1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 */
9
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
12
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
22
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
39 #include <net/flow.h>
40 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
41 #include <linux/netfilter/nf_conntrack_common.h>
42 #endif
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 struct bpf_prog;
247 union bpf_attr;
248 struct skb_ext;
249
250 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
251 struct nf_bridge_info {
252 enum {
253 BRNF_PROTO_UNCHANGED,
254 BRNF_PROTO_8021Q,
255 BRNF_PROTO_PPPOE
256 } orig_proto:8;
257 u8 pkt_otherhost:1;
258 u8 in_prerouting:1;
259 u8 bridged_dnat:1;
260 __u16 frag_max_size;
261 struct net_device *physindev;
262
263 /* always valid & non-NULL from FORWARD on, for physdev match */
264 struct net_device *physoutdev;
265 union {
266 /* prerouting: detect dnat in orig/reply direction */
267 __be32 ipv4_daddr;
268 struct in6_addr ipv6_daddr;
269
270 /* after prerouting + nat detected: store original source
271 * mac since neigh resolution overwrites it, only used while
272 * skb is out in neigh layer.
273 */
274 char neigh_header[8];
275 };
276 };
277 #endif
278
279 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
280 /* Chain in tc_skb_ext will be used to share the tc chain with
281 * ovs recirc_id. It will be set to the current chain by tc
282 * and read by ovs to recirc_id.
283 */
284 struct tc_skb_ext {
285 __u32 chain;
286 };
287 #endif
288
289 struct sk_buff_head {
290 /* These two members must be first. */
291 struct sk_buff *next;
292 struct sk_buff *prev;
293
294 __u32 qlen;
295 spinlock_t lock;
296 };
297
298 struct sk_buff;
299
300 /* To allow 64K frame to be packed as single skb without frag_list we
301 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
302 * buffers which do not start on a page boundary.
303 *
304 * Since GRO uses frags we allocate at least 16 regardless of page
305 * size.
306 */
307 #if (65536/PAGE_SIZE + 1) < 16
308 #define MAX_SKB_FRAGS 16UL
309 #else
310 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
311 #endif
312 extern int sysctl_max_skb_frags;
313
314 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
315 * segment using its current segmentation instead.
316 */
317 #define GSO_BY_FRAGS 0xFFFF
318
319 typedef struct bio_vec skb_frag_t;
320
321 /**
322 * skb_frag_size() - Returns the size of a skb fragment
323 * @frag: skb fragment
324 */
skb_frag_size(const skb_frag_t * frag)325 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326 {
327 return frag->bv_len;
328 }
329
330 /**
331 * skb_frag_size_set() - Sets the size of a skb fragment
332 * @frag: skb fragment
333 * @size: size of fragment
334 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)335 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
336 {
337 frag->bv_len = size;
338 }
339
340 /**
341 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
342 * @frag: skb fragment
343 * @delta: value to add
344 */
skb_frag_size_add(skb_frag_t * frag,int delta)345 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
346 {
347 frag->bv_len += delta;
348 }
349
350 /**
351 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
352 * @frag: skb fragment
353 * @delta: value to subtract
354 */
skb_frag_size_sub(skb_frag_t * frag,int delta)355 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
356 {
357 frag->bv_len -= delta;
358 }
359
360 /**
361 * skb_frag_must_loop - Test if %p is a high memory page
362 * @p: fragment's page
363 */
skb_frag_must_loop(struct page * p)364 static inline bool skb_frag_must_loop(struct page *p)
365 {
366 #if defined(CONFIG_HIGHMEM)
367 if (PageHighMem(p))
368 return true;
369 #endif
370 return false;
371 }
372
373 /**
374 * skb_frag_foreach_page - loop over pages in a fragment
375 *
376 * @f: skb frag to operate on
377 * @f_off: offset from start of f->bv_page
378 * @f_len: length from f_off to loop over
379 * @p: (temp var) current page
380 * @p_off: (temp var) offset from start of current page,
381 * non-zero only on first page.
382 * @p_len: (temp var) length in current page,
383 * < PAGE_SIZE only on first and last page.
384 * @copied: (temp var) length so far, excluding current p_len.
385 *
386 * A fragment can hold a compound page, in which case per-page
387 * operations, notably kmap_atomic, must be called for each
388 * regular page.
389 */
390 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
391 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
392 p_off = (f_off) & (PAGE_SIZE - 1), \
393 p_len = skb_frag_must_loop(p) ? \
394 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
395 copied = 0; \
396 copied < f_len; \
397 copied += p_len, p++, p_off = 0, \
398 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
399
400 #define HAVE_HW_TIME_STAMP
401
402 /**
403 * struct skb_shared_hwtstamps - hardware time stamps
404 * @hwtstamp: hardware time stamp transformed into duration
405 * since arbitrary point in time
406 *
407 * Software time stamps generated by ktime_get_real() are stored in
408 * skb->tstamp.
409 *
410 * hwtstamps can only be compared against other hwtstamps from
411 * the same device.
412 *
413 * This structure is attached to packets as part of the
414 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
415 */
416 struct skb_shared_hwtstamps {
417 ktime_t hwtstamp;
418 };
419
420 /* Definitions for tx_flags in struct skb_shared_info */
421 enum {
422 /* generate hardware time stamp */
423 SKBTX_HW_TSTAMP = 1 << 0,
424
425 /* generate software time stamp when queueing packet to NIC */
426 SKBTX_SW_TSTAMP = 1 << 1,
427
428 /* device driver is going to provide hardware time stamp */
429 SKBTX_IN_PROGRESS = 1 << 2,
430
431 /* device driver supports TX zero-copy buffers */
432 SKBTX_DEV_ZEROCOPY = 1 << 3,
433
434 /* generate wifi status information (where possible) */
435 SKBTX_WIFI_STATUS = 1 << 4,
436
437 /* This indicates at least one fragment might be overwritten
438 * (as in vmsplice(), sendfile() ...)
439 * If we need to compute a TX checksum, we'll need to copy
440 * all frags to avoid possible bad checksum
441 */
442 SKBTX_SHARED_FRAG = 1 << 5,
443
444 /* generate software time stamp when entering packet scheduling */
445 SKBTX_SCHED_TSTAMP = 1 << 6,
446 };
447
448 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
449 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
450 SKBTX_SCHED_TSTAMP)
451 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
452
453 /*
454 * The callback notifies userspace to release buffers when skb DMA is done in
455 * lower device, the skb last reference should be 0 when calling this.
456 * The zerocopy_success argument is true if zero copy transmit occurred,
457 * false on data copy or out of memory error caused by data copy attempt.
458 * The ctx field is used to track device context.
459 * The desc field is used to track userspace buffer index.
460 */
461 struct ubuf_info {
462 void (*callback)(struct ubuf_info *, bool zerocopy_success);
463 union {
464 struct {
465 unsigned long desc;
466 void *ctx;
467 };
468 struct {
469 u32 id;
470 u16 len;
471 u16 zerocopy:1;
472 u32 bytelen;
473 };
474 };
475 refcount_t refcnt;
476
477 struct mmpin {
478 struct user_struct *user;
479 unsigned int num_pg;
480 } mmp;
481 };
482
483 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
484
485 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
486 void mm_unaccount_pinned_pages(struct mmpin *mmp);
487
488 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
489 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
490 struct ubuf_info *uarg);
491
sock_zerocopy_get(struct ubuf_info * uarg)492 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
493 {
494 refcount_inc(&uarg->refcnt);
495 }
496
497 void sock_zerocopy_put(struct ubuf_info *uarg);
498 void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
499
500 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
501
502 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
503 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
504 struct msghdr *msg, int len,
505 struct ubuf_info *uarg);
506
507 /* This data is invariant across clones and lives at
508 * the end of the header data, ie. at skb->end.
509 */
510 struct skb_shared_info {
511 __u8 __unused;
512 __u8 meta_len;
513 __u8 nr_frags;
514 __u8 tx_flags;
515 unsigned short gso_size;
516 /* Warning: this field is not always filled in (UFO)! */
517 unsigned short gso_segs;
518 struct sk_buff *frag_list;
519 struct skb_shared_hwtstamps hwtstamps;
520 unsigned int gso_type;
521 u32 tskey;
522
523 /*
524 * Warning : all fields before dataref are cleared in __alloc_skb()
525 */
526 atomic_t dataref;
527
528 /* Intermediate layers must ensure that destructor_arg
529 * remains valid until skb destructor */
530 void * destructor_arg;
531
532 /* must be last field, see pskb_expand_head() */
533 skb_frag_t frags[MAX_SKB_FRAGS];
534 };
535
536 /* We divide dataref into two halves. The higher 16 bits hold references
537 * to the payload part of skb->data. The lower 16 bits hold references to
538 * the entire skb->data. A clone of a headerless skb holds the length of
539 * the header in skb->hdr_len.
540 *
541 * All users must obey the rule that the skb->data reference count must be
542 * greater than or equal to the payload reference count.
543 *
544 * Holding a reference to the payload part means that the user does not
545 * care about modifications to the header part of skb->data.
546 */
547 #define SKB_DATAREF_SHIFT 16
548 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
549
550
551 enum {
552 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
553 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
554 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
555 };
556
557 enum {
558 SKB_GSO_TCPV4 = 1 << 0,
559
560 /* This indicates the skb is from an untrusted source. */
561 SKB_GSO_DODGY = 1 << 1,
562
563 /* This indicates the tcp segment has CWR set. */
564 SKB_GSO_TCP_ECN = 1 << 2,
565
566 SKB_GSO_TCP_FIXEDID = 1 << 3,
567
568 SKB_GSO_TCPV6 = 1 << 4,
569
570 SKB_GSO_FCOE = 1 << 5,
571
572 SKB_GSO_GRE = 1 << 6,
573
574 SKB_GSO_GRE_CSUM = 1 << 7,
575
576 SKB_GSO_IPXIP4 = 1 << 8,
577
578 SKB_GSO_IPXIP6 = 1 << 9,
579
580 SKB_GSO_UDP_TUNNEL = 1 << 10,
581
582 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
583
584 SKB_GSO_PARTIAL = 1 << 12,
585
586 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
587
588 SKB_GSO_SCTP = 1 << 14,
589
590 SKB_GSO_ESP = 1 << 15,
591
592 SKB_GSO_UDP = 1 << 16,
593
594 SKB_GSO_UDP_L4 = 1 << 17,
595 };
596
597 #if BITS_PER_LONG > 32
598 #define NET_SKBUFF_DATA_USES_OFFSET 1
599 #endif
600
601 #ifdef NET_SKBUFF_DATA_USES_OFFSET
602 typedef unsigned int sk_buff_data_t;
603 #else
604 typedef unsigned char *sk_buff_data_t;
605 #endif
606
607 /**
608 * struct sk_buff - socket buffer
609 * @next: Next buffer in list
610 * @prev: Previous buffer in list
611 * @tstamp: Time we arrived/left
612 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
613 * @sk: Socket we are owned by
614 * @dev: Device we arrived on/are leaving by
615 * @cb: Control buffer. Free for use by every layer. Put private vars here
616 * @_skb_refdst: destination entry (with norefcount bit)
617 * @sp: the security path, used for xfrm
618 * @len: Length of actual data
619 * @data_len: Data length
620 * @mac_len: Length of link layer header
621 * @hdr_len: writable header length of cloned skb
622 * @csum: Checksum (must include start/offset pair)
623 * @csum_start: Offset from skb->head where checksumming should start
624 * @csum_offset: Offset from csum_start where checksum should be stored
625 * @priority: Packet queueing priority
626 * @ignore_df: allow local fragmentation
627 * @cloned: Head may be cloned (check refcnt to be sure)
628 * @ip_summed: Driver fed us an IP checksum
629 * @nohdr: Payload reference only, must not modify header
630 * @pkt_type: Packet class
631 * @fclone: skbuff clone status
632 * @ipvs_property: skbuff is owned by ipvs
633 * @offload_fwd_mark: Packet was L2-forwarded in hardware
634 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
635 * @tc_skip_classify: do not classify packet. set by IFB device
636 * @tc_at_ingress: used within tc_classify to distinguish in/egress
637 * @tc_redirected: packet was redirected by a tc action
638 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
639 * @peeked: this packet has been seen already, so stats have been
640 * done for it, don't do them again
641 * @nf_trace: netfilter packet trace flag
642 * @protocol: Packet protocol from driver
643 * @destructor: Destruct function
644 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
645 * @_nfct: Associated connection, if any (with nfctinfo bits)
646 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
647 * @skb_iif: ifindex of device we arrived on
648 * @tc_index: Traffic control index
649 * @hash: the packet hash
650 * @queue_mapping: Queue mapping for multiqueue devices
651 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
652 * @active_extensions: active extensions (skb_ext_id types)
653 * @ndisc_nodetype: router type (from link layer)
654 * @ooo_okay: allow the mapping of a socket to a queue to be changed
655 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
656 * ports.
657 * @sw_hash: indicates hash was computed in software stack
658 * @wifi_acked_valid: wifi_acked was set
659 * @wifi_acked: whether frame was acked on wifi or not
660 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
661 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
662 * @dst_pending_confirm: need to confirm neighbour
663 * @decrypted: Decrypted SKB
664 * @napi_id: id of the NAPI struct this skb came from
665 * @secmark: security marking
666 * @mark: Generic packet mark
667 * @vlan_proto: vlan encapsulation protocol
668 * @vlan_tci: vlan tag control information
669 * @inner_protocol: Protocol (encapsulation)
670 * @inner_transport_header: Inner transport layer header (encapsulation)
671 * @inner_network_header: Network layer header (encapsulation)
672 * @inner_mac_header: Link layer header (encapsulation)
673 * @transport_header: Transport layer header
674 * @network_header: Network layer header
675 * @mac_header: Link layer header
676 * @tail: Tail pointer
677 * @end: End pointer
678 * @head: Head of buffer
679 * @data: Data head pointer
680 * @truesize: Buffer size
681 * @users: User count - see {datagram,tcp}.c
682 * @extensions: allocated extensions, valid if active_extensions is nonzero
683 */
684
685 struct sk_buff {
686 union {
687 struct {
688 /* These two members must be first. */
689 struct sk_buff *next;
690 struct sk_buff *prev;
691
692 union {
693 struct net_device *dev;
694 /* Some protocols might use this space to store information,
695 * while device pointer would be NULL.
696 * UDP receive path is one user.
697 */
698 unsigned long dev_scratch;
699 };
700 };
701 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
702 struct list_head list;
703 };
704
705 union {
706 struct sock *sk;
707 int ip_defrag_offset;
708 };
709
710 union {
711 ktime_t tstamp;
712 u64 skb_mstamp_ns; /* earliest departure time */
713 };
714 /*
715 * This is the control buffer. It is free to use for every
716 * layer. Please put your private variables there. If you
717 * want to keep them across layers you have to do a skb_clone()
718 * first. This is owned by whoever has the skb queued ATM.
719 */
720 char cb[48] __aligned(8);
721
722 union {
723 struct {
724 unsigned long _skb_refdst;
725 void (*destructor)(struct sk_buff *skb);
726 };
727 struct list_head tcp_tsorted_anchor;
728 };
729
730 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
731 unsigned long _nfct;
732 #endif
733 unsigned int len,
734 data_len;
735 __u16 mac_len,
736 hdr_len;
737
738 /* Following fields are _not_ copied in __copy_skb_header()
739 * Note that queue_mapping is here mostly to fill a hole.
