 struct sk_buff_head *list);
1690 static inline void __skb_insert(struct sk_buff *newsk,
1691 struct sk_buff *prev, struct sk_buff *next,
1692 struct sk_buff_head *list)
1693 {
1694 newsk->next = next;
1695 newsk->prev = prev;
1696 next->prev = prev->next = newsk;
1697 list->qlen++;
1698 }
1699
1700 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1701 struct sk_buff *prev,
1702 struct sk_buff *next)
1703 {
1704 struct sk_buff *first = list->next;
1705 struct sk_buff *last = list->prev;
1706
1707 first->prev = prev;
1708 prev->next = first;
1709
1710 last->next = next;
1711 next->prev = last;
1712 }
1713
1714 /**
1715 * skb_queue_splice - join two skb lists, this is designed for stacks
1716 * @list: the new list to add
1717 * @head: the place to add it in the first list
1718 */
1719 static inline void skb_queue_splice(const struct sk_buff_head *list,
1720 struct sk_buff_head *head)
1721 {
1722 if (!skb_queue_empty(list)) {
1723 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1724 head->qlen += list->qlen;
1725 }
1726 }
1727
1728 /**
1729 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1730 * @list: the new list to add
1731 * @head: the place to add it in the first list
1732 *
1733 * The list at @list is reinitialised
1734 */
1735 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1736 struct sk_buff_head *head)
1737 {
1738 if (!skb_queue_empty(list)) {
1739 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1740 head->qlen += list->qlen;
1741 __skb_queue_head_init(list);
1742 }
1743 }
1744
1745 /**
1746 * skb_queue_splice_tail - join two skb lists, each list being a queue
1747 * @list: the new list to add
1748 * @head: the place to add it in the first list
1749 */
1750 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1751 struct sk_buff_head *head)
1752 {
1753 if (!skb_queue_empty(list)) {
1754 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1755 head->qlen += list->qlen;
1756 }
1757 }
1758
1759 /**
1760 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1761 * @list: the new list to add
1762 * @head: the place to add it in the first list
1763 *
1764 * Each of the lists is a queue.
1765 * The list at @list is reinitialised
1766 */
1767 static inline void skb_queue_splice_tail_init(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 __skb_queue_head_init(list);
1774 }
1775 }
1776
1777 /**
1778 * __skb_queue_after - queue a buffer at the list head
1779 * @list: list to use
1780 * @prev: place after this buffer
1781 * @newsk: buffer to queue
1782 *
1783 * Queue a buffer int the middle of a list. This function takes no locks
1784 * and you must therefore hold required locks before calling it.
1785 *
1786 * A buffer cannot be placed on two lists at the same time.
1787 */
1788 static inline void __skb_queue_after(struct sk_buff_head *list,
1789 struct sk_buff *prev,
1790 struct sk_buff *newsk)
1791 {
1792 __skb_insert(newsk, prev, prev->next, list);
1793 }
1794
1795 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1796 struct sk_buff_head *list);
1797
1798 static inline void __skb_queue_before(struct sk_buff_head *list,
1799 struct sk_buff *next,
1800 struct sk_buff *newsk)
1801 {
1802 __skb_insert(newsk, next->prev, next, list);
1803 }
1804
1805 /**
1806 * __skb_queue_head - queue a buffer at the list head
1807 * @list: list to use
1808 * @newsk: buffer to queue
1809 *
1810 * Queue a buffer at the start of a list. This function takes no locks
1811 * and you must therefore hold required locks before calling it.
1812 *
1813 * A buffer cannot be placed on two lists at the same time.
1814 */
1815 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1816 static inline void __skb_queue_head(struct sk_buff_head *list,
1817 struct sk_buff *newsk)
1818 {
1819 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1820 }
1821
1822 /**
1823 * __skb_queue_tail - queue a buffer at the list tail
1824 * @list: list to use
1825 * @newsk: buffer to queue
1826 *
1827 * Queue a buffer at the end 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_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1833 static inline void __skb_queue_tail(struct sk_buff_head *list,
1834 struct sk_buff *newsk)
1835 {
1836 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1837 }
1838
1839 /*
1840 * remove sk_buff from list. _Must_ be called atomically, and with
1841 * the list known..
1842 */
1843 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1844 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1845 {
1846 struct sk_buff *next, *prev;
1847
1848 list->qlen--;
1849 next = skb->next;
1850 prev = skb->prev;
1851 skb->next = skb->prev = NULL;
1852 next->prev = prev;
1853 prev->next = next;
1854 }
1855
1856 /**
1857 * __skb_dequeue - remove from the head of the queue
1858 * @list: list to dequeue from
1859 *
1860 * Remove the head of the list. This function does not take any locks
1861 * so must be used with appropriate locks held only. The head item is
1862 * returned or %NULL if the list is empty.
1863 */
1864 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1865 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1866 {
1867 struct sk_buff *skb = skb_peek(list);
1868 if (skb)
1869 __skb_unlink(skb, list);
1870 return skb;
1871 }
1872
1873 /**
1874 * __skb_dequeue_tail - remove from the tail of the queue
1875 * @list: list to dequeue from
1876 *
1877 * Remove the tail of the list. This function does not take any locks
1878 * so must be used with appropriate locks held only. The tail item is
1879 * returned or %NULL if the list is empty.
1880 */
1881 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1882 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1883 {
1884 struct sk_buff *skb = skb_peek_tail(list);
1885 if (skb)
1886 __skb_unlink(skb, list);
1887 return skb;
1888 }
1889
1890
1891 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1892 {
1893 return skb->data_len;
1894 }
1895
1896 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1897 {
1898 return skb->len - skb->data_len;
1899 }
1900
1901 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1902 {
1903 unsigned int i, len = 0;
1904
1905 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
1906 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1907 return len;
1908 }
1909
1910 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1911 {
1912 return skb_headlen(skb) + __skb_pagelen(skb);
1913 }
1914
1915 /**
1916 * __skb_fill_page_desc - initialise a paged fragment in an skb
1917 * @skb: buffer containing fragment to be initialised
1918 * @i: paged fragment index to initialise
1919 * @page: the page to use for this fragment
1920 * @off: the offset to the data with @page
1921 * @size: the length of the data
1922 *
1923 * Initialises the @i'th fragment of @skb to point to &size bytes at
1924 * offset @off within @page.
1925 *
1926 * Does not take any additional reference on the fragment.
1927 */
1928 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1929 struct page *page, int off, int size)
1930 {
1931 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1932
1933 /*
1934 * Propagate page pfmemalloc to the skb if we can. The problem is
1935 * that not all callers have unique ownership of the page but rely
1936 * on page_is_pfmemalloc doing the right thing(tm).
1937 */
1938 frag->page.p = page;
1939 frag->page_offset = off;
1940 skb_frag_size_set(frag, size);
1941
1942 page = compound_head(page);
1943 if (page_is_pfmemalloc(page))
1944 skb->pfmemalloc = true;
1945 }
1946
1947 /**
1948 * skb_fill_page_desc - initialise a paged fragment in an skb
1949 * @skb: buffer containing fragment to be initialised
1950 * @i: paged fragment index to initialise
1951 * @page: the page to use for this fragment
1952 * @off: the offset to the data with @page
1953 * @size: the length of the data
1954 *
1955 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1956 * @skb to point to @size bytes at offset @off within @page. In
1957 * addition updates @skb such that @i is the last fragment.
1958 *
1959 * Does not take any additional reference on the fragment.