740 */
741 __u16 queue_mapping;
742
743 /* if you move cloned around you also must adapt those constants */
744 #ifdef __BIG_ENDIAN_BITFIELD
745 #define CLONED_MASK (1 << 7)
746 #else
747 #define CLONED_MASK 1
748 #endif
749 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
750
751 __u8 __cloned_offset[0];
752 __u8 cloned:1,
753 nohdr:1,
754 fclone:2,
755 peeked:1,
756 head_frag:1,
757 pfmemalloc:1;
758 #ifdef CONFIG_SKB_EXTENSIONS
759 __u8 active_extensions;
760 #endif
761 /* fields enclosed in headers_start/headers_end are copied
762 * using a single memcpy() in __copy_skb_header()
763 */
764 /* private: */
765 __u32 headers_start[0];
766 /* public: */
767
768 /* if you move pkt_type around you also must adapt those constants */
769 #ifdef __BIG_ENDIAN_BITFIELD
770 #define PKT_TYPE_MAX (7 << 5)
771 #else
772 #define PKT_TYPE_MAX 7
773 #endif
774 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
775
776 __u8 __pkt_type_offset[0];
777 __u8 pkt_type:3;
778 __u8 ignore_df:1;
779 __u8 nf_trace:1;
780 __u8 ip_summed:2;
781 __u8 ooo_okay:1;
782
783 __u8 l4_hash:1;
784 __u8 sw_hash:1;
785 __u8 wifi_acked_valid:1;
786 __u8 wifi_acked:1;
787 __u8 no_fcs:1;
788 /* Indicates the inner headers are valid in the skbuff. */
789 __u8 encapsulation:1;
790 __u8 encap_hdr_csum:1;
791 __u8 csum_valid:1;
792
793 #ifdef __BIG_ENDIAN_BITFIELD
794 #define PKT_VLAN_PRESENT_BIT 7
795 #else
796 #define PKT_VLAN_PRESENT_BIT 0
797 #endif
798 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
799 __u8 __pkt_vlan_present_offset[0];
800 __u8 vlan_present:1;
801 __u8 csum_complete_sw:1;
802 __u8 csum_level:2;
803 __u8 csum_not_inet:1;
804 __u8 dst_pending_confirm:1;
805 #ifdef CONFIG_IPV6_NDISC_NODETYPE
806 __u8 ndisc_nodetype:2;
807 #endif
808
809 __u8 ipvs_property:1;
810 __u8 inner_protocol_type:1;
811 __u8 remcsum_offload:1;
812 #ifdef CONFIG_NET_SWITCHDEV
813 __u8 offload_fwd_mark:1;
814 __u8 offload_l3_fwd_mark:1;
815 #endif
816 #ifdef CONFIG_NET_CLS_ACT
817 __u8 tc_skip_classify:1;
818 __u8 tc_at_ingress:1;
819 __u8 tc_redirected:1;
820 __u8 tc_from_ingress:1;
821 #endif
822 #ifdef CONFIG_TLS_DEVICE
823 __u8 decrypted:1;
824 #endif
825
826 #ifdef CONFIG_NET_SCHED
827 __u16 tc_index; /* traffic control index */
828 #endif
829
830 union {
831 __wsum csum;
832 struct {
833 __u16 csum_start;
834 __u16 csum_offset;
835 };
836 };
837 __u32 priority;
838 int skb_iif;
839 __u32 hash;
840 __be16 vlan_proto;
841 __u16 vlan_tci;
842 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
843 union {
844 unsigned int napi_id;
845 unsigned int sender_cpu;
846 };
847 #endif
848 #ifdef CONFIG_NETWORK_SECMARK
849 __u32 secmark;
850 #endif
851
852 union {
853 __u32 mark;
854 __u32 reserved_tailroom;
855 };
856
857 union {
858 __be16 inner_protocol;
859 __u8 inner_ipproto;
860 };
861
862 __u16 inner_transport_header;
863 __u16 inner_network_header;
864 __u16 inner_mac_header;
865
866 __be16 protocol;
867 __u16 transport_header;
868 __u16 network_header;
869 __u16 mac_header;
870
871 /* private: */
872 __u32 headers_end[0];
873 /* public: */
874
875 /* These elements must be at the end, see alloc_skb() for details. */
876 sk_buff_data_t tail;
877 sk_buff_data_t end;
878 unsigned char *head,
879 *data;
880 unsigned int truesize;
881 refcount_t users;
882
883 #ifdef CONFIG_SKB_EXTENSIONS
884 /* only useable after checking ->active_extensions != 0 */
885 struct skb_ext *extensions;
886 #endif
887 };
888
889 #ifdef __KERNEL__
890 /*
891 * Handling routines are only of interest to the kernel
892 */
893
894 #define SKB_ALLOC_FCLONE 0x01
895 #define SKB_ALLOC_RX 0x02
896 #define SKB_ALLOC_NAPI 0x04
897
898 /**
899 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
900 * @skb: buffer
901 */
skb_pfmemalloc(const struct sk_buff * skb)902 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
903 {
904 return unlikely(skb->pfmemalloc);
905 }
906
907 /*
908 * skb might have a dst pointer attached, refcounted or not.
909 * _skb_refdst low order bit is set if refcount was _not_ taken
910 */
911 #define SKB_DST_NOREF 1UL
912 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
913
914 /**
915 * skb_dst - returns skb dst_entry
916 * @skb: buffer
917 *
918 * Returns skb dst_entry, regardless of reference taken or not.
919 */
skb_dst(const struct sk_buff * skb)920 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
921 {
922 /* If refdst was not refcounted, check we still are in a
923 * rcu_read_lock section
924 */
925 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
926 !rcu_read_lock_held() &&
927 !rcu_read_lock_bh_held());
928 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
929 }
930
931 /**
932 * skb_dst_set - sets skb dst
933 * @skb: buffer
934 * @dst: dst entry
935 *
936 * Sets skb dst, assuming a reference was taken on dst and should
937 * be released by skb_dst_drop()
938 */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)939 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
940 {
941 skb->_skb_refdst = (unsigned long)dst;
942 }
943
944 /**
945 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
946 * @skb: buffer
947 * @dst: dst entry
948 *
949 * Sets skb dst, assuming a reference was not taken on dst.
950 * If dst entry is cached, we do not take reference and dst_release
951 * will be avoided by refdst_drop. If dst entry is not cached, we take
952 * reference, so that last dst_release can destroy the dst immediately.
953 */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)954 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
955 {
956 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
957 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
958 }
959
960 /**
961 * skb_dst_is_noref - Test if skb dst isn't refcounted
962 * @skb: buffer
963 */
skb_dst_is_noref(const struct sk_buff * skb)964 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
965 {
966 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
967 }
968
969 /**
970 * skb_rtable - Returns the skb &rtable
971 * @skb: buffer
972 */
skb_rtable(const struct sk_buff * skb)973 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
974 {
975 return (struct rtable *)skb_dst(skb);
976 }
977
978 /* For mangling skb->pkt_type from user space side from applications
979 * such as nft, tc, etc, we only allow a conservative subset of
980 * possible pkt_types to be set.
981 */
skb_pkt_type_ok(u32 ptype)982 static inline bool skb_pkt_type_ok(u32 ptype)
983 {
984 return ptype <= PACKET_OTHERHOST;
985 }
986
987 /**
988 * skb_napi_id - Returns the skb's NAPI id
989 * @skb: buffer
990 */
skb_napi_id(const struct sk_buff * skb)991 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
992 {
993 #ifdef CONFIG_NET_RX_BUSY_POLL
994 return skb->napi_id;
995 #else
996 return 0;
997 #endif
998 }
999
1000 /**
1001 * skb_unref - decrement the skb's reference count
1002 * @skb: buffer
1003 *
1004 * Returns true if we can free the skb.
1005 */
skb_unref(struct sk_buff * skb)1006 static inline bool skb_unref(struct sk_buff *skb)
1007 {
1008 if (unlikely(!skb))
1009 return false;
1010 if (likely(refcount_read(&skb->users) == 1))
1011 smp_rmb();
1012 else if (likely(!refcount_dec_and_test(&skb->users)))
1013 return false;
1014
1015 return true;
1016 }
1017
1018 void skb_release_head_state(struct sk_buff *skb);
1019 void kfree_skb(struct sk_buff *skb);
1020 void kfree_skb_list(struct sk_buff *segs);
1021 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1022 void skb_tx_error(struct sk_buff *skb);
1023 void consume_skb(struct sk_buff *skb);
1024 void __consume_stateless_skb(struct sk_buff *skb);
1025 void __kfree_skb(struct sk_buff *skb);
1026 extern struct kmem_cache *skbuff_head_cache;
1027
1028 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1029 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1030 bool *fragstolen, int *delta_truesize);
1031
1032 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1033 int node);
1034 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1035 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1036 struct sk_buff *build_skb_around(struct sk_buff *skb,
1037 void *data, unsigned int frag_size);
1038
1039 /**
1040 * alloc_skb - allocate a network buffer
1041 * @size: size to allocate
1042 * @priority: allocation mask
1043 *
1044 * This function is a convenient wrapper around __alloc_skb().
1045 */
alloc_skb(unsigned int size,gfp_t priority)1046 static inline struct sk_buff *alloc_skb(unsigned int size,
1047 gfp_t priority)
1048 {
1049 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1050 }
1051
1052 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1053 unsigned long data_len,
1054 int max_page_order,
1055 int *errcode,
1056 gfp_t gfp_mask);
1057 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1058
1059 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1060 struct sk_buff_fclones {
1061 struct sk_buff skb1;
1062
1063 struct sk_buff skb2;
1064
1065 refcount_t fclone_ref;
1066 };
1067
1068 /**
1069 * skb_fclone_busy - check if fclone is busy
1070 * @sk: socket
1071 * @skb: buffer
1072 *
1073 * Returns true if skb is a fast clone, and its clone is not freed.
1074 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1075 * so we also check that this didnt happen.
1076 */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1077 static inline bool skb_fclone_busy(const struct sock *sk,
1078 const struct sk_buff *skb)
1079 {
1080 const struct sk_buff_fclones *fclones;
1081
1082 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1083
1084 return skb->fclone == SKB_FCLONE_ORIG &&
1085 refcount_read(&fclones->fclone_ref) > 1 &&
1086 fclones->skb2.sk == sk;
1087 }
1088
1089 /**
1090 * alloc_skb_fclone - allocate a network buffer from fclone cache
1091 * @size: size to allocate
1092 * @priority: allocation mask
1093 *
1094 * This function is a convenient wrapper around __alloc_skb().
1095 */
alloc_skb_fclone(unsigned int size,gfp_t priority)1096 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1097 gfp_t priority)
1098 {
1099 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1100 }
1101
1102 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1103 void skb_headers_offset_update(struct sk_buff *skb, int off);
1104 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1105 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1106 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1107 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1108 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1109 gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1110 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1111 gfp_t gfp_mask)
1112 {
1113 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1114 }
1115
1116 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1117 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1118 unsigned int headroom);
1119 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1120 int newtailroom, gfp_t priority);
1121 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1122 int offset, int len);
1123 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1124 int offset, int len);
1125 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1126 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1127
1128 /**
1129 * skb_pad - zero pad the tail of an skb
1130 * @skb: buffer to pad
1131 * @pad: space to pad
1132 *
1133 * Ensure that a buffer is followed by a padding area that is zero
1134 * filled. Used by network drivers which may DMA or transfer data
1135 * beyond the buffer end onto the wire.
1136 *
1137 * May return error in out of memory cases. The skb is freed on error.
1138 */
skb_pad(struct sk_buff * skb,int pad)1139 static inline int skb_pad(struct sk_buff *skb, int pad)
1140 {
1141 return __skb_pad(skb, pad, true);
1142 }
1143 #define dev_kfree_skb(a) consume_skb(a)
1144
1145 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1146 int offset, size_t size);
1147
1148 struct skb_seq_state {
1149 __u32 lower_offset;
1150 __u32 upper_offset;
1151 __u32 frag_idx;
1152 __u32 stepped_offset;
1153 struct sk_buff *root_skb;
1154 struct sk_buff *cur_skb;
1155 __u8 *frag_data;
1156 };
1157
1158 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1159 unsigned int to, struct skb_seq_state *st);
1160 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1161 struct skb_seq_state *st);
1162 void skb_abort_seq_read(struct skb_seq_state *st);
1163
1164 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1165 unsigned int to, struct ts_config *config);
1166
1167 /*
1168 * Packet hash types specify the type of hash in skb_set_hash.
1169 *
1170 * Hash types refer to the protocol layer addresses which are used to
1171 * construct a packet's hash. The hashes are used to differentiate or identify
1172 * flows of the protocol layer for the hash type. Hash types are either
1173 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1174 *
1175 * Properties of hashes:
1176 *
1177 * 1) Two packets in different flows have different hash values
1178 * 2) Two packets in the same flow should have the same hash value
1179 *
1180 * A hash at a higher layer is considered to be more specific. A driver should
1181 * set the most specific hash possible.
1182 *
1183 * A driver cannot indicate a more specific hash than the layer at which a hash
1184 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1185 *
1186 * A driver may indicate a hash level which is less specific than the
1187 * actual layer the hash was computed on. For instance, a hash computed
1188 * at L4 may be considered an L3 hash. This should only be done if the
1189 * driver can't unambiguously determine that the HW computed the hash at
1190 * the higher layer. Note that the "should" in the second property above
1191 * permits this.
1192 */
1193 enum pkt_hash_types {
1194 PKT_HASH_TYPE_NONE, /* Undefined type */
1195 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1196 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1197 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1198 };
1199
skb_clear_hash(struct sk_buff * skb)1200 static inline void skb_clear_hash(struct sk_buff *skb)
1201 {
1202 skb->hash = 0;
1203 skb->sw_hash = 0;
1204 skb->l4_hash = 0;
1205 }
1206
skb_clear_hash_if_not_l4(struct sk_buff * skb)1207 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1208 {
1209 if (!skb->l4_hash)
1210 skb_clear_hash(skb);
1211 }
1212
1213 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1214 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1215 {
1216 skb->l4_hash = is_l4;
1217 skb->sw_hash = is_sw;
1218 skb->hash = hash;
1219 }
1220
1221 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1222 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1223 {
1224 /* Used by drivers to set hash from HW */
1225 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1226 }
1227
1228 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1229 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1230 {
1231 __skb_set_hash(skb, hash, true, is_l4);
1232 }
1233
1234 void __skb_get_hash(struct sk_buff *skb);
1235 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1236 u32 skb_get_poff(const struct sk_buff *skb);
1237 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1238 const struct flow_keys_basic *keys, int hlen);
1239 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1240 void *data, int hlen_proto);
1241
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1242 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1243 int thoff, u8 ip_proto)
1244 {
1245 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1246 }
1247
1248 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1249 const struct flow_dissector_key *key,
1250 unsigned int key_count);
1251
1252 #ifdef CONFIG_NET
1253 int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1254 union bpf_attr __user *uattr);
1255 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1256 struct bpf_prog *prog);
1257
1258 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
1259 #else
skb_flow_dissector_prog_query(const union bpf_attr * attr,union bpf_attr __user * uattr)1260 static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1261 union bpf_attr __user *uattr)
1262 {
1263 return -EOPNOTSUPP;
1264 }
1265
skb_flow_dissector_bpf_prog_attach(const union bpf_attr * attr,struct bpf_prog * prog)1266 static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1267 struct bpf_prog *prog)
1268 {
1269 return -EOPNOTSUPP;
1270 }
1271
skb_flow_dissector_bpf_prog_detach(const union bpf_attr * attr)1272 static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1273 {
1274 return -EOPNOTSUPP;
1275 }
1276 #endif
1277
1278 struct bpf_flow_dissector;
1279 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1280 __be16 proto, int nhoff, int hlen, unsigned int flags);
1281
1282 bool __skb_flow_dissect(const struct net *net,
1283 const struct sk_buff *skb,
1284 struct flow_dissector *flow_dissector,
1285 void *target_container,
1286 void *data, __be16 proto, int nhoff, int hlen,
1287 unsigned int flags);
1288
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1289 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1290 struct flow_dissector *flow_dissector,
1291 void *target_container, unsigned int flags)
1292 {
1293 return __skb_flow_dissect(NULL, skb, flow_dissector,
1294 target_container, NULL, 0, 0, 0, flags);
1295 }
1296
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1297 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1298 struct flow_keys *flow,
1299 unsigned int flags)
1300 {
1301 memset(flow, 0, sizeof(*flow));
1302 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1303 flow, NULL, 0, 0, 0, flags);
1304 }
1305
1306 static inline bool
skb_flow_dissect_flow_keys_basic(const struct net * net,const struct sk_buff * skb,struct flow_keys_basic * flow,void * data,__be16 proto,int nhoff,int hlen,unsigned int flags)1307 skb_flow_dissect_flow_keys_basic(const struct net *net,
1308 const struct sk_buff *skb,
1309 struct flow_keys_basic *flow, void *data,
1310 __be16 proto, int nhoff, int hlen,
1311 unsigned int flags)
1312 {
1313 memset(flow, 0, sizeof(*flow));
1314 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1315 data, proto, nhoff, hlen, flags);
1316 }
1317
1318 void skb_flow_dissect_meta(const struct sk_buff *skb,
1319 struct flow_dissector *flow_dissector,
1320 void *target_container);
1321
1322 /* Gets a skb connection tracking info, ctinfo map should be a
1323 * a map of mapsize to translate enum ip_conntrack_info states
1324 * to user states.