1960 */
1961 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1962 struct page *page, int off, int size)
1963 {
1964 __skb_fill_page_desc(skb, i, page, off, size);
1965 skb_shinfo(skb)->nr_frags = i + 1;
1966 }
1967
1968 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1969 int size, unsigned int truesize);
1970
1971 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1972 unsigned int truesize);
1973
1974 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1975 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1976 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1977
1978 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1979 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1980 {
1981 return skb->head + skb->tail;
1982 }
1983
1984 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1985 {
1986 skb->tail = skb->data - skb->head;
1987 }
1988
1989 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1990 {
1991 skb_reset_tail_pointer(skb);
1992 skb->tail += offset;
1993 }
1994
1995 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1996 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1997 {
1998 return skb->tail;
1999 }
2000
2001 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2002 {
2003 skb->tail = skb->data;
2004 }
2005
2006 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2007 {
2008 skb->tail = skb->data + offset;
2009 }
2010
2011 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2012
2013 /*
2014 * Add data to an sk_buff
2015 */
2016 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2017 void *skb_put(struct sk_buff *skb, unsigned int len);
2018 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2019 {
2020 void *tmp = skb_tail_pointer(skb);
2021 SKB_LINEAR_ASSERT(skb);
2022 skb->tail += len;
2023 skb->len += len;
2024 return tmp;
2025 }
2026
2027 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2028 {
2029 void *tmp = __skb_put(skb, len);
2030
2031 memset(tmp, 0, len);
2032 return tmp;
2033 }
2034
2035 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2036 unsigned int len)
2037 {
2038 void *tmp = __skb_put(skb, len);
2039
2040 memcpy(tmp, data, len);
2041 return tmp;
2042 }
2043
2044 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2045 {
2046 *(u8 *)__skb_put(skb, 1) = val;
2047 }
2048
2049 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2050 {
2051 void *tmp = skb_put(skb, len);
2052
2053 memset(tmp, 0, len);
2054
2055 return tmp;
2056 }
2057
2058 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2059 unsigned int len)
2060 {
2061 void *tmp = skb_put(skb, len);
2062
2063 memcpy(tmp, data, len);
2064
2065 return tmp;
2066 }
2067
2068 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2069 {
2070 *(u8 *)skb_put(skb, 1) = val;
2071 }
2072
2073 void *skb_push(struct sk_buff *skb, unsigned int len);
2074 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2075 {
2076 skb->data -= len;
2077 skb->len += len;
2078 return skb->data;
2079 }
2080
2081 void *skb_pull(struct sk_buff *skb, unsigned int len);
2082 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2083 {
2084 skb->len -= len;
2085 BUG_ON(skb->len < skb->data_len);
2086 return skb->data += len;
2087 }
2088
2089 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2090 {
2091 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2092 }
2093
2094 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2095
2096 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2097 {
2098 if (len > skb_headlen(skb) &&
2099 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2100 return NULL;
2101 skb->len -= len;
2102 return skb->data += len;
2103 }
2104
2105 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2106 {
2107 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2108 }
2109
2110 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2111 {
2112 if (likely(len <= skb_headlen(skb)))
2113 return 1;
2114 if (unlikely(len > skb->len))
2115 return 0;
2116 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2117 }
2118
2119 void skb_condense(struct sk_buff *skb);
2120
2121 /**
2122 * skb_headroom - bytes at buffer head
2123 * @skb: buffer to check
2124 *
2125 * Return the number of bytes of free space at the head of an &sk_buff.
2126 */
2127 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2128 {
2129 return skb->data - skb->head;
2130 }
2131
2132 /**
2133 * skb_tailroom - bytes at buffer end
2134 * @skb: buffer to check
2135 *
2136 * Return the number of bytes of free space at the tail of an sk_buff
2137 */
2138 static inline int skb_tailroom(const struct sk_buff *skb)
2139 {
2140 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2141 }
2142
2143 /**
2144 * skb_availroom - bytes at buffer end
2145 * @skb: buffer to check
2146 *
2147 * Return the number of bytes of free space at the tail of an sk_buff
2148 * allocated by sk_stream_alloc()
2149 */
2150 static inline int skb_availroom(const struct sk_buff *skb)
2151 {
2152 if (skb_is_nonlinear(skb))
2153 return 0;
2154
2155 return skb->end - skb->tail - skb->reserved_tailroom;
2156 }
2157
2158 /**
2159 * skb_reserve - adjust headroom
2160 * @skb: buffer to alter
2161 * @len: bytes to move
2162 *
2163 * Increase the headroom of an empty &sk_buff by reducing the tail
2164 * room. This is only allowed for an empty buffer.
2165 */
2166 static inline void skb_reserve(struct sk_buff *skb, int len)
2167 {
2168 skb->data += len;
2169 skb->tail += len;
2170 }
2171
2172 /**
2173 * skb_tailroom_reserve - adjust reserved_tailroom
2174 * @skb: buffer to alter
2175 * @mtu: maximum amount of headlen permitted
2176 * @needed_tailroom: minimum amount of reserved_tailroom
2177 *
2178 * Set reserved_tailroom so that headlen can be as large as possible but
2179 * not larger than mtu and tailroom cannot be smaller than
2180 * needed_tailroom.
2181 * The required headroom should already have been reserved before using
2182 * this function.
2183 */
2184 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2185 unsigned int needed_tailroom)
2186 {
2187 SKB_LINEAR_ASSERT(skb);
2188 if (mtu < skb_tailroom(skb) - needed_tailroom)
2189 /* use at most mtu */
2190 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2191 else
2192 /* use up to all available space */
2193 skb->reserved_tailroom = needed_tailroom;
2194 }
2195
2196 #define ENCAP_TYPE_ETHER 0
2197 #define ENCAP_TYPE_IPPROTO 1
2198
2199 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2200 __be16 protocol)
2201 {
2202 skb->inner_protocol = protocol;
2203 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2204 }
2205
2206 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2207 __u8 ipproto)
2208 {
2209 skb->inner_ipproto = ipproto;
2210 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2211 }
2212
2213 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2214 {
2215 skb->inner_mac_header = skb->mac_header;
2216 skb->inner_network_header = skb->network_header;
2217 skb->inner_transport_header = skb->transport_header;
2218 }
2219
2220 static inline void skb_reset_mac_len(struct sk_buff *skb)
2221 {
2222 skb->mac_len = skb->network_header - skb->mac_header;
2223 }
2224
2225 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2226 *skb)
2227 {
2228 return skb->head + skb->inner_transport_header;
2229 }
2230
2231 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2232 {
2233 return skb_inner_transport_header(skb) - skb->data;
2234 }
2235
2236 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2237 {
2238 skb->inner_transport_header = skb->data - skb->head;
2239 }
2240
2241 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2242 const int offset)
2243 {
2244 skb_reset_inner_transport_header(skb);
2245 skb->inner_transport_header += offset;
2246 }
2247
2248 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2249 {
2250 return skb->head + skb->inner_network_header;
2251 }
2252
2253 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2254 {
2255 skb->inner_network_header = skb->data - skb->head;
2256 }
2257
2258 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2259 const int offset)
2260 {
2261 skb_reset_inner_network_header(skb);
2262 skb->inner_network_header += offset;
2263 }
2264
2265 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2266 {
2267 return skb->head + skb->inner_mac_header;
2268 }
2269
2270 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2271 {
2272 skb->inner_mac_header = skb->data - skb->head;
2273 }
2274
2275 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2276 const int offset)
2277 {
2278 skb_reset_inner_mac_header(skb);
2279 skb->inner_mac_header += offset;
2280 }
2281 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2282 {
2283 return skb->transport_header != (typeof(skb->transport_header))~0U;
2284 }
2285
2286 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2287 {
2288 return skb->head + skb->transport_header;
2289 }
2290
2291 static inline void skb_reset_transport_header(struct sk_buff *skb)
2292 {
2293 skb->transport_header = skb->data - skb->head;
2294 }
2295
2296 static inline void skb_set_transport_header(struct sk_buff *skb,
2297 const int offset)
2298 {
2299 skb_reset_transport_header(skb);
2300 skb->transport_header += offset;
2301 }
2302
2303 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2304 {
2305 return skb->head + skb->network_header;
2306 }
2307
2308 static inline void skb_reset_network_header(struct sk_buff *skb)
2309 {
2310 skb->network_header = skb->data - skb->head;
2311 }
2312
2313 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2314 {
2315 skb_reset_network_header(skb);
2316 skb->network_header += offset;
2317 }
2318
2319 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2320 {
2321 return skb->head + skb->mac_header;
2322 }
2323
2324 static inline int skb_mac_offset(const struct sk_buff *skb)
2325 {
2326 return skb_mac_header(skb) - skb->data;
2327 }
2328
2329 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2330 {
2331 return skb->network_header - skb->mac_header;
2332 }
2333
2334 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2335 {
2336 return skb->mac_header != (typeof(skb->mac_header))~0U;
2337 }
2338
2339 static inline void skb_reset_mac_header(struct sk_buff *skb)
2340 {
2341 skb->mac_header = skb->data - skb->head;
2342 }
2343
2344 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2345 {
2346 skb_reset_mac_header(skb);
2347 skb->mac_header += offset;
2348 }
2349
2350 static inline void skb_pop_mac_header(struct sk_buff *skb)
2351 {
2352 skb->mac_header = skb->network_header;
2353 }
2354
2355 static inline void skb_probe_transport_header(struct sk_buff *skb,
2356 const int offset_hint)
2357 {
2358 struct flow_keys keys;
2359
2360 if (skb_transport_header_was_set(skb))
2361 return;
2362 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
2363 skb_set_transport_header(skb, keys.control.thoff);
2364 else
2365 skb_set_transport_header(skb, offset_hint);
2366 }
2367
2368 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2369 {
2370 if (skb_mac_header_was_set(skb)) {
2371 const unsigned char *old_mac = skb_mac_header(skb);
2372
2373 skb_set_mac_header(skb, -skb->mac_len);
2374 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2375 }
2376 }
2377
2378 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2379 {
2380 return skb->csum_start - skb_headroom(skb);
2381 }
2382
2383 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2384 {
2385 return skb->head + skb->csum_start;
2386 }
2387
2388 static inline int skb_transport_offset(const struct sk_buff *skb)
2389 {
2390 return skb_transport_header(skb) - skb->data;
2391 }
2392
2393 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2394 {
2395 return skb->transport_header - skb->network_header;
2396 }
2397
2398 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2399 {
2400 return skb->inner_transport_header - skb->inner_network_header;
2401 }
2402
2403 static inline int skb_network_offset(const struct sk_buff *skb)
2404 {
2405 return skb_network_header(skb) - skb->data;
2406 }
2407
2408 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2409 {
2410 return skb_inner_network_header(skb) - skb->data;
2411 }
2412
2413 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2414 {
2415 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2416 }
2417
2418 /*
2419 * CPUs often take a performance hit when accessing unaligned memory
2420 * locations. The actual performance hit varies, it can be small if the
2421 * hardware handles it or large if we have to take an exception and fix it
2422 * in software.