1325 */
1326 void
1327 skb_flow_dissect_ct(const struct sk_buff *skb,
1328 struct flow_dissector *flow_dissector,
1329 void *target_container,
1330 u16 *ctinfo_map,
1331 size_t mapsize);
1332 void
1333 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1334 struct flow_dissector *flow_dissector,
1335 void *target_container);
1336
skb_get_hash(struct sk_buff * skb)1337 static inline __u32 skb_get_hash(struct sk_buff *skb)
1338 {
1339 if (!skb->l4_hash && !skb->sw_hash)
1340 __skb_get_hash(skb);
1341
1342 return skb->hash;
1343 }
1344
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1345 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1346 {
1347 if (!skb->l4_hash && !skb->sw_hash) {
1348 struct flow_keys keys;
1349 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1350
1351 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1352 }
1353
1354 return skb->hash;
1355 }
1356
1357 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1358 const siphash_key_t *perturb);
1359
skb_get_hash_raw(const struct sk_buff * skb)1360 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1361 {
1362 return skb->hash;
1363 }
1364
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1365 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1366 {
1367 to->hash = from->hash;
1368 to->sw_hash = from->sw_hash;
1369 to->l4_hash = from->l4_hash;
1370 };
1371
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1372 static inline void skb_copy_decrypted(struct sk_buff *to,
1373 const struct sk_buff *from)
1374 {
1375 #ifdef CONFIG_TLS_DEVICE
1376 to->decrypted = from->decrypted;
1377 #endif
1378 }
1379
1380 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1381 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1382 {
1383 return skb->head + skb->end;
1384 }
1385
skb_end_offset(const struct sk_buff * skb)1386 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1387 {
1388 return skb->end;
1389 }
1390 #else
skb_end_pointer(const struct sk_buff * skb)1391 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1392 {
1393 return skb->end;
1394 }
1395
skb_end_offset(const struct sk_buff * skb)1396 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1397 {
1398 return skb->end - skb->head;
1399 }
1400 #endif
1401
1402 /* Internal */
1403 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1404
skb_hwtstamps(struct sk_buff * skb)1405 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1406 {
1407 return &skb_shinfo(skb)->hwtstamps;
1408 }
1409
skb_zcopy(struct sk_buff * skb)1410 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1411 {
1412 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1413
1414 return is_zcopy ? skb_uarg(skb) : NULL;
1415 }
1416
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1417 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1418 bool *have_ref)
1419 {
1420 if (skb && uarg && !skb_zcopy(skb)) {
1421 if (unlikely(have_ref && *have_ref))
1422 *have_ref = false;
1423 else
1424 sock_zerocopy_get(uarg);
1425 skb_shinfo(skb)->destructor_arg = uarg;
1426 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1427 }
1428 }
1429
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1430 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1431 {
1432 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1433 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1434 }
1435
skb_zcopy_is_nouarg(struct sk_buff * skb)1436 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1437 {
1438 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1439 }
1440
skb_zcopy_get_nouarg(struct sk_buff * skb)1441 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1442 {
1443 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1444 }
1445
1446 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy)1447 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1448 {
1449 struct ubuf_info *uarg = skb_zcopy(skb);
1450
1451 if (uarg) {
1452 if (skb_zcopy_is_nouarg(skb)) {
1453 /* no notification callback */
1454 } else if (uarg->callback == sock_zerocopy_callback) {
1455 uarg->zerocopy = uarg->zerocopy && zerocopy;
1456 sock_zerocopy_put(uarg);
1457 } else {
1458 uarg->callback(uarg, zerocopy);
1459 }
1460
1461 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1462 }
1463 }
1464
1465 /* Abort a zerocopy operation and revert zckey on error in send syscall */
skb_zcopy_abort(struct sk_buff * skb)1466 static inline void skb_zcopy_abort(struct sk_buff *skb)
1467 {
1468 struct ubuf_info *uarg = skb_zcopy(skb);
1469
1470 if (uarg) {
1471 sock_zerocopy_put_abort(uarg, false);
1472 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1473 }
1474 }
1475
skb_mark_not_on_list(struct sk_buff * skb)1476 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1477 {
1478 skb->next = NULL;
1479 }
1480
skb_list_del_init(struct sk_buff * skb)1481 static inline void skb_list_del_init(struct sk_buff *skb)
1482 {
1483 __list_del_entry(&skb->list);
1484 skb_mark_not_on_list(skb);
1485 }
1486
1487 /**
1488 * skb_queue_empty - check if a queue is empty
1489 * @list: queue head
1490 *
1491 * Returns true if the queue is empty, false otherwise.
1492 */
skb_queue_empty(const struct sk_buff_head * list)1493 static inline int skb_queue_empty(const struct sk_buff_head *list)
1494 {
1495 return list->next == (const struct sk_buff *) list;
1496 }
1497
1498 /**
1499 * skb_queue_empty_lockless - check if a queue is empty
1500 * @list: queue head
1501 *
1502 * Returns true if the queue is empty, false otherwise.
1503 * This variant can be used in lockless contexts.
1504 */
skb_queue_empty_lockless(const struct sk_buff_head * list)1505 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1506 {
1507 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1508 }
1509
1510
1511 /**
1512 * skb_queue_is_last - check if skb is the last entry in the queue
1513 * @list: queue head
1514 * @skb: buffer
1515 *
1516 * Returns true if @skb is the last buffer on the list.
1517 */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1518 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1519 const struct sk_buff *skb)
1520 {
1521 return skb->next == (const struct sk_buff *) list;
1522 }
1523
1524 /**
1525 * skb_queue_is_first - check if skb is the first entry in the queue
1526 * @list: queue head
1527 * @skb: buffer
1528 *
1529 * Returns true if @skb is the first buffer on the list.
1530 */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1531 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1532 const struct sk_buff *skb)
1533 {
1534 return skb->prev == (const struct sk_buff *) list;
1535 }
1536
1537 /**
1538 * skb_queue_next - return the next packet in the queue
1539 * @list: queue head
1540 * @skb: current buffer
1541 *
1542 * Return the next packet in @list after @skb. It is only valid to
1543 * call this if skb_queue_is_last() evaluates to false.
1544 */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1545 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1546 const struct sk_buff *skb)
1547 {
1548 /* This BUG_ON may seem severe, but if we just return then we
1549 * are going to dereference garbage.
1550 */
1551 BUG_ON(skb_queue_is_last(list, skb));
1552 return skb->next;
1553 }
1554
1555 /**
1556 * skb_queue_prev - return the prev packet in the queue
1557 * @list: queue head
1558 * @skb: current buffer
1559 *
1560 * Return the prev packet in @list before @skb. It is only valid to
1561 * call this if skb_queue_is_first() evaluates to false.
1562 */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1563 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1564 const struct sk_buff *skb)
1565 {
1566 /* This BUG_ON may seem severe, but if we just return then we
1567 * are going to dereference garbage.
1568 */
1569 BUG_ON(skb_queue_is_first(list, skb));
1570 return skb->prev;
1571 }
1572
1573 /**
1574 * skb_get - reference buffer
1575 * @skb: buffer to reference
1576 *
1577 * Makes another reference to a socket buffer and returns a pointer
1578 * to the buffer.
1579 */
skb_get(struct sk_buff * skb)1580 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1581 {
1582 refcount_inc(&skb->users);
1583 return skb;
1584 }
1585
1586 /*
1587 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1588 */
1589
1590 /**
1591 * skb_cloned - is the buffer a clone
1592 * @skb: buffer to check
1593 *
1594 * Returns true if the buffer was generated with skb_clone() and is
1595 * one of multiple shared copies of the buffer. Cloned buffers are
1596 * shared data so must not be written to under normal circumstances.
1597 */
skb_cloned(const struct sk_buff * skb)1598 static inline int skb_cloned(const struct sk_buff *skb)
1599 {
1600 return skb->cloned &&
1601 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1602 }
1603
skb_unclone(struct sk_buff * skb,gfp_t pri)1604 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1605 {
1606 might_sleep_if(gfpflags_allow_blocking(pri));
1607
1608 if (skb_cloned(skb))
1609 return pskb_expand_head(skb, 0, 0, pri);
1610
1611 return 0;
1612 }
1613
1614 /**
1615 * skb_header_cloned - is the header a clone
1616 * @skb: buffer to check
1617 *
1618 * Returns true if modifying the header part of the buffer requires
1619 * the data to be copied.
1620 */
skb_header_cloned(const struct sk_buff * skb)1621 static inline int skb_header_cloned(const struct sk_buff *skb)
1622 {
1623 int dataref;
1624
1625 if (!skb->cloned)
1626 return 0;
1627
1628 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1629 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1630 return dataref != 1;
1631 }
1632
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1633 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1634 {
1635 might_sleep_if(gfpflags_allow_blocking(pri));
1636
1637 if (skb_header_cloned(skb))
1638 return pskb_expand_head(skb, 0, 0, pri);
1639
1640 return 0;
1641 }
1642
1643 /**
1644 * __skb_header_release - release reference to header
1645 * @skb: buffer to operate on
1646 */
__skb_header_release(struct sk_buff * skb)1647 static inline void __skb_header_release(struct sk_buff *skb)
1648 {
1649 skb->nohdr = 1;
1650 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1651 }
1652
1653
1654 /**
1655 * skb_shared - is the buffer shared
1656 * @skb: buffer to check
1657 *
1658 * Returns true if more than one person has a reference to this
1659 * buffer.
1660 */
skb_shared(const struct sk_buff * skb)1661 static inline int skb_shared(const struct sk_buff *skb)
1662 {
1663 return refcount_read(&skb->users) != 1;
1664 }
1665
1666 /**
1667 * skb_share_check - check if buffer is shared and if so clone it
1668 * @skb: buffer to check
1669 * @pri: priority for memory allocation
1670 *
1671 * If the buffer is shared the buffer is cloned and the old copy
1672 * drops a reference. A new clone with a single reference is returned.
1673 * If the buffer is not shared the original buffer is returned. When
1674 * being called from interrupt status or with spinlocks held pri must
1675 * be GFP_ATOMIC.
1676 *
1677 * NULL is returned on a memory allocation failure.
1678 */
skb_share_check(struct sk_buff * skb,gfp_t pri)1679 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1680 {
1681 might_sleep_if(gfpflags_allow_blocking(pri));
1682 if (skb_shared(skb)) {
1683 struct sk_buff *nskb = skb_clone(skb, pri);
1684
1685 if (likely(nskb))
1686 consume_skb(skb);
1687 else
1688 kfree_skb(skb);
1689 skb = nskb;
1690 }
1691 return skb;
1692 }
1693
1694 /*
1695 * Copy shared buffers into a new sk_buff. We effectively do COW on
1696 * packets to handle cases where we have a local reader and forward
1697 * and a couple of other messy ones. The normal one is tcpdumping
1698 * a packet thats being forwarded.
1699 */
1700
1701 /**
1702 * skb_unshare - make a copy of a shared buffer
1703 * @skb: buffer to check
1704 * @pri: priority for memory allocation
1705 *
1706 * If the socket buffer is a clone then this function creates a new
1707 * copy of the data, drops a reference count on the old copy and returns
1708 * the new copy with the reference count at 1. If the buffer is not a clone
1709 * the original buffer is returned. When called with a spinlock held or
1710 * from interrupt state @pri must be %GFP_ATOMIC
1711 *
1712 * %NULL is returned on a memory allocation failure.
1713 */
skb_unshare(struct sk_buff * skb,gfp_t pri)1714 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1715 gfp_t pri)
1716 {
1717 might_sleep_if(gfpflags_allow_blocking(pri));
1718 if (skb_cloned(skb)) {
1719 struct sk_buff *nskb = skb_copy(skb, pri);
1720
1721 /* Free our shared copy */
1722 if (likely(nskb))
1723 consume_skb(skb);
1724 else
1725 kfree_skb(skb);
1726 skb = nskb;
1727 }
1728 return skb;
1729 }
1730
1731 /**
1732 * skb_peek - peek at the head of an &sk_buff_head
1733 * @list_: list to peek at
1734 *
1735 * Peek an &sk_buff. Unlike most other operations you _MUST_
1736 * be careful with this one. A peek leaves the buffer on the
1737 * list and someone else may run off with it. You must hold
1738 * the appropriate locks or have a private queue to do this.
1739 *
1740 * Returns %NULL for an empty list or a pointer to the head element.
1741 * The reference count is not incremented and the reference is therefore
1742 * volatile. Use with caution.
1743 */
skb_peek(const struct sk_buff_head * list_)1744 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1745 {
1746 struct sk_buff *skb = list_->next;
1747
1748 if (skb == (struct sk_buff *)list_)
1749 skb = NULL;
1750 return skb;
1751 }
1752
1753 /**
1754 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1755 * @list_: list to peek at
1756 *
1757 * Like skb_peek(), but the caller knows that the list is not empty.
1758 */
__skb_peek(const struct sk_buff_head * list_)1759 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1760 {
1761 return list_->next;
1762 }
1763
1764 /**
1765 * skb_peek_next - peek skb following the given one from a queue
1766 * @skb: skb to start from
1767 * @list_: list to peek at
1768 *
1769 * Returns %NULL when the end of the list is met or a pointer to the
1770 * next element. The reference count is not incremented and the
1771 * reference is therefore volatile. Use with caution.
1772 */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1773 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1774 const struct sk_buff_head *list_)
1775 {
1776 struct sk_buff *next = skb->next;
1777
1778 if (next == (struct sk_buff *)list_)
1779 next = NULL;
1780 return next;
1781 }
1782
1783 /**
1784 * skb_peek_tail - peek at the tail of an &sk_buff_head
1785 * @list_: list to peek at
1786 *
1787 * Peek an &sk_buff. Unlike most other operations you _MUST_
1788 * be careful with this one. A peek leaves the buffer on the
1789 * list and someone else may run off with it. You must hold
1790 * the appropriate locks or have a private queue to do this.
1791 *
1792 * Returns %NULL for an empty list or a pointer to the tail element.
1793 * The reference count is not incremented and the reference is therefore
1794 * volatile. Use with caution.
1795 */
skb_peek_tail(const struct sk_buff_head * list_)1796 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1797 {
1798 struct sk_buff *skb = READ_ONCE(list_->prev);
1799
1800 if (skb == (struct sk_buff *)list_)
1801 skb = NULL;
1802 return skb;
1803
1804 }
1805
1806 /**
1807 * skb_queue_len - get queue length
1808 * @list_: list to measure
1809 *
1810 * Return the length of an &sk_buff queue.
1811 */
skb_queue_len(const struct sk_buff_head * list_)1812 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1813 {
1814 return list_->qlen;
1815 }
1816
1817 /**
1818 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1819 * @list: queue to initialize
1820 *
1821 * This initializes only the list and queue length aspects of
1822 * an sk_buff_head object. This allows to initialize the list
1823 * aspects of an sk_buff_head without reinitializing things like
1824 * the spinlock. It can also be used for on-stack sk_buff_head
1825 * objects where the spinlock is known to not be used.
1826 */
__skb_queue_head_init(struct sk_buff_head * list)1827 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1828 {
1829 list->prev = list->next = (struct sk_buff *)list;
1830 list->qlen = 0;
1831 }
1832
1833 /*
1834 * This function creates a split out lock class for each invocation;
1835 * this is needed for now since a whole lot of users of the skb-queue
1836 * infrastructure in drivers have different locking usage (in hardirq)
1837 * than the networking core (in softirq only). In the long run either the
1838 * network layer or drivers should need annotation to consolidate the
1839 * main types of usage into 3 classes.
1840 */
skb_queue_head_init(struct sk_buff_head * list)1841 static inline void skb_queue_head_init(struct sk_buff_head *list)
1842 {
1843 spin_lock_init(&list->lock);
1844 __skb_queue_head_init(list);
1845 }
1846
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)1847 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1848 struct lock_class_key *class)
1849 {
1850 skb_queue_head_init(list);
1851 lockdep_set_class(&list->lock, class);
1852 }
1853
1854 /*
1855 * Insert an sk_buff on a list.
1856 *
1857 * The "__skb_xxxx()" functions are the non-atomic ones that
1858 * can only be called with interrupts disabled.
1859 */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)1860 static inline void __skb_insert(struct sk_buff *newsk,
1861 struct sk_buff *prev, struct sk_buff *next,
1862 struct sk_buff_head *list)
1863 {
1864 /* See skb_queue_empty_lockless() and skb_peek_tail()
1865 * for the opposite READ_ONCE()
1866 */
1867 WRITE_ONCE(newsk->next, next);
1868 WRITE_ONCE(newsk->prev, prev);
1869 WRITE_ONCE(next->prev, newsk);
1870 WRITE_ONCE(prev->next, newsk);
1871 list->qlen++;
1872 }
1873
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)1874 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1875 struct sk_buff *prev,
1876 struct sk_buff *next)
1877 {
1878 struct sk_buff *first = list->next;
1879 struct sk_buff *last = list->prev;
1880
1881 WRITE_ONCE(first->prev, prev);
1882 WRITE_ONCE(prev->next, first);
1883
1884 WRITE_ONCE(last->next, next);
1885 WRITE_ONCE(next->prev, last);
1886 }
1887
1888 /**
1889 * skb_queue_splice - join two skb lists, this is designed for stacks
1890 * @list: the new list to add
1891 * @head: the place to add it in the first list
1892 */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)1893 static inline void skb_queue_splice(const struct sk_buff_head *list,
1894 struct sk_buff_head *head)
1895 {
1896 if (!skb_queue_empty(list)) {
1897 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1898 head->qlen += list->qlen;
1899 }
1900 }
1901
1902 /**
1903 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1904 * @list: the new list to add
1905 * @head: the place to add it in the first list
1906 *
1907 * The list at @list is reinitialised
1908 */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)1909 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1910 struct sk_buff_head *head)
1911 {
1912 if (!skb_queue_empty(list)) {
1913 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1914 head->qlen += list->qlen;
1915 __skb_queue_head_init(list);
1916 }
1917 }
1918
1919 /**
1920 * skb_queue_splice_tail - join two skb lists, each list being a queue
1921 * @list: the new list to add
1922 * @head: the place to add it in the first list
1923 */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)1924 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1925 struct sk_buff_head *head)
1926 {
1927 if (!skb_queue_empty(list)) {
1928 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1929 head->qlen += list->qlen;
1930 }
1931 }
1932
1933 /**
1934 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1935 * @list: the new list to add
1936 * @head: the place to add it in the first list
1937 *
1938 * Each of the lists is a queue.
1939 * The list at @list is reinitialised
1940 */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)1941 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1942 struct sk_buff_head *head)
1943 {
1944 if (!skb_queue_empty(list)) {
1945 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1946 head->qlen += list->qlen;
1947 __skb_queue_head_init(list);
1948 }
1949 }
1950
1951 /**
1952 * __skb_queue_after - queue a buffer at the list head
1953 * @list: list to use
1954 * @prev: place after this buffer
1955 * @newsk: buffer to queue
1956 *
1957 * Queue a buffer int the middle of a list. This function takes no locks
1958 * and you must therefore hold required locks before calling it.
1959 *
1960 * A buffer cannot be placed on two lists at the same time.
1961 */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)1962 static inline void __skb_queue_after(struct sk_buff_head *list,
1963 struct sk_buff *prev,
1964 struct sk_buff *newsk)
1965 {
1966 __skb_insert(newsk, prev, prev->next, list);
1967 }
1968
1969 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1970 struct sk_buff_head *list);
1971
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)1972 static inline void __skb_queue_before(struct sk_buff_head *list,
1973 struct sk_buff *next,
1974 struct sk_buff *newsk)
1975 {
1976 __skb_insert(newsk, next->prev, next, list);
1977 }
1978
1979 /**
1980 * __skb_queue_head - queue a buffer at the list head
1981 * @list: list to use
1982 * @newsk: buffer to queue
1983 *
1984 * Queue a buffer at the start of a list. This function takes no locks
1985 * and you must therefore hold required locks before calling it.
1986 *
1987 * A buffer cannot be placed on two lists at the same time.
1988 */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)1989 static inline void __skb_queue_head(struct sk_buff_head *list,
1990 struct sk_buff *newsk)
1991 {
1992 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1993 }
1994 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1995
1996 /**
1997 * __skb_queue_tail - queue a buffer at the list tail
1998 * @list: list to use
1999 * @newsk: buffer to queue
2000 *
2001 * Queue a buffer at the end of a list. This function takes no locks
2002 * and you must therefore hold required locks before calling it.
2003 *
2004 * A buffer cannot be placed on two lists at the same time.
2005 */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2006 static inline void __skb_queue_tail(struct sk_buff_head *list,
2007 struct sk_buff *newsk)
2008 {
2009 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2010 }
2011 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2012
2013 /*
2014 * remove sk_buff from list. _Must_ be called atomically, and with
2015 * the list known..
2016 */
2017 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2018 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2019 {
2020 struct sk_buff *next, *prev;
2021
2022 list->qlen--;
2023 next = skb->next;
2024 prev = skb->prev;
2025 skb->next = skb->prev = NULL;
2026 WRITE_ONCE(next->prev, prev);
2027 WRITE_ONCE(prev->next, next);
2028 }
2029
2030 /**
2031 * __skb_dequeue - remove from the head of the queue
2032 * @list: list to dequeue from
2033 *
2034 * Remove the head of the list. This function does not take any locks
2035 * so must be used with appropriate locks held only. The head item is
2036 * returned or %NULL if the list is empty.
2037 */
__skb_dequeue(struct sk_buff_head * list)2038 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2039 {
2040 struct sk_buff *skb = skb_peek(list);
2041 if (skb)
2042 __skb_unlink(skb, list);
2043 return skb;
2044 }
2045 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2046
2047 /**
2048 * __skb_dequeue_tail - remove from the tail of the queue
2049 * @list: list to dequeue from
2050 *
2051 * Remove the tail of the list. This function does not take any locks
2052 * so must be used with appropriate locks held only. The tail item is
2053 * returned or %NULL if the list is empty.
2054 */
__skb_dequeue_tail(struct sk_buff_head * list)2055 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2056 {
2057 struct sk_buff *skb = skb_peek_tail(list);
2058 if (skb)
2059 __skb_unlink(skb, list);
2060 return skb;
2061 }
2062 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2063
2064
skb_is_nonlinear(const struct sk_buff * skb)2065 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2066 {
2067 return skb->data_len;
2068 }
2069
skb_headlen(const struct sk_buff * skb)2070 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2071 {
2072 return skb->len - skb->data_len;
2073 }
2074
__skb_pagelen(const struct sk_buff * skb)2075 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2076 {
2077 unsigned int i, len = 0;
2078
2079 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2080 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2081 return len;
2082 }
2083
skb_pagelen(const struct sk_buff * skb)2084 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2085 {
2086 return skb_headlen(skb) + __skb_pagelen(skb);
2087 }
2088
2089 /**
2090 * __skb_fill_page_desc - initialise a paged fragment in an skb
2091 * @skb: buffer containing fragment to be initialised
2092 * @i: paged fragment index to initialise
2093 * @page: the page to use for this fragment
2094 * @off: the offset to the data with @page
2095 * @size: the length of the data
2096 *
2097 * Initialises the @i'th fragment of @skb to point to &size bytes at
2098 * offset @off within @page.
2099 *
2100 * Does not take any additional reference on the fragment.
2101 */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2102 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2103 struct page *page, int off, int size)
2104 {
2105 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2106
2107 /*
2108 * Propagate page pfmemalloc to the skb if we can. The problem is
2109 * that not all callers have unique ownership of the page but rely
2110 * on page_is_pfmemalloc doing the right thing(tm).
2111 */
2112 frag->bv_page = page;
2113 frag->bv_offset = off;
2114 skb_frag_size_set(frag, size);
2115
2116 page = compound_head(page);
2117 if (page_is_pfmemalloc(page))
2118 skb->pfmemalloc = true;
2119 }
2120
2121 /**
2122 * skb_fill_page_desc - initialise a paged fragment in an skb
2123 * @skb: buffer containing fragment to be initialised
2124 * @i: paged fragment index to initialise
2125 * @page: the page to use for this fragment
2126 * @off: the offset to the data with @page
2127 * @size: the length of the data
2128 *
2129 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2130 * @skb to point to @size bytes at offset @off within @page. In
2131 * addition updates @skb such that @i is the last fragment.
2132 *
2133 * Does not take any additional reference on the fragment.
2134 */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2135 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2136 struct page *page, int off, int size)
2137 {
2138 __skb_fill_page_desc(skb, i, page, off, size);
2139 skb_shinfo(skb)->nr_frags = i + 1;
2140 }
2141
2142 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2143 int size, unsigned int truesize);
2144
2145 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2146 unsigned int truesize);
2147
2148 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2149
2150 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2151 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2152 {
2153 return skb->head + skb->tail;
2154 }
2155
skb_reset_tail_pointer(struct sk_buff * skb)2156 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2157 {
2158 skb->tail = skb->data - skb->head;
2159 }
2160
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2161 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2162 {
2163 skb_reset_tail_pointer(skb);
2164 skb->tail += offset;
2165 }
2166
2167 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2168 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2169 {
2170 return skb->tail;
2171 }
2172
skb_reset_tail_pointer(struct sk_buff * skb)2173 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2174 {
2175 skb->tail = skb->data;
2176 }
2177
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2178 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2179 {
2180 skb->tail = skb->data + offset;
2181 }
2182
2183 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2184
2185 /*
2186 * Add data to an sk_buff
2187 */
2188 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2189 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2190 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2191 {
2192 void *tmp = skb_tail_pointer(skb);
2193 SKB_LINEAR_ASSERT(skb);
2194 skb->tail += len;
2195 skb->len += len;
2196 return tmp;
2197 }
2198
__skb_put_zero(struct sk_buff * skb,unsigned int len)2199 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2200 {
2201 void *tmp = __skb_put(skb, len);
2202
2203 memset(tmp, 0, len);
2204 return tmp;
2205 }
2206
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2207 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2208 unsigned int len)
2209 {
2210 void *tmp = __skb_put(skb, len);
2211
2212 memcpy(tmp, data, len);
2213 return tmp;
2214 }
2215
__skb_put_u8(struct sk_buff * skb,u8 val)2216 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2217 {
2218 *(u8 *)__skb_put(skb, 1) = val;
2219 }
2220
skb_put_zero(struct sk_buff * skb,unsigned int len)2221 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2222 {
2223 void *tmp = skb_put(skb, len);
2224
2225 memset(tmp, 0, len);
2226
2227 return tmp;
2228 }
2229
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2230 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2231 unsigned int len)
2232 {
2233 void *tmp = skb_put(skb, len);
2234
2235 memcpy(tmp, data, len);
2236
2237 return tmp;
2238 }
2239
skb_put_u8(struct sk_buff * skb,u8 val)2240 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2241 {
2242 *(u8 *)skb_put(skb, 1) = val;
2243 }
2244
2245 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2246 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2247 {
2248 skb->data -= len;
2249 skb->len += len;
2250 return skb->data;
2251 }
2252
2253 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2254 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2255 {
2256 skb->len -= len;
2257 BUG_ON(skb->len < skb->data_len);
2258 return skb->data += len;
2259 }
2260
skb_pull_inline(struct sk_buff * skb,unsigned int len)2261 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2262 {
2263 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2264 }
2265
2266 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2267
__pskb_pull(struct sk_buff * skb,unsigned int len)2268 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2269 {
2270 if (len > skb_headlen(skb) &&
2271 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2272 return NULL;
2273 skb->len -= len;
2274 return skb->data += len;
2275 }
2276
pskb_pull(struct sk_buff * skb,unsigned int len)2277 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2278 {
2279 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2280 }
2281
pskb_may_pull(struct sk_buff * skb,unsigned int len)2282 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2283 {
2284 if (likely(len <= skb_headlen(skb)))
2285 return 1;
2286 if (unlikely(len > skb->len))
2287 return 0;
2288 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2289 }
2290
2291 void skb_condense(struct sk_buff *skb);
2292
2293 /**
2294 * skb_headroom - bytes at buffer head
2295 * @skb: buffer to check
2296 *
2297 * Return the number of bytes of free space at the head of an &sk_buff.
2298 */
skb_headroom(const struct sk_buff * skb)2299 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2300 {
2301 return skb->data - skb->head;
2302 }
2303
2304 /**
2305 * skb_tailroom - bytes at buffer end
2306 * @skb: buffer to check
2307 *
2308 * Return the number of bytes of free space at the tail of an sk_buff
2309 */
skb_tailroom(const struct sk_buff * skb)2310 static inline int skb_tailroom(const struct sk_buff *skb)
2311 {
2312 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2313 }
2314
2315 /**
2316 * skb_availroom - bytes at buffer end
2317 * @skb: buffer to check
2318 *
2319 * Return the number of bytes of free space at the tail of an sk_buff
2320 * allocated by sk_stream_alloc()
2321 */
skb_availroom(const struct sk_buff * skb)2322 static inline int skb_availroom(const struct sk_buff *skb)
2323 {
2324 if (skb_is_nonlinear(skb))
2325 return 0;
2326
2327 return skb->end - skb->tail - skb->reserved_tailroom;
2328 }
2329
2330 /**
2331 * skb_reserve - adjust headroom
2332 * @skb: buffer to alter
2333 * @len: bytes to move
2334 *
2335 * Increase the headroom of an empty &sk_buff by reducing the tail
2336 * room. This is only allowed for an empty buffer.
2337 */
skb_reserve(struct sk_buff * skb,int len)2338 static inline void skb_reserve(struct sk_buff *skb, int len)
2339 {
2340 skb->data += len;
2341 skb->tail += len;
2342 }
2343
2344 /**
2345 * skb_tailroom_reserve - adjust reserved_tailroom
2346 * @skb: buffer to alter
2347 * @mtu: maximum amount of headlen permitted
2348 * @needed_tailroom: minimum amount of reserved_tailroom
2349 *
2350 * Set reserved_tailroom so that headlen can be as large as possible but
2351 * not larger than mtu and tailroom cannot be smaller than
2352 * needed_tailroom.
2353 * The required headroom should already have been reserved before using
2354 * this function.