2423 *
2424 * Since an ethernet header is 14 bytes network drivers often end up with
2425 * the IP header at an unaligned offset. The IP header can be aligned by
2426 * shifting the start of the packet by 2 bytes. Drivers should do this
2427 * with:
2428 *
2429 * skb_reserve(skb, NET_IP_ALIGN);
2430 *
2431 * The downside to this alignment of the IP header is that the DMA is now
2432 * unaligned. On some architectures the cost of an unaligned DMA is high
2433 * and this cost outweighs the gains made by aligning the IP header.
2434 *
2435 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2436 * to be overridden.
2437 */
2438 #ifndef NET_IP_ALIGN
2439 #define NET_IP_ALIGN 2
2440 #endif
2441
2442 /*
2443 * The networking layer reserves some headroom in skb data (via
2444 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2445 * the header has to grow. In the default case, if the header has to grow
2446 * 32 bytes or less we avoid the reallocation.
2447 *
2448 * Unfortunately this headroom changes the DMA alignment of the resulting
2449 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2450 * on some architectures. An architecture can override this value,
2451 * perhaps setting it to a cacheline in size (since that will maintain
2452 * cacheline alignment of the DMA). It must be a power of 2.
2453 *
2454 * Various parts of the networking layer expect at least 32 bytes of
2455 * headroom, you should not reduce this.
2456 *
2457 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2458 * to reduce average number of cache lines per packet.
2459 * get_rps_cpus() for example only access one 64 bytes aligned block :
2460 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2461 */
2462 #ifndef NET_SKB_PAD
2463 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2464 #endif
2465
2466 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2467
2468 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2469 {
2470 if (unlikely(skb_is_nonlinear(skb))) {
2471 WARN_ON(1);
2472 return;
2473 }
2474 skb->len = len;
2475 skb_set_tail_pointer(skb, len);
2476 }
2477
2478 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2479 {
2480 __skb_set_length(skb, len);
2481 }
2482
2483 void skb_trim(struct sk_buff *skb, unsigned int len);
2484
2485 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2486 {
2487 if (skb->data_len)
2488 return ___pskb_trim(skb, len);
2489 __skb_trim(skb, len);
2490 return 0;
2491 }
2492
2493 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2494 {
2495 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2496 }
2497
2498 /**
2499 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2500 * @skb: buffer to alter
2501 * @len: new length
2502 *
2503 * This is identical to pskb_trim except that the caller knows that
2504 * the skb is not cloned so we should never get an error due to out-
2505 * of-memory.
2506 */
2507 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2508 {
2509 int err = pskb_trim(skb, len);
2510 BUG_ON(err);
2511 }
2512
2513 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2514 {
2515 unsigned int diff = len - skb->len;
2516
2517 if (skb_tailroom(skb) < diff) {
2518 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2519 GFP_ATOMIC);
2520 if (ret)
2521 return ret;
2522 }
2523 __skb_set_length(skb, len);
2524 return 0;
2525 }
2526
2527 /**
2528 * skb_orphan - orphan a buffer
2529 * @skb: buffer to orphan
2530 *
2531 * If a buffer currently has an owner then we call the owner's
2532 * destructor function and make the @skb unowned. The buffer continues
2533 * to exist but is no longer charged to its former owner.
2534 */
2535 static inline void skb_orphan(struct sk_buff *skb)
2536 {
2537 if (skb->destructor) {
2538 skb->destructor(skb);
2539 skb->destructor = NULL;
2540 skb->sk = NULL;
2541 } else {
2542 BUG_ON(skb->sk);
2543 }
2544 }
2545
2546 /**
2547 * skb_orphan_frags - orphan the frags contained in a buffer
2548 * @skb: buffer to orphan frags from
2549 * @gfp_mask: allocation mask for replacement pages
2550 *
2551 * For each frag in the SKB which needs a destructor (i.e. has an
2552 * owner) create a copy of that frag and release the original
2553 * page by calling the destructor.
2554 */
2555 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2556 {
2557 if (likely(!skb_zcopy(skb)))
2558 return 0;
2559 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2560 return 0;
2561 return skb_copy_ubufs(skb, gfp_mask);
2562 }
2563
2564 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2565 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2566 {
2567 if (likely(!skb_zcopy(skb)))
2568 return 0;
2569 return skb_copy_ubufs(skb, gfp_mask);
2570 }
2571
2572 /**
2573 * __skb_queue_purge - empty a list
2574 * @list: list to empty
2575 *
2576 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2577 * the list and one reference dropped. This function does not take the
2578 * list lock and the caller must hold the relevant locks to use it.
2579 */
2580 void skb_queue_purge(struct sk_buff_head *list);
2581 static inline void __skb_queue_purge(struct sk_buff_head *list)
2582 {
2583 struct sk_buff *skb;
2584 while ((skb = __skb_dequeue(list)) != NULL)
2585 kfree_skb(skb);
2586 }
2587
2588 unsigned int skb_rbtree_purge(struct rb_root *root);
2589
2590 void *netdev_alloc_frag(unsigned int fragsz);
2591
2592 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2593 gfp_t gfp_mask);
2594
2595 /**
2596 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2597 * @dev: network device to receive on
2598 * @length: length to allocate
2599 *
2600 * Allocate a new &sk_buff and assign it a usage count of one. The
2601 * buffer has unspecified headroom built in. Users should allocate
2602 * the headroom they think they need without accounting for the
2603 * built in space. The built in space is used for optimisations.
2604 *
2605 * %NULL is returned if there is no free memory. Although this function
2606 * allocates memory it can be called from an interrupt.
2607 */
2608 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2609 unsigned int length)
2610 {
2611 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2612 }
2613
2614 /* legacy helper around __netdev_alloc_skb() */
2615 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2616 gfp_t gfp_mask)
2617 {
2618 return __netdev_alloc_skb(NULL, length, gfp_mask);
2619 }
2620
2621 /* legacy helper around netdev_alloc_skb() */
2622 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2623 {
2624 return netdev_alloc_skb(NULL, length);
2625 }
2626
2627
2628 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2629 unsigned int length, gfp_t gfp)
2630 {
2631 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2632
2633 if (NET_IP_ALIGN && skb)
2634 skb_reserve(skb, NET_IP_ALIGN);
2635 return skb;
2636 }
2637
2638 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2639 unsigned int length)
2640 {
2641 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2642 }
2643
2644 static inline void skb_free_frag(void *addr)
2645 {
2646 page_frag_free(addr);
2647 }
2648
2649 void *napi_alloc_frag(unsigned int fragsz);
2650 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2651 unsigned int length, gfp_t gfp_mask);
2652 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2653 unsigned int length)
2654 {
2655 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2656 }
2657 void napi_consume_skb(struct sk_buff *skb, int budget);
2658
2659 void __kfree_skb_flush(void);
2660 void __kfree_skb_defer(struct sk_buff *skb);
2661
2662 /**
2663 * __dev_alloc_pages - allocate page for network Rx
2664 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2665 * @order: size of the allocation
2666 *
2667 * Allocate a new page.
2668 *
2669 * %NULL is returned if there is no free memory.
2670 */
2671 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2672 unsigned int order)
2673 {
2674 /* This piece of code contains several assumptions.
2675 * 1. This is for device Rx, therefor a cold page is preferred.
2676 * 2. The expectation is the user wants a compound page.
2677 * 3. If requesting a order 0 page it will not be compound
2678 * due to the check to see if order has a value in prep_new_page
2679 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2680 * code in gfp_to_alloc_flags that should be enforcing this.
2681 */
2682 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2683
2684 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2685 }
2686
2687 static inline struct page *dev_alloc_pages(unsigned int order)
2688 {
2689 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2690 }
2691
2692 /**
2693 * __dev_alloc_page - allocate a page for network Rx
2694 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2695 *
2696 * Allocate a new page.
2697 *
2698 * %NULL is returned if there is no free memory.
2699 */
2700 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2701 {
2702 return __dev_alloc_pages(gfp_mask, 0);
2703 }
2704
2705 static inline struct page *dev_alloc_page(void)
2706 {
2707 return dev_alloc_pages(0);
2708 }
2709
2710 /**
2711 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2712 * @page: The page that was allocated from skb_alloc_page
2713 * @skb: The skb that may need pfmemalloc set
2714 */
2715 static inline void skb_propagate_pfmemalloc(struct page *page,
2716 struct sk_buff *skb)
2717 {
2718 if (page_is_pfmemalloc(page))
2719 skb->pfmemalloc = true;
2720 }
2721
2722 /**
2723 * skb_frag_page - retrieve the page referred to by a paged fragment
2724 * @frag: the paged fragment
2725 *
2726 * Returns the &struct page associated with @frag.
2727 */
2728 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2729 {
2730 return frag->page.p;
2731 }
2732
2733 /**
2734 * __skb_frag_ref - take an addition reference on a paged fragment.
2735 * @frag: the paged fragment
2736 *
2737 * Takes an additional reference on the paged fragment @frag.
2738 */
2739 static inline void __skb_frag_ref(skb_frag_t *frag)
2740 {
2741 get_page(skb_frag_page(frag));
2742 }
2743
2744 /**
2745 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2746 * @skb: the buffer
2747 * @f: the fragment offset.
2748 *
2749 * Takes an additional reference on the @f'th paged fragment of @skb.
2750 */
2751 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2752 {
2753 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2754 }
2755
2756 /**
2757 * __skb_frag_unref - release a reference on a paged fragment.
2758 * @frag: the paged fragment
2759 *
2760 * Releases a reference on the paged fragment @frag.
2761 */
2762 static inline void __skb_frag_unref(skb_frag_t *frag)
2763 {
2764 put_page(skb_frag_page(frag));
2765 }
2766
2767 /**
2768 * skb_frag_unref - release a reference on a paged fragment of an skb.
2769 * @skb: the buffer
2770 * @f: the fragment offset
2771 *
2772 * Releases a reference on the @f'th paged fragment of @skb.
2773 */
2774 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2775 {
2776 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2777 }
2778
2779 /**
2780 * skb_frag_address - gets the address of the data contained in a paged fragment
2781 * @frag: the paged fragment buffer
2782 *
2783 * Returns the address of the data within @frag. The page must already
2784 * be mapped.
2785 */
2786 static inline void *skb_frag_address(const skb_frag_t *frag)
2787 {
2788 return page_address(skb_frag_page(frag)) + frag->page_offset;
2789 }
2790
2791 /**
2792 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2793 * @frag: the paged fragment buffer
2794 *
2795 * Returns the address of the data within @frag. Checks that the page
2796 * is mapped and returns %NULL otherwise.
2797 */
2798 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2799 {
2800 void *ptr = page_address(skb_frag_page(frag));
2801 if (unlikely(!ptr))
2802 return NULL;
2803
2804 return ptr + frag->page_offset;
2805 }
2806
2807 /**
2808 * __skb_frag_set_page - sets the page contained in a paged fragment
2809 * @frag: the paged fragment
2810 * @page: the page to set
2811 *
2812 * Sets the fragment @frag to contain @page.
2813 */
2814 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2815 {
2816 frag->page.p = page;
2817 }
2818
2819 /**
2820 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2821 * @skb: the buffer
2822 * @f: the fragment offset
2823 * @page: the page to set
2824 *
2825 * Sets the @f'th fragment of @skb to contain @page.
2826 */
2827 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2828 struct page *page)
2829 {
2830 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2831 }
2832
2833 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2834
2835 /**
2836 * skb_frag_dma_map - maps a paged fragment via the DMA API
2837 * @dev: the device to map the fragment to
2838 * @frag: the paged fragment to map
2839 * @offset: the offset within the fragment (starting at the
2840 * fragment's own offset)
2841 * @size: the number of bytes to map
2842 * @dir: the direction of the mapping (``PCI_DMA_*``)
2843 *
2844 * Maps the page associated with @frag to @device.
2845 */
2846 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2847 const skb_frag_t *frag,
2848 size_t offset, size_t size,
2849 enum dma_data_direction dir)
2850 {
2851 return dma_map_page(dev, skb_frag_page(frag),
2852 frag->page_offset + offset, size, dir);
2853 }
2854
2855 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2856 gfp_t gfp_mask)
2857 {
2858 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2859 }
2860
2861
2862 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2863 gfp_t gfp_mask)
2864 {
2865 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2866 }
2867
2868
2869 /**
2870 * skb_clone_writable - is the header of a clone writable
2871 * @skb: buffer to check
2872 * @len: length up to which to write
2873 *
2874 * Returns true if modifying the header part of the cloned buffer
2875 * does not requires the data to be copied.
2876 */
2877 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2878 {
2879 return !skb_header_cloned(skb) &&
2880 skb_headroom(skb) + len <= skb->hdr_len;
2881 }
2882
2883 static inline int skb_try_make_writable(struct sk_buff *skb,
2884 unsigned int write_len)
2885 {
2886 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2887 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2888 }
2889
2890 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2891 int cloned)
2892 {
2893 int delta = 0;
2894
2895 if (headroom > skb_headroom(skb))
2896 delta = headroom - skb_headroom(skb);
2897
2898 if (delta || cloned)
2899 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2900 GFP_ATOMIC);
2901 return 0;
2902 }
2903
2904 /**
2905 * skb_cow - copy header of skb when it is required
2906 * @skb: buffer to cow
2907 * @headroom: needed headroom
2908 *
2909 * If the skb passed lacks sufficient headroom or its data part
2910 * is shared, data is reallocated. If reallocation fails, an error
2911 * is returned and original skb is not changed.
2912 *
2913 * The result is skb with writable area skb->head...skb->tail
2914 * and at least @headroom of space at head.
2915 */
2916 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2917 {
2918 return __skb_cow(skb, headroom, skb_cloned(skb));
2919 }
2920
2921 /**
2922 * skb_cow_head - skb_cow but only making the head writable
2923 * @skb: buffer to cow
2924 * @headroom: needed headroom
2925 *
2926 * This function is identical to skb_cow except that we replace the
2927 * skb_cloned check by skb_header_cloned. It should be used when
2928 * you only need to push on some header and do not need to modify
2929 * the data.
2930 */
2931 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2932 {
2933 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2934 }
2935
2936 /**
2937 * skb_padto - pad an skbuff up to a minimal size
2938 * @skb: buffer to pad
2939 * @len: minimal length
2940 *
2941 * Pads up a buffer to ensure the trailing bytes exist and are
2942 * blanked. If the buffer already contains sufficient data it
2943 * is untouched. Otherwise it is extended. Returns zero on
2944 * success. The skb is freed on error.
2945 */
2946 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2947 {
2948 unsigned int size = skb->len;
2949 if (likely(size >= len))
2950 return 0;
2951 return skb_pad(skb, len - size);
2952 }
2953
2954 /**
2955 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2956 * @skb: buffer to pad
2957 * @len: minimal length
2958 * @free_on_error: free buffer on error
2959 *
2960 * Pads up a buffer to ensure the trailing bytes exist and are
2961 * blanked. If the buffer already contains sufficient data it
2962 * is untouched. Otherwise it is extended. Returns zero on
2963 * success. The skb is freed on error if @free_on_error is true.