2355 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2356 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2357 unsigned int needed_tailroom)
2358 {
2359 SKB_LINEAR_ASSERT(skb);
2360 if (mtu < skb_tailroom(skb) - needed_tailroom)
2361 /* use at most mtu */
2362 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2363 else
2364 /* use up to all available space */
2365 skb->reserved_tailroom = needed_tailroom;
2366 }
2367
2368 #define ENCAP_TYPE_ETHER 0
2369 #define ENCAP_TYPE_IPPROTO 1
2370
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2371 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2372 __be16 protocol)
2373 {
2374 skb->inner_protocol = protocol;
2375 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2376 }
2377
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2378 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2379 __u8 ipproto)
2380 {
2381 skb->inner_ipproto = ipproto;
2382 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2383 }
2384
skb_reset_inner_headers(struct sk_buff * skb)2385 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2386 {
2387 skb->inner_mac_header = skb->mac_header;
2388 skb->inner_network_header = skb->network_header;
2389 skb->inner_transport_header = skb->transport_header;
2390 }
2391
skb_reset_mac_len(struct sk_buff * skb)2392 static inline void skb_reset_mac_len(struct sk_buff *skb)
2393 {
2394 skb->mac_len = skb->network_header - skb->mac_header;
2395 }
2396
skb_inner_transport_header(const struct sk_buff * skb)2397 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2398 *skb)
2399 {
2400 return skb->head + skb->inner_transport_header;
2401 }
2402
skb_inner_transport_offset(const struct sk_buff * skb)2403 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2404 {
2405 return skb_inner_transport_header(skb) - skb->data;
2406 }
2407
skb_reset_inner_transport_header(struct sk_buff * skb)2408 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2409 {
2410 skb->inner_transport_header = skb->data - skb->head;
2411 }
2412
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2413 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2414 const int offset)
2415 {
2416 skb_reset_inner_transport_header(skb);
2417 skb->inner_transport_header += offset;
2418 }
2419
skb_inner_network_header(const struct sk_buff * skb)2420 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2421 {
2422 return skb->head + skb->inner_network_header;
2423 }
2424
skb_reset_inner_network_header(struct sk_buff * skb)2425 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2426 {
2427 skb->inner_network_header = skb->data - skb->head;
2428 }
2429
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2430 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2431 const int offset)
2432 {
2433 skb_reset_inner_network_header(skb);
2434 skb->inner_network_header += offset;
2435 }
2436
skb_inner_mac_header(const struct sk_buff * skb)2437 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2438 {
2439 return skb->head + skb->inner_mac_header;
2440 }
2441
skb_reset_inner_mac_header(struct sk_buff * skb)2442 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2443 {
2444 skb->inner_mac_header = skb->data - skb->head;
2445 }
2446
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2447 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2448 const int offset)
2449 {
2450 skb_reset_inner_mac_header(skb);
2451 skb->inner_mac_header += offset;
2452 }
skb_transport_header_was_set(const struct sk_buff * skb)2453 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2454 {
2455 return skb->transport_header != (typeof(skb->transport_header))~0U;
2456 }
2457
skb_transport_header(const struct sk_buff * skb)2458 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2459 {
2460 return skb->head + skb->transport_header;
2461 }
2462
skb_reset_transport_header(struct sk_buff * skb)2463 static inline void skb_reset_transport_header(struct sk_buff *skb)
2464 {
2465 skb->transport_header = skb->data - skb->head;
2466 }
2467
skb_set_transport_header(struct sk_buff * skb,const int offset)2468 static inline void skb_set_transport_header(struct sk_buff *skb,
2469 const int offset)
2470 {
2471 skb_reset_transport_header(skb);
2472 skb->transport_header += offset;
2473 }
2474
skb_network_header(const struct sk_buff * skb)2475 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2476 {
2477 return skb->head + skb->network_header;
2478 }
2479
skb_reset_network_header(struct sk_buff * skb)2480 static inline void skb_reset_network_header(struct sk_buff *skb)
2481 {
2482 skb->network_header = skb->data - skb->head;
2483 }
2484
skb_set_network_header(struct sk_buff * skb,const int offset)2485 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2486 {
2487 skb_reset_network_header(skb);
2488 skb->network_header += offset;
2489 }
2490
skb_mac_header(const struct sk_buff * skb)2491 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2492 {
2493 return skb->head + skb->mac_header;
2494 }
2495
skb_mac_offset(const struct sk_buff * skb)2496 static inline int skb_mac_offset(const struct sk_buff *skb)
2497 {
2498 return skb_mac_header(skb) - skb->data;
2499 }
2500
skb_mac_header_len(const struct sk_buff * skb)2501 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2502 {
2503 return skb->network_header - skb->mac_header;
2504 }
2505
skb_mac_header_was_set(const struct sk_buff * skb)2506 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2507 {
2508 return skb->mac_header != (typeof(skb->mac_header))~0U;
2509 }
2510
skb_reset_mac_header(struct sk_buff * skb)2511 static inline void skb_reset_mac_header(struct sk_buff *skb)
2512 {
2513 skb->mac_header = skb->data - skb->head;
2514 }
2515
skb_set_mac_header(struct sk_buff * skb,const int offset)2516 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2517 {
2518 skb_reset_mac_header(skb);
2519 skb->mac_header += offset;
2520 }
2521
skb_pop_mac_header(struct sk_buff * skb)2522 static inline void skb_pop_mac_header(struct sk_buff *skb)
2523 {
2524 skb->mac_header = skb->network_header;
2525 }
2526
skb_probe_transport_header(struct sk_buff * skb)2527 static inline void skb_probe_transport_header(struct sk_buff *skb)
2528 {
2529 struct flow_keys_basic keys;
2530
2531 if (skb_transport_header_was_set(skb))
2532 return;
2533
2534 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2535 NULL, 0, 0, 0, 0))
2536 skb_set_transport_header(skb, keys.control.thoff);
2537 }
2538
skb_mac_header_rebuild(struct sk_buff * skb)2539 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2540 {
2541 if (skb_mac_header_was_set(skb)) {
2542 const unsigned char *old_mac = skb_mac_header(skb);
2543
2544 skb_set_mac_header(skb, -skb->mac_len);
2545 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2546 }
2547 }
2548
skb_checksum_start_offset(const struct sk_buff * skb)2549 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2550 {
2551 return skb->csum_start - skb_headroom(skb);
2552 }
2553
skb_checksum_start(const struct sk_buff * skb)2554 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2555 {
2556 return skb->head + skb->csum_start;
2557 }
2558
skb_transport_offset(const struct sk_buff * skb)2559 static inline int skb_transport_offset(const struct sk_buff *skb)
2560 {
2561 return skb_transport_header(skb) - skb->data;
2562 }
2563
skb_network_header_len(const struct sk_buff * skb)2564 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2565 {
2566 return skb->transport_header - skb->network_header;
2567 }
2568
skb_inner_network_header_len(const struct sk_buff * skb)2569 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2570 {
2571 return skb->inner_transport_header - skb->inner_network_header;
2572 }
2573
skb_network_offset(const struct sk_buff * skb)2574 static inline int skb_network_offset(const struct sk_buff *skb)
2575 {
2576 return skb_network_header(skb) - skb->data;
2577 }
2578
skb_inner_network_offset(const struct sk_buff * skb)2579 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2580 {
2581 return skb_inner_network_header(skb) - skb->data;
2582 }
2583
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2584 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2585 {
2586 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2587 }
2588
2589 /*
2590 * CPUs often take a performance hit when accessing unaligned memory
2591 * locations. The actual performance hit varies, it can be small if the
2592 * hardware handles it or large if we have to take an exception and fix it
2593 * in software.
2594 *
2595 * Since an ethernet header is 14 bytes network drivers often end up with
2596 * the IP header at an unaligned offset. The IP header can be aligned by
2597 * shifting the start of the packet by 2 bytes. Drivers should do this
2598 * with:
2599 *
2600 * skb_reserve(skb, NET_IP_ALIGN);
2601 *
2602 * The downside to this alignment of the IP header is that the DMA is now
2603 * unaligned. On some architectures the cost of an unaligned DMA is high
2604 * and this cost outweighs the gains made by aligning the IP header.
2605 *
2606 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2607 * to be overridden.
2608 */
2609 #ifndef NET_IP_ALIGN
2610 #define NET_IP_ALIGN 2
2611 #endif
2612
2613 /*
2614 * The networking layer reserves some headroom in skb data (via
2615 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2616 * the header has to grow. In the default case, if the header has to grow
2617 * 32 bytes or less we avoid the reallocation.
2618 *
2619 * Unfortunately this headroom changes the DMA alignment of the resulting
2620 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2621 * on some architectures. An architecture can override this value,
2622 * perhaps setting it to a cacheline in size (since that will maintain
2623 * cacheline alignment of the DMA). It must be a power of 2.
2624 *
2625 * Various parts of the networking layer expect at least 32 bytes of
2626 * headroom, you should not reduce this.
2627 *
2628 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2629 * to reduce average number of cache lines per packet.
2630 * get_rps_cpus() for example only access one 64 bytes aligned block :
2631 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2632 */
2633 #ifndef NET_SKB_PAD
2634 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2635 #endif
2636
2637 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2638
__skb_set_length(struct sk_buff * skb,unsigned int len)2639 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2640 {
2641 if (WARN_ON(skb_is_nonlinear(skb)))
2642 return;
2643 skb->len = len;
2644 skb_set_tail_pointer(skb, len);
2645 }
2646
__skb_trim(struct sk_buff * skb,unsigned int len)2647 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2648 {
2649 __skb_set_length(skb, len);
2650 }
2651
2652 void skb_trim(struct sk_buff *skb, unsigned int len);
2653
__pskb_trim(struct sk_buff * skb,unsigned int len)2654 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2655 {
2656 if (skb->data_len)
2657 return ___pskb_trim(skb, len);
2658 __skb_trim(skb, len);
2659 return 0;
2660 }
2661
pskb_trim(struct sk_buff * skb,unsigned int len)2662 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2663 {
2664 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2665 }
2666
2667 /**
2668 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2669 * @skb: buffer to alter
2670 * @len: new length
2671 *
2672 * This is identical to pskb_trim except that the caller knows that
2673 * the skb is not cloned so we should never get an error due to out-
2674 * of-memory.
2675 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2676 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2677 {
2678 int err = pskb_trim(skb, len);
2679 BUG_ON(err);
2680 }
2681
__skb_grow(struct sk_buff * skb,unsigned int len)2682 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2683 {
2684 unsigned int diff = len - skb->len;
2685
2686 if (skb_tailroom(skb) < diff) {
2687 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2688 GFP_ATOMIC);
2689 if (ret)
2690 return ret;
2691 }
2692 __skb_set_length(skb, len);
2693 return 0;
2694 }
2695
2696 /**
2697 * skb_orphan - orphan a buffer
2698 * @skb: buffer to orphan
2699 *
2700 * If a buffer currently has an owner then we call the owner's
2701 * destructor function and make the @skb unowned. The buffer continues
2702 * to exist but is no longer charged to its former owner.
2703 */
skb_orphan(struct sk_buff * skb)2704 static inline void skb_orphan(struct sk_buff *skb)
2705 {
2706 if (skb->destructor) {
2707 skb->destructor(skb);
2708 skb->destructor = NULL;
2709 skb->sk = NULL;
2710 } else {
2711 BUG_ON(skb->sk);
2712 }
2713 }
2714
2715 /**
2716 * skb_orphan_frags - orphan the frags contained in a buffer
2717 * @skb: buffer to orphan frags from
2718 * @gfp_mask: allocation mask for replacement pages
2719 *
2720 * For each frag in the SKB which needs a destructor (i.e. has an
2721 * owner) create a copy of that frag and release the original
2722 * page by calling the destructor.
2723 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2724 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2725 {
2726 if (likely(!skb_zcopy(skb)))
2727 return 0;
2728 if (!skb_zcopy_is_nouarg(skb) &&
2729 skb_uarg(skb)->callback == sock_zerocopy_callback)
2730 return 0;
2731 return skb_copy_ubufs(skb, gfp_mask);
2732 }
2733
2734 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
skb_orphan_frags_rx(struct sk_buff * skb,gfp_t gfp_mask)2735 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2736 {
2737 if (likely(!skb_zcopy(skb)))
2738 return 0;
2739 return skb_copy_ubufs(skb, gfp_mask);
2740 }
2741
2742 /**
2743 * __skb_queue_purge - empty a list
2744 * @list: list to empty
2745 *
2746 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2747 * the list and one reference dropped. This function does not take the
2748 * list lock and the caller must hold the relevant locks to use it.
2749 */
__skb_queue_purge(struct sk_buff_head * list)2750 static inline void __skb_queue_purge(struct sk_buff_head *list)
2751 {
2752 struct sk_buff *skb;
2753 while ((skb = __skb_dequeue(list)) != NULL)
2754 kfree_skb(skb);
2755 }
2756 void skb_queue_purge(struct sk_buff_head *list);
2757
2758 unsigned int skb_rbtree_purge(struct rb_root *root);
2759
2760 void *netdev_alloc_frag(unsigned int fragsz);
2761
2762 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2763 gfp_t gfp_mask);
2764
2765 /**
2766 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2767 * @dev: network device to receive on
2768 * @length: length to allocate
2769 *
2770 * Allocate a new &sk_buff and assign it a usage count of one. The
2771 * buffer has unspecified headroom built in. Users should allocate
2772 * the headroom they think they need without accounting for the
2773 * built in space. The built in space is used for optimisations.
2774 *
2775 * %NULL is returned if there is no free memory. Although this function
2776 * allocates memory it can be called from an interrupt.
2777 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)2778 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2779 unsigned int length)
2780 {
2781 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2782 }
2783
2784 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)2785 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2786 gfp_t gfp_mask)
2787 {
2788 return __netdev_alloc_skb(NULL, length, gfp_mask);
2789 }
2790
2791 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)2792 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2793 {
2794 return netdev_alloc_skb(NULL, length);
2795 }
2796
2797
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)2798 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2799 unsigned int length, gfp_t gfp)
2800 {
2801 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2802
2803 if (NET_IP_ALIGN && skb)
2804 skb_reserve(skb, NET_IP_ALIGN);
2805 return skb;
2806 }
2807
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)2808 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2809 unsigned int length)
2810 {
2811 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2812 }
2813
skb_free_frag(void * addr)2814 static inline void skb_free_frag(void *addr)
2815 {
2816 page_frag_free(addr);
2817 }
2818
2819 void *napi_alloc_frag(unsigned int fragsz);
2820 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2821 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)2822 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2823 unsigned int length)
2824 {
2825 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2826 }
2827 void napi_consume_skb(struct sk_buff *skb, int budget);
2828
2829 void __kfree_skb_flush(void);
2830 void __kfree_skb_defer(struct sk_buff *skb);
2831
2832 /**
2833 * __dev_alloc_pages - allocate page for network Rx
2834 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2835 * @order: size of the allocation
2836 *
2837 * Allocate a new page.
2838 *
2839 * %NULL is returned if there is no free memory.
2840 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)2841 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2842 unsigned int order)
2843 {
2844 /* This piece of code contains several assumptions.
2845 * 1. This is for device Rx, therefor a cold page is preferred.
2846 * 2. The expectation is the user wants a compound page.
2847 * 3. If requesting a order 0 page it will not be compound
2848 * due to the check to see if order has a value in prep_new_page
2849 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2850 * code in gfp_to_alloc_flags that should be enforcing this.
2851 */
2852 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2853
2854 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2855 }
2856
dev_alloc_pages(unsigned int order)2857 static inline struct page *dev_alloc_pages(unsigned int order)
2858 {
2859 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2860 }
2861
2862 /**
2863 * __dev_alloc_page - allocate a page for network Rx
2864 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2865 *
2866 * Allocate a new page.
2867 *
2868 * %NULL is returned if there is no free memory.