2964 */
2965 static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
2966 bool free_on_error)
2967 {
2968 unsigned int size = skb->len;
2969
2970 if (unlikely(size < len)) {
2971 len -= size;
2972 if (__skb_pad(skb, len, free_on_error))
2973 return -ENOMEM;
2974 __skb_put(skb, len);
2975 }
2976 return 0;
2977 }
2978
2979 /**
2980 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2981 * @skb: buffer to pad
2982 * @len: minimal length
2983 *
2984 * Pads up a buffer to ensure the trailing bytes exist and are
2985 * blanked. If the buffer already contains sufficient data it
2986 * is untouched. Otherwise it is extended. Returns zero on
2987 * success. The skb is freed on error.
2988 */
2989 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2990 {
2991 return __skb_put_padto(skb, len, true);
2992 }
2993
2994 static inline int skb_add_data(struct sk_buff *skb,
2995 struct iov_iter *from, int copy)
2996 {
2997 const int off = skb->len;
2998
2999 if (skb->ip_summed == CHECKSUM_NONE) {
3000 __wsum csum = 0;
3001 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3002 &csum, from)) {
3003 skb->csum = csum_block_add(skb->csum, csum, off);
3004 return 0;
3005 }
3006 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3007 return 0;
3008
3009 __skb_trim(skb, off);
3010 return -EFAULT;
3011 }
3012
3013 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3014 const struct page *page, int off)
3015 {
3016 if (skb_zcopy(skb))
3017 return false;
3018 if (i) {
3019 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
3020
3021 return page == skb_frag_page(frag) &&
3022 off == frag->page_offset + skb_frag_size(frag);
3023 }
3024 return false;
3025 }
3026
3027 static inline int __skb_linearize(struct sk_buff *skb)
3028 {
3029 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3030 }
3031
3032 /**
3033 * skb_linearize - convert paged skb to linear one
3034 * @skb: buffer to linarize
3035 *
3036 * If there is no free memory -ENOMEM is returned, otherwise zero
3037 * is returned and the old skb data released.
3038 */
3039 static inline int skb_linearize(struct sk_buff *skb)
3040 {
3041 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3042 }
3043
3044 /**
3045 * skb_has_shared_frag - can any frag be overwritten
3046 * @skb: buffer to test
3047 *
3048 * Return true if the skb has at least one frag that might be modified
3049 * by an external entity (as in vmsplice()/sendfile())
3050 */
3051 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3052 {
3053 return skb_is_nonlinear(skb) &&
3054 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3055 }
3056
3057 /**
3058 * skb_linearize_cow - make sure skb is linear and writable
3059 * @skb: buffer to process
3060 *
3061 * If there is no free memory -ENOMEM is returned, otherwise zero
3062 * is returned and the old skb data released.
3063 */
3064 static inline int skb_linearize_cow(struct sk_buff *skb)
3065 {
3066 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3067 __skb_linearize(skb) : 0;
3068 }
3069
3070 static __always_inline void
3071 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3072 unsigned int off)
3073 {
3074 if (skb->ip_summed == CHECKSUM_COMPLETE)
3075 skb->csum = csum_block_sub(skb->csum,
3076 csum_partial(start, len, 0), off);
3077 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3078 skb_checksum_start_offset(skb) < 0)
3079 skb->ip_summed = CHECKSUM_NONE;
3080 }
3081
3082 /**
3083 * skb_postpull_rcsum - update checksum for received skb after pull
3084 * @skb: buffer to update
3085 * @start: start of data before pull
3086 * @len: length of data pulled
3087 *
3088 * After doing a pull on a received packet, you need to call this to
3089 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3090 * CHECKSUM_NONE so that it can be recomputed from scratch.
3091 */
3092 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3093 const void *start, unsigned int len)
3094 {
3095 __skb_postpull_rcsum(skb, start, len, 0);
3096 }
3097
3098 static __always_inline void
3099 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3100 unsigned int off)
3101 {
3102 if (skb->ip_summed == CHECKSUM_COMPLETE)
3103 skb->csum = csum_block_add(skb->csum,
3104 csum_partial(start, len, 0), off);
3105 }
3106
3107 /**
3108 * skb_postpush_rcsum - update checksum for received skb after push
3109 * @skb: buffer to update
3110 * @start: start of data after push
3111 * @len: length of data pushed
3112 *
3113 * After doing a push on a received packet, you need to call this to
3114 * update the CHECKSUM_COMPLETE checksum.
3115 */
3116 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3117 const void *start, unsigned int len)
3118 {
3119 __skb_postpush_rcsum(skb, start, len, 0);
3120 }
3121
3122 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3123
3124 /**
3125 * skb_push_rcsum - push skb and update receive checksum
3126 * @skb: buffer to update
3127 * @len: length of data pulled
3128 *
3129 * This function performs an skb_push on the packet and updates
3130 * the CHECKSUM_COMPLETE checksum. It should be used on
3131 * receive path processing instead of skb_push unless you know
3132 * that the checksum difference is zero (e.g., a valid IP header)
3133 * or you are setting ip_summed to CHECKSUM_NONE.
3134 */
3135 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3136 {
3137 skb_push(skb, len);
3138 skb_postpush_rcsum(skb, skb->data, len);
3139 return skb->data;
3140 }
3141
3142 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3143 /**
3144 * pskb_trim_rcsum - trim received skb and update checksum
3145 * @skb: buffer to trim
3146 * @len: new length
3147 *
3148 * This is exactly the same as pskb_trim except that it ensures the
3149 * checksum of received packets are still valid after the operation.
3150 * It can change skb pointers.
3151 */
3152
3153 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3154 {
3155 if (likely(len >= skb->len))
3156 return 0;
3157 return pskb_trim_rcsum_slow(skb, len);
3158 }
3159
3160 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3161 {
3162 if (skb->ip_summed == CHECKSUM_COMPLETE)
3163 skb->ip_summed = CHECKSUM_NONE;
3164 __skb_trim(skb, len);
3165 return 0;
3166 }
3167
3168 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3169 {
3170 if (skb->ip_summed == CHECKSUM_COMPLETE)
3171 skb->ip_summed = CHECKSUM_NONE;
3172 return __skb_grow(skb, len);
3173 }
3174
3175 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3176 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3177 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3178 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3179 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3180
3181 #define skb_queue_walk(queue, skb) \
3182 for (skb = (queue)->next; \
3183 skb != (struct sk_buff *)(queue); \
3184 skb = skb->next)
3185
3186 #define skb_queue_walk_safe(queue, skb, tmp) \
3187 for (skb = (queue)->next, tmp = skb->next; \
3188 skb != (struct sk_buff *)(queue); \
3189 skb = tmp, tmp = skb->next)
3190
3191 #define skb_queue_walk_from(queue, skb) \
3192 for (; skb != (struct sk_buff *)(queue); \
3193 skb = skb->next)
3194
3195 #define skb_rbtree_walk(skb, root) \
3196 for (skb = skb_rb_first(root); skb != NULL; \
3197 skb = skb_rb_next(skb))
3198
3199 #define skb_rbtree_walk_from(skb) \
3200 for (; skb != NULL; \
3201 skb = skb_rb_next(skb))
3202
3203 #define skb_rbtree_walk_from_safe(skb, tmp) \
3204 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3205 skb = tmp)
3206
3207 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3208 for (tmp = skb->next; \
3209 skb != (struct sk_buff *)(queue); \
3210 skb = tmp, tmp = skb->next)
3211
3212 #define skb_queue_reverse_walk(queue, skb) \
3213 for (skb = (queue)->prev; \
3214 skb != (struct sk_buff *)(queue); \
3215 skb = skb->prev)
3216
3217 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3218 for (skb = (queue)->prev, tmp = skb->prev; \
3219 skb != (struct sk_buff *)(queue); \
3220 skb = tmp, tmp = skb->prev)
3221
3222 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3223 for (tmp = skb->prev; \
3224 skb != (struct sk_buff *)(queue); \
3225 skb = tmp, tmp = skb->prev)
3226
3227 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3228 {
3229 return skb_shinfo(skb)->frag_list != NULL;
3230 }
3231
3232 static inline void skb_frag_list_init(struct sk_buff *skb)
3233 {
3234 skb_shinfo(skb)->frag_list = NULL;
3235 }
3236
3237 #define skb_walk_frags(skb, iter) \
3238 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3239
3240
3241 int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3242 const struct sk_buff *skb);
3243 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3244 struct sk_buff_head *queue,
3245 unsigned int flags,
3246 void (*destructor)(struct sock *sk,
3247 struct sk_buff *skb),
3248 int *peeked, int *off, int *err,
3249 struct sk_buff **last);
3250 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3251 void (*destructor)(struct sock *sk,
3252 struct sk_buff *skb),
3253 int *peeked, int *off, int *err,
3254 struct sk_buff **last);
3255 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3256 void (*destructor)(struct sock *sk,
3257 struct sk_buff *skb),
3258 int *peeked, int *off, int *err);
3259 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3260 int *err);
3261 unsigned int datagram_poll(struct file *file, struct socket *sock,
3262 struct poll_table_struct *wait);
3263 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3264 struct iov_iter *to, int size);