2869 */
__dev_alloc_page(gfp_t gfp_mask)2870 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2871 {
2872 return __dev_alloc_pages(gfp_mask, 0);
2873 }
2874
dev_alloc_page(void)2875 static inline struct page *dev_alloc_page(void)
2876 {
2877 return dev_alloc_pages(0);
2878 }
2879
2880 /**
2881 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2882 * @page: The page that was allocated from skb_alloc_page
2883 * @skb: The skb that may need pfmemalloc set
2884 */
skb_propagate_pfmemalloc(struct page * page,struct sk_buff * skb)2885 static inline void skb_propagate_pfmemalloc(struct page *page,
2886 struct sk_buff *skb)
2887 {
2888 if (page_is_pfmemalloc(page))
2889 skb->pfmemalloc = true;
2890 }
2891
2892 /**
2893 * skb_frag_off() - Returns the offset of a skb fragment
2894 * @frag: the paged fragment
2895 */
skb_frag_off(const skb_frag_t * frag)2896 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2897 {
2898 return frag->bv_offset;
2899 }
2900
2901 /**
2902 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2903 * @frag: skb fragment
2904 * @delta: value to add
2905 */
skb_frag_off_add(skb_frag_t * frag,int delta)2906 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2907 {
2908 frag->bv_offset += delta;
2909 }
2910
2911 /**
2912 * skb_frag_off_set() - Sets the offset of a skb fragment
2913 * @frag: skb fragment
2914 * @offset: offset of fragment
2915 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)2916 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2917 {
2918 frag->bv_offset = offset;
2919 }
2920
2921 /**
2922 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2923 * @fragto: skb fragment where offset is set
2924 * @fragfrom: skb fragment offset is copied from
2925 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)2926 static inline void skb_frag_off_copy(skb_frag_t *fragto,
2927 const skb_frag_t *fragfrom)
2928 {
2929 fragto->bv_offset = fragfrom->bv_offset;
2930 }
2931
2932 /**
2933 * skb_frag_page - retrieve the page referred to by a paged fragment
2934 * @frag: the paged fragment
2935 *
2936 * Returns the &struct page associated with @frag.
2937 */
skb_frag_page(const skb_frag_t * frag)2938 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2939 {
2940 return frag->bv_page;
2941 }
2942
2943 /**
2944 * __skb_frag_ref - take an addition reference on a paged fragment.
2945 * @frag: the paged fragment
2946 *
2947 * Takes an additional reference on the paged fragment @frag.
2948 */
__skb_frag_ref(skb_frag_t * frag)2949 static inline void __skb_frag_ref(skb_frag_t *frag)
2950 {
2951 get_page(skb_frag_page(frag));
2952 }
2953
2954 /**
2955 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2956 * @skb: the buffer
2957 * @f: the fragment offset.
2958 *
2959 * Takes an additional reference on the @f'th paged fragment of @skb.
2960 */
skb_frag_ref(struct sk_buff * skb,int f)2961 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2962 {
2963 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2964 }
2965
2966 /**
2967 * __skb_frag_unref - release a reference on a paged fragment.
2968 * @frag: the paged fragment
2969 *
2970 * Releases a reference on the paged fragment @frag.
2971 */
__skb_frag_unref(skb_frag_t * frag)2972 static inline void __skb_frag_unref(skb_frag_t *frag)
2973 {
2974 put_page(skb_frag_page(frag));
2975 }
2976
2977 /**
2978 * skb_frag_unref - release a reference on a paged fragment of an skb.
2979 * @skb: the buffer
2980 * @f: the fragment offset
2981 *
2982 * Releases a reference on the @f'th paged fragment of @skb.
2983 */
skb_frag_unref(struct sk_buff * skb,int f)2984 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2985 {
2986 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2987 }
2988
2989 /**
2990 * skb_frag_address - gets the address of the data contained in a paged fragment
2991 * @frag: the paged fragment buffer
2992 *
2993 * Returns the address of the data within @frag. The page must already
2994 * be mapped.
2995 */
skb_frag_address(const skb_frag_t * frag)2996 static inline void *skb_frag_address(const skb_frag_t *frag)
2997 {
2998 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
2999 }
3000
3001 /**
3002 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3003 * @frag: the paged fragment buffer
3004 *
3005 * Returns the address of the data within @frag. Checks that the page
3006 * is mapped and returns %NULL otherwise.
3007 */
skb_frag_address_safe(const skb_frag_t * frag)3008 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3009 {
3010 void *ptr = page_address(skb_frag_page(frag));
3011 if (unlikely(!ptr))
3012 return NULL;
3013
3014 return ptr + skb_frag_off(frag);
3015 }
3016
3017 /**
3018 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3019 * @fragto: skb fragment where page is set
3020 * @fragfrom: skb fragment page is copied from
3021 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3022 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3023 const skb_frag_t *fragfrom)
3024 {
3025 fragto->bv_page = fragfrom->bv_page;
3026 }
3027
3028 /**
3029 * __skb_frag_set_page - sets the page contained in a paged fragment
3030 * @frag: the paged fragment
3031 * @page: the page to set
3032 *
3033 * Sets the fragment @frag to contain @page.
3034 */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3035 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3036 {
3037 frag->bv_page = page;
3038 }
3039
3040 /**
3041 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3042 * @skb: the buffer
3043 * @f: the fragment offset
3044 * @page: the page to set
3045 *
3046 * Sets the @f'th fragment of @skb to contain @page.
3047 */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3048 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3049 struct page *page)
3050 {
3051 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3052 }
3053
3054 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3055
3056 /**
3057 * skb_frag_dma_map - maps a paged fragment via the DMA API
3058 * @dev: the device to map the fragment to
3059 * @frag: the paged fragment to map
3060 * @offset: the offset within the fragment (starting at the
3061 * fragment's own offset)
3062 * @size: the number of bytes to map
3063 * @dir: the direction of the mapping (``PCI_DMA_*``)
3064 *
3065 * Maps the page associated with @frag to @device.
3066 */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3067 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3068 const skb_frag_t *frag,
3069 size_t offset, size_t size,
3070 enum dma_data_direction dir)
3071 {
3072 return dma_map_page(dev, skb_frag_page(frag),
3073 skb_frag_off(frag) + offset, size, dir);
3074 }
3075
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3076 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3077 gfp_t gfp_mask)
3078 {
3079 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3080 }
3081
3082
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3083 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3084 gfp_t gfp_mask)
3085 {
3086 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3087 }
3088
3089
3090 /**
3091 * skb_clone_writable - is the header of a clone writable
3092 * @skb: buffer to check
3093 * @len: length up to which to write
3094 *
3095 * Returns true if modifying the header part of the cloned buffer
3096 * does not requires the data to be copied.
3097 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3098 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3099 {
3100 return !skb_header_cloned(skb) &&
3101 skb_headroom(skb) + len <= skb->hdr_len;
3102 }
3103
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3104 static inline int skb_try_make_writable(struct sk_buff *skb,
3105 unsigned int write_len)
3106 {
3107 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3108 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3109 }
3110
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3111 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3112 int cloned)
3113 {
3114 int delta = 0;
3115
3116 if (headroom > skb_headroom(skb))
3117 delta = headroom - skb_headroom(skb);
3118
3119 if (delta || cloned)
3120 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3121 GFP_ATOMIC);
3122 return 0;
3123 }
3124
3125 /**
3126 * skb_cow - copy header of skb when it is required
3127 * @skb: buffer to cow
3128 * @headroom: needed headroom
3129 *
3130 * If the skb passed lacks sufficient headroom or its data part
3131 * is shared, data is reallocated. If reallocation fails, an error
3132 * is returned and original skb is not changed.
3133 *
3134 * The result is skb with writable area skb->head...skb->tail
3135 * and at least @headroom of space at head.
3136 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3137 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3138 {
3139 return __skb_cow(skb, headroom, skb_cloned(skb));
3140 }
3141
3142 /**
3143 * skb_cow_head - skb_cow but only making the head writable
3144 * @skb: buffer to cow
3145 * @headroom: needed headroom
3146 *
3147 * This function is identical to skb_cow except that we replace the
3148 * skb_cloned check by skb_header_cloned. It should be used when
3149 * you only need to push on some header and do not need to modify
3150 * the data.
3151 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3152 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3153 {
3154 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3155 }
3156
3157 /**
3158 * skb_padto - pad an skbuff up to a minimal size
3159 * @skb: buffer to pad
3160 * @len: minimal length
3161 *
3162 * Pads up a buffer to ensure the trailing bytes exist and are
3163 * blanked. If the buffer already contains sufficient data it
3164 * is untouched. Otherwise it is extended. Returns zero on
3165 * success. The skb is freed on error.
3166 */
skb_padto(struct sk_buff * skb,unsigned int len)3167 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3168 {
3169 unsigned int size = skb->len;
3170 if (likely(size >= len))
3171 return 0;
3172 return skb_pad(skb, len - size);
3173 }
3174
3175 /**
3176 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3177 * @skb: buffer to pad
3178 * @len: minimal length
3179 * @free_on_error: free buffer on error
3180 *
3181 * Pads up a buffer to ensure the trailing bytes exist and are
3182 * blanked. If the buffer already contains sufficient data it
3183 * is untouched. Otherwise it is extended. Returns zero on
3184 * success. The skb is freed on error if @free_on_error is true.
3185 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3186 static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3187 bool free_on_error)
3188 {
3189 unsigned int size = skb->len;
3190
3191 if (unlikely(size < len)) {
3192 len -= size;
3193 if (__skb_pad(skb, len, free_on_error))
3194 return -ENOMEM;
3195 __skb_put(skb, len);
3196 }
3197 return 0;
3198 }
3199
3200 /**
3201 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3202 * @skb: buffer to pad
3203 * @len: minimal length
3204 *
3205 * Pads up a buffer to ensure the trailing bytes exist and are
3206 * blanked. If the buffer already contains sufficient data it
3207 * is untouched. Otherwise it is extended. Returns zero on
3208 * success. The skb is freed on error.
3209 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3210 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3211 {
3212 return __skb_put_padto(skb, len, true);
3213 }
3214
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3215 static inline int skb_add_data(struct sk_buff *skb,
3216 struct iov_iter *from, int copy)
3217 {
3218 const int off = skb->len;
3219
3220 if (skb->ip_summed == CHECKSUM_NONE) {
3221 __wsum csum = 0;
3222 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3223 &csum, from)) {
3224 skb->csum = csum_block_add(skb->csum, csum, off);
3225 return 0;
3226 }
3227 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3228 return 0;
3229
3230 __skb_trim(skb, off);
3231 return -EFAULT;
3232 }
3233
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3234 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3235 const struct page *page, int off)
3236 {
3237 if (skb_zcopy(skb))
3238 return false;
3239 if (i) {
3240 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3241
3242 return page == skb_frag_page(frag) &&
3243 off == skb_frag_off(frag) + skb_frag_size(frag);
3244 }
3245 return false;
3246 }
3247
__skb_linearize(struct sk_buff * skb)3248 static inline int __skb_linearize(struct sk_buff *skb)
3249 {
3250 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3251 }
3252
3253 /**
3254 * skb_linearize - convert paged skb to linear one
3255 * @skb: buffer to linarize
3256 *
3257 * If there is no free memory -ENOMEM is returned, otherwise zero
3258 * is returned and the old skb data released.
3259 */
skb_linearize(struct sk_buff * skb)3260 static inline int skb_linearize(struct sk_buff *skb)
3261 {
3262 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3263 }
3264
3265 /**
3266 * skb_has_shared_frag - can any frag be overwritten
3267 * @skb: buffer to test
3268 *
3269 * Return true if the skb has at least one frag that might be modified
3270 * by an external entity (as in vmsplice()/sendfile())
3271 */
skb_has_shared_frag(const struct sk_buff * skb)3272 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3273 {
3274 return skb_is_nonlinear(skb) &&
3275 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3276 }
3277
3278 /**
3279 * skb_linearize_cow - make sure skb is linear and writable
3280 * @skb: buffer to process
3281 *
3282 * If there is no free memory -ENOMEM is returned, otherwise zero
3283 * is returned and the old skb data released.
3284 */
skb_linearize_cow(struct sk_buff * skb)3285 static inline int skb_linearize_cow(struct sk_buff *skb)
3286 {
3287 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3288 __skb_linearize(skb) : 0;
3289 }
3290
3291 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3292 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3293 unsigned int off)
3294 {
3295 if (skb->ip_summed == CHECKSUM_COMPLETE)
3296 skb->csum = csum_block_sub(skb->csum,
3297 csum_partial(start, len, 0), off);
3298 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3299 skb_checksum_start_offset(skb) < 0)
3300 skb->ip_summed = CHECKSUM_NONE;
3301 }
3302
3303 /**
3304 * skb_postpull_rcsum - update checksum for received skb after pull
3305 * @skb: buffer to update
3306 * @start: start of data before pull
3307 * @len: length of data pulled
3308 *
3309 * After doing a pull on a received packet, you need to call this to
3310 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3311 * CHECKSUM_NONE so that it can be recomputed from scratch.
3312 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3313 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3314 const void *start, unsigned int len)
3315 {
3316 __skb_postpull_rcsum(skb, start, len, 0);
3317 }
3318
3319 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3320 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3321 unsigned int off)
3322 {
3323 if (skb->ip_summed == CHECKSUM_COMPLETE)
3324 skb->csum = csum_block_add(skb->csum,
3325 csum_partial(start, len, 0), off);
3326 }
3327
3328 /**
3329 * skb_postpush_rcsum - update checksum for received skb after push
3330 * @skb: buffer to update
3331 * @start: start of data after push
3332 * @len: length of data pushed
3333 *
3334 * After doing a push on a received packet, you need to call this to
3335 * update the CHECKSUM_COMPLETE checksum.
3336 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3337 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3338 const void *start, unsigned int len)
3339 {
3340 __skb_postpush_rcsum(skb, start, len, 0);
3341 }
3342
3343 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3344
3345 /**
3346 * skb_push_rcsum - push skb and update receive checksum
3347 * @skb: buffer to update
3348 * @len: length of data pulled
3349 *
3350 * This function performs an skb_push on the packet and updates
3351 * the CHECKSUM_COMPLETE checksum. It should be used on
3352 * receive path processing instead of skb_push unless you know
3353 * that the checksum difference is zero (e.g., a valid IP header)
3354 * or you are setting ip_summed to CHECKSUM_NONE.
3355 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3356 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3357 {
3358 skb_push(skb, len);
3359 skb_postpush_rcsum(skb, skb->data, len);
3360 return skb->data;
3361 }
3362
3363 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3364 /**
3365 * pskb_trim_rcsum - trim received skb and update checksum
3366 * @skb: buffer to trim
3367 * @len: new length
3368 *
3369 * This is exactly the same as pskb_trim except that it ensures the
3370 * checksum of received packets are still valid after the operation.
3371 * It can change skb pointers.