3265 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3266 struct msghdr *msg, int size)
3267 {
3268 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3269 }
3270 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3271 struct msghdr *msg);
3272 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3273 struct iov_iter *from, int len);
3274 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3275 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3276 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3277 static inline void skb_free_datagram_locked(struct sock *sk,
3278 struct sk_buff *skb)
3279 {
3280 __skb_free_datagram_locked(sk, skb, 0);
3281 }
3282 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3283 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3284 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3285 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3286 int len, __wsum csum);
3287 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3288 struct pipe_inode_info *pipe, unsigned int len,
3289 unsigned int flags);
3290 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3291 int len);
3292 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3293 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3294 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3295 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3296 int len, int hlen);
3297 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3298 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3299 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3300 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
3301 bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
3302 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3303 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3304 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3305 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3306 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3307 int skb_vlan_pop(struct sk_buff *skb);
3308 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3309 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3310 gfp_t gfp);
3311
3312 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3313 {
3314 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3315 }
3316
3317 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3318 {
3319 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3320 }
3321
3322 struct skb_checksum_ops {
3323 __wsum (*update)(const void *mem, int len, __wsum wsum);
3324 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3325 };
3326
3327 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3328
3329 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3330 __wsum csum, const struct skb_checksum_ops *ops);
3331 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3332 __wsum csum);
3333
3334 static inline void * __must_check
3335 __skb_header_pointer(const struct sk_buff *skb, int offset,
3336 int len, void *data, int hlen, void *buffer)
3337 {
3338 if (hlen - offset >= len)
3339 return data + offset;
3340
3341 if (!skb ||
3342 skb_copy_bits(skb, offset, buffer, len) < 0)
3343 return NULL;
3344
3345 return buffer;
3346 }
3347
3348 static inline void * __must_check
3349 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3350 {
3351 return __skb_header_pointer(skb, offset, len, skb->data,
3352 skb_headlen(skb), buffer);
3353 }
3354
3355 /**
3356 * skb_needs_linearize - check if we need to linearize a given skb
3357 * depending on the given device features.
3358 * @skb: socket buffer to check
3359 * @features: net device features
3360 *
3361 * Returns true if either:
3362 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3363 * 2. skb is fragmented and the device does not support SG.
3364 */
3365 static inline bool skb_needs_linearize(struct sk_buff *skb,
3366 netdev_features_t features)
3367 {
3368 return skb_is_nonlinear(skb) &&
3369 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3370 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3371 }
3372
3373 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3374 void *to,
3375 const unsigned int len)
3376 {
3377 memcpy(to, skb->data, len);
3378 }
3379
3380 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3381 const int offset, void *to,
3382 const unsigned int len)
3383 {
3384 memcpy(to, skb->data + offset, len);
3385 }
3386
3387 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3388 const void *from,
3389 const unsigned int len)
3390 {
3391 memcpy(skb->data, from, len);
3392 }
3393
3394 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3395 const int offset,
3396 const void *from,
3397 const unsigned int len)
3398 {
3399 memcpy(skb->data + offset, from, len);
3400 }
3401
3402 void skb_init(void);
3403
3404 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3405 {
3406 return skb->tstamp;
3407 }
3408
3409 /**
3410 * skb_get_timestamp - get timestamp from a skb
3411 * @skb: skb to get stamp from
3412 * @stamp: pointer to struct timeval to store stamp in
3413 *
3414 * Timestamps are stored in the skb as offsets to a base timestamp.
3415 * This function converts the offset back to a struct timeval and stores
3416 * it in stamp.
3417 */
3418 static inline void skb_get_timestamp(const struct sk_buff *skb,
3419 struct timeval *stamp)
3420 {
3421 *stamp = ktime_to_timeval(skb->tstamp);
3422 }
3423
3424 static inline void skb_get_timestampns(const struct sk_buff *skb,
3425 struct timespec *stamp)
3426 {
3427 *stamp = ktime_to_timespec(skb->tstamp);
3428 }
3429
3430 static inline void __net_timestamp(struct sk_buff *skb)
3431 {
3432 skb->tstamp = ktime_get_real();
3433 }
3434
3435 static inline ktime_t net_timedelta(ktime_t t)
3436 {
3437 return ktime_sub(ktime_get_real(), t);
3438 }
3439
3440 static inline ktime_t net_invalid_timestamp(void)
3441 {
3442 return 0;
3443 }
3444
3445 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3446 {
3447 return skb_shinfo(skb)->meta_len;
3448 }
3449
3450 static inline void *skb_metadata_end(const struct sk_buff *skb)
3451 {
3452 return skb_mac_header(skb);
3453 }
3454
3455 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3456 const struct sk_buff *skb_b,
3457 u8 meta_len)
3458 {
3459 const void *a = skb_metadata_end(skb_a);
3460 const void *b = skb_metadata_end(skb_b);
3461 /* Using more efficient varaiant than plain call to memcmp(). */
3462 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3463 u64 diffs = 0;
3464
3465 switch (meta_len) {
3466 #define __it(x, op) (x -= sizeof(u##op))
3467 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3468 case 32: diffs |= __it_diff(a, b, 64);
3469 case 24: diffs |= __it_diff(a, b, 64);
3470 case 16: diffs |= __it_diff(a, b, 64);
3471 case 8: diffs |= __it_diff(a, b, 64);
3472 break;
3473 case 28: diffs |= __it_diff(a, b, 64);
3474 case 20: diffs |= __it_diff(a, b, 64);
3475 case 12: diffs |= __it_diff(a, b, 64);
3476 case 4: diffs |= __it_diff(a, b, 32);
3477 break;
3478 }
3479 return diffs;
3480 #else
3481 return memcmp(a - meta_len, b - meta_len, meta_len);
3482 #endif
3483 }
3484
3485 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3486 const struct sk_buff *skb_b)
3487 {
3488 u8 len_a = skb_metadata_len(skb_a);
3489 u8 len_b = skb_metadata_len(skb_b);
3490
3491 if (!(len_a | len_b))
3492 return false;
3493
3494 return len_a != len_b ?
3495 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3496 }
3497
3498 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3499 {
3500 skb_shinfo(skb)->meta_len = meta_len;
3501 }
3502
3503 static inline void skb_metadata_clear(struct sk_buff *skb)
3504 {
3505 skb_metadata_set(skb, 0);
3506 }
3507
3508 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3509
3510 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3511
3512 void skb_clone_tx_timestamp(struct sk_buff *skb);
3513 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3514
3515 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3516
3517 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3518 {
3519 }
3520
3521 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3522 {
3523 return false;
3524 }
3525
3526 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3527
3528 /**
3529 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3530 *
3531 * PHY drivers may accept clones of transmitted packets for
3532 * timestamping via their phy_driver.txtstamp method. These drivers
3533 * must call this function to return the skb back to the stack with a
3534 * timestamp.
3535 *
3536 * @skb: clone of the the original outgoing packet
3537 * @hwtstamps: hardware time stamps
3538 *
3539 */
3540 void skb_complete_tx_timestamp(struct sk_buff *skb,
3541 struct skb_shared_hwtstamps *hwtstamps);
3542
3543 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3544 struct skb_shared_hwtstamps *hwtstamps,
3545 struct sock *sk, int tstype);
3546
3547 /**
3548 * skb_tstamp_tx - queue clone of skb with send time stamps
3549 * @orig_skb: the original outgoing packet
3550 * @hwtstamps: hardware time stamps, may be NULL if not available
3551 *
3552 * If the skb has a socket associated, then this function clones the
3553 * skb (thus sharing the actual data and optional structures), stores
3554 * the optional hardware time stamping information (if non NULL) or
3555 * generates a software time stamp (otherwise), then queues the clone
3556 * to the error queue of the socket. Errors are silently ignored.
3557 */
3558 void skb_tstamp_tx(struct sk_buff *orig_skb,
3559 struct skb_shared_hwtstamps *hwtstamps);
3560
3561 /**
3562 * skb_tx_timestamp() - Driver hook for transmit timestamping
3563 *
3564 * Ethernet MAC Drivers should call this function in their hard_xmit()
3565 * function immediately before giving the sk_buff to the MAC hardware.
3566 *
3567 * Specifically, one should make absolutely sure that this function is
3568 * called before TX completion of this packet can trigger. Otherwise
3569 * the packet could potentially already be freed.
3570 *
3571 * @skb: A socket buffer.
3572 */
3573 static inline void skb_tx_timestamp(struct sk_buff *skb)
3574 {
3575 skb_clone_tx_timestamp(skb);
3576 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3577 skb_tstamp_tx(skb, NULL);
3578 }
3579
3580 /**
3581 * skb_complete_wifi_ack - deliver skb with wifi status
3582 *
3583 * @skb: the original outgoing packet
3584 * @acked: ack status
3585 *
3586 */
3587 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3588
3589 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3590 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3591
3592 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3593 {
3594 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3595 skb->csum_valid ||
3596 (skb->ip_summed == CHECKSUM_PARTIAL &&
3597 skb_checksum_start_offset(skb) >= 0));
3598 }
3599
3600 /**
3601 * skb_checksum_complete - Calculate checksum of an entire packet
3602 * @skb: packet to process
3603 *
3604 * This function calculates the checksum over the entire packet plus
3605 * the value of skb->csum. The latter can be used to supply the
3606 * checksum of a pseudo header as used by TCP/UDP. It returns the
3607 * checksum.
3608 *
3609 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3610 * this function can be used to verify that checksum on received
3611 * packets. In that case the function should return zero if the
3612 * checksum is correct. In particular, this function will return zero
3613 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3614 * hardware has already verified the correctness of the checksum.
3615 */
3616 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3617 {
3618 return skb_csum_unnecessary(skb) ?
3619 0 : __skb_checksum_complete(skb);
3620 }
3621
3622 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3623 {
3624 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3625 if (skb->csum_level == 0)
3626 skb->ip_summed = CHECKSUM_NONE;
3627 else
3628 skb->csum_level--;
3629 }
3630 }
3631
3632 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3633 {
3634 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3635 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3636 skb->csum_level++;
3637 } else if (skb->ip_summed == CHECKSUM_NONE) {
3638 skb->ip_summed = CHECKSUM_UNNECESSARY;
3639 skb->csum_level = 0;
3640 }
3641 }
3642
3643 /* Check if we need to perform checksum complete validation.
3644 *
3645 * Returns true if checksum complete is needed, false otherwise
3646 * (either checksum is unnecessary or zero checksum is allowed).
3647 */
3648 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3649 bool zero_okay,
3650 __sum16 check)
3651 {
3652 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3653 skb->csum_valid = 1;
3654 __skb_decr_checksum_unnecessary(skb);
3655 return false;
3656 }
3657
3658 return true;
3659 }
3660
3661 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3662 * in checksum_init.
3663 */
3664 #define CHECKSUM_BREAK 76
3665
3666 /* Unset checksum-complete
3667 *
3668 * Unset checksum complete can be done when packet is being modified
3669 * (uncompressed for instance) and checksum-complete value is
3670 * invalidated.
3671 */
3672 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3673 {
3674 if (skb->ip_summed == CHECKSUM_COMPLETE)
3675 skb->ip_summed = CHECKSUM_NONE;
3676 }
3677
3678 /* Validate (init) checksum based on checksum complete.
3679 *
3680 * Return values:
3681 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3682 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3683 * checksum is stored in skb->csum for use in __skb_checksum_complete
3684 * non-zero: value of invalid checksum
3685 *
3686 */
3687 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3688 bool complete,
3689 __wsum psum)
3690 {
3691 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3692 if (!csum_fold(csum_add(psum, skb->csum))) {
3693 skb->csum_valid = 1;
3694 return 0;
3695 }
3696 }
3697
3698 skb->csum = psum;
3699
3700 if (complete || skb->len <= CHECKSUM_BREAK) {
3701 __sum16 csum;
3702
3703 csum = __skb_checksum_complete(skb);
3704 skb->csum_valid = !csum;
3705 return csum;
3706 }
3707
3708 return 0;
3709 }
3710
3711 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3712 {
3713 return 0;
3714 }
3715
3716 /* Perform checksum validate (init). Note that this is a macro since we only
3717 * want to calculate the pseudo header which is an input function if necessary.
3718 * First we try to validate without any computation (checksum unnecessary) and
3719 * then calculate based on checksum complete calling the function to compute
3720 * pseudo header.
3721 *
3722 * Return values:
3723 * 0: checksum is validated or try to in skb_checksum_complete
3724 * non-zero: value of invalid checksum
3725 */
3726 #define __skb_checksum_validate(skb, proto, complete, \
3727 zero_okay, check, compute_pseudo) \
3728 ({ \
3729 __sum16 __ret = 0; \
3730 skb->csum_valid = 0; \
3731 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3732 __ret = __skb_checksum_validate_complete(skb, \
3733 complete, compute_pseudo(skb, proto)); \
3734 __ret; \
3735 })
3736
3737 #define skb_checksum_init(skb, proto, compute_pseudo) \
3738 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3739
3740 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3741 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3742
3743 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3744 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3745
3746 #define skb_checksum_validate_zero_check(skb, proto, check, \
3747 compute_pseudo) \
3748 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3749
3750 #define skb_checksum_simple_validate(skb) \
3751 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3752
3753 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3754 {
3755 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
3756 }
3757
3758 static inline void __skb_checksum_convert(struct sk_buff *skb,
3759 __sum16 check, __wsum pseudo)
3760 {
3761 skb->csum = ~pseudo;
3762 skb->ip_summed = CHECKSUM_COMPLETE;
3763 }
3764
3765 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3766 do { \
3767 if (__skb_checksum_convert_check(skb)) \
3768 __skb_checksum_convert(skb, check, \
3769 compute_pseudo(skb, proto)); \
3770 } while (0)
3771
3772 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3773 u16 start, u16 offset)
3774 {
3775 skb->ip_summed = CHECKSUM_PARTIAL;
3776 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3777 skb->csum_offset = offset - start;
3778 }
3779
3780 /* Update skbuf and packet to reflect the remote checksum offload operation.