3372 */
3373
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3374 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3375 {
3376 if (likely(len >= skb->len))
3377 return 0;
3378 return pskb_trim_rcsum_slow(skb, len);
3379 }
3380
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3381 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3382 {
3383 if (skb->ip_summed == CHECKSUM_COMPLETE)
3384 skb->ip_summed = CHECKSUM_NONE;
3385 __skb_trim(skb, len);
3386 return 0;
3387 }
3388
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3389 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3390 {
3391 if (skb->ip_summed == CHECKSUM_COMPLETE)
3392 skb->ip_summed = CHECKSUM_NONE;
3393 return __skb_grow(skb, len);
3394 }
3395
3396 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3397 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3398 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3399 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3400 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3401
3402 #define skb_queue_walk(queue, skb) \
3403 for (skb = (queue)->next; \
3404 skb != (struct sk_buff *)(queue); \
3405 skb = skb->next)
3406
3407 #define skb_queue_walk_safe(queue, skb, tmp) \
3408 for (skb = (queue)->next, tmp = skb->next; \
3409 skb != (struct sk_buff *)(queue); \
3410 skb = tmp, tmp = skb->next)
3411
3412 #define skb_queue_walk_from(queue, skb) \
3413 for (; skb != (struct sk_buff *)(queue); \
3414 skb = skb->next)
3415
3416 #define skb_rbtree_walk(skb, root) \
3417 for (skb = skb_rb_first(root); skb != NULL; \
3418 skb = skb_rb_next(skb))
3419
3420 #define skb_rbtree_walk_from(skb) \
3421 for (; skb != NULL; \
3422 skb = skb_rb_next(skb))
3423
3424 #define skb_rbtree_walk_from_safe(skb, tmp) \
3425 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3426 skb = tmp)
3427
3428 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3429 for (tmp = skb->next; \
3430 skb != (struct sk_buff *)(queue); \
3431 skb = tmp, tmp = skb->next)
3432
3433 #define skb_queue_reverse_walk(queue, skb) \
3434 for (skb = (queue)->prev; \
3435 skb != (struct sk_buff *)(queue); \
3436 skb = skb->prev)
3437
3438 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3439 for (skb = (queue)->prev, tmp = skb->prev; \
3440 skb != (struct sk_buff *)(queue); \
3441 skb = tmp, tmp = skb->prev)
3442
3443 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3444 for (tmp = skb->prev; \
3445 skb != (struct sk_buff *)(queue); \
3446 skb = tmp, tmp = skb->prev)
3447
skb_has_frag_list(const struct sk_buff * skb)3448 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3449 {
3450 return skb_shinfo(skb)->frag_list != NULL;
3451 }
3452
skb_frag_list_init(struct sk_buff * skb)3453 static inline void skb_frag_list_init(struct sk_buff *skb)
3454 {
3455 skb_shinfo(skb)->frag_list = NULL;
3456 }
3457
3458 #define skb_walk_frags(skb, iter) \
3459 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3460
3461
3462 int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3463 const struct sk_buff *skb);
3464 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3465 struct sk_buff_head *queue,
3466 unsigned int flags,
3467 void (*destructor)(struct sock *sk,
3468 struct sk_buff *skb),
3469 int *off, int *err,
3470 struct sk_buff **last);
3471 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3472 void (*destructor)(struct sock *sk,
3473 struct sk_buff *skb),
3474 int *off, int *err,
3475 struct sk_buff **last);
3476 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3477 void (*destructor)(struct sock *sk,
3478 struct sk_buff *skb),
3479 int *off, int *err);
3480 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3481 int *err);
3482 __poll_t datagram_poll(struct file *file, struct socket *sock,
3483 struct poll_table_struct *wait);
3484 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3485 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3486 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3487 struct msghdr *msg, int size)
3488 {
3489 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3490 }
3491 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3492 struct msghdr *msg);
3493 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3494 struct iov_iter *to, int len,
3495 struct ahash_request *hash);
3496 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3497 struct iov_iter *from, int len);
3498 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3499 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3500 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
skb_free_datagram_locked(struct sock * sk,struct sk_buff * skb)3501 static inline void skb_free_datagram_locked(struct sock *sk,
3502 struct sk_buff *skb)
3503 {
3504 __skb_free_datagram_locked(sk, skb, 0);
3505 }
3506 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3507 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3508 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3509 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3510 int len, __wsum csum);
3511 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3512 struct pipe_inode_info *pipe, unsigned int len,
3513 unsigned int flags);
3514 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3515 int len);
3516 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3517 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3518 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3519 int len, int hlen);
3520 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3521 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3522 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3523 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3524 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3525 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3526 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3527 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3528 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3529 int skb_vlan_pop(struct sk_buff *skb);
3530 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3531 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3532 int mac_len, bool ethernet);
3533 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3534 bool ethernet);
3535 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3536 int skb_mpls_dec_ttl(struct sk_buff *skb);
3537 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3538 gfp_t gfp);
3539
memcpy_from_msg(void * data,struct msghdr * msg,int len)3540 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3541 {
3542 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3543 }
3544
memcpy_to_msg(struct msghdr * msg,void * data,int len)3545 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3546 {
3547 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3548 }
3549
3550 struct skb_checksum_ops {
3551 __wsum (*update)(const void *mem, int len, __wsum wsum);
3552 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3553 };
3554
3555 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3556
3557 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3558 __wsum csum, const struct skb_checksum_ops *ops);
3559 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3560 __wsum csum);
3561
3562 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * data,int hlen,void * buffer)3563 __skb_header_pointer(const struct sk_buff *skb, int offset,
3564 int len, void *data, int hlen, void *buffer)
3565 {
3566 if (hlen - offset >= len)
3567 return data + offset;
3568
3569 if (!skb ||
3570 skb_copy_bits(skb, offset, buffer, len) < 0)
3571 return NULL;
3572
3573 return buffer;
3574 }
3575
3576 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3577 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3578 {
3579 return __skb_header_pointer(skb, offset, len, skb->data,
3580 skb_headlen(skb), buffer);
3581 }
3582
3583 /**
3584 * skb_needs_linearize - check if we need to linearize a given skb
3585 * depending on the given device features.
3586 * @skb: socket buffer to check
3587 * @features: net device features
3588 *
3589 * Returns true if either:
3590 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3591 * 2. skb is fragmented and the device does not support SG.
3592 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3593 static inline bool skb_needs_linearize(struct sk_buff *skb,
3594 netdev_features_t features)
3595 {
3596 return skb_is_nonlinear(skb) &&
3597 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3598 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3599 }
3600
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3601 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3602 void *to,
3603 const unsigned int len)
3604 {
3605 memcpy(to, skb->data, len);
3606 }
3607
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3608 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3609 const int offset, void *to,
3610 const unsigned int len)
3611 {
3612 memcpy(to, skb->data + offset, len);
3613 }
3614
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3615 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3616 const void *from,
3617 const unsigned int len)
3618 {
3619 memcpy(skb->data, from, len);
3620 }
3621
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3622 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3623 const int offset,
3624 const void *from,
3625 const unsigned int len)
3626 {
3627 memcpy(skb->data + offset, from, len);
3628 }
3629
3630 void skb_init(void);
3631
skb_get_ktime(const struct sk_buff * skb)3632 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3633 {
3634 return skb->tstamp;
3635 }
3636
3637 /**
3638 * skb_get_timestamp - get timestamp from a skb
3639 * @skb: skb to get stamp from
3640 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3641 *
3642 * Timestamps are stored in the skb as offsets to a base timestamp.
3643 * This function converts the offset back to a struct timeval and stores
3644 * it in stamp.
3645 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3646 static inline void skb_get_timestamp(const struct sk_buff *skb,
3647 struct __kernel_old_timeval *stamp)
3648 {
3649 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3650 }
3651
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3652 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3653 struct __kernel_sock_timeval *stamp)
3654 {
3655 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3656
3657 stamp->tv_sec = ts.tv_sec;
3658 stamp->tv_usec = ts.tv_nsec / 1000;
3659 }
3660
skb_get_timestampns(const struct sk_buff * skb,struct timespec * stamp)3661 static inline void skb_get_timestampns(const struct sk_buff *skb,
3662 struct timespec *stamp)
3663 {
3664 *stamp = ktime_to_timespec(skb->tstamp);
3665 }
3666
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3667 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3668 struct __kernel_timespec *stamp)
3669 {
3670 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3671
3672 stamp->tv_sec = ts.tv_sec;
3673 stamp->tv_nsec = ts.tv_nsec;
3674 }
3675
__net_timestamp(struct sk_buff * skb)3676 static inline void __net_timestamp(struct sk_buff *skb)
3677 {
3678 skb->tstamp = ktime_get_real();
3679 }
3680
net_timedelta(ktime_t t)3681 static inline ktime_t net_timedelta(ktime_t t)
3682 {
3683 return ktime_sub(ktime_get_real(), t);
3684 }
3685
net_invalid_timestamp(void)3686 static inline ktime_t net_invalid_timestamp(void)
3687 {
3688 return 0;
3689 }
3690
skb_metadata_len(const struct sk_buff * skb)3691 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3692 {
3693 return skb_shinfo(skb)->meta_len;
3694 }
3695
skb_metadata_end(const struct sk_buff * skb)3696 static inline void *skb_metadata_end(const struct sk_buff *skb)
3697 {
3698 return skb_mac_header(skb);
3699 }
3700
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3701 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3702 const struct sk_buff *skb_b,
3703 u8 meta_len)
3704 {
3705 const void *a = skb_metadata_end(skb_a);
3706 const void *b = skb_metadata_end(skb_b);
3707 /* Using more efficient varaiant than plain call to memcmp(). */
3708 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3709 u64 diffs = 0;
3710
3711 switch (meta_len) {
3712 #define __it(x, op) (x -= sizeof(u##op))
3713 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3714 case 32: diffs |= __it_diff(a, b, 64);
3715 /* fall through */
3716 case 24: diffs |= __it_diff(a, b, 64);
3717 /* fall through */
3718 case 16: diffs |= __it_diff(a, b, 64);
3719 /* fall through */
3720 case 8: diffs |= __it_diff(a, b, 64);
3721 break;
3722 case 28: diffs |= __it_diff(a, b, 64);
3723 /* fall through */
3724 case 20: diffs |= __it_diff(a, b, 64);
3725 /* fall through */
3726 case 12: diffs |= __it_diff(a, b, 64);
3727 /* fall through */
3728 case 4: diffs |= __it_diff(a, b, 32);
3729 break;
3730 }
3731 return diffs;
3732 #else
3733 return memcmp(a - meta_len, b - meta_len, meta_len);
3734 #endif
3735 }
3736
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)3737 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3738 const struct sk_buff *skb_b)
3739 {
3740 u8 len_a = skb_metadata_len(skb_a);
3741 u8 len_b = skb_metadata_len(skb_b);
3742
3743 if (!(len_a | len_b))
3744 return false;
3745
3746 return len_a != len_b ?
3747 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3748 }
3749
skb_metadata_set(struct sk_buff * skb,u8 meta_len)3750 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3751 {
3752 skb_shinfo(skb)->meta_len = meta_len;
3753 }
3754
skb_metadata_clear(struct sk_buff * skb)3755 static inline void skb_metadata_clear(struct sk_buff *skb)
3756 {
3757 skb_metadata_set(skb, 0);
3758 }
3759
3760 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3761
3762 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3763
3764 void skb_clone_tx_timestamp(struct sk_buff *skb);
3765 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3766
3767 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3768
skb_clone_tx_timestamp(struct sk_buff * skb)3769 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3770 {
3771 }
3772
skb_defer_rx_timestamp(struct sk_buff * skb)3773 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3774 {
3775 return false;
3776 }
3777
3778 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3779
3780 /**
3781 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3782 *
3783 * PHY drivers may accept clones of transmitted packets for
3784 * timestamping via their phy_driver.txtstamp method. These drivers
3785 * must call this function to return the skb back to the stack with a
3786 * timestamp.
3787 *
3788 * @skb: clone of the the original outgoing packet
3789 * @hwtstamps: hardware time stamps
3790 *
3791 */
3792 void skb_complete_tx_timestamp(struct sk_buff *skb,
3793 struct skb_shared_hwtstamps *hwtstamps);
3794
3795 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3796 struct skb_shared_hwtstamps *hwtstamps,
3797 struct sock *sk, int tstype);
3798
3799 /**
3800 * skb_tstamp_tx - queue clone of skb with send time stamps
3801 * @orig_skb: the original outgoing packet
3802 * @hwtstamps: hardware time stamps, may be NULL if not available
3803 *
3804 * If the skb has a socket associated, then this function clones the
3805 * skb (thus sharing the actual data and optional structures), stores
3806 * the optional hardware time stamping information (if non NULL) or
3807 * generates a software time stamp (otherwise), then queues the clone
3808 * to the error queue of the socket. Errors are silently ignored.
3809 */
3810 void skb_tstamp_tx(struct sk_buff *orig_skb,
3811 struct skb_shared_hwtstamps *hwtstamps);
3812
3813 /**
3814 * skb_tx_timestamp() - Driver hook for transmit timestamping
3815 *
3816 * Ethernet MAC Drivers should call this function in their hard_xmit()
3817 * function immediately before giving the sk_buff to the MAC hardware.
3818 *
3819 * Specifically, one should make absolutely sure that this function is
3820 * called before TX completion of this packet can trigger. Otherwise
3821 * the packet could potentially already be freed.
3822 *
3823 * @skb: A socket buffer.
3824 */
skb_tx_timestamp(struct sk_buff * skb)3825 static inline void skb_tx_timestamp(struct sk_buff *skb)
3826 {
3827 skb_clone_tx_timestamp(skb);
3828 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3829 skb_tstamp_tx(skb, NULL);
3830 }
3831
3832 /**
3833 * skb_complete_wifi_ack - deliver skb with wifi status
3834 *
3835 * @skb: the original outgoing packet
3836 * @acked: ack status
3837 *
3838 */
3839 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3840
3841 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3842 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3843
skb_csum_unnecessary(const struct sk_buff * skb)3844 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3845 {
3846 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3847 skb->csum_valid ||
3848 (skb->ip_summed == CHECKSUM_PARTIAL &&
3849 skb_checksum_start_offset(skb) >= 0));
3850 }
3851
3852 /**
3853 * skb_checksum_complete - Calculate checksum of an entire packet
3854 * @skb: packet to process
3855 *
3856 * This function calculates the checksum over the entire packet plus
3857 * the value of skb->csum. The latter can be used to supply the
3858 * checksum of a pseudo header as used by TCP/UDP. It returns the
3859 * checksum.
3860 *
3861 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3862 * this function can be used to verify that checksum on received
3863 * packets. In that case the function should return zero if the
3864 * checksum is correct. In particular, this function will return zero
3865 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3866 * hardware has already verified the correctness of the checksum.
3867 */
skb_checksum_complete(struct sk_buff * skb)3868 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3869 {
3870 return skb_csum_unnecessary(skb) ?
3871 0 : __skb_checksum_complete(skb);
3872 }
3873
__skb_decr_checksum_unnecessary(struct sk_buff * skb)3874 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3875 {
3876 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3877 if (skb->csum_level == 0)
3878 skb->ip_summed = CHECKSUM_NONE;
3879 else
3880 skb->csum_level--;
3881 }
3882 }
3883
__skb_incr_checksum_unnecessary(struct sk_buff * skb)3884 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3885 {
3886 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3887 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3888 skb->csum_level++;
3889 } else if (skb->ip_summed == CHECKSUM_NONE) {
3890 skb->ip_summed = CHECKSUM_UNNECESSARY;
3891 skb->csum_level = 0;
3892 }
3893 }
3894
3895 /* Check if we need to perform checksum complete validation.
3896 *
3897 * Returns true if checksum complete is needed, false otherwise
3898 * (either checksum is unnecessary or zero checksum is allowed).
3899 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3900 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3901 bool zero_okay,
3902 __sum16 check)
3903 {
3904 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3905 skb->csum_valid = 1;
3906 __skb_decr_checksum_unnecessary(skb);
3907 return false;
3908 }
3909
3910 return true;
3911 }
3912
3913 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
3914 * in checksum_init.
3915 */
3916 #define CHECKSUM_BREAK 76
3917
3918 /* Unset checksum-complete
3919 *
3920 * Unset checksum complete can be done when packet is being modified
3921 * (uncompressed for instance) and checksum-complete value is
3922 * invalidated.
3923 */
skb_checksum_complete_unset(struct sk_buff * skb)3924 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3925 {
3926 if (skb->ip_summed == CHECKSUM_COMPLETE)
3927 skb->ip_summed = CHECKSUM_NONE;
3928 }
3929
3930 /* Validate (init) checksum based on checksum complete.