3781 * When called, ptr indicates the starting point for skb->csum when
3782 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3783 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3784 */
3785 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3786 int start, int offset, bool nopartial)
3787 {
3788 __wsum delta;
3789
3790 if (!nopartial) {
3791 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3792 return;
3793 }
3794
3795 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3796 __skb_checksum_complete(skb);
3797 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3798 }
3799
3800 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3801
3802 /* Adjust skb->csum since we changed the packet */
3803 skb->csum = csum_add(skb->csum, delta);
3804 }
3805
3806 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3807 {
3808 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3809 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
3810 #else
3811 return NULL;
3812 #endif
3813 }
3814
3815 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3816 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3817 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3818 {
3819 if (nfct && atomic_dec_and_test(&nfct->use))
3820 nf_conntrack_destroy(nfct);
3821 }
3822 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3823 {
3824 if (nfct)
3825 atomic_inc(&nfct->use);
3826 }
3827 #endif
3828 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3829 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3830 {
3831 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
3832 kfree(nf_bridge);
3833 }
3834 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3835 {
3836 if (nf_bridge)
3837 refcount_inc(&nf_bridge->use);
3838 }
3839 #endif /* CONFIG_BRIDGE_NETFILTER */
3840 static inline void nf_reset(struct sk_buff *skb)
3841 {
3842 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3843 nf_conntrack_put(skb_nfct(skb));
3844 skb->_nfct = 0;
3845 #endif
3846 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3847 nf_bridge_put(skb->nf_bridge);
3848 skb->nf_bridge = NULL;
3849 #endif
3850 }
3851
3852 static inline void nf_reset_trace(struct sk_buff *skb)
3853 {
3854 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3855 skb->nf_trace = 0;
3856 #endif
3857 }
3858
3859 static inline void ipvs_reset(struct sk_buff *skb)
3860 {
3861 #if IS_ENABLED(CONFIG_IP_VS)
3862 skb->ipvs_property = 0;
3863 #endif
3864 }
3865
3866 /* Note: This doesn't put any conntrack and bridge info in dst. */
3867 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3868 bool copy)
3869 {
3870 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3871 dst->_nfct = src->_nfct;
3872 nf_conntrack_get(skb_nfct(src));
3873 #endif
3874 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3875 dst->nf_bridge = src->nf_bridge;
3876 nf_bridge_get(src->nf_bridge);
3877 #endif
3878 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3879 if (copy)
3880 dst->nf_trace = src->nf_trace;
3881 #endif
3882 }
3883
3884 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3885 {
3886 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3887 nf_conntrack_put(skb_nfct(dst));
3888 #endif
3889 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3890 nf_bridge_put(dst->nf_bridge);
3891 #endif
3892 __nf_copy(dst, src, true);
3893 }
3894
3895 #ifdef CONFIG_NETWORK_SECMARK
3896 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3897 {
3898 to->secmark = from->secmark;
3899 }
3900
3901 static inline void skb_init_secmark(struct sk_buff *skb)
3902 {
3903 skb->secmark = 0;
3904 }
3905 #else
3906 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3907 { }
3908
3909 static inline void skb_init_secmark(struct sk_buff *skb)
3910 { }
3911 #endif
3912
3913 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3914 {
3915 return !skb->destructor &&
3916 #if IS_ENABLED(CONFIG_XFRM)
3917 !skb->sp &&
3918 #endif
3919 !skb_nfct(skb) &&
3920 !skb->_skb_refdst &&
3921 !skb_has_frag_list(skb);
3922 }
3923
3924 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3925 {
3926 skb->queue_mapping = queue_mapping;
3927 }
3928
3929 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3930 {
3931 return skb->queue_mapping;
3932 }
3933
3934 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3935 {
3936 to->queue_mapping = from->queue_mapping;
3937 }
3938
3939 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3940 {
3941 skb->queue_mapping = rx_queue + 1;
3942 }
3943
3944 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3945 {
3946 return skb->queue_mapping - 1;
3947 }
3948
3949 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3950 {
3951 return skb->queue_mapping != 0;
3952 }
3953
3954 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3955 {
3956 skb->dst_pending_confirm = val;
3957 }
3958
3959 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3960 {
3961 return skb->dst_pending_confirm != 0;
3962 }
3963
3964 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3965 {
3966 #ifdef CONFIG_XFRM
3967 return skb->sp;
3968 #else
3969 return NULL;
3970 #endif
3971 }
3972
3973 /* Keeps track of mac header offset relative to skb->head.
3974 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3975 * For non-tunnel skb it points to skb_mac_header() and for
3976 * tunnel skb it points to outer mac header.
3977 * Keeps track of level of encapsulation of network headers.
3978 */
3979 struct skb_gso_cb {
3980 union {
3981 int mac_offset;
3982 int data_offset;
3983 };
3984 int encap_level;
3985 __wsum csum;
3986 __u16 csum_start;
3987 };
3988 #define SKB_SGO_CB_OFFSET 32
3989 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
3990
3991 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3992 {
3993 return (skb_mac_header(inner_skb) - inner_skb->head) -
3994 SKB_GSO_CB(inner_skb)->mac_offset;
3995 }
3996
3997 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3998 {
3999 int new_headroom, headroom;
4000 int ret;
4001
4002 headroom = skb_headroom(skb);
4003 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4004 if (ret)
4005 return ret;
4006
4007 new_headroom = skb_headroom(skb);
4008 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4009 return 0;
4010 }
4011
4012 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4013 {
4014 /* Do not update partial checksums if remote checksum is enabled. */
4015 if (skb->remcsum_offload)
4016 return;
4017
4018 SKB_GSO_CB(skb)->csum = res;
4019 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4020 }
4021
4022 /* Compute the checksum for a gso segment. First compute the checksum value
4023 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4024 * then add in skb->csum (checksum from csum_start to end of packet).
4025 * skb->csum and csum_start are then updated to reflect the checksum of the
4026 * resultant packet starting from the transport header-- the resultant checksum
4027 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4028 * header.
4029 */
4030 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4031 {
4032 unsigned char *csum_start = skb_transport_header(skb);
4033 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4034 __wsum partial = SKB_GSO_CB(skb)->csum;
4035
4036 SKB_GSO_CB(skb)->csum = res;
4037 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4038
4039 return csum_fold(csum_partial(csum_start, plen, partial));
4040 }
4041
4042 static inline bool skb_is_gso(const struct sk_buff *skb)
4043 {
4044 return skb_shinfo(skb)->gso_size;
4045 }
4046
4047 /* Note: Should be called only if skb_is_gso(skb) is true */
4048 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4049 {
4050 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4051 }
4052
4053 static inline void skb_gso_reset(struct sk_buff *skb)
4054 {
4055 skb_shinfo(skb)->gso_size = 0;
4056 skb_shinfo(skb)->gso_segs = 0;
4057 skb_shinfo(skb)->gso_type = 0;
4058 }
4059
4060 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4061
4062 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4063 {
4064 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4065 * wanted then gso_type will be set. */
4066 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4067
4068 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4069 unlikely(shinfo->gso_type == 0)) {
4070 __skb_warn_lro_forwarding(skb);
4071 return true;
4072 }
4073 return false;
4074 }
4075
4076 static inline void skb_forward_csum(struct sk_buff *skb)
4077 {
4078 /* Unfortunately we don't support this one. Any brave souls? */
4079 if (skb->ip_summed == CHECKSUM_COMPLETE)
4080 skb->ip_summed = CHECKSUM_NONE;
4081 }
4082
4083 /**
4084 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4085 * @skb: skb to check
4086 *
4087 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4088 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4089 * use this helper, to document places where we make this assertion.
4090 */
4091 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4092 {
4093 #ifdef DEBUG
4094 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4095 #endif
4096 }
4097
4098 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4099
4100 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4101 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4102 unsigned int transport_len,
4103 __sum16(*skb_chkf)(struct sk_buff *skb));
4104
4105 /**
4106 * skb_head_is_locked - Determine if the skb->head is locked down
4107 * @skb: skb to check
4108 *
4109 * The head on skbs build around a head frag can be removed if they are
4110 * not cloned. This function returns true if the skb head is locked down
4111 * due to either being allocated via kmalloc, or by being a clone with
4112 * multiple references to the head.
4113 */
4114 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4115 {
4116 return !skb->head_frag || skb_cloned(skb);
4117 }
4118
4119 /**
4120 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4121 *
4122 * @skb: GSO skb
4123 *
4124 * skb_gso_network_seglen is used to determine the real size of the
4125 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4126 *
4127 * The MAC/L2 header is not accounted for.
4128 */
4129 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
4130 {
4131 unsigned int hdr_len = skb_transport_header(skb) -
4132 skb_network_header(skb);
4133 return hdr_len + skb_gso_transport_seglen(skb);
4134 }
4135
4136 /**
4137 * skb_gso_mac_seglen - Return length of individual segments of a gso packet
4138 *
4139 * @skb: GSO skb
4140 *
4141 * skb_gso_mac_seglen is used to determine the real size of the
4142 * individual segments, including MAC/L2, Layer3 (IP, IPv6) and L4
4143 * headers (TCP/UDP).
4144 */
4145 static inline unsigned int skb_gso_mac_seglen(const struct sk_buff *skb)
4146 {
4147 unsigned int hdr_len = skb_transport_header(skb) - skb_mac_header(skb);
4148 return hdr_len + skb_gso_transport_seglen(skb);
4149 }
4150
4151 /* Local Checksum Offload.
4152 * Compute outer checksum based on the assumption that the
4153 * inner checksum will be offloaded later.
4154 * See Documentation/networking/checksum-offloads.txt for
4155 * explanation of how this works.
4156 * Fill in outer checksum adjustment (e.g. with sum of outer
4157 * pseudo-header) before calling.
4158 * Also ensure that inner checksum is in linear data area.
4159 */
4160 static inline __wsum lco_csum(struct sk_buff *skb)
4161 {
4162 unsigned char *csum_start = skb_checksum_start(skb);
4163 unsigned char *l4_hdr = skb_transport_header(skb);
4164 __wsum partial;
4165
4166 /* Start with complement of inner checksum adjustment */
4167 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4168 skb->csum_offset));
4169
4170 /* Add in checksum of our headers (incl. outer checksum
4171 * adjustment filled in by caller) and return result.
4172 */
4173 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4174 }
4175
4176 #endif /* __KERNEL__ */
4177 #endif /* _LINUX_SKBUFF_H */