3931 *
3932 * Return values:
3933 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3934 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3935 * checksum is stored in skb->csum for use in __skb_checksum_complete
3936 * non-zero: value of invalid checksum
3937 *
3938 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)3939 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3940 bool complete,
3941 __wsum psum)
3942 {
3943 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3944 if (!csum_fold(csum_add(psum, skb->csum))) {
3945 skb->csum_valid = 1;
3946 return 0;
3947 }
3948 }
3949
3950 skb->csum = psum;
3951
3952 if (complete || skb->len <= CHECKSUM_BREAK) {
3953 __sum16 csum;
3954
3955 csum = __skb_checksum_complete(skb);
3956 skb->csum_valid = !csum;
3957 return csum;
3958 }
3959
3960 return 0;
3961 }
3962
null_compute_pseudo(struct sk_buff * skb,int proto)3963 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3964 {
3965 return 0;
3966 }
3967
3968 /* Perform checksum validate (init). Note that this is a macro since we only
3969 * want to calculate the pseudo header which is an input function if necessary.
3970 * First we try to validate without any computation (checksum unnecessary) and
3971 * then calculate based on checksum complete calling the function to compute
3972 * pseudo header.
3973 *
3974 * Return values:
3975 * 0: checksum is validated or try to in skb_checksum_complete
3976 * non-zero: value of invalid checksum
3977 */
3978 #define __skb_checksum_validate(skb, proto, complete, \
3979 zero_okay, check, compute_pseudo) \
3980 ({ \
3981 __sum16 __ret = 0; \
3982 skb->csum_valid = 0; \
3983 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3984 __ret = __skb_checksum_validate_complete(skb, \
3985 complete, compute_pseudo(skb, proto)); \
3986 __ret; \
3987 })
3988
3989 #define skb_checksum_init(skb, proto, compute_pseudo) \
3990 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3991
3992 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3993 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3994
3995 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3996 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3997
3998 #define skb_checksum_validate_zero_check(skb, proto, check, \
3999 compute_pseudo) \
4000 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4001
4002 #define skb_checksum_simple_validate(skb) \
4003 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4004
__skb_checksum_convert_check(struct sk_buff * skb)4005 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4006 {
4007 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4008 }
4009
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4010 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4011 {
4012 skb->csum = ~pseudo;
4013 skb->ip_summed = CHECKSUM_COMPLETE;
4014 }
4015
4016 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4017 do { \
4018 if (__skb_checksum_convert_check(skb)) \
4019 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4020 } while (0)
4021
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4022 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4023 u16 start, u16 offset)
4024 {
4025 skb->ip_summed = CHECKSUM_PARTIAL;
4026 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4027 skb->csum_offset = offset - start;
4028 }
4029
4030 /* Update skbuf and packet to reflect the remote checksum offload operation.
4031 * When called, ptr indicates the starting point for skb->csum when
4032 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4033 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4034 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4035 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4036 int start, int offset, bool nopartial)
4037 {
4038 __wsum delta;
4039
4040 if (!nopartial) {
4041 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4042 return;
4043 }
4044
4045 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4046 __skb_checksum_complete(skb);
4047 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4048 }
4049
4050 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4051
4052 /* Adjust skb->csum since we changed the packet */
4053 skb->csum = csum_add(skb->csum, delta);
4054 }
4055
skb_nfct(const struct sk_buff * skb)4056 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4057 {
4058 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4059 return (void *)(skb->_nfct & NFCT_PTRMASK);
4060 #else
4061 return NULL;
4062 #endif
4063 }
4064
skb_get_nfct(const struct sk_buff * skb)4065 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4066 {
4067 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4068 return skb->_nfct;
4069 #else
4070 return 0UL;
4071 #endif
4072 }
4073
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4074 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4075 {
4076 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4077 skb->_nfct = nfct;
4078 #endif
4079 }
4080
4081 #ifdef CONFIG_SKB_EXTENSIONS
4082 enum skb_ext_id {
4083 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4084 SKB_EXT_BRIDGE_NF,
4085 #endif
4086 #ifdef CONFIG_XFRM
4087 SKB_EXT_SEC_PATH,
4088 #endif
4089 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4090 TC_SKB_EXT,
4091 #endif
4092 SKB_EXT_NUM, /* must be last */
4093 };
4094
4095 /**
4096 * struct skb_ext - sk_buff extensions
4097 * @refcnt: 1 on allocation, deallocated on 0
4098 * @offset: offset to add to @data to obtain extension address
4099 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4100 * @data: start of extension data, variable sized
4101 *
4102 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4103 * to use 'u8' types while allowing up to 2kb worth of extension data.
4104 */
4105 struct skb_ext {
4106 refcount_t refcnt;
4107 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4108 u8 chunks; /* same */
4109 char data[0] __aligned(8);
4110 };
4111
4112 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4113 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4114 void __skb_ext_put(struct skb_ext *ext);
4115
skb_ext_put(struct sk_buff * skb)4116 static inline void skb_ext_put(struct sk_buff *skb)
4117 {
4118 if (skb->active_extensions)
4119 __skb_ext_put(skb->extensions);
4120 }
4121
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4122 static inline void __skb_ext_copy(struct sk_buff *dst,
4123 const struct sk_buff *src)
4124 {
4125 dst->active_extensions = src->active_extensions;
4126
4127 if (src->active_extensions) {
4128 struct skb_ext *ext = src->extensions;
4129
4130 refcount_inc(&ext->refcnt);
4131 dst->extensions = ext;
4132 }
4133 }
4134
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4135 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4136 {
4137 skb_ext_put(dst);
4138 __skb_ext_copy(dst, src);
4139 }
4140
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4141 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4142 {
4143 return !!ext->offset[i];
4144 }
4145
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4146 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4147 {
4148 return skb->active_extensions & (1 << id);
4149 }
4150
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4151 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4152 {
4153 if (skb_ext_exist(skb, id))
4154 __skb_ext_del(skb, id);
4155 }
4156
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4157 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4158 {
4159 if (skb_ext_exist(skb, id)) {
4160 struct skb_ext *ext = skb->extensions;
4161
4162 return (void *)ext + (ext->offset[id] << 3);
4163 }
4164
4165 return NULL;
4166 }
4167
skb_ext_reset(struct sk_buff * skb)4168 static inline void skb_ext_reset(struct sk_buff *skb)
4169 {
4170 if (unlikely(skb->active_extensions)) {
4171 __skb_ext_put(skb->extensions);
4172 skb->active_extensions = 0;
4173 }
4174 }
4175
skb_has_extensions(struct sk_buff * skb)4176 static inline bool skb_has_extensions(struct sk_buff *skb)
4177 {
4178 return unlikely(skb->active_extensions);
4179 }
4180 #else
skb_ext_put(struct sk_buff * skb)4181 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4182 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4183 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4184 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * s)4185 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4186 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4187 #endif /* CONFIG_SKB_EXTENSIONS */
4188
nf_reset_ct(struct sk_buff * skb)4189 static inline void nf_reset_ct(struct sk_buff *skb)
4190 {
4191 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4192 nf_conntrack_put(skb_nfct(skb));
4193 skb->_nfct = 0;
4194 #endif
4195 }
4196
nf_reset_trace(struct sk_buff * skb)4197 static inline void nf_reset_trace(struct sk_buff *skb)
4198 {
4199 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4200 skb->nf_trace = 0;
4201 #endif
4202 }
4203
ipvs_reset(struct sk_buff * skb)4204 static inline void ipvs_reset(struct sk_buff *skb)
4205 {
4206 #if IS_ENABLED(CONFIG_IP_VS)
4207 skb->ipvs_property = 0;
4208 #endif
4209 }
4210
4211 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4212 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4213 bool copy)
4214 {
4215 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4216 dst->_nfct = src->_nfct;
4217 nf_conntrack_get(skb_nfct(src));
4218 #endif
4219 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4220 if (copy)
4221 dst->nf_trace = src->nf_trace;
4222 #endif
4223 }
4224
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4225 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4226 {
4227 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4228 nf_conntrack_put(skb_nfct(dst));
4229 #endif
4230 __nf_copy(dst, src, true);
4231 }
4232
4233 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4234 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4235 {
4236 to->secmark = from->secmark;
4237 }
4238
skb_init_secmark(struct sk_buff * skb)4239 static inline void skb_init_secmark(struct sk_buff *skb)
4240 {
4241 skb->secmark = 0;
4242 }
4243 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4244 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4245 { }
4246
skb_init_secmark(struct sk_buff * skb)4247 static inline void skb_init_secmark(struct sk_buff *skb)
4248 { }
4249 #endif
4250
secpath_exists(const struct sk_buff * skb)4251 static inline int secpath_exists(const struct sk_buff *skb)
4252 {
4253 #ifdef CONFIG_XFRM
4254 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4255 #else
4256 return 0;
4257 #endif
4258 }
4259
skb_irq_freeable(const struct sk_buff * skb)4260 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4261 {
4262 return !skb->destructor &&
4263 !secpath_exists(skb) &&
4264 !skb_nfct(skb) &&
4265 !skb->_skb_refdst &&
4266 !skb_has_frag_list(skb);
4267 }
4268
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4269 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4270 {
4271 skb->queue_mapping = queue_mapping;
4272 }
4273
skb_get_queue_mapping(const struct sk_buff * skb)4274 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4275 {
4276 return skb->queue_mapping;
4277 }
4278
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4279 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4280 {
4281 to->queue_mapping = from->queue_mapping;
4282 }
4283
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4284 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4285 {
4286 skb->queue_mapping = rx_queue + 1;
4287 }
4288
skb_get_rx_queue(const struct sk_buff * skb)4289 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4290 {
4291 return skb->queue_mapping - 1;
4292 }
4293
skb_rx_queue_recorded(const struct sk_buff * skb)4294 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4295 {
4296 return skb->queue_mapping != 0;
4297 }
4298
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4299 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4300 {
4301 skb->dst_pending_confirm = val;
4302 }
4303
skb_get_dst_pending_confirm(const struct sk_buff * skb)4304 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4305 {
4306 return skb->dst_pending_confirm != 0;
4307 }
4308
skb_sec_path(const struct sk_buff * skb)4309 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4310 {
4311 #ifdef CONFIG_XFRM
4312 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4313 #else
4314 return NULL;
4315 #endif
4316 }
4317
4318 /* Keeps track of mac header offset relative to skb->head.
4319 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4320 * For non-tunnel skb it points to skb_mac_header() and for
4321 * tunnel skb it points to outer mac header.
4322 * Keeps track of level of encapsulation of network headers.
4323 */
4324 struct skb_gso_cb {
4325 union {
4326 int mac_offset;
4327 int data_offset;
4328 };
4329 int encap_level;
4330 __wsum csum;
4331 __u16 csum_start;
4332 };
4333 #define SKB_SGO_CB_OFFSET 32
4334 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
4335
skb_tnl_header_len(const struct sk_buff * inner_skb)4336 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4337 {
4338 return (skb_mac_header(inner_skb) - inner_skb->head) -
4339 SKB_GSO_CB(inner_skb)->mac_offset;
4340 }
4341
gso_pskb_expand_head(struct sk_buff * skb,int extra)4342 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4343 {
4344 int new_headroom, headroom;
4345 int ret;
4346
4347 headroom = skb_headroom(skb);
4348 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4349 if (ret)
4350 return ret;
4351
4352 new_headroom = skb_headroom(skb);
4353 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4354 return 0;
4355 }
4356
gso_reset_checksum(struct sk_buff * skb,__wsum res)4357 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4358 {
4359 /* Do not update partial checksums if remote checksum is enabled. */
4360 if (skb->remcsum_offload)
4361 return;
4362
4363 SKB_GSO_CB(skb)->csum = res;
4364 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4365 }
4366
4367 /* Compute the checksum for a gso segment. First compute the checksum value
4368 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4369 * then add in skb->csum (checksum from csum_start to end of packet).
4370 * skb->csum and csum_start are then updated to reflect the checksum of the
4371 * resultant packet starting from the transport header-- the resultant checksum
4372 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4373 * header.
4374 */
gso_make_checksum(struct sk_buff * skb,__wsum res)4375 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4376 {
4377 unsigned char *csum_start = skb_transport_header(skb);
4378 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4379 __wsum partial = SKB_GSO_CB(skb)->csum;
4380
4381 SKB_GSO_CB(skb)->csum = res;
4382 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4383
4384 return csum_fold(csum_partial(csum_start, plen, partial));
4385 }
4386
skb_is_gso(const struct sk_buff * skb)4387 static inline bool skb_is_gso(const struct sk_buff *skb)
4388 {
4389 return skb_shinfo(skb)->gso_size;
4390 }
4391
4392 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4393 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4394 {
4395 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4396 }
4397
4398 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4399 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4400 {
4401 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4402 }
4403
4404 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4405 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4406 {
4407 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4408 }
4409
skb_gso_reset(struct sk_buff * skb)4410 static inline void skb_gso_reset(struct sk_buff *skb)
4411 {
4412 skb_shinfo(skb)->gso_size = 0;
4413 skb_shinfo(skb)->gso_segs = 0;
4414 skb_shinfo(skb)->gso_type = 0;
4415 }
4416
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4417 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4418 u16 increment)
4419 {
4420 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4421 return;
4422 shinfo->gso_size += increment;
4423 }
4424
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4425 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4426 u16 decrement)
4427 {
4428 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4429 return;
4430 shinfo->gso_size -= decrement;
4431 }
4432
4433 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4434
skb_warn_if_lro(const struct sk_buff * skb)4435 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4436 {
4437 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4438 * wanted then gso_type will be set. */
4439 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4440
4441 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4442 unlikely(shinfo->gso_type == 0)) {
4443 __skb_warn_lro_forwarding(skb);
4444 return true;
4445 }
4446 return false;
4447 }
4448
skb_forward_csum(struct sk_buff * skb)4449 static inline void skb_forward_csum(struct sk_buff *skb)
4450 {
4451 /* Unfortunately we don't support this one. Any brave souls? */
4452 if (skb->ip_summed == CHECKSUM_COMPLETE)
4453 skb->ip_summed = CHECKSUM_NONE;
4454 }
4455
4456 /**
4457 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4458 * @skb: skb to check
4459 *
4460 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4461 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4462 * use this helper, to document places where we make this assertion.
4463 */
skb_checksum_none_assert(const struct sk_buff * skb)4464 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4465 {
4466 #ifdef DEBUG
4467 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4468 #endif
4469 }
4470
4471 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4472
4473 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4474 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4475 unsigned int transport_len,
4476 __sum16(*skb_chkf)(struct sk_buff *skb));
4477
4478 /**
4479 * skb_head_is_locked - Determine if the skb->head is locked down
4480 * @skb: skb to check
4481 *
4482 * The head on skbs build around a head frag can be removed if they are
4483 * not cloned. This function returns true if the skb head is locked down
4484 * due to either being allocated via kmalloc, or by being a clone with
4485 * multiple references to the head.
4486 */
skb_head_is_locked(const struct sk_buff * skb)4487 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4488 {
4489 return !skb->head_frag || skb_cloned(skb);
4490 }
4491
4492 /* Local Checksum Offload.
4493 * Compute outer checksum based on the assumption that the
4494 * inner checksum will be offloaded later.
4495 * See Documentation/networking/checksum-offloads.rst for
4496 * explanation of how this works.
4497 * Fill in outer checksum adjustment (e.g. with sum of outer
4498 * pseudo-header) before calling.
4499 * Also ensure that inner checksum is in linear data area.
4500 */
lco_csum(struct sk_buff * skb)4501 static inline __wsum lco_csum(struct sk_buff *skb)
4502 {
4503 unsigned char *csum_start = skb_checksum_start(skb);
4504 unsigned char *l4_hdr = skb_transport_header(skb);
4505 __wsum partial;
4506
4507 /* Start with complement of inner checksum adjustment */
4508 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4509 skb->csum_offset));
4510
4511 /* Add in checksum of our headers (incl. outer checksum
4512 * adjustment filled in by caller) and return result.
4513 */
4514 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4515 }
4516
4517 #endif /* __KERNEL__ */
4518 #endif /* _LINUX_SKBUFF_H */
4519